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, int optlen)
2034 if (optlen < sizeof(int))
2037 if (get_user(val, (int __user *)optval))
2040 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2045 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2048 if (optlen > sizeof(struct sctp_event_subscribe))
2050 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2055 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2057 * This socket option is applicable to the UDP-style socket only. When
2058 * set it will cause associations that are idle for more than the
2059 * specified number of seconds to automatically close. An association
2060 * being idle is defined an association that has NOT sent or received
2061 * user data. The special value of '0' indicates that no automatic
2062 * close of any associations should be performed. The option expects an
2063 * integer defining the number of seconds of idle time before an
2064 * association is closed.
2066 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2069 struct sctp_sock *sp = sctp_sk(sk);
2071 /* Applicable to UDP-style socket only */
2072 if (sctp_style(sk, TCP))
2074 if (optlen != sizeof(int))
2076 if (copy_from_user(&sp->autoclose, optval, optlen))
2082 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2084 * Applications can enable or disable heartbeats for any peer address of
2085 * an association, modify an address's heartbeat interval, force a
2086 * heartbeat to be sent immediately, and adjust the address's maximum
2087 * number of retransmissions sent before an address is considered
2088 * unreachable. The following structure is used to access and modify an
2089 * address's parameters:
2091 * struct sctp_paddrparams {
2092 * sctp_assoc_t spp_assoc_id;
2093 * struct sockaddr_storage spp_address;
2094 * uint32_t spp_hbinterval;
2095 * uint16_t spp_pathmaxrxt;
2096 * uint32_t spp_pathmtu;
2097 * uint32_t spp_sackdelay;
2098 * uint32_t spp_flags;
2101 * spp_assoc_id - (one-to-many style socket) This is filled in the
2102 * application, and identifies the association for
2104 * spp_address - This specifies which address is of interest.
2105 * spp_hbinterval - This contains the value of the heartbeat interval,
2106 * in milliseconds. If a value of zero
2107 * is present in this field then no changes are to
2108 * be made to this parameter.
2109 * spp_pathmaxrxt - This contains the maximum number of
2110 * retransmissions before this address shall be
2111 * considered unreachable. If a value of zero
2112 * is present in this field then no changes are to
2113 * be made to this parameter.
2114 * spp_pathmtu - When Path MTU discovery is disabled the value
2115 * specified here will be the "fixed" path mtu.
2116 * Note that if the spp_address field is empty
2117 * then all associations on this address will
2118 * have this fixed path mtu set upon them.
2120 * spp_sackdelay - When delayed sack is enabled, this value specifies
2121 * the number of milliseconds that sacks will be delayed
2122 * for. This value will apply to all addresses of an
2123 * association if the spp_address field is empty. Note
2124 * also, that if delayed sack is enabled and this
2125 * value is set to 0, no change is made to the last
2126 * recorded delayed sack timer value.
2128 * spp_flags - These flags are used to control various features
2129 * on an association. The flag field may contain
2130 * zero or more of the following options.
2132 * SPP_HB_ENABLE - Enable heartbeats on the
2133 * specified address. Note that if the address
2134 * field is empty all addresses for the association
2135 * have heartbeats enabled upon them.
2137 * SPP_HB_DISABLE - Disable heartbeats on the
2138 * speicifed address. Note that if the address
2139 * field is empty all addresses for the association
2140 * will have their heartbeats disabled. Note also
2141 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2142 * mutually exclusive, only one of these two should
2143 * be specified. Enabling both fields will have
2144 * undetermined results.
2146 * SPP_HB_DEMAND - Request a user initiated heartbeat
2147 * to be made immediately.
2149 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2150 * heartbeat delayis to be set to the value of 0
2153 * SPP_PMTUD_ENABLE - This field will enable PMTU
2154 * discovery upon the specified address. Note that
2155 * if the address feild is empty then all addresses
2156 * on the association are effected.
2158 * SPP_PMTUD_DISABLE - This field will disable PMTU
2159 * discovery upon the specified address. Note that
2160 * if the address feild is empty then all addresses
2161 * on the association are effected. Not also that
2162 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2163 * exclusive. Enabling both will have undetermined
2166 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2167 * on delayed sack. The time specified in spp_sackdelay
2168 * is used to specify the sack delay for this address. Note
2169 * that if spp_address is empty then all addresses will
2170 * enable delayed sack and take on the sack delay
2171 * value specified in spp_sackdelay.
2172 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2173 * off delayed sack. If the spp_address field is blank then
2174 * delayed sack is disabled for the entire association. Note
2175 * also that this field is mutually exclusive to
2176 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2179 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2180 struct sctp_transport *trans,
2181 struct sctp_association *asoc,
2182 struct sctp_sock *sp,
2185 int sackdelay_change)
2189 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2190 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2195 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2196 * this field is ignored. Note also that a value of zero indicates
2197 * the current setting should be left unchanged.
2199 if (params->spp_flags & SPP_HB_ENABLE) {
2201 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2202 * set. This lets us use 0 value when this flag
2205 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2206 params->spp_hbinterval = 0;
2208 if (params->spp_hbinterval ||
2209 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2212 msecs_to_jiffies(params->spp_hbinterval);
2215 msecs_to_jiffies(params->spp_hbinterval);
2217 sp->hbinterval = params->spp_hbinterval;
2224 trans->param_flags =
2225 (trans->param_flags & ~SPP_HB) | hb_change;
2228 (asoc->param_flags & ~SPP_HB) | hb_change;
2231 (sp->param_flags & ~SPP_HB) | hb_change;
2235 /* When Path MTU discovery is disabled the value specified here will
2236 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2237 * include the flag SPP_PMTUD_DISABLE for this field to have any
2240 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2242 trans->pathmtu = params->spp_pathmtu;
2243 sctp_assoc_sync_pmtu(asoc);
2245 asoc->pathmtu = params->spp_pathmtu;
2246 sctp_frag_point(sp, params->spp_pathmtu);
2248 sp->pathmtu = params->spp_pathmtu;
2254 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2255 (params->spp_flags & SPP_PMTUD_ENABLE);
2256 trans->param_flags =
2257 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2259 sctp_transport_pmtu(trans);
2260 sctp_assoc_sync_pmtu(asoc);
2264 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2267 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2271 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2272 * value of this field is ignored. Note also that a value of zero
2273 * indicates the current setting should be left unchanged.
2275 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2278 msecs_to_jiffies(params->spp_sackdelay);
2281 msecs_to_jiffies(params->spp_sackdelay);
2283 sp->sackdelay = params->spp_sackdelay;
2287 if (sackdelay_change) {
2289 trans->param_flags =
2290 (trans->param_flags & ~SPP_SACKDELAY) |
2294 (asoc->param_flags & ~SPP_SACKDELAY) |
2298 (sp->param_flags & ~SPP_SACKDELAY) |
2303 /* Note that unless the spp_flag is set to SPP_PMTUD_ENABLE the value
2304 * of this field is ignored. Note also that a value of zero
2305 * indicates the current setting should be left unchanged.
2307 if ((params->spp_flags & SPP_PMTUD_ENABLE) && params->spp_pathmaxrxt) {
2309 trans->pathmaxrxt = params->spp_pathmaxrxt;
2311 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2313 sp->pathmaxrxt = params->spp_pathmaxrxt;
2320 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2321 char __user *optval, int optlen)
2323 struct sctp_paddrparams params;
2324 struct sctp_transport *trans = NULL;
2325 struct sctp_association *asoc = NULL;
2326 struct sctp_sock *sp = sctp_sk(sk);
2328 int hb_change, pmtud_change, sackdelay_change;
2330 if (optlen != sizeof(struct sctp_paddrparams))
2333 if (copy_from_user(¶ms, optval, optlen))
2336 /* Validate flags and value parameters. */
2337 hb_change = params.spp_flags & SPP_HB;
2338 pmtud_change = params.spp_flags & SPP_PMTUD;
2339 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2341 if (hb_change == SPP_HB ||
2342 pmtud_change == SPP_PMTUD ||
2343 sackdelay_change == SPP_SACKDELAY ||
2344 params.spp_sackdelay > 500 ||
2346 && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2349 /* If an address other than INADDR_ANY is specified, and
2350 * no transport is found, then the request is invalid.
2352 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
2353 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2354 params.spp_assoc_id);
2359 /* Get association, if assoc_id != 0 and the socket is a one
2360 * to many style socket, and an association was not found, then
2361 * the id was invalid.
2363 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2364 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2367 /* Heartbeat demand can only be sent on a transport or
2368 * association, but not a socket.
2370 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2373 /* Process parameters. */
2374 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2375 hb_change, pmtud_change,
2381 /* If changes are for association, also apply parameters to each
2384 if (!trans && asoc) {
2385 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2387 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2388 hb_change, pmtud_change,
2397 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2399 * This option will effect the way delayed acks are performed. This
2400 * option allows you to get or set the delayed ack time, in
2401 * milliseconds. It also allows changing the delayed ack frequency.
2402 * Changing the frequency to 1 disables the delayed sack algorithm. If
2403 * the assoc_id is 0, then this sets or gets the endpoints default
2404 * values. If the assoc_id field is non-zero, then the set or get
2405 * effects the specified association for the one to many model (the
2406 * assoc_id field is ignored by the one to one model). Note that if
2407 * sack_delay or sack_freq are 0 when setting this option, then the
2408 * current values will remain unchanged.
2410 * struct sctp_sack_info {
2411 * sctp_assoc_t sack_assoc_id;
2412 * uint32_t sack_delay;
2413 * uint32_t sack_freq;
2416 * sack_assoc_id - This parameter, indicates which association the user
2417 * is performing an action upon. Note that if this field's value is
2418 * zero then the endpoints default value is changed (effecting future
2419 * associations only).
2421 * sack_delay - This parameter contains the number of milliseconds that
2422 * the user is requesting the delayed ACK timer be set to. Note that
2423 * this value is defined in the standard to be between 200 and 500
2426 * sack_freq - This parameter contains the number of packets that must
2427 * be received before a sack is sent without waiting for the delay
2428 * timer to expire. The default value for this is 2, setting this
2429 * value to 1 will disable the delayed sack algorithm.
2432 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2433 char __user *optval, int optlen)
2435 struct sctp_sack_info params;
2436 struct sctp_transport *trans = NULL;
2437 struct sctp_association *asoc = NULL;
2438 struct sctp_sock *sp = sctp_sk(sk);
2440 if (optlen == sizeof(struct sctp_sack_info)) {
2441 if (copy_from_user(¶ms, optval, optlen))
2444 if (params.sack_delay == 0 && params.sack_freq == 0)
2446 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2447 printk(KERN_WARNING "SCTP: Use of struct sctp_assoc_value "
2448 "in delayed_ack socket option deprecated\n");
2449 printk(KERN_WARNING "SCTP: Use struct sctp_sack_info instead\n");
2450 if (copy_from_user(¶ms, optval, optlen))
2453 if (params.sack_delay == 0)
2454 params.sack_freq = 1;
2456 params.sack_freq = 0;
2460 /* Validate value parameter. */
2461 if (params.sack_delay > 500)
2464 /* Get association, if sack_assoc_id != 0 and the socket is a one
2465 * to many style socket, and an association was not found, then
2466 * the id was invalid.
2468 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2469 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2472 if (params.sack_delay) {
2475 msecs_to_jiffies(params.sack_delay);
2477 (asoc->param_flags & ~SPP_SACKDELAY) |
2478 SPP_SACKDELAY_ENABLE;
2480 sp->sackdelay = params.sack_delay;
2482 (sp->param_flags & ~SPP_SACKDELAY) |
2483 SPP_SACKDELAY_ENABLE;
2487 if (params.sack_freq == 1) {
2490 (asoc->param_flags & ~SPP_SACKDELAY) |
2491 SPP_SACKDELAY_DISABLE;
2494 (sp->param_flags & ~SPP_SACKDELAY) |
2495 SPP_SACKDELAY_DISABLE;
2497 } else if (params.sack_freq > 1) {
2499 asoc->sackfreq = params.sack_freq;
2501 (asoc->param_flags & ~SPP_SACKDELAY) |
2502 SPP_SACKDELAY_ENABLE;
2504 sp->sackfreq = params.sack_freq;
2506 (sp->param_flags & ~SPP_SACKDELAY) |
2507 SPP_SACKDELAY_ENABLE;
2511 /* If change is for association, also apply to each transport. */
2513 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2515 if (params.sack_delay) {
2517 msecs_to_jiffies(params.sack_delay);
2518 trans->param_flags =
2519 (trans->param_flags & ~SPP_SACKDELAY) |
2520 SPP_SACKDELAY_ENABLE;
2522 if (params.sack_freq == 1) {
2523 trans->param_flags =
2524 (trans->param_flags & ~SPP_SACKDELAY) |
2525 SPP_SACKDELAY_DISABLE;
2526 } else if (params.sack_freq > 1) {
2527 trans->sackfreq = params.sack_freq;
2528 trans->param_flags =
2529 (trans->param_flags & ~SPP_SACKDELAY) |
2530 SPP_SACKDELAY_ENABLE;
2538 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2540 * Applications can specify protocol parameters for the default association
2541 * initialization. The option name argument to setsockopt() and getsockopt()
2544 * Setting initialization parameters is effective only on an unconnected
2545 * socket (for UDP-style sockets only future associations are effected
2546 * by the change). With TCP-style sockets, this option is inherited by
2547 * sockets derived from a listener socket.
2549 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen)
2551 struct sctp_initmsg sinit;
2552 struct sctp_sock *sp = sctp_sk(sk);
2554 if (optlen != sizeof(struct sctp_initmsg))
2556 if (copy_from_user(&sinit, optval, optlen))
2559 if (sinit.sinit_num_ostreams)
2560 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2561 if (sinit.sinit_max_instreams)
2562 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2563 if (sinit.sinit_max_attempts)
2564 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2565 if (sinit.sinit_max_init_timeo)
2566 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2572 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2574 * Applications that wish to use the sendto() system call may wish to
2575 * specify a default set of parameters that would normally be supplied
2576 * through the inclusion of ancillary data. This socket option allows
2577 * such an application to set the default sctp_sndrcvinfo structure.
2578 * The application that wishes to use this socket option simply passes
2579 * in to this call the sctp_sndrcvinfo structure defined in Section
2580 * 5.2.2) The input parameters accepted by this call include
2581 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2582 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2583 * to this call if the caller is using the UDP model.
2585 static int sctp_setsockopt_default_send_param(struct sock *sk,
2586 char __user *optval, int optlen)
2588 struct sctp_sndrcvinfo info;
2589 struct sctp_association *asoc;
2590 struct sctp_sock *sp = sctp_sk(sk);
2592 if (optlen != sizeof(struct sctp_sndrcvinfo))
2594 if (copy_from_user(&info, optval, optlen))
2597 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2598 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2602 asoc->default_stream = info.sinfo_stream;
2603 asoc->default_flags = info.sinfo_flags;
2604 asoc->default_ppid = info.sinfo_ppid;
2605 asoc->default_context = info.sinfo_context;
2606 asoc->default_timetolive = info.sinfo_timetolive;
2608 sp->default_stream = info.sinfo_stream;
2609 sp->default_flags = info.sinfo_flags;
2610 sp->default_ppid = info.sinfo_ppid;
2611 sp->default_context = info.sinfo_context;
2612 sp->default_timetolive = info.sinfo_timetolive;
2618 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2620 * Requests that the local SCTP stack use the enclosed peer address as
2621 * the association primary. The enclosed address must be one of the
2622 * association peer's addresses.
2624 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2627 struct sctp_prim prim;
2628 struct sctp_transport *trans;
2630 if (optlen != sizeof(struct sctp_prim))
2633 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2636 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2640 sctp_assoc_set_primary(trans->asoc, trans);
2646 * 7.1.5 SCTP_NODELAY
2648 * Turn on/off any Nagle-like algorithm. This means that packets are
2649 * generally sent as soon as possible and no unnecessary delays are
2650 * introduced, at the cost of more packets in the network. Expects an
2651 * integer boolean flag.
2653 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2658 if (optlen < sizeof(int))
2660 if (get_user(val, (int __user *)optval))
2663 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2669 * 7.1.1 SCTP_RTOINFO
2671 * The protocol parameters used to initialize and bound retransmission
2672 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2673 * and modify these parameters.
2674 * All parameters are time values, in milliseconds. A value of 0, when
2675 * modifying the parameters, indicates that the current value should not
2679 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) {
2680 struct sctp_rtoinfo rtoinfo;
2681 struct sctp_association *asoc;
2683 if (optlen != sizeof (struct sctp_rtoinfo))
2686 if (copy_from_user(&rtoinfo, optval, optlen))
2689 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2691 /* Set the values to the specific association */
2692 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2696 if (rtoinfo.srto_initial != 0)
2698 msecs_to_jiffies(rtoinfo.srto_initial);
2699 if (rtoinfo.srto_max != 0)
2700 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2701 if (rtoinfo.srto_min != 0)
2702 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2704 /* If there is no association or the association-id = 0
2705 * set the values to the endpoint.
2707 struct sctp_sock *sp = sctp_sk(sk);
2709 if (rtoinfo.srto_initial != 0)
2710 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2711 if (rtoinfo.srto_max != 0)
2712 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2713 if (rtoinfo.srto_min != 0)
2714 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2722 * 7.1.2 SCTP_ASSOCINFO
2724 * This option is used to tune the maximum retransmission attempts
2725 * of the association.
2726 * Returns an error if the new association retransmission value is
2727 * greater than the sum of the retransmission value of the peer.
2728 * See [SCTP] for more information.
2731 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen)
2734 struct sctp_assocparams assocparams;
2735 struct sctp_association *asoc;
2737 if (optlen != sizeof(struct sctp_assocparams))
2739 if (copy_from_user(&assocparams, optval, optlen))
2742 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2744 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2747 /* Set the values to the specific association */
2749 if (assocparams.sasoc_asocmaxrxt != 0) {
2752 struct sctp_transport *peer_addr;
2754 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2756 path_sum += peer_addr->pathmaxrxt;
2760 /* Only validate asocmaxrxt if we have more than
2761 * one path/transport. We do this because path
2762 * retransmissions are only counted when we have more
2766 assocparams.sasoc_asocmaxrxt > path_sum)
2769 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2772 if (assocparams.sasoc_cookie_life != 0) {
2773 asoc->cookie_life.tv_sec =
2774 assocparams.sasoc_cookie_life / 1000;
2775 asoc->cookie_life.tv_usec =
2776 (assocparams.sasoc_cookie_life % 1000)
2780 /* Set the values to the endpoint */
2781 struct sctp_sock *sp = sctp_sk(sk);
2783 if (assocparams.sasoc_asocmaxrxt != 0)
2784 sp->assocparams.sasoc_asocmaxrxt =
2785 assocparams.sasoc_asocmaxrxt;
2786 if (assocparams.sasoc_cookie_life != 0)
2787 sp->assocparams.sasoc_cookie_life =
2788 assocparams.sasoc_cookie_life;
2794 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2796 * This socket option is a boolean flag which turns on or off mapped V4
2797 * addresses. If this option is turned on and the socket is type
2798 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2799 * If this option is turned off, then no mapping will be done of V4
2800 * addresses and a user will receive both PF_INET6 and PF_INET type
2801 * addresses on the socket.
2803 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen)
2806 struct sctp_sock *sp = sctp_sk(sk);
2808 if (optlen < sizeof(int))
2810 if (get_user(val, (int __user *)optval))
2821 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2822 * This option will get or set the maximum size to put in any outgoing
2823 * SCTP DATA chunk. If a message is larger than this size it will be
2824 * fragmented by SCTP into the specified size. Note that the underlying
2825 * SCTP implementation may fragment into smaller sized chunks when the
2826 * PMTU of the underlying association is smaller than the value set by
2827 * the user. The default value for this option is '0' which indicates
2828 * the user is NOT limiting fragmentation and only the PMTU will effect
2829 * SCTP's choice of DATA chunk size. Note also that values set larger
2830 * than the maximum size of an IP datagram will effectively let SCTP
2831 * control fragmentation (i.e. the same as setting this option to 0).
2833 * The following structure is used to access and modify this parameter:
2835 * struct sctp_assoc_value {
2836 * sctp_assoc_t assoc_id;
2837 * uint32_t assoc_value;
2840 * assoc_id: This parameter is ignored for one-to-one style sockets.
2841 * For one-to-many style sockets this parameter indicates which
2842 * association the user is performing an action upon. Note that if
2843 * this field's value is zero then the endpoints default value is
2844 * changed (effecting future associations only).
2845 * assoc_value: This parameter specifies the maximum size in bytes.
2847 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen)
2849 struct sctp_assoc_value params;
2850 struct sctp_association *asoc;
2851 struct sctp_sock *sp = sctp_sk(sk);
2854 if (optlen == sizeof(int)) {
2856 "SCTP: Use of int in maxseg socket option deprecated\n");
2858 "SCTP: Use struct sctp_assoc_value instead\n");
2859 if (copy_from_user(&val, optval, optlen))
2861 params.assoc_id = 0;
2862 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2863 if (copy_from_user(¶ms, optval, optlen))
2865 val = params.assoc_value;
2869 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2872 asoc = sctp_id2assoc(sk, params.assoc_id);
2873 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2878 val = asoc->pathmtu;
2879 val -= sp->pf->af->net_header_len;
2880 val -= sizeof(struct sctphdr) +
2881 sizeof(struct sctp_data_chunk);
2884 asoc->frag_point = val;
2886 sp->user_frag = val;
2888 /* Update the frag_point of the existing associations. */
2889 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
2890 asoc->frag_point = sctp_frag_point(sp, asoc->pathmtu);
2899 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2901 * Requests that the peer mark the enclosed address as the association
2902 * primary. The enclosed address must be one of the association's
2903 * locally bound addresses. The following structure is used to make a
2904 * set primary request:
2906 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2909 struct sctp_sock *sp;
2910 struct sctp_endpoint *ep;
2911 struct sctp_association *asoc = NULL;
2912 struct sctp_setpeerprim prim;
2913 struct sctp_chunk *chunk;
2919 if (!sctp_addip_enable)
2922 if (optlen != sizeof(struct sctp_setpeerprim))
2925 if (copy_from_user(&prim, optval, optlen))
2928 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2932 if (!asoc->peer.asconf_capable)
2935 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2938 if (!sctp_state(asoc, ESTABLISHED))
2941 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2942 return -EADDRNOTAVAIL;
2944 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2945 chunk = sctp_make_asconf_set_prim(asoc,
2946 (union sctp_addr *)&prim.sspp_addr);
2950 err = sctp_send_asconf(asoc, chunk);
2952 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2957 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2960 struct sctp_setadaptation adaptation;
2962 if (optlen != sizeof(struct sctp_setadaptation))
2964 if (copy_from_user(&adaptation, optval, optlen))
2967 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2973 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
2975 * The context field in the sctp_sndrcvinfo structure is normally only
2976 * used when a failed message is retrieved holding the value that was
2977 * sent down on the actual send call. This option allows the setting of
2978 * a default context on an association basis that will be received on
2979 * reading messages from the peer. This is especially helpful in the
2980 * one-2-many model for an application to keep some reference to an
2981 * internal state machine that is processing messages on the
2982 * association. Note that the setting of this value only effects
2983 * received messages from the peer and does not effect the value that is
2984 * saved with outbound messages.
2986 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
2989 struct sctp_assoc_value params;
2990 struct sctp_sock *sp;
2991 struct sctp_association *asoc;
2993 if (optlen != sizeof(struct sctp_assoc_value))
2995 if (copy_from_user(¶ms, optval, optlen))
3000 if (params.assoc_id != 0) {
3001 asoc = sctp_id2assoc(sk, params.assoc_id);
3004 asoc->default_rcv_context = params.assoc_value;
3006 sp->default_rcv_context = params.assoc_value;
3013 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3015 * This options will at a minimum specify if the implementation is doing
3016 * fragmented interleave. Fragmented interleave, for a one to many
3017 * socket, is when subsequent calls to receive a message may return
3018 * parts of messages from different associations. Some implementations
3019 * may allow you to turn this value on or off. If so, when turned off,
3020 * no fragment interleave will occur (which will cause a head of line
3021 * blocking amongst multiple associations sharing the same one to many
3022 * socket). When this option is turned on, then each receive call may
3023 * come from a different association (thus the user must receive data
3024 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3025 * association each receive belongs to.
3027 * This option takes a boolean value. A non-zero value indicates that
3028 * fragmented interleave is on. A value of zero indicates that
3029 * fragmented interleave is off.
3031 * Note that it is important that an implementation that allows this
3032 * option to be turned on, have it off by default. Otherwise an unaware
3033 * application using the one to many model may become confused and act
3036 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3037 char __user *optval,
3042 if (optlen != sizeof(int))
3044 if (get_user(val, (int __user *)optval))
3047 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3053 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3054 * (SCTP_PARTIAL_DELIVERY_POINT)
3056 * This option will set or get the SCTP partial delivery point. This
3057 * point is the size of a message where the partial delivery API will be
3058 * invoked to help free up rwnd space for the peer. Setting this to a
3059 * lower value will cause partial deliveries to happen more often. The
3060 * calls argument is an integer that sets or gets the partial delivery
3061 * point. Note also that the call will fail if the user attempts to set
3062 * this value larger than the socket receive buffer size.
3064 * Note that any single message having a length smaller than or equal to
3065 * the SCTP partial delivery point will be delivered in one single read
3066 * call as long as the user provided buffer is large enough to hold the
3069 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3070 char __user *optval,
3075 if (optlen != sizeof(u32))
3077 if (get_user(val, (int __user *)optval))
3080 /* Note: We double the receive buffer from what the user sets
3081 * it to be, also initial rwnd is based on rcvbuf/2.
3083 if (val > (sk->sk_rcvbuf >> 1))
3086 sctp_sk(sk)->pd_point = val;
3088 return 0; /* is this the right error code? */
3092 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3094 * This option will allow a user to change the maximum burst of packets
3095 * that can be emitted by this association. Note that the default value
3096 * is 4, and some implementations may restrict this setting so that it
3097 * can only be lowered.
3099 * NOTE: This text doesn't seem right. Do this on a socket basis with
3100 * future associations inheriting the socket value.
3102 static int sctp_setsockopt_maxburst(struct sock *sk,
3103 char __user *optval,
3106 struct sctp_assoc_value params;
3107 struct sctp_sock *sp;
3108 struct sctp_association *asoc;
3112 if (optlen == sizeof(int)) {
3114 "SCTP: Use of int in max_burst socket option deprecated\n");
3116 "SCTP: Use struct sctp_assoc_value instead\n");
3117 if (copy_from_user(&val, optval, optlen))
3119 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3120 if (copy_from_user(¶ms, optval, optlen))
3122 val = params.assoc_value;
3123 assoc_id = params.assoc_id;
3129 if (assoc_id != 0) {
3130 asoc = sctp_id2assoc(sk, assoc_id);
3133 asoc->max_burst = val;
3135 sp->max_burst = val;
3141 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3143 * This set option adds a chunk type that the user is requesting to be
3144 * received only in an authenticated way. Changes to the list of chunks
3145 * will only effect future associations on the socket.
3147 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3148 char __user *optval,
3151 struct sctp_authchunk val;
3153 if (!sctp_auth_enable)
3156 if (optlen != sizeof(struct sctp_authchunk))
3158 if (copy_from_user(&val, optval, optlen))
3161 switch (val.sauth_chunk) {
3163 case SCTP_CID_INIT_ACK:
3164 case SCTP_CID_SHUTDOWN_COMPLETE:
3169 /* add this chunk id to the endpoint */
3170 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3174 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3176 * This option gets or sets the list of HMAC algorithms that the local
3177 * endpoint requires the peer to use.
3179 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3180 char __user *optval,
3183 struct sctp_hmacalgo *hmacs;
3187 if (!sctp_auth_enable)
3190 if (optlen < sizeof(struct sctp_hmacalgo))
3193 hmacs = kmalloc(optlen, GFP_KERNEL);
3197 if (copy_from_user(hmacs, optval, optlen)) {
3202 idents = hmacs->shmac_num_idents;
3203 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3204 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3209 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3216 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3218 * This option will set a shared secret key which is used to build an
3219 * association shared key.
3221 static int sctp_setsockopt_auth_key(struct sock *sk,
3222 char __user *optval,
3225 struct sctp_authkey *authkey;
3226 struct sctp_association *asoc;
3229 if (!sctp_auth_enable)
3232 if (optlen <= sizeof(struct sctp_authkey))
3235 authkey = kmalloc(optlen, GFP_KERNEL);
3239 if (copy_from_user(authkey, optval, optlen)) {
3244 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3249 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3250 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3255 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3262 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3264 * This option will get or set the active shared key to be used to build
3265 * the association shared key.
3267 static int sctp_setsockopt_active_key(struct sock *sk,
3268 char __user *optval,
3271 struct sctp_authkeyid val;
3272 struct sctp_association *asoc;
3274 if (!sctp_auth_enable)
3277 if (optlen != sizeof(struct sctp_authkeyid))
3279 if (copy_from_user(&val, optval, optlen))
3282 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3283 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3286 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3287 val.scact_keynumber);
3291 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3293 * This set option will delete a shared secret key from use.
3295 static int sctp_setsockopt_del_key(struct sock *sk,
3296 char __user *optval,
3299 struct sctp_authkeyid val;
3300 struct sctp_association *asoc;
3302 if (!sctp_auth_enable)
3305 if (optlen != sizeof(struct sctp_authkeyid))
3307 if (copy_from_user(&val, optval, optlen))
3310 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3311 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3314 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3315 val.scact_keynumber);
3320 /* API 6.2 setsockopt(), getsockopt()
3322 * Applications use setsockopt() and getsockopt() to set or retrieve
3323 * socket options. Socket options are used to change the default
3324 * behavior of sockets calls. They are described in Section 7.
3328 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3329 * int __user *optlen);
3330 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3333 * sd - the socket descript.
3334 * level - set to IPPROTO_SCTP for all SCTP options.
3335 * optname - the option name.
3336 * optval - the buffer to store the value of the option.
3337 * optlen - the size of the buffer.
3339 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3340 char __user *optval, int optlen)
3344 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3347 /* I can hardly begin to describe how wrong this is. This is
3348 * so broken as to be worse than useless. The API draft
3349 * REALLY is NOT helpful here... I am not convinced that the
3350 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3351 * are at all well-founded.
3353 if (level != SOL_SCTP) {
3354 struct sctp_af *af = sctp_sk(sk)->pf->af;
3355 retval = af->setsockopt(sk, level, optname, optval, optlen);
3362 case SCTP_SOCKOPT_BINDX_ADD:
3363 /* 'optlen' is the size of the addresses buffer. */
3364 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3365 optlen, SCTP_BINDX_ADD_ADDR);
3368 case SCTP_SOCKOPT_BINDX_REM:
3369 /* 'optlen' is the size of the addresses buffer. */
3370 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3371 optlen, SCTP_BINDX_REM_ADDR);
3374 case SCTP_SOCKOPT_CONNECTX_OLD:
3375 /* 'optlen' is the size of the addresses buffer. */
3376 retval = sctp_setsockopt_connectx_old(sk,
3377 (struct sockaddr __user *)optval,
3381 case SCTP_SOCKOPT_CONNECTX:
3382 /* 'optlen' is the size of the addresses buffer. */
3383 retval = sctp_setsockopt_connectx(sk,
3384 (struct sockaddr __user *)optval,
3388 case SCTP_DISABLE_FRAGMENTS:
3389 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3393 retval = sctp_setsockopt_events(sk, optval, optlen);
3396 case SCTP_AUTOCLOSE:
3397 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3400 case SCTP_PEER_ADDR_PARAMS:
3401 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3404 case SCTP_DELAYED_ACK:
3405 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3407 case SCTP_PARTIAL_DELIVERY_POINT:
3408 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3412 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3414 case SCTP_DEFAULT_SEND_PARAM:
3415 retval = sctp_setsockopt_default_send_param(sk, optval,
3418 case SCTP_PRIMARY_ADDR:
3419 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3421 case SCTP_SET_PEER_PRIMARY_ADDR:
3422 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3425 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3428 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3430 case SCTP_ASSOCINFO:
3431 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3433 case SCTP_I_WANT_MAPPED_V4_ADDR:
3434 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3437 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3439 case SCTP_ADAPTATION_LAYER:
3440 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3443 retval = sctp_setsockopt_context(sk, optval, optlen);
3445 case SCTP_FRAGMENT_INTERLEAVE:
3446 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3448 case SCTP_MAX_BURST:
3449 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3451 case SCTP_AUTH_CHUNK:
3452 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3454 case SCTP_HMAC_IDENT:
3455 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3458 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3460 case SCTP_AUTH_ACTIVE_KEY:
3461 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3463 case SCTP_AUTH_DELETE_KEY:
3464 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3467 retval = -ENOPROTOOPT;
3471 sctp_release_sock(sk);
3477 /* API 3.1.6 connect() - UDP Style Syntax
3479 * An application may use the connect() call in the UDP model to initiate an
3480 * association without sending data.
3484 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3486 * sd: the socket descriptor to have a new association added to.
3488 * nam: the address structure (either struct sockaddr_in or struct
3489 * sockaddr_in6 defined in RFC2553 [7]).
3491 * len: the size of the address.
3493 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3501 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3502 __func__, sk, addr, addr_len);
3504 /* Validate addr_len before calling common connect/connectx routine. */
3505 af = sctp_get_af_specific(addr->sa_family);
3506 if (!af || addr_len < af->sockaddr_len) {
3509 /* Pass correct addr len to common routine (so it knows there
3510 * is only one address being passed.
3512 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3515 sctp_release_sock(sk);
3519 /* FIXME: Write comments. */
3520 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3522 return -EOPNOTSUPP; /* STUB */
3525 /* 4.1.4 accept() - TCP Style Syntax
3527 * Applications use accept() call to remove an established SCTP
3528 * association from the accept queue of the endpoint. A new socket
3529 * descriptor will be returned from accept() to represent the newly
3530 * formed association.
3532 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3534 struct sctp_sock *sp;
3535 struct sctp_endpoint *ep;
3536 struct sock *newsk = NULL;
3537 struct sctp_association *asoc;
3546 if (!sctp_style(sk, TCP)) {
3547 error = -EOPNOTSUPP;
3551 if (!sctp_sstate(sk, LISTENING)) {
3556 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3558 error = sctp_wait_for_accept(sk, timeo);
3562 /* We treat the list of associations on the endpoint as the accept
3563 * queue and pick the first association on the list.
3565 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3567 newsk = sp->pf->create_accept_sk(sk, asoc);
3573 /* Populate the fields of the newsk from the oldsk and migrate the
3574 * asoc to the newsk.
3576 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3579 sctp_release_sock(sk);
3584 /* The SCTP ioctl handler. */
3585 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3587 return -ENOIOCTLCMD;
3590 /* This is the function which gets called during socket creation to
3591 * initialized the SCTP-specific portion of the sock.
3592 * The sock structure should already be zero-filled memory.
3594 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3596 struct sctp_endpoint *ep;
3597 struct sctp_sock *sp;
3599 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3603 /* Initialize the SCTP per socket area. */
3604 switch (sk->sk_type) {
3605 case SOCK_SEQPACKET:
3606 sp->type = SCTP_SOCKET_UDP;
3609 sp->type = SCTP_SOCKET_TCP;
3612 return -ESOCKTNOSUPPORT;
3615 /* Initialize default send parameters. These parameters can be
3616 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3618 sp->default_stream = 0;
3619 sp->default_ppid = 0;
3620 sp->default_flags = 0;
3621 sp->default_context = 0;
3622 sp->default_timetolive = 0;
3624 sp->default_rcv_context = 0;
3625 sp->max_burst = sctp_max_burst;
3627 /* Initialize default setup parameters. These parameters
3628 * can be modified with the SCTP_INITMSG socket option or
3629 * overridden by the SCTP_INIT CMSG.
3631 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3632 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3633 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3634 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3636 /* Initialize default RTO related parameters. These parameters can
3637 * be modified for with the SCTP_RTOINFO socket option.
3639 sp->rtoinfo.srto_initial = sctp_rto_initial;
3640 sp->rtoinfo.srto_max = sctp_rto_max;
3641 sp->rtoinfo.srto_min = sctp_rto_min;
3643 /* Initialize default association related parameters. These parameters
3644 * can be modified with the SCTP_ASSOCINFO socket option.
3646 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3647 sp->assocparams.sasoc_number_peer_destinations = 0;
3648 sp->assocparams.sasoc_peer_rwnd = 0;
3649 sp->assocparams.sasoc_local_rwnd = 0;
3650 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3652 /* Initialize default event subscriptions. By default, all the
3655 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3657 /* Default Peer Address Parameters. These defaults can
3658 * be modified via SCTP_PEER_ADDR_PARAMS
3660 sp->hbinterval = sctp_hb_interval;
3661 sp->pathmaxrxt = sctp_max_retrans_path;
3662 sp->pathmtu = 0; // allow default discovery
3663 sp->sackdelay = sctp_sack_timeout;
3665 sp->param_flags = SPP_HB_ENABLE |
3667 SPP_SACKDELAY_ENABLE;
3669 /* If enabled no SCTP message fragmentation will be performed.
3670 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3672 sp->disable_fragments = 0;
3674 /* Enable Nagle algorithm by default. */
3677 /* Enable by default. */
3680 /* Auto-close idle associations after the configured
3681 * number of seconds. A value of 0 disables this
3682 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3683 * for UDP-style sockets only.
3687 /* User specified fragmentation limit. */
3690 sp->adaptation_ind = 0;
3692 sp->pf = sctp_get_pf_specific(sk->sk_family);
3694 /* Control variables for partial data delivery. */
3695 atomic_set(&sp->pd_mode, 0);
3696 skb_queue_head_init(&sp->pd_lobby);
3697 sp->frag_interleave = 0;
3699 /* Create a per socket endpoint structure. Even if we
3700 * change the data structure relationships, this may still
3701 * be useful for storing pre-connect address information.
3703 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3710 SCTP_DBG_OBJCNT_INC(sock);
3711 percpu_counter_inc(&sctp_sockets_allocated);
3714 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3720 /* Cleanup any SCTP per socket resources. */
3721 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3723 struct sctp_endpoint *ep;
3725 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3727 /* Release our hold on the endpoint. */
3728 ep = sctp_sk(sk)->ep;
3729 sctp_endpoint_free(ep);
3730 percpu_counter_dec(&sctp_sockets_allocated);
3732 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
3736 /* API 4.1.7 shutdown() - TCP Style Syntax
3737 * int shutdown(int socket, int how);
3739 * sd - the socket descriptor of the association to be closed.
3740 * how - Specifies the type of shutdown. The values are
3743 * Disables further receive operations. No SCTP
3744 * protocol action is taken.
3746 * Disables further send operations, and initiates
3747 * the SCTP shutdown sequence.
3749 * Disables further send and receive operations
3750 * and initiates the SCTP shutdown sequence.
3752 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3754 struct sctp_endpoint *ep;
3755 struct sctp_association *asoc;
3757 if (!sctp_style(sk, TCP))
3760 if (how & SEND_SHUTDOWN) {
3761 ep = sctp_sk(sk)->ep;
3762 if (!list_empty(&ep->asocs)) {
3763 asoc = list_entry(ep->asocs.next,
3764 struct sctp_association, asocs);
3765 sctp_primitive_SHUTDOWN(asoc, NULL);
3770 /* 7.2.1 Association Status (SCTP_STATUS)
3772 * Applications can retrieve current status information about an
3773 * association, including association state, peer receiver window size,
3774 * number of unacked data chunks, and number of data chunks pending
3775 * receipt. This information is read-only.
3777 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3778 char __user *optval,
3781 struct sctp_status status;
3782 struct sctp_association *asoc = NULL;
3783 struct sctp_transport *transport;
3784 sctp_assoc_t associd;
3787 if (len < sizeof(status)) {
3792 len = sizeof(status);
3793 if (copy_from_user(&status, optval, len)) {
3798 associd = status.sstat_assoc_id;
3799 asoc = sctp_id2assoc(sk, associd);
3805 transport = asoc->peer.primary_path;
3807 status.sstat_assoc_id = sctp_assoc2id(asoc);
3808 status.sstat_state = asoc->state;
3809 status.sstat_rwnd = asoc->peer.rwnd;
3810 status.sstat_unackdata = asoc->unack_data;
3812 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3813 status.sstat_instrms = asoc->c.sinit_max_instreams;
3814 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3815 status.sstat_fragmentation_point = asoc->frag_point;
3816 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3817 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3818 transport->af_specific->sockaddr_len);
3819 /* Map ipv4 address into v4-mapped-on-v6 address. */
3820 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3821 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3822 status.sstat_primary.spinfo_state = transport->state;
3823 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3824 status.sstat_primary.spinfo_srtt = transport->srtt;
3825 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3826 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3828 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3829 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3831 if (put_user(len, optlen)) {
3836 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3837 len, status.sstat_state, status.sstat_rwnd,
3838 status.sstat_assoc_id);
3840 if (copy_to_user(optval, &status, len)) {
3850 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3852 * Applications can retrieve information about a specific peer address
3853 * of an association, including its reachability state, congestion
3854 * window, and retransmission timer values. This information is
3857 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3858 char __user *optval,
3861 struct sctp_paddrinfo pinfo;
3862 struct sctp_transport *transport;
3865 if (len < sizeof(pinfo)) {
3870 len = sizeof(pinfo);
3871 if (copy_from_user(&pinfo, optval, len)) {
3876 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3877 pinfo.spinfo_assoc_id);
3881 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3882 pinfo.spinfo_state = transport->state;
3883 pinfo.spinfo_cwnd = transport->cwnd;
3884 pinfo.spinfo_srtt = transport->srtt;
3885 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3886 pinfo.spinfo_mtu = transport->pathmtu;
3888 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3889 pinfo.spinfo_state = SCTP_ACTIVE;
3891 if (put_user(len, optlen)) {
3896 if (copy_to_user(optval, &pinfo, len)) {
3905 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3907 * This option is a on/off flag. If enabled no SCTP message
3908 * fragmentation will be performed. Instead if a message being sent
3909 * exceeds the current PMTU size, the message will NOT be sent and
3910 * instead a error will be indicated to the user.
3912 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3913 char __user *optval, int __user *optlen)
3917 if (len < sizeof(int))
3921 val = (sctp_sk(sk)->disable_fragments == 1);
3922 if (put_user(len, optlen))
3924 if (copy_to_user(optval, &val, len))
3929 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3931 * This socket option is used to specify various notifications and
3932 * ancillary data the user wishes to receive.
3934 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3937 if (len < sizeof(struct sctp_event_subscribe))
3939 len = sizeof(struct sctp_event_subscribe);
3940 if (put_user(len, optlen))
3942 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3947 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3949 * This socket option is applicable to the UDP-style socket only. When
3950 * set it will cause associations that are idle for more than the
3951 * specified number of seconds to automatically close. An association
3952 * being idle is defined an association that has NOT sent or received
3953 * user data. The special value of '0' indicates that no automatic
3954 * close of any associations should be performed. The option expects an
3955 * integer defining the number of seconds of idle time before an
3956 * association is closed.
3958 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3960 /* Applicable to UDP-style socket only */
3961 if (sctp_style(sk, TCP))
3963 if (len < sizeof(int))
3966 if (put_user(len, optlen))
3968 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
3973 /* Helper routine to branch off an association to a new socket. */
3974 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3975 struct socket **sockp)
3977 struct sock *sk = asoc->base.sk;
3978 struct socket *sock;
3982 /* An association cannot be branched off from an already peeled-off
3983 * socket, nor is this supported for tcp style sockets.
3985 if (!sctp_style(sk, UDP))
3988 /* Create a new socket. */
3989 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3993 sctp_copy_sock(sock->sk, sk, asoc);
3995 /* Make peeled-off sockets more like 1-1 accepted sockets.
3996 * Set the daddr and initialize id to something more random
3998 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
3999 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4001 /* Populate the fields of the newsk from the oldsk and migrate the
4002 * asoc to the newsk.
4004 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4011 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4013 sctp_peeloff_arg_t peeloff;
4014 struct socket *newsock;
4016 struct sctp_association *asoc;
4018 if (len < sizeof(sctp_peeloff_arg_t))
4020 len = sizeof(sctp_peeloff_arg_t);
4021 if (copy_from_user(&peeloff, optval, len))
4024 asoc = sctp_id2assoc(sk, peeloff.associd);
4030 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
4032 retval = sctp_do_peeloff(asoc, &newsock);
4036 /* Map the socket to an unused fd that can be returned to the user. */
4037 retval = sock_map_fd(newsock, 0);
4039 sock_release(newsock);
4043 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
4044 __func__, sk, asoc, newsock->sk, retval);
4046 /* Return the fd mapped to the new socket. */
4047 peeloff.sd = retval;
4048 if (put_user(len, optlen))
4050 if (copy_to_user(optval, &peeloff, len))
4057 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4059 * Applications can enable or disable heartbeats for any peer address of
4060 * an association, modify an address's heartbeat interval, force a
4061 * heartbeat to be sent immediately, and adjust the address's maximum
4062 * number of retransmissions sent before an address is considered
4063 * unreachable. The following structure is used to access and modify an
4064 * address's parameters:
4066 * struct sctp_paddrparams {
4067 * sctp_assoc_t spp_assoc_id;
4068 * struct sockaddr_storage spp_address;
4069 * uint32_t spp_hbinterval;
4070 * uint16_t spp_pathmaxrxt;
4071 * uint32_t spp_pathmtu;
4072 * uint32_t spp_sackdelay;
4073 * uint32_t spp_flags;
4076 * spp_assoc_id - (one-to-many style socket) This is filled in the
4077 * application, and identifies the association for
4079 * spp_address - This specifies which address is of interest.
4080 * spp_hbinterval - This contains the value of the heartbeat interval,
4081 * in milliseconds. If a value of zero
4082 * is present in this field then no changes are to
4083 * be made to this parameter.
4084 * spp_pathmaxrxt - This contains the maximum number of
4085 * retransmissions before this address shall be
4086 * considered unreachable. If a value of zero
4087 * is present in this field then no changes are to
4088 * be made to this parameter.
4089 * spp_pathmtu - When Path MTU discovery is disabled the value
4090 * specified here will be the "fixed" path mtu.
4091 * Note that if the spp_address field is empty
4092 * then all associations on this address will
4093 * have this fixed path mtu set upon them.
4095 * spp_sackdelay - When delayed sack is enabled, this value specifies
4096 * the number of milliseconds that sacks will be delayed
4097 * for. This value will apply to all addresses of an
4098 * association if the spp_address field is empty. Note
4099 * also, that if delayed sack is enabled and this
4100 * value is set to 0, no change is made to the last
4101 * recorded delayed sack timer value.
4103 * spp_flags - These flags are used to control various features
4104 * on an association. The flag field may contain
4105 * zero or more of the following options.
4107 * SPP_HB_ENABLE - Enable heartbeats on the
4108 * specified address. Note that if the address
4109 * field is empty all addresses for the association
4110 * have heartbeats enabled upon them.
4112 * SPP_HB_DISABLE - Disable heartbeats on the
4113 * speicifed address. Note that if the address
4114 * field is empty all addresses for the association
4115 * will have their heartbeats disabled. Note also
4116 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4117 * mutually exclusive, only one of these two should
4118 * be specified. Enabling both fields will have
4119 * undetermined results.
4121 * SPP_HB_DEMAND - Request a user initiated heartbeat
4122 * to be made immediately.
4124 * SPP_PMTUD_ENABLE - This field will enable PMTU
4125 * discovery upon the specified address. Note that
4126 * if the address feild is empty then all addresses
4127 * on the association are effected.
4129 * SPP_PMTUD_DISABLE - This field will disable PMTU
4130 * discovery upon the specified address. Note that
4131 * if the address feild is empty then all addresses
4132 * on the association are effected. Not also that
4133 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4134 * exclusive. Enabling both will have undetermined
4137 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4138 * on delayed sack. The time specified in spp_sackdelay
4139 * is used to specify the sack delay for this address. Note
4140 * that if spp_address is empty then all addresses will
4141 * enable delayed sack and take on the sack delay
4142 * value specified in spp_sackdelay.
4143 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4144 * off delayed sack. If the spp_address field is blank then
4145 * delayed sack is disabled for the entire association. Note
4146 * also that this field is mutually exclusive to
4147 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4150 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4151 char __user *optval, int __user *optlen)
4153 struct sctp_paddrparams params;
4154 struct sctp_transport *trans = NULL;
4155 struct sctp_association *asoc = NULL;
4156 struct sctp_sock *sp = sctp_sk(sk);
4158 if (len < sizeof(struct sctp_paddrparams))
4160 len = sizeof(struct sctp_paddrparams);
4161 if (copy_from_user(¶ms, optval, len))
4164 /* If an address other than INADDR_ANY is specified, and
4165 * no transport is found, then the request is invalid.
4167 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
4168 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4169 params.spp_assoc_id);
4171 SCTP_DEBUG_PRINTK("Failed no transport\n");
4176 /* Get association, if assoc_id != 0 and the socket is a one
4177 * to many style socket, and an association was not found, then
4178 * the id was invalid.
4180 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4181 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4182 SCTP_DEBUG_PRINTK("Failed no association\n");
4187 /* Fetch transport values. */
4188 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4189 params.spp_pathmtu = trans->pathmtu;
4190 params.spp_pathmaxrxt = trans->pathmaxrxt;
4191 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4193 /*draft-11 doesn't say what to return in spp_flags*/
4194 params.spp_flags = trans->param_flags;
4196 /* Fetch association values. */
4197 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4198 params.spp_pathmtu = asoc->pathmtu;
4199 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4200 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4202 /*draft-11 doesn't say what to return in spp_flags*/
4203 params.spp_flags = asoc->param_flags;
4205 /* Fetch socket values. */
4206 params.spp_hbinterval = sp->hbinterval;
4207 params.spp_pathmtu = sp->pathmtu;
4208 params.spp_sackdelay = sp->sackdelay;
4209 params.spp_pathmaxrxt = sp->pathmaxrxt;
4211 /*draft-11 doesn't say what to return in spp_flags*/
4212 params.spp_flags = sp->param_flags;
4215 if (copy_to_user(optval, ¶ms, len))
4218 if (put_user(len, optlen))
4225 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4227 * This option will effect the way delayed acks are performed. This
4228 * option allows you to get or set the delayed ack time, in
4229 * milliseconds. It also allows changing the delayed ack frequency.
4230 * Changing the frequency to 1 disables the delayed sack algorithm. If
4231 * the assoc_id is 0, then this sets or gets the endpoints default
4232 * values. If the assoc_id field is non-zero, then the set or get
4233 * effects the specified association for the one to many model (the
4234 * assoc_id field is ignored by the one to one model). Note that if
4235 * sack_delay or sack_freq are 0 when setting this option, then the
4236 * current values will remain unchanged.
4238 * struct sctp_sack_info {
4239 * sctp_assoc_t sack_assoc_id;
4240 * uint32_t sack_delay;
4241 * uint32_t sack_freq;
4244 * sack_assoc_id - This parameter, indicates which association the user
4245 * is performing an action upon. Note that if this field's value is
4246 * zero then the endpoints default value is changed (effecting future
4247 * associations only).
4249 * sack_delay - This parameter contains the number of milliseconds that
4250 * the user is requesting the delayed ACK timer be set to. Note that
4251 * this value is defined in the standard to be between 200 and 500
4254 * sack_freq - This parameter contains the number of packets that must
4255 * be received before a sack is sent without waiting for the delay
4256 * timer to expire. The default value for this is 2, setting this
4257 * value to 1 will disable the delayed sack algorithm.
4259 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4260 char __user *optval,
4263 struct sctp_sack_info params;
4264 struct sctp_association *asoc = NULL;
4265 struct sctp_sock *sp = sctp_sk(sk);
4267 if (len >= sizeof(struct sctp_sack_info)) {
4268 len = sizeof(struct sctp_sack_info);
4270 if (copy_from_user(¶ms, optval, len))
4272 } else if (len == sizeof(struct sctp_assoc_value)) {
4273 printk(KERN_WARNING "SCTP: Use of struct sctp_assoc_value "
4274 "in delayed_ack socket option deprecated\n");
4275 printk(KERN_WARNING "SCTP: Use struct sctp_sack_info instead\n");
4276 if (copy_from_user(¶ms, optval, len))
4281 /* Get association, if sack_assoc_id != 0 and the socket is a one
4282 * to many style socket, and an association was not found, then
4283 * the id was invalid.
4285 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4286 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4290 /* Fetch association values. */
4291 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4292 params.sack_delay = jiffies_to_msecs(
4294 params.sack_freq = asoc->sackfreq;
4297 params.sack_delay = 0;
4298 params.sack_freq = 1;
4301 /* Fetch socket values. */
4302 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4303 params.sack_delay = sp->sackdelay;
4304 params.sack_freq = sp->sackfreq;
4306 params.sack_delay = 0;
4307 params.sack_freq = 1;
4311 if (copy_to_user(optval, ¶ms, len))
4314 if (put_user(len, optlen))
4320 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4322 * Applications can specify protocol parameters for the default association
4323 * initialization. The option name argument to setsockopt() and getsockopt()
4326 * Setting initialization parameters is effective only on an unconnected
4327 * socket (for UDP-style sockets only future associations are effected
4328 * by the change). With TCP-style sockets, this option is inherited by
4329 * sockets derived from a listener socket.
4331 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4333 if (len < sizeof(struct sctp_initmsg))
4335 len = sizeof(struct sctp_initmsg);
4336 if (put_user(len, optlen))
4338 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4343 static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
4344 char __user *optval,
4348 struct sctp_association *asoc;
4349 struct list_head *pos;
4352 if (len < sizeof(sctp_assoc_t))
4355 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4358 printk(KERN_WARNING "SCTP: Use of SCTP_GET_PEER_ADDRS_NUM_OLD "
4359 "socket option deprecated\n");
4360 /* For UDP-style sockets, id specifies the association to query. */
4361 asoc = sctp_id2assoc(sk, id);
4365 list_for_each(pos, &asoc->peer.transport_addr_list) {
4373 * Old API for getting list of peer addresses. Does not work for 32-bit
4374 * programs running on a 64-bit kernel
4376 static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
4377 char __user *optval,
4380 struct sctp_association *asoc;
4382 struct sctp_getaddrs_old getaddrs;
4383 struct sctp_transport *from;
4385 union sctp_addr temp;
4386 struct sctp_sock *sp = sctp_sk(sk);
4389 if (len < sizeof(struct sctp_getaddrs_old))
4392 len = sizeof(struct sctp_getaddrs_old);
4394 if (copy_from_user(&getaddrs, optval, len))
4397 if (getaddrs.addr_num <= 0) return -EINVAL;
4399 printk(KERN_WARNING "SCTP: Use of SCTP_GET_PEER_ADDRS_OLD "
4400 "socket option deprecated\n");
4402 /* For UDP-style sockets, id specifies the association to query. */
4403 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4407 to = (void __user *)getaddrs.addrs;
4408 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4410 memcpy(&temp, &from->ipaddr, sizeof(temp));
4411 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4412 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4413 if (copy_to_user(to, &temp, addrlen))
4417 if (cnt >= getaddrs.addr_num) break;
4419 getaddrs.addr_num = cnt;
4420 if (put_user(len, optlen))
4422 if (copy_to_user(optval, &getaddrs, len))
4428 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4429 char __user *optval, int __user *optlen)
4431 struct sctp_association *asoc;
4433 struct sctp_getaddrs getaddrs;
4434 struct sctp_transport *from;
4436 union sctp_addr temp;
4437 struct sctp_sock *sp = sctp_sk(sk);
4442 if (len < sizeof(struct sctp_getaddrs))
4445 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4448 /* For UDP-style sockets, id specifies the association to query. */
4449 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4453 to = optval + offsetof(struct sctp_getaddrs,addrs);
4454 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4456 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4458 memcpy(&temp, &from->ipaddr, sizeof(temp));
4459 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4460 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4461 if (space_left < addrlen)
4463 if (copy_to_user(to, &temp, addrlen))
4467 space_left -= addrlen;
4470 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4472 bytes_copied = ((char __user *)to) - optval;
4473 if (put_user(bytes_copied, optlen))
4479 static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
4480 char __user *optval,
4484 struct sctp_bind_addr *bp;
4485 struct sctp_association *asoc;
4486 struct sctp_sockaddr_entry *addr;
4489 if (len < sizeof(sctp_assoc_t))
4492 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4495 printk(KERN_WARNING "SCTP: Use of SCTP_GET_LOCAL_ADDRS_NUM_OLD "
4496 "socket option deprecated\n");
4499 * For UDP-style sockets, id specifies the association to query.
4500 * If the id field is set to the value '0' then the locally bound
4501 * addresses are returned without regard to any particular
4505 bp = &sctp_sk(sk)->ep->base.bind_addr;
4507 asoc = sctp_id2assoc(sk, id);
4510 bp = &asoc->base.bind_addr;
4513 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
4514 * addresses from the global local address list.
4516 if (sctp_list_single_entry(&bp->address_list)) {
4517 addr = list_entry(bp->address_list.next,
4518 struct sctp_sockaddr_entry, list);
4519 if (sctp_is_any(sk, &addr->a)) {
4521 list_for_each_entry_rcu(addr,
4522 &sctp_local_addr_list, list) {
4526 if ((PF_INET == sk->sk_family) &&
4527 (AF_INET6 == addr->a.sa.sa_family))
4530 if ((PF_INET6 == sk->sk_family) &&
4531 inet_v6_ipv6only(sk) &&
4532 (AF_INET == addr->a.sa.sa_family))
4544 /* Protection on the bound address list is not needed,
4545 * since in the socket option context we hold the socket lock,
4546 * so there is no way that the bound address list can change.
4548 list_for_each_entry(addr, &bp->address_list, list) {
4555 /* Helper function that copies local addresses to user and returns the number
4556 * of addresses copied.
4558 static int sctp_copy_laddrs_old(struct sock *sk, __u16 port,
4559 int max_addrs, void *to,
4562 struct sctp_sockaddr_entry *addr;
4563 union sctp_addr temp;
4568 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4572 if ((PF_INET == sk->sk_family) &&
4573 (AF_INET6 == addr->a.sa.sa_family))
4575 if ((PF_INET6 == sk->sk_family) &&
4576 inet_v6_ipv6only(sk) &&
4577 (AF_INET == addr->a.sa.sa_family))
4579 memcpy(&temp, &addr->a, sizeof(temp));
4580 if (!temp.v4.sin_port)
4581 temp.v4.sin_port = htons(port);
4583 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4585 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4586 memcpy(to, &temp, addrlen);
4589 *bytes_copied += addrlen;
4591 if (cnt >= max_addrs) break;
4598 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4599 size_t space_left, int *bytes_copied)
4601 struct sctp_sockaddr_entry *addr;
4602 union sctp_addr temp;
4607 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4611 if ((PF_INET == sk->sk_family) &&
4612 (AF_INET6 == addr->a.sa.sa_family))
4614 if ((PF_INET6 == sk->sk_family) &&
4615 inet_v6_ipv6only(sk) &&
4616 (AF_INET == addr->a.sa.sa_family))
4618 memcpy(&temp, &addr->a, sizeof(temp));
4619 if (!temp.v4.sin_port)
4620 temp.v4.sin_port = htons(port);
4622 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4624 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4625 if (space_left < addrlen) {
4629 memcpy(to, &temp, addrlen);
4633 space_left -= addrlen;
4634 *bytes_copied += addrlen;
4641 /* Old API for getting list of local addresses. Does not work for 32-bit
4642 * programs running on a 64-bit kernel
4644 static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
4645 char __user *optval, int __user *optlen)
4647 struct sctp_bind_addr *bp;
4648 struct sctp_association *asoc;
4650 struct sctp_getaddrs_old getaddrs;
4651 struct sctp_sockaddr_entry *addr;
4653 union sctp_addr temp;
4654 struct sctp_sock *sp = sctp_sk(sk);
4659 int bytes_copied = 0;
4661 if (len < sizeof(struct sctp_getaddrs_old))
4664 len = sizeof(struct sctp_getaddrs_old);
4665 if (copy_from_user(&getaddrs, optval, len))
4668 if (getaddrs.addr_num <= 0 ||
4669 getaddrs.addr_num >= (INT_MAX / sizeof(union sctp_addr)))
4672 printk(KERN_WARNING "SCTP: Use of SCTP_GET_LOCAL_ADDRS_OLD "
4673 "socket option deprecated\n");
4676 * For UDP-style sockets, id specifies the association to query.
4677 * If the id field is set to the value '0' then the locally bound
4678 * addresses are returned without regard to any particular
4681 if (0 == getaddrs.assoc_id) {
4682 bp = &sctp_sk(sk)->ep->base.bind_addr;
4684 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4687 bp = &asoc->base.bind_addr;
4690 to = getaddrs.addrs;
4692 /* Allocate space for a local instance of packed array to hold all
4693 * the data. We store addresses here first and then put write them
4694 * to the user in one shot.
4696 addrs = kmalloc(sizeof(union sctp_addr) * getaddrs.addr_num,
4701 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4702 * addresses from the global local address list.
4704 if (sctp_list_single_entry(&bp->address_list)) {
4705 addr = list_entry(bp->address_list.next,
4706 struct sctp_sockaddr_entry, list);
4707 if (sctp_is_any(sk, &addr->a)) {
4708 cnt = sctp_copy_laddrs_old(sk, bp->port,
4710 addrs, &bytes_copied);
4716 /* Protection on the bound address list is not needed since
4717 * in the socket option context we hold a socket lock and
4718 * thus the bound address list can't change.
4720 list_for_each_entry(addr, &bp->address_list, list) {
4721 memcpy(&temp, &addr->a, sizeof(temp));
4722 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4723 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4724 memcpy(buf, &temp, addrlen);
4726 bytes_copied += addrlen;
4728 if (cnt >= getaddrs.addr_num) break;
4732 /* copy the entire address list into the user provided space */
4733 if (copy_to_user(to, addrs, bytes_copied)) {
4738 /* copy the leading structure back to user */
4739 getaddrs.addr_num = cnt;
4740 if (copy_to_user(optval, &getaddrs, len))
4748 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4749 char __user *optval, int __user *optlen)
4751 struct sctp_bind_addr *bp;
4752 struct sctp_association *asoc;
4754 struct sctp_getaddrs getaddrs;
4755 struct sctp_sockaddr_entry *addr;
4757 union sctp_addr temp;
4758 struct sctp_sock *sp = sctp_sk(sk);
4762 int bytes_copied = 0;
4766 if (len < sizeof(struct sctp_getaddrs))
4769 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4773 * For UDP-style sockets, id specifies the association to query.
4774 * If the id field is set to the value '0' then the locally bound
4775 * addresses are returned without regard to any particular
4778 if (0 == getaddrs.assoc_id) {
4779 bp = &sctp_sk(sk)->ep->base.bind_addr;
4781 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4784 bp = &asoc->base.bind_addr;
4787 to = optval + offsetof(struct sctp_getaddrs,addrs);
4788 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4790 addrs = kmalloc(space_left, GFP_KERNEL);
4794 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4795 * addresses from the global local address list.
4797 if (sctp_list_single_entry(&bp->address_list)) {
4798 addr = list_entry(bp->address_list.next,
4799 struct sctp_sockaddr_entry, list);
4800 if (sctp_is_any(sk, &addr->a)) {
4801 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4802 space_left, &bytes_copied);
4812 /* Protection on the bound address list is not needed since
4813 * in the socket option context we hold a socket lock and
4814 * thus the bound address list can't change.
4816 list_for_each_entry(addr, &bp->address_list, list) {
4817 memcpy(&temp, &addr->a, sizeof(temp));
4818 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4819 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4820 if (space_left < addrlen) {
4821 err = -ENOMEM; /*fixme: right error?*/
4824 memcpy(buf, &temp, addrlen);
4826 bytes_copied += addrlen;
4828 space_left -= addrlen;
4832 if (copy_to_user(to, addrs, bytes_copied)) {
4836 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4840 if (put_user(bytes_copied, optlen))
4847 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4849 * Requests that the local SCTP stack use the enclosed peer address as
4850 * the association primary. The enclosed address must be one of the
4851 * association peer's addresses.
4853 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4854 char __user *optval, int __user *optlen)
4856 struct sctp_prim prim;
4857 struct sctp_association *asoc;
4858 struct sctp_sock *sp = sctp_sk(sk);
4860 if (len < sizeof(struct sctp_prim))
4863 len = sizeof(struct sctp_prim);
4865 if (copy_from_user(&prim, optval, len))
4868 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4872 if (!asoc->peer.primary_path)
4875 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4876 asoc->peer.primary_path->af_specific->sockaddr_len);
4878 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4879 (union sctp_addr *)&prim.ssp_addr);
4881 if (put_user(len, optlen))
4883 if (copy_to_user(optval, &prim, len))
4890 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4892 * Requests that the local endpoint set the specified Adaptation Layer
4893 * Indication parameter for all future INIT and INIT-ACK exchanges.
4895 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4896 char __user *optval, int __user *optlen)
4898 struct sctp_setadaptation adaptation;
4900 if (len < sizeof(struct sctp_setadaptation))
4903 len = sizeof(struct sctp_setadaptation);
4905 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4907 if (put_user(len, optlen))
4909 if (copy_to_user(optval, &adaptation, len))
4917 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4919 * Applications that wish to use the sendto() system call may wish to
4920 * specify a default set of parameters that would normally be supplied
4921 * through the inclusion of ancillary data. This socket option allows
4922 * such an application to set the default sctp_sndrcvinfo structure.
4925 * The application that wishes to use this socket option simply passes
4926 * in to this call the sctp_sndrcvinfo structure defined in Section
4927 * 5.2.2) The input parameters accepted by this call include
4928 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4929 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4930 * to this call if the caller is using the UDP model.
4932 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4934 static int sctp_getsockopt_default_send_param(struct sock *sk,
4935 int len, char __user *optval,
4938 struct sctp_sndrcvinfo info;
4939 struct sctp_association *asoc;
4940 struct sctp_sock *sp = sctp_sk(sk);
4942 if (len < sizeof(struct sctp_sndrcvinfo))
4945 len = sizeof(struct sctp_sndrcvinfo);
4947 if (copy_from_user(&info, optval, len))
4950 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4951 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4955 info.sinfo_stream = asoc->default_stream;
4956 info.sinfo_flags = asoc->default_flags;
4957 info.sinfo_ppid = asoc->default_ppid;
4958 info.sinfo_context = asoc->default_context;
4959 info.sinfo_timetolive = asoc->default_timetolive;
4961 info.sinfo_stream = sp->default_stream;
4962 info.sinfo_flags = sp->default_flags;
4963 info.sinfo_ppid = sp->default_ppid;
4964 info.sinfo_context = sp->default_context;
4965 info.sinfo_timetolive = sp->default_timetolive;
4968 if (put_user(len, optlen))
4970 if (copy_to_user(optval, &info, len))
4978 * 7.1.5 SCTP_NODELAY
4980 * Turn on/off any Nagle-like algorithm. This means that packets are
4981 * generally sent as soon as possible and no unnecessary delays are
4982 * introduced, at the cost of more packets in the network. Expects an
4983 * integer boolean flag.
4986 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4987 char __user *optval, int __user *optlen)
4991 if (len < sizeof(int))
4995 val = (sctp_sk(sk)->nodelay == 1);
4996 if (put_user(len, optlen))
4998 if (copy_to_user(optval, &val, len))
5005 * 7.1.1 SCTP_RTOINFO
5007 * The protocol parameters used to initialize and bound retransmission
5008 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5009 * and modify these parameters.
5010 * All parameters are time values, in milliseconds. A value of 0, when
5011 * modifying the parameters, indicates that the current value should not
5015 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5016 char __user *optval,
5017 int __user *optlen) {
5018 struct sctp_rtoinfo rtoinfo;
5019 struct sctp_association *asoc;
5021 if (len < sizeof (struct sctp_rtoinfo))
5024 len = sizeof(struct sctp_rtoinfo);
5026 if (copy_from_user(&rtoinfo, optval, len))
5029 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5031 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5034 /* Values corresponding to the specific association. */
5036 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5037 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5038 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5040 /* Values corresponding to the endpoint. */
5041 struct sctp_sock *sp = sctp_sk(sk);
5043 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5044 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5045 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5048 if (put_user(len, optlen))
5051 if (copy_to_user(optval, &rtoinfo, len))
5059 * 7.1.2 SCTP_ASSOCINFO
5061 * This option is used to tune the maximum retransmission attempts
5062 * of the association.
5063 * Returns an error if the new association retransmission value is
5064 * greater than the sum of the retransmission value of the peer.
5065 * See [SCTP] for more information.
5068 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5069 char __user *optval,
5073 struct sctp_assocparams assocparams;
5074 struct sctp_association *asoc;
5075 struct list_head *pos;
5078 if (len < sizeof (struct sctp_assocparams))
5081 len = sizeof(struct sctp_assocparams);
5083 if (copy_from_user(&assocparams, optval, len))
5086 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5088 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5091 /* Values correspoinding to the specific association */
5093 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5094 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5095 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5096 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
5098 (asoc->cookie_life.tv_usec
5101 list_for_each(pos, &asoc->peer.transport_addr_list) {
5105 assocparams.sasoc_number_peer_destinations = cnt;
5107 /* Values corresponding to the endpoint */
5108 struct sctp_sock *sp = sctp_sk(sk);
5110 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5111 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5112 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5113 assocparams.sasoc_cookie_life =
5114 sp->assocparams.sasoc_cookie_life;
5115 assocparams.sasoc_number_peer_destinations =
5117 sasoc_number_peer_destinations;
5120 if (put_user(len, optlen))
5123 if (copy_to_user(optval, &assocparams, len))
5130 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5132 * This socket option is a boolean flag which turns on or off mapped V4
5133 * addresses. If this option is turned on and the socket is type
5134 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5135 * If this option is turned off, then no mapping will be done of V4
5136 * addresses and a user will receive both PF_INET6 and PF_INET type
5137 * addresses on the socket.
5139 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5140 char __user *optval, int __user *optlen)
5143 struct sctp_sock *sp = sctp_sk(sk);
5145 if (len < sizeof(int))
5150 if (put_user(len, optlen))
5152 if (copy_to_user(optval, &val, len))
5159 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5160 * (chapter and verse is quoted at sctp_setsockopt_context())
5162 static int sctp_getsockopt_context(struct sock *sk, int len,
5163 char __user *optval, int __user *optlen)
5165 struct sctp_assoc_value params;
5166 struct sctp_sock *sp;
5167 struct sctp_association *asoc;
5169 if (len < sizeof(struct sctp_assoc_value))
5172 len = sizeof(struct sctp_assoc_value);
5174 if (copy_from_user(¶ms, optval, len))
5179 if (params.assoc_id != 0) {
5180 asoc = sctp_id2assoc(sk, params.assoc_id);
5183 params.assoc_value = asoc->default_rcv_context;
5185 params.assoc_value = sp->default_rcv_context;
5188 if (put_user(len, optlen))
5190 if (copy_to_user(optval, ¶ms, len))
5197 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5198 * This option will get or set the maximum size to put in any outgoing
5199 * SCTP DATA chunk. If a message is larger than this size it will be
5200 * fragmented by SCTP into the specified size. Note that the underlying
5201 * SCTP implementation may fragment into smaller sized chunks when the
5202 * PMTU of the underlying association is smaller than the value set by
5203 * the user. The default value for this option is '0' which indicates
5204 * the user is NOT limiting fragmentation and only the PMTU will effect
5205 * SCTP's choice of DATA chunk size. Note also that values set larger
5206 * than the maximum size of an IP datagram will effectively let SCTP
5207 * control fragmentation (i.e. the same as setting this option to 0).
5209 * The following structure is used to access and modify this parameter:
5211 * struct sctp_assoc_value {
5212 * sctp_assoc_t assoc_id;
5213 * uint32_t assoc_value;
5216 * assoc_id: This parameter is ignored for one-to-one style sockets.
5217 * For one-to-many style sockets this parameter indicates which
5218 * association the user is performing an action upon. Note that if
5219 * this field's value is zero then the endpoints default value is
5220 * changed (effecting future associations only).
5221 * assoc_value: This parameter specifies the maximum size in bytes.
5223 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5224 char __user *optval, int __user *optlen)
5226 struct sctp_assoc_value params;
5227 struct sctp_association *asoc;
5229 if (len == sizeof(int)) {
5231 "SCTP: Use of int in maxseg socket option deprecated\n");
5233 "SCTP: Use struct sctp_assoc_value instead\n");
5234 params.assoc_id = 0;
5235 } else if (len >= sizeof(struct sctp_assoc_value)) {
5236 len = sizeof(struct sctp_assoc_value);
5237 if (copy_from_user(¶ms, optval, sizeof(params)))
5242 asoc = sctp_id2assoc(sk, params.assoc_id);
5243 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5247 params.assoc_value = asoc->frag_point;
5249 params.assoc_value = sctp_sk(sk)->user_frag;
5251 if (put_user(len, optlen))
5253 if (len == sizeof(int)) {
5254 if (copy_to_user(optval, ¶ms.assoc_value, len))
5257 if (copy_to_user(optval, ¶ms, len))
5265 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5266 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5268 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5269 char __user *optval, int __user *optlen)
5273 if (len < sizeof(int))
5278 val = sctp_sk(sk)->frag_interleave;
5279 if (put_user(len, optlen))
5281 if (copy_to_user(optval, &val, len))
5288 * 7.1.25. Set or Get the sctp partial delivery point
5289 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5291 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5292 char __user *optval,
5297 if (len < sizeof(u32))
5302 val = sctp_sk(sk)->pd_point;
5303 if (put_user(len, optlen))
5305 if (copy_to_user(optval, &val, len))
5312 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5313 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5315 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5316 char __user *optval,
5319 struct sctp_assoc_value params;
5320 struct sctp_sock *sp;
5321 struct sctp_association *asoc;
5323 if (len == sizeof(int)) {
5325 "SCTP: Use of int in max_burst socket option deprecated\n");
5327 "SCTP: Use struct sctp_assoc_value instead\n");
5328 params.assoc_id = 0;
5329 } else if (len >= sizeof(struct sctp_assoc_value)) {
5330 len = sizeof(struct sctp_assoc_value);
5331 if (copy_from_user(¶ms, optval, len))
5338 if (params.assoc_id != 0) {
5339 asoc = sctp_id2assoc(sk, params.assoc_id);
5342 params.assoc_value = asoc->max_burst;
5344 params.assoc_value = sp->max_burst;
5346 if (len == sizeof(int)) {
5347 if (copy_to_user(optval, ¶ms.assoc_value, len))
5350 if (copy_to_user(optval, ¶ms, len))
5358 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5359 char __user *optval, int __user *optlen)
5361 struct sctp_hmacalgo __user *p = (void __user *)optval;
5362 struct sctp_hmac_algo_param *hmacs;
5366 if (!sctp_auth_enable)
5369 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5370 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5372 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5375 len = sizeof(struct sctp_hmacalgo) + data_len;
5376 num_idents = data_len / sizeof(u16);
5378 if (put_user(len, optlen))
5380 if (put_user(num_idents, &p->shmac_num_idents))
5382 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5387 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5388 char __user *optval, int __user *optlen)
5390 struct sctp_authkeyid val;
5391 struct sctp_association *asoc;
5393 if (!sctp_auth_enable)
5396 if (len < sizeof(struct sctp_authkeyid))
5398 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5401 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5402 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5406 val.scact_keynumber = asoc->active_key_id;
5408 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5410 len = sizeof(struct sctp_authkeyid);
5411 if (put_user(len, optlen))
5413 if (copy_to_user(optval, &val, len))
5419 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5420 char __user *optval, int __user *optlen)
5422 struct sctp_authchunks __user *p = (void __user *)optval;
5423 struct sctp_authchunks val;
5424 struct sctp_association *asoc;
5425 struct sctp_chunks_param *ch;
5429 if (!sctp_auth_enable)
5432 if (len < sizeof(struct sctp_authchunks))
5435 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5438 to = p->gauth_chunks;
5439 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5443 ch = asoc->peer.peer_chunks;
5447 /* See if the user provided enough room for all the data */
5448 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5449 if (len < num_chunks)
5452 if (copy_to_user(to, ch->chunks, num_chunks))
5455 len = sizeof(struct sctp_authchunks) + num_chunks;
5456 if (put_user(len, optlen)) return -EFAULT;
5457 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5462 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5463 char __user *optval, int __user *optlen)
5465 struct sctp_authchunks __user *p = (void __user *)optval;
5466 struct sctp_authchunks val;
5467 struct sctp_association *asoc;
5468 struct sctp_chunks_param *ch;
5472 if (!sctp_auth_enable)
5475 if (len < sizeof(struct sctp_authchunks))
5478 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5481 to = p->gauth_chunks;
5482 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5483 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5487 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5489 ch = sctp_sk(sk)->ep->auth_chunk_list;
5494 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5495 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5498 if (copy_to_user(to, ch->chunks, num_chunks))
5501 len = sizeof(struct sctp_authchunks) + num_chunks;
5502 if (put_user(len, optlen))
5504 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5511 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5512 * This option gets the current number of associations that are attached
5513 * to a one-to-many style socket. The option value is an uint32_t.
5515 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5516 char __user *optval, int __user *optlen)
5518 struct sctp_sock *sp = sctp_sk(sk);
5519 struct sctp_association *asoc;
5522 if (sctp_style(sk, TCP))
5525 if (len < sizeof(u32))
5530 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5534 if (put_user(len, optlen))
5536 if (copy_to_user(optval, &val, len))
5542 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5543 char __user *optval, int __user *optlen)
5548 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5551 /* I can hardly begin to describe how wrong this is. This is
5552 * so broken as to be worse than useless. The API draft
5553 * REALLY is NOT helpful here... I am not convinced that the
5554 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5555 * are at all well-founded.
5557 if (level != SOL_SCTP) {
5558 struct sctp_af *af = sctp_sk(sk)->pf->af;
5560 retval = af->getsockopt(sk, level, optname, optval, optlen);
5564 if (get_user(len, optlen))
5571 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5573 case SCTP_DISABLE_FRAGMENTS:
5574 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5578 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5580 case SCTP_AUTOCLOSE:
5581 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5583 case SCTP_SOCKOPT_PEELOFF:
5584 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5586 case SCTP_PEER_ADDR_PARAMS:
5587 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5590 case SCTP_DELAYED_ACK:
5591 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5595 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5597 case SCTP_GET_PEER_ADDRS_NUM_OLD:
5598 retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
5601 case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
5602 retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
5605 case SCTP_GET_PEER_ADDRS_OLD:
5606 retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
5609 case SCTP_GET_LOCAL_ADDRS_OLD:
5610 retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
5613 case SCTP_GET_PEER_ADDRS:
5614 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5617 case SCTP_GET_LOCAL_ADDRS:
5618 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5621 case SCTP_SOCKOPT_CONNECTX3:
5622 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5624 case SCTP_DEFAULT_SEND_PARAM:
5625 retval = sctp_getsockopt_default_send_param(sk, len,
5628 case SCTP_PRIMARY_ADDR:
5629 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5632 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5635 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5637 case SCTP_ASSOCINFO:
5638 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5640 case SCTP_I_WANT_MAPPED_V4_ADDR:
5641 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5644 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5646 case SCTP_GET_PEER_ADDR_INFO:
5647 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5650 case SCTP_ADAPTATION_LAYER:
5651 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5655 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5657 case SCTP_FRAGMENT_INTERLEAVE:
5658 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5661 case SCTP_PARTIAL_DELIVERY_POINT:
5662 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5665 case SCTP_MAX_BURST:
5666 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5669 case SCTP_AUTH_CHUNK:
5670 case SCTP_AUTH_DELETE_KEY:
5671 retval = -EOPNOTSUPP;
5673 case SCTP_HMAC_IDENT:
5674 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5676 case SCTP_AUTH_ACTIVE_KEY:
5677 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5679 case SCTP_PEER_AUTH_CHUNKS:
5680 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5683 case SCTP_LOCAL_AUTH_CHUNKS:
5684 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5687 case SCTP_GET_ASSOC_NUMBER:
5688 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5691 retval = -ENOPROTOOPT;
5695 sctp_release_sock(sk);
5699 static void sctp_hash(struct sock *sk)
5704 static void sctp_unhash(struct sock *sk)
5709 /* Check if port is acceptable. Possibly find first available port.
5711 * The port hash table (contained in the 'global' SCTP protocol storage
5712 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5713 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5714 * list (the list number is the port number hashed out, so as you
5715 * would expect from a hash function, all the ports in a given list have
5716 * such a number that hashes out to the same list number; you were
5717 * expecting that, right?); so each list has a set of ports, with a
5718 * link to the socket (struct sock) that uses it, the port number and
5719 * a fastreuse flag (FIXME: NPI ipg).
5721 static struct sctp_bind_bucket *sctp_bucket_create(
5722 struct sctp_bind_hashbucket *head, unsigned short snum);
5724 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5726 struct sctp_bind_hashbucket *head; /* hash list */
5727 struct sctp_bind_bucket *pp; /* hash list port iterator */
5728 struct hlist_node *node;
5729 unsigned short snum;
5732 snum = ntohs(addr->v4.sin_port);
5734 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5735 sctp_local_bh_disable();
5738 /* Search for an available port. */
5739 int low, high, remaining, index;
5742 inet_get_local_port_range(&low, &high);
5743 remaining = (high - low) + 1;
5744 rover = net_random() % remaining + low;
5748 if ((rover < low) || (rover > high))
5750 index = sctp_phashfn(rover);
5751 head = &sctp_port_hashtable[index];
5752 sctp_spin_lock(&head->lock);
5753 sctp_for_each_hentry(pp, node, &head->chain)
5754 if (pp->port == rover)
5758 sctp_spin_unlock(&head->lock);
5759 } while (--remaining > 0);
5761 /* Exhausted local port range during search? */
5766 /* OK, here is the one we will use. HEAD (the port
5767 * hash table list entry) is non-NULL and we hold it's
5772 /* We are given an specific port number; we verify
5773 * that it is not being used. If it is used, we will
5774 * exahust the search in the hash list corresponding
5775 * to the port number (snum) - we detect that with the
5776 * port iterator, pp being NULL.
5778 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5779 sctp_spin_lock(&head->lock);
5780 sctp_for_each_hentry(pp, node, &head->chain) {
5781 if (pp->port == snum)
5788 if (!hlist_empty(&pp->owner)) {
5789 /* We had a port hash table hit - there is an
5790 * available port (pp != NULL) and it is being
5791 * used by other socket (pp->owner not empty); that other
5792 * socket is going to be sk2.
5794 int reuse = sk->sk_reuse;
5796 struct hlist_node *node;
5798 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5799 if (pp->fastreuse && sk->sk_reuse &&
5800 sk->sk_state != SCTP_SS_LISTENING)
5803 /* Run through the list of sockets bound to the port
5804 * (pp->port) [via the pointers bind_next and
5805 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5806 * we get the endpoint they describe and run through
5807 * the endpoint's list of IP (v4 or v6) addresses,
5808 * comparing each of the addresses with the address of
5809 * the socket sk. If we find a match, then that means
5810 * that this port/socket (sk) combination are already
5813 sk_for_each_bound(sk2, node, &pp->owner) {
5814 struct sctp_endpoint *ep2;
5815 ep2 = sctp_sk(sk2)->ep;
5818 (reuse && sk2->sk_reuse &&
5819 sk2->sk_state != SCTP_SS_LISTENING))
5822 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5823 sctp_sk(sk2), sctp_sk(sk))) {
5828 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5831 /* If there was a hash table miss, create a new port. */
5833 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5836 /* In either case (hit or miss), make sure fastreuse is 1 only
5837 * if sk->sk_reuse is too (that is, if the caller requested
5838 * SO_REUSEADDR on this socket -sk-).
5840 if (hlist_empty(&pp->owner)) {
5841 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5845 } else if (pp->fastreuse &&
5846 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5849 /* We are set, so fill up all the data in the hash table
5850 * entry, tie the socket list information with the rest of the
5851 * sockets FIXME: Blurry, NPI (ipg).
5854 if (!sctp_sk(sk)->bind_hash) {
5855 inet_sk(sk)->num = snum;
5856 sk_add_bind_node(sk, &pp->owner);
5857 sctp_sk(sk)->bind_hash = pp;
5862 sctp_spin_unlock(&head->lock);
5865 sctp_local_bh_enable();
5869 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5870 * port is requested.
5872 static int sctp_get_port(struct sock *sk, unsigned short snum)
5875 union sctp_addr addr;
5876 struct sctp_af *af = sctp_sk(sk)->pf->af;
5878 /* Set up a dummy address struct from the sk. */
5879 af->from_sk(&addr, sk);
5880 addr.v4.sin_port = htons(snum);
5882 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5883 ret = sctp_get_port_local(sk, &addr);
5885 return (ret ? 1 : 0);
5889 * Move a socket to LISTENING state.
5891 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5893 struct sctp_sock *sp = sctp_sk(sk);
5894 struct sctp_endpoint *ep = sp->ep;
5895 struct crypto_hash *tfm = NULL;
5897 /* Allocate HMAC for generating cookie. */
5898 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5899 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5901 if (net_ratelimit()) {
5903 "SCTP: failed to load transform for %s: %ld\n",
5904 sctp_hmac_alg, PTR_ERR(tfm));
5908 sctp_sk(sk)->hmac = tfm;
5912 * If a bind() or sctp_bindx() is not called prior to a listen()
5913 * call that allows new associations to be accepted, the system
5914 * picks an ephemeral port and will choose an address set equivalent
5915 * to binding with a wildcard address.
5917 * This is not currently spelled out in the SCTP sockets
5918 * extensions draft, but follows the practice as seen in TCP
5922 sk->sk_state = SCTP_SS_LISTENING;
5923 if (!ep->base.bind_addr.port) {
5924 if (sctp_autobind(sk))
5927 if (sctp_get_port(sk, inet_sk(sk)->num)) {
5928 sk->sk_state = SCTP_SS_CLOSED;
5933 sk->sk_max_ack_backlog = backlog;
5934 sctp_hash_endpoint(ep);
5939 * 4.1.3 / 5.1.3 listen()
5941 * By default, new associations are not accepted for UDP style sockets.
5942 * An application uses listen() to mark a socket as being able to
5943 * accept new associations.
5945 * On TCP style sockets, applications use listen() to ready the SCTP
5946 * endpoint for accepting inbound associations.
5948 * On both types of endpoints a backlog of '0' disables listening.
5950 * Move a socket to LISTENING state.
5952 int sctp_inet_listen(struct socket *sock, int backlog)
5954 struct sock *sk = sock->sk;
5955 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5958 if (unlikely(backlog < 0))
5963 /* Peeled-off sockets are not allowed to listen(). */
5964 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
5967 if (sock->state != SS_UNCONNECTED)
5970 /* If backlog is zero, disable listening. */
5972 if (sctp_sstate(sk, CLOSED))
5976 sctp_unhash_endpoint(ep);
5977 sk->sk_state = SCTP_SS_CLOSED;
5979 sctp_sk(sk)->bind_hash->fastreuse = 1;
5983 /* If we are already listening, just update the backlog */
5984 if (sctp_sstate(sk, LISTENING))
5985 sk->sk_max_ack_backlog = backlog;
5987 err = sctp_listen_start(sk, backlog);
5994 sctp_release_sock(sk);
5999 * This function is done by modeling the current datagram_poll() and the
6000 * tcp_poll(). Note that, based on these implementations, we don't
6001 * lock the socket in this function, even though it seems that,
6002 * ideally, locking or some other mechanisms can be used to ensure
6003 * the integrity of the counters (sndbuf and wmem_alloc) used
6004 * in this place. We assume that we don't need locks either until proven
6007 * Another thing to note is that we include the Async I/O support
6008 * here, again, by modeling the current TCP/UDP code. We don't have
6009 * a good way to test with it yet.
6011 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6013 struct sock *sk = sock->sk;
6014 struct sctp_sock *sp = sctp_sk(sk);
6017 poll_wait(file, sk->sk_sleep, wait);
6019 /* A TCP-style listening socket becomes readable when the accept queue
6022 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6023 return (!list_empty(&sp->ep->asocs)) ?
6024 (POLLIN | POLLRDNORM) : 0;
6028 /* Is there any exceptional events? */
6029 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6031 if (sk->sk_shutdown & RCV_SHUTDOWN)
6033 if (sk->sk_shutdown == SHUTDOWN_MASK)
6036 /* Is it readable? Reconsider this code with TCP-style support. */
6037 if (!skb_queue_empty(&sk->sk_receive_queue) ||
6038 (sk->sk_shutdown & RCV_SHUTDOWN))
6039 mask |= POLLIN | POLLRDNORM;
6041 /* The association is either gone or not ready. */
6042 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6045 /* Is it writable? */
6046 if (sctp_writeable(sk)) {
6047 mask |= POLLOUT | POLLWRNORM;
6049 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6051 * Since the socket is not locked, the buffer
6052 * might be made available after the writeable check and
6053 * before the bit is set. This could cause a lost I/O
6054 * signal. tcp_poll() has a race breaker for this race
6055 * condition. Based on their implementation, we put
6056 * in the following code to cover it as well.
6058 if (sctp_writeable(sk))
6059 mask |= POLLOUT | POLLWRNORM;
6064 /********************************************************************
6065 * 2nd Level Abstractions
6066 ********************************************************************/
6068 static struct sctp_bind_bucket *sctp_bucket_create(
6069 struct sctp_bind_hashbucket *head, unsigned short snum)
6071 struct sctp_bind_bucket *pp;
6073 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6075 SCTP_DBG_OBJCNT_INC(bind_bucket);
6078 INIT_HLIST_HEAD(&pp->owner);
6079 hlist_add_head(&pp->node, &head->chain);
6084 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6085 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6087 if (pp && hlist_empty(&pp->owner)) {
6088 __hlist_del(&pp->node);
6089 kmem_cache_free(sctp_bucket_cachep, pp);
6090 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6094 /* Release this socket's reference to a local port. */
6095 static inline void __sctp_put_port(struct sock *sk)
6097 struct sctp_bind_hashbucket *head =
6098 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
6099 struct sctp_bind_bucket *pp;
6101 sctp_spin_lock(&head->lock);
6102 pp = sctp_sk(sk)->bind_hash;
6103 __sk_del_bind_node(sk);
6104 sctp_sk(sk)->bind_hash = NULL;
6105 inet_sk(sk)->num = 0;
6106 sctp_bucket_destroy(pp);
6107 sctp_spin_unlock(&head->lock);
6110 void sctp_put_port(struct sock *sk)
6112 sctp_local_bh_disable();
6113 __sctp_put_port(sk);
6114 sctp_local_bh_enable();
6118 * The system picks an ephemeral port and choose an address set equivalent
6119 * to binding with a wildcard address.
6120 * One of those addresses will be the primary address for the association.
6121 * This automatically enables the multihoming capability of SCTP.
6123 static int sctp_autobind(struct sock *sk)
6125 union sctp_addr autoaddr;
6129 /* Initialize a local sockaddr structure to INADDR_ANY. */
6130 af = sctp_sk(sk)->pf->af;
6132 port = htons(inet_sk(sk)->num);
6133 af->inaddr_any(&autoaddr, port);
6135 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6138 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6141 * 4.2 The cmsghdr Structure *
6143 * When ancillary data is sent or received, any number of ancillary data
6144 * objects can be specified by the msg_control and msg_controllen members of
6145 * the msghdr structure, because each object is preceded by
6146 * a cmsghdr structure defining the object's length (the cmsg_len member).
6147 * Historically Berkeley-derived implementations have passed only one object
6148 * at a time, but this API allows multiple objects to be
6149 * passed in a single call to sendmsg() or recvmsg(). The following example
6150 * shows two ancillary data objects in a control buffer.
6152 * |<--------------------------- msg_controllen -------------------------->|
6155 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6157 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6160 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6162 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6165 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6166 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6168 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6170 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6177 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
6178 sctp_cmsgs_t *cmsgs)
6180 struct cmsghdr *cmsg;
6181 struct msghdr *my_msg = (struct msghdr *)msg;
6183 for (cmsg = CMSG_FIRSTHDR(msg);
6185 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6186 if (!CMSG_OK(my_msg, cmsg))
6189 /* Should we parse this header or ignore? */
6190 if (cmsg->cmsg_level != IPPROTO_SCTP)
6193 /* Strictly check lengths following example in SCM code. */
6194 switch (cmsg->cmsg_type) {
6196 /* SCTP Socket API Extension
6197 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6199 * This cmsghdr structure provides information for
6200 * initializing new SCTP associations with sendmsg().
6201 * The SCTP_INITMSG socket option uses this same data
6202 * structure. This structure is not used for
6205 * cmsg_level cmsg_type cmsg_data[]
6206 * ------------ ------------ ----------------------
6207 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6209 if (cmsg->cmsg_len !=
6210 CMSG_LEN(sizeof(struct sctp_initmsg)))
6212 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6216 /* SCTP Socket API Extension
6217 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6219 * This cmsghdr structure specifies SCTP options for
6220 * sendmsg() and describes SCTP header information
6221 * about a received message through recvmsg().
6223 * cmsg_level cmsg_type cmsg_data[]
6224 * ------------ ------------ ----------------------
6225 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6227 if (cmsg->cmsg_len !=
6228 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6232 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6234 /* Minimally, validate the sinfo_flags. */
6235 if (cmsgs->info->sinfo_flags &
6236 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6237 SCTP_ABORT | SCTP_EOF))
6249 * Wait for a packet..
6250 * Note: This function is the same function as in core/datagram.c
6251 * with a few modifications to make lksctp work.
6253 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6258 prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
6260 /* Socket errors? */
6261 error = sock_error(sk);
6265 if (!skb_queue_empty(&sk->sk_receive_queue))
6268 /* Socket shut down? */
6269 if (sk->sk_shutdown & RCV_SHUTDOWN)
6272 /* Sequenced packets can come disconnected. If so we report the
6277 /* Is there a good reason to think that we may receive some data? */
6278 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6281 /* Handle signals. */
6282 if (signal_pending(current))
6285 /* Let another process have a go. Since we are going to sleep
6286 * anyway. Note: This may cause odd behaviors if the message
6287 * does not fit in the user's buffer, but this seems to be the
6288 * only way to honor MSG_DONTWAIT realistically.
6290 sctp_release_sock(sk);
6291 *timeo_p = schedule_timeout(*timeo_p);
6295 finish_wait(sk->sk_sleep, &wait);
6299 error = sock_intr_errno(*timeo_p);
6302 finish_wait(sk->sk_sleep, &wait);
6307 /* Receive a datagram.
6308 * Note: This is pretty much the same routine as in core/datagram.c
6309 * with a few changes to make lksctp work.
6311 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6312 int noblock, int *err)
6315 struct sk_buff *skb;
6318 timeo = sock_rcvtimeo(sk, noblock);
6320 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6321 timeo, MAX_SCHEDULE_TIMEOUT);
6324 /* Again only user level code calls this function,
6325 * so nothing interrupt level
6326 * will suddenly eat the receive_queue.
6328 * Look at current nfs client by the way...
6329 * However, this function was corrent in any case. 8)
6331 if (flags & MSG_PEEK) {
6332 spin_lock_bh(&sk->sk_receive_queue.lock);
6333 skb = skb_peek(&sk->sk_receive_queue);
6335 atomic_inc(&skb->users);
6336 spin_unlock_bh(&sk->sk_receive_queue.lock);
6338 skb = skb_dequeue(&sk->sk_receive_queue);
6344 /* Caller is allowed not to check sk->sk_err before calling. */
6345 error = sock_error(sk);
6349 if (sk->sk_shutdown & RCV_SHUTDOWN)
6352 /* User doesn't want to wait. */
6356 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6365 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6366 static void __sctp_write_space(struct sctp_association *asoc)
6368 struct sock *sk = asoc->base.sk;
6369 struct socket *sock = sk->sk_socket;
6371 if ((sctp_wspace(asoc) > 0) && sock) {
6372 if (waitqueue_active(&asoc->wait))
6373 wake_up_interruptible(&asoc->wait);
6375 if (sctp_writeable(sk)) {
6376 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
6377 wake_up_interruptible(sk->sk_sleep);
6379 /* Note that we try to include the Async I/O support
6380 * here by modeling from the current TCP/UDP code.
6381 * We have not tested with it yet.
6383 if (sock->fasync_list &&
6384 !(sk->sk_shutdown & SEND_SHUTDOWN))
6385 sock_wake_async(sock,
6386 SOCK_WAKE_SPACE, POLL_OUT);
6391 /* Do accounting for the sndbuf space.
6392 * Decrement the used sndbuf space of the corresponding association by the
6393 * data size which was just transmitted(freed).
6395 static void sctp_wfree(struct sk_buff *skb)
6397 struct sctp_association *asoc;
6398 struct sctp_chunk *chunk;
6401 /* Get the saved chunk pointer. */
6402 chunk = *((struct sctp_chunk **)(skb->cb));
6405 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6406 sizeof(struct sk_buff) +
6407 sizeof(struct sctp_chunk);
6409 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6412 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6414 sk->sk_wmem_queued -= skb->truesize;
6415 sk_mem_uncharge(sk, skb->truesize);
6418 __sctp_write_space(asoc);
6420 sctp_association_put(asoc);
6423 /* Do accounting for the receive space on the socket.
6424 * Accounting for the association is done in ulpevent.c
6425 * We set this as a destructor for the cloned data skbs so that
6426 * accounting is done at the correct time.
6428 void sctp_sock_rfree(struct sk_buff *skb)
6430 struct sock *sk = skb->sk;
6431 struct sctp_ulpevent *event = sctp_skb2event(skb);
6433 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6436 * Mimic the behavior of sock_rfree
6438 sk_mem_uncharge(sk, event->rmem_len);
6442 /* Helper function to wait for space in the sndbuf. */
6443 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6446 struct sock *sk = asoc->base.sk;
6448 long current_timeo = *timeo_p;
6451 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6452 asoc, (long)(*timeo_p), msg_len);
6454 /* Increment the association's refcnt. */
6455 sctp_association_hold(asoc);
6457 /* Wait on the association specific sndbuf space. */
6459 prepare_to_wait_exclusive(&asoc->wait, &wait,
6460 TASK_INTERRUPTIBLE);
6463 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6466 if (signal_pending(current))
6467 goto do_interrupted;
6468 if (msg_len <= sctp_wspace(asoc))
6471 /* Let another process have a go. Since we are going
6474 sctp_release_sock(sk);
6475 current_timeo = schedule_timeout(current_timeo);
6476 BUG_ON(sk != asoc->base.sk);
6479 *timeo_p = current_timeo;
6483 finish_wait(&asoc->wait, &wait);
6485 /* Release the association's refcnt. */
6486 sctp_association_put(asoc);
6495 err = sock_intr_errno(*timeo_p);
6503 /* If socket sndbuf has changed, wake up all per association waiters. */
6504 void sctp_write_space(struct sock *sk)
6506 struct sctp_association *asoc;
6508 /* Wake up the tasks in each wait queue. */
6509 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6510 __sctp_write_space(asoc);
6514 /* Is there any sndbuf space available on the socket?
6516 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6517 * associations on the same socket. For a UDP-style socket with
6518 * multiple associations, it is possible for it to be "unwriteable"
6519 * prematurely. I assume that this is acceptable because
6520 * a premature "unwriteable" is better than an accidental "writeable" which
6521 * would cause an unwanted block under certain circumstances. For the 1-1
6522 * UDP-style sockets or TCP-style sockets, this code should work.
6525 static int sctp_writeable(struct sock *sk)
6529 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6535 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6536 * returns immediately with EINPROGRESS.
6538 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6540 struct sock *sk = asoc->base.sk;
6542 long current_timeo = *timeo_p;
6545 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6548 /* Increment the association's refcnt. */
6549 sctp_association_hold(asoc);
6552 prepare_to_wait_exclusive(&asoc->wait, &wait,
6553 TASK_INTERRUPTIBLE);
6556 if (sk->sk_shutdown & RCV_SHUTDOWN)
6558 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6561 if (signal_pending(current))
6562 goto do_interrupted;
6564 if (sctp_state(asoc, ESTABLISHED))
6567 /* Let another process have a go. Since we are going
6570 sctp_release_sock(sk);
6571 current_timeo = schedule_timeout(current_timeo);
6574 *timeo_p = current_timeo;
6578 finish_wait(&asoc->wait, &wait);
6580 /* Release the association's refcnt. */
6581 sctp_association_put(asoc);
6586 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6589 err = -ECONNREFUSED;
6593 err = sock_intr_errno(*timeo_p);
6601 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6603 struct sctp_endpoint *ep;
6607 ep = sctp_sk(sk)->ep;
6611 prepare_to_wait_exclusive(sk->sk_sleep, &wait,
6612 TASK_INTERRUPTIBLE);
6614 if (list_empty(&ep->asocs)) {
6615 sctp_release_sock(sk);
6616 timeo = schedule_timeout(timeo);
6621 if (!sctp_sstate(sk, LISTENING))
6625 if (!list_empty(&ep->asocs))
6628 err = sock_intr_errno(timeo);
6629 if (signal_pending(current))
6637 finish_wait(sk->sk_sleep, &wait);
6642 static void sctp_wait_for_close(struct sock *sk, long timeout)
6647 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
6648 if (list_empty(&sctp_sk(sk)->ep->asocs))
6650 sctp_release_sock(sk);
6651 timeout = schedule_timeout(timeout);
6653 } while (!signal_pending(current) && timeout);
6655 finish_wait(sk->sk_sleep, &wait);
6658 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6660 struct sk_buff *frag;
6665 /* Don't forget the fragments. */
6666 skb_walk_frags(skb, frag)
6667 sctp_skb_set_owner_r_frag(frag, sk);
6670 sctp_skb_set_owner_r(skb, sk);
6673 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6674 struct sctp_association *asoc)
6676 struct inet_sock *inet = inet_sk(sk);
6677 struct inet_sock *newinet = inet_sk(newsk);
6679 newsk->sk_type = sk->sk_type;
6680 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6681 newsk->sk_flags = sk->sk_flags;
6682 newsk->sk_no_check = sk->sk_no_check;
6683 newsk->sk_reuse = sk->sk_reuse;
6685 newsk->sk_shutdown = sk->sk_shutdown;
6686 newsk->sk_destruct = inet_sock_destruct;
6687 newsk->sk_family = sk->sk_family;
6688 newsk->sk_protocol = IPPROTO_SCTP;
6689 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6690 newsk->sk_sndbuf = sk->sk_sndbuf;
6691 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6692 newsk->sk_lingertime = sk->sk_lingertime;
6693 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6694 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6696 newinet = inet_sk(newsk);
6698 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6699 * getsockname() and getpeername()
6701 newinet->sport = inet->sport;
6702 newinet->saddr = inet->saddr;
6703 newinet->rcv_saddr = inet->rcv_saddr;
6704 newinet->dport = htons(asoc->peer.port);
6705 newinet->pmtudisc = inet->pmtudisc;
6706 newinet->id = asoc->next_tsn ^ jiffies;
6708 newinet->uc_ttl = inet->uc_ttl;
6709 newinet->mc_loop = 1;
6710 newinet->mc_ttl = 1;
6711 newinet->mc_index = 0;
6712 newinet->mc_list = NULL;
6715 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6716 * and its messages to the newsk.
6718 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6719 struct sctp_association *assoc,
6720 sctp_socket_type_t type)
6722 struct sctp_sock *oldsp = sctp_sk(oldsk);
6723 struct sctp_sock *newsp = sctp_sk(newsk);
6724 struct sctp_bind_bucket *pp; /* hash list port iterator */
6725 struct sctp_endpoint *newep = newsp->ep;
6726 struct sk_buff *skb, *tmp;
6727 struct sctp_ulpevent *event;
6728 struct sctp_bind_hashbucket *head;
6730 /* Migrate socket buffer sizes and all the socket level options to the
6733 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6734 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6735 /* Brute force copy old sctp opt. */
6736 inet_sk_copy_descendant(newsk, oldsk);
6738 /* Restore the ep value that was overwritten with the above structure
6744 /* Hook this new socket in to the bind_hash list. */
6745 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->num)];
6746 sctp_local_bh_disable();
6747 sctp_spin_lock(&head->lock);
6748 pp = sctp_sk(oldsk)->bind_hash;
6749 sk_add_bind_node(newsk, &pp->owner);
6750 sctp_sk(newsk)->bind_hash = pp;
6751 inet_sk(newsk)->num = inet_sk(oldsk)->num;
6752 sctp_spin_unlock(&head->lock);
6753 sctp_local_bh_enable();
6755 /* Copy the bind_addr list from the original endpoint to the new
6756 * endpoint so that we can handle restarts properly
6758 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6759 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6761 /* Move any messages in the old socket's receive queue that are for the
6762 * peeled off association to the new socket's receive queue.
6764 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6765 event = sctp_skb2event(skb);
6766 if (event->asoc == assoc) {
6767 __skb_unlink(skb, &oldsk->sk_receive_queue);
6768 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6769 sctp_skb_set_owner_r_frag(skb, newsk);
6773 /* Clean up any messages pending delivery due to partial
6774 * delivery. Three cases:
6775 * 1) No partial deliver; no work.
6776 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6777 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6779 skb_queue_head_init(&newsp->pd_lobby);
6780 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6782 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6783 struct sk_buff_head *queue;
6785 /* Decide which queue to move pd_lobby skbs to. */
6786 if (assoc->ulpq.pd_mode) {
6787 queue = &newsp->pd_lobby;
6789 queue = &newsk->sk_receive_queue;
6791 /* Walk through the pd_lobby, looking for skbs that
6792 * need moved to the new socket.
6794 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6795 event = sctp_skb2event(skb);
6796 if (event->asoc == assoc) {
6797 __skb_unlink(skb, &oldsp->pd_lobby);
6798 __skb_queue_tail(queue, skb);
6799 sctp_skb_set_owner_r_frag(skb, newsk);
6803 /* Clear up any skbs waiting for the partial
6804 * delivery to finish.
6806 if (assoc->ulpq.pd_mode)
6807 sctp_clear_pd(oldsk, NULL);
6811 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6812 sctp_skb_set_owner_r_frag(skb, newsk);
6814 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6815 sctp_skb_set_owner_r_frag(skb, newsk);
6817 /* Set the type of socket to indicate that it is peeled off from the
6818 * original UDP-style socket or created with the accept() call on a
6819 * TCP-style socket..
6823 /* Mark the new socket "in-use" by the user so that any packets
6824 * that may arrive on the association after we've moved it are
6825 * queued to the backlog. This prevents a potential race between
6826 * backlog processing on the old socket and new-packet processing
6827 * on the new socket.
6829 * The caller has just allocated newsk so we can guarantee that other
6830 * paths won't try to lock it and then oldsk.
6832 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6833 sctp_assoc_migrate(assoc, newsk);
6835 /* If the association on the newsk is already closed before accept()
6836 * is called, set RCV_SHUTDOWN flag.
6838 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6839 newsk->sk_shutdown |= RCV_SHUTDOWN;
6841 newsk->sk_state = SCTP_SS_ESTABLISHED;
6842 sctp_release_sock(newsk);
6846 /* This proto struct describes the ULP interface for SCTP. */
6847 struct proto sctp_prot = {
6849 .owner = THIS_MODULE,
6850 .close = sctp_close,
6851 .connect = sctp_connect,
6852 .disconnect = sctp_disconnect,
6853 .accept = sctp_accept,
6854 .ioctl = sctp_ioctl,
6855 .init = sctp_init_sock,
6856 .destroy = sctp_destroy_sock,
6857 .shutdown = sctp_shutdown,
6858 .setsockopt = sctp_setsockopt,
6859 .getsockopt = sctp_getsockopt,
6860 .sendmsg = sctp_sendmsg,
6861 .recvmsg = sctp_recvmsg,
6863 .backlog_rcv = sctp_backlog_rcv,
6865 .unhash = sctp_unhash,
6866 .get_port = sctp_get_port,
6867 .obj_size = sizeof(struct sctp_sock),
6868 .sysctl_mem = sysctl_sctp_mem,
6869 .sysctl_rmem = sysctl_sctp_rmem,
6870 .sysctl_wmem = sysctl_sctp_wmem,
6871 .memory_pressure = &sctp_memory_pressure,
6872 .enter_memory_pressure = sctp_enter_memory_pressure,
6873 .memory_allocated = &sctp_memory_allocated,
6874 .sockets_allocated = &sctp_sockets_allocated,
6877 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6879 struct proto sctpv6_prot = {
6881 .owner = THIS_MODULE,
6882 .close = sctp_close,
6883 .connect = sctp_connect,
6884 .disconnect = sctp_disconnect,
6885 .accept = sctp_accept,
6886 .ioctl = sctp_ioctl,
6887 .init = sctp_init_sock,
6888 .destroy = sctp_destroy_sock,
6889 .shutdown = sctp_shutdown,
6890 .setsockopt = sctp_setsockopt,
6891 .getsockopt = sctp_getsockopt,
6892 .sendmsg = sctp_sendmsg,
6893 .recvmsg = sctp_recvmsg,
6895 .backlog_rcv = sctp_backlog_rcv,
6897 .unhash = sctp_unhash,
6898 .get_port = sctp_get_port,
6899 .obj_size = sizeof(struct sctp6_sock),
6900 .sysctl_mem = sysctl_sctp_mem,
6901 .sysctl_rmem = sysctl_sctp_rmem,
6902 .sysctl_wmem = sysctl_sctp_wmem,
6903 .memory_pressure = &sctp_memory_pressure,
6904 .enter_memory_pressure = sctp_enter_memory_pressure,
6905 .memory_allocated = &sctp_memory_allocated,
6906 .sockets_allocated = &sctp_sockets_allocated,
6908 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */