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 static atomic_t sctp_sockets_allocated;
119 static void sctp_enter_memory_pressure(void)
121 sctp_memory_pressure = 1;
125 /* Get the sndbuf space available at the time on the association. */
126 static inline int sctp_wspace(struct sctp_association *asoc)
130 if (asoc->ep->sndbuf_policy)
131 amt = asoc->sndbuf_used;
133 amt = atomic_read(&asoc->base.sk->sk_wmem_alloc);
135 if (amt >= asoc->base.sk->sk_sndbuf) {
136 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
139 amt = sk_stream_wspace(asoc->base.sk);
144 amt = asoc->base.sk->sk_sndbuf - amt;
149 /* Increment the used sndbuf space count of the corresponding association by
150 * the size of the outgoing data chunk.
151 * Also, set the skb destructor for sndbuf accounting later.
153 * Since it is always 1-1 between chunk and skb, and also a new skb is always
154 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
155 * destructor in the data chunk skb for the purpose of the sndbuf space
158 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
160 struct sctp_association *asoc = chunk->asoc;
161 struct sock *sk = asoc->base.sk;
163 /* The sndbuf space is tracked per association. */
164 sctp_association_hold(asoc);
166 skb_set_owner_w(chunk->skb, sk);
168 chunk->skb->destructor = sctp_wfree;
169 /* Save the chunk pointer in skb for sctp_wfree to use later. */
170 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
172 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
173 sizeof(struct sk_buff) +
174 sizeof(struct sctp_chunk);
176 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
177 sk->sk_wmem_queued += chunk->skb->truesize;
178 sk_mem_charge(sk, chunk->skb->truesize);
181 /* Verify that this is a valid address. */
182 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
187 /* Verify basic sockaddr. */
188 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
192 /* Is this a valid SCTP address? */
193 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
196 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
202 /* Look up the association by its id. If this is not a UDP-style
203 * socket, the ID field is always ignored.
205 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
207 struct sctp_association *asoc = NULL;
209 /* If this is not a UDP-style socket, assoc id should be ignored. */
210 if (!sctp_style(sk, UDP)) {
211 /* Return NULL if the socket state is not ESTABLISHED. It
212 * could be a TCP-style listening socket or a socket which
213 * hasn't yet called connect() to establish an association.
215 if (!sctp_sstate(sk, ESTABLISHED))
218 /* Get the first and the only association from the list. */
219 if (!list_empty(&sctp_sk(sk)->ep->asocs))
220 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
221 struct sctp_association, asocs);
225 /* Otherwise this is a UDP-style socket. */
226 if (!id || (id == (sctp_assoc_t)-1))
229 spin_lock_bh(&sctp_assocs_id_lock);
230 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
231 spin_unlock_bh(&sctp_assocs_id_lock);
233 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
239 /* Look up the transport from an address and an assoc id. If both address and
240 * id are specified, the associations matching the address and the id should be
243 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
244 struct sockaddr_storage *addr,
247 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
248 struct sctp_transport *transport;
249 union sctp_addr *laddr = (union sctp_addr *)addr;
251 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
258 id_asoc = sctp_id2assoc(sk, id);
259 if (id_asoc && (id_asoc != addr_asoc))
262 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
263 (union sctp_addr *)addr);
268 /* API 3.1.2 bind() - UDP Style Syntax
269 * The syntax of bind() is,
271 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
273 * sd - the socket descriptor returned by socket().
274 * addr - the address structure (struct sockaddr_in or struct
275 * sockaddr_in6 [RFC 2553]),
276 * addr_len - the size of the address structure.
278 SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
284 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
287 /* Disallow binding twice. */
288 if (!sctp_sk(sk)->ep->base.bind_addr.port)
289 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
294 sctp_release_sock(sk);
299 static long sctp_get_port_local(struct sock *, union sctp_addr *);
301 /* Verify this is a valid sockaddr. */
302 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
303 union sctp_addr *addr, int len)
307 /* Check minimum size. */
308 if (len < sizeof (struct sockaddr))
311 /* Does this PF support this AF? */
312 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
315 /* If we get this far, af is valid. */
316 af = sctp_get_af_specific(addr->sa.sa_family);
318 if (len < af->sockaddr_len)
324 /* Bind a local address either to an endpoint or to an association. */
325 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
327 struct sctp_sock *sp = sctp_sk(sk);
328 struct sctp_endpoint *ep = sp->ep;
329 struct sctp_bind_addr *bp = &ep->base.bind_addr;
334 /* Common sockaddr verification. */
335 af = sctp_sockaddr_af(sp, addr, len);
337 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
342 snum = ntohs(addr->v4.sin_port);
344 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
345 ", port: %d, new port: %d, len: %d)\n",
351 /* PF specific bind() address verification. */
352 if (!sp->pf->bind_verify(sp, addr))
353 return -EADDRNOTAVAIL;
355 /* We must either be unbound, or bind to the same port.
356 * It's OK to allow 0 ports if we are already bound.
357 * We'll just inhert an already bound port in this case
362 else if (snum != bp->port) {
363 SCTP_DEBUG_PRINTK("sctp_do_bind:"
364 " New port %d does not match existing port "
365 "%d.\n", snum, bp->port);
370 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
373 /* Make sure we are allowed to bind here.
374 * The function sctp_get_port_local() does duplicate address
377 addr->v4.sin_port = htons(snum);
378 if ((ret = sctp_get_port_local(sk, addr))) {
379 if (ret == (long) sk) {
380 /* This endpoint has a conflicting address. */
387 /* Refresh ephemeral port. */
389 bp->port = inet_sk(sk)->num;
391 /* Add the address to the bind address list.
392 * Use GFP_ATOMIC since BHs will be disabled.
394 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
396 /* Copy back into socket for getsockname() use. */
398 inet_sk(sk)->sport = htons(inet_sk(sk)->num);
399 af->to_sk_saddr(addr, sk);
405 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
407 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
408 * at any one time. If a sender, after sending an ASCONF chunk, decides
409 * it needs to transfer another ASCONF Chunk, it MUST wait until the
410 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
411 * subsequent ASCONF. Note this restriction binds each side, so at any
412 * time two ASCONF may be in-transit on any given association (one sent
413 * from each endpoint).
415 static int sctp_send_asconf(struct sctp_association *asoc,
416 struct sctp_chunk *chunk)
420 /* If there is an outstanding ASCONF chunk, queue it for later
423 if (asoc->addip_last_asconf) {
424 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
428 /* Hold the chunk until an ASCONF_ACK is received. */
429 sctp_chunk_hold(chunk);
430 retval = sctp_primitive_ASCONF(asoc, chunk);
432 sctp_chunk_free(chunk);
434 asoc->addip_last_asconf = chunk;
440 /* Add a list of addresses as bind addresses to local endpoint or
443 * Basically run through each address specified in the addrs/addrcnt
444 * array/length pair, determine if it is IPv6 or IPv4 and call
445 * sctp_do_bind() on it.
447 * If any of them fails, then the operation will be reversed and the
448 * ones that were added will be removed.
450 * Only sctp_setsockopt_bindx() is supposed to call this function.
452 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
457 struct sockaddr *sa_addr;
460 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
464 for (cnt = 0; cnt < addrcnt; cnt++) {
465 /* The list may contain either IPv4 or IPv6 address;
466 * determine the address length for walking thru the list.
468 sa_addr = (struct sockaddr *)addr_buf;
469 af = sctp_get_af_specific(sa_addr->sa_family);
475 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
478 addr_buf += af->sockaddr_len;
482 /* Failed. Cleanup the ones that have been added */
484 sctp_bindx_rem(sk, addrs, cnt);
492 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
493 * associations that are part of the endpoint indicating that a list of local
494 * addresses are added to the endpoint.
496 * If any of the addresses is already in the bind address list of the
497 * association, we do not send the chunk for that association. But it will not
498 * affect other associations.
500 * Only sctp_setsockopt_bindx() is supposed to call this function.
502 static int sctp_send_asconf_add_ip(struct sock *sk,
503 struct sockaddr *addrs,
506 struct sctp_sock *sp;
507 struct sctp_endpoint *ep;
508 struct sctp_association *asoc;
509 struct sctp_bind_addr *bp;
510 struct sctp_chunk *chunk;
511 struct sctp_sockaddr_entry *laddr;
512 union sctp_addr *addr;
513 union sctp_addr saveaddr;
520 if (!sctp_addip_enable)
526 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
527 __func__, sk, addrs, addrcnt);
529 list_for_each_entry(asoc, &ep->asocs, asocs) {
531 if (!asoc->peer.asconf_capable)
534 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
537 if (!sctp_state(asoc, ESTABLISHED))
540 /* Check if any address in the packed array of addresses is
541 * in the bind address list of the association. If so,
542 * do not send the asconf chunk to its peer, but continue with
543 * other associations.
546 for (i = 0; i < addrcnt; i++) {
547 addr = (union sctp_addr *)addr_buf;
548 af = sctp_get_af_specific(addr->v4.sin_family);
554 if (sctp_assoc_lookup_laddr(asoc, addr))
557 addr_buf += af->sockaddr_len;
562 /* Use the first valid address in bind addr list of
563 * association as Address Parameter of ASCONF CHUNK.
565 bp = &asoc->base.bind_addr;
566 p = bp->address_list.next;
567 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
568 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
569 addrcnt, SCTP_PARAM_ADD_IP);
575 retval = sctp_send_asconf(asoc, chunk);
579 /* Add the new addresses to the bind address list with
580 * use_as_src set to 0.
583 for (i = 0; i < addrcnt; i++) {
584 addr = (union sctp_addr *)addr_buf;
585 af = sctp_get_af_specific(addr->v4.sin_family);
586 memcpy(&saveaddr, addr, af->sockaddr_len);
587 retval = sctp_add_bind_addr(bp, &saveaddr,
588 SCTP_ADDR_NEW, GFP_ATOMIC);
589 addr_buf += af->sockaddr_len;
597 /* Remove a list of addresses from bind addresses list. Do not remove the
600 * Basically run through each address specified in the addrs/addrcnt
601 * array/length pair, determine if it is IPv6 or IPv4 and call
602 * sctp_del_bind() on it.
604 * If any of them fails, then the operation will be reversed and the
605 * ones that were removed will be added back.
607 * At least one address has to be left; if only one address is
608 * available, the operation will return -EBUSY.
610 * Only sctp_setsockopt_bindx() is supposed to call this function.
612 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
614 struct sctp_sock *sp = sctp_sk(sk);
615 struct sctp_endpoint *ep = sp->ep;
617 struct sctp_bind_addr *bp = &ep->base.bind_addr;
620 union sctp_addr *sa_addr;
623 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
627 for (cnt = 0; cnt < addrcnt; cnt++) {
628 /* If the bind address list is empty or if there is only one
629 * bind address, there is nothing more to be removed (we need
630 * at least one address here).
632 if (list_empty(&bp->address_list) ||
633 (sctp_list_single_entry(&bp->address_list))) {
638 sa_addr = (union sctp_addr *)addr_buf;
639 af = sctp_get_af_specific(sa_addr->sa.sa_family);
645 if (!af->addr_valid(sa_addr, sp, NULL)) {
646 retval = -EADDRNOTAVAIL;
650 if (sa_addr->v4.sin_port != htons(bp->port)) {
655 /* FIXME - There is probably a need to check if sk->sk_saddr and
656 * sk->sk_rcv_addr are currently set to one of the addresses to
657 * be removed. This is something which needs to be looked into
658 * when we are fixing the outstanding issues with multi-homing
659 * socket routing and failover schemes. Refer to comments in
660 * sctp_do_bind(). -daisy
662 retval = sctp_del_bind_addr(bp, sa_addr);
664 addr_buf += af->sockaddr_len;
667 /* Failed. Add the ones that has been removed back */
669 sctp_bindx_add(sk, addrs, cnt);
677 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
678 * the associations that are part of the endpoint indicating that a list of
679 * local addresses are removed from the endpoint.
681 * If any of the addresses is already in the bind address list of the
682 * association, we do not send the chunk for that association. But it will not
683 * affect other associations.
685 * Only sctp_setsockopt_bindx() is supposed to call this function.
687 static int sctp_send_asconf_del_ip(struct sock *sk,
688 struct sockaddr *addrs,
691 struct sctp_sock *sp;
692 struct sctp_endpoint *ep;
693 struct sctp_association *asoc;
694 struct sctp_transport *transport;
695 struct sctp_bind_addr *bp;
696 struct sctp_chunk *chunk;
697 union sctp_addr *laddr;
700 struct sctp_sockaddr_entry *saddr;
704 if (!sctp_addip_enable)
710 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
711 __func__, sk, addrs, addrcnt);
713 list_for_each_entry(asoc, &ep->asocs, asocs) {
715 if (!asoc->peer.asconf_capable)
718 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
721 if (!sctp_state(asoc, ESTABLISHED))
724 /* Check if any address in the packed array of addresses is
725 * not present in the bind address list of the association.
726 * If so, do not send the asconf chunk to its peer, but
727 * continue with other associations.
730 for (i = 0; i < addrcnt; i++) {
731 laddr = (union sctp_addr *)addr_buf;
732 af = sctp_get_af_specific(laddr->v4.sin_family);
738 if (!sctp_assoc_lookup_laddr(asoc, laddr))
741 addr_buf += af->sockaddr_len;
746 /* Find one address in the association's bind address list
747 * that is not in the packed array of addresses. This is to
748 * make sure that we do not delete all the addresses in the
751 bp = &asoc->base.bind_addr;
752 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
757 /* We do not need RCU protection throughout this loop
758 * because this is done under a socket lock from the
761 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
768 /* Reset use_as_src flag for the addresses in the bind address
769 * list that are to be deleted.
772 for (i = 0; i < addrcnt; i++) {
773 laddr = (union sctp_addr *)addr_buf;
774 af = sctp_get_af_specific(laddr->v4.sin_family);
775 list_for_each_entry(saddr, &bp->address_list, list) {
776 if (sctp_cmp_addr_exact(&saddr->a, laddr))
777 saddr->state = SCTP_ADDR_DEL;
779 addr_buf += af->sockaddr_len;
782 /* Update the route and saddr entries for all the transports
783 * as some of the addresses in the bind address list are
784 * about to be deleted and cannot be used as source addresses.
786 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
788 dst_release(transport->dst);
789 sctp_transport_route(transport, NULL,
790 sctp_sk(asoc->base.sk));
793 retval = sctp_send_asconf(asoc, chunk);
799 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
802 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
805 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
806 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
809 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
810 * Section 3.1.2 for this usage.
812 * addrs is a pointer to an array of one or more socket addresses. Each
813 * address is contained in its appropriate structure (i.e. struct
814 * sockaddr_in or struct sockaddr_in6) the family of the address type
815 * must be used to distinguish the address length (note that this
816 * representation is termed a "packed array" of addresses). The caller
817 * specifies the number of addresses in the array with addrcnt.
819 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
820 * -1, and sets errno to the appropriate error code.
822 * For SCTP, the port given in each socket address must be the same, or
823 * sctp_bindx() will fail, setting errno to EINVAL.
825 * The flags parameter is formed from the bitwise OR of zero or more of
826 * the following currently defined flags:
828 * SCTP_BINDX_ADD_ADDR
830 * SCTP_BINDX_REM_ADDR
832 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
833 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
834 * addresses from the association. The two flags are mutually exclusive;
835 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
836 * not remove all addresses from an association; sctp_bindx() will
837 * reject such an attempt with EINVAL.
839 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
840 * additional addresses with an endpoint after calling bind(). Or use
841 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
842 * socket is associated with so that no new association accepted will be
843 * associated with those addresses. If the endpoint supports dynamic
844 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
845 * endpoint to send the appropriate message to the peer to change the
846 * peers address lists.
848 * Adding and removing addresses from a connected association is
849 * optional functionality. Implementations that do not support this
850 * functionality should return EOPNOTSUPP.
852 * Basically do nothing but copying the addresses from user to kernel
853 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
854 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
857 * We don't use copy_from_user() for optimization: we first do the
858 * sanity checks (buffer size -fast- and access check-healthy
859 * pointer); if all of those succeed, then we can alloc the memory
860 * (expensive operation) needed to copy the data to kernel. Then we do
861 * the copying without checking the user space area
862 * (__copy_from_user()).
864 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
867 * sk The sk of the socket
868 * addrs The pointer to the addresses in user land
869 * addrssize Size of the addrs buffer
870 * op Operation to perform (add or remove, see the flags of
873 * Returns 0 if ok, <0 errno code on error.
875 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
876 struct sockaddr __user *addrs,
877 int addrs_size, int op)
879 struct sockaddr *kaddrs;
883 struct sockaddr *sa_addr;
887 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
888 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
890 if (unlikely(addrs_size <= 0))
893 /* Check the user passed a healthy pointer. */
894 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
897 /* Alloc space for the address array in kernel memory. */
898 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
899 if (unlikely(!kaddrs))
902 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
907 /* Walk through the addrs buffer and count the number of addresses. */
909 while (walk_size < addrs_size) {
910 sa_addr = (struct sockaddr *)addr_buf;
911 af = sctp_get_af_specific(sa_addr->sa_family);
913 /* If the address family is not supported or if this address
914 * causes the address buffer to overflow return EINVAL.
916 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
921 addr_buf += af->sockaddr_len;
922 walk_size += af->sockaddr_len;
927 case SCTP_BINDX_ADD_ADDR:
928 err = sctp_bindx_add(sk, kaddrs, addrcnt);
931 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
934 case SCTP_BINDX_REM_ADDR:
935 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
938 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
952 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
954 * Common routine for handling connect() and sctp_connectx().
955 * Connect will come in with just a single address.
957 static int __sctp_connect(struct sock* sk,
958 struct sockaddr *kaddrs,
961 struct sctp_sock *sp;
962 struct sctp_endpoint *ep;
963 struct sctp_association *asoc = NULL;
964 struct sctp_association *asoc2;
965 struct sctp_transport *transport;
973 union sctp_addr *sa_addr = NULL;
976 unsigned int f_flags = 0;
981 /* connect() cannot be done on a socket that is already in ESTABLISHED
982 * state - UDP-style peeled off socket or a TCP-style socket that
983 * is already connected.
984 * It cannot be done even on a TCP-style listening socket.
986 if (sctp_sstate(sk, ESTABLISHED) ||
987 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
992 /* Walk through the addrs buffer and count the number of addresses. */
994 while (walk_size < addrs_size) {
995 sa_addr = (union sctp_addr *)addr_buf;
996 af = sctp_get_af_specific(sa_addr->sa.sa_family);
997 port = ntohs(sa_addr->v4.sin_port);
999 /* If the address family is not supported or if this address
1000 * causes the address buffer to overflow return EINVAL.
1002 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1007 /* Save current address so we can work with it */
1008 memcpy(&to, sa_addr, af->sockaddr_len);
1010 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1014 /* Make sure the destination port is correctly set
1017 if (asoc && asoc->peer.port && asoc->peer.port != port)
1021 /* Check if there already is a matching association on the
1022 * endpoint (other than the one created here).
1024 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1025 if (asoc2 && asoc2 != asoc) {
1026 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1033 /* If we could not find a matching association on the endpoint,
1034 * make sure that there is no peeled-off association matching
1035 * the peer address even on another socket.
1037 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1038 err = -EADDRNOTAVAIL;
1043 /* If a bind() or sctp_bindx() is not called prior to
1044 * an sctp_connectx() call, the system picks an
1045 * ephemeral port and will choose an address set
1046 * equivalent to binding with a wildcard address.
1048 if (!ep->base.bind_addr.port) {
1049 if (sctp_autobind(sk)) {
1055 * If an unprivileged user inherits a 1-many
1056 * style socket with open associations on a
1057 * privileged port, it MAY be permitted to
1058 * accept new associations, but it SHOULD NOT
1059 * be permitted to open new associations.
1061 if (ep->base.bind_addr.port < PROT_SOCK &&
1062 !capable(CAP_NET_BIND_SERVICE)) {
1068 scope = sctp_scope(&to);
1069 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1076 /* Prime the peer's transport structures. */
1077 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1085 addr_buf += af->sockaddr_len;
1086 walk_size += af->sockaddr_len;
1089 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1094 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1099 /* Initialize sk's dport and daddr for getpeername() */
1100 inet_sk(sk)->dport = htons(asoc->peer.port);
1101 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1102 af->to_sk_daddr(sa_addr, sk);
1105 /* in-kernel sockets don't generally have a file allocated to them
1106 * if all they do is call sock_create_kern().
1108 if (sk->sk_socket->file)
1109 f_flags = sk->sk_socket->file->f_flags;
1111 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1113 err = sctp_wait_for_connect(asoc, &timeo);
1115 /* Don't free association on exit. */
1120 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1121 " kaddrs: %p err: %d\n",
1124 sctp_association_free(asoc);
1128 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1131 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt);
1133 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1134 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1135 * or IPv6 addresses.
1137 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1138 * Section 3.1.2 for this usage.
1140 * addrs is a pointer to an array of one or more socket addresses. Each
1141 * address is contained in its appropriate structure (i.e. struct
1142 * sockaddr_in or struct sockaddr_in6) the family of the address type
1143 * must be used to distengish the address length (note that this
1144 * representation is termed a "packed array" of addresses). The caller
1145 * specifies the number of addresses in the array with addrcnt.
1147 * On success, sctp_connectx() returns 0. On failure, sctp_connectx() returns
1148 * -1, and sets errno to the appropriate error code.
1150 * For SCTP, the port given in each socket address must be the same, or
1151 * sctp_connectx() will fail, setting errno to EINVAL.
1153 * An application can use sctp_connectx to initiate an association with
1154 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1155 * allows a caller to specify multiple addresses at which a peer can be
1156 * reached. The way the SCTP stack uses the list of addresses to set up
1157 * the association is implementation dependant. This function only
1158 * specifies that the stack will try to make use of all the addresses in
1159 * the list when needed.
1161 * Note that the list of addresses passed in is only used for setting up
1162 * the association. It does not necessarily equal the set of addresses
1163 * the peer uses for the resulting association. If the caller wants to
1164 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1165 * retrieve them after the association has been set up.
1167 * Basically do nothing but copying the addresses from user to kernel
1168 * land and invoking either sctp_connectx(). This is used for tunneling
1169 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1171 * We don't use copy_from_user() for optimization: we first do the
1172 * sanity checks (buffer size -fast- and access check-healthy
1173 * pointer); if all of those succeed, then we can alloc the memory
1174 * (expensive operation) needed to copy the data to kernel. Then we do
1175 * the copying without checking the user space area
1176 * (__copy_from_user()).
1178 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1181 * sk The sk of the socket
1182 * addrs The pointer to the addresses in user land
1183 * addrssize Size of the addrs buffer
1185 * Returns 0 if ok, <0 errno code on error.
1187 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1188 struct sockaddr __user *addrs,
1192 struct sockaddr *kaddrs;
1194 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1195 __func__, sk, addrs, addrs_size);
1197 if (unlikely(addrs_size <= 0))
1200 /* Check the user passed a healthy pointer. */
1201 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1204 /* Alloc space for the address array in kernel memory. */
1205 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1206 if (unlikely(!kaddrs))
1209 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1212 err = __sctp_connect(sk, kaddrs, addrs_size);
1219 /* API 3.1.4 close() - UDP Style Syntax
1220 * Applications use close() to perform graceful shutdown (as described in
1221 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1222 * by a UDP-style socket.
1226 * ret = close(int sd);
1228 * sd - the socket descriptor of the associations to be closed.
1230 * To gracefully shutdown a specific association represented by the
1231 * UDP-style socket, an application should use the sendmsg() call,
1232 * passing no user data, but including the appropriate flag in the
1233 * ancillary data (see Section xxxx).
1235 * If sd in the close() call is a branched-off socket representing only
1236 * one association, the shutdown is performed on that association only.
1238 * 4.1.6 close() - TCP Style Syntax
1240 * Applications use close() to gracefully close down an association.
1244 * int close(int sd);
1246 * sd - the socket descriptor of the association to be closed.
1248 * After an application calls close() on a socket descriptor, no further
1249 * socket operations will succeed on that descriptor.
1251 * API 7.1.4 SO_LINGER
1253 * An application using the TCP-style socket can use this option to
1254 * perform the SCTP ABORT primitive. The linger option structure is:
1257 * int l_onoff; // option on/off
1258 * int l_linger; // linger time
1261 * To enable the option, set l_onoff to 1. If the l_linger value is set
1262 * to 0, calling close() is the same as the ABORT primitive. If the
1263 * value is set to a negative value, the setsockopt() call will return
1264 * an error. If the value is set to a positive value linger_time, the
1265 * close() can be blocked for at most linger_time ms. If the graceful
1266 * shutdown phase does not finish during this period, close() will
1267 * return but the graceful shutdown phase continues in the system.
1269 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1271 struct sctp_endpoint *ep;
1272 struct sctp_association *asoc;
1273 struct list_head *pos, *temp;
1275 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1278 sk->sk_shutdown = SHUTDOWN_MASK;
1280 ep = sctp_sk(sk)->ep;
1282 /* Walk all associations on an endpoint. */
1283 list_for_each_safe(pos, temp, &ep->asocs) {
1284 asoc = list_entry(pos, struct sctp_association, asocs);
1286 if (sctp_style(sk, TCP)) {
1287 /* A closed association can still be in the list if
1288 * it belongs to a TCP-style listening socket that is
1289 * not yet accepted. If so, free it. If not, send an
1290 * ABORT or SHUTDOWN based on the linger options.
1292 if (sctp_state(asoc, CLOSED)) {
1293 sctp_unhash_established(asoc);
1294 sctp_association_free(asoc);
1299 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1300 struct sctp_chunk *chunk;
1302 chunk = sctp_make_abort_user(asoc, NULL, 0);
1304 sctp_primitive_ABORT(asoc, chunk);
1306 sctp_primitive_SHUTDOWN(asoc, NULL);
1309 /* Clean up any skbs sitting on the receive queue. */
1310 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1311 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1313 /* On a TCP-style socket, block for at most linger_time if set. */
1314 if (sctp_style(sk, TCP) && timeout)
1315 sctp_wait_for_close(sk, timeout);
1317 /* This will run the backlog queue. */
1318 sctp_release_sock(sk);
1320 /* Supposedly, no process has access to the socket, but
1321 * the net layers still may.
1323 sctp_local_bh_disable();
1324 sctp_bh_lock_sock(sk);
1326 /* Hold the sock, since sk_common_release() will put sock_put()
1327 * and we have just a little more cleanup.
1330 sk_common_release(sk);
1332 sctp_bh_unlock_sock(sk);
1333 sctp_local_bh_enable();
1337 SCTP_DBG_OBJCNT_DEC(sock);
1340 /* Handle EPIPE error. */
1341 static int sctp_error(struct sock *sk, int flags, int err)
1344 err = sock_error(sk) ? : -EPIPE;
1345 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1346 send_sig(SIGPIPE, current, 0);
1350 /* API 3.1.3 sendmsg() - UDP Style Syntax
1352 * An application uses sendmsg() and recvmsg() calls to transmit data to
1353 * and receive data from its peer.
1355 * ssize_t sendmsg(int socket, const struct msghdr *message,
1358 * socket - the socket descriptor of the endpoint.
1359 * message - pointer to the msghdr structure which contains a single
1360 * user message and possibly some ancillary data.
1362 * See Section 5 for complete description of the data
1365 * flags - flags sent or received with the user message, see Section
1366 * 5 for complete description of the flags.
1368 * Note: This function could use a rewrite especially when explicit
1369 * connect support comes in.
1371 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1373 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1375 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1376 struct msghdr *msg, size_t msg_len)
1378 struct sctp_sock *sp;
1379 struct sctp_endpoint *ep;
1380 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1381 struct sctp_transport *transport, *chunk_tp;
1382 struct sctp_chunk *chunk;
1384 struct sockaddr *msg_name = NULL;
1385 struct sctp_sndrcvinfo default_sinfo = { 0 };
1386 struct sctp_sndrcvinfo *sinfo;
1387 struct sctp_initmsg *sinit;
1388 sctp_assoc_t associd = 0;
1389 sctp_cmsgs_t cmsgs = { NULL };
1393 __u16 sinfo_flags = 0;
1394 struct sctp_datamsg *datamsg;
1395 int msg_flags = msg->msg_flags;
1397 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1404 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1406 /* We cannot send a message over a TCP-style listening socket. */
1407 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1412 /* Parse out the SCTP CMSGs. */
1413 err = sctp_msghdr_parse(msg, &cmsgs);
1416 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1420 /* Fetch the destination address for this packet. This
1421 * address only selects the association--it is not necessarily
1422 * the address we will send to.
1423 * For a peeled-off socket, msg_name is ignored.
1425 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1426 int msg_namelen = msg->msg_namelen;
1428 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1433 if (msg_namelen > sizeof(to))
1434 msg_namelen = sizeof(to);
1435 memcpy(&to, msg->msg_name, msg_namelen);
1436 msg_name = msg->msg_name;
1442 /* Did the user specify SNDRCVINFO? */
1444 sinfo_flags = sinfo->sinfo_flags;
1445 associd = sinfo->sinfo_assoc_id;
1448 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1449 msg_len, sinfo_flags);
1451 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1452 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1457 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1458 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1459 * If SCTP_ABORT is set, the message length could be non zero with
1460 * the msg_iov set to the user abort reason.
1462 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1463 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1468 /* If SCTP_ADDR_OVER is set, there must be an address
1469 * specified in msg_name.
1471 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1478 SCTP_DEBUG_PRINTK("About to look up association.\n");
1482 /* If a msg_name has been specified, assume this is to be used. */
1484 /* Look for a matching association on the endpoint. */
1485 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1487 /* If we could not find a matching association on the
1488 * endpoint, make sure that it is not a TCP-style
1489 * socket that already has an association or there is
1490 * no peeled-off association on another socket.
1492 if ((sctp_style(sk, TCP) &&
1493 sctp_sstate(sk, ESTABLISHED)) ||
1494 sctp_endpoint_is_peeled_off(ep, &to)) {
1495 err = -EADDRNOTAVAIL;
1500 asoc = sctp_id2assoc(sk, associd);
1508 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1510 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1511 * socket that has an association in CLOSED state. This can
1512 * happen when an accepted socket has an association that is
1515 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1520 if (sinfo_flags & SCTP_EOF) {
1521 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1523 sctp_primitive_SHUTDOWN(asoc, NULL);
1527 if (sinfo_flags & SCTP_ABORT) {
1529 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1535 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1536 sctp_primitive_ABORT(asoc, chunk);
1542 /* Do we need to create the association? */
1544 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1546 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1551 /* Check for invalid stream against the stream counts,
1552 * either the default or the user specified stream counts.
1555 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1556 /* Check against the defaults. */
1557 if (sinfo->sinfo_stream >=
1558 sp->initmsg.sinit_num_ostreams) {
1563 /* Check against the requested. */
1564 if (sinfo->sinfo_stream >=
1565 sinit->sinit_num_ostreams) {
1573 * API 3.1.2 bind() - UDP Style Syntax
1574 * If a bind() or sctp_bindx() is not called prior to a
1575 * sendmsg() call that initiates a new association, the
1576 * system picks an ephemeral port and will choose an address
1577 * set equivalent to binding with a wildcard address.
1579 if (!ep->base.bind_addr.port) {
1580 if (sctp_autobind(sk)) {
1586 * If an unprivileged user inherits a one-to-many
1587 * style socket with open associations on a privileged
1588 * port, it MAY be permitted to accept new associations,
1589 * but it SHOULD NOT be permitted to open new
1592 if (ep->base.bind_addr.port < PROT_SOCK &&
1593 !capable(CAP_NET_BIND_SERVICE)) {
1599 scope = sctp_scope(&to);
1600 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1607 /* If the SCTP_INIT ancillary data is specified, set all
1608 * the association init values accordingly.
1611 if (sinit->sinit_num_ostreams) {
1612 asoc->c.sinit_num_ostreams =
1613 sinit->sinit_num_ostreams;
1615 if (sinit->sinit_max_instreams) {
1616 asoc->c.sinit_max_instreams =
1617 sinit->sinit_max_instreams;
1619 if (sinit->sinit_max_attempts) {
1620 asoc->max_init_attempts
1621 = sinit->sinit_max_attempts;
1623 if (sinit->sinit_max_init_timeo) {
1624 asoc->max_init_timeo =
1625 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1629 /* Prime the peer's transport structures. */
1630 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1635 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1642 /* ASSERT: we have a valid association at this point. */
1643 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1646 /* If the user didn't specify SNDRCVINFO, make up one with
1649 default_sinfo.sinfo_stream = asoc->default_stream;
1650 default_sinfo.sinfo_flags = asoc->default_flags;
1651 default_sinfo.sinfo_ppid = asoc->default_ppid;
1652 default_sinfo.sinfo_context = asoc->default_context;
1653 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1654 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1655 sinfo = &default_sinfo;
1658 /* API 7.1.7, the sndbuf size per association bounds the
1659 * maximum size of data that can be sent in a single send call.
1661 if (msg_len > sk->sk_sndbuf) {
1666 if (asoc->pmtu_pending)
1667 sctp_assoc_pending_pmtu(asoc);
1669 /* If fragmentation is disabled and the message length exceeds the
1670 * association fragmentation point, return EMSGSIZE. The I-D
1671 * does not specify what this error is, but this looks like
1674 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1680 /* Check for invalid stream. */
1681 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1687 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1688 if (!sctp_wspace(asoc)) {
1689 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1694 /* If an address is passed with the sendto/sendmsg call, it is used
1695 * to override the primary destination address in the TCP model, or
1696 * when SCTP_ADDR_OVER flag is set in the UDP model.
1698 if ((sctp_style(sk, TCP) && msg_name) ||
1699 (sinfo_flags & SCTP_ADDR_OVER)) {
1700 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1708 /* Auto-connect, if we aren't connected already. */
1709 if (sctp_state(asoc, CLOSED)) {
1710 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1713 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1716 /* Break the message into multiple chunks of maximum size. */
1717 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1723 /* Now send the (possibly) fragmented message. */
1724 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1725 sctp_chunk_hold(chunk);
1727 /* Do accounting for the write space. */
1728 sctp_set_owner_w(chunk);
1730 chunk->transport = chunk_tp;
1732 /* Send it to the lower layers. Note: all chunks
1733 * must either fail or succeed. The lower layer
1734 * works that way today. Keep it that way or this
1737 err = sctp_primitive_SEND(asoc, chunk);
1738 /* Did the lower layer accept the chunk? */
1740 sctp_chunk_free(chunk);
1741 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1744 sctp_datamsg_put(datamsg);
1750 /* If we are already past ASSOCIATE, the lower
1751 * layers are responsible for association cleanup.
1757 sctp_association_free(asoc);
1759 sctp_release_sock(sk);
1762 return sctp_error(sk, msg_flags, err);
1769 err = sock_error(sk);
1779 /* This is an extended version of skb_pull() that removes the data from the
1780 * start of a skb even when data is spread across the list of skb's in the
1781 * frag_list. len specifies the total amount of data that needs to be removed.
1782 * when 'len' bytes could be removed from the skb, it returns 0.
1783 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1784 * could not be removed.
1786 static int sctp_skb_pull(struct sk_buff *skb, int len)
1788 struct sk_buff *list;
1789 int skb_len = skb_headlen(skb);
1792 if (len <= skb_len) {
1793 __skb_pull(skb, len);
1797 __skb_pull(skb, skb_len);
1799 for (list = skb_shinfo(skb)->frag_list; list; list = list->next) {
1800 rlen = sctp_skb_pull(list, len);
1801 skb->len -= (len-rlen);
1802 skb->data_len -= (len-rlen);
1813 /* API 3.1.3 recvmsg() - UDP Style Syntax
1815 * ssize_t recvmsg(int socket, struct msghdr *message,
1818 * socket - the socket descriptor of the endpoint.
1819 * message - pointer to the msghdr structure which contains a single
1820 * user message and possibly some ancillary data.
1822 * See Section 5 for complete description of the data
1825 * flags - flags sent or received with the user message, see Section
1826 * 5 for complete description of the flags.
1828 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1830 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1831 struct msghdr *msg, size_t len, int noblock,
1832 int flags, int *addr_len)
1834 struct sctp_ulpevent *event = NULL;
1835 struct sctp_sock *sp = sctp_sk(sk);
1836 struct sk_buff *skb;
1841 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1842 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1843 "len", len, "knoblauch", noblock,
1844 "flags", flags, "addr_len", addr_len);
1848 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1853 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1857 /* Get the total length of the skb including any skb's in the
1866 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1868 event = sctp_skb2event(skb);
1873 sock_recv_timestamp(msg, sk, skb);
1874 if (sctp_ulpevent_is_notification(event)) {
1875 msg->msg_flags |= MSG_NOTIFICATION;
1876 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1878 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1881 /* Check if we allow SCTP_SNDRCVINFO. */
1882 if (sp->subscribe.sctp_data_io_event)
1883 sctp_ulpevent_read_sndrcvinfo(event, msg);
1885 /* FIXME: we should be calling IP/IPv6 layers. */
1886 if (sk->sk_protinfo.af_inet.cmsg_flags)
1887 ip_cmsg_recv(msg, skb);
1892 /* If skb's length exceeds the user's buffer, update the skb and
1893 * push it back to the receive_queue so that the next call to
1894 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1896 if (skb_len > copied) {
1897 msg->msg_flags &= ~MSG_EOR;
1898 if (flags & MSG_PEEK)
1900 sctp_skb_pull(skb, copied);
1901 skb_queue_head(&sk->sk_receive_queue, skb);
1903 /* When only partial message is copied to the user, increase
1904 * rwnd by that amount. If all the data in the skb is read,
1905 * rwnd is updated when the event is freed.
1907 if (!sctp_ulpevent_is_notification(event))
1908 sctp_assoc_rwnd_increase(event->asoc, copied);
1910 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
1911 (event->msg_flags & MSG_EOR))
1912 msg->msg_flags |= MSG_EOR;
1914 msg->msg_flags &= ~MSG_EOR;
1917 if (flags & MSG_PEEK) {
1918 /* Release the skb reference acquired after peeking the skb in
1919 * sctp_skb_recv_datagram().
1923 /* Free the event which includes releasing the reference to
1924 * the owner of the skb, freeing the skb and updating the
1927 sctp_ulpevent_free(event);
1930 sctp_release_sock(sk);
1934 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
1936 * This option is a on/off flag. If enabled no SCTP message
1937 * fragmentation will be performed. Instead if a message being sent
1938 * exceeds the current PMTU size, the message will NOT be sent and
1939 * instead a error will be indicated to the user.
1941 static int sctp_setsockopt_disable_fragments(struct sock *sk,
1942 char __user *optval, int optlen)
1946 if (optlen < sizeof(int))
1949 if (get_user(val, (int __user *)optval))
1952 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
1957 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
1960 if (optlen > sizeof(struct sctp_event_subscribe))
1962 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
1967 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
1969 * This socket option is applicable to the UDP-style socket only. When
1970 * set it will cause associations that are idle for more than the
1971 * specified number of seconds to automatically close. An association
1972 * being idle is defined an association that has NOT sent or received
1973 * user data. The special value of '0' indicates that no automatic
1974 * close of any associations should be performed. The option expects an
1975 * integer defining the number of seconds of idle time before an
1976 * association is closed.
1978 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
1981 struct sctp_sock *sp = sctp_sk(sk);
1983 /* Applicable to UDP-style socket only */
1984 if (sctp_style(sk, TCP))
1986 if (optlen != sizeof(int))
1988 if (copy_from_user(&sp->autoclose, optval, optlen))
1994 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
1996 * Applications can enable or disable heartbeats for any peer address of
1997 * an association, modify an address's heartbeat interval, force a
1998 * heartbeat to be sent immediately, and adjust the address's maximum
1999 * number of retransmissions sent before an address is considered
2000 * unreachable. The following structure is used to access and modify an
2001 * address's parameters:
2003 * struct sctp_paddrparams {
2004 * sctp_assoc_t spp_assoc_id;
2005 * struct sockaddr_storage spp_address;
2006 * uint32_t spp_hbinterval;
2007 * uint16_t spp_pathmaxrxt;
2008 * uint32_t spp_pathmtu;
2009 * uint32_t spp_sackdelay;
2010 * uint32_t spp_flags;
2013 * spp_assoc_id - (one-to-many style socket) This is filled in the
2014 * application, and identifies the association for
2016 * spp_address - This specifies which address is of interest.
2017 * spp_hbinterval - This contains the value of the heartbeat interval,
2018 * in milliseconds. If a value of zero
2019 * is present in this field then no changes are to
2020 * be made to this parameter.
2021 * spp_pathmaxrxt - This contains the maximum number of
2022 * retransmissions before this address shall be
2023 * considered unreachable. If a value of zero
2024 * is present in this field then no changes are to
2025 * be made to this parameter.
2026 * spp_pathmtu - When Path MTU discovery is disabled the value
2027 * specified here will be the "fixed" path mtu.
2028 * Note that if the spp_address field is empty
2029 * then all associations on this address will
2030 * have this fixed path mtu set upon them.
2032 * spp_sackdelay - When delayed sack is enabled, this value specifies
2033 * the number of milliseconds that sacks will be delayed
2034 * for. This value will apply to all addresses of an
2035 * association if the spp_address field is empty. Note
2036 * also, that if delayed sack is enabled and this
2037 * value is set to 0, no change is made to the last
2038 * recorded delayed sack timer value.
2040 * spp_flags - These flags are used to control various features
2041 * on an association. The flag field may contain
2042 * zero or more of the following options.
2044 * SPP_HB_ENABLE - Enable heartbeats on the
2045 * specified address. Note that if the address
2046 * field is empty all addresses for the association
2047 * have heartbeats enabled upon them.
2049 * SPP_HB_DISABLE - Disable heartbeats on the
2050 * speicifed address. Note that if the address
2051 * field is empty all addresses for the association
2052 * will have their heartbeats disabled. Note also
2053 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2054 * mutually exclusive, only one of these two should
2055 * be specified. Enabling both fields will have
2056 * undetermined results.
2058 * SPP_HB_DEMAND - Request a user initiated heartbeat
2059 * to be made immediately.
2061 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2062 * heartbeat delayis to be set to the value of 0
2065 * SPP_PMTUD_ENABLE - This field will enable PMTU
2066 * discovery upon the specified address. Note that
2067 * if the address feild is empty then all addresses
2068 * on the association are effected.
2070 * SPP_PMTUD_DISABLE - This field will disable PMTU
2071 * discovery upon the specified address. Note that
2072 * if the address feild is empty then all addresses
2073 * on the association are effected. Not also that
2074 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2075 * exclusive. Enabling both will have undetermined
2078 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2079 * on delayed sack. The time specified in spp_sackdelay
2080 * is used to specify the sack delay for this address. Note
2081 * that if spp_address is empty then all addresses will
2082 * enable delayed sack and take on the sack delay
2083 * value specified in spp_sackdelay.
2084 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2085 * off delayed sack. If the spp_address field is blank then
2086 * delayed sack is disabled for the entire association. Note
2087 * also that this field is mutually exclusive to
2088 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2091 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2092 struct sctp_transport *trans,
2093 struct sctp_association *asoc,
2094 struct sctp_sock *sp,
2097 int sackdelay_change)
2101 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2102 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2107 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2108 * this field is ignored. Note also that a value of zero indicates
2109 * the current setting should be left unchanged.
2111 if (params->spp_flags & SPP_HB_ENABLE) {
2113 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2114 * set. This lets us use 0 value when this flag
2117 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2118 params->spp_hbinterval = 0;
2120 if (params->spp_hbinterval ||
2121 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2124 msecs_to_jiffies(params->spp_hbinterval);
2127 msecs_to_jiffies(params->spp_hbinterval);
2129 sp->hbinterval = params->spp_hbinterval;
2136 trans->param_flags =
2137 (trans->param_flags & ~SPP_HB) | hb_change;
2140 (asoc->param_flags & ~SPP_HB) | hb_change;
2143 (sp->param_flags & ~SPP_HB) | hb_change;
2147 /* When Path MTU discovery is disabled the value specified here will
2148 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2149 * include the flag SPP_PMTUD_DISABLE for this field to have any
2152 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2154 trans->pathmtu = params->spp_pathmtu;
2155 sctp_assoc_sync_pmtu(asoc);
2157 asoc->pathmtu = params->spp_pathmtu;
2158 sctp_frag_point(sp, params->spp_pathmtu);
2160 sp->pathmtu = params->spp_pathmtu;
2166 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2167 (params->spp_flags & SPP_PMTUD_ENABLE);
2168 trans->param_flags =
2169 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2171 sctp_transport_pmtu(trans);
2172 sctp_assoc_sync_pmtu(asoc);
2176 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2179 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2183 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2184 * value of this field is ignored. Note also that a value of zero
2185 * indicates the current setting should be left unchanged.
2187 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2190 msecs_to_jiffies(params->spp_sackdelay);
2193 msecs_to_jiffies(params->spp_sackdelay);
2195 sp->sackdelay = params->spp_sackdelay;
2199 if (sackdelay_change) {
2201 trans->param_flags =
2202 (trans->param_flags & ~SPP_SACKDELAY) |
2206 (asoc->param_flags & ~SPP_SACKDELAY) |
2210 (sp->param_flags & ~SPP_SACKDELAY) |
2215 /* Note that unless the spp_flag is set to SPP_PMTUD_ENABLE the value
2216 * of this field is ignored. Note also that a value of zero
2217 * indicates the current setting should be left unchanged.
2219 if ((params->spp_flags & SPP_PMTUD_ENABLE) && params->spp_pathmaxrxt) {
2221 trans->pathmaxrxt = params->spp_pathmaxrxt;
2223 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2225 sp->pathmaxrxt = params->spp_pathmaxrxt;
2232 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2233 char __user *optval, int optlen)
2235 struct sctp_paddrparams params;
2236 struct sctp_transport *trans = NULL;
2237 struct sctp_association *asoc = NULL;
2238 struct sctp_sock *sp = sctp_sk(sk);
2240 int hb_change, pmtud_change, sackdelay_change;
2242 if (optlen != sizeof(struct sctp_paddrparams))
2245 if (copy_from_user(¶ms, optval, optlen))
2248 /* Validate flags and value parameters. */
2249 hb_change = params.spp_flags & SPP_HB;
2250 pmtud_change = params.spp_flags & SPP_PMTUD;
2251 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2253 if (hb_change == SPP_HB ||
2254 pmtud_change == SPP_PMTUD ||
2255 sackdelay_change == SPP_SACKDELAY ||
2256 params.spp_sackdelay > 500 ||
2258 && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2261 /* If an address other than INADDR_ANY is specified, and
2262 * no transport is found, then the request is invalid.
2264 if (!sctp_is_any(( union sctp_addr *)¶ms.spp_address)) {
2265 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2266 params.spp_assoc_id);
2271 /* Get association, if assoc_id != 0 and the socket is a one
2272 * to many style socket, and an association was not found, then
2273 * the id was invalid.
2275 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2276 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2279 /* Heartbeat demand can only be sent on a transport or
2280 * association, but not a socket.
2282 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2285 /* Process parameters. */
2286 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2287 hb_change, pmtud_change,
2293 /* If changes are for association, also apply parameters to each
2296 if (!trans && asoc) {
2297 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2299 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2300 hb_change, pmtud_change,
2309 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2311 * This option will effect the way delayed acks are performed. This
2312 * option allows you to get or set the delayed ack time, in
2313 * milliseconds. It also allows changing the delayed ack frequency.
2314 * Changing the frequency to 1 disables the delayed sack algorithm. If
2315 * the assoc_id is 0, then this sets or gets the endpoints default
2316 * values. If the assoc_id field is non-zero, then the set or get
2317 * effects the specified association for the one to many model (the
2318 * assoc_id field is ignored by the one to one model). Note that if
2319 * sack_delay or sack_freq are 0 when setting this option, then the
2320 * current values will remain unchanged.
2322 * struct sctp_sack_info {
2323 * sctp_assoc_t sack_assoc_id;
2324 * uint32_t sack_delay;
2325 * uint32_t sack_freq;
2328 * sack_assoc_id - This parameter, indicates which association the user
2329 * is performing an action upon. Note that if this field's value is
2330 * zero then the endpoints default value is changed (effecting future
2331 * associations only).
2333 * sack_delay - This parameter contains the number of milliseconds that
2334 * the user is requesting the delayed ACK timer be set to. Note that
2335 * this value is defined in the standard to be between 200 and 500
2338 * sack_freq - This parameter contains the number of packets that must
2339 * be received before a sack is sent without waiting for the delay
2340 * timer to expire. The default value for this is 2, setting this
2341 * value to 1 will disable the delayed sack algorithm.
2344 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2345 char __user *optval, int optlen)
2347 struct sctp_sack_info params;
2348 struct sctp_transport *trans = NULL;
2349 struct sctp_association *asoc = NULL;
2350 struct sctp_sock *sp = sctp_sk(sk);
2352 if (optlen == sizeof(struct sctp_sack_info)) {
2353 if (copy_from_user(¶ms, optval, optlen))
2356 if (params.sack_delay == 0 && params.sack_freq == 0)
2358 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2359 printk(KERN_WARNING "SCTP: Use of struct sctp_sack_info "
2360 "in delayed_ack socket option deprecated\n");
2361 printk(KERN_WARNING "SCTP: struct sctp_sack_info instead\n");
2362 if (copy_from_user(¶ms, optval, optlen))
2365 if (params.sack_delay == 0)
2366 params.sack_freq = 1;
2368 params.sack_freq = 0;
2372 /* Validate value parameter. */
2373 if (params.sack_delay > 500)
2376 /* Get association, if sack_assoc_id != 0 and the socket is a one
2377 * to many style socket, and an association was not found, then
2378 * the id was invalid.
2380 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2381 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2384 if (params.sack_delay) {
2387 msecs_to_jiffies(params.sack_delay);
2389 (asoc->param_flags & ~SPP_SACKDELAY) |
2390 SPP_SACKDELAY_ENABLE;
2392 sp->sackdelay = params.sack_delay;
2394 (sp->param_flags & ~SPP_SACKDELAY) |
2395 SPP_SACKDELAY_ENABLE;
2399 if (params.sack_freq == 1) {
2402 (asoc->param_flags & ~SPP_SACKDELAY) |
2403 SPP_SACKDELAY_DISABLE;
2406 (sp->param_flags & ~SPP_SACKDELAY) |
2407 SPP_SACKDELAY_DISABLE;
2409 } else if (params.sack_freq > 1) {
2411 asoc->sackfreq = params.sack_freq;
2413 (asoc->param_flags & ~SPP_SACKDELAY) |
2414 SPP_SACKDELAY_ENABLE;
2416 sp->sackfreq = params.sack_freq;
2418 (sp->param_flags & ~SPP_SACKDELAY) |
2419 SPP_SACKDELAY_ENABLE;
2423 /* If change is for association, also apply to each transport. */
2425 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2427 if (params.sack_delay) {
2429 msecs_to_jiffies(params.sack_delay);
2430 trans->param_flags =
2431 (trans->param_flags & ~SPP_SACKDELAY) |
2432 SPP_SACKDELAY_ENABLE;
2434 if (params.sack_delay == 1) {
2435 trans->param_flags =
2436 (trans->param_flags & ~SPP_SACKDELAY) |
2437 SPP_SACKDELAY_DISABLE;
2438 } else if (params.sack_freq > 1) {
2439 trans->sackfreq = params.sack_freq;
2440 trans->param_flags =
2441 (trans->param_flags & ~SPP_SACKDELAY) |
2442 SPP_SACKDELAY_ENABLE;
2450 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2452 * Applications can specify protocol parameters for the default association
2453 * initialization. The option name argument to setsockopt() and getsockopt()
2456 * Setting initialization parameters is effective only on an unconnected
2457 * socket (for UDP-style sockets only future associations are effected
2458 * by the change). With TCP-style sockets, this option is inherited by
2459 * sockets derived from a listener socket.
2461 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen)
2463 struct sctp_initmsg sinit;
2464 struct sctp_sock *sp = sctp_sk(sk);
2466 if (optlen != sizeof(struct sctp_initmsg))
2468 if (copy_from_user(&sinit, optval, optlen))
2471 if (sinit.sinit_num_ostreams)
2472 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2473 if (sinit.sinit_max_instreams)
2474 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2475 if (sinit.sinit_max_attempts)
2476 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2477 if (sinit.sinit_max_init_timeo)
2478 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2484 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2486 * Applications that wish to use the sendto() system call may wish to
2487 * specify a default set of parameters that would normally be supplied
2488 * through the inclusion of ancillary data. This socket option allows
2489 * such an application to set the default sctp_sndrcvinfo structure.
2490 * The application that wishes to use this socket option simply passes
2491 * in to this call the sctp_sndrcvinfo structure defined in Section
2492 * 5.2.2) The input parameters accepted by this call include
2493 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2494 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2495 * to this call if the caller is using the UDP model.
2497 static int sctp_setsockopt_default_send_param(struct sock *sk,
2498 char __user *optval, int optlen)
2500 struct sctp_sndrcvinfo info;
2501 struct sctp_association *asoc;
2502 struct sctp_sock *sp = sctp_sk(sk);
2504 if (optlen != sizeof(struct sctp_sndrcvinfo))
2506 if (copy_from_user(&info, optval, optlen))
2509 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2510 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2514 asoc->default_stream = info.sinfo_stream;
2515 asoc->default_flags = info.sinfo_flags;
2516 asoc->default_ppid = info.sinfo_ppid;
2517 asoc->default_context = info.sinfo_context;
2518 asoc->default_timetolive = info.sinfo_timetolive;
2520 sp->default_stream = info.sinfo_stream;
2521 sp->default_flags = info.sinfo_flags;
2522 sp->default_ppid = info.sinfo_ppid;
2523 sp->default_context = info.sinfo_context;
2524 sp->default_timetolive = info.sinfo_timetolive;
2530 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2532 * Requests that the local SCTP stack use the enclosed peer address as
2533 * the association primary. The enclosed address must be one of the
2534 * association peer's addresses.
2536 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2539 struct sctp_prim prim;
2540 struct sctp_transport *trans;
2542 if (optlen != sizeof(struct sctp_prim))
2545 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2548 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2552 sctp_assoc_set_primary(trans->asoc, trans);
2558 * 7.1.5 SCTP_NODELAY
2560 * Turn on/off any Nagle-like algorithm. This means that packets are
2561 * generally sent as soon as possible and no unnecessary delays are
2562 * introduced, at the cost of more packets in the network. Expects an
2563 * integer boolean flag.
2565 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2570 if (optlen < sizeof(int))
2572 if (get_user(val, (int __user *)optval))
2575 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2581 * 7.1.1 SCTP_RTOINFO
2583 * The protocol parameters used to initialize and bound retransmission
2584 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2585 * and modify these parameters.
2586 * All parameters are time values, in milliseconds. A value of 0, when
2587 * modifying the parameters, indicates that the current value should not
2591 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) {
2592 struct sctp_rtoinfo rtoinfo;
2593 struct sctp_association *asoc;
2595 if (optlen != sizeof (struct sctp_rtoinfo))
2598 if (copy_from_user(&rtoinfo, optval, optlen))
2601 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2603 /* Set the values to the specific association */
2604 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2608 if (rtoinfo.srto_initial != 0)
2610 msecs_to_jiffies(rtoinfo.srto_initial);
2611 if (rtoinfo.srto_max != 0)
2612 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2613 if (rtoinfo.srto_min != 0)
2614 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2616 /* If there is no association or the association-id = 0
2617 * set the values to the endpoint.
2619 struct sctp_sock *sp = sctp_sk(sk);
2621 if (rtoinfo.srto_initial != 0)
2622 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2623 if (rtoinfo.srto_max != 0)
2624 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2625 if (rtoinfo.srto_min != 0)
2626 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2634 * 7.1.2 SCTP_ASSOCINFO
2636 * This option is used to tune the maximum retransmission attempts
2637 * of the association.
2638 * Returns an error if the new association retransmission value is
2639 * greater than the sum of the retransmission value of the peer.
2640 * See [SCTP] for more information.
2643 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen)
2646 struct sctp_assocparams assocparams;
2647 struct sctp_association *asoc;
2649 if (optlen != sizeof(struct sctp_assocparams))
2651 if (copy_from_user(&assocparams, optval, optlen))
2654 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2656 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2659 /* Set the values to the specific association */
2661 if (assocparams.sasoc_asocmaxrxt != 0) {
2664 struct sctp_transport *peer_addr;
2666 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2668 path_sum += peer_addr->pathmaxrxt;
2672 /* Only validate asocmaxrxt if we have more then
2673 * one path/transport. We do this because path
2674 * retransmissions are only counted when we have more
2678 assocparams.sasoc_asocmaxrxt > path_sum)
2681 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2684 if (assocparams.sasoc_cookie_life != 0) {
2685 asoc->cookie_life.tv_sec =
2686 assocparams.sasoc_cookie_life / 1000;
2687 asoc->cookie_life.tv_usec =
2688 (assocparams.sasoc_cookie_life % 1000)
2692 /* Set the values to the endpoint */
2693 struct sctp_sock *sp = sctp_sk(sk);
2695 if (assocparams.sasoc_asocmaxrxt != 0)
2696 sp->assocparams.sasoc_asocmaxrxt =
2697 assocparams.sasoc_asocmaxrxt;
2698 if (assocparams.sasoc_cookie_life != 0)
2699 sp->assocparams.sasoc_cookie_life =
2700 assocparams.sasoc_cookie_life;
2706 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2708 * This socket option is a boolean flag which turns on or off mapped V4
2709 * addresses. If this option is turned on and the socket is type
2710 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2711 * If this option is turned off, then no mapping will be done of V4
2712 * addresses and a user will receive both PF_INET6 and PF_INET type
2713 * addresses on the socket.
2715 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen)
2718 struct sctp_sock *sp = sctp_sk(sk);
2720 if (optlen < sizeof(int))
2722 if (get_user(val, (int __user *)optval))
2733 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
2735 * This socket option specifies the maximum size to put in any outgoing
2736 * SCTP chunk. If a message is larger than this size it will be
2737 * fragmented by SCTP into the specified size. Note that the underlying
2738 * SCTP implementation may fragment into smaller sized chunks when the
2739 * PMTU of the underlying association is smaller than the value set by
2742 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen)
2744 struct sctp_association *asoc;
2745 struct sctp_sock *sp = sctp_sk(sk);
2748 if (optlen < sizeof(int))
2750 if (get_user(val, (int __user *)optval))
2752 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2754 sp->user_frag = val;
2756 /* Update the frag_point of the existing associations. */
2757 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
2758 asoc->frag_point = sctp_frag_point(sp, asoc->pathmtu);
2766 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2768 * Requests that the peer mark the enclosed address as the association
2769 * primary. The enclosed address must be one of the association's
2770 * locally bound addresses. The following structure is used to make a
2771 * set primary request:
2773 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2776 struct sctp_sock *sp;
2777 struct sctp_endpoint *ep;
2778 struct sctp_association *asoc = NULL;
2779 struct sctp_setpeerprim prim;
2780 struct sctp_chunk *chunk;
2786 if (!sctp_addip_enable)
2789 if (optlen != sizeof(struct sctp_setpeerprim))
2792 if (copy_from_user(&prim, optval, optlen))
2795 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2799 if (!asoc->peer.asconf_capable)
2802 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2805 if (!sctp_state(asoc, ESTABLISHED))
2808 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2809 return -EADDRNOTAVAIL;
2811 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2812 chunk = sctp_make_asconf_set_prim(asoc,
2813 (union sctp_addr *)&prim.sspp_addr);
2817 err = sctp_send_asconf(asoc, chunk);
2819 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2824 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2827 struct sctp_setadaptation adaptation;
2829 if (optlen != sizeof(struct sctp_setadaptation))
2831 if (copy_from_user(&adaptation, optval, optlen))
2834 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2840 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
2842 * The context field in the sctp_sndrcvinfo structure is normally only
2843 * used when a failed message is retrieved holding the value that was
2844 * sent down on the actual send call. This option allows the setting of
2845 * a default context on an association basis that will be received on
2846 * reading messages from the peer. This is especially helpful in the
2847 * one-2-many model for an application to keep some reference to an
2848 * internal state machine that is processing messages on the
2849 * association. Note that the setting of this value only effects
2850 * received messages from the peer and does not effect the value that is
2851 * saved with outbound messages.
2853 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
2856 struct sctp_assoc_value params;
2857 struct sctp_sock *sp;
2858 struct sctp_association *asoc;
2860 if (optlen != sizeof(struct sctp_assoc_value))
2862 if (copy_from_user(¶ms, optval, optlen))
2867 if (params.assoc_id != 0) {
2868 asoc = sctp_id2assoc(sk, params.assoc_id);
2871 asoc->default_rcv_context = params.assoc_value;
2873 sp->default_rcv_context = params.assoc_value;
2880 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
2882 * This options will at a minimum specify if the implementation is doing
2883 * fragmented interleave. Fragmented interleave, for a one to many
2884 * socket, is when subsequent calls to receive a message may return
2885 * parts of messages from different associations. Some implementations
2886 * may allow you to turn this value on or off. If so, when turned off,
2887 * no fragment interleave will occur (which will cause a head of line
2888 * blocking amongst multiple associations sharing the same one to many
2889 * socket). When this option is turned on, then each receive call may
2890 * come from a different association (thus the user must receive data
2891 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
2892 * association each receive belongs to.
2894 * This option takes a boolean value. A non-zero value indicates that
2895 * fragmented interleave is on. A value of zero indicates that
2896 * fragmented interleave is off.
2898 * Note that it is important that an implementation that allows this
2899 * option to be turned on, have it off by default. Otherwise an unaware
2900 * application using the one to many model may become confused and act
2903 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
2904 char __user *optval,
2909 if (optlen != sizeof(int))
2911 if (get_user(val, (int __user *)optval))
2914 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
2920 * 7.1.25. Set or Get the sctp partial delivery point
2921 * (SCTP_PARTIAL_DELIVERY_POINT)
2922 * This option will set or get the SCTP partial delivery point. This
2923 * point is the size of a message where the partial delivery API will be
2924 * invoked to help free up rwnd space for the peer. Setting this to a
2925 * lower value will cause partial delivery's to happen more often. The
2926 * calls argument is an integer that sets or gets the partial delivery
2929 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
2930 char __user *optval,
2935 if (optlen != sizeof(u32))
2937 if (get_user(val, (int __user *)optval))
2940 sctp_sk(sk)->pd_point = val;
2942 return 0; /* is this the right error code? */
2946 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
2948 * This option will allow a user to change the maximum burst of packets
2949 * that can be emitted by this association. Note that the default value
2950 * is 4, and some implementations may restrict this setting so that it
2951 * can only be lowered.
2953 * NOTE: This text doesn't seem right. Do this on a socket basis with
2954 * future associations inheriting the socket value.
2956 static int sctp_setsockopt_maxburst(struct sock *sk,
2957 char __user *optval,
2960 struct sctp_assoc_value params;
2961 struct sctp_sock *sp;
2962 struct sctp_association *asoc;
2966 if (optlen < sizeof(int))
2969 if (optlen == sizeof(int)) {
2971 "SCTP: Use of int in max_burst socket option deprecated\n");
2973 "SCTP: Use struct sctp_assoc_value instead\n");
2974 if (copy_from_user(&val, optval, optlen))
2976 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2977 if (copy_from_user(¶ms, optval, optlen))
2979 val = params.assoc_value;
2980 assoc_id = params.assoc_id;
2986 if (assoc_id != 0) {
2987 asoc = sctp_id2assoc(sk, assoc_id);
2990 asoc->max_burst = val;
2992 sp->max_burst = val;
2998 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3000 * This set option adds a chunk type that the user is requesting to be
3001 * received only in an authenticated way. Changes to the list of chunks
3002 * will only effect future associations on the socket.
3004 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3005 char __user *optval,
3008 struct sctp_authchunk val;
3010 if (optlen != sizeof(struct sctp_authchunk))
3012 if (copy_from_user(&val, optval, optlen))
3015 switch (val.sauth_chunk) {
3017 case SCTP_CID_INIT_ACK:
3018 case SCTP_CID_SHUTDOWN_COMPLETE:
3023 /* add this chunk id to the endpoint */
3024 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3028 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3030 * This option gets or sets the list of HMAC algorithms that the local
3031 * endpoint requires the peer to use.
3033 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3034 char __user *optval,
3037 struct sctp_hmacalgo *hmacs;
3040 if (optlen < sizeof(struct sctp_hmacalgo))
3043 hmacs = kmalloc(optlen, GFP_KERNEL);
3047 if (copy_from_user(hmacs, optval, optlen)) {
3052 if (hmacs->shmac_num_idents == 0 ||
3053 hmacs->shmac_num_idents > SCTP_AUTH_NUM_HMACS) {
3058 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3065 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3067 * This option will set a shared secret key which is used to build an
3068 * association shared key.
3070 static int sctp_setsockopt_auth_key(struct sock *sk,
3071 char __user *optval,
3074 struct sctp_authkey *authkey;
3075 struct sctp_association *asoc;
3078 if (optlen <= sizeof(struct sctp_authkey))
3081 authkey = kmalloc(optlen, GFP_KERNEL);
3085 if (copy_from_user(authkey, optval, optlen)) {
3090 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3091 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3096 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3103 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3105 * This option will get or set the active shared key to be used to build
3106 * the association shared key.
3108 static int sctp_setsockopt_active_key(struct sock *sk,
3109 char __user *optval,
3112 struct sctp_authkeyid val;
3113 struct sctp_association *asoc;
3115 if (optlen != sizeof(struct sctp_authkeyid))
3117 if (copy_from_user(&val, optval, optlen))
3120 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3121 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3124 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3125 val.scact_keynumber);
3129 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3131 * This set option will delete a shared secret key from use.
3133 static int sctp_setsockopt_del_key(struct sock *sk,
3134 char __user *optval,
3137 struct sctp_authkeyid val;
3138 struct sctp_association *asoc;
3140 if (optlen != sizeof(struct sctp_authkeyid))
3142 if (copy_from_user(&val, optval, optlen))
3145 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3146 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3149 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3150 val.scact_keynumber);
3155 /* API 6.2 setsockopt(), getsockopt()
3157 * Applications use setsockopt() and getsockopt() to set or retrieve
3158 * socket options. Socket options are used to change the default
3159 * behavior of sockets calls. They are described in Section 7.
3163 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3164 * int __user *optlen);
3165 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3168 * sd - the socket descript.
3169 * level - set to IPPROTO_SCTP for all SCTP options.
3170 * optname - the option name.
3171 * optval - the buffer to store the value of the option.
3172 * optlen - the size of the buffer.
3174 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3175 char __user *optval, int optlen)
3179 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3182 /* I can hardly begin to describe how wrong this is. This is
3183 * so broken as to be worse than useless. The API draft
3184 * REALLY is NOT helpful here... I am not convinced that the
3185 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3186 * are at all well-founded.
3188 if (level != SOL_SCTP) {
3189 struct sctp_af *af = sctp_sk(sk)->pf->af;
3190 retval = af->setsockopt(sk, level, optname, optval, optlen);
3197 case SCTP_SOCKOPT_BINDX_ADD:
3198 /* 'optlen' is the size of the addresses buffer. */
3199 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3200 optlen, SCTP_BINDX_ADD_ADDR);
3203 case SCTP_SOCKOPT_BINDX_REM:
3204 /* 'optlen' is the size of the addresses buffer. */
3205 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3206 optlen, SCTP_BINDX_REM_ADDR);
3209 case SCTP_SOCKOPT_CONNECTX:
3210 /* 'optlen' is the size of the addresses buffer. */
3211 retval = sctp_setsockopt_connectx(sk, (struct sockaddr __user *)optval,
3215 case SCTP_DISABLE_FRAGMENTS:
3216 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3220 retval = sctp_setsockopt_events(sk, optval, optlen);
3223 case SCTP_AUTOCLOSE:
3224 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3227 case SCTP_PEER_ADDR_PARAMS:
3228 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3231 case SCTP_DELAYED_ACK:
3232 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3234 case SCTP_PARTIAL_DELIVERY_POINT:
3235 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3239 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3241 case SCTP_DEFAULT_SEND_PARAM:
3242 retval = sctp_setsockopt_default_send_param(sk, optval,
3245 case SCTP_PRIMARY_ADDR:
3246 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3248 case SCTP_SET_PEER_PRIMARY_ADDR:
3249 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3252 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3255 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3257 case SCTP_ASSOCINFO:
3258 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3260 case SCTP_I_WANT_MAPPED_V4_ADDR:
3261 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3264 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3266 case SCTP_ADAPTATION_LAYER:
3267 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3270 retval = sctp_setsockopt_context(sk, optval, optlen);
3272 case SCTP_FRAGMENT_INTERLEAVE:
3273 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3275 case SCTP_MAX_BURST:
3276 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3278 case SCTP_AUTH_CHUNK:
3279 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3281 case SCTP_HMAC_IDENT:
3282 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3285 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3287 case SCTP_AUTH_ACTIVE_KEY:
3288 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3290 case SCTP_AUTH_DELETE_KEY:
3291 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3294 retval = -ENOPROTOOPT;
3298 sctp_release_sock(sk);
3304 /* API 3.1.6 connect() - UDP Style Syntax
3306 * An application may use the connect() call in the UDP model to initiate an
3307 * association without sending data.
3311 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3313 * sd: the socket descriptor to have a new association added to.
3315 * nam: the address structure (either struct sockaddr_in or struct
3316 * sockaddr_in6 defined in RFC2553 [7]).
3318 * len: the size of the address.
3320 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3328 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3329 __func__, sk, addr, addr_len);
3331 /* Validate addr_len before calling common connect/connectx routine. */
3332 af = sctp_get_af_specific(addr->sa_family);
3333 if (!af || addr_len < af->sockaddr_len) {
3336 /* Pass correct addr len to common routine (so it knows there
3337 * is only one address being passed.
3339 err = __sctp_connect(sk, addr, af->sockaddr_len);
3342 sctp_release_sock(sk);
3346 /* FIXME: Write comments. */
3347 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3349 return -EOPNOTSUPP; /* STUB */
3352 /* 4.1.4 accept() - TCP Style Syntax
3354 * Applications use accept() call to remove an established SCTP
3355 * association from the accept queue of the endpoint. A new socket
3356 * descriptor will be returned from accept() to represent the newly
3357 * formed association.
3359 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3361 struct sctp_sock *sp;
3362 struct sctp_endpoint *ep;
3363 struct sock *newsk = NULL;
3364 struct sctp_association *asoc;
3373 if (!sctp_style(sk, TCP)) {
3374 error = -EOPNOTSUPP;
3378 if (!sctp_sstate(sk, LISTENING)) {
3383 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3385 error = sctp_wait_for_accept(sk, timeo);
3389 /* We treat the list of associations on the endpoint as the accept
3390 * queue and pick the first association on the list.
3392 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3394 newsk = sp->pf->create_accept_sk(sk, asoc);
3400 /* Populate the fields of the newsk from the oldsk and migrate the
3401 * asoc to the newsk.
3403 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3406 sctp_release_sock(sk);
3411 /* The SCTP ioctl handler. */
3412 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3414 return -ENOIOCTLCMD;
3417 /* This is the function which gets called during socket creation to
3418 * initialized the SCTP-specific portion of the sock.
3419 * The sock structure should already be zero-filled memory.
3421 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3423 struct sctp_endpoint *ep;
3424 struct sctp_sock *sp;
3426 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3430 /* Initialize the SCTP per socket area. */
3431 switch (sk->sk_type) {
3432 case SOCK_SEQPACKET:
3433 sp->type = SCTP_SOCKET_UDP;
3436 sp->type = SCTP_SOCKET_TCP;
3439 return -ESOCKTNOSUPPORT;
3442 /* Initialize default send parameters. These parameters can be
3443 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3445 sp->default_stream = 0;
3446 sp->default_ppid = 0;
3447 sp->default_flags = 0;
3448 sp->default_context = 0;
3449 sp->default_timetolive = 0;
3451 sp->default_rcv_context = 0;
3452 sp->max_burst = sctp_max_burst;
3454 /* Initialize default setup parameters. These parameters
3455 * can be modified with the SCTP_INITMSG socket option or
3456 * overridden by the SCTP_INIT CMSG.
3458 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3459 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3460 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3461 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3463 /* Initialize default RTO related parameters. These parameters can
3464 * be modified for with the SCTP_RTOINFO socket option.
3466 sp->rtoinfo.srto_initial = sctp_rto_initial;
3467 sp->rtoinfo.srto_max = sctp_rto_max;
3468 sp->rtoinfo.srto_min = sctp_rto_min;
3470 /* Initialize default association related parameters. These parameters
3471 * can be modified with the SCTP_ASSOCINFO socket option.
3473 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3474 sp->assocparams.sasoc_number_peer_destinations = 0;
3475 sp->assocparams.sasoc_peer_rwnd = 0;
3476 sp->assocparams.sasoc_local_rwnd = 0;
3477 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3479 /* Initialize default event subscriptions. By default, all the
3482 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3484 /* Default Peer Address Parameters. These defaults can
3485 * be modified via SCTP_PEER_ADDR_PARAMS
3487 sp->hbinterval = sctp_hb_interval;
3488 sp->pathmaxrxt = sctp_max_retrans_path;
3489 sp->pathmtu = 0; // allow default discovery
3490 sp->sackdelay = sctp_sack_timeout;
3492 sp->param_flags = SPP_HB_ENABLE |
3494 SPP_SACKDELAY_ENABLE;
3496 /* If enabled no SCTP message fragmentation will be performed.
3497 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3499 sp->disable_fragments = 0;
3501 /* Enable Nagle algorithm by default. */
3504 /* Enable by default. */
3507 /* Auto-close idle associations after the configured
3508 * number of seconds. A value of 0 disables this
3509 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3510 * for UDP-style sockets only.
3514 /* User specified fragmentation limit. */
3517 sp->adaptation_ind = 0;
3519 sp->pf = sctp_get_pf_specific(sk->sk_family);
3521 /* Control variables for partial data delivery. */
3522 atomic_set(&sp->pd_mode, 0);
3523 skb_queue_head_init(&sp->pd_lobby);
3524 sp->frag_interleave = 0;
3526 /* Create a per socket endpoint structure. Even if we
3527 * change the data structure relationships, this may still
3528 * be useful for storing pre-connect address information.
3530 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3537 SCTP_DBG_OBJCNT_INC(sock);
3538 atomic_inc(&sctp_sockets_allocated);
3542 /* Cleanup any SCTP per socket resources. */
3543 SCTP_STATIC int sctp_destroy_sock(struct sock *sk)
3545 struct sctp_endpoint *ep;
3547 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3549 /* Release our hold on the endpoint. */
3550 ep = sctp_sk(sk)->ep;
3551 sctp_endpoint_free(ep);
3552 atomic_dec(&sctp_sockets_allocated);
3556 /* API 4.1.7 shutdown() - TCP Style Syntax
3557 * int shutdown(int socket, int how);
3559 * sd - the socket descriptor of the association to be closed.
3560 * how - Specifies the type of shutdown. The values are
3563 * Disables further receive operations. No SCTP
3564 * protocol action is taken.
3566 * Disables further send operations, and initiates
3567 * the SCTP shutdown sequence.
3569 * Disables further send and receive operations
3570 * and initiates the SCTP shutdown sequence.
3572 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3574 struct sctp_endpoint *ep;
3575 struct sctp_association *asoc;
3577 if (!sctp_style(sk, TCP))
3580 if (how & SEND_SHUTDOWN) {
3581 ep = sctp_sk(sk)->ep;
3582 if (!list_empty(&ep->asocs)) {
3583 asoc = list_entry(ep->asocs.next,
3584 struct sctp_association, asocs);
3585 sctp_primitive_SHUTDOWN(asoc, NULL);
3590 /* 7.2.1 Association Status (SCTP_STATUS)
3592 * Applications can retrieve current status information about an
3593 * association, including association state, peer receiver window size,
3594 * number of unacked data chunks, and number of data chunks pending
3595 * receipt. This information is read-only.
3597 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3598 char __user *optval,
3601 struct sctp_status status;
3602 struct sctp_association *asoc = NULL;
3603 struct sctp_transport *transport;
3604 sctp_assoc_t associd;
3607 if (len < sizeof(status)) {
3612 len = sizeof(status);
3613 if (copy_from_user(&status, optval, len)) {
3618 associd = status.sstat_assoc_id;
3619 asoc = sctp_id2assoc(sk, associd);
3625 transport = asoc->peer.primary_path;
3627 status.sstat_assoc_id = sctp_assoc2id(asoc);
3628 status.sstat_state = asoc->state;
3629 status.sstat_rwnd = asoc->peer.rwnd;
3630 status.sstat_unackdata = asoc->unack_data;
3632 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3633 status.sstat_instrms = asoc->c.sinit_max_instreams;
3634 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3635 status.sstat_fragmentation_point = asoc->frag_point;
3636 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3637 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3638 transport->af_specific->sockaddr_len);
3639 /* Map ipv4 address into v4-mapped-on-v6 address. */
3640 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3641 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3642 status.sstat_primary.spinfo_state = transport->state;
3643 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3644 status.sstat_primary.spinfo_srtt = transport->srtt;
3645 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3646 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3648 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3649 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3651 if (put_user(len, optlen)) {
3656 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3657 len, status.sstat_state, status.sstat_rwnd,
3658 status.sstat_assoc_id);
3660 if (copy_to_user(optval, &status, len)) {
3670 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3672 * Applications can retrieve information about a specific peer address
3673 * of an association, including its reachability state, congestion
3674 * window, and retransmission timer values. This information is
3677 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3678 char __user *optval,
3681 struct sctp_paddrinfo pinfo;
3682 struct sctp_transport *transport;
3685 if (len < sizeof(pinfo)) {
3690 len = sizeof(pinfo);
3691 if (copy_from_user(&pinfo, optval, len)) {
3696 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3697 pinfo.spinfo_assoc_id);
3701 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3702 pinfo.spinfo_state = transport->state;
3703 pinfo.spinfo_cwnd = transport->cwnd;
3704 pinfo.spinfo_srtt = transport->srtt;
3705 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3706 pinfo.spinfo_mtu = transport->pathmtu;
3708 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3709 pinfo.spinfo_state = SCTP_ACTIVE;
3711 if (put_user(len, optlen)) {
3716 if (copy_to_user(optval, &pinfo, len)) {
3725 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3727 * This option is a on/off flag. If enabled no SCTP message
3728 * fragmentation will be performed. Instead if a message being sent
3729 * exceeds the current PMTU size, the message will NOT be sent and
3730 * instead a error will be indicated to the user.
3732 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3733 char __user *optval, int __user *optlen)
3737 if (len < sizeof(int))
3741 val = (sctp_sk(sk)->disable_fragments == 1);
3742 if (put_user(len, optlen))
3744 if (copy_to_user(optval, &val, len))
3749 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3751 * This socket option is used to specify various notifications and
3752 * ancillary data the user wishes to receive.
3754 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3757 if (len < sizeof(struct sctp_event_subscribe))
3759 len = sizeof(struct sctp_event_subscribe);
3760 if (put_user(len, optlen))
3762 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3767 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3769 * This socket option is applicable to the UDP-style socket only. When
3770 * set it will cause associations that are idle for more than the
3771 * specified number of seconds to automatically close. An association
3772 * being idle is defined an association that has NOT sent or received
3773 * user data. The special value of '0' indicates that no automatic
3774 * close of any associations should be performed. The option expects an
3775 * integer defining the number of seconds of idle time before an
3776 * association is closed.
3778 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3780 /* Applicable to UDP-style socket only */
3781 if (sctp_style(sk, TCP))
3783 if (len < sizeof(int))
3786 if (put_user(len, optlen))
3788 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
3793 /* Helper routine to branch off an association to a new socket. */
3794 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3795 struct socket **sockp)
3797 struct sock *sk = asoc->base.sk;
3798 struct socket *sock;
3799 struct inet_sock *inetsk;
3803 /* An association cannot be branched off from an already peeled-off
3804 * socket, nor is this supported for tcp style sockets.
3806 if (!sctp_style(sk, UDP))
3809 /* Create a new socket. */
3810 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3814 /* Populate the fields of the newsk from the oldsk and migrate the
3815 * asoc to the newsk.
3817 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
3819 /* Make peeled-off sockets more like 1-1 accepted sockets.
3820 * Set the daddr and initialize id to something more random
3822 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
3823 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
3824 inetsk = inet_sk(sock->sk);
3825 inetsk->id = asoc->next_tsn ^ jiffies;
3832 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
3834 sctp_peeloff_arg_t peeloff;
3835 struct socket *newsock;
3837 struct sctp_association *asoc;
3839 if (len < sizeof(sctp_peeloff_arg_t))
3841 len = sizeof(sctp_peeloff_arg_t);
3842 if (copy_from_user(&peeloff, optval, len))
3845 asoc = sctp_id2assoc(sk, peeloff.associd);
3851 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
3853 retval = sctp_do_peeloff(asoc, &newsock);
3857 /* Map the socket to an unused fd that can be returned to the user. */
3858 retval = sock_map_fd(newsock);
3860 sock_release(newsock);
3864 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3865 __func__, sk, asoc, newsock->sk, retval);
3867 /* Return the fd mapped to the new socket. */
3868 peeloff.sd = retval;
3869 if (put_user(len, optlen))
3871 if (copy_to_user(optval, &peeloff, len))
3878 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
3880 * Applications can enable or disable heartbeats for any peer address of
3881 * an association, modify an address's heartbeat interval, force a
3882 * heartbeat to be sent immediately, and adjust the address's maximum
3883 * number of retransmissions sent before an address is considered
3884 * unreachable. The following structure is used to access and modify an
3885 * address's parameters:
3887 * struct sctp_paddrparams {
3888 * sctp_assoc_t spp_assoc_id;
3889 * struct sockaddr_storage spp_address;
3890 * uint32_t spp_hbinterval;
3891 * uint16_t spp_pathmaxrxt;
3892 * uint32_t spp_pathmtu;
3893 * uint32_t spp_sackdelay;
3894 * uint32_t spp_flags;
3897 * spp_assoc_id - (one-to-many style socket) This is filled in the
3898 * application, and identifies the association for
3900 * spp_address - This specifies which address is of interest.
3901 * spp_hbinterval - This contains the value of the heartbeat interval,
3902 * in milliseconds. If a value of zero
3903 * is present in this field then no changes are to
3904 * be made to this parameter.
3905 * spp_pathmaxrxt - This contains the maximum number of
3906 * retransmissions before this address shall be
3907 * considered unreachable. If a value of zero
3908 * is present in this field then no changes are to
3909 * be made to this parameter.
3910 * spp_pathmtu - When Path MTU discovery is disabled the value
3911 * specified here will be the "fixed" path mtu.
3912 * Note that if the spp_address field is empty
3913 * then all associations on this address will
3914 * have this fixed path mtu set upon them.
3916 * spp_sackdelay - When delayed sack is enabled, this value specifies
3917 * the number of milliseconds that sacks will be delayed
3918 * for. This value will apply to all addresses of an
3919 * association if the spp_address field is empty. Note
3920 * also, that if delayed sack is enabled and this
3921 * value is set to 0, no change is made to the last
3922 * recorded delayed sack timer value.
3924 * spp_flags - These flags are used to control various features
3925 * on an association. The flag field may contain
3926 * zero or more of the following options.
3928 * SPP_HB_ENABLE - Enable heartbeats on the
3929 * specified address. Note that if the address
3930 * field is empty all addresses for the association
3931 * have heartbeats enabled upon them.
3933 * SPP_HB_DISABLE - Disable heartbeats on the
3934 * speicifed address. Note that if the address
3935 * field is empty all addresses for the association
3936 * will have their heartbeats disabled. Note also
3937 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
3938 * mutually exclusive, only one of these two should
3939 * be specified. Enabling both fields will have
3940 * undetermined results.
3942 * SPP_HB_DEMAND - Request a user initiated heartbeat
3943 * to be made immediately.
3945 * SPP_PMTUD_ENABLE - This field will enable PMTU
3946 * discovery upon the specified address. Note that
3947 * if the address feild is empty then all addresses
3948 * on the association are effected.
3950 * SPP_PMTUD_DISABLE - This field will disable PMTU
3951 * discovery upon the specified address. Note that
3952 * if the address feild is empty then all addresses
3953 * on the association are effected. Not also that
3954 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
3955 * exclusive. Enabling both will have undetermined
3958 * SPP_SACKDELAY_ENABLE - Setting this flag turns
3959 * on delayed sack. The time specified in spp_sackdelay
3960 * is used to specify the sack delay for this address. Note
3961 * that if spp_address is empty then all addresses will
3962 * enable delayed sack and take on the sack delay
3963 * value specified in spp_sackdelay.
3964 * SPP_SACKDELAY_DISABLE - Setting this flag turns
3965 * off delayed sack. If the spp_address field is blank then
3966 * delayed sack is disabled for the entire association. Note
3967 * also that this field is mutually exclusive to
3968 * SPP_SACKDELAY_ENABLE, setting both will have undefined
3971 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
3972 char __user *optval, int __user *optlen)
3974 struct sctp_paddrparams params;
3975 struct sctp_transport *trans = NULL;
3976 struct sctp_association *asoc = NULL;
3977 struct sctp_sock *sp = sctp_sk(sk);
3979 if (len < sizeof(struct sctp_paddrparams))
3981 len = sizeof(struct sctp_paddrparams);
3982 if (copy_from_user(¶ms, optval, len))
3985 /* If an address other than INADDR_ANY is specified, and
3986 * no transport is found, then the request is invalid.
3988 if (!sctp_is_any(( union sctp_addr *)¶ms.spp_address)) {
3989 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
3990 params.spp_assoc_id);
3992 SCTP_DEBUG_PRINTK("Failed no transport\n");
3997 /* Get association, if assoc_id != 0 and the socket is a one
3998 * to many style socket, and an association was not found, then
3999 * the id was invalid.
4001 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4002 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4003 SCTP_DEBUG_PRINTK("Failed no association\n");
4008 /* Fetch transport values. */
4009 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4010 params.spp_pathmtu = trans->pathmtu;
4011 params.spp_pathmaxrxt = trans->pathmaxrxt;
4012 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4014 /*draft-11 doesn't say what to return in spp_flags*/
4015 params.spp_flags = trans->param_flags;
4017 /* Fetch association values. */
4018 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4019 params.spp_pathmtu = asoc->pathmtu;
4020 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4021 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4023 /*draft-11 doesn't say what to return in spp_flags*/
4024 params.spp_flags = asoc->param_flags;
4026 /* Fetch socket values. */
4027 params.spp_hbinterval = sp->hbinterval;
4028 params.spp_pathmtu = sp->pathmtu;
4029 params.spp_sackdelay = sp->sackdelay;
4030 params.spp_pathmaxrxt = sp->pathmaxrxt;
4032 /*draft-11 doesn't say what to return in spp_flags*/
4033 params.spp_flags = sp->param_flags;
4036 if (copy_to_user(optval, ¶ms, len))
4039 if (put_user(len, optlen))
4046 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4048 * This option will effect the way delayed acks are performed. This
4049 * option allows you to get or set the delayed ack time, in
4050 * milliseconds. It also allows changing the delayed ack frequency.
4051 * Changing the frequency to 1 disables the delayed sack algorithm. If
4052 * the assoc_id is 0, then this sets or gets the endpoints default
4053 * values. If the assoc_id field is non-zero, then the set or get
4054 * effects the specified association for the one to many model (the
4055 * assoc_id field is ignored by the one to one model). Note that if
4056 * sack_delay or sack_freq are 0 when setting this option, then the
4057 * current values will remain unchanged.
4059 * struct sctp_sack_info {
4060 * sctp_assoc_t sack_assoc_id;
4061 * uint32_t sack_delay;
4062 * uint32_t sack_freq;
4065 * sack_assoc_id - This parameter, indicates which association the user
4066 * is performing an action upon. Note that if this field's value is
4067 * zero then the endpoints default value is changed (effecting future
4068 * associations only).
4070 * sack_delay - This parameter contains the number of milliseconds that
4071 * the user is requesting the delayed ACK timer be set to. Note that
4072 * this value is defined in the standard to be between 200 and 500
4075 * sack_freq - This parameter contains the number of packets that must
4076 * be received before a sack is sent without waiting for the delay
4077 * timer to expire. The default value for this is 2, setting this
4078 * value to 1 will disable the delayed sack algorithm.
4080 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4081 char __user *optval,
4084 struct sctp_sack_info params;
4085 struct sctp_association *asoc = NULL;
4086 struct sctp_sock *sp = sctp_sk(sk);
4088 if (len >= sizeof(struct sctp_sack_info)) {
4089 len = sizeof(struct sctp_sack_info);
4091 if (copy_from_user(¶ms, optval, len))
4093 } else if (len == sizeof(struct sctp_assoc_value)) {
4094 printk(KERN_WARNING "SCTP: Use of struct sctp_sack_info "
4095 "in delayed_ack socket option deprecated\n");
4096 printk(KERN_WARNING "SCTP: struct sctp_sack_info instead\n");
4097 if (copy_from_user(¶ms, optval, len))
4102 /* Get association, if sack_assoc_id != 0 and the socket is a one
4103 * to many style socket, and an association was not found, then
4104 * the id was invalid.
4106 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4107 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4111 /* Fetch association values. */
4112 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4113 params.sack_delay = jiffies_to_msecs(
4115 params.sack_freq = asoc->sackfreq;
4118 params.sack_delay = 0;
4119 params.sack_freq = 1;
4122 /* Fetch socket values. */
4123 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4124 params.sack_delay = sp->sackdelay;
4125 params.sack_freq = sp->sackfreq;
4127 params.sack_delay = 0;
4128 params.sack_freq = 1;
4132 if (copy_to_user(optval, ¶ms, len))
4135 if (put_user(len, optlen))
4141 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4143 * Applications can specify protocol parameters for the default association
4144 * initialization. The option name argument to setsockopt() and getsockopt()
4147 * Setting initialization parameters is effective only on an unconnected
4148 * socket (for UDP-style sockets only future associations are effected
4149 * by the change). With TCP-style sockets, this option is inherited by
4150 * sockets derived from a listener socket.
4152 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4154 if (len < sizeof(struct sctp_initmsg))
4156 len = sizeof(struct sctp_initmsg);
4157 if (put_user(len, optlen))
4159 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4164 static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
4165 char __user *optval,
4169 struct sctp_association *asoc;
4170 struct list_head *pos;
4173 if (len < sizeof(sctp_assoc_t))
4176 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4179 /* For UDP-style sockets, id specifies the association to query. */
4180 asoc = sctp_id2assoc(sk, id);
4184 list_for_each(pos, &asoc->peer.transport_addr_list) {
4192 * Old API for getting list of peer addresses. Does not work for 32-bit
4193 * programs running on a 64-bit kernel
4195 static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
4196 char __user *optval,
4199 struct sctp_association *asoc;
4201 struct sctp_getaddrs_old getaddrs;
4202 struct sctp_transport *from;
4204 union sctp_addr temp;
4205 struct sctp_sock *sp = sctp_sk(sk);
4208 if (len < sizeof(struct sctp_getaddrs_old))
4211 len = sizeof(struct sctp_getaddrs_old);
4213 if (copy_from_user(&getaddrs, optval, len))
4216 if (getaddrs.addr_num <= 0) return -EINVAL;
4218 /* For UDP-style sockets, id specifies the association to query. */
4219 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4223 to = (void __user *)getaddrs.addrs;
4224 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4226 memcpy(&temp, &from->ipaddr, sizeof(temp));
4227 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4228 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4229 if (copy_to_user(to, &temp, addrlen))
4233 if (cnt >= getaddrs.addr_num) break;
4235 getaddrs.addr_num = cnt;
4236 if (put_user(len, optlen))
4238 if (copy_to_user(optval, &getaddrs, len))
4244 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4245 char __user *optval, int __user *optlen)
4247 struct sctp_association *asoc;
4249 struct sctp_getaddrs getaddrs;
4250 struct sctp_transport *from;
4252 union sctp_addr temp;
4253 struct sctp_sock *sp = sctp_sk(sk);
4258 if (len < sizeof(struct sctp_getaddrs))
4261 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4264 /* For UDP-style sockets, id specifies the association to query. */
4265 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4269 to = optval + offsetof(struct sctp_getaddrs,addrs);
4270 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4272 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4274 memcpy(&temp, &from->ipaddr, sizeof(temp));
4275 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4276 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4277 if (space_left < addrlen)
4279 if (copy_to_user(to, &temp, addrlen))
4283 space_left -= addrlen;
4286 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4288 bytes_copied = ((char __user *)to) - optval;
4289 if (put_user(bytes_copied, optlen))
4295 static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
4296 char __user *optval,
4300 struct sctp_bind_addr *bp;
4301 struct sctp_association *asoc;
4302 struct sctp_sockaddr_entry *addr;
4305 if (len < sizeof(sctp_assoc_t))
4308 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4312 * For UDP-style sockets, id specifies the association to query.
4313 * If the id field is set to the value '0' then the locally bound
4314 * addresses are returned without regard to any particular
4318 bp = &sctp_sk(sk)->ep->base.bind_addr;
4320 asoc = sctp_id2assoc(sk, id);
4323 bp = &asoc->base.bind_addr;
4326 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
4327 * addresses from the global local address list.
4329 if (sctp_list_single_entry(&bp->address_list)) {
4330 addr = list_entry(bp->address_list.next,
4331 struct sctp_sockaddr_entry, list);
4332 if (sctp_is_any(&addr->a)) {
4334 list_for_each_entry_rcu(addr,
4335 &sctp_local_addr_list, list) {
4339 if ((PF_INET == sk->sk_family) &&
4340 (AF_INET6 == addr->a.sa.sa_family))
4352 /* Protection on the bound address list is not needed,
4353 * since in the socket option context we hold the socket lock,
4354 * so there is no way that the bound address list can change.
4356 list_for_each_entry(addr, &bp->address_list, list) {
4363 /* Helper function that copies local addresses to user and returns the number
4364 * of addresses copied.
4366 static int sctp_copy_laddrs_old(struct sock *sk, __u16 port,
4367 int max_addrs, void *to,
4370 struct sctp_sockaddr_entry *addr;
4371 union sctp_addr temp;
4376 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4380 if ((PF_INET == sk->sk_family) &&
4381 (AF_INET6 == addr->a.sa.sa_family))
4383 memcpy(&temp, &addr->a, sizeof(temp));
4384 if (!temp.v4.sin_port)
4385 temp.v4.sin_port = htons(port);
4387 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4389 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4390 memcpy(to, &temp, addrlen);
4393 *bytes_copied += addrlen;
4395 if (cnt >= max_addrs) break;
4402 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4403 size_t space_left, int *bytes_copied)
4405 struct sctp_sockaddr_entry *addr;
4406 union sctp_addr temp;
4411 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4415 if ((PF_INET == sk->sk_family) &&
4416 (AF_INET6 == addr->a.sa.sa_family))
4418 memcpy(&temp, &addr->a, sizeof(temp));
4419 if (!temp.v4.sin_port)
4420 temp.v4.sin_port = htons(port);
4422 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4424 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4425 if (space_left < addrlen) {
4429 memcpy(to, &temp, addrlen);
4433 space_left -= addrlen;
4434 *bytes_copied += addrlen;
4441 /* Old API for getting list of local addresses. Does not work for 32-bit
4442 * programs running on a 64-bit kernel
4444 static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
4445 char __user *optval, int __user *optlen)
4447 struct sctp_bind_addr *bp;
4448 struct sctp_association *asoc;
4450 struct sctp_getaddrs_old getaddrs;
4451 struct sctp_sockaddr_entry *addr;
4453 union sctp_addr temp;
4454 struct sctp_sock *sp = sctp_sk(sk);
4459 int bytes_copied = 0;
4461 if (len < sizeof(struct sctp_getaddrs_old))
4464 len = sizeof(struct sctp_getaddrs_old);
4465 if (copy_from_user(&getaddrs, optval, len))
4468 if (getaddrs.addr_num <= 0) return -EINVAL;
4470 * For UDP-style sockets, id specifies the association to query.
4471 * If the id field is set to the value '0' then the locally bound
4472 * addresses are returned without regard to any particular
4475 if (0 == getaddrs.assoc_id) {
4476 bp = &sctp_sk(sk)->ep->base.bind_addr;
4478 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4481 bp = &asoc->base.bind_addr;
4484 to = getaddrs.addrs;
4486 /* Allocate space for a local instance of packed array to hold all
4487 * the data. We store addresses here first and then put write them
4488 * to the user in one shot.
4490 addrs = kmalloc(sizeof(union sctp_addr) * getaddrs.addr_num,
4495 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4496 * addresses from the global local address list.
4498 if (sctp_list_single_entry(&bp->address_list)) {
4499 addr = list_entry(bp->address_list.next,
4500 struct sctp_sockaddr_entry, list);
4501 if (sctp_is_any(&addr->a)) {
4502 cnt = sctp_copy_laddrs_old(sk, bp->port,
4504 addrs, &bytes_copied);
4510 /* Protection on the bound address list is not needed since
4511 * in the socket option context we hold a socket lock and
4512 * thus the bound address list can't change.
4514 list_for_each_entry(addr, &bp->address_list, list) {
4515 memcpy(&temp, &addr->a, sizeof(temp));
4516 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4517 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4518 memcpy(buf, &temp, addrlen);
4520 bytes_copied += addrlen;
4522 if (cnt >= getaddrs.addr_num) break;
4526 /* copy the entire address list into the user provided space */
4527 if (copy_to_user(to, addrs, bytes_copied)) {
4532 /* copy the leading structure back to user */
4533 getaddrs.addr_num = cnt;
4534 if (copy_to_user(optval, &getaddrs, len))
4542 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4543 char __user *optval, int __user *optlen)
4545 struct sctp_bind_addr *bp;
4546 struct sctp_association *asoc;
4548 struct sctp_getaddrs getaddrs;
4549 struct sctp_sockaddr_entry *addr;
4551 union sctp_addr temp;
4552 struct sctp_sock *sp = sctp_sk(sk);
4556 int bytes_copied = 0;
4560 if (len < sizeof(struct sctp_getaddrs))
4563 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4567 * For UDP-style sockets, id specifies the association to query.
4568 * If the id field is set to the value '0' then the locally bound
4569 * addresses are returned without regard to any particular
4572 if (0 == getaddrs.assoc_id) {
4573 bp = &sctp_sk(sk)->ep->base.bind_addr;
4575 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4578 bp = &asoc->base.bind_addr;
4581 to = optval + offsetof(struct sctp_getaddrs,addrs);
4582 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4584 addrs = kmalloc(space_left, GFP_KERNEL);
4588 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4589 * addresses from the global local address list.
4591 if (sctp_list_single_entry(&bp->address_list)) {
4592 addr = list_entry(bp->address_list.next,
4593 struct sctp_sockaddr_entry, list);
4594 if (sctp_is_any(&addr->a)) {
4595 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4596 space_left, &bytes_copied);
4606 /* Protection on the bound address list is not needed since
4607 * in the socket option context we hold a socket lock and
4608 * thus the bound address list can't change.
4610 list_for_each_entry(addr, &bp->address_list, list) {
4611 memcpy(&temp, &addr->a, sizeof(temp));
4612 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4613 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4614 if (space_left < addrlen) {
4615 err = -ENOMEM; /*fixme: right error?*/
4618 memcpy(buf, &temp, addrlen);
4620 bytes_copied += addrlen;
4622 space_left -= addrlen;
4626 if (copy_to_user(to, addrs, bytes_copied)) {
4630 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4634 if (put_user(bytes_copied, optlen))
4641 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4643 * Requests that the local SCTP stack use the enclosed peer address as
4644 * the association primary. The enclosed address must be one of the
4645 * association peer's addresses.
4647 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4648 char __user *optval, int __user *optlen)
4650 struct sctp_prim prim;
4651 struct sctp_association *asoc;
4652 struct sctp_sock *sp = sctp_sk(sk);
4654 if (len < sizeof(struct sctp_prim))
4657 len = sizeof(struct sctp_prim);
4659 if (copy_from_user(&prim, optval, len))
4662 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4666 if (!asoc->peer.primary_path)
4669 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4670 asoc->peer.primary_path->af_specific->sockaddr_len);
4672 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4673 (union sctp_addr *)&prim.ssp_addr);
4675 if (put_user(len, optlen))
4677 if (copy_to_user(optval, &prim, len))
4684 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4686 * Requests that the local endpoint set the specified Adaptation Layer
4687 * Indication parameter for all future INIT and INIT-ACK exchanges.
4689 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4690 char __user *optval, int __user *optlen)
4692 struct sctp_setadaptation adaptation;
4694 if (len < sizeof(struct sctp_setadaptation))
4697 len = sizeof(struct sctp_setadaptation);
4699 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4701 if (put_user(len, optlen))
4703 if (copy_to_user(optval, &adaptation, len))
4711 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4713 * Applications that wish to use the sendto() system call may wish to
4714 * specify a default set of parameters that would normally be supplied
4715 * through the inclusion of ancillary data. This socket option allows
4716 * such an application to set the default sctp_sndrcvinfo structure.
4719 * The application that wishes to use this socket option simply passes
4720 * in to this call the sctp_sndrcvinfo structure defined in Section
4721 * 5.2.2) The input parameters accepted by this call include
4722 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4723 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4724 * to this call if the caller is using the UDP model.
4726 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4728 static int sctp_getsockopt_default_send_param(struct sock *sk,
4729 int len, char __user *optval,
4732 struct sctp_sndrcvinfo info;
4733 struct sctp_association *asoc;
4734 struct sctp_sock *sp = sctp_sk(sk);
4736 if (len < sizeof(struct sctp_sndrcvinfo))
4739 len = sizeof(struct sctp_sndrcvinfo);
4741 if (copy_from_user(&info, optval, len))
4744 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4745 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4749 info.sinfo_stream = asoc->default_stream;
4750 info.sinfo_flags = asoc->default_flags;
4751 info.sinfo_ppid = asoc->default_ppid;
4752 info.sinfo_context = asoc->default_context;
4753 info.sinfo_timetolive = asoc->default_timetolive;
4755 info.sinfo_stream = sp->default_stream;
4756 info.sinfo_flags = sp->default_flags;
4757 info.sinfo_ppid = sp->default_ppid;
4758 info.sinfo_context = sp->default_context;
4759 info.sinfo_timetolive = sp->default_timetolive;
4762 if (put_user(len, optlen))
4764 if (copy_to_user(optval, &info, len))
4772 * 7.1.5 SCTP_NODELAY
4774 * Turn on/off any Nagle-like algorithm. This means that packets are
4775 * generally sent as soon as possible and no unnecessary delays are
4776 * introduced, at the cost of more packets in the network. Expects an
4777 * integer boolean flag.
4780 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4781 char __user *optval, int __user *optlen)
4785 if (len < sizeof(int))
4789 val = (sctp_sk(sk)->nodelay == 1);
4790 if (put_user(len, optlen))
4792 if (copy_to_user(optval, &val, len))
4799 * 7.1.1 SCTP_RTOINFO
4801 * The protocol parameters used to initialize and bound retransmission
4802 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4803 * and modify these parameters.
4804 * All parameters are time values, in milliseconds. A value of 0, when
4805 * modifying the parameters, indicates that the current value should not
4809 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4810 char __user *optval,
4811 int __user *optlen) {
4812 struct sctp_rtoinfo rtoinfo;
4813 struct sctp_association *asoc;
4815 if (len < sizeof (struct sctp_rtoinfo))
4818 len = sizeof(struct sctp_rtoinfo);
4820 if (copy_from_user(&rtoinfo, optval, len))
4823 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4825 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4828 /* Values corresponding to the specific association. */
4830 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4831 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4832 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4834 /* Values corresponding to the endpoint. */
4835 struct sctp_sock *sp = sctp_sk(sk);
4837 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4838 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4839 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4842 if (put_user(len, optlen))
4845 if (copy_to_user(optval, &rtoinfo, len))
4853 * 7.1.2 SCTP_ASSOCINFO
4855 * This option is used to tune the maximum retransmission attempts
4856 * of the association.
4857 * Returns an error if the new association retransmission value is
4858 * greater than the sum of the retransmission value of the peer.
4859 * See [SCTP] for more information.
4862 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4863 char __user *optval,
4867 struct sctp_assocparams assocparams;
4868 struct sctp_association *asoc;
4869 struct list_head *pos;
4872 if (len < sizeof (struct sctp_assocparams))
4875 len = sizeof(struct sctp_assocparams);
4877 if (copy_from_user(&assocparams, optval, len))
4880 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4882 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4885 /* Values correspoinding to the specific association */
4887 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4888 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4889 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4890 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4892 (asoc->cookie_life.tv_usec
4895 list_for_each(pos, &asoc->peer.transport_addr_list) {
4899 assocparams.sasoc_number_peer_destinations = cnt;
4901 /* Values corresponding to the endpoint */
4902 struct sctp_sock *sp = sctp_sk(sk);
4904 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4905 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4906 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4907 assocparams.sasoc_cookie_life =
4908 sp->assocparams.sasoc_cookie_life;
4909 assocparams.sasoc_number_peer_destinations =
4911 sasoc_number_peer_destinations;
4914 if (put_user(len, optlen))
4917 if (copy_to_user(optval, &assocparams, len))
4924 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4926 * This socket option is a boolean flag which turns on or off mapped V4
4927 * addresses. If this option is turned on and the socket is type
4928 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4929 * If this option is turned off, then no mapping will be done of V4
4930 * addresses and a user will receive both PF_INET6 and PF_INET type
4931 * addresses on the socket.
4933 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4934 char __user *optval, int __user *optlen)
4937 struct sctp_sock *sp = sctp_sk(sk);
4939 if (len < sizeof(int))
4944 if (put_user(len, optlen))
4946 if (copy_to_user(optval, &val, len))
4953 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
4954 * (chapter and verse is quoted at sctp_setsockopt_context())
4956 static int sctp_getsockopt_context(struct sock *sk, int len,
4957 char __user *optval, int __user *optlen)
4959 struct sctp_assoc_value params;
4960 struct sctp_sock *sp;
4961 struct sctp_association *asoc;
4963 if (len < sizeof(struct sctp_assoc_value))
4966 len = sizeof(struct sctp_assoc_value);
4968 if (copy_from_user(¶ms, optval, len))
4973 if (params.assoc_id != 0) {
4974 asoc = sctp_id2assoc(sk, params.assoc_id);
4977 params.assoc_value = asoc->default_rcv_context;
4979 params.assoc_value = sp->default_rcv_context;
4982 if (put_user(len, optlen))
4984 if (copy_to_user(optval, ¶ms, len))
4991 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
4993 * This socket option specifies the maximum size to put in any outgoing
4994 * SCTP chunk. If a message is larger than this size it will be
4995 * fragmented by SCTP into the specified size. Note that the underlying
4996 * SCTP implementation may fragment into smaller sized chunks when the
4997 * PMTU of the underlying association is smaller than the value set by
5000 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5001 char __user *optval, int __user *optlen)
5005 if (len < sizeof(int))
5010 val = sctp_sk(sk)->user_frag;
5011 if (put_user(len, optlen))
5013 if (copy_to_user(optval, &val, len))
5020 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5021 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5023 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5024 char __user *optval, int __user *optlen)
5028 if (len < sizeof(int))
5033 val = sctp_sk(sk)->frag_interleave;
5034 if (put_user(len, optlen))
5036 if (copy_to_user(optval, &val, len))
5043 * 7.1.25. Set or Get the sctp partial delivery point
5044 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5046 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5047 char __user *optval,
5052 if (len < sizeof(u32))
5057 val = sctp_sk(sk)->pd_point;
5058 if (put_user(len, optlen))
5060 if (copy_to_user(optval, &val, len))
5067 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5068 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5070 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5071 char __user *optval,
5074 struct sctp_assoc_value params;
5075 struct sctp_sock *sp;
5076 struct sctp_association *asoc;
5078 if (len < sizeof(int))
5081 if (len == sizeof(int)) {
5083 "SCTP: Use of int in max_burst socket option deprecated\n");
5085 "SCTP: Use struct sctp_assoc_value instead\n");
5086 params.assoc_id = 0;
5087 } else if (len == sizeof (struct sctp_assoc_value)) {
5088 if (copy_from_user(¶ms, optval, len))
5095 if (params.assoc_id != 0) {
5096 asoc = sctp_id2assoc(sk, params.assoc_id);
5099 params.assoc_value = asoc->max_burst;
5101 params.assoc_value = sp->max_burst;
5103 if (len == sizeof(int)) {
5104 if (copy_to_user(optval, ¶ms.assoc_value, len))
5107 if (copy_to_user(optval, ¶ms, len))
5115 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5116 char __user *optval, int __user *optlen)
5118 struct sctp_hmac_algo_param *hmacs;
5121 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5122 param_len = ntohs(hmacs->param_hdr.length);
5124 if (len < param_len)
5126 if (put_user(len, optlen))
5128 if (copy_to_user(optval, hmacs->hmac_ids, len))
5134 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5135 char __user *optval, int __user *optlen)
5137 struct sctp_authkeyid val;
5138 struct sctp_association *asoc;
5140 if (len < sizeof(struct sctp_authkeyid))
5142 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5145 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5146 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5150 val.scact_keynumber = asoc->active_key_id;
5152 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5157 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5158 char __user *optval, int __user *optlen)
5160 struct sctp_authchunks __user *p = (void __user *)optval;
5161 struct sctp_authchunks val;
5162 struct sctp_association *asoc;
5163 struct sctp_chunks_param *ch;
5167 if (len <= sizeof(struct sctp_authchunks))
5170 if (copy_from_user(&val, p, sizeof(struct sctp_authchunks)))
5173 to = p->gauth_chunks;
5174 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5178 ch = asoc->peer.peer_chunks;
5180 /* See if the user provided enough room for all the data */
5181 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5182 if (len < num_chunks)
5186 if (put_user(len, optlen))
5188 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5190 if (copy_to_user(to, ch->chunks, len))
5196 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5197 char __user *optval, int __user *optlen)
5199 struct sctp_authchunks __user *p = (void __user *)optval;
5200 struct sctp_authchunks val;
5201 struct sctp_association *asoc;
5202 struct sctp_chunks_param *ch;
5206 if (len <= sizeof(struct sctp_authchunks))
5209 if (copy_from_user(&val, p, sizeof(struct sctp_authchunks)))
5212 to = p->gauth_chunks;
5213 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5214 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5218 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5220 ch = sctp_sk(sk)->ep->auth_chunk_list;
5222 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5223 if (len < num_chunks)
5227 if (put_user(len, optlen))
5229 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5231 if (copy_to_user(to, ch->chunks, len))
5237 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5238 char __user *optval, int __user *optlen)
5243 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5246 /* I can hardly begin to describe how wrong this is. This is
5247 * so broken as to be worse than useless. The API draft
5248 * REALLY is NOT helpful here... I am not convinced that the
5249 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5250 * are at all well-founded.
5252 if (level != SOL_SCTP) {
5253 struct sctp_af *af = sctp_sk(sk)->pf->af;
5255 retval = af->getsockopt(sk, level, optname, optval, optlen);
5259 if (get_user(len, optlen))
5266 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5268 case SCTP_DISABLE_FRAGMENTS:
5269 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5273 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5275 case SCTP_AUTOCLOSE:
5276 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5278 case SCTP_SOCKOPT_PEELOFF:
5279 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5281 case SCTP_PEER_ADDR_PARAMS:
5282 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5285 case SCTP_DELAYED_ACK:
5286 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5290 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5292 case SCTP_GET_PEER_ADDRS_NUM_OLD:
5293 retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
5296 case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
5297 retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
5300 case SCTP_GET_PEER_ADDRS_OLD:
5301 retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
5304 case SCTP_GET_LOCAL_ADDRS_OLD:
5305 retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
5308 case SCTP_GET_PEER_ADDRS:
5309 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5312 case SCTP_GET_LOCAL_ADDRS:
5313 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5316 case SCTP_DEFAULT_SEND_PARAM:
5317 retval = sctp_getsockopt_default_send_param(sk, len,
5320 case SCTP_PRIMARY_ADDR:
5321 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5324 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5327 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5329 case SCTP_ASSOCINFO:
5330 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5332 case SCTP_I_WANT_MAPPED_V4_ADDR:
5333 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5336 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5338 case SCTP_GET_PEER_ADDR_INFO:
5339 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5342 case SCTP_ADAPTATION_LAYER:
5343 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5347 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5349 case SCTP_FRAGMENT_INTERLEAVE:
5350 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5353 case SCTP_PARTIAL_DELIVERY_POINT:
5354 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5357 case SCTP_MAX_BURST:
5358 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5361 case SCTP_AUTH_CHUNK:
5362 case SCTP_AUTH_DELETE_KEY:
5363 retval = -EOPNOTSUPP;
5365 case SCTP_HMAC_IDENT:
5366 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5368 case SCTP_AUTH_ACTIVE_KEY:
5369 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5371 case SCTP_PEER_AUTH_CHUNKS:
5372 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5375 case SCTP_LOCAL_AUTH_CHUNKS:
5376 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5380 retval = -ENOPROTOOPT;
5384 sctp_release_sock(sk);
5388 static void sctp_hash(struct sock *sk)
5393 static void sctp_unhash(struct sock *sk)
5398 /* Check if port is acceptable. Possibly find first available port.
5400 * The port hash table (contained in the 'global' SCTP protocol storage
5401 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5402 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5403 * list (the list number is the port number hashed out, so as you
5404 * would expect from a hash function, all the ports in a given list have
5405 * such a number that hashes out to the same list number; you were
5406 * expecting that, right?); so each list has a set of ports, with a
5407 * link to the socket (struct sock) that uses it, the port number and
5408 * a fastreuse flag (FIXME: NPI ipg).
5410 static struct sctp_bind_bucket *sctp_bucket_create(
5411 struct sctp_bind_hashbucket *head, unsigned short snum);
5413 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5415 struct sctp_bind_hashbucket *head; /* hash list */
5416 struct sctp_bind_bucket *pp; /* hash list port iterator */
5417 struct hlist_node *node;
5418 unsigned short snum;
5421 snum = ntohs(addr->v4.sin_port);
5423 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5424 sctp_local_bh_disable();
5427 /* Search for an available port. */
5428 int low, high, remaining, index;
5431 inet_get_local_port_range(&low, &high);
5432 remaining = (high - low) + 1;
5433 rover = net_random() % remaining + low;
5437 if ((rover < low) || (rover > high))
5439 index = sctp_phashfn(rover);
5440 head = &sctp_port_hashtable[index];
5441 sctp_spin_lock(&head->lock);
5442 sctp_for_each_hentry(pp, node, &head->chain)
5443 if (pp->port == rover)
5447 sctp_spin_unlock(&head->lock);
5448 } while (--remaining > 0);
5450 /* Exhausted local port range during search? */
5455 /* OK, here is the one we will use. HEAD (the port
5456 * hash table list entry) is non-NULL and we hold it's
5461 /* We are given an specific port number; we verify
5462 * that it is not being used. If it is used, we will
5463 * exahust the search in the hash list corresponding
5464 * to the port number (snum) - we detect that with the
5465 * port iterator, pp being NULL.
5467 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5468 sctp_spin_lock(&head->lock);
5469 sctp_for_each_hentry(pp, node, &head->chain) {
5470 if (pp->port == snum)
5477 if (!hlist_empty(&pp->owner)) {
5478 /* We had a port hash table hit - there is an
5479 * available port (pp != NULL) and it is being
5480 * used by other socket (pp->owner not empty); that other
5481 * socket is going to be sk2.
5483 int reuse = sk->sk_reuse;
5485 struct hlist_node *node;
5487 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5488 if (pp->fastreuse && sk->sk_reuse &&
5489 sk->sk_state != SCTP_SS_LISTENING)
5492 /* Run through the list of sockets bound to the port
5493 * (pp->port) [via the pointers bind_next and
5494 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5495 * we get the endpoint they describe and run through
5496 * the endpoint's list of IP (v4 or v6) addresses,
5497 * comparing each of the addresses with the address of
5498 * the socket sk. If we find a match, then that means
5499 * that this port/socket (sk) combination are already
5502 sk_for_each_bound(sk2, node, &pp->owner) {
5503 struct sctp_endpoint *ep2;
5504 ep2 = sctp_sk(sk2)->ep;
5506 if (reuse && sk2->sk_reuse &&
5507 sk2->sk_state != SCTP_SS_LISTENING)
5510 if (sctp_bind_addr_match(&ep2->base.bind_addr, addr,
5516 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5519 /* If there was a hash table miss, create a new port. */
5521 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5524 /* In either case (hit or miss), make sure fastreuse is 1 only
5525 * if sk->sk_reuse is too (that is, if the caller requested
5526 * SO_REUSEADDR on this socket -sk-).
5528 if (hlist_empty(&pp->owner)) {
5529 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5533 } else if (pp->fastreuse &&
5534 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5537 /* We are set, so fill up all the data in the hash table
5538 * entry, tie the socket list information with the rest of the
5539 * sockets FIXME: Blurry, NPI (ipg).
5542 if (!sctp_sk(sk)->bind_hash) {
5543 inet_sk(sk)->num = snum;
5544 sk_add_bind_node(sk, &pp->owner);
5545 sctp_sk(sk)->bind_hash = pp;
5550 sctp_spin_unlock(&head->lock);
5553 sctp_local_bh_enable();
5557 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5558 * port is requested.
5560 static int sctp_get_port(struct sock *sk, unsigned short snum)
5563 union sctp_addr addr;
5564 struct sctp_af *af = sctp_sk(sk)->pf->af;
5566 /* Set up a dummy address struct from the sk. */
5567 af->from_sk(&addr, sk);
5568 addr.v4.sin_port = htons(snum);
5570 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5571 ret = sctp_get_port_local(sk, &addr);
5573 return (ret ? 1 : 0);
5577 * 3.1.3 listen() - UDP Style Syntax
5579 * By default, new associations are not accepted for UDP style sockets.
5580 * An application uses listen() to mark a socket as being able to
5581 * accept new associations.
5583 SCTP_STATIC int sctp_seqpacket_listen(struct sock *sk, int backlog)
5585 struct sctp_sock *sp = sctp_sk(sk);
5586 struct sctp_endpoint *ep = sp->ep;
5588 /* Only UDP style sockets that are not peeled off are allowed to
5591 if (!sctp_style(sk, UDP))
5594 /* If backlog is zero, disable listening. */
5596 if (sctp_sstate(sk, CLOSED))
5599 sctp_unhash_endpoint(ep);
5600 sk->sk_state = SCTP_SS_CLOSED;
5604 /* Return if we are already listening. */
5605 if (sctp_sstate(sk, LISTENING))
5609 * If a bind() or sctp_bindx() is not called prior to a listen()
5610 * call that allows new associations to be accepted, the system
5611 * picks an ephemeral port and will choose an address set equivalent
5612 * to binding with a wildcard address.
5614 * This is not currently spelled out in the SCTP sockets
5615 * extensions draft, but follows the practice as seen in TCP
5618 * Additionally, turn off fastreuse flag since we are not listening
5620 sk->sk_state = SCTP_SS_LISTENING;
5621 if (!ep->base.bind_addr.port) {
5622 if (sctp_autobind(sk))
5625 sctp_sk(sk)->bind_hash->fastreuse = 0;
5627 sctp_hash_endpoint(ep);
5632 * 4.1.3 listen() - TCP Style Syntax
5634 * Applications uses listen() to ready the SCTP endpoint for accepting
5635 * inbound associations.
5637 SCTP_STATIC int sctp_stream_listen(struct sock *sk, int backlog)
5639 struct sctp_sock *sp = sctp_sk(sk);
5640 struct sctp_endpoint *ep = sp->ep;
5642 /* If backlog is zero, disable listening. */
5644 if (sctp_sstate(sk, CLOSED))
5647 sctp_unhash_endpoint(ep);
5648 sk->sk_state = SCTP_SS_CLOSED;
5652 if (sctp_sstate(sk, LISTENING))
5656 * If a bind() or sctp_bindx() is not called prior to a listen()
5657 * call that allows new associations to be accepted, the system
5658 * picks an ephemeral port and will choose an address set equivalent
5659 * to binding with a wildcard address.
5661 * This is not currently spelled out in the SCTP sockets
5662 * extensions draft, but follows the practice as seen in TCP
5665 sk->sk_state = SCTP_SS_LISTENING;
5666 if (!ep->base.bind_addr.port) {
5667 if (sctp_autobind(sk))
5670 sctp_sk(sk)->bind_hash->fastreuse = 0;
5672 sk->sk_max_ack_backlog = backlog;
5673 sctp_hash_endpoint(ep);
5678 * Move a socket to LISTENING state.
5680 int sctp_inet_listen(struct socket *sock, int backlog)
5682 struct sock *sk = sock->sk;
5683 struct crypto_hash *tfm = NULL;
5686 if (unlikely(backlog < 0))
5691 if (sock->state != SS_UNCONNECTED)
5694 /* Allocate HMAC for generating cookie. */
5695 if (sctp_hmac_alg) {
5696 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5698 if (net_ratelimit()) {
5700 "SCTP: failed to load transform for %s: %ld\n",
5701 sctp_hmac_alg, PTR_ERR(tfm));
5708 switch (sock->type) {
5709 case SOCK_SEQPACKET:
5710 err = sctp_seqpacket_listen(sk, backlog);
5713 err = sctp_stream_listen(sk, backlog);
5722 /* Store away the transform reference. */
5723 sctp_sk(sk)->hmac = tfm;
5725 sctp_release_sock(sk);
5728 crypto_free_hash(tfm);
5733 * This function is done by modeling the current datagram_poll() and the
5734 * tcp_poll(). Note that, based on these implementations, we don't
5735 * lock the socket in this function, even though it seems that,
5736 * ideally, locking or some other mechanisms can be used to ensure
5737 * the integrity of the counters (sndbuf and wmem_alloc) used
5738 * in this place. We assume that we don't need locks either until proven
5741 * Another thing to note is that we include the Async I/O support
5742 * here, again, by modeling the current TCP/UDP code. We don't have
5743 * a good way to test with it yet.
5745 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5747 struct sock *sk = sock->sk;
5748 struct sctp_sock *sp = sctp_sk(sk);
5751 poll_wait(file, sk->sk_sleep, wait);
5753 /* A TCP-style listening socket becomes readable when the accept queue
5756 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5757 return (!list_empty(&sp->ep->asocs)) ?
5758 (POLLIN | POLLRDNORM) : 0;
5762 /* Is there any exceptional events? */
5763 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5765 if (sk->sk_shutdown & RCV_SHUTDOWN)
5767 if (sk->sk_shutdown == SHUTDOWN_MASK)
5770 /* Is it readable? Reconsider this code with TCP-style support. */
5771 if (!skb_queue_empty(&sk->sk_receive_queue) ||
5772 (sk->sk_shutdown & RCV_SHUTDOWN))
5773 mask |= POLLIN | POLLRDNORM;
5775 /* The association is either gone or not ready. */
5776 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5779 /* Is it writable? */
5780 if (sctp_writeable(sk)) {
5781 mask |= POLLOUT | POLLWRNORM;
5783 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5785 * Since the socket is not locked, the buffer
5786 * might be made available after the writeable check and
5787 * before the bit is set. This could cause a lost I/O
5788 * signal. tcp_poll() has a race breaker for this race
5789 * condition. Based on their implementation, we put
5790 * in the following code to cover it as well.
5792 if (sctp_writeable(sk))
5793 mask |= POLLOUT | POLLWRNORM;
5798 /********************************************************************
5799 * 2nd Level Abstractions
5800 ********************************************************************/
5802 static struct sctp_bind_bucket *sctp_bucket_create(
5803 struct sctp_bind_hashbucket *head, unsigned short snum)
5805 struct sctp_bind_bucket *pp;
5807 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
5809 SCTP_DBG_OBJCNT_INC(bind_bucket);
5812 INIT_HLIST_HEAD(&pp->owner);
5813 hlist_add_head(&pp->node, &head->chain);
5818 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5819 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5821 if (pp && hlist_empty(&pp->owner)) {
5822 __hlist_del(&pp->node);
5823 kmem_cache_free(sctp_bucket_cachep, pp);
5824 SCTP_DBG_OBJCNT_DEC(bind_bucket);
5828 /* Release this socket's reference to a local port. */
5829 static inline void __sctp_put_port(struct sock *sk)
5831 struct sctp_bind_hashbucket *head =
5832 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
5833 struct sctp_bind_bucket *pp;
5835 sctp_spin_lock(&head->lock);
5836 pp = sctp_sk(sk)->bind_hash;
5837 __sk_del_bind_node(sk);
5838 sctp_sk(sk)->bind_hash = NULL;
5839 inet_sk(sk)->num = 0;
5840 sctp_bucket_destroy(pp);
5841 sctp_spin_unlock(&head->lock);
5844 void sctp_put_port(struct sock *sk)
5846 sctp_local_bh_disable();
5847 __sctp_put_port(sk);
5848 sctp_local_bh_enable();
5852 * The system picks an ephemeral port and choose an address set equivalent
5853 * to binding with a wildcard address.
5854 * One of those addresses will be the primary address for the association.
5855 * This automatically enables the multihoming capability of SCTP.
5857 static int sctp_autobind(struct sock *sk)
5859 union sctp_addr autoaddr;
5863 /* Initialize a local sockaddr structure to INADDR_ANY. */
5864 af = sctp_sk(sk)->pf->af;
5866 port = htons(inet_sk(sk)->num);
5867 af->inaddr_any(&autoaddr, port);
5869 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5872 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5875 * 4.2 The cmsghdr Structure *
5877 * When ancillary data is sent or received, any number of ancillary data
5878 * objects can be specified by the msg_control and msg_controllen members of
5879 * the msghdr structure, because each object is preceded by
5880 * a cmsghdr structure defining the object's length (the cmsg_len member).
5881 * Historically Berkeley-derived implementations have passed only one object
5882 * at a time, but this API allows multiple objects to be
5883 * passed in a single call to sendmsg() or recvmsg(). The following example
5884 * shows two ancillary data objects in a control buffer.
5886 * |<--------------------------- msg_controllen -------------------------->|
5889 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5891 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5894 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5896 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5899 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5900 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5902 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5904 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5911 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5912 sctp_cmsgs_t *cmsgs)
5914 struct cmsghdr *cmsg;
5915 struct msghdr *my_msg = (struct msghdr *)msg;
5917 for (cmsg = CMSG_FIRSTHDR(msg);
5919 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
5920 if (!CMSG_OK(my_msg, cmsg))
5923 /* Should we parse this header or ignore? */
5924 if (cmsg->cmsg_level != IPPROTO_SCTP)
5927 /* Strictly check lengths following example in SCM code. */
5928 switch (cmsg->cmsg_type) {
5930 /* SCTP Socket API Extension
5931 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5933 * This cmsghdr structure provides information for
5934 * initializing new SCTP associations with sendmsg().
5935 * The SCTP_INITMSG socket option uses this same data
5936 * structure. This structure is not used for
5939 * cmsg_level cmsg_type cmsg_data[]
5940 * ------------ ------------ ----------------------
5941 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5943 if (cmsg->cmsg_len !=
5944 CMSG_LEN(sizeof(struct sctp_initmsg)))
5946 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5950 /* SCTP Socket API Extension
5951 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5953 * This cmsghdr structure specifies SCTP options for
5954 * sendmsg() and describes SCTP header information
5955 * about a received message through recvmsg().
5957 * cmsg_level cmsg_type cmsg_data[]
5958 * ------------ ------------ ----------------------
5959 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5961 if (cmsg->cmsg_len !=
5962 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
5966 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
5968 /* Minimally, validate the sinfo_flags. */
5969 if (cmsgs->info->sinfo_flags &
5970 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
5971 SCTP_ABORT | SCTP_EOF))
5983 * Wait for a packet..
5984 * Note: This function is the same function as in core/datagram.c
5985 * with a few modifications to make lksctp work.
5987 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
5992 prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5994 /* Socket errors? */
5995 error = sock_error(sk);
5999 if (!skb_queue_empty(&sk->sk_receive_queue))
6002 /* Socket shut down? */
6003 if (sk->sk_shutdown & RCV_SHUTDOWN)
6006 /* Sequenced packets can come disconnected. If so we report the
6011 /* Is there a good reason to think that we may receive some data? */
6012 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6015 /* Handle signals. */
6016 if (signal_pending(current))
6019 /* Let another process have a go. Since we are going to sleep
6020 * anyway. Note: This may cause odd behaviors if the message
6021 * does not fit in the user's buffer, but this seems to be the
6022 * only way to honor MSG_DONTWAIT realistically.
6024 sctp_release_sock(sk);
6025 *timeo_p = schedule_timeout(*timeo_p);
6029 finish_wait(sk->sk_sleep, &wait);
6033 error = sock_intr_errno(*timeo_p);
6036 finish_wait(sk->sk_sleep, &wait);
6041 /* Receive a datagram.
6042 * Note: This is pretty much the same routine as in core/datagram.c
6043 * with a few changes to make lksctp work.
6045 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6046 int noblock, int *err)
6049 struct sk_buff *skb;
6052 timeo = sock_rcvtimeo(sk, noblock);
6054 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6055 timeo, MAX_SCHEDULE_TIMEOUT);
6058 /* Again only user level code calls this function,
6059 * so nothing interrupt level
6060 * will suddenly eat the receive_queue.
6062 * Look at current nfs client by the way...
6063 * However, this function was corrent in any case. 8)
6065 if (flags & MSG_PEEK) {
6066 spin_lock_bh(&sk->sk_receive_queue.lock);
6067 skb = skb_peek(&sk->sk_receive_queue);
6069 atomic_inc(&skb->users);
6070 spin_unlock_bh(&sk->sk_receive_queue.lock);
6072 skb = skb_dequeue(&sk->sk_receive_queue);
6078 /* Caller is allowed not to check sk->sk_err before calling. */
6079 error = sock_error(sk);
6083 if (sk->sk_shutdown & RCV_SHUTDOWN)
6086 /* User doesn't want to wait. */
6090 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6099 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6100 static void __sctp_write_space(struct sctp_association *asoc)
6102 struct sock *sk = asoc->base.sk;
6103 struct socket *sock = sk->sk_socket;
6105 if ((sctp_wspace(asoc) > 0) && sock) {
6106 if (waitqueue_active(&asoc->wait))
6107 wake_up_interruptible(&asoc->wait);
6109 if (sctp_writeable(sk)) {
6110 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
6111 wake_up_interruptible(sk->sk_sleep);
6113 /* Note that we try to include the Async I/O support
6114 * here by modeling from the current TCP/UDP code.
6115 * We have not tested with it yet.
6117 if (sock->fasync_list &&
6118 !(sk->sk_shutdown & SEND_SHUTDOWN))
6119 sock_wake_async(sock,
6120 SOCK_WAKE_SPACE, POLL_OUT);
6125 /* Do accounting for the sndbuf space.
6126 * Decrement the used sndbuf space of the corresponding association by the
6127 * data size which was just transmitted(freed).
6129 static void sctp_wfree(struct sk_buff *skb)
6131 struct sctp_association *asoc;
6132 struct sctp_chunk *chunk;
6135 /* Get the saved chunk pointer. */
6136 chunk = *((struct sctp_chunk **)(skb->cb));
6139 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6140 sizeof(struct sk_buff) +
6141 sizeof(struct sctp_chunk);
6143 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6146 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6148 sk->sk_wmem_queued -= skb->truesize;
6149 sk_mem_uncharge(sk, skb->truesize);
6152 __sctp_write_space(asoc);
6154 sctp_association_put(asoc);
6157 /* Do accounting for the receive space on the socket.
6158 * Accounting for the association is done in ulpevent.c
6159 * We set this as a destructor for the cloned data skbs so that
6160 * accounting is done at the correct time.
6162 void sctp_sock_rfree(struct sk_buff *skb)
6164 struct sock *sk = skb->sk;
6165 struct sctp_ulpevent *event = sctp_skb2event(skb);
6167 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6170 * Mimic the behavior of sock_rfree
6172 sk_mem_uncharge(sk, event->rmem_len);
6176 /* Helper function to wait for space in the sndbuf. */
6177 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6180 struct sock *sk = asoc->base.sk;
6182 long current_timeo = *timeo_p;
6185 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6186 asoc, (long)(*timeo_p), msg_len);
6188 /* Increment the association's refcnt. */
6189 sctp_association_hold(asoc);
6191 /* Wait on the association specific sndbuf space. */
6193 prepare_to_wait_exclusive(&asoc->wait, &wait,
6194 TASK_INTERRUPTIBLE);
6197 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6200 if (signal_pending(current))
6201 goto do_interrupted;
6202 if (msg_len <= sctp_wspace(asoc))
6205 /* Let another process have a go. Since we are going
6208 sctp_release_sock(sk);
6209 current_timeo = schedule_timeout(current_timeo);
6210 BUG_ON(sk != asoc->base.sk);
6213 *timeo_p = current_timeo;
6217 finish_wait(&asoc->wait, &wait);
6219 /* Release the association's refcnt. */
6220 sctp_association_put(asoc);
6229 err = sock_intr_errno(*timeo_p);
6237 /* If socket sndbuf has changed, wake up all per association waiters. */
6238 void sctp_write_space(struct sock *sk)
6240 struct sctp_association *asoc;
6242 /* Wake up the tasks in each wait queue. */
6243 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6244 __sctp_write_space(asoc);
6248 /* Is there any sndbuf space available on the socket?
6250 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6251 * associations on the same socket. For a UDP-style socket with
6252 * multiple associations, it is possible for it to be "unwriteable"
6253 * prematurely. I assume that this is acceptable because
6254 * a premature "unwriteable" is better than an accidental "writeable" which
6255 * would cause an unwanted block under certain circumstances. For the 1-1
6256 * UDP-style sockets or TCP-style sockets, this code should work.
6259 static int sctp_writeable(struct sock *sk)
6263 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
6269 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6270 * returns immediately with EINPROGRESS.
6272 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6274 struct sock *sk = asoc->base.sk;
6276 long current_timeo = *timeo_p;
6279 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6282 /* Increment the association's refcnt. */
6283 sctp_association_hold(asoc);
6286 prepare_to_wait_exclusive(&asoc->wait, &wait,
6287 TASK_INTERRUPTIBLE);
6290 if (sk->sk_shutdown & RCV_SHUTDOWN)
6292 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6295 if (signal_pending(current))
6296 goto do_interrupted;
6298 if (sctp_state(asoc, ESTABLISHED))
6301 /* Let another process have a go. Since we are going
6304 sctp_release_sock(sk);
6305 current_timeo = schedule_timeout(current_timeo);
6308 *timeo_p = current_timeo;
6312 finish_wait(&asoc->wait, &wait);
6314 /* Release the association's refcnt. */
6315 sctp_association_put(asoc);
6320 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6323 err = -ECONNREFUSED;
6327 err = sock_intr_errno(*timeo_p);
6335 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6337 struct sctp_endpoint *ep;
6341 ep = sctp_sk(sk)->ep;
6345 prepare_to_wait_exclusive(sk->sk_sleep, &wait,
6346 TASK_INTERRUPTIBLE);
6348 if (list_empty(&ep->asocs)) {
6349 sctp_release_sock(sk);
6350 timeo = schedule_timeout(timeo);
6355 if (!sctp_sstate(sk, LISTENING))
6359 if (!list_empty(&ep->asocs))
6362 err = sock_intr_errno(timeo);
6363 if (signal_pending(current))
6371 finish_wait(sk->sk_sleep, &wait);
6376 static void sctp_wait_for_close(struct sock *sk, long timeout)
6381 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
6382 if (list_empty(&sctp_sk(sk)->ep->asocs))
6384 sctp_release_sock(sk);
6385 timeout = schedule_timeout(timeout);
6387 } while (!signal_pending(current) && timeout);
6389 finish_wait(sk->sk_sleep, &wait);
6392 static void sctp_sock_rfree_frag(struct sk_buff *skb)
6394 struct sk_buff *frag;
6399 /* Don't forget the fragments. */
6400 for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
6401 sctp_sock_rfree_frag(frag);
6404 sctp_sock_rfree(skb);
6407 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6409 struct sk_buff *frag;
6414 /* Don't forget the fragments. */
6415 for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
6416 sctp_skb_set_owner_r_frag(frag, sk);
6419 sctp_skb_set_owner_r(skb, sk);
6422 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6423 * and its messages to the newsk.
6425 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6426 struct sctp_association *assoc,
6427 sctp_socket_type_t type)
6429 struct sctp_sock *oldsp = sctp_sk(oldsk);
6430 struct sctp_sock *newsp = sctp_sk(newsk);
6431 struct sctp_bind_bucket *pp; /* hash list port iterator */
6432 struct sctp_endpoint *newep = newsp->ep;
6433 struct sk_buff *skb, *tmp;
6434 struct sctp_ulpevent *event;
6435 struct sctp_bind_hashbucket *head;
6437 /* Migrate socket buffer sizes and all the socket level options to the
6440 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6441 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6442 /* Brute force copy old sctp opt. */
6443 inet_sk_copy_descendant(newsk, oldsk);
6445 /* Restore the ep value that was overwritten with the above structure
6451 /* Hook this new socket in to the bind_hash list. */
6452 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->num)];
6453 sctp_local_bh_disable();
6454 sctp_spin_lock(&head->lock);
6455 pp = sctp_sk(oldsk)->bind_hash;
6456 sk_add_bind_node(newsk, &pp->owner);
6457 sctp_sk(newsk)->bind_hash = pp;
6458 inet_sk(newsk)->num = inet_sk(oldsk)->num;
6459 sctp_spin_unlock(&head->lock);
6460 sctp_local_bh_enable();
6462 /* Copy the bind_addr list from the original endpoint to the new
6463 * endpoint so that we can handle restarts properly
6465 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6466 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6468 /* Move any messages in the old socket's receive queue that are for the
6469 * peeled off association to the new socket's receive queue.
6471 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6472 event = sctp_skb2event(skb);
6473 if (event->asoc == assoc) {
6474 sctp_sock_rfree_frag(skb);
6475 __skb_unlink(skb, &oldsk->sk_receive_queue);
6476 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6477 sctp_skb_set_owner_r_frag(skb, newsk);
6481 /* Clean up any messages pending delivery due to partial
6482 * delivery. Three cases:
6483 * 1) No partial deliver; no work.
6484 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6485 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6487 skb_queue_head_init(&newsp->pd_lobby);
6488 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6490 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6491 struct sk_buff_head *queue;
6493 /* Decide which queue to move pd_lobby skbs to. */
6494 if (assoc->ulpq.pd_mode) {
6495 queue = &newsp->pd_lobby;
6497 queue = &newsk->sk_receive_queue;
6499 /* Walk through the pd_lobby, looking for skbs that
6500 * need moved to the new socket.
6502 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6503 event = sctp_skb2event(skb);
6504 if (event->asoc == assoc) {
6505 sctp_sock_rfree_frag(skb);
6506 __skb_unlink(skb, &oldsp->pd_lobby);
6507 __skb_queue_tail(queue, skb);
6508 sctp_skb_set_owner_r_frag(skb, newsk);
6512 /* Clear up any skbs waiting for the partial
6513 * delivery to finish.
6515 if (assoc->ulpq.pd_mode)
6516 sctp_clear_pd(oldsk, NULL);
6520 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp) {
6521 sctp_sock_rfree_frag(skb);
6522 sctp_skb_set_owner_r_frag(skb, newsk);
6525 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp) {
6526 sctp_sock_rfree_frag(skb);
6527 sctp_skb_set_owner_r_frag(skb, newsk);
6530 /* Set the type of socket to indicate that it is peeled off from the
6531 * original UDP-style socket or created with the accept() call on a
6532 * TCP-style socket..
6536 /* Mark the new socket "in-use" by the user so that any packets
6537 * that may arrive on the association after we've moved it are
6538 * queued to the backlog. This prevents a potential race between
6539 * backlog processing on the old socket and new-packet processing
6540 * on the new socket.
6542 * The caller has just allocated newsk so we can guarantee that other
6543 * paths won't try to lock it and then oldsk.
6545 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6546 sctp_assoc_migrate(assoc, newsk);
6548 /* If the association on the newsk is already closed before accept()
6549 * is called, set RCV_SHUTDOWN flag.
6551 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6552 newsk->sk_shutdown |= RCV_SHUTDOWN;
6554 newsk->sk_state = SCTP_SS_ESTABLISHED;
6555 sctp_release_sock(newsk);
6559 /* This proto struct describes the ULP interface for SCTP. */
6560 struct proto sctp_prot = {
6562 .owner = THIS_MODULE,
6563 .close = sctp_close,
6564 .connect = sctp_connect,
6565 .disconnect = sctp_disconnect,
6566 .accept = sctp_accept,
6567 .ioctl = sctp_ioctl,
6568 .init = sctp_init_sock,
6569 .destroy = sctp_destroy_sock,
6570 .shutdown = sctp_shutdown,
6571 .setsockopt = sctp_setsockopt,
6572 .getsockopt = sctp_getsockopt,
6573 .sendmsg = sctp_sendmsg,
6574 .recvmsg = sctp_recvmsg,
6576 .backlog_rcv = sctp_backlog_rcv,
6578 .unhash = sctp_unhash,
6579 .get_port = sctp_get_port,
6580 .obj_size = sizeof(struct sctp_sock),
6581 .sysctl_mem = sysctl_sctp_mem,
6582 .sysctl_rmem = sysctl_sctp_rmem,
6583 .sysctl_wmem = sysctl_sctp_wmem,
6584 .memory_pressure = &sctp_memory_pressure,
6585 .enter_memory_pressure = sctp_enter_memory_pressure,
6586 .memory_allocated = &sctp_memory_allocated,
6587 .sockets_allocated = &sctp_sockets_allocated,
6590 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6592 struct proto sctpv6_prot = {
6594 .owner = THIS_MODULE,
6595 .close = sctp_close,
6596 .connect = sctp_connect,
6597 .disconnect = sctp_disconnect,
6598 .accept = sctp_accept,
6599 .ioctl = sctp_ioctl,
6600 .init = sctp_init_sock,
6601 .destroy = sctp_destroy_sock,
6602 .shutdown = sctp_shutdown,
6603 .setsockopt = sctp_setsockopt,
6604 .getsockopt = sctp_getsockopt,
6605 .sendmsg = sctp_sendmsg,
6606 .recvmsg = sctp_recvmsg,
6608 .backlog_rcv = sctp_backlog_rcv,
6610 .unhash = sctp_unhash,
6611 .get_port = sctp_get_port,
6612 .obj_size = sizeof(struct sctp6_sock),
6613 .sysctl_mem = sysctl_sctp_mem,
6614 .sysctl_rmem = sysctl_sctp_rmem,
6615 .sysctl_wmem = sysctl_sctp_wmem,
6616 .memory_pressure = &sctp_memory_pressure,
6617 .enter_memory_pressure = sctp_enter_memory_pressure,
6618 .memory_allocated = &sctp_memory_allocated,
6619 .sockets_allocated = &sctp_sockets_allocated,
6621 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */