1 /* SCTP kernel reference Implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
9 * This file is part of the SCTP kernel reference Implementation
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
18 * The SCTP reference implementation is free software;
19 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
24 * The SCTP reference implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, write to
32 * the Free Software Foundation, 59 Temple Place - Suite 330,
33 * Boston, MA 02111-1307, USA.
35 * Please send any bug reports or fixes you make to the
37 * lksctp developers <lksctp-developers@lists.sourceforge.net>
39 * Or submit a bug report through the following website:
40 * http://www.sf.net/projects/lksctp
42 * Written or modified by:
43 * La Monte H.P. Yarroll <piggy@acm.org>
44 * Narasimha Budihal <narsi@refcode.org>
45 * Karl Knutson <karl@athena.chicago.il.us>
46 * Jon Grimm <jgrimm@us.ibm.com>
47 * Xingang Guo <xingang.guo@intel.com>
48 * Daisy Chang <daisyc@us.ibm.com>
49 * Sridhar Samudrala <samudrala@us.ibm.com>
50 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
51 * Ardelle Fan <ardelle.fan@intel.com>
52 * Ryan Layer <rmlayer@us.ibm.com>
53 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
54 * Kevin Gao <kevin.gao@intel.com>
56 * Any bugs reported given to us we will try to fix... any fixes shared will
57 * be incorporated into the next SCTP release.
60 #include <linux/config.h>
61 #include <linux/types.h>
62 #include <linux/kernel.h>
63 #include <linux/wait.h>
64 #include <linux/time.h>
66 #include <linux/capability.h>
67 #include <linux/fcntl.h>
68 #include <linux/poll.h>
69 #include <linux/init.h>
70 #include <linux/crypto.h>
74 #include <net/route.h>
76 #include <net/inet_common.h>
78 #include <linux/socket.h> /* for sa_family_t */
80 #include <net/sctp/sctp.h>
81 #include <net/sctp/sm.h>
83 /* WARNING: Please do not remove the SCTP_STATIC attribute to
84 * any of the functions below as they are used to export functions
85 * used by a project regression testsuite.
88 /* Forward declarations for internal helper functions. */
89 static int sctp_writeable(struct sock *sk);
90 static void sctp_wfree(struct sk_buff *skb);
91 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
93 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
94 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
95 static int sctp_wait_for_accept(struct sock *sk, long timeo);
96 static void sctp_wait_for_close(struct sock *sk, long timeo);
97 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
98 union sctp_addr *addr, int len);
99 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
100 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
101 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
102 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
103 static int sctp_send_asconf(struct sctp_association *asoc,
104 struct sctp_chunk *chunk);
105 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
106 static int sctp_autobind(struct sock *sk);
107 static void sctp_sock_migrate(struct sock *, struct sock *,
108 struct sctp_association *, sctp_socket_type_t);
109 static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
111 extern kmem_cache_t *sctp_bucket_cachep;
113 /* Get the sndbuf space available at the time on the association. */
114 static inline int sctp_wspace(struct sctp_association *asoc)
116 struct sock *sk = asoc->base.sk;
119 if (asoc->ep->sndbuf_policy) {
120 /* make sure that no association uses more than sk_sndbuf */
121 amt = sk->sk_sndbuf - asoc->sndbuf_used;
123 /* do socket level accounting */
124 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
133 /* Increment the used sndbuf space count of the corresponding association by
134 * the size of the outgoing data chunk.
135 * Also, set the skb destructor for sndbuf accounting later.
137 * Since it is always 1-1 between chunk and skb, and also a new skb is always
138 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
139 * destructor in the data chunk skb for the purpose of the sndbuf space
142 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
144 struct sctp_association *asoc = chunk->asoc;
145 struct sock *sk = asoc->base.sk;
147 /* The sndbuf space is tracked per association. */
148 sctp_association_hold(asoc);
150 skb_set_owner_w(chunk->skb, sk);
152 chunk->skb->destructor = sctp_wfree;
153 /* Save the chunk pointer in skb for sctp_wfree to use later. */
154 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
156 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
157 sizeof(struct sk_buff) +
158 sizeof(struct sctp_chunk);
160 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
163 /* Verify that this is a valid address. */
164 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
169 /* Verify basic sockaddr. */
170 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
174 /* Is this a valid SCTP address? */
175 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
178 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
184 /* Look up the association by its id. If this is not a UDP-style
185 * socket, the ID field is always ignored.
187 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
189 struct sctp_association *asoc = NULL;
191 /* If this is not a UDP-style socket, assoc id should be ignored. */
192 if (!sctp_style(sk, UDP)) {
193 /* Return NULL if the socket state is not ESTABLISHED. It
194 * could be a TCP-style listening socket or a socket which
195 * hasn't yet called connect() to establish an association.
197 if (!sctp_sstate(sk, ESTABLISHED))
200 /* Get the first and the only association from the list. */
201 if (!list_empty(&sctp_sk(sk)->ep->asocs))
202 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
203 struct sctp_association, asocs);
207 /* Otherwise this is a UDP-style socket. */
208 if (!id || (id == (sctp_assoc_t)-1))
211 spin_lock_bh(&sctp_assocs_id_lock);
212 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
213 spin_unlock_bh(&sctp_assocs_id_lock);
215 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
221 /* Look up the transport from an address and an assoc id. If both address and
222 * id are specified, the associations matching the address and the id should be
225 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
226 struct sockaddr_storage *addr,
229 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
230 struct sctp_transport *transport;
231 union sctp_addr *laddr = (union sctp_addr *)addr;
233 laddr->v4.sin_port = ntohs(laddr->v4.sin_port);
234 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
235 (union sctp_addr *)addr,
237 laddr->v4.sin_port = htons(laddr->v4.sin_port);
242 id_asoc = sctp_id2assoc(sk, id);
243 if (id_asoc && (id_asoc != addr_asoc))
246 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
247 (union sctp_addr *)addr);
252 /* API 3.1.2 bind() - UDP Style Syntax
253 * The syntax of bind() is,
255 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
257 * sd - the socket descriptor returned by socket().
258 * addr - the address structure (struct sockaddr_in or struct
259 * sockaddr_in6 [RFC 2553]),
260 * addr_len - the size of the address structure.
262 SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
268 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
271 /* Disallow binding twice. */
272 if (!sctp_sk(sk)->ep->base.bind_addr.port)
273 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
278 sctp_release_sock(sk);
283 static long sctp_get_port_local(struct sock *, union sctp_addr *);
285 /* Verify this is a valid sockaddr. */
286 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
287 union sctp_addr *addr, int len)
291 /* Check minimum size. */
292 if (len < sizeof (struct sockaddr))
295 /* Does this PF support this AF? */
296 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
299 /* If we get this far, af is valid. */
300 af = sctp_get_af_specific(addr->sa.sa_family);
302 if (len < af->sockaddr_len)
308 /* Bind a local address either to an endpoint or to an association. */
309 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
311 struct sctp_sock *sp = sctp_sk(sk);
312 struct sctp_endpoint *ep = sp->ep;
313 struct sctp_bind_addr *bp = &ep->base.bind_addr;
318 /* Common sockaddr verification. */
319 af = sctp_sockaddr_af(sp, addr, len);
321 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
326 snum = ntohs(addr->v4.sin_port);
328 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
329 ", port: %d, new port: %d, len: %d)\n",
335 /* PF specific bind() address verification. */
336 if (!sp->pf->bind_verify(sp, addr))
337 return -EADDRNOTAVAIL;
339 /* We must either be unbound, or bind to the same port. */
340 if (bp->port && (snum != bp->port)) {
341 SCTP_DEBUG_PRINTK("sctp_do_bind:"
342 " New port %d does not match existing port "
343 "%d.\n", snum, bp->port);
347 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
350 /* Make sure we are allowed to bind here.
351 * The function sctp_get_port_local() does duplicate address
354 if ((ret = sctp_get_port_local(sk, addr))) {
355 if (ret == (long) sk) {
356 /* This endpoint has a conflicting address. */
363 /* Refresh ephemeral port. */
365 bp->port = inet_sk(sk)->num;
367 /* Add the address to the bind address list. */
368 sctp_local_bh_disable();
369 sctp_write_lock(&ep->base.addr_lock);
371 /* Use GFP_ATOMIC since BHs are disabled. */
372 addr->v4.sin_port = ntohs(addr->v4.sin_port);
373 ret = sctp_add_bind_addr(bp, addr, GFP_ATOMIC);
374 addr->v4.sin_port = htons(addr->v4.sin_port);
375 sctp_write_unlock(&ep->base.addr_lock);
376 sctp_local_bh_enable();
378 /* Copy back into socket for getsockname() use. */
380 inet_sk(sk)->sport = htons(inet_sk(sk)->num);
381 af->to_sk_saddr(addr, sk);
387 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
389 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
390 * at any one time. If a sender, after sending an ASCONF chunk, decides
391 * it needs to transfer another ASCONF Chunk, it MUST wait until the
392 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
393 * subsequent ASCONF. Note this restriction binds each side, so at any
394 * time two ASCONF may be in-transit on any given association (one sent
395 * from each endpoint).
397 static int sctp_send_asconf(struct sctp_association *asoc,
398 struct sctp_chunk *chunk)
402 /* If there is an outstanding ASCONF chunk, queue it for later
405 if (asoc->addip_last_asconf) {
406 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
410 /* Hold the chunk until an ASCONF_ACK is received. */
411 sctp_chunk_hold(chunk);
412 retval = sctp_primitive_ASCONF(asoc, chunk);
414 sctp_chunk_free(chunk);
416 asoc->addip_last_asconf = chunk;
422 /* Add a list of addresses as bind addresses to local endpoint or
425 * Basically run through each address specified in the addrs/addrcnt
426 * array/length pair, determine if it is IPv6 or IPv4 and call
427 * sctp_do_bind() on it.
429 * If any of them fails, then the operation will be reversed and the
430 * ones that were added will be removed.
432 * Only sctp_setsockopt_bindx() is supposed to call this function.
434 int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
439 struct sockaddr *sa_addr;
442 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
446 for (cnt = 0; cnt < addrcnt; cnt++) {
447 /* The list may contain either IPv4 or IPv6 address;
448 * determine the address length for walking thru the list.
450 sa_addr = (struct sockaddr *)addr_buf;
451 af = sctp_get_af_specific(sa_addr->sa_family);
457 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
460 addr_buf += af->sockaddr_len;
464 /* Failed. Cleanup the ones that have been added */
466 sctp_bindx_rem(sk, addrs, cnt);
474 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
475 * associations that are part of the endpoint indicating that a list of local
476 * addresses are added to the endpoint.
478 * If any of the addresses is already in the bind address list of the
479 * association, we do not send the chunk for that association. But it will not
480 * affect other associations.
482 * Only sctp_setsockopt_bindx() is supposed to call this function.
484 static int sctp_send_asconf_add_ip(struct sock *sk,
485 struct sockaddr *addrs,
488 struct sctp_sock *sp;
489 struct sctp_endpoint *ep;
490 struct sctp_association *asoc;
491 struct sctp_bind_addr *bp;
492 struct sctp_chunk *chunk;
493 struct sctp_sockaddr_entry *laddr;
494 union sctp_addr *addr;
497 struct list_head *pos;
502 if (!sctp_addip_enable)
508 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
509 __FUNCTION__, sk, addrs, addrcnt);
511 list_for_each(pos, &ep->asocs) {
512 asoc = list_entry(pos, struct sctp_association, asocs);
514 if (!asoc->peer.asconf_capable)
517 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
520 if (!sctp_state(asoc, ESTABLISHED))
523 /* Check if any address in the packed array of addresses is
524 * in the bind address list of the association. If so,
525 * do not send the asconf chunk to its peer, but continue with
526 * other associations.
529 for (i = 0; i < addrcnt; i++) {
530 addr = (union sctp_addr *)addr_buf;
531 af = sctp_get_af_specific(addr->v4.sin_family);
537 if (sctp_assoc_lookup_laddr(asoc, addr))
540 addr_buf += af->sockaddr_len;
545 /* Use the first address in bind addr list of association as
546 * Address Parameter of ASCONF CHUNK.
548 sctp_read_lock(&asoc->base.addr_lock);
549 bp = &asoc->base.bind_addr;
550 p = bp->address_list.next;
551 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
552 sctp_read_unlock(&asoc->base.addr_lock);
554 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
555 addrcnt, SCTP_PARAM_ADD_IP);
561 retval = sctp_send_asconf(asoc, chunk);
563 /* FIXME: After sending the add address ASCONF chunk, we
564 * cannot append the address to the association's binding
565 * address list, because the new address may be used as the
566 * source of a message sent to the peer before the ASCONF
567 * chunk is received by the peer. So we should wait until
568 * ASCONF_ACK is received.
576 /* Remove a list of addresses from bind addresses list. Do not remove the
579 * Basically run through each address specified in the addrs/addrcnt
580 * array/length pair, determine if it is IPv6 or IPv4 and call
581 * sctp_del_bind() on it.
583 * If any of them fails, then the operation will be reversed and the
584 * ones that were removed will be added back.
586 * At least one address has to be left; if only one address is
587 * available, the operation will return -EBUSY.
589 * Only sctp_setsockopt_bindx() is supposed to call this function.
591 int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
593 struct sctp_sock *sp = sctp_sk(sk);
594 struct sctp_endpoint *ep = sp->ep;
596 struct sctp_bind_addr *bp = &ep->base.bind_addr;
598 union sctp_addr saveaddr;
600 struct sockaddr *sa_addr;
603 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
607 for (cnt = 0; cnt < addrcnt; cnt++) {
608 /* If the bind address list is empty or if there is only one
609 * bind address, there is nothing more to be removed (we need
610 * at least one address here).
612 if (list_empty(&bp->address_list) ||
613 (sctp_list_single_entry(&bp->address_list))) {
618 /* The list may contain either IPv4 or IPv6 address;
619 * determine the address length to copy the address to
622 sa_addr = (struct sockaddr *)addr_buf;
623 af = sctp_get_af_specific(sa_addr->sa_family);
628 memcpy(&saveaddr, sa_addr, af->sockaddr_len);
629 saveaddr.v4.sin_port = ntohs(saveaddr.v4.sin_port);
630 if (saveaddr.v4.sin_port != bp->port) {
635 /* FIXME - There is probably a need to check if sk->sk_saddr and
636 * sk->sk_rcv_addr are currently set to one of the addresses to
637 * be removed. This is something which needs to be looked into
638 * when we are fixing the outstanding issues with multi-homing
639 * socket routing and failover schemes. Refer to comments in
640 * sctp_do_bind(). -daisy
642 sctp_local_bh_disable();
643 sctp_write_lock(&ep->base.addr_lock);
645 retval = sctp_del_bind_addr(bp, &saveaddr);
647 sctp_write_unlock(&ep->base.addr_lock);
648 sctp_local_bh_enable();
650 addr_buf += af->sockaddr_len;
653 /* Failed. Add the ones that has been removed back */
655 sctp_bindx_add(sk, addrs, cnt);
663 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
664 * the associations that are part of the endpoint indicating that a list of
665 * local addresses are removed from the endpoint.
667 * If any of the addresses is already in the bind address list of the
668 * association, we do not send the chunk for that association. But it will not
669 * affect other associations.
671 * Only sctp_setsockopt_bindx() is supposed to call this function.
673 static int sctp_send_asconf_del_ip(struct sock *sk,
674 struct sockaddr *addrs,
677 struct sctp_sock *sp;
678 struct sctp_endpoint *ep;
679 struct sctp_association *asoc;
680 struct sctp_bind_addr *bp;
681 struct sctp_chunk *chunk;
682 union sctp_addr *laddr;
685 struct list_head *pos;
689 if (!sctp_addip_enable)
695 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
696 __FUNCTION__, sk, addrs, addrcnt);
698 list_for_each(pos, &ep->asocs) {
699 asoc = list_entry(pos, struct sctp_association, asocs);
701 if (!asoc->peer.asconf_capable)
704 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
707 if (!sctp_state(asoc, ESTABLISHED))
710 /* Check if any address in the packed array of addresses is
711 * not present in the bind address list of the association.
712 * If so, do not send the asconf chunk to its peer, but
713 * continue with other associations.
716 for (i = 0; i < addrcnt; i++) {
717 laddr = (union sctp_addr *)addr_buf;
718 af = sctp_get_af_specific(laddr->v4.sin_family);
724 if (!sctp_assoc_lookup_laddr(asoc, laddr))
727 addr_buf += af->sockaddr_len;
732 /* Find one address in the association's bind address list
733 * that is not in the packed array of addresses. This is to
734 * make sure that we do not delete all the addresses in the
737 sctp_read_lock(&asoc->base.addr_lock);
738 bp = &asoc->base.bind_addr;
739 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
741 sctp_read_unlock(&asoc->base.addr_lock);
745 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
752 retval = sctp_send_asconf(asoc, chunk);
754 /* FIXME: After sending the delete address ASCONF chunk, we
755 * cannot remove the addresses from the association's bind
756 * address list, because there maybe some packet send to
757 * the delete addresses, so we should wait until ASCONF_ACK
758 * packet is received.
765 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
768 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
771 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
772 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
775 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
776 * Section 3.1.2 for this usage.
778 * addrs is a pointer to an array of one or more socket addresses. Each
779 * address is contained in its appropriate structure (i.e. struct
780 * sockaddr_in or struct sockaddr_in6) the family of the address type
781 * must be used to distengish the address length (note that this
782 * representation is termed a "packed array" of addresses). The caller
783 * specifies the number of addresses in the array with addrcnt.
785 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
786 * -1, and sets errno to the appropriate error code.
788 * For SCTP, the port given in each socket address must be the same, or
789 * sctp_bindx() will fail, setting errno to EINVAL.
791 * The flags parameter is formed from the bitwise OR of zero or more of
792 * the following currently defined flags:
794 * SCTP_BINDX_ADD_ADDR
796 * SCTP_BINDX_REM_ADDR
798 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
799 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
800 * addresses from the association. The two flags are mutually exclusive;
801 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
802 * not remove all addresses from an association; sctp_bindx() will
803 * reject such an attempt with EINVAL.
805 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
806 * additional addresses with an endpoint after calling bind(). Or use
807 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
808 * socket is associated with so that no new association accepted will be
809 * associated with those addresses. If the endpoint supports dynamic
810 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
811 * endpoint to send the appropriate message to the peer to change the
812 * peers address lists.
814 * Adding and removing addresses from a connected association is
815 * optional functionality. Implementations that do not support this
816 * functionality should return EOPNOTSUPP.
818 * Basically do nothing but copying the addresses from user to kernel
819 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
820 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
823 * We don't use copy_from_user() for optimization: we first do the
824 * sanity checks (buffer size -fast- and access check-healthy
825 * pointer); if all of those succeed, then we can alloc the memory
826 * (expensive operation) needed to copy the data to kernel. Then we do
827 * the copying without checking the user space area
828 * (__copy_from_user()).
830 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
833 * sk The sk of the socket
834 * addrs The pointer to the addresses in user land
835 * addrssize Size of the addrs buffer
836 * op Operation to perform (add or remove, see the flags of
839 * Returns 0 if ok, <0 errno code on error.
841 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
842 struct sockaddr __user *addrs,
843 int addrs_size, int op)
845 struct sockaddr *kaddrs;
849 struct sockaddr *sa_addr;
853 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
854 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
856 if (unlikely(addrs_size <= 0))
859 /* Check the user passed a healthy pointer. */
860 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
863 /* Alloc space for the address array in kernel memory. */
864 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
865 if (unlikely(!kaddrs))
868 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
873 /* Walk through the addrs buffer and count the number of addresses. */
875 while (walk_size < addrs_size) {
876 sa_addr = (struct sockaddr *)addr_buf;
877 af = sctp_get_af_specific(sa_addr->sa_family);
879 /* If the address family is not supported or if this address
880 * causes the address buffer to overflow return EINVAL.
882 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
887 addr_buf += af->sockaddr_len;
888 walk_size += af->sockaddr_len;
893 case SCTP_BINDX_ADD_ADDR:
894 err = sctp_bindx_add(sk, kaddrs, addrcnt);
897 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
900 case SCTP_BINDX_REM_ADDR:
901 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
904 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
918 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
920 * Common routine for handling connect() and sctp_connectx().
921 * Connect will come in with just a single address.
923 static int __sctp_connect(struct sock* sk,
924 struct sockaddr *kaddrs,
927 struct sctp_sock *sp;
928 struct sctp_endpoint *ep;
929 struct sctp_association *asoc = NULL;
930 struct sctp_association *asoc2;
931 struct sctp_transport *transport;
939 struct sockaddr *sa_addr;
945 /* connect() cannot be done on a socket that is already in ESTABLISHED
946 * state - UDP-style peeled off socket or a TCP-style socket that
947 * is already connected.
948 * It cannot be done even on a TCP-style listening socket.
950 if (sctp_sstate(sk, ESTABLISHED) ||
951 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
956 /* Walk through the addrs buffer and count the number of addresses. */
958 while (walk_size < addrs_size) {
959 sa_addr = (struct sockaddr *)addr_buf;
960 af = sctp_get_af_specific(sa_addr->sa_family);
962 /* If the address family is not supported or if this address
963 * causes the address buffer to overflow return EINVAL.
965 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
970 err = sctp_verify_addr(sk, (union sctp_addr *)sa_addr,
975 memcpy(&to, sa_addr, af->sockaddr_len);
976 to.v4.sin_port = ntohs(to.v4.sin_port);
978 /* Check if there already is a matching association on the
979 * endpoint (other than the one created here).
981 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
982 if (asoc2 && asoc2 != asoc) {
983 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
990 /* If we could not find a matching association on the endpoint,
991 * make sure that there is no peeled-off association matching
992 * the peer address even on another socket.
994 if (sctp_endpoint_is_peeled_off(ep, &to)) {
995 err = -EADDRNOTAVAIL;
1000 /* If a bind() or sctp_bindx() is not called prior to
1001 * an sctp_connectx() call, the system picks an
1002 * ephemeral port and will choose an address set
1003 * equivalent to binding with a wildcard address.
1005 if (!ep->base.bind_addr.port) {
1006 if (sctp_autobind(sk)) {
1012 * If an unprivileged user inherits a 1-many
1013 * style socket with open associations on a
1014 * privileged port, it MAY be permitted to
1015 * accept new associations, but it SHOULD NOT
1016 * be permitted to open new associations.
1018 if (ep->base.bind_addr.port < PROT_SOCK &&
1019 !capable(CAP_NET_BIND_SERVICE)) {
1025 scope = sctp_scope(&to);
1026 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1033 /* Prime the peer's transport structures. */
1034 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1042 addr_buf += af->sockaddr_len;
1043 walk_size += af->sockaddr_len;
1046 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1051 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1056 /* Initialize sk's dport and daddr for getpeername() */
1057 inet_sk(sk)->dport = htons(asoc->peer.port);
1058 af = sctp_get_af_specific(to.sa.sa_family);
1059 af->to_sk_daddr(&to, sk);
1062 timeo = sock_sndtimeo(sk, sk->sk_socket->file->f_flags & O_NONBLOCK);
1063 err = sctp_wait_for_connect(asoc, &timeo);
1065 /* Don't free association on exit. */
1070 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1071 " kaddrs: %p err: %d\n",
1074 sctp_association_free(asoc);
1078 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1081 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt);
1083 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1084 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1085 * or IPv6 addresses.
1087 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1088 * Section 3.1.2 for this usage.
1090 * addrs is a pointer to an array of one or more socket addresses. Each
1091 * address is contained in its appropriate structure (i.e. struct
1092 * sockaddr_in or struct sockaddr_in6) the family of the address type
1093 * must be used to distengish the address length (note that this
1094 * representation is termed a "packed array" of addresses). The caller
1095 * specifies the number of addresses in the array with addrcnt.
1097 * On success, sctp_connectx() returns 0. On failure, sctp_connectx() returns
1098 * -1, and sets errno to the appropriate error code.
1100 * For SCTP, the port given in each socket address must be the same, or
1101 * sctp_connectx() will fail, setting errno to EINVAL.
1103 * An application can use sctp_connectx to initiate an association with
1104 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1105 * allows a caller to specify multiple addresses at which a peer can be
1106 * reached. The way the SCTP stack uses the list of addresses to set up
1107 * the association is implementation dependant. This function only
1108 * specifies that the stack will try to make use of all the addresses in
1109 * the list when needed.
1111 * Note that the list of addresses passed in is only used for setting up
1112 * the association. It does not necessarily equal the set of addresses
1113 * the peer uses for the resulting association. If the caller wants to
1114 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1115 * retrieve them after the association has been set up.
1117 * Basically do nothing but copying the addresses from user to kernel
1118 * land and invoking either sctp_connectx(). This is used for tunneling
1119 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1121 * We don't use copy_from_user() for optimization: we first do the
1122 * sanity checks (buffer size -fast- and access check-healthy
1123 * pointer); if all of those succeed, then we can alloc the memory
1124 * (expensive operation) needed to copy the data to kernel. Then we do
1125 * the copying without checking the user space area
1126 * (__copy_from_user()).
1128 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1131 * sk The sk of the socket
1132 * addrs The pointer to the addresses in user land
1133 * addrssize Size of the addrs buffer
1135 * Returns 0 if ok, <0 errno code on error.
1137 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1138 struct sockaddr __user *addrs,
1142 struct sockaddr *kaddrs;
1144 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1145 __FUNCTION__, sk, addrs, addrs_size);
1147 if (unlikely(addrs_size <= 0))
1150 /* Check the user passed a healthy pointer. */
1151 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1154 /* Alloc space for the address array in kernel memory. */
1155 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1156 if (unlikely(!kaddrs))
1159 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1162 err = __sctp_connect(sk, kaddrs, addrs_size);
1169 /* API 3.1.4 close() - UDP Style Syntax
1170 * Applications use close() to perform graceful shutdown (as described in
1171 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1172 * by a UDP-style socket.
1176 * ret = close(int sd);
1178 * sd - the socket descriptor of the associations to be closed.
1180 * To gracefully shutdown a specific association represented by the
1181 * UDP-style socket, an application should use the sendmsg() call,
1182 * passing no user data, but including the appropriate flag in the
1183 * ancillary data (see Section xxxx).
1185 * If sd in the close() call is a branched-off socket representing only
1186 * one association, the shutdown is performed on that association only.
1188 * 4.1.6 close() - TCP Style Syntax
1190 * Applications use close() to gracefully close down an association.
1194 * int close(int sd);
1196 * sd - the socket descriptor of the association to be closed.
1198 * After an application calls close() on a socket descriptor, no further
1199 * socket operations will succeed on that descriptor.
1201 * API 7.1.4 SO_LINGER
1203 * An application using the TCP-style socket can use this option to
1204 * perform the SCTP ABORT primitive. The linger option structure is:
1207 * int l_onoff; // option on/off
1208 * int l_linger; // linger time
1211 * To enable the option, set l_onoff to 1. If the l_linger value is set
1212 * to 0, calling close() is the same as the ABORT primitive. If the
1213 * value is set to a negative value, the setsockopt() call will return
1214 * an error. If the value is set to a positive value linger_time, the
1215 * close() can be blocked for at most linger_time ms. If the graceful
1216 * shutdown phase does not finish during this period, close() will
1217 * return but the graceful shutdown phase continues in the system.
1219 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1221 struct sctp_endpoint *ep;
1222 struct sctp_association *asoc;
1223 struct list_head *pos, *temp;
1225 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1228 sk->sk_shutdown = SHUTDOWN_MASK;
1230 ep = sctp_sk(sk)->ep;
1232 /* Walk all associations on an endpoint. */
1233 list_for_each_safe(pos, temp, &ep->asocs) {
1234 asoc = list_entry(pos, struct sctp_association, asocs);
1236 if (sctp_style(sk, TCP)) {
1237 /* A closed association can still be in the list if
1238 * it belongs to a TCP-style listening socket that is
1239 * not yet accepted. If so, free it. If not, send an
1240 * ABORT or SHUTDOWN based on the linger options.
1242 if (sctp_state(asoc, CLOSED)) {
1243 sctp_unhash_established(asoc);
1244 sctp_association_free(asoc);
1249 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1250 struct sctp_chunk *chunk;
1252 chunk = sctp_make_abort_user(asoc, NULL, 0);
1254 sctp_primitive_ABORT(asoc, chunk);
1256 sctp_primitive_SHUTDOWN(asoc, NULL);
1259 /* Clean up any skbs sitting on the receive queue. */
1260 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1261 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1263 /* On a TCP-style socket, block for at most linger_time if set. */
1264 if (sctp_style(sk, TCP) && timeout)
1265 sctp_wait_for_close(sk, timeout);
1267 /* This will run the backlog queue. */
1268 sctp_release_sock(sk);
1270 /* Supposedly, no process has access to the socket, but
1271 * the net layers still may.
1273 sctp_local_bh_disable();
1274 sctp_bh_lock_sock(sk);
1276 /* Hold the sock, since sk_common_release() will put sock_put()
1277 * and we have just a little more cleanup.
1280 sk_common_release(sk);
1282 sctp_bh_unlock_sock(sk);
1283 sctp_local_bh_enable();
1287 SCTP_DBG_OBJCNT_DEC(sock);
1290 /* Handle EPIPE error. */
1291 static int sctp_error(struct sock *sk, int flags, int err)
1294 err = sock_error(sk) ? : -EPIPE;
1295 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1296 send_sig(SIGPIPE, current, 0);
1300 /* API 3.1.3 sendmsg() - UDP Style Syntax
1302 * An application uses sendmsg() and recvmsg() calls to transmit data to
1303 * and receive data from its peer.
1305 * ssize_t sendmsg(int socket, const struct msghdr *message,
1308 * socket - the socket descriptor of the endpoint.
1309 * message - pointer to the msghdr structure which contains a single
1310 * user message and possibly some ancillary data.
1312 * See Section 5 for complete description of the data
1315 * flags - flags sent or received with the user message, see Section
1316 * 5 for complete description of the flags.
1318 * Note: This function could use a rewrite especially when explicit
1319 * connect support comes in.
1321 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1323 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1325 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1326 struct msghdr *msg, size_t msg_len)
1328 struct sctp_sock *sp;
1329 struct sctp_endpoint *ep;
1330 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1331 struct sctp_transport *transport, *chunk_tp;
1332 struct sctp_chunk *chunk;
1334 struct sockaddr *msg_name = NULL;
1335 struct sctp_sndrcvinfo default_sinfo = { 0 };
1336 struct sctp_sndrcvinfo *sinfo;
1337 struct sctp_initmsg *sinit;
1338 sctp_assoc_t associd = 0;
1339 sctp_cmsgs_t cmsgs = { NULL };
1343 __u16 sinfo_flags = 0;
1344 struct sctp_datamsg *datamsg;
1345 struct list_head *pos;
1346 int msg_flags = msg->msg_flags;
1348 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1355 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1357 /* We cannot send a message over a TCP-style listening socket. */
1358 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1363 /* Parse out the SCTP CMSGs. */
1364 err = sctp_msghdr_parse(msg, &cmsgs);
1367 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1371 /* Fetch the destination address for this packet. This
1372 * address only selects the association--it is not necessarily
1373 * the address we will send to.
1374 * For a peeled-off socket, msg_name is ignored.
1376 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1377 int msg_namelen = msg->msg_namelen;
1379 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1384 if (msg_namelen > sizeof(to))
1385 msg_namelen = sizeof(to);
1386 memcpy(&to, msg->msg_name, msg_namelen);
1387 SCTP_DEBUG_PRINTK("Just memcpy'd. msg_name is "
1389 to.v4.sin_addr.s_addr, to.v4.sin_port);
1391 to.v4.sin_port = ntohs(to.v4.sin_port);
1392 msg_name = msg->msg_name;
1398 /* Did the user specify SNDRCVINFO? */
1400 sinfo_flags = sinfo->sinfo_flags;
1401 associd = sinfo->sinfo_assoc_id;
1404 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1405 msg_len, sinfo_flags);
1407 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1408 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1413 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1414 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1415 * If SCTP_ABORT is set, the message length could be non zero with
1416 * the msg_iov set to the user abort reason.
1418 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1419 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1424 /* If SCTP_ADDR_OVER is set, there must be an address
1425 * specified in msg_name.
1427 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1434 SCTP_DEBUG_PRINTK("About to look up association.\n");
1438 /* If a msg_name has been specified, assume this is to be used. */
1440 /* Look for a matching association on the endpoint. */
1441 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1443 /* If we could not find a matching association on the
1444 * endpoint, make sure that it is not a TCP-style
1445 * socket that already has an association or there is
1446 * no peeled-off association on another socket.
1448 if ((sctp_style(sk, TCP) &&
1449 sctp_sstate(sk, ESTABLISHED)) ||
1450 sctp_endpoint_is_peeled_off(ep, &to)) {
1451 err = -EADDRNOTAVAIL;
1456 asoc = sctp_id2assoc(sk, associd);
1464 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1466 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1467 * socket that has an association in CLOSED state. This can
1468 * happen when an accepted socket has an association that is
1471 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1476 if (sinfo_flags & SCTP_EOF) {
1477 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1479 sctp_primitive_SHUTDOWN(asoc, NULL);
1483 if (sinfo_flags & SCTP_ABORT) {
1484 struct sctp_chunk *chunk;
1486 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1492 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1493 sctp_primitive_ABORT(asoc, chunk);
1499 /* Do we need to create the association? */
1501 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1503 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1508 /* Check for invalid stream against the stream counts,
1509 * either the default or the user specified stream counts.
1512 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1513 /* Check against the defaults. */
1514 if (sinfo->sinfo_stream >=
1515 sp->initmsg.sinit_num_ostreams) {
1520 /* Check against the requested. */
1521 if (sinfo->sinfo_stream >=
1522 sinit->sinit_num_ostreams) {
1530 * API 3.1.2 bind() - UDP Style Syntax
1531 * If a bind() or sctp_bindx() is not called prior to a
1532 * sendmsg() call that initiates a new association, the
1533 * system picks an ephemeral port and will choose an address
1534 * set equivalent to binding with a wildcard address.
1536 if (!ep->base.bind_addr.port) {
1537 if (sctp_autobind(sk)) {
1543 * If an unprivileged user inherits a one-to-many
1544 * style socket with open associations on a privileged
1545 * port, it MAY be permitted to accept new associations,
1546 * but it SHOULD NOT be permitted to open new
1549 if (ep->base.bind_addr.port < PROT_SOCK &&
1550 !capable(CAP_NET_BIND_SERVICE)) {
1556 scope = sctp_scope(&to);
1557 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1564 /* If the SCTP_INIT ancillary data is specified, set all
1565 * the association init values accordingly.
1568 if (sinit->sinit_num_ostreams) {
1569 asoc->c.sinit_num_ostreams =
1570 sinit->sinit_num_ostreams;
1572 if (sinit->sinit_max_instreams) {
1573 asoc->c.sinit_max_instreams =
1574 sinit->sinit_max_instreams;
1576 if (sinit->sinit_max_attempts) {
1577 asoc->max_init_attempts
1578 = sinit->sinit_max_attempts;
1580 if (sinit->sinit_max_init_timeo) {
1581 asoc->max_init_timeo =
1582 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1586 /* Prime the peer's transport structures. */
1587 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1592 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1599 /* ASSERT: we have a valid association at this point. */
1600 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1603 /* If the user didn't specify SNDRCVINFO, make up one with
1606 default_sinfo.sinfo_stream = asoc->default_stream;
1607 default_sinfo.sinfo_flags = asoc->default_flags;
1608 default_sinfo.sinfo_ppid = asoc->default_ppid;
1609 default_sinfo.sinfo_context = asoc->default_context;
1610 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1611 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1612 sinfo = &default_sinfo;
1615 /* API 7.1.7, the sndbuf size per association bounds the
1616 * maximum size of data that can be sent in a single send call.
1618 if (msg_len > sk->sk_sndbuf) {
1623 /* If fragmentation is disabled and the message length exceeds the
1624 * association fragmentation point, return EMSGSIZE. The I-D
1625 * does not specify what this error is, but this looks like
1628 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1634 /* Check for invalid stream. */
1635 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1641 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1642 if (!sctp_wspace(asoc)) {
1643 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1648 /* If an address is passed with the sendto/sendmsg call, it is used
1649 * to override the primary destination address in the TCP model, or
1650 * when SCTP_ADDR_OVER flag is set in the UDP model.
1652 if ((sctp_style(sk, TCP) && msg_name) ||
1653 (sinfo_flags & SCTP_ADDR_OVER)) {
1654 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1662 /* Auto-connect, if we aren't connected already. */
1663 if (sctp_state(asoc, CLOSED)) {
1664 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1667 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1670 /* Break the message into multiple chunks of maximum size. */
1671 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1677 /* Now send the (possibly) fragmented message. */
1678 list_for_each(pos, &datamsg->chunks) {
1679 chunk = list_entry(pos, struct sctp_chunk, frag_list);
1680 sctp_datamsg_track(chunk);
1682 /* Do accounting for the write space. */
1683 sctp_set_owner_w(chunk);
1685 chunk->transport = chunk_tp;
1687 /* Send it to the lower layers. Note: all chunks
1688 * must either fail or succeed. The lower layer
1689 * works that way today. Keep it that way or this
1692 err = sctp_primitive_SEND(asoc, chunk);
1693 /* Did the lower layer accept the chunk? */
1695 sctp_chunk_free(chunk);
1696 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1699 sctp_datamsg_free(datamsg);
1705 /* If we are already past ASSOCIATE, the lower
1706 * layers are responsible for association cleanup.
1712 sctp_association_free(asoc);
1714 sctp_release_sock(sk);
1717 return sctp_error(sk, msg_flags, err);
1724 err = sock_error(sk);
1734 /* This is an extended version of skb_pull() that removes the data from the
1735 * start of a skb even when data is spread across the list of skb's in the
1736 * frag_list. len specifies the total amount of data that needs to be removed.
1737 * when 'len' bytes could be removed from the skb, it returns 0.
1738 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1739 * could not be removed.
1741 static int sctp_skb_pull(struct sk_buff *skb, int len)
1743 struct sk_buff *list;
1744 int skb_len = skb_headlen(skb);
1747 if (len <= skb_len) {
1748 __skb_pull(skb, len);
1752 __skb_pull(skb, skb_len);
1754 for (list = skb_shinfo(skb)->frag_list; list; list = list->next) {
1755 rlen = sctp_skb_pull(list, len);
1756 skb->len -= (len-rlen);
1757 skb->data_len -= (len-rlen);
1768 /* API 3.1.3 recvmsg() - UDP Style Syntax
1770 * ssize_t recvmsg(int socket, struct msghdr *message,
1773 * socket - the socket descriptor of the endpoint.
1774 * message - pointer to the msghdr structure which contains a single
1775 * user message and possibly some ancillary data.
1777 * See Section 5 for complete description of the data
1780 * flags - flags sent or received with the user message, see Section
1781 * 5 for complete description of the flags.
1783 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1785 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1786 struct msghdr *msg, size_t len, int noblock,
1787 int flags, int *addr_len)
1789 struct sctp_ulpevent *event = NULL;
1790 struct sctp_sock *sp = sctp_sk(sk);
1791 struct sk_buff *skb;
1796 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1797 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1798 "len", len, "knoblauch", noblock,
1799 "flags", flags, "addr_len", addr_len);
1803 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1808 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1812 /* Get the total length of the skb including any skb's in the
1821 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1823 event = sctp_skb2event(skb);
1828 sock_recv_timestamp(msg, sk, skb);
1829 if (sctp_ulpevent_is_notification(event)) {
1830 msg->msg_flags |= MSG_NOTIFICATION;
1831 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1833 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1836 /* Check if we allow SCTP_SNDRCVINFO. */
1837 if (sp->subscribe.sctp_data_io_event)
1838 sctp_ulpevent_read_sndrcvinfo(event, msg);
1840 /* FIXME: we should be calling IP/IPv6 layers. */
1841 if (sk->sk_protinfo.af_inet.cmsg_flags)
1842 ip_cmsg_recv(msg, skb);
1847 /* If skb's length exceeds the user's buffer, update the skb and
1848 * push it back to the receive_queue so that the next call to
1849 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1851 if (skb_len > copied) {
1852 msg->msg_flags &= ~MSG_EOR;
1853 if (flags & MSG_PEEK)
1855 sctp_skb_pull(skb, copied);
1856 skb_queue_head(&sk->sk_receive_queue, skb);
1858 /* When only partial message is copied to the user, increase
1859 * rwnd by that amount. If all the data in the skb is read,
1860 * rwnd is updated when the event is freed.
1862 sctp_assoc_rwnd_increase(event->asoc, copied);
1864 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
1865 (event->msg_flags & MSG_EOR))
1866 msg->msg_flags |= MSG_EOR;
1868 msg->msg_flags &= ~MSG_EOR;
1871 if (flags & MSG_PEEK) {
1872 /* Release the skb reference acquired after peeking the skb in
1873 * sctp_skb_recv_datagram().
1877 /* Free the event which includes releasing the reference to
1878 * the owner of the skb, freeing the skb and updating the
1881 sctp_ulpevent_free(event);
1884 sctp_release_sock(sk);
1888 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
1890 * This option is a on/off flag. If enabled no SCTP message
1891 * fragmentation will be performed. Instead if a message being sent
1892 * exceeds the current PMTU size, the message will NOT be sent and
1893 * instead a error will be indicated to the user.
1895 static int sctp_setsockopt_disable_fragments(struct sock *sk,
1896 char __user *optval, int optlen)
1900 if (optlen < sizeof(int))
1903 if (get_user(val, (int __user *)optval))
1906 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
1911 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
1914 if (optlen != sizeof(struct sctp_event_subscribe))
1916 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
1921 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
1923 * This socket option is applicable to the UDP-style socket only. When
1924 * set it will cause associations that are idle for more than the
1925 * specified number of seconds to automatically close. An association
1926 * being idle is defined an association that has NOT sent or received
1927 * user data. The special value of '0' indicates that no automatic
1928 * close of any associations should be performed. The option expects an
1929 * integer defining the number of seconds of idle time before an
1930 * association is closed.
1932 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
1935 struct sctp_sock *sp = sctp_sk(sk);
1937 /* Applicable to UDP-style socket only */
1938 if (sctp_style(sk, TCP))
1940 if (optlen != sizeof(int))
1942 if (copy_from_user(&sp->autoclose, optval, optlen))
1948 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
1950 * Applications can enable or disable heartbeats for any peer address of
1951 * an association, modify an address's heartbeat interval, force a
1952 * heartbeat to be sent immediately, and adjust the address's maximum
1953 * number of retransmissions sent before an address is considered
1954 * unreachable. The following structure is used to access and modify an
1955 * address's parameters:
1957 * struct sctp_paddrparams {
1958 * sctp_assoc_t spp_assoc_id;
1959 * struct sockaddr_storage spp_address;
1960 * uint32_t spp_hbinterval;
1961 * uint16_t spp_pathmaxrxt;
1962 * uint32_t spp_pathmtu;
1963 * uint32_t spp_sackdelay;
1964 * uint32_t spp_flags;
1967 * spp_assoc_id - (one-to-many style socket) This is filled in the
1968 * application, and identifies the association for
1970 * spp_address - This specifies which address is of interest.
1971 * spp_hbinterval - This contains the value of the heartbeat interval,
1972 * in milliseconds. If a value of zero
1973 * is present in this field then no changes are to
1974 * be made to this parameter.
1975 * spp_pathmaxrxt - This contains the maximum number of
1976 * retransmissions before this address shall be
1977 * considered unreachable. If a value of zero
1978 * is present in this field then no changes are to
1979 * be made to this parameter.
1980 * spp_pathmtu - When Path MTU discovery is disabled the value
1981 * specified here will be the "fixed" path mtu.
1982 * Note that if the spp_address field is empty
1983 * then all associations on this address will
1984 * have this fixed path mtu set upon them.
1986 * spp_sackdelay - When delayed sack is enabled, this value specifies
1987 * the number of milliseconds that sacks will be delayed
1988 * for. This value will apply to all addresses of an
1989 * association if the spp_address field is empty. Note
1990 * also, that if delayed sack is enabled and this
1991 * value is set to 0, no change is made to the last
1992 * recorded delayed sack timer value.
1994 * spp_flags - These flags are used to control various features
1995 * on an association. The flag field may contain
1996 * zero or more of the following options.
1998 * SPP_HB_ENABLE - Enable heartbeats on the
1999 * specified address. Note that if the address
2000 * field is empty all addresses for the association
2001 * have heartbeats enabled upon them.
2003 * SPP_HB_DISABLE - Disable heartbeats on the
2004 * speicifed address. Note that if the address
2005 * field is empty all addresses for the association
2006 * will have their heartbeats disabled. Note also
2007 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2008 * mutually exclusive, only one of these two should
2009 * be specified. Enabling both fields will have
2010 * undetermined results.
2012 * SPP_HB_DEMAND - Request a user initiated heartbeat
2013 * to be made immediately.
2015 * SPP_PMTUD_ENABLE - This field will enable PMTU
2016 * discovery upon the specified address. Note that
2017 * if the address feild is empty then all addresses
2018 * on the association are effected.
2020 * SPP_PMTUD_DISABLE - This field will disable PMTU
2021 * discovery upon the specified address. Note that
2022 * if the address feild is empty then all addresses
2023 * on the association are effected. Not also that
2024 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2025 * exclusive. Enabling both will have undetermined
2028 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2029 * on delayed sack. The time specified in spp_sackdelay
2030 * is used to specify the sack delay for this address. Note
2031 * that if spp_address is empty then all addresses will
2032 * enable delayed sack and take on the sack delay
2033 * value specified in spp_sackdelay.
2034 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2035 * off delayed sack. If the spp_address field is blank then
2036 * delayed sack is disabled for the entire association. Note
2037 * also that this field is mutually exclusive to
2038 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2041 int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2042 struct sctp_transport *trans,
2043 struct sctp_association *asoc,
2044 struct sctp_sock *sp,
2047 int sackdelay_change)
2051 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2052 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2057 if (params->spp_hbinterval) {
2059 trans->hbinterval = msecs_to_jiffies(params->spp_hbinterval);
2061 asoc->hbinterval = msecs_to_jiffies(params->spp_hbinterval);
2063 sp->hbinterval = params->spp_hbinterval;
2069 trans->param_flags =
2070 (trans->param_flags & ~SPP_HB) | hb_change;
2073 (asoc->param_flags & ~SPP_HB) | hb_change;
2076 (sp->param_flags & ~SPP_HB) | hb_change;
2080 if (params->spp_pathmtu) {
2082 trans->pathmtu = params->spp_pathmtu;
2083 sctp_assoc_sync_pmtu(asoc);
2085 asoc->pathmtu = params->spp_pathmtu;
2086 sctp_frag_point(sp, params->spp_pathmtu);
2088 sp->pathmtu = params->spp_pathmtu;
2094 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2095 (params->spp_flags & SPP_PMTUD_ENABLE);
2096 trans->param_flags =
2097 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2099 sctp_transport_pmtu(trans);
2100 sctp_assoc_sync_pmtu(asoc);
2104 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2107 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2111 if (params->spp_sackdelay) {
2114 msecs_to_jiffies(params->spp_sackdelay);
2117 msecs_to_jiffies(params->spp_sackdelay);
2119 sp->sackdelay = params->spp_sackdelay;
2123 if (sackdelay_change) {
2125 trans->param_flags =
2126 (trans->param_flags & ~SPP_SACKDELAY) |
2130 (asoc->param_flags & ~SPP_SACKDELAY) |
2134 (sp->param_flags & ~SPP_SACKDELAY) |
2139 if (params->spp_pathmaxrxt) {
2141 trans->pathmaxrxt = params->spp_pathmaxrxt;
2143 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2145 sp->pathmaxrxt = params->spp_pathmaxrxt;
2152 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2153 char __user *optval, int optlen)
2155 struct sctp_paddrparams params;
2156 struct sctp_transport *trans = NULL;
2157 struct sctp_association *asoc = NULL;
2158 struct sctp_sock *sp = sctp_sk(sk);
2160 int hb_change, pmtud_change, sackdelay_change;
2162 if (optlen != sizeof(struct sctp_paddrparams))
2165 if (copy_from_user(¶ms, optval, optlen))
2168 /* Validate flags and value parameters. */
2169 hb_change = params.spp_flags & SPP_HB;
2170 pmtud_change = params.spp_flags & SPP_PMTUD;
2171 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2173 if (hb_change == SPP_HB ||
2174 pmtud_change == SPP_PMTUD ||
2175 sackdelay_change == SPP_SACKDELAY ||
2176 params.spp_sackdelay > 500 ||
2178 && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2181 /* If an address other than INADDR_ANY is specified, and
2182 * no transport is found, then the request is invalid.
2184 if (!sctp_is_any(( union sctp_addr *)¶ms.spp_address)) {
2185 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2186 params.spp_assoc_id);
2191 /* Get association, if assoc_id != 0 and the socket is a one
2192 * to many style socket, and an association was not found, then
2193 * the id was invalid.
2195 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2196 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2199 /* Heartbeat demand can only be sent on a transport or
2200 * association, but not a socket.
2202 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2205 /* Process parameters. */
2206 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2207 hb_change, pmtud_change,
2213 /* If changes are for association, also apply parameters to each
2216 if (!trans && asoc) {
2217 struct list_head *pos;
2219 list_for_each(pos, &asoc->peer.transport_addr_list) {
2220 trans = list_entry(pos, struct sctp_transport,
2222 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2223 hb_change, pmtud_change,
2231 /* 7.1.24. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
2233 * This options will get or set the delayed ack timer. The time is set
2234 * in milliseconds. If the assoc_id is 0, then this sets or gets the
2235 * endpoints default delayed ack timer value. If the assoc_id field is
2236 * non-zero, then the set or get effects the specified association.
2238 * struct sctp_assoc_value {
2239 * sctp_assoc_t assoc_id;
2240 * uint32_t assoc_value;
2243 * assoc_id - This parameter, indicates which association the
2244 * user is preforming an action upon. Note that if
2245 * this field's value is zero then the endpoints
2246 * default value is changed (effecting future
2247 * associations only).
2249 * assoc_value - This parameter contains the number of milliseconds
2250 * that the user is requesting the delayed ACK timer
2251 * be set to. Note that this value is defined in
2252 * the standard to be between 200 and 500 milliseconds.
2254 * Note: a value of zero will leave the value alone,
2255 * but disable SACK delay. A non-zero value will also
2256 * enable SACK delay.
2259 static int sctp_setsockopt_delayed_ack_time(struct sock *sk,
2260 char __user *optval, int optlen)
2262 struct sctp_assoc_value params;
2263 struct sctp_transport *trans = NULL;
2264 struct sctp_association *asoc = NULL;
2265 struct sctp_sock *sp = sctp_sk(sk);
2267 if (optlen != sizeof(struct sctp_assoc_value))
2270 if (copy_from_user(¶ms, optval, optlen))
2273 /* Validate value parameter. */
2274 if (params.assoc_value > 500)
2277 /* Get association, if assoc_id != 0 and the socket is a one
2278 * to many style socket, and an association was not found, then
2279 * the id was invalid.
2281 asoc = sctp_id2assoc(sk, params.assoc_id);
2282 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2285 if (params.assoc_value) {
2288 msecs_to_jiffies(params.assoc_value);
2290 (asoc->param_flags & ~SPP_SACKDELAY) |
2291 SPP_SACKDELAY_ENABLE;
2293 sp->sackdelay = params.assoc_value;
2295 (sp->param_flags & ~SPP_SACKDELAY) |
2296 SPP_SACKDELAY_ENABLE;
2301 (asoc->param_flags & ~SPP_SACKDELAY) |
2302 SPP_SACKDELAY_DISABLE;
2305 (sp->param_flags & ~SPP_SACKDELAY) |
2306 SPP_SACKDELAY_DISABLE;
2310 /* If change is for association, also apply to each transport. */
2312 struct list_head *pos;
2314 list_for_each(pos, &asoc->peer.transport_addr_list) {
2315 trans = list_entry(pos, struct sctp_transport,
2317 if (params.assoc_value) {
2319 msecs_to_jiffies(params.assoc_value);
2320 trans->param_flags =
2321 (trans->param_flags & ~SPP_SACKDELAY) |
2322 SPP_SACKDELAY_ENABLE;
2324 trans->param_flags =
2325 (trans->param_flags & ~SPP_SACKDELAY) |
2326 SPP_SACKDELAY_DISABLE;
2334 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2336 * Applications can specify protocol parameters for the default association
2337 * initialization. The option name argument to setsockopt() and getsockopt()
2340 * Setting initialization parameters is effective only on an unconnected
2341 * socket (for UDP-style sockets only future associations are effected
2342 * by the change). With TCP-style sockets, this option is inherited by
2343 * sockets derived from a listener socket.
2345 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen)
2347 struct sctp_initmsg sinit;
2348 struct sctp_sock *sp = sctp_sk(sk);
2350 if (optlen != sizeof(struct sctp_initmsg))
2352 if (copy_from_user(&sinit, optval, optlen))
2355 if (sinit.sinit_num_ostreams)
2356 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2357 if (sinit.sinit_max_instreams)
2358 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2359 if (sinit.sinit_max_attempts)
2360 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2361 if (sinit.sinit_max_init_timeo)
2362 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2368 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2370 * Applications that wish to use the sendto() system call may wish to
2371 * specify a default set of parameters that would normally be supplied
2372 * through the inclusion of ancillary data. This socket option allows
2373 * such an application to set the default sctp_sndrcvinfo structure.
2374 * The application that wishes to use this socket option simply passes
2375 * in to this call the sctp_sndrcvinfo structure defined in Section
2376 * 5.2.2) The input parameters accepted by this call include
2377 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2378 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2379 * to this call if the caller is using the UDP model.
2381 static int sctp_setsockopt_default_send_param(struct sock *sk,
2382 char __user *optval, int optlen)
2384 struct sctp_sndrcvinfo info;
2385 struct sctp_association *asoc;
2386 struct sctp_sock *sp = sctp_sk(sk);
2388 if (optlen != sizeof(struct sctp_sndrcvinfo))
2390 if (copy_from_user(&info, optval, optlen))
2393 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2394 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2398 asoc->default_stream = info.sinfo_stream;
2399 asoc->default_flags = info.sinfo_flags;
2400 asoc->default_ppid = info.sinfo_ppid;
2401 asoc->default_context = info.sinfo_context;
2402 asoc->default_timetolive = info.sinfo_timetolive;
2404 sp->default_stream = info.sinfo_stream;
2405 sp->default_flags = info.sinfo_flags;
2406 sp->default_ppid = info.sinfo_ppid;
2407 sp->default_context = info.sinfo_context;
2408 sp->default_timetolive = info.sinfo_timetolive;
2414 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2416 * Requests that the local SCTP stack use the enclosed peer address as
2417 * the association primary. The enclosed address must be one of the
2418 * association peer's addresses.
2420 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2423 struct sctp_prim prim;
2424 struct sctp_transport *trans;
2426 if (optlen != sizeof(struct sctp_prim))
2429 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2432 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2436 sctp_assoc_set_primary(trans->asoc, trans);
2442 * 7.1.5 SCTP_NODELAY
2444 * Turn on/off any Nagle-like algorithm. This means that packets are
2445 * generally sent as soon as possible and no unnecessary delays are
2446 * introduced, at the cost of more packets in the network. Expects an
2447 * integer boolean flag.
2449 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2454 if (optlen < sizeof(int))
2456 if (get_user(val, (int __user *)optval))
2459 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2465 * 7.1.1 SCTP_RTOINFO
2467 * The protocol parameters used to initialize and bound retransmission
2468 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2469 * and modify these parameters.
2470 * All parameters are time values, in milliseconds. A value of 0, when
2471 * modifying the parameters, indicates that the current value should not
2475 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) {
2476 struct sctp_rtoinfo rtoinfo;
2477 struct sctp_association *asoc;
2479 if (optlen != sizeof (struct sctp_rtoinfo))
2482 if (copy_from_user(&rtoinfo, optval, optlen))
2485 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2487 /* Set the values to the specific association */
2488 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2492 if (rtoinfo.srto_initial != 0)
2494 msecs_to_jiffies(rtoinfo.srto_initial);
2495 if (rtoinfo.srto_max != 0)
2496 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2497 if (rtoinfo.srto_min != 0)
2498 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2500 /* If there is no association or the association-id = 0
2501 * set the values to the endpoint.
2503 struct sctp_sock *sp = sctp_sk(sk);
2505 if (rtoinfo.srto_initial != 0)
2506 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2507 if (rtoinfo.srto_max != 0)
2508 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2509 if (rtoinfo.srto_min != 0)
2510 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2518 * 7.1.2 SCTP_ASSOCINFO
2520 * This option is used to tune the the maximum retransmission attempts
2521 * of the association.
2522 * Returns an error if the new association retransmission value is
2523 * greater than the sum of the retransmission value of the peer.
2524 * See [SCTP] for more information.
2527 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen)
2530 struct sctp_assocparams assocparams;
2531 struct sctp_association *asoc;
2533 if (optlen != sizeof(struct sctp_assocparams))
2535 if (copy_from_user(&assocparams, optval, optlen))
2538 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2540 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2543 /* Set the values to the specific association */
2545 if (assocparams.sasoc_asocmaxrxt != 0) {
2548 struct list_head *pos;
2549 struct sctp_transport *peer_addr;
2551 list_for_each(pos, &asoc->peer.transport_addr_list) {
2552 peer_addr = list_entry(pos,
2553 struct sctp_transport,
2555 path_sum += peer_addr->pathmaxrxt;
2559 /* Only validate asocmaxrxt if we have more then
2560 * one path/transport. We do this because path
2561 * retransmissions are only counted when we have more
2565 assocparams.sasoc_asocmaxrxt > path_sum)
2568 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2571 if (assocparams.sasoc_cookie_life != 0) {
2572 asoc->cookie_life.tv_sec =
2573 assocparams.sasoc_cookie_life / 1000;
2574 asoc->cookie_life.tv_usec =
2575 (assocparams.sasoc_cookie_life % 1000)
2579 /* Set the values to the endpoint */
2580 struct sctp_sock *sp = sctp_sk(sk);
2582 if (assocparams.sasoc_asocmaxrxt != 0)
2583 sp->assocparams.sasoc_asocmaxrxt =
2584 assocparams.sasoc_asocmaxrxt;
2585 if (assocparams.sasoc_cookie_life != 0)
2586 sp->assocparams.sasoc_cookie_life =
2587 assocparams.sasoc_cookie_life;
2593 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2595 * This socket option is a boolean flag which turns on or off mapped V4
2596 * addresses. If this option is turned on and the socket is type
2597 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2598 * If this option is turned off, then no mapping will be done of V4
2599 * addresses and a user will receive both PF_INET6 and PF_INET type
2600 * addresses on the socket.
2602 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen)
2605 struct sctp_sock *sp = sctp_sk(sk);
2607 if (optlen < sizeof(int))
2609 if (get_user(val, (int __user *)optval))
2620 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
2622 * This socket option specifies the maximum size to put in any outgoing
2623 * SCTP chunk. If a message is larger than this size it will be
2624 * fragmented by SCTP into the specified size. Note that the underlying
2625 * SCTP implementation may fragment into smaller sized chunks when the
2626 * PMTU of the underlying association is smaller than the value set by
2629 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen)
2631 struct sctp_association *asoc;
2632 struct list_head *pos;
2633 struct sctp_sock *sp = sctp_sk(sk);
2636 if (optlen < sizeof(int))
2638 if (get_user(val, (int __user *)optval))
2640 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2642 sp->user_frag = val;
2644 /* Update the frag_point of the existing associations. */
2645 list_for_each(pos, &(sp->ep->asocs)) {
2646 asoc = list_entry(pos, struct sctp_association, asocs);
2647 asoc->frag_point = sctp_frag_point(sp, asoc->pathmtu);
2655 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2657 * Requests that the peer mark the enclosed address as the association
2658 * primary. The enclosed address must be one of the association's
2659 * locally bound addresses. The following structure is used to make a
2660 * set primary request:
2662 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2665 struct sctp_sock *sp;
2666 struct sctp_endpoint *ep;
2667 struct sctp_association *asoc = NULL;
2668 struct sctp_setpeerprim prim;
2669 struct sctp_chunk *chunk;
2675 if (!sctp_addip_enable)
2678 if (optlen != sizeof(struct sctp_setpeerprim))
2681 if (copy_from_user(&prim, optval, optlen))
2684 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2688 if (!asoc->peer.asconf_capable)
2691 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2694 if (!sctp_state(asoc, ESTABLISHED))
2697 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2698 return -EADDRNOTAVAIL;
2700 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2701 chunk = sctp_make_asconf_set_prim(asoc,
2702 (union sctp_addr *)&prim.sspp_addr);
2706 err = sctp_send_asconf(asoc, chunk);
2708 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2713 static int sctp_setsockopt_adaption_layer(struct sock *sk, char __user *optval,
2716 struct sctp_setadaption adaption;
2718 if (optlen != sizeof(struct sctp_setadaption))
2720 if (copy_from_user(&adaption, optval, optlen))
2723 sctp_sk(sk)->adaption_ind = adaption.ssb_adaption_ind;
2728 /* API 6.2 setsockopt(), getsockopt()
2730 * Applications use setsockopt() and getsockopt() to set or retrieve
2731 * socket options. Socket options are used to change the default
2732 * behavior of sockets calls. They are described in Section 7.
2736 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
2737 * int __user *optlen);
2738 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
2741 * sd - the socket descript.
2742 * level - set to IPPROTO_SCTP for all SCTP options.
2743 * optname - the option name.
2744 * optval - the buffer to store the value of the option.
2745 * optlen - the size of the buffer.
2747 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
2748 char __user *optval, int optlen)
2752 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
2755 /* I can hardly begin to describe how wrong this is. This is
2756 * so broken as to be worse than useless. The API draft
2757 * REALLY is NOT helpful here... I am not convinced that the
2758 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
2759 * are at all well-founded.
2761 if (level != SOL_SCTP) {
2762 struct sctp_af *af = sctp_sk(sk)->pf->af;
2763 retval = af->setsockopt(sk, level, optname, optval, optlen);
2770 case SCTP_SOCKOPT_BINDX_ADD:
2771 /* 'optlen' is the size of the addresses buffer. */
2772 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2773 optlen, SCTP_BINDX_ADD_ADDR);
2776 case SCTP_SOCKOPT_BINDX_REM:
2777 /* 'optlen' is the size of the addresses buffer. */
2778 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2779 optlen, SCTP_BINDX_REM_ADDR);
2782 case SCTP_SOCKOPT_CONNECTX:
2783 /* 'optlen' is the size of the addresses buffer. */
2784 retval = sctp_setsockopt_connectx(sk, (struct sockaddr __user *)optval,
2788 case SCTP_DISABLE_FRAGMENTS:
2789 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
2793 retval = sctp_setsockopt_events(sk, optval, optlen);
2796 case SCTP_AUTOCLOSE:
2797 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
2800 case SCTP_PEER_ADDR_PARAMS:
2801 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
2804 case SCTP_DELAYED_ACK_TIME:
2805 retval = sctp_setsockopt_delayed_ack_time(sk, optval, optlen);
2809 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
2811 case SCTP_DEFAULT_SEND_PARAM:
2812 retval = sctp_setsockopt_default_send_param(sk, optval,
2815 case SCTP_PRIMARY_ADDR:
2816 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
2818 case SCTP_SET_PEER_PRIMARY_ADDR:
2819 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
2822 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
2825 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
2827 case SCTP_ASSOCINFO:
2828 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
2830 case SCTP_I_WANT_MAPPED_V4_ADDR:
2831 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
2834 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
2836 case SCTP_ADAPTION_LAYER:
2837 retval = sctp_setsockopt_adaption_layer(sk, optval, optlen);
2841 retval = -ENOPROTOOPT;
2845 sctp_release_sock(sk);
2851 /* API 3.1.6 connect() - UDP Style Syntax
2853 * An application may use the connect() call in the UDP model to initiate an
2854 * association without sending data.
2858 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
2860 * sd: the socket descriptor to have a new association added to.
2862 * nam: the address structure (either struct sockaddr_in or struct
2863 * sockaddr_in6 defined in RFC2553 [7]).
2865 * len: the size of the address.
2867 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
2875 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
2876 __FUNCTION__, sk, addr, addr_len);
2878 /* Validate addr_len before calling common connect/connectx routine. */
2879 af = sctp_get_af_specific(addr->sa_family);
2880 if (!af || addr_len < af->sockaddr_len) {
2883 /* Pass correct addr len to common routine (so it knows there
2884 * is only one address being passed.
2886 err = __sctp_connect(sk, addr, af->sockaddr_len);
2889 sctp_release_sock(sk);
2893 /* FIXME: Write comments. */
2894 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
2896 return -EOPNOTSUPP; /* STUB */
2899 /* 4.1.4 accept() - TCP Style Syntax
2901 * Applications use accept() call to remove an established SCTP
2902 * association from the accept queue of the endpoint. A new socket
2903 * descriptor will be returned from accept() to represent the newly
2904 * formed association.
2906 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
2908 struct sctp_sock *sp;
2909 struct sctp_endpoint *ep;
2910 struct sock *newsk = NULL;
2911 struct sctp_association *asoc;
2920 if (!sctp_style(sk, TCP)) {
2921 error = -EOPNOTSUPP;
2925 if (!sctp_sstate(sk, LISTENING)) {
2930 timeo = sock_rcvtimeo(sk, sk->sk_socket->file->f_flags & O_NONBLOCK);
2932 error = sctp_wait_for_accept(sk, timeo);
2936 /* We treat the list of associations on the endpoint as the accept
2937 * queue and pick the first association on the list.
2939 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
2941 newsk = sp->pf->create_accept_sk(sk, asoc);
2947 /* Populate the fields of the newsk from the oldsk and migrate the
2948 * asoc to the newsk.
2950 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
2953 sctp_release_sock(sk);
2958 /* The SCTP ioctl handler. */
2959 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
2961 return -ENOIOCTLCMD;
2964 /* This is the function which gets called during socket creation to
2965 * initialized the SCTP-specific portion of the sock.
2966 * The sock structure should already be zero-filled memory.
2968 SCTP_STATIC int sctp_init_sock(struct sock *sk)
2970 struct sctp_endpoint *ep;
2971 struct sctp_sock *sp;
2973 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
2977 /* Initialize the SCTP per socket area. */
2978 switch (sk->sk_type) {
2979 case SOCK_SEQPACKET:
2980 sp->type = SCTP_SOCKET_UDP;
2983 sp->type = SCTP_SOCKET_TCP;
2986 return -ESOCKTNOSUPPORT;
2989 /* Initialize default send parameters. These parameters can be
2990 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
2992 sp->default_stream = 0;
2993 sp->default_ppid = 0;
2994 sp->default_flags = 0;
2995 sp->default_context = 0;
2996 sp->default_timetolive = 0;
2998 /* Initialize default setup parameters. These parameters
2999 * can be modified with the SCTP_INITMSG socket option or
3000 * overridden by the SCTP_INIT CMSG.
3002 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3003 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3004 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3005 sp->initmsg.sinit_max_init_timeo = jiffies_to_msecs(sctp_rto_max);
3007 /* Initialize default RTO related parameters. These parameters can
3008 * be modified for with the SCTP_RTOINFO socket option.
3010 sp->rtoinfo.srto_initial = jiffies_to_msecs(sctp_rto_initial);
3011 sp->rtoinfo.srto_max = jiffies_to_msecs(sctp_rto_max);
3012 sp->rtoinfo.srto_min = jiffies_to_msecs(sctp_rto_min);
3014 /* Initialize default association related parameters. These parameters
3015 * can be modified with the SCTP_ASSOCINFO socket option.
3017 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3018 sp->assocparams.sasoc_number_peer_destinations = 0;
3019 sp->assocparams.sasoc_peer_rwnd = 0;
3020 sp->assocparams.sasoc_local_rwnd = 0;
3021 sp->assocparams.sasoc_cookie_life =
3022 jiffies_to_msecs(sctp_valid_cookie_life);
3024 /* Initialize default event subscriptions. By default, all the
3027 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3029 /* Default Peer Address Parameters. These defaults can
3030 * be modified via SCTP_PEER_ADDR_PARAMS
3032 sp->hbinterval = jiffies_to_msecs(sctp_hb_interval);
3033 sp->pathmaxrxt = sctp_max_retrans_path;
3034 sp->pathmtu = 0; // allow default discovery
3035 sp->sackdelay = jiffies_to_msecs(sctp_sack_timeout);
3036 sp->param_flags = SPP_HB_ENABLE |
3038 SPP_SACKDELAY_ENABLE;
3040 /* If enabled no SCTP message fragmentation will be performed.
3041 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3043 sp->disable_fragments = 0;
3045 /* Turn on/off any Nagle-like algorithm. */
3048 /* Enable by default. */
3051 /* Auto-close idle associations after the configured
3052 * number of seconds. A value of 0 disables this
3053 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3054 * for UDP-style sockets only.
3058 /* User specified fragmentation limit. */
3061 sp->adaption_ind = 0;
3063 sp->pf = sctp_get_pf_specific(sk->sk_family);
3065 /* Control variables for partial data delivery. */
3067 skb_queue_head_init(&sp->pd_lobby);
3069 /* Create a per socket endpoint structure. Even if we
3070 * change the data structure relationships, this may still
3071 * be useful for storing pre-connect address information.
3073 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3080 SCTP_DBG_OBJCNT_INC(sock);
3084 /* Cleanup any SCTP per socket resources. */
3085 SCTP_STATIC int sctp_destroy_sock(struct sock *sk)
3087 struct sctp_endpoint *ep;
3089 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3091 /* Release our hold on the endpoint. */
3092 ep = sctp_sk(sk)->ep;
3093 sctp_endpoint_free(ep);
3098 /* API 4.1.7 shutdown() - TCP Style Syntax
3099 * int shutdown(int socket, int how);
3101 * sd - the socket descriptor of the association to be closed.
3102 * how - Specifies the type of shutdown. The values are
3105 * Disables further receive operations. No SCTP
3106 * protocol action is taken.
3108 * Disables further send operations, and initiates
3109 * the SCTP shutdown sequence.
3111 * Disables further send and receive operations
3112 * and initiates the SCTP shutdown sequence.
3114 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3116 struct sctp_endpoint *ep;
3117 struct sctp_association *asoc;
3119 if (!sctp_style(sk, TCP))
3122 if (how & SEND_SHUTDOWN) {
3123 ep = sctp_sk(sk)->ep;
3124 if (!list_empty(&ep->asocs)) {
3125 asoc = list_entry(ep->asocs.next,
3126 struct sctp_association, asocs);
3127 sctp_primitive_SHUTDOWN(asoc, NULL);
3132 /* 7.2.1 Association Status (SCTP_STATUS)
3134 * Applications can retrieve current status information about an
3135 * association, including association state, peer receiver window size,
3136 * number of unacked data chunks, and number of data chunks pending
3137 * receipt. This information is read-only.
3139 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3140 char __user *optval,
3143 struct sctp_status status;
3144 struct sctp_association *asoc = NULL;
3145 struct sctp_transport *transport;
3146 sctp_assoc_t associd;
3149 if (len != sizeof(status)) {
3154 if (copy_from_user(&status, optval, sizeof(status))) {
3159 associd = status.sstat_assoc_id;
3160 asoc = sctp_id2assoc(sk, associd);
3166 transport = asoc->peer.primary_path;
3168 status.sstat_assoc_id = sctp_assoc2id(asoc);
3169 status.sstat_state = asoc->state;
3170 status.sstat_rwnd = asoc->peer.rwnd;
3171 status.sstat_unackdata = asoc->unack_data;
3173 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3174 status.sstat_instrms = asoc->c.sinit_max_instreams;
3175 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3176 status.sstat_fragmentation_point = asoc->frag_point;
3177 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3178 memcpy(&status.sstat_primary.spinfo_address,
3179 &(transport->ipaddr), sizeof(union sctp_addr));
3180 /* Map ipv4 address into v4-mapped-on-v6 address. */
3181 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3182 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3183 status.sstat_primary.spinfo_state = transport->state;
3184 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3185 status.sstat_primary.spinfo_srtt = transport->srtt;
3186 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3187 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3189 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3190 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3192 if (put_user(len, optlen)) {
3197 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3198 len, status.sstat_state, status.sstat_rwnd,
3199 status.sstat_assoc_id);
3201 if (copy_to_user(optval, &status, len)) {
3211 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3213 * Applications can retrieve information about a specific peer address
3214 * of an association, including its reachability state, congestion
3215 * window, and retransmission timer values. This information is
3218 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3219 char __user *optval,
3222 struct sctp_paddrinfo pinfo;
3223 struct sctp_transport *transport;
3226 if (len != sizeof(pinfo)) {
3231 if (copy_from_user(&pinfo, optval, sizeof(pinfo))) {
3236 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3237 pinfo.spinfo_assoc_id);
3241 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3242 pinfo.spinfo_state = transport->state;
3243 pinfo.spinfo_cwnd = transport->cwnd;
3244 pinfo.spinfo_srtt = transport->srtt;
3245 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3246 pinfo.spinfo_mtu = transport->pathmtu;
3248 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3249 pinfo.spinfo_state = SCTP_ACTIVE;
3251 if (put_user(len, optlen)) {
3256 if (copy_to_user(optval, &pinfo, len)) {
3265 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3267 * This option is a on/off flag. If enabled no SCTP message
3268 * fragmentation will be performed. Instead if a message being sent
3269 * exceeds the current PMTU size, the message will NOT be sent and
3270 * instead a error will be indicated to the user.
3272 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3273 char __user *optval, int __user *optlen)
3277 if (len < sizeof(int))
3281 val = (sctp_sk(sk)->disable_fragments == 1);
3282 if (put_user(len, optlen))
3284 if (copy_to_user(optval, &val, len))
3289 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3291 * This socket option is used to specify various notifications and
3292 * ancillary data the user wishes to receive.
3294 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3297 if (len != sizeof(struct sctp_event_subscribe))
3299 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3304 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3306 * This socket option is applicable to the UDP-style socket only. When
3307 * set it will cause associations that are idle for more than the
3308 * specified number of seconds to automatically close. An association
3309 * being idle is defined an association that has NOT sent or received
3310 * user data. The special value of '0' indicates that no automatic
3311 * close of any associations should be performed. The option expects an
3312 * integer defining the number of seconds of idle time before an
3313 * association is closed.
3315 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3317 /* Applicable to UDP-style socket only */
3318 if (sctp_style(sk, TCP))
3320 if (len != sizeof(int))
3322 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, len))
3327 /* Helper routine to branch off an association to a new socket. */
3328 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3329 struct socket **sockp)
3331 struct sock *sk = asoc->base.sk;
3332 struct socket *sock;
3335 /* An association cannot be branched off from an already peeled-off
3336 * socket, nor is this supported for tcp style sockets.
3338 if (!sctp_style(sk, UDP))
3341 /* Create a new socket. */
3342 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3346 /* Populate the fields of the newsk from the oldsk and migrate the
3347 * asoc to the newsk.
3349 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
3355 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
3357 sctp_peeloff_arg_t peeloff;
3358 struct socket *newsock;
3360 struct sctp_association *asoc;
3362 if (len != sizeof(sctp_peeloff_arg_t))
3364 if (copy_from_user(&peeloff, optval, len))
3367 asoc = sctp_id2assoc(sk, peeloff.associd);
3373 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __FUNCTION__, sk, asoc);
3375 retval = sctp_do_peeloff(asoc, &newsock);
3379 /* Map the socket to an unused fd that can be returned to the user. */
3380 retval = sock_map_fd(newsock);
3382 sock_release(newsock);
3386 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3387 __FUNCTION__, sk, asoc, newsock->sk, retval);
3389 /* Return the fd mapped to the new socket. */
3390 peeloff.sd = retval;
3391 if (copy_to_user(optval, &peeloff, len))
3398 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
3400 * Applications can enable or disable heartbeats for any peer address of
3401 * an association, modify an address's heartbeat interval, force a
3402 * heartbeat to be sent immediately, and adjust the address's maximum
3403 * number of retransmissions sent before an address is considered
3404 * unreachable. The following structure is used to access and modify an
3405 * address's parameters:
3407 * struct sctp_paddrparams {
3408 * sctp_assoc_t spp_assoc_id;
3409 * struct sockaddr_storage spp_address;
3410 * uint32_t spp_hbinterval;
3411 * uint16_t spp_pathmaxrxt;
3412 * uint32_t spp_pathmtu;
3413 * uint32_t spp_sackdelay;
3414 * uint32_t spp_flags;
3417 * spp_assoc_id - (one-to-many style socket) This is filled in the
3418 * application, and identifies the association for
3420 * spp_address - This specifies which address is of interest.
3421 * spp_hbinterval - This contains the value of the heartbeat interval,
3422 * in milliseconds. If a value of zero
3423 * is present in this field then no changes are to
3424 * be made to this parameter.
3425 * spp_pathmaxrxt - This contains the maximum number of
3426 * retransmissions before this address shall be
3427 * considered unreachable. If a value of zero
3428 * is present in this field then no changes are to
3429 * be made to this parameter.
3430 * spp_pathmtu - When Path MTU discovery is disabled the value
3431 * specified here will be the "fixed" path mtu.
3432 * Note that if the spp_address field is empty
3433 * then all associations on this address will
3434 * have this fixed path mtu set upon them.
3436 * spp_sackdelay - When delayed sack is enabled, this value specifies
3437 * the number of milliseconds that sacks will be delayed
3438 * for. This value will apply to all addresses of an
3439 * association if the spp_address field is empty. Note
3440 * also, that if delayed sack is enabled and this
3441 * value is set to 0, no change is made to the last
3442 * recorded delayed sack timer value.
3444 * spp_flags - These flags are used to control various features
3445 * on an association. The flag field may contain
3446 * zero or more of the following options.
3448 * SPP_HB_ENABLE - Enable heartbeats on the
3449 * specified address. Note that if the address
3450 * field is empty all addresses for the association
3451 * have heartbeats enabled upon them.
3453 * SPP_HB_DISABLE - Disable heartbeats on the
3454 * speicifed address. Note that if the address
3455 * field is empty all addresses for the association
3456 * will have their heartbeats disabled. Note also
3457 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
3458 * mutually exclusive, only one of these two should
3459 * be specified. Enabling both fields will have
3460 * undetermined results.
3462 * SPP_HB_DEMAND - Request a user initiated heartbeat
3463 * to be made immediately.
3465 * SPP_PMTUD_ENABLE - This field will enable PMTU
3466 * discovery upon the specified address. Note that
3467 * if the address feild is empty then all addresses
3468 * on the association are effected.
3470 * SPP_PMTUD_DISABLE - This field will disable PMTU
3471 * discovery upon the specified address. Note that
3472 * if the address feild is empty then all addresses
3473 * on the association are effected. Not also that
3474 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
3475 * exclusive. Enabling both will have undetermined
3478 * SPP_SACKDELAY_ENABLE - Setting this flag turns
3479 * on delayed sack. The time specified in spp_sackdelay
3480 * is used to specify the sack delay for this address. Note
3481 * that if spp_address is empty then all addresses will
3482 * enable delayed sack and take on the sack delay
3483 * value specified in spp_sackdelay.
3484 * SPP_SACKDELAY_DISABLE - Setting this flag turns
3485 * off delayed sack. If the spp_address field is blank then
3486 * delayed sack is disabled for the entire association. Note
3487 * also that this field is mutually exclusive to
3488 * SPP_SACKDELAY_ENABLE, setting both will have undefined
3491 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
3492 char __user *optval, int __user *optlen)
3494 struct sctp_paddrparams params;
3495 struct sctp_transport *trans = NULL;
3496 struct sctp_association *asoc = NULL;
3497 struct sctp_sock *sp = sctp_sk(sk);
3499 if (len != sizeof(struct sctp_paddrparams))
3502 if (copy_from_user(¶ms, optval, len))
3505 /* If an address other than INADDR_ANY is specified, and
3506 * no transport is found, then the request is invalid.
3508 if (!sctp_is_any(( union sctp_addr *)¶ms.spp_address)) {
3509 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
3510 params.spp_assoc_id);
3512 SCTP_DEBUG_PRINTK("Failed no transport\n");
3517 /* Get association, if assoc_id != 0 and the socket is a one
3518 * to many style socket, and an association was not found, then
3519 * the id was invalid.
3521 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
3522 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
3523 SCTP_DEBUG_PRINTK("Failed no association\n");
3528 /* Fetch transport values. */
3529 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
3530 params.spp_pathmtu = trans->pathmtu;
3531 params.spp_pathmaxrxt = trans->pathmaxrxt;
3532 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
3534 /*draft-11 doesn't say what to return in spp_flags*/
3535 params.spp_flags = trans->param_flags;
3537 /* Fetch association values. */
3538 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
3539 params.spp_pathmtu = asoc->pathmtu;
3540 params.spp_pathmaxrxt = asoc->pathmaxrxt;
3541 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
3543 /*draft-11 doesn't say what to return in spp_flags*/
3544 params.spp_flags = asoc->param_flags;
3546 /* Fetch socket values. */
3547 params.spp_hbinterval = sp->hbinterval;
3548 params.spp_pathmtu = sp->pathmtu;
3549 params.spp_sackdelay = sp->sackdelay;
3550 params.spp_pathmaxrxt = sp->pathmaxrxt;
3552 /*draft-11 doesn't say what to return in spp_flags*/
3553 params.spp_flags = sp->param_flags;
3556 if (copy_to_user(optval, ¶ms, len))
3559 if (put_user(len, optlen))
3565 /* 7.1.24. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
3567 * This options will get or set the delayed ack timer. The time is set
3568 * in milliseconds. If the assoc_id is 0, then this sets or gets the
3569 * endpoints default delayed ack timer value. If the assoc_id field is
3570 * non-zero, then the set or get effects the specified association.
3572 * struct sctp_assoc_value {
3573 * sctp_assoc_t assoc_id;
3574 * uint32_t assoc_value;
3577 * assoc_id - This parameter, indicates which association the
3578 * user is preforming an action upon. Note that if
3579 * this field's value is zero then the endpoints
3580 * default value is changed (effecting future
3581 * associations only).
3583 * assoc_value - This parameter contains the number of milliseconds
3584 * that the user is requesting the delayed ACK timer
3585 * be set to. Note that this value is defined in
3586 * the standard to be between 200 and 500 milliseconds.
3588 * Note: a value of zero will leave the value alone,
3589 * but disable SACK delay. A non-zero value will also
3590 * enable SACK delay.
3592 static int sctp_getsockopt_delayed_ack_time(struct sock *sk, int len,
3593 char __user *optval,
3596 struct sctp_assoc_value params;
3597 struct sctp_association *asoc = NULL;
3598 struct sctp_sock *sp = sctp_sk(sk);
3600 if (len != sizeof(struct sctp_assoc_value))
3603 if (copy_from_user(¶ms, optval, len))
3606 /* Get association, if assoc_id != 0 and the socket is a one
3607 * to many style socket, and an association was not found, then
3608 * the id was invalid.
3610 asoc = sctp_id2assoc(sk, params.assoc_id);
3611 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3615 /* Fetch association values. */
3616 if (asoc->param_flags & SPP_SACKDELAY_ENABLE)
3617 params.assoc_value = jiffies_to_msecs(
3620 params.assoc_value = 0;
3622 /* Fetch socket values. */
3623 if (sp->param_flags & SPP_SACKDELAY_ENABLE)
3624 params.assoc_value = sp->sackdelay;
3626 params.assoc_value = 0;
3629 if (copy_to_user(optval, ¶ms, len))
3632 if (put_user(len, optlen))
3638 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
3640 * Applications can specify protocol parameters for the default association
3641 * initialization. The option name argument to setsockopt() and getsockopt()
3644 * Setting initialization parameters is effective only on an unconnected
3645 * socket (for UDP-style sockets only future associations are effected
3646 * by the change). With TCP-style sockets, this option is inherited by
3647 * sockets derived from a listener socket.
3649 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
3651 if (len != sizeof(struct sctp_initmsg))
3653 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
3658 static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
3659 char __user *optval,
3663 struct sctp_association *asoc;
3664 struct list_head *pos;
3667 if (len != sizeof(sctp_assoc_t))
3670 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3673 /* For UDP-style sockets, id specifies the association to query. */
3674 asoc = sctp_id2assoc(sk, id);
3678 list_for_each(pos, &asoc->peer.transport_addr_list) {
3686 * Old API for getting list of peer addresses. Does not work for 32-bit
3687 * programs running on a 64-bit kernel
3689 static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
3690 char __user *optval,
3693 struct sctp_association *asoc;
3694 struct list_head *pos;
3696 struct sctp_getaddrs_old getaddrs;
3697 struct sctp_transport *from;
3699 union sctp_addr temp;
3700 struct sctp_sock *sp = sctp_sk(sk);
3703 if (len != sizeof(struct sctp_getaddrs_old))
3706 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs_old)))
3709 if (getaddrs.addr_num <= 0) return -EINVAL;
3711 /* For UDP-style sockets, id specifies the association to query. */
3712 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3716 to = (void __user *)getaddrs.addrs;
3717 list_for_each(pos, &asoc->peer.transport_addr_list) {
3718 from = list_entry(pos, struct sctp_transport, transports);
3719 memcpy(&temp, &from->ipaddr, sizeof(temp));
3720 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3721 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3722 temp.v4.sin_port = htons(temp.v4.sin_port);
3723 if (copy_to_user(to, &temp, addrlen))
3727 if (cnt >= getaddrs.addr_num) break;
3729 getaddrs.addr_num = cnt;
3730 if (copy_to_user(optval, &getaddrs, sizeof(struct sctp_getaddrs_old)))
3736 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
3737 char __user *optval, int __user *optlen)
3739 struct sctp_association *asoc;
3740 struct list_head *pos;
3742 struct sctp_getaddrs getaddrs;
3743 struct sctp_transport *from;
3745 union sctp_addr temp;
3746 struct sctp_sock *sp = sctp_sk(sk);
3751 if (len < sizeof(struct sctp_getaddrs))
3754 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
3757 /* For UDP-style sockets, id specifies the association to query. */
3758 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3762 to = optval + offsetof(struct sctp_getaddrs,addrs);
3763 space_left = len - sizeof(struct sctp_getaddrs) -
3764 offsetof(struct sctp_getaddrs,addrs);
3766 list_for_each(pos, &asoc->peer.transport_addr_list) {
3767 from = list_entry(pos, struct sctp_transport, transports);
3768 memcpy(&temp, &from->ipaddr, sizeof(temp));
3769 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3770 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3771 if(space_left < addrlen)
3773 temp.v4.sin_port = htons(temp.v4.sin_port);
3774 if (copy_to_user(to, &temp, addrlen))
3778 space_left -= addrlen;
3781 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
3783 bytes_copied = ((char __user *)to) - optval;
3784 if (put_user(bytes_copied, optlen))
3790 static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
3791 char __user *optval,
3795 struct sctp_bind_addr *bp;
3796 struct sctp_association *asoc;
3797 struct list_head *pos;
3798 struct sctp_sockaddr_entry *addr;
3799 rwlock_t *addr_lock;
3800 unsigned long flags;
3803 if (len != sizeof(sctp_assoc_t))
3806 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3810 * For UDP-style sockets, id specifies the association to query.
3811 * If the id field is set to the value '0' then the locally bound
3812 * addresses are returned without regard to any particular
3816 bp = &sctp_sk(sk)->ep->base.bind_addr;
3817 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
3819 asoc = sctp_id2assoc(sk, id);
3822 bp = &asoc->base.bind_addr;
3823 addr_lock = &asoc->base.addr_lock;
3826 sctp_read_lock(addr_lock);
3828 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
3829 * addresses from the global local address list.
3831 if (sctp_list_single_entry(&bp->address_list)) {
3832 addr = list_entry(bp->address_list.next,
3833 struct sctp_sockaddr_entry, list);
3834 if (sctp_is_any(&addr->a)) {
3835 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
3836 list_for_each(pos, &sctp_local_addr_list) {
3837 addr = list_entry(pos,
3838 struct sctp_sockaddr_entry,
3840 if ((PF_INET == sk->sk_family) &&
3841 (AF_INET6 == addr->a.sa.sa_family))
3845 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock,
3853 list_for_each(pos, &bp->address_list) {
3858 sctp_read_unlock(addr_lock);
3862 /* Helper function that copies local addresses to user and returns the number
3863 * of addresses copied.
3865 static int sctp_copy_laddrs_to_user_old(struct sock *sk, __u16 port, int max_addrs,
3868 struct list_head *pos;
3869 struct sctp_sockaddr_entry *addr;
3870 unsigned long flags;
3871 union sctp_addr temp;
3875 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
3876 list_for_each(pos, &sctp_local_addr_list) {
3877 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
3878 if ((PF_INET == sk->sk_family) &&
3879 (AF_INET6 == addr->a.sa.sa_family))
3881 memcpy(&temp, &addr->a, sizeof(temp));
3882 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3884 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
3885 temp.v4.sin_port = htons(port);
3886 if (copy_to_user(to, &temp, addrlen)) {
3887 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock,
3893 if (cnt >= max_addrs) break;
3895 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock, flags);
3900 static int sctp_copy_laddrs_to_user(struct sock *sk, __u16 port,
3901 void __user **to, size_t space_left)
3903 struct list_head *pos;
3904 struct sctp_sockaddr_entry *addr;
3905 unsigned long flags;
3906 union sctp_addr temp;
3910 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
3911 list_for_each(pos, &sctp_local_addr_list) {
3912 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
3913 if ((PF_INET == sk->sk_family) &&
3914 (AF_INET6 == addr->a.sa.sa_family))
3916 memcpy(&temp, &addr->a, sizeof(temp));
3917 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3919 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
3920 if(space_left<addrlen)
3922 temp.v4.sin_port = htons(port);
3923 if (copy_to_user(*to, &temp, addrlen)) {
3924 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock,
3930 space_left -= addrlen;
3932 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock, flags);
3937 /* Old API for getting list of local addresses. Does not work for 32-bit
3938 * programs running on a 64-bit kernel
3940 static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
3941 char __user *optval, int __user *optlen)
3943 struct sctp_bind_addr *bp;
3944 struct sctp_association *asoc;
3945 struct list_head *pos;
3947 struct sctp_getaddrs_old getaddrs;
3948 struct sctp_sockaddr_entry *addr;
3950 union sctp_addr temp;
3951 struct sctp_sock *sp = sctp_sk(sk);
3953 rwlock_t *addr_lock;
3956 if (len != sizeof(struct sctp_getaddrs_old))
3959 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs_old)))
3962 if (getaddrs.addr_num <= 0) return -EINVAL;
3964 * For UDP-style sockets, id specifies the association to query.
3965 * If the id field is set to the value '0' then the locally bound
3966 * addresses are returned without regard to any particular
3969 if (0 == getaddrs.assoc_id) {
3970 bp = &sctp_sk(sk)->ep->base.bind_addr;
3971 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
3973 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3976 bp = &asoc->base.bind_addr;
3977 addr_lock = &asoc->base.addr_lock;
3980 to = getaddrs.addrs;
3982 sctp_read_lock(addr_lock);
3984 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
3985 * addresses from the global local address list.
3987 if (sctp_list_single_entry(&bp->address_list)) {
3988 addr = list_entry(bp->address_list.next,
3989 struct sctp_sockaddr_entry, list);
3990 if (sctp_is_any(&addr->a)) {
3991 cnt = sctp_copy_laddrs_to_user_old(sk, bp->port,
4002 list_for_each(pos, &bp->address_list) {
4003 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4004 memcpy(&temp, &addr->a, sizeof(temp));
4005 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4006 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4007 temp.v4.sin_port = htons(temp.v4.sin_port);
4008 if (copy_to_user(to, &temp, addrlen)) {
4014 if (cnt >= getaddrs.addr_num) break;
4018 getaddrs.addr_num = cnt;
4019 if (copy_to_user(optval, &getaddrs, sizeof(struct sctp_getaddrs_old)))
4023 sctp_read_unlock(addr_lock);
4027 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4028 char __user *optval, int __user *optlen)
4030 struct sctp_bind_addr *bp;
4031 struct sctp_association *asoc;
4032 struct list_head *pos;
4034 struct sctp_getaddrs getaddrs;
4035 struct sctp_sockaddr_entry *addr;
4037 union sctp_addr temp;
4038 struct sctp_sock *sp = sctp_sk(sk);
4040 rwlock_t *addr_lock;
4045 if (len <= sizeof(struct sctp_getaddrs))
4048 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4052 * For UDP-style sockets, id specifies the association to query.
4053 * If the id field is set to the value '0' then the locally bound
4054 * addresses are returned without regard to any particular
4057 if (0 == getaddrs.assoc_id) {
4058 bp = &sctp_sk(sk)->ep->base.bind_addr;
4059 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4061 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4064 bp = &asoc->base.bind_addr;
4065 addr_lock = &asoc->base.addr_lock;
4068 to = optval + offsetof(struct sctp_getaddrs,addrs);
4069 space_left = len - sizeof(struct sctp_getaddrs) -
4070 offsetof(struct sctp_getaddrs,addrs);
4072 sctp_read_lock(addr_lock);
4074 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4075 * addresses from the global local address list.
4077 if (sctp_list_single_entry(&bp->address_list)) {
4078 addr = list_entry(bp->address_list.next,
4079 struct sctp_sockaddr_entry, list);
4080 if (sctp_is_any(&addr->a)) {
4081 cnt = sctp_copy_laddrs_to_user(sk, bp->port,
4091 list_for_each(pos, &bp->address_list) {
4092 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4093 memcpy(&temp, &addr->a, sizeof(temp));
4094 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4095 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4096 if(space_left < addrlen)
4097 return -ENOMEM; /*fixme: right error?*/
4098 temp.v4.sin_port = htons(temp.v4.sin_port);
4099 if (copy_to_user(to, &temp, addrlen)) {
4105 space_left -= addrlen;
4109 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4111 bytes_copied = ((char __user *)to) - optval;
4112 if (put_user(bytes_copied, optlen))
4116 sctp_read_unlock(addr_lock);
4120 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4122 * Requests that the local SCTP stack use the enclosed peer address as
4123 * the association primary. The enclosed address must be one of the
4124 * association peer's addresses.
4126 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4127 char __user *optval, int __user *optlen)
4129 struct sctp_prim prim;
4130 struct sctp_association *asoc;
4131 struct sctp_sock *sp = sctp_sk(sk);
4133 if (len != sizeof(struct sctp_prim))
4136 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
4139 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4143 if (!asoc->peer.primary_path)
4146 asoc->peer.primary_path->ipaddr.v4.sin_port =
4147 htons(asoc->peer.primary_path->ipaddr.v4.sin_port);
4148 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4149 sizeof(union sctp_addr));
4150 asoc->peer.primary_path->ipaddr.v4.sin_port =
4151 ntohs(asoc->peer.primary_path->ipaddr.v4.sin_port);
4153 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4154 (union sctp_addr *)&prim.ssp_addr);
4156 if (copy_to_user(optval, &prim, sizeof(struct sctp_prim)))
4163 * 7.1.11 Set Adaption Layer Indicator (SCTP_ADAPTION_LAYER)
4165 * Requests that the local endpoint set the specified Adaption Layer
4166 * Indication parameter for all future INIT and INIT-ACK exchanges.
4168 static int sctp_getsockopt_adaption_layer(struct sock *sk, int len,
4169 char __user *optval, int __user *optlen)
4171 struct sctp_setadaption adaption;
4173 if (len != sizeof(struct sctp_setadaption))
4176 adaption.ssb_adaption_ind = sctp_sk(sk)->adaption_ind;
4177 if (copy_to_user(optval, &adaption, len))
4185 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4187 * Applications that wish to use the sendto() system call may wish to
4188 * specify a default set of parameters that would normally be supplied
4189 * through the inclusion of ancillary data. This socket option allows
4190 * such an application to set the default sctp_sndrcvinfo structure.
4193 * The application that wishes to use this socket option simply passes
4194 * in to this call the sctp_sndrcvinfo structure defined in Section
4195 * 5.2.2) The input parameters accepted by this call include
4196 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4197 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4198 * to this call if the caller is using the UDP model.
4200 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4202 static int sctp_getsockopt_default_send_param(struct sock *sk,
4203 int len, char __user *optval,
4206 struct sctp_sndrcvinfo info;
4207 struct sctp_association *asoc;
4208 struct sctp_sock *sp = sctp_sk(sk);
4210 if (len != sizeof(struct sctp_sndrcvinfo))
4212 if (copy_from_user(&info, optval, sizeof(struct sctp_sndrcvinfo)))
4215 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4216 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4220 info.sinfo_stream = asoc->default_stream;
4221 info.sinfo_flags = asoc->default_flags;
4222 info.sinfo_ppid = asoc->default_ppid;
4223 info.sinfo_context = asoc->default_context;
4224 info.sinfo_timetolive = asoc->default_timetolive;
4226 info.sinfo_stream = sp->default_stream;
4227 info.sinfo_flags = sp->default_flags;
4228 info.sinfo_ppid = sp->default_ppid;
4229 info.sinfo_context = sp->default_context;
4230 info.sinfo_timetolive = sp->default_timetolive;
4233 if (copy_to_user(optval, &info, sizeof(struct sctp_sndrcvinfo)))
4241 * 7.1.5 SCTP_NODELAY
4243 * Turn on/off any Nagle-like algorithm. This means that packets are
4244 * generally sent as soon as possible and no unnecessary delays are
4245 * introduced, at the cost of more packets in the network. Expects an
4246 * integer boolean flag.
4249 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4250 char __user *optval, int __user *optlen)
4254 if (len < sizeof(int))
4258 val = (sctp_sk(sk)->nodelay == 1);
4259 if (put_user(len, optlen))
4261 if (copy_to_user(optval, &val, len))
4268 * 7.1.1 SCTP_RTOINFO
4270 * The protocol parameters used to initialize and bound retransmission
4271 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4272 * and modify these parameters.
4273 * All parameters are time values, in milliseconds. A value of 0, when
4274 * modifying the parameters, indicates that the current value should not
4278 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4279 char __user *optval,
4280 int __user *optlen) {
4281 struct sctp_rtoinfo rtoinfo;
4282 struct sctp_association *asoc;
4284 if (len != sizeof (struct sctp_rtoinfo))
4287 if (copy_from_user(&rtoinfo, optval, sizeof (struct sctp_rtoinfo)))
4290 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4292 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4295 /* Values corresponding to the specific association. */
4297 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4298 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4299 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4301 /* Values corresponding to the endpoint. */
4302 struct sctp_sock *sp = sctp_sk(sk);
4304 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4305 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4306 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4309 if (put_user(len, optlen))
4312 if (copy_to_user(optval, &rtoinfo, len))
4320 * 7.1.2 SCTP_ASSOCINFO
4322 * This option is used to tune the the maximum retransmission attempts
4323 * of the association.
4324 * Returns an error if the new association retransmission value is
4325 * greater than the sum of the retransmission value of the peer.
4326 * See [SCTP] for more information.
4329 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4330 char __user *optval,
4334 struct sctp_assocparams assocparams;
4335 struct sctp_association *asoc;
4336 struct list_head *pos;
4339 if (len != sizeof (struct sctp_assocparams))
4342 if (copy_from_user(&assocparams, optval,
4343 sizeof (struct sctp_assocparams)))
4346 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4348 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4351 /* Values correspoinding to the specific association */
4353 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4354 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4355 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4356 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4358 (asoc->cookie_life.tv_usec
4361 list_for_each(pos, &asoc->peer.transport_addr_list) {
4365 assocparams.sasoc_number_peer_destinations = cnt;
4367 /* Values corresponding to the endpoint */
4368 struct sctp_sock *sp = sctp_sk(sk);
4370 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4371 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4372 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4373 assocparams.sasoc_cookie_life =
4374 sp->assocparams.sasoc_cookie_life;
4375 assocparams.sasoc_number_peer_destinations =
4377 sasoc_number_peer_destinations;
4380 if (put_user(len, optlen))
4383 if (copy_to_user(optval, &assocparams, len))
4390 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4392 * This socket option is a boolean flag which turns on or off mapped V4
4393 * addresses. If this option is turned on and the socket is type
4394 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4395 * If this option is turned off, then no mapping will be done of V4
4396 * addresses and a user will receive both PF_INET6 and PF_INET type
4397 * addresses on the socket.
4399 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4400 char __user *optval, int __user *optlen)
4403 struct sctp_sock *sp = sctp_sk(sk);
4405 if (len < sizeof(int))
4410 if (put_user(len, optlen))
4412 if (copy_to_user(optval, &val, len))
4419 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
4421 * This socket option specifies the maximum size to put in any outgoing
4422 * SCTP chunk. If a message is larger than this size it will be
4423 * fragmented by SCTP into the specified size. Note that the underlying
4424 * SCTP implementation may fragment into smaller sized chunks when the
4425 * PMTU of the underlying association is smaller than the value set by
4428 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4429 char __user *optval, int __user *optlen)
4433 if (len < sizeof(int))
4438 val = sctp_sk(sk)->user_frag;
4439 if (put_user(len, optlen))
4441 if (copy_to_user(optval, &val, len))
4447 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
4448 char __user *optval, int __user *optlen)
4453 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
4456 /* I can hardly begin to describe how wrong this is. This is
4457 * so broken as to be worse than useless. The API draft
4458 * REALLY is NOT helpful here... I am not convinced that the
4459 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
4460 * are at all well-founded.
4462 if (level != SOL_SCTP) {
4463 struct sctp_af *af = sctp_sk(sk)->pf->af;
4465 retval = af->getsockopt(sk, level, optname, optval, optlen);
4469 if (get_user(len, optlen))
4476 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
4478 case SCTP_DISABLE_FRAGMENTS:
4479 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
4483 retval = sctp_getsockopt_events(sk, len, optval, optlen);
4485 case SCTP_AUTOCLOSE:
4486 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
4488 case SCTP_SOCKOPT_PEELOFF:
4489 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
4491 case SCTP_PEER_ADDR_PARAMS:
4492 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
4495 case SCTP_DELAYED_ACK_TIME:
4496 retval = sctp_getsockopt_delayed_ack_time(sk, len, optval,
4500 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
4502 case SCTP_GET_PEER_ADDRS_NUM_OLD:
4503 retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
4506 case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
4507 retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
4510 case SCTP_GET_PEER_ADDRS_OLD:
4511 retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
4514 case SCTP_GET_LOCAL_ADDRS_OLD:
4515 retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
4518 case SCTP_GET_PEER_ADDRS:
4519 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
4522 case SCTP_GET_LOCAL_ADDRS:
4523 retval = sctp_getsockopt_local_addrs(sk, len, optval,
4526 case SCTP_DEFAULT_SEND_PARAM:
4527 retval = sctp_getsockopt_default_send_param(sk, len,
4530 case SCTP_PRIMARY_ADDR:
4531 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
4534 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
4537 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
4539 case SCTP_ASSOCINFO:
4540 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
4542 case SCTP_I_WANT_MAPPED_V4_ADDR:
4543 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
4546 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
4548 case SCTP_GET_PEER_ADDR_INFO:
4549 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
4552 case SCTP_ADAPTION_LAYER:
4553 retval = sctp_getsockopt_adaption_layer(sk, len, optval,
4557 retval = -ENOPROTOOPT;
4561 sctp_release_sock(sk);
4565 static void sctp_hash(struct sock *sk)
4570 static void sctp_unhash(struct sock *sk)
4575 /* Check if port is acceptable. Possibly find first available port.
4577 * The port hash table (contained in the 'global' SCTP protocol storage
4578 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
4579 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
4580 * list (the list number is the port number hashed out, so as you
4581 * would expect from a hash function, all the ports in a given list have
4582 * such a number that hashes out to the same list number; you were
4583 * expecting that, right?); so each list has a set of ports, with a
4584 * link to the socket (struct sock) that uses it, the port number and
4585 * a fastreuse flag (FIXME: NPI ipg).
4587 static struct sctp_bind_bucket *sctp_bucket_create(
4588 struct sctp_bind_hashbucket *head, unsigned short snum);
4590 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
4592 struct sctp_bind_hashbucket *head; /* hash list */
4593 struct sctp_bind_bucket *pp; /* hash list port iterator */
4594 unsigned short snum;
4597 /* NOTE: Remember to put this back to net order. */
4598 addr->v4.sin_port = ntohs(addr->v4.sin_port);
4599 snum = addr->v4.sin_port;
4601 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
4602 sctp_local_bh_disable();
4605 /* Search for an available port.
4607 * 'sctp_port_rover' was the last port assigned, so
4608 * we start to search from 'sctp_port_rover +
4609 * 1'. What we do is first check if port 'rover' is
4610 * already in the hash table; if not, we use that; if
4611 * it is, we try next.
4613 int low = sysctl_local_port_range[0];
4614 int high = sysctl_local_port_range[1];
4615 int remaining = (high - low) + 1;
4619 sctp_spin_lock(&sctp_port_alloc_lock);
4620 rover = sctp_port_rover;
4623 if ((rover < low) || (rover > high))
4625 index = sctp_phashfn(rover);
4626 head = &sctp_port_hashtable[index];
4627 sctp_spin_lock(&head->lock);
4628 for (pp = head->chain; pp; pp = pp->next)
4629 if (pp->port == rover)
4633 sctp_spin_unlock(&head->lock);
4634 } while (--remaining > 0);
4635 sctp_port_rover = rover;
4636 sctp_spin_unlock(&sctp_port_alloc_lock);
4638 /* Exhausted local port range during search? */
4643 /* OK, here is the one we will use. HEAD (the port
4644 * hash table list entry) is non-NULL and we hold it's
4649 /* We are given an specific port number; we verify
4650 * that it is not being used. If it is used, we will
4651 * exahust the search in the hash list corresponding
4652 * to the port number (snum) - we detect that with the
4653 * port iterator, pp being NULL.
4655 head = &sctp_port_hashtable[sctp_phashfn(snum)];
4656 sctp_spin_lock(&head->lock);
4657 for (pp = head->chain; pp; pp = pp->next) {
4658 if (pp->port == snum)
4665 if (!hlist_empty(&pp->owner)) {
4666 /* We had a port hash table hit - there is an
4667 * available port (pp != NULL) and it is being
4668 * used by other socket (pp->owner not empty); that other
4669 * socket is going to be sk2.
4671 int reuse = sk->sk_reuse;
4673 struct hlist_node *node;
4675 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
4676 if (pp->fastreuse && sk->sk_reuse)
4679 /* Run through the list of sockets bound to the port
4680 * (pp->port) [via the pointers bind_next and
4681 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
4682 * we get the endpoint they describe and run through
4683 * the endpoint's list of IP (v4 or v6) addresses,
4684 * comparing each of the addresses with the address of
4685 * the socket sk. If we find a match, then that means
4686 * that this port/socket (sk) combination are already
4689 sk_for_each_bound(sk2, node, &pp->owner) {
4690 struct sctp_endpoint *ep2;
4691 ep2 = sctp_sk(sk2)->ep;
4693 if (reuse && sk2->sk_reuse)
4696 if (sctp_bind_addr_match(&ep2->base.bind_addr, addr,
4702 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
4705 /* If there was a hash table miss, create a new port. */
4707 if (!pp && !(pp = sctp_bucket_create(head, snum)))
4710 /* In either case (hit or miss), make sure fastreuse is 1 only
4711 * if sk->sk_reuse is too (that is, if the caller requested
4712 * SO_REUSEADDR on this socket -sk-).
4714 if (hlist_empty(&pp->owner))
4715 pp->fastreuse = sk->sk_reuse ? 1 : 0;
4716 else if (pp->fastreuse && !sk->sk_reuse)
4719 /* We are set, so fill up all the data in the hash table
4720 * entry, tie the socket list information with the rest of the
4721 * sockets FIXME: Blurry, NPI (ipg).
4724 inet_sk(sk)->num = snum;
4725 if (!sctp_sk(sk)->bind_hash) {
4726 sk_add_bind_node(sk, &pp->owner);
4727 sctp_sk(sk)->bind_hash = pp;
4732 sctp_spin_unlock(&head->lock);
4735 sctp_local_bh_enable();
4736 addr->v4.sin_port = htons(addr->v4.sin_port);
4740 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
4741 * port is requested.
4743 static int sctp_get_port(struct sock *sk, unsigned short snum)
4746 union sctp_addr addr;
4747 struct sctp_af *af = sctp_sk(sk)->pf->af;
4749 /* Set up a dummy address struct from the sk. */
4750 af->from_sk(&addr, sk);
4751 addr.v4.sin_port = htons(snum);
4753 /* Note: sk->sk_num gets filled in if ephemeral port request. */
4754 ret = sctp_get_port_local(sk, &addr);
4756 return (ret ? 1 : 0);
4760 * 3.1.3 listen() - UDP Style Syntax
4762 * By default, new associations are not accepted for UDP style sockets.
4763 * An application uses listen() to mark a socket as being able to
4764 * accept new associations.
4766 SCTP_STATIC int sctp_seqpacket_listen(struct sock *sk, int backlog)
4768 struct sctp_sock *sp = sctp_sk(sk);
4769 struct sctp_endpoint *ep = sp->ep;
4771 /* Only UDP style sockets that are not peeled off are allowed to
4774 if (!sctp_style(sk, UDP))
4777 /* If backlog is zero, disable listening. */
4779 if (sctp_sstate(sk, CLOSED))
4782 sctp_unhash_endpoint(ep);
4783 sk->sk_state = SCTP_SS_CLOSED;
4786 /* Return if we are already listening. */
4787 if (sctp_sstate(sk, LISTENING))
4791 * If a bind() or sctp_bindx() is not called prior to a listen()
4792 * call that allows new associations to be accepted, the system
4793 * picks an ephemeral port and will choose an address set equivalent
4794 * to binding with a wildcard address.
4796 * This is not currently spelled out in the SCTP sockets
4797 * extensions draft, but follows the practice as seen in TCP
4800 if (!ep->base.bind_addr.port) {
4801 if (sctp_autobind(sk))
4804 sk->sk_state = SCTP_SS_LISTENING;
4805 sctp_hash_endpoint(ep);
4810 * 4.1.3 listen() - TCP Style Syntax
4812 * Applications uses listen() to ready the SCTP endpoint for accepting
4813 * inbound associations.
4815 SCTP_STATIC int sctp_stream_listen(struct sock *sk, int backlog)
4817 struct sctp_sock *sp = sctp_sk(sk);
4818 struct sctp_endpoint *ep = sp->ep;
4820 /* If backlog is zero, disable listening. */
4822 if (sctp_sstate(sk, CLOSED))
4825 sctp_unhash_endpoint(ep);
4826 sk->sk_state = SCTP_SS_CLOSED;
4829 if (sctp_sstate(sk, LISTENING))
4833 * If a bind() or sctp_bindx() is not called prior to a listen()
4834 * call that allows new associations to be accepted, the system
4835 * picks an ephemeral port and will choose an address set equivalent
4836 * to binding with a wildcard address.
4838 * This is not currently spelled out in the SCTP sockets
4839 * extensions draft, but follows the practice as seen in TCP
4842 if (!ep->base.bind_addr.port) {
4843 if (sctp_autobind(sk))
4846 sk->sk_state = SCTP_SS_LISTENING;
4847 sk->sk_max_ack_backlog = backlog;
4848 sctp_hash_endpoint(ep);
4853 * Move a socket to LISTENING state.
4855 int sctp_inet_listen(struct socket *sock, int backlog)
4857 struct sock *sk = sock->sk;
4858 struct crypto_tfm *tfm=NULL;
4861 if (unlikely(backlog < 0))
4866 if (sock->state != SS_UNCONNECTED)
4869 /* Allocate HMAC for generating cookie. */
4870 if (sctp_hmac_alg) {
4871 tfm = sctp_crypto_alloc_tfm(sctp_hmac_alg, 0);
4878 switch (sock->type) {
4879 case SOCK_SEQPACKET:
4880 err = sctp_seqpacket_listen(sk, backlog);
4883 err = sctp_stream_listen(sk, backlog);
4891 /* Store away the transform reference. */
4892 sctp_sk(sk)->hmac = tfm;
4894 sctp_release_sock(sk);
4897 sctp_crypto_free_tfm(tfm);
4902 * This function is done by modeling the current datagram_poll() and the
4903 * tcp_poll(). Note that, based on these implementations, we don't
4904 * lock the socket in this function, even though it seems that,
4905 * ideally, locking or some other mechanisms can be used to ensure
4906 * the integrity of the counters (sndbuf and wmem_alloc) used
4907 * in this place. We assume that we don't need locks either until proven
4910 * Another thing to note is that we include the Async I/O support
4911 * here, again, by modeling the current TCP/UDP code. We don't have
4912 * a good way to test with it yet.
4914 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
4916 struct sock *sk = sock->sk;
4917 struct sctp_sock *sp = sctp_sk(sk);
4920 poll_wait(file, sk->sk_sleep, wait);
4922 /* A TCP-style listening socket becomes readable when the accept queue
4925 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4926 return (!list_empty(&sp->ep->asocs)) ?
4927 (POLLIN | POLLRDNORM) : 0;
4931 /* Is there any exceptional events? */
4932 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
4934 if (sk->sk_shutdown & RCV_SHUTDOWN)
4936 if (sk->sk_shutdown == SHUTDOWN_MASK)
4939 /* Is it readable? Reconsider this code with TCP-style support. */
4940 if (!skb_queue_empty(&sk->sk_receive_queue) ||
4941 (sk->sk_shutdown & RCV_SHUTDOWN))
4942 mask |= POLLIN | POLLRDNORM;
4944 /* The association is either gone or not ready. */
4945 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
4948 /* Is it writable? */
4949 if (sctp_writeable(sk)) {
4950 mask |= POLLOUT | POLLWRNORM;
4952 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
4954 * Since the socket is not locked, the buffer
4955 * might be made available after the writeable check and
4956 * before the bit is set. This could cause a lost I/O
4957 * signal. tcp_poll() has a race breaker for this race
4958 * condition. Based on their implementation, we put
4959 * in the following code to cover it as well.
4961 if (sctp_writeable(sk))
4962 mask |= POLLOUT | POLLWRNORM;
4967 /********************************************************************
4968 * 2nd Level Abstractions
4969 ********************************************************************/
4971 static struct sctp_bind_bucket *sctp_bucket_create(
4972 struct sctp_bind_hashbucket *head, unsigned short snum)
4974 struct sctp_bind_bucket *pp;
4976 pp = kmem_cache_alloc(sctp_bucket_cachep, SLAB_ATOMIC);
4977 SCTP_DBG_OBJCNT_INC(bind_bucket);
4981 INIT_HLIST_HEAD(&pp->owner);
4982 if ((pp->next = head->chain) != NULL)
4983 pp->next->pprev = &pp->next;
4985 pp->pprev = &head->chain;
4990 /* Caller must hold hashbucket lock for this tb with local BH disabled */
4991 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
4993 if (hlist_empty(&pp->owner)) {
4995 pp->next->pprev = pp->pprev;
4996 *(pp->pprev) = pp->next;
4997 kmem_cache_free(sctp_bucket_cachep, pp);
4998 SCTP_DBG_OBJCNT_DEC(bind_bucket);
5002 /* Release this socket's reference to a local port. */
5003 static inline void __sctp_put_port(struct sock *sk)
5005 struct sctp_bind_hashbucket *head =
5006 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
5007 struct sctp_bind_bucket *pp;
5009 sctp_spin_lock(&head->lock);
5010 pp = sctp_sk(sk)->bind_hash;
5011 __sk_del_bind_node(sk);
5012 sctp_sk(sk)->bind_hash = NULL;
5013 inet_sk(sk)->num = 0;
5014 sctp_bucket_destroy(pp);
5015 sctp_spin_unlock(&head->lock);
5018 void sctp_put_port(struct sock *sk)
5020 sctp_local_bh_disable();
5021 __sctp_put_port(sk);
5022 sctp_local_bh_enable();
5026 * The system picks an ephemeral port and choose an address set equivalent
5027 * to binding with a wildcard address.
5028 * One of those addresses will be the primary address for the association.
5029 * This automatically enables the multihoming capability of SCTP.
5031 static int sctp_autobind(struct sock *sk)
5033 union sctp_addr autoaddr;
5035 unsigned short port;
5037 /* Initialize a local sockaddr structure to INADDR_ANY. */
5038 af = sctp_sk(sk)->pf->af;
5040 port = htons(inet_sk(sk)->num);
5041 af->inaddr_any(&autoaddr, port);
5043 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5046 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5049 * 4.2 The cmsghdr Structure *
5051 * When ancillary data is sent or received, any number of ancillary data
5052 * objects can be specified by the msg_control and msg_controllen members of
5053 * the msghdr structure, because each object is preceded by
5054 * a cmsghdr structure defining the object's length (the cmsg_len member).
5055 * Historically Berkeley-derived implementations have passed only one object
5056 * at a time, but this API allows multiple objects to be
5057 * passed in a single call to sendmsg() or recvmsg(). The following example
5058 * shows two ancillary data objects in a control buffer.
5060 * |<--------------------------- msg_controllen -------------------------->|
5063 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5065 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5068 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5070 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5073 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5074 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5076 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5078 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5085 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5086 sctp_cmsgs_t *cmsgs)
5088 struct cmsghdr *cmsg;
5090 for (cmsg = CMSG_FIRSTHDR(msg);
5092 cmsg = CMSG_NXTHDR((struct msghdr*)msg, cmsg)) {
5093 if (!CMSG_OK(msg, cmsg))
5096 /* Should we parse this header or ignore? */
5097 if (cmsg->cmsg_level != IPPROTO_SCTP)
5100 /* Strictly check lengths following example in SCM code. */
5101 switch (cmsg->cmsg_type) {
5103 /* SCTP Socket API Extension
5104 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5106 * This cmsghdr structure provides information for
5107 * initializing new SCTP associations with sendmsg().
5108 * The SCTP_INITMSG socket option uses this same data
5109 * structure. This structure is not used for
5112 * cmsg_level cmsg_type cmsg_data[]
5113 * ------------ ------------ ----------------------
5114 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5116 if (cmsg->cmsg_len !=
5117 CMSG_LEN(sizeof(struct sctp_initmsg)))
5119 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5123 /* SCTP Socket API Extension
5124 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5126 * This cmsghdr structure specifies SCTP options for
5127 * sendmsg() and describes SCTP header information
5128 * about a received message through recvmsg().
5130 * cmsg_level cmsg_type cmsg_data[]
5131 * ------------ ------------ ----------------------
5132 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5134 if (cmsg->cmsg_len !=
5135 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
5139 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
5141 /* Minimally, validate the sinfo_flags. */
5142 if (cmsgs->info->sinfo_flags &
5143 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
5144 SCTP_ABORT | SCTP_EOF))
5156 * Wait for a packet..
5157 * Note: This function is the same function as in core/datagram.c
5158 * with a few modifications to make lksctp work.
5160 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
5165 prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5167 /* Socket errors? */
5168 error = sock_error(sk);
5172 if (!skb_queue_empty(&sk->sk_receive_queue))
5175 /* Socket shut down? */
5176 if (sk->sk_shutdown & RCV_SHUTDOWN)
5179 /* Sequenced packets can come disconnected. If so we report the
5184 /* Is there a good reason to think that we may receive some data? */
5185 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
5188 /* Handle signals. */
5189 if (signal_pending(current))
5192 /* Let another process have a go. Since we are going to sleep
5193 * anyway. Note: This may cause odd behaviors if the message
5194 * does not fit in the user's buffer, but this seems to be the
5195 * only way to honor MSG_DONTWAIT realistically.
5197 sctp_release_sock(sk);
5198 *timeo_p = schedule_timeout(*timeo_p);
5202 finish_wait(sk->sk_sleep, &wait);
5206 error = sock_intr_errno(*timeo_p);
5209 finish_wait(sk->sk_sleep, &wait);
5214 /* Receive a datagram.
5215 * Note: This is pretty much the same routine as in core/datagram.c
5216 * with a few changes to make lksctp work.
5218 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
5219 int noblock, int *err)
5222 struct sk_buff *skb;
5225 timeo = sock_rcvtimeo(sk, noblock);
5227 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
5228 timeo, MAX_SCHEDULE_TIMEOUT);
5231 /* Again only user level code calls this function,
5232 * so nothing interrupt level
5233 * will suddenly eat the receive_queue.
5235 * Look at current nfs client by the way...
5236 * However, this function was corrent in any case. 8)
5238 if (flags & MSG_PEEK) {
5239 spin_lock_bh(&sk->sk_receive_queue.lock);
5240 skb = skb_peek(&sk->sk_receive_queue);
5242 atomic_inc(&skb->users);
5243 spin_unlock_bh(&sk->sk_receive_queue.lock);
5245 skb = skb_dequeue(&sk->sk_receive_queue);
5251 /* Caller is allowed not to check sk->sk_err before calling. */
5252 error = sock_error(sk);
5256 if (sk->sk_shutdown & RCV_SHUTDOWN)
5259 /* User doesn't want to wait. */
5263 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
5272 /* If sndbuf has changed, wake up per association sndbuf waiters. */
5273 static void __sctp_write_space(struct sctp_association *asoc)
5275 struct sock *sk = asoc->base.sk;
5276 struct socket *sock = sk->sk_socket;
5278 if ((sctp_wspace(asoc) > 0) && sock) {
5279 if (waitqueue_active(&asoc->wait))
5280 wake_up_interruptible(&asoc->wait);
5282 if (sctp_writeable(sk)) {
5283 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
5284 wake_up_interruptible(sk->sk_sleep);
5286 /* Note that we try to include the Async I/O support
5287 * here by modeling from the current TCP/UDP code.
5288 * We have not tested with it yet.
5290 if (sock->fasync_list &&
5291 !(sk->sk_shutdown & SEND_SHUTDOWN))
5292 sock_wake_async(sock, 2, POLL_OUT);
5297 /* Do accounting for the sndbuf space.
5298 * Decrement the used sndbuf space of the corresponding association by the
5299 * data size which was just transmitted(freed).
5301 static void sctp_wfree(struct sk_buff *skb)
5303 struct sctp_association *asoc;
5304 struct sctp_chunk *chunk;
5307 /* Get the saved chunk pointer. */
5308 chunk = *((struct sctp_chunk **)(skb->cb));
5311 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
5312 sizeof(struct sk_buff) +
5313 sizeof(struct sctp_chunk);
5315 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
5318 __sctp_write_space(asoc);
5320 sctp_association_put(asoc);
5323 /* Helper function to wait for space in the sndbuf. */
5324 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
5327 struct sock *sk = asoc->base.sk;
5329 long current_timeo = *timeo_p;
5332 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
5333 asoc, (long)(*timeo_p), msg_len);
5335 /* Increment the association's refcnt. */
5336 sctp_association_hold(asoc);
5338 /* Wait on the association specific sndbuf space. */
5340 prepare_to_wait_exclusive(&asoc->wait, &wait,
5341 TASK_INTERRUPTIBLE);
5344 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
5347 if (signal_pending(current))
5348 goto do_interrupted;
5349 if (msg_len <= sctp_wspace(asoc))
5352 /* Let another process have a go. Since we are going
5355 sctp_release_sock(sk);
5356 current_timeo = schedule_timeout(current_timeo);
5357 BUG_ON(sk != asoc->base.sk);
5360 *timeo_p = current_timeo;
5364 finish_wait(&asoc->wait, &wait);
5366 /* Release the association's refcnt. */
5367 sctp_association_put(asoc);
5376 err = sock_intr_errno(*timeo_p);
5384 /* If socket sndbuf has changed, wake up all per association waiters. */
5385 void sctp_write_space(struct sock *sk)
5387 struct sctp_association *asoc;
5388 struct list_head *pos;
5390 /* Wake up the tasks in each wait queue. */
5391 list_for_each(pos, &((sctp_sk(sk))->ep->asocs)) {
5392 asoc = list_entry(pos, struct sctp_association, asocs);
5393 __sctp_write_space(asoc);
5397 /* Is there any sndbuf space available on the socket?
5399 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
5400 * associations on the same socket. For a UDP-style socket with
5401 * multiple associations, it is possible for it to be "unwriteable"
5402 * prematurely. I assume that this is acceptable because
5403 * a premature "unwriteable" is better than an accidental "writeable" which
5404 * would cause an unwanted block under certain circumstances. For the 1-1
5405 * UDP-style sockets or TCP-style sockets, this code should work.
5408 static int sctp_writeable(struct sock *sk)
5412 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
5418 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
5419 * returns immediately with EINPROGRESS.
5421 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
5423 struct sock *sk = asoc->base.sk;
5425 long current_timeo = *timeo_p;
5428 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __FUNCTION__, asoc,
5431 /* Increment the association's refcnt. */
5432 sctp_association_hold(asoc);
5435 prepare_to_wait_exclusive(&asoc->wait, &wait,
5436 TASK_INTERRUPTIBLE);
5439 if (sk->sk_shutdown & RCV_SHUTDOWN)
5441 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
5444 if (signal_pending(current))
5445 goto do_interrupted;
5447 if (sctp_state(asoc, ESTABLISHED))
5450 /* Let another process have a go. Since we are going
5453 sctp_release_sock(sk);
5454 current_timeo = schedule_timeout(current_timeo);
5457 *timeo_p = current_timeo;
5461 finish_wait(&asoc->wait, &wait);
5463 /* Release the association's refcnt. */
5464 sctp_association_put(asoc);
5469 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
5472 err = -ECONNREFUSED;
5476 err = sock_intr_errno(*timeo_p);
5484 static int sctp_wait_for_accept(struct sock *sk, long timeo)
5486 struct sctp_endpoint *ep;
5490 ep = sctp_sk(sk)->ep;
5494 prepare_to_wait_exclusive(sk->sk_sleep, &wait,
5495 TASK_INTERRUPTIBLE);
5497 if (list_empty(&ep->asocs)) {
5498 sctp_release_sock(sk);
5499 timeo = schedule_timeout(timeo);
5504 if (!sctp_sstate(sk, LISTENING))
5508 if (!list_empty(&ep->asocs))
5511 err = sock_intr_errno(timeo);
5512 if (signal_pending(current))
5520 finish_wait(sk->sk_sleep, &wait);
5525 void sctp_wait_for_close(struct sock *sk, long timeout)
5530 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5531 if (list_empty(&sctp_sk(sk)->ep->asocs))
5533 sctp_release_sock(sk);
5534 timeout = schedule_timeout(timeout);
5536 } while (!signal_pending(current) && timeout);
5538 finish_wait(sk->sk_sleep, &wait);
5541 /* Populate the fields of the newsk from the oldsk and migrate the assoc
5542 * and its messages to the newsk.
5544 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
5545 struct sctp_association *assoc,
5546 sctp_socket_type_t type)
5548 struct sctp_sock *oldsp = sctp_sk(oldsk);
5549 struct sctp_sock *newsp = sctp_sk(newsk);
5550 struct sctp_bind_bucket *pp; /* hash list port iterator */
5551 struct sctp_endpoint *newep = newsp->ep;
5552 struct sk_buff *skb, *tmp;
5553 struct sctp_ulpevent *event;
5556 /* Migrate socket buffer sizes and all the socket level options to the
5559 newsk->sk_sndbuf = oldsk->sk_sndbuf;
5560 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
5561 /* Brute force copy old sctp opt. */
5562 inet_sk_copy_descendant(newsk, oldsk);
5564 /* Restore the ep value that was overwritten with the above structure
5570 /* Hook this new socket in to the bind_hash list. */
5571 pp = sctp_sk(oldsk)->bind_hash;
5572 sk_add_bind_node(newsk, &pp->owner);
5573 sctp_sk(newsk)->bind_hash = pp;
5574 inet_sk(newsk)->num = inet_sk(oldsk)->num;
5576 /* Copy the bind_addr list from the original endpoint to the new
5577 * endpoint so that we can handle restarts properly
5579 if (assoc->peer.ipv4_address)
5580 flags |= SCTP_ADDR4_PEERSUPP;
5581 if (assoc->peer.ipv6_address)
5582 flags |= SCTP_ADDR6_PEERSUPP;
5583 sctp_bind_addr_copy(&newsp->ep->base.bind_addr,
5584 &oldsp->ep->base.bind_addr,
5585 SCTP_SCOPE_GLOBAL, GFP_KERNEL, flags);
5587 /* Move any messages in the old socket's receive queue that are for the
5588 * peeled off association to the new socket's receive queue.
5590 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
5591 event = sctp_skb2event(skb);
5592 if (event->asoc == assoc) {
5594 __skb_unlink(skb, &oldsk->sk_receive_queue);
5595 __skb_queue_tail(&newsk->sk_receive_queue, skb);
5596 skb_set_owner_r(skb, newsk);
5600 /* Clean up any messages pending delivery due to partial
5601 * delivery. Three cases:
5602 * 1) No partial deliver; no work.
5603 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
5604 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
5606 skb_queue_head_init(&newsp->pd_lobby);
5607 sctp_sk(newsk)->pd_mode = assoc->ulpq.pd_mode;
5609 if (sctp_sk(oldsk)->pd_mode) {
5610 struct sk_buff_head *queue;
5612 /* Decide which queue to move pd_lobby skbs to. */
5613 if (assoc->ulpq.pd_mode) {
5614 queue = &newsp->pd_lobby;
5616 queue = &newsk->sk_receive_queue;
5618 /* Walk through the pd_lobby, looking for skbs that
5619 * need moved to the new socket.
5621 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
5622 event = sctp_skb2event(skb);
5623 if (event->asoc == assoc) {
5625 __skb_unlink(skb, &oldsp->pd_lobby);
5626 __skb_queue_tail(queue, skb);
5627 skb_set_owner_r(skb, newsk);
5631 /* Clear up any skbs waiting for the partial
5632 * delivery to finish.
5634 if (assoc->ulpq.pd_mode)
5635 sctp_clear_pd(oldsk);
5639 /* Set the type of socket to indicate that it is peeled off from the
5640 * original UDP-style socket or created with the accept() call on a
5641 * TCP-style socket..
5645 /* Mark the new socket "in-use" by the user so that any packets
5646 * that may arrive on the association after we've moved it are
5647 * queued to the backlog. This prevents a potential race between
5648 * backlog processing on the old socket and new-packet processing
5649 * on the new socket.
5651 sctp_lock_sock(newsk);
5652 sctp_assoc_migrate(assoc, newsk);
5654 /* If the association on the newsk is already closed before accept()
5655 * is called, set RCV_SHUTDOWN flag.
5657 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
5658 newsk->sk_shutdown |= RCV_SHUTDOWN;
5660 newsk->sk_state = SCTP_SS_ESTABLISHED;
5661 sctp_release_sock(newsk);
5664 /* This proto struct describes the ULP interface for SCTP. */
5665 struct proto sctp_prot = {
5667 .owner = THIS_MODULE,
5668 .close = sctp_close,
5669 .connect = sctp_connect,
5670 .disconnect = sctp_disconnect,
5671 .accept = sctp_accept,
5672 .ioctl = sctp_ioctl,
5673 .init = sctp_init_sock,
5674 .destroy = sctp_destroy_sock,
5675 .shutdown = sctp_shutdown,
5676 .setsockopt = sctp_setsockopt,
5677 .getsockopt = sctp_getsockopt,
5678 .sendmsg = sctp_sendmsg,
5679 .recvmsg = sctp_recvmsg,
5681 .backlog_rcv = sctp_backlog_rcv,
5683 .unhash = sctp_unhash,
5684 .get_port = sctp_get_port,
5685 .obj_size = sizeof(struct sctp_sock),
5688 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
5689 struct proto sctpv6_prot = {
5691 .owner = THIS_MODULE,
5692 .close = sctp_close,
5693 .connect = sctp_connect,
5694 .disconnect = sctp_disconnect,
5695 .accept = sctp_accept,
5696 .ioctl = sctp_ioctl,
5697 .init = sctp_init_sock,
5698 .destroy = sctp_destroy_sock,
5699 .shutdown = sctp_shutdown,
5700 .setsockopt = sctp_setsockopt,
5701 .getsockopt = sctp_getsockopt,
5702 .sendmsg = sctp_sendmsg,
5703 .recvmsg = sctp_recvmsg,
5705 .backlog_rcv = sctp_backlog_rcv,
5707 .unhash = sctp_unhash,
5708 .get_port = sctp_get_port,
5709 .obj_size = sizeof(struct sctp6_sock),
5711 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */