1 /* SCTP kernel implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
9 * This file is part of the SCTP kernel implementation
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
18 * This SCTP implementation is free software;
19 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
24 * This SCTP implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, write to
32 * the Free Software Foundation, 59 Temple Place - Suite 330,
33 * Boston, MA 02111-1307, USA.
35 * Please send any bug reports or fixes you make to the
37 * lksctp developers <lksctp-developers@lists.sourceforge.net>
39 * Or submit a bug report through the following website:
40 * http://www.sf.net/projects/lksctp
42 * Written or modified by:
43 * La Monte H.P. Yarroll <piggy@acm.org>
44 * Narasimha Budihal <narsi@refcode.org>
45 * Karl Knutson <karl@athena.chicago.il.us>
46 * Jon Grimm <jgrimm@us.ibm.com>
47 * Xingang Guo <xingang.guo@intel.com>
48 * Daisy Chang <daisyc@us.ibm.com>
49 * Sridhar Samudrala <samudrala@us.ibm.com>
50 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
51 * Ardelle Fan <ardelle.fan@intel.com>
52 * Ryan Layer <rmlayer@us.ibm.com>
53 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
54 * Kevin Gao <kevin.gao@intel.com>
56 * Any bugs reported given to us we will try to fix... any fixes shared will
57 * be incorporated into the next SCTP release.
60 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
62 #include <linux/types.h>
63 #include <linux/kernel.h>
64 #include <linux/wait.h>
65 #include <linux/time.h>
67 #include <linux/capability.h>
68 #include <linux/fcntl.h>
69 #include <linux/poll.h>
70 #include <linux/init.h>
71 #include <linux/crypto.h>
72 #include <linux/slab.h>
76 #include <net/route.h>
78 #include <net/inet_common.h>
80 #include <linux/socket.h> /* for sa_family_t */
82 #include <net/sctp/sctp.h>
83 #include <net/sctp/sm.h>
85 /* WARNING: Please do not remove the SCTP_STATIC attribute to
86 * any of the functions below as they are used to export functions
87 * used by a project regression testsuite.
90 /* Forward declarations for internal helper functions. */
91 static int sctp_writeable(struct sock *sk);
92 static void sctp_wfree(struct sk_buff *skb);
93 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
95 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
96 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
97 static int sctp_wait_for_accept(struct sock *sk, long timeo);
98 static void sctp_wait_for_close(struct sock *sk, long timeo);
99 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
100 union sctp_addr *addr, int len);
101 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
102 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
103 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
104 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
105 static int sctp_send_asconf(struct sctp_association *asoc,
106 struct sctp_chunk *chunk);
107 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
108 static int sctp_autobind(struct sock *sk);
109 static void sctp_sock_migrate(struct sock *, struct sock *,
110 struct sctp_association *, sctp_socket_type_t);
111 static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
113 extern struct kmem_cache *sctp_bucket_cachep;
114 extern long sysctl_sctp_mem[3];
115 extern int sysctl_sctp_rmem[3];
116 extern int sysctl_sctp_wmem[3];
118 static int sctp_memory_pressure;
119 static atomic_long_t sctp_memory_allocated;
120 struct percpu_counter sctp_sockets_allocated;
122 static void sctp_enter_memory_pressure(struct sock *sk)
124 sctp_memory_pressure = 1;
128 /* Get the sndbuf space available at the time on the association. */
129 static inline int sctp_wspace(struct sctp_association *asoc)
133 if (asoc->ep->sndbuf_policy)
134 amt = asoc->sndbuf_used;
136 amt = sk_wmem_alloc_get(asoc->base.sk);
138 if (amt >= asoc->base.sk->sk_sndbuf) {
139 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
142 amt = sk_stream_wspace(asoc->base.sk);
147 amt = asoc->base.sk->sk_sndbuf - amt;
152 /* Increment the used sndbuf space count of the corresponding association by
153 * the size of the outgoing data chunk.
154 * Also, set the skb destructor for sndbuf accounting later.
156 * Since it is always 1-1 between chunk and skb, and also a new skb is always
157 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
158 * destructor in the data chunk skb for the purpose of the sndbuf space
161 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
163 struct sctp_association *asoc = chunk->asoc;
164 struct sock *sk = asoc->base.sk;
166 /* The sndbuf space is tracked per association. */
167 sctp_association_hold(asoc);
169 skb_set_owner_w(chunk->skb, sk);
171 chunk->skb->destructor = sctp_wfree;
172 /* Save the chunk pointer in skb for sctp_wfree to use later. */
173 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
175 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
176 sizeof(struct sk_buff) +
177 sizeof(struct sctp_chunk);
179 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
180 sk->sk_wmem_queued += chunk->skb->truesize;
181 sk_mem_charge(sk, chunk->skb->truesize);
184 /* Verify that this is a valid address. */
185 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
190 /* Verify basic sockaddr. */
191 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
195 /* Is this a valid SCTP address? */
196 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
199 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
205 /* Look up the association by its id. If this is not a UDP-style
206 * socket, the ID field is always ignored.
208 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
210 struct sctp_association *asoc = NULL;
212 /* If this is not a UDP-style socket, assoc id should be ignored. */
213 if (!sctp_style(sk, UDP)) {
214 /* Return NULL if the socket state is not ESTABLISHED. It
215 * could be a TCP-style listening socket or a socket which
216 * hasn't yet called connect() to establish an association.
218 if (!sctp_sstate(sk, ESTABLISHED))
221 /* Get the first and the only association from the list. */
222 if (!list_empty(&sctp_sk(sk)->ep->asocs))
223 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
224 struct sctp_association, asocs);
228 /* Otherwise this is a UDP-style socket. */
229 if (!id || (id == (sctp_assoc_t)-1))
232 spin_lock_bh(&sctp_assocs_id_lock);
233 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
234 spin_unlock_bh(&sctp_assocs_id_lock);
236 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
242 /* Look up the transport from an address and an assoc id. If both address and
243 * id are specified, the associations matching the address and the id should be
246 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
247 struct sockaddr_storage *addr,
250 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
251 struct sctp_transport *transport;
252 union sctp_addr *laddr = (union sctp_addr *)addr;
254 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
261 id_asoc = sctp_id2assoc(sk, id);
262 if (id_asoc && (id_asoc != addr_asoc))
265 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
266 (union sctp_addr *)addr);
271 /* API 3.1.2 bind() - UDP Style Syntax
272 * The syntax of bind() is,
274 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
276 * sd - the socket descriptor returned by socket().
277 * addr - the address structure (struct sockaddr_in or struct
278 * sockaddr_in6 [RFC 2553]),
279 * addr_len - the size of the address structure.
281 SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
287 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
290 /* Disallow binding twice. */
291 if (!sctp_sk(sk)->ep->base.bind_addr.port)
292 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
297 sctp_release_sock(sk);
302 static long sctp_get_port_local(struct sock *, union sctp_addr *);
304 /* Verify this is a valid sockaddr. */
305 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
306 union sctp_addr *addr, int len)
310 /* Check minimum size. */
311 if (len < sizeof (struct sockaddr))
314 /* V4 mapped address are really of AF_INET family */
315 if (addr->sa.sa_family == AF_INET6 &&
316 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
317 if (!opt->pf->af_supported(AF_INET, opt))
320 /* Does this PF support this AF? */
321 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
325 /* If we get this far, af is valid. */
326 af = sctp_get_af_specific(addr->sa.sa_family);
328 if (len < af->sockaddr_len)
334 /* Bind a local address either to an endpoint or to an association. */
335 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
337 struct sctp_sock *sp = sctp_sk(sk);
338 struct sctp_endpoint *ep = sp->ep;
339 struct sctp_bind_addr *bp = &ep->base.bind_addr;
344 /* Common sockaddr verification. */
345 af = sctp_sockaddr_af(sp, addr, len);
347 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
352 snum = ntohs(addr->v4.sin_port);
354 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
355 ", port: %d, new port: %d, len: %d)\n",
361 /* PF specific bind() address verification. */
362 if (!sp->pf->bind_verify(sp, addr))
363 return -EADDRNOTAVAIL;
365 /* We must either be unbound, or bind to the same port.
366 * It's OK to allow 0 ports if we are already bound.
367 * We'll just inhert an already bound port in this case
372 else if (snum != bp->port) {
373 SCTP_DEBUG_PRINTK("sctp_do_bind:"
374 " New port %d does not match existing port "
375 "%d.\n", snum, bp->port);
380 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
383 /* See if the address matches any of the addresses we may have
384 * already bound before checking against other endpoints.
386 if (sctp_bind_addr_match(bp, addr, sp))
389 /* Make sure we are allowed to bind here.
390 * The function sctp_get_port_local() does duplicate address
393 addr->v4.sin_port = htons(snum);
394 if ((ret = sctp_get_port_local(sk, addr))) {
398 /* Refresh ephemeral port. */
400 bp->port = inet_sk(sk)->inet_num;
402 /* Add the address to the bind address list.
403 * Use GFP_ATOMIC since BHs will be disabled.
405 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
407 /* Copy back into socket for getsockname() use. */
409 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
410 af->to_sk_saddr(addr, sk);
416 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
418 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
419 * at any one time. If a sender, after sending an ASCONF chunk, decides
420 * it needs to transfer another ASCONF Chunk, it MUST wait until the
421 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
422 * subsequent ASCONF. Note this restriction binds each side, so at any
423 * time two ASCONF may be in-transit on any given association (one sent
424 * from each endpoint).
426 static int sctp_send_asconf(struct sctp_association *asoc,
427 struct sctp_chunk *chunk)
431 /* If there is an outstanding ASCONF chunk, queue it for later
434 if (asoc->addip_last_asconf) {
435 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
439 /* Hold the chunk until an ASCONF_ACK is received. */
440 sctp_chunk_hold(chunk);
441 retval = sctp_primitive_ASCONF(asoc, chunk);
443 sctp_chunk_free(chunk);
445 asoc->addip_last_asconf = chunk;
451 /* Add a list of addresses as bind addresses to local endpoint or
454 * Basically run through each address specified in the addrs/addrcnt
455 * array/length pair, determine if it is IPv6 or IPv4 and call
456 * sctp_do_bind() on it.
458 * If any of them fails, then the operation will be reversed and the
459 * ones that were added will be removed.
461 * Only sctp_setsockopt_bindx() is supposed to call this function.
463 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
468 struct sockaddr *sa_addr;
471 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
475 for (cnt = 0; cnt < addrcnt; cnt++) {
476 /* The list may contain either IPv4 or IPv6 address;
477 * determine the address length for walking thru the list.
479 sa_addr = (struct sockaddr *)addr_buf;
480 af = sctp_get_af_specific(sa_addr->sa_family);
486 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
489 addr_buf += af->sockaddr_len;
493 /* Failed. Cleanup the ones that have been added */
495 sctp_bindx_rem(sk, addrs, cnt);
503 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
504 * associations that are part of the endpoint indicating that a list of local
505 * addresses are added to the endpoint.
507 * If any of the addresses is already in the bind address list of the
508 * association, we do not send the chunk for that association. But it will not
509 * affect other associations.
511 * Only sctp_setsockopt_bindx() is supposed to call this function.
513 static int sctp_send_asconf_add_ip(struct sock *sk,
514 struct sockaddr *addrs,
517 struct sctp_sock *sp;
518 struct sctp_endpoint *ep;
519 struct sctp_association *asoc;
520 struct sctp_bind_addr *bp;
521 struct sctp_chunk *chunk;
522 struct sctp_sockaddr_entry *laddr;
523 union sctp_addr *addr;
524 union sctp_addr saveaddr;
531 if (!sctp_addip_enable)
537 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
538 __func__, sk, addrs, addrcnt);
540 list_for_each_entry(asoc, &ep->asocs, asocs) {
542 if (!asoc->peer.asconf_capable)
545 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
548 if (!sctp_state(asoc, ESTABLISHED))
551 /* Check if any address in the packed array of addresses is
552 * in the bind address list of the association. If so,
553 * do not send the asconf chunk to its peer, but continue with
554 * other associations.
557 for (i = 0; i < addrcnt; i++) {
558 addr = (union sctp_addr *)addr_buf;
559 af = sctp_get_af_specific(addr->v4.sin_family);
565 if (sctp_assoc_lookup_laddr(asoc, addr))
568 addr_buf += af->sockaddr_len;
573 /* Use the first valid address in bind addr list of
574 * association as Address Parameter of ASCONF CHUNK.
576 bp = &asoc->base.bind_addr;
577 p = bp->address_list.next;
578 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
579 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
580 addrcnt, SCTP_PARAM_ADD_IP);
586 retval = sctp_send_asconf(asoc, chunk);
590 /* Add the new addresses to the bind address list with
591 * use_as_src set to 0.
594 for (i = 0; i < addrcnt; i++) {
595 addr = (union sctp_addr *)addr_buf;
596 af = sctp_get_af_specific(addr->v4.sin_family);
597 memcpy(&saveaddr, addr, af->sockaddr_len);
598 retval = sctp_add_bind_addr(bp, &saveaddr,
599 SCTP_ADDR_NEW, GFP_ATOMIC);
600 addr_buf += af->sockaddr_len;
608 /* Remove a list of addresses from bind addresses list. Do not remove the
611 * Basically run through each address specified in the addrs/addrcnt
612 * array/length pair, determine if it is IPv6 or IPv4 and call
613 * sctp_del_bind() on it.
615 * If any of them fails, then the operation will be reversed and the
616 * ones that were removed will be added back.
618 * At least one address has to be left; if only one address is
619 * available, the operation will return -EBUSY.
621 * Only sctp_setsockopt_bindx() is supposed to call this function.
623 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
625 struct sctp_sock *sp = sctp_sk(sk);
626 struct sctp_endpoint *ep = sp->ep;
628 struct sctp_bind_addr *bp = &ep->base.bind_addr;
631 union sctp_addr *sa_addr;
634 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
638 for (cnt = 0; cnt < addrcnt; cnt++) {
639 /* If the bind address list is empty or if there is only one
640 * bind address, there is nothing more to be removed (we need
641 * at least one address here).
643 if (list_empty(&bp->address_list) ||
644 (sctp_list_single_entry(&bp->address_list))) {
649 sa_addr = (union sctp_addr *)addr_buf;
650 af = sctp_get_af_specific(sa_addr->sa.sa_family);
656 if (!af->addr_valid(sa_addr, sp, NULL)) {
657 retval = -EADDRNOTAVAIL;
661 if (sa_addr->v4.sin_port &&
662 sa_addr->v4.sin_port != htons(bp->port)) {
667 if (!sa_addr->v4.sin_port)
668 sa_addr->v4.sin_port = htons(bp->port);
670 /* FIXME - There is probably a need to check if sk->sk_saddr and
671 * sk->sk_rcv_addr are currently set to one of the addresses to
672 * be removed. This is something which needs to be looked into
673 * when we are fixing the outstanding issues with multi-homing
674 * socket routing and failover schemes. Refer to comments in
675 * sctp_do_bind(). -daisy
677 retval = sctp_del_bind_addr(bp, sa_addr);
679 addr_buf += af->sockaddr_len;
682 /* Failed. Add the ones that has been removed back */
684 sctp_bindx_add(sk, addrs, cnt);
692 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
693 * the associations that are part of the endpoint indicating that a list of
694 * local addresses are removed from the endpoint.
696 * If any of the addresses is already in the bind address list of the
697 * association, we do not send the chunk for that association. But it will not
698 * affect other associations.
700 * Only sctp_setsockopt_bindx() is supposed to call this function.
702 static int sctp_send_asconf_del_ip(struct sock *sk,
703 struct sockaddr *addrs,
706 struct sctp_sock *sp;
707 struct sctp_endpoint *ep;
708 struct sctp_association *asoc;
709 struct sctp_transport *transport;
710 struct sctp_bind_addr *bp;
711 struct sctp_chunk *chunk;
712 union sctp_addr *laddr;
715 struct sctp_sockaddr_entry *saddr;
719 if (!sctp_addip_enable)
725 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
726 __func__, sk, addrs, addrcnt);
728 list_for_each_entry(asoc, &ep->asocs, asocs) {
730 if (!asoc->peer.asconf_capable)
733 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
736 if (!sctp_state(asoc, ESTABLISHED))
739 /* Check if any address in the packed array of addresses is
740 * not present in the bind address list of the association.
741 * If so, do not send the asconf chunk to its peer, but
742 * continue with other associations.
745 for (i = 0; i < addrcnt; i++) {
746 laddr = (union sctp_addr *)addr_buf;
747 af = sctp_get_af_specific(laddr->v4.sin_family);
753 if (!sctp_assoc_lookup_laddr(asoc, laddr))
756 addr_buf += af->sockaddr_len;
761 /* Find one address in the association's bind address list
762 * that is not in the packed array of addresses. This is to
763 * make sure that we do not delete all the addresses in the
766 bp = &asoc->base.bind_addr;
767 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
772 /* We do not need RCU protection throughout this loop
773 * because this is done under a socket lock from the
776 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
783 /* Reset use_as_src flag for the addresses in the bind address
784 * list that are to be deleted.
787 for (i = 0; i < addrcnt; i++) {
788 laddr = (union sctp_addr *)addr_buf;
789 af = sctp_get_af_specific(laddr->v4.sin_family);
790 list_for_each_entry(saddr, &bp->address_list, list) {
791 if (sctp_cmp_addr_exact(&saddr->a, laddr))
792 saddr->state = SCTP_ADDR_DEL;
794 addr_buf += af->sockaddr_len;
797 /* Update the route and saddr entries for all the transports
798 * as some of the addresses in the bind address list are
799 * about to be deleted and cannot be used as source addresses.
801 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
803 dst_release(transport->dst);
804 sctp_transport_route(transport, NULL,
805 sctp_sk(asoc->base.sk));
808 retval = sctp_send_asconf(asoc, chunk);
814 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
817 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
820 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
821 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
824 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
825 * Section 3.1.2 for this usage.
827 * addrs is a pointer to an array of one or more socket addresses. Each
828 * address is contained in its appropriate structure (i.e. struct
829 * sockaddr_in or struct sockaddr_in6) the family of the address type
830 * must be used to distinguish the address length (note that this
831 * representation is termed a "packed array" of addresses). The caller
832 * specifies the number of addresses in the array with addrcnt.
834 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
835 * -1, and sets errno to the appropriate error code.
837 * For SCTP, the port given in each socket address must be the same, or
838 * sctp_bindx() will fail, setting errno to EINVAL.
840 * The flags parameter is formed from the bitwise OR of zero or more of
841 * the following currently defined flags:
843 * SCTP_BINDX_ADD_ADDR
845 * SCTP_BINDX_REM_ADDR
847 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
848 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
849 * addresses from the association. The two flags are mutually exclusive;
850 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
851 * not remove all addresses from an association; sctp_bindx() will
852 * reject such an attempt with EINVAL.
854 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
855 * additional addresses with an endpoint after calling bind(). Or use
856 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
857 * socket is associated with so that no new association accepted will be
858 * associated with those addresses. If the endpoint supports dynamic
859 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
860 * endpoint to send the appropriate message to the peer to change the
861 * peers address lists.
863 * Adding and removing addresses from a connected association is
864 * optional functionality. Implementations that do not support this
865 * functionality should return EOPNOTSUPP.
867 * Basically do nothing but copying the addresses from user to kernel
868 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
869 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
872 * We don't use copy_from_user() for optimization: we first do the
873 * sanity checks (buffer size -fast- and access check-healthy
874 * pointer); if all of those succeed, then we can alloc the memory
875 * (expensive operation) needed to copy the data to kernel. Then we do
876 * the copying without checking the user space area
877 * (__copy_from_user()).
879 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
882 * sk The sk of the socket
883 * addrs The pointer to the addresses in user land
884 * addrssize Size of the addrs buffer
885 * op Operation to perform (add or remove, see the flags of
888 * Returns 0 if ok, <0 errno code on error.
890 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
891 struct sockaddr __user *addrs,
892 int addrs_size, int op)
894 struct sockaddr *kaddrs;
898 struct sockaddr *sa_addr;
902 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
903 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
905 if (unlikely(addrs_size <= 0))
908 /* Check the user passed a healthy pointer. */
909 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
912 /* Alloc space for the address array in kernel memory. */
913 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
914 if (unlikely(!kaddrs))
917 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
922 /* Walk through the addrs buffer and count the number of addresses. */
924 while (walk_size < addrs_size) {
925 if (walk_size + sizeof(sa_family_t) > addrs_size) {
930 sa_addr = (struct sockaddr *)addr_buf;
931 af = sctp_get_af_specific(sa_addr->sa_family);
933 /* If the address family is not supported or if this address
934 * causes the address buffer to overflow return EINVAL.
936 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
941 addr_buf += af->sockaddr_len;
942 walk_size += af->sockaddr_len;
947 case SCTP_BINDX_ADD_ADDR:
948 err = sctp_bindx_add(sk, kaddrs, addrcnt);
951 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
954 case SCTP_BINDX_REM_ADDR:
955 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
958 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
972 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
974 * Common routine for handling connect() and sctp_connectx().
975 * Connect will come in with just a single address.
977 static int __sctp_connect(struct sock* sk,
978 struct sockaddr *kaddrs,
980 sctp_assoc_t *assoc_id)
982 struct sctp_sock *sp;
983 struct sctp_endpoint *ep;
984 struct sctp_association *asoc = NULL;
985 struct sctp_association *asoc2;
986 struct sctp_transport *transport;
994 union sctp_addr *sa_addr = NULL;
997 unsigned int f_flags = 0;
1002 /* connect() cannot be done on a socket that is already in ESTABLISHED
1003 * state - UDP-style peeled off socket or a TCP-style socket that
1004 * is already connected.
1005 * It cannot be done even on a TCP-style listening socket.
1007 if (sctp_sstate(sk, ESTABLISHED) ||
1008 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1013 /* Walk through the addrs buffer and count the number of addresses. */
1015 while (walk_size < addrs_size) {
1016 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1021 sa_addr = (union sctp_addr *)addr_buf;
1022 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1024 /* If the address family is not supported or if this address
1025 * causes the address buffer to overflow return EINVAL.
1027 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1032 port = ntohs(sa_addr->v4.sin_port);
1034 /* Save current address so we can work with it */
1035 memcpy(&to, sa_addr, af->sockaddr_len);
1037 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1041 /* Make sure the destination port is correctly set
1044 if (asoc && asoc->peer.port && asoc->peer.port != port)
1048 /* Check if there already is a matching association on the
1049 * endpoint (other than the one created here).
1051 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1052 if (asoc2 && asoc2 != asoc) {
1053 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1060 /* If we could not find a matching association on the endpoint,
1061 * make sure that there is no peeled-off association matching
1062 * the peer address even on another socket.
1064 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1065 err = -EADDRNOTAVAIL;
1070 /* If a bind() or sctp_bindx() is not called prior to
1071 * an sctp_connectx() call, the system picks an
1072 * ephemeral port and will choose an address set
1073 * equivalent to binding with a wildcard address.
1075 if (!ep->base.bind_addr.port) {
1076 if (sctp_autobind(sk)) {
1082 * If an unprivileged user inherits a 1-many
1083 * style socket with open associations on a
1084 * privileged port, it MAY be permitted to
1085 * accept new associations, but it SHOULD NOT
1086 * be permitted to open new associations.
1088 if (ep->base.bind_addr.port < PROT_SOCK &&
1089 !capable(CAP_NET_BIND_SERVICE)) {
1095 scope = sctp_scope(&to);
1096 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1102 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1110 /* Prime the peer's transport structures. */
1111 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1119 addr_buf += af->sockaddr_len;
1120 walk_size += af->sockaddr_len;
1123 /* In case the user of sctp_connectx() wants an association
1124 * id back, assign one now.
1127 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1132 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1137 /* Initialize sk's dport and daddr for getpeername() */
1138 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1139 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1140 af->to_sk_daddr(sa_addr, sk);
1143 /* in-kernel sockets don't generally have a file allocated to them
1144 * if all they do is call sock_create_kern().
1146 if (sk->sk_socket->file)
1147 f_flags = sk->sk_socket->file->f_flags;
1149 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1151 err = sctp_wait_for_connect(asoc, &timeo);
1152 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1153 *assoc_id = asoc->assoc_id;
1155 /* Don't free association on exit. */
1160 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1161 " kaddrs: %p err: %d\n",
1164 sctp_association_free(asoc);
1168 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1171 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1172 * sctp_assoc_t *asoc);
1174 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1175 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1176 * or IPv6 addresses.
1178 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1179 * Section 3.1.2 for this usage.
1181 * addrs is a pointer to an array of one or more socket addresses. Each
1182 * address is contained in its appropriate structure (i.e. struct
1183 * sockaddr_in or struct sockaddr_in6) the family of the address type
1184 * must be used to distengish the address length (note that this
1185 * representation is termed a "packed array" of addresses). The caller
1186 * specifies the number of addresses in the array with addrcnt.
1188 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1189 * the association id of the new association. On failure, sctp_connectx()
1190 * returns -1, and sets errno to the appropriate error code. The assoc_id
1191 * is not touched by the kernel.
1193 * For SCTP, the port given in each socket address must be the same, or
1194 * sctp_connectx() will fail, setting errno to EINVAL.
1196 * An application can use sctp_connectx to initiate an association with
1197 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1198 * allows a caller to specify multiple addresses at which a peer can be
1199 * reached. The way the SCTP stack uses the list of addresses to set up
1200 * the association is implementation dependent. This function only
1201 * specifies that the stack will try to make use of all the addresses in
1202 * the list when needed.
1204 * Note that the list of addresses passed in is only used for setting up
1205 * the association. It does not necessarily equal the set of addresses
1206 * the peer uses for the resulting association. If the caller wants to
1207 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1208 * retrieve them after the association has been set up.
1210 * Basically do nothing but copying the addresses from user to kernel
1211 * land and invoking either sctp_connectx(). This is used for tunneling
1212 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1214 * We don't use copy_from_user() for optimization: we first do the
1215 * sanity checks (buffer size -fast- and access check-healthy
1216 * pointer); if all of those succeed, then we can alloc the memory
1217 * (expensive operation) needed to copy the data to kernel. Then we do
1218 * the copying without checking the user space area
1219 * (__copy_from_user()).
1221 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1224 * sk The sk of the socket
1225 * addrs The pointer to the addresses in user land
1226 * addrssize Size of the addrs buffer
1228 * Returns >=0 if ok, <0 errno code on error.
1230 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1231 struct sockaddr __user *addrs,
1233 sctp_assoc_t *assoc_id)
1236 struct sockaddr *kaddrs;
1238 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1239 __func__, sk, addrs, addrs_size);
1241 if (unlikely(addrs_size <= 0))
1244 /* Check the user passed a healthy pointer. */
1245 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1248 /* Alloc space for the address array in kernel memory. */
1249 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1250 if (unlikely(!kaddrs))
1253 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1256 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1265 * This is an older interface. It's kept for backward compatibility
1266 * to the option that doesn't provide association id.
1268 SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1269 struct sockaddr __user *addrs,
1272 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1276 * New interface for the API. The since the API is done with a socket
1277 * option, to make it simple we feed back the association id is as a return
1278 * indication to the call. Error is always negative and association id is
1281 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1282 struct sockaddr __user *addrs,
1285 sctp_assoc_t assoc_id = 0;
1288 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1297 * New (hopefully final) interface for the API.
1298 * We use the sctp_getaddrs_old structure so that use-space library
1299 * can avoid any unnecessary allocations. The only defferent part
1300 * is that we store the actual length of the address buffer into the
1301 * addrs_num structure member. That way we can re-use the existing
1304 SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1305 char __user *optval,
1308 struct sctp_getaddrs_old param;
1309 sctp_assoc_t assoc_id = 0;
1312 if (len < sizeof(param))
1315 if (copy_from_user(¶m, optval, sizeof(param)))
1318 err = __sctp_setsockopt_connectx(sk,
1319 (struct sockaddr __user *)param.addrs,
1320 param.addr_num, &assoc_id);
1322 if (err == 0 || err == -EINPROGRESS) {
1323 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1325 if (put_user(sizeof(assoc_id), optlen))
1332 /* API 3.1.4 close() - UDP Style Syntax
1333 * Applications use close() to perform graceful shutdown (as described in
1334 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1335 * by a UDP-style socket.
1339 * ret = close(int sd);
1341 * sd - the socket descriptor of the associations to be closed.
1343 * To gracefully shutdown a specific association represented by the
1344 * UDP-style socket, an application should use the sendmsg() call,
1345 * passing no user data, but including the appropriate flag in the
1346 * ancillary data (see Section xxxx).
1348 * If sd in the close() call is a branched-off socket representing only
1349 * one association, the shutdown is performed on that association only.
1351 * 4.1.6 close() - TCP Style Syntax
1353 * Applications use close() to gracefully close down an association.
1357 * int close(int sd);
1359 * sd - the socket descriptor of the association to be closed.
1361 * After an application calls close() on a socket descriptor, no further
1362 * socket operations will succeed on that descriptor.
1364 * API 7.1.4 SO_LINGER
1366 * An application using the TCP-style socket can use this option to
1367 * perform the SCTP ABORT primitive. The linger option structure is:
1370 * int l_onoff; // option on/off
1371 * int l_linger; // linger time
1374 * To enable the option, set l_onoff to 1. If the l_linger value is set
1375 * to 0, calling close() is the same as the ABORT primitive. If the
1376 * value is set to a negative value, the setsockopt() call will return
1377 * an error. If the value is set to a positive value linger_time, the
1378 * close() can be blocked for at most linger_time ms. If the graceful
1379 * shutdown phase does not finish during this period, close() will
1380 * return but the graceful shutdown phase continues in the system.
1382 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1384 struct sctp_endpoint *ep;
1385 struct sctp_association *asoc;
1386 struct list_head *pos, *temp;
1387 unsigned int data_was_unread;
1389 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1392 sk->sk_shutdown = SHUTDOWN_MASK;
1393 sk->sk_state = SCTP_SS_CLOSING;
1395 ep = sctp_sk(sk)->ep;
1397 /* Clean up any skbs sitting on the receive queue. */
1398 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1399 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1401 /* Walk all associations on an endpoint. */
1402 list_for_each_safe(pos, temp, &ep->asocs) {
1403 asoc = list_entry(pos, struct sctp_association, asocs);
1405 if (sctp_style(sk, TCP)) {
1406 /* A closed association can still be in the list if
1407 * it belongs to a TCP-style listening socket that is
1408 * not yet accepted. If so, free it. If not, send an
1409 * ABORT or SHUTDOWN based on the linger options.
1411 if (sctp_state(asoc, CLOSED)) {
1412 sctp_unhash_established(asoc);
1413 sctp_association_free(asoc);
1418 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1419 !skb_queue_empty(&asoc->ulpq.reasm) ||
1420 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1421 struct sctp_chunk *chunk;
1423 chunk = sctp_make_abort_user(asoc, NULL, 0);
1425 sctp_primitive_ABORT(asoc, chunk);
1427 sctp_primitive_SHUTDOWN(asoc, NULL);
1430 /* On a TCP-style socket, block for at most linger_time if set. */
1431 if (sctp_style(sk, TCP) && timeout)
1432 sctp_wait_for_close(sk, timeout);
1434 /* This will run the backlog queue. */
1435 sctp_release_sock(sk);
1437 /* Supposedly, no process has access to the socket, but
1438 * the net layers still may.
1440 sctp_local_bh_disable();
1441 sctp_bh_lock_sock(sk);
1443 /* Hold the sock, since sk_common_release() will put sock_put()
1444 * and we have just a little more cleanup.
1447 sk_common_release(sk);
1449 sctp_bh_unlock_sock(sk);
1450 sctp_local_bh_enable();
1454 SCTP_DBG_OBJCNT_DEC(sock);
1457 /* Handle EPIPE error. */
1458 static int sctp_error(struct sock *sk, int flags, int err)
1461 err = sock_error(sk) ? : -EPIPE;
1462 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1463 send_sig(SIGPIPE, current, 0);
1467 /* API 3.1.3 sendmsg() - UDP Style Syntax
1469 * An application uses sendmsg() and recvmsg() calls to transmit data to
1470 * and receive data from its peer.
1472 * ssize_t sendmsg(int socket, const struct msghdr *message,
1475 * socket - the socket descriptor of the endpoint.
1476 * message - pointer to the msghdr structure which contains a single
1477 * user message and possibly some ancillary data.
1479 * See Section 5 for complete description of the data
1482 * flags - flags sent or received with the user message, see Section
1483 * 5 for complete description of the flags.
1485 * Note: This function could use a rewrite especially when explicit
1486 * connect support comes in.
1488 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1490 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1492 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1493 struct msghdr *msg, size_t msg_len)
1495 struct sctp_sock *sp;
1496 struct sctp_endpoint *ep;
1497 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1498 struct sctp_transport *transport, *chunk_tp;
1499 struct sctp_chunk *chunk;
1501 struct sockaddr *msg_name = NULL;
1502 struct sctp_sndrcvinfo default_sinfo;
1503 struct sctp_sndrcvinfo *sinfo;
1504 struct sctp_initmsg *sinit;
1505 sctp_assoc_t associd = 0;
1506 sctp_cmsgs_t cmsgs = { NULL };
1510 __u16 sinfo_flags = 0;
1511 struct sctp_datamsg *datamsg;
1512 int msg_flags = msg->msg_flags;
1514 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1521 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1523 /* We cannot send a message over a TCP-style listening socket. */
1524 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1529 /* Parse out the SCTP CMSGs. */
1530 err = sctp_msghdr_parse(msg, &cmsgs);
1533 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1537 /* Fetch the destination address for this packet. This
1538 * address only selects the association--it is not necessarily
1539 * the address we will send to.
1540 * For a peeled-off socket, msg_name is ignored.
1542 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1543 int msg_namelen = msg->msg_namelen;
1545 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1550 if (msg_namelen > sizeof(to))
1551 msg_namelen = sizeof(to);
1552 memcpy(&to, msg->msg_name, msg_namelen);
1553 msg_name = msg->msg_name;
1559 /* Did the user specify SNDRCVINFO? */
1561 sinfo_flags = sinfo->sinfo_flags;
1562 associd = sinfo->sinfo_assoc_id;
1565 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1566 msg_len, sinfo_flags);
1568 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1569 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1574 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1575 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1576 * If SCTP_ABORT is set, the message length could be non zero with
1577 * the msg_iov set to the user abort reason.
1579 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1580 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1585 /* If SCTP_ADDR_OVER is set, there must be an address
1586 * specified in msg_name.
1588 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1595 SCTP_DEBUG_PRINTK("About to look up association.\n");
1599 /* If a msg_name has been specified, assume this is to be used. */
1601 /* Look for a matching association on the endpoint. */
1602 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1604 /* If we could not find a matching association on the
1605 * endpoint, make sure that it is not a TCP-style
1606 * socket that already has an association or there is
1607 * no peeled-off association on another socket.
1609 if ((sctp_style(sk, TCP) &&
1610 sctp_sstate(sk, ESTABLISHED)) ||
1611 sctp_endpoint_is_peeled_off(ep, &to)) {
1612 err = -EADDRNOTAVAIL;
1617 asoc = sctp_id2assoc(sk, associd);
1625 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1627 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1628 * socket that has an association in CLOSED state. This can
1629 * happen when an accepted socket has an association that is
1632 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1637 if (sinfo_flags & SCTP_EOF) {
1638 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1640 sctp_primitive_SHUTDOWN(asoc, NULL);
1644 if (sinfo_flags & SCTP_ABORT) {
1646 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1652 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1653 sctp_primitive_ABORT(asoc, chunk);
1659 /* Do we need to create the association? */
1661 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1663 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1668 /* Check for invalid stream against the stream counts,
1669 * either the default or the user specified stream counts.
1672 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1673 /* Check against the defaults. */
1674 if (sinfo->sinfo_stream >=
1675 sp->initmsg.sinit_num_ostreams) {
1680 /* Check against the requested. */
1681 if (sinfo->sinfo_stream >=
1682 sinit->sinit_num_ostreams) {
1690 * API 3.1.2 bind() - UDP Style Syntax
1691 * If a bind() or sctp_bindx() is not called prior to a
1692 * sendmsg() call that initiates a new association, the
1693 * system picks an ephemeral port and will choose an address
1694 * set equivalent to binding with a wildcard address.
1696 if (!ep->base.bind_addr.port) {
1697 if (sctp_autobind(sk)) {
1703 * If an unprivileged user inherits a one-to-many
1704 * style socket with open associations on a privileged
1705 * port, it MAY be permitted to accept new associations,
1706 * but it SHOULD NOT be permitted to open new
1709 if (ep->base.bind_addr.port < PROT_SOCK &&
1710 !capable(CAP_NET_BIND_SERVICE)) {
1716 scope = sctp_scope(&to);
1717 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1723 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1729 /* If the SCTP_INIT ancillary data is specified, set all
1730 * the association init values accordingly.
1733 if (sinit->sinit_num_ostreams) {
1734 asoc->c.sinit_num_ostreams =
1735 sinit->sinit_num_ostreams;
1737 if (sinit->sinit_max_instreams) {
1738 asoc->c.sinit_max_instreams =
1739 sinit->sinit_max_instreams;
1741 if (sinit->sinit_max_attempts) {
1742 asoc->max_init_attempts
1743 = sinit->sinit_max_attempts;
1745 if (sinit->sinit_max_init_timeo) {
1746 asoc->max_init_timeo =
1747 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1751 /* Prime the peer's transport structures. */
1752 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1759 /* ASSERT: we have a valid association at this point. */
1760 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1763 /* If the user didn't specify SNDRCVINFO, make up one with
1766 memset(&default_sinfo, 0, sizeof(default_sinfo));
1767 default_sinfo.sinfo_stream = asoc->default_stream;
1768 default_sinfo.sinfo_flags = asoc->default_flags;
1769 default_sinfo.sinfo_ppid = asoc->default_ppid;
1770 default_sinfo.sinfo_context = asoc->default_context;
1771 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1772 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1773 sinfo = &default_sinfo;
1776 /* API 7.1.7, the sndbuf size per association bounds the
1777 * maximum size of data that can be sent in a single send call.
1779 if (msg_len > sk->sk_sndbuf) {
1784 if (asoc->pmtu_pending)
1785 sctp_assoc_pending_pmtu(asoc);
1787 /* If fragmentation is disabled and the message length exceeds the
1788 * association fragmentation point, return EMSGSIZE. The I-D
1789 * does not specify what this error is, but this looks like
1792 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1797 /* Check for invalid stream. */
1798 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1803 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1804 if (!sctp_wspace(asoc)) {
1805 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1810 /* If an address is passed with the sendto/sendmsg call, it is used
1811 * to override the primary destination address in the TCP model, or
1812 * when SCTP_ADDR_OVER flag is set in the UDP model.
1814 if ((sctp_style(sk, TCP) && msg_name) ||
1815 (sinfo_flags & SCTP_ADDR_OVER)) {
1816 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1824 /* Auto-connect, if we aren't connected already. */
1825 if (sctp_state(asoc, CLOSED)) {
1826 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1829 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1832 /* Break the message into multiple chunks of maximum size. */
1833 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1839 /* Now send the (possibly) fragmented message. */
1840 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1841 sctp_chunk_hold(chunk);
1843 /* Do accounting for the write space. */
1844 sctp_set_owner_w(chunk);
1846 chunk->transport = chunk_tp;
1849 /* Send it to the lower layers. Note: all chunks
1850 * must either fail or succeed. The lower layer
1851 * works that way today. Keep it that way or this
1854 err = sctp_primitive_SEND(asoc, datamsg);
1855 /* Did the lower layer accept the chunk? */
1857 sctp_datamsg_free(datamsg);
1859 sctp_datamsg_put(datamsg);
1861 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1868 /* If we are already past ASSOCIATE, the lower
1869 * layers are responsible for association cleanup.
1875 sctp_association_free(asoc);
1877 sctp_release_sock(sk);
1880 return sctp_error(sk, msg_flags, err);
1887 err = sock_error(sk);
1897 /* This is an extended version of skb_pull() that removes the data from the
1898 * start of a skb even when data is spread across the list of skb's in the
1899 * frag_list. len specifies the total amount of data that needs to be removed.
1900 * when 'len' bytes could be removed from the skb, it returns 0.
1901 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1902 * could not be removed.
1904 static int sctp_skb_pull(struct sk_buff *skb, int len)
1906 struct sk_buff *list;
1907 int skb_len = skb_headlen(skb);
1910 if (len <= skb_len) {
1911 __skb_pull(skb, len);
1915 __skb_pull(skb, skb_len);
1917 skb_walk_frags(skb, list) {
1918 rlen = sctp_skb_pull(list, len);
1919 skb->len -= (len-rlen);
1920 skb->data_len -= (len-rlen);
1931 /* API 3.1.3 recvmsg() - UDP Style Syntax
1933 * ssize_t recvmsg(int socket, struct msghdr *message,
1936 * socket - the socket descriptor of the endpoint.
1937 * message - pointer to the msghdr structure which contains a single
1938 * user message and possibly some ancillary data.
1940 * See Section 5 for complete description of the data
1943 * flags - flags sent or received with the user message, see Section
1944 * 5 for complete description of the flags.
1946 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1948 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1949 struct msghdr *msg, size_t len, int noblock,
1950 int flags, int *addr_len)
1952 struct sctp_ulpevent *event = NULL;
1953 struct sctp_sock *sp = sctp_sk(sk);
1954 struct sk_buff *skb;
1959 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1960 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1961 "len", len, "knoblauch", noblock,
1962 "flags", flags, "addr_len", addr_len);
1966 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1971 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1975 /* Get the total length of the skb including any skb's in the
1984 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1986 event = sctp_skb2event(skb);
1991 sock_recv_ts_and_drops(msg, sk, skb);
1992 if (sctp_ulpevent_is_notification(event)) {
1993 msg->msg_flags |= MSG_NOTIFICATION;
1994 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1996 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1999 /* Check if we allow SCTP_SNDRCVINFO. */
2000 if (sp->subscribe.sctp_data_io_event)
2001 sctp_ulpevent_read_sndrcvinfo(event, msg);
2003 /* FIXME: we should be calling IP/IPv6 layers. */
2004 if (sk->sk_protinfo.af_inet.cmsg_flags)
2005 ip_cmsg_recv(msg, skb);
2010 /* If skb's length exceeds the user's buffer, update the skb and
2011 * push it back to the receive_queue so that the next call to
2012 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2014 if (skb_len > copied) {
2015 msg->msg_flags &= ~MSG_EOR;
2016 if (flags & MSG_PEEK)
2018 sctp_skb_pull(skb, copied);
2019 skb_queue_head(&sk->sk_receive_queue, skb);
2021 /* When only partial message is copied to the user, increase
2022 * rwnd by that amount. If all the data in the skb is read,
2023 * rwnd is updated when the event is freed.
2025 if (!sctp_ulpevent_is_notification(event))
2026 sctp_assoc_rwnd_increase(event->asoc, copied);
2028 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2029 (event->msg_flags & MSG_EOR))
2030 msg->msg_flags |= MSG_EOR;
2032 msg->msg_flags &= ~MSG_EOR;
2035 if (flags & MSG_PEEK) {
2036 /* Release the skb reference acquired after peeking the skb in
2037 * sctp_skb_recv_datagram().
2041 /* Free the event which includes releasing the reference to
2042 * the owner of the skb, freeing the skb and updating the
2045 sctp_ulpevent_free(event);
2048 sctp_release_sock(sk);
2052 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2054 * This option is a on/off flag. If enabled no SCTP message
2055 * fragmentation will be performed. Instead if a message being sent
2056 * exceeds the current PMTU size, the message will NOT be sent and
2057 * instead a error will be indicated to the user.
2059 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2060 char __user *optval,
2061 unsigned int optlen)
2065 if (optlen < sizeof(int))
2068 if (get_user(val, (int __user *)optval))
2071 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2076 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2077 unsigned int optlen)
2079 struct sctp_association *asoc;
2080 struct sctp_ulpevent *event;
2082 if (optlen > sizeof(struct sctp_event_subscribe))
2084 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2088 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2089 * if there is no data to be sent or retransmit, the stack will
2090 * immediately send up this notification.
2092 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2093 &sctp_sk(sk)->subscribe)) {
2094 asoc = sctp_id2assoc(sk, 0);
2096 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2097 event = sctp_ulpevent_make_sender_dry_event(asoc,
2102 sctp_ulpq_tail_event(&asoc->ulpq, event);
2109 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2111 * This socket option is applicable to the UDP-style socket only. When
2112 * set it will cause associations that are idle for more than the
2113 * specified number of seconds to automatically close. An association
2114 * being idle is defined an association that has NOT sent or received
2115 * user data. The special value of '0' indicates that no automatic
2116 * close of any associations should be performed. The option expects an
2117 * integer defining the number of seconds of idle time before an
2118 * association is closed.
2120 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2121 unsigned int optlen)
2123 struct sctp_sock *sp = sctp_sk(sk);
2125 /* Applicable to UDP-style socket only */
2126 if (sctp_style(sk, TCP))
2128 if (optlen != sizeof(int))
2130 if (copy_from_user(&sp->autoclose, optval, optlen))
2132 /* make sure it won't exceed MAX_SCHEDULE_TIMEOUT */
2133 sp->autoclose = min_t(long, sp->autoclose, MAX_SCHEDULE_TIMEOUT / HZ);
2138 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2140 * Applications can enable or disable heartbeats for any peer address of
2141 * an association, modify an address's heartbeat interval, force a
2142 * heartbeat to be sent immediately, and adjust the address's maximum
2143 * number of retransmissions sent before an address is considered
2144 * unreachable. The following structure is used to access and modify an
2145 * address's parameters:
2147 * struct sctp_paddrparams {
2148 * sctp_assoc_t spp_assoc_id;
2149 * struct sockaddr_storage spp_address;
2150 * uint32_t spp_hbinterval;
2151 * uint16_t spp_pathmaxrxt;
2152 * uint32_t spp_pathmtu;
2153 * uint32_t spp_sackdelay;
2154 * uint32_t spp_flags;
2157 * spp_assoc_id - (one-to-many style socket) This is filled in the
2158 * application, and identifies the association for
2160 * spp_address - This specifies which address is of interest.
2161 * spp_hbinterval - This contains the value of the heartbeat interval,
2162 * in milliseconds. If a value of zero
2163 * is present in this field then no changes are to
2164 * be made to this parameter.
2165 * spp_pathmaxrxt - This contains the maximum number of
2166 * retransmissions before this address shall be
2167 * considered unreachable. If a value of zero
2168 * is present in this field then no changes are to
2169 * be made to this parameter.
2170 * spp_pathmtu - When Path MTU discovery is disabled the value
2171 * specified here will be the "fixed" path mtu.
2172 * Note that if the spp_address field is empty
2173 * then all associations on this address will
2174 * have this fixed path mtu set upon them.
2176 * spp_sackdelay - When delayed sack is enabled, this value specifies
2177 * the number of milliseconds that sacks will be delayed
2178 * for. This value will apply to all addresses of an
2179 * association if the spp_address field is empty. Note
2180 * also, that if delayed sack is enabled and this
2181 * value is set to 0, no change is made to the last
2182 * recorded delayed sack timer value.
2184 * spp_flags - These flags are used to control various features
2185 * on an association. The flag field may contain
2186 * zero or more of the following options.
2188 * SPP_HB_ENABLE - Enable heartbeats on the
2189 * specified address. Note that if the address
2190 * field is empty all addresses for the association
2191 * have heartbeats enabled upon them.
2193 * SPP_HB_DISABLE - Disable heartbeats on the
2194 * speicifed address. Note that if the address
2195 * field is empty all addresses for the association
2196 * will have their heartbeats disabled. Note also
2197 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2198 * mutually exclusive, only one of these two should
2199 * be specified. Enabling both fields will have
2200 * undetermined results.
2202 * SPP_HB_DEMAND - Request a user initiated heartbeat
2203 * to be made immediately.
2205 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2206 * heartbeat delayis to be set to the value of 0
2209 * SPP_PMTUD_ENABLE - This field will enable PMTU
2210 * discovery upon the specified address. Note that
2211 * if the address feild is empty then all addresses
2212 * on the association are effected.
2214 * SPP_PMTUD_DISABLE - This field will disable PMTU
2215 * discovery upon the specified address. Note that
2216 * if the address feild is empty then all addresses
2217 * on the association are effected. Not also that
2218 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2219 * exclusive. Enabling both will have undetermined
2222 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2223 * on delayed sack. The time specified in spp_sackdelay
2224 * is used to specify the sack delay for this address. Note
2225 * that if spp_address is empty then all addresses will
2226 * enable delayed sack and take on the sack delay
2227 * value specified in spp_sackdelay.
2228 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2229 * off delayed sack. If the spp_address field is blank then
2230 * delayed sack is disabled for the entire association. Note
2231 * also that this field is mutually exclusive to
2232 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2235 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2236 struct sctp_transport *trans,
2237 struct sctp_association *asoc,
2238 struct sctp_sock *sp,
2241 int sackdelay_change)
2245 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2246 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2251 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2252 * this field is ignored. Note also that a value of zero indicates
2253 * the current setting should be left unchanged.
2255 if (params->spp_flags & SPP_HB_ENABLE) {
2257 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2258 * set. This lets us use 0 value when this flag
2261 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2262 params->spp_hbinterval = 0;
2264 if (params->spp_hbinterval ||
2265 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2268 msecs_to_jiffies(params->spp_hbinterval);
2271 msecs_to_jiffies(params->spp_hbinterval);
2273 sp->hbinterval = params->spp_hbinterval;
2280 trans->param_flags =
2281 (trans->param_flags & ~SPP_HB) | hb_change;
2284 (asoc->param_flags & ~SPP_HB) | hb_change;
2287 (sp->param_flags & ~SPP_HB) | hb_change;
2291 /* When Path MTU discovery is disabled the value specified here will
2292 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2293 * include the flag SPP_PMTUD_DISABLE for this field to have any
2296 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2298 trans->pathmtu = params->spp_pathmtu;
2299 sctp_assoc_sync_pmtu(asoc);
2301 asoc->pathmtu = params->spp_pathmtu;
2302 sctp_frag_point(asoc, params->spp_pathmtu);
2304 sp->pathmtu = params->spp_pathmtu;
2310 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2311 (params->spp_flags & SPP_PMTUD_ENABLE);
2312 trans->param_flags =
2313 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2315 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2316 sctp_assoc_sync_pmtu(asoc);
2320 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2323 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2327 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2328 * value of this field is ignored. Note also that a value of zero
2329 * indicates the current setting should be left unchanged.
2331 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2334 msecs_to_jiffies(params->spp_sackdelay);
2337 msecs_to_jiffies(params->spp_sackdelay);
2339 sp->sackdelay = params->spp_sackdelay;
2343 if (sackdelay_change) {
2345 trans->param_flags =
2346 (trans->param_flags & ~SPP_SACKDELAY) |
2350 (asoc->param_flags & ~SPP_SACKDELAY) |
2354 (sp->param_flags & ~SPP_SACKDELAY) |
2359 /* Note that a value of zero indicates the current setting should be
2362 if (params->spp_pathmaxrxt) {
2364 trans->pathmaxrxt = params->spp_pathmaxrxt;
2366 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2368 sp->pathmaxrxt = params->spp_pathmaxrxt;
2375 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2376 char __user *optval,
2377 unsigned int optlen)
2379 struct sctp_paddrparams params;
2380 struct sctp_transport *trans = NULL;
2381 struct sctp_association *asoc = NULL;
2382 struct sctp_sock *sp = sctp_sk(sk);
2384 int hb_change, pmtud_change, sackdelay_change;
2386 if (optlen != sizeof(struct sctp_paddrparams))
2389 if (copy_from_user(¶ms, optval, optlen))
2392 /* Validate flags and value parameters. */
2393 hb_change = params.spp_flags & SPP_HB;
2394 pmtud_change = params.spp_flags & SPP_PMTUD;
2395 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2397 if (hb_change == SPP_HB ||
2398 pmtud_change == SPP_PMTUD ||
2399 sackdelay_change == SPP_SACKDELAY ||
2400 params.spp_sackdelay > 500 ||
2401 (params.spp_pathmtu &&
2402 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2405 /* If an address other than INADDR_ANY is specified, and
2406 * no transport is found, then the request is invalid.
2408 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
2409 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2410 params.spp_assoc_id);
2415 /* Get association, if assoc_id != 0 and the socket is a one
2416 * to many style socket, and an association was not found, then
2417 * the id was invalid.
2419 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2420 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2423 /* Heartbeat demand can only be sent on a transport or
2424 * association, but not a socket.
2426 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2429 /* Process parameters. */
2430 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2431 hb_change, pmtud_change,
2437 /* If changes are for association, also apply parameters to each
2440 if (!trans && asoc) {
2441 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2443 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2444 hb_change, pmtud_change,
2453 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2455 * This option will effect the way delayed acks are performed. This
2456 * option allows you to get or set the delayed ack time, in
2457 * milliseconds. It also allows changing the delayed ack frequency.
2458 * Changing the frequency to 1 disables the delayed sack algorithm. If
2459 * the assoc_id is 0, then this sets or gets the endpoints default
2460 * values. If the assoc_id field is non-zero, then the set or get
2461 * effects the specified association for the one to many model (the
2462 * assoc_id field is ignored by the one to one model). Note that if
2463 * sack_delay or sack_freq are 0 when setting this option, then the
2464 * current values will remain unchanged.
2466 * struct sctp_sack_info {
2467 * sctp_assoc_t sack_assoc_id;
2468 * uint32_t sack_delay;
2469 * uint32_t sack_freq;
2472 * sack_assoc_id - This parameter, indicates which association the user
2473 * is performing an action upon. Note that if this field's value is
2474 * zero then the endpoints default value is changed (effecting future
2475 * associations only).
2477 * sack_delay - This parameter contains the number of milliseconds that
2478 * the user is requesting the delayed ACK timer be set to. Note that
2479 * this value is defined in the standard to be between 200 and 500
2482 * sack_freq - This parameter contains the number of packets that must
2483 * be received before a sack is sent without waiting for the delay
2484 * timer to expire. The default value for this is 2, setting this
2485 * value to 1 will disable the delayed sack algorithm.
2488 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2489 char __user *optval, unsigned int optlen)
2491 struct sctp_sack_info params;
2492 struct sctp_transport *trans = NULL;
2493 struct sctp_association *asoc = NULL;
2494 struct sctp_sock *sp = sctp_sk(sk);
2496 if (optlen == sizeof(struct sctp_sack_info)) {
2497 if (copy_from_user(¶ms, optval, optlen))
2500 if (params.sack_delay == 0 && params.sack_freq == 0)
2502 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2503 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2504 pr_warn("Use struct sctp_sack_info instead\n");
2505 if (copy_from_user(¶ms, optval, optlen))
2508 if (params.sack_delay == 0)
2509 params.sack_freq = 1;
2511 params.sack_freq = 0;
2515 /* Validate value parameter. */
2516 if (params.sack_delay > 500)
2519 /* Get association, if sack_assoc_id != 0 and the socket is a one
2520 * to many style socket, and an association was not found, then
2521 * the id was invalid.
2523 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2524 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2527 if (params.sack_delay) {
2530 msecs_to_jiffies(params.sack_delay);
2532 (asoc->param_flags & ~SPP_SACKDELAY) |
2533 SPP_SACKDELAY_ENABLE;
2535 sp->sackdelay = params.sack_delay;
2537 (sp->param_flags & ~SPP_SACKDELAY) |
2538 SPP_SACKDELAY_ENABLE;
2542 if (params.sack_freq == 1) {
2545 (asoc->param_flags & ~SPP_SACKDELAY) |
2546 SPP_SACKDELAY_DISABLE;
2549 (sp->param_flags & ~SPP_SACKDELAY) |
2550 SPP_SACKDELAY_DISABLE;
2552 } else if (params.sack_freq > 1) {
2554 asoc->sackfreq = params.sack_freq;
2556 (asoc->param_flags & ~SPP_SACKDELAY) |
2557 SPP_SACKDELAY_ENABLE;
2559 sp->sackfreq = params.sack_freq;
2561 (sp->param_flags & ~SPP_SACKDELAY) |
2562 SPP_SACKDELAY_ENABLE;
2566 /* If change is for association, also apply to each transport. */
2568 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2570 if (params.sack_delay) {
2572 msecs_to_jiffies(params.sack_delay);
2573 trans->param_flags =
2574 (trans->param_flags & ~SPP_SACKDELAY) |
2575 SPP_SACKDELAY_ENABLE;
2577 if (params.sack_freq == 1) {
2578 trans->param_flags =
2579 (trans->param_flags & ~SPP_SACKDELAY) |
2580 SPP_SACKDELAY_DISABLE;
2581 } else if (params.sack_freq > 1) {
2582 trans->sackfreq = params.sack_freq;
2583 trans->param_flags =
2584 (trans->param_flags & ~SPP_SACKDELAY) |
2585 SPP_SACKDELAY_ENABLE;
2593 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2595 * Applications can specify protocol parameters for the default association
2596 * initialization. The option name argument to setsockopt() and getsockopt()
2599 * Setting initialization parameters is effective only on an unconnected
2600 * socket (for UDP-style sockets only future associations are effected
2601 * by the change). With TCP-style sockets, this option is inherited by
2602 * sockets derived from a listener socket.
2604 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2606 struct sctp_initmsg sinit;
2607 struct sctp_sock *sp = sctp_sk(sk);
2609 if (optlen != sizeof(struct sctp_initmsg))
2611 if (copy_from_user(&sinit, optval, optlen))
2614 if (sinit.sinit_num_ostreams)
2615 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2616 if (sinit.sinit_max_instreams)
2617 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2618 if (sinit.sinit_max_attempts)
2619 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2620 if (sinit.sinit_max_init_timeo)
2621 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2627 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2629 * Applications that wish to use the sendto() system call may wish to
2630 * specify a default set of parameters that would normally be supplied
2631 * through the inclusion of ancillary data. This socket option allows
2632 * such an application to set the default sctp_sndrcvinfo structure.
2633 * The application that wishes to use this socket option simply passes
2634 * in to this call the sctp_sndrcvinfo structure defined in Section
2635 * 5.2.2) The input parameters accepted by this call include
2636 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2637 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2638 * to this call if the caller is using the UDP model.
2640 static int sctp_setsockopt_default_send_param(struct sock *sk,
2641 char __user *optval,
2642 unsigned int optlen)
2644 struct sctp_sndrcvinfo info;
2645 struct sctp_association *asoc;
2646 struct sctp_sock *sp = sctp_sk(sk);
2648 if (optlen != sizeof(struct sctp_sndrcvinfo))
2650 if (copy_from_user(&info, optval, optlen))
2653 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2654 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2658 asoc->default_stream = info.sinfo_stream;
2659 asoc->default_flags = info.sinfo_flags;
2660 asoc->default_ppid = info.sinfo_ppid;
2661 asoc->default_context = info.sinfo_context;
2662 asoc->default_timetolive = info.sinfo_timetolive;
2664 sp->default_stream = info.sinfo_stream;
2665 sp->default_flags = info.sinfo_flags;
2666 sp->default_ppid = info.sinfo_ppid;
2667 sp->default_context = info.sinfo_context;
2668 sp->default_timetolive = info.sinfo_timetolive;
2674 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2676 * Requests that the local SCTP stack use the enclosed peer address as
2677 * the association primary. The enclosed address must be one of the
2678 * association peer's addresses.
2680 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2681 unsigned int optlen)
2683 struct sctp_prim prim;
2684 struct sctp_transport *trans;
2686 if (optlen != sizeof(struct sctp_prim))
2689 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2692 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2696 sctp_assoc_set_primary(trans->asoc, trans);
2702 * 7.1.5 SCTP_NODELAY
2704 * Turn on/off any Nagle-like algorithm. This means that packets are
2705 * generally sent as soon as possible and no unnecessary delays are
2706 * introduced, at the cost of more packets in the network. Expects an
2707 * integer boolean flag.
2709 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2710 unsigned int optlen)
2714 if (optlen < sizeof(int))
2716 if (get_user(val, (int __user *)optval))
2719 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2725 * 7.1.1 SCTP_RTOINFO
2727 * The protocol parameters used to initialize and bound retransmission
2728 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2729 * and modify these parameters.
2730 * All parameters are time values, in milliseconds. A value of 0, when
2731 * modifying the parameters, indicates that the current value should not
2735 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2737 struct sctp_rtoinfo rtoinfo;
2738 struct sctp_association *asoc;
2740 if (optlen != sizeof (struct sctp_rtoinfo))
2743 if (copy_from_user(&rtoinfo, optval, optlen))
2746 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2748 /* Set the values to the specific association */
2749 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2753 if (rtoinfo.srto_initial != 0)
2755 msecs_to_jiffies(rtoinfo.srto_initial);
2756 if (rtoinfo.srto_max != 0)
2757 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2758 if (rtoinfo.srto_min != 0)
2759 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2761 /* If there is no association or the association-id = 0
2762 * set the values to the endpoint.
2764 struct sctp_sock *sp = sctp_sk(sk);
2766 if (rtoinfo.srto_initial != 0)
2767 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2768 if (rtoinfo.srto_max != 0)
2769 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2770 if (rtoinfo.srto_min != 0)
2771 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2779 * 7.1.2 SCTP_ASSOCINFO
2781 * This option is used to tune the maximum retransmission attempts
2782 * of the association.
2783 * Returns an error if the new association retransmission value is
2784 * greater than the sum of the retransmission value of the peer.
2785 * See [SCTP] for more information.
2788 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2791 struct sctp_assocparams assocparams;
2792 struct sctp_association *asoc;
2794 if (optlen != sizeof(struct sctp_assocparams))
2796 if (copy_from_user(&assocparams, optval, optlen))
2799 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2801 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2804 /* Set the values to the specific association */
2806 if (assocparams.sasoc_asocmaxrxt != 0) {
2809 struct sctp_transport *peer_addr;
2811 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2813 path_sum += peer_addr->pathmaxrxt;
2817 /* Only validate asocmaxrxt if we have more than
2818 * one path/transport. We do this because path
2819 * retransmissions are only counted when we have more
2823 assocparams.sasoc_asocmaxrxt > path_sum)
2826 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2829 if (assocparams.sasoc_cookie_life != 0) {
2830 asoc->cookie_life.tv_sec =
2831 assocparams.sasoc_cookie_life / 1000;
2832 asoc->cookie_life.tv_usec =
2833 (assocparams.sasoc_cookie_life % 1000)
2837 /* Set the values to the endpoint */
2838 struct sctp_sock *sp = sctp_sk(sk);
2840 if (assocparams.sasoc_asocmaxrxt != 0)
2841 sp->assocparams.sasoc_asocmaxrxt =
2842 assocparams.sasoc_asocmaxrxt;
2843 if (assocparams.sasoc_cookie_life != 0)
2844 sp->assocparams.sasoc_cookie_life =
2845 assocparams.sasoc_cookie_life;
2851 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2853 * This socket option is a boolean flag which turns on or off mapped V4
2854 * addresses. If this option is turned on and the socket is type
2855 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2856 * If this option is turned off, then no mapping will be done of V4
2857 * addresses and a user will receive both PF_INET6 and PF_INET type
2858 * addresses on the socket.
2860 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2863 struct sctp_sock *sp = sctp_sk(sk);
2865 if (optlen < sizeof(int))
2867 if (get_user(val, (int __user *)optval))
2878 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2879 * This option will get or set the maximum size to put in any outgoing
2880 * SCTP DATA chunk. If a message is larger than this size it will be
2881 * fragmented by SCTP into the specified size. Note that the underlying
2882 * SCTP implementation may fragment into smaller sized chunks when the
2883 * PMTU of the underlying association is smaller than the value set by
2884 * the user. The default value for this option is '0' which indicates
2885 * the user is NOT limiting fragmentation and only the PMTU will effect
2886 * SCTP's choice of DATA chunk size. Note also that values set larger
2887 * than the maximum size of an IP datagram will effectively let SCTP
2888 * control fragmentation (i.e. the same as setting this option to 0).
2890 * The following structure is used to access and modify this parameter:
2892 * struct sctp_assoc_value {
2893 * sctp_assoc_t assoc_id;
2894 * uint32_t assoc_value;
2897 * assoc_id: This parameter is ignored for one-to-one style sockets.
2898 * For one-to-many style sockets this parameter indicates which
2899 * association the user is performing an action upon. Note that if
2900 * this field's value is zero then the endpoints default value is
2901 * changed (effecting future associations only).
2902 * assoc_value: This parameter specifies the maximum size in bytes.
2904 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2906 struct sctp_assoc_value params;
2907 struct sctp_association *asoc;
2908 struct sctp_sock *sp = sctp_sk(sk);
2911 if (optlen == sizeof(int)) {
2912 pr_warn("Use of int in maxseg socket option deprecated\n");
2913 pr_warn("Use struct sctp_assoc_value instead\n");
2914 if (copy_from_user(&val, optval, optlen))
2916 params.assoc_id = 0;
2917 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2918 if (copy_from_user(¶ms, optval, optlen))
2920 val = params.assoc_value;
2924 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2927 asoc = sctp_id2assoc(sk, params.assoc_id);
2928 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2933 val = asoc->pathmtu;
2934 val -= sp->pf->af->net_header_len;
2935 val -= sizeof(struct sctphdr) +
2936 sizeof(struct sctp_data_chunk);
2938 asoc->user_frag = val;
2939 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
2941 sp->user_frag = val;
2949 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2951 * Requests that the peer mark the enclosed address as the association
2952 * primary. The enclosed address must be one of the association's
2953 * locally bound addresses. The following structure is used to make a
2954 * set primary request:
2956 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2957 unsigned int optlen)
2959 struct sctp_sock *sp;
2960 struct sctp_association *asoc = NULL;
2961 struct sctp_setpeerprim prim;
2962 struct sctp_chunk *chunk;
2968 if (!sctp_addip_enable)
2971 if (optlen != sizeof(struct sctp_setpeerprim))
2974 if (copy_from_user(&prim, optval, optlen))
2977 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2981 if (!asoc->peer.asconf_capable)
2984 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2987 if (!sctp_state(asoc, ESTABLISHED))
2990 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
2994 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
2995 return -EADDRNOTAVAIL;
2997 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2998 return -EADDRNOTAVAIL;
3000 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3001 chunk = sctp_make_asconf_set_prim(asoc,
3002 (union sctp_addr *)&prim.sspp_addr);
3006 err = sctp_send_asconf(asoc, chunk);
3008 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
3013 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3014 unsigned int optlen)
3016 struct sctp_setadaptation adaptation;
3018 if (optlen != sizeof(struct sctp_setadaptation))
3020 if (copy_from_user(&adaptation, optval, optlen))
3023 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3029 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3031 * The context field in the sctp_sndrcvinfo structure is normally only
3032 * used when a failed message is retrieved holding the value that was
3033 * sent down on the actual send call. This option allows the setting of
3034 * a default context on an association basis that will be received on
3035 * reading messages from the peer. This is especially helpful in the
3036 * one-2-many model for an application to keep some reference to an
3037 * internal state machine that is processing messages on the
3038 * association. Note that the setting of this value only effects
3039 * received messages from the peer and does not effect the value that is
3040 * saved with outbound messages.
3042 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3043 unsigned int optlen)
3045 struct sctp_assoc_value params;
3046 struct sctp_sock *sp;
3047 struct sctp_association *asoc;
3049 if (optlen != sizeof(struct sctp_assoc_value))
3051 if (copy_from_user(¶ms, optval, optlen))
3056 if (params.assoc_id != 0) {
3057 asoc = sctp_id2assoc(sk, params.assoc_id);
3060 asoc->default_rcv_context = params.assoc_value;
3062 sp->default_rcv_context = params.assoc_value;
3069 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3071 * This options will at a minimum specify if the implementation is doing
3072 * fragmented interleave. Fragmented interleave, for a one to many
3073 * socket, is when subsequent calls to receive a message may return
3074 * parts of messages from different associations. Some implementations
3075 * may allow you to turn this value on or off. If so, when turned off,
3076 * no fragment interleave will occur (which will cause a head of line
3077 * blocking amongst multiple associations sharing the same one to many
3078 * socket). When this option is turned on, then each receive call may
3079 * come from a different association (thus the user must receive data
3080 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3081 * association each receive belongs to.
3083 * This option takes a boolean value. A non-zero value indicates that
3084 * fragmented interleave is on. A value of zero indicates that
3085 * fragmented interleave is off.
3087 * Note that it is important that an implementation that allows this
3088 * option to be turned on, have it off by default. Otherwise an unaware
3089 * application using the one to many model may become confused and act
3092 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3093 char __user *optval,
3094 unsigned int optlen)
3098 if (optlen != sizeof(int))
3100 if (get_user(val, (int __user *)optval))
3103 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3109 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3110 * (SCTP_PARTIAL_DELIVERY_POINT)
3112 * This option will set or get the SCTP partial delivery point. This
3113 * point is the size of a message where the partial delivery API will be
3114 * invoked to help free up rwnd space for the peer. Setting this to a
3115 * lower value will cause partial deliveries to happen more often. The
3116 * calls argument is an integer that sets or gets the partial delivery
3117 * point. Note also that the call will fail if the user attempts to set
3118 * this value larger than the socket receive buffer size.
3120 * Note that any single message having a length smaller than or equal to
3121 * the SCTP partial delivery point will be delivered in one single read
3122 * call as long as the user provided buffer is large enough to hold the
3125 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3126 char __user *optval,
3127 unsigned int optlen)
3131 if (optlen != sizeof(u32))
3133 if (get_user(val, (int __user *)optval))
3136 /* Note: We double the receive buffer from what the user sets
3137 * it to be, also initial rwnd is based on rcvbuf/2.
3139 if (val > (sk->sk_rcvbuf >> 1))
3142 sctp_sk(sk)->pd_point = val;
3144 return 0; /* is this the right error code? */
3148 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3150 * This option will allow a user to change the maximum burst of packets
3151 * that can be emitted by this association. Note that the default value
3152 * is 4, and some implementations may restrict this setting so that it
3153 * can only be lowered.
3155 * NOTE: This text doesn't seem right. Do this on a socket basis with
3156 * future associations inheriting the socket value.
3158 static int sctp_setsockopt_maxburst(struct sock *sk,
3159 char __user *optval,
3160 unsigned int optlen)
3162 struct sctp_assoc_value params;
3163 struct sctp_sock *sp;
3164 struct sctp_association *asoc;
3168 if (optlen == sizeof(int)) {
3169 pr_warn("Use of int in max_burst socket option deprecated\n");
3170 pr_warn("Use struct sctp_assoc_value instead\n");
3171 if (copy_from_user(&val, optval, optlen))
3173 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3174 if (copy_from_user(¶ms, optval, optlen))
3176 val = params.assoc_value;
3177 assoc_id = params.assoc_id;
3183 if (assoc_id != 0) {
3184 asoc = sctp_id2assoc(sk, assoc_id);
3187 asoc->max_burst = val;
3189 sp->max_burst = val;
3195 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3197 * This set option adds a chunk type that the user is requesting to be
3198 * received only in an authenticated way. Changes to the list of chunks
3199 * will only effect future associations on the socket.
3201 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3202 char __user *optval,
3203 unsigned int optlen)
3205 struct sctp_authchunk val;
3207 if (!sctp_auth_enable)
3210 if (optlen != sizeof(struct sctp_authchunk))
3212 if (copy_from_user(&val, optval, optlen))
3215 switch (val.sauth_chunk) {
3217 case SCTP_CID_INIT_ACK:
3218 case SCTP_CID_SHUTDOWN_COMPLETE:
3223 /* add this chunk id to the endpoint */
3224 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3228 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3230 * This option gets or sets the list of HMAC algorithms that the local
3231 * endpoint requires the peer to use.
3233 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3234 char __user *optval,
3235 unsigned int optlen)
3237 struct sctp_hmacalgo *hmacs;
3241 if (!sctp_auth_enable)
3244 if (optlen < sizeof(struct sctp_hmacalgo))
3247 hmacs= memdup_user(optval, optlen);
3249 return PTR_ERR(hmacs);
3251 idents = hmacs->shmac_num_idents;
3252 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3253 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3258 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3265 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3267 * This option will set a shared secret key which is used to build an
3268 * association shared key.
3270 static int sctp_setsockopt_auth_key(struct sock *sk,
3271 char __user *optval,
3272 unsigned int optlen)
3274 struct sctp_authkey *authkey;
3275 struct sctp_association *asoc;
3278 if (!sctp_auth_enable)
3281 if (optlen <= sizeof(struct sctp_authkey))
3284 authkey= memdup_user(optval, optlen);
3285 if (IS_ERR(authkey))
3286 return PTR_ERR(authkey);
3288 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3293 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3294 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3299 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3306 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3308 * This option will get or set the active shared key to be used to build
3309 * the association shared key.
3311 static int sctp_setsockopt_active_key(struct sock *sk,
3312 char __user *optval,
3313 unsigned int optlen)
3315 struct sctp_authkeyid val;
3316 struct sctp_association *asoc;
3318 if (!sctp_auth_enable)
3321 if (optlen != sizeof(struct sctp_authkeyid))
3323 if (copy_from_user(&val, optval, optlen))
3326 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3327 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3330 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3331 val.scact_keynumber);
3335 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3337 * This set option will delete a shared secret key from use.
3339 static int sctp_setsockopt_del_key(struct sock *sk,
3340 char __user *optval,
3341 unsigned int optlen)
3343 struct sctp_authkeyid val;
3344 struct sctp_association *asoc;
3346 if (!sctp_auth_enable)
3349 if (optlen != sizeof(struct sctp_authkeyid))
3351 if (copy_from_user(&val, optval, optlen))
3354 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3355 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3358 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3359 val.scact_keynumber);
3364 /* API 6.2 setsockopt(), getsockopt()
3366 * Applications use setsockopt() and getsockopt() to set or retrieve
3367 * socket options. Socket options are used to change the default
3368 * behavior of sockets calls. They are described in Section 7.
3372 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3373 * int __user *optlen);
3374 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3377 * sd - the socket descript.
3378 * level - set to IPPROTO_SCTP for all SCTP options.
3379 * optname - the option name.
3380 * optval - the buffer to store the value of the option.
3381 * optlen - the size of the buffer.
3383 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3384 char __user *optval, unsigned int optlen)
3388 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3391 /* I can hardly begin to describe how wrong this is. This is
3392 * so broken as to be worse than useless. The API draft
3393 * REALLY is NOT helpful here... I am not convinced that the
3394 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3395 * are at all well-founded.
3397 if (level != SOL_SCTP) {
3398 struct sctp_af *af = sctp_sk(sk)->pf->af;
3399 retval = af->setsockopt(sk, level, optname, optval, optlen);
3406 case SCTP_SOCKOPT_BINDX_ADD:
3407 /* 'optlen' is the size of the addresses buffer. */
3408 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3409 optlen, SCTP_BINDX_ADD_ADDR);
3412 case SCTP_SOCKOPT_BINDX_REM:
3413 /* 'optlen' is the size of the addresses buffer. */
3414 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3415 optlen, SCTP_BINDX_REM_ADDR);
3418 case SCTP_SOCKOPT_CONNECTX_OLD:
3419 /* 'optlen' is the size of the addresses buffer. */
3420 retval = sctp_setsockopt_connectx_old(sk,
3421 (struct sockaddr __user *)optval,
3425 case SCTP_SOCKOPT_CONNECTX:
3426 /* 'optlen' is the size of the addresses buffer. */
3427 retval = sctp_setsockopt_connectx(sk,
3428 (struct sockaddr __user *)optval,
3432 case SCTP_DISABLE_FRAGMENTS:
3433 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3437 retval = sctp_setsockopt_events(sk, optval, optlen);
3440 case SCTP_AUTOCLOSE:
3441 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3444 case SCTP_PEER_ADDR_PARAMS:
3445 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3448 case SCTP_DELAYED_SACK:
3449 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3451 case SCTP_PARTIAL_DELIVERY_POINT:
3452 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3456 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3458 case SCTP_DEFAULT_SEND_PARAM:
3459 retval = sctp_setsockopt_default_send_param(sk, optval,
3462 case SCTP_PRIMARY_ADDR:
3463 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3465 case SCTP_SET_PEER_PRIMARY_ADDR:
3466 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3469 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3472 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3474 case SCTP_ASSOCINFO:
3475 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3477 case SCTP_I_WANT_MAPPED_V4_ADDR:
3478 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3481 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3483 case SCTP_ADAPTATION_LAYER:
3484 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3487 retval = sctp_setsockopt_context(sk, optval, optlen);
3489 case SCTP_FRAGMENT_INTERLEAVE:
3490 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3492 case SCTP_MAX_BURST:
3493 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3495 case SCTP_AUTH_CHUNK:
3496 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3498 case SCTP_HMAC_IDENT:
3499 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3502 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3504 case SCTP_AUTH_ACTIVE_KEY:
3505 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3507 case SCTP_AUTH_DELETE_KEY:
3508 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3511 retval = -ENOPROTOOPT;
3515 sctp_release_sock(sk);
3521 /* API 3.1.6 connect() - UDP Style Syntax
3523 * An application may use the connect() call in the UDP model to initiate an
3524 * association without sending data.
3528 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3530 * sd: the socket descriptor to have a new association added to.
3532 * nam: the address structure (either struct sockaddr_in or struct
3533 * sockaddr_in6 defined in RFC2553 [7]).
3535 * len: the size of the address.
3537 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3545 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3546 __func__, sk, addr, addr_len);
3548 /* Validate addr_len before calling common connect/connectx routine. */
3549 af = sctp_get_af_specific(addr->sa_family);
3550 if (!af || addr_len < af->sockaddr_len) {
3553 /* Pass correct addr len to common routine (so it knows there
3554 * is only one address being passed.
3556 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3559 sctp_release_sock(sk);
3563 /* FIXME: Write comments. */
3564 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3566 return -EOPNOTSUPP; /* STUB */
3569 /* 4.1.4 accept() - TCP Style Syntax
3571 * Applications use accept() call to remove an established SCTP
3572 * association from the accept queue of the endpoint. A new socket
3573 * descriptor will be returned from accept() to represent the newly
3574 * formed association.
3576 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3578 struct sctp_sock *sp;
3579 struct sctp_endpoint *ep;
3580 struct sock *newsk = NULL;
3581 struct sctp_association *asoc;
3590 if (!sctp_style(sk, TCP)) {
3591 error = -EOPNOTSUPP;
3595 if (!sctp_sstate(sk, LISTENING)) {
3600 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3602 error = sctp_wait_for_accept(sk, timeo);
3606 /* We treat the list of associations on the endpoint as the accept
3607 * queue and pick the first association on the list.
3609 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3611 newsk = sp->pf->create_accept_sk(sk, asoc);
3617 /* Populate the fields of the newsk from the oldsk and migrate the
3618 * asoc to the newsk.
3620 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3623 sctp_release_sock(sk);
3628 /* The SCTP ioctl handler. */
3629 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3636 * SEQPACKET-style sockets in LISTENING state are valid, for
3637 * SCTP, so only discard TCP-style sockets in LISTENING state.
3639 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3644 struct sk_buff *skb;
3645 unsigned int amount = 0;
3647 skb = skb_peek(&sk->sk_receive_queue);
3650 * We will only return the amount of this packet since
3651 * that is all that will be read.
3655 rc = put_user(amount, (int __user *)arg);
3663 sctp_release_sock(sk);
3667 /* This is the function which gets called during socket creation to
3668 * initialized the SCTP-specific portion of the sock.
3669 * The sock structure should already be zero-filled memory.
3671 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3673 struct sctp_endpoint *ep;
3674 struct sctp_sock *sp;
3676 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3680 /* Initialize the SCTP per socket area. */
3681 switch (sk->sk_type) {
3682 case SOCK_SEQPACKET:
3683 sp->type = SCTP_SOCKET_UDP;
3686 sp->type = SCTP_SOCKET_TCP;
3689 return -ESOCKTNOSUPPORT;
3692 /* Initialize default send parameters. These parameters can be
3693 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3695 sp->default_stream = 0;
3696 sp->default_ppid = 0;
3697 sp->default_flags = 0;
3698 sp->default_context = 0;
3699 sp->default_timetolive = 0;
3701 sp->default_rcv_context = 0;
3702 sp->max_burst = sctp_max_burst;
3704 /* Initialize default setup parameters. These parameters
3705 * can be modified with the SCTP_INITMSG socket option or
3706 * overridden by the SCTP_INIT CMSG.
3708 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3709 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3710 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3711 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3713 /* Initialize default RTO related parameters. These parameters can
3714 * be modified for with the SCTP_RTOINFO socket option.
3716 sp->rtoinfo.srto_initial = sctp_rto_initial;
3717 sp->rtoinfo.srto_max = sctp_rto_max;
3718 sp->rtoinfo.srto_min = sctp_rto_min;
3720 /* Initialize default association related parameters. These parameters
3721 * can be modified with the SCTP_ASSOCINFO socket option.
3723 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3724 sp->assocparams.sasoc_number_peer_destinations = 0;
3725 sp->assocparams.sasoc_peer_rwnd = 0;
3726 sp->assocparams.sasoc_local_rwnd = 0;
3727 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3729 /* Initialize default event subscriptions. By default, all the
3732 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3734 /* Default Peer Address Parameters. These defaults can
3735 * be modified via SCTP_PEER_ADDR_PARAMS
3737 sp->hbinterval = sctp_hb_interval;
3738 sp->pathmaxrxt = sctp_max_retrans_path;
3739 sp->pathmtu = 0; // allow default discovery
3740 sp->sackdelay = sctp_sack_timeout;
3742 sp->param_flags = SPP_HB_ENABLE |
3744 SPP_SACKDELAY_ENABLE;
3746 /* If enabled no SCTP message fragmentation will be performed.
3747 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3749 sp->disable_fragments = 0;
3751 /* Enable Nagle algorithm by default. */
3754 /* Enable by default. */
3757 /* Auto-close idle associations after the configured
3758 * number of seconds. A value of 0 disables this
3759 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3760 * for UDP-style sockets only.
3764 /* User specified fragmentation limit. */
3767 sp->adaptation_ind = 0;
3769 sp->pf = sctp_get_pf_specific(sk->sk_family);
3771 /* Control variables for partial data delivery. */
3772 atomic_set(&sp->pd_mode, 0);
3773 skb_queue_head_init(&sp->pd_lobby);
3774 sp->frag_interleave = 0;
3776 /* Create a per socket endpoint structure. Even if we
3777 * change the data structure relationships, this may still
3778 * be useful for storing pre-connect address information.
3780 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3787 SCTP_DBG_OBJCNT_INC(sock);
3790 percpu_counter_inc(&sctp_sockets_allocated);
3791 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3797 /* Cleanup any SCTP per socket resources. */
3798 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3800 struct sctp_endpoint *ep;
3802 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3804 /* Release our hold on the endpoint. */
3805 ep = sctp_sk(sk)->ep;
3806 sctp_endpoint_free(ep);
3808 percpu_counter_dec(&sctp_sockets_allocated);
3809 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
3813 /* API 4.1.7 shutdown() - TCP Style Syntax
3814 * int shutdown(int socket, int how);
3816 * sd - the socket descriptor of the association to be closed.
3817 * how - Specifies the type of shutdown. The values are
3820 * Disables further receive operations. No SCTP
3821 * protocol action is taken.
3823 * Disables further send operations, and initiates
3824 * the SCTP shutdown sequence.
3826 * Disables further send and receive operations
3827 * and initiates the SCTP shutdown sequence.
3829 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3831 struct sctp_endpoint *ep;
3832 struct sctp_association *asoc;
3834 if (!sctp_style(sk, TCP))
3837 if (how & SEND_SHUTDOWN) {
3838 ep = sctp_sk(sk)->ep;
3839 if (!list_empty(&ep->asocs)) {
3840 asoc = list_entry(ep->asocs.next,
3841 struct sctp_association, asocs);
3842 sctp_primitive_SHUTDOWN(asoc, NULL);
3847 /* 7.2.1 Association Status (SCTP_STATUS)
3849 * Applications can retrieve current status information about an
3850 * association, including association state, peer receiver window size,
3851 * number of unacked data chunks, and number of data chunks pending
3852 * receipt. This information is read-only.
3854 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3855 char __user *optval,
3858 struct sctp_status status;
3859 struct sctp_association *asoc = NULL;
3860 struct sctp_transport *transport;
3861 sctp_assoc_t associd;
3864 if (len < sizeof(status)) {
3869 len = sizeof(status);
3870 if (copy_from_user(&status, optval, len)) {
3875 associd = status.sstat_assoc_id;
3876 asoc = sctp_id2assoc(sk, associd);
3882 transport = asoc->peer.primary_path;
3884 status.sstat_assoc_id = sctp_assoc2id(asoc);
3885 status.sstat_state = asoc->state;
3886 status.sstat_rwnd = asoc->peer.rwnd;
3887 status.sstat_unackdata = asoc->unack_data;
3889 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3890 status.sstat_instrms = asoc->c.sinit_max_instreams;
3891 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3892 status.sstat_fragmentation_point = asoc->frag_point;
3893 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3894 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3895 transport->af_specific->sockaddr_len);
3896 /* Map ipv4 address into v4-mapped-on-v6 address. */
3897 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3898 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3899 status.sstat_primary.spinfo_state = transport->state;
3900 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3901 status.sstat_primary.spinfo_srtt = transport->srtt;
3902 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3903 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3905 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3906 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3908 if (put_user(len, optlen)) {
3913 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3914 len, status.sstat_state, status.sstat_rwnd,
3915 status.sstat_assoc_id);
3917 if (copy_to_user(optval, &status, len)) {
3927 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3929 * Applications can retrieve information about a specific peer address
3930 * of an association, including its reachability state, congestion
3931 * window, and retransmission timer values. This information is
3934 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3935 char __user *optval,
3938 struct sctp_paddrinfo pinfo;
3939 struct sctp_transport *transport;
3942 if (len < sizeof(pinfo)) {
3947 len = sizeof(pinfo);
3948 if (copy_from_user(&pinfo, optval, len)) {
3953 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3954 pinfo.spinfo_assoc_id);
3958 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3959 pinfo.spinfo_state = transport->state;
3960 pinfo.spinfo_cwnd = transport->cwnd;
3961 pinfo.spinfo_srtt = transport->srtt;
3962 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3963 pinfo.spinfo_mtu = transport->pathmtu;
3965 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3966 pinfo.spinfo_state = SCTP_ACTIVE;
3968 if (put_user(len, optlen)) {
3973 if (copy_to_user(optval, &pinfo, len)) {
3982 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3984 * This option is a on/off flag. If enabled no SCTP message
3985 * fragmentation will be performed. Instead if a message being sent
3986 * exceeds the current PMTU size, the message will NOT be sent and
3987 * instead a error will be indicated to the user.
3989 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3990 char __user *optval, int __user *optlen)
3994 if (len < sizeof(int))
3998 val = (sctp_sk(sk)->disable_fragments == 1);
3999 if (put_user(len, optlen))
4001 if (copy_to_user(optval, &val, len))
4006 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4008 * This socket option is used to specify various notifications and
4009 * ancillary data the user wishes to receive.
4011 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4014 if (len < sizeof(struct sctp_event_subscribe))
4016 len = sizeof(struct sctp_event_subscribe);
4017 if (put_user(len, optlen))
4019 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4024 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4026 * This socket option is applicable to the UDP-style socket only. When
4027 * set it will cause associations that are idle for more than the
4028 * specified number of seconds to automatically close. An association
4029 * being idle is defined an association that has NOT sent or received
4030 * user data. The special value of '0' indicates that no automatic
4031 * close of any associations should be performed. The option expects an
4032 * integer defining the number of seconds of idle time before an
4033 * association is closed.
4035 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4037 /* Applicable to UDP-style socket only */
4038 if (sctp_style(sk, TCP))
4040 if (len < sizeof(int))
4043 if (put_user(len, optlen))
4045 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4050 /* Helper routine to branch off an association to a new socket. */
4051 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
4052 struct socket **sockp)
4054 struct sock *sk = asoc->base.sk;
4055 struct socket *sock;
4059 /* An association cannot be branched off from an already peeled-off
4060 * socket, nor is this supported for tcp style sockets.
4062 if (!sctp_style(sk, UDP))
4065 /* Create a new socket. */
4066 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4070 sctp_copy_sock(sock->sk, sk, asoc);
4072 /* Make peeled-off sockets more like 1-1 accepted sockets.
4073 * Set the daddr and initialize id to something more random
4075 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4076 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4078 /* Populate the fields of the newsk from the oldsk and migrate the
4079 * asoc to the newsk.
4081 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4088 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4090 sctp_peeloff_arg_t peeloff;
4091 struct socket *newsock;
4093 struct sctp_association *asoc;
4095 if (len < sizeof(sctp_peeloff_arg_t))
4097 len = sizeof(sctp_peeloff_arg_t);
4098 if (copy_from_user(&peeloff, optval, len))
4101 asoc = sctp_id2assoc(sk, peeloff.associd);
4107 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
4109 retval = sctp_do_peeloff(asoc, &newsock);
4113 /* Map the socket to an unused fd that can be returned to the user. */
4114 retval = sock_map_fd(newsock, 0);
4116 sock_release(newsock);
4120 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
4121 __func__, sk, asoc, newsock->sk, retval);
4123 /* Return the fd mapped to the new socket. */
4124 peeloff.sd = retval;
4125 if (put_user(len, optlen))
4127 if (copy_to_user(optval, &peeloff, len))
4134 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4136 * Applications can enable or disable heartbeats for any peer address of
4137 * an association, modify an address's heartbeat interval, force a
4138 * heartbeat to be sent immediately, and adjust the address's maximum
4139 * number of retransmissions sent before an address is considered
4140 * unreachable. The following structure is used to access and modify an
4141 * address's parameters:
4143 * struct sctp_paddrparams {
4144 * sctp_assoc_t spp_assoc_id;
4145 * struct sockaddr_storage spp_address;
4146 * uint32_t spp_hbinterval;
4147 * uint16_t spp_pathmaxrxt;
4148 * uint32_t spp_pathmtu;
4149 * uint32_t spp_sackdelay;
4150 * uint32_t spp_flags;
4153 * spp_assoc_id - (one-to-many style socket) This is filled in the
4154 * application, and identifies the association for
4156 * spp_address - This specifies which address is of interest.
4157 * spp_hbinterval - This contains the value of the heartbeat interval,
4158 * in milliseconds. If a value of zero
4159 * is present in this field then no changes are to
4160 * be made to this parameter.
4161 * spp_pathmaxrxt - This contains the maximum number of
4162 * retransmissions before this address shall be
4163 * considered unreachable. If a value of zero
4164 * is present in this field then no changes are to
4165 * be made to this parameter.
4166 * spp_pathmtu - When Path MTU discovery is disabled the value
4167 * specified here will be the "fixed" path mtu.
4168 * Note that if the spp_address field is empty
4169 * then all associations on this address will
4170 * have this fixed path mtu set upon them.
4172 * spp_sackdelay - When delayed sack is enabled, this value specifies
4173 * the number of milliseconds that sacks will be delayed
4174 * for. This value will apply to all addresses of an
4175 * association if the spp_address field is empty. Note
4176 * also, that if delayed sack is enabled and this
4177 * value is set to 0, no change is made to the last
4178 * recorded delayed sack timer value.
4180 * spp_flags - These flags are used to control various features
4181 * on an association. The flag field may contain
4182 * zero or more of the following options.
4184 * SPP_HB_ENABLE - Enable heartbeats on the
4185 * specified address. Note that if the address
4186 * field is empty all addresses for the association
4187 * have heartbeats enabled upon them.
4189 * SPP_HB_DISABLE - Disable heartbeats on the
4190 * speicifed address. Note that if the address
4191 * field is empty all addresses for the association
4192 * will have their heartbeats disabled. Note also
4193 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4194 * mutually exclusive, only one of these two should
4195 * be specified. Enabling both fields will have
4196 * undetermined results.
4198 * SPP_HB_DEMAND - Request a user initiated heartbeat
4199 * to be made immediately.
4201 * SPP_PMTUD_ENABLE - This field will enable PMTU
4202 * discovery upon the specified address. Note that
4203 * if the address feild is empty then all addresses
4204 * on the association are effected.
4206 * SPP_PMTUD_DISABLE - This field will disable PMTU
4207 * discovery upon the specified address. Note that
4208 * if the address feild is empty then all addresses
4209 * on the association are effected. Not also that
4210 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4211 * exclusive. Enabling both will have undetermined
4214 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4215 * on delayed sack. The time specified in spp_sackdelay
4216 * is used to specify the sack delay for this address. Note
4217 * that if spp_address is empty then all addresses will
4218 * enable delayed sack and take on the sack delay
4219 * value specified in spp_sackdelay.
4220 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4221 * off delayed sack. If the spp_address field is blank then
4222 * delayed sack is disabled for the entire association. Note
4223 * also that this field is mutually exclusive to
4224 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4227 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4228 char __user *optval, int __user *optlen)
4230 struct sctp_paddrparams params;
4231 struct sctp_transport *trans = NULL;
4232 struct sctp_association *asoc = NULL;
4233 struct sctp_sock *sp = sctp_sk(sk);
4235 if (len < sizeof(struct sctp_paddrparams))
4237 len = sizeof(struct sctp_paddrparams);
4238 if (copy_from_user(¶ms, optval, len))
4241 /* If an address other than INADDR_ANY is specified, and
4242 * no transport is found, then the request is invalid.
4244 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
4245 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4246 params.spp_assoc_id);
4248 SCTP_DEBUG_PRINTK("Failed no transport\n");
4253 /* Get association, if assoc_id != 0 and the socket is a one
4254 * to many style socket, and an association was not found, then
4255 * the id was invalid.
4257 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4258 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4259 SCTP_DEBUG_PRINTK("Failed no association\n");
4264 /* Fetch transport values. */
4265 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4266 params.spp_pathmtu = trans->pathmtu;
4267 params.spp_pathmaxrxt = trans->pathmaxrxt;
4268 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4270 /*draft-11 doesn't say what to return in spp_flags*/
4271 params.spp_flags = trans->param_flags;
4273 /* Fetch association values. */
4274 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4275 params.spp_pathmtu = asoc->pathmtu;
4276 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4277 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4279 /*draft-11 doesn't say what to return in spp_flags*/
4280 params.spp_flags = asoc->param_flags;
4282 /* Fetch socket values. */
4283 params.spp_hbinterval = sp->hbinterval;
4284 params.spp_pathmtu = sp->pathmtu;
4285 params.spp_sackdelay = sp->sackdelay;
4286 params.spp_pathmaxrxt = sp->pathmaxrxt;
4288 /*draft-11 doesn't say what to return in spp_flags*/
4289 params.spp_flags = sp->param_flags;
4292 if (copy_to_user(optval, ¶ms, len))
4295 if (put_user(len, optlen))
4302 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4304 * This option will effect the way delayed acks are performed. This
4305 * option allows you to get or set the delayed ack time, in
4306 * milliseconds. It also allows changing the delayed ack frequency.
4307 * Changing the frequency to 1 disables the delayed sack algorithm. If
4308 * the assoc_id is 0, then this sets or gets the endpoints default
4309 * values. If the assoc_id field is non-zero, then the set or get
4310 * effects the specified association for the one to many model (the
4311 * assoc_id field is ignored by the one to one model). Note that if
4312 * sack_delay or sack_freq are 0 when setting this option, then the
4313 * current values will remain unchanged.
4315 * struct sctp_sack_info {
4316 * sctp_assoc_t sack_assoc_id;
4317 * uint32_t sack_delay;
4318 * uint32_t sack_freq;
4321 * sack_assoc_id - This parameter, indicates which association the user
4322 * is performing an action upon. Note that if this field's value is
4323 * zero then the endpoints default value is changed (effecting future
4324 * associations only).
4326 * sack_delay - This parameter contains the number of milliseconds that
4327 * the user is requesting the delayed ACK timer be set to. Note that
4328 * this value is defined in the standard to be between 200 and 500
4331 * sack_freq - This parameter contains the number of packets that must
4332 * be received before a sack is sent without waiting for the delay
4333 * timer to expire. The default value for this is 2, setting this
4334 * value to 1 will disable the delayed sack algorithm.
4336 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4337 char __user *optval,
4340 struct sctp_sack_info params;
4341 struct sctp_association *asoc = NULL;
4342 struct sctp_sock *sp = sctp_sk(sk);
4344 if (len >= sizeof(struct sctp_sack_info)) {
4345 len = sizeof(struct sctp_sack_info);
4347 if (copy_from_user(¶ms, optval, len))
4349 } else if (len == sizeof(struct sctp_assoc_value)) {
4350 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4351 pr_warn("Use struct sctp_sack_info instead\n");
4352 if (copy_from_user(¶ms, optval, len))
4357 /* Get association, if sack_assoc_id != 0 and the socket is a one
4358 * to many style socket, and an association was not found, then
4359 * the id was invalid.
4361 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4362 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4366 /* Fetch association values. */
4367 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4368 params.sack_delay = jiffies_to_msecs(
4370 params.sack_freq = asoc->sackfreq;
4373 params.sack_delay = 0;
4374 params.sack_freq = 1;
4377 /* Fetch socket values. */
4378 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4379 params.sack_delay = sp->sackdelay;
4380 params.sack_freq = sp->sackfreq;
4382 params.sack_delay = 0;
4383 params.sack_freq = 1;
4387 if (copy_to_user(optval, ¶ms, len))
4390 if (put_user(len, optlen))
4396 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4398 * Applications can specify protocol parameters for the default association
4399 * initialization. The option name argument to setsockopt() and getsockopt()
4402 * Setting initialization parameters is effective only on an unconnected
4403 * socket (for UDP-style sockets only future associations are effected
4404 * by the change). With TCP-style sockets, this option is inherited by
4405 * sockets derived from a listener socket.
4407 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4409 if (len < sizeof(struct sctp_initmsg))
4411 len = sizeof(struct sctp_initmsg);
4412 if (put_user(len, optlen))
4414 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4420 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4421 char __user *optval, int __user *optlen)
4423 struct sctp_association *asoc;
4425 struct sctp_getaddrs getaddrs;
4426 struct sctp_transport *from;
4428 union sctp_addr temp;
4429 struct sctp_sock *sp = sctp_sk(sk);
4434 if (len < sizeof(struct sctp_getaddrs))
4437 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4440 /* For UDP-style sockets, id specifies the association to query. */
4441 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4445 to = optval + offsetof(struct sctp_getaddrs,addrs);
4446 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4448 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4450 memcpy(&temp, &from->ipaddr, sizeof(temp));
4451 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4452 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4453 if (space_left < addrlen)
4455 if (copy_to_user(to, &temp, addrlen))
4459 space_left -= addrlen;
4462 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4464 bytes_copied = ((char __user *)to) - optval;
4465 if (put_user(bytes_copied, optlen))
4471 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4472 size_t space_left, int *bytes_copied)
4474 struct sctp_sockaddr_entry *addr;
4475 union sctp_addr temp;
4480 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4484 if ((PF_INET == sk->sk_family) &&
4485 (AF_INET6 == addr->a.sa.sa_family))
4487 if ((PF_INET6 == sk->sk_family) &&
4488 inet_v6_ipv6only(sk) &&
4489 (AF_INET == addr->a.sa.sa_family))
4491 memcpy(&temp, &addr->a, sizeof(temp));
4492 if (!temp.v4.sin_port)
4493 temp.v4.sin_port = htons(port);
4495 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4497 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4498 if (space_left < addrlen) {
4502 memcpy(to, &temp, addrlen);
4506 space_left -= addrlen;
4507 *bytes_copied += addrlen;
4515 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4516 char __user *optval, int __user *optlen)
4518 struct sctp_bind_addr *bp;
4519 struct sctp_association *asoc;
4521 struct sctp_getaddrs getaddrs;
4522 struct sctp_sockaddr_entry *addr;
4524 union sctp_addr temp;
4525 struct sctp_sock *sp = sctp_sk(sk);
4529 int bytes_copied = 0;
4533 if (len < sizeof(struct sctp_getaddrs))
4536 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4540 * For UDP-style sockets, id specifies the association to query.
4541 * If the id field is set to the value '0' then the locally bound
4542 * addresses are returned without regard to any particular
4545 if (0 == getaddrs.assoc_id) {
4546 bp = &sctp_sk(sk)->ep->base.bind_addr;
4548 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4551 bp = &asoc->base.bind_addr;
4554 to = optval + offsetof(struct sctp_getaddrs,addrs);
4555 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4557 addrs = kmalloc(space_left, GFP_KERNEL);
4561 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4562 * addresses from the global local address list.
4564 if (sctp_list_single_entry(&bp->address_list)) {
4565 addr = list_entry(bp->address_list.next,
4566 struct sctp_sockaddr_entry, list);
4567 if (sctp_is_any(sk, &addr->a)) {
4568 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4569 space_left, &bytes_copied);
4579 /* Protection on the bound address list is not needed since
4580 * in the socket option context we hold a socket lock and
4581 * thus the bound address list can't change.
4583 list_for_each_entry(addr, &bp->address_list, list) {
4584 memcpy(&temp, &addr->a, sizeof(temp));
4585 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4586 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4587 if (space_left < addrlen) {
4588 err = -ENOMEM; /*fixme: right error?*/
4591 memcpy(buf, &temp, addrlen);
4593 bytes_copied += addrlen;
4595 space_left -= addrlen;
4599 if (copy_to_user(to, addrs, bytes_copied)) {
4603 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4607 if (put_user(bytes_copied, optlen))
4614 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4616 * Requests that the local SCTP stack use the enclosed peer address as
4617 * the association primary. The enclosed address must be one of the
4618 * association peer's addresses.
4620 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4621 char __user *optval, int __user *optlen)
4623 struct sctp_prim prim;
4624 struct sctp_association *asoc;
4625 struct sctp_sock *sp = sctp_sk(sk);
4627 if (len < sizeof(struct sctp_prim))
4630 len = sizeof(struct sctp_prim);
4632 if (copy_from_user(&prim, optval, len))
4635 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4639 if (!asoc->peer.primary_path)
4642 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4643 asoc->peer.primary_path->af_specific->sockaddr_len);
4645 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4646 (union sctp_addr *)&prim.ssp_addr);
4648 if (put_user(len, optlen))
4650 if (copy_to_user(optval, &prim, len))
4657 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4659 * Requests that the local endpoint set the specified Adaptation Layer
4660 * Indication parameter for all future INIT and INIT-ACK exchanges.
4662 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4663 char __user *optval, int __user *optlen)
4665 struct sctp_setadaptation adaptation;
4667 if (len < sizeof(struct sctp_setadaptation))
4670 len = sizeof(struct sctp_setadaptation);
4672 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4674 if (put_user(len, optlen))
4676 if (copy_to_user(optval, &adaptation, len))
4684 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4686 * Applications that wish to use the sendto() system call may wish to
4687 * specify a default set of parameters that would normally be supplied
4688 * through the inclusion of ancillary data. This socket option allows
4689 * such an application to set the default sctp_sndrcvinfo structure.
4692 * The application that wishes to use this socket option simply passes
4693 * in to this call the sctp_sndrcvinfo structure defined in Section
4694 * 5.2.2) The input parameters accepted by this call include
4695 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4696 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4697 * to this call if the caller is using the UDP model.
4699 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4701 static int sctp_getsockopt_default_send_param(struct sock *sk,
4702 int len, char __user *optval,
4705 struct sctp_sndrcvinfo info;
4706 struct sctp_association *asoc;
4707 struct sctp_sock *sp = sctp_sk(sk);
4709 if (len < sizeof(struct sctp_sndrcvinfo))
4712 len = sizeof(struct sctp_sndrcvinfo);
4714 if (copy_from_user(&info, optval, len))
4717 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4718 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4722 info.sinfo_stream = asoc->default_stream;
4723 info.sinfo_flags = asoc->default_flags;
4724 info.sinfo_ppid = asoc->default_ppid;
4725 info.sinfo_context = asoc->default_context;
4726 info.sinfo_timetolive = asoc->default_timetolive;
4728 info.sinfo_stream = sp->default_stream;
4729 info.sinfo_flags = sp->default_flags;
4730 info.sinfo_ppid = sp->default_ppid;
4731 info.sinfo_context = sp->default_context;
4732 info.sinfo_timetolive = sp->default_timetolive;
4735 if (put_user(len, optlen))
4737 if (copy_to_user(optval, &info, len))
4745 * 7.1.5 SCTP_NODELAY
4747 * Turn on/off any Nagle-like algorithm. This means that packets are
4748 * generally sent as soon as possible and no unnecessary delays are
4749 * introduced, at the cost of more packets in the network. Expects an
4750 * integer boolean flag.
4753 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4754 char __user *optval, int __user *optlen)
4758 if (len < sizeof(int))
4762 val = (sctp_sk(sk)->nodelay == 1);
4763 if (put_user(len, optlen))
4765 if (copy_to_user(optval, &val, len))
4772 * 7.1.1 SCTP_RTOINFO
4774 * The protocol parameters used to initialize and bound retransmission
4775 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4776 * and modify these parameters.
4777 * All parameters are time values, in milliseconds. A value of 0, when
4778 * modifying the parameters, indicates that the current value should not
4782 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4783 char __user *optval,
4784 int __user *optlen) {
4785 struct sctp_rtoinfo rtoinfo;
4786 struct sctp_association *asoc;
4788 if (len < sizeof (struct sctp_rtoinfo))
4791 len = sizeof(struct sctp_rtoinfo);
4793 if (copy_from_user(&rtoinfo, optval, len))
4796 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4798 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4801 /* Values corresponding to the specific association. */
4803 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4804 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4805 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4807 /* Values corresponding to the endpoint. */
4808 struct sctp_sock *sp = sctp_sk(sk);
4810 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4811 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4812 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4815 if (put_user(len, optlen))
4818 if (copy_to_user(optval, &rtoinfo, len))
4826 * 7.1.2 SCTP_ASSOCINFO
4828 * This option is used to tune the maximum retransmission attempts
4829 * of the association.
4830 * Returns an error if the new association retransmission value is
4831 * greater than the sum of the retransmission value of the peer.
4832 * See [SCTP] for more information.
4835 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4836 char __user *optval,
4840 struct sctp_assocparams assocparams;
4841 struct sctp_association *asoc;
4842 struct list_head *pos;
4845 if (len < sizeof (struct sctp_assocparams))
4848 len = sizeof(struct sctp_assocparams);
4850 if (copy_from_user(&assocparams, optval, len))
4853 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4855 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4858 /* Values correspoinding to the specific association */
4860 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4861 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4862 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4863 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4865 (asoc->cookie_life.tv_usec
4868 list_for_each(pos, &asoc->peer.transport_addr_list) {
4872 assocparams.sasoc_number_peer_destinations = cnt;
4874 /* Values corresponding to the endpoint */
4875 struct sctp_sock *sp = sctp_sk(sk);
4877 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4878 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4879 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4880 assocparams.sasoc_cookie_life =
4881 sp->assocparams.sasoc_cookie_life;
4882 assocparams.sasoc_number_peer_destinations =
4884 sasoc_number_peer_destinations;
4887 if (put_user(len, optlen))
4890 if (copy_to_user(optval, &assocparams, len))
4897 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4899 * This socket option is a boolean flag which turns on or off mapped V4
4900 * addresses. If this option is turned on and the socket is type
4901 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4902 * If this option is turned off, then no mapping will be done of V4
4903 * addresses and a user will receive both PF_INET6 and PF_INET type
4904 * addresses on the socket.
4906 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4907 char __user *optval, int __user *optlen)
4910 struct sctp_sock *sp = sctp_sk(sk);
4912 if (len < sizeof(int))
4917 if (put_user(len, optlen))
4919 if (copy_to_user(optval, &val, len))
4926 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
4927 * (chapter and verse is quoted at sctp_setsockopt_context())
4929 static int sctp_getsockopt_context(struct sock *sk, int len,
4930 char __user *optval, int __user *optlen)
4932 struct sctp_assoc_value params;
4933 struct sctp_sock *sp;
4934 struct sctp_association *asoc;
4936 if (len < sizeof(struct sctp_assoc_value))
4939 len = sizeof(struct sctp_assoc_value);
4941 if (copy_from_user(¶ms, optval, len))
4946 if (params.assoc_id != 0) {
4947 asoc = sctp_id2assoc(sk, params.assoc_id);
4950 params.assoc_value = asoc->default_rcv_context;
4952 params.assoc_value = sp->default_rcv_context;
4955 if (put_user(len, optlen))
4957 if (copy_to_user(optval, ¶ms, len))
4964 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
4965 * This option will get or set the maximum size to put in any outgoing
4966 * SCTP DATA chunk. If a message is larger than this size it will be
4967 * fragmented by SCTP into the specified size. Note that the underlying
4968 * SCTP implementation may fragment into smaller sized chunks when the
4969 * PMTU of the underlying association is smaller than the value set by
4970 * the user. The default value for this option is '0' which indicates
4971 * the user is NOT limiting fragmentation and only the PMTU will effect
4972 * SCTP's choice of DATA chunk size. Note also that values set larger
4973 * than the maximum size of an IP datagram will effectively let SCTP
4974 * control fragmentation (i.e. the same as setting this option to 0).
4976 * The following structure is used to access and modify this parameter:
4978 * struct sctp_assoc_value {
4979 * sctp_assoc_t assoc_id;
4980 * uint32_t assoc_value;
4983 * assoc_id: This parameter is ignored for one-to-one style sockets.
4984 * For one-to-many style sockets this parameter indicates which
4985 * association the user is performing an action upon. Note that if
4986 * this field's value is zero then the endpoints default value is
4987 * changed (effecting future associations only).
4988 * assoc_value: This parameter specifies the maximum size in bytes.
4990 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4991 char __user *optval, int __user *optlen)
4993 struct sctp_assoc_value params;
4994 struct sctp_association *asoc;
4996 if (len == sizeof(int)) {
4997 pr_warn("Use of int in maxseg socket option deprecated\n");
4998 pr_warn("Use struct sctp_assoc_value instead\n");
4999 params.assoc_id = 0;
5000 } else if (len >= sizeof(struct sctp_assoc_value)) {
5001 len = sizeof(struct sctp_assoc_value);
5002 if (copy_from_user(¶ms, optval, sizeof(params)))
5007 asoc = sctp_id2assoc(sk, params.assoc_id);
5008 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5012 params.assoc_value = asoc->frag_point;
5014 params.assoc_value = sctp_sk(sk)->user_frag;
5016 if (put_user(len, optlen))
5018 if (len == sizeof(int)) {
5019 if (copy_to_user(optval, ¶ms.assoc_value, len))
5022 if (copy_to_user(optval, ¶ms, len))
5030 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5031 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5033 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5034 char __user *optval, int __user *optlen)
5038 if (len < sizeof(int))
5043 val = sctp_sk(sk)->frag_interleave;
5044 if (put_user(len, optlen))
5046 if (copy_to_user(optval, &val, len))
5053 * 7.1.25. Set or Get the sctp partial delivery point
5054 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5056 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5057 char __user *optval,
5062 if (len < sizeof(u32))
5067 val = sctp_sk(sk)->pd_point;
5068 if (put_user(len, optlen))
5070 if (copy_to_user(optval, &val, len))
5077 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5078 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5080 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5081 char __user *optval,
5084 struct sctp_assoc_value params;
5085 struct sctp_sock *sp;
5086 struct sctp_association *asoc;
5088 if (len == sizeof(int)) {
5089 pr_warn("Use of int in max_burst socket option deprecated\n");
5090 pr_warn("Use struct sctp_assoc_value instead\n");
5091 params.assoc_id = 0;
5092 } else if (len >= sizeof(struct sctp_assoc_value)) {
5093 len = sizeof(struct sctp_assoc_value);
5094 if (copy_from_user(¶ms, optval, len))
5101 if (params.assoc_id != 0) {
5102 asoc = sctp_id2assoc(sk, params.assoc_id);
5105 params.assoc_value = asoc->max_burst;
5107 params.assoc_value = sp->max_burst;
5109 if (len == sizeof(int)) {
5110 if (copy_to_user(optval, ¶ms.assoc_value, len))
5113 if (copy_to_user(optval, ¶ms, len))
5121 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5122 char __user *optval, int __user *optlen)
5124 struct sctp_hmacalgo __user *p = (void __user *)optval;
5125 struct sctp_hmac_algo_param *hmacs;
5129 if (!sctp_auth_enable)
5132 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5133 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5135 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5138 len = sizeof(struct sctp_hmacalgo) + data_len;
5139 num_idents = data_len / sizeof(u16);
5141 if (put_user(len, optlen))
5143 if (put_user(num_idents, &p->shmac_num_idents))
5145 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5150 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5151 char __user *optval, int __user *optlen)
5153 struct sctp_authkeyid val;
5154 struct sctp_association *asoc;
5156 if (!sctp_auth_enable)
5159 if (len < sizeof(struct sctp_authkeyid))
5161 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5164 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5165 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5169 val.scact_keynumber = asoc->active_key_id;
5171 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5173 len = sizeof(struct sctp_authkeyid);
5174 if (put_user(len, optlen))
5176 if (copy_to_user(optval, &val, len))
5182 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5183 char __user *optval, int __user *optlen)
5185 struct sctp_authchunks __user *p = (void __user *)optval;
5186 struct sctp_authchunks val;
5187 struct sctp_association *asoc;
5188 struct sctp_chunks_param *ch;
5192 if (!sctp_auth_enable)
5195 if (len < sizeof(struct sctp_authchunks))
5198 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5201 to = p->gauth_chunks;
5202 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5206 ch = asoc->peer.peer_chunks;
5210 /* See if the user provided enough room for all the data */
5211 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5212 if (len < num_chunks)
5215 if (copy_to_user(to, ch->chunks, num_chunks))
5218 len = sizeof(struct sctp_authchunks) + num_chunks;
5219 if (put_user(len, optlen)) return -EFAULT;
5220 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5225 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5226 char __user *optval, int __user *optlen)
5228 struct sctp_authchunks __user *p = (void __user *)optval;
5229 struct sctp_authchunks val;
5230 struct sctp_association *asoc;
5231 struct sctp_chunks_param *ch;
5235 if (!sctp_auth_enable)
5238 if (len < sizeof(struct sctp_authchunks))
5241 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5244 to = p->gauth_chunks;
5245 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5246 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5250 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5252 ch = sctp_sk(sk)->ep->auth_chunk_list;
5257 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5258 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5261 if (copy_to_user(to, ch->chunks, num_chunks))
5264 len = sizeof(struct sctp_authchunks) + num_chunks;
5265 if (put_user(len, optlen))
5267 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5274 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5275 * This option gets the current number of associations that are attached
5276 * to a one-to-many style socket. The option value is an uint32_t.
5278 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5279 char __user *optval, int __user *optlen)
5281 struct sctp_sock *sp = sctp_sk(sk);
5282 struct sctp_association *asoc;
5285 if (sctp_style(sk, TCP))
5288 if (len < sizeof(u32))
5293 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5297 if (put_user(len, optlen))
5299 if (copy_to_user(optval, &val, len))
5306 * 8.2.6. Get the Current Identifiers of Associations
5307 * (SCTP_GET_ASSOC_ID_LIST)
5309 * This option gets the current list of SCTP association identifiers of
5310 * the SCTP associations handled by a one-to-many style socket.
5312 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5313 char __user *optval, int __user *optlen)
5315 struct sctp_sock *sp = sctp_sk(sk);
5316 struct sctp_association *asoc;
5317 struct sctp_assoc_ids *ids;
5320 if (sctp_style(sk, TCP))
5323 if (len < sizeof(struct sctp_assoc_ids))
5326 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5330 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5333 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5335 ids = kmalloc(len, GFP_KERNEL);
5339 ids->gaids_number_of_ids = num;
5341 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5342 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5345 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5354 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5355 char __user *optval, int __user *optlen)
5360 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5363 /* I can hardly begin to describe how wrong this is. This is
5364 * so broken as to be worse than useless. The API draft
5365 * REALLY is NOT helpful here... I am not convinced that the
5366 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5367 * are at all well-founded.
5369 if (level != SOL_SCTP) {
5370 struct sctp_af *af = sctp_sk(sk)->pf->af;
5372 retval = af->getsockopt(sk, level, optname, optval, optlen);
5376 if (get_user(len, optlen))
5383 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5385 case SCTP_DISABLE_FRAGMENTS:
5386 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5390 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5392 case SCTP_AUTOCLOSE:
5393 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5395 case SCTP_SOCKOPT_PEELOFF:
5396 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5398 case SCTP_PEER_ADDR_PARAMS:
5399 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5402 case SCTP_DELAYED_SACK:
5403 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5407 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5409 case SCTP_GET_PEER_ADDRS:
5410 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5413 case SCTP_GET_LOCAL_ADDRS:
5414 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5417 case SCTP_SOCKOPT_CONNECTX3:
5418 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5420 case SCTP_DEFAULT_SEND_PARAM:
5421 retval = sctp_getsockopt_default_send_param(sk, len,
5424 case SCTP_PRIMARY_ADDR:
5425 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5428 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5431 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5433 case SCTP_ASSOCINFO:
5434 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5436 case SCTP_I_WANT_MAPPED_V4_ADDR:
5437 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5440 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5442 case SCTP_GET_PEER_ADDR_INFO:
5443 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5446 case SCTP_ADAPTATION_LAYER:
5447 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5451 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5453 case SCTP_FRAGMENT_INTERLEAVE:
5454 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5457 case SCTP_PARTIAL_DELIVERY_POINT:
5458 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5461 case SCTP_MAX_BURST:
5462 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5465 case SCTP_AUTH_CHUNK:
5466 case SCTP_AUTH_DELETE_KEY:
5467 retval = -EOPNOTSUPP;
5469 case SCTP_HMAC_IDENT:
5470 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5472 case SCTP_AUTH_ACTIVE_KEY:
5473 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5475 case SCTP_PEER_AUTH_CHUNKS:
5476 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5479 case SCTP_LOCAL_AUTH_CHUNKS:
5480 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5483 case SCTP_GET_ASSOC_NUMBER:
5484 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5486 case SCTP_GET_ASSOC_ID_LIST:
5487 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5490 retval = -ENOPROTOOPT;
5494 sctp_release_sock(sk);
5498 static void sctp_hash(struct sock *sk)
5503 static void sctp_unhash(struct sock *sk)
5508 /* Check if port is acceptable. Possibly find first available port.
5510 * The port hash table (contained in the 'global' SCTP protocol storage
5511 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5512 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5513 * list (the list number is the port number hashed out, so as you
5514 * would expect from a hash function, all the ports in a given list have
5515 * such a number that hashes out to the same list number; you were
5516 * expecting that, right?); so each list has a set of ports, with a
5517 * link to the socket (struct sock) that uses it, the port number and
5518 * a fastreuse flag (FIXME: NPI ipg).
5520 static struct sctp_bind_bucket *sctp_bucket_create(
5521 struct sctp_bind_hashbucket *head, unsigned short snum);
5523 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5525 struct sctp_bind_hashbucket *head; /* hash list */
5526 struct sctp_bind_bucket *pp; /* hash list port iterator */
5527 struct hlist_node *node;
5528 unsigned short snum;
5531 snum = ntohs(addr->v4.sin_port);
5533 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5534 sctp_local_bh_disable();
5537 /* Search for an available port. */
5538 int low, high, remaining, index;
5541 inet_get_local_port_range(&low, &high);
5542 remaining = (high - low) + 1;
5543 rover = net_random() % remaining + low;
5547 if ((rover < low) || (rover > high))
5549 if (inet_is_reserved_local_port(rover))
5551 index = sctp_phashfn(rover);
5552 head = &sctp_port_hashtable[index];
5553 sctp_spin_lock(&head->lock);
5554 sctp_for_each_hentry(pp, node, &head->chain)
5555 if (pp->port == rover)
5559 sctp_spin_unlock(&head->lock);
5560 } while (--remaining > 0);
5562 /* Exhausted local port range during search? */
5567 /* OK, here is the one we will use. HEAD (the port
5568 * hash table list entry) is non-NULL and we hold it's
5573 /* We are given an specific port number; we verify
5574 * that it is not being used. If it is used, we will
5575 * exahust the search in the hash list corresponding
5576 * to the port number (snum) - we detect that with the
5577 * port iterator, pp being NULL.
5579 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5580 sctp_spin_lock(&head->lock);
5581 sctp_for_each_hentry(pp, node, &head->chain) {
5582 if (pp->port == snum)
5589 if (!hlist_empty(&pp->owner)) {
5590 /* We had a port hash table hit - there is an
5591 * available port (pp != NULL) and it is being
5592 * used by other socket (pp->owner not empty); that other
5593 * socket is going to be sk2.
5595 int reuse = sk->sk_reuse;
5598 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5599 if (pp->fastreuse && sk->sk_reuse &&
5600 sk->sk_state != SCTP_SS_LISTENING)
5603 /* Run through the list of sockets bound to the port
5604 * (pp->port) [via the pointers bind_next and
5605 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5606 * we get the endpoint they describe and run through
5607 * the endpoint's list of IP (v4 or v6) addresses,
5608 * comparing each of the addresses with the address of
5609 * the socket sk. If we find a match, then that means
5610 * that this port/socket (sk) combination are already
5613 sk_for_each_bound(sk2, node, &pp->owner) {
5614 struct sctp_endpoint *ep2;
5615 ep2 = sctp_sk(sk2)->ep;
5618 (reuse && sk2->sk_reuse &&
5619 sk2->sk_state != SCTP_SS_LISTENING))
5622 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5623 sctp_sk(sk2), sctp_sk(sk))) {
5628 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5631 /* If there was a hash table miss, create a new port. */
5633 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5636 /* In either case (hit or miss), make sure fastreuse is 1 only
5637 * if sk->sk_reuse is too (that is, if the caller requested
5638 * SO_REUSEADDR on this socket -sk-).
5640 if (hlist_empty(&pp->owner)) {
5641 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5645 } else if (pp->fastreuse &&
5646 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5649 /* We are set, so fill up all the data in the hash table
5650 * entry, tie the socket list information with the rest of the
5651 * sockets FIXME: Blurry, NPI (ipg).
5654 if (!sctp_sk(sk)->bind_hash) {
5655 inet_sk(sk)->inet_num = snum;
5656 sk_add_bind_node(sk, &pp->owner);
5657 sctp_sk(sk)->bind_hash = pp;
5662 sctp_spin_unlock(&head->lock);
5665 sctp_local_bh_enable();
5669 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5670 * port is requested.
5672 static int sctp_get_port(struct sock *sk, unsigned short snum)
5675 union sctp_addr addr;
5676 struct sctp_af *af = sctp_sk(sk)->pf->af;
5678 /* Set up a dummy address struct from the sk. */
5679 af->from_sk(&addr, sk);
5680 addr.v4.sin_port = htons(snum);
5682 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5683 ret = sctp_get_port_local(sk, &addr);
5689 * Move a socket to LISTENING state.
5691 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5693 struct sctp_sock *sp = sctp_sk(sk);
5694 struct sctp_endpoint *ep = sp->ep;
5695 struct crypto_hash *tfm = NULL;
5697 /* Allocate HMAC for generating cookie. */
5698 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5699 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5701 if (net_ratelimit()) {
5702 pr_info("failed to load transform for %s: %ld\n",
5703 sctp_hmac_alg, PTR_ERR(tfm));
5707 sctp_sk(sk)->hmac = tfm;
5711 * If a bind() or sctp_bindx() is not called prior to a listen()
5712 * call that allows new associations to be accepted, the system
5713 * picks an ephemeral port and will choose an address set equivalent
5714 * to binding with a wildcard address.
5716 * This is not currently spelled out in the SCTP sockets
5717 * extensions draft, but follows the practice as seen in TCP
5721 sk->sk_state = SCTP_SS_LISTENING;
5722 if (!ep->base.bind_addr.port) {
5723 if (sctp_autobind(sk))
5726 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
5727 sk->sk_state = SCTP_SS_CLOSED;
5732 sk->sk_max_ack_backlog = backlog;
5733 sctp_hash_endpoint(ep);
5738 * 4.1.3 / 5.1.3 listen()
5740 * By default, new associations are not accepted for UDP style sockets.
5741 * An application uses listen() to mark a socket as being able to
5742 * accept new associations.
5744 * On TCP style sockets, applications use listen() to ready the SCTP
5745 * endpoint for accepting inbound associations.
5747 * On both types of endpoints a backlog of '0' disables listening.
5749 * Move a socket to LISTENING state.
5751 int sctp_inet_listen(struct socket *sock, int backlog)
5753 struct sock *sk = sock->sk;
5754 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5757 if (unlikely(backlog < 0))
5762 /* Peeled-off sockets are not allowed to listen(). */
5763 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
5766 if (sock->state != SS_UNCONNECTED)
5769 /* If backlog is zero, disable listening. */
5771 if (sctp_sstate(sk, CLOSED))
5775 sctp_unhash_endpoint(ep);
5776 sk->sk_state = SCTP_SS_CLOSED;
5778 sctp_sk(sk)->bind_hash->fastreuse = 1;
5782 /* If we are already listening, just update the backlog */
5783 if (sctp_sstate(sk, LISTENING))
5784 sk->sk_max_ack_backlog = backlog;
5786 err = sctp_listen_start(sk, backlog);
5793 sctp_release_sock(sk);
5798 * This function is done by modeling the current datagram_poll() and the
5799 * tcp_poll(). Note that, based on these implementations, we don't
5800 * lock the socket in this function, even though it seems that,
5801 * ideally, locking or some other mechanisms can be used to ensure
5802 * the integrity of the counters (sndbuf and wmem_alloc) used
5803 * in this place. We assume that we don't need locks either until proven
5806 * Another thing to note is that we include the Async I/O support
5807 * here, again, by modeling the current TCP/UDP code. We don't have
5808 * a good way to test with it yet.
5810 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5812 struct sock *sk = sock->sk;
5813 struct sctp_sock *sp = sctp_sk(sk);
5816 poll_wait(file, sk_sleep(sk), wait);
5818 /* A TCP-style listening socket becomes readable when the accept queue
5821 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5822 return (!list_empty(&sp->ep->asocs)) ?
5823 (POLLIN | POLLRDNORM) : 0;
5827 /* Is there any exceptional events? */
5828 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5830 if (sk->sk_shutdown & RCV_SHUTDOWN)
5831 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
5832 if (sk->sk_shutdown == SHUTDOWN_MASK)
5835 /* Is it readable? Reconsider this code with TCP-style support. */
5836 if (!skb_queue_empty(&sk->sk_receive_queue))
5837 mask |= POLLIN | POLLRDNORM;
5839 /* The association is either gone or not ready. */
5840 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5843 /* Is it writable? */
5844 if (sctp_writeable(sk)) {
5845 mask |= POLLOUT | POLLWRNORM;
5847 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5849 * Since the socket is not locked, the buffer
5850 * might be made available after the writeable check and
5851 * before the bit is set. This could cause a lost I/O
5852 * signal. tcp_poll() has a race breaker for this race
5853 * condition. Based on their implementation, we put
5854 * in the following code to cover it as well.
5856 if (sctp_writeable(sk))
5857 mask |= POLLOUT | POLLWRNORM;
5862 /********************************************************************
5863 * 2nd Level Abstractions
5864 ********************************************************************/
5866 static struct sctp_bind_bucket *sctp_bucket_create(
5867 struct sctp_bind_hashbucket *head, unsigned short snum)
5869 struct sctp_bind_bucket *pp;
5871 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
5873 SCTP_DBG_OBJCNT_INC(bind_bucket);
5876 INIT_HLIST_HEAD(&pp->owner);
5877 hlist_add_head(&pp->node, &head->chain);
5882 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5883 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5885 if (pp && hlist_empty(&pp->owner)) {
5886 __hlist_del(&pp->node);
5887 kmem_cache_free(sctp_bucket_cachep, pp);
5888 SCTP_DBG_OBJCNT_DEC(bind_bucket);
5892 /* Release this socket's reference to a local port. */
5893 static inline void __sctp_put_port(struct sock *sk)
5895 struct sctp_bind_hashbucket *head =
5896 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->inet_num)];
5897 struct sctp_bind_bucket *pp;
5899 sctp_spin_lock(&head->lock);
5900 pp = sctp_sk(sk)->bind_hash;
5901 __sk_del_bind_node(sk);
5902 sctp_sk(sk)->bind_hash = NULL;
5903 inet_sk(sk)->inet_num = 0;
5904 sctp_bucket_destroy(pp);
5905 sctp_spin_unlock(&head->lock);
5908 void sctp_put_port(struct sock *sk)
5910 sctp_local_bh_disable();
5911 __sctp_put_port(sk);
5912 sctp_local_bh_enable();
5916 * The system picks an ephemeral port and choose an address set equivalent
5917 * to binding with a wildcard address.
5918 * One of those addresses will be the primary address for the association.
5919 * This automatically enables the multihoming capability of SCTP.
5921 static int sctp_autobind(struct sock *sk)
5923 union sctp_addr autoaddr;
5927 /* Initialize a local sockaddr structure to INADDR_ANY. */
5928 af = sctp_sk(sk)->pf->af;
5930 port = htons(inet_sk(sk)->inet_num);
5931 af->inaddr_any(&autoaddr, port);
5933 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5936 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5939 * 4.2 The cmsghdr Structure *
5941 * When ancillary data is sent or received, any number of ancillary data
5942 * objects can be specified by the msg_control and msg_controllen members of
5943 * the msghdr structure, because each object is preceded by
5944 * a cmsghdr structure defining the object's length (the cmsg_len member).
5945 * Historically Berkeley-derived implementations have passed only one object
5946 * at a time, but this API allows multiple objects to be
5947 * passed in a single call to sendmsg() or recvmsg(). The following example
5948 * shows two ancillary data objects in a control buffer.
5950 * |<--------------------------- msg_controllen -------------------------->|
5953 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5955 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5958 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5960 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5963 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5964 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5966 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5968 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5975 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5976 sctp_cmsgs_t *cmsgs)
5978 struct cmsghdr *cmsg;
5979 struct msghdr *my_msg = (struct msghdr *)msg;
5981 for (cmsg = CMSG_FIRSTHDR(msg);
5983 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
5984 if (!CMSG_OK(my_msg, cmsg))
5987 /* Should we parse this header or ignore? */
5988 if (cmsg->cmsg_level != IPPROTO_SCTP)
5991 /* Strictly check lengths following example in SCM code. */
5992 switch (cmsg->cmsg_type) {
5994 /* SCTP Socket API Extension
5995 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5997 * This cmsghdr structure provides information for
5998 * initializing new SCTP associations with sendmsg().
5999 * The SCTP_INITMSG socket option uses this same data
6000 * structure. This structure is not used for
6003 * cmsg_level cmsg_type cmsg_data[]
6004 * ------------ ------------ ----------------------
6005 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6007 if (cmsg->cmsg_len !=
6008 CMSG_LEN(sizeof(struct sctp_initmsg)))
6010 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6014 /* SCTP Socket API Extension
6015 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6017 * This cmsghdr structure specifies SCTP options for
6018 * sendmsg() and describes SCTP header information
6019 * about a received message through recvmsg().
6021 * cmsg_level cmsg_type cmsg_data[]
6022 * ------------ ------------ ----------------------
6023 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6025 if (cmsg->cmsg_len !=
6026 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6030 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6032 /* Minimally, validate the sinfo_flags. */
6033 if (cmsgs->info->sinfo_flags &
6034 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6035 SCTP_ABORT | SCTP_EOF))
6047 * Wait for a packet..
6048 * Note: This function is the same function as in core/datagram.c
6049 * with a few modifications to make lksctp work.
6051 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6056 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6058 /* Socket errors? */
6059 error = sock_error(sk);
6063 if (!skb_queue_empty(&sk->sk_receive_queue))
6066 /* Socket shut down? */
6067 if (sk->sk_shutdown & RCV_SHUTDOWN)
6070 /* Sequenced packets can come disconnected. If so we report the
6075 /* Is there a good reason to think that we may receive some data? */
6076 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6079 /* Handle signals. */
6080 if (signal_pending(current))
6083 /* Let another process have a go. Since we are going to sleep
6084 * anyway. Note: This may cause odd behaviors if the message
6085 * does not fit in the user's buffer, but this seems to be the
6086 * only way to honor MSG_DONTWAIT realistically.
6088 sctp_release_sock(sk);
6089 *timeo_p = schedule_timeout(*timeo_p);
6093 finish_wait(sk_sleep(sk), &wait);
6097 error = sock_intr_errno(*timeo_p);
6100 finish_wait(sk_sleep(sk), &wait);
6105 /* Receive a datagram.
6106 * Note: This is pretty much the same routine as in core/datagram.c
6107 * with a few changes to make lksctp work.
6109 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6110 int noblock, int *err)
6113 struct sk_buff *skb;
6116 timeo = sock_rcvtimeo(sk, noblock);
6118 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6119 timeo, MAX_SCHEDULE_TIMEOUT);
6122 /* Again only user level code calls this function,
6123 * so nothing interrupt level
6124 * will suddenly eat the receive_queue.
6126 * Look at current nfs client by the way...
6127 * However, this function was correct in any case. 8)
6129 if (flags & MSG_PEEK) {
6130 spin_lock_bh(&sk->sk_receive_queue.lock);
6131 skb = skb_peek(&sk->sk_receive_queue);
6133 atomic_inc(&skb->users);
6134 spin_unlock_bh(&sk->sk_receive_queue.lock);
6136 skb = skb_dequeue(&sk->sk_receive_queue);
6142 /* Caller is allowed not to check sk->sk_err before calling. */
6143 error = sock_error(sk);
6147 if (sk->sk_shutdown & RCV_SHUTDOWN)
6150 /* User doesn't want to wait. */
6154 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6163 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6164 static void __sctp_write_space(struct sctp_association *asoc)
6166 struct sock *sk = asoc->base.sk;
6167 struct socket *sock = sk->sk_socket;
6169 if ((sctp_wspace(asoc) > 0) && sock) {
6170 if (waitqueue_active(&asoc->wait))
6171 wake_up_interruptible(&asoc->wait);
6173 if (sctp_writeable(sk)) {
6174 wait_queue_head_t *wq = sk_sleep(sk);
6176 if (wq && waitqueue_active(wq))
6177 wake_up_interruptible(wq);
6179 /* Note that we try to include the Async I/O support
6180 * here by modeling from the current TCP/UDP code.
6181 * We have not tested with it yet.
6183 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6184 sock_wake_async(sock,
6185 SOCK_WAKE_SPACE, POLL_OUT);
6190 /* Do accounting for the sndbuf space.
6191 * Decrement the used sndbuf space of the corresponding association by the
6192 * data size which was just transmitted(freed).
6194 static void sctp_wfree(struct sk_buff *skb)
6196 struct sctp_association *asoc;
6197 struct sctp_chunk *chunk;
6200 /* Get the saved chunk pointer. */
6201 chunk = *((struct sctp_chunk **)(skb->cb));
6204 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6205 sizeof(struct sk_buff) +
6206 sizeof(struct sctp_chunk);
6208 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6211 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6213 sk->sk_wmem_queued -= skb->truesize;
6214 sk_mem_uncharge(sk, skb->truesize);
6217 __sctp_write_space(asoc);
6219 sctp_association_put(asoc);
6222 /* Do accounting for the receive space on the socket.
6223 * Accounting for the association is done in ulpevent.c
6224 * We set this as a destructor for the cloned data skbs so that
6225 * accounting is done at the correct time.
6227 void sctp_sock_rfree(struct sk_buff *skb)
6229 struct sock *sk = skb->sk;
6230 struct sctp_ulpevent *event = sctp_skb2event(skb);
6232 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6235 * Mimic the behavior of sock_rfree
6237 sk_mem_uncharge(sk, event->rmem_len);
6241 /* Helper function to wait for space in the sndbuf. */
6242 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6245 struct sock *sk = asoc->base.sk;
6247 long current_timeo = *timeo_p;
6250 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6251 asoc, (long)(*timeo_p), msg_len);
6253 /* Increment the association's refcnt. */
6254 sctp_association_hold(asoc);
6256 /* Wait on the association specific sndbuf space. */
6258 prepare_to_wait_exclusive(&asoc->wait, &wait,
6259 TASK_INTERRUPTIBLE);
6262 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6265 if (signal_pending(current))
6266 goto do_interrupted;
6267 if (msg_len <= sctp_wspace(asoc))
6270 /* Let another process have a go. Since we are going
6273 sctp_release_sock(sk);
6274 current_timeo = schedule_timeout(current_timeo);
6275 BUG_ON(sk != asoc->base.sk);
6278 *timeo_p = current_timeo;
6282 finish_wait(&asoc->wait, &wait);
6284 /* Release the association's refcnt. */
6285 sctp_association_put(asoc);
6294 err = sock_intr_errno(*timeo_p);
6302 void sctp_data_ready(struct sock *sk, int len)
6304 struct socket_wq *wq;
6307 wq = rcu_dereference(sk->sk_wq);
6308 if (wq_has_sleeper(wq))
6309 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6310 POLLRDNORM | POLLRDBAND);
6311 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6315 /* If socket sndbuf has changed, wake up all per association waiters. */
6316 void sctp_write_space(struct sock *sk)
6318 struct sctp_association *asoc;
6320 /* Wake up the tasks in each wait queue. */
6321 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6322 __sctp_write_space(asoc);
6326 /* Is there any sndbuf space available on the socket?
6328 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6329 * associations on the same socket. For a UDP-style socket with
6330 * multiple associations, it is possible for it to be "unwriteable"
6331 * prematurely. I assume that this is acceptable because
6332 * a premature "unwriteable" is better than an accidental "writeable" which
6333 * would cause an unwanted block under certain circumstances. For the 1-1
6334 * UDP-style sockets or TCP-style sockets, this code should work.
6337 static int sctp_writeable(struct sock *sk)
6341 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6347 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6348 * returns immediately with EINPROGRESS.
6350 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6352 struct sock *sk = asoc->base.sk;
6354 long current_timeo = *timeo_p;
6357 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6360 /* Increment the association's refcnt. */
6361 sctp_association_hold(asoc);
6364 prepare_to_wait_exclusive(&asoc->wait, &wait,
6365 TASK_INTERRUPTIBLE);
6368 if (sk->sk_shutdown & RCV_SHUTDOWN)
6370 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6373 if (signal_pending(current))
6374 goto do_interrupted;
6376 if (sctp_state(asoc, ESTABLISHED))
6379 /* Let another process have a go. Since we are going
6382 sctp_release_sock(sk);
6383 current_timeo = schedule_timeout(current_timeo);
6386 *timeo_p = current_timeo;
6390 finish_wait(&asoc->wait, &wait);
6392 /* Release the association's refcnt. */
6393 sctp_association_put(asoc);
6398 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6401 err = -ECONNREFUSED;
6405 err = sock_intr_errno(*timeo_p);
6413 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6415 struct sctp_endpoint *ep;
6419 ep = sctp_sk(sk)->ep;
6423 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6424 TASK_INTERRUPTIBLE);
6426 if (list_empty(&ep->asocs)) {
6427 sctp_release_sock(sk);
6428 timeo = schedule_timeout(timeo);
6433 if (!sctp_sstate(sk, LISTENING))
6437 if (!list_empty(&ep->asocs))
6440 err = sock_intr_errno(timeo);
6441 if (signal_pending(current))
6449 finish_wait(sk_sleep(sk), &wait);
6454 static void sctp_wait_for_close(struct sock *sk, long timeout)
6459 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6460 if (list_empty(&sctp_sk(sk)->ep->asocs))
6462 sctp_release_sock(sk);
6463 timeout = schedule_timeout(timeout);
6465 } while (!signal_pending(current) && timeout);
6467 finish_wait(sk_sleep(sk), &wait);
6470 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6472 struct sk_buff *frag;
6477 /* Don't forget the fragments. */
6478 skb_walk_frags(skb, frag)
6479 sctp_skb_set_owner_r_frag(frag, sk);
6482 sctp_skb_set_owner_r(skb, sk);
6485 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6486 struct sctp_association *asoc)
6488 struct inet_sock *inet = inet_sk(sk);
6489 struct inet_sock *newinet;
6491 newsk->sk_type = sk->sk_type;
6492 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6493 newsk->sk_flags = sk->sk_flags;
6494 newsk->sk_no_check = sk->sk_no_check;
6495 newsk->sk_reuse = sk->sk_reuse;
6497 newsk->sk_shutdown = sk->sk_shutdown;
6498 newsk->sk_destruct = inet_sock_destruct;
6499 newsk->sk_family = sk->sk_family;
6500 newsk->sk_protocol = IPPROTO_SCTP;
6501 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6502 newsk->sk_sndbuf = sk->sk_sndbuf;
6503 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6504 newsk->sk_lingertime = sk->sk_lingertime;
6505 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6506 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6508 newinet = inet_sk(newsk);
6510 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6511 * getsockname() and getpeername()
6513 newinet->inet_sport = inet->inet_sport;
6514 newinet->inet_saddr = inet->inet_saddr;
6515 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6516 newinet->inet_dport = htons(asoc->peer.port);
6517 newinet->pmtudisc = inet->pmtudisc;
6518 newinet->inet_id = asoc->next_tsn ^ jiffies;
6520 newinet->uc_ttl = inet->uc_ttl;
6521 newinet->mc_loop = 1;
6522 newinet->mc_ttl = 1;
6523 newinet->mc_index = 0;
6524 newinet->mc_list = NULL;
6527 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6528 * and its messages to the newsk.
6530 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6531 struct sctp_association *assoc,
6532 sctp_socket_type_t type)
6534 struct sctp_sock *oldsp = sctp_sk(oldsk);
6535 struct sctp_sock *newsp = sctp_sk(newsk);
6536 struct sctp_bind_bucket *pp; /* hash list port iterator */
6537 struct sctp_endpoint *newep = newsp->ep;
6538 struct sk_buff *skb, *tmp;
6539 struct sctp_ulpevent *event;
6540 struct sctp_bind_hashbucket *head;
6542 /* Migrate socket buffer sizes and all the socket level options to the
6545 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6546 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6547 /* Brute force copy old sctp opt. */
6548 inet_sk_copy_descendant(newsk, oldsk);
6550 /* Restore the ep value that was overwritten with the above structure
6556 /* Hook this new socket in to the bind_hash list. */
6557 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->inet_num)];
6558 sctp_local_bh_disable();
6559 sctp_spin_lock(&head->lock);
6560 pp = sctp_sk(oldsk)->bind_hash;
6561 sk_add_bind_node(newsk, &pp->owner);
6562 sctp_sk(newsk)->bind_hash = pp;
6563 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6564 sctp_spin_unlock(&head->lock);
6565 sctp_local_bh_enable();
6567 /* Copy the bind_addr list from the original endpoint to the new
6568 * endpoint so that we can handle restarts properly
6570 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6571 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6573 /* Move any messages in the old socket's receive queue that are for the
6574 * peeled off association to the new socket's receive queue.
6576 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6577 event = sctp_skb2event(skb);
6578 if (event->asoc == assoc) {
6579 __skb_unlink(skb, &oldsk->sk_receive_queue);
6580 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6581 sctp_skb_set_owner_r_frag(skb, newsk);
6585 /* Clean up any messages pending delivery due to partial
6586 * delivery. Three cases:
6587 * 1) No partial deliver; no work.
6588 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6589 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6591 skb_queue_head_init(&newsp->pd_lobby);
6592 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6594 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6595 struct sk_buff_head *queue;
6597 /* Decide which queue to move pd_lobby skbs to. */
6598 if (assoc->ulpq.pd_mode) {
6599 queue = &newsp->pd_lobby;
6601 queue = &newsk->sk_receive_queue;
6603 /* Walk through the pd_lobby, looking for skbs that
6604 * need moved to the new socket.
6606 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6607 event = sctp_skb2event(skb);
6608 if (event->asoc == assoc) {
6609 __skb_unlink(skb, &oldsp->pd_lobby);
6610 __skb_queue_tail(queue, skb);
6611 sctp_skb_set_owner_r_frag(skb, newsk);
6615 /* Clear up any skbs waiting for the partial
6616 * delivery to finish.
6618 if (assoc->ulpq.pd_mode)
6619 sctp_clear_pd(oldsk, NULL);
6623 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6624 sctp_skb_set_owner_r_frag(skb, newsk);
6626 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6627 sctp_skb_set_owner_r_frag(skb, newsk);
6629 /* Set the type of socket to indicate that it is peeled off from the
6630 * original UDP-style socket or created with the accept() call on a
6631 * TCP-style socket..
6635 /* Mark the new socket "in-use" by the user so that any packets
6636 * that may arrive on the association after we've moved it are
6637 * queued to the backlog. This prevents a potential race between
6638 * backlog processing on the old socket and new-packet processing
6639 * on the new socket.
6641 * The caller has just allocated newsk so we can guarantee that other
6642 * paths won't try to lock it and then oldsk.
6644 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6645 sctp_assoc_migrate(assoc, newsk);
6647 /* If the association on the newsk is already closed before accept()
6648 * is called, set RCV_SHUTDOWN flag.
6650 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6651 newsk->sk_shutdown |= RCV_SHUTDOWN;
6653 newsk->sk_state = SCTP_SS_ESTABLISHED;
6654 sctp_release_sock(newsk);
6658 /* This proto struct describes the ULP interface for SCTP. */
6659 struct proto sctp_prot = {
6661 .owner = THIS_MODULE,
6662 .close = sctp_close,
6663 .connect = sctp_connect,
6664 .disconnect = sctp_disconnect,
6665 .accept = sctp_accept,
6666 .ioctl = sctp_ioctl,
6667 .init = sctp_init_sock,
6668 .destroy = sctp_destroy_sock,
6669 .shutdown = sctp_shutdown,
6670 .setsockopt = sctp_setsockopt,
6671 .getsockopt = sctp_getsockopt,
6672 .sendmsg = sctp_sendmsg,
6673 .recvmsg = sctp_recvmsg,
6675 .backlog_rcv = sctp_backlog_rcv,
6677 .unhash = sctp_unhash,
6678 .get_port = sctp_get_port,
6679 .obj_size = sizeof(struct sctp_sock),
6680 .sysctl_mem = sysctl_sctp_mem,
6681 .sysctl_rmem = sysctl_sctp_rmem,
6682 .sysctl_wmem = sysctl_sctp_wmem,
6683 .memory_pressure = &sctp_memory_pressure,
6684 .enter_memory_pressure = sctp_enter_memory_pressure,
6685 .memory_allocated = &sctp_memory_allocated,
6686 .sockets_allocated = &sctp_sockets_allocated,
6689 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6691 struct proto sctpv6_prot = {
6693 .owner = THIS_MODULE,
6694 .close = sctp_close,
6695 .connect = sctp_connect,
6696 .disconnect = sctp_disconnect,
6697 .accept = sctp_accept,
6698 .ioctl = sctp_ioctl,
6699 .init = sctp_init_sock,
6700 .destroy = sctp_destroy_sock,
6701 .shutdown = sctp_shutdown,
6702 .setsockopt = sctp_setsockopt,
6703 .getsockopt = sctp_getsockopt,
6704 .sendmsg = sctp_sendmsg,
6705 .recvmsg = sctp_recvmsg,
6707 .backlog_rcv = sctp_backlog_rcv,
6709 .unhash = sctp_unhash,
6710 .get_port = sctp_get_port,
6711 .obj_size = sizeof(struct sctp6_sock),
6712 .sysctl_mem = sysctl_sctp_mem,
6713 .sysctl_rmem = sysctl_sctp_rmem,
6714 .sysctl_wmem = sysctl_sctp_wmem,
6715 .memory_pressure = &sctp_memory_pressure,
6716 .enter_memory_pressure = sctp_enter_memory_pressure,
6717 .memory_allocated = &sctp_memory_allocated,
6718 .sockets_allocated = &sctp_sockets_allocated,
6720 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */