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
39 * Or submit a bug report through the following website:
40 * http://www.sf.net/projects/lksctp
42 * Written or modified by:
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>
73 #include <linux/file.h>
77 #include <net/route.h>
79 #include <net/inet_common.h>
81 #include <linux/socket.h> /* for sa_family_t */
82 #include <linux/export.h>
84 #include <net/sctp/sctp.h>
85 #include <net/sctp/sm.h>
87 /* WARNING: Please do not remove the SCTP_STATIC attribute to
88 * any of the functions below as they are used to export functions
89 * used by a project regression testsuite.
92 /* Forward declarations for internal helper functions. */
93 static int sctp_writeable(struct sock *sk);
94 static void sctp_wfree(struct sk_buff *skb);
95 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
97 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
98 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
99 static int sctp_wait_for_accept(struct sock *sk, long timeo);
100 static void sctp_wait_for_close(struct sock *sk, long timeo);
101 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
102 union sctp_addr *addr, int len);
103 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
104 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
105 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
106 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
107 static int sctp_send_asconf(struct sctp_association *asoc,
108 struct sctp_chunk *chunk);
109 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
110 static int sctp_autobind(struct sock *sk);
111 static void sctp_sock_migrate(struct sock *, struct sock *,
112 struct sctp_association *, sctp_socket_type_t);
114 extern struct kmem_cache *sctp_bucket_cachep;
115 extern long sysctl_sctp_mem[3];
116 extern int sysctl_sctp_rmem[3];
117 extern int sysctl_sctp_wmem[3];
119 static int sctp_memory_pressure;
120 static atomic_long_t sctp_memory_allocated;
121 struct percpu_counter sctp_sockets_allocated;
123 static void sctp_enter_memory_pressure(struct sock *sk)
125 sctp_memory_pressure = 1;
129 /* Get the sndbuf space available at the time on the association. */
130 static inline int sctp_wspace(struct sctp_association *asoc)
134 if (asoc->ep->sndbuf_policy)
135 amt = asoc->sndbuf_used;
137 amt = sk_wmem_alloc_get(asoc->base.sk);
139 if (amt >= asoc->base.sk->sk_sndbuf) {
140 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
143 amt = sk_stream_wspace(asoc->base.sk);
148 amt = asoc->base.sk->sk_sndbuf - amt;
153 /* Increment the used sndbuf space count of the corresponding association by
154 * the size of the outgoing data chunk.
155 * Also, set the skb destructor for sndbuf accounting later.
157 * Since it is always 1-1 between chunk and skb, and also a new skb is always
158 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
159 * destructor in the data chunk skb for the purpose of the sndbuf space
162 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
164 struct sctp_association *asoc = chunk->asoc;
165 struct sock *sk = asoc->base.sk;
167 /* The sndbuf space is tracked per association. */
168 sctp_association_hold(asoc);
170 skb_set_owner_w(chunk->skb, sk);
172 chunk->skb->destructor = sctp_wfree;
173 /* Save the chunk pointer in skb for sctp_wfree to use later. */
174 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
176 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
177 sizeof(struct sk_buff) +
178 sizeof(struct sctp_chunk);
180 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
181 sk->sk_wmem_queued += chunk->skb->truesize;
182 sk_mem_charge(sk, chunk->skb->truesize);
185 /* Verify that this is a valid address. */
186 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
191 /* Verify basic sockaddr. */
192 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
196 /* Is this a valid SCTP address? */
197 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
200 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
206 /* Look up the association by its id. If this is not a UDP-style
207 * socket, the ID field is always ignored.
209 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
211 struct sctp_association *asoc = NULL;
213 /* If this is not a UDP-style socket, assoc id should be ignored. */
214 if (!sctp_style(sk, UDP)) {
215 /* Return NULL if the socket state is not ESTABLISHED. It
216 * could be a TCP-style listening socket or a socket which
217 * hasn't yet called connect() to establish an association.
219 if (!sctp_sstate(sk, ESTABLISHED))
222 /* Get the first and the only association from the list. */
223 if (!list_empty(&sctp_sk(sk)->ep->asocs))
224 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
225 struct sctp_association, asocs);
229 /* Otherwise this is a UDP-style socket. */
230 if (!id || (id == (sctp_assoc_t)-1))
233 spin_lock_bh(&sctp_assocs_id_lock);
234 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
235 spin_unlock_bh(&sctp_assocs_id_lock);
237 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
243 /* Look up the transport from an address and an assoc id. If both address and
244 * id are specified, the associations matching the address and the id should be
247 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
248 struct sockaddr_storage *addr,
251 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
252 struct sctp_transport *transport;
253 union sctp_addr *laddr = (union sctp_addr *)addr;
255 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
262 id_asoc = sctp_id2assoc(sk, id);
263 if (id_asoc && (id_asoc != addr_asoc))
266 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
267 (union sctp_addr *)addr);
272 /* API 3.1.2 bind() - UDP Style Syntax
273 * The syntax of bind() is,
275 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
277 * sd - the socket descriptor returned by socket().
278 * addr - the address structure (struct sockaddr_in or struct
279 * sockaddr_in6 [RFC 2553]),
280 * addr_len - the size of the address structure.
282 SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
288 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
291 /* Disallow binding twice. */
292 if (!sctp_sk(sk)->ep->base.bind_addr.port)
293 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
298 sctp_release_sock(sk);
303 static long sctp_get_port_local(struct sock *, union sctp_addr *);
305 /* Verify this is a valid sockaddr. */
306 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
307 union sctp_addr *addr, int len)
311 /* Check minimum size. */
312 if (len < sizeof (struct sockaddr))
315 /* V4 mapped address are really of AF_INET family */
316 if (addr->sa.sa_family == AF_INET6 &&
317 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
318 if (!opt->pf->af_supported(AF_INET, opt))
321 /* Does this PF support this AF? */
322 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
326 /* If we get this far, af is valid. */
327 af = sctp_get_af_specific(addr->sa.sa_family);
329 if (len < af->sockaddr_len)
335 /* Bind a local address either to an endpoint or to an association. */
336 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
338 struct sctp_sock *sp = sctp_sk(sk);
339 struct sctp_endpoint *ep = sp->ep;
340 struct sctp_bind_addr *bp = &ep->base.bind_addr;
345 /* Common sockaddr verification. */
346 af = sctp_sockaddr_af(sp, addr, len);
348 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
353 snum = ntohs(addr->v4.sin_port);
355 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
356 ", port: %d, new port: %d, len: %d)\n",
362 /* PF specific bind() address verification. */
363 if (!sp->pf->bind_verify(sp, addr))
364 return -EADDRNOTAVAIL;
366 /* We must either be unbound, or bind to the same port.
367 * It's OK to allow 0 ports if we are already bound.
368 * We'll just inhert an already bound port in this case
373 else if (snum != bp->port) {
374 SCTP_DEBUG_PRINTK("sctp_do_bind:"
375 " New port %d does not match existing port "
376 "%d.\n", snum, bp->port);
381 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
384 /* See if the address matches any of the addresses we may have
385 * already bound before checking against other endpoints.
387 if (sctp_bind_addr_match(bp, addr, sp))
390 /* Make sure we are allowed to bind here.
391 * The function sctp_get_port_local() does duplicate address
394 addr->v4.sin_port = htons(snum);
395 if ((ret = sctp_get_port_local(sk, addr))) {
399 /* Refresh ephemeral port. */
401 bp->port = inet_sk(sk)->inet_num;
403 /* Add the address to the bind address list.
404 * Use GFP_ATOMIC since BHs will be disabled.
406 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
408 /* Copy back into socket for getsockname() use. */
410 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
411 af->to_sk_saddr(addr, sk);
417 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
419 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
420 * at any one time. If a sender, after sending an ASCONF chunk, decides
421 * it needs to transfer another ASCONF Chunk, it MUST wait until the
422 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
423 * subsequent ASCONF. Note this restriction binds each side, so at any
424 * time two ASCONF may be in-transit on any given association (one sent
425 * from each endpoint).
427 static int sctp_send_asconf(struct sctp_association *asoc,
428 struct sctp_chunk *chunk)
430 struct net *net = sock_net(asoc->base.sk);
433 /* If there is an outstanding ASCONF chunk, queue it for later
436 if (asoc->addip_last_asconf) {
437 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
441 /* Hold the chunk until an ASCONF_ACK is received. */
442 sctp_chunk_hold(chunk);
443 retval = sctp_primitive_ASCONF(net, asoc, chunk);
445 sctp_chunk_free(chunk);
447 asoc->addip_last_asconf = chunk;
453 /* Add a list of addresses as bind addresses to local endpoint or
456 * Basically run through each address specified in the addrs/addrcnt
457 * array/length pair, determine if it is IPv6 or IPv4 and call
458 * sctp_do_bind() on it.
460 * If any of them fails, then the operation will be reversed and the
461 * ones that were added will be removed.
463 * Only sctp_setsockopt_bindx() is supposed to call this function.
465 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
470 struct sockaddr *sa_addr;
473 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
477 for (cnt = 0; cnt < addrcnt; cnt++) {
478 /* The list may contain either IPv4 or IPv6 address;
479 * determine the address length for walking thru the list.
482 af = sctp_get_af_specific(sa_addr->sa_family);
488 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
491 addr_buf += af->sockaddr_len;
495 /* Failed. Cleanup the ones that have been added */
497 sctp_bindx_rem(sk, addrs, cnt);
505 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
506 * associations that are part of the endpoint indicating that a list of local
507 * addresses are added to the endpoint.
509 * If any of the addresses is already in the bind address list of the
510 * association, we do not send the chunk for that association. But it will not
511 * affect other associations.
513 * Only sctp_setsockopt_bindx() is supposed to call this function.
515 static int sctp_send_asconf_add_ip(struct sock *sk,
516 struct sockaddr *addrs,
519 struct net *net = sock_net(sk);
520 struct sctp_sock *sp;
521 struct sctp_endpoint *ep;
522 struct sctp_association *asoc;
523 struct sctp_bind_addr *bp;
524 struct sctp_chunk *chunk;
525 struct sctp_sockaddr_entry *laddr;
526 union sctp_addr *addr;
527 union sctp_addr saveaddr;
534 if (!net->sctp.addip_enable)
540 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
541 __func__, sk, addrs, addrcnt);
543 list_for_each_entry(asoc, &ep->asocs, asocs) {
545 if (!asoc->peer.asconf_capable)
548 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
551 if (!sctp_state(asoc, ESTABLISHED))
554 /* Check if any address in the packed array of addresses is
555 * in the bind address list of the association. If so,
556 * do not send the asconf chunk to its peer, but continue with
557 * other associations.
560 for (i = 0; i < addrcnt; i++) {
562 af = sctp_get_af_specific(addr->v4.sin_family);
568 if (sctp_assoc_lookup_laddr(asoc, addr))
571 addr_buf += af->sockaddr_len;
576 /* Use the first valid address in bind addr list of
577 * association as Address Parameter of ASCONF CHUNK.
579 bp = &asoc->base.bind_addr;
580 p = bp->address_list.next;
581 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
582 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
583 addrcnt, SCTP_PARAM_ADD_IP);
589 /* Add the new addresses to the bind address list with
590 * use_as_src set to 0.
593 for (i = 0; i < addrcnt; i++) {
595 af = sctp_get_af_specific(addr->v4.sin_family);
596 memcpy(&saveaddr, addr, af->sockaddr_len);
597 retval = sctp_add_bind_addr(bp, &saveaddr,
598 SCTP_ADDR_NEW, GFP_ATOMIC);
599 addr_buf += af->sockaddr_len;
601 if (asoc->src_out_of_asoc_ok) {
602 struct sctp_transport *trans;
604 list_for_each_entry(trans,
605 &asoc->peer.transport_addr_list, transports) {
606 /* Clear the source and route cache */
607 dst_release(trans->dst);
608 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
609 2*asoc->pathmtu, 4380));
610 trans->ssthresh = asoc->peer.i.a_rwnd;
611 trans->rto = asoc->rto_initial;
612 trans->rtt = trans->srtt = trans->rttvar = 0;
613 sctp_transport_route(trans, NULL,
614 sctp_sk(asoc->base.sk));
617 retval = sctp_send_asconf(asoc, chunk);
624 /* Remove a list of addresses from bind addresses list. Do not remove the
627 * Basically run through each address specified in the addrs/addrcnt
628 * array/length pair, determine if it is IPv6 or IPv4 and call
629 * sctp_del_bind() on it.
631 * If any of them fails, then the operation will be reversed and the
632 * ones that were removed will be added back.
634 * At least one address has to be left; if only one address is
635 * available, the operation will return -EBUSY.
637 * Only sctp_setsockopt_bindx() is supposed to call this function.
639 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
641 struct sctp_sock *sp = sctp_sk(sk);
642 struct sctp_endpoint *ep = sp->ep;
644 struct sctp_bind_addr *bp = &ep->base.bind_addr;
647 union sctp_addr *sa_addr;
650 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
654 for (cnt = 0; cnt < addrcnt; cnt++) {
655 /* If the bind address list is empty or if there is only one
656 * bind address, there is nothing more to be removed (we need
657 * at least one address here).
659 if (list_empty(&bp->address_list) ||
660 (sctp_list_single_entry(&bp->address_list))) {
666 af = sctp_get_af_specific(sa_addr->sa.sa_family);
672 if (!af->addr_valid(sa_addr, sp, NULL)) {
673 retval = -EADDRNOTAVAIL;
677 if (sa_addr->v4.sin_port &&
678 sa_addr->v4.sin_port != htons(bp->port)) {
683 if (!sa_addr->v4.sin_port)
684 sa_addr->v4.sin_port = htons(bp->port);
686 /* FIXME - There is probably a need to check if sk->sk_saddr and
687 * sk->sk_rcv_addr are currently set to one of the addresses to
688 * be removed. This is something which needs to be looked into
689 * when we are fixing the outstanding issues with multi-homing
690 * socket routing and failover schemes. Refer to comments in
691 * sctp_do_bind(). -daisy
693 retval = sctp_del_bind_addr(bp, sa_addr);
695 addr_buf += af->sockaddr_len;
698 /* Failed. Add the ones that has been removed back */
700 sctp_bindx_add(sk, addrs, cnt);
708 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
709 * the associations that are part of the endpoint indicating that a list of
710 * local addresses are removed from the endpoint.
712 * If any of the addresses is already in the bind address list of the
713 * association, we do not send the chunk for that association. But it will not
714 * affect other associations.
716 * Only sctp_setsockopt_bindx() is supposed to call this function.
718 static int sctp_send_asconf_del_ip(struct sock *sk,
719 struct sockaddr *addrs,
722 struct net *net = sock_net(sk);
723 struct sctp_sock *sp;
724 struct sctp_endpoint *ep;
725 struct sctp_association *asoc;
726 struct sctp_transport *transport;
727 struct sctp_bind_addr *bp;
728 struct sctp_chunk *chunk;
729 union sctp_addr *laddr;
732 struct sctp_sockaddr_entry *saddr;
738 if (!net->sctp.addip_enable)
744 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
745 __func__, sk, addrs, addrcnt);
747 list_for_each_entry(asoc, &ep->asocs, asocs) {
749 if (!asoc->peer.asconf_capable)
752 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
755 if (!sctp_state(asoc, ESTABLISHED))
758 /* Check if any address in the packed array of addresses is
759 * not present in the bind address list of the association.
760 * If so, do not send the asconf chunk to its peer, but
761 * continue with other associations.
764 for (i = 0; i < addrcnt; i++) {
766 af = sctp_get_af_specific(laddr->v4.sin_family);
772 if (!sctp_assoc_lookup_laddr(asoc, laddr))
775 addr_buf += af->sockaddr_len;
780 /* Find one address in the association's bind address list
781 * that is not in the packed array of addresses. This is to
782 * make sure that we do not delete all the addresses in the
785 bp = &asoc->base.bind_addr;
786 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
788 if ((laddr == NULL) && (addrcnt == 1)) {
789 if (asoc->asconf_addr_del_pending)
791 asoc->asconf_addr_del_pending =
792 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
793 if (asoc->asconf_addr_del_pending == NULL) {
797 asoc->asconf_addr_del_pending->sa.sa_family =
799 asoc->asconf_addr_del_pending->v4.sin_port =
801 if (addrs->sa_family == AF_INET) {
802 struct sockaddr_in *sin;
804 sin = (struct sockaddr_in *)addrs;
805 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
806 } else if (addrs->sa_family == AF_INET6) {
807 struct sockaddr_in6 *sin6;
809 sin6 = (struct sockaddr_in6 *)addrs;
810 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
812 SCTP_DEBUG_PRINTK_IPADDR("send_asconf_del_ip: keep the last address asoc: %p ",
813 " at %p\n", asoc, asoc->asconf_addr_del_pending,
814 asoc->asconf_addr_del_pending);
815 asoc->src_out_of_asoc_ok = 1;
820 /* We do not need RCU protection throughout this loop
821 * because this is done under a socket lock from the
824 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
832 /* Reset use_as_src flag for the addresses in the bind address
833 * list that are to be deleted.
836 for (i = 0; i < addrcnt; i++) {
838 af = sctp_get_af_specific(laddr->v4.sin_family);
839 list_for_each_entry(saddr, &bp->address_list, list) {
840 if (sctp_cmp_addr_exact(&saddr->a, laddr))
841 saddr->state = SCTP_ADDR_DEL;
843 addr_buf += af->sockaddr_len;
846 /* Update the route and saddr entries for all the transports
847 * as some of the addresses in the bind address list are
848 * about to be deleted and cannot be used as source addresses.
850 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
852 dst_release(transport->dst);
853 sctp_transport_route(transport, NULL,
854 sctp_sk(asoc->base.sk));
858 /* We don't need to transmit ASCONF */
860 retval = sctp_send_asconf(asoc, chunk);
866 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
867 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
869 struct sock *sk = sctp_opt2sk(sp);
870 union sctp_addr *addr;
873 /* It is safe to write port space in caller. */
875 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
876 af = sctp_get_af_specific(addr->sa.sa_family);
879 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
882 if (addrw->state == SCTP_ADDR_NEW)
883 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
885 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
888 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
891 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
894 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
895 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
898 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
899 * Section 3.1.2 for this usage.
901 * addrs is a pointer to an array of one or more socket addresses. Each
902 * address is contained in its appropriate structure (i.e. struct
903 * sockaddr_in or struct sockaddr_in6) the family of the address type
904 * must be used to distinguish the address length (note that this
905 * representation is termed a "packed array" of addresses). The caller
906 * specifies the number of addresses in the array with addrcnt.
908 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
909 * -1, and sets errno to the appropriate error code.
911 * For SCTP, the port given in each socket address must be the same, or
912 * sctp_bindx() will fail, setting errno to EINVAL.
914 * The flags parameter is formed from the bitwise OR of zero or more of
915 * the following currently defined flags:
917 * SCTP_BINDX_ADD_ADDR
919 * SCTP_BINDX_REM_ADDR
921 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
922 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
923 * addresses from the association. The two flags are mutually exclusive;
924 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
925 * not remove all addresses from an association; sctp_bindx() will
926 * reject such an attempt with EINVAL.
928 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
929 * additional addresses with an endpoint after calling bind(). Or use
930 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
931 * socket is associated with so that no new association accepted will be
932 * associated with those addresses. If the endpoint supports dynamic
933 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
934 * endpoint to send the appropriate message to the peer to change the
935 * peers address lists.
937 * Adding and removing addresses from a connected association is
938 * optional functionality. Implementations that do not support this
939 * functionality should return EOPNOTSUPP.
941 * Basically do nothing but copying the addresses from user to kernel
942 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
943 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
946 * We don't use copy_from_user() for optimization: we first do the
947 * sanity checks (buffer size -fast- and access check-healthy
948 * pointer); if all of those succeed, then we can alloc the memory
949 * (expensive operation) needed to copy the data to kernel. Then we do
950 * the copying without checking the user space area
951 * (__copy_from_user()).
953 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
956 * sk The sk of the socket
957 * addrs The pointer to the addresses in user land
958 * addrssize Size of the addrs buffer
959 * op Operation to perform (add or remove, see the flags of
962 * Returns 0 if ok, <0 errno code on error.
964 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
965 struct sockaddr __user *addrs,
966 int addrs_size, int op)
968 struct sockaddr *kaddrs;
972 struct sockaddr *sa_addr;
976 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
977 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
979 if (unlikely(addrs_size <= 0))
982 /* Check the user passed a healthy pointer. */
983 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
986 /* Alloc space for the address array in kernel memory. */
987 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
988 if (unlikely(!kaddrs))
991 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
996 /* Walk through the addrs buffer and count the number of addresses. */
998 while (walk_size < addrs_size) {
999 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1005 af = sctp_get_af_specific(sa_addr->sa_family);
1007 /* If the address family is not supported or if this address
1008 * causes the address buffer to overflow return EINVAL.
1010 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1015 addr_buf += af->sockaddr_len;
1016 walk_size += af->sockaddr_len;
1021 case SCTP_BINDX_ADD_ADDR:
1022 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1025 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1028 case SCTP_BINDX_REM_ADDR:
1029 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1032 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1046 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1048 * Common routine for handling connect() and sctp_connectx().
1049 * Connect will come in with just a single address.
1051 static int __sctp_connect(struct sock* sk,
1052 struct sockaddr *kaddrs,
1054 sctp_assoc_t *assoc_id)
1056 struct net *net = sock_net(sk);
1057 struct sctp_sock *sp;
1058 struct sctp_endpoint *ep;
1059 struct sctp_association *asoc = NULL;
1060 struct sctp_association *asoc2;
1061 struct sctp_transport *transport;
1069 union sctp_addr *sa_addr = NULL;
1071 unsigned short port;
1072 unsigned int f_flags = 0;
1077 /* connect() cannot be done on a socket that is already in ESTABLISHED
1078 * state - UDP-style peeled off socket or a TCP-style socket that
1079 * is already connected.
1080 * It cannot be done even on a TCP-style listening socket.
1082 if (sctp_sstate(sk, ESTABLISHED) ||
1083 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1088 /* Walk through the addrs buffer and count the number of addresses. */
1090 while (walk_size < addrs_size) {
1091 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1097 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1099 /* If the address family is not supported or if this address
1100 * causes the address buffer to overflow return EINVAL.
1102 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1107 port = ntohs(sa_addr->v4.sin_port);
1109 /* Save current address so we can work with it */
1110 memcpy(&to, sa_addr, af->sockaddr_len);
1112 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1116 /* Make sure the destination port is correctly set
1119 if (asoc && asoc->peer.port && asoc->peer.port != port)
1123 /* Check if there already is a matching association on the
1124 * endpoint (other than the one created here).
1126 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1127 if (asoc2 && asoc2 != asoc) {
1128 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1135 /* If we could not find a matching association on the endpoint,
1136 * make sure that there is no peeled-off association matching
1137 * the peer address even on another socket.
1139 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1140 err = -EADDRNOTAVAIL;
1145 /* If a bind() or sctp_bindx() is not called prior to
1146 * an sctp_connectx() call, the system picks an
1147 * ephemeral port and will choose an address set
1148 * equivalent to binding with a wildcard address.
1150 if (!ep->base.bind_addr.port) {
1151 if (sctp_autobind(sk)) {
1157 * If an unprivileged user inherits a 1-many
1158 * style socket with open associations on a
1159 * privileged port, it MAY be permitted to
1160 * accept new associations, but it SHOULD NOT
1161 * be permitted to open new associations.
1163 if (ep->base.bind_addr.port < PROT_SOCK &&
1164 !capable(CAP_NET_BIND_SERVICE)) {
1170 scope = sctp_scope(&to);
1171 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1177 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1185 /* Prime the peer's transport structures. */
1186 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1194 addr_buf += af->sockaddr_len;
1195 walk_size += af->sockaddr_len;
1198 /* In case the user of sctp_connectx() wants an association
1199 * id back, assign one now.
1202 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1207 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1212 /* Initialize sk's dport and daddr for getpeername() */
1213 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1214 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1215 af->to_sk_daddr(sa_addr, sk);
1218 /* in-kernel sockets don't generally have a file allocated to them
1219 * if all they do is call sock_create_kern().
1221 if (sk->sk_socket->file)
1222 f_flags = sk->sk_socket->file->f_flags;
1224 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1226 err = sctp_wait_for_connect(asoc, &timeo);
1227 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1228 *assoc_id = asoc->assoc_id;
1230 /* Don't free association on exit. */
1235 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1236 " kaddrs: %p err: %d\n",
1239 /* sctp_primitive_ASSOCIATE may have added this association
1240 * To the hash table, try to unhash it, just in case, its a noop
1241 * if it wasn't hashed so we're safe
1243 sctp_unhash_established(asoc);
1244 sctp_association_free(asoc);
1249 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1252 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1253 * sctp_assoc_t *asoc);
1255 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1256 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1257 * or IPv6 addresses.
1259 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1260 * Section 3.1.2 for this usage.
1262 * addrs is a pointer to an array of one or more socket addresses. Each
1263 * address is contained in its appropriate structure (i.e. struct
1264 * sockaddr_in or struct sockaddr_in6) the family of the address type
1265 * must be used to distengish the address length (note that this
1266 * representation is termed a "packed array" of addresses). The caller
1267 * specifies the number of addresses in the array with addrcnt.
1269 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1270 * the association id of the new association. On failure, sctp_connectx()
1271 * returns -1, and sets errno to the appropriate error code. The assoc_id
1272 * is not touched by the kernel.
1274 * For SCTP, the port given in each socket address must be the same, or
1275 * sctp_connectx() will fail, setting errno to EINVAL.
1277 * An application can use sctp_connectx to initiate an association with
1278 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1279 * allows a caller to specify multiple addresses at which a peer can be
1280 * reached. The way the SCTP stack uses the list of addresses to set up
1281 * the association is implementation dependent. This function only
1282 * specifies that the stack will try to make use of all the addresses in
1283 * the list when needed.
1285 * Note that the list of addresses passed in is only used for setting up
1286 * the association. It does not necessarily equal the set of addresses
1287 * the peer uses for the resulting association. If the caller wants to
1288 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1289 * retrieve them after the association has been set up.
1291 * Basically do nothing but copying the addresses from user to kernel
1292 * land and invoking either sctp_connectx(). This is used for tunneling
1293 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1295 * We don't use copy_from_user() for optimization: we first do the
1296 * sanity checks (buffer size -fast- and access check-healthy
1297 * pointer); if all of those succeed, then we can alloc the memory
1298 * (expensive operation) needed to copy the data to kernel. Then we do
1299 * the copying without checking the user space area
1300 * (__copy_from_user()).
1302 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1305 * sk The sk of the socket
1306 * addrs The pointer to the addresses in user land
1307 * addrssize Size of the addrs buffer
1309 * Returns >=0 if ok, <0 errno code on error.
1311 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1312 struct sockaddr __user *addrs,
1314 sctp_assoc_t *assoc_id)
1317 struct sockaddr *kaddrs;
1319 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1320 __func__, sk, addrs, addrs_size);
1322 if (unlikely(addrs_size <= 0))
1325 /* Check the user passed a healthy pointer. */
1326 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1329 /* Alloc space for the address array in kernel memory. */
1330 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1331 if (unlikely(!kaddrs))
1334 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1337 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1346 * This is an older interface. It's kept for backward compatibility
1347 * to the option that doesn't provide association id.
1349 SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1350 struct sockaddr __user *addrs,
1353 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1357 * New interface for the API. The since the API is done with a socket
1358 * option, to make it simple we feed back the association id is as a return
1359 * indication to the call. Error is always negative and association id is
1362 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1363 struct sockaddr __user *addrs,
1366 sctp_assoc_t assoc_id = 0;
1369 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1378 * New (hopefully final) interface for the API.
1379 * We use the sctp_getaddrs_old structure so that use-space library
1380 * can avoid any unnecessary allocations. The only defferent part
1381 * is that we store the actual length of the address buffer into the
1382 * addrs_num structure member. That way we can re-use the existing
1385 SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1386 char __user *optval,
1389 struct sctp_getaddrs_old param;
1390 sctp_assoc_t assoc_id = 0;
1393 if (len < sizeof(param))
1396 if (copy_from_user(¶m, optval, sizeof(param)))
1399 err = __sctp_setsockopt_connectx(sk,
1400 (struct sockaddr __user *)param.addrs,
1401 param.addr_num, &assoc_id);
1403 if (err == 0 || err == -EINPROGRESS) {
1404 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1406 if (put_user(sizeof(assoc_id), optlen))
1413 /* API 3.1.4 close() - UDP Style Syntax
1414 * Applications use close() to perform graceful shutdown (as described in
1415 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1416 * by a UDP-style socket.
1420 * ret = close(int sd);
1422 * sd - the socket descriptor of the associations to be closed.
1424 * To gracefully shutdown a specific association represented by the
1425 * UDP-style socket, an application should use the sendmsg() call,
1426 * passing no user data, but including the appropriate flag in the
1427 * ancillary data (see Section xxxx).
1429 * If sd in the close() call is a branched-off socket representing only
1430 * one association, the shutdown is performed on that association only.
1432 * 4.1.6 close() - TCP Style Syntax
1434 * Applications use close() to gracefully close down an association.
1438 * int close(int sd);
1440 * sd - the socket descriptor of the association to be closed.
1442 * After an application calls close() on a socket descriptor, no further
1443 * socket operations will succeed on that descriptor.
1445 * API 7.1.4 SO_LINGER
1447 * An application using the TCP-style socket can use this option to
1448 * perform the SCTP ABORT primitive. The linger option structure is:
1451 * int l_onoff; // option on/off
1452 * int l_linger; // linger time
1455 * To enable the option, set l_onoff to 1. If the l_linger value is set
1456 * to 0, calling close() is the same as the ABORT primitive. If the
1457 * value is set to a negative value, the setsockopt() call will return
1458 * an error. If the value is set to a positive value linger_time, the
1459 * close() can be blocked for at most linger_time ms. If the graceful
1460 * shutdown phase does not finish during this period, close() will
1461 * return but the graceful shutdown phase continues in the system.
1463 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1465 struct net *net = sock_net(sk);
1466 struct sctp_endpoint *ep;
1467 struct sctp_association *asoc;
1468 struct list_head *pos, *temp;
1469 unsigned int data_was_unread;
1471 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1474 sk->sk_shutdown = SHUTDOWN_MASK;
1475 sk->sk_state = SCTP_SS_CLOSING;
1477 ep = sctp_sk(sk)->ep;
1479 /* Clean up any skbs sitting on the receive queue. */
1480 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1481 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1483 /* Walk all associations on an endpoint. */
1484 list_for_each_safe(pos, temp, &ep->asocs) {
1485 asoc = list_entry(pos, struct sctp_association, asocs);
1487 if (sctp_style(sk, TCP)) {
1488 /* A closed association can still be in the list if
1489 * it belongs to a TCP-style listening socket that is
1490 * not yet accepted. If so, free it. If not, send an
1491 * ABORT or SHUTDOWN based on the linger options.
1493 if (sctp_state(asoc, CLOSED)) {
1494 sctp_unhash_established(asoc);
1495 sctp_association_free(asoc);
1500 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1501 !skb_queue_empty(&asoc->ulpq.reasm) ||
1502 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1503 struct sctp_chunk *chunk;
1505 chunk = sctp_make_abort_user(asoc, NULL, 0);
1507 sctp_primitive_ABORT(net, asoc, chunk);
1509 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1512 /* On a TCP-style socket, block for at most linger_time if set. */
1513 if (sctp_style(sk, TCP) && timeout)
1514 sctp_wait_for_close(sk, timeout);
1516 /* This will run the backlog queue. */
1517 sctp_release_sock(sk);
1519 /* Supposedly, no process has access to the socket, but
1520 * the net layers still may.
1522 sctp_local_bh_disable();
1523 sctp_bh_lock_sock(sk);
1525 /* Hold the sock, since sk_common_release() will put sock_put()
1526 * and we have just a little more cleanup.
1529 sk_common_release(sk);
1531 sctp_bh_unlock_sock(sk);
1532 sctp_local_bh_enable();
1536 SCTP_DBG_OBJCNT_DEC(sock);
1539 /* Handle EPIPE error. */
1540 static int sctp_error(struct sock *sk, int flags, int err)
1543 err = sock_error(sk) ? : -EPIPE;
1544 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1545 send_sig(SIGPIPE, current, 0);
1549 /* API 3.1.3 sendmsg() - UDP Style Syntax
1551 * An application uses sendmsg() and recvmsg() calls to transmit data to
1552 * and receive data from its peer.
1554 * ssize_t sendmsg(int socket, const struct msghdr *message,
1557 * socket - the socket descriptor of the endpoint.
1558 * message - pointer to the msghdr structure which contains a single
1559 * user message and possibly some ancillary data.
1561 * See Section 5 for complete description of the data
1564 * flags - flags sent or received with the user message, see Section
1565 * 5 for complete description of the flags.
1567 * Note: This function could use a rewrite especially when explicit
1568 * connect support comes in.
1570 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1572 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1574 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1575 struct msghdr *msg, size_t msg_len)
1577 struct net *net = sock_net(sk);
1578 struct sctp_sock *sp;
1579 struct sctp_endpoint *ep;
1580 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1581 struct sctp_transport *transport, *chunk_tp;
1582 struct sctp_chunk *chunk;
1584 struct sockaddr *msg_name = NULL;
1585 struct sctp_sndrcvinfo default_sinfo;
1586 struct sctp_sndrcvinfo *sinfo;
1587 struct sctp_initmsg *sinit;
1588 sctp_assoc_t associd = 0;
1589 sctp_cmsgs_t cmsgs = { NULL };
1593 __u16 sinfo_flags = 0;
1594 struct sctp_datamsg *datamsg;
1595 int msg_flags = msg->msg_flags;
1597 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1604 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1606 /* We cannot send a message over a TCP-style listening socket. */
1607 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1612 /* Parse out the SCTP CMSGs. */
1613 err = sctp_msghdr_parse(msg, &cmsgs);
1616 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1620 /* Fetch the destination address for this packet. This
1621 * address only selects the association--it is not necessarily
1622 * the address we will send to.
1623 * For a peeled-off socket, msg_name is ignored.
1625 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1626 int msg_namelen = msg->msg_namelen;
1628 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1633 if (msg_namelen > sizeof(to))
1634 msg_namelen = sizeof(to);
1635 memcpy(&to, msg->msg_name, msg_namelen);
1636 msg_name = msg->msg_name;
1642 /* Did the user specify SNDRCVINFO? */
1644 sinfo_flags = sinfo->sinfo_flags;
1645 associd = sinfo->sinfo_assoc_id;
1648 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1649 msg_len, sinfo_flags);
1651 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1652 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1657 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1658 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1659 * If SCTP_ABORT is set, the message length could be non zero with
1660 * the msg_iov set to the user abort reason.
1662 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1663 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1668 /* If SCTP_ADDR_OVER is set, there must be an address
1669 * specified in msg_name.
1671 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1678 SCTP_DEBUG_PRINTK("About to look up association.\n");
1682 /* If a msg_name has been specified, assume this is to be used. */
1684 /* Look for a matching association on the endpoint. */
1685 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1687 /* If we could not find a matching association on the
1688 * endpoint, make sure that it is not a TCP-style
1689 * socket that already has an association or there is
1690 * no peeled-off association on another socket.
1692 if ((sctp_style(sk, TCP) &&
1693 sctp_sstate(sk, ESTABLISHED)) ||
1694 sctp_endpoint_is_peeled_off(ep, &to)) {
1695 err = -EADDRNOTAVAIL;
1700 asoc = sctp_id2assoc(sk, associd);
1708 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1710 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1711 * socket that has an association in CLOSED state. This can
1712 * happen when an accepted socket has an association that is
1715 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1720 if (sinfo_flags & SCTP_EOF) {
1721 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1723 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1727 if (sinfo_flags & SCTP_ABORT) {
1729 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1735 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1736 sctp_primitive_ABORT(net, asoc, chunk);
1742 /* Do we need to create the association? */
1744 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1746 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1751 /* Check for invalid stream against the stream counts,
1752 * either the default or the user specified stream counts.
1755 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1756 /* Check against the defaults. */
1757 if (sinfo->sinfo_stream >=
1758 sp->initmsg.sinit_num_ostreams) {
1763 /* Check against the requested. */
1764 if (sinfo->sinfo_stream >=
1765 sinit->sinit_num_ostreams) {
1773 * API 3.1.2 bind() - UDP Style Syntax
1774 * If a bind() or sctp_bindx() is not called prior to a
1775 * sendmsg() call that initiates a new association, the
1776 * system picks an ephemeral port and will choose an address
1777 * set equivalent to binding with a wildcard address.
1779 if (!ep->base.bind_addr.port) {
1780 if (sctp_autobind(sk)) {
1786 * If an unprivileged user inherits a one-to-many
1787 * style socket with open associations on a privileged
1788 * port, it MAY be permitted to accept new associations,
1789 * but it SHOULD NOT be permitted to open new
1792 if (ep->base.bind_addr.port < PROT_SOCK &&
1793 !capable(CAP_NET_BIND_SERVICE)) {
1799 scope = sctp_scope(&to);
1800 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1806 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1812 /* If the SCTP_INIT ancillary data is specified, set all
1813 * the association init values accordingly.
1816 if (sinit->sinit_num_ostreams) {
1817 asoc->c.sinit_num_ostreams =
1818 sinit->sinit_num_ostreams;
1820 if (sinit->sinit_max_instreams) {
1821 asoc->c.sinit_max_instreams =
1822 sinit->sinit_max_instreams;
1824 if (sinit->sinit_max_attempts) {
1825 asoc->max_init_attempts
1826 = sinit->sinit_max_attempts;
1828 if (sinit->sinit_max_init_timeo) {
1829 asoc->max_init_timeo =
1830 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1834 /* Prime the peer's transport structures. */
1835 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1842 /* ASSERT: we have a valid association at this point. */
1843 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1846 /* If the user didn't specify SNDRCVINFO, make up one with
1849 memset(&default_sinfo, 0, sizeof(default_sinfo));
1850 default_sinfo.sinfo_stream = asoc->default_stream;
1851 default_sinfo.sinfo_flags = asoc->default_flags;
1852 default_sinfo.sinfo_ppid = asoc->default_ppid;
1853 default_sinfo.sinfo_context = asoc->default_context;
1854 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1855 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1856 sinfo = &default_sinfo;
1859 /* API 7.1.7, the sndbuf size per association bounds the
1860 * maximum size of data that can be sent in a single send call.
1862 if (msg_len > sk->sk_sndbuf) {
1867 if (asoc->pmtu_pending)
1868 sctp_assoc_pending_pmtu(sk, asoc);
1870 /* If fragmentation is disabled and the message length exceeds the
1871 * association fragmentation point, return EMSGSIZE. The I-D
1872 * does not specify what this error is, but this looks like
1875 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1880 /* Check for invalid stream. */
1881 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1886 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1887 if (!sctp_wspace(asoc)) {
1888 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1893 /* If an address is passed with the sendto/sendmsg call, it is used
1894 * to override the primary destination address in the TCP model, or
1895 * when SCTP_ADDR_OVER flag is set in the UDP model.
1897 if ((sctp_style(sk, TCP) && msg_name) ||
1898 (sinfo_flags & SCTP_ADDR_OVER)) {
1899 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1907 /* Auto-connect, if we aren't connected already. */
1908 if (sctp_state(asoc, CLOSED)) {
1909 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1912 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1915 /* Break the message into multiple chunks of maximum size. */
1916 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1922 /* Now send the (possibly) fragmented message. */
1923 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1924 sctp_chunk_hold(chunk);
1926 /* Do accounting for the write space. */
1927 sctp_set_owner_w(chunk);
1929 chunk->transport = chunk_tp;
1932 /* Send it to the lower layers. Note: all chunks
1933 * must either fail or succeed. The lower layer
1934 * works that way today. Keep it that way or this
1937 err = sctp_primitive_SEND(net, asoc, datamsg);
1938 /* Did the lower layer accept the chunk? */
1940 sctp_datamsg_free(datamsg);
1942 sctp_datamsg_put(datamsg);
1944 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1951 /* If we are already past ASSOCIATE, the lower
1952 * layers are responsible for association cleanup.
1958 sctp_unhash_established(asoc);
1959 sctp_association_free(asoc);
1962 sctp_release_sock(sk);
1965 return sctp_error(sk, msg_flags, err);
1972 err = sock_error(sk);
1982 /* This is an extended version of skb_pull() that removes the data from the
1983 * start of a skb even when data is spread across the list of skb's in the
1984 * frag_list. len specifies the total amount of data that needs to be removed.
1985 * when 'len' bytes could be removed from the skb, it returns 0.
1986 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1987 * could not be removed.
1989 static int sctp_skb_pull(struct sk_buff *skb, int len)
1991 struct sk_buff *list;
1992 int skb_len = skb_headlen(skb);
1995 if (len <= skb_len) {
1996 __skb_pull(skb, len);
2000 __skb_pull(skb, skb_len);
2002 skb_walk_frags(skb, list) {
2003 rlen = sctp_skb_pull(list, len);
2004 skb->len -= (len-rlen);
2005 skb->data_len -= (len-rlen);
2016 /* API 3.1.3 recvmsg() - UDP Style Syntax
2018 * ssize_t recvmsg(int socket, struct msghdr *message,
2021 * socket - the socket descriptor of the endpoint.
2022 * message - pointer to the msghdr structure which contains a single
2023 * user message and possibly some ancillary data.
2025 * See Section 5 for complete description of the data
2028 * flags - flags sent or received with the user message, see Section
2029 * 5 for complete description of the flags.
2031 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2033 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2034 struct msghdr *msg, size_t len, int noblock,
2035 int flags, int *addr_len)
2037 struct sctp_ulpevent *event = NULL;
2038 struct sctp_sock *sp = sctp_sk(sk);
2039 struct sk_buff *skb;
2044 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
2045 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
2046 "len", len, "knoblauch", noblock,
2047 "flags", flags, "addr_len", addr_len);
2051 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2056 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2060 /* Get the total length of the skb including any skb's in the
2069 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2071 event = sctp_skb2event(skb);
2076 sock_recv_ts_and_drops(msg, sk, skb);
2077 if (sctp_ulpevent_is_notification(event)) {
2078 msg->msg_flags |= MSG_NOTIFICATION;
2079 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2081 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2084 /* Check if we allow SCTP_SNDRCVINFO. */
2085 if (sp->subscribe.sctp_data_io_event)
2086 sctp_ulpevent_read_sndrcvinfo(event, msg);
2088 /* FIXME: we should be calling IP/IPv6 layers. */
2089 if (sk->sk_protinfo.af_inet.cmsg_flags)
2090 ip_cmsg_recv(msg, skb);
2095 /* If skb's length exceeds the user's buffer, update the skb and
2096 * push it back to the receive_queue so that the next call to
2097 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2099 if (skb_len > copied) {
2100 msg->msg_flags &= ~MSG_EOR;
2101 if (flags & MSG_PEEK)
2103 sctp_skb_pull(skb, copied);
2104 skb_queue_head(&sk->sk_receive_queue, skb);
2106 /* When only partial message is copied to the user, increase
2107 * rwnd by that amount. If all the data in the skb is read,
2108 * rwnd is updated when the event is freed.
2110 if (!sctp_ulpevent_is_notification(event))
2111 sctp_assoc_rwnd_increase(event->asoc, copied);
2113 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2114 (event->msg_flags & MSG_EOR))
2115 msg->msg_flags |= MSG_EOR;
2117 msg->msg_flags &= ~MSG_EOR;
2120 if (flags & MSG_PEEK) {
2121 /* Release the skb reference acquired after peeking the skb in
2122 * sctp_skb_recv_datagram().
2126 /* Free the event which includes releasing the reference to
2127 * the owner of the skb, freeing the skb and updating the
2130 sctp_ulpevent_free(event);
2133 sctp_release_sock(sk);
2137 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2139 * This option is a on/off flag. If enabled no SCTP message
2140 * fragmentation will be performed. Instead if a message being sent
2141 * exceeds the current PMTU size, the message will NOT be sent and
2142 * instead a error will be indicated to the user.
2144 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2145 char __user *optval,
2146 unsigned int optlen)
2150 if (optlen < sizeof(int))
2153 if (get_user(val, (int __user *)optval))
2156 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2161 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2162 unsigned int optlen)
2164 struct sctp_association *asoc;
2165 struct sctp_ulpevent *event;
2167 if (optlen > sizeof(struct sctp_event_subscribe))
2169 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2173 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2174 * if there is no data to be sent or retransmit, the stack will
2175 * immediately send up this notification.
2177 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2178 &sctp_sk(sk)->subscribe)) {
2179 asoc = sctp_id2assoc(sk, 0);
2181 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2182 event = sctp_ulpevent_make_sender_dry_event(asoc,
2187 sctp_ulpq_tail_event(&asoc->ulpq, event);
2194 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2196 * This socket option is applicable to the UDP-style socket only. When
2197 * set it will cause associations that are idle for more than the
2198 * specified number of seconds to automatically close. An association
2199 * being idle is defined an association that has NOT sent or received
2200 * user data. The special value of '0' indicates that no automatic
2201 * close of any associations should be performed. The option expects an
2202 * integer defining the number of seconds of idle time before an
2203 * association is closed.
2205 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2206 unsigned int optlen)
2208 struct sctp_sock *sp = sctp_sk(sk);
2210 /* Applicable to UDP-style socket only */
2211 if (sctp_style(sk, TCP))
2213 if (optlen != sizeof(int))
2215 if (copy_from_user(&sp->autoclose, optval, optlen))
2221 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2223 * Applications can enable or disable heartbeats for any peer address of
2224 * an association, modify an address's heartbeat interval, force a
2225 * heartbeat to be sent immediately, and adjust the address's maximum
2226 * number of retransmissions sent before an address is considered
2227 * unreachable. The following structure is used to access and modify an
2228 * address's parameters:
2230 * struct sctp_paddrparams {
2231 * sctp_assoc_t spp_assoc_id;
2232 * struct sockaddr_storage spp_address;
2233 * uint32_t spp_hbinterval;
2234 * uint16_t spp_pathmaxrxt;
2235 * uint32_t spp_pathmtu;
2236 * uint32_t spp_sackdelay;
2237 * uint32_t spp_flags;
2240 * spp_assoc_id - (one-to-many style socket) This is filled in the
2241 * application, and identifies the association for
2243 * spp_address - This specifies which address is of interest.
2244 * spp_hbinterval - This contains the value of the heartbeat interval,
2245 * in milliseconds. If a value of zero
2246 * is present in this field then no changes are to
2247 * be made to this parameter.
2248 * spp_pathmaxrxt - This contains the maximum number of
2249 * retransmissions before this address shall be
2250 * considered unreachable. If a value of zero
2251 * is present in this field then no changes are to
2252 * be made to this parameter.
2253 * spp_pathmtu - When Path MTU discovery is disabled the value
2254 * specified here will be the "fixed" path mtu.
2255 * Note that if the spp_address field is empty
2256 * then all associations on this address will
2257 * have this fixed path mtu set upon them.
2259 * spp_sackdelay - When delayed sack is enabled, this value specifies
2260 * the number of milliseconds that sacks will be delayed
2261 * for. This value will apply to all addresses of an
2262 * association if the spp_address field is empty. Note
2263 * also, that if delayed sack is enabled and this
2264 * value is set to 0, no change is made to the last
2265 * recorded delayed sack timer value.
2267 * spp_flags - These flags are used to control various features
2268 * on an association. The flag field may contain
2269 * zero or more of the following options.
2271 * SPP_HB_ENABLE - Enable heartbeats on the
2272 * specified address. Note that if the address
2273 * field is empty all addresses for the association
2274 * have heartbeats enabled upon them.
2276 * SPP_HB_DISABLE - Disable heartbeats on the
2277 * speicifed address. Note that if the address
2278 * field is empty all addresses for the association
2279 * will have their heartbeats disabled. Note also
2280 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2281 * mutually exclusive, only one of these two should
2282 * be specified. Enabling both fields will have
2283 * undetermined results.
2285 * SPP_HB_DEMAND - Request a user initiated heartbeat
2286 * to be made immediately.
2288 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2289 * heartbeat delayis to be set to the value of 0
2292 * SPP_PMTUD_ENABLE - This field will enable PMTU
2293 * discovery upon the specified address. Note that
2294 * if the address feild is empty then all addresses
2295 * on the association are effected.
2297 * SPP_PMTUD_DISABLE - This field will disable PMTU
2298 * discovery upon the specified address. Note that
2299 * if the address feild is empty then all addresses
2300 * on the association are effected. Not also that
2301 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2302 * exclusive. Enabling both will have undetermined
2305 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2306 * on delayed sack. The time specified in spp_sackdelay
2307 * is used to specify the sack delay for this address. Note
2308 * that if spp_address is empty then all addresses will
2309 * enable delayed sack and take on the sack delay
2310 * value specified in spp_sackdelay.
2311 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2312 * off delayed sack. If the spp_address field is blank then
2313 * delayed sack is disabled for the entire association. Note
2314 * also that this field is mutually exclusive to
2315 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2318 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2319 struct sctp_transport *trans,
2320 struct sctp_association *asoc,
2321 struct sctp_sock *sp,
2324 int sackdelay_change)
2328 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2329 struct net *net = sock_net(trans->asoc->base.sk);
2331 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2336 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2337 * this field is ignored. Note also that a value of zero indicates
2338 * the current setting should be left unchanged.
2340 if (params->spp_flags & SPP_HB_ENABLE) {
2342 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2343 * set. This lets us use 0 value when this flag
2346 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2347 params->spp_hbinterval = 0;
2349 if (params->spp_hbinterval ||
2350 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2353 msecs_to_jiffies(params->spp_hbinterval);
2356 msecs_to_jiffies(params->spp_hbinterval);
2358 sp->hbinterval = params->spp_hbinterval;
2365 trans->param_flags =
2366 (trans->param_flags & ~SPP_HB) | hb_change;
2369 (asoc->param_flags & ~SPP_HB) | hb_change;
2372 (sp->param_flags & ~SPP_HB) | hb_change;
2376 /* When Path MTU discovery is disabled the value specified here will
2377 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2378 * include the flag SPP_PMTUD_DISABLE for this field to have any
2381 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2383 trans->pathmtu = params->spp_pathmtu;
2384 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2386 asoc->pathmtu = params->spp_pathmtu;
2387 sctp_frag_point(asoc, params->spp_pathmtu);
2389 sp->pathmtu = params->spp_pathmtu;
2395 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2396 (params->spp_flags & SPP_PMTUD_ENABLE);
2397 trans->param_flags =
2398 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2400 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2401 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2405 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2408 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2412 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2413 * value of this field is ignored. Note also that a value of zero
2414 * indicates the current setting should be left unchanged.
2416 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2419 msecs_to_jiffies(params->spp_sackdelay);
2422 msecs_to_jiffies(params->spp_sackdelay);
2424 sp->sackdelay = params->spp_sackdelay;
2428 if (sackdelay_change) {
2430 trans->param_flags =
2431 (trans->param_flags & ~SPP_SACKDELAY) |
2435 (asoc->param_flags & ~SPP_SACKDELAY) |
2439 (sp->param_flags & ~SPP_SACKDELAY) |
2444 /* Note that a value of zero indicates the current setting should be
2447 if (params->spp_pathmaxrxt) {
2449 trans->pathmaxrxt = params->spp_pathmaxrxt;
2451 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2453 sp->pathmaxrxt = params->spp_pathmaxrxt;
2460 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2461 char __user *optval,
2462 unsigned int optlen)
2464 struct sctp_paddrparams params;
2465 struct sctp_transport *trans = NULL;
2466 struct sctp_association *asoc = NULL;
2467 struct sctp_sock *sp = sctp_sk(sk);
2469 int hb_change, pmtud_change, sackdelay_change;
2471 if (optlen != sizeof(struct sctp_paddrparams))
2474 if (copy_from_user(¶ms, optval, optlen))
2477 /* Validate flags and value parameters. */
2478 hb_change = params.spp_flags & SPP_HB;
2479 pmtud_change = params.spp_flags & SPP_PMTUD;
2480 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2482 if (hb_change == SPP_HB ||
2483 pmtud_change == SPP_PMTUD ||
2484 sackdelay_change == SPP_SACKDELAY ||
2485 params.spp_sackdelay > 500 ||
2486 (params.spp_pathmtu &&
2487 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2490 /* If an address other than INADDR_ANY is specified, and
2491 * no transport is found, then the request is invalid.
2493 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
2494 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2495 params.spp_assoc_id);
2500 /* Get association, if assoc_id != 0 and the socket is a one
2501 * to many style socket, and an association was not found, then
2502 * the id was invalid.
2504 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2505 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2508 /* Heartbeat demand can only be sent on a transport or
2509 * association, but not a socket.
2511 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2514 /* Process parameters. */
2515 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2516 hb_change, pmtud_change,
2522 /* If changes are for association, also apply parameters to each
2525 if (!trans && asoc) {
2526 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2528 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2529 hb_change, pmtud_change,
2538 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2540 * This option will effect the way delayed acks are performed. This
2541 * option allows you to get or set the delayed ack time, in
2542 * milliseconds. It also allows changing the delayed ack frequency.
2543 * Changing the frequency to 1 disables the delayed sack algorithm. If
2544 * the assoc_id is 0, then this sets or gets the endpoints default
2545 * values. If the assoc_id field is non-zero, then the set or get
2546 * effects the specified association for the one to many model (the
2547 * assoc_id field is ignored by the one to one model). Note that if
2548 * sack_delay or sack_freq are 0 when setting this option, then the
2549 * current values will remain unchanged.
2551 * struct sctp_sack_info {
2552 * sctp_assoc_t sack_assoc_id;
2553 * uint32_t sack_delay;
2554 * uint32_t sack_freq;
2557 * sack_assoc_id - This parameter, indicates which association the user
2558 * is performing an action upon. Note that if this field's value is
2559 * zero then the endpoints default value is changed (effecting future
2560 * associations only).
2562 * sack_delay - This parameter contains the number of milliseconds that
2563 * the user is requesting the delayed ACK timer be set to. Note that
2564 * this value is defined in the standard to be between 200 and 500
2567 * sack_freq - This parameter contains the number of packets that must
2568 * be received before a sack is sent without waiting for the delay
2569 * timer to expire. The default value for this is 2, setting this
2570 * value to 1 will disable the delayed sack algorithm.
2573 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2574 char __user *optval, unsigned int optlen)
2576 struct sctp_sack_info params;
2577 struct sctp_transport *trans = NULL;
2578 struct sctp_association *asoc = NULL;
2579 struct sctp_sock *sp = sctp_sk(sk);
2581 if (optlen == sizeof(struct sctp_sack_info)) {
2582 if (copy_from_user(¶ms, optval, optlen))
2585 if (params.sack_delay == 0 && params.sack_freq == 0)
2587 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2588 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2589 pr_warn("Use struct sctp_sack_info instead\n");
2590 if (copy_from_user(¶ms, optval, optlen))
2593 if (params.sack_delay == 0)
2594 params.sack_freq = 1;
2596 params.sack_freq = 0;
2600 /* Validate value parameter. */
2601 if (params.sack_delay > 500)
2604 /* Get association, if sack_assoc_id != 0 and the socket is a one
2605 * to many style socket, and an association was not found, then
2606 * the id was invalid.
2608 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2609 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2612 if (params.sack_delay) {
2615 msecs_to_jiffies(params.sack_delay);
2617 (asoc->param_flags & ~SPP_SACKDELAY) |
2618 SPP_SACKDELAY_ENABLE;
2620 sp->sackdelay = params.sack_delay;
2622 (sp->param_flags & ~SPP_SACKDELAY) |
2623 SPP_SACKDELAY_ENABLE;
2627 if (params.sack_freq == 1) {
2630 (asoc->param_flags & ~SPP_SACKDELAY) |
2631 SPP_SACKDELAY_DISABLE;
2634 (sp->param_flags & ~SPP_SACKDELAY) |
2635 SPP_SACKDELAY_DISABLE;
2637 } else if (params.sack_freq > 1) {
2639 asoc->sackfreq = params.sack_freq;
2641 (asoc->param_flags & ~SPP_SACKDELAY) |
2642 SPP_SACKDELAY_ENABLE;
2644 sp->sackfreq = params.sack_freq;
2646 (sp->param_flags & ~SPP_SACKDELAY) |
2647 SPP_SACKDELAY_ENABLE;
2651 /* If change is for association, also apply to each transport. */
2653 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2655 if (params.sack_delay) {
2657 msecs_to_jiffies(params.sack_delay);
2658 trans->param_flags =
2659 (trans->param_flags & ~SPP_SACKDELAY) |
2660 SPP_SACKDELAY_ENABLE;
2662 if (params.sack_freq == 1) {
2663 trans->param_flags =
2664 (trans->param_flags & ~SPP_SACKDELAY) |
2665 SPP_SACKDELAY_DISABLE;
2666 } else if (params.sack_freq > 1) {
2667 trans->sackfreq = params.sack_freq;
2668 trans->param_flags =
2669 (trans->param_flags & ~SPP_SACKDELAY) |
2670 SPP_SACKDELAY_ENABLE;
2678 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2680 * Applications can specify protocol parameters for the default association
2681 * initialization. The option name argument to setsockopt() and getsockopt()
2684 * Setting initialization parameters is effective only on an unconnected
2685 * socket (for UDP-style sockets only future associations are effected
2686 * by the change). With TCP-style sockets, this option is inherited by
2687 * sockets derived from a listener socket.
2689 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2691 struct sctp_initmsg sinit;
2692 struct sctp_sock *sp = sctp_sk(sk);
2694 if (optlen != sizeof(struct sctp_initmsg))
2696 if (copy_from_user(&sinit, optval, optlen))
2699 if (sinit.sinit_num_ostreams)
2700 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2701 if (sinit.sinit_max_instreams)
2702 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2703 if (sinit.sinit_max_attempts)
2704 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2705 if (sinit.sinit_max_init_timeo)
2706 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2712 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2714 * Applications that wish to use the sendto() system call may wish to
2715 * specify a default set of parameters that would normally be supplied
2716 * through the inclusion of ancillary data. This socket option allows
2717 * such an application to set the default sctp_sndrcvinfo structure.
2718 * The application that wishes to use this socket option simply passes
2719 * in to this call the sctp_sndrcvinfo structure defined in Section
2720 * 5.2.2) The input parameters accepted by this call include
2721 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2722 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2723 * to this call if the caller is using the UDP model.
2725 static int sctp_setsockopt_default_send_param(struct sock *sk,
2726 char __user *optval,
2727 unsigned int optlen)
2729 struct sctp_sndrcvinfo info;
2730 struct sctp_association *asoc;
2731 struct sctp_sock *sp = sctp_sk(sk);
2733 if (optlen != sizeof(struct sctp_sndrcvinfo))
2735 if (copy_from_user(&info, optval, optlen))
2738 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2739 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2743 asoc->default_stream = info.sinfo_stream;
2744 asoc->default_flags = info.sinfo_flags;
2745 asoc->default_ppid = info.sinfo_ppid;
2746 asoc->default_context = info.sinfo_context;
2747 asoc->default_timetolive = info.sinfo_timetolive;
2749 sp->default_stream = info.sinfo_stream;
2750 sp->default_flags = info.sinfo_flags;
2751 sp->default_ppid = info.sinfo_ppid;
2752 sp->default_context = info.sinfo_context;
2753 sp->default_timetolive = info.sinfo_timetolive;
2759 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2761 * Requests that the local SCTP stack use the enclosed peer address as
2762 * the association primary. The enclosed address must be one of the
2763 * association peer's addresses.
2765 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2766 unsigned int optlen)
2768 struct sctp_prim prim;
2769 struct sctp_transport *trans;
2771 if (optlen != sizeof(struct sctp_prim))
2774 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2777 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2781 sctp_assoc_set_primary(trans->asoc, trans);
2787 * 7.1.5 SCTP_NODELAY
2789 * Turn on/off any Nagle-like algorithm. This means that packets are
2790 * generally sent as soon as possible and no unnecessary delays are
2791 * introduced, at the cost of more packets in the network. Expects an
2792 * integer boolean flag.
2794 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2795 unsigned int optlen)
2799 if (optlen < sizeof(int))
2801 if (get_user(val, (int __user *)optval))
2804 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2810 * 7.1.1 SCTP_RTOINFO
2812 * The protocol parameters used to initialize and bound retransmission
2813 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2814 * and modify these parameters.
2815 * All parameters are time values, in milliseconds. A value of 0, when
2816 * modifying the parameters, indicates that the current value should not
2820 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2822 struct sctp_rtoinfo rtoinfo;
2823 struct sctp_association *asoc;
2825 if (optlen != sizeof (struct sctp_rtoinfo))
2828 if (copy_from_user(&rtoinfo, optval, optlen))
2831 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2833 /* Set the values to the specific association */
2834 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2838 if (rtoinfo.srto_initial != 0)
2840 msecs_to_jiffies(rtoinfo.srto_initial);
2841 if (rtoinfo.srto_max != 0)
2842 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2843 if (rtoinfo.srto_min != 0)
2844 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2846 /* If there is no association or the association-id = 0
2847 * set the values to the endpoint.
2849 struct sctp_sock *sp = sctp_sk(sk);
2851 if (rtoinfo.srto_initial != 0)
2852 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2853 if (rtoinfo.srto_max != 0)
2854 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2855 if (rtoinfo.srto_min != 0)
2856 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2864 * 7.1.2 SCTP_ASSOCINFO
2866 * This option is used to tune the maximum retransmission attempts
2867 * of the association.
2868 * Returns an error if the new association retransmission value is
2869 * greater than the sum of the retransmission value of the peer.
2870 * See [SCTP] for more information.
2873 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2876 struct sctp_assocparams assocparams;
2877 struct sctp_association *asoc;
2879 if (optlen != sizeof(struct sctp_assocparams))
2881 if (copy_from_user(&assocparams, optval, optlen))
2884 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2886 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2889 /* Set the values to the specific association */
2891 if (assocparams.sasoc_asocmaxrxt != 0) {
2894 struct sctp_transport *peer_addr;
2896 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2898 path_sum += peer_addr->pathmaxrxt;
2902 /* Only validate asocmaxrxt if we have more than
2903 * one path/transport. We do this because path
2904 * retransmissions are only counted when we have more
2908 assocparams.sasoc_asocmaxrxt > path_sum)
2911 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2914 if (assocparams.sasoc_cookie_life != 0) {
2915 asoc->cookie_life.tv_sec =
2916 assocparams.sasoc_cookie_life / 1000;
2917 asoc->cookie_life.tv_usec =
2918 (assocparams.sasoc_cookie_life % 1000)
2922 /* Set the values to the endpoint */
2923 struct sctp_sock *sp = sctp_sk(sk);
2925 if (assocparams.sasoc_asocmaxrxt != 0)
2926 sp->assocparams.sasoc_asocmaxrxt =
2927 assocparams.sasoc_asocmaxrxt;
2928 if (assocparams.sasoc_cookie_life != 0)
2929 sp->assocparams.sasoc_cookie_life =
2930 assocparams.sasoc_cookie_life;
2936 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2938 * This socket option is a boolean flag which turns on or off mapped V4
2939 * addresses. If this option is turned on and the socket is type
2940 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2941 * If this option is turned off, then no mapping will be done of V4
2942 * addresses and a user will receive both PF_INET6 and PF_INET type
2943 * addresses on the socket.
2945 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2948 struct sctp_sock *sp = sctp_sk(sk);
2950 if (optlen < sizeof(int))
2952 if (get_user(val, (int __user *)optval))
2963 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2964 * This option will get or set the maximum size to put in any outgoing
2965 * SCTP DATA chunk. If a message is larger than this size it will be
2966 * fragmented by SCTP into the specified size. Note that the underlying
2967 * SCTP implementation may fragment into smaller sized chunks when the
2968 * PMTU of the underlying association is smaller than the value set by
2969 * the user. The default value for this option is '0' which indicates
2970 * the user is NOT limiting fragmentation and only the PMTU will effect
2971 * SCTP's choice of DATA chunk size. Note also that values set larger
2972 * than the maximum size of an IP datagram will effectively let SCTP
2973 * control fragmentation (i.e. the same as setting this option to 0).
2975 * The following structure is used to access and modify this parameter:
2977 * struct sctp_assoc_value {
2978 * sctp_assoc_t assoc_id;
2979 * uint32_t assoc_value;
2982 * assoc_id: This parameter is ignored for one-to-one style sockets.
2983 * For one-to-many style sockets this parameter indicates which
2984 * association the user is performing an action upon. Note that if
2985 * this field's value is zero then the endpoints default value is
2986 * changed (effecting future associations only).
2987 * assoc_value: This parameter specifies the maximum size in bytes.
2989 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2991 struct sctp_assoc_value params;
2992 struct sctp_association *asoc;
2993 struct sctp_sock *sp = sctp_sk(sk);
2996 if (optlen == sizeof(int)) {
2997 pr_warn("Use of int in maxseg socket option deprecated\n");
2998 pr_warn("Use struct sctp_assoc_value instead\n");
2999 if (copy_from_user(&val, optval, optlen))
3001 params.assoc_id = 0;
3002 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3003 if (copy_from_user(¶ms, optval, optlen))
3005 val = params.assoc_value;
3009 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3012 asoc = sctp_id2assoc(sk, params.assoc_id);
3013 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3018 val = asoc->pathmtu;
3019 val -= sp->pf->af->net_header_len;
3020 val -= sizeof(struct sctphdr) +
3021 sizeof(struct sctp_data_chunk);
3023 asoc->user_frag = val;
3024 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3026 sp->user_frag = val;
3034 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3036 * Requests that the peer mark the enclosed address as the association
3037 * primary. The enclosed address must be one of the association's
3038 * locally bound addresses. The following structure is used to make a
3039 * set primary request:
3041 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3042 unsigned int optlen)
3044 struct net *net = sock_net(sk);
3045 struct sctp_sock *sp;
3046 struct sctp_association *asoc = NULL;
3047 struct sctp_setpeerprim prim;
3048 struct sctp_chunk *chunk;
3054 if (!net->sctp.addip_enable)
3057 if (optlen != sizeof(struct sctp_setpeerprim))
3060 if (copy_from_user(&prim, optval, optlen))
3063 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3067 if (!asoc->peer.asconf_capable)
3070 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3073 if (!sctp_state(asoc, ESTABLISHED))
3076 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3080 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3081 return -EADDRNOTAVAIL;
3083 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3084 return -EADDRNOTAVAIL;
3086 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3087 chunk = sctp_make_asconf_set_prim(asoc,
3088 (union sctp_addr *)&prim.sspp_addr);
3092 err = sctp_send_asconf(asoc, chunk);
3094 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
3099 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3100 unsigned int optlen)
3102 struct sctp_setadaptation adaptation;
3104 if (optlen != sizeof(struct sctp_setadaptation))
3106 if (copy_from_user(&adaptation, optval, optlen))
3109 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3115 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3117 * The context field in the sctp_sndrcvinfo structure is normally only
3118 * used when a failed message is retrieved holding the value that was
3119 * sent down on the actual send call. This option allows the setting of
3120 * a default context on an association basis that will be received on
3121 * reading messages from the peer. This is especially helpful in the
3122 * one-2-many model for an application to keep some reference to an
3123 * internal state machine that is processing messages on the
3124 * association. Note that the setting of this value only effects
3125 * received messages from the peer and does not effect the value that is
3126 * saved with outbound messages.
3128 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3129 unsigned int optlen)
3131 struct sctp_assoc_value params;
3132 struct sctp_sock *sp;
3133 struct sctp_association *asoc;
3135 if (optlen != sizeof(struct sctp_assoc_value))
3137 if (copy_from_user(¶ms, optval, optlen))
3142 if (params.assoc_id != 0) {
3143 asoc = sctp_id2assoc(sk, params.assoc_id);
3146 asoc->default_rcv_context = params.assoc_value;
3148 sp->default_rcv_context = params.assoc_value;
3155 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3157 * This options will at a minimum specify if the implementation is doing
3158 * fragmented interleave. Fragmented interleave, for a one to many
3159 * socket, is when subsequent calls to receive a message may return
3160 * parts of messages from different associations. Some implementations
3161 * may allow you to turn this value on or off. If so, when turned off,
3162 * no fragment interleave will occur (which will cause a head of line
3163 * blocking amongst multiple associations sharing the same one to many
3164 * socket). When this option is turned on, then each receive call may
3165 * come from a different association (thus the user must receive data
3166 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3167 * association each receive belongs to.
3169 * This option takes a boolean value. A non-zero value indicates that
3170 * fragmented interleave is on. A value of zero indicates that
3171 * fragmented interleave is off.
3173 * Note that it is important that an implementation that allows this
3174 * option to be turned on, have it off by default. Otherwise an unaware
3175 * application using the one to many model may become confused and act
3178 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3179 char __user *optval,
3180 unsigned int optlen)
3184 if (optlen != sizeof(int))
3186 if (get_user(val, (int __user *)optval))
3189 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3195 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3196 * (SCTP_PARTIAL_DELIVERY_POINT)
3198 * This option will set or get the SCTP partial delivery point. This
3199 * point is the size of a message where the partial delivery API will be
3200 * invoked to help free up rwnd space for the peer. Setting this to a
3201 * lower value will cause partial deliveries to happen more often. The
3202 * calls argument is an integer that sets or gets the partial delivery
3203 * point. Note also that the call will fail if the user attempts to set
3204 * this value larger than the socket receive buffer size.
3206 * Note that any single message having a length smaller than or equal to
3207 * the SCTP partial delivery point will be delivered in one single read
3208 * call as long as the user provided buffer is large enough to hold the
3211 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3212 char __user *optval,
3213 unsigned int optlen)
3217 if (optlen != sizeof(u32))
3219 if (get_user(val, (int __user *)optval))
3222 /* Note: We double the receive buffer from what the user sets
3223 * it to be, also initial rwnd is based on rcvbuf/2.
3225 if (val > (sk->sk_rcvbuf >> 1))
3228 sctp_sk(sk)->pd_point = val;
3230 return 0; /* is this the right error code? */
3234 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3236 * This option will allow a user to change the maximum burst of packets
3237 * that can be emitted by this association. Note that the default value
3238 * is 4, and some implementations may restrict this setting so that it
3239 * can only be lowered.
3241 * NOTE: This text doesn't seem right. Do this on a socket basis with
3242 * future associations inheriting the socket value.
3244 static int sctp_setsockopt_maxburst(struct sock *sk,
3245 char __user *optval,
3246 unsigned int optlen)
3248 struct sctp_assoc_value params;
3249 struct sctp_sock *sp;
3250 struct sctp_association *asoc;
3254 if (optlen == sizeof(int)) {
3255 pr_warn("Use of int in max_burst socket option deprecated\n");
3256 pr_warn("Use struct sctp_assoc_value instead\n");
3257 if (copy_from_user(&val, optval, optlen))
3259 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3260 if (copy_from_user(¶ms, optval, optlen))
3262 val = params.assoc_value;
3263 assoc_id = params.assoc_id;
3269 if (assoc_id != 0) {
3270 asoc = sctp_id2assoc(sk, assoc_id);
3273 asoc->max_burst = val;
3275 sp->max_burst = val;
3281 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3283 * This set option adds a chunk type that the user is requesting to be
3284 * received only in an authenticated way. Changes to the list of chunks
3285 * will only effect future associations on the socket.
3287 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3288 char __user *optval,
3289 unsigned int optlen)
3291 struct net *net = sock_net(sk);
3292 struct sctp_authchunk val;
3294 if (!net->sctp.auth_enable)
3297 if (optlen != sizeof(struct sctp_authchunk))
3299 if (copy_from_user(&val, optval, optlen))
3302 switch (val.sauth_chunk) {
3304 case SCTP_CID_INIT_ACK:
3305 case SCTP_CID_SHUTDOWN_COMPLETE:
3310 /* add this chunk id to the endpoint */
3311 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3315 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3317 * This option gets or sets the list of HMAC algorithms that the local
3318 * endpoint requires the peer to use.
3320 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3321 char __user *optval,
3322 unsigned int optlen)
3324 struct net *net = sock_net(sk);
3325 struct sctp_hmacalgo *hmacs;
3329 if (!net->sctp.auth_enable)
3332 if (optlen < sizeof(struct sctp_hmacalgo))
3335 hmacs= memdup_user(optval, optlen);
3337 return PTR_ERR(hmacs);
3339 idents = hmacs->shmac_num_idents;
3340 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3341 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3346 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3353 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3355 * This option will set a shared secret key which is used to build an
3356 * association shared key.
3358 static int sctp_setsockopt_auth_key(struct sock *sk,
3359 char __user *optval,
3360 unsigned int optlen)
3362 struct net *net = sock_net(sk);
3363 struct sctp_authkey *authkey;
3364 struct sctp_association *asoc;
3367 if (!net->sctp.auth_enable)
3370 if (optlen <= sizeof(struct sctp_authkey))
3373 authkey= memdup_user(optval, optlen);
3374 if (IS_ERR(authkey))
3375 return PTR_ERR(authkey);
3377 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3382 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3383 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3388 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3395 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3397 * This option will get or set the active shared key to be used to build
3398 * the association shared key.
3400 static int sctp_setsockopt_active_key(struct sock *sk,
3401 char __user *optval,
3402 unsigned int optlen)
3404 struct net *net = sock_net(sk);
3405 struct sctp_authkeyid val;
3406 struct sctp_association *asoc;
3408 if (!net->sctp.auth_enable)
3411 if (optlen != sizeof(struct sctp_authkeyid))
3413 if (copy_from_user(&val, optval, optlen))
3416 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3417 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3420 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3421 val.scact_keynumber);
3425 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3427 * This set option will delete a shared secret key from use.
3429 static int sctp_setsockopt_del_key(struct sock *sk,
3430 char __user *optval,
3431 unsigned int optlen)
3433 struct net *net = sock_net(sk);
3434 struct sctp_authkeyid val;
3435 struct sctp_association *asoc;
3437 if (!net->sctp.auth_enable)
3440 if (optlen != sizeof(struct sctp_authkeyid))
3442 if (copy_from_user(&val, optval, optlen))
3445 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3446 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3449 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3450 val.scact_keynumber);
3455 * 8.1.23 SCTP_AUTO_ASCONF
3457 * This option will enable or disable the use of the automatic generation of
3458 * ASCONF chunks to add and delete addresses to an existing association. Note
3459 * that this option has two caveats namely: a) it only affects sockets that
3460 * are bound to all addresses available to the SCTP stack, and b) the system
3461 * administrator may have an overriding control that turns the ASCONF feature
3462 * off no matter what setting the socket option may have.
3463 * This option expects an integer boolean flag, where a non-zero value turns on
3464 * the option, and a zero value turns off the option.
3465 * Note. In this implementation, socket operation overrides default parameter
3466 * being set by sysctl as well as FreeBSD implementation
3468 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3469 unsigned int optlen)
3472 struct sctp_sock *sp = sctp_sk(sk);
3474 if (optlen < sizeof(int))
3476 if (get_user(val, (int __user *)optval))
3478 if (!sctp_is_ep_boundall(sk) && val)
3480 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3483 if (val == 0 && sp->do_auto_asconf) {
3484 list_del(&sp->auto_asconf_list);
3485 sp->do_auto_asconf = 0;
3486 } else if (val && !sp->do_auto_asconf) {
3487 list_add_tail(&sp->auto_asconf_list,
3488 &sock_net(sk)->sctp.auto_asconf_splist);
3489 sp->do_auto_asconf = 1;
3496 * SCTP_PEER_ADDR_THLDS
3498 * This option allows us to alter the partially failed threshold for one or all
3499 * transports in an association. See Section 6.1 of:
3500 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3502 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3503 char __user *optval,
3504 unsigned int optlen)
3506 struct sctp_paddrthlds val;
3507 struct sctp_transport *trans;
3508 struct sctp_association *asoc;
3510 if (optlen < sizeof(struct sctp_paddrthlds))
3512 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3513 sizeof(struct sctp_paddrthlds)))
3517 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3518 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3521 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3523 if (val.spt_pathmaxrxt)
3524 trans->pathmaxrxt = val.spt_pathmaxrxt;
3525 trans->pf_retrans = val.spt_pathpfthld;
3528 if (val.spt_pathmaxrxt)
3529 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3530 asoc->pf_retrans = val.spt_pathpfthld;
3532 trans = sctp_addr_id2transport(sk, &val.spt_address,
3537 if (val.spt_pathmaxrxt)
3538 trans->pathmaxrxt = val.spt_pathmaxrxt;
3539 trans->pf_retrans = val.spt_pathpfthld;
3545 /* API 6.2 setsockopt(), getsockopt()
3547 * Applications use setsockopt() and getsockopt() to set or retrieve
3548 * socket options. Socket options are used to change the default
3549 * behavior of sockets calls. They are described in Section 7.
3553 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3554 * int __user *optlen);
3555 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3558 * sd - the socket descript.
3559 * level - set to IPPROTO_SCTP for all SCTP options.
3560 * optname - the option name.
3561 * optval - the buffer to store the value of the option.
3562 * optlen - the size of the buffer.
3564 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3565 char __user *optval, unsigned int optlen)
3569 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3572 /* I can hardly begin to describe how wrong this is. This is
3573 * so broken as to be worse than useless. The API draft
3574 * REALLY is NOT helpful here... I am not convinced that the
3575 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3576 * are at all well-founded.
3578 if (level != SOL_SCTP) {
3579 struct sctp_af *af = sctp_sk(sk)->pf->af;
3580 retval = af->setsockopt(sk, level, optname, optval, optlen);
3587 case SCTP_SOCKOPT_BINDX_ADD:
3588 /* 'optlen' is the size of the addresses buffer. */
3589 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3590 optlen, SCTP_BINDX_ADD_ADDR);
3593 case SCTP_SOCKOPT_BINDX_REM:
3594 /* 'optlen' is the size of the addresses buffer. */
3595 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3596 optlen, SCTP_BINDX_REM_ADDR);
3599 case SCTP_SOCKOPT_CONNECTX_OLD:
3600 /* 'optlen' is the size of the addresses buffer. */
3601 retval = sctp_setsockopt_connectx_old(sk,
3602 (struct sockaddr __user *)optval,
3606 case SCTP_SOCKOPT_CONNECTX:
3607 /* 'optlen' is the size of the addresses buffer. */
3608 retval = sctp_setsockopt_connectx(sk,
3609 (struct sockaddr __user *)optval,
3613 case SCTP_DISABLE_FRAGMENTS:
3614 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3618 retval = sctp_setsockopt_events(sk, optval, optlen);
3621 case SCTP_AUTOCLOSE:
3622 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3625 case SCTP_PEER_ADDR_PARAMS:
3626 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3629 case SCTP_DELAYED_SACK:
3630 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3632 case SCTP_PARTIAL_DELIVERY_POINT:
3633 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3637 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3639 case SCTP_DEFAULT_SEND_PARAM:
3640 retval = sctp_setsockopt_default_send_param(sk, optval,
3643 case SCTP_PRIMARY_ADDR:
3644 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3646 case SCTP_SET_PEER_PRIMARY_ADDR:
3647 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3650 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3653 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3655 case SCTP_ASSOCINFO:
3656 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3658 case SCTP_I_WANT_MAPPED_V4_ADDR:
3659 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3662 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3664 case SCTP_ADAPTATION_LAYER:
3665 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3668 retval = sctp_setsockopt_context(sk, optval, optlen);
3670 case SCTP_FRAGMENT_INTERLEAVE:
3671 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3673 case SCTP_MAX_BURST:
3674 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3676 case SCTP_AUTH_CHUNK:
3677 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3679 case SCTP_HMAC_IDENT:
3680 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3683 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3685 case SCTP_AUTH_ACTIVE_KEY:
3686 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3688 case SCTP_AUTH_DELETE_KEY:
3689 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3691 case SCTP_AUTO_ASCONF:
3692 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3694 case SCTP_PEER_ADDR_THLDS:
3695 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3698 retval = -ENOPROTOOPT;
3702 sctp_release_sock(sk);
3708 /* API 3.1.6 connect() - UDP Style Syntax
3710 * An application may use the connect() call in the UDP model to initiate an
3711 * association without sending data.
3715 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3717 * sd: the socket descriptor to have a new association added to.
3719 * nam: the address structure (either struct sockaddr_in or struct
3720 * sockaddr_in6 defined in RFC2553 [7]).
3722 * len: the size of the address.
3724 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3732 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3733 __func__, sk, addr, addr_len);
3735 /* Validate addr_len before calling common connect/connectx routine. */
3736 af = sctp_get_af_specific(addr->sa_family);
3737 if (!af || addr_len < af->sockaddr_len) {
3740 /* Pass correct addr len to common routine (so it knows there
3741 * is only one address being passed.
3743 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3746 sctp_release_sock(sk);
3750 /* FIXME: Write comments. */
3751 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3753 return -EOPNOTSUPP; /* STUB */
3756 /* 4.1.4 accept() - TCP Style Syntax
3758 * Applications use accept() call to remove an established SCTP
3759 * association from the accept queue of the endpoint. A new socket
3760 * descriptor will be returned from accept() to represent the newly
3761 * formed association.
3763 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3765 struct sctp_sock *sp;
3766 struct sctp_endpoint *ep;
3767 struct sock *newsk = NULL;
3768 struct sctp_association *asoc;
3777 if (!sctp_style(sk, TCP)) {
3778 error = -EOPNOTSUPP;
3782 if (!sctp_sstate(sk, LISTENING)) {
3787 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3789 error = sctp_wait_for_accept(sk, timeo);
3793 /* We treat the list of associations on the endpoint as the accept
3794 * queue and pick the first association on the list.
3796 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3798 newsk = sp->pf->create_accept_sk(sk, asoc);
3804 /* Populate the fields of the newsk from the oldsk and migrate the
3805 * asoc to the newsk.
3807 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3810 sctp_release_sock(sk);
3815 /* The SCTP ioctl handler. */
3816 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3823 * SEQPACKET-style sockets in LISTENING state are valid, for
3824 * SCTP, so only discard TCP-style sockets in LISTENING state.
3826 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3831 struct sk_buff *skb;
3832 unsigned int amount = 0;
3834 skb = skb_peek(&sk->sk_receive_queue);
3837 * We will only return the amount of this packet since
3838 * that is all that will be read.
3842 rc = put_user(amount, (int __user *)arg);
3850 sctp_release_sock(sk);
3854 /* This is the function which gets called during socket creation to
3855 * initialized the SCTP-specific portion of the sock.
3856 * The sock structure should already be zero-filled memory.
3858 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3860 struct net *net = sock_net(sk);
3861 struct sctp_endpoint *ep;
3862 struct sctp_sock *sp;
3864 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3868 /* Initialize the SCTP per socket area. */
3869 switch (sk->sk_type) {
3870 case SOCK_SEQPACKET:
3871 sp->type = SCTP_SOCKET_UDP;
3874 sp->type = SCTP_SOCKET_TCP;
3877 return -ESOCKTNOSUPPORT;
3880 /* Initialize default send parameters. These parameters can be
3881 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3883 sp->default_stream = 0;
3884 sp->default_ppid = 0;
3885 sp->default_flags = 0;
3886 sp->default_context = 0;
3887 sp->default_timetolive = 0;
3889 sp->default_rcv_context = 0;
3890 sp->max_burst = net->sctp.max_burst;
3892 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
3894 /* Initialize default setup parameters. These parameters
3895 * can be modified with the SCTP_INITMSG socket option or
3896 * overridden by the SCTP_INIT CMSG.
3898 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3899 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3900 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
3901 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
3903 /* Initialize default RTO related parameters. These parameters can
3904 * be modified for with the SCTP_RTOINFO socket option.
3906 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
3907 sp->rtoinfo.srto_max = net->sctp.rto_max;
3908 sp->rtoinfo.srto_min = net->sctp.rto_min;
3910 /* Initialize default association related parameters. These parameters
3911 * can be modified with the SCTP_ASSOCINFO socket option.
3913 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
3914 sp->assocparams.sasoc_number_peer_destinations = 0;
3915 sp->assocparams.sasoc_peer_rwnd = 0;
3916 sp->assocparams.sasoc_local_rwnd = 0;
3917 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
3919 /* Initialize default event subscriptions. By default, all the
3922 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3924 /* Default Peer Address Parameters. These defaults can
3925 * be modified via SCTP_PEER_ADDR_PARAMS
3927 sp->hbinterval = net->sctp.hb_interval;
3928 sp->pathmaxrxt = net->sctp.max_retrans_path;
3929 sp->pathmtu = 0; // allow default discovery
3930 sp->sackdelay = net->sctp.sack_timeout;
3932 sp->param_flags = SPP_HB_ENABLE |
3934 SPP_SACKDELAY_ENABLE;
3936 /* If enabled no SCTP message fragmentation will be performed.
3937 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3939 sp->disable_fragments = 0;
3941 /* Enable Nagle algorithm by default. */
3944 /* Enable by default. */
3947 /* Auto-close idle associations after the configured
3948 * number of seconds. A value of 0 disables this
3949 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3950 * for UDP-style sockets only.
3954 /* User specified fragmentation limit. */
3957 sp->adaptation_ind = 0;
3959 sp->pf = sctp_get_pf_specific(sk->sk_family);
3961 /* Control variables for partial data delivery. */
3962 atomic_set(&sp->pd_mode, 0);
3963 skb_queue_head_init(&sp->pd_lobby);
3964 sp->frag_interleave = 0;
3966 /* Create a per socket endpoint structure. Even if we
3967 * change the data structure relationships, this may still
3968 * be useful for storing pre-connect address information.
3970 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3977 SCTP_DBG_OBJCNT_INC(sock);
3980 percpu_counter_inc(&sctp_sockets_allocated);
3981 sock_prot_inuse_add(net, sk->sk_prot, 1);
3982 if (net->sctp.default_auto_asconf) {
3983 list_add_tail(&sp->auto_asconf_list,
3984 &net->sctp.auto_asconf_splist);
3985 sp->do_auto_asconf = 1;
3987 sp->do_auto_asconf = 0;
3993 /* Cleanup any SCTP per socket resources. */
3994 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3996 struct sctp_sock *sp;
3998 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
4000 /* Release our hold on the endpoint. */
4002 if (sp->do_auto_asconf) {
4003 sp->do_auto_asconf = 0;
4004 list_del(&sp->auto_asconf_list);
4006 sctp_endpoint_free(sp->ep);
4008 percpu_counter_dec(&sctp_sockets_allocated);
4009 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4013 /* API 4.1.7 shutdown() - TCP Style Syntax
4014 * int shutdown(int socket, int how);
4016 * sd - the socket descriptor of the association to be closed.
4017 * how - Specifies the type of shutdown. The values are
4020 * Disables further receive operations. No SCTP
4021 * protocol action is taken.
4023 * Disables further send operations, and initiates
4024 * the SCTP shutdown sequence.
4026 * Disables further send and receive operations
4027 * and initiates the SCTP shutdown sequence.
4029 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
4031 struct net *net = sock_net(sk);
4032 struct sctp_endpoint *ep;
4033 struct sctp_association *asoc;
4035 if (!sctp_style(sk, TCP))
4038 if (how & SEND_SHUTDOWN) {
4039 ep = sctp_sk(sk)->ep;
4040 if (!list_empty(&ep->asocs)) {
4041 asoc = list_entry(ep->asocs.next,
4042 struct sctp_association, asocs);
4043 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4048 /* 7.2.1 Association Status (SCTP_STATUS)
4050 * Applications can retrieve current status information about an
4051 * association, including association state, peer receiver window size,
4052 * number of unacked data chunks, and number of data chunks pending
4053 * receipt. This information is read-only.
4055 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4056 char __user *optval,
4059 struct sctp_status status;
4060 struct sctp_association *asoc = NULL;
4061 struct sctp_transport *transport;
4062 sctp_assoc_t associd;
4065 if (len < sizeof(status)) {
4070 len = sizeof(status);
4071 if (copy_from_user(&status, optval, len)) {
4076 associd = status.sstat_assoc_id;
4077 asoc = sctp_id2assoc(sk, associd);
4083 transport = asoc->peer.primary_path;
4085 status.sstat_assoc_id = sctp_assoc2id(asoc);
4086 status.sstat_state = asoc->state;
4087 status.sstat_rwnd = asoc->peer.rwnd;
4088 status.sstat_unackdata = asoc->unack_data;
4090 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4091 status.sstat_instrms = asoc->c.sinit_max_instreams;
4092 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4093 status.sstat_fragmentation_point = asoc->frag_point;
4094 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4095 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4096 transport->af_specific->sockaddr_len);
4097 /* Map ipv4 address into v4-mapped-on-v6 address. */
4098 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4099 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4100 status.sstat_primary.spinfo_state = transport->state;
4101 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4102 status.sstat_primary.spinfo_srtt = transport->srtt;
4103 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4104 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4106 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4107 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4109 if (put_user(len, optlen)) {
4114 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
4115 len, status.sstat_state, status.sstat_rwnd,
4116 status.sstat_assoc_id);
4118 if (copy_to_user(optval, &status, len)) {
4128 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4130 * Applications can retrieve information about a specific peer address
4131 * of an association, including its reachability state, congestion
4132 * window, and retransmission timer values. This information is
4135 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4136 char __user *optval,
4139 struct sctp_paddrinfo pinfo;
4140 struct sctp_transport *transport;
4143 if (len < sizeof(pinfo)) {
4148 len = sizeof(pinfo);
4149 if (copy_from_user(&pinfo, optval, len)) {
4154 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4155 pinfo.spinfo_assoc_id);
4159 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4160 pinfo.spinfo_state = transport->state;
4161 pinfo.spinfo_cwnd = transport->cwnd;
4162 pinfo.spinfo_srtt = transport->srtt;
4163 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4164 pinfo.spinfo_mtu = transport->pathmtu;
4166 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4167 pinfo.spinfo_state = SCTP_ACTIVE;
4169 if (put_user(len, optlen)) {
4174 if (copy_to_user(optval, &pinfo, len)) {
4183 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4185 * This option is a on/off flag. If enabled no SCTP message
4186 * fragmentation will be performed. Instead if a message being sent
4187 * exceeds the current PMTU size, the message will NOT be sent and
4188 * instead a error will be indicated to the user.
4190 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4191 char __user *optval, int __user *optlen)
4195 if (len < sizeof(int))
4199 val = (sctp_sk(sk)->disable_fragments == 1);
4200 if (put_user(len, optlen))
4202 if (copy_to_user(optval, &val, len))
4207 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4209 * This socket option is used to specify various notifications and
4210 * ancillary data the user wishes to receive.
4212 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4217 if (len > sizeof(struct sctp_event_subscribe))
4218 len = sizeof(struct sctp_event_subscribe);
4219 if (put_user(len, optlen))
4221 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4226 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4228 * This socket option is applicable to the UDP-style socket only. When
4229 * set it will cause associations that are idle for more than the
4230 * specified number of seconds to automatically close. An association
4231 * being idle is defined an association that has NOT sent or received
4232 * user data. The special value of '0' indicates that no automatic
4233 * close of any associations should be performed. The option expects an
4234 * integer defining the number of seconds of idle time before an
4235 * association is closed.
4237 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4239 /* Applicable to UDP-style socket only */
4240 if (sctp_style(sk, TCP))
4242 if (len < sizeof(int))
4245 if (put_user(len, optlen))
4247 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4252 /* Helper routine to branch off an association to a new socket. */
4253 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4255 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4256 struct socket *sock;
4263 /* An association cannot be branched off from an already peeled-off
4264 * socket, nor is this supported for tcp style sockets.
4266 if (!sctp_style(sk, UDP))
4269 /* Create a new socket. */
4270 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4274 sctp_copy_sock(sock->sk, sk, asoc);
4276 /* Make peeled-off sockets more like 1-1 accepted sockets.
4277 * Set the daddr and initialize id to something more random
4279 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4280 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4282 /* Populate the fields of the newsk from the oldsk and migrate the
4283 * asoc to the newsk.
4285 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4291 EXPORT_SYMBOL(sctp_do_peeloff);
4293 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4295 sctp_peeloff_arg_t peeloff;
4296 struct socket *newsock;
4297 struct file *newfile;
4300 if (len < sizeof(sctp_peeloff_arg_t))
4302 len = sizeof(sctp_peeloff_arg_t);
4303 if (copy_from_user(&peeloff, optval, len))
4306 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4310 /* Map the socket to an unused fd that can be returned to the user. */
4311 retval = get_unused_fd();
4313 sock_release(newsock);
4317 newfile = sock_alloc_file(newsock, 0, NULL);
4318 if (unlikely(IS_ERR(newfile))) {
4319 put_unused_fd(retval);
4320 sock_release(newsock);
4321 return PTR_ERR(newfile);
4324 SCTP_DEBUG_PRINTK("%s: sk: %p newsk: %p sd: %d\n",
4325 __func__, sk, newsock->sk, retval);
4327 /* Return the fd mapped to the new socket. */
4328 if (put_user(len, optlen)) {
4330 put_unused_fd(retval);
4333 peeloff.sd = retval;
4334 if (copy_to_user(optval, &peeloff, len)) {
4336 put_unused_fd(retval);
4339 fd_install(retval, newfile);
4344 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4346 * Applications can enable or disable heartbeats for any peer address of
4347 * an association, modify an address's heartbeat interval, force a
4348 * heartbeat to be sent immediately, and adjust the address's maximum
4349 * number of retransmissions sent before an address is considered
4350 * unreachable. The following structure is used to access and modify an
4351 * address's parameters:
4353 * struct sctp_paddrparams {
4354 * sctp_assoc_t spp_assoc_id;
4355 * struct sockaddr_storage spp_address;
4356 * uint32_t spp_hbinterval;
4357 * uint16_t spp_pathmaxrxt;
4358 * uint32_t spp_pathmtu;
4359 * uint32_t spp_sackdelay;
4360 * uint32_t spp_flags;
4363 * spp_assoc_id - (one-to-many style socket) This is filled in the
4364 * application, and identifies the association for
4366 * spp_address - This specifies which address is of interest.
4367 * spp_hbinterval - This contains the value of the heartbeat interval,
4368 * in milliseconds. If a value of zero
4369 * is present in this field then no changes are to
4370 * be made to this parameter.
4371 * spp_pathmaxrxt - This contains the maximum number of
4372 * retransmissions before this address shall be
4373 * considered unreachable. If a value of zero
4374 * is present in this field then no changes are to
4375 * be made to this parameter.
4376 * spp_pathmtu - When Path MTU discovery is disabled the value
4377 * specified here will be the "fixed" path mtu.
4378 * Note that if the spp_address field is empty
4379 * then all associations on this address will
4380 * have this fixed path mtu set upon them.
4382 * spp_sackdelay - When delayed sack is enabled, this value specifies
4383 * the number of milliseconds that sacks will be delayed
4384 * for. This value will apply to all addresses of an
4385 * association if the spp_address field is empty. Note
4386 * also, that if delayed sack is enabled and this
4387 * value is set to 0, no change is made to the last
4388 * recorded delayed sack timer value.
4390 * spp_flags - These flags are used to control various features
4391 * on an association. The flag field may contain
4392 * zero or more of the following options.
4394 * SPP_HB_ENABLE - Enable heartbeats on the
4395 * specified address. Note that if the address
4396 * field is empty all addresses for the association
4397 * have heartbeats enabled upon them.
4399 * SPP_HB_DISABLE - Disable heartbeats on the
4400 * speicifed address. Note that if the address
4401 * field is empty all addresses for the association
4402 * will have their heartbeats disabled. Note also
4403 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4404 * mutually exclusive, only one of these two should
4405 * be specified. Enabling both fields will have
4406 * undetermined results.
4408 * SPP_HB_DEMAND - Request a user initiated heartbeat
4409 * to be made immediately.
4411 * SPP_PMTUD_ENABLE - This field will enable PMTU
4412 * discovery upon the specified address. Note that
4413 * if the address feild is empty then all addresses
4414 * on the association are effected.
4416 * SPP_PMTUD_DISABLE - This field will disable PMTU
4417 * discovery upon the specified address. Note that
4418 * if the address feild is empty then all addresses
4419 * on the association are effected. Not also that
4420 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4421 * exclusive. Enabling both will have undetermined
4424 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4425 * on delayed sack. The time specified in spp_sackdelay
4426 * is used to specify the sack delay for this address. Note
4427 * that if spp_address is empty then all addresses will
4428 * enable delayed sack and take on the sack delay
4429 * value specified in spp_sackdelay.
4430 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4431 * off delayed sack. If the spp_address field is blank then
4432 * delayed sack is disabled for the entire association. Note
4433 * also that this field is mutually exclusive to
4434 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4437 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4438 char __user *optval, int __user *optlen)
4440 struct sctp_paddrparams params;
4441 struct sctp_transport *trans = NULL;
4442 struct sctp_association *asoc = NULL;
4443 struct sctp_sock *sp = sctp_sk(sk);
4445 if (len < sizeof(struct sctp_paddrparams))
4447 len = sizeof(struct sctp_paddrparams);
4448 if (copy_from_user(¶ms, optval, len))
4451 /* If an address other than INADDR_ANY is specified, and
4452 * no transport is found, then the request is invalid.
4454 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
4455 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4456 params.spp_assoc_id);
4458 SCTP_DEBUG_PRINTK("Failed no transport\n");
4463 /* Get association, if assoc_id != 0 and the socket is a one
4464 * to many style socket, and an association was not found, then
4465 * the id was invalid.
4467 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4468 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4469 SCTP_DEBUG_PRINTK("Failed no association\n");
4474 /* Fetch transport values. */
4475 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4476 params.spp_pathmtu = trans->pathmtu;
4477 params.spp_pathmaxrxt = trans->pathmaxrxt;
4478 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4480 /*draft-11 doesn't say what to return in spp_flags*/
4481 params.spp_flags = trans->param_flags;
4483 /* Fetch association values. */
4484 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4485 params.spp_pathmtu = asoc->pathmtu;
4486 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4487 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4489 /*draft-11 doesn't say what to return in spp_flags*/
4490 params.spp_flags = asoc->param_flags;
4492 /* Fetch socket values. */
4493 params.spp_hbinterval = sp->hbinterval;
4494 params.spp_pathmtu = sp->pathmtu;
4495 params.spp_sackdelay = sp->sackdelay;
4496 params.spp_pathmaxrxt = sp->pathmaxrxt;
4498 /*draft-11 doesn't say what to return in spp_flags*/
4499 params.spp_flags = sp->param_flags;
4502 if (copy_to_user(optval, ¶ms, len))
4505 if (put_user(len, optlen))
4512 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4514 * This option will effect the way delayed acks are performed. This
4515 * option allows you to get or set the delayed ack time, in
4516 * milliseconds. It also allows changing the delayed ack frequency.
4517 * Changing the frequency to 1 disables the delayed sack algorithm. If
4518 * the assoc_id is 0, then this sets or gets the endpoints default
4519 * values. If the assoc_id field is non-zero, then the set or get
4520 * effects the specified association for the one to many model (the
4521 * assoc_id field is ignored by the one to one model). Note that if
4522 * sack_delay or sack_freq are 0 when setting this option, then the
4523 * current values will remain unchanged.
4525 * struct sctp_sack_info {
4526 * sctp_assoc_t sack_assoc_id;
4527 * uint32_t sack_delay;
4528 * uint32_t sack_freq;
4531 * sack_assoc_id - This parameter, indicates which association the user
4532 * is performing an action upon. Note that if this field's value is
4533 * zero then the endpoints default value is changed (effecting future
4534 * associations only).
4536 * sack_delay - This parameter contains the number of milliseconds that
4537 * the user is requesting the delayed ACK timer be set to. Note that
4538 * this value is defined in the standard to be between 200 and 500
4541 * sack_freq - This parameter contains the number of packets that must
4542 * be received before a sack is sent without waiting for the delay
4543 * timer to expire. The default value for this is 2, setting this
4544 * value to 1 will disable the delayed sack algorithm.
4546 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4547 char __user *optval,
4550 struct sctp_sack_info params;
4551 struct sctp_association *asoc = NULL;
4552 struct sctp_sock *sp = sctp_sk(sk);
4554 if (len >= sizeof(struct sctp_sack_info)) {
4555 len = sizeof(struct sctp_sack_info);
4557 if (copy_from_user(¶ms, optval, len))
4559 } else if (len == sizeof(struct sctp_assoc_value)) {
4560 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4561 pr_warn("Use struct sctp_sack_info instead\n");
4562 if (copy_from_user(¶ms, optval, len))
4567 /* Get association, if sack_assoc_id != 0 and the socket is a one
4568 * to many style socket, and an association was not found, then
4569 * the id was invalid.
4571 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4572 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4576 /* Fetch association values. */
4577 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4578 params.sack_delay = jiffies_to_msecs(
4580 params.sack_freq = asoc->sackfreq;
4583 params.sack_delay = 0;
4584 params.sack_freq = 1;
4587 /* Fetch socket values. */
4588 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4589 params.sack_delay = sp->sackdelay;
4590 params.sack_freq = sp->sackfreq;
4592 params.sack_delay = 0;
4593 params.sack_freq = 1;
4597 if (copy_to_user(optval, ¶ms, len))
4600 if (put_user(len, optlen))
4606 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4608 * Applications can specify protocol parameters for the default association
4609 * initialization. The option name argument to setsockopt() and getsockopt()
4612 * Setting initialization parameters is effective only on an unconnected
4613 * socket (for UDP-style sockets only future associations are effected
4614 * by the change). With TCP-style sockets, this option is inherited by
4615 * sockets derived from a listener socket.
4617 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4619 if (len < sizeof(struct sctp_initmsg))
4621 len = sizeof(struct sctp_initmsg);
4622 if (put_user(len, optlen))
4624 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4630 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4631 char __user *optval, int __user *optlen)
4633 struct sctp_association *asoc;
4635 struct sctp_getaddrs getaddrs;
4636 struct sctp_transport *from;
4638 union sctp_addr temp;
4639 struct sctp_sock *sp = sctp_sk(sk);
4644 if (len < sizeof(struct sctp_getaddrs))
4647 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4650 /* For UDP-style sockets, id specifies the association to query. */
4651 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4655 to = optval + offsetof(struct sctp_getaddrs,addrs);
4656 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4658 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4660 memcpy(&temp, &from->ipaddr, sizeof(temp));
4661 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4662 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4663 if (space_left < addrlen)
4665 if (copy_to_user(to, &temp, addrlen))
4669 space_left -= addrlen;
4672 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4674 bytes_copied = ((char __user *)to) - optval;
4675 if (put_user(bytes_copied, optlen))
4681 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4682 size_t space_left, int *bytes_copied)
4684 struct sctp_sockaddr_entry *addr;
4685 union sctp_addr temp;
4688 struct net *net = sock_net(sk);
4691 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4695 if ((PF_INET == sk->sk_family) &&
4696 (AF_INET6 == addr->a.sa.sa_family))
4698 if ((PF_INET6 == sk->sk_family) &&
4699 inet_v6_ipv6only(sk) &&
4700 (AF_INET == addr->a.sa.sa_family))
4702 memcpy(&temp, &addr->a, sizeof(temp));
4703 if (!temp.v4.sin_port)
4704 temp.v4.sin_port = htons(port);
4706 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4708 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4709 if (space_left < addrlen) {
4713 memcpy(to, &temp, addrlen);
4717 space_left -= addrlen;
4718 *bytes_copied += addrlen;
4726 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4727 char __user *optval, int __user *optlen)
4729 struct sctp_bind_addr *bp;
4730 struct sctp_association *asoc;
4732 struct sctp_getaddrs getaddrs;
4733 struct sctp_sockaddr_entry *addr;
4735 union sctp_addr temp;
4736 struct sctp_sock *sp = sctp_sk(sk);
4740 int bytes_copied = 0;
4744 if (len < sizeof(struct sctp_getaddrs))
4747 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4751 * For UDP-style sockets, id specifies the association to query.
4752 * If the id field is set to the value '0' then the locally bound
4753 * addresses are returned without regard to any particular
4756 if (0 == getaddrs.assoc_id) {
4757 bp = &sctp_sk(sk)->ep->base.bind_addr;
4759 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4762 bp = &asoc->base.bind_addr;
4765 to = optval + offsetof(struct sctp_getaddrs,addrs);
4766 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4768 addrs = kmalloc(space_left, GFP_KERNEL);
4772 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4773 * addresses from the global local address list.
4775 if (sctp_list_single_entry(&bp->address_list)) {
4776 addr = list_entry(bp->address_list.next,
4777 struct sctp_sockaddr_entry, list);
4778 if (sctp_is_any(sk, &addr->a)) {
4779 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4780 space_left, &bytes_copied);
4790 /* Protection on the bound address list is not needed since
4791 * in the socket option context we hold a socket lock and
4792 * thus the bound address list can't change.
4794 list_for_each_entry(addr, &bp->address_list, list) {
4795 memcpy(&temp, &addr->a, sizeof(temp));
4796 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4797 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4798 if (space_left < addrlen) {
4799 err = -ENOMEM; /*fixme: right error?*/
4802 memcpy(buf, &temp, addrlen);
4804 bytes_copied += addrlen;
4806 space_left -= addrlen;
4810 if (copy_to_user(to, addrs, bytes_copied)) {
4814 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4818 if (put_user(bytes_copied, optlen))
4825 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4827 * Requests that the local SCTP stack use the enclosed peer address as
4828 * the association primary. The enclosed address must be one of the
4829 * association peer's addresses.
4831 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4832 char __user *optval, int __user *optlen)
4834 struct sctp_prim prim;
4835 struct sctp_association *asoc;
4836 struct sctp_sock *sp = sctp_sk(sk);
4838 if (len < sizeof(struct sctp_prim))
4841 len = sizeof(struct sctp_prim);
4843 if (copy_from_user(&prim, optval, len))
4846 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4850 if (!asoc->peer.primary_path)
4853 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4854 asoc->peer.primary_path->af_specific->sockaddr_len);
4856 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4857 (union sctp_addr *)&prim.ssp_addr);
4859 if (put_user(len, optlen))
4861 if (copy_to_user(optval, &prim, len))
4868 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4870 * Requests that the local endpoint set the specified Adaptation Layer
4871 * Indication parameter for all future INIT and INIT-ACK exchanges.
4873 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4874 char __user *optval, int __user *optlen)
4876 struct sctp_setadaptation adaptation;
4878 if (len < sizeof(struct sctp_setadaptation))
4881 len = sizeof(struct sctp_setadaptation);
4883 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4885 if (put_user(len, optlen))
4887 if (copy_to_user(optval, &adaptation, len))
4895 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4897 * Applications that wish to use the sendto() system call may wish to
4898 * specify a default set of parameters that would normally be supplied
4899 * through the inclusion of ancillary data. This socket option allows
4900 * such an application to set the default sctp_sndrcvinfo structure.
4903 * The application that wishes to use this socket option simply passes
4904 * in to this call the sctp_sndrcvinfo structure defined in Section
4905 * 5.2.2) The input parameters accepted by this call include
4906 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4907 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4908 * to this call if the caller is using the UDP model.
4910 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4912 static int sctp_getsockopt_default_send_param(struct sock *sk,
4913 int len, char __user *optval,
4916 struct sctp_sndrcvinfo info;
4917 struct sctp_association *asoc;
4918 struct sctp_sock *sp = sctp_sk(sk);
4920 if (len < sizeof(struct sctp_sndrcvinfo))
4923 len = sizeof(struct sctp_sndrcvinfo);
4925 if (copy_from_user(&info, optval, len))
4928 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4929 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4933 info.sinfo_stream = asoc->default_stream;
4934 info.sinfo_flags = asoc->default_flags;
4935 info.sinfo_ppid = asoc->default_ppid;
4936 info.sinfo_context = asoc->default_context;
4937 info.sinfo_timetolive = asoc->default_timetolive;
4939 info.sinfo_stream = sp->default_stream;
4940 info.sinfo_flags = sp->default_flags;
4941 info.sinfo_ppid = sp->default_ppid;
4942 info.sinfo_context = sp->default_context;
4943 info.sinfo_timetolive = sp->default_timetolive;
4946 if (put_user(len, optlen))
4948 if (copy_to_user(optval, &info, len))
4956 * 7.1.5 SCTP_NODELAY
4958 * Turn on/off any Nagle-like algorithm. This means that packets are
4959 * generally sent as soon as possible and no unnecessary delays are
4960 * introduced, at the cost of more packets in the network. Expects an
4961 * integer boolean flag.
4964 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4965 char __user *optval, int __user *optlen)
4969 if (len < sizeof(int))
4973 val = (sctp_sk(sk)->nodelay == 1);
4974 if (put_user(len, optlen))
4976 if (copy_to_user(optval, &val, len))
4983 * 7.1.1 SCTP_RTOINFO
4985 * The protocol parameters used to initialize and bound retransmission
4986 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4987 * and modify these parameters.
4988 * All parameters are time values, in milliseconds. A value of 0, when
4989 * modifying the parameters, indicates that the current value should not
4993 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4994 char __user *optval,
4995 int __user *optlen) {
4996 struct sctp_rtoinfo rtoinfo;
4997 struct sctp_association *asoc;
4999 if (len < sizeof (struct sctp_rtoinfo))
5002 len = sizeof(struct sctp_rtoinfo);
5004 if (copy_from_user(&rtoinfo, optval, len))
5007 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5009 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5012 /* Values corresponding to the specific association. */
5014 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5015 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5016 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5018 /* Values corresponding to the endpoint. */
5019 struct sctp_sock *sp = sctp_sk(sk);
5021 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5022 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5023 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5026 if (put_user(len, optlen))
5029 if (copy_to_user(optval, &rtoinfo, len))
5037 * 7.1.2 SCTP_ASSOCINFO
5039 * This option is used to tune the maximum retransmission attempts
5040 * of the association.
5041 * Returns an error if the new association retransmission value is
5042 * greater than the sum of the retransmission value of the peer.
5043 * See [SCTP] for more information.
5046 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5047 char __user *optval,
5051 struct sctp_assocparams assocparams;
5052 struct sctp_association *asoc;
5053 struct list_head *pos;
5056 if (len < sizeof (struct sctp_assocparams))
5059 len = sizeof(struct sctp_assocparams);
5061 if (copy_from_user(&assocparams, optval, len))
5064 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5066 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5069 /* Values correspoinding to the specific association */
5071 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5072 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5073 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5074 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
5076 (asoc->cookie_life.tv_usec
5079 list_for_each(pos, &asoc->peer.transport_addr_list) {
5083 assocparams.sasoc_number_peer_destinations = cnt;
5085 /* Values corresponding to the endpoint */
5086 struct sctp_sock *sp = sctp_sk(sk);
5088 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5089 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5090 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5091 assocparams.sasoc_cookie_life =
5092 sp->assocparams.sasoc_cookie_life;
5093 assocparams.sasoc_number_peer_destinations =
5095 sasoc_number_peer_destinations;
5098 if (put_user(len, optlen))
5101 if (copy_to_user(optval, &assocparams, len))
5108 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5110 * This socket option is a boolean flag which turns on or off mapped V4
5111 * addresses. If this option is turned on and the socket is type
5112 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5113 * If this option is turned off, then no mapping will be done of V4
5114 * addresses and a user will receive both PF_INET6 and PF_INET type
5115 * addresses on the socket.
5117 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5118 char __user *optval, int __user *optlen)
5121 struct sctp_sock *sp = sctp_sk(sk);
5123 if (len < sizeof(int))
5128 if (put_user(len, optlen))
5130 if (copy_to_user(optval, &val, len))
5137 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5138 * (chapter and verse is quoted at sctp_setsockopt_context())
5140 static int sctp_getsockopt_context(struct sock *sk, int len,
5141 char __user *optval, int __user *optlen)
5143 struct sctp_assoc_value params;
5144 struct sctp_sock *sp;
5145 struct sctp_association *asoc;
5147 if (len < sizeof(struct sctp_assoc_value))
5150 len = sizeof(struct sctp_assoc_value);
5152 if (copy_from_user(¶ms, optval, len))
5157 if (params.assoc_id != 0) {
5158 asoc = sctp_id2assoc(sk, params.assoc_id);
5161 params.assoc_value = asoc->default_rcv_context;
5163 params.assoc_value = sp->default_rcv_context;
5166 if (put_user(len, optlen))
5168 if (copy_to_user(optval, ¶ms, len))
5175 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5176 * This option will get or set the maximum size to put in any outgoing
5177 * SCTP DATA chunk. If a message is larger than this size it will be
5178 * fragmented by SCTP into the specified size. Note that the underlying
5179 * SCTP implementation may fragment into smaller sized chunks when the
5180 * PMTU of the underlying association is smaller than the value set by
5181 * the user. The default value for this option is '0' which indicates
5182 * the user is NOT limiting fragmentation and only the PMTU will effect
5183 * SCTP's choice of DATA chunk size. Note also that values set larger
5184 * than the maximum size of an IP datagram will effectively let SCTP
5185 * control fragmentation (i.e. the same as setting this option to 0).
5187 * The following structure is used to access and modify this parameter:
5189 * struct sctp_assoc_value {
5190 * sctp_assoc_t assoc_id;
5191 * uint32_t assoc_value;
5194 * assoc_id: This parameter is ignored for one-to-one style sockets.
5195 * For one-to-many style sockets this parameter indicates which
5196 * association the user is performing an action upon. Note that if
5197 * this field's value is zero then the endpoints default value is
5198 * changed (effecting future associations only).
5199 * assoc_value: This parameter specifies the maximum size in bytes.
5201 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5202 char __user *optval, int __user *optlen)
5204 struct sctp_assoc_value params;
5205 struct sctp_association *asoc;
5207 if (len == sizeof(int)) {
5208 pr_warn("Use of int in maxseg socket option deprecated\n");
5209 pr_warn("Use struct sctp_assoc_value instead\n");
5210 params.assoc_id = 0;
5211 } else if (len >= sizeof(struct sctp_assoc_value)) {
5212 len = sizeof(struct sctp_assoc_value);
5213 if (copy_from_user(¶ms, optval, sizeof(params)))
5218 asoc = sctp_id2assoc(sk, params.assoc_id);
5219 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5223 params.assoc_value = asoc->frag_point;
5225 params.assoc_value = sctp_sk(sk)->user_frag;
5227 if (put_user(len, optlen))
5229 if (len == sizeof(int)) {
5230 if (copy_to_user(optval, ¶ms.assoc_value, len))
5233 if (copy_to_user(optval, ¶ms, len))
5241 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5242 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5244 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5245 char __user *optval, int __user *optlen)
5249 if (len < sizeof(int))
5254 val = sctp_sk(sk)->frag_interleave;
5255 if (put_user(len, optlen))
5257 if (copy_to_user(optval, &val, len))
5264 * 7.1.25. Set or Get the sctp partial delivery point
5265 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5267 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5268 char __user *optval,
5273 if (len < sizeof(u32))
5278 val = sctp_sk(sk)->pd_point;
5279 if (put_user(len, optlen))
5281 if (copy_to_user(optval, &val, len))
5288 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5289 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5291 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5292 char __user *optval,
5295 struct sctp_assoc_value params;
5296 struct sctp_sock *sp;
5297 struct sctp_association *asoc;
5299 if (len == sizeof(int)) {
5300 pr_warn("Use of int in max_burst socket option deprecated\n");
5301 pr_warn("Use struct sctp_assoc_value instead\n");
5302 params.assoc_id = 0;
5303 } else if (len >= sizeof(struct sctp_assoc_value)) {
5304 len = sizeof(struct sctp_assoc_value);
5305 if (copy_from_user(¶ms, optval, len))
5312 if (params.assoc_id != 0) {
5313 asoc = sctp_id2assoc(sk, params.assoc_id);
5316 params.assoc_value = asoc->max_burst;
5318 params.assoc_value = sp->max_burst;
5320 if (len == sizeof(int)) {
5321 if (copy_to_user(optval, ¶ms.assoc_value, len))
5324 if (copy_to_user(optval, ¶ms, len))
5332 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5333 char __user *optval, int __user *optlen)
5335 struct net *net = sock_net(sk);
5336 struct sctp_hmacalgo __user *p = (void __user *)optval;
5337 struct sctp_hmac_algo_param *hmacs;
5341 if (!net->sctp.auth_enable)
5344 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5345 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5347 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5350 len = sizeof(struct sctp_hmacalgo) + data_len;
5351 num_idents = data_len / sizeof(u16);
5353 if (put_user(len, optlen))
5355 if (put_user(num_idents, &p->shmac_num_idents))
5357 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5362 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5363 char __user *optval, int __user *optlen)
5365 struct net *net = sock_net(sk);
5366 struct sctp_authkeyid val;
5367 struct sctp_association *asoc;
5369 if (!net->sctp.auth_enable)
5372 if (len < sizeof(struct sctp_authkeyid))
5374 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5377 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5378 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5382 val.scact_keynumber = asoc->active_key_id;
5384 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5386 len = sizeof(struct sctp_authkeyid);
5387 if (put_user(len, optlen))
5389 if (copy_to_user(optval, &val, len))
5395 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5396 char __user *optval, int __user *optlen)
5398 struct net *net = sock_net(sk);
5399 struct sctp_authchunks __user *p = (void __user *)optval;
5400 struct sctp_authchunks val;
5401 struct sctp_association *asoc;
5402 struct sctp_chunks_param *ch;
5406 if (!net->sctp.auth_enable)
5409 if (len < sizeof(struct sctp_authchunks))
5412 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5415 to = p->gauth_chunks;
5416 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5420 ch = asoc->peer.peer_chunks;
5424 /* See if the user provided enough room for all the data */
5425 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5426 if (len < num_chunks)
5429 if (copy_to_user(to, ch->chunks, num_chunks))
5432 len = sizeof(struct sctp_authchunks) + num_chunks;
5433 if (put_user(len, optlen)) return -EFAULT;
5434 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5439 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5440 char __user *optval, int __user *optlen)
5442 struct net *net = sock_net(sk);
5443 struct sctp_authchunks __user *p = (void __user *)optval;
5444 struct sctp_authchunks val;
5445 struct sctp_association *asoc;
5446 struct sctp_chunks_param *ch;
5450 if (!net->sctp.auth_enable)
5453 if (len < sizeof(struct sctp_authchunks))
5456 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5459 to = p->gauth_chunks;
5460 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5461 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5465 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5467 ch = sctp_sk(sk)->ep->auth_chunk_list;
5472 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5473 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5476 if (copy_to_user(to, ch->chunks, num_chunks))
5479 len = sizeof(struct sctp_authchunks) + num_chunks;
5480 if (put_user(len, optlen))
5482 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5489 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5490 * This option gets the current number of associations that are attached
5491 * to a one-to-many style socket. The option value is an uint32_t.
5493 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5494 char __user *optval, int __user *optlen)
5496 struct sctp_sock *sp = sctp_sk(sk);
5497 struct sctp_association *asoc;
5500 if (sctp_style(sk, TCP))
5503 if (len < sizeof(u32))
5508 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5512 if (put_user(len, optlen))
5514 if (copy_to_user(optval, &val, len))
5521 * 8.1.23 SCTP_AUTO_ASCONF
5522 * See the corresponding setsockopt entry as description
5524 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5525 char __user *optval, int __user *optlen)
5529 if (len < sizeof(int))
5533 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5535 if (put_user(len, optlen))
5537 if (copy_to_user(optval, &val, len))
5543 * 8.2.6. Get the Current Identifiers of Associations
5544 * (SCTP_GET_ASSOC_ID_LIST)
5546 * This option gets the current list of SCTP association identifiers of
5547 * the SCTP associations handled by a one-to-many style socket.
5549 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5550 char __user *optval, int __user *optlen)
5552 struct sctp_sock *sp = sctp_sk(sk);
5553 struct sctp_association *asoc;
5554 struct sctp_assoc_ids *ids;
5557 if (sctp_style(sk, TCP))
5560 if (len < sizeof(struct sctp_assoc_ids))
5563 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5567 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5570 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5572 ids = kmalloc(len, GFP_KERNEL);
5576 ids->gaids_number_of_ids = num;
5578 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5579 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5582 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5592 * SCTP_PEER_ADDR_THLDS
5594 * This option allows us to fetch the partially failed threshold for one or all
5595 * transports in an association. See Section 6.1 of:
5596 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5598 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5599 char __user *optval,
5603 struct sctp_paddrthlds val;
5604 struct sctp_transport *trans;
5605 struct sctp_association *asoc;
5607 if (len < sizeof(struct sctp_paddrthlds))
5609 len = sizeof(struct sctp_paddrthlds);
5610 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5613 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5614 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5618 val.spt_pathpfthld = asoc->pf_retrans;
5619 val.spt_pathmaxrxt = asoc->pathmaxrxt;
5621 trans = sctp_addr_id2transport(sk, &val.spt_address,
5626 val.spt_pathmaxrxt = trans->pathmaxrxt;
5627 val.spt_pathpfthld = trans->pf_retrans;
5630 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5636 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5637 char __user *optval, int __user *optlen)
5642 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5645 /* I can hardly begin to describe how wrong this is. This is
5646 * so broken as to be worse than useless. The API draft
5647 * REALLY is NOT helpful here... I am not convinced that the
5648 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5649 * are at all well-founded.
5651 if (level != SOL_SCTP) {
5652 struct sctp_af *af = sctp_sk(sk)->pf->af;
5654 retval = af->getsockopt(sk, level, optname, optval, optlen);
5658 if (get_user(len, optlen))
5665 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5667 case SCTP_DISABLE_FRAGMENTS:
5668 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5672 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5674 case SCTP_AUTOCLOSE:
5675 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5677 case SCTP_SOCKOPT_PEELOFF:
5678 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5680 case SCTP_PEER_ADDR_PARAMS:
5681 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5684 case SCTP_DELAYED_SACK:
5685 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5689 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5691 case SCTP_GET_PEER_ADDRS:
5692 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5695 case SCTP_GET_LOCAL_ADDRS:
5696 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5699 case SCTP_SOCKOPT_CONNECTX3:
5700 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5702 case SCTP_DEFAULT_SEND_PARAM:
5703 retval = sctp_getsockopt_default_send_param(sk, len,
5706 case SCTP_PRIMARY_ADDR:
5707 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5710 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5713 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5715 case SCTP_ASSOCINFO:
5716 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5718 case SCTP_I_WANT_MAPPED_V4_ADDR:
5719 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5722 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5724 case SCTP_GET_PEER_ADDR_INFO:
5725 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5728 case SCTP_ADAPTATION_LAYER:
5729 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5733 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5735 case SCTP_FRAGMENT_INTERLEAVE:
5736 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5739 case SCTP_PARTIAL_DELIVERY_POINT:
5740 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5743 case SCTP_MAX_BURST:
5744 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5747 case SCTP_AUTH_CHUNK:
5748 case SCTP_AUTH_DELETE_KEY:
5749 retval = -EOPNOTSUPP;
5751 case SCTP_HMAC_IDENT:
5752 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5754 case SCTP_AUTH_ACTIVE_KEY:
5755 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5757 case SCTP_PEER_AUTH_CHUNKS:
5758 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5761 case SCTP_LOCAL_AUTH_CHUNKS:
5762 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5765 case SCTP_GET_ASSOC_NUMBER:
5766 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5768 case SCTP_GET_ASSOC_ID_LIST:
5769 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5771 case SCTP_AUTO_ASCONF:
5772 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5774 case SCTP_PEER_ADDR_THLDS:
5775 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
5778 retval = -ENOPROTOOPT;
5782 sctp_release_sock(sk);
5786 static void sctp_hash(struct sock *sk)
5791 static void sctp_unhash(struct sock *sk)
5796 /* Check if port is acceptable. Possibly find first available port.
5798 * The port hash table (contained in the 'global' SCTP protocol storage
5799 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5800 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5801 * list (the list number is the port number hashed out, so as you
5802 * would expect from a hash function, all the ports in a given list have
5803 * such a number that hashes out to the same list number; you were
5804 * expecting that, right?); so each list has a set of ports, with a
5805 * link to the socket (struct sock) that uses it, the port number and
5806 * a fastreuse flag (FIXME: NPI ipg).
5808 static struct sctp_bind_bucket *sctp_bucket_create(
5809 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
5811 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5813 struct sctp_bind_hashbucket *head; /* hash list */
5814 struct sctp_bind_bucket *pp; /* hash list port iterator */
5815 struct hlist_node *node;
5816 unsigned short snum;
5819 snum = ntohs(addr->v4.sin_port);
5821 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5822 sctp_local_bh_disable();
5825 /* Search for an available port. */
5826 int low, high, remaining, index;
5829 inet_get_local_port_range(&low, &high);
5830 remaining = (high - low) + 1;
5831 rover = net_random() % remaining + low;
5835 if ((rover < low) || (rover > high))
5837 if (inet_is_reserved_local_port(rover))
5839 index = sctp_phashfn(sock_net(sk), rover);
5840 head = &sctp_port_hashtable[index];
5841 sctp_spin_lock(&head->lock);
5842 sctp_for_each_hentry(pp, node, &head->chain)
5843 if ((pp->port == rover) &&
5844 net_eq(sock_net(sk), pp->net))
5848 sctp_spin_unlock(&head->lock);
5849 } while (--remaining > 0);
5851 /* Exhausted local port range during search? */
5856 /* OK, here is the one we will use. HEAD (the port
5857 * hash table list entry) is non-NULL and we hold it's
5862 /* We are given an specific port number; we verify
5863 * that it is not being used. If it is used, we will
5864 * exahust the search in the hash list corresponding
5865 * to the port number (snum) - we detect that with the
5866 * port iterator, pp being NULL.
5868 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
5869 sctp_spin_lock(&head->lock);
5870 sctp_for_each_hentry(pp, node, &head->chain) {
5871 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
5878 if (!hlist_empty(&pp->owner)) {
5879 /* We had a port hash table hit - there is an
5880 * available port (pp != NULL) and it is being
5881 * used by other socket (pp->owner not empty); that other
5882 * socket is going to be sk2.
5884 int reuse = sk->sk_reuse;
5887 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5888 if (pp->fastreuse && sk->sk_reuse &&
5889 sk->sk_state != SCTP_SS_LISTENING)
5892 /* Run through the list of sockets bound to the port
5893 * (pp->port) [via the pointers bind_next and
5894 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5895 * we get the endpoint they describe and run through
5896 * the endpoint's list of IP (v4 or v6) addresses,
5897 * comparing each of the addresses with the address of
5898 * the socket sk. If we find a match, then that means
5899 * that this port/socket (sk) combination are already
5902 sk_for_each_bound(sk2, node, &pp->owner) {
5903 struct sctp_endpoint *ep2;
5904 ep2 = sctp_sk(sk2)->ep;
5907 (reuse && sk2->sk_reuse &&
5908 sk2->sk_state != SCTP_SS_LISTENING))
5911 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5912 sctp_sk(sk2), sctp_sk(sk))) {
5917 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5920 /* If there was a hash table miss, create a new port. */
5922 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
5925 /* In either case (hit or miss), make sure fastreuse is 1 only
5926 * if sk->sk_reuse is too (that is, if the caller requested
5927 * SO_REUSEADDR on this socket -sk-).
5929 if (hlist_empty(&pp->owner)) {
5930 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5934 } else if (pp->fastreuse &&
5935 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5938 /* We are set, so fill up all the data in the hash table
5939 * entry, tie the socket list information with the rest of the
5940 * sockets FIXME: Blurry, NPI (ipg).
5943 if (!sctp_sk(sk)->bind_hash) {
5944 inet_sk(sk)->inet_num = snum;
5945 sk_add_bind_node(sk, &pp->owner);
5946 sctp_sk(sk)->bind_hash = pp;
5951 sctp_spin_unlock(&head->lock);
5954 sctp_local_bh_enable();
5958 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5959 * port is requested.
5961 static int sctp_get_port(struct sock *sk, unsigned short snum)
5964 union sctp_addr addr;
5965 struct sctp_af *af = sctp_sk(sk)->pf->af;
5967 /* Set up a dummy address struct from the sk. */
5968 af->from_sk(&addr, sk);
5969 addr.v4.sin_port = htons(snum);
5971 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5972 ret = sctp_get_port_local(sk, &addr);
5978 * Move a socket to LISTENING state.
5980 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5982 struct sctp_sock *sp = sctp_sk(sk);
5983 struct sctp_endpoint *ep = sp->ep;
5984 struct crypto_hash *tfm = NULL;
5987 /* Allocate HMAC for generating cookie. */
5988 if (!sp->hmac && sp->sctp_hmac_alg) {
5989 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
5990 tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
5992 net_info_ratelimited("failed to load transform for %s: %ld\n",
5993 sp->sctp_hmac_alg, PTR_ERR(tfm));
5996 sctp_sk(sk)->hmac = tfm;
6000 * If a bind() or sctp_bindx() is not called prior to a listen()
6001 * call that allows new associations to be accepted, the system
6002 * picks an ephemeral port and will choose an address set equivalent
6003 * to binding with a wildcard address.
6005 * This is not currently spelled out in the SCTP sockets
6006 * extensions draft, but follows the practice as seen in TCP
6010 sk->sk_state = SCTP_SS_LISTENING;
6011 if (!ep->base.bind_addr.port) {
6012 if (sctp_autobind(sk))
6015 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6016 sk->sk_state = SCTP_SS_CLOSED;
6021 sk->sk_max_ack_backlog = backlog;
6022 sctp_hash_endpoint(ep);
6027 * 4.1.3 / 5.1.3 listen()
6029 * By default, new associations are not accepted for UDP style sockets.
6030 * An application uses listen() to mark a socket as being able to
6031 * accept new associations.
6033 * On TCP style sockets, applications use listen() to ready the SCTP
6034 * endpoint for accepting inbound associations.
6036 * On both types of endpoints a backlog of '0' disables listening.
6038 * Move a socket to LISTENING state.
6040 int sctp_inet_listen(struct socket *sock, int backlog)
6042 struct sock *sk = sock->sk;
6043 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6046 if (unlikely(backlog < 0))
6051 /* Peeled-off sockets are not allowed to listen(). */
6052 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6055 if (sock->state != SS_UNCONNECTED)
6058 /* If backlog is zero, disable listening. */
6060 if (sctp_sstate(sk, CLOSED))
6064 sctp_unhash_endpoint(ep);
6065 sk->sk_state = SCTP_SS_CLOSED;
6067 sctp_sk(sk)->bind_hash->fastreuse = 1;
6071 /* If we are already listening, just update the backlog */
6072 if (sctp_sstate(sk, LISTENING))
6073 sk->sk_max_ack_backlog = backlog;
6075 err = sctp_listen_start(sk, backlog);
6082 sctp_release_sock(sk);
6087 * This function is done by modeling the current datagram_poll() and the
6088 * tcp_poll(). Note that, based on these implementations, we don't
6089 * lock the socket in this function, even though it seems that,
6090 * ideally, locking or some other mechanisms can be used to ensure
6091 * the integrity of the counters (sndbuf and wmem_alloc) used
6092 * in this place. We assume that we don't need locks either until proven
6095 * Another thing to note is that we include the Async I/O support
6096 * here, again, by modeling the current TCP/UDP code. We don't have
6097 * a good way to test with it yet.
6099 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6101 struct sock *sk = sock->sk;
6102 struct sctp_sock *sp = sctp_sk(sk);
6105 poll_wait(file, sk_sleep(sk), wait);
6107 /* A TCP-style listening socket becomes readable when the accept queue
6110 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6111 return (!list_empty(&sp->ep->asocs)) ?
6112 (POLLIN | POLLRDNORM) : 0;
6116 /* Is there any exceptional events? */
6117 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6119 if (sk->sk_shutdown & RCV_SHUTDOWN)
6120 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6121 if (sk->sk_shutdown == SHUTDOWN_MASK)
6124 /* Is it readable? Reconsider this code with TCP-style support. */
6125 if (!skb_queue_empty(&sk->sk_receive_queue))
6126 mask |= POLLIN | POLLRDNORM;
6128 /* The association is either gone or not ready. */
6129 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6132 /* Is it writable? */
6133 if (sctp_writeable(sk)) {
6134 mask |= POLLOUT | POLLWRNORM;
6136 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6138 * Since the socket is not locked, the buffer
6139 * might be made available after the writeable check and
6140 * before the bit is set. This could cause a lost I/O
6141 * signal. tcp_poll() has a race breaker for this race
6142 * condition. Based on their implementation, we put
6143 * in the following code to cover it as well.
6145 if (sctp_writeable(sk))
6146 mask |= POLLOUT | POLLWRNORM;
6151 /********************************************************************
6152 * 2nd Level Abstractions
6153 ********************************************************************/
6155 static struct sctp_bind_bucket *sctp_bucket_create(
6156 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6158 struct sctp_bind_bucket *pp;
6160 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6162 SCTP_DBG_OBJCNT_INC(bind_bucket);
6165 INIT_HLIST_HEAD(&pp->owner);
6167 hlist_add_head(&pp->node, &head->chain);
6172 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6173 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6175 if (pp && hlist_empty(&pp->owner)) {
6176 __hlist_del(&pp->node);
6177 kmem_cache_free(sctp_bucket_cachep, pp);
6178 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6182 /* Release this socket's reference to a local port. */
6183 static inline void __sctp_put_port(struct sock *sk)
6185 struct sctp_bind_hashbucket *head =
6186 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6187 inet_sk(sk)->inet_num)];
6188 struct sctp_bind_bucket *pp;
6190 sctp_spin_lock(&head->lock);
6191 pp = sctp_sk(sk)->bind_hash;
6192 __sk_del_bind_node(sk);
6193 sctp_sk(sk)->bind_hash = NULL;
6194 inet_sk(sk)->inet_num = 0;
6195 sctp_bucket_destroy(pp);
6196 sctp_spin_unlock(&head->lock);
6199 void sctp_put_port(struct sock *sk)
6201 sctp_local_bh_disable();
6202 __sctp_put_port(sk);
6203 sctp_local_bh_enable();
6207 * The system picks an ephemeral port and choose an address set equivalent
6208 * to binding with a wildcard address.
6209 * One of those addresses will be the primary address for the association.
6210 * This automatically enables the multihoming capability of SCTP.
6212 static int sctp_autobind(struct sock *sk)
6214 union sctp_addr autoaddr;
6218 /* Initialize a local sockaddr structure to INADDR_ANY. */
6219 af = sctp_sk(sk)->pf->af;
6221 port = htons(inet_sk(sk)->inet_num);
6222 af->inaddr_any(&autoaddr, port);
6224 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6227 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6230 * 4.2 The cmsghdr Structure *
6232 * When ancillary data is sent or received, any number of ancillary data
6233 * objects can be specified by the msg_control and msg_controllen members of
6234 * the msghdr structure, because each object is preceded by
6235 * a cmsghdr structure defining the object's length (the cmsg_len member).
6236 * Historically Berkeley-derived implementations have passed only one object
6237 * at a time, but this API allows multiple objects to be
6238 * passed in a single call to sendmsg() or recvmsg(). The following example
6239 * shows two ancillary data objects in a control buffer.
6241 * |<--------------------------- msg_controllen -------------------------->|
6244 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6246 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6249 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6251 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6254 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6255 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6257 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6259 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6266 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
6267 sctp_cmsgs_t *cmsgs)
6269 struct cmsghdr *cmsg;
6270 struct msghdr *my_msg = (struct msghdr *)msg;
6272 for (cmsg = CMSG_FIRSTHDR(msg);
6274 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6275 if (!CMSG_OK(my_msg, cmsg))
6278 /* Should we parse this header or ignore? */
6279 if (cmsg->cmsg_level != IPPROTO_SCTP)
6282 /* Strictly check lengths following example in SCM code. */
6283 switch (cmsg->cmsg_type) {
6285 /* SCTP Socket API Extension
6286 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6288 * This cmsghdr structure provides information for
6289 * initializing new SCTP associations with sendmsg().
6290 * The SCTP_INITMSG socket option uses this same data
6291 * structure. This structure is not used for
6294 * cmsg_level cmsg_type cmsg_data[]
6295 * ------------ ------------ ----------------------
6296 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6298 if (cmsg->cmsg_len !=
6299 CMSG_LEN(sizeof(struct sctp_initmsg)))
6301 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6305 /* SCTP Socket API Extension
6306 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6308 * This cmsghdr structure specifies SCTP options for
6309 * sendmsg() and describes SCTP header information
6310 * about a received message through recvmsg().
6312 * cmsg_level cmsg_type cmsg_data[]
6313 * ------------ ------------ ----------------------
6314 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6316 if (cmsg->cmsg_len !=
6317 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6321 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6323 /* Minimally, validate the sinfo_flags. */
6324 if (cmsgs->info->sinfo_flags &
6325 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6326 SCTP_ABORT | SCTP_EOF))
6338 * Wait for a packet..
6339 * Note: This function is the same function as in core/datagram.c
6340 * with a few modifications to make lksctp work.
6342 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6347 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6349 /* Socket errors? */
6350 error = sock_error(sk);
6354 if (!skb_queue_empty(&sk->sk_receive_queue))
6357 /* Socket shut down? */
6358 if (sk->sk_shutdown & RCV_SHUTDOWN)
6361 /* Sequenced packets can come disconnected. If so we report the
6366 /* Is there a good reason to think that we may receive some data? */
6367 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6370 /* Handle signals. */
6371 if (signal_pending(current))
6374 /* Let another process have a go. Since we are going to sleep
6375 * anyway. Note: This may cause odd behaviors if the message
6376 * does not fit in the user's buffer, but this seems to be the
6377 * only way to honor MSG_DONTWAIT realistically.
6379 sctp_release_sock(sk);
6380 *timeo_p = schedule_timeout(*timeo_p);
6384 finish_wait(sk_sleep(sk), &wait);
6388 error = sock_intr_errno(*timeo_p);
6391 finish_wait(sk_sleep(sk), &wait);
6396 /* Receive a datagram.
6397 * Note: This is pretty much the same routine as in core/datagram.c
6398 * with a few changes to make lksctp work.
6400 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6401 int noblock, int *err)
6404 struct sk_buff *skb;
6407 timeo = sock_rcvtimeo(sk, noblock);
6409 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6410 timeo, MAX_SCHEDULE_TIMEOUT);
6413 /* Again only user level code calls this function,
6414 * so nothing interrupt level
6415 * will suddenly eat the receive_queue.
6417 * Look at current nfs client by the way...
6418 * However, this function was correct in any case. 8)
6420 if (flags & MSG_PEEK) {
6421 spin_lock_bh(&sk->sk_receive_queue.lock);
6422 skb = skb_peek(&sk->sk_receive_queue);
6424 atomic_inc(&skb->users);
6425 spin_unlock_bh(&sk->sk_receive_queue.lock);
6427 skb = skb_dequeue(&sk->sk_receive_queue);
6433 /* Caller is allowed not to check sk->sk_err before calling. */
6434 error = sock_error(sk);
6438 if (sk->sk_shutdown & RCV_SHUTDOWN)
6441 /* User doesn't want to wait. */
6445 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6454 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6455 static void __sctp_write_space(struct sctp_association *asoc)
6457 struct sock *sk = asoc->base.sk;
6458 struct socket *sock = sk->sk_socket;
6460 if ((sctp_wspace(asoc) > 0) && sock) {
6461 if (waitqueue_active(&asoc->wait))
6462 wake_up_interruptible(&asoc->wait);
6464 if (sctp_writeable(sk)) {
6465 wait_queue_head_t *wq = sk_sleep(sk);
6467 if (wq && waitqueue_active(wq))
6468 wake_up_interruptible(wq);
6470 /* Note that we try to include the Async I/O support
6471 * here by modeling from the current TCP/UDP code.
6472 * We have not tested with it yet.
6474 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6475 sock_wake_async(sock,
6476 SOCK_WAKE_SPACE, POLL_OUT);
6481 /* Do accounting for the sndbuf space.
6482 * Decrement the used sndbuf space of the corresponding association by the
6483 * data size which was just transmitted(freed).
6485 static void sctp_wfree(struct sk_buff *skb)
6487 struct sctp_association *asoc;
6488 struct sctp_chunk *chunk;
6491 /* Get the saved chunk pointer. */
6492 chunk = *((struct sctp_chunk **)(skb->cb));
6495 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6496 sizeof(struct sk_buff) +
6497 sizeof(struct sctp_chunk);
6499 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6502 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6504 sk->sk_wmem_queued -= skb->truesize;
6505 sk_mem_uncharge(sk, skb->truesize);
6508 __sctp_write_space(asoc);
6510 sctp_association_put(asoc);
6513 /* Do accounting for the receive space on the socket.
6514 * Accounting for the association is done in ulpevent.c
6515 * We set this as a destructor for the cloned data skbs so that
6516 * accounting is done at the correct time.
6518 void sctp_sock_rfree(struct sk_buff *skb)
6520 struct sock *sk = skb->sk;
6521 struct sctp_ulpevent *event = sctp_skb2event(skb);
6523 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6526 * Mimic the behavior of sock_rfree
6528 sk_mem_uncharge(sk, event->rmem_len);
6532 /* Helper function to wait for space in the sndbuf. */
6533 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6536 struct sock *sk = asoc->base.sk;
6538 long current_timeo = *timeo_p;
6541 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6542 asoc, (long)(*timeo_p), msg_len);
6544 /* Increment the association's refcnt. */
6545 sctp_association_hold(asoc);
6547 /* Wait on the association specific sndbuf space. */
6549 prepare_to_wait_exclusive(&asoc->wait, &wait,
6550 TASK_INTERRUPTIBLE);
6553 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6556 if (signal_pending(current))
6557 goto do_interrupted;
6558 if (msg_len <= sctp_wspace(asoc))
6561 /* Let another process have a go. Since we are going
6564 sctp_release_sock(sk);
6565 current_timeo = schedule_timeout(current_timeo);
6566 BUG_ON(sk != asoc->base.sk);
6569 *timeo_p = current_timeo;
6573 finish_wait(&asoc->wait, &wait);
6575 /* Release the association's refcnt. */
6576 sctp_association_put(asoc);
6585 err = sock_intr_errno(*timeo_p);
6593 void sctp_data_ready(struct sock *sk, int len)
6595 struct socket_wq *wq;
6598 wq = rcu_dereference(sk->sk_wq);
6599 if (wq_has_sleeper(wq))
6600 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6601 POLLRDNORM | POLLRDBAND);
6602 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6606 /* If socket sndbuf has changed, wake up all per association waiters. */
6607 void sctp_write_space(struct sock *sk)
6609 struct sctp_association *asoc;
6611 /* Wake up the tasks in each wait queue. */
6612 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6613 __sctp_write_space(asoc);
6617 /* Is there any sndbuf space available on the socket?
6619 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6620 * associations on the same socket. For a UDP-style socket with
6621 * multiple associations, it is possible for it to be "unwriteable"
6622 * prematurely. I assume that this is acceptable because
6623 * a premature "unwriteable" is better than an accidental "writeable" which
6624 * would cause an unwanted block under certain circumstances. For the 1-1
6625 * UDP-style sockets or TCP-style sockets, this code should work.
6628 static int sctp_writeable(struct sock *sk)
6632 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6638 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6639 * returns immediately with EINPROGRESS.
6641 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6643 struct sock *sk = asoc->base.sk;
6645 long current_timeo = *timeo_p;
6648 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6651 /* Increment the association's refcnt. */
6652 sctp_association_hold(asoc);
6655 prepare_to_wait_exclusive(&asoc->wait, &wait,
6656 TASK_INTERRUPTIBLE);
6659 if (sk->sk_shutdown & RCV_SHUTDOWN)
6661 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6664 if (signal_pending(current))
6665 goto do_interrupted;
6667 if (sctp_state(asoc, ESTABLISHED))
6670 /* Let another process have a go. Since we are going
6673 sctp_release_sock(sk);
6674 current_timeo = schedule_timeout(current_timeo);
6677 *timeo_p = current_timeo;
6681 finish_wait(&asoc->wait, &wait);
6683 /* Release the association's refcnt. */
6684 sctp_association_put(asoc);
6689 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6692 err = -ECONNREFUSED;
6696 err = sock_intr_errno(*timeo_p);
6704 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6706 struct sctp_endpoint *ep;
6710 ep = sctp_sk(sk)->ep;
6714 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6715 TASK_INTERRUPTIBLE);
6717 if (list_empty(&ep->asocs)) {
6718 sctp_release_sock(sk);
6719 timeo = schedule_timeout(timeo);
6724 if (!sctp_sstate(sk, LISTENING))
6728 if (!list_empty(&ep->asocs))
6731 err = sock_intr_errno(timeo);
6732 if (signal_pending(current))
6740 finish_wait(sk_sleep(sk), &wait);
6745 static void sctp_wait_for_close(struct sock *sk, long timeout)
6750 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6751 if (list_empty(&sctp_sk(sk)->ep->asocs))
6753 sctp_release_sock(sk);
6754 timeout = schedule_timeout(timeout);
6756 } while (!signal_pending(current) && timeout);
6758 finish_wait(sk_sleep(sk), &wait);
6761 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6763 struct sk_buff *frag;
6768 /* Don't forget the fragments. */
6769 skb_walk_frags(skb, frag)
6770 sctp_skb_set_owner_r_frag(frag, sk);
6773 sctp_skb_set_owner_r(skb, sk);
6776 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6777 struct sctp_association *asoc)
6779 struct inet_sock *inet = inet_sk(sk);
6780 struct inet_sock *newinet;
6782 newsk->sk_type = sk->sk_type;
6783 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6784 newsk->sk_flags = sk->sk_flags;
6785 newsk->sk_no_check = sk->sk_no_check;
6786 newsk->sk_reuse = sk->sk_reuse;
6788 newsk->sk_shutdown = sk->sk_shutdown;
6789 newsk->sk_destruct = inet_sock_destruct;
6790 newsk->sk_family = sk->sk_family;
6791 newsk->sk_protocol = IPPROTO_SCTP;
6792 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6793 newsk->sk_sndbuf = sk->sk_sndbuf;
6794 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6795 newsk->sk_lingertime = sk->sk_lingertime;
6796 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6797 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6799 newinet = inet_sk(newsk);
6801 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6802 * getsockname() and getpeername()
6804 newinet->inet_sport = inet->inet_sport;
6805 newinet->inet_saddr = inet->inet_saddr;
6806 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6807 newinet->inet_dport = htons(asoc->peer.port);
6808 newinet->pmtudisc = inet->pmtudisc;
6809 newinet->inet_id = asoc->next_tsn ^ jiffies;
6811 newinet->uc_ttl = inet->uc_ttl;
6812 newinet->mc_loop = 1;
6813 newinet->mc_ttl = 1;
6814 newinet->mc_index = 0;
6815 newinet->mc_list = NULL;
6818 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6819 * and its messages to the newsk.
6821 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6822 struct sctp_association *assoc,
6823 sctp_socket_type_t type)
6825 struct sctp_sock *oldsp = sctp_sk(oldsk);
6826 struct sctp_sock *newsp = sctp_sk(newsk);
6827 struct sctp_bind_bucket *pp; /* hash list port iterator */
6828 struct sctp_endpoint *newep = newsp->ep;
6829 struct sk_buff *skb, *tmp;
6830 struct sctp_ulpevent *event;
6831 struct sctp_bind_hashbucket *head;
6832 struct list_head tmplist;
6834 /* Migrate socket buffer sizes and all the socket level options to the
6837 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6838 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6839 /* Brute force copy old sctp opt. */
6840 if (oldsp->do_auto_asconf) {
6841 memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
6842 inet_sk_copy_descendant(newsk, oldsk);
6843 memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
6845 inet_sk_copy_descendant(newsk, oldsk);
6847 /* Restore the ep value that was overwritten with the above structure
6853 /* Hook this new socket in to the bind_hash list. */
6854 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
6855 inet_sk(oldsk)->inet_num)];
6856 sctp_local_bh_disable();
6857 sctp_spin_lock(&head->lock);
6858 pp = sctp_sk(oldsk)->bind_hash;
6859 sk_add_bind_node(newsk, &pp->owner);
6860 sctp_sk(newsk)->bind_hash = pp;
6861 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6862 sctp_spin_unlock(&head->lock);
6863 sctp_local_bh_enable();
6865 /* Copy the bind_addr list from the original endpoint to the new
6866 * endpoint so that we can handle restarts properly
6868 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6869 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6871 /* Move any messages in the old socket's receive queue that are for the
6872 * peeled off association to the new socket's receive queue.
6874 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6875 event = sctp_skb2event(skb);
6876 if (event->asoc == assoc) {
6877 __skb_unlink(skb, &oldsk->sk_receive_queue);
6878 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6879 sctp_skb_set_owner_r_frag(skb, newsk);
6883 /* Clean up any messages pending delivery due to partial
6884 * delivery. Three cases:
6885 * 1) No partial deliver; no work.
6886 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6887 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6889 skb_queue_head_init(&newsp->pd_lobby);
6890 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6892 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6893 struct sk_buff_head *queue;
6895 /* Decide which queue to move pd_lobby skbs to. */
6896 if (assoc->ulpq.pd_mode) {
6897 queue = &newsp->pd_lobby;
6899 queue = &newsk->sk_receive_queue;
6901 /* Walk through the pd_lobby, looking for skbs that
6902 * need moved to the new socket.
6904 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6905 event = sctp_skb2event(skb);
6906 if (event->asoc == assoc) {
6907 __skb_unlink(skb, &oldsp->pd_lobby);
6908 __skb_queue_tail(queue, skb);
6909 sctp_skb_set_owner_r_frag(skb, newsk);
6913 /* Clear up any skbs waiting for the partial
6914 * delivery to finish.
6916 if (assoc->ulpq.pd_mode)
6917 sctp_clear_pd(oldsk, NULL);
6921 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6922 sctp_skb_set_owner_r_frag(skb, newsk);
6924 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6925 sctp_skb_set_owner_r_frag(skb, newsk);
6927 /* Set the type of socket to indicate that it is peeled off from the
6928 * original UDP-style socket or created with the accept() call on a
6929 * TCP-style socket..
6933 /* Mark the new socket "in-use" by the user so that any packets
6934 * that may arrive on the association after we've moved it are
6935 * queued to the backlog. This prevents a potential race between
6936 * backlog processing on the old socket and new-packet processing
6937 * on the new socket.
6939 * The caller has just allocated newsk so we can guarantee that other
6940 * paths won't try to lock it and then oldsk.
6942 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6943 sctp_assoc_migrate(assoc, newsk);
6945 /* If the association on the newsk is already closed before accept()
6946 * is called, set RCV_SHUTDOWN flag.
6948 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6949 newsk->sk_shutdown |= RCV_SHUTDOWN;
6951 newsk->sk_state = SCTP_SS_ESTABLISHED;
6952 sctp_release_sock(newsk);
6956 /* This proto struct describes the ULP interface for SCTP. */
6957 struct proto sctp_prot = {
6959 .owner = THIS_MODULE,
6960 .close = sctp_close,
6961 .connect = sctp_connect,
6962 .disconnect = sctp_disconnect,
6963 .accept = sctp_accept,
6964 .ioctl = sctp_ioctl,
6965 .init = sctp_init_sock,
6966 .destroy = sctp_destroy_sock,
6967 .shutdown = sctp_shutdown,
6968 .setsockopt = sctp_setsockopt,
6969 .getsockopt = sctp_getsockopt,
6970 .sendmsg = sctp_sendmsg,
6971 .recvmsg = sctp_recvmsg,
6973 .backlog_rcv = sctp_backlog_rcv,
6975 .unhash = sctp_unhash,
6976 .get_port = sctp_get_port,
6977 .obj_size = sizeof(struct sctp_sock),
6978 .sysctl_mem = sysctl_sctp_mem,
6979 .sysctl_rmem = sysctl_sctp_rmem,
6980 .sysctl_wmem = sysctl_sctp_wmem,
6981 .memory_pressure = &sctp_memory_pressure,
6982 .enter_memory_pressure = sctp_enter_memory_pressure,
6983 .memory_allocated = &sctp_memory_allocated,
6984 .sockets_allocated = &sctp_sockets_allocated,
6987 #if IS_ENABLED(CONFIG_IPV6)
6989 struct proto sctpv6_prot = {
6991 .owner = THIS_MODULE,
6992 .close = sctp_close,
6993 .connect = sctp_connect,
6994 .disconnect = sctp_disconnect,
6995 .accept = sctp_accept,
6996 .ioctl = sctp_ioctl,
6997 .init = sctp_init_sock,
6998 .destroy = sctp_destroy_sock,
6999 .shutdown = sctp_shutdown,
7000 .setsockopt = sctp_setsockopt,
7001 .getsockopt = sctp_getsockopt,
7002 .sendmsg = sctp_sendmsg,
7003 .recvmsg = sctp_recvmsg,
7005 .backlog_rcv = sctp_backlog_rcv,
7007 .unhash = sctp_unhash,
7008 .get_port = sctp_get_port,
7009 .obj_size = sizeof(struct sctp6_sock),
7010 .sysctl_mem = sysctl_sctp_mem,
7011 .sysctl_rmem = sysctl_sctp_rmem,
7012 .sysctl_wmem = sysctl_sctp_wmem,
7013 .memory_pressure = &sctp_memory_pressure,
7014 .enter_memory_pressure = sctp_enter_memory_pressure,
7015 .memory_allocated = &sctp_memory_allocated,
7016 .sockets_allocated = &sctp_sockets_allocated,
7018 #endif /* IS_ENABLED(CONFIG_IPV6) */