1 /* SCTP kernel implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
9 * This file is part of the SCTP kernel implementation
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
18 * This SCTP implementation is free software;
19 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
24 * This SCTP implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, write to
32 * the Free Software Foundation, 59 Temple Place - Suite 330,
33 * Boston, MA 02111-1307, USA.
35 * Please send any bug reports or fixes you make to the
37 * lksctp developers <lksctp-developers@lists.sourceforge.net>
39 * Or submit a bug report through the following website:
40 * http://www.sf.net/projects/lksctp
42 * Written or modified by:
43 * La Monte H.P. Yarroll <piggy@acm.org>
44 * Narasimha Budihal <narsi@refcode.org>
45 * Karl Knutson <karl@athena.chicago.il.us>
46 * Jon Grimm <jgrimm@us.ibm.com>
47 * Xingang Guo <xingang.guo@intel.com>
48 * Daisy Chang <daisyc@us.ibm.com>
49 * Sridhar Samudrala <samudrala@us.ibm.com>
50 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
51 * Ardelle Fan <ardelle.fan@intel.com>
52 * Ryan Layer <rmlayer@us.ibm.com>
53 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
54 * Kevin Gao <kevin.gao@intel.com>
56 * Any bugs reported given to us we will try to fix... any fixes shared will
57 * be incorporated into the next SCTP release.
60 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
62 #include <linux/types.h>
63 #include <linux/kernel.h>
64 #include <linux/wait.h>
65 #include <linux/time.h>
67 #include <linux/capability.h>
68 #include <linux/fcntl.h>
69 #include <linux/poll.h>
70 #include <linux/init.h>
71 #include <linux/crypto.h>
72 #include <linux/slab.h>
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 net *net = sock_net(sk);
339 struct sctp_sock *sp = sctp_sk(sk);
340 struct sctp_endpoint *ep = sp->ep;
341 struct sctp_bind_addr *bp = &ep->base.bind_addr;
346 /* Common sockaddr verification. */
347 af = sctp_sockaddr_af(sp, addr, len);
349 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
354 snum = ntohs(addr->v4.sin_port);
356 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
357 ", port: %d, new port: %d, len: %d)\n",
363 /* PF specific bind() address verification. */
364 if (!sp->pf->bind_verify(sp, addr))
365 return -EADDRNOTAVAIL;
367 /* We must either be unbound, or bind to the same port.
368 * It's OK to allow 0 ports if we are already bound.
369 * We'll just inhert an already bound port in this case
374 else if (snum != bp->port) {
375 SCTP_DEBUG_PRINTK("sctp_do_bind:"
376 " New port %d does not match existing port "
377 "%d.\n", snum, bp->port);
382 if (snum && snum < PROT_SOCK &&
383 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
386 /* See if the address matches any of the addresses we may have
387 * already bound before checking against other endpoints.
389 if (sctp_bind_addr_match(bp, addr, sp))
392 /* Make sure we are allowed to bind here.
393 * The function sctp_get_port_local() does duplicate address
396 addr->v4.sin_port = htons(snum);
397 if ((ret = sctp_get_port_local(sk, addr))) {
401 /* Refresh ephemeral port. */
403 bp->port = inet_sk(sk)->inet_num;
405 /* Add the address to the bind address list.
406 * Use GFP_ATOMIC since BHs will be disabled.
408 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
410 /* Copy back into socket for getsockname() use. */
412 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
413 af->to_sk_saddr(addr, sk);
419 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
421 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
422 * at any one time. If a sender, after sending an ASCONF chunk, decides
423 * it needs to transfer another ASCONF Chunk, it MUST wait until the
424 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
425 * subsequent ASCONF. Note this restriction binds each side, so at any
426 * time two ASCONF may be in-transit on any given association (one sent
427 * from each endpoint).
429 static int sctp_send_asconf(struct sctp_association *asoc,
430 struct sctp_chunk *chunk)
432 struct net *net = sock_net(asoc->base.sk);
435 /* If there is an outstanding ASCONF chunk, queue it for later
438 if (asoc->addip_last_asconf) {
439 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
443 /* Hold the chunk until an ASCONF_ACK is received. */
444 sctp_chunk_hold(chunk);
445 retval = sctp_primitive_ASCONF(net, asoc, chunk);
447 sctp_chunk_free(chunk);
449 asoc->addip_last_asconf = chunk;
455 /* Add a list of addresses as bind addresses to local endpoint or
458 * Basically run through each address specified in the addrs/addrcnt
459 * array/length pair, determine if it is IPv6 or IPv4 and call
460 * sctp_do_bind() on it.
462 * If any of them fails, then the operation will be reversed and the
463 * ones that were added will be removed.
465 * Only sctp_setsockopt_bindx() is supposed to call this function.
467 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
472 struct sockaddr *sa_addr;
475 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
479 for (cnt = 0; cnt < addrcnt; cnt++) {
480 /* The list may contain either IPv4 or IPv6 address;
481 * determine the address length for walking thru the list.
484 af = sctp_get_af_specific(sa_addr->sa_family);
490 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
493 addr_buf += af->sockaddr_len;
497 /* Failed. Cleanup the ones that have been added */
499 sctp_bindx_rem(sk, addrs, cnt);
507 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
508 * associations that are part of the endpoint indicating that a list of local
509 * addresses are added to the endpoint.
511 * If any of the addresses is already in the bind address list of the
512 * association, we do not send the chunk for that association. But it will not
513 * affect other associations.
515 * Only sctp_setsockopt_bindx() is supposed to call this function.
517 static int sctp_send_asconf_add_ip(struct sock *sk,
518 struct sockaddr *addrs,
521 struct net *net = sock_net(sk);
522 struct sctp_sock *sp;
523 struct sctp_endpoint *ep;
524 struct sctp_association *asoc;
525 struct sctp_bind_addr *bp;
526 struct sctp_chunk *chunk;
527 struct sctp_sockaddr_entry *laddr;
528 union sctp_addr *addr;
529 union sctp_addr saveaddr;
536 if (!net->sctp.addip_enable)
542 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
543 __func__, sk, addrs, addrcnt);
545 list_for_each_entry(asoc, &ep->asocs, asocs) {
547 if (!asoc->peer.asconf_capable)
550 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
553 if (!sctp_state(asoc, ESTABLISHED))
556 /* Check if any address in the packed array of addresses is
557 * in the bind address list of the association. If so,
558 * do not send the asconf chunk to its peer, but continue with
559 * other associations.
562 for (i = 0; i < addrcnt; i++) {
564 af = sctp_get_af_specific(addr->v4.sin_family);
570 if (sctp_assoc_lookup_laddr(asoc, addr))
573 addr_buf += af->sockaddr_len;
578 /* Use the first valid address in bind addr list of
579 * association as Address Parameter of ASCONF CHUNK.
581 bp = &asoc->base.bind_addr;
582 p = bp->address_list.next;
583 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
584 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
585 addrcnt, SCTP_PARAM_ADD_IP);
591 /* Add the new addresses to the bind address list with
592 * use_as_src set to 0.
595 for (i = 0; i < addrcnt; i++) {
597 af = sctp_get_af_specific(addr->v4.sin_family);
598 memcpy(&saveaddr, addr, af->sockaddr_len);
599 retval = sctp_add_bind_addr(bp, &saveaddr,
600 SCTP_ADDR_NEW, GFP_ATOMIC);
601 addr_buf += af->sockaddr_len;
603 if (asoc->src_out_of_asoc_ok) {
604 struct sctp_transport *trans;
606 list_for_each_entry(trans,
607 &asoc->peer.transport_addr_list, transports) {
608 /* Clear the source and route cache */
609 dst_release(trans->dst);
610 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
611 2*asoc->pathmtu, 4380));
612 trans->ssthresh = asoc->peer.i.a_rwnd;
613 trans->rto = asoc->rto_initial;
614 trans->rtt = trans->srtt = trans->rttvar = 0;
615 sctp_transport_route(trans, NULL,
616 sctp_sk(asoc->base.sk));
619 retval = sctp_send_asconf(asoc, chunk);
626 /* Remove a list of addresses from bind addresses list. Do not remove the
629 * Basically run through each address specified in the addrs/addrcnt
630 * array/length pair, determine if it is IPv6 or IPv4 and call
631 * sctp_del_bind() on it.
633 * If any of them fails, then the operation will be reversed and the
634 * ones that were removed will be added back.
636 * At least one address has to be left; if only one address is
637 * available, the operation will return -EBUSY.
639 * Only sctp_setsockopt_bindx() is supposed to call this function.
641 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
643 struct sctp_sock *sp = sctp_sk(sk);
644 struct sctp_endpoint *ep = sp->ep;
646 struct sctp_bind_addr *bp = &ep->base.bind_addr;
649 union sctp_addr *sa_addr;
652 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
656 for (cnt = 0; cnt < addrcnt; cnt++) {
657 /* If the bind address list is empty or if there is only one
658 * bind address, there is nothing more to be removed (we need
659 * at least one address here).
661 if (list_empty(&bp->address_list) ||
662 (sctp_list_single_entry(&bp->address_list))) {
668 af = sctp_get_af_specific(sa_addr->sa.sa_family);
674 if (!af->addr_valid(sa_addr, sp, NULL)) {
675 retval = -EADDRNOTAVAIL;
679 if (sa_addr->v4.sin_port &&
680 sa_addr->v4.sin_port != htons(bp->port)) {
685 if (!sa_addr->v4.sin_port)
686 sa_addr->v4.sin_port = htons(bp->port);
688 /* FIXME - There is probably a need to check if sk->sk_saddr and
689 * sk->sk_rcv_addr are currently set to one of the addresses to
690 * be removed. This is something which needs to be looked into
691 * when we are fixing the outstanding issues with multi-homing
692 * socket routing and failover schemes. Refer to comments in
693 * sctp_do_bind(). -daisy
695 retval = sctp_del_bind_addr(bp, sa_addr);
697 addr_buf += af->sockaddr_len;
700 /* Failed. Add the ones that has been removed back */
702 sctp_bindx_add(sk, addrs, cnt);
710 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
711 * the associations that are part of the endpoint indicating that a list of
712 * local addresses are removed from the endpoint.
714 * If any of the addresses is already in the bind address list of the
715 * association, we do not send the chunk for that association. But it will not
716 * affect other associations.
718 * Only sctp_setsockopt_bindx() is supposed to call this function.
720 static int sctp_send_asconf_del_ip(struct sock *sk,
721 struct sockaddr *addrs,
724 struct net *net = sock_net(sk);
725 struct sctp_sock *sp;
726 struct sctp_endpoint *ep;
727 struct sctp_association *asoc;
728 struct sctp_transport *transport;
729 struct sctp_bind_addr *bp;
730 struct sctp_chunk *chunk;
731 union sctp_addr *laddr;
734 struct sctp_sockaddr_entry *saddr;
740 if (!net->sctp.addip_enable)
746 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
747 __func__, sk, addrs, addrcnt);
749 list_for_each_entry(asoc, &ep->asocs, asocs) {
751 if (!asoc->peer.asconf_capable)
754 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
757 if (!sctp_state(asoc, ESTABLISHED))
760 /* Check if any address in the packed array of addresses is
761 * not present in the bind address list of the association.
762 * If so, do not send the asconf chunk to its peer, but
763 * continue with other associations.
766 for (i = 0; i < addrcnt; i++) {
768 af = sctp_get_af_specific(laddr->v4.sin_family);
774 if (!sctp_assoc_lookup_laddr(asoc, laddr))
777 addr_buf += af->sockaddr_len;
782 /* Find one address in the association's bind address list
783 * that is not in the packed array of addresses. This is to
784 * make sure that we do not delete all the addresses in the
787 bp = &asoc->base.bind_addr;
788 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
790 if ((laddr == NULL) && (addrcnt == 1)) {
791 if (asoc->asconf_addr_del_pending)
793 asoc->asconf_addr_del_pending =
794 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
795 if (asoc->asconf_addr_del_pending == NULL) {
799 asoc->asconf_addr_del_pending->sa.sa_family =
801 asoc->asconf_addr_del_pending->v4.sin_port =
803 if (addrs->sa_family == AF_INET) {
804 struct sockaddr_in *sin;
806 sin = (struct sockaddr_in *)addrs;
807 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
808 } else if (addrs->sa_family == AF_INET6) {
809 struct sockaddr_in6 *sin6;
811 sin6 = (struct sockaddr_in6 *)addrs;
812 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
814 SCTP_DEBUG_PRINTK_IPADDR("send_asconf_del_ip: keep the last address asoc: %p ",
815 " at %p\n", asoc, asoc->asconf_addr_del_pending,
816 asoc->asconf_addr_del_pending);
817 asoc->src_out_of_asoc_ok = 1;
822 /* We do not need RCU protection throughout this loop
823 * because this is done under a socket lock from the
826 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
834 /* Reset use_as_src flag for the addresses in the bind address
835 * list that are to be deleted.
838 for (i = 0; i < addrcnt; i++) {
840 af = sctp_get_af_specific(laddr->v4.sin_family);
841 list_for_each_entry(saddr, &bp->address_list, list) {
842 if (sctp_cmp_addr_exact(&saddr->a, laddr))
843 saddr->state = SCTP_ADDR_DEL;
845 addr_buf += af->sockaddr_len;
848 /* Update the route and saddr entries for all the transports
849 * as some of the addresses in the bind address list are
850 * about to be deleted and cannot be used as source addresses.
852 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
854 dst_release(transport->dst);
855 sctp_transport_route(transport, NULL,
856 sctp_sk(asoc->base.sk));
860 /* We don't need to transmit ASCONF */
862 retval = sctp_send_asconf(asoc, chunk);
868 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
869 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
871 struct sock *sk = sctp_opt2sk(sp);
872 union sctp_addr *addr;
875 /* It is safe to write port space in caller. */
877 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
878 af = sctp_get_af_specific(addr->sa.sa_family);
881 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
884 if (addrw->state == SCTP_ADDR_NEW)
885 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
887 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
890 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
893 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
896 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
897 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
900 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
901 * Section 3.1.2 for this usage.
903 * addrs is a pointer to an array of one or more socket addresses. Each
904 * address is contained in its appropriate structure (i.e. struct
905 * sockaddr_in or struct sockaddr_in6) the family of the address type
906 * must be used to distinguish the address length (note that this
907 * representation is termed a "packed array" of addresses). The caller
908 * specifies the number of addresses in the array with addrcnt.
910 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
911 * -1, and sets errno to the appropriate error code.
913 * For SCTP, the port given in each socket address must be the same, or
914 * sctp_bindx() will fail, setting errno to EINVAL.
916 * The flags parameter is formed from the bitwise OR of zero or more of
917 * the following currently defined flags:
919 * SCTP_BINDX_ADD_ADDR
921 * SCTP_BINDX_REM_ADDR
923 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
924 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
925 * addresses from the association. The two flags are mutually exclusive;
926 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
927 * not remove all addresses from an association; sctp_bindx() will
928 * reject such an attempt with EINVAL.
930 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
931 * additional addresses with an endpoint after calling bind(). Or use
932 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
933 * socket is associated with so that no new association accepted will be
934 * associated with those addresses. If the endpoint supports dynamic
935 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
936 * endpoint to send the appropriate message to the peer to change the
937 * peers address lists.
939 * Adding and removing addresses from a connected association is
940 * optional functionality. Implementations that do not support this
941 * functionality should return EOPNOTSUPP.
943 * Basically do nothing but copying the addresses from user to kernel
944 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
945 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
948 * We don't use copy_from_user() for optimization: we first do the
949 * sanity checks (buffer size -fast- and access check-healthy
950 * pointer); if all of those succeed, then we can alloc the memory
951 * (expensive operation) needed to copy the data to kernel. Then we do
952 * the copying without checking the user space area
953 * (__copy_from_user()).
955 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
958 * sk The sk of the socket
959 * addrs The pointer to the addresses in user land
960 * addrssize Size of the addrs buffer
961 * op Operation to perform (add or remove, see the flags of
964 * Returns 0 if ok, <0 errno code on error.
966 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
967 struct sockaddr __user *addrs,
968 int addrs_size, int op)
970 struct sockaddr *kaddrs;
974 struct sockaddr *sa_addr;
978 SCTP_DEBUG_PRINTK("sctp_setsockopt_bindx: sk %p addrs %p"
979 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
981 if (unlikely(addrs_size <= 0))
984 /* Check the user passed a healthy pointer. */
985 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
988 /* Alloc space for the address array in kernel memory. */
989 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
990 if (unlikely(!kaddrs))
993 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
998 /* Walk through the addrs buffer and count the number of addresses. */
1000 while (walk_size < addrs_size) {
1001 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1007 af = sctp_get_af_specific(sa_addr->sa_family);
1009 /* If the address family is not supported or if this address
1010 * causes the address buffer to overflow return EINVAL.
1012 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1017 addr_buf += af->sockaddr_len;
1018 walk_size += af->sockaddr_len;
1023 case SCTP_BINDX_ADD_ADDR:
1024 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1027 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1030 case SCTP_BINDX_REM_ADDR:
1031 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1034 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1048 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1050 * Common routine for handling connect() and sctp_connectx().
1051 * Connect will come in with just a single address.
1053 static int __sctp_connect(struct sock* sk,
1054 struct sockaddr *kaddrs,
1056 sctp_assoc_t *assoc_id)
1058 struct net *net = sock_net(sk);
1059 struct sctp_sock *sp;
1060 struct sctp_endpoint *ep;
1061 struct sctp_association *asoc = NULL;
1062 struct sctp_association *asoc2;
1063 struct sctp_transport *transport;
1071 union sctp_addr *sa_addr = NULL;
1073 unsigned short port;
1074 unsigned int f_flags = 0;
1079 /* connect() cannot be done on a socket that is already in ESTABLISHED
1080 * state - UDP-style peeled off socket or a TCP-style socket that
1081 * is already connected.
1082 * It cannot be done even on a TCP-style listening socket.
1084 if (sctp_sstate(sk, ESTABLISHED) ||
1085 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1090 /* Walk through the addrs buffer and count the number of addresses. */
1092 while (walk_size < addrs_size) {
1093 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1099 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1101 /* If the address family is not supported or if this address
1102 * causes the address buffer to overflow return EINVAL.
1104 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1109 port = ntohs(sa_addr->v4.sin_port);
1111 /* Save current address so we can work with it */
1112 memcpy(&to, sa_addr, af->sockaddr_len);
1114 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1118 /* Make sure the destination port is correctly set
1121 if (asoc && asoc->peer.port && asoc->peer.port != port)
1125 /* Check if there already is a matching association on the
1126 * endpoint (other than the one created here).
1128 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1129 if (asoc2 && asoc2 != asoc) {
1130 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1137 /* If we could not find a matching association on the endpoint,
1138 * make sure that there is no peeled-off association matching
1139 * the peer address even on another socket.
1141 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1142 err = -EADDRNOTAVAIL;
1147 /* If a bind() or sctp_bindx() is not called prior to
1148 * an sctp_connectx() call, the system picks an
1149 * ephemeral port and will choose an address set
1150 * equivalent to binding with a wildcard address.
1152 if (!ep->base.bind_addr.port) {
1153 if (sctp_autobind(sk)) {
1159 * If an unprivileged user inherits a 1-many
1160 * style socket with open associations on a
1161 * privileged port, it MAY be permitted to
1162 * accept new associations, but it SHOULD NOT
1163 * be permitted to open new associations.
1165 if (ep->base.bind_addr.port < PROT_SOCK &&
1166 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1172 scope = sctp_scope(&to);
1173 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1179 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1187 /* Prime the peer's transport structures. */
1188 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1196 addr_buf += af->sockaddr_len;
1197 walk_size += af->sockaddr_len;
1200 /* In case the user of sctp_connectx() wants an association
1201 * id back, assign one now.
1204 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1209 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1214 /* Initialize sk's dport and daddr for getpeername() */
1215 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1216 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1217 af->to_sk_daddr(sa_addr, sk);
1220 /* in-kernel sockets don't generally have a file allocated to them
1221 * if all they do is call sock_create_kern().
1223 if (sk->sk_socket->file)
1224 f_flags = sk->sk_socket->file->f_flags;
1226 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1228 err = sctp_wait_for_connect(asoc, &timeo);
1229 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1230 *assoc_id = asoc->assoc_id;
1232 /* Don't free association on exit. */
1237 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1238 " kaddrs: %p err: %d\n",
1241 /* sctp_primitive_ASSOCIATE may have added this association
1242 * To the hash table, try to unhash it, just in case, its a noop
1243 * if it wasn't hashed so we're safe
1245 sctp_unhash_established(asoc);
1246 sctp_association_free(asoc);
1251 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1254 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1255 * sctp_assoc_t *asoc);
1257 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1258 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1259 * or IPv6 addresses.
1261 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1262 * Section 3.1.2 for this usage.
1264 * addrs is a pointer to an array of one or more socket addresses. Each
1265 * address is contained in its appropriate structure (i.e. struct
1266 * sockaddr_in or struct sockaddr_in6) the family of the address type
1267 * must be used to distengish the address length (note that this
1268 * representation is termed a "packed array" of addresses). The caller
1269 * specifies the number of addresses in the array with addrcnt.
1271 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1272 * the association id of the new association. On failure, sctp_connectx()
1273 * returns -1, and sets errno to the appropriate error code. The assoc_id
1274 * is not touched by the kernel.
1276 * For SCTP, the port given in each socket address must be the same, or
1277 * sctp_connectx() will fail, setting errno to EINVAL.
1279 * An application can use sctp_connectx to initiate an association with
1280 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1281 * allows a caller to specify multiple addresses at which a peer can be
1282 * reached. The way the SCTP stack uses the list of addresses to set up
1283 * the association is implementation dependent. This function only
1284 * specifies that the stack will try to make use of all the addresses in
1285 * the list when needed.
1287 * Note that the list of addresses passed in is only used for setting up
1288 * the association. It does not necessarily equal the set of addresses
1289 * the peer uses for the resulting association. If the caller wants to
1290 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1291 * retrieve them after the association has been set up.
1293 * Basically do nothing but copying the addresses from user to kernel
1294 * land and invoking either sctp_connectx(). This is used for tunneling
1295 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1297 * We don't use copy_from_user() for optimization: we first do the
1298 * sanity checks (buffer size -fast- and access check-healthy
1299 * pointer); if all of those succeed, then we can alloc the memory
1300 * (expensive operation) needed to copy the data to kernel. Then we do
1301 * the copying without checking the user space area
1302 * (__copy_from_user()).
1304 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1307 * sk The sk of the socket
1308 * addrs The pointer to the addresses in user land
1309 * addrssize Size of the addrs buffer
1311 * Returns >=0 if ok, <0 errno code on error.
1313 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1314 struct sockaddr __user *addrs,
1316 sctp_assoc_t *assoc_id)
1319 struct sockaddr *kaddrs;
1321 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1322 __func__, sk, addrs, addrs_size);
1324 if (unlikely(addrs_size <= 0))
1327 /* Check the user passed a healthy pointer. */
1328 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1331 /* Alloc space for the address array in kernel memory. */
1332 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1333 if (unlikely(!kaddrs))
1336 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1339 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1348 * This is an older interface. It's kept for backward compatibility
1349 * to the option that doesn't provide association id.
1351 SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1352 struct sockaddr __user *addrs,
1355 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1359 * New interface for the API. The since the API is done with a socket
1360 * option, to make it simple we feed back the association id is as a return
1361 * indication to the call. Error is always negative and association id is
1364 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1365 struct sockaddr __user *addrs,
1368 sctp_assoc_t assoc_id = 0;
1371 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1380 * New (hopefully final) interface for the API.
1381 * We use the sctp_getaddrs_old structure so that use-space library
1382 * can avoid any unnecessary allocations. The only defferent part
1383 * is that we store the actual length of the address buffer into the
1384 * addrs_num structure member. That way we can re-use the existing
1387 SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1388 char __user *optval,
1391 struct sctp_getaddrs_old param;
1392 sctp_assoc_t assoc_id = 0;
1395 if (len < sizeof(param))
1398 if (copy_from_user(¶m, optval, sizeof(param)))
1401 err = __sctp_setsockopt_connectx(sk,
1402 (struct sockaddr __user *)param.addrs,
1403 param.addr_num, &assoc_id);
1405 if (err == 0 || err == -EINPROGRESS) {
1406 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1408 if (put_user(sizeof(assoc_id), optlen))
1415 /* API 3.1.4 close() - UDP Style Syntax
1416 * Applications use close() to perform graceful shutdown (as described in
1417 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1418 * by a UDP-style socket.
1422 * ret = close(int sd);
1424 * sd - the socket descriptor of the associations to be closed.
1426 * To gracefully shutdown a specific association represented by the
1427 * UDP-style socket, an application should use the sendmsg() call,
1428 * passing no user data, but including the appropriate flag in the
1429 * ancillary data (see Section xxxx).
1431 * If sd in the close() call is a branched-off socket representing only
1432 * one association, the shutdown is performed on that association only.
1434 * 4.1.6 close() - TCP Style Syntax
1436 * Applications use close() to gracefully close down an association.
1440 * int close(int sd);
1442 * sd - the socket descriptor of the association to be closed.
1444 * After an application calls close() on a socket descriptor, no further
1445 * socket operations will succeed on that descriptor.
1447 * API 7.1.4 SO_LINGER
1449 * An application using the TCP-style socket can use this option to
1450 * perform the SCTP ABORT primitive. The linger option structure is:
1453 * int l_onoff; // option on/off
1454 * int l_linger; // linger time
1457 * To enable the option, set l_onoff to 1. If the l_linger value is set
1458 * to 0, calling close() is the same as the ABORT primitive. If the
1459 * value is set to a negative value, the setsockopt() call will return
1460 * an error. If the value is set to a positive value linger_time, the
1461 * close() can be blocked for at most linger_time ms. If the graceful
1462 * shutdown phase does not finish during this period, close() will
1463 * return but the graceful shutdown phase continues in the system.
1465 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1467 struct net *net = sock_net(sk);
1468 struct sctp_endpoint *ep;
1469 struct sctp_association *asoc;
1470 struct list_head *pos, *temp;
1471 unsigned int data_was_unread;
1473 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1476 sk->sk_shutdown = SHUTDOWN_MASK;
1477 sk->sk_state = SCTP_SS_CLOSING;
1479 ep = sctp_sk(sk)->ep;
1481 /* Clean up any skbs sitting on the receive queue. */
1482 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1483 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1485 /* Walk all associations on an endpoint. */
1486 list_for_each_safe(pos, temp, &ep->asocs) {
1487 asoc = list_entry(pos, struct sctp_association, asocs);
1489 if (sctp_style(sk, TCP)) {
1490 /* A closed association can still be in the list if
1491 * it belongs to a TCP-style listening socket that is
1492 * not yet accepted. If so, free it. If not, send an
1493 * ABORT or SHUTDOWN based on the linger options.
1495 if (sctp_state(asoc, CLOSED)) {
1496 sctp_unhash_established(asoc);
1497 sctp_association_free(asoc);
1502 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1503 !skb_queue_empty(&asoc->ulpq.reasm) ||
1504 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1505 struct sctp_chunk *chunk;
1507 chunk = sctp_make_abort_user(asoc, NULL, 0);
1509 sctp_primitive_ABORT(net, asoc, chunk);
1511 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1514 /* On a TCP-style socket, block for at most linger_time if set. */
1515 if (sctp_style(sk, TCP) && timeout)
1516 sctp_wait_for_close(sk, timeout);
1518 /* This will run the backlog queue. */
1519 sctp_release_sock(sk);
1521 /* Supposedly, no process has access to the socket, but
1522 * the net layers still may.
1524 sctp_local_bh_disable();
1525 sctp_bh_lock_sock(sk);
1527 /* Hold the sock, since sk_common_release() will put sock_put()
1528 * and we have just a little more cleanup.
1531 sk_common_release(sk);
1533 sctp_bh_unlock_sock(sk);
1534 sctp_local_bh_enable();
1538 SCTP_DBG_OBJCNT_DEC(sock);
1541 /* Handle EPIPE error. */
1542 static int sctp_error(struct sock *sk, int flags, int err)
1545 err = sock_error(sk) ? : -EPIPE;
1546 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1547 send_sig(SIGPIPE, current, 0);
1551 /* API 3.1.3 sendmsg() - UDP Style Syntax
1553 * An application uses sendmsg() and recvmsg() calls to transmit data to
1554 * and receive data from its peer.
1556 * ssize_t sendmsg(int socket, const struct msghdr *message,
1559 * socket - the socket descriptor of the endpoint.
1560 * message - pointer to the msghdr structure which contains a single
1561 * user message and possibly some ancillary data.
1563 * See Section 5 for complete description of the data
1566 * flags - flags sent or received with the user message, see Section
1567 * 5 for complete description of the flags.
1569 * Note: This function could use a rewrite especially when explicit
1570 * connect support comes in.
1572 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1574 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1576 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1577 struct msghdr *msg, size_t msg_len)
1579 struct net *net = sock_net(sk);
1580 struct sctp_sock *sp;
1581 struct sctp_endpoint *ep;
1582 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1583 struct sctp_transport *transport, *chunk_tp;
1584 struct sctp_chunk *chunk;
1586 struct sockaddr *msg_name = NULL;
1587 struct sctp_sndrcvinfo default_sinfo;
1588 struct sctp_sndrcvinfo *sinfo;
1589 struct sctp_initmsg *sinit;
1590 sctp_assoc_t associd = 0;
1591 sctp_cmsgs_t cmsgs = { NULL };
1595 __u16 sinfo_flags = 0;
1596 struct sctp_datamsg *datamsg;
1597 int msg_flags = msg->msg_flags;
1599 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1606 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1608 /* We cannot send a message over a TCP-style listening socket. */
1609 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1614 /* Parse out the SCTP CMSGs. */
1615 err = sctp_msghdr_parse(msg, &cmsgs);
1618 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1622 /* Fetch the destination address for this packet. This
1623 * address only selects the association--it is not necessarily
1624 * the address we will send to.
1625 * For a peeled-off socket, msg_name is ignored.
1627 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1628 int msg_namelen = msg->msg_namelen;
1630 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1635 if (msg_namelen > sizeof(to))
1636 msg_namelen = sizeof(to);
1637 memcpy(&to, msg->msg_name, msg_namelen);
1638 msg_name = msg->msg_name;
1644 /* Did the user specify SNDRCVINFO? */
1646 sinfo_flags = sinfo->sinfo_flags;
1647 associd = sinfo->sinfo_assoc_id;
1650 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1651 msg_len, sinfo_flags);
1653 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1654 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1659 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1660 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1661 * If SCTP_ABORT is set, the message length could be non zero with
1662 * the msg_iov set to the user abort reason.
1664 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1665 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1670 /* If SCTP_ADDR_OVER is set, there must be an address
1671 * specified in msg_name.
1673 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1680 SCTP_DEBUG_PRINTK("About to look up association.\n");
1684 /* If a msg_name has been specified, assume this is to be used. */
1686 /* Look for a matching association on the endpoint. */
1687 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1689 /* If we could not find a matching association on the
1690 * endpoint, make sure that it is not a TCP-style
1691 * socket that already has an association or there is
1692 * no peeled-off association on another socket.
1694 if ((sctp_style(sk, TCP) &&
1695 sctp_sstate(sk, ESTABLISHED)) ||
1696 sctp_endpoint_is_peeled_off(ep, &to)) {
1697 err = -EADDRNOTAVAIL;
1702 asoc = sctp_id2assoc(sk, associd);
1710 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1712 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1713 * socket that has an association in CLOSED state. This can
1714 * happen when an accepted socket has an association that is
1717 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1722 if (sinfo_flags & SCTP_EOF) {
1723 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1725 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1729 if (sinfo_flags & SCTP_ABORT) {
1731 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1737 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1738 sctp_primitive_ABORT(net, asoc, chunk);
1744 /* Do we need to create the association? */
1746 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1748 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1753 /* Check for invalid stream against the stream counts,
1754 * either the default or the user specified stream counts.
1757 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1758 /* Check against the defaults. */
1759 if (sinfo->sinfo_stream >=
1760 sp->initmsg.sinit_num_ostreams) {
1765 /* Check against the requested. */
1766 if (sinfo->sinfo_stream >=
1767 sinit->sinit_num_ostreams) {
1775 * API 3.1.2 bind() - UDP Style Syntax
1776 * If a bind() or sctp_bindx() is not called prior to a
1777 * sendmsg() call that initiates a new association, the
1778 * system picks an ephemeral port and will choose an address
1779 * set equivalent to binding with a wildcard address.
1781 if (!ep->base.bind_addr.port) {
1782 if (sctp_autobind(sk)) {
1788 * If an unprivileged user inherits a one-to-many
1789 * style socket with open associations on a privileged
1790 * port, it MAY be permitted to accept new associations,
1791 * but it SHOULD NOT be permitted to open new
1794 if (ep->base.bind_addr.port < PROT_SOCK &&
1795 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1801 scope = sctp_scope(&to);
1802 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1808 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1814 /* If the SCTP_INIT ancillary data is specified, set all
1815 * the association init values accordingly.
1818 if (sinit->sinit_num_ostreams) {
1819 asoc->c.sinit_num_ostreams =
1820 sinit->sinit_num_ostreams;
1822 if (sinit->sinit_max_instreams) {
1823 asoc->c.sinit_max_instreams =
1824 sinit->sinit_max_instreams;
1826 if (sinit->sinit_max_attempts) {
1827 asoc->max_init_attempts
1828 = sinit->sinit_max_attempts;
1830 if (sinit->sinit_max_init_timeo) {
1831 asoc->max_init_timeo =
1832 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1836 /* Prime the peer's transport structures. */
1837 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1844 /* ASSERT: we have a valid association at this point. */
1845 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1848 /* If the user didn't specify SNDRCVINFO, make up one with
1851 memset(&default_sinfo, 0, sizeof(default_sinfo));
1852 default_sinfo.sinfo_stream = asoc->default_stream;
1853 default_sinfo.sinfo_flags = asoc->default_flags;
1854 default_sinfo.sinfo_ppid = asoc->default_ppid;
1855 default_sinfo.sinfo_context = asoc->default_context;
1856 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1857 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1858 sinfo = &default_sinfo;
1861 /* API 7.1.7, the sndbuf size per association bounds the
1862 * maximum size of data that can be sent in a single send call.
1864 if (msg_len > sk->sk_sndbuf) {
1869 if (asoc->pmtu_pending)
1870 sctp_assoc_pending_pmtu(sk, asoc);
1872 /* If fragmentation is disabled and the message length exceeds the
1873 * association fragmentation point, return EMSGSIZE. The I-D
1874 * does not specify what this error is, but this looks like
1877 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1882 /* Check for invalid stream. */
1883 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1888 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1889 if (!sctp_wspace(asoc)) {
1890 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1895 /* If an address is passed with the sendto/sendmsg call, it is used
1896 * to override the primary destination address in the TCP model, or
1897 * when SCTP_ADDR_OVER flag is set in the UDP model.
1899 if ((sctp_style(sk, TCP) && msg_name) ||
1900 (sinfo_flags & SCTP_ADDR_OVER)) {
1901 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1909 /* Auto-connect, if we aren't connected already. */
1910 if (sctp_state(asoc, CLOSED)) {
1911 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1914 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1917 /* Break the message into multiple chunks of maximum size. */
1918 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1924 /* Now send the (possibly) fragmented message. */
1925 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1926 sctp_chunk_hold(chunk);
1928 /* Do accounting for the write space. */
1929 sctp_set_owner_w(chunk);
1931 chunk->transport = chunk_tp;
1934 /* Send it to the lower layers. Note: all chunks
1935 * must either fail or succeed. The lower layer
1936 * works that way today. Keep it that way or this
1939 err = sctp_primitive_SEND(net, asoc, datamsg);
1940 /* Did the lower layer accept the chunk? */
1942 sctp_datamsg_free(datamsg);
1944 sctp_datamsg_put(datamsg);
1946 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1953 /* If we are already past ASSOCIATE, the lower
1954 * layers are responsible for association cleanup.
1960 sctp_unhash_established(asoc);
1961 sctp_association_free(asoc);
1964 sctp_release_sock(sk);
1967 return sctp_error(sk, msg_flags, err);
1974 err = sock_error(sk);
1984 /* This is an extended version of skb_pull() that removes the data from the
1985 * start of a skb even when data is spread across the list of skb's in the
1986 * frag_list. len specifies the total amount of data that needs to be removed.
1987 * when 'len' bytes could be removed from the skb, it returns 0.
1988 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1989 * could not be removed.
1991 static int sctp_skb_pull(struct sk_buff *skb, int len)
1993 struct sk_buff *list;
1994 int skb_len = skb_headlen(skb);
1997 if (len <= skb_len) {
1998 __skb_pull(skb, len);
2002 __skb_pull(skb, skb_len);
2004 skb_walk_frags(skb, list) {
2005 rlen = sctp_skb_pull(list, len);
2006 skb->len -= (len-rlen);
2007 skb->data_len -= (len-rlen);
2018 /* API 3.1.3 recvmsg() - UDP Style Syntax
2020 * ssize_t recvmsg(int socket, struct msghdr *message,
2023 * socket - the socket descriptor of the endpoint.
2024 * message - pointer to the msghdr structure which contains a single
2025 * user message and possibly some ancillary data.
2027 * See Section 5 for complete description of the data
2030 * flags - flags sent or received with the user message, see Section
2031 * 5 for complete description of the flags.
2033 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2035 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2036 struct msghdr *msg, size_t len, int noblock,
2037 int flags, int *addr_len)
2039 struct sctp_ulpevent *event = NULL;
2040 struct sctp_sock *sp = sctp_sk(sk);
2041 struct sk_buff *skb;
2046 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
2047 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
2048 "len", len, "knoblauch", noblock,
2049 "flags", flags, "addr_len", addr_len);
2053 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2058 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2062 /* Get the total length of the skb including any skb's in the
2071 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2073 event = sctp_skb2event(skb);
2078 sock_recv_ts_and_drops(msg, sk, skb);
2079 if (sctp_ulpevent_is_notification(event)) {
2080 msg->msg_flags |= MSG_NOTIFICATION;
2081 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2083 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2086 /* Check if we allow SCTP_SNDRCVINFO. */
2087 if (sp->subscribe.sctp_data_io_event)
2088 sctp_ulpevent_read_sndrcvinfo(event, msg);
2090 /* FIXME: we should be calling IP/IPv6 layers. */
2091 if (sk->sk_protinfo.af_inet.cmsg_flags)
2092 ip_cmsg_recv(msg, skb);
2097 /* If skb's length exceeds the user's buffer, update the skb and
2098 * push it back to the receive_queue so that the next call to
2099 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2101 if (skb_len > copied) {
2102 msg->msg_flags &= ~MSG_EOR;
2103 if (flags & MSG_PEEK)
2105 sctp_skb_pull(skb, copied);
2106 skb_queue_head(&sk->sk_receive_queue, skb);
2108 /* When only partial message is copied to the user, increase
2109 * rwnd by that amount. If all the data in the skb is read,
2110 * rwnd is updated when the event is freed.
2112 if (!sctp_ulpevent_is_notification(event))
2113 sctp_assoc_rwnd_increase(event->asoc, copied);
2115 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2116 (event->msg_flags & MSG_EOR))
2117 msg->msg_flags |= MSG_EOR;
2119 msg->msg_flags &= ~MSG_EOR;
2122 if (flags & MSG_PEEK) {
2123 /* Release the skb reference acquired after peeking the skb in
2124 * sctp_skb_recv_datagram().
2128 /* Free the event which includes releasing the reference to
2129 * the owner of the skb, freeing the skb and updating the
2132 sctp_ulpevent_free(event);
2135 sctp_release_sock(sk);
2139 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2141 * This option is a on/off flag. If enabled no SCTP message
2142 * fragmentation will be performed. Instead if a message being sent
2143 * exceeds the current PMTU size, the message will NOT be sent and
2144 * instead a error will be indicated to the user.
2146 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2147 char __user *optval,
2148 unsigned int optlen)
2152 if (optlen < sizeof(int))
2155 if (get_user(val, (int __user *)optval))
2158 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2163 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2164 unsigned int optlen)
2166 struct sctp_association *asoc;
2167 struct sctp_ulpevent *event;
2169 if (optlen > sizeof(struct sctp_event_subscribe))
2171 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2175 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2176 * if there is no data to be sent or retransmit, the stack will
2177 * immediately send up this notification.
2179 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2180 &sctp_sk(sk)->subscribe)) {
2181 asoc = sctp_id2assoc(sk, 0);
2183 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2184 event = sctp_ulpevent_make_sender_dry_event(asoc,
2189 sctp_ulpq_tail_event(&asoc->ulpq, event);
2196 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2198 * This socket option is applicable to the UDP-style socket only. When
2199 * set it will cause associations that are idle for more than the
2200 * specified number of seconds to automatically close. An association
2201 * being idle is defined an association that has NOT sent or received
2202 * user data. The special value of '0' indicates that no automatic
2203 * close of any associations should be performed. The option expects an
2204 * integer defining the number of seconds of idle time before an
2205 * association is closed.
2207 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2208 unsigned int optlen)
2210 struct sctp_sock *sp = sctp_sk(sk);
2212 /* Applicable to UDP-style socket only */
2213 if (sctp_style(sk, TCP))
2215 if (optlen != sizeof(int))
2217 if (copy_from_user(&sp->autoclose, optval, optlen))
2223 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2225 * Applications can enable or disable heartbeats for any peer address of
2226 * an association, modify an address's heartbeat interval, force a
2227 * heartbeat to be sent immediately, and adjust the address's maximum
2228 * number of retransmissions sent before an address is considered
2229 * unreachable. The following structure is used to access and modify an
2230 * address's parameters:
2232 * struct sctp_paddrparams {
2233 * sctp_assoc_t spp_assoc_id;
2234 * struct sockaddr_storage spp_address;
2235 * uint32_t spp_hbinterval;
2236 * uint16_t spp_pathmaxrxt;
2237 * uint32_t spp_pathmtu;
2238 * uint32_t spp_sackdelay;
2239 * uint32_t spp_flags;
2242 * spp_assoc_id - (one-to-many style socket) This is filled in the
2243 * application, and identifies the association for
2245 * spp_address - This specifies which address is of interest.
2246 * spp_hbinterval - This contains the value of the heartbeat interval,
2247 * in milliseconds. If a value of zero
2248 * is present in this field then no changes are to
2249 * be made to this parameter.
2250 * spp_pathmaxrxt - This contains the maximum number of
2251 * retransmissions before this address shall be
2252 * considered unreachable. If a value of zero
2253 * is present in this field then no changes are to
2254 * be made to this parameter.
2255 * spp_pathmtu - When Path MTU discovery is disabled the value
2256 * specified here will be the "fixed" path mtu.
2257 * Note that if the spp_address field is empty
2258 * then all associations on this address will
2259 * have this fixed path mtu set upon them.
2261 * spp_sackdelay - When delayed sack is enabled, this value specifies
2262 * the number of milliseconds that sacks will be delayed
2263 * for. This value will apply to all addresses of an
2264 * association if the spp_address field is empty. Note
2265 * also, that if delayed sack is enabled and this
2266 * value is set to 0, no change is made to the last
2267 * recorded delayed sack timer value.
2269 * spp_flags - These flags are used to control various features
2270 * on an association. The flag field may contain
2271 * zero or more of the following options.
2273 * SPP_HB_ENABLE - Enable heartbeats on the
2274 * specified address. Note that if the address
2275 * field is empty all addresses for the association
2276 * have heartbeats enabled upon them.
2278 * SPP_HB_DISABLE - Disable heartbeats on the
2279 * speicifed address. Note that if the address
2280 * field is empty all addresses for the association
2281 * will have their heartbeats disabled. Note also
2282 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2283 * mutually exclusive, only one of these two should
2284 * be specified. Enabling both fields will have
2285 * undetermined results.
2287 * SPP_HB_DEMAND - Request a user initiated heartbeat
2288 * to be made immediately.
2290 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2291 * heartbeat delayis to be set to the value of 0
2294 * SPP_PMTUD_ENABLE - This field will enable PMTU
2295 * discovery upon the specified address. Note that
2296 * if the address feild is empty then all addresses
2297 * on the association are effected.
2299 * SPP_PMTUD_DISABLE - This field will disable PMTU
2300 * discovery upon the specified address. Note that
2301 * if the address feild is empty then all addresses
2302 * on the association are effected. Not also that
2303 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2304 * exclusive. Enabling both will have undetermined
2307 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2308 * on delayed sack. The time specified in spp_sackdelay
2309 * is used to specify the sack delay for this address. Note
2310 * that if spp_address is empty then all addresses will
2311 * enable delayed sack and take on the sack delay
2312 * value specified in spp_sackdelay.
2313 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2314 * off delayed sack. If the spp_address field is blank then
2315 * delayed sack is disabled for the entire association. Note
2316 * also that this field is mutually exclusive to
2317 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2320 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2321 struct sctp_transport *trans,
2322 struct sctp_association *asoc,
2323 struct sctp_sock *sp,
2326 int sackdelay_change)
2330 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2331 struct net *net = sock_net(trans->asoc->base.sk);
2333 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2338 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2339 * this field is ignored. Note also that a value of zero indicates
2340 * the current setting should be left unchanged.
2342 if (params->spp_flags & SPP_HB_ENABLE) {
2344 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2345 * set. This lets us use 0 value when this flag
2348 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2349 params->spp_hbinterval = 0;
2351 if (params->spp_hbinterval ||
2352 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2355 msecs_to_jiffies(params->spp_hbinterval);
2358 msecs_to_jiffies(params->spp_hbinterval);
2360 sp->hbinterval = params->spp_hbinterval;
2367 trans->param_flags =
2368 (trans->param_flags & ~SPP_HB) | hb_change;
2371 (asoc->param_flags & ~SPP_HB) | hb_change;
2374 (sp->param_flags & ~SPP_HB) | hb_change;
2378 /* When Path MTU discovery is disabled the value specified here will
2379 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2380 * include the flag SPP_PMTUD_DISABLE for this field to have any
2383 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2385 trans->pathmtu = params->spp_pathmtu;
2386 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2388 asoc->pathmtu = params->spp_pathmtu;
2389 sctp_frag_point(asoc, params->spp_pathmtu);
2391 sp->pathmtu = params->spp_pathmtu;
2397 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2398 (params->spp_flags & SPP_PMTUD_ENABLE);
2399 trans->param_flags =
2400 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2402 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2403 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2407 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2410 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2414 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2415 * value of this field is ignored. Note also that a value of zero
2416 * indicates the current setting should be left unchanged.
2418 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2421 msecs_to_jiffies(params->spp_sackdelay);
2424 msecs_to_jiffies(params->spp_sackdelay);
2426 sp->sackdelay = params->spp_sackdelay;
2430 if (sackdelay_change) {
2432 trans->param_flags =
2433 (trans->param_flags & ~SPP_SACKDELAY) |
2437 (asoc->param_flags & ~SPP_SACKDELAY) |
2441 (sp->param_flags & ~SPP_SACKDELAY) |
2446 /* Note that a value of zero indicates the current setting should be
2449 if (params->spp_pathmaxrxt) {
2451 trans->pathmaxrxt = params->spp_pathmaxrxt;
2453 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2455 sp->pathmaxrxt = params->spp_pathmaxrxt;
2462 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2463 char __user *optval,
2464 unsigned int optlen)
2466 struct sctp_paddrparams params;
2467 struct sctp_transport *trans = NULL;
2468 struct sctp_association *asoc = NULL;
2469 struct sctp_sock *sp = sctp_sk(sk);
2471 int hb_change, pmtud_change, sackdelay_change;
2473 if (optlen != sizeof(struct sctp_paddrparams))
2476 if (copy_from_user(¶ms, optval, optlen))
2479 /* Validate flags and value parameters. */
2480 hb_change = params.spp_flags & SPP_HB;
2481 pmtud_change = params.spp_flags & SPP_PMTUD;
2482 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2484 if (hb_change == SPP_HB ||
2485 pmtud_change == SPP_PMTUD ||
2486 sackdelay_change == SPP_SACKDELAY ||
2487 params.spp_sackdelay > 500 ||
2488 (params.spp_pathmtu &&
2489 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2492 /* If an address other than INADDR_ANY is specified, and
2493 * no transport is found, then the request is invalid.
2495 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
2496 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2497 params.spp_assoc_id);
2502 /* Get association, if assoc_id != 0 and the socket is a one
2503 * to many style socket, and an association was not found, then
2504 * the id was invalid.
2506 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2507 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2510 /* Heartbeat demand can only be sent on a transport or
2511 * association, but not a socket.
2513 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2516 /* Process parameters. */
2517 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2518 hb_change, pmtud_change,
2524 /* If changes are for association, also apply parameters to each
2527 if (!trans && asoc) {
2528 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2530 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2531 hb_change, pmtud_change,
2540 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2542 * This option will effect the way delayed acks are performed. This
2543 * option allows you to get or set the delayed ack time, in
2544 * milliseconds. It also allows changing the delayed ack frequency.
2545 * Changing the frequency to 1 disables the delayed sack algorithm. If
2546 * the assoc_id is 0, then this sets or gets the endpoints default
2547 * values. If the assoc_id field is non-zero, then the set or get
2548 * effects the specified association for the one to many model (the
2549 * assoc_id field is ignored by the one to one model). Note that if
2550 * sack_delay or sack_freq are 0 when setting this option, then the
2551 * current values will remain unchanged.
2553 * struct sctp_sack_info {
2554 * sctp_assoc_t sack_assoc_id;
2555 * uint32_t sack_delay;
2556 * uint32_t sack_freq;
2559 * sack_assoc_id - This parameter, indicates which association the user
2560 * is performing an action upon. Note that if this field's value is
2561 * zero then the endpoints default value is changed (effecting future
2562 * associations only).
2564 * sack_delay - This parameter contains the number of milliseconds that
2565 * the user is requesting the delayed ACK timer be set to. Note that
2566 * this value is defined in the standard to be between 200 and 500
2569 * sack_freq - This parameter contains the number of packets that must
2570 * be received before a sack is sent without waiting for the delay
2571 * timer to expire. The default value for this is 2, setting this
2572 * value to 1 will disable the delayed sack algorithm.
2575 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2576 char __user *optval, unsigned int optlen)
2578 struct sctp_sack_info params;
2579 struct sctp_transport *trans = NULL;
2580 struct sctp_association *asoc = NULL;
2581 struct sctp_sock *sp = sctp_sk(sk);
2583 if (optlen == sizeof(struct sctp_sack_info)) {
2584 if (copy_from_user(¶ms, optval, optlen))
2587 if (params.sack_delay == 0 && params.sack_freq == 0)
2589 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2590 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2591 pr_warn("Use struct sctp_sack_info instead\n");
2592 if (copy_from_user(¶ms, optval, optlen))
2595 if (params.sack_delay == 0)
2596 params.sack_freq = 1;
2598 params.sack_freq = 0;
2602 /* Validate value parameter. */
2603 if (params.sack_delay > 500)
2606 /* Get association, if sack_assoc_id != 0 and the socket is a one
2607 * to many style socket, and an association was not found, then
2608 * the id was invalid.
2610 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2611 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2614 if (params.sack_delay) {
2617 msecs_to_jiffies(params.sack_delay);
2619 (asoc->param_flags & ~SPP_SACKDELAY) |
2620 SPP_SACKDELAY_ENABLE;
2622 sp->sackdelay = params.sack_delay;
2624 (sp->param_flags & ~SPP_SACKDELAY) |
2625 SPP_SACKDELAY_ENABLE;
2629 if (params.sack_freq == 1) {
2632 (asoc->param_flags & ~SPP_SACKDELAY) |
2633 SPP_SACKDELAY_DISABLE;
2636 (sp->param_flags & ~SPP_SACKDELAY) |
2637 SPP_SACKDELAY_DISABLE;
2639 } else if (params.sack_freq > 1) {
2641 asoc->sackfreq = params.sack_freq;
2643 (asoc->param_flags & ~SPP_SACKDELAY) |
2644 SPP_SACKDELAY_ENABLE;
2646 sp->sackfreq = params.sack_freq;
2648 (sp->param_flags & ~SPP_SACKDELAY) |
2649 SPP_SACKDELAY_ENABLE;
2653 /* If change is for association, also apply to each transport. */
2655 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2657 if (params.sack_delay) {
2659 msecs_to_jiffies(params.sack_delay);
2660 trans->param_flags =
2661 (trans->param_flags & ~SPP_SACKDELAY) |
2662 SPP_SACKDELAY_ENABLE;
2664 if (params.sack_freq == 1) {
2665 trans->param_flags =
2666 (trans->param_flags & ~SPP_SACKDELAY) |
2667 SPP_SACKDELAY_DISABLE;
2668 } else if (params.sack_freq > 1) {
2669 trans->sackfreq = params.sack_freq;
2670 trans->param_flags =
2671 (trans->param_flags & ~SPP_SACKDELAY) |
2672 SPP_SACKDELAY_ENABLE;
2680 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2682 * Applications can specify protocol parameters for the default association
2683 * initialization. The option name argument to setsockopt() and getsockopt()
2686 * Setting initialization parameters is effective only on an unconnected
2687 * socket (for UDP-style sockets only future associations are effected
2688 * by the change). With TCP-style sockets, this option is inherited by
2689 * sockets derived from a listener socket.
2691 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2693 struct sctp_initmsg sinit;
2694 struct sctp_sock *sp = sctp_sk(sk);
2696 if (optlen != sizeof(struct sctp_initmsg))
2698 if (copy_from_user(&sinit, optval, optlen))
2701 if (sinit.sinit_num_ostreams)
2702 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2703 if (sinit.sinit_max_instreams)
2704 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2705 if (sinit.sinit_max_attempts)
2706 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2707 if (sinit.sinit_max_init_timeo)
2708 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2714 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2716 * Applications that wish to use the sendto() system call may wish to
2717 * specify a default set of parameters that would normally be supplied
2718 * through the inclusion of ancillary data. This socket option allows
2719 * such an application to set the default sctp_sndrcvinfo structure.
2720 * The application that wishes to use this socket option simply passes
2721 * in to this call the sctp_sndrcvinfo structure defined in Section
2722 * 5.2.2) The input parameters accepted by this call include
2723 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2724 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2725 * to this call if the caller is using the UDP model.
2727 static int sctp_setsockopt_default_send_param(struct sock *sk,
2728 char __user *optval,
2729 unsigned int optlen)
2731 struct sctp_sndrcvinfo info;
2732 struct sctp_association *asoc;
2733 struct sctp_sock *sp = sctp_sk(sk);
2735 if (optlen != sizeof(struct sctp_sndrcvinfo))
2737 if (copy_from_user(&info, optval, optlen))
2740 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2741 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2745 asoc->default_stream = info.sinfo_stream;
2746 asoc->default_flags = info.sinfo_flags;
2747 asoc->default_ppid = info.sinfo_ppid;
2748 asoc->default_context = info.sinfo_context;
2749 asoc->default_timetolive = info.sinfo_timetolive;
2751 sp->default_stream = info.sinfo_stream;
2752 sp->default_flags = info.sinfo_flags;
2753 sp->default_ppid = info.sinfo_ppid;
2754 sp->default_context = info.sinfo_context;
2755 sp->default_timetolive = info.sinfo_timetolive;
2761 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2763 * Requests that the local SCTP stack use the enclosed peer address as
2764 * the association primary. The enclosed address must be one of the
2765 * association peer's addresses.
2767 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2768 unsigned int optlen)
2770 struct sctp_prim prim;
2771 struct sctp_transport *trans;
2773 if (optlen != sizeof(struct sctp_prim))
2776 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2779 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2783 sctp_assoc_set_primary(trans->asoc, trans);
2789 * 7.1.5 SCTP_NODELAY
2791 * Turn on/off any Nagle-like algorithm. This means that packets are
2792 * generally sent as soon as possible and no unnecessary delays are
2793 * introduced, at the cost of more packets in the network. Expects an
2794 * integer boolean flag.
2796 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2797 unsigned int optlen)
2801 if (optlen < sizeof(int))
2803 if (get_user(val, (int __user *)optval))
2806 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2812 * 7.1.1 SCTP_RTOINFO
2814 * The protocol parameters used to initialize and bound retransmission
2815 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2816 * and modify these parameters.
2817 * All parameters are time values, in milliseconds. A value of 0, when
2818 * modifying the parameters, indicates that the current value should not
2822 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2824 struct sctp_rtoinfo rtoinfo;
2825 struct sctp_association *asoc;
2827 if (optlen != sizeof (struct sctp_rtoinfo))
2830 if (copy_from_user(&rtoinfo, optval, optlen))
2833 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2835 /* Set the values to the specific association */
2836 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2840 if (rtoinfo.srto_initial != 0)
2842 msecs_to_jiffies(rtoinfo.srto_initial);
2843 if (rtoinfo.srto_max != 0)
2844 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2845 if (rtoinfo.srto_min != 0)
2846 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2848 /* If there is no association or the association-id = 0
2849 * set the values to the endpoint.
2851 struct sctp_sock *sp = sctp_sk(sk);
2853 if (rtoinfo.srto_initial != 0)
2854 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2855 if (rtoinfo.srto_max != 0)
2856 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2857 if (rtoinfo.srto_min != 0)
2858 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2866 * 7.1.2 SCTP_ASSOCINFO
2868 * This option is used to tune the maximum retransmission attempts
2869 * of the association.
2870 * Returns an error if the new association retransmission value is
2871 * greater than the sum of the retransmission value of the peer.
2872 * See [SCTP] for more information.
2875 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2878 struct sctp_assocparams assocparams;
2879 struct sctp_association *asoc;
2881 if (optlen != sizeof(struct sctp_assocparams))
2883 if (copy_from_user(&assocparams, optval, optlen))
2886 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2888 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2891 /* Set the values to the specific association */
2893 if (assocparams.sasoc_asocmaxrxt != 0) {
2896 struct sctp_transport *peer_addr;
2898 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2900 path_sum += peer_addr->pathmaxrxt;
2904 /* Only validate asocmaxrxt if we have more than
2905 * one path/transport. We do this because path
2906 * retransmissions are only counted when we have more
2910 assocparams.sasoc_asocmaxrxt > path_sum)
2913 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2916 if (assocparams.sasoc_cookie_life != 0) {
2917 asoc->cookie_life.tv_sec =
2918 assocparams.sasoc_cookie_life / 1000;
2919 asoc->cookie_life.tv_usec =
2920 (assocparams.sasoc_cookie_life % 1000)
2924 /* Set the values to the endpoint */
2925 struct sctp_sock *sp = sctp_sk(sk);
2927 if (assocparams.sasoc_asocmaxrxt != 0)
2928 sp->assocparams.sasoc_asocmaxrxt =
2929 assocparams.sasoc_asocmaxrxt;
2930 if (assocparams.sasoc_cookie_life != 0)
2931 sp->assocparams.sasoc_cookie_life =
2932 assocparams.sasoc_cookie_life;
2938 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2940 * This socket option is a boolean flag which turns on or off mapped V4
2941 * addresses. If this option is turned on and the socket is type
2942 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2943 * If this option is turned off, then no mapping will be done of V4
2944 * addresses and a user will receive both PF_INET6 and PF_INET type
2945 * addresses on the socket.
2947 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2950 struct sctp_sock *sp = sctp_sk(sk);
2952 if (optlen < sizeof(int))
2954 if (get_user(val, (int __user *)optval))
2965 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2966 * This option will get or set the maximum size to put in any outgoing
2967 * SCTP DATA chunk. If a message is larger than this size it will be
2968 * fragmented by SCTP into the specified size. Note that the underlying
2969 * SCTP implementation may fragment into smaller sized chunks when the
2970 * PMTU of the underlying association is smaller than the value set by
2971 * the user. The default value for this option is '0' which indicates
2972 * the user is NOT limiting fragmentation and only the PMTU will effect
2973 * SCTP's choice of DATA chunk size. Note also that values set larger
2974 * than the maximum size of an IP datagram will effectively let SCTP
2975 * control fragmentation (i.e. the same as setting this option to 0).
2977 * The following structure is used to access and modify this parameter:
2979 * struct sctp_assoc_value {
2980 * sctp_assoc_t assoc_id;
2981 * uint32_t assoc_value;
2984 * assoc_id: This parameter is ignored for one-to-one style sockets.
2985 * For one-to-many style sockets this parameter indicates which
2986 * association the user is performing an action upon. Note that if
2987 * this field's value is zero then the endpoints default value is
2988 * changed (effecting future associations only).
2989 * assoc_value: This parameter specifies the maximum size in bytes.
2991 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2993 struct sctp_assoc_value params;
2994 struct sctp_association *asoc;
2995 struct sctp_sock *sp = sctp_sk(sk);
2998 if (optlen == sizeof(int)) {
2999 pr_warn("Use of int in maxseg socket option deprecated\n");
3000 pr_warn("Use struct sctp_assoc_value instead\n");
3001 if (copy_from_user(&val, optval, optlen))
3003 params.assoc_id = 0;
3004 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3005 if (copy_from_user(¶ms, optval, optlen))
3007 val = params.assoc_value;
3011 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3014 asoc = sctp_id2assoc(sk, params.assoc_id);
3015 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3020 val = asoc->pathmtu;
3021 val -= sp->pf->af->net_header_len;
3022 val -= sizeof(struct sctphdr) +
3023 sizeof(struct sctp_data_chunk);
3025 asoc->user_frag = val;
3026 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3028 sp->user_frag = val;
3036 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3038 * Requests that the peer mark the enclosed address as the association
3039 * primary. The enclosed address must be one of the association's
3040 * locally bound addresses. The following structure is used to make a
3041 * set primary request:
3043 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3044 unsigned int optlen)
3046 struct net *net = sock_net(sk);
3047 struct sctp_sock *sp;
3048 struct sctp_association *asoc = NULL;
3049 struct sctp_setpeerprim prim;
3050 struct sctp_chunk *chunk;
3056 if (!net->sctp.addip_enable)
3059 if (optlen != sizeof(struct sctp_setpeerprim))
3062 if (copy_from_user(&prim, optval, optlen))
3065 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3069 if (!asoc->peer.asconf_capable)
3072 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3075 if (!sctp_state(asoc, ESTABLISHED))
3078 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3082 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3083 return -EADDRNOTAVAIL;
3085 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3086 return -EADDRNOTAVAIL;
3088 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3089 chunk = sctp_make_asconf_set_prim(asoc,
3090 (union sctp_addr *)&prim.sspp_addr);
3094 err = sctp_send_asconf(asoc, chunk);
3096 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
3101 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3102 unsigned int optlen)
3104 struct sctp_setadaptation adaptation;
3106 if (optlen != sizeof(struct sctp_setadaptation))
3108 if (copy_from_user(&adaptation, optval, optlen))
3111 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3117 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3119 * The context field in the sctp_sndrcvinfo structure is normally only
3120 * used when a failed message is retrieved holding the value that was
3121 * sent down on the actual send call. This option allows the setting of
3122 * a default context on an association basis that will be received on
3123 * reading messages from the peer. This is especially helpful in the
3124 * one-2-many model for an application to keep some reference to an
3125 * internal state machine that is processing messages on the
3126 * association. Note that the setting of this value only effects
3127 * received messages from the peer and does not effect the value that is
3128 * saved with outbound messages.
3130 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3131 unsigned int optlen)
3133 struct sctp_assoc_value params;
3134 struct sctp_sock *sp;
3135 struct sctp_association *asoc;
3137 if (optlen != sizeof(struct sctp_assoc_value))
3139 if (copy_from_user(¶ms, optval, optlen))
3144 if (params.assoc_id != 0) {
3145 asoc = sctp_id2assoc(sk, params.assoc_id);
3148 asoc->default_rcv_context = params.assoc_value;
3150 sp->default_rcv_context = params.assoc_value;
3157 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3159 * This options will at a minimum specify if the implementation is doing
3160 * fragmented interleave. Fragmented interleave, for a one to many
3161 * socket, is when subsequent calls to receive a message may return
3162 * parts of messages from different associations. Some implementations
3163 * may allow you to turn this value on or off. If so, when turned off,
3164 * no fragment interleave will occur (which will cause a head of line
3165 * blocking amongst multiple associations sharing the same one to many
3166 * socket). When this option is turned on, then each receive call may
3167 * come from a different association (thus the user must receive data
3168 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3169 * association each receive belongs to.
3171 * This option takes a boolean value. A non-zero value indicates that
3172 * fragmented interleave is on. A value of zero indicates that
3173 * fragmented interleave is off.
3175 * Note that it is important that an implementation that allows this
3176 * option to be turned on, have it off by default. Otherwise an unaware
3177 * application using the one to many model may become confused and act
3180 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3181 char __user *optval,
3182 unsigned int optlen)
3186 if (optlen != sizeof(int))
3188 if (get_user(val, (int __user *)optval))
3191 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3197 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3198 * (SCTP_PARTIAL_DELIVERY_POINT)
3200 * This option will set or get the SCTP partial delivery point. This
3201 * point is the size of a message where the partial delivery API will be
3202 * invoked to help free up rwnd space for the peer. Setting this to a
3203 * lower value will cause partial deliveries to happen more often. The
3204 * calls argument is an integer that sets or gets the partial delivery
3205 * point. Note also that the call will fail if the user attempts to set
3206 * this value larger than the socket receive buffer size.
3208 * Note that any single message having a length smaller than or equal to
3209 * the SCTP partial delivery point will be delivered in one single read
3210 * call as long as the user provided buffer is large enough to hold the
3213 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3214 char __user *optval,
3215 unsigned int optlen)
3219 if (optlen != sizeof(u32))
3221 if (get_user(val, (int __user *)optval))
3224 /* Note: We double the receive buffer from what the user sets
3225 * it to be, also initial rwnd is based on rcvbuf/2.
3227 if (val > (sk->sk_rcvbuf >> 1))
3230 sctp_sk(sk)->pd_point = val;
3232 return 0; /* is this the right error code? */
3236 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3238 * This option will allow a user to change the maximum burst of packets
3239 * that can be emitted by this association. Note that the default value
3240 * is 4, and some implementations may restrict this setting so that it
3241 * can only be lowered.
3243 * NOTE: This text doesn't seem right. Do this on a socket basis with
3244 * future associations inheriting the socket value.
3246 static int sctp_setsockopt_maxburst(struct sock *sk,
3247 char __user *optval,
3248 unsigned int optlen)
3250 struct sctp_assoc_value params;
3251 struct sctp_sock *sp;
3252 struct sctp_association *asoc;
3256 if (optlen == sizeof(int)) {
3257 pr_warn("Use of int in max_burst socket option deprecated\n");
3258 pr_warn("Use struct sctp_assoc_value instead\n");
3259 if (copy_from_user(&val, optval, optlen))
3261 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3262 if (copy_from_user(¶ms, optval, optlen))
3264 val = params.assoc_value;
3265 assoc_id = params.assoc_id;
3271 if (assoc_id != 0) {
3272 asoc = sctp_id2assoc(sk, assoc_id);
3275 asoc->max_burst = val;
3277 sp->max_burst = val;
3283 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3285 * This set option adds a chunk type that the user is requesting to be
3286 * received only in an authenticated way. Changes to the list of chunks
3287 * will only effect future associations on the socket.
3289 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3290 char __user *optval,
3291 unsigned int optlen)
3293 struct net *net = sock_net(sk);
3294 struct sctp_authchunk val;
3296 if (!net->sctp.auth_enable)
3299 if (optlen != sizeof(struct sctp_authchunk))
3301 if (copy_from_user(&val, optval, optlen))
3304 switch (val.sauth_chunk) {
3306 case SCTP_CID_INIT_ACK:
3307 case SCTP_CID_SHUTDOWN_COMPLETE:
3312 /* add this chunk id to the endpoint */
3313 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3317 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3319 * This option gets or sets the list of HMAC algorithms that the local
3320 * endpoint requires the peer to use.
3322 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3323 char __user *optval,
3324 unsigned int optlen)
3326 struct net *net = sock_net(sk);
3327 struct sctp_hmacalgo *hmacs;
3331 if (!net->sctp.auth_enable)
3334 if (optlen < sizeof(struct sctp_hmacalgo))
3337 hmacs= memdup_user(optval, optlen);
3339 return PTR_ERR(hmacs);
3341 idents = hmacs->shmac_num_idents;
3342 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3343 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3348 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3355 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3357 * This option will set a shared secret key which is used to build an
3358 * association shared key.
3360 static int sctp_setsockopt_auth_key(struct sock *sk,
3361 char __user *optval,
3362 unsigned int optlen)
3364 struct net *net = sock_net(sk);
3365 struct sctp_authkey *authkey;
3366 struct sctp_association *asoc;
3369 if (!net->sctp.auth_enable)
3372 if (optlen <= sizeof(struct sctp_authkey))
3375 authkey= memdup_user(optval, optlen);
3376 if (IS_ERR(authkey))
3377 return PTR_ERR(authkey);
3379 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3384 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3385 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3390 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3397 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3399 * This option will get or set the active shared key to be used to build
3400 * the association shared key.
3402 static int sctp_setsockopt_active_key(struct sock *sk,
3403 char __user *optval,
3404 unsigned int optlen)
3406 struct net *net = sock_net(sk);
3407 struct sctp_authkeyid val;
3408 struct sctp_association *asoc;
3410 if (!net->sctp.auth_enable)
3413 if (optlen != sizeof(struct sctp_authkeyid))
3415 if (copy_from_user(&val, optval, optlen))
3418 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3419 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3422 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3423 val.scact_keynumber);
3427 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3429 * This set option will delete a shared secret key from use.
3431 static int sctp_setsockopt_del_key(struct sock *sk,
3432 char __user *optval,
3433 unsigned int optlen)
3435 struct net *net = sock_net(sk);
3436 struct sctp_authkeyid val;
3437 struct sctp_association *asoc;
3439 if (!net->sctp.auth_enable)
3442 if (optlen != sizeof(struct sctp_authkeyid))
3444 if (copy_from_user(&val, optval, optlen))
3447 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3448 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3451 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3452 val.scact_keynumber);
3457 * 8.1.23 SCTP_AUTO_ASCONF
3459 * This option will enable or disable the use of the automatic generation of
3460 * ASCONF chunks to add and delete addresses to an existing association. Note
3461 * that this option has two caveats namely: a) it only affects sockets that
3462 * are bound to all addresses available to the SCTP stack, and b) the system
3463 * administrator may have an overriding control that turns the ASCONF feature
3464 * off no matter what setting the socket option may have.
3465 * This option expects an integer boolean flag, where a non-zero value turns on
3466 * the option, and a zero value turns off the option.
3467 * Note. In this implementation, socket operation overrides default parameter
3468 * being set by sysctl as well as FreeBSD implementation
3470 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3471 unsigned int optlen)
3474 struct sctp_sock *sp = sctp_sk(sk);
3476 if (optlen < sizeof(int))
3478 if (get_user(val, (int __user *)optval))
3480 if (!sctp_is_ep_boundall(sk) && val)
3482 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3485 if (val == 0 && sp->do_auto_asconf) {
3486 list_del(&sp->auto_asconf_list);
3487 sp->do_auto_asconf = 0;
3488 } else if (val && !sp->do_auto_asconf) {
3489 list_add_tail(&sp->auto_asconf_list,
3490 &sock_net(sk)->sctp.auto_asconf_splist);
3491 sp->do_auto_asconf = 1;
3498 * SCTP_PEER_ADDR_THLDS
3500 * This option allows us to alter the partially failed threshold for one or all
3501 * transports in an association. See Section 6.1 of:
3502 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3504 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3505 char __user *optval,
3506 unsigned int optlen)
3508 struct sctp_paddrthlds val;
3509 struct sctp_transport *trans;
3510 struct sctp_association *asoc;
3512 if (optlen < sizeof(struct sctp_paddrthlds))
3514 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3515 sizeof(struct sctp_paddrthlds)))
3519 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3520 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3523 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3525 if (val.spt_pathmaxrxt)
3526 trans->pathmaxrxt = val.spt_pathmaxrxt;
3527 trans->pf_retrans = val.spt_pathpfthld;
3530 if (val.spt_pathmaxrxt)
3531 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3532 asoc->pf_retrans = val.spt_pathpfthld;
3534 trans = sctp_addr_id2transport(sk, &val.spt_address,
3539 if (val.spt_pathmaxrxt)
3540 trans->pathmaxrxt = val.spt_pathmaxrxt;
3541 trans->pf_retrans = val.spt_pathpfthld;
3547 /* API 6.2 setsockopt(), getsockopt()
3549 * Applications use setsockopt() and getsockopt() to set or retrieve
3550 * socket options. Socket options are used to change the default
3551 * behavior of sockets calls. They are described in Section 7.
3555 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3556 * int __user *optlen);
3557 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3560 * sd - the socket descript.
3561 * level - set to IPPROTO_SCTP for all SCTP options.
3562 * optname - the option name.
3563 * optval - the buffer to store the value of the option.
3564 * optlen - the size of the buffer.
3566 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3567 char __user *optval, unsigned int optlen)
3571 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3574 /* I can hardly begin to describe how wrong this is. This is
3575 * so broken as to be worse than useless. The API draft
3576 * REALLY is NOT helpful here... I am not convinced that the
3577 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3578 * are at all well-founded.
3580 if (level != SOL_SCTP) {
3581 struct sctp_af *af = sctp_sk(sk)->pf->af;
3582 retval = af->setsockopt(sk, level, optname, optval, optlen);
3589 case SCTP_SOCKOPT_BINDX_ADD:
3590 /* 'optlen' is the size of the addresses buffer. */
3591 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3592 optlen, SCTP_BINDX_ADD_ADDR);
3595 case SCTP_SOCKOPT_BINDX_REM:
3596 /* 'optlen' is the size of the addresses buffer. */
3597 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3598 optlen, SCTP_BINDX_REM_ADDR);
3601 case SCTP_SOCKOPT_CONNECTX_OLD:
3602 /* 'optlen' is the size of the addresses buffer. */
3603 retval = sctp_setsockopt_connectx_old(sk,
3604 (struct sockaddr __user *)optval,
3608 case SCTP_SOCKOPT_CONNECTX:
3609 /* 'optlen' is the size of the addresses buffer. */
3610 retval = sctp_setsockopt_connectx(sk,
3611 (struct sockaddr __user *)optval,
3615 case SCTP_DISABLE_FRAGMENTS:
3616 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3620 retval = sctp_setsockopt_events(sk, optval, optlen);
3623 case SCTP_AUTOCLOSE:
3624 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3627 case SCTP_PEER_ADDR_PARAMS:
3628 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3631 case SCTP_DELAYED_SACK:
3632 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3634 case SCTP_PARTIAL_DELIVERY_POINT:
3635 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3639 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3641 case SCTP_DEFAULT_SEND_PARAM:
3642 retval = sctp_setsockopt_default_send_param(sk, optval,
3645 case SCTP_PRIMARY_ADDR:
3646 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3648 case SCTP_SET_PEER_PRIMARY_ADDR:
3649 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3652 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3655 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3657 case SCTP_ASSOCINFO:
3658 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3660 case SCTP_I_WANT_MAPPED_V4_ADDR:
3661 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3664 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3666 case SCTP_ADAPTATION_LAYER:
3667 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3670 retval = sctp_setsockopt_context(sk, optval, optlen);
3672 case SCTP_FRAGMENT_INTERLEAVE:
3673 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3675 case SCTP_MAX_BURST:
3676 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3678 case SCTP_AUTH_CHUNK:
3679 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3681 case SCTP_HMAC_IDENT:
3682 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3685 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3687 case SCTP_AUTH_ACTIVE_KEY:
3688 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3690 case SCTP_AUTH_DELETE_KEY:
3691 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3693 case SCTP_AUTO_ASCONF:
3694 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3696 case SCTP_PEER_ADDR_THLDS:
3697 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3700 retval = -ENOPROTOOPT;
3704 sctp_release_sock(sk);
3710 /* API 3.1.6 connect() - UDP Style Syntax
3712 * An application may use the connect() call in the UDP model to initiate an
3713 * association without sending data.
3717 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3719 * sd: the socket descriptor to have a new association added to.
3721 * nam: the address structure (either struct sockaddr_in or struct
3722 * sockaddr_in6 defined in RFC2553 [7]).
3724 * len: the size of the address.
3726 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3734 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3735 __func__, sk, addr, addr_len);
3737 /* Validate addr_len before calling common connect/connectx routine. */
3738 af = sctp_get_af_specific(addr->sa_family);
3739 if (!af || addr_len < af->sockaddr_len) {
3742 /* Pass correct addr len to common routine (so it knows there
3743 * is only one address being passed.
3745 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3748 sctp_release_sock(sk);
3752 /* FIXME: Write comments. */
3753 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3755 return -EOPNOTSUPP; /* STUB */
3758 /* 4.1.4 accept() - TCP Style Syntax
3760 * Applications use accept() call to remove an established SCTP
3761 * association from the accept queue of the endpoint. A new socket
3762 * descriptor will be returned from accept() to represent the newly
3763 * formed association.
3765 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3767 struct sctp_sock *sp;
3768 struct sctp_endpoint *ep;
3769 struct sock *newsk = NULL;
3770 struct sctp_association *asoc;
3779 if (!sctp_style(sk, TCP)) {
3780 error = -EOPNOTSUPP;
3784 if (!sctp_sstate(sk, LISTENING)) {
3789 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3791 error = sctp_wait_for_accept(sk, timeo);
3795 /* We treat the list of associations on the endpoint as the accept
3796 * queue and pick the first association on the list.
3798 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3800 newsk = sp->pf->create_accept_sk(sk, asoc);
3806 /* Populate the fields of the newsk from the oldsk and migrate the
3807 * asoc to the newsk.
3809 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3812 sctp_release_sock(sk);
3817 /* The SCTP ioctl handler. */
3818 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3825 * SEQPACKET-style sockets in LISTENING state are valid, for
3826 * SCTP, so only discard TCP-style sockets in LISTENING state.
3828 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3833 struct sk_buff *skb;
3834 unsigned int amount = 0;
3836 skb = skb_peek(&sk->sk_receive_queue);
3839 * We will only return the amount of this packet since
3840 * that is all that will be read.
3844 rc = put_user(amount, (int __user *)arg);
3852 sctp_release_sock(sk);
3856 /* This is the function which gets called during socket creation to
3857 * initialized the SCTP-specific portion of the sock.
3858 * The sock structure should already be zero-filled memory.
3860 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3862 struct net *net = sock_net(sk);
3863 struct sctp_endpoint *ep;
3864 struct sctp_sock *sp;
3866 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3870 /* Initialize the SCTP per socket area. */
3871 switch (sk->sk_type) {
3872 case SOCK_SEQPACKET:
3873 sp->type = SCTP_SOCKET_UDP;
3876 sp->type = SCTP_SOCKET_TCP;
3879 return -ESOCKTNOSUPPORT;
3882 /* Initialize default send parameters. These parameters can be
3883 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3885 sp->default_stream = 0;
3886 sp->default_ppid = 0;
3887 sp->default_flags = 0;
3888 sp->default_context = 0;
3889 sp->default_timetolive = 0;
3891 sp->default_rcv_context = 0;
3892 sp->max_burst = net->sctp.max_burst;
3894 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
3896 /* Initialize default setup parameters. These parameters
3897 * can be modified with the SCTP_INITMSG socket option or
3898 * overridden by the SCTP_INIT CMSG.
3900 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3901 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3902 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
3903 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
3905 /* Initialize default RTO related parameters. These parameters can
3906 * be modified for with the SCTP_RTOINFO socket option.
3908 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
3909 sp->rtoinfo.srto_max = net->sctp.rto_max;
3910 sp->rtoinfo.srto_min = net->sctp.rto_min;
3912 /* Initialize default association related parameters. These parameters
3913 * can be modified with the SCTP_ASSOCINFO socket option.
3915 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
3916 sp->assocparams.sasoc_number_peer_destinations = 0;
3917 sp->assocparams.sasoc_peer_rwnd = 0;
3918 sp->assocparams.sasoc_local_rwnd = 0;
3919 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
3921 /* Initialize default event subscriptions. By default, all the
3924 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3926 /* Default Peer Address Parameters. These defaults can
3927 * be modified via SCTP_PEER_ADDR_PARAMS
3929 sp->hbinterval = net->sctp.hb_interval;
3930 sp->pathmaxrxt = net->sctp.max_retrans_path;
3931 sp->pathmtu = 0; // allow default discovery
3932 sp->sackdelay = net->sctp.sack_timeout;
3934 sp->param_flags = SPP_HB_ENABLE |
3936 SPP_SACKDELAY_ENABLE;
3938 /* If enabled no SCTP message fragmentation will be performed.
3939 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3941 sp->disable_fragments = 0;
3943 /* Enable Nagle algorithm by default. */
3946 /* Enable by default. */
3949 /* Auto-close idle associations after the configured
3950 * number of seconds. A value of 0 disables this
3951 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3952 * for UDP-style sockets only.
3956 /* User specified fragmentation limit. */
3959 sp->adaptation_ind = 0;
3961 sp->pf = sctp_get_pf_specific(sk->sk_family);
3963 /* Control variables for partial data delivery. */
3964 atomic_set(&sp->pd_mode, 0);
3965 skb_queue_head_init(&sp->pd_lobby);
3966 sp->frag_interleave = 0;
3968 /* Create a per socket endpoint structure. Even if we
3969 * change the data structure relationships, this may still
3970 * be useful for storing pre-connect address information.
3972 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3979 SCTP_DBG_OBJCNT_INC(sock);
3982 percpu_counter_inc(&sctp_sockets_allocated);
3983 sock_prot_inuse_add(net, sk->sk_prot, 1);
3984 if (net->sctp.default_auto_asconf) {
3985 list_add_tail(&sp->auto_asconf_list,
3986 &net->sctp.auto_asconf_splist);
3987 sp->do_auto_asconf = 1;
3989 sp->do_auto_asconf = 0;
3995 /* Cleanup any SCTP per socket resources. */
3996 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3998 struct sctp_sock *sp;
4000 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
4002 /* Release our hold on the endpoint. */
4004 if (sp->do_auto_asconf) {
4005 sp->do_auto_asconf = 0;
4006 list_del(&sp->auto_asconf_list);
4008 sctp_endpoint_free(sp->ep);
4010 percpu_counter_dec(&sctp_sockets_allocated);
4011 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4015 /* API 4.1.7 shutdown() - TCP Style Syntax
4016 * int shutdown(int socket, int how);
4018 * sd - the socket descriptor of the association to be closed.
4019 * how - Specifies the type of shutdown. The values are
4022 * Disables further receive operations. No SCTP
4023 * protocol action is taken.
4025 * Disables further send operations, and initiates
4026 * the SCTP shutdown sequence.
4028 * Disables further send and receive operations
4029 * and initiates the SCTP shutdown sequence.
4031 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
4033 struct net *net = sock_net(sk);
4034 struct sctp_endpoint *ep;
4035 struct sctp_association *asoc;
4037 if (!sctp_style(sk, TCP))
4040 if (how & SEND_SHUTDOWN) {
4041 ep = sctp_sk(sk)->ep;
4042 if (!list_empty(&ep->asocs)) {
4043 asoc = list_entry(ep->asocs.next,
4044 struct sctp_association, asocs);
4045 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4050 /* 7.2.1 Association Status (SCTP_STATUS)
4052 * Applications can retrieve current status information about an
4053 * association, including association state, peer receiver window size,
4054 * number of unacked data chunks, and number of data chunks pending
4055 * receipt. This information is read-only.
4057 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4058 char __user *optval,
4061 struct sctp_status status;
4062 struct sctp_association *asoc = NULL;
4063 struct sctp_transport *transport;
4064 sctp_assoc_t associd;
4067 if (len < sizeof(status)) {
4072 len = sizeof(status);
4073 if (copy_from_user(&status, optval, len)) {
4078 associd = status.sstat_assoc_id;
4079 asoc = sctp_id2assoc(sk, associd);
4085 transport = asoc->peer.primary_path;
4087 status.sstat_assoc_id = sctp_assoc2id(asoc);
4088 status.sstat_state = asoc->state;
4089 status.sstat_rwnd = asoc->peer.rwnd;
4090 status.sstat_unackdata = asoc->unack_data;
4092 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4093 status.sstat_instrms = asoc->c.sinit_max_instreams;
4094 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4095 status.sstat_fragmentation_point = asoc->frag_point;
4096 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4097 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4098 transport->af_specific->sockaddr_len);
4099 /* Map ipv4 address into v4-mapped-on-v6 address. */
4100 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4101 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4102 status.sstat_primary.spinfo_state = transport->state;
4103 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4104 status.sstat_primary.spinfo_srtt = transport->srtt;
4105 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4106 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4108 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4109 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4111 if (put_user(len, optlen)) {
4116 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
4117 len, status.sstat_state, status.sstat_rwnd,
4118 status.sstat_assoc_id);
4120 if (copy_to_user(optval, &status, len)) {
4130 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4132 * Applications can retrieve information about a specific peer address
4133 * of an association, including its reachability state, congestion
4134 * window, and retransmission timer values. This information is
4137 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4138 char __user *optval,
4141 struct sctp_paddrinfo pinfo;
4142 struct sctp_transport *transport;
4145 if (len < sizeof(pinfo)) {
4150 len = sizeof(pinfo);
4151 if (copy_from_user(&pinfo, optval, len)) {
4156 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4157 pinfo.spinfo_assoc_id);
4161 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4162 pinfo.spinfo_state = transport->state;
4163 pinfo.spinfo_cwnd = transport->cwnd;
4164 pinfo.spinfo_srtt = transport->srtt;
4165 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4166 pinfo.spinfo_mtu = transport->pathmtu;
4168 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4169 pinfo.spinfo_state = SCTP_ACTIVE;
4171 if (put_user(len, optlen)) {
4176 if (copy_to_user(optval, &pinfo, len)) {
4185 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4187 * This option is a on/off flag. If enabled no SCTP message
4188 * fragmentation will be performed. Instead if a message being sent
4189 * exceeds the current PMTU size, the message will NOT be sent and
4190 * instead a error will be indicated to the user.
4192 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4193 char __user *optval, int __user *optlen)
4197 if (len < sizeof(int))
4201 val = (sctp_sk(sk)->disable_fragments == 1);
4202 if (put_user(len, optlen))
4204 if (copy_to_user(optval, &val, len))
4209 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4211 * This socket option is used to specify various notifications and
4212 * ancillary data the user wishes to receive.
4214 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4219 if (len > sizeof(struct sctp_event_subscribe))
4220 len = sizeof(struct sctp_event_subscribe);
4221 if (put_user(len, optlen))
4223 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4228 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4230 * This socket option is applicable to the UDP-style socket only. When
4231 * set it will cause associations that are idle for more than the
4232 * specified number of seconds to automatically close. An association
4233 * being idle is defined an association that has NOT sent or received
4234 * user data. The special value of '0' indicates that no automatic
4235 * close of any associations should be performed. The option expects an
4236 * integer defining the number of seconds of idle time before an
4237 * association is closed.
4239 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4241 /* Applicable to UDP-style socket only */
4242 if (sctp_style(sk, TCP))
4244 if (len < sizeof(int))
4247 if (put_user(len, optlen))
4249 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4254 /* Helper routine to branch off an association to a new socket. */
4255 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4257 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4258 struct socket *sock;
4265 /* An association cannot be branched off from an already peeled-off
4266 * socket, nor is this supported for tcp style sockets.
4268 if (!sctp_style(sk, UDP))
4271 /* Create a new socket. */
4272 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4276 sctp_copy_sock(sock->sk, sk, asoc);
4278 /* Make peeled-off sockets more like 1-1 accepted sockets.
4279 * Set the daddr and initialize id to something more random
4281 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4282 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4284 /* Populate the fields of the newsk from the oldsk and migrate the
4285 * asoc to the newsk.
4287 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4293 EXPORT_SYMBOL(sctp_do_peeloff);
4295 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4297 sctp_peeloff_arg_t peeloff;
4298 struct socket *newsock;
4299 struct file *newfile;
4302 if (len < sizeof(sctp_peeloff_arg_t))
4304 len = sizeof(sctp_peeloff_arg_t);
4305 if (copy_from_user(&peeloff, optval, len))
4308 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4312 /* Map the socket to an unused fd that can be returned to the user. */
4313 retval = get_unused_fd();
4315 sock_release(newsock);
4319 newfile = sock_alloc_file(newsock, 0, NULL);
4320 if (unlikely(IS_ERR(newfile))) {
4321 put_unused_fd(retval);
4322 sock_release(newsock);
4323 return PTR_ERR(newfile);
4326 SCTP_DEBUG_PRINTK("%s: sk: %p newsk: %p sd: %d\n",
4327 __func__, sk, newsock->sk, retval);
4329 /* Return the fd mapped to the new socket. */
4330 if (put_user(len, optlen)) {
4332 put_unused_fd(retval);
4335 peeloff.sd = retval;
4336 if (copy_to_user(optval, &peeloff, len)) {
4338 put_unused_fd(retval);
4341 fd_install(retval, newfile);
4346 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4348 * Applications can enable or disable heartbeats for any peer address of
4349 * an association, modify an address's heartbeat interval, force a
4350 * heartbeat to be sent immediately, and adjust the address's maximum
4351 * number of retransmissions sent before an address is considered
4352 * unreachable. The following structure is used to access and modify an
4353 * address's parameters:
4355 * struct sctp_paddrparams {
4356 * sctp_assoc_t spp_assoc_id;
4357 * struct sockaddr_storage spp_address;
4358 * uint32_t spp_hbinterval;
4359 * uint16_t spp_pathmaxrxt;
4360 * uint32_t spp_pathmtu;
4361 * uint32_t spp_sackdelay;
4362 * uint32_t spp_flags;
4365 * spp_assoc_id - (one-to-many style socket) This is filled in the
4366 * application, and identifies the association for
4368 * spp_address - This specifies which address is of interest.
4369 * spp_hbinterval - This contains the value of the heartbeat interval,
4370 * in milliseconds. If a value of zero
4371 * is present in this field then no changes are to
4372 * be made to this parameter.
4373 * spp_pathmaxrxt - This contains the maximum number of
4374 * retransmissions before this address shall be
4375 * considered unreachable. If a value of zero
4376 * is present in this field then no changes are to
4377 * be made to this parameter.
4378 * spp_pathmtu - When Path MTU discovery is disabled the value
4379 * specified here will be the "fixed" path mtu.
4380 * Note that if the spp_address field is empty
4381 * then all associations on this address will
4382 * have this fixed path mtu set upon them.
4384 * spp_sackdelay - When delayed sack is enabled, this value specifies
4385 * the number of milliseconds that sacks will be delayed
4386 * for. This value will apply to all addresses of an
4387 * association if the spp_address field is empty. Note
4388 * also, that if delayed sack is enabled and this
4389 * value is set to 0, no change is made to the last
4390 * recorded delayed sack timer value.
4392 * spp_flags - These flags are used to control various features
4393 * on an association. The flag field may contain
4394 * zero or more of the following options.
4396 * SPP_HB_ENABLE - Enable heartbeats on the
4397 * specified address. Note that if the address
4398 * field is empty all addresses for the association
4399 * have heartbeats enabled upon them.
4401 * SPP_HB_DISABLE - Disable heartbeats on the
4402 * speicifed address. Note that if the address
4403 * field is empty all addresses for the association
4404 * will have their heartbeats disabled. Note also
4405 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4406 * mutually exclusive, only one of these two should
4407 * be specified. Enabling both fields will have
4408 * undetermined results.
4410 * SPP_HB_DEMAND - Request a user initiated heartbeat
4411 * to be made immediately.
4413 * SPP_PMTUD_ENABLE - This field will enable PMTU
4414 * discovery upon the specified address. Note that
4415 * if the address feild is empty then all addresses
4416 * on the association are effected.
4418 * SPP_PMTUD_DISABLE - This field will disable PMTU
4419 * discovery upon the specified address. Note that
4420 * if the address feild is empty then all addresses
4421 * on the association are effected. Not also that
4422 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4423 * exclusive. Enabling both will have undetermined
4426 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4427 * on delayed sack. The time specified in spp_sackdelay
4428 * is used to specify the sack delay for this address. Note
4429 * that if spp_address is empty then all addresses will
4430 * enable delayed sack and take on the sack delay
4431 * value specified in spp_sackdelay.
4432 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4433 * off delayed sack. If the spp_address field is blank then
4434 * delayed sack is disabled for the entire association. Note
4435 * also that this field is mutually exclusive to
4436 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4439 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4440 char __user *optval, int __user *optlen)
4442 struct sctp_paddrparams params;
4443 struct sctp_transport *trans = NULL;
4444 struct sctp_association *asoc = NULL;
4445 struct sctp_sock *sp = sctp_sk(sk);
4447 if (len < sizeof(struct sctp_paddrparams))
4449 len = sizeof(struct sctp_paddrparams);
4450 if (copy_from_user(¶ms, optval, len))
4453 /* If an address other than INADDR_ANY is specified, and
4454 * no transport is found, then the request is invalid.
4456 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
4457 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4458 params.spp_assoc_id);
4460 SCTP_DEBUG_PRINTK("Failed no transport\n");
4465 /* Get association, if assoc_id != 0 and the socket is a one
4466 * to many style socket, and an association was not found, then
4467 * the id was invalid.
4469 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4470 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4471 SCTP_DEBUG_PRINTK("Failed no association\n");
4476 /* Fetch transport values. */
4477 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4478 params.spp_pathmtu = trans->pathmtu;
4479 params.spp_pathmaxrxt = trans->pathmaxrxt;
4480 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4482 /*draft-11 doesn't say what to return in spp_flags*/
4483 params.spp_flags = trans->param_flags;
4485 /* Fetch association values. */
4486 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4487 params.spp_pathmtu = asoc->pathmtu;
4488 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4489 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4491 /*draft-11 doesn't say what to return in spp_flags*/
4492 params.spp_flags = asoc->param_flags;
4494 /* Fetch socket values. */
4495 params.spp_hbinterval = sp->hbinterval;
4496 params.spp_pathmtu = sp->pathmtu;
4497 params.spp_sackdelay = sp->sackdelay;
4498 params.spp_pathmaxrxt = sp->pathmaxrxt;
4500 /*draft-11 doesn't say what to return in spp_flags*/
4501 params.spp_flags = sp->param_flags;
4504 if (copy_to_user(optval, ¶ms, len))
4507 if (put_user(len, optlen))
4514 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4516 * This option will effect the way delayed acks are performed. This
4517 * option allows you to get or set the delayed ack time, in
4518 * milliseconds. It also allows changing the delayed ack frequency.
4519 * Changing the frequency to 1 disables the delayed sack algorithm. If
4520 * the assoc_id is 0, then this sets or gets the endpoints default
4521 * values. If the assoc_id field is non-zero, then the set or get
4522 * effects the specified association for the one to many model (the
4523 * assoc_id field is ignored by the one to one model). Note that if
4524 * sack_delay or sack_freq are 0 when setting this option, then the
4525 * current values will remain unchanged.
4527 * struct sctp_sack_info {
4528 * sctp_assoc_t sack_assoc_id;
4529 * uint32_t sack_delay;
4530 * uint32_t sack_freq;
4533 * sack_assoc_id - This parameter, indicates which association the user
4534 * is performing an action upon. Note that if this field's value is
4535 * zero then the endpoints default value is changed (effecting future
4536 * associations only).
4538 * sack_delay - This parameter contains the number of milliseconds that
4539 * the user is requesting the delayed ACK timer be set to. Note that
4540 * this value is defined in the standard to be between 200 and 500
4543 * sack_freq - This parameter contains the number of packets that must
4544 * be received before a sack is sent without waiting for the delay
4545 * timer to expire. The default value for this is 2, setting this
4546 * value to 1 will disable the delayed sack algorithm.
4548 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4549 char __user *optval,
4552 struct sctp_sack_info params;
4553 struct sctp_association *asoc = NULL;
4554 struct sctp_sock *sp = sctp_sk(sk);
4556 if (len >= sizeof(struct sctp_sack_info)) {
4557 len = sizeof(struct sctp_sack_info);
4559 if (copy_from_user(¶ms, optval, len))
4561 } else if (len == sizeof(struct sctp_assoc_value)) {
4562 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4563 pr_warn("Use struct sctp_sack_info instead\n");
4564 if (copy_from_user(¶ms, optval, len))
4569 /* Get association, if sack_assoc_id != 0 and the socket is a one
4570 * to many style socket, and an association was not found, then
4571 * the id was invalid.
4573 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4574 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4578 /* Fetch association values. */
4579 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4580 params.sack_delay = jiffies_to_msecs(
4582 params.sack_freq = asoc->sackfreq;
4585 params.sack_delay = 0;
4586 params.sack_freq = 1;
4589 /* Fetch socket values. */
4590 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4591 params.sack_delay = sp->sackdelay;
4592 params.sack_freq = sp->sackfreq;
4594 params.sack_delay = 0;
4595 params.sack_freq = 1;
4599 if (copy_to_user(optval, ¶ms, len))
4602 if (put_user(len, optlen))
4608 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4610 * Applications can specify protocol parameters for the default association
4611 * initialization. The option name argument to setsockopt() and getsockopt()
4614 * Setting initialization parameters is effective only on an unconnected
4615 * socket (for UDP-style sockets only future associations are effected
4616 * by the change). With TCP-style sockets, this option is inherited by
4617 * sockets derived from a listener socket.
4619 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4621 if (len < sizeof(struct sctp_initmsg))
4623 len = sizeof(struct sctp_initmsg);
4624 if (put_user(len, optlen))
4626 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4632 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4633 char __user *optval, int __user *optlen)
4635 struct sctp_association *asoc;
4637 struct sctp_getaddrs getaddrs;
4638 struct sctp_transport *from;
4640 union sctp_addr temp;
4641 struct sctp_sock *sp = sctp_sk(sk);
4646 if (len < sizeof(struct sctp_getaddrs))
4649 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4652 /* For UDP-style sockets, id specifies the association to query. */
4653 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4657 to = optval + offsetof(struct sctp_getaddrs,addrs);
4658 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4660 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4662 memcpy(&temp, &from->ipaddr, sizeof(temp));
4663 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4664 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4665 if (space_left < addrlen)
4667 if (copy_to_user(to, &temp, addrlen))
4671 space_left -= addrlen;
4674 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4676 bytes_copied = ((char __user *)to) - optval;
4677 if (put_user(bytes_copied, optlen))
4683 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4684 size_t space_left, int *bytes_copied)
4686 struct sctp_sockaddr_entry *addr;
4687 union sctp_addr temp;
4690 struct net *net = sock_net(sk);
4693 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4697 if ((PF_INET == sk->sk_family) &&
4698 (AF_INET6 == addr->a.sa.sa_family))
4700 if ((PF_INET6 == sk->sk_family) &&
4701 inet_v6_ipv6only(sk) &&
4702 (AF_INET == addr->a.sa.sa_family))
4704 memcpy(&temp, &addr->a, sizeof(temp));
4705 if (!temp.v4.sin_port)
4706 temp.v4.sin_port = htons(port);
4708 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4710 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4711 if (space_left < addrlen) {
4715 memcpy(to, &temp, addrlen);
4719 space_left -= addrlen;
4720 *bytes_copied += addrlen;
4728 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4729 char __user *optval, int __user *optlen)
4731 struct sctp_bind_addr *bp;
4732 struct sctp_association *asoc;
4734 struct sctp_getaddrs getaddrs;
4735 struct sctp_sockaddr_entry *addr;
4737 union sctp_addr temp;
4738 struct sctp_sock *sp = sctp_sk(sk);
4742 int bytes_copied = 0;
4746 if (len < sizeof(struct sctp_getaddrs))
4749 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4753 * For UDP-style sockets, id specifies the association to query.
4754 * If the id field is set to the value '0' then the locally bound
4755 * addresses are returned without regard to any particular
4758 if (0 == getaddrs.assoc_id) {
4759 bp = &sctp_sk(sk)->ep->base.bind_addr;
4761 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4764 bp = &asoc->base.bind_addr;
4767 to = optval + offsetof(struct sctp_getaddrs,addrs);
4768 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4770 addrs = kmalloc(space_left, GFP_KERNEL);
4774 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4775 * addresses from the global local address list.
4777 if (sctp_list_single_entry(&bp->address_list)) {
4778 addr = list_entry(bp->address_list.next,
4779 struct sctp_sockaddr_entry, list);
4780 if (sctp_is_any(sk, &addr->a)) {
4781 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4782 space_left, &bytes_copied);
4792 /* Protection on the bound address list is not needed since
4793 * in the socket option context we hold a socket lock and
4794 * thus the bound address list can't change.
4796 list_for_each_entry(addr, &bp->address_list, list) {
4797 memcpy(&temp, &addr->a, sizeof(temp));
4798 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4799 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4800 if (space_left < addrlen) {
4801 err = -ENOMEM; /*fixme: right error?*/
4804 memcpy(buf, &temp, addrlen);
4806 bytes_copied += addrlen;
4808 space_left -= addrlen;
4812 if (copy_to_user(to, addrs, bytes_copied)) {
4816 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4820 if (put_user(bytes_copied, optlen))
4827 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4829 * Requests that the local SCTP stack use the enclosed peer address as
4830 * the association primary. The enclosed address must be one of the
4831 * association peer's addresses.
4833 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4834 char __user *optval, int __user *optlen)
4836 struct sctp_prim prim;
4837 struct sctp_association *asoc;
4838 struct sctp_sock *sp = sctp_sk(sk);
4840 if (len < sizeof(struct sctp_prim))
4843 len = sizeof(struct sctp_prim);
4845 if (copy_from_user(&prim, optval, len))
4848 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4852 if (!asoc->peer.primary_path)
4855 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4856 asoc->peer.primary_path->af_specific->sockaddr_len);
4858 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4859 (union sctp_addr *)&prim.ssp_addr);
4861 if (put_user(len, optlen))
4863 if (copy_to_user(optval, &prim, len))
4870 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4872 * Requests that the local endpoint set the specified Adaptation Layer
4873 * Indication parameter for all future INIT and INIT-ACK exchanges.
4875 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4876 char __user *optval, int __user *optlen)
4878 struct sctp_setadaptation adaptation;
4880 if (len < sizeof(struct sctp_setadaptation))
4883 len = sizeof(struct sctp_setadaptation);
4885 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4887 if (put_user(len, optlen))
4889 if (copy_to_user(optval, &adaptation, len))
4897 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4899 * Applications that wish to use the sendto() system call may wish to
4900 * specify a default set of parameters that would normally be supplied
4901 * through the inclusion of ancillary data. This socket option allows
4902 * such an application to set the default sctp_sndrcvinfo structure.
4905 * The application that wishes to use this socket option simply passes
4906 * in to this call the sctp_sndrcvinfo structure defined in Section
4907 * 5.2.2) The input parameters accepted by this call include
4908 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4909 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4910 * to this call if the caller is using the UDP model.
4912 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4914 static int sctp_getsockopt_default_send_param(struct sock *sk,
4915 int len, char __user *optval,
4918 struct sctp_sndrcvinfo info;
4919 struct sctp_association *asoc;
4920 struct sctp_sock *sp = sctp_sk(sk);
4922 if (len < sizeof(struct sctp_sndrcvinfo))
4925 len = sizeof(struct sctp_sndrcvinfo);
4927 if (copy_from_user(&info, optval, len))
4930 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4931 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4935 info.sinfo_stream = asoc->default_stream;
4936 info.sinfo_flags = asoc->default_flags;
4937 info.sinfo_ppid = asoc->default_ppid;
4938 info.sinfo_context = asoc->default_context;
4939 info.sinfo_timetolive = asoc->default_timetolive;
4941 info.sinfo_stream = sp->default_stream;
4942 info.sinfo_flags = sp->default_flags;
4943 info.sinfo_ppid = sp->default_ppid;
4944 info.sinfo_context = sp->default_context;
4945 info.sinfo_timetolive = sp->default_timetolive;
4948 if (put_user(len, optlen))
4950 if (copy_to_user(optval, &info, len))
4958 * 7.1.5 SCTP_NODELAY
4960 * Turn on/off any Nagle-like algorithm. This means that packets are
4961 * generally sent as soon as possible and no unnecessary delays are
4962 * introduced, at the cost of more packets in the network. Expects an
4963 * integer boolean flag.
4966 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4967 char __user *optval, int __user *optlen)
4971 if (len < sizeof(int))
4975 val = (sctp_sk(sk)->nodelay == 1);
4976 if (put_user(len, optlen))
4978 if (copy_to_user(optval, &val, len))
4985 * 7.1.1 SCTP_RTOINFO
4987 * The protocol parameters used to initialize and bound retransmission
4988 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4989 * and modify these parameters.
4990 * All parameters are time values, in milliseconds. A value of 0, when
4991 * modifying the parameters, indicates that the current value should not
4995 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4996 char __user *optval,
4997 int __user *optlen) {
4998 struct sctp_rtoinfo rtoinfo;
4999 struct sctp_association *asoc;
5001 if (len < sizeof (struct sctp_rtoinfo))
5004 len = sizeof(struct sctp_rtoinfo);
5006 if (copy_from_user(&rtoinfo, optval, len))
5009 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5011 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5014 /* Values corresponding to the specific association. */
5016 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5017 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5018 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5020 /* Values corresponding to the endpoint. */
5021 struct sctp_sock *sp = sctp_sk(sk);
5023 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5024 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5025 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5028 if (put_user(len, optlen))
5031 if (copy_to_user(optval, &rtoinfo, len))
5039 * 7.1.2 SCTP_ASSOCINFO
5041 * This option is used to tune the maximum retransmission attempts
5042 * of the association.
5043 * Returns an error if the new association retransmission value is
5044 * greater than the sum of the retransmission value of the peer.
5045 * See [SCTP] for more information.
5048 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5049 char __user *optval,
5053 struct sctp_assocparams assocparams;
5054 struct sctp_association *asoc;
5055 struct list_head *pos;
5058 if (len < sizeof (struct sctp_assocparams))
5061 len = sizeof(struct sctp_assocparams);
5063 if (copy_from_user(&assocparams, optval, len))
5066 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5068 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5071 /* Values correspoinding to the specific association */
5073 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5074 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5075 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5076 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
5078 (asoc->cookie_life.tv_usec
5081 list_for_each(pos, &asoc->peer.transport_addr_list) {
5085 assocparams.sasoc_number_peer_destinations = cnt;
5087 /* Values corresponding to the endpoint */
5088 struct sctp_sock *sp = sctp_sk(sk);
5090 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5091 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5092 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5093 assocparams.sasoc_cookie_life =
5094 sp->assocparams.sasoc_cookie_life;
5095 assocparams.sasoc_number_peer_destinations =
5097 sasoc_number_peer_destinations;
5100 if (put_user(len, optlen))
5103 if (copy_to_user(optval, &assocparams, len))
5110 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5112 * This socket option is a boolean flag which turns on or off mapped V4
5113 * addresses. If this option is turned on and the socket is type
5114 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5115 * If this option is turned off, then no mapping will be done of V4
5116 * addresses and a user will receive both PF_INET6 and PF_INET type
5117 * addresses on the socket.
5119 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5120 char __user *optval, int __user *optlen)
5123 struct sctp_sock *sp = sctp_sk(sk);
5125 if (len < sizeof(int))
5130 if (put_user(len, optlen))
5132 if (copy_to_user(optval, &val, len))
5139 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5140 * (chapter and verse is quoted at sctp_setsockopt_context())
5142 static int sctp_getsockopt_context(struct sock *sk, int len,
5143 char __user *optval, int __user *optlen)
5145 struct sctp_assoc_value params;
5146 struct sctp_sock *sp;
5147 struct sctp_association *asoc;
5149 if (len < sizeof(struct sctp_assoc_value))
5152 len = sizeof(struct sctp_assoc_value);
5154 if (copy_from_user(¶ms, optval, len))
5159 if (params.assoc_id != 0) {
5160 asoc = sctp_id2assoc(sk, params.assoc_id);
5163 params.assoc_value = asoc->default_rcv_context;
5165 params.assoc_value = sp->default_rcv_context;
5168 if (put_user(len, optlen))
5170 if (copy_to_user(optval, ¶ms, len))
5177 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5178 * This option will get or set the maximum size to put in any outgoing
5179 * SCTP DATA chunk. If a message is larger than this size it will be
5180 * fragmented by SCTP into the specified size. Note that the underlying
5181 * SCTP implementation may fragment into smaller sized chunks when the
5182 * PMTU of the underlying association is smaller than the value set by
5183 * the user. The default value for this option is '0' which indicates
5184 * the user is NOT limiting fragmentation and only the PMTU will effect
5185 * SCTP's choice of DATA chunk size. Note also that values set larger
5186 * than the maximum size of an IP datagram will effectively let SCTP
5187 * control fragmentation (i.e. the same as setting this option to 0).
5189 * The following structure is used to access and modify this parameter:
5191 * struct sctp_assoc_value {
5192 * sctp_assoc_t assoc_id;
5193 * uint32_t assoc_value;
5196 * assoc_id: This parameter is ignored for one-to-one style sockets.
5197 * For one-to-many style sockets this parameter indicates which
5198 * association the user is performing an action upon. Note that if
5199 * this field's value is zero then the endpoints default value is
5200 * changed (effecting future associations only).
5201 * assoc_value: This parameter specifies the maximum size in bytes.
5203 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5204 char __user *optval, int __user *optlen)
5206 struct sctp_assoc_value params;
5207 struct sctp_association *asoc;
5209 if (len == sizeof(int)) {
5210 pr_warn("Use of int in maxseg socket option deprecated\n");
5211 pr_warn("Use struct sctp_assoc_value instead\n");
5212 params.assoc_id = 0;
5213 } else if (len >= sizeof(struct sctp_assoc_value)) {
5214 len = sizeof(struct sctp_assoc_value);
5215 if (copy_from_user(¶ms, optval, sizeof(params)))
5220 asoc = sctp_id2assoc(sk, params.assoc_id);
5221 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5225 params.assoc_value = asoc->frag_point;
5227 params.assoc_value = sctp_sk(sk)->user_frag;
5229 if (put_user(len, optlen))
5231 if (len == sizeof(int)) {
5232 if (copy_to_user(optval, ¶ms.assoc_value, len))
5235 if (copy_to_user(optval, ¶ms, len))
5243 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5244 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5246 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5247 char __user *optval, int __user *optlen)
5251 if (len < sizeof(int))
5256 val = sctp_sk(sk)->frag_interleave;
5257 if (put_user(len, optlen))
5259 if (copy_to_user(optval, &val, len))
5266 * 7.1.25. Set or Get the sctp partial delivery point
5267 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5269 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5270 char __user *optval,
5275 if (len < sizeof(u32))
5280 val = sctp_sk(sk)->pd_point;
5281 if (put_user(len, optlen))
5283 if (copy_to_user(optval, &val, len))
5290 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5291 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5293 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5294 char __user *optval,
5297 struct sctp_assoc_value params;
5298 struct sctp_sock *sp;
5299 struct sctp_association *asoc;
5301 if (len == sizeof(int)) {
5302 pr_warn("Use of int in max_burst socket option deprecated\n");
5303 pr_warn("Use struct sctp_assoc_value instead\n");
5304 params.assoc_id = 0;
5305 } else if (len >= sizeof(struct sctp_assoc_value)) {
5306 len = sizeof(struct sctp_assoc_value);
5307 if (copy_from_user(¶ms, optval, len))
5314 if (params.assoc_id != 0) {
5315 asoc = sctp_id2assoc(sk, params.assoc_id);
5318 params.assoc_value = asoc->max_burst;
5320 params.assoc_value = sp->max_burst;
5322 if (len == sizeof(int)) {
5323 if (copy_to_user(optval, ¶ms.assoc_value, len))
5326 if (copy_to_user(optval, ¶ms, len))
5334 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5335 char __user *optval, int __user *optlen)
5337 struct net *net = sock_net(sk);
5338 struct sctp_hmacalgo __user *p = (void __user *)optval;
5339 struct sctp_hmac_algo_param *hmacs;
5343 if (!net->sctp.auth_enable)
5346 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5347 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5349 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5352 len = sizeof(struct sctp_hmacalgo) + data_len;
5353 num_idents = data_len / sizeof(u16);
5355 if (put_user(len, optlen))
5357 if (put_user(num_idents, &p->shmac_num_idents))
5359 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5364 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5365 char __user *optval, int __user *optlen)
5367 struct net *net = sock_net(sk);
5368 struct sctp_authkeyid val;
5369 struct sctp_association *asoc;
5371 if (!net->sctp.auth_enable)
5374 if (len < sizeof(struct sctp_authkeyid))
5376 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5379 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5380 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5384 val.scact_keynumber = asoc->active_key_id;
5386 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5388 len = sizeof(struct sctp_authkeyid);
5389 if (put_user(len, optlen))
5391 if (copy_to_user(optval, &val, len))
5397 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5398 char __user *optval, int __user *optlen)
5400 struct net *net = sock_net(sk);
5401 struct sctp_authchunks __user *p = (void __user *)optval;
5402 struct sctp_authchunks val;
5403 struct sctp_association *asoc;
5404 struct sctp_chunks_param *ch;
5408 if (!net->sctp.auth_enable)
5411 if (len < sizeof(struct sctp_authchunks))
5414 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5417 to = p->gauth_chunks;
5418 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5422 ch = asoc->peer.peer_chunks;
5426 /* See if the user provided enough room for all the data */
5427 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5428 if (len < num_chunks)
5431 if (copy_to_user(to, ch->chunks, num_chunks))
5434 len = sizeof(struct sctp_authchunks) + num_chunks;
5435 if (put_user(len, optlen)) return -EFAULT;
5436 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5441 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5442 char __user *optval, int __user *optlen)
5444 struct net *net = sock_net(sk);
5445 struct sctp_authchunks __user *p = (void __user *)optval;
5446 struct sctp_authchunks val;
5447 struct sctp_association *asoc;
5448 struct sctp_chunks_param *ch;
5452 if (!net->sctp.auth_enable)
5455 if (len < sizeof(struct sctp_authchunks))
5458 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5461 to = p->gauth_chunks;
5462 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5463 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5467 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5469 ch = sctp_sk(sk)->ep->auth_chunk_list;
5474 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5475 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5478 if (copy_to_user(to, ch->chunks, num_chunks))
5481 len = sizeof(struct sctp_authchunks) + num_chunks;
5482 if (put_user(len, optlen))
5484 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5491 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5492 * This option gets the current number of associations that are attached
5493 * to a one-to-many style socket. The option value is an uint32_t.
5495 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5496 char __user *optval, int __user *optlen)
5498 struct sctp_sock *sp = sctp_sk(sk);
5499 struct sctp_association *asoc;
5502 if (sctp_style(sk, TCP))
5505 if (len < sizeof(u32))
5510 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5514 if (put_user(len, optlen))
5516 if (copy_to_user(optval, &val, len))
5523 * 8.1.23 SCTP_AUTO_ASCONF
5524 * See the corresponding setsockopt entry as description
5526 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5527 char __user *optval, int __user *optlen)
5531 if (len < sizeof(int))
5535 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5537 if (put_user(len, optlen))
5539 if (copy_to_user(optval, &val, len))
5545 * 8.2.6. Get the Current Identifiers of Associations
5546 * (SCTP_GET_ASSOC_ID_LIST)
5548 * This option gets the current list of SCTP association identifiers of
5549 * the SCTP associations handled by a one-to-many style socket.
5551 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5552 char __user *optval, int __user *optlen)
5554 struct sctp_sock *sp = sctp_sk(sk);
5555 struct sctp_association *asoc;
5556 struct sctp_assoc_ids *ids;
5559 if (sctp_style(sk, TCP))
5562 if (len < sizeof(struct sctp_assoc_ids))
5565 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5569 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5572 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5574 ids = kmalloc(len, GFP_KERNEL);
5578 ids->gaids_number_of_ids = num;
5580 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5581 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5584 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5594 * SCTP_PEER_ADDR_THLDS
5596 * This option allows us to fetch the partially failed threshold for one or all
5597 * transports in an association. See Section 6.1 of:
5598 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5600 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5601 char __user *optval,
5605 struct sctp_paddrthlds val;
5606 struct sctp_transport *trans;
5607 struct sctp_association *asoc;
5609 if (len < sizeof(struct sctp_paddrthlds))
5611 len = sizeof(struct sctp_paddrthlds);
5612 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5615 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5616 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5620 val.spt_pathpfthld = asoc->pf_retrans;
5621 val.spt_pathmaxrxt = asoc->pathmaxrxt;
5623 trans = sctp_addr_id2transport(sk, &val.spt_address,
5628 val.spt_pathmaxrxt = trans->pathmaxrxt;
5629 val.spt_pathpfthld = trans->pf_retrans;
5632 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5638 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5639 char __user *optval, int __user *optlen)
5644 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5647 /* I can hardly begin to describe how wrong this is. This is
5648 * so broken as to be worse than useless. The API draft
5649 * REALLY is NOT helpful here... I am not convinced that the
5650 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5651 * are at all well-founded.
5653 if (level != SOL_SCTP) {
5654 struct sctp_af *af = sctp_sk(sk)->pf->af;
5656 retval = af->getsockopt(sk, level, optname, optval, optlen);
5660 if (get_user(len, optlen))
5667 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5669 case SCTP_DISABLE_FRAGMENTS:
5670 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5674 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5676 case SCTP_AUTOCLOSE:
5677 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5679 case SCTP_SOCKOPT_PEELOFF:
5680 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5682 case SCTP_PEER_ADDR_PARAMS:
5683 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5686 case SCTP_DELAYED_SACK:
5687 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5691 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5693 case SCTP_GET_PEER_ADDRS:
5694 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5697 case SCTP_GET_LOCAL_ADDRS:
5698 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5701 case SCTP_SOCKOPT_CONNECTX3:
5702 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5704 case SCTP_DEFAULT_SEND_PARAM:
5705 retval = sctp_getsockopt_default_send_param(sk, len,
5708 case SCTP_PRIMARY_ADDR:
5709 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5712 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5715 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5717 case SCTP_ASSOCINFO:
5718 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5720 case SCTP_I_WANT_MAPPED_V4_ADDR:
5721 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5724 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5726 case SCTP_GET_PEER_ADDR_INFO:
5727 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5730 case SCTP_ADAPTATION_LAYER:
5731 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5735 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5737 case SCTP_FRAGMENT_INTERLEAVE:
5738 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5741 case SCTP_PARTIAL_DELIVERY_POINT:
5742 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5745 case SCTP_MAX_BURST:
5746 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5749 case SCTP_AUTH_CHUNK:
5750 case SCTP_AUTH_DELETE_KEY:
5751 retval = -EOPNOTSUPP;
5753 case SCTP_HMAC_IDENT:
5754 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5756 case SCTP_AUTH_ACTIVE_KEY:
5757 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5759 case SCTP_PEER_AUTH_CHUNKS:
5760 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5763 case SCTP_LOCAL_AUTH_CHUNKS:
5764 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5767 case SCTP_GET_ASSOC_NUMBER:
5768 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5770 case SCTP_GET_ASSOC_ID_LIST:
5771 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5773 case SCTP_AUTO_ASCONF:
5774 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5776 case SCTP_PEER_ADDR_THLDS:
5777 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
5780 retval = -ENOPROTOOPT;
5784 sctp_release_sock(sk);
5788 static void sctp_hash(struct sock *sk)
5793 static void sctp_unhash(struct sock *sk)
5798 /* Check if port is acceptable. Possibly find first available port.
5800 * The port hash table (contained in the 'global' SCTP protocol storage
5801 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5802 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5803 * list (the list number is the port number hashed out, so as you
5804 * would expect from a hash function, all the ports in a given list have
5805 * such a number that hashes out to the same list number; you were
5806 * expecting that, right?); so each list has a set of ports, with a
5807 * link to the socket (struct sock) that uses it, the port number and
5808 * a fastreuse flag (FIXME: NPI ipg).
5810 static struct sctp_bind_bucket *sctp_bucket_create(
5811 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
5813 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5815 struct sctp_bind_hashbucket *head; /* hash list */
5816 struct sctp_bind_bucket *pp; /* hash list port iterator */
5817 struct hlist_node *node;
5818 unsigned short snum;
5821 snum = ntohs(addr->v4.sin_port);
5823 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5824 sctp_local_bh_disable();
5827 /* Search for an available port. */
5828 int low, high, remaining, index;
5831 inet_get_local_port_range(&low, &high);
5832 remaining = (high - low) + 1;
5833 rover = net_random() % remaining + low;
5837 if ((rover < low) || (rover > high))
5839 if (inet_is_reserved_local_port(rover))
5841 index = sctp_phashfn(sock_net(sk), rover);
5842 head = &sctp_port_hashtable[index];
5843 sctp_spin_lock(&head->lock);
5844 sctp_for_each_hentry(pp, node, &head->chain)
5845 if ((pp->port == rover) &&
5846 net_eq(sock_net(sk), pp->net))
5850 sctp_spin_unlock(&head->lock);
5851 } while (--remaining > 0);
5853 /* Exhausted local port range during search? */
5858 /* OK, here is the one we will use. HEAD (the port
5859 * hash table list entry) is non-NULL and we hold it's
5864 /* We are given an specific port number; we verify
5865 * that it is not being used. If it is used, we will
5866 * exahust the search in the hash list corresponding
5867 * to the port number (snum) - we detect that with the
5868 * port iterator, pp being NULL.
5870 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
5871 sctp_spin_lock(&head->lock);
5872 sctp_for_each_hentry(pp, node, &head->chain) {
5873 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
5880 if (!hlist_empty(&pp->owner)) {
5881 /* We had a port hash table hit - there is an
5882 * available port (pp != NULL) and it is being
5883 * used by other socket (pp->owner not empty); that other
5884 * socket is going to be sk2.
5886 int reuse = sk->sk_reuse;
5889 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5890 if (pp->fastreuse && sk->sk_reuse &&
5891 sk->sk_state != SCTP_SS_LISTENING)
5894 /* Run through the list of sockets bound to the port
5895 * (pp->port) [via the pointers bind_next and
5896 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5897 * we get the endpoint they describe and run through
5898 * the endpoint's list of IP (v4 or v6) addresses,
5899 * comparing each of the addresses with the address of
5900 * the socket sk. If we find a match, then that means
5901 * that this port/socket (sk) combination are already
5904 sk_for_each_bound(sk2, node, &pp->owner) {
5905 struct sctp_endpoint *ep2;
5906 ep2 = sctp_sk(sk2)->ep;
5909 (reuse && sk2->sk_reuse &&
5910 sk2->sk_state != SCTP_SS_LISTENING))
5913 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5914 sctp_sk(sk2), sctp_sk(sk))) {
5919 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5922 /* If there was a hash table miss, create a new port. */
5924 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
5927 /* In either case (hit or miss), make sure fastreuse is 1 only
5928 * if sk->sk_reuse is too (that is, if the caller requested
5929 * SO_REUSEADDR on this socket -sk-).
5931 if (hlist_empty(&pp->owner)) {
5932 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5936 } else if (pp->fastreuse &&
5937 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5940 /* We are set, so fill up all the data in the hash table
5941 * entry, tie the socket list information with the rest of the
5942 * sockets FIXME: Blurry, NPI (ipg).
5945 if (!sctp_sk(sk)->bind_hash) {
5946 inet_sk(sk)->inet_num = snum;
5947 sk_add_bind_node(sk, &pp->owner);
5948 sctp_sk(sk)->bind_hash = pp;
5953 sctp_spin_unlock(&head->lock);
5956 sctp_local_bh_enable();
5960 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5961 * port is requested.
5963 static int sctp_get_port(struct sock *sk, unsigned short snum)
5966 union sctp_addr addr;
5967 struct sctp_af *af = sctp_sk(sk)->pf->af;
5969 /* Set up a dummy address struct from the sk. */
5970 af->from_sk(&addr, sk);
5971 addr.v4.sin_port = htons(snum);
5973 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5974 ret = sctp_get_port_local(sk, &addr);
5980 * Move a socket to LISTENING state.
5982 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5984 struct sctp_sock *sp = sctp_sk(sk);
5985 struct sctp_endpoint *ep = sp->ep;
5986 struct crypto_hash *tfm = NULL;
5989 /* Allocate HMAC for generating cookie. */
5990 if (!sp->hmac && sp->sctp_hmac_alg) {
5991 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
5992 tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
5994 net_info_ratelimited("failed to load transform for %s: %ld\n",
5995 sp->sctp_hmac_alg, PTR_ERR(tfm));
5998 sctp_sk(sk)->hmac = tfm;
6002 * If a bind() or sctp_bindx() is not called prior to a listen()
6003 * call that allows new associations to be accepted, the system
6004 * picks an ephemeral port and will choose an address set equivalent
6005 * to binding with a wildcard address.
6007 * This is not currently spelled out in the SCTP sockets
6008 * extensions draft, but follows the practice as seen in TCP
6012 sk->sk_state = SCTP_SS_LISTENING;
6013 if (!ep->base.bind_addr.port) {
6014 if (sctp_autobind(sk))
6017 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6018 sk->sk_state = SCTP_SS_CLOSED;
6023 sk->sk_max_ack_backlog = backlog;
6024 sctp_hash_endpoint(ep);
6029 * 4.1.3 / 5.1.3 listen()
6031 * By default, new associations are not accepted for UDP style sockets.
6032 * An application uses listen() to mark a socket as being able to
6033 * accept new associations.
6035 * On TCP style sockets, applications use listen() to ready the SCTP
6036 * endpoint for accepting inbound associations.
6038 * On both types of endpoints a backlog of '0' disables listening.
6040 * Move a socket to LISTENING state.
6042 int sctp_inet_listen(struct socket *sock, int backlog)
6044 struct sock *sk = sock->sk;
6045 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6048 if (unlikely(backlog < 0))
6053 /* Peeled-off sockets are not allowed to listen(). */
6054 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6057 if (sock->state != SS_UNCONNECTED)
6060 /* If backlog is zero, disable listening. */
6062 if (sctp_sstate(sk, CLOSED))
6066 sctp_unhash_endpoint(ep);
6067 sk->sk_state = SCTP_SS_CLOSED;
6069 sctp_sk(sk)->bind_hash->fastreuse = 1;
6073 /* If we are already listening, just update the backlog */
6074 if (sctp_sstate(sk, LISTENING))
6075 sk->sk_max_ack_backlog = backlog;
6077 err = sctp_listen_start(sk, backlog);
6084 sctp_release_sock(sk);
6089 * This function is done by modeling the current datagram_poll() and the
6090 * tcp_poll(). Note that, based on these implementations, we don't
6091 * lock the socket in this function, even though it seems that,
6092 * ideally, locking or some other mechanisms can be used to ensure
6093 * the integrity of the counters (sndbuf and wmem_alloc) used
6094 * in this place. We assume that we don't need locks either until proven
6097 * Another thing to note is that we include the Async I/O support
6098 * here, again, by modeling the current TCP/UDP code. We don't have
6099 * a good way to test with it yet.
6101 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6103 struct sock *sk = sock->sk;
6104 struct sctp_sock *sp = sctp_sk(sk);
6107 poll_wait(file, sk_sleep(sk), wait);
6109 /* A TCP-style listening socket becomes readable when the accept queue
6112 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6113 return (!list_empty(&sp->ep->asocs)) ?
6114 (POLLIN | POLLRDNORM) : 0;
6118 /* Is there any exceptional events? */
6119 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6121 if (sk->sk_shutdown & RCV_SHUTDOWN)
6122 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6123 if (sk->sk_shutdown == SHUTDOWN_MASK)
6126 /* Is it readable? Reconsider this code with TCP-style support. */
6127 if (!skb_queue_empty(&sk->sk_receive_queue))
6128 mask |= POLLIN | POLLRDNORM;
6130 /* The association is either gone or not ready. */
6131 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6134 /* Is it writable? */
6135 if (sctp_writeable(sk)) {
6136 mask |= POLLOUT | POLLWRNORM;
6138 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6140 * Since the socket is not locked, the buffer
6141 * might be made available after the writeable check and
6142 * before the bit is set. This could cause a lost I/O
6143 * signal. tcp_poll() has a race breaker for this race
6144 * condition. Based on their implementation, we put
6145 * in the following code to cover it as well.
6147 if (sctp_writeable(sk))
6148 mask |= POLLOUT | POLLWRNORM;
6153 /********************************************************************
6154 * 2nd Level Abstractions
6155 ********************************************************************/
6157 static struct sctp_bind_bucket *sctp_bucket_create(
6158 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6160 struct sctp_bind_bucket *pp;
6162 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6164 SCTP_DBG_OBJCNT_INC(bind_bucket);
6167 INIT_HLIST_HEAD(&pp->owner);
6169 hlist_add_head(&pp->node, &head->chain);
6174 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6175 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6177 if (pp && hlist_empty(&pp->owner)) {
6178 __hlist_del(&pp->node);
6179 kmem_cache_free(sctp_bucket_cachep, pp);
6180 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6184 /* Release this socket's reference to a local port. */
6185 static inline void __sctp_put_port(struct sock *sk)
6187 struct sctp_bind_hashbucket *head =
6188 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6189 inet_sk(sk)->inet_num)];
6190 struct sctp_bind_bucket *pp;
6192 sctp_spin_lock(&head->lock);
6193 pp = sctp_sk(sk)->bind_hash;
6194 __sk_del_bind_node(sk);
6195 sctp_sk(sk)->bind_hash = NULL;
6196 inet_sk(sk)->inet_num = 0;
6197 sctp_bucket_destroy(pp);
6198 sctp_spin_unlock(&head->lock);
6201 void sctp_put_port(struct sock *sk)
6203 sctp_local_bh_disable();
6204 __sctp_put_port(sk);
6205 sctp_local_bh_enable();
6209 * The system picks an ephemeral port and choose an address set equivalent
6210 * to binding with a wildcard address.
6211 * One of those addresses will be the primary address for the association.
6212 * This automatically enables the multihoming capability of SCTP.
6214 static int sctp_autobind(struct sock *sk)
6216 union sctp_addr autoaddr;
6220 /* Initialize a local sockaddr structure to INADDR_ANY. */
6221 af = sctp_sk(sk)->pf->af;
6223 port = htons(inet_sk(sk)->inet_num);
6224 af->inaddr_any(&autoaddr, port);
6226 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6229 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6232 * 4.2 The cmsghdr Structure *
6234 * When ancillary data is sent or received, any number of ancillary data
6235 * objects can be specified by the msg_control and msg_controllen members of
6236 * the msghdr structure, because each object is preceded by
6237 * a cmsghdr structure defining the object's length (the cmsg_len member).
6238 * Historically Berkeley-derived implementations have passed only one object
6239 * at a time, but this API allows multiple objects to be
6240 * passed in a single call to sendmsg() or recvmsg(). The following example
6241 * shows two ancillary data objects in a control buffer.
6243 * |<--------------------------- msg_controllen -------------------------->|
6246 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6248 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6251 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6253 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6256 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6257 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6259 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6261 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6268 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
6269 sctp_cmsgs_t *cmsgs)
6271 struct cmsghdr *cmsg;
6272 struct msghdr *my_msg = (struct msghdr *)msg;
6274 for (cmsg = CMSG_FIRSTHDR(msg);
6276 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6277 if (!CMSG_OK(my_msg, cmsg))
6280 /* Should we parse this header or ignore? */
6281 if (cmsg->cmsg_level != IPPROTO_SCTP)
6284 /* Strictly check lengths following example in SCM code. */
6285 switch (cmsg->cmsg_type) {
6287 /* SCTP Socket API Extension
6288 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6290 * This cmsghdr structure provides information for
6291 * initializing new SCTP associations with sendmsg().
6292 * The SCTP_INITMSG socket option uses this same data
6293 * structure. This structure is not used for
6296 * cmsg_level cmsg_type cmsg_data[]
6297 * ------------ ------------ ----------------------
6298 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6300 if (cmsg->cmsg_len !=
6301 CMSG_LEN(sizeof(struct sctp_initmsg)))
6303 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6307 /* SCTP Socket API Extension
6308 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6310 * This cmsghdr structure specifies SCTP options for
6311 * sendmsg() and describes SCTP header information
6312 * about a received message through recvmsg().
6314 * cmsg_level cmsg_type cmsg_data[]
6315 * ------------ ------------ ----------------------
6316 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6318 if (cmsg->cmsg_len !=
6319 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6323 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6325 /* Minimally, validate the sinfo_flags. */
6326 if (cmsgs->info->sinfo_flags &
6327 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6328 SCTP_ABORT | SCTP_EOF))
6340 * Wait for a packet..
6341 * Note: This function is the same function as in core/datagram.c
6342 * with a few modifications to make lksctp work.
6344 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6349 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6351 /* Socket errors? */
6352 error = sock_error(sk);
6356 if (!skb_queue_empty(&sk->sk_receive_queue))
6359 /* Socket shut down? */
6360 if (sk->sk_shutdown & RCV_SHUTDOWN)
6363 /* Sequenced packets can come disconnected. If so we report the
6368 /* Is there a good reason to think that we may receive some data? */
6369 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6372 /* Handle signals. */
6373 if (signal_pending(current))
6376 /* Let another process have a go. Since we are going to sleep
6377 * anyway. Note: This may cause odd behaviors if the message
6378 * does not fit in the user's buffer, but this seems to be the
6379 * only way to honor MSG_DONTWAIT realistically.
6381 sctp_release_sock(sk);
6382 *timeo_p = schedule_timeout(*timeo_p);
6386 finish_wait(sk_sleep(sk), &wait);
6390 error = sock_intr_errno(*timeo_p);
6393 finish_wait(sk_sleep(sk), &wait);
6398 /* Receive a datagram.
6399 * Note: This is pretty much the same routine as in core/datagram.c
6400 * with a few changes to make lksctp work.
6402 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6403 int noblock, int *err)
6406 struct sk_buff *skb;
6409 timeo = sock_rcvtimeo(sk, noblock);
6411 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6412 timeo, MAX_SCHEDULE_TIMEOUT);
6415 /* Again only user level code calls this function,
6416 * so nothing interrupt level
6417 * will suddenly eat the receive_queue.
6419 * Look at current nfs client by the way...
6420 * However, this function was correct in any case. 8)
6422 if (flags & MSG_PEEK) {
6423 spin_lock_bh(&sk->sk_receive_queue.lock);
6424 skb = skb_peek(&sk->sk_receive_queue);
6426 atomic_inc(&skb->users);
6427 spin_unlock_bh(&sk->sk_receive_queue.lock);
6429 skb = skb_dequeue(&sk->sk_receive_queue);
6435 /* Caller is allowed not to check sk->sk_err before calling. */
6436 error = sock_error(sk);
6440 if (sk->sk_shutdown & RCV_SHUTDOWN)
6443 /* User doesn't want to wait. */
6447 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6456 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6457 static void __sctp_write_space(struct sctp_association *asoc)
6459 struct sock *sk = asoc->base.sk;
6460 struct socket *sock = sk->sk_socket;
6462 if ((sctp_wspace(asoc) > 0) && sock) {
6463 if (waitqueue_active(&asoc->wait))
6464 wake_up_interruptible(&asoc->wait);
6466 if (sctp_writeable(sk)) {
6467 wait_queue_head_t *wq = sk_sleep(sk);
6469 if (wq && waitqueue_active(wq))
6470 wake_up_interruptible(wq);
6472 /* Note that we try to include the Async I/O support
6473 * here by modeling from the current TCP/UDP code.
6474 * We have not tested with it yet.
6476 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6477 sock_wake_async(sock,
6478 SOCK_WAKE_SPACE, POLL_OUT);
6483 /* Do accounting for the sndbuf space.
6484 * Decrement the used sndbuf space of the corresponding association by the
6485 * data size which was just transmitted(freed).
6487 static void sctp_wfree(struct sk_buff *skb)
6489 struct sctp_association *asoc;
6490 struct sctp_chunk *chunk;
6493 /* Get the saved chunk pointer. */
6494 chunk = *((struct sctp_chunk **)(skb->cb));
6497 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6498 sizeof(struct sk_buff) +
6499 sizeof(struct sctp_chunk);
6501 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6504 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6506 sk->sk_wmem_queued -= skb->truesize;
6507 sk_mem_uncharge(sk, skb->truesize);
6510 __sctp_write_space(asoc);
6512 sctp_association_put(asoc);
6515 /* Do accounting for the receive space on the socket.
6516 * Accounting for the association is done in ulpevent.c
6517 * We set this as a destructor for the cloned data skbs so that
6518 * accounting is done at the correct time.
6520 void sctp_sock_rfree(struct sk_buff *skb)
6522 struct sock *sk = skb->sk;
6523 struct sctp_ulpevent *event = sctp_skb2event(skb);
6525 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6528 * Mimic the behavior of sock_rfree
6530 sk_mem_uncharge(sk, event->rmem_len);
6534 /* Helper function to wait for space in the sndbuf. */
6535 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6538 struct sock *sk = asoc->base.sk;
6540 long current_timeo = *timeo_p;
6543 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6544 asoc, (long)(*timeo_p), msg_len);
6546 /* Increment the association's refcnt. */
6547 sctp_association_hold(asoc);
6549 /* Wait on the association specific sndbuf space. */
6551 prepare_to_wait_exclusive(&asoc->wait, &wait,
6552 TASK_INTERRUPTIBLE);
6555 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6558 if (signal_pending(current))
6559 goto do_interrupted;
6560 if (msg_len <= sctp_wspace(asoc))
6563 /* Let another process have a go. Since we are going
6566 sctp_release_sock(sk);
6567 current_timeo = schedule_timeout(current_timeo);
6568 BUG_ON(sk != asoc->base.sk);
6571 *timeo_p = current_timeo;
6575 finish_wait(&asoc->wait, &wait);
6577 /* Release the association's refcnt. */
6578 sctp_association_put(asoc);
6587 err = sock_intr_errno(*timeo_p);
6595 void sctp_data_ready(struct sock *sk, int len)
6597 struct socket_wq *wq;
6600 wq = rcu_dereference(sk->sk_wq);
6601 if (wq_has_sleeper(wq))
6602 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6603 POLLRDNORM | POLLRDBAND);
6604 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6608 /* If socket sndbuf has changed, wake up all per association waiters. */
6609 void sctp_write_space(struct sock *sk)
6611 struct sctp_association *asoc;
6613 /* Wake up the tasks in each wait queue. */
6614 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6615 __sctp_write_space(asoc);
6619 /* Is there any sndbuf space available on the socket?
6621 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6622 * associations on the same socket. For a UDP-style socket with
6623 * multiple associations, it is possible for it to be "unwriteable"
6624 * prematurely. I assume that this is acceptable because
6625 * a premature "unwriteable" is better than an accidental "writeable" which
6626 * would cause an unwanted block under certain circumstances. For the 1-1
6627 * UDP-style sockets or TCP-style sockets, this code should work.
6630 static int sctp_writeable(struct sock *sk)
6634 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6640 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6641 * returns immediately with EINPROGRESS.
6643 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6645 struct sock *sk = asoc->base.sk;
6647 long current_timeo = *timeo_p;
6650 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6653 /* Increment the association's refcnt. */
6654 sctp_association_hold(asoc);
6657 prepare_to_wait_exclusive(&asoc->wait, &wait,
6658 TASK_INTERRUPTIBLE);
6661 if (sk->sk_shutdown & RCV_SHUTDOWN)
6663 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6666 if (signal_pending(current))
6667 goto do_interrupted;
6669 if (sctp_state(asoc, ESTABLISHED))
6672 /* Let another process have a go. Since we are going
6675 sctp_release_sock(sk);
6676 current_timeo = schedule_timeout(current_timeo);
6679 *timeo_p = current_timeo;
6683 finish_wait(&asoc->wait, &wait);
6685 /* Release the association's refcnt. */
6686 sctp_association_put(asoc);
6691 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6694 err = -ECONNREFUSED;
6698 err = sock_intr_errno(*timeo_p);
6706 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6708 struct sctp_endpoint *ep;
6712 ep = sctp_sk(sk)->ep;
6716 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6717 TASK_INTERRUPTIBLE);
6719 if (list_empty(&ep->asocs)) {
6720 sctp_release_sock(sk);
6721 timeo = schedule_timeout(timeo);
6726 if (!sctp_sstate(sk, LISTENING))
6730 if (!list_empty(&ep->asocs))
6733 err = sock_intr_errno(timeo);
6734 if (signal_pending(current))
6742 finish_wait(sk_sleep(sk), &wait);
6747 static void sctp_wait_for_close(struct sock *sk, long timeout)
6752 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6753 if (list_empty(&sctp_sk(sk)->ep->asocs))
6755 sctp_release_sock(sk);
6756 timeout = schedule_timeout(timeout);
6758 } while (!signal_pending(current) && timeout);
6760 finish_wait(sk_sleep(sk), &wait);
6763 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6765 struct sk_buff *frag;
6770 /* Don't forget the fragments. */
6771 skb_walk_frags(skb, frag)
6772 sctp_skb_set_owner_r_frag(frag, sk);
6775 sctp_skb_set_owner_r(skb, sk);
6778 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6779 struct sctp_association *asoc)
6781 struct inet_sock *inet = inet_sk(sk);
6782 struct inet_sock *newinet;
6784 newsk->sk_type = sk->sk_type;
6785 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6786 newsk->sk_flags = sk->sk_flags;
6787 newsk->sk_no_check = sk->sk_no_check;
6788 newsk->sk_reuse = sk->sk_reuse;
6790 newsk->sk_shutdown = sk->sk_shutdown;
6791 newsk->sk_destruct = inet_sock_destruct;
6792 newsk->sk_family = sk->sk_family;
6793 newsk->sk_protocol = IPPROTO_SCTP;
6794 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6795 newsk->sk_sndbuf = sk->sk_sndbuf;
6796 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6797 newsk->sk_lingertime = sk->sk_lingertime;
6798 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6799 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6801 newinet = inet_sk(newsk);
6803 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6804 * getsockname() and getpeername()
6806 newinet->inet_sport = inet->inet_sport;
6807 newinet->inet_saddr = inet->inet_saddr;
6808 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6809 newinet->inet_dport = htons(asoc->peer.port);
6810 newinet->pmtudisc = inet->pmtudisc;
6811 newinet->inet_id = asoc->next_tsn ^ jiffies;
6813 newinet->uc_ttl = inet->uc_ttl;
6814 newinet->mc_loop = 1;
6815 newinet->mc_ttl = 1;
6816 newinet->mc_index = 0;
6817 newinet->mc_list = NULL;
6820 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6821 * and its messages to the newsk.
6823 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6824 struct sctp_association *assoc,
6825 sctp_socket_type_t type)
6827 struct sctp_sock *oldsp = sctp_sk(oldsk);
6828 struct sctp_sock *newsp = sctp_sk(newsk);
6829 struct sctp_bind_bucket *pp; /* hash list port iterator */
6830 struct sctp_endpoint *newep = newsp->ep;
6831 struct sk_buff *skb, *tmp;
6832 struct sctp_ulpevent *event;
6833 struct sctp_bind_hashbucket *head;
6834 struct list_head tmplist;
6836 /* Migrate socket buffer sizes and all the socket level options to the
6839 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6840 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6841 /* Brute force copy old sctp opt. */
6842 if (oldsp->do_auto_asconf) {
6843 memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
6844 inet_sk_copy_descendant(newsk, oldsk);
6845 memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
6847 inet_sk_copy_descendant(newsk, oldsk);
6849 /* Restore the ep value that was overwritten with the above structure
6855 /* Hook this new socket in to the bind_hash list. */
6856 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
6857 inet_sk(oldsk)->inet_num)];
6858 sctp_local_bh_disable();
6859 sctp_spin_lock(&head->lock);
6860 pp = sctp_sk(oldsk)->bind_hash;
6861 sk_add_bind_node(newsk, &pp->owner);
6862 sctp_sk(newsk)->bind_hash = pp;
6863 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6864 sctp_spin_unlock(&head->lock);
6865 sctp_local_bh_enable();
6867 /* Copy the bind_addr list from the original endpoint to the new
6868 * endpoint so that we can handle restarts properly
6870 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6871 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6873 /* Move any messages in the old socket's receive queue that are for the
6874 * peeled off association to the new socket's receive queue.
6876 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6877 event = sctp_skb2event(skb);
6878 if (event->asoc == assoc) {
6879 __skb_unlink(skb, &oldsk->sk_receive_queue);
6880 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6881 sctp_skb_set_owner_r_frag(skb, newsk);
6885 /* Clean up any messages pending delivery due to partial
6886 * delivery. Three cases:
6887 * 1) No partial deliver; no work.
6888 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6889 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6891 skb_queue_head_init(&newsp->pd_lobby);
6892 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6894 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6895 struct sk_buff_head *queue;
6897 /* Decide which queue to move pd_lobby skbs to. */
6898 if (assoc->ulpq.pd_mode) {
6899 queue = &newsp->pd_lobby;
6901 queue = &newsk->sk_receive_queue;
6903 /* Walk through the pd_lobby, looking for skbs that
6904 * need moved to the new socket.
6906 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6907 event = sctp_skb2event(skb);
6908 if (event->asoc == assoc) {
6909 __skb_unlink(skb, &oldsp->pd_lobby);
6910 __skb_queue_tail(queue, skb);
6911 sctp_skb_set_owner_r_frag(skb, newsk);
6915 /* Clear up any skbs waiting for the partial
6916 * delivery to finish.
6918 if (assoc->ulpq.pd_mode)
6919 sctp_clear_pd(oldsk, NULL);
6923 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6924 sctp_skb_set_owner_r_frag(skb, newsk);
6926 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6927 sctp_skb_set_owner_r_frag(skb, newsk);
6929 /* Set the type of socket to indicate that it is peeled off from the
6930 * original UDP-style socket or created with the accept() call on a
6931 * TCP-style socket..
6935 /* Mark the new socket "in-use" by the user so that any packets
6936 * that may arrive on the association after we've moved it are
6937 * queued to the backlog. This prevents a potential race between
6938 * backlog processing on the old socket and new-packet processing
6939 * on the new socket.
6941 * The caller has just allocated newsk so we can guarantee that other
6942 * paths won't try to lock it and then oldsk.
6944 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6945 sctp_assoc_migrate(assoc, newsk);
6947 /* If the association on the newsk is already closed before accept()
6948 * is called, set RCV_SHUTDOWN flag.
6950 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6951 newsk->sk_shutdown |= RCV_SHUTDOWN;
6953 newsk->sk_state = SCTP_SS_ESTABLISHED;
6954 sctp_release_sock(newsk);
6958 /* This proto struct describes the ULP interface for SCTP. */
6959 struct proto sctp_prot = {
6961 .owner = THIS_MODULE,
6962 .close = sctp_close,
6963 .connect = sctp_connect,
6964 .disconnect = sctp_disconnect,
6965 .accept = sctp_accept,
6966 .ioctl = sctp_ioctl,
6967 .init = sctp_init_sock,
6968 .destroy = sctp_destroy_sock,
6969 .shutdown = sctp_shutdown,
6970 .setsockopt = sctp_setsockopt,
6971 .getsockopt = sctp_getsockopt,
6972 .sendmsg = sctp_sendmsg,
6973 .recvmsg = sctp_recvmsg,
6975 .backlog_rcv = sctp_backlog_rcv,
6977 .unhash = sctp_unhash,
6978 .get_port = sctp_get_port,
6979 .obj_size = sizeof(struct sctp_sock),
6980 .sysctl_mem = sysctl_sctp_mem,
6981 .sysctl_rmem = sysctl_sctp_rmem,
6982 .sysctl_wmem = sysctl_sctp_wmem,
6983 .memory_pressure = &sctp_memory_pressure,
6984 .enter_memory_pressure = sctp_enter_memory_pressure,
6985 .memory_allocated = &sctp_memory_allocated,
6986 .sockets_allocated = &sctp_sockets_allocated,
6989 #if IS_ENABLED(CONFIG_IPV6)
6991 struct proto sctpv6_prot = {
6993 .owner = THIS_MODULE,
6994 .close = sctp_close,
6995 .connect = sctp_connect,
6996 .disconnect = sctp_disconnect,
6997 .accept = sctp_accept,
6998 .ioctl = sctp_ioctl,
6999 .init = sctp_init_sock,
7000 .destroy = sctp_destroy_sock,
7001 .shutdown = sctp_shutdown,
7002 .setsockopt = sctp_setsockopt,
7003 .getsockopt = sctp_getsockopt,
7004 .sendmsg = sctp_sendmsg,
7005 .recvmsg = sctp_recvmsg,
7007 .backlog_rcv = sctp_backlog_rcv,
7009 .unhash = sctp_unhash,
7010 .get_port = sctp_get_port,
7011 .obj_size = sizeof(struct sctp6_sock),
7012 .sysctl_mem = sysctl_sctp_mem,
7013 .sysctl_rmem = sysctl_sctp_rmem,
7014 .sysctl_wmem = sysctl_sctp_wmem,
7015 .memory_pressure = &sctp_memory_pressure,
7016 .enter_memory_pressure = sctp_enter_memory_pressure,
7017 .memory_allocated = &sctp_memory_allocated,
7018 .sockets_allocated = &sctp_sockets_allocated,
7020 #endif /* IS_ENABLED(CONFIG_IPV6) */