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/compat.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);
113 static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
115 extern struct kmem_cache *sctp_bucket_cachep;
116 extern long sysctl_sctp_mem[3];
117 extern int sysctl_sctp_rmem[3];
118 extern int sysctl_sctp_wmem[3];
120 static int sctp_memory_pressure;
121 static atomic_long_t sctp_memory_allocated;
122 struct percpu_counter sctp_sockets_allocated;
124 static void sctp_enter_memory_pressure(struct sock *sk)
126 sctp_memory_pressure = 1;
130 /* Get the sndbuf space available at the time on the association. */
131 static inline int sctp_wspace(struct sctp_association *asoc)
135 if (asoc->ep->sndbuf_policy)
136 amt = asoc->sndbuf_used;
138 amt = sk_wmem_alloc_get(asoc->base.sk);
140 if (amt >= asoc->base.sk->sk_sndbuf) {
141 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
144 amt = sk_stream_wspace(asoc->base.sk);
149 amt = asoc->base.sk->sk_sndbuf - amt;
154 /* Increment the used sndbuf space count of the corresponding association by
155 * the size of the outgoing data chunk.
156 * Also, set the skb destructor for sndbuf accounting later.
158 * Since it is always 1-1 between chunk and skb, and also a new skb is always
159 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
160 * destructor in the data chunk skb for the purpose of the sndbuf space
163 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
165 struct sctp_association *asoc = chunk->asoc;
166 struct sock *sk = asoc->base.sk;
168 /* The sndbuf space is tracked per association. */
169 sctp_association_hold(asoc);
171 skb_set_owner_w(chunk->skb, sk);
173 chunk->skb->destructor = sctp_wfree;
174 /* Save the chunk pointer in skb for sctp_wfree to use later. */
175 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
177 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
178 sizeof(struct sk_buff) +
179 sizeof(struct sctp_chunk);
181 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
182 sk->sk_wmem_queued += chunk->skb->truesize;
183 sk_mem_charge(sk, chunk->skb->truesize);
186 /* Verify that this is a valid address. */
187 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
192 /* Verify basic sockaddr. */
193 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
197 /* Is this a valid SCTP address? */
198 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
201 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
207 /* Look up the association by its id. If this is not a UDP-style
208 * socket, the ID field is always ignored.
210 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
212 struct sctp_association *asoc = NULL;
214 /* If this is not a UDP-style socket, assoc id should be ignored. */
215 if (!sctp_style(sk, UDP)) {
216 /* Return NULL if the socket state is not ESTABLISHED. It
217 * could be a TCP-style listening socket or a socket which
218 * hasn't yet called connect() to establish an association.
220 if (!sctp_sstate(sk, ESTABLISHED))
223 /* Get the first and the only association from the list. */
224 if (!list_empty(&sctp_sk(sk)->ep->asocs))
225 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
226 struct sctp_association, asocs);
230 /* Otherwise this is a UDP-style socket. */
231 if (!id || (id == (sctp_assoc_t)-1))
234 spin_lock_bh(&sctp_assocs_id_lock);
235 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
236 spin_unlock_bh(&sctp_assocs_id_lock);
238 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
244 /* Look up the transport from an address and an assoc id. If both address and
245 * id are specified, the associations matching the address and the id should be
248 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
249 struct sockaddr_storage *addr,
252 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
253 struct sctp_transport *transport;
254 union sctp_addr *laddr = (union sctp_addr *)addr;
256 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
263 id_asoc = sctp_id2assoc(sk, id);
264 if (id_asoc && (id_asoc != addr_asoc))
267 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
268 (union sctp_addr *)addr);
273 /* API 3.1.2 bind() - UDP Style Syntax
274 * The syntax of bind() is,
276 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
278 * sd - the socket descriptor returned by socket().
279 * addr - the address structure (struct sockaddr_in or struct
280 * sockaddr_in6 [RFC 2553]),
281 * addr_len - the size of the address structure.
283 SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
289 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
292 /* Disallow binding twice. */
293 if (!sctp_sk(sk)->ep->base.bind_addr.port)
294 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
299 sctp_release_sock(sk);
304 static long sctp_get_port_local(struct sock *, union sctp_addr *);
306 /* Verify this is a valid sockaddr. */
307 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
308 union sctp_addr *addr, int len)
312 /* Check minimum size. */
313 if (len < sizeof (struct sockaddr))
316 /* V4 mapped address are really of AF_INET family */
317 if (addr->sa.sa_family == AF_INET6 &&
318 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
319 if (!opt->pf->af_supported(AF_INET, opt))
322 /* Does this PF support this AF? */
323 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
327 /* If we get this far, af is valid. */
328 af = sctp_get_af_specific(addr->sa.sa_family);
330 if (len < af->sockaddr_len)
336 /* Bind a local address either to an endpoint or to an association. */
337 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
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 && !capable(CAP_NET_BIND_SERVICE))
385 /* See if the address matches any of the addresses we may have
386 * already bound before checking against other endpoints.
388 if (sctp_bind_addr_match(bp, addr, sp))
391 /* Make sure we are allowed to bind here.
392 * The function sctp_get_port_local() does duplicate address
395 addr->v4.sin_port = htons(snum);
396 if ((ret = sctp_get_port_local(sk, addr))) {
400 /* Refresh ephemeral port. */
402 bp->port = inet_sk(sk)->inet_num;
404 /* Add the address to the bind address list.
405 * Use GFP_ATOMIC since BHs will be disabled.
407 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
409 /* Copy back into socket for getsockname() use. */
411 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
412 af->to_sk_saddr(addr, sk);
418 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
420 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
421 * at any one time. If a sender, after sending an ASCONF chunk, decides
422 * it needs to transfer another ASCONF Chunk, it MUST wait until the
423 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
424 * subsequent ASCONF. Note this restriction binds each side, so at any
425 * time two ASCONF may be in-transit on any given association (one sent
426 * from each endpoint).
428 static int sctp_send_asconf(struct sctp_association *asoc,
429 struct sctp_chunk *chunk)
433 /* If there is an outstanding ASCONF chunk, queue it for later
436 if (asoc->addip_last_asconf) {
437 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
441 /* Hold the chunk until an ASCONF_ACK is received. */
442 sctp_chunk_hold(chunk);
443 retval = sctp_primitive_ASCONF(asoc, chunk);
445 sctp_chunk_free(chunk);
447 asoc->addip_last_asconf = chunk;
453 /* Add a list of addresses as bind addresses to local endpoint or
456 * Basically run through each address specified in the addrs/addrcnt
457 * array/length pair, determine if it is IPv6 or IPv4 and call
458 * sctp_do_bind() on it.
460 * If any of them fails, then the operation will be reversed and the
461 * ones that were added will be removed.
463 * Only sctp_setsockopt_bindx() is supposed to call this function.
465 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
470 struct sockaddr *sa_addr;
473 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
477 for (cnt = 0; cnt < addrcnt; cnt++) {
478 /* The list may contain either IPv4 or IPv6 address;
479 * determine the address length for walking thru the list.
482 af = sctp_get_af_specific(sa_addr->sa_family);
488 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
491 addr_buf += af->sockaddr_len;
495 /* Failed. Cleanup the ones that have been added */
497 sctp_bindx_rem(sk, addrs, cnt);
505 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
506 * associations that are part of the endpoint indicating that a list of local
507 * addresses are added to the endpoint.
509 * If any of the addresses is already in the bind address list of the
510 * association, we do not send the chunk for that association. But it will not
511 * affect other associations.
513 * Only sctp_setsockopt_bindx() is supposed to call this function.
515 static int sctp_send_asconf_add_ip(struct sock *sk,
516 struct sockaddr *addrs,
519 struct sctp_sock *sp;
520 struct sctp_endpoint *ep;
521 struct sctp_association *asoc;
522 struct sctp_bind_addr *bp;
523 struct sctp_chunk *chunk;
524 struct sctp_sockaddr_entry *laddr;
525 union sctp_addr *addr;
526 union sctp_addr saveaddr;
533 if (!sctp_addip_enable)
539 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
540 __func__, sk, addrs, addrcnt);
542 list_for_each_entry(asoc, &ep->asocs, asocs) {
544 if (!asoc->peer.asconf_capable)
547 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
550 if (!sctp_state(asoc, ESTABLISHED))
553 /* Check if any address in the packed array of addresses is
554 * in the bind address list of the association. If so,
555 * do not send the asconf chunk to its peer, but continue with
556 * other associations.
559 for (i = 0; i < addrcnt; i++) {
561 af = sctp_get_af_specific(addr->v4.sin_family);
567 if (sctp_assoc_lookup_laddr(asoc, addr))
570 addr_buf += af->sockaddr_len;
575 /* Use the first valid address in bind addr list of
576 * association as Address Parameter of ASCONF CHUNK.
578 bp = &asoc->base.bind_addr;
579 p = bp->address_list.next;
580 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
581 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
582 addrcnt, SCTP_PARAM_ADD_IP);
588 /* Add the new addresses to the bind address list with
589 * use_as_src set to 0.
592 for (i = 0; i < addrcnt; i++) {
594 af = sctp_get_af_specific(addr->v4.sin_family);
595 memcpy(&saveaddr, addr, af->sockaddr_len);
596 retval = sctp_add_bind_addr(bp, &saveaddr,
597 SCTP_ADDR_NEW, GFP_ATOMIC);
598 addr_buf += af->sockaddr_len;
600 if (asoc->src_out_of_asoc_ok) {
601 struct sctp_transport *trans;
603 list_for_each_entry(trans,
604 &asoc->peer.transport_addr_list, transports) {
605 /* Clear the source and route cache */
606 dst_release(trans->dst);
607 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
608 2*asoc->pathmtu, 4380));
609 trans->ssthresh = asoc->peer.i.a_rwnd;
610 trans->rto = asoc->rto_initial;
611 trans->rtt = trans->srtt = trans->rttvar = 0;
612 sctp_transport_route(trans, NULL,
613 sctp_sk(asoc->base.sk));
616 retval = sctp_send_asconf(asoc, chunk);
623 /* Remove a list of addresses from bind addresses list. Do not remove the
626 * Basically run through each address specified in the addrs/addrcnt
627 * array/length pair, determine if it is IPv6 or IPv4 and call
628 * sctp_del_bind() on it.
630 * If any of them fails, then the operation will be reversed and the
631 * ones that were removed will be added back.
633 * At least one address has to be left; if only one address is
634 * available, the operation will return -EBUSY.
636 * Only sctp_setsockopt_bindx() is supposed to call this function.
638 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
640 struct sctp_sock *sp = sctp_sk(sk);
641 struct sctp_endpoint *ep = sp->ep;
643 struct sctp_bind_addr *bp = &ep->base.bind_addr;
646 union sctp_addr *sa_addr;
649 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
653 for (cnt = 0; cnt < addrcnt; cnt++) {
654 /* If the bind address list is empty or if there is only one
655 * bind address, there is nothing more to be removed (we need
656 * at least one address here).
658 if (list_empty(&bp->address_list) ||
659 (sctp_list_single_entry(&bp->address_list))) {
665 af = sctp_get_af_specific(sa_addr->sa.sa_family);
671 if (!af->addr_valid(sa_addr, sp, NULL)) {
672 retval = -EADDRNOTAVAIL;
676 if (sa_addr->v4.sin_port &&
677 sa_addr->v4.sin_port != htons(bp->port)) {
682 if (!sa_addr->v4.sin_port)
683 sa_addr->v4.sin_port = htons(bp->port);
685 /* FIXME - There is probably a need to check if sk->sk_saddr and
686 * sk->sk_rcv_addr are currently set to one of the addresses to
687 * be removed. This is something which needs to be looked into
688 * when we are fixing the outstanding issues with multi-homing
689 * socket routing and failover schemes. Refer to comments in
690 * sctp_do_bind(). -daisy
692 retval = sctp_del_bind_addr(bp, sa_addr);
694 addr_buf += af->sockaddr_len;
697 /* Failed. Add the ones that has been removed back */
699 sctp_bindx_add(sk, addrs, cnt);
707 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
708 * the associations that are part of the endpoint indicating that a list of
709 * local addresses are removed from the endpoint.
711 * If any of the addresses is already in the bind address list of the
712 * association, we do not send the chunk for that association. But it will not
713 * affect other associations.
715 * Only sctp_setsockopt_bindx() is supposed to call this function.
717 static int sctp_send_asconf_del_ip(struct sock *sk,
718 struct sockaddr *addrs,
721 struct sctp_sock *sp;
722 struct sctp_endpoint *ep;
723 struct sctp_association *asoc;
724 struct sctp_transport *transport;
725 struct sctp_bind_addr *bp;
726 struct sctp_chunk *chunk;
727 union sctp_addr *laddr;
730 struct sctp_sockaddr_entry *saddr;
736 if (!sctp_addip_enable)
742 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
743 __func__, sk, addrs, addrcnt);
745 list_for_each_entry(asoc, &ep->asocs, asocs) {
747 if (!asoc->peer.asconf_capable)
750 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
753 if (!sctp_state(asoc, ESTABLISHED))
756 /* Check if any address in the packed array of addresses is
757 * not present in the bind address list of the association.
758 * If so, do not send the asconf chunk to its peer, but
759 * continue with other associations.
762 for (i = 0; i < addrcnt; i++) {
764 af = sctp_get_af_specific(laddr->v4.sin_family);
770 if (!sctp_assoc_lookup_laddr(asoc, laddr))
773 addr_buf += af->sockaddr_len;
778 /* Find one address in the association's bind address list
779 * that is not in the packed array of addresses. This is to
780 * make sure that we do not delete all the addresses in the
783 bp = &asoc->base.bind_addr;
784 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
786 if ((laddr == NULL) && (addrcnt == 1)) {
787 if (asoc->asconf_addr_del_pending)
789 asoc->asconf_addr_del_pending =
790 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
791 if (asoc->asconf_addr_del_pending == NULL) {
795 asoc->asconf_addr_del_pending->sa.sa_family =
797 asoc->asconf_addr_del_pending->v4.sin_port =
799 if (addrs->sa_family == AF_INET) {
800 struct sockaddr_in *sin;
802 sin = (struct sockaddr_in *)addrs;
803 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
804 } else if (addrs->sa_family == AF_INET6) {
805 struct sockaddr_in6 *sin6;
807 sin6 = (struct sockaddr_in6 *)addrs;
808 ipv6_addr_copy(&asoc->asconf_addr_del_pending->v6.sin6_addr, &sin6->sin6_addr);
810 SCTP_DEBUG_PRINTK_IPADDR("send_asconf_del_ip: keep the last address asoc: %p ",
811 " at %p\n", asoc, asoc->asconf_addr_del_pending,
812 asoc->asconf_addr_del_pending);
813 asoc->src_out_of_asoc_ok = 1;
821 /* We do not need RCU protection throughout this loop
822 * because this is done under a socket lock from the
825 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
833 /* Reset use_as_src flag for the addresses in the bind address
834 * list that are to be deleted.
837 for (i = 0; i < addrcnt; i++) {
839 af = sctp_get_af_specific(laddr->v4.sin_family);
840 list_for_each_entry(saddr, &bp->address_list, list) {
841 if (sctp_cmp_addr_exact(&saddr->a, laddr))
842 saddr->state = SCTP_ADDR_DEL;
844 addr_buf += af->sockaddr_len;
847 /* Update the route and saddr entries for all the transports
848 * as some of the addresses in the bind address list are
849 * about to be deleted and cannot be used as source addresses.
851 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
853 dst_release(transport->dst);
854 sctp_transport_route(transport, NULL,
855 sctp_sk(asoc->base.sk));
859 /* We don't need to transmit ASCONF */
861 retval = sctp_send_asconf(asoc, chunk);
867 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
868 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
870 struct sock *sk = sctp_opt2sk(sp);
871 union sctp_addr *addr;
874 /* It is safe to write port space in caller. */
876 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
877 af = sctp_get_af_specific(addr->sa.sa_family);
880 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
883 if (addrw->state == SCTP_ADDR_NEW)
884 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
886 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
889 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
892 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
895 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
896 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
899 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
900 * Section 3.1.2 for this usage.
902 * addrs is a pointer to an array of one or more socket addresses. Each
903 * address is contained in its appropriate structure (i.e. struct
904 * sockaddr_in or struct sockaddr_in6) the family of the address type
905 * must be used to distinguish the address length (note that this
906 * representation is termed a "packed array" of addresses). The caller
907 * specifies the number of addresses in the array with addrcnt.
909 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
910 * -1, and sets errno to the appropriate error code.
912 * For SCTP, the port given in each socket address must be the same, or
913 * sctp_bindx() will fail, setting errno to EINVAL.
915 * The flags parameter is formed from the bitwise OR of zero or more of
916 * the following currently defined flags:
918 * SCTP_BINDX_ADD_ADDR
920 * SCTP_BINDX_REM_ADDR
922 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
923 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
924 * addresses from the association. The two flags are mutually exclusive;
925 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
926 * not remove all addresses from an association; sctp_bindx() will
927 * reject such an attempt with EINVAL.
929 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
930 * additional addresses with an endpoint after calling bind(). Or use
931 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
932 * socket is associated with so that no new association accepted will be
933 * associated with those addresses. If the endpoint supports dynamic
934 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
935 * endpoint to send the appropriate message to the peer to change the
936 * peers address lists.
938 * Adding and removing addresses from a connected association is
939 * optional functionality. Implementations that do not support this
940 * functionality should return EOPNOTSUPP.
942 * Basically do nothing but copying the addresses from user to kernel
943 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
944 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
947 * We don't use copy_from_user() for optimization: we first do the
948 * sanity checks (buffer size -fast- and access check-healthy
949 * pointer); if all of those succeed, then we can alloc the memory
950 * (expensive operation) needed to copy the data to kernel. Then we do
951 * the copying without checking the user space area
952 * (__copy_from_user()).
954 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
957 * sk The sk of the socket
958 * addrs The pointer to the addresses in user land
959 * addrssize Size of the addrs buffer
960 * op Operation to perform (add or remove, see the flags of
963 * Returns 0 if ok, <0 errno code on error.
965 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
966 struct sockaddr __user *addrs,
967 int addrs_size, int op)
969 struct sockaddr *kaddrs;
973 struct sockaddr *sa_addr;
977 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
978 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
980 if (unlikely(addrs_size <= 0))
983 /* Check the user passed a healthy pointer. */
984 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
987 /* Alloc space for the address array in kernel memory. */
988 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
989 if (unlikely(!kaddrs))
992 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
997 /* Walk through the addrs buffer and count the number of addresses. */
999 while (walk_size < addrs_size) {
1000 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1006 af = sctp_get_af_specific(sa_addr->sa_family);
1008 /* If the address family is not supported or if this address
1009 * causes the address buffer to overflow return EINVAL.
1011 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1016 addr_buf += af->sockaddr_len;
1017 walk_size += af->sockaddr_len;
1022 case SCTP_BINDX_ADD_ADDR:
1023 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1026 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1029 case SCTP_BINDX_REM_ADDR:
1030 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1033 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1047 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1049 * Common routine for handling connect() and sctp_connectx().
1050 * Connect will come in with just a single address.
1052 static int __sctp_connect(struct sock* sk,
1053 struct sockaddr *kaddrs,
1055 sctp_assoc_t *assoc_id)
1057 struct sctp_sock *sp;
1058 struct sctp_endpoint *ep;
1059 struct sctp_association *asoc = NULL;
1060 struct sctp_association *asoc2;
1061 struct sctp_transport *transport;
1069 union sctp_addr *sa_addr = NULL;
1071 unsigned short port;
1072 unsigned int f_flags = 0;
1077 /* connect() cannot be done on a socket that is already in ESTABLISHED
1078 * state - UDP-style peeled off socket or a TCP-style socket that
1079 * is already connected.
1080 * It cannot be done even on a TCP-style listening socket.
1082 if (sctp_sstate(sk, ESTABLISHED) ||
1083 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1088 /* Walk through the addrs buffer and count the number of addresses. */
1090 while (walk_size < addrs_size) {
1091 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1097 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1099 /* If the address family is not supported or if this address
1100 * causes the address buffer to overflow return EINVAL.
1102 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1107 port = ntohs(sa_addr->v4.sin_port);
1109 /* Save current address so we can work with it */
1110 memcpy(&to, sa_addr, af->sockaddr_len);
1112 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1116 /* Make sure the destination port is correctly set
1119 if (asoc && asoc->peer.port && asoc->peer.port != port)
1123 /* Check if there already is a matching association on the
1124 * endpoint (other than the one created here).
1126 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1127 if (asoc2 && asoc2 != asoc) {
1128 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1135 /* If we could not find a matching association on the endpoint,
1136 * make sure that there is no peeled-off association matching
1137 * the peer address even on another socket.
1139 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1140 err = -EADDRNOTAVAIL;
1145 /* If a bind() or sctp_bindx() is not called prior to
1146 * an sctp_connectx() call, the system picks an
1147 * ephemeral port and will choose an address set
1148 * equivalent to binding with a wildcard address.
1150 if (!ep->base.bind_addr.port) {
1151 if (sctp_autobind(sk)) {
1157 * If an unprivileged user inherits a 1-many
1158 * style socket with open associations on a
1159 * privileged port, it MAY be permitted to
1160 * accept new associations, but it SHOULD NOT
1161 * be permitted to open new associations.
1163 if (ep->base.bind_addr.port < PROT_SOCK &&
1164 !capable(CAP_NET_BIND_SERVICE)) {
1170 scope = sctp_scope(&to);
1171 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1177 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1185 /* Prime the peer's transport structures. */
1186 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1194 addr_buf += af->sockaddr_len;
1195 walk_size += af->sockaddr_len;
1198 /* In case the user of sctp_connectx() wants an association
1199 * id back, assign one now.
1202 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1207 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1212 /* Initialize sk's dport and daddr for getpeername() */
1213 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1214 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1215 af->to_sk_daddr(sa_addr, sk);
1218 /* in-kernel sockets don't generally have a file allocated to them
1219 * if all they do is call sock_create_kern().
1221 if (sk->sk_socket->file)
1222 f_flags = sk->sk_socket->file->f_flags;
1224 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1226 err = sctp_wait_for_connect(asoc, &timeo);
1227 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1228 *assoc_id = asoc->assoc_id;
1230 /* Don't free association on exit. */
1235 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1236 " kaddrs: %p err: %d\n",
1239 /* sctp_primitive_ASSOCIATE may have added this association
1240 * To the hash table, try to unhash it, just in case, its a noop
1241 * if it wasn't hashed so we're safe
1243 sctp_unhash_established(asoc);
1244 sctp_association_free(asoc);
1249 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1252 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1253 * sctp_assoc_t *asoc);
1255 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1256 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1257 * or IPv6 addresses.
1259 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1260 * Section 3.1.2 for this usage.
1262 * addrs is a pointer to an array of one or more socket addresses. Each
1263 * address is contained in its appropriate structure (i.e. struct
1264 * sockaddr_in or struct sockaddr_in6) the family of the address type
1265 * must be used to distengish the address length (note that this
1266 * representation is termed a "packed array" of addresses). The caller
1267 * specifies the number of addresses in the array with addrcnt.
1269 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1270 * the association id of the new association. On failure, sctp_connectx()
1271 * returns -1, and sets errno to the appropriate error code. The assoc_id
1272 * is not touched by the kernel.
1274 * For SCTP, the port given in each socket address must be the same, or
1275 * sctp_connectx() will fail, setting errno to EINVAL.
1277 * An application can use sctp_connectx to initiate an association with
1278 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1279 * allows a caller to specify multiple addresses at which a peer can be
1280 * reached. The way the SCTP stack uses the list of addresses to set up
1281 * the association is implementation dependent. This function only
1282 * specifies that the stack will try to make use of all the addresses in
1283 * the list when needed.
1285 * Note that the list of addresses passed in is only used for setting up
1286 * the association. It does not necessarily equal the set of addresses
1287 * the peer uses for the resulting association. If the caller wants to
1288 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1289 * retrieve them after the association has been set up.
1291 * Basically do nothing but copying the addresses from user to kernel
1292 * land and invoking either sctp_connectx(). This is used for tunneling
1293 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1295 * We don't use copy_from_user() for optimization: we first do the
1296 * sanity checks (buffer size -fast- and access check-healthy
1297 * pointer); if all of those succeed, then we can alloc the memory
1298 * (expensive operation) needed to copy the data to kernel. Then we do
1299 * the copying without checking the user space area
1300 * (__copy_from_user()).
1302 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1305 * sk The sk of the socket
1306 * addrs The pointer to the addresses in user land
1307 * addrssize Size of the addrs buffer
1309 * Returns >=0 if ok, <0 errno code on error.
1311 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1312 struct sockaddr __user *addrs,
1314 sctp_assoc_t *assoc_id)
1317 struct sockaddr *kaddrs;
1319 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1320 __func__, sk, addrs, addrs_size);
1322 if (unlikely(addrs_size <= 0))
1325 /* Check the user passed a healthy pointer. */
1326 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1329 /* Alloc space for the address array in kernel memory. */
1330 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1331 if (unlikely(!kaddrs))
1334 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1337 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1346 * This is an older interface. It's kept for backward compatibility
1347 * to the option that doesn't provide association id.
1349 SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1350 struct sockaddr __user *addrs,
1353 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1357 * New interface for the API. The since the API is done with a socket
1358 * option, to make it simple we feed back the association id is as a return
1359 * indication to the call. Error is always negative and association id is
1362 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1363 struct sockaddr __user *addrs,
1366 sctp_assoc_t assoc_id = 0;
1369 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1378 * New (hopefully final) interface for the API.
1379 * We use the sctp_getaddrs_old structure so that use-space library
1380 * can avoid any unnecessary allocations. The only different part
1381 * is that we store the actual length of the address buffer into the
1382 * addrs_num structure member. That way we can re-use the existing
1385 #ifdef CONFIG_COMPAT
1386 struct compat_sctp_getaddrs_old {
1387 sctp_assoc_t assoc_id;
1389 compat_uptr_t addrs; /* struct sockaddr * */
1393 SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1394 char __user *optval,
1397 struct sctp_getaddrs_old param;
1398 sctp_assoc_t assoc_id = 0;
1401 #ifdef CONFIG_COMPAT
1402 if (is_compat_task()) {
1403 struct compat_sctp_getaddrs_old param32;
1405 if (len < sizeof(param32))
1407 if (copy_from_user(¶m32, optval, sizeof(param32)))
1410 param.assoc_id = param32.assoc_id;
1411 param.addr_num = param32.addr_num;
1412 param.addrs = compat_ptr(param32.addrs);
1416 if (len < sizeof(param))
1418 if (copy_from_user(¶m, optval, sizeof(param)))
1422 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1423 param.addrs, param.addr_num,
1425 if (err == 0 || err == -EINPROGRESS) {
1426 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1428 if (put_user(sizeof(assoc_id), optlen))
1435 /* API 3.1.4 close() - UDP Style Syntax
1436 * Applications use close() to perform graceful shutdown (as described in
1437 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1438 * by a UDP-style socket.
1442 * ret = close(int sd);
1444 * sd - the socket descriptor of the associations to be closed.
1446 * To gracefully shutdown a specific association represented by the
1447 * UDP-style socket, an application should use the sendmsg() call,
1448 * passing no user data, but including the appropriate flag in the
1449 * ancillary data (see Section xxxx).
1451 * If sd in the close() call is a branched-off socket representing only
1452 * one association, the shutdown is performed on that association only.
1454 * 4.1.6 close() - TCP Style Syntax
1456 * Applications use close() to gracefully close down an association.
1460 * int close(int sd);
1462 * sd - the socket descriptor of the association to be closed.
1464 * After an application calls close() on a socket descriptor, no further
1465 * socket operations will succeed on that descriptor.
1467 * API 7.1.4 SO_LINGER
1469 * An application using the TCP-style socket can use this option to
1470 * perform the SCTP ABORT primitive. The linger option structure is:
1473 * int l_onoff; // option on/off
1474 * int l_linger; // linger time
1477 * To enable the option, set l_onoff to 1. If the l_linger value is set
1478 * to 0, calling close() is the same as the ABORT primitive. If the
1479 * value is set to a negative value, the setsockopt() call will return
1480 * an error. If the value is set to a positive value linger_time, the
1481 * close() can be blocked for at most linger_time ms. If the graceful
1482 * shutdown phase does not finish during this period, close() will
1483 * return but the graceful shutdown phase continues in the system.
1485 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1487 struct sctp_endpoint *ep;
1488 struct sctp_association *asoc;
1489 struct list_head *pos, *temp;
1490 unsigned int data_was_unread;
1492 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1495 sk->sk_shutdown = SHUTDOWN_MASK;
1496 sk->sk_state = SCTP_SS_CLOSING;
1498 ep = sctp_sk(sk)->ep;
1500 /* Clean up any skbs sitting on the receive queue. */
1501 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1502 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1504 /* Walk all associations on an endpoint. */
1505 list_for_each_safe(pos, temp, &ep->asocs) {
1506 asoc = list_entry(pos, struct sctp_association, asocs);
1508 if (sctp_style(sk, TCP)) {
1509 /* A closed association can still be in the list if
1510 * it belongs to a TCP-style listening socket that is
1511 * not yet accepted. If so, free it. If not, send an
1512 * ABORT or SHUTDOWN based on the linger options.
1514 if (sctp_state(asoc, CLOSED)) {
1515 sctp_unhash_established(asoc);
1516 sctp_association_free(asoc);
1521 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1522 !skb_queue_empty(&asoc->ulpq.reasm) ||
1523 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1524 struct sctp_chunk *chunk;
1526 chunk = sctp_make_abort_user(asoc, NULL, 0);
1528 sctp_primitive_ABORT(asoc, chunk);
1530 sctp_primitive_SHUTDOWN(asoc, NULL);
1533 /* On a TCP-style socket, block for at most linger_time if set. */
1534 if (sctp_style(sk, TCP) && timeout)
1535 sctp_wait_for_close(sk, timeout);
1537 /* This will run the backlog queue. */
1538 sctp_release_sock(sk);
1540 /* Supposedly, no process has access to the socket, but
1541 * the net layers still may.
1542 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1543 * held and that should be grabbed before socket lock.
1545 spin_lock_bh(&sctp_globals.addr_wq_lock);
1546 sctp_bh_lock_sock(sk);
1548 /* Hold the sock, since sk_common_release() will put sock_put()
1549 * and we have just a little more cleanup.
1552 sk_common_release(sk);
1554 sctp_bh_unlock_sock(sk);
1555 spin_unlock_bh(&sctp_globals.addr_wq_lock);
1559 SCTP_DBG_OBJCNT_DEC(sock);
1562 /* Handle EPIPE error. */
1563 static int sctp_error(struct sock *sk, int flags, int err)
1566 err = sock_error(sk) ? : -EPIPE;
1567 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1568 send_sig(SIGPIPE, current, 0);
1572 /* API 3.1.3 sendmsg() - UDP Style Syntax
1574 * An application uses sendmsg() and recvmsg() calls to transmit data to
1575 * and receive data from its peer.
1577 * ssize_t sendmsg(int socket, const struct msghdr *message,
1580 * socket - the socket descriptor of the endpoint.
1581 * message - pointer to the msghdr structure which contains a single
1582 * user message and possibly some ancillary data.
1584 * See Section 5 for complete description of the data
1587 * flags - flags sent or received with the user message, see Section
1588 * 5 for complete description of the flags.
1590 * Note: This function could use a rewrite especially when explicit
1591 * connect support comes in.
1593 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1595 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1597 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1598 struct msghdr *msg, size_t msg_len)
1600 struct sctp_sock *sp;
1601 struct sctp_endpoint *ep;
1602 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1603 struct sctp_transport *transport, *chunk_tp;
1604 struct sctp_chunk *chunk;
1606 struct sockaddr *msg_name = NULL;
1607 struct sctp_sndrcvinfo default_sinfo;
1608 struct sctp_sndrcvinfo *sinfo;
1609 struct sctp_initmsg *sinit;
1610 sctp_assoc_t associd = 0;
1611 sctp_cmsgs_t cmsgs = { NULL };
1615 __u16 sinfo_flags = 0;
1616 bool wait_connect = false;
1617 struct sctp_datamsg *datamsg;
1618 int msg_flags = msg->msg_flags;
1620 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1627 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1629 /* We cannot send a message over a TCP-style listening socket. */
1630 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1635 /* Parse out the SCTP CMSGs. */
1636 err = sctp_msghdr_parse(msg, &cmsgs);
1639 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1643 /* Fetch the destination address for this packet. This
1644 * address only selects the association--it is not necessarily
1645 * the address we will send to.
1646 * For a peeled-off socket, msg_name is ignored.
1648 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1649 int msg_namelen = msg->msg_namelen;
1651 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1656 if (msg_namelen > sizeof(to))
1657 msg_namelen = sizeof(to);
1658 memcpy(&to, msg->msg_name, msg_namelen);
1659 msg_name = msg->msg_name;
1665 /* Did the user specify SNDRCVINFO? */
1667 sinfo_flags = sinfo->sinfo_flags;
1668 associd = sinfo->sinfo_assoc_id;
1671 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1672 msg_len, sinfo_flags);
1674 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1675 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1680 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1681 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1682 * If SCTP_ABORT is set, the message length could be non zero with
1683 * the msg_iov set to the user abort reason.
1685 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1686 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1691 /* If SCTP_ADDR_OVER is set, there must be an address
1692 * specified in msg_name.
1694 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1701 SCTP_DEBUG_PRINTK("About to look up association.\n");
1705 /* If a msg_name has been specified, assume this is to be used. */
1707 /* Look for a matching association on the endpoint. */
1708 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1710 /* If we could not find a matching association on the
1711 * endpoint, make sure that it is not a TCP-style
1712 * socket that already has an association or there is
1713 * no peeled-off association on another socket.
1715 if ((sctp_style(sk, TCP) &&
1716 sctp_sstate(sk, ESTABLISHED)) ||
1717 sctp_endpoint_is_peeled_off(ep, &to)) {
1718 err = -EADDRNOTAVAIL;
1723 asoc = sctp_id2assoc(sk, associd);
1731 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1733 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1734 * socket that has an association in CLOSED state. This can
1735 * happen when an accepted socket has an association that is
1738 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1743 if (sinfo_flags & SCTP_EOF) {
1744 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1746 sctp_primitive_SHUTDOWN(asoc, NULL);
1750 if (sinfo_flags & SCTP_ABORT) {
1752 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1758 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1759 sctp_primitive_ABORT(asoc, chunk);
1765 /* Do we need to create the association? */
1767 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1769 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1774 /* Check for invalid stream against the stream counts,
1775 * either the default or the user specified stream counts.
1778 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1779 /* Check against the defaults. */
1780 if (sinfo->sinfo_stream >=
1781 sp->initmsg.sinit_num_ostreams) {
1786 /* Check against the requested. */
1787 if (sinfo->sinfo_stream >=
1788 sinit->sinit_num_ostreams) {
1796 * API 3.1.2 bind() - UDP Style Syntax
1797 * If a bind() or sctp_bindx() is not called prior to a
1798 * sendmsg() call that initiates a new association, the
1799 * system picks an ephemeral port and will choose an address
1800 * set equivalent to binding with a wildcard address.
1802 if (!ep->base.bind_addr.port) {
1803 if (sctp_autobind(sk)) {
1809 * If an unprivileged user inherits a one-to-many
1810 * style socket with open associations on a privileged
1811 * port, it MAY be permitted to accept new associations,
1812 * but it SHOULD NOT be permitted to open new
1815 if (ep->base.bind_addr.port < PROT_SOCK &&
1816 !capable(CAP_NET_BIND_SERVICE)) {
1822 scope = sctp_scope(&to);
1823 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1829 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1835 /* If the SCTP_INIT ancillary data is specified, set all
1836 * the association init values accordingly.
1839 if (sinit->sinit_num_ostreams) {
1840 asoc->c.sinit_num_ostreams =
1841 sinit->sinit_num_ostreams;
1843 if (sinit->sinit_max_instreams) {
1844 asoc->c.sinit_max_instreams =
1845 sinit->sinit_max_instreams;
1847 if (sinit->sinit_max_attempts) {
1848 asoc->max_init_attempts
1849 = sinit->sinit_max_attempts;
1851 if (sinit->sinit_max_init_timeo) {
1852 asoc->max_init_timeo =
1853 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1857 /* Prime the peer's transport structures. */
1858 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1865 /* ASSERT: we have a valid association at this point. */
1866 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1869 /* If the user didn't specify SNDRCVINFO, make up one with
1872 memset(&default_sinfo, 0, sizeof(default_sinfo));
1873 default_sinfo.sinfo_stream = asoc->default_stream;
1874 default_sinfo.sinfo_flags = asoc->default_flags;
1875 default_sinfo.sinfo_ppid = asoc->default_ppid;
1876 default_sinfo.sinfo_context = asoc->default_context;
1877 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1878 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1879 sinfo = &default_sinfo;
1882 /* API 7.1.7, the sndbuf size per association bounds the
1883 * maximum size of data that can be sent in a single send call.
1885 if (msg_len > sk->sk_sndbuf) {
1890 if (asoc->pmtu_pending)
1891 sctp_assoc_pending_pmtu(asoc);
1893 /* If fragmentation is disabled and the message length exceeds the
1894 * association fragmentation point, return EMSGSIZE. The I-D
1895 * does not specify what this error is, but this looks like
1898 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1903 /* Check for invalid stream. */
1904 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1909 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1910 if (!sctp_wspace(asoc)) {
1911 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1916 /* If an address is passed with the sendto/sendmsg call, it is used
1917 * to override the primary destination address in the TCP model, or
1918 * when SCTP_ADDR_OVER flag is set in the UDP model.
1920 if ((sctp_style(sk, TCP) && msg_name) ||
1921 (sinfo_flags & SCTP_ADDR_OVER)) {
1922 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1930 /* Auto-connect, if we aren't connected already. */
1931 if (sctp_state(asoc, CLOSED)) {
1932 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1935 wait_connect = true;
1936 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1939 /* Break the message into multiple chunks of maximum size. */
1940 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1941 if (IS_ERR(datamsg)) {
1942 err = PTR_ERR(datamsg);
1946 /* Now send the (possibly) fragmented message. */
1947 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1948 sctp_chunk_hold(chunk);
1950 /* Do accounting for the write space. */
1951 sctp_set_owner_w(chunk);
1953 chunk->transport = chunk_tp;
1956 /* Send it to the lower layers. Note: all chunks
1957 * must either fail or succeed. The lower layer
1958 * works that way today. Keep it that way or this
1961 err = sctp_primitive_SEND(asoc, datamsg);
1962 /* Did the lower layer accept the chunk? */
1964 sctp_datamsg_free(datamsg);
1966 sctp_datamsg_put(datamsg);
1968 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1975 if (unlikely(wait_connect)) {
1976 timeo = sock_sndtimeo(sk, msg_flags & MSG_DONTWAIT);
1977 sctp_wait_for_connect(asoc, &timeo);
1980 /* If we are already past ASSOCIATE, the lower
1981 * layers are responsible for association cleanup.
1987 sctp_unhash_established(asoc);
1988 sctp_association_free(asoc);
1991 sctp_release_sock(sk);
1994 return sctp_error(sk, msg_flags, err);
2001 err = sock_error(sk);
2011 /* This is an extended version of skb_pull() that removes the data from the
2012 * start of a skb even when data is spread across the list of skb's in the
2013 * frag_list. len specifies the total amount of data that needs to be removed.
2014 * when 'len' bytes could be removed from the skb, it returns 0.
2015 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2016 * could not be removed.
2018 static int sctp_skb_pull(struct sk_buff *skb, int len)
2020 struct sk_buff *list;
2021 int skb_len = skb_headlen(skb);
2024 if (len <= skb_len) {
2025 __skb_pull(skb, len);
2029 __skb_pull(skb, skb_len);
2031 skb_walk_frags(skb, list) {
2032 rlen = sctp_skb_pull(list, len);
2033 skb->len -= (len-rlen);
2034 skb->data_len -= (len-rlen);
2045 /* API 3.1.3 recvmsg() - UDP Style Syntax
2047 * ssize_t recvmsg(int socket, struct msghdr *message,
2050 * socket - the socket descriptor of the endpoint.
2051 * message - pointer to the msghdr structure which contains a single
2052 * user message and possibly some ancillary data.
2054 * See Section 5 for complete description of the data
2057 * flags - flags sent or received with the user message, see Section
2058 * 5 for complete description of the flags.
2060 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2062 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2063 struct msghdr *msg, size_t len, int noblock,
2064 int flags, int *addr_len)
2066 struct sctp_ulpevent *event = NULL;
2067 struct sctp_sock *sp = sctp_sk(sk);
2068 struct sk_buff *skb;
2073 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
2074 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
2075 "len", len, "knoblauch", noblock,
2076 "flags", flags, "addr_len", addr_len);
2080 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2085 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2089 /* Get the total length of the skb including any skb's in the
2098 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2100 event = sctp_skb2event(skb);
2105 sock_recv_ts_and_drops(msg, sk, skb);
2106 if (sctp_ulpevent_is_notification(event)) {
2107 msg->msg_flags |= MSG_NOTIFICATION;
2108 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2110 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2113 /* Check if we allow SCTP_SNDRCVINFO. */
2114 if (sp->subscribe.sctp_data_io_event)
2115 sctp_ulpevent_read_sndrcvinfo(event, msg);
2117 /* FIXME: we should be calling IP/IPv6 layers. */
2118 if (sk->sk_protinfo.af_inet.cmsg_flags)
2119 ip_cmsg_recv(msg, skb);
2124 /* If skb's length exceeds the user's buffer, update the skb and
2125 * push it back to the receive_queue so that the next call to
2126 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2128 if (skb_len > copied) {
2129 msg->msg_flags &= ~MSG_EOR;
2130 if (flags & MSG_PEEK)
2132 sctp_skb_pull(skb, copied);
2133 skb_queue_head(&sk->sk_receive_queue, skb);
2135 /* When only partial message is copied to the user, increase
2136 * rwnd by that amount. If all the data in the skb is read,
2137 * rwnd is updated when the event is freed.
2139 if (!sctp_ulpevent_is_notification(event))
2140 sctp_assoc_rwnd_increase(event->asoc, copied);
2142 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2143 (event->msg_flags & MSG_EOR))
2144 msg->msg_flags |= MSG_EOR;
2146 msg->msg_flags &= ~MSG_EOR;
2149 if (flags & MSG_PEEK) {
2150 /* Release the skb reference acquired after peeking the skb in
2151 * sctp_skb_recv_datagram().
2155 /* Free the event which includes releasing the reference to
2156 * the owner of the skb, freeing the skb and updating the
2159 sctp_ulpevent_free(event);
2162 sctp_release_sock(sk);
2166 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2168 * This option is a on/off flag. If enabled no SCTP message
2169 * fragmentation will be performed. Instead if a message being sent
2170 * exceeds the current PMTU size, the message will NOT be sent and
2171 * instead a error will be indicated to the user.
2173 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2174 char __user *optval,
2175 unsigned int optlen)
2179 if (optlen < sizeof(int))
2182 if (get_user(val, (int __user *)optval))
2185 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2190 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2191 unsigned int optlen)
2193 struct sctp_association *asoc;
2194 struct sctp_ulpevent *event;
2196 if (optlen > sizeof(struct sctp_event_subscribe))
2198 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2202 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2203 * if there is no data to be sent or retransmit, the stack will
2204 * immediately send up this notification.
2206 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2207 &sctp_sk(sk)->subscribe)) {
2208 asoc = sctp_id2assoc(sk, 0);
2210 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2211 event = sctp_ulpevent_make_sender_dry_event(asoc,
2216 sctp_ulpq_tail_event(&asoc->ulpq, event);
2223 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2225 * This socket option is applicable to the UDP-style socket only. When
2226 * set it will cause associations that are idle for more than the
2227 * specified number of seconds to automatically close. An association
2228 * being idle is defined an association that has NOT sent or received
2229 * user data. The special value of '0' indicates that no automatic
2230 * close of any associations should be performed. The option expects an
2231 * integer defining the number of seconds of idle time before an
2232 * association is closed.
2234 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2235 unsigned int optlen)
2237 struct sctp_sock *sp = sctp_sk(sk);
2239 /* Applicable to UDP-style socket only */
2240 if (sctp_style(sk, TCP))
2242 if (optlen != sizeof(int))
2244 if (copy_from_user(&sp->autoclose, optval, optlen))
2250 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2252 * Applications can enable or disable heartbeats for any peer address of
2253 * an association, modify an address's heartbeat interval, force a
2254 * heartbeat to be sent immediately, and adjust the address's maximum
2255 * number of retransmissions sent before an address is considered
2256 * unreachable. The following structure is used to access and modify an
2257 * address's parameters:
2259 * struct sctp_paddrparams {
2260 * sctp_assoc_t spp_assoc_id;
2261 * struct sockaddr_storage spp_address;
2262 * uint32_t spp_hbinterval;
2263 * uint16_t spp_pathmaxrxt;
2264 * uint32_t spp_pathmtu;
2265 * uint32_t spp_sackdelay;
2266 * uint32_t spp_flags;
2269 * spp_assoc_id - (one-to-many style socket) This is filled in the
2270 * application, and identifies the association for
2272 * spp_address - This specifies which address is of interest.
2273 * spp_hbinterval - This contains the value of the heartbeat interval,
2274 * in milliseconds. If a value of zero
2275 * is present in this field then no changes are to
2276 * be made to this parameter.
2277 * spp_pathmaxrxt - This contains the maximum number of
2278 * retransmissions before this address shall be
2279 * considered unreachable. If a value of zero
2280 * is present in this field then no changes are to
2281 * be made to this parameter.
2282 * spp_pathmtu - When Path MTU discovery is disabled the value
2283 * specified here will be the "fixed" path mtu.
2284 * Note that if the spp_address field is empty
2285 * then all associations on this address will
2286 * have this fixed path mtu set upon them.
2288 * spp_sackdelay - When delayed sack is enabled, this value specifies
2289 * the number of milliseconds that sacks will be delayed
2290 * for. This value will apply to all addresses of an
2291 * association if the spp_address field is empty. Note
2292 * also, that if delayed sack is enabled and this
2293 * value is set to 0, no change is made to the last
2294 * recorded delayed sack timer value.
2296 * spp_flags - These flags are used to control various features
2297 * on an association. The flag field may contain
2298 * zero or more of the following options.
2300 * SPP_HB_ENABLE - Enable heartbeats on the
2301 * specified address. Note that if the address
2302 * field is empty all addresses for the association
2303 * have heartbeats enabled upon them.
2305 * SPP_HB_DISABLE - Disable heartbeats on the
2306 * speicifed address. Note that if the address
2307 * field is empty all addresses for the association
2308 * will have their heartbeats disabled. Note also
2309 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2310 * mutually exclusive, only one of these two should
2311 * be specified. Enabling both fields will have
2312 * undetermined results.
2314 * SPP_HB_DEMAND - Request a user initiated heartbeat
2315 * to be made immediately.
2317 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2318 * heartbeat delayis to be set to the value of 0
2321 * SPP_PMTUD_ENABLE - This field will enable PMTU
2322 * discovery upon the specified address. Note that
2323 * if the address feild is empty then all addresses
2324 * on the association are effected.
2326 * SPP_PMTUD_DISABLE - This field will disable PMTU
2327 * discovery upon the specified address. Note that
2328 * if the address feild is empty then all addresses
2329 * on the association are effected. Not also that
2330 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2331 * exclusive. Enabling both will have undetermined
2334 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2335 * on delayed sack. The time specified in spp_sackdelay
2336 * is used to specify the sack delay for this address. Note
2337 * that if spp_address is empty then all addresses will
2338 * enable delayed sack and take on the sack delay
2339 * value specified in spp_sackdelay.
2340 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2341 * off delayed sack. If the spp_address field is blank then
2342 * delayed sack is disabled for the entire association. Note
2343 * also that this field is mutually exclusive to
2344 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2347 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2348 struct sctp_transport *trans,
2349 struct sctp_association *asoc,
2350 struct sctp_sock *sp,
2353 int sackdelay_change)
2357 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2358 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2363 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2364 * this field is ignored. Note also that a value of zero indicates
2365 * the current setting should be left unchanged.
2367 if (params->spp_flags & SPP_HB_ENABLE) {
2369 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2370 * set. This lets us use 0 value when this flag
2373 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2374 params->spp_hbinterval = 0;
2376 if (params->spp_hbinterval ||
2377 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2380 msecs_to_jiffies(params->spp_hbinterval);
2383 msecs_to_jiffies(params->spp_hbinterval);
2385 sp->hbinterval = params->spp_hbinterval;
2392 trans->param_flags =
2393 (trans->param_flags & ~SPP_HB) | hb_change;
2396 (asoc->param_flags & ~SPP_HB) | hb_change;
2399 (sp->param_flags & ~SPP_HB) | hb_change;
2403 /* When Path MTU discovery is disabled the value specified here will
2404 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2405 * include the flag SPP_PMTUD_DISABLE for this field to have any
2408 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2410 trans->pathmtu = params->spp_pathmtu;
2411 sctp_assoc_sync_pmtu(asoc);
2413 asoc->pathmtu = params->spp_pathmtu;
2414 sctp_frag_point(asoc, params->spp_pathmtu);
2416 sp->pathmtu = params->spp_pathmtu;
2422 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2423 (params->spp_flags & SPP_PMTUD_ENABLE);
2424 trans->param_flags =
2425 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2427 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2428 sctp_assoc_sync_pmtu(asoc);
2432 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2435 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2439 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2440 * value of this field is ignored. Note also that a value of zero
2441 * indicates the current setting should be left unchanged.
2443 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2446 msecs_to_jiffies(params->spp_sackdelay);
2449 msecs_to_jiffies(params->spp_sackdelay);
2451 sp->sackdelay = params->spp_sackdelay;
2455 if (sackdelay_change) {
2457 trans->param_flags =
2458 (trans->param_flags & ~SPP_SACKDELAY) |
2462 (asoc->param_flags & ~SPP_SACKDELAY) |
2466 (sp->param_flags & ~SPP_SACKDELAY) |
2471 /* Note that a value of zero indicates the current setting should be
2474 if (params->spp_pathmaxrxt) {
2476 trans->pathmaxrxt = params->spp_pathmaxrxt;
2478 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2480 sp->pathmaxrxt = params->spp_pathmaxrxt;
2487 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2488 char __user *optval,
2489 unsigned int optlen)
2491 struct sctp_paddrparams params;
2492 struct sctp_transport *trans = NULL;
2493 struct sctp_association *asoc = NULL;
2494 struct sctp_sock *sp = sctp_sk(sk);
2496 int hb_change, pmtud_change, sackdelay_change;
2498 if (optlen != sizeof(struct sctp_paddrparams))
2501 if (copy_from_user(¶ms, optval, optlen))
2504 /* Validate flags and value parameters. */
2505 hb_change = params.spp_flags & SPP_HB;
2506 pmtud_change = params.spp_flags & SPP_PMTUD;
2507 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2509 if (hb_change == SPP_HB ||
2510 pmtud_change == SPP_PMTUD ||
2511 sackdelay_change == SPP_SACKDELAY ||
2512 params.spp_sackdelay > 500 ||
2513 (params.spp_pathmtu &&
2514 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2517 /* If an address other than INADDR_ANY is specified, and
2518 * no transport is found, then the request is invalid.
2520 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
2521 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2522 params.spp_assoc_id);
2527 /* Get association, if assoc_id != 0 and the socket is a one
2528 * to many style socket, and an association was not found, then
2529 * the id was invalid.
2531 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2532 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2535 /* Heartbeat demand can only be sent on a transport or
2536 * association, but not a socket.
2538 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2541 /* Process parameters. */
2542 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2543 hb_change, pmtud_change,
2549 /* If changes are for association, also apply parameters to each
2552 if (!trans && asoc) {
2553 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2555 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2556 hb_change, pmtud_change,
2565 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2567 * This option will effect the way delayed acks are performed. This
2568 * option allows you to get or set the delayed ack time, in
2569 * milliseconds. It also allows changing the delayed ack frequency.
2570 * Changing the frequency to 1 disables the delayed sack algorithm. If
2571 * the assoc_id is 0, then this sets or gets the endpoints default
2572 * values. If the assoc_id field is non-zero, then the set or get
2573 * effects the specified association for the one to many model (the
2574 * assoc_id field is ignored by the one to one model). Note that if
2575 * sack_delay or sack_freq are 0 when setting this option, then the
2576 * current values will remain unchanged.
2578 * struct sctp_sack_info {
2579 * sctp_assoc_t sack_assoc_id;
2580 * uint32_t sack_delay;
2581 * uint32_t sack_freq;
2584 * sack_assoc_id - This parameter, indicates which association the user
2585 * is performing an action upon. Note that if this field's value is
2586 * zero then the endpoints default value is changed (effecting future
2587 * associations only).
2589 * sack_delay - This parameter contains the number of milliseconds that
2590 * the user is requesting the delayed ACK timer be set to. Note that
2591 * this value is defined in the standard to be between 200 and 500
2594 * sack_freq - This parameter contains the number of packets that must
2595 * be received before a sack is sent without waiting for the delay
2596 * timer to expire. The default value for this is 2, setting this
2597 * value to 1 will disable the delayed sack algorithm.
2600 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2601 char __user *optval, unsigned int optlen)
2603 struct sctp_sack_info params;
2604 struct sctp_transport *trans = NULL;
2605 struct sctp_association *asoc = NULL;
2606 struct sctp_sock *sp = sctp_sk(sk);
2608 if (optlen == sizeof(struct sctp_sack_info)) {
2609 if (copy_from_user(¶ms, optval, optlen))
2612 if (params.sack_delay == 0 && params.sack_freq == 0)
2614 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2615 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2616 pr_warn("Use struct sctp_sack_info instead\n");
2617 if (copy_from_user(¶ms, optval, optlen))
2620 if (params.sack_delay == 0)
2621 params.sack_freq = 1;
2623 params.sack_freq = 0;
2627 /* Validate value parameter. */
2628 if (params.sack_delay > 500)
2631 /* Get association, if sack_assoc_id != 0 and the socket is a one
2632 * to many style socket, and an association was not found, then
2633 * the id was invalid.
2635 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2636 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2639 if (params.sack_delay) {
2642 msecs_to_jiffies(params.sack_delay);
2644 (asoc->param_flags & ~SPP_SACKDELAY) |
2645 SPP_SACKDELAY_ENABLE;
2647 sp->sackdelay = params.sack_delay;
2649 (sp->param_flags & ~SPP_SACKDELAY) |
2650 SPP_SACKDELAY_ENABLE;
2654 if (params.sack_freq == 1) {
2657 (asoc->param_flags & ~SPP_SACKDELAY) |
2658 SPP_SACKDELAY_DISABLE;
2661 (sp->param_flags & ~SPP_SACKDELAY) |
2662 SPP_SACKDELAY_DISABLE;
2664 } else if (params.sack_freq > 1) {
2666 asoc->sackfreq = params.sack_freq;
2668 (asoc->param_flags & ~SPP_SACKDELAY) |
2669 SPP_SACKDELAY_ENABLE;
2671 sp->sackfreq = params.sack_freq;
2673 (sp->param_flags & ~SPP_SACKDELAY) |
2674 SPP_SACKDELAY_ENABLE;
2678 /* If change is for association, also apply to each transport. */
2680 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2682 if (params.sack_delay) {
2684 msecs_to_jiffies(params.sack_delay);
2685 trans->param_flags =
2686 (trans->param_flags & ~SPP_SACKDELAY) |
2687 SPP_SACKDELAY_ENABLE;
2689 if (params.sack_freq == 1) {
2690 trans->param_flags =
2691 (trans->param_flags & ~SPP_SACKDELAY) |
2692 SPP_SACKDELAY_DISABLE;
2693 } else if (params.sack_freq > 1) {
2694 trans->sackfreq = params.sack_freq;
2695 trans->param_flags =
2696 (trans->param_flags & ~SPP_SACKDELAY) |
2697 SPP_SACKDELAY_ENABLE;
2705 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2707 * Applications can specify protocol parameters for the default association
2708 * initialization. The option name argument to setsockopt() and getsockopt()
2711 * Setting initialization parameters is effective only on an unconnected
2712 * socket (for UDP-style sockets only future associations are effected
2713 * by the change). With TCP-style sockets, this option is inherited by
2714 * sockets derived from a listener socket.
2716 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2718 struct sctp_initmsg sinit;
2719 struct sctp_sock *sp = sctp_sk(sk);
2721 if (optlen != sizeof(struct sctp_initmsg))
2723 if (copy_from_user(&sinit, optval, optlen))
2726 if (sinit.sinit_num_ostreams)
2727 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2728 if (sinit.sinit_max_instreams)
2729 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2730 if (sinit.sinit_max_attempts)
2731 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2732 if (sinit.sinit_max_init_timeo)
2733 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2739 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2741 * Applications that wish to use the sendto() system call may wish to
2742 * specify a default set of parameters that would normally be supplied
2743 * through the inclusion of ancillary data. This socket option allows
2744 * such an application to set the default sctp_sndrcvinfo structure.
2745 * The application that wishes to use this socket option simply passes
2746 * in to this call the sctp_sndrcvinfo structure defined in Section
2747 * 5.2.2) The input parameters accepted by this call include
2748 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2749 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2750 * to this call if the caller is using the UDP model.
2752 static int sctp_setsockopt_default_send_param(struct sock *sk,
2753 char __user *optval,
2754 unsigned int optlen)
2756 struct sctp_sndrcvinfo info;
2757 struct sctp_association *asoc;
2758 struct sctp_sock *sp = sctp_sk(sk);
2760 if (optlen != sizeof(struct sctp_sndrcvinfo))
2762 if (copy_from_user(&info, optval, optlen))
2765 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2766 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2770 asoc->default_stream = info.sinfo_stream;
2771 asoc->default_flags = info.sinfo_flags;
2772 asoc->default_ppid = info.sinfo_ppid;
2773 asoc->default_context = info.sinfo_context;
2774 asoc->default_timetolive = info.sinfo_timetolive;
2776 sp->default_stream = info.sinfo_stream;
2777 sp->default_flags = info.sinfo_flags;
2778 sp->default_ppid = info.sinfo_ppid;
2779 sp->default_context = info.sinfo_context;
2780 sp->default_timetolive = info.sinfo_timetolive;
2786 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2788 * Requests that the local SCTP stack use the enclosed peer address as
2789 * the association primary. The enclosed address must be one of the
2790 * association peer's addresses.
2792 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2793 unsigned int optlen)
2795 struct sctp_prim prim;
2796 struct sctp_transport *trans;
2798 if (optlen != sizeof(struct sctp_prim))
2801 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2804 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2808 sctp_assoc_set_primary(trans->asoc, trans);
2814 * 7.1.5 SCTP_NODELAY
2816 * Turn on/off any Nagle-like algorithm. This means that packets are
2817 * generally sent as soon as possible and no unnecessary delays are
2818 * introduced, at the cost of more packets in the network. Expects an
2819 * integer boolean flag.
2821 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2822 unsigned int optlen)
2826 if (optlen < sizeof(int))
2828 if (get_user(val, (int __user *)optval))
2831 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2837 * 7.1.1 SCTP_RTOINFO
2839 * The protocol parameters used to initialize and bound retransmission
2840 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2841 * and modify these parameters.
2842 * All parameters are time values, in milliseconds. A value of 0, when
2843 * modifying the parameters, indicates that the current value should not
2847 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2849 struct sctp_rtoinfo rtoinfo;
2850 struct sctp_association *asoc;
2852 if (optlen != sizeof (struct sctp_rtoinfo))
2855 if (copy_from_user(&rtoinfo, optval, optlen))
2858 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2860 /* Set the values to the specific association */
2861 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2865 if (rtoinfo.srto_initial != 0)
2867 msecs_to_jiffies(rtoinfo.srto_initial);
2868 if (rtoinfo.srto_max != 0)
2869 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2870 if (rtoinfo.srto_min != 0)
2871 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2873 /* If there is no association or the association-id = 0
2874 * set the values to the endpoint.
2876 struct sctp_sock *sp = sctp_sk(sk);
2878 if (rtoinfo.srto_initial != 0)
2879 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2880 if (rtoinfo.srto_max != 0)
2881 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2882 if (rtoinfo.srto_min != 0)
2883 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2891 * 7.1.2 SCTP_ASSOCINFO
2893 * This option is used to tune the maximum retransmission attempts
2894 * of the association.
2895 * Returns an error if the new association retransmission value is
2896 * greater than the sum of the retransmission value of the peer.
2897 * See [SCTP] for more information.
2900 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2903 struct sctp_assocparams assocparams;
2904 struct sctp_association *asoc;
2906 if (optlen != sizeof(struct sctp_assocparams))
2908 if (copy_from_user(&assocparams, optval, optlen))
2911 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2913 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2916 /* Set the values to the specific association */
2918 if (assocparams.sasoc_asocmaxrxt != 0) {
2921 struct sctp_transport *peer_addr;
2923 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2925 path_sum += peer_addr->pathmaxrxt;
2929 /* Only validate asocmaxrxt if we have more than
2930 * one path/transport. We do this because path
2931 * retransmissions are only counted when we have more
2935 assocparams.sasoc_asocmaxrxt > path_sum)
2938 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2941 if (assocparams.sasoc_cookie_life != 0) {
2942 asoc->cookie_life.tv_sec =
2943 assocparams.sasoc_cookie_life / 1000;
2944 asoc->cookie_life.tv_usec =
2945 (assocparams.sasoc_cookie_life % 1000)
2949 /* Set the values to the endpoint */
2950 struct sctp_sock *sp = sctp_sk(sk);
2952 if (assocparams.sasoc_asocmaxrxt != 0)
2953 sp->assocparams.sasoc_asocmaxrxt =
2954 assocparams.sasoc_asocmaxrxt;
2955 if (assocparams.sasoc_cookie_life != 0)
2956 sp->assocparams.sasoc_cookie_life =
2957 assocparams.sasoc_cookie_life;
2963 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2965 * This socket option is a boolean flag which turns on or off mapped V4
2966 * addresses. If this option is turned on and the socket is type
2967 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2968 * If this option is turned off, then no mapping will be done of V4
2969 * addresses and a user will receive both PF_INET6 and PF_INET type
2970 * addresses on the socket.
2972 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2975 struct sctp_sock *sp = sctp_sk(sk);
2977 if (optlen < sizeof(int))
2979 if (get_user(val, (int __user *)optval))
2990 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2991 * This option will get or set the maximum size to put in any outgoing
2992 * SCTP DATA chunk. If a message is larger than this size it will be
2993 * fragmented by SCTP into the specified size. Note that the underlying
2994 * SCTP implementation may fragment into smaller sized chunks when the
2995 * PMTU of the underlying association is smaller than the value set by
2996 * the user. The default value for this option is '0' which indicates
2997 * the user is NOT limiting fragmentation and only the PMTU will effect
2998 * SCTP's choice of DATA chunk size. Note also that values set larger
2999 * than the maximum size of an IP datagram will effectively let SCTP
3000 * control fragmentation (i.e. the same as setting this option to 0).
3002 * The following structure is used to access and modify this parameter:
3004 * struct sctp_assoc_value {
3005 * sctp_assoc_t assoc_id;
3006 * uint32_t assoc_value;
3009 * assoc_id: This parameter is ignored for one-to-one style sockets.
3010 * For one-to-many style sockets this parameter indicates which
3011 * association the user is performing an action upon. Note that if
3012 * this field's value is zero then the endpoints default value is
3013 * changed (effecting future associations only).
3014 * assoc_value: This parameter specifies the maximum size in bytes.
3016 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3018 struct sctp_assoc_value params;
3019 struct sctp_association *asoc;
3020 struct sctp_sock *sp = sctp_sk(sk);
3023 if (optlen == sizeof(int)) {
3024 pr_warn("Use of int in maxseg socket option deprecated\n");
3025 pr_warn("Use struct sctp_assoc_value instead\n");
3026 if (copy_from_user(&val, optval, optlen))
3028 params.assoc_id = 0;
3029 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3030 if (copy_from_user(¶ms, optval, optlen))
3032 val = params.assoc_value;
3036 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3039 asoc = sctp_id2assoc(sk, params.assoc_id);
3040 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3045 val = asoc->pathmtu;
3046 val -= sp->pf->af->net_header_len;
3047 val -= sizeof(struct sctphdr) +
3048 sizeof(struct sctp_data_chunk);
3050 asoc->user_frag = val;
3051 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3053 sp->user_frag = val;
3061 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3063 * Requests that the peer mark the enclosed address as the association
3064 * primary. The enclosed address must be one of the association's
3065 * locally bound addresses. The following structure is used to make a
3066 * set primary request:
3068 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3069 unsigned int optlen)
3071 struct sctp_sock *sp;
3072 struct sctp_association *asoc = NULL;
3073 struct sctp_setpeerprim prim;
3074 struct sctp_chunk *chunk;
3080 if (!sctp_addip_enable)
3083 if (optlen != sizeof(struct sctp_setpeerprim))
3086 if (copy_from_user(&prim, optval, optlen))
3089 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3093 if (!asoc->peer.asconf_capable)
3096 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3099 if (!sctp_state(asoc, ESTABLISHED))
3102 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3106 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3107 return -EADDRNOTAVAIL;
3109 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3110 return -EADDRNOTAVAIL;
3112 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3113 chunk = sctp_make_asconf_set_prim(asoc,
3114 (union sctp_addr *)&prim.sspp_addr);
3118 err = sctp_send_asconf(asoc, chunk);
3120 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
3125 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3126 unsigned int optlen)
3128 struct sctp_setadaptation adaptation;
3130 if (optlen != sizeof(struct sctp_setadaptation))
3132 if (copy_from_user(&adaptation, optval, optlen))
3135 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3141 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3143 * The context field in the sctp_sndrcvinfo structure is normally only
3144 * used when a failed message is retrieved holding the value that was
3145 * sent down on the actual send call. This option allows the setting of
3146 * a default context on an association basis that will be received on
3147 * reading messages from the peer. This is especially helpful in the
3148 * one-2-many model for an application to keep some reference to an
3149 * internal state machine that is processing messages on the
3150 * association. Note that the setting of this value only effects
3151 * received messages from the peer and does not effect the value that is
3152 * saved with outbound messages.
3154 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3155 unsigned int optlen)
3157 struct sctp_assoc_value params;
3158 struct sctp_sock *sp;
3159 struct sctp_association *asoc;
3161 if (optlen != sizeof(struct sctp_assoc_value))
3163 if (copy_from_user(¶ms, optval, optlen))
3168 if (params.assoc_id != 0) {
3169 asoc = sctp_id2assoc(sk, params.assoc_id);
3172 asoc->default_rcv_context = params.assoc_value;
3174 sp->default_rcv_context = params.assoc_value;
3181 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3183 * This options will at a minimum specify if the implementation is doing
3184 * fragmented interleave. Fragmented interleave, for a one to many
3185 * socket, is when subsequent calls to receive a message may return
3186 * parts of messages from different associations. Some implementations
3187 * may allow you to turn this value on or off. If so, when turned off,
3188 * no fragment interleave will occur (which will cause a head of line
3189 * blocking amongst multiple associations sharing the same one to many
3190 * socket). When this option is turned on, then each receive call may
3191 * come from a different association (thus the user must receive data
3192 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3193 * association each receive belongs to.
3195 * This option takes a boolean value. A non-zero value indicates that
3196 * fragmented interleave is on. A value of zero indicates that
3197 * fragmented interleave is off.
3199 * Note that it is important that an implementation that allows this
3200 * option to be turned on, have it off by default. Otherwise an unaware
3201 * application using the one to many model may become confused and act
3204 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3205 char __user *optval,
3206 unsigned int optlen)
3210 if (optlen != sizeof(int))
3212 if (get_user(val, (int __user *)optval))
3215 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3221 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3222 * (SCTP_PARTIAL_DELIVERY_POINT)
3224 * This option will set or get the SCTP partial delivery point. This
3225 * point is the size of a message where the partial delivery API will be
3226 * invoked to help free up rwnd space for the peer. Setting this to a
3227 * lower value will cause partial deliveries to happen more often. The
3228 * calls argument is an integer that sets or gets the partial delivery
3229 * point. Note also that the call will fail if the user attempts to set
3230 * this value larger than the socket receive buffer size.
3232 * Note that any single message having a length smaller than or equal to
3233 * the SCTP partial delivery point will be delivered in one single read
3234 * call as long as the user provided buffer is large enough to hold the
3237 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3238 char __user *optval,
3239 unsigned int optlen)
3243 if (optlen != sizeof(u32))
3245 if (get_user(val, (int __user *)optval))
3248 /* Note: We double the receive buffer from what the user sets
3249 * it to be, also initial rwnd is based on rcvbuf/2.
3251 if (val > (sk->sk_rcvbuf >> 1))
3254 sctp_sk(sk)->pd_point = val;
3256 return 0; /* is this the right error code? */
3260 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3262 * This option will allow a user to change the maximum burst of packets
3263 * that can be emitted by this association. Note that the default value
3264 * is 4, and some implementations may restrict this setting so that it
3265 * can only be lowered.
3267 * NOTE: This text doesn't seem right. Do this on a socket basis with
3268 * future associations inheriting the socket value.
3270 static int sctp_setsockopt_maxburst(struct sock *sk,
3271 char __user *optval,
3272 unsigned int optlen)
3274 struct sctp_assoc_value params;
3275 struct sctp_sock *sp;
3276 struct sctp_association *asoc;
3280 if (optlen == sizeof(int)) {
3281 pr_warn("Use of int in max_burst socket option deprecated\n");
3282 pr_warn("Use struct sctp_assoc_value instead\n");
3283 if (copy_from_user(&val, optval, optlen))
3285 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3286 if (copy_from_user(¶ms, optval, optlen))
3288 val = params.assoc_value;
3289 assoc_id = params.assoc_id;
3295 if (assoc_id != 0) {
3296 asoc = sctp_id2assoc(sk, assoc_id);
3299 asoc->max_burst = val;
3301 sp->max_burst = val;
3307 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3309 * This set option adds a chunk type that the user is requesting to be
3310 * received only in an authenticated way. Changes to the list of chunks
3311 * will only effect future associations on the socket.
3313 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3314 char __user *optval,
3315 unsigned int optlen)
3317 struct sctp_authchunk val;
3319 if (!sctp_auth_enable)
3322 if (optlen != sizeof(struct sctp_authchunk))
3324 if (copy_from_user(&val, optval, optlen))
3327 switch (val.sauth_chunk) {
3329 case SCTP_CID_INIT_ACK:
3330 case SCTP_CID_SHUTDOWN_COMPLETE:
3335 /* add this chunk id to the endpoint */
3336 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3340 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3342 * This option gets or sets the list of HMAC algorithms that the local
3343 * endpoint requires the peer to use.
3345 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3346 char __user *optval,
3347 unsigned int optlen)
3349 struct sctp_hmacalgo *hmacs;
3353 if (!sctp_auth_enable)
3356 if (optlen < sizeof(struct sctp_hmacalgo))
3359 hmacs= memdup_user(optval, optlen);
3361 return PTR_ERR(hmacs);
3363 idents = hmacs->shmac_num_idents;
3364 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3365 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3370 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3377 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3379 * This option will set a shared secret key which is used to build an
3380 * association shared key.
3382 static int sctp_setsockopt_auth_key(struct sock *sk,
3383 char __user *optval,
3384 unsigned int optlen)
3386 struct sctp_authkey *authkey;
3387 struct sctp_association *asoc;
3390 if (!sctp_auth_enable)
3393 if (optlen <= sizeof(struct sctp_authkey))
3396 authkey= memdup_user(optval, optlen);
3397 if (IS_ERR(authkey))
3398 return PTR_ERR(authkey);
3400 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3405 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3406 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3411 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3418 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3420 * This option will get or set the active shared key to be used to build
3421 * the association shared key.
3423 static int sctp_setsockopt_active_key(struct sock *sk,
3424 char __user *optval,
3425 unsigned int optlen)
3427 struct sctp_authkeyid val;
3428 struct sctp_association *asoc;
3430 if (!sctp_auth_enable)
3433 if (optlen != sizeof(struct sctp_authkeyid))
3435 if (copy_from_user(&val, optval, optlen))
3438 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3439 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3442 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3443 val.scact_keynumber);
3447 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3449 * This set option will delete a shared secret key from use.
3451 static int sctp_setsockopt_del_key(struct sock *sk,
3452 char __user *optval,
3453 unsigned int optlen)
3455 struct sctp_authkeyid val;
3456 struct sctp_association *asoc;
3458 if (!sctp_auth_enable)
3461 if (optlen != sizeof(struct sctp_authkeyid))
3463 if (copy_from_user(&val, optval, optlen))
3466 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3467 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3470 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3471 val.scact_keynumber);
3476 * 8.1.23 SCTP_AUTO_ASCONF
3478 * This option will enable or disable the use of the automatic generation of
3479 * ASCONF chunks to add and delete addresses to an existing association. Note
3480 * that this option has two caveats namely: a) it only affects sockets that
3481 * are bound to all addresses available to the SCTP stack, and b) the system
3482 * administrator may have an overriding control that turns the ASCONF feature
3483 * off no matter what setting the socket option may have.
3484 * This option expects an integer boolean flag, where a non-zero value turns on
3485 * the option, and a zero value turns off the option.
3486 * Note. In this implementation, socket operation overrides default parameter
3487 * being set by sysctl as well as FreeBSD implementation
3489 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3490 unsigned int optlen)
3493 struct sctp_sock *sp = sctp_sk(sk);
3495 if (optlen < sizeof(int))
3497 if (get_user(val, (int __user *)optval))
3499 if (!sctp_is_ep_boundall(sk) && val)
3501 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3504 spin_lock_bh(&sctp_globals.addr_wq_lock);
3505 if (val == 0 && sp->do_auto_asconf) {
3506 list_del(&sp->auto_asconf_list);
3507 sp->do_auto_asconf = 0;
3508 } else if (val && !sp->do_auto_asconf) {
3509 list_add_tail(&sp->auto_asconf_list,
3510 &sctp_auto_asconf_splist);
3511 sp->do_auto_asconf = 1;
3513 spin_unlock_bh(&sctp_globals.addr_wq_lock);
3518 /* API 6.2 setsockopt(), getsockopt()
3520 * Applications use setsockopt() and getsockopt() to set or retrieve
3521 * socket options. Socket options are used to change the default
3522 * behavior of sockets calls. They are described in Section 7.
3526 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3527 * int __user *optlen);
3528 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3531 * sd - the socket descript.
3532 * level - set to IPPROTO_SCTP for all SCTP options.
3533 * optname - the option name.
3534 * optval - the buffer to store the value of the option.
3535 * optlen - the size of the buffer.
3537 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3538 char __user *optval, unsigned int optlen)
3542 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3545 /* I can hardly begin to describe how wrong this is. This is
3546 * so broken as to be worse than useless. The API draft
3547 * REALLY is NOT helpful here... I am not convinced that the
3548 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3549 * are at all well-founded.
3551 if (level != SOL_SCTP) {
3552 struct sctp_af *af = sctp_sk(sk)->pf->af;
3553 retval = af->setsockopt(sk, level, optname, optval, optlen);
3560 case SCTP_SOCKOPT_BINDX_ADD:
3561 /* 'optlen' is the size of the addresses buffer. */
3562 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3563 optlen, SCTP_BINDX_ADD_ADDR);
3566 case SCTP_SOCKOPT_BINDX_REM:
3567 /* 'optlen' is the size of the addresses buffer. */
3568 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3569 optlen, SCTP_BINDX_REM_ADDR);
3572 case SCTP_SOCKOPT_CONNECTX_OLD:
3573 /* 'optlen' is the size of the addresses buffer. */
3574 retval = sctp_setsockopt_connectx_old(sk,
3575 (struct sockaddr __user *)optval,
3579 case SCTP_SOCKOPT_CONNECTX:
3580 /* 'optlen' is the size of the addresses buffer. */
3581 retval = sctp_setsockopt_connectx(sk,
3582 (struct sockaddr __user *)optval,
3586 case SCTP_DISABLE_FRAGMENTS:
3587 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3591 retval = sctp_setsockopt_events(sk, optval, optlen);
3594 case SCTP_AUTOCLOSE:
3595 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3598 case SCTP_PEER_ADDR_PARAMS:
3599 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3602 case SCTP_DELAYED_SACK:
3603 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3605 case SCTP_PARTIAL_DELIVERY_POINT:
3606 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3610 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3612 case SCTP_DEFAULT_SEND_PARAM:
3613 retval = sctp_setsockopt_default_send_param(sk, optval,
3616 case SCTP_PRIMARY_ADDR:
3617 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3619 case SCTP_SET_PEER_PRIMARY_ADDR:
3620 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3623 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3626 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3628 case SCTP_ASSOCINFO:
3629 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3631 case SCTP_I_WANT_MAPPED_V4_ADDR:
3632 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3635 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3637 case SCTP_ADAPTATION_LAYER:
3638 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3641 retval = sctp_setsockopt_context(sk, optval, optlen);
3643 case SCTP_FRAGMENT_INTERLEAVE:
3644 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3646 case SCTP_MAX_BURST:
3647 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3649 case SCTP_AUTH_CHUNK:
3650 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3652 case SCTP_HMAC_IDENT:
3653 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3656 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3658 case SCTP_AUTH_ACTIVE_KEY:
3659 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3661 case SCTP_AUTH_DELETE_KEY:
3662 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3664 case SCTP_AUTO_ASCONF:
3665 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3668 retval = -ENOPROTOOPT;
3672 sctp_release_sock(sk);
3678 /* API 3.1.6 connect() - UDP Style Syntax
3680 * An application may use the connect() call in the UDP model to initiate an
3681 * association without sending data.
3685 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3687 * sd: the socket descriptor to have a new association added to.
3689 * nam: the address structure (either struct sockaddr_in or struct
3690 * sockaddr_in6 defined in RFC2553 [7]).
3692 * len: the size of the address.
3694 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3702 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3703 __func__, sk, addr, addr_len);
3705 /* Validate addr_len before calling common connect/connectx routine. */
3706 af = sctp_get_af_specific(addr->sa_family);
3707 if (!af || addr_len < af->sockaddr_len) {
3710 /* Pass correct addr len to common routine (so it knows there
3711 * is only one address being passed.
3713 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3716 sctp_release_sock(sk);
3720 /* FIXME: Write comments. */
3721 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3723 return -EOPNOTSUPP; /* STUB */
3726 /* 4.1.4 accept() - TCP Style Syntax
3728 * Applications use accept() call to remove an established SCTP
3729 * association from the accept queue of the endpoint. A new socket
3730 * descriptor will be returned from accept() to represent the newly
3731 * formed association.
3733 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3735 struct sctp_sock *sp;
3736 struct sctp_endpoint *ep;
3737 struct sock *newsk = NULL;
3738 struct sctp_association *asoc;
3747 if (!sctp_style(sk, TCP)) {
3748 error = -EOPNOTSUPP;
3752 if (!sctp_sstate(sk, LISTENING)) {
3757 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3759 error = sctp_wait_for_accept(sk, timeo);
3763 /* We treat the list of associations on the endpoint as the accept
3764 * queue and pick the first association on the list.
3766 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3768 newsk = sp->pf->create_accept_sk(sk, asoc);
3774 /* Populate the fields of the newsk from the oldsk and migrate the
3775 * asoc to the newsk.
3777 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3780 sctp_release_sock(sk);
3785 /* The SCTP ioctl handler. */
3786 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3793 * SEQPACKET-style sockets in LISTENING state are valid, for
3794 * SCTP, so only discard TCP-style sockets in LISTENING state.
3796 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3801 struct sk_buff *skb;
3802 unsigned int amount = 0;
3804 skb = skb_peek(&sk->sk_receive_queue);
3807 * We will only return the amount of this packet since
3808 * that is all that will be read.
3812 rc = put_user(amount, (int __user *)arg);
3820 sctp_release_sock(sk);
3824 /* This is the function which gets called during socket creation to
3825 * initialized the SCTP-specific portion of the sock.
3826 * The sock structure should already be zero-filled memory.
3828 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3830 struct sctp_endpoint *ep;
3831 struct sctp_sock *sp;
3833 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3837 /* Initialize the SCTP per socket area. */
3838 switch (sk->sk_type) {
3839 case SOCK_SEQPACKET:
3840 sp->type = SCTP_SOCKET_UDP;
3843 sp->type = SCTP_SOCKET_TCP;
3846 return -ESOCKTNOSUPPORT;
3849 /* Initialize default send parameters. These parameters can be
3850 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3852 sp->default_stream = 0;
3853 sp->default_ppid = 0;
3854 sp->default_flags = 0;
3855 sp->default_context = 0;
3856 sp->default_timetolive = 0;
3858 sp->default_rcv_context = 0;
3859 sp->max_burst = sctp_max_burst;
3861 /* Initialize default setup parameters. These parameters
3862 * can be modified with the SCTP_INITMSG socket option or
3863 * overridden by the SCTP_INIT CMSG.
3865 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3866 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3867 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3868 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3870 /* Initialize default RTO related parameters. These parameters can
3871 * be modified for with the SCTP_RTOINFO socket option.
3873 sp->rtoinfo.srto_initial = sctp_rto_initial;
3874 sp->rtoinfo.srto_max = sctp_rto_max;
3875 sp->rtoinfo.srto_min = sctp_rto_min;
3877 /* Initialize default association related parameters. These parameters
3878 * can be modified with the SCTP_ASSOCINFO socket option.
3880 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3881 sp->assocparams.sasoc_number_peer_destinations = 0;
3882 sp->assocparams.sasoc_peer_rwnd = 0;
3883 sp->assocparams.sasoc_local_rwnd = 0;
3884 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3886 /* Initialize default event subscriptions. By default, all the
3889 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3891 /* Default Peer Address Parameters. These defaults can
3892 * be modified via SCTP_PEER_ADDR_PARAMS
3894 sp->hbinterval = sctp_hb_interval;
3895 sp->pathmaxrxt = sctp_max_retrans_path;
3896 sp->pathmtu = 0; // allow default discovery
3897 sp->sackdelay = sctp_sack_timeout;
3899 sp->param_flags = SPP_HB_ENABLE |
3901 SPP_SACKDELAY_ENABLE;
3903 /* If enabled no SCTP message fragmentation will be performed.
3904 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3906 sp->disable_fragments = 0;
3908 /* Enable Nagle algorithm by default. */
3911 /* Enable by default. */
3914 /* Auto-close idle associations after the configured
3915 * number of seconds. A value of 0 disables this
3916 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3917 * for UDP-style sockets only.
3921 /* User specified fragmentation limit. */
3924 sp->adaptation_ind = 0;
3926 sp->pf = sctp_get_pf_specific(sk->sk_family);
3928 /* Control variables for partial data delivery. */
3929 atomic_set(&sp->pd_mode, 0);
3930 skb_queue_head_init(&sp->pd_lobby);
3931 sp->frag_interleave = 0;
3933 /* Create a per socket endpoint structure. Even if we
3934 * change the data structure relationships, this may still
3935 * be useful for storing pre-connect address information.
3937 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3944 SCTP_DBG_OBJCNT_INC(sock);
3947 percpu_counter_inc(&sctp_sockets_allocated);
3948 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3950 /* Nothing can fail after this block, otherwise
3951 * sctp_destroy_sock() will be called without addr_wq_lock held
3953 if (sctp_default_auto_asconf) {
3954 spin_lock(&sctp_globals.addr_wq_lock);
3955 list_add_tail(&sp->auto_asconf_list,
3956 &sctp_auto_asconf_splist);
3957 sp->do_auto_asconf = 1;
3958 spin_unlock(&sctp_globals.addr_wq_lock);
3960 sp->do_auto_asconf = 0;
3968 /* Cleanup any SCTP per socket resources. Must be called with
3969 * sctp_globals.addr_wq_lock held if sp->do_auto_asconf is true
3971 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3973 struct sctp_sock *sp;
3975 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3977 /* Release our hold on the endpoint. */
3979 /* This could happen during socket init, thus we bail out
3980 * early, since the rest of the below is not setup either.
3985 if (sp->do_auto_asconf) {
3986 sp->do_auto_asconf = 0;
3987 list_del(&sp->auto_asconf_list);
3989 sctp_endpoint_free(sp->ep);
3991 percpu_counter_dec(&sctp_sockets_allocated);
3992 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
3996 /* API 4.1.7 shutdown() - TCP Style Syntax
3997 * int shutdown(int socket, int how);
3999 * sd - the socket descriptor of the association to be closed.
4000 * how - Specifies the type of shutdown. The values are
4003 * Disables further receive operations. No SCTP
4004 * protocol action is taken.
4006 * Disables further send operations, and initiates
4007 * the SCTP shutdown sequence.
4009 * Disables further send and receive operations
4010 * and initiates the SCTP shutdown sequence.
4012 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
4014 struct sctp_endpoint *ep;
4015 struct sctp_association *asoc;
4017 if (!sctp_style(sk, TCP))
4020 if (how & SEND_SHUTDOWN) {
4021 ep = sctp_sk(sk)->ep;
4022 if (!list_empty(&ep->asocs)) {
4023 asoc = list_entry(ep->asocs.next,
4024 struct sctp_association, asocs);
4025 sctp_primitive_SHUTDOWN(asoc, NULL);
4030 /* 7.2.1 Association Status (SCTP_STATUS)
4032 * Applications can retrieve current status information about an
4033 * association, including association state, peer receiver window size,
4034 * number of unacked data chunks, and number of data chunks pending
4035 * receipt. This information is read-only.
4037 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4038 char __user *optval,
4041 struct sctp_status status;
4042 struct sctp_association *asoc = NULL;
4043 struct sctp_transport *transport;
4044 sctp_assoc_t associd;
4047 if (len < sizeof(status)) {
4052 len = sizeof(status);
4053 if (copy_from_user(&status, optval, len)) {
4058 associd = status.sstat_assoc_id;
4059 asoc = sctp_id2assoc(sk, associd);
4065 transport = asoc->peer.primary_path;
4067 status.sstat_assoc_id = sctp_assoc2id(asoc);
4068 status.sstat_state = asoc->state;
4069 status.sstat_rwnd = asoc->peer.rwnd;
4070 status.sstat_unackdata = asoc->unack_data;
4072 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4073 status.sstat_instrms = asoc->c.sinit_max_instreams;
4074 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4075 status.sstat_fragmentation_point = asoc->frag_point;
4076 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4077 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4078 transport->af_specific->sockaddr_len);
4079 /* Map ipv4 address into v4-mapped-on-v6 address. */
4080 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4081 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4082 status.sstat_primary.spinfo_state = transport->state;
4083 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4084 status.sstat_primary.spinfo_srtt = transport->srtt;
4085 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4086 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4088 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4089 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4091 if (put_user(len, optlen)) {
4096 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
4097 len, status.sstat_state, status.sstat_rwnd,
4098 status.sstat_assoc_id);
4100 if (copy_to_user(optval, &status, len)) {
4110 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4112 * Applications can retrieve information about a specific peer address
4113 * of an association, including its reachability state, congestion
4114 * window, and retransmission timer values. This information is
4117 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4118 char __user *optval,
4121 struct sctp_paddrinfo pinfo;
4122 struct sctp_transport *transport;
4125 if (len < sizeof(pinfo)) {
4130 len = sizeof(pinfo);
4131 if (copy_from_user(&pinfo, optval, len)) {
4136 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4137 pinfo.spinfo_assoc_id);
4141 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4142 pinfo.spinfo_state = transport->state;
4143 pinfo.spinfo_cwnd = transport->cwnd;
4144 pinfo.spinfo_srtt = transport->srtt;
4145 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4146 pinfo.spinfo_mtu = transport->pathmtu;
4148 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4149 pinfo.spinfo_state = SCTP_ACTIVE;
4151 if (put_user(len, optlen)) {
4156 if (copy_to_user(optval, &pinfo, len)) {
4165 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4167 * This option is a on/off flag. If enabled no SCTP message
4168 * fragmentation will be performed. Instead if a message being sent
4169 * exceeds the current PMTU size, the message will NOT be sent and
4170 * instead a error will be indicated to the user.
4172 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4173 char __user *optval, int __user *optlen)
4177 if (len < sizeof(int))
4181 val = (sctp_sk(sk)->disable_fragments == 1);
4182 if (put_user(len, optlen))
4184 if (copy_to_user(optval, &val, len))
4189 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4191 * This socket option is used to specify various notifications and
4192 * ancillary data the user wishes to receive.
4194 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4199 if (len > sizeof(struct sctp_event_subscribe))
4200 len = sizeof(struct sctp_event_subscribe);
4201 if (put_user(len, optlen))
4203 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4208 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4210 * This socket option is applicable to the UDP-style socket only. When
4211 * set it will cause associations that are idle for more than the
4212 * specified number of seconds to automatically close. An association
4213 * being idle is defined an association that has NOT sent or received
4214 * user data. The special value of '0' indicates that no automatic
4215 * close of any associations should be performed. The option expects an
4216 * integer defining the number of seconds of idle time before an
4217 * association is closed.
4219 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4221 /* Applicable to UDP-style socket only */
4222 if (sctp_style(sk, TCP))
4224 if (len < sizeof(int))
4227 if (put_user(len, optlen))
4229 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4234 /* Helper routine to branch off an association to a new socket. */
4235 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
4236 struct socket **sockp)
4238 struct sock *sk = asoc->base.sk;
4239 struct socket *sock;
4243 /* An association cannot be branched off from an already peeled-off
4244 * socket, nor is this supported for tcp style sockets.
4246 if (!sctp_style(sk, UDP))
4249 /* Create a new socket. */
4250 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4254 sctp_copy_sock(sock->sk, sk, asoc);
4256 /* Make peeled-off sockets more like 1-1 accepted sockets.
4257 * Set the daddr and initialize id to something more random
4259 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4260 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4262 /* Populate the fields of the newsk from the oldsk and migrate the
4263 * asoc to the newsk.
4265 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4272 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4274 sctp_peeloff_arg_t peeloff;
4275 struct socket *newsock;
4277 struct sctp_association *asoc;
4279 if (len < sizeof(sctp_peeloff_arg_t))
4281 len = sizeof(sctp_peeloff_arg_t);
4282 if (copy_from_user(&peeloff, optval, len))
4285 asoc = sctp_id2assoc(sk, peeloff.associd);
4291 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
4293 retval = sctp_do_peeloff(asoc, &newsock);
4297 /* Map the socket to an unused fd that can be returned to the user. */
4298 retval = sock_map_fd(newsock, 0);
4300 sock_release(newsock);
4304 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
4305 __func__, sk, asoc, newsock->sk, retval);
4307 /* Return the fd mapped to the new socket. */
4308 peeloff.sd = retval;
4309 if (put_user(len, optlen))
4311 if (copy_to_user(optval, &peeloff, len))
4318 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4320 * Applications can enable or disable heartbeats for any peer address of
4321 * an association, modify an address's heartbeat interval, force a
4322 * heartbeat to be sent immediately, and adjust the address's maximum
4323 * number of retransmissions sent before an address is considered
4324 * unreachable. The following structure is used to access and modify an
4325 * address's parameters:
4327 * struct sctp_paddrparams {
4328 * sctp_assoc_t spp_assoc_id;
4329 * struct sockaddr_storage spp_address;
4330 * uint32_t spp_hbinterval;
4331 * uint16_t spp_pathmaxrxt;
4332 * uint32_t spp_pathmtu;
4333 * uint32_t spp_sackdelay;
4334 * uint32_t spp_flags;
4337 * spp_assoc_id - (one-to-many style socket) This is filled in the
4338 * application, and identifies the association for
4340 * spp_address - This specifies which address is of interest.
4341 * spp_hbinterval - This contains the value of the heartbeat interval,
4342 * in milliseconds. If a value of zero
4343 * is present in this field then no changes are to
4344 * be made to this parameter.
4345 * spp_pathmaxrxt - This contains the maximum number of
4346 * retransmissions before this address shall be
4347 * considered unreachable. If a value of zero
4348 * is present in this field then no changes are to
4349 * be made to this parameter.
4350 * spp_pathmtu - When Path MTU discovery is disabled the value
4351 * specified here will be the "fixed" path mtu.
4352 * Note that if the spp_address field is empty
4353 * then all associations on this address will
4354 * have this fixed path mtu set upon them.
4356 * spp_sackdelay - When delayed sack is enabled, this value specifies
4357 * the number of milliseconds that sacks will be delayed
4358 * for. This value will apply to all addresses of an
4359 * association if the spp_address field is empty. Note
4360 * also, that if delayed sack is enabled and this
4361 * value is set to 0, no change is made to the last
4362 * recorded delayed sack timer value.
4364 * spp_flags - These flags are used to control various features
4365 * on an association. The flag field may contain
4366 * zero or more of the following options.
4368 * SPP_HB_ENABLE - Enable heartbeats on the
4369 * specified address. Note that if the address
4370 * field is empty all addresses for the association
4371 * have heartbeats enabled upon them.
4373 * SPP_HB_DISABLE - Disable heartbeats on the
4374 * speicifed address. Note that if the address
4375 * field is empty all addresses for the association
4376 * will have their heartbeats disabled. Note also
4377 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4378 * mutually exclusive, only one of these two should
4379 * be specified. Enabling both fields will have
4380 * undetermined results.
4382 * SPP_HB_DEMAND - Request a user initiated heartbeat
4383 * to be made immediately.
4385 * SPP_PMTUD_ENABLE - This field will enable PMTU
4386 * discovery upon the specified address. Note that
4387 * if the address feild is empty then all addresses
4388 * on the association are effected.
4390 * SPP_PMTUD_DISABLE - This field will disable PMTU
4391 * discovery upon the specified address. Note that
4392 * if the address feild is empty then all addresses
4393 * on the association are effected. Not also that
4394 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4395 * exclusive. Enabling both will have undetermined
4398 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4399 * on delayed sack. The time specified in spp_sackdelay
4400 * is used to specify the sack delay for this address. Note
4401 * that if spp_address is empty then all addresses will
4402 * enable delayed sack and take on the sack delay
4403 * value specified in spp_sackdelay.
4404 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4405 * off delayed sack. If the spp_address field is blank then
4406 * delayed sack is disabled for the entire association. Note
4407 * also that this field is mutually exclusive to
4408 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4411 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4412 char __user *optval, int __user *optlen)
4414 struct sctp_paddrparams params;
4415 struct sctp_transport *trans = NULL;
4416 struct sctp_association *asoc = NULL;
4417 struct sctp_sock *sp = sctp_sk(sk);
4419 if (len < sizeof(struct sctp_paddrparams))
4421 len = sizeof(struct sctp_paddrparams);
4422 if (copy_from_user(¶ms, optval, len))
4425 /* If an address other than INADDR_ANY is specified, and
4426 * no transport is found, then the request is invalid.
4428 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
4429 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4430 params.spp_assoc_id);
4432 SCTP_DEBUG_PRINTK("Failed no transport\n");
4437 /* Get association, if assoc_id != 0 and the socket is a one
4438 * to many style socket, and an association was not found, then
4439 * the id was invalid.
4441 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4442 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4443 SCTP_DEBUG_PRINTK("Failed no association\n");
4448 /* Fetch transport values. */
4449 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4450 params.spp_pathmtu = trans->pathmtu;
4451 params.spp_pathmaxrxt = trans->pathmaxrxt;
4452 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4454 /*draft-11 doesn't say what to return in spp_flags*/
4455 params.spp_flags = trans->param_flags;
4457 /* Fetch association values. */
4458 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4459 params.spp_pathmtu = asoc->pathmtu;
4460 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4461 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4463 /*draft-11 doesn't say what to return in spp_flags*/
4464 params.spp_flags = asoc->param_flags;
4466 /* Fetch socket values. */
4467 params.spp_hbinterval = sp->hbinterval;
4468 params.spp_pathmtu = sp->pathmtu;
4469 params.spp_sackdelay = sp->sackdelay;
4470 params.spp_pathmaxrxt = sp->pathmaxrxt;
4472 /*draft-11 doesn't say what to return in spp_flags*/
4473 params.spp_flags = sp->param_flags;
4476 if (copy_to_user(optval, ¶ms, len))
4479 if (put_user(len, optlen))
4486 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4488 * This option will effect the way delayed acks are performed. This
4489 * option allows you to get or set the delayed ack time, in
4490 * milliseconds. It also allows changing the delayed ack frequency.
4491 * Changing the frequency to 1 disables the delayed sack algorithm. If
4492 * the assoc_id is 0, then this sets or gets the endpoints default
4493 * values. If the assoc_id field is non-zero, then the set or get
4494 * effects the specified association for the one to many model (the
4495 * assoc_id field is ignored by the one to one model). Note that if
4496 * sack_delay or sack_freq are 0 when setting this option, then the
4497 * current values will remain unchanged.
4499 * struct sctp_sack_info {
4500 * sctp_assoc_t sack_assoc_id;
4501 * uint32_t sack_delay;
4502 * uint32_t sack_freq;
4505 * sack_assoc_id - This parameter, indicates which association the user
4506 * is performing an action upon. Note that if this field's value is
4507 * zero then the endpoints default value is changed (effecting future
4508 * associations only).
4510 * sack_delay - This parameter contains the number of milliseconds that
4511 * the user is requesting the delayed ACK timer be set to. Note that
4512 * this value is defined in the standard to be between 200 and 500
4515 * sack_freq - This parameter contains the number of packets that must
4516 * be received before a sack is sent without waiting for the delay
4517 * timer to expire. The default value for this is 2, setting this
4518 * value to 1 will disable the delayed sack algorithm.
4520 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4521 char __user *optval,
4524 struct sctp_sack_info params;
4525 struct sctp_association *asoc = NULL;
4526 struct sctp_sock *sp = sctp_sk(sk);
4528 if (len >= sizeof(struct sctp_sack_info)) {
4529 len = sizeof(struct sctp_sack_info);
4531 if (copy_from_user(¶ms, optval, len))
4533 } else if (len == sizeof(struct sctp_assoc_value)) {
4534 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4535 pr_warn("Use struct sctp_sack_info instead\n");
4536 if (copy_from_user(¶ms, optval, len))
4541 /* Get association, if sack_assoc_id != 0 and the socket is a one
4542 * to many style socket, and an association was not found, then
4543 * the id was invalid.
4545 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4546 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4550 /* Fetch association values. */
4551 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4552 params.sack_delay = jiffies_to_msecs(
4554 params.sack_freq = asoc->sackfreq;
4557 params.sack_delay = 0;
4558 params.sack_freq = 1;
4561 /* Fetch socket values. */
4562 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4563 params.sack_delay = sp->sackdelay;
4564 params.sack_freq = sp->sackfreq;
4566 params.sack_delay = 0;
4567 params.sack_freq = 1;
4571 if (copy_to_user(optval, ¶ms, len))
4574 if (put_user(len, optlen))
4580 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4582 * Applications can specify protocol parameters for the default association
4583 * initialization. The option name argument to setsockopt() and getsockopt()
4586 * Setting initialization parameters is effective only on an unconnected
4587 * socket (for UDP-style sockets only future associations are effected
4588 * by the change). With TCP-style sockets, this option is inherited by
4589 * sockets derived from a listener socket.
4591 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4593 if (len < sizeof(struct sctp_initmsg))
4595 len = sizeof(struct sctp_initmsg);
4596 if (put_user(len, optlen))
4598 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4604 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4605 char __user *optval, int __user *optlen)
4607 struct sctp_association *asoc;
4609 struct sctp_getaddrs getaddrs;
4610 struct sctp_transport *from;
4612 union sctp_addr temp;
4613 struct sctp_sock *sp = sctp_sk(sk);
4618 if (len < sizeof(struct sctp_getaddrs))
4621 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4624 /* For UDP-style sockets, id specifies the association to query. */
4625 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4629 to = optval + offsetof(struct sctp_getaddrs,addrs);
4630 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4632 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4634 memcpy(&temp, &from->ipaddr, sizeof(temp));
4635 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4636 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4637 if (space_left < addrlen)
4639 if (copy_to_user(to, &temp, addrlen))
4643 space_left -= addrlen;
4646 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4648 bytes_copied = ((char __user *)to) - optval;
4649 if (put_user(bytes_copied, optlen))
4655 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4656 size_t space_left, int *bytes_copied)
4658 struct sctp_sockaddr_entry *addr;
4659 union sctp_addr temp;
4664 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4668 if ((PF_INET == sk->sk_family) &&
4669 (AF_INET6 == addr->a.sa.sa_family))
4671 if ((PF_INET6 == sk->sk_family) &&
4672 inet_v6_ipv6only(sk) &&
4673 (AF_INET == addr->a.sa.sa_family))
4675 memcpy(&temp, &addr->a, sizeof(temp));
4676 if (!temp.v4.sin_port)
4677 temp.v4.sin_port = htons(port);
4679 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4681 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4682 if (space_left < addrlen) {
4686 memcpy(to, &temp, addrlen);
4690 space_left -= addrlen;
4691 *bytes_copied += addrlen;
4699 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4700 char __user *optval, int __user *optlen)
4702 struct sctp_bind_addr *bp;
4703 struct sctp_association *asoc;
4705 struct sctp_getaddrs getaddrs;
4706 struct sctp_sockaddr_entry *addr;
4708 union sctp_addr temp;
4709 struct sctp_sock *sp = sctp_sk(sk);
4713 int bytes_copied = 0;
4717 if (len < sizeof(struct sctp_getaddrs))
4720 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4724 * For UDP-style sockets, id specifies the association to query.
4725 * If the id field is set to the value '0' then the locally bound
4726 * addresses are returned without regard to any particular
4729 if (0 == getaddrs.assoc_id) {
4730 bp = &sctp_sk(sk)->ep->base.bind_addr;
4732 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4735 bp = &asoc->base.bind_addr;
4738 to = optval + offsetof(struct sctp_getaddrs,addrs);
4739 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4741 addrs = kmalloc(space_left, GFP_KERNEL);
4745 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4746 * addresses from the global local address list.
4748 if (sctp_list_single_entry(&bp->address_list)) {
4749 addr = list_entry(bp->address_list.next,
4750 struct sctp_sockaddr_entry, list);
4751 if (sctp_is_any(sk, &addr->a)) {
4752 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4753 space_left, &bytes_copied);
4763 /* Protection on the bound address list is not needed since
4764 * in the socket option context we hold a socket lock and
4765 * thus the bound address list can't change.
4767 list_for_each_entry(addr, &bp->address_list, list) {
4768 memcpy(&temp, &addr->a, sizeof(temp));
4769 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4770 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4771 if (space_left < addrlen) {
4772 err = -ENOMEM; /*fixme: right error?*/
4775 memcpy(buf, &temp, addrlen);
4777 bytes_copied += addrlen;
4779 space_left -= addrlen;
4783 if (copy_to_user(to, addrs, bytes_copied)) {
4787 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4791 if (put_user(bytes_copied, optlen))
4798 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4800 * Requests that the local SCTP stack use the enclosed peer address as
4801 * the association primary. The enclosed address must be one of the
4802 * association peer's addresses.
4804 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4805 char __user *optval, int __user *optlen)
4807 struct sctp_prim prim;
4808 struct sctp_association *asoc;
4809 struct sctp_sock *sp = sctp_sk(sk);
4811 if (len < sizeof(struct sctp_prim))
4814 len = sizeof(struct sctp_prim);
4816 if (copy_from_user(&prim, optval, len))
4819 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4823 if (!asoc->peer.primary_path)
4826 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4827 asoc->peer.primary_path->af_specific->sockaddr_len);
4829 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4830 (union sctp_addr *)&prim.ssp_addr);
4832 if (put_user(len, optlen))
4834 if (copy_to_user(optval, &prim, len))
4841 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4843 * Requests that the local endpoint set the specified Adaptation Layer
4844 * Indication parameter for all future INIT and INIT-ACK exchanges.
4846 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4847 char __user *optval, int __user *optlen)
4849 struct sctp_setadaptation adaptation;
4851 if (len < sizeof(struct sctp_setadaptation))
4854 len = sizeof(struct sctp_setadaptation);
4856 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4858 if (put_user(len, optlen))
4860 if (copy_to_user(optval, &adaptation, len))
4868 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4870 * Applications that wish to use the sendto() system call may wish to
4871 * specify a default set of parameters that would normally be supplied
4872 * through the inclusion of ancillary data. This socket option allows
4873 * such an application to set the default sctp_sndrcvinfo structure.
4876 * The application that wishes to use this socket option simply passes
4877 * in to this call the sctp_sndrcvinfo structure defined in Section
4878 * 5.2.2) The input parameters accepted by this call include
4879 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4880 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4881 * to this call if the caller is using the UDP model.
4883 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4885 static int sctp_getsockopt_default_send_param(struct sock *sk,
4886 int len, char __user *optval,
4889 struct sctp_sndrcvinfo info;
4890 struct sctp_association *asoc;
4891 struct sctp_sock *sp = sctp_sk(sk);
4893 if (len < sizeof(struct sctp_sndrcvinfo))
4896 len = sizeof(struct sctp_sndrcvinfo);
4898 if (copy_from_user(&info, optval, len))
4901 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4902 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4906 info.sinfo_stream = asoc->default_stream;
4907 info.sinfo_flags = asoc->default_flags;
4908 info.sinfo_ppid = asoc->default_ppid;
4909 info.sinfo_context = asoc->default_context;
4910 info.sinfo_timetolive = asoc->default_timetolive;
4912 info.sinfo_stream = sp->default_stream;
4913 info.sinfo_flags = sp->default_flags;
4914 info.sinfo_ppid = sp->default_ppid;
4915 info.sinfo_context = sp->default_context;
4916 info.sinfo_timetolive = sp->default_timetolive;
4919 if (put_user(len, optlen))
4921 if (copy_to_user(optval, &info, len))
4929 * 7.1.5 SCTP_NODELAY
4931 * Turn on/off any Nagle-like algorithm. This means that packets are
4932 * generally sent as soon as possible and no unnecessary delays are
4933 * introduced, at the cost of more packets in the network. Expects an
4934 * integer boolean flag.
4937 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4938 char __user *optval, int __user *optlen)
4942 if (len < sizeof(int))
4946 val = (sctp_sk(sk)->nodelay == 1);
4947 if (put_user(len, optlen))
4949 if (copy_to_user(optval, &val, len))
4956 * 7.1.1 SCTP_RTOINFO
4958 * The protocol parameters used to initialize and bound retransmission
4959 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4960 * and modify these parameters.
4961 * All parameters are time values, in milliseconds. A value of 0, when
4962 * modifying the parameters, indicates that the current value should not
4966 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4967 char __user *optval,
4968 int __user *optlen) {
4969 struct sctp_rtoinfo rtoinfo;
4970 struct sctp_association *asoc;
4972 if (len < sizeof (struct sctp_rtoinfo))
4975 len = sizeof(struct sctp_rtoinfo);
4977 if (copy_from_user(&rtoinfo, optval, len))
4980 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4982 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4985 /* Values corresponding to the specific association. */
4987 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4988 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4989 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4991 /* Values corresponding to the endpoint. */
4992 struct sctp_sock *sp = sctp_sk(sk);
4994 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4995 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4996 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4999 if (put_user(len, optlen))
5002 if (copy_to_user(optval, &rtoinfo, len))
5010 * 7.1.2 SCTP_ASSOCINFO
5012 * This option is used to tune the maximum retransmission attempts
5013 * of the association.
5014 * Returns an error if the new association retransmission value is
5015 * greater than the sum of the retransmission value of the peer.
5016 * See [SCTP] for more information.
5019 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5020 char __user *optval,
5024 struct sctp_assocparams assocparams;
5025 struct sctp_association *asoc;
5026 struct list_head *pos;
5029 if (len < sizeof (struct sctp_assocparams))
5032 len = sizeof(struct sctp_assocparams);
5034 if (copy_from_user(&assocparams, optval, len))
5037 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5039 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5042 /* Values correspoinding to the specific association */
5044 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5045 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5046 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5047 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
5049 (asoc->cookie_life.tv_usec
5052 list_for_each(pos, &asoc->peer.transport_addr_list) {
5056 assocparams.sasoc_number_peer_destinations = cnt;
5058 /* Values corresponding to the endpoint */
5059 struct sctp_sock *sp = sctp_sk(sk);
5061 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5062 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5063 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5064 assocparams.sasoc_cookie_life =
5065 sp->assocparams.sasoc_cookie_life;
5066 assocparams.sasoc_number_peer_destinations =
5068 sasoc_number_peer_destinations;
5071 if (put_user(len, optlen))
5074 if (copy_to_user(optval, &assocparams, len))
5081 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5083 * This socket option is a boolean flag which turns on or off mapped V4
5084 * addresses. If this option is turned on and the socket is type
5085 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5086 * If this option is turned off, then no mapping will be done of V4
5087 * addresses and a user will receive both PF_INET6 and PF_INET type
5088 * addresses on the socket.
5090 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5091 char __user *optval, int __user *optlen)
5094 struct sctp_sock *sp = sctp_sk(sk);
5096 if (len < sizeof(int))
5101 if (put_user(len, optlen))
5103 if (copy_to_user(optval, &val, len))
5110 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5111 * (chapter and verse is quoted at sctp_setsockopt_context())
5113 static int sctp_getsockopt_context(struct sock *sk, int len,
5114 char __user *optval, int __user *optlen)
5116 struct sctp_assoc_value params;
5117 struct sctp_sock *sp;
5118 struct sctp_association *asoc;
5120 if (len < sizeof(struct sctp_assoc_value))
5123 len = sizeof(struct sctp_assoc_value);
5125 if (copy_from_user(¶ms, optval, len))
5130 if (params.assoc_id != 0) {
5131 asoc = sctp_id2assoc(sk, params.assoc_id);
5134 params.assoc_value = asoc->default_rcv_context;
5136 params.assoc_value = sp->default_rcv_context;
5139 if (put_user(len, optlen))
5141 if (copy_to_user(optval, ¶ms, len))
5148 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5149 * This option will get or set the maximum size to put in any outgoing
5150 * SCTP DATA chunk. If a message is larger than this size it will be
5151 * fragmented by SCTP into the specified size. Note that the underlying
5152 * SCTP implementation may fragment into smaller sized chunks when the
5153 * PMTU of the underlying association is smaller than the value set by
5154 * the user. The default value for this option is '0' which indicates
5155 * the user is NOT limiting fragmentation and only the PMTU will effect
5156 * SCTP's choice of DATA chunk size. Note also that values set larger
5157 * than the maximum size of an IP datagram will effectively let SCTP
5158 * control fragmentation (i.e. the same as setting this option to 0).
5160 * The following structure is used to access and modify this parameter:
5162 * struct sctp_assoc_value {
5163 * sctp_assoc_t assoc_id;
5164 * uint32_t assoc_value;
5167 * assoc_id: This parameter is ignored for one-to-one style sockets.
5168 * For one-to-many style sockets this parameter indicates which
5169 * association the user is performing an action upon. Note that if
5170 * this field's value is zero then the endpoints default value is
5171 * changed (effecting future associations only).
5172 * assoc_value: This parameter specifies the maximum size in bytes.
5174 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5175 char __user *optval, int __user *optlen)
5177 struct sctp_assoc_value params;
5178 struct sctp_association *asoc;
5180 if (len == sizeof(int)) {
5181 pr_warn("Use of int in maxseg socket option deprecated\n");
5182 pr_warn("Use struct sctp_assoc_value instead\n");
5183 params.assoc_id = 0;
5184 } else if (len >= sizeof(struct sctp_assoc_value)) {
5185 len = sizeof(struct sctp_assoc_value);
5186 if (copy_from_user(¶ms, optval, sizeof(params)))
5191 asoc = sctp_id2assoc(sk, params.assoc_id);
5192 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5196 params.assoc_value = asoc->frag_point;
5198 params.assoc_value = sctp_sk(sk)->user_frag;
5200 if (put_user(len, optlen))
5202 if (len == sizeof(int)) {
5203 if (copy_to_user(optval, ¶ms.assoc_value, len))
5206 if (copy_to_user(optval, ¶ms, len))
5214 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5215 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5217 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5218 char __user *optval, int __user *optlen)
5222 if (len < sizeof(int))
5227 val = sctp_sk(sk)->frag_interleave;
5228 if (put_user(len, optlen))
5230 if (copy_to_user(optval, &val, len))
5237 * 7.1.25. Set or Get the sctp partial delivery point
5238 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5240 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5241 char __user *optval,
5246 if (len < sizeof(u32))
5251 val = sctp_sk(sk)->pd_point;
5252 if (put_user(len, optlen))
5254 if (copy_to_user(optval, &val, len))
5261 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5262 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5264 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5265 char __user *optval,
5268 struct sctp_assoc_value params;
5269 struct sctp_sock *sp;
5270 struct sctp_association *asoc;
5272 if (len == sizeof(int)) {
5273 pr_warn("Use of int in max_burst socket option deprecated\n");
5274 pr_warn("Use struct sctp_assoc_value instead\n");
5275 params.assoc_id = 0;
5276 } else if (len >= sizeof(struct sctp_assoc_value)) {
5277 len = sizeof(struct sctp_assoc_value);
5278 if (copy_from_user(¶ms, optval, len))
5285 if (params.assoc_id != 0) {
5286 asoc = sctp_id2assoc(sk, params.assoc_id);
5289 params.assoc_value = asoc->max_burst;
5291 params.assoc_value = sp->max_burst;
5293 if (len == sizeof(int)) {
5294 if (copy_to_user(optval, ¶ms.assoc_value, len))
5297 if (copy_to_user(optval, ¶ms, len))
5305 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5306 char __user *optval, int __user *optlen)
5308 struct sctp_hmacalgo __user *p = (void __user *)optval;
5309 struct sctp_hmac_algo_param *hmacs;
5313 if (!sctp_auth_enable)
5316 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5317 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5319 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5322 len = sizeof(struct sctp_hmacalgo) + data_len;
5323 num_idents = data_len / sizeof(u16);
5325 if (put_user(len, optlen))
5327 if (put_user(num_idents, &p->shmac_num_idents))
5329 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5334 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5335 char __user *optval, int __user *optlen)
5337 struct sctp_authkeyid val;
5338 struct sctp_association *asoc;
5340 if (!sctp_auth_enable)
5343 if (len < sizeof(struct sctp_authkeyid))
5345 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5348 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5349 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5353 val.scact_keynumber = asoc->active_key_id;
5355 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5357 len = sizeof(struct sctp_authkeyid);
5358 if (put_user(len, optlen))
5360 if (copy_to_user(optval, &val, len))
5366 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5367 char __user *optval, int __user *optlen)
5369 struct sctp_authchunks __user *p = (void __user *)optval;
5370 struct sctp_authchunks val;
5371 struct sctp_association *asoc;
5372 struct sctp_chunks_param *ch;
5376 if (!sctp_auth_enable)
5379 if (len < sizeof(struct sctp_authchunks))
5382 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5385 to = p->gauth_chunks;
5386 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5390 ch = asoc->peer.peer_chunks;
5394 /* See if the user provided enough room for all the data */
5395 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5396 if (len < num_chunks)
5399 if (copy_to_user(to, ch->chunks, num_chunks))
5402 len = sizeof(struct sctp_authchunks) + num_chunks;
5403 if (put_user(len, optlen)) return -EFAULT;
5404 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5409 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5410 char __user *optval, int __user *optlen)
5412 struct sctp_authchunks __user *p = (void __user *)optval;
5413 struct sctp_authchunks val;
5414 struct sctp_association *asoc;
5415 struct sctp_chunks_param *ch;
5419 if (!sctp_auth_enable)
5422 if (len < sizeof(struct sctp_authchunks))
5425 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5428 to = p->gauth_chunks;
5429 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5430 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5434 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5436 ch = sctp_sk(sk)->ep->auth_chunk_list;
5441 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5442 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5445 if (copy_to_user(to, ch->chunks, num_chunks))
5448 len = sizeof(struct sctp_authchunks) + num_chunks;
5449 if (put_user(len, optlen))
5451 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5458 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5459 * This option gets the current number of associations that are attached
5460 * to a one-to-many style socket. The option value is an uint32_t.
5462 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5463 char __user *optval, int __user *optlen)
5465 struct sctp_sock *sp = sctp_sk(sk);
5466 struct sctp_association *asoc;
5469 if (sctp_style(sk, TCP))
5472 if (len < sizeof(u32))
5477 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5481 if (put_user(len, optlen))
5483 if (copy_to_user(optval, &val, len))
5490 * 8.1.23 SCTP_AUTO_ASCONF
5491 * See the corresponding setsockopt entry as description
5493 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5494 char __user *optval, int __user *optlen)
5498 if (len < sizeof(int))
5502 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5504 if (put_user(len, optlen))
5506 if (copy_to_user(optval, &val, len))
5512 * 8.2.6. Get the Current Identifiers of Associations
5513 * (SCTP_GET_ASSOC_ID_LIST)
5515 * This option gets the current list of SCTP association identifiers of
5516 * the SCTP associations handled by a one-to-many style socket.
5518 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5519 char __user *optval, int __user *optlen)
5521 struct sctp_sock *sp = sctp_sk(sk);
5522 struct sctp_association *asoc;
5523 struct sctp_assoc_ids *ids;
5526 if (sctp_style(sk, TCP))
5529 if (len < sizeof(struct sctp_assoc_ids))
5532 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5536 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5539 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5541 ids = kmalloc(len, GFP_KERNEL);
5545 ids->gaids_number_of_ids = num;
5547 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5548 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5551 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5560 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5561 char __user *optval, int __user *optlen)
5566 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5569 /* I can hardly begin to describe how wrong this is. This is
5570 * so broken as to be worse than useless. The API draft
5571 * REALLY is NOT helpful here... I am not convinced that the
5572 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5573 * are at all well-founded.
5575 if (level != SOL_SCTP) {
5576 struct sctp_af *af = sctp_sk(sk)->pf->af;
5578 retval = af->getsockopt(sk, level, optname, optval, optlen);
5582 if (get_user(len, optlen))
5589 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5591 case SCTP_DISABLE_FRAGMENTS:
5592 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5596 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5598 case SCTP_AUTOCLOSE:
5599 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5601 case SCTP_SOCKOPT_PEELOFF:
5602 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5604 case SCTP_PEER_ADDR_PARAMS:
5605 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5608 case SCTP_DELAYED_SACK:
5609 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5613 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5615 case SCTP_GET_PEER_ADDRS:
5616 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5619 case SCTP_GET_LOCAL_ADDRS:
5620 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5623 case SCTP_SOCKOPT_CONNECTX3:
5624 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5626 case SCTP_DEFAULT_SEND_PARAM:
5627 retval = sctp_getsockopt_default_send_param(sk, len,
5630 case SCTP_PRIMARY_ADDR:
5631 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5634 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5637 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5639 case SCTP_ASSOCINFO:
5640 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5642 case SCTP_I_WANT_MAPPED_V4_ADDR:
5643 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5646 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5648 case SCTP_GET_PEER_ADDR_INFO:
5649 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5652 case SCTP_ADAPTATION_LAYER:
5653 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5657 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5659 case SCTP_FRAGMENT_INTERLEAVE:
5660 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5663 case SCTP_PARTIAL_DELIVERY_POINT:
5664 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5667 case SCTP_MAX_BURST:
5668 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5671 case SCTP_AUTH_CHUNK:
5672 case SCTP_AUTH_DELETE_KEY:
5673 retval = -EOPNOTSUPP;
5675 case SCTP_HMAC_IDENT:
5676 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5678 case SCTP_AUTH_ACTIVE_KEY:
5679 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5681 case SCTP_PEER_AUTH_CHUNKS:
5682 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5685 case SCTP_LOCAL_AUTH_CHUNKS:
5686 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5689 case SCTP_GET_ASSOC_NUMBER:
5690 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5692 case SCTP_GET_ASSOC_ID_LIST:
5693 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5695 case SCTP_AUTO_ASCONF:
5696 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5699 retval = -ENOPROTOOPT;
5703 sctp_release_sock(sk);
5707 static void sctp_hash(struct sock *sk)
5712 static void sctp_unhash(struct sock *sk)
5717 /* Check if port is acceptable. Possibly find first available port.
5719 * The port hash table (contained in the 'global' SCTP protocol storage
5720 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5721 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5722 * list (the list number is the port number hashed out, so as you
5723 * would expect from a hash function, all the ports in a given list have
5724 * such a number that hashes out to the same list number; you were
5725 * expecting that, right?); so each list has a set of ports, with a
5726 * link to the socket (struct sock) that uses it, the port number and
5727 * a fastreuse flag (FIXME: NPI ipg).
5729 static struct sctp_bind_bucket *sctp_bucket_create(
5730 struct sctp_bind_hashbucket *head, unsigned short snum);
5732 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5734 struct sctp_bind_hashbucket *head; /* hash list */
5735 struct sctp_bind_bucket *pp; /* hash list port iterator */
5736 struct hlist_node *node;
5737 unsigned short snum;
5740 snum = ntohs(addr->v4.sin_port);
5742 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5743 sctp_local_bh_disable();
5746 /* Search for an available port. */
5747 int low, high, remaining, index;
5750 inet_get_local_port_range(&low, &high);
5751 remaining = (high - low) + 1;
5752 rover = net_random() % remaining + low;
5756 if ((rover < low) || (rover > high))
5758 if (inet_is_reserved_local_port(rover))
5760 index = sctp_phashfn(rover);
5761 head = &sctp_port_hashtable[index];
5762 sctp_spin_lock(&head->lock);
5763 sctp_for_each_hentry(pp, node, &head->chain)
5764 if (pp->port == rover)
5768 sctp_spin_unlock(&head->lock);
5769 } while (--remaining > 0);
5771 /* Exhausted local port range during search? */
5776 /* OK, here is the one we will use. HEAD (the port
5777 * hash table list entry) is non-NULL and we hold it's
5782 /* We are given an specific port number; we verify
5783 * that it is not being used. If it is used, we will
5784 * exahust the search in the hash list corresponding
5785 * to the port number (snum) - we detect that with the
5786 * port iterator, pp being NULL.
5788 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5789 sctp_spin_lock(&head->lock);
5790 sctp_for_each_hentry(pp, node, &head->chain) {
5791 if (pp->port == snum)
5798 if (!hlist_empty(&pp->owner)) {
5799 /* We had a port hash table hit - there is an
5800 * available port (pp != NULL) and it is being
5801 * used by other socket (pp->owner not empty); that other
5802 * socket is going to be sk2.
5804 int reuse = sk->sk_reuse;
5807 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5808 if (pp->fastreuse && sk->sk_reuse &&
5809 sk->sk_state != SCTP_SS_LISTENING)
5812 /* Run through the list of sockets bound to the port
5813 * (pp->port) [via the pointers bind_next and
5814 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5815 * we get the endpoint they describe and run through
5816 * the endpoint's list of IP (v4 or v6) addresses,
5817 * comparing each of the addresses with the address of
5818 * the socket sk. If we find a match, then that means
5819 * that this port/socket (sk) combination are already
5822 sk_for_each_bound(sk2, node, &pp->owner) {
5823 struct sctp_endpoint *ep2;
5824 ep2 = sctp_sk(sk2)->ep;
5827 (reuse && sk2->sk_reuse &&
5828 sk2->sk_state != SCTP_SS_LISTENING))
5831 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5832 sctp_sk(sk2), sctp_sk(sk))) {
5837 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5840 /* If there was a hash table miss, create a new port. */
5842 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5845 /* In either case (hit or miss), make sure fastreuse is 1 only
5846 * if sk->sk_reuse is too (that is, if the caller requested
5847 * SO_REUSEADDR on this socket -sk-).
5849 if (hlist_empty(&pp->owner)) {
5850 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5854 } else if (pp->fastreuse &&
5855 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5858 /* We are set, so fill up all the data in the hash table
5859 * entry, tie the socket list information with the rest of the
5860 * sockets FIXME: Blurry, NPI (ipg).
5863 if (!sctp_sk(sk)->bind_hash) {
5864 inet_sk(sk)->inet_num = snum;
5865 sk_add_bind_node(sk, &pp->owner);
5866 sctp_sk(sk)->bind_hash = pp;
5871 sctp_spin_unlock(&head->lock);
5874 sctp_local_bh_enable();
5878 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5879 * port is requested.
5881 static int sctp_get_port(struct sock *sk, unsigned short snum)
5884 union sctp_addr addr;
5885 struct sctp_af *af = sctp_sk(sk)->pf->af;
5887 /* Set up a dummy address struct from the sk. */
5888 af->from_sk(&addr, sk);
5889 addr.v4.sin_port = htons(snum);
5891 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5892 ret = sctp_get_port_local(sk, &addr);
5898 * Move a socket to LISTENING state.
5900 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5902 struct sctp_sock *sp = sctp_sk(sk);
5903 struct sctp_endpoint *ep = sp->ep;
5904 struct crypto_hash *tfm = NULL;
5906 /* Allocate HMAC for generating cookie. */
5907 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5908 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5910 if (net_ratelimit()) {
5911 pr_info("failed to load transform for %s: %ld\n",
5912 sctp_hmac_alg, PTR_ERR(tfm));
5916 sctp_sk(sk)->hmac = tfm;
5920 * If a bind() or sctp_bindx() is not called prior to a listen()
5921 * call that allows new associations to be accepted, the system
5922 * picks an ephemeral port and will choose an address set equivalent
5923 * to binding with a wildcard address.
5925 * This is not currently spelled out in the SCTP sockets
5926 * extensions draft, but follows the practice as seen in TCP
5930 sk->sk_state = SCTP_SS_LISTENING;
5931 if (!ep->base.bind_addr.port) {
5932 if (sctp_autobind(sk))
5935 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
5936 sk->sk_state = SCTP_SS_CLOSED;
5941 sk->sk_max_ack_backlog = backlog;
5942 sctp_hash_endpoint(ep);
5947 * 4.1.3 / 5.1.3 listen()
5949 * By default, new associations are not accepted for UDP style sockets.
5950 * An application uses listen() to mark a socket as being able to
5951 * accept new associations.
5953 * On TCP style sockets, applications use listen() to ready the SCTP
5954 * endpoint for accepting inbound associations.
5956 * On both types of endpoints a backlog of '0' disables listening.
5958 * Move a socket to LISTENING state.
5960 int sctp_inet_listen(struct socket *sock, int backlog)
5962 struct sock *sk = sock->sk;
5963 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5966 if (unlikely(backlog < 0))
5971 /* Peeled-off sockets are not allowed to listen(). */
5972 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
5975 if (sock->state != SS_UNCONNECTED)
5978 /* If backlog is zero, disable listening. */
5980 if (sctp_sstate(sk, CLOSED))
5984 sctp_unhash_endpoint(ep);
5985 sk->sk_state = SCTP_SS_CLOSED;
5987 sctp_sk(sk)->bind_hash->fastreuse = 1;
5991 /* If we are already listening, just update the backlog */
5992 if (sctp_sstate(sk, LISTENING))
5993 sk->sk_max_ack_backlog = backlog;
5995 err = sctp_listen_start(sk, backlog);
6002 sctp_release_sock(sk);
6007 * This function is done by modeling the current datagram_poll() and the
6008 * tcp_poll(). Note that, based on these implementations, we don't
6009 * lock the socket in this function, even though it seems that,
6010 * ideally, locking or some other mechanisms can be used to ensure
6011 * the integrity of the counters (sndbuf and wmem_alloc) used
6012 * in this place. We assume that we don't need locks either until proven
6015 * Another thing to note is that we include the Async I/O support
6016 * here, again, by modeling the current TCP/UDP code. We don't have
6017 * a good way to test with it yet.
6019 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6021 struct sock *sk = sock->sk;
6022 struct sctp_sock *sp = sctp_sk(sk);
6025 poll_wait(file, sk_sleep(sk), wait);
6027 /* A TCP-style listening socket becomes readable when the accept queue
6030 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6031 return (!list_empty(&sp->ep->asocs)) ?
6032 (POLLIN | POLLRDNORM) : 0;
6036 /* Is there any exceptional events? */
6037 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6039 if (sk->sk_shutdown & RCV_SHUTDOWN)
6040 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6041 if (sk->sk_shutdown == SHUTDOWN_MASK)
6044 /* Is it readable? Reconsider this code with TCP-style support. */
6045 if (!skb_queue_empty(&sk->sk_receive_queue))
6046 mask |= POLLIN | POLLRDNORM;
6048 /* The association is either gone or not ready. */
6049 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6052 /* Is it writable? */
6053 if (sctp_writeable(sk)) {
6054 mask |= POLLOUT | POLLWRNORM;
6056 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6058 * Since the socket is not locked, the buffer
6059 * might be made available after the writeable check and
6060 * before the bit is set. This could cause a lost I/O
6061 * signal. tcp_poll() has a race breaker for this race
6062 * condition. Based on their implementation, we put
6063 * in the following code to cover it as well.
6065 if (sctp_writeable(sk))
6066 mask |= POLLOUT | POLLWRNORM;
6071 /********************************************************************
6072 * 2nd Level Abstractions
6073 ********************************************************************/
6075 static struct sctp_bind_bucket *sctp_bucket_create(
6076 struct sctp_bind_hashbucket *head, unsigned short snum)
6078 struct sctp_bind_bucket *pp;
6080 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6082 SCTP_DBG_OBJCNT_INC(bind_bucket);
6085 INIT_HLIST_HEAD(&pp->owner);
6086 hlist_add_head(&pp->node, &head->chain);
6091 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6092 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6094 if (pp && hlist_empty(&pp->owner)) {
6095 __hlist_del(&pp->node);
6096 kmem_cache_free(sctp_bucket_cachep, pp);
6097 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6101 /* Release this socket's reference to a local port. */
6102 static inline void __sctp_put_port(struct sock *sk)
6104 struct sctp_bind_hashbucket *head =
6105 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->inet_num)];
6106 struct sctp_bind_bucket *pp;
6108 sctp_spin_lock(&head->lock);
6109 pp = sctp_sk(sk)->bind_hash;
6110 __sk_del_bind_node(sk);
6111 sctp_sk(sk)->bind_hash = NULL;
6112 inet_sk(sk)->inet_num = 0;
6113 sctp_bucket_destroy(pp);
6114 sctp_spin_unlock(&head->lock);
6117 void sctp_put_port(struct sock *sk)
6119 sctp_local_bh_disable();
6120 __sctp_put_port(sk);
6121 sctp_local_bh_enable();
6125 * The system picks an ephemeral port and choose an address set equivalent
6126 * to binding with a wildcard address.
6127 * One of those addresses will be the primary address for the association.
6128 * This automatically enables the multihoming capability of SCTP.
6130 static int sctp_autobind(struct sock *sk)
6132 union sctp_addr autoaddr;
6136 /* Initialize a local sockaddr structure to INADDR_ANY. */
6137 af = sctp_sk(sk)->pf->af;
6139 port = htons(inet_sk(sk)->inet_num);
6140 af->inaddr_any(&autoaddr, port);
6142 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6145 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6148 * 4.2 The cmsghdr Structure *
6150 * When ancillary data is sent or received, any number of ancillary data
6151 * objects can be specified by the msg_control and msg_controllen members of
6152 * the msghdr structure, because each object is preceded by
6153 * a cmsghdr structure defining the object's length (the cmsg_len member).
6154 * Historically Berkeley-derived implementations have passed only one object
6155 * at a time, but this API allows multiple objects to be
6156 * passed in a single call to sendmsg() or recvmsg(). The following example
6157 * shows two ancillary data objects in a control buffer.
6159 * |<--------------------------- msg_controllen -------------------------->|
6162 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6164 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6167 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6169 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6172 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6173 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6175 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6177 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6184 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
6185 sctp_cmsgs_t *cmsgs)
6187 struct cmsghdr *cmsg;
6188 struct msghdr *my_msg = (struct msghdr *)msg;
6190 for (cmsg = CMSG_FIRSTHDR(msg);
6192 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6193 if (!CMSG_OK(my_msg, cmsg))
6196 /* Should we parse this header or ignore? */
6197 if (cmsg->cmsg_level != IPPROTO_SCTP)
6200 /* Strictly check lengths following example in SCM code. */
6201 switch (cmsg->cmsg_type) {
6203 /* SCTP Socket API Extension
6204 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6206 * This cmsghdr structure provides information for
6207 * initializing new SCTP associations with sendmsg().
6208 * The SCTP_INITMSG socket option uses this same data
6209 * structure. This structure is not used for
6212 * cmsg_level cmsg_type cmsg_data[]
6213 * ------------ ------------ ----------------------
6214 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6216 if (cmsg->cmsg_len !=
6217 CMSG_LEN(sizeof(struct sctp_initmsg)))
6219 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6223 /* SCTP Socket API Extension
6224 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6226 * This cmsghdr structure specifies SCTP options for
6227 * sendmsg() and describes SCTP header information
6228 * about a received message through recvmsg().
6230 * cmsg_level cmsg_type cmsg_data[]
6231 * ------------ ------------ ----------------------
6232 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6234 if (cmsg->cmsg_len !=
6235 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6239 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6241 /* Minimally, validate the sinfo_flags. */
6242 if (cmsgs->info->sinfo_flags &
6243 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6244 SCTP_ABORT | SCTP_EOF))
6256 * Wait for a packet..
6257 * Note: This function is the same function as in core/datagram.c
6258 * with a few modifications to make lksctp work.
6260 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6265 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6267 /* Socket errors? */
6268 error = sock_error(sk);
6272 if (!skb_queue_empty(&sk->sk_receive_queue))
6275 /* Socket shut down? */
6276 if (sk->sk_shutdown & RCV_SHUTDOWN)
6279 /* Sequenced packets can come disconnected. If so we report the
6284 /* Is there a good reason to think that we may receive some data? */
6285 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6288 /* Handle signals. */
6289 if (signal_pending(current))
6292 /* Let another process have a go. Since we are going to sleep
6293 * anyway. Note: This may cause odd behaviors if the message
6294 * does not fit in the user's buffer, but this seems to be the
6295 * only way to honor MSG_DONTWAIT realistically.
6297 sctp_release_sock(sk);
6298 *timeo_p = schedule_timeout(*timeo_p);
6302 finish_wait(sk_sleep(sk), &wait);
6306 error = sock_intr_errno(*timeo_p);
6309 finish_wait(sk_sleep(sk), &wait);
6314 /* Receive a datagram.
6315 * Note: This is pretty much the same routine as in core/datagram.c
6316 * with a few changes to make lksctp work.
6318 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6319 int noblock, int *err)
6322 struct sk_buff *skb;
6325 timeo = sock_rcvtimeo(sk, noblock);
6327 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6328 timeo, MAX_SCHEDULE_TIMEOUT);
6331 /* Again only user level code calls this function,
6332 * so nothing interrupt level
6333 * will suddenly eat the receive_queue.
6335 * Look at current nfs client by the way...
6336 * However, this function was correct in any case. 8)
6338 if (flags & MSG_PEEK) {
6339 spin_lock_bh(&sk->sk_receive_queue.lock);
6340 skb = skb_peek(&sk->sk_receive_queue);
6342 atomic_inc(&skb->users);
6343 spin_unlock_bh(&sk->sk_receive_queue.lock);
6345 skb = skb_dequeue(&sk->sk_receive_queue);
6351 /* Caller is allowed not to check sk->sk_err before calling. */
6352 error = sock_error(sk);
6356 if (sk->sk_shutdown & RCV_SHUTDOWN)
6359 /* User doesn't want to wait. */
6363 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6372 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6373 static void __sctp_write_space(struct sctp_association *asoc)
6375 struct sock *sk = asoc->base.sk;
6376 struct socket *sock = sk->sk_socket;
6378 if ((sctp_wspace(asoc) > 0) && sock) {
6379 if (waitqueue_active(&asoc->wait))
6380 wake_up_interruptible(&asoc->wait);
6382 if (sctp_writeable(sk)) {
6383 wait_queue_head_t *wq = sk_sleep(sk);
6385 if (wq && waitqueue_active(wq))
6386 wake_up_interruptible(wq);
6388 /* Note that we try to include the Async I/O support
6389 * here by modeling from the current TCP/UDP code.
6390 * We have not tested with it yet.
6392 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6393 sock_wake_async(sock,
6394 SOCK_WAKE_SPACE, POLL_OUT);
6399 /* Do accounting for the sndbuf space.
6400 * Decrement the used sndbuf space of the corresponding association by the
6401 * data size which was just transmitted(freed).
6403 static void sctp_wfree(struct sk_buff *skb)
6405 struct sctp_association *asoc;
6406 struct sctp_chunk *chunk;
6409 /* Get the saved chunk pointer. */
6410 chunk = *((struct sctp_chunk **)(skb->cb));
6413 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6414 sizeof(struct sk_buff) +
6415 sizeof(struct sctp_chunk);
6417 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6420 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6422 sk->sk_wmem_queued -= skb->truesize;
6423 sk_mem_uncharge(sk, skb->truesize);
6426 __sctp_write_space(asoc);
6428 sctp_association_put(asoc);
6431 /* Do accounting for the receive space on the socket.
6432 * Accounting for the association is done in ulpevent.c
6433 * We set this as a destructor for the cloned data skbs so that
6434 * accounting is done at the correct time.
6436 void sctp_sock_rfree(struct sk_buff *skb)
6438 struct sock *sk = skb->sk;
6439 struct sctp_ulpevent *event = sctp_skb2event(skb);
6441 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6444 * Mimic the behavior of sock_rfree
6446 sk_mem_uncharge(sk, event->rmem_len);
6450 /* Helper function to wait for space in the sndbuf. */
6451 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6454 struct sock *sk = asoc->base.sk;
6456 long current_timeo = *timeo_p;
6459 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6460 asoc, (long)(*timeo_p), msg_len);
6462 /* Increment the association's refcnt. */
6463 sctp_association_hold(asoc);
6465 /* Wait on the association specific sndbuf space. */
6467 prepare_to_wait_exclusive(&asoc->wait, &wait,
6468 TASK_INTERRUPTIBLE);
6471 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6474 if (signal_pending(current))
6475 goto do_interrupted;
6476 if (msg_len <= sctp_wspace(asoc))
6479 /* Let another process have a go. Since we are going
6482 sctp_release_sock(sk);
6483 current_timeo = schedule_timeout(current_timeo);
6484 BUG_ON(sk != asoc->base.sk);
6487 *timeo_p = current_timeo;
6491 finish_wait(&asoc->wait, &wait);
6493 /* Release the association's refcnt. */
6494 sctp_association_put(asoc);
6503 err = sock_intr_errno(*timeo_p);
6511 void sctp_data_ready(struct sock *sk, int len)
6513 struct socket_wq *wq;
6516 wq = rcu_dereference(sk->sk_wq);
6517 if (wq_has_sleeper(wq))
6518 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6519 POLLRDNORM | POLLRDBAND);
6520 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6524 /* If socket sndbuf has changed, wake up all per association waiters. */
6525 void sctp_write_space(struct sock *sk)
6527 struct sctp_association *asoc;
6529 /* Wake up the tasks in each wait queue. */
6530 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6531 __sctp_write_space(asoc);
6535 /* Is there any sndbuf space available on the socket?
6537 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6538 * associations on the same socket. For a UDP-style socket with
6539 * multiple associations, it is possible for it to be "unwriteable"
6540 * prematurely. I assume that this is acceptable because
6541 * a premature "unwriteable" is better than an accidental "writeable" which
6542 * would cause an unwanted block under certain circumstances. For the 1-1
6543 * UDP-style sockets or TCP-style sockets, this code should work.
6546 static int sctp_writeable(struct sock *sk)
6550 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6556 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6557 * returns immediately with EINPROGRESS.
6559 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6561 struct sock *sk = asoc->base.sk;
6563 long current_timeo = *timeo_p;
6566 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6569 /* Increment the association's refcnt. */
6570 sctp_association_hold(asoc);
6573 prepare_to_wait_exclusive(&asoc->wait, &wait,
6574 TASK_INTERRUPTIBLE);
6577 if (sk->sk_shutdown & RCV_SHUTDOWN)
6579 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6582 if (signal_pending(current))
6583 goto do_interrupted;
6585 if (sctp_state(asoc, ESTABLISHED))
6588 /* Let another process have a go. Since we are going
6591 sctp_release_sock(sk);
6592 current_timeo = schedule_timeout(current_timeo);
6595 *timeo_p = current_timeo;
6599 finish_wait(&asoc->wait, &wait);
6601 /* Release the association's refcnt. */
6602 sctp_association_put(asoc);
6607 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6610 err = -ECONNREFUSED;
6614 err = sock_intr_errno(*timeo_p);
6622 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6624 struct sctp_endpoint *ep;
6628 ep = sctp_sk(sk)->ep;
6632 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6633 TASK_INTERRUPTIBLE);
6635 if (list_empty(&ep->asocs)) {
6636 sctp_release_sock(sk);
6637 timeo = schedule_timeout(timeo);
6642 if (!sctp_sstate(sk, LISTENING))
6646 if (!list_empty(&ep->asocs))
6649 err = sock_intr_errno(timeo);
6650 if (signal_pending(current))
6658 finish_wait(sk_sleep(sk), &wait);
6663 static void sctp_wait_for_close(struct sock *sk, long timeout)
6668 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6669 if (list_empty(&sctp_sk(sk)->ep->asocs))
6671 sctp_release_sock(sk);
6672 timeout = schedule_timeout(timeout);
6674 } while (!signal_pending(current) && timeout);
6676 finish_wait(sk_sleep(sk), &wait);
6679 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6681 struct sk_buff *frag;
6686 /* Don't forget the fragments. */
6687 skb_walk_frags(skb, frag)
6688 sctp_skb_set_owner_r_frag(frag, sk);
6691 sctp_skb_set_owner_r(skb, sk);
6694 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6695 struct sctp_association *asoc)
6697 struct inet_sock *inet = inet_sk(sk);
6698 struct inet_sock *newinet;
6700 newsk->sk_type = sk->sk_type;
6701 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6702 newsk->sk_flags = sk->sk_flags;
6703 newsk->sk_no_check = sk->sk_no_check;
6704 newsk->sk_reuse = sk->sk_reuse;
6706 newsk->sk_shutdown = sk->sk_shutdown;
6707 newsk->sk_destruct = inet_sock_destruct;
6708 newsk->sk_family = sk->sk_family;
6709 newsk->sk_protocol = IPPROTO_SCTP;
6710 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6711 newsk->sk_sndbuf = sk->sk_sndbuf;
6712 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6713 newsk->sk_lingertime = sk->sk_lingertime;
6714 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6715 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6717 newinet = inet_sk(newsk);
6719 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6720 * getsockname() and getpeername()
6722 newinet->inet_sport = inet->inet_sport;
6723 newinet->inet_saddr = inet->inet_saddr;
6724 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6725 newinet->inet_dport = htons(asoc->peer.port);
6726 newinet->pmtudisc = inet->pmtudisc;
6727 newinet->inet_id = asoc->next_tsn ^ jiffies;
6729 newinet->uc_ttl = inet->uc_ttl;
6730 newinet->mc_loop = 1;
6731 newinet->mc_ttl = 1;
6732 newinet->mc_index = 0;
6733 newinet->mc_list = NULL;
6736 static inline void sctp_copy_descendant(struct sock *sk_to,
6737 const struct sock *sk_from)
6739 int ancestor_size = sizeof(struct inet_sock) +
6740 sizeof(struct sctp_sock) -
6741 offsetof(struct sctp_sock, auto_asconf_list);
6743 if (sk_from->sk_family == PF_INET6)
6744 ancestor_size += sizeof(struct ipv6_pinfo);
6746 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
6749 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6750 * and its messages to the newsk.
6752 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6753 struct sctp_association *assoc,
6754 sctp_socket_type_t type)
6756 struct sctp_sock *oldsp = sctp_sk(oldsk);
6757 struct sctp_sock *newsp = sctp_sk(newsk);
6758 struct sctp_bind_bucket *pp; /* hash list port iterator */
6759 struct sctp_endpoint *newep = newsp->ep;
6760 struct sk_buff *skb, *tmp;
6761 struct sctp_ulpevent *event;
6762 struct sctp_bind_hashbucket *head;
6764 /* Migrate socket buffer sizes and all the socket level options to the
6767 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6768 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6769 /* Brute force copy old sctp opt. */
6770 sctp_copy_descendant(newsk, oldsk);
6772 /* Restore the ep value that was overwritten with the above structure
6778 /* Hook this new socket in to the bind_hash list. */
6779 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->inet_num)];
6780 sctp_local_bh_disable();
6781 sctp_spin_lock(&head->lock);
6782 pp = sctp_sk(oldsk)->bind_hash;
6783 sk_add_bind_node(newsk, &pp->owner);
6784 sctp_sk(newsk)->bind_hash = pp;
6785 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6786 sctp_spin_unlock(&head->lock);
6787 sctp_local_bh_enable();
6789 /* Copy the bind_addr list from the original endpoint to the new
6790 * endpoint so that we can handle restarts properly
6792 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6793 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6795 /* Move any messages in the old socket's receive queue that are for the
6796 * peeled off association to the new socket's receive queue.
6798 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6799 event = sctp_skb2event(skb);
6800 if (event->asoc == assoc) {
6801 __skb_unlink(skb, &oldsk->sk_receive_queue);
6802 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6803 sctp_skb_set_owner_r_frag(skb, newsk);
6807 /* Clean up any messages pending delivery due to partial
6808 * delivery. Three cases:
6809 * 1) No partial deliver; no work.
6810 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6811 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6813 skb_queue_head_init(&newsp->pd_lobby);
6814 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6816 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6817 struct sk_buff_head *queue;
6819 /* Decide which queue to move pd_lobby skbs to. */
6820 if (assoc->ulpq.pd_mode) {
6821 queue = &newsp->pd_lobby;
6823 queue = &newsk->sk_receive_queue;
6825 /* Walk through the pd_lobby, looking for skbs that
6826 * need moved to the new socket.
6828 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6829 event = sctp_skb2event(skb);
6830 if (event->asoc == assoc) {
6831 __skb_unlink(skb, &oldsp->pd_lobby);
6832 __skb_queue_tail(queue, skb);
6833 sctp_skb_set_owner_r_frag(skb, newsk);
6837 /* Clear up any skbs waiting for the partial
6838 * delivery to finish.
6840 if (assoc->ulpq.pd_mode)
6841 sctp_clear_pd(oldsk, NULL);
6845 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6846 sctp_skb_set_owner_r_frag(skb, newsk);
6848 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6849 sctp_skb_set_owner_r_frag(skb, newsk);
6851 /* Set the type of socket to indicate that it is peeled off from the
6852 * original UDP-style socket or created with the accept() call on a
6853 * TCP-style socket..
6857 /* Mark the new socket "in-use" by the user so that any packets
6858 * that may arrive on the association after we've moved it are
6859 * queued to the backlog. This prevents a potential race between
6860 * backlog processing on the old socket and new-packet processing
6861 * on the new socket.
6863 * The caller has just allocated newsk so we can guarantee that other
6864 * paths won't try to lock it and then oldsk.
6866 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6867 sctp_assoc_migrate(assoc, newsk);
6869 /* If the association on the newsk is already closed before accept()
6870 * is called, set RCV_SHUTDOWN flag.
6872 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6873 newsk->sk_shutdown |= RCV_SHUTDOWN;
6875 newsk->sk_state = SCTP_SS_ESTABLISHED;
6876 sctp_release_sock(newsk);
6880 /* This proto struct describes the ULP interface for SCTP. */
6881 struct proto sctp_prot = {
6883 .owner = THIS_MODULE,
6884 .close = sctp_close,
6885 .connect = sctp_connect,
6886 .disconnect = sctp_disconnect,
6887 .accept = sctp_accept,
6888 .ioctl = sctp_ioctl,
6889 .init = sctp_init_sock,
6890 .destroy = sctp_destroy_sock,
6891 .shutdown = sctp_shutdown,
6892 .setsockopt = sctp_setsockopt,
6893 .getsockopt = sctp_getsockopt,
6894 .sendmsg = sctp_sendmsg,
6895 .recvmsg = sctp_recvmsg,
6897 .backlog_rcv = sctp_backlog_rcv,
6899 .unhash = sctp_unhash,
6900 .get_port = sctp_get_port,
6901 .obj_size = sizeof(struct sctp_sock),
6902 .sysctl_mem = sysctl_sctp_mem,
6903 .sysctl_rmem = sysctl_sctp_rmem,
6904 .sysctl_wmem = sysctl_sctp_wmem,
6905 .memory_pressure = &sctp_memory_pressure,
6906 .enter_memory_pressure = sctp_enter_memory_pressure,
6907 .memory_allocated = &sctp_memory_allocated,
6908 .sockets_allocated = &sctp_sockets_allocated,
6911 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6913 #include <net/transp_v6.h>
6914 static void sctp_v6_destroy_sock(struct sock *sk)
6916 sctp_destroy_sock(sk);
6917 inet6_destroy_sock(sk);
6920 struct proto sctpv6_prot = {
6922 .owner = THIS_MODULE,
6923 .close = sctp_close,
6924 .connect = sctp_connect,
6925 .disconnect = sctp_disconnect,
6926 .accept = sctp_accept,
6927 .ioctl = sctp_ioctl,
6928 .init = sctp_init_sock,
6929 .destroy = sctp_v6_destroy_sock,
6930 .shutdown = sctp_shutdown,
6931 .setsockopt = sctp_setsockopt,
6932 .getsockopt = sctp_getsockopt,
6933 .sendmsg = sctp_sendmsg,
6934 .recvmsg = sctp_recvmsg,
6936 .backlog_rcv = sctp_backlog_rcv,
6938 .unhash = sctp_unhash,
6939 .get_port = sctp_get_port,
6940 .obj_size = sizeof(struct sctp6_sock),
6941 .sysctl_mem = sysctl_sctp_mem,
6942 .sysctl_rmem = sysctl_sctp_rmem,
6943 .sysctl_wmem = sysctl_sctp_wmem,
6944 .memory_pressure = &sctp_memory_pressure,
6945 .enter_memory_pressure = sctp_enter_memory_pressure,
6946 .memory_allocated = &sctp_memory_allocated,
6947 .sockets_allocated = &sctp_sockets_allocated,
6949 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */