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);
1527 sctp_primitive_ABORT(asoc, chunk);
1529 sctp_primitive_SHUTDOWN(asoc, NULL);
1532 /* On a TCP-style socket, block for at most linger_time if set. */
1533 if (sctp_style(sk, TCP) && timeout)
1534 sctp_wait_for_close(sk, timeout);
1536 /* This will run the backlog queue. */
1537 sctp_release_sock(sk);
1539 /* Supposedly, no process has access to the socket, but
1540 * the net layers still may.
1541 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1542 * held and that should be grabbed before socket lock.
1544 spin_lock_bh(&sctp_globals.addr_wq_lock);
1545 sctp_bh_lock_sock(sk);
1547 /* Hold the sock, since sk_common_release() will put sock_put()
1548 * and we have just a little more cleanup.
1551 sk_common_release(sk);
1553 sctp_bh_unlock_sock(sk);
1554 spin_unlock_bh(&sctp_globals.addr_wq_lock);
1558 SCTP_DBG_OBJCNT_DEC(sock);
1561 /* Handle EPIPE error. */
1562 static int sctp_error(struct sock *sk, int flags, int err)
1565 err = sock_error(sk) ? : -EPIPE;
1566 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1567 send_sig(SIGPIPE, current, 0);
1571 /* API 3.1.3 sendmsg() - UDP Style Syntax
1573 * An application uses sendmsg() and recvmsg() calls to transmit data to
1574 * and receive data from its peer.
1576 * ssize_t sendmsg(int socket, const struct msghdr *message,
1579 * socket - the socket descriptor of the endpoint.
1580 * message - pointer to the msghdr structure which contains a single
1581 * user message and possibly some ancillary data.
1583 * See Section 5 for complete description of the data
1586 * flags - flags sent or received with the user message, see Section
1587 * 5 for complete description of the flags.
1589 * Note: This function could use a rewrite especially when explicit
1590 * connect support comes in.
1592 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1594 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1596 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1597 struct msghdr *msg, size_t msg_len)
1599 struct sctp_sock *sp;
1600 struct sctp_endpoint *ep;
1601 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1602 struct sctp_transport *transport, *chunk_tp;
1603 struct sctp_chunk *chunk;
1605 struct sockaddr *msg_name = NULL;
1606 struct sctp_sndrcvinfo default_sinfo;
1607 struct sctp_sndrcvinfo *sinfo;
1608 struct sctp_initmsg *sinit;
1609 sctp_assoc_t associd = 0;
1610 sctp_cmsgs_t cmsgs = { NULL };
1614 __u16 sinfo_flags = 0;
1615 bool wait_connect = false;
1616 struct sctp_datamsg *datamsg;
1617 int msg_flags = msg->msg_flags;
1619 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1626 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1628 /* We cannot send a message over a TCP-style listening socket. */
1629 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1634 /* Parse out the SCTP CMSGs. */
1635 err = sctp_msghdr_parse(msg, &cmsgs);
1638 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1642 /* Fetch the destination address for this packet. This
1643 * address only selects the association--it is not necessarily
1644 * the address we will send to.
1645 * For a peeled-off socket, msg_name is ignored.
1647 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1648 int msg_namelen = msg->msg_namelen;
1650 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1655 if (msg_namelen > sizeof(to))
1656 msg_namelen = sizeof(to);
1657 memcpy(&to, msg->msg_name, msg_namelen);
1658 msg_name = msg->msg_name;
1664 /* Did the user specify SNDRCVINFO? */
1666 sinfo_flags = sinfo->sinfo_flags;
1667 associd = sinfo->sinfo_assoc_id;
1670 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1671 msg_len, sinfo_flags);
1673 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1674 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1679 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1680 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1681 * If SCTP_ABORT is set, the message length could be non zero with
1682 * the msg_iov set to the user abort reason.
1684 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1685 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1690 /* If SCTP_ADDR_OVER is set, there must be an address
1691 * specified in msg_name.
1693 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1700 SCTP_DEBUG_PRINTK("About to look up association.\n");
1704 /* If a msg_name has been specified, assume this is to be used. */
1706 /* Look for a matching association on the endpoint. */
1707 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1709 /* If we could not find a matching association on the
1710 * endpoint, make sure that it is not a TCP-style
1711 * socket that already has an association or there is
1712 * no peeled-off association on another socket.
1714 if ((sctp_style(sk, TCP) &&
1715 sctp_sstate(sk, ESTABLISHED)) ||
1716 sctp_endpoint_is_peeled_off(ep, &to)) {
1717 err = -EADDRNOTAVAIL;
1722 asoc = sctp_id2assoc(sk, associd);
1730 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1732 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1733 * socket that has an association in CLOSED state. This can
1734 * happen when an accepted socket has an association that is
1737 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1742 if (sinfo_flags & SCTP_EOF) {
1743 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1745 sctp_primitive_SHUTDOWN(asoc, NULL);
1749 if (sinfo_flags & SCTP_ABORT) {
1751 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1757 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1758 sctp_primitive_ABORT(asoc, chunk);
1764 /* Do we need to create the association? */
1766 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1768 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1773 /* Check for invalid stream against the stream counts,
1774 * either the default or the user specified stream counts.
1777 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1778 /* Check against the defaults. */
1779 if (sinfo->sinfo_stream >=
1780 sp->initmsg.sinit_num_ostreams) {
1785 /* Check against the requested. */
1786 if (sinfo->sinfo_stream >=
1787 sinit->sinit_num_ostreams) {
1795 * API 3.1.2 bind() - UDP Style Syntax
1796 * If a bind() or sctp_bindx() is not called prior to a
1797 * sendmsg() call that initiates a new association, the
1798 * system picks an ephemeral port and will choose an address
1799 * set equivalent to binding with a wildcard address.
1801 if (!ep->base.bind_addr.port) {
1802 if (sctp_autobind(sk)) {
1808 * If an unprivileged user inherits a one-to-many
1809 * style socket with open associations on a privileged
1810 * port, it MAY be permitted to accept new associations,
1811 * but it SHOULD NOT be permitted to open new
1814 if (ep->base.bind_addr.port < PROT_SOCK &&
1815 !capable(CAP_NET_BIND_SERVICE)) {
1821 scope = sctp_scope(&to);
1822 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1828 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1834 /* If the SCTP_INIT ancillary data is specified, set all
1835 * the association init values accordingly.
1838 if (sinit->sinit_num_ostreams) {
1839 asoc->c.sinit_num_ostreams =
1840 sinit->sinit_num_ostreams;
1842 if (sinit->sinit_max_instreams) {
1843 asoc->c.sinit_max_instreams =
1844 sinit->sinit_max_instreams;
1846 if (sinit->sinit_max_attempts) {
1847 asoc->max_init_attempts
1848 = sinit->sinit_max_attempts;
1850 if (sinit->sinit_max_init_timeo) {
1851 asoc->max_init_timeo =
1852 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1856 /* Prime the peer's transport structures. */
1857 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1864 /* ASSERT: we have a valid association at this point. */
1865 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1868 /* If the user didn't specify SNDRCVINFO, make up one with
1871 memset(&default_sinfo, 0, sizeof(default_sinfo));
1872 default_sinfo.sinfo_stream = asoc->default_stream;
1873 default_sinfo.sinfo_flags = asoc->default_flags;
1874 default_sinfo.sinfo_ppid = asoc->default_ppid;
1875 default_sinfo.sinfo_context = asoc->default_context;
1876 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1877 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1878 sinfo = &default_sinfo;
1881 /* API 7.1.7, the sndbuf size per association bounds the
1882 * maximum size of data that can be sent in a single send call.
1884 if (msg_len > sk->sk_sndbuf) {
1889 if (asoc->pmtu_pending)
1890 sctp_assoc_pending_pmtu(asoc);
1892 /* If fragmentation is disabled and the message length exceeds the
1893 * association fragmentation point, return EMSGSIZE. The I-D
1894 * does not specify what this error is, but this looks like
1897 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1902 /* Check for invalid stream. */
1903 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1908 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1909 if (!sctp_wspace(asoc)) {
1910 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1915 /* If an address is passed with the sendto/sendmsg call, it is used
1916 * to override the primary destination address in the TCP model, or
1917 * when SCTP_ADDR_OVER flag is set in the UDP model.
1919 if ((sctp_style(sk, TCP) && msg_name) ||
1920 (sinfo_flags & SCTP_ADDR_OVER)) {
1921 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1929 /* Auto-connect, if we aren't connected already. */
1930 if (sctp_state(asoc, CLOSED)) {
1931 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1934 wait_connect = true;
1935 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1938 /* Break the message into multiple chunks of maximum size. */
1939 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1940 if (IS_ERR(datamsg)) {
1941 err = PTR_ERR(datamsg);
1945 /* Now send the (possibly) fragmented message. */
1946 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1947 sctp_chunk_hold(chunk);
1949 /* Do accounting for the write space. */
1950 sctp_set_owner_w(chunk);
1952 chunk->transport = chunk_tp;
1955 /* Send it to the lower layers. Note: all chunks
1956 * must either fail or succeed. The lower layer
1957 * works that way today. Keep it that way or this
1960 err = sctp_primitive_SEND(asoc, datamsg);
1961 /* Did the lower layer accept the chunk? */
1963 sctp_datamsg_free(datamsg);
1965 sctp_datamsg_put(datamsg);
1967 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1974 if (unlikely(wait_connect)) {
1975 timeo = sock_sndtimeo(sk, msg_flags & MSG_DONTWAIT);
1976 sctp_wait_for_connect(asoc, &timeo);
1979 /* If we are already past ASSOCIATE, the lower
1980 * layers are responsible for association cleanup.
1986 sctp_unhash_established(asoc);
1987 sctp_association_free(asoc);
1990 sctp_release_sock(sk);
1993 return sctp_error(sk, msg_flags, err);
2000 err = sock_error(sk);
2010 /* This is an extended version of skb_pull() that removes the data from the
2011 * start of a skb even when data is spread across the list of skb's in the
2012 * frag_list. len specifies the total amount of data that needs to be removed.
2013 * when 'len' bytes could be removed from the skb, it returns 0.
2014 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2015 * could not be removed.
2017 static int sctp_skb_pull(struct sk_buff *skb, int len)
2019 struct sk_buff *list;
2020 int skb_len = skb_headlen(skb);
2023 if (len <= skb_len) {
2024 __skb_pull(skb, len);
2028 __skb_pull(skb, skb_len);
2030 skb_walk_frags(skb, list) {
2031 rlen = sctp_skb_pull(list, len);
2032 skb->len -= (len-rlen);
2033 skb->data_len -= (len-rlen);
2044 /* API 3.1.3 recvmsg() - UDP Style Syntax
2046 * ssize_t recvmsg(int socket, struct msghdr *message,
2049 * socket - the socket descriptor of the endpoint.
2050 * message - pointer to the msghdr structure which contains a single
2051 * user message and possibly some ancillary data.
2053 * See Section 5 for complete description of the data
2056 * flags - flags sent or received with the user message, see Section
2057 * 5 for complete description of the flags.
2059 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2061 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2062 struct msghdr *msg, size_t len, int noblock,
2063 int flags, int *addr_len)
2065 struct sctp_ulpevent *event = NULL;
2066 struct sctp_sock *sp = sctp_sk(sk);
2067 struct sk_buff *skb;
2072 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
2073 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
2074 "len", len, "knoblauch", noblock,
2075 "flags", flags, "addr_len", addr_len);
2079 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2084 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2088 /* Get the total length of the skb including any skb's in the
2097 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2099 event = sctp_skb2event(skb);
2104 sock_recv_ts_and_drops(msg, sk, skb);
2105 if (sctp_ulpevent_is_notification(event)) {
2106 msg->msg_flags |= MSG_NOTIFICATION;
2107 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2109 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2112 /* Check if we allow SCTP_SNDRCVINFO. */
2113 if (sp->subscribe.sctp_data_io_event)
2114 sctp_ulpevent_read_sndrcvinfo(event, msg);
2116 /* FIXME: we should be calling IP/IPv6 layers. */
2117 if (sk->sk_protinfo.af_inet.cmsg_flags)
2118 ip_cmsg_recv(msg, skb);
2123 /* If skb's length exceeds the user's buffer, update the skb and
2124 * push it back to the receive_queue so that the next call to
2125 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2127 if (skb_len > copied) {
2128 msg->msg_flags &= ~MSG_EOR;
2129 if (flags & MSG_PEEK)
2131 sctp_skb_pull(skb, copied);
2132 skb_queue_head(&sk->sk_receive_queue, skb);
2134 /* When only partial message is copied to the user, increase
2135 * rwnd by that amount. If all the data in the skb is read,
2136 * rwnd is updated when the event is freed.
2138 if (!sctp_ulpevent_is_notification(event))
2139 sctp_assoc_rwnd_increase(event->asoc, copied);
2141 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2142 (event->msg_flags & MSG_EOR))
2143 msg->msg_flags |= MSG_EOR;
2145 msg->msg_flags &= ~MSG_EOR;
2148 if (flags & MSG_PEEK) {
2149 /* Release the skb reference acquired after peeking the skb in
2150 * sctp_skb_recv_datagram().
2154 /* Free the event which includes releasing the reference to
2155 * the owner of the skb, freeing the skb and updating the
2158 sctp_ulpevent_free(event);
2161 sctp_release_sock(sk);
2165 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2167 * This option is a on/off flag. If enabled no SCTP message
2168 * fragmentation will be performed. Instead if a message being sent
2169 * exceeds the current PMTU size, the message will NOT be sent and
2170 * instead a error will be indicated to the user.
2172 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2173 char __user *optval,
2174 unsigned int optlen)
2178 if (optlen < sizeof(int))
2181 if (get_user(val, (int __user *)optval))
2184 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2189 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2190 unsigned int optlen)
2192 struct sctp_association *asoc;
2193 struct sctp_ulpevent *event;
2195 if (optlen > sizeof(struct sctp_event_subscribe))
2197 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2201 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2202 * if there is no data to be sent or retransmit, the stack will
2203 * immediately send up this notification.
2205 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2206 &sctp_sk(sk)->subscribe)) {
2207 asoc = sctp_id2assoc(sk, 0);
2209 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2210 event = sctp_ulpevent_make_sender_dry_event(asoc,
2215 sctp_ulpq_tail_event(&asoc->ulpq, event);
2222 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2224 * This socket option is applicable to the UDP-style socket only. When
2225 * set it will cause associations that are idle for more than the
2226 * specified number of seconds to automatically close. An association
2227 * being idle is defined an association that has NOT sent or received
2228 * user data. The special value of '0' indicates that no automatic
2229 * close of any associations should be performed. The option expects an
2230 * integer defining the number of seconds of idle time before an
2231 * association is closed.
2233 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2234 unsigned int optlen)
2236 struct sctp_sock *sp = sctp_sk(sk);
2238 /* Applicable to UDP-style socket only */
2239 if (sctp_style(sk, TCP))
2241 if (optlen != sizeof(int))
2243 if (copy_from_user(&sp->autoclose, optval, optlen))
2249 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2251 * Applications can enable or disable heartbeats for any peer address of
2252 * an association, modify an address's heartbeat interval, force a
2253 * heartbeat to be sent immediately, and adjust the address's maximum
2254 * number of retransmissions sent before an address is considered
2255 * unreachable. The following structure is used to access and modify an
2256 * address's parameters:
2258 * struct sctp_paddrparams {
2259 * sctp_assoc_t spp_assoc_id;
2260 * struct sockaddr_storage spp_address;
2261 * uint32_t spp_hbinterval;
2262 * uint16_t spp_pathmaxrxt;
2263 * uint32_t spp_pathmtu;
2264 * uint32_t spp_sackdelay;
2265 * uint32_t spp_flags;
2268 * spp_assoc_id - (one-to-many style socket) This is filled in the
2269 * application, and identifies the association for
2271 * spp_address - This specifies which address is of interest.
2272 * spp_hbinterval - This contains the value of the heartbeat interval,
2273 * in milliseconds. If a value of zero
2274 * is present in this field then no changes are to
2275 * be made to this parameter.
2276 * spp_pathmaxrxt - This contains the maximum number of
2277 * retransmissions before this address shall be
2278 * considered unreachable. If a value of zero
2279 * is present in this field then no changes are to
2280 * be made to this parameter.
2281 * spp_pathmtu - When Path MTU discovery is disabled the value
2282 * specified here will be the "fixed" path mtu.
2283 * Note that if the spp_address field is empty
2284 * then all associations on this address will
2285 * have this fixed path mtu set upon them.
2287 * spp_sackdelay - When delayed sack is enabled, this value specifies
2288 * the number of milliseconds that sacks will be delayed
2289 * for. This value will apply to all addresses of an
2290 * association if the spp_address field is empty. Note
2291 * also, that if delayed sack is enabled and this
2292 * value is set to 0, no change is made to the last
2293 * recorded delayed sack timer value.
2295 * spp_flags - These flags are used to control various features
2296 * on an association. The flag field may contain
2297 * zero or more of the following options.
2299 * SPP_HB_ENABLE - Enable heartbeats on the
2300 * specified address. Note that if the address
2301 * field is empty all addresses for the association
2302 * have heartbeats enabled upon them.
2304 * SPP_HB_DISABLE - Disable heartbeats on the
2305 * speicifed address. Note that if the address
2306 * field is empty all addresses for the association
2307 * will have their heartbeats disabled. Note also
2308 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2309 * mutually exclusive, only one of these two should
2310 * be specified. Enabling both fields will have
2311 * undetermined results.
2313 * SPP_HB_DEMAND - Request a user initiated heartbeat
2314 * to be made immediately.
2316 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2317 * heartbeat delayis to be set to the value of 0
2320 * SPP_PMTUD_ENABLE - This field will enable PMTU
2321 * discovery upon the specified address. Note that
2322 * if the address feild is empty then all addresses
2323 * on the association are effected.
2325 * SPP_PMTUD_DISABLE - This field will disable PMTU
2326 * discovery upon the specified address. Note that
2327 * if the address feild is empty then all addresses
2328 * on the association are effected. Not also that
2329 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2330 * exclusive. Enabling both will have undetermined
2333 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2334 * on delayed sack. The time specified in spp_sackdelay
2335 * is used to specify the sack delay for this address. Note
2336 * that if spp_address is empty then all addresses will
2337 * enable delayed sack and take on the sack delay
2338 * value specified in spp_sackdelay.
2339 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2340 * off delayed sack. If the spp_address field is blank then
2341 * delayed sack is disabled for the entire association. Note
2342 * also that this field is mutually exclusive to
2343 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2346 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2347 struct sctp_transport *trans,
2348 struct sctp_association *asoc,
2349 struct sctp_sock *sp,
2352 int sackdelay_change)
2356 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2357 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2362 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2363 * this field is ignored. Note also that a value of zero indicates
2364 * the current setting should be left unchanged.
2366 if (params->spp_flags & SPP_HB_ENABLE) {
2368 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2369 * set. This lets us use 0 value when this flag
2372 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2373 params->spp_hbinterval = 0;
2375 if (params->spp_hbinterval ||
2376 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2379 msecs_to_jiffies(params->spp_hbinterval);
2382 msecs_to_jiffies(params->spp_hbinterval);
2384 sp->hbinterval = params->spp_hbinterval;
2391 trans->param_flags =
2392 (trans->param_flags & ~SPP_HB) | hb_change;
2395 (asoc->param_flags & ~SPP_HB) | hb_change;
2398 (sp->param_flags & ~SPP_HB) | hb_change;
2402 /* When Path MTU discovery is disabled the value specified here will
2403 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2404 * include the flag SPP_PMTUD_DISABLE for this field to have any
2407 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2409 trans->pathmtu = params->spp_pathmtu;
2410 sctp_assoc_sync_pmtu(asoc);
2412 asoc->pathmtu = params->spp_pathmtu;
2413 sctp_frag_point(asoc, params->spp_pathmtu);
2415 sp->pathmtu = params->spp_pathmtu;
2421 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2422 (params->spp_flags & SPP_PMTUD_ENABLE);
2423 trans->param_flags =
2424 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2426 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2427 sctp_assoc_sync_pmtu(asoc);
2431 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2434 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2438 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2439 * value of this field is ignored. Note also that a value of zero
2440 * indicates the current setting should be left unchanged.
2442 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2445 msecs_to_jiffies(params->spp_sackdelay);
2448 msecs_to_jiffies(params->spp_sackdelay);
2450 sp->sackdelay = params->spp_sackdelay;
2454 if (sackdelay_change) {
2456 trans->param_flags =
2457 (trans->param_flags & ~SPP_SACKDELAY) |
2461 (asoc->param_flags & ~SPP_SACKDELAY) |
2465 (sp->param_flags & ~SPP_SACKDELAY) |
2470 /* Note that a value of zero indicates the current setting should be
2473 if (params->spp_pathmaxrxt) {
2475 trans->pathmaxrxt = params->spp_pathmaxrxt;
2477 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2479 sp->pathmaxrxt = params->spp_pathmaxrxt;
2486 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2487 char __user *optval,
2488 unsigned int optlen)
2490 struct sctp_paddrparams params;
2491 struct sctp_transport *trans = NULL;
2492 struct sctp_association *asoc = NULL;
2493 struct sctp_sock *sp = sctp_sk(sk);
2495 int hb_change, pmtud_change, sackdelay_change;
2497 if (optlen != sizeof(struct sctp_paddrparams))
2500 if (copy_from_user(¶ms, optval, optlen))
2503 /* Validate flags and value parameters. */
2504 hb_change = params.spp_flags & SPP_HB;
2505 pmtud_change = params.spp_flags & SPP_PMTUD;
2506 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2508 if (hb_change == SPP_HB ||
2509 pmtud_change == SPP_PMTUD ||
2510 sackdelay_change == SPP_SACKDELAY ||
2511 params.spp_sackdelay > 500 ||
2512 (params.spp_pathmtu &&
2513 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2516 /* If an address other than INADDR_ANY is specified, and
2517 * no transport is found, then the request is invalid.
2519 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
2520 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2521 params.spp_assoc_id);
2526 /* Get association, if assoc_id != 0 and the socket is a one
2527 * to many style socket, and an association was not found, then
2528 * the id was invalid.
2530 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2531 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2534 /* Heartbeat demand can only be sent on a transport or
2535 * association, but not a socket.
2537 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2540 /* Process parameters. */
2541 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2542 hb_change, pmtud_change,
2548 /* If changes are for association, also apply parameters to each
2551 if (!trans && asoc) {
2552 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2554 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2555 hb_change, pmtud_change,
2564 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2566 * This option will effect the way delayed acks are performed. This
2567 * option allows you to get or set the delayed ack time, in
2568 * milliseconds. It also allows changing the delayed ack frequency.
2569 * Changing the frequency to 1 disables the delayed sack algorithm. If
2570 * the assoc_id is 0, then this sets or gets the endpoints default
2571 * values. If the assoc_id field is non-zero, then the set or get
2572 * effects the specified association for the one to many model (the
2573 * assoc_id field is ignored by the one to one model). Note that if
2574 * sack_delay or sack_freq are 0 when setting this option, then the
2575 * current values will remain unchanged.
2577 * struct sctp_sack_info {
2578 * sctp_assoc_t sack_assoc_id;
2579 * uint32_t sack_delay;
2580 * uint32_t sack_freq;
2583 * sack_assoc_id - This parameter, indicates which association the user
2584 * is performing an action upon. Note that if this field's value is
2585 * zero then the endpoints default value is changed (effecting future
2586 * associations only).
2588 * sack_delay - This parameter contains the number of milliseconds that
2589 * the user is requesting the delayed ACK timer be set to. Note that
2590 * this value is defined in the standard to be between 200 and 500
2593 * sack_freq - This parameter contains the number of packets that must
2594 * be received before a sack is sent without waiting for the delay
2595 * timer to expire. The default value for this is 2, setting this
2596 * value to 1 will disable the delayed sack algorithm.
2599 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2600 char __user *optval, unsigned int optlen)
2602 struct sctp_sack_info params;
2603 struct sctp_transport *trans = NULL;
2604 struct sctp_association *asoc = NULL;
2605 struct sctp_sock *sp = sctp_sk(sk);
2607 if (optlen == sizeof(struct sctp_sack_info)) {
2608 if (copy_from_user(¶ms, optval, optlen))
2611 if (params.sack_delay == 0 && params.sack_freq == 0)
2613 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2614 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2615 pr_warn("Use struct sctp_sack_info instead\n");
2616 if (copy_from_user(¶ms, optval, optlen))
2619 if (params.sack_delay == 0)
2620 params.sack_freq = 1;
2622 params.sack_freq = 0;
2626 /* Validate value parameter. */
2627 if (params.sack_delay > 500)
2630 /* Get association, if sack_assoc_id != 0 and the socket is a one
2631 * to many style socket, and an association was not found, then
2632 * the id was invalid.
2634 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2635 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2638 if (params.sack_delay) {
2641 msecs_to_jiffies(params.sack_delay);
2643 (asoc->param_flags & ~SPP_SACKDELAY) |
2644 SPP_SACKDELAY_ENABLE;
2646 sp->sackdelay = params.sack_delay;
2648 (sp->param_flags & ~SPP_SACKDELAY) |
2649 SPP_SACKDELAY_ENABLE;
2653 if (params.sack_freq == 1) {
2656 (asoc->param_flags & ~SPP_SACKDELAY) |
2657 SPP_SACKDELAY_DISABLE;
2660 (sp->param_flags & ~SPP_SACKDELAY) |
2661 SPP_SACKDELAY_DISABLE;
2663 } else if (params.sack_freq > 1) {
2665 asoc->sackfreq = params.sack_freq;
2667 (asoc->param_flags & ~SPP_SACKDELAY) |
2668 SPP_SACKDELAY_ENABLE;
2670 sp->sackfreq = params.sack_freq;
2672 (sp->param_flags & ~SPP_SACKDELAY) |
2673 SPP_SACKDELAY_ENABLE;
2677 /* If change is for association, also apply to each transport. */
2679 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2681 if (params.sack_delay) {
2683 msecs_to_jiffies(params.sack_delay);
2684 trans->param_flags =
2685 (trans->param_flags & ~SPP_SACKDELAY) |
2686 SPP_SACKDELAY_ENABLE;
2688 if (params.sack_freq == 1) {
2689 trans->param_flags =
2690 (trans->param_flags & ~SPP_SACKDELAY) |
2691 SPP_SACKDELAY_DISABLE;
2692 } else if (params.sack_freq > 1) {
2693 trans->sackfreq = params.sack_freq;
2694 trans->param_flags =
2695 (trans->param_flags & ~SPP_SACKDELAY) |
2696 SPP_SACKDELAY_ENABLE;
2704 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2706 * Applications can specify protocol parameters for the default association
2707 * initialization. The option name argument to setsockopt() and getsockopt()
2710 * Setting initialization parameters is effective only on an unconnected
2711 * socket (for UDP-style sockets only future associations are effected
2712 * by the change). With TCP-style sockets, this option is inherited by
2713 * sockets derived from a listener socket.
2715 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2717 struct sctp_initmsg sinit;
2718 struct sctp_sock *sp = sctp_sk(sk);
2720 if (optlen != sizeof(struct sctp_initmsg))
2722 if (copy_from_user(&sinit, optval, optlen))
2725 if (sinit.sinit_num_ostreams)
2726 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2727 if (sinit.sinit_max_instreams)
2728 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2729 if (sinit.sinit_max_attempts)
2730 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2731 if (sinit.sinit_max_init_timeo)
2732 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2738 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2740 * Applications that wish to use the sendto() system call may wish to
2741 * specify a default set of parameters that would normally be supplied
2742 * through the inclusion of ancillary data. This socket option allows
2743 * such an application to set the default sctp_sndrcvinfo structure.
2744 * The application that wishes to use this socket option simply passes
2745 * in to this call the sctp_sndrcvinfo structure defined in Section
2746 * 5.2.2) The input parameters accepted by this call include
2747 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2748 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2749 * to this call if the caller is using the UDP model.
2751 static int sctp_setsockopt_default_send_param(struct sock *sk,
2752 char __user *optval,
2753 unsigned int optlen)
2755 struct sctp_sndrcvinfo info;
2756 struct sctp_association *asoc;
2757 struct sctp_sock *sp = sctp_sk(sk);
2759 if (optlen != sizeof(struct sctp_sndrcvinfo))
2761 if (copy_from_user(&info, optval, optlen))
2764 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2765 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2769 asoc->default_stream = info.sinfo_stream;
2770 asoc->default_flags = info.sinfo_flags;
2771 asoc->default_ppid = info.sinfo_ppid;
2772 asoc->default_context = info.sinfo_context;
2773 asoc->default_timetolive = info.sinfo_timetolive;
2775 sp->default_stream = info.sinfo_stream;
2776 sp->default_flags = info.sinfo_flags;
2777 sp->default_ppid = info.sinfo_ppid;
2778 sp->default_context = info.sinfo_context;
2779 sp->default_timetolive = info.sinfo_timetolive;
2785 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2787 * Requests that the local SCTP stack use the enclosed peer address as
2788 * the association primary. The enclosed address must be one of the
2789 * association peer's addresses.
2791 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2792 unsigned int optlen)
2794 struct sctp_prim prim;
2795 struct sctp_transport *trans;
2797 if (optlen != sizeof(struct sctp_prim))
2800 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2803 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2807 sctp_assoc_set_primary(trans->asoc, trans);
2813 * 7.1.5 SCTP_NODELAY
2815 * Turn on/off any Nagle-like algorithm. This means that packets are
2816 * generally sent as soon as possible and no unnecessary delays are
2817 * introduced, at the cost of more packets in the network. Expects an
2818 * integer boolean flag.
2820 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2821 unsigned int optlen)
2825 if (optlen < sizeof(int))
2827 if (get_user(val, (int __user *)optval))
2830 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2836 * 7.1.1 SCTP_RTOINFO
2838 * The protocol parameters used to initialize and bound retransmission
2839 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2840 * and modify these parameters.
2841 * All parameters are time values, in milliseconds. A value of 0, when
2842 * modifying the parameters, indicates that the current value should not
2846 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2848 struct sctp_rtoinfo rtoinfo;
2849 struct sctp_association *asoc;
2851 if (optlen != sizeof (struct sctp_rtoinfo))
2854 if (copy_from_user(&rtoinfo, optval, optlen))
2857 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2859 /* Set the values to the specific association */
2860 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2864 if (rtoinfo.srto_initial != 0)
2866 msecs_to_jiffies(rtoinfo.srto_initial);
2867 if (rtoinfo.srto_max != 0)
2868 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2869 if (rtoinfo.srto_min != 0)
2870 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2872 /* If there is no association or the association-id = 0
2873 * set the values to the endpoint.
2875 struct sctp_sock *sp = sctp_sk(sk);
2877 if (rtoinfo.srto_initial != 0)
2878 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2879 if (rtoinfo.srto_max != 0)
2880 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2881 if (rtoinfo.srto_min != 0)
2882 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2890 * 7.1.2 SCTP_ASSOCINFO
2892 * This option is used to tune the maximum retransmission attempts
2893 * of the association.
2894 * Returns an error if the new association retransmission value is
2895 * greater than the sum of the retransmission value of the peer.
2896 * See [SCTP] for more information.
2899 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2902 struct sctp_assocparams assocparams;
2903 struct sctp_association *asoc;
2905 if (optlen != sizeof(struct sctp_assocparams))
2907 if (copy_from_user(&assocparams, optval, optlen))
2910 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2912 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2915 /* Set the values to the specific association */
2917 if (assocparams.sasoc_asocmaxrxt != 0) {
2920 struct sctp_transport *peer_addr;
2922 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2924 path_sum += peer_addr->pathmaxrxt;
2928 /* Only validate asocmaxrxt if we have more than
2929 * one path/transport. We do this because path
2930 * retransmissions are only counted when we have more
2934 assocparams.sasoc_asocmaxrxt > path_sum)
2937 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2940 if (assocparams.sasoc_cookie_life != 0) {
2941 asoc->cookie_life.tv_sec =
2942 assocparams.sasoc_cookie_life / 1000;
2943 asoc->cookie_life.tv_usec =
2944 (assocparams.sasoc_cookie_life % 1000)
2948 /* Set the values to the endpoint */
2949 struct sctp_sock *sp = sctp_sk(sk);
2951 if (assocparams.sasoc_asocmaxrxt != 0)
2952 sp->assocparams.sasoc_asocmaxrxt =
2953 assocparams.sasoc_asocmaxrxt;
2954 if (assocparams.sasoc_cookie_life != 0)
2955 sp->assocparams.sasoc_cookie_life =
2956 assocparams.sasoc_cookie_life;
2962 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2964 * This socket option is a boolean flag which turns on or off mapped V4
2965 * addresses. If this option is turned on and the socket is type
2966 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2967 * If this option is turned off, then no mapping will be done of V4
2968 * addresses and a user will receive both PF_INET6 and PF_INET type
2969 * addresses on the socket.
2971 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2974 struct sctp_sock *sp = sctp_sk(sk);
2976 if (optlen < sizeof(int))
2978 if (get_user(val, (int __user *)optval))
2989 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2990 * This option will get or set the maximum size to put in any outgoing
2991 * SCTP DATA chunk. If a message is larger than this size it will be
2992 * fragmented by SCTP into the specified size. Note that the underlying
2993 * SCTP implementation may fragment into smaller sized chunks when the
2994 * PMTU of the underlying association is smaller than the value set by
2995 * the user. The default value for this option is '0' which indicates
2996 * the user is NOT limiting fragmentation and only the PMTU will effect
2997 * SCTP's choice of DATA chunk size. Note also that values set larger
2998 * than the maximum size of an IP datagram will effectively let SCTP
2999 * control fragmentation (i.e. the same as setting this option to 0).
3001 * The following structure is used to access and modify this parameter:
3003 * struct sctp_assoc_value {
3004 * sctp_assoc_t assoc_id;
3005 * uint32_t assoc_value;
3008 * assoc_id: This parameter is ignored for one-to-one style sockets.
3009 * For one-to-many style sockets this parameter indicates which
3010 * association the user is performing an action upon. Note that if
3011 * this field's value is zero then the endpoints default value is
3012 * changed (effecting future associations only).
3013 * assoc_value: This parameter specifies the maximum size in bytes.
3015 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3017 struct sctp_assoc_value params;
3018 struct sctp_association *asoc;
3019 struct sctp_sock *sp = sctp_sk(sk);
3022 if (optlen == sizeof(int)) {
3023 pr_warn("Use of int in maxseg socket option deprecated\n");
3024 pr_warn("Use struct sctp_assoc_value instead\n");
3025 if (copy_from_user(&val, optval, optlen))
3027 params.assoc_id = 0;
3028 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3029 if (copy_from_user(¶ms, optval, optlen))
3031 val = params.assoc_value;
3035 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3038 asoc = sctp_id2assoc(sk, params.assoc_id);
3039 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3044 val = asoc->pathmtu;
3045 val -= sp->pf->af->net_header_len;
3046 val -= sizeof(struct sctphdr) +
3047 sizeof(struct sctp_data_chunk);
3049 asoc->user_frag = val;
3050 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3052 sp->user_frag = val;
3060 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3062 * Requests that the peer mark the enclosed address as the association
3063 * primary. The enclosed address must be one of the association's
3064 * locally bound addresses. The following structure is used to make a
3065 * set primary request:
3067 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3068 unsigned int optlen)
3070 struct sctp_sock *sp;
3071 struct sctp_association *asoc = NULL;
3072 struct sctp_setpeerprim prim;
3073 struct sctp_chunk *chunk;
3079 if (!sctp_addip_enable)
3082 if (optlen != sizeof(struct sctp_setpeerprim))
3085 if (copy_from_user(&prim, optval, optlen))
3088 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3092 if (!asoc->peer.asconf_capable)
3095 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3098 if (!sctp_state(asoc, ESTABLISHED))
3101 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3105 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3106 return -EADDRNOTAVAIL;
3108 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3109 return -EADDRNOTAVAIL;
3111 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3112 chunk = sctp_make_asconf_set_prim(asoc,
3113 (union sctp_addr *)&prim.sspp_addr);
3117 err = sctp_send_asconf(asoc, chunk);
3119 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
3124 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3125 unsigned int optlen)
3127 struct sctp_setadaptation adaptation;
3129 if (optlen != sizeof(struct sctp_setadaptation))
3131 if (copy_from_user(&adaptation, optval, optlen))
3134 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3140 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3142 * The context field in the sctp_sndrcvinfo structure is normally only
3143 * used when a failed message is retrieved holding the value that was
3144 * sent down on the actual send call. This option allows the setting of
3145 * a default context on an association basis that will be received on
3146 * reading messages from the peer. This is especially helpful in the
3147 * one-2-many model for an application to keep some reference to an
3148 * internal state machine that is processing messages on the
3149 * association. Note that the setting of this value only effects
3150 * received messages from the peer and does not effect the value that is
3151 * saved with outbound messages.
3153 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3154 unsigned int optlen)
3156 struct sctp_assoc_value params;
3157 struct sctp_sock *sp;
3158 struct sctp_association *asoc;
3160 if (optlen != sizeof(struct sctp_assoc_value))
3162 if (copy_from_user(¶ms, optval, optlen))
3167 if (params.assoc_id != 0) {
3168 asoc = sctp_id2assoc(sk, params.assoc_id);
3171 asoc->default_rcv_context = params.assoc_value;
3173 sp->default_rcv_context = params.assoc_value;
3180 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3182 * This options will at a minimum specify if the implementation is doing
3183 * fragmented interleave. Fragmented interleave, for a one to many
3184 * socket, is when subsequent calls to receive a message may return
3185 * parts of messages from different associations. Some implementations
3186 * may allow you to turn this value on or off. If so, when turned off,
3187 * no fragment interleave will occur (which will cause a head of line
3188 * blocking amongst multiple associations sharing the same one to many
3189 * socket). When this option is turned on, then each receive call may
3190 * come from a different association (thus the user must receive data
3191 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3192 * association each receive belongs to.
3194 * This option takes a boolean value. A non-zero value indicates that
3195 * fragmented interleave is on. A value of zero indicates that
3196 * fragmented interleave is off.
3198 * Note that it is important that an implementation that allows this
3199 * option to be turned on, have it off by default. Otherwise an unaware
3200 * application using the one to many model may become confused and act
3203 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3204 char __user *optval,
3205 unsigned int optlen)
3209 if (optlen != sizeof(int))
3211 if (get_user(val, (int __user *)optval))
3214 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3220 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3221 * (SCTP_PARTIAL_DELIVERY_POINT)
3223 * This option will set or get the SCTP partial delivery point. This
3224 * point is the size of a message where the partial delivery API will be
3225 * invoked to help free up rwnd space for the peer. Setting this to a
3226 * lower value will cause partial deliveries to happen more often. The
3227 * calls argument is an integer that sets or gets the partial delivery
3228 * point. Note also that the call will fail if the user attempts to set
3229 * this value larger than the socket receive buffer size.
3231 * Note that any single message having a length smaller than or equal to
3232 * the SCTP partial delivery point will be delivered in one single read
3233 * call as long as the user provided buffer is large enough to hold the
3236 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3237 char __user *optval,
3238 unsigned int optlen)
3242 if (optlen != sizeof(u32))
3244 if (get_user(val, (int __user *)optval))
3247 /* Note: We double the receive buffer from what the user sets
3248 * it to be, also initial rwnd is based on rcvbuf/2.
3250 if (val > (sk->sk_rcvbuf >> 1))
3253 sctp_sk(sk)->pd_point = val;
3255 return 0; /* is this the right error code? */
3259 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3261 * This option will allow a user to change the maximum burst of packets
3262 * that can be emitted by this association. Note that the default value
3263 * is 4, and some implementations may restrict this setting so that it
3264 * can only be lowered.
3266 * NOTE: This text doesn't seem right. Do this on a socket basis with
3267 * future associations inheriting the socket value.
3269 static int sctp_setsockopt_maxburst(struct sock *sk,
3270 char __user *optval,
3271 unsigned int optlen)
3273 struct sctp_assoc_value params;
3274 struct sctp_sock *sp;
3275 struct sctp_association *asoc;
3279 if (optlen == sizeof(int)) {
3280 pr_warn("Use of int in max_burst socket option deprecated\n");
3281 pr_warn("Use struct sctp_assoc_value instead\n");
3282 if (copy_from_user(&val, optval, optlen))
3284 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3285 if (copy_from_user(¶ms, optval, optlen))
3287 val = params.assoc_value;
3288 assoc_id = params.assoc_id;
3294 if (assoc_id != 0) {
3295 asoc = sctp_id2assoc(sk, assoc_id);
3298 asoc->max_burst = val;
3300 sp->max_burst = val;
3306 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3308 * This set option adds a chunk type that the user is requesting to be
3309 * received only in an authenticated way. Changes to the list of chunks
3310 * will only effect future associations on the socket.
3312 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3313 char __user *optval,
3314 unsigned int optlen)
3316 struct sctp_authchunk val;
3318 if (!sctp_auth_enable)
3321 if (optlen != sizeof(struct sctp_authchunk))
3323 if (copy_from_user(&val, optval, optlen))
3326 switch (val.sauth_chunk) {
3328 case SCTP_CID_INIT_ACK:
3329 case SCTP_CID_SHUTDOWN_COMPLETE:
3334 /* add this chunk id to the endpoint */
3335 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3339 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3341 * This option gets or sets the list of HMAC algorithms that the local
3342 * endpoint requires the peer to use.
3344 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3345 char __user *optval,
3346 unsigned int optlen)
3348 struct sctp_hmacalgo *hmacs;
3352 if (!sctp_auth_enable)
3355 if (optlen < sizeof(struct sctp_hmacalgo))
3358 hmacs= memdup_user(optval, optlen);
3360 return PTR_ERR(hmacs);
3362 idents = hmacs->shmac_num_idents;
3363 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3364 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3369 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3376 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3378 * This option will set a shared secret key which is used to build an
3379 * association shared key.
3381 static int sctp_setsockopt_auth_key(struct sock *sk,
3382 char __user *optval,
3383 unsigned int optlen)
3385 struct sctp_authkey *authkey;
3386 struct sctp_association *asoc;
3389 if (!sctp_auth_enable)
3392 if (optlen <= sizeof(struct sctp_authkey))
3395 authkey= memdup_user(optval, optlen);
3396 if (IS_ERR(authkey))
3397 return PTR_ERR(authkey);
3399 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3404 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3405 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3410 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3417 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3419 * This option will get or set the active shared key to be used to build
3420 * the association shared key.
3422 static int sctp_setsockopt_active_key(struct sock *sk,
3423 char __user *optval,
3424 unsigned int optlen)
3426 struct sctp_authkeyid val;
3427 struct sctp_association *asoc;
3429 if (!sctp_auth_enable)
3432 if (optlen != sizeof(struct sctp_authkeyid))
3434 if (copy_from_user(&val, optval, optlen))
3437 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3438 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3441 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3442 val.scact_keynumber);
3446 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3448 * This set option will delete a shared secret key from use.
3450 static int sctp_setsockopt_del_key(struct sock *sk,
3451 char __user *optval,
3452 unsigned int optlen)
3454 struct sctp_authkeyid val;
3455 struct sctp_association *asoc;
3457 if (!sctp_auth_enable)
3460 if (optlen != sizeof(struct sctp_authkeyid))
3462 if (copy_from_user(&val, optval, optlen))
3465 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3466 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3469 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3470 val.scact_keynumber);
3475 * 8.1.23 SCTP_AUTO_ASCONF
3477 * This option will enable or disable the use of the automatic generation of
3478 * ASCONF chunks to add and delete addresses to an existing association. Note
3479 * that this option has two caveats namely: a) it only affects sockets that
3480 * are bound to all addresses available to the SCTP stack, and b) the system
3481 * administrator may have an overriding control that turns the ASCONF feature
3482 * off no matter what setting the socket option may have.
3483 * This option expects an integer boolean flag, where a non-zero value turns on
3484 * the option, and a zero value turns off the option.
3485 * Note. In this implementation, socket operation overrides default parameter
3486 * being set by sysctl as well as FreeBSD implementation
3488 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3489 unsigned int optlen)
3492 struct sctp_sock *sp = sctp_sk(sk);
3494 if (optlen < sizeof(int))
3496 if (get_user(val, (int __user *)optval))
3498 if (!sctp_is_ep_boundall(sk) && val)
3500 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3503 spin_lock_bh(&sctp_globals.addr_wq_lock);
3504 if (val == 0 && sp->do_auto_asconf) {
3505 list_del(&sp->auto_asconf_list);
3506 sp->do_auto_asconf = 0;
3507 } else if (val && !sp->do_auto_asconf) {
3508 list_add_tail(&sp->auto_asconf_list,
3509 &sctp_auto_asconf_splist);
3510 sp->do_auto_asconf = 1;
3512 spin_unlock_bh(&sctp_globals.addr_wq_lock);
3517 /* API 6.2 setsockopt(), getsockopt()
3519 * Applications use setsockopt() and getsockopt() to set or retrieve
3520 * socket options. Socket options are used to change the default
3521 * behavior of sockets calls. They are described in Section 7.
3525 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3526 * int __user *optlen);
3527 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3530 * sd - the socket descript.
3531 * level - set to IPPROTO_SCTP for all SCTP options.
3532 * optname - the option name.
3533 * optval - the buffer to store the value of the option.
3534 * optlen - the size of the buffer.
3536 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3537 char __user *optval, unsigned int optlen)
3541 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3544 /* I can hardly begin to describe how wrong this is. This is
3545 * so broken as to be worse than useless. The API draft
3546 * REALLY is NOT helpful here... I am not convinced that the
3547 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3548 * are at all well-founded.
3550 if (level != SOL_SCTP) {
3551 struct sctp_af *af = sctp_sk(sk)->pf->af;
3552 retval = af->setsockopt(sk, level, optname, optval, optlen);
3559 case SCTP_SOCKOPT_BINDX_ADD:
3560 /* 'optlen' is the size of the addresses buffer. */
3561 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3562 optlen, SCTP_BINDX_ADD_ADDR);
3565 case SCTP_SOCKOPT_BINDX_REM:
3566 /* 'optlen' is the size of the addresses buffer. */
3567 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3568 optlen, SCTP_BINDX_REM_ADDR);
3571 case SCTP_SOCKOPT_CONNECTX_OLD:
3572 /* 'optlen' is the size of the addresses buffer. */
3573 retval = sctp_setsockopt_connectx_old(sk,
3574 (struct sockaddr __user *)optval,
3578 case SCTP_SOCKOPT_CONNECTX:
3579 /* 'optlen' is the size of the addresses buffer. */
3580 retval = sctp_setsockopt_connectx(sk,
3581 (struct sockaddr __user *)optval,
3585 case SCTP_DISABLE_FRAGMENTS:
3586 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3590 retval = sctp_setsockopt_events(sk, optval, optlen);
3593 case SCTP_AUTOCLOSE:
3594 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3597 case SCTP_PEER_ADDR_PARAMS:
3598 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3601 case SCTP_DELAYED_SACK:
3602 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3604 case SCTP_PARTIAL_DELIVERY_POINT:
3605 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3609 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3611 case SCTP_DEFAULT_SEND_PARAM:
3612 retval = sctp_setsockopt_default_send_param(sk, optval,
3615 case SCTP_PRIMARY_ADDR:
3616 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3618 case SCTP_SET_PEER_PRIMARY_ADDR:
3619 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3622 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3625 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3627 case SCTP_ASSOCINFO:
3628 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3630 case SCTP_I_WANT_MAPPED_V4_ADDR:
3631 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3634 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3636 case SCTP_ADAPTATION_LAYER:
3637 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3640 retval = sctp_setsockopt_context(sk, optval, optlen);
3642 case SCTP_FRAGMENT_INTERLEAVE:
3643 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3645 case SCTP_MAX_BURST:
3646 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3648 case SCTP_AUTH_CHUNK:
3649 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3651 case SCTP_HMAC_IDENT:
3652 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3655 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3657 case SCTP_AUTH_ACTIVE_KEY:
3658 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3660 case SCTP_AUTH_DELETE_KEY:
3661 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3663 case SCTP_AUTO_ASCONF:
3664 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3667 retval = -ENOPROTOOPT;
3671 sctp_release_sock(sk);
3677 /* API 3.1.6 connect() - UDP Style Syntax
3679 * An application may use the connect() call in the UDP model to initiate an
3680 * association without sending data.
3684 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3686 * sd: the socket descriptor to have a new association added to.
3688 * nam: the address structure (either struct sockaddr_in or struct
3689 * sockaddr_in6 defined in RFC2553 [7]).
3691 * len: the size of the address.
3693 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3701 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3702 __func__, sk, addr, addr_len);
3704 /* Validate addr_len before calling common connect/connectx routine. */
3705 af = sctp_get_af_specific(addr->sa_family);
3706 if (!af || addr_len < af->sockaddr_len) {
3709 /* Pass correct addr len to common routine (so it knows there
3710 * is only one address being passed.
3712 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3715 sctp_release_sock(sk);
3719 /* FIXME: Write comments. */
3720 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3722 return -EOPNOTSUPP; /* STUB */
3725 /* 4.1.4 accept() - TCP Style Syntax
3727 * Applications use accept() call to remove an established SCTP
3728 * association from the accept queue of the endpoint. A new socket
3729 * descriptor will be returned from accept() to represent the newly
3730 * formed association.
3732 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3734 struct sctp_sock *sp;
3735 struct sctp_endpoint *ep;
3736 struct sock *newsk = NULL;
3737 struct sctp_association *asoc;
3746 if (!sctp_style(sk, TCP)) {
3747 error = -EOPNOTSUPP;
3751 if (!sctp_sstate(sk, LISTENING)) {
3756 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3758 error = sctp_wait_for_accept(sk, timeo);
3762 /* We treat the list of associations on the endpoint as the accept
3763 * queue and pick the first association on the list.
3765 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3767 newsk = sp->pf->create_accept_sk(sk, asoc);
3773 /* Populate the fields of the newsk from the oldsk and migrate the
3774 * asoc to the newsk.
3776 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3779 sctp_release_sock(sk);
3784 /* The SCTP ioctl handler. */
3785 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3792 * SEQPACKET-style sockets in LISTENING state are valid, for
3793 * SCTP, so only discard TCP-style sockets in LISTENING state.
3795 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3800 struct sk_buff *skb;
3801 unsigned int amount = 0;
3803 skb = skb_peek(&sk->sk_receive_queue);
3806 * We will only return the amount of this packet since
3807 * that is all that will be read.
3811 rc = put_user(amount, (int __user *)arg);
3819 sctp_release_sock(sk);
3823 /* This is the function which gets called during socket creation to
3824 * initialized the SCTP-specific portion of the sock.
3825 * The sock structure should already be zero-filled memory.
3827 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3829 struct sctp_endpoint *ep;
3830 struct sctp_sock *sp;
3832 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3836 /* Initialize the SCTP per socket area. */
3837 switch (sk->sk_type) {
3838 case SOCK_SEQPACKET:
3839 sp->type = SCTP_SOCKET_UDP;
3842 sp->type = SCTP_SOCKET_TCP;
3845 return -ESOCKTNOSUPPORT;
3848 /* Initialize default send parameters. These parameters can be
3849 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3851 sp->default_stream = 0;
3852 sp->default_ppid = 0;
3853 sp->default_flags = 0;
3854 sp->default_context = 0;
3855 sp->default_timetolive = 0;
3857 sp->default_rcv_context = 0;
3858 sp->max_burst = sctp_max_burst;
3860 /* Initialize default setup parameters. These parameters
3861 * can be modified with the SCTP_INITMSG socket option or
3862 * overridden by the SCTP_INIT CMSG.
3864 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3865 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3866 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3867 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3869 /* Initialize default RTO related parameters. These parameters can
3870 * be modified for with the SCTP_RTOINFO socket option.
3872 sp->rtoinfo.srto_initial = sctp_rto_initial;
3873 sp->rtoinfo.srto_max = sctp_rto_max;
3874 sp->rtoinfo.srto_min = sctp_rto_min;
3876 /* Initialize default association related parameters. These parameters
3877 * can be modified with the SCTP_ASSOCINFO socket option.
3879 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3880 sp->assocparams.sasoc_number_peer_destinations = 0;
3881 sp->assocparams.sasoc_peer_rwnd = 0;
3882 sp->assocparams.sasoc_local_rwnd = 0;
3883 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3885 /* Initialize default event subscriptions. By default, all the
3888 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3890 /* Default Peer Address Parameters. These defaults can
3891 * be modified via SCTP_PEER_ADDR_PARAMS
3893 sp->hbinterval = sctp_hb_interval;
3894 sp->pathmaxrxt = sctp_max_retrans_path;
3895 sp->pathmtu = 0; // allow default discovery
3896 sp->sackdelay = sctp_sack_timeout;
3898 sp->param_flags = SPP_HB_ENABLE |
3900 SPP_SACKDELAY_ENABLE;
3902 /* If enabled no SCTP message fragmentation will be performed.
3903 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3905 sp->disable_fragments = 0;
3907 /* Enable Nagle algorithm by default. */
3910 /* Enable by default. */
3913 /* Auto-close idle associations after the configured
3914 * number of seconds. A value of 0 disables this
3915 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3916 * for UDP-style sockets only.
3920 /* User specified fragmentation limit. */
3923 sp->adaptation_ind = 0;
3925 sp->pf = sctp_get_pf_specific(sk->sk_family);
3927 /* Control variables for partial data delivery. */
3928 atomic_set(&sp->pd_mode, 0);
3929 skb_queue_head_init(&sp->pd_lobby);
3930 sp->frag_interleave = 0;
3932 /* Create a per socket endpoint structure. Even if we
3933 * change the data structure relationships, this may still
3934 * be useful for storing pre-connect address information.
3936 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3943 SCTP_DBG_OBJCNT_INC(sock);
3946 percpu_counter_inc(&sctp_sockets_allocated);
3947 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3949 /* Nothing can fail after this block, otherwise
3950 * sctp_destroy_sock() will be called without addr_wq_lock held
3952 if (sctp_default_auto_asconf) {
3953 spin_lock(&sctp_globals.addr_wq_lock);
3954 list_add_tail(&sp->auto_asconf_list,
3955 &sctp_auto_asconf_splist);
3956 sp->do_auto_asconf = 1;
3957 spin_unlock(&sctp_globals.addr_wq_lock);
3959 sp->do_auto_asconf = 0;
3967 /* Cleanup any SCTP per socket resources. Must be called with
3968 * sctp_globals.addr_wq_lock held if sp->do_auto_asconf is true
3970 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3972 struct sctp_sock *sp;
3974 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3976 /* Release our hold on the endpoint. */
3978 /* This could happen during socket init, thus we bail out
3979 * early, since the rest of the below is not setup either.
3984 if (sp->do_auto_asconf) {
3985 sp->do_auto_asconf = 0;
3986 list_del(&sp->auto_asconf_list);
3988 sctp_endpoint_free(sp->ep);
3990 percpu_counter_dec(&sctp_sockets_allocated);
3991 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
3995 /* API 4.1.7 shutdown() - TCP Style Syntax
3996 * int shutdown(int socket, int how);
3998 * sd - the socket descriptor of the association to be closed.
3999 * how - Specifies the type of shutdown. The values are
4002 * Disables further receive operations. No SCTP
4003 * protocol action is taken.
4005 * Disables further send operations, and initiates
4006 * the SCTP shutdown sequence.
4008 * Disables further send and receive operations
4009 * and initiates the SCTP shutdown sequence.
4011 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
4013 struct sctp_endpoint *ep;
4014 struct sctp_association *asoc;
4016 if (!sctp_style(sk, TCP))
4019 if (how & SEND_SHUTDOWN) {
4020 ep = sctp_sk(sk)->ep;
4021 if (!list_empty(&ep->asocs)) {
4022 asoc = list_entry(ep->asocs.next,
4023 struct sctp_association, asocs);
4024 sctp_primitive_SHUTDOWN(asoc, NULL);
4029 /* 7.2.1 Association Status (SCTP_STATUS)
4031 * Applications can retrieve current status information about an
4032 * association, including association state, peer receiver window size,
4033 * number of unacked data chunks, and number of data chunks pending
4034 * receipt. This information is read-only.
4036 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4037 char __user *optval,
4040 struct sctp_status status;
4041 struct sctp_association *asoc = NULL;
4042 struct sctp_transport *transport;
4043 sctp_assoc_t associd;
4046 if (len < sizeof(status)) {
4051 len = sizeof(status);
4052 if (copy_from_user(&status, optval, len)) {
4057 associd = status.sstat_assoc_id;
4058 asoc = sctp_id2assoc(sk, associd);
4064 transport = asoc->peer.primary_path;
4066 status.sstat_assoc_id = sctp_assoc2id(asoc);
4067 status.sstat_state = asoc->state;
4068 status.sstat_rwnd = asoc->peer.rwnd;
4069 status.sstat_unackdata = asoc->unack_data;
4071 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4072 status.sstat_instrms = asoc->c.sinit_max_instreams;
4073 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4074 status.sstat_fragmentation_point = asoc->frag_point;
4075 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4076 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4077 transport->af_specific->sockaddr_len);
4078 /* Map ipv4 address into v4-mapped-on-v6 address. */
4079 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4080 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4081 status.sstat_primary.spinfo_state = transport->state;
4082 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4083 status.sstat_primary.spinfo_srtt = transport->srtt;
4084 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4085 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4087 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4088 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4090 if (put_user(len, optlen)) {
4095 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
4096 len, status.sstat_state, status.sstat_rwnd,
4097 status.sstat_assoc_id);
4099 if (copy_to_user(optval, &status, len)) {
4109 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4111 * Applications can retrieve information about a specific peer address
4112 * of an association, including its reachability state, congestion
4113 * window, and retransmission timer values. This information is
4116 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4117 char __user *optval,
4120 struct sctp_paddrinfo pinfo;
4121 struct sctp_transport *transport;
4124 if (len < sizeof(pinfo)) {
4129 len = sizeof(pinfo);
4130 if (copy_from_user(&pinfo, optval, len)) {
4135 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4136 pinfo.spinfo_assoc_id);
4140 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4141 pinfo.spinfo_state = transport->state;
4142 pinfo.spinfo_cwnd = transport->cwnd;
4143 pinfo.spinfo_srtt = transport->srtt;
4144 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4145 pinfo.spinfo_mtu = transport->pathmtu;
4147 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4148 pinfo.spinfo_state = SCTP_ACTIVE;
4150 if (put_user(len, optlen)) {
4155 if (copy_to_user(optval, &pinfo, len)) {
4164 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4166 * This option is a on/off flag. If enabled no SCTP message
4167 * fragmentation will be performed. Instead if a message being sent
4168 * exceeds the current PMTU size, the message will NOT be sent and
4169 * instead a error will be indicated to the user.
4171 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4172 char __user *optval, int __user *optlen)
4176 if (len < sizeof(int))
4180 val = (sctp_sk(sk)->disable_fragments == 1);
4181 if (put_user(len, optlen))
4183 if (copy_to_user(optval, &val, len))
4188 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4190 * This socket option is used to specify various notifications and
4191 * ancillary data the user wishes to receive.
4193 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4198 if (len > sizeof(struct sctp_event_subscribe))
4199 len = sizeof(struct sctp_event_subscribe);
4200 if (put_user(len, optlen))
4202 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4207 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4209 * This socket option is applicable to the UDP-style socket only. When
4210 * set it will cause associations that are idle for more than the
4211 * specified number of seconds to automatically close. An association
4212 * being idle is defined an association that has NOT sent or received
4213 * user data. The special value of '0' indicates that no automatic
4214 * close of any associations should be performed. The option expects an
4215 * integer defining the number of seconds of idle time before an
4216 * association is closed.
4218 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4220 /* Applicable to UDP-style socket only */
4221 if (sctp_style(sk, TCP))
4223 if (len < sizeof(int))
4226 if (put_user(len, optlen))
4228 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4233 /* Helper routine to branch off an association to a new socket. */
4234 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
4235 struct socket **sockp)
4237 struct sock *sk = asoc->base.sk;
4238 struct socket *sock;
4242 /* An association cannot be branched off from an already peeled-off
4243 * socket, nor is this supported for tcp style sockets.
4245 if (!sctp_style(sk, UDP))
4248 /* Create a new socket. */
4249 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4253 sctp_copy_sock(sock->sk, sk, asoc);
4255 /* Make peeled-off sockets more like 1-1 accepted sockets.
4256 * Set the daddr and initialize id to something more random
4258 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4259 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4261 /* Populate the fields of the newsk from the oldsk and migrate the
4262 * asoc to the newsk.
4264 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4271 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4273 sctp_peeloff_arg_t peeloff;
4274 struct socket *newsock;
4276 struct sctp_association *asoc;
4278 if (len < sizeof(sctp_peeloff_arg_t))
4280 len = sizeof(sctp_peeloff_arg_t);
4281 if (copy_from_user(&peeloff, optval, len))
4284 asoc = sctp_id2assoc(sk, peeloff.associd);
4290 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
4292 retval = sctp_do_peeloff(asoc, &newsock);
4296 /* Map the socket to an unused fd that can be returned to the user. */
4297 retval = sock_map_fd(newsock, 0);
4299 sock_release(newsock);
4303 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
4304 __func__, sk, asoc, newsock->sk, retval);
4306 /* Return the fd mapped to the new socket. */
4307 peeloff.sd = retval;
4308 if (put_user(len, optlen))
4310 if (copy_to_user(optval, &peeloff, len))
4317 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4319 * Applications can enable or disable heartbeats for any peer address of
4320 * an association, modify an address's heartbeat interval, force a
4321 * heartbeat to be sent immediately, and adjust the address's maximum
4322 * number of retransmissions sent before an address is considered
4323 * unreachable. The following structure is used to access and modify an
4324 * address's parameters:
4326 * struct sctp_paddrparams {
4327 * sctp_assoc_t spp_assoc_id;
4328 * struct sockaddr_storage spp_address;
4329 * uint32_t spp_hbinterval;
4330 * uint16_t spp_pathmaxrxt;
4331 * uint32_t spp_pathmtu;
4332 * uint32_t spp_sackdelay;
4333 * uint32_t spp_flags;
4336 * spp_assoc_id - (one-to-many style socket) This is filled in the
4337 * application, and identifies the association for
4339 * spp_address - This specifies which address is of interest.
4340 * spp_hbinterval - This contains the value of the heartbeat interval,
4341 * in milliseconds. If a value of zero
4342 * is present in this field then no changes are to
4343 * be made to this parameter.
4344 * spp_pathmaxrxt - This contains the maximum number of
4345 * retransmissions before this address shall be
4346 * considered unreachable. If a value of zero
4347 * is present in this field then no changes are to
4348 * be made to this parameter.
4349 * spp_pathmtu - When Path MTU discovery is disabled the value
4350 * specified here will be the "fixed" path mtu.
4351 * Note that if the spp_address field is empty
4352 * then all associations on this address will
4353 * have this fixed path mtu set upon them.
4355 * spp_sackdelay - When delayed sack is enabled, this value specifies
4356 * the number of milliseconds that sacks will be delayed
4357 * for. This value will apply to all addresses of an
4358 * association if the spp_address field is empty. Note
4359 * also, that if delayed sack is enabled and this
4360 * value is set to 0, no change is made to the last
4361 * recorded delayed sack timer value.
4363 * spp_flags - These flags are used to control various features
4364 * on an association. The flag field may contain
4365 * zero or more of the following options.
4367 * SPP_HB_ENABLE - Enable heartbeats on the
4368 * specified address. Note that if the address
4369 * field is empty all addresses for the association
4370 * have heartbeats enabled upon them.
4372 * SPP_HB_DISABLE - Disable heartbeats on the
4373 * speicifed address. Note that if the address
4374 * field is empty all addresses for the association
4375 * will have their heartbeats disabled. Note also
4376 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4377 * mutually exclusive, only one of these two should
4378 * be specified. Enabling both fields will have
4379 * undetermined results.
4381 * SPP_HB_DEMAND - Request a user initiated heartbeat
4382 * to be made immediately.
4384 * SPP_PMTUD_ENABLE - This field will enable PMTU
4385 * discovery upon the specified address. Note that
4386 * if the address feild is empty then all addresses
4387 * on the association are effected.
4389 * SPP_PMTUD_DISABLE - This field will disable PMTU
4390 * discovery upon the specified address. Note that
4391 * if the address feild is empty then all addresses
4392 * on the association are effected. Not also that
4393 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4394 * exclusive. Enabling both will have undetermined
4397 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4398 * on delayed sack. The time specified in spp_sackdelay
4399 * is used to specify the sack delay for this address. Note
4400 * that if spp_address is empty then all addresses will
4401 * enable delayed sack and take on the sack delay
4402 * value specified in spp_sackdelay.
4403 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4404 * off delayed sack. If the spp_address field is blank then
4405 * delayed sack is disabled for the entire association. Note
4406 * also that this field is mutually exclusive to
4407 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4410 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4411 char __user *optval, int __user *optlen)
4413 struct sctp_paddrparams params;
4414 struct sctp_transport *trans = NULL;
4415 struct sctp_association *asoc = NULL;
4416 struct sctp_sock *sp = sctp_sk(sk);
4418 if (len < sizeof(struct sctp_paddrparams))
4420 len = sizeof(struct sctp_paddrparams);
4421 if (copy_from_user(¶ms, optval, len))
4424 /* If an address other than INADDR_ANY is specified, and
4425 * no transport is found, then the request is invalid.
4427 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
4428 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4429 params.spp_assoc_id);
4431 SCTP_DEBUG_PRINTK("Failed no transport\n");
4436 /* Get association, if assoc_id != 0 and the socket is a one
4437 * to many style socket, and an association was not found, then
4438 * the id was invalid.
4440 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4441 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4442 SCTP_DEBUG_PRINTK("Failed no association\n");
4447 /* Fetch transport values. */
4448 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4449 params.spp_pathmtu = trans->pathmtu;
4450 params.spp_pathmaxrxt = trans->pathmaxrxt;
4451 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4453 /*draft-11 doesn't say what to return in spp_flags*/
4454 params.spp_flags = trans->param_flags;
4456 /* Fetch association values. */
4457 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4458 params.spp_pathmtu = asoc->pathmtu;
4459 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4460 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4462 /*draft-11 doesn't say what to return in spp_flags*/
4463 params.spp_flags = asoc->param_flags;
4465 /* Fetch socket values. */
4466 params.spp_hbinterval = sp->hbinterval;
4467 params.spp_pathmtu = sp->pathmtu;
4468 params.spp_sackdelay = sp->sackdelay;
4469 params.spp_pathmaxrxt = sp->pathmaxrxt;
4471 /*draft-11 doesn't say what to return in spp_flags*/
4472 params.spp_flags = sp->param_flags;
4475 if (copy_to_user(optval, ¶ms, len))
4478 if (put_user(len, optlen))
4485 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4487 * This option will effect the way delayed acks are performed. This
4488 * option allows you to get or set the delayed ack time, in
4489 * milliseconds. It also allows changing the delayed ack frequency.
4490 * Changing the frequency to 1 disables the delayed sack algorithm. If
4491 * the assoc_id is 0, then this sets or gets the endpoints default
4492 * values. If the assoc_id field is non-zero, then the set or get
4493 * effects the specified association for the one to many model (the
4494 * assoc_id field is ignored by the one to one model). Note that if
4495 * sack_delay or sack_freq are 0 when setting this option, then the
4496 * current values will remain unchanged.
4498 * struct sctp_sack_info {
4499 * sctp_assoc_t sack_assoc_id;
4500 * uint32_t sack_delay;
4501 * uint32_t sack_freq;
4504 * sack_assoc_id - This parameter, indicates which association the user
4505 * is performing an action upon. Note that if this field's value is
4506 * zero then the endpoints default value is changed (effecting future
4507 * associations only).
4509 * sack_delay - This parameter contains the number of milliseconds that
4510 * the user is requesting the delayed ACK timer be set to. Note that
4511 * this value is defined in the standard to be between 200 and 500
4514 * sack_freq - This parameter contains the number of packets that must
4515 * be received before a sack is sent without waiting for the delay
4516 * timer to expire. The default value for this is 2, setting this
4517 * value to 1 will disable the delayed sack algorithm.
4519 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4520 char __user *optval,
4523 struct sctp_sack_info params;
4524 struct sctp_association *asoc = NULL;
4525 struct sctp_sock *sp = sctp_sk(sk);
4527 if (len >= sizeof(struct sctp_sack_info)) {
4528 len = sizeof(struct sctp_sack_info);
4530 if (copy_from_user(¶ms, optval, len))
4532 } else if (len == sizeof(struct sctp_assoc_value)) {
4533 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4534 pr_warn("Use struct sctp_sack_info instead\n");
4535 if (copy_from_user(¶ms, optval, len))
4540 /* Get association, if sack_assoc_id != 0 and the socket is a one
4541 * to many style socket, and an association was not found, then
4542 * the id was invalid.
4544 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4545 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4549 /* Fetch association values. */
4550 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4551 params.sack_delay = jiffies_to_msecs(
4553 params.sack_freq = asoc->sackfreq;
4556 params.sack_delay = 0;
4557 params.sack_freq = 1;
4560 /* Fetch socket values. */
4561 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4562 params.sack_delay = sp->sackdelay;
4563 params.sack_freq = sp->sackfreq;
4565 params.sack_delay = 0;
4566 params.sack_freq = 1;
4570 if (copy_to_user(optval, ¶ms, len))
4573 if (put_user(len, optlen))
4579 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4581 * Applications can specify protocol parameters for the default association
4582 * initialization. The option name argument to setsockopt() and getsockopt()
4585 * Setting initialization parameters is effective only on an unconnected
4586 * socket (for UDP-style sockets only future associations are effected
4587 * by the change). With TCP-style sockets, this option is inherited by
4588 * sockets derived from a listener socket.
4590 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4592 if (len < sizeof(struct sctp_initmsg))
4594 len = sizeof(struct sctp_initmsg);
4595 if (put_user(len, optlen))
4597 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4603 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4604 char __user *optval, int __user *optlen)
4606 struct sctp_association *asoc;
4608 struct sctp_getaddrs getaddrs;
4609 struct sctp_transport *from;
4611 union sctp_addr temp;
4612 struct sctp_sock *sp = sctp_sk(sk);
4617 if (len < sizeof(struct sctp_getaddrs))
4620 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4623 /* For UDP-style sockets, id specifies the association to query. */
4624 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4628 to = optval + offsetof(struct sctp_getaddrs,addrs);
4629 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4631 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4633 memcpy(&temp, &from->ipaddr, sizeof(temp));
4634 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4635 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4636 if (space_left < addrlen)
4638 if (copy_to_user(to, &temp, addrlen))
4642 space_left -= addrlen;
4645 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4647 bytes_copied = ((char __user *)to) - optval;
4648 if (put_user(bytes_copied, optlen))
4654 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4655 size_t space_left, int *bytes_copied)
4657 struct sctp_sockaddr_entry *addr;
4658 union sctp_addr temp;
4663 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4667 if ((PF_INET == sk->sk_family) &&
4668 (AF_INET6 == addr->a.sa.sa_family))
4670 if ((PF_INET6 == sk->sk_family) &&
4671 inet_v6_ipv6only(sk) &&
4672 (AF_INET == addr->a.sa.sa_family))
4674 memcpy(&temp, &addr->a, sizeof(temp));
4675 if (!temp.v4.sin_port)
4676 temp.v4.sin_port = htons(port);
4678 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4680 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4681 if (space_left < addrlen) {
4685 memcpy(to, &temp, addrlen);
4689 space_left -= addrlen;
4690 *bytes_copied += addrlen;
4698 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4699 char __user *optval, int __user *optlen)
4701 struct sctp_bind_addr *bp;
4702 struct sctp_association *asoc;
4704 struct sctp_getaddrs getaddrs;
4705 struct sctp_sockaddr_entry *addr;
4707 union sctp_addr temp;
4708 struct sctp_sock *sp = sctp_sk(sk);
4712 int bytes_copied = 0;
4716 if (len < sizeof(struct sctp_getaddrs))
4719 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4723 * For UDP-style sockets, id specifies the association to query.
4724 * If the id field is set to the value '0' then the locally bound
4725 * addresses are returned without regard to any particular
4728 if (0 == getaddrs.assoc_id) {
4729 bp = &sctp_sk(sk)->ep->base.bind_addr;
4731 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4734 bp = &asoc->base.bind_addr;
4737 to = optval + offsetof(struct sctp_getaddrs,addrs);
4738 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4740 addrs = kmalloc(space_left, GFP_KERNEL);
4744 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4745 * addresses from the global local address list.
4747 if (sctp_list_single_entry(&bp->address_list)) {
4748 addr = list_entry(bp->address_list.next,
4749 struct sctp_sockaddr_entry, list);
4750 if (sctp_is_any(sk, &addr->a)) {
4751 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4752 space_left, &bytes_copied);
4762 /* Protection on the bound address list is not needed since
4763 * in the socket option context we hold a socket lock and
4764 * thus the bound address list can't change.
4766 list_for_each_entry(addr, &bp->address_list, list) {
4767 memcpy(&temp, &addr->a, sizeof(temp));
4768 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4769 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4770 if (space_left < addrlen) {
4771 err = -ENOMEM; /*fixme: right error?*/
4774 memcpy(buf, &temp, addrlen);
4776 bytes_copied += addrlen;
4778 space_left -= addrlen;
4782 if (copy_to_user(to, addrs, bytes_copied)) {
4786 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4790 if (put_user(bytes_copied, optlen))
4797 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4799 * Requests that the local SCTP stack use the enclosed peer address as
4800 * the association primary. The enclosed address must be one of the
4801 * association peer's addresses.
4803 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4804 char __user *optval, int __user *optlen)
4806 struct sctp_prim prim;
4807 struct sctp_association *asoc;
4808 struct sctp_sock *sp = sctp_sk(sk);
4810 if (len < sizeof(struct sctp_prim))
4813 len = sizeof(struct sctp_prim);
4815 if (copy_from_user(&prim, optval, len))
4818 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4822 if (!asoc->peer.primary_path)
4825 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4826 asoc->peer.primary_path->af_specific->sockaddr_len);
4828 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4829 (union sctp_addr *)&prim.ssp_addr);
4831 if (put_user(len, optlen))
4833 if (copy_to_user(optval, &prim, len))
4840 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4842 * Requests that the local endpoint set the specified Adaptation Layer
4843 * Indication parameter for all future INIT and INIT-ACK exchanges.
4845 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4846 char __user *optval, int __user *optlen)
4848 struct sctp_setadaptation adaptation;
4850 if (len < sizeof(struct sctp_setadaptation))
4853 len = sizeof(struct sctp_setadaptation);
4855 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4857 if (put_user(len, optlen))
4859 if (copy_to_user(optval, &adaptation, len))
4867 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4869 * Applications that wish to use the sendto() system call may wish to
4870 * specify a default set of parameters that would normally be supplied
4871 * through the inclusion of ancillary data. This socket option allows
4872 * such an application to set the default sctp_sndrcvinfo structure.
4875 * The application that wishes to use this socket option simply passes
4876 * in to this call the sctp_sndrcvinfo structure defined in Section
4877 * 5.2.2) The input parameters accepted by this call include
4878 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4879 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4880 * to this call if the caller is using the UDP model.
4882 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4884 static int sctp_getsockopt_default_send_param(struct sock *sk,
4885 int len, char __user *optval,
4888 struct sctp_sndrcvinfo info;
4889 struct sctp_association *asoc;
4890 struct sctp_sock *sp = sctp_sk(sk);
4892 if (len < sizeof(struct sctp_sndrcvinfo))
4895 len = sizeof(struct sctp_sndrcvinfo);
4897 if (copy_from_user(&info, optval, len))
4900 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4901 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4905 info.sinfo_stream = asoc->default_stream;
4906 info.sinfo_flags = asoc->default_flags;
4907 info.sinfo_ppid = asoc->default_ppid;
4908 info.sinfo_context = asoc->default_context;
4909 info.sinfo_timetolive = asoc->default_timetolive;
4911 info.sinfo_stream = sp->default_stream;
4912 info.sinfo_flags = sp->default_flags;
4913 info.sinfo_ppid = sp->default_ppid;
4914 info.sinfo_context = sp->default_context;
4915 info.sinfo_timetolive = sp->default_timetolive;
4918 if (put_user(len, optlen))
4920 if (copy_to_user(optval, &info, len))
4928 * 7.1.5 SCTP_NODELAY
4930 * Turn on/off any Nagle-like algorithm. This means that packets are
4931 * generally sent as soon as possible and no unnecessary delays are
4932 * introduced, at the cost of more packets in the network. Expects an
4933 * integer boolean flag.
4936 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4937 char __user *optval, int __user *optlen)
4941 if (len < sizeof(int))
4945 val = (sctp_sk(sk)->nodelay == 1);
4946 if (put_user(len, optlen))
4948 if (copy_to_user(optval, &val, len))
4955 * 7.1.1 SCTP_RTOINFO
4957 * The protocol parameters used to initialize and bound retransmission
4958 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4959 * and modify these parameters.
4960 * All parameters are time values, in milliseconds. A value of 0, when
4961 * modifying the parameters, indicates that the current value should not
4965 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4966 char __user *optval,
4967 int __user *optlen) {
4968 struct sctp_rtoinfo rtoinfo;
4969 struct sctp_association *asoc;
4971 if (len < sizeof (struct sctp_rtoinfo))
4974 len = sizeof(struct sctp_rtoinfo);
4976 if (copy_from_user(&rtoinfo, optval, len))
4979 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4981 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4984 /* Values corresponding to the specific association. */
4986 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4987 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4988 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4990 /* Values corresponding to the endpoint. */
4991 struct sctp_sock *sp = sctp_sk(sk);
4993 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4994 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4995 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4998 if (put_user(len, optlen))
5001 if (copy_to_user(optval, &rtoinfo, len))
5009 * 7.1.2 SCTP_ASSOCINFO
5011 * This option is used to tune the maximum retransmission attempts
5012 * of the association.
5013 * Returns an error if the new association retransmission value is
5014 * greater than the sum of the retransmission value of the peer.
5015 * See [SCTP] for more information.
5018 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5019 char __user *optval,
5023 struct sctp_assocparams assocparams;
5024 struct sctp_association *asoc;
5025 struct list_head *pos;
5028 if (len < sizeof (struct sctp_assocparams))
5031 len = sizeof(struct sctp_assocparams);
5033 if (copy_from_user(&assocparams, optval, len))
5036 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5038 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5041 /* Values correspoinding to the specific association */
5043 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5044 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5045 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5046 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
5048 (asoc->cookie_life.tv_usec
5051 list_for_each(pos, &asoc->peer.transport_addr_list) {
5055 assocparams.sasoc_number_peer_destinations = cnt;
5057 /* Values corresponding to the endpoint */
5058 struct sctp_sock *sp = sctp_sk(sk);
5060 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5061 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5062 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5063 assocparams.sasoc_cookie_life =
5064 sp->assocparams.sasoc_cookie_life;
5065 assocparams.sasoc_number_peer_destinations =
5067 sasoc_number_peer_destinations;
5070 if (put_user(len, optlen))
5073 if (copy_to_user(optval, &assocparams, len))
5080 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5082 * This socket option is a boolean flag which turns on or off mapped V4
5083 * addresses. If this option is turned on and the socket is type
5084 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5085 * If this option is turned off, then no mapping will be done of V4
5086 * addresses and a user will receive both PF_INET6 and PF_INET type
5087 * addresses on the socket.
5089 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5090 char __user *optval, int __user *optlen)
5093 struct sctp_sock *sp = sctp_sk(sk);
5095 if (len < sizeof(int))
5100 if (put_user(len, optlen))
5102 if (copy_to_user(optval, &val, len))
5109 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5110 * (chapter and verse is quoted at sctp_setsockopt_context())
5112 static int sctp_getsockopt_context(struct sock *sk, int len,
5113 char __user *optval, int __user *optlen)
5115 struct sctp_assoc_value params;
5116 struct sctp_sock *sp;
5117 struct sctp_association *asoc;
5119 if (len < sizeof(struct sctp_assoc_value))
5122 len = sizeof(struct sctp_assoc_value);
5124 if (copy_from_user(¶ms, optval, len))
5129 if (params.assoc_id != 0) {
5130 asoc = sctp_id2assoc(sk, params.assoc_id);
5133 params.assoc_value = asoc->default_rcv_context;
5135 params.assoc_value = sp->default_rcv_context;
5138 if (put_user(len, optlen))
5140 if (copy_to_user(optval, ¶ms, len))
5147 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5148 * This option will get or set the maximum size to put in any outgoing
5149 * SCTP DATA chunk. If a message is larger than this size it will be
5150 * fragmented by SCTP into the specified size. Note that the underlying
5151 * SCTP implementation may fragment into smaller sized chunks when the
5152 * PMTU of the underlying association is smaller than the value set by
5153 * the user. The default value for this option is '0' which indicates
5154 * the user is NOT limiting fragmentation and only the PMTU will effect
5155 * SCTP's choice of DATA chunk size. Note also that values set larger
5156 * than the maximum size of an IP datagram will effectively let SCTP
5157 * control fragmentation (i.e. the same as setting this option to 0).
5159 * The following structure is used to access and modify this parameter:
5161 * struct sctp_assoc_value {
5162 * sctp_assoc_t assoc_id;
5163 * uint32_t assoc_value;
5166 * assoc_id: This parameter is ignored for one-to-one style sockets.
5167 * For one-to-many style sockets this parameter indicates which
5168 * association the user is performing an action upon. Note that if
5169 * this field's value is zero then the endpoints default value is
5170 * changed (effecting future associations only).
5171 * assoc_value: This parameter specifies the maximum size in bytes.
5173 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5174 char __user *optval, int __user *optlen)
5176 struct sctp_assoc_value params;
5177 struct sctp_association *asoc;
5179 if (len == sizeof(int)) {
5180 pr_warn("Use of int in maxseg socket option deprecated\n");
5181 pr_warn("Use struct sctp_assoc_value instead\n");
5182 params.assoc_id = 0;
5183 } else if (len >= sizeof(struct sctp_assoc_value)) {
5184 len = sizeof(struct sctp_assoc_value);
5185 if (copy_from_user(¶ms, optval, sizeof(params)))
5190 asoc = sctp_id2assoc(sk, params.assoc_id);
5191 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5195 params.assoc_value = asoc->frag_point;
5197 params.assoc_value = sctp_sk(sk)->user_frag;
5199 if (put_user(len, optlen))
5201 if (len == sizeof(int)) {
5202 if (copy_to_user(optval, ¶ms.assoc_value, len))
5205 if (copy_to_user(optval, ¶ms, len))
5213 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5214 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5216 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5217 char __user *optval, int __user *optlen)
5221 if (len < sizeof(int))
5226 val = sctp_sk(sk)->frag_interleave;
5227 if (put_user(len, optlen))
5229 if (copy_to_user(optval, &val, len))
5236 * 7.1.25. Set or Get the sctp partial delivery point
5237 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5239 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5240 char __user *optval,
5245 if (len < sizeof(u32))
5250 val = sctp_sk(sk)->pd_point;
5251 if (put_user(len, optlen))
5253 if (copy_to_user(optval, &val, len))
5260 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5261 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5263 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5264 char __user *optval,
5267 struct sctp_assoc_value params;
5268 struct sctp_sock *sp;
5269 struct sctp_association *asoc;
5271 if (len == sizeof(int)) {
5272 pr_warn("Use of int in max_burst socket option deprecated\n");
5273 pr_warn("Use struct sctp_assoc_value instead\n");
5274 params.assoc_id = 0;
5275 } else if (len >= sizeof(struct sctp_assoc_value)) {
5276 len = sizeof(struct sctp_assoc_value);
5277 if (copy_from_user(¶ms, optval, len))
5284 if (params.assoc_id != 0) {
5285 asoc = sctp_id2assoc(sk, params.assoc_id);
5288 params.assoc_value = asoc->max_burst;
5290 params.assoc_value = sp->max_burst;
5292 if (len == sizeof(int)) {
5293 if (copy_to_user(optval, ¶ms.assoc_value, len))
5296 if (copy_to_user(optval, ¶ms, len))
5304 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5305 char __user *optval, int __user *optlen)
5307 struct sctp_hmacalgo __user *p = (void __user *)optval;
5308 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 for (i = 0; i < num_idents; i++) {
5330 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
5332 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
5338 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5339 char __user *optval, int __user *optlen)
5341 struct sctp_authkeyid val;
5342 struct sctp_association *asoc;
5344 if (!sctp_auth_enable)
5347 if (len < sizeof(struct sctp_authkeyid))
5349 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5352 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5353 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5357 val.scact_keynumber = asoc->active_key_id;
5359 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5361 len = sizeof(struct sctp_authkeyid);
5362 if (put_user(len, optlen))
5364 if (copy_to_user(optval, &val, len))
5370 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5371 char __user *optval, int __user *optlen)
5373 struct sctp_authchunks __user *p = (void __user *)optval;
5374 struct sctp_authchunks val;
5375 struct sctp_association *asoc;
5376 struct sctp_chunks_param *ch;
5380 if (!sctp_auth_enable)
5383 if (len < sizeof(struct sctp_authchunks))
5386 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5389 to = p->gauth_chunks;
5390 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5394 ch = asoc->peer.peer_chunks;
5398 /* See if the user provided enough room for all the data */
5399 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5400 if (len < num_chunks)
5403 if (copy_to_user(to, ch->chunks, num_chunks))
5406 len = sizeof(struct sctp_authchunks) + num_chunks;
5407 if (put_user(len, optlen)) return -EFAULT;
5408 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5413 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5414 char __user *optval, int __user *optlen)
5416 struct sctp_authchunks __user *p = (void __user *)optval;
5417 struct sctp_authchunks val;
5418 struct sctp_association *asoc;
5419 struct sctp_chunks_param *ch;
5423 if (!sctp_auth_enable)
5426 if (len < sizeof(struct sctp_authchunks))
5429 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5432 to = p->gauth_chunks;
5433 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5434 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5438 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5440 ch = sctp_sk(sk)->ep->auth_chunk_list;
5445 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5446 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5449 if (copy_to_user(to, ch->chunks, num_chunks))
5452 len = sizeof(struct sctp_authchunks) + num_chunks;
5453 if (put_user(len, optlen))
5455 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5462 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5463 * This option gets the current number of associations that are attached
5464 * to a one-to-many style socket. The option value is an uint32_t.
5466 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5467 char __user *optval, int __user *optlen)
5469 struct sctp_sock *sp = sctp_sk(sk);
5470 struct sctp_association *asoc;
5473 if (sctp_style(sk, TCP))
5476 if (len < sizeof(u32))
5481 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5485 if (put_user(len, optlen))
5487 if (copy_to_user(optval, &val, len))
5494 * 8.1.23 SCTP_AUTO_ASCONF
5495 * See the corresponding setsockopt entry as description
5497 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5498 char __user *optval, int __user *optlen)
5502 if (len < sizeof(int))
5506 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5508 if (put_user(len, optlen))
5510 if (copy_to_user(optval, &val, len))
5516 * 8.2.6. Get the Current Identifiers of Associations
5517 * (SCTP_GET_ASSOC_ID_LIST)
5519 * This option gets the current list of SCTP association identifiers of
5520 * the SCTP associations handled by a one-to-many style socket.
5522 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5523 char __user *optval, int __user *optlen)
5525 struct sctp_sock *sp = sctp_sk(sk);
5526 struct sctp_association *asoc;
5527 struct sctp_assoc_ids *ids;
5530 if (sctp_style(sk, TCP))
5533 if (len < sizeof(struct sctp_assoc_ids))
5536 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5540 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5543 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5545 ids = kmalloc(len, GFP_KERNEL);
5549 ids->gaids_number_of_ids = num;
5551 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5552 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5555 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5564 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5565 char __user *optval, int __user *optlen)
5570 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5573 /* I can hardly begin to describe how wrong this is. This is
5574 * so broken as to be worse than useless. The API draft
5575 * REALLY is NOT helpful here... I am not convinced that the
5576 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5577 * are at all well-founded.
5579 if (level != SOL_SCTP) {
5580 struct sctp_af *af = sctp_sk(sk)->pf->af;
5582 retval = af->getsockopt(sk, level, optname, optval, optlen);
5586 if (get_user(len, optlen))
5593 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5595 case SCTP_DISABLE_FRAGMENTS:
5596 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5600 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5602 case SCTP_AUTOCLOSE:
5603 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5605 case SCTP_SOCKOPT_PEELOFF:
5606 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5608 case SCTP_PEER_ADDR_PARAMS:
5609 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5612 case SCTP_DELAYED_SACK:
5613 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5617 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5619 case SCTP_GET_PEER_ADDRS:
5620 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5623 case SCTP_GET_LOCAL_ADDRS:
5624 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5627 case SCTP_SOCKOPT_CONNECTX3:
5628 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5630 case SCTP_DEFAULT_SEND_PARAM:
5631 retval = sctp_getsockopt_default_send_param(sk, len,
5634 case SCTP_PRIMARY_ADDR:
5635 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5638 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5641 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5643 case SCTP_ASSOCINFO:
5644 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5646 case SCTP_I_WANT_MAPPED_V4_ADDR:
5647 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5650 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5652 case SCTP_GET_PEER_ADDR_INFO:
5653 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5656 case SCTP_ADAPTATION_LAYER:
5657 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5661 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5663 case SCTP_FRAGMENT_INTERLEAVE:
5664 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5667 case SCTP_PARTIAL_DELIVERY_POINT:
5668 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5671 case SCTP_MAX_BURST:
5672 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5675 case SCTP_AUTH_CHUNK:
5676 case SCTP_AUTH_DELETE_KEY:
5677 retval = -EOPNOTSUPP;
5679 case SCTP_HMAC_IDENT:
5680 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5682 case SCTP_AUTH_ACTIVE_KEY:
5683 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5685 case SCTP_PEER_AUTH_CHUNKS:
5686 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5689 case SCTP_LOCAL_AUTH_CHUNKS:
5690 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5693 case SCTP_GET_ASSOC_NUMBER:
5694 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5696 case SCTP_GET_ASSOC_ID_LIST:
5697 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5699 case SCTP_AUTO_ASCONF:
5700 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5703 retval = -ENOPROTOOPT;
5707 sctp_release_sock(sk);
5711 static void sctp_hash(struct sock *sk)
5716 static void sctp_unhash(struct sock *sk)
5721 /* Check if port is acceptable. Possibly find first available port.
5723 * The port hash table (contained in the 'global' SCTP protocol storage
5724 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5725 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5726 * list (the list number is the port number hashed out, so as you
5727 * would expect from a hash function, all the ports in a given list have
5728 * such a number that hashes out to the same list number; you were
5729 * expecting that, right?); so each list has a set of ports, with a
5730 * link to the socket (struct sock) that uses it, the port number and
5731 * a fastreuse flag (FIXME: NPI ipg).
5733 static struct sctp_bind_bucket *sctp_bucket_create(
5734 struct sctp_bind_hashbucket *head, unsigned short snum);
5736 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5738 struct sctp_bind_hashbucket *head; /* hash list */
5739 struct sctp_bind_bucket *pp; /* hash list port iterator */
5740 struct hlist_node *node;
5741 unsigned short snum;
5744 snum = ntohs(addr->v4.sin_port);
5746 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5747 sctp_local_bh_disable();
5750 /* Search for an available port. */
5751 int low, high, remaining, index;
5754 inet_get_local_port_range(&low, &high);
5755 remaining = (high - low) + 1;
5756 rover = net_random() % remaining + low;
5760 if ((rover < low) || (rover > high))
5762 if (inet_is_reserved_local_port(rover))
5764 index = sctp_phashfn(rover);
5765 head = &sctp_port_hashtable[index];
5766 sctp_spin_lock(&head->lock);
5767 sctp_for_each_hentry(pp, node, &head->chain)
5768 if (pp->port == rover)
5772 sctp_spin_unlock(&head->lock);
5773 } while (--remaining > 0);
5775 /* Exhausted local port range during search? */
5780 /* OK, here is the one we will use. HEAD (the port
5781 * hash table list entry) is non-NULL and we hold it's
5786 /* We are given an specific port number; we verify
5787 * that it is not being used. If it is used, we will
5788 * exahust the search in the hash list corresponding
5789 * to the port number (snum) - we detect that with the
5790 * port iterator, pp being NULL.
5792 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5793 sctp_spin_lock(&head->lock);
5794 sctp_for_each_hentry(pp, node, &head->chain) {
5795 if (pp->port == snum)
5802 if (!hlist_empty(&pp->owner)) {
5803 /* We had a port hash table hit - there is an
5804 * available port (pp != NULL) and it is being
5805 * used by other socket (pp->owner not empty); that other
5806 * socket is going to be sk2.
5808 int reuse = sk->sk_reuse;
5811 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5812 if (pp->fastreuse && sk->sk_reuse &&
5813 sk->sk_state != SCTP_SS_LISTENING)
5816 /* Run through the list of sockets bound to the port
5817 * (pp->port) [via the pointers bind_next and
5818 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5819 * we get the endpoint they describe and run through
5820 * the endpoint's list of IP (v4 or v6) addresses,
5821 * comparing each of the addresses with the address of
5822 * the socket sk. If we find a match, then that means
5823 * that this port/socket (sk) combination are already
5826 sk_for_each_bound(sk2, node, &pp->owner) {
5827 struct sctp_endpoint *ep2;
5828 ep2 = sctp_sk(sk2)->ep;
5831 (reuse && sk2->sk_reuse &&
5832 sk2->sk_state != SCTP_SS_LISTENING))
5835 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5836 sctp_sk(sk2), sctp_sk(sk))) {
5841 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5844 /* If there was a hash table miss, create a new port. */
5846 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5849 /* In either case (hit or miss), make sure fastreuse is 1 only
5850 * if sk->sk_reuse is too (that is, if the caller requested
5851 * SO_REUSEADDR on this socket -sk-).
5853 if (hlist_empty(&pp->owner)) {
5854 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5858 } else if (pp->fastreuse &&
5859 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5862 /* We are set, so fill up all the data in the hash table
5863 * entry, tie the socket list information with the rest of the
5864 * sockets FIXME: Blurry, NPI (ipg).
5867 if (!sctp_sk(sk)->bind_hash) {
5868 inet_sk(sk)->inet_num = snum;
5869 sk_add_bind_node(sk, &pp->owner);
5870 sctp_sk(sk)->bind_hash = pp;
5875 sctp_spin_unlock(&head->lock);
5878 sctp_local_bh_enable();
5882 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5883 * port is requested.
5885 static int sctp_get_port(struct sock *sk, unsigned short snum)
5888 union sctp_addr addr;
5889 struct sctp_af *af = sctp_sk(sk)->pf->af;
5891 /* Set up a dummy address struct from the sk. */
5892 af->from_sk(&addr, sk);
5893 addr.v4.sin_port = htons(snum);
5895 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5896 ret = sctp_get_port_local(sk, &addr);
5902 * Move a socket to LISTENING state.
5904 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5906 struct sctp_sock *sp = sctp_sk(sk);
5907 struct sctp_endpoint *ep = sp->ep;
5908 struct crypto_hash *tfm = NULL;
5910 /* Allocate HMAC for generating cookie. */
5911 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5912 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5914 if (net_ratelimit()) {
5915 pr_info("failed to load transform for %s: %ld\n",
5916 sctp_hmac_alg, PTR_ERR(tfm));
5920 sctp_sk(sk)->hmac = tfm;
5924 * If a bind() or sctp_bindx() is not called prior to a listen()
5925 * call that allows new associations to be accepted, the system
5926 * picks an ephemeral port and will choose an address set equivalent
5927 * to binding with a wildcard address.
5929 * This is not currently spelled out in the SCTP sockets
5930 * extensions draft, but follows the practice as seen in TCP
5934 sk->sk_state = SCTP_SS_LISTENING;
5935 if (!ep->base.bind_addr.port) {
5936 if (sctp_autobind(sk))
5939 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
5940 sk->sk_state = SCTP_SS_CLOSED;
5945 sk->sk_max_ack_backlog = backlog;
5946 sctp_hash_endpoint(ep);
5951 * 4.1.3 / 5.1.3 listen()
5953 * By default, new associations are not accepted for UDP style sockets.
5954 * An application uses listen() to mark a socket as being able to
5955 * accept new associations.
5957 * On TCP style sockets, applications use listen() to ready the SCTP
5958 * endpoint for accepting inbound associations.
5960 * On both types of endpoints a backlog of '0' disables listening.
5962 * Move a socket to LISTENING state.
5964 int sctp_inet_listen(struct socket *sock, int backlog)
5966 struct sock *sk = sock->sk;
5967 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5970 if (unlikely(backlog < 0))
5975 /* Peeled-off sockets are not allowed to listen(). */
5976 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
5979 if (sock->state != SS_UNCONNECTED)
5982 /* If backlog is zero, disable listening. */
5984 if (sctp_sstate(sk, CLOSED))
5988 sctp_unhash_endpoint(ep);
5989 sk->sk_state = SCTP_SS_CLOSED;
5991 sctp_sk(sk)->bind_hash->fastreuse = 1;
5995 /* If we are already listening, just update the backlog */
5996 if (sctp_sstate(sk, LISTENING))
5997 sk->sk_max_ack_backlog = backlog;
5999 err = sctp_listen_start(sk, backlog);
6006 sctp_release_sock(sk);
6011 * This function is done by modeling the current datagram_poll() and the
6012 * tcp_poll(). Note that, based on these implementations, we don't
6013 * lock the socket in this function, even though it seems that,
6014 * ideally, locking or some other mechanisms can be used to ensure
6015 * the integrity of the counters (sndbuf and wmem_alloc) used
6016 * in this place. We assume that we don't need locks either until proven
6019 * Another thing to note is that we include the Async I/O support
6020 * here, again, by modeling the current TCP/UDP code. We don't have
6021 * a good way to test with it yet.
6023 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6025 struct sock *sk = sock->sk;
6026 struct sctp_sock *sp = sctp_sk(sk);
6029 poll_wait(file, sk_sleep(sk), wait);
6031 /* A TCP-style listening socket becomes readable when the accept queue
6034 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6035 return (!list_empty(&sp->ep->asocs)) ?
6036 (POLLIN | POLLRDNORM) : 0;
6040 /* Is there any exceptional events? */
6041 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6043 if (sk->sk_shutdown & RCV_SHUTDOWN)
6044 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6045 if (sk->sk_shutdown == SHUTDOWN_MASK)
6048 /* Is it readable? Reconsider this code with TCP-style support. */
6049 if (!skb_queue_empty(&sk->sk_receive_queue))
6050 mask |= POLLIN | POLLRDNORM;
6052 /* The association is either gone or not ready. */
6053 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6056 /* Is it writable? */
6057 if (sctp_writeable(sk)) {
6058 mask |= POLLOUT | POLLWRNORM;
6060 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6062 * Since the socket is not locked, the buffer
6063 * might be made available after the writeable check and
6064 * before the bit is set. This could cause a lost I/O
6065 * signal. tcp_poll() has a race breaker for this race
6066 * condition. Based on their implementation, we put
6067 * in the following code to cover it as well.
6069 if (sctp_writeable(sk))
6070 mask |= POLLOUT | POLLWRNORM;
6075 /********************************************************************
6076 * 2nd Level Abstractions
6077 ********************************************************************/
6079 static struct sctp_bind_bucket *sctp_bucket_create(
6080 struct sctp_bind_hashbucket *head, unsigned short snum)
6082 struct sctp_bind_bucket *pp;
6084 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6086 SCTP_DBG_OBJCNT_INC(bind_bucket);
6089 INIT_HLIST_HEAD(&pp->owner);
6090 hlist_add_head(&pp->node, &head->chain);
6095 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6096 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6098 if (pp && hlist_empty(&pp->owner)) {
6099 __hlist_del(&pp->node);
6100 kmem_cache_free(sctp_bucket_cachep, pp);
6101 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6105 /* Release this socket's reference to a local port. */
6106 static inline void __sctp_put_port(struct sock *sk)
6108 struct sctp_bind_hashbucket *head =
6109 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->inet_num)];
6110 struct sctp_bind_bucket *pp;
6112 sctp_spin_lock(&head->lock);
6113 pp = sctp_sk(sk)->bind_hash;
6114 __sk_del_bind_node(sk);
6115 sctp_sk(sk)->bind_hash = NULL;
6116 inet_sk(sk)->inet_num = 0;
6117 sctp_bucket_destroy(pp);
6118 sctp_spin_unlock(&head->lock);
6121 void sctp_put_port(struct sock *sk)
6123 sctp_local_bh_disable();
6124 __sctp_put_port(sk);
6125 sctp_local_bh_enable();
6129 * The system picks an ephemeral port and choose an address set equivalent
6130 * to binding with a wildcard address.
6131 * One of those addresses will be the primary address for the association.
6132 * This automatically enables the multihoming capability of SCTP.
6134 static int sctp_autobind(struct sock *sk)
6136 union sctp_addr autoaddr;
6140 /* Initialize a local sockaddr structure to INADDR_ANY. */
6141 af = sctp_sk(sk)->pf->af;
6143 port = htons(inet_sk(sk)->inet_num);
6144 af->inaddr_any(&autoaddr, port);
6146 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6149 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6152 * 4.2 The cmsghdr Structure *
6154 * When ancillary data is sent or received, any number of ancillary data
6155 * objects can be specified by the msg_control and msg_controllen members of
6156 * the msghdr structure, because each object is preceded by
6157 * a cmsghdr structure defining the object's length (the cmsg_len member).
6158 * Historically Berkeley-derived implementations have passed only one object
6159 * at a time, but this API allows multiple objects to be
6160 * passed in a single call to sendmsg() or recvmsg(). The following example
6161 * shows two ancillary data objects in a control buffer.
6163 * |<--------------------------- msg_controllen -------------------------->|
6166 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6168 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6171 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6173 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6176 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6177 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6179 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6181 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6188 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
6189 sctp_cmsgs_t *cmsgs)
6191 struct cmsghdr *cmsg;
6192 struct msghdr *my_msg = (struct msghdr *)msg;
6194 for (cmsg = CMSG_FIRSTHDR(msg);
6196 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6197 if (!CMSG_OK(my_msg, cmsg))
6200 /* Should we parse this header or ignore? */
6201 if (cmsg->cmsg_level != IPPROTO_SCTP)
6204 /* Strictly check lengths following example in SCM code. */
6205 switch (cmsg->cmsg_type) {
6207 /* SCTP Socket API Extension
6208 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6210 * This cmsghdr structure provides information for
6211 * initializing new SCTP associations with sendmsg().
6212 * The SCTP_INITMSG socket option uses this same data
6213 * structure. This structure is not used for
6216 * cmsg_level cmsg_type cmsg_data[]
6217 * ------------ ------------ ----------------------
6218 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6220 if (cmsg->cmsg_len !=
6221 CMSG_LEN(sizeof(struct sctp_initmsg)))
6223 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6227 /* SCTP Socket API Extension
6228 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6230 * This cmsghdr structure specifies SCTP options for
6231 * sendmsg() and describes SCTP header information
6232 * about a received message through recvmsg().
6234 * cmsg_level cmsg_type cmsg_data[]
6235 * ------------ ------------ ----------------------
6236 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6238 if (cmsg->cmsg_len !=
6239 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6243 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6245 /* Minimally, validate the sinfo_flags. */
6246 if (cmsgs->info->sinfo_flags &
6247 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6248 SCTP_SACK_IMMEDIATELY |
6249 SCTP_ABORT | SCTP_EOF))
6261 * Wait for a packet..
6262 * Note: This function is the same function as in core/datagram.c
6263 * with a few modifications to make lksctp work.
6265 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6270 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6272 /* Socket errors? */
6273 error = sock_error(sk);
6277 if (!skb_queue_empty(&sk->sk_receive_queue))
6280 /* Socket shut down? */
6281 if (sk->sk_shutdown & RCV_SHUTDOWN)
6284 /* Sequenced packets can come disconnected. If so we report the
6289 /* Is there a good reason to think that we may receive some data? */
6290 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6293 /* Handle signals. */
6294 if (signal_pending(current))
6297 /* Let another process have a go. Since we are going to sleep
6298 * anyway. Note: This may cause odd behaviors if the message
6299 * does not fit in the user's buffer, but this seems to be the
6300 * only way to honor MSG_DONTWAIT realistically.
6302 sctp_release_sock(sk);
6303 *timeo_p = schedule_timeout(*timeo_p);
6307 finish_wait(sk_sleep(sk), &wait);
6311 error = sock_intr_errno(*timeo_p);
6314 finish_wait(sk_sleep(sk), &wait);
6319 /* Receive a datagram.
6320 * Note: This is pretty much the same routine as in core/datagram.c
6321 * with a few changes to make lksctp work.
6323 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6324 int noblock, int *err)
6327 struct sk_buff *skb;
6330 timeo = sock_rcvtimeo(sk, noblock);
6332 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6333 timeo, MAX_SCHEDULE_TIMEOUT);
6336 /* Again only user level code calls this function,
6337 * so nothing interrupt level
6338 * will suddenly eat the receive_queue.
6340 * Look at current nfs client by the way...
6341 * However, this function was correct in any case. 8)
6343 if (flags & MSG_PEEK) {
6344 spin_lock_bh(&sk->sk_receive_queue.lock);
6345 skb = skb_peek(&sk->sk_receive_queue);
6347 atomic_inc(&skb->users);
6348 spin_unlock_bh(&sk->sk_receive_queue.lock);
6350 skb = skb_dequeue(&sk->sk_receive_queue);
6356 /* Caller is allowed not to check sk->sk_err before calling. */
6357 error = sock_error(sk);
6361 if (sk->sk_shutdown & RCV_SHUTDOWN)
6364 /* User doesn't want to wait. */
6368 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6377 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6378 static void __sctp_write_space(struct sctp_association *asoc)
6380 struct sock *sk = asoc->base.sk;
6381 struct socket *sock = sk->sk_socket;
6383 if ((sctp_wspace(asoc) > 0) && sock) {
6384 if (waitqueue_active(&asoc->wait))
6385 wake_up_interruptible(&asoc->wait);
6387 if (sctp_writeable(sk)) {
6388 wait_queue_head_t *wq = sk_sleep(sk);
6390 if (wq && waitqueue_active(wq))
6391 wake_up_interruptible(wq);
6393 /* Note that we try to include the Async I/O support
6394 * here by modeling from the current TCP/UDP code.
6395 * We have not tested with it yet.
6397 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6398 sock_wake_async(sock,
6399 SOCK_WAKE_SPACE, POLL_OUT);
6404 /* Do accounting for the sndbuf space.
6405 * Decrement the used sndbuf space of the corresponding association by the
6406 * data size which was just transmitted(freed).
6408 static void sctp_wfree(struct sk_buff *skb)
6410 struct sctp_association *asoc;
6411 struct sctp_chunk *chunk;
6414 /* Get the saved chunk pointer. */
6415 chunk = *((struct sctp_chunk **)(skb->cb));
6418 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6419 sizeof(struct sk_buff) +
6420 sizeof(struct sctp_chunk);
6422 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6425 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6427 sk->sk_wmem_queued -= skb->truesize;
6428 sk_mem_uncharge(sk, skb->truesize);
6431 __sctp_write_space(asoc);
6433 sctp_association_put(asoc);
6436 /* Do accounting for the receive space on the socket.
6437 * Accounting for the association is done in ulpevent.c
6438 * We set this as a destructor for the cloned data skbs so that
6439 * accounting is done at the correct time.
6441 void sctp_sock_rfree(struct sk_buff *skb)
6443 struct sock *sk = skb->sk;
6444 struct sctp_ulpevent *event = sctp_skb2event(skb);
6446 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6449 * Mimic the behavior of sock_rfree
6451 sk_mem_uncharge(sk, event->rmem_len);
6455 /* Helper function to wait for space in the sndbuf. */
6456 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6459 struct sock *sk = asoc->base.sk;
6461 long current_timeo = *timeo_p;
6464 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6465 asoc, (long)(*timeo_p), msg_len);
6467 /* Increment the association's refcnt. */
6468 sctp_association_hold(asoc);
6470 /* Wait on the association specific sndbuf space. */
6472 prepare_to_wait_exclusive(&asoc->wait, &wait,
6473 TASK_INTERRUPTIBLE);
6476 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6479 if (signal_pending(current))
6480 goto do_interrupted;
6481 if (msg_len <= sctp_wspace(asoc))
6484 /* Let another process have a go. Since we are going
6487 sctp_release_sock(sk);
6488 current_timeo = schedule_timeout(current_timeo);
6489 BUG_ON(sk != asoc->base.sk);
6492 *timeo_p = current_timeo;
6496 finish_wait(&asoc->wait, &wait);
6498 /* Release the association's refcnt. */
6499 sctp_association_put(asoc);
6508 err = sock_intr_errno(*timeo_p);
6516 void sctp_data_ready(struct sock *sk, int len)
6518 struct socket_wq *wq;
6521 wq = rcu_dereference(sk->sk_wq);
6522 if (wq_has_sleeper(wq))
6523 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6524 POLLRDNORM | POLLRDBAND);
6525 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6529 /* If socket sndbuf has changed, wake up all per association waiters. */
6530 void sctp_write_space(struct sock *sk)
6532 struct sctp_association *asoc;
6534 /* Wake up the tasks in each wait queue. */
6535 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6536 __sctp_write_space(asoc);
6540 /* Is there any sndbuf space available on the socket?
6542 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6543 * associations on the same socket. For a UDP-style socket with
6544 * multiple associations, it is possible for it to be "unwriteable"
6545 * prematurely. I assume that this is acceptable because
6546 * a premature "unwriteable" is better than an accidental "writeable" which
6547 * would cause an unwanted block under certain circumstances. For the 1-1
6548 * UDP-style sockets or TCP-style sockets, this code should work.
6551 static int sctp_writeable(struct sock *sk)
6555 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6561 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6562 * returns immediately with EINPROGRESS.
6564 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6566 struct sock *sk = asoc->base.sk;
6568 long current_timeo = *timeo_p;
6571 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6574 /* Increment the association's refcnt. */
6575 sctp_association_hold(asoc);
6578 prepare_to_wait_exclusive(&asoc->wait, &wait,
6579 TASK_INTERRUPTIBLE);
6582 if (sk->sk_shutdown & RCV_SHUTDOWN)
6584 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6587 if (signal_pending(current))
6588 goto do_interrupted;
6590 if (sctp_state(asoc, ESTABLISHED))
6593 /* Let another process have a go. Since we are going
6596 sctp_release_sock(sk);
6597 current_timeo = schedule_timeout(current_timeo);
6600 *timeo_p = current_timeo;
6604 finish_wait(&asoc->wait, &wait);
6606 /* Release the association's refcnt. */
6607 sctp_association_put(asoc);
6612 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6615 err = -ECONNREFUSED;
6619 err = sock_intr_errno(*timeo_p);
6627 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6629 struct sctp_endpoint *ep;
6633 ep = sctp_sk(sk)->ep;
6637 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6638 TASK_INTERRUPTIBLE);
6640 if (list_empty(&ep->asocs)) {
6641 sctp_release_sock(sk);
6642 timeo = schedule_timeout(timeo);
6647 if (!sctp_sstate(sk, LISTENING))
6651 if (!list_empty(&ep->asocs))
6654 err = sock_intr_errno(timeo);
6655 if (signal_pending(current))
6663 finish_wait(sk_sleep(sk), &wait);
6668 static void sctp_wait_for_close(struct sock *sk, long timeout)
6673 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6674 if (list_empty(&sctp_sk(sk)->ep->asocs))
6676 sctp_release_sock(sk);
6677 timeout = schedule_timeout(timeout);
6679 } while (!signal_pending(current) && timeout);
6681 finish_wait(sk_sleep(sk), &wait);
6684 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6686 struct sk_buff *frag;
6691 /* Don't forget the fragments. */
6692 skb_walk_frags(skb, frag)
6693 sctp_skb_set_owner_r_frag(frag, sk);
6696 sctp_skb_set_owner_r(skb, sk);
6699 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6700 struct sctp_association *asoc)
6702 struct inet_sock *inet = inet_sk(sk);
6703 struct inet_sock *newinet;
6705 newsk->sk_type = sk->sk_type;
6706 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6707 newsk->sk_flags = sk->sk_flags;
6708 newsk->sk_no_check = sk->sk_no_check;
6709 newsk->sk_reuse = sk->sk_reuse;
6711 newsk->sk_shutdown = sk->sk_shutdown;
6712 newsk->sk_destruct = inet_sock_destruct;
6713 newsk->sk_family = sk->sk_family;
6714 newsk->sk_protocol = IPPROTO_SCTP;
6715 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6716 newsk->sk_sndbuf = sk->sk_sndbuf;
6717 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6718 newsk->sk_lingertime = sk->sk_lingertime;
6719 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6720 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6722 newinet = inet_sk(newsk);
6724 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6725 * getsockname() and getpeername()
6727 newinet->inet_sport = inet->inet_sport;
6728 newinet->inet_saddr = inet->inet_saddr;
6729 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6730 newinet->inet_dport = htons(asoc->peer.port);
6731 newinet->pmtudisc = inet->pmtudisc;
6732 newinet->inet_id = asoc->next_tsn ^ jiffies;
6734 newinet->uc_ttl = inet->uc_ttl;
6735 newinet->mc_loop = 1;
6736 newinet->mc_ttl = 1;
6737 newinet->mc_index = 0;
6738 newinet->mc_list = NULL;
6740 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
6741 net_enable_timestamp();
6744 static inline void sctp_copy_descendant(struct sock *sk_to,
6745 const struct sock *sk_from)
6747 int ancestor_size = sizeof(struct inet_sock) +
6748 sizeof(struct sctp_sock) -
6749 offsetof(struct sctp_sock, auto_asconf_list);
6751 if (sk_from->sk_family == PF_INET6)
6752 ancestor_size += sizeof(struct ipv6_pinfo);
6754 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
6757 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6758 * and its messages to the newsk.
6760 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6761 struct sctp_association *assoc,
6762 sctp_socket_type_t type)
6764 struct sctp_sock *oldsp = sctp_sk(oldsk);
6765 struct sctp_sock *newsp = sctp_sk(newsk);
6766 struct sctp_bind_bucket *pp; /* hash list port iterator */
6767 struct sctp_endpoint *newep = newsp->ep;
6768 struct sk_buff *skb, *tmp;
6769 struct sctp_ulpevent *event;
6770 struct sctp_bind_hashbucket *head;
6772 /* Migrate socket buffer sizes and all the socket level options to the
6775 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6776 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6777 /* Brute force copy old sctp opt. */
6778 sctp_copy_descendant(newsk, oldsk);
6780 /* Restore the ep value that was overwritten with the above structure
6786 /* Hook this new socket in to the bind_hash list. */
6787 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->inet_num)];
6788 sctp_local_bh_disable();
6789 sctp_spin_lock(&head->lock);
6790 pp = sctp_sk(oldsk)->bind_hash;
6791 sk_add_bind_node(newsk, &pp->owner);
6792 sctp_sk(newsk)->bind_hash = pp;
6793 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6794 sctp_spin_unlock(&head->lock);
6795 sctp_local_bh_enable();
6797 /* Copy the bind_addr list from the original endpoint to the new
6798 * endpoint so that we can handle restarts properly
6800 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6801 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6803 /* Move any messages in the old socket's receive queue that are for the
6804 * peeled off association to the new socket's receive queue.
6806 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6807 event = sctp_skb2event(skb);
6808 if (event->asoc == assoc) {
6809 __skb_unlink(skb, &oldsk->sk_receive_queue);
6810 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6811 sctp_skb_set_owner_r_frag(skb, newsk);
6815 /* Clean up any messages pending delivery due to partial
6816 * delivery. Three cases:
6817 * 1) No partial deliver; no work.
6818 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6819 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6821 skb_queue_head_init(&newsp->pd_lobby);
6822 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6824 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6825 struct sk_buff_head *queue;
6827 /* Decide which queue to move pd_lobby skbs to. */
6828 if (assoc->ulpq.pd_mode) {
6829 queue = &newsp->pd_lobby;
6831 queue = &newsk->sk_receive_queue;
6833 /* Walk through the pd_lobby, looking for skbs that
6834 * need moved to the new socket.
6836 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6837 event = sctp_skb2event(skb);
6838 if (event->asoc == assoc) {
6839 __skb_unlink(skb, &oldsp->pd_lobby);
6840 __skb_queue_tail(queue, skb);
6841 sctp_skb_set_owner_r_frag(skb, newsk);
6845 /* Clear up any skbs waiting for the partial
6846 * delivery to finish.
6848 if (assoc->ulpq.pd_mode)
6849 sctp_clear_pd(oldsk, NULL);
6853 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6854 sctp_skb_set_owner_r_frag(skb, newsk);
6856 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6857 sctp_skb_set_owner_r_frag(skb, newsk);
6859 /* Set the type of socket to indicate that it is peeled off from the
6860 * original UDP-style socket or created with the accept() call on a
6861 * TCP-style socket..
6865 /* Mark the new socket "in-use" by the user so that any packets
6866 * that may arrive on the association after we've moved it are
6867 * queued to the backlog. This prevents a potential race between
6868 * backlog processing on the old socket and new-packet processing
6869 * on the new socket.
6871 * The caller has just allocated newsk so we can guarantee that other
6872 * paths won't try to lock it and then oldsk.
6874 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6875 sctp_assoc_migrate(assoc, newsk);
6877 /* If the association on the newsk is already closed before accept()
6878 * is called, set RCV_SHUTDOWN flag.
6880 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6881 newsk->sk_shutdown |= RCV_SHUTDOWN;
6883 newsk->sk_state = SCTP_SS_ESTABLISHED;
6884 sctp_release_sock(newsk);
6888 /* This proto struct describes the ULP interface for SCTP. */
6889 struct proto sctp_prot = {
6891 .owner = THIS_MODULE,
6892 .close = sctp_close,
6893 .connect = sctp_connect,
6894 .disconnect = sctp_disconnect,
6895 .accept = sctp_accept,
6896 .ioctl = sctp_ioctl,
6897 .init = sctp_init_sock,
6898 .destroy = sctp_destroy_sock,
6899 .shutdown = sctp_shutdown,
6900 .setsockopt = sctp_setsockopt,
6901 .getsockopt = sctp_getsockopt,
6902 .sendmsg = sctp_sendmsg,
6903 .recvmsg = sctp_recvmsg,
6905 .backlog_rcv = sctp_backlog_rcv,
6907 .unhash = sctp_unhash,
6908 .get_port = sctp_get_port,
6909 .obj_size = sizeof(struct sctp_sock),
6910 .sysctl_mem = sysctl_sctp_mem,
6911 .sysctl_rmem = sysctl_sctp_rmem,
6912 .sysctl_wmem = sysctl_sctp_wmem,
6913 .memory_pressure = &sctp_memory_pressure,
6914 .enter_memory_pressure = sctp_enter_memory_pressure,
6915 .memory_allocated = &sctp_memory_allocated,
6916 .sockets_allocated = &sctp_sockets_allocated,
6919 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6921 #include <net/transp_v6.h>
6922 static void sctp_v6_destroy_sock(struct sock *sk)
6924 sctp_destroy_sock(sk);
6925 inet6_destroy_sock(sk);
6928 struct proto sctpv6_prot = {
6930 .owner = THIS_MODULE,
6931 .close = sctp_close,
6932 .connect = sctp_connect,
6933 .disconnect = sctp_disconnect,
6934 .accept = sctp_accept,
6935 .ioctl = sctp_ioctl,
6936 .init = sctp_init_sock,
6937 .destroy = sctp_v6_destroy_sock,
6938 .shutdown = sctp_shutdown,
6939 .setsockopt = sctp_setsockopt,
6940 .getsockopt = sctp_getsockopt,
6941 .sendmsg = sctp_sendmsg,
6942 .recvmsg = sctp_recvmsg,
6944 .backlog_rcv = sctp_backlog_rcv,
6946 .unhash = sctp_unhash,
6947 .get_port = sctp_get_port,
6948 .obj_size = sizeof(struct sctp6_sock),
6949 .sysctl_mem = sysctl_sctp_mem,
6950 .sysctl_rmem = sysctl_sctp_rmem,
6951 .sysctl_wmem = sysctl_sctp_wmem,
6952 .memory_pressure = &sctp_memory_pressure,
6953 .enter_memory_pressure = sctp_enter_memory_pressure,
6954 .memory_allocated = &sctp_memory_allocated,
6955 .sockets_allocated = &sctp_sockets_allocated,
6957 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */