1 /* SCTP kernel implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
9 * This file is part of the SCTP kernel implementation
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
18 * This SCTP implementation is free software;
19 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
24 * This SCTP implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, write to
32 * the Free Software Foundation, 59 Temple Place - Suite 330,
33 * Boston, MA 02111-1307, USA.
35 * Please send any bug reports or fixes you make to the
37 * lksctp developers <lksctp-developers@lists.sourceforge.net>
39 * Or submit a bug report through the following website:
40 * http://www.sf.net/projects/lksctp
42 * Written or modified by:
43 * La Monte H.P. Yarroll <piggy@acm.org>
44 * Narasimha Budihal <narsi@refcode.org>
45 * Karl Knutson <karl@athena.chicago.il.us>
46 * Jon Grimm <jgrimm@us.ibm.com>
47 * Xingang Guo <xingang.guo@intel.com>
48 * Daisy Chang <daisyc@us.ibm.com>
49 * Sridhar Samudrala <samudrala@us.ibm.com>
50 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
51 * Ardelle Fan <ardelle.fan@intel.com>
52 * Ryan Layer <rmlayer@us.ibm.com>
53 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
54 * Kevin Gao <kevin.gao@intel.com>
56 * Any bugs reported given to us we will try to fix... any fixes shared will
57 * be incorporated into the next SCTP release.
60 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
62 #include <linux/types.h>
63 #include <linux/kernel.h>
64 #include <linux/wait.h>
65 #include <linux/time.h>
67 #include <linux/capability.h>
68 #include <linux/fcntl.h>
69 #include <linux/poll.h>
70 #include <linux/init.h>
71 #include <linux/crypto.h>
72 #include <linux/slab.h>
73 #include <linux/file.h>
77 #include <net/route.h>
79 #include <net/inet_common.h>
81 #include <linux/socket.h> /* for sa_family_t */
82 #include <linux/export.h>
84 #include <net/sctp/sctp.h>
85 #include <net/sctp/sm.h>
87 /* WARNING: Please do not remove the SCTP_STATIC attribute to
88 * any of the functions below as they are used to export functions
89 * used by a project regression testsuite.
92 /* Forward declarations for internal helper functions. */
93 static int sctp_writeable(struct sock *sk);
94 static void sctp_wfree(struct sk_buff *skb);
95 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
97 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
98 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
99 static int sctp_wait_for_accept(struct sock *sk, long timeo);
100 static void sctp_wait_for_close(struct sock *sk, long timeo);
101 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
102 union sctp_addr *addr, int len);
103 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
104 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
105 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
106 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
107 static int sctp_send_asconf(struct sctp_association *asoc,
108 struct sctp_chunk *chunk);
109 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
110 static int sctp_autobind(struct sock *sk);
111 static void sctp_sock_migrate(struct sock *, struct sock *,
112 struct sctp_association *, sctp_socket_type_t);
114 extern struct kmem_cache *sctp_bucket_cachep;
115 extern long sysctl_sctp_mem[3];
116 extern int sysctl_sctp_rmem[3];
117 extern int sysctl_sctp_wmem[3];
119 static int sctp_memory_pressure;
120 static atomic_long_t sctp_memory_allocated;
121 struct percpu_counter sctp_sockets_allocated;
123 static void sctp_enter_memory_pressure(struct sock *sk)
125 sctp_memory_pressure = 1;
129 /* Get the sndbuf space available at the time on the association. */
130 static inline int sctp_wspace(struct sctp_association *asoc)
134 if (asoc->ep->sndbuf_policy)
135 amt = asoc->sndbuf_used;
137 amt = sk_wmem_alloc_get(asoc->base.sk);
139 if (amt >= asoc->base.sk->sk_sndbuf) {
140 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
143 amt = sk_stream_wspace(asoc->base.sk);
148 amt = asoc->base.sk->sk_sndbuf - amt;
153 /* Increment the used sndbuf space count of the corresponding association by
154 * the size of the outgoing data chunk.
155 * Also, set the skb destructor for sndbuf accounting later.
157 * Since it is always 1-1 between chunk and skb, and also a new skb is always
158 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
159 * destructor in the data chunk skb for the purpose of the sndbuf space
162 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
164 struct sctp_association *asoc = chunk->asoc;
165 struct sock *sk = asoc->base.sk;
167 /* The sndbuf space is tracked per association. */
168 sctp_association_hold(asoc);
170 skb_set_owner_w(chunk->skb, sk);
172 chunk->skb->destructor = sctp_wfree;
173 /* Save the chunk pointer in skb for sctp_wfree to use later. */
174 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
176 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
177 sizeof(struct sk_buff) +
178 sizeof(struct sctp_chunk);
180 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
181 sk->sk_wmem_queued += chunk->skb->truesize;
182 sk_mem_charge(sk, chunk->skb->truesize);
185 /* Verify that this is a valid address. */
186 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
191 /* Verify basic sockaddr. */
192 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
196 /* Is this a valid SCTP address? */
197 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
200 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
206 /* Look up the association by its id. If this is not a UDP-style
207 * socket, the ID field is always ignored.
209 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
211 struct sctp_association *asoc = NULL;
213 /* If this is not a UDP-style socket, assoc id should be ignored. */
214 if (!sctp_style(sk, UDP)) {
215 /* Return NULL if the socket state is not ESTABLISHED. It
216 * could be a TCP-style listening socket or a socket which
217 * hasn't yet called connect() to establish an association.
219 if (!sctp_sstate(sk, ESTABLISHED))
222 /* Get the first and the only association from the list. */
223 if (!list_empty(&sctp_sk(sk)->ep->asocs))
224 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
225 struct sctp_association, asocs);
229 /* Otherwise this is a UDP-style socket. */
230 if (!id || (id == (sctp_assoc_t)-1))
233 spin_lock_bh(&sctp_assocs_id_lock);
234 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
235 spin_unlock_bh(&sctp_assocs_id_lock);
237 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
243 /* Look up the transport from an address and an assoc id. If both address and
244 * id are specified, the associations matching the address and the id should be
247 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
248 struct sockaddr_storage *addr,
251 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
252 struct sctp_transport *transport;
253 union sctp_addr *laddr = (union sctp_addr *)addr;
255 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
262 id_asoc = sctp_id2assoc(sk, id);
263 if (id_asoc && (id_asoc != addr_asoc))
266 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
267 (union sctp_addr *)addr);
272 /* API 3.1.2 bind() - UDP Style Syntax
273 * The syntax of bind() is,
275 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
277 * sd - the socket descriptor returned by socket().
278 * addr - the address structure (struct sockaddr_in or struct
279 * sockaddr_in6 [RFC 2553]),
280 * addr_len - the size of the address structure.
282 SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
288 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
291 /* Disallow binding twice. */
292 if (!sctp_sk(sk)->ep->base.bind_addr.port)
293 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
298 sctp_release_sock(sk);
303 static long sctp_get_port_local(struct sock *, union sctp_addr *);
305 /* Verify this is a valid sockaddr. */
306 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
307 union sctp_addr *addr, int len)
311 /* Check minimum size. */
312 if (len < sizeof (struct sockaddr))
315 /* V4 mapped address are really of AF_INET family */
316 if (addr->sa.sa_family == AF_INET6 &&
317 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
318 if (!opt->pf->af_supported(AF_INET, opt))
321 /* Does this PF support this AF? */
322 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
326 /* If we get this far, af is valid. */
327 af = sctp_get_af_specific(addr->sa.sa_family);
329 if (len < af->sockaddr_len)
335 /* Bind a local address either to an endpoint or to an association. */
336 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
338 struct net *net = sock_net(sk);
339 struct sctp_sock *sp = sctp_sk(sk);
340 struct sctp_endpoint *ep = sp->ep;
341 struct sctp_bind_addr *bp = &ep->base.bind_addr;
346 /* Common sockaddr verification. */
347 af = sctp_sockaddr_af(sp, addr, len);
349 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
354 snum = ntohs(addr->v4.sin_port);
356 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
357 ", port: %d, new port: %d, len: %d)\n",
363 /* PF specific bind() address verification. */
364 if (!sp->pf->bind_verify(sp, addr))
365 return -EADDRNOTAVAIL;
367 /* We must either be unbound, or bind to the same port.
368 * It's OK to allow 0 ports if we are already bound.
369 * We'll just inhert an already bound port in this case
374 else if (snum != bp->port) {
375 SCTP_DEBUG_PRINTK("sctp_do_bind:"
376 " New port %d does not match existing port "
377 "%d.\n", snum, bp->port);
382 if (snum && snum < PROT_SOCK &&
383 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
386 /* See if the address matches any of the addresses we may have
387 * already bound before checking against other endpoints.
389 if (sctp_bind_addr_match(bp, addr, sp))
392 /* Make sure we are allowed to bind here.
393 * The function sctp_get_port_local() does duplicate address
396 addr->v4.sin_port = htons(snum);
397 if ((ret = sctp_get_port_local(sk, addr))) {
401 /* Refresh ephemeral port. */
403 bp->port = inet_sk(sk)->inet_num;
405 /* Add the address to the bind address list.
406 * Use GFP_ATOMIC since BHs will be disabled.
408 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
410 /* Copy back into socket for getsockname() use. */
412 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
413 af->to_sk_saddr(addr, sk);
419 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
421 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
422 * at any one time. If a sender, after sending an ASCONF chunk, decides
423 * it needs to transfer another ASCONF Chunk, it MUST wait until the
424 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
425 * subsequent ASCONF. Note this restriction binds each side, so at any
426 * time two ASCONF may be in-transit on any given association (one sent
427 * from each endpoint).
429 static int sctp_send_asconf(struct sctp_association *asoc,
430 struct sctp_chunk *chunk)
432 struct net *net = sock_net(asoc->base.sk);
435 /* If there is an outstanding ASCONF chunk, queue it for later
438 if (asoc->addip_last_asconf) {
439 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
443 /* Hold the chunk until an ASCONF_ACK is received. */
444 sctp_chunk_hold(chunk);
445 retval = sctp_primitive_ASCONF(net, asoc, chunk);
447 sctp_chunk_free(chunk);
449 asoc->addip_last_asconf = chunk;
455 /* Add a list of addresses as bind addresses to local endpoint or
458 * Basically run through each address specified in the addrs/addrcnt
459 * array/length pair, determine if it is IPv6 or IPv4 and call
460 * sctp_do_bind() on it.
462 * If any of them fails, then the operation will be reversed and the
463 * ones that were added will be removed.
465 * Only sctp_setsockopt_bindx() is supposed to call this function.
467 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
472 struct sockaddr *sa_addr;
475 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
479 for (cnt = 0; cnt < addrcnt; cnt++) {
480 /* The list may contain either IPv4 or IPv6 address;
481 * determine the address length for walking thru the list.
484 af = sctp_get_af_specific(sa_addr->sa_family);
490 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
493 addr_buf += af->sockaddr_len;
497 /* Failed. Cleanup the ones that have been added */
499 sctp_bindx_rem(sk, addrs, cnt);
507 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
508 * associations that are part of the endpoint indicating that a list of local
509 * addresses are added to the endpoint.
511 * If any of the addresses is already in the bind address list of the
512 * association, we do not send the chunk for that association. But it will not
513 * affect other associations.
515 * Only sctp_setsockopt_bindx() is supposed to call this function.
517 static int sctp_send_asconf_add_ip(struct sock *sk,
518 struct sockaddr *addrs,
521 struct net *net = sock_net(sk);
522 struct sctp_sock *sp;
523 struct sctp_endpoint *ep;
524 struct sctp_association *asoc;
525 struct sctp_bind_addr *bp;
526 struct sctp_chunk *chunk;
527 struct sctp_sockaddr_entry *laddr;
528 union sctp_addr *addr;
529 union sctp_addr saveaddr;
536 if (!net->sctp.addip_enable)
542 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
543 __func__, sk, addrs, addrcnt);
545 list_for_each_entry(asoc, &ep->asocs, asocs) {
547 if (!asoc->peer.asconf_capable)
550 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
553 if (!sctp_state(asoc, ESTABLISHED))
556 /* Check if any address in the packed array of addresses is
557 * in the bind address list of the association. If so,
558 * do not send the asconf chunk to its peer, but continue with
559 * other associations.
562 for (i = 0; i < addrcnt; i++) {
564 af = sctp_get_af_specific(addr->v4.sin_family);
570 if (sctp_assoc_lookup_laddr(asoc, addr))
573 addr_buf += af->sockaddr_len;
578 /* Use the first valid address in bind addr list of
579 * association as Address Parameter of ASCONF CHUNK.
581 bp = &asoc->base.bind_addr;
582 p = bp->address_list.next;
583 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
584 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
585 addrcnt, SCTP_PARAM_ADD_IP);
591 /* Add the new addresses to the bind address list with
592 * use_as_src set to 0.
595 for (i = 0; i < addrcnt; i++) {
597 af = sctp_get_af_specific(addr->v4.sin_family);
598 memcpy(&saveaddr, addr, af->sockaddr_len);
599 retval = sctp_add_bind_addr(bp, &saveaddr,
600 SCTP_ADDR_NEW, GFP_ATOMIC);
601 addr_buf += af->sockaddr_len;
603 if (asoc->src_out_of_asoc_ok) {
604 struct sctp_transport *trans;
606 list_for_each_entry(trans,
607 &asoc->peer.transport_addr_list, transports) {
608 /* Clear the source and route cache */
609 dst_release(trans->dst);
610 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
611 2*asoc->pathmtu, 4380));
612 trans->ssthresh = asoc->peer.i.a_rwnd;
613 trans->rto = asoc->rto_initial;
614 sctp_max_rto(asoc, trans);
615 trans->rtt = trans->srtt = trans->rttvar = 0;
616 sctp_transport_route(trans, NULL,
617 sctp_sk(asoc->base.sk));
620 retval = sctp_send_asconf(asoc, chunk);
627 /* Remove a list of addresses from bind addresses list. Do not remove the
630 * Basically run through each address specified in the addrs/addrcnt
631 * array/length pair, determine if it is IPv6 or IPv4 and call
632 * sctp_del_bind() on it.
634 * If any of them fails, then the operation will be reversed and the
635 * ones that were removed will be added back.
637 * At least one address has to be left; if only one address is
638 * available, the operation will return -EBUSY.
640 * Only sctp_setsockopt_bindx() is supposed to call this function.
642 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
644 struct sctp_sock *sp = sctp_sk(sk);
645 struct sctp_endpoint *ep = sp->ep;
647 struct sctp_bind_addr *bp = &ep->base.bind_addr;
650 union sctp_addr *sa_addr;
653 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
657 for (cnt = 0; cnt < addrcnt; cnt++) {
658 /* If the bind address list is empty or if there is only one
659 * bind address, there is nothing more to be removed (we need
660 * at least one address here).
662 if (list_empty(&bp->address_list) ||
663 (sctp_list_single_entry(&bp->address_list))) {
669 af = sctp_get_af_specific(sa_addr->sa.sa_family);
675 if (!af->addr_valid(sa_addr, sp, NULL)) {
676 retval = -EADDRNOTAVAIL;
680 if (sa_addr->v4.sin_port &&
681 sa_addr->v4.sin_port != htons(bp->port)) {
686 if (!sa_addr->v4.sin_port)
687 sa_addr->v4.sin_port = htons(bp->port);
689 /* FIXME - There is probably a need to check if sk->sk_saddr and
690 * sk->sk_rcv_addr are currently set to one of the addresses to
691 * be removed. This is something which needs to be looked into
692 * when we are fixing the outstanding issues with multi-homing
693 * socket routing and failover schemes. Refer to comments in
694 * sctp_do_bind(). -daisy
696 retval = sctp_del_bind_addr(bp, sa_addr);
698 addr_buf += af->sockaddr_len;
701 /* Failed. Add the ones that has been removed back */
703 sctp_bindx_add(sk, addrs, cnt);
711 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
712 * the associations that are part of the endpoint indicating that a list of
713 * local addresses are removed from the endpoint.
715 * If any of the addresses is already in the bind address list of the
716 * association, we do not send the chunk for that association. But it will not
717 * affect other associations.
719 * Only sctp_setsockopt_bindx() is supposed to call this function.
721 static int sctp_send_asconf_del_ip(struct sock *sk,
722 struct sockaddr *addrs,
725 struct net *net = sock_net(sk);
726 struct sctp_sock *sp;
727 struct sctp_endpoint *ep;
728 struct sctp_association *asoc;
729 struct sctp_transport *transport;
730 struct sctp_bind_addr *bp;
731 struct sctp_chunk *chunk;
732 union sctp_addr *laddr;
735 struct sctp_sockaddr_entry *saddr;
741 if (!net->sctp.addip_enable)
747 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
748 __func__, sk, addrs, addrcnt);
750 list_for_each_entry(asoc, &ep->asocs, asocs) {
752 if (!asoc->peer.asconf_capable)
755 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
758 if (!sctp_state(asoc, ESTABLISHED))
761 /* Check if any address in the packed array of addresses is
762 * not present in the bind address list of the association.
763 * If so, do not send the asconf chunk to its peer, but
764 * continue with other associations.
767 for (i = 0; i < addrcnt; i++) {
769 af = sctp_get_af_specific(laddr->v4.sin_family);
775 if (!sctp_assoc_lookup_laddr(asoc, laddr))
778 addr_buf += af->sockaddr_len;
783 /* Find one address in the association's bind address list
784 * that is not in the packed array of addresses. This is to
785 * make sure that we do not delete all the addresses in the
788 bp = &asoc->base.bind_addr;
789 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
791 if ((laddr == NULL) && (addrcnt == 1)) {
792 if (asoc->asconf_addr_del_pending)
794 asoc->asconf_addr_del_pending =
795 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
796 if (asoc->asconf_addr_del_pending == NULL) {
800 asoc->asconf_addr_del_pending->sa.sa_family =
802 asoc->asconf_addr_del_pending->v4.sin_port =
804 if (addrs->sa_family == AF_INET) {
805 struct sockaddr_in *sin;
807 sin = (struct sockaddr_in *)addrs;
808 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
809 } else if (addrs->sa_family == AF_INET6) {
810 struct sockaddr_in6 *sin6;
812 sin6 = (struct sockaddr_in6 *)addrs;
813 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
815 SCTP_DEBUG_PRINTK_IPADDR("send_asconf_del_ip: keep the last address asoc: %p ",
816 " at %p\n", asoc, asoc->asconf_addr_del_pending,
817 asoc->asconf_addr_del_pending);
818 asoc->src_out_of_asoc_ok = 1;
826 /* We do not need RCU protection throughout this loop
827 * because this is done under a socket lock from the
830 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
838 /* Reset use_as_src flag for the addresses in the bind address
839 * list that are to be deleted.
842 for (i = 0; i < addrcnt; i++) {
844 af = sctp_get_af_specific(laddr->v4.sin_family);
845 list_for_each_entry(saddr, &bp->address_list, list) {
846 if (sctp_cmp_addr_exact(&saddr->a, laddr))
847 saddr->state = SCTP_ADDR_DEL;
849 addr_buf += af->sockaddr_len;
852 /* Update the route and saddr entries for all the transports
853 * as some of the addresses in the bind address list are
854 * about to be deleted and cannot be used as source addresses.
856 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
858 dst_release(transport->dst);
859 sctp_transport_route(transport, NULL,
860 sctp_sk(asoc->base.sk));
864 /* We don't need to transmit ASCONF */
866 retval = sctp_send_asconf(asoc, chunk);
872 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
873 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
875 struct sock *sk = sctp_opt2sk(sp);
876 union sctp_addr *addr;
879 /* It is safe to write port space in caller. */
881 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
882 af = sctp_get_af_specific(addr->sa.sa_family);
885 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
888 if (addrw->state == SCTP_ADDR_NEW)
889 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
891 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
894 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
897 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
900 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
901 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
904 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
905 * Section 3.1.2 for this usage.
907 * addrs is a pointer to an array of one or more socket addresses. Each
908 * address is contained in its appropriate structure (i.e. struct
909 * sockaddr_in or struct sockaddr_in6) the family of the address type
910 * must be used to distinguish the address length (note that this
911 * representation is termed a "packed array" of addresses). The caller
912 * specifies the number of addresses in the array with addrcnt.
914 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
915 * -1, and sets errno to the appropriate error code.
917 * For SCTP, the port given in each socket address must be the same, or
918 * sctp_bindx() will fail, setting errno to EINVAL.
920 * The flags parameter is formed from the bitwise OR of zero or more of
921 * the following currently defined flags:
923 * SCTP_BINDX_ADD_ADDR
925 * SCTP_BINDX_REM_ADDR
927 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
928 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
929 * addresses from the association. The two flags are mutually exclusive;
930 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
931 * not remove all addresses from an association; sctp_bindx() will
932 * reject such an attempt with EINVAL.
934 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
935 * additional addresses with an endpoint after calling bind(). Or use
936 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
937 * socket is associated with so that no new association accepted will be
938 * associated with those addresses. If the endpoint supports dynamic
939 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
940 * endpoint to send the appropriate message to the peer to change the
941 * peers address lists.
943 * Adding and removing addresses from a connected association is
944 * optional functionality. Implementations that do not support this
945 * functionality should return EOPNOTSUPP.
947 * Basically do nothing but copying the addresses from user to kernel
948 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
949 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
952 * We don't use copy_from_user() for optimization: we first do the
953 * sanity checks (buffer size -fast- and access check-healthy
954 * pointer); if all of those succeed, then we can alloc the memory
955 * (expensive operation) needed to copy the data to kernel. Then we do
956 * the copying without checking the user space area
957 * (__copy_from_user()).
959 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
962 * sk The sk of the socket
963 * addrs The pointer to the addresses in user land
964 * addrssize Size of the addrs buffer
965 * op Operation to perform (add or remove, see the flags of
968 * Returns 0 if ok, <0 errno code on error.
970 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
971 struct sockaddr __user *addrs,
972 int addrs_size, int op)
974 struct sockaddr *kaddrs;
978 struct sockaddr *sa_addr;
982 SCTP_DEBUG_PRINTK("sctp_setsockopt_bindx: sk %p addrs %p"
983 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
985 if (unlikely(addrs_size <= 0))
988 /* Check the user passed a healthy pointer. */
989 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
992 /* Alloc space for the address array in kernel memory. */
993 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
994 if (unlikely(!kaddrs))
997 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1002 /* Walk through the addrs buffer and count the number of addresses. */
1004 while (walk_size < addrs_size) {
1005 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1011 af = sctp_get_af_specific(sa_addr->sa_family);
1013 /* If the address family is not supported or if this address
1014 * causes the address buffer to overflow return EINVAL.
1016 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1021 addr_buf += af->sockaddr_len;
1022 walk_size += af->sockaddr_len;
1027 case SCTP_BINDX_ADD_ADDR:
1028 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1031 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1034 case SCTP_BINDX_REM_ADDR:
1035 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1038 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1052 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1054 * Common routine for handling connect() and sctp_connectx().
1055 * Connect will come in with just a single address.
1057 static int __sctp_connect(struct sock* sk,
1058 struct sockaddr *kaddrs,
1060 sctp_assoc_t *assoc_id)
1062 struct net *net = sock_net(sk);
1063 struct sctp_sock *sp;
1064 struct sctp_endpoint *ep;
1065 struct sctp_association *asoc = NULL;
1066 struct sctp_association *asoc2;
1067 struct sctp_transport *transport;
1075 union sctp_addr *sa_addr = NULL;
1077 unsigned short port;
1078 unsigned int f_flags = 0;
1083 /* connect() cannot be done on a socket that is already in ESTABLISHED
1084 * state - UDP-style peeled off socket or a TCP-style socket that
1085 * is already connected.
1086 * It cannot be done even on a TCP-style listening socket.
1088 if (sctp_sstate(sk, ESTABLISHED) ||
1089 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1094 /* Walk through the addrs buffer and count the number of addresses. */
1096 while (walk_size < addrs_size) {
1097 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1103 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1105 /* If the address family is not supported or if this address
1106 * causes the address buffer to overflow return EINVAL.
1108 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1113 port = ntohs(sa_addr->v4.sin_port);
1115 /* Save current address so we can work with it */
1116 memcpy(&to, sa_addr, af->sockaddr_len);
1118 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1122 /* Make sure the destination port is correctly set
1125 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1130 /* Check if there already is a matching association on the
1131 * endpoint (other than the one created here).
1133 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1134 if (asoc2 && asoc2 != asoc) {
1135 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1142 /* If we could not find a matching association on the endpoint,
1143 * make sure that there is no peeled-off association matching
1144 * the peer address even on another socket.
1146 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1147 err = -EADDRNOTAVAIL;
1152 /* If a bind() or sctp_bindx() is not called prior to
1153 * an sctp_connectx() call, the system picks an
1154 * ephemeral port and will choose an address set
1155 * equivalent to binding with a wildcard address.
1157 if (!ep->base.bind_addr.port) {
1158 if (sctp_autobind(sk)) {
1164 * If an unprivileged user inherits a 1-many
1165 * style socket with open associations on a
1166 * privileged port, it MAY be permitted to
1167 * accept new associations, but it SHOULD NOT
1168 * be permitted to open new associations.
1170 if (ep->base.bind_addr.port < PROT_SOCK &&
1171 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1177 scope = sctp_scope(&to);
1178 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1184 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1192 /* Prime the peer's transport structures. */
1193 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1201 addr_buf += af->sockaddr_len;
1202 walk_size += af->sockaddr_len;
1205 /* In case the user of sctp_connectx() wants an association
1206 * id back, assign one now.
1209 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1214 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1219 /* Initialize sk's dport and daddr for getpeername() */
1220 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1221 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1222 af->to_sk_daddr(sa_addr, sk);
1225 /* in-kernel sockets don't generally have a file allocated to them
1226 * if all they do is call sock_create_kern().
1228 if (sk->sk_socket->file)
1229 f_flags = sk->sk_socket->file->f_flags;
1231 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1233 err = sctp_wait_for_connect(asoc, &timeo);
1234 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1235 *assoc_id = asoc->assoc_id;
1237 /* Don't free association on exit. */
1242 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1243 " kaddrs: %p err: %d\n",
1246 /* sctp_primitive_ASSOCIATE may have added this association
1247 * To the hash table, try to unhash it, just in case, its a noop
1248 * if it wasn't hashed so we're safe
1250 sctp_unhash_established(asoc);
1251 sctp_association_free(asoc);
1256 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1259 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1260 * sctp_assoc_t *asoc);
1262 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1263 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1264 * or IPv6 addresses.
1266 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1267 * Section 3.1.2 for this usage.
1269 * addrs is a pointer to an array of one or more socket addresses. Each
1270 * address is contained in its appropriate structure (i.e. struct
1271 * sockaddr_in or struct sockaddr_in6) the family of the address type
1272 * must be used to distengish the address length (note that this
1273 * representation is termed a "packed array" of addresses). The caller
1274 * specifies the number of addresses in the array with addrcnt.
1276 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1277 * the association id of the new association. On failure, sctp_connectx()
1278 * returns -1, and sets errno to the appropriate error code. The assoc_id
1279 * is not touched by the kernel.
1281 * For SCTP, the port given in each socket address must be the same, or
1282 * sctp_connectx() will fail, setting errno to EINVAL.
1284 * An application can use sctp_connectx to initiate an association with
1285 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1286 * allows a caller to specify multiple addresses at which a peer can be
1287 * reached. The way the SCTP stack uses the list of addresses to set up
1288 * the association is implementation dependent. This function only
1289 * specifies that the stack will try to make use of all the addresses in
1290 * the list when needed.
1292 * Note that the list of addresses passed in is only used for setting up
1293 * the association. It does not necessarily equal the set of addresses
1294 * the peer uses for the resulting association. If the caller wants to
1295 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1296 * retrieve them after the association has been set up.
1298 * Basically do nothing but copying the addresses from user to kernel
1299 * land and invoking either sctp_connectx(). This is used for tunneling
1300 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1302 * We don't use copy_from_user() for optimization: we first do the
1303 * sanity checks (buffer size -fast- and access check-healthy
1304 * pointer); if all of those succeed, then we can alloc the memory
1305 * (expensive operation) needed to copy the data to kernel. Then we do
1306 * the copying without checking the user space area
1307 * (__copy_from_user()).
1309 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1312 * sk The sk of the socket
1313 * addrs The pointer to the addresses in user land
1314 * addrssize Size of the addrs buffer
1316 * Returns >=0 if ok, <0 errno code on error.
1318 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1319 struct sockaddr __user *addrs,
1321 sctp_assoc_t *assoc_id)
1324 struct sockaddr *kaddrs;
1326 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1327 __func__, sk, addrs, addrs_size);
1329 if (unlikely(addrs_size <= 0))
1332 /* Check the user passed a healthy pointer. */
1333 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1336 /* Alloc space for the address array in kernel memory. */
1337 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1338 if (unlikely(!kaddrs))
1341 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1344 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1353 * This is an older interface. It's kept for backward compatibility
1354 * to the option that doesn't provide association id.
1356 SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1357 struct sockaddr __user *addrs,
1360 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1364 * New interface for the API. The since the API is done with a socket
1365 * option, to make it simple we feed back the association id is as a return
1366 * indication to the call. Error is always negative and association id is
1369 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1370 struct sockaddr __user *addrs,
1373 sctp_assoc_t assoc_id = 0;
1376 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1385 * New (hopefully final) interface for the API.
1386 * We use the sctp_getaddrs_old structure so that use-space library
1387 * can avoid any unnecessary allocations. The only defferent part
1388 * is that we store the actual length of the address buffer into the
1389 * addrs_num structure member. That way we can re-use the existing
1392 SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1393 char __user *optval,
1396 struct sctp_getaddrs_old param;
1397 sctp_assoc_t assoc_id = 0;
1400 if (len < sizeof(param))
1403 if (copy_from_user(¶m, optval, sizeof(param)))
1406 err = __sctp_setsockopt_connectx(sk,
1407 (struct sockaddr __user *)param.addrs,
1408 param.addr_num, &assoc_id);
1410 if (err == 0 || err == -EINPROGRESS) {
1411 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1413 if (put_user(sizeof(assoc_id), optlen))
1420 /* API 3.1.4 close() - UDP Style Syntax
1421 * Applications use close() to perform graceful shutdown (as described in
1422 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1423 * by a UDP-style socket.
1427 * ret = close(int sd);
1429 * sd - the socket descriptor of the associations to be closed.
1431 * To gracefully shutdown a specific association represented by the
1432 * UDP-style socket, an application should use the sendmsg() call,
1433 * passing no user data, but including the appropriate flag in the
1434 * ancillary data (see Section xxxx).
1436 * If sd in the close() call is a branched-off socket representing only
1437 * one association, the shutdown is performed on that association only.
1439 * 4.1.6 close() - TCP Style Syntax
1441 * Applications use close() to gracefully close down an association.
1445 * int close(int sd);
1447 * sd - the socket descriptor of the association to be closed.
1449 * After an application calls close() on a socket descriptor, no further
1450 * socket operations will succeed on that descriptor.
1452 * API 7.1.4 SO_LINGER
1454 * An application using the TCP-style socket can use this option to
1455 * perform the SCTP ABORT primitive. The linger option structure is:
1458 * int l_onoff; // option on/off
1459 * int l_linger; // linger time
1462 * To enable the option, set l_onoff to 1. If the l_linger value is set
1463 * to 0, calling close() is the same as the ABORT primitive. If the
1464 * value is set to a negative value, the setsockopt() call will return
1465 * an error. If the value is set to a positive value linger_time, the
1466 * close() can be blocked for at most linger_time ms. If the graceful
1467 * shutdown phase does not finish during this period, close() will
1468 * return but the graceful shutdown phase continues in the system.
1470 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1472 struct net *net = sock_net(sk);
1473 struct sctp_endpoint *ep;
1474 struct sctp_association *asoc;
1475 struct list_head *pos, *temp;
1476 unsigned int data_was_unread;
1478 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1481 sk->sk_shutdown = SHUTDOWN_MASK;
1482 sk->sk_state = SCTP_SS_CLOSING;
1484 ep = sctp_sk(sk)->ep;
1486 /* Clean up any skbs sitting on the receive queue. */
1487 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1488 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1490 /* Walk all associations on an endpoint. */
1491 list_for_each_safe(pos, temp, &ep->asocs) {
1492 asoc = list_entry(pos, struct sctp_association, asocs);
1494 if (sctp_style(sk, TCP)) {
1495 /* A closed association can still be in the list if
1496 * it belongs to a TCP-style listening socket that is
1497 * not yet accepted. If so, free it. If not, send an
1498 * ABORT or SHUTDOWN based on the linger options.
1500 if (sctp_state(asoc, CLOSED)) {
1501 sctp_unhash_established(asoc);
1502 sctp_association_free(asoc);
1507 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1508 !skb_queue_empty(&asoc->ulpq.reasm) ||
1509 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1510 struct sctp_chunk *chunk;
1512 chunk = sctp_make_abort_user(asoc, NULL, 0);
1514 sctp_primitive_ABORT(net, asoc, chunk);
1516 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1519 /* On a TCP-style socket, block for at most linger_time if set. */
1520 if (sctp_style(sk, TCP) && timeout)
1521 sctp_wait_for_close(sk, timeout);
1523 /* This will run the backlog queue. */
1524 sctp_release_sock(sk);
1526 /* Supposedly, no process has access to the socket, but
1527 * the net layers still may.
1529 sctp_local_bh_disable();
1530 sctp_bh_lock_sock(sk);
1532 /* Hold the sock, since sk_common_release() will put sock_put()
1533 * and we have just a little more cleanup.
1536 sk_common_release(sk);
1538 sctp_bh_unlock_sock(sk);
1539 sctp_local_bh_enable();
1543 SCTP_DBG_OBJCNT_DEC(sock);
1546 /* Handle EPIPE error. */
1547 static int sctp_error(struct sock *sk, int flags, int err)
1550 err = sock_error(sk) ? : -EPIPE;
1551 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1552 send_sig(SIGPIPE, current, 0);
1556 /* API 3.1.3 sendmsg() - UDP Style Syntax
1558 * An application uses sendmsg() and recvmsg() calls to transmit data to
1559 * and receive data from its peer.
1561 * ssize_t sendmsg(int socket, const struct msghdr *message,
1564 * socket - the socket descriptor of the endpoint.
1565 * message - pointer to the msghdr structure which contains a single
1566 * user message and possibly some ancillary data.
1568 * See Section 5 for complete description of the data
1571 * flags - flags sent or received with the user message, see Section
1572 * 5 for complete description of the flags.
1574 * Note: This function could use a rewrite especially when explicit
1575 * connect support comes in.
1577 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1579 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1581 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1582 struct msghdr *msg, size_t msg_len)
1584 struct net *net = sock_net(sk);
1585 struct sctp_sock *sp;
1586 struct sctp_endpoint *ep;
1587 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1588 struct sctp_transport *transport, *chunk_tp;
1589 struct sctp_chunk *chunk;
1591 struct sockaddr *msg_name = NULL;
1592 struct sctp_sndrcvinfo default_sinfo;
1593 struct sctp_sndrcvinfo *sinfo;
1594 struct sctp_initmsg *sinit;
1595 sctp_assoc_t associd = 0;
1596 sctp_cmsgs_t cmsgs = { NULL };
1600 __u16 sinfo_flags = 0;
1601 struct sctp_datamsg *datamsg;
1602 int msg_flags = msg->msg_flags;
1604 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1611 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1613 /* We cannot send a message over a TCP-style listening socket. */
1614 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1619 /* Parse out the SCTP CMSGs. */
1620 err = sctp_msghdr_parse(msg, &cmsgs);
1623 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1627 /* Fetch the destination address for this packet. This
1628 * address only selects the association--it is not necessarily
1629 * the address we will send to.
1630 * For a peeled-off socket, msg_name is ignored.
1632 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1633 int msg_namelen = msg->msg_namelen;
1635 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1640 if (msg_namelen > sizeof(to))
1641 msg_namelen = sizeof(to);
1642 memcpy(&to, msg->msg_name, msg_namelen);
1643 msg_name = msg->msg_name;
1649 /* Did the user specify SNDRCVINFO? */
1651 sinfo_flags = sinfo->sinfo_flags;
1652 associd = sinfo->sinfo_assoc_id;
1655 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1656 msg_len, sinfo_flags);
1658 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1659 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1664 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1665 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1666 * If SCTP_ABORT is set, the message length could be non zero with
1667 * the msg_iov set to the user abort reason.
1669 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1670 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1675 /* If SCTP_ADDR_OVER is set, there must be an address
1676 * specified in msg_name.
1678 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1685 SCTP_DEBUG_PRINTK("About to look up association.\n");
1689 /* If a msg_name has been specified, assume this is to be used. */
1691 /* Look for a matching association on the endpoint. */
1692 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1694 /* If we could not find a matching association on the
1695 * endpoint, make sure that it is not a TCP-style
1696 * socket that already has an association or there is
1697 * no peeled-off association on another socket.
1699 if ((sctp_style(sk, TCP) &&
1700 sctp_sstate(sk, ESTABLISHED)) ||
1701 sctp_endpoint_is_peeled_off(ep, &to)) {
1702 err = -EADDRNOTAVAIL;
1707 asoc = sctp_id2assoc(sk, associd);
1715 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1717 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1718 * socket that has an association in CLOSED state. This can
1719 * happen when an accepted socket has an association that is
1722 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1727 if (sinfo_flags & SCTP_EOF) {
1728 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1730 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1734 if (sinfo_flags & SCTP_ABORT) {
1736 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1742 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1743 sctp_primitive_ABORT(net, asoc, chunk);
1749 /* Do we need to create the association? */
1751 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1753 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1758 /* Check for invalid stream against the stream counts,
1759 * either the default or the user specified stream counts.
1762 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1763 /* Check against the defaults. */
1764 if (sinfo->sinfo_stream >=
1765 sp->initmsg.sinit_num_ostreams) {
1770 /* Check against the requested. */
1771 if (sinfo->sinfo_stream >=
1772 sinit->sinit_num_ostreams) {
1780 * API 3.1.2 bind() - UDP Style Syntax
1781 * If a bind() or sctp_bindx() is not called prior to a
1782 * sendmsg() call that initiates a new association, the
1783 * system picks an ephemeral port and will choose an address
1784 * set equivalent to binding with a wildcard address.
1786 if (!ep->base.bind_addr.port) {
1787 if (sctp_autobind(sk)) {
1793 * If an unprivileged user inherits a one-to-many
1794 * style socket with open associations on a privileged
1795 * port, it MAY be permitted to accept new associations,
1796 * but it SHOULD NOT be permitted to open new
1799 if (ep->base.bind_addr.port < PROT_SOCK &&
1800 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1806 scope = sctp_scope(&to);
1807 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1813 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1819 /* If the SCTP_INIT ancillary data is specified, set all
1820 * the association init values accordingly.
1823 if (sinit->sinit_num_ostreams) {
1824 asoc->c.sinit_num_ostreams =
1825 sinit->sinit_num_ostreams;
1827 if (sinit->sinit_max_instreams) {
1828 asoc->c.sinit_max_instreams =
1829 sinit->sinit_max_instreams;
1831 if (sinit->sinit_max_attempts) {
1832 asoc->max_init_attempts
1833 = sinit->sinit_max_attempts;
1835 if (sinit->sinit_max_init_timeo) {
1836 asoc->max_init_timeo =
1837 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1841 /* Prime the peer's transport structures. */
1842 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1849 /* ASSERT: we have a valid association at this point. */
1850 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1853 /* If the user didn't specify SNDRCVINFO, make up one with
1856 memset(&default_sinfo, 0, sizeof(default_sinfo));
1857 default_sinfo.sinfo_stream = asoc->default_stream;
1858 default_sinfo.sinfo_flags = asoc->default_flags;
1859 default_sinfo.sinfo_ppid = asoc->default_ppid;
1860 default_sinfo.sinfo_context = asoc->default_context;
1861 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1862 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1863 sinfo = &default_sinfo;
1866 /* API 7.1.7, the sndbuf size per association bounds the
1867 * maximum size of data that can be sent in a single send call.
1869 if (msg_len > sk->sk_sndbuf) {
1874 if (asoc->pmtu_pending)
1875 sctp_assoc_pending_pmtu(sk, asoc);
1877 /* If fragmentation is disabled and the message length exceeds the
1878 * association fragmentation point, return EMSGSIZE. The I-D
1879 * does not specify what this error is, but this looks like
1882 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1887 /* Check for invalid stream. */
1888 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1893 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1894 if (!sctp_wspace(asoc)) {
1895 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1900 /* If an address is passed with the sendto/sendmsg call, it is used
1901 * to override the primary destination address in the TCP model, or
1902 * when SCTP_ADDR_OVER flag is set in the UDP model.
1904 if ((sctp_style(sk, TCP) && msg_name) ||
1905 (sinfo_flags & SCTP_ADDR_OVER)) {
1906 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1914 /* Auto-connect, if we aren't connected already. */
1915 if (sctp_state(asoc, CLOSED)) {
1916 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1919 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1922 /* Break the message into multiple chunks of maximum size. */
1923 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1924 if (IS_ERR(datamsg)) {
1925 err = PTR_ERR(datamsg);
1929 /* Now send the (possibly) fragmented message. */
1930 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1931 sctp_chunk_hold(chunk);
1933 /* Do accounting for the write space. */
1934 sctp_set_owner_w(chunk);
1936 chunk->transport = chunk_tp;
1939 /* Send it to the lower layers. Note: all chunks
1940 * must either fail or succeed. The lower layer
1941 * works that way today. Keep it that way or this
1944 err = sctp_primitive_SEND(net, asoc, datamsg);
1945 /* Did the lower layer accept the chunk? */
1947 sctp_datamsg_free(datamsg);
1949 sctp_datamsg_put(datamsg);
1951 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1958 /* If we are already past ASSOCIATE, the lower
1959 * layers are responsible for association cleanup.
1965 sctp_unhash_established(asoc);
1966 sctp_association_free(asoc);
1969 sctp_release_sock(sk);
1972 return sctp_error(sk, msg_flags, err);
1979 err = sock_error(sk);
1989 /* This is an extended version of skb_pull() that removes the data from the
1990 * start of a skb even when data is spread across the list of skb's in the
1991 * frag_list. len specifies the total amount of data that needs to be removed.
1992 * when 'len' bytes could be removed from the skb, it returns 0.
1993 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1994 * could not be removed.
1996 static int sctp_skb_pull(struct sk_buff *skb, int len)
1998 struct sk_buff *list;
1999 int skb_len = skb_headlen(skb);
2002 if (len <= skb_len) {
2003 __skb_pull(skb, len);
2007 __skb_pull(skb, skb_len);
2009 skb_walk_frags(skb, list) {
2010 rlen = sctp_skb_pull(list, len);
2011 skb->len -= (len-rlen);
2012 skb->data_len -= (len-rlen);
2023 /* API 3.1.3 recvmsg() - UDP Style Syntax
2025 * ssize_t recvmsg(int socket, struct msghdr *message,
2028 * socket - the socket descriptor of the endpoint.
2029 * message - pointer to the msghdr structure which contains a single
2030 * user message and possibly some ancillary data.
2032 * See Section 5 for complete description of the data
2035 * flags - flags sent or received with the user message, see Section
2036 * 5 for complete description of the flags.
2038 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2040 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2041 struct msghdr *msg, size_t len, int noblock,
2042 int flags, int *addr_len)
2044 struct sctp_ulpevent *event = NULL;
2045 struct sctp_sock *sp = sctp_sk(sk);
2046 struct sk_buff *skb;
2051 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
2052 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
2053 "len", len, "knoblauch", noblock,
2054 "flags", flags, "addr_len", addr_len);
2058 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2063 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2067 /* Get the total length of the skb including any skb's in the
2076 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2078 event = sctp_skb2event(skb);
2083 sock_recv_ts_and_drops(msg, sk, skb);
2084 if (sctp_ulpevent_is_notification(event)) {
2085 msg->msg_flags |= MSG_NOTIFICATION;
2086 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2088 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2091 /* Check if we allow SCTP_SNDRCVINFO. */
2092 if (sp->subscribe.sctp_data_io_event)
2093 sctp_ulpevent_read_sndrcvinfo(event, msg);
2095 /* FIXME: we should be calling IP/IPv6 layers. */
2096 if (sk->sk_protinfo.af_inet.cmsg_flags)
2097 ip_cmsg_recv(msg, skb);
2102 /* If skb's length exceeds the user's buffer, update the skb and
2103 * push it back to the receive_queue so that the next call to
2104 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2106 if (skb_len > copied) {
2107 msg->msg_flags &= ~MSG_EOR;
2108 if (flags & MSG_PEEK)
2110 sctp_skb_pull(skb, copied);
2111 skb_queue_head(&sk->sk_receive_queue, skb);
2113 /* When only partial message is copied to the user, increase
2114 * rwnd by that amount. If all the data in the skb is read,
2115 * rwnd is updated when the event is freed.
2117 if (!sctp_ulpevent_is_notification(event))
2118 sctp_assoc_rwnd_increase(event->asoc, copied);
2120 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2121 (event->msg_flags & MSG_EOR))
2122 msg->msg_flags |= MSG_EOR;
2124 msg->msg_flags &= ~MSG_EOR;
2127 if (flags & MSG_PEEK) {
2128 /* Release the skb reference acquired after peeking the skb in
2129 * sctp_skb_recv_datagram().
2133 /* Free the event which includes releasing the reference to
2134 * the owner of the skb, freeing the skb and updating the
2137 sctp_ulpevent_free(event);
2140 sctp_release_sock(sk);
2144 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2146 * This option is a on/off flag. If enabled no SCTP message
2147 * fragmentation will be performed. Instead if a message being sent
2148 * exceeds the current PMTU size, the message will NOT be sent and
2149 * instead a error will be indicated to the user.
2151 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2152 char __user *optval,
2153 unsigned int optlen)
2157 if (optlen < sizeof(int))
2160 if (get_user(val, (int __user *)optval))
2163 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2168 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2169 unsigned int optlen)
2171 struct sctp_association *asoc;
2172 struct sctp_ulpevent *event;
2174 if (optlen > sizeof(struct sctp_event_subscribe))
2176 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2180 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2181 * if there is no data to be sent or retransmit, the stack will
2182 * immediately send up this notification.
2184 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2185 &sctp_sk(sk)->subscribe)) {
2186 asoc = sctp_id2assoc(sk, 0);
2188 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2189 event = sctp_ulpevent_make_sender_dry_event(asoc,
2194 sctp_ulpq_tail_event(&asoc->ulpq, event);
2201 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2203 * This socket option is applicable to the UDP-style socket only. When
2204 * set it will cause associations that are idle for more than the
2205 * specified number of seconds to automatically close. An association
2206 * being idle is defined an association that has NOT sent or received
2207 * user data. The special value of '0' indicates that no automatic
2208 * close of any associations should be performed. The option expects an
2209 * integer defining the number of seconds of idle time before an
2210 * association is closed.
2212 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2213 unsigned int optlen)
2215 struct sctp_sock *sp = sctp_sk(sk);
2217 /* Applicable to UDP-style socket only */
2218 if (sctp_style(sk, TCP))
2220 if (optlen != sizeof(int))
2222 if (copy_from_user(&sp->autoclose, optval, optlen))
2228 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2230 * Applications can enable or disable heartbeats for any peer address of
2231 * an association, modify an address's heartbeat interval, force a
2232 * heartbeat to be sent immediately, and adjust the address's maximum
2233 * number of retransmissions sent before an address is considered
2234 * unreachable. The following structure is used to access and modify an
2235 * address's parameters:
2237 * struct sctp_paddrparams {
2238 * sctp_assoc_t spp_assoc_id;
2239 * struct sockaddr_storage spp_address;
2240 * uint32_t spp_hbinterval;
2241 * uint16_t spp_pathmaxrxt;
2242 * uint32_t spp_pathmtu;
2243 * uint32_t spp_sackdelay;
2244 * uint32_t spp_flags;
2247 * spp_assoc_id - (one-to-many style socket) This is filled in the
2248 * application, and identifies the association for
2250 * spp_address - This specifies which address is of interest.
2251 * spp_hbinterval - This contains the value of the heartbeat interval,
2252 * in milliseconds. If a value of zero
2253 * is present in this field then no changes are to
2254 * be made to this parameter.
2255 * spp_pathmaxrxt - This contains the maximum number of
2256 * retransmissions before this address shall be
2257 * considered unreachable. If a value of zero
2258 * is present in this field then no changes are to
2259 * be made to this parameter.
2260 * spp_pathmtu - When Path MTU discovery is disabled the value
2261 * specified here will be the "fixed" path mtu.
2262 * Note that if the spp_address field is empty
2263 * then all associations on this address will
2264 * have this fixed path mtu set upon them.
2266 * spp_sackdelay - When delayed sack is enabled, this value specifies
2267 * the number of milliseconds that sacks will be delayed
2268 * for. This value will apply to all addresses of an
2269 * association if the spp_address field is empty. Note
2270 * also, that if delayed sack is enabled and this
2271 * value is set to 0, no change is made to the last
2272 * recorded delayed sack timer value.
2274 * spp_flags - These flags are used to control various features
2275 * on an association. The flag field may contain
2276 * zero or more of the following options.
2278 * SPP_HB_ENABLE - Enable heartbeats on the
2279 * specified address. Note that if the address
2280 * field is empty all addresses for the association
2281 * have heartbeats enabled upon them.
2283 * SPP_HB_DISABLE - Disable heartbeats on the
2284 * speicifed address. Note that if the address
2285 * field is empty all addresses for the association
2286 * will have their heartbeats disabled. Note also
2287 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2288 * mutually exclusive, only one of these two should
2289 * be specified. Enabling both fields will have
2290 * undetermined results.
2292 * SPP_HB_DEMAND - Request a user initiated heartbeat
2293 * to be made immediately.
2295 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2296 * heartbeat delayis to be set to the value of 0
2299 * SPP_PMTUD_ENABLE - This field will enable PMTU
2300 * discovery upon the specified address. Note that
2301 * if the address feild is empty then all addresses
2302 * on the association are effected.
2304 * SPP_PMTUD_DISABLE - This field will disable PMTU
2305 * discovery upon the specified address. Note that
2306 * if the address feild is empty then all addresses
2307 * on the association are effected. Not also that
2308 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2309 * exclusive. Enabling both will have undetermined
2312 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2313 * on delayed sack. The time specified in spp_sackdelay
2314 * is used to specify the sack delay for this address. Note
2315 * that if spp_address is empty then all addresses will
2316 * enable delayed sack and take on the sack delay
2317 * value specified in spp_sackdelay.
2318 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2319 * off delayed sack. If the spp_address field is blank then
2320 * delayed sack is disabled for the entire association. Note
2321 * also that this field is mutually exclusive to
2322 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2325 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2326 struct sctp_transport *trans,
2327 struct sctp_association *asoc,
2328 struct sctp_sock *sp,
2331 int sackdelay_change)
2335 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2336 struct net *net = sock_net(trans->asoc->base.sk);
2338 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2343 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2344 * this field is ignored. Note also that a value of zero indicates
2345 * the current setting should be left unchanged.
2347 if (params->spp_flags & SPP_HB_ENABLE) {
2349 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2350 * set. This lets us use 0 value when this flag
2353 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2354 params->spp_hbinterval = 0;
2356 if (params->spp_hbinterval ||
2357 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2360 msecs_to_jiffies(params->spp_hbinterval);
2363 msecs_to_jiffies(params->spp_hbinterval);
2365 sp->hbinterval = params->spp_hbinterval;
2372 trans->param_flags =
2373 (trans->param_flags & ~SPP_HB) | hb_change;
2376 (asoc->param_flags & ~SPP_HB) | hb_change;
2379 (sp->param_flags & ~SPP_HB) | hb_change;
2383 /* When Path MTU discovery is disabled the value specified here will
2384 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2385 * include the flag SPP_PMTUD_DISABLE for this field to have any
2388 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2390 trans->pathmtu = params->spp_pathmtu;
2391 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2393 asoc->pathmtu = params->spp_pathmtu;
2394 sctp_frag_point(asoc, params->spp_pathmtu);
2396 sp->pathmtu = params->spp_pathmtu;
2402 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2403 (params->spp_flags & SPP_PMTUD_ENABLE);
2404 trans->param_flags =
2405 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2407 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2408 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2412 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2415 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2419 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2420 * value of this field is ignored. Note also that a value of zero
2421 * indicates the current setting should be left unchanged.
2423 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2426 msecs_to_jiffies(params->spp_sackdelay);
2429 msecs_to_jiffies(params->spp_sackdelay);
2431 sp->sackdelay = params->spp_sackdelay;
2435 if (sackdelay_change) {
2437 trans->param_flags =
2438 (trans->param_flags & ~SPP_SACKDELAY) |
2442 (asoc->param_flags & ~SPP_SACKDELAY) |
2446 (sp->param_flags & ~SPP_SACKDELAY) |
2451 /* Note that a value of zero indicates the current setting should be
2454 if (params->spp_pathmaxrxt) {
2456 trans->pathmaxrxt = params->spp_pathmaxrxt;
2458 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2460 sp->pathmaxrxt = params->spp_pathmaxrxt;
2467 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2468 char __user *optval,
2469 unsigned int optlen)
2471 struct sctp_paddrparams params;
2472 struct sctp_transport *trans = NULL;
2473 struct sctp_association *asoc = NULL;
2474 struct sctp_sock *sp = sctp_sk(sk);
2476 int hb_change, pmtud_change, sackdelay_change;
2478 if (optlen != sizeof(struct sctp_paddrparams))
2481 if (copy_from_user(¶ms, optval, optlen))
2484 /* Validate flags and value parameters. */
2485 hb_change = params.spp_flags & SPP_HB;
2486 pmtud_change = params.spp_flags & SPP_PMTUD;
2487 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2489 if (hb_change == SPP_HB ||
2490 pmtud_change == SPP_PMTUD ||
2491 sackdelay_change == SPP_SACKDELAY ||
2492 params.spp_sackdelay > 500 ||
2493 (params.spp_pathmtu &&
2494 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2497 /* If an address other than INADDR_ANY is specified, and
2498 * no transport is found, then the request is invalid.
2500 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
2501 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2502 params.spp_assoc_id);
2507 /* Get association, if assoc_id != 0 and the socket is a one
2508 * to many style socket, and an association was not found, then
2509 * the id was invalid.
2511 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2512 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2515 /* Heartbeat demand can only be sent on a transport or
2516 * association, but not a socket.
2518 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2521 /* Process parameters. */
2522 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2523 hb_change, pmtud_change,
2529 /* If changes are for association, also apply parameters to each
2532 if (!trans && asoc) {
2533 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2535 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2536 hb_change, pmtud_change,
2545 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2547 * This option will effect the way delayed acks are performed. This
2548 * option allows you to get or set the delayed ack time, in
2549 * milliseconds. It also allows changing the delayed ack frequency.
2550 * Changing the frequency to 1 disables the delayed sack algorithm. If
2551 * the assoc_id is 0, then this sets or gets the endpoints default
2552 * values. If the assoc_id field is non-zero, then the set or get
2553 * effects the specified association for the one to many model (the
2554 * assoc_id field is ignored by the one to one model). Note that if
2555 * sack_delay or sack_freq are 0 when setting this option, then the
2556 * current values will remain unchanged.
2558 * struct sctp_sack_info {
2559 * sctp_assoc_t sack_assoc_id;
2560 * uint32_t sack_delay;
2561 * uint32_t sack_freq;
2564 * sack_assoc_id - This parameter, indicates which association the user
2565 * is performing an action upon. Note that if this field's value is
2566 * zero then the endpoints default value is changed (effecting future
2567 * associations only).
2569 * sack_delay - This parameter contains the number of milliseconds that
2570 * the user is requesting the delayed ACK timer be set to. Note that
2571 * this value is defined in the standard to be between 200 and 500
2574 * sack_freq - This parameter contains the number of packets that must
2575 * be received before a sack is sent without waiting for the delay
2576 * timer to expire. The default value for this is 2, setting this
2577 * value to 1 will disable the delayed sack algorithm.
2580 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2581 char __user *optval, unsigned int optlen)
2583 struct sctp_sack_info params;
2584 struct sctp_transport *trans = NULL;
2585 struct sctp_association *asoc = NULL;
2586 struct sctp_sock *sp = sctp_sk(sk);
2588 if (optlen == sizeof(struct sctp_sack_info)) {
2589 if (copy_from_user(¶ms, optval, optlen))
2592 if (params.sack_delay == 0 && params.sack_freq == 0)
2594 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2595 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2596 pr_warn("Use struct sctp_sack_info instead\n");
2597 if (copy_from_user(¶ms, optval, optlen))
2600 if (params.sack_delay == 0)
2601 params.sack_freq = 1;
2603 params.sack_freq = 0;
2607 /* Validate value parameter. */
2608 if (params.sack_delay > 500)
2611 /* Get association, if sack_assoc_id != 0 and the socket is a one
2612 * to many style socket, and an association was not found, then
2613 * the id was invalid.
2615 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2616 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2619 if (params.sack_delay) {
2622 msecs_to_jiffies(params.sack_delay);
2624 (asoc->param_flags & ~SPP_SACKDELAY) |
2625 SPP_SACKDELAY_ENABLE;
2627 sp->sackdelay = params.sack_delay;
2629 (sp->param_flags & ~SPP_SACKDELAY) |
2630 SPP_SACKDELAY_ENABLE;
2634 if (params.sack_freq == 1) {
2637 (asoc->param_flags & ~SPP_SACKDELAY) |
2638 SPP_SACKDELAY_DISABLE;
2641 (sp->param_flags & ~SPP_SACKDELAY) |
2642 SPP_SACKDELAY_DISABLE;
2644 } else if (params.sack_freq > 1) {
2646 asoc->sackfreq = params.sack_freq;
2648 (asoc->param_flags & ~SPP_SACKDELAY) |
2649 SPP_SACKDELAY_ENABLE;
2651 sp->sackfreq = params.sack_freq;
2653 (sp->param_flags & ~SPP_SACKDELAY) |
2654 SPP_SACKDELAY_ENABLE;
2658 /* If change is for association, also apply to each transport. */
2660 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2662 if (params.sack_delay) {
2664 msecs_to_jiffies(params.sack_delay);
2665 trans->param_flags =
2666 (trans->param_flags & ~SPP_SACKDELAY) |
2667 SPP_SACKDELAY_ENABLE;
2669 if (params.sack_freq == 1) {
2670 trans->param_flags =
2671 (trans->param_flags & ~SPP_SACKDELAY) |
2672 SPP_SACKDELAY_DISABLE;
2673 } else if (params.sack_freq > 1) {
2674 trans->sackfreq = params.sack_freq;
2675 trans->param_flags =
2676 (trans->param_flags & ~SPP_SACKDELAY) |
2677 SPP_SACKDELAY_ENABLE;
2685 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2687 * Applications can specify protocol parameters for the default association
2688 * initialization. The option name argument to setsockopt() and getsockopt()
2691 * Setting initialization parameters is effective only on an unconnected
2692 * socket (for UDP-style sockets only future associations are effected
2693 * by the change). With TCP-style sockets, this option is inherited by
2694 * sockets derived from a listener socket.
2696 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2698 struct sctp_initmsg sinit;
2699 struct sctp_sock *sp = sctp_sk(sk);
2701 if (optlen != sizeof(struct sctp_initmsg))
2703 if (copy_from_user(&sinit, optval, optlen))
2706 if (sinit.sinit_num_ostreams)
2707 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2708 if (sinit.sinit_max_instreams)
2709 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2710 if (sinit.sinit_max_attempts)
2711 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2712 if (sinit.sinit_max_init_timeo)
2713 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2719 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2721 * Applications that wish to use the sendto() system call may wish to
2722 * specify a default set of parameters that would normally be supplied
2723 * through the inclusion of ancillary data. This socket option allows
2724 * such an application to set the default sctp_sndrcvinfo structure.
2725 * The application that wishes to use this socket option simply passes
2726 * in to this call the sctp_sndrcvinfo structure defined in Section
2727 * 5.2.2) The input parameters accepted by this call include
2728 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2729 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2730 * to this call if the caller is using the UDP model.
2732 static int sctp_setsockopt_default_send_param(struct sock *sk,
2733 char __user *optval,
2734 unsigned int optlen)
2736 struct sctp_sndrcvinfo info;
2737 struct sctp_association *asoc;
2738 struct sctp_sock *sp = sctp_sk(sk);
2740 if (optlen != sizeof(struct sctp_sndrcvinfo))
2742 if (copy_from_user(&info, optval, optlen))
2745 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2746 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2750 asoc->default_stream = info.sinfo_stream;
2751 asoc->default_flags = info.sinfo_flags;
2752 asoc->default_ppid = info.sinfo_ppid;
2753 asoc->default_context = info.sinfo_context;
2754 asoc->default_timetolive = info.sinfo_timetolive;
2756 sp->default_stream = info.sinfo_stream;
2757 sp->default_flags = info.sinfo_flags;
2758 sp->default_ppid = info.sinfo_ppid;
2759 sp->default_context = info.sinfo_context;
2760 sp->default_timetolive = info.sinfo_timetolive;
2766 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2768 * Requests that the local SCTP stack use the enclosed peer address as
2769 * the association primary. The enclosed address must be one of the
2770 * association peer's addresses.
2772 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2773 unsigned int optlen)
2775 struct sctp_prim prim;
2776 struct sctp_transport *trans;
2778 if (optlen != sizeof(struct sctp_prim))
2781 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2784 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2788 sctp_assoc_set_primary(trans->asoc, trans);
2794 * 7.1.5 SCTP_NODELAY
2796 * Turn on/off any Nagle-like algorithm. This means that packets are
2797 * generally sent as soon as possible and no unnecessary delays are
2798 * introduced, at the cost of more packets in the network. Expects an
2799 * integer boolean flag.
2801 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2802 unsigned int optlen)
2806 if (optlen < sizeof(int))
2808 if (get_user(val, (int __user *)optval))
2811 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2817 * 7.1.1 SCTP_RTOINFO
2819 * The protocol parameters used to initialize and bound retransmission
2820 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2821 * and modify these parameters.
2822 * All parameters are time values, in milliseconds. A value of 0, when
2823 * modifying the parameters, indicates that the current value should not
2827 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2829 struct sctp_rtoinfo rtoinfo;
2830 struct sctp_association *asoc;
2832 if (optlen != sizeof (struct sctp_rtoinfo))
2835 if (copy_from_user(&rtoinfo, optval, optlen))
2838 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2840 /* Set the values to the specific association */
2841 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2845 if (rtoinfo.srto_initial != 0)
2847 msecs_to_jiffies(rtoinfo.srto_initial);
2848 if (rtoinfo.srto_max != 0)
2849 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2850 if (rtoinfo.srto_min != 0)
2851 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2853 /* If there is no association or the association-id = 0
2854 * set the values to the endpoint.
2856 struct sctp_sock *sp = sctp_sk(sk);
2858 if (rtoinfo.srto_initial != 0)
2859 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2860 if (rtoinfo.srto_max != 0)
2861 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2862 if (rtoinfo.srto_min != 0)
2863 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2871 * 7.1.2 SCTP_ASSOCINFO
2873 * This option is used to tune the maximum retransmission attempts
2874 * of the association.
2875 * Returns an error if the new association retransmission value is
2876 * greater than the sum of the retransmission value of the peer.
2877 * See [SCTP] for more information.
2880 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2883 struct sctp_assocparams assocparams;
2884 struct sctp_association *asoc;
2886 if (optlen != sizeof(struct sctp_assocparams))
2888 if (copy_from_user(&assocparams, optval, optlen))
2891 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2893 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2896 /* Set the values to the specific association */
2898 if (assocparams.sasoc_asocmaxrxt != 0) {
2901 struct sctp_transport *peer_addr;
2903 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2905 path_sum += peer_addr->pathmaxrxt;
2909 /* Only validate asocmaxrxt if we have more than
2910 * one path/transport. We do this because path
2911 * retransmissions are only counted when we have more
2915 assocparams.sasoc_asocmaxrxt > path_sum)
2918 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2921 if (assocparams.sasoc_cookie_life != 0) {
2922 asoc->cookie_life.tv_sec =
2923 assocparams.sasoc_cookie_life / 1000;
2924 asoc->cookie_life.tv_usec =
2925 (assocparams.sasoc_cookie_life % 1000)
2929 /* Set the values to the endpoint */
2930 struct sctp_sock *sp = sctp_sk(sk);
2932 if (assocparams.sasoc_asocmaxrxt != 0)
2933 sp->assocparams.sasoc_asocmaxrxt =
2934 assocparams.sasoc_asocmaxrxt;
2935 if (assocparams.sasoc_cookie_life != 0)
2936 sp->assocparams.sasoc_cookie_life =
2937 assocparams.sasoc_cookie_life;
2943 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2945 * This socket option is a boolean flag which turns on or off mapped V4
2946 * addresses. If this option is turned on and the socket is type
2947 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2948 * If this option is turned off, then no mapping will be done of V4
2949 * addresses and a user will receive both PF_INET6 and PF_INET type
2950 * addresses on the socket.
2952 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2955 struct sctp_sock *sp = sctp_sk(sk);
2957 if (optlen < sizeof(int))
2959 if (get_user(val, (int __user *)optval))
2970 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2971 * This option will get or set the maximum size to put in any outgoing
2972 * SCTP DATA chunk. If a message is larger than this size it will be
2973 * fragmented by SCTP into the specified size. Note that the underlying
2974 * SCTP implementation may fragment into smaller sized chunks when the
2975 * PMTU of the underlying association is smaller than the value set by
2976 * the user. The default value for this option is '0' which indicates
2977 * the user is NOT limiting fragmentation and only the PMTU will effect
2978 * SCTP's choice of DATA chunk size. Note also that values set larger
2979 * than the maximum size of an IP datagram will effectively let SCTP
2980 * control fragmentation (i.e. the same as setting this option to 0).
2982 * The following structure is used to access and modify this parameter:
2984 * struct sctp_assoc_value {
2985 * sctp_assoc_t assoc_id;
2986 * uint32_t assoc_value;
2989 * assoc_id: This parameter is ignored for one-to-one style sockets.
2990 * For one-to-many style sockets this parameter indicates which
2991 * association the user is performing an action upon. Note that if
2992 * this field's value is zero then the endpoints default value is
2993 * changed (effecting future associations only).
2994 * assoc_value: This parameter specifies the maximum size in bytes.
2996 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2998 struct sctp_assoc_value params;
2999 struct sctp_association *asoc;
3000 struct sctp_sock *sp = sctp_sk(sk);
3003 if (optlen == sizeof(int)) {
3004 pr_warn("Use of int in maxseg socket option deprecated\n");
3005 pr_warn("Use struct sctp_assoc_value instead\n");
3006 if (copy_from_user(&val, optval, optlen))
3008 params.assoc_id = 0;
3009 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3010 if (copy_from_user(¶ms, optval, optlen))
3012 val = params.assoc_value;
3016 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3019 asoc = sctp_id2assoc(sk, params.assoc_id);
3020 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3025 val = asoc->pathmtu;
3026 val -= sp->pf->af->net_header_len;
3027 val -= sizeof(struct sctphdr) +
3028 sizeof(struct sctp_data_chunk);
3030 asoc->user_frag = val;
3031 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3033 sp->user_frag = val;
3041 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3043 * Requests that the peer mark the enclosed address as the association
3044 * primary. The enclosed address must be one of the association's
3045 * locally bound addresses. The following structure is used to make a
3046 * set primary request:
3048 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3049 unsigned int optlen)
3051 struct net *net = sock_net(sk);
3052 struct sctp_sock *sp;
3053 struct sctp_association *asoc = NULL;
3054 struct sctp_setpeerprim prim;
3055 struct sctp_chunk *chunk;
3061 if (!net->sctp.addip_enable)
3064 if (optlen != sizeof(struct sctp_setpeerprim))
3067 if (copy_from_user(&prim, optval, optlen))
3070 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3074 if (!asoc->peer.asconf_capable)
3077 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3080 if (!sctp_state(asoc, ESTABLISHED))
3083 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3087 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3088 return -EADDRNOTAVAIL;
3090 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3091 return -EADDRNOTAVAIL;
3093 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3094 chunk = sctp_make_asconf_set_prim(asoc,
3095 (union sctp_addr *)&prim.sspp_addr);
3099 err = sctp_send_asconf(asoc, chunk);
3101 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
3106 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3107 unsigned int optlen)
3109 struct sctp_setadaptation adaptation;
3111 if (optlen != sizeof(struct sctp_setadaptation))
3113 if (copy_from_user(&adaptation, optval, optlen))
3116 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3122 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3124 * The context field in the sctp_sndrcvinfo structure is normally only
3125 * used when a failed message is retrieved holding the value that was
3126 * sent down on the actual send call. This option allows the setting of
3127 * a default context on an association basis that will be received on
3128 * reading messages from the peer. This is especially helpful in the
3129 * one-2-many model for an application to keep some reference to an
3130 * internal state machine that is processing messages on the
3131 * association. Note that the setting of this value only effects
3132 * received messages from the peer and does not effect the value that is
3133 * saved with outbound messages.
3135 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3136 unsigned int optlen)
3138 struct sctp_assoc_value params;
3139 struct sctp_sock *sp;
3140 struct sctp_association *asoc;
3142 if (optlen != sizeof(struct sctp_assoc_value))
3144 if (copy_from_user(¶ms, optval, optlen))
3149 if (params.assoc_id != 0) {
3150 asoc = sctp_id2assoc(sk, params.assoc_id);
3153 asoc->default_rcv_context = params.assoc_value;
3155 sp->default_rcv_context = params.assoc_value;
3162 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3164 * This options will at a minimum specify if the implementation is doing
3165 * fragmented interleave. Fragmented interleave, for a one to many
3166 * socket, is when subsequent calls to receive a message may return
3167 * parts of messages from different associations. Some implementations
3168 * may allow you to turn this value on or off. If so, when turned off,
3169 * no fragment interleave will occur (which will cause a head of line
3170 * blocking amongst multiple associations sharing the same one to many
3171 * socket). When this option is turned on, then each receive call may
3172 * come from a different association (thus the user must receive data
3173 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3174 * association each receive belongs to.
3176 * This option takes a boolean value. A non-zero value indicates that
3177 * fragmented interleave is on. A value of zero indicates that
3178 * fragmented interleave is off.
3180 * Note that it is important that an implementation that allows this
3181 * option to be turned on, have it off by default. Otherwise an unaware
3182 * application using the one to many model may become confused and act
3185 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3186 char __user *optval,
3187 unsigned int optlen)
3191 if (optlen != sizeof(int))
3193 if (get_user(val, (int __user *)optval))
3196 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3202 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3203 * (SCTP_PARTIAL_DELIVERY_POINT)
3205 * This option will set or get the SCTP partial delivery point. This
3206 * point is the size of a message where the partial delivery API will be
3207 * invoked to help free up rwnd space for the peer. Setting this to a
3208 * lower value will cause partial deliveries to happen more often. The
3209 * calls argument is an integer that sets or gets the partial delivery
3210 * point. Note also that the call will fail if the user attempts to set
3211 * this value larger than the socket receive buffer size.
3213 * Note that any single message having a length smaller than or equal to
3214 * the SCTP partial delivery point will be delivered in one single read
3215 * call as long as the user provided buffer is large enough to hold the
3218 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3219 char __user *optval,
3220 unsigned int optlen)
3224 if (optlen != sizeof(u32))
3226 if (get_user(val, (int __user *)optval))
3229 /* Note: We double the receive buffer from what the user sets
3230 * it to be, also initial rwnd is based on rcvbuf/2.
3232 if (val > (sk->sk_rcvbuf >> 1))
3235 sctp_sk(sk)->pd_point = val;
3237 return 0; /* is this the right error code? */
3241 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3243 * This option will allow a user to change the maximum burst of packets
3244 * that can be emitted by this association. Note that the default value
3245 * is 4, and some implementations may restrict this setting so that it
3246 * can only be lowered.
3248 * NOTE: This text doesn't seem right. Do this on a socket basis with
3249 * future associations inheriting the socket value.
3251 static int sctp_setsockopt_maxburst(struct sock *sk,
3252 char __user *optval,
3253 unsigned int optlen)
3255 struct sctp_assoc_value params;
3256 struct sctp_sock *sp;
3257 struct sctp_association *asoc;
3261 if (optlen == sizeof(int)) {
3262 pr_warn("Use of int in max_burst socket option deprecated\n");
3263 pr_warn("Use struct sctp_assoc_value instead\n");
3264 if (copy_from_user(&val, optval, optlen))
3266 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3267 if (copy_from_user(¶ms, optval, optlen))
3269 val = params.assoc_value;
3270 assoc_id = params.assoc_id;
3276 if (assoc_id != 0) {
3277 asoc = sctp_id2assoc(sk, assoc_id);
3280 asoc->max_burst = val;
3282 sp->max_burst = val;
3288 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3290 * This set option adds a chunk type that the user is requesting to be
3291 * received only in an authenticated way. Changes to the list of chunks
3292 * will only effect future associations on the socket.
3294 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3295 char __user *optval,
3296 unsigned int optlen)
3298 struct net *net = sock_net(sk);
3299 struct sctp_authchunk val;
3301 if (!net->sctp.auth_enable)
3304 if (optlen != sizeof(struct sctp_authchunk))
3306 if (copy_from_user(&val, optval, optlen))
3309 switch (val.sauth_chunk) {
3311 case SCTP_CID_INIT_ACK:
3312 case SCTP_CID_SHUTDOWN_COMPLETE:
3317 /* add this chunk id to the endpoint */
3318 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3322 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3324 * This option gets or sets the list of HMAC algorithms that the local
3325 * endpoint requires the peer to use.
3327 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3328 char __user *optval,
3329 unsigned int optlen)
3331 struct net *net = sock_net(sk);
3332 struct sctp_hmacalgo *hmacs;
3336 if (!net->sctp.auth_enable)
3339 if (optlen < sizeof(struct sctp_hmacalgo))
3342 hmacs= memdup_user(optval, optlen);
3344 return PTR_ERR(hmacs);
3346 idents = hmacs->shmac_num_idents;
3347 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3348 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3353 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3360 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3362 * This option will set a shared secret key which is used to build an
3363 * association shared key.
3365 static int sctp_setsockopt_auth_key(struct sock *sk,
3366 char __user *optval,
3367 unsigned int optlen)
3369 struct net *net = sock_net(sk);
3370 struct sctp_authkey *authkey;
3371 struct sctp_association *asoc;
3374 if (!net->sctp.auth_enable)
3377 if (optlen <= sizeof(struct sctp_authkey))
3380 authkey= memdup_user(optval, optlen);
3381 if (IS_ERR(authkey))
3382 return PTR_ERR(authkey);
3384 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3389 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3390 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3395 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3402 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3404 * This option will get or set the active shared key to be used to build
3405 * the association shared key.
3407 static int sctp_setsockopt_active_key(struct sock *sk,
3408 char __user *optval,
3409 unsigned int optlen)
3411 struct net *net = sock_net(sk);
3412 struct sctp_authkeyid val;
3413 struct sctp_association *asoc;
3415 if (!net->sctp.auth_enable)
3418 if (optlen != sizeof(struct sctp_authkeyid))
3420 if (copy_from_user(&val, optval, optlen))
3423 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3424 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3427 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3428 val.scact_keynumber);
3432 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3434 * This set option will delete a shared secret key from use.
3436 static int sctp_setsockopt_del_key(struct sock *sk,
3437 char __user *optval,
3438 unsigned int optlen)
3440 struct net *net = sock_net(sk);
3441 struct sctp_authkeyid val;
3442 struct sctp_association *asoc;
3444 if (!net->sctp.auth_enable)
3447 if (optlen != sizeof(struct sctp_authkeyid))
3449 if (copy_from_user(&val, optval, optlen))
3452 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3453 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3456 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3457 val.scact_keynumber);
3462 * 8.1.23 SCTP_AUTO_ASCONF
3464 * This option will enable or disable the use of the automatic generation of
3465 * ASCONF chunks to add and delete addresses to an existing association. Note
3466 * that this option has two caveats namely: a) it only affects sockets that
3467 * are bound to all addresses available to the SCTP stack, and b) the system
3468 * administrator may have an overriding control that turns the ASCONF feature
3469 * off no matter what setting the socket option may have.
3470 * This option expects an integer boolean flag, where a non-zero value turns on
3471 * the option, and a zero value turns off the option.
3472 * Note. In this implementation, socket operation overrides default parameter
3473 * being set by sysctl as well as FreeBSD implementation
3475 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3476 unsigned int optlen)
3479 struct sctp_sock *sp = sctp_sk(sk);
3481 if (optlen < sizeof(int))
3483 if (get_user(val, (int __user *)optval))
3485 if (!sctp_is_ep_boundall(sk) && val)
3487 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3490 if (val == 0 && sp->do_auto_asconf) {
3491 list_del(&sp->auto_asconf_list);
3492 sp->do_auto_asconf = 0;
3493 } else if (val && !sp->do_auto_asconf) {
3494 list_add_tail(&sp->auto_asconf_list,
3495 &sock_net(sk)->sctp.auto_asconf_splist);
3496 sp->do_auto_asconf = 1;
3503 * SCTP_PEER_ADDR_THLDS
3505 * This option allows us to alter the partially failed threshold for one or all
3506 * transports in an association. See Section 6.1 of:
3507 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3509 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3510 char __user *optval,
3511 unsigned int optlen)
3513 struct sctp_paddrthlds val;
3514 struct sctp_transport *trans;
3515 struct sctp_association *asoc;
3517 if (optlen < sizeof(struct sctp_paddrthlds))
3519 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3520 sizeof(struct sctp_paddrthlds)))
3524 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3525 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3528 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3530 if (val.spt_pathmaxrxt)
3531 trans->pathmaxrxt = val.spt_pathmaxrxt;
3532 trans->pf_retrans = val.spt_pathpfthld;
3535 if (val.spt_pathmaxrxt)
3536 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3537 asoc->pf_retrans = val.spt_pathpfthld;
3539 trans = sctp_addr_id2transport(sk, &val.spt_address,
3544 if (val.spt_pathmaxrxt)
3545 trans->pathmaxrxt = val.spt_pathmaxrxt;
3546 trans->pf_retrans = val.spt_pathpfthld;
3552 /* API 6.2 setsockopt(), getsockopt()
3554 * Applications use setsockopt() and getsockopt() to set or retrieve
3555 * socket options. Socket options are used to change the default
3556 * behavior of sockets calls. They are described in Section 7.
3560 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3561 * int __user *optlen);
3562 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3565 * sd - the socket descript.
3566 * level - set to IPPROTO_SCTP for all SCTP options.
3567 * optname - the option name.
3568 * optval - the buffer to store the value of the option.
3569 * optlen - the size of the buffer.
3571 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3572 char __user *optval, unsigned int optlen)
3576 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3579 /* I can hardly begin to describe how wrong this is. This is
3580 * so broken as to be worse than useless. The API draft
3581 * REALLY is NOT helpful here... I am not convinced that the
3582 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3583 * are at all well-founded.
3585 if (level != SOL_SCTP) {
3586 struct sctp_af *af = sctp_sk(sk)->pf->af;
3587 retval = af->setsockopt(sk, level, optname, optval, optlen);
3594 case SCTP_SOCKOPT_BINDX_ADD:
3595 /* 'optlen' is the size of the addresses buffer. */
3596 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3597 optlen, SCTP_BINDX_ADD_ADDR);
3600 case SCTP_SOCKOPT_BINDX_REM:
3601 /* 'optlen' is the size of the addresses buffer. */
3602 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3603 optlen, SCTP_BINDX_REM_ADDR);
3606 case SCTP_SOCKOPT_CONNECTX_OLD:
3607 /* 'optlen' is the size of the addresses buffer. */
3608 retval = sctp_setsockopt_connectx_old(sk,
3609 (struct sockaddr __user *)optval,
3613 case SCTP_SOCKOPT_CONNECTX:
3614 /* 'optlen' is the size of the addresses buffer. */
3615 retval = sctp_setsockopt_connectx(sk,
3616 (struct sockaddr __user *)optval,
3620 case SCTP_DISABLE_FRAGMENTS:
3621 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3625 retval = sctp_setsockopt_events(sk, optval, optlen);
3628 case SCTP_AUTOCLOSE:
3629 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3632 case SCTP_PEER_ADDR_PARAMS:
3633 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3636 case SCTP_DELAYED_SACK:
3637 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3639 case SCTP_PARTIAL_DELIVERY_POINT:
3640 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3644 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3646 case SCTP_DEFAULT_SEND_PARAM:
3647 retval = sctp_setsockopt_default_send_param(sk, optval,
3650 case SCTP_PRIMARY_ADDR:
3651 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3653 case SCTP_SET_PEER_PRIMARY_ADDR:
3654 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3657 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3660 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3662 case SCTP_ASSOCINFO:
3663 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3665 case SCTP_I_WANT_MAPPED_V4_ADDR:
3666 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3669 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3671 case SCTP_ADAPTATION_LAYER:
3672 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3675 retval = sctp_setsockopt_context(sk, optval, optlen);
3677 case SCTP_FRAGMENT_INTERLEAVE:
3678 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3680 case SCTP_MAX_BURST:
3681 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3683 case SCTP_AUTH_CHUNK:
3684 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3686 case SCTP_HMAC_IDENT:
3687 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3690 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3692 case SCTP_AUTH_ACTIVE_KEY:
3693 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3695 case SCTP_AUTH_DELETE_KEY:
3696 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3698 case SCTP_AUTO_ASCONF:
3699 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3701 case SCTP_PEER_ADDR_THLDS:
3702 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3705 retval = -ENOPROTOOPT;
3709 sctp_release_sock(sk);
3715 /* API 3.1.6 connect() - UDP Style Syntax
3717 * An application may use the connect() call in the UDP model to initiate an
3718 * association without sending data.
3722 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3724 * sd: the socket descriptor to have a new association added to.
3726 * nam: the address structure (either struct sockaddr_in or struct
3727 * sockaddr_in6 defined in RFC2553 [7]).
3729 * len: the size of the address.
3731 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3739 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3740 __func__, sk, addr, addr_len);
3742 /* Validate addr_len before calling common connect/connectx routine. */
3743 af = sctp_get_af_specific(addr->sa_family);
3744 if (!af || addr_len < af->sockaddr_len) {
3747 /* Pass correct addr len to common routine (so it knows there
3748 * is only one address being passed.
3750 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3753 sctp_release_sock(sk);
3757 /* FIXME: Write comments. */
3758 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3760 return -EOPNOTSUPP; /* STUB */
3763 /* 4.1.4 accept() - TCP Style Syntax
3765 * Applications use accept() call to remove an established SCTP
3766 * association from the accept queue of the endpoint. A new socket
3767 * descriptor will be returned from accept() to represent the newly
3768 * formed association.
3770 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3772 struct sctp_sock *sp;
3773 struct sctp_endpoint *ep;
3774 struct sock *newsk = NULL;
3775 struct sctp_association *asoc;
3784 if (!sctp_style(sk, TCP)) {
3785 error = -EOPNOTSUPP;
3789 if (!sctp_sstate(sk, LISTENING)) {
3794 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3796 error = sctp_wait_for_accept(sk, timeo);
3800 /* We treat the list of associations on the endpoint as the accept
3801 * queue and pick the first association on the list.
3803 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3805 newsk = sp->pf->create_accept_sk(sk, asoc);
3811 /* Populate the fields of the newsk from the oldsk and migrate the
3812 * asoc to the newsk.
3814 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3817 sctp_release_sock(sk);
3822 /* The SCTP ioctl handler. */
3823 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3830 * SEQPACKET-style sockets in LISTENING state are valid, for
3831 * SCTP, so only discard TCP-style sockets in LISTENING state.
3833 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3838 struct sk_buff *skb;
3839 unsigned int amount = 0;
3841 skb = skb_peek(&sk->sk_receive_queue);
3844 * We will only return the amount of this packet since
3845 * that is all that will be read.
3849 rc = put_user(amount, (int __user *)arg);
3857 sctp_release_sock(sk);
3861 /* This is the function which gets called during socket creation to
3862 * initialized the SCTP-specific portion of the sock.
3863 * The sock structure should already be zero-filled memory.
3865 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3867 struct net *net = sock_net(sk);
3868 struct sctp_endpoint *ep;
3869 struct sctp_sock *sp;
3871 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3875 /* Initialize the SCTP per socket area. */
3876 switch (sk->sk_type) {
3877 case SOCK_SEQPACKET:
3878 sp->type = SCTP_SOCKET_UDP;
3881 sp->type = SCTP_SOCKET_TCP;
3884 return -ESOCKTNOSUPPORT;
3887 /* Initialize default send parameters. These parameters can be
3888 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3890 sp->default_stream = 0;
3891 sp->default_ppid = 0;
3892 sp->default_flags = 0;
3893 sp->default_context = 0;
3894 sp->default_timetolive = 0;
3896 sp->default_rcv_context = 0;
3897 sp->max_burst = net->sctp.max_burst;
3899 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
3901 /* Initialize default setup parameters. These parameters
3902 * can be modified with the SCTP_INITMSG socket option or
3903 * overridden by the SCTP_INIT CMSG.
3905 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3906 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3907 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
3908 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
3910 /* Initialize default RTO related parameters. These parameters can
3911 * be modified for with the SCTP_RTOINFO socket option.
3913 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
3914 sp->rtoinfo.srto_max = net->sctp.rto_max;
3915 sp->rtoinfo.srto_min = net->sctp.rto_min;
3917 /* Initialize default association related parameters. These parameters
3918 * can be modified with the SCTP_ASSOCINFO socket option.
3920 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
3921 sp->assocparams.sasoc_number_peer_destinations = 0;
3922 sp->assocparams.sasoc_peer_rwnd = 0;
3923 sp->assocparams.sasoc_local_rwnd = 0;
3924 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
3926 /* Initialize default event subscriptions. By default, all the
3929 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3931 /* Default Peer Address Parameters. These defaults can
3932 * be modified via SCTP_PEER_ADDR_PARAMS
3934 sp->hbinterval = net->sctp.hb_interval;
3935 sp->pathmaxrxt = net->sctp.max_retrans_path;
3936 sp->pathmtu = 0; // allow default discovery
3937 sp->sackdelay = net->sctp.sack_timeout;
3939 sp->param_flags = SPP_HB_ENABLE |
3941 SPP_SACKDELAY_ENABLE;
3943 /* If enabled no SCTP message fragmentation will be performed.
3944 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3946 sp->disable_fragments = 0;
3948 /* Enable Nagle algorithm by default. */
3951 /* Enable by default. */
3954 /* Auto-close idle associations after the configured
3955 * number of seconds. A value of 0 disables this
3956 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3957 * for UDP-style sockets only.
3961 /* User specified fragmentation limit. */
3964 sp->adaptation_ind = 0;
3966 sp->pf = sctp_get_pf_specific(sk->sk_family);
3968 /* Control variables for partial data delivery. */
3969 atomic_set(&sp->pd_mode, 0);
3970 skb_queue_head_init(&sp->pd_lobby);
3971 sp->frag_interleave = 0;
3973 /* Create a per socket endpoint structure. Even if we
3974 * change the data structure relationships, this may still
3975 * be useful for storing pre-connect address information.
3977 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3984 SCTP_DBG_OBJCNT_INC(sock);
3987 percpu_counter_inc(&sctp_sockets_allocated);
3988 sock_prot_inuse_add(net, sk->sk_prot, 1);
3989 if (net->sctp.default_auto_asconf) {
3990 list_add_tail(&sp->auto_asconf_list,
3991 &net->sctp.auto_asconf_splist);
3992 sp->do_auto_asconf = 1;
3994 sp->do_auto_asconf = 0;
4000 /* Cleanup any SCTP per socket resources. */
4001 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
4003 struct sctp_sock *sp;
4005 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
4007 /* Release our hold on the endpoint. */
4009 /* This could happen during socket init, thus we bail out
4010 * early, since the rest of the below is not setup either.
4015 if (sp->do_auto_asconf) {
4016 sp->do_auto_asconf = 0;
4017 list_del(&sp->auto_asconf_list);
4019 sctp_endpoint_free(sp->ep);
4021 percpu_counter_dec(&sctp_sockets_allocated);
4022 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4026 /* API 4.1.7 shutdown() - TCP Style Syntax
4027 * int shutdown(int socket, int how);
4029 * sd - the socket descriptor of the association to be closed.
4030 * how - Specifies the type of shutdown. The values are
4033 * Disables further receive operations. No SCTP
4034 * protocol action is taken.
4036 * Disables further send operations, and initiates
4037 * the SCTP shutdown sequence.
4039 * Disables further send and receive operations
4040 * and initiates the SCTP shutdown sequence.
4042 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
4044 struct net *net = sock_net(sk);
4045 struct sctp_endpoint *ep;
4046 struct sctp_association *asoc;
4048 if (!sctp_style(sk, TCP))
4051 if (how & SEND_SHUTDOWN) {
4052 ep = sctp_sk(sk)->ep;
4053 if (!list_empty(&ep->asocs)) {
4054 asoc = list_entry(ep->asocs.next,
4055 struct sctp_association, asocs);
4056 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4061 /* 7.2.1 Association Status (SCTP_STATUS)
4063 * Applications can retrieve current status information about an
4064 * association, including association state, peer receiver window size,
4065 * number of unacked data chunks, and number of data chunks pending
4066 * receipt. This information is read-only.
4068 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4069 char __user *optval,
4072 struct sctp_status status;
4073 struct sctp_association *asoc = NULL;
4074 struct sctp_transport *transport;
4075 sctp_assoc_t associd;
4078 if (len < sizeof(status)) {
4083 len = sizeof(status);
4084 if (copy_from_user(&status, optval, len)) {
4089 associd = status.sstat_assoc_id;
4090 asoc = sctp_id2assoc(sk, associd);
4096 transport = asoc->peer.primary_path;
4098 status.sstat_assoc_id = sctp_assoc2id(asoc);
4099 status.sstat_state = asoc->state;
4100 status.sstat_rwnd = asoc->peer.rwnd;
4101 status.sstat_unackdata = asoc->unack_data;
4103 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4104 status.sstat_instrms = asoc->c.sinit_max_instreams;
4105 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4106 status.sstat_fragmentation_point = asoc->frag_point;
4107 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4108 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4109 transport->af_specific->sockaddr_len);
4110 /* Map ipv4 address into v4-mapped-on-v6 address. */
4111 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4112 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4113 status.sstat_primary.spinfo_state = transport->state;
4114 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4115 status.sstat_primary.spinfo_srtt = transport->srtt;
4116 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4117 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4119 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4120 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4122 if (put_user(len, optlen)) {
4127 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
4128 len, status.sstat_state, status.sstat_rwnd,
4129 status.sstat_assoc_id);
4131 if (copy_to_user(optval, &status, len)) {
4141 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4143 * Applications can retrieve information about a specific peer address
4144 * of an association, including its reachability state, congestion
4145 * window, and retransmission timer values. This information is
4148 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4149 char __user *optval,
4152 struct sctp_paddrinfo pinfo;
4153 struct sctp_transport *transport;
4156 if (len < sizeof(pinfo)) {
4161 len = sizeof(pinfo);
4162 if (copy_from_user(&pinfo, optval, len)) {
4167 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4168 pinfo.spinfo_assoc_id);
4172 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4173 pinfo.spinfo_state = transport->state;
4174 pinfo.spinfo_cwnd = transport->cwnd;
4175 pinfo.spinfo_srtt = transport->srtt;
4176 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4177 pinfo.spinfo_mtu = transport->pathmtu;
4179 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4180 pinfo.spinfo_state = SCTP_ACTIVE;
4182 if (put_user(len, optlen)) {
4187 if (copy_to_user(optval, &pinfo, len)) {
4196 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4198 * This option is a on/off flag. If enabled no SCTP message
4199 * fragmentation will be performed. Instead if a message being sent
4200 * exceeds the current PMTU size, the message will NOT be sent and
4201 * instead a error will be indicated to the user.
4203 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4204 char __user *optval, int __user *optlen)
4208 if (len < sizeof(int))
4212 val = (sctp_sk(sk)->disable_fragments == 1);
4213 if (put_user(len, optlen))
4215 if (copy_to_user(optval, &val, len))
4220 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4222 * This socket option is used to specify various notifications and
4223 * ancillary data the user wishes to receive.
4225 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4230 if (len > sizeof(struct sctp_event_subscribe))
4231 len = sizeof(struct sctp_event_subscribe);
4232 if (put_user(len, optlen))
4234 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4239 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4241 * This socket option is applicable to the UDP-style socket only. When
4242 * set it will cause associations that are idle for more than the
4243 * specified number of seconds to automatically close. An association
4244 * being idle is defined an association that has NOT sent or received
4245 * user data. The special value of '0' indicates that no automatic
4246 * close of any associations should be performed. The option expects an
4247 * integer defining the number of seconds of idle time before an
4248 * association is closed.
4250 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4252 /* Applicable to UDP-style socket only */
4253 if (sctp_style(sk, TCP))
4255 if (len < sizeof(int))
4258 if (put_user(len, optlen))
4260 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4265 /* Helper routine to branch off an association to a new socket. */
4266 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4268 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4269 struct socket *sock;
4276 /* An association cannot be branched off from an already peeled-off
4277 * socket, nor is this supported for tcp style sockets.
4279 if (!sctp_style(sk, UDP))
4282 /* Create a new socket. */
4283 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4287 sctp_copy_sock(sock->sk, sk, asoc);
4289 /* Make peeled-off sockets more like 1-1 accepted sockets.
4290 * Set the daddr and initialize id to something more random
4292 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4293 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4295 /* Populate the fields of the newsk from the oldsk and migrate the
4296 * asoc to the newsk.
4298 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4304 EXPORT_SYMBOL(sctp_do_peeloff);
4306 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4308 sctp_peeloff_arg_t peeloff;
4309 struct socket *newsock;
4310 struct file *newfile;
4313 if (len < sizeof(sctp_peeloff_arg_t))
4315 len = sizeof(sctp_peeloff_arg_t);
4316 if (copy_from_user(&peeloff, optval, len))
4319 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4323 /* Map the socket to an unused fd that can be returned to the user. */
4324 retval = get_unused_fd();
4326 sock_release(newsock);
4330 newfile = sock_alloc_file(newsock, 0, NULL);
4331 if (unlikely(IS_ERR(newfile))) {
4332 put_unused_fd(retval);
4333 sock_release(newsock);
4334 return PTR_ERR(newfile);
4337 SCTP_DEBUG_PRINTK("%s: sk: %p newsk: %p sd: %d\n",
4338 __func__, sk, newsock->sk, retval);
4340 /* Return the fd mapped to the new socket. */
4341 if (put_user(len, optlen)) {
4343 put_unused_fd(retval);
4346 peeloff.sd = retval;
4347 if (copy_to_user(optval, &peeloff, len)) {
4349 put_unused_fd(retval);
4352 fd_install(retval, newfile);
4357 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4359 * Applications can enable or disable heartbeats for any peer address of
4360 * an association, modify an address's heartbeat interval, force a
4361 * heartbeat to be sent immediately, and adjust the address's maximum
4362 * number of retransmissions sent before an address is considered
4363 * unreachable. The following structure is used to access and modify an
4364 * address's parameters:
4366 * struct sctp_paddrparams {
4367 * sctp_assoc_t spp_assoc_id;
4368 * struct sockaddr_storage spp_address;
4369 * uint32_t spp_hbinterval;
4370 * uint16_t spp_pathmaxrxt;
4371 * uint32_t spp_pathmtu;
4372 * uint32_t spp_sackdelay;
4373 * uint32_t spp_flags;
4376 * spp_assoc_id - (one-to-many style socket) This is filled in the
4377 * application, and identifies the association for
4379 * spp_address - This specifies which address is of interest.
4380 * spp_hbinterval - This contains the value of the heartbeat interval,
4381 * in milliseconds. If a value of zero
4382 * is present in this field then no changes are to
4383 * be made to this parameter.
4384 * spp_pathmaxrxt - This contains the maximum number of
4385 * retransmissions before this address shall be
4386 * considered unreachable. If a value of zero
4387 * is present in this field then no changes are to
4388 * be made to this parameter.
4389 * spp_pathmtu - When Path MTU discovery is disabled the value
4390 * specified here will be the "fixed" path mtu.
4391 * Note that if the spp_address field is empty
4392 * then all associations on this address will
4393 * have this fixed path mtu set upon them.
4395 * spp_sackdelay - When delayed sack is enabled, this value specifies
4396 * the number of milliseconds that sacks will be delayed
4397 * for. This value will apply to all addresses of an
4398 * association if the spp_address field is empty. Note
4399 * also, that if delayed sack is enabled and this
4400 * value is set to 0, no change is made to the last
4401 * recorded delayed sack timer value.
4403 * spp_flags - These flags are used to control various features
4404 * on an association. The flag field may contain
4405 * zero or more of the following options.
4407 * SPP_HB_ENABLE - Enable heartbeats on the
4408 * specified address. Note that if the address
4409 * field is empty all addresses for the association
4410 * have heartbeats enabled upon them.
4412 * SPP_HB_DISABLE - Disable heartbeats on the
4413 * speicifed address. Note that if the address
4414 * field is empty all addresses for the association
4415 * will have their heartbeats disabled. Note also
4416 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4417 * mutually exclusive, only one of these two should
4418 * be specified. Enabling both fields will have
4419 * undetermined results.
4421 * SPP_HB_DEMAND - Request a user initiated heartbeat
4422 * to be made immediately.
4424 * SPP_PMTUD_ENABLE - This field will enable PMTU
4425 * discovery upon the specified address. Note that
4426 * if the address feild is empty then all addresses
4427 * on the association are effected.
4429 * SPP_PMTUD_DISABLE - This field will disable PMTU
4430 * discovery upon the specified address. Note that
4431 * if the address feild is empty then all addresses
4432 * on the association are effected. Not also that
4433 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4434 * exclusive. Enabling both will have undetermined
4437 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4438 * on delayed sack. The time specified in spp_sackdelay
4439 * is used to specify the sack delay for this address. Note
4440 * that if spp_address is empty then all addresses will
4441 * enable delayed sack and take on the sack delay
4442 * value specified in spp_sackdelay.
4443 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4444 * off delayed sack. If the spp_address field is blank then
4445 * delayed sack is disabled for the entire association. Note
4446 * also that this field is mutually exclusive to
4447 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4450 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4451 char __user *optval, int __user *optlen)
4453 struct sctp_paddrparams params;
4454 struct sctp_transport *trans = NULL;
4455 struct sctp_association *asoc = NULL;
4456 struct sctp_sock *sp = sctp_sk(sk);
4458 if (len < sizeof(struct sctp_paddrparams))
4460 len = sizeof(struct sctp_paddrparams);
4461 if (copy_from_user(¶ms, optval, len))
4464 /* If an address other than INADDR_ANY is specified, and
4465 * no transport is found, then the request is invalid.
4467 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
4468 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4469 params.spp_assoc_id);
4471 SCTP_DEBUG_PRINTK("Failed no transport\n");
4476 /* Get association, if assoc_id != 0 and the socket is a one
4477 * to many style socket, and an association was not found, then
4478 * the id was invalid.
4480 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4481 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4482 SCTP_DEBUG_PRINTK("Failed no association\n");
4487 /* Fetch transport values. */
4488 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4489 params.spp_pathmtu = trans->pathmtu;
4490 params.spp_pathmaxrxt = trans->pathmaxrxt;
4491 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4493 /*draft-11 doesn't say what to return in spp_flags*/
4494 params.spp_flags = trans->param_flags;
4496 /* Fetch association values. */
4497 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4498 params.spp_pathmtu = asoc->pathmtu;
4499 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4500 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4502 /*draft-11 doesn't say what to return in spp_flags*/
4503 params.spp_flags = asoc->param_flags;
4505 /* Fetch socket values. */
4506 params.spp_hbinterval = sp->hbinterval;
4507 params.spp_pathmtu = sp->pathmtu;
4508 params.spp_sackdelay = sp->sackdelay;
4509 params.spp_pathmaxrxt = sp->pathmaxrxt;
4511 /*draft-11 doesn't say what to return in spp_flags*/
4512 params.spp_flags = sp->param_flags;
4515 if (copy_to_user(optval, ¶ms, len))
4518 if (put_user(len, optlen))
4525 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4527 * This option will effect the way delayed acks are performed. This
4528 * option allows you to get or set the delayed ack time, in
4529 * milliseconds. It also allows changing the delayed ack frequency.
4530 * Changing the frequency to 1 disables the delayed sack algorithm. If
4531 * the assoc_id is 0, then this sets or gets the endpoints default
4532 * values. If the assoc_id field is non-zero, then the set or get
4533 * effects the specified association for the one to many model (the
4534 * assoc_id field is ignored by the one to one model). Note that if
4535 * sack_delay or sack_freq are 0 when setting this option, then the
4536 * current values will remain unchanged.
4538 * struct sctp_sack_info {
4539 * sctp_assoc_t sack_assoc_id;
4540 * uint32_t sack_delay;
4541 * uint32_t sack_freq;
4544 * sack_assoc_id - This parameter, indicates which association the user
4545 * is performing an action upon. Note that if this field's value is
4546 * zero then the endpoints default value is changed (effecting future
4547 * associations only).
4549 * sack_delay - This parameter contains the number of milliseconds that
4550 * the user is requesting the delayed ACK timer be set to. Note that
4551 * this value is defined in the standard to be between 200 and 500
4554 * sack_freq - This parameter contains the number of packets that must
4555 * be received before a sack is sent without waiting for the delay
4556 * timer to expire. The default value for this is 2, setting this
4557 * value to 1 will disable the delayed sack algorithm.
4559 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4560 char __user *optval,
4563 struct sctp_sack_info params;
4564 struct sctp_association *asoc = NULL;
4565 struct sctp_sock *sp = sctp_sk(sk);
4567 if (len >= sizeof(struct sctp_sack_info)) {
4568 len = sizeof(struct sctp_sack_info);
4570 if (copy_from_user(¶ms, optval, len))
4572 } else if (len == sizeof(struct sctp_assoc_value)) {
4573 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4574 pr_warn("Use struct sctp_sack_info instead\n");
4575 if (copy_from_user(¶ms, optval, len))
4580 /* Get association, if sack_assoc_id != 0 and the socket is a one
4581 * to many style socket, and an association was not found, then
4582 * the id was invalid.
4584 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4585 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4589 /* Fetch association values. */
4590 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4591 params.sack_delay = jiffies_to_msecs(
4593 params.sack_freq = asoc->sackfreq;
4596 params.sack_delay = 0;
4597 params.sack_freq = 1;
4600 /* Fetch socket values. */
4601 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4602 params.sack_delay = sp->sackdelay;
4603 params.sack_freq = sp->sackfreq;
4605 params.sack_delay = 0;
4606 params.sack_freq = 1;
4610 if (copy_to_user(optval, ¶ms, len))
4613 if (put_user(len, optlen))
4619 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4621 * Applications can specify protocol parameters for the default association
4622 * initialization. The option name argument to setsockopt() and getsockopt()
4625 * Setting initialization parameters is effective only on an unconnected
4626 * socket (for UDP-style sockets only future associations are effected
4627 * by the change). With TCP-style sockets, this option is inherited by
4628 * sockets derived from a listener socket.
4630 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4632 if (len < sizeof(struct sctp_initmsg))
4634 len = sizeof(struct sctp_initmsg);
4635 if (put_user(len, optlen))
4637 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4643 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4644 char __user *optval, int __user *optlen)
4646 struct sctp_association *asoc;
4648 struct sctp_getaddrs getaddrs;
4649 struct sctp_transport *from;
4651 union sctp_addr temp;
4652 struct sctp_sock *sp = sctp_sk(sk);
4657 if (len < sizeof(struct sctp_getaddrs))
4660 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4663 /* For UDP-style sockets, id specifies the association to query. */
4664 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4668 to = optval + offsetof(struct sctp_getaddrs,addrs);
4669 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4671 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4673 memcpy(&temp, &from->ipaddr, sizeof(temp));
4674 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4675 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4676 if (space_left < addrlen)
4678 if (copy_to_user(to, &temp, addrlen))
4682 space_left -= addrlen;
4685 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4687 bytes_copied = ((char __user *)to) - optval;
4688 if (put_user(bytes_copied, optlen))
4694 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4695 size_t space_left, int *bytes_copied)
4697 struct sctp_sockaddr_entry *addr;
4698 union sctp_addr temp;
4701 struct net *net = sock_net(sk);
4704 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4708 if ((PF_INET == sk->sk_family) &&
4709 (AF_INET6 == addr->a.sa.sa_family))
4711 if ((PF_INET6 == sk->sk_family) &&
4712 inet_v6_ipv6only(sk) &&
4713 (AF_INET == addr->a.sa.sa_family))
4715 memcpy(&temp, &addr->a, sizeof(temp));
4716 if (!temp.v4.sin_port)
4717 temp.v4.sin_port = htons(port);
4719 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4721 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4722 if (space_left < addrlen) {
4726 memcpy(to, &temp, addrlen);
4730 space_left -= addrlen;
4731 *bytes_copied += addrlen;
4739 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4740 char __user *optval, int __user *optlen)
4742 struct sctp_bind_addr *bp;
4743 struct sctp_association *asoc;
4745 struct sctp_getaddrs getaddrs;
4746 struct sctp_sockaddr_entry *addr;
4748 union sctp_addr temp;
4749 struct sctp_sock *sp = sctp_sk(sk);
4753 int bytes_copied = 0;
4757 if (len < sizeof(struct sctp_getaddrs))
4760 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4764 * For UDP-style sockets, id specifies the association to query.
4765 * If the id field is set to the value '0' then the locally bound
4766 * addresses are returned without regard to any particular
4769 if (0 == getaddrs.assoc_id) {
4770 bp = &sctp_sk(sk)->ep->base.bind_addr;
4772 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4775 bp = &asoc->base.bind_addr;
4778 to = optval + offsetof(struct sctp_getaddrs,addrs);
4779 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4781 addrs = kmalloc(space_left, GFP_KERNEL);
4785 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4786 * addresses from the global local address list.
4788 if (sctp_list_single_entry(&bp->address_list)) {
4789 addr = list_entry(bp->address_list.next,
4790 struct sctp_sockaddr_entry, list);
4791 if (sctp_is_any(sk, &addr->a)) {
4792 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4793 space_left, &bytes_copied);
4803 /* Protection on the bound address list is not needed since
4804 * in the socket option context we hold a socket lock and
4805 * thus the bound address list can't change.
4807 list_for_each_entry(addr, &bp->address_list, list) {
4808 memcpy(&temp, &addr->a, sizeof(temp));
4809 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4810 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4811 if (space_left < addrlen) {
4812 err = -ENOMEM; /*fixme: right error?*/
4815 memcpy(buf, &temp, addrlen);
4817 bytes_copied += addrlen;
4819 space_left -= addrlen;
4823 if (copy_to_user(to, addrs, bytes_copied)) {
4827 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4831 if (put_user(bytes_copied, optlen))
4838 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4840 * Requests that the local SCTP stack use the enclosed peer address as
4841 * the association primary. The enclosed address must be one of the
4842 * association peer's addresses.
4844 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4845 char __user *optval, int __user *optlen)
4847 struct sctp_prim prim;
4848 struct sctp_association *asoc;
4849 struct sctp_sock *sp = sctp_sk(sk);
4851 if (len < sizeof(struct sctp_prim))
4854 len = sizeof(struct sctp_prim);
4856 if (copy_from_user(&prim, optval, len))
4859 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4863 if (!asoc->peer.primary_path)
4866 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4867 asoc->peer.primary_path->af_specific->sockaddr_len);
4869 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4870 (union sctp_addr *)&prim.ssp_addr);
4872 if (put_user(len, optlen))
4874 if (copy_to_user(optval, &prim, len))
4881 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4883 * Requests that the local endpoint set the specified Adaptation Layer
4884 * Indication parameter for all future INIT and INIT-ACK exchanges.
4886 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4887 char __user *optval, int __user *optlen)
4889 struct sctp_setadaptation adaptation;
4891 if (len < sizeof(struct sctp_setadaptation))
4894 len = sizeof(struct sctp_setadaptation);
4896 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4898 if (put_user(len, optlen))
4900 if (copy_to_user(optval, &adaptation, len))
4908 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4910 * Applications that wish to use the sendto() system call may wish to
4911 * specify a default set of parameters that would normally be supplied
4912 * through the inclusion of ancillary data. This socket option allows
4913 * such an application to set the default sctp_sndrcvinfo structure.
4916 * The application that wishes to use this socket option simply passes
4917 * in to this call the sctp_sndrcvinfo structure defined in Section
4918 * 5.2.2) The input parameters accepted by this call include
4919 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4920 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4921 * to this call if the caller is using the UDP model.
4923 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4925 static int sctp_getsockopt_default_send_param(struct sock *sk,
4926 int len, char __user *optval,
4929 struct sctp_sndrcvinfo info;
4930 struct sctp_association *asoc;
4931 struct sctp_sock *sp = sctp_sk(sk);
4933 if (len < sizeof(struct sctp_sndrcvinfo))
4936 len = sizeof(struct sctp_sndrcvinfo);
4938 if (copy_from_user(&info, optval, len))
4941 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4942 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4946 info.sinfo_stream = asoc->default_stream;
4947 info.sinfo_flags = asoc->default_flags;
4948 info.sinfo_ppid = asoc->default_ppid;
4949 info.sinfo_context = asoc->default_context;
4950 info.sinfo_timetolive = asoc->default_timetolive;
4952 info.sinfo_stream = sp->default_stream;
4953 info.sinfo_flags = sp->default_flags;
4954 info.sinfo_ppid = sp->default_ppid;
4955 info.sinfo_context = sp->default_context;
4956 info.sinfo_timetolive = sp->default_timetolive;
4959 if (put_user(len, optlen))
4961 if (copy_to_user(optval, &info, len))
4969 * 7.1.5 SCTP_NODELAY
4971 * Turn on/off any Nagle-like algorithm. This means that packets are
4972 * generally sent as soon as possible and no unnecessary delays are
4973 * introduced, at the cost of more packets in the network. Expects an
4974 * integer boolean flag.
4977 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4978 char __user *optval, int __user *optlen)
4982 if (len < sizeof(int))
4986 val = (sctp_sk(sk)->nodelay == 1);
4987 if (put_user(len, optlen))
4989 if (copy_to_user(optval, &val, len))
4996 * 7.1.1 SCTP_RTOINFO
4998 * The protocol parameters used to initialize and bound retransmission
4999 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5000 * and modify these parameters.
5001 * All parameters are time values, in milliseconds. A value of 0, when
5002 * modifying the parameters, indicates that the current value should not
5006 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5007 char __user *optval,
5008 int __user *optlen) {
5009 struct sctp_rtoinfo rtoinfo;
5010 struct sctp_association *asoc;
5012 if (len < sizeof (struct sctp_rtoinfo))
5015 len = sizeof(struct sctp_rtoinfo);
5017 if (copy_from_user(&rtoinfo, optval, len))
5020 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5022 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5025 /* Values corresponding to the specific association. */
5027 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5028 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5029 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5031 /* Values corresponding to the endpoint. */
5032 struct sctp_sock *sp = sctp_sk(sk);
5034 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5035 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5036 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5039 if (put_user(len, optlen))
5042 if (copy_to_user(optval, &rtoinfo, len))
5050 * 7.1.2 SCTP_ASSOCINFO
5052 * This option is used to tune the maximum retransmission attempts
5053 * of the association.
5054 * Returns an error if the new association retransmission value is
5055 * greater than the sum of the retransmission value of the peer.
5056 * See [SCTP] for more information.
5059 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5060 char __user *optval,
5064 struct sctp_assocparams assocparams;
5065 struct sctp_association *asoc;
5066 struct list_head *pos;
5069 if (len < sizeof (struct sctp_assocparams))
5072 len = sizeof(struct sctp_assocparams);
5074 if (copy_from_user(&assocparams, optval, len))
5077 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5079 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5082 /* Values correspoinding to the specific association */
5084 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5085 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5086 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5087 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
5089 (asoc->cookie_life.tv_usec
5092 list_for_each(pos, &asoc->peer.transport_addr_list) {
5096 assocparams.sasoc_number_peer_destinations = cnt;
5098 /* Values corresponding to the endpoint */
5099 struct sctp_sock *sp = sctp_sk(sk);
5101 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5102 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5103 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5104 assocparams.sasoc_cookie_life =
5105 sp->assocparams.sasoc_cookie_life;
5106 assocparams.sasoc_number_peer_destinations =
5108 sasoc_number_peer_destinations;
5111 if (put_user(len, optlen))
5114 if (copy_to_user(optval, &assocparams, len))
5121 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5123 * This socket option is a boolean flag which turns on or off mapped V4
5124 * addresses. If this option is turned on and the socket is type
5125 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5126 * If this option is turned off, then no mapping will be done of V4
5127 * addresses and a user will receive both PF_INET6 and PF_INET type
5128 * addresses on the socket.
5130 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5131 char __user *optval, int __user *optlen)
5134 struct sctp_sock *sp = sctp_sk(sk);
5136 if (len < sizeof(int))
5141 if (put_user(len, optlen))
5143 if (copy_to_user(optval, &val, len))
5150 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5151 * (chapter and verse is quoted at sctp_setsockopt_context())
5153 static int sctp_getsockopt_context(struct sock *sk, int len,
5154 char __user *optval, int __user *optlen)
5156 struct sctp_assoc_value params;
5157 struct sctp_sock *sp;
5158 struct sctp_association *asoc;
5160 if (len < sizeof(struct sctp_assoc_value))
5163 len = sizeof(struct sctp_assoc_value);
5165 if (copy_from_user(¶ms, optval, len))
5170 if (params.assoc_id != 0) {
5171 asoc = sctp_id2assoc(sk, params.assoc_id);
5174 params.assoc_value = asoc->default_rcv_context;
5176 params.assoc_value = sp->default_rcv_context;
5179 if (put_user(len, optlen))
5181 if (copy_to_user(optval, ¶ms, len))
5188 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5189 * This option will get or set the maximum size to put in any outgoing
5190 * SCTP DATA chunk. If a message is larger than this size it will be
5191 * fragmented by SCTP into the specified size. Note that the underlying
5192 * SCTP implementation may fragment into smaller sized chunks when the
5193 * PMTU of the underlying association is smaller than the value set by
5194 * the user. The default value for this option is '0' which indicates
5195 * the user is NOT limiting fragmentation and only the PMTU will effect
5196 * SCTP's choice of DATA chunk size. Note also that values set larger
5197 * than the maximum size of an IP datagram will effectively let SCTP
5198 * control fragmentation (i.e. the same as setting this option to 0).
5200 * The following structure is used to access and modify this parameter:
5202 * struct sctp_assoc_value {
5203 * sctp_assoc_t assoc_id;
5204 * uint32_t assoc_value;
5207 * assoc_id: This parameter is ignored for one-to-one style sockets.
5208 * For one-to-many style sockets this parameter indicates which
5209 * association the user is performing an action upon. Note that if
5210 * this field's value is zero then the endpoints default value is
5211 * changed (effecting future associations only).
5212 * assoc_value: This parameter specifies the maximum size in bytes.
5214 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5215 char __user *optval, int __user *optlen)
5217 struct sctp_assoc_value params;
5218 struct sctp_association *asoc;
5220 if (len == sizeof(int)) {
5221 pr_warn("Use of int in maxseg socket option deprecated\n");
5222 pr_warn("Use struct sctp_assoc_value instead\n");
5223 params.assoc_id = 0;
5224 } else if (len >= sizeof(struct sctp_assoc_value)) {
5225 len = sizeof(struct sctp_assoc_value);
5226 if (copy_from_user(¶ms, optval, sizeof(params)))
5231 asoc = sctp_id2assoc(sk, params.assoc_id);
5232 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5236 params.assoc_value = asoc->frag_point;
5238 params.assoc_value = sctp_sk(sk)->user_frag;
5240 if (put_user(len, optlen))
5242 if (len == sizeof(int)) {
5243 if (copy_to_user(optval, ¶ms.assoc_value, len))
5246 if (copy_to_user(optval, ¶ms, len))
5254 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5255 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5257 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5258 char __user *optval, int __user *optlen)
5262 if (len < sizeof(int))
5267 val = sctp_sk(sk)->frag_interleave;
5268 if (put_user(len, optlen))
5270 if (copy_to_user(optval, &val, len))
5277 * 7.1.25. Set or Get the sctp partial delivery point
5278 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5280 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5281 char __user *optval,
5286 if (len < sizeof(u32))
5291 val = sctp_sk(sk)->pd_point;
5292 if (put_user(len, optlen))
5294 if (copy_to_user(optval, &val, len))
5301 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5302 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5304 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5305 char __user *optval,
5308 struct sctp_assoc_value params;
5309 struct sctp_sock *sp;
5310 struct sctp_association *asoc;
5312 if (len == sizeof(int)) {
5313 pr_warn("Use of int in max_burst socket option deprecated\n");
5314 pr_warn("Use struct sctp_assoc_value instead\n");
5315 params.assoc_id = 0;
5316 } else if (len >= sizeof(struct sctp_assoc_value)) {
5317 len = sizeof(struct sctp_assoc_value);
5318 if (copy_from_user(¶ms, optval, len))
5325 if (params.assoc_id != 0) {
5326 asoc = sctp_id2assoc(sk, params.assoc_id);
5329 params.assoc_value = asoc->max_burst;
5331 params.assoc_value = sp->max_burst;
5333 if (len == sizeof(int)) {
5334 if (copy_to_user(optval, ¶ms.assoc_value, len))
5337 if (copy_to_user(optval, ¶ms, len))
5345 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5346 char __user *optval, int __user *optlen)
5348 struct net *net = sock_net(sk);
5349 struct sctp_hmacalgo __user *p = (void __user *)optval;
5350 struct sctp_hmac_algo_param *hmacs;
5354 if (!net->sctp.auth_enable)
5357 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5358 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5360 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5363 len = sizeof(struct sctp_hmacalgo) + data_len;
5364 num_idents = data_len / sizeof(u16);
5366 if (put_user(len, optlen))
5368 if (put_user(num_idents, &p->shmac_num_idents))
5370 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5375 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5376 char __user *optval, int __user *optlen)
5378 struct net *net = sock_net(sk);
5379 struct sctp_authkeyid val;
5380 struct sctp_association *asoc;
5382 if (!net->sctp.auth_enable)
5385 if (len < sizeof(struct sctp_authkeyid))
5387 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5390 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5391 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5395 val.scact_keynumber = asoc->active_key_id;
5397 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5399 len = sizeof(struct sctp_authkeyid);
5400 if (put_user(len, optlen))
5402 if (copy_to_user(optval, &val, len))
5408 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5409 char __user *optval, int __user *optlen)
5411 struct net *net = sock_net(sk);
5412 struct sctp_authchunks __user *p = (void __user *)optval;
5413 struct sctp_authchunks val;
5414 struct sctp_association *asoc;
5415 struct sctp_chunks_param *ch;
5419 if (!net->sctp.auth_enable)
5422 if (len < sizeof(struct sctp_authchunks))
5425 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5428 to = p->gauth_chunks;
5429 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5433 ch = asoc->peer.peer_chunks;
5437 /* See if the user provided enough room for all the data */
5438 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5439 if (len < num_chunks)
5442 if (copy_to_user(to, ch->chunks, num_chunks))
5445 len = sizeof(struct sctp_authchunks) + num_chunks;
5446 if (put_user(len, optlen)) return -EFAULT;
5447 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5452 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5453 char __user *optval, int __user *optlen)
5455 struct net *net = sock_net(sk);
5456 struct sctp_authchunks __user *p = (void __user *)optval;
5457 struct sctp_authchunks val;
5458 struct sctp_association *asoc;
5459 struct sctp_chunks_param *ch;
5463 if (!net->sctp.auth_enable)
5466 if (len < sizeof(struct sctp_authchunks))
5469 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5472 to = p->gauth_chunks;
5473 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5474 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5478 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5480 ch = sctp_sk(sk)->ep->auth_chunk_list;
5485 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5486 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5489 if (copy_to_user(to, ch->chunks, num_chunks))
5492 len = sizeof(struct sctp_authchunks) + num_chunks;
5493 if (put_user(len, optlen))
5495 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5502 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5503 * This option gets the current number of associations that are attached
5504 * to a one-to-many style socket. The option value is an uint32_t.
5506 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5507 char __user *optval, int __user *optlen)
5509 struct sctp_sock *sp = sctp_sk(sk);
5510 struct sctp_association *asoc;
5513 if (sctp_style(sk, TCP))
5516 if (len < sizeof(u32))
5521 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5525 if (put_user(len, optlen))
5527 if (copy_to_user(optval, &val, len))
5534 * 8.1.23 SCTP_AUTO_ASCONF
5535 * See the corresponding setsockopt entry as description
5537 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5538 char __user *optval, int __user *optlen)
5542 if (len < sizeof(int))
5546 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5548 if (put_user(len, optlen))
5550 if (copy_to_user(optval, &val, len))
5556 * 8.2.6. Get the Current Identifiers of Associations
5557 * (SCTP_GET_ASSOC_ID_LIST)
5559 * This option gets the current list of SCTP association identifiers of
5560 * the SCTP associations handled by a one-to-many style socket.
5562 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5563 char __user *optval, int __user *optlen)
5565 struct sctp_sock *sp = sctp_sk(sk);
5566 struct sctp_association *asoc;
5567 struct sctp_assoc_ids *ids;
5570 if (sctp_style(sk, TCP))
5573 if (len < sizeof(struct sctp_assoc_ids))
5576 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5580 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5583 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5585 ids = kmalloc(len, GFP_KERNEL);
5589 ids->gaids_number_of_ids = num;
5591 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5592 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5595 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5605 * SCTP_PEER_ADDR_THLDS
5607 * This option allows us to fetch the partially failed threshold for one or all
5608 * transports in an association. See Section 6.1 of:
5609 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5611 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5612 char __user *optval,
5616 struct sctp_paddrthlds val;
5617 struct sctp_transport *trans;
5618 struct sctp_association *asoc;
5620 if (len < sizeof(struct sctp_paddrthlds))
5622 len = sizeof(struct sctp_paddrthlds);
5623 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5626 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5627 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5631 val.spt_pathpfthld = asoc->pf_retrans;
5632 val.spt_pathmaxrxt = asoc->pathmaxrxt;
5634 trans = sctp_addr_id2transport(sk, &val.spt_address,
5639 val.spt_pathmaxrxt = trans->pathmaxrxt;
5640 val.spt_pathpfthld = trans->pf_retrans;
5643 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5650 * SCTP_GET_ASSOC_STATS
5652 * This option retrieves local per endpoint statistics. It is modeled
5653 * after OpenSolaris' implementation
5655 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
5656 char __user *optval,
5659 struct sctp_assoc_stats sas;
5660 struct sctp_association *asoc = NULL;
5662 /* User must provide at least the assoc id */
5663 if (len < sizeof(sctp_assoc_t))
5666 /* Allow the struct to grow and fill in as much as possible */
5667 len = min_t(size_t, len, sizeof(sas));
5669 if (copy_from_user(&sas, optval, len))
5672 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
5676 sas.sas_rtxchunks = asoc->stats.rtxchunks;
5677 sas.sas_gapcnt = asoc->stats.gapcnt;
5678 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
5679 sas.sas_osacks = asoc->stats.osacks;
5680 sas.sas_isacks = asoc->stats.isacks;
5681 sas.sas_octrlchunks = asoc->stats.octrlchunks;
5682 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
5683 sas.sas_oodchunks = asoc->stats.oodchunks;
5684 sas.sas_iodchunks = asoc->stats.iodchunks;
5685 sas.sas_ouodchunks = asoc->stats.ouodchunks;
5686 sas.sas_iuodchunks = asoc->stats.iuodchunks;
5687 sas.sas_idupchunks = asoc->stats.idupchunks;
5688 sas.sas_opackets = asoc->stats.opackets;
5689 sas.sas_ipackets = asoc->stats.ipackets;
5691 /* New high max rto observed, will return 0 if not a single
5692 * RTO update took place. obs_rto_ipaddr will be bogus
5695 sas.sas_maxrto = asoc->stats.max_obs_rto;
5696 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
5697 sizeof(struct sockaddr_storage));
5699 /* Mark beginning of a new observation period */
5700 asoc->stats.max_obs_rto = asoc->rto_min;
5702 if (put_user(len, optlen))
5705 SCTP_DEBUG_PRINTK("sctp_getsockopt_assoc_stat(%d): %d\n",
5706 len, sas.sas_assoc_id);
5708 if (copy_to_user(optval, &sas, len))
5714 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5715 char __user *optval, int __user *optlen)
5720 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5723 /* I can hardly begin to describe how wrong this is. This is
5724 * so broken as to be worse than useless. The API draft
5725 * REALLY is NOT helpful here... I am not convinced that the
5726 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5727 * are at all well-founded.
5729 if (level != SOL_SCTP) {
5730 struct sctp_af *af = sctp_sk(sk)->pf->af;
5732 retval = af->getsockopt(sk, level, optname, optval, optlen);
5736 if (get_user(len, optlen))
5743 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5745 case SCTP_DISABLE_FRAGMENTS:
5746 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5750 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5752 case SCTP_AUTOCLOSE:
5753 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5755 case SCTP_SOCKOPT_PEELOFF:
5756 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5758 case SCTP_PEER_ADDR_PARAMS:
5759 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5762 case SCTP_DELAYED_SACK:
5763 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5767 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5769 case SCTP_GET_PEER_ADDRS:
5770 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5773 case SCTP_GET_LOCAL_ADDRS:
5774 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5777 case SCTP_SOCKOPT_CONNECTX3:
5778 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5780 case SCTP_DEFAULT_SEND_PARAM:
5781 retval = sctp_getsockopt_default_send_param(sk, len,
5784 case SCTP_PRIMARY_ADDR:
5785 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5788 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5791 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5793 case SCTP_ASSOCINFO:
5794 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5796 case SCTP_I_WANT_MAPPED_V4_ADDR:
5797 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5800 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5802 case SCTP_GET_PEER_ADDR_INFO:
5803 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5806 case SCTP_ADAPTATION_LAYER:
5807 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5811 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5813 case SCTP_FRAGMENT_INTERLEAVE:
5814 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5817 case SCTP_PARTIAL_DELIVERY_POINT:
5818 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5821 case SCTP_MAX_BURST:
5822 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5825 case SCTP_AUTH_CHUNK:
5826 case SCTP_AUTH_DELETE_KEY:
5827 retval = -EOPNOTSUPP;
5829 case SCTP_HMAC_IDENT:
5830 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5832 case SCTP_AUTH_ACTIVE_KEY:
5833 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5835 case SCTP_PEER_AUTH_CHUNKS:
5836 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5839 case SCTP_LOCAL_AUTH_CHUNKS:
5840 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5843 case SCTP_GET_ASSOC_NUMBER:
5844 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5846 case SCTP_GET_ASSOC_ID_LIST:
5847 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5849 case SCTP_AUTO_ASCONF:
5850 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5852 case SCTP_PEER_ADDR_THLDS:
5853 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
5855 case SCTP_GET_ASSOC_STATS:
5856 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
5859 retval = -ENOPROTOOPT;
5863 sctp_release_sock(sk);
5867 static void sctp_hash(struct sock *sk)
5872 static void sctp_unhash(struct sock *sk)
5877 /* Check if port is acceptable. Possibly find first available port.
5879 * The port hash table (contained in the 'global' SCTP protocol storage
5880 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5881 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5882 * list (the list number is the port number hashed out, so as you
5883 * would expect from a hash function, all the ports in a given list have
5884 * such a number that hashes out to the same list number; you were
5885 * expecting that, right?); so each list has a set of ports, with a
5886 * link to the socket (struct sock) that uses it, the port number and
5887 * a fastreuse flag (FIXME: NPI ipg).
5889 static struct sctp_bind_bucket *sctp_bucket_create(
5890 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
5892 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5894 struct sctp_bind_hashbucket *head; /* hash list */
5895 struct sctp_bind_bucket *pp;
5896 unsigned short snum;
5899 snum = ntohs(addr->v4.sin_port);
5901 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5902 sctp_local_bh_disable();
5905 /* Search for an available port. */
5906 int low, high, remaining, index;
5909 inet_get_local_port_range(&low, &high);
5910 remaining = (high - low) + 1;
5911 rover = net_random() % remaining + low;
5915 if ((rover < low) || (rover > high))
5917 if (inet_is_reserved_local_port(rover))
5919 index = sctp_phashfn(sock_net(sk), rover);
5920 head = &sctp_port_hashtable[index];
5921 sctp_spin_lock(&head->lock);
5922 sctp_for_each_hentry(pp, &head->chain)
5923 if ((pp->port == rover) &&
5924 net_eq(sock_net(sk), pp->net))
5928 sctp_spin_unlock(&head->lock);
5929 } while (--remaining > 0);
5931 /* Exhausted local port range during search? */
5936 /* OK, here is the one we will use. HEAD (the port
5937 * hash table list entry) is non-NULL and we hold it's
5942 /* We are given an specific port number; we verify
5943 * that it is not being used. If it is used, we will
5944 * exahust the search in the hash list corresponding
5945 * to the port number (snum) - we detect that with the
5946 * port iterator, pp being NULL.
5948 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
5949 sctp_spin_lock(&head->lock);
5950 sctp_for_each_hentry(pp, &head->chain) {
5951 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
5958 if (!hlist_empty(&pp->owner)) {
5959 /* We had a port hash table hit - there is an
5960 * available port (pp != NULL) and it is being
5961 * used by other socket (pp->owner not empty); that other
5962 * socket is going to be sk2.
5964 int reuse = sk->sk_reuse;
5967 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5968 if (pp->fastreuse && sk->sk_reuse &&
5969 sk->sk_state != SCTP_SS_LISTENING)
5972 /* Run through the list of sockets bound to the port
5973 * (pp->port) [via the pointers bind_next and
5974 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5975 * we get the endpoint they describe and run through
5976 * the endpoint's list of IP (v4 or v6) addresses,
5977 * comparing each of the addresses with the address of
5978 * the socket sk. If we find a match, then that means
5979 * that this port/socket (sk) combination are already
5982 sk_for_each_bound(sk2, &pp->owner) {
5983 struct sctp_endpoint *ep2;
5984 ep2 = sctp_sk(sk2)->ep;
5987 (reuse && sk2->sk_reuse &&
5988 sk2->sk_state != SCTP_SS_LISTENING))
5991 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5992 sctp_sk(sk2), sctp_sk(sk))) {
5997 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
6000 /* If there was a hash table miss, create a new port. */
6002 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6005 /* In either case (hit or miss), make sure fastreuse is 1 only
6006 * if sk->sk_reuse is too (that is, if the caller requested
6007 * SO_REUSEADDR on this socket -sk-).
6009 if (hlist_empty(&pp->owner)) {
6010 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6014 } else if (pp->fastreuse &&
6015 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6018 /* We are set, so fill up all the data in the hash table
6019 * entry, tie the socket list information with the rest of the
6020 * sockets FIXME: Blurry, NPI (ipg).
6023 if (!sctp_sk(sk)->bind_hash) {
6024 inet_sk(sk)->inet_num = snum;
6025 sk_add_bind_node(sk, &pp->owner);
6026 sctp_sk(sk)->bind_hash = pp;
6031 sctp_spin_unlock(&head->lock);
6034 sctp_local_bh_enable();
6038 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6039 * port is requested.
6041 static int sctp_get_port(struct sock *sk, unsigned short snum)
6044 union sctp_addr addr;
6045 struct sctp_af *af = sctp_sk(sk)->pf->af;
6047 /* Set up a dummy address struct from the sk. */
6048 af->from_sk(&addr, sk);
6049 addr.v4.sin_port = htons(snum);
6051 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6052 ret = sctp_get_port_local(sk, &addr);
6058 * Move a socket to LISTENING state.
6060 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
6062 struct sctp_sock *sp = sctp_sk(sk);
6063 struct sctp_endpoint *ep = sp->ep;
6064 struct crypto_hash *tfm = NULL;
6067 /* Allocate HMAC for generating cookie. */
6068 if (!sp->hmac && sp->sctp_hmac_alg) {
6069 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6070 tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
6072 net_info_ratelimited("failed to load transform for %s: %ld\n",
6073 sp->sctp_hmac_alg, PTR_ERR(tfm));
6076 sctp_sk(sk)->hmac = tfm;
6080 * If a bind() or sctp_bindx() is not called prior to a listen()
6081 * call that allows new associations to be accepted, the system
6082 * picks an ephemeral port and will choose an address set equivalent
6083 * to binding with a wildcard address.
6085 * This is not currently spelled out in the SCTP sockets
6086 * extensions draft, but follows the practice as seen in TCP
6090 sk->sk_state = SCTP_SS_LISTENING;
6091 if (!ep->base.bind_addr.port) {
6092 if (sctp_autobind(sk))
6095 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6096 sk->sk_state = SCTP_SS_CLOSED;
6101 sk->sk_max_ack_backlog = backlog;
6102 sctp_hash_endpoint(ep);
6107 * 4.1.3 / 5.1.3 listen()
6109 * By default, new associations are not accepted for UDP style sockets.
6110 * An application uses listen() to mark a socket as being able to
6111 * accept new associations.
6113 * On TCP style sockets, applications use listen() to ready the SCTP
6114 * endpoint for accepting inbound associations.
6116 * On both types of endpoints a backlog of '0' disables listening.
6118 * Move a socket to LISTENING state.
6120 int sctp_inet_listen(struct socket *sock, int backlog)
6122 struct sock *sk = sock->sk;
6123 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6126 if (unlikely(backlog < 0))
6131 /* Peeled-off sockets are not allowed to listen(). */
6132 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6135 if (sock->state != SS_UNCONNECTED)
6138 /* If backlog is zero, disable listening. */
6140 if (sctp_sstate(sk, CLOSED))
6144 sctp_unhash_endpoint(ep);
6145 sk->sk_state = SCTP_SS_CLOSED;
6147 sctp_sk(sk)->bind_hash->fastreuse = 1;
6151 /* If we are already listening, just update the backlog */
6152 if (sctp_sstate(sk, LISTENING))
6153 sk->sk_max_ack_backlog = backlog;
6155 err = sctp_listen_start(sk, backlog);
6162 sctp_release_sock(sk);
6167 * This function is done by modeling the current datagram_poll() and the
6168 * tcp_poll(). Note that, based on these implementations, we don't
6169 * lock the socket in this function, even though it seems that,
6170 * ideally, locking or some other mechanisms can be used to ensure
6171 * the integrity of the counters (sndbuf and wmem_alloc) used
6172 * in this place. We assume that we don't need locks either until proven
6175 * Another thing to note is that we include the Async I/O support
6176 * here, again, by modeling the current TCP/UDP code. We don't have
6177 * a good way to test with it yet.
6179 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6181 struct sock *sk = sock->sk;
6182 struct sctp_sock *sp = sctp_sk(sk);
6185 poll_wait(file, sk_sleep(sk), wait);
6187 /* A TCP-style listening socket becomes readable when the accept queue
6190 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6191 return (!list_empty(&sp->ep->asocs)) ?
6192 (POLLIN | POLLRDNORM) : 0;
6196 /* Is there any exceptional events? */
6197 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6199 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6200 if (sk->sk_shutdown & RCV_SHUTDOWN)
6201 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6202 if (sk->sk_shutdown == SHUTDOWN_MASK)
6205 /* Is it readable? Reconsider this code with TCP-style support. */
6206 if (!skb_queue_empty(&sk->sk_receive_queue))
6207 mask |= POLLIN | POLLRDNORM;
6209 /* The association is either gone or not ready. */
6210 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6213 /* Is it writable? */
6214 if (sctp_writeable(sk)) {
6215 mask |= POLLOUT | POLLWRNORM;
6217 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6219 * Since the socket is not locked, the buffer
6220 * might be made available after the writeable check and
6221 * before the bit is set. This could cause a lost I/O
6222 * signal. tcp_poll() has a race breaker for this race
6223 * condition. Based on their implementation, we put
6224 * in the following code to cover it as well.
6226 if (sctp_writeable(sk))
6227 mask |= POLLOUT | POLLWRNORM;
6232 /********************************************************************
6233 * 2nd Level Abstractions
6234 ********************************************************************/
6236 static struct sctp_bind_bucket *sctp_bucket_create(
6237 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6239 struct sctp_bind_bucket *pp;
6241 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6243 SCTP_DBG_OBJCNT_INC(bind_bucket);
6246 INIT_HLIST_HEAD(&pp->owner);
6248 hlist_add_head(&pp->node, &head->chain);
6253 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6254 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6256 if (pp && hlist_empty(&pp->owner)) {
6257 __hlist_del(&pp->node);
6258 kmem_cache_free(sctp_bucket_cachep, pp);
6259 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6263 /* Release this socket's reference to a local port. */
6264 static inline void __sctp_put_port(struct sock *sk)
6266 struct sctp_bind_hashbucket *head =
6267 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6268 inet_sk(sk)->inet_num)];
6269 struct sctp_bind_bucket *pp;
6271 sctp_spin_lock(&head->lock);
6272 pp = sctp_sk(sk)->bind_hash;
6273 __sk_del_bind_node(sk);
6274 sctp_sk(sk)->bind_hash = NULL;
6275 inet_sk(sk)->inet_num = 0;
6276 sctp_bucket_destroy(pp);
6277 sctp_spin_unlock(&head->lock);
6280 void sctp_put_port(struct sock *sk)
6282 sctp_local_bh_disable();
6283 __sctp_put_port(sk);
6284 sctp_local_bh_enable();
6288 * The system picks an ephemeral port and choose an address set equivalent
6289 * to binding with a wildcard address.
6290 * One of those addresses will be the primary address for the association.
6291 * This automatically enables the multihoming capability of SCTP.
6293 static int sctp_autobind(struct sock *sk)
6295 union sctp_addr autoaddr;
6299 /* Initialize a local sockaddr structure to INADDR_ANY. */
6300 af = sctp_sk(sk)->pf->af;
6302 port = htons(inet_sk(sk)->inet_num);
6303 af->inaddr_any(&autoaddr, port);
6305 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6308 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6311 * 4.2 The cmsghdr Structure *
6313 * When ancillary data is sent or received, any number of ancillary data
6314 * objects can be specified by the msg_control and msg_controllen members of
6315 * the msghdr structure, because each object is preceded by
6316 * a cmsghdr structure defining the object's length (the cmsg_len member).
6317 * Historically Berkeley-derived implementations have passed only one object
6318 * at a time, but this API allows multiple objects to be
6319 * passed in a single call to sendmsg() or recvmsg(). The following example
6320 * shows two ancillary data objects in a control buffer.
6322 * |<--------------------------- msg_controllen -------------------------->|
6325 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6327 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6330 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6332 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6335 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6336 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6338 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6340 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6347 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
6348 sctp_cmsgs_t *cmsgs)
6350 struct cmsghdr *cmsg;
6351 struct msghdr *my_msg = (struct msghdr *)msg;
6353 for (cmsg = CMSG_FIRSTHDR(msg);
6355 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6356 if (!CMSG_OK(my_msg, cmsg))
6359 /* Should we parse this header or ignore? */
6360 if (cmsg->cmsg_level != IPPROTO_SCTP)
6363 /* Strictly check lengths following example in SCM code. */
6364 switch (cmsg->cmsg_type) {
6366 /* SCTP Socket API Extension
6367 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6369 * This cmsghdr structure provides information for
6370 * initializing new SCTP associations with sendmsg().
6371 * The SCTP_INITMSG socket option uses this same data
6372 * structure. This structure is not used for
6375 * cmsg_level cmsg_type cmsg_data[]
6376 * ------------ ------------ ----------------------
6377 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6379 if (cmsg->cmsg_len !=
6380 CMSG_LEN(sizeof(struct sctp_initmsg)))
6382 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6386 /* SCTP Socket API Extension
6387 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6389 * This cmsghdr structure specifies SCTP options for
6390 * sendmsg() and describes SCTP header information
6391 * about a received message through recvmsg().
6393 * cmsg_level cmsg_type cmsg_data[]
6394 * ------------ ------------ ----------------------
6395 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6397 if (cmsg->cmsg_len !=
6398 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6402 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6404 /* Minimally, validate the sinfo_flags. */
6405 if (cmsgs->info->sinfo_flags &
6406 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6407 SCTP_ABORT | SCTP_EOF))
6419 * Wait for a packet..
6420 * Note: This function is the same function as in core/datagram.c
6421 * with a few modifications to make lksctp work.
6423 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6428 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6430 /* Socket errors? */
6431 error = sock_error(sk);
6435 if (!skb_queue_empty(&sk->sk_receive_queue))
6438 /* Socket shut down? */
6439 if (sk->sk_shutdown & RCV_SHUTDOWN)
6442 /* Sequenced packets can come disconnected. If so we report the
6447 /* Is there a good reason to think that we may receive some data? */
6448 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6451 /* Handle signals. */
6452 if (signal_pending(current))
6455 /* Let another process have a go. Since we are going to sleep
6456 * anyway. Note: This may cause odd behaviors if the message
6457 * does not fit in the user's buffer, but this seems to be the
6458 * only way to honor MSG_DONTWAIT realistically.
6460 sctp_release_sock(sk);
6461 *timeo_p = schedule_timeout(*timeo_p);
6465 finish_wait(sk_sleep(sk), &wait);
6469 error = sock_intr_errno(*timeo_p);
6472 finish_wait(sk_sleep(sk), &wait);
6477 /* Receive a datagram.
6478 * Note: This is pretty much the same routine as in core/datagram.c
6479 * with a few changes to make lksctp work.
6481 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6482 int noblock, int *err)
6485 struct sk_buff *skb;
6488 timeo = sock_rcvtimeo(sk, noblock);
6490 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6491 timeo, MAX_SCHEDULE_TIMEOUT);
6494 /* Again only user level code calls this function,
6495 * so nothing interrupt level
6496 * will suddenly eat the receive_queue.
6498 * Look at current nfs client by the way...
6499 * However, this function was correct in any case. 8)
6501 if (flags & MSG_PEEK) {
6502 spin_lock_bh(&sk->sk_receive_queue.lock);
6503 skb = skb_peek(&sk->sk_receive_queue);
6505 atomic_inc(&skb->users);
6506 spin_unlock_bh(&sk->sk_receive_queue.lock);
6508 skb = skb_dequeue(&sk->sk_receive_queue);
6514 /* Caller is allowed not to check sk->sk_err before calling. */
6515 error = sock_error(sk);
6519 if (sk->sk_shutdown & RCV_SHUTDOWN)
6522 /* User doesn't want to wait. */
6526 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6535 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6536 static void __sctp_write_space(struct sctp_association *asoc)
6538 struct sock *sk = asoc->base.sk;
6539 struct socket *sock = sk->sk_socket;
6541 if ((sctp_wspace(asoc) > 0) && sock) {
6542 if (waitqueue_active(&asoc->wait))
6543 wake_up_interruptible(&asoc->wait);
6545 if (sctp_writeable(sk)) {
6546 wait_queue_head_t *wq = sk_sleep(sk);
6548 if (wq && waitqueue_active(wq))
6549 wake_up_interruptible(wq);
6551 /* Note that we try to include the Async I/O support
6552 * here by modeling from the current TCP/UDP code.
6553 * We have not tested with it yet.
6555 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6556 sock_wake_async(sock,
6557 SOCK_WAKE_SPACE, POLL_OUT);
6562 /* Do accounting for the sndbuf space.
6563 * Decrement the used sndbuf space of the corresponding association by the
6564 * data size which was just transmitted(freed).
6566 static void sctp_wfree(struct sk_buff *skb)
6568 struct sctp_association *asoc;
6569 struct sctp_chunk *chunk;
6572 /* Get the saved chunk pointer. */
6573 chunk = *((struct sctp_chunk **)(skb->cb));
6576 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6577 sizeof(struct sk_buff) +
6578 sizeof(struct sctp_chunk);
6580 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6583 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6585 sk->sk_wmem_queued -= skb->truesize;
6586 sk_mem_uncharge(sk, skb->truesize);
6589 __sctp_write_space(asoc);
6591 sctp_association_put(asoc);
6594 /* Do accounting for the receive space on the socket.
6595 * Accounting for the association is done in ulpevent.c
6596 * We set this as a destructor for the cloned data skbs so that
6597 * accounting is done at the correct time.
6599 void sctp_sock_rfree(struct sk_buff *skb)
6601 struct sock *sk = skb->sk;
6602 struct sctp_ulpevent *event = sctp_skb2event(skb);
6604 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6607 * Mimic the behavior of sock_rfree
6609 sk_mem_uncharge(sk, event->rmem_len);
6613 /* Helper function to wait for space in the sndbuf. */
6614 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6617 struct sock *sk = asoc->base.sk;
6619 long current_timeo = *timeo_p;
6622 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6623 asoc, (long)(*timeo_p), msg_len);
6625 /* Increment the association's refcnt. */
6626 sctp_association_hold(asoc);
6628 /* Wait on the association specific sndbuf space. */
6630 prepare_to_wait_exclusive(&asoc->wait, &wait,
6631 TASK_INTERRUPTIBLE);
6634 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6637 if (signal_pending(current))
6638 goto do_interrupted;
6639 if (msg_len <= sctp_wspace(asoc))
6642 /* Let another process have a go. Since we are going
6645 sctp_release_sock(sk);
6646 current_timeo = schedule_timeout(current_timeo);
6647 BUG_ON(sk != asoc->base.sk);
6650 *timeo_p = current_timeo;
6654 finish_wait(&asoc->wait, &wait);
6656 /* Release the association's refcnt. */
6657 sctp_association_put(asoc);
6666 err = sock_intr_errno(*timeo_p);
6674 void sctp_data_ready(struct sock *sk, int len)
6676 struct socket_wq *wq;
6679 wq = rcu_dereference(sk->sk_wq);
6680 if (wq_has_sleeper(wq))
6681 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6682 POLLRDNORM | POLLRDBAND);
6683 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6687 /* If socket sndbuf has changed, wake up all per association waiters. */
6688 void sctp_write_space(struct sock *sk)
6690 struct sctp_association *asoc;
6692 /* Wake up the tasks in each wait queue. */
6693 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6694 __sctp_write_space(asoc);
6698 /* Is there any sndbuf space available on the socket?
6700 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6701 * associations on the same socket. For a UDP-style socket with
6702 * multiple associations, it is possible for it to be "unwriteable"
6703 * prematurely. I assume that this is acceptable because
6704 * a premature "unwriteable" is better than an accidental "writeable" which
6705 * would cause an unwanted block under certain circumstances. For the 1-1
6706 * UDP-style sockets or TCP-style sockets, this code should work.
6709 static int sctp_writeable(struct sock *sk)
6713 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6719 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6720 * returns immediately with EINPROGRESS.
6722 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6724 struct sock *sk = asoc->base.sk;
6726 long current_timeo = *timeo_p;
6729 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6732 /* Increment the association's refcnt. */
6733 sctp_association_hold(asoc);
6736 prepare_to_wait_exclusive(&asoc->wait, &wait,
6737 TASK_INTERRUPTIBLE);
6740 if (sk->sk_shutdown & RCV_SHUTDOWN)
6742 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6745 if (signal_pending(current))
6746 goto do_interrupted;
6748 if (sctp_state(asoc, ESTABLISHED))
6751 /* Let another process have a go. Since we are going
6754 sctp_release_sock(sk);
6755 current_timeo = schedule_timeout(current_timeo);
6758 *timeo_p = current_timeo;
6762 finish_wait(&asoc->wait, &wait);
6764 /* Release the association's refcnt. */
6765 sctp_association_put(asoc);
6770 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6773 err = -ECONNREFUSED;
6777 err = sock_intr_errno(*timeo_p);
6785 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6787 struct sctp_endpoint *ep;
6791 ep = sctp_sk(sk)->ep;
6795 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6796 TASK_INTERRUPTIBLE);
6798 if (list_empty(&ep->asocs)) {
6799 sctp_release_sock(sk);
6800 timeo = schedule_timeout(timeo);
6805 if (!sctp_sstate(sk, LISTENING))
6809 if (!list_empty(&ep->asocs))
6812 err = sock_intr_errno(timeo);
6813 if (signal_pending(current))
6821 finish_wait(sk_sleep(sk), &wait);
6826 static void sctp_wait_for_close(struct sock *sk, long timeout)
6831 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6832 if (list_empty(&sctp_sk(sk)->ep->asocs))
6834 sctp_release_sock(sk);
6835 timeout = schedule_timeout(timeout);
6837 } while (!signal_pending(current) && timeout);
6839 finish_wait(sk_sleep(sk), &wait);
6842 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6844 struct sk_buff *frag;
6849 /* Don't forget the fragments. */
6850 skb_walk_frags(skb, frag)
6851 sctp_skb_set_owner_r_frag(frag, sk);
6854 sctp_skb_set_owner_r(skb, sk);
6857 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6858 struct sctp_association *asoc)
6860 struct inet_sock *inet = inet_sk(sk);
6861 struct inet_sock *newinet;
6863 newsk->sk_type = sk->sk_type;
6864 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6865 newsk->sk_flags = sk->sk_flags;
6866 newsk->sk_no_check = sk->sk_no_check;
6867 newsk->sk_reuse = sk->sk_reuse;
6869 newsk->sk_shutdown = sk->sk_shutdown;
6870 newsk->sk_destruct = inet_sock_destruct;
6871 newsk->sk_family = sk->sk_family;
6872 newsk->sk_protocol = IPPROTO_SCTP;
6873 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6874 newsk->sk_sndbuf = sk->sk_sndbuf;
6875 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6876 newsk->sk_lingertime = sk->sk_lingertime;
6877 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6878 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6880 newinet = inet_sk(newsk);
6882 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6883 * getsockname() and getpeername()
6885 newinet->inet_sport = inet->inet_sport;
6886 newinet->inet_saddr = inet->inet_saddr;
6887 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6888 newinet->inet_dport = htons(asoc->peer.port);
6889 newinet->pmtudisc = inet->pmtudisc;
6890 newinet->inet_id = asoc->next_tsn ^ jiffies;
6892 newinet->uc_ttl = inet->uc_ttl;
6893 newinet->mc_loop = 1;
6894 newinet->mc_ttl = 1;
6895 newinet->mc_index = 0;
6896 newinet->mc_list = NULL;
6899 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6900 * and its messages to the newsk.
6902 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6903 struct sctp_association *assoc,
6904 sctp_socket_type_t type)
6906 struct sctp_sock *oldsp = sctp_sk(oldsk);
6907 struct sctp_sock *newsp = sctp_sk(newsk);
6908 struct sctp_bind_bucket *pp; /* hash list port iterator */
6909 struct sctp_endpoint *newep = newsp->ep;
6910 struct sk_buff *skb, *tmp;
6911 struct sctp_ulpevent *event;
6912 struct sctp_bind_hashbucket *head;
6913 struct list_head tmplist;
6915 /* Migrate socket buffer sizes and all the socket level options to the
6918 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6919 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6920 /* Brute force copy old sctp opt. */
6921 if (oldsp->do_auto_asconf) {
6922 memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
6923 inet_sk_copy_descendant(newsk, oldsk);
6924 memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
6926 inet_sk_copy_descendant(newsk, oldsk);
6928 /* Restore the ep value that was overwritten with the above structure
6934 /* Hook this new socket in to the bind_hash list. */
6935 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
6936 inet_sk(oldsk)->inet_num)];
6937 sctp_local_bh_disable();
6938 sctp_spin_lock(&head->lock);
6939 pp = sctp_sk(oldsk)->bind_hash;
6940 sk_add_bind_node(newsk, &pp->owner);
6941 sctp_sk(newsk)->bind_hash = pp;
6942 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6943 sctp_spin_unlock(&head->lock);
6944 sctp_local_bh_enable();
6946 /* Copy the bind_addr list from the original endpoint to the new
6947 * endpoint so that we can handle restarts properly
6949 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6950 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6952 /* Move any messages in the old socket's receive queue that are for the
6953 * peeled off association to the new socket's receive queue.
6955 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6956 event = sctp_skb2event(skb);
6957 if (event->asoc == assoc) {
6958 __skb_unlink(skb, &oldsk->sk_receive_queue);
6959 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6960 sctp_skb_set_owner_r_frag(skb, newsk);
6964 /* Clean up any messages pending delivery due to partial
6965 * delivery. Three cases:
6966 * 1) No partial deliver; no work.
6967 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6968 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6970 skb_queue_head_init(&newsp->pd_lobby);
6971 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6973 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6974 struct sk_buff_head *queue;
6976 /* Decide which queue to move pd_lobby skbs to. */
6977 if (assoc->ulpq.pd_mode) {
6978 queue = &newsp->pd_lobby;
6980 queue = &newsk->sk_receive_queue;
6982 /* Walk through the pd_lobby, looking for skbs that
6983 * need moved to the new socket.
6985 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6986 event = sctp_skb2event(skb);
6987 if (event->asoc == assoc) {
6988 __skb_unlink(skb, &oldsp->pd_lobby);
6989 __skb_queue_tail(queue, skb);
6990 sctp_skb_set_owner_r_frag(skb, newsk);
6994 /* Clear up any skbs waiting for the partial
6995 * delivery to finish.
6997 if (assoc->ulpq.pd_mode)
6998 sctp_clear_pd(oldsk, NULL);
7002 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7003 sctp_skb_set_owner_r_frag(skb, newsk);
7005 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7006 sctp_skb_set_owner_r_frag(skb, newsk);
7008 /* Set the type of socket to indicate that it is peeled off from the
7009 * original UDP-style socket or created with the accept() call on a
7010 * TCP-style socket..
7014 /* Mark the new socket "in-use" by the user so that any packets
7015 * that may arrive on the association after we've moved it are
7016 * queued to the backlog. This prevents a potential race between
7017 * backlog processing on the old socket and new-packet processing
7018 * on the new socket.
7020 * The caller has just allocated newsk so we can guarantee that other
7021 * paths won't try to lock it and then oldsk.
7023 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7024 sctp_assoc_migrate(assoc, newsk);
7026 /* If the association on the newsk is already closed before accept()
7027 * is called, set RCV_SHUTDOWN flag.
7029 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
7030 newsk->sk_shutdown |= RCV_SHUTDOWN;
7032 newsk->sk_state = SCTP_SS_ESTABLISHED;
7033 sctp_release_sock(newsk);
7037 /* This proto struct describes the ULP interface for SCTP. */
7038 struct proto sctp_prot = {
7040 .owner = THIS_MODULE,
7041 .close = sctp_close,
7042 .connect = sctp_connect,
7043 .disconnect = sctp_disconnect,
7044 .accept = sctp_accept,
7045 .ioctl = sctp_ioctl,
7046 .init = sctp_init_sock,
7047 .destroy = sctp_destroy_sock,
7048 .shutdown = sctp_shutdown,
7049 .setsockopt = sctp_setsockopt,
7050 .getsockopt = sctp_getsockopt,
7051 .sendmsg = sctp_sendmsg,
7052 .recvmsg = sctp_recvmsg,
7054 .backlog_rcv = sctp_backlog_rcv,
7056 .unhash = sctp_unhash,
7057 .get_port = sctp_get_port,
7058 .obj_size = sizeof(struct sctp_sock),
7059 .sysctl_mem = sysctl_sctp_mem,
7060 .sysctl_rmem = sysctl_sctp_rmem,
7061 .sysctl_wmem = sysctl_sctp_wmem,
7062 .memory_pressure = &sctp_memory_pressure,
7063 .enter_memory_pressure = sctp_enter_memory_pressure,
7064 .memory_allocated = &sctp_memory_allocated,
7065 .sockets_allocated = &sctp_sockets_allocated,
7068 #if IS_ENABLED(CONFIG_IPV6)
7070 struct proto sctpv6_prot = {
7072 .owner = THIS_MODULE,
7073 .close = sctp_close,
7074 .connect = sctp_connect,
7075 .disconnect = sctp_disconnect,
7076 .accept = sctp_accept,
7077 .ioctl = sctp_ioctl,
7078 .init = sctp_init_sock,
7079 .destroy = sctp_destroy_sock,
7080 .shutdown = sctp_shutdown,
7081 .setsockopt = sctp_setsockopt,
7082 .getsockopt = sctp_getsockopt,
7083 .sendmsg = sctp_sendmsg,
7084 .recvmsg = sctp_recvmsg,
7086 .backlog_rcv = sctp_backlog_rcv,
7088 .unhash = sctp_unhash,
7089 .get_port = sctp_get_port,
7090 .obj_size = sizeof(struct sctp6_sock),
7091 .sysctl_mem = sysctl_sctp_mem,
7092 .sysctl_rmem = sysctl_sctp_rmem,
7093 .sysctl_wmem = sysctl_sctp_wmem,
7094 .memory_pressure = &sctp_memory_pressure,
7095 .enter_memory_pressure = sctp_enter_memory_pressure,
7096 .memory_allocated = &sctp_memory_allocated,
7097 .sockets_allocated = &sctp_sockets_allocated,
7099 #endif /* IS_ENABLED(CONFIG_IPV6) */