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, see
32 * <http://www.gnu.org/licenses/>.
34 * Please send any bug reports or fixes you make to the
36 * lksctp developers <linux-sctp@vger.kernel.org>
38 * Written or modified by:
39 * La Monte H.P. Yarroll <piggy@acm.org>
40 * Narasimha Budihal <narsi@refcode.org>
41 * Karl Knutson <karl@athena.chicago.il.us>
42 * Jon Grimm <jgrimm@us.ibm.com>
43 * Xingang Guo <xingang.guo@intel.com>
44 * Daisy Chang <daisyc@us.ibm.com>
45 * Sridhar Samudrala <samudrala@us.ibm.com>
46 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
47 * Ardelle Fan <ardelle.fan@intel.com>
48 * Ryan Layer <rmlayer@us.ibm.com>
49 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
50 * Kevin Gao <kevin.gao@intel.com>
53 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
55 #include <linux/types.h>
56 #include <linux/kernel.h>
57 #include <linux/wait.h>
58 #include <linux/time.h>
60 #include <linux/capability.h>
61 #include <linux/fcntl.h>
62 #include <linux/poll.h>
63 #include <linux/init.h>
64 #include <linux/crypto.h>
65 #include <linux/slab.h>
66 #include <linux/file.h>
67 #include <linux/compat.h>
71 #include <net/route.h>
73 #include <net/inet_common.h>
75 #include <linux/socket.h> /* for sa_family_t */
76 #include <linux/export.h>
78 #include <net/sctp/sctp.h>
79 #include <net/sctp/sm.h>
81 /* Forward declarations for internal helper functions. */
82 static int sctp_writeable(struct sock *sk);
83 static void sctp_wfree(struct sk_buff *skb);
84 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
86 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p);
87 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
88 static int sctp_wait_for_accept(struct sock *sk, long timeo);
89 static void sctp_wait_for_close(struct sock *sk, long timeo);
90 static void sctp_destruct_sock(struct sock *sk);
91 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
92 union sctp_addr *addr, int len);
93 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
94 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
95 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
96 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
97 static int sctp_send_asconf(struct sctp_association *asoc,
98 struct sctp_chunk *chunk);
99 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
100 static int sctp_autobind(struct sock *sk);
101 static void sctp_sock_migrate(struct sock *, struct sock *,
102 struct sctp_association *, sctp_socket_type_t);
104 extern struct kmem_cache *sctp_bucket_cachep;
105 extern long sysctl_sctp_mem[3];
106 extern int sysctl_sctp_rmem[3];
107 extern int sysctl_sctp_wmem[3];
109 static int sctp_memory_pressure;
110 static atomic_long_t sctp_memory_allocated;
111 struct percpu_counter sctp_sockets_allocated;
113 static void sctp_enter_memory_pressure(struct sock *sk)
115 sctp_memory_pressure = 1;
119 /* Get the sndbuf space available at the time on the association. */
120 static inline int sctp_wspace(struct sctp_association *asoc)
124 if (asoc->ep->sndbuf_policy)
125 amt = asoc->sndbuf_used;
127 amt = sk_wmem_alloc_get(asoc->base.sk);
129 if (amt >= asoc->base.sk->sk_sndbuf) {
130 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
133 amt = sk_stream_wspace(asoc->base.sk);
138 amt = asoc->base.sk->sk_sndbuf - amt;
143 /* Increment the used sndbuf space count of the corresponding association by
144 * the size of the outgoing data chunk.
145 * Also, set the skb destructor for sndbuf accounting later.
147 * Since it is always 1-1 between chunk and skb, and also a new skb is always
148 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
149 * destructor in the data chunk skb for the purpose of the sndbuf space
152 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
154 struct sctp_association *asoc = chunk->asoc;
155 struct sock *sk = asoc->base.sk;
157 /* The sndbuf space is tracked per association. */
158 sctp_association_hold(asoc);
160 skb_set_owner_w(chunk->skb, sk);
162 chunk->skb->destructor = sctp_wfree;
163 /* Save the chunk pointer in skb for sctp_wfree to use later. */
164 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
166 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
167 sizeof(struct sk_buff) +
168 sizeof(struct sctp_chunk);
170 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
171 sk->sk_wmem_queued += chunk->skb->truesize;
172 sk_mem_charge(sk, chunk->skb->truesize);
175 /* Verify that this is a valid address. */
176 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
181 /* Verify basic sockaddr. */
182 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
186 /* Is this a valid SCTP address? */
187 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
190 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
196 /* Look up the association by its id. If this is not a UDP-style
197 * socket, the ID field is always ignored.
199 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
201 struct sctp_association *asoc = NULL;
203 /* If this is not a UDP-style socket, assoc id should be ignored. */
204 if (!sctp_style(sk, UDP)) {
205 /* Return NULL if the socket state is not ESTABLISHED. It
206 * could be a TCP-style listening socket or a socket which
207 * hasn't yet called connect() to establish an association.
209 if (!sctp_sstate(sk, ESTABLISHED))
212 /* Get the first and the only association from the list. */
213 if (!list_empty(&sctp_sk(sk)->ep->asocs))
214 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
215 struct sctp_association, asocs);
219 /* Otherwise this is a UDP-style socket. */
220 if (!id || (id == (sctp_assoc_t)-1))
223 spin_lock_bh(&sctp_assocs_id_lock);
224 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
225 spin_unlock_bh(&sctp_assocs_id_lock);
227 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
233 /* Look up the transport from an address and an assoc id. If both address and
234 * id are specified, the associations matching the address and the id should be
237 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
238 struct sockaddr_storage *addr,
241 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
242 struct sctp_transport *transport;
243 union sctp_addr *laddr = (union sctp_addr *)addr;
245 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
252 id_asoc = sctp_id2assoc(sk, id);
253 if (id_asoc && (id_asoc != addr_asoc))
256 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
257 (union sctp_addr *)addr);
262 /* API 3.1.2 bind() - UDP Style Syntax
263 * The syntax of bind() is,
265 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
267 * sd - the socket descriptor returned by socket().
268 * addr - the address structure (struct sockaddr_in or struct
269 * sockaddr_in6 [RFC 2553]),
270 * addr_len - the size of the address structure.
272 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
278 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
281 /* Disallow binding twice. */
282 if (!sctp_sk(sk)->ep->base.bind_addr.port)
283 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
293 static long sctp_get_port_local(struct sock *, union sctp_addr *);
295 /* Verify this is a valid sockaddr. */
296 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
297 union sctp_addr *addr, int len)
301 /* Check minimum size. */
302 if (len < sizeof (struct sockaddr))
305 /* V4 mapped address are really of AF_INET family */
306 if (addr->sa.sa_family == AF_INET6 &&
307 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
308 if (!opt->pf->af_supported(AF_INET, opt))
311 /* Does this PF support this AF? */
312 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
316 /* If we get this far, af is valid. */
317 af = sctp_get_af_specific(addr->sa.sa_family);
319 if (len < af->sockaddr_len)
325 /* Bind a local address either to an endpoint or to an association. */
326 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
328 struct net *net = sock_net(sk);
329 struct sctp_sock *sp = sctp_sk(sk);
330 struct sctp_endpoint *ep = sp->ep;
331 struct sctp_bind_addr *bp = &ep->base.bind_addr;
336 /* Common sockaddr verification. */
337 af = sctp_sockaddr_af(sp, addr, len);
339 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
340 __func__, sk, addr, len);
344 snum = ntohs(addr->v4.sin_port);
346 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
347 __func__, sk, &addr->sa, bp->port, snum, len);
349 /* PF specific bind() address verification. */
350 if (!sp->pf->bind_verify(sp, addr))
351 return -EADDRNOTAVAIL;
353 /* We must either be unbound, or bind to the same port.
354 * It's OK to allow 0 ports if we are already bound.
355 * We'll just inhert an already bound port in this case
360 else if (snum != bp->port) {
361 pr_debug("%s: new port %d doesn't match existing port "
362 "%d\n", __func__, snum, bp->port);
367 if (snum && snum < PROT_SOCK &&
368 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
371 /* See if the address matches any of the addresses we may have
372 * already bound before checking against other endpoints.
374 if (sctp_bind_addr_match(bp, addr, sp))
377 /* Make sure we are allowed to bind here.
378 * The function sctp_get_port_local() does duplicate address
381 addr->v4.sin_port = htons(snum);
382 if ((ret = sctp_get_port_local(sk, addr))) {
386 /* Refresh ephemeral port. */
388 bp->port = inet_sk(sk)->inet_num;
390 /* Add the address to the bind address list.
391 * Use GFP_ATOMIC since BHs will be disabled.
393 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
395 /* Copy back into socket for getsockname() use. */
397 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
398 af->to_sk_saddr(addr, sk);
404 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
406 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
407 * at any one time. If a sender, after sending an ASCONF chunk, decides
408 * it needs to transfer another ASCONF Chunk, it MUST wait until the
409 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
410 * subsequent ASCONF. Note this restriction binds each side, so at any
411 * time two ASCONF may be in-transit on any given association (one sent
412 * from each endpoint).
414 static int sctp_send_asconf(struct sctp_association *asoc,
415 struct sctp_chunk *chunk)
417 struct net *net = sock_net(asoc->base.sk);
420 /* If there is an outstanding ASCONF chunk, queue it for later
423 if (asoc->addip_last_asconf) {
424 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
428 /* Hold the chunk until an ASCONF_ACK is received. */
429 sctp_chunk_hold(chunk);
430 retval = sctp_primitive_ASCONF(net, asoc, chunk);
432 sctp_chunk_free(chunk);
434 asoc->addip_last_asconf = chunk;
440 /* Add a list of addresses as bind addresses to local endpoint or
443 * Basically run through each address specified in the addrs/addrcnt
444 * array/length pair, determine if it is IPv6 or IPv4 and call
445 * sctp_do_bind() on it.
447 * If any of them fails, then the operation will be reversed and the
448 * ones that were added will be removed.
450 * Only sctp_setsockopt_bindx() is supposed to call this function.
452 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
457 struct sockaddr *sa_addr;
460 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
464 for (cnt = 0; cnt < addrcnt; cnt++) {
465 /* The list may contain either IPv4 or IPv6 address;
466 * determine the address length for walking thru the list.
469 af = sctp_get_af_specific(sa_addr->sa_family);
475 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
478 addr_buf += af->sockaddr_len;
482 /* Failed. Cleanup the ones that have been added */
484 sctp_bindx_rem(sk, addrs, cnt);
492 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
493 * associations that are part of the endpoint indicating that a list of local
494 * addresses are added to the endpoint.
496 * If any of the addresses is already in the bind address list of the
497 * association, we do not send the chunk for that association. But it will not
498 * affect other associations.
500 * Only sctp_setsockopt_bindx() is supposed to call this function.
502 static int sctp_send_asconf_add_ip(struct sock *sk,
503 struct sockaddr *addrs,
506 struct net *net = sock_net(sk);
507 struct sctp_sock *sp;
508 struct sctp_endpoint *ep;
509 struct sctp_association *asoc;
510 struct sctp_bind_addr *bp;
511 struct sctp_chunk *chunk;
512 struct sctp_sockaddr_entry *laddr;
513 union sctp_addr *addr;
514 union sctp_addr saveaddr;
521 if (!net->sctp.addip_enable)
527 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
528 __func__, sk, addrs, addrcnt);
530 list_for_each_entry(asoc, &ep->asocs, asocs) {
531 if (!asoc->peer.asconf_capable)
534 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
537 if (!sctp_state(asoc, ESTABLISHED))
540 /* Check if any address in the packed array of addresses is
541 * in the bind address list of the association. If so,
542 * do not send the asconf chunk to its peer, but continue with
543 * other associations.
546 for (i = 0; i < addrcnt; i++) {
548 af = sctp_get_af_specific(addr->v4.sin_family);
554 if (sctp_assoc_lookup_laddr(asoc, addr))
557 addr_buf += af->sockaddr_len;
562 /* Use the first valid address in bind addr list of
563 * association as Address Parameter of ASCONF CHUNK.
565 bp = &asoc->base.bind_addr;
566 p = bp->address_list.next;
567 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
568 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
569 addrcnt, SCTP_PARAM_ADD_IP);
575 /* Add the new addresses to the bind address list with
576 * use_as_src set to 0.
579 for (i = 0; i < addrcnt; i++) {
581 af = sctp_get_af_specific(addr->v4.sin_family);
582 memcpy(&saveaddr, addr, af->sockaddr_len);
583 retval = sctp_add_bind_addr(bp, &saveaddr,
584 SCTP_ADDR_NEW, GFP_ATOMIC);
585 addr_buf += af->sockaddr_len;
587 if (asoc->src_out_of_asoc_ok) {
588 struct sctp_transport *trans;
590 list_for_each_entry(trans,
591 &asoc->peer.transport_addr_list, transports) {
592 /* Clear the source and route cache */
593 dst_release(trans->dst);
594 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
595 2*asoc->pathmtu, 4380));
596 trans->ssthresh = asoc->peer.i.a_rwnd;
597 trans->rto = asoc->rto_initial;
598 sctp_max_rto(asoc, trans);
599 trans->rtt = trans->srtt = trans->rttvar = 0;
600 sctp_transport_route(trans, NULL,
601 sctp_sk(asoc->base.sk));
604 retval = sctp_send_asconf(asoc, chunk);
611 /* Remove a list of addresses from bind addresses list. Do not remove the
614 * Basically run through each address specified in the addrs/addrcnt
615 * array/length pair, determine if it is IPv6 or IPv4 and call
616 * sctp_del_bind() on it.
618 * If any of them fails, then the operation will be reversed and the
619 * ones that were removed will be added back.
621 * At least one address has to be left; if only one address is
622 * available, the operation will return -EBUSY.
624 * Only sctp_setsockopt_bindx() is supposed to call this function.
626 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
628 struct sctp_sock *sp = sctp_sk(sk);
629 struct sctp_endpoint *ep = sp->ep;
631 struct sctp_bind_addr *bp = &ep->base.bind_addr;
634 union sctp_addr *sa_addr;
637 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
638 __func__, sk, addrs, addrcnt);
641 for (cnt = 0; cnt < addrcnt; cnt++) {
642 /* If the bind address list is empty or if there is only one
643 * bind address, there is nothing more to be removed (we need
644 * at least one address here).
646 if (list_empty(&bp->address_list) ||
647 (sctp_list_single_entry(&bp->address_list))) {
653 af = sctp_get_af_specific(sa_addr->sa.sa_family);
659 if (!af->addr_valid(sa_addr, sp, NULL)) {
660 retval = -EADDRNOTAVAIL;
664 if (sa_addr->v4.sin_port &&
665 sa_addr->v4.sin_port != htons(bp->port)) {
670 if (!sa_addr->v4.sin_port)
671 sa_addr->v4.sin_port = htons(bp->port);
673 /* FIXME - There is probably a need to check if sk->sk_saddr and
674 * sk->sk_rcv_addr are currently set to one of the addresses to
675 * be removed. This is something which needs to be looked into
676 * when we are fixing the outstanding issues with multi-homing
677 * socket routing and failover schemes. Refer to comments in
678 * sctp_do_bind(). -daisy
680 retval = sctp_del_bind_addr(bp, sa_addr);
682 addr_buf += af->sockaddr_len;
685 /* Failed. Add the ones that has been removed back */
687 sctp_bindx_add(sk, addrs, cnt);
695 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
696 * the associations that are part of the endpoint indicating that a list of
697 * local addresses are removed from the endpoint.
699 * If any of the addresses is already in the bind address list of the
700 * association, we do not send the chunk for that association. But it will not
701 * affect other associations.
703 * Only sctp_setsockopt_bindx() is supposed to call this function.
705 static int sctp_send_asconf_del_ip(struct sock *sk,
706 struct sockaddr *addrs,
709 struct net *net = sock_net(sk);
710 struct sctp_sock *sp;
711 struct sctp_endpoint *ep;
712 struct sctp_association *asoc;
713 struct sctp_transport *transport;
714 struct sctp_bind_addr *bp;
715 struct sctp_chunk *chunk;
716 union sctp_addr *laddr;
719 struct sctp_sockaddr_entry *saddr;
725 if (!net->sctp.addip_enable)
731 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
732 __func__, sk, addrs, addrcnt);
734 list_for_each_entry(asoc, &ep->asocs, asocs) {
736 if (!asoc->peer.asconf_capable)
739 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
742 if (!sctp_state(asoc, ESTABLISHED))
745 /* Check if any address in the packed array of addresses is
746 * not present in the bind address list of the association.
747 * If so, do not send the asconf chunk to its peer, but
748 * continue with other associations.
751 for (i = 0; i < addrcnt; i++) {
753 af = sctp_get_af_specific(laddr->v4.sin_family);
759 if (!sctp_assoc_lookup_laddr(asoc, laddr))
762 addr_buf += af->sockaddr_len;
767 /* Find one address in the association's bind address list
768 * that is not in the packed array of addresses. This is to
769 * make sure that we do not delete all the addresses in the
772 bp = &asoc->base.bind_addr;
773 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
775 if ((laddr == NULL) && (addrcnt == 1)) {
776 if (asoc->asconf_addr_del_pending)
778 asoc->asconf_addr_del_pending =
779 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
780 if (asoc->asconf_addr_del_pending == NULL) {
784 asoc->asconf_addr_del_pending->sa.sa_family =
786 asoc->asconf_addr_del_pending->v4.sin_port =
788 if (addrs->sa_family == AF_INET) {
789 struct sockaddr_in *sin;
791 sin = (struct sockaddr_in *)addrs;
792 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
793 } else if (addrs->sa_family == AF_INET6) {
794 struct sockaddr_in6 *sin6;
796 sin6 = (struct sockaddr_in6 *)addrs;
797 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
800 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
801 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
802 asoc->asconf_addr_del_pending);
804 asoc->src_out_of_asoc_ok = 1;
812 /* We do not need RCU protection throughout this loop
813 * because this is done under a socket lock from the
816 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
824 /* Reset use_as_src flag for the addresses in the bind address
825 * list that are to be deleted.
828 for (i = 0; i < addrcnt; i++) {
830 af = sctp_get_af_specific(laddr->v4.sin_family);
831 list_for_each_entry(saddr, &bp->address_list, list) {
832 if (sctp_cmp_addr_exact(&saddr->a, laddr))
833 saddr->state = SCTP_ADDR_DEL;
835 addr_buf += af->sockaddr_len;
838 /* Update the route and saddr entries for all the transports
839 * as some of the addresses in the bind address list are
840 * about to be deleted and cannot be used as source addresses.
842 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
844 dst_release(transport->dst);
845 sctp_transport_route(transport, NULL,
846 sctp_sk(asoc->base.sk));
850 /* We don't need to transmit ASCONF */
852 retval = sctp_send_asconf(asoc, chunk);
858 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
859 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
861 struct sock *sk = sctp_opt2sk(sp);
862 union sctp_addr *addr;
865 /* It is safe to write port space in caller. */
867 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
868 af = sctp_get_af_specific(addr->sa.sa_family);
871 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
874 if (addrw->state == SCTP_ADDR_NEW)
875 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
877 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
880 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
883 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
886 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
887 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
890 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
891 * Section 3.1.2 for this usage.
893 * addrs is a pointer to an array of one or more socket addresses. Each
894 * address is contained in its appropriate structure (i.e. struct
895 * sockaddr_in or struct sockaddr_in6) the family of the address type
896 * must be used to distinguish the address length (note that this
897 * representation is termed a "packed array" of addresses). The caller
898 * specifies the number of addresses in the array with addrcnt.
900 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
901 * -1, and sets errno to the appropriate error code.
903 * For SCTP, the port given in each socket address must be the same, or
904 * sctp_bindx() will fail, setting errno to EINVAL.
906 * The flags parameter is formed from the bitwise OR of zero or more of
907 * the following currently defined flags:
909 * SCTP_BINDX_ADD_ADDR
911 * SCTP_BINDX_REM_ADDR
913 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
914 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
915 * addresses from the association. The two flags are mutually exclusive;
916 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
917 * not remove all addresses from an association; sctp_bindx() will
918 * reject such an attempt with EINVAL.
920 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
921 * additional addresses with an endpoint after calling bind(). Or use
922 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
923 * socket is associated with so that no new association accepted will be
924 * associated with those addresses. If the endpoint supports dynamic
925 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
926 * endpoint to send the appropriate message to the peer to change the
927 * peers address lists.
929 * Adding and removing addresses from a connected association is
930 * optional functionality. Implementations that do not support this
931 * functionality should return EOPNOTSUPP.
933 * Basically do nothing but copying the addresses from user to kernel
934 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
935 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
938 * We don't use copy_from_user() for optimization: we first do the
939 * sanity checks (buffer size -fast- and access check-healthy
940 * pointer); if all of those succeed, then we can alloc the memory
941 * (expensive operation) needed to copy the data to kernel. Then we do
942 * the copying without checking the user space area
943 * (__copy_from_user()).
945 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
948 * sk The sk of the socket
949 * addrs The pointer to the addresses in user land
950 * addrssize Size of the addrs buffer
951 * op Operation to perform (add or remove, see the flags of
954 * Returns 0 if ok, <0 errno code on error.
956 static int sctp_setsockopt_bindx(struct sock *sk,
957 struct sockaddr __user *addrs,
958 int addrs_size, int op)
960 struct sockaddr *kaddrs;
964 struct sockaddr *sa_addr;
968 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
969 __func__, sk, addrs, addrs_size, op);
971 if (unlikely(addrs_size <= 0))
974 /* Check the user passed a healthy pointer. */
975 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
978 /* Alloc space for the address array in kernel memory. */
979 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
980 if (unlikely(!kaddrs))
983 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
988 /* Walk through the addrs buffer and count the number of addresses. */
990 while (walk_size < addrs_size) {
991 if (walk_size + sizeof(sa_family_t) > addrs_size) {
997 af = sctp_get_af_specific(sa_addr->sa_family);
999 /* If the address family is not supported or if this address
1000 * causes the address buffer to overflow return EINVAL.
1002 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1007 addr_buf += af->sockaddr_len;
1008 walk_size += af->sockaddr_len;
1013 case SCTP_BINDX_ADD_ADDR:
1014 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1017 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1020 case SCTP_BINDX_REM_ADDR:
1021 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1024 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1038 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1040 * Common routine for handling connect() and sctp_connectx().
1041 * Connect will come in with just a single address.
1043 static int __sctp_connect(struct sock *sk,
1044 struct sockaddr *kaddrs,
1046 sctp_assoc_t *assoc_id)
1048 struct net *net = sock_net(sk);
1049 struct sctp_sock *sp;
1050 struct sctp_endpoint *ep;
1051 struct sctp_association *asoc = NULL;
1052 struct sctp_association *asoc2;
1053 struct sctp_transport *transport;
1061 union sctp_addr *sa_addr = NULL;
1063 unsigned short port;
1064 unsigned int f_flags = 0;
1069 /* connect() cannot be done on a socket that is already in ESTABLISHED
1070 * state - UDP-style peeled off socket or a TCP-style socket that
1071 * is already connected.
1072 * It cannot be done even on a TCP-style listening socket.
1074 if (sctp_sstate(sk, ESTABLISHED) ||
1075 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1080 /* Walk through the addrs buffer and count the number of addresses. */
1082 while (walk_size < addrs_size) {
1083 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1089 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1091 /* If the address family is not supported or if this address
1092 * causes the address buffer to overflow return EINVAL.
1094 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1099 port = ntohs(sa_addr->v4.sin_port);
1101 /* Save current address so we can work with it */
1102 memcpy(&to, sa_addr, af->sockaddr_len);
1104 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1108 /* Make sure the destination port is correctly set
1111 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1116 /* Check if there already is a matching association on the
1117 * endpoint (other than the one created here).
1119 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1120 if (asoc2 && asoc2 != asoc) {
1121 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1128 /* If we could not find a matching association on the endpoint,
1129 * make sure that there is no peeled-off association matching
1130 * the peer address even on another socket.
1132 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1133 err = -EADDRNOTAVAIL;
1138 /* If a bind() or sctp_bindx() is not called prior to
1139 * an sctp_connectx() call, the system picks an
1140 * ephemeral port and will choose an address set
1141 * equivalent to binding with a wildcard address.
1143 if (!ep->base.bind_addr.port) {
1144 if (sctp_autobind(sk)) {
1150 * If an unprivileged user inherits a 1-many
1151 * style socket with open associations on a
1152 * privileged port, it MAY be permitted to
1153 * accept new associations, but it SHOULD NOT
1154 * be permitted to open new associations.
1156 if (ep->base.bind_addr.port < PROT_SOCK &&
1157 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1163 scope = sctp_scope(&to);
1164 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1170 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1178 /* Prime the peer's transport structures. */
1179 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1187 addr_buf += af->sockaddr_len;
1188 walk_size += af->sockaddr_len;
1191 /* In case the user of sctp_connectx() wants an association
1192 * id back, assign one now.
1195 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1200 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1205 /* Initialize sk's dport and daddr for getpeername() */
1206 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1207 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1208 af->to_sk_daddr(sa_addr, sk);
1211 /* in-kernel sockets don't generally have a file allocated to them
1212 * if all they do is call sock_create_kern().
1214 if (sk->sk_socket->file)
1215 f_flags = sk->sk_socket->file->f_flags;
1217 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1219 err = sctp_wait_for_connect(asoc, &timeo);
1220 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1221 *assoc_id = asoc->assoc_id;
1223 /* Don't free association on exit. */
1227 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1228 __func__, asoc, kaddrs, err);
1231 /* sctp_primitive_ASSOCIATE may have added this association
1232 * To the hash table, try to unhash it, just in case, its a noop
1233 * if it wasn't hashed so we're safe
1235 sctp_unhash_established(asoc);
1236 sctp_association_free(asoc);
1241 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1244 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1245 * sctp_assoc_t *asoc);
1247 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1248 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1249 * or IPv6 addresses.
1251 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1252 * Section 3.1.2 for this usage.
1254 * addrs is a pointer to an array of one or more socket addresses. Each
1255 * address is contained in its appropriate structure (i.e. struct
1256 * sockaddr_in or struct sockaddr_in6) the family of the address type
1257 * must be used to distengish the address length (note that this
1258 * representation is termed a "packed array" of addresses). The caller
1259 * specifies the number of addresses in the array with addrcnt.
1261 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1262 * the association id of the new association. On failure, sctp_connectx()
1263 * returns -1, and sets errno to the appropriate error code. The assoc_id
1264 * is not touched by the kernel.
1266 * For SCTP, the port given in each socket address must be the same, or
1267 * sctp_connectx() will fail, setting errno to EINVAL.
1269 * An application can use sctp_connectx to initiate an association with
1270 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1271 * allows a caller to specify multiple addresses at which a peer can be
1272 * reached. The way the SCTP stack uses the list of addresses to set up
1273 * the association is implementation dependent. This function only
1274 * specifies that the stack will try to make use of all the addresses in
1275 * the list when needed.
1277 * Note that the list of addresses passed in is only used for setting up
1278 * the association. It does not necessarily equal the set of addresses
1279 * the peer uses for the resulting association. If the caller wants to
1280 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1281 * retrieve them after the association has been set up.
1283 * Basically do nothing but copying the addresses from user to kernel
1284 * land and invoking either sctp_connectx(). This is used for tunneling
1285 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1287 * We don't use copy_from_user() for optimization: we first do the
1288 * sanity checks (buffer size -fast- and access check-healthy
1289 * pointer); if all of those succeed, then we can alloc the memory
1290 * (expensive operation) needed to copy the data to kernel. Then we do
1291 * the copying without checking the user space area
1292 * (__copy_from_user()).
1294 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1297 * sk The sk of the socket
1298 * addrs The pointer to the addresses in user land
1299 * addrssize Size of the addrs buffer
1301 * Returns >=0 if ok, <0 errno code on error.
1303 static int __sctp_setsockopt_connectx(struct sock *sk,
1304 struct sockaddr __user *addrs,
1306 sctp_assoc_t *assoc_id)
1309 struct sockaddr *kaddrs;
1311 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1312 __func__, sk, addrs, addrs_size);
1314 if (unlikely(addrs_size <= 0))
1317 /* Check the user passed a healthy pointer. */
1318 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1321 /* Alloc space for the address array in kernel memory. */
1322 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1323 if (unlikely(!kaddrs))
1326 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1329 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1338 * This is an older interface. It's kept for backward compatibility
1339 * to the option that doesn't provide association id.
1341 static int sctp_setsockopt_connectx_old(struct sock *sk,
1342 struct sockaddr __user *addrs,
1345 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1349 * New interface for the API. The since the API is done with a socket
1350 * option, to make it simple we feed back the association id is as a return
1351 * indication to the call. Error is always negative and association id is
1354 static int sctp_setsockopt_connectx(struct sock *sk,
1355 struct sockaddr __user *addrs,
1358 sctp_assoc_t assoc_id = 0;
1361 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1370 * New (hopefully final) interface for the API.
1371 * We use the sctp_getaddrs_old structure so that use-space library
1372 * can avoid any unnecessary allocations. The only different part
1373 * is that we store the actual length of the address buffer into the
1374 * addrs_num structure member. That way we can re-use the existing
1377 #ifdef CONFIG_COMPAT
1378 struct compat_sctp_getaddrs_old {
1379 sctp_assoc_t assoc_id;
1381 compat_uptr_t addrs; /* struct sockaddr * */
1385 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1386 char __user *optval,
1389 struct sctp_getaddrs_old param;
1390 sctp_assoc_t assoc_id = 0;
1393 #ifdef CONFIG_COMPAT
1394 if (is_compat_task()) {
1395 struct compat_sctp_getaddrs_old param32;
1397 if (len < sizeof(param32))
1399 if (copy_from_user(¶m32, optval, sizeof(param32)))
1402 param.assoc_id = param32.assoc_id;
1403 param.addr_num = param32.addr_num;
1404 param.addrs = compat_ptr(param32.addrs);
1408 if (len < sizeof(param))
1410 if (copy_from_user(¶m, optval, sizeof(param)))
1414 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1415 param.addrs, param.addr_num,
1417 if (err == 0 || err == -EINPROGRESS) {
1418 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1420 if (put_user(sizeof(assoc_id), optlen))
1427 /* API 3.1.4 close() - UDP Style Syntax
1428 * Applications use close() to perform graceful shutdown (as described in
1429 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1430 * by a UDP-style socket.
1434 * ret = close(int sd);
1436 * sd - the socket descriptor of the associations to be closed.
1438 * To gracefully shutdown a specific association represented by the
1439 * UDP-style socket, an application should use the sendmsg() call,
1440 * passing no user data, but including the appropriate flag in the
1441 * ancillary data (see Section xxxx).
1443 * If sd in the close() call is a branched-off socket representing only
1444 * one association, the shutdown is performed on that association only.
1446 * 4.1.6 close() - TCP Style Syntax
1448 * Applications use close() to gracefully close down an association.
1452 * int close(int sd);
1454 * sd - the socket descriptor of the association to be closed.
1456 * After an application calls close() on a socket descriptor, no further
1457 * socket operations will succeed on that descriptor.
1459 * API 7.1.4 SO_LINGER
1461 * An application using the TCP-style socket can use this option to
1462 * perform the SCTP ABORT primitive. The linger option structure is:
1465 * int l_onoff; // option on/off
1466 * int l_linger; // linger time
1469 * To enable the option, set l_onoff to 1. If the l_linger value is set
1470 * to 0, calling close() is the same as the ABORT primitive. If the
1471 * value is set to a negative value, the setsockopt() call will return
1472 * an error. If the value is set to a positive value linger_time, the
1473 * close() can be blocked for at most linger_time ms. If the graceful
1474 * shutdown phase does not finish during this period, close() will
1475 * return but the graceful shutdown phase continues in the system.
1477 static void sctp_close(struct sock *sk, long timeout)
1479 struct net *net = sock_net(sk);
1480 struct sctp_endpoint *ep;
1481 struct sctp_association *asoc;
1482 struct list_head *pos, *temp;
1483 unsigned int data_was_unread;
1485 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1488 sk->sk_shutdown = SHUTDOWN_MASK;
1489 sk->sk_state = SCTP_SS_CLOSING;
1491 ep = sctp_sk(sk)->ep;
1493 /* Clean up any skbs sitting on the receive queue. */
1494 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1495 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1497 /* Walk all associations on an endpoint. */
1498 list_for_each_safe(pos, temp, &ep->asocs) {
1499 asoc = list_entry(pos, struct sctp_association, asocs);
1501 if (sctp_style(sk, TCP)) {
1502 /* A closed association can still be in the list if
1503 * it belongs to a TCP-style listening socket that is
1504 * not yet accepted. If so, free it. If not, send an
1505 * ABORT or SHUTDOWN based on the linger options.
1507 if (sctp_state(asoc, CLOSED)) {
1508 sctp_unhash_established(asoc);
1509 sctp_association_free(asoc);
1514 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1515 !skb_queue_empty(&asoc->ulpq.reasm) ||
1516 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1517 struct sctp_chunk *chunk;
1519 chunk = sctp_make_abort_user(asoc, NULL, 0);
1521 sctp_primitive_ABORT(net, asoc, chunk);
1523 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1526 /* On a TCP-style socket, block for at most linger_time if set. */
1527 if (sctp_style(sk, TCP) && timeout)
1528 sctp_wait_for_close(sk, timeout);
1530 /* This will run the backlog queue. */
1533 /* Supposedly, no process has access to the socket, but
1534 * the net layers still may.
1539 /* Hold the sock, since sk_common_release() will put sock_put()
1540 * and we have just a little more cleanup.
1543 sk_common_release(sk);
1550 SCTP_DBG_OBJCNT_DEC(sock);
1553 /* Handle EPIPE error. */
1554 static int sctp_error(struct sock *sk, int flags, int err)
1557 err = sock_error(sk) ? : -EPIPE;
1558 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1559 send_sig(SIGPIPE, current, 0);
1563 /* API 3.1.3 sendmsg() - UDP Style Syntax
1565 * An application uses sendmsg() and recvmsg() calls to transmit data to
1566 * and receive data from its peer.
1568 * ssize_t sendmsg(int socket, const struct msghdr *message,
1571 * socket - the socket descriptor of the endpoint.
1572 * message - pointer to the msghdr structure which contains a single
1573 * user message and possibly some ancillary data.
1575 * See Section 5 for complete description of the data
1578 * flags - flags sent or received with the user message, see Section
1579 * 5 for complete description of the flags.
1581 * Note: This function could use a rewrite especially when explicit
1582 * connect support comes in.
1584 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1586 static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1588 static int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1589 struct msghdr *msg, size_t msg_len)
1591 struct net *net = sock_net(sk);
1592 struct sctp_sock *sp;
1593 struct sctp_endpoint *ep;
1594 struct sctp_association *new_asoc = NULL, *asoc = NULL;
1595 struct sctp_transport *transport, *chunk_tp;
1596 struct sctp_chunk *chunk;
1598 struct sockaddr *msg_name = NULL;
1599 struct sctp_sndrcvinfo default_sinfo;
1600 struct sctp_sndrcvinfo *sinfo;
1601 struct sctp_initmsg *sinit;
1602 sctp_assoc_t associd = 0;
1603 sctp_cmsgs_t cmsgs = { NULL };
1607 __u16 sinfo_flags = 0;
1608 struct sctp_datamsg *datamsg;
1609 int msg_flags = msg->msg_flags;
1615 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1618 /* We cannot send a message over a TCP-style listening socket. */
1619 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1624 /* Parse out the SCTP CMSGs. */
1625 err = sctp_msghdr_parse(msg, &cmsgs);
1627 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1631 /* Fetch the destination address for this packet. This
1632 * address only selects the association--it is not necessarily
1633 * the address we will send to.
1634 * For a peeled-off socket, msg_name is ignored.
1636 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1637 int msg_namelen = msg->msg_namelen;
1639 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1644 if (msg_namelen > sizeof(to))
1645 msg_namelen = sizeof(to);
1646 memcpy(&to, msg->msg_name, msg_namelen);
1647 msg_name = msg->msg_name;
1653 /* Did the user specify SNDRCVINFO? */
1655 sinfo_flags = sinfo->sinfo_flags;
1656 associd = sinfo->sinfo_assoc_id;
1659 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1660 msg_len, sinfo_flags);
1662 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1663 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1668 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1669 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1670 * If SCTP_ABORT is set, the message length could be non zero with
1671 * the msg_iov set to the user abort reason.
1673 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1674 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1679 /* If SCTP_ADDR_OVER is set, there must be an address
1680 * specified in msg_name.
1682 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1689 pr_debug("%s: about to look up association\n", __func__);
1693 /* If a msg_name has been specified, assume this is to be used. */
1695 /* Look for a matching association on the endpoint. */
1696 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1698 /* If we could not find a matching association on the
1699 * endpoint, make sure that it is not a TCP-style
1700 * socket that already has an association or there is
1701 * no peeled-off association on another socket.
1703 if ((sctp_style(sk, TCP) &&
1704 sctp_sstate(sk, ESTABLISHED)) ||
1705 sctp_endpoint_is_peeled_off(ep, &to)) {
1706 err = -EADDRNOTAVAIL;
1711 asoc = sctp_id2assoc(sk, associd);
1719 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1721 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1722 * socket that has an association in CLOSED state. This can
1723 * happen when an accepted socket has an association that is
1726 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1731 if (sinfo_flags & SCTP_EOF) {
1732 pr_debug("%s: shutting down association:%p\n",
1735 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1739 if (sinfo_flags & SCTP_ABORT) {
1741 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1747 pr_debug("%s: aborting association:%p\n",
1750 sctp_primitive_ABORT(net, asoc, chunk);
1756 /* Do we need to create the association? */
1758 pr_debug("%s: there is no association yet\n", __func__);
1760 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1765 /* Check for invalid stream against the stream counts,
1766 * either the default or the user specified stream counts.
1769 if (!sinit || !sinit->sinit_num_ostreams) {
1770 /* Check against the defaults. */
1771 if (sinfo->sinfo_stream >=
1772 sp->initmsg.sinit_num_ostreams) {
1777 /* Check against the requested. */
1778 if (sinfo->sinfo_stream >=
1779 sinit->sinit_num_ostreams) {
1787 * API 3.1.2 bind() - UDP Style Syntax
1788 * If a bind() or sctp_bindx() is not called prior to a
1789 * sendmsg() call that initiates a new association, the
1790 * system picks an ephemeral port and will choose an address
1791 * set equivalent to binding with a wildcard address.
1793 if (!ep->base.bind_addr.port) {
1794 if (sctp_autobind(sk)) {
1800 * If an unprivileged user inherits a one-to-many
1801 * style socket with open associations on a privileged
1802 * port, it MAY be permitted to accept new associations,
1803 * but it SHOULD NOT be permitted to open new
1806 if (ep->base.bind_addr.port < PROT_SOCK &&
1807 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1813 scope = sctp_scope(&to);
1814 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1820 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1826 /* If the SCTP_INIT ancillary data is specified, set all
1827 * the association init values accordingly.
1830 if (sinit->sinit_num_ostreams) {
1831 asoc->c.sinit_num_ostreams =
1832 sinit->sinit_num_ostreams;
1834 if (sinit->sinit_max_instreams) {
1835 asoc->c.sinit_max_instreams =
1836 sinit->sinit_max_instreams;
1838 if (sinit->sinit_max_attempts) {
1839 asoc->max_init_attempts
1840 = sinit->sinit_max_attempts;
1842 if (sinit->sinit_max_init_timeo) {
1843 asoc->max_init_timeo =
1844 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1848 /* Prime the peer's transport structures. */
1849 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1856 /* ASSERT: we have a valid association at this point. */
1857 pr_debug("%s: we have a valid association\n", __func__);
1860 /* If the user didn't specify SNDRCVINFO, make up one with
1863 memset(&default_sinfo, 0, sizeof(default_sinfo));
1864 default_sinfo.sinfo_stream = asoc->default_stream;
1865 default_sinfo.sinfo_flags = asoc->default_flags;
1866 default_sinfo.sinfo_ppid = asoc->default_ppid;
1867 default_sinfo.sinfo_context = asoc->default_context;
1868 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1869 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1870 sinfo = &default_sinfo;
1873 /* API 7.1.7, the sndbuf size per association bounds the
1874 * maximum size of data that can be sent in a single send call.
1876 if (msg_len > sk->sk_sndbuf) {
1881 if (asoc->pmtu_pending)
1882 sctp_assoc_pending_pmtu(sk, asoc);
1884 /* If fragmentation is disabled and the message length exceeds the
1885 * association fragmentation point, return EMSGSIZE. The I-D
1886 * does not specify what this error is, but this looks like
1889 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1894 /* Check for invalid stream. */
1895 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1900 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1901 if (!sctp_wspace(asoc)) {
1902 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1907 /* If an address is passed with the sendto/sendmsg call, it is used
1908 * to override the primary destination address in the TCP model, or
1909 * when SCTP_ADDR_OVER flag is set in the UDP model.
1911 if ((sctp_style(sk, TCP) && msg_name) ||
1912 (sinfo_flags & SCTP_ADDR_OVER)) {
1913 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1921 /* Auto-connect, if we aren't connected already. */
1922 if (sctp_state(asoc, CLOSED)) {
1923 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1927 pr_debug("%s: we associated primitively\n", __func__);
1930 /* Break the message into multiple chunks of maximum size. */
1931 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1932 if (IS_ERR(datamsg)) {
1933 err = PTR_ERR(datamsg);
1937 /* Now send the (possibly) fragmented message. */
1938 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1939 sctp_chunk_hold(chunk);
1941 /* Do accounting for the write space. */
1942 sctp_set_owner_w(chunk);
1944 chunk->transport = chunk_tp;
1947 /* Send it to the lower layers. Note: all chunks
1948 * must either fail or succeed. The lower layer
1949 * works that way today. Keep it that way or this
1952 err = sctp_primitive_SEND(net, asoc, datamsg);
1953 /* Did the lower layer accept the chunk? */
1955 sctp_datamsg_free(datamsg);
1959 pr_debug("%s: we sent primitively\n", __func__);
1961 sctp_datamsg_put(datamsg);
1964 /* If we are already past ASSOCIATE, the lower
1965 * layers are responsible for association cleanup.
1971 sctp_unhash_established(asoc);
1972 sctp_association_free(asoc);
1978 return sctp_error(sk, msg_flags, err);
1985 err = sock_error(sk);
1995 /* This is an extended version of skb_pull() that removes the data from the
1996 * start of a skb even when data is spread across the list of skb's in the
1997 * frag_list. len specifies the total amount of data that needs to be removed.
1998 * when 'len' bytes could be removed from the skb, it returns 0.
1999 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2000 * could not be removed.
2002 static int sctp_skb_pull(struct sk_buff *skb, int len)
2004 struct sk_buff *list;
2005 int skb_len = skb_headlen(skb);
2008 if (len <= skb_len) {
2009 __skb_pull(skb, len);
2013 __skb_pull(skb, skb_len);
2015 skb_walk_frags(skb, list) {
2016 rlen = sctp_skb_pull(list, len);
2017 skb->len -= (len-rlen);
2018 skb->data_len -= (len-rlen);
2029 /* API 3.1.3 recvmsg() - UDP Style Syntax
2031 * ssize_t recvmsg(int socket, struct msghdr *message,
2034 * socket - the socket descriptor of the endpoint.
2035 * message - pointer to the msghdr structure which contains a single
2036 * user message and possibly some ancillary data.
2038 * See Section 5 for complete description of the data
2041 * flags - flags sent or received with the user message, see Section
2042 * 5 for complete description of the flags.
2044 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2046 static int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2047 struct msghdr *msg, size_t len, int noblock,
2048 int flags, int *addr_len)
2050 struct sctp_ulpevent *event = NULL;
2051 struct sctp_sock *sp = sctp_sk(sk);
2052 struct sk_buff *skb;
2057 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2058 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2063 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2068 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2072 /* Get the total length of the skb including any skb's in the
2081 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2083 event = sctp_skb2event(skb);
2088 sock_recv_ts_and_drops(msg, sk, skb);
2089 if (sctp_ulpevent_is_notification(event)) {
2090 msg->msg_flags |= MSG_NOTIFICATION;
2091 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2093 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2096 /* Check if we allow SCTP_SNDRCVINFO. */
2097 if (sp->subscribe.sctp_data_io_event)
2098 sctp_ulpevent_read_sndrcvinfo(event, msg);
2100 /* FIXME: we should be calling IP/IPv6 layers. */
2101 if (sk->sk_protinfo.af_inet.cmsg_flags)
2102 ip_cmsg_recv(msg, skb);
2107 /* If skb's length exceeds the user's buffer, update the skb and
2108 * push it back to the receive_queue so that the next call to
2109 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2111 if (skb_len > copied) {
2112 msg->msg_flags &= ~MSG_EOR;
2113 if (flags & MSG_PEEK)
2115 sctp_skb_pull(skb, copied);
2116 skb_queue_head(&sk->sk_receive_queue, skb);
2118 /* When only partial message is copied to the user, increase
2119 * rwnd by that amount. If all the data in the skb is read,
2120 * rwnd is updated when the event is freed.
2122 if (!sctp_ulpevent_is_notification(event))
2123 sctp_assoc_rwnd_increase(event->asoc, copied);
2125 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2126 (event->msg_flags & MSG_EOR))
2127 msg->msg_flags |= MSG_EOR;
2129 msg->msg_flags &= ~MSG_EOR;
2132 if (flags & MSG_PEEK) {
2133 /* Release the skb reference acquired after peeking the skb in
2134 * sctp_skb_recv_datagram().
2138 /* Free the event which includes releasing the reference to
2139 * the owner of the skb, freeing the skb and updating the
2142 sctp_ulpevent_free(event);
2149 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2151 * This option is a on/off flag. If enabled no SCTP message
2152 * fragmentation will be performed. Instead if a message being sent
2153 * exceeds the current PMTU size, the message will NOT be sent and
2154 * instead a error will be indicated to the user.
2156 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2157 char __user *optval,
2158 unsigned int optlen)
2162 if (optlen < sizeof(int))
2165 if (get_user(val, (int __user *)optval))
2168 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2173 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2174 unsigned int optlen)
2176 struct sctp_association *asoc;
2177 struct sctp_ulpevent *event;
2179 if (optlen > sizeof(struct sctp_event_subscribe))
2181 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2185 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2186 * if there is no data to be sent or retransmit, the stack will
2187 * immediately send up this notification.
2189 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2190 &sctp_sk(sk)->subscribe)) {
2191 asoc = sctp_id2assoc(sk, 0);
2193 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2194 event = sctp_ulpevent_make_sender_dry_event(asoc,
2199 sctp_ulpq_tail_event(&asoc->ulpq, event);
2206 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2208 * This socket option is applicable to the UDP-style socket only. When
2209 * set it will cause associations that are idle for more than the
2210 * specified number of seconds to automatically close. An association
2211 * being idle is defined an association that has NOT sent or received
2212 * user data. The special value of '0' indicates that no automatic
2213 * close of any associations should be performed. The option expects an
2214 * integer defining the number of seconds of idle time before an
2215 * association is closed.
2217 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2218 unsigned int optlen)
2220 struct sctp_sock *sp = sctp_sk(sk);
2221 struct net *net = sock_net(sk);
2223 /* Applicable to UDP-style socket only */
2224 if (sctp_style(sk, TCP))
2226 if (optlen != sizeof(int))
2228 if (copy_from_user(&sp->autoclose, optval, optlen))
2231 if (sp->autoclose > net->sctp.max_autoclose)
2232 sp->autoclose = net->sctp.max_autoclose;
2237 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2239 * Applications can enable or disable heartbeats for any peer address of
2240 * an association, modify an address's heartbeat interval, force a
2241 * heartbeat to be sent immediately, and adjust the address's maximum
2242 * number of retransmissions sent before an address is considered
2243 * unreachable. The following structure is used to access and modify an
2244 * address's parameters:
2246 * struct sctp_paddrparams {
2247 * sctp_assoc_t spp_assoc_id;
2248 * struct sockaddr_storage spp_address;
2249 * uint32_t spp_hbinterval;
2250 * uint16_t spp_pathmaxrxt;
2251 * uint32_t spp_pathmtu;
2252 * uint32_t spp_sackdelay;
2253 * uint32_t spp_flags;
2256 * spp_assoc_id - (one-to-many style socket) This is filled in the
2257 * application, and identifies the association for
2259 * spp_address - This specifies which address is of interest.
2260 * spp_hbinterval - This contains the value of the heartbeat interval,
2261 * in milliseconds. If a value of zero
2262 * is present in this field then no changes are to
2263 * be made to this parameter.
2264 * spp_pathmaxrxt - This contains the maximum number of
2265 * retransmissions before this address shall be
2266 * considered unreachable. If a value of zero
2267 * is present in this field then no changes are to
2268 * be made to this parameter.
2269 * spp_pathmtu - When Path MTU discovery is disabled the value
2270 * specified here will be the "fixed" path mtu.
2271 * Note that if the spp_address field is empty
2272 * then all associations on this address will
2273 * have this fixed path mtu set upon them.
2275 * spp_sackdelay - When delayed sack is enabled, this value specifies
2276 * the number of milliseconds that sacks will be delayed
2277 * for. This value will apply to all addresses of an
2278 * association if the spp_address field is empty. Note
2279 * also, that if delayed sack is enabled and this
2280 * value is set to 0, no change is made to the last
2281 * recorded delayed sack timer value.
2283 * spp_flags - These flags are used to control various features
2284 * on an association. The flag field may contain
2285 * zero or more of the following options.
2287 * SPP_HB_ENABLE - Enable heartbeats on the
2288 * specified address. Note that if the address
2289 * field is empty all addresses for the association
2290 * have heartbeats enabled upon them.
2292 * SPP_HB_DISABLE - Disable heartbeats on the
2293 * speicifed address. Note that if the address
2294 * field is empty all addresses for the association
2295 * will have their heartbeats disabled. Note also
2296 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2297 * mutually exclusive, only one of these two should
2298 * be specified. Enabling both fields will have
2299 * undetermined results.
2301 * SPP_HB_DEMAND - Request a user initiated heartbeat
2302 * to be made immediately.
2304 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2305 * heartbeat delayis to be set to the value of 0
2308 * SPP_PMTUD_ENABLE - This field will enable PMTU
2309 * discovery upon the specified address. Note that
2310 * if the address feild is empty then all addresses
2311 * on the association are effected.
2313 * SPP_PMTUD_DISABLE - This field will disable PMTU
2314 * discovery upon the specified address. Note that
2315 * if the address feild is empty then all addresses
2316 * on the association are effected. Not also that
2317 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2318 * exclusive. Enabling both will have undetermined
2321 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2322 * on delayed sack. The time specified in spp_sackdelay
2323 * is used to specify the sack delay for this address. Note
2324 * that if spp_address is empty then all addresses will
2325 * enable delayed sack and take on the sack delay
2326 * value specified in spp_sackdelay.
2327 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2328 * off delayed sack. If the spp_address field is blank then
2329 * delayed sack is disabled for the entire association. Note
2330 * also that this field is mutually exclusive to
2331 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2334 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2335 struct sctp_transport *trans,
2336 struct sctp_association *asoc,
2337 struct sctp_sock *sp,
2340 int sackdelay_change)
2344 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2345 struct net *net = sock_net(trans->asoc->base.sk);
2347 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2352 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2353 * this field is ignored. Note also that a value of zero indicates
2354 * the current setting should be left unchanged.
2356 if (params->spp_flags & SPP_HB_ENABLE) {
2358 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2359 * set. This lets us use 0 value when this flag
2362 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2363 params->spp_hbinterval = 0;
2365 if (params->spp_hbinterval ||
2366 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2369 msecs_to_jiffies(params->spp_hbinterval);
2372 msecs_to_jiffies(params->spp_hbinterval);
2374 sp->hbinterval = params->spp_hbinterval;
2381 trans->param_flags =
2382 (trans->param_flags & ~SPP_HB) | hb_change;
2385 (asoc->param_flags & ~SPP_HB) | hb_change;
2388 (sp->param_flags & ~SPP_HB) | hb_change;
2392 /* When Path MTU discovery is disabled the value specified here will
2393 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2394 * include the flag SPP_PMTUD_DISABLE for this field to have any
2397 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2399 trans->pathmtu = params->spp_pathmtu;
2400 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2402 asoc->pathmtu = params->spp_pathmtu;
2403 sctp_frag_point(asoc, params->spp_pathmtu);
2405 sp->pathmtu = params->spp_pathmtu;
2411 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2412 (params->spp_flags & SPP_PMTUD_ENABLE);
2413 trans->param_flags =
2414 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2416 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2417 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2421 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2424 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2428 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2429 * value of this field is ignored. Note also that a value of zero
2430 * indicates the current setting should be left unchanged.
2432 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2435 msecs_to_jiffies(params->spp_sackdelay);
2438 msecs_to_jiffies(params->spp_sackdelay);
2440 sp->sackdelay = params->spp_sackdelay;
2444 if (sackdelay_change) {
2446 trans->param_flags =
2447 (trans->param_flags & ~SPP_SACKDELAY) |
2451 (asoc->param_flags & ~SPP_SACKDELAY) |
2455 (sp->param_flags & ~SPP_SACKDELAY) |
2460 /* Note that a value of zero indicates the current setting should be
2463 if (params->spp_pathmaxrxt) {
2465 trans->pathmaxrxt = params->spp_pathmaxrxt;
2467 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2469 sp->pathmaxrxt = params->spp_pathmaxrxt;
2476 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2477 char __user *optval,
2478 unsigned int optlen)
2480 struct sctp_paddrparams params;
2481 struct sctp_transport *trans = NULL;
2482 struct sctp_association *asoc = NULL;
2483 struct sctp_sock *sp = sctp_sk(sk);
2485 int hb_change, pmtud_change, sackdelay_change;
2487 if (optlen != sizeof(struct sctp_paddrparams))
2490 if (copy_from_user(¶ms, optval, optlen))
2493 /* Validate flags and value parameters. */
2494 hb_change = params.spp_flags & SPP_HB;
2495 pmtud_change = params.spp_flags & SPP_PMTUD;
2496 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2498 if (hb_change == SPP_HB ||
2499 pmtud_change == SPP_PMTUD ||
2500 sackdelay_change == SPP_SACKDELAY ||
2501 params.spp_sackdelay > 500 ||
2502 (params.spp_pathmtu &&
2503 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2506 /* If an address other than INADDR_ANY is specified, and
2507 * no transport is found, then the request is invalid.
2509 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
2510 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2511 params.spp_assoc_id);
2516 /* Get association, if assoc_id != 0 and the socket is a one
2517 * to many style socket, and an association was not found, then
2518 * the id was invalid.
2520 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2521 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2524 /* Heartbeat demand can only be sent on a transport or
2525 * association, but not a socket.
2527 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2530 /* Process parameters. */
2531 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2532 hb_change, pmtud_change,
2538 /* If changes are for association, also apply parameters to each
2541 if (!trans && asoc) {
2542 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2544 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2545 hb_change, pmtud_change,
2553 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2555 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2558 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2560 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2564 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2566 * This option will effect the way delayed acks are performed. This
2567 * option allows you to get or set the delayed ack time, in
2568 * milliseconds. It also allows changing the delayed ack frequency.
2569 * Changing the frequency to 1 disables the delayed sack algorithm. If
2570 * the assoc_id is 0, then this sets or gets the endpoints default
2571 * values. If the assoc_id field is non-zero, then the set or get
2572 * effects the specified association for the one to many model (the
2573 * assoc_id field is ignored by the one to one model). Note that if
2574 * sack_delay or sack_freq are 0 when setting this option, then the
2575 * current values will remain unchanged.
2577 * struct sctp_sack_info {
2578 * sctp_assoc_t sack_assoc_id;
2579 * uint32_t sack_delay;
2580 * uint32_t sack_freq;
2583 * sack_assoc_id - This parameter, indicates which association the user
2584 * is performing an action upon. Note that if this field's value is
2585 * zero then the endpoints default value is changed (effecting future
2586 * associations only).
2588 * sack_delay - This parameter contains the number of milliseconds that
2589 * the user is requesting the delayed ACK timer be set to. Note that
2590 * this value is defined in the standard to be between 200 and 500
2593 * sack_freq - This parameter contains the number of packets that must
2594 * be received before a sack is sent without waiting for the delay
2595 * timer to expire. The default value for this is 2, setting this
2596 * value to 1 will disable the delayed sack algorithm.
2599 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2600 char __user *optval, unsigned int optlen)
2602 struct sctp_sack_info params;
2603 struct sctp_transport *trans = NULL;
2604 struct sctp_association *asoc = NULL;
2605 struct sctp_sock *sp = sctp_sk(sk);
2607 if (optlen == sizeof(struct sctp_sack_info)) {
2608 if (copy_from_user(¶ms, optval, optlen))
2611 if (params.sack_delay == 0 && params.sack_freq == 0)
2613 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2614 pr_warn_ratelimited(DEPRECATED
2616 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2617 "Use struct sctp_sack_info instead\n",
2618 current->comm, task_pid_nr(current));
2619 if (copy_from_user(¶ms, optval, optlen))
2622 if (params.sack_delay == 0)
2623 params.sack_freq = 1;
2625 params.sack_freq = 0;
2629 /* Validate value parameter. */
2630 if (params.sack_delay > 500)
2633 /* Get association, if sack_assoc_id != 0 and the socket is a one
2634 * to many style socket, and an association was not found, then
2635 * the id was invalid.
2637 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2638 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2641 if (params.sack_delay) {
2644 msecs_to_jiffies(params.sack_delay);
2646 sctp_spp_sackdelay_enable(asoc->param_flags);
2648 sp->sackdelay = params.sack_delay;
2650 sctp_spp_sackdelay_enable(sp->param_flags);
2654 if (params.sack_freq == 1) {
2657 sctp_spp_sackdelay_disable(asoc->param_flags);
2660 sctp_spp_sackdelay_disable(sp->param_flags);
2662 } else if (params.sack_freq > 1) {
2664 asoc->sackfreq = params.sack_freq;
2666 sctp_spp_sackdelay_enable(asoc->param_flags);
2668 sp->sackfreq = params.sack_freq;
2670 sctp_spp_sackdelay_enable(sp->param_flags);
2674 /* If change is for association, also apply to each transport. */
2676 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2678 if (params.sack_delay) {
2680 msecs_to_jiffies(params.sack_delay);
2681 trans->param_flags =
2682 sctp_spp_sackdelay_enable(trans->param_flags);
2684 if (params.sack_freq == 1) {
2685 trans->param_flags =
2686 sctp_spp_sackdelay_disable(trans->param_flags);
2687 } else if (params.sack_freq > 1) {
2688 trans->sackfreq = params.sack_freq;
2689 trans->param_flags =
2690 sctp_spp_sackdelay_enable(trans->param_flags);
2698 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2700 * Applications can specify protocol parameters for the default association
2701 * initialization. The option name argument to setsockopt() and getsockopt()
2704 * Setting initialization parameters is effective only on an unconnected
2705 * socket (for UDP-style sockets only future associations are effected
2706 * by the change). With TCP-style sockets, this option is inherited by
2707 * sockets derived from a listener socket.
2709 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2711 struct sctp_initmsg sinit;
2712 struct sctp_sock *sp = sctp_sk(sk);
2714 if (optlen != sizeof(struct sctp_initmsg))
2716 if (copy_from_user(&sinit, optval, optlen))
2719 if (sinit.sinit_num_ostreams)
2720 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2721 if (sinit.sinit_max_instreams)
2722 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2723 if (sinit.sinit_max_attempts)
2724 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2725 if (sinit.sinit_max_init_timeo)
2726 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2732 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2734 * Applications that wish to use the sendto() system call may wish to
2735 * specify a default set of parameters that would normally be supplied
2736 * through the inclusion of ancillary data. This socket option allows
2737 * such an application to set the default sctp_sndrcvinfo structure.
2738 * The application that wishes to use this socket option simply passes
2739 * in to this call the sctp_sndrcvinfo structure defined in Section
2740 * 5.2.2) The input parameters accepted by this call include
2741 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2742 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2743 * to this call if the caller is using the UDP model.
2745 static int sctp_setsockopt_default_send_param(struct sock *sk,
2746 char __user *optval,
2747 unsigned int optlen)
2749 struct sctp_sndrcvinfo info;
2750 struct sctp_association *asoc;
2751 struct sctp_sock *sp = sctp_sk(sk);
2753 if (optlen != sizeof(struct sctp_sndrcvinfo))
2755 if (copy_from_user(&info, optval, optlen))
2758 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2759 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2763 asoc->default_stream = info.sinfo_stream;
2764 asoc->default_flags = info.sinfo_flags;
2765 asoc->default_ppid = info.sinfo_ppid;
2766 asoc->default_context = info.sinfo_context;
2767 asoc->default_timetolive = info.sinfo_timetolive;
2769 sp->default_stream = info.sinfo_stream;
2770 sp->default_flags = info.sinfo_flags;
2771 sp->default_ppid = info.sinfo_ppid;
2772 sp->default_context = info.sinfo_context;
2773 sp->default_timetolive = info.sinfo_timetolive;
2779 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2781 * Requests that the local SCTP stack use the enclosed peer address as
2782 * the association primary. The enclosed address must be one of the
2783 * association peer's addresses.
2785 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2786 unsigned int optlen)
2788 struct sctp_prim prim;
2789 struct sctp_transport *trans;
2791 if (optlen != sizeof(struct sctp_prim))
2794 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2797 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2801 sctp_assoc_set_primary(trans->asoc, trans);
2807 * 7.1.5 SCTP_NODELAY
2809 * Turn on/off any Nagle-like algorithm. This means that packets are
2810 * generally sent as soon as possible and no unnecessary delays are
2811 * introduced, at the cost of more packets in the network. Expects an
2812 * integer boolean flag.
2814 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2815 unsigned int optlen)
2819 if (optlen < sizeof(int))
2821 if (get_user(val, (int __user *)optval))
2824 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2830 * 7.1.1 SCTP_RTOINFO
2832 * The protocol parameters used to initialize and bound retransmission
2833 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2834 * and modify these parameters.
2835 * All parameters are time values, in milliseconds. A value of 0, when
2836 * modifying the parameters, indicates that the current value should not
2840 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2842 struct sctp_rtoinfo rtoinfo;
2843 struct sctp_association *asoc;
2844 unsigned long rto_min, rto_max;
2845 struct sctp_sock *sp = sctp_sk(sk);
2847 if (optlen != sizeof (struct sctp_rtoinfo))
2850 if (copy_from_user(&rtoinfo, optval, optlen))
2853 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2855 /* Set the values to the specific association */
2856 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2859 rto_max = rtoinfo.srto_max;
2860 rto_min = rtoinfo.srto_min;
2863 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2865 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2868 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2870 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2872 if (rto_min > rto_max)
2876 if (rtoinfo.srto_initial != 0)
2878 msecs_to_jiffies(rtoinfo.srto_initial);
2879 asoc->rto_max = rto_max;
2880 asoc->rto_min = rto_min;
2882 /* If there is no association or the association-id = 0
2883 * set the values to the endpoint.
2885 if (rtoinfo.srto_initial != 0)
2886 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2887 sp->rtoinfo.srto_max = rto_max;
2888 sp->rtoinfo.srto_min = rto_min;
2896 * 7.1.2 SCTP_ASSOCINFO
2898 * This option is used to tune the maximum retransmission attempts
2899 * of the association.
2900 * Returns an error if the new association retransmission value is
2901 * greater than the sum of the retransmission value of the peer.
2902 * See [SCTP] for more information.
2905 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2908 struct sctp_assocparams assocparams;
2909 struct sctp_association *asoc;
2911 if (optlen != sizeof(struct sctp_assocparams))
2913 if (copy_from_user(&assocparams, optval, optlen))
2916 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2918 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2921 /* Set the values to the specific association */
2923 if (assocparams.sasoc_asocmaxrxt != 0) {
2926 struct sctp_transport *peer_addr;
2928 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2930 path_sum += peer_addr->pathmaxrxt;
2934 /* Only validate asocmaxrxt if we have more than
2935 * one path/transport. We do this because path
2936 * retransmissions are only counted when we have more
2940 assocparams.sasoc_asocmaxrxt > path_sum)
2943 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2946 if (assocparams.sasoc_cookie_life != 0)
2947 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
2949 /* Set the values to the endpoint */
2950 struct sctp_sock *sp = sctp_sk(sk);
2952 if (assocparams.sasoc_asocmaxrxt != 0)
2953 sp->assocparams.sasoc_asocmaxrxt =
2954 assocparams.sasoc_asocmaxrxt;
2955 if (assocparams.sasoc_cookie_life != 0)
2956 sp->assocparams.sasoc_cookie_life =
2957 assocparams.sasoc_cookie_life;
2963 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2965 * This socket option is a boolean flag which turns on or off mapped V4
2966 * addresses. If this option is turned on and the socket is type
2967 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2968 * If this option is turned off, then no mapping will be done of V4
2969 * addresses and a user will receive both PF_INET6 and PF_INET type
2970 * addresses on the socket.
2972 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2975 struct sctp_sock *sp = sctp_sk(sk);
2977 if (optlen < sizeof(int))
2979 if (get_user(val, (int __user *)optval))
2990 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2991 * This option will get or set the maximum size to put in any outgoing
2992 * SCTP DATA chunk. If a message is larger than this size it will be
2993 * fragmented by SCTP into the specified size. Note that the underlying
2994 * SCTP implementation may fragment into smaller sized chunks when the
2995 * PMTU of the underlying association is smaller than the value set by
2996 * the user. The default value for this option is '0' which indicates
2997 * the user is NOT limiting fragmentation and only the PMTU will effect
2998 * SCTP's choice of DATA chunk size. Note also that values set larger
2999 * than the maximum size of an IP datagram will effectively let SCTP
3000 * control fragmentation (i.e. the same as setting this option to 0).
3002 * The following structure is used to access and modify this parameter:
3004 * struct sctp_assoc_value {
3005 * sctp_assoc_t assoc_id;
3006 * uint32_t assoc_value;
3009 * assoc_id: This parameter is ignored for one-to-one style sockets.
3010 * For one-to-many style sockets this parameter indicates which
3011 * association the user is performing an action upon. Note that if
3012 * this field's value is zero then the endpoints default value is
3013 * changed (effecting future associations only).
3014 * assoc_value: This parameter specifies the maximum size in bytes.
3016 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3018 struct sctp_assoc_value params;
3019 struct sctp_association *asoc;
3020 struct sctp_sock *sp = sctp_sk(sk);
3023 if (optlen == sizeof(int)) {
3024 pr_warn_ratelimited(DEPRECATED
3026 "Use of int in maxseg socket option.\n"
3027 "Use struct sctp_assoc_value instead\n",
3028 current->comm, task_pid_nr(current));
3029 if (copy_from_user(&val, optval, optlen))
3031 params.assoc_id = 0;
3032 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3033 if (copy_from_user(¶ms, optval, optlen))
3035 val = params.assoc_value;
3039 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3042 asoc = sctp_id2assoc(sk, params.assoc_id);
3043 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3048 val = asoc->pathmtu;
3049 val -= sp->pf->af->net_header_len;
3050 val -= sizeof(struct sctphdr) +
3051 sizeof(struct sctp_data_chunk);
3053 asoc->user_frag = val;
3054 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3056 sp->user_frag = val;
3064 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3066 * Requests that the peer mark the enclosed address as the association
3067 * primary. The enclosed address must be one of the association's
3068 * locally bound addresses. The following structure is used to make a
3069 * set primary request:
3071 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3072 unsigned int optlen)
3074 struct net *net = sock_net(sk);
3075 struct sctp_sock *sp;
3076 struct sctp_association *asoc = NULL;
3077 struct sctp_setpeerprim prim;
3078 struct sctp_chunk *chunk;
3084 if (!net->sctp.addip_enable)
3087 if (optlen != sizeof(struct sctp_setpeerprim))
3090 if (copy_from_user(&prim, optval, optlen))
3093 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3097 if (!asoc->peer.asconf_capable)
3100 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3103 if (!sctp_state(asoc, ESTABLISHED))
3106 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3110 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3111 return -EADDRNOTAVAIL;
3113 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3114 return -EADDRNOTAVAIL;
3116 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3117 chunk = sctp_make_asconf_set_prim(asoc,
3118 (union sctp_addr *)&prim.sspp_addr);
3122 err = sctp_send_asconf(asoc, chunk);
3124 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3129 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3130 unsigned int optlen)
3132 struct sctp_setadaptation adaptation;
3134 if (optlen != sizeof(struct sctp_setadaptation))
3136 if (copy_from_user(&adaptation, optval, optlen))
3139 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3145 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3147 * The context field in the sctp_sndrcvinfo structure is normally only
3148 * used when a failed message is retrieved holding the value that was
3149 * sent down on the actual send call. This option allows the setting of
3150 * a default context on an association basis that will be received on
3151 * reading messages from the peer. This is especially helpful in the
3152 * one-2-many model for an application to keep some reference to an
3153 * internal state machine that is processing messages on the
3154 * association. Note that the setting of this value only effects
3155 * received messages from the peer and does not effect the value that is
3156 * saved with outbound messages.
3158 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3159 unsigned int optlen)
3161 struct sctp_assoc_value params;
3162 struct sctp_sock *sp;
3163 struct sctp_association *asoc;
3165 if (optlen != sizeof(struct sctp_assoc_value))
3167 if (copy_from_user(¶ms, optval, optlen))
3172 if (params.assoc_id != 0) {
3173 asoc = sctp_id2assoc(sk, params.assoc_id);
3176 asoc->default_rcv_context = params.assoc_value;
3178 sp->default_rcv_context = params.assoc_value;
3185 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3187 * This options will at a minimum specify if the implementation is doing
3188 * fragmented interleave. Fragmented interleave, for a one to many
3189 * socket, is when subsequent calls to receive a message may return
3190 * parts of messages from different associations. Some implementations
3191 * may allow you to turn this value on or off. If so, when turned off,
3192 * no fragment interleave will occur (which will cause a head of line
3193 * blocking amongst multiple associations sharing the same one to many
3194 * socket). When this option is turned on, then each receive call may
3195 * come from a different association (thus the user must receive data
3196 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3197 * association each receive belongs to.
3199 * This option takes a boolean value. A non-zero value indicates that
3200 * fragmented interleave is on. A value of zero indicates that
3201 * fragmented interleave is off.
3203 * Note that it is important that an implementation that allows this
3204 * option to be turned on, have it off by default. Otherwise an unaware
3205 * application using the one to many model may become confused and act
3208 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3209 char __user *optval,
3210 unsigned int optlen)
3214 if (optlen != sizeof(int))
3216 if (get_user(val, (int __user *)optval))
3219 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3225 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3226 * (SCTP_PARTIAL_DELIVERY_POINT)
3228 * This option will set or get the SCTP partial delivery point. This
3229 * point is the size of a message where the partial delivery API will be
3230 * invoked to help free up rwnd space for the peer. Setting this to a
3231 * lower value will cause partial deliveries to happen more often. The
3232 * calls argument is an integer that sets or gets the partial delivery
3233 * point. Note also that the call will fail if the user attempts to set
3234 * this value larger than the socket receive buffer size.
3236 * Note that any single message having a length smaller than or equal to
3237 * the SCTP partial delivery point will be delivered in one single read
3238 * call as long as the user provided buffer is large enough to hold the
3241 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3242 char __user *optval,
3243 unsigned int optlen)
3247 if (optlen != sizeof(u32))
3249 if (get_user(val, (int __user *)optval))
3252 /* Note: We double the receive buffer from what the user sets
3253 * it to be, also initial rwnd is based on rcvbuf/2.
3255 if (val > (sk->sk_rcvbuf >> 1))
3258 sctp_sk(sk)->pd_point = val;
3260 return 0; /* is this the right error code? */
3264 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3266 * This option will allow a user to change the maximum burst of packets
3267 * that can be emitted by this association. Note that the default value
3268 * is 4, and some implementations may restrict this setting so that it
3269 * can only be lowered.
3271 * NOTE: This text doesn't seem right. Do this on a socket basis with
3272 * future associations inheriting the socket value.
3274 static int sctp_setsockopt_maxburst(struct sock *sk,
3275 char __user *optval,
3276 unsigned int optlen)
3278 struct sctp_assoc_value params;
3279 struct sctp_sock *sp;
3280 struct sctp_association *asoc;
3284 if (optlen == sizeof(int)) {
3285 pr_warn_ratelimited(DEPRECATED
3287 "Use of int in max_burst socket option deprecated.\n"
3288 "Use struct sctp_assoc_value instead\n",
3289 current->comm, task_pid_nr(current));
3290 if (copy_from_user(&val, optval, optlen))
3292 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3293 if (copy_from_user(¶ms, optval, optlen))
3295 val = params.assoc_value;
3296 assoc_id = params.assoc_id;
3302 if (assoc_id != 0) {
3303 asoc = sctp_id2assoc(sk, assoc_id);
3306 asoc->max_burst = val;
3308 sp->max_burst = val;
3314 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3316 * This set option adds a chunk type that the user is requesting to be
3317 * received only in an authenticated way. Changes to the list of chunks
3318 * will only effect future associations on the socket.
3320 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3321 char __user *optval,
3322 unsigned int optlen)
3324 struct net *net = sock_net(sk);
3325 struct sctp_authchunk val;
3327 if (!net->sctp.auth_enable)
3330 if (optlen != sizeof(struct sctp_authchunk))
3332 if (copy_from_user(&val, optval, optlen))
3335 switch (val.sauth_chunk) {
3337 case SCTP_CID_INIT_ACK:
3338 case SCTP_CID_SHUTDOWN_COMPLETE:
3343 /* add this chunk id to the endpoint */
3344 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3348 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3350 * This option gets or sets the list of HMAC algorithms that the local
3351 * endpoint requires the peer to use.
3353 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3354 char __user *optval,
3355 unsigned int optlen)
3357 struct net *net = sock_net(sk);
3358 struct sctp_hmacalgo *hmacs;
3362 if (!net->sctp.auth_enable)
3365 if (optlen < sizeof(struct sctp_hmacalgo))
3368 hmacs = memdup_user(optval, optlen);
3370 return PTR_ERR(hmacs);
3372 idents = hmacs->shmac_num_idents;
3373 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3374 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3379 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3386 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3388 * This option will set a shared secret key which is used to build an
3389 * association shared key.
3391 static int sctp_setsockopt_auth_key(struct sock *sk,
3392 char __user *optval,
3393 unsigned int optlen)
3395 struct net *net = sock_net(sk);
3396 struct sctp_authkey *authkey;
3397 struct sctp_association *asoc;
3400 if (!net->sctp.auth_enable)
3403 if (optlen <= sizeof(struct sctp_authkey))
3406 authkey = memdup_user(optval, optlen);
3407 if (IS_ERR(authkey))
3408 return PTR_ERR(authkey);
3410 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3415 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3416 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3421 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3428 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3430 * This option will get or set the active shared key to be used to build
3431 * the association shared key.
3433 static int sctp_setsockopt_active_key(struct sock *sk,
3434 char __user *optval,
3435 unsigned int optlen)
3437 struct net *net = sock_net(sk);
3438 struct sctp_authkeyid val;
3439 struct sctp_association *asoc;
3441 if (!net->sctp.auth_enable)
3444 if (optlen != sizeof(struct sctp_authkeyid))
3446 if (copy_from_user(&val, optval, optlen))
3449 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3450 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3453 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3454 val.scact_keynumber);
3458 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3460 * This set option will delete a shared secret key from use.
3462 static int sctp_setsockopt_del_key(struct sock *sk,
3463 char __user *optval,
3464 unsigned int optlen)
3466 struct net *net = sock_net(sk);
3467 struct sctp_authkeyid val;
3468 struct sctp_association *asoc;
3470 if (!net->sctp.auth_enable)
3473 if (optlen != sizeof(struct sctp_authkeyid))
3475 if (copy_from_user(&val, optval, optlen))
3478 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3479 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3482 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3483 val.scact_keynumber);
3488 * 8.1.23 SCTP_AUTO_ASCONF
3490 * This option will enable or disable the use of the automatic generation of
3491 * ASCONF chunks to add and delete addresses to an existing association. Note
3492 * that this option has two caveats namely: a) it only affects sockets that
3493 * are bound to all addresses available to the SCTP stack, and b) the system
3494 * administrator may have an overriding control that turns the ASCONF feature
3495 * off no matter what setting the socket option may have.
3496 * This option expects an integer boolean flag, where a non-zero value turns on
3497 * the option, and a zero value turns off the option.
3498 * Note. In this implementation, socket operation overrides default parameter
3499 * being set by sysctl as well as FreeBSD implementation
3501 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3502 unsigned int optlen)
3505 struct sctp_sock *sp = sctp_sk(sk);
3507 if (optlen < sizeof(int))
3509 if (get_user(val, (int __user *)optval))
3511 if (!sctp_is_ep_boundall(sk) && val)
3513 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3516 if (val == 0 && sp->do_auto_asconf) {
3517 list_del(&sp->auto_asconf_list);
3518 sp->do_auto_asconf = 0;
3519 } else if (val && !sp->do_auto_asconf) {
3520 list_add_tail(&sp->auto_asconf_list,
3521 &sock_net(sk)->sctp.auto_asconf_splist);
3522 sp->do_auto_asconf = 1;
3529 * SCTP_PEER_ADDR_THLDS
3531 * This option allows us to alter the partially failed threshold for one or all
3532 * transports in an association. See Section 6.1 of:
3533 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3535 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3536 char __user *optval,
3537 unsigned int optlen)
3539 struct sctp_paddrthlds val;
3540 struct sctp_transport *trans;
3541 struct sctp_association *asoc;
3543 if (optlen < sizeof(struct sctp_paddrthlds))
3545 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3546 sizeof(struct sctp_paddrthlds)))
3550 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3551 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3554 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3556 if (val.spt_pathmaxrxt)
3557 trans->pathmaxrxt = val.spt_pathmaxrxt;
3558 trans->pf_retrans = val.spt_pathpfthld;
3561 if (val.spt_pathmaxrxt)
3562 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3563 asoc->pf_retrans = val.spt_pathpfthld;
3565 trans = sctp_addr_id2transport(sk, &val.spt_address,
3570 if (val.spt_pathmaxrxt)
3571 trans->pathmaxrxt = val.spt_pathmaxrxt;
3572 trans->pf_retrans = val.spt_pathpfthld;
3578 /* API 6.2 setsockopt(), getsockopt()
3580 * Applications use setsockopt() and getsockopt() to set or retrieve
3581 * socket options. Socket options are used to change the default
3582 * behavior of sockets calls. They are described in Section 7.
3586 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3587 * int __user *optlen);
3588 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3591 * sd - the socket descript.
3592 * level - set to IPPROTO_SCTP for all SCTP options.
3593 * optname - the option name.
3594 * optval - the buffer to store the value of the option.
3595 * optlen - the size of the buffer.
3597 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3598 char __user *optval, unsigned int optlen)
3602 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3604 /* I can hardly begin to describe how wrong this is. This is
3605 * so broken as to be worse than useless. The API draft
3606 * REALLY is NOT helpful here... I am not convinced that the
3607 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3608 * are at all well-founded.
3610 if (level != SOL_SCTP) {
3611 struct sctp_af *af = sctp_sk(sk)->pf->af;
3612 retval = af->setsockopt(sk, level, optname, optval, optlen);
3619 case SCTP_SOCKOPT_BINDX_ADD:
3620 /* 'optlen' is the size of the addresses buffer. */
3621 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3622 optlen, SCTP_BINDX_ADD_ADDR);
3625 case SCTP_SOCKOPT_BINDX_REM:
3626 /* 'optlen' is the size of the addresses buffer. */
3627 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3628 optlen, SCTP_BINDX_REM_ADDR);
3631 case SCTP_SOCKOPT_CONNECTX_OLD:
3632 /* 'optlen' is the size of the addresses buffer. */
3633 retval = sctp_setsockopt_connectx_old(sk,
3634 (struct sockaddr __user *)optval,
3638 case SCTP_SOCKOPT_CONNECTX:
3639 /* 'optlen' is the size of the addresses buffer. */
3640 retval = sctp_setsockopt_connectx(sk,
3641 (struct sockaddr __user *)optval,
3645 case SCTP_DISABLE_FRAGMENTS:
3646 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3650 retval = sctp_setsockopt_events(sk, optval, optlen);
3653 case SCTP_AUTOCLOSE:
3654 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3657 case SCTP_PEER_ADDR_PARAMS:
3658 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3661 case SCTP_DELAYED_SACK:
3662 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3664 case SCTP_PARTIAL_DELIVERY_POINT:
3665 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3669 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3671 case SCTP_DEFAULT_SEND_PARAM:
3672 retval = sctp_setsockopt_default_send_param(sk, optval,
3675 case SCTP_PRIMARY_ADDR:
3676 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3678 case SCTP_SET_PEER_PRIMARY_ADDR:
3679 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3682 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3685 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3687 case SCTP_ASSOCINFO:
3688 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3690 case SCTP_I_WANT_MAPPED_V4_ADDR:
3691 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3694 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3696 case SCTP_ADAPTATION_LAYER:
3697 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3700 retval = sctp_setsockopt_context(sk, optval, optlen);
3702 case SCTP_FRAGMENT_INTERLEAVE:
3703 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3705 case SCTP_MAX_BURST:
3706 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3708 case SCTP_AUTH_CHUNK:
3709 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3711 case SCTP_HMAC_IDENT:
3712 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3715 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3717 case SCTP_AUTH_ACTIVE_KEY:
3718 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3720 case SCTP_AUTH_DELETE_KEY:
3721 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3723 case SCTP_AUTO_ASCONF:
3724 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3726 case SCTP_PEER_ADDR_THLDS:
3727 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3730 retval = -ENOPROTOOPT;
3740 /* API 3.1.6 connect() - UDP Style Syntax
3742 * An application may use the connect() call in the UDP model to initiate an
3743 * association without sending data.
3747 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3749 * sd: the socket descriptor to have a new association added to.
3751 * nam: the address structure (either struct sockaddr_in or struct
3752 * sockaddr_in6 defined in RFC2553 [7]).
3754 * len: the size of the address.
3756 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3764 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
3767 /* Validate addr_len before calling common connect/connectx routine. */
3768 af = sctp_get_af_specific(addr->sa_family);
3769 if (!af || addr_len < af->sockaddr_len) {
3772 /* Pass correct addr len to common routine (so it knows there
3773 * is only one address being passed.
3775 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3782 /* FIXME: Write comments. */
3783 static int sctp_disconnect(struct sock *sk, int flags)
3785 return -EOPNOTSUPP; /* STUB */
3788 /* 4.1.4 accept() - TCP Style Syntax
3790 * Applications use accept() call to remove an established SCTP
3791 * association from the accept queue of the endpoint. A new socket
3792 * descriptor will be returned from accept() to represent the newly
3793 * formed association.
3795 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3797 struct sctp_sock *sp;
3798 struct sctp_endpoint *ep;
3799 struct sock *newsk = NULL;
3800 struct sctp_association *asoc;
3809 if (!sctp_style(sk, TCP)) {
3810 error = -EOPNOTSUPP;
3814 if (!sctp_sstate(sk, LISTENING)) {
3819 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3821 error = sctp_wait_for_accept(sk, timeo);
3825 /* We treat the list of associations on the endpoint as the accept
3826 * queue and pick the first association on the list.
3828 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3830 newsk = sp->pf->create_accept_sk(sk, asoc);
3836 /* Populate the fields of the newsk from the oldsk and migrate the
3837 * asoc to the newsk.
3839 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3847 /* The SCTP ioctl handler. */
3848 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3855 * SEQPACKET-style sockets in LISTENING state are valid, for
3856 * SCTP, so only discard TCP-style sockets in LISTENING state.
3858 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3863 struct sk_buff *skb;
3864 unsigned int amount = 0;
3866 skb = skb_peek(&sk->sk_receive_queue);
3869 * We will only return the amount of this packet since
3870 * that is all that will be read.
3874 rc = put_user(amount, (int __user *)arg);
3886 /* This is the function which gets called during socket creation to
3887 * initialized the SCTP-specific portion of the sock.
3888 * The sock structure should already be zero-filled memory.
3890 static int sctp_init_sock(struct sock *sk)
3892 struct net *net = sock_net(sk);
3893 struct sctp_sock *sp;
3895 pr_debug("%s: sk:%p\n", __func__, sk);
3899 /* Initialize the SCTP per socket area. */
3900 switch (sk->sk_type) {
3901 case SOCK_SEQPACKET:
3902 sp->type = SCTP_SOCKET_UDP;
3905 sp->type = SCTP_SOCKET_TCP;
3908 return -ESOCKTNOSUPPORT;
3911 /* Initialize default send parameters. These parameters can be
3912 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3914 sp->default_stream = 0;
3915 sp->default_ppid = 0;
3916 sp->default_flags = 0;
3917 sp->default_context = 0;
3918 sp->default_timetolive = 0;
3920 sp->default_rcv_context = 0;
3921 sp->max_burst = net->sctp.max_burst;
3923 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
3925 /* Initialize default setup parameters. These parameters
3926 * can be modified with the SCTP_INITMSG socket option or
3927 * overridden by the SCTP_INIT CMSG.
3929 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3930 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3931 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
3932 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
3934 /* Initialize default RTO related parameters. These parameters can
3935 * be modified for with the SCTP_RTOINFO socket option.
3937 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
3938 sp->rtoinfo.srto_max = net->sctp.rto_max;
3939 sp->rtoinfo.srto_min = net->sctp.rto_min;
3941 /* Initialize default association related parameters. These parameters
3942 * can be modified with the SCTP_ASSOCINFO socket option.
3944 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
3945 sp->assocparams.sasoc_number_peer_destinations = 0;
3946 sp->assocparams.sasoc_peer_rwnd = 0;
3947 sp->assocparams.sasoc_local_rwnd = 0;
3948 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
3950 /* Initialize default event subscriptions. By default, all the
3953 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3955 /* Default Peer Address Parameters. These defaults can
3956 * be modified via SCTP_PEER_ADDR_PARAMS
3958 sp->hbinterval = net->sctp.hb_interval;
3959 sp->pathmaxrxt = net->sctp.max_retrans_path;
3960 sp->pathmtu = 0; /* allow default discovery */
3961 sp->sackdelay = net->sctp.sack_timeout;
3963 sp->param_flags = SPP_HB_ENABLE |
3965 SPP_SACKDELAY_ENABLE;
3967 /* If enabled no SCTP message fragmentation will be performed.
3968 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3970 sp->disable_fragments = 0;
3972 /* Enable Nagle algorithm by default. */
3975 /* Enable by default. */
3978 /* Auto-close idle associations after the configured
3979 * number of seconds. A value of 0 disables this
3980 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3981 * for UDP-style sockets only.
3985 /* User specified fragmentation limit. */
3988 sp->adaptation_ind = 0;
3990 sp->pf = sctp_get_pf_specific(sk->sk_family);
3992 /* Control variables for partial data delivery. */
3993 atomic_set(&sp->pd_mode, 0);
3994 skb_queue_head_init(&sp->pd_lobby);
3995 sp->frag_interleave = 0;
3997 /* Create a per socket endpoint structure. Even if we
3998 * change the data structure relationships, this may still
3999 * be useful for storing pre-connect address information.
4001 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4007 sk->sk_destruct = sctp_destruct_sock;
4009 SCTP_DBG_OBJCNT_INC(sock);
4012 percpu_counter_inc(&sctp_sockets_allocated);
4013 sock_prot_inuse_add(net, sk->sk_prot, 1);
4014 if (net->sctp.default_auto_asconf) {
4015 list_add_tail(&sp->auto_asconf_list,
4016 &net->sctp.auto_asconf_splist);
4017 sp->do_auto_asconf = 1;
4019 sp->do_auto_asconf = 0;
4025 /* Cleanup any SCTP per socket resources. */
4026 static void sctp_destroy_sock(struct sock *sk)
4028 struct sctp_sock *sp;
4030 pr_debug("%s: sk:%p\n", __func__, sk);
4032 /* Release our hold on the endpoint. */
4034 /* This could happen during socket init, thus we bail out
4035 * early, since the rest of the below is not setup either.
4040 if (sp->do_auto_asconf) {
4041 sp->do_auto_asconf = 0;
4042 list_del(&sp->auto_asconf_list);
4044 sctp_endpoint_free(sp->ep);
4046 percpu_counter_dec(&sctp_sockets_allocated);
4047 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4051 /* Triggered when there are no references on the socket anymore */
4052 static void sctp_destruct_sock(struct sock *sk)
4054 struct sctp_sock *sp = sctp_sk(sk);
4056 /* Free up the HMAC transform. */
4057 crypto_free_hash(sp->hmac);
4059 inet_sock_destruct(sk);
4062 /* API 4.1.7 shutdown() - TCP Style Syntax
4063 * int shutdown(int socket, int how);
4065 * sd - the socket descriptor of the association to be closed.
4066 * how - Specifies the type of shutdown. The values are
4069 * Disables further receive operations. No SCTP
4070 * protocol action is taken.
4072 * Disables further send operations, and initiates
4073 * the SCTP shutdown sequence.
4075 * Disables further send and receive operations
4076 * and initiates the SCTP shutdown sequence.
4078 static void sctp_shutdown(struct sock *sk, int how)
4080 struct net *net = sock_net(sk);
4081 struct sctp_endpoint *ep;
4082 struct sctp_association *asoc;
4084 if (!sctp_style(sk, TCP))
4087 if (how & SEND_SHUTDOWN) {
4088 ep = sctp_sk(sk)->ep;
4089 if (!list_empty(&ep->asocs)) {
4090 asoc = list_entry(ep->asocs.next,
4091 struct sctp_association, asocs);
4092 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4097 /* 7.2.1 Association Status (SCTP_STATUS)
4099 * Applications can retrieve current status information about an
4100 * association, including association state, peer receiver window size,
4101 * number of unacked data chunks, and number of data chunks pending
4102 * receipt. This information is read-only.
4104 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4105 char __user *optval,
4108 struct sctp_status status;
4109 struct sctp_association *asoc = NULL;
4110 struct sctp_transport *transport;
4111 sctp_assoc_t associd;
4114 if (len < sizeof(status)) {
4119 len = sizeof(status);
4120 if (copy_from_user(&status, optval, len)) {
4125 associd = status.sstat_assoc_id;
4126 asoc = sctp_id2assoc(sk, associd);
4132 transport = asoc->peer.primary_path;
4134 status.sstat_assoc_id = sctp_assoc2id(asoc);
4135 status.sstat_state = asoc->state;
4136 status.sstat_rwnd = asoc->peer.rwnd;
4137 status.sstat_unackdata = asoc->unack_data;
4139 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4140 status.sstat_instrms = asoc->c.sinit_max_instreams;
4141 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4142 status.sstat_fragmentation_point = asoc->frag_point;
4143 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4144 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4145 transport->af_specific->sockaddr_len);
4146 /* Map ipv4 address into v4-mapped-on-v6 address. */
4147 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4148 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4149 status.sstat_primary.spinfo_state = transport->state;
4150 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4151 status.sstat_primary.spinfo_srtt = transport->srtt;
4152 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4153 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4155 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4156 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4158 if (put_user(len, optlen)) {
4163 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4164 __func__, len, status.sstat_state, status.sstat_rwnd,
4165 status.sstat_assoc_id);
4167 if (copy_to_user(optval, &status, len)) {
4177 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4179 * Applications can retrieve information about a specific peer address
4180 * of an association, including its reachability state, congestion
4181 * window, and retransmission timer values. This information is
4184 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4185 char __user *optval,
4188 struct sctp_paddrinfo pinfo;
4189 struct sctp_transport *transport;
4192 if (len < sizeof(pinfo)) {
4197 len = sizeof(pinfo);
4198 if (copy_from_user(&pinfo, optval, len)) {
4203 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4204 pinfo.spinfo_assoc_id);
4208 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4209 pinfo.spinfo_state = transport->state;
4210 pinfo.spinfo_cwnd = transport->cwnd;
4211 pinfo.spinfo_srtt = transport->srtt;
4212 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4213 pinfo.spinfo_mtu = transport->pathmtu;
4215 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4216 pinfo.spinfo_state = SCTP_ACTIVE;
4218 if (put_user(len, optlen)) {
4223 if (copy_to_user(optval, &pinfo, len)) {
4232 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4234 * This option is a on/off flag. If enabled no SCTP message
4235 * fragmentation will be performed. Instead if a message being sent
4236 * exceeds the current PMTU size, the message will NOT be sent and
4237 * instead a error will be indicated to the user.
4239 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4240 char __user *optval, int __user *optlen)
4244 if (len < sizeof(int))
4248 val = (sctp_sk(sk)->disable_fragments == 1);
4249 if (put_user(len, optlen))
4251 if (copy_to_user(optval, &val, len))
4256 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4258 * This socket option is used to specify various notifications and
4259 * ancillary data the user wishes to receive.
4261 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4266 if (len > sizeof(struct sctp_event_subscribe))
4267 len = sizeof(struct sctp_event_subscribe);
4268 if (put_user(len, optlen))
4270 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4275 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4277 * This socket option is applicable to the UDP-style socket only. When
4278 * set it will cause associations that are idle for more than the
4279 * specified number of seconds to automatically close. An association
4280 * being idle is defined an association that has NOT sent or received
4281 * user data. The special value of '0' indicates that no automatic
4282 * close of any associations should be performed. The option expects an
4283 * integer defining the number of seconds of idle time before an
4284 * association is closed.
4286 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4288 /* Applicable to UDP-style socket only */
4289 if (sctp_style(sk, TCP))
4291 if (len < sizeof(int))
4294 if (put_user(len, optlen))
4296 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4301 /* Helper routine to branch off an association to a new socket. */
4302 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4304 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4305 struct socket *sock;
4312 /* An association cannot be branched off from an already peeled-off
4313 * socket, nor is this supported for tcp style sockets.
4315 if (!sctp_style(sk, UDP))
4318 /* Create a new socket. */
4319 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4323 sctp_copy_sock(sock->sk, sk, asoc);
4325 /* Make peeled-off sockets more like 1-1 accepted sockets.
4326 * Set the daddr and initialize id to something more random
4328 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4329 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4331 /* Populate the fields of the newsk from the oldsk and migrate the
4332 * asoc to the newsk.
4334 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4340 EXPORT_SYMBOL(sctp_do_peeloff);
4342 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4344 sctp_peeloff_arg_t peeloff;
4345 struct socket *newsock;
4346 struct file *newfile;
4349 if (len < sizeof(sctp_peeloff_arg_t))
4351 len = sizeof(sctp_peeloff_arg_t);
4352 if (copy_from_user(&peeloff, optval, len))
4355 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4359 /* Map the socket to an unused fd that can be returned to the user. */
4360 retval = get_unused_fd_flags(0);
4362 sock_release(newsock);
4366 newfile = sock_alloc_file(newsock, 0, NULL);
4367 if (unlikely(IS_ERR(newfile))) {
4368 put_unused_fd(retval);
4369 sock_release(newsock);
4370 return PTR_ERR(newfile);
4373 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4376 /* Return the fd mapped to the new socket. */
4377 if (put_user(len, optlen)) {
4379 put_unused_fd(retval);
4382 peeloff.sd = retval;
4383 if (copy_to_user(optval, &peeloff, len)) {
4385 put_unused_fd(retval);
4388 fd_install(retval, newfile);
4393 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4395 * Applications can enable or disable heartbeats for any peer address of
4396 * an association, modify an address's heartbeat interval, force a
4397 * heartbeat to be sent immediately, and adjust the address's maximum
4398 * number of retransmissions sent before an address is considered
4399 * unreachable. The following structure is used to access and modify an
4400 * address's parameters:
4402 * struct sctp_paddrparams {
4403 * sctp_assoc_t spp_assoc_id;
4404 * struct sockaddr_storage spp_address;
4405 * uint32_t spp_hbinterval;
4406 * uint16_t spp_pathmaxrxt;
4407 * uint32_t spp_pathmtu;
4408 * uint32_t spp_sackdelay;
4409 * uint32_t spp_flags;
4412 * spp_assoc_id - (one-to-many style socket) This is filled in the
4413 * application, and identifies the association for
4415 * spp_address - This specifies which address is of interest.
4416 * spp_hbinterval - This contains the value of the heartbeat interval,
4417 * in milliseconds. If a value of zero
4418 * is present in this field then no changes are to
4419 * be made to this parameter.
4420 * spp_pathmaxrxt - This contains the maximum number of
4421 * retransmissions before this address shall be
4422 * considered unreachable. If a value of zero
4423 * is present in this field then no changes are to
4424 * be made to this parameter.
4425 * spp_pathmtu - When Path MTU discovery is disabled the value
4426 * specified here will be the "fixed" path mtu.
4427 * Note that if the spp_address field is empty
4428 * then all associations on this address will
4429 * have this fixed path mtu set upon them.
4431 * spp_sackdelay - When delayed sack is enabled, this value specifies
4432 * the number of milliseconds that sacks will be delayed
4433 * for. This value will apply to all addresses of an
4434 * association if the spp_address field is empty. Note
4435 * also, that if delayed sack is enabled and this
4436 * value is set to 0, no change is made to the last
4437 * recorded delayed sack timer value.
4439 * spp_flags - These flags are used to control various features
4440 * on an association. The flag field may contain
4441 * zero or more of the following options.
4443 * SPP_HB_ENABLE - Enable heartbeats on the
4444 * specified address. Note that if the address
4445 * field is empty all addresses for the association
4446 * have heartbeats enabled upon them.
4448 * SPP_HB_DISABLE - Disable heartbeats on the
4449 * speicifed address. Note that if the address
4450 * field is empty all addresses for the association
4451 * will have their heartbeats disabled. Note also
4452 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4453 * mutually exclusive, only one of these two should
4454 * be specified. Enabling both fields will have
4455 * undetermined results.
4457 * SPP_HB_DEMAND - Request a user initiated heartbeat
4458 * to be made immediately.
4460 * SPP_PMTUD_ENABLE - This field will enable PMTU
4461 * discovery upon the specified address. Note that
4462 * if the address feild is empty then all addresses
4463 * on the association are effected.
4465 * SPP_PMTUD_DISABLE - This field will disable PMTU
4466 * discovery upon the specified address. Note that
4467 * if the address feild is empty then all addresses
4468 * on the association are effected. Not also that
4469 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4470 * exclusive. Enabling both will have undetermined
4473 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4474 * on delayed sack. The time specified in spp_sackdelay
4475 * is used to specify the sack delay for this address. Note
4476 * that if spp_address is empty then all addresses will
4477 * enable delayed sack and take on the sack delay
4478 * value specified in spp_sackdelay.
4479 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4480 * off delayed sack. If the spp_address field is blank then
4481 * delayed sack is disabled for the entire association. Note
4482 * also that this field is mutually exclusive to
4483 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4486 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4487 char __user *optval, int __user *optlen)
4489 struct sctp_paddrparams params;
4490 struct sctp_transport *trans = NULL;
4491 struct sctp_association *asoc = NULL;
4492 struct sctp_sock *sp = sctp_sk(sk);
4494 if (len < sizeof(struct sctp_paddrparams))
4496 len = sizeof(struct sctp_paddrparams);
4497 if (copy_from_user(¶ms, optval, len))
4500 /* If an address other than INADDR_ANY is specified, and
4501 * no transport is found, then the request is invalid.
4503 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
4504 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4505 params.spp_assoc_id);
4507 pr_debug("%s: failed no transport\n", __func__);
4512 /* Get association, if assoc_id != 0 and the socket is a one
4513 * to many style socket, and an association was not found, then
4514 * the id was invalid.
4516 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4517 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4518 pr_debug("%s: failed no association\n", __func__);
4523 /* Fetch transport values. */
4524 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4525 params.spp_pathmtu = trans->pathmtu;
4526 params.spp_pathmaxrxt = trans->pathmaxrxt;
4527 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4529 /*draft-11 doesn't say what to return in spp_flags*/
4530 params.spp_flags = trans->param_flags;
4532 /* Fetch association values. */
4533 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4534 params.spp_pathmtu = asoc->pathmtu;
4535 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4536 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4538 /*draft-11 doesn't say what to return in spp_flags*/
4539 params.spp_flags = asoc->param_flags;
4541 /* Fetch socket values. */
4542 params.spp_hbinterval = sp->hbinterval;
4543 params.spp_pathmtu = sp->pathmtu;
4544 params.spp_sackdelay = sp->sackdelay;
4545 params.spp_pathmaxrxt = sp->pathmaxrxt;
4547 /*draft-11 doesn't say what to return in spp_flags*/
4548 params.spp_flags = sp->param_flags;
4551 if (copy_to_user(optval, ¶ms, len))
4554 if (put_user(len, optlen))
4561 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4563 * This option will effect the way delayed acks are performed. This
4564 * option allows you to get or set the delayed ack time, in
4565 * milliseconds. It also allows changing the delayed ack frequency.
4566 * Changing the frequency to 1 disables the delayed sack algorithm. If
4567 * the assoc_id is 0, then this sets or gets the endpoints default
4568 * values. If the assoc_id field is non-zero, then the set or get
4569 * effects the specified association for the one to many model (the
4570 * assoc_id field is ignored by the one to one model). Note that if
4571 * sack_delay or sack_freq are 0 when setting this option, then the
4572 * current values will remain unchanged.
4574 * struct sctp_sack_info {
4575 * sctp_assoc_t sack_assoc_id;
4576 * uint32_t sack_delay;
4577 * uint32_t sack_freq;
4580 * sack_assoc_id - This parameter, indicates which association the user
4581 * is performing an action upon. Note that if this field's value is
4582 * zero then the endpoints default value is changed (effecting future
4583 * associations only).
4585 * sack_delay - This parameter contains the number of milliseconds that
4586 * the user is requesting the delayed ACK timer be set to. Note that
4587 * this value is defined in the standard to be between 200 and 500
4590 * sack_freq - This parameter contains the number of packets that must
4591 * be received before a sack is sent without waiting for the delay
4592 * timer to expire. The default value for this is 2, setting this
4593 * value to 1 will disable the delayed sack algorithm.
4595 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4596 char __user *optval,
4599 struct sctp_sack_info params;
4600 struct sctp_association *asoc = NULL;
4601 struct sctp_sock *sp = sctp_sk(sk);
4603 if (len >= sizeof(struct sctp_sack_info)) {
4604 len = sizeof(struct sctp_sack_info);
4606 if (copy_from_user(¶ms, optval, len))
4608 } else if (len == sizeof(struct sctp_assoc_value)) {
4609 pr_warn_ratelimited(DEPRECATED
4611 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
4612 "Use struct sctp_sack_info instead\n",
4613 current->comm, task_pid_nr(current));
4614 if (copy_from_user(¶ms, optval, len))
4619 /* Get association, if sack_assoc_id != 0 and the socket is a one
4620 * to many style socket, and an association was not found, then
4621 * the id was invalid.
4623 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4624 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4628 /* Fetch association values. */
4629 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4630 params.sack_delay = jiffies_to_msecs(
4632 params.sack_freq = asoc->sackfreq;
4635 params.sack_delay = 0;
4636 params.sack_freq = 1;
4639 /* Fetch socket values. */
4640 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4641 params.sack_delay = sp->sackdelay;
4642 params.sack_freq = sp->sackfreq;
4644 params.sack_delay = 0;
4645 params.sack_freq = 1;
4649 if (copy_to_user(optval, ¶ms, len))
4652 if (put_user(len, optlen))
4658 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4660 * Applications can specify protocol parameters for the default association
4661 * initialization. The option name argument to setsockopt() and getsockopt()
4664 * Setting initialization parameters is effective only on an unconnected
4665 * socket (for UDP-style sockets only future associations are effected
4666 * by the change). With TCP-style sockets, this option is inherited by
4667 * sockets derived from a listener socket.
4669 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4671 if (len < sizeof(struct sctp_initmsg))
4673 len = sizeof(struct sctp_initmsg);
4674 if (put_user(len, optlen))
4676 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4682 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4683 char __user *optval, int __user *optlen)
4685 struct sctp_association *asoc;
4687 struct sctp_getaddrs getaddrs;
4688 struct sctp_transport *from;
4690 union sctp_addr temp;
4691 struct sctp_sock *sp = sctp_sk(sk);
4696 if (len < sizeof(struct sctp_getaddrs))
4699 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4702 /* For UDP-style sockets, id specifies the association to query. */
4703 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4707 to = optval + offsetof(struct sctp_getaddrs, addrs);
4708 space_left = len - offsetof(struct sctp_getaddrs, addrs);
4710 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4712 memcpy(&temp, &from->ipaddr, sizeof(temp));
4713 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4714 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4715 if (space_left < addrlen)
4717 if (copy_to_user(to, &temp, addrlen))
4721 space_left -= addrlen;
4724 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4726 bytes_copied = ((char __user *)to) - optval;
4727 if (put_user(bytes_copied, optlen))
4733 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4734 size_t space_left, int *bytes_copied)
4736 struct sctp_sockaddr_entry *addr;
4737 union sctp_addr temp;
4740 struct net *net = sock_net(sk);
4743 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4747 if ((PF_INET == sk->sk_family) &&
4748 (AF_INET6 == addr->a.sa.sa_family))
4750 if ((PF_INET6 == sk->sk_family) &&
4751 inet_v6_ipv6only(sk) &&
4752 (AF_INET == addr->a.sa.sa_family))
4754 memcpy(&temp, &addr->a, sizeof(temp));
4755 if (!temp.v4.sin_port)
4756 temp.v4.sin_port = htons(port);
4758 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4760 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4761 if (space_left < addrlen) {
4765 memcpy(to, &temp, addrlen);
4769 space_left -= addrlen;
4770 *bytes_copied += addrlen;
4778 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4779 char __user *optval, int __user *optlen)
4781 struct sctp_bind_addr *bp;
4782 struct sctp_association *asoc;
4784 struct sctp_getaddrs getaddrs;
4785 struct sctp_sockaddr_entry *addr;
4787 union sctp_addr temp;
4788 struct sctp_sock *sp = sctp_sk(sk);
4792 int bytes_copied = 0;
4796 if (len < sizeof(struct sctp_getaddrs))
4799 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4803 * For UDP-style sockets, id specifies the association to query.
4804 * If the id field is set to the value '0' then the locally bound
4805 * addresses are returned without regard to any particular
4808 if (0 == getaddrs.assoc_id) {
4809 bp = &sctp_sk(sk)->ep->base.bind_addr;
4811 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4814 bp = &asoc->base.bind_addr;
4817 to = optval + offsetof(struct sctp_getaddrs, addrs);
4818 space_left = len - offsetof(struct sctp_getaddrs, addrs);
4820 addrs = kmalloc(space_left, GFP_KERNEL);
4824 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4825 * addresses from the global local address list.
4827 if (sctp_list_single_entry(&bp->address_list)) {
4828 addr = list_entry(bp->address_list.next,
4829 struct sctp_sockaddr_entry, list);
4830 if (sctp_is_any(sk, &addr->a)) {
4831 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4832 space_left, &bytes_copied);
4842 /* Protection on the bound address list is not needed since
4843 * in the socket option context we hold a socket lock and
4844 * thus the bound address list can't change.
4846 list_for_each_entry(addr, &bp->address_list, list) {
4847 memcpy(&temp, &addr->a, sizeof(temp));
4848 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4849 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4850 if (space_left < addrlen) {
4851 err = -ENOMEM; /*fixme: right error?*/
4854 memcpy(buf, &temp, addrlen);
4856 bytes_copied += addrlen;
4858 space_left -= addrlen;
4862 if (copy_to_user(to, addrs, bytes_copied)) {
4866 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4870 if (put_user(bytes_copied, optlen))
4877 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4879 * Requests that the local SCTP stack use the enclosed peer address as
4880 * the association primary. The enclosed address must be one of the
4881 * association peer's addresses.
4883 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4884 char __user *optval, int __user *optlen)
4886 struct sctp_prim prim;
4887 struct sctp_association *asoc;
4888 struct sctp_sock *sp = sctp_sk(sk);
4890 if (len < sizeof(struct sctp_prim))
4893 len = sizeof(struct sctp_prim);
4895 if (copy_from_user(&prim, optval, len))
4898 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4902 if (!asoc->peer.primary_path)
4905 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4906 asoc->peer.primary_path->af_specific->sockaddr_len);
4908 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4909 (union sctp_addr *)&prim.ssp_addr);
4911 if (put_user(len, optlen))
4913 if (copy_to_user(optval, &prim, len))
4920 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4922 * Requests that the local endpoint set the specified Adaptation Layer
4923 * Indication parameter for all future INIT and INIT-ACK exchanges.
4925 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4926 char __user *optval, int __user *optlen)
4928 struct sctp_setadaptation adaptation;
4930 if (len < sizeof(struct sctp_setadaptation))
4933 len = sizeof(struct sctp_setadaptation);
4935 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4937 if (put_user(len, optlen))
4939 if (copy_to_user(optval, &adaptation, len))
4947 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4949 * Applications that wish to use the sendto() system call may wish to
4950 * specify a default set of parameters that would normally be supplied
4951 * through the inclusion of ancillary data. This socket option allows
4952 * such an application to set the default sctp_sndrcvinfo structure.
4955 * The application that wishes to use this socket option simply passes
4956 * in to this call the sctp_sndrcvinfo structure defined in Section
4957 * 5.2.2) The input parameters accepted by this call include
4958 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4959 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4960 * to this call if the caller is using the UDP model.
4962 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4964 static int sctp_getsockopt_default_send_param(struct sock *sk,
4965 int len, char __user *optval,
4968 struct sctp_sndrcvinfo info;
4969 struct sctp_association *asoc;
4970 struct sctp_sock *sp = sctp_sk(sk);
4972 if (len < sizeof(struct sctp_sndrcvinfo))
4975 len = sizeof(struct sctp_sndrcvinfo);
4977 if (copy_from_user(&info, optval, len))
4980 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4981 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4985 info.sinfo_stream = asoc->default_stream;
4986 info.sinfo_flags = asoc->default_flags;
4987 info.sinfo_ppid = asoc->default_ppid;
4988 info.sinfo_context = asoc->default_context;
4989 info.sinfo_timetolive = asoc->default_timetolive;
4991 info.sinfo_stream = sp->default_stream;
4992 info.sinfo_flags = sp->default_flags;
4993 info.sinfo_ppid = sp->default_ppid;
4994 info.sinfo_context = sp->default_context;
4995 info.sinfo_timetolive = sp->default_timetolive;
4998 if (put_user(len, optlen))
5000 if (copy_to_user(optval, &info, len))
5008 * 7.1.5 SCTP_NODELAY
5010 * Turn on/off any Nagle-like algorithm. This means that packets are
5011 * generally sent as soon as possible and no unnecessary delays are
5012 * introduced, at the cost of more packets in the network. Expects an
5013 * integer boolean flag.
5016 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5017 char __user *optval, int __user *optlen)
5021 if (len < sizeof(int))
5025 val = (sctp_sk(sk)->nodelay == 1);
5026 if (put_user(len, optlen))
5028 if (copy_to_user(optval, &val, len))
5035 * 7.1.1 SCTP_RTOINFO
5037 * The protocol parameters used to initialize and bound retransmission
5038 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5039 * and modify these parameters.
5040 * All parameters are time values, in milliseconds. A value of 0, when
5041 * modifying the parameters, indicates that the current value should not
5045 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5046 char __user *optval,
5047 int __user *optlen) {
5048 struct sctp_rtoinfo rtoinfo;
5049 struct sctp_association *asoc;
5051 if (len < sizeof (struct sctp_rtoinfo))
5054 len = sizeof(struct sctp_rtoinfo);
5056 if (copy_from_user(&rtoinfo, optval, len))
5059 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5061 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5064 /* Values corresponding to the specific association. */
5066 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5067 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5068 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5070 /* Values corresponding to the endpoint. */
5071 struct sctp_sock *sp = sctp_sk(sk);
5073 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5074 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5075 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5078 if (put_user(len, optlen))
5081 if (copy_to_user(optval, &rtoinfo, len))
5089 * 7.1.2 SCTP_ASSOCINFO
5091 * This option is used to tune the maximum retransmission attempts
5092 * of the association.
5093 * Returns an error if the new association retransmission value is
5094 * greater than the sum of the retransmission value of the peer.
5095 * See [SCTP] for more information.
5098 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5099 char __user *optval,
5103 struct sctp_assocparams assocparams;
5104 struct sctp_association *asoc;
5105 struct list_head *pos;
5108 if (len < sizeof (struct sctp_assocparams))
5111 len = sizeof(struct sctp_assocparams);
5113 if (copy_from_user(&assocparams, optval, len))
5116 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5118 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5121 /* Values correspoinding to the specific association */
5123 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5124 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5125 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5126 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5128 list_for_each(pos, &asoc->peer.transport_addr_list) {
5132 assocparams.sasoc_number_peer_destinations = cnt;
5134 /* Values corresponding to the endpoint */
5135 struct sctp_sock *sp = sctp_sk(sk);
5137 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5138 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5139 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5140 assocparams.sasoc_cookie_life =
5141 sp->assocparams.sasoc_cookie_life;
5142 assocparams.sasoc_number_peer_destinations =
5144 sasoc_number_peer_destinations;
5147 if (put_user(len, optlen))
5150 if (copy_to_user(optval, &assocparams, len))
5157 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5159 * This socket option is a boolean flag which turns on or off mapped V4
5160 * addresses. If this option is turned on and the socket is type
5161 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5162 * If this option is turned off, then no mapping will be done of V4
5163 * addresses and a user will receive both PF_INET6 and PF_INET type
5164 * addresses on the socket.
5166 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5167 char __user *optval, int __user *optlen)
5170 struct sctp_sock *sp = sctp_sk(sk);
5172 if (len < sizeof(int))
5177 if (put_user(len, optlen))
5179 if (copy_to_user(optval, &val, len))
5186 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5187 * (chapter and verse is quoted at sctp_setsockopt_context())
5189 static int sctp_getsockopt_context(struct sock *sk, int len,
5190 char __user *optval, int __user *optlen)
5192 struct sctp_assoc_value params;
5193 struct sctp_sock *sp;
5194 struct sctp_association *asoc;
5196 if (len < sizeof(struct sctp_assoc_value))
5199 len = sizeof(struct sctp_assoc_value);
5201 if (copy_from_user(¶ms, optval, len))
5206 if (params.assoc_id != 0) {
5207 asoc = sctp_id2assoc(sk, params.assoc_id);
5210 params.assoc_value = asoc->default_rcv_context;
5212 params.assoc_value = sp->default_rcv_context;
5215 if (put_user(len, optlen))
5217 if (copy_to_user(optval, ¶ms, len))
5224 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5225 * This option will get or set the maximum size to put in any outgoing
5226 * SCTP DATA chunk. If a message is larger than this size it will be
5227 * fragmented by SCTP into the specified size. Note that the underlying
5228 * SCTP implementation may fragment into smaller sized chunks when the
5229 * PMTU of the underlying association is smaller than the value set by
5230 * the user. The default value for this option is '0' which indicates
5231 * the user is NOT limiting fragmentation and only the PMTU will effect
5232 * SCTP's choice of DATA chunk size. Note also that values set larger
5233 * than the maximum size of an IP datagram will effectively let SCTP
5234 * control fragmentation (i.e. the same as setting this option to 0).
5236 * The following structure is used to access and modify this parameter:
5238 * struct sctp_assoc_value {
5239 * sctp_assoc_t assoc_id;
5240 * uint32_t assoc_value;
5243 * assoc_id: This parameter is ignored for one-to-one style sockets.
5244 * For one-to-many style sockets this parameter indicates which
5245 * association the user is performing an action upon. Note that if
5246 * this field's value is zero then the endpoints default value is
5247 * changed (effecting future associations only).
5248 * assoc_value: This parameter specifies the maximum size in bytes.
5250 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5251 char __user *optval, int __user *optlen)
5253 struct sctp_assoc_value params;
5254 struct sctp_association *asoc;
5256 if (len == sizeof(int)) {
5257 pr_warn_ratelimited(DEPRECATED
5259 "Use of int in maxseg socket option.\n"
5260 "Use struct sctp_assoc_value instead\n",
5261 current->comm, task_pid_nr(current));
5262 params.assoc_id = 0;
5263 } else if (len >= sizeof(struct sctp_assoc_value)) {
5264 len = sizeof(struct sctp_assoc_value);
5265 if (copy_from_user(¶ms, optval, sizeof(params)))
5270 asoc = sctp_id2assoc(sk, params.assoc_id);
5271 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5275 params.assoc_value = asoc->frag_point;
5277 params.assoc_value = sctp_sk(sk)->user_frag;
5279 if (put_user(len, optlen))
5281 if (len == sizeof(int)) {
5282 if (copy_to_user(optval, ¶ms.assoc_value, len))
5285 if (copy_to_user(optval, ¶ms, len))
5293 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5294 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5296 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5297 char __user *optval, int __user *optlen)
5301 if (len < sizeof(int))
5306 val = sctp_sk(sk)->frag_interleave;
5307 if (put_user(len, optlen))
5309 if (copy_to_user(optval, &val, len))
5316 * 7.1.25. Set or Get the sctp partial delivery point
5317 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5319 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5320 char __user *optval,
5325 if (len < sizeof(u32))
5330 val = sctp_sk(sk)->pd_point;
5331 if (put_user(len, optlen))
5333 if (copy_to_user(optval, &val, len))
5340 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5341 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5343 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5344 char __user *optval,
5347 struct sctp_assoc_value params;
5348 struct sctp_sock *sp;
5349 struct sctp_association *asoc;
5351 if (len == sizeof(int)) {
5352 pr_warn_ratelimited(DEPRECATED
5354 "Use of int in max_burst socket option.\n"
5355 "Use struct sctp_assoc_value instead\n",
5356 current->comm, task_pid_nr(current));
5357 params.assoc_id = 0;
5358 } else if (len >= sizeof(struct sctp_assoc_value)) {
5359 len = sizeof(struct sctp_assoc_value);
5360 if (copy_from_user(¶ms, optval, len))
5367 if (params.assoc_id != 0) {
5368 asoc = sctp_id2assoc(sk, params.assoc_id);
5371 params.assoc_value = asoc->max_burst;
5373 params.assoc_value = sp->max_burst;
5375 if (len == sizeof(int)) {
5376 if (copy_to_user(optval, ¶ms.assoc_value, len))
5379 if (copy_to_user(optval, ¶ms, len))
5387 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5388 char __user *optval, int __user *optlen)
5390 struct net *net = sock_net(sk);
5391 struct sctp_hmacalgo __user *p = (void __user *)optval;
5392 struct sctp_hmac_algo_param *hmacs;
5396 if (!net->sctp.auth_enable)
5399 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5400 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5402 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5405 len = sizeof(struct sctp_hmacalgo) + data_len;
5406 num_idents = data_len / sizeof(u16);
5408 if (put_user(len, optlen))
5410 if (put_user(num_idents, &p->shmac_num_idents))
5412 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5417 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5418 char __user *optval, int __user *optlen)
5420 struct net *net = sock_net(sk);
5421 struct sctp_authkeyid val;
5422 struct sctp_association *asoc;
5424 if (!net->sctp.auth_enable)
5427 if (len < sizeof(struct sctp_authkeyid))
5429 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5432 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5433 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5437 val.scact_keynumber = asoc->active_key_id;
5439 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5441 len = sizeof(struct sctp_authkeyid);
5442 if (put_user(len, optlen))
5444 if (copy_to_user(optval, &val, len))
5450 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5451 char __user *optval, int __user *optlen)
5453 struct net *net = sock_net(sk);
5454 struct sctp_authchunks __user *p = (void __user *)optval;
5455 struct sctp_authchunks val;
5456 struct sctp_association *asoc;
5457 struct sctp_chunks_param *ch;
5461 if (!net->sctp.auth_enable)
5464 if (len < sizeof(struct sctp_authchunks))
5467 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5470 to = p->gauth_chunks;
5471 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5475 ch = asoc->peer.peer_chunks;
5479 /* See if the user provided enough room for all the data */
5480 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5481 if (len < num_chunks)
5484 if (copy_to_user(to, ch->chunks, num_chunks))
5487 len = sizeof(struct sctp_authchunks) + num_chunks;
5488 if (put_user(len, optlen))
5490 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5495 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5496 char __user *optval, int __user *optlen)
5498 struct net *net = sock_net(sk);
5499 struct sctp_authchunks __user *p = (void __user *)optval;
5500 struct sctp_authchunks val;
5501 struct sctp_association *asoc;
5502 struct sctp_chunks_param *ch;
5506 if (!net->sctp.auth_enable)
5509 if (len < sizeof(struct sctp_authchunks))
5512 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5515 to = p->gauth_chunks;
5516 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5517 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5521 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
5523 ch = sctp_sk(sk)->ep->auth_chunk_list;
5528 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5529 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5532 if (copy_to_user(to, ch->chunks, num_chunks))
5535 len = sizeof(struct sctp_authchunks) + num_chunks;
5536 if (put_user(len, optlen))
5538 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5545 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5546 * This option gets the current number of associations that are attached
5547 * to a one-to-many style socket. The option value is an uint32_t.
5549 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5550 char __user *optval, int __user *optlen)
5552 struct sctp_sock *sp = sctp_sk(sk);
5553 struct sctp_association *asoc;
5556 if (sctp_style(sk, TCP))
5559 if (len < sizeof(u32))
5564 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5568 if (put_user(len, optlen))
5570 if (copy_to_user(optval, &val, len))
5577 * 8.1.23 SCTP_AUTO_ASCONF
5578 * See the corresponding setsockopt entry as description
5580 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5581 char __user *optval, int __user *optlen)
5585 if (len < sizeof(int))
5589 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5591 if (put_user(len, optlen))
5593 if (copy_to_user(optval, &val, len))
5599 * 8.2.6. Get the Current Identifiers of Associations
5600 * (SCTP_GET_ASSOC_ID_LIST)
5602 * This option gets the current list of SCTP association identifiers of
5603 * the SCTP associations handled by a one-to-many style socket.
5605 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5606 char __user *optval, int __user *optlen)
5608 struct sctp_sock *sp = sctp_sk(sk);
5609 struct sctp_association *asoc;
5610 struct sctp_assoc_ids *ids;
5613 if (sctp_style(sk, TCP))
5616 if (len < sizeof(struct sctp_assoc_ids))
5619 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5623 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5626 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5628 ids = kmalloc(len, GFP_KERNEL);
5632 ids->gaids_number_of_ids = num;
5634 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5635 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5638 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5648 * SCTP_PEER_ADDR_THLDS
5650 * This option allows us to fetch the partially failed threshold for one or all
5651 * transports in an association. See Section 6.1 of:
5652 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5654 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5655 char __user *optval,
5659 struct sctp_paddrthlds val;
5660 struct sctp_transport *trans;
5661 struct sctp_association *asoc;
5663 if (len < sizeof(struct sctp_paddrthlds))
5665 len = sizeof(struct sctp_paddrthlds);
5666 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5669 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5670 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5674 val.spt_pathpfthld = asoc->pf_retrans;
5675 val.spt_pathmaxrxt = asoc->pathmaxrxt;
5677 trans = sctp_addr_id2transport(sk, &val.spt_address,
5682 val.spt_pathmaxrxt = trans->pathmaxrxt;
5683 val.spt_pathpfthld = trans->pf_retrans;
5686 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5693 * SCTP_GET_ASSOC_STATS
5695 * This option retrieves local per endpoint statistics. It is modeled
5696 * after OpenSolaris' implementation
5698 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
5699 char __user *optval,
5702 struct sctp_assoc_stats sas;
5703 struct sctp_association *asoc = NULL;
5705 /* User must provide at least the assoc id */
5706 if (len < sizeof(sctp_assoc_t))
5709 /* Allow the struct to grow and fill in as much as possible */
5710 len = min_t(size_t, len, sizeof(sas));
5712 if (copy_from_user(&sas, optval, len))
5715 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
5719 sas.sas_rtxchunks = asoc->stats.rtxchunks;
5720 sas.sas_gapcnt = asoc->stats.gapcnt;
5721 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
5722 sas.sas_osacks = asoc->stats.osacks;
5723 sas.sas_isacks = asoc->stats.isacks;
5724 sas.sas_octrlchunks = asoc->stats.octrlchunks;
5725 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
5726 sas.sas_oodchunks = asoc->stats.oodchunks;
5727 sas.sas_iodchunks = asoc->stats.iodchunks;
5728 sas.sas_ouodchunks = asoc->stats.ouodchunks;
5729 sas.sas_iuodchunks = asoc->stats.iuodchunks;
5730 sas.sas_idupchunks = asoc->stats.idupchunks;
5731 sas.sas_opackets = asoc->stats.opackets;
5732 sas.sas_ipackets = asoc->stats.ipackets;
5734 /* New high max rto observed, will return 0 if not a single
5735 * RTO update took place. obs_rto_ipaddr will be bogus
5738 sas.sas_maxrto = asoc->stats.max_obs_rto;
5739 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
5740 sizeof(struct sockaddr_storage));
5742 /* Mark beginning of a new observation period */
5743 asoc->stats.max_obs_rto = asoc->rto_min;
5745 if (put_user(len, optlen))
5748 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
5750 if (copy_to_user(optval, &sas, len))
5756 static int sctp_getsockopt(struct sock *sk, int level, int optname,
5757 char __user *optval, int __user *optlen)
5762 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
5764 /* I can hardly begin to describe how wrong this is. This is
5765 * so broken as to be worse than useless. The API draft
5766 * REALLY is NOT helpful here... I am not convinced that the
5767 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5768 * are at all well-founded.
5770 if (level != SOL_SCTP) {
5771 struct sctp_af *af = sctp_sk(sk)->pf->af;
5773 retval = af->getsockopt(sk, level, optname, optval, optlen);
5777 if (get_user(len, optlen))
5784 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5786 case SCTP_DISABLE_FRAGMENTS:
5787 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5791 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5793 case SCTP_AUTOCLOSE:
5794 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5796 case SCTP_SOCKOPT_PEELOFF:
5797 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5799 case SCTP_PEER_ADDR_PARAMS:
5800 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5803 case SCTP_DELAYED_SACK:
5804 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5808 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5810 case SCTP_GET_PEER_ADDRS:
5811 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5814 case SCTP_GET_LOCAL_ADDRS:
5815 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5818 case SCTP_SOCKOPT_CONNECTX3:
5819 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5821 case SCTP_DEFAULT_SEND_PARAM:
5822 retval = sctp_getsockopt_default_send_param(sk, len,
5825 case SCTP_PRIMARY_ADDR:
5826 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5829 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5832 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5834 case SCTP_ASSOCINFO:
5835 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5837 case SCTP_I_WANT_MAPPED_V4_ADDR:
5838 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5841 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5843 case SCTP_GET_PEER_ADDR_INFO:
5844 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5847 case SCTP_ADAPTATION_LAYER:
5848 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5852 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5854 case SCTP_FRAGMENT_INTERLEAVE:
5855 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5858 case SCTP_PARTIAL_DELIVERY_POINT:
5859 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5862 case SCTP_MAX_BURST:
5863 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5866 case SCTP_AUTH_CHUNK:
5867 case SCTP_AUTH_DELETE_KEY:
5868 retval = -EOPNOTSUPP;
5870 case SCTP_HMAC_IDENT:
5871 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5873 case SCTP_AUTH_ACTIVE_KEY:
5874 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5876 case SCTP_PEER_AUTH_CHUNKS:
5877 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5880 case SCTP_LOCAL_AUTH_CHUNKS:
5881 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5884 case SCTP_GET_ASSOC_NUMBER:
5885 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5887 case SCTP_GET_ASSOC_ID_LIST:
5888 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5890 case SCTP_AUTO_ASCONF:
5891 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5893 case SCTP_PEER_ADDR_THLDS:
5894 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
5896 case SCTP_GET_ASSOC_STATS:
5897 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
5900 retval = -ENOPROTOOPT;
5908 static void sctp_hash(struct sock *sk)
5913 static void sctp_unhash(struct sock *sk)
5918 /* Check if port is acceptable. Possibly find first available port.
5920 * The port hash table (contained in the 'global' SCTP protocol storage
5921 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5922 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5923 * list (the list number is the port number hashed out, so as you
5924 * would expect from a hash function, all the ports in a given list have
5925 * such a number that hashes out to the same list number; you were
5926 * expecting that, right?); so each list has a set of ports, with a
5927 * link to the socket (struct sock) that uses it, the port number and
5928 * a fastreuse flag (FIXME: NPI ipg).
5930 static struct sctp_bind_bucket *sctp_bucket_create(
5931 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
5933 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5935 struct sctp_bind_hashbucket *head; /* hash list */
5936 struct sctp_bind_bucket *pp;
5937 unsigned short snum;
5940 snum = ntohs(addr->v4.sin_port);
5942 pr_debug("%s: begins, snum:%d\n", __func__, snum);
5947 /* Search for an available port. */
5948 int low, high, remaining, index;
5951 inet_get_local_port_range(sock_net(sk), &low, &high);
5952 remaining = (high - low) + 1;
5953 rover = prandom_u32() % remaining + low;
5957 if ((rover < low) || (rover > high))
5959 if (inet_is_reserved_local_port(rover))
5961 index = sctp_phashfn(sock_net(sk), rover);
5962 head = &sctp_port_hashtable[index];
5963 spin_lock(&head->lock);
5964 sctp_for_each_hentry(pp, &head->chain)
5965 if ((pp->port == rover) &&
5966 net_eq(sock_net(sk), pp->net))
5970 spin_unlock(&head->lock);
5971 } while (--remaining > 0);
5973 /* Exhausted local port range during search? */
5978 /* OK, here is the one we will use. HEAD (the port
5979 * hash table list entry) is non-NULL and we hold it's
5984 /* We are given an specific port number; we verify
5985 * that it is not being used. If it is used, we will
5986 * exahust the search in the hash list corresponding
5987 * to the port number (snum) - we detect that with the
5988 * port iterator, pp being NULL.
5990 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
5991 spin_lock(&head->lock);
5992 sctp_for_each_hentry(pp, &head->chain) {
5993 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
6000 if (!hlist_empty(&pp->owner)) {
6001 /* We had a port hash table hit - there is an
6002 * available port (pp != NULL) and it is being
6003 * used by other socket (pp->owner not empty); that other
6004 * socket is going to be sk2.
6006 int reuse = sk->sk_reuse;
6009 pr_debug("%s: found a possible match\n", __func__);
6011 if (pp->fastreuse && sk->sk_reuse &&
6012 sk->sk_state != SCTP_SS_LISTENING)
6015 /* Run through the list of sockets bound to the port
6016 * (pp->port) [via the pointers bind_next and
6017 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6018 * we get the endpoint they describe and run through
6019 * the endpoint's list of IP (v4 or v6) addresses,
6020 * comparing each of the addresses with the address of
6021 * the socket sk. If we find a match, then that means
6022 * that this port/socket (sk) combination are already
6025 sk_for_each_bound(sk2, &pp->owner) {
6026 struct sctp_endpoint *ep2;
6027 ep2 = sctp_sk(sk2)->ep;
6030 (reuse && sk2->sk_reuse &&
6031 sk2->sk_state != SCTP_SS_LISTENING))
6034 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
6035 sctp_sk(sk2), sctp_sk(sk))) {
6041 pr_debug("%s: found a match\n", __func__);
6044 /* If there was a hash table miss, create a new port. */
6046 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6049 /* In either case (hit or miss), make sure fastreuse is 1 only
6050 * if sk->sk_reuse is too (that is, if the caller requested
6051 * SO_REUSEADDR on this socket -sk-).
6053 if (hlist_empty(&pp->owner)) {
6054 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6058 } else if (pp->fastreuse &&
6059 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6062 /* We are set, so fill up all the data in the hash table
6063 * entry, tie the socket list information with the rest of the
6064 * sockets FIXME: Blurry, NPI (ipg).
6067 if (!sctp_sk(sk)->bind_hash) {
6068 inet_sk(sk)->inet_num = snum;
6069 sk_add_bind_node(sk, &pp->owner);
6070 sctp_sk(sk)->bind_hash = pp;
6075 spin_unlock(&head->lock);
6082 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6083 * port is requested.
6085 static int sctp_get_port(struct sock *sk, unsigned short snum)
6087 union sctp_addr addr;
6088 struct sctp_af *af = sctp_sk(sk)->pf->af;
6090 /* Set up a dummy address struct from the sk. */
6091 af->from_sk(&addr, sk);
6092 addr.v4.sin_port = htons(snum);
6094 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6095 return !!sctp_get_port_local(sk, &addr);
6099 * Move a socket to LISTENING state.
6101 static int sctp_listen_start(struct sock *sk, int backlog)
6103 struct sctp_sock *sp = sctp_sk(sk);
6104 struct sctp_endpoint *ep = sp->ep;
6105 struct crypto_hash *tfm = NULL;
6108 /* Allocate HMAC for generating cookie. */
6109 if (!sp->hmac && sp->sctp_hmac_alg) {
6110 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6111 tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
6113 net_info_ratelimited("failed to load transform for %s: %ld\n",
6114 sp->sctp_hmac_alg, PTR_ERR(tfm));
6117 sctp_sk(sk)->hmac = tfm;
6121 * If a bind() or sctp_bindx() is not called prior to a listen()
6122 * call that allows new associations to be accepted, the system
6123 * picks an ephemeral port and will choose an address set equivalent
6124 * to binding with a wildcard address.
6126 * This is not currently spelled out in the SCTP sockets
6127 * extensions draft, but follows the practice as seen in TCP
6131 sk->sk_state = SCTP_SS_LISTENING;
6132 if (!ep->base.bind_addr.port) {
6133 if (sctp_autobind(sk))
6136 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6137 sk->sk_state = SCTP_SS_CLOSED;
6142 sk->sk_max_ack_backlog = backlog;
6143 sctp_hash_endpoint(ep);
6148 * 4.1.3 / 5.1.3 listen()
6150 * By default, new associations are not accepted for UDP style sockets.
6151 * An application uses listen() to mark a socket as being able to
6152 * accept new associations.
6154 * On TCP style sockets, applications use listen() to ready the SCTP
6155 * endpoint for accepting inbound associations.
6157 * On both types of endpoints a backlog of '0' disables listening.
6159 * Move a socket to LISTENING state.
6161 int sctp_inet_listen(struct socket *sock, int backlog)
6163 struct sock *sk = sock->sk;
6164 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6167 if (unlikely(backlog < 0))
6172 /* Peeled-off sockets are not allowed to listen(). */
6173 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6176 if (sock->state != SS_UNCONNECTED)
6179 /* If backlog is zero, disable listening. */
6181 if (sctp_sstate(sk, CLOSED))
6185 sctp_unhash_endpoint(ep);
6186 sk->sk_state = SCTP_SS_CLOSED;
6188 sctp_sk(sk)->bind_hash->fastreuse = 1;
6192 /* If we are already listening, just update the backlog */
6193 if (sctp_sstate(sk, LISTENING))
6194 sk->sk_max_ack_backlog = backlog;
6196 err = sctp_listen_start(sk, backlog);
6208 * This function is done by modeling the current datagram_poll() and the
6209 * tcp_poll(). Note that, based on these implementations, we don't
6210 * lock the socket in this function, even though it seems that,
6211 * ideally, locking or some other mechanisms can be used to ensure
6212 * the integrity of the counters (sndbuf and wmem_alloc) used
6213 * in this place. We assume that we don't need locks either until proven
6216 * Another thing to note is that we include the Async I/O support
6217 * here, again, by modeling the current TCP/UDP code. We don't have
6218 * a good way to test with it yet.
6220 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6222 struct sock *sk = sock->sk;
6223 struct sctp_sock *sp = sctp_sk(sk);
6226 poll_wait(file, sk_sleep(sk), wait);
6228 /* A TCP-style listening socket becomes readable when the accept queue
6231 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6232 return (!list_empty(&sp->ep->asocs)) ?
6233 (POLLIN | POLLRDNORM) : 0;
6237 /* Is there any exceptional events? */
6238 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6240 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6241 if (sk->sk_shutdown & RCV_SHUTDOWN)
6242 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6243 if (sk->sk_shutdown == SHUTDOWN_MASK)
6246 /* Is it readable? Reconsider this code with TCP-style support. */
6247 if (!skb_queue_empty(&sk->sk_receive_queue))
6248 mask |= POLLIN | POLLRDNORM;
6250 /* The association is either gone or not ready. */
6251 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6254 /* Is it writable? */
6255 if (sctp_writeable(sk)) {
6256 mask |= POLLOUT | POLLWRNORM;
6258 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6260 * Since the socket is not locked, the buffer
6261 * might be made available after the writeable check and
6262 * before the bit is set. This could cause a lost I/O
6263 * signal. tcp_poll() has a race breaker for this race
6264 * condition. Based on their implementation, we put
6265 * in the following code to cover it as well.
6267 if (sctp_writeable(sk))
6268 mask |= POLLOUT | POLLWRNORM;
6273 /********************************************************************
6274 * 2nd Level Abstractions
6275 ********************************************************************/
6277 static struct sctp_bind_bucket *sctp_bucket_create(
6278 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6280 struct sctp_bind_bucket *pp;
6282 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6284 SCTP_DBG_OBJCNT_INC(bind_bucket);
6287 INIT_HLIST_HEAD(&pp->owner);
6289 hlist_add_head(&pp->node, &head->chain);
6294 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6295 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6297 if (pp && hlist_empty(&pp->owner)) {
6298 __hlist_del(&pp->node);
6299 kmem_cache_free(sctp_bucket_cachep, pp);
6300 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6304 /* Release this socket's reference to a local port. */
6305 static inline void __sctp_put_port(struct sock *sk)
6307 struct sctp_bind_hashbucket *head =
6308 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6309 inet_sk(sk)->inet_num)];
6310 struct sctp_bind_bucket *pp;
6312 spin_lock(&head->lock);
6313 pp = sctp_sk(sk)->bind_hash;
6314 __sk_del_bind_node(sk);
6315 sctp_sk(sk)->bind_hash = NULL;
6316 inet_sk(sk)->inet_num = 0;
6317 sctp_bucket_destroy(pp);
6318 spin_unlock(&head->lock);
6321 void sctp_put_port(struct sock *sk)
6324 __sctp_put_port(sk);
6329 * The system picks an ephemeral port and choose an address set equivalent
6330 * to binding with a wildcard address.
6331 * One of those addresses will be the primary address for the association.
6332 * This automatically enables the multihoming capability of SCTP.
6334 static int sctp_autobind(struct sock *sk)
6336 union sctp_addr autoaddr;
6340 /* Initialize a local sockaddr structure to INADDR_ANY. */
6341 af = sctp_sk(sk)->pf->af;
6343 port = htons(inet_sk(sk)->inet_num);
6344 af->inaddr_any(&autoaddr, port);
6346 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6349 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6352 * 4.2 The cmsghdr Structure *
6354 * When ancillary data is sent or received, any number of ancillary data
6355 * objects can be specified by the msg_control and msg_controllen members of
6356 * the msghdr structure, because each object is preceded by
6357 * a cmsghdr structure defining the object's length (the cmsg_len member).
6358 * Historically Berkeley-derived implementations have passed only one object
6359 * at a time, but this API allows multiple objects to be
6360 * passed in a single call to sendmsg() or recvmsg(). The following example
6361 * shows two ancillary data objects in a control buffer.
6363 * |<--------------------------- msg_controllen -------------------------->|
6366 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6368 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6371 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6373 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6376 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6377 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6379 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6381 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6388 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
6390 struct cmsghdr *cmsg;
6391 struct msghdr *my_msg = (struct msghdr *)msg;
6393 for (cmsg = CMSG_FIRSTHDR(msg);
6395 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6396 if (!CMSG_OK(my_msg, cmsg))
6399 /* Should we parse this header or ignore? */
6400 if (cmsg->cmsg_level != IPPROTO_SCTP)
6403 /* Strictly check lengths following example in SCM code. */
6404 switch (cmsg->cmsg_type) {
6406 /* SCTP Socket API Extension
6407 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6409 * This cmsghdr structure provides information for
6410 * initializing new SCTP associations with sendmsg().
6411 * The SCTP_INITMSG socket option uses this same data
6412 * structure. This structure is not used for
6415 * cmsg_level cmsg_type cmsg_data[]
6416 * ------------ ------------ ----------------------
6417 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6419 if (cmsg->cmsg_len !=
6420 CMSG_LEN(sizeof(struct sctp_initmsg)))
6422 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6426 /* SCTP Socket API Extension
6427 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6429 * This cmsghdr structure specifies SCTP options for
6430 * sendmsg() and describes SCTP header information
6431 * about a received message through recvmsg().
6433 * cmsg_level cmsg_type cmsg_data[]
6434 * ------------ ------------ ----------------------
6435 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6437 if (cmsg->cmsg_len !=
6438 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6442 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6444 /* Minimally, validate the sinfo_flags. */
6445 if (cmsgs->info->sinfo_flags &
6446 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6447 SCTP_ABORT | SCTP_EOF))
6459 * Wait for a packet..
6460 * Note: This function is the same function as in core/datagram.c
6461 * with a few modifications to make lksctp work.
6463 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
6468 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6470 /* Socket errors? */
6471 error = sock_error(sk);
6475 if (!skb_queue_empty(&sk->sk_receive_queue))
6478 /* Socket shut down? */
6479 if (sk->sk_shutdown & RCV_SHUTDOWN)
6482 /* Sequenced packets can come disconnected. If so we report the
6487 /* Is there a good reason to think that we may receive some data? */
6488 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6491 /* Handle signals. */
6492 if (signal_pending(current))
6495 /* Let another process have a go. Since we are going to sleep
6496 * anyway. Note: This may cause odd behaviors if the message
6497 * does not fit in the user's buffer, but this seems to be the
6498 * only way to honor MSG_DONTWAIT realistically.
6501 *timeo_p = schedule_timeout(*timeo_p);
6505 finish_wait(sk_sleep(sk), &wait);
6509 error = sock_intr_errno(*timeo_p);
6512 finish_wait(sk_sleep(sk), &wait);
6517 /* Receive a datagram.
6518 * Note: This is pretty much the same routine as in core/datagram.c
6519 * with a few changes to make lksctp work.
6521 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6522 int noblock, int *err)
6525 struct sk_buff *skb;
6528 timeo = sock_rcvtimeo(sk, noblock);
6530 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
6531 MAX_SCHEDULE_TIMEOUT);
6534 /* Again only user level code calls this function,
6535 * so nothing interrupt level
6536 * will suddenly eat the receive_queue.
6538 * Look at current nfs client by the way...
6539 * However, this function was correct in any case. 8)
6541 if (flags & MSG_PEEK) {
6542 spin_lock_bh(&sk->sk_receive_queue.lock);
6543 skb = skb_peek(&sk->sk_receive_queue);
6545 atomic_inc(&skb->users);
6546 spin_unlock_bh(&sk->sk_receive_queue.lock);
6548 skb = skb_dequeue(&sk->sk_receive_queue);
6554 /* Caller is allowed not to check sk->sk_err before calling. */
6555 error = sock_error(sk);
6559 if (sk->sk_shutdown & RCV_SHUTDOWN)
6562 /* User doesn't want to wait. */
6566 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6575 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6576 static void __sctp_write_space(struct sctp_association *asoc)
6578 struct sock *sk = asoc->base.sk;
6579 struct socket *sock = sk->sk_socket;
6581 if ((sctp_wspace(asoc) > 0) && sock) {
6582 if (waitqueue_active(&asoc->wait))
6583 wake_up_interruptible(&asoc->wait);
6585 if (sctp_writeable(sk)) {
6586 wait_queue_head_t *wq = sk_sleep(sk);
6588 if (wq && waitqueue_active(wq))
6589 wake_up_interruptible(wq);
6591 /* Note that we try to include the Async I/O support
6592 * here by modeling from the current TCP/UDP code.
6593 * We have not tested with it yet.
6595 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6596 sock_wake_async(sock,
6597 SOCK_WAKE_SPACE, POLL_OUT);
6602 static void sctp_wake_up_waiters(struct sock *sk,
6603 struct sctp_association *asoc)
6605 struct sctp_association *tmp = asoc;
6607 /* We do accounting for the sndbuf space per association,
6608 * so we only need to wake our own association.
6610 if (asoc->ep->sndbuf_policy)
6611 return __sctp_write_space(asoc);
6613 /* If association goes down and is just flushing its
6614 * outq, then just normally notify others.
6616 if (asoc->base.dead)
6617 return sctp_write_space(sk);
6619 /* Accounting for the sndbuf space is per socket, so we
6620 * need to wake up others, try to be fair and in case of
6621 * other associations, let them have a go first instead
6622 * of just doing a sctp_write_space() call.
6624 * Note that we reach sctp_wake_up_waiters() only when
6625 * associations free up queued chunks, thus we are under
6626 * lock and the list of associations on a socket is
6627 * guaranteed not to change.
6629 for (tmp = list_next_entry(tmp, asocs); 1;
6630 tmp = list_next_entry(tmp, asocs)) {
6631 /* Manually skip the head element. */
6632 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
6634 /* Wake up association. */
6635 __sctp_write_space(tmp);
6636 /* We've reached the end. */
6642 /* Do accounting for the sndbuf space.
6643 * Decrement the used sndbuf space of the corresponding association by the
6644 * data size which was just transmitted(freed).
6646 static void sctp_wfree(struct sk_buff *skb)
6648 struct sctp_association *asoc;
6649 struct sctp_chunk *chunk;
6652 /* Get the saved chunk pointer. */
6653 chunk = *((struct sctp_chunk **)(skb->cb));
6656 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6657 sizeof(struct sk_buff) +
6658 sizeof(struct sctp_chunk);
6660 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6663 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6665 sk->sk_wmem_queued -= skb->truesize;
6666 sk_mem_uncharge(sk, skb->truesize);
6669 sctp_wake_up_waiters(sk, asoc);
6671 sctp_association_put(asoc);
6674 /* Do accounting for the receive space on the socket.
6675 * Accounting for the association is done in ulpevent.c
6676 * We set this as a destructor for the cloned data skbs so that
6677 * accounting is done at the correct time.
6679 void sctp_sock_rfree(struct sk_buff *skb)
6681 struct sock *sk = skb->sk;
6682 struct sctp_ulpevent *event = sctp_skb2event(skb);
6684 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6687 * Mimic the behavior of sock_rfree
6689 sk_mem_uncharge(sk, event->rmem_len);
6693 /* Helper function to wait for space in the sndbuf. */
6694 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6697 struct sock *sk = asoc->base.sk;
6699 long current_timeo = *timeo_p;
6702 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
6705 /* Increment the association's refcnt. */
6706 sctp_association_hold(asoc);
6708 /* Wait on the association specific sndbuf space. */
6710 prepare_to_wait_exclusive(&asoc->wait, &wait,
6711 TASK_INTERRUPTIBLE);
6714 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6717 if (signal_pending(current))
6718 goto do_interrupted;
6719 if (msg_len <= sctp_wspace(asoc))
6722 /* Let another process have a go. Since we are going
6726 current_timeo = schedule_timeout(current_timeo);
6727 BUG_ON(sk != asoc->base.sk);
6730 *timeo_p = current_timeo;
6734 finish_wait(&asoc->wait, &wait);
6736 /* Release the association's refcnt. */
6737 sctp_association_put(asoc);
6746 err = sock_intr_errno(*timeo_p);
6754 void sctp_data_ready(struct sock *sk)
6756 struct socket_wq *wq;
6759 wq = rcu_dereference(sk->sk_wq);
6760 if (wq_has_sleeper(wq))
6761 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6762 POLLRDNORM | POLLRDBAND);
6763 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6767 /* If socket sndbuf has changed, wake up all per association waiters. */
6768 void sctp_write_space(struct sock *sk)
6770 struct sctp_association *asoc;
6772 /* Wake up the tasks in each wait queue. */
6773 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6774 __sctp_write_space(asoc);
6778 /* Is there any sndbuf space available on the socket?
6780 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6781 * associations on the same socket. For a UDP-style socket with
6782 * multiple associations, it is possible for it to be "unwriteable"
6783 * prematurely. I assume that this is acceptable because
6784 * a premature "unwriteable" is better than an accidental "writeable" which
6785 * would cause an unwanted block under certain circumstances. For the 1-1
6786 * UDP-style sockets or TCP-style sockets, this code should work.
6789 static int sctp_writeable(struct sock *sk)
6793 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6799 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6800 * returns immediately with EINPROGRESS.
6802 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6804 struct sock *sk = asoc->base.sk;
6806 long current_timeo = *timeo_p;
6809 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
6811 /* Increment the association's refcnt. */
6812 sctp_association_hold(asoc);
6815 prepare_to_wait_exclusive(&asoc->wait, &wait,
6816 TASK_INTERRUPTIBLE);
6819 if (sk->sk_shutdown & RCV_SHUTDOWN)
6821 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6824 if (signal_pending(current))
6825 goto do_interrupted;
6827 if (sctp_state(asoc, ESTABLISHED))
6830 /* Let another process have a go. Since we are going
6834 current_timeo = schedule_timeout(current_timeo);
6837 *timeo_p = current_timeo;
6841 finish_wait(&asoc->wait, &wait);
6843 /* Release the association's refcnt. */
6844 sctp_association_put(asoc);
6849 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6852 err = -ECONNREFUSED;
6856 err = sock_intr_errno(*timeo_p);
6864 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6866 struct sctp_endpoint *ep;
6870 ep = sctp_sk(sk)->ep;
6874 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6875 TASK_INTERRUPTIBLE);
6877 if (list_empty(&ep->asocs)) {
6879 timeo = schedule_timeout(timeo);
6884 if (!sctp_sstate(sk, LISTENING))
6888 if (!list_empty(&ep->asocs))
6891 err = sock_intr_errno(timeo);
6892 if (signal_pending(current))
6900 finish_wait(sk_sleep(sk), &wait);
6905 static void sctp_wait_for_close(struct sock *sk, long timeout)
6910 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6911 if (list_empty(&sctp_sk(sk)->ep->asocs))
6914 timeout = schedule_timeout(timeout);
6916 } while (!signal_pending(current) && timeout);
6918 finish_wait(sk_sleep(sk), &wait);
6921 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6923 struct sk_buff *frag;
6928 /* Don't forget the fragments. */
6929 skb_walk_frags(skb, frag)
6930 sctp_skb_set_owner_r_frag(frag, sk);
6933 sctp_skb_set_owner_r(skb, sk);
6936 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6937 struct sctp_association *asoc)
6939 struct inet_sock *inet = inet_sk(sk);
6940 struct inet_sock *newinet;
6942 newsk->sk_type = sk->sk_type;
6943 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6944 newsk->sk_flags = sk->sk_flags;
6945 newsk->sk_no_check = sk->sk_no_check;
6946 newsk->sk_reuse = sk->sk_reuse;
6948 newsk->sk_shutdown = sk->sk_shutdown;
6949 newsk->sk_destruct = sctp_destruct_sock;
6950 newsk->sk_family = sk->sk_family;
6951 newsk->sk_protocol = IPPROTO_SCTP;
6952 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6953 newsk->sk_sndbuf = sk->sk_sndbuf;
6954 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6955 newsk->sk_lingertime = sk->sk_lingertime;
6956 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6957 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6959 newinet = inet_sk(newsk);
6961 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6962 * getsockname() and getpeername()
6964 newinet->inet_sport = inet->inet_sport;
6965 newinet->inet_saddr = inet->inet_saddr;
6966 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6967 newinet->inet_dport = htons(asoc->peer.port);
6968 newinet->pmtudisc = inet->pmtudisc;
6969 newinet->inet_id = asoc->next_tsn ^ jiffies;
6971 newinet->uc_ttl = inet->uc_ttl;
6972 newinet->mc_loop = 1;
6973 newinet->mc_ttl = 1;
6974 newinet->mc_index = 0;
6975 newinet->mc_list = NULL;
6978 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6979 * and its messages to the newsk.
6981 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6982 struct sctp_association *assoc,
6983 sctp_socket_type_t type)
6985 struct sctp_sock *oldsp = sctp_sk(oldsk);
6986 struct sctp_sock *newsp = sctp_sk(newsk);
6987 struct sctp_bind_bucket *pp; /* hash list port iterator */
6988 struct sctp_endpoint *newep = newsp->ep;
6989 struct sk_buff *skb, *tmp;
6990 struct sctp_ulpevent *event;
6991 struct sctp_bind_hashbucket *head;
6992 struct list_head tmplist;
6994 /* Migrate socket buffer sizes and all the socket level options to the
6997 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6998 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6999 /* Brute force copy old sctp opt. */
7000 if (oldsp->do_auto_asconf) {
7001 memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
7002 inet_sk_copy_descendant(newsk, oldsk);
7003 memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
7005 inet_sk_copy_descendant(newsk, oldsk);
7007 /* Restore the ep value that was overwritten with the above structure
7013 /* Hook this new socket in to the bind_hash list. */
7014 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
7015 inet_sk(oldsk)->inet_num)];
7017 spin_lock(&head->lock);
7018 pp = sctp_sk(oldsk)->bind_hash;
7019 sk_add_bind_node(newsk, &pp->owner);
7020 sctp_sk(newsk)->bind_hash = pp;
7021 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
7022 spin_unlock(&head->lock);
7025 /* Copy the bind_addr list from the original endpoint to the new
7026 * endpoint so that we can handle restarts properly
7028 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
7029 &oldsp->ep->base.bind_addr, GFP_KERNEL);
7031 /* Move any messages in the old socket's receive queue that are for the
7032 * peeled off association to the new socket's receive queue.
7034 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
7035 event = sctp_skb2event(skb);
7036 if (event->asoc == assoc) {
7037 __skb_unlink(skb, &oldsk->sk_receive_queue);
7038 __skb_queue_tail(&newsk->sk_receive_queue, skb);
7039 sctp_skb_set_owner_r_frag(skb, newsk);
7043 /* Clean up any messages pending delivery due to partial
7044 * delivery. Three cases:
7045 * 1) No partial deliver; no work.
7046 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7047 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7049 skb_queue_head_init(&newsp->pd_lobby);
7050 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
7052 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
7053 struct sk_buff_head *queue;
7055 /* Decide which queue to move pd_lobby skbs to. */
7056 if (assoc->ulpq.pd_mode) {
7057 queue = &newsp->pd_lobby;
7059 queue = &newsk->sk_receive_queue;
7061 /* Walk through the pd_lobby, looking for skbs that
7062 * need moved to the new socket.
7064 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
7065 event = sctp_skb2event(skb);
7066 if (event->asoc == assoc) {
7067 __skb_unlink(skb, &oldsp->pd_lobby);
7068 __skb_queue_tail(queue, skb);
7069 sctp_skb_set_owner_r_frag(skb, newsk);
7073 /* Clear up any skbs waiting for the partial
7074 * delivery to finish.
7076 if (assoc->ulpq.pd_mode)
7077 sctp_clear_pd(oldsk, NULL);
7081 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7082 sctp_skb_set_owner_r_frag(skb, newsk);
7084 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7085 sctp_skb_set_owner_r_frag(skb, newsk);
7087 /* Set the type of socket to indicate that it is peeled off from the
7088 * original UDP-style socket or created with the accept() call on a
7089 * TCP-style socket..
7093 /* Mark the new socket "in-use" by the user so that any packets
7094 * that may arrive on the association after we've moved it are
7095 * queued to the backlog. This prevents a potential race between
7096 * backlog processing on the old socket and new-packet processing
7097 * on the new socket.
7099 * The caller has just allocated newsk so we can guarantee that other
7100 * paths won't try to lock it and then oldsk.
7102 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7103 sctp_assoc_migrate(assoc, newsk);
7105 /* If the association on the newsk is already closed before accept()
7106 * is called, set RCV_SHUTDOWN flag.
7108 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
7109 newsk->sk_shutdown |= RCV_SHUTDOWN;
7111 newsk->sk_state = SCTP_SS_ESTABLISHED;
7112 release_sock(newsk);
7116 /* This proto struct describes the ULP interface for SCTP. */
7117 struct proto sctp_prot = {
7119 .owner = THIS_MODULE,
7120 .close = sctp_close,
7121 .connect = sctp_connect,
7122 .disconnect = sctp_disconnect,
7123 .accept = sctp_accept,
7124 .ioctl = sctp_ioctl,
7125 .init = sctp_init_sock,
7126 .destroy = sctp_destroy_sock,
7127 .shutdown = sctp_shutdown,
7128 .setsockopt = sctp_setsockopt,
7129 .getsockopt = sctp_getsockopt,
7130 .sendmsg = sctp_sendmsg,
7131 .recvmsg = sctp_recvmsg,
7133 .backlog_rcv = sctp_backlog_rcv,
7135 .unhash = sctp_unhash,
7136 .get_port = sctp_get_port,
7137 .obj_size = sizeof(struct sctp_sock),
7138 .sysctl_mem = sysctl_sctp_mem,
7139 .sysctl_rmem = sysctl_sctp_rmem,
7140 .sysctl_wmem = sysctl_sctp_wmem,
7141 .memory_pressure = &sctp_memory_pressure,
7142 .enter_memory_pressure = sctp_enter_memory_pressure,
7143 .memory_allocated = &sctp_memory_allocated,
7144 .sockets_allocated = &sctp_sockets_allocated,
7147 #if IS_ENABLED(CONFIG_IPV6)
7149 struct proto sctpv6_prot = {
7151 .owner = THIS_MODULE,
7152 .close = sctp_close,
7153 .connect = sctp_connect,
7154 .disconnect = sctp_disconnect,
7155 .accept = sctp_accept,
7156 .ioctl = sctp_ioctl,
7157 .init = sctp_init_sock,
7158 .destroy = sctp_destroy_sock,
7159 .shutdown = sctp_shutdown,
7160 .setsockopt = sctp_setsockopt,
7161 .getsockopt = sctp_getsockopt,
7162 .sendmsg = sctp_sendmsg,
7163 .recvmsg = sctp_recvmsg,
7165 .backlog_rcv = sctp_backlog_rcv,
7167 .unhash = sctp_unhash,
7168 .get_port = sctp_get_port,
7169 .obj_size = sizeof(struct sctp6_sock),
7170 .sysctl_mem = sysctl_sctp_mem,
7171 .sysctl_rmem = sysctl_sctp_rmem,
7172 .sysctl_wmem = sysctl_sctp_wmem,
7173 .memory_pressure = &sctp_memory_pressure,
7174 .enter_memory_pressure = sctp_enter_memory_pressure,
7175 .memory_allocated = &sctp_memory_allocated,
7176 .sockets_allocated = &sctp_sockets_allocated,
7178 #endif /* IS_ENABLED(CONFIG_IPV6) */