1 /* SCTP kernel implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
9 * This file is part of the SCTP kernel implementation
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
18 * This SCTP implementation is free software;
19 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
24 * This SCTP implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, write to
32 * the Free Software Foundation, 59 Temple Place - Suite 330,
33 * Boston, MA 02111-1307, USA.
35 * Please send any bug reports or fixes you make to the
37 * lksctp developers <linux-sctp@vger.kernel.org>
39 * Written or modified by:
40 * La Monte H.P. Yarroll <piggy@acm.org>
41 * Narasimha Budihal <narsi@refcode.org>
42 * Karl Knutson <karl@athena.chicago.il.us>
43 * Jon Grimm <jgrimm@us.ibm.com>
44 * Xingang Guo <xingang.guo@intel.com>
45 * Daisy Chang <daisyc@us.ibm.com>
46 * Sridhar Samudrala <samudrala@us.ibm.com>
47 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
48 * Ardelle Fan <ardelle.fan@intel.com>
49 * Ryan Layer <rmlayer@us.ibm.com>
50 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
51 * Kevin Gao <kevin.gao@intel.com>
54 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
56 #include <linux/types.h>
57 #include <linux/kernel.h>
58 #include <linux/wait.h>
59 #include <linux/time.h>
61 #include <linux/capability.h>
62 #include <linux/fcntl.h>
63 #include <linux/poll.h>
64 #include <linux/init.h>
65 #include <linux/crypto.h>
66 #include <linux/slab.h>
67 #include <linux/file.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,
288 sctp_release_sock(sk);
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 defferent 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 static int sctp_getsockopt_connectx3(struct sock* sk, int len,
1378 char __user *optval,
1381 struct sctp_getaddrs_old param;
1382 sctp_assoc_t assoc_id = 0;
1385 if (len < sizeof(param))
1388 if (copy_from_user(¶m, optval, sizeof(param)))
1391 err = __sctp_setsockopt_connectx(sk,
1392 (struct sockaddr __user *)param.addrs,
1393 param.addr_num, &assoc_id);
1395 if (err == 0 || err == -EINPROGRESS) {
1396 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1398 if (put_user(sizeof(assoc_id), optlen))
1405 /* API 3.1.4 close() - UDP Style Syntax
1406 * Applications use close() to perform graceful shutdown (as described in
1407 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1408 * by a UDP-style socket.
1412 * ret = close(int sd);
1414 * sd - the socket descriptor of the associations to be closed.
1416 * To gracefully shutdown a specific association represented by the
1417 * UDP-style socket, an application should use the sendmsg() call,
1418 * passing no user data, but including the appropriate flag in the
1419 * ancillary data (see Section xxxx).
1421 * If sd in the close() call is a branched-off socket representing only
1422 * one association, the shutdown is performed on that association only.
1424 * 4.1.6 close() - TCP Style Syntax
1426 * Applications use close() to gracefully close down an association.
1430 * int close(int sd);
1432 * sd - the socket descriptor of the association to be closed.
1434 * After an application calls close() on a socket descriptor, no further
1435 * socket operations will succeed on that descriptor.
1437 * API 7.1.4 SO_LINGER
1439 * An application using the TCP-style socket can use this option to
1440 * perform the SCTP ABORT primitive. The linger option structure is:
1443 * int l_onoff; // option on/off
1444 * int l_linger; // linger time
1447 * To enable the option, set l_onoff to 1. If the l_linger value is set
1448 * to 0, calling close() is the same as the ABORT primitive. If the
1449 * value is set to a negative value, the setsockopt() call will return
1450 * an error. If the value is set to a positive value linger_time, the
1451 * close() can be blocked for at most linger_time ms. If the graceful
1452 * shutdown phase does not finish during this period, close() will
1453 * return but the graceful shutdown phase continues in the system.
1455 static void sctp_close(struct sock *sk, long timeout)
1457 struct net *net = sock_net(sk);
1458 struct sctp_endpoint *ep;
1459 struct sctp_association *asoc;
1460 struct list_head *pos, *temp;
1461 unsigned int data_was_unread;
1463 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1466 sk->sk_shutdown = SHUTDOWN_MASK;
1467 sk->sk_state = SCTP_SS_CLOSING;
1469 ep = sctp_sk(sk)->ep;
1471 /* Clean up any skbs sitting on the receive queue. */
1472 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1473 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1475 /* Walk all associations on an endpoint. */
1476 list_for_each_safe(pos, temp, &ep->asocs) {
1477 asoc = list_entry(pos, struct sctp_association, asocs);
1479 if (sctp_style(sk, TCP)) {
1480 /* A closed association can still be in the list if
1481 * it belongs to a TCP-style listening socket that is
1482 * not yet accepted. If so, free it. If not, send an
1483 * ABORT or SHUTDOWN based on the linger options.
1485 if (sctp_state(asoc, CLOSED)) {
1486 sctp_unhash_established(asoc);
1487 sctp_association_free(asoc);
1492 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1493 !skb_queue_empty(&asoc->ulpq.reasm) ||
1494 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1495 struct sctp_chunk *chunk;
1497 chunk = sctp_make_abort_user(asoc, NULL, 0);
1499 sctp_primitive_ABORT(net, asoc, chunk);
1501 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1504 /* On a TCP-style socket, block for at most linger_time if set. */
1505 if (sctp_style(sk, TCP) && timeout)
1506 sctp_wait_for_close(sk, timeout);
1508 /* This will run the backlog queue. */
1509 sctp_release_sock(sk);
1511 /* Supposedly, no process has access to the socket, but
1512 * the net layers still may.
1514 sctp_local_bh_disable();
1515 sctp_bh_lock_sock(sk);
1517 /* Hold the sock, since sk_common_release() will put sock_put()
1518 * and we have just a little more cleanup.
1521 sk_common_release(sk);
1523 sctp_bh_unlock_sock(sk);
1524 sctp_local_bh_enable();
1528 SCTP_DBG_OBJCNT_DEC(sock);
1531 /* Handle EPIPE error. */
1532 static int sctp_error(struct sock *sk, int flags, int err)
1535 err = sock_error(sk) ? : -EPIPE;
1536 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1537 send_sig(SIGPIPE, current, 0);
1541 /* API 3.1.3 sendmsg() - UDP Style Syntax
1543 * An application uses sendmsg() and recvmsg() calls to transmit data to
1544 * and receive data from its peer.
1546 * ssize_t sendmsg(int socket, const struct msghdr *message,
1549 * socket - the socket descriptor of the endpoint.
1550 * message - pointer to the msghdr structure which contains a single
1551 * user message and possibly some ancillary data.
1553 * See Section 5 for complete description of the data
1556 * flags - flags sent or received with the user message, see Section
1557 * 5 for complete description of the flags.
1559 * Note: This function could use a rewrite especially when explicit
1560 * connect support comes in.
1562 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1564 static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1566 static int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1567 struct msghdr *msg, size_t msg_len)
1569 struct net *net = sock_net(sk);
1570 struct sctp_sock *sp;
1571 struct sctp_endpoint *ep;
1572 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1573 struct sctp_transport *transport, *chunk_tp;
1574 struct sctp_chunk *chunk;
1576 struct sockaddr *msg_name = NULL;
1577 struct sctp_sndrcvinfo default_sinfo;
1578 struct sctp_sndrcvinfo *sinfo;
1579 struct sctp_initmsg *sinit;
1580 sctp_assoc_t associd = 0;
1581 sctp_cmsgs_t cmsgs = { NULL };
1585 __u16 sinfo_flags = 0;
1586 struct sctp_datamsg *datamsg;
1587 int msg_flags = msg->msg_flags;
1593 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1596 /* We cannot send a message over a TCP-style listening socket. */
1597 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1602 /* Parse out the SCTP CMSGs. */
1603 err = sctp_msghdr_parse(msg, &cmsgs);
1605 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1609 /* Fetch the destination address for this packet. This
1610 * address only selects the association--it is not necessarily
1611 * the address we will send to.
1612 * For a peeled-off socket, msg_name is ignored.
1614 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1615 int msg_namelen = msg->msg_namelen;
1617 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1622 if (msg_namelen > sizeof(to))
1623 msg_namelen = sizeof(to);
1624 memcpy(&to, msg->msg_name, msg_namelen);
1625 msg_name = msg->msg_name;
1631 /* Did the user specify SNDRCVINFO? */
1633 sinfo_flags = sinfo->sinfo_flags;
1634 associd = sinfo->sinfo_assoc_id;
1637 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1638 msg_len, sinfo_flags);
1640 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1641 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1646 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1647 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1648 * If SCTP_ABORT is set, the message length could be non zero with
1649 * the msg_iov set to the user abort reason.
1651 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1652 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1657 /* If SCTP_ADDR_OVER is set, there must be an address
1658 * specified in msg_name.
1660 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1667 pr_debug("%s: about to look up association\n", __func__);
1671 /* If a msg_name has been specified, assume this is to be used. */
1673 /* Look for a matching association on the endpoint. */
1674 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1676 /* If we could not find a matching association on the
1677 * endpoint, make sure that it is not a TCP-style
1678 * socket that already has an association or there is
1679 * no peeled-off association on another socket.
1681 if ((sctp_style(sk, TCP) &&
1682 sctp_sstate(sk, ESTABLISHED)) ||
1683 sctp_endpoint_is_peeled_off(ep, &to)) {
1684 err = -EADDRNOTAVAIL;
1689 asoc = sctp_id2assoc(sk, associd);
1697 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1699 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1700 * socket that has an association in CLOSED state. This can
1701 * happen when an accepted socket has an association that is
1704 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1709 if (sinfo_flags & SCTP_EOF) {
1710 pr_debug("%s: shutting down association:%p\n",
1713 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1717 if (sinfo_flags & SCTP_ABORT) {
1719 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1725 pr_debug("%s: aborting association:%p\n",
1728 sctp_primitive_ABORT(net, asoc, chunk);
1734 /* Do we need to create the association? */
1736 pr_debug("%s: there is no association yet\n", __func__);
1738 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1743 /* Check for invalid stream against the stream counts,
1744 * either the default or the user specified stream counts.
1747 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1748 /* Check against the defaults. */
1749 if (sinfo->sinfo_stream >=
1750 sp->initmsg.sinit_num_ostreams) {
1755 /* Check against the requested. */
1756 if (sinfo->sinfo_stream >=
1757 sinit->sinit_num_ostreams) {
1765 * API 3.1.2 bind() - UDP Style Syntax
1766 * If a bind() or sctp_bindx() is not called prior to a
1767 * sendmsg() call that initiates a new association, the
1768 * system picks an ephemeral port and will choose an address
1769 * set equivalent to binding with a wildcard address.
1771 if (!ep->base.bind_addr.port) {
1772 if (sctp_autobind(sk)) {
1778 * If an unprivileged user inherits a one-to-many
1779 * style socket with open associations on a privileged
1780 * port, it MAY be permitted to accept new associations,
1781 * but it SHOULD NOT be permitted to open new
1784 if (ep->base.bind_addr.port < PROT_SOCK &&
1785 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1791 scope = sctp_scope(&to);
1792 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1798 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1804 /* If the SCTP_INIT ancillary data is specified, set all
1805 * the association init values accordingly.
1808 if (sinit->sinit_num_ostreams) {
1809 asoc->c.sinit_num_ostreams =
1810 sinit->sinit_num_ostreams;
1812 if (sinit->sinit_max_instreams) {
1813 asoc->c.sinit_max_instreams =
1814 sinit->sinit_max_instreams;
1816 if (sinit->sinit_max_attempts) {
1817 asoc->max_init_attempts
1818 = sinit->sinit_max_attempts;
1820 if (sinit->sinit_max_init_timeo) {
1821 asoc->max_init_timeo =
1822 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1826 /* Prime the peer's transport structures. */
1827 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1834 /* ASSERT: we have a valid association at this point. */
1835 pr_debug("%s: we have a valid association\n", __func__);
1838 /* If the user didn't specify SNDRCVINFO, make up one with
1841 memset(&default_sinfo, 0, sizeof(default_sinfo));
1842 default_sinfo.sinfo_stream = asoc->default_stream;
1843 default_sinfo.sinfo_flags = asoc->default_flags;
1844 default_sinfo.sinfo_ppid = asoc->default_ppid;
1845 default_sinfo.sinfo_context = asoc->default_context;
1846 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1847 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1848 sinfo = &default_sinfo;
1851 /* API 7.1.7, the sndbuf size per association bounds the
1852 * maximum size of data that can be sent in a single send call.
1854 if (msg_len > sk->sk_sndbuf) {
1859 if (asoc->pmtu_pending)
1860 sctp_assoc_pending_pmtu(sk, asoc);
1862 /* If fragmentation is disabled and the message length exceeds the
1863 * association fragmentation point, return EMSGSIZE. The I-D
1864 * does not specify what this error is, but this looks like
1867 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1872 /* Check for invalid stream. */
1873 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1878 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1879 if (!sctp_wspace(asoc)) {
1880 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1885 /* If an address is passed with the sendto/sendmsg call, it is used
1886 * to override the primary destination address in the TCP model, or
1887 * when SCTP_ADDR_OVER flag is set in the UDP model.
1889 if ((sctp_style(sk, TCP) && msg_name) ||
1890 (sinfo_flags & SCTP_ADDR_OVER)) {
1891 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1899 /* Auto-connect, if we aren't connected already. */
1900 if (sctp_state(asoc, CLOSED)) {
1901 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1905 pr_debug("%s: we associated primitively\n", __func__);
1908 /* Break the message into multiple chunks of maximum size. */
1909 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1910 if (IS_ERR(datamsg)) {
1911 err = PTR_ERR(datamsg);
1915 /* Now send the (possibly) fragmented message. */
1916 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1917 sctp_chunk_hold(chunk);
1919 /* Do accounting for the write space. */
1920 sctp_set_owner_w(chunk);
1922 chunk->transport = chunk_tp;
1925 /* Send it to the lower layers. Note: all chunks
1926 * must either fail or succeed. The lower layer
1927 * works that way today. Keep it that way or this
1930 err = sctp_primitive_SEND(net, asoc, datamsg);
1931 /* Did the lower layer accept the chunk? */
1933 sctp_datamsg_free(datamsg);
1937 pr_debug("%s: we sent primitively\n", __func__);
1939 sctp_datamsg_put(datamsg);
1942 /* If we are already past ASSOCIATE, the lower
1943 * layers are responsible for association cleanup.
1949 sctp_unhash_established(asoc);
1950 sctp_association_free(asoc);
1953 sctp_release_sock(sk);
1956 return sctp_error(sk, msg_flags, err);
1963 err = sock_error(sk);
1973 /* This is an extended version of skb_pull() that removes the data from the
1974 * start of a skb even when data is spread across the list of skb's in the
1975 * frag_list. len specifies the total amount of data that needs to be removed.
1976 * when 'len' bytes could be removed from the skb, it returns 0.
1977 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1978 * could not be removed.
1980 static int sctp_skb_pull(struct sk_buff *skb, int len)
1982 struct sk_buff *list;
1983 int skb_len = skb_headlen(skb);
1986 if (len <= skb_len) {
1987 __skb_pull(skb, len);
1991 __skb_pull(skb, skb_len);
1993 skb_walk_frags(skb, list) {
1994 rlen = sctp_skb_pull(list, len);
1995 skb->len -= (len-rlen);
1996 skb->data_len -= (len-rlen);
2007 /* API 3.1.3 recvmsg() - UDP Style Syntax
2009 * ssize_t recvmsg(int socket, struct msghdr *message,
2012 * socket - the socket descriptor of the endpoint.
2013 * message - pointer to the msghdr structure which contains a single
2014 * user message and possibly some ancillary data.
2016 * See Section 5 for complete description of the data
2019 * flags - flags sent or received with the user message, see Section
2020 * 5 for complete description of the flags.
2022 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2024 static int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2025 struct msghdr *msg, size_t len, int noblock,
2026 int flags, int *addr_len)
2028 struct sctp_ulpevent *event = NULL;
2029 struct sctp_sock *sp = sctp_sk(sk);
2030 struct sk_buff *skb;
2035 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2036 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2041 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2046 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2050 /* Get the total length of the skb including any skb's in the
2059 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2061 event = sctp_skb2event(skb);
2066 sock_recv_ts_and_drops(msg, sk, skb);
2067 if (sctp_ulpevent_is_notification(event)) {
2068 msg->msg_flags |= MSG_NOTIFICATION;
2069 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2071 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2074 /* Check if we allow SCTP_SNDRCVINFO. */
2075 if (sp->subscribe.sctp_data_io_event)
2076 sctp_ulpevent_read_sndrcvinfo(event, msg);
2078 /* FIXME: we should be calling IP/IPv6 layers. */
2079 if (sk->sk_protinfo.af_inet.cmsg_flags)
2080 ip_cmsg_recv(msg, skb);
2085 /* If skb's length exceeds the user's buffer, update the skb and
2086 * push it back to the receive_queue so that the next call to
2087 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2089 if (skb_len > copied) {
2090 msg->msg_flags &= ~MSG_EOR;
2091 if (flags & MSG_PEEK)
2093 sctp_skb_pull(skb, copied);
2094 skb_queue_head(&sk->sk_receive_queue, skb);
2096 /* When only partial message is copied to the user, increase
2097 * rwnd by that amount. If all the data in the skb is read,
2098 * rwnd is updated when the event is freed.
2100 if (!sctp_ulpevent_is_notification(event))
2101 sctp_assoc_rwnd_increase(event->asoc, copied);
2103 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2104 (event->msg_flags & MSG_EOR))
2105 msg->msg_flags |= MSG_EOR;
2107 msg->msg_flags &= ~MSG_EOR;
2110 if (flags & MSG_PEEK) {
2111 /* Release the skb reference acquired after peeking the skb in
2112 * sctp_skb_recv_datagram().
2116 /* Free the event which includes releasing the reference to
2117 * the owner of the skb, freeing the skb and updating the
2120 sctp_ulpevent_free(event);
2123 sctp_release_sock(sk);
2127 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2129 * This option is a on/off flag. If enabled no SCTP message
2130 * fragmentation will be performed. Instead if a message being sent
2131 * exceeds the current PMTU size, the message will NOT be sent and
2132 * instead a error will be indicated to the user.
2134 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2135 char __user *optval,
2136 unsigned int optlen)
2140 if (optlen < sizeof(int))
2143 if (get_user(val, (int __user *)optval))
2146 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2151 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2152 unsigned int optlen)
2154 struct sctp_association *asoc;
2155 struct sctp_ulpevent *event;
2157 if (optlen > sizeof(struct sctp_event_subscribe))
2159 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2163 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2164 * if there is no data to be sent or retransmit, the stack will
2165 * immediately send up this notification.
2167 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2168 &sctp_sk(sk)->subscribe)) {
2169 asoc = sctp_id2assoc(sk, 0);
2171 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2172 event = sctp_ulpevent_make_sender_dry_event(asoc,
2177 sctp_ulpq_tail_event(&asoc->ulpq, event);
2184 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2186 * This socket option is applicable to the UDP-style socket only. When
2187 * set it will cause associations that are idle for more than the
2188 * specified number of seconds to automatically close. An association
2189 * being idle is defined an association that has NOT sent or received
2190 * user data. The special value of '0' indicates that no automatic
2191 * close of any associations should be performed. The option expects an
2192 * integer defining the number of seconds of idle time before an
2193 * association is closed.
2195 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2196 unsigned int optlen)
2198 struct sctp_sock *sp = sctp_sk(sk);
2199 struct net *net = sock_net(sk);
2201 /* Applicable to UDP-style socket only */
2202 if (sctp_style(sk, TCP))
2204 if (optlen != sizeof(int))
2206 if (copy_from_user(&sp->autoclose, optval, optlen))
2209 if (sp->autoclose > net->sctp.max_autoclose)
2210 sp->autoclose = net->sctp.max_autoclose;
2215 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2217 * Applications can enable or disable heartbeats for any peer address of
2218 * an association, modify an address's heartbeat interval, force a
2219 * heartbeat to be sent immediately, and adjust the address's maximum
2220 * number of retransmissions sent before an address is considered
2221 * unreachable. The following structure is used to access and modify an
2222 * address's parameters:
2224 * struct sctp_paddrparams {
2225 * sctp_assoc_t spp_assoc_id;
2226 * struct sockaddr_storage spp_address;
2227 * uint32_t spp_hbinterval;
2228 * uint16_t spp_pathmaxrxt;
2229 * uint32_t spp_pathmtu;
2230 * uint32_t spp_sackdelay;
2231 * uint32_t spp_flags;
2234 * spp_assoc_id - (one-to-many style socket) This is filled in the
2235 * application, and identifies the association for
2237 * spp_address - This specifies which address is of interest.
2238 * spp_hbinterval - This contains the value of the heartbeat interval,
2239 * in milliseconds. If a value of zero
2240 * is present in this field then no changes are to
2241 * be made to this parameter.
2242 * spp_pathmaxrxt - This contains the maximum number of
2243 * retransmissions before this address shall be
2244 * considered unreachable. If a value of zero
2245 * is present in this field then no changes are to
2246 * be made to this parameter.
2247 * spp_pathmtu - When Path MTU discovery is disabled the value
2248 * specified here will be the "fixed" path mtu.
2249 * Note that if the spp_address field is empty
2250 * then all associations on this address will
2251 * have this fixed path mtu set upon them.
2253 * spp_sackdelay - When delayed sack is enabled, this value specifies
2254 * the number of milliseconds that sacks will be delayed
2255 * for. This value will apply to all addresses of an
2256 * association if the spp_address field is empty. Note
2257 * also, that if delayed sack is enabled and this
2258 * value is set to 0, no change is made to the last
2259 * recorded delayed sack timer value.
2261 * spp_flags - These flags are used to control various features
2262 * on an association. The flag field may contain
2263 * zero or more of the following options.
2265 * SPP_HB_ENABLE - Enable heartbeats on the
2266 * specified address. Note that if the address
2267 * field is empty all addresses for the association
2268 * have heartbeats enabled upon them.
2270 * SPP_HB_DISABLE - Disable heartbeats on the
2271 * speicifed address. Note that if the address
2272 * field is empty all addresses for the association
2273 * will have their heartbeats disabled. Note also
2274 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2275 * mutually exclusive, only one of these two should
2276 * be specified. Enabling both fields will have
2277 * undetermined results.
2279 * SPP_HB_DEMAND - Request a user initiated heartbeat
2280 * to be made immediately.
2282 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2283 * heartbeat delayis to be set to the value of 0
2286 * SPP_PMTUD_ENABLE - This field will enable PMTU
2287 * discovery upon the specified address. Note that
2288 * if the address feild is empty then all addresses
2289 * on the association are effected.
2291 * SPP_PMTUD_DISABLE - This field will disable PMTU
2292 * discovery upon the specified address. Note that
2293 * if the address feild is empty then all addresses
2294 * on the association are effected. Not also that
2295 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2296 * exclusive. Enabling both will have undetermined
2299 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2300 * on delayed sack. The time specified in spp_sackdelay
2301 * is used to specify the sack delay for this address. Note
2302 * that if spp_address is empty then all addresses will
2303 * enable delayed sack and take on the sack delay
2304 * value specified in spp_sackdelay.
2305 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2306 * off delayed sack. If the spp_address field is blank then
2307 * delayed sack is disabled for the entire association. Note
2308 * also that this field is mutually exclusive to
2309 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2312 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2313 struct sctp_transport *trans,
2314 struct sctp_association *asoc,
2315 struct sctp_sock *sp,
2318 int sackdelay_change)
2322 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2323 struct net *net = sock_net(trans->asoc->base.sk);
2325 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2330 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2331 * this field is ignored. Note also that a value of zero indicates
2332 * the current setting should be left unchanged.
2334 if (params->spp_flags & SPP_HB_ENABLE) {
2336 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2337 * set. This lets us use 0 value when this flag
2340 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2341 params->spp_hbinterval = 0;
2343 if (params->spp_hbinterval ||
2344 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2347 msecs_to_jiffies(params->spp_hbinterval);
2350 msecs_to_jiffies(params->spp_hbinterval);
2352 sp->hbinterval = params->spp_hbinterval;
2359 trans->param_flags =
2360 (trans->param_flags & ~SPP_HB) | hb_change;
2363 (asoc->param_flags & ~SPP_HB) | hb_change;
2366 (sp->param_flags & ~SPP_HB) | hb_change;
2370 /* When Path MTU discovery is disabled the value specified here will
2371 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2372 * include the flag SPP_PMTUD_DISABLE for this field to have any
2375 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2377 trans->pathmtu = params->spp_pathmtu;
2378 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2380 asoc->pathmtu = params->spp_pathmtu;
2381 sctp_frag_point(asoc, params->spp_pathmtu);
2383 sp->pathmtu = params->spp_pathmtu;
2389 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2390 (params->spp_flags & SPP_PMTUD_ENABLE);
2391 trans->param_flags =
2392 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2394 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2395 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2399 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2402 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2406 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2407 * value of this field is ignored. Note also that a value of zero
2408 * indicates the current setting should be left unchanged.
2410 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2413 msecs_to_jiffies(params->spp_sackdelay);
2416 msecs_to_jiffies(params->spp_sackdelay);
2418 sp->sackdelay = params->spp_sackdelay;
2422 if (sackdelay_change) {
2424 trans->param_flags =
2425 (trans->param_flags & ~SPP_SACKDELAY) |
2429 (asoc->param_flags & ~SPP_SACKDELAY) |
2433 (sp->param_flags & ~SPP_SACKDELAY) |
2438 /* Note that a value of zero indicates the current setting should be
2441 if (params->spp_pathmaxrxt) {
2443 trans->pathmaxrxt = params->spp_pathmaxrxt;
2445 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2447 sp->pathmaxrxt = params->spp_pathmaxrxt;
2454 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2455 char __user *optval,
2456 unsigned int optlen)
2458 struct sctp_paddrparams params;
2459 struct sctp_transport *trans = NULL;
2460 struct sctp_association *asoc = NULL;
2461 struct sctp_sock *sp = sctp_sk(sk);
2463 int hb_change, pmtud_change, sackdelay_change;
2465 if (optlen != sizeof(struct sctp_paddrparams))
2468 if (copy_from_user(¶ms, optval, optlen))
2471 /* Validate flags and value parameters. */
2472 hb_change = params.spp_flags & SPP_HB;
2473 pmtud_change = params.spp_flags & SPP_PMTUD;
2474 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2476 if (hb_change == SPP_HB ||
2477 pmtud_change == SPP_PMTUD ||
2478 sackdelay_change == SPP_SACKDELAY ||
2479 params.spp_sackdelay > 500 ||
2480 (params.spp_pathmtu &&
2481 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2484 /* If an address other than INADDR_ANY is specified, and
2485 * no transport is found, then the request is invalid.
2487 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
2488 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2489 params.spp_assoc_id);
2494 /* Get association, if assoc_id != 0 and the socket is a one
2495 * to many style socket, and an association was not found, then
2496 * the id was invalid.
2498 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2499 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2502 /* Heartbeat demand can only be sent on a transport or
2503 * association, but not a socket.
2505 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2508 /* Process parameters. */
2509 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2510 hb_change, pmtud_change,
2516 /* If changes are for association, also apply parameters to each
2519 if (!trans && asoc) {
2520 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2522 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2523 hb_change, pmtud_change,
2532 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2534 * This option will effect the way delayed acks are performed. This
2535 * option allows you to get or set the delayed ack time, in
2536 * milliseconds. It also allows changing the delayed ack frequency.
2537 * Changing the frequency to 1 disables the delayed sack algorithm. If
2538 * the assoc_id is 0, then this sets or gets the endpoints default
2539 * values. If the assoc_id field is non-zero, then the set or get
2540 * effects the specified association for the one to many model (the
2541 * assoc_id field is ignored by the one to one model). Note that if
2542 * sack_delay or sack_freq are 0 when setting this option, then the
2543 * current values will remain unchanged.
2545 * struct sctp_sack_info {
2546 * sctp_assoc_t sack_assoc_id;
2547 * uint32_t sack_delay;
2548 * uint32_t sack_freq;
2551 * sack_assoc_id - This parameter, indicates which association the user
2552 * is performing an action upon. Note that if this field's value is
2553 * zero then the endpoints default value is changed (effecting future
2554 * associations only).
2556 * sack_delay - This parameter contains the number of milliseconds that
2557 * the user is requesting the delayed ACK timer be set to. Note that
2558 * this value is defined in the standard to be between 200 and 500
2561 * sack_freq - This parameter contains the number of packets that must
2562 * be received before a sack is sent without waiting for the delay
2563 * timer to expire. The default value for this is 2, setting this
2564 * value to 1 will disable the delayed sack algorithm.
2567 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2568 char __user *optval, unsigned int optlen)
2570 struct sctp_sack_info params;
2571 struct sctp_transport *trans = NULL;
2572 struct sctp_association *asoc = NULL;
2573 struct sctp_sock *sp = sctp_sk(sk);
2575 if (optlen == sizeof(struct sctp_sack_info)) {
2576 if (copy_from_user(¶ms, optval, optlen))
2579 if (params.sack_delay == 0 && params.sack_freq == 0)
2581 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2582 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2583 pr_warn("Use struct sctp_sack_info instead\n");
2584 if (copy_from_user(¶ms, optval, optlen))
2587 if (params.sack_delay == 0)
2588 params.sack_freq = 1;
2590 params.sack_freq = 0;
2594 /* Validate value parameter. */
2595 if (params.sack_delay > 500)
2598 /* Get association, if sack_assoc_id != 0 and the socket is a one
2599 * to many style socket, and an association was not found, then
2600 * the id was invalid.
2602 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2603 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2606 if (params.sack_delay) {
2609 msecs_to_jiffies(params.sack_delay);
2611 (asoc->param_flags & ~SPP_SACKDELAY) |
2612 SPP_SACKDELAY_ENABLE;
2614 sp->sackdelay = params.sack_delay;
2616 (sp->param_flags & ~SPP_SACKDELAY) |
2617 SPP_SACKDELAY_ENABLE;
2621 if (params.sack_freq == 1) {
2624 (asoc->param_flags & ~SPP_SACKDELAY) |
2625 SPP_SACKDELAY_DISABLE;
2628 (sp->param_flags & ~SPP_SACKDELAY) |
2629 SPP_SACKDELAY_DISABLE;
2631 } else if (params.sack_freq > 1) {
2633 asoc->sackfreq = params.sack_freq;
2635 (asoc->param_flags & ~SPP_SACKDELAY) |
2636 SPP_SACKDELAY_ENABLE;
2638 sp->sackfreq = params.sack_freq;
2640 (sp->param_flags & ~SPP_SACKDELAY) |
2641 SPP_SACKDELAY_ENABLE;
2645 /* If change is for association, also apply to each transport. */
2647 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2649 if (params.sack_delay) {
2651 msecs_to_jiffies(params.sack_delay);
2652 trans->param_flags =
2653 (trans->param_flags & ~SPP_SACKDELAY) |
2654 SPP_SACKDELAY_ENABLE;
2656 if (params.sack_freq == 1) {
2657 trans->param_flags =
2658 (trans->param_flags & ~SPP_SACKDELAY) |
2659 SPP_SACKDELAY_DISABLE;
2660 } else if (params.sack_freq > 1) {
2661 trans->sackfreq = params.sack_freq;
2662 trans->param_flags =
2663 (trans->param_flags & ~SPP_SACKDELAY) |
2664 SPP_SACKDELAY_ENABLE;
2672 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2674 * Applications can specify protocol parameters for the default association
2675 * initialization. The option name argument to setsockopt() and getsockopt()
2678 * Setting initialization parameters is effective only on an unconnected
2679 * socket (for UDP-style sockets only future associations are effected
2680 * by the change). With TCP-style sockets, this option is inherited by
2681 * sockets derived from a listener socket.
2683 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2685 struct sctp_initmsg sinit;
2686 struct sctp_sock *sp = sctp_sk(sk);
2688 if (optlen != sizeof(struct sctp_initmsg))
2690 if (copy_from_user(&sinit, optval, optlen))
2693 if (sinit.sinit_num_ostreams)
2694 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2695 if (sinit.sinit_max_instreams)
2696 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2697 if (sinit.sinit_max_attempts)
2698 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2699 if (sinit.sinit_max_init_timeo)
2700 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2706 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2708 * Applications that wish to use the sendto() system call may wish to
2709 * specify a default set of parameters that would normally be supplied
2710 * through the inclusion of ancillary data. This socket option allows
2711 * such an application to set the default sctp_sndrcvinfo structure.
2712 * The application that wishes to use this socket option simply passes
2713 * in to this call the sctp_sndrcvinfo structure defined in Section
2714 * 5.2.2) The input parameters accepted by this call include
2715 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2716 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2717 * to this call if the caller is using the UDP model.
2719 static int sctp_setsockopt_default_send_param(struct sock *sk,
2720 char __user *optval,
2721 unsigned int optlen)
2723 struct sctp_sndrcvinfo info;
2724 struct sctp_association *asoc;
2725 struct sctp_sock *sp = sctp_sk(sk);
2727 if (optlen != sizeof(struct sctp_sndrcvinfo))
2729 if (copy_from_user(&info, optval, optlen))
2732 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2733 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2737 asoc->default_stream = info.sinfo_stream;
2738 asoc->default_flags = info.sinfo_flags;
2739 asoc->default_ppid = info.sinfo_ppid;
2740 asoc->default_context = info.sinfo_context;
2741 asoc->default_timetolive = info.sinfo_timetolive;
2743 sp->default_stream = info.sinfo_stream;
2744 sp->default_flags = info.sinfo_flags;
2745 sp->default_ppid = info.sinfo_ppid;
2746 sp->default_context = info.sinfo_context;
2747 sp->default_timetolive = info.sinfo_timetolive;
2753 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2755 * Requests that the local SCTP stack use the enclosed peer address as
2756 * the association primary. The enclosed address must be one of the
2757 * association peer's addresses.
2759 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2760 unsigned int optlen)
2762 struct sctp_prim prim;
2763 struct sctp_transport *trans;
2765 if (optlen != sizeof(struct sctp_prim))
2768 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2771 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2775 sctp_assoc_set_primary(trans->asoc, trans);
2781 * 7.1.5 SCTP_NODELAY
2783 * Turn on/off any Nagle-like algorithm. This means that packets are
2784 * generally sent as soon as possible and no unnecessary delays are
2785 * introduced, at the cost of more packets in the network. Expects an
2786 * integer boolean flag.
2788 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2789 unsigned int optlen)
2793 if (optlen < sizeof(int))
2795 if (get_user(val, (int __user *)optval))
2798 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2804 * 7.1.1 SCTP_RTOINFO
2806 * The protocol parameters used to initialize and bound retransmission
2807 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2808 * and modify these parameters.
2809 * All parameters are time values, in milliseconds. A value of 0, when
2810 * modifying the parameters, indicates that the current value should not
2814 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2816 struct sctp_rtoinfo rtoinfo;
2817 struct sctp_association *asoc;
2818 unsigned long rto_min, rto_max;
2819 struct sctp_sock *sp = sctp_sk(sk);
2821 if (optlen != sizeof (struct sctp_rtoinfo))
2824 if (copy_from_user(&rtoinfo, optval, optlen))
2827 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2829 /* Set the values to the specific association */
2830 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2833 rto_max = rtoinfo.srto_max;
2834 rto_min = rtoinfo.srto_min;
2837 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2839 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2842 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2844 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2846 if (rto_min > rto_max)
2850 if (rtoinfo.srto_initial != 0)
2852 msecs_to_jiffies(rtoinfo.srto_initial);
2853 asoc->rto_max = rto_max;
2854 asoc->rto_min = rto_min;
2856 /* If there is no association or the association-id = 0
2857 * set the values to the endpoint.
2859 if (rtoinfo.srto_initial != 0)
2860 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2861 sp->rtoinfo.srto_max = rto_max;
2862 sp->rtoinfo.srto_min = rto_min;
2870 * 7.1.2 SCTP_ASSOCINFO
2872 * This option is used to tune the maximum retransmission attempts
2873 * of the association.
2874 * Returns an error if the new association retransmission value is
2875 * greater than the sum of the retransmission value of the peer.
2876 * See [SCTP] for more information.
2879 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2882 struct sctp_assocparams assocparams;
2883 struct sctp_association *asoc;
2885 if (optlen != sizeof(struct sctp_assocparams))
2887 if (copy_from_user(&assocparams, optval, optlen))
2890 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2892 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2895 /* Set the values to the specific association */
2897 if (assocparams.sasoc_asocmaxrxt != 0) {
2900 struct sctp_transport *peer_addr;
2902 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2904 path_sum += peer_addr->pathmaxrxt;
2908 /* Only validate asocmaxrxt if we have more than
2909 * one path/transport. We do this because path
2910 * retransmissions are only counted when we have more
2914 assocparams.sasoc_asocmaxrxt > path_sum)
2917 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2920 if (assocparams.sasoc_cookie_life != 0)
2921 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
2923 /* Set the values to the endpoint */
2924 struct sctp_sock *sp = sctp_sk(sk);
2926 if (assocparams.sasoc_asocmaxrxt != 0)
2927 sp->assocparams.sasoc_asocmaxrxt =
2928 assocparams.sasoc_asocmaxrxt;
2929 if (assocparams.sasoc_cookie_life != 0)
2930 sp->assocparams.sasoc_cookie_life =
2931 assocparams.sasoc_cookie_life;
2937 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2939 * This socket option is a boolean flag which turns on or off mapped V4
2940 * addresses. If this option is turned on and the socket is type
2941 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2942 * If this option is turned off, then no mapping will be done of V4
2943 * addresses and a user will receive both PF_INET6 and PF_INET type
2944 * addresses on the socket.
2946 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2949 struct sctp_sock *sp = sctp_sk(sk);
2951 if (optlen < sizeof(int))
2953 if (get_user(val, (int __user *)optval))
2964 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2965 * This option will get or set the maximum size to put in any outgoing
2966 * SCTP DATA chunk. If a message is larger than this size it will be
2967 * fragmented by SCTP into the specified size. Note that the underlying
2968 * SCTP implementation may fragment into smaller sized chunks when the
2969 * PMTU of the underlying association is smaller than the value set by
2970 * the user. The default value for this option is '0' which indicates
2971 * the user is NOT limiting fragmentation and only the PMTU will effect
2972 * SCTP's choice of DATA chunk size. Note also that values set larger
2973 * than the maximum size of an IP datagram will effectively let SCTP
2974 * control fragmentation (i.e. the same as setting this option to 0).
2976 * The following structure is used to access and modify this parameter:
2978 * struct sctp_assoc_value {
2979 * sctp_assoc_t assoc_id;
2980 * uint32_t assoc_value;
2983 * assoc_id: This parameter is ignored for one-to-one style sockets.
2984 * For one-to-many style sockets this parameter indicates which
2985 * association the user is performing an action upon. Note that if
2986 * this field's value is zero then the endpoints default value is
2987 * changed (effecting future associations only).
2988 * assoc_value: This parameter specifies the maximum size in bytes.
2990 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2992 struct sctp_assoc_value params;
2993 struct sctp_association *asoc;
2994 struct sctp_sock *sp = sctp_sk(sk);
2997 if (optlen == sizeof(int)) {
2998 pr_warn("Use of int in maxseg socket option deprecated\n");
2999 pr_warn("Use struct sctp_assoc_value instead\n");
3000 if (copy_from_user(&val, optval, optlen))
3002 params.assoc_id = 0;
3003 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3004 if (copy_from_user(¶ms, optval, optlen))
3006 val = params.assoc_value;
3010 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3013 asoc = sctp_id2assoc(sk, params.assoc_id);
3014 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3019 val = asoc->pathmtu;
3020 val -= sp->pf->af->net_header_len;
3021 val -= sizeof(struct sctphdr) +
3022 sizeof(struct sctp_data_chunk);
3024 asoc->user_frag = val;
3025 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3027 sp->user_frag = val;
3035 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3037 * Requests that the peer mark the enclosed address as the association
3038 * primary. The enclosed address must be one of the association's
3039 * locally bound addresses. The following structure is used to make a
3040 * set primary request:
3042 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3043 unsigned int optlen)
3045 struct net *net = sock_net(sk);
3046 struct sctp_sock *sp;
3047 struct sctp_association *asoc = NULL;
3048 struct sctp_setpeerprim prim;
3049 struct sctp_chunk *chunk;
3055 if (!net->sctp.addip_enable)
3058 if (optlen != sizeof(struct sctp_setpeerprim))
3061 if (copy_from_user(&prim, optval, optlen))
3064 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3068 if (!asoc->peer.asconf_capable)
3071 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3074 if (!sctp_state(asoc, ESTABLISHED))
3077 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3081 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3082 return -EADDRNOTAVAIL;
3084 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3085 return -EADDRNOTAVAIL;
3087 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3088 chunk = sctp_make_asconf_set_prim(asoc,
3089 (union sctp_addr *)&prim.sspp_addr);
3093 err = sctp_send_asconf(asoc, chunk);
3095 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3100 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3101 unsigned int optlen)
3103 struct sctp_setadaptation adaptation;
3105 if (optlen != sizeof(struct sctp_setadaptation))
3107 if (copy_from_user(&adaptation, optval, optlen))
3110 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3116 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3118 * The context field in the sctp_sndrcvinfo structure is normally only
3119 * used when a failed message is retrieved holding the value that was
3120 * sent down on the actual send call. This option allows the setting of
3121 * a default context on an association basis that will be received on
3122 * reading messages from the peer. This is especially helpful in the
3123 * one-2-many model for an application to keep some reference to an
3124 * internal state machine that is processing messages on the
3125 * association. Note that the setting of this value only effects
3126 * received messages from the peer and does not effect the value that is
3127 * saved with outbound messages.
3129 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3130 unsigned int optlen)
3132 struct sctp_assoc_value params;
3133 struct sctp_sock *sp;
3134 struct sctp_association *asoc;
3136 if (optlen != sizeof(struct sctp_assoc_value))
3138 if (copy_from_user(¶ms, optval, optlen))
3143 if (params.assoc_id != 0) {
3144 asoc = sctp_id2assoc(sk, params.assoc_id);
3147 asoc->default_rcv_context = params.assoc_value;
3149 sp->default_rcv_context = params.assoc_value;
3156 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3158 * This options will at a minimum specify if the implementation is doing
3159 * fragmented interleave. Fragmented interleave, for a one to many
3160 * socket, is when subsequent calls to receive a message may return
3161 * parts of messages from different associations. Some implementations
3162 * may allow you to turn this value on or off. If so, when turned off,
3163 * no fragment interleave will occur (which will cause a head of line
3164 * blocking amongst multiple associations sharing the same one to many
3165 * socket). When this option is turned on, then each receive call may
3166 * come from a different association (thus the user must receive data
3167 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3168 * association each receive belongs to.
3170 * This option takes a boolean value. A non-zero value indicates that
3171 * fragmented interleave is on. A value of zero indicates that
3172 * fragmented interleave is off.
3174 * Note that it is important that an implementation that allows this
3175 * option to be turned on, have it off by default. Otherwise an unaware
3176 * application using the one to many model may become confused and act
3179 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3180 char __user *optval,
3181 unsigned int optlen)
3185 if (optlen != sizeof(int))
3187 if (get_user(val, (int __user *)optval))
3190 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3196 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3197 * (SCTP_PARTIAL_DELIVERY_POINT)
3199 * This option will set or get the SCTP partial delivery point. This
3200 * point is the size of a message where the partial delivery API will be
3201 * invoked to help free up rwnd space for the peer. Setting this to a
3202 * lower value will cause partial deliveries to happen more often. The
3203 * calls argument is an integer that sets or gets the partial delivery
3204 * point. Note also that the call will fail if the user attempts to set
3205 * this value larger than the socket receive buffer size.
3207 * Note that any single message having a length smaller than or equal to
3208 * the SCTP partial delivery point will be delivered in one single read
3209 * call as long as the user provided buffer is large enough to hold the
3212 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3213 char __user *optval,
3214 unsigned int optlen)
3218 if (optlen != sizeof(u32))
3220 if (get_user(val, (int __user *)optval))
3223 /* Note: We double the receive buffer from what the user sets
3224 * it to be, also initial rwnd is based on rcvbuf/2.
3226 if (val > (sk->sk_rcvbuf >> 1))
3229 sctp_sk(sk)->pd_point = val;
3231 return 0; /* is this the right error code? */
3235 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3237 * This option will allow a user to change the maximum burst of packets
3238 * that can be emitted by this association. Note that the default value
3239 * is 4, and some implementations may restrict this setting so that it
3240 * can only be lowered.
3242 * NOTE: This text doesn't seem right. Do this on a socket basis with
3243 * future associations inheriting the socket value.
3245 static int sctp_setsockopt_maxburst(struct sock *sk,
3246 char __user *optval,
3247 unsigned int optlen)
3249 struct sctp_assoc_value params;
3250 struct sctp_sock *sp;
3251 struct sctp_association *asoc;
3255 if (optlen == sizeof(int)) {
3256 pr_warn("Use of int in max_burst socket option deprecated\n");
3257 pr_warn("Use struct sctp_assoc_value instead\n");
3258 if (copy_from_user(&val, optval, optlen))
3260 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3261 if (copy_from_user(¶ms, optval, optlen))
3263 val = params.assoc_value;
3264 assoc_id = params.assoc_id;
3270 if (assoc_id != 0) {
3271 asoc = sctp_id2assoc(sk, assoc_id);
3274 asoc->max_burst = val;
3276 sp->max_burst = val;
3282 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3284 * This set option adds a chunk type that the user is requesting to be
3285 * received only in an authenticated way. Changes to the list of chunks
3286 * will only effect future associations on the socket.
3288 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3289 char __user *optval,
3290 unsigned int optlen)
3292 struct net *net = sock_net(sk);
3293 struct sctp_authchunk val;
3295 if (!net->sctp.auth_enable)
3298 if (optlen != sizeof(struct sctp_authchunk))
3300 if (copy_from_user(&val, optval, optlen))
3303 switch (val.sauth_chunk) {
3305 case SCTP_CID_INIT_ACK:
3306 case SCTP_CID_SHUTDOWN_COMPLETE:
3311 /* add this chunk id to the endpoint */
3312 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3316 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3318 * This option gets or sets the list of HMAC algorithms that the local
3319 * endpoint requires the peer to use.
3321 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3322 char __user *optval,
3323 unsigned int optlen)
3325 struct net *net = sock_net(sk);
3326 struct sctp_hmacalgo *hmacs;
3330 if (!net->sctp.auth_enable)
3333 if (optlen < sizeof(struct sctp_hmacalgo))
3336 hmacs= memdup_user(optval, optlen);
3338 return PTR_ERR(hmacs);
3340 idents = hmacs->shmac_num_idents;
3341 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3342 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3347 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3354 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3356 * This option will set a shared secret key which is used to build an
3357 * association shared key.
3359 static int sctp_setsockopt_auth_key(struct sock *sk,
3360 char __user *optval,
3361 unsigned int optlen)
3363 struct net *net = sock_net(sk);
3364 struct sctp_authkey *authkey;
3365 struct sctp_association *asoc;
3368 if (!net->sctp.auth_enable)
3371 if (optlen <= sizeof(struct sctp_authkey))
3374 authkey= memdup_user(optval, optlen);
3375 if (IS_ERR(authkey))
3376 return PTR_ERR(authkey);
3378 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3383 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3384 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3389 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3396 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3398 * This option will get or set the active shared key to be used to build
3399 * the association shared key.
3401 static int sctp_setsockopt_active_key(struct sock *sk,
3402 char __user *optval,
3403 unsigned int optlen)
3405 struct net *net = sock_net(sk);
3406 struct sctp_authkeyid val;
3407 struct sctp_association *asoc;
3409 if (!net->sctp.auth_enable)
3412 if (optlen != sizeof(struct sctp_authkeyid))
3414 if (copy_from_user(&val, optval, optlen))
3417 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3418 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3421 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3422 val.scact_keynumber);
3426 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3428 * This set option will delete a shared secret key from use.
3430 static int sctp_setsockopt_del_key(struct sock *sk,
3431 char __user *optval,
3432 unsigned int optlen)
3434 struct net *net = sock_net(sk);
3435 struct sctp_authkeyid val;
3436 struct sctp_association *asoc;
3438 if (!net->sctp.auth_enable)
3441 if (optlen != sizeof(struct sctp_authkeyid))
3443 if (copy_from_user(&val, optval, optlen))
3446 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3447 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3450 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3451 val.scact_keynumber);
3456 * 8.1.23 SCTP_AUTO_ASCONF
3458 * This option will enable or disable the use of the automatic generation of
3459 * ASCONF chunks to add and delete addresses to an existing association. Note
3460 * that this option has two caveats namely: a) it only affects sockets that
3461 * are bound to all addresses available to the SCTP stack, and b) the system
3462 * administrator may have an overriding control that turns the ASCONF feature
3463 * off no matter what setting the socket option may have.
3464 * This option expects an integer boolean flag, where a non-zero value turns on
3465 * the option, and a zero value turns off the option.
3466 * Note. In this implementation, socket operation overrides default parameter
3467 * being set by sysctl as well as FreeBSD implementation
3469 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3470 unsigned int optlen)
3473 struct sctp_sock *sp = sctp_sk(sk);
3475 if (optlen < sizeof(int))
3477 if (get_user(val, (int __user *)optval))
3479 if (!sctp_is_ep_boundall(sk) && val)
3481 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3484 if (val == 0 && sp->do_auto_asconf) {
3485 list_del(&sp->auto_asconf_list);
3486 sp->do_auto_asconf = 0;
3487 } else if (val && !sp->do_auto_asconf) {
3488 list_add_tail(&sp->auto_asconf_list,
3489 &sock_net(sk)->sctp.auto_asconf_splist);
3490 sp->do_auto_asconf = 1;
3497 * SCTP_PEER_ADDR_THLDS
3499 * This option allows us to alter the partially failed threshold for one or all
3500 * transports in an association. See Section 6.1 of:
3501 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3503 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3504 char __user *optval,
3505 unsigned int optlen)
3507 struct sctp_paddrthlds val;
3508 struct sctp_transport *trans;
3509 struct sctp_association *asoc;
3511 if (optlen < sizeof(struct sctp_paddrthlds))
3513 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3514 sizeof(struct sctp_paddrthlds)))
3518 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3519 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3522 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3524 if (val.spt_pathmaxrxt)
3525 trans->pathmaxrxt = val.spt_pathmaxrxt;
3526 trans->pf_retrans = val.spt_pathpfthld;
3529 if (val.spt_pathmaxrxt)
3530 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3531 asoc->pf_retrans = val.spt_pathpfthld;
3533 trans = sctp_addr_id2transport(sk, &val.spt_address,
3538 if (val.spt_pathmaxrxt)
3539 trans->pathmaxrxt = val.spt_pathmaxrxt;
3540 trans->pf_retrans = val.spt_pathpfthld;
3546 /* API 6.2 setsockopt(), getsockopt()
3548 * Applications use setsockopt() and getsockopt() to set or retrieve
3549 * socket options. Socket options are used to change the default
3550 * behavior of sockets calls. They are described in Section 7.
3554 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3555 * int __user *optlen);
3556 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3559 * sd - the socket descript.
3560 * level - set to IPPROTO_SCTP for all SCTP options.
3561 * optname - the option name.
3562 * optval - the buffer to store the value of the option.
3563 * optlen - the size of the buffer.
3565 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3566 char __user *optval, unsigned int optlen)
3570 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3572 /* I can hardly begin to describe how wrong this is. This is
3573 * so broken as to be worse than useless. The API draft
3574 * REALLY is NOT helpful here... I am not convinced that the
3575 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3576 * are at all well-founded.
3578 if (level != SOL_SCTP) {
3579 struct sctp_af *af = sctp_sk(sk)->pf->af;
3580 retval = af->setsockopt(sk, level, optname, optval, optlen);
3587 case SCTP_SOCKOPT_BINDX_ADD:
3588 /* 'optlen' is the size of the addresses buffer. */
3589 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3590 optlen, SCTP_BINDX_ADD_ADDR);
3593 case SCTP_SOCKOPT_BINDX_REM:
3594 /* 'optlen' is the size of the addresses buffer. */
3595 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3596 optlen, SCTP_BINDX_REM_ADDR);
3599 case SCTP_SOCKOPT_CONNECTX_OLD:
3600 /* 'optlen' is the size of the addresses buffer. */
3601 retval = sctp_setsockopt_connectx_old(sk,
3602 (struct sockaddr __user *)optval,
3606 case SCTP_SOCKOPT_CONNECTX:
3607 /* 'optlen' is the size of the addresses buffer. */
3608 retval = sctp_setsockopt_connectx(sk,
3609 (struct sockaddr __user *)optval,
3613 case SCTP_DISABLE_FRAGMENTS:
3614 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3618 retval = sctp_setsockopt_events(sk, optval, optlen);
3621 case SCTP_AUTOCLOSE:
3622 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3625 case SCTP_PEER_ADDR_PARAMS:
3626 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3629 case SCTP_DELAYED_SACK:
3630 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3632 case SCTP_PARTIAL_DELIVERY_POINT:
3633 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3637 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3639 case SCTP_DEFAULT_SEND_PARAM:
3640 retval = sctp_setsockopt_default_send_param(sk, optval,
3643 case SCTP_PRIMARY_ADDR:
3644 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3646 case SCTP_SET_PEER_PRIMARY_ADDR:
3647 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3650 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3653 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3655 case SCTP_ASSOCINFO:
3656 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3658 case SCTP_I_WANT_MAPPED_V4_ADDR:
3659 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3662 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3664 case SCTP_ADAPTATION_LAYER:
3665 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3668 retval = sctp_setsockopt_context(sk, optval, optlen);
3670 case SCTP_FRAGMENT_INTERLEAVE:
3671 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3673 case SCTP_MAX_BURST:
3674 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3676 case SCTP_AUTH_CHUNK:
3677 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3679 case SCTP_HMAC_IDENT:
3680 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3683 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3685 case SCTP_AUTH_ACTIVE_KEY:
3686 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3688 case SCTP_AUTH_DELETE_KEY:
3689 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3691 case SCTP_AUTO_ASCONF:
3692 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3694 case SCTP_PEER_ADDR_THLDS:
3695 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3698 retval = -ENOPROTOOPT;
3702 sctp_release_sock(sk);
3708 /* API 3.1.6 connect() - UDP Style Syntax
3710 * An application may use the connect() call in the UDP model to initiate an
3711 * association without sending data.
3715 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3717 * sd: the socket descriptor to have a new association added to.
3719 * nam: the address structure (either struct sockaddr_in or struct
3720 * sockaddr_in6 defined in RFC2553 [7]).
3722 * len: the size of the address.
3724 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3732 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
3735 /* Validate addr_len before calling common connect/connectx routine. */
3736 af = sctp_get_af_specific(addr->sa_family);
3737 if (!af || addr_len < af->sockaddr_len) {
3740 /* Pass correct addr len to common routine (so it knows there
3741 * is only one address being passed.
3743 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3746 sctp_release_sock(sk);
3750 /* FIXME: Write comments. */
3751 static int sctp_disconnect(struct sock *sk, int flags)
3753 return -EOPNOTSUPP; /* STUB */
3756 /* 4.1.4 accept() - TCP Style Syntax
3758 * Applications use accept() call to remove an established SCTP
3759 * association from the accept queue of the endpoint. A new socket
3760 * descriptor will be returned from accept() to represent the newly
3761 * formed association.
3763 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3765 struct sctp_sock *sp;
3766 struct sctp_endpoint *ep;
3767 struct sock *newsk = NULL;
3768 struct sctp_association *asoc;
3777 if (!sctp_style(sk, TCP)) {
3778 error = -EOPNOTSUPP;
3782 if (!sctp_sstate(sk, LISTENING)) {
3787 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3789 error = sctp_wait_for_accept(sk, timeo);
3793 /* We treat the list of associations on the endpoint as the accept
3794 * queue and pick the first association on the list.
3796 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3798 newsk = sp->pf->create_accept_sk(sk, asoc);
3804 /* Populate the fields of the newsk from the oldsk and migrate the
3805 * asoc to the newsk.
3807 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3810 sctp_release_sock(sk);
3815 /* The SCTP ioctl handler. */
3816 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3823 * SEQPACKET-style sockets in LISTENING state are valid, for
3824 * SCTP, so only discard TCP-style sockets in LISTENING state.
3826 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3831 struct sk_buff *skb;
3832 unsigned int amount = 0;
3834 skb = skb_peek(&sk->sk_receive_queue);
3837 * We will only return the amount of this packet since
3838 * that is all that will be read.
3842 rc = put_user(amount, (int __user *)arg);
3850 sctp_release_sock(sk);
3854 /* This is the function which gets called during socket creation to
3855 * initialized the SCTP-specific portion of the sock.
3856 * The sock structure should already be zero-filled memory.
3858 static int sctp_init_sock(struct sock *sk)
3860 struct net *net = sock_net(sk);
3861 struct sctp_sock *sp;
3863 pr_debug("%s: sk:%p\n", __func__, sk);
3867 /* Initialize the SCTP per socket area. */
3868 switch (sk->sk_type) {
3869 case SOCK_SEQPACKET:
3870 sp->type = SCTP_SOCKET_UDP;
3873 sp->type = SCTP_SOCKET_TCP;
3876 return -ESOCKTNOSUPPORT;
3879 /* Initialize default send parameters. These parameters can be
3880 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3882 sp->default_stream = 0;
3883 sp->default_ppid = 0;
3884 sp->default_flags = 0;
3885 sp->default_context = 0;
3886 sp->default_timetolive = 0;
3888 sp->default_rcv_context = 0;
3889 sp->max_burst = net->sctp.max_burst;
3891 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
3893 /* Initialize default setup parameters. These parameters
3894 * can be modified with the SCTP_INITMSG socket option or
3895 * overridden by the SCTP_INIT CMSG.
3897 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3898 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3899 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
3900 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
3902 /* Initialize default RTO related parameters. These parameters can
3903 * be modified for with the SCTP_RTOINFO socket option.
3905 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
3906 sp->rtoinfo.srto_max = net->sctp.rto_max;
3907 sp->rtoinfo.srto_min = net->sctp.rto_min;
3909 /* Initialize default association related parameters. These parameters
3910 * can be modified with the SCTP_ASSOCINFO socket option.
3912 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
3913 sp->assocparams.sasoc_number_peer_destinations = 0;
3914 sp->assocparams.sasoc_peer_rwnd = 0;
3915 sp->assocparams.sasoc_local_rwnd = 0;
3916 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
3918 /* Initialize default event subscriptions. By default, all the
3921 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3923 /* Default Peer Address Parameters. These defaults can
3924 * be modified via SCTP_PEER_ADDR_PARAMS
3926 sp->hbinterval = net->sctp.hb_interval;
3927 sp->pathmaxrxt = net->sctp.max_retrans_path;
3928 sp->pathmtu = 0; // allow default discovery
3929 sp->sackdelay = net->sctp.sack_timeout;
3931 sp->param_flags = SPP_HB_ENABLE |
3933 SPP_SACKDELAY_ENABLE;
3935 /* If enabled no SCTP message fragmentation will be performed.
3936 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3938 sp->disable_fragments = 0;
3940 /* Enable Nagle algorithm by default. */
3943 /* Enable by default. */
3946 /* Auto-close idle associations after the configured
3947 * number of seconds. A value of 0 disables this
3948 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3949 * for UDP-style sockets only.
3953 /* User specified fragmentation limit. */
3956 sp->adaptation_ind = 0;
3958 sp->pf = sctp_get_pf_specific(sk->sk_family);
3960 /* Control variables for partial data delivery. */
3961 atomic_set(&sp->pd_mode, 0);
3962 skb_queue_head_init(&sp->pd_lobby);
3963 sp->frag_interleave = 0;
3965 /* Create a per socket endpoint structure. Even if we
3966 * change the data structure relationships, this may still
3967 * be useful for storing pre-connect address information.
3969 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
3975 sk->sk_destruct = sctp_destruct_sock;
3977 SCTP_DBG_OBJCNT_INC(sock);
3980 percpu_counter_inc(&sctp_sockets_allocated);
3981 sock_prot_inuse_add(net, sk->sk_prot, 1);
3982 if (net->sctp.default_auto_asconf) {
3983 list_add_tail(&sp->auto_asconf_list,
3984 &net->sctp.auto_asconf_splist);
3985 sp->do_auto_asconf = 1;
3987 sp->do_auto_asconf = 0;
3993 /* Cleanup any SCTP per socket resources. */
3994 static void sctp_destroy_sock(struct sock *sk)
3996 struct sctp_sock *sp;
3998 pr_debug("%s: sk:%p\n", __func__, sk);
4000 /* Release our hold on the endpoint. */
4002 /* This could happen during socket init, thus we bail out
4003 * early, since the rest of the below is not setup either.
4008 if (sp->do_auto_asconf) {
4009 sp->do_auto_asconf = 0;
4010 list_del(&sp->auto_asconf_list);
4012 sctp_endpoint_free(sp->ep);
4014 percpu_counter_dec(&sctp_sockets_allocated);
4015 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4019 /* Triggered when there are no references on the socket anymore */
4020 static void sctp_destruct_sock(struct sock *sk)
4022 struct sctp_sock *sp = sctp_sk(sk);
4024 /* Free up the HMAC transform. */
4025 crypto_free_hash(sp->hmac);
4027 inet_sock_destruct(sk);
4030 /* API 4.1.7 shutdown() - TCP Style Syntax
4031 * int shutdown(int socket, int how);
4033 * sd - the socket descriptor of the association to be closed.
4034 * how - Specifies the type of shutdown. The values are
4037 * Disables further receive operations. No SCTP
4038 * protocol action is taken.
4040 * Disables further send operations, and initiates
4041 * the SCTP shutdown sequence.
4043 * Disables further send and receive operations
4044 * and initiates the SCTP shutdown sequence.
4046 static void sctp_shutdown(struct sock *sk, int how)
4048 struct net *net = sock_net(sk);
4049 struct sctp_endpoint *ep;
4050 struct sctp_association *asoc;
4052 if (!sctp_style(sk, TCP))
4055 if (how & SEND_SHUTDOWN) {
4056 ep = sctp_sk(sk)->ep;
4057 if (!list_empty(&ep->asocs)) {
4058 asoc = list_entry(ep->asocs.next,
4059 struct sctp_association, asocs);
4060 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4065 /* 7.2.1 Association Status (SCTP_STATUS)
4067 * Applications can retrieve current status information about an
4068 * association, including association state, peer receiver window size,
4069 * number of unacked data chunks, and number of data chunks pending
4070 * receipt. This information is read-only.
4072 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4073 char __user *optval,
4076 struct sctp_status status;
4077 struct sctp_association *asoc = NULL;
4078 struct sctp_transport *transport;
4079 sctp_assoc_t associd;
4082 if (len < sizeof(status)) {
4087 len = sizeof(status);
4088 if (copy_from_user(&status, optval, len)) {
4093 associd = status.sstat_assoc_id;
4094 asoc = sctp_id2assoc(sk, associd);
4100 transport = asoc->peer.primary_path;
4102 status.sstat_assoc_id = sctp_assoc2id(asoc);
4103 status.sstat_state = asoc->state;
4104 status.sstat_rwnd = asoc->peer.rwnd;
4105 status.sstat_unackdata = asoc->unack_data;
4107 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4108 status.sstat_instrms = asoc->c.sinit_max_instreams;
4109 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4110 status.sstat_fragmentation_point = asoc->frag_point;
4111 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4112 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4113 transport->af_specific->sockaddr_len);
4114 /* Map ipv4 address into v4-mapped-on-v6 address. */
4115 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4116 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4117 status.sstat_primary.spinfo_state = transport->state;
4118 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4119 status.sstat_primary.spinfo_srtt = transport->srtt;
4120 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4121 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4123 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4124 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4126 if (put_user(len, optlen)) {
4131 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4132 __func__, len, status.sstat_state, status.sstat_rwnd,
4133 status.sstat_assoc_id);
4135 if (copy_to_user(optval, &status, len)) {
4145 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4147 * Applications can retrieve information about a specific peer address
4148 * of an association, including its reachability state, congestion
4149 * window, and retransmission timer values. This information is
4152 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4153 char __user *optval,
4156 struct sctp_paddrinfo pinfo;
4157 struct sctp_transport *transport;
4160 if (len < sizeof(pinfo)) {
4165 len = sizeof(pinfo);
4166 if (copy_from_user(&pinfo, optval, len)) {
4171 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4172 pinfo.spinfo_assoc_id);
4176 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4177 pinfo.spinfo_state = transport->state;
4178 pinfo.spinfo_cwnd = transport->cwnd;
4179 pinfo.spinfo_srtt = transport->srtt;
4180 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4181 pinfo.spinfo_mtu = transport->pathmtu;
4183 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4184 pinfo.spinfo_state = SCTP_ACTIVE;
4186 if (put_user(len, optlen)) {
4191 if (copy_to_user(optval, &pinfo, len)) {
4200 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4202 * This option is a on/off flag. If enabled no SCTP message
4203 * fragmentation will be performed. Instead if a message being sent
4204 * exceeds the current PMTU size, the message will NOT be sent and
4205 * instead a error will be indicated to the user.
4207 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4208 char __user *optval, int __user *optlen)
4212 if (len < sizeof(int))
4216 val = (sctp_sk(sk)->disable_fragments == 1);
4217 if (put_user(len, optlen))
4219 if (copy_to_user(optval, &val, len))
4224 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4226 * This socket option is used to specify various notifications and
4227 * ancillary data the user wishes to receive.
4229 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4234 if (len > sizeof(struct sctp_event_subscribe))
4235 len = sizeof(struct sctp_event_subscribe);
4236 if (put_user(len, optlen))
4238 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4243 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4245 * This socket option is applicable to the UDP-style socket only. When
4246 * set it will cause associations that are idle for more than the
4247 * specified number of seconds to automatically close. An association
4248 * being idle is defined an association that has NOT sent or received
4249 * user data. The special value of '0' indicates that no automatic
4250 * close of any associations should be performed. The option expects an
4251 * integer defining the number of seconds of idle time before an
4252 * association is closed.
4254 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4256 /* Applicable to UDP-style socket only */
4257 if (sctp_style(sk, TCP))
4259 if (len < sizeof(int))
4262 if (put_user(len, optlen))
4264 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4269 /* Helper routine to branch off an association to a new socket. */
4270 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4272 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4273 struct socket *sock;
4280 /* An association cannot be branched off from an already peeled-off
4281 * socket, nor is this supported for tcp style sockets.
4283 if (!sctp_style(sk, UDP))
4286 /* Create a new socket. */
4287 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4291 sctp_copy_sock(sock->sk, sk, asoc);
4293 /* Make peeled-off sockets more like 1-1 accepted sockets.
4294 * Set the daddr and initialize id to something more random
4296 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4297 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4299 /* Populate the fields of the newsk from the oldsk and migrate the
4300 * asoc to the newsk.
4302 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4308 EXPORT_SYMBOL(sctp_do_peeloff);
4310 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4312 sctp_peeloff_arg_t peeloff;
4313 struct socket *newsock;
4314 struct file *newfile;
4317 if (len < sizeof(sctp_peeloff_arg_t))
4319 len = sizeof(sctp_peeloff_arg_t);
4320 if (copy_from_user(&peeloff, optval, len))
4323 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4327 /* Map the socket to an unused fd that can be returned to the user. */
4328 retval = get_unused_fd_flags(0);
4330 sock_release(newsock);
4334 newfile = sock_alloc_file(newsock, 0, NULL);
4335 if (unlikely(IS_ERR(newfile))) {
4336 put_unused_fd(retval);
4337 sock_release(newsock);
4338 return PTR_ERR(newfile);
4341 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4344 /* Return the fd mapped to the new socket. */
4345 if (put_user(len, optlen)) {
4347 put_unused_fd(retval);
4350 peeloff.sd = retval;
4351 if (copy_to_user(optval, &peeloff, len)) {
4353 put_unused_fd(retval);
4356 fd_install(retval, newfile);
4361 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4363 * Applications can enable or disable heartbeats for any peer address of
4364 * an association, modify an address's heartbeat interval, force a
4365 * heartbeat to be sent immediately, and adjust the address's maximum
4366 * number of retransmissions sent before an address is considered
4367 * unreachable. The following structure is used to access and modify an
4368 * address's parameters:
4370 * struct sctp_paddrparams {
4371 * sctp_assoc_t spp_assoc_id;
4372 * struct sockaddr_storage spp_address;
4373 * uint32_t spp_hbinterval;
4374 * uint16_t spp_pathmaxrxt;
4375 * uint32_t spp_pathmtu;
4376 * uint32_t spp_sackdelay;
4377 * uint32_t spp_flags;
4380 * spp_assoc_id - (one-to-many style socket) This is filled in the
4381 * application, and identifies the association for
4383 * spp_address - This specifies which address is of interest.
4384 * spp_hbinterval - This contains the value of the heartbeat interval,
4385 * in milliseconds. If a value of zero
4386 * is present in this field then no changes are to
4387 * be made to this parameter.
4388 * spp_pathmaxrxt - This contains the maximum number of
4389 * retransmissions before this address shall be
4390 * considered unreachable. If a value of zero
4391 * is present in this field then no changes are to
4392 * be made to this parameter.
4393 * spp_pathmtu - When Path MTU discovery is disabled the value
4394 * specified here will be the "fixed" path mtu.
4395 * Note that if the spp_address field is empty
4396 * then all associations on this address will
4397 * have this fixed path mtu set upon them.
4399 * spp_sackdelay - When delayed sack is enabled, this value specifies
4400 * the number of milliseconds that sacks will be delayed
4401 * for. This value will apply to all addresses of an
4402 * association if the spp_address field is empty. Note
4403 * also, that if delayed sack is enabled and this
4404 * value is set to 0, no change is made to the last
4405 * recorded delayed sack timer value.
4407 * spp_flags - These flags are used to control various features
4408 * on an association. The flag field may contain
4409 * zero or more of the following options.
4411 * SPP_HB_ENABLE - Enable heartbeats on the
4412 * specified address. Note that if the address
4413 * field is empty all addresses for the association
4414 * have heartbeats enabled upon them.
4416 * SPP_HB_DISABLE - Disable heartbeats on the
4417 * speicifed address. Note that if the address
4418 * field is empty all addresses for the association
4419 * will have their heartbeats disabled. Note also
4420 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4421 * mutually exclusive, only one of these two should
4422 * be specified. Enabling both fields will have
4423 * undetermined results.
4425 * SPP_HB_DEMAND - Request a user initiated heartbeat
4426 * to be made immediately.
4428 * SPP_PMTUD_ENABLE - This field will enable PMTU
4429 * discovery upon the specified address. Note that
4430 * if the address feild is empty then all addresses
4431 * on the association are effected.
4433 * SPP_PMTUD_DISABLE - This field will disable PMTU
4434 * discovery upon the specified address. Note that
4435 * if the address feild is empty then all addresses
4436 * on the association are effected. Not also that
4437 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4438 * exclusive. Enabling both will have undetermined
4441 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4442 * on delayed sack. The time specified in spp_sackdelay
4443 * is used to specify the sack delay for this address. Note
4444 * that if spp_address is empty then all addresses will
4445 * enable delayed sack and take on the sack delay
4446 * value specified in spp_sackdelay.
4447 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4448 * off delayed sack. If the spp_address field is blank then
4449 * delayed sack is disabled for the entire association. Note
4450 * also that this field is mutually exclusive to
4451 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4454 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4455 char __user *optval, int __user *optlen)
4457 struct sctp_paddrparams params;
4458 struct sctp_transport *trans = NULL;
4459 struct sctp_association *asoc = NULL;
4460 struct sctp_sock *sp = sctp_sk(sk);
4462 if (len < sizeof(struct sctp_paddrparams))
4464 len = sizeof(struct sctp_paddrparams);
4465 if (copy_from_user(¶ms, optval, len))
4468 /* If an address other than INADDR_ANY is specified, and
4469 * no transport is found, then the request is invalid.
4471 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
4472 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4473 params.spp_assoc_id);
4475 pr_debug("%s: failed no transport\n", __func__);
4480 /* Get association, if assoc_id != 0 and the socket is a one
4481 * to many style socket, and an association was not found, then
4482 * the id was invalid.
4484 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4485 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4486 pr_debug("%s: failed no association\n", __func__);
4491 /* Fetch transport values. */
4492 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4493 params.spp_pathmtu = trans->pathmtu;
4494 params.spp_pathmaxrxt = trans->pathmaxrxt;
4495 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4497 /*draft-11 doesn't say what to return in spp_flags*/
4498 params.spp_flags = trans->param_flags;
4500 /* Fetch association values. */
4501 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4502 params.spp_pathmtu = asoc->pathmtu;
4503 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4504 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4506 /*draft-11 doesn't say what to return in spp_flags*/
4507 params.spp_flags = asoc->param_flags;
4509 /* Fetch socket values. */
4510 params.spp_hbinterval = sp->hbinterval;
4511 params.spp_pathmtu = sp->pathmtu;
4512 params.spp_sackdelay = sp->sackdelay;
4513 params.spp_pathmaxrxt = sp->pathmaxrxt;
4515 /*draft-11 doesn't say what to return in spp_flags*/
4516 params.spp_flags = sp->param_flags;
4519 if (copy_to_user(optval, ¶ms, len))
4522 if (put_user(len, optlen))
4529 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4531 * This option will effect the way delayed acks are performed. This
4532 * option allows you to get or set the delayed ack time, in
4533 * milliseconds. It also allows changing the delayed ack frequency.
4534 * Changing the frequency to 1 disables the delayed sack algorithm. If
4535 * the assoc_id is 0, then this sets or gets the endpoints default
4536 * values. If the assoc_id field is non-zero, then the set or get
4537 * effects the specified association for the one to many model (the
4538 * assoc_id field is ignored by the one to one model). Note that if
4539 * sack_delay or sack_freq are 0 when setting this option, then the
4540 * current values will remain unchanged.
4542 * struct sctp_sack_info {
4543 * sctp_assoc_t sack_assoc_id;
4544 * uint32_t sack_delay;
4545 * uint32_t sack_freq;
4548 * sack_assoc_id - This parameter, indicates which association the user
4549 * is performing an action upon. Note that if this field's value is
4550 * zero then the endpoints default value is changed (effecting future
4551 * associations only).
4553 * sack_delay - This parameter contains the number of milliseconds that
4554 * the user is requesting the delayed ACK timer be set to. Note that
4555 * this value is defined in the standard to be between 200 and 500
4558 * sack_freq - This parameter contains the number of packets that must
4559 * be received before a sack is sent without waiting for the delay
4560 * timer to expire. The default value for this is 2, setting this
4561 * value to 1 will disable the delayed sack algorithm.
4563 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4564 char __user *optval,
4567 struct sctp_sack_info params;
4568 struct sctp_association *asoc = NULL;
4569 struct sctp_sock *sp = sctp_sk(sk);
4571 if (len >= sizeof(struct sctp_sack_info)) {
4572 len = sizeof(struct sctp_sack_info);
4574 if (copy_from_user(¶ms, optval, len))
4576 } else if (len == sizeof(struct sctp_assoc_value)) {
4577 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4578 pr_warn("Use struct sctp_sack_info instead\n");
4579 if (copy_from_user(¶ms, optval, len))
4584 /* Get association, if sack_assoc_id != 0 and the socket is a one
4585 * to many style socket, and an association was not found, then
4586 * the id was invalid.
4588 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4589 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4593 /* Fetch association values. */
4594 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4595 params.sack_delay = jiffies_to_msecs(
4597 params.sack_freq = asoc->sackfreq;
4600 params.sack_delay = 0;
4601 params.sack_freq = 1;
4604 /* Fetch socket values. */
4605 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4606 params.sack_delay = sp->sackdelay;
4607 params.sack_freq = sp->sackfreq;
4609 params.sack_delay = 0;
4610 params.sack_freq = 1;
4614 if (copy_to_user(optval, ¶ms, len))
4617 if (put_user(len, optlen))
4623 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4625 * Applications can specify protocol parameters for the default association
4626 * initialization. The option name argument to setsockopt() and getsockopt()
4629 * Setting initialization parameters is effective only on an unconnected
4630 * socket (for UDP-style sockets only future associations are effected
4631 * by the change). With TCP-style sockets, this option is inherited by
4632 * sockets derived from a listener socket.
4634 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4636 if (len < sizeof(struct sctp_initmsg))
4638 len = sizeof(struct sctp_initmsg);
4639 if (put_user(len, optlen))
4641 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4647 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4648 char __user *optval, int __user *optlen)
4650 struct sctp_association *asoc;
4652 struct sctp_getaddrs getaddrs;
4653 struct sctp_transport *from;
4655 union sctp_addr temp;
4656 struct sctp_sock *sp = sctp_sk(sk);
4661 if (len < sizeof(struct sctp_getaddrs))
4664 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4667 /* For UDP-style sockets, id specifies the association to query. */
4668 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4672 to = optval + offsetof(struct sctp_getaddrs,addrs);
4673 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4675 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4677 memcpy(&temp, &from->ipaddr, sizeof(temp));
4678 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4679 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4680 if (space_left < addrlen)
4682 if (copy_to_user(to, &temp, addrlen))
4686 space_left -= addrlen;
4689 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4691 bytes_copied = ((char __user *)to) - optval;
4692 if (put_user(bytes_copied, optlen))
4698 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4699 size_t space_left, int *bytes_copied)
4701 struct sctp_sockaddr_entry *addr;
4702 union sctp_addr temp;
4705 struct net *net = sock_net(sk);
4708 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4712 if ((PF_INET == sk->sk_family) &&
4713 (AF_INET6 == addr->a.sa.sa_family))
4715 if ((PF_INET6 == sk->sk_family) &&
4716 inet_v6_ipv6only(sk) &&
4717 (AF_INET == addr->a.sa.sa_family))
4719 memcpy(&temp, &addr->a, sizeof(temp));
4720 if (!temp.v4.sin_port)
4721 temp.v4.sin_port = htons(port);
4723 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4725 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4726 if (space_left < addrlen) {
4730 memcpy(to, &temp, addrlen);
4734 space_left -= addrlen;
4735 *bytes_copied += addrlen;
4743 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4744 char __user *optval, int __user *optlen)
4746 struct sctp_bind_addr *bp;
4747 struct sctp_association *asoc;
4749 struct sctp_getaddrs getaddrs;
4750 struct sctp_sockaddr_entry *addr;
4752 union sctp_addr temp;
4753 struct sctp_sock *sp = sctp_sk(sk);
4757 int bytes_copied = 0;
4761 if (len < sizeof(struct sctp_getaddrs))
4764 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4768 * For UDP-style sockets, id specifies the association to query.
4769 * If the id field is set to the value '0' then the locally bound
4770 * addresses are returned without regard to any particular
4773 if (0 == getaddrs.assoc_id) {
4774 bp = &sctp_sk(sk)->ep->base.bind_addr;
4776 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4779 bp = &asoc->base.bind_addr;
4782 to = optval + offsetof(struct sctp_getaddrs,addrs);
4783 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4785 addrs = kmalloc(space_left, GFP_KERNEL);
4789 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4790 * addresses from the global local address list.
4792 if (sctp_list_single_entry(&bp->address_list)) {
4793 addr = list_entry(bp->address_list.next,
4794 struct sctp_sockaddr_entry, list);
4795 if (sctp_is_any(sk, &addr->a)) {
4796 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4797 space_left, &bytes_copied);
4807 /* Protection on the bound address list is not needed since
4808 * in the socket option context we hold a socket lock and
4809 * thus the bound address list can't change.
4811 list_for_each_entry(addr, &bp->address_list, list) {
4812 memcpy(&temp, &addr->a, sizeof(temp));
4813 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4814 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4815 if (space_left < addrlen) {
4816 err = -ENOMEM; /*fixme: right error?*/
4819 memcpy(buf, &temp, addrlen);
4821 bytes_copied += addrlen;
4823 space_left -= addrlen;
4827 if (copy_to_user(to, addrs, bytes_copied)) {
4831 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4835 if (put_user(bytes_copied, optlen))
4842 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4844 * Requests that the local SCTP stack use the enclosed peer address as
4845 * the association primary. The enclosed address must be one of the
4846 * association peer's addresses.
4848 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4849 char __user *optval, int __user *optlen)
4851 struct sctp_prim prim;
4852 struct sctp_association *asoc;
4853 struct sctp_sock *sp = sctp_sk(sk);
4855 if (len < sizeof(struct sctp_prim))
4858 len = sizeof(struct sctp_prim);
4860 if (copy_from_user(&prim, optval, len))
4863 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4867 if (!asoc->peer.primary_path)
4870 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4871 asoc->peer.primary_path->af_specific->sockaddr_len);
4873 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4874 (union sctp_addr *)&prim.ssp_addr);
4876 if (put_user(len, optlen))
4878 if (copy_to_user(optval, &prim, len))
4885 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4887 * Requests that the local endpoint set the specified Adaptation Layer
4888 * Indication parameter for all future INIT and INIT-ACK exchanges.
4890 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4891 char __user *optval, int __user *optlen)
4893 struct sctp_setadaptation adaptation;
4895 if (len < sizeof(struct sctp_setadaptation))
4898 len = sizeof(struct sctp_setadaptation);
4900 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4902 if (put_user(len, optlen))
4904 if (copy_to_user(optval, &adaptation, len))
4912 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4914 * Applications that wish to use the sendto() system call may wish to
4915 * specify a default set of parameters that would normally be supplied
4916 * through the inclusion of ancillary data. This socket option allows
4917 * such an application to set the default sctp_sndrcvinfo structure.
4920 * The application that wishes to use this socket option simply passes
4921 * in to this call the sctp_sndrcvinfo structure defined in Section
4922 * 5.2.2) The input parameters accepted by this call include
4923 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4924 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4925 * to this call if the caller is using the UDP model.
4927 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4929 static int sctp_getsockopt_default_send_param(struct sock *sk,
4930 int len, char __user *optval,
4933 struct sctp_sndrcvinfo info;
4934 struct sctp_association *asoc;
4935 struct sctp_sock *sp = sctp_sk(sk);
4937 if (len < sizeof(struct sctp_sndrcvinfo))
4940 len = sizeof(struct sctp_sndrcvinfo);
4942 if (copy_from_user(&info, optval, len))
4945 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4946 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4950 info.sinfo_stream = asoc->default_stream;
4951 info.sinfo_flags = asoc->default_flags;
4952 info.sinfo_ppid = asoc->default_ppid;
4953 info.sinfo_context = asoc->default_context;
4954 info.sinfo_timetolive = asoc->default_timetolive;
4956 info.sinfo_stream = sp->default_stream;
4957 info.sinfo_flags = sp->default_flags;
4958 info.sinfo_ppid = sp->default_ppid;
4959 info.sinfo_context = sp->default_context;
4960 info.sinfo_timetolive = sp->default_timetolive;
4963 if (put_user(len, optlen))
4965 if (copy_to_user(optval, &info, len))
4973 * 7.1.5 SCTP_NODELAY
4975 * Turn on/off any Nagle-like algorithm. This means that packets are
4976 * generally sent as soon as possible and no unnecessary delays are
4977 * introduced, at the cost of more packets in the network. Expects an
4978 * integer boolean flag.
4981 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4982 char __user *optval, int __user *optlen)
4986 if (len < sizeof(int))
4990 val = (sctp_sk(sk)->nodelay == 1);
4991 if (put_user(len, optlen))
4993 if (copy_to_user(optval, &val, len))
5000 * 7.1.1 SCTP_RTOINFO
5002 * The protocol parameters used to initialize and bound retransmission
5003 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5004 * and modify these parameters.
5005 * All parameters are time values, in milliseconds. A value of 0, when
5006 * modifying the parameters, indicates that the current value should not
5010 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5011 char __user *optval,
5012 int __user *optlen) {
5013 struct sctp_rtoinfo rtoinfo;
5014 struct sctp_association *asoc;
5016 if (len < sizeof (struct sctp_rtoinfo))
5019 len = sizeof(struct sctp_rtoinfo);
5021 if (copy_from_user(&rtoinfo, optval, len))
5024 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5026 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5029 /* Values corresponding to the specific association. */
5031 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5032 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5033 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5035 /* Values corresponding to the endpoint. */
5036 struct sctp_sock *sp = sctp_sk(sk);
5038 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5039 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5040 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5043 if (put_user(len, optlen))
5046 if (copy_to_user(optval, &rtoinfo, len))
5054 * 7.1.2 SCTP_ASSOCINFO
5056 * This option is used to tune the maximum retransmission attempts
5057 * of the association.
5058 * Returns an error if the new association retransmission value is
5059 * greater than the sum of the retransmission value of the peer.
5060 * See [SCTP] for more information.
5063 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5064 char __user *optval,
5068 struct sctp_assocparams assocparams;
5069 struct sctp_association *asoc;
5070 struct list_head *pos;
5073 if (len < sizeof (struct sctp_assocparams))
5076 len = sizeof(struct sctp_assocparams);
5078 if (copy_from_user(&assocparams, optval, len))
5081 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5083 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5086 /* Values correspoinding to the specific association */
5088 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5089 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5090 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5091 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5093 list_for_each(pos, &asoc->peer.transport_addr_list) {
5097 assocparams.sasoc_number_peer_destinations = cnt;
5099 /* Values corresponding to the endpoint */
5100 struct sctp_sock *sp = sctp_sk(sk);
5102 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5103 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5104 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5105 assocparams.sasoc_cookie_life =
5106 sp->assocparams.sasoc_cookie_life;
5107 assocparams.sasoc_number_peer_destinations =
5109 sasoc_number_peer_destinations;
5112 if (put_user(len, optlen))
5115 if (copy_to_user(optval, &assocparams, len))
5122 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5124 * This socket option is a boolean flag which turns on or off mapped V4
5125 * addresses. If this option is turned on and the socket is type
5126 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5127 * If this option is turned off, then no mapping will be done of V4
5128 * addresses and a user will receive both PF_INET6 and PF_INET type
5129 * addresses on the socket.
5131 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5132 char __user *optval, int __user *optlen)
5135 struct sctp_sock *sp = sctp_sk(sk);
5137 if (len < sizeof(int))
5142 if (put_user(len, optlen))
5144 if (copy_to_user(optval, &val, len))
5151 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5152 * (chapter and verse is quoted at sctp_setsockopt_context())
5154 static int sctp_getsockopt_context(struct sock *sk, int len,
5155 char __user *optval, int __user *optlen)
5157 struct sctp_assoc_value params;
5158 struct sctp_sock *sp;
5159 struct sctp_association *asoc;
5161 if (len < sizeof(struct sctp_assoc_value))
5164 len = sizeof(struct sctp_assoc_value);
5166 if (copy_from_user(¶ms, optval, len))
5171 if (params.assoc_id != 0) {
5172 asoc = sctp_id2assoc(sk, params.assoc_id);
5175 params.assoc_value = asoc->default_rcv_context;
5177 params.assoc_value = sp->default_rcv_context;
5180 if (put_user(len, optlen))
5182 if (copy_to_user(optval, ¶ms, len))
5189 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5190 * This option will get or set the maximum size to put in any outgoing
5191 * SCTP DATA chunk. If a message is larger than this size it will be
5192 * fragmented by SCTP into the specified size. Note that the underlying
5193 * SCTP implementation may fragment into smaller sized chunks when the
5194 * PMTU of the underlying association is smaller than the value set by
5195 * the user. The default value for this option is '0' which indicates
5196 * the user is NOT limiting fragmentation and only the PMTU will effect
5197 * SCTP's choice of DATA chunk size. Note also that values set larger
5198 * than the maximum size of an IP datagram will effectively let SCTP
5199 * control fragmentation (i.e. the same as setting this option to 0).
5201 * The following structure is used to access and modify this parameter:
5203 * struct sctp_assoc_value {
5204 * sctp_assoc_t assoc_id;
5205 * uint32_t assoc_value;
5208 * assoc_id: This parameter is ignored for one-to-one style sockets.
5209 * For one-to-many style sockets this parameter indicates which
5210 * association the user is performing an action upon. Note that if
5211 * this field's value is zero then the endpoints default value is
5212 * changed (effecting future associations only).
5213 * assoc_value: This parameter specifies the maximum size in bytes.
5215 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5216 char __user *optval, int __user *optlen)
5218 struct sctp_assoc_value params;
5219 struct sctp_association *asoc;
5221 if (len == sizeof(int)) {
5222 pr_warn("Use of int in maxseg socket option deprecated\n");
5223 pr_warn("Use struct sctp_assoc_value instead\n");
5224 params.assoc_id = 0;
5225 } else if (len >= sizeof(struct sctp_assoc_value)) {
5226 len = sizeof(struct sctp_assoc_value);
5227 if (copy_from_user(¶ms, optval, sizeof(params)))
5232 asoc = sctp_id2assoc(sk, params.assoc_id);
5233 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5237 params.assoc_value = asoc->frag_point;
5239 params.assoc_value = sctp_sk(sk)->user_frag;
5241 if (put_user(len, optlen))
5243 if (len == sizeof(int)) {
5244 if (copy_to_user(optval, ¶ms.assoc_value, len))
5247 if (copy_to_user(optval, ¶ms, len))
5255 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5256 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5258 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5259 char __user *optval, int __user *optlen)
5263 if (len < sizeof(int))
5268 val = sctp_sk(sk)->frag_interleave;
5269 if (put_user(len, optlen))
5271 if (copy_to_user(optval, &val, len))
5278 * 7.1.25. Set or Get the sctp partial delivery point
5279 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5281 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5282 char __user *optval,
5287 if (len < sizeof(u32))
5292 val = sctp_sk(sk)->pd_point;
5293 if (put_user(len, optlen))
5295 if (copy_to_user(optval, &val, len))
5302 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5303 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5305 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5306 char __user *optval,
5309 struct sctp_assoc_value params;
5310 struct sctp_sock *sp;
5311 struct sctp_association *asoc;
5313 if (len == sizeof(int)) {
5314 pr_warn("Use of int in max_burst socket option deprecated\n");
5315 pr_warn("Use struct sctp_assoc_value instead\n");
5316 params.assoc_id = 0;
5317 } else if (len >= sizeof(struct sctp_assoc_value)) {
5318 len = sizeof(struct sctp_assoc_value);
5319 if (copy_from_user(¶ms, optval, len))
5326 if (params.assoc_id != 0) {
5327 asoc = sctp_id2assoc(sk, params.assoc_id);
5330 params.assoc_value = asoc->max_burst;
5332 params.assoc_value = sp->max_burst;
5334 if (len == sizeof(int)) {
5335 if (copy_to_user(optval, ¶ms.assoc_value, len))
5338 if (copy_to_user(optval, ¶ms, len))
5346 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5347 char __user *optval, int __user *optlen)
5349 struct net *net = sock_net(sk);
5350 struct sctp_hmacalgo __user *p = (void __user *)optval;
5351 struct sctp_hmac_algo_param *hmacs;
5355 if (!net->sctp.auth_enable)
5358 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5359 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5361 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5364 len = sizeof(struct sctp_hmacalgo) + data_len;
5365 num_idents = data_len / sizeof(u16);
5367 if (put_user(len, optlen))
5369 if (put_user(num_idents, &p->shmac_num_idents))
5371 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5376 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5377 char __user *optval, int __user *optlen)
5379 struct net *net = sock_net(sk);
5380 struct sctp_authkeyid val;
5381 struct sctp_association *asoc;
5383 if (!net->sctp.auth_enable)
5386 if (len < sizeof(struct sctp_authkeyid))
5388 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5391 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5392 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5396 val.scact_keynumber = asoc->active_key_id;
5398 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5400 len = sizeof(struct sctp_authkeyid);
5401 if (put_user(len, optlen))
5403 if (copy_to_user(optval, &val, len))
5409 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5410 char __user *optval, int __user *optlen)
5412 struct net *net = sock_net(sk);
5413 struct sctp_authchunks __user *p = (void __user *)optval;
5414 struct sctp_authchunks val;
5415 struct sctp_association *asoc;
5416 struct sctp_chunks_param *ch;
5420 if (!net->sctp.auth_enable)
5423 if (len < sizeof(struct sctp_authchunks))
5426 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5429 to = p->gauth_chunks;
5430 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5434 ch = asoc->peer.peer_chunks;
5438 /* See if the user provided enough room for all the data */
5439 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5440 if (len < num_chunks)
5443 if (copy_to_user(to, ch->chunks, num_chunks))
5446 len = sizeof(struct sctp_authchunks) + num_chunks;
5447 if (put_user(len, optlen)) return -EFAULT;
5448 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5453 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5454 char __user *optval, int __user *optlen)
5456 struct net *net = sock_net(sk);
5457 struct sctp_authchunks __user *p = (void __user *)optval;
5458 struct sctp_authchunks val;
5459 struct sctp_association *asoc;
5460 struct sctp_chunks_param *ch;
5464 if (!net->sctp.auth_enable)
5467 if (len < sizeof(struct sctp_authchunks))
5470 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5473 to = p->gauth_chunks;
5474 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5475 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5479 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5481 ch = sctp_sk(sk)->ep->auth_chunk_list;
5486 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5487 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5490 if (copy_to_user(to, ch->chunks, num_chunks))
5493 len = sizeof(struct sctp_authchunks) + num_chunks;
5494 if (put_user(len, optlen))
5496 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5503 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5504 * This option gets the current number of associations that are attached
5505 * to a one-to-many style socket. The option value is an uint32_t.
5507 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5508 char __user *optval, int __user *optlen)
5510 struct sctp_sock *sp = sctp_sk(sk);
5511 struct sctp_association *asoc;
5514 if (sctp_style(sk, TCP))
5517 if (len < sizeof(u32))
5522 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5526 if (put_user(len, optlen))
5528 if (copy_to_user(optval, &val, len))
5535 * 8.1.23 SCTP_AUTO_ASCONF
5536 * See the corresponding setsockopt entry as description
5538 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5539 char __user *optval, int __user *optlen)
5543 if (len < sizeof(int))
5547 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5549 if (put_user(len, optlen))
5551 if (copy_to_user(optval, &val, len))
5557 * 8.2.6. Get the Current Identifiers of Associations
5558 * (SCTP_GET_ASSOC_ID_LIST)
5560 * This option gets the current list of SCTP association identifiers of
5561 * the SCTP associations handled by a one-to-many style socket.
5563 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5564 char __user *optval, int __user *optlen)
5566 struct sctp_sock *sp = sctp_sk(sk);
5567 struct sctp_association *asoc;
5568 struct sctp_assoc_ids *ids;
5571 if (sctp_style(sk, TCP))
5574 if (len < sizeof(struct sctp_assoc_ids))
5577 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5581 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5584 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5586 ids = kmalloc(len, GFP_KERNEL);
5590 ids->gaids_number_of_ids = num;
5592 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5593 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5596 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5606 * SCTP_PEER_ADDR_THLDS
5608 * This option allows us to fetch the partially failed threshold for one or all
5609 * transports in an association. See Section 6.1 of:
5610 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5612 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5613 char __user *optval,
5617 struct sctp_paddrthlds val;
5618 struct sctp_transport *trans;
5619 struct sctp_association *asoc;
5621 if (len < sizeof(struct sctp_paddrthlds))
5623 len = sizeof(struct sctp_paddrthlds);
5624 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5627 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5628 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5632 val.spt_pathpfthld = asoc->pf_retrans;
5633 val.spt_pathmaxrxt = asoc->pathmaxrxt;
5635 trans = sctp_addr_id2transport(sk, &val.spt_address,
5640 val.spt_pathmaxrxt = trans->pathmaxrxt;
5641 val.spt_pathpfthld = trans->pf_retrans;
5644 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5651 * SCTP_GET_ASSOC_STATS
5653 * This option retrieves local per endpoint statistics. It is modeled
5654 * after OpenSolaris' implementation
5656 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
5657 char __user *optval,
5660 struct sctp_assoc_stats sas;
5661 struct sctp_association *asoc = NULL;
5663 /* User must provide at least the assoc id */
5664 if (len < sizeof(sctp_assoc_t))
5667 /* Allow the struct to grow and fill in as much as possible */
5668 len = min_t(size_t, len, sizeof(sas));
5670 if (copy_from_user(&sas, optval, len))
5673 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
5677 sas.sas_rtxchunks = asoc->stats.rtxchunks;
5678 sas.sas_gapcnt = asoc->stats.gapcnt;
5679 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
5680 sas.sas_osacks = asoc->stats.osacks;
5681 sas.sas_isacks = asoc->stats.isacks;
5682 sas.sas_octrlchunks = asoc->stats.octrlchunks;
5683 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
5684 sas.sas_oodchunks = asoc->stats.oodchunks;
5685 sas.sas_iodchunks = asoc->stats.iodchunks;
5686 sas.sas_ouodchunks = asoc->stats.ouodchunks;
5687 sas.sas_iuodchunks = asoc->stats.iuodchunks;
5688 sas.sas_idupchunks = asoc->stats.idupchunks;
5689 sas.sas_opackets = asoc->stats.opackets;
5690 sas.sas_ipackets = asoc->stats.ipackets;
5692 /* New high max rto observed, will return 0 if not a single
5693 * RTO update took place. obs_rto_ipaddr will be bogus
5696 sas.sas_maxrto = asoc->stats.max_obs_rto;
5697 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
5698 sizeof(struct sockaddr_storage));
5700 /* Mark beginning of a new observation period */
5701 asoc->stats.max_obs_rto = asoc->rto_min;
5703 if (put_user(len, optlen))
5706 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
5708 if (copy_to_user(optval, &sas, len))
5714 static int sctp_getsockopt(struct sock *sk, int level, int optname,
5715 char __user *optval, int __user *optlen)
5720 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
5722 /* I can hardly begin to describe how wrong this is. This is
5723 * so broken as to be worse than useless. The API draft
5724 * REALLY is NOT helpful here... I am not convinced that the
5725 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5726 * are at all well-founded.
5728 if (level != SOL_SCTP) {
5729 struct sctp_af *af = sctp_sk(sk)->pf->af;
5731 retval = af->getsockopt(sk, level, optname, optval, optlen);
5735 if (get_user(len, optlen))
5742 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5744 case SCTP_DISABLE_FRAGMENTS:
5745 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5749 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5751 case SCTP_AUTOCLOSE:
5752 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5754 case SCTP_SOCKOPT_PEELOFF:
5755 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5757 case SCTP_PEER_ADDR_PARAMS:
5758 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5761 case SCTP_DELAYED_SACK:
5762 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5766 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5768 case SCTP_GET_PEER_ADDRS:
5769 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5772 case SCTP_GET_LOCAL_ADDRS:
5773 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5776 case SCTP_SOCKOPT_CONNECTX3:
5777 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5779 case SCTP_DEFAULT_SEND_PARAM:
5780 retval = sctp_getsockopt_default_send_param(sk, len,
5783 case SCTP_PRIMARY_ADDR:
5784 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5787 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5790 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5792 case SCTP_ASSOCINFO:
5793 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5795 case SCTP_I_WANT_MAPPED_V4_ADDR:
5796 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5799 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5801 case SCTP_GET_PEER_ADDR_INFO:
5802 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5805 case SCTP_ADAPTATION_LAYER:
5806 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5810 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5812 case SCTP_FRAGMENT_INTERLEAVE:
5813 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5816 case SCTP_PARTIAL_DELIVERY_POINT:
5817 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5820 case SCTP_MAX_BURST:
5821 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5824 case SCTP_AUTH_CHUNK:
5825 case SCTP_AUTH_DELETE_KEY:
5826 retval = -EOPNOTSUPP;
5828 case SCTP_HMAC_IDENT:
5829 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5831 case SCTP_AUTH_ACTIVE_KEY:
5832 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5834 case SCTP_PEER_AUTH_CHUNKS:
5835 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5838 case SCTP_LOCAL_AUTH_CHUNKS:
5839 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5842 case SCTP_GET_ASSOC_NUMBER:
5843 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5845 case SCTP_GET_ASSOC_ID_LIST:
5846 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5848 case SCTP_AUTO_ASCONF:
5849 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5851 case SCTP_PEER_ADDR_THLDS:
5852 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
5854 case SCTP_GET_ASSOC_STATS:
5855 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
5858 retval = -ENOPROTOOPT;
5862 sctp_release_sock(sk);
5866 static void sctp_hash(struct sock *sk)
5871 static void sctp_unhash(struct sock *sk)
5876 /* Check if port is acceptable. Possibly find first available port.
5878 * The port hash table (contained in the 'global' SCTP protocol storage
5879 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5880 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5881 * list (the list number is the port number hashed out, so as you
5882 * would expect from a hash function, all the ports in a given list have
5883 * such a number that hashes out to the same list number; you were
5884 * expecting that, right?); so each list has a set of ports, with a
5885 * link to the socket (struct sock) that uses it, the port number and
5886 * a fastreuse flag (FIXME: NPI ipg).
5888 static struct sctp_bind_bucket *sctp_bucket_create(
5889 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
5891 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5893 struct sctp_bind_hashbucket *head; /* hash list */
5894 struct sctp_bind_bucket *pp;
5895 unsigned short snum;
5898 snum = ntohs(addr->v4.sin_port);
5900 pr_debug("%s: begins, snum:%d\n", __func__, snum);
5902 sctp_local_bh_disable();
5905 /* Search for an available port. */
5906 int low, high, remaining, index;
5909 inet_get_local_port_range(sock_net(sk), &low, &high);
5910 remaining = (high - low) + 1;
5911 rover = net_random() % remaining + low;
5915 if ((rover < low) || (rover > high))
5917 if (inet_is_reserved_local_port(rover))
5919 index = sctp_phashfn(sock_net(sk), rover);
5920 head = &sctp_port_hashtable[index];
5921 sctp_spin_lock(&head->lock);
5922 sctp_for_each_hentry(pp, &head->chain)
5923 if ((pp->port == rover) &&
5924 net_eq(sock_net(sk), pp->net))
5928 sctp_spin_unlock(&head->lock);
5929 } while (--remaining > 0);
5931 /* Exhausted local port range during search? */
5936 /* OK, here is the one we will use. HEAD (the port
5937 * hash table list entry) is non-NULL and we hold it's
5942 /* We are given an specific port number; we verify
5943 * that it is not being used. If it is used, we will
5944 * exahust the search in the hash list corresponding
5945 * to the port number (snum) - we detect that with the
5946 * port iterator, pp being NULL.
5948 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
5949 sctp_spin_lock(&head->lock);
5950 sctp_for_each_hentry(pp, &head->chain) {
5951 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
5958 if (!hlist_empty(&pp->owner)) {
5959 /* We had a port hash table hit - there is an
5960 * available port (pp != NULL) and it is being
5961 * used by other socket (pp->owner not empty); that other
5962 * socket is going to be sk2.
5964 int reuse = sk->sk_reuse;
5967 pr_debug("%s: found a possible match\n", __func__);
5969 if (pp->fastreuse && sk->sk_reuse &&
5970 sk->sk_state != SCTP_SS_LISTENING)
5973 /* Run through the list of sockets bound to the port
5974 * (pp->port) [via the pointers bind_next and
5975 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5976 * we get the endpoint they describe and run through
5977 * the endpoint's list of IP (v4 or v6) addresses,
5978 * comparing each of the addresses with the address of
5979 * the socket sk. If we find a match, then that means
5980 * that this port/socket (sk) combination are already
5983 sk_for_each_bound(sk2, &pp->owner) {
5984 struct sctp_endpoint *ep2;
5985 ep2 = sctp_sk(sk2)->ep;
5988 (reuse && sk2->sk_reuse &&
5989 sk2->sk_state != SCTP_SS_LISTENING))
5992 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5993 sctp_sk(sk2), sctp_sk(sk))) {
5999 pr_debug("%s: found a match\n", __func__);
6002 /* If there was a hash table miss, create a new port. */
6004 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6007 /* In either case (hit or miss), make sure fastreuse is 1 only
6008 * if sk->sk_reuse is too (that is, if the caller requested
6009 * SO_REUSEADDR on this socket -sk-).
6011 if (hlist_empty(&pp->owner)) {
6012 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6016 } else if (pp->fastreuse &&
6017 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6020 /* We are set, so fill up all the data in the hash table
6021 * entry, tie the socket list information with the rest of the
6022 * sockets FIXME: Blurry, NPI (ipg).
6025 if (!sctp_sk(sk)->bind_hash) {
6026 inet_sk(sk)->inet_num = snum;
6027 sk_add_bind_node(sk, &pp->owner);
6028 sctp_sk(sk)->bind_hash = pp;
6033 sctp_spin_unlock(&head->lock);
6036 sctp_local_bh_enable();
6040 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6041 * port is requested.
6043 static int sctp_get_port(struct sock *sk, unsigned short snum)
6045 union sctp_addr addr;
6046 struct sctp_af *af = sctp_sk(sk)->pf->af;
6048 /* Set up a dummy address struct from the sk. */
6049 af->from_sk(&addr, sk);
6050 addr.v4.sin_port = htons(snum);
6052 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6053 return !!sctp_get_port_local(sk, &addr);
6057 * Move a socket to LISTENING state.
6059 static int sctp_listen_start(struct sock *sk, int backlog)
6061 struct sctp_sock *sp = sctp_sk(sk);
6062 struct sctp_endpoint *ep = sp->ep;
6063 struct crypto_hash *tfm = NULL;
6066 /* Allocate HMAC for generating cookie. */
6067 if (!sp->hmac && sp->sctp_hmac_alg) {
6068 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6069 tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
6071 net_info_ratelimited("failed to load transform for %s: %ld\n",
6072 sp->sctp_hmac_alg, PTR_ERR(tfm));
6075 sctp_sk(sk)->hmac = tfm;
6079 * If a bind() or sctp_bindx() is not called prior to a listen()
6080 * call that allows new associations to be accepted, the system
6081 * picks an ephemeral port and will choose an address set equivalent
6082 * to binding with a wildcard address.
6084 * This is not currently spelled out in the SCTP sockets
6085 * extensions draft, but follows the practice as seen in TCP
6089 sk->sk_state = SCTP_SS_LISTENING;
6090 if (!ep->base.bind_addr.port) {
6091 if (sctp_autobind(sk))
6094 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6095 sk->sk_state = SCTP_SS_CLOSED;
6100 sk->sk_max_ack_backlog = backlog;
6101 sctp_hash_endpoint(ep);
6106 * 4.1.3 / 5.1.3 listen()
6108 * By default, new associations are not accepted for UDP style sockets.
6109 * An application uses listen() to mark a socket as being able to
6110 * accept new associations.
6112 * On TCP style sockets, applications use listen() to ready the SCTP
6113 * endpoint for accepting inbound associations.
6115 * On both types of endpoints a backlog of '0' disables listening.
6117 * Move a socket to LISTENING state.
6119 int sctp_inet_listen(struct socket *sock, int backlog)
6121 struct sock *sk = sock->sk;
6122 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6125 if (unlikely(backlog < 0))
6130 /* Peeled-off sockets are not allowed to listen(). */
6131 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6134 if (sock->state != SS_UNCONNECTED)
6137 /* If backlog is zero, disable listening. */
6139 if (sctp_sstate(sk, CLOSED))
6143 sctp_unhash_endpoint(ep);
6144 sk->sk_state = SCTP_SS_CLOSED;
6146 sctp_sk(sk)->bind_hash->fastreuse = 1;
6150 /* If we are already listening, just update the backlog */
6151 if (sctp_sstate(sk, LISTENING))
6152 sk->sk_max_ack_backlog = backlog;
6154 err = sctp_listen_start(sk, backlog);
6161 sctp_release_sock(sk);
6166 * This function is done by modeling the current datagram_poll() and the
6167 * tcp_poll(). Note that, based on these implementations, we don't
6168 * lock the socket in this function, even though it seems that,
6169 * ideally, locking or some other mechanisms can be used to ensure
6170 * the integrity of the counters (sndbuf and wmem_alloc) used
6171 * in this place. We assume that we don't need locks either until proven
6174 * Another thing to note is that we include the Async I/O support
6175 * here, again, by modeling the current TCP/UDP code. We don't have
6176 * a good way to test with it yet.
6178 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6180 struct sock *sk = sock->sk;
6181 struct sctp_sock *sp = sctp_sk(sk);
6184 poll_wait(file, sk_sleep(sk), wait);
6186 /* A TCP-style listening socket becomes readable when the accept queue
6189 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6190 return (!list_empty(&sp->ep->asocs)) ?
6191 (POLLIN | POLLRDNORM) : 0;
6195 /* Is there any exceptional events? */
6196 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6198 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6199 if (sk->sk_shutdown & RCV_SHUTDOWN)
6200 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6201 if (sk->sk_shutdown == SHUTDOWN_MASK)
6204 /* Is it readable? Reconsider this code with TCP-style support. */
6205 if (!skb_queue_empty(&sk->sk_receive_queue))
6206 mask |= POLLIN | POLLRDNORM;
6208 /* The association is either gone or not ready. */
6209 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6212 /* Is it writable? */
6213 if (sctp_writeable(sk)) {
6214 mask |= POLLOUT | POLLWRNORM;
6216 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6218 * Since the socket is not locked, the buffer
6219 * might be made available after the writeable check and
6220 * before the bit is set. This could cause a lost I/O
6221 * signal. tcp_poll() has a race breaker for this race
6222 * condition. Based on their implementation, we put
6223 * in the following code to cover it as well.
6225 if (sctp_writeable(sk))
6226 mask |= POLLOUT | POLLWRNORM;
6231 /********************************************************************
6232 * 2nd Level Abstractions
6233 ********************************************************************/
6235 static struct sctp_bind_bucket *sctp_bucket_create(
6236 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6238 struct sctp_bind_bucket *pp;
6240 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6242 SCTP_DBG_OBJCNT_INC(bind_bucket);
6245 INIT_HLIST_HEAD(&pp->owner);
6247 hlist_add_head(&pp->node, &head->chain);
6252 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6253 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6255 if (pp && hlist_empty(&pp->owner)) {
6256 __hlist_del(&pp->node);
6257 kmem_cache_free(sctp_bucket_cachep, pp);
6258 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6262 /* Release this socket's reference to a local port. */
6263 static inline void __sctp_put_port(struct sock *sk)
6265 struct sctp_bind_hashbucket *head =
6266 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6267 inet_sk(sk)->inet_num)];
6268 struct sctp_bind_bucket *pp;
6270 sctp_spin_lock(&head->lock);
6271 pp = sctp_sk(sk)->bind_hash;
6272 __sk_del_bind_node(sk);
6273 sctp_sk(sk)->bind_hash = NULL;
6274 inet_sk(sk)->inet_num = 0;
6275 sctp_bucket_destroy(pp);
6276 sctp_spin_unlock(&head->lock);
6279 void sctp_put_port(struct sock *sk)
6281 sctp_local_bh_disable();
6282 __sctp_put_port(sk);
6283 sctp_local_bh_enable();
6287 * The system picks an ephemeral port and choose an address set equivalent
6288 * to binding with a wildcard address.
6289 * One of those addresses will be the primary address for the association.
6290 * This automatically enables the multihoming capability of SCTP.
6292 static int sctp_autobind(struct sock *sk)
6294 union sctp_addr autoaddr;
6298 /* Initialize a local sockaddr structure to INADDR_ANY. */
6299 af = sctp_sk(sk)->pf->af;
6301 port = htons(inet_sk(sk)->inet_num);
6302 af->inaddr_any(&autoaddr, port);
6304 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6307 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6310 * 4.2 The cmsghdr Structure *
6312 * When ancillary data is sent or received, any number of ancillary data
6313 * objects can be specified by the msg_control and msg_controllen members of
6314 * the msghdr structure, because each object is preceded by
6315 * a cmsghdr structure defining the object's length (the cmsg_len member).
6316 * Historically Berkeley-derived implementations have passed only one object
6317 * at a time, but this API allows multiple objects to be
6318 * passed in a single call to sendmsg() or recvmsg(). The following example
6319 * shows two ancillary data objects in a control buffer.
6321 * |<--------------------------- msg_controllen -------------------------->|
6324 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6326 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6329 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6331 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6334 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6335 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6337 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6339 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6346 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
6348 struct cmsghdr *cmsg;
6349 struct msghdr *my_msg = (struct msghdr *)msg;
6351 for (cmsg = CMSG_FIRSTHDR(msg);
6353 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6354 if (!CMSG_OK(my_msg, cmsg))
6357 /* Should we parse this header or ignore? */
6358 if (cmsg->cmsg_level != IPPROTO_SCTP)
6361 /* Strictly check lengths following example in SCM code. */
6362 switch (cmsg->cmsg_type) {
6364 /* SCTP Socket API Extension
6365 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6367 * This cmsghdr structure provides information for
6368 * initializing new SCTP associations with sendmsg().
6369 * The SCTP_INITMSG socket option uses this same data
6370 * structure. This structure is not used for
6373 * cmsg_level cmsg_type cmsg_data[]
6374 * ------------ ------------ ----------------------
6375 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6377 if (cmsg->cmsg_len !=
6378 CMSG_LEN(sizeof(struct sctp_initmsg)))
6380 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6384 /* SCTP Socket API Extension
6385 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6387 * This cmsghdr structure specifies SCTP options for
6388 * sendmsg() and describes SCTP header information
6389 * about a received message through recvmsg().
6391 * cmsg_level cmsg_type cmsg_data[]
6392 * ------------ ------------ ----------------------
6393 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6395 if (cmsg->cmsg_len !=
6396 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6400 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6402 /* Minimally, validate the sinfo_flags. */
6403 if (cmsgs->info->sinfo_flags &
6404 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6405 SCTP_ABORT | SCTP_EOF))
6417 * Wait for a packet..
6418 * Note: This function is the same function as in core/datagram.c
6419 * with a few modifications to make lksctp work.
6421 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6426 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6428 /* Socket errors? */
6429 error = sock_error(sk);
6433 if (!skb_queue_empty(&sk->sk_receive_queue))
6436 /* Socket shut down? */
6437 if (sk->sk_shutdown & RCV_SHUTDOWN)
6440 /* Sequenced packets can come disconnected. If so we report the
6445 /* Is there a good reason to think that we may receive some data? */
6446 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6449 /* Handle signals. */
6450 if (signal_pending(current))
6453 /* Let another process have a go. Since we are going to sleep
6454 * anyway. Note: This may cause odd behaviors if the message
6455 * does not fit in the user's buffer, but this seems to be the
6456 * only way to honor MSG_DONTWAIT realistically.
6458 sctp_release_sock(sk);
6459 *timeo_p = schedule_timeout(*timeo_p);
6463 finish_wait(sk_sleep(sk), &wait);
6467 error = sock_intr_errno(*timeo_p);
6470 finish_wait(sk_sleep(sk), &wait);
6475 /* Receive a datagram.
6476 * Note: This is pretty much the same routine as in core/datagram.c
6477 * with a few changes to make lksctp work.
6479 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6480 int noblock, int *err)
6483 struct sk_buff *skb;
6486 timeo = sock_rcvtimeo(sk, noblock);
6488 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
6489 MAX_SCHEDULE_TIMEOUT);
6492 /* Again only user level code calls this function,
6493 * so nothing interrupt level
6494 * will suddenly eat the receive_queue.
6496 * Look at current nfs client by the way...
6497 * However, this function was correct in any case. 8)
6499 if (flags & MSG_PEEK) {
6500 spin_lock_bh(&sk->sk_receive_queue.lock);
6501 skb = skb_peek(&sk->sk_receive_queue);
6503 atomic_inc(&skb->users);
6504 spin_unlock_bh(&sk->sk_receive_queue.lock);
6506 skb = skb_dequeue(&sk->sk_receive_queue);
6512 /* Caller is allowed not to check sk->sk_err before calling. */
6513 error = sock_error(sk);
6517 if (sk->sk_shutdown & RCV_SHUTDOWN)
6520 /* User doesn't want to wait. */
6524 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6533 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6534 static void __sctp_write_space(struct sctp_association *asoc)
6536 struct sock *sk = asoc->base.sk;
6537 struct socket *sock = sk->sk_socket;
6539 if ((sctp_wspace(asoc) > 0) && sock) {
6540 if (waitqueue_active(&asoc->wait))
6541 wake_up_interruptible(&asoc->wait);
6543 if (sctp_writeable(sk)) {
6544 wait_queue_head_t *wq = sk_sleep(sk);
6546 if (wq && waitqueue_active(wq))
6547 wake_up_interruptible(wq);
6549 /* Note that we try to include the Async I/O support
6550 * here by modeling from the current TCP/UDP code.
6551 * We have not tested with it yet.
6553 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6554 sock_wake_async(sock,
6555 SOCK_WAKE_SPACE, POLL_OUT);
6560 /* Do accounting for the sndbuf space.
6561 * Decrement the used sndbuf space of the corresponding association by the
6562 * data size which was just transmitted(freed).
6564 static void sctp_wfree(struct sk_buff *skb)
6566 struct sctp_association *asoc;
6567 struct sctp_chunk *chunk;
6570 /* Get the saved chunk pointer. */
6571 chunk = *((struct sctp_chunk **)(skb->cb));
6574 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6575 sizeof(struct sk_buff) +
6576 sizeof(struct sctp_chunk);
6578 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6581 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6583 sk->sk_wmem_queued -= skb->truesize;
6584 sk_mem_uncharge(sk, skb->truesize);
6587 __sctp_write_space(asoc);
6589 sctp_association_put(asoc);
6592 /* Do accounting for the receive space on the socket.
6593 * Accounting for the association is done in ulpevent.c
6594 * We set this as a destructor for the cloned data skbs so that
6595 * accounting is done at the correct time.
6597 void sctp_sock_rfree(struct sk_buff *skb)
6599 struct sock *sk = skb->sk;
6600 struct sctp_ulpevent *event = sctp_skb2event(skb);
6602 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6605 * Mimic the behavior of sock_rfree
6607 sk_mem_uncharge(sk, event->rmem_len);
6611 /* Helper function to wait for space in the sndbuf. */
6612 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6615 struct sock *sk = asoc->base.sk;
6617 long current_timeo = *timeo_p;
6620 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
6623 /* Increment the association's refcnt. */
6624 sctp_association_hold(asoc);
6626 /* Wait on the association specific sndbuf space. */
6628 prepare_to_wait_exclusive(&asoc->wait, &wait,
6629 TASK_INTERRUPTIBLE);
6632 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6635 if (signal_pending(current))
6636 goto do_interrupted;
6637 if (msg_len <= sctp_wspace(asoc))
6640 /* Let another process have a go. Since we are going
6643 sctp_release_sock(sk);
6644 current_timeo = schedule_timeout(current_timeo);
6645 BUG_ON(sk != asoc->base.sk);
6648 *timeo_p = current_timeo;
6652 finish_wait(&asoc->wait, &wait);
6654 /* Release the association's refcnt. */
6655 sctp_association_put(asoc);
6664 err = sock_intr_errno(*timeo_p);
6672 void sctp_data_ready(struct sock *sk, int len)
6674 struct socket_wq *wq;
6677 wq = rcu_dereference(sk->sk_wq);
6678 if (wq_has_sleeper(wq))
6679 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6680 POLLRDNORM | POLLRDBAND);
6681 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6685 /* If socket sndbuf has changed, wake up all per association waiters. */
6686 void sctp_write_space(struct sock *sk)
6688 struct sctp_association *asoc;
6690 /* Wake up the tasks in each wait queue. */
6691 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6692 __sctp_write_space(asoc);
6696 /* Is there any sndbuf space available on the socket?
6698 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6699 * associations on the same socket. For a UDP-style socket with
6700 * multiple associations, it is possible for it to be "unwriteable"
6701 * prematurely. I assume that this is acceptable because
6702 * a premature "unwriteable" is better than an accidental "writeable" which
6703 * would cause an unwanted block under certain circumstances. For the 1-1
6704 * UDP-style sockets or TCP-style sockets, this code should work.
6707 static int sctp_writeable(struct sock *sk)
6711 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6717 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6718 * returns immediately with EINPROGRESS.
6720 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6722 struct sock *sk = asoc->base.sk;
6724 long current_timeo = *timeo_p;
6727 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
6729 /* Increment the association's refcnt. */
6730 sctp_association_hold(asoc);
6733 prepare_to_wait_exclusive(&asoc->wait, &wait,
6734 TASK_INTERRUPTIBLE);
6737 if (sk->sk_shutdown & RCV_SHUTDOWN)
6739 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6742 if (signal_pending(current))
6743 goto do_interrupted;
6745 if (sctp_state(asoc, ESTABLISHED))
6748 /* Let another process have a go. Since we are going
6751 sctp_release_sock(sk);
6752 current_timeo = schedule_timeout(current_timeo);
6755 *timeo_p = current_timeo;
6759 finish_wait(&asoc->wait, &wait);
6761 /* Release the association's refcnt. */
6762 sctp_association_put(asoc);
6767 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6770 err = -ECONNREFUSED;
6774 err = sock_intr_errno(*timeo_p);
6782 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6784 struct sctp_endpoint *ep;
6788 ep = sctp_sk(sk)->ep;
6792 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6793 TASK_INTERRUPTIBLE);
6795 if (list_empty(&ep->asocs)) {
6796 sctp_release_sock(sk);
6797 timeo = schedule_timeout(timeo);
6802 if (!sctp_sstate(sk, LISTENING))
6806 if (!list_empty(&ep->asocs))
6809 err = sock_intr_errno(timeo);
6810 if (signal_pending(current))
6818 finish_wait(sk_sleep(sk), &wait);
6823 static void sctp_wait_for_close(struct sock *sk, long timeout)
6828 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6829 if (list_empty(&sctp_sk(sk)->ep->asocs))
6831 sctp_release_sock(sk);
6832 timeout = schedule_timeout(timeout);
6834 } while (!signal_pending(current) && timeout);
6836 finish_wait(sk_sleep(sk), &wait);
6839 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6841 struct sk_buff *frag;
6846 /* Don't forget the fragments. */
6847 skb_walk_frags(skb, frag)
6848 sctp_skb_set_owner_r_frag(frag, sk);
6851 sctp_skb_set_owner_r(skb, sk);
6854 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6855 struct sctp_association *asoc)
6857 struct inet_sock *inet = inet_sk(sk);
6858 struct inet_sock *newinet;
6860 newsk->sk_type = sk->sk_type;
6861 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6862 newsk->sk_flags = sk->sk_flags;
6863 newsk->sk_no_check = sk->sk_no_check;
6864 newsk->sk_reuse = sk->sk_reuse;
6866 newsk->sk_shutdown = sk->sk_shutdown;
6867 newsk->sk_destruct = sctp_destruct_sock;
6868 newsk->sk_family = sk->sk_family;
6869 newsk->sk_protocol = IPPROTO_SCTP;
6870 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6871 newsk->sk_sndbuf = sk->sk_sndbuf;
6872 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6873 newsk->sk_lingertime = sk->sk_lingertime;
6874 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6875 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6877 newinet = inet_sk(newsk);
6879 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6880 * getsockname() and getpeername()
6882 newinet->inet_sport = inet->inet_sport;
6883 newinet->inet_saddr = inet->inet_saddr;
6884 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6885 newinet->inet_dport = htons(asoc->peer.port);
6886 newinet->pmtudisc = inet->pmtudisc;
6887 newinet->inet_id = asoc->next_tsn ^ jiffies;
6889 newinet->uc_ttl = inet->uc_ttl;
6890 newinet->mc_loop = 1;
6891 newinet->mc_ttl = 1;
6892 newinet->mc_index = 0;
6893 newinet->mc_list = NULL;
6896 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6897 * and its messages to the newsk.
6899 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6900 struct sctp_association *assoc,
6901 sctp_socket_type_t type)
6903 struct sctp_sock *oldsp = sctp_sk(oldsk);
6904 struct sctp_sock *newsp = sctp_sk(newsk);
6905 struct sctp_bind_bucket *pp; /* hash list port iterator */
6906 struct sctp_endpoint *newep = newsp->ep;
6907 struct sk_buff *skb, *tmp;
6908 struct sctp_ulpevent *event;
6909 struct sctp_bind_hashbucket *head;
6910 struct list_head tmplist;
6912 /* Migrate socket buffer sizes and all the socket level options to the
6915 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6916 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6917 /* Brute force copy old sctp opt. */
6918 if (oldsp->do_auto_asconf) {
6919 memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
6920 inet_sk_copy_descendant(newsk, oldsk);
6921 memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
6923 inet_sk_copy_descendant(newsk, oldsk);
6925 /* Restore the ep value that was overwritten with the above structure
6931 /* Hook this new socket in to the bind_hash list. */
6932 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
6933 inet_sk(oldsk)->inet_num)];
6934 sctp_local_bh_disable();
6935 sctp_spin_lock(&head->lock);
6936 pp = sctp_sk(oldsk)->bind_hash;
6937 sk_add_bind_node(newsk, &pp->owner);
6938 sctp_sk(newsk)->bind_hash = pp;
6939 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6940 sctp_spin_unlock(&head->lock);
6941 sctp_local_bh_enable();
6943 /* Copy the bind_addr list from the original endpoint to the new
6944 * endpoint so that we can handle restarts properly
6946 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6947 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6949 /* Move any messages in the old socket's receive queue that are for the
6950 * peeled off association to the new socket's receive queue.
6952 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6953 event = sctp_skb2event(skb);
6954 if (event->asoc == assoc) {
6955 __skb_unlink(skb, &oldsk->sk_receive_queue);
6956 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6957 sctp_skb_set_owner_r_frag(skb, newsk);
6961 /* Clean up any messages pending delivery due to partial
6962 * delivery. Three cases:
6963 * 1) No partial deliver; no work.
6964 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6965 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6967 skb_queue_head_init(&newsp->pd_lobby);
6968 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6970 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6971 struct sk_buff_head *queue;
6973 /* Decide which queue to move pd_lobby skbs to. */
6974 if (assoc->ulpq.pd_mode) {
6975 queue = &newsp->pd_lobby;
6977 queue = &newsk->sk_receive_queue;
6979 /* Walk through the pd_lobby, looking for skbs that
6980 * need moved to the new socket.
6982 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6983 event = sctp_skb2event(skb);
6984 if (event->asoc == assoc) {
6985 __skb_unlink(skb, &oldsp->pd_lobby);
6986 __skb_queue_tail(queue, skb);
6987 sctp_skb_set_owner_r_frag(skb, newsk);
6991 /* Clear up any skbs waiting for the partial
6992 * delivery to finish.
6994 if (assoc->ulpq.pd_mode)
6995 sctp_clear_pd(oldsk, NULL);
6999 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7000 sctp_skb_set_owner_r_frag(skb, newsk);
7002 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7003 sctp_skb_set_owner_r_frag(skb, newsk);
7005 /* Set the type of socket to indicate that it is peeled off from the
7006 * original UDP-style socket or created with the accept() call on a
7007 * TCP-style socket..
7011 /* Mark the new socket "in-use" by the user so that any packets
7012 * that may arrive on the association after we've moved it are
7013 * queued to the backlog. This prevents a potential race between
7014 * backlog processing on the old socket and new-packet processing
7015 * on the new socket.
7017 * The caller has just allocated newsk so we can guarantee that other
7018 * paths won't try to lock it and then oldsk.
7020 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7021 sctp_assoc_migrate(assoc, newsk);
7023 /* If the association on the newsk is already closed before accept()
7024 * is called, set RCV_SHUTDOWN flag.
7026 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
7027 newsk->sk_shutdown |= RCV_SHUTDOWN;
7029 newsk->sk_state = SCTP_SS_ESTABLISHED;
7030 sctp_release_sock(newsk);
7034 /* This proto struct describes the ULP interface for SCTP. */
7035 struct proto sctp_prot = {
7037 .owner = THIS_MODULE,
7038 .close = sctp_close,
7039 .connect = sctp_connect,
7040 .disconnect = sctp_disconnect,
7041 .accept = sctp_accept,
7042 .ioctl = sctp_ioctl,
7043 .init = sctp_init_sock,
7044 .destroy = sctp_destroy_sock,
7045 .shutdown = sctp_shutdown,
7046 .setsockopt = sctp_setsockopt,
7047 .getsockopt = sctp_getsockopt,
7048 .sendmsg = sctp_sendmsg,
7049 .recvmsg = sctp_recvmsg,
7051 .backlog_rcv = sctp_backlog_rcv,
7053 .unhash = sctp_unhash,
7054 .get_port = sctp_get_port,
7055 .obj_size = sizeof(struct sctp_sock),
7056 .sysctl_mem = sysctl_sctp_mem,
7057 .sysctl_rmem = sysctl_sctp_rmem,
7058 .sysctl_wmem = sysctl_sctp_wmem,
7059 .memory_pressure = &sctp_memory_pressure,
7060 .enter_memory_pressure = sctp_enter_memory_pressure,
7061 .memory_allocated = &sctp_memory_allocated,
7062 .sockets_allocated = &sctp_sockets_allocated,
7065 #if IS_ENABLED(CONFIG_IPV6)
7067 struct proto sctpv6_prot = {
7069 .owner = THIS_MODULE,
7070 .close = sctp_close,
7071 .connect = sctp_connect,
7072 .disconnect = sctp_disconnect,
7073 .accept = sctp_accept,
7074 .ioctl = sctp_ioctl,
7075 .init = sctp_init_sock,
7076 .destroy = sctp_destroy_sock,
7077 .shutdown = sctp_shutdown,
7078 .setsockopt = sctp_setsockopt,
7079 .getsockopt = sctp_getsockopt,
7080 .sendmsg = sctp_sendmsg,
7081 .recvmsg = sctp_recvmsg,
7083 .backlog_rcv = sctp_backlog_rcv,
7085 .unhash = sctp_unhash,
7086 .get_port = sctp_get_port,
7087 .obj_size = sizeof(struct sctp6_sock),
7088 .sysctl_mem = sysctl_sctp_mem,
7089 .sysctl_rmem = sysctl_sctp_rmem,
7090 .sysctl_wmem = sysctl_sctp_wmem,
7091 .memory_pressure = &sctp_memory_pressure,
7092 .enter_memory_pressure = sctp_enter_memory_pressure,
7093 .memory_allocated = &sctp_memory_allocated,
7094 .sockets_allocated = &sctp_sockets_allocated,
7096 #endif /* IS_ENABLED(CONFIG_IPV6) */