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;
809 /* We do not need RCU protection throughout this loop
810 * because this is done under a socket lock from the
813 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
821 /* Reset use_as_src flag for the addresses in the bind address
822 * list that are to be deleted.
825 for (i = 0; i < addrcnt; i++) {
827 af = sctp_get_af_specific(laddr->v4.sin_family);
828 list_for_each_entry(saddr, &bp->address_list, list) {
829 if (sctp_cmp_addr_exact(&saddr->a, laddr))
830 saddr->state = SCTP_ADDR_DEL;
832 addr_buf += af->sockaddr_len;
835 /* Update the route and saddr entries for all the transports
836 * as some of the addresses in the bind address list are
837 * about to be deleted and cannot be used as source addresses.
839 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
841 dst_release(transport->dst);
842 sctp_transport_route(transport, NULL,
843 sctp_sk(asoc->base.sk));
847 /* We don't need to transmit ASCONF */
849 retval = sctp_send_asconf(asoc, chunk);
855 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
856 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
858 struct sock *sk = sctp_opt2sk(sp);
859 union sctp_addr *addr;
862 /* It is safe to write port space in caller. */
864 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
865 af = sctp_get_af_specific(addr->sa.sa_family);
868 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
871 if (addrw->state == SCTP_ADDR_NEW)
872 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
874 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
877 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
880 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
883 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
884 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
887 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
888 * Section 3.1.2 for this usage.
890 * addrs is a pointer to an array of one or more socket addresses. Each
891 * address is contained in its appropriate structure (i.e. struct
892 * sockaddr_in or struct sockaddr_in6) the family of the address type
893 * must be used to distinguish the address length (note that this
894 * representation is termed a "packed array" of addresses). The caller
895 * specifies the number of addresses in the array with addrcnt.
897 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
898 * -1, and sets errno to the appropriate error code.
900 * For SCTP, the port given in each socket address must be the same, or
901 * sctp_bindx() will fail, setting errno to EINVAL.
903 * The flags parameter is formed from the bitwise OR of zero or more of
904 * the following currently defined flags:
906 * SCTP_BINDX_ADD_ADDR
908 * SCTP_BINDX_REM_ADDR
910 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
911 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
912 * addresses from the association. The two flags are mutually exclusive;
913 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
914 * not remove all addresses from an association; sctp_bindx() will
915 * reject such an attempt with EINVAL.
917 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
918 * additional addresses with an endpoint after calling bind(). Or use
919 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
920 * socket is associated with so that no new association accepted will be
921 * associated with those addresses. If the endpoint supports dynamic
922 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
923 * endpoint to send the appropriate message to the peer to change the
924 * peers address lists.
926 * Adding and removing addresses from a connected association is
927 * optional functionality. Implementations that do not support this
928 * functionality should return EOPNOTSUPP.
930 * Basically do nothing but copying the addresses from user to kernel
931 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
932 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
935 * We don't use copy_from_user() for optimization: we first do the
936 * sanity checks (buffer size -fast- and access check-healthy
937 * pointer); if all of those succeed, then we can alloc the memory
938 * (expensive operation) needed to copy the data to kernel. Then we do
939 * the copying without checking the user space area
940 * (__copy_from_user()).
942 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
945 * sk The sk of the socket
946 * addrs The pointer to the addresses in user land
947 * addrssize Size of the addrs buffer
948 * op Operation to perform (add or remove, see the flags of
951 * Returns 0 if ok, <0 errno code on error.
953 static int sctp_setsockopt_bindx(struct sock* sk,
954 struct sockaddr __user *addrs,
955 int addrs_size, int op)
957 struct sockaddr *kaddrs;
961 struct sockaddr *sa_addr;
965 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
966 __func__, sk, addrs, addrs_size, op);
968 if (unlikely(addrs_size <= 0))
971 /* Check the user passed a healthy pointer. */
972 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
975 /* Alloc space for the address array in kernel memory. */
976 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
977 if (unlikely(!kaddrs))
980 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
985 /* Walk through the addrs buffer and count the number of addresses. */
987 while (walk_size < addrs_size) {
988 if (walk_size + sizeof(sa_family_t) > addrs_size) {
994 af = sctp_get_af_specific(sa_addr->sa_family);
996 /* If the address family is not supported or if this address
997 * causes the address buffer to overflow return EINVAL.
999 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1004 addr_buf += af->sockaddr_len;
1005 walk_size += af->sockaddr_len;
1010 case SCTP_BINDX_ADD_ADDR:
1011 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1014 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1017 case SCTP_BINDX_REM_ADDR:
1018 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1021 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1035 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1037 * Common routine for handling connect() and sctp_connectx().
1038 * Connect will come in with just a single address.
1040 static int __sctp_connect(struct sock* sk,
1041 struct sockaddr *kaddrs,
1043 sctp_assoc_t *assoc_id)
1045 struct net *net = sock_net(sk);
1046 struct sctp_sock *sp;
1047 struct sctp_endpoint *ep;
1048 struct sctp_association *asoc = NULL;
1049 struct sctp_association *asoc2;
1050 struct sctp_transport *transport;
1058 union sctp_addr *sa_addr = NULL;
1060 unsigned short port;
1061 unsigned int f_flags = 0;
1066 /* connect() cannot be done on a socket that is already in ESTABLISHED
1067 * state - UDP-style peeled off socket or a TCP-style socket that
1068 * is already connected.
1069 * It cannot be done even on a TCP-style listening socket.
1071 if (sctp_sstate(sk, ESTABLISHED) ||
1072 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1077 /* Walk through the addrs buffer and count the number of addresses. */
1079 while (walk_size < addrs_size) {
1080 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1086 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1088 /* If the address family is not supported or if this address
1089 * causes the address buffer to overflow return EINVAL.
1091 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1096 port = ntohs(sa_addr->v4.sin_port);
1098 /* Save current address so we can work with it */
1099 memcpy(&to, sa_addr, af->sockaddr_len);
1101 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1105 /* Make sure the destination port is correctly set
1108 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1113 /* Check if there already is a matching association on the
1114 * endpoint (other than the one created here).
1116 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1117 if (asoc2 && asoc2 != asoc) {
1118 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1125 /* If we could not find a matching association on the endpoint,
1126 * make sure that there is no peeled-off association matching
1127 * the peer address even on another socket.
1129 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1130 err = -EADDRNOTAVAIL;
1135 /* If a bind() or sctp_bindx() is not called prior to
1136 * an sctp_connectx() call, the system picks an
1137 * ephemeral port and will choose an address set
1138 * equivalent to binding with a wildcard address.
1140 if (!ep->base.bind_addr.port) {
1141 if (sctp_autobind(sk)) {
1147 * If an unprivileged user inherits a 1-many
1148 * style socket with open associations on a
1149 * privileged port, it MAY be permitted to
1150 * accept new associations, but it SHOULD NOT
1151 * be permitted to open new associations.
1153 if (ep->base.bind_addr.port < PROT_SOCK &&
1154 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1160 scope = sctp_scope(&to);
1161 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1167 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1175 /* Prime the peer's transport structures. */
1176 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1184 addr_buf += af->sockaddr_len;
1185 walk_size += af->sockaddr_len;
1188 /* In case the user of sctp_connectx() wants an association
1189 * id back, assign one now.
1192 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1197 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1202 /* Initialize sk's dport and daddr for getpeername() */
1203 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1204 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1205 af->to_sk_daddr(sa_addr, sk);
1208 /* in-kernel sockets don't generally have a file allocated to them
1209 * if all they do is call sock_create_kern().
1211 if (sk->sk_socket->file)
1212 f_flags = sk->sk_socket->file->f_flags;
1214 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1216 err = sctp_wait_for_connect(asoc, &timeo);
1217 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1218 *assoc_id = asoc->assoc_id;
1220 /* Don't free association on exit. */
1224 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1225 __func__, asoc, kaddrs, err);
1228 /* sctp_primitive_ASSOCIATE may have added this association
1229 * To the hash table, try to unhash it, just in case, its a noop
1230 * if it wasn't hashed so we're safe
1232 sctp_unhash_established(asoc);
1233 sctp_association_free(asoc);
1238 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1241 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1242 * sctp_assoc_t *asoc);
1244 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1245 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1246 * or IPv6 addresses.
1248 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1249 * Section 3.1.2 for this usage.
1251 * addrs is a pointer to an array of one or more socket addresses. Each
1252 * address is contained in its appropriate structure (i.e. struct
1253 * sockaddr_in or struct sockaddr_in6) the family of the address type
1254 * must be used to distengish the address length (note that this
1255 * representation is termed a "packed array" of addresses). The caller
1256 * specifies the number of addresses in the array with addrcnt.
1258 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1259 * the association id of the new association. On failure, sctp_connectx()
1260 * returns -1, and sets errno to the appropriate error code. The assoc_id
1261 * is not touched by the kernel.
1263 * For SCTP, the port given in each socket address must be the same, or
1264 * sctp_connectx() will fail, setting errno to EINVAL.
1266 * An application can use sctp_connectx to initiate an association with
1267 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1268 * allows a caller to specify multiple addresses at which a peer can be
1269 * reached. The way the SCTP stack uses the list of addresses to set up
1270 * the association is implementation dependent. This function only
1271 * specifies that the stack will try to make use of all the addresses in
1272 * the list when needed.
1274 * Note that the list of addresses passed in is only used for setting up
1275 * the association. It does not necessarily equal the set of addresses
1276 * the peer uses for the resulting association. If the caller wants to
1277 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1278 * retrieve them after the association has been set up.
1280 * Basically do nothing but copying the addresses from user to kernel
1281 * land and invoking either sctp_connectx(). This is used for tunneling
1282 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1284 * We don't use copy_from_user() for optimization: we first do the
1285 * sanity checks (buffer size -fast- and access check-healthy
1286 * pointer); if all of those succeed, then we can alloc the memory
1287 * (expensive operation) needed to copy the data to kernel. Then we do
1288 * the copying without checking the user space area
1289 * (__copy_from_user()).
1291 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1294 * sk The sk of the socket
1295 * addrs The pointer to the addresses in user land
1296 * addrssize Size of the addrs buffer
1298 * Returns >=0 if ok, <0 errno code on error.
1300 static int __sctp_setsockopt_connectx(struct sock* sk,
1301 struct sockaddr __user *addrs,
1303 sctp_assoc_t *assoc_id)
1306 struct sockaddr *kaddrs;
1308 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1309 __func__, sk, addrs, addrs_size);
1311 if (unlikely(addrs_size <= 0))
1314 /* Check the user passed a healthy pointer. */
1315 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1318 /* Alloc space for the address array in kernel memory. */
1319 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1320 if (unlikely(!kaddrs))
1323 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1326 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1335 * This is an older interface. It's kept for backward compatibility
1336 * to the option that doesn't provide association id.
1338 static int sctp_setsockopt_connectx_old(struct sock* sk,
1339 struct sockaddr __user *addrs,
1342 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1346 * New interface for the API. The since the API is done with a socket
1347 * option, to make it simple we feed back the association id is as a return
1348 * indication to the call. Error is always negative and association id is
1351 static int sctp_setsockopt_connectx(struct sock* sk,
1352 struct sockaddr __user *addrs,
1355 sctp_assoc_t assoc_id = 0;
1358 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1367 * New (hopefully final) interface for the API.
1368 * We use the sctp_getaddrs_old structure so that use-space library
1369 * can avoid any unnecessary allocations. The only defferent part
1370 * is that we store the actual length of the address buffer into the
1371 * addrs_num structure member. That way we can re-use the existing
1374 static int sctp_getsockopt_connectx3(struct sock* sk, int len,
1375 char __user *optval,
1378 struct sctp_getaddrs_old param;
1379 sctp_assoc_t assoc_id = 0;
1382 if (len < sizeof(param))
1385 if (copy_from_user(¶m, optval, sizeof(param)))
1388 err = __sctp_setsockopt_connectx(sk,
1389 (struct sockaddr __user *)param.addrs,
1390 param.addr_num, &assoc_id);
1392 if (err == 0 || err == -EINPROGRESS) {
1393 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1395 if (put_user(sizeof(assoc_id), optlen))
1402 /* API 3.1.4 close() - UDP Style Syntax
1403 * Applications use close() to perform graceful shutdown (as described in
1404 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1405 * by a UDP-style socket.
1409 * ret = close(int sd);
1411 * sd - the socket descriptor of the associations to be closed.
1413 * To gracefully shutdown a specific association represented by the
1414 * UDP-style socket, an application should use the sendmsg() call,
1415 * passing no user data, but including the appropriate flag in the
1416 * ancillary data (see Section xxxx).
1418 * If sd in the close() call is a branched-off socket representing only
1419 * one association, the shutdown is performed on that association only.
1421 * 4.1.6 close() - TCP Style Syntax
1423 * Applications use close() to gracefully close down an association.
1427 * int close(int sd);
1429 * sd - the socket descriptor of the association to be closed.
1431 * After an application calls close() on a socket descriptor, no further
1432 * socket operations will succeed on that descriptor.
1434 * API 7.1.4 SO_LINGER
1436 * An application using the TCP-style socket can use this option to
1437 * perform the SCTP ABORT primitive. The linger option structure is:
1440 * int l_onoff; // option on/off
1441 * int l_linger; // linger time
1444 * To enable the option, set l_onoff to 1. If the l_linger value is set
1445 * to 0, calling close() is the same as the ABORT primitive. If the
1446 * value is set to a negative value, the setsockopt() call will return
1447 * an error. If the value is set to a positive value linger_time, the
1448 * close() can be blocked for at most linger_time ms. If the graceful
1449 * shutdown phase does not finish during this period, close() will
1450 * return but the graceful shutdown phase continues in the system.
1452 static void sctp_close(struct sock *sk, long timeout)
1454 struct net *net = sock_net(sk);
1455 struct sctp_endpoint *ep;
1456 struct sctp_association *asoc;
1457 struct list_head *pos, *temp;
1458 unsigned int data_was_unread;
1460 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1463 sk->sk_shutdown = SHUTDOWN_MASK;
1464 sk->sk_state = SCTP_SS_CLOSING;
1466 ep = sctp_sk(sk)->ep;
1468 /* Clean up any skbs sitting on the receive queue. */
1469 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1470 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1472 /* Walk all associations on an endpoint. */
1473 list_for_each_safe(pos, temp, &ep->asocs) {
1474 asoc = list_entry(pos, struct sctp_association, asocs);
1476 if (sctp_style(sk, TCP)) {
1477 /* A closed association can still be in the list if
1478 * it belongs to a TCP-style listening socket that is
1479 * not yet accepted. If so, free it. If not, send an
1480 * ABORT or SHUTDOWN based on the linger options.
1482 if (sctp_state(asoc, CLOSED)) {
1483 sctp_unhash_established(asoc);
1484 sctp_association_free(asoc);
1489 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1490 !skb_queue_empty(&asoc->ulpq.reasm) ||
1491 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1492 struct sctp_chunk *chunk;
1494 chunk = sctp_make_abort_user(asoc, NULL, 0);
1496 sctp_primitive_ABORT(net, asoc, chunk);
1498 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1501 /* On a TCP-style socket, block for at most linger_time if set. */
1502 if (sctp_style(sk, TCP) && timeout)
1503 sctp_wait_for_close(sk, timeout);
1505 /* This will run the backlog queue. */
1506 sctp_release_sock(sk);
1508 /* Supposedly, no process has access to the socket, but
1509 * the net layers still may.
1511 sctp_local_bh_disable();
1512 sctp_bh_lock_sock(sk);
1514 /* Hold the sock, since sk_common_release() will put sock_put()
1515 * and we have just a little more cleanup.
1518 sk_common_release(sk);
1520 sctp_bh_unlock_sock(sk);
1521 sctp_local_bh_enable();
1525 SCTP_DBG_OBJCNT_DEC(sock);
1528 /* Handle EPIPE error. */
1529 static int sctp_error(struct sock *sk, int flags, int err)
1532 err = sock_error(sk) ? : -EPIPE;
1533 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1534 send_sig(SIGPIPE, current, 0);
1538 /* API 3.1.3 sendmsg() - UDP Style Syntax
1540 * An application uses sendmsg() and recvmsg() calls to transmit data to
1541 * and receive data from its peer.
1543 * ssize_t sendmsg(int socket, const struct msghdr *message,
1546 * socket - the socket descriptor of the endpoint.
1547 * message - pointer to the msghdr structure which contains a single
1548 * user message and possibly some ancillary data.
1550 * See Section 5 for complete description of the data
1553 * flags - flags sent or received with the user message, see Section
1554 * 5 for complete description of the flags.
1556 * Note: This function could use a rewrite especially when explicit
1557 * connect support comes in.
1559 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1561 static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1563 static int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1564 struct msghdr *msg, size_t msg_len)
1566 struct net *net = sock_net(sk);
1567 struct sctp_sock *sp;
1568 struct sctp_endpoint *ep;
1569 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1570 struct sctp_transport *transport, *chunk_tp;
1571 struct sctp_chunk *chunk;
1573 struct sockaddr *msg_name = NULL;
1574 struct sctp_sndrcvinfo default_sinfo;
1575 struct sctp_sndrcvinfo *sinfo;
1576 struct sctp_initmsg *sinit;
1577 sctp_assoc_t associd = 0;
1578 sctp_cmsgs_t cmsgs = { NULL };
1582 __u16 sinfo_flags = 0;
1583 struct sctp_datamsg *datamsg;
1584 int msg_flags = msg->msg_flags;
1590 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1593 /* We cannot send a message over a TCP-style listening socket. */
1594 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1599 /* Parse out the SCTP CMSGs. */
1600 err = sctp_msghdr_parse(msg, &cmsgs);
1602 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1606 /* Fetch the destination address for this packet. This
1607 * address only selects the association--it is not necessarily
1608 * the address we will send to.
1609 * For a peeled-off socket, msg_name is ignored.
1611 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1612 int msg_namelen = msg->msg_namelen;
1614 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1619 if (msg_namelen > sizeof(to))
1620 msg_namelen = sizeof(to);
1621 memcpy(&to, msg->msg_name, msg_namelen);
1622 msg_name = msg->msg_name;
1628 /* Did the user specify SNDRCVINFO? */
1630 sinfo_flags = sinfo->sinfo_flags;
1631 associd = sinfo->sinfo_assoc_id;
1634 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1635 msg_len, sinfo_flags);
1637 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1638 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1643 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1644 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1645 * If SCTP_ABORT is set, the message length could be non zero with
1646 * the msg_iov set to the user abort reason.
1648 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1649 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1654 /* If SCTP_ADDR_OVER is set, there must be an address
1655 * specified in msg_name.
1657 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1664 pr_debug("%s: about to look up association\n", __func__);
1668 /* If a msg_name has been specified, assume this is to be used. */
1670 /* Look for a matching association on the endpoint. */
1671 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1673 /* If we could not find a matching association on the
1674 * endpoint, make sure that it is not a TCP-style
1675 * socket that already has an association or there is
1676 * no peeled-off association on another socket.
1678 if ((sctp_style(sk, TCP) &&
1679 sctp_sstate(sk, ESTABLISHED)) ||
1680 sctp_endpoint_is_peeled_off(ep, &to)) {
1681 err = -EADDRNOTAVAIL;
1686 asoc = sctp_id2assoc(sk, associd);
1694 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1696 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1697 * socket that has an association in CLOSED state. This can
1698 * happen when an accepted socket has an association that is
1701 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1706 if (sinfo_flags & SCTP_EOF) {
1707 pr_debug("%s: shutting down association:%p\n",
1710 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1714 if (sinfo_flags & SCTP_ABORT) {
1716 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1722 pr_debug("%s: aborting association:%p\n",
1725 sctp_primitive_ABORT(net, asoc, chunk);
1731 /* Do we need to create the association? */
1733 pr_debug("%s: there is no association yet\n", __func__);
1735 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1740 /* Check for invalid stream against the stream counts,
1741 * either the default or the user specified stream counts.
1744 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1745 /* Check against the defaults. */
1746 if (sinfo->sinfo_stream >=
1747 sp->initmsg.sinit_num_ostreams) {
1752 /* Check against the requested. */
1753 if (sinfo->sinfo_stream >=
1754 sinit->sinit_num_ostreams) {
1762 * API 3.1.2 bind() - UDP Style Syntax
1763 * If a bind() or sctp_bindx() is not called prior to a
1764 * sendmsg() call that initiates a new association, the
1765 * system picks an ephemeral port and will choose an address
1766 * set equivalent to binding with a wildcard address.
1768 if (!ep->base.bind_addr.port) {
1769 if (sctp_autobind(sk)) {
1775 * If an unprivileged user inherits a one-to-many
1776 * style socket with open associations on a privileged
1777 * port, it MAY be permitted to accept new associations,
1778 * but it SHOULD NOT be permitted to open new
1781 if (ep->base.bind_addr.port < PROT_SOCK &&
1782 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1788 scope = sctp_scope(&to);
1789 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1795 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1801 /* If the SCTP_INIT ancillary data is specified, set all
1802 * the association init values accordingly.
1805 if (sinit->sinit_num_ostreams) {
1806 asoc->c.sinit_num_ostreams =
1807 sinit->sinit_num_ostreams;
1809 if (sinit->sinit_max_instreams) {
1810 asoc->c.sinit_max_instreams =
1811 sinit->sinit_max_instreams;
1813 if (sinit->sinit_max_attempts) {
1814 asoc->max_init_attempts
1815 = sinit->sinit_max_attempts;
1817 if (sinit->sinit_max_init_timeo) {
1818 asoc->max_init_timeo =
1819 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1823 /* Prime the peer's transport structures. */
1824 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1831 /* ASSERT: we have a valid association at this point. */
1832 pr_debug("%s: we have a valid association\n", __func__);
1835 /* If the user didn't specify SNDRCVINFO, make up one with
1838 memset(&default_sinfo, 0, sizeof(default_sinfo));
1839 default_sinfo.sinfo_stream = asoc->default_stream;
1840 default_sinfo.sinfo_flags = asoc->default_flags;
1841 default_sinfo.sinfo_ppid = asoc->default_ppid;
1842 default_sinfo.sinfo_context = asoc->default_context;
1843 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1844 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1845 sinfo = &default_sinfo;
1848 /* API 7.1.7, the sndbuf size per association bounds the
1849 * maximum size of data that can be sent in a single send call.
1851 if (msg_len > sk->sk_sndbuf) {
1856 if (asoc->pmtu_pending)
1857 sctp_assoc_pending_pmtu(sk, asoc);
1859 /* If fragmentation is disabled and the message length exceeds the
1860 * association fragmentation point, return EMSGSIZE. The I-D
1861 * does not specify what this error is, but this looks like
1864 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1869 /* Check for invalid stream. */
1870 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1875 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1876 if (!sctp_wspace(asoc)) {
1877 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1882 /* If an address is passed with the sendto/sendmsg call, it is used
1883 * to override the primary destination address in the TCP model, or
1884 * when SCTP_ADDR_OVER flag is set in the UDP model.
1886 if ((sctp_style(sk, TCP) && msg_name) ||
1887 (sinfo_flags & SCTP_ADDR_OVER)) {
1888 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1896 /* Auto-connect, if we aren't connected already. */
1897 if (sctp_state(asoc, CLOSED)) {
1898 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1902 pr_debug("%s: we associated primitively\n", __func__);
1905 /* Break the message into multiple chunks of maximum size. */
1906 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1907 if (IS_ERR(datamsg)) {
1908 err = PTR_ERR(datamsg);
1912 /* Now send the (possibly) fragmented message. */
1913 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1914 sctp_chunk_hold(chunk);
1916 /* Do accounting for the write space. */
1917 sctp_set_owner_w(chunk);
1919 chunk->transport = chunk_tp;
1922 /* Send it to the lower layers. Note: all chunks
1923 * must either fail or succeed. The lower layer
1924 * works that way today. Keep it that way or this
1927 err = sctp_primitive_SEND(net, asoc, datamsg);
1928 /* Did the lower layer accept the chunk? */
1930 sctp_datamsg_free(datamsg);
1934 pr_debug("%s: we sent primitively\n", __func__);
1936 sctp_datamsg_put(datamsg);
1939 /* If we are already past ASSOCIATE, the lower
1940 * layers are responsible for association cleanup.
1946 sctp_unhash_established(asoc);
1947 sctp_association_free(asoc);
1950 sctp_release_sock(sk);
1953 return sctp_error(sk, msg_flags, err);
1960 err = sock_error(sk);
1970 /* This is an extended version of skb_pull() that removes the data from the
1971 * start of a skb even when data is spread across the list of skb's in the
1972 * frag_list. len specifies the total amount of data that needs to be removed.
1973 * when 'len' bytes could be removed from the skb, it returns 0.
1974 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1975 * could not be removed.
1977 static int sctp_skb_pull(struct sk_buff *skb, int len)
1979 struct sk_buff *list;
1980 int skb_len = skb_headlen(skb);
1983 if (len <= skb_len) {
1984 __skb_pull(skb, len);
1988 __skb_pull(skb, skb_len);
1990 skb_walk_frags(skb, list) {
1991 rlen = sctp_skb_pull(list, len);
1992 skb->len -= (len-rlen);
1993 skb->data_len -= (len-rlen);
2004 /* API 3.1.3 recvmsg() - UDP Style Syntax
2006 * ssize_t recvmsg(int socket, struct msghdr *message,
2009 * socket - the socket descriptor of the endpoint.
2010 * message - pointer to the msghdr structure which contains a single
2011 * user message and possibly some ancillary data.
2013 * See Section 5 for complete description of the data
2016 * flags - flags sent or received with the user message, see Section
2017 * 5 for complete description of the flags.
2019 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2021 static int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2022 struct msghdr *msg, size_t len, int noblock,
2023 int flags, int *addr_len)
2025 struct sctp_ulpevent *event = NULL;
2026 struct sctp_sock *sp = sctp_sk(sk);
2027 struct sk_buff *skb;
2032 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2033 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2038 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2043 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2047 /* Get the total length of the skb including any skb's in the
2056 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2058 event = sctp_skb2event(skb);
2063 sock_recv_ts_and_drops(msg, sk, skb);
2064 if (sctp_ulpevent_is_notification(event)) {
2065 msg->msg_flags |= MSG_NOTIFICATION;
2066 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2068 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2071 /* Check if we allow SCTP_SNDRCVINFO. */
2072 if (sp->subscribe.sctp_data_io_event)
2073 sctp_ulpevent_read_sndrcvinfo(event, msg);
2075 /* FIXME: we should be calling IP/IPv6 layers. */
2076 if (sk->sk_protinfo.af_inet.cmsg_flags)
2077 ip_cmsg_recv(msg, skb);
2082 /* If skb's length exceeds the user's buffer, update the skb and
2083 * push it back to the receive_queue so that the next call to
2084 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2086 if (skb_len > copied) {
2087 msg->msg_flags &= ~MSG_EOR;
2088 if (flags & MSG_PEEK)
2090 sctp_skb_pull(skb, copied);
2091 skb_queue_head(&sk->sk_receive_queue, skb);
2093 /* When only partial message is copied to the user, increase
2094 * rwnd by that amount. If all the data in the skb is read,
2095 * rwnd is updated when the event is freed.
2097 if (!sctp_ulpevent_is_notification(event))
2098 sctp_assoc_rwnd_increase(event->asoc, copied);
2100 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2101 (event->msg_flags & MSG_EOR))
2102 msg->msg_flags |= MSG_EOR;
2104 msg->msg_flags &= ~MSG_EOR;
2107 if (flags & MSG_PEEK) {
2108 /* Release the skb reference acquired after peeking the skb in
2109 * sctp_skb_recv_datagram().
2113 /* Free the event which includes releasing the reference to
2114 * the owner of the skb, freeing the skb and updating the
2117 sctp_ulpevent_free(event);
2120 sctp_release_sock(sk);
2124 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2126 * This option is a on/off flag. If enabled no SCTP message
2127 * fragmentation will be performed. Instead if a message being sent
2128 * exceeds the current PMTU size, the message will NOT be sent and
2129 * instead a error will be indicated to the user.
2131 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2132 char __user *optval,
2133 unsigned int optlen)
2137 if (optlen < sizeof(int))
2140 if (get_user(val, (int __user *)optval))
2143 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2148 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2149 unsigned int optlen)
2151 struct sctp_association *asoc;
2152 struct sctp_ulpevent *event;
2154 if (optlen > sizeof(struct sctp_event_subscribe))
2156 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2160 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2161 * if there is no data to be sent or retransmit, the stack will
2162 * immediately send up this notification.
2164 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2165 &sctp_sk(sk)->subscribe)) {
2166 asoc = sctp_id2assoc(sk, 0);
2168 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2169 event = sctp_ulpevent_make_sender_dry_event(asoc,
2174 sctp_ulpq_tail_event(&asoc->ulpq, event);
2181 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2183 * This socket option is applicable to the UDP-style socket only. When
2184 * set it will cause associations that are idle for more than the
2185 * specified number of seconds to automatically close. An association
2186 * being idle is defined an association that has NOT sent or received
2187 * user data. The special value of '0' indicates that no automatic
2188 * close of any associations should be performed. The option expects an
2189 * integer defining the number of seconds of idle time before an
2190 * association is closed.
2192 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2193 unsigned int optlen)
2195 struct sctp_sock *sp = sctp_sk(sk);
2197 /* Applicable to UDP-style socket only */
2198 if (sctp_style(sk, TCP))
2200 if (optlen != sizeof(int))
2202 if (copy_from_user(&sp->autoclose, optval, optlen))
2208 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2210 * Applications can enable or disable heartbeats for any peer address of
2211 * an association, modify an address's heartbeat interval, force a
2212 * heartbeat to be sent immediately, and adjust the address's maximum
2213 * number of retransmissions sent before an address is considered
2214 * unreachable. The following structure is used to access and modify an
2215 * address's parameters:
2217 * struct sctp_paddrparams {
2218 * sctp_assoc_t spp_assoc_id;
2219 * struct sockaddr_storage spp_address;
2220 * uint32_t spp_hbinterval;
2221 * uint16_t spp_pathmaxrxt;
2222 * uint32_t spp_pathmtu;
2223 * uint32_t spp_sackdelay;
2224 * uint32_t spp_flags;
2227 * spp_assoc_id - (one-to-many style socket) This is filled in the
2228 * application, and identifies the association for
2230 * spp_address - This specifies which address is of interest.
2231 * spp_hbinterval - This contains the value of the heartbeat interval,
2232 * in milliseconds. If a value of zero
2233 * is present in this field then no changes are to
2234 * be made to this parameter.
2235 * spp_pathmaxrxt - This contains the maximum number of
2236 * retransmissions before this address shall be
2237 * considered unreachable. If a value of zero
2238 * is present in this field then no changes are to
2239 * be made to this parameter.
2240 * spp_pathmtu - When Path MTU discovery is disabled the value
2241 * specified here will be the "fixed" path mtu.
2242 * Note that if the spp_address field is empty
2243 * then all associations on this address will
2244 * have this fixed path mtu set upon them.
2246 * spp_sackdelay - When delayed sack is enabled, this value specifies
2247 * the number of milliseconds that sacks will be delayed
2248 * for. This value will apply to all addresses of an
2249 * association if the spp_address field is empty. Note
2250 * also, that if delayed sack is enabled and this
2251 * value is set to 0, no change is made to the last
2252 * recorded delayed sack timer value.
2254 * spp_flags - These flags are used to control various features
2255 * on an association. The flag field may contain
2256 * zero or more of the following options.
2258 * SPP_HB_ENABLE - Enable heartbeats on the
2259 * specified address. Note that if the address
2260 * field is empty all addresses for the association
2261 * have heartbeats enabled upon them.
2263 * SPP_HB_DISABLE - Disable heartbeats on the
2264 * speicifed address. Note that if the address
2265 * field is empty all addresses for the association
2266 * will have their heartbeats disabled. Note also
2267 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2268 * mutually exclusive, only one of these two should
2269 * be specified. Enabling both fields will have
2270 * undetermined results.
2272 * SPP_HB_DEMAND - Request a user initiated heartbeat
2273 * to be made immediately.
2275 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2276 * heartbeat delayis to be set to the value of 0
2279 * SPP_PMTUD_ENABLE - This field will enable PMTU
2280 * discovery upon the specified address. Note that
2281 * if the address feild is empty then all addresses
2282 * on the association are effected.
2284 * SPP_PMTUD_DISABLE - This field will disable PMTU
2285 * discovery upon the specified address. Note that
2286 * if the address feild is empty then all addresses
2287 * on the association are effected. Not also that
2288 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2289 * exclusive. Enabling both will have undetermined
2292 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2293 * on delayed sack. The time specified in spp_sackdelay
2294 * is used to specify the sack delay for this address. Note
2295 * that if spp_address is empty then all addresses will
2296 * enable delayed sack and take on the sack delay
2297 * value specified in spp_sackdelay.
2298 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2299 * off delayed sack. If the spp_address field is blank then
2300 * delayed sack is disabled for the entire association. Note
2301 * also that this field is mutually exclusive to
2302 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2305 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2306 struct sctp_transport *trans,
2307 struct sctp_association *asoc,
2308 struct sctp_sock *sp,
2311 int sackdelay_change)
2315 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2316 struct net *net = sock_net(trans->asoc->base.sk);
2318 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2323 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2324 * this field is ignored. Note also that a value of zero indicates
2325 * the current setting should be left unchanged.
2327 if (params->spp_flags & SPP_HB_ENABLE) {
2329 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2330 * set. This lets us use 0 value when this flag
2333 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2334 params->spp_hbinterval = 0;
2336 if (params->spp_hbinterval ||
2337 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2340 msecs_to_jiffies(params->spp_hbinterval);
2343 msecs_to_jiffies(params->spp_hbinterval);
2345 sp->hbinterval = params->spp_hbinterval;
2352 trans->param_flags =
2353 (trans->param_flags & ~SPP_HB) | hb_change;
2356 (asoc->param_flags & ~SPP_HB) | hb_change;
2359 (sp->param_flags & ~SPP_HB) | hb_change;
2363 /* When Path MTU discovery is disabled the value specified here will
2364 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2365 * include the flag SPP_PMTUD_DISABLE for this field to have any
2368 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2370 trans->pathmtu = params->spp_pathmtu;
2371 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2373 asoc->pathmtu = params->spp_pathmtu;
2374 sctp_frag_point(asoc, params->spp_pathmtu);
2376 sp->pathmtu = params->spp_pathmtu;
2382 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2383 (params->spp_flags & SPP_PMTUD_ENABLE);
2384 trans->param_flags =
2385 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2387 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2388 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2392 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2395 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2399 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2400 * value of this field is ignored. Note also that a value of zero
2401 * indicates the current setting should be left unchanged.
2403 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2406 msecs_to_jiffies(params->spp_sackdelay);
2409 msecs_to_jiffies(params->spp_sackdelay);
2411 sp->sackdelay = params->spp_sackdelay;
2415 if (sackdelay_change) {
2417 trans->param_flags =
2418 (trans->param_flags & ~SPP_SACKDELAY) |
2422 (asoc->param_flags & ~SPP_SACKDELAY) |
2426 (sp->param_flags & ~SPP_SACKDELAY) |
2431 /* Note that a value of zero indicates the current setting should be
2434 if (params->spp_pathmaxrxt) {
2436 trans->pathmaxrxt = params->spp_pathmaxrxt;
2438 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2440 sp->pathmaxrxt = params->spp_pathmaxrxt;
2447 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2448 char __user *optval,
2449 unsigned int optlen)
2451 struct sctp_paddrparams params;
2452 struct sctp_transport *trans = NULL;
2453 struct sctp_association *asoc = NULL;
2454 struct sctp_sock *sp = sctp_sk(sk);
2456 int hb_change, pmtud_change, sackdelay_change;
2458 if (optlen != sizeof(struct sctp_paddrparams))
2461 if (copy_from_user(¶ms, optval, optlen))
2464 /* Validate flags and value parameters. */
2465 hb_change = params.spp_flags & SPP_HB;
2466 pmtud_change = params.spp_flags & SPP_PMTUD;
2467 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2469 if (hb_change == SPP_HB ||
2470 pmtud_change == SPP_PMTUD ||
2471 sackdelay_change == SPP_SACKDELAY ||
2472 params.spp_sackdelay > 500 ||
2473 (params.spp_pathmtu &&
2474 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2477 /* If an address other than INADDR_ANY is specified, and
2478 * no transport is found, then the request is invalid.
2480 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
2481 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2482 params.spp_assoc_id);
2487 /* Get association, if assoc_id != 0 and the socket is a one
2488 * to many style socket, and an association was not found, then
2489 * the id was invalid.
2491 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2492 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2495 /* Heartbeat demand can only be sent on a transport or
2496 * association, but not a socket.
2498 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2501 /* Process parameters. */
2502 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2503 hb_change, pmtud_change,
2509 /* If changes are for association, also apply parameters to each
2512 if (!trans && asoc) {
2513 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2515 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2516 hb_change, pmtud_change,
2525 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2527 * This option will effect the way delayed acks are performed. This
2528 * option allows you to get or set the delayed ack time, in
2529 * milliseconds. It also allows changing the delayed ack frequency.
2530 * Changing the frequency to 1 disables the delayed sack algorithm. If
2531 * the assoc_id is 0, then this sets or gets the endpoints default
2532 * values. If the assoc_id field is non-zero, then the set or get
2533 * effects the specified association for the one to many model (the
2534 * assoc_id field is ignored by the one to one model). Note that if
2535 * sack_delay or sack_freq are 0 when setting this option, then the
2536 * current values will remain unchanged.
2538 * struct sctp_sack_info {
2539 * sctp_assoc_t sack_assoc_id;
2540 * uint32_t sack_delay;
2541 * uint32_t sack_freq;
2544 * sack_assoc_id - This parameter, indicates which association the user
2545 * is performing an action upon. Note that if this field's value is
2546 * zero then the endpoints default value is changed (effecting future
2547 * associations only).
2549 * sack_delay - This parameter contains the number of milliseconds that
2550 * the user is requesting the delayed ACK timer be set to. Note that
2551 * this value is defined in the standard to be between 200 and 500
2554 * sack_freq - This parameter contains the number of packets that must
2555 * be received before a sack is sent without waiting for the delay
2556 * timer to expire. The default value for this is 2, setting this
2557 * value to 1 will disable the delayed sack algorithm.
2560 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2561 char __user *optval, unsigned int optlen)
2563 struct sctp_sack_info params;
2564 struct sctp_transport *trans = NULL;
2565 struct sctp_association *asoc = NULL;
2566 struct sctp_sock *sp = sctp_sk(sk);
2568 if (optlen == sizeof(struct sctp_sack_info)) {
2569 if (copy_from_user(¶ms, optval, optlen))
2572 if (params.sack_delay == 0 && params.sack_freq == 0)
2574 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2575 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2576 pr_warn("Use struct sctp_sack_info instead\n");
2577 if (copy_from_user(¶ms, optval, optlen))
2580 if (params.sack_delay == 0)
2581 params.sack_freq = 1;
2583 params.sack_freq = 0;
2587 /* Validate value parameter. */
2588 if (params.sack_delay > 500)
2591 /* Get association, if sack_assoc_id != 0 and the socket is a one
2592 * to many style socket, and an association was not found, then
2593 * the id was invalid.
2595 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2596 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2599 if (params.sack_delay) {
2602 msecs_to_jiffies(params.sack_delay);
2604 (asoc->param_flags & ~SPP_SACKDELAY) |
2605 SPP_SACKDELAY_ENABLE;
2607 sp->sackdelay = params.sack_delay;
2609 (sp->param_flags & ~SPP_SACKDELAY) |
2610 SPP_SACKDELAY_ENABLE;
2614 if (params.sack_freq == 1) {
2617 (asoc->param_flags & ~SPP_SACKDELAY) |
2618 SPP_SACKDELAY_DISABLE;
2621 (sp->param_flags & ~SPP_SACKDELAY) |
2622 SPP_SACKDELAY_DISABLE;
2624 } else if (params.sack_freq > 1) {
2626 asoc->sackfreq = params.sack_freq;
2628 (asoc->param_flags & ~SPP_SACKDELAY) |
2629 SPP_SACKDELAY_ENABLE;
2631 sp->sackfreq = params.sack_freq;
2633 (sp->param_flags & ~SPP_SACKDELAY) |
2634 SPP_SACKDELAY_ENABLE;
2638 /* If change is for association, also apply to each transport. */
2640 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2642 if (params.sack_delay) {
2644 msecs_to_jiffies(params.sack_delay);
2645 trans->param_flags =
2646 (trans->param_flags & ~SPP_SACKDELAY) |
2647 SPP_SACKDELAY_ENABLE;
2649 if (params.sack_freq == 1) {
2650 trans->param_flags =
2651 (trans->param_flags & ~SPP_SACKDELAY) |
2652 SPP_SACKDELAY_DISABLE;
2653 } else if (params.sack_freq > 1) {
2654 trans->sackfreq = params.sack_freq;
2655 trans->param_flags =
2656 (trans->param_flags & ~SPP_SACKDELAY) |
2657 SPP_SACKDELAY_ENABLE;
2665 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2667 * Applications can specify protocol parameters for the default association
2668 * initialization. The option name argument to setsockopt() and getsockopt()
2671 * Setting initialization parameters is effective only on an unconnected
2672 * socket (for UDP-style sockets only future associations are effected
2673 * by the change). With TCP-style sockets, this option is inherited by
2674 * sockets derived from a listener socket.
2676 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2678 struct sctp_initmsg sinit;
2679 struct sctp_sock *sp = sctp_sk(sk);
2681 if (optlen != sizeof(struct sctp_initmsg))
2683 if (copy_from_user(&sinit, optval, optlen))
2686 if (sinit.sinit_num_ostreams)
2687 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2688 if (sinit.sinit_max_instreams)
2689 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2690 if (sinit.sinit_max_attempts)
2691 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2692 if (sinit.sinit_max_init_timeo)
2693 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2699 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2701 * Applications that wish to use the sendto() system call may wish to
2702 * specify a default set of parameters that would normally be supplied
2703 * through the inclusion of ancillary data. This socket option allows
2704 * such an application to set the default sctp_sndrcvinfo structure.
2705 * The application that wishes to use this socket option simply passes
2706 * in to this call the sctp_sndrcvinfo structure defined in Section
2707 * 5.2.2) The input parameters accepted by this call include
2708 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2709 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2710 * to this call if the caller is using the UDP model.
2712 static int sctp_setsockopt_default_send_param(struct sock *sk,
2713 char __user *optval,
2714 unsigned int optlen)
2716 struct sctp_sndrcvinfo info;
2717 struct sctp_association *asoc;
2718 struct sctp_sock *sp = sctp_sk(sk);
2720 if (optlen != sizeof(struct sctp_sndrcvinfo))
2722 if (copy_from_user(&info, optval, optlen))
2725 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2726 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2730 asoc->default_stream = info.sinfo_stream;
2731 asoc->default_flags = info.sinfo_flags;
2732 asoc->default_ppid = info.sinfo_ppid;
2733 asoc->default_context = info.sinfo_context;
2734 asoc->default_timetolive = info.sinfo_timetolive;
2736 sp->default_stream = info.sinfo_stream;
2737 sp->default_flags = info.sinfo_flags;
2738 sp->default_ppid = info.sinfo_ppid;
2739 sp->default_context = info.sinfo_context;
2740 sp->default_timetolive = info.sinfo_timetolive;
2746 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2748 * Requests that the local SCTP stack use the enclosed peer address as
2749 * the association primary. The enclosed address must be one of the
2750 * association peer's addresses.
2752 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2753 unsigned int optlen)
2755 struct sctp_prim prim;
2756 struct sctp_transport *trans;
2758 if (optlen != sizeof(struct sctp_prim))
2761 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2764 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2768 sctp_assoc_set_primary(trans->asoc, trans);
2774 * 7.1.5 SCTP_NODELAY
2776 * Turn on/off any Nagle-like algorithm. This means that packets are
2777 * generally sent as soon as possible and no unnecessary delays are
2778 * introduced, at the cost of more packets in the network. Expects an
2779 * integer boolean flag.
2781 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2782 unsigned int optlen)
2786 if (optlen < sizeof(int))
2788 if (get_user(val, (int __user *)optval))
2791 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2797 * 7.1.1 SCTP_RTOINFO
2799 * The protocol parameters used to initialize and bound retransmission
2800 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2801 * and modify these parameters.
2802 * All parameters are time values, in milliseconds. A value of 0, when
2803 * modifying the parameters, indicates that the current value should not
2807 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2809 struct sctp_rtoinfo rtoinfo;
2810 struct sctp_association *asoc;
2812 if (optlen != sizeof (struct sctp_rtoinfo))
2815 if (copy_from_user(&rtoinfo, optval, optlen))
2818 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2820 /* Set the values to the specific association */
2821 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2825 if (rtoinfo.srto_initial != 0)
2827 msecs_to_jiffies(rtoinfo.srto_initial);
2828 if (rtoinfo.srto_max != 0)
2829 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2830 if (rtoinfo.srto_min != 0)
2831 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2833 /* If there is no association or the association-id = 0
2834 * set the values to the endpoint.
2836 struct sctp_sock *sp = sctp_sk(sk);
2838 if (rtoinfo.srto_initial != 0)
2839 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2840 if (rtoinfo.srto_max != 0)
2841 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2842 if (rtoinfo.srto_min != 0)
2843 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2851 * 7.1.2 SCTP_ASSOCINFO
2853 * This option is used to tune the maximum retransmission attempts
2854 * of the association.
2855 * Returns an error if the new association retransmission value is
2856 * greater than the sum of the retransmission value of the peer.
2857 * See [SCTP] for more information.
2860 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2863 struct sctp_assocparams assocparams;
2864 struct sctp_association *asoc;
2866 if (optlen != sizeof(struct sctp_assocparams))
2868 if (copy_from_user(&assocparams, optval, optlen))
2871 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2873 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2876 /* Set the values to the specific association */
2878 if (assocparams.sasoc_asocmaxrxt != 0) {
2881 struct sctp_transport *peer_addr;
2883 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2885 path_sum += peer_addr->pathmaxrxt;
2889 /* Only validate asocmaxrxt if we have more than
2890 * one path/transport. We do this because path
2891 * retransmissions are only counted when we have more
2895 assocparams.sasoc_asocmaxrxt > path_sum)
2898 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2901 if (assocparams.sasoc_cookie_life != 0)
2902 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
2904 /* Set the values to the endpoint */
2905 struct sctp_sock *sp = sctp_sk(sk);
2907 if (assocparams.sasoc_asocmaxrxt != 0)
2908 sp->assocparams.sasoc_asocmaxrxt =
2909 assocparams.sasoc_asocmaxrxt;
2910 if (assocparams.sasoc_cookie_life != 0)
2911 sp->assocparams.sasoc_cookie_life =
2912 assocparams.sasoc_cookie_life;
2918 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2920 * This socket option is a boolean flag which turns on or off mapped V4
2921 * addresses. If this option is turned on and the socket is type
2922 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2923 * If this option is turned off, then no mapping will be done of V4
2924 * addresses and a user will receive both PF_INET6 and PF_INET type
2925 * addresses on the socket.
2927 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2930 struct sctp_sock *sp = sctp_sk(sk);
2932 if (optlen < sizeof(int))
2934 if (get_user(val, (int __user *)optval))
2945 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2946 * This option will get or set the maximum size to put in any outgoing
2947 * SCTP DATA chunk. If a message is larger than this size it will be
2948 * fragmented by SCTP into the specified size. Note that the underlying
2949 * SCTP implementation may fragment into smaller sized chunks when the
2950 * PMTU of the underlying association is smaller than the value set by
2951 * the user. The default value for this option is '0' which indicates
2952 * the user is NOT limiting fragmentation and only the PMTU will effect
2953 * SCTP's choice of DATA chunk size. Note also that values set larger
2954 * than the maximum size of an IP datagram will effectively let SCTP
2955 * control fragmentation (i.e. the same as setting this option to 0).
2957 * The following structure is used to access and modify this parameter:
2959 * struct sctp_assoc_value {
2960 * sctp_assoc_t assoc_id;
2961 * uint32_t assoc_value;
2964 * assoc_id: This parameter is ignored for one-to-one style sockets.
2965 * For one-to-many style sockets this parameter indicates which
2966 * association the user is performing an action upon. Note that if
2967 * this field's value is zero then the endpoints default value is
2968 * changed (effecting future associations only).
2969 * assoc_value: This parameter specifies the maximum size in bytes.
2971 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2973 struct sctp_assoc_value params;
2974 struct sctp_association *asoc;
2975 struct sctp_sock *sp = sctp_sk(sk);
2978 if (optlen == sizeof(int)) {
2979 pr_warn("Use of int in maxseg socket option deprecated\n");
2980 pr_warn("Use struct sctp_assoc_value instead\n");
2981 if (copy_from_user(&val, optval, optlen))
2983 params.assoc_id = 0;
2984 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2985 if (copy_from_user(¶ms, optval, optlen))
2987 val = params.assoc_value;
2991 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2994 asoc = sctp_id2assoc(sk, params.assoc_id);
2995 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3000 val = asoc->pathmtu;
3001 val -= sp->pf->af->net_header_len;
3002 val -= sizeof(struct sctphdr) +
3003 sizeof(struct sctp_data_chunk);
3005 asoc->user_frag = val;
3006 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3008 sp->user_frag = val;
3016 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3018 * Requests that the peer mark the enclosed address as the association
3019 * primary. The enclosed address must be one of the association's
3020 * locally bound addresses. The following structure is used to make a
3021 * set primary request:
3023 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3024 unsigned int optlen)
3026 struct net *net = sock_net(sk);
3027 struct sctp_sock *sp;
3028 struct sctp_association *asoc = NULL;
3029 struct sctp_setpeerprim prim;
3030 struct sctp_chunk *chunk;
3036 if (!net->sctp.addip_enable)
3039 if (optlen != sizeof(struct sctp_setpeerprim))
3042 if (copy_from_user(&prim, optval, optlen))
3045 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3049 if (!asoc->peer.asconf_capable)
3052 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3055 if (!sctp_state(asoc, ESTABLISHED))
3058 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3062 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3063 return -EADDRNOTAVAIL;
3065 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3066 return -EADDRNOTAVAIL;
3068 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3069 chunk = sctp_make_asconf_set_prim(asoc,
3070 (union sctp_addr *)&prim.sspp_addr);
3074 err = sctp_send_asconf(asoc, chunk);
3076 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3081 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3082 unsigned int optlen)
3084 struct sctp_setadaptation adaptation;
3086 if (optlen != sizeof(struct sctp_setadaptation))
3088 if (copy_from_user(&adaptation, optval, optlen))
3091 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3097 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3099 * The context field in the sctp_sndrcvinfo structure is normally only
3100 * used when a failed message is retrieved holding the value that was
3101 * sent down on the actual send call. This option allows the setting of
3102 * a default context on an association basis that will be received on
3103 * reading messages from the peer. This is especially helpful in the
3104 * one-2-many model for an application to keep some reference to an
3105 * internal state machine that is processing messages on the
3106 * association. Note that the setting of this value only effects
3107 * received messages from the peer and does not effect the value that is
3108 * saved with outbound messages.
3110 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3111 unsigned int optlen)
3113 struct sctp_assoc_value params;
3114 struct sctp_sock *sp;
3115 struct sctp_association *asoc;
3117 if (optlen != sizeof(struct sctp_assoc_value))
3119 if (copy_from_user(¶ms, optval, optlen))
3124 if (params.assoc_id != 0) {
3125 asoc = sctp_id2assoc(sk, params.assoc_id);
3128 asoc->default_rcv_context = params.assoc_value;
3130 sp->default_rcv_context = params.assoc_value;
3137 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3139 * This options will at a minimum specify if the implementation is doing
3140 * fragmented interleave. Fragmented interleave, for a one to many
3141 * socket, is when subsequent calls to receive a message may return
3142 * parts of messages from different associations. Some implementations
3143 * may allow you to turn this value on or off. If so, when turned off,
3144 * no fragment interleave will occur (which will cause a head of line
3145 * blocking amongst multiple associations sharing the same one to many
3146 * socket). When this option is turned on, then each receive call may
3147 * come from a different association (thus the user must receive data
3148 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3149 * association each receive belongs to.
3151 * This option takes a boolean value. A non-zero value indicates that
3152 * fragmented interleave is on. A value of zero indicates that
3153 * fragmented interleave is off.
3155 * Note that it is important that an implementation that allows this
3156 * option to be turned on, have it off by default. Otherwise an unaware
3157 * application using the one to many model may become confused and act
3160 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3161 char __user *optval,
3162 unsigned int optlen)
3166 if (optlen != sizeof(int))
3168 if (get_user(val, (int __user *)optval))
3171 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3177 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3178 * (SCTP_PARTIAL_DELIVERY_POINT)
3180 * This option will set or get the SCTP partial delivery point. This
3181 * point is the size of a message where the partial delivery API will be
3182 * invoked to help free up rwnd space for the peer. Setting this to a
3183 * lower value will cause partial deliveries to happen more often. The
3184 * calls argument is an integer that sets or gets the partial delivery
3185 * point. Note also that the call will fail if the user attempts to set
3186 * this value larger than the socket receive buffer size.
3188 * Note that any single message having a length smaller than or equal to
3189 * the SCTP partial delivery point will be delivered in one single read
3190 * call as long as the user provided buffer is large enough to hold the
3193 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3194 char __user *optval,
3195 unsigned int optlen)
3199 if (optlen != sizeof(u32))
3201 if (get_user(val, (int __user *)optval))
3204 /* Note: We double the receive buffer from what the user sets
3205 * it to be, also initial rwnd is based on rcvbuf/2.
3207 if (val > (sk->sk_rcvbuf >> 1))
3210 sctp_sk(sk)->pd_point = val;
3212 return 0; /* is this the right error code? */
3216 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3218 * This option will allow a user to change the maximum burst of packets
3219 * that can be emitted by this association. Note that the default value
3220 * is 4, and some implementations may restrict this setting so that it
3221 * can only be lowered.
3223 * NOTE: This text doesn't seem right. Do this on a socket basis with
3224 * future associations inheriting the socket value.
3226 static int sctp_setsockopt_maxburst(struct sock *sk,
3227 char __user *optval,
3228 unsigned int optlen)
3230 struct sctp_assoc_value params;
3231 struct sctp_sock *sp;
3232 struct sctp_association *asoc;
3236 if (optlen == sizeof(int)) {
3237 pr_warn("Use of int in max_burst socket option deprecated\n");
3238 pr_warn("Use struct sctp_assoc_value instead\n");
3239 if (copy_from_user(&val, optval, optlen))
3241 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3242 if (copy_from_user(¶ms, optval, optlen))
3244 val = params.assoc_value;
3245 assoc_id = params.assoc_id;
3251 if (assoc_id != 0) {
3252 asoc = sctp_id2assoc(sk, assoc_id);
3255 asoc->max_burst = val;
3257 sp->max_burst = val;
3263 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3265 * This set option adds a chunk type that the user is requesting to be
3266 * received only in an authenticated way. Changes to the list of chunks
3267 * will only effect future associations on the socket.
3269 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3270 char __user *optval,
3271 unsigned int optlen)
3273 struct net *net = sock_net(sk);
3274 struct sctp_authchunk val;
3276 if (!net->sctp.auth_enable)
3279 if (optlen != sizeof(struct sctp_authchunk))
3281 if (copy_from_user(&val, optval, optlen))
3284 switch (val.sauth_chunk) {
3286 case SCTP_CID_INIT_ACK:
3287 case SCTP_CID_SHUTDOWN_COMPLETE:
3292 /* add this chunk id to the endpoint */
3293 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3297 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3299 * This option gets or sets the list of HMAC algorithms that the local
3300 * endpoint requires the peer to use.
3302 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3303 char __user *optval,
3304 unsigned int optlen)
3306 struct net *net = sock_net(sk);
3307 struct sctp_hmacalgo *hmacs;
3311 if (!net->sctp.auth_enable)
3314 if (optlen < sizeof(struct sctp_hmacalgo))
3317 hmacs= memdup_user(optval, optlen);
3319 return PTR_ERR(hmacs);
3321 idents = hmacs->shmac_num_idents;
3322 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3323 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3328 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3335 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3337 * This option will set a shared secret key which is used to build an
3338 * association shared key.
3340 static int sctp_setsockopt_auth_key(struct sock *sk,
3341 char __user *optval,
3342 unsigned int optlen)
3344 struct net *net = sock_net(sk);
3345 struct sctp_authkey *authkey;
3346 struct sctp_association *asoc;
3349 if (!net->sctp.auth_enable)
3352 if (optlen <= sizeof(struct sctp_authkey))
3355 authkey= memdup_user(optval, optlen);
3356 if (IS_ERR(authkey))
3357 return PTR_ERR(authkey);
3359 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3364 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3365 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3370 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3377 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3379 * This option will get or set the active shared key to be used to build
3380 * the association shared key.
3382 static int sctp_setsockopt_active_key(struct sock *sk,
3383 char __user *optval,
3384 unsigned int optlen)
3386 struct net *net = sock_net(sk);
3387 struct sctp_authkeyid val;
3388 struct sctp_association *asoc;
3390 if (!net->sctp.auth_enable)
3393 if (optlen != sizeof(struct sctp_authkeyid))
3395 if (copy_from_user(&val, optval, optlen))
3398 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3399 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3402 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3403 val.scact_keynumber);
3407 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3409 * This set option will delete a shared secret key from use.
3411 static int sctp_setsockopt_del_key(struct sock *sk,
3412 char __user *optval,
3413 unsigned int optlen)
3415 struct net *net = sock_net(sk);
3416 struct sctp_authkeyid val;
3417 struct sctp_association *asoc;
3419 if (!net->sctp.auth_enable)
3422 if (optlen != sizeof(struct sctp_authkeyid))
3424 if (copy_from_user(&val, optval, optlen))
3427 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3428 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3431 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3432 val.scact_keynumber);
3437 * 8.1.23 SCTP_AUTO_ASCONF
3439 * This option will enable or disable the use of the automatic generation of
3440 * ASCONF chunks to add and delete addresses to an existing association. Note
3441 * that this option has two caveats namely: a) it only affects sockets that
3442 * are bound to all addresses available to the SCTP stack, and b) the system
3443 * administrator may have an overriding control that turns the ASCONF feature
3444 * off no matter what setting the socket option may have.
3445 * This option expects an integer boolean flag, where a non-zero value turns on
3446 * the option, and a zero value turns off the option.
3447 * Note. In this implementation, socket operation overrides default parameter
3448 * being set by sysctl as well as FreeBSD implementation
3450 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3451 unsigned int optlen)
3454 struct sctp_sock *sp = sctp_sk(sk);
3456 if (optlen < sizeof(int))
3458 if (get_user(val, (int __user *)optval))
3460 if (!sctp_is_ep_boundall(sk) && val)
3462 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3465 if (val == 0 && sp->do_auto_asconf) {
3466 list_del(&sp->auto_asconf_list);
3467 sp->do_auto_asconf = 0;
3468 } else if (val && !sp->do_auto_asconf) {
3469 list_add_tail(&sp->auto_asconf_list,
3470 &sock_net(sk)->sctp.auto_asconf_splist);
3471 sp->do_auto_asconf = 1;
3478 * SCTP_PEER_ADDR_THLDS
3480 * This option allows us to alter the partially failed threshold for one or all
3481 * transports in an association. See Section 6.1 of:
3482 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3484 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3485 char __user *optval,
3486 unsigned int optlen)
3488 struct sctp_paddrthlds val;
3489 struct sctp_transport *trans;
3490 struct sctp_association *asoc;
3492 if (optlen < sizeof(struct sctp_paddrthlds))
3494 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3495 sizeof(struct sctp_paddrthlds)))
3499 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3500 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3503 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3505 if (val.spt_pathmaxrxt)
3506 trans->pathmaxrxt = val.spt_pathmaxrxt;
3507 trans->pf_retrans = val.spt_pathpfthld;
3510 if (val.spt_pathmaxrxt)
3511 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3512 asoc->pf_retrans = val.spt_pathpfthld;
3514 trans = sctp_addr_id2transport(sk, &val.spt_address,
3519 if (val.spt_pathmaxrxt)
3520 trans->pathmaxrxt = val.spt_pathmaxrxt;
3521 trans->pf_retrans = val.spt_pathpfthld;
3527 /* API 6.2 setsockopt(), getsockopt()
3529 * Applications use setsockopt() and getsockopt() to set or retrieve
3530 * socket options. Socket options are used to change the default
3531 * behavior of sockets calls. They are described in Section 7.
3535 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3536 * int __user *optlen);
3537 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3540 * sd - the socket descript.
3541 * level - set to IPPROTO_SCTP for all SCTP options.
3542 * optname - the option name.
3543 * optval - the buffer to store the value of the option.
3544 * optlen - the size of the buffer.
3546 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3547 char __user *optval, unsigned int optlen)
3551 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3553 /* I can hardly begin to describe how wrong this is. This is
3554 * so broken as to be worse than useless. The API draft
3555 * REALLY is NOT helpful here... I am not convinced that the
3556 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3557 * are at all well-founded.
3559 if (level != SOL_SCTP) {
3560 struct sctp_af *af = sctp_sk(sk)->pf->af;
3561 retval = af->setsockopt(sk, level, optname, optval, optlen);
3568 case SCTP_SOCKOPT_BINDX_ADD:
3569 /* 'optlen' is the size of the addresses buffer. */
3570 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3571 optlen, SCTP_BINDX_ADD_ADDR);
3574 case SCTP_SOCKOPT_BINDX_REM:
3575 /* 'optlen' is the size of the addresses buffer. */
3576 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3577 optlen, SCTP_BINDX_REM_ADDR);
3580 case SCTP_SOCKOPT_CONNECTX_OLD:
3581 /* 'optlen' is the size of the addresses buffer. */
3582 retval = sctp_setsockopt_connectx_old(sk,
3583 (struct sockaddr __user *)optval,
3587 case SCTP_SOCKOPT_CONNECTX:
3588 /* 'optlen' is the size of the addresses buffer. */
3589 retval = sctp_setsockopt_connectx(sk,
3590 (struct sockaddr __user *)optval,
3594 case SCTP_DISABLE_FRAGMENTS:
3595 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3599 retval = sctp_setsockopt_events(sk, optval, optlen);
3602 case SCTP_AUTOCLOSE:
3603 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3606 case SCTP_PEER_ADDR_PARAMS:
3607 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3610 case SCTP_DELAYED_SACK:
3611 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3613 case SCTP_PARTIAL_DELIVERY_POINT:
3614 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3618 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3620 case SCTP_DEFAULT_SEND_PARAM:
3621 retval = sctp_setsockopt_default_send_param(sk, optval,
3624 case SCTP_PRIMARY_ADDR:
3625 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3627 case SCTP_SET_PEER_PRIMARY_ADDR:
3628 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3631 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3634 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3636 case SCTP_ASSOCINFO:
3637 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3639 case SCTP_I_WANT_MAPPED_V4_ADDR:
3640 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3643 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3645 case SCTP_ADAPTATION_LAYER:
3646 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3649 retval = sctp_setsockopt_context(sk, optval, optlen);
3651 case SCTP_FRAGMENT_INTERLEAVE:
3652 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3654 case SCTP_MAX_BURST:
3655 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3657 case SCTP_AUTH_CHUNK:
3658 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3660 case SCTP_HMAC_IDENT:
3661 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3664 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3666 case SCTP_AUTH_ACTIVE_KEY:
3667 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3669 case SCTP_AUTH_DELETE_KEY:
3670 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3672 case SCTP_AUTO_ASCONF:
3673 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3675 case SCTP_PEER_ADDR_THLDS:
3676 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3679 retval = -ENOPROTOOPT;
3683 sctp_release_sock(sk);
3689 /* API 3.1.6 connect() - UDP Style Syntax
3691 * An application may use the connect() call in the UDP model to initiate an
3692 * association without sending data.
3696 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3698 * sd: the socket descriptor to have a new association added to.
3700 * nam: the address structure (either struct sockaddr_in or struct
3701 * sockaddr_in6 defined in RFC2553 [7]).
3703 * len: the size of the address.
3705 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3713 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
3716 /* Validate addr_len before calling common connect/connectx routine. */
3717 af = sctp_get_af_specific(addr->sa_family);
3718 if (!af || addr_len < af->sockaddr_len) {
3721 /* Pass correct addr len to common routine (so it knows there
3722 * is only one address being passed.
3724 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3727 sctp_release_sock(sk);
3731 /* FIXME: Write comments. */
3732 static int sctp_disconnect(struct sock *sk, int flags)
3734 return -EOPNOTSUPP; /* STUB */
3737 /* 4.1.4 accept() - TCP Style Syntax
3739 * Applications use accept() call to remove an established SCTP
3740 * association from the accept queue of the endpoint. A new socket
3741 * descriptor will be returned from accept() to represent the newly
3742 * formed association.
3744 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3746 struct sctp_sock *sp;
3747 struct sctp_endpoint *ep;
3748 struct sock *newsk = NULL;
3749 struct sctp_association *asoc;
3758 if (!sctp_style(sk, TCP)) {
3759 error = -EOPNOTSUPP;
3763 if (!sctp_sstate(sk, LISTENING)) {
3768 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3770 error = sctp_wait_for_accept(sk, timeo);
3774 /* We treat the list of associations on the endpoint as the accept
3775 * queue and pick the first association on the list.
3777 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3779 newsk = sp->pf->create_accept_sk(sk, asoc);
3785 /* Populate the fields of the newsk from the oldsk and migrate the
3786 * asoc to the newsk.
3788 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3791 sctp_release_sock(sk);
3796 /* The SCTP ioctl handler. */
3797 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3804 * SEQPACKET-style sockets in LISTENING state are valid, for
3805 * SCTP, so only discard TCP-style sockets in LISTENING state.
3807 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3812 struct sk_buff *skb;
3813 unsigned int amount = 0;
3815 skb = skb_peek(&sk->sk_receive_queue);
3818 * We will only return the amount of this packet since
3819 * that is all that will be read.
3823 rc = put_user(amount, (int __user *)arg);
3831 sctp_release_sock(sk);
3835 /* This is the function which gets called during socket creation to
3836 * initialized the SCTP-specific portion of the sock.
3837 * The sock structure should already be zero-filled memory.
3839 static int sctp_init_sock(struct sock *sk)
3841 struct net *net = sock_net(sk);
3842 struct sctp_sock *sp;
3844 pr_debug("%s: sk:%p\n", __func__, sk);
3848 /* Initialize the SCTP per socket area. */
3849 switch (sk->sk_type) {
3850 case SOCK_SEQPACKET:
3851 sp->type = SCTP_SOCKET_UDP;
3854 sp->type = SCTP_SOCKET_TCP;
3857 return -ESOCKTNOSUPPORT;
3860 /* Initialize default send parameters. These parameters can be
3861 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3863 sp->default_stream = 0;
3864 sp->default_ppid = 0;
3865 sp->default_flags = 0;
3866 sp->default_context = 0;
3867 sp->default_timetolive = 0;
3869 sp->default_rcv_context = 0;
3870 sp->max_burst = net->sctp.max_burst;
3872 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
3874 /* Initialize default setup parameters. These parameters
3875 * can be modified with the SCTP_INITMSG socket option or
3876 * overridden by the SCTP_INIT CMSG.
3878 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3879 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3880 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
3881 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
3883 /* Initialize default RTO related parameters. These parameters can
3884 * be modified for with the SCTP_RTOINFO socket option.
3886 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
3887 sp->rtoinfo.srto_max = net->sctp.rto_max;
3888 sp->rtoinfo.srto_min = net->sctp.rto_min;
3890 /* Initialize default association related parameters. These parameters
3891 * can be modified with the SCTP_ASSOCINFO socket option.
3893 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
3894 sp->assocparams.sasoc_number_peer_destinations = 0;
3895 sp->assocparams.sasoc_peer_rwnd = 0;
3896 sp->assocparams.sasoc_local_rwnd = 0;
3897 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
3899 /* Initialize default event subscriptions. By default, all the
3902 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3904 /* Default Peer Address Parameters. These defaults can
3905 * be modified via SCTP_PEER_ADDR_PARAMS
3907 sp->hbinterval = net->sctp.hb_interval;
3908 sp->pathmaxrxt = net->sctp.max_retrans_path;
3909 sp->pathmtu = 0; // allow default discovery
3910 sp->sackdelay = net->sctp.sack_timeout;
3912 sp->param_flags = SPP_HB_ENABLE |
3914 SPP_SACKDELAY_ENABLE;
3916 /* If enabled no SCTP message fragmentation will be performed.
3917 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3919 sp->disable_fragments = 0;
3921 /* Enable Nagle algorithm by default. */
3924 /* Enable by default. */
3927 /* Auto-close idle associations after the configured
3928 * number of seconds. A value of 0 disables this
3929 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3930 * for UDP-style sockets only.
3934 /* User specified fragmentation limit. */
3937 sp->adaptation_ind = 0;
3939 sp->pf = sctp_get_pf_specific(sk->sk_family);
3941 /* Control variables for partial data delivery. */
3942 atomic_set(&sp->pd_mode, 0);
3943 skb_queue_head_init(&sp->pd_lobby);
3944 sp->frag_interleave = 0;
3946 /* Create a per socket endpoint structure. Even if we
3947 * change the data structure relationships, this may still
3948 * be useful for storing pre-connect address information.
3950 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
3956 sk->sk_destruct = sctp_destruct_sock;
3958 SCTP_DBG_OBJCNT_INC(sock);
3961 percpu_counter_inc(&sctp_sockets_allocated);
3962 sock_prot_inuse_add(net, sk->sk_prot, 1);
3963 if (net->sctp.default_auto_asconf) {
3964 list_add_tail(&sp->auto_asconf_list,
3965 &net->sctp.auto_asconf_splist);
3966 sp->do_auto_asconf = 1;
3968 sp->do_auto_asconf = 0;
3974 /* Cleanup any SCTP per socket resources. */
3975 static void sctp_destroy_sock(struct sock *sk)
3977 struct sctp_sock *sp;
3979 pr_debug("%s: sk:%p\n", __func__, sk);
3981 /* Release our hold on the endpoint. */
3983 /* This could happen during socket init, thus we bail out
3984 * early, since the rest of the below is not setup either.
3989 if (sp->do_auto_asconf) {
3990 sp->do_auto_asconf = 0;
3991 list_del(&sp->auto_asconf_list);
3993 sctp_endpoint_free(sp->ep);
3995 percpu_counter_dec(&sctp_sockets_allocated);
3996 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4000 /* Triggered when there are no references on the socket anymore */
4001 static void sctp_destruct_sock(struct sock *sk)
4003 struct sctp_sock *sp = sctp_sk(sk);
4005 /* Free up the HMAC transform. */
4006 crypto_free_hash(sp->hmac);
4008 inet_sock_destruct(sk);
4011 /* API 4.1.7 shutdown() - TCP Style Syntax
4012 * int shutdown(int socket, int how);
4014 * sd - the socket descriptor of the association to be closed.
4015 * how - Specifies the type of shutdown. The values are
4018 * Disables further receive operations. No SCTP
4019 * protocol action is taken.
4021 * Disables further send operations, and initiates
4022 * the SCTP shutdown sequence.
4024 * Disables further send and receive operations
4025 * and initiates the SCTP shutdown sequence.
4027 static void sctp_shutdown(struct sock *sk, int how)
4029 struct net *net = sock_net(sk);
4030 struct sctp_endpoint *ep;
4031 struct sctp_association *asoc;
4033 if (!sctp_style(sk, TCP))
4036 if (how & SEND_SHUTDOWN) {
4037 ep = sctp_sk(sk)->ep;
4038 if (!list_empty(&ep->asocs)) {
4039 asoc = list_entry(ep->asocs.next,
4040 struct sctp_association, asocs);
4041 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4046 /* 7.2.1 Association Status (SCTP_STATUS)
4048 * Applications can retrieve current status information about an
4049 * association, including association state, peer receiver window size,
4050 * number of unacked data chunks, and number of data chunks pending
4051 * receipt. This information is read-only.
4053 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4054 char __user *optval,
4057 struct sctp_status status;
4058 struct sctp_association *asoc = NULL;
4059 struct sctp_transport *transport;
4060 sctp_assoc_t associd;
4063 if (len < sizeof(status)) {
4068 len = sizeof(status);
4069 if (copy_from_user(&status, optval, len)) {
4074 associd = status.sstat_assoc_id;
4075 asoc = sctp_id2assoc(sk, associd);
4081 transport = asoc->peer.primary_path;
4083 status.sstat_assoc_id = sctp_assoc2id(asoc);
4084 status.sstat_state = asoc->state;
4085 status.sstat_rwnd = asoc->peer.rwnd;
4086 status.sstat_unackdata = asoc->unack_data;
4088 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4089 status.sstat_instrms = asoc->c.sinit_max_instreams;
4090 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4091 status.sstat_fragmentation_point = asoc->frag_point;
4092 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4093 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4094 transport->af_specific->sockaddr_len);
4095 /* Map ipv4 address into v4-mapped-on-v6 address. */
4096 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4097 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4098 status.sstat_primary.spinfo_state = transport->state;
4099 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4100 status.sstat_primary.spinfo_srtt = transport->srtt;
4101 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4102 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4104 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4105 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4107 if (put_user(len, optlen)) {
4112 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4113 __func__, len, status.sstat_state, status.sstat_rwnd,
4114 status.sstat_assoc_id);
4116 if (copy_to_user(optval, &status, len)) {
4126 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4128 * Applications can retrieve information about a specific peer address
4129 * of an association, including its reachability state, congestion
4130 * window, and retransmission timer values. This information is
4133 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4134 char __user *optval,
4137 struct sctp_paddrinfo pinfo;
4138 struct sctp_transport *transport;
4141 if (len < sizeof(pinfo)) {
4146 len = sizeof(pinfo);
4147 if (copy_from_user(&pinfo, optval, len)) {
4152 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4153 pinfo.spinfo_assoc_id);
4157 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4158 pinfo.spinfo_state = transport->state;
4159 pinfo.spinfo_cwnd = transport->cwnd;
4160 pinfo.spinfo_srtt = transport->srtt;
4161 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4162 pinfo.spinfo_mtu = transport->pathmtu;
4164 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4165 pinfo.spinfo_state = SCTP_ACTIVE;
4167 if (put_user(len, optlen)) {
4172 if (copy_to_user(optval, &pinfo, len)) {
4181 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4183 * This option is a on/off flag. If enabled no SCTP message
4184 * fragmentation will be performed. Instead if a message being sent
4185 * exceeds the current PMTU size, the message will NOT be sent and
4186 * instead a error will be indicated to the user.
4188 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4189 char __user *optval, int __user *optlen)
4193 if (len < sizeof(int))
4197 val = (sctp_sk(sk)->disable_fragments == 1);
4198 if (put_user(len, optlen))
4200 if (copy_to_user(optval, &val, len))
4205 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4207 * This socket option is used to specify various notifications and
4208 * ancillary data the user wishes to receive.
4210 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4215 if (len > sizeof(struct sctp_event_subscribe))
4216 len = sizeof(struct sctp_event_subscribe);
4217 if (put_user(len, optlen))
4219 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4224 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4226 * This socket option is applicable to the UDP-style socket only. When
4227 * set it will cause associations that are idle for more than the
4228 * specified number of seconds to automatically close. An association
4229 * being idle is defined an association that has NOT sent or received
4230 * user data. The special value of '0' indicates that no automatic
4231 * close of any associations should be performed. The option expects an
4232 * integer defining the number of seconds of idle time before an
4233 * association is closed.
4235 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4237 /* Applicable to UDP-style socket only */
4238 if (sctp_style(sk, TCP))
4240 if (len < sizeof(int))
4243 if (put_user(len, optlen))
4245 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4250 /* Helper routine to branch off an association to a new socket. */
4251 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4253 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4254 struct socket *sock;
4261 /* An association cannot be branched off from an already peeled-off
4262 * socket, nor is this supported for tcp style sockets.
4264 if (!sctp_style(sk, UDP))
4267 /* Create a new socket. */
4268 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4272 sctp_copy_sock(sock->sk, sk, asoc);
4274 /* Make peeled-off sockets more like 1-1 accepted sockets.
4275 * Set the daddr and initialize id to something more random
4277 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4278 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4280 /* Populate the fields of the newsk from the oldsk and migrate the
4281 * asoc to the newsk.
4283 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4289 EXPORT_SYMBOL(sctp_do_peeloff);
4291 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4293 sctp_peeloff_arg_t peeloff;
4294 struct socket *newsock;
4295 struct file *newfile;
4298 if (len < sizeof(sctp_peeloff_arg_t))
4300 len = sizeof(sctp_peeloff_arg_t);
4301 if (copy_from_user(&peeloff, optval, len))
4304 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4308 /* Map the socket to an unused fd that can be returned to the user. */
4309 retval = get_unused_fd_flags(0);
4311 sock_release(newsock);
4315 newfile = sock_alloc_file(newsock, 0, NULL);
4316 if (unlikely(IS_ERR(newfile))) {
4317 put_unused_fd(retval);
4318 sock_release(newsock);
4319 return PTR_ERR(newfile);
4322 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4325 /* Return the fd mapped to the new socket. */
4326 if (put_user(len, optlen)) {
4328 put_unused_fd(retval);
4331 peeloff.sd = retval;
4332 if (copy_to_user(optval, &peeloff, len)) {
4334 put_unused_fd(retval);
4337 fd_install(retval, newfile);
4342 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4344 * Applications can enable or disable heartbeats for any peer address of
4345 * an association, modify an address's heartbeat interval, force a
4346 * heartbeat to be sent immediately, and adjust the address's maximum
4347 * number of retransmissions sent before an address is considered
4348 * unreachable. The following structure is used to access and modify an
4349 * address's parameters:
4351 * struct sctp_paddrparams {
4352 * sctp_assoc_t spp_assoc_id;
4353 * struct sockaddr_storage spp_address;
4354 * uint32_t spp_hbinterval;
4355 * uint16_t spp_pathmaxrxt;
4356 * uint32_t spp_pathmtu;
4357 * uint32_t spp_sackdelay;
4358 * uint32_t spp_flags;
4361 * spp_assoc_id - (one-to-many style socket) This is filled in the
4362 * application, and identifies the association for
4364 * spp_address - This specifies which address is of interest.
4365 * spp_hbinterval - This contains the value of the heartbeat interval,
4366 * in milliseconds. If a value of zero
4367 * is present in this field then no changes are to
4368 * be made to this parameter.
4369 * spp_pathmaxrxt - This contains the maximum number of
4370 * retransmissions before this address shall be
4371 * considered unreachable. If a value of zero
4372 * is present in this field then no changes are to
4373 * be made to this parameter.
4374 * spp_pathmtu - When Path MTU discovery is disabled the value
4375 * specified here will be the "fixed" path mtu.
4376 * Note that if the spp_address field is empty
4377 * then all associations on this address will
4378 * have this fixed path mtu set upon them.
4380 * spp_sackdelay - When delayed sack is enabled, this value specifies
4381 * the number of milliseconds that sacks will be delayed
4382 * for. This value will apply to all addresses of an
4383 * association if the spp_address field is empty. Note
4384 * also, that if delayed sack is enabled and this
4385 * value is set to 0, no change is made to the last
4386 * recorded delayed sack timer value.
4388 * spp_flags - These flags are used to control various features
4389 * on an association. The flag field may contain
4390 * zero or more of the following options.
4392 * SPP_HB_ENABLE - Enable heartbeats on the
4393 * specified address. Note that if the address
4394 * field is empty all addresses for the association
4395 * have heartbeats enabled upon them.
4397 * SPP_HB_DISABLE - Disable heartbeats on the
4398 * speicifed address. Note that if the address
4399 * field is empty all addresses for the association
4400 * will have their heartbeats disabled. Note also
4401 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4402 * mutually exclusive, only one of these two should
4403 * be specified. Enabling both fields will have
4404 * undetermined results.
4406 * SPP_HB_DEMAND - Request a user initiated heartbeat
4407 * to be made immediately.
4409 * SPP_PMTUD_ENABLE - This field will enable PMTU
4410 * discovery upon the specified address. Note that
4411 * if the address feild is empty then all addresses
4412 * on the association are effected.
4414 * SPP_PMTUD_DISABLE - This field will disable PMTU
4415 * discovery upon the specified address. Note that
4416 * if the address feild is empty then all addresses
4417 * on the association are effected. Not also that
4418 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4419 * exclusive. Enabling both will have undetermined
4422 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4423 * on delayed sack. The time specified in spp_sackdelay
4424 * is used to specify the sack delay for this address. Note
4425 * that if spp_address is empty then all addresses will
4426 * enable delayed sack and take on the sack delay
4427 * value specified in spp_sackdelay.
4428 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4429 * off delayed sack. If the spp_address field is blank then
4430 * delayed sack is disabled for the entire association. Note
4431 * also that this field is mutually exclusive to
4432 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4435 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4436 char __user *optval, int __user *optlen)
4438 struct sctp_paddrparams params;
4439 struct sctp_transport *trans = NULL;
4440 struct sctp_association *asoc = NULL;
4441 struct sctp_sock *sp = sctp_sk(sk);
4443 if (len < sizeof(struct sctp_paddrparams))
4445 len = sizeof(struct sctp_paddrparams);
4446 if (copy_from_user(¶ms, optval, len))
4449 /* If an address other than INADDR_ANY is specified, and
4450 * no transport is found, then the request is invalid.
4452 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
4453 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4454 params.spp_assoc_id);
4456 pr_debug("%s: failed no transport\n", __func__);
4461 /* Get association, if assoc_id != 0 and the socket is a one
4462 * to many style socket, and an association was not found, then
4463 * the id was invalid.
4465 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4466 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4467 pr_debug("%s: failed no association\n", __func__);
4472 /* Fetch transport values. */
4473 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4474 params.spp_pathmtu = trans->pathmtu;
4475 params.spp_pathmaxrxt = trans->pathmaxrxt;
4476 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4478 /*draft-11 doesn't say what to return in spp_flags*/
4479 params.spp_flags = trans->param_flags;
4481 /* Fetch association values. */
4482 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4483 params.spp_pathmtu = asoc->pathmtu;
4484 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4485 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4487 /*draft-11 doesn't say what to return in spp_flags*/
4488 params.spp_flags = asoc->param_flags;
4490 /* Fetch socket values. */
4491 params.spp_hbinterval = sp->hbinterval;
4492 params.spp_pathmtu = sp->pathmtu;
4493 params.spp_sackdelay = sp->sackdelay;
4494 params.spp_pathmaxrxt = sp->pathmaxrxt;
4496 /*draft-11 doesn't say what to return in spp_flags*/
4497 params.spp_flags = sp->param_flags;
4500 if (copy_to_user(optval, ¶ms, len))
4503 if (put_user(len, optlen))
4510 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4512 * This option will effect the way delayed acks are performed. This
4513 * option allows you to get or set the delayed ack time, in
4514 * milliseconds. It also allows changing the delayed ack frequency.
4515 * Changing the frequency to 1 disables the delayed sack algorithm. If
4516 * the assoc_id is 0, then this sets or gets the endpoints default
4517 * values. If the assoc_id field is non-zero, then the set or get
4518 * effects the specified association for the one to many model (the
4519 * assoc_id field is ignored by the one to one model). Note that if
4520 * sack_delay or sack_freq are 0 when setting this option, then the
4521 * current values will remain unchanged.
4523 * struct sctp_sack_info {
4524 * sctp_assoc_t sack_assoc_id;
4525 * uint32_t sack_delay;
4526 * uint32_t sack_freq;
4529 * sack_assoc_id - This parameter, indicates which association the user
4530 * is performing an action upon. Note that if this field's value is
4531 * zero then the endpoints default value is changed (effecting future
4532 * associations only).
4534 * sack_delay - This parameter contains the number of milliseconds that
4535 * the user is requesting the delayed ACK timer be set to. Note that
4536 * this value is defined in the standard to be between 200 and 500
4539 * sack_freq - This parameter contains the number of packets that must
4540 * be received before a sack is sent without waiting for the delay
4541 * timer to expire. The default value for this is 2, setting this
4542 * value to 1 will disable the delayed sack algorithm.
4544 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4545 char __user *optval,
4548 struct sctp_sack_info params;
4549 struct sctp_association *asoc = NULL;
4550 struct sctp_sock *sp = sctp_sk(sk);
4552 if (len >= sizeof(struct sctp_sack_info)) {
4553 len = sizeof(struct sctp_sack_info);
4555 if (copy_from_user(¶ms, optval, len))
4557 } else if (len == sizeof(struct sctp_assoc_value)) {
4558 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4559 pr_warn("Use struct sctp_sack_info instead\n");
4560 if (copy_from_user(¶ms, optval, len))
4565 /* Get association, if sack_assoc_id != 0 and the socket is a one
4566 * to many style socket, and an association was not found, then
4567 * the id was invalid.
4569 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4570 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4574 /* Fetch association values. */
4575 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4576 params.sack_delay = jiffies_to_msecs(
4578 params.sack_freq = asoc->sackfreq;
4581 params.sack_delay = 0;
4582 params.sack_freq = 1;
4585 /* Fetch socket values. */
4586 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4587 params.sack_delay = sp->sackdelay;
4588 params.sack_freq = sp->sackfreq;
4590 params.sack_delay = 0;
4591 params.sack_freq = 1;
4595 if (copy_to_user(optval, ¶ms, len))
4598 if (put_user(len, optlen))
4604 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4606 * Applications can specify protocol parameters for the default association
4607 * initialization. The option name argument to setsockopt() and getsockopt()
4610 * Setting initialization parameters is effective only on an unconnected
4611 * socket (for UDP-style sockets only future associations are effected
4612 * by the change). With TCP-style sockets, this option is inherited by
4613 * sockets derived from a listener socket.
4615 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4617 if (len < sizeof(struct sctp_initmsg))
4619 len = sizeof(struct sctp_initmsg);
4620 if (put_user(len, optlen))
4622 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4628 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4629 char __user *optval, int __user *optlen)
4631 struct sctp_association *asoc;
4633 struct sctp_getaddrs getaddrs;
4634 struct sctp_transport *from;
4636 union sctp_addr temp;
4637 struct sctp_sock *sp = sctp_sk(sk);
4642 if (len < sizeof(struct sctp_getaddrs))
4645 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4648 /* For UDP-style sockets, id specifies the association to query. */
4649 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4653 to = optval + offsetof(struct sctp_getaddrs,addrs);
4654 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4656 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4658 memcpy(&temp, &from->ipaddr, sizeof(temp));
4659 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4660 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4661 if (space_left < addrlen)
4663 if (copy_to_user(to, &temp, addrlen))
4667 space_left -= addrlen;
4670 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4672 bytes_copied = ((char __user *)to) - optval;
4673 if (put_user(bytes_copied, optlen))
4679 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4680 size_t space_left, int *bytes_copied)
4682 struct sctp_sockaddr_entry *addr;
4683 union sctp_addr temp;
4686 struct net *net = sock_net(sk);
4689 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4693 if ((PF_INET == sk->sk_family) &&
4694 (AF_INET6 == addr->a.sa.sa_family))
4696 if ((PF_INET6 == sk->sk_family) &&
4697 inet_v6_ipv6only(sk) &&
4698 (AF_INET == addr->a.sa.sa_family))
4700 memcpy(&temp, &addr->a, sizeof(temp));
4701 if (!temp.v4.sin_port)
4702 temp.v4.sin_port = htons(port);
4704 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4706 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4707 if (space_left < addrlen) {
4711 memcpy(to, &temp, addrlen);
4715 space_left -= addrlen;
4716 *bytes_copied += addrlen;
4724 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4725 char __user *optval, int __user *optlen)
4727 struct sctp_bind_addr *bp;
4728 struct sctp_association *asoc;
4730 struct sctp_getaddrs getaddrs;
4731 struct sctp_sockaddr_entry *addr;
4733 union sctp_addr temp;
4734 struct sctp_sock *sp = sctp_sk(sk);
4738 int bytes_copied = 0;
4742 if (len < sizeof(struct sctp_getaddrs))
4745 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4749 * For UDP-style sockets, id specifies the association to query.
4750 * If the id field is set to the value '0' then the locally bound
4751 * addresses are returned without regard to any particular
4754 if (0 == getaddrs.assoc_id) {
4755 bp = &sctp_sk(sk)->ep->base.bind_addr;
4757 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4760 bp = &asoc->base.bind_addr;
4763 to = optval + offsetof(struct sctp_getaddrs,addrs);
4764 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4766 addrs = kmalloc(space_left, GFP_KERNEL);
4770 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4771 * addresses from the global local address list.
4773 if (sctp_list_single_entry(&bp->address_list)) {
4774 addr = list_entry(bp->address_list.next,
4775 struct sctp_sockaddr_entry, list);
4776 if (sctp_is_any(sk, &addr->a)) {
4777 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4778 space_left, &bytes_copied);
4788 /* Protection on the bound address list is not needed since
4789 * in the socket option context we hold a socket lock and
4790 * thus the bound address list can't change.
4792 list_for_each_entry(addr, &bp->address_list, list) {
4793 memcpy(&temp, &addr->a, sizeof(temp));
4794 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4795 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4796 if (space_left < addrlen) {
4797 err = -ENOMEM; /*fixme: right error?*/
4800 memcpy(buf, &temp, addrlen);
4802 bytes_copied += addrlen;
4804 space_left -= addrlen;
4808 if (copy_to_user(to, addrs, bytes_copied)) {
4812 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4816 if (put_user(bytes_copied, optlen))
4823 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4825 * Requests that the local SCTP stack use the enclosed peer address as
4826 * the association primary. The enclosed address must be one of the
4827 * association peer's addresses.
4829 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4830 char __user *optval, int __user *optlen)
4832 struct sctp_prim prim;
4833 struct sctp_association *asoc;
4834 struct sctp_sock *sp = sctp_sk(sk);
4836 if (len < sizeof(struct sctp_prim))
4839 len = sizeof(struct sctp_prim);
4841 if (copy_from_user(&prim, optval, len))
4844 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4848 if (!asoc->peer.primary_path)
4851 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4852 asoc->peer.primary_path->af_specific->sockaddr_len);
4854 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4855 (union sctp_addr *)&prim.ssp_addr);
4857 if (put_user(len, optlen))
4859 if (copy_to_user(optval, &prim, len))
4866 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4868 * Requests that the local endpoint set the specified Adaptation Layer
4869 * Indication parameter for all future INIT and INIT-ACK exchanges.
4871 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4872 char __user *optval, int __user *optlen)
4874 struct sctp_setadaptation adaptation;
4876 if (len < sizeof(struct sctp_setadaptation))
4879 len = sizeof(struct sctp_setadaptation);
4881 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4883 if (put_user(len, optlen))
4885 if (copy_to_user(optval, &adaptation, len))
4893 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4895 * Applications that wish to use the sendto() system call may wish to
4896 * specify a default set of parameters that would normally be supplied
4897 * through the inclusion of ancillary data. This socket option allows
4898 * such an application to set the default sctp_sndrcvinfo structure.
4901 * The application that wishes to use this socket option simply passes
4902 * in to this call the sctp_sndrcvinfo structure defined in Section
4903 * 5.2.2) The input parameters accepted by this call include
4904 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4905 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4906 * to this call if the caller is using the UDP model.
4908 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4910 static int sctp_getsockopt_default_send_param(struct sock *sk,
4911 int len, char __user *optval,
4914 struct sctp_sndrcvinfo info;
4915 struct sctp_association *asoc;
4916 struct sctp_sock *sp = sctp_sk(sk);
4918 if (len < sizeof(struct sctp_sndrcvinfo))
4921 len = sizeof(struct sctp_sndrcvinfo);
4923 if (copy_from_user(&info, optval, len))
4926 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4927 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4931 info.sinfo_stream = asoc->default_stream;
4932 info.sinfo_flags = asoc->default_flags;
4933 info.sinfo_ppid = asoc->default_ppid;
4934 info.sinfo_context = asoc->default_context;
4935 info.sinfo_timetolive = asoc->default_timetolive;
4937 info.sinfo_stream = sp->default_stream;
4938 info.sinfo_flags = sp->default_flags;
4939 info.sinfo_ppid = sp->default_ppid;
4940 info.sinfo_context = sp->default_context;
4941 info.sinfo_timetolive = sp->default_timetolive;
4944 if (put_user(len, optlen))
4946 if (copy_to_user(optval, &info, len))
4954 * 7.1.5 SCTP_NODELAY
4956 * Turn on/off any Nagle-like algorithm. This means that packets are
4957 * generally sent as soon as possible and no unnecessary delays are
4958 * introduced, at the cost of more packets in the network. Expects an
4959 * integer boolean flag.
4962 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4963 char __user *optval, int __user *optlen)
4967 if (len < sizeof(int))
4971 val = (sctp_sk(sk)->nodelay == 1);
4972 if (put_user(len, optlen))
4974 if (copy_to_user(optval, &val, len))
4981 * 7.1.1 SCTP_RTOINFO
4983 * The protocol parameters used to initialize and bound retransmission
4984 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4985 * and modify these parameters.
4986 * All parameters are time values, in milliseconds. A value of 0, when
4987 * modifying the parameters, indicates that the current value should not
4991 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4992 char __user *optval,
4993 int __user *optlen) {
4994 struct sctp_rtoinfo rtoinfo;
4995 struct sctp_association *asoc;
4997 if (len < sizeof (struct sctp_rtoinfo))
5000 len = sizeof(struct sctp_rtoinfo);
5002 if (copy_from_user(&rtoinfo, optval, len))
5005 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5007 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5010 /* Values corresponding to the specific association. */
5012 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5013 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5014 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5016 /* Values corresponding to the endpoint. */
5017 struct sctp_sock *sp = sctp_sk(sk);
5019 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5020 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5021 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5024 if (put_user(len, optlen))
5027 if (copy_to_user(optval, &rtoinfo, len))
5035 * 7.1.2 SCTP_ASSOCINFO
5037 * This option is used to tune the maximum retransmission attempts
5038 * of the association.
5039 * Returns an error if the new association retransmission value is
5040 * greater than the sum of the retransmission value of the peer.
5041 * See [SCTP] for more information.
5044 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5045 char __user *optval,
5049 struct sctp_assocparams assocparams;
5050 struct sctp_association *asoc;
5051 struct list_head *pos;
5054 if (len < sizeof (struct sctp_assocparams))
5057 len = sizeof(struct sctp_assocparams);
5059 if (copy_from_user(&assocparams, optval, len))
5062 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5064 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5067 /* Values correspoinding to the specific association */
5069 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5070 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5071 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5072 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5074 list_for_each(pos, &asoc->peer.transport_addr_list) {
5078 assocparams.sasoc_number_peer_destinations = cnt;
5080 /* Values corresponding to the endpoint */
5081 struct sctp_sock *sp = sctp_sk(sk);
5083 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5084 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5085 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5086 assocparams.sasoc_cookie_life =
5087 sp->assocparams.sasoc_cookie_life;
5088 assocparams.sasoc_number_peer_destinations =
5090 sasoc_number_peer_destinations;
5093 if (put_user(len, optlen))
5096 if (copy_to_user(optval, &assocparams, len))
5103 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5105 * This socket option is a boolean flag which turns on or off mapped V4
5106 * addresses. If this option is turned on and the socket is type
5107 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5108 * If this option is turned off, then no mapping will be done of V4
5109 * addresses and a user will receive both PF_INET6 and PF_INET type
5110 * addresses on the socket.
5112 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5113 char __user *optval, int __user *optlen)
5116 struct sctp_sock *sp = sctp_sk(sk);
5118 if (len < sizeof(int))
5123 if (put_user(len, optlen))
5125 if (copy_to_user(optval, &val, len))
5132 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5133 * (chapter and verse is quoted at sctp_setsockopt_context())
5135 static int sctp_getsockopt_context(struct sock *sk, int len,
5136 char __user *optval, int __user *optlen)
5138 struct sctp_assoc_value params;
5139 struct sctp_sock *sp;
5140 struct sctp_association *asoc;
5142 if (len < sizeof(struct sctp_assoc_value))
5145 len = sizeof(struct sctp_assoc_value);
5147 if (copy_from_user(¶ms, optval, len))
5152 if (params.assoc_id != 0) {
5153 asoc = sctp_id2assoc(sk, params.assoc_id);
5156 params.assoc_value = asoc->default_rcv_context;
5158 params.assoc_value = sp->default_rcv_context;
5161 if (put_user(len, optlen))
5163 if (copy_to_user(optval, ¶ms, len))
5170 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5171 * This option will get or set the maximum size to put in any outgoing
5172 * SCTP DATA chunk. If a message is larger than this size it will be
5173 * fragmented by SCTP into the specified size. Note that the underlying
5174 * SCTP implementation may fragment into smaller sized chunks when the
5175 * PMTU of the underlying association is smaller than the value set by
5176 * the user. The default value for this option is '0' which indicates
5177 * the user is NOT limiting fragmentation and only the PMTU will effect
5178 * SCTP's choice of DATA chunk size. Note also that values set larger
5179 * than the maximum size of an IP datagram will effectively let SCTP
5180 * control fragmentation (i.e. the same as setting this option to 0).
5182 * The following structure is used to access and modify this parameter:
5184 * struct sctp_assoc_value {
5185 * sctp_assoc_t assoc_id;
5186 * uint32_t assoc_value;
5189 * assoc_id: This parameter is ignored for one-to-one style sockets.
5190 * For one-to-many style sockets this parameter indicates which
5191 * association the user is performing an action upon. Note that if
5192 * this field's value is zero then the endpoints default value is
5193 * changed (effecting future associations only).
5194 * assoc_value: This parameter specifies the maximum size in bytes.
5196 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5197 char __user *optval, int __user *optlen)
5199 struct sctp_assoc_value params;
5200 struct sctp_association *asoc;
5202 if (len == sizeof(int)) {
5203 pr_warn("Use of int in maxseg socket option deprecated\n");
5204 pr_warn("Use struct sctp_assoc_value instead\n");
5205 params.assoc_id = 0;
5206 } else if (len >= sizeof(struct sctp_assoc_value)) {
5207 len = sizeof(struct sctp_assoc_value);
5208 if (copy_from_user(¶ms, optval, sizeof(params)))
5213 asoc = sctp_id2assoc(sk, params.assoc_id);
5214 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5218 params.assoc_value = asoc->frag_point;
5220 params.assoc_value = sctp_sk(sk)->user_frag;
5222 if (put_user(len, optlen))
5224 if (len == sizeof(int)) {
5225 if (copy_to_user(optval, ¶ms.assoc_value, len))
5228 if (copy_to_user(optval, ¶ms, len))
5236 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5237 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5239 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5240 char __user *optval, int __user *optlen)
5244 if (len < sizeof(int))
5249 val = sctp_sk(sk)->frag_interleave;
5250 if (put_user(len, optlen))
5252 if (copy_to_user(optval, &val, len))
5259 * 7.1.25. Set or Get the sctp partial delivery point
5260 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5262 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5263 char __user *optval,
5268 if (len < sizeof(u32))
5273 val = sctp_sk(sk)->pd_point;
5274 if (put_user(len, optlen))
5276 if (copy_to_user(optval, &val, len))
5283 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5284 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5286 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5287 char __user *optval,
5290 struct sctp_assoc_value params;
5291 struct sctp_sock *sp;
5292 struct sctp_association *asoc;
5294 if (len == sizeof(int)) {
5295 pr_warn("Use of int in max_burst socket option deprecated\n");
5296 pr_warn("Use struct sctp_assoc_value instead\n");
5297 params.assoc_id = 0;
5298 } else if (len >= sizeof(struct sctp_assoc_value)) {
5299 len = sizeof(struct sctp_assoc_value);
5300 if (copy_from_user(¶ms, optval, len))
5307 if (params.assoc_id != 0) {
5308 asoc = sctp_id2assoc(sk, params.assoc_id);
5311 params.assoc_value = asoc->max_burst;
5313 params.assoc_value = sp->max_burst;
5315 if (len == sizeof(int)) {
5316 if (copy_to_user(optval, ¶ms.assoc_value, len))
5319 if (copy_to_user(optval, ¶ms, len))
5327 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5328 char __user *optval, int __user *optlen)
5330 struct net *net = sock_net(sk);
5331 struct sctp_hmacalgo __user *p = (void __user *)optval;
5332 struct sctp_hmac_algo_param *hmacs;
5336 if (!net->sctp.auth_enable)
5339 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5340 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5342 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5345 len = sizeof(struct sctp_hmacalgo) + data_len;
5346 num_idents = data_len / sizeof(u16);
5348 if (put_user(len, optlen))
5350 if (put_user(num_idents, &p->shmac_num_idents))
5352 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5357 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5358 char __user *optval, int __user *optlen)
5360 struct net *net = sock_net(sk);
5361 struct sctp_authkeyid val;
5362 struct sctp_association *asoc;
5364 if (!net->sctp.auth_enable)
5367 if (len < sizeof(struct sctp_authkeyid))
5369 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5372 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5373 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5377 val.scact_keynumber = asoc->active_key_id;
5379 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5381 len = sizeof(struct sctp_authkeyid);
5382 if (put_user(len, optlen))
5384 if (copy_to_user(optval, &val, len))
5390 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5391 char __user *optval, int __user *optlen)
5393 struct net *net = sock_net(sk);
5394 struct sctp_authchunks __user *p = (void __user *)optval;
5395 struct sctp_authchunks val;
5396 struct sctp_association *asoc;
5397 struct sctp_chunks_param *ch;
5401 if (!net->sctp.auth_enable)
5404 if (len < sizeof(struct sctp_authchunks))
5407 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5410 to = p->gauth_chunks;
5411 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5415 ch = asoc->peer.peer_chunks;
5419 /* See if the user provided enough room for all the data */
5420 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5421 if (len < num_chunks)
5424 if (copy_to_user(to, ch->chunks, num_chunks))
5427 len = sizeof(struct sctp_authchunks) + num_chunks;
5428 if (put_user(len, optlen)) return -EFAULT;
5429 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5434 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5435 char __user *optval, int __user *optlen)
5437 struct net *net = sock_net(sk);
5438 struct sctp_authchunks __user *p = (void __user *)optval;
5439 struct sctp_authchunks val;
5440 struct sctp_association *asoc;
5441 struct sctp_chunks_param *ch;
5445 if (!net->sctp.auth_enable)
5448 if (len < sizeof(struct sctp_authchunks))
5451 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5454 to = p->gauth_chunks;
5455 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5456 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5460 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5462 ch = sctp_sk(sk)->ep->auth_chunk_list;
5467 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5468 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5471 if (copy_to_user(to, ch->chunks, num_chunks))
5474 len = sizeof(struct sctp_authchunks) + num_chunks;
5475 if (put_user(len, optlen))
5477 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5484 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5485 * This option gets the current number of associations that are attached
5486 * to a one-to-many style socket. The option value is an uint32_t.
5488 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5489 char __user *optval, int __user *optlen)
5491 struct sctp_sock *sp = sctp_sk(sk);
5492 struct sctp_association *asoc;
5495 if (sctp_style(sk, TCP))
5498 if (len < sizeof(u32))
5503 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5507 if (put_user(len, optlen))
5509 if (copy_to_user(optval, &val, len))
5516 * 8.1.23 SCTP_AUTO_ASCONF
5517 * See the corresponding setsockopt entry as description
5519 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5520 char __user *optval, int __user *optlen)
5524 if (len < sizeof(int))
5528 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5530 if (put_user(len, optlen))
5532 if (copy_to_user(optval, &val, len))
5538 * 8.2.6. Get the Current Identifiers of Associations
5539 * (SCTP_GET_ASSOC_ID_LIST)
5541 * This option gets the current list of SCTP association identifiers of
5542 * the SCTP associations handled by a one-to-many style socket.
5544 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5545 char __user *optval, int __user *optlen)
5547 struct sctp_sock *sp = sctp_sk(sk);
5548 struct sctp_association *asoc;
5549 struct sctp_assoc_ids *ids;
5552 if (sctp_style(sk, TCP))
5555 if (len < sizeof(struct sctp_assoc_ids))
5558 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5562 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5565 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5567 ids = kmalloc(len, GFP_KERNEL);
5571 ids->gaids_number_of_ids = num;
5573 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5574 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5577 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5587 * SCTP_PEER_ADDR_THLDS
5589 * This option allows us to fetch the partially failed threshold for one or all
5590 * transports in an association. See Section 6.1 of:
5591 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5593 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5594 char __user *optval,
5598 struct sctp_paddrthlds val;
5599 struct sctp_transport *trans;
5600 struct sctp_association *asoc;
5602 if (len < sizeof(struct sctp_paddrthlds))
5604 len = sizeof(struct sctp_paddrthlds);
5605 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5608 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5609 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5613 val.spt_pathpfthld = asoc->pf_retrans;
5614 val.spt_pathmaxrxt = asoc->pathmaxrxt;
5616 trans = sctp_addr_id2transport(sk, &val.spt_address,
5621 val.spt_pathmaxrxt = trans->pathmaxrxt;
5622 val.spt_pathpfthld = trans->pf_retrans;
5625 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5632 * SCTP_GET_ASSOC_STATS
5634 * This option retrieves local per endpoint statistics. It is modeled
5635 * after OpenSolaris' implementation
5637 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
5638 char __user *optval,
5641 struct sctp_assoc_stats sas;
5642 struct sctp_association *asoc = NULL;
5644 /* User must provide at least the assoc id */
5645 if (len < sizeof(sctp_assoc_t))
5648 /* Allow the struct to grow and fill in as much as possible */
5649 len = min_t(size_t, len, sizeof(sas));
5651 if (copy_from_user(&sas, optval, len))
5654 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
5658 sas.sas_rtxchunks = asoc->stats.rtxchunks;
5659 sas.sas_gapcnt = asoc->stats.gapcnt;
5660 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
5661 sas.sas_osacks = asoc->stats.osacks;
5662 sas.sas_isacks = asoc->stats.isacks;
5663 sas.sas_octrlchunks = asoc->stats.octrlchunks;
5664 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
5665 sas.sas_oodchunks = asoc->stats.oodchunks;
5666 sas.sas_iodchunks = asoc->stats.iodchunks;
5667 sas.sas_ouodchunks = asoc->stats.ouodchunks;
5668 sas.sas_iuodchunks = asoc->stats.iuodchunks;
5669 sas.sas_idupchunks = asoc->stats.idupchunks;
5670 sas.sas_opackets = asoc->stats.opackets;
5671 sas.sas_ipackets = asoc->stats.ipackets;
5673 /* New high max rto observed, will return 0 if not a single
5674 * RTO update took place. obs_rto_ipaddr will be bogus
5677 sas.sas_maxrto = asoc->stats.max_obs_rto;
5678 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
5679 sizeof(struct sockaddr_storage));
5681 /* Mark beginning of a new observation period */
5682 asoc->stats.max_obs_rto = asoc->rto_min;
5684 if (put_user(len, optlen))
5687 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
5689 if (copy_to_user(optval, &sas, len))
5695 static int sctp_getsockopt(struct sock *sk, int level, int optname,
5696 char __user *optval, int __user *optlen)
5701 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
5703 /* I can hardly begin to describe how wrong this is. This is
5704 * so broken as to be worse than useless. The API draft
5705 * REALLY is NOT helpful here... I am not convinced that the
5706 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5707 * are at all well-founded.
5709 if (level != SOL_SCTP) {
5710 struct sctp_af *af = sctp_sk(sk)->pf->af;
5712 retval = af->getsockopt(sk, level, optname, optval, optlen);
5716 if (get_user(len, optlen))
5723 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5725 case SCTP_DISABLE_FRAGMENTS:
5726 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5730 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5732 case SCTP_AUTOCLOSE:
5733 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5735 case SCTP_SOCKOPT_PEELOFF:
5736 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5738 case SCTP_PEER_ADDR_PARAMS:
5739 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5742 case SCTP_DELAYED_SACK:
5743 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5747 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5749 case SCTP_GET_PEER_ADDRS:
5750 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5753 case SCTP_GET_LOCAL_ADDRS:
5754 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5757 case SCTP_SOCKOPT_CONNECTX3:
5758 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5760 case SCTP_DEFAULT_SEND_PARAM:
5761 retval = sctp_getsockopt_default_send_param(sk, len,
5764 case SCTP_PRIMARY_ADDR:
5765 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5768 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5771 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5773 case SCTP_ASSOCINFO:
5774 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5776 case SCTP_I_WANT_MAPPED_V4_ADDR:
5777 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5780 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5782 case SCTP_GET_PEER_ADDR_INFO:
5783 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5786 case SCTP_ADAPTATION_LAYER:
5787 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5791 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5793 case SCTP_FRAGMENT_INTERLEAVE:
5794 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5797 case SCTP_PARTIAL_DELIVERY_POINT:
5798 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5801 case SCTP_MAX_BURST:
5802 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5805 case SCTP_AUTH_CHUNK:
5806 case SCTP_AUTH_DELETE_KEY:
5807 retval = -EOPNOTSUPP;
5809 case SCTP_HMAC_IDENT:
5810 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5812 case SCTP_AUTH_ACTIVE_KEY:
5813 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5815 case SCTP_PEER_AUTH_CHUNKS:
5816 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5819 case SCTP_LOCAL_AUTH_CHUNKS:
5820 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5823 case SCTP_GET_ASSOC_NUMBER:
5824 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5826 case SCTP_GET_ASSOC_ID_LIST:
5827 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5829 case SCTP_AUTO_ASCONF:
5830 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5832 case SCTP_PEER_ADDR_THLDS:
5833 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
5835 case SCTP_GET_ASSOC_STATS:
5836 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
5839 retval = -ENOPROTOOPT;
5843 sctp_release_sock(sk);
5847 static void sctp_hash(struct sock *sk)
5852 static void sctp_unhash(struct sock *sk)
5857 /* Check if port is acceptable. Possibly find first available port.
5859 * The port hash table (contained in the 'global' SCTP protocol storage
5860 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5861 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5862 * list (the list number is the port number hashed out, so as you
5863 * would expect from a hash function, all the ports in a given list have
5864 * such a number that hashes out to the same list number; you were
5865 * expecting that, right?); so each list has a set of ports, with a
5866 * link to the socket (struct sock) that uses it, the port number and
5867 * a fastreuse flag (FIXME: NPI ipg).
5869 static struct sctp_bind_bucket *sctp_bucket_create(
5870 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
5872 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5874 struct sctp_bind_hashbucket *head; /* hash list */
5875 struct sctp_bind_bucket *pp;
5876 unsigned short snum;
5879 snum = ntohs(addr->v4.sin_port);
5881 pr_debug("%s: begins, snum:%d\n", __func__, snum);
5883 sctp_local_bh_disable();
5886 /* Search for an available port. */
5887 int low, high, remaining, index;
5890 inet_get_local_port_range(&low, &high);
5891 remaining = (high - low) + 1;
5892 rover = net_random() % remaining + low;
5896 if ((rover < low) || (rover > high))
5898 if (inet_is_reserved_local_port(rover))
5900 index = sctp_phashfn(sock_net(sk), rover);
5901 head = &sctp_port_hashtable[index];
5902 sctp_spin_lock(&head->lock);
5903 sctp_for_each_hentry(pp, &head->chain)
5904 if ((pp->port == rover) &&
5905 net_eq(sock_net(sk), pp->net))
5909 sctp_spin_unlock(&head->lock);
5910 } while (--remaining > 0);
5912 /* Exhausted local port range during search? */
5917 /* OK, here is the one we will use. HEAD (the port
5918 * hash table list entry) is non-NULL and we hold it's
5923 /* We are given an specific port number; we verify
5924 * that it is not being used. If it is used, we will
5925 * exahust the search in the hash list corresponding
5926 * to the port number (snum) - we detect that with the
5927 * port iterator, pp being NULL.
5929 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
5930 sctp_spin_lock(&head->lock);
5931 sctp_for_each_hentry(pp, &head->chain) {
5932 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
5939 if (!hlist_empty(&pp->owner)) {
5940 /* We had a port hash table hit - there is an
5941 * available port (pp != NULL) and it is being
5942 * used by other socket (pp->owner not empty); that other
5943 * socket is going to be sk2.
5945 int reuse = sk->sk_reuse;
5948 pr_debug("%s: found a possible match\n", __func__);
5950 if (pp->fastreuse && sk->sk_reuse &&
5951 sk->sk_state != SCTP_SS_LISTENING)
5954 /* Run through the list of sockets bound to the port
5955 * (pp->port) [via the pointers bind_next and
5956 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5957 * we get the endpoint they describe and run through
5958 * the endpoint's list of IP (v4 or v6) addresses,
5959 * comparing each of the addresses with the address of
5960 * the socket sk. If we find a match, then that means
5961 * that this port/socket (sk) combination are already
5964 sk_for_each_bound(sk2, &pp->owner) {
5965 struct sctp_endpoint *ep2;
5966 ep2 = sctp_sk(sk2)->ep;
5969 (reuse && sk2->sk_reuse &&
5970 sk2->sk_state != SCTP_SS_LISTENING))
5973 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5974 sctp_sk(sk2), sctp_sk(sk))) {
5980 pr_debug("%s: found a match\n", __func__);
5983 /* If there was a hash table miss, create a new port. */
5985 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
5988 /* In either case (hit or miss), make sure fastreuse is 1 only
5989 * if sk->sk_reuse is too (that is, if the caller requested
5990 * SO_REUSEADDR on this socket -sk-).
5992 if (hlist_empty(&pp->owner)) {
5993 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5997 } else if (pp->fastreuse &&
5998 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6001 /* We are set, so fill up all the data in the hash table
6002 * entry, tie the socket list information with the rest of the
6003 * sockets FIXME: Blurry, NPI (ipg).
6006 if (!sctp_sk(sk)->bind_hash) {
6007 inet_sk(sk)->inet_num = snum;
6008 sk_add_bind_node(sk, &pp->owner);
6009 sctp_sk(sk)->bind_hash = pp;
6014 sctp_spin_unlock(&head->lock);
6017 sctp_local_bh_enable();
6021 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6022 * port is requested.
6024 static int sctp_get_port(struct sock *sk, unsigned short snum)
6026 union sctp_addr addr;
6027 struct sctp_af *af = sctp_sk(sk)->pf->af;
6029 /* Set up a dummy address struct from the sk. */
6030 af->from_sk(&addr, sk);
6031 addr.v4.sin_port = htons(snum);
6033 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6034 return !!sctp_get_port_local(sk, &addr);
6038 * Move a socket to LISTENING state.
6040 static int sctp_listen_start(struct sock *sk, int backlog)
6042 struct sctp_sock *sp = sctp_sk(sk);
6043 struct sctp_endpoint *ep = sp->ep;
6044 struct crypto_hash *tfm = NULL;
6047 /* Allocate HMAC for generating cookie. */
6048 if (!sp->hmac && sp->sctp_hmac_alg) {
6049 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6050 tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
6052 net_info_ratelimited("failed to load transform for %s: %ld\n",
6053 sp->sctp_hmac_alg, PTR_ERR(tfm));
6056 sctp_sk(sk)->hmac = tfm;
6060 * If a bind() or sctp_bindx() is not called prior to a listen()
6061 * call that allows new associations to be accepted, the system
6062 * picks an ephemeral port and will choose an address set equivalent
6063 * to binding with a wildcard address.
6065 * This is not currently spelled out in the SCTP sockets
6066 * extensions draft, but follows the practice as seen in TCP
6070 sk->sk_state = SCTP_SS_LISTENING;
6071 if (!ep->base.bind_addr.port) {
6072 if (sctp_autobind(sk))
6075 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6076 sk->sk_state = SCTP_SS_CLOSED;
6081 sk->sk_max_ack_backlog = backlog;
6082 sctp_hash_endpoint(ep);
6087 * 4.1.3 / 5.1.3 listen()
6089 * By default, new associations are not accepted for UDP style sockets.
6090 * An application uses listen() to mark a socket as being able to
6091 * accept new associations.
6093 * On TCP style sockets, applications use listen() to ready the SCTP
6094 * endpoint for accepting inbound associations.
6096 * On both types of endpoints a backlog of '0' disables listening.
6098 * Move a socket to LISTENING state.
6100 int sctp_inet_listen(struct socket *sock, int backlog)
6102 struct sock *sk = sock->sk;
6103 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6106 if (unlikely(backlog < 0))
6111 /* Peeled-off sockets are not allowed to listen(). */
6112 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6115 if (sock->state != SS_UNCONNECTED)
6118 /* If backlog is zero, disable listening. */
6120 if (sctp_sstate(sk, CLOSED))
6124 sctp_unhash_endpoint(ep);
6125 sk->sk_state = SCTP_SS_CLOSED;
6127 sctp_sk(sk)->bind_hash->fastreuse = 1;
6131 /* If we are already listening, just update the backlog */
6132 if (sctp_sstate(sk, LISTENING))
6133 sk->sk_max_ack_backlog = backlog;
6135 err = sctp_listen_start(sk, backlog);
6142 sctp_release_sock(sk);
6147 * This function is done by modeling the current datagram_poll() and the
6148 * tcp_poll(). Note that, based on these implementations, we don't
6149 * lock the socket in this function, even though it seems that,
6150 * ideally, locking or some other mechanisms can be used to ensure
6151 * the integrity of the counters (sndbuf and wmem_alloc) used
6152 * in this place. We assume that we don't need locks either until proven
6155 * Another thing to note is that we include the Async I/O support
6156 * here, again, by modeling the current TCP/UDP code. We don't have
6157 * a good way to test with it yet.
6159 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6161 struct sock *sk = sock->sk;
6162 struct sctp_sock *sp = sctp_sk(sk);
6165 poll_wait(file, sk_sleep(sk), wait);
6167 /* A TCP-style listening socket becomes readable when the accept queue
6170 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6171 return (!list_empty(&sp->ep->asocs)) ?
6172 (POLLIN | POLLRDNORM) : 0;
6176 /* Is there any exceptional events? */
6177 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6179 sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0;
6180 if (sk->sk_shutdown & RCV_SHUTDOWN)
6181 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6182 if (sk->sk_shutdown == SHUTDOWN_MASK)
6185 /* Is it readable? Reconsider this code with TCP-style support. */
6186 if (!skb_queue_empty(&sk->sk_receive_queue))
6187 mask |= POLLIN | POLLRDNORM;
6189 /* The association is either gone or not ready. */
6190 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6193 /* Is it writable? */
6194 if (sctp_writeable(sk)) {
6195 mask |= POLLOUT | POLLWRNORM;
6197 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6199 * Since the socket is not locked, the buffer
6200 * might be made available after the writeable check and
6201 * before the bit is set. This could cause a lost I/O
6202 * signal. tcp_poll() has a race breaker for this race
6203 * condition. Based on their implementation, we put
6204 * in the following code to cover it as well.
6206 if (sctp_writeable(sk))
6207 mask |= POLLOUT | POLLWRNORM;
6212 /********************************************************************
6213 * 2nd Level Abstractions
6214 ********************************************************************/
6216 static struct sctp_bind_bucket *sctp_bucket_create(
6217 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6219 struct sctp_bind_bucket *pp;
6221 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6223 SCTP_DBG_OBJCNT_INC(bind_bucket);
6226 INIT_HLIST_HEAD(&pp->owner);
6228 hlist_add_head(&pp->node, &head->chain);
6233 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6234 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6236 if (pp && hlist_empty(&pp->owner)) {
6237 __hlist_del(&pp->node);
6238 kmem_cache_free(sctp_bucket_cachep, pp);
6239 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6243 /* Release this socket's reference to a local port. */
6244 static inline void __sctp_put_port(struct sock *sk)
6246 struct sctp_bind_hashbucket *head =
6247 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6248 inet_sk(sk)->inet_num)];
6249 struct sctp_bind_bucket *pp;
6251 sctp_spin_lock(&head->lock);
6252 pp = sctp_sk(sk)->bind_hash;
6253 __sk_del_bind_node(sk);
6254 sctp_sk(sk)->bind_hash = NULL;
6255 inet_sk(sk)->inet_num = 0;
6256 sctp_bucket_destroy(pp);
6257 sctp_spin_unlock(&head->lock);
6260 void sctp_put_port(struct sock *sk)
6262 sctp_local_bh_disable();
6263 __sctp_put_port(sk);
6264 sctp_local_bh_enable();
6268 * The system picks an ephemeral port and choose an address set equivalent
6269 * to binding with a wildcard address.
6270 * One of those addresses will be the primary address for the association.
6271 * This automatically enables the multihoming capability of SCTP.
6273 static int sctp_autobind(struct sock *sk)
6275 union sctp_addr autoaddr;
6279 /* Initialize a local sockaddr structure to INADDR_ANY. */
6280 af = sctp_sk(sk)->pf->af;
6282 port = htons(inet_sk(sk)->inet_num);
6283 af->inaddr_any(&autoaddr, port);
6285 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6288 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6291 * 4.2 The cmsghdr Structure *
6293 * When ancillary data is sent or received, any number of ancillary data
6294 * objects can be specified by the msg_control and msg_controllen members of
6295 * the msghdr structure, because each object is preceded by
6296 * a cmsghdr structure defining the object's length (the cmsg_len member).
6297 * Historically Berkeley-derived implementations have passed only one object
6298 * at a time, but this API allows multiple objects to be
6299 * passed in a single call to sendmsg() or recvmsg(). The following example
6300 * shows two ancillary data objects in a control buffer.
6302 * |<--------------------------- msg_controllen -------------------------->|
6305 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6307 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6310 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6312 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6315 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6316 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6318 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6320 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6327 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
6329 struct cmsghdr *cmsg;
6330 struct msghdr *my_msg = (struct msghdr *)msg;
6332 for (cmsg = CMSG_FIRSTHDR(msg);
6334 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6335 if (!CMSG_OK(my_msg, cmsg))
6338 /* Should we parse this header or ignore? */
6339 if (cmsg->cmsg_level != IPPROTO_SCTP)
6342 /* Strictly check lengths following example in SCM code. */
6343 switch (cmsg->cmsg_type) {
6345 /* SCTP Socket API Extension
6346 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6348 * This cmsghdr structure provides information for
6349 * initializing new SCTP associations with sendmsg().
6350 * The SCTP_INITMSG socket option uses this same data
6351 * structure. This structure is not used for
6354 * cmsg_level cmsg_type cmsg_data[]
6355 * ------------ ------------ ----------------------
6356 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6358 if (cmsg->cmsg_len !=
6359 CMSG_LEN(sizeof(struct sctp_initmsg)))
6361 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6365 /* SCTP Socket API Extension
6366 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6368 * This cmsghdr structure specifies SCTP options for
6369 * sendmsg() and describes SCTP header information
6370 * about a received message through recvmsg().
6372 * cmsg_level cmsg_type cmsg_data[]
6373 * ------------ ------------ ----------------------
6374 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6376 if (cmsg->cmsg_len !=
6377 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6381 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6383 /* Minimally, validate the sinfo_flags. */
6384 if (cmsgs->info->sinfo_flags &
6385 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6386 SCTP_ABORT | SCTP_EOF))
6398 * Wait for a packet..
6399 * Note: This function is the same function as in core/datagram.c
6400 * with a few modifications to make lksctp work.
6402 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6407 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6409 /* Socket errors? */
6410 error = sock_error(sk);
6414 if (!skb_queue_empty(&sk->sk_receive_queue))
6417 /* Socket shut down? */
6418 if (sk->sk_shutdown & RCV_SHUTDOWN)
6421 /* Sequenced packets can come disconnected. If so we report the
6426 /* Is there a good reason to think that we may receive some data? */
6427 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6430 /* Handle signals. */
6431 if (signal_pending(current))
6434 /* Let another process have a go. Since we are going to sleep
6435 * anyway. Note: This may cause odd behaviors if the message
6436 * does not fit in the user's buffer, but this seems to be the
6437 * only way to honor MSG_DONTWAIT realistically.
6439 sctp_release_sock(sk);
6440 *timeo_p = schedule_timeout(*timeo_p);
6444 finish_wait(sk_sleep(sk), &wait);
6448 error = sock_intr_errno(*timeo_p);
6451 finish_wait(sk_sleep(sk), &wait);
6456 /* Receive a datagram.
6457 * Note: This is pretty much the same routine as in core/datagram.c
6458 * with a few changes to make lksctp work.
6460 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6461 int noblock, int *err)
6464 struct sk_buff *skb;
6467 timeo = sock_rcvtimeo(sk, noblock);
6469 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
6470 MAX_SCHEDULE_TIMEOUT);
6473 /* Again only user level code calls this function,
6474 * so nothing interrupt level
6475 * will suddenly eat the receive_queue.
6477 * Look at current nfs client by the way...
6478 * However, this function was correct in any case. 8)
6480 if (flags & MSG_PEEK) {
6481 spin_lock_bh(&sk->sk_receive_queue.lock);
6482 skb = skb_peek(&sk->sk_receive_queue);
6484 atomic_inc(&skb->users);
6485 spin_unlock_bh(&sk->sk_receive_queue.lock);
6487 skb = skb_dequeue(&sk->sk_receive_queue);
6493 /* Caller is allowed not to check sk->sk_err before calling. */
6494 error = sock_error(sk);
6498 if (sk->sk_shutdown & RCV_SHUTDOWN)
6501 /* User doesn't want to wait. */
6505 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6514 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6515 static void __sctp_write_space(struct sctp_association *asoc)
6517 struct sock *sk = asoc->base.sk;
6518 struct socket *sock = sk->sk_socket;
6520 if ((sctp_wspace(asoc) > 0) && sock) {
6521 if (waitqueue_active(&asoc->wait))
6522 wake_up_interruptible(&asoc->wait);
6524 if (sctp_writeable(sk)) {
6525 wait_queue_head_t *wq = sk_sleep(sk);
6527 if (wq && waitqueue_active(wq))
6528 wake_up_interruptible(wq);
6530 /* Note that we try to include the Async I/O support
6531 * here by modeling from the current TCP/UDP code.
6532 * We have not tested with it yet.
6534 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6535 sock_wake_async(sock,
6536 SOCK_WAKE_SPACE, POLL_OUT);
6541 /* Do accounting for the sndbuf space.
6542 * Decrement the used sndbuf space of the corresponding association by the
6543 * data size which was just transmitted(freed).
6545 static void sctp_wfree(struct sk_buff *skb)
6547 struct sctp_association *asoc;
6548 struct sctp_chunk *chunk;
6551 /* Get the saved chunk pointer. */
6552 chunk = *((struct sctp_chunk **)(skb->cb));
6555 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6556 sizeof(struct sk_buff) +
6557 sizeof(struct sctp_chunk);
6559 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6562 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6564 sk->sk_wmem_queued -= skb->truesize;
6565 sk_mem_uncharge(sk, skb->truesize);
6568 __sctp_write_space(asoc);
6570 sctp_association_put(asoc);
6573 /* Do accounting for the receive space on the socket.
6574 * Accounting for the association is done in ulpevent.c
6575 * We set this as a destructor for the cloned data skbs so that
6576 * accounting is done at the correct time.
6578 void sctp_sock_rfree(struct sk_buff *skb)
6580 struct sock *sk = skb->sk;
6581 struct sctp_ulpevent *event = sctp_skb2event(skb);
6583 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6586 * Mimic the behavior of sock_rfree
6588 sk_mem_uncharge(sk, event->rmem_len);
6592 /* Helper function to wait for space in the sndbuf. */
6593 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6596 struct sock *sk = asoc->base.sk;
6598 long current_timeo = *timeo_p;
6601 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
6604 /* Increment the association's refcnt. */
6605 sctp_association_hold(asoc);
6607 /* Wait on the association specific sndbuf space. */
6609 prepare_to_wait_exclusive(&asoc->wait, &wait,
6610 TASK_INTERRUPTIBLE);
6613 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6616 if (signal_pending(current))
6617 goto do_interrupted;
6618 if (msg_len <= sctp_wspace(asoc))
6621 /* Let another process have a go. Since we are going
6624 sctp_release_sock(sk);
6625 current_timeo = schedule_timeout(current_timeo);
6626 BUG_ON(sk != asoc->base.sk);
6629 *timeo_p = current_timeo;
6633 finish_wait(&asoc->wait, &wait);
6635 /* Release the association's refcnt. */
6636 sctp_association_put(asoc);
6645 err = sock_intr_errno(*timeo_p);
6653 void sctp_data_ready(struct sock *sk, int len)
6655 struct socket_wq *wq;
6658 wq = rcu_dereference(sk->sk_wq);
6659 if (wq_has_sleeper(wq))
6660 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6661 POLLRDNORM | POLLRDBAND);
6662 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6666 /* If socket sndbuf has changed, wake up all per association waiters. */
6667 void sctp_write_space(struct sock *sk)
6669 struct sctp_association *asoc;
6671 /* Wake up the tasks in each wait queue. */
6672 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6673 __sctp_write_space(asoc);
6677 /* Is there any sndbuf space available on the socket?
6679 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6680 * associations on the same socket. For a UDP-style socket with
6681 * multiple associations, it is possible for it to be "unwriteable"
6682 * prematurely. I assume that this is acceptable because
6683 * a premature "unwriteable" is better than an accidental "writeable" which
6684 * would cause an unwanted block under certain circumstances. For the 1-1
6685 * UDP-style sockets or TCP-style sockets, this code should work.
6688 static int sctp_writeable(struct sock *sk)
6692 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6698 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6699 * returns immediately with EINPROGRESS.
6701 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6703 struct sock *sk = asoc->base.sk;
6705 long current_timeo = *timeo_p;
6708 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
6710 /* Increment the association's refcnt. */
6711 sctp_association_hold(asoc);
6714 prepare_to_wait_exclusive(&asoc->wait, &wait,
6715 TASK_INTERRUPTIBLE);
6718 if (sk->sk_shutdown & RCV_SHUTDOWN)
6720 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6723 if (signal_pending(current))
6724 goto do_interrupted;
6726 if (sctp_state(asoc, ESTABLISHED))
6729 /* Let another process have a go. Since we are going
6732 sctp_release_sock(sk);
6733 current_timeo = schedule_timeout(current_timeo);
6736 *timeo_p = current_timeo;
6740 finish_wait(&asoc->wait, &wait);
6742 /* Release the association's refcnt. */
6743 sctp_association_put(asoc);
6748 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6751 err = -ECONNREFUSED;
6755 err = sock_intr_errno(*timeo_p);
6763 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6765 struct sctp_endpoint *ep;
6769 ep = sctp_sk(sk)->ep;
6773 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6774 TASK_INTERRUPTIBLE);
6776 if (list_empty(&ep->asocs)) {
6777 sctp_release_sock(sk);
6778 timeo = schedule_timeout(timeo);
6783 if (!sctp_sstate(sk, LISTENING))
6787 if (!list_empty(&ep->asocs))
6790 err = sock_intr_errno(timeo);
6791 if (signal_pending(current))
6799 finish_wait(sk_sleep(sk), &wait);
6804 static void sctp_wait_for_close(struct sock *sk, long timeout)
6809 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6810 if (list_empty(&sctp_sk(sk)->ep->asocs))
6812 sctp_release_sock(sk);
6813 timeout = schedule_timeout(timeout);
6815 } while (!signal_pending(current) && timeout);
6817 finish_wait(sk_sleep(sk), &wait);
6820 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6822 struct sk_buff *frag;
6827 /* Don't forget the fragments. */
6828 skb_walk_frags(skb, frag)
6829 sctp_skb_set_owner_r_frag(frag, sk);
6832 sctp_skb_set_owner_r(skb, sk);
6835 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6836 struct sctp_association *asoc)
6838 struct inet_sock *inet = inet_sk(sk);
6839 struct inet_sock *newinet;
6841 newsk->sk_type = sk->sk_type;
6842 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6843 newsk->sk_flags = sk->sk_flags;
6844 newsk->sk_no_check = sk->sk_no_check;
6845 newsk->sk_reuse = sk->sk_reuse;
6847 newsk->sk_shutdown = sk->sk_shutdown;
6848 newsk->sk_destruct = sctp_destruct_sock;
6849 newsk->sk_family = sk->sk_family;
6850 newsk->sk_protocol = IPPROTO_SCTP;
6851 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6852 newsk->sk_sndbuf = sk->sk_sndbuf;
6853 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6854 newsk->sk_lingertime = sk->sk_lingertime;
6855 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6856 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6858 newinet = inet_sk(newsk);
6860 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6861 * getsockname() and getpeername()
6863 newinet->inet_sport = inet->inet_sport;
6864 newinet->inet_saddr = inet->inet_saddr;
6865 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6866 newinet->inet_dport = htons(asoc->peer.port);
6867 newinet->pmtudisc = inet->pmtudisc;
6868 newinet->inet_id = asoc->next_tsn ^ jiffies;
6870 newinet->uc_ttl = inet->uc_ttl;
6871 newinet->mc_loop = 1;
6872 newinet->mc_ttl = 1;
6873 newinet->mc_index = 0;
6874 newinet->mc_list = NULL;
6877 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6878 * and its messages to the newsk.
6880 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6881 struct sctp_association *assoc,
6882 sctp_socket_type_t type)
6884 struct sctp_sock *oldsp = sctp_sk(oldsk);
6885 struct sctp_sock *newsp = sctp_sk(newsk);
6886 struct sctp_bind_bucket *pp; /* hash list port iterator */
6887 struct sctp_endpoint *newep = newsp->ep;
6888 struct sk_buff *skb, *tmp;
6889 struct sctp_ulpevent *event;
6890 struct sctp_bind_hashbucket *head;
6891 struct list_head tmplist;
6893 /* Migrate socket buffer sizes and all the socket level options to the
6896 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6897 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6898 /* Brute force copy old sctp opt. */
6899 if (oldsp->do_auto_asconf) {
6900 memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
6901 inet_sk_copy_descendant(newsk, oldsk);
6902 memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
6904 inet_sk_copy_descendant(newsk, oldsk);
6906 /* Restore the ep value that was overwritten with the above structure
6912 /* Hook this new socket in to the bind_hash list. */
6913 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
6914 inet_sk(oldsk)->inet_num)];
6915 sctp_local_bh_disable();
6916 sctp_spin_lock(&head->lock);
6917 pp = sctp_sk(oldsk)->bind_hash;
6918 sk_add_bind_node(newsk, &pp->owner);
6919 sctp_sk(newsk)->bind_hash = pp;
6920 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6921 sctp_spin_unlock(&head->lock);
6922 sctp_local_bh_enable();
6924 /* Copy the bind_addr list from the original endpoint to the new
6925 * endpoint so that we can handle restarts properly
6927 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6928 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6930 /* Move any messages in the old socket's receive queue that are for the
6931 * peeled off association to the new socket's receive queue.
6933 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6934 event = sctp_skb2event(skb);
6935 if (event->asoc == assoc) {
6936 __skb_unlink(skb, &oldsk->sk_receive_queue);
6937 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6938 sctp_skb_set_owner_r_frag(skb, newsk);
6942 /* Clean up any messages pending delivery due to partial
6943 * delivery. Three cases:
6944 * 1) No partial deliver; no work.
6945 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6946 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6948 skb_queue_head_init(&newsp->pd_lobby);
6949 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6951 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6952 struct sk_buff_head *queue;
6954 /* Decide which queue to move pd_lobby skbs to. */
6955 if (assoc->ulpq.pd_mode) {
6956 queue = &newsp->pd_lobby;
6958 queue = &newsk->sk_receive_queue;
6960 /* Walk through the pd_lobby, looking for skbs that
6961 * need moved to the new socket.
6963 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6964 event = sctp_skb2event(skb);
6965 if (event->asoc == assoc) {
6966 __skb_unlink(skb, &oldsp->pd_lobby);
6967 __skb_queue_tail(queue, skb);
6968 sctp_skb_set_owner_r_frag(skb, newsk);
6972 /* Clear up any skbs waiting for the partial
6973 * delivery to finish.
6975 if (assoc->ulpq.pd_mode)
6976 sctp_clear_pd(oldsk, NULL);
6980 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6981 sctp_skb_set_owner_r_frag(skb, newsk);
6983 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6984 sctp_skb_set_owner_r_frag(skb, newsk);
6986 /* Set the type of socket to indicate that it is peeled off from the
6987 * original UDP-style socket or created with the accept() call on a
6988 * TCP-style socket..
6992 /* Mark the new socket "in-use" by the user so that any packets
6993 * that may arrive on the association after we've moved it are
6994 * queued to the backlog. This prevents a potential race between
6995 * backlog processing on the old socket and new-packet processing
6996 * on the new socket.
6998 * The caller has just allocated newsk so we can guarantee that other
6999 * paths won't try to lock it and then oldsk.
7001 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7002 sctp_assoc_migrate(assoc, newsk);
7004 /* If the association on the newsk is already closed before accept()
7005 * is called, set RCV_SHUTDOWN flag.
7007 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
7008 newsk->sk_shutdown |= RCV_SHUTDOWN;
7010 newsk->sk_state = SCTP_SS_ESTABLISHED;
7011 sctp_release_sock(newsk);
7015 /* This proto struct describes the ULP interface for SCTP. */
7016 struct proto sctp_prot = {
7018 .owner = THIS_MODULE,
7019 .close = sctp_close,
7020 .connect = sctp_connect,
7021 .disconnect = sctp_disconnect,
7022 .accept = sctp_accept,
7023 .ioctl = sctp_ioctl,
7024 .init = sctp_init_sock,
7025 .destroy = sctp_destroy_sock,
7026 .shutdown = sctp_shutdown,
7027 .setsockopt = sctp_setsockopt,
7028 .getsockopt = sctp_getsockopt,
7029 .sendmsg = sctp_sendmsg,
7030 .recvmsg = sctp_recvmsg,
7032 .backlog_rcv = sctp_backlog_rcv,
7034 .unhash = sctp_unhash,
7035 .get_port = sctp_get_port,
7036 .obj_size = sizeof(struct sctp_sock),
7037 .sysctl_mem = sysctl_sctp_mem,
7038 .sysctl_rmem = sysctl_sctp_rmem,
7039 .sysctl_wmem = sysctl_sctp_wmem,
7040 .memory_pressure = &sctp_memory_pressure,
7041 .enter_memory_pressure = sctp_enter_memory_pressure,
7042 .memory_allocated = &sctp_memory_allocated,
7043 .sockets_allocated = &sctp_sockets_allocated,
7046 #if IS_ENABLED(CONFIG_IPV6)
7048 struct proto sctpv6_prot = {
7050 .owner = THIS_MODULE,
7051 .close = sctp_close,
7052 .connect = sctp_connect,
7053 .disconnect = sctp_disconnect,
7054 .accept = sctp_accept,
7055 .ioctl = sctp_ioctl,
7056 .init = sctp_init_sock,
7057 .destroy = sctp_destroy_sock,
7058 .shutdown = sctp_shutdown,
7059 .setsockopt = sctp_setsockopt,
7060 .getsockopt = sctp_getsockopt,
7061 .sendmsg = sctp_sendmsg,
7062 .recvmsg = sctp_recvmsg,
7064 .backlog_rcv = sctp_backlog_rcv,
7066 .unhash = sctp_unhash,
7067 .get_port = sctp_get_port,
7068 .obj_size = sizeof(struct sctp6_sock),
7069 .sysctl_mem = sysctl_sctp_mem,
7070 .sysctl_rmem = sysctl_sctp_rmem,
7071 .sysctl_wmem = sysctl_sctp_wmem,
7072 .memory_pressure = &sctp_memory_pressure,
7073 .enter_memory_pressure = sctp_enter_memory_pressure,
7074 .memory_allocated = &sctp_memory_allocated,
7075 .sockets_allocated = &sctp_sockets_allocated,
7077 #endif /* IS_ENABLED(CONFIG_IPV6) */