1 /* SCTP kernel implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
9 * This file is part of the SCTP kernel implementation
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
18 * This SCTP implementation is free software;
19 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
24 * This SCTP implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, see
32 * <http://www.gnu.org/licenses/>.
34 * Please send any bug reports or fixes you make to the
38 * Written or modified by:
53 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
55 #include <linux/types.h>
56 #include <linux/kernel.h>
57 #include <linux/wait.h>
58 #include <linux/time.h>
60 #include <linux/capability.h>
61 #include <linux/fcntl.h>
62 #include <linux/poll.h>
63 #include <linux/init.h>
64 #include <linux/crypto.h>
65 #include <linux/slab.h>
66 #include <linux/file.h>
67 #include <linux/compat.h>
71 #include <net/route.h>
73 #include <net/inet_common.h>
74 #include <net/busy_poll.h>
76 #include <linux/socket.h> /* for sa_family_t */
77 #include <linux/export.h>
79 #include <net/sctp/sctp.h>
80 #include <net/sctp/sm.h>
82 /* Forward declarations for internal helper functions. */
83 static int sctp_writeable(struct sock *sk);
84 static void sctp_wfree(struct sk_buff *skb);
85 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
87 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p);
88 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
89 static int sctp_wait_for_accept(struct sock *sk, long timeo);
90 static void sctp_wait_for_close(struct sock *sk, long timeo);
91 static void sctp_destruct_sock(struct sock *sk);
92 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
93 union sctp_addr *addr, int len);
94 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
95 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
96 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
97 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
98 static int sctp_send_asconf(struct sctp_association *asoc,
99 struct sctp_chunk *chunk);
100 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
101 static int sctp_autobind(struct sock *sk);
102 static void sctp_sock_migrate(struct sock *, struct sock *,
103 struct sctp_association *, sctp_socket_type_t);
105 static int sctp_memory_pressure;
106 static atomic_long_t sctp_memory_allocated;
107 struct percpu_counter sctp_sockets_allocated;
109 static void sctp_enter_memory_pressure(struct sock *sk)
111 sctp_memory_pressure = 1;
115 /* Get the sndbuf space available at the time on the association. */
116 static inline int sctp_wspace(struct sctp_association *asoc)
120 if (asoc->ep->sndbuf_policy)
121 amt = asoc->sndbuf_used;
123 amt = sk_wmem_alloc_get(asoc->base.sk);
125 if (amt >= asoc->base.sk->sk_sndbuf) {
126 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
129 amt = sk_stream_wspace(asoc->base.sk);
134 amt = asoc->base.sk->sk_sndbuf - amt;
139 /* Increment the used sndbuf space count of the corresponding association by
140 * the size of the outgoing data chunk.
141 * Also, set the skb destructor for sndbuf accounting later.
143 * Since it is always 1-1 between chunk and skb, and also a new skb is always
144 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
145 * destructor in the data chunk skb for the purpose of the sndbuf space
148 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
150 struct sctp_association *asoc = chunk->asoc;
151 struct sock *sk = asoc->base.sk;
153 /* The sndbuf space is tracked per association. */
154 sctp_association_hold(asoc);
156 skb_set_owner_w(chunk->skb, sk);
158 chunk->skb->destructor = sctp_wfree;
159 /* Save the chunk pointer in skb for sctp_wfree to use later. */
160 skb_shinfo(chunk->skb)->destructor_arg = chunk;
162 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
163 sizeof(struct sk_buff) +
164 sizeof(struct sctp_chunk);
166 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
167 sk->sk_wmem_queued += chunk->skb->truesize;
168 sk_mem_charge(sk, chunk->skb->truesize);
171 /* Verify that this is a valid address. */
172 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
177 /* Verify basic sockaddr. */
178 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
182 /* Is this a valid SCTP address? */
183 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
186 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
192 /* Look up the association by its id. If this is not a UDP-style
193 * socket, the ID field is always ignored.
195 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
197 struct sctp_association *asoc = NULL;
199 /* If this is not a UDP-style socket, assoc id should be ignored. */
200 if (!sctp_style(sk, UDP)) {
201 /* Return NULL if the socket state is not ESTABLISHED. It
202 * could be a TCP-style listening socket or a socket which
203 * hasn't yet called connect() to establish an association.
205 if (!sctp_sstate(sk, ESTABLISHED))
208 /* Get the first and the only association from the list. */
209 if (!list_empty(&sctp_sk(sk)->ep->asocs))
210 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
211 struct sctp_association, asocs);
215 /* Otherwise this is a UDP-style socket. */
216 if (!id || (id == (sctp_assoc_t)-1))
219 spin_lock_bh(&sctp_assocs_id_lock);
220 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
221 spin_unlock_bh(&sctp_assocs_id_lock);
223 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
229 /* Look up the transport from an address and an assoc id. If both address and
230 * id are specified, the associations matching the address and the id should be
233 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
234 struct sockaddr_storage *addr,
237 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
238 struct sctp_transport *transport;
239 union sctp_addr *laddr = (union sctp_addr *)addr;
241 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
248 id_asoc = sctp_id2assoc(sk, id);
249 if (id_asoc && (id_asoc != addr_asoc))
252 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
253 (union sctp_addr *)addr);
258 /* API 3.1.2 bind() - UDP Style Syntax
259 * The syntax of bind() is,
261 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
263 * sd - the socket descriptor returned by socket().
264 * addr - the address structure (struct sockaddr_in or struct
265 * sockaddr_in6 [RFC 2553]),
266 * addr_len - the size of the address structure.
268 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
274 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
277 /* Disallow binding twice. */
278 if (!sctp_sk(sk)->ep->base.bind_addr.port)
279 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
289 static long sctp_get_port_local(struct sock *, union sctp_addr *);
291 /* Verify this is a valid sockaddr. */
292 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
293 union sctp_addr *addr, int len)
297 /* Check minimum size. */
298 if (len < sizeof (struct sockaddr))
301 /* V4 mapped address are really of AF_INET family */
302 if (addr->sa.sa_family == AF_INET6 &&
303 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
304 if (!opt->pf->af_supported(AF_INET, opt))
307 /* Does this PF support this AF? */
308 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
312 /* If we get this far, af is valid. */
313 af = sctp_get_af_specific(addr->sa.sa_family);
315 if (len < af->sockaddr_len)
321 /* Bind a local address either to an endpoint or to an association. */
322 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
324 struct net *net = sock_net(sk);
325 struct sctp_sock *sp = sctp_sk(sk);
326 struct sctp_endpoint *ep = sp->ep;
327 struct sctp_bind_addr *bp = &ep->base.bind_addr;
332 /* Common sockaddr verification. */
333 af = sctp_sockaddr_af(sp, addr, len);
335 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
336 __func__, sk, addr, len);
340 snum = ntohs(addr->v4.sin_port);
342 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
343 __func__, sk, &addr->sa, bp->port, snum, len);
345 /* PF specific bind() address verification. */
346 if (!sp->pf->bind_verify(sp, addr))
347 return -EADDRNOTAVAIL;
349 /* We must either be unbound, or bind to the same port.
350 * It's OK to allow 0 ports if we are already bound.
351 * We'll just inhert an already bound port in this case
356 else if (snum != bp->port) {
357 pr_debug("%s: new port %d doesn't match existing port "
358 "%d\n", __func__, snum, bp->port);
363 if (snum && snum < PROT_SOCK &&
364 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
367 /* See if the address matches any of the addresses we may have
368 * already bound before checking against other endpoints.
370 if (sctp_bind_addr_match(bp, addr, sp))
373 /* Make sure we are allowed to bind here.
374 * The function sctp_get_port_local() does duplicate address
377 addr->v4.sin_port = htons(snum);
378 if ((ret = sctp_get_port_local(sk, addr))) {
382 /* Refresh ephemeral port. */
384 bp->port = inet_sk(sk)->inet_num;
386 /* Add the address to the bind address list.
387 * Use GFP_ATOMIC since BHs will be disabled.
389 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
391 /* Copy back into socket for getsockname() use. */
393 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
394 sp->pf->to_sk_saddr(addr, sk);
400 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
402 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
403 * at any one time. If a sender, after sending an ASCONF chunk, decides
404 * it needs to transfer another ASCONF Chunk, it MUST wait until the
405 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
406 * subsequent ASCONF. Note this restriction binds each side, so at any
407 * time two ASCONF may be in-transit on any given association (one sent
408 * from each endpoint).
410 static int sctp_send_asconf(struct sctp_association *asoc,
411 struct sctp_chunk *chunk)
413 struct net *net = sock_net(asoc->base.sk);
416 /* If there is an outstanding ASCONF chunk, queue it for later
419 if (asoc->addip_last_asconf) {
420 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
424 /* Hold the chunk until an ASCONF_ACK is received. */
425 sctp_chunk_hold(chunk);
426 retval = sctp_primitive_ASCONF(net, asoc, chunk);
428 sctp_chunk_free(chunk);
430 asoc->addip_last_asconf = chunk;
436 /* Add a list of addresses as bind addresses to local endpoint or
439 * Basically run through each address specified in the addrs/addrcnt
440 * array/length pair, determine if it is IPv6 or IPv4 and call
441 * sctp_do_bind() on it.
443 * If any of them fails, then the operation will be reversed and the
444 * ones that were added will be removed.
446 * Only sctp_setsockopt_bindx() is supposed to call this function.
448 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
453 struct sockaddr *sa_addr;
456 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
460 for (cnt = 0; cnt < addrcnt; cnt++) {
461 /* The list may contain either IPv4 or IPv6 address;
462 * determine the address length for walking thru the list.
465 af = sctp_get_af_specific(sa_addr->sa_family);
471 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
474 addr_buf += af->sockaddr_len;
478 /* Failed. Cleanup the ones that have been added */
480 sctp_bindx_rem(sk, addrs, cnt);
488 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
489 * associations that are part of the endpoint indicating that a list of local
490 * addresses are added to the endpoint.
492 * If any of the addresses is already in the bind address list of the
493 * association, we do not send the chunk for that association. But it will not
494 * affect other associations.
496 * Only sctp_setsockopt_bindx() is supposed to call this function.
498 static int sctp_send_asconf_add_ip(struct sock *sk,
499 struct sockaddr *addrs,
502 struct net *net = sock_net(sk);
503 struct sctp_sock *sp;
504 struct sctp_endpoint *ep;
505 struct sctp_association *asoc;
506 struct sctp_bind_addr *bp;
507 struct sctp_chunk *chunk;
508 struct sctp_sockaddr_entry *laddr;
509 union sctp_addr *addr;
510 union sctp_addr saveaddr;
517 if (!net->sctp.addip_enable)
523 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
524 __func__, sk, addrs, addrcnt);
526 list_for_each_entry(asoc, &ep->asocs, asocs) {
527 if (!asoc->peer.asconf_capable)
530 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
533 if (!sctp_state(asoc, ESTABLISHED))
536 /* Check if any address in the packed array of addresses is
537 * in the bind address list of the association. If so,
538 * do not send the asconf chunk to its peer, but continue with
539 * other associations.
542 for (i = 0; i < addrcnt; i++) {
544 af = sctp_get_af_specific(addr->v4.sin_family);
550 if (sctp_assoc_lookup_laddr(asoc, addr))
553 addr_buf += af->sockaddr_len;
558 /* Use the first valid address in bind addr list of
559 * association as Address Parameter of ASCONF CHUNK.
561 bp = &asoc->base.bind_addr;
562 p = bp->address_list.next;
563 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
564 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
565 addrcnt, SCTP_PARAM_ADD_IP);
571 /* Add the new addresses to the bind address list with
572 * use_as_src set to 0.
575 for (i = 0; i < addrcnt; i++) {
577 af = sctp_get_af_specific(addr->v4.sin_family);
578 memcpy(&saveaddr, addr, af->sockaddr_len);
579 retval = sctp_add_bind_addr(bp, &saveaddr,
580 SCTP_ADDR_NEW, GFP_ATOMIC);
581 addr_buf += af->sockaddr_len;
583 if (asoc->src_out_of_asoc_ok) {
584 struct sctp_transport *trans;
586 list_for_each_entry(trans,
587 &asoc->peer.transport_addr_list, transports) {
588 /* Clear the source and route cache */
589 dst_release(trans->dst);
590 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
591 2*asoc->pathmtu, 4380));
592 trans->ssthresh = asoc->peer.i.a_rwnd;
593 trans->rto = asoc->rto_initial;
594 sctp_max_rto(asoc, trans);
595 trans->rtt = trans->srtt = trans->rttvar = 0;
596 sctp_transport_route(trans, NULL,
597 sctp_sk(asoc->base.sk));
600 retval = sctp_send_asconf(asoc, chunk);
607 /* Remove a list of addresses from bind addresses list. Do not remove the
610 * Basically run through each address specified in the addrs/addrcnt
611 * array/length pair, determine if it is IPv6 or IPv4 and call
612 * sctp_del_bind() on it.
614 * If any of them fails, then the operation will be reversed and the
615 * ones that were removed will be added back.
617 * At least one address has to be left; if only one address is
618 * available, the operation will return -EBUSY.
620 * Only sctp_setsockopt_bindx() is supposed to call this function.
622 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
624 struct sctp_sock *sp = sctp_sk(sk);
625 struct sctp_endpoint *ep = sp->ep;
627 struct sctp_bind_addr *bp = &ep->base.bind_addr;
630 union sctp_addr *sa_addr;
633 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
634 __func__, sk, addrs, addrcnt);
637 for (cnt = 0; cnt < addrcnt; cnt++) {
638 /* If the bind address list is empty or if there is only one
639 * bind address, there is nothing more to be removed (we need
640 * at least one address here).
642 if (list_empty(&bp->address_list) ||
643 (sctp_list_single_entry(&bp->address_list))) {
649 af = sctp_get_af_specific(sa_addr->sa.sa_family);
655 if (!af->addr_valid(sa_addr, sp, NULL)) {
656 retval = -EADDRNOTAVAIL;
660 if (sa_addr->v4.sin_port &&
661 sa_addr->v4.sin_port != htons(bp->port)) {
666 if (!sa_addr->v4.sin_port)
667 sa_addr->v4.sin_port = htons(bp->port);
669 /* FIXME - There is probably a need to check if sk->sk_saddr and
670 * sk->sk_rcv_addr are currently set to one of the addresses to
671 * be removed. This is something which needs to be looked into
672 * when we are fixing the outstanding issues with multi-homing
673 * socket routing and failover schemes. Refer to comments in
674 * sctp_do_bind(). -daisy
676 retval = sctp_del_bind_addr(bp, sa_addr);
678 addr_buf += af->sockaddr_len;
681 /* Failed. Add the ones that has been removed back */
683 sctp_bindx_add(sk, addrs, cnt);
691 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
692 * the associations that are part of the endpoint indicating that a list of
693 * local addresses are removed from the endpoint.
695 * If any of the addresses is already in the bind address list of the
696 * association, we do not send the chunk for that association. But it will not
697 * affect other associations.
699 * Only sctp_setsockopt_bindx() is supposed to call this function.
701 static int sctp_send_asconf_del_ip(struct sock *sk,
702 struct sockaddr *addrs,
705 struct net *net = sock_net(sk);
706 struct sctp_sock *sp;
707 struct sctp_endpoint *ep;
708 struct sctp_association *asoc;
709 struct sctp_transport *transport;
710 struct sctp_bind_addr *bp;
711 struct sctp_chunk *chunk;
712 union sctp_addr *laddr;
715 struct sctp_sockaddr_entry *saddr;
721 if (!net->sctp.addip_enable)
727 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
728 __func__, sk, addrs, addrcnt);
730 list_for_each_entry(asoc, &ep->asocs, asocs) {
732 if (!asoc->peer.asconf_capable)
735 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
738 if (!sctp_state(asoc, ESTABLISHED))
741 /* Check if any address in the packed array of addresses is
742 * not present in the bind address list of the association.
743 * If so, do not send the asconf chunk to its peer, but
744 * continue with other associations.
747 for (i = 0; i < addrcnt; i++) {
749 af = sctp_get_af_specific(laddr->v4.sin_family);
755 if (!sctp_assoc_lookup_laddr(asoc, laddr))
758 addr_buf += af->sockaddr_len;
763 /* Find one address in the association's bind address list
764 * that is not in the packed array of addresses. This is to
765 * make sure that we do not delete all the addresses in the
768 bp = &asoc->base.bind_addr;
769 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
771 if ((laddr == NULL) && (addrcnt == 1)) {
772 if (asoc->asconf_addr_del_pending)
774 asoc->asconf_addr_del_pending =
775 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
776 if (asoc->asconf_addr_del_pending == NULL) {
780 asoc->asconf_addr_del_pending->sa.sa_family =
782 asoc->asconf_addr_del_pending->v4.sin_port =
784 if (addrs->sa_family == AF_INET) {
785 struct sockaddr_in *sin;
787 sin = (struct sockaddr_in *)addrs;
788 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
789 } else if (addrs->sa_family == AF_INET6) {
790 struct sockaddr_in6 *sin6;
792 sin6 = (struct sockaddr_in6 *)addrs;
793 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
796 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
797 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
798 asoc->asconf_addr_del_pending);
800 asoc->src_out_of_asoc_ok = 1;
808 /* We do not need RCU protection throughout this loop
809 * because this is done under a socket lock from the
812 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
820 /* Reset use_as_src flag for the addresses in the bind address
821 * list that are to be deleted.
824 for (i = 0; i < addrcnt; i++) {
826 af = sctp_get_af_specific(laddr->v4.sin_family);
827 list_for_each_entry(saddr, &bp->address_list, list) {
828 if (sctp_cmp_addr_exact(&saddr->a, laddr))
829 saddr->state = SCTP_ADDR_DEL;
831 addr_buf += af->sockaddr_len;
834 /* Update the route and saddr entries for all the transports
835 * as some of the addresses in the bind address list are
836 * about to be deleted and cannot be used as source addresses.
838 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
840 dst_release(transport->dst);
841 sctp_transport_route(transport, NULL,
842 sctp_sk(asoc->base.sk));
846 /* We don't need to transmit ASCONF */
848 retval = sctp_send_asconf(asoc, chunk);
854 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
855 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
857 struct sock *sk = sctp_opt2sk(sp);
858 union sctp_addr *addr;
861 /* It is safe to write port space in caller. */
863 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
864 af = sctp_get_af_specific(addr->sa.sa_family);
867 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
870 if (addrw->state == SCTP_ADDR_NEW)
871 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
873 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
876 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
879 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
882 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
883 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
886 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
887 * Section 3.1.2 for this usage.
889 * addrs is a pointer to an array of one or more socket addresses. Each
890 * address is contained in its appropriate structure (i.e. struct
891 * sockaddr_in or struct sockaddr_in6) the family of the address type
892 * must be used to distinguish the address length (note that this
893 * representation is termed a "packed array" of addresses). The caller
894 * specifies the number of addresses in the array with addrcnt.
896 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
897 * -1, and sets errno to the appropriate error code.
899 * For SCTP, the port given in each socket address must be the same, or
900 * sctp_bindx() will fail, setting errno to EINVAL.
902 * The flags parameter is formed from the bitwise OR of zero or more of
903 * the following currently defined flags:
905 * SCTP_BINDX_ADD_ADDR
907 * SCTP_BINDX_REM_ADDR
909 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
910 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
911 * addresses from the association. The two flags are mutually exclusive;
912 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
913 * not remove all addresses from an association; sctp_bindx() will
914 * reject such an attempt with EINVAL.
916 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
917 * additional addresses with an endpoint after calling bind(). Or use
918 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
919 * socket is associated with so that no new association accepted will be
920 * associated with those addresses. If the endpoint supports dynamic
921 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
922 * endpoint to send the appropriate message to the peer to change the
923 * peers address lists.
925 * Adding and removing addresses from a connected association is
926 * optional functionality. Implementations that do not support this
927 * functionality should return EOPNOTSUPP.
929 * Basically do nothing but copying the addresses from user to kernel
930 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
931 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
934 * We don't use copy_from_user() for optimization: we first do the
935 * sanity checks (buffer size -fast- and access check-healthy
936 * pointer); if all of those succeed, then we can alloc the memory
937 * (expensive operation) needed to copy the data to kernel. Then we do
938 * the copying without checking the user space area
939 * (__copy_from_user()).
941 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
944 * sk The sk of the socket
945 * addrs The pointer to the addresses in user land
946 * addrssize Size of the addrs buffer
947 * op Operation to perform (add or remove, see the flags of
950 * Returns 0 if ok, <0 errno code on error.
952 static int sctp_setsockopt_bindx(struct sock *sk,
953 struct sockaddr __user *addrs,
954 int addrs_size, int op)
956 struct sockaddr *kaddrs;
960 struct sockaddr *sa_addr;
964 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
965 __func__, sk, addrs, addrs_size, op);
967 if (unlikely(addrs_size <= 0))
970 /* Check the user passed a healthy pointer. */
971 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
974 /* Alloc space for the address array in kernel memory. */
975 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
976 if (unlikely(!kaddrs))
979 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
984 /* Walk through the addrs buffer and count the number of addresses. */
986 while (walk_size < addrs_size) {
987 if (walk_size + sizeof(sa_family_t) > addrs_size) {
993 af = sctp_get_af_specific(sa_addr->sa_family);
995 /* If the address family is not supported or if this address
996 * causes the address buffer to overflow return EINVAL.
998 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1003 addr_buf += af->sockaddr_len;
1004 walk_size += af->sockaddr_len;
1009 case SCTP_BINDX_ADD_ADDR:
1010 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1013 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1016 case SCTP_BINDX_REM_ADDR:
1017 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1020 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1034 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1036 * Common routine for handling connect() and sctp_connectx().
1037 * Connect will come in with just a single address.
1039 static int __sctp_connect(struct sock *sk,
1040 struct sockaddr *kaddrs,
1042 sctp_assoc_t *assoc_id)
1044 struct net *net = sock_net(sk);
1045 struct sctp_sock *sp;
1046 struct sctp_endpoint *ep;
1047 struct sctp_association *asoc = NULL;
1048 struct sctp_association *asoc2;
1049 struct sctp_transport *transport;
1056 union sctp_addr *sa_addr = NULL;
1058 unsigned short port;
1059 unsigned int f_flags = 0;
1064 /* connect() cannot be done on a socket that is already in ESTABLISHED
1065 * state - UDP-style peeled off socket or a TCP-style socket that
1066 * is already connected.
1067 * It cannot be done even on a TCP-style listening socket.
1069 if (sctp_sstate(sk, ESTABLISHED) ||
1070 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1075 /* Walk through the addrs buffer and count the number of addresses. */
1077 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 sp->pf->to_sk_daddr(sa_addr, sk);
1207 /* in-kernel sockets don't generally have a file allocated to them
1208 * if all they do is call sock_create_kern().
1210 if (sk->sk_socket->file)
1211 f_flags = sk->sk_socket->file->f_flags;
1213 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1215 err = sctp_wait_for_connect(asoc, &timeo);
1216 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1217 *assoc_id = asoc->assoc_id;
1219 /* Don't free association on exit. */
1223 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1224 __func__, asoc, kaddrs, err);
1227 /* sctp_primitive_ASSOCIATE may have added this association
1228 * To the hash table, try to unhash it, just in case, its a noop
1229 * if it wasn't hashed so we're safe
1231 sctp_unhash_established(asoc);
1232 sctp_association_free(asoc);
1237 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1240 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1241 * sctp_assoc_t *asoc);
1243 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1244 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1245 * or IPv6 addresses.
1247 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1248 * Section 3.1.2 for this usage.
1250 * addrs is a pointer to an array of one or more socket addresses. Each
1251 * address is contained in its appropriate structure (i.e. struct
1252 * sockaddr_in or struct sockaddr_in6) the family of the address type
1253 * must be used to distengish the address length (note that this
1254 * representation is termed a "packed array" of addresses). The caller
1255 * specifies the number of addresses in the array with addrcnt.
1257 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1258 * the association id of the new association. On failure, sctp_connectx()
1259 * returns -1, and sets errno to the appropriate error code. The assoc_id
1260 * is not touched by the kernel.
1262 * For SCTP, the port given in each socket address must be the same, or
1263 * sctp_connectx() will fail, setting errno to EINVAL.
1265 * An application can use sctp_connectx to initiate an association with
1266 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1267 * allows a caller to specify multiple addresses at which a peer can be
1268 * reached. The way the SCTP stack uses the list of addresses to set up
1269 * the association is implementation dependent. This function only
1270 * specifies that the stack will try to make use of all the addresses in
1271 * the list when needed.
1273 * Note that the list of addresses passed in is only used for setting up
1274 * the association. It does not necessarily equal the set of addresses
1275 * the peer uses for the resulting association. If the caller wants to
1276 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1277 * retrieve them after the association has been set up.
1279 * Basically do nothing but copying the addresses from user to kernel
1280 * land and invoking either sctp_connectx(). This is used for tunneling
1281 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1283 * We don't use copy_from_user() for optimization: we first do the
1284 * sanity checks (buffer size -fast- and access check-healthy
1285 * pointer); if all of those succeed, then we can alloc the memory
1286 * (expensive operation) needed to copy the data to kernel. Then we do
1287 * the copying without checking the user space area
1288 * (__copy_from_user()).
1290 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1293 * sk The sk of the socket
1294 * addrs The pointer to the addresses in user land
1295 * addrssize Size of the addrs buffer
1297 * Returns >=0 if ok, <0 errno code on error.
1299 static int __sctp_setsockopt_connectx(struct sock *sk,
1300 struct sockaddr __user *addrs,
1302 sctp_assoc_t *assoc_id)
1305 struct sockaddr *kaddrs;
1307 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1308 __func__, sk, addrs, addrs_size);
1310 if (unlikely(addrs_size <= 0))
1313 /* Check the user passed a healthy pointer. */
1314 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1317 /* Alloc space for the address array in kernel memory. */
1318 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1319 if (unlikely(!kaddrs))
1322 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1325 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1334 * This is an older interface. It's kept for backward compatibility
1335 * to the option that doesn't provide association id.
1337 static int sctp_setsockopt_connectx_old(struct sock *sk,
1338 struct sockaddr __user *addrs,
1341 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1345 * New interface for the API. The since the API is done with a socket
1346 * option, to make it simple we feed back the association id is as a return
1347 * indication to the call. Error is always negative and association id is
1350 static int sctp_setsockopt_connectx(struct sock *sk,
1351 struct sockaddr __user *addrs,
1354 sctp_assoc_t assoc_id = 0;
1357 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1366 * New (hopefully final) interface for the API.
1367 * We use the sctp_getaddrs_old structure so that use-space library
1368 * can avoid any unnecessary allocations. The only different part
1369 * is that we store the actual length of the address buffer into the
1370 * addrs_num structure member. That way we can re-use the existing
1373 #ifdef CONFIG_COMPAT
1374 struct compat_sctp_getaddrs_old {
1375 sctp_assoc_t assoc_id;
1377 compat_uptr_t addrs; /* struct sockaddr * */
1381 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1382 char __user *optval,
1385 struct sctp_getaddrs_old param;
1386 sctp_assoc_t assoc_id = 0;
1389 #ifdef CONFIG_COMPAT
1390 if (is_compat_task()) {
1391 struct compat_sctp_getaddrs_old param32;
1393 if (len < sizeof(param32))
1395 if (copy_from_user(¶m32, optval, sizeof(param32)))
1398 param.assoc_id = param32.assoc_id;
1399 param.addr_num = param32.addr_num;
1400 param.addrs = compat_ptr(param32.addrs);
1404 if (len < sizeof(param))
1406 if (copy_from_user(¶m, optval, sizeof(param)))
1410 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1411 param.addrs, param.addr_num,
1413 if (err == 0 || err == -EINPROGRESS) {
1414 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1416 if (put_user(sizeof(assoc_id), optlen))
1423 /* API 3.1.4 close() - UDP Style Syntax
1424 * Applications use close() to perform graceful shutdown (as described in
1425 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1426 * by a UDP-style socket.
1430 * ret = close(int sd);
1432 * sd - the socket descriptor of the associations to be closed.
1434 * To gracefully shutdown a specific association represented by the
1435 * UDP-style socket, an application should use the sendmsg() call,
1436 * passing no user data, but including the appropriate flag in the
1437 * ancillary data (see Section xxxx).
1439 * If sd in the close() call is a branched-off socket representing only
1440 * one association, the shutdown is performed on that association only.
1442 * 4.1.6 close() - TCP Style Syntax
1444 * Applications use close() to gracefully close down an association.
1448 * int close(int sd);
1450 * sd - the socket descriptor of the association to be closed.
1452 * After an application calls close() on a socket descriptor, no further
1453 * socket operations will succeed on that descriptor.
1455 * API 7.1.4 SO_LINGER
1457 * An application using the TCP-style socket can use this option to
1458 * perform the SCTP ABORT primitive. The linger option structure is:
1461 * int l_onoff; // option on/off
1462 * int l_linger; // linger time
1465 * To enable the option, set l_onoff to 1. If the l_linger value is set
1466 * to 0, calling close() is the same as the ABORT primitive. If the
1467 * value is set to a negative value, the setsockopt() call will return
1468 * an error. If the value is set to a positive value linger_time, the
1469 * close() can be blocked for at most linger_time ms. If the graceful
1470 * shutdown phase does not finish during this period, close() will
1471 * return but the graceful shutdown phase continues in the system.
1473 static void sctp_close(struct sock *sk, long timeout)
1475 struct net *net = sock_net(sk);
1476 struct sctp_endpoint *ep;
1477 struct sctp_association *asoc;
1478 struct list_head *pos, *temp;
1479 unsigned int data_was_unread;
1481 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1484 sk->sk_shutdown = SHUTDOWN_MASK;
1485 sk->sk_state = SCTP_SS_CLOSING;
1487 ep = sctp_sk(sk)->ep;
1489 /* Clean up any skbs sitting on the receive queue. */
1490 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1491 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1493 /* Walk all associations on an endpoint. */
1494 list_for_each_safe(pos, temp, &ep->asocs) {
1495 asoc = list_entry(pos, struct sctp_association, asocs);
1497 if (sctp_style(sk, TCP)) {
1498 /* A closed association can still be in the list if
1499 * it belongs to a TCP-style listening socket that is
1500 * not yet accepted. If so, free it. If not, send an
1501 * ABORT or SHUTDOWN based on the linger options.
1503 if (sctp_state(asoc, CLOSED)) {
1504 sctp_unhash_established(asoc);
1505 sctp_association_free(asoc);
1510 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1511 !skb_queue_empty(&asoc->ulpq.reasm) ||
1512 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1513 struct sctp_chunk *chunk;
1515 chunk = sctp_make_abort_user(asoc, NULL, 0);
1517 sctp_primitive_ABORT(net, asoc, chunk);
1519 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1522 /* On a TCP-style socket, block for at most linger_time if set. */
1523 if (sctp_style(sk, TCP) && timeout)
1524 sctp_wait_for_close(sk, timeout);
1526 /* This will run the backlog queue. */
1529 /* Supposedly, no process has access to the socket, but
1530 * the net layers still may.
1535 /* Hold the sock, since sk_common_release() will put sock_put()
1536 * and we have just a little more cleanup.
1539 sk_common_release(sk);
1546 SCTP_DBG_OBJCNT_DEC(sock);
1549 /* Handle EPIPE error. */
1550 static int sctp_error(struct sock *sk, int flags, int err)
1553 err = sock_error(sk) ? : -EPIPE;
1554 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1555 send_sig(SIGPIPE, current, 0);
1559 /* API 3.1.3 sendmsg() - UDP Style Syntax
1561 * An application uses sendmsg() and recvmsg() calls to transmit data to
1562 * and receive data from its peer.
1564 * ssize_t sendmsg(int socket, const struct msghdr *message,
1567 * socket - the socket descriptor of the endpoint.
1568 * message - pointer to the msghdr structure which contains a single
1569 * user message and possibly some ancillary data.
1571 * See Section 5 for complete description of the data
1574 * flags - flags sent or received with the user message, see Section
1575 * 5 for complete description of the flags.
1577 * Note: This function could use a rewrite especially when explicit
1578 * connect support comes in.
1580 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1582 static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1584 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
1586 struct net *net = sock_net(sk);
1587 struct sctp_sock *sp;
1588 struct sctp_endpoint *ep;
1589 struct sctp_association *new_asoc = NULL, *asoc = NULL;
1590 struct sctp_transport *transport, *chunk_tp;
1591 struct sctp_chunk *chunk;
1593 struct sockaddr *msg_name = NULL;
1594 struct sctp_sndrcvinfo default_sinfo;
1595 struct sctp_sndrcvinfo *sinfo;
1596 struct sctp_initmsg *sinit;
1597 sctp_assoc_t associd = 0;
1598 sctp_cmsgs_t cmsgs = { NULL };
1600 bool fill_sinfo_ttl = false, wait_connect = false;
1601 struct sctp_datamsg *datamsg;
1602 int msg_flags = msg->msg_flags;
1603 __u16 sinfo_flags = 0;
1611 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1614 /* We cannot send a message over a TCP-style listening socket. */
1615 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1620 /* Parse out the SCTP CMSGs. */
1621 err = sctp_msghdr_parse(msg, &cmsgs);
1623 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1627 /* Fetch the destination address for this packet. This
1628 * address only selects the association--it is not necessarily
1629 * the address we will send to.
1630 * For a peeled-off socket, msg_name is ignored.
1632 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1633 int msg_namelen = msg->msg_namelen;
1635 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1640 if (msg_namelen > sizeof(to))
1641 msg_namelen = sizeof(to);
1642 memcpy(&to, msg->msg_name, msg_namelen);
1643 msg_name = msg->msg_name;
1647 if (cmsgs.sinfo != NULL) {
1648 memset(&default_sinfo, 0, sizeof(default_sinfo));
1649 default_sinfo.sinfo_stream = cmsgs.sinfo->snd_sid;
1650 default_sinfo.sinfo_flags = cmsgs.sinfo->snd_flags;
1651 default_sinfo.sinfo_ppid = cmsgs.sinfo->snd_ppid;
1652 default_sinfo.sinfo_context = cmsgs.sinfo->snd_context;
1653 default_sinfo.sinfo_assoc_id = cmsgs.sinfo->snd_assoc_id;
1655 sinfo = &default_sinfo;
1656 fill_sinfo_ttl = true;
1658 sinfo = cmsgs.srinfo;
1660 /* Did the user specify SNDINFO/SNDRCVINFO? */
1662 sinfo_flags = sinfo->sinfo_flags;
1663 associd = sinfo->sinfo_assoc_id;
1666 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1667 msg_len, sinfo_flags);
1669 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1670 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1675 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1676 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1677 * If SCTP_ABORT is set, the message length could be non zero with
1678 * the msg_iov set to the user abort reason.
1680 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1681 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1686 /* If SCTP_ADDR_OVER is set, there must be an address
1687 * specified in msg_name.
1689 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1696 pr_debug("%s: about to look up association\n", __func__);
1700 /* If a msg_name has been specified, assume this is to be used. */
1702 /* Look for a matching association on the endpoint. */
1703 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1705 /* If we could not find a matching association on the
1706 * endpoint, make sure that it is not a TCP-style
1707 * socket that already has an association or there is
1708 * no peeled-off association on another socket.
1710 if ((sctp_style(sk, TCP) &&
1711 sctp_sstate(sk, ESTABLISHED)) ||
1712 sctp_endpoint_is_peeled_off(ep, &to)) {
1713 err = -EADDRNOTAVAIL;
1718 asoc = sctp_id2assoc(sk, associd);
1726 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1728 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1729 * socket that has an association in CLOSED state. This can
1730 * happen when an accepted socket has an association that is
1733 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1738 if (sinfo_flags & SCTP_EOF) {
1739 pr_debug("%s: shutting down association:%p\n",
1742 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1746 if (sinfo_flags & SCTP_ABORT) {
1748 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1754 pr_debug("%s: aborting association:%p\n",
1757 sctp_primitive_ABORT(net, asoc, chunk);
1763 /* Do we need to create the association? */
1765 pr_debug("%s: there is no association yet\n", __func__);
1767 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1772 /* Check for invalid stream against the stream counts,
1773 * either the default or the user specified stream counts.
1776 if (!sinit || !sinit->sinit_num_ostreams) {
1777 /* Check against the defaults. */
1778 if (sinfo->sinfo_stream >=
1779 sp->initmsg.sinit_num_ostreams) {
1784 /* Check against the requested. */
1785 if (sinfo->sinfo_stream >=
1786 sinit->sinit_num_ostreams) {
1794 * API 3.1.2 bind() - UDP Style Syntax
1795 * If a bind() or sctp_bindx() is not called prior to a
1796 * sendmsg() call that initiates a new association, the
1797 * system picks an ephemeral port and will choose an address
1798 * set equivalent to binding with a wildcard address.
1800 if (!ep->base.bind_addr.port) {
1801 if (sctp_autobind(sk)) {
1807 * If an unprivileged user inherits a one-to-many
1808 * style socket with open associations on a privileged
1809 * port, it MAY be permitted to accept new associations,
1810 * but it SHOULD NOT be permitted to open new
1813 if (ep->base.bind_addr.port < PROT_SOCK &&
1814 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1820 scope = sctp_scope(&to);
1821 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1827 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1833 /* If the SCTP_INIT ancillary data is specified, set all
1834 * the association init values accordingly.
1837 if (sinit->sinit_num_ostreams) {
1838 asoc->c.sinit_num_ostreams =
1839 sinit->sinit_num_ostreams;
1841 if (sinit->sinit_max_instreams) {
1842 asoc->c.sinit_max_instreams =
1843 sinit->sinit_max_instreams;
1845 if (sinit->sinit_max_attempts) {
1846 asoc->max_init_attempts
1847 = sinit->sinit_max_attempts;
1849 if (sinit->sinit_max_init_timeo) {
1850 asoc->max_init_timeo =
1851 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1855 /* Prime the peer's transport structures. */
1856 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1863 /* ASSERT: we have a valid association at this point. */
1864 pr_debug("%s: we have a valid association\n", __func__);
1867 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1868 * one with some defaults.
1870 memset(&default_sinfo, 0, sizeof(default_sinfo));
1871 default_sinfo.sinfo_stream = asoc->default_stream;
1872 default_sinfo.sinfo_flags = asoc->default_flags;
1873 default_sinfo.sinfo_ppid = asoc->default_ppid;
1874 default_sinfo.sinfo_context = asoc->default_context;
1875 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1876 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1878 sinfo = &default_sinfo;
1879 } else if (fill_sinfo_ttl) {
1880 /* In case SNDINFO was specified, we still need to fill
1881 * it with a default ttl from the assoc here.
1883 sinfo->sinfo_timetolive = asoc->default_timetolive;
1886 /* API 7.1.7, the sndbuf size per association bounds the
1887 * maximum size of data that can be sent in a single send call.
1889 if (msg_len > sk->sk_sndbuf) {
1894 if (asoc->pmtu_pending)
1895 sctp_assoc_pending_pmtu(sk, asoc);
1897 /* If fragmentation is disabled and the message length exceeds the
1898 * association fragmentation point, return EMSGSIZE. The I-D
1899 * does not specify what this error is, but this looks like
1902 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1907 /* Check for invalid stream. */
1908 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1913 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1914 if (!sctp_wspace(asoc)) {
1915 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1920 /* If an address is passed with the sendto/sendmsg call, it is used
1921 * to override the primary destination address in the TCP model, or
1922 * when SCTP_ADDR_OVER flag is set in the UDP model.
1924 if ((sctp_style(sk, TCP) && msg_name) ||
1925 (sinfo_flags & SCTP_ADDR_OVER)) {
1926 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1934 /* Auto-connect, if we aren't connected already. */
1935 if (sctp_state(asoc, CLOSED)) {
1936 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1940 wait_connect = true;
1941 pr_debug("%s: we associated primitively\n", __func__);
1944 /* Break the message into multiple chunks of maximum size. */
1945 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
1946 if (IS_ERR(datamsg)) {
1947 err = PTR_ERR(datamsg);
1951 /* Now send the (possibly) fragmented message. */
1952 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1953 sctp_chunk_hold(chunk);
1955 /* Do accounting for the write space. */
1956 sctp_set_owner_w(chunk);
1958 chunk->transport = chunk_tp;
1961 /* Send it to the lower layers. Note: all chunks
1962 * must either fail or succeed. The lower layer
1963 * works that way today. Keep it that way or this
1966 err = sctp_primitive_SEND(net, asoc, datamsg);
1967 /* Did the lower layer accept the chunk? */
1969 sctp_datamsg_free(datamsg);
1973 pr_debug("%s: we sent primitively\n", __func__);
1975 sctp_datamsg_put(datamsg);
1978 if (unlikely(wait_connect)) {
1979 timeo = sock_sndtimeo(sk, msg_flags & MSG_DONTWAIT);
1980 sctp_wait_for_connect(asoc, &timeo);
1983 /* If we are already past ASSOCIATE, the lower
1984 * layers are responsible for association cleanup.
1990 sctp_unhash_established(asoc);
1991 sctp_association_free(asoc);
1997 return sctp_error(sk, msg_flags, err);
2004 err = sock_error(sk);
2014 /* This is an extended version of skb_pull() that removes the data from the
2015 * start of a skb even when data is spread across the list of skb's in the
2016 * frag_list. len specifies the total amount of data that needs to be removed.
2017 * when 'len' bytes could be removed from the skb, it returns 0.
2018 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2019 * could not be removed.
2021 static int sctp_skb_pull(struct sk_buff *skb, int len)
2023 struct sk_buff *list;
2024 int skb_len = skb_headlen(skb);
2027 if (len <= skb_len) {
2028 __skb_pull(skb, len);
2032 __skb_pull(skb, skb_len);
2034 skb_walk_frags(skb, list) {
2035 rlen = sctp_skb_pull(list, len);
2036 skb->len -= (len-rlen);
2037 skb->data_len -= (len-rlen);
2048 /* API 3.1.3 recvmsg() - UDP Style Syntax
2050 * ssize_t recvmsg(int socket, struct msghdr *message,
2053 * socket - the socket descriptor of the endpoint.
2054 * message - pointer to the msghdr structure which contains a single
2055 * user message and possibly some ancillary data.
2057 * See Section 5 for complete description of the data
2060 * flags - flags sent or received with the user message, see Section
2061 * 5 for complete description of the flags.
2063 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2064 int noblock, int flags, int *addr_len)
2066 struct sctp_ulpevent *event = NULL;
2067 struct sctp_sock *sp = sctp_sk(sk);
2068 struct sk_buff *skb;
2073 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2074 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2079 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2084 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2088 /* Get the total length of the skb including any skb's in the
2097 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2099 event = sctp_skb2event(skb);
2104 sock_recv_ts_and_drops(msg, sk, skb);
2105 if (sctp_ulpevent_is_notification(event)) {
2106 msg->msg_flags |= MSG_NOTIFICATION;
2107 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2109 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2112 /* Check if we allow SCTP_NXTINFO. */
2113 if (sp->recvnxtinfo)
2114 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2115 /* Check if we allow SCTP_RCVINFO. */
2116 if (sp->recvrcvinfo)
2117 sctp_ulpevent_read_rcvinfo(event, msg);
2118 /* Check if we allow SCTP_SNDRCVINFO. */
2119 if (sp->subscribe.sctp_data_io_event)
2120 sctp_ulpevent_read_sndrcvinfo(event, msg);
2123 /* FIXME: we should be calling IP/IPv6 layers. */
2124 if (sk->sk_protinfo.af_inet.cmsg_flags)
2125 ip_cmsg_recv(msg, skb);
2130 /* If skb's length exceeds the user's buffer, update the skb and
2131 * push it back to the receive_queue so that the next call to
2132 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2134 if (skb_len > copied) {
2135 msg->msg_flags &= ~MSG_EOR;
2136 if (flags & MSG_PEEK)
2138 sctp_skb_pull(skb, copied);
2139 skb_queue_head(&sk->sk_receive_queue, skb);
2141 /* When only partial message is copied to the user, increase
2142 * rwnd by that amount. If all the data in the skb is read,
2143 * rwnd is updated when the event is freed.
2145 if (!sctp_ulpevent_is_notification(event))
2146 sctp_assoc_rwnd_increase(event->asoc, copied);
2148 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2149 (event->msg_flags & MSG_EOR))
2150 msg->msg_flags |= MSG_EOR;
2152 msg->msg_flags &= ~MSG_EOR;
2155 if (flags & MSG_PEEK) {
2156 /* Release the skb reference acquired after peeking the skb in
2157 * sctp_skb_recv_datagram().
2161 /* Free the event which includes releasing the reference to
2162 * the owner of the skb, freeing the skb and updating the
2165 sctp_ulpevent_free(event);
2172 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2174 * This option is a on/off flag. If enabled no SCTP message
2175 * fragmentation will be performed. Instead if a message being sent
2176 * exceeds the current PMTU size, the message will NOT be sent and
2177 * instead a error will be indicated to the user.
2179 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2180 char __user *optval,
2181 unsigned int optlen)
2185 if (optlen < sizeof(int))
2188 if (get_user(val, (int __user *)optval))
2191 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2196 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2197 unsigned int optlen)
2199 struct sctp_association *asoc;
2200 struct sctp_ulpevent *event;
2202 if (optlen > sizeof(struct sctp_event_subscribe))
2204 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2207 if (sctp_sk(sk)->subscribe.sctp_data_io_event)
2208 pr_warn_ratelimited(DEPRECATED "%s (pid %d) "
2209 "Requested SCTP_SNDRCVINFO event.\n"
2210 "Use SCTP_RCVINFO through SCTP_RECVRCVINFO option instead.\n",
2211 current->comm, task_pid_nr(current));
2213 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2214 * if there is no data to be sent or retransmit, the stack will
2215 * immediately send up this notification.
2217 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2218 &sctp_sk(sk)->subscribe)) {
2219 asoc = sctp_id2assoc(sk, 0);
2221 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2222 event = sctp_ulpevent_make_sender_dry_event(asoc,
2227 sctp_ulpq_tail_event(&asoc->ulpq, event);
2234 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2236 * This socket option is applicable to the UDP-style socket only. When
2237 * set it will cause associations that are idle for more than the
2238 * specified number of seconds to automatically close. An association
2239 * being idle is defined an association that has NOT sent or received
2240 * user data. The special value of '0' indicates that no automatic
2241 * close of any associations should be performed. The option expects an
2242 * integer defining the number of seconds of idle time before an
2243 * association is closed.
2245 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2246 unsigned int optlen)
2248 struct sctp_sock *sp = sctp_sk(sk);
2249 struct net *net = sock_net(sk);
2251 /* Applicable to UDP-style socket only */
2252 if (sctp_style(sk, TCP))
2254 if (optlen != sizeof(int))
2256 if (copy_from_user(&sp->autoclose, optval, optlen))
2259 if (sp->autoclose > net->sctp.max_autoclose)
2260 sp->autoclose = net->sctp.max_autoclose;
2265 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2267 * Applications can enable or disable heartbeats for any peer address of
2268 * an association, modify an address's heartbeat interval, force a
2269 * heartbeat to be sent immediately, and adjust the address's maximum
2270 * number of retransmissions sent before an address is considered
2271 * unreachable. The following structure is used to access and modify an
2272 * address's parameters:
2274 * struct sctp_paddrparams {
2275 * sctp_assoc_t spp_assoc_id;
2276 * struct sockaddr_storage spp_address;
2277 * uint32_t spp_hbinterval;
2278 * uint16_t spp_pathmaxrxt;
2279 * uint32_t spp_pathmtu;
2280 * uint32_t spp_sackdelay;
2281 * uint32_t spp_flags;
2284 * spp_assoc_id - (one-to-many style socket) This is filled in the
2285 * application, and identifies the association for
2287 * spp_address - This specifies which address is of interest.
2288 * spp_hbinterval - This contains the value of the heartbeat interval,
2289 * in milliseconds. If a value of zero
2290 * is present in this field then no changes are to
2291 * be made to this parameter.
2292 * spp_pathmaxrxt - This contains the maximum number of
2293 * retransmissions before this address shall be
2294 * considered unreachable. If a value of zero
2295 * is present in this field then no changes are to
2296 * be made to this parameter.
2297 * spp_pathmtu - When Path MTU discovery is disabled the value
2298 * specified here will be the "fixed" path mtu.
2299 * Note that if the spp_address field is empty
2300 * then all associations on this address will
2301 * have this fixed path mtu set upon them.
2303 * spp_sackdelay - When delayed sack is enabled, this value specifies
2304 * the number of milliseconds that sacks will be delayed
2305 * for. This value will apply to all addresses of an
2306 * association if the spp_address field is empty. Note
2307 * also, that if delayed sack is enabled and this
2308 * value is set to 0, no change is made to the last
2309 * recorded delayed sack timer value.
2311 * spp_flags - These flags are used to control various features
2312 * on an association. The flag field may contain
2313 * zero or more of the following options.
2315 * SPP_HB_ENABLE - Enable heartbeats on the
2316 * specified address. Note that if the address
2317 * field is empty all addresses for the association
2318 * have heartbeats enabled upon them.
2320 * SPP_HB_DISABLE - Disable heartbeats on the
2321 * speicifed address. Note that if the address
2322 * field is empty all addresses for the association
2323 * will have their heartbeats disabled. Note also
2324 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2325 * mutually exclusive, only one of these two should
2326 * be specified. Enabling both fields will have
2327 * undetermined results.
2329 * SPP_HB_DEMAND - Request a user initiated heartbeat
2330 * to be made immediately.
2332 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2333 * heartbeat delayis to be set to the value of 0
2336 * SPP_PMTUD_ENABLE - This field will enable PMTU
2337 * discovery upon the specified address. Note that
2338 * if the address feild is empty then all addresses
2339 * on the association are effected.
2341 * SPP_PMTUD_DISABLE - This field will disable PMTU
2342 * discovery upon the specified address. Note that
2343 * if the address feild is empty then all addresses
2344 * on the association are effected. Not also that
2345 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2346 * exclusive. Enabling both will have undetermined
2349 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2350 * on delayed sack. The time specified in spp_sackdelay
2351 * is used to specify the sack delay for this address. Note
2352 * that if spp_address is empty then all addresses will
2353 * enable delayed sack and take on the sack delay
2354 * value specified in spp_sackdelay.
2355 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2356 * off delayed sack. If the spp_address field is blank then
2357 * delayed sack is disabled for the entire association. Note
2358 * also that this field is mutually exclusive to
2359 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2362 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2363 struct sctp_transport *trans,
2364 struct sctp_association *asoc,
2365 struct sctp_sock *sp,
2368 int sackdelay_change)
2372 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2373 struct net *net = sock_net(trans->asoc->base.sk);
2375 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2380 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2381 * this field is ignored. Note also that a value of zero indicates
2382 * the current setting should be left unchanged.
2384 if (params->spp_flags & SPP_HB_ENABLE) {
2386 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2387 * set. This lets us use 0 value when this flag
2390 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2391 params->spp_hbinterval = 0;
2393 if (params->spp_hbinterval ||
2394 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2397 msecs_to_jiffies(params->spp_hbinterval);
2400 msecs_to_jiffies(params->spp_hbinterval);
2402 sp->hbinterval = params->spp_hbinterval;
2409 trans->param_flags =
2410 (trans->param_flags & ~SPP_HB) | hb_change;
2413 (asoc->param_flags & ~SPP_HB) | hb_change;
2416 (sp->param_flags & ~SPP_HB) | hb_change;
2420 /* When Path MTU discovery is disabled the value specified here will
2421 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2422 * include the flag SPP_PMTUD_DISABLE for this field to have any
2425 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2427 trans->pathmtu = params->spp_pathmtu;
2428 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2430 asoc->pathmtu = params->spp_pathmtu;
2431 sctp_frag_point(asoc, params->spp_pathmtu);
2433 sp->pathmtu = params->spp_pathmtu;
2439 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2440 (params->spp_flags & SPP_PMTUD_ENABLE);
2441 trans->param_flags =
2442 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2444 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2445 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2449 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2452 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2456 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2457 * value of this field is ignored. Note also that a value of zero
2458 * indicates the current setting should be left unchanged.
2460 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2463 msecs_to_jiffies(params->spp_sackdelay);
2466 msecs_to_jiffies(params->spp_sackdelay);
2468 sp->sackdelay = params->spp_sackdelay;
2472 if (sackdelay_change) {
2474 trans->param_flags =
2475 (trans->param_flags & ~SPP_SACKDELAY) |
2479 (asoc->param_flags & ~SPP_SACKDELAY) |
2483 (sp->param_flags & ~SPP_SACKDELAY) |
2488 /* Note that a value of zero indicates the current setting should be
2491 if (params->spp_pathmaxrxt) {
2493 trans->pathmaxrxt = params->spp_pathmaxrxt;
2495 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2497 sp->pathmaxrxt = params->spp_pathmaxrxt;
2504 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2505 char __user *optval,
2506 unsigned int optlen)
2508 struct sctp_paddrparams params;
2509 struct sctp_transport *trans = NULL;
2510 struct sctp_association *asoc = NULL;
2511 struct sctp_sock *sp = sctp_sk(sk);
2513 int hb_change, pmtud_change, sackdelay_change;
2515 if (optlen != sizeof(struct sctp_paddrparams))
2518 if (copy_from_user(¶ms, optval, optlen))
2521 /* Validate flags and value parameters. */
2522 hb_change = params.spp_flags & SPP_HB;
2523 pmtud_change = params.spp_flags & SPP_PMTUD;
2524 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2526 if (hb_change == SPP_HB ||
2527 pmtud_change == SPP_PMTUD ||
2528 sackdelay_change == SPP_SACKDELAY ||
2529 params.spp_sackdelay > 500 ||
2530 (params.spp_pathmtu &&
2531 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2534 /* If an address other than INADDR_ANY is specified, and
2535 * no transport is found, then the request is invalid.
2537 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
2538 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2539 params.spp_assoc_id);
2544 /* Get association, if assoc_id != 0 and the socket is a one
2545 * to many style socket, and an association was not found, then
2546 * the id was invalid.
2548 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2549 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2552 /* Heartbeat demand can only be sent on a transport or
2553 * association, but not a socket.
2555 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2558 /* Process parameters. */
2559 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2560 hb_change, pmtud_change,
2566 /* If changes are for association, also apply parameters to each
2569 if (!trans && asoc) {
2570 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2572 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2573 hb_change, pmtud_change,
2581 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2583 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2586 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2588 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2592 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2594 * This option will effect the way delayed acks are performed. This
2595 * option allows you to get or set the delayed ack time, in
2596 * milliseconds. It also allows changing the delayed ack frequency.
2597 * Changing the frequency to 1 disables the delayed sack algorithm. If
2598 * the assoc_id is 0, then this sets or gets the endpoints default
2599 * values. If the assoc_id field is non-zero, then the set or get
2600 * effects the specified association for the one to many model (the
2601 * assoc_id field is ignored by the one to one model). Note that if
2602 * sack_delay or sack_freq are 0 when setting this option, then the
2603 * current values will remain unchanged.
2605 * struct sctp_sack_info {
2606 * sctp_assoc_t sack_assoc_id;
2607 * uint32_t sack_delay;
2608 * uint32_t sack_freq;
2611 * sack_assoc_id - This parameter, indicates which association the user
2612 * is performing an action upon. Note that if this field's value is
2613 * zero then the endpoints default value is changed (effecting future
2614 * associations only).
2616 * sack_delay - This parameter contains the number of milliseconds that
2617 * the user is requesting the delayed ACK timer be set to. Note that
2618 * this value is defined in the standard to be between 200 and 500
2621 * sack_freq - This parameter contains the number of packets that must
2622 * be received before a sack is sent without waiting for the delay
2623 * timer to expire. The default value for this is 2, setting this
2624 * value to 1 will disable the delayed sack algorithm.
2627 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2628 char __user *optval, unsigned int optlen)
2630 struct sctp_sack_info params;
2631 struct sctp_transport *trans = NULL;
2632 struct sctp_association *asoc = NULL;
2633 struct sctp_sock *sp = sctp_sk(sk);
2635 if (optlen == sizeof(struct sctp_sack_info)) {
2636 if (copy_from_user(¶ms, optval, optlen))
2639 if (params.sack_delay == 0 && params.sack_freq == 0)
2641 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2642 pr_warn_ratelimited(DEPRECATED
2644 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2645 "Use struct sctp_sack_info instead\n",
2646 current->comm, task_pid_nr(current));
2647 if (copy_from_user(¶ms, optval, optlen))
2650 if (params.sack_delay == 0)
2651 params.sack_freq = 1;
2653 params.sack_freq = 0;
2657 /* Validate value parameter. */
2658 if (params.sack_delay > 500)
2661 /* Get association, if sack_assoc_id != 0 and the socket is a one
2662 * to many style socket, and an association was not found, then
2663 * the id was invalid.
2665 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2666 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2669 if (params.sack_delay) {
2672 msecs_to_jiffies(params.sack_delay);
2674 sctp_spp_sackdelay_enable(asoc->param_flags);
2676 sp->sackdelay = params.sack_delay;
2678 sctp_spp_sackdelay_enable(sp->param_flags);
2682 if (params.sack_freq == 1) {
2685 sctp_spp_sackdelay_disable(asoc->param_flags);
2688 sctp_spp_sackdelay_disable(sp->param_flags);
2690 } else if (params.sack_freq > 1) {
2692 asoc->sackfreq = params.sack_freq;
2694 sctp_spp_sackdelay_enable(asoc->param_flags);
2696 sp->sackfreq = params.sack_freq;
2698 sctp_spp_sackdelay_enable(sp->param_flags);
2702 /* If change is for association, also apply to each transport. */
2704 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2706 if (params.sack_delay) {
2708 msecs_to_jiffies(params.sack_delay);
2709 trans->param_flags =
2710 sctp_spp_sackdelay_enable(trans->param_flags);
2712 if (params.sack_freq == 1) {
2713 trans->param_flags =
2714 sctp_spp_sackdelay_disable(trans->param_flags);
2715 } else if (params.sack_freq > 1) {
2716 trans->sackfreq = params.sack_freq;
2717 trans->param_flags =
2718 sctp_spp_sackdelay_enable(trans->param_flags);
2726 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2728 * Applications can specify protocol parameters for the default association
2729 * initialization. The option name argument to setsockopt() and getsockopt()
2732 * Setting initialization parameters is effective only on an unconnected
2733 * socket (for UDP-style sockets only future associations are effected
2734 * by the change). With TCP-style sockets, this option is inherited by
2735 * sockets derived from a listener socket.
2737 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2739 struct sctp_initmsg sinit;
2740 struct sctp_sock *sp = sctp_sk(sk);
2742 if (optlen != sizeof(struct sctp_initmsg))
2744 if (copy_from_user(&sinit, optval, optlen))
2747 if (sinit.sinit_num_ostreams)
2748 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2749 if (sinit.sinit_max_instreams)
2750 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2751 if (sinit.sinit_max_attempts)
2752 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2753 if (sinit.sinit_max_init_timeo)
2754 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2760 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2762 * Applications that wish to use the sendto() system call may wish to
2763 * specify a default set of parameters that would normally be supplied
2764 * through the inclusion of ancillary data. This socket option allows
2765 * such an application to set the default sctp_sndrcvinfo structure.
2766 * The application that wishes to use this socket option simply passes
2767 * in to this call the sctp_sndrcvinfo structure defined in Section
2768 * 5.2.2) The input parameters accepted by this call include
2769 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2770 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2771 * to this call if the caller is using the UDP model.
2773 static int sctp_setsockopt_default_send_param(struct sock *sk,
2774 char __user *optval,
2775 unsigned int optlen)
2777 struct sctp_sock *sp = sctp_sk(sk);
2778 struct sctp_association *asoc;
2779 struct sctp_sndrcvinfo info;
2781 if (optlen != sizeof(info))
2783 if (copy_from_user(&info, optval, optlen))
2785 if (info.sinfo_flags &
2786 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2787 SCTP_ABORT | SCTP_EOF))
2790 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2791 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2794 asoc->default_stream = info.sinfo_stream;
2795 asoc->default_flags = info.sinfo_flags;
2796 asoc->default_ppid = info.sinfo_ppid;
2797 asoc->default_context = info.sinfo_context;
2798 asoc->default_timetolive = info.sinfo_timetolive;
2800 sp->default_stream = info.sinfo_stream;
2801 sp->default_flags = info.sinfo_flags;
2802 sp->default_ppid = info.sinfo_ppid;
2803 sp->default_context = info.sinfo_context;
2804 sp->default_timetolive = info.sinfo_timetolive;
2810 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2811 * (SCTP_DEFAULT_SNDINFO)
2813 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
2814 char __user *optval,
2815 unsigned int optlen)
2817 struct sctp_sock *sp = sctp_sk(sk);
2818 struct sctp_association *asoc;
2819 struct sctp_sndinfo info;
2821 if (optlen != sizeof(info))
2823 if (copy_from_user(&info, optval, optlen))
2825 if (info.snd_flags &
2826 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2827 SCTP_ABORT | SCTP_EOF))
2830 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
2831 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
2834 asoc->default_stream = info.snd_sid;
2835 asoc->default_flags = info.snd_flags;
2836 asoc->default_ppid = info.snd_ppid;
2837 asoc->default_context = info.snd_context;
2839 sp->default_stream = info.snd_sid;
2840 sp->default_flags = info.snd_flags;
2841 sp->default_ppid = info.snd_ppid;
2842 sp->default_context = info.snd_context;
2848 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2850 * Requests that the local SCTP stack use the enclosed peer address as
2851 * the association primary. The enclosed address must be one of the
2852 * association peer's addresses.
2854 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2855 unsigned int optlen)
2857 struct sctp_prim prim;
2858 struct sctp_transport *trans;
2860 if (optlen != sizeof(struct sctp_prim))
2863 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2866 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2870 sctp_assoc_set_primary(trans->asoc, trans);
2876 * 7.1.5 SCTP_NODELAY
2878 * Turn on/off any Nagle-like algorithm. This means that packets are
2879 * generally sent as soon as possible and no unnecessary delays are
2880 * introduced, at the cost of more packets in the network. Expects an
2881 * integer boolean flag.
2883 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2884 unsigned int optlen)
2888 if (optlen < sizeof(int))
2890 if (get_user(val, (int __user *)optval))
2893 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2899 * 7.1.1 SCTP_RTOINFO
2901 * The protocol parameters used to initialize and bound retransmission
2902 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2903 * and modify these parameters.
2904 * All parameters are time values, in milliseconds. A value of 0, when
2905 * modifying the parameters, indicates that the current value should not
2909 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2911 struct sctp_rtoinfo rtoinfo;
2912 struct sctp_association *asoc;
2913 unsigned long rto_min, rto_max;
2914 struct sctp_sock *sp = sctp_sk(sk);
2916 if (optlen != sizeof (struct sctp_rtoinfo))
2919 if (copy_from_user(&rtoinfo, optval, optlen))
2922 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2924 /* Set the values to the specific association */
2925 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2928 rto_max = rtoinfo.srto_max;
2929 rto_min = rtoinfo.srto_min;
2932 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2934 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2937 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2939 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2941 if (rto_min > rto_max)
2945 if (rtoinfo.srto_initial != 0)
2947 msecs_to_jiffies(rtoinfo.srto_initial);
2948 asoc->rto_max = rto_max;
2949 asoc->rto_min = rto_min;
2951 /* If there is no association or the association-id = 0
2952 * set the values to the endpoint.
2954 if (rtoinfo.srto_initial != 0)
2955 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2956 sp->rtoinfo.srto_max = rto_max;
2957 sp->rtoinfo.srto_min = rto_min;
2965 * 7.1.2 SCTP_ASSOCINFO
2967 * This option is used to tune the maximum retransmission attempts
2968 * of the association.
2969 * Returns an error if the new association retransmission value is
2970 * greater than the sum of the retransmission value of the peer.
2971 * See [SCTP] for more information.
2974 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2977 struct sctp_assocparams assocparams;
2978 struct sctp_association *asoc;
2980 if (optlen != sizeof(struct sctp_assocparams))
2982 if (copy_from_user(&assocparams, optval, optlen))
2985 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2987 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2990 /* Set the values to the specific association */
2992 if (assocparams.sasoc_asocmaxrxt != 0) {
2995 struct sctp_transport *peer_addr;
2997 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2999 path_sum += peer_addr->pathmaxrxt;
3003 /* Only validate asocmaxrxt if we have more than
3004 * one path/transport. We do this because path
3005 * retransmissions are only counted when we have more
3009 assocparams.sasoc_asocmaxrxt > path_sum)
3012 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3015 if (assocparams.sasoc_cookie_life != 0)
3016 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3018 /* Set the values to the endpoint */
3019 struct sctp_sock *sp = sctp_sk(sk);
3021 if (assocparams.sasoc_asocmaxrxt != 0)
3022 sp->assocparams.sasoc_asocmaxrxt =
3023 assocparams.sasoc_asocmaxrxt;
3024 if (assocparams.sasoc_cookie_life != 0)
3025 sp->assocparams.sasoc_cookie_life =
3026 assocparams.sasoc_cookie_life;
3032 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3034 * This socket option is a boolean flag which turns on or off mapped V4
3035 * addresses. If this option is turned on and the socket is type
3036 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3037 * If this option is turned off, then no mapping will be done of V4
3038 * addresses and a user will receive both PF_INET6 and PF_INET type
3039 * addresses on the socket.
3041 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3044 struct sctp_sock *sp = sctp_sk(sk);
3046 if (optlen < sizeof(int))
3048 if (get_user(val, (int __user *)optval))
3059 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3060 * This option will get or set the maximum size to put in any outgoing
3061 * SCTP DATA chunk. If a message is larger than this size it will be
3062 * fragmented by SCTP into the specified size. Note that the underlying
3063 * SCTP implementation may fragment into smaller sized chunks when the
3064 * PMTU of the underlying association is smaller than the value set by
3065 * the user. The default value for this option is '0' which indicates
3066 * the user is NOT limiting fragmentation and only the PMTU will effect
3067 * SCTP's choice of DATA chunk size. Note also that values set larger
3068 * than the maximum size of an IP datagram will effectively let SCTP
3069 * control fragmentation (i.e. the same as setting this option to 0).
3071 * The following structure is used to access and modify this parameter:
3073 * struct sctp_assoc_value {
3074 * sctp_assoc_t assoc_id;
3075 * uint32_t assoc_value;
3078 * assoc_id: This parameter is ignored for one-to-one style sockets.
3079 * For one-to-many style sockets this parameter indicates which
3080 * association the user is performing an action upon. Note that if
3081 * this field's value is zero then the endpoints default value is
3082 * changed (effecting future associations only).
3083 * assoc_value: This parameter specifies the maximum size in bytes.
3085 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3087 struct sctp_assoc_value params;
3088 struct sctp_association *asoc;
3089 struct sctp_sock *sp = sctp_sk(sk);
3092 if (optlen == sizeof(int)) {
3093 pr_warn_ratelimited(DEPRECATED
3095 "Use of int in maxseg socket option.\n"
3096 "Use struct sctp_assoc_value instead\n",
3097 current->comm, task_pid_nr(current));
3098 if (copy_from_user(&val, optval, optlen))
3100 params.assoc_id = 0;
3101 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3102 if (copy_from_user(¶ms, optval, optlen))
3104 val = params.assoc_value;
3108 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3111 asoc = sctp_id2assoc(sk, params.assoc_id);
3112 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3117 val = asoc->pathmtu;
3118 val -= sp->pf->af->net_header_len;
3119 val -= sizeof(struct sctphdr) +
3120 sizeof(struct sctp_data_chunk);
3122 asoc->user_frag = val;
3123 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3125 sp->user_frag = val;
3133 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3135 * Requests that the peer mark the enclosed address as the association
3136 * primary. The enclosed address must be one of the association's
3137 * locally bound addresses. The following structure is used to make a
3138 * set primary request:
3140 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3141 unsigned int optlen)
3143 struct net *net = sock_net(sk);
3144 struct sctp_sock *sp;
3145 struct sctp_association *asoc = NULL;
3146 struct sctp_setpeerprim prim;
3147 struct sctp_chunk *chunk;
3153 if (!net->sctp.addip_enable)
3156 if (optlen != sizeof(struct sctp_setpeerprim))
3159 if (copy_from_user(&prim, optval, optlen))
3162 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3166 if (!asoc->peer.asconf_capable)
3169 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3172 if (!sctp_state(asoc, ESTABLISHED))
3175 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3179 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3180 return -EADDRNOTAVAIL;
3182 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3183 return -EADDRNOTAVAIL;
3185 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3186 chunk = sctp_make_asconf_set_prim(asoc,
3187 (union sctp_addr *)&prim.sspp_addr);
3191 err = sctp_send_asconf(asoc, chunk);
3193 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3198 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3199 unsigned int optlen)
3201 struct sctp_setadaptation adaptation;
3203 if (optlen != sizeof(struct sctp_setadaptation))
3205 if (copy_from_user(&adaptation, optval, optlen))
3208 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3214 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3216 * The context field in the sctp_sndrcvinfo structure is normally only
3217 * used when a failed message is retrieved holding the value that was
3218 * sent down on the actual send call. This option allows the setting of
3219 * a default context on an association basis that will be received on
3220 * reading messages from the peer. This is especially helpful in the
3221 * one-2-many model for an application to keep some reference to an
3222 * internal state machine that is processing messages on the
3223 * association. Note that the setting of this value only effects
3224 * received messages from the peer and does not effect the value that is
3225 * saved with outbound messages.
3227 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3228 unsigned int optlen)
3230 struct sctp_assoc_value params;
3231 struct sctp_sock *sp;
3232 struct sctp_association *asoc;
3234 if (optlen != sizeof(struct sctp_assoc_value))
3236 if (copy_from_user(¶ms, optval, optlen))
3241 if (params.assoc_id != 0) {
3242 asoc = sctp_id2assoc(sk, params.assoc_id);
3245 asoc->default_rcv_context = params.assoc_value;
3247 sp->default_rcv_context = params.assoc_value;
3254 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3256 * This options will at a minimum specify if the implementation is doing
3257 * fragmented interleave. Fragmented interleave, for a one to many
3258 * socket, is when subsequent calls to receive a message may return
3259 * parts of messages from different associations. Some implementations
3260 * may allow you to turn this value on or off. If so, when turned off,
3261 * no fragment interleave will occur (which will cause a head of line
3262 * blocking amongst multiple associations sharing the same one to many
3263 * socket). When this option is turned on, then each receive call may
3264 * come from a different association (thus the user must receive data
3265 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3266 * association each receive belongs to.
3268 * This option takes a boolean value. A non-zero value indicates that
3269 * fragmented interleave is on. A value of zero indicates that
3270 * fragmented interleave is off.
3272 * Note that it is important that an implementation that allows this
3273 * option to be turned on, have it off by default. Otherwise an unaware
3274 * application using the one to many model may become confused and act
3277 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3278 char __user *optval,
3279 unsigned int optlen)
3283 if (optlen != sizeof(int))
3285 if (get_user(val, (int __user *)optval))
3288 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3294 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3295 * (SCTP_PARTIAL_DELIVERY_POINT)
3297 * This option will set or get the SCTP partial delivery point. This
3298 * point is the size of a message where the partial delivery API will be
3299 * invoked to help free up rwnd space for the peer. Setting this to a
3300 * lower value will cause partial deliveries to happen more often. The
3301 * calls argument is an integer that sets or gets the partial delivery
3302 * point. Note also that the call will fail if the user attempts to set
3303 * this value larger than the socket receive buffer size.
3305 * Note that any single message having a length smaller than or equal to
3306 * the SCTP partial delivery point will be delivered in one single read
3307 * call as long as the user provided buffer is large enough to hold the
3310 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3311 char __user *optval,
3312 unsigned int optlen)
3316 if (optlen != sizeof(u32))
3318 if (get_user(val, (int __user *)optval))
3321 /* Note: We double the receive buffer from what the user sets
3322 * it to be, also initial rwnd is based on rcvbuf/2.
3324 if (val > (sk->sk_rcvbuf >> 1))
3327 sctp_sk(sk)->pd_point = val;
3329 return 0; /* is this the right error code? */
3333 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3335 * This option will allow a user to change the maximum burst of packets
3336 * that can be emitted by this association. Note that the default value
3337 * is 4, and some implementations may restrict this setting so that it
3338 * can only be lowered.
3340 * NOTE: This text doesn't seem right. Do this on a socket basis with
3341 * future associations inheriting the socket value.
3343 static int sctp_setsockopt_maxburst(struct sock *sk,
3344 char __user *optval,
3345 unsigned int optlen)
3347 struct sctp_assoc_value params;
3348 struct sctp_sock *sp;
3349 struct sctp_association *asoc;
3353 if (optlen == sizeof(int)) {
3354 pr_warn_ratelimited(DEPRECATED
3356 "Use of int in max_burst socket option deprecated.\n"
3357 "Use struct sctp_assoc_value instead\n",
3358 current->comm, task_pid_nr(current));
3359 if (copy_from_user(&val, optval, optlen))
3361 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3362 if (copy_from_user(¶ms, optval, optlen))
3364 val = params.assoc_value;
3365 assoc_id = params.assoc_id;
3371 if (assoc_id != 0) {
3372 asoc = sctp_id2assoc(sk, assoc_id);
3375 asoc->max_burst = val;
3377 sp->max_burst = val;
3383 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3385 * This set option adds a chunk type that the user is requesting to be
3386 * received only in an authenticated way. Changes to the list of chunks
3387 * will only effect future associations on the socket.
3389 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3390 char __user *optval,
3391 unsigned int optlen)
3393 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3394 struct sctp_authchunk val;
3396 if (!ep->auth_enable)
3399 if (optlen != sizeof(struct sctp_authchunk))
3401 if (copy_from_user(&val, optval, optlen))
3404 switch (val.sauth_chunk) {
3406 case SCTP_CID_INIT_ACK:
3407 case SCTP_CID_SHUTDOWN_COMPLETE:
3412 /* add this chunk id to the endpoint */
3413 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3417 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3419 * This option gets or sets the list of HMAC algorithms that the local
3420 * endpoint requires the peer to use.
3422 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3423 char __user *optval,
3424 unsigned int optlen)
3426 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3427 struct sctp_hmacalgo *hmacs;
3431 if (!ep->auth_enable)
3434 if (optlen < sizeof(struct sctp_hmacalgo))
3437 hmacs = memdup_user(optval, optlen);
3439 return PTR_ERR(hmacs);
3441 idents = hmacs->shmac_num_idents;
3442 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3443 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3448 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3455 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3457 * This option will set a shared secret key which is used to build an
3458 * association shared key.
3460 static int sctp_setsockopt_auth_key(struct sock *sk,
3461 char __user *optval,
3462 unsigned int optlen)
3464 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3465 struct sctp_authkey *authkey;
3466 struct sctp_association *asoc;
3469 if (!ep->auth_enable)
3472 if (optlen <= sizeof(struct sctp_authkey))
3475 authkey = memdup_user(optval, optlen);
3476 if (IS_ERR(authkey))
3477 return PTR_ERR(authkey);
3479 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3484 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3485 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3490 ret = sctp_auth_set_key(ep, asoc, authkey);
3497 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3499 * This option will get or set the active shared key to be used to build
3500 * the association shared key.
3502 static int sctp_setsockopt_active_key(struct sock *sk,
3503 char __user *optval,
3504 unsigned int optlen)
3506 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3507 struct sctp_authkeyid val;
3508 struct sctp_association *asoc;
3510 if (!ep->auth_enable)
3513 if (optlen != sizeof(struct sctp_authkeyid))
3515 if (copy_from_user(&val, optval, optlen))
3518 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3519 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3522 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3526 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3528 * This set option will delete a shared secret key from use.
3530 static int sctp_setsockopt_del_key(struct sock *sk,
3531 char __user *optval,
3532 unsigned int optlen)
3534 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3535 struct sctp_authkeyid val;
3536 struct sctp_association *asoc;
3538 if (!ep->auth_enable)
3541 if (optlen != sizeof(struct sctp_authkeyid))
3543 if (copy_from_user(&val, optval, optlen))
3546 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3547 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3550 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3555 * 8.1.23 SCTP_AUTO_ASCONF
3557 * This option will enable or disable the use of the automatic generation of
3558 * ASCONF chunks to add and delete addresses to an existing association. Note
3559 * that this option has two caveats namely: a) it only affects sockets that
3560 * are bound to all addresses available to the SCTP stack, and b) the system
3561 * administrator may have an overriding control that turns the ASCONF feature
3562 * off no matter what setting the socket option may have.
3563 * This option expects an integer boolean flag, where a non-zero value turns on
3564 * the option, and a zero value turns off the option.
3565 * Note. In this implementation, socket operation overrides default parameter
3566 * being set by sysctl as well as FreeBSD implementation
3568 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3569 unsigned int optlen)
3572 struct sctp_sock *sp = sctp_sk(sk);
3574 if (optlen < sizeof(int))
3576 if (get_user(val, (int __user *)optval))
3578 if (!sctp_is_ep_boundall(sk) && val)
3580 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3583 if (val == 0 && sp->do_auto_asconf) {
3584 list_del(&sp->auto_asconf_list);
3585 sp->do_auto_asconf = 0;
3586 } else if (val && !sp->do_auto_asconf) {
3587 list_add_tail(&sp->auto_asconf_list,
3588 &sock_net(sk)->sctp.auto_asconf_splist);
3589 sp->do_auto_asconf = 1;
3595 * SCTP_PEER_ADDR_THLDS
3597 * This option allows us to alter the partially failed threshold for one or all
3598 * transports in an association. See Section 6.1 of:
3599 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3601 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3602 char __user *optval,
3603 unsigned int optlen)
3605 struct sctp_paddrthlds val;
3606 struct sctp_transport *trans;
3607 struct sctp_association *asoc;
3609 if (optlen < sizeof(struct sctp_paddrthlds))
3611 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3612 sizeof(struct sctp_paddrthlds)))
3616 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3617 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3620 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3622 if (val.spt_pathmaxrxt)
3623 trans->pathmaxrxt = val.spt_pathmaxrxt;
3624 trans->pf_retrans = val.spt_pathpfthld;
3627 if (val.spt_pathmaxrxt)
3628 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3629 asoc->pf_retrans = val.spt_pathpfthld;
3631 trans = sctp_addr_id2transport(sk, &val.spt_address,
3636 if (val.spt_pathmaxrxt)
3637 trans->pathmaxrxt = val.spt_pathmaxrxt;
3638 trans->pf_retrans = val.spt_pathpfthld;
3644 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3645 char __user *optval,
3646 unsigned int optlen)
3650 if (optlen < sizeof(int))
3652 if (get_user(val, (int __user *) optval))
3655 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3660 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3661 char __user *optval,
3662 unsigned int optlen)
3666 if (optlen < sizeof(int))
3668 if (get_user(val, (int __user *) optval))
3671 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3676 /* API 6.2 setsockopt(), getsockopt()
3678 * Applications use setsockopt() and getsockopt() to set or retrieve
3679 * socket options. Socket options are used to change the default
3680 * behavior of sockets calls. They are described in Section 7.
3684 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3685 * int __user *optlen);
3686 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3689 * sd - the socket descript.
3690 * level - set to IPPROTO_SCTP for all SCTP options.
3691 * optname - the option name.
3692 * optval - the buffer to store the value of the option.
3693 * optlen - the size of the buffer.
3695 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3696 char __user *optval, unsigned int optlen)
3700 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3702 /* I can hardly begin to describe how wrong this is. This is
3703 * so broken as to be worse than useless. The API draft
3704 * REALLY is NOT helpful here... I am not convinced that the
3705 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3706 * are at all well-founded.
3708 if (level != SOL_SCTP) {
3709 struct sctp_af *af = sctp_sk(sk)->pf->af;
3710 retval = af->setsockopt(sk, level, optname, optval, optlen);
3717 case SCTP_SOCKOPT_BINDX_ADD:
3718 /* 'optlen' is the size of the addresses buffer. */
3719 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3720 optlen, SCTP_BINDX_ADD_ADDR);
3723 case SCTP_SOCKOPT_BINDX_REM:
3724 /* 'optlen' is the size of the addresses buffer. */
3725 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3726 optlen, SCTP_BINDX_REM_ADDR);
3729 case SCTP_SOCKOPT_CONNECTX_OLD:
3730 /* 'optlen' is the size of the addresses buffer. */
3731 retval = sctp_setsockopt_connectx_old(sk,
3732 (struct sockaddr __user *)optval,
3736 case SCTP_SOCKOPT_CONNECTX:
3737 /* 'optlen' is the size of the addresses buffer. */
3738 retval = sctp_setsockopt_connectx(sk,
3739 (struct sockaddr __user *)optval,
3743 case SCTP_DISABLE_FRAGMENTS:
3744 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3748 retval = sctp_setsockopt_events(sk, optval, optlen);
3751 case SCTP_AUTOCLOSE:
3752 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3755 case SCTP_PEER_ADDR_PARAMS:
3756 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3759 case SCTP_DELAYED_SACK:
3760 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3762 case SCTP_PARTIAL_DELIVERY_POINT:
3763 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3767 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3769 case SCTP_DEFAULT_SEND_PARAM:
3770 retval = sctp_setsockopt_default_send_param(sk, optval,
3773 case SCTP_DEFAULT_SNDINFO:
3774 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
3776 case SCTP_PRIMARY_ADDR:
3777 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3779 case SCTP_SET_PEER_PRIMARY_ADDR:
3780 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3783 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3786 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3788 case SCTP_ASSOCINFO:
3789 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3791 case SCTP_I_WANT_MAPPED_V4_ADDR:
3792 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3795 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3797 case SCTP_ADAPTATION_LAYER:
3798 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3801 retval = sctp_setsockopt_context(sk, optval, optlen);
3803 case SCTP_FRAGMENT_INTERLEAVE:
3804 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3806 case SCTP_MAX_BURST:
3807 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3809 case SCTP_AUTH_CHUNK:
3810 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3812 case SCTP_HMAC_IDENT:
3813 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3816 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3818 case SCTP_AUTH_ACTIVE_KEY:
3819 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3821 case SCTP_AUTH_DELETE_KEY:
3822 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3824 case SCTP_AUTO_ASCONF:
3825 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3827 case SCTP_PEER_ADDR_THLDS:
3828 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3830 case SCTP_RECVRCVINFO:
3831 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
3833 case SCTP_RECVNXTINFO:
3834 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
3837 retval = -ENOPROTOOPT;
3847 /* API 3.1.6 connect() - UDP Style Syntax
3849 * An application may use the connect() call in the UDP model to initiate an
3850 * association without sending data.
3854 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3856 * sd: the socket descriptor to have a new association added to.
3858 * nam: the address structure (either struct sockaddr_in or struct
3859 * sockaddr_in6 defined in RFC2553 [7]).
3861 * len: the size of the address.
3863 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3871 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
3874 /* Validate addr_len before calling common connect/connectx routine. */
3875 af = sctp_get_af_specific(addr->sa_family);
3876 if (!af || addr_len < af->sockaddr_len) {
3879 /* Pass correct addr len to common routine (so it knows there
3880 * is only one address being passed.
3882 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3889 /* FIXME: Write comments. */
3890 static int sctp_disconnect(struct sock *sk, int flags)
3892 return -EOPNOTSUPP; /* STUB */
3895 /* 4.1.4 accept() - TCP Style Syntax
3897 * Applications use accept() call to remove an established SCTP
3898 * association from the accept queue of the endpoint. A new socket
3899 * descriptor will be returned from accept() to represent the newly
3900 * formed association.
3902 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3904 struct sctp_sock *sp;
3905 struct sctp_endpoint *ep;
3906 struct sock *newsk = NULL;
3907 struct sctp_association *asoc;
3916 if (!sctp_style(sk, TCP)) {
3917 error = -EOPNOTSUPP;
3921 if (!sctp_sstate(sk, LISTENING)) {
3926 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3928 error = sctp_wait_for_accept(sk, timeo);
3932 /* We treat the list of associations on the endpoint as the accept
3933 * queue and pick the first association on the list.
3935 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3937 newsk = sp->pf->create_accept_sk(sk, asoc);
3943 /* Populate the fields of the newsk from the oldsk and migrate the
3944 * asoc to the newsk.
3946 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3954 /* The SCTP ioctl handler. */
3955 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3962 * SEQPACKET-style sockets in LISTENING state are valid, for
3963 * SCTP, so only discard TCP-style sockets in LISTENING state.
3965 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3970 struct sk_buff *skb;
3971 unsigned int amount = 0;
3973 skb = skb_peek(&sk->sk_receive_queue);
3976 * We will only return the amount of this packet since
3977 * that is all that will be read.
3981 rc = put_user(amount, (int __user *)arg);
3993 /* This is the function which gets called during socket creation to
3994 * initialized the SCTP-specific portion of the sock.
3995 * The sock structure should already be zero-filled memory.
3997 static int sctp_init_sock(struct sock *sk)
3999 struct net *net = sock_net(sk);
4000 struct sctp_sock *sp;
4002 pr_debug("%s: sk:%p\n", __func__, sk);
4006 /* Initialize the SCTP per socket area. */
4007 switch (sk->sk_type) {
4008 case SOCK_SEQPACKET:
4009 sp->type = SCTP_SOCKET_UDP;
4012 sp->type = SCTP_SOCKET_TCP;
4015 return -ESOCKTNOSUPPORT;
4018 /* Initialize default send parameters. These parameters can be
4019 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4021 sp->default_stream = 0;
4022 sp->default_ppid = 0;
4023 sp->default_flags = 0;
4024 sp->default_context = 0;
4025 sp->default_timetolive = 0;
4027 sp->default_rcv_context = 0;
4028 sp->max_burst = net->sctp.max_burst;
4030 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4032 /* Initialize default setup parameters. These parameters
4033 * can be modified with the SCTP_INITMSG socket option or
4034 * overridden by the SCTP_INIT CMSG.
4036 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4037 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4038 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4039 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4041 /* Initialize default RTO related parameters. These parameters can
4042 * be modified for with the SCTP_RTOINFO socket option.
4044 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4045 sp->rtoinfo.srto_max = net->sctp.rto_max;
4046 sp->rtoinfo.srto_min = net->sctp.rto_min;
4048 /* Initialize default association related parameters. These parameters
4049 * can be modified with the SCTP_ASSOCINFO socket option.
4051 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4052 sp->assocparams.sasoc_number_peer_destinations = 0;
4053 sp->assocparams.sasoc_peer_rwnd = 0;
4054 sp->assocparams.sasoc_local_rwnd = 0;
4055 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4057 /* Initialize default event subscriptions. By default, all the
4060 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4062 /* Default Peer Address Parameters. These defaults can
4063 * be modified via SCTP_PEER_ADDR_PARAMS
4065 sp->hbinterval = net->sctp.hb_interval;
4066 sp->pathmaxrxt = net->sctp.max_retrans_path;
4067 sp->pathmtu = 0; /* allow default discovery */
4068 sp->sackdelay = net->sctp.sack_timeout;
4070 sp->param_flags = SPP_HB_ENABLE |
4072 SPP_SACKDELAY_ENABLE;
4074 /* If enabled no SCTP message fragmentation will be performed.
4075 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4077 sp->disable_fragments = 0;
4079 /* Enable Nagle algorithm by default. */
4082 sp->recvrcvinfo = 0;
4083 sp->recvnxtinfo = 0;
4085 /* Enable by default. */
4088 /* Auto-close idle associations after the configured
4089 * number of seconds. A value of 0 disables this
4090 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4091 * for UDP-style sockets only.
4095 /* User specified fragmentation limit. */
4098 sp->adaptation_ind = 0;
4100 sp->pf = sctp_get_pf_specific(sk->sk_family);
4102 /* Control variables for partial data delivery. */
4103 atomic_set(&sp->pd_mode, 0);
4104 skb_queue_head_init(&sp->pd_lobby);
4105 sp->frag_interleave = 0;
4107 /* Create a per socket endpoint structure. Even if we
4108 * change the data structure relationships, this may still
4109 * be useful for storing pre-connect address information.
4111 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4117 sk->sk_destruct = sctp_destruct_sock;
4119 SCTP_DBG_OBJCNT_INC(sock);
4122 percpu_counter_inc(&sctp_sockets_allocated);
4123 sock_prot_inuse_add(net, sk->sk_prot, 1);
4124 if (net->sctp.default_auto_asconf) {
4125 list_add_tail(&sp->auto_asconf_list,
4126 &net->sctp.auto_asconf_splist);
4127 sp->do_auto_asconf = 1;
4129 sp->do_auto_asconf = 0;
4135 /* Cleanup any SCTP per socket resources. */
4136 static void sctp_destroy_sock(struct sock *sk)
4138 struct sctp_sock *sp;
4140 pr_debug("%s: sk:%p\n", __func__, sk);
4142 /* Release our hold on the endpoint. */
4144 /* This could happen during socket init, thus we bail out
4145 * early, since the rest of the below is not setup either.
4150 if (sp->do_auto_asconf) {
4151 sp->do_auto_asconf = 0;
4152 list_del(&sp->auto_asconf_list);
4154 sctp_endpoint_free(sp->ep);
4156 percpu_counter_dec(&sctp_sockets_allocated);
4157 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4161 /* Triggered when there are no references on the socket anymore */
4162 static void sctp_destruct_sock(struct sock *sk)
4164 struct sctp_sock *sp = sctp_sk(sk);
4166 /* Free up the HMAC transform. */
4167 crypto_free_hash(sp->hmac);
4169 inet_sock_destruct(sk);
4172 /* API 4.1.7 shutdown() - TCP Style Syntax
4173 * int shutdown(int socket, int how);
4175 * sd - the socket descriptor of the association to be closed.
4176 * how - Specifies the type of shutdown. The values are
4179 * Disables further receive operations. No SCTP
4180 * protocol action is taken.
4182 * Disables further send operations, and initiates
4183 * the SCTP shutdown sequence.
4185 * Disables further send and receive operations
4186 * and initiates the SCTP shutdown sequence.
4188 static void sctp_shutdown(struct sock *sk, int how)
4190 struct net *net = sock_net(sk);
4191 struct sctp_endpoint *ep;
4192 struct sctp_association *asoc;
4194 if (!sctp_style(sk, TCP))
4197 if (how & SEND_SHUTDOWN) {
4198 ep = sctp_sk(sk)->ep;
4199 if (!list_empty(&ep->asocs)) {
4200 asoc = list_entry(ep->asocs.next,
4201 struct sctp_association, asocs);
4202 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4207 /* 7.2.1 Association Status (SCTP_STATUS)
4209 * Applications can retrieve current status information about an
4210 * association, including association state, peer receiver window size,
4211 * number of unacked data chunks, and number of data chunks pending
4212 * receipt. This information is read-only.
4214 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4215 char __user *optval,
4218 struct sctp_status status;
4219 struct sctp_association *asoc = NULL;
4220 struct sctp_transport *transport;
4221 sctp_assoc_t associd;
4224 if (len < sizeof(status)) {
4229 len = sizeof(status);
4230 if (copy_from_user(&status, optval, len)) {
4235 associd = status.sstat_assoc_id;
4236 asoc = sctp_id2assoc(sk, associd);
4242 transport = asoc->peer.primary_path;
4244 status.sstat_assoc_id = sctp_assoc2id(asoc);
4245 status.sstat_state = sctp_assoc_to_state(asoc);
4246 status.sstat_rwnd = asoc->peer.rwnd;
4247 status.sstat_unackdata = asoc->unack_data;
4249 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4250 status.sstat_instrms = asoc->c.sinit_max_instreams;
4251 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4252 status.sstat_fragmentation_point = asoc->frag_point;
4253 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4254 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4255 transport->af_specific->sockaddr_len);
4256 /* Map ipv4 address into v4-mapped-on-v6 address. */
4257 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
4258 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4259 status.sstat_primary.spinfo_state = transport->state;
4260 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4261 status.sstat_primary.spinfo_srtt = transport->srtt;
4262 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4263 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4265 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4266 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4268 if (put_user(len, optlen)) {
4273 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4274 __func__, len, status.sstat_state, status.sstat_rwnd,
4275 status.sstat_assoc_id);
4277 if (copy_to_user(optval, &status, len)) {
4287 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4289 * Applications can retrieve information about a specific peer address
4290 * of an association, including its reachability state, congestion
4291 * window, and retransmission timer values. This information is
4294 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4295 char __user *optval,
4298 struct sctp_paddrinfo pinfo;
4299 struct sctp_transport *transport;
4302 if (len < sizeof(pinfo)) {
4307 len = sizeof(pinfo);
4308 if (copy_from_user(&pinfo, optval, len)) {
4313 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4314 pinfo.spinfo_assoc_id);
4318 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4319 pinfo.spinfo_state = transport->state;
4320 pinfo.spinfo_cwnd = transport->cwnd;
4321 pinfo.spinfo_srtt = transport->srtt;
4322 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4323 pinfo.spinfo_mtu = transport->pathmtu;
4325 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4326 pinfo.spinfo_state = SCTP_ACTIVE;
4328 if (put_user(len, optlen)) {
4333 if (copy_to_user(optval, &pinfo, len)) {
4342 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4344 * This option is a on/off flag. If enabled no SCTP message
4345 * fragmentation will be performed. Instead if a message being sent
4346 * exceeds the current PMTU size, the message will NOT be sent and
4347 * instead a error will be indicated to the user.
4349 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4350 char __user *optval, int __user *optlen)
4354 if (len < sizeof(int))
4358 val = (sctp_sk(sk)->disable_fragments == 1);
4359 if (put_user(len, optlen))
4361 if (copy_to_user(optval, &val, len))
4366 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4368 * This socket option is used to specify various notifications and
4369 * ancillary data the user wishes to receive.
4371 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4376 if (len > sizeof(struct sctp_event_subscribe))
4377 len = sizeof(struct sctp_event_subscribe);
4378 if (put_user(len, optlen))
4380 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4385 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4387 * This socket option is applicable to the UDP-style socket only. When
4388 * set it will cause associations that are idle for more than the
4389 * specified number of seconds to automatically close. An association
4390 * being idle is defined an association that has NOT sent or received
4391 * user data. The special value of '0' indicates that no automatic
4392 * close of any associations should be performed. The option expects an
4393 * integer defining the number of seconds of idle time before an
4394 * association is closed.
4396 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4398 /* Applicable to UDP-style socket only */
4399 if (sctp_style(sk, TCP))
4401 if (len < sizeof(int))
4404 if (put_user(len, optlen))
4406 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4411 /* Helper routine to branch off an association to a new socket. */
4412 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4414 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4415 struct sctp_sock *sp = sctp_sk(sk);
4416 struct socket *sock;
4422 /* An association cannot be branched off from an already peeled-off
4423 * socket, nor is this supported for tcp style sockets.
4425 if (!sctp_style(sk, UDP))
4428 /* Create a new socket. */
4429 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4433 sctp_copy_sock(sock->sk, sk, asoc);
4435 /* Make peeled-off sockets more like 1-1 accepted sockets.
4436 * Set the daddr and initialize id to something more random
4438 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
4440 /* Populate the fields of the newsk from the oldsk and migrate the
4441 * asoc to the newsk.
4443 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4449 EXPORT_SYMBOL(sctp_do_peeloff);
4451 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4453 sctp_peeloff_arg_t peeloff;
4454 struct socket *newsock;
4455 struct file *newfile;
4458 if (len < sizeof(sctp_peeloff_arg_t))
4460 len = sizeof(sctp_peeloff_arg_t);
4461 if (copy_from_user(&peeloff, optval, len))
4464 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4468 /* Map the socket to an unused fd that can be returned to the user. */
4469 retval = get_unused_fd_flags(0);
4471 sock_release(newsock);
4475 newfile = sock_alloc_file(newsock, 0, NULL);
4476 if (unlikely(IS_ERR(newfile))) {
4477 put_unused_fd(retval);
4478 sock_release(newsock);
4479 return PTR_ERR(newfile);
4482 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4485 /* Return the fd mapped to the new socket. */
4486 if (put_user(len, optlen)) {
4488 put_unused_fd(retval);
4491 peeloff.sd = retval;
4492 if (copy_to_user(optval, &peeloff, len)) {
4494 put_unused_fd(retval);
4497 fd_install(retval, newfile);
4502 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4504 * Applications can enable or disable heartbeats for any peer address of
4505 * an association, modify an address's heartbeat interval, force a
4506 * heartbeat to be sent immediately, and adjust the address's maximum
4507 * number of retransmissions sent before an address is considered
4508 * unreachable. The following structure is used to access and modify an
4509 * address's parameters:
4511 * struct sctp_paddrparams {
4512 * sctp_assoc_t spp_assoc_id;
4513 * struct sockaddr_storage spp_address;
4514 * uint32_t spp_hbinterval;
4515 * uint16_t spp_pathmaxrxt;
4516 * uint32_t spp_pathmtu;
4517 * uint32_t spp_sackdelay;
4518 * uint32_t spp_flags;
4521 * spp_assoc_id - (one-to-many style socket) This is filled in the
4522 * application, and identifies the association for
4524 * spp_address - This specifies which address is of interest.
4525 * spp_hbinterval - This contains the value of the heartbeat interval,
4526 * in milliseconds. If a value of zero
4527 * is present in this field then no changes are to
4528 * be made to this parameter.
4529 * spp_pathmaxrxt - This contains the maximum number of
4530 * retransmissions before this address shall be
4531 * considered unreachable. If a value of zero
4532 * is present in this field then no changes are to
4533 * be made to this parameter.
4534 * spp_pathmtu - When Path MTU discovery is disabled the value
4535 * specified here will be the "fixed" path mtu.
4536 * Note that if the spp_address field is empty
4537 * then all associations on this address will
4538 * have this fixed path mtu set upon them.
4540 * spp_sackdelay - When delayed sack is enabled, this value specifies
4541 * the number of milliseconds that sacks will be delayed
4542 * for. This value will apply to all addresses of an
4543 * association if the spp_address field is empty. Note
4544 * also, that if delayed sack is enabled and this
4545 * value is set to 0, no change is made to the last
4546 * recorded delayed sack timer value.
4548 * spp_flags - These flags are used to control various features
4549 * on an association. The flag field may contain
4550 * zero or more of the following options.
4552 * SPP_HB_ENABLE - Enable heartbeats on the
4553 * specified address. Note that if the address
4554 * field is empty all addresses for the association
4555 * have heartbeats enabled upon them.
4557 * SPP_HB_DISABLE - Disable heartbeats on the
4558 * speicifed address. Note that if the address
4559 * field is empty all addresses for the association
4560 * will have their heartbeats disabled. Note also
4561 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4562 * mutually exclusive, only one of these two should
4563 * be specified. Enabling both fields will have
4564 * undetermined results.
4566 * SPP_HB_DEMAND - Request a user initiated heartbeat
4567 * to be made immediately.
4569 * SPP_PMTUD_ENABLE - This field will enable PMTU
4570 * discovery upon the specified address. Note that
4571 * if the address feild is empty then all addresses
4572 * on the association are effected.
4574 * SPP_PMTUD_DISABLE - This field will disable PMTU
4575 * discovery upon the specified address. Note that
4576 * if the address feild is empty then all addresses
4577 * on the association are effected. Not also that
4578 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4579 * exclusive. Enabling both will have undetermined
4582 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4583 * on delayed sack. The time specified in spp_sackdelay
4584 * is used to specify the sack delay for this address. Note
4585 * that if spp_address is empty then all addresses will
4586 * enable delayed sack and take on the sack delay
4587 * value specified in spp_sackdelay.
4588 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4589 * off delayed sack. If the spp_address field is blank then
4590 * delayed sack is disabled for the entire association. Note
4591 * also that this field is mutually exclusive to
4592 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4595 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4596 char __user *optval, int __user *optlen)
4598 struct sctp_paddrparams params;
4599 struct sctp_transport *trans = NULL;
4600 struct sctp_association *asoc = NULL;
4601 struct sctp_sock *sp = sctp_sk(sk);
4603 if (len < sizeof(struct sctp_paddrparams))
4605 len = sizeof(struct sctp_paddrparams);
4606 if (copy_from_user(¶ms, optval, len))
4609 /* If an address other than INADDR_ANY is specified, and
4610 * no transport is found, then the request is invalid.
4612 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
4613 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4614 params.spp_assoc_id);
4616 pr_debug("%s: failed no transport\n", __func__);
4621 /* Get association, if assoc_id != 0 and the socket is a one
4622 * to many style socket, and an association was not found, then
4623 * the id was invalid.
4625 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4626 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4627 pr_debug("%s: failed no association\n", __func__);
4632 /* Fetch transport values. */
4633 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4634 params.spp_pathmtu = trans->pathmtu;
4635 params.spp_pathmaxrxt = trans->pathmaxrxt;
4636 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4638 /*draft-11 doesn't say what to return in spp_flags*/
4639 params.spp_flags = trans->param_flags;
4641 /* Fetch association values. */
4642 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4643 params.spp_pathmtu = asoc->pathmtu;
4644 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4645 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4647 /*draft-11 doesn't say what to return in spp_flags*/
4648 params.spp_flags = asoc->param_flags;
4650 /* Fetch socket values. */
4651 params.spp_hbinterval = sp->hbinterval;
4652 params.spp_pathmtu = sp->pathmtu;
4653 params.spp_sackdelay = sp->sackdelay;
4654 params.spp_pathmaxrxt = sp->pathmaxrxt;
4656 /*draft-11 doesn't say what to return in spp_flags*/
4657 params.spp_flags = sp->param_flags;
4660 if (copy_to_user(optval, ¶ms, len))
4663 if (put_user(len, optlen))
4670 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4672 * This option will effect the way delayed acks are performed. This
4673 * option allows you to get or set the delayed ack time, in
4674 * milliseconds. It also allows changing the delayed ack frequency.
4675 * Changing the frequency to 1 disables the delayed sack algorithm. If
4676 * the assoc_id is 0, then this sets or gets the endpoints default
4677 * values. If the assoc_id field is non-zero, then the set or get
4678 * effects the specified association for the one to many model (the
4679 * assoc_id field is ignored by the one to one model). Note that if
4680 * sack_delay or sack_freq are 0 when setting this option, then the
4681 * current values will remain unchanged.
4683 * struct sctp_sack_info {
4684 * sctp_assoc_t sack_assoc_id;
4685 * uint32_t sack_delay;
4686 * uint32_t sack_freq;
4689 * sack_assoc_id - This parameter, indicates which association the user
4690 * is performing an action upon. Note that if this field's value is
4691 * zero then the endpoints default value is changed (effecting future
4692 * associations only).
4694 * sack_delay - This parameter contains the number of milliseconds that
4695 * the user is requesting the delayed ACK timer be set to. Note that
4696 * this value is defined in the standard to be between 200 and 500
4699 * sack_freq - This parameter contains the number of packets that must
4700 * be received before a sack is sent without waiting for the delay
4701 * timer to expire. The default value for this is 2, setting this
4702 * value to 1 will disable the delayed sack algorithm.
4704 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4705 char __user *optval,
4708 struct sctp_sack_info params;
4709 struct sctp_association *asoc = NULL;
4710 struct sctp_sock *sp = sctp_sk(sk);
4712 if (len >= sizeof(struct sctp_sack_info)) {
4713 len = sizeof(struct sctp_sack_info);
4715 if (copy_from_user(¶ms, optval, len))
4717 } else if (len == sizeof(struct sctp_assoc_value)) {
4718 pr_warn_ratelimited(DEPRECATED
4720 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
4721 "Use struct sctp_sack_info instead\n",
4722 current->comm, task_pid_nr(current));
4723 if (copy_from_user(¶ms, optval, len))
4728 /* Get association, if sack_assoc_id != 0 and the socket is a one
4729 * to many style socket, and an association was not found, then
4730 * the id was invalid.
4732 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4733 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4737 /* Fetch association values. */
4738 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4739 params.sack_delay = jiffies_to_msecs(
4741 params.sack_freq = asoc->sackfreq;
4744 params.sack_delay = 0;
4745 params.sack_freq = 1;
4748 /* Fetch socket values. */
4749 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4750 params.sack_delay = sp->sackdelay;
4751 params.sack_freq = sp->sackfreq;
4753 params.sack_delay = 0;
4754 params.sack_freq = 1;
4758 if (copy_to_user(optval, ¶ms, len))
4761 if (put_user(len, optlen))
4767 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4769 * Applications can specify protocol parameters for the default association
4770 * initialization. The option name argument to setsockopt() and getsockopt()
4773 * Setting initialization parameters is effective only on an unconnected
4774 * socket (for UDP-style sockets only future associations are effected
4775 * by the change). With TCP-style sockets, this option is inherited by
4776 * sockets derived from a listener socket.
4778 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4780 if (len < sizeof(struct sctp_initmsg))
4782 len = sizeof(struct sctp_initmsg);
4783 if (put_user(len, optlen))
4785 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4791 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4792 char __user *optval, int __user *optlen)
4794 struct sctp_association *asoc;
4796 struct sctp_getaddrs getaddrs;
4797 struct sctp_transport *from;
4799 union sctp_addr temp;
4800 struct sctp_sock *sp = sctp_sk(sk);
4805 if (len < sizeof(struct sctp_getaddrs))
4808 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4811 /* For UDP-style sockets, id specifies the association to query. */
4812 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4816 to = optval + offsetof(struct sctp_getaddrs, addrs);
4817 space_left = len - offsetof(struct sctp_getaddrs, addrs);
4819 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4821 memcpy(&temp, &from->ipaddr, sizeof(temp));
4822 addrlen = sctp_get_pf_specific(sk->sk_family)
4823 ->addr_to_user(sp, &temp);
4824 if (space_left < addrlen)
4826 if (copy_to_user(to, &temp, addrlen))
4830 space_left -= addrlen;
4833 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4835 bytes_copied = ((char __user *)to) - optval;
4836 if (put_user(bytes_copied, optlen))
4842 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4843 size_t space_left, int *bytes_copied)
4845 struct sctp_sockaddr_entry *addr;
4846 union sctp_addr temp;
4849 struct net *net = sock_net(sk);
4852 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4856 if ((PF_INET == sk->sk_family) &&
4857 (AF_INET6 == addr->a.sa.sa_family))
4859 if ((PF_INET6 == sk->sk_family) &&
4860 inet_v6_ipv6only(sk) &&
4861 (AF_INET == addr->a.sa.sa_family))
4863 memcpy(&temp, &addr->a, sizeof(temp));
4864 if (!temp.v4.sin_port)
4865 temp.v4.sin_port = htons(port);
4867 addrlen = sctp_get_pf_specific(sk->sk_family)
4868 ->addr_to_user(sctp_sk(sk), &temp);
4870 if (space_left < addrlen) {
4874 memcpy(to, &temp, addrlen);
4878 space_left -= addrlen;
4879 *bytes_copied += addrlen;
4887 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4888 char __user *optval, int __user *optlen)
4890 struct sctp_bind_addr *bp;
4891 struct sctp_association *asoc;
4893 struct sctp_getaddrs getaddrs;
4894 struct sctp_sockaddr_entry *addr;
4896 union sctp_addr temp;
4897 struct sctp_sock *sp = sctp_sk(sk);
4901 int bytes_copied = 0;
4905 if (len < sizeof(struct sctp_getaddrs))
4908 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4912 * For UDP-style sockets, id specifies the association to query.
4913 * If the id field is set to the value '0' then the locally bound
4914 * addresses are returned without regard to any particular
4917 if (0 == getaddrs.assoc_id) {
4918 bp = &sctp_sk(sk)->ep->base.bind_addr;
4920 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4923 bp = &asoc->base.bind_addr;
4926 to = optval + offsetof(struct sctp_getaddrs, addrs);
4927 space_left = len - offsetof(struct sctp_getaddrs, addrs);
4929 addrs = kmalloc(space_left, GFP_KERNEL);
4933 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4934 * addresses from the global local address list.
4936 if (sctp_list_single_entry(&bp->address_list)) {
4937 addr = list_entry(bp->address_list.next,
4938 struct sctp_sockaddr_entry, list);
4939 if (sctp_is_any(sk, &addr->a)) {
4940 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4941 space_left, &bytes_copied);
4951 /* Protection on the bound address list is not needed since
4952 * in the socket option context we hold a socket lock and
4953 * thus the bound address list can't change.
4955 list_for_each_entry(addr, &bp->address_list, list) {
4956 memcpy(&temp, &addr->a, sizeof(temp));
4957 addrlen = sctp_get_pf_specific(sk->sk_family)
4958 ->addr_to_user(sp, &temp);
4959 if (space_left < addrlen) {
4960 err = -ENOMEM; /*fixme: right error?*/
4963 memcpy(buf, &temp, addrlen);
4965 bytes_copied += addrlen;
4967 space_left -= addrlen;
4971 if (copy_to_user(to, addrs, bytes_copied)) {
4975 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4979 if (put_user(bytes_copied, optlen))
4986 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4988 * Requests that the local SCTP stack use the enclosed peer address as
4989 * the association primary. The enclosed address must be one of the
4990 * association peer's addresses.
4992 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4993 char __user *optval, int __user *optlen)
4995 struct sctp_prim prim;
4996 struct sctp_association *asoc;
4997 struct sctp_sock *sp = sctp_sk(sk);
4999 if (len < sizeof(struct sctp_prim))
5002 len = sizeof(struct sctp_prim);
5004 if (copy_from_user(&prim, optval, len))
5007 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
5011 if (!asoc->peer.primary_path)
5014 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
5015 asoc->peer.primary_path->af_specific->sockaddr_len);
5017 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
5018 (union sctp_addr *)&prim.ssp_addr);
5020 if (put_user(len, optlen))
5022 if (copy_to_user(optval, &prim, len))
5029 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5031 * Requests that the local endpoint set the specified Adaptation Layer
5032 * Indication parameter for all future INIT and INIT-ACK exchanges.
5034 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
5035 char __user *optval, int __user *optlen)
5037 struct sctp_setadaptation adaptation;
5039 if (len < sizeof(struct sctp_setadaptation))
5042 len = sizeof(struct sctp_setadaptation);
5044 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
5046 if (put_user(len, optlen))
5048 if (copy_to_user(optval, &adaptation, len))
5056 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5058 * Applications that wish to use the sendto() system call may wish to
5059 * specify a default set of parameters that would normally be supplied
5060 * through the inclusion of ancillary data. This socket option allows
5061 * such an application to set the default sctp_sndrcvinfo structure.
5064 * The application that wishes to use this socket option simply passes
5065 * in to this call the sctp_sndrcvinfo structure defined in Section
5066 * 5.2.2) The input parameters accepted by this call include
5067 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5068 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5069 * to this call if the caller is using the UDP model.
5071 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5073 static int sctp_getsockopt_default_send_param(struct sock *sk,
5074 int len, char __user *optval,
5077 struct sctp_sock *sp = sctp_sk(sk);
5078 struct sctp_association *asoc;
5079 struct sctp_sndrcvinfo info;
5081 if (len < sizeof(info))
5086 if (copy_from_user(&info, optval, len))
5089 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
5090 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
5093 info.sinfo_stream = asoc->default_stream;
5094 info.sinfo_flags = asoc->default_flags;
5095 info.sinfo_ppid = asoc->default_ppid;
5096 info.sinfo_context = asoc->default_context;
5097 info.sinfo_timetolive = asoc->default_timetolive;
5099 info.sinfo_stream = sp->default_stream;
5100 info.sinfo_flags = sp->default_flags;
5101 info.sinfo_ppid = sp->default_ppid;
5102 info.sinfo_context = sp->default_context;
5103 info.sinfo_timetolive = sp->default_timetolive;
5106 if (put_user(len, optlen))
5108 if (copy_to_user(optval, &info, len))
5114 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5115 * (SCTP_DEFAULT_SNDINFO)
5117 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
5118 char __user *optval,
5121 struct sctp_sock *sp = sctp_sk(sk);
5122 struct sctp_association *asoc;
5123 struct sctp_sndinfo info;
5125 if (len < sizeof(info))
5130 if (copy_from_user(&info, optval, len))
5133 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
5134 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
5137 info.snd_sid = asoc->default_stream;
5138 info.snd_flags = asoc->default_flags;
5139 info.snd_ppid = asoc->default_ppid;
5140 info.snd_context = asoc->default_context;
5142 info.snd_sid = sp->default_stream;
5143 info.snd_flags = sp->default_flags;
5144 info.snd_ppid = sp->default_ppid;
5145 info.snd_context = sp->default_context;
5148 if (put_user(len, optlen))
5150 if (copy_to_user(optval, &info, len))
5158 * 7.1.5 SCTP_NODELAY
5160 * Turn on/off any Nagle-like algorithm. This means that packets are
5161 * generally sent as soon as possible and no unnecessary delays are
5162 * introduced, at the cost of more packets in the network. Expects an
5163 * integer boolean flag.
5166 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5167 char __user *optval, int __user *optlen)
5171 if (len < sizeof(int))
5175 val = (sctp_sk(sk)->nodelay == 1);
5176 if (put_user(len, optlen))
5178 if (copy_to_user(optval, &val, len))
5185 * 7.1.1 SCTP_RTOINFO
5187 * The protocol parameters used to initialize and bound retransmission
5188 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5189 * and modify these parameters.
5190 * All parameters are time values, in milliseconds. A value of 0, when
5191 * modifying the parameters, indicates that the current value should not
5195 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5196 char __user *optval,
5197 int __user *optlen) {
5198 struct sctp_rtoinfo rtoinfo;
5199 struct sctp_association *asoc;
5201 if (len < sizeof (struct sctp_rtoinfo))
5204 len = sizeof(struct sctp_rtoinfo);
5206 if (copy_from_user(&rtoinfo, optval, len))
5209 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5211 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5214 /* Values corresponding to the specific association. */
5216 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5217 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5218 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5220 /* Values corresponding to the endpoint. */
5221 struct sctp_sock *sp = sctp_sk(sk);
5223 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5224 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5225 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5228 if (put_user(len, optlen))
5231 if (copy_to_user(optval, &rtoinfo, len))
5239 * 7.1.2 SCTP_ASSOCINFO
5241 * This option is used to tune the maximum retransmission attempts
5242 * of the association.
5243 * Returns an error if the new association retransmission value is
5244 * greater than the sum of the retransmission value of the peer.
5245 * See [SCTP] for more information.
5248 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5249 char __user *optval,
5253 struct sctp_assocparams assocparams;
5254 struct sctp_association *asoc;
5255 struct list_head *pos;
5258 if (len < sizeof (struct sctp_assocparams))
5261 len = sizeof(struct sctp_assocparams);
5263 if (copy_from_user(&assocparams, optval, len))
5266 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5268 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5271 /* Values correspoinding to the specific association */
5273 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5274 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5275 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5276 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5278 list_for_each(pos, &asoc->peer.transport_addr_list) {
5282 assocparams.sasoc_number_peer_destinations = cnt;
5284 /* Values corresponding to the endpoint */
5285 struct sctp_sock *sp = sctp_sk(sk);
5287 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5288 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5289 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5290 assocparams.sasoc_cookie_life =
5291 sp->assocparams.sasoc_cookie_life;
5292 assocparams.sasoc_number_peer_destinations =
5294 sasoc_number_peer_destinations;
5297 if (put_user(len, optlen))
5300 if (copy_to_user(optval, &assocparams, len))
5307 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5309 * This socket option is a boolean flag which turns on or off mapped V4
5310 * addresses. If this option is turned on and the socket is type
5311 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5312 * If this option is turned off, then no mapping will be done of V4
5313 * addresses and a user will receive both PF_INET6 and PF_INET type
5314 * addresses on the socket.
5316 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5317 char __user *optval, int __user *optlen)
5320 struct sctp_sock *sp = sctp_sk(sk);
5322 if (len < sizeof(int))
5327 if (put_user(len, optlen))
5329 if (copy_to_user(optval, &val, len))
5336 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5337 * (chapter and verse is quoted at sctp_setsockopt_context())
5339 static int sctp_getsockopt_context(struct sock *sk, int len,
5340 char __user *optval, int __user *optlen)
5342 struct sctp_assoc_value params;
5343 struct sctp_sock *sp;
5344 struct sctp_association *asoc;
5346 if (len < sizeof(struct sctp_assoc_value))
5349 len = sizeof(struct sctp_assoc_value);
5351 if (copy_from_user(¶ms, optval, len))
5356 if (params.assoc_id != 0) {
5357 asoc = sctp_id2assoc(sk, params.assoc_id);
5360 params.assoc_value = asoc->default_rcv_context;
5362 params.assoc_value = sp->default_rcv_context;
5365 if (put_user(len, optlen))
5367 if (copy_to_user(optval, ¶ms, len))
5374 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5375 * This option will get or set the maximum size to put in any outgoing
5376 * SCTP DATA chunk. If a message is larger than this size it will be
5377 * fragmented by SCTP into the specified size. Note that the underlying
5378 * SCTP implementation may fragment into smaller sized chunks when the
5379 * PMTU of the underlying association is smaller than the value set by
5380 * the user. The default value for this option is '0' which indicates
5381 * the user is NOT limiting fragmentation and only the PMTU will effect
5382 * SCTP's choice of DATA chunk size. Note also that values set larger
5383 * than the maximum size of an IP datagram will effectively let SCTP
5384 * control fragmentation (i.e. the same as setting this option to 0).
5386 * The following structure is used to access and modify this parameter:
5388 * struct sctp_assoc_value {
5389 * sctp_assoc_t assoc_id;
5390 * uint32_t assoc_value;
5393 * assoc_id: This parameter is ignored for one-to-one style sockets.
5394 * For one-to-many style sockets this parameter indicates which
5395 * association the user is performing an action upon. Note that if
5396 * this field's value is zero then the endpoints default value is
5397 * changed (effecting future associations only).
5398 * assoc_value: This parameter specifies the maximum size in bytes.
5400 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5401 char __user *optval, int __user *optlen)
5403 struct sctp_assoc_value params;
5404 struct sctp_association *asoc;
5406 if (len == sizeof(int)) {
5407 pr_warn_ratelimited(DEPRECATED
5409 "Use of int in maxseg socket option.\n"
5410 "Use struct sctp_assoc_value instead\n",
5411 current->comm, task_pid_nr(current));
5412 params.assoc_id = 0;
5413 } else if (len >= sizeof(struct sctp_assoc_value)) {
5414 len = sizeof(struct sctp_assoc_value);
5415 if (copy_from_user(¶ms, optval, sizeof(params)))
5420 asoc = sctp_id2assoc(sk, params.assoc_id);
5421 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5425 params.assoc_value = asoc->frag_point;
5427 params.assoc_value = sctp_sk(sk)->user_frag;
5429 if (put_user(len, optlen))
5431 if (len == sizeof(int)) {
5432 if (copy_to_user(optval, ¶ms.assoc_value, len))
5435 if (copy_to_user(optval, ¶ms, len))
5443 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5444 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5446 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5447 char __user *optval, int __user *optlen)
5451 if (len < sizeof(int))
5456 val = sctp_sk(sk)->frag_interleave;
5457 if (put_user(len, optlen))
5459 if (copy_to_user(optval, &val, len))
5466 * 7.1.25. Set or Get the sctp partial delivery point
5467 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5469 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5470 char __user *optval,
5475 if (len < sizeof(u32))
5480 val = sctp_sk(sk)->pd_point;
5481 if (put_user(len, optlen))
5483 if (copy_to_user(optval, &val, len))
5490 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5491 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5493 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5494 char __user *optval,
5497 struct sctp_assoc_value params;
5498 struct sctp_sock *sp;
5499 struct sctp_association *asoc;
5501 if (len == sizeof(int)) {
5502 pr_warn_ratelimited(DEPRECATED
5504 "Use of int in max_burst socket option.\n"
5505 "Use struct sctp_assoc_value instead\n",
5506 current->comm, task_pid_nr(current));
5507 params.assoc_id = 0;
5508 } else if (len >= sizeof(struct sctp_assoc_value)) {
5509 len = sizeof(struct sctp_assoc_value);
5510 if (copy_from_user(¶ms, optval, len))
5517 if (params.assoc_id != 0) {
5518 asoc = sctp_id2assoc(sk, params.assoc_id);
5521 params.assoc_value = asoc->max_burst;
5523 params.assoc_value = sp->max_burst;
5525 if (len == sizeof(int)) {
5526 if (copy_to_user(optval, ¶ms.assoc_value, len))
5529 if (copy_to_user(optval, ¶ms, len))
5537 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5538 char __user *optval, int __user *optlen)
5540 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5541 struct sctp_hmacalgo __user *p = (void __user *)optval;
5542 struct sctp_hmac_algo_param *hmacs;
5546 if (!ep->auth_enable)
5549 hmacs = ep->auth_hmacs_list;
5550 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5552 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5555 len = sizeof(struct sctp_hmacalgo) + data_len;
5556 num_idents = data_len / sizeof(u16);
5558 if (put_user(len, optlen))
5560 if (put_user(num_idents, &p->shmac_num_idents))
5562 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5567 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5568 char __user *optval, int __user *optlen)
5570 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5571 struct sctp_authkeyid val;
5572 struct sctp_association *asoc;
5574 if (!ep->auth_enable)
5577 if (len < sizeof(struct sctp_authkeyid))
5579 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5582 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5583 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5587 val.scact_keynumber = asoc->active_key_id;
5589 val.scact_keynumber = ep->active_key_id;
5591 len = sizeof(struct sctp_authkeyid);
5592 if (put_user(len, optlen))
5594 if (copy_to_user(optval, &val, len))
5600 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5601 char __user *optval, int __user *optlen)
5603 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5604 struct sctp_authchunks __user *p = (void __user *)optval;
5605 struct sctp_authchunks val;
5606 struct sctp_association *asoc;
5607 struct sctp_chunks_param *ch;
5611 if (!ep->auth_enable)
5614 if (len < sizeof(struct sctp_authchunks))
5617 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5620 to = p->gauth_chunks;
5621 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5625 ch = asoc->peer.peer_chunks;
5629 /* See if the user provided enough room for all the data */
5630 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5631 if (len < num_chunks)
5634 if (copy_to_user(to, ch->chunks, num_chunks))
5637 len = sizeof(struct sctp_authchunks) + num_chunks;
5638 if (put_user(len, optlen))
5640 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5645 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5646 char __user *optval, int __user *optlen)
5648 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5649 struct sctp_authchunks __user *p = (void __user *)optval;
5650 struct sctp_authchunks val;
5651 struct sctp_association *asoc;
5652 struct sctp_chunks_param *ch;
5656 if (!ep->auth_enable)
5659 if (len < sizeof(struct sctp_authchunks))
5662 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5665 to = p->gauth_chunks;
5666 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5667 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5671 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
5673 ch = ep->auth_chunk_list;
5678 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5679 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5682 if (copy_to_user(to, ch->chunks, num_chunks))
5685 len = sizeof(struct sctp_authchunks) + num_chunks;
5686 if (put_user(len, optlen))
5688 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5695 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5696 * This option gets the current number of associations that are attached
5697 * to a one-to-many style socket. The option value is an uint32_t.
5699 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5700 char __user *optval, int __user *optlen)
5702 struct sctp_sock *sp = sctp_sk(sk);
5703 struct sctp_association *asoc;
5706 if (sctp_style(sk, TCP))
5709 if (len < sizeof(u32))
5714 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5718 if (put_user(len, optlen))
5720 if (copy_to_user(optval, &val, len))
5727 * 8.1.23 SCTP_AUTO_ASCONF
5728 * See the corresponding setsockopt entry as description
5730 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5731 char __user *optval, int __user *optlen)
5735 if (len < sizeof(int))
5739 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5741 if (put_user(len, optlen))
5743 if (copy_to_user(optval, &val, len))
5749 * 8.2.6. Get the Current Identifiers of Associations
5750 * (SCTP_GET_ASSOC_ID_LIST)
5752 * This option gets the current list of SCTP association identifiers of
5753 * the SCTP associations handled by a one-to-many style socket.
5755 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5756 char __user *optval, int __user *optlen)
5758 struct sctp_sock *sp = sctp_sk(sk);
5759 struct sctp_association *asoc;
5760 struct sctp_assoc_ids *ids;
5763 if (sctp_style(sk, TCP))
5766 if (len < sizeof(struct sctp_assoc_ids))
5769 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5773 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5776 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5778 ids = kmalloc(len, GFP_KERNEL);
5782 ids->gaids_number_of_ids = num;
5784 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5785 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5788 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5798 * SCTP_PEER_ADDR_THLDS
5800 * This option allows us to fetch the partially failed threshold for one or all
5801 * transports in an association. See Section 6.1 of:
5802 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5804 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5805 char __user *optval,
5809 struct sctp_paddrthlds val;
5810 struct sctp_transport *trans;
5811 struct sctp_association *asoc;
5813 if (len < sizeof(struct sctp_paddrthlds))
5815 len = sizeof(struct sctp_paddrthlds);
5816 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5819 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5820 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5824 val.spt_pathpfthld = asoc->pf_retrans;
5825 val.spt_pathmaxrxt = asoc->pathmaxrxt;
5827 trans = sctp_addr_id2transport(sk, &val.spt_address,
5832 val.spt_pathmaxrxt = trans->pathmaxrxt;
5833 val.spt_pathpfthld = trans->pf_retrans;
5836 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5843 * SCTP_GET_ASSOC_STATS
5845 * This option retrieves local per endpoint statistics. It is modeled
5846 * after OpenSolaris' implementation
5848 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
5849 char __user *optval,
5852 struct sctp_assoc_stats sas;
5853 struct sctp_association *asoc = NULL;
5855 /* User must provide at least the assoc id */
5856 if (len < sizeof(sctp_assoc_t))
5859 /* Allow the struct to grow and fill in as much as possible */
5860 len = min_t(size_t, len, sizeof(sas));
5862 if (copy_from_user(&sas, optval, len))
5865 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
5869 sas.sas_rtxchunks = asoc->stats.rtxchunks;
5870 sas.sas_gapcnt = asoc->stats.gapcnt;
5871 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
5872 sas.sas_osacks = asoc->stats.osacks;
5873 sas.sas_isacks = asoc->stats.isacks;
5874 sas.sas_octrlchunks = asoc->stats.octrlchunks;
5875 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
5876 sas.sas_oodchunks = asoc->stats.oodchunks;
5877 sas.sas_iodchunks = asoc->stats.iodchunks;
5878 sas.sas_ouodchunks = asoc->stats.ouodchunks;
5879 sas.sas_iuodchunks = asoc->stats.iuodchunks;
5880 sas.sas_idupchunks = asoc->stats.idupchunks;
5881 sas.sas_opackets = asoc->stats.opackets;
5882 sas.sas_ipackets = asoc->stats.ipackets;
5884 /* New high max rto observed, will return 0 if not a single
5885 * RTO update took place. obs_rto_ipaddr will be bogus
5888 sas.sas_maxrto = asoc->stats.max_obs_rto;
5889 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
5890 sizeof(struct sockaddr_storage));
5892 /* Mark beginning of a new observation period */
5893 asoc->stats.max_obs_rto = asoc->rto_min;
5895 if (put_user(len, optlen))
5898 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
5900 if (copy_to_user(optval, &sas, len))
5906 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
5907 char __user *optval,
5912 if (len < sizeof(int))
5916 if (sctp_sk(sk)->recvrcvinfo)
5918 if (put_user(len, optlen))
5920 if (copy_to_user(optval, &val, len))
5926 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
5927 char __user *optval,
5932 if (len < sizeof(int))
5936 if (sctp_sk(sk)->recvnxtinfo)
5938 if (put_user(len, optlen))
5940 if (copy_to_user(optval, &val, len))
5946 static int sctp_getsockopt(struct sock *sk, int level, int optname,
5947 char __user *optval, int __user *optlen)
5952 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
5954 /* I can hardly begin to describe how wrong this is. This is
5955 * so broken as to be worse than useless. The API draft
5956 * REALLY is NOT helpful here... I am not convinced that the
5957 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5958 * are at all well-founded.
5960 if (level != SOL_SCTP) {
5961 struct sctp_af *af = sctp_sk(sk)->pf->af;
5963 retval = af->getsockopt(sk, level, optname, optval, optlen);
5967 if (get_user(len, optlen))
5974 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5976 case SCTP_DISABLE_FRAGMENTS:
5977 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5981 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5983 case SCTP_AUTOCLOSE:
5984 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5986 case SCTP_SOCKOPT_PEELOFF:
5987 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5989 case SCTP_PEER_ADDR_PARAMS:
5990 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5993 case SCTP_DELAYED_SACK:
5994 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5998 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
6000 case SCTP_GET_PEER_ADDRS:
6001 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
6004 case SCTP_GET_LOCAL_ADDRS:
6005 retval = sctp_getsockopt_local_addrs(sk, len, optval,
6008 case SCTP_SOCKOPT_CONNECTX3:
6009 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
6011 case SCTP_DEFAULT_SEND_PARAM:
6012 retval = sctp_getsockopt_default_send_param(sk, len,
6015 case SCTP_DEFAULT_SNDINFO:
6016 retval = sctp_getsockopt_default_sndinfo(sk, len,
6019 case SCTP_PRIMARY_ADDR:
6020 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
6023 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
6026 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
6028 case SCTP_ASSOCINFO:
6029 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
6031 case SCTP_I_WANT_MAPPED_V4_ADDR:
6032 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
6035 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
6037 case SCTP_GET_PEER_ADDR_INFO:
6038 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
6041 case SCTP_ADAPTATION_LAYER:
6042 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
6046 retval = sctp_getsockopt_context(sk, len, optval, optlen);
6048 case SCTP_FRAGMENT_INTERLEAVE:
6049 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
6052 case SCTP_PARTIAL_DELIVERY_POINT:
6053 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
6056 case SCTP_MAX_BURST:
6057 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
6060 case SCTP_AUTH_CHUNK:
6061 case SCTP_AUTH_DELETE_KEY:
6062 retval = -EOPNOTSUPP;
6064 case SCTP_HMAC_IDENT:
6065 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
6067 case SCTP_AUTH_ACTIVE_KEY:
6068 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
6070 case SCTP_PEER_AUTH_CHUNKS:
6071 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
6074 case SCTP_LOCAL_AUTH_CHUNKS:
6075 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
6078 case SCTP_GET_ASSOC_NUMBER:
6079 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
6081 case SCTP_GET_ASSOC_ID_LIST:
6082 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
6084 case SCTP_AUTO_ASCONF:
6085 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
6087 case SCTP_PEER_ADDR_THLDS:
6088 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
6090 case SCTP_GET_ASSOC_STATS:
6091 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
6093 case SCTP_RECVRCVINFO:
6094 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
6096 case SCTP_RECVNXTINFO:
6097 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
6100 retval = -ENOPROTOOPT;
6108 static void sctp_hash(struct sock *sk)
6113 static void sctp_unhash(struct sock *sk)
6118 /* Check if port is acceptable. Possibly find first available port.
6120 * The port hash table (contained in the 'global' SCTP protocol storage
6121 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
6122 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
6123 * list (the list number is the port number hashed out, so as you
6124 * would expect from a hash function, all the ports in a given list have
6125 * such a number that hashes out to the same list number; you were
6126 * expecting that, right?); so each list has a set of ports, with a
6127 * link to the socket (struct sock) that uses it, the port number and
6128 * a fastreuse flag (FIXME: NPI ipg).
6130 static struct sctp_bind_bucket *sctp_bucket_create(
6131 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
6133 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
6135 struct sctp_bind_hashbucket *head; /* hash list */
6136 struct sctp_bind_bucket *pp;
6137 unsigned short snum;
6140 snum = ntohs(addr->v4.sin_port);
6142 pr_debug("%s: begins, snum:%d\n", __func__, snum);
6147 /* Search for an available port. */
6148 int low, high, remaining, index;
6150 struct net *net = sock_net(sk);
6152 inet_get_local_port_range(net, &low, &high);
6153 remaining = (high - low) + 1;
6154 rover = prandom_u32() % remaining + low;
6158 if ((rover < low) || (rover > high))
6160 if (inet_is_local_reserved_port(net, rover))
6162 index = sctp_phashfn(sock_net(sk), rover);
6163 head = &sctp_port_hashtable[index];
6164 spin_lock(&head->lock);
6165 sctp_for_each_hentry(pp, &head->chain)
6166 if ((pp->port == rover) &&
6167 net_eq(sock_net(sk), pp->net))
6171 spin_unlock(&head->lock);
6172 } while (--remaining > 0);
6174 /* Exhausted local port range during search? */
6179 /* OK, here is the one we will use. HEAD (the port
6180 * hash table list entry) is non-NULL and we hold it's
6185 /* We are given an specific port number; we verify
6186 * that it is not being used. If it is used, we will
6187 * exahust the search in the hash list corresponding
6188 * to the port number (snum) - we detect that with the
6189 * port iterator, pp being NULL.
6191 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
6192 spin_lock(&head->lock);
6193 sctp_for_each_hentry(pp, &head->chain) {
6194 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
6201 if (!hlist_empty(&pp->owner)) {
6202 /* We had a port hash table hit - there is an
6203 * available port (pp != NULL) and it is being
6204 * used by other socket (pp->owner not empty); that other
6205 * socket is going to be sk2.
6207 int reuse = sk->sk_reuse;
6210 pr_debug("%s: found a possible match\n", __func__);
6212 if (pp->fastreuse && sk->sk_reuse &&
6213 sk->sk_state != SCTP_SS_LISTENING)
6216 /* Run through the list of sockets bound to the port
6217 * (pp->port) [via the pointers bind_next and
6218 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6219 * we get the endpoint they describe and run through
6220 * the endpoint's list of IP (v4 or v6) addresses,
6221 * comparing each of the addresses with the address of
6222 * the socket sk. If we find a match, then that means
6223 * that this port/socket (sk) combination are already
6226 sk_for_each_bound(sk2, &pp->owner) {
6227 struct sctp_endpoint *ep2;
6228 ep2 = sctp_sk(sk2)->ep;
6231 (reuse && sk2->sk_reuse &&
6232 sk2->sk_state != SCTP_SS_LISTENING))
6235 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
6236 sctp_sk(sk2), sctp_sk(sk))) {
6242 pr_debug("%s: found a match\n", __func__);
6245 /* If there was a hash table miss, create a new port. */
6247 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6250 /* In either case (hit or miss), make sure fastreuse is 1 only
6251 * if sk->sk_reuse is too (that is, if the caller requested
6252 * SO_REUSEADDR on this socket -sk-).
6254 if (hlist_empty(&pp->owner)) {
6255 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6259 } else if (pp->fastreuse &&
6260 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6263 /* We are set, so fill up all the data in the hash table
6264 * entry, tie the socket list information with the rest of the
6265 * sockets FIXME: Blurry, NPI (ipg).
6268 if (!sctp_sk(sk)->bind_hash) {
6269 inet_sk(sk)->inet_num = snum;
6270 sk_add_bind_node(sk, &pp->owner);
6271 sctp_sk(sk)->bind_hash = pp;
6276 spin_unlock(&head->lock);
6283 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6284 * port is requested.
6286 static int sctp_get_port(struct sock *sk, unsigned short snum)
6288 union sctp_addr addr;
6289 struct sctp_af *af = sctp_sk(sk)->pf->af;
6291 /* Set up a dummy address struct from the sk. */
6292 af->from_sk(&addr, sk);
6293 addr.v4.sin_port = htons(snum);
6295 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6296 return !!sctp_get_port_local(sk, &addr);
6300 * Move a socket to LISTENING state.
6302 static int sctp_listen_start(struct sock *sk, int backlog)
6304 struct sctp_sock *sp = sctp_sk(sk);
6305 struct sctp_endpoint *ep = sp->ep;
6306 struct crypto_hash *tfm = NULL;
6309 /* Allocate HMAC for generating cookie. */
6310 if (!sp->hmac && sp->sctp_hmac_alg) {
6311 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6312 tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
6314 net_info_ratelimited("failed to load transform for %s: %ld\n",
6315 sp->sctp_hmac_alg, PTR_ERR(tfm));
6318 sctp_sk(sk)->hmac = tfm;
6322 * If a bind() or sctp_bindx() is not called prior to a listen()
6323 * call that allows new associations to be accepted, the system
6324 * picks an ephemeral port and will choose an address set equivalent
6325 * to binding with a wildcard address.
6327 * This is not currently spelled out in the SCTP sockets
6328 * extensions draft, but follows the practice as seen in TCP
6332 sk->sk_state = SCTP_SS_LISTENING;
6333 if (!ep->base.bind_addr.port) {
6334 if (sctp_autobind(sk))
6337 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6338 sk->sk_state = SCTP_SS_CLOSED;
6343 sk->sk_max_ack_backlog = backlog;
6344 sctp_hash_endpoint(ep);
6349 * 4.1.3 / 5.1.3 listen()
6351 * By default, new associations are not accepted for UDP style sockets.
6352 * An application uses listen() to mark a socket as being able to
6353 * accept new associations.
6355 * On TCP style sockets, applications use listen() to ready the SCTP
6356 * endpoint for accepting inbound associations.
6358 * On both types of endpoints a backlog of '0' disables listening.
6360 * Move a socket to LISTENING state.
6362 int sctp_inet_listen(struct socket *sock, int backlog)
6364 struct sock *sk = sock->sk;
6365 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6368 if (unlikely(backlog < 0))
6373 /* Peeled-off sockets are not allowed to listen(). */
6374 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6377 if (sock->state != SS_UNCONNECTED)
6380 /* If backlog is zero, disable listening. */
6382 if (sctp_sstate(sk, CLOSED))
6386 sctp_unhash_endpoint(ep);
6387 sk->sk_state = SCTP_SS_CLOSED;
6389 sctp_sk(sk)->bind_hash->fastreuse = 1;
6393 /* If we are already listening, just update the backlog */
6394 if (sctp_sstate(sk, LISTENING))
6395 sk->sk_max_ack_backlog = backlog;
6397 err = sctp_listen_start(sk, backlog);
6409 * This function is done by modeling the current datagram_poll() and the
6410 * tcp_poll(). Note that, based on these implementations, we don't
6411 * lock the socket in this function, even though it seems that,
6412 * ideally, locking or some other mechanisms can be used to ensure
6413 * the integrity of the counters (sndbuf and wmem_alloc) used
6414 * in this place. We assume that we don't need locks either until proven
6417 * Another thing to note is that we include the Async I/O support
6418 * here, again, by modeling the current TCP/UDP code. We don't have
6419 * a good way to test with it yet.
6421 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6423 struct sock *sk = sock->sk;
6424 struct sctp_sock *sp = sctp_sk(sk);
6427 poll_wait(file, sk_sleep(sk), wait);
6429 /* A TCP-style listening socket becomes readable when the accept queue
6432 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6433 return (!list_empty(&sp->ep->asocs)) ?
6434 (POLLIN | POLLRDNORM) : 0;
6438 /* Is there any exceptional events? */
6439 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6441 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6442 if (sk->sk_shutdown & RCV_SHUTDOWN)
6443 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6444 if (sk->sk_shutdown == SHUTDOWN_MASK)
6447 /* Is it readable? Reconsider this code with TCP-style support. */
6448 if (!skb_queue_empty(&sk->sk_receive_queue))
6449 mask |= POLLIN | POLLRDNORM;
6451 /* The association is either gone or not ready. */
6452 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6455 /* Is it writable? */
6456 if (sctp_writeable(sk)) {
6457 mask |= POLLOUT | POLLWRNORM;
6459 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6461 * Since the socket is not locked, the buffer
6462 * might be made available after the writeable check and
6463 * before the bit is set. This could cause a lost I/O
6464 * signal. tcp_poll() has a race breaker for this race
6465 * condition. Based on their implementation, we put
6466 * in the following code to cover it as well.
6468 if (sctp_writeable(sk))
6469 mask |= POLLOUT | POLLWRNORM;
6474 /********************************************************************
6475 * 2nd Level Abstractions
6476 ********************************************************************/
6478 static struct sctp_bind_bucket *sctp_bucket_create(
6479 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6481 struct sctp_bind_bucket *pp;
6483 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6485 SCTP_DBG_OBJCNT_INC(bind_bucket);
6488 INIT_HLIST_HEAD(&pp->owner);
6490 hlist_add_head(&pp->node, &head->chain);
6495 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6496 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6498 if (pp && hlist_empty(&pp->owner)) {
6499 __hlist_del(&pp->node);
6500 kmem_cache_free(sctp_bucket_cachep, pp);
6501 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6505 /* Release this socket's reference to a local port. */
6506 static inline void __sctp_put_port(struct sock *sk)
6508 struct sctp_bind_hashbucket *head =
6509 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6510 inet_sk(sk)->inet_num)];
6511 struct sctp_bind_bucket *pp;
6513 spin_lock(&head->lock);
6514 pp = sctp_sk(sk)->bind_hash;
6515 __sk_del_bind_node(sk);
6516 sctp_sk(sk)->bind_hash = NULL;
6517 inet_sk(sk)->inet_num = 0;
6518 sctp_bucket_destroy(pp);
6519 spin_unlock(&head->lock);
6522 void sctp_put_port(struct sock *sk)
6525 __sctp_put_port(sk);
6530 * The system picks an ephemeral port and choose an address set equivalent
6531 * to binding with a wildcard address.
6532 * One of those addresses will be the primary address for the association.
6533 * This automatically enables the multihoming capability of SCTP.
6535 static int sctp_autobind(struct sock *sk)
6537 union sctp_addr autoaddr;
6541 /* Initialize a local sockaddr structure to INADDR_ANY. */
6542 af = sctp_sk(sk)->pf->af;
6544 port = htons(inet_sk(sk)->inet_num);
6545 af->inaddr_any(&autoaddr, port);
6547 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6550 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6553 * 4.2 The cmsghdr Structure *
6555 * When ancillary data is sent or received, any number of ancillary data
6556 * objects can be specified by the msg_control and msg_controllen members of
6557 * the msghdr structure, because each object is preceded by
6558 * a cmsghdr structure defining the object's length (the cmsg_len member).
6559 * Historically Berkeley-derived implementations have passed only one object
6560 * at a time, but this API allows multiple objects to be
6561 * passed in a single call to sendmsg() or recvmsg(). The following example
6562 * shows two ancillary data objects in a control buffer.
6564 * |<--------------------------- msg_controllen -------------------------->|
6567 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6569 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6572 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6574 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6577 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6578 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6580 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6582 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6589 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
6591 struct cmsghdr *cmsg;
6592 struct msghdr *my_msg = (struct msghdr *)msg;
6594 for_each_cmsghdr(cmsg, my_msg) {
6595 if (!CMSG_OK(my_msg, cmsg))
6598 /* Should we parse this header or ignore? */
6599 if (cmsg->cmsg_level != IPPROTO_SCTP)
6602 /* Strictly check lengths following example in SCM code. */
6603 switch (cmsg->cmsg_type) {
6605 /* SCTP Socket API Extension
6606 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
6608 * This cmsghdr structure provides information for
6609 * initializing new SCTP associations with sendmsg().
6610 * The SCTP_INITMSG socket option uses this same data
6611 * structure. This structure is not used for
6614 * cmsg_level cmsg_type cmsg_data[]
6615 * ------------ ------------ ----------------------
6616 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6618 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
6621 cmsgs->init = CMSG_DATA(cmsg);
6625 /* SCTP Socket API Extension
6626 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
6628 * This cmsghdr structure specifies SCTP options for
6629 * sendmsg() and describes SCTP header information
6630 * about a received message through recvmsg().
6632 * cmsg_level cmsg_type cmsg_data[]
6633 * ------------ ------------ ----------------------
6634 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6636 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6639 cmsgs->srinfo = CMSG_DATA(cmsg);
6641 if (cmsgs->srinfo->sinfo_flags &
6642 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6643 SCTP_ABORT | SCTP_EOF))
6648 /* SCTP Socket API Extension
6649 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
6651 * This cmsghdr structure specifies SCTP options for
6652 * sendmsg(). This structure and SCTP_RCVINFO replaces
6653 * SCTP_SNDRCV which has been deprecated.
6655 * cmsg_level cmsg_type cmsg_data[]
6656 * ------------ ------------ ---------------------
6657 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
6659 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
6662 cmsgs->sinfo = CMSG_DATA(cmsg);
6664 if (cmsgs->sinfo->snd_flags &
6665 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6666 SCTP_ABORT | SCTP_EOF))
6678 * Wait for a packet..
6679 * Note: This function is the same function as in core/datagram.c
6680 * with a few modifications to make lksctp work.
6682 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
6687 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6689 /* Socket errors? */
6690 error = sock_error(sk);
6694 if (!skb_queue_empty(&sk->sk_receive_queue))
6697 /* Socket shut down? */
6698 if (sk->sk_shutdown & RCV_SHUTDOWN)
6701 /* Sequenced packets can come disconnected. If so we report the
6706 /* Is there a good reason to think that we may receive some data? */
6707 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6710 /* Handle signals. */
6711 if (signal_pending(current))
6714 /* Let another process have a go. Since we are going to sleep
6715 * anyway. Note: This may cause odd behaviors if the message
6716 * does not fit in the user's buffer, but this seems to be the
6717 * only way to honor MSG_DONTWAIT realistically.
6720 *timeo_p = schedule_timeout(*timeo_p);
6724 finish_wait(sk_sleep(sk), &wait);
6728 error = sock_intr_errno(*timeo_p);
6731 finish_wait(sk_sleep(sk), &wait);
6736 /* Receive a datagram.
6737 * Note: This is pretty much the same routine as in core/datagram.c
6738 * with a few changes to make lksctp work.
6740 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6741 int noblock, int *err)
6744 struct sk_buff *skb;
6747 timeo = sock_rcvtimeo(sk, noblock);
6749 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
6750 MAX_SCHEDULE_TIMEOUT);
6753 /* Again only user level code calls this function,
6754 * so nothing interrupt level
6755 * will suddenly eat the receive_queue.
6757 * Look at current nfs client by the way...
6758 * However, this function was correct in any case. 8)
6760 if (flags & MSG_PEEK) {
6761 spin_lock_bh(&sk->sk_receive_queue.lock);
6762 skb = skb_peek(&sk->sk_receive_queue);
6764 atomic_inc(&skb->users);
6765 spin_unlock_bh(&sk->sk_receive_queue.lock);
6767 skb = skb_dequeue(&sk->sk_receive_queue);
6773 /* Caller is allowed not to check sk->sk_err before calling. */
6774 error = sock_error(sk);
6778 if (sk->sk_shutdown & RCV_SHUTDOWN)
6781 if (sk_can_busy_loop(sk) &&
6782 sk_busy_loop(sk, noblock))
6785 /* User doesn't want to wait. */
6789 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6798 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6799 static void __sctp_write_space(struct sctp_association *asoc)
6801 struct sock *sk = asoc->base.sk;
6802 struct socket *sock = sk->sk_socket;
6804 if ((sctp_wspace(asoc) > 0) && sock) {
6805 if (waitqueue_active(&asoc->wait))
6806 wake_up_interruptible(&asoc->wait);
6808 if (sctp_writeable(sk)) {
6809 wait_queue_head_t *wq = sk_sleep(sk);
6811 if (wq && waitqueue_active(wq))
6812 wake_up_interruptible(wq);
6814 /* Note that we try to include the Async I/O support
6815 * here by modeling from the current TCP/UDP code.
6816 * We have not tested with it yet.
6818 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6819 sock_wake_async(sock,
6820 SOCK_WAKE_SPACE, POLL_OUT);
6825 static void sctp_wake_up_waiters(struct sock *sk,
6826 struct sctp_association *asoc)
6828 struct sctp_association *tmp = asoc;
6830 /* We do accounting for the sndbuf space per association,
6831 * so we only need to wake our own association.
6833 if (asoc->ep->sndbuf_policy)
6834 return __sctp_write_space(asoc);
6836 /* If association goes down and is just flushing its
6837 * outq, then just normally notify others.
6839 if (asoc->base.dead)
6840 return sctp_write_space(sk);
6842 /* Accounting for the sndbuf space is per socket, so we
6843 * need to wake up others, try to be fair and in case of
6844 * other associations, let them have a go first instead
6845 * of just doing a sctp_write_space() call.
6847 * Note that we reach sctp_wake_up_waiters() only when
6848 * associations free up queued chunks, thus we are under
6849 * lock and the list of associations on a socket is
6850 * guaranteed not to change.
6852 for (tmp = list_next_entry(tmp, asocs); 1;
6853 tmp = list_next_entry(tmp, asocs)) {
6854 /* Manually skip the head element. */
6855 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
6857 /* Wake up association. */
6858 __sctp_write_space(tmp);
6859 /* We've reached the end. */
6865 /* Do accounting for the sndbuf space.
6866 * Decrement the used sndbuf space of the corresponding association by the
6867 * data size which was just transmitted(freed).
6869 static void sctp_wfree(struct sk_buff *skb)
6871 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
6872 struct sctp_association *asoc = chunk->asoc;
6873 struct sock *sk = asoc->base.sk;
6875 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6876 sizeof(struct sk_buff) +
6877 sizeof(struct sctp_chunk);
6879 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6882 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6884 sk->sk_wmem_queued -= skb->truesize;
6885 sk_mem_uncharge(sk, skb->truesize);
6888 sctp_wake_up_waiters(sk, asoc);
6890 sctp_association_put(asoc);
6893 /* Do accounting for the receive space on the socket.
6894 * Accounting for the association is done in ulpevent.c
6895 * We set this as a destructor for the cloned data skbs so that
6896 * accounting is done at the correct time.
6898 void sctp_sock_rfree(struct sk_buff *skb)
6900 struct sock *sk = skb->sk;
6901 struct sctp_ulpevent *event = sctp_skb2event(skb);
6903 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6906 * Mimic the behavior of sock_rfree
6908 sk_mem_uncharge(sk, event->rmem_len);
6912 /* Helper function to wait for space in the sndbuf. */
6913 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6916 struct sock *sk = asoc->base.sk;
6918 long current_timeo = *timeo_p;
6921 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
6924 /* Increment the association's refcnt. */
6925 sctp_association_hold(asoc);
6927 /* Wait on the association specific sndbuf space. */
6929 prepare_to_wait_exclusive(&asoc->wait, &wait,
6930 TASK_INTERRUPTIBLE);
6933 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6936 if (signal_pending(current))
6937 goto do_interrupted;
6938 if (msg_len <= sctp_wspace(asoc))
6941 /* Let another process have a go. Since we are going
6945 current_timeo = schedule_timeout(current_timeo);
6946 BUG_ON(sk != asoc->base.sk);
6949 *timeo_p = current_timeo;
6953 finish_wait(&asoc->wait, &wait);
6955 /* Release the association's refcnt. */
6956 sctp_association_put(asoc);
6965 err = sock_intr_errno(*timeo_p);
6973 void sctp_data_ready(struct sock *sk)
6975 struct socket_wq *wq;
6978 wq = rcu_dereference(sk->sk_wq);
6979 if (wq_has_sleeper(wq))
6980 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6981 POLLRDNORM | POLLRDBAND);
6982 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6986 /* If socket sndbuf has changed, wake up all per association waiters. */
6987 void sctp_write_space(struct sock *sk)
6989 struct sctp_association *asoc;
6991 /* Wake up the tasks in each wait queue. */
6992 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6993 __sctp_write_space(asoc);
6997 /* Is there any sndbuf space available on the socket?
6999 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
7000 * associations on the same socket. For a UDP-style socket with
7001 * multiple associations, it is possible for it to be "unwriteable"
7002 * prematurely. I assume that this is acceptable because
7003 * a premature "unwriteable" is better than an accidental "writeable" which
7004 * would cause an unwanted block under certain circumstances. For the 1-1
7005 * UDP-style sockets or TCP-style sockets, this code should work.
7008 static int sctp_writeable(struct sock *sk)
7012 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
7018 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
7019 * returns immediately with EINPROGRESS.
7021 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
7023 struct sock *sk = asoc->base.sk;
7025 long current_timeo = *timeo_p;
7028 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
7030 /* Increment the association's refcnt. */
7031 sctp_association_hold(asoc);
7034 prepare_to_wait_exclusive(&asoc->wait, &wait,
7035 TASK_INTERRUPTIBLE);
7038 if (sk->sk_shutdown & RCV_SHUTDOWN)
7040 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7043 if (signal_pending(current))
7044 goto do_interrupted;
7046 if (sctp_state(asoc, ESTABLISHED))
7049 /* Let another process have a go. Since we are going
7053 current_timeo = schedule_timeout(current_timeo);
7056 *timeo_p = current_timeo;
7060 finish_wait(&asoc->wait, &wait);
7062 /* Release the association's refcnt. */
7063 sctp_association_put(asoc);
7068 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
7071 err = -ECONNREFUSED;
7075 err = sock_intr_errno(*timeo_p);
7083 static int sctp_wait_for_accept(struct sock *sk, long timeo)
7085 struct sctp_endpoint *ep;
7089 ep = sctp_sk(sk)->ep;
7093 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
7094 TASK_INTERRUPTIBLE);
7096 if (list_empty(&ep->asocs)) {
7098 timeo = schedule_timeout(timeo);
7103 if (!sctp_sstate(sk, LISTENING))
7107 if (!list_empty(&ep->asocs))
7110 err = sock_intr_errno(timeo);
7111 if (signal_pending(current))
7119 finish_wait(sk_sleep(sk), &wait);
7124 static void sctp_wait_for_close(struct sock *sk, long timeout)
7129 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7130 if (list_empty(&sctp_sk(sk)->ep->asocs))
7133 timeout = schedule_timeout(timeout);
7135 } while (!signal_pending(current) && timeout);
7137 finish_wait(sk_sleep(sk), &wait);
7140 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
7142 struct sk_buff *frag;
7147 /* Don't forget the fragments. */
7148 skb_walk_frags(skb, frag)
7149 sctp_skb_set_owner_r_frag(frag, sk);
7152 sctp_skb_set_owner_r(skb, sk);
7155 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
7156 struct sctp_association *asoc)
7158 struct inet_sock *inet = inet_sk(sk);
7159 struct inet_sock *newinet;
7161 newsk->sk_type = sk->sk_type;
7162 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
7163 newsk->sk_flags = sk->sk_flags;
7164 newsk->sk_no_check_tx = sk->sk_no_check_tx;
7165 newsk->sk_no_check_rx = sk->sk_no_check_rx;
7166 newsk->sk_reuse = sk->sk_reuse;
7168 newsk->sk_shutdown = sk->sk_shutdown;
7169 newsk->sk_destruct = sctp_destruct_sock;
7170 newsk->sk_family = sk->sk_family;
7171 newsk->sk_protocol = IPPROTO_SCTP;
7172 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
7173 newsk->sk_sndbuf = sk->sk_sndbuf;
7174 newsk->sk_rcvbuf = sk->sk_rcvbuf;
7175 newsk->sk_lingertime = sk->sk_lingertime;
7176 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
7177 newsk->sk_sndtimeo = sk->sk_sndtimeo;
7179 newinet = inet_sk(newsk);
7181 /* Initialize sk's sport, dport, rcv_saddr and daddr for
7182 * getsockname() and getpeername()
7184 newinet->inet_sport = inet->inet_sport;
7185 newinet->inet_saddr = inet->inet_saddr;
7186 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
7187 newinet->inet_dport = htons(asoc->peer.port);
7188 newinet->pmtudisc = inet->pmtudisc;
7189 newinet->inet_id = asoc->next_tsn ^ jiffies;
7191 newinet->uc_ttl = inet->uc_ttl;
7192 newinet->mc_loop = 1;
7193 newinet->mc_ttl = 1;
7194 newinet->mc_index = 0;
7195 newinet->mc_list = NULL;
7198 /* Populate the fields of the newsk from the oldsk and migrate the assoc
7199 * and its messages to the newsk.
7201 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
7202 struct sctp_association *assoc,
7203 sctp_socket_type_t type)
7205 struct sctp_sock *oldsp = sctp_sk(oldsk);
7206 struct sctp_sock *newsp = sctp_sk(newsk);
7207 struct sctp_bind_bucket *pp; /* hash list port iterator */
7208 struct sctp_endpoint *newep = newsp->ep;
7209 struct sk_buff *skb, *tmp;
7210 struct sctp_ulpevent *event;
7211 struct sctp_bind_hashbucket *head;
7212 struct list_head tmplist;
7214 /* Migrate socket buffer sizes and all the socket level options to the
7217 newsk->sk_sndbuf = oldsk->sk_sndbuf;
7218 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
7219 /* Brute force copy old sctp opt. */
7220 if (oldsp->do_auto_asconf) {
7221 memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
7222 inet_sk_copy_descendant(newsk, oldsk);
7223 memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
7225 inet_sk_copy_descendant(newsk, oldsk);
7227 /* Restore the ep value that was overwritten with the above structure
7233 /* Hook this new socket in to the bind_hash list. */
7234 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
7235 inet_sk(oldsk)->inet_num)];
7237 spin_lock(&head->lock);
7238 pp = sctp_sk(oldsk)->bind_hash;
7239 sk_add_bind_node(newsk, &pp->owner);
7240 sctp_sk(newsk)->bind_hash = pp;
7241 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
7242 spin_unlock(&head->lock);
7245 /* Copy the bind_addr list from the original endpoint to the new
7246 * endpoint so that we can handle restarts properly
7248 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
7249 &oldsp->ep->base.bind_addr, GFP_KERNEL);
7251 /* Move any messages in the old socket's receive queue that are for the
7252 * peeled off association to the new socket's receive queue.
7254 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
7255 event = sctp_skb2event(skb);
7256 if (event->asoc == assoc) {
7257 __skb_unlink(skb, &oldsk->sk_receive_queue);
7258 __skb_queue_tail(&newsk->sk_receive_queue, skb);
7259 sctp_skb_set_owner_r_frag(skb, newsk);
7263 /* Clean up any messages pending delivery due to partial
7264 * delivery. Three cases:
7265 * 1) No partial deliver; no work.
7266 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7267 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7269 skb_queue_head_init(&newsp->pd_lobby);
7270 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
7272 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
7273 struct sk_buff_head *queue;
7275 /* Decide which queue to move pd_lobby skbs to. */
7276 if (assoc->ulpq.pd_mode) {
7277 queue = &newsp->pd_lobby;
7279 queue = &newsk->sk_receive_queue;
7281 /* Walk through the pd_lobby, looking for skbs that
7282 * need moved to the new socket.
7284 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
7285 event = sctp_skb2event(skb);
7286 if (event->asoc == assoc) {
7287 __skb_unlink(skb, &oldsp->pd_lobby);
7288 __skb_queue_tail(queue, skb);
7289 sctp_skb_set_owner_r_frag(skb, newsk);
7293 /* Clear up any skbs waiting for the partial
7294 * delivery to finish.
7296 if (assoc->ulpq.pd_mode)
7297 sctp_clear_pd(oldsk, NULL);
7301 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7302 sctp_skb_set_owner_r_frag(skb, newsk);
7304 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7305 sctp_skb_set_owner_r_frag(skb, newsk);
7307 /* Set the type of socket to indicate that it is peeled off from the
7308 * original UDP-style socket or created with the accept() call on a
7309 * TCP-style socket..
7313 /* Mark the new socket "in-use" by the user so that any packets
7314 * that may arrive on the association after we've moved it are
7315 * queued to the backlog. This prevents a potential race between
7316 * backlog processing on the old socket and new-packet processing
7317 * on the new socket.
7319 * The caller has just allocated newsk so we can guarantee that other
7320 * paths won't try to lock it and then oldsk.
7322 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7323 sctp_assoc_migrate(assoc, newsk);
7325 /* If the association on the newsk is already closed before accept()
7326 * is called, set RCV_SHUTDOWN flag.
7328 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
7329 newsk->sk_shutdown |= RCV_SHUTDOWN;
7331 newsk->sk_state = SCTP_SS_ESTABLISHED;
7332 release_sock(newsk);
7336 /* This proto struct describes the ULP interface for SCTP. */
7337 struct proto sctp_prot = {
7339 .owner = THIS_MODULE,
7340 .close = sctp_close,
7341 .connect = sctp_connect,
7342 .disconnect = sctp_disconnect,
7343 .accept = sctp_accept,
7344 .ioctl = sctp_ioctl,
7345 .init = sctp_init_sock,
7346 .destroy = sctp_destroy_sock,
7347 .shutdown = sctp_shutdown,
7348 .setsockopt = sctp_setsockopt,
7349 .getsockopt = sctp_getsockopt,
7350 .sendmsg = sctp_sendmsg,
7351 .recvmsg = sctp_recvmsg,
7353 .backlog_rcv = sctp_backlog_rcv,
7355 .unhash = sctp_unhash,
7356 .get_port = sctp_get_port,
7357 .obj_size = sizeof(struct sctp_sock),
7358 .sysctl_mem = sysctl_sctp_mem,
7359 .sysctl_rmem = sysctl_sctp_rmem,
7360 .sysctl_wmem = sysctl_sctp_wmem,
7361 .memory_pressure = &sctp_memory_pressure,
7362 .enter_memory_pressure = sctp_enter_memory_pressure,
7363 .memory_allocated = &sctp_memory_allocated,
7364 .sockets_allocated = &sctp_sockets_allocated,
7367 #if IS_ENABLED(CONFIG_IPV6)
7369 struct proto sctpv6_prot = {
7371 .owner = THIS_MODULE,
7372 .close = sctp_close,
7373 .connect = sctp_connect,
7374 .disconnect = sctp_disconnect,
7375 .accept = sctp_accept,
7376 .ioctl = sctp_ioctl,
7377 .init = sctp_init_sock,
7378 .destroy = sctp_destroy_sock,
7379 .shutdown = sctp_shutdown,
7380 .setsockopt = sctp_setsockopt,
7381 .getsockopt = sctp_getsockopt,
7382 .sendmsg = sctp_sendmsg,
7383 .recvmsg = sctp_recvmsg,
7385 .backlog_rcv = sctp_backlog_rcv,
7387 .unhash = sctp_unhash,
7388 .get_port = sctp_get_port,
7389 .obj_size = sizeof(struct sctp6_sock),
7390 .sysctl_mem = sysctl_sctp_mem,
7391 .sysctl_rmem = sysctl_sctp_rmem,
7392 .sysctl_wmem = sysctl_sctp_wmem,
7393 .memory_pressure = &sctp_memory_pressure,
7394 .enter_memory_pressure = sctp_enter_memory_pressure,
7395 .memory_allocated = &sctp_memory_allocated,
7396 .sockets_allocated = &sctp_sockets_allocated,
7398 #endif /* IS_ENABLED(CONFIG_IPV6) */