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 <crypto/hash.h>
56 #include <linux/types.h>
57 #include <linux/kernel.h>
58 #include <linux/wait.h>
59 #include <linux/time.h>
61 #include <linux/capability.h>
62 #include <linux/fcntl.h>
63 #include <linux/poll.h>
64 #include <linux/init.h>
65 #include <linux/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) && !sctp_sstate(sk, CLOSING))
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, af->sockaddr_len,
390 SCTP_ADDR_SRC, GFP_ATOMIC);
392 /* Copy back into socket for getsockname() use. */
394 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
395 sp->pf->to_sk_saddr(addr, sk);
401 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
403 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
404 * at any one time. If a sender, after sending an ASCONF chunk, decides
405 * it needs to transfer another ASCONF Chunk, it MUST wait until the
406 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
407 * subsequent ASCONF. Note this restriction binds each side, so at any
408 * time two ASCONF may be in-transit on any given association (one sent
409 * from each endpoint).
411 static int sctp_send_asconf(struct sctp_association *asoc,
412 struct sctp_chunk *chunk)
414 struct net *net = sock_net(asoc->base.sk);
417 /* If there is an outstanding ASCONF chunk, queue it for later
420 if (asoc->addip_last_asconf) {
421 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
425 /* Hold the chunk until an ASCONF_ACK is received. */
426 sctp_chunk_hold(chunk);
427 retval = sctp_primitive_ASCONF(net, asoc, chunk);
429 sctp_chunk_free(chunk);
431 asoc->addip_last_asconf = chunk;
437 /* Add a list of addresses as bind addresses to local endpoint or
440 * Basically run through each address specified in the addrs/addrcnt
441 * array/length pair, determine if it is IPv6 or IPv4 and call
442 * sctp_do_bind() on it.
444 * If any of them fails, then the operation will be reversed and the
445 * ones that were added will be removed.
447 * Only sctp_setsockopt_bindx() is supposed to call this function.
449 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
454 struct sockaddr *sa_addr;
457 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
461 for (cnt = 0; cnt < addrcnt; cnt++) {
462 /* The list may contain either IPv4 or IPv6 address;
463 * determine the address length for walking thru the list.
466 af = sctp_get_af_specific(sa_addr->sa_family);
472 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
475 addr_buf += af->sockaddr_len;
479 /* Failed. Cleanup the ones that have been added */
481 sctp_bindx_rem(sk, addrs, cnt);
489 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
490 * associations that are part of the endpoint indicating that a list of local
491 * addresses are added to the endpoint.
493 * If any of the addresses is already in the bind address list of the
494 * association, we do not send the chunk for that association. But it will not
495 * affect other associations.
497 * Only sctp_setsockopt_bindx() is supposed to call this function.
499 static int sctp_send_asconf_add_ip(struct sock *sk,
500 struct sockaddr *addrs,
503 struct net *net = sock_net(sk);
504 struct sctp_sock *sp;
505 struct sctp_endpoint *ep;
506 struct sctp_association *asoc;
507 struct sctp_bind_addr *bp;
508 struct sctp_chunk *chunk;
509 struct sctp_sockaddr_entry *laddr;
510 union sctp_addr *addr;
511 union sctp_addr saveaddr;
518 if (!net->sctp.addip_enable)
524 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
525 __func__, sk, addrs, addrcnt);
527 list_for_each_entry(asoc, &ep->asocs, asocs) {
528 if (!asoc->peer.asconf_capable)
531 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
534 if (!sctp_state(asoc, ESTABLISHED))
537 /* Check if any address in the packed array of addresses is
538 * in the bind address list of the association. If so,
539 * do not send the asconf chunk to its peer, but continue with
540 * other associations.
543 for (i = 0; i < addrcnt; i++) {
545 af = sctp_get_af_specific(addr->v4.sin_family);
551 if (sctp_assoc_lookup_laddr(asoc, addr))
554 addr_buf += af->sockaddr_len;
559 /* Use the first valid address in bind addr list of
560 * association as Address Parameter of ASCONF CHUNK.
562 bp = &asoc->base.bind_addr;
563 p = bp->address_list.next;
564 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
565 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
566 addrcnt, SCTP_PARAM_ADD_IP);
572 /* Add the new addresses to the bind address list with
573 * use_as_src set to 0.
576 for (i = 0; i < addrcnt; i++) {
578 af = sctp_get_af_specific(addr->v4.sin_family);
579 memcpy(&saveaddr, addr, af->sockaddr_len);
580 retval = sctp_add_bind_addr(bp, &saveaddr,
582 SCTP_ADDR_NEW, GFP_ATOMIC);
583 addr_buf += af->sockaddr_len;
585 if (asoc->src_out_of_asoc_ok) {
586 struct sctp_transport *trans;
588 list_for_each_entry(trans,
589 &asoc->peer.transport_addr_list, transports) {
590 /* Clear the source and route cache */
591 dst_release(trans->dst);
592 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
593 2*asoc->pathmtu, 4380));
594 trans->ssthresh = asoc->peer.i.a_rwnd;
595 trans->rto = asoc->rto_initial;
596 sctp_max_rto(asoc, trans);
597 trans->rtt = trans->srtt = trans->rttvar = 0;
598 sctp_transport_route(trans, NULL,
599 sctp_sk(asoc->base.sk));
602 retval = sctp_send_asconf(asoc, chunk);
609 /* Remove a list of addresses from bind addresses list. Do not remove the
612 * Basically run through each address specified in the addrs/addrcnt
613 * array/length pair, determine if it is IPv6 or IPv4 and call
614 * sctp_del_bind() on it.
616 * If any of them fails, then the operation will be reversed and the
617 * ones that were removed will be added back.
619 * At least one address has to be left; if only one address is
620 * available, the operation will return -EBUSY.
622 * Only sctp_setsockopt_bindx() is supposed to call this function.
624 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
626 struct sctp_sock *sp = sctp_sk(sk);
627 struct sctp_endpoint *ep = sp->ep;
629 struct sctp_bind_addr *bp = &ep->base.bind_addr;
632 union sctp_addr *sa_addr;
635 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
636 __func__, sk, addrs, addrcnt);
639 for (cnt = 0; cnt < addrcnt; cnt++) {
640 /* If the bind address list is empty or if there is only one
641 * bind address, there is nothing more to be removed (we need
642 * at least one address here).
644 if (list_empty(&bp->address_list) ||
645 (sctp_list_single_entry(&bp->address_list))) {
651 af = sctp_get_af_specific(sa_addr->sa.sa_family);
657 if (!af->addr_valid(sa_addr, sp, NULL)) {
658 retval = -EADDRNOTAVAIL;
662 if (sa_addr->v4.sin_port &&
663 sa_addr->v4.sin_port != htons(bp->port)) {
668 if (!sa_addr->v4.sin_port)
669 sa_addr->v4.sin_port = htons(bp->port);
671 /* FIXME - There is probably a need to check if sk->sk_saddr and
672 * sk->sk_rcv_addr are currently set to one of the addresses to
673 * be removed. This is something which needs to be looked into
674 * when we are fixing the outstanding issues with multi-homing
675 * socket routing and failover schemes. Refer to comments in
676 * sctp_do_bind(). -daisy
678 retval = sctp_del_bind_addr(bp, sa_addr);
680 addr_buf += af->sockaddr_len;
683 /* Failed. Add the ones that has been removed back */
685 sctp_bindx_add(sk, addrs, cnt);
693 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
694 * the associations that are part of the endpoint indicating that a list of
695 * local addresses are removed from the endpoint.
697 * If any of the addresses is already in the bind address list of the
698 * association, we do not send the chunk for that association. But it will not
699 * affect other associations.
701 * Only sctp_setsockopt_bindx() is supposed to call this function.
703 static int sctp_send_asconf_del_ip(struct sock *sk,
704 struct sockaddr *addrs,
707 struct net *net = sock_net(sk);
708 struct sctp_sock *sp;
709 struct sctp_endpoint *ep;
710 struct sctp_association *asoc;
711 struct sctp_transport *transport;
712 struct sctp_bind_addr *bp;
713 struct sctp_chunk *chunk;
714 union sctp_addr *laddr;
717 struct sctp_sockaddr_entry *saddr;
723 if (!net->sctp.addip_enable)
729 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
730 __func__, sk, addrs, addrcnt);
732 list_for_each_entry(asoc, &ep->asocs, asocs) {
734 if (!asoc->peer.asconf_capable)
737 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
740 if (!sctp_state(asoc, ESTABLISHED))
743 /* Check if any address in the packed array of addresses is
744 * not present in the bind address list of the association.
745 * If so, do not send the asconf chunk to its peer, but
746 * continue with other associations.
749 for (i = 0; i < addrcnt; i++) {
751 af = sctp_get_af_specific(laddr->v4.sin_family);
757 if (!sctp_assoc_lookup_laddr(asoc, laddr))
760 addr_buf += af->sockaddr_len;
765 /* Find one address in the association's bind address list
766 * that is not in the packed array of addresses. This is to
767 * make sure that we do not delete all the addresses in the
770 bp = &asoc->base.bind_addr;
771 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
773 if ((laddr == NULL) && (addrcnt == 1)) {
774 if (asoc->asconf_addr_del_pending)
776 asoc->asconf_addr_del_pending =
777 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
778 if (asoc->asconf_addr_del_pending == NULL) {
782 asoc->asconf_addr_del_pending->sa.sa_family =
784 asoc->asconf_addr_del_pending->v4.sin_port =
786 if (addrs->sa_family == AF_INET) {
787 struct sockaddr_in *sin;
789 sin = (struct sockaddr_in *)addrs;
790 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
791 } else if (addrs->sa_family == AF_INET6) {
792 struct sockaddr_in6 *sin6;
794 sin6 = (struct sockaddr_in6 *)addrs;
795 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
798 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
799 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
800 asoc->asconf_addr_del_pending);
802 asoc->src_out_of_asoc_ok = 1;
810 /* We do not need RCU protection throughout this loop
811 * because this is done under a socket lock from the
814 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
822 /* Reset use_as_src flag for the addresses in the bind address
823 * list that are to be deleted.
826 for (i = 0; i < addrcnt; i++) {
828 af = sctp_get_af_specific(laddr->v4.sin_family);
829 list_for_each_entry(saddr, &bp->address_list, list) {
830 if (sctp_cmp_addr_exact(&saddr->a, laddr))
831 saddr->state = SCTP_ADDR_DEL;
833 addr_buf += af->sockaddr_len;
836 /* Update the route and saddr entries for all the transports
837 * as some of the addresses in the bind address list are
838 * about to be deleted and cannot be used as source addresses.
840 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
842 dst_release(transport->dst);
843 sctp_transport_route(transport, NULL,
844 sctp_sk(asoc->base.sk));
848 /* We don't need to transmit ASCONF */
850 retval = sctp_send_asconf(asoc, chunk);
856 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
857 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
859 struct sock *sk = sctp_opt2sk(sp);
860 union sctp_addr *addr;
863 /* It is safe to write port space in caller. */
865 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
866 af = sctp_get_af_specific(addr->sa.sa_family);
869 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
872 if (addrw->state == SCTP_ADDR_NEW)
873 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
875 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
878 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
881 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
884 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
885 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
888 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
889 * Section 3.1.2 for this usage.
891 * addrs is a pointer to an array of one or more socket addresses. Each
892 * address is contained in its appropriate structure (i.e. struct
893 * sockaddr_in or struct sockaddr_in6) the family of the address type
894 * must be used to distinguish the address length (note that this
895 * representation is termed a "packed array" of addresses). The caller
896 * specifies the number of addresses in the array with addrcnt.
898 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
899 * -1, and sets errno to the appropriate error code.
901 * For SCTP, the port given in each socket address must be the same, or
902 * sctp_bindx() will fail, setting errno to EINVAL.
904 * The flags parameter is formed from the bitwise OR of zero or more of
905 * the following currently defined flags:
907 * SCTP_BINDX_ADD_ADDR
909 * SCTP_BINDX_REM_ADDR
911 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
912 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
913 * addresses from the association. The two flags are mutually exclusive;
914 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
915 * not remove all addresses from an association; sctp_bindx() will
916 * reject such an attempt with EINVAL.
918 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
919 * additional addresses with an endpoint after calling bind(). Or use
920 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
921 * socket is associated with so that no new association accepted will be
922 * associated with those addresses. If the endpoint supports dynamic
923 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
924 * endpoint to send the appropriate message to the peer to change the
925 * peers address lists.
927 * Adding and removing addresses from a connected association is
928 * optional functionality. Implementations that do not support this
929 * functionality should return EOPNOTSUPP.
931 * Basically do nothing but copying the addresses from user to kernel
932 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
933 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
936 * We don't use copy_from_user() for optimization: we first do the
937 * sanity checks (buffer size -fast- and access check-healthy
938 * pointer); if all of those succeed, then we can alloc the memory
939 * (expensive operation) needed to copy the data to kernel. Then we do
940 * the copying without checking the user space area
941 * (__copy_from_user()).
943 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
946 * sk The sk of the socket
947 * addrs The pointer to the addresses in user land
948 * addrssize Size of the addrs buffer
949 * op Operation to perform (add or remove, see the flags of
952 * Returns 0 if ok, <0 errno code on error.
954 static int sctp_setsockopt_bindx(struct sock *sk,
955 struct sockaddr __user *addrs,
956 int addrs_size, int op)
958 struct sockaddr *kaddrs;
962 struct sockaddr *sa_addr;
966 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
967 __func__, sk, addrs, addrs_size, op);
969 if (unlikely(addrs_size <= 0))
972 /* Check the user passed a healthy pointer. */
973 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
976 /* Alloc space for the address array in kernel memory. */
977 kaddrs = kmalloc(addrs_size, GFP_USER | __GFP_NOWARN);
978 if (unlikely(!kaddrs))
981 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
986 /* Walk through the addrs buffer and count the number of addresses. */
988 while (walk_size < addrs_size) {
989 if (walk_size + sizeof(sa_family_t) > addrs_size) {
995 af = sctp_get_af_specific(sa_addr->sa_family);
997 /* If the address family is not supported or if this address
998 * causes the address buffer to overflow return EINVAL.
1000 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1005 addr_buf += af->sockaddr_len;
1006 walk_size += af->sockaddr_len;
1011 case SCTP_BINDX_ADD_ADDR:
1012 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1015 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1018 case SCTP_BINDX_REM_ADDR:
1019 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1022 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1036 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1038 * Common routine for handling connect() and sctp_connectx().
1039 * Connect will come in with just a single address.
1041 static int __sctp_connect(struct sock *sk,
1042 struct sockaddr *kaddrs,
1044 sctp_assoc_t *assoc_id)
1046 struct net *net = sock_net(sk);
1047 struct sctp_sock *sp;
1048 struct sctp_endpoint *ep;
1049 struct sctp_association *asoc = NULL;
1050 struct sctp_association *asoc2;
1051 struct sctp_transport *transport;
1058 union sctp_addr *sa_addr = NULL;
1060 unsigned short port;
1061 unsigned int f_flags = 0;
1066 /* connect() cannot be done on a socket that is already in ESTABLISHED
1067 * state - UDP-style peeled off socket or a TCP-style socket that
1068 * is already connected.
1069 * It cannot be done even on a TCP-style listening socket.
1071 if (sctp_sstate(sk, ESTABLISHED) || sctp_sstate(sk, CLOSING) ||
1072 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1077 /* Walk through the addrs buffer and count the number of addresses. */
1079 while (walk_size < addrs_size) {
1082 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1088 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1090 /* If the address family is not supported or if this address
1091 * causes the address buffer to overflow return EINVAL.
1093 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1098 port = ntohs(sa_addr->v4.sin_port);
1100 /* Save current address so we can work with it */
1101 memcpy(&to, sa_addr, af->sockaddr_len);
1103 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1107 /* Make sure the destination port is correctly set
1110 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1115 /* Check if there already is a matching association on the
1116 * endpoint (other than the one created here).
1118 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1119 if (asoc2 && asoc2 != asoc) {
1120 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1127 /* If we could not find a matching association on the endpoint,
1128 * make sure that there is no peeled-off association matching
1129 * the peer address even on another socket.
1131 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1132 err = -EADDRNOTAVAIL;
1137 /* If a bind() or sctp_bindx() is not called prior to
1138 * an sctp_connectx() call, the system picks an
1139 * ephemeral port and will choose an address set
1140 * equivalent to binding with a wildcard address.
1142 if (!ep->base.bind_addr.port) {
1143 if (sctp_autobind(sk)) {
1149 * If an unprivileged user inherits a 1-many
1150 * style socket with open associations on a
1151 * privileged port, it MAY be permitted to
1152 * accept new associations, but it SHOULD NOT
1153 * be permitted to open new associations.
1155 if (ep->base.bind_addr.port < PROT_SOCK &&
1156 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1162 scope = sctp_scope(&to);
1163 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1169 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1177 /* Prime the peer's transport structures. */
1178 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1186 addr_buf += af->sockaddr_len;
1187 walk_size += af->sockaddr_len;
1190 /* In case the user of sctp_connectx() wants an association
1191 * id back, assign one now.
1194 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1199 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1204 /* Initialize sk's dport and daddr for getpeername() */
1205 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1206 sp->pf->to_sk_daddr(sa_addr, sk);
1209 /* in-kernel sockets don't generally have a file allocated to them
1210 * if all they do is call sock_create_kern().
1212 if (sk->sk_socket->file)
1213 f_flags = sk->sk_socket->file->f_flags;
1215 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1217 err = sctp_wait_for_connect(asoc, &timeo);
1218 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1219 *assoc_id = asoc->assoc_id;
1221 /* Don't free association on exit. */
1225 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1226 __func__, asoc, kaddrs, err);
1229 /* sctp_primitive_ASSOCIATE may have added this association
1230 * To the hash table, try to unhash it, just in case, its a noop
1231 * if it wasn't hashed so we're safe
1233 sctp_association_free(asoc);
1238 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1241 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1242 * sctp_assoc_t *asoc);
1244 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1245 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1246 * or IPv6 addresses.
1248 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1249 * Section 3.1.2 for this usage.
1251 * addrs is a pointer to an array of one or more socket addresses. Each
1252 * address is contained in its appropriate structure (i.e. struct
1253 * sockaddr_in or struct sockaddr_in6) the family of the address type
1254 * must be used to distengish the address length (note that this
1255 * representation is termed a "packed array" of addresses). The caller
1256 * specifies the number of addresses in the array with addrcnt.
1258 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1259 * the association id of the new association. On failure, sctp_connectx()
1260 * returns -1, and sets errno to the appropriate error code. The assoc_id
1261 * is not touched by the kernel.
1263 * For SCTP, the port given in each socket address must be the same, or
1264 * sctp_connectx() will fail, setting errno to EINVAL.
1266 * An application can use sctp_connectx to initiate an association with
1267 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1268 * allows a caller to specify multiple addresses at which a peer can be
1269 * reached. The way the SCTP stack uses the list of addresses to set up
1270 * the association is implementation dependent. This function only
1271 * specifies that the stack will try to make use of all the addresses in
1272 * the list when needed.
1274 * Note that the list of addresses passed in is only used for setting up
1275 * the association. It does not necessarily equal the set of addresses
1276 * the peer uses for the resulting association. If the caller wants to
1277 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1278 * retrieve them after the association has been set up.
1280 * Basically do nothing but copying the addresses from user to kernel
1281 * land and invoking either sctp_connectx(). This is used for tunneling
1282 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1284 * We don't use copy_from_user() for optimization: we first do the
1285 * sanity checks (buffer size -fast- and access check-healthy
1286 * pointer); if all of those succeed, then we can alloc the memory
1287 * (expensive operation) needed to copy the data to kernel. Then we do
1288 * the copying without checking the user space area
1289 * (__copy_from_user()).
1291 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1294 * sk The sk of the socket
1295 * addrs The pointer to the addresses in user land
1296 * addrssize Size of the addrs buffer
1298 * Returns >=0 if ok, <0 errno code on error.
1300 static int __sctp_setsockopt_connectx(struct sock *sk,
1301 struct sockaddr __user *addrs,
1303 sctp_assoc_t *assoc_id)
1305 struct sockaddr *kaddrs;
1306 gfp_t gfp = GFP_KERNEL;
1309 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1310 __func__, sk, addrs, addrs_size);
1312 if (unlikely(addrs_size <= 0))
1315 /* Check the user passed a healthy pointer. */
1316 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1319 /* Alloc space for the address array in kernel memory. */
1320 if (sk->sk_socket->file)
1321 gfp = GFP_USER | __GFP_NOWARN;
1322 kaddrs = kmalloc(addrs_size, gfp);
1323 if (unlikely(!kaddrs))
1326 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1329 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1338 * This is an older interface. It's kept for backward compatibility
1339 * to the option that doesn't provide association id.
1341 static int sctp_setsockopt_connectx_old(struct sock *sk,
1342 struct sockaddr __user *addrs,
1345 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1349 * New interface for the API. The since the API is done with a socket
1350 * option, to make it simple we feed back the association id is as a return
1351 * indication to the call. Error is always negative and association id is
1354 static int sctp_setsockopt_connectx(struct sock *sk,
1355 struct sockaddr __user *addrs,
1358 sctp_assoc_t assoc_id = 0;
1361 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1370 * New (hopefully final) interface for the API.
1371 * We use the sctp_getaddrs_old structure so that use-space library
1372 * can avoid any unnecessary allocations. The only different part
1373 * is that we store the actual length of the address buffer into the
1374 * addrs_num structure member. That way we can re-use the existing
1377 #ifdef CONFIG_COMPAT
1378 struct compat_sctp_getaddrs_old {
1379 sctp_assoc_t assoc_id;
1381 compat_uptr_t addrs; /* struct sockaddr * */
1385 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1386 char __user *optval,
1389 struct sctp_getaddrs_old param;
1390 sctp_assoc_t assoc_id = 0;
1393 #ifdef CONFIG_COMPAT
1394 if (in_compat_syscall()) {
1395 struct compat_sctp_getaddrs_old param32;
1397 if (len < sizeof(param32))
1399 if (copy_from_user(¶m32, optval, sizeof(param32)))
1402 param.assoc_id = param32.assoc_id;
1403 param.addr_num = param32.addr_num;
1404 param.addrs = compat_ptr(param32.addrs);
1408 if (len < sizeof(param))
1410 if (copy_from_user(¶m, optval, sizeof(param)))
1414 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1415 param.addrs, param.addr_num,
1417 if (err == 0 || err == -EINPROGRESS) {
1418 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1420 if (put_user(sizeof(assoc_id), optlen))
1427 /* API 3.1.4 close() - UDP Style Syntax
1428 * Applications use close() to perform graceful shutdown (as described in
1429 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1430 * by a UDP-style socket.
1434 * ret = close(int sd);
1436 * sd - the socket descriptor of the associations to be closed.
1438 * To gracefully shutdown a specific association represented by the
1439 * UDP-style socket, an application should use the sendmsg() call,
1440 * passing no user data, but including the appropriate flag in the
1441 * ancillary data (see Section xxxx).
1443 * If sd in the close() call is a branched-off socket representing only
1444 * one association, the shutdown is performed on that association only.
1446 * 4.1.6 close() - TCP Style Syntax
1448 * Applications use close() to gracefully close down an association.
1452 * int close(int sd);
1454 * sd - the socket descriptor of the association to be closed.
1456 * After an application calls close() on a socket descriptor, no further
1457 * socket operations will succeed on that descriptor.
1459 * API 7.1.4 SO_LINGER
1461 * An application using the TCP-style socket can use this option to
1462 * perform the SCTP ABORT primitive. The linger option structure is:
1465 * int l_onoff; // option on/off
1466 * int l_linger; // linger time
1469 * To enable the option, set l_onoff to 1. If the l_linger value is set
1470 * to 0, calling close() is the same as the ABORT primitive. If the
1471 * value is set to a negative value, the setsockopt() call will return
1472 * an error. If the value is set to a positive value linger_time, the
1473 * close() can be blocked for at most linger_time ms. If the graceful
1474 * shutdown phase does not finish during this period, close() will
1475 * return but the graceful shutdown phase continues in the system.
1477 static void sctp_close(struct sock *sk, long timeout)
1479 struct net *net = sock_net(sk);
1480 struct sctp_endpoint *ep;
1481 struct sctp_association *asoc;
1482 struct list_head *pos, *temp;
1483 unsigned int data_was_unread;
1485 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1488 sk->sk_shutdown = SHUTDOWN_MASK;
1489 sk->sk_state = SCTP_SS_CLOSING;
1491 ep = sctp_sk(sk)->ep;
1493 /* Clean up any skbs sitting on the receive queue. */
1494 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1495 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1497 /* Walk all associations on an endpoint. */
1498 list_for_each_safe(pos, temp, &ep->asocs) {
1499 asoc = list_entry(pos, struct sctp_association, asocs);
1501 if (sctp_style(sk, TCP)) {
1502 /* A closed association can still be in the list if
1503 * it belongs to a TCP-style listening socket that is
1504 * not yet accepted. If so, free it. If not, send an
1505 * ABORT or SHUTDOWN based on the linger options.
1507 if (sctp_state(asoc, CLOSED)) {
1508 sctp_association_free(asoc);
1513 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1514 !skb_queue_empty(&asoc->ulpq.reasm) ||
1515 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1516 struct sctp_chunk *chunk;
1518 chunk = sctp_make_abort_user(asoc, NULL, 0);
1519 sctp_primitive_ABORT(net, asoc, chunk);
1521 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1524 /* On a TCP-style socket, block for at most linger_time if set. */
1525 if (sctp_style(sk, TCP) && timeout)
1526 sctp_wait_for_close(sk, timeout);
1528 /* This will run the backlog queue. */
1531 /* Supposedly, no process has access to the socket, but
1532 * the net layers still may.
1533 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1534 * held and that should be grabbed before socket lock.
1536 spin_lock_bh(&net->sctp.addr_wq_lock);
1539 /* Hold the sock, since sk_common_release() will put sock_put()
1540 * and we have just a little more cleanup.
1543 sk_common_release(sk);
1546 spin_unlock_bh(&net->sctp.addr_wq_lock);
1550 SCTP_DBG_OBJCNT_DEC(sock);
1553 /* Handle EPIPE error. */
1554 static int sctp_error(struct sock *sk, int flags, int err)
1557 err = sock_error(sk) ? : -EPIPE;
1558 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1559 send_sig(SIGPIPE, current, 0);
1563 /* API 3.1.3 sendmsg() - UDP Style Syntax
1565 * An application uses sendmsg() and recvmsg() calls to transmit data to
1566 * and receive data from its peer.
1568 * ssize_t sendmsg(int socket, const struct msghdr *message,
1571 * socket - the socket descriptor of the endpoint.
1572 * message - pointer to the msghdr structure which contains a single
1573 * user message and possibly some ancillary data.
1575 * See Section 5 for complete description of the data
1578 * flags - flags sent or received with the user message, see Section
1579 * 5 for complete description of the flags.
1581 * Note: This function could use a rewrite especially when explicit
1582 * connect support comes in.
1584 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1586 static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1588 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
1590 struct net *net = sock_net(sk);
1591 struct sctp_sock *sp;
1592 struct sctp_endpoint *ep;
1593 struct sctp_association *new_asoc = NULL, *asoc = NULL;
1594 struct sctp_transport *transport, *chunk_tp;
1595 struct sctp_chunk *chunk;
1597 struct sockaddr *msg_name = NULL;
1598 struct sctp_sndrcvinfo default_sinfo;
1599 struct sctp_sndrcvinfo *sinfo;
1600 struct sctp_initmsg *sinit;
1601 sctp_assoc_t associd = 0;
1602 sctp_cmsgs_t cmsgs = { NULL };
1604 bool fill_sinfo_ttl = false, wait_connect = false;
1605 struct sctp_datamsg *datamsg;
1606 int msg_flags = msg->msg_flags;
1607 __u16 sinfo_flags = 0;
1615 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1618 /* We cannot send a message over a TCP-style listening socket. */
1619 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1624 /* Parse out the SCTP CMSGs. */
1625 err = sctp_msghdr_parse(msg, &cmsgs);
1627 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1631 /* Fetch the destination address for this packet. This
1632 * address only selects the association--it is not necessarily
1633 * the address we will send to.
1634 * For a peeled-off socket, msg_name is ignored.
1636 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1637 int msg_namelen = msg->msg_namelen;
1639 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1644 if (msg_namelen > sizeof(to))
1645 msg_namelen = sizeof(to);
1646 memcpy(&to, msg->msg_name, msg_namelen);
1647 msg_name = msg->msg_name;
1651 if (cmsgs.sinfo != NULL) {
1652 memset(&default_sinfo, 0, sizeof(default_sinfo));
1653 default_sinfo.sinfo_stream = cmsgs.sinfo->snd_sid;
1654 default_sinfo.sinfo_flags = cmsgs.sinfo->snd_flags;
1655 default_sinfo.sinfo_ppid = cmsgs.sinfo->snd_ppid;
1656 default_sinfo.sinfo_context = cmsgs.sinfo->snd_context;
1657 default_sinfo.sinfo_assoc_id = cmsgs.sinfo->snd_assoc_id;
1659 sinfo = &default_sinfo;
1660 fill_sinfo_ttl = true;
1662 sinfo = cmsgs.srinfo;
1664 /* Did the user specify SNDINFO/SNDRCVINFO? */
1666 sinfo_flags = sinfo->sinfo_flags;
1667 associd = sinfo->sinfo_assoc_id;
1670 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1671 msg_len, sinfo_flags);
1673 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1674 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1679 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1680 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1681 * If SCTP_ABORT is set, the message length could be non zero with
1682 * the msg_iov set to the user abort reason.
1684 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1685 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1690 /* If SCTP_ADDR_OVER is set, there must be an address
1691 * specified in msg_name.
1693 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1700 pr_debug("%s: about to look up association\n", __func__);
1704 /* If a msg_name has been specified, assume this is to be used. */
1706 /* Look for a matching association on the endpoint. */
1707 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1709 /* If we could not find a matching association on the
1710 * endpoint, make sure that it is not a TCP-style
1711 * socket that already has an association or there is
1712 * no peeled-off association on another socket.
1715 ((sctp_style(sk, TCP) &&
1716 (sctp_sstate(sk, ESTABLISHED) ||
1717 sctp_sstate(sk, CLOSING))) ||
1718 sctp_endpoint_is_peeled_off(ep, &to))) {
1719 err = -EADDRNOTAVAIL;
1723 asoc = sctp_id2assoc(sk, associd);
1731 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1733 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1734 * socket that has an association in CLOSED state. This can
1735 * happen when an accepted socket has an association that is
1738 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1743 if (sinfo_flags & SCTP_EOF) {
1744 pr_debug("%s: shutting down association:%p\n",
1747 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1751 if (sinfo_flags & SCTP_ABORT) {
1753 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1759 pr_debug("%s: aborting association:%p\n",
1762 sctp_primitive_ABORT(net, asoc, chunk);
1768 /* Do we need to create the association? */
1770 pr_debug("%s: there is no association yet\n", __func__);
1772 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1777 /* Check for invalid stream against the stream counts,
1778 * either the default or the user specified stream counts.
1781 if (!sinit || !sinit->sinit_num_ostreams) {
1782 /* Check against the defaults. */
1783 if (sinfo->sinfo_stream >=
1784 sp->initmsg.sinit_num_ostreams) {
1789 /* Check against the requested. */
1790 if (sinfo->sinfo_stream >=
1791 sinit->sinit_num_ostreams) {
1799 * API 3.1.2 bind() - UDP Style Syntax
1800 * If a bind() or sctp_bindx() is not called prior to a
1801 * sendmsg() call that initiates a new association, the
1802 * system picks an ephemeral port and will choose an address
1803 * set equivalent to binding with a wildcard address.
1805 if (!ep->base.bind_addr.port) {
1806 if (sctp_autobind(sk)) {
1812 * If an unprivileged user inherits a one-to-many
1813 * style socket with open associations on a privileged
1814 * port, it MAY be permitted to accept new associations,
1815 * but it SHOULD NOT be permitted to open new
1818 if (ep->base.bind_addr.port < PROT_SOCK &&
1819 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1825 scope = sctp_scope(&to);
1826 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1832 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1838 /* If the SCTP_INIT ancillary data is specified, set all
1839 * the association init values accordingly.
1842 if (sinit->sinit_num_ostreams) {
1843 asoc->c.sinit_num_ostreams =
1844 sinit->sinit_num_ostreams;
1846 if (sinit->sinit_max_instreams) {
1847 asoc->c.sinit_max_instreams =
1848 sinit->sinit_max_instreams;
1850 if (sinit->sinit_max_attempts) {
1851 asoc->max_init_attempts
1852 = sinit->sinit_max_attempts;
1854 if (sinit->sinit_max_init_timeo) {
1855 asoc->max_init_timeo =
1856 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1860 /* Prime the peer's transport structures. */
1861 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1868 /* ASSERT: we have a valid association at this point. */
1869 pr_debug("%s: we have a valid association\n", __func__);
1872 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1873 * one with some defaults.
1875 memset(&default_sinfo, 0, sizeof(default_sinfo));
1876 default_sinfo.sinfo_stream = asoc->default_stream;
1877 default_sinfo.sinfo_flags = asoc->default_flags;
1878 default_sinfo.sinfo_ppid = asoc->default_ppid;
1879 default_sinfo.sinfo_context = asoc->default_context;
1880 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1881 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1883 sinfo = &default_sinfo;
1884 } else if (fill_sinfo_ttl) {
1885 /* In case SNDINFO was specified, we still need to fill
1886 * it with a default ttl from the assoc here.
1888 sinfo->sinfo_timetolive = asoc->default_timetolive;
1891 /* API 7.1.7, the sndbuf size per association bounds the
1892 * maximum size of data that can be sent in a single send call.
1894 if (msg_len > sk->sk_sndbuf) {
1899 if (asoc->pmtu_pending)
1900 sctp_assoc_pending_pmtu(sk, asoc);
1902 /* If fragmentation is disabled and the message length exceeds the
1903 * association fragmentation point, return EMSGSIZE. The I-D
1904 * does not specify what this error is, but this looks like
1907 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1912 /* Check for invalid stream. */
1913 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1918 if (sctp_wspace(asoc) < msg_len)
1919 sctp_prsctp_prune(asoc, sinfo, msg_len - sctp_wspace(asoc));
1921 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1922 if (!sctp_wspace(asoc)) {
1923 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1928 /* If an address is passed with the sendto/sendmsg call, it is used
1929 * to override the primary destination address in the TCP model, or
1930 * when SCTP_ADDR_OVER flag is set in the UDP model.
1932 if ((sctp_style(sk, TCP) && msg_name) ||
1933 (sinfo_flags & SCTP_ADDR_OVER)) {
1934 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1942 /* Auto-connect, if we aren't connected already. */
1943 if (sctp_state(asoc, CLOSED)) {
1944 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1948 wait_connect = true;
1949 pr_debug("%s: we associated primitively\n", __func__);
1952 /* Break the message into multiple chunks of maximum size. */
1953 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
1954 if (IS_ERR(datamsg)) {
1955 err = PTR_ERR(datamsg);
1959 /* Now send the (possibly) fragmented message. */
1960 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1961 sctp_chunk_hold(chunk);
1963 /* Do accounting for the write space. */
1964 sctp_set_owner_w(chunk);
1966 chunk->transport = chunk_tp;
1969 /* Send it to the lower layers. Note: all chunks
1970 * must either fail or succeed. The lower layer
1971 * works that way today. Keep it that way or this
1974 err = sctp_primitive_SEND(net, asoc, datamsg);
1975 /* Did the lower layer accept the chunk? */
1977 sctp_datamsg_free(datamsg);
1981 pr_debug("%s: we sent primitively\n", __func__);
1983 sctp_datamsg_put(datamsg);
1986 if (unlikely(wait_connect)) {
1987 timeo = sock_sndtimeo(sk, msg_flags & MSG_DONTWAIT);
1988 sctp_wait_for_connect(asoc, &timeo);
1991 /* If we are already past ASSOCIATE, the lower
1992 * layers are responsible for association cleanup.
1998 sctp_association_free(asoc);
2003 return sctp_error(sk, msg_flags, err);
2010 err = sock_error(sk);
2020 /* This is an extended version of skb_pull() that removes the data from the
2021 * start of a skb even when data is spread across the list of skb's in the
2022 * frag_list. len specifies the total amount of data that needs to be removed.
2023 * when 'len' bytes could be removed from the skb, it returns 0.
2024 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2025 * could not be removed.
2027 static int sctp_skb_pull(struct sk_buff *skb, int len)
2029 struct sk_buff *list;
2030 int skb_len = skb_headlen(skb);
2033 if (len <= skb_len) {
2034 __skb_pull(skb, len);
2038 __skb_pull(skb, skb_len);
2040 skb_walk_frags(skb, list) {
2041 rlen = sctp_skb_pull(list, len);
2042 skb->len -= (len-rlen);
2043 skb->data_len -= (len-rlen);
2054 /* API 3.1.3 recvmsg() - UDP Style Syntax
2056 * ssize_t recvmsg(int socket, struct msghdr *message,
2059 * socket - the socket descriptor of the endpoint.
2060 * message - pointer to the msghdr structure which contains a single
2061 * user message and possibly some ancillary data.
2063 * See Section 5 for complete description of the data
2066 * flags - flags sent or received with the user message, see Section
2067 * 5 for complete description of the flags.
2069 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2070 int noblock, int flags, int *addr_len)
2072 struct sctp_ulpevent *event = NULL;
2073 struct sctp_sock *sp = sctp_sk(sk);
2074 struct sk_buff *skb, *head_skb;
2079 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2080 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2085 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED) &&
2086 !sctp_sstate(sk, CLOSING) && !sctp_sstate(sk, CLOSED)) {
2091 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2095 /* Get the total length of the skb including any skb's in the
2104 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2106 event = sctp_skb2event(skb);
2111 if (event->chunk && event->chunk->head_skb)
2112 head_skb = event->chunk->head_skb;
2115 sock_recv_ts_and_drops(msg, sk, head_skb);
2116 if (sctp_ulpevent_is_notification(event)) {
2117 msg->msg_flags |= MSG_NOTIFICATION;
2118 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2120 sp->pf->skb_msgname(head_skb, msg->msg_name, addr_len);
2123 /* Check if we allow SCTP_NXTINFO. */
2124 if (sp->recvnxtinfo)
2125 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2126 /* Check if we allow SCTP_RCVINFO. */
2127 if (sp->recvrcvinfo)
2128 sctp_ulpevent_read_rcvinfo(event, msg);
2129 /* Check if we allow SCTP_SNDRCVINFO. */
2130 if (sp->subscribe.sctp_data_io_event)
2131 sctp_ulpevent_read_sndrcvinfo(event, msg);
2135 /* If skb's length exceeds the user's buffer, update the skb and
2136 * push it back to the receive_queue so that the next call to
2137 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2139 if (skb_len > copied) {
2140 msg->msg_flags &= ~MSG_EOR;
2141 if (flags & MSG_PEEK)
2143 sctp_skb_pull(skb, copied);
2144 skb_queue_head(&sk->sk_receive_queue, skb);
2146 /* When only partial message is copied to the user, increase
2147 * rwnd by that amount. If all the data in the skb is read,
2148 * rwnd is updated when the event is freed.
2150 if (!sctp_ulpevent_is_notification(event))
2151 sctp_assoc_rwnd_increase(event->asoc, copied);
2153 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2154 (event->msg_flags & MSG_EOR))
2155 msg->msg_flags |= MSG_EOR;
2157 msg->msg_flags &= ~MSG_EOR;
2160 if (flags & MSG_PEEK) {
2161 /* Release the skb reference acquired after peeking the skb in
2162 * sctp_skb_recv_datagram().
2166 /* Free the event which includes releasing the reference to
2167 * the owner of the skb, freeing the skb and updating the
2170 sctp_ulpevent_free(event);
2177 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2179 * This option is a on/off flag. If enabled no SCTP message
2180 * fragmentation will be performed. Instead if a message being sent
2181 * exceeds the current PMTU size, the message will NOT be sent and
2182 * instead a error will be indicated to the user.
2184 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2185 char __user *optval,
2186 unsigned int optlen)
2190 if (optlen < sizeof(int))
2193 if (get_user(val, (int __user *)optval))
2196 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2201 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2202 unsigned int optlen)
2204 struct sctp_association *asoc;
2205 struct sctp_ulpevent *event;
2207 if (optlen > sizeof(struct sctp_event_subscribe))
2209 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2212 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2213 * if there is no data to be sent or retransmit, the stack will
2214 * immediately send up this notification.
2216 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2217 &sctp_sk(sk)->subscribe)) {
2218 asoc = sctp_id2assoc(sk, 0);
2220 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2221 event = sctp_ulpevent_make_sender_dry_event(asoc,
2226 sctp_ulpq_tail_event(&asoc->ulpq, event);
2233 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2235 * This socket option is applicable to the UDP-style socket only. When
2236 * set it will cause associations that are idle for more than the
2237 * specified number of seconds to automatically close. An association
2238 * being idle is defined an association that has NOT sent or received
2239 * user data. The special value of '0' indicates that no automatic
2240 * close of any associations should be performed. The option expects an
2241 * integer defining the number of seconds of idle time before an
2242 * association is closed.
2244 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2245 unsigned int optlen)
2247 struct sctp_sock *sp = sctp_sk(sk);
2248 struct net *net = sock_net(sk);
2250 /* Applicable to UDP-style socket only */
2251 if (sctp_style(sk, TCP))
2253 if (optlen != sizeof(int))
2255 if (copy_from_user(&sp->autoclose, optval, optlen))
2258 if (sp->autoclose > net->sctp.max_autoclose)
2259 sp->autoclose = net->sctp.max_autoclose;
2264 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2266 * Applications can enable or disable heartbeats for any peer address of
2267 * an association, modify an address's heartbeat interval, force a
2268 * heartbeat to be sent immediately, and adjust the address's maximum
2269 * number of retransmissions sent before an address is considered
2270 * unreachable. The following structure is used to access and modify an
2271 * address's parameters:
2273 * struct sctp_paddrparams {
2274 * sctp_assoc_t spp_assoc_id;
2275 * struct sockaddr_storage spp_address;
2276 * uint32_t spp_hbinterval;
2277 * uint16_t spp_pathmaxrxt;
2278 * uint32_t spp_pathmtu;
2279 * uint32_t spp_sackdelay;
2280 * uint32_t spp_flags;
2283 * spp_assoc_id - (one-to-many style socket) This is filled in the
2284 * application, and identifies the association for
2286 * spp_address - This specifies which address is of interest.
2287 * spp_hbinterval - This contains the value of the heartbeat interval,
2288 * in milliseconds. If a value of zero
2289 * is present in this field then no changes are to
2290 * be made to this parameter.
2291 * spp_pathmaxrxt - This contains the maximum number of
2292 * retransmissions before this address shall be
2293 * considered unreachable. If a value of zero
2294 * is present in this field then no changes are to
2295 * be made to this parameter.
2296 * spp_pathmtu - When Path MTU discovery is disabled the value
2297 * specified here will be the "fixed" path mtu.
2298 * Note that if the spp_address field is empty
2299 * then all associations on this address will
2300 * have this fixed path mtu set upon them.
2302 * spp_sackdelay - When delayed sack is enabled, this value specifies
2303 * the number of milliseconds that sacks will be delayed
2304 * for. This value will apply to all addresses of an
2305 * association if the spp_address field is empty. Note
2306 * also, that if delayed sack is enabled and this
2307 * value is set to 0, no change is made to the last
2308 * recorded delayed sack timer value.
2310 * spp_flags - These flags are used to control various features
2311 * on an association. The flag field may contain
2312 * zero or more of the following options.
2314 * SPP_HB_ENABLE - Enable heartbeats on the
2315 * specified address. Note that if the address
2316 * field is empty all addresses for the association
2317 * have heartbeats enabled upon them.
2319 * SPP_HB_DISABLE - Disable heartbeats on the
2320 * speicifed address. Note that if the address
2321 * field is empty all addresses for the association
2322 * will have their heartbeats disabled. Note also
2323 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2324 * mutually exclusive, only one of these two should
2325 * be specified. Enabling both fields will have
2326 * undetermined results.
2328 * SPP_HB_DEMAND - Request a user initiated heartbeat
2329 * to be made immediately.
2331 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2332 * heartbeat delayis to be set to the value of 0
2335 * SPP_PMTUD_ENABLE - This field will enable PMTU
2336 * discovery upon the specified address. Note that
2337 * if the address feild is empty then all addresses
2338 * on the association are effected.
2340 * SPP_PMTUD_DISABLE - This field will disable PMTU
2341 * discovery upon the specified address. Note that
2342 * if the address feild is empty then all addresses
2343 * on the association are effected. Not also that
2344 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2345 * exclusive. Enabling both will have undetermined
2348 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2349 * on delayed sack. The time specified in spp_sackdelay
2350 * is used to specify the sack delay for this address. Note
2351 * that if spp_address is empty then all addresses will
2352 * enable delayed sack and take on the sack delay
2353 * value specified in spp_sackdelay.
2354 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2355 * off delayed sack. If the spp_address field is blank then
2356 * delayed sack is disabled for the entire association. Note
2357 * also that this field is mutually exclusive to
2358 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2361 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2362 struct sctp_transport *trans,
2363 struct sctp_association *asoc,
2364 struct sctp_sock *sp,
2367 int sackdelay_change)
2371 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2372 struct net *net = sock_net(trans->asoc->base.sk);
2374 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2379 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2380 * this field is ignored. Note also that a value of zero indicates
2381 * the current setting should be left unchanged.
2383 if (params->spp_flags & SPP_HB_ENABLE) {
2385 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2386 * set. This lets us use 0 value when this flag
2389 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2390 params->spp_hbinterval = 0;
2392 if (params->spp_hbinterval ||
2393 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2396 msecs_to_jiffies(params->spp_hbinterval);
2399 msecs_to_jiffies(params->spp_hbinterval);
2401 sp->hbinterval = params->spp_hbinterval;
2408 trans->param_flags =
2409 (trans->param_flags & ~SPP_HB) | hb_change;
2412 (asoc->param_flags & ~SPP_HB) | hb_change;
2415 (sp->param_flags & ~SPP_HB) | hb_change;
2419 /* When Path MTU discovery is disabled the value specified here will
2420 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2421 * include the flag SPP_PMTUD_DISABLE for this field to have any
2424 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2426 trans->pathmtu = params->spp_pathmtu;
2427 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2429 asoc->pathmtu = params->spp_pathmtu;
2430 sctp_frag_point(asoc, params->spp_pathmtu);
2432 sp->pathmtu = params->spp_pathmtu;
2438 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2439 (params->spp_flags & SPP_PMTUD_ENABLE);
2440 trans->param_flags =
2441 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2443 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2444 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2448 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2451 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2455 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2456 * value of this field is ignored. Note also that a value of zero
2457 * indicates the current setting should be left unchanged.
2459 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2462 msecs_to_jiffies(params->spp_sackdelay);
2465 msecs_to_jiffies(params->spp_sackdelay);
2467 sp->sackdelay = params->spp_sackdelay;
2471 if (sackdelay_change) {
2473 trans->param_flags =
2474 (trans->param_flags & ~SPP_SACKDELAY) |
2478 (asoc->param_flags & ~SPP_SACKDELAY) |
2482 (sp->param_flags & ~SPP_SACKDELAY) |
2487 /* Note that a value of zero indicates the current setting should be
2490 if (params->spp_pathmaxrxt) {
2492 trans->pathmaxrxt = params->spp_pathmaxrxt;
2494 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2496 sp->pathmaxrxt = params->spp_pathmaxrxt;
2503 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2504 char __user *optval,
2505 unsigned int optlen)
2507 struct sctp_paddrparams params;
2508 struct sctp_transport *trans = NULL;
2509 struct sctp_association *asoc = NULL;
2510 struct sctp_sock *sp = sctp_sk(sk);
2512 int hb_change, pmtud_change, sackdelay_change;
2514 if (optlen != sizeof(struct sctp_paddrparams))
2517 if (copy_from_user(¶ms, optval, optlen))
2520 /* Validate flags and value parameters. */
2521 hb_change = params.spp_flags & SPP_HB;
2522 pmtud_change = params.spp_flags & SPP_PMTUD;
2523 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2525 if (hb_change == SPP_HB ||
2526 pmtud_change == SPP_PMTUD ||
2527 sackdelay_change == SPP_SACKDELAY ||
2528 params.spp_sackdelay > 500 ||
2529 (params.spp_pathmtu &&
2530 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2533 /* If an address other than INADDR_ANY is specified, and
2534 * no transport is found, then the request is invalid.
2536 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
2537 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2538 params.spp_assoc_id);
2543 /* Get association, if assoc_id != 0 and the socket is a one
2544 * to many style socket, and an association was not found, then
2545 * the id was invalid.
2547 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2548 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2551 /* Heartbeat demand can only be sent on a transport or
2552 * association, but not a socket.
2554 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2557 /* Process parameters. */
2558 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2559 hb_change, pmtud_change,
2565 /* If changes are for association, also apply parameters to each
2568 if (!trans && asoc) {
2569 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2571 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2572 hb_change, pmtud_change,
2580 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2582 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2585 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2587 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2591 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2593 * This option will effect the way delayed acks are performed. This
2594 * option allows you to get or set the delayed ack time, in
2595 * milliseconds. It also allows changing the delayed ack frequency.
2596 * Changing the frequency to 1 disables the delayed sack algorithm. If
2597 * the assoc_id is 0, then this sets or gets the endpoints default
2598 * values. If the assoc_id field is non-zero, then the set or get
2599 * effects the specified association for the one to many model (the
2600 * assoc_id field is ignored by the one to one model). Note that if
2601 * sack_delay or sack_freq are 0 when setting this option, then the
2602 * current values will remain unchanged.
2604 * struct sctp_sack_info {
2605 * sctp_assoc_t sack_assoc_id;
2606 * uint32_t sack_delay;
2607 * uint32_t sack_freq;
2610 * sack_assoc_id - This parameter, indicates which association the user
2611 * is performing an action upon. Note that if this field's value is
2612 * zero then the endpoints default value is changed (effecting future
2613 * associations only).
2615 * sack_delay - This parameter contains the number of milliseconds that
2616 * the user is requesting the delayed ACK timer be set to. Note that
2617 * this value is defined in the standard to be between 200 and 500
2620 * sack_freq - This parameter contains the number of packets that must
2621 * be received before a sack is sent without waiting for the delay
2622 * timer to expire. The default value for this is 2, setting this
2623 * value to 1 will disable the delayed sack algorithm.
2626 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2627 char __user *optval, unsigned int optlen)
2629 struct sctp_sack_info params;
2630 struct sctp_transport *trans = NULL;
2631 struct sctp_association *asoc = NULL;
2632 struct sctp_sock *sp = sctp_sk(sk);
2634 if (optlen == sizeof(struct sctp_sack_info)) {
2635 if (copy_from_user(¶ms, optval, optlen))
2638 if (params.sack_delay == 0 && params.sack_freq == 0)
2640 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2641 pr_warn_ratelimited(DEPRECATED
2643 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2644 "Use struct sctp_sack_info instead\n",
2645 current->comm, task_pid_nr(current));
2646 if (copy_from_user(¶ms, optval, optlen))
2649 if (params.sack_delay == 0)
2650 params.sack_freq = 1;
2652 params.sack_freq = 0;
2656 /* Validate value parameter. */
2657 if (params.sack_delay > 500)
2660 /* Get association, if sack_assoc_id != 0 and the socket is a one
2661 * to many style socket, and an association was not found, then
2662 * the id was invalid.
2664 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2665 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2668 if (params.sack_delay) {
2671 msecs_to_jiffies(params.sack_delay);
2673 sctp_spp_sackdelay_enable(asoc->param_flags);
2675 sp->sackdelay = params.sack_delay;
2677 sctp_spp_sackdelay_enable(sp->param_flags);
2681 if (params.sack_freq == 1) {
2684 sctp_spp_sackdelay_disable(asoc->param_flags);
2687 sctp_spp_sackdelay_disable(sp->param_flags);
2689 } else if (params.sack_freq > 1) {
2691 asoc->sackfreq = params.sack_freq;
2693 sctp_spp_sackdelay_enable(asoc->param_flags);
2695 sp->sackfreq = params.sack_freq;
2697 sctp_spp_sackdelay_enable(sp->param_flags);
2701 /* If change is for association, also apply to each transport. */
2703 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2705 if (params.sack_delay) {
2707 msecs_to_jiffies(params.sack_delay);
2708 trans->param_flags =
2709 sctp_spp_sackdelay_enable(trans->param_flags);
2711 if (params.sack_freq == 1) {
2712 trans->param_flags =
2713 sctp_spp_sackdelay_disable(trans->param_flags);
2714 } else if (params.sack_freq > 1) {
2715 trans->sackfreq = params.sack_freq;
2716 trans->param_flags =
2717 sctp_spp_sackdelay_enable(trans->param_flags);
2725 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2727 * Applications can specify protocol parameters for the default association
2728 * initialization. The option name argument to setsockopt() and getsockopt()
2731 * Setting initialization parameters is effective only on an unconnected
2732 * socket (for UDP-style sockets only future associations are effected
2733 * by the change). With TCP-style sockets, this option is inherited by
2734 * sockets derived from a listener socket.
2736 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2738 struct sctp_initmsg sinit;
2739 struct sctp_sock *sp = sctp_sk(sk);
2741 if (optlen != sizeof(struct sctp_initmsg))
2743 if (copy_from_user(&sinit, optval, optlen))
2746 if (sinit.sinit_num_ostreams)
2747 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2748 if (sinit.sinit_max_instreams)
2749 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2750 if (sinit.sinit_max_attempts)
2751 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2752 if (sinit.sinit_max_init_timeo)
2753 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2759 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2761 * Applications that wish to use the sendto() system call may wish to
2762 * specify a default set of parameters that would normally be supplied
2763 * through the inclusion of ancillary data. This socket option allows
2764 * such an application to set the default sctp_sndrcvinfo structure.
2765 * The application that wishes to use this socket option simply passes
2766 * in to this call the sctp_sndrcvinfo structure defined in Section
2767 * 5.2.2) The input parameters accepted by this call include
2768 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2769 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2770 * to this call if the caller is using the UDP model.
2772 static int sctp_setsockopt_default_send_param(struct sock *sk,
2773 char __user *optval,
2774 unsigned int optlen)
2776 struct sctp_sock *sp = sctp_sk(sk);
2777 struct sctp_association *asoc;
2778 struct sctp_sndrcvinfo info;
2780 if (optlen != sizeof(info))
2782 if (copy_from_user(&info, optval, optlen))
2784 if (info.sinfo_flags &
2785 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2786 SCTP_ABORT | SCTP_EOF))
2789 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2790 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2793 asoc->default_stream = info.sinfo_stream;
2794 asoc->default_flags = info.sinfo_flags;
2795 asoc->default_ppid = info.sinfo_ppid;
2796 asoc->default_context = info.sinfo_context;
2797 asoc->default_timetolive = info.sinfo_timetolive;
2799 sp->default_stream = info.sinfo_stream;
2800 sp->default_flags = info.sinfo_flags;
2801 sp->default_ppid = info.sinfo_ppid;
2802 sp->default_context = info.sinfo_context;
2803 sp->default_timetolive = info.sinfo_timetolive;
2809 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2810 * (SCTP_DEFAULT_SNDINFO)
2812 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
2813 char __user *optval,
2814 unsigned int optlen)
2816 struct sctp_sock *sp = sctp_sk(sk);
2817 struct sctp_association *asoc;
2818 struct sctp_sndinfo info;
2820 if (optlen != sizeof(info))
2822 if (copy_from_user(&info, optval, optlen))
2824 if (info.snd_flags &
2825 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2826 SCTP_ABORT | SCTP_EOF))
2829 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
2830 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
2833 asoc->default_stream = info.snd_sid;
2834 asoc->default_flags = info.snd_flags;
2835 asoc->default_ppid = info.snd_ppid;
2836 asoc->default_context = info.snd_context;
2838 sp->default_stream = info.snd_sid;
2839 sp->default_flags = info.snd_flags;
2840 sp->default_ppid = info.snd_ppid;
2841 sp->default_context = info.snd_context;
2847 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2849 * Requests that the local SCTP stack use the enclosed peer address as
2850 * the association primary. The enclosed address must be one of the
2851 * association peer's addresses.
2853 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2854 unsigned int optlen)
2856 struct sctp_prim prim;
2857 struct sctp_transport *trans;
2859 if (optlen != sizeof(struct sctp_prim))
2862 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2865 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2869 sctp_assoc_set_primary(trans->asoc, trans);
2875 * 7.1.5 SCTP_NODELAY
2877 * Turn on/off any Nagle-like algorithm. This means that packets are
2878 * generally sent as soon as possible and no unnecessary delays are
2879 * introduced, at the cost of more packets in the network. Expects an
2880 * integer boolean flag.
2882 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2883 unsigned int optlen)
2887 if (optlen < sizeof(int))
2889 if (get_user(val, (int __user *)optval))
2892 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2898 * 7.1.1 SCTP_RTOINFO
2900 * The protocol parameters used to initialize and bound retransmission
2901 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2902 * and modify these parameters.
2903 * All parameters are time values, in milliseconds. A value of 0, when
2904 * modifying the parameters, indicates that the current value should not
2908 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2910 struct sctp_rtoinfo rtoinfo;
2911 struct sctp_association *asoc;
2912 unsigned long rto_min, rto_max;
2913 struct sctp_sock *sp = sctp_sk(sk);
2915 if (optlen != sizeof (struct sctp_rtoinfo))
2918 if (copy_from_user(&rtoinfo, optval, optlen))
2921 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2923 /* Set the values to the specific association */
2924 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2927 rto_max = rtoinfo.srto_max;
2928 rto_min = rtoinfo.srto_min;
2931 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2933 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2936 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2938 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2940 if (rto_min > rto_max)
2944 if (rtoinfo.srto_initial != 0)
2946 msecs_to_jiffies(rtoinfo.srto_initial);
2947 asoc->rto_max = rto_max;
2948 asoc->rto_min = rto_min;
2950 /* If there is no association or the association-id = 0
2951 * set the values to the endpoint.
2953 if (rtoinfo.srto_initial != 0)
2954 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2955 sp->rtoinfo.srto_max = rto_max;
2956 sp->rtoinfo.srto_min = rto_min;
2964 * 7.1.2 SCTP_ASSOCINFO
2966 * This option is used to tune the maximum retransmission attempts
2967 * of the association.
2968 * Returns an error if the new association retransmission value is
2969 * greater than the sum of the retransmission value of the peer.
2970 * See [SCTP] for more information.
2973 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2976 struct sctp_assocparams assocparams;
2977 struct sctp_association *asoc;
2979 if (optlen != sizeof(struct sctp_assocparams))
2981 if (copy_from_user(&assocparams, optval, optlen))
2984 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2986 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2989 /* Set the values to the specific association */
2991 if (assocparams.sasoc_asocmaxrxt != 0) {
2994 struct sctp_transport *peer_addr;
2996 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2998 path_sum += peer_addr->pathmaxrxt;
3002 /* Only validate asocmaxrxt if we have more than
3003 * one path/transport. We do this because path
3004 * retransmissions are only counted when we have more
3008 assocparams.sasoc_asocmaxrxt > path_sum)
3011 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3014 if (assocparams.sasoc_cookie_life != 0)
3015 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3017 /* Set the values to the endpoint */
3018 struct sctp_sock *sp = sctp_sk(sk);
3020 if (assocparams.sasoc_asocmaxrxt != 0)
3021 sp->assocparams.sasoc_asocmaxrxt =
3022 assocparams.sasoc_asocmaxrxt;
3023 if (assocparams.sasoc_cookie_life != 0)
3024 sp->assocparams.sasoc_cookie_life =
3025 assocparams.sasoc_cookie_life;
3031 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3033 * This socket option is a boolean flag which turns on or off mapped V4
3034 * addresses. If this option is turned on and the socket is type
3035 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3036 * If this option is turned off, then no mapping will be done of V4
3037 * addresses and a user will receive both PF_INET6 and PF_INET type
3038 * addresses on the socket.
3040 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3043 struct sctp_sock *sp = sctp_sk(sk);
3045 if (optlen < sizeof(int))
3047 if (get_user(val, (int __user *)optval))
3058 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3059 * This option will get or set the maximum size to put in any outgoing
3060 * SCTP DATA chunk. If a message is larger than this size it will be
3061 * fragmented by SCTP into the specified size. Note that the underlying
3062 * SCTP implementation may fragment into smaller sized chunks when the
3063 * PMTU of the underlying association is smaller than the value set by
3064 * the user. The default value for this option is '0' which indicates
3065 * the user is NOT limiting fragmentation and only the PMTU will effect
3066 * SCTP's choice of DATA chunk size. Note also that values set larger
3067 * than the maximum size of an IP datagram will effectively let SCTP
3068 * control fragmentation (i.e. the same as setting this option to 0).
3070 * The following structure is used to access and modify this parameter:
3072 * struct sctp_assoc_value {
3073 * sctp_assoc_t assoc_id;
3074 * uint32_t assoc_value;
3077 * assoc_id: This parameter is ignored for one-to-one style sockets.
3078 * For one-to-many style sockets this parameter indicates which
3079 * association the user is performing an action upon. Note that if
3080 * this field's value is zero then the endpoints default value is
3081 * changed (effecting future associations only).
3082 * assoc_value: This parameter specifies the maximum size in bytes.
3084 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3086 struct sctp_assoc_value params;
3087 struct sctp_association *asoc;
3088 struct sctp_sock *sp = sctp_sk(sk);
3091 if (optlen == sizeof(int)) {
3092 pr_warn_ratelimited(DEPRECATED
3094 "Use of int in maxseg socket option.\n"
3095 "Use struct sctp_assoc_value instead\n",
3096 current->comm, task_pid_nr(current));
3097 if (copy_from_user(&val, optval, optlen))
3099 params.assoc_id = 0;
3100 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3101 if (copy_from_user(¶ms, optval, optlen))
3103 val = params.assoc_value;
3107 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3110 asoc = sctp_id2assoc(sk, params.assoc_id);
3111 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3116 val = asoc->pathmtu;
3117 val -= sp->pf->af->net_header_len;
3118 val -= sizeof(struct sctphdr) +
3119 sizeof(struct sctp_data_chunk);
3121 asoc->user_frag = val;
3122 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3124 sp->user_frag = val;
3132 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3134 * Requests that the peer mark the enclosed address as the association
3135 * primary. The enclosed address must be one of the association's
3136 * locally bound addresses. The following structure is used to make a
3137 * set primary request:
3139 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3140 unsigned int optlen)
3142 struct net *net = sock_net(sk);
3143 struct sctp_sock *sp;
3144 struct sctp_association *asoc = NULL;
3145 struct sctp_setpeerprim prim;
3146 struct sctp_chunk *chunk;
3152 if (!net->sctp.addip_enable)
3155 if (optlen != sizeof(struct sctp_setpeerprim))
3158 if (copy_from_user(&prim, optval, optlen))
3161 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3165 if (!asoc->peer.asconf_capable)
3168 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3171 if (!sctp_state(asoc, ESTABLISHED))
3174 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3178 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3179 return -EADDRNOTAVAIL;
3181 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3182 return -EADDRNOTAVAIL;
3184 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3185 chunk = sctp_make_asconf_set_prim(asoc,
3186 (union sctp_addr *)&prim.sspp_addr);
3190 err = sctp_send_asconf(asoc, chunk);
3192 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3197 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3198 unsigned int optlen)
3200 struct sctp_setadaptation adaptation;
3202 if (optlen != sizeof(struct sctp_setadaptation))
3204 if (copy_from_user(&adaptation, optval, optlen))
3207 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3213 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3215 * The context field in the sctp_sndrcvinfo structure is normally only
3216 * used when a failed message is retrieved holding the value that was
3217 * sent down on the actual send call. This option allows the setting of
3218 * a default context on an association basis that will be received on
3219 * reading messages from the peer. This is especially helpful in the
3220 * one-2-many model for an application to keep some reference to an
3221 * internal state machine that is processing messages on the
3222 * association. Note that the setting of this value only effects
3223 * received messages from the peer and does not effect the value that is
3224 * saved with outbound messages.
3226 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3227 unsigned int optlen)
3229 struct sctp_assoc_value params;
3230 struct sctp_sock *sp;
3231 struct sctp_association *asoc;
3233 if (optlen != sizeof(struct sctp_assoc_value))
3235 if (copy_from_user(¶ms, optval, optlen))
3240 if (params.assoc_id != 0) {
3241 asoc = sctp_id2assoc(sk, params.assoc_id);
3244 asoc->default_rcv_context = params.assoc_value;
3246 sp->default_rcv_context = params.assoc_value;
3253 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3255 * This options will at a minimum specify if the implementation is doing
3256 * fragmented interleave. Fragmented interleave, for a one to many
3257 * socket, is when subsequent calls to receive a message may return
3258 * parts of messages from different associations. Some implementations
3259 * may allow you to turn this value on or off. If so, when turned off,
3260 * no fragment interleave will occur (which will cause a head of line
3261 * blocking amongst multiple associations sharing the same one to many
3262 * socket). When this option is turned on, then each receive call may
3263 * come from a different association (thus the user must receive data
3264 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3265 * association each receive belongs to.
3267 * This option takes a boolean value. A non-zero value indicates that
3268 * fragmented interleave is on. A value of zero indicates that
3269 * fragmented interleave is off.
3271 * Note that it is important that an implementation that allows this
3272 * option to be turned on, have it off by default. Otherwise an unaware
3273 * application using the one to many model may become confused and act
3276 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3277 char __user *optval,
3278 unsigned int optlen)
3282 if (optlen != sizeof(int))
3284 if (get_user(val, (int __user *)optval))
3287 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3293 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3294 * (SCTP_PARTIAL_DELIVERY_POINT)
3296 * This option will set or get the SCTP partial delivery point. This
3297 * point is the size of a message where the partial delivery API will be
3298 * invoked to help free up rwnd space for the peer. Setting this to a
3299 * lower value will cause partial deliveries to happen more often. The
3300 * calls argument is an integer that sets or gets the partial delivery
3301 * point. Note also that the call will fail if the user attempts to set
3302 * this value larger than the socket receive buffer size.
3304 * Note that any single message having a length smaller than or equal to
3305 * the SCTP partial delivery point will be delivered in one single read
3306 * call as long as the user provided buffer is large enough to hold the
3309 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3310 char __user *optval,
3311 unsigned int optlen)
3315 if (optlen != sizeof(u32))
3317 if (get_user(val, (int __user *)optval))
3320 /* Note: We double the receive buffer from what the user sets
3321 * it to be, also initial rwnd is based on rcvbuf/2.
3323 if (val > (sk->sk_rcvbuf >> 1))
3326 sctp_sk(sk)->pd_point = val;
3328 return 0; /* is this the right error code? */
3332 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3334 * This option will allow a user to change the maximum burst of packets
3335 * that can be emitted by this association. Note that the default value
3336 * is 4, and some implementations may restrict this setting so that it
3337 * can only be lowered.
3339 * NOTE: This text doesn't seem right. Do this on a socket basis with
3340 * future associations inheriting the socket value.
3342 static int sctp_setsockopt_maxburst(struct sock *sk,
3343 char __user *optval,
3344 unsigned int optlen)
3346 struct sctp_assoc_value params;
3347 struct sctp_sock *sp;
3348 struct sctp_association *asoc;
3352 if (optlen == sizeof(int)) {
3353 pr_warn_ratelimited(DEPRECATED
3355 "Use of int in max_burst socket option deprecated.\n"
3356 "Use struct sctp_assoc_value instead\n",
3357 current->comm, task_pid_nr(current));
3358 if (copy_from_user(&val, optval, optlen))
3360 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3361 if (copy_from_user(¶ms, optval, optlen))
3363 val = params.assoc_value;
3364 assoc_id = params.assoc_id;
3370 if (assoc_id != 0) {
3371 asoc = sctp_id2assoc(sk, assoc_id);
3374 asoc->max_burst = val;
3376 sp->max_burst = val;
3382 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3384 * This set option adds a chunk type that the user is requesting to be
3385 * received only in an authenticated way. Changes to the list of chunks
3386 * will only effect future associations on the socket.
3388 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3389 char __user *optval,
3390 unsigned int optlen)
3392 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3393 struct sctp_authchunk val;
3395 if (!ep->auth_enable)
3398 if (optlen != sizeof(struct sctp_authchunk))
3400 if (copy_from_user(&val, optval, optlen))
3403 switch (val.sauth_chunk) {
3405 case SCTP_CID_INIT_ACK:
3406 case SCTP_CID_SHUTDOWN_COMPLETE:
3411 /* add this chunk id to the endpoint */
3412 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3416 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3418 * This option gets or sets the list of HMAC algorithms that the local
3419 * endpoint requires the peer to use.
3421 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3422 char __user *optval,
3423 unsigned int optlen)
3425 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3426 struct sctp_hmacalgo *hmacs;
3430 if (!ep->auth_enable)
3433 if (optlen < sizeof(struct sctp_hmacalgo))
3436 hmacs = memdup_user(optval, optlen);
3438 return PTR_ERR(hmacs);
3440 idents = hmacs->shmac_num_idents;
3441 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3442 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3447 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3454 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3456 * This option will set a shared secret key which is used to build an
3457 * association shared key.
3459 static int sctp_setsockopt_auth_key(struct sock *sk,
3460 char __user *optval,
3461 unsigned int optlen)
3463 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3464 struct sctp_authkey *authkey;
3465 struct sctp_association *asoc;
3468 if (!ep->auth_enable)
3471 if (optlen <= sizeof(struct sctp_authkey))
3474 authkey = memdup_user(optval, optlen);
3475 if (IS_ERR(authkey))
3476 return PTR_ERR(authkey);
3478 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3483 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3484 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3489 ret = sctp_auth_set_key(ep, asoc, authkey);
3496 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3498 * This option will get or set the active shared key to be used to build
3499 * the association shared key.
3501 static int sctp_setsockopt_active_key(struct sock *sk,
3502 char __user *optval,
3503 unsigned int optlen)
3505 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3506 struct sctp_authkeyid val;
3507 struct sctp_association *asoc;
3509 if (!ep->auth_enable)
3512 if (optlen != sizeof(struct sctp_authkeyid))
3514 if (copy_from_user(&val, optval, optlen))
3517 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3518 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3521 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3525 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3527 * This set option will delete a shared secret key from use.
3529 static int sctp_setsockopt_del_key(struct sock *sk,
3530 char __user *optval,
3531 unsigned int optlen)
3533 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3534 struct sctp_authkeyid val;
3535 struct sctp_association *asoc;
3537 if (!ep->auth_enable)
3540 if (optlen != sizeof(struct sctp_authkeyid))
3542 if (copy_from_user(&val, optval, optlen))
3545 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3546 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3549 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3554 * 8.1.23 SCTP_AUTO_ASCONF
3556 * This option will enable or disable the use of the automatic generation of
3557 * ASCONF chunks to add and delete addresses to an existing association. Note
3558 * that this option has two caveats namely: a) it only affects sockets that
3559 * are bound to all addresses available to the SCTP stack, and b) the system
3560 * administrator may have an overriding control that turns the ASCONF feature
3561 * off no matter what setting the socket option may have.
3562 * This option expects an integer boolean flag, where a non-zero value turns on
3563 * the option, and a zero value turns off the option.
3564 * Note. In this implementation, socket operation overrides default parameter
3565 * being set by sysctl as well as FreeBSD implementation
3567 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3568 unsigned int optlen)
3571 struct sctp_sock *sp = sctp_sk(sk);
3573 if (optlen < sizeof(int))
3575 if (get_user(val, (int __user *)optval))
3577 if (!sctp_is_ep_boundall(sk) && val)
3579 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3582 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
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;
3591 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3596 * SCTP_PEER_ADDR_THLDS
3598 * This option allows us to alter the partially failed threshold for one or all
3599 * transports in an association. See Section 6.1 of:
3600 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3602 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3603 char __user *optval,
3604 unsigned int optlen)
3606 struct sctp_paddrthlds val;
3607 struct sctp_transport *trans;
3608 struct sctp_association *asoc;
3610 if (optlen < sizeof(struct sctp_paddrthlds))
3612 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3613 sizeof(struct sctp_paddrthlds)))
3617 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3618 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3621 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3623 if (val.spt_pathmaxrxt)
3624 trans->pathmaxrxt = val.spt_pathmaxrxt;
3625 trans->pf_retrans = val.spt_pathpfthld;
3628 if (val.spt_pathmaxrxt)
3629 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3630 asoc->pf_retrans = val.spt_pathpfthld;
3632 trans = sctp_addr_id2transport(sk, &val.spt_address,
3637 if (val.spt_pathmaxrxt)
3638 trans->pathmaxrxt = val.spt_pathmaxrxt;
3639 trans->pf_retrans = val.spt_pathpfthld;
3645 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3646 char __user *optval,
3647 unsigned int optlen)
3651 if (optlen < sizeof(int))
3653 if (get_user(val, (int __user *) optval))
3656 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3661 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3662 char __user *optval,
3663 unsigned int optlen)
3667 if (optlen < sizeof(int))
3669 if (get_user(val, (int __user *) optval))
3672 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3677 static int sctp_setsockopt_pr_supported(struct sock *sk,
3678 char __user *optval,
3679 unsigned int optlen)
3681 struct sctp_assoc_value params;
3682 struct sctp_association *asoc;
3683 int retval = -EINVAL;
3685 if (optlen != sizeof(params))
3688 if (copy_from_user(¶ms, optval, optlen)) {
3693 asoc = sctp_id2assoc(sk, params.assoc_id);
3695 asoc->prsctp_enable = !!params.assoc_value;
3696 } else if (!params.assoc_id) {
3697 struct sctp_sock *sp = sctp_sk(sk);
3699 sp->ep->prsctp_enable = !!params.assoc_value;
3710 static int sctp_setsockopt_default_prinfo(struct sock *sk,
3711 char __user *optval,
3712 unsigned int optlen)
3714 struct sctp_default_prinfo info;
3715 struct sctp_association *asoc;
3716 int retval = -EINVAL;
3718 if (optlen != sizeof(info))
3721 if (copy_from_user(&info, optval, sizeof(info))) {
3726 if (info.pr_policy & ~SCTP_PR_SCTP_MASK)
3729 if (info.pr_policy == SCTP_PR_SCTP_NONE)
3732 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
3734 SCTP_PR_SET_POLICY(asoc->default_flags, info.pr_policy);
3735 asoc->default_timetolive = info.pr_value;
3736 } else if (!info.pr_assoc_id) {
3737 struct sctp_sock *sp = sctp_sk(sk);
3739 SCTP_PR_SET_POLICY(sp->default_flags, info.pr_policy);
3740 sp->default_timetolive = info.pr_value;
3751 /* API 6.2 setsockopt(), getsockopt()
3753 * Applications use setsockopt() and getsockopt() to set or retrieve
3754 * socket options. Socket options are used to change the default
3755 * behavior of sockets calls. They are described in Section 7.
3759 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3760 * int __user *optlen);
3761 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3764 * sd - the socket descript.
3765 * level - set to IPPROTO_SCTP for all SCTP options.
3766 * optname - the option name.
3767 * optval - the buffer to store the value of the option.
3768 * optlen - the size of the buffer.
3770 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3771 char __user *optval, unsigned int optlen)
3775 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3777 /* I can hardly begin to describe how wrong this is. This is
3778 * so broken as to be worse than useless. The API draft
3779 * REALLY is NOT helpful here... I am not convinced that the
3780 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3781 * are at all well-founded.
3783 if (level != SOL_SCTP) {
3784 struct sctp_af *af = sctp_sk(sk)->pf->af;
3785 retval = af->setsockopt(sk, level, optname, optval, optlen);
3792 case SCTP_SOCKOPT_BINDX_ADD:
3793 /* 'optlen' is the size of the addresses buffer. */
3794 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3795 optlen, SCTP_BINDX_ADD_ADDR);
3798 case SCTP_SOCKOPT_BINDX_REM:
3799 /* 'optlen' is the size of the addresses buffer. */
3800 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3801 optlen, SCTP_BINDX_REM_ADDR);
3804 case SCTP_SOCKOPT_CONNECTX_OLD:
3805 /* 'optlen' is the size of the addresses buffer. */
3806 retval = sctp_setsockopt_connectx_old(sk,
3807 (struct sockaddr __user *)optval,
3811 case SCTP_SOCKOPT_CONNECTX:
3812 /* 'optlen' is the size of the addresses buffer. */
3813 retval = sctp_setsockopt_connectx(sk,
3814 (struct sockaddr __user *)optval,
3818 case SCTP_DISABLE_FRAGMENTS:
3819 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3823 retval = sctp_setsockopt_events(sk, optval, optlen);
3826 case SCTP_AUTOCLOSE:
3827 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3830 case SCTP_PEER_ADDR_PARAMS:
3831 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3834 case SCTP_DELAYED_SACK:
3835 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3837 case SCTP_PARTIAL_DELIVERY_POINT:
3838 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3842 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3844 case SCTP_DEFAULT_SEND_PARAM:
3845 retval = sctp_setsockopt_default_send_param(sk, optval,
3848 case SCTP_DEFAULT_SNDINFO:
3849 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
3851 case SCTP_PRIMARY_ADDR:
3852 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3854 case SCTP_SET_PEER_PRIMARY_ADDR:
3855 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3858 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3861 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3863 case SCTP_ASSOCINFO:
3864 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3866 case SCTP_I_WANT_MAPPED_V4_ADDR:
3867 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3870 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3872 case SCTP_ADAPTATION_LAYER:
3873 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3876 retval = sctp_setsockopt_context(sk, optval, optlen);
3878 case SCTP_FRAGMENT_INTERLEAVE:
3879 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3881 case SCTP_MAX_BURST:
3882 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3884 case SCTP_AUTH_CHUNK:
3885 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3887 case SCTP_HMAC_IDENT:
3888 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3891 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3893 case SCTP_AUTH_ACTIVE_KEY:
3894 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3896 case SCTP_AUTH_DELETE_KEY:
3897 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3899 case SCTP_AUTO_ASCONF:
3900 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3902 case SCTP_PEER_ADDR_THLDS:
3903 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3905 case SCTP_RECVRCVINFO:
3906 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
3908 case SCTP_RECVNXTINFO:
3909 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
3911 case SCTP_PR_SUPPORTED:
3912 retval = sctp_setsockopt_pr_supported(sk, optval, optlen);
3914 case SCTP_DEFAULT_PRINFO:
3915 retval = sctp_setsockopt_default_prinfo(sk, optval, optlen);
3918 retval = -ENOPROTOOPT;
3928 /* API 3.1.6 connect() - UDP Style Syntax
3930 * An application may use the connect() call in the UDP model to initiate an
3931 * association without sending data.
3935 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3937 * sd: the socket descriptor to have a new association added to.
3939 * nam: the address structure (either struct sockaddr_in or struct
3940 * sockaddr_in6 defined in RFC2553 [7]).
3942 * len: the size of the address.
3944 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3952 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
3955 /* Validate addr_len before calling common connect/connectx routine. */
3956 af = sctp_get_af_specific(addr->sa_family);
3957 if (!af || addr_len < af->sockaddr_len) {
3960 /* Pass correct addr len to common routine (so it knows there
3961 * is only one address being passed.
3963 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3970 /* FIXME: Write comments. */
3971 static int sctp_disconnect(struct sock *sk, int flags)
3973 return -EOPNOTSUPP; /* STUB */
3976 /* 4.1.4 accept() - TCP Style Syntax
3978 * Applications use accept() call to remove an established SCTP
3979 * association from the accept queue of the endpoint. A new socket
3980 * descriptor will be returned from accept() to represent the newly
3981 * formed association.
3983 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3985 struct sctp_sock *sp;
3986 struct sctp_endpoint *ep;
3987 struct sock *newsk = NULL;
3988 struct sctp_association *asoc;
3997 if (!sctp_style(sk, TCP)) {
3998 error = -EOPNOTSUPP;
4002 if (!sctp_sstate(sk, LISTENING)) {
4007 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
4009 error = sctp_wait_for_accept(sk, timeo);
4013 /* We treat the list of associations on the endpoint as the accept
4014 * queue and pick the first association on the list.
4016 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
4018 newsk = sp->pf->create_accept_sk(sk, asoc);
4024 /* Populate the fields of the newsk from the oldsk and migrate the
4025 * asoc to the newsk.
4027 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
4035 /* The SCTP ioctl handler. */
4036 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
4043 * SEQPACKET-style sockets in LISTENING state are valid, for
4044 * SCTP, so only discard TCP-style sockets in LISTENING state.
4046 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4051 struct sk_buff *skb;
4052 unsigned int amount = 0;
4054 skb = skb_peek(&sk->sk_receive_queue);
4057 * We will only return the amount of this packet since
4058 * that is all that will be read.
4062 rc = put_user(amount, (int __user *)arg);
4074 /* This is the function which gets called during socket creation to
4075 * initialized the SCTP-specific portion of the sock.
4076 * The sock structure should already be zero-filled memory.
4078 static int sctp_init_sock(struct sock *sk)
4080 struct net *net = sock_net(sk);
4081 struct sctp_sock *sp;
4083 pr_debug("%s: sk:%p\n", __func__, sk);
4087 /* Initialize the SCTP per socket area. */
4088 switch (sk->sk_type) {
4089 case SOCK_SEQPACKET:
4090 sp->type = SCTP_SOCKET_UDP;
4093 sp->type = SCTP_SOCKET_TCP;
4096 return -ESOCKTNOSUPPORT;
4099 sk->sk_gso_type = SKB_GSO_SCTP;
4101 /* Initialize default send parameters. These parameters can be
4102 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4104 sp->default_stream = 0;
4105 sp->default_ppid = 0;
4106 sp->default_flags = 0;
4107 sp->default_context = 0;
4108 sp->default_timetolive = 0;
4110 sp->default_rcv_context = 0;
4111 sp->max_burst = net->sctp.max_burst;
4113 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4115 /* Initialize default setup parameters. These parameters
4116 * can be modified with the SCTP_INITMSG socket option or
4117 * overridden by the SCTP_INIT CMSG.
4119 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4120 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4121 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4122 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4124 /* Initialize default RTO related parameters. These parameters can
4125 * be modified for with the SCTP_RTOINFO socket option.
4127 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4128 sp->rtoinfo.srto_max = net->sctp.rto_max;
4129 sp->rtoinfo.srto_min = net->sctp.rto_min;
4131 /* Initialize default association related parameters. These parameters
4132 * can be modified with the SCTP_ASSOCINFO socket option.
4134 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4135 sp->assocparams.sasoc_number_peer_destinations = 0;
4136 sp->assocparams.sasoc_peer_rwnd = 0;
4137 sp->assocparams.sasoc_local_rwnd = 0;
4138 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4140 /* Initialize default event subscriptions. By default, all the
4143 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4145 /* Default Peer Address Parameters. These defaults can
4146 * be modified via SCTP_PEER_ADDR_PARAMS
4148 sp->hbinterval = net->sctp.hb_interval;
4149 sp->pathmaxrxt = net->sctp.max_retrans_path;
4150 sp->pathmtu = 0; /* allow default discovery */
4151 sp->sackdelay = net->sctp.sack_timeout;
4153 sp->param_flags = SPP_HB_ENABLE |
4155 SPP_SACKDELAY_ENABLE;
4157 /* If enabled no SCTP message fragmentation will be performed.
4158 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4160 sp->disable_fragments = 0;
4162 /* Enable Nagle algorithm by default. */
4165 sp->recvrcvinfo = 0;
4166 sp->recvnxtinfo = 0;
4168 /* Enable by default. */
4171 /* Auto-close idle associations after the configured
4172 * number of seconds. A value of 0 disables this
4173 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4174 * for UDP-style sockets only.
4178 /* User specified fragmentation limit. */
4181 sp->adaptation_ind = 0;
4183 sp->pf = sctp_get_pf_specific(sk->sk_family);
4185 /* Control variables for partial data delivery. */
4186 atomic_set(&sp->pd_mode, 0);
4187 skb_queue_head_init(&sp->pd_lobby);
4188 sp->frag_interleave = 0;
4190 /* Create a per socket endpoint structure. Even if we
4191 * change the data structure relationships, this may still
4192 * be useful for storing pre-connect address information.
4194 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4200 sk->sk_destruct = sctp_destruct_sock;
4202 SCTP_DBG_OBJCNT_INC(sock);
4205 percpu_counter_inc(&sctp_sockets_allocated);
4206 sock_prot_inuse_add(net, sk->sk_prot, 1);
4208 /* Nothing can fail after this block, otherwise
4209 * sctp_destroy_sock() will be called without addr_wq_lock held
4211 if (net->sctp.default_auto_asconf) {
4212 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
4213 list_add_tail(&sp->auto_asconf_list,
4214 &net->sctp.auto_asconf_splist);
4215 sp->do_auto_asconf = 1;
4216 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4218 sp->do_auto_asconf = 0;
4226 /* Cleanup any SCTP per socket resources. Must be called with
4227 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4229 static void sctp_destroy_sock(struct sock *sk)
4231 struct sctp_sock *sp;
4233 pr_debug("%s: sk:%p\n", __func__, sk);
4235 /* Release our hold on the endpoint. */
4237 /* This could happen during socket init, thus we bail out
4238 * early, since the rest of the below is not setup either.
4243 if (sp->do_auto_asconf) {
4244 sp->do_auto_asconf = 0;
4245 list_del(&sp->auto_asconf_list);
4247 sctp_endpoint_free(sp->ep);
4249 percpu_counter_dec(&sctp_sockets_allocated);
4250 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4254 /* Triggered when there are no references on the socket anymore */
4255 static void sctp_destruct_sock(struct sock *sk)
4257 struct sctp_sock *sp = sctp_sk(sk);
4259 /* Free up the HMAC transform. */
4260 crypto_free_shash(sp->hmac);
4262 inet_sock_destruct(sk);
4265 /* API 4.1.7 shutdown() - TCP Style Syntax
4266 * int shutdown(int socket, int how);
4268 * sd - the socket descriptor of the association to be closed.
4269 * how - Specifies the type of shutdown. The values are
4272 * Disables further receive operations. No SCTP
4273 * protocol action is taken.
4275 * Disables further send operations, and initiates
4276 * the SCTP shutdown sequence.
4278 * Disables further send and receive operations
4279 * and initiates the SCTP shutdown sequence.
4281 static void sctp_shutdown(struct sock *sk, int how)
4283 struct net *net = sock_net(sk);
4284 struct sctp_endpoint *ep;
4285 struct sctp_association *asoc;
4287 if (!sctp_style(sk, TCP))
4290 if (how & SEND_SHUTDOWN) {
4291 sk->sk_state = SCTP_SS_CLOSING;
4292 ep = sctp_sk(sk)->ep;
4293 if (!list_empty(&ep->asocs)) {
4294 asoc = list_entry(ep->asocs.next,
4295 struct sctp_association, asocs);
4296 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4301 int sctp_get_sctp_info(struct sock *sk, struct sctp_association *asoc,
4302 struct sctp_info *info)
4304 struct sctp_transport *prim;
4305 struct list_head *pos;
4308 memset(info, 0, sizeof(*info));
4310 struct sctp_sock *sp = sctp_sk(sk);
4312 info->sctpi_s_autoclose = sp->autoclose;
4313 info->sctpi_s_adaptation_ind = sp->adaptation_ind;
4314 info->sctpi_s_pd_point = sp->pd_point;
4315 info->sctpi_s_nodelay = sp->nodelay;
4316 info->sctpi_s_disable_fragments = sp->disable_fragments;
4317 info->sctpi_s_v4mapped = sp->v4mapped;
4318 info->sctpi_s_frag_interleave = sp->frag_interleave;
4319 info->sctpi_s_type = sp->type;
4324 info->sctpi_tag = asoc->c.my_vtag;
4325 info->sctpi_state = asoc->state;
4326 info->sctpi_rwnd = asoc->a_rwnd;
4327 info->sctpi_unackdata = asoc->unack_data;
4328 info->sctpi_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4329 info->sctpi_instrms = asoc->c.sinit_max_instreams;
4330 info->sctpi_outstrms = asoc->c.sinit_num_ostreams;
4331 list_for_each(pos, &asoc->base.inqueue.in_chunk_list)
4332 info->sctpi_inqueue++;
4333 list_for_each(pos, &asoc->outqueue.out_chunk_list)
4334 info->sctpi_outqueue++;
4335 info->sctpi_overall_error = asoc->overall_error_count;
4336 info->sctpi_max_burst = asoc->max_burst;
4337 info->sctpi_maxseg = asoc->frag_point;
4338 info->sctpi_peer_rwnd = asoc->peer.rwnd;
4339 info->sctpi_peer_tag = asoc->c.peer_vtag;
4341 mask = asoc->peer.ecn_capable << 1;
4342 mask = (mask | asoc->peer.ipv4_address) << 1;
4343 mask = (mask | asoc->peer.ipv6_address) << 1;
4344 mask = (mask | asoc->peer.hostname_address) << 1;
4345 mask = (mask | asoc->peer.asconf_capable) << 1;
4346 mask = (mask | asoc->peer.prsctp_capable) << 1;
4347 mask = (mask | asoc->peer.auth_capable);
4348 info->sctpi_peer_capable = mask;
4349 mask = asoc->peer.sack_needed << 1;
4350 mask = (mask | asoc->peer.sack_generation) << 1;
4351 mask = (mask | asoc->peer.zero_window_announced);
4352 info->sctpi_peer_sack = mask;
4354 info->sctpi_isacks = asoc->stats.isacks;
4355 info->sctpi_osacks = asoc->stats.osacks;
4356 info->sctpi_opackets = asoc->stats.opackets;
4357 info->sctpi_ipackets = asoc->stats.ipackets;
4358 info->sctpi_rtxchunks = asoc->stats.rtxchunks;
4359 info->sctpi_outofseqtsns = asoc->stats.outofseqtsns;
4360 info->sctpi_idupchunks = asoc->stats.idupchunks;
4361 info->sctpi_gapcnt = asoc->stats.gapcnt;
4362 info->sctpi_ouodchunks = asoc->stats.ouodchunks;
4363 info->sctpi_iuodchunks = asoc->stats.iuodchunks;
4364 info->sctpi_oodchunks = asoc->stats.oodchunks;
4365 info->sctpi_iodchunks = asoc->stats.iodchunks;
4366 info->sctpi_octrlchunks = asoc->stats.octrlchunks;
4367 info->sctpi_ictrlchunks = asoc->stats.ictrlchunks;
4369 prim = asoc->peer.primary_path;
4370 memcpy(&info->sctpi_p_address, &prim->ipaddr,
4371 sizeof(struct sockaddr_storage));
4372 info->sctpi_p_state = prim->state;
4373 info->sctpi_p_cwnd = prim->cwnd;
4374 info->sctpi_p_srtt = prim->srtt;
4375 info->sctpi_p_rto = jiffies_to_msecs(prim->rto);
4376 info->sctpi_p_hbinterval = prim->hbinterval;
4377 info->sctpi_p_pathmaxrxt = prim->pathmaxrxt;
4378 info->sctpi_p_sackdelay = jiffies_to_msecs(prim->sackdelay);
4379 info->sctpi_p_ssthresh = prim->ssthresh;
4380 info->sctpi_p_partial_bytes_acked = prim->partial_bytes_acked;
4381 info->sctpi_p_flight_size = prim->flight_size;
4382 info->sctpi_p_error = prim->error_count;
4386 EXPORT_SYMBOL_GPL(sctp_get_sctp_info);
4388 /* use callback to avoid exporting the core structure */
4389 int sctp_transport_walk_start(struct rhashtable_iter *iter)
4393 err = rhashtable_walk_init(&sctp_transport_hashtable, iter,
4398 err = rhashtable_walk_start(iter);
4399 if (err && err != -EAGAIN) {
4400 rhashtable_walk_stop(iter);
4401 rhashtable_walk_exit(iter);
4408 void sctp_transport_walk_stop(struct rhashtable_iter *iter)
4410 rhashtable_walk_stop(iter);
4411 rhashtable_walk_exit(iter);
4414 struct sctp_transport *sctp_transport_get_next(struct net *net,
4415 struct rhashtable_iter *iter)
4417 struct sctp_transport *t;
4419 t = rhashtable_walk_next(iter);
4420 for (; t; t = rhashtable_walk_next(iter)) {
4422 if (PTR_ERR(t) == -EAGAIN)
4427 if (net_eq(sock_net(t->asoc->base.sk), net) &&
4428 t->asoc->peer.primary_path == t)
4435 struct sctp_transport *sctp_transport_get_idx(struct net *net,
4436 struct rhashtable_iter *iter,
4439 void *obj = SEQ_START_TOKEN;
4441 while (pos && (obj = sctp_transport_get_next(net, iter)) &&
4448 int sctp_for_each_endpoint(int (*cb)(struct sctp_endpoint *, void *),
4452 struct sctp_ep_common *epb;
4453 struct sctp_hashbucket *head;
4455 for (head = sctp_ep_hashtable; hash < sctp_ep_hashsize;
4457 read_lock(&head->lock);
4458 sctp_for_each_hentry(epb, &head->chain) {
4459 err = cb(sctp_ep(epb), p);
4463 read_unlock(&head->lock);
4468 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint);
4470 int sctp_transport_lookup_process(int (*cb)(struct sctp_transport *, void *),
4472 const union sctp_addr *laddr,
4473 const union sctp_addr *paddr, void *p)
4475 struct sctp_transport *transport;
4479 transport = sctp_addrs_lookup_transport(net, laddr, paddr);
4480 if (!transport || !sctp_transport_hold(transport))
4483 sctp_association_hold(transport->asoc);
4484 sctp_transport_put(transport);
4487 err = cb(transport, p);
4488 sctp_association_put(transport->asoc);
4493 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process);
4495 int sctp_for_each_transport(int (*cb)(struct sctp_transport *, void *),
4496 struct net *net, int pos, void *p) {
4497 struct rhashtable_iter hti;
4501 err = sctp_transport_walk_start(&hti);
4505 sctp_transport_get_idx(net, &hti, pos);
4506 obj = sctp_transport_get_next(net, &hti);
4507 for (; obj && !IS_ERR(obj); obj = sctp_transport_get_next(net, &hti)) {
4508 struct sctp_transport *transport = obj;
4510 if (!sctp_transport_hold(transport))
4512 err = cb(transport, p);
4513 sctp_transport_put(transport);
4517 sctp_transport_walk_stop(&hti);
4521 EXPORT_SYMBOL_GPL(sctp_for_each_transport);
4523 /* 7.2.1 Association Status (SCTP_STATUS)
4525 * Applications can retrieve current status information about an
4526 * association, including association state, peer receiver window size,
4527 * number of unacked data chunks, and number of data chunks pending
4528 * receipt. This information is read-only.
4530 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4531 char __user *optval,
4534 struct sctp_status status;
4535 struct sctp_association *asoc = NULL;
4536 struct sctp_transport *transport;
4537 sctp_assoc_t associd;
4540 if (len < sizeof(status)) {
4545 len = sizeof(status);
4546 if (copy_from_user(&status, optval, len)) {
4551 associd = status.sstat_assoc_id;
4552 asoc = sctp_id2assoc(sk, associd);
4558 transport = asoc->peer.primary_path;
4560 status.sstat_assoc_id = sctp_assoc2id(asoc);
4561 status.sstat_state = sctp_assoc_to_state(asoc);
4562 status.sstat_rwnd = asoc->peer.rwnd;
4563 status.sstat_unackdata = asoc->unack_data;
4565 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4566 status.sstat_instrms = asoc->c.sinit_max_instreams;
4567 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4568 status.sstat_fragmentation_point = asoc->frag_point;
4569 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4570 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4571 transport->af_specific->sockaddr_len);
4572 /* Map ipv4 address into v4-mapped-on-v6 address. */
4573 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
4574 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4575 status.sstat_primary.spinfo_state = transport->state;
4576 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4577 status.sstat_primary.spinfo_srtt = transport->srtt;
4578 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4579 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4581 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4582 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4584 if (put_user(len, optlen)) {
4589 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4590 __func__, len, status.sstat_state, status.sstat_rwnd,
4591 status.sstat_assoc_id);
4593 if (copy_to_user(optval, &status, len)) {
4603 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4605 * Applications can retrieve information about a specific peer address
4606 * of an association, including its reachability state, congestion
4607 * window, and retransmission timer values. This information is
4610 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4611 char __user *optval,
4614 struct sctp_paddrinfo pinfo;
4615 struct sctp_transport *transport;
4618 if (len < sizeof(pinfo)) {
4623 len = sizeof(pinfo);
4624 if (copy_from_user(&pinfo, optval, len)) {
4629 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4630 pinfo.spinfo_assoc_id);
4634 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4635 pinfo.spinfo_state = transport->state;
4636 pinfo.spinfo_cwnd = transport->cwnd;
4637 pinfo.spinfo_srtt = transport->srtt;
4638 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4639 pinfo.spinfo_mtu = transport->pathmtu;
4641 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4642 pinfo.spinfo_state = SCTP_ACTIVE;
4644 if (put_user(len, optlen)) {
4649 if (copy_to_user(optval, &pinfo, len)) {
4658 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4660 * This option is a on/off flag. If enabled no SCTP message
4661 * fragmentation will be performed. Instead if a message being sent
4662 * exceeds the current PMTU size, the message will NOT be sent and
4663 * instead a error will be indicated to the user.
4665 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4666 char __user *optval, int __user *optlen)
4670 if (len < sizeof(int))
4674 val = (sctp_sk(sk)->disable_fragments == 1);
4675 if (put_user(len, optlen))
4677 if (copy_to_user(optval, &val, len))
4682 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4684 * This socket option is used to specify various notifications and
4685 * ancillary data the user wishes to receive.
4687 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4692 if (len > sizeof(struct sctp_event_subscribe))
4693 len = sizeof(struct sctp_event_subscribe);
4694 if (put_user(len, optlen))
4696 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4701 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4703 * This socket option is applicable to the UDP-style socket only. When
4704 * set it will cause associations that are idle for more than the
4705 * specified number of seconds to automatically close. An association
4706 * being idle is defined an association that has NOT sent or received
4707 * user data. The special value of '0' indicates that no automatic
4708 * close of any associations should be performed. The option expects an
4709 * integer defining the number of seconds of idle time before an
4710 * association is closed.
4712 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4714 /* Applicable to UDP-style socket only */
4715 if (sctp_style(sk, TCP))
4717 if (len < sizeof(int))
4720 if (put_user(len, optlen))
4722 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4727 /* Helper routine to branch off an association to a new socket. */
4728 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4730 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4731 struct sctp_sock *sp = sctp_sk(sk);
4732 struct socket *sock;
4738 /* An association cannot be branched off from an already peeled-off
4739 * socket, nor is this supported for tcp style sockets.
4741 if (!sctp_style(sk, UDP))
4744 /* Create a new socket. */
4745 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4749 sctp_copy_sock(sock->sk, sk, asoc);
4751 /* Make peeled-off sockets more like 1-1 accepted sockets.
4752 * Set the daddr and initialize id to something more random
4754 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
4756 /* Populate the fields of the newsk from the oldsk and migrate the
4757 * asoc to the newsk.
4759 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4765 EXPORT_SYMBOL(sctp_do_peeloff);
4767 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4769 sctp_peeloff_arg_t peeloff;
4770 struct socket *newsock;
4771 struct file *newfile;
4774 if (len < sizeof(sctp_peeloff_arg_t))
4776 len = sizeof(sctp_peeloff_arg_t);
4777 if (copy_from_user(&peeloff, optval, len))
4780 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4784 /* Map the socket to an unused fd that can be returned to the user. */
4785 retval = get_unused_fd_flags(0);
4787 sock_release(newsock);
4791 newfile = sock_alloc_file(newsock, 0, NULL);
4792 if (IS_ERR(newfile)) {
4793 put_unused_fd(retval);
4794 sock_release(newsock);
4795 return PTR_ERR(newfile);
4798 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4801 /* Return the fd mapped to the new socket. */
4802 if (put_user(len, optlen)) {
4804 put_unused_fd(retval);
4807 peeloff.sd = retval;
4808 if (copy_to_user(optval, &peeloff, len)) {
4810 put_unused_fd(retval);
4813 fd_install(retval, newfile);
4818 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4820 * Applications can enable or disable heartbeats for any peer address of
4821 * an association, modify an address's heartbeat interval, force a
4822 * heartbeat to be sent immediately, and adjust the address's maximum
4823 * number of retransmissions sent before an address is considered
4824 * unreachable. The following structure is used to access and modify an
4825 * address's parameters:
4827 * struct sctp_paddrparams {
4828 * sctp_assoc_t spp_assoc_id;
4829 * struct sockaddr_storage spp_address;
4830 * uint32_t spp_hbinterval;
4831 * uint16_t spp_pathmaxrxt;
4832 * uint32_t spp_pathmtu;
4833 * uint32_t spp_sackdelay;
4834 * uint32_t spp_flags;
4837 * spp_assoc_id - (one-to-many style socket) This is filled in the
4838 * application, and identifies the association for
4840 * spp_address - This specifies which address is of interest.
4841 * spp_hbinterval - This contains the value of the heartbeat interval,
4842 * in milliseconds. If a value of zero
4843 * is present in this field then no changes are to
4844 * be made to this parameter.
4845 * spp_pathmaxrxt - This contains the maximum number of
4846 * retransmissions before this address shall be
4847 * considered unreachable. If a value of zero
4848 * is present in this field then no changes are to
4849 * be made to this parameter.
4850 * spp_pathmtu - When Path MTU discovery is disabled the value
4851 * specified here will be the "fixed" path mtu.
4852 * Note that if the spp_address field is empty
4853 * then all associations on this address will
4854 * have this fixed path mtu set upon them.
4856 * spp_sackdelay - When delayed sack is enabled, this value specifies
4857 * the number of milliseconds that sacks will be delayed
4858 * for. This value will apply to all addresses of an
4859 * association if the spp_address field is empty. Note
4860 * also, that if delayed sack is enabled and this
4861 * value is set to 0, no change is made to the last
4862 * recorded delayed sack timer value.
4864 * spp_flags - These flags are used to control various features
4865 * on an association. The flag field may contain
4866 * zero or more of the following options.
4868 * SPP_HB_ENABLE - Enable heartbeats on the
4869 * specified address. Note that if the address
4870 * field is empty all addresses for the association
4871 * have heartbeats enabled upon them.
4873 * SPP_HB_DISABLE - Disable heartbeats on the
4874 * speicifed address. Note that if the address
4875 * field is empty all addresses for the association
4876 * will have their heartbeats disabled. Note also
4877 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4878 * mutually exclusive, only one of these two should
4879 * be specified. Enabling both fields will have
4880 * undetermined results.
4882 * SPP_HB_DEMAND - Request a user initiated heartbeat
4883 * to be made immediately.
4885 * SPP_PMTUD_ENABLE - This field will enable PMTU
4886 * discovery upon the specified address. Note that
4887 * if the address feild is empty then all addresses
4888 * on the association are effected.
4890 * SPP_PMTUD_DISABLE - This field will disable PMTU
4891 * discovery upon the specified address. Note that
4892 * if the address feild is empty then all addresses
4893 * on the association are effected. Not also that
4894 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4895 * exclusive. Enabling both will have undetermined
4898 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4899 * on delayed sack. The time specified in spp_sackdelay
4900 * is used to specify the sack delay for this address. Note
4901 * that if spp_address is empty then all addresses will
4902 * enable delayed sack and take on the sack delay
4903 * value specified in spp_sackdelay.
4904 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4905 * off delayed sack. If the spp_address field is blank then
4906 * delayed sack is disabled for the entire association. Note
4907 * also that this field is mutually exclusive to
4908 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4911 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4912 char __user *optval, int __user *optlen)
4914 struct sctp_paddrparams params;
4915 struct sctp_transport *trans = NULL;
4916 struct sctp_association *asoc = NULL;
4917 struct sctp_sock *sp = sctp_sk(sk);
4919 if (len < sizeof(struct sctp_paddrparams))
4921 len = sizeof(struct sctp_paddrparams);
4922 if (copy_from_user(¶ms, optval, len))
4925 /* If an address other than INADDR_ANY is specified, and
4926 * no transport is found, then the request is invalid.
4928 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
4929 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4930 params.spp_assoc_id);
4932 pr_debug("%s: failed no transport\n", __func__);
4937 /* Get association, if assoc_id != 0 and the socket is a one
4938 * to many style socket, and an association was not found, then
4939 * the id was invalid.
4941 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4942 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4943 pr_debug("%s: failed no association\n", __func__);
4948 /* Fetch transport values. */
4949 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4950 params.spp_pathmtu = trans->pathmtu;
4951 params.spp_pathmaxrxt = trans->pathmaxrxt;
4952 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4954 /*draft-11 doesn't say what to return in spp_flags*/
4955 params.spp_flags = trans->param_flags;
4957 /* Fetch association values. */
4958 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4959 params.spp_pathmtu = asoc->pathmtu;
4960 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4961 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4963 /*draft-11 doesn't say what to return in spp_flags*/
4964 params.spp_flags = asoc->param_flags;
4966 /* Fetch socket values. */
4967 params.spp_hbinterval = sp->hbinterval;
4968 params.spp_pathmtu = sp->pathmtu;
4969 params.spp_sackdelay = sp->sackdelay;
4970 params.spp_pathmaxrxt = sp->pathmaxrxt;
4972 /*draft-11 doesn't say what to return in spp_flags*/
4973 params.spp_flags = sp->param_flags;
4976 if (copy_to_user(optval, ¶ms, len))
4979 if (put_user(len, optlen))
4986 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4988 * This option will effect the way delayed acks are performed. This
4989 * option allows you to get or set the delayed ack time, in
4990 * milliseconds. It also allows changing the delayed ack frequency.
4991 * Changing the frequency to 1 disables the delayed sack algorithm. If
4992 * the assoc_id is 0, then this sets or gets the endpoints default
4993 * values. If the assoc_id field is non-zero, then the set or get
4994 * effects the specified association for the one to many model (the
4995 * assoc_id field is ignored by the one to one model). Note that if
4996 * sack_delay or sack_freq are 0 when setting this option, then the
4997 * current values will remain unchanged.
4999 * struct sctp_sack_info {
5000 * sctp_assoc_t sack_assoc_id;
5001 * uint32_t sack_delay;
5002 * uint32_t sack_freq;
5005 * sack_assoc_id - This parameter, indicates which association the user
5006 * is performing an action upon. Note that if this field's value is
5007 * zero then the endpoints default value is changed (effecting future
5008 * associations only).
5010 * sack_delay - This parameter contains the number of milliseconds that
5011 * the user is requesting the delayed ACK timer be set to. Note that
5012 * this value is defined in the standard to be between 200 and 500
5015 * sack_freq - This parameter contains the number of packets that must
5016 * be received before a sack is sent without waiting for the delay
5017 * timer to expire. The default value for this is 2, setting this
5018 * value to 1 will disable the delayed sack algorithm.
5020 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
5021 char __user *optval,
5024 struct sctp_sack_info params;
5025 struct sctp_association *asoc = NULL;
5026 struct sctp_sock *sp = sctp_sk(sk);
5028 if (len >= sizeof(struct sctp_sack_info)) {
5029 len = sizeof(struct sctp_sack_info);
5031 if (copy_from_user(¶ms, optval, len))
5033 } else if (len == sizeof(struct sctp_assoc_value)) {
5034 pr_warn_ratelimited(DEPRECATED
5036 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
5037 "Use struct sctp_sack_info instead\n",
5038 current->comm, task_pid_nr(current));
5039 if (copy_from_user(¶ms, optval, len))
5044 /* Get association, if sack_assoc_id != 0 and the socket is a one
5045 * to many style socket, and an association was not found, then
5046 * the id was invalid.
5048 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
5049 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
5053 /* Fetch association values. */
5054 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
5055 params.sack_delay = jiffies_to_msecs(
5057 params.sack_freq = asoc->sackfreq;
5060 params.sack_delay = 0;
5061 params.sack_freq = 1;
5064 /* Fetch socket values. */
5065 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
5066 params.sack_delay = sp->sackdelay;
5067 params.sack_freq = sp->sackfreq;
5069 params.sack_delay = 0;
5070 params.sack_freq = 1;
5074 if (copy_to_user(optval, ¶ms, len))
5077 if (put_user(len, optlen))
5083 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5085 * Applications can specify protocol parameters for the default association
5086 * initialization. The option name argument to setsockopt() and getsockopt()
5089 * Setting initialization parameters is effective only on an unconnected
5090 * socket (for UDP-style sockets only future associations are effected
5091 * by the change). With TCP-style sockets, this option is inherited by
5092 * sockets derived from a listener socket.
5094 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
5096 if (len < sizeof(struct sctp_initmsg))
5098 len = sizeof(struct sctp_initmsg);
5099 if (put_user(len, optlen))
5101 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
5107 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
5108 char __user *optval, int __user *optlen)
5110 struct sctp_association *asoc;
5112 struct sctp_getaddrs getaddrs;
5113 struct sctp_transport *from;
5115 union sctp_addr temp;
5116 struct sctp_sock *sp = sctp_sk(sk);
5121 if (len < sizeof(struct sctp_getaddrs))
5124 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5127 /* For UDP-style sockets, id specifies the association to query. */
5128 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5132 to = optval + offsetof(struct sctp_getaddrs, addrs);
5133 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5135 list_for_each_entry(from, &asoc->peer.transport_addr_list,
5137 memcpy(&temp, &from->ipaddr, sizeof(temp));
5138 addrlen = sctp_get_pf_specific(sk->sk_family)
5139 ->addr_to_user(sp, &temp);
5140 if (space_left < addrlen)
5142 if (copy_to_user(to, &temp, addrlen))
5146 space_left -= addrlen;
5149 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
5151 bytes_copied = ((char __user *)to) - optval;
5152 if (put_user(bytes_copied, optlen))
5158 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
5159 size_t space_left, int *bytes_copied)
5161 struct sctp_sockaddr_entry *addr;
5162 union sctp_addr temp;
5165 struct net *net = sock_net(sk);
5168 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
5172 if ((PF_INET == sk->sk_family) &&
5173 (AF_INET6 == addr->a.sa.sa_family))
5175 if ((PF_INET6 == sk->sk_family) &&
5176 inet_v6_ipv6only(sk) &&
5177 (AF_INET == addr->a.sa.sa_family))
5179 memcpy(&temp, &addr->a, sizeof(temp));
5180 if (!temp.v4.sin_port)
5181 temp.v4.sin_port = htons(port);
5183 addrlen = sctp_get_pf_specific(sk->sk_family)
5184 ->addr_to_user(sctp_sk(sk), &temp);
5186 if (space_left < addrlen) {
5190 memcpy(to, &temp, addrlen);
5194 space_left -= addrlen;
5195 *bytes_copied += addrlen;
5203 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
5204 char __user *optval, int __user *optlen)
5206 struct sctp_bind_addr *bp;
5207 struct sctp_association *asoc;
5209 struct sctp_getaddrs getaddrs;
5210 struct sctp_sockaddr_entry *addr;
5212 union sctp_addr temp;
5213 struct sctp_sock *sp = sctp_sk(sk);
5217 int bytes_copied = 0;
5221 if (len < sizeof(struct sctp_getaddrs))
5224 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5228 * For UDP-style sockets, id specifies the association to query.
5229 * If the id field is set to the value '0' then the locally bound
5230 * addresses are returned without regard to any particular
5233 if (0 == getaddrs.assoc_id) {
5234 bp = &sctp_sk(sk)->ep->base.bind_addr;
5236 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5239 bp = &asoc->base.bind_addr;
5242 to = optval + offsetof(struct sctp_getaddrs, addrs);
5243 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5245 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN);
5249 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5250 * addresses from the global local address list.
5252 if (sctp_list_single_entry(&bp->address_list)) {
5253 addr = list_entry(bp->address_list.next,
5254 struct sctp_sockaddr_entry, list);
5255 if (sctp_is_any(sk, &addr->a)) {
5256 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
5257 space_left, &bytes_copied);
5267 /* Protection on the bound address list is not needed since
5268 * in the socket option context we hold a socket lock and
5269 * thus the bound address list can't change.
5271 list_for_each_entry(addr, &bp->address_list, list) {
5272 memcpy(&temp, &addr->a, sizeof(temp));
5273 addrlen = sctp_get_pf_specific(sk->sk_family)
5274 ->addr_to_user(sp, &temp);
5275 if (space_left < addrlen) {
5276 err = -ENOMEM; /*fixme: right error?*/
5279 memcpy(buf, &temp, addrlen);
5281 bytes_copied += addrlen;
5283 space_left -= addrlen;
5287 if (copy_to_user(to, addrs, bytes_copied)) {
5291 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
5295 if (put_user(bytes_copied, optlen))
5302 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
5304 * Requests that the local SCTP stack use the enclosed peer address as
5305 * the association primary. The enclosed address must be one of the
5306 * association peer's addresses.
5308 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
5309 char __user *optval, int __user *optlen)
5311 struct sctp_prim prim;
5312 struct sctp_association *asoc;
5313 struct sctp_sock *sp = sctp_sk(sk);
5315 if (len < sizeof(struct sctp_prim))
5318 len = sizeof(struct sctp_prim);
5320 if (copy_from_user(&prim, optval, len))
5323 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
5327 if (!asoc->peer.primary_path)
5330 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
5331 asoc->peer.primary_path->af_specific->sockaddr_len);
5333 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
5334 (union sctp_addr *)&prim.ssp_addr);
5336 if (put_user(len, optlen))
5338 if (copy_to_user(optval, &prim, len))
5345 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5347 * Requests that the local endpoint set the specified Adaptation Layer
5348 * Indication parameter for all future INIT and INIT-ACK exchanges.
5350 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
5351 char __user *optval, int __user *optlen)
5353 struct sctp_setadaptation adaptation;
5355 if (len < sizeof(struct sctp_setadaptation))
5358 len = sizeof(struct sctp_setadaptation);
5360 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
5362 if (put_user(len, optlen))
5364 if (copy_to_user(optval, &adaptation, len))
5372 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5374 * Applications that wish to use the sendto() system call may wish to
5375 * specify a default set of parameters that would normally be supplied
5376 * through the inclusion of ancillary data. This socket option allows
5377 * such an application to set the default sctp_sndrcvinfo structure.
5380 * The application that wishes to use this socket option simply passes
5381 * in to this call the sctp_sndrcvinfo structure defined in Section
5382 * 5.2.2) The input parameters accepted by this call include
5383 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5384 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5385 * to this call if the caller is using the UDP model.
5387 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5389 static int sctp_getsockopt_default_send_param(struct sock *sk,
5390 int len, char __user *optval,
5393 struct sctp_sock *sp = sctp_sk(sk);
5394 struct sctp_association *asoc;
5395 struct sctp_sndrcvinfo info;
5397 if (len < sizeof(info))
5402 if (copy_from_user(&info, optval, len))
5405 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
5406 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
5409 info.sinfo_stream = asoc->default_stream;
5410 info.sinfo_flags = asoc->default_flags;
5411 info.sinfo_ppid = asoc->default_ppid;
5412 info.sinfo_context = asoc->default_context;
5413 info.sinfo_timetolive = asoc->default_timetolive;
5415 info.sinfo_stream = sp->default_stream;
5416 info.sinfo_flags = sp->default_flags;
5417 info.sinfo_ppid = sp->default_ppid;
5418 info.sinfo_context = sp->default_context;
5419 info.sinfo_timetolive = sp->default_timetolive;
5422 if (put_user(len, optlen))
5424 if (copy_to_user(optval, &info, len))
5430 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5431 * (SCTP_DEFAULT_SNDINFO)
5433 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
5434 char __user *optval,
5437 struct sctp_sock *sp = sctp_sk(sk);
5438 struct sctp_association *asoc;
5439 struct sctp_sndinfo info;
5441 if (len < sizeof(info))
5446 if (copy_from_user(&info, optval, len))
5449 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
5450 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
5453 info.snd_sid = asoc->default_stream;
5454 info.snd_flags = asoc->default_flags;
5455 info.snd_ppid = asoc->default_ppid;
5456 info.snd_context = asoc->default_context;
5458 info.snd_sid = sp->default_stream;
5459 info.snd_flags = sp->default_flags;
5460 info.snd_ppid = sp->default_ppid;
5461 info.snd_context = sp->default_context;
5464 if (put_user(len, optlen))
5466 if (copy_to_user(optval, &info, len))
5474 * 7.1.5 SCTP_NODELAY
5476 * Turn on/off any Nagle-like algorithm. This means that packets are
5477 * generally sent as soon as possible and no unnecessary delays are
5478 * introduced, at the cost of more packets in the network. Expects an
5479 * integer boolean flag.
5482 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5483 char __user *optval, int __user *optlen)
5487 if (len < sizeof(int))
5491 val = (sctp_sk(sk)->nodelay == 1);
5492 if (put_user(len, optlen))
5494 if (copy_to_user(optval, &val, len))
5501 * 7.1.1 SCTP_RTOINFO
5503 * The protocol parameters used to initialize and bound retransmission
5504 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5505 * and modify these parameters.
5506 * All parameters are time values, in milliseconds. A value of 0, when
5507 * modifying the parameters, indicates that the current value should not
5511 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5512 char __user *optval,
5513 int __user *optlen) {
5514 struct sctp_rtoinfo rtoinfo;
5515 struct sctp_association *asoc;
5517 if (len < sizeof (struct sctp_rtoinfo))
5520 len = sizeof(struct sctp_rtoinfo);
5522 if (copy_from_user(&rtoinfo, optval, len))
5525 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5527 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5530 /* Values corresponding to the specific association. */
5532 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5533 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5534 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5536 /* Values corresponding to the endpoint. */
5537 struct sctp_sock *sp = sctp_sk(sk);
5539 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5540 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5541 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5544 if (put_user(len, optlen))
5547 if (copy_to_user(optval, &rtoinfo, len))
5555 * 7.1.2 SCTP_ASSOCINFO
5557 * This option is used to tune the maximum retransmission attempts
5558 * of the association.
5559 * Returns an error if the new association retransmission value is
5560 * greater than the sum of the retransmission value of the peer.
5561 * See [SCTP] for more information.
5564 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5565 char __user *optval,
5569 struct sctp_assocparams assocparams;
5570 struct sctp_association *asoc;
5571 struct list_head *pos;
5574 if (len < sizeof (struct sctp_assocparams))
5577 len = sizeof(struct sctp_assocparams);
5579 if (copy_from_user(&assocparams, optval, len))
5582 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5584 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5587 /* Values correspoinding to the specific association */
5589 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5590 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5591 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5592 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5594 list_for_each(pos, &asoc->peer.transport_addr_list) {
5598 assocparams.sasoc_number_peer_destinations = cnt;
5600 /* Values corresponding to the endpoint */
5601 struct sctp_sock *sp = sctp_sk(sk);
5603 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5604 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5605 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5606 assocparams.sasoc_cookie_life =
5607 sp->assocparams.sasoc_cookie_life;
5608 assocparams.sasoc_number_peer_destinations =
5610 sasoc_number_peer_destinations;
5613 if (put_user(len, optlen))
5616 if (copy_to_user(optval, &assocparams, len))
5623 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5625 * This socket option is a boolean flag which turns on or off mapped V4
5626 * addresses. If this option is turned on and the socket is type
5627 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5628 * If this option is turned off, then no mapping will be done of V4
5629 * addresses and a user will receive both PF_INET6 and PF_INET type
5630 * addresses on the socket.
5632 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5633 char __user *optval, int __user *optlen)
5636 struct sctp_sock *sp = sctp_sk(sk);
5638 if (len < sizeof(int))
5643 if (put_user(len, optlen))
5645 if (copy_to_user(optval, &val, len))
5652 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5653 * (chapter and verse is quoted at sctp_setsockopt_context())
5655 static int sctp_getsockopt_context(struct sock *sk, int len,
5656 char __user *optval, int __user *optlen)
5658 struct sctp_assoc_value params;
5659 struct sctp_sock *sp;
5660 struct sctp_association *asoc;
5662 if (len < sizeof(struct sctp_assoc_value))
5665 len = sizeof(struct sctp_assoc_value);
5667 if (copy_from_user(¶ms, optval, len))
5672 if (params.assoc_id != 0) {
5673 asoc = sctp_id2assoc(sk, params.assoc_id);
5676 params.assoc_value = asoc->default_rcv_context;
5678 params.assoc_value = sp->default_rcv_context;
5681 if (put_user(len, optlen))
5683 if (copy_to_user(optval, ¶ms, len))
5690 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5691 * This option will get or set the maximum size to put in any outgoing
5692 * SCTP DATA chunk. If a message is larger than this size it will be
5693 * fragmented by SCTP into the specified size. Note that the underlying
5694 * SCTP implementation may fragment into smaller sized chunks when the
5695 * PMTU of the underlying association is smaller than the value set by
5696 * the user. The default value for this option is '0' which indicates
5697 * the user is NOT limiting fragmentation and only the PMTU will effect
5698 * SCTP's choice of DATA chunk size. Note also that values set larger
5699 * than the maximum size of an IP datagram will effectively let SCTP
5700 * control fragmentation (i.e. the same as setting this option to 0).
5702 * The following structure is used to access and modify this parameter:
5704 * struct sctp_assoc_value {
5705 * sctp_assoc_t assoc_id;
5706 * uint32_t assoc_value;
5709 * assoc_id: This parameter is ignored for one-to-one style sockets.
5710 * For one-to-many style sockets this parameter indicates which
5711 * association the user is performing an action upon. Note that if
5712 * this field's value is zero then the endpoints default value is
5713 * changed (effecting future associations only).
5714 * assoc_value: This parameter specifies the maximum size in bytes.
5716 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5717 char __user *optval, int __user *optlen)
5719 struct sctp_assoc_value params;
5720 struct sctp_association *asoc;
5722 if (len == sizeof(int)) {
5723 pr_warn_ratelimited(DEPRECATED
5725 "Use of int in maxseg socket option.\n"
5726 "Use struct sctp_assoc_value instead\n",
5727 current->comm, task_pid_nr(current));
5728 params.assoc_id = 0;
5729 } else if (len >= sizeof(struct sctp_assoc_value)) {
5730 len = sizeof(struct sctp_assoc_value);
5731 if (copy_from_user(¶ms, optval, sizeof(params)))
5736 asoc = sctp_id2assoc(sk, params.assoc_id);
5737 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5741 params.assoc_value = asoc->frag_point;
5743 params.assoc_value = sctp_sk(sk)->user_frag;
5745 if (put_user(len, optlen))
5747 if (len == sizeof(int)) {
5748 if (copy_to_user(optval, ¶ms.assoc_value, len))
5751 if (copy_to_user(optval, ¶ms, len))
5759 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5760 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5762 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5763 char __user *optval, int __user *optlen)
5767 if (len < sizeof(int))
5772 val = sctp_sk(sk)->frag_interleave;
5773 if (put_user(len, optlen))
5775 if (copy_to_user(optval, &val, len))
5782 * 7.1.25. Set or Get the sctp partial delivery point
5783 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5785 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5786 char __user *optval,
5791 if (len < sizeof(u32))
5796 val = sctp_sk(sk)->pd_point;
5797 if (put_user(len, optlen))
5799 if (copy_to_user(optval, &val, len))
5806 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5807 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5809 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5810 char __user *optval,
5813 struct sctp_assoc_value params;
5814 struct sctp_sock *sp;
5815 struct sctp_association *asoc;
5817 if (len == sizeof(int)) {
5818 pr_warn_ratelimited(DEPRECATED
5820 "Use of int in max_burst socket option.\n"
5821 "Use struct sctp_assoc_value instead\n",
5822 current->comm, task_pid_nr(current));
5823 params.assoc_id = 0;
5824 } else if (len >= sizeof(struct sctp_assoc_value)) {
5825 len = sizeof(struct sctp_assoc_value);
5826 if (copy_from_user(¶ms, optval, len))
5833 if (params.assoc_id != 0) {
5834 asoc = sctp_id2assoc(sk, params.assoc_id);
5837 params.assoc_value = asoc->max_burst;
5839 params.assoc_value = sp->max_burst;
5841 if (len == sizeof(int)) {
5842 if (copy_to_user(optval, ¶ms.assoc_value, len))
5845 if (copy_to_user(optval, ¶ms, len))
5853 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5854 char __user *optval, int __user *optlen)
5856 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5857 struct sctp_hmacalgo __user *p = (void __user *)optval;
5858 struct sctp_hmac_algo_param *hmacs;
5863 if (!ep->auth_enable)
5866 hmacs = ep->auth_hmacs_list;
5867 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5869 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5872 len = sizeof(struct sctp_hmacalgo) + data_len;
5873 num_idents = data_len / sizeof(u16);
5875 if (put_user(len, optlen))
5877 if (put_user(num_idents, &p->shmac_num_idents))
5879 for (i = 0; i < num_idents; i++) {
5880 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
5882 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
5888 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5889 char __user *optval, int __user *optlen)
5891 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5892 struct sctp_authkeyid val;
5893 struct sctp_association *asoc;
5895 if (!ep->auth_enable)
5898 if (len < sizeof(struct sctp_authkeyid))
5900 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5903 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5904 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5908 val.scact_keynumber = asoc->active_key_id;
5910 val.scact_keynumber = ep->active_key_id;
5912 len = sizeof(struct sctp_authkeyid);
5913 if (put_user(len, optlen))
5915 if (copy_to_user(optval, &val, len))
5921 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5922 char __user *optval, int __user *optlen)
5924 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5925 struct sctp_authchunks __user *p = (void __user *)optval;
5926 struct sctp_authchunks val;
5927 struct sctp_association *asoc;
5928 struct sctp_chunks_param *ch;
5932 if (!ep->auth_enable)
5935 if (len < sizeof(struct sctp_authchunks))
5938 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5941 to = p->gauth_chunks;
5942 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5946 ch = asoc->peer.peer_chunks;
5950 /* See if the user provided enough room for all the data */
5951 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5952 if (len < num_chunks)
5955 if (copy_to_user(to, ch->chunks, num_chunks))
5958 len = sizeof(struct sctp_authchunks) + num_chunks;
5959 if (put_user(len, optlen))
5961 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5966 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5967 char __user *optval, int __user *optlen)
5969 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5970 struct sctp_authchunks __user *p = (void __user *)optval;
5971 struct sctp_authchunks val;
5972 struct sctp_association *asoc;
5973 struct sctp_chunks_param *ch;
5977 if (!ep->auth_enable)
5980 if (len < sizeof(struct sctp_authchunks))
5983 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5986 to = p->gauth_chunks;
5987 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5988 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5992 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
5994 ch = ep->auth_chunk_list;
5999 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
6000 if (len < sizeof(struct sctp_authchunks) + num_chunks)
6003 if (copy_to_user(to, ch->chunks, num_chunks))
6006 len = sizeof(struct sctp_authchunks) + num_chunks;
6007 if (put_user(len, optlen))
6009 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6016 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
6017 * This option gets the current number of associations that are attached
6018 * to a one-to-many style socket. The option value is an uint32_t.
6020 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
6021 char __user *optval, int __user *optlen)
6023 struct sctp_sock *sp = sctp_sk(sk);
6024 struct sctp_association *asoc;
6027 if (sctp_style(sk, TCP))
6030 if (len < sizeof(u32))
6035 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6039 if (put_user(len, optlen))
6041 if (copy_to_user(optval, &val, len))
6048 * 8.1.23 SCTP_AUTO_ASCONF
6049 * See the corresponding setsockopt entry as description
6051 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
6052 char __user *optval, int __user *optlen)
6056 if (len < sizeof(int))
6060 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
6062 if (put_user(len, optlen))
6064 if (copy_to_user(optval, &val, len))
6070 * 8.2.6. Get the Current Identifiers of Associations
6071 * (SCTP_GET_ASSOC_ID_LIST)
6073 * This option gets the current list of SCTP association identifiers of
6074 * the SCTP associations handled by a one-to-many style socket.
6076 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
6077 char __user *optval, int __user *optlen)
6079 struct sctp_sock *sp = sctp_sk(sk);
6080 struct sctp_association *asoc;
6081 struct sctp_assoc_ids *ids;
6084 if (sctp_style(sk, TCP))
6087 if (len < sizeof(struct sctp_assoc_ids))
6090 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6094 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
6097 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
6099 ids = kmalloc(len, GFP_USER | __GFP_NOWARN);
6103 ids->gaids_number_of_ids = num;
6105 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6106 ids->gaids_assoc_id[num++] = asoc->assoc_id;
6109 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
6119 * SCTP_PEER_ADDR_THLDS
6121 * This option allows us to fetch the partially failed threshold for one or all
6122 * transports in an association. See Section 6.1 of:
6123 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6125 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
6126 char __user *optval,
6130 struct sctp_paddrthlds val;
6131 struct sctp_transport *trans;
6132 struct sctp_association *asoc;
6134 if (len < sizeof(struct sctp_paddrthlds))
6136 len = sizeof(struct sctp_paddrthlds);
6137 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
6140 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
6141 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
6145 val.spt_pathpfthld = asoc->pf_retrans;
6146 val.spt_pathmaxrxt = asoc->pathmaxrxt;
6148 trans = sctp_addr_id2transport(sk, &val.spt_address,
6153 val.spt_pathmaxrxt = trans->pathmaxrxt;
6154 val.spt_pathpfthld = trans->pf_retrans;
6157 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
6164 * SCTP_GET_ASSOC_STATS
6166 * This option retrieves local per endpoint statistics. It is modeled
6167 * after OpenSolaris' implementation
6169 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
6170 char __user *optval,
6173 struct sctp_assoc_stats sas;
6174 struct sctp_association *asoc = NULL;
6176 /* User must provide at least the assoc id */
6177 if (len < sizeof(sctp_assoc_t))
6180 /* Allow the struct to grow and fill in as much as possible */
6181 len = min_t(size_t, len, sizeof(sas));
6183 if (copy_from_user(&sas, optval, len))
6186 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
6190 sas.sas_rtxchunks = asoc->stats.rtxchunks;
6191 sas.sas_gapcnt = asoc->stats.gapcnt;
6192 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
6193 sas.sas_osacks = asoc->stats.osacks;
6194 sas.sas_isacks = asoc->stats.isacks;
6195 sas.sas_octrlchunks = asoc->stats.octrlchunks;
6196 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
6197 sas.sas_oodchunks = asoc->stats.oodchunks;
6198 sas.sas_iodchunks = asoc->stats.iodchunks;
6199 sas.sas_ouodchunks = asoc->stats.ouodchunks;
6200 sas.sas_iuodchunks = asoc->stats.iuodchunks;
6201 sas.sas_idupchunks = asoc->stats.idupchunks;
6202 sas.sas_opackets = asoc->stats.opackets;
6203 sas.sas_ipackets = asoc->stats.ipackets;
6205 /* New high max rto observed, will return 0 if not a single
6206 * RTO update took place. obs_rto_ipaddr will be bogus
6209 sas.sas_maxrto = asoc->stats.max_obs_rto;
6210 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
6211 sizeof(struct sockaddr_storage));
6213 /* Mark beginning of a new observation period */
6214 asoc->stats.max_obs_rto = asoc->rto_min;
6216 if (put_user(len, optlen))
6219 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
6221 if (copy_to_user(optval, &sas, len))
6227 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
6228 char __user *optval,
6233 if (len < sizeof(int))
6237 if (sctp_sk(sk)->recvrcvinfo)
6239 if (put_user(len, optlen))
6241 if (copy_to_user(optval, &val, len))
6247 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
6248 char __user *optval,
6253 if (len < sizeof(int))
6257 if (sctp_sk(sk)->recvnxtinfo)
6259 if (put_user(len, optlen))
6261 if (copy_to_user(optval, &val, len))
6267 static int sctp_getsockopt_pr_supported(struct sock *sk, int len,
6268 char __user *optval,
6271 struct sctp_assoc_value params;
6272 struct sctp_association *asoc;
6273 int retval = -EFAULT;
6275 if (len < sizeof(params)) {
6280 len = sizeof(params);
6281 if (copy_from_user(¶ms, optval, len))
6284 asoc = sctp_id2assoc(sk, params.assoc_id);
6286 params.assoc_value = asoc->prsctp_enable;
6287 } else if (!params.assoc_id) {
6288 struct sctp_sock *sp = sctp_sk(sk);
6290 params.assoc_value = sp->ep->prsctp_enable;
6296 if (put_user(len, optlen))
6299 if (copy_to_user(optval, ¶ms, len))
6308 static int sctp_getsockopt_default_prinfo(struct sock *sk, int len,
6309 char __user *optval,
6312 struct sctp_default_prinfo info;
6313 struct sctp_association *asoc;
6314 int retval = -EFAULT;
6316 if (len < sizeof(info)) {
6322 if (copy_from_user(&info, optval, len))
6325 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
6327 info.pr_policy = SCTP_PR_POLICY(asoc->default_flags);
6328 info.pr_value = asoc->default_timetolive;
6329 } else if (!info.pr_assoc_id) {
6330 struct sctp_sock *sp = sctp_sk(sk);
6332 info.pr_policy = SCTP_PR_POLICY(sp->default_flags);
6333 info.pr_value = sp->default_timetolive;
6339 if (put_user(len, optlen))
6342 if (copy_to_user(optval, &info, len))
6351 static int sctp_getsockopt_pr_assocstatus(struct sock *sk, int len,
6352 char __user *optval,
6355 struct sctp_prstatus params;
6356 struct sctp_association *asoc;
6358 int retval = -EINVAL;
6360 if (len < sizeof(params))
6363 len = sizeof(params);
6364 if (copy_from_user(¶ms, optval, len)) {
6369 policy = params.sprstat_policy;
6370 if (policy & ~SCTP_PR_SCTP_MASK)
6373 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
6377 if (policy == SCTP_PR_SCTP_NONE) {
6378 params.sprstat_abandoned_unsent = 0;
6379 params.sprstat_abandoned_sent = 0;
6380 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
6381 params.sprstat_abandoned_unsent +=
6382 asoc->abandoned_unsent[policy];
6383 params.sprstat_abandoned_sent +=
6384 asoc->abandoned_sent[policy];
6387 params.sprstat_abandoned_unsent =
6388 asoc->abandoned_unsent[__SCTP_PR_INDEX(policy)];
6389 params.sprstat_abandoned_sent =
6390 asoc->abandoned_sent[__SCTP_PR_INDEX(policy)];
6393 if (put_user(len, optlen)) {
6398 if (copy_to_user(optval, ¶ms, len)) {
6409 static int sctp_getsockopt(struct sock *sk, int level, int optname,
6410 char __user *optval, int __user *optlen)
6415 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
6417 /* I can hardly begin to describe how wrong this is. This is
6418 * so broken as to be worse than useless. The API draft
6419 * REALLY is NOT helpful here... I am not convinced that the
6420 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
6421 * are at all well-founded.
6423 if (level != SOL_SCTP) {
6424 struct sctp_af *af = sctp_sk(sk)->pf->af;
6426 retval = af->getsockopt(sk, level, optname, optval, optlen);
6430 if (get_user(len, optlen))
6437 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
6439 case SCTP_DISABLE_FRAGMENTS:
6440 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
6444 retval = sctp_getsockopt_events(sk, len, optval, optlen);
6446 case SCTP_AUTOCLOSE:
6447 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
6449 case SCTP_SOCKOPT_PEELOFF:
6450 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
6452 case SCTP_PEER_ADDR_PARAMS:
6453 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
6456 case SCTP_DELAYED_SACK:
6457 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
6461 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
6463 case SCTP_GET_PEER_ADDRS:
6464 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
6467 case SCTP_GET_LOCAL_ADDRS:
6468 retval = sctp_getsockopt_local_addrs(sk, len, optval,
6471 case SCTP_SOCKOPT_CONNECTX3:
6472 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
6474 case SCTP_DEFAULT_SEND_PARAM:
6475 retval = sctp_getsockopt_default_send_param(sk, len,
6478 case SCTP_DEFAULT_SNDINFO:
6479 retval = sctp_getsockopt_default_sndinfo(sk, len,
6482 case SCTP_PRIMARY_ADDR:
6483 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
6486 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
6489 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
6491 case SCTP_ASSOCINFO:
6492 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
6494 case SCTP_I_WANT_MAPPED_V4_ADDR:
6495 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
6498 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
6500 case SCTP_GET_PEER_ADDR_INFO:
6501 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
6504 case SCTP_ADAPTATION_LAYER:
6505 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
6509 retval = sctp_getsockopt_context(sk, len, optval, optlen);
6511 case SCTP_FRAGMENT_INTERLEAVE:
6512 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
6515 case SCTP_PARTIAL_DELIVERY_POINT:
6516 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
6519 case SCTP_MAX_BURST:
6520 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
6523 case SCTP_AUTH_CHUNK:
6524 case SCTP_AUTH_DELETE_KEY:
6525 retval = -EOPNOTSUPP;
6527 case SCTP_HMAC_IDENT:
6528 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
6530 case SCTP_AUTH_ACTIVE_KEY:
6531 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
6533 case SCTP_PEER_AUTH_CHUNKS:
6534 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
6537 case SCTP_LOCAL_AUTH_CHUNKS:
6538 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
6541 case SCTP_GET_ASSOC_NUMBER:
6542 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
6544 case SCTP_GET_ASSOC_ID_LIST:
6545 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
6547 case SCTP_AUTO_ASCONF:
6548 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
6550 case SCTP_PEER_ADDR_THLDS:
6551 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
6553 case SCTP_GET_ASSOC_STATS:
6554 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
6556 case SCTP_RECVRCVINFO:
6557 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
6559 case SCTP_RECVNXTINFO:
6560 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
6562 case SCTP_PR_SUPPORTED:
6563 retval = sctp_getsockopt_pr_supported(sk, len, optval, optlen);
6565 case SCTP_DEFAULT_PRINFO:
6566 retval = sctp_getsockopt_default_prinfo(sk, len, optval,
6569 case SCTP_PR_ASSOC_STATUS:
6570 retval = sctp_getsockopt_pr_assocstatus(sk, len, optval,
6574 retval = -ENOPROTOOPT;
6582 static int sctp_hash(struct sock *sk)
6588 static void sctp_unhash(struct sock *sk)
6593 /* Check if port is acceptable. Possibly find first available port.
6595 * The port hash table (contained in the 'global' SCTP protocol storage
6596 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
6597 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
6598 * list (the list number is the port number hashed out, so as you
6599 * would expect from a hash function, all the ports in a given list have
6600 * such a number that hashes out to the same list number; you were
6601 * expecting that, right?); so each list has a set of ports, with a
6602 * link to the socket (struct sock) that uses it, the port number and
6603 * a fastreuse flag (FIXME: NPI ipg).
6605 static struct sctp_bind_bucket *sctp_bucket_create(
6606 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
6608 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
6610 struct sctp_bind_hashbucket *head; /* hash list */
6611 struct sctp_bind_bucket *pp;
6612 unsigned short snum;
6615 snum = ntohs(addr->v4.sin_port);
6617 pr_debug("%s: begins, snum:%d\n", __func__, snum);
6622 /* Search for an available port. */
6623 int low, high, remaining, index;
6625 struct net *net = sock_net(sk);
6627 inet_get_local_port_range(net, &low, &high);
6628 remaining = (high - low) + 1;
6629 rover = prandom_u32() % remaining + low;
6633 if ((rover < low) || (rover > high))
6635 if (inet_is_local_reserved_port(net, rover))
6637 index = sctp_phashfn(sock_net(sk), rover);
6638 head = &sctp_port_hashtable[index];
6639 spin_lock(&head->lock);
6640 sctp_for_each_hentry(pp, &head->chain)
6641 if ((pp->port == rover) &&
6642 net_eq(sock_net(sk), pp->net))
6646 spin_unlock(&head->lock);
6647 } while (--remaining > 0);
6649 /* Exhausted local port range during search? */
6654 /* OK, here is the one we will use. HEAD (the port
6655 * hash table list entry) is non-NULL and we hold it's
6660 /* We are given an specific port number; we verify
6661 * that it is not being used. If it is used, we will
6662 * exahust the search in the hash list corresponding
6663 * to the port number (snum) - we detect that with the
6664 * port iterator, pp being NULL.
6666 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
6667 spin_lock(&head->lock);
6668 sctp_for_each_hentry(pp, &head->chain) {
6669 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
6676 if (!hlist_empty(&pp->owner)) {
6677 /* We had a port hash table hit - there is an
6678 * available port (pp != NULL) and it is being
6679 * used by other socket (pp->owner not empty); that other
6680 * socket is going to be sk2.
6682 int reuse = sk->sk_reuse;
6685 pr_debug("%s: found a possible match\n", __func__);
6687 if (pp->fastreuse && sk->sk_reuse &&
6688 sk->sk_state != SCTP_SS_LISTENING)
6691 /* Run through the list of sockets bound to the port
6692 * (pp->port) [via the pointers bind_next and
6693 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6694 * we get the endpoint they describe and run through
6695 * the endpoint's list of IP (v4 or v6) addresses,
6696 * comparing each of the addresses with the address of
6697 * the socket sk. If we find a match, then that means
6698 * that this port/socket (sk) combination are already
6701 sk_for_each_bound(sk2, &pp->owner) {
6702 struct sctp_endpoint *ep2;
6703 ep2 = sctp_sk(sk2)->ep;
6706 (reuse && sk2->sk_reuse &&
6707 sk2->sk_state != SCTP_SS_LISTENING))
6710 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
6711 sctp_sk(sk2), sctp_sk(sk))) {
6717 pr_debug("%s: found a match\n", __func__);
6720 /* If there was a hash table miss, create a new port. */
6722 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6725 /* In either case (hit or miss), make sure fastreuse is 1 only
6726 * if sk->sk_reuse is too (that is, if the caller requested
6727 * SO_REUSEADDR on this socket -sk-).
6729 if (hlist_empty(&pp->owner)) {
6730 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6734 } else if (pp->fastreuse &&
6735 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6738 /* We are set, so fill up all the data in the hash table
6739 * entry, tie the socket list information with the rest of the
6740 * sockets FIXME: Blurry, NPI (ipg).
6743 if (!sctp_sk(sk)->bind_hash) {
6744 inet_sk(sk)->inet_num = snum;
6745 sk_add_bind_node(sk, &pp->owner);
6746 sctp_sk(sk)->bind_hash = pp;
6751 spin_unlock(&head->lock);
6758 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6759 * port is requested.
6761 static int sctp_get_port(struct sock *sk, unsigned short snum)
6763 union sctp_addr addr;
6764 struct sctp_af *af = sctp_sk(sk)->pf->af;
6766 /* Set up a dummy address struct from the sk. */
6767 af->from_sk(&addr, sk);
6768 addr.v4.sin_port = htons(snum);
6770 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6771 return !!sctp_get_port_local(sk, &addr);
6775 * Move a socket to LISTENING state.
6777 static int sctp_listen_start(struct sock *sk, int backlog)
6779 struct sctp_sock *sp = sctp_sk(sk);
6780 struct sctp_endpoint *ep = sp->ep;
6781 struct crypto_shash *tfm = NULL;
6784 /* Allocate HMAC for generating cookie. */
6785 if (!sp->hmac && sp->sctp_hmac_alg) {
6786 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6787 tfm = crypto_alloc_shash(alg, 0, 0);
6789 net_info_ratelimited("failed to load transform for %s: %ld\n",
6790 sp->sctp_hmac_alg, PTR_ERR(tfm));
6793 sctp_sk(sk)->hmac = tfm;
6797 * If a bind() or sctp_bindx() is not called prior to a listen()
6798 * call that allows new associations to be accepted, the system
6799 * picks an ephemeral port and will choose an address set equivalent
6800 * to binding with a wildcard address.
6802 * This is not currently spelled out in the SCTP sockets
6803 * extensions draft, but follows the practice as seen in TCP
6807 sk->sk_state = SCTP_SS_LISTENING;
6808 if (!ep->base.bind_addr.port) {
6809 if (sctp_autobind(sk))
6812 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6813 sk->sk_state = SCTP_SS_CLOSED;
6818 sk->sk_max_ack_backlog = backlog;
6819 sctp_hash_endpoint(ep);
6824 * 4.1.3 / 5.1.3 listen()
6826 * By default, new associations are not accepted for UDP style sockets.
6827 * An application uses listen() to mark a socket as being able to
6828 * accept new associations.
6830 * On TCP style sockets, applications use listen() to ready the SCTP
6831 * endpoint for accepting inbound associations.
6833 * On both types of endpoints a backlog of '0' disables listening.
6835 * Move a socket to LISTENING state.
6837 int sctp_inet_listen(struct socket *sock, int backlog)
6839 struct sock *sk = sock->sk;
6840 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6843 if (unlikely(backlog < 0))
6848 /* Peeled-off sockets are not allowed to listen(). */
6849 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6852 if (sock->state != SS_UNCONNECTED)
6855 /* If backlog is zero, disable listening. */
6857 if (sctp_sstate(sk, CLOSED))
6861 sctp_unhash_endpoint(ep);
6862 sk->sk_state = SCTP_SS_CLOSED;
6864 sctp_sk(sk)->bind_hash->fastreuse = 1;
6868 /* If we are already listening, just update the backlog */
6869 if (sctp_sstate(sk, LISTENING))
6870 sk->sk_max_ack_backlog = backlog;
6872 err = sctp_listen_start(sk, backlog);
6884 * This function is done by modeling the current datagram_poll() and the
6885 * tcp_poll(). Note that, based on these implementations, we don't
6886 * lock the socket in this function, even though it seems that,
6887 * ideally, locking or some other mechanisms can be used to ensure
6888 * the integrity of the counters (sndbuf and wmem_alloc) used
6889 * in this place. We assume that we don't need locks either until proven
6892 * Another thing to note is that we include the Async I/O support
6893 * here, again, by modeling the current TCP/UDP code. We don't have
6894 * a good way to test with it yet.
6896 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6898 struct sock *sk = sock->sk;
6899 struct sctp_sock *sp = sctp_sk(sk);
6902 poll_wait(file, sk_sleep(sk), wait);
6904 sock_rps_record_flow(sk);
6906 /* A TCP-style listening socket becomes readable when the accept queue
6909 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6910 return (!list_empty(&sp->ep->asocs)) ?
6911 (POLLIN | POLLRDNORM) : 0;
6915 /* Is there any exceptional events? */
6916 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6918 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6919 if (sk->sk_shutdown & RCV_SHUTDOWN)
6920 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6921 if (sk->sk_shutdown == SHUTDOWN_MASK)
6924 /* Is it readable? Reconsider this code with TCP-style support. */
6925 if (!skb_queue_empty(&sk->sk_receive_queue))
6926 mask |= POLLIN | POLLRDNORM;
6928 /* The association is either gone or not ready. */
6929 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6932 /* Is it writable? */
6933 if (sctp_writeable(sk)) {
6934 mask |= POLLOUT | POLLWRNORM;
6936 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
6938 * Since the socket is not locked, the buffer
6939 * might be made available after the writeable check and
6940 * before the bit is set. This could cause a lost I/O
6941 * signal. tcp_poll() has a race breaker for this race
6942 * condition. Based on their implementation, we put
6943 * in the following code to cover it as well.
6945 if (sctp_writeable(sk))
6946 mask |= POLLOUT | POLLWRNORM;
6951 /********************************************************************
6952 * 2nd Level Abstractions
6953 ********************************************************************/
6955 static struct sctp_bind_bucket *sctp_bucket_create(
6956 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6958 struct sctp_bind_bucket *pp;
6960 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6962 SCTP_DBG_OBJCNT_INC(bind_bucket);
6965 INIT_HLIST_HEAD(&pp->owner);
6967 hlist_add_head(&pp->node, &head->chain);
6972 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6973 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6975 if (pp && hlist_empty(&pp->owner)) {
6976 __hlist_del(&pp->node);
6977 kmem_cache_free(sctp_bucket_cachep, pp);
6978 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6982 /* Release this socket's reference to a local port. */
6983 static inline void __sctp_put_port(struct sock *sk)
6985 struct sctp_bind_hashbucket *head =
6986 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6987 inet_sk(sk)->inet_num)];
6988 struct sctp_bind_bucket *pp;
6990 spin_lock(&head->lock);
6991 pp = sctp_sk(sk)->bind_hash;
6992 __sk_del_bind_node(sk);
6993 sctp_sk(sk)->bind_hash = NULL;
6994 inet_sk(sk)->inet_num = 0;
6995 sctp_bucket_destroy(pp);
6996 spin_unlock(&head->lock);
6999 void sctp_put_port(struct sock *sk)
7002 __sctp_put_port(sk);
7007 * The system picks an ephemeral port and choose an address set equivalent
7008 * to binding with a wildcard address.
7009 * One of those addresses will be the primary address for the association.
7010 * This automatically enables the multihoming capability of SCTP.
7012 static int sctp_autobind(struct sock *sk)
7014 union sctp_addr autoaddr;
7018 /* Initialize a local sockaddr structure to INADDR_ANY. */
7019 af = sctp_sk(sk)->pf->af;
7021 port = htons(inet_sk(sk)->inet_num);
7022 af->inaddr_any(&autoaddr, port);
7024 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
7027 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
7030 * 4.2 The cmsghdr Structure *
7032 * When ancillary data is sent or received, any number of ancillary data
7033 * objects can be specified by the msg_control and msg_controllen members of
7034 * the msghdr structure, because each object is preceded by
7035 * a cmsghdr structure defining the object's length (the cmsg_len member).
7036 * Historically Berkeley-derived implementations have passed only one object
7037 * at a time, but this API allows multiple objects to be
7038 * passed in a single call to sendmsg() or recvmsg(). The following example
7039 * shows two ancillary data objects in a control buffer.
7041 * |<--------------------------- msg_controllen -------------------------->|
7044 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
7046 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
7049 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
7051 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
7054 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7055 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
7057 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
7059 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7066 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
7068 struct cmsghdr *cmsg;
7069 struct msghdr *my_msg = (struct msghdr *)msg;
7071 for_each_cmsghdr(cmsg, my_msg) {
7072 if (!CMSG_OK(my_msg, cmsg))
7075 /* Should we parse this header or ignore? */
7076 if (cmsg->cmsg_level != IPPROTO_SCTP)
7079 /* Strictly check lengths following example in SCM code. */
7080 switch (cmsg->cmsg_type) {
7082 /* SCTP Socket API Extension
7083 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
7085 * This cmsghdr structure provides information for
7086 * initializing new SCTP associations with sendmsg().
7087 * The SCTP_INITMSG socket option uses this same data
7088 * structure. This structure is not used for
7091 * cmsg_level cmsg_type cmsg_data[]
7092 * ------------ ------------ ----------------------
7093 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
7095 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
7098 cmsgs->init = CMSG_DATA(cmsg);
7102 /* SCTP Socket API Extension
7103 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
7105 * This cmsghdr structure specifies SCTP options for
7106 * sendmsg() and describes SCTP header information
7107 * about a received message through recvmsg().
7109 * cmsg_level cmsg_type cmsg_data[]
7110 * ------------ ------------ ----------------------
7111 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
7113 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
7116 cmsgs->srinfo = CMSG_DATA(cmsg);
7118 if (cmsgs->srinfo->sinfo_flags &
7119 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7120 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7121 SCTP_ABORT | SCTP_EOF))
7126 /* SCTP Socket API Extension
7127 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
7129 * This cmsghdr structure specifies SCTP options for
7130 * sendmsg(). This structure and SCTP_RCVINFO replaces
7131 * SCTP_SNDRCV which has been deprecated.
7133 * cmsg_level cmsg_type cmsg_data[]
7134 * ------------ ------------ ---------------------
7135 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
7137 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
7140 cmsgs->sinfo = CMSG_DATA(cmsg);
7142 if (cmsgs->sinfo->snd_flags &
7143 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7144 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7145 SCTP_ABORT | SCTP_EOF))
7157 * Wait for a packet..
7158 * Note: This function is the same function as in core/datagram.c
7159 * with a few modifications to make lksctp work.
7161 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
7166 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7168 /* Socket errors? */
7169 error = sock_error(sk);
7173 if (!skb_queue_empty(&sk->sk_receive_queue))
7176 /* Socket shut down? */
7177 if (sk->sk_shutdown & RCV_SHUTDOWN)
7180 /* Sequenced packets can come disconnected. If so we report the
7185 /* Is there a good reason to think that we may receive some data? */
7186 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
7189 /* Handle signals. */
7190 if (signal_pending(current))
7193 /* Let another process have a go. Since we are going to sleep
7194 * anyway. Note: This may cause odd behaviors if the message
7195 * does not fit in the user's buffer, but this seems to be the
7196 * only way to honor MSG_DONTWAIT realistically.
7199 *timeo_p = schedule_timeout(*timeo_p);
7203 finish_wait(sk_sleep(sk), &wait);
7207 error = sock_intr_errno(*timeo_p);
7210 finish_wait(sk_sleep(sk), &wait);
7215 /* Receive a datagram.
7216 * Note: This is pretty much the same routine as in core/datagram.c
7217 * with a few changes to make lksctp work.
7219 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
7220 int noblock, int *err)
7223 struct sk_buff *skb;
7226 timeo = sock_rcvtimeo(sk, noblock);
7228 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
7229 MAX_SCHEDULE_TIMEOUT);
7232 /* Again only user level code calls this function,
7233 * so nothing interrupt level
7234 * will suddenly eat the receive_queue.
7236 * Look at current nfs client by the way...
7237 * However, this function was correct in any case. 8)
7239 if (flags & MSG_PEEK) {
7240 skb = skb_peek(&sk->sk_receive_queue);
7242 atomic_inc(&skb->users);
7244 skb = __skb_dequeue(&sk->sk_receive_queue);
7250 /* Caller is allowed not to check sk->sk_err before calling. */
7251 error = sock_error(sk);
7255 if (sk->sk_shutdown & RCV_SHUTDOWN)
7258 if (sk_can_busy_loop(sk) &&
7259 sk_busy_loop(sk, noblock))
7262 /* User doesn't want to wait. */
7266 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
7275 /* If sndbuf has changed, wake up per association sndbuf waiters. */
7276 static void __sctp_write_space(struct sctp_association *asoc)
7278 struct sock *sk = asoc->base.sk;
7280 if (sctp_wspace(asoc) <= 0)
7283 if (waitqueue_active(&asoc->wait))
7284 wake_up_interruptible(&asoc->wait);
7286 if (sctp_writeable(sk)) {
7287 struct socket_wq *wq;
7290 wq = rcu_dereference(sk->sk_wq);
7292 if (waitqueue_active(&wq->wait))
7293 wake_up_interruptible(&wq->wait);
7295 /* Note that we try to include the Async I/O support
7296 * here by modeling from the current TCP/UDP code.
7297 * We have not tested with it yet.
7299 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
7300 sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT);
7306 static void sctp_wake_up_waiters(struct sock *sk,
7307 struct sctp_association *asoc)
7309 struct sctp_association *tmp = asoc;
7311 /* We do accounting for the sndbuf space per association,
7312 * so we only need to wake our own association.
7314 if (asoc->ep->sndbuf_policy)
7315 return __sctp_write_space(asoc);
7317 /* If association goes down and is just flushing its
7318 * outq, then just normally notify others.
7320 if (asoc->base.dead)
7321 return sctp_write_space(sk);
7323 /* Accounting for the sndbuf space is per socket, so we
7324 * need to wake up others, try to be fair and in case of
7325 * other associations, let them have a go first instead
7326 * of just doing a sctp_write_space() call.
7328 * Note that we reach sctp_wake_up_waiters() only when
7329 * associations free up queued chunks, thus we are under
7330 * lock and the list of associations on a socket is
7331 * guaranteed not to change.
7333 for (tmp = list_next_entry(tmp, asocs); 1;
7334 tmp = list_next_entry(tmp, asocs)) {
7335 /* Manually skip the head element. */
7336 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
7338 /* Wake up association. */
7339 __sctp_write_space(tmp);
7340 /* We've reached the end. */
7346 /* Do accounting for the sndbuf space.
7347 * Decrement the used sndbuf space of the corresponding association by the
7348 * data size which was just transmitted(freed).
7350 static void sctp_wfree(struct sk_buff *skb)
7352 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
7353 struct sctp_association *asoc = chunk->asoc;
7354 struct sock *sk = asoc->base.sk;
7356 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
7357 sizeof(struct sk_buff) +
7358 sizeof(struct sctp_chunk);
7360 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
7363 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
7365 sk->sk_wmem_queued -= skb->truesize;
7366 sk_mem_uncharge(sk, skb->truesize);
7369 sctp_wake_up_waiters(sk, asoc);
7371 sctp_association_put(asoc);
7374 /* Do accounting for the receive space on the socket.
7375 * Accounting for the association is done in ulpevent.c
7376 * We set this as a destructor for the cloned data skbs so that
7377 * accounting is done at the correct time.
7379 void sctp_sock_rfree(struct sk_buff *skb)
7381 struct sock *sk = skb->sk;
7382 struct sctp_ulpevent *event = sctp_skb2event(skb);
7384 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
7387 * Mimic the behavior of sock_rfree
7389 sk_mem_uncharge(sk, event->rmem_len);
7393 /* Helper function to wait for space in the sndbuf. */
7394 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
7397 struct sock *sk = asoc->base.sk;
7399 long current_timeo = *timeo_p;
7402 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
7405 /* Increment the association's refcnt. */
7406 sctp_association_hold(asoc);
7408 /* Wait on the association specific sndbuf space. */
7410 prepare_to_wait_exclusive(&asoc->wait, &wait,
7411 TASK_INTERRUPTIBLE);
7414 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7417 if (signal_pending(current))
7418 goto do_interrupted;
7419 if (msg_len <= sctp_wspace(asoc))
7422 /* Let another process have a go. Since we are going
7426 current_timeo = schedule_timeout(current_timeo);
7427 BUG_ON(sk != asoc->base.sk);
7430 *timeo_p = current_timeo;
7434 finish_wait(&asoc->wait, &wait);
7436 /* Release the association's refcnt. */
7437 sctp_association_put(asoc);
7446 err = sock_intr_errno(*timeo_p);
7454 void sctp_data_ready(struct sock *sk)
7456 struct socket_wq *wq;
7459 wq = rcu_dereference(sk->sk_wq);
7460 if (skwq_has_sleeper(wq))
7461 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
7462 POLLRDNORM | POLLRDBAND);
7463 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
7467 /* If socket sndbuf has changed, wake up all per association waiters. */
7468 void sctp_write_space(struct sock *sk)
7470 struct sctp_association *asoc;
7472 /* Wake up the tasks in each wait queue. */
7473 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
7474 __sctp_write_space(asoc);
7478 /* Is there any sndbuf space available on the socket?
7480 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
7481 * associations on the same socket. For a UDP-style socket with
7482 * multiple associations, it is possible for it to be "unwriteable"
7483 * prematurely. I assume that this is acceptable because
7484 * a premature "unwriteable" is better than an accidental "writeable" which
7485 * would cause an unwanted block under certain circumstances. For the 1-1
7486 * UDP-style sockets or TCP-style sockets, this code should work.
7489 static int sctp_writeable(struct sock *sk)
7493 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
7499 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
7500 * returns immediately with EINPROGRESS.
7502 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
7504 struct sock *sk = asoc->base.sk;
7506 long current_timeo = *timeo_p;
7509 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
7511 /* Increment the association's refcnt. */
7512 sctp_association_hold(asoc);
7515 prepare_to_wait_exclusive(&asoc->wait, &wait,
7516 TASK_INTERRUPTIBLE);
7519 if (sk->sk_shutdown & RCV_SHUTDOWN)
7521 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7524 if (signal_pending(current))
7525 goto do_interrupted;
7527 if (sctp_state(asoc, ESTABLISHED))
7530 /* Let another process have a go. Since we are going
7534 current_timeo = schedule_timeout(current_timeo);
7537 *timeo_p = current_timeo;
7541 finish_wait(&asoc->wait, &wait);
7543 /* Release the association's refcnt. */
7544 sctp_association_put(asoc);
7549 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
7552 err = -ECONNREFUSED;
7556 err = sock_intr_errno(*timeo_p);
7564 static int sctp_wait_for_accept(struct sock *sk, long timeo)
7566 struct sctp_endpoint *ep;
7570 ep = sctp_sk(sk)->ep;
7574 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
7575 TASK_INTERRUPTIBLE);
7577 if (list_empty(&ep->asocs)) {
7579 timeo = schedule_timeout(timeo);
7584 if (!sctp_sstate(sk, LISTENING))
7588 if (!list_empty(&ep->asocs))
7591 err = sock_intr_errno(timeo);
7592 if (signal_pending(current))
7600 finish_wait(sk_sleep(sk), &wait);
7605 static void sctp_wait_for_close(struct sock *sk, long timeout)
7610 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7611 if (list_empty(&sctp_sk(sk)->ep->asocs))
7614 timeout = schedule_timeout(timeout);
7616 } while (!signal_pending(current) && timeout);
7618 finish_wait(sk_sleep(sk), &wait);
7621 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
7623 struct sk_buff *frag;
7628 /* Don't forget the fragments. */
7629 skb_walk_frags(skb, frag)
7630 sctp_skb_set_owner_r_frag(frag, sk);
7633 sctp_skb_set_owner_r(skb, sk);
7636 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
7637 struct sctp_association *asoc)
7639 struct inet_sock *inet = inet_sk(sk);
7640 struct inet_sock *newinet;
7642 newsk->sk_type = sk->sk_type;
7643 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
7644 newsk->sk_flags = sk->sk_flags;
7645 newsk->sk_tsflags = sk->sk_tsflags;
7646 newsk->sk_no_check_tx = sk->sk_no_check_tx;
7647 newsk->sk_no_check_rx = sk->sk_no_check_rx;
7648 newsk->sk_reuse = sk->sk_reuse;
7650 newsk->sk_shutdown = sk->sk_shutdown;
7651 newsk->sk_destruct = sctp_destruct_sock;
7652 newsk->sk_family = sk->sk_family;
7653 newsk->sk_protocol = IPPROTO_SCTP;
7654 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
7655 newsk->sk_sndbuf = sk->sk_sndbuf;
7656 newsk->sk_rcvbuf = sk->sk_rcvbuf;
7657 newsk->sk_lingertime = sk->sk_lingertime;
7658 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
7659 newsk->sk_sndtimeo = sk->sk_sndtimeo;
7660 newsk->sk_rxhash = sk->sk_rxhash;
7662 newinet = inet_sk(newsk);
7664 /* Initialize sk's sport, dport, rcv_saddr and daddr for
7665 * getsockname() and getpeername()
7667 newinet->inet_sport = inet->inet_sport;
7668 newinet->inet_saddr = inet->inet_saddr;
7669 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
7670 newinet->inet_dport = htons(asoc->peer.port);
7671 newinet->pmtudisc = inet->pmtudisc;
7672 newinet->inet_id = asoc->next_tsn ^ jiffies;
7674 newinet->uc_ttl = inet->uc_ttl;
7675 newinet->mc_loop = 1;
7676 newinet->mc_ttl = 1;
7677 newinet->mc_index = 0;
7678 newinet->mc_list = NULL;
7680 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
7681 net_enable_timestamp();
7683 security_sk_clone(sk, newsk);
7686 static inline void sctp_copy_descendant(struct sock *sk_to,
7687 const struct sock *sk_from)
7689 int ancestor_size = sizeof(struct inet_sock) +
7690 sizeof(struct sctp_sock) -
7691 offsetof(struct sctp_sock, auto_asconf_list);
7693 if (sk_from->sk_family == PF_INET6)
7694 ancestor_size += sizeof(struct ipv6_pinfo);
7696 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
7699 /* Populate the fields of the newsk from the oldsk and migrate the assoc
7700 * and its messages to the newsk.
7702 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
7703 struct sctp_association *assoc,
7704 sctp_socket_type_t type)
7706 struct sctp_sock *oldsp = sctp_sk(oldsk);
7707 struct sctp_sock *newsp = sctp_sk(newsk);
7708 struct sctp_bind_bucket *pp; /* hash list port iterator */
7709 struct sctp_endpoint *newep = newsp->ep;
7710 struct sk_buff *skb, *tmp;
7711 struct sctp_ulpevent *event;
7712 struct sctp_bind_hashbucket *head;
7714 /* Migrate socket buffer sizes and all the socket level options to the
7717 newsk->sk_sndbuf = oldsk->sk_sndbuf;
7718 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
7719 /* Brute force copy old sctp opt. */
7720 sctp_copy_descendant(newsk, oldsk);
7722 /* Restore the ep value that was overwritten with the above structure
7728 /* Hook this new socket in to the bind_hash list. */
7729 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
7730 inet_sk(oldsk)->inet_num)];
7731 spin_lock_bh(&head->lock);
7732 pp = sctp_sk(oldsk)->bind_hash;
7733 sk_add_bind_node(newsk, &pp->owner);
7734 sctp_sk(newsk)->bind_hash = pp;
7735 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
7736 spin_unlock_bh(&head->lock);
7738 /* Copy the bind_addr list from the original endpoint to the new
7739 * endpoint so that we can handle restarts properly
7741 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
7742 &oldsp->ep->base.bind_addr, GFP_KERNEL);
7744 /* Move any messages in the old socket's receive queue that are for the
7745 * peeled off association to the new socket's receive queue.
7747 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
7748 event = sctp_skb2event(skb);
7749 if (event->asoc == assoc) {
7750 __skb_unlink(skb, &oldsk->sk_receive_queue);
7751 __skb_queue_tail(&newsk->sk_receive_queue, skb);
7752 sctp_skb_set_owner_r_frag(skb, newsk);
7756 /* Clean up any messages pending delivery due to partial
7757 * delivery. Three cases:
7758 * 1) No partial deliver; no work.
7759 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7760 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7762 skb_queue_head_init(&newsp->pd_lobby);
7763 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
7765 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
7766 struct sk_buff_head *queue;
7768 /* Decide which queue to move pd_lobby skbs to. */
7769 if (assoc->ulpq.pd_mode) {
7770 queue = &newsp->pd_lobby;
7772 queue = &newsk->sk_receive_queue;
7774 /* Walk through the pd_lobby, looking for skbs that
7775 * need moved to the new socket.
7777 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
7778 event = sctp_skb2event(skb);
7779 if (event->asoc == assoc) {
7780 __skb_unlink(skb, &oldsp->pd_lobby);
7781 __skb_queue_tail(queue, skb);
7782 sctp_skb_set_owner_r_frag(skb, newsk);
7786 /* Clear up any skbs waiting for the partial
7787 * delivery to finish.
7789 if (assoc->ulpq.pd_mode)
7790 sctp_clear_pd(oldsk, NULL);
7794 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7795 sctp_skb_set_owner_r_frag(skb, newsk);
7797 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7798 sctp_skb_set_owner_r_frag(skb, newsk);
7800 /* Set the type of socket to indicate that it is peeled off from the
7801 * original UDP-style socket or created with the accept() call on a
7802 * TCP-style socket..
7806 /* Mark the new socket "in-use" by the user so that any packets
7807 * that may arrive on the association after we've moved it are
7808 * queued to the backlog. This prevents a potential race between
7809 * backlog processing on the old socket and new-packet processing
7810 * on the new socket.
7812 * The caller has just allocated newsk so we can guarantee that other
7813 * paths won't try to lock it and then oldsk.
7815 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7816 sctp_assoc_migrate(assoc, newsk);
7818 /* If the association on the newsk is already closed before accept()
7819 * is called, set RCV_SHUTDOWN flag.
7821 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP)) {
7822 newsk->sk_state = SCTP_SS_CLOSED;
7823 newsk->sk_shutdown |= RCV_SHUTDOWN;
7825 newsk->sk_state = SCTP_SS_ESTABLISHED;
7828 release_sock(newsk);
7832 /* This proto struct describes the ULP interface for SCTP. */
7833 struct proto sctp_prot = {
7835 .owner = THIS_MODULE,
7836 .close = sctp_close,
7837 .connect = sctp_connect,
7838 .disconnect = sctp_disconnect,
7839 .accept = sctp_accept,
7840 .ioctl = sctp_ioctl,
7841 .init = sctp_init_sock,
7842 .destroy = sctp_destroy_sock,
7843 .shutdown = sctp_shutdown,
7844 .setsockopt = sctp_setsockopt,
7845 .getsockopt = sctp_getsockopt,
7846 .sendmsg = sctp_sendmsg,
7847 .recvmsg = sctp_recvmsg,
7849 .backlog_rcv = sctp_backlog_rcv,
7851 .unhash = sctp_unhash,
7852 .get_port = sctp_get_port,
7853 .obj_size = sizeof(struct sctp_sock),
7854 .sysctl_mem = sysctl_sctp_mem,
7855 .sysctl_rmem = sysctl_sctp_rmem,
7856 .sysctl_wmem = sysctl_sctp_wmem,
7857 .memory_pressure = &sctp_memory_pressure,
7858 .enter_memory_pressure = sctp_enter_memory_pressure,
7859 .memory_allocated = &sctp_memory_allocated,
7860 .sockets_allocated = &sctp_sockets_allocated,
7863 #if IS_ENABLED(CONFIG_IPV6)
7865 #include <net/transp_v6.h>
7866 static void sctp_v6_destroy_sock(struct sock *sk)
7868 sctp_destroy_sock(sk);
7869 inet6_destroy_sock(sk);
7872 struct proto sctpv6_prot = {
7874 .owner = THIS_MODULE,
7875 .close = sctp_close,
7876 .connect = sctp_connect,
7877 .disconnect = sctp_disconnect,
7878 .accept = sctp_accept,
7879 .ioctl = sctp_ioctl,
7880 .init = sctp_init_sock,
7881 .destroy = sctp_v6_destroy_sock,
7882 .shutdown = sctp_shutdown,
7883 .setsockopt = sctp_setsockopt,
7884 .getsockopt = sctp_getsockopt,
7885 .sendmsg = sctp_sendmsg,
7886 .recvmsg = sctp_recvmsg,
7888 .backlog_rcv = sctp_backlog_rcv,
7890 .unhash = sctp_unhash,
7891 .get_port = sctp_get_port,
7892 .obj_size = sizeof(struct sctp6_sock),
7893 .sysctl_mem = sysctl_sctp_mem,
7894 .sysctl_rmem = sysctl_sctp_rmem,
7895 .sysctl_wmem = sysctl_sctp_wmem,
7896 .memory_pressure = &sctp_memory_pressure,
7897 .enter_memory_pressure = sctp_enter_memory_pressure,
7898 .memory_allocated = &sctp_memory_allocated,
7899 .sockets_allocated = &sctp_sockets_allocated,
7901 #endif /* IS_ENABLED(CONFIG_IPV6) */