1 /* SCTP kernel implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
9 * This file is part of the SCTP kernel implementation
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
18 * This SCTP implementation is free software;
19 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
24 * This SCTP implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, see
32 * <http://www.gnu.org/licenses/>.
34 * Please send any bug reports or fixes you make to the
36 * lksctp developers <linux-sctp@vger.kernel.org>
38 * Written or modified by:
39 * La Monte H.P. Yarroll <piggy@acm.org>
40 * Narasimha Budihal <narsi@refcode.org>
41 * Karl Knutson <karl@athena.chicago.il.us>
42 * Jon Grimm <jgrimm@us.ibm.com>
43 * Xingang Guo <xingang.guo@intel.com>
44 * Daisy Chang <daisyc@us.ibm.com>
45 * Sridhar Samudrala <samudrala@us.ibm.com>
46 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
47 * Ardelle Fan <ardelle.fan@intel.com>
48 * Ryan Layer <rmlayer@us.ibm.com>
49 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
50 * Kevin Gao <kevin.gao@intel.com>
53 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
55 #include <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 static void sctp_clear_owner_w(struct sctp_chunk *chunk)
173 skb_orphan(chunk->skb);
176 static void sctp_for_each_tx_datachunk(struct sctp_association *asoc,
177 void (*cb)(struct sctp_chunk *))
180 struct sctp_outq *q = &asoc->outqueue;
181 struct sctp_transport *t;
182 struct sctp_chunk *chunk;
184 list_for_each_entry(t, &asoc->peer.transport_addr_list, transports)
185 list_for_each_entry(chunk, &t->transmitted, transmitted_list)
188 list_for_each_entry(chunk, &q->retransmit, list)
191 list_for_each_entry(chunk, &q->sacked, list)
194 list_for_each_entry(chunk, &q->abandoned, list)
197 list_for_each_entry(chunk, &q->out_chunk_list, list)
201 /* Verify that this is a valid address. */
202 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
207 /* Verify basic sockaddr. */
208 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
212 /* Is this a valid SCTP address? */
213 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
216 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
222 /* Look up the association by its id. If this is not a UDP-style
223 * socket, the ID field is always ignored.
225 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
227 struct sctp_association *asoc = NULL;
229 /* If this is not a UDP-style socket, assoc id should be ignored. */
230 if (!sctp_style(sk, UDP)) {
231 /* Return NULL if the socket state is not ESTABLISHED. It
232 * could be a TCP-style listening socket or a socket which
233 * hasn't yet called connect() to establish an association.
235 if (!sctp_sstate(sk, ESTABLISHED) && !sctp_sstate(sk, CLOSING))
238 /* Get the first and the only association from the list. */
239 if (!list_empty(&sctp_sk(sk)->ep->asocs))
240 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
241 struct sctp_association, asocs);
245 /* Otherwise this is a UDP-style socket. */
246 if (!id || (id == (sctp_assoc_t)-1))
249 spin_lock_bh(&sctp_assocs_id_lock);
250 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
251 spin_unlock_bh(&sctp_assocs_id_lock);
253 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
259 /* Look up the transport from an address and an assoc id. If both address and
260 * id are specified, the associations matching the address and the id should be
263 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
264 struct sockaddr_storage *addr,
267 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
268 struct sctp_af *af = sctp_get_af_specific(addr->ss_family);
269 union sctp_addr *laddr = (union sctp_addr *)addr;
270 struct sctp_transport *transport;
272 if (!af || sctp_verify_addr(sk, laddr, af->sockaddr_len))
275 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
282 id_asoc = sctp_id2assoc(sk, id);
283 if (id_asoc && (id_asoc != addr_asoc))
286 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
287 (union sctp_addr *)addr);
292 /* API 3.1.2 bind() - UDP Style Syntax
293 * The syntax of bind() is,
295 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
297 * sd - the socket descriptor returned by socket().
298 * addr - the address structure (struct sockaddr_in or struct
299 * sockaddr_in6 [RFC 2553]),
300 * addr_len - the size of the address structure.
302 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
308 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
311 /* Disallow binding twice. */
312 if (!sctp_sk(sk)->ep->base.bind_addr.port)
313 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
323 static long sctp_get_port_local(struct sock *, union sctp_addr *);
325 /* Verify this is a valid sockaddr. */
326 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
327 union sctp_addr *addr, int len)
331 /* Check minimum size. */
332 if (len < sizeof (struct sockaddr))
335 /* V4 mapped address are really of AF_INET family */
336 if (addr->sa.sa_family == AF_INET6 &&
337 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
338 if (!opt->pf->af_supported(AF_INET, opt))
341 /* Does this PF support this AF? */
342 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
346 /* If we get this far, af is valid. */
347 af = sctp_get_af_specific(addr->sa.sa_family);
349 if (len < af->sockaddr_len)
355 /* Bind a local address either to an endpoint or to an association. */
356 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
358 struct net *net = sock_net(sk);
359 struct sctp_sock *sp = sctp_sk(sk);
360 struct sctp_endpoint *ep = sp->ep;
361 struct sctp_bind_addr *bp = &ep->base.bind_addr;
366 /* Common sockaddr verification. */
367 af = sctp_sockaddr_af(sp, addr, len);
369 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
370 __func__, sk, addr, len);
374 snum = ntohs(addr->v4.sin_port);
376 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
377 __func__, sk, &addr->sa, bp->port, snum, len);
379 /* PF specific bind() address verification. */
380 if (!sp->pf->bind_verify(sp, addr))
381 return -EADDRNOTAVAIL;
383 /* We must either be unbound, or bind to the same port.
384 * It's OK to allow 0 ports if we are already bound.
385 * We'll just inhert an already bound port in this case
390 else if (snum != bp->port) {
391 pr_debug("%s: new port %d doesn't match existing port "
392 "%d\n", __func__, snum, bp->port);
397 if (snum && snum < PROT_SOCK &&
398 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
401 /* See if the address matches any of the addresses we may have
402 * already bound before checking against other endpoints.
404 if (sctp_bind_addr_match(bp, addr, sp))
407 /* Make sure we are allowed to bind here.
408 * The function sctp_get_port_local() does duplicate address
411 addr->v4.sin_port = htons(snum);
412 if ((ret = sctp_get_port_local(sk, addr))) {
416 /* Refresh ephemeral port. */
418 bp->port = inet_sk(sk)->inet_num;
420 /* Add the address to the bind address list.
421 * Use GFP_ATOMIC since BHs will be disabled.
423 ret = sctp_add_bind_addr(bp, addr, af->sockaddr_len,
424 SCTP_ADDR_SRC, GFP_ATOMIC);
426 /* Copy back into socket for getsockname() use. */
428 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
429 sp->pf->to_sk_saddr(addr, sk);
435 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
437 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
438 * at any one time. If a sender, after sending an ASCONF chunk, decides
439 * it needs to transfer another ASCONF Chunk, it MUST wait until the
440 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
441 * subsequent ASCONF. Note this restriction binds each side, so at any
442 * time two ASCONF may be in-transit on any given association (one sent
443 * from each endpoint).
445 static int sctp_send_asconf(struct sctp_association *asoc,
446 struct sctp_chunk *chunk)
448 struct net *net = sock_net(asoc->base.sk);
451 /* If there is an outstanding ASCONF chunk, queue it for later
454 if (asoc->addip_last_asconf) {
455 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
459 /* Hold the chunk until an ASCONF_ACK is received. */
460 sctp_chunk_hold(chunk);
461 retval = sctp_primitive_ASCONF(net, asoc, chunk);
463 sctp_chunk_free(chunk);
465 asoc->addip_last_asconf = chunk;
471 /* Add a list of addresses as bind addresses to local endpoint or
474 * Basically run through each address specified in the addrs/addrcnt
475 * array/length pair, determine if it is IPv6 or IPv4 and call
476 * sctp_do_bind() on it.
478 * If any of them fails, then the operation will be reversed and the
479 * ones that were added will be removed.
481 * Only sctp_setsockopt_bindx() is supposed to call this function.
483 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
488 struct sockaddr *sa_addr;
491 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
495 for (cnt = 0; cnt < addrcnt; cnt++) {
496 /* The list may contain either IPv4 or IPv6 address;
497 * determine the address length for walking thru the list.
500 af = sctp_get_af_specific(sa_addr->sa_family);
506 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
509 addr_buf += af->sockaddr_len;
513 /* Failed. Cleanup the ones that have been added */
515 sctp_bindx_rem(sk, addrs, cnt);
523 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
524 * associations that are part of the endpoint indicating that a list of local
525 * addresses are added to the endpoint.
527 * If any of the addresses is already in the bind address list of the
528 * association, we do not send the chunk for that association. But it will not
529 * affect other associations.
531 * Only sctp_setsockopt_bindx() is supposed to call this function.
533 static int sctp_send_asconf_add_ip(struct sock *sk,
534 struct sockaddr *addrs,
537 struct net *net = sock_net(sk);
538 struct sctp_sock *sp;
539 struct sctp_endpoint *ep;
540 struct sctp_association *asoc;
541 struct sctp_bind_addr *bp;
542 struct sctp_chunk *chunk;
543 struct sctp_sockaddr_entry *laddr;
544 union sctp_addr *addr;
545 union sctp_addr saveaddr;
552 if (!net->sctp.addip_enable)
558 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
559 __func__, sk, addrs, addrcnt);
561 list_for_each_entry(asoc, &ep->asocs, asocs) {
562 if (!asoc->peer.asconf_capable)
565 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
568 if (!sctp_state(asoc, ESTABLISHED))
571 /* Check if any address in the packed array of addresses is
572 * in the bind address list of the association. If so,
573 * do not send the asconf chunk to its peer, but continue with
574 * other associations.
577 for (i = 0; i < addrcnt; i++) {
579 af = sctp_get_af_specific(addr->v4.sin_family);
585 if (sctp_assoc_lookup_laddr(asoc, addr))
588 addr_buf += af->sockaddr_len;
593 /* Use the first valid address in bind addr list of
594 * association as Address Parameter of ASCONF CHUNK.
596 bp = &asoc->base.bind_addr;
597 p = bp->address_list.next;
598 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
599 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
600 addrcnt, SCTP_PARAM_ADD_IP);
606 /* Add the new addresses to the bind address list with
607 * use_as_src set to 0.
610 for (i = 0; i < addrcnt; i++) {
612 af = sctp_get_af_specific(addr->v4.sin_family);
613 memcpy(&saveaddr, addr, af->sockaddr_len);
614 retval = sctp_add_bind_addr(bp, &saveaddr,
616 SCTP_ADDR_NEW, GFP_ATOMIC);
617 addr_buf += af->sockaddr_len;
619 if (asoc->src_out_of_asoc_ok) {
620 struct sctp_transport *trans;
622 list_for_each_entry(trans,
623 &asoc->peer.transport_addr_list, transports) {
624 /* Clear the source and route cache */
625 dst_release(trans->dst);
626 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
627 2*asoc->pathmtu, 4380));
628 trans->ssthresh = asoc->peer.i.a_rwnd;
629 trans->rto = asoc->rto_initial;
630 sctp_max_rto(asoc, trans);
631 trans->rtt = trans->srtt = trans->rttvar = 0;
632 sctp_transport_route(trans, NULL,
633 sctp_sk(asoc->base.sk));
636 retval = sctp_send_asconf(asoc, chunk);
643 /* Remove a list of addresses from bind addresses list. Do not remove the
646 * Basically run through each address specified in the addrs/addrcnt
647 * array/length pair, determine if it is IPv6 or IPv4 and call
648 * sctp_del_bind() on it.
650 * If any of them fails, then the operation will be reversed and the
651 * ones that were removed will be added back.
653 * At least one address has to be left; if only one address is
654 * available, the operation will return -EBUSY.
656 * Only sctp_setsockopt_bindx() is supposed to call this function.
658 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
660 struct sctp_sock *sp = sctp_sk(sk);
661 struct sctp_endpoint *ep = sp->ep;
663 struct sctp_bind_addr *bp = &ep->base.bind_addr;
666 union sctp_addr *sa_addr;
669 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
670 __func__, sk, addrs, addrcnt);
673 for (cnt = 0; cnt < addrcnt; cnt++) {
674 /* If the bind address list is empty or if there is only one
675 * bind address, there is nothing more to be removed (we need
676 * at least one address here).
678 if (list_empty(&bp->address_list) ||
679 (sctp_list_single_entry(&bp->address_list))) {
685 af = sctp_get_af_specific(sa_addr->sa.sa_family);
691 if (!af->addr_valid(sa_addr, sp, NULL)) {
692 retval = -EADDRNOTAVAIL;
696 if (sa_addr->v4.sin_port &&
697 sa_addr->v4.sin_port != htons(bp->port)) {
702 if (!sa_addr->v4.sin_port)
703 sa_addr->v4.sin_port = htons(bp->port);
705 /* FIXME - There is probably a need to check if sk->sk_saddr and
706 * sk->sk_rcv_addr are currently set to one of the addresses to
707 * be removed. This is something which needs to be looked into
708 * when we are fixing the outstanding issues with multi-homing
709 * socket routing and failover schemes. Refer to comments in
710 * sctp_do_bind(). -daisy
712 retval = sctp_del_bind_addr(bp, sa_addr);
714 addr_buf += af->sockaddr_len;
717 /* Failed. Add the ones that has been removed back */
719 sctp_bindx_add(sk, addrs, cnt);
727 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
728 * the associations that are part of the endpoint indicating that a list of
729 * local addresses are removed from the endpoint.
731 * If any of the addresses is already in the bind address list of the
732 * association, we do not send the chunk for that association. But it will not
733 * affect other associations.
735 * Only sctp_setsockopt_bindx() is supposed to call this function.
737 static int sctp_send_asconf_del_ip(struct sock *sk,
738 struct sockaddr *addrs,
741 struct net *net = sock_net(sk);
742 struct sctp_sock *sp;
743 struct sctp_endpoint *ep;
744 struct sctp_association *asoc;
745 struct sctp_transport *transport;
746 struct sctp_bind_addr *bp;
747 struct sctp_chunk *chunk;
748 union sctp_addr *laddr;
751 struct sctp_sockaddr_entry *saddr;
757 if (!net->sctp.addip_enable)
763 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
764 __func__, sk, addrs, addrcnt);
766 list_for_each_entry(asoc, &ep->asocs, asocs) {
768 if (!asoc->peer.asconf_capable)
771 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
774 if (!sctp_state(asoc, ESTABLISHED))
777 /* Check if any address in the packed array of addresses is
778 * not present in the bind address list of the association.
779 * If so, do not send the asconf chunk to its peer, but
780 * continue with other associations.
783 for (i = 0; i < addrcnt; i++) {
785 af = sctp_get_af_specific(laddr->v4.sin_family);
791 if (!sctp_assoc_lookup_laddr(asoc, laddr))
794 addr_buf += af->sockaddr_len;
799 /* Find one address in the association's bind address list
800 * that is not in the packed array of addresses. This is to
801 * make sure that we do not delete all the addresses in the
804 bp = &asoc->base.bind_addr;
805 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
807 if ((laddr == NULL) && (addrcnt == 1)) {
808 if (asoc->asconf_addr_del_pending)
810 asoc->asconf_addr_del_pending =
811 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
812 if (asoc->asconf_addr_del_pending == NULL) {
816 asoc->asconf_addr_del_pending->sa.sa_family =
818 asoc->asconf_addr_del_pending->v4.sin_port =
820 if (addrs->sa_family == AF_INET) {
821 struct sockaddr_in *sin;
823 sin = (struct sockaddr_in *)addrs;
824 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
825 } else if (addrs->sa_family == AF_INET6) {
826 struct sockaddr_in6 *sin6;
828 sin6 = (struct sockaddr_in6 *)addrs;
829 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
832 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
833 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
834 asoc->asconf_addr_del_pending);
836 asoc->src_out_of_asoc_ok = 1;
844 /* We do not need RCU protection throughout this loop
845 * because this is done under a socket lock from the
848 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
856 /* Reset use_as_src flag for the addresses in the bind address
857 * list that are to be deleted.
860 for (i = 0; i < addrcnt; i++) {
862 af = sctp_get_af_specific(laddr->v4.sin_family);
863 list_for_each_entry(saddr, &bp->address_list, list) {
864 if (sctp_cmp_addr_exact(&saddr->a, laddr))
865 saddr->state = SCTP_ADDR_DEL;
867 addr_buf += af->sockaddr_len;
870 /* Update the route and saddr entries for all the transports
871 * as some of the addresses in the bind address list are
872 * about to be deleted and cannot be used as source addresses.
874 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
876 dst_release(transport->dst);
877 sctp_transport_route(transport, NULL,
878 sctp_sk(asoc->base.sk));
882 /* We don't need to transmit ASCONF */
884 retval = sctp_send_asconf(asoc, chunk);
890 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
891 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
893 struct sock *sk = sctp_opt2sk(sp);
894 union sctp_addr *addr;
897 /* It is safe to write port space in caller. */
899 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
900 af = sctp_get_af_specific(addr->sa.sa_family);
903 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
906 if (addrw->state == SCTP_ADDR_NEW)
907 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
909 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
912 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
915 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
918 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
919 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
922 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
923 * Section 3.1.2 for this usage.
925 * addrs is a pointer to an array of one or more socket addresses. Each
926 * address is contained in its appropriate structure (i.e. struct
927 * sockaddr_in or struct sockaddr_in6) the family of the address type
928 * must be used to distinguish the address length (note that this
929 * representation is termed a "packed array" of addresses). The caller
930 * specifies the number of addresses in the array with addrcnt.
932 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
933 * -1, and sets errno to the appropriate error code.
935 * For SCTP, the port given in each socket address must be the same, or
936 * sctp_bindx() will fail, setting errno to EINVAL.
938 * The flags parameter is formed from the bitwise OR of zero or more of
939 * the following currently defined flags:
941 * SCTP_BINDX_ADD_ADDR
943 * SCTP_BINDX_REM_ADDR
945 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
946 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
947 * addresses from the association. The two flags are mutually exclusive;
948 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
949 * not remove all addresses from an association; sctp_bindx() will
950 * reject such an attempt with EINVAL.
952 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
953 * additional addresses with an endpoint after calling bind(). Or use
954 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
955 * socket is associated with so that no new association accepted will be
956 * associated with those addresses. If the endpoint supports dynamic
957 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
958 * endpoint to send the appropriate message to the peer to change the
959 * peers address lists.
961 * Adding and removing addresses from a connected association is
962 * optional functionality. Implementations that do not support this
963 * functionality should return EOPNOTSUPP.
965 * Basically do nothing but copying the addresses from user to kernel
966 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
967 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
970 * We don't use copy_from_user() for optimization: we first do the
971 * sanity checks (buffer size -fast- and access check-healthy
972 * pointer); if all of those succeed, then we can alloc the memory
973 * (expensive operation) needed to copy the data to kernel. Then we do
974 * the copying without checking the user space area
975 * (__copy_from_user()).
977 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
980 * sk The sk of the socket
981 * addrs The pointer to the addresses in user land
982 * addrssize Size of the addrs buffer
983 * op Operation to perform (add or remove, see the flags of
986 * Returns 0 if ok, <0 errno code on error.
988 static int sctp_setsockopt_bindx(struct sock *sk,
989 struct sockaddr __user *addrs,
990 int addrs_size, int op)
992 struct sockaddr *kaddrs;
996 struct sockaddr *sa_addr;
1000 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
1001 __func__, sk, addrs, addrs_size, op);
1003 if (unlikely(addrs_size <= 0))
1006 /* Check the user passed a healthy pointer. */
1007 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1010 /* Alloc space for the address array in kernel memory. */
1011 kaddrs = kmalloc(addrs_size, GFP_USER | __GFP_NOWARN);
1012 if (unlikely(!kaddrs))
1015 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1020 /* Walk through the addrs buffer and count the number of addresses. */
1022 while (walk_size < addrs_size) {
1023 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1029 af = sctp_get_af_specific(sa_addr->sa_family);
1031 /* If the address family is not supported or if this address
1032 * causes the address buffer to overflow return EINVAL.
1034 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1039 addr_buf += af->sockaddr_len;
1040 walk_size += af->sockaddr_len;
1045 case SCTP_BINDX_ADD_ADDR:
1046 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1049 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1052 case SCTP_BINDX_REM_ADDR:
1053 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1056 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1070 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1072 * Common routine for handling connect() and sctp_connectx().
1073 * Connect will come in with just a single address.
1075 static int __sctp_connect(struct sock *sk,
1076 struct sockaddr *kaddrs,
1078 sctp_assoc_t *assoc_id)
1080 struct net *net = sock_net(sk);
1081 struct sctp_sock *sp;
1082 struct sctp_endpoint *ep;
1083 struct sctp_association *asoc = NULL;
1084 struct sctp_association *asoc2;
1085 struct sctp_transport *transport;
1092 union sctp_addr *sa_addr = NULL;
1094 unsigned short port;
1095 unsigned int f_flags = 0;
1100 /* connect() cannot be done on a socket that is already in ESTABLISHED
1101 * state - UDP-style peeled off socket or a TCP-style socket that
1102 * is already connected.
1103 * It cannot be done even on a TCP-style listening socket.
1105 if (sctp_sstate(sk, ESTABLISHED) || sctp_sstate(sk, CLOSING) ||
1106 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1111 /* Walk through the addrs buffer and count the number of addresses. */
1113 while (walk_size < addrs_size) {
1116 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1122 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1124 /* If the address family is not supported or if this address
1125 * causes the address buffer to overflow return EINVAL.
1127 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1132 port = ntohs(sa_addr->v4.sin_port);
1134 /* Save current address so we can work with it */
1135 memcpy(&to, sa_addr, af->sockaddr_len);
1137 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1141 /* Make sure the destination port is correctly set
1144 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1149 /* Check if there already is a matching association on the
1150 * endpoint (other than the one created here).
1152 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1153 if (asoc2 && asoc2 != asoc) {
1154 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1161 /* If we could not find a matching association on the endpoint,
1162 * make sure that there is no peeled-off association matching
1163 * the peer address even on another socket.
1165 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1166 err = -EADDRNOTAVAIL;
1171 /* If a bind() or sctp_bindx() is not called prior to
1172 * an sctp_connectx() call, the system picks an
1173 * ephemeral port and will choose an address set
1174 * equivalent to binding with a wildcard address.
1176 if (!ep->base.bind_addr.port) {
1177 if (sctp_autobind(sk)) {
1183 * If an unprivileged user inherits a 1-many
1184 * style socket with open associations on a
1185 * privileged port, it MAY be permitted to
1186 * accept new associations, but it SHOULD NOT
1187 * be permitted to open new associations.
1189 if (ep->base.bind_addr.port < PROT_SOCK &&
1190 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1196 scope = sctp_scope(&to);
1197 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1203 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1211 /* Prime the peer's transport structures. */
1212 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1220 addr_buf += af->sockaddr_len;
1221 walk_size += af->sockaddr_len;
1224 /* In case the user of sctp_connectx() wants an association
1225 * id back, assign one now.
1228 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1233 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1238 /* Initialize sk's dport and daddr for getpeername() */
1239 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1240 sp->pf->to_sk_daddr(sa_addr, sk);
1243 /* in-kernel sockets don't generally have a file allocated to them
1244 * if all they do is call sock_create_kern().
1246 if (sk->sk_socket->file)
1247 f_flags = sk->sk_socket->file->f_flags;
1249 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1252 *assoc_id = asoc->assoc_id;
1253 err = sctp_wait_for_connect(asoc, &timeo);
1254 /* Note: the asoc may be freed after the return of
1255 * sctp_wait_for_connect.
1258 /* Don't free association on exit. */
1262 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1263 __func__, asoc, kaddrs, err);
1266 /* sctp_primitive_ASSOCIATE may have added this association
1267 * To the hash table, try to unhash it, just in case, its a noop
1268 * if it wasn't hashed so we're safe
1270 sctp_association_free(asoc);
1275 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1278 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1279 * sctp_assoc_t *asoc);
1281 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1282 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1283 * or IPv6 addresses.
1285 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1286 * Section 3.1.2 for this usage.
1288 * addrs is a pointer to an array of one or more socket addresses. Each
1289 * address is contained in its appropriate structure (i.e. struct
1290 * sockaddr_in or struct sockaddr_in6) the family of the address type
1291 * must be used to distengish the address length (note that this
1292 * representation is termed a "packed array" of addresses). The caller
1293 * specifies the number of addresses in the array with addrcnt.
1295 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1296 * the association id of the new association. On failure, sctp_connectx()
1297 * returns -1, and sets errno to the appropriate error code. The assoc_id
1298 * is not touched by the kernel.
1300 * For SCTP, the port given in each socket address must be the same, or
1301 * sctp_connectx() will fail, setting errno to EINVAL.
1303 * An application can use sctp_connectx to initiate an association with
1304 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1305 * allows a caller to specify multiple addresses at which a peer can be
1306 * reached. The way the SCTP stack uses the list of addresses to set up
1307 * the association is implementation dependent. This function only
1308 * specifies that the stack will try to make use of all the addresses in
1309 * the list when needed.
1311 * Note that the list of addresses passed in is only used for setting up
1312 * the association. It does not necessarily equal the set of addresses
1313 * the peer uses for the resulting association. If the caller wants to
1314 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1315 * retrieve them after the association has been set up.
1317 * Basically do nothing but copying the addresses from user to kernel
1318 * land and invoking either sctp_connectx(). This is used for tunneling
1319 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1321 * We don't use copy_from_user() for optimization: we first do the
1322 * sanity checks (buffer size -fast- and access check-healthy
1323 * pointer); if all of those succeed, then we can alloc the memory
1324 * (expensive operation) needed to copy the data to kernel. Then we do
1325 * the copying without checking the user space area
1326 * (__copy_from_user()).
1328 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1331 * sk The sk of the socket
1332 * addrs The pointer to the addresses in user land
1333 * addrssize Size of the addrs buffer
1335 * Returns >=0 if ok, <0 errno code on error.
1337 static int __sctp_setsockopt_connectx(struct sock *sk,
1338 struct sockaddr __user *addrs,
1340 sctp_assoc_t *assoc_id)
1342 struct sockaddr *kaddrs;
1343 gfp_t gfp = GFP_KERNEL;
1346 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1347 __func__, sk, addrs, addrs_size);
1349 if (unlikely(addrs_size <= 0))
1352 /* Check the user passed a healthy pointer. */
1353 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1356 /* Alloc space for the address array in kernel memory. */
1357 if (sk->sk_socket->file)
1358 gfp = GFP_USER | __GFP_NOWARN;
1359 kaddrs = kmalloc(addrs_size, gfp);
1360 if (unlikely(!kaddrs))
1363 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1366 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1375 * This is an older interface. It's kept for backward compatibility
1376 * to the option that doesn't provide association id.
1378 static int sctp_setsockopt_connectx_old(struct sock *sk,
1379 struct sockaddr __user *addrs,
1382 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1386 * New interface for the API. The since the API is done with a socket
1387 * option, to make it simple we feed back the association id is as a return
1388 * indication to the call. Error is always negative and association id is
1391 static int sctp_setsockopt_connectx(struct sock *sk,
1392 struct sockaddr __user *addrs,
1395 sctp_assoc_t assoc_id = 0;
1398 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1407 * New (hopefully final) interface for the API.
1408 * We use the sctp_getaddrs_old structure so that use-space library
1409 * can avoid any unnecessary allocations. The only different part
1410 * is that we store the actual length of the address buffer into the
1411 * addrs_num structure member. That way we can re-use the existing
1414 #ifdef CONFIG_COMPAT
1415 struct compat_sctp_getaddrs_old {
1416 sctp_assoc_t assoc_id;
1418 compat_uptr_t addrs; /* struct sockaddr * */
1422 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1423 char __user *optval,
1426 struct sctp_getaddrs_old param;
1427 sctp_assoc_t assoc_id = 0;
1430 #ifdef CONFIG_COMPAT
1431 if (in_compat_syscall()) {
1432 struct compat_sctp_getaddrs_old param32;
1434 if (len < sizeof(param32))
1436 if (copy_from_user(¶m32, optval, sizeof(param32)))
1439 param.assoc_id = param32.assoc_id;
1440 param.addr_num = param32.addr_num;
1441 param.addrs = compat_ptr(param32.addrs);
1445 if (len < sizeof(param))
1447 if (copy_from_user(¶m, optval, sizeof(param)))
1451 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1452 param.addrs, param.addr_num,
1454 if (err == 0 || err == -EINPROGRESS) {
1455 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1457 if (put_user(sizeof(assoc_id), optlen))
1464 /* API 3.1.4 close() - UDP Style Syntax
1465 * Applications use close() to perform graceful shutdown (as described in
1466 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1467 * by a UDP-style socket.
1471 * ret = close(int sd);
1473 * sd - the socket descriptor of the associations to be closed.
1475 * To gracefully shutdown a specific association represented by the
1476 * UDP-style socket, an application should use the sendmsg() call,
1477 * passing no user data, but including the appropriate flag in the
1478 * ancillary data (see Section xxxx).
1480 * If sd in the close() call is a branched-off socket representing only
1481 * one association, the shutdown is performed on that association only.
1483 * 4.1.6 close() - TCP Style Syntax
1485 * Applications use close() to gracefully close down an association.
1489 * int close(int sd);
1491 * sd - the socket descriptor of the association to be closed.
1493 * After an application calls close() on a socket descriptor, no further
1494 * socket operations will succeed on that descriptor.
1496 * API 7.1.4 SO_LINGER
1498 * An application using the TCP-style socket can use this option to
1499 * perform the SCTP ABORT primitive. The linger option structure is:
1502 * int l_onoff; // option on/off
1503 * int l_linger; // linger time
1506 * To enable the option, set l_onoff to 1. If the l_linger value is set
1507 * to 0, calling close() is the same as the ABORT primitive. If the
1508 * value is set to a negative value, the setsockopt() call will return
1509 * an error. If the value is set to a positive value linger_time, the
1510 * close() can be blocked for at most linger_time ms. If the graceful
1511 * shutdown phase does not finish during this period, close() will
1512 * return but the graceful shutdown phase continues in the system.
1514 static void sctp_close(struct sock *sk, long timeout)
1516 struct net *net = sock_net(sk);
1517 struct sctp_endpoint *ep;
1518 struct sctp_association *asoc;
1519 struct list_head *pos, *temp;
1520 unsigned int data_was_unread;
1522 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1525 sk->sk_shutdown = SHUTDOWN_MASK;
1526 sk->sk_state = SCTP_SS_CLOSING;
1528 ep = sctp_sk(sk)->ep;
1530 /* Clean up any skbs sitting on the receive queue. */
1531 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1532 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1534 /* Walk all associations on an endpoint. */
1535 list_for_each_safe(pos, temp, &ep->asocs) {
1536 asoc = list_entry(pos, struct sctp_association, asocs);
1538 if (sctp_style(sk, TCP)) {
1539 /* A closed association can still be in the list if
1540 * it belongs to a TCP-style listening socket that is
1541 * not yet accepted. If so, free it. If not, send an
1542 * ABORT or SHUTDOWN based on the linger options.
1544 if (sctp_state(asoc, CLOSED)) {
1545 sctp_association_free(asoc);
1550 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1551 !skb_queue_empty(&asoc->ulpq.reasm) ||
1552 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1553 struct sctp_chunk *chunk;
1555 chunk = sctp_make_abort_user(asoc, NULL, 0);
1556 sctp_primitive_ABORT(net, asoc, chunk);
1558 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1561 /* On a TCP-style socket, block for at most linger_time if set. */
1562 if (sctp_style(sk, TCP) && timeout)
1563 sctp_wait_for_close(sk, timeout);
1565 /* This will run the backlog queue. */
1568 /* Supposedly, no process has access to the socket, but
1569 * the net layers still may.
1570 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1571 * held and that should be grabbed before socket lock.
1573 spin_lock_bh(&net->sctp.addr_wq_lock);
1576 /* Hold the sock, since sk_common_release() will put sock_put()
1577 * and we have just a little more cleanup.
1580 sk_common_release(sk);
1583 spin_unlock_bh(&net->sctp.addr_wq_lock);
1587 SCTP_DBG_OBJCNT_DEC(sock);
1590 /* Handle EPIPE error. */
1591 static int sctp_error(struct sock *sk, int flags, int err)
1594 err = sock_error(sk) ? : -EPIPE;
1595 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1596 send_sig(SIGPIPE, current, 0);
1600 /* API 3.1.3 sendmsg() - UDP Style Syntax
1602 * An application uses sendmsg() and recvmsg() calls to transmit data to
1603 * and receive data from its peer.
1605 * ssize_t sendmsg(int socket, const struct msghdr *message,
1608 * socket - the socket descriptor of the endpoint.
1609 * message - pointer to the msghdr structure which contains a single
1610 * user message and possibly some ancillary data.
1612 * See Section 5 for complete description of the data
1615 * flags - flags sent or received with the user message, see Section
1616 * 5 for complete description of the flags.
1618 * Note: This function could use a rewrite especially when explicit
1619 * connect support comes in.
1621 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1623 static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1625 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
1627 struct net *net = sock_net(sk);
1628 struct sctp_sock *sp;
1629 struct sctp_endpoint *ep;
1630 struct sctp_association *new_asoc = NULL, *asoc = NULL;
1631 struct sctp_transport *transport, *chunk_tp;
1632 struct sctp_chunk *chunk;
1634 struct sockaddr *msg_name = NULL;
1635 struct sctp_sndrcvinfo default_sinfo;
1636 struct sctp_sndrcvinfo *sinfo;
1637 struct sctp_initmsg *sinit;
1638 sctp_assoc_t associd = 0;
1639 sctp_cmsgs_t cmsgs = { NULL };
1641 bool fill_sinfo_ttl = false, wait_connect = false;
1642 struct sctp_datamsg *datamsg;
1643 int msg_flags = msg->msg_flags;
1644 __u16 sinfo_flags = 0;
1652 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1655 /* We cannot send a message over a TCP-style listening socket. */
1656 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1661 /* Parse out the SCTP CMSGs. */
1662 err = sctp_msghdr_parse(msg, &cmsgs);
1664 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1668 /* Fetch the destination address for this packet. This
1669 * address only selects the association--it is not necessarily
1670 * the address we will send to.
1671 * For a peeled-off socket, msg_name is ignored.
1673 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1674 int msg_namelen = msg->msg_namelen;
1676 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1681 if (msg_namelen > sizeof(to))
1682 msg_namelen = sizeof(to);
1683 memcpy(&to, msg->msg_name, msg_namelen);
1684 msg_name = msg->msg_name;
1688 if (cmsgs.sinfo != NULL) {
1689 memset(&default_sinfo, 0, sizeof(default_sinfo));
1690 default_sinfo.sinfo_stream = cmsgs.sinfo->snd_sid;
1691 default_sinfo.sinfo_flags = cmsgs.sinfo->snd_flags;
1692 default_sinfo.sinfo_ppid = cmsgs.sinfo->snd_ppid;
1693 default_sinfo.sinfo_context = cmsgs.sinfo->snd_context;
1694 default_sinfo.sinfo_assoc_id = cmsgs.sinfo->snd_assoc_id;
1696 sinfo = &default_sinfo;
1697 fill_sinfo_ttl = true;
1699 sinfo = cmsgs.srinfo;
1701 /* Did the user specify SNDINFO/SNDRCVINFO? */
1703 sinfo_flags = sinfo->sinfo_flags;
1704 associd = sinfo->sinfo_assoc_id;
1707 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1708 msg_len, sinfo_flags);
1710 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1711 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1716 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1717 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1718 * If SCTP_ABORT is set, the message length could be non zero with
1719 * the msg_iov set to the user abort reason.
1721 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1722 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1727 /* If SCTP_ADDR_OVER is set, there must be an address
1728 * specified in msg_name.
1730 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1737 pr_debug("%s: about to look up association\n", __func__);
1741 /* If a msg_name has been specified, assume this is to be used. */
1743 /* Look for a matching association on the endpoint. */
1744 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1746 /* If we could not find a matching association on the
1747 * endpoint, make sure that it is not a TCP-style
1748 * socket that already has an association or there is
1749 * no peeled-off association on another socket.
1752 ((sctp_style(sk, TCP) &&
1753 (sctp_sstate(sk, ESTABLISHED) ||
1754 sctp_sstate(sk, CLOSING))) ||
1755 sctp_endpoint_is_peeled_off(ep, &to))) {
1756 err = -EADDRNOTAVAIL;
1760 asoc = sctp_id2assoc(sk, associd);
1768 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1770 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1771 * socket that has an association in CLOSED state. This can
1772 * happen when an accepted socket has an association that is
1775 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1780 if (sinfo_flags & SCTP_EOF) {
1781 pr_debug("%s: shutting down association:%p\n",
1784 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1788 if (sinfo_flags & SCTP_ABORT) {
1790 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1796 pr_debug("%s: aborting association:%p\n",
1799 sctp_primitive_ABORT(net, asoc, chunk);
1805 /* Do we need to create the association? */
1807 pr_debug("%s: there is no association yet\n", __func__);
1809 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1814 /* Check for invalid stream against the stream counts,
1815 * either the default or the user specified stream counts.
1818 if (!sinit || !sinit->sinit_num_ostreams) {
1819 /* Check against the defaults. */
1820 if (sinfo->sinfo_stream >=
1821 sp->initmsg.sinit_num_ostreams) {
1826 /* Check against the requested. */
1827 if (sinfo->sinfo_stream >=
1828 sinit->sinit_num_ostreams) {
1836 * API 3.1.2 bind() - UDP Style Syntax
1837 * If a bind() or sctp_bindx() is not called prior to a
1838 * sendmsg() call that initiates a new association, the
1839 * system picks an ephemeral port and will choose an address
1840 * set equivalent to binding with a wildcard address.
1842 if (!ep->base.bind_addr.port) {
1843 if (sctp_autobind(sk)) {
1849 * If an unprivileged user inherits a one-to-many
1850 * style socket with open associations on a privileged
1851 * port, it MAY be permitted to accept new associations,
1852 * but it SHOULD NOT be permitted to open new
1855 if (ep->base.bind_addr.port < PROT_SOCK &&
1856 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1862 scope = sctp_scope(&to);
1863 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1869 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1875 /* If the SCTP_INIT ancillary data is specified, set all
1876 * the association init values accordingly.
1879 if (sinit->sinit_num_ostreams) {
1880 asoc->c.sinit_num_ostreams =
1881 sinit->sinit_num_ostreams;
1883 if (sinit->sinit_max_instreams) {
1884 asoc->c.sinit_max_instreams =
1885 sinit->sinit_max_instreams;
1887 if (sinit->sinit_max_attempts) {
1888 asoc->max_init_attempts
1889 = sinit->sinit_max_attempts;
1891 if (sinit->sinit_max_init_timeo) {
1892 asoc->max_init_timeo =
1893 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1897 /* Prime the peer's transport structures. */
1898 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1905 /* ASSERT: we have a valid association at this point. */
1906 pr_debug("%s: we have a valid association\n", __func__);
1909 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1910 * one with some defaults.
1912 memset(&default_sinfo, 0, sizeof(default_sinfo));
1913 default_sinfo.sinfo_stream = asoc->default_stream;
1914 default_sinfo.sinfo_flags = asoc->default_flags;
1915 default_sinfo.sinfo_ppid = asoc->default_ppid;
1916 default_sinfo.sinfo_context = asoc->default_context;
1917 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1918 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1920 sinfo = &default_sinfo;
1921 } else if (fill_sinfo_ttl) {
1922 /* In case SNDINFO was specified, we still need to fill
1923 * it with a default ttl from the assoc here.
1925 sinfo->sinfo_timetolive = asoc->default_timetolive;
1928 /* API 7.1.7, the sndbuf size per association bounds the
1929 * maximum size of data that can be sent in a single send call.
1931 if (msg_len > sk->sk_sndbuf) {
1936 if (asoc->pmtu_pending)
1937 sctp_assoc_pending_pmtu(sk, asoc);
1939 /* If fragmentation is disabled and the message length exceeds the
1940 * association fragmentation point, return EMSGSIZE. The I-D
1941 * does not specify what this error is, but this looks like
1944 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1949 /* Check for invalid stream. */
1950 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1955 if (sctp_wspace(asoc) < msg_len)
1956 sctp_prsctp_prune(asoc, sinfo, msg_len - sctp_wspace(asoc));
1958 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1959 if (!sctp_wspace(asoc)) {
1960 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1965 /* If an address is passed with the sendto/sendmsg call, it is used
1966 * to override the primary destination address in the TCP model, or
1967 * when SCTP_ADDR_OVER flag is set in the UDP model.
1969 if ((sctp_style(sk, TCP) && msg_name) ||
1970 (sinfo_flags & SCTP_ADDR_OVER)) {
1971 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1979 /* Auto-connect, if we aren't connected already. */
1980 if (sctp_state(asoc, CLOSED)) {
1981 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1985 wait_connect = true;
1986 pr_debug("%s: we associated primitively\n", __func__);
1989 /* Break the message into multiple chunks of maximum size. */
1990 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
1991 if (IS_ERR(datamsg)) {
1992 err = PTR_ERR(datamsg);
1996 /* Now send the (possibly) fragmented message. */
1997 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1998 sctp_chunk_hold(chunk);
2000 /* Do accounting for the write space. */
2001 sctp_set_owner_w(chunk);
2003 chunk->transport = chunk_tp;
2006 /* Send it to the lower layers. Note: all chunks
2007 * must either fail or succeed. The lower layer
2008 * works that way today. Keep it that way or this
2011 err = sctp_primitive_SEND(net, asoc, datamsg);
2012 /* Did the lower layer accept the chunk? */
2014 sctp_datamsg_free(datamsg);
2018 pr_debug("%s: we sent primitively\n", __func__);
2020 sctp_datamsg_put(datamsg);
2023 if (unlikely(wait_connect)) {
2024 timeo = sock_sndtimeo(sk, msg_flags & MSG_DONTWAIT);
2025 sctp_wait_for_connect(asoc, &timeo);
2028 /* If we are already past ASSOCIATE, the lower
2029 * layers are responsible for association cleanup.
2035 sctp_association_free(asoc);
2040 return sctp_error(sk, msg_flags, err);
2047 err = sock_error(sk);
2057 /* This is an extended version of skb_pull() that removes the data from the
2058 * start of a skb even when data is spread across the list of skb's in the
2059 * frag_list. len specifies the total amount of data that needs to be removed.
2060 * when 'len' bytes could be removed from the skb, it returns 0.
2061 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2062 * could not be removed.
2064 static int sctp_skb_pull(struct sk_buff *skb, int len)
2066 struct sk_buff *list;
2067 int skb_len = skb_headlen(skb);
2070 if (len <= skb_len) {
2071 __skb_pull(skb, len);
2075 __skb_pull(skb, skb_len);
2077 skb_walk_frags(skb, list) {
2078 rlen = sctp_skb_pull(list, len);
2079 skb->len -= (len-rlen);
2080 skb->data_len -= (len-rlen);
2091 /* API 3.1.3 recvmsg() - UDP Style Syntax
2093 * ssize_t recvmsg(int socket, struct msghdr *message,
2096 * socket - the socket descriptor of the endpoint.
2097 * message - pointer to the msghdr structure which contains a single
2098 * user message and possibly some ancillary data.
2100 * See Section 5 for complete description of the data
2103 * flags - flags sent or received with the user message, see Section
2104 * 5 for complete description of the flags.
2106 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2107 int noblock, int flags, int *addr_len)
2109 struct sctp_ulpevent *event = NULL;
2110 struct sctp_sock *sp = sctp_sk(sk);
2111 struct sk_buff *skb, *head_skb;
2116 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2117 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2122 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED) &&
2123 !sctp_sstate(sk, CLOSING) && !sctp_sstate(sk, CLOSED)) {
2128 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2132 /* Get the total length of the skb including any skb's in the
2141 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2143 event = sctp_skb2event(skb);
2148 if (event->chunk && event->chunk->head_skb)
2149 head_skb = event->chunk->head_skb;
2152 sock_recv_ts_and_drops(msg, sk, head_skb);
2153 if (sctp_ulpevent_is_notification(event)) {
2154 msg->msg_flags |= MSG_NOTIFICATION;
2155 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2157 sp->pf->skb_msgname(head_skb, msg->msg_name, addr_len);
2160 /* Check if we allow SCTP_NXTINFO. */
2161 if (sp->recvnxtinfo)
2162 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2163 /* Check if we allow SCTP_RCVINFO. */
2164 if (sp->recvrcvinfo)
2165 sctp_ulpevent_read_rcvinfo(event, msg);
2166 /* Check if we allow SCTP_SNDRCVINFO. */
2167 if (sp->subscribe.sctp_data_io_event)
2168 sctp_ulpevent_read_sndrcvinfo(event, msg);
2172 /* If skb's length exceeds the user's buffer, update the skb and
2173 * push it back to the receive_queue so that the next call to
2174 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2176 if (skb_len > copied) {
2177 msg->msg_flags &= ~MSG_EOR;
2178 if (flags & MSG_PEEK)
2180 sctp_skb_pull(skb, copied);
2181 skb_queue_head(&sk->sk_receive_queue, skb);
2183 /* When only partial message is copied to the user, increase
2184 * rwnd by that amount. If all the data in the skb is read,
2185 * rwnd is updated when the event is freed.
2187 if (!sctp_ulpevent_is_notification(event))
2188 sctp_assoc_rwnd_increase(event->asoc, copied);
2190 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2191 (event->msg_flags & MSG_EOR))
2192 msg->msg_flags |= MSG_EOR;
2194 msg->msg_flags &= ~MSG_EOR;
2197 if (flags & MSG_PEEK) {
2198 /* Release the skb reference acquired after peeking the skb in
2199 * sctp_skb_recv_datagram().
2203 /* Free the event which includes releasing the reference to
2204 * the owner of the skb, freeing the skb and updating the
2207 sctp_ulpevent_free(event);
2214 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2216 * This option is a on/off flag. If enabled no SCTP message
2217 * fragmentation will be performed. Instead if a message being sent
2218 * exceeds the current PMTU size, the message will NOT be sent and
2219 * instead a error will be indicated to the user.
2221 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2222 char __user *optval,
2223 unsigned int optlen)
2227 if (optlen < sizeof(int))
2230 if (get_user(val, (int __user *)optval))
2233 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2238 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2239 unsigned int optlen)
2241 struct sctp_association *asoc;
2242 struct sctp_ulpevent *event;
2244 if (optlen > sizeof(struct sctp_event_subscribe))
2246 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2249 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2250 * if there is no data to be sent or retransmit, the stack will
2251 * immediately send up this notification.
2253 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2254 &sctp_sk(sk)->subscribe)) {
2255 asoc = sctp_id2assoc(sk, 0);
2257 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2258 event = sctp_ulpevent_make_sender_dry_event(asoc,
2263 sctp_ulpq_tail_event(&asoc->ulpq, event);
2270 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2272 * This socket option is applicable to the UDP-style socket only. When
2273 * set it will cause associations that are idle for more than the
2274 * specified number of seconds to automatically close. An association
2275 * being idle is defined an association that has NOT sent or received
2276 * user data. The special value of '0' indicates that no automatic
2277 * close of any associations should be performed. The option expects an
2278 * integer defining the number of seconds of idle time before an
2279 * association is closed.
2281 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2282 unsigned int optlen)
2284 struct sctp_sock *sp = sctp_sk(sk);
2285 struct net *net = sock_net(sk);
2287 /* Applicable to UDP-style socket only */
2288 if (sctp_style(sk, TCP))
2290 if (optlen != sizeof(int))
2292 if (copy_from_user(&sp->autoclose, optval, optlen))
2295 if (sp->autoclose > net->sctp.max_autoclose)
2296 sp->autoclose = net->sctp.max_autoclose;
2301 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2303 * Applications can enable or disable heartbeats for any peer address of
2304 * an association, modify an address's heartbeat interval, force a
2305 * heartbeat to be sent immediately, and adjust the address's maximum
2306 * number of retransmissions sent before an address is considered
2307 * unreachable. The following structure is used to access and modify an
2308 * address's parameters:
2310 * struct sctp_paddrparams {
2311 * sctp_assoc_t spp_assoc_id;
2312 * struct sockaddr_storage spp_address;
2313 * uint32_t spp_hbinterval;
2314 * uint16_t spp_pathmaxrxt;
2315 * uint32_t spp_pathmtu;
2316 * uint32_t spp_sackdelay;
2317 * uint32_t spp_flags;
2320 * spp_assoc_id - (one-to-many style socket) This is filled in the
2321 * application, and identifies the association for
2323 * spp_address - This specifies which address is of interest.
2324 * spp_hbinterval - This contains the value of the heartbeat interval,
2325 * in milliseconds. If a value of zero
2326 * is present in this field then no changes are to
2327 * be made to this parameter.
2328 * spp_pathmaxrxt - This contains the maximum number of
2329 * retransmissions before this address shall be
2330 * considered unreachable. If a value of zero
2331 * is present in this field then no changes are to
2332 * be made to this parameter.
2333 * spp_pathmtu - When Path MTU discovery is disabled the value
2334 * specified here will be the "fixed" path mtu.
2335 * Note that if the spp_address field is empty
2336 * then all associations on this address will
2337 * have this fixed path mtu set upon them.
2339 * spp_sackdelay - When delayed sack is enabled, this value specifies
2340 * the number of milliseconds that sacks will be delayed
2341 * for. This value will apply to all addresses of an
2342 * association if the spp_address field is empty. Note
2343 * also, that if delayed sack is enabled and this
2344 * value is set to 0, no change is made to the last
2345 * recorded delayed sack timer value.
2347 * spp_flags - These flags are used to control various features
2348 * on an association. The flag field may contain
2349 * zero or more of the following options.
2351 * SPP_HB_ENABLE - Enable heartbeats on the
2352 * specified address. Note that if the address
2353 * field is empty all addresses for the association
2354 * have heartbeats enabled upon them.
2356 * SPP_HB_DISABLE - Disable heartbeats on the
2357 * speicifed address. Note that if the address
2358 * field is empty all addresses for the association
2359 * will have their heartbeats disabled. Note also
2360 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2361 * mutually exclusive, only one of these two should
2362 * be specified. Enabling both fields will have
2363 * undetermined results.
2365 * SPP_HB_DEMAND - Request a user initiated heartbeat
2366 * to be made immediately.
2368 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2369 * heartbeat delayis to be set to the value of 0
2372 * SPP_PMTUD_ENABLE - This field will enable PMTU
2373 * discovery upon the specified address. Note that
2374 * if the address feild is empty then all addresses
2375 * on the association are effected.
2377 * SPP_PMTUD_DISABLE - This field will disable PMTU
2378 * discovery upon the specified address. Note that
2379 * if the address feild is empty then all addresses
2380 * on the association are effected. Not also that
2381 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2382 * exclusive. Enabling both will have undetermined
2385 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2386 * on delayed sack. The time specified in spp_sackdelay
2387 * is used to specify the sack delay for this address. Note
2388 * that if spp_address is empty then all addresses will
2389 * enable delayed sack and take on the sack delay
2390 * value specified in spp_sackdelay.
2391 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2392 * off delayed sack. If the spp_address field is blank then
2393 * delayed sack is disabled for the entire association. Note
2394 * also that this field is mutually exclusive to
2395 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2398 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2399 struct sctp_transport *trans,
2400 struct sctp_association *asoc,
2401 struct sctp_sock *sp,
2404 int sackdelay_change)
2408 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2409 struct net *net = sock_net(trans->asoc->base.sk);
2411 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2416 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2417 * this field is ignored. Note also that a value of zero indicates
2418 * the current setting should be left unchanged.
2420 if (params->spp_flags & SPP_HB_ENABLE) {
2422 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2423 * set. This lets us use 0 value when this flag
2426 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2427 params->spp_hbinterval = 0;
2429 if (params->spp_hbinterval ||
2430 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2433 msecs_to_jiffies(params->spp_hbinterval);
2436 msecs_to_jiffies(params->spp_hbinterval);
2438 sp->hbinterval = params->spp_hbinterval;
2445 trans->param_flags =
2446 (trans->param_flags & ~SPP_HB) | hb_change;
2449 (asoc->param_flags & ~SPP_HB) | hb_change;
2452 (sp->param_flags & ~SPP_HB) | hb_change;
2456 /* When Path MTU discovery is disabled the value specified here will
2457 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2458 * include the flag SPP_PMTUD_DISABLE for this field to have any
2461 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2463 trans->pathmtu = params->spp_pathmtu;
2464 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2466 asoc->pathmtu = params->spp_pathmtu;
2467 sctp_frag_point(asoc, params->spp_pathmtu);
2469 sp->pathmtu = params->spp_pathmtu;
2475 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2476 (params->spp_flags & SPP_PMTUD_ENABLE);
2477 trans->param_flags =
2478 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2480 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2481 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2485 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2488 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2492 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2493 * value of this field is ignored. Note also that a value of zero
2494 * indicates the current setting should be left unchanged.
2496 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2499 msecs_to_jiffies(params->spp_sackdelay);
2502 msecs_to_jiffies(params->spp_sackdelay);
2504 sp->sackdelay = params->spp_sackdelay;
2508 if (sackdelay_change) {
2510 trans->param_flags =
2511 (trans->param_flags & ~SPP_SACKDELAY) |
2515 (asoc->param_flags & ~SPP_SACKDELAY) |
2519 (sp->param_flags & ~SPP_SACKDELAY) |
2524 /* Note that a value of zero indicates the current setting should be
2527 if (params->spp_pathmaxrxt) {
2529 trans->pathmaxrxt = params->spp_pathmaxrxt;
2531 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2533 sp->pathmaxrxt = params->spp_pathmaxrxt;
2540 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2541 char __user *optval,
2542 unsigned int optlen)
2544 struct sctp_paddrparams params;
2545 struct sctp_transport *trans = NULL;
2546 struct sctp_association *asoc = NULL;
2547 struct sctp_sock *sp = sctp_sk(sk);
2549 int hb_change, pmtud_change, sackdelay_change;
2551 if (optlen != sizeof(struct sctp_paddrparams))
2554 if (copy_from_user(¶ms, optval, optlen))
2557 /* Validate flags and value parameters. */
2558 hb_change = params.spp_flags & SPP_HB;
2559 pmtud_change = params.spp_flags & SPP_PMTUD;
2560 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2562 if (hb_change == SPP_HB ||
2563 pmtud_change == SPP_PMTUD ||
2564 sackdelay_change == SPP_SACKDELAY ||
2565 params.spp_sackdelay > 500 ||
2566 (params.spp_pathmtu &&
2567 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2570 /* If an address other than INADDR_ANY is specified, and
2571 * no transport is found, then the request is invalid.
2573 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
2574 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2575 params.spp_assoc_id);
2580 /* Get association, if assoc_id != 0 and the socket is a one
2581 * to many style socket, and an association was not found, then
2582 * the id was invalid.
2584 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2585 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2588 /* Heartbeat demand can only be sent on a transport or
2589 * association, but not a socket.
2591 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2594 /* Process parameters. */
2595 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2596 hb_change, pmtud_change,
2602 /* If changes are for association, also apply parameters to each
2605 if (!trans && asoc) {
2606 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2608 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2609 hb_change, pmtud_change,
2617 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2619 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2622 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2624 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2628 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2630 * This option will effect the way delayed acks are performed. This
2631 * option allows you to get or set the delayed ack time, in
2632 * milliseconds. It also allows changing the delayed ack frequency.
2633 * Changing the frequency to 1 disables the delayed sack algorithm. If
2634 * the assoc_id is 0, then this sets or gets the endpoints default
2635 * values. If the assoc_id field is non-zero, then the set or get
2636 * effects the specified association for the one to many model (the
2637 * assoc_id field is ignored by the one to one model). Note that if
2638 * sack_delay or sack_freq are 0 when setting this option, then the
2639 * current values will remain unchanged.
2641 * struct sctp_sack_info {
2642 * sctp_assoc_t sack_assoc_id;
2643 * uint32_t sack_delay;
2644 * uint32_t sack_freq;
2647 * sack_assoc_id - This parameter, indicates which association the user
2648 * is performing an action upon. Note that if this field's value is
2649 * zero then the endpoints default value is changed (effecting future
2650 * associations only).
2652 * sack_delay - This parameter contains the number of milliseconds that
2653 * the user is requesting the delayed ACK timer be set to. Note that
2654 * this value is defined in the standard to be between 200 and 500
2657 * sack_freq - This parameter contains the number of packets that must
2658 * be received before a sack is sent without waiting for the delay
2659 * timer to expire. The default value for this is 2, setting this
2660 * value to 1 will disable the delayed sack algorithm.
2663 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2664 char __user *optval, unsigned int optlen)
2666 struct sctp_sack_info params;
2667 struct sctp_transport *trans = NULL;
2668 struct sctp_association *asoc = NULL;
2669 struct sctp_sock *sp = sctp_sk(sk);
2671 if (optlen == sizeof(struct sctp_sack_info)) {
2672 if (copy_from_user(¶ms, optval, optlen))
2675 if (params.sack_delay == 0 && params.sack_freq == 0)
2677 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2678 pr_warn_ratelimited(DEPRECATED
2680 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2681 "Use struct sctp_sack_info instead\n",
2682 current->comm, task_pid_nr(current));
2683 if (copy_from_user(¶ms, optval, optlen))
2686 if (params.sack_delay == 0)
2687 params.sack_freq = 1;
2689 params.sack_freq = 0;
2693 /* Validate value parameter. */
2694 if (params.sack_delay > 500)
2697 /* Get association, if sack_assoc_id != 0 and the socket is a one
2698 * to many style socket, and an association was not found, then
2699 * the id was invalid.
2701 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2702 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2705 if (params.sack_delay) {
2708 msecs_to_jiffies(params.sack_delay);
2710 sctp_spp_sackdelay_enable(asoc->param_flags);
2712 sp->sackdelay = params.sack_delay;
2714 sctp_spp_sackdelay_enable(sp->param_flags);
2718 if (params.sack_freq == 1) {
2721 sctp_spp_sackdelay_disable(asoc->param_flags);
2724 sctp_spp_sackdelay_disable(sp->param_flags);
2726 } else if (params.sack_freq > 1) {
2728 asoc->sackfreq = params.sack_freq;
2730 sctp_spp_sackdelay_enable(asoc->param_flags);
2732 sp->sackfreq = params.sack_freq;
2734 sctp_spp_sackdelay_enable(sp->param_flags);
2738 /* If change is for association, also apply to each transport. */
2740 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2742 if (params.sack_delay) {
2744 msecs_to_jiffies(params.sack_delay);
2745 trans->param_flags =
2746 sctp_spp_sackdelay_enable(trans->param_flags);
2748 if (params.sack_freq == 1) {
2749 trans->param_flags =
2750 sctp_spp_sackdelay_disable(trans->param_flags);
2751 } else if (params.sack_freq > 1) {
2752 trans->sackfreq = params.sack_freq;
2753 trans->param_flags =
2754 sctp_spp_sackdelay_enable(trans->param_flags);
2762 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2764 * Applications can specify protocol parameters for the default association
2765 * initialization. The option name argument to setsockopt() and getsockopt()
2768 * Setting initialization parameters is effective only on an unconnected
2769 * socket (for UDP-style sockets only future associations are effected
2770 * by the change). With TCP-style sockets, this option is inherited by
2771 * sockets derived from a listener socket.
2773 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2775 struct sctp_initmsg sinit;
2776 struct sctp_sock *sp = sctp_sk(sk);
2778 if (optlen != sizeof(struct sctp_initmsg))
2780 if (copy_from_user(&sinit, optval, optlen))
2783 if (sinit.sinit_num_ostreams)
2784 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2785 if (sinit.sinit_max_instreams)
2786 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2787 if (sinit.sinit_max_attempts)
2788 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2789 if (sinit.sinit_max_init_timeo)
2790 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2796 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2798 * Applications that wish to use the sendto() system call may wish to
2799 * specify a default set of parameters that would normally be supplied
2800 * through the inclusion of ancillary data. This socket option allows
2801 * such an application to set the default sctp_sndrcvinfo structure.
2802 * The application that wishes to use this socket option simply passes
2803 * in to this call the sctp_sndrcvinfo structure defined in Section
2804 * 5.2.2) The input parameters accepted by this call include
2805 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2806 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2807 * to this call if the caller is using the UDP model.
2809 static int sctp_setsockopt_default_send_param(struct sock *sk,
2810 char __user *optval,
2811 unsigned int optlen)
2813 struct sctp_sock *sp = sctp_sk(sk);
2814 struct sctp_association *asoc;
2815 struct sctp_sndrcvinfo info;
2817 if (optlen != sizeof(info))
2819 if (copy_from_user(&info, optval, optlen))
2821 if (info.sinfo_flags &
2822 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2823 SCTP_ABORT | SCTP_EOF))
2826 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2827 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2830 asoc->default_stream = info.sinfo_stream;
2831 asoc->default_flags = info.sinfo_flags;
2832 asoc->default_ppid = info.sinfo_ppid;
2833 asoc->default_context = info.sinfo_context;
2834 asoc->default_timetolive = info.sinfo_timetolive;
2836 sp->default_stream = info.sinfo_stream;
2837 sp->default_flags = info.sinfo_flags;
2838 sp->default_ppid = info.sinfo_ppid;
2839 sp->default_context = info.sinfo_context;
2840 sp->default_timetolive = info.sinfo_timetolive;
2846 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2847 * (SCTP_DEFAULT_SNDINFO)
2849 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
2850 char __user *optval,
2851 unsigned int optlen)
2853 struct sctp_sock *sp = sctp_sk(sk);
2854 struct sctp_association *asoc;
2855 struct sctp_sndinfo info;
2857 if (optlen != sizeof(info))
2859 if (copy_from_user(&info, optval, optlen))
2861 if (info.snd_flags &
2862 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2863 SCTP_ABORT | SCTP_EOF))
2866 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
2867 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
2870 asoc->default_stream = info.snd_sid;
2871 asoc->default_flags = info.snd_flags;
2872 asoc->default_ppid = info.snd_ppid;
2873 asoc->default_context = info.snd_context;
2875 sp->default_stream = info.snd_sid;
2876 sp->default_flags = info.snd_flags;
2877 sp->default_ppid = info.snd_ppid;
2878 sp->default_context = info.snd_context;
2884 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2886 * Requests that the local SCTP stack use the enclosed peer address as
2887 * the association primary. The enclosed address must be one of the
2888 * association peer's addresses.
2890 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2891 unsigned int optlen)
2893 struct sctp_prim prim;
2894 struct sctp_transport *trans;
2896 if (optlen != sizeof(struct sctp_prim))
2899 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2902 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2906 sctp_assoc_set_primary(trans->asoc, trans);
2912 * 7.1.5 SCTP_NODELAY
2914 * Turn on/off any Nagle-like algorithm. This means that packets are
2915 * generally sent as soon as possible and no unnecessary delays are
2916 * introduced, at the cost of more packets in the network. Expects an
2917 * integer boolean flag.
2919 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2920 unsigned int optlen)
2924 if (optlen < sizeof(int))
2926 if (get_user(val, (int __user *)optval))
2929 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2935 * 7.1.1 SCTP_RTOINFO
2937 * The protocol parameters used to initialize and bound retransmission
2938 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2939 * and modify these parameters.
2940 * All parameters are time values, in milliseconds. A value of 0, when
2941 * modifying the parameters, indicates that the current value should not
2945 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2947 struct sctp_rtoinfo rtoinfo;
2948 struct sctp_association *asoc;
2949 unsigned long rto_min, rto_max;
2950 struct sctp_sock *sp = sctp_sk(sk);
2952 if (optlen != sizeof (struct sctp_rtoinfo))
2955 if (copy_from_user(&rtoinfo, optval, optlen))
2958 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2960 /* Set the values to the specific association */
2961 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2964 rto_max = rtoinfo.srto_max;
2965 rto_min = rtoinfo.srto_min;
2968 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2970 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2973 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2975 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2977 if (rto_min > rto_max)
2981 if (rtoinfo.srto_initial != 0)
2983 msecs_to_jiffies(rtoinfo.srto_initial);
2984 asoc->rto_max = rto_max;
2985 asoc->rto_min = rto_min;
2987 /* If there is no association or the association-id = 0
2988 * set the values to the endpoint.
2990 if (rtoinfo.srto_initial != 0)
2991 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2992 sp->rtoinfo.srto_max = rto_max;
2993 sp->rtoinfo.srto_min = rto_min;
3001 * 7.1.2 SCTP_ASSOCINFO
3003 * This option is used to tune the maximum retransmission attempts
3004 * of the association.
3005 * Returns an error if the new association retransmission value is
3006 * greater than the sum of the retransmission value of the peer.
3007 * See [SCTP] for more information.
3010 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
3013 struct sctp_assocparams assocparams;
3014 struct sctp_association *asoc;
3016 if (optlen != sizeof(struct sctp_assocparams))
3018 if (copy_from_user(&assocparams, optval, optlen))
3021 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
3023 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
3026 /* Set the values to the specific association */
3028 if (assocparams.sasoc_asocmaxrxt != 0) {
3031 struct sctp_transport *peer_addr;
3033 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
3035 path_sum += peer_addr->pathmaxrxt;
3039 /* Only validate asocmaxrxt if we have more than
3040 * one path/transport. We do this because path
3041 * retransmissions are only counted when we have more
3045 assocparams.sasoc_asocmaxrxt > path_sum)
3048 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3051 if (assocparams.sasoc_cookie_life != 0)
3052 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3054 /* Set the values to the endpoint */
3055 struct sctp_sock *sp = sctp_sk(sk);
3057 if (assocparams.sasoc_asocmaxrxt != 0)
3058 sp->assocparams.sasoc_asocmaxrxt =
3059 assocparams.sasoc_asocmaxrxt;
3060 if (assocparams.sasoc_cookie_life != 0)
3061 sp->assocparams.sasoc_cookie_life =
3062 assocparams.sasoc_cookie_life;
3068 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3070 * This socket option is a boolean flag which turns on or off mapped V4
3071 * addresses. If this option is turned on and the socket is type
3072 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3073 * If this option is turned off, then no mapping will be done of V4
3074 * addresses and a user will receive both PF_INET6 and PF_INET type
3075 * addresses on the socket.
3077 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3080 struct sctp_sock *sp = sctp_sk(sk);
3082 if (optlen < sizeof(int))
3084 if (get_user(val, (int __user *)optval))
3095 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3096 * This option will get or set the maximum size to put in any outgoing
3097 * SCTP DATA chunk. If a message is larger than this size it will be
3098 * fragmented by SCTP into the specified size. Note that the underlying
3099 * SCTP implementation may fragment into smaller sized chunks when the
3100 * PMTU of the underlying association is smaller than the value set by
3101 * the user. The default value for this option is '0' which indicates
3102 * the user is NOT limiting fragmentation and only the PMTU will effect
3103 * SCTP's choice of DATA chunk size. Note also that values set larger
3104 * than the maximum size of an IP datagram will effectively let SCTP
3105 * control fragmentation (i.e. the same as setting this option to 0).
3107 * The following structure is used to access and modify this parameter:
3109 * struct sctp_assoc_value {
3110 * sctp_assoc_t assoc_id;
3111 * uint32_t assoc_value;
3114 * assoc_id: This parameter is ignored for one-to-one style sockets.
3115 * For one-to-many style sockets this parameter indicates which
3116 * association the user is performing an action upon. Note that if
3117 * this field's value is zero then the endpoints default value is
3118 * changed (effecting future associations only).
3119 * assoc_value: This parameter specifies the maximum size in bytes.
3121 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3123 struct sctp_assoc_value params;
3124 struct sctp_association *asoc;
3125 struct sctp_sock *sp = sctp_sk(sk);
3128 if (optlen == sizeof(int)) {
3129 pr_warn_ratelimited(DEPRECATED
3131 "Use of int in maxseg socket option.\n"
3132 "Use struct sctp_assoc_value instead\n",
3133 current->comm, task_pid_nr(current));
3134 if (copy_from_user(&val, optval, optlen))
3136 params.assoc_id = 0;
3137 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3138 if (copy_from_user(¶ms, optval, optlen))
3140 val = params.assoc_value;
3144 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3147 asoc = sctp_id2assoc(sk, params.assoc_id);
3148 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3153 val = asoc->pathmtu;
3154 val -= sp->pf->af->net_header_len;
3155 val -= sizeof(struct sctphdr) +
3156 sizeof(struct sctp_data_chunk);
3158 asoc->user_frag = val;
3159 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3161 sp->user_frag = val;
3169 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3171 * Requests that the peer mark the enclosed address as the association
3172 * primary. The enclosed address must be one of the association's
3173 * locally bound addresses. The following structure is used to make a
3174 * set primary request:
3176 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3177 unsigned int optlen)
3179 struct net *net = sock_net(sk);
3180 struct sctp_sock *sp;
3181 struct sctp_association *asoc = NULL;
3182 struct sctp_setpeerprim prim;
3183 struct sctp_chunk *chunk;
3189 if (!net->sctp.addip_enable)
3192 if (optlen != sizeof(struct sctp_setpeerprim))
3195 if (copy_from_user(&prim, optval, optlen))
3198 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3202 if (!asoc->peer.asconf_capable)
3205 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3208 if (!sctp_state(asoc, ESTABLISHED))
3211 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3215 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3216 return -EADDRNOTAVAIL;
3218 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3219 return -EADDRNOTAVAIL;
3221 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3222 chunk = sctp_make_asconf_set_prim(asoc,
3223 (union sctp_addr *)&prim.sspp_addr);
3227 err = sctp_send_asconf(asoc, chunk);
3229 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3234 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3235 unsigned int optlen)
3237 struct sctp_setadaptation adaptation;
3239 if (optlen != sizeof(struct sctp_setadaptation))
3241 if (copy_from_user(&adaptation, optval, optlen))
3244 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3250 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3252 * The context field in the sctp_sndrcvinfo structure is normally only
3253 * used when a failed message is retrieved holding the value that was
3254 * sent down on the actual send call. This option allows the setting of
3255 * a default context on an association basis that will be received on
3256 * reading messages from the peer. This is especially helpful in the
3257 * one-2-many model for an application to keep some reference to an
3258 * internal state machine that is processing messages on the
3259 * association. Note that the setting of this value only effects
3260 * received messages from the peer and does not effect the value that is
3261 * saved with outbound messages.
3263 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3264 unsigned int optlen)
3266 struct sctp_assoc_value params;
3267 struct sctp_sock *sp;
3268 struct sctp_association *asoc;
3270 if (optlen != sizeof(struct sctp_assoc_value))
3272 if (copy_from_user(¶ms, optval, optlen))
3277 if (params.assoc_id != 0) {
3278 asoc = sctp_id2assoc(sk, params.assoc_id);
3281 asoc->default_rcv_context = params.assoc_value;
3283 sp->default_rcv_context = params.assoc_value;
3290 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3292 * This options will at a minimum specify if the implementation is doing
3293 * fragmented interleave. Fragmented interleave, for a one to many
3294 * socket, is when subsequent calls to receive a message may return
3295 * parts of messages from different associations. Some implementations
3296 * may allow you to turn this value on or off. If so, when turned off,
3297 * no fragment interleave will occur (which will cause a head of line
3298 * blocking amongst multiple associations sharing the same one to many
3299 * socket). When this option is turned on, then each receive call may
3300 * come from a different association (thus the user must receive data
3301 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3302 * association each receive belongs to.
3304 * This option takes a boolean value. A non-zero value indicates that
3305 * fragmented interleave is on. A value of zero indicates that
3306 * fragmented interleave is off.
3308 * Note that it is important that an implementation that allows this
3309 * option to be turned on, have it off by default. Otherwise an unaware
3310 * application using the one to many model may become confused and act
3313 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3314 char __user *optval,
3315 unsigned int optlen)
3319 if (optlen != sizeof(int))
3321 if (get_user(val, (int __user *)optval))
3324 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3330 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3331 * (SCTP_PARTIAL_DELIVERY_POINT)
3333 * This option will set or get the SCTP partial delivery point. This
3334 * point is the size of a message where the partial delivery API will be
3335 * invoked to help free up rwnd space for the peer. Setting this to a
3336 * lower value will cause partial deliveries to happen more often. The
3337 * calls argument is an integer that sets or gets the partial delivery
3338 * point. Note also that the call will fail if the user attempts to set
3339 * this value larger than the socket receive buffer size.
3341 * Note that any single message having a length smaller than or equal to
3342 * the SCTP partial delivery point will be delivered in one single read
3343 * call as long as the user provided buffer is large enough to hold the
3346 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3347 char __user *optval,
3348 unsigned int optlen)
3352 if (optlen != sizeof(u32))
3354 if (get_user(val, (int __user *)optval))
3357 /* Note: We double the receive buffer from what the user sets
3358 * it to be, also initial rwnd is based on rcvbuf/2.
3360 if (val > (sk->sk_rcvbuf >> 1))
3363 sctp_sk(sk)->pd_point = val;
3365 return 0; /* is this the right error code? */
3369 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3371 * This option will allow a user to change the maximum burst of packets
3372 * that can be emitted by this association. Note that the default value
3373 * is 4, and some implementations may restrict this setting so that it
3374 * can only be lowered.
3376 * NOTE: This text doesn't seem right. Do this on a socket basis with
3377 * future associations inheriting the socket value.
3379 static int sctp_setsockopt_maxburst(struct sock *sk,
3380 char __user *optval,
3381 unsigned int optlen)
3383 struct sctp_assoc_value params;
3384 struct sctp_sock *sp;
3385 struct sctp_association *asoc;
3389 if (optlen == sizeof(int)) {
3390 pr_warn_ratelimited(DEPRECATED
3392 "Use of int in max_burst socket option deprecated.\n"
3393 "Use struct sctp_assoc_value instead\n",
3394 current->comm, task_pid_nr(current));
3395 if (copy_from_user(&val, optval, optlen))
3397 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3398 if (copy_from_user(¶ms, optval, optlen))
3400 val = params.assoc_value;
3401 assoc_id = params.assoc_id;
3407 if (assoc_id != 0) {
3408 asoc = sctp_id2assoc(sk, assoc_id);
3411 asoc->max_burst = val;
3413 sp->max_burst = val;
3419 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3421 * This set option adds a chunk type that the user is requesting to be
3422 * received only in an authenticated way. Changes to the list of chunks
3423 * will only effect future associations on the socket.
3425 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3426 char __user *optval,
3427 unsigned int optlen)
3429 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3430 struct sctp_authchunk val;
3432 if (!ep->auth_enable)
3435 if (optlen != sizeof(struct sctp_authchunk))
3437 if (copy_from_user(&val, optval, optlen))
3440 switch (val.sauth_chunk) {
3442 case SCTP_CID_INIT_ACK:
3443 case SCTP_CID_SHUTDOWN_COMPLETE:
3448 /* add this chunk id to the endpoint */
3449 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3453 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3455 * This option gets or sets the list of HMAC algorithms that the local
3456 * endpoint requires the peer to use.
3458 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3459 char __user *optval,
3460 unsigned int optlen)
3462 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3463 struct sctp_hmacalgo *hmacs;
3467 if (!ep->auth_enable)
3470 if (optlen < sizeof(struct sctp_hmacalgo))
3473 hmacs = memdup_user(optval, optlen);
3475 return PTR_ERR(hmacs);
3477 idents = hmacs->shmac_num_idents;
3478 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3479 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3484 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3491 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3493 * This option will set a shared secret key which is used to build an
3494 * association shared key.
3496 static int sctp_setsockopt_auth_key(struct sock *sk,
3497 char __user *optval,
3498 unsigned int optlen)
3500 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3501 struct sctp_authkey *authkey;
3502 struct sctp_association *asoc;
3505 if (!ep->auth_enable)
3508 if (optlen <= sizeof(struct sctp_authkey))
3511 authkey = memdup_user(optval, optlen);
3512 if (IS_ERR(authkey))
3513 return PTR_ERR(authkey);
3515 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3520 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3521 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3526 ret = sctp_auth_set_key(ep, asoc, authkey);
3533 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3535 * This option will get or set the active shared key to be used to build
3536 * the association shared key.
3538 static int sctp_setsockopt_active_key(struct sock *sk,
3539 char __user *optval,
3540 unsigned int optlen)
3542 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3543 struct sctp_authkeyid val;
3544 struct sctp_association *asoc;
3546 if (!ep->auth_enable)
3549 if (optlen != sizeof(struct sctp_authkeyid))
3551 if (copy_from_user(&val, optval, optlen))
3554 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3555 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3558 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3562 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3564 * This set option will delete a shared secret key from use.
3566 static int sctp_setsockopt_del_key(struct sock *sk,
3567 char __user *optval,
3568 unsigned int optlen)
3570 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3571 struct sctp_authkeyid val;
3572 struct sctp_association *asoc;
3574 if (!ep->auth_enable)
3577 if (optlen != sizeof(struct sctp_authkeyid))
3579 if (copy_from_user(&val, optval, optlen))
3582 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3583 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3586 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3591 * 8.1.23 SCTP_AUTO_ASCONF
3593 * This option will enable or disable the use of the automatic generation of
3594 * ASCONF chunks to add and delete addresses to an existing association. Note
3595 * that this option has two caveats namely: a) it only affects sockets that
3596 * are bound to all addresses available to the SCTP stack, and b) the system
3597 * administrator may have an overriding control that turns the ASCONF feature
3598 * off no matter what setting the socket option may have.
3599 * This option expects an integer boolean flag, where a non-zero value turns on
3600 * the option, and a zero value turns off the option.
3601 * Note. In this implementation, socket operation overrides default parameter
3602 * being set by sysctl as well as FreeBSD implementation
3604 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3605 unsigned int optlen)
3608 struct sctp_sock *sp = sctp_sk(sk);
3610 if (optlen < sizeof(int))
3612 if (get_user(val, (int __user *)optval))
3614 if (!sctp_is_ep_boundall(sk) && val)
3616 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3619 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3620 if (val == 0 && sp->do_auto_asconf) {
3621 list_del(&sp->auto_asconf_list);
3622 sp->do_auto_asconf = 0;
3623 } else if (val && !sp->do_auto_asconf) {
3624 list_add_tail(&sp->auto_asconf_list,
3625 &sock_net(sk)->sctp.auto_asconf_splist);
3626 sp->do_auto_asconf = 1;
3628 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3633 * SCTP_PEER_ADDR_THLDS
3635 * This option allows us to alter the partially failed threshold for one or all
3636 * transports in an association. See Section 6.1 of:
3637 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3639 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3640 char __user *optval,
3641 unsigned int optlen)
3643 struct sctp_paddrthlds val;
3644 struct sctp_transport *trans;
3645 struct sctp_association *asoc;
3647 if (optlen < sizeof(struct sctp_paddrthlds))
3649 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3650 sizeof(struct sctp_paddrthlds)))
3654 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3655 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3658 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3660 if (val.spt_pathmaxrxt)
3661 trans->pathmaxrxt = val.spt_pathmaxrxt;
3662 trans->pf_retrans = val.spt_pathpfthld;
3665 if (val.spt_pathmaxrxt)
3666 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3667 asoc->pf_retrans = val.spt_pathpfthld;
3669 trans = sctp_addr_id2transport(sk, &val.spt_address,
3674 if (val.spt_pathmaxrxt)
3675 trans->pathmaxrxt = val.spt_pathmaxrxt;
3676 trans->pf_retrans = val.spt_pathpfthld;
3682 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3683 char __user *optval,
3684 unsigned int optlen)
3688 if (optlen < sizeof(int))
3690 if (get_user(val, (int __user *) optval))
3693 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3698 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3699 char __user *optval,
3700 unsigned int optlen)
3704 if (optlen < sizeof(int))
3706 if (get_user(val, (int __user *) optval))
3709 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3714 static int sctp_setsockopt_pr_supported(struct sock *sk,
3715 char __user *optval,
3716 unsigned int optlen)
3718 struct sctp_assoc_value params;
3719 struct sctp_association *asoc;
3720 int retval = -EINVAL;
3722 if (optlen != sizeof(params))
3725 if (copy_from_user(¶ms, optval, optlen)) {
3730 asoc = sctp_id2assoc(sk, params.assoc_id);
3732 asoc->prsctp_enable = !!params.assoc_value;
3733 } else if (!params.assoc_id) {
3734 struct sctp_sock *sp = sctp_sk(sk);
3736 sp->ep->prsctp_enable = !!params.assoc_value;
3747 static int sctp_setsockopt_default_prinfo(struct sock *sk,
3748 char __user *optval,
3749 unsigned int optlen)
3751 struct sctp_default_prinfo info;
3752 struct sctp_association *asoc;
3753 int retval = -EINVAL;
3755 if (optlen != sizeof(info))
3758 if (copy_from_user(&info, optval, sizeof(info))) {
3763 if (info.pr_policy & ~SCTP_PR_SCTP_MASK)
3766 if (info.pr_policy == SCTP_PR_SCTP_NONE)
3769 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
3771 SCTP_PR_SET_POLICY(asoc->default_flags, info.pr_policy);
3772 asoc->default_timetolive = info.pr_value;
3773 } else if (!info.pr_assoc_id) {
3774 struct sctp_sock *sp = sctp_sk(sk);
3776 SCTP_PR_SET_POLICY(sp->default_flags, info.pr_policy);
3777 sp->default_timetolive = info.pr_value;
3788 /* API 6.2 setsockopt(), getsockopt()
3790 * Applications use setsockopt() and getsockopt() to set or retrieve
3791 * socket options. Socket options are used to change the default
3792 * behavior of sockets calls. They are described in Section 7.
3796 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3797 * int __user *optlen);
3798 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3801 * sd - the socket descript.
3802 * level - set to IPPROTO_SCTP for all SCTP options.
3803 * optname - the option name.
3804 * optval - the buffer to store the value of the option.
3805 * optlen - the size of the buffer.
3807 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3808 char __user *optval, unsigned int optlen)
3812 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3814 /* I can hardly begin to describe how wrong this is. This is
3815 * so broken as to be worse than useless. The API draft
3816 * REALLY is NOT helpful here... I am not convinced that the
3817 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3818 * are at all well-founded.
3820 if (level != SOL_SCTP) {
3821 struct sctp_af *af = sctp_sk(sk)->pf->af;
3822 retval = af->setsockopt(sk, level, optname, optval, optlen);
3829 case SCTP_SOCKOPT_BINDX_ADD:
3830 /* 'optlen' is the size of the addresses buffer. */
3831 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3832 optlen, SCTP_BINDX_ADD_ADDR);
3835 case SCTP_SOCKOPT_BINDX_REM:
3836 /* 'optlen' is the size of the addresses buffer. */
3837 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3838 optlen, SCTP_BINDX_REM_ADDR);
3841 case SCTP_SOCKOPT_CONNECTX_OLD:
3842 /* 'optlen' is the size of the addresses buffer. */
3843 retval = sctp_setsockopt_connectx_old(sk,
3844 (struct sockaddr __user *)optval,
3848 case SCTP_SOCKOPT_CONNECTX:
3849 /* 'optlen' is the size of the addresses buffer. */
3850 retval = sctp_setsockopt_connectx(sk,
3851 (struct sockaddr __user *)optval,
3855 case SCTP_DISABLE_FRAGMENTS:
3856 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3860 retval = sctp_setsockopt_events(sk, optval, optlen);
3863 case SCTP_AUTOCLOSE:
3864 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3867 case SCTP_PEER_ADDR_PARAMS:
3868 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3871 case SCTP_DELAYED_SACK:
3872 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3874 case SCTP_PARTIAL_DELIVERY_POINT:
3875 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3879 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3881 case SCTP_DEFAULT_SEND_PARAM:
3882 retval = sctp_setsockopt_default_send_param(sk, optval,
3885 case SCTP_DEFAULT_SNDINFO:
3886 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
3888 case SCTP_PRIMARY_ADDR:
3889 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3891 case SCTP_SET_PEER_PRIMARY_ADDR:
3892 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3895 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3898 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3900 case SCTP_ASSOCINFO:
3901 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3903 case SCTP_I_WANT_MAPPED_V4_ADDR:
3904 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3907 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3909 case SCTP_ADAPTATION_LAYER:
3910 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3913 retval = sctp_setsockopt_context(sk, optval, optlen);
3915 case SCTP_FRAGMENT_INTERLEAVE:
3916 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3918 case SCTP_MAX_BURST:
3919 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3921 case SCTP_AUTH_CHUNK:
3922 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3924 case SCTP_HMAC_IDENT:
3925 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3928 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3930 case SCTP_AUTH_ACTIVE_KEY:
3931 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3933 case SCTP_AUTH_DELETE_KEY:
3934 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3936 case SCTP_AUTO_ASCONF:
3937 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3939 case SCTP_PEER_ADDR_THLDS:
3940 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3942 case SCTP_RECVRCVINFO:
3943 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
3945 case SCTP_RECVNXTINFO:
3946 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
3948 case SCTP_PR_SUPPORTED:
3949 retval = sctp_setsockopt_pr_supported(sk, optval, optlen);
3951 case SCTP_DEFAULT_PRINFO:
3952 retval = sctp_setsockopt_default_prinfo(sk, optval, optlen);
3955 retval = -ENOPROTOOPT;
3965 /* API 3.1.6 connect() - UDP Style Syntax
3967 * An application may use the connect() call in the UDP model to initiate an
3968 * association without sending data.
3972 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3974 * sd: the socket descriptor to have a new association added to.
3976 * nam: the address structure (either struct sockaddr_in or struct
3977 * sockaddr_in6 defined in RFC2553 [7]).
3979 * len: the size of the address.
3981 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3989 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
3992 /* Validate addr_len before calling common connect/connectx routine. */
3993 af = sctp_get_af_specific(addr->sa_family);
3994 if (!af || addr_len < af->sockaddr_len) {
3997 /* Pass correct addr len to common routine (so it knows there
3998 * is only one address being passed.
4000 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
4007 /* FIXME: Write comments. */
4008 static int sctp_disconnect(struct sock *sk, int flags)
4010 return -EOPNOTSUPP; /* STUB */
4013 /* 4.1.4 accept() - TCP Style Syntax
4015 * Applications use accept() call to remove an established SCTP
4016 * association from the accept queue of the endpoint. A new socket
4017 * descriptor will be returned from accept() to represent the newly
4018 * formed association.
4020 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
4022 struct sctp_sock *sp;
4023 struct sctp_endpoint *ep;
4024 struct sock *newsk = NULL;
4025 struct sctp_association *asoc;
4034 if (!sctp_style(sk, TCP)) {
4035 error = -EOPNOTSUPP;
4039 if (!sctp_sstate(sk, LISTENING)) {
4044 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
4046 error = sctp_wait_for_accept(sk, timeo);
4050 /* We treat the list of associations on the endpoint as the accept
4051 * queue and pick the first association on the list.
4053 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
4055 newsk = sp->pf->create_accept_sk(sk, asoc);
4061 /* Populate the fields of the newsk from the oldsk and migrate the
4062 * asoc to the newsk.
4064 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
4072 /* The SCTP ioctl handler. */
4073 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
4080 * SEQPACKET-style sockets in LISTENING state are valid, for
4081 * SCTP, so only discard TCP-style sockets in LISTENING state.
4083 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4088 struct sk_buff *skb;
4089 unsigned int amount = 0;
4091 skb = skb_peek(&sk->sk_receive_queue);
4094 * We will only return the amount of this packet since
4095 * that is all that will be read.
4099 rc = put_user(amount, (int __user *)arg);
4111 /* This is the function which gets called during socket creation to
4112 * initialized the SCTP-specific portion of the sock.
4113 * The sock structure should already be zero-filled memory.
4115 static int sctp_init_sock(struct sock *sk)
4117 struct net *net = sock_net(sk);
4118 struct sctp_sock *sp;
4120 pr_debug("%s: sk:%p\n", __func__, sk);
4124 /* Initialize the SCTP per socket area. */
4125 switch (sk->sk_type) {
4126 case SOCK_SEQPACKET:
4127 sp->type = SCTP_SOCKET_UDP;
4130 sp->type = SCTP_SOCKET_TCP;
4133 return -ESOCKTNOSUPPORT;
4136 sk->sk_gso_type = SKB_GSO_SCTP;
4138 /* Initialize default send parameters. These parameters can be
4139 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4141 sp->default_stream = 0;
4142 sp->default_ppid = 0;
4143 sp->default_flags = 0;
4144 sp->default_context = 0;
4145 sp->default_timetolive = 0;
4147 sp->default_rcv_context = 0;
4148 sp->max_burst = net->sctp.max_burst;
4150 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4152 /* Initialize default setup parameters. These parameters
4153 * can be modified with the SCTP_INITMSG socket option or
4154 * overridden by the SCTP_INIT CMSG.
4156 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4157 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4158 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4159 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4161 /* Initialize default RTO related parameters. These parameters can
4162 * be modified for with the SCTP_RTOINFO socket option.
4164 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4165 sp->rtoinfo.srto_max = net->sctp.rto_max;
4166 sp->rtoinfo.srto_min = net->sctp.rto_min;
4168 /* Initialize default association related parameters. These parameters
4169 * can be modified with the SCTP_ASSOCINFO socket option.
4171 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4172 sp->assocparams.sasoc_number_peer_destinations = 0;
4173 sp->assocparams.sasoc_peer_rwnd = 0;
4174 sp->assocparams.sasoc_local_rwnd = 0;
4175 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4177 /* Initialize default event subscriptions. By default, all the
4180 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4182 /* Default Peer Address Parameters. These defaults can
4183 * be modified via SCTP_PEER_ADDR_PARAMS
4185 sp->hbinterval = net->sctp.hb_interval;
4186 sp->pathmaxrxt = net->sctp.max_retrans_path;
4187 sp->pathmtu = 0; /* allow default discovery */
4188 sp->sackdelay = net->sctp.sack_timeout;
4190 sp->param_flags = SPP_HB_ENABLE |
4192 SPP_SACKDELAY_ENABLE;
4194 /* If enabled no SCTP message fragmentation will be performed.
4195 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4197 sp->disable_fragments = 0;
4199 /* Enable Nagle algorithm by default. */
4202 sp->recvrcvinfo = 0;
4203 sp->recvnxtinfo = 0;
4205 /* Enable by default. */
4208 /* Auto-close idle associations after the configured
4209 * number of seconds. A value of 0 disables this
4210 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4211 * for UDP-style sockets only.
4215 /* User specified fragmentation limit. */
4218 sp->adaptation_ind = 0;
4220 sp->pf = sctp_get_pf_specific(sk->sk_family);
4222 /* Control variables for partial data delivery. */
4223 atomic_set(&sp->pd_mode, 0);
4224 skb_queue_head_init(&sp->pd_lobby);
4225 sp->frag_interleave = 0;
4227 /* Create a per socket endpoint structure. Even if we
4228 * change the data structure relationships, this may still
4229 * be useful for storing pre-connect address information.
4231 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4237 sk->sk_destruct = sctp_destruct_sock;
4239 SCTP_DBG_OBJCNT_INC(sock);
4242 percpu_counter_inc(&sctp_sockets_allocated);
4243 sock_prot_inuse_add(net, sk->sk_prot, 1);
4245 /* Nothing can fail after this block, otherwise
4246 * sctp_destroy_sock() will be called without addr_wq_lock held
4248 if (net->sctp.default_auto_asconf) {
4249 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
4250 list_add_tail(&sp->auto_asconf_list,
4251 &net->sctp.auto_asconf_splist);
4252 sp->do_auto_asconf = 1;
4253 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4255 sp->do_auto_asconf = 0;
4263 /* Cleanup any SCTP per socket resources. Must be called with
4264 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4266 static void sctp_destroy_sock(struct sock *sk)
4268 struct sctp_sock *sp;
4270 pr_debug("%s: sk:%p\n", __func__, sk);
4272 /* Release our hold on the endpoint. */
4274 /* This could happen during socket init, thus we bail out
4275 * early, since the rest of the below is not setup either.
4280 if (sp->do_auto_asconf) {
4281 sp->do_auto_asconf = 0;
4282 list_del(&sp->auto_asconf_list);
4284 sctp_endpoint_free(sp->ep);
4286 percpu_counter_dec(&sctp_sockets_allocated);
4287 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4291 /* Triggered when there are no references on the socket anymore */
4292 static void sctp_destruct_sock(struct sock *sk)
4294 struct sctp_sock *sp = sctp_sk(sk);
4296 /* Free up the HMAC transform. */
4297 crypto_free_shash(sp->hmac);
4299 inet_sock_destruct(sk);
4302 /* API 4.1.7 shutdown() - TCP Style Syntax
4303 * int shutdown(int socket, int how);
4305 * sd - the socket descriptor of the association to be closed.
4306 * how - Specifies the type of shutdown. The values are
4309 * Disables further receive operations. No SCTP
4310 * protocol action is taken.
4312 * Disables further send operations, and initiates
4313 * the SCTP shutdown sequence.
4315 * Disables further send and receive operations
4316 * and initiates the SCTP shutdown sequence.
4318 static void sctp_shutdown(struct sock *sk, int how)
4320 struct net *net = sock_net(sk);
4321 struct sctp_endpoint *ep;
4323 if (!sctp_style(sk, TCP))
4326 ep = sctp_sk(sk)->ep;
4327 if (how & SEND_SHUTDOWN && !list_empty(&ep->asocs)) {
4328 struct sctp_association *asoc;
4330 sk->sk_state = SCTP_SS_CLOSING;
4331 asoc = list_entry(ep->asocs.next,
4332 struct sctp_association, asocs);
4333 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4337 int sctp_get_sctp_info(struct sock *sk, struct sctp_association *asoc,
4338 struct sctp_info *info)
4340 struct sctp_transport *prim;
4341 struct list_head *pos;
4344 memset(info, 0, sizeof(*info));
4346 struct sctp_sock *sp = sctp_sk(sk);
4348 info->sctpi_s_autoclose = sp->autoclose;
4349 info->sctpi_s_adaptation_ind = sp->adaptation_ind;
4350 info->sctpi_s_pd_point = sp->pd_point;
4351 info->sctpi_s_nodelay = sp->nodelay;
4352 info->sctpi_s_disable_fragments = sp->disable_fragments;
4353 info->sctpi_s_v4mapped = sp->v4mapped;
4354 info->sctpi_s_frag_interleave = sp->frag_interleave;
4355 info->sctpi_s_type = sp->type;
4360 info->sctpi_tag = asoc->c.my_vtag;
4361 info->sctpi_state = asoc->state;
4362 info->sctpi_rwnd = asoc->a_rwnd;
4363 info->sctpi_unackdata = asoc->unack_data;
4364 info->sctpi_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4365 info->sctpi_instrms = asoc->c.sinit_max_instreams;
4366 info->sctpi_outstrms = asoc->c.sinit_num_ostreams;
4367 list_for_each(pos, &asoc->base.inqueue.in_chunk_list)
4368 info->sctpi_inqueue++;
4369 list_for_each(pos, &asoc->outqueue.out_chunk_list)
4370 info->sctpi_outqueue++;
4371 info->sctpi_overall_error = asoc->overall_error_count;
4372 info->sctpi_max_burst = asoc->max_burst;
4373 info->sctpi_maxseg = asoc->frag_point;
4374 info->sctpi_peer_rwnd = asoc->peer.rwnd;
4375 info->sctpi_peer_tag = asoc->c.peer_vtag;
4377 mask = asoc->peer.ecn_capable << 1;
4378 mask = (mask | asoc->peer.ipv4_address) << 1;
4379 mask = (mask | asoc->peer.ipv6_address) << 1;
4380 mask = (mask | asoc->peer.hostname_address) << 1;
4381 mask = (mask | asoc->peer.asconf_capable) << 1;
4382 mask = (mask | asoc->peer.prsctp_capable) << 1;
4383 mask = (mask | asoc->peer.auth_capable);
4384 info->sctpi_peer_capable = mask;
4385 mask = asoc->peer.sack_needed << 1;
4386 mask = (mask | asoc->peer.sack_generation) << 1;
4387 mask = (mask | asoc->peer.zero_window_announced);
4388 info->sctpi_peer_sack = mask;
4390 info->sctpi_isacks = asoc->stats.isacks;
4391 info->sctpi_osacks = asoc->stats.osacks;
4392 info->sctpi_opackets = asoc->stats.opackets;
4393 info->sctpi_ipackets = asoc->stats.ipackets;
4394 info->sctpi_rtxchunks = asoc->stats.rtxchunks;
4395 info->sctpi_outofseqtsns = asoc->stats.outofseqtsns;
4396 info->sctpi_idupchunks = asoc->stats.idupchunks;
4397 info->sctpi_gapcnt = asoc->stats.gapcnt;
4398 info->sctpi_ouodchunks = asoc->stats.ouodchunks;
4399 info->sctpi_iuodchunks = asoc->stats.iuodchunks;
4400 info->sctpi_oodchunks = asoc->stats.oodchunks;
4401 info->sctpi_iodchunks = asoc->stats.iodchunks;
4402 info->sctpi_octrlchunks = asoc->stats.octrlchunks;
4403 info->sctpi_ictrlchunks = asoc->stats.ictrlchunks;
4405 prim = asoc->peer.primary_path;
4406 memcpy(&info->sctpi_p_address, &prim->ipaddr, sizeof(prim->ipaddr));
4407 info->sctpi_p_state = prim->state;
4408 info->sctpi_p_cwnd = prim->cwnd;
4409 info->sctpi_p_srtt = prim->srtt;
4410 info->sctpi_p_rto = jiffies_to_msecs(prim->rto);
4411 info->sctpi_p_hbinterval = prim->hbinterval;
4412 info->sctpi_p_pathmaxrxt = prim->pathmaxrxt;
4413 info->sctpi_p_sackdelay = jiffies_to_msecs(prim->sackdelay);
4414 info->sctpi_p_ssthresh = prim->ssthresh;
4415 info->sctpi_p_partial_bytes_acked = prim->partial_bytes_acked;
4416 info->sctpi_p_flight_size = prim->flight_size;
4417 info->sctpi_p_error = prim->error_count;
4421 EXPORT_SYMBOL_GPL(sctp_get_sctp_info);
4423 /* use callback to avoid exporting the core structure */
4424 int sctp_transport_walk_start(struct rhashtable_iter *iter)
4428 err = rhashtable_walk_init(&sctp_transport_hashtable, iter,
4433 err = rhashtable_walk_start(iter);
4434 if (err && err != -EAGAIN) {
4435 rhashtable_walk_stop(iter);
4436 rhashtable_walk_exit(iter);
4443 void sctp_transport_walk_stop(struct rhashtable_iter *iter)
4445 rhashtable_walk_stop(iter);
4446 rhashtable_walk_exit(iter);
4449 struct sctp_transport *sctp_transport_get_next(struct net *net,
4450 struct rhashtable_iter *iter)
4452 struct sctp_transport *t;
4454 t = rhashtable_walk_next(iter);
4455 for (; t; t = rhashtable_walk_next(iter)) {
4457 if (PTR_ERR(t) == -EAGAIN)
4462 if (net_eq(sock_net(t->asoc->base.sk), net) &&
4463 t->asoc->peer.primary_path == t)
4470 struct sctp_transport *sctp_transport_get_idx(struct net *net,
4471 struct rhashtable_iter *iter,
4474 void *obj = SEQ_START_TOKEN;
4476 while (pos && (obj = sctp_transport_get_next(net, iter)) &&
4483 int sctp_for_each_endpoint(int (*cb)(struct sctp_endpoint *, void *),
4487 struct sctp_ep_common *epb;
4488 struct sctp_hashbucket *head;
4490 for (head = sctp_ep_hashtable; hash < sctp_ep_hashsize;
4492 read_lock_bh(&head->lock);
4493 sctp_for_each_hentry(epb, &head->chain) {
4494 err = cb(sctp_ep(epb), p);
4498 read_unlock_bh(&head->lock);
4503 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint);
4505 int sctp_transport_lookup_process(int (*cb)(struct sctp_transport *, void *),
4507 const union sctp_addr *laddr,
4508 const union sctp_addr *paddr, void *p)
4510 struct sctp_transport *transport;
4514 transport = sctp_addrs_lookup_transport(net, laddr, paddr);
4515 if (!transport || !sctp_transport_hold(transport)) {
4520 err = cb(transport, p);
4521 sctp_transport_put(transport);
4526 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process);
4528 int sctp_for_each_transport(int (*cb)(struct sctp_transport *, void *),
4529 struct net *net, int pos, void *p) {
4530 struct rhashtable_iter hti;
4534 err = sctp_transport_walk_start(&hti);
4538 obj = sctp_transport_get_idx(net, &hti, pos + 1);
4539 for (; !IS_ERR_OR_NULL(obj); obj = sctp_transport_get_next(net, &hti)) {
4540 struct sctp_transport *transport = obj;
4542 if (!sctp_transport_hold(transport))
4544 err = cb(transport, p);
4545 sctp_transport_put(transport);
4549 sctp_transport_walk_stop(&hti);
4553 EXPORT_SYMBOL_GPL(sctp_for_each_transport);
4555 /* 7.2.1 Association Status (SCTP_STATUS)
4557 * Applications can retrieve current status information about an
4558 * association, including association state, peer receiver window size,
4559 * number of unacked data chunks, and number of data chunks pending
4560 * receipt. This information is read-only.
4562 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4563 char __user *optval,
4566 struct sctp_status status;
4567 struct sctp_association *asoc = NULL;
4568 struct sctp_transport *transport;
4569 sctp_assoc_t associd;
4572 if (len < sizeof(status)) {
4577 len = sizeof(status);
4578 if (copy_from_user(&status, optval, len)) {
4583 associd = status.sstat_assoc_id;
4584 asoc = sctp_id2assoc(sk, associd);
4590 transport = asoc->peer.primary_path;
4592 status.sstat_assoc_id = sctp_assoc2id(asoc);
4593 status.sstat_state = sctp_assoc_to_state(asoc);
4594 status.sstat_rwnd = asoc->peer.rwnd;
4595 status.sstat_unackdata = asoc->unack_data;
4597 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4598 status.sstat_instrms = asoc->c.sinit_max_instreams;
4599 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4600 status.sstat_fragmentation_point = asoc->frag_point;
4601 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4602 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4603 transport->af_specific->sockaddr_len);
4604 /* Map ipv4 address into v4-mapped-on-v6 address. */
4605 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
4606 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4607 status.sstat_primary.spinfo_state = transport->state;
4608 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4609 status.sstat_primary.spinfo_srtt = transport->srtt;
4610 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4611 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4613 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4614 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4616 if (put_user(len, optlen)) {
4621 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4622 __func__, len, status.sstat_state, status.sstat_rwnd,
4623 status.sstat_assoc_id);
4625 if (copy_to_user(optval, &status, len)) {
4635 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4637 * Applications can retrieve information about a specific peer address
4638 * of an association, including its reachability state, congestion
4639 * window, and retransmission timer values. This information is
4642 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4643 char __user *optval,
4646 struct sctp_paddrinfo pinfo;
4647 struct sctp_transport *transport;
4650 if (len < sizeof(pinfo)) {
4655 len = sizeof(pinfo);
4656 if (copy_from_user(&pinfo, optval, len)) {
4661 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4662 pinfo.spinfo_assoc_id);
4666 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4667 pinfo.spinfo_state = transport->state;
4668 pinfo.spinfo_cwnd = transport->cwnd;
4669 pinfo.spinfo_srtt = transport->srtt;
4670 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4671 pinfo.spinfo_mtu = transport->pathmtu;
4673 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4674 pinfo.spinfo_state = SCTP_ACTIVE;
4676 if (put_user(len, optlen)) {
4681 if (copy_to_user(optval, &pinfo, len)) {
4690 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4692 * This option is a on/off flag. If enabled no SCTP message
4693 * fragmentation will be performed. Instead if a message being sent
4694 * exceeds the current PMTU size, the message will NOT be sent and
4695 * instead a error will be indicated to the user.
4697 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4698 char __user *optval, int __user *optlen)
4702 if (len < sizeof(int))
4706 val = (sctp_sk(sk)->disable_fragments == 1);
4707 if (put_user(len, optlen))
4709 if (copy_to_user(optval, &val, len))
4714 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4716 * This socket option is used to specify various notifications and
4717 * ancillary data the user wishes to receive.
4719 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4724 if (len > sizeof(struct sctp_event_subscribe))
4725 len = sizeof(struct sctp_event_subscribe);
4726 if (put_user(len, optlen))
4728 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4733 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4735 * This socket option is applicable to the UDP-style socket only. When
4736 * set it will cause associations that are idle for more than the
4737 * specified number of seconds to automatically close. An association
4738 * being idle is defined an association that has NOT sent or received
4739 * user data. The special value of '0' indicates that no automatic
4740 * close of any associations should be performed. The option expects an
4741 * integer defining the number of seconds of idle time before an
4742 * association is closed.
4744 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4746 /* Applicable to UDP-style socket only */
4747 if (sctp_style(sk, TCP))
4749 if (len < sizeof(int))
4752 if (put_user(len, optlen))
4754 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4759 /* Helper routine to branch off an association to a new socket. */
4760 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4762 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4763 struct sctp_sock *sp = sctp_sk(sk);
4764 struct socket *sock;
4767 /* Do not peel off from one netns to another one. */
4768 if (!net_eq(current->nsproxy->net_ns, sock_net(sk)))
4774 /* If there is a thread waiting on more sndbuf space for
4775 * sending on this asoc, it cannot be peeled.
4777 if (waitqueue_active(&asoc->wait))
4780 /* An association cannot be branched off from an already peeled-off
4781 * socket, nor is this supported for tcp style sockets.
4783 if (!sctp_style(sk, UDP))
4786 /* Create a new socket. */
4787 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4791 sctp_copy_sock(sock->sk, sk, asoc);
4793 /* Make peeled-off sockets more like 1-1 accepted sockets.
4794 * Set the daddr and initialize id to something more random
4796 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
4798 /* Populate the fields of the newsk from the oldsk and migrate the
4799 * asoc to the newsk.
4801 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4807 EXPORT_SYMBOL(sctp_do_peeloff);
4809 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4811 sctp_peeloff_arg_t peeloff;
4812 struct socket *newsock;
4813 struct file *newfile;
4816 if (len < sizeof(sctp_peeloff_arg_t))
4818 len = sizeof(sctp_peeloff_arg_t);
4819 if (copy_from_user(&peeloff, optval, len))
4822 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4826 /* Map the socket to an unused fd that can be returned to the user. */
4827 retval = get_unused_fd_flags(0);
4829 sock_release(newsock);
4833 newfile = sock_alloc_file(newsock, 0, NULL);
4834 if (IS_ERR(newfile)) {
4835 put_unused_fd(retval);
4836 sock_release(newsock);
4837 return PTR_ERR(newfile);
4840 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4843 /* Return the fd mapped to the new socket. */
4844 if (put_user(len, optlen)) {
4846 put_unused_fd(retval);
4849 peeloff.sd = retval;
4850 if (copy_to_user(optval, &peeloff, len)) {
4852 put_unused_fd(retval);
4855 fd_install(retval, newfile);
4860 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4862 * Applications can enable or disable heartbeats for any peer address of
4863 * an association, modify an address's heartbeat interval, force a
4864 * heartbeat to be sent immediately, and adjust the address's maximum
4865 * number of retransmissions sent before an address is considered
4866 * unreachable. The following structure is used to access and modify an
4867 * address's parameters:
4869 * struct sctp_paddrparams {
4870 * sctp_assoc_t spp_assoc_id;
4871 * struct sockaddr_storage spp_address;
4872 * uint32_t spp_hbinterval;
4873 * uint16_t spp_pathmaxrxt;
4874 * uint32_t spp_pathmtu;
4875 * uint32_t spp_sackdelay;
4876 * uint32_t spp_flags;
4879 * spp_assoc_id - (one-to-many style socket) This is filled in the
4880 * application, and identifies the association for
4882 * spp_address - This specifies which address is of interest.
4883 * spp_hbinterval - This contains the value of the heartbeat interval,
4884 * in milliseconds. If a value of zero
4885 * is present in this field then no changes are to
4886 * be made to this parameter.
4887 * spp_pathmaxrxt - This contains the maximum number of
4888 * retransmissions before this address shall be
4889 * considered unreachable. If a value of zero
4890 * is present in this field then no changes are to
4891 * be made to this parameter.
4892 * spp_pathmtu - When Path MTU discovery is disabled the value
4893 * specified here will be the "fixed" path mtu.
4894 * Note that if the spp_address field is empty
4895 * then all associations on this address will
4896 * have this fixed path mtu set upon them.
4898 * spp_sackdelay - When delayed sack is enabled, this value specifies
4899 * the number of milliseconds that sacks will be delayed
4900 * for. This value will apply to all addresses of an
4901 * association if the spp_address field is empty. Note
4902 * also, that if delayed sack is enabled and this
4903 * value is set to 0, no change is made to the last
4904 * recorded delayed sack timer value.
4906 * spp_flags - These flags are used to control various features
4907 * on an association. The flag field may contain
4908 * zero or more of the following options.
4910 * SPP_HB_ENABLE - Enable heartbeats on the
4911 * specified address. Note that if the address
4912 * field is empty all addresses for the association
4913 * have heartbeats enabled upon them.
4915 * SPP_HB_DISABLE - Disable heartbeats on the
4916 * speicifed address. Note that if the address
4917 * field is empty all addresses for the association
4918 * will have their heartbeats disabled. Note also
4919 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4920 * mutually exclusive, only one of these two should
4921 * be specified. Enabling both fields will have
4922 * undetermined results.
4924 * SPP_HB_DEMAND - Request a user initiated heartbeat
4925 * to be made immediately.
4927 * SPP_PMTUD_ENABLE - This field will enable PMTU
4928 * discovery upon the specified address. Note that
4929 * if the address feild is empty then all addresses
4930 * on the association are effected.
4932 * SPP_PMTUD_DISABLE - This field will disable PMTU
4933 * discovery upon the specified address. Note that
4934 * if the address feild is empty then all addresses
4935 * on the association are effected. Not also that
4936 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4937 * exclusive. Enabling both will have undetermined
4940 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4941 * on delayed sack. The time specified in spp_sackdelay
4942 * is used to specify the sack delay for this address. Note
4943 * that if spp_address is empty then all addresses will
4944 * enable delayed sack and take on the sack delay
4945 * value specified in spp_sackdelay.
4946 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4947 * off delayed sack. If the spp_address field is blank then
4948 * delayed sack is disabled for the entire association. Note
4949 * also that this field is mutually exclusive to
4950 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4953 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4954 char __user *optval, int __user *optlen)
4956 struct sctp_paddrparams params;
4957 struct sctp_transport *trans = NULL;
4958 struct sctp_association *asoc = NULL;
4959 struct sctp_sock *sp = sctp_sk(sk);
4961 if (len < sizeof(struct sctp_paddrparams))
4963 len = sizeof(struct sctp_paddrparams);
4964 if (copy_from_user(¶ms, optval, len))
4967 /* If an address other than INADDR_ANY is specified, and
4968 * no transport is found, then the request is invalid.
4970 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
4971 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4972 params.spp_assoc_id);
4974 pr_debug("%s: failed no transport\n", __func__);
4979 /* Get association, if assoc_id != 0 and the socket is a one
4980 * to many style socket, and an association was not found, then
4981 * the id was invalid.
4983 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4984 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4985 pr_debug("%s: failed no association\n", __func__);
4990 /* Fetch transport values. */
4991 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4992 params.spp_pathmtu = trans->pathmtu;
4993 params.spp_pathmaxrxt = trans->pathmaxrxt;
4994 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4996 /*draft-11 doesn't say what to return in spp_flags*/
4997 params.spp_flags = trans->param_flags;
4999 /* Fetch association values. */
5000 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
5001 params.spp_pathmtu = asoc->pathmtu;
5002 params.spp_pathmaxrxt = asoc->pathmaxrxt;
5003 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
5005 /*draft-11 doesn't say what to return in spp_flags*/
5006 params.spp_flags = asoc->param_flags;
5008 /* Fetch socket values. */
5009 params.spp_hbinterval = sp->hbinterval;
5010 params.spp_pathmtu = sp->pathmtu;
5011 params.spp_sackdelay = sp->sackdelay;
5012 params.spp_pathmaxrxt = sp->pathmaxrxt;
5014 /*draft-11 doesn't say what to return in spp_flags*/
5015 params.spp_flags = sp->param_flags;
5018 if (copy_to_user(optval, ¶ms, len))
5021 if (put_user(len, optlen))
5028 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
5030 * This option will effect the way delayed acks are performed. This
5031 * option allows you to get or set the delayed ack time, in
5032 * milliseconds. It also allows changing the delayed ack frequency.
5033 * Changing the frequency to 1 disables the delayed sack algorithm. If
5034 * the assoc_id is 0, then this sets or gets the endpoints default
5035 * values. If the assoc_id field is non-zero, then the set or get
5036 * effects the specified association for the one to many model (the
5037 * assoc_id field is ignored by the one to one model). Note that if
5038 * sack_delay or sack_freq are 0 when setting this option, then the
5039 * current values will remain unchanged.
5041 * struct sctp_sack_info {
5042 * sctp_assoc_t sack_assoc_id;
5043 * uint32_t sack_delay;
5044 * uint32_t sack_freq;
5047 * sack_assoc_id - This parameter, indicates which association the user
5048 * is performing an action upon. Note that if this field's value is
5049 * zero then the endpoints default value is changed (effecting future
5050 * associations only).
5052 * sack_delay - This parameter contains the number of milliseconds that
5053 * the user is requesting the delayed ACK timer be set to. Note that
5054 * this value is defined in the standard to be between 200 and 500
5057 * sack_freq - This parameter contains the number of packets that must
5058 * be received before a sack is sent without waiting for the delay
5059 * timer to expire. The default value for this is 2, setting this
5060 * value to 1 will disable the delayed sack algorithm.
5062 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
5063 char __user *optval,
5066 struct sctp_sack_info params;
5067 struct sctp_association *asoc = NULL;
5068 struct sctp_sock *sp = sctp_sk(sk);
5070 if (len >= sizeof(struct sctp_sack_info)) {
5071 len = sizeof(struct sctp_sack_info);
5073 if (copy_from_user(¶ms, optval, len))
5075 } else if (len == sizeof(struct sctp_assoc_value)) {
5076 pr_warn_ratelimited(DEPRECATED
5078 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
5079 "Use struct sctp_sack_info instead\n",
5080 current->comm, task_pid_nr(current));
5081 if (copy_from_user(¶ms, optval, len))
5086 /* Get association, if sack_assoc_id != 0 and the socket is a one
5087 * to many style socket, and an association was not found, then
5088 * the id was invalid.
5090 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
5091 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
5095 /* Fetch association values. */
5096 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
5097 params.sack_delay = jiffies_to_msecs(
5099 params.sack_freq = asoc->sackfreq;
5102 params.sack_delay = 0;
5103 params.sack_freq = 1;
5106 /* Fetch socket values. */
5107 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
5108 params.sack_delay = sp->sackdelay;
5109 params.sack_freq = sp->sackfreq;
5111 params.sack_delay = 0;
5112 params.sack_freq = 1;
5116 if (copy_to_user(optval, ¶ms, len))
5119 if (put_user(len, optlen))
5125 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5127 * Applications can specify protocol parameters for the default association
5128 * initialization. The option name argument to setsockopt() and getsockopt()
5131 * Setting initialization parameters is effective only on an unconnected
5132 * socket (for UDP-style sockets only future associations are effected
5133 * by the change). With TCP-style sockets, this option is inherited by
5134 * sockets derived from a listener socket.
5136 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
5138 if (len < sizeof(struct sctp_initmsg))
5140 len = sizeof(struct sctp_initmsg);
5141 if (put_user(len, optlen))
5143 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
5149 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
5150 char __user *optval, int __user *optlen)
5152 struct sctp_association *asoc;
5154 struct sctp_getaddrs getaddrs;
5155 struct sctp_transport *from;
5157 union sctp_addr temp;
5158 struct sctp_sock *sp = sctp_sk(sk);
5163 if (len < sizeof(struct sctp_getaddrs))
5166 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5169 /* For UDP-style sockets, id specifies the association to query. */
5170 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5174 to = optval + offsetof(struct sctp_getaddrs, addrs);
5175 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5177 list_for_each_entry(from, &asoc->peer.transport_addr_list,
5179 memcpy(&temp, &from->ipaddr, sizeof(temp));
5180 addrlen = sctp_get_pf_specific(sk->sk_family)
5181 ->addr_to_user(sp, &temp);
5182 if (space_left < addrlen)
5184 if (copy_to_user(to, &temp, addrlen))
5188 space_left -= addrlen;
5191 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
5193 bytes_copied = ((char __user *)to) - optval;
5194 if (put_user(bytes_copied, optlen))
5200 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
5201 size_t space_left, int *bytes_copied)
5203 struct sctp_sockaddr_entry *addr;
5204 union sctp_addr temp;
5207 struct net *net = sock_net(sk);
5210 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
5214 if ((PF_INET == sk->sk_family) &&
5215 (AF_INET6 == addr->a.sa.sa_family))
5217 if ((PF_INET6 == sk->sk_family) &&
5218 inet_v6_ipv6only(sk) &&
5219 (AF_INET == addr->a.sa.sa_family))
5221 memcpy(&temp, &addr->a, sizeof(temp));
5222 if (!temp.v4.sin_port)
5223 temp.v4.sin_port = htons(port);
5225 addrlen = sctp_get_pf_specific(sk->sk_family)
5226 ->addr_to_user(sctp_sk(sk), &temp);
5228 if (space_left < addrlen) {
5232 memcpy(to, &temp, addrlen);
5236 space_left -= addrlen;
5237 *bytes_copied += addrlen;
5245 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
5246 char __user *optval, int __user *optlen)
5248 struct sctp_bind_addr *bp;
5249 struct sctp_association *asoc;
5251 struct sctp_getaddrs getaddrs;
5252 struct sctp_sockaddr_entry *addr;
5254 union sctp_addr temp;
5255 struct sctp_sock *sp = sctp_sk(sk);
5259 int bytes_copied = 0;
5263 if (len < sizeof(struct sctp_getaddrs))
5266 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5270 * For UDP-style sockets, id specifies the association to query.
5271 * If the id field is set to the value '0' then the locally bound
5272 * addresses are returned without regard to any particular
5275 if (0 == getaddrs.assoc_id) {
5276 bp = &sctp_sk(sk)->ep->base.bind_addr;
5278 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5281 bp = &asoc->base.bind_addr;
5284 to = optval + offsetof(struct sctp_getaddrs, addrs);
5285 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5287 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN);
5291 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5292 * addresses from the global local address list.
5294 if (sctp_list_single_entry(&bp->address_list)) {
5295 addr = list_entry(bp->address_list.next,
5296 struct sctp_sockaddr_entry, list);
5297 if (sctp_is_any(sk, &addr->a)) {
5298 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
5299 space_left, &bytes_copied);
5309 /* Protection on the bound address list is not needed since
5310 * in the socket option context we hold a socket lock and
5311 * thus the bound address list can't change.
5313 list_for_each_entry(addr, &bp->address_list, list) {
5314 memcpy(&temp, &addr->a, sizeof(temp));
5315 addrlen = sctp_get_pf_specific(sk->sk_family)
5316 ->addr_to_user(sp, &temp);
5317 if (space_left < addrlen) {
5318 err = -ENOMEM; /*fixme: right error?*/
5321 memcpy(buf, &temp, addrlen);
5323 bytes_copied += addrlen;
5325 space_left -= addrlen;
5329 if (copy_to_user(to, addrs, bytes_copied)) {
5333 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
5337 if (put_user(bytes_copied, optlen))
5344 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
5346 * Requests that the local SCTP stack use the enclosed peer address as
5347 * the association primary. The enclosed address must be one of the
5348 * association peer's addresses.
5350 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
5351 char __user *optval, int __user *optlen)
5353 struct sctp_prim prim;
5354 struct sctp_association *asoc;
5355 struct sctp_sock *sp = sctp_sk(sk);
5357 if (len < sizeof(struct sctp_prim))
5360 len = sizeof(struct sctp_prim);
5362 if (copy_from_user(&prim, optval, len))
5365 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
5369 if (!asoc->peer.primary_path)
5372 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
5373 asoc->peer.primary_path->af_specific->sockaddr_len);
5375 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
5376 (union sctp_addr *)&prim.ssp_addr);
5378 if (put_user(len, optlen))
5380 if (copy_to_user(optval, &prim, len))
5387 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5389 * Requests that the local endpoint set the specified Adaptation Layer
5390 * Indication parameter for all future INIT and INIT-ACK exchanges.
5392 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
5393 char __user *optval, int __user *optlen)
5395 struct sctp_setadaptation adaptation;
5397 if (len < sizeof(struct sctp_setadaptation))
5400 len = sizeof(struct sctp_setadaptation);
5402 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
5404 if (put_user(len, optlen))
5406 if (copy_to_user(optval, &adaptation, len))
5414 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5416 * Applications that wish to use the sendto() system call may wish to
5417 * specify a default set of parameters that would normally be supplied
5418 * through the inclusion of ancillary data. This socket option allows
5419 * such an application to set the default sctp_sndrcvinfo structure.
5422 * The application that wishes to use this socket option simply passes
5423 * in to this call the sctp_sndrcvinfo structure defined in Section
5424 * 5.2.2) The input parameters accepted by this call include
5425 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5426 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5427 * to this call if the caller is using the UDP model.
5429 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5431 static int sctp_getsockopt_default_send_param(struct sock *sk,
5432 int len, char __user *optval,
5435 struct sctp_sock *sp = sctp_sk(sk);
5436 struct sctp_association *asoc;
5437 struct sctp_sndrcvinfo info;
5439 if (len < sizeof(info))
5444 if (copy_from_user(&info, optval, len))
5447 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
5448 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
5451 info.sinfo_stream = asoc->default_stream;
5452 info.sinfo_flags = asoc->default_flags;
5453 info.sinfo_ppid = asoc->default_ppid;
5454 info.sinfo_context = asoc->default_context;
5455 info.sinfo_timetolive = asoc->default_timetolive;
5457 info.sinfo_stream = sp->default_stream;
5458 info.sinfo_flags = sp->default_flags;
5459 info.sinfo_ppid = sp->default_ppid;
5460 info.sinfo_context = sp->default_context;
5461 info.sinfo_timetolive = sp->default_timetolive;
5464 if (put_user(len, optlen))
5466 if (copy_to_user(optval, &info, len))
5472 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5473 * (SCTP_DEFAULT_SNDINFO)
5475 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
5476 char __user *optval,
5479 struct sctp_sock *sp = sctp_sk(sk);
5480 struct sctp_association *asoc;
5481 struct sctp_sndinfo info;
5483 if (len < sizeof(info))
5488 if (copy_from_user(&info, optval, len))
5491 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
5492 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
5495 info.snd_sid = asoc->default_stream;
5496 info.snd_flags = asoc->default_flags;
5497 info.snd_ppid = asoc->default_ppid;
5498 info.snd_context = asoc->default_context;
5500 info.snd_sid = sp->default_stream;
5501 info.snd_flags = sp->default_flags;
5502 info.snd_ppid = sp->default_ppid;
5503 info.snd_context = sp->default_context;
5506 if (put_user(len, optlen))
5508 if (copy_to_user(optval, &info, len))
5516 * 7.1.5 SCTP_NODELAY
5518 * Turn on/off any Nagle-like algorithm. This means that packets are
5519 * generally sent as soon as possible and no unnecessary delays are
5520 * introduced, at the cost of more packets in the network. Expects an
5521 * integer boolean flag.
5524 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5525 char __user *optval, int __user *optlen)
5529 if (len < sizeof(int))
5533 val = (sctp_sk(sk)->nodelay == 1);
5534 if (put_user(len, optlen))
5536 if (copy_to_user(optval, &val, len))
5543 * 7.1.1 SCTP_RTOINFO
5545 * The protocol parameters used to initialize and bound retransmission
5546 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5547 * and modify these parameters.
5548 * All parameters are time values, in milliseconds. A value of 0, when
5549 * modifying the parameters, indicates that the current value should not
5553 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5554 char __user *optval,
5555 int __user *optlen) {
5556 struct sctp_rtoinfo rtoinfo;
5557 struct sctp_association *asoc;
5559 if (len < sizeof (struct sctp_rtoinfo))
5562 len = sizeof(struct sctp_rtoinfo);
5564 if (copy_from_user(&rtoinfo, optval, len))
5567 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5569 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5572 /* Values corresponding to the specific association. */
5574 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5575 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5576 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5578 /* Values corresponding to the endpoint. */
5579 struct sctp_sock *sp = sctp_sk(sk);
5581 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5582 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5583 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5586 if (put_user(len, optlen))
5589 if (copy_to_user(optval, &rtoinfo, len))
5597 * 7.1.2 SCTP_ASSOCINFO
5599 * This option is used to tune the maximum retransmission attempts
5600 * of the association.
5601 * Returns an error if the new association retransmission value is
5602 * greater than the sum of the retransmission value of the peer.
5603 * See [SCTP] for more information.
5606 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5607 char __user *optval,
5611 struct sctp_assocparams assocparams;
5612 struct sctp_association *asoc;
5613 struct list_head *pos;
5616 if (len < sizeof (struct sctp_assocparams))
5619 len = sizeof(struct sctp_assocparams);
5621 if (copy_from_user(&assocparams, optval, len))
5624 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5626 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5629 /* Values correspoinding to the specific association */
5631 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5632 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5633 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5634 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5636 list_for_each(pos, &asoc->peer.transport_addr_list) {
5640 assocparams.sasoc_number_peer_destinations = cnt;
5642 /* Values corresponding to the endpoint */
5643 struct sctp_sock *sp = sctp_sk(sk);
5645 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5646 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5647 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5648 assocparams.sasoc_cookie_life =
5649 sp->assocparams.sasoc_cookie_life;
5650 assocparams.sasoc_number_peer_destinations =
5652 sasoc_number_peer_destinations;
5655 if (put_user(len, optlen))
5658 if (copy_to_user(optval, &assocparams, len))
5665 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5667 * This socket option is a boolean flag which turns on or off mapped V4
5668 * addresses. If this option is turned on and the socket is type
5669 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5670 * If this option is turned off, then no mapping will be done of V4
5671 * addresses and a user will receive both PF_INET6 and PF_INET type
5672 * addresses on the socket.
5674 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5675 char __user *optval, int __user *optlen)
5678 struct sctp_sock *sp = sctp_sk(sk);
5680 if (len < sizeof(int))
5685 if (put_user(len, optlen))
5687 if (copy_to_user(optval, &val, len))
5694 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5695 * (chapter and verse is quoted at sctp_setsockopt_context())
5697 static int sctp_getsockopt_context(struct sock *sk, int len,
5698 char __user *optval, int __user *optlen)
5700 struct sctp_assoc_value params;
5701 struct sctp_sock *sp;
5702 struct sctp_association *asoc;
5704 if (len < sizeof(struct sctp_assoc_value))
5707 len = sizeof(struct sctp_assoc_value);
5709 if (copy_from_user(¶ms, optval, len))
5714 if (params.assoc_id != 0) {
5715 asoc = sctp_id2assoc(sk, params.assoc_id);
5718 params.assoc_value = asoc->default_rcv_context;
5720 params.assoc_value = sp->default_rcv_context;
5723 if (put_user(len, optlen))
5725 if (copy_to_user(optval, ¶ms, len))
5732 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5733 * This option will get or set the maximum size to put in any outgoing
5734 * SCTP DATA chunk. If a message is larger than this size it will be
5735 * fragmented by SCTP into the specified size. Note that the underlying
5736 * SCTP implementation may fragment into smaller sized chunks when the
5737 * PMTU of the underlying association is smaller than the value set by
5738 * the user. The default value for this option is '0' which indicates
5739 * the user is NOT limiting fragmentation and only the PMTU will effect
5740 * SCTP's choice of DATA chunk size. Note also that values set larger
5741 * than the maximum size of an IP datagram will effectively let SCTP
5742 * control fragmentation (i.e. the same as setting this option to 0).
5744 * The following structure is used to access and modify this parameter:
5746 * struct sctp_assoc_value {
5747 * sctp_assoc_t assoc_id;
5748 * uint32_t assoc_value;
5751 * assoc_id: This parameter is ignored for one-to-one style sockets.
5752 * For one-to-many style sockets this parameter indicates which
5753 * association the user is performing an action upon. Note that if
5754 * this field's value is zero then the endpoints default value is
5755 * changed (effecting future associations only).
5756 * assoc_value: This parameter specifies the maximum size in bytes.
5758 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5759 char __user *optval, int __user *optlen)
5761 struct sctp_assoc_value params;
5762 struct sctp_association *asoc;
5764 if (len == sizeof(int)) {
5765 pr_warn_ratelimited(DEPRECATED
5767 "Use of int in maxseg socket option.\n"
5768 "Use struct sctp_assoc_value instead\n",
5769 current->comm, task_pid_nr(current));
5770 params.assoc_id = 0;
5771 } else if (len >= sizeof(struct sctp_assoc_value)) {
5772 len = sizeof(struct sctp_assoc_value);
5773 if (copy_from_user(¶ms, optval, sizeof(params)))
5778 asoc = sctp_id2assoc(sk, params.assoc_id);
5779 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5783 params.assoc_value = asoc->frag_point;
5785 params.assoc_value = sctp_sk(sk)->user_frag;
5787 if (put_user(len, optlen))
5789 if (len == sizeof(int)) {
5790 if (copy_to_user(optval, ¶ms.assoc_value, len))
5793 if (copy_to_user(optval, ¶ms, len))
5801 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5802 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5804 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5805 char __user *optval, int __user *optlen)
5809 if (len < sizeof(int))
5814 val = sctp_sk(sk)->frag_interleave;
5815 if (put_user(len, optlen))
5817 if (copy_to_user(optval, &val, len))
5824 * 7.1.25. Set or Get the sctp partial delivery point
5825 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5827 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5828 char __user *optval,
5833 if (len < sizeof(u32))
5838 val = sctp_sk(sk)->pd_point;
5839 if (put_user(len, optlen))
5841 if (copy_to_user(optval, &val, len))
5848 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5849 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5851 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5852 char __user *optval,
5855 struct sctp_assoc_value params;
5856 struct sctp_sock *sp;
5857 struct sctp_association *asoc;
5859 if (len == sizeof(int)) {
5860 pr_warn_ratelimited(DEPRECATED
5862 "Use of int in max_burst socket option.\n"
5863 "Use struct sctp_assoc_value instead\n",
5864 current->comm, task_pid_nr(current));
5865 params.assoc_id = 0;
5866 } else if (len >= sizeof(struct sctp_assoc_value)) {
5867 len = sizeof(struct sctp_assoc_value);
5868 if (copy_from_user(¶ms, optval, len))
5875 if (params.assoc_id != 0) {
5876 asoc = sctp_id2assoc(sk, params.assoc_id);
5879 params.assoc_value = asoc->max_burst;
5881 params.assoc_value = sp->max_burst;
5883 if (len == sizeof(int)) {
5884 if (copy_to_user(optval, ¶ms.assoc_value, len))
5887 if (copy_to_user(optval, ¶ms, len))
5895 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5896 char __user *optval, int __user *optlen)
5898 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5899 struct sctp_hmacalgo __user *p = (void __user *)optval;
5900 struct sctp_hmac_algo_param *hmacs;
5905 if (!ep->auth_enable)
5908 hmacs = ep->auth_hmacs_list;
5909 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5911 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5914 len = sizeof(struct sctp_hmacalgo) + data_len;
5915 num_idents = data_len / sizeof(u16);
5917 if (put_user(len, optlen))
5919 if (put_user(num_idents, &p->shmac_num_idents))
5921 for (i = 0; i < num_idents; i++) {
5922 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
5924 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
5930 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5931 char __user *optval, int __user *optlen)
5933 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5934 struct sctp_authkeyid val;
5935 struct sctp_association *asoc;
5937 if (!ep->auth_enable)
5940 if (len < sizeof(struct sctp_authkeyid))
5942 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5945 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5946 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5950 val.scact_keynumber = asoc->active_key_id;
5952 val.scact_keynumber = ep->active_key_id;
5954 len = sizeof(struct sctp_authkeyid);
5955 if (put_user(len, optlen))
5957 if (copy_to_user(optval, &val, len))
5963 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5964 char __user *optval, int __user *optlen)
5966 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5967 struct sctp_authchunks __user *p = (void __user *)optval;
5968 struct sctp_authchunks val;
5969 struct sctp_association *asoc;
5970 struct sctp_chunks_param *ch;
5974 if (!ep->auth_enable)
5977 if (len < sizeof(struct sctp_authchunks))
5980 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5983 to = p->gauth_chunks;
5984 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5988 ch = asoc->peer.peer_chunks;
5992 /* See if the user provided enough room for all the data */
5993 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5994 if (len < num_chunks)
5997 if (copy_to_user(to, ch->chunks, num_chunks))
6000 len = sizeof(struct sctp_authchunks) + num_chunks;
6001 if (put_user(len, optlen))
6003 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6008 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
6009 char __user *optval, int __user *optlen)
6011 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6012 struct sctp_authchunks __user *p = (void __user *)optval;
6013 struct sctp_authchunks val;
6014 struct sctp_association *asoc;
6015 struct sctp_chunks_param *ch;
6019 if (!ep->auth_enable)
6022 if (len < sizeof(struct sctp_authchunks))
6025 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
6028 to = p->gauth_chunks;
6029 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6030 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
6034 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
6036 ch = ep->auth_chunk_list;
6041 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
6042 if (len < sizeof(struct sctp_authchunks) + num_chunks)
6045 if (copy_to_user(to, ch->chunks, num_chunks))
6048 len = sizeof(struct sctp_authchunks) + num_chunks;
6049 if (put_user(len, optlen))
6051 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6058 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
6059 * This option gets the current number of associations that are attached
6060 * to a one-to-many style socket. The option value is an uint32_t.
6062 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
6063 char __user *optval, int __user *optlen)
6065 struct sctp_sock *sp = sctp_sk(sk);
6066 struct sctp_association *asoc;
6069 if (sctp_style(sk, TCP))
6072 if (len < sizeof(u32))
6077 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6081 if (put_user(len, optlen))
6083 if (copy_to_user(optval, &val, len))
6090 * 8.1.23 SCTP_AUTO_ASCONF
6091 * See the corresponding setsockopt entry as description
6093 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
6094 char __user *optval, int __user *optlen)
6098 if (len < sizeof(int))
6102 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
6104 if (put_user(len, optlen))
6106 if (copy_to_user(optval, &val, len))
6112 * 8.2.6. Get the Current Identifiers of Associations
6113 * (SCTP_GET_ASSOC_ID_LIST)
6115 * This option gets the current list of SCTP association identifiers of
6116 * the SCTP associations handled by a one-to-many style socket.
6118 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
6119 char __user *optval, int __user *optlen)
6121 struct sctp_sock *sp = sctp_sk(sk);
6122 struct sctp_association *asoc;
6123 struct sctp_assoc_ids *ids;
6126 if (sctp_style(sk, TCP))
6129 if (len < sizeof(struct sctp_assoc_ids))
6132 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6136 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
6139 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
6141 ids = kmalloc(len, GFP_USER | __GFP_NOWARN);
6145 ids->gaids_number_of_ids = num;
6147 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6148 ids->gaids_assoc_id[num++] = asoc->assoc_id;
6151 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
6161 * SCTP_PEER_ADDR_THLDS
6163 * This option allows us to fetch the partially failed threshold for one or all
6164 * transports in an association. See Section 6.1 of:
6165 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6167 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
6168 char __user *optval,
6172 struct sctp_paddrthlds val;
6173 struct sctp_transport *trans;
6174 struct sctp_association *asoc;
6176 if (len < sizeof(struct sctp_paddrthlds))
6178 len = sizeof(struct sctp_paddrthlds);
6179 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
6182 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
6183 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
6187 val.spt_pathpfthld = asoc->pf_retrans;
6188 val.spt_pathmaxrxt = asoc->pathmaxrxt;
6190 trans = sctp_addr_id2transport(sk, &val.spt_address,
6195 val.spt_pathmaxrxt = trans->pathmaxrxt;
6196 val.spt_pathpfthld = trans->pf_retrans;
6199 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
6206 * SCTP_GET_ASSOC_STATS
6208 * This option retrieves local per endpoint statistics. It is modeled
6209 * after OpenSolaris' implementation
6211 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
6212 char __user *optval,
6215 struct sctp_assoc_stats sas;
6216 struct sctp_association *asoc = NULL;
6218 /* User must provide at least the assoc id */
6219 if (len < sizeof(sctp_assoc_t))
6222 /* Allow the struct to grow and fill in as much as possible */
6223 len = min_t(size_t, len, sizeof(sas));
6225 if (copy_from_user(&sas, optval, len))
6228 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
6232 sas.sas_rtxchunks = asoc->stats.rtxchunks;
6233 sas.sas_gapcnt = asoc->stats.gapcnt;
6234 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
6235 sas.sas_osacks = asoc->stats.osacks;
6236 sas.sas_isacks = asoc->stats.isacks;
6237 sas.sas_octrlchunks = asoc->stats.octrlchunks;
6238 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
6239 sas.sas_oodchunks = asoc->stats.oodchunks;
6240 sas.sas_iodchunks = asoc->stats.iodchunks;
6241 sas.sas_ouodchunks = asoc->stats.ouodchunks;
6242 sas.sas_iuodchunks = asoc->stats.iuodchunks;
6243 sas.sas_idupchunks = asoc->stats.idupchunks;
6244 sas.sas_opackets = asoc->stats.opackets;
6245 sas.sas_ipackets = asoc->stats.ipackets;
6247 /* New high max rto observed, will return 0 if not a single
6248 * RTO update took place. obs_rto_ipaddr will be bogus
6251 sas.sas_maxrto = asoc->stats.max_obs_rto;
6252 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
6253 sizeof(struct sockaddr_storage));
6255 /* Mark beginning of a new observation period */
6256 asoc->stats.max_obs_rto = asoc->rto_min;
6258 if (put_user(len, optlen))
6261 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
6263 if (copy_to_user(optval, &sas, len))
6269 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
6270 char __user *optval,
6275 if (len < sizeof(int))
6279 if (sctp_sk(sk)->recvrcvinfo)
6281 if (put_user(len, optlen))
6283 if (copy_to_user(optval, &val, len))
6289 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
6290 char __user *optval,
6295 if (len < sizeof(int))
6299 if (sctp_sk(sk)->recvnxtinfo)
6301 if (put_user(len, optlen))
6303 if (copy_to_user(optval, &val, len))
6309 static int sctp_getsockopt_pr_supported(struct sock *sk, int len,
6310 char __user *optval,
6313 struct sctp_assoc_value params;
6314 struct sctp_association *asoc;
6315 int retval = -EFAULT;
6317 if (len < sizeof(params)) {
6322 len = sizeof(params);
6323 if (copy_from_user(¶ms, optval, len))
6326 asoc = sctp_id2assoc(sk, params.assoc_id);
6328 params.assoc_value = asoc->prsctp_enable;
6329 } else if (!params.assoc_id) {
6330 struct sctp_sock *sp = sctp_sk(sk);
6332 params.assoc_value = sp->ep->prsctp_enable;
6338 if (put_user(len, optlen))
6341 if (copy_to_user(optval, ¶ms, len))
6350 static int sctp_getsockopt_default_prinfo(struct sock *sk, int len,
6351 char __user *optval,
6354 struct sctp_default_prinfo info;
6355 struct sctp_association *asoc;
6356 int retval = -EFAULT;
6358 if (len < sizeof(info)) {
6364 if (copy_from_user(&info, optval, len))
6367 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
6369 info.pr_policy = SCTP_PR_POLICY(asoc->default_flags);
6370 info.pr_value = asoc->default_timetolive;
6371 } else if (!info.pr_assoc_id) {
6372 struct sctp_sock *sp = sctp_sk(sk);
6374 info.pr_policy = SCTP_PR_POLICY(sp->default_flags);
6375 info.pr_value = sp->default_timetolive;
6381 if (put_user(len, optlen))
6384 if (copy_to_user(optval, &info, len))
6393 static int sctp_getsockopt_pr_assocstatus(struct sock *sk, int len,
6394 char __user *optval,
6397 struct sctp_prstatus params;
6398 struct sctp_association *asoc;
6400 int retval = -EINVAL;
6402 if (len < sizeof(params))
6405 len = sizeof(params);
6406 if (copy_from_user(¶ms, optval, len)) {
6411 policy = params.sprstat_policy;
6412 if (policy & ~SCTP_PR_SCTP_MASK)
6415 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
6419 if (policy == SCTP_PR_SCTP_NONE) {
6420 params.sprstat_abandoned_unsent = 0;
6421 params.sprstat_abandoned_sent = 0;
6422 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
6423 params.sprstat_abandoned_unsent +=
6424 asoc->abandoned_unsent[policy];
6425 params.sprstat_abandoned_sent +=
6426 asoc->abandoned_sent[policy];
6429 params.sprstat_abandoned_unsent =
6430 asoc->abandoned_unsent[__SCTP_PR_INDEX(policy)];
6431 params.sprstat_abandoned_sent =
6432 asoc->abandoned_sent[__SCTP_PR_INDEX(policy)];
6435 if (put_user(len, optlen)) {
6440 if (copy_to_user(optval, ¶ms, len)) {
6451 static int sctp_getsockopt(struct sock *sk, int level, int optname,
6452 char __user *optval, int __user *optlen)
6457 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
6459 /* I can hardly begin to describe how wrong this is. This is
6460 * so broken as to be worse than useless. The API draft
6461 * REALLY is NOT helpful here... I am not convinced that the
6462 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
6463 * are at all well-founded.
6465 if (level != SOL_SCTP) {
6466 struct sctp_af *af = sctp_sk(sk)->pf->af;
6468 retval = af->getsockopt(sk, level, optname, optval, optlen);
6472 if (get_user(len, optlen))
6482 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
6484 case SCTP_DISABLE_FRAGMENTS:
6485 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
6489 retval = sctp_getsockopt_events(sk, len, optval, optlen);
6491 case SCTP_AUTOCLOSE:
6492 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
6494 case SCTP_SOCKOPT_PEELOFF:
6495 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
6497 case SCTP_PEER_ADDR_PARAMS:
6498 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
6501 case SCTP_DELAYED_SACK:
6502 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
6506 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
6508 case SCTP_GET_PEER_ADDRS:
6509 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
6512 case SCTP_GET_LOCAL_ADDRS:
6513 retval = sctp_getsockopt_local_addrs(sk, len, optval,
6516 case SCTP_SOCKOPT_CONNECTX3:
6517 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
6519 case SCTP_DEFAULT_SEND_PARAM:
6520 retval = sctp_getsockopt_default_send_param(sk, len,
6523 case SCTP_DEFAULT_SNDINFO:
6524 retval = sctp_getsockopt_default_sndinfo(sk, len,
6527 case SCTP_PRIMARY_ADDR:
6528 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
6531 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
6534 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
6536 case SCTP_ASSOCINFO:
6537 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
6539 case SCTP_I_WANT_MAPPED_V4_ADDR:
6540 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
6543 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
6545 case SCTP_GET_PEER_ADDR_INFO:
6546 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
6549 case SCTP_ADAPTATION_LAYER:
6550 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
6554 retval = sctp_getsockopt_context(sk, len, optval, optlen);
6556 case SCTP_FRAGMENT_INTERLEAVE:
6557 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
6560 case SCTP_PARTIAL_DELIVERY_POINT:
6561 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
6564 case SCTP_MAX_BURST:
6565 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
6568 case SCTP_AUTH_CHUNK:
6569 case SCTP_AUTH_DELETE_KEY:
6570 retval = -EOPNOTSUPP;
6572 case SCTP_HMAC_IDENT:
6573 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
6575 case SCTP_AUTH_ACTIVE_KEY:
6576 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
6578 case SCTP_PEER_AUTH_CHUNKS:
6579 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
6582 case SCTP_LOCAL_AUTH_CHUNKS:
6583 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
6586 case SCTP_GET_ASSOC_NUMBER:
6587 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
6589 case SCTP_GET_ASSOC_ID_LIST:
6590 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
6592 case SCTP_AUTO_ASCONF:
6593 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
6595 case SCTP_PEER_ADDR_THLDS:
6596 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
6598 case SCTP_GET_ASSOC_STATS:
6599 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
6601 case SCTP_RECVRCVINFO:
6602 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
6604 case SCTP_RECVNXTINFO:
6605 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
6607 case SCTP_PR_SUPPORTED:
6608 retval = sctp_getsockopt_pr_supported(sk, len, optval, optlen);
6610 case SCTP_DEFAULT_PRINFO:
6611 retval = sctp_getsockopt_default_prinfo(sk, len, optval,
6614 case SCTP_PR_ASSOC_STATUS:
6615 retval = sctp_getsockopt_pr_assocstatus(sk, len, optval,
6619 retval = -ENOPROTOOPT;
6627 static int sctp_hash(struct sock *sk)
6633 static void sctp_unhash(struct sock *sk)
6638 /* Check if port is acceptable. Possibly find first available port.
6640 * The port hash table (contained in the 'global' SCTP protocol storage
6641 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
6642 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
6643 * list (the list number is the port number hashed out, so as you
6644 * would expect from a hash function, all the ports in a given list have
6645 * such a number that hashes out to the same list number; you were
6646 * expecting that, right?); so each list has a set of ports, with a
6647 * link to the socket (struct sock) that uses it, the port number and
6648 * a fastreuse flag (FIXME: NPI ipg).
6650 static struct sctp_bind_bucket *sctp_bucket_create(
6651 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
6653 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
6655 struct sctp_bind_hashbucket *head; /* hash list */
6656 struct sctp_bind_bucket *pp;
6657 unsigned short snum;
6660 snum = ntohs(addr->v4.sin_port);
6662 pr_debug("%s: begins, snum:%d\n", __func__, snum);
6667 /* Search for an available port. */
6668 int low, high, remaining, index;
6670 struct net *net = sock_net(sk);
6672 inet_get_local_port_range(net, &low, &high);
6673 remaining = (high - low) + 1;
6674 rover = prandom_u32() % remaining + low;
6678 if ((rover < low) || (rover > high))
6680 if (inet_is_local_reserved_port(net, rover))
6682 index = sctp_phashfn(sock_net(sk), rover);
6683 head = &sctp_port_hashtable[index];
6684 spin_lock(&head->lock);
6685 sctp_for_each_hentry(pp, &head->chain)
6686 if ((pp->port == rover) &&
6687 net_eq(sock_net(sk), pp->net))
6691 spin_unlock(&head->lock);
6692 } while (--remaining > 0);
6694 /* Exhausted local port range during search? */
6699 /* OK, here is the one we will use. HEAD (the port
6700 * hash table list entry) is non-NULL and we hold it's
6705 /* We are given an specific port number; we verify
6706 * that it is not being used. If it is used, we will
6707 * exahust the search in the hash list corresponding
6708 * to the port number (snum) - we detect that with the
6709 * port iterator, pp being NULL.
6711 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
6712 spin_lock(&head->lock);
6713 sctp_for_each_hentry(pp, &head->chain) {
6714 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
6721 if (!hlist_empty(&pp->owner)) {
6722 /* We had a port hash table hit - there is an
6723 * available port (pp != NULL) and it is being
6724 * used by other socket (pp->owner not empty); that other
6725 * socket is going to be sk2.
6727 int reuse = sk->sk_reuse;
6730 pr_debug("%s: found a possible match\n", __func__);
6732 if (pp->fastreuse && sk->sk_reuse &&
6733 sk->sk_state != SCTP_SS_LISTENING)
6736 /* Run through the list of sockets bound to the port
6737 * (pp->port) [via the pointers bind_next and
6738 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6739 * we get the endpoint they describe and run through
6740 * the endpoint's list of IP (v4 or v6) addresses,
6741 * comparing each of the addresses with the address of
6742 * the socket sk. If we find a match, then that means
6743 * that this port/socket (sk) combination are already
6746 sk_for_each_bound(sk2, &pp->owner) {
6747 struct sctp_endpoint *ep2;
6748 ep2 = sctp_sk(sk2)->ep;
6751 (reuse && sk2->sk_reuse &&
6752 sk2->sk_state != SCTP_SS_LISTENING))
6755 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
6756 sctp_sk(sk2), sctp_sk(sk))) {
6762 pr_debug("%s: found a match\n", __func__);
6765 /* If there was a hash table miss, create a new port. */
6767 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6770 /* In either case (hit or miss), make sure fastreuse is 1 only
6771 * if sk->sk_reuse is too (that is, if the caller requested
6772 * SO_REUSEADDR on this socket -sk-).
6774 if (hlist_empty(&pp->owner)) {
6775 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6779 } else if (pp->fastreuse &&
6780 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6783 /* We are set, so fill up all the data in the hash table
6784 * entry, tie the socket list information with the rest of the
6785 * sockets FIXME: Blurry, NPI (ipg).
6788 if (!sctp_sk(sk)->bind_hash) {
6789 inet_sk(sk)->inet_num = snum;
6790 sk_add_bind_node(sk, &pp->owner);
6791 sctp_sk(sk)->bind_hash = pp;
6796 spin_unlock(&head->lock);
6803 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6804 * port is requested.
6806 static int sctp_get_port(struct sock *sk, unsigned short snum)
6808 union sctp_addr addr;
6809 struct sctp_af *af = sctp_sk(sk)->pf->af;
6811 /* Set up a dummy address struct from the sk. */
6812 af->from_sk(&addr, sk);
6813 addr.v4.sin_port = htons(snum);
6815 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6816 return !!sctp_get_port_local(sk, &addr);
6820 * Move a socket to LISTENING state.
6822 static int sctp_listen_start(struct sock *sk, int backlog)
6824 struct sctp_sock *sp = sctp_sk(sk);
6825 struct sctp_endpoint *ep = sp->ep;
6826 struct crypto_shash *tfm = NULL;
6829 /* Allocate HMAC for generating cookie. */
6830 if (!sp->hmac && sp->sctp_hmac_alg) {
6831 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6832 tfm = crypto_alloc_shash(alg, 0, 0);
6834 net_info_ratelimited("failed to load transform for %s: %ld\n",
6835 sp->sctp_hmac_alg, PTR_ERR(tfm));
6838 sctp_sk(sk)->hmac = tfm;
6842 * If a bind() or sctp_bindx() is not called prior to a listen()
6843 * call that allows new associations to be accepted, the system
6844 * picks an ephemeral port and will choose an address set equivalent
6845 * to binding with a wildcard address.
6847 * This is not currently spelled out in the SCTP sockets
6848 * extensions draft, but follows the practice as seen in TCP
6852 sk->sk_state = SCTP_SS_LISTENING;
6853 if (!ep->base.bind_addr.port) {
6854 if (sctp_autobind(sk))
6857 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6858 sk->sk_state = SCTP_SS_CLOSED;
6863 sk->sk_max_ack_backlog = backlog;
6864 sctp_hash_endpoint(ep);
6869 * 4.1.3 / 5.1.3 listen()
6871 * By default, new associations are not accepted for UDP style sockets.
6872 * An application uses listen() to mark a socket as being able to
6873 * accept new associations.
6875 * On TCP style sockets, applications use listen() to ready the SCTP
6876 * endpoint for accepting inbound associations.
6878 * On both types of endpoints a backlog of '0' disables listening.
6880 * Move a socket to LISTENING state.
6882 int sctp_inet_listen(struct socket *sock, int backlog)
6884 struct sock *sk = sock->sk;
6885 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6888 if (unlikely(backlog < 0))
6893 /* Peeled-off sockets are not allowed to listen(). */
6894 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6897 if (sock->state != SS_UNCONNECTED)
6900 if (!sctp_sstate(sk, LISTENING) && !sctp_sstate(sk, CLOSED))
6903 /* If backlog is zero, disable listening. */
6905 if (sctp_sstate(sk, CLOSED))
6909 sctp_unhash_endpoint(ep);
6910 sk->sk_state = SCTP_SS_CLOSED;
6912 sctp_sk(sk)->bind_hash->fastreuse = 1;
6916 /* If we are already listening, just update the backlog */
6917 if (sctp_sstate(sk, LISTENING))
6918 sk->sk_max_ack_backlog = backlog;
6920 err = sctp_listen_start(sk, backlog);
6932 * This function is done by modeling the current datagram_poll() and the
6933 * tcp_poll(). Note that, based on these implementations, we don't
6934 * lock the socket in this function, even though it seems that,
6935 * ideally, locking or some other mechanisms can be used to ensure
6936 * the integrity of the counters (sndbuf and wmem_alloc) used
6937 * in this place. We assume that we don't need locks either until proven
6940 * Another thing to note is that we include the Async I/O support
6941 * here, again, by modeling the current TCP/UDP code. We don't have
6942 * a good way to test with it yet.
6944 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6946 struct sock *sk = sock->sk;
6947 struct sctp_sock *sp = sctp_sk(sk);
6950 poll_wait(file, sk_sleep(sk), wait);
6952 sock_rps_record_flow(sk);
6954 /* A TCP-style listening socket becomes readable when the accept queue
6957 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6958 return (!list_empty(&sp->ep->asocs)) ?
6959 (POLLIN | POLLRDNORM) : 0;
6963 /* Is there any exceptional events? */
6964 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6966 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6967 if (sk->sk_shutdown & RCV_SHUTDOWN)
6968 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6969 if (sk->sk_shutdown == SHUTDOWN_MASK)
6972 /* Is it readable? Reconsider this code with TCP-style support. */
6973 if (!skb_queue_empty(&sk->sk_receive_queue))
6974 mask |= POLLIN | POLLRDNORM;
6976 /* The association is either gone or not ready. */
6977 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6980 /* Is it writable? */
6981 if (sctp_writeable(sk)) {
6982 mask |= POLLOUT | POLLWRNORM;
6984 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
6986 * Since the socket is not locked, the buffer
6987 * might be made available after the writeable check and
6988 * before the bit is set. This could cause a lost I/O
6989 * signal. tcp_poll() has a race breaker for this race
6990 * condition. Based on their implementation, we put
6991 * in the following code to cover it as well.
6993 if (sctp_writeable(sk))
6994 mask |= POLLOUT | POLLWRNORM;
6999 /********************************************************************
7000 * 2nd Level Abstractions
7001 ********************************************************************/
7003 static struct sctp_bind_bucket *sctp_bucket_create(
7004 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
7006 struct sctp_bind_bucket *pp;
7008 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
7010 SCTP_DBG_OBJCNT_INC(bind_bucket);
7013 INIT_HLIST_HEAD(&pp->owner);
7015 hlist_add_head(&pp->node, &head->chain);
7020 /* Caller must hold hashbucket lock for this tb with local BH disabled */
7021 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
7023 if (pp && hlist_empty(&pp->owner)) {
7024 __hlist_del(&pp->node);
7025 kmem_cache_free(sctp_bucket_cachep, pp);
7026 SCTP_DBG_OBJCNT_DEC(bind_bucket);
7030 /* Release this socket's reference to a local port. */
7031 static inline void __sctp_put_port(struct sock *sk)
7033 struct sctp_bind_hashbucket *head =
7034 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
7035 inet_sk(sk)->inet_num)];
7036 struct sctp_bind_bucket *pp;
7038 spin_lock(&head->lock);
7039 pp = sctp_sk(sk)->bind_hash;
7040 __sk_del_bind_node(sk);
7041 sctp_sk(sk)->bind_hash = NULL;
7042 inet_sk(sk)->inet_num = 0;
7043 sctp_bucket_destroy(pp);
7044 spin_unlock(&head->lock);
7047 void sctp_put_port(struct sock *sk)
7050 __sctp_put_port(sk);
7055 * The system picks an ephemeral port and choose an address set equivalent
7056 * to binding with a wildcard address.
7057 * One of those addresses will be the primary address for the association.
7058 * This automatically enables the multihoming capability of SCTP.
7060 static int sctp_autobind(struct sock *sk)
7062 union sctp_addr autoaddr;
7066 /* Initialize a local sockaddr structure to INADDR_ANY. */
7067 af = sctp_sk(sk)->pf->af;
7069 port = htons(inet_sk(sk)->inet_num);
7070 af->inaddr_any(&autoaddr, port);
7072 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
7075 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
7078 * 4.2 The cmsghdr Structure *
7080 * When ancillary data is sent or received, any number of ancillary data
7081 * objects can be specified by the msg_control and msg_controllen members of
7082 * the msghdr structure, because each object is preceded by
7083 * a cmsghdr structure defining the object's length (the cmsg_len member).
7084 * Historically Berkeley-derived implementations have passed only one object
7085 * at a time, but this API allows multiple objects to be
7086 * passed in a single call to sendmsg() or recvmsg(). The following example
7087 * shows two ancillary data objects in a control buffer.
7089 * |<--------------------------- msg_controllen -------------------------->|
7092 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
7094 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
7097 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
7099 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
7102 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7103 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
7105 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
7107 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7114 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
7116 struct cmsghdr *cmsg;
7117 struct msghdr *my_msg = (struct msghdr *)msg;
7119 for_each_cmsghdr(cmsg, my_msg) {
7120 if (!CMSG_OK(my_msg, cmsg))
7123 /* Should we parse this header or ignore? */
7124 if (cmsg->cmsg_level != IPPROTO_SCTP)
7127 /* Strictly check lengths following example in SCM code. */
7128 switch (cmsg->cmsg_type) {
7130 /* SCTP Socket API Extension
7131 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
7133 * This cmsghdr structure provides information for
7134 * initializing new SCTP associations with sendmsg().
7135 * The SCTP_INITMSG socket option uses this same data
7136 * structure. This structure is not used for
7139 * cmsg_level cmsg_type cmsg_data[]
7140 * ------------ ------------ ----------------------
7141 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
7143 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
7146 cmsgs->init = CMSG_DATA(cmsg);
7150 /* SCTP Socket API Extension
7151 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
7153 * This cmsghdr structure specifies SCTP options for
7154 * sendmsg() and describes SCTP header information
7155 * about a received message through recvmsg().
7157 * cmsg_level cmsg_type cmsg_data[]
7158 * ------------ ------------ ----------------------
7159 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
7161 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
7164 cmsgs->srinfo = CMSG_DATA(cmsg);
7166 if (cmsgs->srinfo->sinfo_flags &
7167 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7168 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7169 SCTP_ABORT | SCTP_EOF))
7174 /* SCTP Socket API Extension
7175 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
7177 * This cmsghdr structure specifies SCTP options for
7178 * sendmsg(). This structure and SCTP_RCVINFO replaces
7179 * SCTP_SNDRCV which has been deprecated.
7181 * cmsg_level cmsg_type cmsg_data[]
7182 * ------------ ------------ ---------------------
7183 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
7185 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
7188 cmsgs->sinfo = CMSG_DATA(cmsg);
7190 if (cmsgs->sinfo->snd_flags &
7191 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7192 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7193 SCTP_ABORT | SCTP_EOF))
7205 * Wait for a packet..
7206 * Note: This function is the same function as in core/datagram.c
7207 * with a few modifications to make lksctp work.
7209 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
7214 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7216 /* Socket errors? */
7217 error = sock_error(sk);
7221 if (!skb_queue_empty(&sk->sk_receive_queue))
7224 /* Socket shut down? */
7225 if (sk->sk_shutdown & RCV_SHUTDOWN)
7228 /* Sequenced packets can come disconnected. If so we report the
7233 /* Is there a good reason to think that we may receive some data? */
7234 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
7237 /* Handle signals. */
7238 if (signal_pending(current))
7241 /* Let another process have a go. Since we are going to sleep
7242 * anyway. Note: This may cause odd behaviors if the message
7243 * does not fit in the user's buffer, but this seems to be the
7244 * only way to honor MSG_DONTWAIT realistically.
7247 *timeo_p = schedule_timeout(*timeo_p);
7251 finish_wait(sk_sleep(sk), &wait);
7255 error = sock_intr_errno(*timeo_p);
7258 finish_wait(sk_sleep(sk), &wait);
7263 /* Receive a datagram.
7264 * Note: This is pretty much the same routine as in core/datagram.c
7265 * with a few changes to make lksctp work.
7267 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
7268 int noblock, int *err)
7271 struct sk_buff *skb;
7274 timeo = sock_rcvtimeo(sk, noblock);
7276 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
7277 MAX_SCHEDULE_TIMEOUT);
7280 /* Again only user level code calls this function,
7281 * so nothing interrupt level
7282 * will suddenly eat the receive_queue.
7284 * Look at current nfs client by the way...
7285 * However, this function was correct in any case. 8)
7287 if (flags & MSG_PEEK) {
7288 skb = skb_peek(&sk->sk_receive_queue);
7290 atomic_inc(&skb->users);
7292 skb = __skb_dequeue(&sk->sk_receive_queue);
7298 /* Caller is allowed not to check sk->sk_err before calling. */
7299 error = sock_error(sk);
7303 if (sk->sk_shutdown & RCV_SHUTDOWN)
7306 if (sk_can_busy_loop(sk) &&
7307 sk_busy_loop(sk, noblock))
7310 /* User doesn't want to wait. */
7314 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
7323 /* If sndbuf has changed, wake up per association sndbuf waiters. */
7324 static void __sctp_write_space(struct sctp_association *asoc)
7326 struct sock *sk = asoc->base.sk;
7328 if (sctp_wspace(asoc) <= 0)
7331 if (waitqueue_active(&asoc->wait))
7332 wake_up_interruptible(&asoc->wait);
7334 if (sctp_writeable(sk)) {
7335 struct socket_wq *wq;
7338 wq = rcu_dereference(sk->sk_wq);
7340 if (waitqueue_active(&wq->wait))
7341 wake_up_interruptible(&wq->wait);
7343 /* Note that we try to include the Async I/O support
7344 * here by modeling from the current TCP/UDP code.
7345 * We have not tested with it yet.
7347 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
7348 sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT);
7354 static void sctp_wake_up_waiters(struct sock *sk,
7355 struct sctp_association *asoc)
7357 struct sctp_association *tmp = asoc;
7359 /* We do accounting for the sndbuf space per association,
7360 * so we only need to wake our own association.
7362 if (asoc->ep->sndbuf_policy)
7363 return __sctp_write_space(asoc);
7365 /* If association goes down and is just flushing its
7366 * outq, then just normally notify others.
7368 if (asoc->base.dead)
7369 return sctp_write_space(sk);
7371 /* Accounting for the sndbuf space is per socket, so we
7372 * need to wake up others, try to be fair and in case of
7373 * other associations, let them have a go first instead
7374 * of just doing a sctp_write_space() call.
7376 * Note that we reach sctp_wake_up_waiters() only when
7377 * associations free up queued chunks, thus we are under
7378 * lock and the list of associations on a socket is
7379 * guaranteed not to change.
7381 for (tmp = list_next_entry(tmp, asocs); 1;
7382 tmp = list_next_entry(tmp, asocs)) {
7383 /* Manually skip the head element. */
7384 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
7386 /* Wake up association. */
7387 __sctp_write_space(tmp);
7388 /* We've reached the end. */
7394 /* Do accounting for the sndbuf space.
7395 * Decrement the used sndbuf space of the corresponding association by the
7396 * data size which was just transmitted(freed).
7398 static void sctp_wfree(struct sk_buff *skb)
7400 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
7401 struct sctp_association *asoc = chunk->asoc;
7402 struct sock *sk = asoc->base.sk;
7404 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
7405 sizeof(struct sk_buff) +
7406 sizeof(struct sctp_chunk);
7408 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
7411 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
7413 sk->sk_wmem_queued -= skb->truesize;
7414 sk_mem_uncharge(sk, skb->truesize);
7417 sctp_wake_up_waiters(sk, asoc);
7419 sctp_association_put(asoc);
7422 /* Do accounting for the receive space on the socket.
7423 * Accounting for the association is done in ulpevent.c
7424 * We set this as a destructor for the cloned data skbs so that
7425 * accounting is done at the correct time.
7427 void sctp_sock_rfree(struct sk_buff *skb)
7429 struct sock *sk = skb->sk;
7430 struct sctp_ulpevent *event = sctp_skb2event(skb);
7432 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
7435 * Mimic the behavior of sock_rfree
7437 sk_mem_uncharge(sk, event->rmem_len);
7441 /* Helper function to wait for space in the sndbuf. */
7442 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
7445 struct sock *sk = asoc->base.sk;
7447 long current_timeo = *timeo_p;
7450 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
7453 /* Increment the association's refcnt. */
7454 sctp_association_hold(asoc);
7456 /* Wait on the association specific sndbuf space. */
7458 prepare_to_wait_exclusive(&asoc->wait, &wait,
7459 TASK_INTERRUPTIBLE);
7462 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7465 if (signal_pending(current))
7466 goto do_interrupted;
7467 if (msg_len <= sctp_wspace(asoc))
7470 /* Let another process have a go. Since we are going
7474 current_timeo = schedule_timeout(current_timeo);
7477 *timeo_p = current_timeo;
7481 finish_wait(&asoc->wait, &wait);
7483 /* Release the association's refcnt. */
7484 sctp_association_put(asoc);
7493 err = sock_intr_errno(*timeo_p);
7501 void sctp_data_ready(struct sock *sk)
7503 struct socket_wq *wq;
7506 wq = rcu_dereference(sk->sk_wq);
7507 if (skwq_has_sleeper(wq))
7508 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
7509 POLLRDNORM | POLLRDBAND);
7510 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
7514 /* If socket sndbuf has changed, wake up all per association waiters. */
7515 void sctp_write_space(struct sock *sk)
7517 struct sctp_association *asoc;
7519 /* Wake up the tasks in each wait queue. */
7520 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
7521 __sctp_write_space(asoc);
7525 /* Is there any sndbuf space available on the socket?
7527 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
7528 * associations on the same socket. For a UDP-style socket with
7529 * multiple associations, it is possible for it to be "unwriteable"
7530 * prematurely. I assume that this is acceptable because
7531 * a premature "unwriteable" is better than an accidental "writeable" which
7532 * would cause an unwanted block under certain circumstances. For the 1-1
7533 * UDP-style sockets or TCP-style sockets, this code should work.
7536 static int sctp_writeable(struct sock *sk)
7540 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
7546 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
7547 * returns immediately with EINPROGRESS.
7549 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
7551 struct sock *sk = asoc->base.sk;
7553 long current_timeo = *timeo_p;
7556 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
7558 /* Increment the association's refcnt. */
7559 sctp_association_hold(asoc);
7562 prepare_to_wait_exclusive(&asoc->wait, &wait,
7563 TASK_INTERRUPTIBLE);
7566 if (sk->sk_shutdown & RCV_SHUTDOWN)
7568 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7571 if (signal_pending(current))
7572 goto do_interrupted;
7574 if (sctp_state(asoc, ESTABLISHED))
7577 /* Let another process have a go. Since we are going
7581 current_timeo = schedule_timeout(current_timeo);
7584 *timeo_p = current_timeo;
7588 finish_wait(&asoc->wait, &wait);
7590 /* Release the association's refcnt. */
7591 sctp_association_put(asoc);
7596 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
7599 err = -ECONNREFUSED;
7603 err = sock_intr_errno(*timeo_p);
7611 static int sctp_wait_for_accept(struct sock *sk, long timeo)
7613 struct sctp_endpoint *ep;
7617 ep = sctp_sk(sk)->ep;
7621 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
7622 TASK_INTERRUPTIBLE);
7624 if (list_empty(&ep->asocs)) {
7626 timeo = schedule_timeout(timeo);
7631 if (!sctp_sstate(sk, LISTENING))
7635 if (!list_empty(&ep->asocs))
7638 err = sock_intr_errno(timeo);
7639 if (signal_pending(current))
7647 finish_wait(sk_sleep(sk), &wait);
7652 static void sctp_wait_for_close(struct sock *sk, long timeout)
7657 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7658 if (list_empty(&sctp_sk(sk)->ep->asocs))
7661 timeout = schedule_timeout(timeout);
7663 } while (!signal_pending(current) && timeout);
7665 finish_wait(sk_sleep(sk), &wait);
7668 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
7670 struct sk_buff *frag;
7675 /* Don't forget the fragments. */
7676 skb_walk_frags(skb, frag)
7677 sctp_skb_set_owner_r_frag(frag, sk);
7680 sctp_skb_set_owner_r(skb, sk);
7683 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
7684 struct sctp_association *asoc)
7686 struct inet_sock *inet = inet_sk(sk);
7687 struct inet_sock *newinet;
7689 newsk->sk_type = sk->sk_type;
7690 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
7691 newsk->sk_flags = sk->sk_flags;
7692 newsk->sk_tsflags = sk->sk_tsflags;
7693 newsk->sk_no_check_tx = sk->sk_no_check_tx;
7694 newsk->sk_no_check_rx = sk->sk_no_check_rx;
7695 newsk->sk_reuse = sk->sk_reuse;
7697 newsk->sk_shutdown = sk->sk_shutdown;
7698 newsk->sk_destruct = sctp_destruct_sock;
7699 newsk->sk_family = sk->sk_family;
7700 newsk->sk_protocol = IPPROTO_SCTP;
7701 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
7702 newsk->sk_sndbuf = sk->sk_sndbuf;
7703 newsk->sk_rcvbuf = sk->sk_rcvbuf;
7704 newsk->sk_lingertime = sk->sk_lingertime;
7705 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
7706 newsk->sk_sndtimeo = sk->sk_sndtimeo;
7707 newsk->sk_rxhash = sk->sk_rxhash;
7709 newinet = inet_sk(newsk);
7711 /* Initialize sk's sport, dport, rcv_saddr and daddr for
7712 * getsockname() and getpeername()
7714 newinet->inet_sport = inet->inet_sport;
7715 newinet->inet_saddr = inet->inet_saddr;
7716 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
7717 newinet->inet_dport = htons(asoc->peer.port);
7718 newinet->pmtudisc = inet->pmtudisc;
7719 newinet->inet_id = asoc->next_tsn ^ jiffies;
7721 newinet->uc_ttl = inet->uc_ttl;
7722 newinet->mc_loop = 1;
7723 newinet->mc_ttl = 1;
7724 newinet->mc_index = 0;
7725 newinet->mc_list = NULL;
7727 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
7728 net_enable_timestamp();
7730 security_sk_clone(sk, newsk);
7733 static inline void sctp_copy_descendant(struct sock *sk_to,
7734 const struct sock *sk_from)
7736 int ancestor_size = sizeof(struct inet_sock) +
7737 sizeof(struct sctp_sock) -
7738 offsetof(struct sctp_sock, auto_asconf_list);
7740 if (sk_from->sk_family == PF_INET6)
7741 ancestor_size += sizeof(struct ipv6_pinfo);
7743 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
7746 /* Populate the fields of the newsk from the oldsk and migrate the assoc
7747 * and its messages to the newsk.
7749 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
7750 struct sctp_association *assoc,
7751 sctp_socket_type_t type)
7753 struct sctp_sock *oldsp = sctp_sk(oldsk);
7754 struct sctp_sock *newsp = sctp_sk(newsk);
7755 struct sctp_bind_bucket *pp; /* hash list port iterator */
7756 struct sctp_endpoint *newep = newsp->ep;
7757 struct sk_buff *skb, *tmp;
7758 struct sctp_ulpevent *event;
7759 struct sctp_bind_hashbucket *head;
7761 /* Migrate socket buffer sizes and all the socket level options to the
7764 newsk->sk_sndbuf = oldsk->sk_sndbuf;
7765 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
7766 /* Brute force copy old sctp opt. */
7767 sctp_copy_descendant(newsk, oldsk);
7769 /* Restore the ep value that was overwritten with the above structure
7775 /* Hook this new socket in to the bind_hash list. */
7776 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
7777 inet_sk(oldsk)->inet_num)];
7778 spin_lock_bh(&head->lock);
7779 pp = sctp_sk(oldsk)->bind_hash;
7780 sk_add_bind_node(newsk, &pp->owner);
7781 sctp_sk(newsk)->bind_hash = pp;
7782 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
7783 spin_unlock_bh(&head->lock);
7785 /* Copy the bind_addr list from the original endpoint to the new
7786 * endpoint so that we can handle restarts properly
7788 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
7789 &oldsp->ep->base.bind_addr, GFP_KERNEL);
7791 /* Move any messages in the old socket's receive queue that are for the
7792 * peeled off association to the new socket's receive queue.
7794 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
7795 event = sctp_skb2event(skb);
7796 if (event->asoc == assoc) {
7797 __skb_unlink(skb, &oldsk->sk_receive_queue);
7798 __skb_queue_tail(&newsk->sk_receive_queue, skb);
7799 sctp_skb_set_owner_r_frag(skb, newsk);
7803 /* Clean up any messages pending delivery due to partial
7804 * delivery. Three cases:
7805 * 1) No partial deliver; no work.
7806 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7807 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7809 skb_queue_head_init(&newsp->pd_lobby);
7810 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
7812 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
7813 struct sk_buff_head *queue;
7815 /* Decide which queue to move pd_lobby skbs to. */
7816 if (assoc->ulpq.pd_mode) {
7817 queue = &newsp->pd_lobby;
7819 queue = &newsk->sk_receive_queue;
7821 /* Walk through the pd_lobby, looking for skbs that
7822 * need moved to the new socket.
7824 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
7825 event = sctp_skb2event(skb);
7826 if (event->asoc == assoc) {
7827 __skb_unlink(skb, &oldsp->pd_lobby);
7828 __skb_queue_tail(queue, skb);
7829 sctp_skb_set_owner_r_frag(skb, newsk);
7833 /* Clear up any skbs waiting for the partial
7834 * delivery to finish.
7836 if (assoc->ulpq.pd_mode)
7837 sctp_clear_pd(oldsk, NULL);
7841 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7842 sctp_skb_set_owner_r_frag(skb, newsk);
7844 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7845 sctp_skb_set_owner_r_frag(skb, newsk);
7847 /* Set the type of socket to indicate that it is peeled off from the
7848 * original UDP-style socket or created with the accept() call on a
7849 * TCP-style socket..
7853 /* Mark the new socket "in-use" by the user so that any packets
7854 * that may arrive on the association after we've moved it are
7855 * queued to the backlog. This prevents a potential race between
7856 * backlog processing on the old socket and new-packet processing
7857 * on the new socket.
7859 * The caller has just allocated newsk so we can guarantee that other
7860 * paths won't try to lock it and then oldsk.
7862 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7863 sctp_for_each_tx_datachunk(assoc, sctp_clear_owner_w);
7864 sctp_assoc_migrate(assoc, newsk);
7865 sctp_for_each_tx_datachunk(assoc, sctp_set_owner_w);
7867 /* If the association on the newsk is already closed before accept()
7868 * is called, set RCV_SHUTDOWN flag.
7870 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP)) {
7871 newsk->sk_state = SCTP_SS_CLOSED;
7872 newsk->sk_shutdown |= RCV_SHUTDOWN;
7874 newsk->sk_state = SCTP_SS_ESTABLISHED;
7877 release_sock(newsk);
7881 /* This proto struct describes the ULP interface for SCTP. */
7882 struct proto sctp_prot = {
7884 .owner = THIS_MODULE,
7885 .close = sctp_close,
7886 .connect = sctp_connect,
7887 .disconnect = sctp_disconnect,
7888 .accept = sctp_accept,
7889 .ioctl = sctp_ioctl,
7890 .init = sctp_init_sock,
7891 .destroy = sctp_destroy_sock,
7892 .shutdown = sctp_shutdown,
7893 .setsockopt = sctp_setsockopt,
7894 .getsockopt = sctp_getsockopt,
7895 .sendmsg = sctp_sendmsg,
7896 .recvmsg = sctp_recvmsg,
7898 .backlog_rcv = sctp_backlog_rcv,
7900 .unhash = sctp_unhash,
7901 .get_port = sctp_get_port,
7902 .obj_size = sizeof(struct sctp_sock),
7903 .sysctl_mem = sysctl_sctp_mem,
7904 .sysctl_rmem = sysctl_sctp_rmem,
7905 .sysctl_wmem = sysctl_sctp_wmem,
7906 .memory_pressure = &sctp_memory_pressure,
7907 .enter_memory_pressure = sctp_enter_memory_pressure,
7908 .memory_allocated = &sctp_memory_allocated,
7909 .sockets_allocated = &sctp_sockets_allocated,
7912 #if IS_ENABLED(CONFIG_IPV6)
7914 #include <net/transp_v6.h>
7915 static void sctp_v6_destroy_sock(struct sock *sk)
7917 sctp_destroy_sock(sk);
7918 inet6_destroy_sock(sk);
7921 struct proto sctpv6_prot = {
7923 .owner = THIS_MODULE,
7924 .close = sctp_close,
7925 .connect = sctp_connect,
7926 .disconnect = sctp_disconnect,
7927 .accept = sctp_accept,
7928 .ioctl = sctp_ioctl,
7929 .init = sctp_init_sock,
7930 .destroy = sctp_v6_destroy_sock,
7931 .shutdown = sctp_shutdown,
7932 .setsockopt = sctp_setsockopt,
7933 .getsockopt = sctp_getsockopt,
7934 .sendmsg = sctp_sendmsg,
7935 .recvmsg = sctp_recvmsg,
7937 .backlog_rcv = sctp_backlog_rcv,
7939 .unhash = sctp_unhash,
7940 .get_port = sctp_get_port,
7941 .obj_size = sizeof(struct sctp6_sock),
7942 .sysctl_mem = sysctl_sctp_mem,
7943 .sysctl_rmem = sysctl_sctp_rmem,
7944 .sysctl_wmem = sysctl_sctp_wmem,
7945 .memory_pressure = &sctp_memory_pressure,
7946 .enter_memory_pressure = sctp_enter_memory_pressure,
7947 .memory_allocated = &sctp_memory_allocated,
7948 .sockets_allocated = &sctp_sockets_allocated,
7950 #endif /* IS_ENABLED(CONFIG_IPV6) */