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>
60 #include <linux/sched/signal.h>
62 #include <linux/capability.h>
63 #include <linux/fcntl.h>
64 #include <linux/poll.h>
65 #include <linux/init.h>
66 #include <linux/slab.h>
67 #include <linux/file.h>
68 #include <linux/compat.h>
72 #include <net/route.h>
74 #include <net/inet_common.h>
75 #include <net/busy_poll.h>
77 #include <linux/socket.h> /* for sa_family_t */
78 #include <linux/export.h>
80 #include <net/sctp/sctp.h>
81 #include <net/sctp/sm.h>
83 /* Forward declarations for internal helper functions. */
84 static int sctp_writeable(struct sock *sk);
85 static void sctp_wfree(struct sk_buff *skb);
86 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
88 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p);
89 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
90 static int sctp_wait_for_accept(struct sock *sk, long timeo);
91 static void sctp_wait_for_close(struct sock *sk, long timeo);
92 static void sctp_destruct_sock(struct sock *sk);
93 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
94 union sctp_addr *addr, int len);
95 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
96 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
97 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
98 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
99 static int sctp_send_asconf(struct sctp_association *asoc,
100 struct sctp_chunk *chunk);
101 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
102 static int sctp_autobind(struct sock *sk);
103 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
104 struct sctp_association *assoc,
105 enum sctp_socket_type type);
107 static unsigned long sctp_memory_pressure;
108 static atomic_long_t sctp_memory_allocated;
109 struct percpu_counter sctp_sockets_allocated;
111 static void sctp_enter_memory_pressure(struct sock *sk)
113 sctp_memory_pressure = 1;
117 /* Get the sndbuf space available at the time on the association. */
118 static inline int sctp_wspace(struct sctp_association *asoc)
122 if (asoc->ep->sndbuf_policy)
123 amt = asoc->sndbuf_used;
125 amt = sk_wmem_alloc_get(asoc->base.sk);
127 if (amt >= asoc->base.sk->sk_sndbuf) {
128 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
131 amt = sk_stream_wspace(asoc->base.sk);
136 amt = asoc->base.sk->sk_sndbuf - amt;
141 /* Increment the used sndbuf space count of the corresponding association by
142 * the size of the outgoing data chunk.
143 * Also, set the skb destructor for sndbuf accounting later.
145 * Since it is always 1-1 between chunk and skb, and also a new skb is always
146 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
147 * destructor in the data chunk skb for the purpose of the sndbuf space
150 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
152 struct sctp_association *asoc = chunk->asoc;
153 struct sock *sk = asoc->base.sk;
155 /* The sndbuf space is tracked per association. */
156 sctp_association_hold(asoc);
158 skb_set_owner_w(chunk->skb, sk);
160 chunk->skb->destructor = sctp_wfree;
161 /* Save the chunk pointer in skb for sctp_wfree to use later. */
162 skb_shinfo(chunk->skb)->destructor_arg = chunk;
164 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
165 sizeof(struct sk_buff) +
166 sizeof(struct sctp_chunk);
168 refcount_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
169 sk->sk_wmem_queued += chunk->skb->truesize;
170 sk_mem_charge(sk, chunk->skb->truesize);
173 /* Verify that this is a valid address. */
174 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
179 /* Verify basic sockaddr. */
180 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
184 /* Is this a valid SCTP address? */
185 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
188 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
194 /* Look up the association by its id. If this is not a UDP-style
195 * socket, the ID field is always ignored.
197 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
199 struct sctp_association *asoc = NULL;
201 /* If this is not a UDP-style socket, assoc id should be ignored. */
202 if (!sctp_style(sk, UDP)) {
203 /* Return NULL if the socket state is not ESTABLISHED. It
204 * could be a TCP-style listening socket or a socket which
205 * hasn't yet called connect() to establish an association.
207 if (!sctp_sstate(sk, ESTABLISHED) && !sctp_sstate(sk, CLOSING))
210 /* Get the first and the only association from the list. */
211 if (!list_empty(&sctp_sk(sk)->ep->asocs))
212 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
213 struct sctp_association, asocs);
217 /* Otherwise this is a UDP-style socket. */
218 if (!id || (id == (sctp_assoc_t)-1))
221 spin_lock_bh(&sctp_assocs_id_lock);
222 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
223 spin_unlock_bh(&sctp_assocs_id_lock);
225 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
231 /* Look up the transport from an address and an assoc id. If both address and
232 * id are specified, the associations matching the address and the id should be
235 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
236 struct sockaddr_storage *addr,
239 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
240 struct sctp_af *af = sctp_get_af_specific(addr->ss_family);
241 union sctp_addr *laddr = (union sctp_addr *)addr;
242 struct sctp_transport *transport;
244 if (!af || sctp_verify_addr(sk, laddr, af->sockaddr_len))
247 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
254 id_asoc = sctp_id2assoc(sk, id);
255 if (id_asoc && (id_asoc != addr_asoc))
258 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
259 (union sctp_addr *)addr);
264 /* API 3.1.2 bind() - UDP Style Syntax
265 * The syntax of bind() is,
267 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
269 * sd - the socket descriptor returned by socket().
270 * addr - the address structure (struct sockaddr_in or struct
271 * sockaddr_in6 [RFC 2553]),
272 * addr_len - the size of the address structure.
274 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
280 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
283 /* Disallow binding twice. */
284 if (!sctp_sk(sk)->ep->base.bind_addr.port)
285 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
295 static long sctp_get_port_local(struct sock *, union sctp_addr *);
297 /* Verify this is a valid sockaddr. */
298 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
299 union sctp_addr *addr, int len)
303 /* Check minimum size. */
304 if (len < sizeof (struct sockaddr))
307 /* V4 mapped address are really of AF_INET family */
308 if (addr->sa.sa_family == AF_INET6 &&
309 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
310 if (!opt->pf->af_supported(AF_INET, opt))
313 /* Does this PF support this AF? */
314 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
318 /* If we get this far, af is valid. */
319 af = sctp_get_af_specific(addr->sa.sa_family);
321 if (len < af->sockaddr_len)
327 /* Bind a local address either to an endpoint or to an association. */
328 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
330 struct net *net = sock_net(sk);
331 struct sctp_sock *sp = sctp_sk(sk);
332 struct sctp_endpoint *ep = sp->ep;
333 struct sctp_bind_addr *bp = &ep->base.bind_addr;
338 /* Common sockaddr verification. */
339 af = sctp_sockaddr_af(sp, addr, len);
341 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
342 __func__, sk, addr, len);
346 snum = ntohs(addr->v4.sin_port);
348 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
349 __func__, sk, &addr->sa, bp->port, snum, len);
351 /* PF specific bind() address verification. */
352 if (!sp->pf->bind_verify(sp, addr))
353 return -EADDRNOTAVAIL;
355 /* We must either be unbound, or bind to the same port.
356 * It's OK to allow 0 ports if we are already bound.
357 * We'll just inhert an already bound port in this case
362 else if (snum != bp->port) {
363 pr_debug("%s: new port %d doesn't match existing port "
364 "%d\n", __func__, snum, bp->port);
369 if (snum && snum < inet_prot_sock(net) &&
370 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
373 /* See if the address matches any of the addresses we may have
374 * already bound before checking against other endpoints.
376 if (sctp_bind_addr_match(bp, addr, sp))
379 /* Make sure we are allowed to bind here.
380 * The function sctp_get_port_local() does duplicate address
383 addr->v4.sin_port = htons(snum);
384 if ((ret = sctp_get_port_local(sk, addr))) {
388 /* Refresh ephemeral port. */
390 bp->port = inet_sk(sk)->inet_num;
392 /* Add the address to the bind address list.
393 * Use GFP_ATOMIC since BHs will be disabled.
395 ret = sctp_add_bind_addr(bp, addr, af->sockaddr_len,
396 SCTP_ADDR_SRC, GFP_ATOMIC);
398 /* Copy back into socket for getsockname() use. */
400 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
401 sp->pf->to_sk_saddr(addr, sk);
407 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
409 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
410 * at any one time. If a sender, after sending an ASCONF chunk, decides
411 * it needs to transfer another ASCONF Chunk, it MUST wait until the
412 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
413 * subsequent ASCONF. Note this restriction binds each side, so at any
414 * time two ASCONF may be in-transit on any given association (one sent
415 * from each endpoint).
417 static int sctp_send_asconf(struct sctp_association *asoc,
418 struct sctp_chunk *chunk)
420 struct net *net = sock_net(asoc->base.sk);
423 /* If there is an outstanding ASCONF chunk, queue it for later
426 if (asoc->addip_last_asconf) {
427 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
431 /* Hold the chunk until an ASCONF_ACK is received. */
432 sctp_chunk_hold(chunk);
433 retval = sctp_primitive_ASCONF(net, asoc, chunk);
435 sctp_chunk_free(chunk);
437 asoc->addip_last_asconf = chunk;
443 /* Add a list of addresses as bind addresses to local endpoint or
446 * Basically run through each address specified in the addrs/addrcnt
447 * array/length pair, determine if it is IPv6 or IPv4 and call
448 * sctp_do_bind() on it.
450 * If any of them fails, then the operation will be reversed and the
451 * ones that were added will be removed.
453 * Only sctp_setsockopt_bindx() is supposed to call this function.
455 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
460 struct sockaddr *sa_addr;
463 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
467 for (cnt = 0; cnt < addrcnt; cnt++) {
468 /* The list may contain either IPv4 or IPv6 address;
469 * determine the address length for walking thru the list.
472 af = sctp_get_af_specific(sa_addr->sa_family);
478 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
481 addr_buf += af->sockaddr_len;
485 /* Failed. Cleanup the ones that have been added */
487 sctp_bindx_rem(sk, addrs, cnt);
495 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
496 * associations that are part of the endpoint indicating that a list of local
497 * addresses are added to the endpoint.
499 * If any of the addresses is already in the bind address list of the
500 * association, we do not send the chunk for that association. But it will not
501 * affect other associations.
503 * Only sctp_setsockopt_bindx() is supposed to call this function.
505 static int sctp_send_asconf_add_ip(struct sock *sk,
506 struct sockaddr *addrs,
509 struct net *net = sock_net(sk);
510 struct sctp_sock *sp;
511 struct sctp_endpoint *ep;
512 struct sctp_association *asoc;
513 struct sctp_bind_addr *bp;
514 struct sctp_chunk *chunk;
515 struct sctp_sockaddr_entry *laddr;
516 union sctp_addr *addr;
517 union sctp_addr saveaddr;
524 if (!net->sctp.addip_enable)
530 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
531 __func__, sk, addrs, addrcnt);
533 list_for_each_entry(asoc, &ep->asocs, asocs) {
534 if (!asoc->peer.asconf_capable)
537 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
540 if (!sctp_state(asoc, ESTABLISHED))
543 /* Check if any address in the packed array of addresses is
544 * in the bind address list of the association. If so,
545 * do not send the asconf chunk to its peer, but continue with
546 * other associations.
549 for (i = 0; i < addrcnt; i++) {
551 af = sctp_get_af_specific(addr->v4.sin_family);
557 if (sctp_assoc_lookup_laddr(asoc, addr))
560 addr_buf += af->sockaddr_len;
565 /* Use the first valid address in bind addr list of
566 * association as Address Parameter of ASCONF CHUNK.
568 bp = &asoc->base.bind_addr;
569 p = bp->address_list.next;
570 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
571 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
572 addrcnt, SCTP_PARAM_ADD_IP);
578 /* Add the new addresses to the bind address list with
579 * use_as_src set to 0.
582 for (i = 0; i < addrcnt; i++) {
584 af = sctp_get_af_specific(addr->v4.sin_family);
585 memcpy(&saveaddr, addr, af->sockaddr_len);
586 retval = sctp_add_bind_addr(bp, &saveaddr,
588 SCTP_ADDR_NEW, GFP_ATOMIC);
589 addr_buf += af->sockaddr_len;
591 if (asoc->src_out_of_asoc_ok) {
592 struct sctp_transport *trans;
594 list_for_each_entry(trans,
595 &asoc->peer.transport_addr_list, transports) {
596 /* Clear the source and route cache */
597 sctp_transport_dst_release(trans);
598 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
599 2*asoc->pathmtu, 4380));
600 trans->ssthresh = asoc->peer.i.a_rwnd;
601 trans->rto = asoc->rto_initial;
602 sctp_max_rto(asoc, trans);
603 trans->rtt = trans->srtt = trans->rttvar = 0;
604 sctp_transport_route(trans, NULL,
605 sctp_sk(asoc->base.sk));
608 retval = sctp_send_asconf(asoc, chunk);
615 /* Remove a list of addresses from bind addresses list. Do not remove the
618 * Basically run through each address specified in the addrs/addrcnt
619 * array/length pair, determine if it is IPv6 or IPv4 and call
620 * sctp_del_bind() on it.
622 * If any of them fails, then the operation will be reversed and the
623 * ones that were removed will be added back.
625 * At least one address has to be left; if only one address is
626 * available, the operation will return -EBUSY.
628 * Only sctp_setsockopt_bindx() is supposed to call this function.
630 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
632 struct sctp_sock *sp = sctp_sk(sk);
633 struct sctp_endpoint *ep = sp->ep;
635 struct sctp_bind_addr *bp = &ep->base.bind_addr;
638 union sctp_addr *sa_addr;
641 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
642 __func__, sk, addrs, addrcnt);
645 for (cnt = 0; cnt < addrcnt; cnt++) {
646 /* If the bind address list is empty or if there is only one
647 * bind address, there is nothing more to be removed (we need
648 * at least one address here).
650 if (list_empty(&bp->address_list) ||
651 (sctp_list_single_entry(&bp->address_list))) {
657 af = sctp_get_af_specific(sa_addr->sa.sa_family);
663 if (!af->addr_valid(sa_addr, sp, NULL)) {
664 retval = -EADDRNOTAVAIL;
668 if (sa_addr->v4.sin_port &&
669 sa_addr->v4.sin_port != htons(bp->port)) {
674 if (!sa_addr->v4.sin_port)
675 sa_addr->v4.sin_port = htons(bp->port);
677 /* FIXME - There is probably a need to check if sk->sk_saddr and
678 * sk->sk_rcv_addr are currently set to one of the addresses to
679 * be removed. This is something which needs to be looked into
680 * when we are fixing the outstanding issues with multi-homing
681 * socket routing and failover schemes. Refer to comments in
682 * sctp_do_bind(). -daisy
684 retval = sctp_del_bind_addr(bp, sa_addr);
686 addr_buf += af->sockaddr_len;
689 /* Failed. Add the ones that has been removed back */
691 sctp_bindx_add(sk, addrs, cnt);
699 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
700 * the associations that are part of the endpoint indicating that a list of
701 * local addresses are removed from the endpoint.
703 * If any of the addresses is already in the bind address list of the
704 * association, we do not send the chunk for that association. But it will not
705 * affect other associations.
707 * Only sctp_setsockopt_bindx() is supposed to call this function.
709 static int sctp_send_asconf_del_ip(struct sock *sk,
710 struct sockaddr *addrs,
713 struct net *net = sock_net(sk);
714 struct sctp_sock *sp;
715 struct sctp_endpoint *ep;
716 struct sctp_association *asoc;
717 struct sctp_transport *transport;
718 struct sctp_bind_addr *bp;
719 struct sctp_chunk *chunk;
720 union sctp_addr *laddr;
723 struct sctp_sockaddr_entry *saddr;
729 if (!net->sctp.addip_enable)
735 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
736 __func__, sk, addrs, addrcnt);
738 list_for_each_entry(asoc, &ep->asocs, asocs) {
740 if (!asoc->peer.asconf_capable)
743 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
746 if (!sctp_state(asoc, ESTABLISHED))
749 /* Check if any address in the packed array of addresses is
750 * not present in the bind address list of the association.
751 * If so, do not send the asconf chunk to its peer, but
752 * continue with other associations.
755 for (i = 0; i < addrcnt; i++) {
757 af = sctp_get_af_specific(laddr->v4.sin_family);
763 if (!sctp_assoc_lookup_laddr(asoc, laddr))
766 addr_buf += af->sockaddr_len;
771 /* Find one address in the association's bind address list
772 * that is not in the packed array of addresses. This is to
773 * make sure that we do not delete all the addresses in the
776 bp = &asoc->base.bind_addr;
777 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
779 if ((laddr == NULL) && (addrcnt == 1)) {
780 if (asoc->asconf_addr_del_pending)
782 asoc->asconf_addr_del_pending =
783 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
784 if (asoc->asconf_addr_del_pending == NULL) {
788 asoc->asconf_addr_del_pending->sa.sa_family =
790 asoc->asconf_addr_del_pending->v4.sin_port =
792 if (addrs->sa_family == AF_INET) {
793 struct sockaddr_in *sin;
795 sin = (struct sockaddr_in *)addrs;
796 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
797 } else if (addrs->sa_family == AF_INET6) {
798 struct sockaddr_in6 *sin6;
800 sin6 = (struct sockaddr_in6 *)addrs;
801 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
804 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
805 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
806 asoc->asconf_addr_del_pending);
808 asoc->src_out_of_asoc_ok = 1;
816 /* We do not need RCU protection throughout this loop
817 * because this is done under a socket lock from the
820 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
828 /* Reset use_as_src flag for the addresses in the bind address
829 * list that are to be deleted.
832 for (i = 0; i < addrcnt; i++) {
834 af = sctp_get_af_specific(laddr->v4.sin_family);
835 list_for_each_entry(saddr, &bp->address_list, list) {
836 if (sctp_cmp_addr_exact(&saddr->a, laddr))
837 saddr->state = SCTP_ADDR_DEL;
839 addr_buf += af->sockaddr_len;
842 /* Update the route and saddr entries for all the transports
843 * as some of the addresses in the bind address list are
844 * about to be deleted and cannot be used as source addresses.
846 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
848 sctp_transport_dst_release(transport);
849 sctp_transport_route(transport, NULL,
850 sctp_sk(asoc->base.sk));
854 /* We don't need to transmit ASCONF */
856 retval = sctp_send_asconf(asoc, chunk);
862 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
863 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
865 struct sock *sk = sctp_opt2sk(sp);
866 union sctp_addr *addr;
869 /* It is safe to write port space in caller. */
871 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
872 af = sctp_get_af_specific(addr->sa.sa_family);
875 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
878 if (addrw->state == SCTP_ADDR_NEW)
879 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
881 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
884 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
887 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
890 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
891 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
894 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
895 * Section 3.1.2 for this usage.
897 * addrs is a pointer to an array of one or more socket addresses. Each
898 * address is contained in its appropriate structure (i.e. struct
899 * sockaddr_in or struct sockaddr_in6) the family of the address type
900 * must be used to distinguish the address length (note that this
901 * representation is termed a "packed array" of addresses). The caller
902 * specifies the number of addresses in the array with addrcnt.
904 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
905 * -1, and sets errno to the appropriate error code.
907 * For SCTP, the port given in each socket address must be the same, or
908 * sctp_bindx() will fail, setting errno to EINVAL.
910 * The flags parameter is formed from the bitwise OR of zero or more of
911 * the following currently defined flags:
913 * SCTP_BINDX_ADD_ADDR
915 * SCTP_BINDX_REM_ADDR
917 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
918 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
919 * addresses from the association. The two flags are mutually exclusive;
920 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
921 * not remove all addresses from an association; sctp_bindx() will
922 * reject such an attempt with EINVAL.
924 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
925 * additional addresses with an endpoint after calling bind(). Or use
926 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
927 * socket is associated with so that no new association accepted will be
928 * associated with those addresses. If the endpoint supports dynamic
929 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
930 * endpoint to send the appropriate message to the peer to change the
931 * peers address lists.
933 * Adding and removing addresses from a connected association is
934 * optional functionality. Implementations that do not support this
935 * functionality should return EOPNOTSUPP.
937 * Basically do nothing but copying the addresses from user to kernel
938 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
939 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
942 * We don't use copy_from_user() for optimization: we first do the
943 * sanity checks (buffer size -fast- and access check-healthy
944 * pointer); if all of those succeed, then we can alloc the memory
945 * (expensive operation) needed to copy the data to kernel. Then we do
946 * the copying without checking the user space area
947 * (__copy_from_user()).
949 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
952 * sk The sk of the socket
953 * addrs The pointer to the addresses in user land
954 * addrssize Size of the addrs buffer
955 * op Operation to perform (add or remove, see the flags of
958 * Returns 0 if ok, <0 errno code on error.
960 static int sctp_setsockopt_bindx(struct sock *sk,
961 struct sockaddr __user *addrs,
962 int addrs_size, int op)
964 struct sockaddr *kaddrs;
968 struct sockaddr *sa_addr;
972 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
973 __func__, sk, addrs, addrs_size, op);
975 if (unlikely(addrs_size <= 0))
978 /* Check the user passed a healthy pointer. */
979 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
982 /* Alloc space for the address array in kernel memory. */
983 kaddrs = kmalloc(addrs_size, GFP_USER | __GFP_NOWARN);
984 if (unlikely(!kaddrs))
987 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
992 /* Walk through the addrs buffer and count the number of addresses. */
994 while (walk_size < addrs_size) {
995 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1001 af = sctp_get_af_specific(sa_addr->sa_family);
1003 /* If the address family is not supported or if this address
1004 * causes the address buffer to overflow return EINVAL.
1006 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1011 addr_buf += af->sockaddr_len;
1012 walk_size += af->sockaddr_len;
1017 case SCTP_BINDX_ADD_ADDR:
1018 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1021 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1024 case SCTP_BINDX_REM_ADDR:
1025 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1028 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1042 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1044 * Common routine for handling connect() and sctp_connectx().
1045 * Connect will come in with just a single address.
1047 static int __sctp_connect(struct sock *sk,
1048 struct sockaddr *kaddrs,
1050 sctp_assoc_t *assoc_id)
1052 struct net *net = sock_net(sk);
1053 struct sctp_sock *sp;
1054 struct sctp_endpoint *ep;
1055 struct sctp_association *asoc = NULL;
1056 struct sctp_association *asoc2;
1057 struct sctp_transport *transport;
1059 enum sctp_scope scope;
1064 union sctp_addr *sa_addr = NULL;
1066 unsigned short port;
1067 unsigned int f_flags = 0;
1072 /* connect() cannot be done on a socket that is already in ESTABLISHED
1073 * state - UDP-style peeled off socket or a TCP-style socket that
1074 * is already connected.
1075 * It cannot be done even on a TCP-style listening socket.
1077 if (sctp_sstate(sk, ESTABLISHED) || sctp_sstate(sk, CLOSING) ||
1078 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1083 /* Walk through the addrs buffer and count the number of addresses. */
1085 while (walk_size < addrs_size) {
1088 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1094 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1096 /* If the address family is not supported or if this address
1097 * causes the address buffer to overflow return EINVAL.
1099 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1104 port = ntohs(sa_addr->v4.sin_port);
1106 /* Save current address so we can work with it */
1107 memcpy(&to, sa_addr, af->sockaddr_len);
1109 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1113 /* Make sure the destination port is correctly set
1116 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1121 /* Check if there already is a matching association on the
1122 * endpoint (other than the one created here).
1124 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1125 if (asoc2 && asoc2 != asoc) {
1126 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1133 /* If we could not find a matching association on the endpoint,
1134 * make sure that there is no peeled-off association matching
1135 * the peer address even on another socket.
1137 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1138 err = -EADDRNOTAVAIL;
1143 /* If a bind() or sctp_bindx() is not called prior to
1144 * an sctp_connectx() call, the system picks an
1145 * ephemeral port and will choose an address set
1146 * equivalent to binding with a wildcard address.
1148 if (!ep->base.bind_addr.port) {
1149 if (sctp_autobind(sk)) {
1155 * If an unprivileged user inherits a 1-many
1156 * style socket with open associations on a
1157 * privileged port, it MAY be permitted to
1158 * accept new associations, but it SHOULD NOT
1159 * be permitted to open new associations.
1161 if (ep->base.bind_addr.port <
1162 inet_prot_sock(net) &&
1163 !ns_capable(net->user_ns,
1164 CAP_NET_BIND_SERVICE)) {
1170 scope = sctp_scope(&to);
1171 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1177 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1185 /* Prime the peer's transport structures. */
1186 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1194 addr_buf += af->sockaddr_len;
1195 walk_size += af->sockaddr_len;
1198 /* In case the user of sctp_connectx() wants an association
1199 * id back, assign one now.
1202 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1207 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1212 /* Initialize sk's dport and daddr for getpeername() */
1213 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1214 sp->pf->to_sk_daddr(sa_addr, sk);
1217 /* in-kernel sockets don't generally have a file allocated to them
1218 * if all they do is call sock_create_kern().
1220 if (sk->sk_socket->file)
1221 f_flags = sk->sk_socket->file->f_flags;
1223 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1226 *assoc_id = asoc->assoc_id;
1227 err = sctp_wait_for_connect(asoc, &timeo);
1228 /* Note: the asoc may be freed after the return of
1229 * sctp_wait_for_connect.
1232 /* Don't free association on exit. */
1236 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1237 __func__, asoc, kaddrs, err);
1240 /* sctp_primitive_ASSOCIATE may have added this association
1241 * To the hash table, try to unhash it, just in case, its a noop
1242 * if it wasn't hashed so we're safe
1244 sctp_association_free(asoc);
1249 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1252 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1253 * sctp_assoc_t *asoc);
1255 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1256 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1257 * or IPv6 addresses.
1259 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1260 * Section 3.1.2 for this usage.
1262 * addrs is a pointer to an array of one or more socket addresses. Each
1263 * address is contained in its appropriate structure (i.e. struct
1264 * sockaddr_in or struct sockaddr_in6) the family of the address type
1265 * must be used to distengish the address length (note that this
1266 * representation is termed a "packed array" of addresses). The caller
1267 * specifies the number of addresses in the array with addrcnt.
1269 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1270 * the association id of the new association. On failure, sctp_connectx()
1271 * returns -1, and sets errno to the appropriate error code. The assoc_id
1272 * is not touched by the kernel.
1274 * For SCTP, the port given in each socket address must be the same, or
1275 * sctp_connectx() will fail, setting errno to EINVAL.
1277 * An application can use sctp_connectx to initiate an association with
1278 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1279 * allows a caller to specify multiple addresses at which a peer can be
1280 * reached. The way the SCTP stack uses the list of addresses to set up
1281 * the association is implementation dependent. This function only
1282 * specifies that the stack will try to make use of all the addresses in
1283 * the list when needed.
1285 * Note that the list of addresses passed in is only used for setting up
1286 * the association. It does not necessarily equal the set of addresses
1287 * the peer uses for the resulting association. If the caller wants to
1288 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1289 * retrieve them after the association has been set up.
1291 * Basically do nothing but copying the addresses from user to kernel
1292 * land and invoking either sctp_connectx(). This is used for tunneling
1293 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1295 * We don't use copy_from_user() for optimization: we first do the
1296 * sanity checks (buffer size -fast- and access check-healthy
1297 * pointer); if all of those succeed, then we can alloc the memory
1298 * (expensive operation) needed to copy the data to kernel. Then we do
1299 * the copying without checking the user space area
1300 * (__copy_from_user()).
1302 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1305 * sk The sk of the socket
1306 * addrs The pointer to the addresses in user land
1307 * addrssize Size of the addrs buffer
1309 * Returns >=0 if ok, <0 errno code on error.
1311 static int __sctp_setsockopt_connectx(struct sock *sk,
1312 struct sockaddr __user *addrs,
1314 sctp_assoc_t *assoc_id)
1316 struct sockaddr *kaddrs;
1317 gfp_t gfp = GFP_KERNEL;
1320 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1321 __func__, sk, addrs, addrs_size);
1323 if (unlikely(addrs_size <= 0))
1326 /* Check the user passed a healthy pointer. */
1327 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1330 /* Alloc space for the address array in kernel memory. */
1331 if (sk->sk_socket->file)
1332 gfp = GFP_USER | __GFP_NOWARN;
1333 kaddrs = kmalloc(addrs_size, gfp);
1334 if (unlikely(!kaddrs))
1337 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1340 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1349 * This is an older interface. It's kept for backward compatibility
1350 * to the option that doesn't provide association id.
1352 static int sctp_setsockopt_connectx_old(struct sock *sk,
1353 struct sockaddr __user *addrs,
1356 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1360 * New interface for the API. The since the API is done with a socket
1361 * option, to make it simple we feed back the association id is as a return
1362 * indication to the call. Error is always negative and association id is
1365 static int sctp_setsockopt_connectx(struct sock *sk,
1366 struct sockaddr __user *addrs,
1369 sctp_assoc_t assoc_id = 0;
1372 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1381 * New (hopefully final) interface for the API.
1382 * We use the sctp_getaddrs_old structure so that use-space library
1383 * can avoid any unnecessary allocations. The only different part
1384 * is that we store the actual length of the address buffer into the
1385 * addrs_num structure member. That way we can re-use the existing
1388 #ifdef CONFIG_COMPAT
1389 struct compat_sctp_getaddrs_old {
1390 sctp_assoc_t assoc_id;
1392 compat_uptr_t addrs; /* struct sockaddr * */
1396 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1397 char __user *optval,
1400 struct sctp_getaddrs_old param;
1401 sctp_assoc_t assoc_id = 0;
1404 #ifdef CONFIG_COMPAT
1405 if (in_compat_syscall()) {
1406 struct compat_sctp_getaddrs_old param32;
1408 if (len < sizeof(param32))
1410 if (copy_from_user(¶m32, optval, sizeof(param32)))
1413 param.assoc_id = param32.assoc_id;
1414 param.addr_num = param32.addr_num;
1415 param.addrs = compat_ptr(param32.addrs);
1419 if (len < sizeof(param))
1421 if (copy_from_user(¶m, optval, sizeof(param)))
1425 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1426 param.addrs, param.addr_num,
1428 if (err == 0 || err == -EINPROGRESS) {
1429 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1431 if (put_user(sizeof(assoc_id), optlen))
1438 /* API 3.1.4 close() - UDP Style Syntax
1439 * Applications use close() to perform graceful shutdown (as described in
1440 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1441 * by a UDP-style socket.
1445 * ret = close(int sd);
1447 * sd - the socket descriptor of the associations to be closed.
1449 * To gracefully shutdown a specific association represented by the
1450 * UDP-style socket, an application should use the sendmsg() call,
1451 * passing no user data, but including the appropriate flag in the
1452 * ancillary data (see Section xxxx).
1454 * If sd in the close() call is a branched-off socket representing only
1455 * one association, the shutdown is performed on that association only.
1457 * 4.1.6 close() - TCP Style Syntax
1459 * Applications use close() to gracefully close down an association.
1463 * int close(int sd);
1465 * sd - the socket descriptor of the association to be closed.
1467 * After an application calls close() on a socket descriptor, no further
1468 * socket operations will succeed on that descriptor.
1470 * API 7.1.4 SO_LINGER
1472 * An application using the TCP-style socket can use this option to
1473 * perform the SCTP ABORT primitive. The linger option structure is:
1476 * int l_onoff; // option on/off
1477 * int l_linger; // linger time
1480 * To enable the option, set l_onoff to 1. If the l_linger value is set
1481 * to 0, calling close() is the same as the ABORT primitive. If the
1482 * value is set to a negative value, the setsockopt() call will return
1483 * an error. If the value is set to a positive value linger_time, the
1484 * close() can be blocked for at most linger_time ms. If the graceful
1485 * shutdown phase does not finish during this period, close() will
1486 * return but the graceful shutdown phase continues in the system.
1488 static void sctp_close(struct sock *sk, long timeout)
1490 struct net *net = sock_net(sk);
1491 struct sctp_endpoint *ep;
1492 struct sctp_association *asoc;
1493 struct list_head *pos, *temp;
1494 unsigned int data_was_unread;
1496 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1498 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
1499 sk->sk_shutdown = SHUTDOWN_MASK;
1500 sk->sk_state = SCTP_SS_CLOSING;
1502 ep = sctp_sk(sk)->ep;
1504 /* Clean up any skbs sitting on the receive queue. */
1505 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1506 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1508 /* Walk all associations on an endpoint. */
1509 list_for_each_safe(pos, temp, &ep->asocs) {
1510 asoc = list_entry(pos, struct sctp_association, asocs);
1512 if (sctp_style(sk, TCP)) {
1513 /* A closed association can still be in the list if
1514 * it belongs to a TCP-style listening socket that is
1515 * not yet accepted. If so, free it. If not, send an
1516 * ABORT or SHUTDOWN based on the linger options.
1518 if (sctp_state(asoc, CLOSED)) {
1519 sctp_association_free(asoc);
1524 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1525 !skb_queue_empty(&asoc->ulpq.reasm) ||
1526 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1527 struct sctp_chunk *chunk;
1529 chunk = sctp_make_abort_user(asoc, NULL, 0);
1530 sctp_primitive_ABORT(net, asoc, chunk);
1532 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1535 /* On a TCP-style socket, block for at most linger_time if set. */
1536 if (sctp_style(sk, TCP) && timeout)
1537 sctp_wait_for_close(sk, timeout);
1539 /* This will run the backlog queue. */
1542 /* Supposedly, no process has access to the socket, but
1543 * the net layers still may.
1544 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1545 * held and that should be grabbed before socket lock.
1547 spin_lock_bh(&net->sctp.addr_wq_lock);
1548 bh_lock_sock_nested(sk);
1550 /* Hold the sock, since sk_common_release() will put sock_put()
1551 * and we have just a little more cleanup.
1554 sk_common_release(sk);
1557 spin_unlock_bh(&net->sctp.addr_wq_lock);
1561 SCTP_DBG_OBJCNT_DEC(sock);
1564 /* Handle EPIPE error. */
1565 static int sctp_error(struct sock *sk, int flags, int err)
1568 err = sock_error(sk) ? : -EPIPE;
1569 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1570 send_sig(SIGPIPE, current, 0);
1574 /* API 3.1.3 sendmsg() - UDP Style Syntax
1576 * An application uses sendmsg() and recvmsg() calls to transmit data to
1577 * and receive data from its peer.
1579 * ssize_t sendmsg(int socket, const struct msghdr *message,
1582 * socket - the socket descriptor of the endpoint.
1583 * message - pointer to the msghdr structure which contains a single
1584 * user message and possibly some ancillary data.
1586 * See Section 5 for complete description of the data
1589 * flags - flags sent or received with the user message, see Section
1590 * 5 for complete description of the flags.
1592 * Note: This function could use a rewrite especially when explicit
1593 * connect support comes in.
1595 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1597 static int sctp_msghdr_parse(const struct msghdr *msg,
1598 struct sctp_cmsgs *cmsgs);
1600 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
1602 struct net *net = sock_net(sk);
1603 struct sctp_sock *sp;
1604 struct sctp_endpoint *ep;
1605 struct sctp_association *new_asoc = NULL, *asoc = NULL;
1606 struct sctp_transport *transport, *chunk_tp;
1607 struct sctp_chunk *chunk;
1609 struct sockaddr *msg_name = NULL;
1610 struct sctp_sndrcvinfo default_sinfo;
1611 struct sctp_sndrcvinfo *sinfo;
1612 struct sctp_initmsg *sinit;
1613 sctp_assoc_t associd = 0;
1614 struct sctp_cmsgs cmsgs = { NULL };
1615 enum sctp_scope scope;
1616 bool fill_sinfo_ttl = false, wait_connect = false;
1617 struct sctp_datamsg *datamsg;
1618 int msg_flags = msg->msg_flags;
1619 __u16 sinfo_flags = 0;
1627 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1630 /* We cannot send a message over a TCP-style listening socket. */
1631 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1636 /* Parse out the SCTP CMSGs. */
1637 err = sctp_msghdr_parse(msg, &cmsgs);
1639 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1643 /* Fetch the destination address for this packet. This
1644 * address only selects the association--it is not necessarily
1645 * the address we will send to.
1646 * For a peeled-off socket, msg_name is ignored.
1648 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1649 int msg_namelen = msg->msg_namelen;
1651 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1656 if (msg_namelen > sizeof(to))
1657 msg_namelen = sizeof(to);
1658 memcpy(&to, msg->msg_name, msg_namelen);
1659 msg_name = msg->msg_name;
1663 if (cmsgs.sinfo != NULL) {
1664 memset(&default_sinfo, 0, sizeof(default_sinfo));
1665 default_sinfo.sinfo_stream = cmsgs.sinfo->snd_sid;
1666 default_sinfo.sinfo_flags = cmsgs.sinfo->snd_flags;
1667 default_sinfo.sinfo_ppid = cmsgs.sinfo->snd_ppid;
1668 default_sinfo.sinfo_context = cmsgs.sinfo->snd_context;
1669 default_sinfo.sinfo_assoc_id = cmsgs.sinfo->snd_assoc_id;
1671 sinfo = &default_sinfo;
1672 fill_sinfo_ttl = true;
1674 sinfo = cmsgs.srinfo;
1676 /* Did the user specify SNDINFO/SNDRCVINFO? */
1678 sinfo_flags = sinfo->sinfo_flags;
1679 associd = sinfo->sinfo_assoc_id;
1682 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1683 msg_len, sinfo_flags);
1685 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1686 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1691 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1692 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1693 * If SCTP_ABORT is set, the message length could be non zero with
1694 * the msg_iov set to the user abort reason.
1696 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1697 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1702 /* If SCTP_ADDR_OVER is set, there must be an address
1703 * specified in msg_name.
1705 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1712 pr_debug("%s: about to look up association\n", __func__);
1716 /* If a msg_name has been specified, assume this is to be used. */
1718 /* Look for a matching association on the endpoint. */
1719 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1721 /* If we could not find a matching association on the
1722 * endpoint, make sure that it is not a TCP-style
1723 * socket that already has an association or there is
1724 * no peeled-off association on another socket.
1727 ((sctp_style(sk, TCP) &&
1728 (sctp_sstate(sk, ESTABLISHED) ||
1729 sctp_sstate(sk, CLOSING))) ||
1730 sctp_endpoint_is_peeled_off(ep, &to))) {
1731 err = -EADDRNOTAVAIL;
1735 asoc = sctp_id2assoc(sk, associd);
1743 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1745 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1746 * socket that has an association in CLOSED state. This can
1747 * happen when an accepted socket has an association that is
1750 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1755 if (sinfo_flags & SCTP_EOF) {
1756 pr_debug("%s: shutting down association:%p\n",
1759 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1763 if (sinfo_flags & SCTP_ABORT) {
1765 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1771 pr_debug("%s: aborting association:%p\n",
1774 sctp_primitive_ABORT(net, asoc, chunk);
1780 /* Do we need to create the association? */
1782 pr_debug("%s: there is no association yet\n", __func__);
1784 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1789 /* Check for invalid stream against the stream counts,
1790 * either the default or the user specified stream counts.
1793 if (!sinit || !sinit->sinit_num_ostreams) {
1794 /* Check against the defaults. */
1795 if (sinfo->sinfo_stream >=
1796 sp->initmsg.sinit_num_ostreams) {
1801 /* Check against the requested. */
1802 if (sinfo->sinfo_stream >=
1803 sinit->sinit_num_ostreams) {
1811 * API 3.1.2 bind() - UDP Style Syntax
1812 * If a bind() or sctp_bindx() is not called prior to a
1813 * sendmsg() call that initiates a new association, the
1814 * system picks an ephemeral port and will choose an address
1815 * set equivalent to binding with a wildcard address.
1817 if (!ep->base.bind_addr.port) {
1818 if (sctp_autobind(sk)) {
1824 * If an unprivileged user inherits a one-to-many
1825 * style socket with open associations on a privileged
1826 * port, it MAY be permitted to accept new associations,
1827 * but it SHOULD NOT be permitted to open new
1830 if (ep->base.bind_addr.port < inet_prot_sock(net) &&
1831 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1837 scope = sctp_scope(&to);
1838 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1844 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1850 /* If the SCTP_INIT ancillary data is specified, set all
1851 * the association init values accordingly.
1854 if (sinit->sinit_num_ostreams) {
1855 asoc->c.sinit_num_ostreams =
1856 sinit->sinit_num_ostreams;
1858 if (sinit->sinit_max_instreams) {
1859 asoc->c.sinit_max_instreams =
1860 sinit->sinit_max_instreams;
1862 if (sinit->sinit_max_attempts) {
1863 asoc->max_init_attempts
1864 = sinit->sinit_max_attempts;
1866 if (sinit->sinit_max_init_timeo) {
1867 asoc->max_init_timeo =
1868 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1872 /* Prime the peer's transport structures. */
1873 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1880 /* ASSERT: we have a valid association at this point. */
1881 pr_debug("%s: we have a valid association\n", __func__);
1884 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1885 * one with some defaults.
1887 memset(&default_sinfo, 0, sizeof(default_sinfo));
1888 default_sinfo.sinfo_stream = asoc->default_stream;
1889 default_sinfo.sinfo_flags = asoc->default_flags;
1890 default_sinfo.sinfo_ppid = asoc->default_ppid;
1891 default_sinfo.sinfo_context = asoc->default_context;
1892 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1893 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1895 sinfo = &default_sinfo;
1896 } else if (fill_sinfo_ttl) {
1897 /* In case SNDINFO was specified, we still need to fill
1898 * it with a default ttl from the assoc here.
1900 sinfo->sinfo_timetolive = asoc->default_timetolive;
1903 /* API 7.1.7, the sndbuf size per association bounds the
1904 * maximum size of data that can be sent in a single send call.
1906 if (msg_len > sk->sk_sndbuf) {
1911 if (asoc->pmtu_pending)
1912 sctp_assoc_pending_pmtu(asoc);
1914 /* If fragmentation is disabled and the message length exceeds the
1915 * association fragmentation point, return EMSGSIZE. The I-D
1916 * does not specify what this error is, but this looks like
1919 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1924 /* Check for invalid stream. */
1925 if (sinfo->sinfo_stream >= asoc->stream.outcnt) {
1930 /* Allocate sctp_stream_out_ext if not already done */
1931 if (unlikely(!asoc->stream.out[sinfo->sinfo_stream].ext)) {
1932 err = sctp_stream_init_ext(&asoc->stream, sinfo->sinfo_stream);
1937 if (sctp_wspace(asoc) < msg_len)
1938 sctp_prsctp_prune(asoc, sinfo, msg_len - sctp_wspace(asoc));
1940 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1941 if (!sctp_wspace(asoc)) {
1942 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1947 /* If an address is passed with the sendto/sendmsg call, it is used
1948 * to override the primary destination address in the TCP model, or
1949 * when SCTP_ADDR_OVER flag is set in the UDP model.
1951 if ((sctp_style(sk, TCP) && msg_name) ||
1952 (sinfo_flags & SCTP_ADDR_OVER)) {
1953 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1961 /* Auto-connect, if we aren't connected already. */
1962 if (sctp_state(asoc, CLOSED)) {
1963 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1967 wait_connect = true;
1968 pr_debug("%s: we associated primitively\n", __func__);
1971 /* Break the message into multiple chunks of maximum size. */
1972 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
1973 if (IS_ERR(datamsg)) {
1974 err = PTR_ERR(datamsg);
1977 asoc->force_delay = !!(msg->msg_flags & MSG_MORE);
1979 /* Now send the (possibly) fragmented message. */
1980 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1981 sctp_chunk_hold(chunk);
1983 /* Do accounting for the write space. */
1984 sctp_set_owner_w(chunk);
1986 chunk->transport = chunk_tp;
1989 /* Send it to the lower layers. Note: all chunks
1990 * must either fail or succeed. The lower layer
1991 * works that way today. Keep it that way or this
1994 err = sctp_primitive_SEND(net, asoc, datamsg);
1995 /* Did the lower layer accept the chunk? */
1997 sctp_datamsg_free(datamsg);
2001 pr_debug("%s: we sent primitively\n", __func__);
2003 sctp_datamsg_put(datamsg);
2006 if (unlikely(wait_connect)) {
2007 timeo = sock_sndtimeo(sk, msg_flags & MSG_DONTWAIT);
2008 sctp_wait_for_connect(asoc, &timeo);
2011 /* If we are already past ASSOCIATE, the lower
2012 * layers are responsible for association cleanup.
2018 sctp_association_free(asoc);
2023 return sctp_error(sk, msg_flags, err);
2030 err = sock_error(sk);
2040 /* This is an extended version of skb_pull() that removes the data from the
2041 * start of a skb even when data is spread across the list of skb's in the
2042 * frag_list. len specifies the total amount of data that needs to be removed.
2043 * when 'len' bytes could be removed from the skb, it returns 0.
2044 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2045 * could not be removed.
2047 static int sctp_skb_pull(struct sk_buff *skb, int len)
2049 struct sk_buff *list;
2050 int skb_len = skb_headlen(skb);
2053 if (len <= skb_len) {
2054 __skb_pull(skb, len);
2058 __skb_pull(skb, skb_len);
2060 skb_walk_frags(skb, list) {
2061 rlen = sctp_skb_pull(list, len);
2062 skb->len -= (len-rlen);
2063 skb->data_len -= (len-rlen);
2074 /* API 3.1.3 recvmsg() - UDP Style Syntax
2076 * ssize_t recvmsg(int socket, struct msghdr *message,
2079 * socket - the socket descriptor of the endpoint.
2080 * message - pointer to the msghdr structure which contains a single
2081 * user message and possibly some ancillary data.
2083 * See Section 5 for complete description of the data
2086 * flags - flags sent or received with the user message, see Section
2087 * 5 for complete description of the flags.
2089 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2090 int noblock, int flags, int *addr_len)
2092 struct sctp_ulpevent *event = NULL;
2093 struct sctp_sock *sp = sctp_sk(sk);
2094 struct sk_buff *skb, *head_skb;
2099 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2100 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2105 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED) &&
2106 !sctp_sstate(sk, CLOSING) && !sctp_sstate(sk, CLOSED)) {
2111 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2115 /* Get the total length of the skb including any skb's in the
2124 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2126 event = sctp_skb2event(skb);
2131 if (event->chunk && event->chunk->head_skb)
2132 head_skb = event->chunk->head_skb;
2135 sock_recv_ts_and_drops(msg, sk, head_skb);
2136 if (sctp_ulpevent_is_notification(event)) {
2137 msg->msg_flags |= MSG_NOTIFICATION;
2138 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2140 sp->pf->skb_msgname(head_skb, msg->msg_name, addr_len);
2143 /* Check if we allow SCTP_NXTINFO. */
2144 if (sp->recvnxtinfo)
2145 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2146 /* Check if we allow SCTP_RCVINFO. */
2147 if (sp->recvrcvinfo)
2148 sctp_ulpevent_read_rcvinfo(event, msg);
2149 /* Check if we allow SCTP_SNDRCVINFO. */
2150 if (sp->subscribe.sctp_data_io_event)
2151 sctp_ulpevent_read_sndrcvinfo(event, msg);
2155 /* If skb's length exceeds the user's buffer, update the skb and
2156 * push it back to the receive_queue so that the next call to
2157 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2159 if (skb_len > copied) {
2160 msg->msg_flags &= ~MSG_EOR;
2161 if (flags & MSG_PEEK)
2163 sctp_skb_pull(skb, copied);
2164 skb_queue_head(&sk->sk_receive_queue, skb);
2166 /* When only partial message is copied to the user, increase
2167 * rwnd by that amount. If all the data in the skb is read,
2168 * rwnd is updated when the event is freed.
2170 if (!sctp_ulpevent_is_notification(event))
2171 sctp_assoc_rwnd_increase(event->asoc, copied);
2173 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2174 (event->msg_flags & MSG_EOR))
2175 msg->msg_flags |= MSG_EOR;
2177 msg->msg_flags &= ~MSG_EOR;
2180 if (flags & MSG_PEEK) {
2181 /* Release the skb reference acquired after peeking the skb in
2182 * sctp_skb_recv_datagram().
2186 /* Free the event which includes releasing the reference to
2187 * the owner of the skb, freeing the skb and updating the
2190 sctp_ulpevent_free(event);
2197 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2199 * This option is a on/off flag. If enabled no SCTP message
2200 * fragmentation will be performed. Instead if a message being sent
2201 * exceeds the current PMTU size, the message will NOT be sent and
2202 * instead a error will be indicated to the user.
2204 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2205 char __user *optval,
2206 unsigned int optlen)
2210 if (optlen < sizeof(int))
2213 if (get_user(val, (int __user *)optval))
2216 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2221 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2222 unsigned int optlen)
2224 struct sctp_association *asoc;
2225 struct sctp_ulpevent *event;
2227 if (optlen > sizeof(struct sctp_event_subscribe))
2229 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2232 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2233 * if there is no data to be sent or retransmit, the stack will
2234 * immediately send up this notification.
2236 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2237 &sctp_sk(sk)->subscribe)) {
2238 asoc = sctp_id2assoc(sk, 0);
2240 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2241 event = sctp_ulpevent_make_sender_dry_event(asoc,
2246 sctp_ulpq_tail_event(&asoc->ulpq, event);
2253 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2255 * This socket option is applicable to the UDP-style socket only. When
2256 * set it will cause associations that are idle for more than the
2257 * specified number of seconds to automatically close. An association
2258 * being idle is defined an association that has NOT sent or received
2259 * user data. The special value of '0' indicates that no automatic
2260 * close of any associations should be performed. The option expects an
2261 * integer defining the number of seconds of idle time before an
2262 * association is closed.
2264 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2265 unsigned int optlen)
2267 struct sctp_sock *sp = sctp_sk(sk);
2268 struct net *net = sock_net(sk);
2270 /* Applicable to UDP-style socket only */
2271 if (sctp_style(sk, TCP))
2273 if (optlen != sizeof(int))
2275 if (copy_from_user(&sp->autoclose, optval, optlen))
2278 if (sp->autoclose > net->sctp.max_autoclose)
2279 sp->autoclose = net->sctp.max_autoclose;
2284 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2286 * Applications can enable or disable heartbeats for any peer address of
2287 * an association, modify an address's heartbeat interval, force a
2288 * heartbeat to be sent immediately, and adjust the address's maximum
2289 * number of retransmissions sent before an address is considered
2290 * unreachable. The following structure is used to access and modify an
2291 * address's parameters:
2293 * struct sctp_paddrparams {
2294 * sctp_assoc_t spp_assoc_id;
2295 * struct sockaddr_storage spp_address;
2296 * uint32_t spp_hbinterval;
2297 * uint16_t spp_pathmaxrxt;
2298 * uint32_t spp_pathmtu;
2299 * uint32_t spp_sackdelay;
2300 * uint32_t spp_flags;
2303 * spp_assoc_id - (one-to-many style socket) This is filled in the
2304 * application, and identifies the association for
2306 * spp_address - This specifies which address is of interest.
2307 * spp_hbinterval - This contains the value of the heartbeat interval,
2308 * in milliseconds. If a value of zero
2309 * is present in this field then no changes are to
2310 * be made to this parameter.
2311 * spp_pathmaxrxt - This contains the maximum number of
2312 * retransmissions before this address shall be
2313 * considered unreachable. If a value of zero
2314 * is present in this field then no changes are to
2315 * be made to this parameter.
2316 * spp_pathmtu - When Path MTU discovery is disabled the value
2317 * specified here will be the "fixed" path mtu.
2318 * Note that if the spp_address field is empty
2319 * then all associations on this address will
2320 * have this fixed path mtu set upon them.
2322 * spp_sackdelay - When delayed sack is enabled, this value specifies
2323 * the number of milliseconds that sacks will be delayed
2324 * for. This value will apply to all addresses of an
2325 * association if the spp_address field is empty. Note
2326 * also, that if delayed sack is enabled and this
2327 * value is set to 0, no change is made to the last
2328 * recorded delayed sack timer value.
2330 * spp_flags - These flags are used to control various features
2331 * on an association. The flag field may contain
2332 * zero or more of the following options.
2334 * SPP_HB_ENABLE - Enable heartbeats on the
2335 * specified address. Note that if the address
2336 * field is empty all addresses for the association
2337 * have heartbeats enabled upon them.
2339 * SPP_HB_DISABLE - Disable heartbeats on the
2340 * speicifed address. Note that if the address
2341 * field is empty all addresses for the association
2342 * will have their heartbeats disabled. Note also
2343 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2344 * mutually exclusive, only one of these two should
2345 * be specified. Enabling both fields will have
2346 * undetermined results.
2348 * SPP_HB_DEMAND - Request a user initiated heartbeat
2349 * to be made immediately.
2351 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2352 * heartbeat delayis to be set to the value of 0
2355 * SPP_PMTUD_ENABLE - This field will enable PMTU
2356 * discovery upon the specified address. Note that
2357 * if the address feild is empty then all addresses
2358 * on the association are effected.
2360 * SPP_PMTUD_DISABLE - This field will disable PMTU
2361 * discovery upon the specified address. Note that
2362 * if the address feild is empty then all addresses
2363 * on the association are effected. Not also that
2364 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2365 * exclusive. Enabling both will have undetermined
2368 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2369 * on delayed sack. The time specified in spp_sackdelay
2370 * is used to specify the sack delay for this address. Note
2371 * that if spp_address is empty then all addresses will
2372 * enable delayed sack and take on the sack delay
2373 * value specified in spp_sackdelay.
2374 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2375 * off delayed sack. If the spp_address field is blank then
2376 * delayed sack is disabled for the entire association. Note
2377 * also that this field is mutually exclusive to
2378 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2381 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2382 struct sctp_transport *trans,
2383 struct sctp_association *asoc,
2384 struct sctp_sock *sp,
2387 int sackdelay_change)
2391 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2392 struct net *net = sock_net(trans->asoc->base.sk);
2394 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2399 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2400 * this field is ignored. Note also that a value of zero indicates
2401 * the current setting should be left unchanged.
2403 if (params->spp_flags & SPP_HB_ENABLE) {
2405 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2406 * set. This lets us use 0 value when this flag
2409 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2410 params->spp_hbinterval = 0;
2412 if (params->spp_hbinterval ||
2413 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2416 msecs_to_jiffies(params->spp_hbinterval);
2419 msecs_to_jiffies(params->spp_hbinterval);
2421 sp->hbinterval = params->spp_hbinterval;
2428 trans->param_flags =
2429 (trans->param_flags & ~SPP_HB) | hb_change;
2432 (asoc->param_flags & ~SPP_HB) | hb_change;
2435 (sp->param_flags & ~SPP_HB) | hb_change;
2439 /* When Path MTU discovery is disabled the value specified here will
2440 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2441 * include the flag SPP_PMTUD_DISABLE for this field to have any
2444 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2446 trans->pathmtu = params->spp_pathmtu;
2447 sctp_assoc_sync_pmtu(asoc);
2449 asoc->pathmtu = params->spp_pathmtu;
2451 sp->pathmtu = params->spp_pathmtu;
2457 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2458 (params->spp_flags & SPP_PMTUD_ENABLE);
2459 trans->param_flags =
2460 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2462 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2463 sctp_assoc_sync_pmtu(asoc);
2467 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2470 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2474 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2475 * value of this field is ignored. Note also that a value of zero
2476 * indicates the current setting should be left unchanged.
2478 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2481 msecs_to_jiffies(params->spp_sackdelay);
2484 msecs_to_jiffies(params->spp_sackdelay);
2486 sp->sackdelay = params->spp_sackdelay;
2490 if (sackdelay_change) {
2492 trans->param_flags =
2493 (trans->param_flags & ~SPP_SACKDELAY) |
2497 (asoc->param_flags & ~SPP_SACKDELAY) |
2501 (sp->param_flags & ~SPP_SACKDELAY) |
2506 /* Note that a value of zero indicates the current setting should be
2509 if (params->spp_pathmaxrxt) {
2511 trans->pathmaxrxt = params->spp_pathmaxrxt;
2513 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2515 sp->pathmaxrxt = params->spp_pathmaxrxt;
2522 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2523 char __user *optval,
2524 unsigned int optlen)
2526 struct sctp_paddrparams params;
2527 struct sctp_transport *trans = NULL;
2528 struct sctp_association *asoc = NULL;
2529 struct sctp_sock *sp = sctp_sk(sk);
2531 int hb_change, pmtud_change, sackdelay_change;
2533 if (optlen != sizeof(struct sctp_paddrparams))
2536 if (copy_from_user(¶ms, optval, optlen))
2539 /* Validate flags and value parameters. */
2540 hb_change = params.spp_flags & SPP_HB;
2541 pmtud_change = params.spp_flags & SPP_PMTUD;
2542 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2544 if (hb_change == SPP_HB ||
2545 pmtud_change == SPP_PMTUD ||
2546 sackdelay_change == SPP_SACKDELAY ||
2547 params.spp_sackdelay > 500 ||
2548 (params.spp_pathmtu &&
2549 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2552 /* If an address other than INADDR_ANY is specified, and
2553 * no transport is found, then the request is invalid.
2555 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
2556 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2557 params.spp_assoc_id);
2562 /* Get association, if assoc_id != 0 and the socket is a one
2563 * to many style socket, and an association was not found, then
2564 * the id was invalid.
2566 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2567 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2570 /* Heartbeat demand can only be sent on a transport or
2571 * association, but not a socket.
2573 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2576 /* Process parameters. */
2577 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2578 hb_change, pmtud_change,
2584 /* If changes are for association, also apply parameters to each
2587 if (!trans && asoc) {
2588 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2590 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2591 hb_change, pmtud_change,
2599 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2601 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2604 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2606 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2610 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2612 * This option will effect the way delayed acks are performed. This
2613 * option allows you to get or set the delayed ack time, in
2614 * milliseconds. It also allows changing the delayed ack frequency.
2615 * Changing the frequency to 1 disables the delayed sack algorithm. If
2616 * the assoc_id is 0, then this sets or gets the endpoints default
2617 * values. If the assoc_id field is non-zero, then the set or get
2618 * effects the specified association for the one to many model (the
2619 * assoc_id field is ignored by the one to one model). Note that if
2620 * sack_delay or sack_freq are 0 when setting this option, then the
2621 * current values will remain unchanged.
2623 * struct sctp_sack_info {
2624 * sctp_assoc_t sack_assoc_id;
2625 * uint32_t sack_delay;
2626 * uint32_t sack_freq;
2629 * sack_assoc_id - This parameter, indicates which association the user
2630 * is performing an action upon. Note that if this field's value is
2631 * zero then the endpoints default value is changed (effecting future
2632 * associations only).
2634 * sack_delay - This parameter contains the number of milliseconds that
2635 * the user is requesting the delayed ACK timer be set to. Note that
2636 * this value is defined in the standard to be between 200 and 500
2639 * sack_freq - This parameter contains the number of packets that must
2640 * be received before a sack is sent without waiting for the delay
2641 * timer to expire. The default value for this is 2, setting this
2642 * value to 1 will disable the delayed sack algorithm.
2645 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2646 char __user *optval, unsigned int optlen)
2648 struct sctp_sack_info params;
2649 struct sctp_transport *trans = NULL;
2650 struct sctp_association *asoc = NULL;
2651 struct sctp_sock *sp = sctp_sk(sk);
2653 if (optlen == sizeof(struct sctp_sack_info)) {
2654 if (copy_from_user(¶ms, optval, optlen))
2657 if (params.sack_delay == 0 && params.sack_freq == 0)
2659 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2660 pr_warn_ratelimited(DEPRECATED
2662 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2663 "Use struct sctp_sack_info instead\n",
2664 current->comm, task_pid_nr(current));
2665 if (copy_from_user(¶ms, optval, optlen))
2668 if (params.sack_delay == 0)
2669 params.sack_freq = 1;
2671 params.sack_freq = 0;
2675 /* Validate value parameter. */
2676 if (params.sack_delay > 500)
2679 /* Get association, if sack_assoc_id != 0 and the socket is a one
2680 * to many style socket, and an association was not found, then
2681 * the id was invalid.
2683 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2684 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2687 if (params.sack_delay) {
2690 msecs_to_jiffies(params.sack_delay);
2692 sctp_spp_sackdelay_enable(asoc->param_flags);
2694 sp->sackdelay = params.sack_delay;
2696 sctp_spp_sackdelay_enable(sp->param_flags);
2700 if (params.sack_freq == 1) {
2703 sctp_spp_sackdelay_disable(asoc->param_flags);
2706 sctp_spp_sackdelay_disable(sp->param_flags);
2708 } else if (params.sack_freq > 1) {
2710 asoc->sackfreq = params.sack_freq;
2712 sctp_spp_sackdelay_enable(asoc->param_flags);
2714 sp->sackfreq = params.sack_freq;
2716 sctp_spp_sackdelay_enable(sp->param_flags);
2720 /* If change is for association, also apply to each transport. */
2722 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2724 if (params.sack_delay) {
2726 msecs_to_jiffies(params.sack_delay);
2727 trans->param_flags =
2728 sctp_spp_sackdelay_enable(trans->param_flags);
2730 if (params.sack_freq == 1) {
2731 trans->param_flags =
2732 sctp_spp_sackdelay_disable(trans->param_flags);
2733 } else if (params.sack_freq > 1) {
2734 trans->sackfreq = params.sack_freq;
2735 trans->param_flags =
2736 sctp_spp_sackdelay_enable(trans->param_flags);
2744 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2746 * Applications can specify protocol parameters for the default association
2747 * initialization. The option name argument to setsockopt() and getsockopt()
2750 * Setting initialization parameters is effective only on an unconnected
2751 * socket (for UDP-style sockets only future associations are effected
2752 * by the change). With TCP-style sockets, this option is inherited by
2753 * sockets derived from a listener socket.
2755 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2757 struct sctp_initmsg sinit;
2758 struct sctp_sock *sp = sctp_sk(sk);
2760 if (optlen != sizeof(struct sctp_initmsg))
2762 if (copy_from_user(&sinit, optval, optlen))
2765 if (sinit.sinit_num_ostreams)
2766 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2767 if (sinit.sinit_max_instreams)
2768 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2769 if (sinit.sinit_max_attempts)
2770 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2771 if (sinit.sinit_max_init_timeo)
2772 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2778 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2780 * Applications that wish to use the sendto() system call may wish to
2781 * specify a default set of parameters that would normally be supplied
2782 * through the inclusion of ancillary data. This socket option allows
2783 * such an application to set the default sctp_sndrcvinfo structure.
2784 * The application that wishes to use this socket option simply passes
2785 * in to this call the sctp_sndrcvinfo structure defined in Section
2786 * 5.2.2) The input parameters accepted by this call include
2787 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2788 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2789 * to this call if the caller is using the UDP model.
2791 static int sctp_setsockopt_default_send_param(struct sock *sk,
2792 char __user *optval,
2793 unsigned int optlen)
2795 struct sctp_sock *sp = sctp_sk(sk);
2796 struct sctp_association *asoc;
2797 struct sctp_sndrcvinfo info;
2799 if (optlen != sizeof(info))
2801 if (copy_from_user(&info, optval, optlen))
2803 if (info.sinfo_flags &
2804 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2805 SCTP_ABORT | SCTP_EOF))
2808 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2809 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2812 asoc->default_stream = info.sinfo_stream;
2813 asoc->default_flags = info.sinfo_flags;
2814 asoc->default_ppid = info.sinfo_ppid;
2815 asoc->default_context = info.sinfo_context;
2816 asoc->default_timetolive = info.sinfo_timetolive;
2818 sp->default_stream = info.sinfo_stream;
2819 sp->default_flags = info.sinfo_flags;
2820 sp->default_ppid = info.sinfo_ppid;
2821 sp->default_context = info.sinfo_context;
2822 sp->default_timetolive = info.sinfo_timetolive;
2828 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2829 * (SCTP_DEFAULT_SNDINFO)
2831 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
2832 char __user *optval,
2833 unsigned int optlen)
2835 struct sctp_sock *sp = sctp_sk(sk);
2836 struct sctp_association *asoc;
2837 struct sctp_sndinfo info;
2839 if (optlen != sizeof(info))
2841 if (copy_from_user(&info, optval, optlen))
2843 if (info.snd_flags &
2844 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2845 SCTP_ABORT | SCTP_EOF))
2848 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
2849 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
2852 asoc->default_stream = info.snd_sid;
2853 asoc->default_flags = info.snd_flags;
2854 asoc->default_ppid = info.snd_ppid;
2855 asoc->default_context = info.snd_context;
2857 sp->default_stream = info.snd_sid;
2858 sp->default_flags = info.snd_flags;
2859 sp->default_ppid = info.snd_ppid;
2860 sp->default_context = info.snd_context;
2866 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2868 * Requests that the local SCTP stack use the enclosed peer address as
2869 * the association primary. The enclosed address must be one of the
2870 * association peer's addresses.
2872 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2873 unsigned int optlen)
2875 struct sctp_prim prim;
2876 struct sctp_transport *trans;
2878 if (optlen != sizeof(struct sctp_prim))
2881 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2884 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2888 sctp_assoc_set_primary(trans->asoc, trans);
2894 * 7.1.5 SCTP_NODELAY
2896 * Turn on/off any Nagle-like algorithm. This means that packets are
2897 * generally sent as soon as possible and no unnecessary delays are
2898 * introduced, at the cost of more packets in the network. Expects an
2899 * integer boolean flag.
2901 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2902 unsigned int optlen)
2906 if (optlen < sizeof(int))
2908 if (get_user(val, (int __user *)optval))
2911 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2917 * 7.1.1 SCTP_RTOINFO
2919 * The protocol parameters used to initialize and bound retransmission
2920 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2921 * and modify these parameters.
2922 * All parameters are time values, in milliseconds. A value of 0, when
2923 * modifying the parameters, indicates that the current value should not
2927 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2929 struct sctp_rtoinfo rtoinfo;
2930 struct sctp_association *asoc;
2931 unsigned long rto_min, rto_max;
2932 struct sctp_sock *sp = sctp_sk(sk);
2934 if (optlen != sizeof (struct sctp_rtoinfo))
2937 if (copy_from_user(&rtoinfo, optval, optlen))
2940 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2942 /* Set the values to the specific association */
2943 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2946 rto_max = rtoinfo.srto_max;
2947 rto_min = rtoinfo.srto_min;
2950 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2952 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2955 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2957 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2959 if (rto_min > rto_max)
2963 if (rtoinfo.srto_initial != 0)
2965 msecs_to_jiffies(rtoinfo.srto_initial);
2966 asoc->rto_max = rto_max;
2967 asoc->rto_min = rto_min;
2969 /* If there is no association or the association-id = 0
2970 * set the values to the endpoint.
2972 if (rtoinfo.srto_initial != 0)
2973 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2974 sp->rtoinfo.srto_max = rto_max;
2975 sp->rtoinfo.srto_min = rto_min;
2983 * 7.1.2 SCTP_ASSOCINFO
2985 * This option is used to tune the maximum retransmission attempts
2986 * of the association.
2987 * Returns an error if the new association retransmission value is
2988 * greater than the sum of the retransmission value of the peer.
2989 * See [SCTP] for more information.
2992 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2995 struct sctp_assocparams assocparams;
2996 struct sctp_association *asoc;
2998 if (optlen != sizeof(struct sctp_assocparams))
3000 if (copy_from_user(&assocparams, optval, optlen))
3003 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
3005 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
3008 /* Set the values to the specific association */
3010 if (assocparams.sasoc_asocmaxrxt != 0) {
3013 struct sctp_transport *peer_addr;
3015 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
3017 path_sum += peer_addr->pathmaxrxt;
3021 /* Only validate asocmaxrxt if we have more than
3022 * one path/transport. We do this because path
3023 * retransmissions are only counted when we have more
3027 assocparams.sasoc_asocmaxrxt > path_sum)
3030 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3033 if (assocparams.sasoc_cookie_life != 0)
3034 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3036 /* Set the values to the endpoint */
3037 struct sctp_sock *sp = sctp_sk(sk);
3039 if (assocparams.sasoc_asocmaxrxt != 0)
3040 sp->assocparams.sasoc_asocmaxrxt =
3041 assocparams.sasoc_asocmaxrxt;
3042 if (assocparams.sasoc_cookie_life != 0)
3043 sp->assocparams.sasoc_cookie_life =
3044 assocparams.sasoc_cookie_life;
3050 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3052 * This socket option is a boolean flag which turns on or off mapped V4
3053 * addresses. If this option is turned on and the socket is type
3054 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3055 * If this option is turned off, then no mapping will be done of V4
3056 * addresses and a user will receive both PF_INET6 and PF_INET type
3057 * addresses on the socket.
3059 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3062 struct sctp_sock *sp = sctp_sk(sk);
3064 if (optlen < sizeof(int))
3066 if (get_user(val, (int __user *)optval))
3077 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3078 * This option will get or set the maximum size to put in any outgoing
3079 * SCTP DATA chunk. If a message is larger than this size it will be
3080 * fragmented by SCTP into the specified size. Note that the underlying
3081 * SCTP implementation may fragment into smaller sized chunks when the
3082 * PMTU of the underlying association is smaller than the value set by
3083 * the user. The default value for this option is '0' which indicates
3084 * the user is NOT limiting fragmentation and only the PMTU will effect
3085 * SCTP's choice of DATA chunk size. Note also that values set larger
3086 * than the maximum size of an IP datagram will effectively let SCTP
3087 * control fragmentation (i.e. the same as setting this option to 0).
3089 * The following structure is used to access and modify this parameter:
3091 * struct sctp_assoc_value {
3092 * sctp_assoc_t assoc_id;
3093 * uint32_t assoc_value;
3096 * assoc_id: This parameter is ignored for one-to-one style sockets.
3097 * For one-to-many style sockets this parameter indicates which
3098 * association the user is performing an action upon. Note that if
3099 * this field's value is zero then the endpoints default value is
3100 * changed (effecting future associations only).
3101 * assoc_value: This parameter specifies the maximum size in bytes.
3103 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3105 struct sctp_assoc_value params;
3106 struct sctp_association *asoc;
3107 struct sctp_sock *sp = sctp_sk(sk);
3110 if (optlen == sizeof(int)) {
3111 pr_warn_ratelimited(DEPRECATED
3113 "Use of int in maxseg socket option.\n"
3114 "Use struct sctp_assoc_value instead\n",
3115 current->comm, task_pid_nr(current));
3116 if (copy_from_user(&val, optval, optlen))
3118 params.assoc_id = 0;
3119 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3120 if (copy_from_user(¶ms, optval, optlen))
3122 val = params.assoc_value;
3126 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3129 asoc = sctp_id2assoc(sk, params.assoc_id);
3130 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3135 val = asoc->pathmtu;
3136 val -= sp->pf->af->net_header_len;
3137 val -= sizeof(struct sctphdr) +
3138 sizeof(struct sctp_data_chunk);
3140 asoc->user_frag = val;
3141 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3143 sp->user_frag = val;
3151 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3153 * Requests that the peer mark the enclosed address as the association
3154 * primary. The enclosed address must be one of the association's
3155 * locally bound addresses. The following structure is used to make a
3156 * set primary request:
3158 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3159 unsigned int optlen)
3161 struct net *net = sock_net(sk);
3162 struct sctp_sock *sp;
3163 struct sctp_association *asoc = NULL;
3164 struct sctp_setpeerprim prim;
3165 struct sctp_chunk *chunk;
3171 if (!net->sctp.addip_enable)
3174 if (optlen != sizeof(struct sctp_setpeerprim))
3177 if (copy_from_user(&prim, optval, optlen))
3180 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3184 if (!asoc->peer.asconf_capable)
3187 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3190 if (!sctp_state(asoc, ESTABLISHED))
3193 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3197 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3198 return -EADDRNOTAVAIL;
3200 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3201 return -EADDRNOTAVAIL;
3203 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3204 chunk = sctp_make_asconf_set_prim(asoc,
3205 (union sctp_addr *)&prim.sspp_addr);
3209 err = sctp_send_asconf(asoc, chunk);
3211 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3216 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3217 unsigned int optlen)
3219 struct sctp_setadaptation adaptation;
3221 if (optlen != sizeof(struct sctp_setadaptation))
3223 if (copy_from_user(&adaptation, optval, optlen))
3226 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3232 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3234 * The context field in the sctp_sndrcvinfo structure is normally only
3235 * used when a failed message is retrieved holding the value that was
3236 * sent down on the actual send call. This option allows the setting of
3237 * a default context on an association basis that will be received on
3238 * reading messages from the peer. This is especially helpful in the
3239 * one-2-many model for an application to keep some reference to an
3240 * internal state machine that is processing messages on the
3241 * association. Note that the setting of this value only effects
3242 * received messages from the peer and does not effect the value that is
3243 * saved with outbound messages.
3245 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3246 unsigned int optlen)
3248 struct sctp_assoc_value params;
3249 struct sctp_sock *sp;
3250 struct sctp_association *asoc;
3252 if (optlen != sizeof(struct sctp_assoc_value))
3254 if (copy_from_user(¶ms, optval, optlen))
3259 if (params.assoc_id != 0) {
3260 asoc = sctp_id2assoc(sk, params.assoc_id);
3263 asoc->default_rcv_context = params.assoc_value;
3265 sp->default_rcv_context = params.assoc_value;
3272 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3274 * This options will at a minimum specify if the implementation is doing
3275 * fragmented interleave. Fragmented interleave, for a one to many
3276 * socket, is when subsequent calls to receive a message may return
3277 * parts of messages from different associations. Some implementations
3278 * may allow you to turn this value on or off. If so, when turned off,
3279 * no fragment interleave will occur (which will cause a head of line
3280 * blocking amongst multiple associations sharing the same one to many
3281 * socket). When this option is turned on, then each receive call may
3282 * come from a different association (thus the user must receive data
3283 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3284 * association each receive belongs to.
3286 * This option takes a boolean value. A non-zero value indicates that
3287 * fragmented interleave is on. A value of zero indicates that
3288 * fragmented interleave is off.
3290 * Note that it is important that an implementation that allows this
3291 * option to be turned on, have it off by default. Otherwise an unaware
3292 * application using the one to many model may become confused and act
3295 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3296 char __user *optval,
3297 unsigned int optlen)
3301 if (optlen != sizeof(int))
3303 if (get_user(val, (int __user *)optval))
3306 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3312 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3313 * (SCTP_PARTIAL_DELIVERY_POINT)
3315 * This option will set or get the SCTP partial delivery point. This
3316 * point is the size of a message where the partial delivery API will be
3317 * invoked to help free up rwnd space for the peer. Setting this to a
3318 * lower value will cause partial deliveries to happen more often. The
3319 * calls argument is an integer that sets or gets the partial delivery
3320 * point. Note also that the call will fail if the user attempts to set
3321 * this value larger than the socket receive buffer size.
3323 * Note that any single message having a length smaller than or equal to
3324 * the SCTP partial delivery point will be delivered in one single read
3325 * call as long as the user provided buffer is large enough to hold the
3328 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3329 char __user *optval,
3330 unsigned int optlen)
3334 if (optlen != sizeof(u32))
3336 if (get_user(val, (int __user *)optval))
3339 /* Note: We double the receive buffer from what the user sets
3340 * it to be, also initial rwnd is based on rcvbuf/2.
3342 if (val > (sk->sk_rcvbuf >> 1))
3345 sctp_sk(sk)->pd_point = val;
3347 return 0; /* is this the right error code? */
3351 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3353 * This option will allow a user to change the maximum burst of packets
3354 * that can be emitted by this association. Note that the default value
3355 * is 4, and some implementations may restrict this setting so that it
3356 * can only be lowered.
3358 * NOTE: This text doesn't seem right. Do this on a socket basis with
3359 * future associations inheriting the socket value.
3361 static int sctp_setsockopt_maxburst(struct sock *sk,
3362 char __user *optval,
3363 unsigned int optlen)
3365 struct sctp_assoc_value params;
3366 struct sctp_sock *sp;
3367 struct sctp_association *asoc;
3371 if (optlen == sizeof(int)) {
3372 pr_warn_ratelimited(DEPRECATED
3374 "Use of int in max_burst socket option deprecated.\n"
3375 "Use struct sctp_assoc_value instead\n",
3376 current->comm, task_pid_nr(current));
3377 if (copy_from_user(&val, optval, optlen))
3379 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3380 if (copy_from_user(¶ms, optval, optlen))
3382 val = params.assoc_value;
3383 assoc_id = params.assoc_id;
3389 if (assoc_id != 0) {
3390 asoc = sctp_id2assoc(sk, assoc_id);
3393 asoc->max_burst = val;
3395 sp->max_burst = val;
3401 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3403 * This set option adds a chunk type that the user is requesting to be
3404 * received only in an authenticated way. Changes to the list of chunks
3405 * will only effect future associations on the socket.
3407 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3408 char __user *optval,
3409 unsigned int optlen)
3411 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3412 struct sctp_authchunk val;
3414 if (!ep->auth_enable)
3417 if (optlen != sizeof(struct sctp_authchunk))
3419 if (copy_from_user(&val, optval, optlen))
3422 switch (val.sauth_chunk) {
3424 case SCTP_CID_INIT_ACK:
3425 case SCTP_CID_SHUTDOWN_COMPLETE:
3430 /* add this chunk id to the endpoint */
3431 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3435 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3437 * This option gets or sets the list of HMAC algorithms that the local
3438 * endpoint requires the peer to use.
3440 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3441 char __user *optval,
3442 unsigned int optlen)
3444 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3445 struct sctp_hmacalgo *hmacs;
3449 if (!ep->auth_enable)
3452 if (optlen < sizeof(struct sctp_hmacalgo))
3455 hmacs = memdup_user(optval, optlen);
3457 return PTR_ERR(hmacs);
3459 idents = hmacs->shmac_num_idents;
3460 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3461 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3466 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3473 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3475 * This option will set a shared secret key which is used to build an
3476 * association shared key.
3478 static int sctp_setsockopt_auth_key(struct sock *sk,
3479 char __user *optval,
3480 unsigned int optlen)
3482 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3483 struct sctp_authkey *authkey;
3484 struct sctp_association *asoc;
3487 if (!ep->auth_enable)
3490 if (optlen <= sizeof(struct sctp_authkey))
3493 authkey = memdup_user(optval, optlen);
3494 if (IS_ERR(authkey))
3495 return PTR_ERR(authkey);
3497 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3502 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3503 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3508 ret = sctp_auth_set_key(ep, asoc, authkey);
3515 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3517 * This option will get or set the active shared key to be used to build
3518 * the association shared key.
3520 static int sctp_setsockopt_active_key(struct sock *sk,
3521 char __user *optval,
3522 unsigned int optlen)
3524 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3525 struct sctp_authkeyid val;
3526 struct sctp_association *asoc;
3528 if (!ep->auth_enable)
3531 if (optlen != sizeof(struct sctp_authkeyid))
3533 if (copy_from_user(&val, optval, optlen))
3536 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3537 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3540 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3544 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3546 * This set option will delete a shared secret key from use.
3548 static int sctp_setsockopt_del_key(struct sock *sk,
3549 char __user *optval,
3550 unsigned int optlen)
3552 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3553 struct sctp_authkeyid val;
3554 struct sctp_association *asoc;
3556 if (!ep->auth_enable)
3559 if (optlen != sizeof(struct sctp_authkeyid))
3561 if (copy_from_user(&val, optval, optlen))
3564 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3565 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3568 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3573 * 8.1.23 SCTP_AUTO_ASCONF
3575 * This option will enable or disable the use of the automatic generation of
3576 * ASCONF chunks to add and delete addresses to an existing association. Note
3577 * that this option has two caveats namely: a) it only affects sockets that
3578 * are bound to all addresses available to the SCTP stack, and b) the system
3579 * administrator may have an overriding control that turns the ASCONF feature
3580 * off no matter what setting the socket option may have.
3581 * This option expects an integer boolean flag, where a non-zero value turns on
3582 * the option, and a zero value turns off the option.
3583 * Note. In this implementation, socket operation overrides default parameter
3584 * being set by sysctl as well as FreeBSD implementation
3586 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3587 unsigned int optlen)
3590 struct sctp_sock *sp = sctp_sk(sk);
3592 if (optlen < sizeof(int))
3594 if (get_user(val, (int __user *)optval))
3596 if (!sctp_is_ep_boundall(sk) && val)
3598 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3601 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3602 if (val == 0 && sp->do_auto_asconf) {
3603 list_del(&sp->auto_asconf_list);
3604 sp->do_auto_asconf = 0;
3605 } else if (val && !sp->do_auto_asconf) {
3606 list_add_tail(&sp->auto_asconf_list,
3607 &sock_net(sk)->sctp.auto_asconf_splist);
3608 sp->do_auto_asconf = 1;
3610 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3615 * SCTP_PEER_ADDR_THLDS
3617 * This option allows us to alter the partially failed threshold for one or all
3618 * transports in an association. See Section 6.1 of:
3619 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3621 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3622 char __user *optval,
3623 unsigned int optlen)
3625 struct sctp_paddrthlds val;
3626 struct sctp_transport *trans;
3627 struct sctp_association *asoc;
3629 if (optlen < sizeof(struct sctp_paddrthlds))
3631 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3632 sizeof(struct sctp_paddrthlds)))
3636 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3637 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3640 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3642 if (val.spt_pathmaxrxt)
3643 trans->pathmaxrxt = val.spt_pathmaxrxt;
3644 trans->pf_retrans = val.spt_pathpfthld;
3647 if (val.spt_pathmaxrxt)
3648 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3649 asoc->pf_retrans = val.spt_pathpfthld;
3651 trans = sctp_addr_id2transport(sk, &val.spt_address,
3656 if (val.spt_pathmaxrxt)
3657 trans->pathmaxrxt = val.spt_pathmaxrxt;
3658 trans->pf_retrans = val.spt_pathpfthld;
3664 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3665 char __user *optval,
3666 unsigned int optlen)
3670 if (optlen < sizeof(int))
3672 if (get_user(val, (int __user *) optval))
3675 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3680 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3681 char __user *optval,
3682 unsigned int optlen)
3686 if (optlen < sizeof(int))
3688 if (get_user(val, (int __user *) optval))
3691 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3696 static int sctp_setsockopt_pr_supported(struct sock *sk,
3697 char __user *optval,
3698 unsigned int optlen)
3700 struct sctp_assoc_value params;
3701 struct sctp_association *asoc;
3702 int retval = -EINVAL;
3704 if (optlen != sizeof(params))
3707 if (copy_from_user(¶ms, optval, optlen)) {
3712 asoc = sctp_id2assoc(sk, params.assoc_id);
3714 asoc->prsctp_enable = !!params.assoc_value;
3715 } else if (!params.assoc_id) {
3716 struct sctp_sock *sp = sctp_sk(sk);
3718 sp->ep->prsctp_enable = !!params.assoc_value;
3729 static int sctp_setsockopt_default_prinfo(struct sock *sk,
3730 char __user *optval,
3731 unsigned int optlen)
3733 struct sctp_default_prinfo info;
3734 struct sctp_association *asoc;
3735 int retval = -EINVAL;
3737 if (optlen != sizeof(info))
3740 if (copy_from_user(&info, optval, sizeof(info))) {
3745 if (info.pr_policy & ~SCTP_PR_SCTP_MASK)
3748 if (info.pr_policy == SCTP_PR_SCTP_NONE)
3751 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
3753 SCTP_PR_SET_POLICY(asoc->default_flags, info.pr_policy);
3754 asoc->default_timetolive = info.pr_value;
3755 } else if (!info.pr_assoc_id) {
3756 struct sctp_sock *sp = sctp_sk(sk);
3758 SCTP_PR_SET_POLICY(sp->default_flags, info.pr_policy);
3759 sp->default_timetolive = info.pr_value;
3770 static int sctp_setsockopt_reconfig_supported(struct sock *sk,
3771 char __user *optval,
3772 unsigned int optlen)
3774 struct sctp_assoc_value params;
3775 struct sctp_association *asoc;
3776 int retval = -EINVAL;
3778 if (optlen != sizeof(params))
3781 if (copy_from_user(¶ms, optval, optlen)) {
3786 asoc = sctp_id2assoc(sk, params.assoc_id);
3788 asoc->reconf_enable = !!params.assoc_value;
3789 } else if (!params.assoc_id) {
3790 struct sctp_sock *sp = sctp_sk(sk);
3792 sp->ep->reconf_enable = !!params.assoc_value;
3803 static int sctp_setsockopt_enable_strreset(struct sock *sk,
3804 char __user *optval,
3805 unsigned int optlen)
3807 struct sctp_assoc_value params;
3808 struct sctp_association *asoc;
3809 int retval = -EINVAL;
3811 if (optlen != sizeof(params))
3814 if (copy_from_user(¶ms, optval, optlen)) {
3819 if (params.assoc_value & (~SCTP_ENABLE_STRRESET_MASK))
3822 asoc = sctp_id2assoc(sk, params.assoc_id);
3824 asoc->strreset_enable = params.assoc_value;
3825 } else if (!params.assoc_id) {
3826 struct sctp_sock *sp = sctp_sk(sk);
3828 sp->ep->strreset_enable = params.assoc_value;
3839 static int sctp_setsockopt_reset_streams(struct sock *sk,
3840 char __user *optval,
3841 unsigned int optlen)
3843 struct sctp_reset_streams *params;
3844 struct sctp_association *asoc;
3845 int retval = -EINVAL;
3847 if (optlen < sizeof(struct sctp_reset_streams))
3850 params = memdup_user(optval, optlen);
3852 return PTR_ERR(params);
3854 asoc = sctp_id2assoc(sk, params->srs_assoc_id);
3858 retval = sctp_send_reset_streams(asoc, params);
3865 static int sctp_setsockopt_reset_assoc(struct sock *sk,
3866 char __user *optval,
3867 unsigned int optlen)
3869 struct sctp_association *asoc;
3870 sctp_assoc_t associd;
3871 int retval = -EINVAL;
3873 if (optlen != sizeof(associd))
3876 if (copy_from_user(&associd, optval, optlen)) {
3881 asoc = sctp_id2assoc(sk, associd);
3885 retval = sctp_send_reset_assoc(asoc);
3891 static int sctp_setsockopt_add_streams(struct sock *sk,
3892 char __user *optval,
3893 unsigned int optlen)
3895 struct sctp_association *asoc;
3896 struct sctp_add_streams params;
3897 int retval = -EINVAL;
3899 if (optlen != sizeof(params))
3902 if (copy_from_user(¶ms, optval, optlen)) {
3907 asoc = sctp_id2assoc(sk, params.sas_assoc_id);
3911 retval = sctp_send_add_streams(asoc, ¶ms);
3917 /* API 6.2 setsockopt(), getsockopt()
3919 * Applications use setsockopt() and getsockopt() to set or retrieve
3920 * socket options. Socket options are used to change the default
3921 * behavior of sockets calls. They are described in Section 7.
3925 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3926 * int __user *optlen);
3927 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3930 * sd - the socket descript.
3931 * level - set to IPPROTO_SCTP for all SCTP options.
3932 * optname - the option name.
3933 * optval - the buffer to store the value of the option.
3934 * optlen - the size of the buffer.
3936 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3937 char __user *optval, unsigned int optlen)
3941 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3943 /* I can hardly begin to describe how wrong this is. This is
3944 * so broken as to be worse than useless. The API draft
3945 * REALLY is NOT helpful here... I am not convinced that the
3946 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3947 * are at all well-founded.
3949 if (level != SOL_SCTP) {
3950 struct sctp_af *af = sctp_sk(sk)->pf->af;
3951 retval = af->setsockopt(sk, level, optname, optval, optlen);
3958 case SCTP_SOCKOPT_BINDX_ADD:
3959 /* 'optlen' is the size of the addresses buffer. */
3960 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3961 optlen, SCTP_BINDX_ADD_ADDR);
3964 case SCTP_SOCKOPT_BINDX_REM:
3965 /* 'optlen' is the size of the addresses buffer. */
3966 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3967 optlen, SCTP_BINDX_REM_ADDR);
3970 case SCTP_SOCKOPT_CONNECTX_OLD:
3971 /* 'optlen' is the size of the addresses buffer. */
3972 retval = sctp_setsockopt_connectx_old(sk,
3973 (struct sockaddr __user *)optval,
3977 case SCTP_SOCKOPT_CONNECTX:
3978 /* 'optlen' is the size of the addresses buffer. */
3979 retval = sctp_setsockopt_connectx(sk,
3980 (struct sockaddr __user *)optval,
3984 case SCTP_DISABLE_FRAGMENTS:
3985 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3989 retval = sctp_setsockopt_events(sk, optval, optlen);
3992 case SCTP_AUTOCLOSE:
3993 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3996 case SCTP_PEER_ADDR_PARAMS:
3997 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
4000 case SCTP_DELAYED_SACK:
4001 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
4003 case SCTP_PARTIAL_DELIVERY_POINT:
4004 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
4008 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
4010 case SCTP_DEFAULT_SEND_PARAM:
4011 retval = sctp_setsockopt_default_send_param(sk, optval,
4014 case SCTP_DEFAULT_SNDINFO:
4015 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
4017 case SCTP_PRIMARY_ADDR:
4018 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
4020 case SCTP_SET_PEER_PRIMARY_ADDR:
4021 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
4024 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
4027 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
4029 case SCTP_ASSOCINFO:
4030 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
4032 case SCTP_I_WANT_MAPPED_V4_ADDR:
4033 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
4036 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
4038 case SCTP_ADAPTATION_LAYER:
4039 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
4042 retval = sctp_setsockopt_context(sk, optval, optlen);
4044 case SCTP_FRAGMENT_INTERLEAVE:
4045 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
4047 case SCTP_MAX_BURST:
4048 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
4050 case SCTP_AUTH_CHUNK:
4051 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
4053 case SCTP_HMAC_IDENT:
4054 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
4057 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
4059 case SCTP_AUTH_ACTIVE_KEY:
4060 retval = sctp_setsockopt_active_key(sk, optval, optlen);
4062 case SCTP_AUTH_DELETE_KEY:
4063 retval = sctp_setsockopt_del_key(sk, optval, optlen);
4065 case SCTP_AUTO_ASCONF:
4066 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
4068 case SCTP_PEER_ADDR_THLDS:
4069 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
4071 case SCTP_RECVRCVINFO:
4072 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
4074 case SCTP_RECVNXTINFO:
4075 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
4077 case SCTP_PR_SUPPORTED:
4078 retval = sctp_setsockopt_pr_supported(sk, optval, optlen);
4080 case SCTP_DEFAULT_PRINFO:
4081 retval = sctp_setsockopt_default_prinfo(sk, optval, optlen);
4083 case SCTP_RECONFIG_SUPPORTED:
4084 retval = sctp_setsockopt_reconfig_supported(sk, optval, optlen);
4086 case SCTP_ENABLE_STREAM_RESET:
4087 retval = sctp_setsockopt_enable_strreset(sk, optval, optlen);
4089 case SCTP_RESET_STREAMS:
4090 retval = sctp_setsockopt_reset_streams(sk, optval, optlen);
4092 case SCTP_RESET_ASSOC:
4093 retval = sctp_setsockopt_reset_assoc(sk, optval, optlen);
4095 case SCTP_ADD_STREAMS:
4096 retval = sctp_setsockopt_add_streams(sk, optval, optlen);
4099 retval = -ENOPROTOOPT;
4109 /* API 3.1.6 connect() - UDP Style Syntax
4111 * An application may use the connect() call in the UDP model to initiate an
4112 * association without sending data.
4116 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
4118 * sd: the socket descriptor to have a new association added to.
4120 * nam: the address structure (either struct sockaddr_in or struct
4121 * sockaddr_in6 defined in RFC2553 [7]).
4123 * len: the size of the address.
4125 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
4133 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
4136 /* Validate addr_len before calling common connect/connectx routine. */
4137 af = sctp_get_af_specific(addr->sa_family);
4138 if (!af || addr_len < af->sockaddr_len) {
4141 /* Pass correct addr len to common routine (so it knows there
4142 * is only one address being passed.
4144 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
4151 /* FIXME: Write comments. */
4152 static int sctp_disconnect(struct sock *sk, int flags)
4154 return -EOPNOTSUPP; /* STUB */
4157 /* 4.1.4 accept() - TCP Style Syntax
4159 * Applications use accept() call to remove an established SCTP
4160 * association from the accept queue of the endpoint. A new socket
4161 * descriptor will be returned from accept() to represent the newly
4162 * formed association.
4164 static struct sock *sctp_accept(struct sock *sk, int flags, int *err, bool kern)
4166 struct sctp_sock *sp;
4167 struct sctp_endpoint *ep;
4168 struct sock *newsk = NULL;
4169 struct sctp_association *asoc;
4178 if (!sctp_style(sk, TCP)) {
4179 error = -EOPNOTSUPP;
4183 if (!sctp_sstate(sk, LISTENING)) {
4188 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
4190 error = sctp_wait_for_accept(sk, timeo);
4194 /* We treat the list of associations on the endpoint as the accept
4195 * queue and pick the first association on the list.
4197 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
4199 newsk = sp->pf->create_accept_sk(sk, asoc, kern);
4205 /* Populate the fields of the newsk from the oldsk and migrate the
4206 * asoc to the newsk.
4208 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
4216 /* The SCTP ioctl handler. */
4217 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
4224 * SEQPACKET-style sockets in LISTENING state are valid, for
4225 * SCTP, so only discard TCP-style sockets in LISTENING state.
4227 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4232 struct sk_buff *skb;
4233 unsigned int amount = 0;
4235 skb = skb_peek(&sk->sk_receive_queue);
4238 * We will only return the amount of this packet since
4239 * that is all that will be read.
4243 rc = put_user(amount, (int __user *)arg);
4255 /* This is the function which gets called during socket creation to
4256 * initialized the SCTP-specific portion of the sock.
4257 * The sock structure should already be zero-filled memory.
4259 static int sctp_init_sock(struct sock *sk)
4261 struct net *net = sock_net(sk);
4262 struct sctp_sock *sp;
4264 pr_debug("%s: sk:%p\n", __func__, sk);
4268 /* Initialize the SCTP per socket area. */
4269 switch (sk->sk_type) {
4270 case SOCK_SEQPACKET:
4271 sp->type = SCTP_SOCKET_UDP;
4274 sp->type = SCTP_SOCKET_TCP;
4277 return -ESOCKTNOSUPPORT;
4280 sk->sk_gso_type = SKB_GSO_SCTP;
4282 /* Initialize default send parameters. These parameters can be
4283 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4285 sp->default_stream = 0;
4286 sp->default_ppid = 0;
4287 sp->default_flags = 0;
4288 sp->default_context = 0;
4289 sp->default_timetolive = 0;
4291 sp->default_rcv_context = 0;
4292 sp->max_burst = net->sctp.max_burst;
4294 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4296 /* Initialize default setup parameters. These parameters
4297 * can be modified with the SCTP_INITMSG socket option or
4298 * overridden by the SCTP_INIT CMSG.
4300 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4301 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4302 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4303 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4305 /* Initialize default RTO related parameters. These parameters can
4306 * be modified for with the SCTP_RTOINFO socket option.
4308 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4309 sp->rtoinfo.srto_max = net->sctp.rto_max;
4310 sp->rtoinfo.srto_min = net->sctp.rto_min;
4312 /* Initialize default association related parameters. These parameters
4313 * can be modified with the SCTP_ASSOCINFO socket option.
4315 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4316 sp->assocparams.sasoc_number_peer_destinations = 0;
4317 sp->assocparams.sasoc_peer_rwnd = 0;
4318 sp->assocparams.sasoc_local_rwnd = 0;
4319 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4321 /* Initialize default event subscriptions. By default, all the
4324 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4326 /* Default Peer Address Parameters. These defaults can
4327 * be modified via SCTP_PEER_ADDR_PARAMS
4329 sp->hbinterval = net->sctp.hb_interval;
4330 sp->pathmaxrxt = net->sctp.max_retrans_path;
4331 sp->pathmtu = 0; /* allow default discovery */
4332 sp->sackdelay = net->sctp.sack_timeout;
4334 sp->param_flags = SPP_HB_ENABLE |
4336 SPP_SACKDELAY_ENABLE;
4338 /* If enabled no SCTP message fragmentation will be performed.
4339 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4341 sp->disable_fragments = 0;
4343 /* Enable Nagle algorithm by default. */
4346 sp->recvrcvinfo = 0;
4347 sp->recvnxtinfo = 0;
4349 /* Enable by default. */
4352 /* Auto-close idle associations after the configured
4353 * number of seconds. A value of 0 disables this
4354 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4355 * for UDP-style sockets only.
4359 /* User specified fragmentation limit. */
4362 sp->adaptation_ind = 0;
4364 sp->pf = sctp_get_pf_specific(sk->sk_family);
4366 /* Control variables for partial data delivery. */
4367 atomic_set(&sp->pd_mode, 0);
4368 skb_queue_head_init(&sp->pd_lobby);
4369 sp->frag_interleave = 0;
4371 /* Create a per socket endpoint structure. Even if we
4372 * change the data structure relationships, this may still
4373 * be useful for storing pre-connect address information.
4375 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4381 sk->sk_destruct = sctp_destruct_sock;
4383 SCTP_DBG_OBJCNT_INC(sock);
4386 percpu_counter_inc(&sctp_sockets_allocated);
4387 sock_prot_inuse_add(net, sk->sk_prot, 1);
4389 /* Nothing can fail after this block, otherwise
4390 * sctp_destroy_sock() will be called without addr_wq_lock held
4392 if (net->sctp.default_auto_asconf) {
4393 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
4394 list_add_tail(&sp->auto_asconf_list,
4395 &net->sctp.auto_asconf_splist);
4396 sp->do_auto_asconf = 1;
4397 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4399 sp->do_auto_asconf = 0;
4407 /* Cleanup any SCTP per socket resources. Must be called with
4408 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4410 static void sctp_destroy_sock(struct sock *sk)
4412 struct sctp_sock *sp;
4414 pr_debug("%s: sk:%p\n", __func__, sk);
4416 /* Release our hold on the endpoint. */
4418 /* This could happen during socket init, thus we bail out
4419 * early, since the rest of the below is not setup either.
4424 if (sp->do_auto_asconf) {
4425 sp->do_auto_asconf = 0;
4426 list_del(&sp->auto_asconf_list);
4428 sctp_endpoint_free(sp->ep);
4430 percpu_counter_dec(&sctp_sockets_allocated);
4431 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4435 /* Triggered when there are no references on the socket anymore */
4436 static void sctp_destruct_sock(struct sock *sk)
4438 struct sctp_sock *sp = sctp_sk(sk);
4440 /* Free up the HMAC transform. */
4441 crypto_free_shash(sp->hmac);
4443 inet_sock_destruct(sk);
4446 /* API 4.1.7 shutdown() - TCP Style Syntax
4447 * int shutdown(int socket, int how);
4449 * sd - the socket descriptor of the association to be closed.
4450 * how - Specifies the type of shutdown. The values are
4453 * Disables further receive operations. No SCTP
4454 * protocol action is taken.
4456 * Disables further send operations, and initiates
4457 * the SCTP shutdown sequence.
4459 * Disables further send and receive operations
4460 * and initiates the SCTP shutdown sequence.
4462 static void sctp_shutdown(struct sock *sk, int how)
4464 struct net *net = sock_net(sk);
4465 struct sctp_endpoint *ep;
4467 if (!sctp_style(sk, TCP))
4470 ep = sctp_sk(sk)->ep;
4471 if (how & SEND_SHUTDOWN && !list_empty(&ep->asocs)) {
4472 struct sctp_association *asoc;
4474 sk->sk_state = SCTP_SS_CLOSING;
4475 asoc = list_entry(ep->asocs.next,
4476 struct sctp_association, asocs);
4477 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4481 int sctp_get_sctp_info(struct sock *sk, struct sctp_association *asoc,
4482 struct sctp_info *info)
4484 struct sctp_transport *prim;
4485 struct list_head *pos;
4488 memset(info, 0, sizeof(*info));
4490 struct sctp_sock *sp = sctp_sk(sk);
4492 info->sctpi_s_autoclose = sp->autoclose;
4493 info->sctpi_s_adaptation_ind = sp->adaptation_ind;
4494 info->sctpi_s_pd_point = sp->pd_point;
4495 info->sctpi_s_nodelay = sp->nodelay;
4496 info->sctpi_s_disable_fragments = sp->disable_fragments;
4497 info->sctpi_s_v4mapped = sp->v4mapped;
4498 info->sctpi_s_frag_interleave = sp->frag_interleave;
4499 info->sctpi_s_type = sp->type;
4504 info->sctpi_tag = asoc->c.my_vtag;
4505 info->sctpi_state = asoc->state;
4506 info->sctpi_rwnd = asoc->a_rwnd;
4507 info->sctpi_unackdata = asoc->unack_data;
4508 info->sctpi_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4509 info->sctpi_instrms = asoc->stream.incnt;
4510 info->sctpi_outstrms = asoc->stream.outcnt;
4511 list_for_each(pos, &asoc->base.inqueue.in_chunk_list)
4512 info->sctpi_inqueue++;
4513 list_for_each(pos, &asoc->outqueue.out_chunk_list)
4514 info->sctpi_outqueue++;
4515 info->sctpi_overall_error = asoc->overall_error_count;
4516 info->sctpi_max_burst = asoc->max_burst;
4517 info->sctpi_maxseg = asoc->frag_point;
4518 info->sctpi_peer_rwnd = asoc->peer.rwnd;
4519 info->sctpi_peer_tag = asoc->c.peer_vtag;
4521 mask = asoc->peer.ecn_capable << 1;
4522 mask = (mask | asoc->peer.ipv4_address) << 1;
4523 mask = (mask | asoc->peer.ipv6_address) << 1;
4524 mask = (mask | asoc->peer.hostname_address) << 1;
4525 mask = (mask | asoc->peer.asconf_capable) << 1;
4526 mask = (mask | asoc->peer.prsctp_capable) << 1;
4527 mask = (mask | asoc->peer.auth_capable);
4528 info->sctpi_peer_capable = mask;
4529 mask = asoc->peer.sack_needed << 1;
4530 mask = (mask | asoc->peer.sack_generation) << 1;
4531 mask = (mask | asoc->peer.zero_window_announced);
4532 info->sctpi_peer_sack = mask;
4534 info->sctpi_isacks = asoc->stats.isacks;
4535 info->sctpi_osacks = asoc->stats.osacks;
4536 info->sctpi_opackets = asoc->stats.opackets;
4537 info->sctpi_ipackets = asoc->stats.ipackets;
4538 info->sctpi_rtxchunks = asoc->stats.rtxchunks;
4539 info->sctpi_outofseqtsns = asoc->stats.outofseqtsns;
4540 info->sctpi_idupchunks = asoc->stats.idupchunks;
4541 info->sctpi_gapcnt = asoc->stats.gapcnt;
4542 info->sctpi_ouodchunks = asoc->stats.ouodchunks;
4543 info->sctpi_iuodchunks = asoc->stats.iuodchunks;
4544 info->sctpi_oodchunks = asoc->stats.oodchunks;
4545 info->sctpi_iodchunks = asoc->stats.iodchunks;
4546 info->sctpi_octrlchunks = asoc->stats.octrlchunks;
4547 info->sctpi_ictrlchunks = asoc->stats.ictrlchunks;
4549 prim = asoc->peer.primary_path;
4550 memcpy(&info->sctpi_p_address, &prim->ipaddr, sizeof(prim->ipaddr));
4551 info->sctpi_p_state = prim->state;
4552 info->sctpi_p_cwnd = prim->cwnd;
4553 info->sctpi_p_srtt = prim->srtt;
4554 info->sctpi_p_rto = jiffies_to_msecs(prim->rto);
4555 info->sctpi_p_hbinterval = prim->hbinterval;
4556 info->sctpi_p_pathmaxrxt = prim->pathmaxrxt;
4557 info->sctpi_p_sackdelay = jiffies_to_msecs(prim->sackdelay);
4558 info->sctpi_p_ssthresh = prim->ssthresh;
4559 info->sctpi_p_partial_bytes_acked = prim->partial_bytes_acked;
4560 info->sctpi_p_flight_size = prim->flight_size;
4561 info->sctpi_p_error = prim->error_count;
4565 EXPORT_SYMBOL_GPL(sctp_get_sctp_info);
4567 /* use callback to avoid exporting the core structure */
4568 int sctp_transport_walk_start(struct rhashtable_iter *iter)
4572 rhltable_walk_enter(&sctp_transport_hashtable, iter);
4574 err = rhashtable_walk_start(iter);
4575 if (err && err != -EAGAIN) {
4576 rhashtable_walk_stop(iter);
4577 rhashtable_walk_exit(iter);
4584 void sctp_transport_walk_stop(struct rhashtable_iter *iter)
4586 rhashtable_walk_stop(iter);
4587 rhashtable_walk_exit(iter);
4590 struct sctp_transport *sctp_transport_get_next(struct net *net,
4591 struct rhashtable_iter *iter)
4593 struct sctp_transport *t;
4595 t = rhashtable_walk_next(iter);
4596 for (; t; t = rhashtable_walk_next(iter)) {
4598 if (PTR_ERR(t) == -EAGAIN)
4603 if (net_eq(sock_net(t->asoc->base.sk), net) &&
4604 t->asoc->peer.primary_path == t)
4611 struct sctp_transport *sctp_transport_get_idx(struct net *net,
4612 struct rhashtable_iter *iter,
4615 void *obj = SEQ_START_TOKEN;
4617 while (pos && (obj = sctp_transport_get_next(net, iter)) &&
4624 int sctp_for_each_endpoint(int (*cb)(struct sctp_endpoint *, void *),
4628 struct sctp_ep_common *epb;
4629 struct sctp_hashbucket *head;
4631 for (head = sctp_ep_hashtable; hash < sctp_ep_hashsize;
4633 read_lock_bh(&head->lock);
4634 sctp_for_each_hentry(epb, &head->chain) {
4635 err = cb(sctp_ep(epb), p);
4639 read_unlock_bh(&head->lock);
4644 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint);
4646 int sctp_transport_lookup_process(int (*cb)(struct sctp_transport *, void *),
4648 const union sctp_addr *laddr,
4649 const union sctp_addr *paddr, void *p)
4651 struct sctp_transport *transport;
4655 transport = sctp_addrs_lookup_transport(net, laddr, paddr);
4660 err = cb(transport, p);
4661 sctp_transport_put(transport);
4665 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process);
4667 int sctp_for_each_transport(int (*cb)(struct sctp_transport *, void *),
4668 int (*cb_done)(struct sctp_transport *, void *),
4669 struct net *net, int *pos, void *p) {
4670 struct rhashtable_iter hti;
4671 struct sctp_transport *tsp;
4675 ret = sctp_transport_walk_start(&hti);
4679 tsp = sctp_transport_get_idx(net, &hti, *pos + 1);
4680 for (; !IS_ERR_OR_NULL(tsp); tsp = sctp_transport_get_next(net, &hti)) {
4681 if (!sctp_transport_hold(tsp))
4687 sctp_transport_put(tsp);
4689 sctp_transport_walk_stop(&hti);
4692 if (cb_done && !cb_done(tsp, p)) {
4694 sctp_transport_put(tsp);
4697 sctp_transport_put(tsp);
4702 EXPORT_SYMBOL_GPL(sctp_for_each_transport);
4704 /* 7.2.1 Association Status (SCTP_STATUS)
4706 * Applications can retrieve current status information about an
4707 * association, including association state, peer receiver window size,
4708 * number of unacked data chunks, and number of data chunks pending
4709 * receipt. This information is read-only.
4711 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4712 char __user *optval,
4715 struct sctp_status status;
4716 struct sctp_association *asoc = NULL;
4717 struct sctp_transport *transport;
4718 sctp_assoc_t associd;
4721 if (len < sizeof(status)) {
4726 len = sizeof(status);
4727 if (copy_from_user(&status, optval, len)) {
4732 associd = status.sstat_assoc_id;
4733 asoc = sctp_id2assoc(sk, associd);
4739 transport = asoc->peer.primary_path;
4741 status.sstat_assoc_id = sctp_assoc2id(asoc);
4742 status.sstat_state = sctp_assoc_to_state(asoc);
4743 status.sstat_rwnd = asoc->peer.rwnd;
4744 status.sstat_unackdata = asoc->unack_data;
4746 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4747 status.sstat_instrms = asoc->stream.incnt;
4748 status.sstat_outstrms = asoc->stream.outcnt;
4749 status.sstat_fragmentation_point = asoc->frag_point;
4750 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4751 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4752 transport->af_specific->sockaddr_len);
4753 /* Map ipv4 address into v4-mapped-on-v6 address. */
4754 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
4755 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4756 status.sstat_primary.spinfo_state = transport->state;
4757 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4758 status.sstat_primary.spinfo_srtt = transport->srtt;
4759 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4760 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4762 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4763 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4765 if (put_user(len, optlen)) {
4770 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4771 __func__, len, status.sstat_state, status.sstat_rwnd,
4772 status.sstat_assoc_id);
4774 if (copy_to_user(optval, &status, len)) {
4784 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4786 * Applications can retrieve information about a specific peer address
4787 * of an association, including its reachability state, congestion
4788 * window, and retransmission timer values. This information is
4791 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4792 char __user *optval,
4795 struct sctp_paddrinfo pinfo;
4796 struct sctp_transport *transport;
4799 if (len < sizeof(pinfo)) {
4804 len = sizeof(pinfo);
4805 if (copy_from_user(&pinfo, optval, len)) {
4810 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4811 pinfo.spinfo_assoc_id);
4815 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4816 pinfo.spinfo_state = transport->state;
4817 pinfo.spinfo_cwnd = transport->cwnd;
4818 pinfo.spinfo_srtt = transport->srtt;
4819 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4820 pinfo.spinfo_mtu = transport->pathmtu;
4822 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4823 pinfo.spinfo_state = SCTP_ACTIVE;
4825 if (put_user(len, optlen)) {
4830 if (copy_to_user(optval, &pinfo, len)) {
4839 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4841 * This option is a on/off flag. If enabled no SCTP message
4842 * fragmentation will be performed. Instead if a message being sent
4843 * exceeds the current PMTU size, the message will NOT be sent and
4844 * instead a error will be indicated to the user.
4846 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4847 char __user *optval, int __user *optlen)
4851 if (len < sizeof(int))
4855 val = (sctp_sk(sk)->disable_fragments == 1);
4856 if (put_user(len, optlen))
4858 if (copy_to_user(optval, &val, len))
4863 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4865 * This socket option is used to specify various notifications and
4866 * ancillary data the user wishes to receive.
4868 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4873 if (len > sizeof(struct sctp_event_subscribe))
4874 len = sizeof(struct sctp_event_subscribe);
4875 if (put_user(len, optlen))
4877 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4882 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4884 * This socket option is applicable to the UDP-style socket only. When
4885 * set it will cause associations that are idle for more than the
4886 * specified number of seconds to automatically close. An association
4887 * being idle is defined an association that has NOT sent or received
4888 * user data. The special value of '0' indicates that no automatic
4889 * close of any associations should be performed. The option expects an
4890 * integer defining the number of seconds of idle time before an
4891 * association is closed.
4893 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4895 /* Applicable to UDP-style socket only */
4896 if (sctp_style(sk, TCP))
4898 if (len < sizeof(int))
4901 if (put_user(len, optlen))
4903 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4908 /* Helper routine to branch off an association to a new socket. */
4909 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4911 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4912 struct sctp_sock *sp = sctp_sk(sk);
4913 struct socket *sock;
4919 /* If there is a thread waiting on more sndbuf space for
4920 * sending on this asoc, it cannot be peeled.
4922 if (waitqueue_active(&asoc->wait))
4925 /* An association cannot be branched off from an already peeled-off
4926 * socket, nor is this supported for tcp style sockets.
4928 if (!sctp_style(sk, UDP))
4931 /* Create a new socket. */
4932 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4936 sctp_copy_sock(sock->sk, sk, asoc);
4938 /* Make peeled-off sockets more like 1-1 accepted sockets.
4939 * Set the daddr and initialize id to something more random
4941 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
4943 /* Populate the fields of the newsk from the oldsk and migrate the
4944 * asoc to the newsk.
4946 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4952 EXPORT_SYMBOL(sctp_do_peeloff);
4954 static int sctp_getsockopt_peeloff_common(struct sock *sk, sctp_peeloff_arg_t *peeloff,
4955 struct file **newfile, unsigned flags)
4957 struct socket *newsock;
4960 retval = sctp_do_peeloff(sk, peeloff->associd, &newsock);
4964 /* Map the socket to an unused fd that can be returned to the user. */
4965 retval = get_unused_fd_flags(flags & SOCK_CLOEXEC);
4967 sock_release(newsock);
4971 *newfile = sock_alloc_file(newsock, 0, NULL);
4972 if (IS_ERR(*newfile)) {
4973 put_unused_fd(retval);
4974 sock_release(newsock);
4975 retval = PTR_ERR(*newfile);
4980 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4983 peeloff->sd = retval;
4985 if (flags & SOCK_NONBLOCK)
4986 (*newfile)->f_flags |= O_NONBLOCK;
4991 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4993 sctp_peeloff_arg_t peeloff;
4994 struct file *newfile = NULL;
4997 if (len < sizeof(sctp_peeloff_arg_t))
4999 len = sizeof(sctp_peeloff_arg_t);
5000 if (copy_from_user(&peeloff, optval, len))
5003 retval = sctp_getsockopt_peeloff_common(sk, &peeloff, &newfile, 0);
5007 /* Return the fd mapped to the new socket. */
5008 if (put_user(len, optlen)) {
5010 put_unused_fd(retval);
5014 if (copy_to_user(optval, &peeloff, len)) {
5016 put_unused_fd(retval);
5019 fd_install(retval, newfile);
5024 static int sctp_getsockopt_peeloff_flags(struct sock *sk, int len,
5025 char __user *optval, int __user *optlen)
5027 sctp_peeloff_flags_arg_t peeloff;
5028 struct file *newfile = NULL;
5031 if (len < sizeof(sctp_peeloff_flags_arg_t))
5033 len = sizeof(sctp_peeloff_flags_arg_t);
5034 if (copy_from_user(&peeloff, optval, len))
5037 retval = sctp_getsockopt_peeloff_common(sk, &peeloff.p_arg,
5038 &newfile, peeloff.flags);
5042 /* Return the fd mapped to the new socket. */
5043 if (put_user(len, optlen)) {
5045 put_unused_fd(retval);
5049 if (copy_to_user(optval, &peeloff, len)) {
5051 put_unused_fd(retval);
5054 fd_install(retval, newfile);
5059 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
5061 * Applications can enable or disable heartbeats for any peer address of
5062 * an association, modify an address's heartbeat interval, force a
5063 * heartbeat to be sent immediately, and adjust the address's maximum
5064 * number of retransmissions sent before an address is considered
5065 * unreachable. The following structure is used to access and modify an
5066 * address's parameters:
5068 * struct sctp_paddrparams {
5069 * sctp_assoc_t spp_assoc_id;
5070 * struct sockaddr_storage spp_address;
5071 * uint32_t spp_hbinterval;
5072 * uint16_t spp_pathmaxrxt;
5073 * uint32_t spp_pathmtu;
5074 * uint32_t spp_sackdelay;
5075 * uint32_t spp_flags;
5078 * spp_assoc_id - (one-to-many style socket) This is filled in the
5079 * application, and identifies the association for
5081 * spp_address - This specifies which address is of interest.
5082 * spp_hbinterval - This contains the value of the heartbeat interval,
5083 * in milliseconds. If a value of zero
5084 * is present in this field then no changes are to
5085 * be made to this parameter.
5086 * spp_pathmaxrxt - This contains the maximum number of
5087 * retransmissions before this address shall be
5088 * considered unreachable. If a value of zero
5089 * is present in this field then no changes are to
5090 * be made to this parameter.
5091 * spp_pathmtu - When Path MTU discovery is disabled the value
5092 * specified here will be the "fixed" path mtu.
5093 * Note that if the spp_address field is empty
5094 * then all associations on this address will
5095 * have this fixed path mtu set upon them.
5097 * spp_sackdelay - When delayed sack is enabled, this value specifies
5098 * the number of milliseconds that sacks will be delayed
5099 * for. This value will apply to all addresses of an
5100 * association if the spp_address field is empty. Note
5101 * also, that if delayed sack is enabled and this
5102 * value is set to 0, no change is made to the last
5103 * recorded delayed sack timer value.
5105 * spp_flags - These flags are used to control various features
5106 * on an association. The flag field may contain
5107 * zero or more of the following options.
5109 * SPP_HB_ENABLE - Enable heartbeats on the
5110 * specified address. Note that if the address
5111 * field is empty all addresses for the association
5112 * have heartbeats enabled upon them.
5114 * SPP_HB_DISABLE - Disable heartbeats on the
5115 * speicifed address. Note that if the address
5116 * field is empty all addresses for the association
5117 * will have their heartbeats disabled. Note also
5118 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
5119 * mutually exclusive, only one of these two should
5120 * be specified. Enabling both fields will have
5121 * undetermined results.
5123 * SPP_HB_DEMAND - Request a user initiated heartbeat
5124 * to be made immediately.
5126 * SPP_PMTUD_ENABLE - This field will enable PMTU
5127 * discovery upon the specified address. Note that
5128 * if the address feild is empty then all addresses
5129 * on the association are effected.
5131 * SPP_PMTUD_DISABLE - This field will disable PMTU
5132 * discovery upon the specified address. Note that
5133 * if the address feild is empty then all addresses
5134 * on the association are effected. Not also that
5135 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
5136 * exclusive. Enabling both will have undetermined
5139 * SPP_SACKDELAY_ENABLE - Setting this flag turns
5140 * on delayed sack. The time specified in spp_sackdelay
5141 * is used to specify the sack delay for this address. Note
5142 * that if spp_address is empty then all addresses will
5143 * enable delayed sack and take on the sack delay
5144 * value specified in spp_sackdelay.
5145 * SPP_SACKDELAY_DISABLE - Setting this flag turns
5146 * off delayed sack. If the spp_address field is blank then
5147 * delayed sack is disabled for the entire association. Note
5148 * also that this field is mutually exclusive to
5149 * SPP_SACKDELAY_ENABLE, setting both will have undefined
5152 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
5153 char __user *optval, int __user *optlen)
5155 struct sctp_paddrparams params;
5156 struct sctp_transport *trans = NULL;
5157 struct sctp_association *asoc = NULL;
5158 struct sctp_sock *sp = sctp_sk(sk);
5160 if (len < sizeof(struct sctp_paddrparams))
5162 len = sizeof(struct sctp_paddrparams);
5163 if (copy_from_user(¶ms, optval, len))
5166 /* If an address other than INADDR_ANY is specified, and
5167 * no transport is found, then the request is invalid.
5169 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
5170 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
5171 params.spp_assoc_id);
5173 pr_debug("%s: failed no transport\n", __func__);
5178 /* Get association, if assoc_id != 0 and the socket is a one
5179 * to many style socket, and an association was not found, then
5180 * the id was invalid.
5182 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
5183 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
5184 pr_debug("%s: failed no association\n", __func__);
5189 /* Fetch transport values. */
5190 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
5191 params.spp_pathmtu = trans->pathmtu;
5192 params.spp_pathmaxrxt = trans->pathmaxrxt;
5193 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
5195 /*draft-11 doesn't say what to return in spp_flags*/
5196 params.spp_flags = trans->param_flags;
5198 /* Fetch association values. */
5199 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
5200 params.spp_pathmtu = asoc->pathmtu;
5201 params.spp_pathmaxrxt = asoc->pathmaxrxt;
5202 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
5204 /*draft-11 doesn't say what to return in spp_flags*/
5205 params.spp_flags = asoc->param_flags;
5207 /* Fetch socket values. */
5208 params.spp_hbinterval = sp->hbinterval;
5209 params.spp_pathmtu = sp->pathmtu;
5210 params.spp_sackdelay = sp->sackdelay;
5211 params.spp_pathmaxrxt = sp->pathmaxrxt;
5213 /*draft-11 doesn't say what to return in spp_flags*/
5214 params.spp_flags = sp->param_flags;
5217 if (copy_to_user(optval, ¶ms, len))
5220 if (put_user(len, optlen))
5227 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
5229 * This option will effect the way delayed acks are performed. This
5230 * option allows you to get or set the delayed ack time, in
5231 * milliseconds. It also allows changing the delayed ack frequency.
5232 * Changing the frequency to 1 disables the delayed sack algorithm. If
5233 * the assoc_id is 0, then this sets or gets the endpoints default
5234 * values. If the assoc_id field is non-zero, then the set or get
5235 * effects the specified association for the one to many model (the
5236 * assoc_id field is ignored by the one to one model). Note that if
5237 * sack_delay or sack_freq are 0 when setting this option, then the
5238 * current values will remain unchanged.
5240 * struct sctp_sack_info {
5241 * sctp_assoc_t sack_assoc_id;
5242 * uint32_t sack_delay;
5243 * uint32_t sack_freq;
5246 * sack_assoc_id - This parameter, indicates which association the user
5247 * is performing an action upon. Note that if this field's value is
5248 * zero then the endpoints default value is changed (effecting future
5249 * associations only).
5251 * sack_delay - This parameter contains the number of milliseconds that
5252 * the user is requesting the delayed ACK timer be set to. Note that
5253 * this value is defined in the standard to be between 200 and 500
5256 * sack_freq - This parameter contains the number of packets that must
5257 * be received before a sack is sent without waiting for the delay
5258 * timer to expire. The default value for this is 2, setting this
5259 * value to 1 will disable the delayed sack algorithm.
5261 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
5262 char __user *optval,
5265 struct sctp_sack_info params;
5266 struct sctp_association *asoc = NULL;
5267 struct sctp_sock *sp = sctp_sk(sk);
5269 if (len >= sizeof(struct sctp_sack_info)) {
5270 len = sizeof(struct sctp_sack_info);
5272 if (copy_from_user(¶ms, optval, len))
5274 } else if (len == sizeof(struct sctp_assoc_value)) {
5275 pr_warn_ratelimited(DEPRECATED
5277 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
5278 "Use struct sctp_sack_info instead\n",
5279 current->comm, task_pid_nr(current));
5280 if (copy_from_user(¶ms, optval, len))
5285 /* Get association, if sack_assoc_id != 0 and the socket is a one
5286 * to many style socket, and an association was not found, then
5287 * the id was invalid.
5289 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
5290 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
5294 /* Fetch association values. */
5295 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
5296 params.sack_delay = jiffies_to_msecs(
5298 params.sack_freq = asoc->sackfreq;
5301 params.sack_delay = 0;
5302 params.sack_freq = 1;
5305 /* Fetch socket values. */
5306 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
5307 params.sack_delay = sp->sackdelay;
5308 params.sack_freq = sp->sackfreq;
5310 params.sack_delay = 0;
5311 params.sack_freq = 1;
5315 if (copy_to_user(optval, ¶ms, len))
5318 if (put_user(len, optlen))
5324 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5326 * Applications can specify protocol parameters for the default association
5327 * initialization. The option name argument to setsockopt() and getsockopt()
5330 * Setting initialization parameters is effective only on an unconnected
5331 * socket (for UDP-style sockets only future associations are effected
5332 * by the change). With TCP-style sockets, this option is inherited by
5333 * sockets derived from a listener socket.
5335 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
5337 if (len < sizeof(struct sctp_initmsg))
5339 len = sizeof(struct sctp_initmsg);
5340 if (put_user(len, optlen))
5342 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
5348 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
5349 char __user *optval, int __user *optlen)
5351 struct sctp_association *asoc;
5353 struct sctp_getaddrs getaddrs;
5354 struct sctp_transport *from;
5356 union sctp_addr temp;
5357 struct sctp_sock *sp = sctp_sk(sk);
5362 if (len < sizeof(struct sctp_getaddrs))
5365 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5368 /* For UDP-style sockets, id specifies the association to query. */
5369 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5373 to = optval + offsetof(struct sctp_getaddrs, addrs);
5374 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5376 list_for_each_entry(from, &asoc->peer.transport_addr_list,
5378 memcpy(&temp, &from->ipaddr, sizeof(temp));
5379 addrlen = sctp_get_pf_specific(sk->sk_family)
5380 ->addr_to_user(sp, &temp);
5381 if (space_left < addrlen)
5383 if (copy_to_user(to, &temp, addrlen))
5387 space_left -= addrlen;
5390 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
5392 bytes_copied = ((char __user *)to) - optval;
5393 if (put_user(bytes_copied, optlen))
5399 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
5400 size_t space_left, int *bytes_copied)
5402 struct sctp_sockaddr_entry *addr;
5403 union sctp_addr temp;
5406 struct net *net = sock_net(sk);
5409 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
5413 if ((PF_INET == sk->sk_family) &&
5414 (AF_INET6 == addr->a.sa.sa_family))
5416 if ((PF_INET6 == sk->sk_family) &&
5417 inet_v6_ipv6only(sk) &&
5418 (AF_INET == addr->a.sa.sa_family))
5420 memcpy(&temp, &addr->a, sizeof(temp));
5421 if (!temp.v4.sin_port)
5422 temp.v4.sin_port = htons(port);
5424 addrlen = sctp_get_pf_specific(sk->sk_family)
5425 ->addr_to_user(sctp_sk(sk), &temp);
5427 if (space_left < addrlen) {
5431 memcpy(to, &temp, addrlen);
5435 space_left -= addrlen;
5436 *bytes_copied += addrlen;
5444 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
5445 char __user *optval, int __user *optlen)
5447 struct sctp_bind_addr *bp;
5448 struct sctp_association *asoc;
5450 struct sctp_getaddrs getaddrs;
5451 struct sctp_sockaddr_entry *addr;
5453 union sctp_addr temp;
5454 struct sctp_sock *sp = sctp_sk(sk);
5458 int bytes_copied = 0;
5462 if (len < sizeof(struct sctp_getaddrs))
5465 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5469 * For UDP-style sockets, id specifies the association to query.
5470 * If the id field is set to the value '0' then the locally bound
5471 * addresses are returned without regard to any particular
5474 if (0 == getaddrs.assoc_id) {
5475 bp = &sctp_sk(sk)->ep->base.bind_addr;
5477 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5480 bp = &asoc->base.bind_addr;
5483 to = optval + offsetof(struct sctp_getaddrs, addrs);
5484 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5486 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN);
5490 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5491 * addresses from the global local address list.
5493 if (sctp_list_single_entry(&bp->address_list)) {
5494 addr = list_entry(bp->address_list.next,
5495 struct sctp_sockaddr_entry, list);
5496 if (sctp_is_any(sk, &addr->a)) {
5497 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
5498 space_left, &bytes_copied);
5508 /* Protection on the bound address list is not needed since
5509 * in the socket option context we hold a socket lock and
5510 * thus the bound address list can't change.
5512 list_for_each_entry(addr, &bp->address_list, list) {
5513 memcpy(&temp, &addr->a, sizeof(temp));
5514 addrlen = sctp_get_pf_specific(sk->sk_family)
5515 ->addr_to_user(sp, &temp);
5516 if (space_left < addrlen) {
5517 err = -ENOMEM; /*fixme: right error?*/
5520 memcpy(buf, &temp, addrlen);
5522 bytes_copied += addrlen;
5524 space_left -= addrlen;
5528 if (copy_to_user(to, addrs, bytes_copied)) {
5532 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
5536 if (put_user(bytes_copied, optlen))
5543 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
5545 * Requests that the local SCTP stack use the enclosed peer address as
5546 * the association primary. The enclosed address must be one of the
5547 * association peer's addresses.
5549 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
5550 char __user *optval, int __user *optlen)
5552 struct sctp_prim prim;
5553 struct sctp_association *asoc;
5554 struct sctp_sock *sp = sctp_sk(sk);
5556 if (len < sizeof(struct sctp_prim))
5559 len = sizeof(struct sctp_prim);
5561 if (copy_from_user(&prim, optval, len))
5564 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
5568 if (!asoc->peer.primary_path)
5571 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
5572 asoc->peer.primary_path->af_specific->sockaddr_len);
5574 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
5575 (union sctp_addr *)&prim.ssp_addr);
5577 if (put_user(len, optlen))
5579 if (copy_to_user(optval, &prim, len))
5586 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5588 * Requests that the local endpoint set the specified Adaptation Layer
5589 * Indication parameter for all future INIT and INIT-ACK exchanges.
5591 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
5592 char __user *optval, int __user *optlen)
5594 struct sctp_setadaptation adaptation;
5596 if (len < sizeof(struct sctp_setadaptation))
5599 len = sizeof(struct sctp_setadaptation);
5601 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
5603 if (put_user(len, optlen))
5605 if (copy_to_user(optval, &adaptation, len))
5613 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5615 * Applications that wish to use the sendto() system call may wish to
5616 * specify a default set of parameters that would normally be supplied
5617 * through the inclusion of ancillary data. This socket option allows
5618 * such an application to set the default sctp_sndrcvinfo structure.
5621 * The application that wishes to use this socket option simply passes
5622 * in to this call the sctp_sndrcvinfo structure defined in Section
5623 * 5.2.2) The input parameters accepted by this call include
5624 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5625 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5626 * to this call if the caller is using the UDP model.
5628 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5630 static int sctp_getsockopt_default_send_param(struct sock *sk,
5631 int len, char __user *optval,
5634 struct sctp_sock *sp = sctp_sk(sk);
5635 struct sctp_association *asoc;
5636 struct sctp_sndrcvinfo info;
5638 if (len < sizeof(info))
5643 if (copy_from_user(&info, optval, len))
5646 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
5647 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
5650 info.sinfo_stream = asoc->default_stream;
5651 info.sinfo_flags = asoc->default_flags;
5652 info.sinfo_ppid = asoc->default_ppid;
5653 info.sinfo_context = asoc->default_context;
5654 info.sinfo_timetolive = asoc->default_timetolive;
5656 info.sinfo_stream = sp->default_stream;
5657 info.sinfo_flags = sp->default_flags;
5658 info.sinfo_ppid = sp->default_ppid;
5659 info.sinfo_context = sp->default_context;
5660 info.sinfo_timetolive = sp->default_timetolive;
5663 if (put_user(len, optlen))
5665 if (copy_to_user(optval, &info, len))
5671 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5672 * (SCTP_DEFAULT_SNDINFO)
5674 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
5675 char __user *optval,
5678 struct sctp_sock *sp = sctp_sk(sk);
5679 struct sctp_association *asoc;
5680 struct sctp_sndinfo info;
5682 if (len < sizeof(info))
5687 if (copy_from_user(&info, optval, len))
5690 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
5691 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
5694 info.snd_sid = asoc->default_stream;
5695 info.snd_flags = asoc->default_flags;
5696 info.snd_ppid = asoc->default_ppid;
5697 info.snd_context = asoc->default_context;
5699 info.snd_sid = sp->default_stream;
5700 info.snd_flags = sp->default_flags;
5701 info.snd_ppid = sp->default_ppid;
5702 info.snd_context = sp->default_context;
5705 if (put_user(len, optlen))
5707 if (copy_to_user(optval, &info, len))
5715 * 7.1.5 SCTP_NODELAY
5717 * Turn on/off any Nagle-like algorithm. This means that packets are
5718 * generally sent as soon as possible and no unnecessary delays are
5719 * introduced, at the cost of more packets in the network. Expects an
5720 * integer boolean flag.
5723 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5724 char __user *optval, int __user *optlen)
5728 if (len < sizeof(int))
5732 val = (sctp_sk(sk)->nodelay == 1);
5733 if (put_user(len, optlen))
5735 if (copy_to_user(optval, &val, len))
5742 * 7.1.1 SCTP_RTOINFO
5744 * The protocol parameters used to initialize and bound retransmission
5745 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5746 * and modify these parameters.
5747 * All parameters are time values, in milliseconds. A value of 0, when
5748 * modifying the parameters, indicates that the current value should not
5752 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5753 char __user *optval,
5754 int __user *optlen) {
5755 struct sctp_rtoinfo rtoinfo;
5756 struct sctp_association *asoc;
5758 if (len < sizeof (struct sctp_rtoinfo))
5761 len = sizeof(struct sctp_rtoinfo);
5763 if (copy_from_user(&rtoinfo, optval, len))
5766 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5768 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5771 /* Values corresponding to the specific association. */
5773 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5774 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5775 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5777 /* Values corresponding to the endpoint. */
5778 struct sctp_sock *sp = sctp_sk(sk);
5780 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5781 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5782 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5785 if (put_user(len, optlen))
5788 if (copy_to_user(optval, &rtoinfo, len))
5796 * 7.1.2 SCTP_ASSOCINFO
5798 * This option is used to tune the maximum retransmission attempts
5799 * of the association.
5800 * Returns an error if the new association retransmission value is
5801 * greater than the sum of the retransmission value of the peer.
5802 * See [SCTP] for more information.
5805 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5806 char __user *optval,
5810 struct sctp_assocparams assocparams;
5811 struct sctp_association *asoc;
5812 struct list_head *pos;
5815 if (len < sizeof (struct sctp_assocparams))
5818 len = sizeof(struct sctp_assocparams);
5820 if (copy_from_user(&assocparams, optval, len))
5823 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5825 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5828 /* Values correspoinding to the specific association */
5830 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5831 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5832 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5833 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5835 list_for_each(pos, &asoc->peer.transport_addr_list) {
5839 assocparams.sasoc_number_peer_destinations = cnt;
5841 /* Values corresponding to the endpoint */
5842 struct sctp_sock *sp = sctp_sk(sk);
5844 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5845 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5846 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5847 assocparams.sasoc_cookie_life =
5848 sp->assocparams.sasoc_cookie_life;
5849 assocparams.sasoc_number_peer_destinations =
5851 sasoc_number_peer_destinations;
5854 if (put_user(len, optlen))
5857 if (copy_to_user(optval, &assocparams, len))
5864 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5866 * This socket option is a boolean flag which turns on or off mapped V4
5867 * addresses. If this option is turned on and the socket is type
5868 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5869 * If this option is turned off, then no mapping will be done of V4
5870 * addresses and a user will receive both PF_INET6 and PF_INET type
5871 * addresses on the socket.
5873 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5874 char __user *optval, int __user *optlen)
5877 struct sctp_sock *sp = sctp_sk(sk);
5879 if (len < sizeof(int))
5884 if (put_user(len, optlen))
5886 if (copy_to_user(optval, &val, len))
5893 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5894 * (chapter and verse is quoted at sctp_setsockopt_context())
5896 static int sctp_getsockopt_context(struct sock *sk, int len,
5897 char __user *optval, int __user *optlen)
5899 struct sctp_assoc_value params;
5900 struct sctp_sock *sp;
5901 struct sctp_association *asoc;
5903 if (len < sizeof(struct sctp_assoc_value))
5906 len = sizeof(struct sctp_assoc_value);
5908 if (copy_from_user(¶ms, optval, len))
5913 if (params.assoc_id != 0) {
5914 asoc = sctp_id2assoc(sk, params.assoc_id);
5917 params.assoc_value = asoc->default_rcv_context;
5919 params.assoc_value = sp->default_rcv_context;
5922 if (put_user(len, optlen))
5924 if (copy_to_user(optval, ¶ms, len))
5931 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5932 * This option will get or set the maximum size to put in any outgoing
5933 * SCTP DATA chunk. If a message is larger than this size it will be
5934 * fragmented by SCTP into the specified size. Note that the underlying
5935 * SCTP implementation may fragment into smaller sized chunks when the
5936 * PMTU of the underlying association is smaller than the value set by
5937 * the user. The default value for this option is '0' which indicates
5938 * the user is NOT limiting fragmentation and only the PMTU will effect
5939 * SCTP's choice of DATA chunk size. Note also that values set larger
5940 * than the maximum size of an IP datagram will effectively let SCTP
5941 * control fragmentation (i.e. the same as setting this option to 0).
5943 * The following structure is used to access and modify this parameter:
5945 * struct sctp_assoc_value {
5946 * sctp_assoc_t assoc_id;
5947 * uint32_t assoc_value;
5950 * assoc_id: This parameter is ignored for one-to-one style sockets.
5951 * For one-to-many style sockets this parameter indicates which
5952 * association the user is performing an action upon. Note that if
5953 * this field's value is zero then the endpoints default value is
5954 * changed (effecting future associations only).
5955 * assoc_value: This parameter specifies the maximum size in bytes.
5957 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5958 char __user *optval, int __user *optlen)
5960 struct sctp_assoc_value params;
5961 struct sctp_association *asoc;
5963 if (len == sizeof(int)) {
5964 pr_warn_ratelimited(DEPRECATED
5966 "Use of int in maxseg socket option.\n"
5967 "Use struct sctp_assoc_value instead\n",
5968 current->comm, task_pid_nr(current));
5969 params.assoc_id = 0;
5970 } else if (len >= sizeof(struct sctp_assoc_value)) {
5971 len = sizeof(struct sctp_assoc_value);
5972 if (copy_from_user(¶ms, optval, sizeof(params)))
5977 asoc = sctp_id2assoc(sk, params.assoc_id);
5978 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5982 params.assoc_value = asoc->frag_point;
5984 params.assoc_value = sctp_sk(sk)->user_frag;
5986 if (put_user(len, optlen))
5988 if (len == sizeof(int)) {
5989 if (copy_to_user(optval, ¶ms.assoc_value, len))
5992 if (copy_to_user(optval, ¶ms, len))
6000 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
6001 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
6003 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
6004 char __user *optval, int __user *optlen)
6008 if (len < sizeof(int))
6013 val = sctp_sk(sk)->frag_interleave;
6014 if (put_user(len, optlen))
6016 if (copy_to_user(optval, &val, len))
6023 * 7.1.25. Set or Get the sctp partial delivery point
6024 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
6026 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
6027 char __user *optval,
6032 if (len < sizeof(u32))
6037 val = sctp_sk(sk)->pd_point;
6038 if (put_user(len, optlen))
6040 if (copy_to_user(optval, &val, len))
6047 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
6048 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
6050 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
6051 char __user *optval,
6054 struct sctp_assoc_value params;
6055 struct sctp_sock *sp;
6056 struct sctp_association *asoc;
6058 if (len == sizeof(int)) {
6059 pr_warn_ratelimited(DEPRECATED
6061 "Use of int in max_burst socket option.\n"
6062 "Use struct sctp_assoc_value instead\n",
6063 current->comm, task_pid_nr(current));
6064 params.assoc_id = 0;
6065 } else if (len >= sizeof(struct sctp_assoc_value)) {
6066 len = sizeof(struct sctp_assoc_value);
6067 if (copy_from_user(¶ms, optval, len))
6074 if (params.assoc_id != 0) {
6075 asoc = sctp_id2assoc(sk, params.assoc_id);
6078 params.assoc_value = asoc->max_burst;
6080 params.assoc_value = sp->max_burst;
6082 if (len == sizeof(int)) {
6083 if (copy_to_user(optval, ¶ms.assoc_value, len))
6086 if (copy_to_user(optval, ¶ms, len))
6094 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
6095 char __user *optval, int __user *optlen)
6097 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6098 struct sctp_hmacalgo __user *p = (void __user *)optval;
6099 struct sctp_hmac_algo_param *hmacs;
6104 if (!ep->auth_enable)
6107 hmacs = ep->auth_hmacs_list;
6108 data_len = ntohs(hmacs->param_hdr.length) -
6109 sizeof(struct sctp_paramhdr);
6111 if (len < sizeof(struct sctp_hmacalgo) + data_len)
6114 len = sizeof(struct sctp_hmacalgo) + data_len;
6115 num_idents = data_len / sizeof(u16);
6117 if (put_user(len, optlen))
6119 if (put_user(num_idents, &p->shmac_num_idents))
6121 for (i = 0; i < num_idents; i++) {
6122 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
6124 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
6130 static int sctp_getsockopt_active_key(struct sock *sk, int len,
6131 char __user *optval, int __user *optlen)
6133 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6134 struct sctp_authkeyid val;
6135 struct sctp_association *asoc;
6137 if (!ep->auth_enable)
6140 if (len < sizeof(struct sctp_authkeyid))
6142 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
6145 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
6146 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
6150 val.scact_keynumber = asoc->active_key_id;
6152 val.scact_keynumber = ep->active_key_id;
6154 len = sizeof(struct sctp_authkeyid);
6155 if (put_user(len, optlen))
6157 if (copy_to_user(optval, &val, len))
6163 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
6164 char __user *optval, int __user *optlen)
6166 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6167 struct sctp_authchunks __user *p = (void __user *)optval;
6168 struct sctp_authchunks val;
6169 struct sctp_association *asoc;
6170 struct sctp_chunks_param *ch;
6174 if (!ep->auth_enable)
6177 if (len < sizeof(struct sctp_authchunks))
6180 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
6183 to = p->gauth_chunks;
6184 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6188 ch = asoc->peer.peer_chunks;
6192 /* See if the user provided enough room for all the data */
6193 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6194 if (len < num_chunks)
6197 if (copy_to_user(to, ch->chunks, num_chunks))
6200 len = sizeof(struct sctp_authchunks) + num_chunks;
6201 if (put_user(len, optlen))
6203 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6208 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
6209 char __user *optval, int __user *optlen)
6211 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6212 struct sctp_authchunks __user *p = (void __user *)optval;
6213 struct sctp_authchunks val;
6214 struct sctp_association *asoc;
6215 struct sctp_chunks_param *ch;
6219 if (!ep->auth_enable)
6222 if (len < sizeof(struct sctp_authchunks))
6225 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
6228 to = p->gauth_chunks;
6229 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6230 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
6234 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
6236 ch = ep->auth_chunk_list;
6241 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6242 if (len < sizeof(struct sctp_authchunks) + num_chunks)
6245 if (copy_to_user(to, ch->chunks, num_chunks))
6248 len = sizeof(struct sctp_authchunks) + num_chunks;
6249 if (put_user(len, optlen))
6251 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6258 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
6259 * This option gets the current number of associations that are attached
6260 * to a one-to-many style socket. The option value is an uint32_t.
6262 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
6263 char __user *optval, int __user *optlen)
6265 struct sctp_sock *sp = sctp_sk(sk);
6266 struct sctp_association *asoc;
6269 if (sctp_style(sk, TCP))
6272 if (len < sizeof(u32))
6277 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6281 if (put_user(len, optlen))
6283 if (copy_to_user(optval, &val, len))
6290 * 8.1.23 SCTP_AUTO_ASCONF
6291 * See the corresponding setsockopt entry as description
6293 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
6294 char __user *optval, int __user *optlen)
6298 if (len < sizeof(int))
6302 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
6304 if (put_user(len, optlen))
6306 if (copy_to_user(optval, &val, len))
6312 * 8.2.6. Get the Current Identifiers of Associations
6313 * (SCTP_GET_ASSOC_ID_LIST)
6315 * This option gets the current list of SCTP association identifiers of
6316 * the SCTP associations handled by a one-to-many style socket.
6318 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
6319 char __user *optval, int __user *optlen)
6321 struct sctp_sock *sp = sctp_sk(sk);
6322 struct sctp_association *asoc;
6323 struct sctp_assoc_ids *ids;
6326 if (sctp_style(sk, TCP))
6329 if (len < sizeof(struct sctp_assoc_ids))
6332 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6336 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
6339 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
6341 ids = kmalloc(len, GFP_USER | __GFP_NOWARN);
6345 ids->gaids_number_of_ids = num;
6347 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6348 ids->gaids_assoc_id[num++] = asoc->assoc_id;
6351 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
6361 * SCTP_PEER_ADDR_THLDS
6363 * This option allows us to fetch the partially failed threshold for one or all
6364 * transports in an association. See Section 6.1 of:
6365 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6367 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
6368 char __user *optval,
6372 struct sctp_paddrthlds val;
6373 struct sctp_transport *trans;
6374 struct sctp_association *asoc;
6376 if (len < sizeof(struct sctp_paddrthlds))
6378 len = sizeof(struct sctp_paddrthlds);
6379 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
6382 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
6383 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
6387 val.spt_pathpfthld = asoc->pf_retrans;
6388 val.spt_pathmaxrxt = asoc->pathmaxrxt;
6390 trans = sctp_addr_id2transport(sk, &val.spt_address,
6395 val.spt_pathmaxrxt = trans->pathmaxrxt;
6396 val.spt_pathpfthld = trans->pf_retrans;
6399 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
6406 * SCTP_GET_ASSOC_STATS
6408 * This option retrieves local per endpoint statistics. It is modeled
6409 * after OpenSolaris' implementation
6411 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
6412 char __user *optval,
6415 struct sctp_assoc_stats sas;
6416 struct sctp_association *asoc = NULL;
6418 /* User must provide at least the assoc id */
6419 if (len < sizeof(sctp_assoc_t))
6422 /* Allow the struct to grow and fill in as much as possible */
6423 len = min_t(size_t, len, sizeof(sas));
6425 if (copy_from_user(&sas, optval, len))
6428 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
6432 sas.sas_rtxchunks = asoc->stats.rtxchunks;
6433 sas.sas_gapcnt = asoc->stats.gapcnt;
6434 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
6435 sas.sas_osacks = asoc->stats.osacks;
6436 sas.sas_isacks = asoc->stats.isacks;
6437 sas.sas_octrlchunks = asoc->stats.octrlchunks;
6438 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
6439 sas.sas_oodchunks = asoc->stats.oodchunks;
6440 sas.sas_iodchunks = asoc->stats.iodchunks;
6441 sas.sas_ouodchunks = asoc->stats.ouodchunks;
6442 sas.sas_iuodchunks = asoc->stats.iuodchunks;
6443 sas.sas_idupchunks = asoc->stats.idupchunks;
6444 sas.sas_opackets = asoc->stats.opackets;
6445 sas.sas_ipackets = asoc->stats.ipackets;
6447 /* New high max rto observed, will return 0 if not a single
6448 * RTO update took place. obs_rto_ipaddr will be bogus
6451 sas.sas_maxrto = asoc->stats.max_obs_rto;
6452 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
6453 sizeof(struct sockaddr_storage));
6455 /* Mark beginning of a new observation period */
6456 asoc->stats.max_obs_rto = asoc->rto_min;
6458 if (put_user(len, optlen))
6461 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
6463 if (copy_to_user(optval, &sas, len))
6469 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
6470 char __user *optval,
6475 if (len < sizeof(int))
6479 if (sctp_sk(sk)->recvrcvinfo)
6481 if (put_user(len, optlen))
6483 if (copy_to_user(optval, &val, len))
6489 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
6490 char __user *optval,
6495 if (len < sizeof(int))
6499 if (sctp_sk(sk)->recvnxtinfo)
6501 if (put_user(len, optlen))
6503 if (copy_to_user(optval, &val, len))
6509 static int sctp_getsockopt_pr_supported(struct sock *sk, int len,
6510 char __user *optval,
6513 struct sctp_assoc_value params;
6514 struct sctp_association *asoc;
6515 int retval = -EFAULT;
6517 if (len < sizeof(params)) {
6522 len = sizeof(params);
6523 if (copy_from_user(¶ms, optval, len))
6526 asoc = sctp_id2assoc(sk, params.assoc_id);
6528 params.assoc_value = asoc->prsctp_enable;
6529 } else if (!params.assoc_id) {
6530 struct sctp_sock *sp = sctp_sk(sk);
6532 params.assoc_value = sp->ep->prsctp_enable;
6538 if (put_user(len, optlen))
6541 if (copy_to_user(optval, ¶ms, len))
6550 static int sctp_getsockopt_default_prinfo(struct sock *sk, int len,
6551 char __user *optval,
6554 struct sctp_default_prinfo info;
6555 struct sctp_association *asoc;
6556 int retval = -EFAULT;
6558 if (len < sizeof(info)) {
6564 if (copy_from_user(&info, optval, len))
6567 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
6569 info.pr_policy = SCTP_PR_POLICY(asoc->default_flags);
6570 info.pr_value = asoc->default_timetolive;
6571 } else if (!info.pr_assoc_id) {
6572 struct sctp_sock *sp = sctp_sk(sk);
6574 info.pr_policy = SCTP_PR_POLICY(sp->default_flags);
6575 info.pr_value = sp->default_timetolive;
6581 if (put_user(len, optlen))
6584 if (copy_to_user(optval, &info, len))
6593 static int sctp_getsockopt_pr_assocstatus(struct sock *sk, int len,
6594 char __user *optval,
6597 struct sctp_prstatus params;
6598 struct sctp_association *asoc;
6600 int retval = -EINVAL;
6602 if (len < sizeof(params))
6605 len = sizeof(params);
6606 if (copy_from_user(¶ms, optval, len)) {
6611 policy = params.sprstat_policy;
6612 if (policy & ~SCTP_PR_SCTP_MASK)
6615 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
6619 if (policy == SCTP_PR_SCTP_NONE) {
6620 params.sprstat_abandoned_unsent = 0;
6621 params.sprstat_abandoned_sent = 0;
6622 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
6623 params.sprstat_abandoned_unsent +=
6624 asoc->abandoned_unsent[policy];
6625 params.sprstat_abandoned_sent +=
6626 asoc->abandoned_sent[policy];
6629 params.sprstat_abandoned_unsent =
6630 asoc->abandoned_unsent[__SCTP_PR_INDEX(policy)];
6631 params.sprstat_abandoned_sent =
6632 asoc->abandoned_sent[__SCTP_PR_INDEX(policy)];
6635 if (put_user(len, optlen)) {
6640 if (copy_to_user(optval, ¶ms, len)) {
6651 static int sctp_getsockopt_pr_streamstatus(struct sock *sk, int len,
6652 char __user *optval,
6655 struct sctp_stream_out_ext *streamoute;
6656 struct sctp_association *asoc;
6657 struct sctp_prstatus params;
6658 int retval = -EINVAL;
6661 if (len < sizeof(params))
6664 len = sizeof(params);
6665 if (copy_from_user(¶ms, optval, len)) {
6670 policy = params.sprstat_policy;
6671 if (policy & ~SCTP_PR_SCTP_MASK)
6674 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
6675 if (!asoc || params.sprstat_sid >= asoc->stream.outcnt)
6678 streamoute = asoc->stream.out[params.sprstat_sid].ext;
6680 /* Not allocated yet, means all stats are 0 */
6681 params.sprstat_abandoned_unsent = 0;
6682 params.sprstat_abandoned_sent = 0;
6687 if (policy == SCTP_PR_SCTP_NONE) {
6688 params.sprstat_abandoned_unsent = 0;
6689 params.sprstat_abandoned_sent = 0;
6690 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
6691 params.sprstat_abandoned_unsent +=
6692 streamoute->abandoned_unsent[policy];
6693 params.sprstat_abandoned_sent +=
6694 streamoute->abandoned_sent[policy];
6697 params.sprstat_abandoned_unsent =
6698 streamoute->abandoned_unsent[__SCTP_PR_INDEX(policy)];
6699 params.sprstat_abandoned_sent =
6700 streamoute->abandoned_sent[__SCTP_PR_INDEX(policy)];
6703 if (put_user(len, optlen) || copy_to_user(optval, ¶ms, len)) {
6714 static int sctp_getsockopt_reconfig_supported(struct sock *sk, int len,
6715 char __user *optval,
6718 struct sctp_assoc_value params;
6719 struct sctp_association *asoc;
6720 int retval = -EFAULT;
6722 if (len < sizeof(params)) {
6727 len = sizeof(params);
6728 if (copy_from_user(¶ms, optval, len))
6731 asoc = sctp_id2assoc(sk, params.assoc_id);
6733 params.assoc_value = asoc->reconf_enable;
6734 } else if (!params.assoc_id) {
6735 struct sctp_sock *sp = sctp_sk(sk);
6737 params.assoc_value = sp->ep->reconf_enable;
6743 if (put_user(len, optlen))
6746 if (copy_to_user(optval, ¶ms, len))
6755 static int sctp_getsockopt_enable_strreset(struct sock *sk, int len,
6756 char __user *optval,
6759 struct sctp_assoc_value params;
6760 struct sctp_association *asoc;
6761 int retval = -EFAULT;
6763 if (len < sizeof(params)) {
6768 len = sizeof(params);
6769 if (copy_from_user(¶ms, optval, len))
6772 asoc = sctp_id2assoc(sk, params.assoc_id);
6774 params.assoc_value = asoc->strreset_enable;
6775 } else if (!params.assoc_id) {
6776 struct sctp_sock *sp = sctp_sk(sk);
6778 params.assoc_value = sp->ep->strreset_enable;
6784 if (put_user(len, optlen))
6787 if (copy_to_user(optval, ¶ms, len))
6796 static int sctp_getsockopt(struct sock *sk, int level, int optname,
6797 char __user *optval, int __user *optlen)
6802 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
6804 /* I can hardly begin to describe how wrong this is. This is
6805 * so broken as to be worse than useless. The API draft
6806 * REALLY is NOT helpful here... I am not convinced that the
6807 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
6808 * are at all well-founded.
6810 if (level != SOL_SCTP) {
6811 struct sctp_af *af = sctp_sk(sk)->pf->af;
6813 retval = af->getsockopt(sk, level, optname, optval, optlen);
6817 if (get_user(len, optlen))
6827 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
6829 case SCTP_DISABLE_FRAGMENTS:
6830 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
6834 retval = sctp_getsockopt_events(sk, len, optval, optlen);
6836 case SCTP_AUTOCLOSE:
6837 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
6839 case SCTP_SOCKOPT_PEELOFF:
6840 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
6842 case SCTP_SOCKOPT_PEELOFF_FLAGS:
6843 retval = sctp_getsockopt_peeloff_flags(sk, len, optval, optlen);
6845 case SCTP_PEER_ADDR_PARAMS:
6846 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
6849 case SCTP_DELAYED_SACK:
6850 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
6854 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
6856 case SCTP_GET_PEER_ADDRS:
6857 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
6860 case SCTP_GET_LOCAL_ADDRS:
6861 retval = sctp_getsockopt_local_addrs(sk, len, optval,
6864 case SCTP_SOCKOPT_CONNECTX3:
6865 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
6867 case SCTP_DEFAULT_SEND_PARAM:
6868 retval = sctp_getsockopt_default_send_param(sk, len,
6871 case SCTP_DEFAULT_SNDINFO:
6872 retval = sctp_getsockopt_default_sndinfo(sk, len,
6875 case SCTP_PRIMARY_ADDR:
6876 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
6879 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
6882 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
6884 case SCTP_ASSOCINFO:
6885 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
6887 case SCTP_I_WANT_MAPPED_V4_ADDR:
6888 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
6891 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
6893 case SCTP_GET_PEER_ADDR_INFO:
6894 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
6897 case SCTP_ADAPTATION_LAYER:
6898 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
6902 retval = sctp_getsockopt_context(sk, len, optval, optlen);
6904 case SCTP_FRAGMENT_INTERLEAVE:
6905 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
6908 case SCTP_PARTIAL_DELIVERY_POINT:
6909 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
6912 case SCTP_MAX_BURST:
6913 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
6916 case SCTP_AUTH_CHUNK:
6917 case SCTP_AUTH_DELETE_KEY:
6918 retval = -EOPNOTSUPP;
6920 case SCTP_HMAC_IDENT:
6921 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
6923 case SCTP_AUTH_ACTIVE_KEY:
6924 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
6926 case SCTP_PEER_AUTH_CHUNKS:
6927 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
6930 case SCTP_LOCAL_AUTH_CHUNKS:
6931 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
6934 case SCTP_GET_ASSOC_NUMBER:
6935 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
6937 case SCTP_GET_ASSOC_ID_LIST:
6938 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
6940 case SCTP_AUTO_ASCONF:
6941 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
6943 case SCTP_PEER_ADDR_THLDS:
6944 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
6946 case SCTP_GET_ASSOC_STATS:
6947 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
6949 case SCTP_RECVRCVINFO:
6950 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
6952 case SCTP_RECVNXTINFO:
6953 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
6955 case SCTP_PR_SUPPORTED:
6956 retval = sctp_getsockopt_pr_supported(sk, len, optval, optlen);
6958 case SCTP_DEFAULT_PRINFO:
6959 retval = sctp_getsockopt_default_prinfo(sk, len, optval,
6962 case SCTP_PR_ASSOC_STATUS:
6963 retval = sctp_getsockopt_pr_assocstatus(sk, len, optval,
6966 case SCTP_PR_STREAM_STATUS:
6967 retval = sctp_getsockopt_pr_streamstatus(sk, len, optval,
6970 case SCTP_RECONFIG_SUPPORTED:
6971 retval = sctp_getsockopt_reconfig_supported(sk, len, optval,
6974 case SCTP_ENABLE_STREAM_RESET:
6975 retval = sctp_getsockopt_enable_strreset(sk, len, optval,
6979 retval = -ENOPROTOOPT;
6987 static int sctp_hash(struct sock *sk)
6993 static void sctp_unhash(struct sock *sk)
6998 /* Check if port is acceptable. Possibly find first available port.
7000 * The port hash table (contained in the 'global' SCTP protocol storage
7001 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
7002 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
7003 * list (the list number is the port number hashed out, so as you
7004 * would expect from a hash function, all the ports in a given list have
7005 * such a number that hashes out to the same list number; you were
7006 * expecting that, right?); so each list has a set of ports, with a
7007 * link to the socket (struct sock) that uses it, the port number and
7008 * a fastreuse flag (FIXME: NPI ipg).
7010 static struct sctp_bind_bucket *sctp_bucket_create(
7011 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
7013 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
7015 struct sctp_bind_hashbucket *head; /* hash list */
7016 struct sctp_bind_bucket *pp;
7017 unsigned short snum;
7020 snum = ntohs(addr->v4.sin_port);
7022 pr_debug("%s: begins, snum:%d\n", __func__, snum);
7027 /* Search for an available port. */
7028 int low, high, remaining, index;
7030 struct net *net = sock_net(sk);
7032 inet_get_local_port_range(net, &low, &high);
7033 remaining = (high - low) + 1;
7034 rover = prandom_u32() % remaining + low;
7038 if ((rover < low) || (rover > high))
7040 if (inet_is_local_reserved_port(net, rover))
7042 index = sctp_phashfn(sock_net(sk), rover);
7043 head = &sctp_port_hashtable[index];
7044 spin_lock(&head->lock);
7045 sctp_for_each_hentry(pp, &head->chain)
7046 if ((pp->port == rover) &&
7047 net_eq(sock_net(sk), pp->net))
7051 spin_unlock(&head->lock);
7052 } while (--remaining > 0);
7054 /* Exhausted local port range during search? */
7059 /* OK, here is the one we will use. HEAD (the port
7060 * hash table list entry) is non-NULL and we hold it's
7065 /* We are given an specific port number; we verify
7066 * that it is not being used. If it is used, we will
7067 * exahust the search in the hash list corresponding
7068 * to the port number (snum) - we detect that with the
7069 * port iterator, pp being NULL.
7071 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
7072 spin_lock(&head->lock);
7073 sctp_for_each_hentry(pp, &head->chain) {
7074 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
7081 if (!hlist_empty(&pp->owner)) {
7082 /* We had a port hash table hit - there is an
7083 * available port (pp != NULL) and it is being
7084 * used by other socket (pp->owner not empty); that other
7085 * socket is going to be sk2.
7087 int reuse = sk->sk_reuse;
7090 pr_debug("%s: found a possible match\n", __func__);
7092 if (pp->fastreuse && sk->sk_reuse &&
7093 sk->sk_state != SCTP_SS_LISTENING)
7096 /* Run through the list of sockets bound to the port
7097 * (pp->port) [via the pointers bind_next and
7098 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
7099 * we get the endpoint they describe and run through
7100 * the endpoint's list of IP (v4 or v6) addresses,
7101 * comparing each of the addresses with the address of
7102 * the socket sk. If we find a match, then that means
7103 * that this port/socket (sk) combination are already
7106 sk_for_each_bound(sk2, &pp->owner) {
7107 struct sctp_endpoint *ep2;
7108 ep2 = sctp_sk(sk2)->ep;
7111 (reuse && sk2->sk_reuse &&
7112 sk2->sk_state != SCTP_SS_LISTENING))
7115 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
7116 sctp_sk(sk2), sctp_sk(sk))) {
7122 pr_debug("%s: found a match\n", __func__);
7125 /* If there was a hash table miss, create a new port. */
7127 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
7130 /* In either case (hit or miss), make sure fastreuse is 1 only
7131 * if sk->sk_reuse is too (that is, if the caller requested
7132 * SO_REUSEADDR on this socket -sk-).
7134 if (hlist_empty(&pp->owner)) {
7135 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
7139 } else if (pp->fastreuse &&
7140 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
7143 /* We are set, so fill up all the data in the hash table
7144 * entry, tie the socket list information with the rest of the
7145 * sockets FIXME: Blurry, NPI (ipg).
7148 if (!sctp_sk(sk)->bind_hash) {
7149 inet_sk(sk)->inet_num = snum;
7150 sk_add_bind_node(sk, &pp->owner);
7151 sctp_sk(sk)->bind_hash = pp;
7156 spin_unlock(&head->lock);
7163 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
7164 * port is requested.
7166 static int sctp_get_port(struct sock *sk, unsigned short snum)
7168 union sctp_addr addr;
7169 struct sctp_af *af = sctp_sk(sk)->pf->af;
7171 /* Set up a dummy address struct from the sk. */
7172 af->from_sk(&addr, sk);
7173 addr.v4.sin_port = htons(snum);
7175 /* Note: sk->sk_num gets filled in if ephemeral port request. */
7176 return !!sctp_get_port_local(sk, &addr);
7180 * Move a socket to LISTENING state.
7182 static int sctp_listen_start(struct sock *sk, int backlog)
7184 struct sctp_sock *sp = sctp_sk(sk);
7185 struct sctp_endpoint *ep = sp->ep;
7186 struct crypto_shash *tfm = NULL;
7189 /* Allocate HMAC for generating cookie. */
7190 if (!sp->hmac && sp->sctp_hmac_alg) {
7191 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
7192 tfm = crypto_alloc_shash(alg, 0, 0);
7194 net_info_ratelimited("failed to load transform for %s: %ld\n",
7195 sp->sctp_hmac_alg, PTR_ERR(tfm));
7198 sctp_sk(sk)->hmac = tfm;
7202 * If a bind() or sctp_bindx() is not called prior to a listen()
7203 * call that allows new associations to be accepted, the system
7204 * picks an ephemeral port and will choose an address set equivalent
7205 * to binding with a wildcard address.
7207 * This is not currently spelled out in the SCTP sockets
7208 * extensions draft, but follows the practice as seen in TCP
7212 sk->sk_state = SCTP_SS_LISTENING;
7213 if (!ep->base.bind_addr.port) {
7214 if (sctp_autobind(sk))
7217 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
7218 sk->sk_state = SCTP_SS_CLOSED;
7223 sk->sk_max_ack_backlog = backlog;
7224 sctp_hash_endpoint(ep);
7229 * 4.1.3 / 5.1.3 listen()
7231 * By default, new associations are not accepted for UDP style sockets.
7232 * An application uses listen() to mark a socket as being able to
7233 * accept new associations.
7235 * On TCP style sockets, applications use listen() to ready the SCTP
7236 * endpoint for accepting inbound associations.
7238 * On both types of endpoints a backlog of '0' disables listening.
7240 * Move a socket to LISTENING state.
7242 int sctp_inet_listen(struct socket *sock, int backlog)
7244 struct sock *sk = sock->sk;
7245 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
7248 if (unlikely(backlog < 0))
7253 /* Peeled-off sockets are not allowed to listen(). */
7254 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
7257 if (sock->state != SS_UNCONNECTED)
7260 if (!sctp_sstate(sk, LISTENING) && !sctp_sstate(sk, CLOSED))
7263 /* If backlog is zero, disable listening. */
7265 if (sctp_sstate(sk, CLOSED))
7269 sctp_unhash_endpoint(ep);
7270 sk->sk_state = SCTP_SS_CLOSED;
7272 sctp_sk(sk)->bind_hash->fastreuse = 1;
7276 /* If we are already listening, just update the backlog */
7277 if (sctp_sstate(sk, LISTENING))
7278 sk->sk_max_ack_backlog = backlog;
7280 err = sctp_listen_start(sk, backlog);
7292 * This function is done by modeling the current datagram_poll() and the
7293 * tcp_poll(). Note that, based on these implementations, we don't
7294 * lock the socket in this function, even though it seems that,
7295 * ideally, locking or some other mechanisms can be used to ensure
7296 * the integrity of the counters (sndbuf and wmem_alloc) used
7297 * in this place. We assume that we don't need locks either until proven
7300 * Another thing to note is that we include the Async I/O support
7301 * here, again, by modeling the current TCP/UDP code. We don't have
7302 * a good way to test with it yet.
7304 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
7306 struct sock *sk = sock->sk;
7307 struct sctp_sock *sp = sctp_sk(sk);
7310 poll_wait(file, sk_sleep(sk), wait);
7312 sock_rps_record_flow(sk);
7314 /* A TCP-style listening socket becomes readable when the accept queue
7317 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
7318 return (!list_empty(&sp->ep->asocs)) ?
7319 (POLLIN | POLLRDNORM) : 0;
7323 /* Is there any exceptional events? */
7324 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
7326 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
7327 if (sk->sk_shutdown & RCV_SHUTDOWN)
7328 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
7329 if (sk->sk_shutdown == SHUTDOWN_MASK)
7332 /* Is it readable? Reconsider this code with TCP-style support. */
7333 if (!skb_queue_empty(&sk->sk_receive_queue))
7334 mask |= POLLIN | POLLRDNORM;
7336 /* The association is either gone or not ready. */
7337 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
7340 /* Is it writable? */
7341 if (sctp_writeable(sk)) {
7342 mask |= POLLOUT | POLLWRNORM;
7344 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
7346 * Since the socket is not locked, the buffer
7347 * might be made available after the writeable check and
7348 * before the bit is set. This could cause a lost I/O
7349 * signal. tcp_poll() has a race breaker for this race
7350 * condition. Based on their implementation, we put
7351 * in the following code to cover it as well.
7353 if (sctp_writeable(sk))
7354 mask |= POLLOUT | POLLWRNORM;
7359 /********************************************************************
7360 * 2nd Level Abstractions
7361 ********************************************************************/
7363 static struct sctp_bind_bucket *sctp_bucket_create(
7364 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
7366 struct sctp_bind_bucket *pp;
7368 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
7370 SCTP_DBG_OBJCNT_INC(bind_bucket);
7373 INIT_HLIST_HEAD(&pp->owner);
7375 hlist_add_head(&pp->node, &head->chain);
7380 /* Caller must hold hashbucket lock for this tb with local BH disabled */
7381 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
7383 if (pp && hlist_empty(&pp->owner)) {
7384 __hlist_del(&pp->node);
7385 kmem_cache_free(sctp_bucket_cachep, pp);
7386 SCTP_DBG_OBJCNT_DEC(bind_bucket);
7390 /* Release this socket's reference to a local port. */
7391 static inline void __sctp_put_port(struct sock *sk)
7393 struct sctp_bind_hashbucket *head =
7394 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
7395 inet_sk(sk)->inet_num)];
7396 struct sctp_bind_bucket *pp;
7398 spin_lock(&head->lock);
7399 pp = sctp_sk(sk)->bind_hash;
7400 __sk_del_bind_node(sk);
7401 sctp_sk(sk)->bind_hash = NULL;
7402 inet_sk(sk)->inet_num = 0;
7403 sctp_bucket_destroy(pp);
7404 spin_unlock(&head->lock);
7407 void sctp_put_port(struct sock *sk)
7410 __sctp_put_port(sk);
7415 * The system picks an ephemeral port and choose an address set equivalent
7416 * to binding with a wildcard address.
7417 * One of those addresses will be the primary address for the association.
7418 * This automatically enables the multihoming capability of SCTP.
7420 static int sctp_autobind(struct sock *sk)
7422 union sctp_addr autoaddr;
7426 /* Initialize a local sockaddr structure to INADDR_ANY. */
7427 af = sctp_sk(sk)->pf->af;
7429 port = htons(inet_sk(sk)->inet_num);
7430 af->inaddr_any(&autoaddr, port);
7432 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
7435 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
7438 * 4.2 The cmsghdr Structure *
7440 * When ancillary data is sent or received, any number of ancillary data
7441 * objects can be specified by the msg_control and msg_controllen members of
7442 * the msghdr structure, because each object is preceded by
7443 * a cmsghdr structure defining the object's length (the cmsg_len member).
7444 * Historically Berkeley-derived implementations have passed only one object
7445 * at a time, but this API allows multiple objects to be
7446 * passed in a single call to sendmsg() or recvmsg(). The following example
7447 * shows two ancillary data objects in a control buffer.
7449 * |<--------------------------- msg_controllen -------------------------->|
7452 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
7454 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
7457 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
7459 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
7462 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7463 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
7465 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
7467 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7474 static int sctp_msghdr_parse(const struct msghdr *msg, struct sctp_cmsgs *cmsgs)
7476 struct msghdr *my_msg = (struct msghdr *)msg;
7477 struct cmsghdr *cmsg;
7479 for_each_cmsghdr(cmsg, my_msg) {
7480 if (!CMSG_OK(my_msg, cmsg))
7483 /* Should we parse this header or ignore? */
7484 if (cmsg->cmsg_level != IPPROTO_SCTP)
7487 /* Strictly check lengths following example in SCM code. */
7488 switch (cmsg->cmsg_type) {
7490 /* SCTP Socket API Extension
7491 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
7493 * This cmsghdr structure provides information for
7494 * initializing new SCTP associations with sendmsg().
7495 * The SCTP_INITMSG socket option uses this same data
7496 * structure. This structure is not used for
7499 * cmsg_level cmsg_type cmsg_data[]
7500 * ------------ ------------ ----------------------
7501 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
7503 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
7506 cmsgs->init = CMSG_DATA(cmsg);
7510 /* SCTP Socket API Extension
7511 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
7513 * This cmsghdr structure specifies SCTP options for
7514 * sendmsg() and describes SCTP header information
7515 * about a received message through recvmsg().
7517 * cmsg_level cmsg_type cmsg_data[]
7518 * ------------ ------------ ----------------------
7519 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
7521 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
7524 cmsgs->srinfo = CMSG_DATA(cmsg);
7526 if (cmsgs->srinfo->sinfo_flags &
7527 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7528 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7529 SCTP_ABORT | SCTP_EOF))
7534 /* SCTP Socket API Extension
7535 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
7537 * This cmsghdr structure specifies SCTP options for
7538 * sendmsg(). This structure and SCTP_RCVINFO replaces
7539 * SCTP_SNDRCV which has been deprecated.
7541 * cmsg_level cmsg_type cmsg_data[]
7542 * ------------ ------------ ---------------------
7543 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
7545 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
7548 cmsgs->sinfo = CMSG_DATA(cmsg);
7550 if (cmsgs->sinfo->snd_flags &
7551 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7552 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7553 SCTP_ABORT | SCTP_EOF))
7565 * Wait for a packet..
7566 * Note: This function is the same function as in core/datagram.c
7567 * with a few modifications to make lksctp work.
7569 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
7574 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7576 /* Socket errors? */
7577 error = sock_error(sk);
7581 if (!skb_queue_empty(&sk->sk_receive_queue))
7584 /* Socket shut down? */
7585 if (sk->sk_shutdown & RCV_SHUTDOWN)
7588 /* Sequenced packets can come disconnected. If so we report the
7593 /* Is there a good reason to think that we may receive some data? */
7594 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
7597 /* Handle signals. */
7598 if (signal_pending(current))
7601 /* Let another process have a go. Since we are going to sleep
7602 * anyway. Note: This may cause odd behaviors if the message
7603 * does not fit in the user's buffer, but this seems to be the
7604 * only way to honor MSG_DONTWAIT realistically.
7607 *timeo_p = schedule_timeout(*timeo_p);
7611 finish_wait(sk_sleep(sk), &wait);
7615 error = sock_intr_errno(*timeo_p);
7618 finish_wait(sk_sleep(sk), &wait);
7623 /* Receive a datagram.
7624 * Note: This is pretty much the same routine as in core/datagram.c
7625 * with a few changes to make lksctp work.
7627 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
7628 int noblock, int *err)
7631 struct sk_buff *skb;
7634 timeo = sock_rcvtimeo(sk, noblock);
7636 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
7637 MAX_SCHEDULE_TIMEOUT);
7640 /* Again only user level code calls this function,
7641 * so nothing interrupt level
7642 * will suddenly eat the receive_queue.
7644 * Look at current nfs client by the way...
7645 * However, this function was correct in any case. 8)
7647 if (flags & MSG_PEEK) {
7648 skb = skb_peek(&sk->sk_receive_queue);
7650 refcount_inc(&skb->users);
7652 skb = __skb_dequeue(&sk->sk_receive_queue);
7658 /* Caller is allowed not to check sk->sk_err before calling. */
7659 error = sock_error(sk);
7663 if (sk->sk_shutdown & RCV_SHUTDOWN)
7666 if (sk_can_busy_loop(sk)) {
7667 sk_busy_loop(sk, noblock);
7669 if (!skb_queue_empty(&sk->sk_receive_queue))
7673 /* User doesn't want to wait. */
7677 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
7686 /* If sndbuf has changed, wake up per association sndbuf waiters. */
7687 static void __sctp_write_space(struct sctp_association *asoc)
7689 struct sock *sk = asoc->base.sk;
7691 if (sctp_wspace(asoc) <= 0)
7694 if (waitqueue_active(&asoc->wait))
7695 wake_up_interruptible(&asoc->wait);
7697 if (sctp_writeable(sk)) {
7698 struct socket_wq *wq;
7701 wq = rcu_dereference(sk->sk_wq);
7703 if (waitqueue_active(&wq->wait))
7704 wake_up_interruptible(&wq->wait);
7706 /* Note that we try to include the Async I/O support
7707 * here by modeling from the current TCP/UDP code.
7708 * We have not tested with it yet.
7710 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
7711 sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT);
7717 static void sctp_wake_up_waiters(struct sock *sk,
7718 struct sctp_association *asoc)
7720 struct sctp_association *tmp = asoc;
7722 /* We do accounting for the sndbuf space per association,
7723 * so we only need to wake our own association.
7725 if (asoc->ep->sndbuf_policy)
7726 return __sctp_write_space(asoc);
7728 /* If association goes down and is just flushing its
7729 * outq, then just normally notify others.
7731 if (asoc->base.dead)
7732 return sctp_write_space(sk);
7734 /* Accounting for the sndbuf space is per socket, so we
7735 * need to wake up others, try to be fair and in case of
7736 * other associations, let them have a go first instead
7737 * of just doing a sctp_write_space() call.
7739 * Note that we reach sctp_wake_up_waiters() only when
7740 * associations free up queued chunks, thus we are under
7741 * lock and the list of associations on a socket is
7742 * guaranteed not to change.
7744 for (tmp = list_next_entry(tmp, asocs); 1;
7745 tmp = list_next_entry(tmp, asocs)) {
7746 /* Manually skip the head element. */
7747 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
7749 /* Wake up association. */
7750 __sctp_write_space(tmp);
7751 /* We've reached the end. */
7757 /* Do accounting for the sndbuf space.
7758 * Decrement the used sndbuf space of the corresponding association by the
7759 * data size which was just transmitted(freed).
7761 static void sctp_wfree(struct sk_buff *skb)
7763 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
7764 struct sctp_association *asoc = chunk->asoc;
7765 struct sock *sk = asoc->base.sk;
7767 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
7768 sizeof(struct sk_buff) +
7769 sizeof(struct sctp_chunk);
7771 WARN_ON(refcount_sub_and_test(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc));
7774 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
7776 sk->sk_wmem_queued -= skb->truesize;
7777 sk_mem_uncharge(sk, skb->truesize);
7780 sctp_wake_up_waiters(sk, asoc);
7782 sctp_association_put(asoc);
7785 /* Do accounting for the receive space on the socket.
7786 * Accounting for the association is done in ulpevent.c
7787 * We set this as a destructor for the cloned data skbs so that
7788 * accounting is done at the correct time.
7790 void sctp_sock_rfree(struct sk_buff *skb)
7792 struct sock *sk = skb->sk;
7793 struct sctp_ulpevent *event = sctp_skb2event(skb);
7795 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
7798 * Mimic the behavior of sock_rfree
7800 sk_mem_uncharge(sk, event->rmem_len);
7804 /* Helper function to wait for space in the sndbuf. */
7805 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
7808 struct sock *sk = asoc->base.sk;
7810 long current_timeo = *timeo_p;
7813 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
7816 /* Increment the association's refcnt. */
7817 sctp_association_hold(asoc);
7819 /* Wait on the association specific sndbuf space. */
7821 prepare_to_wait_exclusive(&asoc->wait, &wait,
7822 TASK_INTERRUPTIBLE);
7825 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7828 if (signal_pending(current))
7829 goto do_interrupted;
7830 if (msg_len <= sctp_wspace(asoc))
7833 /* Let another process have a go. Since we are going
7837 current_timeo = schedule_timeout(current_timeo);
7840 *timeo_p = current_timeo;
7844 finish_wait(&asoc->wait, &wait);
7846 /* Release the association's refcnt. */
7847 sctp_association_put(asoc);
7856 err = sock_intr_errno(*timeo_p);
7864 void sctp_data_ready(struct sock *sk)
7866 struct socket_wq *wq;
7869 wq = rcu_dereference(sk->sk_wq);
7870 if (skwq_has_sleeper(wq))
7871 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
7872 POLLRDNORM | POLLRDBAND);
7873 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
7877 /* If socket sndbuf has changed, wake up all per association waiters. */
7878 void sctp_write_space(struct sock *sk)
7880 struct sctp_association *asoc;
7882 /* Wake up the tasks in each wait queue. */
7883 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
7884 __sctp_write_space(asoc);
7888 /* Is there any sndbuf space available on the socket?
7890 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
7891 * associations on the same socket. For a UDP-style socket with
7892 * multiple associations, it is possible for it to be "unwriteable"
7893 * prematurely. I assume that this is acceptable because
7894 * a premature "unwriteable" is better than an accidental "writeable" which
7895 * would cause an unwanted block under certain circumstances. For the 1-1
7896 * UDP-style sockets or TCP-style sockets, this code should work.
7899 static int sctp_writeable(struct sock *sk)
7903 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
7909 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
7910 * returns immediately with EINPROGRESS.
7912 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
7914 struct sock *sk = asoc->base.sk;
7916 long current_timeo = *timeo_p;
7919 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
7921 /* Increment the association's refcnt. */
7922 sctp_association_hold(asoc);
7925 prepare_to_wait_exclusive(&asoc->wait, &wait,
7926 TASK_INTERRUPTIBLE);
7929 if (sk->sk_shutdown & RCV_SHUTDOWN)
7931 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7934 if (signal_pending(current))
7935 goto do_interrupted;
7937 if (sctp_state(asoc, ESTABLISHED))
7940 /* Let another process have a go. Since we are going
7944 current_timeo = schedule_timeout(current_timeo);
7947 *timeo_p = current_timeo;
7951 finish_wait(&asoc->wait, &wait);
7953 /* Release the association's refcnt. */
7954 sctp_association_put(asoc);
7959 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
7962 err = -ECONNREFUSED;
7966 err = sock_intr_errno(*timeo_p);
7974 static int sctp_wait_for_accept(struct sock *sk, long timeo)
7976 struct sctp_endpoint *ep;
7980 ep = sctp_sk(sk)->ep;
7984 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
7985 TASK_INTERRUPTIBLE);
7987 if (list_empty(&ep->asocs)) {
7989 timeo = schedule_timeout(timeo);
7994 if (!sctp_sstate(sk, LISTENING))
7998 if (!list_empty(&ep->asocs))
8001 err = sock_intr_errno(timeo);
8002 if (signal_pending(current))
8010 finish_wait(sk_sleep(sk), &wait);
8015 static void sctp_wait_for_close(struct sock *sk, long timeout)
8020 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
8021 if (list_empty(&sctp_sk(sk)->ep->asocs))
8024 timeout = schedule_timeout(timeout);
8026 } while (!signal_pending(current) && timeout);
8028 finish_wait(sk_sleep(sk), &wait);
8031 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
8033 struct sk_buff *frag;
8038 /* Don't forget the fragments. */
8039 skb_walk_frags(skb, frag)
8040 sctp_skb_set_owner_r_frag(frag, sk);
8043 sctp_skb_set_owner_r(skb, sk);
8046 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
8047 struct sctp_association *asoc)
8049 struct inet_sock *inet = inet_sk(sk);
8050 struct inet_sock *newinet;
8052 newsk->sk_type = sk->sk_type;
8053 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
8054 newsk->sk_flags = sk->sk_flags;
8055 newsk->sk_tsflags = sk->sk_tsflags;
8056 newsk->sk_no_check_tx = sk->sk_no_check_tx;
8057 newsk->sk_no_check_rx = sk->sk_no_check_rx;
8058 newsk->sk_reuse = sk->sk_reuse;
8060 newsk->sk_shutdown = sk->sk_shutdown;
8061 newsk->sk_destruct = sctp_destruct_sock;
8062 newsk->sk_family = sk->sk_family;
8063 newsk->sk_protocol = IPPROTO_SCTP;
8064 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
8065 newsk->sk_sndbuf = sk->sk_sndbuf;
8066 newsk->sk_rcvbuf = sk->sk_rcvbuf;
8067 newsk->sk_lingertime = sk->sk_lingertime;
8068 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
8069 newsk->sk_sndtimeo = sk->sk_sndtimeo;
8070 newsk->sk_rxhash = sk->sk_rxhash;
8072 newinet = inet_sk(newsk);
8074 /* Initialize sk's sport, dport, rcv_saddr and daddr for
8075 * getsockname() and getpeername()
8077 newinet->inet_sport = inet->inet_sport;
8078 newinet->inet_saddr = inet->inet_saddr;
8079 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
8080 newinet->inet_dport = htons(asoc->peer.port);
8081 newinet->pmtudisc = inet->pmtudisc;
8082 newinet->inet_id = asoc->next_tsn ^ jiffies;
8084 newinet->uc_ttl = inet->uc_ttl;
8085 newinet->mc_loop = 1;
8086 newinet->mc_ttl = 1;
8087 newinet->mc_index = 0;
8088 newinet->mc_list = NULL;
8090 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
8091 net_enable_timestamp();
8093 security_sk_clone(sk, newsk);
8096 static inline void sctp_copy_descendant(struct sock *sk_to,
8097 const struct sock *sk_from)
8099 int ancestor_size = sizeof(struct inet_sock) +
8100 sizeof(struct sctp_sock) -
8101 offsetof(struct sctp_sock, auto_asconf_list);
8103 if (sk_from->sk_family == PF_INET6)
8104 ancestor_size += sizeof(struct ipv6_pinfo);
8106 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
8109 /* Populate the fields of the newsk from the oldsk and migrate the assoc
8110 * and its messages to the newsk.
8112 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
8113 struct sctp_association *assoc,
8114 enum sctp_socket_type type)
8116 struct sctp_sock *oldsp = sctp_sk(oldsk);
8117 struct sctp_sock *newsp = sctp_sk(newsk);
8118 struct sctp_bind_bucket *pp; /* hash list port iterator */
8119 struct sctp_endpoint *newep = newsp->ep;
8120 struct sk_buff *skb, *tmp;
8121 struct sctp_ulpevent *event;
8122 struct sctp_bind_hashbucket *head;
8124 /* Migrate socket buffer sizes and all the socket level options to the
8127 newsk->sk_sndbuf = oldsk->sk_sndbuf;
8128 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
8129 /* Brute force copy old sctp opt. */
8130 sctp_copy_descendant(newsk, oldsk);
8132 /* Restore the ep value that was overwritten with the above structure
8138 /* Hook this new socket in to the bind_hash list. */
8139 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
8140 inet_sk(oldsk)->inet_num)];
8141 spin_lock_bh(&head->lock);
8142 pp = sctp_sk(oldsk)->bind_hash;
8143 sk_add_bind_node(newsk, &pp->owner);
8144 sctp_sk(newsk)->bind_hash = pp;
8145 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
8146 spin_unlock_bh(&head->lock);
8148 /* Copy the bind_addr list from the original endpoint to the new
8149 * endpoint so that we can handle restarts properly
8151 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
8152 &oldsp->ep->base.bind_addr, GFP_KERNEL);
8154 /* Move any messages in the old socket's receive queue that are for the
8155 * peeled off association to the new socket's receive queue.
8157 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
8158 event = sctp_skb2event(skb);
8159 if (event->asoc == assoc) {
8160 __skb_unlink(skb, &oldsk->sk_receive_queue);
8161 __skb_queue_tail(&newsk->sk_receive_queue, skb);
8162 sctp_skb_set_owner_r_frag(skb, newsk);
8166 /* Clean up any messages pending delivery due to partial
8167 * delivery. Three cases:
8168 * 1) No partial deliver; no work.
8169 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
8170 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
8172 skb_queue_head_init(&newsp->pd_lobby);
8173 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
8175 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
8176 struct sk_buff_head *queue;
8178 /* Decide which queue to move pd_lobby skbs to. */
8179 if (assoc->ulpq.pd_mode) {
8180 queue = &newsp->pd_lobby;
8182 queue = &newsk->sk_receive_queue;
8184 /* Walk through the pd_lobby, looking for skbs that
8185 * need moved to the new socket.
8187 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
8188 event = sctp_skb2event(skb);
8189 if (event->asoc == assoc) {
8190 __skb_unlink(skb, &oldsp->pd_lobby);
8191 __skb_queue_tail(queue, skb);
8192 sctp_skb_set_owner_r_frag(skb, newsk);
8196 /* Clear up any skbs waiting for the partial
8197 * delivery to finish.
8199 if (assoc->ulpq.pd_mode)
8200 sctp_clear_pd(oldsk, NULL);
8204 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
8205 sctp_skb_set_owner_r_frag(skb, newsk);
8207 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
8208 sctp_skb_set_owner_r_frag(skb, newsk);
8210 /* Set the type of socket to indicate that it is peeled off from the
8211 * original UDP-style socket or created with the accept() call on a
8212 * TCP-style socket..
8216 /* Mark the new socket "in-use" by the user so that any packets
8217 * that may arrive on the association after we've moved it are
8218 * queued to the backlog. This prevents a potential race between
8219 * backlog processing on the old socket and new-packet processing
8220 * on the new socket.
8222 * The caller has just allocated newsk so we can guarantee that other
8223 * paths won't try to lock it and then oldsk.
8225 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
8226 sctp_assoc_migrate(assoc, newsk);
8228 /* If the association on the newsk is already closed before accept()
8229 * is called, set RCV_SHUTDOWN flag.
8231 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP)) {
8232 newsk->sk_state = SCTP_SS_CLOSED;
8233 newsk->sk_shutdown |= RCV_SHUTDOWN;
8235 newsk->sk_state = SCTP_SS_ESTABLISHED;
8238 release_sock(newsk);
8242 /* This proto struct describes the ULP interface for SCTP. */
8243 struct proto sctp_prot = {
8245 .owner = THIS_MODULE,
8246 .close = sctp_close,
8247 .connect = sctp_connect,
8248 .disconnect = sctp_disconnect,
8249 .accept = sctp_accept,
8250 .ioctl = sctp_ioctl,
8251 .init = sctp_init_sock,
8252 .destroy = sctp_destroy_sock,
8253 .shutdown = sctp_shutdown,
8254 .setsockopt = sctp_setsockopt,
8255 .getsockopt = sctp_getsockopt,
8256 .sendmsg = sctp_sendmsg,
8257 .recvmsg = sctp_recvmsg,
8259 .backlog_rcv = sctp_backlog_rcv,
8261 .unhash = sctp_unhash,
8262 .get_port = sctp_get_port,
8263 .obj_size = sizeof(struct sctp_sock),
8264 .sysctl_mem = sysctl_sctp_mem,
8265 .sysctl_rmem = sysctl_sctp_rmem,
8266 .sysctl_wmem = sysctl_sctp_wmem,
8267 .memory_pressure = &sctp_memory_pressure,
8268 .enter_memory_pressure = sctp_enter_memory_pressure,
8269 .memory_allocated = &sctp_memory_allocated,
8270 .sockets_allocated = &sctp_sockets_allocated,
8273 #if IS_ENABLED(CONFIG_IPV6)
8275 #include <net/transp_v6.h>
8276 static void sctp_v6_destroy_sock(struct sock *sk)
8278 sctp_destroy_sock(sk);
8279 inet6_destroy_sock(sk);
8282 struct proto sctpv6_prot = {
8284 .owner = THIS_MODULE,
8285 .close = sctp_close,
8286 .connect = sctp_connect,
8287 .disconnect = sctp_disconnect,
8288 .accept = sctp_accept,
8289 .ioctl = sctp_ioctl,
8290 .init = sctp_init_sock,
8291 .destroy = sctp_v6_destroy_sock,
8292 .shutdown = sctp_shutdown,
8293 .setsockopt = sctp_setsockopt,
8294 .getsockopt = sctp_getsockopt,
8295 .sendmsg = sctp_sendmsg,
8296 .recvmsg = sctp_recvmsg,
8298 .backlog_rcv = sctp_backlog_rcv,
8300 .unhash = sctp_unhash,
8301 .get_port = sctp_get_port,
8302 .obj_size = sizeof(struct sctp6_sock),
8303 .sysctl_mem = sysctl_sctp_mem,
8304 .sysctl_rmem = sysctl_sctp_rmem,
8305 .sysctl_wmem = sysctl_sctp_wmem,
8306 .memory_pressure = &sctp_memory_pressure,
8307 .enter_memory_pressure = sctp_enter_memory_pressure,
8308 .memory_allocated = &sctp_memory_allocated,
8309 .sockets_allocated = &sctp_sockets_allocated,
8311 #endif /* IS_ENABLED(CONFIG_IPV6) */