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>
82 #include <net/sctp/stream_sched.h>
84 /* Forward declarations for internal helper functions. */
85 static int sctp_writeable(struct sock *sk);
86 static void sctp_wfree(struct sk_buff *skb);
87 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
88 size_t msg_len, struct sock **orig_sk);
89 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p);
90 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
91 static int sctp_wait_for_accept(struct sock *sk, long timeo);
92 static void sctp_wait_for_close(struct sock *sk, long timeo);
93 static void sctp_destruct_sock(struct sock *sk);
94 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
95 union sctp_addr *addr, int len);
96 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
97 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
98 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
99 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
100 static int sctp_send_asconf(struct sctp_association *asoc,
101 struct sctp_chunk *chunk);
102 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
103 static int sctp_autobind(struct sock *sk);
104 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
105 struct sctp_association *assoc,
106 enum sctp_socket_type type);
108 static unsigned long sctp_memory_pressure;
109 static atomic_long_t sctp_memory_allocated;
110 struct percpu_counter sctp_sockets_allocated;
112 static void sctp_enter_memory_pressure(struct sock *sk)
114 sctp_memory_pressure = 1;
118 /* Get the sndbuf space available at the time on the association. */
119 static inline int sctp_wspace(struct sctp_association *asoc)
123 if (asoc->ep->sndbuf_policy)
124 amt = asoc->sndbuf_used;
126 amt = sk_wmem_alloc_get(asoc->base.sk);
128 if (amt >= asoc->base.sk->sk_sndbuf) {
129 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
132 amt = sk_stream_wspace(asoc->base.sk);
137 amt = asoc->base.sk->sk_sndbuf - amt;
142 /* Increment the used sndbuf space count of the corresponding association by
143 * the size of the outgoing data chunk.
144 * Also, set the skb destructor for sndbuf accounting later.
146 * Since it is always 1-1 between chunk and skb, and also a new skb is always
147 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
148 * destructor in the data chunk skb for the purpose of the sndbuf space
151 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
153 struct sctp_association *asoc = chunk->asoc;
154 struct sock *sk = asoc->base.sk;
156 /* The sndbuf space is tracked per association. */
157 sctp_association_hold(asoc);
159 skb_set_owner_w(chunk->skb, sk);
161 chunk->skb->destructor = sctp_wfree;
162 /* Save the chunk pointer in skb for sctp_wfree to use later. */
163 skb_shinfo(chunk->skb)->destructor_arg = chunk;
165 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
166 sizeof(struct sk_buff) +
167 sizeof(struct sctp_chunk);
169 refcount_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
170 sk->sk_wmem_queued += chunk->skb->truesize;
171 sk_mem_charge(sk, chunk->skb->truesize);
174 static void sctp_clear_owner_w(struct sctp_chunk *chunk)
176 skb_orphan(chunk->skb);
179 static void sctp_for_each_tx_datachunk(struct sctp_association *asoc,
180 void (*cb)(struct sctp_chunk *))
183 struct sctp_outq *q = &asoc->outqueue;
184 struct sctp_transport *t;
185 struct sctp_chunk *chunk;
187 list_for_each_entry(t, &asoc->peer.transport_addr_list, transports)
188 list_for_each_entry(chunk, &t->transmitted, transmitted_list)
191 list_for_each_entry(chunk, &q->retransmit, transmitted_list)
194 list_for_each_entry(chunk, &q->sacked, transmitted_list)
197 list_for_each_entry(chunk, &q->abandoned, transmitted_list)
200 list_for_each_entry(chunk, &q->out_chunk_list, list)
204 /* Verify that this is a valid address. */
205 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
210 /* Verify basic sockaddr. */
211 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
215 /* Is this a valid SCTP address? */
216 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
219 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
225 /* Look up the association by its id. If this is not a UDP-style
226 * socket, the ID field is always ignored.
228 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
230 struct sctp_association *asoc = NULL;
232 /* If this is not a UDP-style socket, assoc id should be ignored. */
233 if (!sctp_style(sk, UDP)) {
234 /* Return NULL if the socket state is not ESTABLISHED. It
235 * could be a TCP-style listening socket or a socket which
236 * hasn't yet called connect() to establish an association.
238 if (!sctp_sstate(sk, ESTABLISHED) && !sctp_sstate(sk, CLOSING))
241 /* Get the first and the only association from the list. */
242 if (!list_empty(&sctp_sk(sk)->ep->asocs))
243 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
244 struct sctp_association, asocs);
248 /* Otherwise this is a UDP-style socket. */
249 if (!id || (id == (sctp_assoc_t)-1))
252 spin_lock_bh(&sctp_assocs_id_lock);
253 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
254 spin_unlock_bh(&sctp_assocs_id_lock);
256 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
262 /* Look up the transport from an address and an assoc id. If both address and
263 * id are specified, the associations matching the address and the id should be
266 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
267 struct sockaddr_storage *addr,
270 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
271 struct sctp_af *af = sctp_get_af_specific(addr->ss_family);
272 union sctp_addr *laddr = (union sctp_addr *)addr;
273 struct sctp_transport *transport;
275 if (!af || sctp_verify_addr(sk, laddr, af->sockaddr_len))
278 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
285 id_asoc = sctp_id2assoc(sk, id);
286 if (id_asoc && (id_asoc != addr_asoc))
289 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
290 (union sctp_addr *)addr);
295 /* API 3.1.2 bind() - UDP Style Syntax
296 * The syntax of bind() is,
298 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
300 * sd - the socket descriptor returned by socket().
301 * addr - the address structure (struct sockaddr_in or struct
302 * sockaddr_in6 [RFC 2553]),
303 * addr_len - the size of the address structure.
305 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
311 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
314 /* Disallow binding twice. */
315 if (!sctp_sk(sk)->ep->base.bind_addr.port)
316 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
326 static long sctp_get_port_local(struct sock *, union sctp_addr *);
328 /* Verify this is a valid sockaddr. */
329 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
330 union sctp_addr *addr, int len)
334 /* Check minimum size. */
335 if (len < sizeof (struct sockaddr))
338 /* V4 mapped address are really of AF_INET family */
339 if (addr->sa.sa_family == AF_INET6 &&
340 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
341 if (!opt->pf->af_supported(AF_INET, opt))
344 /* Does this PF support this AF? */
345 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
349 /* If we get this far, af is valid. */
350 af = sctp_get_af_specific(addr->sa.sa_family);
352 if (len < af->sockaddr_len)
358 /* Bind a local address either to an endpoint or to an association. */
359 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
361 struct net *net = sock_net(sk);
362 struct sctp_sock *sp = sctp_sk(sk);
363 struct sctp_endpoint *ep = sp->ep;
364 struct sctp_bind_addr *bp = &ep->base.bind_addr;
369 /* Common sockaddr verification. */
370 af = sctp_sockaddr_af(sp, addr, len);
372 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
373 __func__, sk, addr, len);
377 snum = ntohs(addr->v4.sin_port);
379 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
380 __func__, sk, &addr->sa, bp->port, snum, len);
382 /* PF specific bind() address verification. */
383 if (!sp->pf->bind_verify(sp, addr))
384 return -EADDRNOTAVAIL;
386 /* We must either be unbound, or bind to the same port.
387 * It's OK to allow 0 ports if we are already bound.
388 * We'll just inhert an already bound port in this case
393 else if (snum != bp->port) {
394 pr_debug("%s: new port %d doesn't match existing port "
395 "%d\n", __func__, snum, bp->port);
400 if (snum && snum < inet_prot_sock(net) &&
401 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
404 /* See if the address matches any of the addresses we may have
405 * already bound before checking against other endpoints.
407 if (sctp_bind_addr_match(bp, addr, sp))
410 /* Make sure we are allowed to bind here.
411 * The function sctp_get_port_local() does duplicate address
414 addr->v4.sin_port = htons(snum);
415 if ((ret = sctp_get_port_local(sk, addr))) {
419 /* Refresh ephemeral port. */
421 bp->port = inet_sk(sk)->inet_num;
423 /* Add the address to the bind address list.
424 * Use GFP_ATOMIC since BHs will be disabled.
426 ret = sctp_add_bind_addr(bp, addr, af->sockaddr_len,
427 SCTP_ADDR_SRC, GFP_ATOMIC);
429 /* Copy back into socket for getsockname() use. */
431 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
432 sp->pf->to_sk_saddr(addr, sk);
438 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
440 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
441 * at any one time. If a sender, after sending an ASCONF chunk, decides
442 * it needs to transfer another ASCONF Chunk, it MUST wait until the
443 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
444 * subsequent ASCONF. Note this restriction binds each side, so at any
445 * time two ASCONF may be in-transit on any given association (one sent
446 * from each endpoint).
448 static int sctp_send_asconf(struct sctp_association *asoc,
449 struct sctp_chunk *chunk)
451 struct net *net = sock_net(asoc->base.sk);
454 /* If there is an outstanding ASCONF chunk, queue it for later
457 if (asoc->addip_last_asconf) {
458 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
462 /* Hold the chunk until an ASCONF_ACK is received. */
463 sctp_chunk_hold(chunk);
464 retval = sctp_primitive_ASCONF(net, asoc, chunk);
466 sctp_chunk_free(chunk);
468 asoc->addip_last_asconf = chunk;
474 /* Add a list of addresses as bind addresses to local endpoint or
477 * Basically run through each address specified in the addrs/addrcnt
478 * array/length pair, determine if it is IPv6 or IPv4 and call
479 * sctp_do_bind() on it.
481 * If any of them fails, then the operation will be reversed and the
482 * ones that were added will be removed.
484 * Only sctp_setsockopt_bindx() is supposed to call this function.
486 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
491 struct sockaddr *sa_addr;
494 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
498 for (cnt = 0; cnt < addrcnt; cnt++) {
499 /* The list may contain either IPv4 or IPv6 address;
500 * determine the address length for walking thru the list.
503 af = sctp_get_af_specific(sa_addr->sa_family);
509 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
512 addr_buf += af->sockaddr_len;
516 /* Failed. Cleanup the ones that have been added */
518 sctp_bindx_rem(sk, addrs, cnt);
526 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
527 * associations that are part of the endpoint indicating that a list of local
528 * addresses are added to the endpoint.
530 * If any of the addresses is already in the bind address list of the
531 * association, we do not send the chunk for that association. But it will not
532 * affect other associations.
534 * Only sctp_setsockopt_bindx() is supposed to call this function.
536 static int sctp_send_asconf_add_ip(struct sock *sk,
537 struct sockaddr *addrs,
540 struct net *net = sock_net(sk);
541 struct sctp_sock *sp;
542 struct sctp_endpoint *ep;
543 struct sctp_association *asoc;
544 struct sctp_bind_addr *bp;
545 struct sctp_chunk *chunk;
546 struct sctp_sockaddr_entry *laddr;
547 union sctp_addr *addr;
548 union sctp_addr saveaddr;
555 if (!net->sctp.addip_enable)
561 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
562 __func__, sk, addrs, addrcnt);
564 list_for_each_entry(asoc, &ep->asocs, asocs) {
565 if (!asoc->peer.asconf_capable)
568 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
571 if (!sctp_state(asoc, ESTABLISHED))
574 /* Check if any address in the packed array of addresses is
575 * in the bind address list of the association. If so,
576 * do not send the asconf chunk to its peer, but continue with
577 * other associations.
580 for (i = 0; i < addrcnt; i++) {
582 af = sctp_get_af_specific(addr->v4.sin_family);
588 if (sctp_assoc_lookup_laddr(asoc, addr))
591 addr_buf += af->sockaddr_len;
596 /* Use the first valid address in bind addr list of
597 * association as Address Parameter of ASCONF CHUNK.
599 bp = &asoc->base.bind_addr;
600 p = bp->address_list.next;
601 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
602 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
603 addrcnt, SCTP_PARAM_ADD_IP);
609 /* Add the new addresses to the bind address list with
610 * use_as_src set to 0.
613 for (i = 0; i < addrcnt; i++) {
615 af = sctp_get_af_specific(addr->v4.sin_family);
616 memcpy(&saveaddr, addr, af->sockaddr_len);
617 retval = sctp_add_bind_addr(bp, &saveaddr,
619 SCTP_ADDR_NEW, GFP_ATOMIC);
620 addr_buf += af->sockaddr_len;
622 if (asoc->src_out_of_asoc_ok) {
623 struct sctp_transport *trans;
625 list_for_each_entry(trans,
626 &asoc->peer.transport_addr_list, transports) {
627 /* Clear the source and route cache */
628 sctp_transport_dst_release(trans);
629 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
630 2*asoc->pathmtu, 4380));
631 trans->ssthresh = asoc->peer.i.a_rwnd;
632 trans->rto = asoc->rto_initial;
633 sctp_max_rto(asoc, trans);
634 trans->rtt = trans->srtt = trans->rttvar = 0;
635 sctp_transport_route(trans, NULL,
636 sctp_sk(asoc->base.sk));
639 retval = sctp_send_asconf(asoc, chunk);
646 /* Remove a list of addresses from bind addresses list. Do not remove the
649 * Basically run through each address specified in the addrs/addrcnt
650 * array/length pair, determine if it is IPv6 or IPv4 and call
651 * sctp_del_bind() on it.
653 * If any of them fails, then the operation will be reversed and the
654 * ones that were removed will be added back.
656 * At least one address has to be left; if only one address is
657 * available, the operation will return -EBUSY.
659 * Only sctp_setsockopt_bindx() is supposed to call this function.
661 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
663 struct sctp_sock *sp = sctp_sk(sk);
664 struct sctp_endpoint *ep = sp->ep;
666 struct sctp_bind_addr *bp = &ep->base.bind_addr;
669 union sctp_addr *sa_addr;
672 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
673 __func__, sk, addrs, addrcnt);
676 for (cnt = 0; cnt < addrcnt; cnt++) {
677 /* If the bind address list is empty or if there is only one
678 * bind address, there is nothing more to be removed (we need
679 * at least one address here).
681 if (list_empty(&bp->address_list) ||
682 (sctp_list_single_entry(&bp->address_list))) {
688 af = sctp_get_af_specific(sa_addr->sa.sa_family);
694 if (!af->addr_valid(sa_addr, sp, NULL)) {
695 retval = -EADDRNOTAVAIL;
699 if (sa_addr->v4.sin_port &&
700 sa_addr->v4.sin_port != htons(bp->port)) {
705 if (!sa_addr->v4.sin_port)
706 sa_addr->v4.sin_port = htons(bp->port);
708 /* FIXME - There is probably a need to check if sk->sk_saddr and
709 * sk->sk_rcv_addr are currently set to one of the addresses to
710 * be removed. This is something which needs to be looked into
711 * when we are fixing the outstanding issues with multi-homing
712 * socket routing and failover schemes. Refer to comments in
713 * sctp_do_bind(). -daisy
715 retval = sctp_del_bind_addr(bp, sa_addr);
717 addr_buf += af->sockaddr_len;
720 /* Failed. Add the ones that has been removed back */
722 sctp_bindx_add(sk, addrs, cnt);
730 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
731 * the associations that are part of the endpoint indicating that a list of
732 * local addresses are removed from the endpoint.
734 * If any of the addresses is already in the bind address list of the
735 * association, we do not send the chunk for that association. But it will not
736 * affect other associations.
738 * Only sctp_setsockopt_bindx() is supposed to call this function.
740 static int sctp_send_asconf_del_ip(struct sock *sk,
741 struct sockaddr *addrs,
744 struct net *net = sock_net(sk);
745 struct sctp_sock *sp;
746 struct sctp_endpoint *ep;
747 struct sctp_association *asoc;
748 struct sctp_transport *transport;
749 struct sctp_bind_addr *bp;
750 struct sctp_chunk *chunk;
751 union sctp_addr *laddr;
754 struct sctp_sockaddr_entry *saddr;
760 if (!net->sctp.addip_enable)
766 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
767 __func__, sk, addrs, addrcnt);
769 list_for_each_entry(asoc, &ep->asocs, asocs) {
771 if (!asoc->peer.asconf_capable)
774 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
777 if (!sctp_state(asoc, ESTABLISHED))
780 /* Check if any address in the packed array of addresses is
781 * not present in the bind address list of the association.
782 * If so, do not send the asconf chunk to its peer, but
783 * continue with other associations.
786 for (i = 0; i < addrcnt; i++) {
788 af = sctp_get_af_specific(laddr->v4.sin_family);
794 if (!sctp_assoc_lookup_laddr(asoc, laddr))
797 addr_buf += af->sockaddr_len;
802 /* Find one address in the association's bind address list
803 * that is not in the packed array of addresses. This is to
804 * make sure that we do not delete all the addresses in the
807 bp = &asoc->base.bind_addr;
808 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
810 if ((laddr == NULL) && (addrcnt == 1)) {
811 if (asoc->asconf_addr_del_pending)
813 asoc->asconf_addr_del_pending =
814 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
815 if (asoc->asconf_addr_del_pending == NULL) {
819 asoc->asconf_addr_del_pending->sa.sa_family =
821 asoc->asconf_addr_del_pending->v4.sin_port =
823 if (addrs->sa_family == AF_INET) {
824 struct sockaddr_in *sin;
826 sin = (struct sockaddr_in *)addrs;
827 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
828 } else if (addrs->sa_family == AF_INET6) {
829 struct sockaddr_in6 *sin6;
831 sin6 = (struct sockaddr_in6 *)addrs;
832 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
835 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
836 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
837 asoc->asconf_addr_del_pending);
839 asoc->src_out_of_asoc_ok = 1;
847 /* We do not need RCU protection throughout this loop
848 * because this is done under a socket lock from the
851 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
859 /* Reset use_as_src flag for the addresses in the bind address
860 * list that are to be deleted.
863 for (i = 0; i < addrcnt; i++) {
865 af = sctp_get_af_specific(laddr->v4.sin_family);
866 list_for_each_entry(saddr, &bp->address_list, list) {
867 if (sctp_cmp_addr_exact(&saddr->a, laddr))
868 saddr->state = SCTP_ADDR_DEL;
870 addr_buf += af->sockaddr_len;
873 /* Update the route and saddr entries for all the transports
874 * as some of the addresses in the bind address list are
875 * about to be deleted and cannot be used as source addresses.
877 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
879 sctp_transport_dst_release(transport);
880 sctp_transport_route(transport, NULL,
881 sctp_sk(asoc->base.sk));
885 /* We don't need to transmit ASCONF */
887 retval = sctp_send_asconf(asoc, chunk);
893 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
894 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
896 struct sock *sk = sctp_opt2sk(sp);
897 union sctp_addr *addr;
900 /* It is safe to write port space in caller. */
902 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
903 af = sctp_get_af_specific(addr->sa.sa_family);
906 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
909 if (addrw->state == SCTP_ADDR_NEW)
910 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
912 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
915 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
918 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
921 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
922 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
925 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
926 * Section 3.1.2 for this usage.
928 * addrs is a pointer to an array of one or more socket addresses. Each
929 * address is contained in its appropriate structure (i.e. struct
930 * sockaddr_in or struct sockaddr_in6) the family of the address type
931 * must be used to distinguish the address length (note that this
932 * representation is termed a "packed array" of addresses). The caller
933 * specifies the number of addresses in the array with addrcnt.
935 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
936 * -1, and sets errno to the appropriate error code.
938 * For SCTP, the port given in each socket address must be the same, or
939 * sctp_bindx() will fail, setting errno to EINVAL.
941 * The flags parameter is formed from the bitwise OR of zero or more of
942 * the following currently defined flags:
944 * SCTP_BINDX_ADD_ADDR
946 * SCTP_BINDX_REM_ADDR
948 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
949 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
950 * addresses from the association. The two flags are mutually exclusive;
951 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
952 * not remove all addresses from an association; sctp_bindx() will
953 * reject such an attempt with EINVAL.
955 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
956 * additional addresses with an endpoint after calling bind(). Or use
957 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
958 * socket is associated with so that no new association accepted will be
959 * associated with those addresses. If the endpoint supports dynamic
960 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
961 * endpoint to send the appropriate message to the peer to change the
962 * peers address lists.
964 * Adding and removing addresses from a connected association is
965 * optional functionality. Implementations that do not support this
966 * functionality should return EOPNOTSUPP.
968 * Basically do nothing but copying the addresses from user to kernel
969 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
970 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
973 * We don't use copy_from_user() for optimization: we first do the
974 * sanity checks (buffer size -fast- and access check-healthy
975 * pointer); if all of those succeed, then we can alloc the memory
976 * (expensive operation) needed to copy the data to kernel. Then we do
977 * the copying without checking the user space area
978 * (__copy_from_user()).
980 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
983 * sk The sk of the socket
984 * addrs The pointer to the addresses in user land
985 * addrssize Size of the addrs buffer
986 * op Operation to perform (add or remove, see the flags of
989 * Returns 0 if ok, <0 errno code on error.
991 static int sctp_setsockopt_bindx(struct sock *sk,
992 struct sockaddr __user *addrs,
993 int addrs_size, int op)
995 struct sockaddr *kaddrs;
999 struct sockaddr *sa_addr;
1003 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
1004 __func__, sk, addrs, addrs_size, op);
1006 if (unlikely(addrs_size <= 0))
1009 /* Check the user passed a healthy pointer. */
1010 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1013 /* Alloc space for the address array in kernel memory. */
1014 kaddrs = kmalloc(addrs_size, GFP_USER | __GFP_NOWARN);
1015 if (unlikely(!kaddrs))
1018 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1023 /* Walk through the addrs buffer and count the number of addresses. */
1025 while (walk_size < addrs_size) {
1026 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1032 af = sctp_get_af_specific(sa_addr->sa_family);
1034 /* If the address family is not supported or if this address
1035 * causes the address buffer to overflow return EINVAL.
1037 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1042 addr_buf += af->sockaddr_len;
1043 walk_size += af->sockaddr_len;
1048 case SCTP_BINDX_ADD_ADDR:
1049 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1052 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1055 case SCTP_BINDX_REM_ADDR:
1056 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1059 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1073 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1075 * Common routine for handling connect() and sctp_connectx().
1076 * Connect will come in with just a single address.
1078 static int __sctp_connect(struct sock *sk,
1079 struct sockaddr *kaddrs,
1081 sctp_assoc_t *assoc_id)
1083 struct net *net = sock_net(sk);
1084 struct sctp_sock *sp;
1085 struct sctp_endpoint *ep;
1086 struct sctp_association *asoc = NULL;
1087 struct sctp_association *asoc2;
1088 struct sctp_transport *transport;
1090 enum sctp_scope scope;
1095 union sctp_addr *sa_addr = NULL;
1097 unsigned short port;
1098 unsigned int f_flags = 0;
1103 /* connect() cannot be done on a socket that is already in ESTABLISHED
1104 * state - UDP-style peeled off socket or a TCP-style socket that
1105 * is already connected.
1106 * It cannot be done even on a TCP-style listening socket.
1108 if (sctp_sstate(sk, ESTABLISHED) || sctp_sstate(sk, CLOSING) ||
1109 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1114 /* Walk through the addrs buffer and count the number of addresses. */
1116 while (walk_size < addrs_size) {
1119 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1125 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1127 /* If the address family is not supported or if this address
1128 * causes the address buffer to overflow return EINVAL.
1130 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1135 port = ntohs(sa_addr->v4.sin_port);
1137 /* Save current address so we can work with it */
1138 memcpy(&to, sa_addr, af->sockaddr_len);
1140 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1144 /* Make sure the destination port is correctly set
1147 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1152 /* Check if there already is a matching association on the
1153 * endpoint (other than the one created here).
1155 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1156 if (asoc2 && asoc2 != asoc) {
1157 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1164 /* If we could not find a matching association on the endpoint,
1165 * make sure that there is no peeled-off association matching
1166 * the peer address even on another socket.
1168 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1169 err = -EADDRNOTAVAIL;
1174 /* If a bind() or sctp_bindx() is not called prior to
1175 * an sctp_connectx() call, the system picks an
1176 * ephemeral port and will choose an address set
1177 * equivalent to binding with a wildcard address.
1179 if (!ep->base.bind_addr.port) {
1180 if (sctp_autobind(sk)) {
1186 * If an unprivileged user inherits a 1-many
1187 * style socket with open associations on a
1188 * privileged port, it MAY be permitted to
1189 * accept new associations, but it SHOULD NOT
1190 * be permitted to open new associations.
1192 if (ep->base.bind_addr.port <
1193 inet_prot_sock(net) &&
1194 !ns_capable(net->user_ns,
1195 CAP_NET_BIND_SERVICE)) {
1201 scope = sctp_scope(&to);
1202 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1208 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1216 /* Prime the peer's transport structures. */
1217 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1225 addr_buf += af->sockaddr_len;
1226 walk_size += af->sockaddr_len;
1229 /* In case the user of sctp_connectx() wants an association
1230 * id back, assign one now.
1233 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1238 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1243 /* Initialize sk's dport and daddr for getpeername() */
1244 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1245 sp->pf->to_sk_daddr(sa_addr, sk);
1248 /* in-kernel sockets don't generally have a file allocated to them
1249 * if all they do is call sock_create_kern().
1251 if (sk->sk_socket->file)
1252 f_flags = sk->sk_socket->file->f_flags;
1254 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1257 *assoc_id = asoc->assoc_id;
1258 err = sctp_wait_for_connect(asoc, &timeo);
1259 /* Note: the asoc may be freed after the return of
1260 * sctp_wait_for_connect.
1263 /* Don't free association on exit. */
1267 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1268 __func__, asoc, kaddrs, err);
1271 /* sctp_primitive_ASSOCIATE may have added this association
1272 * To the hash table, try to unhash it, just in case, its a noop
1273 * if it wasn't hashed so we're safe
1275 sctp_association_free(asoc);
1280 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1283 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1284 * sctp_assoc_t *asoc);
1286 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1287 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1288 * or IPv6 addresses.
1290 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1291 * Section 3.1.2 for this usage.
1293 * addrs is a pointer to an array of one or more socket addresses. Each
1294 * address is contained in its appropriate structure (i.e. struct
1295 * sockaddr_in or struct sockaddr_in6) the family of the address type
1296 * must be used to distengish the address length (note that this
1297 * representation is termed a "packed array" of addresses). The caller
1298 * specifies the number of addresses in the array with addrcnt.
1300 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1301 * the association id of the new association. On failure, sctp_connectx()
1302 * returns -1, and sets errno to the appropriate error code. The assoc_id
1303 * is not touched by the kernel.
1305 * For SCTP, the port given in each socket address must be the same, or
1306 * sctp_connectx() will fail, setting errno to EINVAL.
1308 * An application can use sctp_connectx to initiate an association with
1309 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1310 * allows a caller to specify multiple addresses at which a peer can be
1311 * reached. The way the SCTP stack uses the list of addresses to set up
1312 * the association is implementation dependent. This function only
1313 * specifies that the stack will try to make use of all the addresses in
1314 * the list when needed.
1316 * Note that the list of addresses passed in is only used for setting up
1317 * the association. It does not necessarily equal the set of addresses
1318 * the peer uses for the resulting association. If the caller wants to
1319 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1320 * retrieve them after the association has been set up.
1322 * Basically do nothing but copying the addresses from user to kernel
1323 * land and invoking either sctp_connectx(). This is used for tunneling
1324 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1326 * We don't use copy_from_user() for optimization: we first do the
1327 * sanity checks (buffer size -fast- and access check-healthy
1328 * pointer); if all of those succeed, then we can alloc the memory
1329 * (expensive operation) needed to copy the data to kernel. Then we do
1330 * the copying without checking the user space area
1331 * (__copy_from_user()).
1333 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1336 * sk The sk of the socket
1337 * addrs The pointer to the addresses in user land
1338 * addrssize Size of the addrs buffer
1340 * Returns >=0 if ok, <0 errno code on error.
1342 static int __sctp_setsockopt_connectx(struct sock *sk,
1343 struct sockaddr __user *addrs,
1345 sctp_assoc_t *assoc_id)
1347 struct sockaddr *kaddrs;
1348 gfp_t gfp = GFP_KERNEL;
1351 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1352 __func__, sk, addrs, addrs_size);
1354 if (unlikely(addrs_size <= 0))
1357 /* Check the user passed a healthy pointer. */
1358 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1361 /* Alloc space for the address array in kernel memory. */
1362 if (sk->sk_socket->file)
1363 gfp = GFP_USER | __GFP_NOWARN;
1364 kaddrs = kmalloc(addrs_size, gfp);
1365 if (unlikely(!kaddrs))
1368 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1371 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1380 * This is an older interface. It's kept for backward compatibility
1381 * to the option that doesn't provide association id.
1383 static int sctp_setsockopt_connectx_old(struct sock *sk,
1384 struct sockaddr __user *addrs,
1387 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1391 * New interface for the API. The since the API is done with a socket
1392 * option, to make it simple we feed back the association id is as a return
1393 * indication to the call. Error is always negative and association id is
1396 static int sctp_setsockopt_connectx(struct sock *sk,
1397 struct sockaddr __user *addrs,
1400 sctp_assoc_t assoc_id = 0;
1403 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1412 * New (hopefully final) interface for the API.
1413 * We use the sctp_getaddrs_old structure so that use-space library
1414 * can avoid any unnecessary allocations. The only different part
1415 * is that we store the actual length of the address buffer into the
1416 * addrs_num structure member. That way we can re-use the existing
1419 #ifdef CONFIG_COMPAT
1420 struct compat_sctp_getaddrs_old {
1421 sctp_assoc_t assoc_id;
1423 compat_uptr_t addrs; /* struct sockaddr * */
1427 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1428 char __user *optval,
1431 struct sctp_getaddrs_old param;
1432 sctp_assoc_t assoc_id = 0;
1435 #ifdef CONFIG_COMPAT
1436 if (in_compat_syscall()) {
1437 struct compat_sctp_getaddrs_old param32;
1439 if (len < sizeof(param32))
1441 if (copy_from_user(¶m32, optval, sizeof(param32)))
1444 param.assoc_id = param32.assoc_id;
1445 param.addr_num = param32.addr_num;
1446 param.addrs = compat_ptr(param32.addrs);
1450 if (len < sizeof(param))
1452 if (copy_from_user(¶m, optval, sizeof(param)))
1456 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1457 param.addrs, param.addr_num,
1459 if (err == 0 || err == -EINPROGRESS) {
1460 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1462 if (put_user(sizeof(assoc_id), optlen))
1469 /* API 3.1.4 close() - UDP Style Syntax
1470 * Applications use close() to perform graceful shutdown (as described in
1471 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1472 * by a UDP-style socket.
1476 * ret = close(int sd);
1478 * sd - the socket descriptor of the associations to be closed.
1480 * To gracefully shutdown a specific association represented by the
1481 * UDP-style socket, an application should use the sendmsg() call,
1482 * passing no user data, but including the appropriate flag in the
1483 * ancillary data (see Section xxxx).
1485 * If sd in the close() call is a branched-off socket representing only
1486 * one association, the shutdown is performed on that association only.
1488 * 4.1.6 close() - TCP Style Syntax
1490 * Applications use close() to gracefully close down an association.
1494 * int close(int sd);
1496 * sd - the socket descriptor of the association to be closed.
1498 * After an application calls close() on a socket descriptor, no further
1499 * socket operations will succeed on that descriptor.
1501 * API 7.1.4 SO_LINGER
1503 * An application using the TCP-style socket can use this option to
1504 * perform the SCTP ABORT primitive. The linger option structure is:
1507 * int l_onoff; // option on/off
1508 * int l_linger; // linger time
1511 * To enable the option, set l_onoff to 1. If the l_linger value is set
1512 * to 0, calling close() is the same as the ABORT primitive. If the
1513 * value is set to a negative value, the setsockopt() call will return
1514 * an error. If the value is set to a positive value linger_time, the
1515 * close() can be blocked for at most linger_time ms. If the graceful
1516 * shutdown phase does not finish during this period, close() will
1517 * return but the graceful shutdown phase continues in the system.
1519 static void sctp_close(struct sock *sk, long timeout)
1521 struct net *net = sock_net(sk);
1522 struct sctp_endpoint *ep;
1523 struct sctp_association *asoc;
1524 struct list_head *pos, *temp;
1525 unsigned int data_was_unread;
1527 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1529 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
1530 sk->sk_shutdown = SHUTDOWN_MASK;
1531 sk->sk_state = SCTP_SS_CLOSING;
1533 ep = sctp_sk(sk)->ep;
1535 /* Clean up any skbs sitting on the receive queue. */
1536 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1537 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1539 /* Walk all associations on an endpoint. */
1540 list_for_each_safe(pos, temp, &ep->asocs) {
1541 asoc = list_entry(pos, struct sctp_association, asocs);
1543 if (sctp_style(sk, TCP)) {
1544 /* A closed association can still be in the list if
1545 * it belongs to a TCP-style listening socket that is
1546 * not yet accepted. If so, free it. If not, send an
1547 * ABORT or SHUTDOWN based on the linger options.
1549 if (sctp_state(asoc, CLOSED)) {
1550 sctp_association_free(asoc);
1555 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1556 !skb_queue_empty(&asoc->ulpq.reasm) ||
1557 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1558 struct sctp_chunk *chunk;
1560 chunk = sctp_make_abort_user(asoc, NULL, 0);
1561 sctp_primitive_ABORT(net, asoc, chunk);
1563 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1566 /* On a TCP-style socket, block for at most linger_time if set. */
1567 if (sctp_style(sk, TCP) && timeout)
1568 sctp_wait_for_close(sk, timeout);
1570 /* This will run the backlog queue. */
1573 /* Supposedly, no process has access to the socket, but
1574 * the net layers still may.
1575 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1576 * held and that should be grabbed before socket lock.
1578 spin_lock_bh(&net->sctp.addr_wq_lock);
1579 bh_lock_sock_nested(sk);
1581 /* Hold the sock, since sk_common_release() will put sock_put()
1582 * and we have just a little more cleanup.
1585 sk_common_release(sk);
1588 spin_unlock_bh(&net->sctp.addr_wq_lock);
1592 SCTP_DBG_OBJCNT_DEC(sock);
1595 /* Handle EPIPE error. */
1596 static int sctp_error(struct sock *sk, int flags, int err)
1599 err = sock_error(sk) ? : -EPIPE;
1600 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1601 send_sig(SIGPIPE, current, 0);
1605 /* API 3.1.3 sendmsg() - UDP Style Syntax
1607 * An application uses sendmsg() and recvmsg() calls to transmit data to
1608 * and receive data from its peer.
1610 * ssize_t sendmsg(int socket, const struct msghdr *message,
1613 * socket - the socket descriptor of the endpoint.
1614 * message - pointer to the msghdr structure which contains a single
1615 * user message and possibly some ancillary data.
1617 * See Section 5 for complete description of the data
1620 * flags - flags sent or received with the user message, see Section
1621 * 5 for complete description of the flags.
1623 * Note: This function could use a rewrite especially when explicit
1624 * connect support comes in.
1626 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1628 static int sctp_msghdr_parse(const struct msghdr *msg,
1629 struct sctp_cmsgs *cmsgs);
1631 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
1633 struct net *net = sock_net(sk);
1634 struct sctp_sock *sp;
1635 struct sctp_endpoint *ep;
1636 struct sctp_association *new_asoc = NULL, *asoc = NULL;
1637 struct sctp_transport *transport, *chunk_tp;
1638 struct sctp_chunk *chunk;
1640 struct sockaddr *msg_name = NULL;
1641 struct sctp_sndrcvinfo default_sinfo;
1642 struct sctp_sndrcvinfo *sinfo;
1643 struct sctp_initmsg *sinit;
1644 sctp_assoc_t associd = 0;
1645 struct sctp_cmsgs cmsgs = { NULL };
1646 enum sctp_scope scope;
1647 bool fill_sinfo_ttl = false, wait_connect = false;
1648 struct sctp_datamsg *datamsg;
1649 int msg_flags = msg->msg_flags;
1650 __u16 sinfo_flags = 0;
1658 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1661 /* We cannot send a message over a TCP-style listening socket. */
1662 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1667 /* Parse out the SCTP CMSGs. */
1668 err = sctp_msghdr_parse(msg, &cmsgs);
1670 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1674 /* Fetch the destination address for this packet. This
1675 * address only selects the association--it is not necessarily
1676 * the address we will send to.
1677 * For a peeled-off socket, msg_name is ignored.
1679 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1680 int msg_namelen = msg->msg_namelen;
1682 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1687 if (msg_namelen > sizeof(to))
1688 msg_namelen = sizeof(to);
1689 memcpy(&to, msg->msg_name, msg_namelen);
1690 msg_name = msg->msg_name;
1694 if (cmsgs.sinfo != NULL) {
1695 memset(&default_sinfo, 0, sizeof(default_sinfo));
1696 default_sinfo.sinfo_stream = cmsgs.sinfo->snd_sid;
1697 default_sinfo.sinfo_flags = cmsgs.sinfo->snd_flags;
1698 default_sinfo.sinfo_ppid = cmsgs.sinfo->snd_ppid;
1699 default_sinfo.sinfo_context = cmsgs.sinfo->snd_context;
1700 default_sinfo.sinfo_assoc_id = cmsgs.sinfo->snd_assoc_id;
1702 sinfo = &default_sinfo;
1703 fill_sinfo_ttl = true;
1705 sinfo = cmsgs.srinfo;
1707 /* Did the user specify SNDINFO/SNDRCVINFO? */
1709 sinfo_flags = sinfo->sinfo_flags;
1710 associd = sinfo->sinfo_assoc_id;
1713 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1714 msg_len, sinfo_flags);
1716 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1717 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1722 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1723 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1724 * If SCTP_ABORT is set, the message length could be non zero with
1725 * the msg_iov set to the user abort reason.
1727 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1728 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1733 /* If SCTP_ADDR_OVER is set, there must be an address
1734 * specified in msg_name.
1736 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1743 pr_debug("%s: about to look up association\n", __func__);
1747 /* If a msg_name has been specified, assume this is to be used. */
1749 /* Look for a matching association on the endpoint. */
1750 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1752 /* If we could not find a matching association on the
1753 * endpoint, make sure that it is not a TCP-style
1754 * socket that already has an association or there is
1755 * no peeled-off association on another socket.
1758 ((sctp_style(sk, TCP) &&
1759 (sctp_sstate(sk, ESTABLISHED) ||
1760 sctp_sstate(sk, CLOSING))) ||
1761 sctp_endpoint_is_peeled_off(ep, &to))) {
1762 err = -EADDRNOTAVAIL;
1766 asoc = sctp_id2assoc(sk, associd);
1774 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1776 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1777 * socket that has an association in CLOSED state. This can
1778 * happen when an accepted socket has an association that is
1781 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1786 if (sinfo_flags & SCTP_EOF) {
1787 pr_debug("%s: shutting down association:%p\n",
1790 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1794 if (sinfo_flags & SCTP_ABORT) {
1796 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1802 pr_debug("%s: aborting association:%p\n",
1805 sctp_primitive_ABORT(net, asoc, chunk);
1811 /* Do we need to create the association? */
1813 pr_debug("%s: there is no association yet\n", __func__);
1815 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1820 /* Check for invalid stream against the stream counts,
1821 * either the default or the user specified stream counts.
1824 if (!sinit || !sinit->sinit_num_ostreams) {
1825 /* Check against the defaults. */
1826 if (sinfo->sinfo_stream >=
1827 sp->initmsg.sinit_num_ostreams) {
1832 /* Check against the requested. */
1833 if (sinfo->sinfo_stream >=
1834 sinit->sinit_num_ostreams) {
1842 * API 3.1.2 bind() - UDP Style Syntax
1843 * If a bind() or sctp_bindx() is not called prior to a
1844 * sendmsg() call that initiates a new association, the
1845 * system picks an ephemeral port and will choose an address
1846 * set equivalent to binding with a wildcard address.
1848 if (!ep->base.bind_addr.port) {
1849 if (sctp_autobind(sk)) {
1855 * If an unprivileged user inherits a one-to-many
1856 * style socket with open associations on a privileged
1857 * port, it MAY be permitted to accept new associations,
1858 * but it SHOULD NOT be permitted to open new
1861 if (ep->base.bind_addr.port < inet_prot_sock(net) &&
1862 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1868 scope = sctp_scope(&to);
1869 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1875 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1881 /* If the SCTP_INIT ancillary data is specified, set all
1882 * the association init values accordingly.
1885 if (sinit->sinit_num_ostreams) {
1886 __u16 outcnt = sinit->sinit_num_ostreams;
1888 asoc->c.sinit_num_ostreams = outcnt;
1889 /* outcnt has been changed, so re-init stream */
1890 err = sctp_stream_init(&asoc->stream, outcnt, 0,
1895 if (sinit->sinit_max_instreams) {
1896 asoc->c.sinit_max_instreams =
1897 sinit->sinit_max_instreams;
1899 if (sinit->sinit_max_attempts) {
1900 asoc->max_init_attempts
1901 = sinit->sinit_max_attempts;
1903 if (sinit->sinit_max_init_timeo) {
1904 asoc->max_init_timeo =
1905 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1909 /* Prime the peer's transport structures. */
1910 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1917 /* ASSERT: we have a valid association at this point. */
1918 pr_debug("%s: we have a valid association\n", __func__);
1921 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1922 * one with some defaults.
1924 memset(&default_sinfo, 0, sizeof(default_sinfo));
1925 default_sinfo.sinfo_stream = asoc->default_stream;
1926 default_sinfo.sinfo_flags = asoc->default_flags;
1927 default_sinfo.sinfo_ppid = asoc->default_ppid;
1928 default_sinfo.sinfo_context = asoc->default_context;
1929 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1930 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1932 sinfo = &default_sinfo;
1933 } else if (fill_sinfo_ttl) {
1934 /* In case SNDINFO was specified, we still need to fill
1935 * it with a default ttl from the assoc here.
1937 sinfo->sinfo_timetolive = asoc->default_timetolive;
1940 /* API 7.1.7, the sndbuf size per association bounds the
1941 * maximum size of data that can be sent in a single send call.
1943 if (msg_len > sk->sk_sndbuf) {
1948 if (asoc->pmtu_pending)
1949 sctp_assoc_pending_pmtu(asoc);
1951 /* If fragmentation is disabled and the message length exceeds the
1952 * association fragmentation point, return EMSGSIZE. The I-D
1953 * does not specify what this error is, but this looks like
1956 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1961 /* Check for invalid stream. */
1962 if (sinfo->sinfo_stream >= asoc->stream.outcnt) {
1967 /* Allocate sctp_stream_out_ext if not already done */
1968 if (unlikely(!asoc->stream.out[sinfo->sinfo_stream].ext)) {
1969 err = sctp_stream_init_ext(&asoc->stream, sinfo->sinfo_stream);
1974 if (sctp_wspace(asoc) < msg_len)
1975 sctp_prsctp_prune(asoc, sinfo, msg_len - sctp_wspace(asoc));
1977 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1978 if (!sctp_wspace(asoc)) {
1979 /* sk can be changed by peel off when waiting for buf. */
1980 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len, &sk);
1982 if (err == -ESRCH) {
1983 /* asoc is already dead. */
1991 /* If an address is passed with the sendto/sendmsg call, it is used
1992 * to override the primary destination address in the TCP model, or
1993 * when SCTP_ADDR_OVER flag is set in the UDP model.
1995 if ((sctp_style(sk, TCP) && msg_name) ||
1996 (sinfo_flags & SCTP_ADDR_OVER)) {
1997 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
2005 /* Auto-connect, if we aren't connected already. */
2006 if (sctp_state(asoc, CLOSED)) {
2007 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
2011 wait_connect = true;
2012 pr_debug("%s: we associated primitively\n", __func__);
2015 /* Break the message into multiple chunks of maximum size. */
2016 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
2017 if (IS_ERR(datamsg)) {
2018 err = PTR_ERR(datamsg);
2021 asoc->force_delay = !!(msg->msg_flags & MSG_MORE);
2023 /* Now send the (possibly) fragmented message. */
2024 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
2025 sctp_chunk_hold(chunk);
2027 /* Do accounting for the write space. */
2028 sctp_set_owner_w(chunk);
2030 chunk->transport = chunk_tp;
2033 /* Send it to the lower layers. Note: all chunks
2034 * must either fail or succeed. The lower layer
2035 * works that way today. Keep it that way or this
2038 err = sctp_primitive_SEND(net, asoc, datamsg);
2039 /* Did the lower layer accept the chunk? */
2041 sctp_datamsg_free(datamsg);
2045 pr_debug("%s: we sent primitively\n", __func__);
2047 sctp_datamsg_put(datamsg);
2050 if (unlikely(wait_connect)) {
2051 timeo = sock_sndtimeo(sk, msg_flags & MSG_DONTWAIT);
2052 sctp_wait_for_connect(asoc, &timeo);
2055 /* If we are already past ASSOCIATE, the lower
2056 * layers are responsible for association cleanup.
2062 sctp_association_free(asoc);
2067 return sctp_error(sk, msg_flags, err);
2074 err = sock_error(sk);
2084 /* This is an extended version of skb_pull() that removes the data from the
2085 * start of a skb even when data is spread across the list of skb's in the
2086 * frag_list. len specifies the total amount of data that needs to be removed.
2087 * when 'len' bytes could be removed from the skb, it returns 0.
2088 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2089 * could not be removed.
2091 static int sctp_skb_pull(struct sk_buff *skb, int len)
2093 struct sk_buff *list;
2094 int skb_len = skb_headlen(skb);
2097 if (len <= skb_len) {
2098 __skb_pull(skb, len);
2102 __skb_pull(skb, skb_len);
2104 skb_walk_frags(skb, list) {
2105 rlen = sctp_skb_pull(list, len);
2106 skb->len -= (len-rlen);
2107 skb->data_len -= (len-rlen);
2118 /* API 3.1.3 recvmsg() - UDP Style Syntax
2120 * ssize_t recvmsg(int socket, struct msghdr *message,
2123 * socket - the socket descriptor of the endpoint.
2124 * message - pointer to the msghdr structure which contains a single
2125 * user message and possibly some ancillary data.
2127 * See Section 5 for complete description of the data
2130 * flags - flags sent or received with the user message, see Section
2131 * 5 for complete description of the flags.
2133 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2134 int noblock, int flags, int *addr_len)
2136 struct sctp_ulpevent *event = NULL;
2137 struct sctp_sock *sp = sctp_sk(sk);
2138 struct sk_buff *skb, *head_skb;
2143 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2144 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2149 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED) &&
2150 !sctp_sstate(sk, CLOSING) && !sctp_sstate(sk, CLOSED)) {
2155 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2159 /* Get the total length of the skb including any skb's in the
2168 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2170 event = sctp_skb2event(skb);
2175 if (event->chunk && event->chunk->head_skb)
2176 head_skb = event->chunk->head_skb;
2179 sock_recv_ts_and_drops(msg, sk, head_skb);
2180 if (sctp_ulpevent_is_notification(event)) {
2181 msg->msg_flags |= MSG_NOTIFICATION;
2182 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2184 sp->pf->skb_msgname(head_skb, msg->msg_name, addr_len);
2187 /* Check if we allow SCTP_NXTINFO. */
2188 if (sp->recvnxtinfo)
2189 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2190 /* Check if we allow SCTP_RCVINFO. */
2191 if (sp->recvrcvinfo)
2192 sctp_ulpevent_read_rcvinfo(event, msg);
2193 /* Check if we allow SCTP_SNDRCVINFO. */
2194 if (sp->subscribe.sctp_data_io_event)
2195 sctp_ulpevent_read_sndrcvinfo(event, msg);
2199 /* If skb's length exceeds the user's buffer, update the skb and
2200 * push it back to the receive_queue so that the next call to
2201 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2203 if (skb_len > copied) {
2204 msg->msg_flags &= ~MSG_EOR;
2205 if (flags & MSG_PEEK)
2207 sctp_skb_pull(skb, copied);
2208 skb_queue_head(&sk->sk_receive_queue, skb);
2210 /* When only partial message is copied to the user, increase
2211 * rwnd by that amount. If all the data in the skb is read,
2212 * rwnd is updated when the event is freed.
2214 if (!sctp_ulpevent_is_notification(event))
2215 sctp_assoc_rwnd_increase(event->asoc, copied);
2217 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2218 (event->msg_flags & MSG_EOR))
2219 msg->msg_flags |= MSG_EOR;
2221 msg->msg_flags &= ~MSG_EOR;
2224 if (flags & MSG_PEEK) {
2225 /* Release the skb reference acquired after peeking the skb in
2226 * sctp_skb_recv_datagram().
2230 /* Free the event which includes releasing the reference to
2231 * the owner of the skb, freeing the skb and updating the
2234 sctp_ulpevent_free(event);
2241 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2243 * This option is a on/off flag. If enabled no SCTP message
2244 * fragmentation will be performed. Instead if a message being sent
2245 * exceeds the current PMTU size, the message will NOT be sent and
2246 * instead a error will be indicated to the user.
2248 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2249 char __user *optval,
2250 unsigned int optlen)
2254 if (optlen < sizeof(int))
2257 if (get_user(val, (int __user *)optval))
2260 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2265 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2266 unsigned int optlen)
2268 struct sctp_association *asoc;
2269 struct sctp_ulpevent *event;
2271 if (optlen > sizeof(struct sctp_event_subscribe))
2273 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2276 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2277 * if there is no data to be sent or retransmit, the stack will
2278 * immediately send up this notification.
2280 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2281 &sctp_sk(sk)->subscribe)) {
2282 asoc = sctp_id2assoc(sk, 0);
2284 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2285 event = sctp_ulpevent_make_sender_dry_event(asoc,
2286 GFP_USER | __GFP_NOWARN);
2290 sctp_ulpq_tail_event(&asoc->ulpq, event);
2297 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2299 * This socket option is applicable to the UDP-style socket only. When
2300 * set it will cause associations that are idle for more than the
2301 * specified number of seconds to automatically close. An association
2302 * being idle is defined an association that has NOT sent or received
2303 * user data. The special value of '0' indicates that no automatic
2304 * close of any associations should be performed. The option expects an
2305 * integer defining the number of seconds of idle time before an
2306 * association is closed.
2308 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2309 unsigned int optlen)
2311 struct sctp_sock *sp = sctp_sk(sk);
2312 struct net *net = sock_net(sk);
2314 /* Applicable to UDP-style socket only */
2315 if (sctp_style(sk, TCP))
2317 if (optlen != sizeof(int))
2319 if (copy_from_user(&sp->autoclose, optval, optlen))
2322 if (sp->autoclose > net->sctp.max_autoclose)
2323 sp->autoclose = net->sctp.max_autoclose;
2328 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2330 * Applications can enable or disable heartbeats for any peer address of
2331 * an association, modify an address's heartbeat interval, force a
2332 * heartbeat to be sent immediately, and adjust the address's maximum
2333 * number of retransmissions sent before an address is considered
2334 * unreachable. The following structure is used to access and modify an
2335 * address's parameters:
2337 * struct sctp_paddrparams {
2338 * sctp_assoc_t spp_assoc_id;
2339 * struct sockaddr_storage spp_address;
2340 * uint32_t spp_hbinterval;
2341 * uint16_t spp_pathmaxrxt;
2342 * uint32_t spp_pathmtu;
2343 * uint32_t spp_sackdelay;
2344 * uint32_t spp_flags;
2347 * spp_assoc_id - (one-to-many style socket) This is filled in the
2348 * application, and identifies the association for
2350 * spp_address - This specifies which address is of interest.
2351 * spp_hbinterval - This contains the value of the heartbeat interval,
2352 * in milliseconds. If a value of zero
2353 * is present in this field then no changes are to
2354 * be made to this parameter.
2355 * spp_pathmaxrxt - This contains the maximum number of
2356 * retransmissions before this address shall be
2357 * considered unreachable. If a value of zero
2358 * is present in this field then no changes are to
2359 * be made to this parameter.
2360 * spp_pathmtu - When Path MTU discovery is disabled the value
2361 * specified here will be the "fixed" path mtu.
2362 * Note that if the spp_address field is empty
2363 * then all associations on this address will
2364 * have this fixed path mtu set upon them.
2366 * spp_sackdelay - When delayed sack is enabled, this value specifies
2367 * the number of milliseconds that sacks will be delayed
2368 * for. This value will apply to all addresses of an
2369 * association if the spp_address field is empty. Note
2370 * also, that if delayed sack is enabled and this
2371 * value is set to 0, no change is made to the last
2372 * recorded delayed sack timer value.
2374 * spp_flags - These flags are used to control various features
2375 * on an association. The flag field may contain
2376 * zero or more of the following options.
2378 * SPP_HB_ENABLE - Enable heartbeats on the
2379 * specified address. Note that if the address
2380 * field is empty all addresses for the association
2381 * have heartbeats enabled upon them.
2383 * SPP_HB_DISABLE - Disable heartbeats on the
2384 * speicifed address. Note that if the address
2385 * field is empty all addresses for the association
2386 * will have their heartbeats disabled. Note also
2387 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2388 * mutually exclusive, only one of these two should
2389 * be specified. Enabling both fields will have
2390 * undetermined results.
2392 * SPP_HB_DEMAND - Request a user initiated heartbeat
2393 * to be made immediately.
2395 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2396 * heartbeat delayis to be set to the value of 0
2399 * SPP_PMTUD_ENABLE - This field will enable PMTU
2400 * discovery upon the specified address. Note that
2401 * if the address feild is empty then all addresses
2402 * on the association are effected.
2404 * SPP_PMTUD_DISABLE - This field will disable PMTU
2405 * discovery upon the specified address. Note that
2406 * if the address feild is empty then all addresses
2407 * on the association are effected. Not also that
2408 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2409 * exclusive. Enabling both will have undetermined
2412 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2413 * on delayed sack. The time specified in spp_sackdelay
2414 * is used to specify the sack delay for this address. Note
2415 * that if spp_address is empty then all addresses will
2416 * enable delayed sack and take on the sack delay
2417 * value specified in spp_sackdelay.
2418 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2419 * off delayed sack. If the spp_address field is blank then
2420 * delayed sack is disabled for the entire association. Note
2421 * also that this field is mutually exclusive to
2422 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2425 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2426 struct sctp_transport *trans,
2427 struct sctp_association *asoc,
2428 struct sctp_sock *sp,
2431 int sackdelay_change)
2435 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2436 struct net *net = sock_net(trans->asoc->base.sk);
2438 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2443 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2444 * this field is ignored. Note also that a value of zero indicates
2445 * the current setting should be left unchanged.
2447 if (params->spp_flags & SPP_HB_ENABLE) {
2449 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2450 * set. This lets us use 0 value when this flag
2453 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2454 params->spp_hbinterval = 0;
2456 if (params->spp_hbinterval ||
2457 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2460 msecs_to_jiffies(params->spp_hbinterval);
2463 msecs_to_jiffies(params->spp_hbinterval);
2465 sp->hbinterval = params->spp_hbinterval;
2472 trans->param_flags =
2473 (trans->param_flags & ~SPP_HB) | hb_change;
2476 (asoc->param_flags & ~SPP_HB) | hb_change;
2479 (sp->param_flags & ~SPP_HB) | hb_change;
2483 /* When Path MTU discovery is disabled the value specified here will
2484 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2485 * include the flag SPP_PMTUD_DISABLE for this field to have any
2488 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2490 trans->pathmtu = params->spp_pathmtu;
2491 sctp_assoc_sync_pmtu(asoc);
2493 asoc->pathmtu = params->spp_pathmtu;
2495 sp->pathmtu = params->spp_pathmtu;
2501 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2502 (params->spp_flags & SPP_PMTUD_ENABLE);
2503 trans->param_flags =
2504 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2506 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2507 sctp_assoc_sync_pmtu(asoc);
2511 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2514 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2518 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2519 * value of this field is ignored. Note also that a value of zero
2520 * indicates the current setting should be left unchanged.
2522 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2525 msecs_to_jiffies(params->spp_sackdelay);
2528 msecs_to_jiffies(params->spp_sackdelay);
2530 sp->sackdelay = params->spp_sackdelay;
2534 if (sackdelay_change) {
2536 trans->param_flags =
2537 (trans->param_flags & ~SPP_SACKDELAY) |
2541 (asoc->param_flags & ~SPP_SACKDELAY) |
2545 (sp->param_flags & ~SPP_SACKDELAY) |
2550 /* Note that a value of zero indicates the current setting should be
2553 if (params->spp_pathmaxrxt) {
2555 trans->pathmaxrxt = params->spp_pathmaxrxt;
2557 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2559 sp->pathmaxrxt = params->spp_pathmaxrxt;
2566 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2567 char __user *optval,
2568 unsigned int optlen)
2570 struct sctp_paddrparams params;
2571 struct sctp_transport *trans = NULL;
2572 struct sctp_association *asoc = NULL;
2573 struct sctp_sock *sp = sctp_sk(sk);
2575 int hb_change, pmtud_change, sackdelay_change;
2577 if (optlen != sizeof(struct sctp_paddrparams))
2580 if (copy_from_user(¶ms, optval, optlen))
2583 /* Validate flags and value parameters. */
2584 hb_change = params.spp_flags & SPP_HB;
2585 pmtud_change = params.spp_flags & SPP_PMTUD;
2586 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2588 if (hb_change == SPP_HB ||
2589 pmtud_change == SPP_PMTUD ||
2590 sackdelay_change == SPP_SACKDELAY ||
2591 params.spp_sackdelay > 500 ||
2592 (params.spp_pathmtu &&
2593 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2596 /* If an address other than INADDR_ANY is specified, and
2597 * no transport is found, then the request is invalid.
2599 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
2600 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2601 params.spp_assoc_id);
2606 /* Get association, if assoc_id != 0 and the socket is a one
2607 * to many style socket, and an association was not found, then
2608 * the id was invalid.
2610 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2611 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2614 /* Heartbeat demand can only be sent on a transport or
2615 * association, but not a socket.
2617 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2620 /* Process parameters. */
2621 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2622 hb_change, pmtud_change,
2628 /* If changes are for association, also apply parameters to each
2631 if (!trans && asoc) {
2632 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2634 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2635 hb_change, pmtud_change,
2643 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2645 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2648 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2650 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2654 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2656 * This option will effect the way delayed acks are performed. This
2657 * option allows you to get or set the delayed ack time, in
2658 * milliseconds. It also allows changing the delayed ack frequency.
2659 * Changing the frequency to 1 disables the delayed sack algorithm. If
2660 * the assoc_id is 0, then this sets or gets the endpoints default
2661 * values. If the assoc_id field is non-zero, then the set or get
2662 * effects the specified association for the one to many model (the
2663 * assoc_id field is ignored by the one to one model). Note that if
2664 * sack_delay or sack_freq are 0 when setting this option, then the
2665 * current values will remain unchanged.
2667 * struct sctp_sack_info {
2668 * sctp_assoc_t sack_assoc_id;
2669 * uint32_t sack_delay;
2670 * uint32_t sack_freq;
2673 * sack_assoc_id - This parameter, indicates which association the user
2674 * is performing an action upon. Note that if this field's value is
2675 * zero then the endpoints default value is changed (effecting future
2676 * associations only).
2678 * sack_delay - This parameter contains the number of milliseconds that
2679 * the user is requesting the delayed ACK timer be set to. Note that
2680 * this value is defined in the standard to be between 200 and 500
2683 * sack_freq - This parameter contains the number of packets that must
2684 * be received before a sack is sent without waiting for the delay
2685 * timer to expire. The default value for this is 2, setting this
2686 * value to 1 will disable the delayed sack algorithm.
2689 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2690 char __user *optval, unsigned int optlen)
2692 struct sctp_sack_info params;
2693 struct sctp_transport *trans = NULL;
2694 struct sctp_association *asoc = NULL;
2695 struct sctp_sock *sp = sctp_sk(sk);
2697 if (optlen == sizeof(struct sctp_sack_info)) {
2698 if (copy_from_user(¶ms, optval, optlen))
2701 if (params.sack_delay == 0 && params.sack_freq == 0)
2703 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2704 pr_warn_ratelimited(DEPRECATED
2706 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2707 "Use struct sctp_sack_info instead\n",
2708 current->comm, task_pid_nr(current));
2709 if (copy_from_user(¶ms, optval, optlen))
2712 if (params.sack_delay == 0)
2713 params.sack_freq = 1;
2715 params.sack_freq = 0;
2719 /* Validate value parameter. */
2720 if (params.sack_delay > 500)
2723 /* Get association, if sack_assoc_id != 0 and the socket is a one
2724 * to many style socket, and an association was not found, then
2725 * the id was invalid.
2727 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2728 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2731 if (params.sack_delay) {
2734 msecs_to_jiffies(params.sack_delay);
2736 sctp_spp_sackdelay_enable(asoc->param_flags);
2738 sp->sackdelay = params.sack_delay;
2740 sctp_spp_sackdelay_enable(sp->param_flags);
2744 if (params.sack_freq == 1) {
2747 sctp_spp_sackdelay_disable(asoc->param_flags);
2750 sctp_spp_sackdelay_disable(sp->param_flags);
2752 } else if (params.sack_freq > 1) {
2754 asoc->sackfreq = params.sack_freq;
2756 sctp_spp_sackdelay_enable(asoc->param_flags);
2758 sp->sackfreq = params.sack_freq;
2760 sctp_spp_sackdelay_enable(sp->param_flags);
2764 /* If change is for association, also apply to each transport. */
2766 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2768 if (params.sack_delay) {
2770 msecs_to_jiffies(params.sack_delay);
2771 trans->param_flags =
2772 sctp_spp_sackdelay_enable(trans->param_flags);
2774 if (params.sack_freq == 1) {
2775 trans->param_flags =
2776 sctp_spp_sackdelay_disable(trans->param_flags);
2777 } else if (params.sack_freq > 1) {
2778 trans->sackfreq = params.sack_freq;
2779 trans->param_flags =
2780 sctp_spp_sackdelay_enable(trans->param_flags);
2788 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2790 * Applications can specify protocol parameters for the default association
2791 * initialization. The option name argument to setsockopt() and getsockopt()
2794 * Setting initialization parameters is effective only on an unconnected
2795 * socket (for UDP-style sockets only future associations are effected
2796 * by the change). With TCP-style sockets, this option is inherited by
2797 * sockets derived from a listener socket.
2799 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2801 struct sctp_initmsg sinit;
2802 struct sctp_sock *sp = sctp_sk(sk);
2804 if (optlen != sizeof(struct sctp_initmsg))
2806 if (copy_from_user(&sinit, optval, optlen))
2809 if (sinit.sinit_num_ostreams)
2810 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2811 if (sinit.sinit_max_instreams)
2812 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2813 if (sinit.sinit_max_attempts)
2814 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2815 if (sinit.sinit_max_init_timeo)
2816 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2822 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2824 * Applications that wish to use the sendto() system call may wish to
2825 * specify a default set of parameters that would normally be supplied
2826 * through the inclusion of ancillary data. This socket option allows
2827 * such an application to set the default sctp_sndrcvinfo structure.
2828 * The application that wishes to use this socket option simply passes
2829 * in to this call the sctp_sndrcvinfo structure defined in Section
2830 * 5.2.2) The input parameters accepted by this call include
2831 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2832 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2833 * to this call if the caller is using the UDP model.
2835 static int sctp_setsockopt_default_send_param(struct sock *sk,
2836 char __user *optval,
2837 unsigned int optlen)
2839 struct sctp_sock *sp = sctp_sk(sk);
2840 struct sctp_association *asoc;
2841 struct sctp_sndrcvinfo info;
2843 if (optlen != sizeof(info))
2845 if (copy_from_user(&info, optval, optlen))
2847 if (info.sinfo_flags &
2848 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2849 SCTP_ABORT | SCTP_EOF))
2852 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2853 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2856 asoc->default_stream = info.sinfo_stream;
2857 asoc->default_flags = info.sinfo_flags;
2858 asoc->default_ppid = info.sinfo_ppid;
2859 asoc->default_context = info.sinfo_context;
2860 asoc->default_timetolive = info.sinfo_timetolive;
2862 sp->default_stream = info.sinfo_stream;
2863 sp->default_flags = info.sinfo_flags;
2864 sp->default_ppid = info.sinfo_ppid;
2865 sp->default_context = info.sinfo_context;
2866 sp->default_timetolive = info.sinfo_timetolive;
2872 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2873 * (SCTP_DEFAULT_SNDINFO)
2875 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
2876 char __user *optval,
2877 unsigned int optlen)
2879 struct sctp_sock *sp = sctp_sk(sk);
2880 struct sctp_association *asoc;
2881 struct sctp_sndinfo info;
2883 if (optlen != sizeof(info))
2885 if (copy_from_user(&info, optval, optlen))
2887 if (info.snd_flags &
2888 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2889 SCTP_ABORT | SCTP_EOF))
2892 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
2893 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
2896 asoc->default_stream = info.snd_sid;
2897 asoc->default_flags = info.snd_flags;
2898 asoc->default_ppid = info.snd_ppid;
2899 asoc->default_context = info.snd_context;
2901 sp->default_stream = info.snd_sid;
2902 sp->default_flags = info.snd_flags;
2903 sp->default_ppid = info.snd_ppid;
2904 sp->default_context = info.snd_context;
2910 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2912 * Requests that the local SCTP stack use the enclosed peer address as
2913 * the association primary. The enclosed address must be one of the
2914 * association peer's addresses.
2916 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2917 unsigned int optlen)
2919 struct sctp_prim prim;
2920 struct sctp_transport *trans;
2922 if (optlen != sizeof(struct sctp_prim))
2925 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2928 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2932 sctp_assoc_set_primary(trans->asoc, trans);
2938 * 7.1.5 SCTP_NODELAY
2940 * Turn on/off any Nagle-like algorithm. This means that packets are
2941 * generally sent as soon as possible and no unnecessary delays are
2942 * introduced, at the cost of more packets in the network. Expects an
2943 * integer boolean flag.
2945 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2946 unsigned int optlen)
2950 if (optlen < sizeof(int))
2952 if (get_user(val, (int __user *)optval))
2955 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2961 * 7.1.1 SCTP_RTOINFO
2963 * The protocol parameters used to initialize and bound retransmission
2964 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2965 * and modify these parameters.
2966 * All parameters are time values, in milliseconds. A value of 0, when
2967 * modifying the parameters, indicates that the current value should not
2971 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2973 struct sctp_rtoinfo rtoinfo;
2974 struct sctp_association *asoc;
2975 unsigned long rto_min, rto_max;
2976 struct sctp_sock *sp = sctp_sk(sk);
2978 if (optlen != sizeof (struct sctp_rtoinfo))
2981 if (copy_from_user(&rtoinfo, optval, optlen))
2984 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2986 /* Set the values to the specific association */
2987 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2990 rto_max = rtoinfo.srto_max;
2991 rto_min = rtoinfo.srto_min;
2994 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2996 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2999 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
3001 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
3003 if (rto_min > rto_max)
3007 if (rtoinfo.srto_initial != 0)
3009 msecs_to_jiffies(rtoinfo.srto_initial);
3010 asoc->rto_max = rto_max;
3011 asoc->rto_min = rto_min;
3013 /* If there is no association or the association-id = 0
3014 * set the values to the endpoint.
3016 if (rtoinfo.srto_initial != 0)
3017 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
3018 sp->rtoinfo.srto_max = rto_max;
3019 sp->rtoinfo.srto_min = rto_min;
3027 * 7.1.2 SCTP_ASSOCINFO
3029 * This option is used to tune the maximum retransmission attempts
3030 * of the association.
3031 * Returns an error if the new association retransmission value is
3032 * greater than the sum of the retransmission value of the peer.
3033 * See [SCTP] for more information.
3036 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
3039 struct sctp_assocparams assocparams;
3040 struct sctp_association *asoc;
3042 if (optlen != sizeof(struct sctp_assocparams))
3044 if (copy_from_user(&assocparams, optval, optlen))
3047 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
3049 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
3052 /* Set the values to the specific association */
3054 if (assocparams.sasoc_asocmaxrxt != 0) {
3057 struct sctp_transport *peer_addr;
3059 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
3061 path_sum += peer_addr->pathmaxrxt;
3065 /* Only validate asocmaxrxt if we have more than
3066 * one path/transport. We do this because path
3067 * retransmissions are only counted when we have more
3071 assocparams.sasoc_asocmaxrxt > path_sum)
3074 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3077 if (assocparams.sasoc_cookie_life != 0)
3078 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3080 /* Set the values to the endpoint */
3081 struct sctp_sock *sp = sctp_sk(sk);
3083 if (assocparams.sasoc_asocmaxrxt != 0)
3084 sp->assocparams.sasoc_asocmaxrxt =
3085 assocparams.sasoc_asocmaxrxt;
3086 if (assocparams.sasoc_cookie_life != 0)
3087 sp->assocparams.sasoc_cookie_life =
3088 assocparams.sasoc_cookie_life;
3094 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3096 * This socket option is a boolean flag which turns on or off mapped V4
3097 * addresses. If this option is turned on and the socket is type
3098 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3099 * If this option is turned off, then no mapping will be done of V4
3100 * addresses and a user will receive both PF_INET6 and PF_INET type
3101 * addresses on the socket.
3103 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3106 struct sctp_sock *sp = sctp_sk(sk);
3108 if (optlen < sizeof(int))
3110 if (get_user(val, (int __user *)optval))
3121 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3122 * This option will get or set the maximum size to put in any outgoing
3123 * SCTP DATA chunk. If a message is larger than this size it will be
3124 * fragmented by SCTP into the specified size. Note that the underlying
3125 * SCTP implementation may fragment into smaller sized chunks when the
3126 * PMTU of the underlying association is smaller than the value set by
3127 * the user. The default value for this option is '0' which indicates
3128 * the user is NOT limiting fragmentation and only the PMTU will effect
3129 * SCTP's choice of DATA chunk size. Note also that values set larger
3130 * than the maximum size of an IP datagram will effectively let SCTP
3131 * control fragmentation (i.e. the same as setting this option to 0).
3133 * The following structure is used to access and modify this parameter:
3135 * struct sctp_assoc_value {
3136 * sctp_assoc_t assoc_id;
3137 * uint32_t assoc_value;
3140 * assoc_id: This parameter is ignored for one-to-one style sockets.
3141 * For one-to-many style sockets this parameter indicates which
3142 * association the user is performing an action upon. Note that if
3143 * this field's value is zero then the endpoints default value is
3144 * changed (effecting future associations only).
3145 * assoc_value: This parameter specifies the maximum size in bytes.
3147 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3149 struct sctp_sock *sp = sctp_sk(sk);
3150 struct sctp_assoc_value params;
3151 struct sctp_association *asoc;
3154 if (optlen == sizeof(int)) {
3155 pr_warn_ratelimited(DEPRECATED
3157 "Use of int in maxseg socket option.\n"
3158 "Use struct sctp_assoc_value instead\n",
3159 current->comm, task_pid_nr(current));
3160 if (copy_from_user(&val, optval, optlen))
3162 params.assoc_id = 0;
3163 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3164 if (copy_from_user(¶ms, optval, optlen))
3166 val = params.assoc_value;
3172 int min_len, max_len;
3174 min_len = SCTP_DEFAULT_MINSEGMENT - sp->pf->af->net_header_len;
3175 min_len -= sizeof(struct sctphdr) +
3176 sizeof(struct sctp_data_chunk);
3178 max_len = SCTP_MAX_CHUNK_LEN - sizeof(struct sctp_data_chunk);
3180 if (val < min_len || val > max_len)
3184 asoc = sctp_id2assoc(sk, params.assoc_id);
3187 val = asoc->pathmtu - sp->pf->af->net_header_len;
3188 val -= sizeof(struct sctphdr) +
3189 sizeof(struct sctp_data_chunk);
3191 asoc->user_frag = val;
3192 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3194 if (params.assoc_id && sctp_style(sk, UDP))
3196 sp->user_frag = val;
3204 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3206 * Requests that the peer mark the enclosed address as the association
3207 * primary. The enclosed address must be one of the association's
3208 * locally bound addresses. The following structure is used to make a
3209 * set primary request:
3211 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3212 unsigned int optlen)
3214 struct net *net = sock_net(sk);
3215 struct sctp_sock *sp;
3216 struct sctp_association *asoc = NULL;
3217 struct sctp_setpeerprim prim;
3218 struct sctp_chunk *chunk;
3224 if (!net->sctp.addip_enable)
3227 if (optlen != sizeof(struct sctp_setpeerprim))
3230 if (copy_from_user(&prim, optval, optlen))
3233 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3237 if (!asoc->peer.asconf_capable)
3240 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3243 if (!sctp_state(asoc, ESTABLISHED))
3246 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3250 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3251 return -EADDRNOTAVAIL;
3253 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3254 return -EADDRNOTAVAIL;
3256 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3257 chunk = sctp_make_asconf_set_prim(asoc,
3258 (union sctp_addr *)&prim.sspp_addr);
3262 err = sctp_send_asconf(asoc, chunk);
3264 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3269 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3270 unsigned int optlen)
3272 struct sctp_setadaptation adaptation;
3274 if (optlen != sizeof(struct sctp_setadaptation))
3276 if (copy_from_user(&adaptation, optval, optlen))
3279 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3285 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3287 * The context field in the sctp_sndrcvinfo structure is normally only
3288 * used when a failed message is retrieved holding the value that was
3289 * sent down on the actual send call. This option allows the setting of
3290 * a default context on an association basis that will be received on
3291 * reading messages from the peer. This is especially helpful in the
3292 * one-2-many model for an application to keep some reference to an
3293 * internal state machine that is processing messages on the
3294 * association. Note that the setting of this value only effects
3295 * received messages from the peer and does not effect the value that is
3296 * saved with outbound messages.
3298 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3299 unsigned int optlen)
3301 struct sctp_assoc_value params;
3302 struct sctp_sock *sp;
3303 struct sctp_association *asoc;
3305 if (optlen != sizeof(struct sctp_assoc_value))
3307 if (copy_from_user(¶ms, optval, optlen))
3312 if (params.assoc_id != 0) {
3313 asoc = sctp_id2assoc(sk, params.assoc_id);
3316 asoc->default_rcv_context = params.assoc_value;
3318 sp->default_rcv_context = params.assoc_value;
3325 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3327 * This options will at a minimum specify if the implementation is doing
3328 * fragmented interleave. Fragmented interleave, for a one to many
3329 * socket, is when subsequent calls to receive a message may return
3330 * parts of messages from different associations. Some implementations
3331 * may allow you to turn this value on or off. If so, when turned off,
3332 * no fragment interleave will occur (which will cause a head of line
3333 * blocking amongst multiple associations sharing the same one to many
3334 * socket). When this option is turned on, then each receive call may
3335 * come from a different association (thus the user must receive data
3336 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3337 * association each receive belongs to.
3339 * This option takes a boolean value. A non-zero value indicates that
3340 * fragmented interleave is on. A value of zero indicates that
3341 * fragmented interleave is off.
3343 * Note that it is important that an implementation that allows this
3344 * option to be turned on, have it off by default. Otherwise an unaware
3345 * application using the one to many model may become confused and act
3348 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3349 char __user *optval,
3350 unsigned int optlen)
3354 if (optlen != sizeof(int))
3356 if (get_user(val, (int __user *)optval))
3359 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3365 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3366 * (SCTP_PARTIAL_DELIVERY_POINT)
3368 * This option will set or get the SCTP partial delivery point. This
3369 * point is the size of a message where the partial delivery API will be
3370 * invoked to help free up rwnd space for the peer. Setting this to a
3371 * lower value will cause partial deliveries to happen more often. The
3372 * calls argument is an integer that sets or gets the partial delivery
3373 * point. Note also that the call will fail if the user attempts to set
3374 * this value larger than the socket receive buffer size.
3376 * Note that any single message having a length smaller than or equal to
3377 * the SCTP partial delivery point will be delivered in one single read
3378 * call as long as the user provided buffer is large enough to hold the
3381 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3382 char __user *optval,
3383 unsigned int optlen)
3387 if (optlen != sizeof(u32))
3389 if (get_user(val, (int __user *)optval))
3392 /* Note: We double the receive buffer from what the user sets
3393 * it to be, also initial rwnd is based on rcvbuf/2.
3395 if (val > (sk->sk_rcvbuf >> 1))
3398 sctp_sk(sk)->pd_point = val;
3400 return 0; /* is this the right error code? */
3404 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3406 * This option will allow a user to change the maximum burst of packets
3407 * that can be emitted by this association. Note that the default value
3408 * is 4, and some implementations may restrict this setting so that it
3409 * can only be lowered.
3411 * NOTE: This text doesn't seem right. Do this on a socket basis with
3412 * future associations inheriting the socket value.
3414 static int sctp_setsockopt_maxburst(struct sock *sk,
3415 char __user *optval,
3416 unsigned int optlen)
3418 struct sctp_assoc_value params;
3419 struct sctp_sock *sp;
3420 struct sctp_association *asoc;
3424 if (optlen == sizeof(int)) {
3425 pr_warn_ratelimited(DEPRECATED
3427 "Use of int in max_burst socket option deprecated.\n"
3428 "Use struct sctp_assoc_value instead\n",
3429 current->comm, task_pid_nr(current));
3430 if (copy_from_user(&val, optval, optlen))
3432 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3433 if (copy_from_user(¶ms, optval, optlen))
3435 val = params.assoc_value;
3436 assoc_id = params.assoc_id;
3442 if (assoc_id != 0) {
3443 asoc = sctp_id2assoc(sk, assoc_id);
3446 asoc->max_burst = val;
3448 sp->max_burst = val;
3454 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3456 * This set option adds a chunk type that the user is requesting to be
3457 * received only in an authenticated way. Changes to the list of chunks
3458 * will only effect future associations on the socket.
3460 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3461 char __user *optval,
3462 unsigned int optlen)
3464 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3465 struct sctp_authchunk val;
3467 if (!ep->auth_enable)
3470 if (optlen != sizeof(struct sctp_authchunk))
3472 if (copy_from_user(&val, optval, optlen))
3475 switch (val.sauth_chunk) {
3477 case SCTP_CID_INIT_ACK:
3478 case SCTP_CID_SHUTDOWN_COMPLETE:
3483 /* add this chunk id to the endpoint */
3484 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3488 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3490 * This option gets or sets the list of HMAC algorithms that the local
3491 * endpoint requires the peer to use.
3493 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3494 char __user *optval,
3495 unsigned int optlen)
3497 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3498 struct sctp_hmacalgo *hmacs;
3502 if (!ep->auth_enable)
3505 if (optlen < sizeof(struct sctp_hmacalgo))
3507 optlen = min_t(unsigned int, optlen, sizeof(struct sctp_hmacalgo) +
3508 SCTP_AUTH_NUM_HMACS * sizeof(u16));
3510 hmacs = memdup_user(optval, optlen);
3512 return PTR_ERR(hmacs);
3514 idents = hmacs->shmac_num_idents;
3515 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3516 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3521 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3528 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3530 * This option will set a shared secret key which is used to build an
3531 * association shared key.
3533 static int sctp_setsockopt_auth_key(struct sock *sk,
3534 char __user *optval,
3535 unsigned int optlen)
3537 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3538 struct sctp_authkey *authkey;
3539 struct sctp_association *asoc;
3542 if (!ep->auth_enable)
3545 if (optlen <= sizeof(struct sctp_authkey))
3547 /* authkey->sca_keylength is u16, so optlen can't be bigger than
3550 optlen = min_t(unsigned int, optlen, USHRT_MAX +
3551 sizeof(struct sctp_authkey));
3553 authkey = memdup_user(optval, optlen);
3554 if (IS_ERR(authkey))
3555 return PTR_ERR(authkey);
3557 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3562 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3563 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3568 ret = sctp_auth_set_key(ep, asoc, authkey);
3575 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3577 * This option will get or set the active shared key to be used to build
3578 * the association shared key.
3580 static int sctp_setsockopt_active_key(struct sock *sk,
3581 char __user *optval,
3582 unsigned int optlen)
3584 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3585 struct sctp_authkeyid val;
3586 struct sctp_association *asoc;
3588 if (!ep->auth_enable)
3591 if (optlen != sizeof(struct sctp_authkeyid))
3593 if (copy_from_user(&val, optval, optlen))
3596 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3597 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3600 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3604 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3606 * This set option will delete a shared secret key from use.
3608 static int sctp_setsockopt_del_key(struct sock *sk,
3609 char __user *optval,
3610 unsigned int optlen)
3612 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3613 struct sctp_authkeyid val;
3614 struct sctp_association *asoc;
3616 if (!ep->auth_enable)
3619 if (optlen != sizeof(struct sctp_authkeyid))
3621 if (copy_from_user(&val, optval, optlen))
3624 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3625 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3628 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3633 * 8.1.23 SCTP_AUTO_ASCONF
3635 * This option will enable or disable the use of the automatic generation of
3636 * ASCONF chunks to add and delete addresses to an existing association. Note
3637 * that this option has two caveats namely: a) it only affects sockets that
3638 * are bound to all addresses available to the SCTP stack, and b) the system
3639 * administrator may have an overriding control that turns the ASCONF feature
3640 * off no matter what setting the socket option may have.
3641 * This option expects an integer boolean flag, where a non-zero value turns on
3642 * the option, and a zero value turns off the option.
3643 * Note. In this implementation, socket operation overrides default parameter
3644 * being set by sysctl as well as FreeBSD implementation
3646 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3647 unsigned int optlen)
3650 struct sctp_sock *sp = sctp_sk(sk);
3652 if (optlen < sizeof(int))
3654 if (get_user(val, (int __user *)optval))
3656 if (!sctp_is_ep_boundall(sk) && val)
3658 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3661 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3662 if (val == 0 && sp->do_auto_asconf) {
3663 list_del(&sp->auto_asconf_list);
3664 sp->do_auto_asconf = 0;
3665 } else if (val && !sp->do_auto_asconf) {
3666 list_add_tail(&sp->auto_asconf_list,
3667 &sock_net(sk)->sctp.auto_asconf_splist);
3668 sp->do_auto_asconf = 1;
3670 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3675 * SCTP_PEER_ADDR_THLDS
3677 * This option allows us to alter the partially failed threshold for one or all
3678 * transports in an association. See Section 6.1 of:
3679 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3681 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3682 char __user *optval,
3683 unsigned int optlen)
3685 struct sctp_paddrthlds val;
3686 struct sctp_transport *trans;
3687 struct sctp_association *asoc;
3689 if (optlen < sizeof(struct sctp_paddrthlds))
3691 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3692 sizeof(struct sctp_paddrthlds)))
3696 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3697 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3700 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3702 if (val.spt_pathmaxrxt)
3703 trans->pathmaxrxt = val.spt_pathmaxrxt;
3704 trans->pf_retrans = val.spt_pathpfthld;
3707 if (val.spt_pathmaxrxt)
3708 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3709 asoc->pf_retrans = val.spt_pathpfthld;
3711 trans = sctp_addr_id2transport(sk, &val.spt_address,
3716 if (val.spt_pathmaxrxt)
3717 trans->pathmaxrxt = val.spt_pathmaxrxt;
3718 trans->pf_retrans = val.spt_pathpfthld;
3724 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3725 char __user *optval,
3726 unsigned int optlen)
3730 if (optlen < sizeof(int))
3732 if (get_user(val, (int __user *) optval))
3735 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3740 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3741 char __user *optval,
3742 unsigned int optlen)
3746 if (optlen < sizeof(int))
3748 if (get_user(val, (int __user *) optval))
3751 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3756 static int sctp_setsockopt_pr_supported(struct sock *sk,
3757 char __user *optval,
3758 unsigned int optlen)
3760 struct sctp_assoc_value params;
3761 struct sctp_association *asoc;
3762 int retval = -EINVAL;
3764 if (optlen != sizeof(params))
3767 if (copy_from_user(¶ms, optval, optlen)) {
3772 asoc = sctp_id2assoc(sk, params.assoc_id);
3774 asoc->prsctp_enable = !!params.assoc_value;
3775 } else if (!params.assoc_id) {
3776 struct sctp_sock *sp = sctp_sk(sk);
3778 sp->ep->prsctp_enable = !!params.assoc_value;
3789 static int sctp_setsockopt_default_prinfo(struct sock *sk,
3790 char __user *optval,
3791 unsigned int optlen)
3793 struct sctp_default_prinfo info;
3794 struct sctp_association *asoc;
3795 int retval = -EINVAL;
3797 if (optlen != sizeof(info))
3800 if (copy_from_user(&info, optval, sizeof(info))) {
3805 if (info.pr_policy & ~SCTP_PR_SCTP_MASK)
3808 if (info.pr_policy == SCTP_PR_SCTP_NONE)
3811 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
3813 SCTP_PR_SET_POLICY(asoc->default_flags, info.pr_policy);
3814 asoc->default_timetolive = info.pr_value;
3815 } else if (!info.pr_assoc_id) {
3816 struct sctp_sock *sp = sctp_sk(sk);
3818 SCTP_PR_SET_POLICY(sp->default_flags, info.pr_policy);
3819 sp->default_timetolive = info.pr_value;
3830 static int sctp_setsockopt_reconfig_supported(struct sock *sk,
3831 char __user *optval,
3832 unsigned int optlen)
3834 struct sctp_assoc_value params;
3835 struct sctp_association *asoc;
3836 int retval = -EINVAL;
3838 if (optlen != sizeof(params))
3841 if (copy_from_user(¶ms, optval, optlen)) {
3846 asoc = sctp_id2assoc(sk, params.assoc_id);
3848 asoc->reconf_enable = !!params.assoc_value;
3849 } else if (!params.assoc_id) {
3850 struct sctp_sock *sp = sctp_sk(sk);
3852 sp->ep->reconf_enable = !!params.assoc_value;
3863 static int sctp_setsockopt_enable_strreset(struct sock *sk,
3864 char __user *optval,
3865 unsigned int optlen)
3867 struct sctp_assoc_value params;
3868 struct sctp_association *asoc;
3869 int retval = -EINVAL;
3871 if (optlen != sizeof(params))
3874 if (copy_from_user(¶ms, optval, optlen)) {
3879 if (params.assoc_value & (~SCTP_ENABLE_STRRESET_MASK))
3882 asoc = sctp_id2assoc(sk, params.assoc_id);
3884 asoc->strreset_enable = params.assoc_value;
3885 } else if (!params.assoc_id) {
3886 struct sctp_sock *sp = sctp_sk(sk);
3888 sp->ep->strreset_enable = params.assoc_value;
3899 static int sctp_setsockopt_reset_streams(struct sock *sk,
3900 char __user *optval,
3901 unsigned int optlen)
3903 struct sctp_reset_streams *params;
3904 struct sctp_association *asoc;
3905 int retval = -EINVAL;
3907 if (optlen < sizeof(*params))
3909 /* srs_number_streams is u16, so optlen can't be bigger than this. */
3910 optlen = min_t(unsigned int, optlen, USHRT_MAX +
3911 sizeof(__u16) * sizeof(*params));
3913 params = memdup_user(optval, optlen);
3915 return PTR_ERR(params);
3917 if (params->srs_number_streams * sizeof(__u16) >
3918 optlen - sizeof(*params))
3921 asoc = sctp_id2assoc(sk, params->srs_assoc_id);
3925 retval = sctp_send_reset_streams(asoc, params);
3932 static int sctp_setsockopt_reset_assoc(struct sock *sk,
3933 char __user *optval,
3934 unsigned int optlen)
3936 struct sctp_association *asoc;
3937 sctp_assoc_t associd;
3938 int retval = -EINVAL;
3940 if (optlen != sizeof(associd))
3943 if (copy_from_user(&associd, optval, optlen)) {
3948 asoc = sctp_id2assoc(sk, associd);
3952 retval = sctp_send_reset_assoc(asoc);
3958 static int sctp_setsockopt_add_streams(struct sock *sk,
3959 char __user *optval,
3960 unsigned int optlen)
3962 struct sctp_association *asoc;
3963 struct sctp_add_streams params;
3964 int retval = -EINVAL;
3966 if (optlen != sizeof(params))
3969 if (copy_from_user(¶ms, optval, optlen)) {
3974 asoc = sctp_id2assoc(sk, params.sas_assoc_id);
3978 retval = sctp_send_add_streams(asoc, ¶ms);
3984 static int sctp_setsockopt_scheduler(struct sock *sk,
3985 char __user *optval,
3986 unsigned int optlen)
3988 struct sctp_association *asoc;
3989 struct sctp_assoc_value params;
3990 int retval = -EINVAL;
3992 if (optlen < sizeof(params))
3995 optlen = sizeof(params);
3996 if (copy_from_user(¶ms, optval, optlen)) {
4001 if (params.assoc_value > SCTP_SS_MAX)
4004 asoc = sctp_id2assoc(sk, params.assoc_id);
4008 retval = sctp_sched_set_sched(asoc, params.assoc_value);
4014 static int sctp_setsockopt_scheduler_value(struct sock *sk,
4015 char __user *optval,
4016 unsigned int optlen)
4018 struct sctp_association *asoc;
4019 struct sctp_stream_value params;
4020 int retval = -EINVAL;
4022 if (optlen < sizeof(params))
4025 optlen = sizeof(params);
4026 if (copy_from_user(¶ms, optval, optlen)) {
4031 asoc = sctp_id2assoc(sk, params.assoc_id);
4035 retval = sctp_sched_set_value(asoc, params.stream_id,
4036 params.stream_value, GFP_KERNEL);
4042 /* API 6.2 setsockopt(), getsockopt()
4044 * Applications use setsockopt() and getsockopt() to set or retrieve
4045 * socket options. Socket options are used to change the default
4046 * behavior of sockets calls. They are described in Section 7.
4050 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
4051 * int __user *optlen);
4052 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
4055 * sd - the socket descript.
4056 * level - set to IPPROTO_SCTP for all SCTP options.
4057 * optname - the option name.
4058 * optval - the buffer to store the value of the option.
4059 * optlen - the size of the buffer.
4061 static int sctp_setsockopt(struct sock *sk, int level, int optname,
4062 char __user *optval, unsigned int optlen)
4066 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
4068 /* I can hardly begin to describe how wrong this is. This is
4069 * so broken as to be worse than useless. The API draft
4070 * REALLY is NOT helpful here... I am not convinced that the
4071 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
4072 * are at all well-founded.
4074 if (level != SOL_SCTP) {
4075 struct sctp_af *af = sctp_sk(sk)->pf->af;
4076 retval = af->setsockopt(sk, level, optname, optval, optlen);
4083 case SCTP_SOCKOPT_BINDX_ADD:
4084 /* 'optlen' is the size of the addresses buffer. */
4085 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
4086 optlen, SCTP_BINDX_ADD_ADDR);
4089 case SCTP_SOCKOPT_BINDX_REM:
4090 /* 'optlen' is the size of the addresses buffer. */
4091 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
4092 optlen, SCTP_BINDX_REM_ADDR);
4095 case SCTP_SOCKOPT_CONNECTX_OLD:
4096 /* 'optlen' is the size of the addresses buffer. */
4097 retval = sctp_setsockopt_connectx_old(sk,
4098 (struct sockaddr __user *)optval,
4102 case SCTP_SOCKOPT_CONNECTX:
4103 /* 'optlen' is the size of the addresses buffer. */
4104 retval = sctp_setsockopt_connectx(sk,
4105 (struct sockaddr __user *)optval,
4109 case SCTP_DISABLE_FRAGMENTS:
4110 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
4114 retval = sctp_setsockopt_events(sk, optval, optlen);
4117 case SCTP_AUTOCLOSE:
4118 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
4121 case SCTP_PEER_ADDR_PARAMS:
4122 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
4125 case SCTP_DELAYED_SACK:
4126 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
4128 case SCTP_PARTIAL_DELIVERY_POINT:
4129 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
4133 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
4135 case SCTP_DEFAULT_SEND_PARAM:
4136 retval = sctp_setsockopt_default_send_param(sk, optval,
4139 case SCTP_DEFAULT_SNDINFO:
4140 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
4142 case SCTP_PRIMARY_ADDR:
4143 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
4145 case SCTP_SET_PEER_PRIMARY_ADDR:
4146 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
4149 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
4152 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
4154 case SCTP_ASSOCINFO:
4155 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
4157 case SCTP_I_WANT_MAPPED_V4_ADDR:
4158 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
4161 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
4163 case SCTP_ADAPTATION_LAYER:
4164 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
4167 retval = sctp_setsockopt_context(sk, optval, optlen);
4169 case SCTP_FRAGMENT_INTERLEAVE:
4170 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
4172 case SCTP_MAX_BURST:
4173 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
4175 case SCTP_AUTH_CHUNK:
4176 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
4178 case SCTP_HMAC_IDENT:
4179 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
4182 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
4184 case SCTP_AUTH_ACTIVE_KEY:
4185 retval = sctp_setsockopt_active_key(sk, optval, optlen);
4187 case SCTP_AUTH_DELETE_KEY:
4188 retval = sctp_setsockopt_del_key(sk, optval, optlen);
4190 case SCTP_AUTO_ASCONF:
4191 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
4193 case SCTP_PEER_ADDR_THLDS:
4194 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
4196 case SCTP_RECVRCVINFO:
4197 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
4199 case SCTP_RECVNXTINFO:
4200 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
4202 case SCTP_PR_SUPPORTED:
4203 retval = sctp_setsockopt_pr_supported(sk, optval, optlen);
4205 case SCTP_DEFAULT_PRINFO:
4206 retval = sctp_setsockopt_default_prinfo(sk, optval, optlen);
4208 case SCTP_RECONFIG_SUPPORTED:
4209 retval = sctp_setsockopt_reconfig_supported(sk, optval, optlen);
4211 case SCTP_ENABLE_STREAM_RESET:
4212 retval = sctp_setsockopt_enable_strreset(sk, optval, optlen);
4214 case SCTP_RESET_STREAMS:
4215 retval = sctp_setsockopt_reset_streams(sk, optval, optlen);
4217 case SCTP_RESET_ASSOC:
4218 retval = sctp_setsockopt_reset_assoc(sk, optval, optlen);
4220 case SCTP_ADD_STREAMS:
4221 retval = sctp_setsockopt_add_streams(sk, optval, optlen);
4223 case SCTP_STREAM_SCHEDULER:
4224 retval = sctp_setsockopt_scheduler(sk, optval, optlen);
4226 case SCTP_STREAM_SCHEDULER_VALUE:
4227 retval = sctp_setsockopt_scheduler_value(sk, optval, optlen);
4230 retval = -ENOPROTOOPT;
4240 /* API 3.1.6 connect() - UDP Style Syntax
4242 * An application may use the connect() call in the UDP model to initiate an
4243 * association without sending data.
4247 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
4249 * sd: the socket descriptor to have a new association added to.
4251 * nam: the address structure (either struct sockaddr_in or struct
4252 * sockaddr_in6 defined in RFC2553 [7]).
4254 * len: the size of the address.
4256 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
4264 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
4267 /* Validate addr_len before calling common connect/connectx routine. */
4268 af = sctp_get_af_specific(addr->sa_family);
4269 if (!af || addr_len < af->sockaddr_len) {
4272 /* Pass correct addr len to common routine (so it knows there
4273 * is only one address being passed.
4275 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
4282 /* FIXME: Write comments. */
4283 static int sctp_disconnect(struct sock *sk, int flags)
4285 return -EOPNOTSUPP; /* STUB */
4288 /* 4.1.4 accept() - TCP Style Syntax
4290 * Applications use accept() call to remove an established SCTP
4291 * association from the accept queue of the endpoint. A new socket
4292 * descriptor will be returned from accept() to represent the newly
4293 * formed association.
4295 static struct sock *sctp_accept(struct sock *sk, int flags, int *err, bool kern)
4297 struct sctp_sock *sp;
4298 struct sctp_endpoint *ep;
4299 struct sock *newsk = NULL;
4300 struct sctp_association *asoc;
4309 if (!sctp_style(sk, TCP)) {
4310 error = -EOPNOTSUPP;
4314 if (!sctp_sstate(sk, LISTENING)) {
4319 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
4321 error = sctp_wait_for_accept(sk, timeo);
4325 /* We treat the list of associations on the endpoint as the accept
4326 * queue and pick the first association on the list.
4328 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
4330 newsk = sp->pf->create_accept_sk(sk, asoc, kern);
4336 /* Populate the fields of the newsk from the oldsk and migrate the
4337 * asoc to the newsk.
4339 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
4347 /* The SCTP ioctl handler. */
4348 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
4355 * SEQPACKET-style sockets in LISTENING state are valid, for
4356 * SCTP, so only discard TCP-style sockets in LISTENING state.
4358 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4363 struct sk_buff *skb;
4364 unsigned int amount = 0;
4366 skb = skb_peek(&sk->sk_receive_queue);
4369 * We will only return the amount of this packet since
4370 * that is all that will be read.
4374 rc = put_user(amount, (int __user *)arg);
4386 /* This is the function which gets called during socket creation to
4387 * initialized the SCTP-specific portion of the sock.
4388 * The sock structure should already be zero-filled memory.
4390 static int sctp_init_sock(struct sock *sk)
4392 struct net *net = sock_net(sk);
4393 struct sctp_sock *sp;
4395 pr_debug("%s: sk:%p\n", __func__, sk);
4399 /* Initialize the SCTP per socket area. */
4400 switch (sk->sk_type) {
4401 case SOCK_SEQPACKET:
4402 sp->type = SCTP_SOCKET_UDP;
4405 sp->type = SCTP_SOCKET_TCP;
4408 return -ESOCKTNOSUPPORT;
4411 sk->sk_gso_type = SKB_GSO_SCTP;
4413 /* Initialize default send parameters. These parameters can be
4414 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4416 sp->default_stream = 0;
4417 sp->default_ppid = 0;
4418 sp->default_flags = 0;
4419 sp->default_context = 0;
4420 sp->default_timetolive = 0;
4422 sp->default_rcv_context = 0;
4423 sp->max_burst = net->sctp.max_burst;
4425 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4427 /* Initialize default setup parameters. These parameters
4428 * can be modified with the SCTP_INITMSG socket option or
4429 * overridden by the SCTP_INIT CMSG.
4431 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4432 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4433 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4434 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4436 /* Initialize default RTO related parameters. These parameters can
4437 * be modified for with the SCTP_RTOINFO socket option.
4439 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4440 sp->rtoinfo.srto_max = net->sctp.rto_max;
4441 sp->rtoinfo.srto_min = net->sctp.rto_min;
4443 /* Initialize default association related parameters. These parameters
4444 * can be modified with the SCTP_ASSOCINFO socket option.
4446 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4447 sp->assocparams.sasoc_number_peer_destinations = 0;
4448 sp->assocparams.sasoc_peer_rwnd = 0;
4449 sp->assocparams.sasoc_local_rwnd = 0;
4450 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4452 /* Initialize default event subscriptions. By default, all the
4455 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4457 /* Default Peer Address Parameters. These defaults can
4458 * be modified via SCTP_PEER_ADDR_PARAMS
4460 sp->hbinterval = net->sctp.hb_interval;
4461 sp->pathmaxrxt = net->sctp.max_retrans_path;
4462 sp->pathmtu = 0; /* allow default discovery */
4463 sp->sackdelay = net->sctp.sack_timeout;
4465 sp->param_flags = SPP_HB_ENABLE |
4467 SPP_SACKDELAY_ENABLE;
4469 /* If enabled no SCTP message fragmentation will be performed.
4470 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4472 sp->disable_fragments = 0;
4474 /* Enable Nagle algorithm by default. */
4477 sp->recvrcvinfo = 0;
4478 sp->recvnxtinfo = 0;
4480 /* Enable by default. */
4483 /* Auto-close idle associations after the configured
4484 * number of seconds. A value of 0 disables this
4485 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4486 * for UDP-style sockets only.
4490 /* User specified fragmentation limit. */
4493 sp->adaptation_ind = 0;
4495 sp->pf = sctp_get_pf_specific(sk->sk_family);
4497 /* Control variables for partial data delivery. */
4498 atomic_set(&sp->pd_mode, 0);
4499 skb_queue_head_init(&sp->pd_lobby);
4500 sp->frag_interleave = 0;
4502 /* Create a per socket endpoint structure. Even if we
4503 * change the data structure relationships, this may still
4504 * be useful for storing pre-connect address information.
4506 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4512 sk->sk_destruct = sctp_destruct_sock;
4514 SCTP_DBG_OBJCNT_INC(sock);
4517 sk_sockets_allocated_inc(sk);
4518 sock_prot_inuse_add(net, sk->sk_prot, 1);
4520 /* Nothing can fail after this block, otherwise
4521 * sctp_destroy_sock() will be called without addr_wq_lock held
4523 if (net->sctp.default_auto_asconf) {
4524 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
4525 list_add_tail(&sp->auto_asconf_list,
4526 &net->sctp.auto_asconf_splist);
4527 sp->do_auto_asconf = 1;
4528 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4530 sp->do_auto_asconf = 0;
4538 /* Cleanup any SCTP per socket resources. Must be called with
4539 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4541 static void sctp_destroy_sock(struct sock *sk)
4543 struct sctp_sock *sp;
4545 pr_debug("%s: sk:%p\n", __func__, sk);
4547 /* Release our hold on the endpoint. */
4549 /* This could happen during socket init, thus we bail out
4550 * early, since the rest of the below is not setup either.
4555 if (sp->do_auto_asconf) {
4556 sp->do_auto_asconf = 0;
4557 list_del(&sp->auto_asconf_list);
4559 sctp_endpoint_free(sp->ep);
4561 sk_sockets_allocated_dec(sk);
4562 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4566 /* Triggered when there are no references on the socket anymore */
4567 static void sctp_destruct_sock(struct sock *sk)
4569 struct sctp_sock *sp = sctp_sk(sk);
4571 /* Free up the HMAC transform. */
4572 crypto_free_shash(sp->hmac);
4574 inet_sock_destruct(sk);
4577 /* API 4.1.7 shutdown() - TCP Style Syntax
4578 * int shutdown(int socket, int how);
4580 * sd - the socket descriptor of the association to be closed.
4581 * how - Specifies the type of shutdown. The values are
4584 * Disables further receive operations. No SCTP
4585 * protocol action is taken.
4587 * Disables further send operations, and initiates
4588 * the SCTP shutdown sequence.
4590 * Disables further send and receive operations
4591 * and initiates the SCTP shutdown sequence.
4593 static void sctp_shutdown(struct sock *sk, int how)
4595 struct net *net = sock_net(sk);
4596 struct sctp_endpoint *ep;
4598 if (!sctp_style(sk, TCP))
4601 ep = sctp_sk(sk)->ep;
4602 if (how & SEND_SHUTDOWN && !list_empty(&ep->asocs)) {
4603 struct sctp_association *asoc;
4605 sk->sk_state = SCTP_SS_CLOSING;
4606 asoc = list_entry(ep->asocs.next,
4607 struct sctp_association, asocs);
4608 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4612 int sctp_get_sctp_info(struct sock *sk, struct sctp_association *asoc,
4613 struct sctp_info *info)
4615 struct sctp_transport *prim;
4616 struct list_head *pos;
4619 memset(info, 0, sizeof(*info));
4621 struct sctp_sock *sp = sctp_sk(sk);
4623 info->sctpi_s_autoclose = sp->autoclose;
4624 info->sctpi_s_adaptation_ind = sp->adaptation_ind;
4625 info->sctpi_s_pd_point = sp->pd_point;
4626 info->sctpi_s_nodelay = sp->nodelay;
4627 info->sctpi_s_disable_fragments = sp->disable_fragments;
4628 info->sctpi_s_v4mapped = sp->v4mapped;
4629 info->sctpi_s_frag_interleave = sp->frag_interleave;
4630 info->sctpi_s_type = sp->type;
4635 info->sctpi_tag = asoc->c.my_vtag;
4636 info->sctpi_state = asoc->state;
4637 info->sctpi_rwnd = asoc->a_rwnd;
4638 info->sctpi_unackdata = asoc->unack_data;
4639 info->sctpi_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4640 info->sctpi_instrms = asoc->stream.incnt;
4641 info->sctpi_outstrms = asoc->stream.outcnt;
4642 list_for_each(pos, &asoc->base.inqueue.in_chunk_list)
4643 info->sctpi_inqueue++;
4644 list_for_each(pos, &asoc->outqueue.out_chunk_list)
4645 info->sctpi_outqueue++;
4646 info->sctpi_overall_error = asoc->overall_error_count;
4647 info->sctpi_max_burst = asoc->max_burst;
4648 info->sctpi_maxseg = asoc->frag_point;
4649 info->sctpi_peer_rwnd = asoc->peer.rwnd;
4650 info->sctpi_peer_tag = asoc->c.peer_vtag;
4652 mask = asoc->peer.ecn_capable << 1;
4653 mask = (mask | asoc->peer.ipv4_address) << 1;
4654 mask = (mask | asoc->peer.ipv6_address) << 1;
4655 mask = (mask | asoc->peer.hostname_address) << 1;
4656 mask = (mask | asoc->peer.asconf_capable) << 1;
4657 mask = (mask | asoc->peer.prsctp_capable) << 1;
4658 mask = (mask | asoc->peer.auth_capable);
4659 info->sctpi_peer_capable = mask;
4660 mask = asoc->peer.sack_needed << 1;
4661 mask = (mask | asoc->peer.sack_generation) << 1;
4662 mask = (mask | asoc->peer.zero_window_announced);
4663 info->sctpi_peer_sack = mask;
4665 info->sctpi_isacks = asoc->stats.isacks;
4666 info->sctpi_osacks = asoc->stats.osacks;
4667 info->sctpi_opackets = asoc->stats.opackets;
4668 info->sctpi_ipackets = asoc->stats.ipackets;
4669 info->sctpi_rtxchunks = asoc->stats.rtxchunks;
4670 info->sctpi_outofseqtsns = asoc->stats.outofseqtsns;
4671 info->sctpi_idupchunks = asoc->stats.idupchunks;
4672 info->sctpi_gapcnt = asoc->stats.gapcnt;
4673 info->sctpi_ouodchunks = asoc->stats.ouodchunks;
4674 info->sctpi_iuodchunks = asoc->stats.iuodchunks;
4675 info->sctpi_oodchunks = asoc->stats.oodchunks;
4676 info->sctpi_iodchunks = asoc->stats.iodchunks;
4677 info->sctpi_octrlchunks = asoc->stats.octrlchunks;
4678 info->sctpi_ictrlchunks = asoc->stats.ictrlchunks;
4680 prim = asoc->peer.primary_path;
4681 memcpy(&info->sctpi_p_address, &prim->ipaddr, sizeof(prim->ipaddr));
4682 info->sctpi_p_state = prim->state;
4683 info->sctpi_p_cwnd = prim->cwnd;
4684 info->sctpi_p_srtt = prim->srtt;
4685 info->sctpi_p_rto = jiffies_to_msecs(prim->rto);
4686 info->sctpi_p_hbinterval = prim->hbinterval;
4687 info->sctpi_p_pathmaxrxt = prim->pathmaxrxt;
4688 info->sctpi_p_sackdelay = jiffies_to_msecs(prim->sackdelay);
4689 info->sctpi_p_ssthresh = prim->ssthresh;
4690 info->sctpi_p_partial_bytes_acked = prim->partial_bytes_acked;
4691 info->sctpi_p_flight_size = prim->flight_size;
4692 info->sctpi_p_error = prim->error_count;
4696 EXPORT_SYMBOL_GPL(sctp_get_sctp_info);
4698 /* use callback to avoid exporting the core structure */
4699 int sctp_transport_walk_start(struct rhashtable_iter *iter)
4703 rhltable_walk_enter(&sctp_transport_hashtable, iter);
4705 err = rhashtable_walk_start(iter);
4706 if (err && err != -EAGAIN) {
4707 rhashtable_walk_stop(iter);
4708 rhashtable_walk_exit(iter);
4715 void sctp_transport_walk_stop(struct rhashtable_iter *iter)
4717 rhashtable_walk_stop(iter);
4718 rhashtable_walk_exit(iter);
4721 struct sctp_transport *sctp_transport_get_next(struct net *net,
4722 struct rhashtable_iter *iter)
4724 struct sctp_transport *t;
4726 t = rhashtable_walk_next(iter);
4727 for (; t; t = rhashtable_walk_next(iter)) {
4729 if (PTR_ERR(t) == -EAGAIN)
4734 if (net_eq(sock_net(t->asoc->base.sk), net) &&
4735 t->asoc->peer.primary_path == t)
4742 struct sctp_transport *sctp_transport_get_idx(struct net *net,
4743 struct rhashtable_iter *iter,
4746 void *obj = SEQ_START_TOKEN;
4748 while (pos && (obj = sctp_transport_get_next(net, iter)) &&
4755 int sctp_for_each_endpoint(int (*cb)(struct sctp_endpoint *, void *),
4759 struct sctp_ep_common *epb;
4760 struct sctp_hashbucket *head;
4762 for (head = sctp_ep_hashtable; hash < sctp_ep_hashsize;
4764 read_lock_bh(&head->lock);
4765 sctp_for_each_hentry(epb, &head->chain) {
4766 err = cb(sctp_ep(epb), p);
4770 read_unlock_bh(&head->lock);
4775 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint);
4777 int sctp_transport_lookup_process(int (*cb)(struct sctp_transport *, void *),
4779 const union sctp_addr *laddr,
4780 const union sctp_addr *paddr, void *p)
4782 struct sctp_transport *transport;
4786 transport = sctp_addrs_lookup_transport(net, laddr, paddr);
4791 err = cb(transport, p);
4792 sctp_transport_put(transport);
4796 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process);
4798 int sctp_for_each_transport(int (*cb)(struct sctp_transport *, void *),
4799 int (*cb_done)(struct sctp_transport *, void *),
4800 struct net *net, int *pos, void *p) {
4801 struct rhashtable_iter hti;
4802 struct sctp_transport *tsp;
4806 ret = sctp_transport_walk_start(&hti);
4810 tsp = sctp_transport_get_idx(net, &hti, *pos + 1);
4811 for (; !IS_ERR_OR_NULL(tsp); tsp = sctp_transport_get_next(net, &hti)) {
4812 if (!sctp_transport_hold(tsp))
4818 sctp_transport_put(tsp);
4820 sctp_transport_walk_stop(&hti);
4823 if (cb_done && !cb_done(tsp, p)) {
4825 sctp_transport_put(tsp);
4828 sctp_transport_put(tsp);
4833 EXPORT_SYMBOL_GPL(sctp_for_each_transport);
4835 /* 7.2.1 Association Status (SCTP_STATUS)
4837 * Applications can retrieve current status information about an
4838 * association, including association state, peer receiver window size,
4839 * number of unacked data chunks, and number of data chunks pending
4840 * receipt. This information is read-only.
4842 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4843 char __user *optval,
4846 struct sctp_status status;
4847 struct sctp_association *asoc = NULL;
4848 struct sctp_transport *transport;
4849 sctp_assoc_t associd;
4852 if (len < sizeof(status)) {
4857 len = sizeof(status);
4858 if (copy_from_user(&status, optval, len)) {
4863 associd = status.sstat_assoc_id;
4864 asoc = sctp_id2assoc(sk, associd);
4870 transport = asoc->peer.primary_path;
4872 status.sstat_assoc_id = sctp_assoc2id(asoc);
4873 status.sstat_state = sctp_assoc_to_state(asoc);
4874 status.sstat_rwnd = asoc->peer.rwnd;
4875 status.sstat_unackdata = asoc->unack_data;
4877 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4878 status.sstat_instrms = asoc->stream.incnt;
4879 status.sstat_outstrms = asoc->stream.outcnt;
4880 status.sstat_fragmentation_point = asoc->frag_point;
4881 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4882 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4883 transport->af_specific->sockaddr_len);
4884 /* Map ipv4 address into v4-mapped-on-v6 address. */
4885 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
4886 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4887 status.sstat_primary.spinfo_state = transport->state;
4888 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4889 status.sstat_primary.spinfo_srtt = transport->srtt;
4890 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4891 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4893 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4894 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4896 if (put_user(len, optlen)) {
4901 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4902 __func__, len, status.sstat_state, status.sstat_rwnd,
4903 status.sstat_assoc_id);
4905 if (copy_to_user(optval, &status, len)) {
4915 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4917 * Applications can retrieve information about a specific peer address
4918 * of an association, including its reachability state, congestion
4919 * window, and retransmission timer values. This information is
4922 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4923 char __user *optval,
4926 struct sctp_paddrinfo pinfo;
4927 struct sctp_transport *transport;
4930 if (len < sizeof(pinfo)) {
4935 len = sizeof(pinfo);
4936 if (copy_from_user(&pinfo, optval, len)) {
4941 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4942 pinfo.spinfo_assoc_id);
4946 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4947 pinfo.spinfo_state = transport->state;
4948 pinfo.spinfo_cwnd = transport->cwnd;
4949 pinfo.spinfo_srtt = transport->srtt;
4950 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4951 pinfo.spinfo_mtu = transport->pathmtu;
4953 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4954 pinfo.spinfo_state = SCTP_ACTIVE;
4956 if (put_user(len, optlen)) {
4961 if (copy_to_user(optval, &pinfo, len)) {
4970 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4972 * This option is a on/off flag. If enabled no SCTP message
4973 * fragmentation will be performed. Instead if a message being sent
4974 * exceeds the current PMTU size, the message will NOT be sent and
4975 * instead a error will be indicated to the user.
4977 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4978 char __user *optval, int __user *optlen)
4982 if (len < sizeof(int))
4986 val = (sctp_sk(sk)->disable_fragments == 1);
4987 if (put_user(len, optlen))
4989 if (copy_to_user(optval, &val, len))
4994 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4996 * This socket option is used to specify various notifications and
4997 * ancillary data the user wishes to receive.
4999 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
5004 if (len > sizeof(struct sctp_event_subscribe))
5005 len = sizeof(struct sctp_event_subscribe);
5006 if (put_user(len, optlen))
5008 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
5013 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
5015 * This socket option is applicable to the UDP-style socket only. When
5016 * set it will cause associations that are idle for more than the
5017 * specified number of seconds to automatically close. An association
5018 * being idle is defined an association that has NOT sent or received
5019 * user data. The special value of '0' indicates that no automatic
5020 * close of any associations should be performed. The option expects an
5021 * integer defining the number of seconds of idle time before an
5022 * association is closed.
5024 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
5026 /* Applicable to UDP-style socket only */
5027 if (sctp_style(sk, TCP))
5029 if (len < sizeof(int))
5032 if (put_user(len, optlen))
5034 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, len))
5039 /* Helper routine to branch off an association to a new socket. */
5040 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
5042 struct sctp_association *asoc = sctp_id2assoc(sk, id);
5043 struct sctp_sock *sp = sctp_sk(sk);
5044 struct socket *sock;
5047 /* Do not peel off from one netns to another one. */
5048 if (!net_eq(current->nsproxy->net_ns, sock_net(sk)))
5054 /* An association cannot be branched off from an already peeled-off
5055 * socket, nor is this supported for tcp style sockets.
5057 if (!sctp_style(sk, UDP))
5060 /* Create a new socket. */
5061 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
5065 sctp_copy_sock(sock->sk, sk, asoc);
5067 /* Make peeled-off sockets more like 1-1 accepted sockets.
5068 * Set the daddr and initialize id to something more random
5070 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
5072 /* Populate the fields of the newsk from the oldsk and migrate the
5073 * asoc to the newsk.
5075 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
5081 EXPORT_SYMBOL(sctp_do_peeloff);
5083 static int sctp_getsockopt_peeloff_common(struct sock *sk, sctp_peeloff_arg_t *peeloff,
5084 struct file **newfile, unsigned flags)
5086 struct socket *newsock;
5089 retval = sctp_do_peeloff(sk, peeloff->associd, &newsock);
5093 /* Map the socket to an unused fd that can be returned to the user. */
5094 retval = get_unused_fd_flags(flags & SOCK_CLOEXEC);
5096 sock_release(newsock);
5100 *newfile = sock_alloc_file(newsock, 0, NULL);
5101 if (IS_ERR(*newfile)) {
5102 put_unused_fd(retval);
5103 retval = PTR_ERR(*newfile);
5108 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
5111 peeloff->sd = retval;
5113 if (flags & SOCK_NONBLOCK)
5114 (*newfile)->f_flags |= O_NONBLOCK;
5119 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
5121 sctp_peeloff_arg_t peeloff;
5122 struct file *newfile = NULL;
5125 if (len < sizeof(sctp_peeloff_arg_t))
5127 len = sizeof(sctp_peeloff_arg_t);
5128 if (copy_from_user(&peeloff, optval, len))
5131 retval = sctp_getsockopt_peeloff_common(sk, &peeloff, &newfile, 0);
5135 /* Return the fd mapped to the new socket. */
5136 if (put_user(len, optlen)) {
5138 put_unused_fd(retval);
5142 if (copy_to_user(optval, &peeloff, len)) {
5144 put_unused_fd(retval);
5147 fd_install(retval, newfile);
5152 static int sctp_getsockopt_peeloff_flags(struct sock *sk, int len,
5153 char __user *optval, int __user *optlen)
5155 sctp_peeloff_flags_arg_t peeloff;
5156 struct file *newfile = NULL;
5159 if (len < sizeof(sctp_peeloff_flags_arg_t))
5161 len = sizeof(sctp_peeloff_flags_arg_t);
5162 if (copy_from_user(&peeloff, optval, len))
5165 retval = sctp_getsockopt_peeloff_common(sk, &peeloff.p_arg,
5166 &newfile, peeloff.flags);
5170 /* Return the fd mapped to the new socket. */
5171 if (put_user(len, optlen)) {
5173 put_unused_fd(retval);
5177 if (copy_to_user(optval, &peeloff, len)) {
5179 put_unused_fd(retval);
5182 fd_install(retval, newfile);
5187 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
5189 * Applications can enable or disable heartbeats for any peer address of
5190 * an association, modify an address's heartbeat interval, force a
5191 * heartbeat to be sent immediately, and adjust the address's maximum
5192 * number of retransmissions sent before an address is considered
5193 * unreachable. The following structure is used to access and modify an
5194 * address's parameters:
5196 * struct sctp_paddrparams {
5197 * sctp_assoc_t spp_assoc_id;
5198 * struct sockaddr_storage spp_address;
5199 * uint32_t spp_hbinterval;
5200 * uint16_t spp_pathmaxrxt;
5201 * uint32_t spp_pathmtu;
5202 * uint32_t spp_sackdelay;
5203 * uint32_t spp_flags;
5206 * spp_assoc_id - (one-to-many style socket) This is filled in the
5207 * application, and identifies the association for
5209 * spp_address - This specifies which address is of interest.
5210 * spp_hbinterval - This contains the value of the heartbeat interval,
5211 * in milliseconds. If a value of zero
5212 * is present in this field then no changes are to
5213 * be made to this parameter.
5214 * spp_pathmaxrxt - This contains the maximum number of
5215 * retransmissions before this address shall be
5216 * considered unreachable. If a value of zero
5217 * is present in this field then no changes are to
5218 * be made to this parameter.
5219 * spp_pathmtu - When Path MTU discovery is disabled the value
5220 * specified here will be the "fixed" path mtu.
5221 * Note that if the spp_address field is empty
5222 * then all associations on this address will
5223 * have this fixed path mtu set upon them.
5225 * spp_sackdelay - When delayed sack is enabled, this value specifies
5226 * the number of milliseconds that sacks will be delayed
5227 * for. This value will apply to all addresses of an
5228 * association if the spp_address field is empty. Note
5229 * also, that if delayed sack is enabled and this
5230 * value is set to 0, no change is made to the last
5231 * recorded delayed sack timer value.
5233 * spp_flags - These flags are used to control various features
5234 * on an association. The flag field may contain
5235 * zero or more of the following options.
5237 * SPP_HB_ENABLE - Enable heartbeats on the
5238 * specified address. Note that if the address
5239 * field is empty all addresses for the association
5240 * have heartbeats enabled upon them.
5242 * SPP_HB_DISABLE - Disable heartbeats on the
5243 * speicifed address. Note that if the address
5244 * field is empty all addresses for the association
5245 * will have their heartbeats disabled. Note also
5246 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
5247 * mutually exclusive, only one of these two should
5248 * be specified. Enabling both fields will have
5249 * undetermined results.
5251 * SPP_HB_DEMAND - Request a user initiated heartbeat
5252 * to be made immediately.
5254 * SPP_PMTUD_ENABLE - This field will enable PMTU
5255 * discovery upon the specified address. Note that
5256 * if the address feild is empty then all addresses
5257 * on the association are effected.
5259 * SPP_PMTUD_DISABLE - This field will disable PMTU
5260 * discovery upon the specified address. Note that
5261 * if the address feild is empty then all addresses
5262 * on the association are effected. Not also that
5263 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
5264 * exclusive. Enabling both will have undetermined
5267 * SPP_SACKDELAY_ENABLE - Setting this flag turns
5268 * on delayed sack. The time specified in spp_sackdelay
5269 * is used to specify the sack delay for this address. Note
5270 * that if spp_address is empty then all addresses will
5271 * enable delayed sack and take on the sack delay
5272 * value specified in spp_sackdelay.
5273 * SPP_SACKDELAY_DISABLE - Setting this flag turns
5274 * off delayed sack. If the spp_address field is blank then
5275 * delayed sack is disabled for the entire association. Note
5276 * also that this field is mutually exclusive to
5277 * SPP_SACKDELAY_ENABLE, setting both will have undefined
5280 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
5281 char __user *optval, int __user *optlen)
5283 struct sctp_paddrparams params;
5284 struct sctp_transport *trans = NULL;
5285 struct sctp_association *asoc = NULL;
5286 struct sctp_sock *sp = sctp_sk(sk);
5288 if (len < sizeof(struct sctp_paddrparams))
5290 len = sizeof(struct sctp_paddrparams);
5291 if (copy_from_user(¶ms, optval, len))
5294 /* If an address other than INADDR_ANY is specified, and
5295 * no transport is found, then the request is invalid.
5297 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
5298 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
5299 params.spp_assoc_id);
5301 pr_debug("%s: failed no transport\n", __func__);
5306 /* Get association, if assoc_id != 0 and the socket is a one
5307 * to many style socket, and an association was not found, then
5308 * the id was invalid.
5310 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
5311 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
5312 pr_debug("%s: failed no association\n", __func__);
5317 /* Fetch transport values. */
5318 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
5319 params.spp_pathmtu = trans->pathmtu;
5320 params.spp_pathmaxrxt = trans->pathmaxrxt;
5321 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
5323 /*draft-11 doesn't say what to return in spp_flags*/
5324 params.spp_flags = trans->param_flags;
5326 /* Fetch association values. */
5327 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
5328 params.spp_pathmtu = asoc->pathmtu;
5329 params.spp_pathmaxrxt = asoc->pathmaxrxt;
5330 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
5332 /*draft-11 doesn't say what to return in spp_flags*/
5333 params.spp_flags = asoc->param_flags;
5335 /* Fetch socket values. */
5336 params.spp_hbinterval = sp->hbinterval;
5337 params.spp_pathmtu = sp->pathmtu;
5338 params.spp_sackdelay = sp->sackdelay;
5339 params.spp_pathmaxrxt = sp->pathmaxrxt;
5341 /*draft-11 doesn't say what to return in spp_flags*/
5342 params.spp_flags = sp->param_flags;
5345 if (copy_to_user(optval, ¶ms, len))
5348 if (put_user(len, optlen))
5355 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
5357 * This option will effect the way delayed acks are performed. This
5358 * option allows you to get or set the delayed ack time, in
5359 * milliseconds. It also allows changing the delayed ack frequency.
5360 * Changing the frequency to 1 disables the delayed sack algorithm. If
5361 * the assoc_id is 0, then this sets or gets the endpoints default
5362 * values. If the assoc_id field is non-zero, then the set or get
5363 * effects the specified association for the one to many model (the
5364 * assoc_id field is ignored by the one to one model). Note that if
5365 * sack_delay or sack_freq are 0 when setting this option, then the
5366 * current values will remain unchanged.
5368 * struct sctp_sack_info {
5369 * sctp_assoc_t sack_assoc_id;
5370 * uint32_t sack_delay;
5371 * uint32_t sack_freq;
5374 * sack_assoc_id - This parameter, indicates which association the user
5375 * is performing an action upon. Note that if this field's value is
5376 * zero then the endpoints default value is changed (effecting future
5377 * associations only).
5379 * sack_delay - This parameter contains the number of milliseconds that
5380 * the user is requesting the delayed ACK timer be set to. Note that
5381 * this value is defined in the standard to be between 200 and 500
5384 * sack_freq - This parameter contains the number of packets that must
5385 * be received before a sack is sent without waiting for the delay
5386 * timer to expire. The default value for this is 2, setting this
5387 * value to 1 will disable the delayed sack algorithm.
5389 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
5390 char __user *optval,
5393 struct sctp_sack_info params;
5394 struct sctp_association *asoc = NULL;
5395 struct sctp_sock *sp = sctp_sk(sk);
5397 if (len >= sizeof(struct sctp_sack_info)) {
5398 len = sizeof(struct sctp_sack_info);
5400 if (copy_from_user(¶ms, optval, len))
5402 } else if (len == sizeof(struct sctp_assoc_value)) {
5403 pr_warn_ratelimited(DEPRECATED
5405 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
5406 "Use struct sctp_sack_info instead\n",
5407 current->comm, task_pid_nr(current));
5408 if (copy_from_user(¶ms, optval, len))
5413 /* Get association, if sack_assoc_id != 0 and the socket is a one
5414 * to many style socket, and an association was not found, then
5415 * the id was invalid.
5417 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
5418 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
5422 /* Fetch association values. */
5423 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
5424 params.sack_delay = jiffies_to_msecs(
5426 params.sack_freq = asoc->sackfreq;
5429 params.sack_delay = 0;
5430 params.sack_freq = 1;
5433 /* Fetch socket values. */
5434 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
5435 params.sack_delay = sp->sackdelay;
5436 params.sack_freq = sp->sackfreq;
5438 params.sack_delay = 0;
5439 params.sack_freq = 1;
5443 if (copy_to_user(optval, ¶ms, len))
5446 if (put_user(len, optlen))
5452 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5454 * Applications can specify protocol parameters for the default association
5455 * initialization. The option name argument to setsockopt() and getsockopt()
5458 * Setting initialization parameters is effective only on an unconnected
5459 * socket (for UDP-style sockets only future associations are effected
5460 * by the change). With TCP-style sockets, this option is inherited by
5461 * sockets derived from a listener socket.
5463 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
5465 if (len < sizeof(struct sctp_initmsg))
5467 len = sizeof(struct sctp_initmsg);
5468 if (put_user(len, optlen))
5470 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
5476 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
5477 char __user *optval, int __user *optlen)
5479 struct sctp_association *asoc;
5481 struct sctp_getaddrs getaddrs;
5482 struct sctp_transport *from;
5484 union sctp_addr temp;
5485 struct sctp_sock *sp = sctp_sk(sk);
5490 if (len < sizeof(struct sctp_getaddrs))
5493 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5496 /* For UDP-style sockets, id specifies the association to query. */
5497 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5501 to = optval + offsetof(struct sctp_getaddrs, addrs);
5502 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5504 list_for_each_entry(from, &asoc->peer.transport_addr_list,
5506 memcpy(&temp, &from->ipaddr, sizeof(temp));
5507 addrlen = sctp_get_pf_specific(sk->sk_family)
5508 ->addr_to_user(sp, &temp);
5509 if (space_left < addrlen)
5511 if (copy_to_user(to, &temp, addrlen))
5515 space_left -= addrlen;
5518 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
5520 bytes_copied = ((char __user *)to) - optval;
5521 if (put_user(bytes_copied, optlen))
5527 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
5528 size_t space_left, int *bytes_copied)
5530 struct sctp_sockaddr_entry *addr;
5531 union sctp_addr temp;
5534 struct net *net = sock_net(sk);
5537 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
5541 if ((PF_INET == sk->sk_family) &&
5542 (AF_INET6 == addr->a.sa.sa_family))
5544 if ((PF_INET6 == sk->sk_family) &&
5545 inet_v6_ipv6only(sk) &&
5546 (AF_INET == addr->a.sa.sa_family))
5548 memcpy(&temp, &addr->a, sizeof(temp));
5549 if (!temp.v4.sin_port)
5550 temp.v4.sin_port = htons(port);
5552 addrlen = sctp_get_pf_specific(sk->sk_family)
5553 ->addr_to_user(sctp_sk(sk), &temp);
5555 if (space_left < addrlen) {
5559 memcpy(to, &temp, addrlen);
5563 space_left -= addrlen;
5564 *bytes_copied += addrlen;
5572 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
5573 char __user *optval, int __user *optlen)
5575 struct sctp_bind_addr *bp;
5576 struct sctp_association *asoc;
5578 struct sctp_getaddrs getaddrs;
5579 struct sctp_sockaddr_entry *addr;
5581 union sctp_addr temp;
5582 struct sctp_sock *sp = sctp_sk(sk);
5586 int bytes_copied = 0;
5590 if (len < sizeof(struct sctp_getaddrs))
5593 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5597 * For UDP-style sockets, id specifies the association to query.
5598 * If the id field is set to the value '0' then the locally bound
5599 * addresses are returned without regard to any particular
5602 if (0 == getaddrs.assoc_id) {
5603 bp = &sctp_sk(sk)->ep->base.bind_addr;
5605 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5608 bp = &asoc->base.bind_addr;
5611 to = optval + offsetof(struct sctp_getaddrs, addrs);
5612 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5614 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN);
5618 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5619 * addresses from the global local address list.
5621 if (sctp_list_single_entry(&bp->address_list)) {
5622 addr = list_entry(bp->address_list.next,
5623 struct sctp_sockaddr_entry, list);
5624 if (sctp_is_any(sk, &addr->a)) {
5625 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
5626 space_left, &bytes_copied);
5636 /* Protection on the bound address list is not needed since
5637 * in the socket option context we hold a socket lock and
5638 * thus the bound address list can't change.
5640 list_for_each_entry(addr, &bp->address_list, list) {
5641 memcpy(&temp, &addr->a, sizeof(temp));
5642 addrlen = sctp_get_pf_specific(sk->sk_family)
5643 ->addr_to_user(sp, &temp);
5644 if (space_left < addrlen) {
5645 err = -ENOMEM; /*fixme: right error?*/
5648 memcpy(buf, &temp, addrlen);
5650 bytes_copied += addrlen;
5652 space_left -= addrlen;
5656 if (copy_to_user(to, addrs, bytes_copied)) {
5660 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
5664 /* XXX: We should have accounted for sizeof(struct sctp_getaddrs) too,
5665 * but we can't change it anymore.
5667 if (put_user(bytes_copied, optlen))
5674 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
5676 * Requests that the local SCTP stack use the enclosed peer address as
5677 * the association primary. The enclosed address must be one of the
5678 * association peer's addresses.
5680 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
5681 char __user *optval, int __user *optlen)
5683 struct sctp_prim prim;
5684 struct sctp_association *asoc;
5685 struct sctp_sock *sp = sctp_sk(sk);
5687 if (len < sizeof(struct sctp_prim))
5690 len = sizeof(struct sctp_prim);
5692 if (copy_from_user(&prim, optval, len))
5695 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
5699 if (!asoc->peer.primary_path)
5702 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
5703 asoc->peer.primary_path->af_specific->sockaddr_len);
5705 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
5706 (union sctp_addr *)&prim.ssp_addr);
5708 if (put_user(len, optlen))
5710 if (copy_to_user(optval, &prim, len))
5717 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5719 * Requests that the local endpoint set the specified Adaptation Layer
5720 * Indication parameter for all future INIT and INIT-ACK exchanges.
5722 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
5723 char __user *optval, int __user *optlen)
5725 struct sctp_setadaptation adaptation;
5727 if (len < sizeof(struct sctp_setadaptation))
5730 len = sizeof(struct sctp_setadaptation);
5732 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
5734 if (put_user(len, optlen))
5736 if (copy_to_user(optval, &adaptation, len))
5744 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5746 * Applications that wish to use the sendto() system call may wish to
5747 * specify a default set of parameters that would normally be supplied
5748 * through the inclusion of ancillary data. This socket option allows
5749 * such an application to set the default sctp_sndrcvinfo structure.
5752 * The application that wishes to use this socket option simply passes
5753 * in to this call the sctp_sndrcvinfo structure defined in Section
5754 * 5.2.2) The input parameters accepted by this call include
5755 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5756 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5757 * to this call if the caller is using the UDP model.
5759 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5761 static int sctp_getsockopt_default_send_param(struct sock *sk,
5762 int len, char __user *optval,
5765 struct sctp_sock *sp = sctp_sk(sk);
5766 struct sctp_association *asoc;
5767 struct sctp_sndrcvinfo info;
5769 if (len < sizeof(info))
5774 if (copy_from_user(&info, optval, len))
5777 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
5778 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
5781 info.sinfo_stream = asoc->default_stream;
5782 info.sinfo_flags = asoc->default_flags;
5783 info.sinfo_ppid = asoc->default_ppid;
5784 info.sinfo_context = asoc->default_context;
5785 info.sinfo_timetolive = asoc->default_timetolive;
5787 info.sinfo_stream = sp->default_stream;
5788 info.sinfo_flags = sp->default_flags;
5789 info.sinfo_ppid = sp->default_ppid;
5790 info.sinfo_context = sp->default_context;
5791 info.sinfo_timetolive = sp->default_timetolive;
5794 if (put_user(len, optlen))
5796 if (copy_to_user(optval, &info, len))
5802 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5803 * (SCTP_DEFAULT_SNDINFO)
5805 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
5806 char __user *optval,
5809 struct sctp_sock *sp = sctp_sk(sk);
5810 struct sctp_association *asoc;
5811 struct sctp_sndinfo info;
5813 if (len < sizeof(info))
5818 if (copy_from_user(&info, optval, len))
5821 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
5822 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
5825 info.snd_sid = asoc->default_stream;
5826 info.snd_flags = asoc->default_flags;
5827 info.snd_ppid = asoc->default_ppid;
5828 info.snd_context = asoc->default_context;
5830 info.snd_sid = sp->default_stream;
5831 info.snd_flags = sp->default_flags;
5832 info.snd_ppid = sp->default_ppid;
5833 info.snd_context = sp->default_context;
5836 if (put_user(len, optlen))
5838 if (copy_to_user(optval, &info, len))
5846 * 7.1.5 SCTP_NODELAY
5848 * Turn on/off any Nagle-like algorithm. This means that packets are
5849 * generally sent as soon as possible and no unnecessary delays are
5850 * introduced, at the cost of more packets in the network. Expects an
5851 * integer boolean flag.
5854 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5855 char __user *optval, int __user *optlen)
5859 if (len < sizeof(int))
5863 val = (sctp_sk(sk)->nodelay == 1);
5864 if (put_user(len, optlen))
5866 if (copy_to_user(optval, &val, len))
5873 * 7.1.1 SCTP_RTOINFO
5875 * The protocol parameters used to initialize and bound retransmission
5876 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5877 * and modify these parameters.
5878 * All parameters are time values, in milliseconds. A value of 0, when
5879 * modifying the parameters, indicates that the current value should not
5883 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5884 char __user *optval,
5885 int __user *optlen) {
5886 struct sctp_rtoinfo rtoinfo;
5887 struct sctp_association *asoc;
5889 if (len < sizeof (struct sctp_rtoinfo))
5892 len = sizeof(struct sctp_rtoinfo);
5894 if (copy_from_user(&rtoinfo, optval, len))
5897 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5899 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5902 /* Values corresponding to the specific association. */
5904 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5905 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5906 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5908 /* Values corresponding to the endpoint. */
5909 struct sctp_sock *sp = sctp_sk(sk);
5911 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5912 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5913 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5916 if (put_user(len, optlen))
5919 if (copy_to_user(optval, &rtoinfo, len))
5927 * 7.1.2 SCTP_ASSOCINFO
5929 * This option is used to tune the maximum retransmission attempts
5930 * of the association.
5931 * Returns an error if the new association retransmission value is
5932 * greater than the sum of the retransmission value of the peer.
5933 * See [SCTP] for more information.
5936 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5937 char __user *optval,
5941 struct sctp_assocparams assocparams;
5942 struct sctp_association *asoc;
5943 struct list_head *pos;
5946 if (len < sizeof (struct sctp_assocparams))
5949 len = sizeof(struct sctp_assocparams);
5951 if (copy_from_user(&assocparams, optval, len))
5954 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5956 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5959 /* Values correspoinding to the specific association */
5961 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5962 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5963 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5964 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5966 list_for_each(pos, &asoc->peer.transport_addr_list) {
5970 assocparams.sasoc_number_peer_destinations = cnt;
5972 /* Values corresponding to the endpoint */
5973 struct sctp_sock *sp = sctp_sk(sk);
5975 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5976 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5977 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5978 assocparams.sasoc_cookie_life =
5979 sp->assocparams.sasoc_cookie_life;
5980 assocparams.sasoc_number_peer_destinations =
5982 sasoc_number_peer_destinations;
5985 if (put_user(len, optlen))
5988 if (copy_to_user(optval, &assocparams, len))
5995 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5997 * This socket option is a boolean flag which turns on or off mapped V4
5998 * addresses. If this option is turned on and the socket is type
5999 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
6000 * If this option is turned off, then no mapping will be done of V4
6001 * addresses and a user will receive both PF_INET6 and PF_INET type
6002 * addresses on the socket.
6004 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
6005 char __user *optval, int __user *optlen)
6008 struct sctp_sock *sp = sctp_sk(sk);
6010 if (len < sizeof(int))
6015 if (put_user(len, optlen))
6017 if (copy_to_user(optval, &val, len))
6024 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
6025 * (chapter and verse is quoted at sctp_setsockopt_context())
6027 static int sctp_getsockopt_context(struct sock *sk, int len,
6028 char __user *optval, int __user *optlen)
6030 struct sctp_assoc_value params;
6031 struct sctp_sock *sp;
6032 struct sctp_association *asoc;
6034 if (len < sizeof(struct sctp_assoc_value))
6037 len = sizeof(struct sctp_assoc_value);
6039 if (copy_from_user(¶ms, optval, len))
6044 if (params.assoc_id != 0) {
6045 asoc = sctp_id2assoc(sk, params.assoc_id);
6048 params.assoc_value = asoc->default_rcv_context;
6050 params.assoc_value = sp->default_rcv_context;
6053 if (put_user(len, optlen))
6055 if (copy_to_user(optval, ¶ms, len))
6062 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
6063 * This option will get or set the maximum size to put in any outgoing
6064 * SCTP DATA chunk. If a message is larger than this size it will be
6065 * fragmented by SCTP into the specified size. Note that the underlying
6066 * SCTP implementation may fragment into smaller sized chunks when the
6067 * PMTU of the underlying association is smaller than the value set by
6068 * the user. The default value for this option is '0' which indicates
6069 * the user is NOT limiting fragmentation and only the PMTU will effect
6070 * SCTP's choice of DATA chunk size. Note also that values set larger
6071 * than the maximum size of an IP datagram will effectively let SCTP
6072 * control fragmentation (i.e. the same as setting this option to 0).
6074 * The following structure is used to access and modify this parameter:
6076 * struct sctp_assoc_value {
6077 * sctp_assoc_t assoc_id;
6078 * uint32_t assoc_value;
6081 * assoc_id: This parameter is ignored for one-to-one style sockets.
6082 * For one-to-many style sockets this parameter indicates which
6083 * association the user is performing an action upon. Note that if
6084 * this field's value is zero then the endpoints default value is
6085 * changed (effecting future associations only).
6086 * assoc_value: This parameter specifies the maximum size in bytes.
6088 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
6089 char __user *optval, int __user *optlen)
6091 struct sctp_assoc_value params;
6092 struct sctp_association *asoc;
6094 if (len == sizeof(int)) {
6095 pr_warn_ratelimited(DEPRECATED
6097 "Use of int in maxseg socket option.\n"
6098 "Use struct sctp_assoc_value instead\n",
6099 current->comm, task_pid_nr(current));
6100 params.assoc_id = 0;
6101 } else if (len >= sizeof(struct sctp_assoc_value)) {
6102 len = sizeof(struct sctp_assoc_value);
6103 if (copy_from_user(¶ms, optval, len))
6108 asoc = sctp_id2assoc(sk, params.assoc_id);
6109 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
6113 params.assoc_value = asoc->frag_point;
6115 params.assoc_value = sctp_sk(sk)->user_frag;
6117 if (put_user(len, optlen))
6119 if (len == sizeof(int)) {
6120 if (copy_to_user(optval, ¶ms.assoc_value, len))
6123 if (copy_to_user(optval, ¶ms, len))
6131 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
6132 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
6134 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
6135 char __user *optval, int __user *optlen)
6139 if (len < sizeof(int))
6144 val = sctp_sk(sk)->frag_interleave;
6145 if (put_user(len, optlen))
6147 if (copy_to_user(optval, &val, len))
6154 * 7.1.25. Set or Get the sctp partial delivery point
6155 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
6157 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
6158 char __user *optval,
6163 if (len < sizeof(u32))
6168 val = sctp_sk(sk)->pd_point;
6169 if (put_user(len, optlen))
6171 if (copy_to_user(optval, &val, len))
6178 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
6179 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
6181 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
6182 char __user *optval,
6185 struct sctp_assoc_value params;
6186 struct sctp_sock *sp;
6187 struct sctp_association *asoc;
6189 if (len == sizeof(int)) {
6190 pr_warn_ratelimited(DEPRECATED
6192 "Use of int in max_burst socket option.\n"
6193 "Use struct sctp_assoc_value instead\n",
6194 current->comm, task_pid_nr(current));
6195 params.assoc_id = 0;
6196 } else if (len >= sizeof(struct sctp_assoc_value)) {
6197 len = sizeof(struct sctp_assoc_value);
6198 if (copy_from_user(¶ms, optval, len))
6205 if (params.assoc_id != 0) {
6206 asoc = sctp_id2assoc(sk, params.assoc_id);
6209 params.assoc_value = asoc->max_burst;
6211 params.assoc_value = sp->max_burst;
6213 if (len == sizeof(int)) {
6214 if (copy_to_user(optval, ¶ms.assoc_value, len))
6217 if (copy_to_user(optval, ¶ms, len))
6225 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
6226 char __user *optval, int __user *optlen)
6228 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6229 struct sctp_hmacalgo __user *p = (void __user *)optval;
6230 struct sctp_hmac_algo_param *hmacs;
6235 if (!ep->auth_enable)
6238 hmacs = ep->auth_hmacs_list;
6239 data_len = ntohs(hmacs->param_hdr.length) -
6240 sizeof(struct sctp_paramhdr);
6242 if (len < sizeof(struct sctp_hmacalgo) + data_len)
6245 len = sizeof(struct sctp_hmacalgo) + data_len;
6246 num_idents = data_len / sizeof(u16);
6248 if (put_user(len, optlen))
6250 if (put_user(num_idents, &p->shmac_num_idents))
6252 for (i = 0; i < num_idents; i++) {
6253 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
6255 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
6261 static int sctp_getsockopt_active_key(struct sock *sk, int len,
6262 char __user *optval, int __user *optlen)
6264 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6265 struct sctp_authkeyid val;
6266 struct sctp_association *asoc;
6268 if (!ep->auth_enable)
6271 if (len < sizeof(struct sctp_authkeyid))
6274 len = sizeof(struct sctp_authkeyid);
6275 if (copy_from_user(&val, optval, len))
6278 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
6279 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
6283 val.scact_keynumber = asoc->active_key_id;
6285 val.scact_keynumber = ep->active_key_id;
6287 if (put_user(len, optlen))
6289 if (copy_to_user(optval, &val, len))
6295 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
6296 char __user *optval, int __user *optlen)
6298 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6299 struct sctp_authchunks __user *p = (void __user *)optval;
6300 struct sctp_authchunks val;
6301 struct sctp_association *asoc;
6302 struct sctp_chunks_param *ch;
6306 if (!ep->auth_enable)
6309 if (len < sizeof(struct sctp_authchunks))
6312 if (copy_from_user(&val, optval, sizeof(val)))
6315 to = p->gauth_chunks;
6316 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6320 ch = asoc->peer.peer_chunks;
6324 /* See if the user provided enough room for all the data */
6325 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6326 if (len < num_chunks)
6329 if (copy_to_user(to, ch->chunks, num_chunks))
6332 len = sizeof(struct sctp_authchunks) + num_chunks;
6333 if (put_user(len, optlen))
6335 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6340 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
6341 char __user *optval, int __user *optlen)
6343 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6344 struct sctp_authchunks __user *p = (void __user *)optval;
6345 struct sctp_authchunks val;
6346 struct sctp_association *asoc;
6347 struct sctp_chunks_param *ch;
6351 if (!ep->auth_enable)
6354 if (len < sizeof(struct sctp_authchunks))
6357 if (copy_from_user(&val, optval, sizeof(val)))
6360 to = p->gauth_chunks;
6361 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6362 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
6366 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
6368 ch = ep->auth_chunk_list;
6373 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6374 if (len < sizeof(struct sctp_authchunks) + num_chunks)
6377 if (copy_to_user(to, ch->chunks, num_chunks))
6380 len = sizeof(struct sctp_authchunks) + num_chunks;
6381 if (put_user(len, optlen))
6383 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6390 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
6391 * This option gets the current number of associations that are attached
6392 * to a one-to-many style socket. The option value is an uint32_t.
6394 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
6395 char __user *optval, int __user *optlen)
6397 struct sctp_sock *sp = sctp_sk(sk);
6398 struct sctp_association *asoc;
6401 if (sctp_style(sk, TCP))
6404 if (len < sizeof(u32))
6409 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6413 if (put_user(len, optlen))
6415 if (copy_to_user(optval, &val, len))
6422 * 8.1.23 SCTP_AUTO_ASCONF
6423 * See the corresponding setsockopt entry as description
6425 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
6426 char __user *optval, int __user *optlen)
6430 if (len < sizeof(int))
6434 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
6436 if (put_user(len, optlen))
6438 if (copy_to_user(optval, &val, len))
6444 * 8.2.6. Get the Current Identifiers of Associations
6445 * (SCTP_GET_ASSOC_ID_LIST)
6447 * This option gets the current list of SCTP association identifiers of
6448 * the SCTP associations handled by a one-to-many style socket.
6450 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
6451 char __user *optval, int __user *optlen)
6453 struct sctp_sock *sp = sctp_sk(sk);
6454 struct sctp_association *asoc;
6455 struct sctp_assoc_ids *ids;
6458 if (sctp_style(sk, TCP))
6461 if (len < sizeof(struct sctp_assoc_ids))
6464 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6468 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
6471 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
6473 ids = kmalloc(len, GFP_USER | __GFP_NOWARN);
6477 ids->gaids_number_of_ids = num;
6479 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6480 ids->gaids_assoc_id[num++] = asoc->assoc_id;
6483 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
6493 * SCTP_PEER_ADDR_THLDS
6495 * This option allows us to fetch the partially failed threshold for one or all
6496 * transports in an association. See Section 6.1 of:
6497 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6499 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
6500 char __user *optval,
6504 struct sctp_paddrthlds val;
6505 struct sctp_transport *trans;
6506 struct sctp_association *asoc;
6508 if (len < sizeof(struct sctp_paddrthlds))
6510 len = sizeof(struct sctp_paddrthlds);
6511 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
6514 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
6515 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
6519 val.spt_pathpfthld = asoc->pf_retrans;
6520 val.spt_pathmaxrxt = asoc->pathmaxrxt;
6522 trans = sctp_addr_id2transport(sk, &val.spt_address,
6527 val.spt_pathmaxrxt = trans->pathmaxrxt;
6528 val.spt_pathpfthld = trans->pf_retrans;
6531 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
6538 * SCTP_GET_ASSOC_STATS
6540 * This option retrieves local per endpoint statistics. It is modeled
6541 * after OpenSolaris' implementation
6543 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
6544 char __user *optval,
6547 struct sctp_assoc_stats sas;
6548 struct sctp_association *asoc = NULL;
6550 /* User must provide at least the assoc id */
6551 if (len < sizeof(sctp_assoc_t))
6554 /* Allow the struct to grow and fill in as much as possible */
6555 len = min_t(size_t, len, sizeof(sas));
6557 if (copy_from_user(&sas, optval, len))
6560 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
6564 sas.sas_rtxchunks = asoc->stats.rtxchunks;
6565 sas.sas_gapcnt = asoc->stats.gapcnt;
6566 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
6567 sas.sas_osacks = asoc->stats.osacks;
6568 sas.sas_isacks = asoc->stats.isacks;
6569 sas.sas_octrlchunks = asoc->stats.octrlchunks;
6570 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
6571 sas.sas_oodchunks = asoc->stats.oodchunks;
6572 sas.sas_iodchunks = asoc->stats.iodchunks;
6573 sas.sas_ouodchunks = asoc->stats.ouodchunks;
6574 sas.sas_iuodchunks = asoc->stats.iuodchunks;
6575 sas.sas_idupchunks = asoc->stats.idupchunks;
6576 sas.sas_opackets = asoc->stats.opackets;
6577 sas.sas_ipackets = asoc->stats.ipackets;
6579 /* New high max rto observed, will return 0 if not a single
6580 * RTO update took place. obs_rto_ipaddr will be bogus
6583 sas.sas_maxrto = asoc->stats.max_obs_rto;
6584 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
6585 sizeof(struct sockaddr_storage));
6587 /* Mark beginning of a new observation period */
6588 asoc->stats.max_obs_rto = asoc->rto_min;
6590 if (put_user(len, optlen))
6593 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
6595 if (copy_to_user(optval, &sas, len))
6601 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
6602 char __user *optval,
6607 if (len < sizeof(int))
6611 if (sctp_sk(sk)->recvrcvinfo)
6613 if (put_user(len, optlen))
6615 if (copy_to_user(optval, &val, len))
6621 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
6622 char __user *optval,
6627 if (len < sizeof(int))
6631 if (sctp_sk(sk)->recvnxtinfo)
6633 if (put_user(len, optlen))
6635 if (copy_to_user(optval, &val, len))
6641 static int sctp_getsockopt_pr_supported(struct sock *sk, int len,
6642 char __user *optval,
6645 struct sctp_assoc_value params;
6646 struct sctp_association *asoc;
6647 int retval = -EFAULT;
6649 if (len < sizeof(params)) {
6654 len = sizeof(params);
6655 if (copy_from_user(¶ms, optval, len))
6658 asoc = sctp_id2assoc(sk, params.assoc_id);
6660 params.assoc_value = asoc->prsctp_enable;
6661 } else if (!params.assoc_id) {
6662 struct sctp_sock *sp = sctp_sk(sk);
6664 params.assoc_value = sp->ep->prsctp_enable;
6670 if (put_user(len, optlen))
6673 if (copy_to_user(optval, ¶ms, len))
6682 static int sctp_getsockopt_default_prinfo(struct sock *sk, int len,
6683 char __user *optval,
6686 struct sctp_default_prinfo info;
6687 struct sctp_association *asoc;
6688 int retval = -EFAULT;
6690 if (len < sizeof(info)) {
6696 if (copy_from_user(&info, optval, len))
6699 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
6701 info.pr_policy = SCTP_PR_POLICY(asoc->default_flags);
6702 info.pr_value = asoc->default_timetolive;
6703 } else if (!info.pr_assoc_id) {
6704 struct sctp_sock *sp = sctp_sk(sk);
6706 info.pr_policy = SCTP_PR_POLICY(sp->default_flags);
6707 info.pr_value = sp->default_timetolive;
6713 if (put_user(len, optlen))
6716 if (copy_to_user(optval, &info, len))
6725 static int sctp_getsockopt_pr_assocstatus(struct sock *sk, int len,
6726 char __user *optval,
6729 struct sctp_prstatus params;
6730 struct sctp_association *asoc;
6732 int retval = -EINVAL;
6734 if (len < sizeof(params))
6737 len = sizeof(params);
6738 if (copy_from_user(¶ms, optval, len)) {
6743 policy = params.sprstat_policy;
6744 if (policy & ~SCTP_PR_SCTP_MASK)
6747 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
6751 if (policy == SCTP_PR_SCTP_NONE) {
6752 params.sprstat_abandoned_unsent = 0;
6753 params.sprstat_abandoned_sent = 0;
6754 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
6755 params.sprstat_abandoned_unsent +=
6756 asoc->abandoned_unsent[policy];
6757 params.sprstat_abandoned_sent +=
6758 asoc->abandoned_sent[policy];
6761 params.sprstat_abandoned_unsent =
6762 asoc->abandoned_unsent[__SCTP_PR_INDEX(policy)];
6763 params.sprstat_abandoned_sent =
6764 asoc->abandoned_sent[__SCTP_PR_INDEX(policy)];
6767 if (put_user(len, optlen)) {
6772 if (copy_to_user(optval, ¶ms, len)) {
6783 static int sctp_getsockopt_pr_streamstatus(struct sock *sk, int len,
6784 char __user *optval,
6787 struct sctp_stream_out_ext *streamoute;
6788 struct sctp_association *asoc;
6789 struct sctp_prstatus params;
6790 int retval = -EINVAL;
6793 if (len < sizeof(params))
6796 len = sizeof(params);
6797 if (copy_from_user(¶ms, optval, len)) {
6802 policy = params.sprstat_policy;
6803 if (policy & ~SCTP_PR_SCTP_MASK)
6806 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
6807 if (!asoc || params.sprstat_sid >= asoc->stream.outcnt)
6810 streamoute = asoc->stream.out[params.sprstat_sid].ext;
6812 /* Not allocated yet, means all stats are 0 */
6813 params.sprstat_abandoned_unsent = 0;
6814 params.sprstat_abandoned_sent = 0;
6819 if (policy == SCTP_PR_SCTP_NONE) {
6820 params.sprstat_abandoned_unsent = 0;
6821 params.sprstat_abandoned_sent = 0;
6822 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
6823 params.sprstat_abandoned_unsent +=
6824 streamoute->abandoned_unsent[policy];
6825 params.sprstat_abandoned_sent +=
6826 streamoute->abandoned_sent[policy];
6829 params.sprstat_abandoned_unsent =
6830 streamoute->abandoned_unsent[__SCTP_PR_INDEX(policy)];
6831 params.sprstat_abandoned_sent =
6832 streamoute->abandoned_sent[__SCTP_PR_INDEX(policy)];
6835 if (put_user(len, optlen) || copy_to_user(optval, ¶ms, len)) {
6846 static int sctp_getsockopt_reconfig_supported(struct sock *sk, int len,
6847 char __user *optval,
6850 struct sctp_assoc_value params;
6851 struct sctp_association *asoc;
6852 int retval = -EFAULT;
6854 if (len < sizeof(params)) {
6859 len = sizeof(params);
6860 if (copy_from_user(¶ms, optval, len))
6863 asoc = sctp_id2assoc(sk, params.assoc_id);
6865 params.assoc_value = asoc->reconf_enable;
6866 } else if (!params.assoc_id) {
6867 struct sctp_sock *sp = sctp_sk(sk);
6869 params.assoc_value = sp->ep->reconf_enable;
6875 if (put_user(len, optlen))
6878 if (copy_to_user(optval, ¶ms, len))
6887 static int sctp_getsockopt_enable_strreset(struct sock *sk, int len,
6888 char __user *optval,
6891 struct sctp_assoc_value params;
6892 struct sctp_association *asoc;
6893 int retval = -EFAULT;
6895 if (len < sizeof(params)) {
6900 len = sizeof(params);
6901 if (copy_from_user(¶ms, optval, len))
6904 asoc = sctp_id2assoc(sk, params.assoc_id);
6906 params.assoc_value = asoc->strreset_enable;
6907 } else if (!params.assoc_id) {
6908 struct sctp_sock *sp = sctp_sk(sk);
6910 params.assoc_value = sp->ep->strreset_enable;
6916 if (put_user(len, optlen))
6919 if (copy_to_user(optval, ¶ms, len))
6928 static int sctp_getsockopt_scheduler(struct sock *sk, int len,
6929 char __user *optval,
6932 struct sctp_assoc_value params;
6933 struct sctp_association *asoc;
6934 int retval = -EFAULT;
6936 if (len < sizeof(params)) {
6941 len = sizeof(params);
6942 if (copy_from_user(¶ms, optval, len))
6945 asoc = sctp_id2assoc(sk, params.assoc_id);
6951 params.assoc_value = sctp_sched_get_sched(asoc);
6953 if (put_user(len, optlen))
6956 if (copy_to_user(optval, ¶ms, len))
6965 static int sctp_getsockopt_scheduler_value(struct sock *sk, int len,
6966 char __user *optval,
6969 struct sctp_stream_value params;
6970 struct sctp_association *asoc;
6971 int retval = -EFAULT;
6973 if (len < sizeof(params)) {
6978 len = sizeof(params);
6979 if (copy_from_user(¶ms, optval, len))
6982 asoc = sctp_id2assoc(sk, params.assoc_id);
6988 retval = sctp_sched_get_value(asoc, params.stream_id,
6989 ¶ms.stream_value);
6993 if (put_user(len, optlen)) {
6998 if (copy_to_user(optval, ¶ms, len)) {
7007 static int sctp_getsockopt(struct sock *sk, int level, int optname,
7008 char __user *optval, int __user *optlen)
7013 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
7015 /* I can hardly begin to describe how wrong this is. This is
7016 * so broken as to be worse than useless. The API draft
7017 * REALLY is NOT helpful here... I am not convinced that the
7018 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
7019 * are at all well-founded.
7021 if (level != SOL_SCTP) {
7022 struct sctp_af *af = sctp_sk(sk)->pf->af;
7024 retval = af->getsockopt(sk, level, optname, optval, optlen);
7028 if (get_user(len, optlen))
7038 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
7040 case SCTP_DISABLE_FRAGMENTS:
7041 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
7045 retval = sctp_getsockopt_events(sk, len, optval, optlen);
7047 case SCTP_AUTOCLOSE:
7048 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
7050 case SCTP_SOCKOPT_PEELOFF:
7051 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
7053 case SCTP_SOCKOPT_PEELOFF_FLAGS:
7054 retval = sctp_getsockopt_peeloff_flags(sk, len, optval, optlen);
7056 case SCTP_PEER_ADDR_PARAMS:
7057 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
7060 case SCTP_DELAYED_SACK:
7061 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
7065 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
7067 case SCTP_GET_PEER_ADDRS:
7068 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
7071 case SCTP_GET_LOCAL_ADDRS:
7072 retval = sctp_getsockopt_local_addrs(sk, len, optval,
7075 case SCTP_SOCKOPT_CONNECTX3:
7076 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
7078 case SCTP_DEFAULT_SEND_PARAM:
7079 retval = sctp_getsockopt_default_send_param(sk, len,
7082 case SCTP_DEFAULT_SNDINFO:
7083 retval = sctp_getsockopt_default_sndinfo(sk, len,
7086 case SCTP_PRIMARY_ADDR:
7087 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
7090 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
7093 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
7095 case SCTP_ASSOCINFO:
7096 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
7098 case SCTP_I_WANT_MAPPED_V4_ADDR:
7099 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
7102 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
7104 case SCTP_GET_PEER_ADDR_INFO:
7105 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
7108 case SCTP_ADAPTATION_LAYER:
7109 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
7113 retval = sctp_getsockopt_context(sk, len, optval, optlen);
7115 case SCTP_FRAGMENT_INTERLEAVE:
7116 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
7119 case SCTP_PARTIAL_DELIVERY_POINT:
7120 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
7123 case SCTP_MAX_BURST:
7124 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
7127 case SCTP_AUTH_CHUNK:
7128 case SCTP_AUTH_DELETE_KEY:
7129 retval = -EOPNOTSUPP;
7131 case SCTP_HMAC_IDENT:
7132 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
7134 case SCTP_AUTH_ACTIVE_KEY:
7135 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
7137 case SCTP_PEER_AUTH_CHUNKS:
7138 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
7141 case SCTP_LOCAL_AUTH_CHUNKS:
7142 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
7145 case SCTP_GET_ASSOC_NUMBER:
7146 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
7148 case SCTP_GET_ASSOC_ID_LIST:
7149 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
7151 case SCTP_AUTO_ASCONF:
7152 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
7154 case SCTP_PEER_ADDR_THLDS:
7155 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
7157 case SCTP_GET_ASSOC_STATS:
7158 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
7160 case SCTP_RECVRCVINFO:
7161 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
7163 case SCTP_RECVNXTINFO:
7164 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
7166 case SCTP_PR_SUPPORTED:
7167 retval = sctp_getsockopt_pr_supported(sk, len, optval, optlen);
7169 case SCTP_DEFAULT_PRINFO:
7170 retval = sctp_getsockopt_default_prinfo(sk, len, optval,
7173 case SCTP_PR_ASSOC_STATUS:
7174 retval = sctp_getsockopt_pr_assocstatus(sk, len, optval,
7177 case SCTP_PR_STREAM_STATUS:
7178 retval = sctp_getsockopt_pr_streamstatus(sk, len, optval,
7181 case SCTP_RECONFIG_SUPPORTED:
7182 retval = sctp_getsockopt_reconfig_supported(sk, len, optval,
7185 case SCTP_ENABLE_STREAM_RESET:
7186 retval = sctp_getsockopt_enable_strreset(sk, len, optval,
7189 case SCTP_STREAM_SCHEDULER:
7190 retval = sctp_getsockopt_scheduler(sk, len, optval,
7193 case SCTP_STREAM_SCHEDULER_VALUE:
7194 retval = sctp_getsockopt_scheduler_value(sk, len, optval,
7198 retval = -ENOPROTOOPT;
7206 static int sctp_hash(struct sock *sk)
7212 static void sctp_unhash(struct sock *sk)
7217 /* Check if port is acceptable. Possibly find first available port.
7219 * The port hash table (contained in the 'global' SCTP protocol storage
7220 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
7221 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
7222 * list (the list number is the port number hashed out, so as you
7223 * would expect from a hash function, all the ports in a given list have
7224 * such a number that hashes out to the same list number; you were
7225 * expecting that, right?); so each list has a set of ports, with a
7226 * link to the socket (struct sock) that uses it, the port number and
7227 * a fastreuse flag (FIXME: NPI ipg).
7229 static struct sctp_bind_bucket *sctp_bucket_create(
7230 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
7232 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
7234 struct sctp_bind_hashbucket *head; /* hash list */
7235 struct sctp_bind_bucket *pp;
7236 unsigned short snum;
7239 snum = ntohs(addr->v4.sin_port);
7241 pr_debug("%s: begins, snum:%d\n", __func__, snum);
7246 /* Search for an available port. */
7247 int low, high, remaining, index;
7249 struct net *net = sock_net(sk);
7251 inet_get_local_port_range(net, &low, &high);
7252 remaining = (high - low) + 1;
7253 rover = prandom_u32() % remaining + low;
7257 if ((rover < low) || (rover > high))
7259 if (inet_is_local_reserved_port(net, rover))
7261 index = sctp_phashfn(sock_net(sk), rover);
7262 head = &sctp_port_hashtable[index];
7263 spin_lock(&head->lock);
7264 sctp_for_each_hentry(pp, &head->chain)
7265 if ((pp->port == rover) &&
7266 net_eq(sock_net(sk), pp->net))
7270 spin_unlock(&head->lock);
7271 } while (--remaining > 0);
7273 /* Exhausted local port range during search? */
7278 /* OK, here is the one we will use. HEAD (the port
7279 * hash table list entry) is non-NULL and we hold it's
7284 /* We are given an specific port number; we verify
7285 * that it is not being used. If it is used, we will
7286 * exahust the search in the hash list corresponding
7287 * to the port number (snum) - we detect that with the
7288 * port iterator, pp being NULL.
7290 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
7291 spin_lock(&head->lock);
7292 sctp_for_each_hentry(pp, &head->chain) {
7293 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
7300 if (!hlist_empty(&pp->owner)) {
7301 /* We had a port hash table hit - there is an
7302 * available port (pp != NULL) and it is being
7303 * used by other socket (pp->owner not empty); that other
7304 * socket is going to be sk2.
7306 int reuse = sk->sk_reuse;
7309 pr_debug("%s: found a possible match\n", __func__);
7311 if (pp->fastreuse && sk->sk_reuse &&
7312 sk->sk_state != SCTP_SS_LISTENING)
7315 /* Run through the list of sockets bound to the port
7316 * (pp->port) [via the pointers bind_next and
7317 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
7318 * we get the endpoint they describe and run through
7319 * the endpoint's list of IP (v4 or v6) addresses,
7320 * comparing each of the addresses with the address of
7321 * the socket sk. If we find a match, then that means
7322 * that this port/socket (sk) combination are already
7325 sk_for_each_bound(sk2, &pp->owner) {
7326 struct sctp_endpoint *ep2;
7327 ep2 = sctp_sk(sk2)->ep;
7330 (reuse && sk2->sk_reuse &&
7331 sk2->sk_state != SCTP_SS_LISTENING))
7334 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
7335 sctp_sk(sk2), sctp_sk(sk))) {
7341 pr_debug("%s: found a match\n", __func__);
7344 /* If there was a hash table miss, create a new port. */
7346 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
7349 /* In either case (hit or miss), make sure fastreuse is 1 only
7350 * if sk->sk_reuse is too (that is, if the caller requested
7351 * SO_REUSEADDR on this socket -sk-).
7353 if (hlist_empty(&pp->owner)) {
7354 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
7358 } else if (pp->fastreuse &&
7359 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
7362 /* We are set, so fill up all the data in the hash table
7363 * entry, tie the socket list information with the rest of the
7364 * sockets FIXME: Blurry, NPI (ipg).
7367 if (!sctp_sk(sk)->bind_hash) {
7368 inet_sk(sk)->inet_num = snum;
7369 sk_add_bind_node(sk, &pp->owner);
7370 sctp_sk(sk)->bind_hash = pp;
7375 spin_unlock(&head->lock);
7382 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
7383 * port is requested.
7385 static int sctp_get_port(struct sock *sk, unsigned short snum)
7387 union sctp_addr addr;
7388 struct sctp_af *af = sctp_sk(sk)->pf->af;
7390 /* Set up a dummy address struct from the sk. */
7391 af->from_sk(&addr, sk);
7392 addr.v4.sin_port = htons(snum);
7394 /* Note: sk->sk_num gets filled in if ephemeral port request. */
7395 return !!sctp_get_port_local(sk, &addr);
7399 * Move a socket to LISTENING state.
7401 static int sctp_listen_start(struct sock *sk, int backlog)
7403 struct sctp_sock *sp = sctp_sk(sk);
7404 struct sctp_endpoint *ep = sp->ep;
7405 struct crypto_shash *tfm = NULL;
7408 /* Allocate HMAC for generating cookie. */
7409 if (!sp->hmac && sp->sctp_hmac_alg) {
7410 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
7411 tfm = crypto_alloc_shash(alg, 0, 0);
7413 net_info_ratelimited("failed to load transform for %s: %ld\n",
7414 sp->sctp_hmac_alg, PTR_ERR(tfm));
7417 sctp_sk(sk)->hmac = tfm;
7421 * If a bind() or sctp_bindx() is not called prior to a listen()
7422 * call that allows new associations to be accepted, the system
7423 * picks an ephemeral port and will choose an address set equivalent
7424 * to binding with a wildcard address.
7426 * This is not currently spelled out in the SCTP sockets
7427 * extensions draft, but follows the practice as seen in TCP
7431 sk->sk_state = SCTP_SS_LISTENING;
7432 if (!ep->base.bind_addr.port) {
7433 if (sctp_autobind(sk))
7436 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
7437 sk->sk_state = SCTP_SS_CLOSED;
7442 sk->sk_max_ack_backlog = backlog;
7443 sctp_hash_endpoint(ep);
7448 * 4.1.3 / 5.1.3 listen()
7450 * By default, new associations are not accepted for UDP style sockets.
7451 * An application uses listen() to mark a socket as being able to
7452 * accept new associations.
7454 * On TCP style sockets, applications use listen() to ready the SCTP
7455 * endpoint for accepting inbound associations.
7457 * On both types of endpoints a backlog of '0' disables listening.
7459 * Move a socket to LISTENING state.
7461 int sctp_inet_listen(struct socket *sock, int backlog)
7463 struct sock *sk = sock->sk;
7464 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
7467 if (unlikely(backlog < 0))
7472 /* Peeled-off sockets are not allowed to listen(). */
7473 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
7476 if (sock->state != SS_UNCONNECTED)
7479 if (!sctp_sstate(sk, LISTENING) && !sctp_sstate(sk, CLOSED))
7482 /* If backlog is zero, disable listening. */
7484 if (sctp_sstate(sk, CLOSED))
7488 sctp_unhash_endpoint(ep);
7489 sk->sk_state = SCTP_SS_CLOSED;
7491 sctp_sk(sk)->bind_hash->fastreuse = 1;
7495 /* If we are already listening, just update the backlog */
7496 if (sctp_sstate(sk, LISTENING))
7497 sk->sk_max_ack_backlog = backlog;
7499 err = sctp_listen_start(sk, backlog);
7511 * This function is done by modeling the current datagram_poll() and the
7512 * tcp_poll(). Note that, based on these implementations, we don't
7513 * lock the socket in this function, even though it seems that,
7514 * ideally, locking or some other mechanisms can be used to ensure
7515 * the integrity of the counters (sndbuf and wmem_alloc) used
7516 * in this place. We assume that we don't need locks either until proven
7519 * Another thing to note is that we include the Async I/O support
7520 * here, again, by modeling the current TCP/UDP code. We don't have
7521 * a good way to test with it yet.
7523 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
7525 struct sock *sk = sock->sk;
7526 struct sctp_sock *sp = sctp_sk(sk);
7529 poll_wait(file, sk_sleep(sk), wait);
7531 sock_rps_record_flow(sk);
7533 /* A TCP-style listening socket becomes readable when the accept queue
7536 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
7537 return (!list_empty(&sp->ep->asocs)) ?
7538 (POLLIN | POLLRDNORM) : 0;
7542 /* Is there any exceptional events? */
7543 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
7545 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
7546 if (sk->sk_shutdown & RCV_SHUTDOWN)
7547 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
7548 if (sk->sk_shutdown == SHUTDOWN_MASK)
7551 /* Is it readable? Reconsider this code with TCP-style support. */
7552 if (!skb_queue_empty(&sk->sk_receive_queue))
7553 mask |= POLLIN | POLLRDNORM;
7555 /* The association is either gone or not ready. */
7556 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
7559 /* Is it writable? */
7560 if (sctp_writeable(sk)) {
7561 mask |= POLLOUT | POLLWRNORM;
7563 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
7565 * Since the socket is not locked, the buffer
7566 * might be made available after the writeable check and
7567 * before the bit is set. This could cause a lost I/O
7568 * signal. tcp_poll() has a race breaker for this race
7569 * condition. Based on their implementation, we put
7570 * in the following code to cover it as well.
7572 if (sctp_writeable(sk))
7573 mask |= POLLOUT | POLLWRNORM;
7578 /********************************************************************
7579 * 2nd Level Abstractions
7580 ********************************************************************/
7582 static struct sctp_bind_bucket *sctp_bucket_create(
7583 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
7585 struct sctp_bind_bucket *pp;
7587 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
7589 SCTP_DBG_OBJCNT_INC(bind_bucket);
7592 INIT_HLIST_HEAD(&pp->owner);
7594 hlist_add_head(&pp->node, &head->chain);
7599 /* Caller must hold hashbucket lock for this tb with local BH disabled */
7600 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
7602 if (pp && hlist_empty(&pp->owner)) {
7603 __hlist_del(&pp->node);
7604 kmem_cache_free(sctp_bucket_cachep, pp);
7605 SCTP_DBG_OBJCNT_DEC(bind_bucket);
7609 /* Release this socket's reference to a local port. */
7610 static inline void __sctp_put_port(struct sock *sk)
7612 struct sctp_bind_hashbucket *head =
7613 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
7614 inet_sk(sk)->inet_num)];
7615 struct sctp_bind_bucket *pp;
7617 spin_lock(&head->lock);
7618 pp = sctp_sk(sk)->bind_hash;
7619 __sk_del_bind_node(sk);
7620 sctp_sk(sk)->bind_hash = NULL;
7621 inet_sk(sk)->inet_num = 0;
7622 sctp_bucket_destroy(pp);
7623 spin_unlock(&head->lock);
7626 void sctp_put_port(struct sock *sk)
7629 __sctp_put_port(sk);
7634 * The system picks an ephemeral port and choose an address set equivalent
7635 * to binding with a wildcard address.
7636 * One of those addresses will be the primary address for the association.
7637 * This automatically enables the multihoming capability of SCTP.
7639 static int sctp_autobind(struct sock *sk)
7641 union sctp_addr autoaddr;
7645 /* Initialize a local sockaddr structure to INADDR_ANY. */
7646 af = sctp_sk(sk)->pf->af;
7648 port = htons(inet_sk(sk)->inet_num);
7649 af->inaddr_any(&autoaddr, port);
7651 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
7654 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
7657 * 4.2 The cmsghdr Structure *
7659 * When ancillary data is sent or received, any number of ancillary data
7660 * objects can be specified by the msg_control and msg_controllen members of
7661 * the msghdr structure, because each object is preceded by
7662 * a cmsghdr structure defining the object's length (the cmsg_len member).
7663 * Historically Berkeley-derived implementations have passed only one object
7664 * at a time, but this API allows multiple objects to be
7665 * passed in a single call to sendmsg() or recvmsg(). The following example
7666 * shows two ancillary data objects in a control buffer.
7668 * |<--------------------------- msg_controllen -------------------------->|
7671 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
7673 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
7676 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
7678 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
7681 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7682 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
7684 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
7686 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7693 static int sctp_msghdr_parse(const struct msghdr *msg, struct sctp_cmsgs *cmsgs)
7695 struct msghdr *my_msg = (struct msghdr *)msg;
7696 struct cmsghdr *cmsg;
7698 for_each_cmsghdr(cmsg, my_msg) {
7699 if (!CMSG_OK(my_msg, cmsg))
7702 /* Should we parse this header or ignore? */
7703 if (cmsg->cmsg_level != IPPROTO_SCTP)
7706 /* Strictly check lengths following example in SCM code. */
7707 switch (cmsg->cmsg_type) {
7709 /* SCTP Socket API Extension
7710 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
7712 * This cmsghdr structure provides information for
7713 * initializing new SCTP associations with sendmsg().
7714 * The SCTP_INITMSG socket option uses this same data
7715 * structure. This structure is not used for
7718 * cmsg_level cmsg_type cmsg_data[]
7719 * ------------ ------------ ----------------------
7720 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
7722 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
7725 cmsgs->init = CMSG_DATA(cmsg);
7729 /* SCTP Socket API Extension
7730 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
7732 * This cmsghdr structure specifies SCTP options for
7733 * sendmsg() and describes SCTP header information
7734 * about a received message through recvmsg().
7736 * cmsg_level cmsg_type cmsg_data[]
7737 * ------------ ------------ ----------------------
7738 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
7740 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
7743 cmsgs->srinfo = CMSG_DATA(cmsg);
7745 if (cmsgs->srinfo->sinfo_flags &
7746 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7747 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7748 SCTP_ABORT | SCTP_EOF))
7753 /* SCTP Socket API Extension
7754 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
7756 * This cmsghdr structure specifies SCTP options for
7757 * sendmsg(). This structure and SCTP_RCVINFO replaces
7758 * SCTP_SNDRCV which has been deprecated.
7760 * cmsg_level cmsg_type cmsg_data[]
7761 * ------------ ------------ ---------------------
7762 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
7764 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
7767 cmsgs->sinfo = CMSG_DATA(cmsg);
7769 if (cmsgs->sinfo->snd_flags &
7770 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7771 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7772 SCTP_ABORT | SCTP_EOF))
7784 * Wait for a packet..
7785 * Note: This function is the same function as in core/datagram.c
7786 * with a few modifications to make lksctp work.
7788 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
7793 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7795 /* Socket errors? */
7796 error = sock_error(sk);
7800 if (!skb_queue_empty(&sk->sk_receive_queue))
7803 /* Socket shut down? */
7804 if (sk->sk_shutdown & RCV_SHUTDOWN)
7807 /* Sequenced packets can come disconnected. If so we report the
7812 /* Is there a good reason to think that we may receive some data? */
7813 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
7816 /* Handle signals. */
7817 if (signal_pending(current))
7820 /* Let another process have a go. Since we are going to sleep
7821 * anyway. Note: This may cause odd behaviors if the message
7822 * does not fit in the user's buffer, but this seems to be the
7823 * only way to honor MSG_DONTWAIT realistically.
7826 *timeo_p = schedule_timeout(*timeo_p);
7830 finish_wait(sk_sleep(sk), &wait);
7834 error = sock_intr_errno(*timeo_p);
7837 finish_wait(sk_sleep(sk), &wait);
7842 /* Receive a datagram.
7843 * Note: This is pretty much the same routine as in core/datagram.c
7844 * with a few changes to make lksctp work.
7846 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
7847 int noblock, int *err)
7850 struct sk_buff *skb;
7853 timeo = sock_rcvtimeo(sk, noblock);
7855 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
7856 MAX_SCHEDULE_TIMEOUT);
7859 /* Again only user level code calls this function,
7860 * so nothing interrupt level
7861 * will suddenly eat the receive_queue.
7863 * Look at current nfs client by the way...
7864 * However, this function was correct in any case. 8)
7866 if (flags & MSG_PEEK) {
7867 skb = skb_peek(&sk->sk_receive_queue);
7869 refcount_inc(&skb->users);
7871 skb = __skb_dequeue(&sk->sk_receive_queue);
7877 /* Caller is allowed not to check sk->sk_err before calling. */
7878 error = sock_error(sk);
7882 if (sk->sk_shutdown & RCV_SHUTDOWN)
7885 if (sk_can_busy_loop(sk)) {
7886 sk_busy_loop(sk, noblock);
7888 if (!skb_queue_empty(&sk->sk_receive_queue))
7892 /* User doesn't want to wait. */
7896 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
7905 /* If sndbuf has changed, wake up per association sndbuf waiters. */
7906 static void __sctp_write_space(struct sctp_association *asoc)
7908 struct sock *sk = asoc->base.sk;
7910 if (sctp_wspace(asoc) <= 0)
7913 if (waitqueue_active(&asoc->wait))
7914 wake_up_interruptible(&asoc->wait);
7916 if (sctp_writeable(sk)) {
7917 struct socket_wq *wq;
7920 wq = rcu_dereference(sk->sk_wq);
7922 if (waitqueue_active(&wq->wait))
7923 wake_up_interruptible(&wq->wait);
7925 /* Note that we try to include the Async I/O support
7926 * here by modeling from the current TCP/UDP code.
7927 * We have not tested with it yet.
7929 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
7930 sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT);
7936 static void sctp_wake_up_waiters(struct sock *sk,
7937 struct sctp_association *asoc)
7939 struct sctp_association *tmp = asoc;
7941 /* We do accounting for the sndbuf space per association,
7942 * so we only need to wake our own association.
7944 if (asoc->ep->sndbuf_policy)
7945 return __sctp_write_space(asoc);
7947 /* If association goes down and is just flushing its
7948 * outq, then just normally notify others.
7950 if (asoc->base.dead)
7951 return sctp_write_space(sk);
7953 /* Accounting for the sndbuf space is per socket, so we
7954 * need to wake up others, try to be fair and in case of
7955 * other associations, let them have a go first instead
7956 * of just doing a sctp_write_space() call.
7958 * Note that we reach sctp_wake_up_waiters() only when
7959 * associations free up queued chunks, thus we are under
7960 * lock and the list of associations on a socket is
7961 * guaranteed not to change.
7963 for (tmp = list_next_entry(tmp, asocs); 1;
7964 tmp = list_next_entry(tmp, asocs)) {
7965 /* Manually skip the head element. */
7966 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
7968 /* Wake up association. */
7969 __sctp_write_space(tmp);
7970 /* We've reached the end. */
7976 /* Do accounting for the sndbuf space.
7977 * Decrement the used sndbuf space of the corresponding association by the
7978 * data size which was just transmitted(freed).
7980 static void sctp_wfree(struct sk_buff *skb)
7982 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
7983 struct sctp_association *asoc = chunk->asoc;
7984 struct sock *sk = asoc->base.sk;
7986 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
7987 sizeof(struct sk_buff) +
7988 sizeof(struct sctp_chunk);
7990 WARN_ON(refcount_sub_and_test(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc));
7993 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
7995 sk->sk_wmem_queued -= skb->truesize;
7996 sk_mem_uncharge(sk, skb->truesize);
7999 sctp_wake_up_waiters(sk, asoc);
8001 sctp_association_put(asoc);
8004 /* Do accounting for the receive space on the socket.
8005 * Accounting for the association is done in ulpevent.c
8006 * We set this as a destructor for the cloned data skbs so that
8007 * accounting is done at the correct time.
8009 void sctp_sock_rfree(struct sk_buff *skb)
8011 struct sock *sk = skb->sk;
8012 struct sctp_ulpevent *event = sctp_skb2event(skb);
8014 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
8017 * Mimic the behavior of sock_rfree
8019 sk_mem_uncharge(sk, event->rmem_len);
8023 /* Helper function to wait for space in the sndbuf. */
8024 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
8025 size_t msg_len, struct sock **orig_sk)
8027 struct sock *sk = asoc->base.sk;
8029 long current_timeo = *timeo_p;
8032 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
8035 /* Increment the association's refcnt. */
8036 sctp_association_hold(asoc);
8038 /* Wait on the association specific sndbuf space. */
8040 prepare_to_wait_exclusive(&asoc->wait, &wait,
8041 TASK_INTERRUPTIBLE);
8042 if (asoc->base.dead)
8046 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING)
8048 if (signal_pending(current))
8049 goto do_interrupted;
8050 if (msg_len <= sctp_wspace(asoc))
8053 /* Let another process have a go. Since we are going
8057 current_timeo = schedule_timeout(current_timeo);
8059 if (sk != asoc->base.sk) {
8065 *timeo_p = current_timeo;
8070 finish_wait(&asoc->wait, &wait);
8072 /* Release the association's refcnt. */
8073 sctp_association_put(asoc);
8086 err = sock_intr_errno(*timeo_p);
8094 void sctp_data_ready(struct sock *sk)
8096 struct socket_wq *wq;
8099 wq = rcu_dereference(sk->sk_wq);
8100 if (skwq_has_sleeper(wq))
8101 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
8102 POLLRDNORM | POLLRDBAND);
8103 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
8107 /* If socket sndbuf has changed, wake up all per association waiters. */
8108 void sctp_write_space(struct sock *sk)
8110 struct sctp_association *asoc;
8112 /* Wake up the tasks in each wait queue. */
8113 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
8114 __sctp_write_space(asoc);
8118 /* Is there any sndbuf space available on the socket?
8120 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
8121 * associations on the same socket. For a UDP-style socket with
8122 * multiple associations, it is possible for it to be "unwriteable"
8123 * prematurely. I assume that this is acceptable because
8124 * a premature "unwriteable" is better than an accidental "writeable" which
8125 * would cause an unwanted block under certain circumstances. For the 1-1
8126 * UDP-style sockets or TCP-style sockets, this code should work.
8129 static int sctp_writeable(struct sock *sk)
8133 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
8139 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
8140 * returns immediately with EINPROGRESS.
8142 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
8144 struct sock *sk = asoc->base.sk;
8146 long current_timeo = *timeo_p;
8149 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
8151 /* Increment the association's refcnt. */
8152 sctp_association_hold(asoc);
8155 prepare_to_wait_exclusive(&asoc->wait, &wait,
8156 TASK_INTERRUPTIBLE);
8159 if (sk->sk_shutdown & RCV_SHUTDOWN)
8161 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
8164 if (signal_pending(current))
8165 goto do_interrupted;
8167 if (sctp_state(asoc, ESTABLISHED))
8170 /* Let another process have a go. Since we are going
8174 current_timeo = schedule_timeout(current_timeo);
8177 *timeo_p = current_timeo;
8181 finish_wait(&asoc->wait, &wait);
8183 /* Release the association's refcnt. */
8184 sctp_association_put(asoc);
8189 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
8192 err = -ECONNREFUSED;
8196 err = sock_intr_errno(*timeo_p);
8204 static int sctp_wait_for_accept(struct sock *sk, long timeo)
8206 struct sctp_endpoint *ep;
8210 ep = sctp_sk(sk)->ep;
8214 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
8215 TASK_INTERRUPTIBLE);
8217 if (list_empty(&ep->asocs)) {
8219 timeo = schedule_timeout(timeo);
8224 if (!sctp_sstate(sk, LISTENING))
8228 if (!list_empty(&ep->asocs))
8231 err = sock_intr_errno(timeo);
8232 if (signal_pending(current))
8240 finish_wait(sk_sleep(sk), &wait);
8245 static void sctp_wait_for_close(struct sock *sk, long timeout)
8250 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
8251 if (list_empty(&sctp_sk(sk)->ep->asocs))
8254 timeout = schedule_timeout(timeout);
8256 } while (!signal_pending(current) && timeout);
8258 finish_wait(sk_sleep(sk), &wait);
8261 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
8263 struct sk_buff *frag;
8268 /* Don't forget the fragments. */
8269 skb_walk_frags(skb, frag)
8270 sctp_skb_set_owner_r_frag(frag, sk);
8273 sctp_skb_set_owner_r(skb, sk);
8276 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
8277 struct sctp_association *asoc)
8279 struct inet_sock *inet = inet_sk(sk);
8280 struct inet_sock *newinet;
8282 newsk->sk_type = sk->sk_type;
8283 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
8284 newsk->sk_flags = sk->sk_flags;
8285 newsk->sk_tsflags = sk->sk_tsflags;
8286 newsk->sk_no_check_tx = sk->sk_no_check_tx;
8287 newsk->sk_no_check_rx = sk->sk_no_check_rx;
8288 newsk->sk_reuse = sk->sk_reuse;
8290 newsk->sk_shutdown = sk->sk_shutdown;
8291 newsk->sk_destruct = sctp_destruct_sock;
8292 newsk->sk_family = sk->sk_family;
8293 newsk->sk_protocol = IPPROTO_SCTP;
8294 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
8295 newsk->sk_sndbuf = sk->sk_sndbuf;
8296 newsk->sk_rcvbuf = sk->sk_rcvbuf;
8297 newsk->sk_lingertime = sk->sk_lingertime;
8298 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
8299 newsk->sk_sndtimeo = sk->sk_sndtimeo;
8300 newsk->sk_rxhash = sk->sk_rxhash;
8302 newinet = inet_sk(newsk);
8304 /* Initialize sk's sport, dport, rcv_saddr and daddr for
8305 * getsockname() and getpeername()
8307 newinet->inet_sport = inet->inet_sport;
8308 newinet->inet_saddr = inet->inet_saddr;
8309 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
8310 newinet->inet_dport = htons(asoc->peer.port);
8311 newinet->pmtudisc = inet->pmtudisc;
8312 newinet->inet_id = asoc->next_tsn ^ jiffies;
8314 newinet->uc_ttl = inet->uc_ttl;
8315 newinet->mc_loop = 1;
8316 newinet->mc_ttl = 1;
8317 newinet->mc_index = 0;
8318 newinet->mc_list = NULL;
8320 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
8321 net_enable_timestamp();
8323 security_sk_clone(sk, newsk);
8326 static inline void sctp_copy_descendant(struct sock *sk_to,
8327 const struct sock *sk_from)
8329 int ancestor_size = sizeof(struct inet_sock) +
8330 sizeof(struct sctp_sock) -
8331 offsetof(struct sctp_sock, auto_asconf_list);
8333 if (sk_from->sk_family == PF_INET6)
8334 ancestor_size += sizeof(struct ipv6_pinfo);
8336 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
8339 /* Populate the fields of the newsk from the oldsk and migrate the assoc
8340 * and its messages to the newsk.
8342 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
8343 struct sctp_association *assoc,
8344 enum sctp_socket_type type)
8346 struct sctp_sock *oldsp = sctp_sk(oldsk);
8347 struct sctp_sock *newsp = sctp_sk(newsk);
8348 struct sctp_bind_bucket *pp; /* hash list port iterator */
8349 struct sctp_endpoint *newep = newsp->ep;
8350 struct sk_buff *skb, *tmp;
8351 struct sctp_ulpevent *event;
8352 struct sctp_bind_hashbucket *head;
8354 /* Migrate socket buffer sizes and all the socket level options to the
8357 newsk->sk_sndbuf = oldsk->sk_sndbuf;
8358 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
8359 /* Brute force copy old sctp opt. */
8360 sctp_copy_descendant(newsk, oldsk);
8362 /* Restore the ep value that was overwritten with the above structure
8368 /* Hook this new socket in to the bind_hash list. */
8369 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
8370 inet_sk(oldsk)->inet_num)];
8371 spin_lock_bh(&head->lock);
8372 pp = sctp_sk(oldsk)->bind_hash;
8373 sk_add_bind_node(newsk, &pp->owner);
8374 sctp_sk(newsk)->bind_hash = pp;
8375 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
8376 spin_unlock_bh(&head->lock);
8378 /* Copy the bind_addr list from the original endpoint to the new
8379 * endpoint so that we can handle restarts properly
8381 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
8382 &oldsp->ep->base.bind_addr, GFP_KERNEL);
8384 /* Move any messages in the old socket's receive queue that are for the
8385 * peeled off association to the new socket's receive queue.
8387 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
8388 event = sctp_skb2event(skb);
8389 if (event->asoc == assoc) {
8390 __skb_unlink(skb, &oldsk->sk_receive_queue);
8391 __skb_queue_tail(&newsk->sk_receive_queue, skb);
8392 sctp_skb_set_owner_r_frag(skb, newsk);
8396 /* Clean up any messages pending delivery due to partial
8397 * delivery. Three cases:
8398 * 1) No partial deliver; no work.
8399 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
8400 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
8402 skb_queue_head_init(&newsp->pd_lobby);
8403 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
8405 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
8406 struct sk_buff_head *queue;
8408 /* Decide which queue to move pd_lobby skbs to. */
8409 if (assoc->ulpq.pd_mode) {
8410 queue = &newsp->pd_lobby;
8412 queue = &newsk->sk_receive_queue;
8414 /* Walk through the pd_lobby, looking for skbs that
8415 * need moved to the new socket.
8417 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
8418 event = sctp_skb2event(skb);
8419 if (event->asoc == assoc) {
8420 __skb_unlink(skb, &oldsp->pd_lobby);
8421 __skb_queue_tail(queue, skb);
8422 sctp_skb_set_owner_r_frag(skb, newsk);
8426 /* Clear up any skbs waiting for the partial
8427 * delivery to finish.
8429 if (assoc->ulpq.pd_mode)
8430 sctp_clear_pd(oldsk, NULL);
8434 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
8435 sctp_skb_set_owner_r_frag(skb, newsk);
8437 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
8438 sctp_skb_set_owner_r_frag(skb, newsk);
8440 /* Set the type of socket to indicate that it is peeled off from the
8441 * original UDP-style socket or created with the accept() call on a
8442 * TCP-style socket..
8446 /* Mark the new socket "in-use" by the user so that any packets
8447 * that may arrive on the association after we've moved it are
8448 * queued to the backlog. This prevents a potential race between
8449 * backlog processing on the old socket and new-packet processing
8450 * on the new socket.
8452 * The caller has just allocated newsk so we can guarantee that other
8453 * paths won't try to lock it and then oldsk.
8455 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
8456 sctp_for_each_tx_datachunk(assoc, sctp_clear_owner_w);
8457 sctp_assoc_migrate(assoc, newsk);
8458 sctp_for_each_tx_datachunk(assoc, sctp_set_owner_w);
8460 /* If the association on the newsk is already closed before accept()
8461 * is called, set RCV_SHUTDOWN flag.
8463 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP)) {
8464 newsk->sk_state = SCTP_SS_CLOSED;
8465 newsk->sk_shutdown |= RCV_SHUTDOWN;
8467 newsk->sk_state = SCTP_SS_ESTABLISHED;
8470 release_sock(newsk);
8474 /* This proto struct describes the ULP interface for SCTP. */
8475 struct proto sctp_prot = {
8477 .owner = THIS_MODULE,
8478 .close = sctp_close,
8479 .connect = sctp_connect,
8480 .disconnect = sctp_disconnect,
8481 .accept = sctp_accept,
8482 .ioctl = sctp_ioctl,
8483 .init = sctp_init_sock,
8484 .destroy = sctp_destroy_sock,
8485 .shutdown = sctp_shutdown,
8486 .setsockopt = sctp_setsockopt,
8487 .getsockopt = sctp_getsockopt,
8488 .sendmsg = sctp_sendmsg,
8489 .recvmsg = sctp_recvmsg,
8491 .backlog_rcv = sctp_backlog_rcv,
8493 .unhash = sctp_unhash,
8494 .get_port = sctp_get_port,
8495 .obj_size = sizeof(struct sctp_sock),
8496 .sysctl_mem = sysctl_sctp_mem,
8497 .sysctl_rmem = sysctl_sctp_rmem,
8498 .sysctl_wmem = sysctl_sctp_wmem,
8499 .memory_pressure = &sctp_memory_pressure,
8500 .enter_memory_pressure = sctp_enter_memory_pressure,
8501 .memory_allocated = &sctp_memory_allocated,
8502 .sockets_allocated = &sctp_sockets_allocated,
8505 #if IS_ENABLED(CONFIG_IPV6)
8507 #include <net/transp_v6.h>
8508 static void sctp_v6_destroy_sock(struct sock *sk)
8510 sctp_destroy_sock(sk);
8511 inet6_destroy_sock(sk);
8514 struct proto sctpv6_prot = {
8516 .owner = THIS_MODULE,
8517 .close = sctp_close,
8518 .connect = sctp_connect,
8519 .disconnect = sctp_disconnect,
8520 .accept = sctp_accept,
8521 .ioctl = sctp_ioctl,
8522 .init = sctp_init_sock,
8523 .destroy = sctp_v6_destroy_sock,
8524 .shutdown = sctp_shutdown,
8525 .setsockopt = sctp_setsockopt,
8526 .getsockopt = sctp_getsockopt,
8527 .sendmsg = sctp_sendmsg,
8528 .recvmsg = sctp_recvmsg,
8530 .backlog_rcv = sctp_backlog_rcv,
8532 .unhash = sctp_unhash,
8533 .get_port = sctp_get_port,
8534 .obj_size = sizeof(struct sctp6_sock),
8535 .sysctl_mem = sysctl_sctp_mem,
8536 .sysctl_rmem = sysctl_sctp_rmem,
8537 .sysctl_wmem = sysctl_sctp_wmem,
8538 .memory_pressure = &sctp_memory_pressure,
8539 .enter_memory_pressure = sctp_enter_memory_pressure,
8540 .memory_allocated = &sctp_memory_allocated,
8541 .sockets_allocated = &sctp_sockets_allocated,
8543 #endif /* IS_ENABLED(CONFIG_IPV6) */