1 /* SCTP kernel implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
9 * This file is part of the SCTP kernel implementation
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
18 * This SCTP implementation is free software;
19 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
24 * This SCTP implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, see
32 * <http://www.gnu.org/licenses/>.
34 * Please send any bug reports or fixes you make to the
36 * lksctp developers <linux-sctp@vger.kernel.org>
38 * Written or modified by:
39 * La Monte H.P. Yarroll <piggy@acm.org>
40 * Narasimha Budihal <narsi@refcode.org>
41 * Karl Knutson <karl@athena.chicago.il.us>
42 * Jon Grimm <jgrimm@us.ibm.com>
43 * Xingang Guo <xingang.guo@intel.com>
44 * Daisy Chang <daisyc@us.ibm.com>
45 * Sridhar Samudrala <samudrala@us.ibm.com>
46 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
47 * Ardelle Fan <ardelle.fan@intel.com>
48 * Ryan Layer <rmlayer@us.ibm.com>
49 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
50 * Kevin Gao <kevin.gao@intel.com>
53 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
55 #include <crypto/hash.h>
56 #include <linux/types.h>
57 #include <linux/kernel.h>
58 #include <linux/wait.h>
59 #include <linux/time.h>
61 #include <linux/capability.h>
62 #include <linux/fcntl.h>
63 #include <linux/poll.h>
64 #include <linux/init.h>
65 #include <linux/slab.h>
66 #include <linux/file.h>
67 #include <linux/compat.h>
71 #include <net/route.h>
73 #include <net/inet_common.h>
74 #include <net/busy_poll.h>
76 #include <linux/socket.h> /* for sa_family_t */
77 #include <linux/export.h>
79 #include <net/sctp/sctp.h>
80 #include <net/sctp/sm.h>
82 /* Forward declarations for internal helper functions. */
83 static int sctp_writeable(struct sock *sk);
84 static void sctp_wfree(struct sk_buff *skb);
85 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
87 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p);
88 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
89 static int sctp_wait_for_accept(struct sock *sk, long timeo);
90 static void sctp_wait_for_close(struct sock *sk, long timeo);
91 static void sctp_destruct_sock(struct sock *sk);
92 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
93 union sctp_addr *addr, int len);
94 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
95 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
96 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
97 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
98 static int sctp_send_asconf(struct sctp_association *asoc,
99 struct sctp_chunk *chunk);
100 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
101 static int sctp_autobind(struct sock *sk);
102 static void sctp_sock_migrate(struct sock *, struct sock *,
103 struct sctp_association *, sctp_socket_type_t);
105 static int sctp_memory_pressure;
106 static atomic_long_t sctp_memory_allocated;
107 struct percpu_counter sctp_sockets_allocated;
109 static void sctp_enter_memory_pressure(struct sock *sk)
111 sctp_memory_pressure = 1;
115 /* Get the sndbuf space available at the time on the association. */
116 static inline int sctp_wspace(struct sctp_association *asoc)
120 if (asoc->ep->sndbuf_policy)
121 amt = asoc->sndbuf_used;
123 amt = sk_wmem_alloc_get(asoc->base.sk);
125 if (amt >= asoc->base.sk->sk_sndbuf) {
126 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
129 amt = sk_stream_wspace(asoc->base.sk);
134 amt = asoc->base.sk->sk_sndbuf - amt;
139 /* Increment the used sndbuf space count of the corresponding association by
140 * the size of the outgoing data chunk.
141 * Also, set the skb destructor for sndbuf accounting later.
143 * Since it is always 1-1 between chunk and skb, and also a new skb is always
144 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
145 * destructor in the data chunk skb for the purpose of the sndbuf space
148 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
150 struct sctp_association *asoc = chunk->asoc;
151 struct sock *sk = asoc->base.sk;
153 /* The sndbuf space is tracked per association. */
154 sctp_association_hold(asoc);
156 skb_set_owner_w(chunk->skb, sk);
158 chunk->skb->destructor = sctp_wfree;
159 /* Save the chunk pointer in skb for sctp_wfree to use later. */
160 skb_shinfo(chunk->skb)->destructor_arg = chunk;
162 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
163 sizeof(struct sk_buff) +
164 sizeof(struct sctp_chunk);
166 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
167 sk->sk_wmem_queued += chunk->skb->truesize;
168 sk_mem_charge(sk, chunk->skb->truesize);
171 /* Verify that this is a valid address. */
172 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
177 /* Verify basic sockaddr. */
178 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
182 /* Is this a valid SCTP address? */
183 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
186 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
192 /* Look up the association by its id. If this is not a UDP-style
193 * socket, the ID field is always ignored.
195 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
197 struct sctp_association *asoc = NULL;
199 /* If this is not a UDP-style socket, assoc id should be ignored. */
200 if (!sctp_style(sk, UDP)) {
201 /* Return NULL if the socket state is not ESTABLISHED. It
202 * could be a TCP-style listening socket or a socket which
203 * hasn't yet called connect() to establish an association.
205 if (!sctp_sstate(sk, ESTABLISHED))
208 /* Get the first and the only association from the list. */
209 if (!list_empty(&sctp_sk(sk)->ep->asocs))
210 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
211 struct sctp_association, asocs);
215 /* Otherwise this is a UDP-style socket. */
216 if (!id || (id == (sctp_assoc_t)-1))
219 spin_lock_bh(&sctp_assocs_id_lock);
220 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
221 spin_unlock_bh(&sctp_assocs_id_lock);
223 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
229 /* Look up the transport from an address and an assoc id. If both address and
230 * id are specified, the associations matching the address and the id should be
233 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
234 struct sockaddr_storage *addr,
237 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
238 struct sctp_transport *transport;
239 union sctp_addr *laddr = (union sctp_addr *)addr;
241 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
248 id_asoc = sctp_id2assoc(sk, id);
249 if (id_asoc && (id_asoc != addr_asoc))
252 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
253 (union sctp_addr *)addr);
258 /* API 3.1.2 bind() - UDP Style Syntax
259 * The syntax of bind() is,
261 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
263 * sd - the socket descriptor returned by socket().
264 * addr - the address structure (struct sockaddr_in or struct
265 * sockaddr_in6 [RFC 2553]),
266 * addr_len - the size of the address structure.
268 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
274 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
277 /* Disallow binding twice. */
278 if (!sctp_sk(sk)->ep->base.bind_addr.port)
279 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
289 static long sctp_get_port_local(struct sock *, union sctp_addr *);
291 /* Verify this is a valid sockaddr. */
292 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
293 union sctp_addr *addr, int len)
297 /* Check minimum size. */
298 if (len < sizeof (struct sockaddr))
301 /* V4 mapped address are really of AF_INET family */
302 if (addr->sa.sa_family == AF_INET6 &&
303 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
304 if (!opt->pf->af_supported(AF_INET, opt))
307 /* Does this PF support this AF? */
308 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
312 /* If we get this far, af is valid. */
313 af = sctp_get_af_specific(addr->sa.sa_family);
315 if (len < af->sockaddr_len)
321 /* Bind a local address either to an endpoint or to an association. */
322 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
324 struct net *net = sock_net(sk);
325 struct sctp_sock *sp = sctp_sk(sk);
326 struct sctp_endpoint *ep = sp->ep;
327 struct sctp_bind_addr *bp = &ep->base.bind_addr;
332 /* Common sockaddr verification. */
333 af = sctp_sockaddr_af(sp, addr, len);
335 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
336 __func__, sk, addr, len);
340 snum = ntohs(addr->v4.sin_port);
342 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
343 __func__, sk, &addr->sa, bp->port, snum, len);
345 /* PF specific bind() address verification. */
346 if (!sp->pf->bind_verify(sp, addr))
347 return -EADDRNOTAVAIL;
349 /* We must either be unbound, or bind to the same port.
350 * It's OK to allow 0 ports if we are already bound.
351 * We'll just inhert an already bound port in this case
356 else if (snum != bp->port) {
357 pr_debug("%s: new port %d doesn't match existing port "
358 "%d\n", __func__, snum, bp->port);
363 if (snum && snum < PROT_SOCK &&
364 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
367 /* See if the address matches any of the addresses we may have
368 * already bound before checking against other endpoints.
370 if (sctp_bind_addr_match(bp, addr, sp))
373 /* Make sure we are allowed to bind here.
374 * The function sctp_get_port_local() does duplicate address
377 addr->v4.sin_port = htons(snum);
378 if ((ret = sctp_get_port_local(sk, addr))) {
382 /* Refresh ephemeral port. */
384 bp->port = inet_sk(sk)->inet_num;
386 /* Add the address to the bind address list.
387 * Use GFP_ATOMIC since BHs will be disabled.
389 ret = sctp_add_bind_addr(bp, addr, af->sockaddr_len,
390 SCTP_ADDR_SRC, GFP_ATOMIC);
392 /* Copy back into socket for getsockname() use. */
394 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
395 sp->pf->to_sk_saddr(addr, sk);
401 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
403 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
404 * at any one time. If a sender, after sending an ASCONF chunk, decides
405 * it needs to transfer another ASCONF Chunk, it MUST wait until the
406 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
407 * subsequent ASCONF. Note this restriction binds each side, so at any
408 * time two ASCONF may be in-transit on any given association (one sent
409 * from each endpoint).
411 static int sctp_send_asconf(struct sctp_association *asoc,
412 struct sctp_chunk *chunk)
414 struct net *net = sock_net(asoc->base.sk);
417 /* If there is an outstanding ASCONF chunk, queue it for later
420 if (asoc->addip_last_asconf) {
421 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
425 /* Hold the chunk until an ASCONF_ACK is received. */
426 sctp_chunk_hold(chunk);
427 retval = sctp_primitive_ASCONF(net, asoc, chunk);
429 sctp_chunk_free(chunk);
431 asoc->addip_last_asconf = chunk;
437 /* Add a list of addresses as bind addresses to local endpoint or
440 * Basically run through each address specified in the addrs/addrcnt
441 * array/length pair, determine if it is IPv6 or IPv4 and call
442 * sctp_do_bind() on it.
444 * If any of them fails, then the operation will be reversed and the
445 * ones that were added will be removed.
447 * Only sctp_setsockopt_bindx() is supposed to call this function.
449 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
454 struct sockaddr *sa_addr;
457 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
461 for (cnt = 0; cnt < addrcnt; cnt++) {
462 /* The list may contain either IPv4 or IPv6 address;
463 * determine the address length for walking thru the list.
466 af = sctp_get_af_specific(sa_addr->sa_family);
472 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
475 addr_buf += af->sockaddr_len;
479 /* Failed. Cleanup the ones that have been added */
481 sctp_bindx_rem(sk, addrs, cnt);
489 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
490 * associations that are part of the endpoint indicating that a list of local
491 * addresses are added to the endpoint.
493 * If any of the addresses is already in the bind address list of the
494 * association, we do not send the chunk for that association. But it will not
495 * affect other associations.
497 * Only sctp_setsockopt_bindx() is supposed to call this function.
499 static int sctp_send_asconf_add_ip(struct sock *sk,
500 struct sockaddr *addrs,
503 struct net *net = sock_net(sk);
504 struct sctp_sock *sp;
505 struct sctp_endpoint *ep;
506 struct sctp_association *asoc;
507 struct sctp_bind_addr *bp;
508 struct sctp_chunk *chunk;
509 struct sctp_sockaddr_entry *laddr;
510 union sctp_addr *addr;
511 union sctp_addr saveaddr;
518 if (!net->sctp.addip_enable)
524 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
525 __func__, sk, addrs, addrcnt);
527 list_for_each_entry(asoc, &ep->asocs, asocs) {
528 if (!asoc->peer.asconf_capable)
531 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
534 if (!sctp_state(asoc, ESTABLISHED))
537 /* Check if any address in the packed array of addresses is
538 * in the bind address list of the association. If so,
539 * do not send the asconf chunk to its peer, but continue with
540 * other associations.
543 for (i = 0; i < addrcnt; i++) {
545 af = sctp_get_af_specific(addr->v4.sin_family);
551 if (sctp_assoc_lookup_laddr(asoc, addr))
554 addr_buf += af->sockaddr_len;
559 /* Use the first valid address in bind addr list of
560 * association as Address Parameter of ASCONF CHUNK.
562 bp = &asoc->base.bind_addr;
563 p = bp->address_list.next;
564 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
565 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
566 addrcnt, SCTP_PARAM_ADD_IP);
572 /* Add the new addresses to the bind address list with
573 * use_as_src set to 0.
576 for (i = 0; i < addrcnt; i++) {
578 af = sctp_get_af_specific(addr->v4.sin_family);
579 memcpy(&saveaddr, addr, af->sockaddr_len);
580 retval = sctp_add_bind_addr(bp, &saveaddr,
582 SCTP_ADDR_NEW, GFP_ATOMIC);
583 addr_buf += af->sockaddr_len;
585 if (asoc->src_out_of_asoc_ok) {
586 struct sctp_transport *trans;
588 list_for_each_entry(trans,
589 &asoc->peer.transport_addr_list, transports) {
590 /* Clear the source and route cache */
591 dst_release(trans->dst);
592 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
593 2*asoc->pathmtu, 4380));
594 trans->ssthresh = asoc->peer.i.a_rwnd;
595 trans->rto = asoc->rto_initial;
596 sctp_max_rto(asoc, trans);
597 trans->rtt = trans->srtt = trans->rttvar = 0;
598 sctp_transport_route(trans, NULL,
599 sctp_sk(asoc->base.sk));
602 retval = sctp_send_asconf(asoc, chunk);
609 /* Remove a list of addresses from bind addresses list. Do not remove the
612 * Basically run through each address specified in the addrs/addrcnt
613 * array/length pair, determine if it is IPv6 or IPv4 and call
614 * sctp_del_bind() on it.
616 * If any of them fails, then the operation will be reversed and the
617 * ones that were removed will be added back.
619 * At least one address has to be left; if only one address is
620 * available, the operation will return -EBUSY.
622 * Only sctp_setsockopt_bindx() is supposed to call this function.
624 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
626 struct sctp_sock *sp = sctp_sk(sk);
627 struct sctp_endpoint *ep = sp->ep;
629 struct sctp_bind_addr *bp = &ep->base.bind_addr;
632 union sctp_addr *sa_addr;
635 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
636 __func__, sk, addrs, addrcnt);
639 for (cnt = 0; cnt < addrcnt; cnt++) {
640 /* If the bind address list is empty or if there is only one
641 * bind address, there is nothing more to be removed (we need
642 * at least one address here).
644 if (list_empty(&bp->address_list) ||
645 (sctp_list_single_entry(&bp->address_list))) {
651 af = sctp_get_af_specific(sa_addr->sa.sa_family);
657 if (!af->addr_valid(sa_addr, sp, NULL)) {
658 retval = -EADDRNOTAVAIL;
662 if (sa_addr->v4.sin_port &&
663 sa_addr->v4.sin_port != htons(bp->port)) {
668 if (!sa_addr->v4.sin_port)
669 sa_addr->v4.sin_port = htons(bp->port);
671 /* FIXME - There is probably a need to check if sk->sk_saddr and
672 * sk->sk_rcv_addr are currently set to one of the addresses to
673 * be removed. This is something which needs to be looked into
674 * when we are fixing the outstanding issues with multi-homing
675 * socket routing and failover schemes. Refer to comments in
676 * sctp_do_bind(). -daisy
678 retval = sctp_del_bind_addr(bp, sa_addr);
680 addr_buf += af->sockaddr_len;
683 /* Failed. Add the ones that has been removed back */
685 sctp_bindx_add(sk, addrs, cnt);
693 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
694 * the associations that are part of the endpoint indicating that a list of
695 * local addresses are removed from the endpoint.
697 * If any of the addresses is already in the bind address list of the
698 * association, we do not send the chunk for that association. But it will not
699 * affect other associations.
701 * Only sctp_setsockopt_bindx() is supposed to call this function.
703 static int sctp_send_asconf_del_ip(struct sock *sk,
704 struct sockaddr *addrs,
707 struct net *net = sock_net(sk);
708 struct sctp_sock *sp;
709 struct sctp_endpoint *ep;
710 struct sctp_association *asoc;
711 struct sctp_transport *transport;
712 struct sctp_bind_addr *bp;
713 struct sctp_chunk *chunk;
714 union sctp_addr *laddr;
717 struct sctp_sockaddr_entry *saddr;
723 if (!net->sctp.addip_enable)
729 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
730 __func__, sk, addrs, addrcnt);
732 list_for_each_entry(asoc, &ep->asocs, asocs) {
734 if (!asoc->peer.asconf_capable)
737 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
740 if (!sctp_state(asoc, ESTABLISHED))
743 /* Check if any address in the packed array of addresses is
744 * not present in the bind address list of the association.
745 * If so, do not send the asconf chunk to its peer, but
746 * continue with other associations.
749 for (i = 0; i < addrcnt; i++) {
751 af = sctp_get_af_specific(laddr->v4.sin_family);
757 if (!sctp_assoc_lookup_laddr(asoc, laddr))
760 addr_buf += af->sockaddr_len;
765 /* Find one address in the association's bind address list
766 * that is not in the packed array of addresses. This is to
767 * make sure that we do not delete all the addresses in the
770 bp = &asoc->base.bind_addr;
771 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
773 if ((laddr == NULL) && (addrcnt == 1)) {
774 if (asoc->asconf_addr_del_pending)
776 asoc->asconf_addr_del_pending =
777 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
778 if (asoc->asconf_addr_del_pending == NULL) {
782 asoc->asconf_addr_del_pending->sa.sa_family =
784 asoc->asconf_addr_del_pending->v4.sin_port =
786 if (addrs->sa_family == AF_INET) {
787 struct sockaddr_in *sin;
789 sin = (struct sockaddr_in *)addrs;
790 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
791 } else if (addrs->sa_family == AF_INET6) {
792 struct sockaddr_in6 *sin6;
794 sin6 = (struct sockaddr_in6 *)addrs;
795 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
798 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
799 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
800 asoc->asconf_addr_del_pending);
802 asoc->src_out_of_asoc_ok = 1;
810 /* We do not need RCU protection throughout this loop
811 * because this is done under a socket lock from the
814 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
822 /* Reset use_as_src flag for the addresses in the bind address
823 * list that are to be deleted.
826 for (i = 0; i < addrcnt; i++) {
828 af = sctp_get_af_specific(laddr->v4.sin_family);
829 list_for_each_entry(saddr, &bp->address_list, list) {
830 if (sctp_cmp_addr_exact(&saddr->a, laddr))
831 saddr->state = SCTP_ADDR_DEL;
833 addr_buf += af->sockaddr_len;
836 /* Update the route and saddr entries for all the transports
837 * as some of the addresses in the bind address list are
838 * about to be deleted and cannot be used as source addresses.
840 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
842 dst_release(transport->dst);
843 sctp_transport_route(transport, NULL,
844 sctp_sk(asoc->base.sk));
848 /* We don't need to transmit ASCONF */
850 retval = sctp_send_asconf(asoc, chunk);
856 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
857 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
859 struct sock *sk = sctp_opt2sk(sp);
860 union sctp_addr *addr;
863 /* It is safe to write port space in caller. */
865 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
866 af = sctp_get_af_specific(addr->sa.sa_family);
869 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
872 if (addrw->state == SCTP_ADDR_NEW)
873 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
875 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
878 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
881 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
884 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
885 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
888 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
889 * Section 3.1.2 for this usage.
891 * addrs is a pointer to an array of one or more socket addresses. Each
892 * address is contained in its appropriate structure (i.e. struct
893 * sockaddr_in or struct sockaddr_in6) the family of the address type
894 * must be used to distinguish the address length (note that this
895 * representation is termed a "packed array" of addresses). The caller
896 * specifies the number of addresses in the array with addrcnt.
898 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
899 * -1, and sets errno to the appropriate error code.
901 * For SCTP, the port given in each socket address must be the same, or
902 * sctp_bindx() will fail, setting errno to EINVAL.
904 * The flags parameter is formed from the bitwise OR of zero or more of
905 * the following currently defined flags:
907 * SCTP_BINDX_ADD_ADDR
909 * SCTP_BINDX_REM_ADDR
911 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
912 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
913 * addresses from the association. The two flags are mutually exclusive;
914 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
915 * not remove all addresses from an association; sctp_bindx() will
916 * reject such an attempt with EINVAL.
918 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
919 * additional addresses with an endpoint after calling bind(). Or use
920 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
921 * socket is associated with so that no new association accepted will be
922 * associated with those addresses. If the endpoint supports dynamic
923 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
924 * endpoint to send the appropriate message to the peer to change the
925 * peers address lists.
927 * Adding and removing addresses from a connected association is
928 * optional functionality. Implementations that do not support this
929 * functionality should return EOPNOTSUPP.
931 * Basically do nothing but copying the addresses from user to kernel
932 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
933 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
936 * We don't use copy_from_user() for optimization: we first do the
937 * sanity checks (buffer size -fast- and access check-healthy
938 * pointer); if all of those succeed, then we can alloc the memory
939 * (expensive operation) needed to copy the data to kernel. Then we do
940 * the copying without checking the user space area
941 * (__copy_from_user()).
943 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
946 * sk The sk of the socket
947 * addrs The pointer to the addresses in user land
948 * addrssize Size of the addrs buffer
949 * op Operation to perform (add or remove, see the flags of
952 * Returns 0 if ok, <0 errno code on error.
954 static int sctp_setsockopt_bindx(struct sock *sk,
955 struct sockaddr __user *addrs,
956 int addrs_size, int op)
958 struct sockaddr *kaddrs;
962 struct sockaddr *sa_addr;
966 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
967 __func__, sk, addrs, addrs_size, op);
969 if (unlikely(addrs_size <= 0))
972 /* Check the user passed a healthy pointer. */
973 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
976 /* Alloc space for the address array in kernel memory. */
977 kaddrs = kmalloc(addrs_size, GFP_USER | __GFP_NOWARN);
978 if (unlikely(!kaddrs))
981 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
986 /* Walk through the addrs buffer and count the number of addresses. */
988 while (walk_size < addrs_size) {
989 if (walk_size + sizeof(sa_family_t) > addrs_size) {
995 af = sctp_get_af_specific(sa_addr->sa_family);
997 /* If the address family is not supported or if this address
998 * causes the address buffer to overflow return EINVAL.
1000 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1005 addr_buf += af->sockaddr_len;
1006 walk_size += af->sockaddr_len;
1011 case SCTP_BINDX_ADD_ADDR:
1012 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1015 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1018 case SCTP_BINDX_REM_ADDR:
1019 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1022 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1036 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1038 * Common routine for handling connect() and sctp_connectx().
1039 * Connect will come in with just a single address.
1041 static int __sctp_connect(struct sock *sk,
1042 struct sockaddr *kaddrs,
1044 sctp_assoc_t *assoc_id)
1046 struct net *net = sock_net(sk);
1047 struct sctp_sock *sp;
1048 struct sctp_endpoint *ep;
1049 struct sctp_association *asoc = NULL;
1050 struct sctp_association *asoc2;
1051 struct sctp_transport *transport;
1058 union sctp_addr *sa_addr = NULL;
1060 unsigned short port;
1061 unsigned int f_flags = 0;
1066 /* connect() cannot be done on a socket that is already in ESTABLISHED
1067 * state - UDP-style peeled off socket or a TCP-style socket that
1068 * is already connected.
1069 * It cannot be done even on a TCP-style listening socket.
1071 if (sctp_sstate(sk, ESTABLISHED) ||
1072 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1077 /* Walk through the addrs buffer and count the number of addresses. */
1079 while (walk_size < addrs_size) {
1082 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1088 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1090 /* If the address family is not supported or if this address
1091 * causes the address buffer to overflow return EINVAL.
1093 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1098 port = ntohs(sa_addr->v4.sin_port);
1100 /* Save current address so we can work with it */
1101 memcpy(&to, sa_addr, af->sockaddr_len);
1103 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1107 /* Make sure the destination port is correctly set
1110 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1115 /* Check if there already is a matching association on the
1116 * endpoint (other than the one created here).
1118 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1119 if (asoc2 && asoc2 != asoc) {
1120 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1127 /* If we could not find a matching association on the endpoint,
1128 * make sure that there is no peeled-off association matching
1129 * the peer address even on another socket.
1131 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1132 err = -EADDRNOTAVAIL;
1137 /* If a bind() or sctp_bindx() is not called prior to
1138 * an sctp_connectx() call, the system picks an
1139 * ephemeral port and will choose an address set
1140 * equivalent to binding with a wildcard address.
1142 if (!ep->base.bind_addr.port) {
1143 if (sctp_autobind(sk)) {
1149 * If an unprivileged user inherits a 1-many
1150 * style socket with open associations on a
1151 * privileged port, it MAY be permitted to
1152 * accept new associations, but it SHOULD NOT
1153 * be permitted to open new associations.
1155 if (ep->base.bind_addr.port < PROT_SOCK &&
1156 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1162 scope = sctp_scope(&to);
1163 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1169 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1177 /* Prime the peer's transport structures. */
1178 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1186 addr_buf += af->sockaddr_len;
1187 walk_size += af->sockaddr_len;
1190 /* In case the user of sctp_connectx() wants an association
1191 * id back, assign one now.
1194 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1199 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1204 /* Initialize sk's dport and daddr for getpeername() */
1205 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1206 sp->pf->to_sk_daddr(sa_addr, sk);
1209 /* in-kernel sockets don't generally have a file allocated to them
1210 * if all they do is call sock_create_kern().
1212 if (sk->sk_socket->file)
1213 f_flags = sk->sk_socket->file->f_flags;
1215 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1217 err = sctp_wait_for_connect(asoc, &timeo);
1218 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1219 *assoc_id = asoc->assoc_id;
1221 /* Don't free association on exit. */
1225 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1226 __func__, asoc, kaddrs, err);
1229 /* sctp_primitive_ASSOCIATE may have added this association
1230 * To the hash table, try to unhash it, just in case, its a noop
1231 * if it wasn't hashed so we're safe
1233 sctp_association_free(asoc);
1238 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1241 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1242 * sctp_assoc_t *asoc);
1244 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1245 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1246 * or IPv6 addresses.
1248 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1249 * Section 3.1.2 for this usage.
1251 * addrs is a pointer to an array of one or more socket addresses. Each
1252 * address is contained in its appropriate structure (i.e. struct
1253 * sockaddr_in or struct sockaddr_in6) the family of the address type
1254 * must be used to distengish the address length (note that this
1255 * representation is termed a "packed array" of addresses). The caller
1256 * specifies the number of addresses in the array with addrcnt.
1258 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1259 * the association id of the new association. On failure, sctp_connectx()
1260 * returns -1, and sets errno to the appropriate error code. The assoc_id
1261 * is not touched by the kernel.
1263 * For SCTP, the port given in each socket address must be the same, or
1264 * sctp_connectx() will fail, setting errno to EINVAL.
1266 * An application can use sctp_connectx to initiate an association with
1267 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1268 * allows a caller to specify multiple addresses at which a peer can be
1269 * reached. The way the SCTP stack uses the list of addresses to set up
1270 * the association is implementation dependent. This function only
1271 * specifies that the stack will try to make use of all the addresses in
1272 * the list when needed.
1274 * Note that the list of addresses passed in is only used for setting up
1275 * the association. It does not necessarily equal the set of addresses
1276 * the peer uses for the resulting association. If the caller wants to
1277 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1278 * retrieve them after the association has been set up.
1280 * Basically do nothing but copying the addresses from user to kernel
1281 * land and invoking either sctp_connectx(). This is used for tunneling
1282 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1284 * We don't use copy_from_user() for optimization: we first do the
1285 * sanity checks (buffer size -fast- and access check-healthy
1286 * pointer); if all of those succeed, then we can alloc the memory
1287 * (expensive operation) needed to copy the data to kernel. Then we do
1288 * the copying without checking the user space area
1289 * (__copy_from_user()).
1291 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1294 * sk The sk of the socket
1295 * addrs The pointer to the addresses in user land
1296 * addrssize Size of the addrs buffer
1298 * Returns >=0 if ok, <0 errno code on error.
1300 static int __sctp_setsockopt_connectx(struct sock *sk,
1301 struct sockaddr __user *addrs,
1303 sctp_assoc_t *assoc_id)
1305 struct sockaddr *kaddrs;
1306 gfp_t gfp = GFP_KERNEL;
1309 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1310 __func__, sk, addrs, addrs_size);
1312 if (unlikely(addrs_size <= 0))
1315 /* Check the user passed a healthy pointer. */
1316 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1319 /* Alloc space for the address array in kernel memory. */
1320 if (sk->sk_socket->file)
1321 gfp = GFP_USER | __GFP_NOWARN;
1322 kaddrs = kmalloc(addrs_size, gfp);
1323 if (unlikely(!kaddrs))
1326 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1329 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1338 * This is an older interface. It's kept for backward compatibility
1339 * to the option that doesn't provide association id.
1341 static int sctp_setsockopt_connectx_old(struct sock *sk,
1342 struct sockaddr __user *addrs,
1345 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1349 * New interface for the API. The since the API is done with a socket
1350 * option, to make it simple we feed back the association id is as a return
1351 * indication to the call. Error is always negative and association id is
1354 static int sctp_setsockopt_connectx(struct sock *sk,
1355 struct sockaddr __user *addrs,
1358 sctp_assoc_t assoc_id = 0;
1361 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1370 * New (hopefully final) interface for the API.
1371 * We use the sctp_getaddrs_old structure so that use-space library
1372 * can avoid any unnecessary allocations. The only different part
1373 * is that we store the actual length of the address buffer into the
1374 * addrs_num structure member. That way we can re-use the existing
1377 #ifdef CONFIG_COMPAT
1378 struct compat_sctp_getaddrs_old {
1379 sctp_assoc_t assoc_id;
1381 compat_uptr_t addrs; /* struct sockaddr * */
1385 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1386 char __user *optval,
1389 struct sctp_getaddrs_old param;
1390 sctp_assoc_t assoc_id = 0;
1393 #ifdef CONFIG_COMPAT
1394 if (in_compat_syscall()) {
1395 struct compat_sctp_getaddrs_old param32;
1397 if (len < sizeof(param32))
1399 if (copy_from_user(¶m32, optval, sizeof(param32)))
1402 param.assoc_id = param32.assoc_id;
1403 param.addr_num = param32.addr_num;
1404 param.addrs = compat_ptr(param32.addrs);
1408 if (len < sizeof(param))
1410 if (copy_from_user(¶m, optval, sizeof(param)))
1414 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1415 param.addrs, param.addr_num,
1417 if (err == 0 || err == -EINPROGRESS) {
1418 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1420 if (put_user(sizeof(assoc_id), optlen))
1427 /* API 3.1.4 close() - UDP Style Syntax
1428 * Applications use close() to perform graceful shutdown (as described in
1429 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1430 * by a UDP-style socket.
1434 * ret = close(int sd);
1436 * sd - the socket descriptor of the associations to be closed.
1438 * To gracefully shutdown a specific association represented by the
1439 * UDP-style socket, an application should use the sendmsg() call,
1440 * passing no user data, but including the appropriate flag in the
1441 * ancillary data (see Section xxxx).
1443 * If sd in the close() call is a branched-off socket representing only
1444 * one association, the shutdown is performed on that association only.
1446 * 4.1.6 close() - TCP Style Syntax
1448 * Applications use close() to gracefully close down an association.
1452 * int close(int sd);
1454 * sd - the socket descriptor of the association to be closed.
1456 * After an application calls close() on a socket descriptor, no further
1457 * socket operations will succeed on that descriptor.
1459 * API 7.1.4 SO_LINGER
1461 * An application using the TCP-style socket can use this option to
1462 * perform the SCTP ABORT primitive. The linger option structure is:
1465 * int l_onoff; // option on/off
1466 * int l_linger; // linger time
1469 * To enable the option, set l_onoff to 1. If the l_linger value is set
1470 * to 0, calling close() is the same as the ABORT primitive. If the
1471 * value is set to a negative value, the setsockopt() call will return
1472 * an error. If the value is set to a positive value linger_time, the
1473 * close() can be blocked for at most linger_time ms. If the graceful
1474 * shutdown phase does not finish during this period, close() will
1475 * return but the graceful shutdown phase continues in the system.
1477 static void sctp_close(struct sock *sk, long timeout)
1479 struct net *net = sock_net(sk);
1480 struct sctp_endpoint *ep;
1481 struct sctp_association *asoc;
1482 struct list_head *pos, *temp;
1483 unsigned int data_was_unread;
1485 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1488 sk->sk_shutdown = SHUTDOWN_MASK;
1489 sk->sk_state = SCTP_SS_CLOSING;
1491 ep = sctp_sk(sk)->ep;
1493 /* Clean up any skbs sitting on the receive queue. */
1494 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1495 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1497 /* Walk all associations on an endpoint. */
1498 list_for_each_safe(pos, temp, &ep->asocs) {
1499 asoc = list_entry(pos, struct sctp_association, asocs);
1501 if (sctp_style(sk, TCP)) {
1502 /* A closed association can still be in the list if
1503 * it belongs to a TCP-style listening socket that is
1504 * not yet accepted. If so, free it. If not, send an
1505 * ABORT or SHUTDOWN based on the linger options.
1507 if (sctp_state(asoc, CLOSED)) {
1508 sctp_association_free(asoc);
1513 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1514 !skb_queue_empty(&asoc->ulpq.reasm) ||
1515 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1516 struct sctp_chunk *chunk;
1518 chunk = sctp_make_abort_user(asoc, NULL, 0);
1519 sctp_primitive_ABORT(net, asoc, chunk);
1521 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1524 /* On a TCP-style socket, block for at most linger_time if set. */
1525 if (sctp_style(sk, TCP) && timeout)
1526 sctp_wait_for_close(sk, timeout);
1528 /* This will run the backlog queue. */
1531 /* Supposedly, no process has access to the socket, but
1532 * the net layers still may.
1533 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1534 * held and that should be grabbed before socket lock.
1536 spin_lock_bh(&net->sctp.addr_wq_lock);
1539 /* Hold the sock, since sk_common_release() will put sock_put()
1540 * and we have just a little more cleanup.
1543 sk_common_release(sk);
1546 spin_unlock_bh(&net->sctp.addr_wq_lock);
1550 SCTP_DBG_OBJCNT_DEC(sock);
1553 /* Handle EPIPE error. */
1554 static int sctp_error(struct sock *sk, int flags, int err)
1557 err = sock_error(sk) ? : -EPIPE;
1558 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1559 send_sig(SIGPIPE, current, 0);
1563 /* API 3.1.3 sendmsg() - UDP Style Syntax
1565 * An application uses sendmsg() and recvmsg() calls to transmit data to
1566 * and receive data from its peer.
1568 * ssize_t sendmsg(int socket, const struct msghdr *message,
1571 * socket - the socket descriptor of the endpoint.
1572 * message - pointer to the msghdr structure which contains a single
1573 * user message and possibly some ancillary data.
1575 * See Section 5 for complete description of the data
1578 * flags - flags sent or received with the user message, see Section
1579 * 5 for complete description of the flags.
1581 * Note: This function could use a rewrite especially when explicit
1582 * connect support comes in.
1584 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1586 static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1588 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
1590 struct net *net = sock_net(sk);
1591 struct sctp_sock *sp;
1592 struct sctp_endpoint *ep;
1593 struct sctp_association *new_asoc = NULL, *asoc = NULL;
1594 struct sctp_transport *transport, *chunk_tp;
1595 struct sctp_chunk *chunk;
1597 struct sockaddr *msg_name = NULL;
1598 struct sctp_sndrcvinfo default_sinfo;
1599 struct sctp_sndrcvinfo *sinfo;
1600 struct sctp_initmsg *sinit;
1601 sctp_assoc_t associd = 0;
1602 sctp_cmsgs_t cmsgs = { NULL };
1604 bool fill_sinfo_ttl = false, wait_connect = false;
1605 struct sctp_datamsg *datamsg;
1606 int msg_flags = msg->msg_flags;
1607 __u16 sinfo_flags = 0;
1615 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1618 /* We cannot send a message over a TCP-style listening socket. */
1619 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1624 /* Parse out the SCTP CMSGs. */
1625 err = sctp_msghdr_parse(msg, &cmsgs);
1627 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1631 /* Fetch the destination address for this packet. This
1632 * address only selects the association--it is not necessarily
1633 * the address we will send to.
1634 * For a peeled-off socket, msg_name is ignored.
1636 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1637 int msg_namelen = msg->msg_namelen;
1639 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1644 if (msg_namelen > sizeof(to))
1645 msg_namelen = sizeof(to);
1646 memcpy(&to, msg->msg_name, msg_namelen);
1647 msg_name = msg->msg_name;
1651 if (cmsgs.sinfo != NULL) {
1652 memset(&default_sinfo, 0, sizeof(default_sinfo));
1653 default_sinfo.sinfo_stream = cmsgs.sinfo->snd_sid;
1654 default_sinfo.sinfo_flags = cmsgs.sinfo->snd_flags;
1655 default_sinfo.sinfo_ppid = cmsgs.sinfo->snd_ppid;
1656 default_sinfo.sinfo_context = cmsgs.sinfo->snd_context;
1657 default_sinfo.sinfo_assoc_id = cmsgs.sinfo->snd_assoc_id;
1659 sinfo = &default_sinfo;
1660 fill_sinfo_ttl = true;
1662 sinfo = cmsgs.srinfo;
1664 /* Did the user specify SNDINFO/SNDRCVINFO? */
1666 sinfo_flags = sinfo->sinfo_flags;
1667 associd = sinfo->sinfo_assoc_id;
1670 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1671 msg_len, sinfo_flags);
1673 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1674 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1679 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1680 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1681 * If SCTP_ABORT is set, the message length could be non zero with
1682 * the msg_iov set to the user abort reason.
1684 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1685 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1690 /* If SCTP_ADDR_OVER is set, there must be an address
1691 * specified in msg_name.
1693 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1700 pr_debug("%s: about to look up association\n", __func__);
1704 /* If a msg_name has been specified, assume this is to be used. */
1706 /* Look for a matching association on the endpoint. */
1707 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1709 /* If we could not find a matching association on the
1710 * endpoint, make sure that it is not a TCP-style
1711 * socket that already has an association or there is
1712 * no peeled-off association on another socket.
1714 if ((sctp_style(sk, TCP) &&
1715 sctp_sstate(sk, ESTABLISHED)) ||
1716 sctp_endpoint_is_peeled_off(ep, &to)) {
1717 err = -EADDRNOTAVAIL;
1722 asoc = sctp_id2assoc(sk, associd);
1730 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1732 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1733 * socket that has an association in CLOSED state. This can
1734 * happen when an accepted socket has an association that is
1737 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1742 if (sinfo_flags & SCTP_EOF) {
1743 pr_debug("%s: shutting down association:%p\n",
1746 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1750 if (sinfo_flags & SCTP_ABORT) {
1752 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1758 pr_debug("%s: aborting association:%p\n",
1761 sctp_primitive_ABORT(net, asoc, chunk);
1767 /* Do we need to create the association? */
1769 pr_debug("%s: there is no association yet\n", __func__);
1771 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1776 /* Check for invalid stream against the stream counts,
1777 * either the default or the user specified stream counts.
1780 if (!sinit || !sinit->sinit_num_ostreams) {
1781 /* Check against the defaults. */
1782 if (sinfo->sinfo_stream >=
1783 sp->initmsg.sinit_num_ostreams) {
1788 /* Check against the requested. */
1789 if (sinfo->sinfo_stream >=
1790 sinit->sinit_num_ostreams) {
1798 * API 3.1.2 bind() - UDP Style Syntax
1799 * If a bind() or sctp_bindx() is not called prior to a
1800 * sendmsg() call that initiates a new association, the
1801 * system picks an ephemeral port and will choose an address
1802 * set equivalent to binding with a wildcard address.
1804 if (!ep->base.bind_addr.port) {
1805 if (sctp_autobind(sk)) {
1811 * If an unprivileged user inherits a one-to-many
1812 * style socket with open associations on a privileged
1813 * port, it MAY be permitted to accept new associations,
1814 * but it SHOULD NOT be permitted to open new
1817 if (ep->base.bind_addr.port < PROT_SOCK &&
1818 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1824 scope = sctp_scope(&to);
1825 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1831 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1837 /* If the SCTP_INIT ancillary data is specified, set all
1838 * the association init values accordingly.
1841 if (sinit->sinit_num_ostreams) {
1842 asoc->c.sinit_num_ostreams =
1843 sinit->sinit_num_ostreams;
1845 if (sinit->sinit_max_instreams) {
1846 asoc->c.sinit_max_instreams =
1847 sinit->sinit_max_instreams;
1849 if (sinit->sinit_max_attempts) {
1850 asoc->max_init_attempts
1851 = sinit->sinit_max_attempts;
1853 if (sinit->sinit_max_init_timeo) {
1854 asoc->max_init_timeo =
1855 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1859 /* Prime the peer's transport structures. */
1860 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1867 /* ASSERT: we have a valid association at this point. */
1868 pr_debug("%s: we have a valid association\n", __func__);
1871 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1872 * one with some defaults.
1874 memset(&default_sinfo, 0, sizeof(default_sinfo));
1875 default_sinfo.sinfo_stream = asoc->default_stream;
1876 default_sinfo.sinfo_flags = asoc->default_flags;
1877 default_sinfo.sinfo_ppid = asoc->default_ppid;
1878 default_sinfo.sinfo_context = asoc->default_context;
1879 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1880 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1882 sinfo = &default_sinfo;
1883 } else if (fill_sinfo_ttl) {
1884 /* In case SNDINFO was specified, we still need to fill
1885 * it with a default ttl from the assoc here.
1887 sinfo->sinfo_timetolive = asoc->default_timetolive;
1890 /* API 7.1.7, the sndbuf size per association bounds the
1891 * maximum size of data that can be sent in a single send call.
1893 if (msg_len > sk->sk_sndbuf) {
1898 if (asoc->pmtu_pending)
1899 sctp_assoc_pending_pmtu(sk, asoc);
1901 /* If fragmentation is disabled and the message length exceeds the
1902 * association fragmentation point, return EMSGSIZE. The I-D
1903 * does not specify what this error is, but this looks like
1906 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1911 /* Check for invalid stream. */
1912 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1917 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1918 if (!sctp_wspace(asoc)) {
1919 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1924 /* If an address is passed with the sendto/sendmsg call, it is used
1925 * to override the primary destination address in the TCP model, or
1926 * when SCTP_ADDR_OVER flag is set in the UDP model.
1928 if ((sctp_style(sk, TCP) && msg_name) ||
1929 (sinfo_flags & SCTP_ADDR_OVER)) {
1930 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1938 /* Auto-connect, if we aren't connected already. */
1939 if (sctp_state(asoc, CLOSED)) {
1940 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1944 wait_connect = true;
1945 pr_debug("%s: we associated primitively\n", __func__);
1948 /* Break the message into multiple chunks of maximum size. */
1949 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
1950 if (IS_ERR(datamsg)) {
1951 err = PTR_ERR(datamsg);
1955 /* Now send the (possibly) fragmented message. */
1956 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1957 /* Do accounting for the write space. */
1958 sctp_set_owner_w(chunk);
1960 chunk->transport = chunk_tp;
1963 /* Send it to the lower layers. Note: all chunks
1964 * must either fail or succeed. The lower layer
1965 * works that way today. Keep it that way or this
1968 err = sctp_primitive_SEND(net, asoc, datamsg);
1969 sctp_datamsg_put(datamsg);
1970 /* Did the lower layer accept the chunk? */
1974 pr_debug("%s: we sent primitively\n", __func__);
1978 if (unlikely(wait_connect)) {
1979 timeo = sock_sndtimeo(sk, msg_flags & MSG_DONTWAIT);
1980 sctp_wait_for_connect(asoc, &timeo);
1983 /* If we are already past ASSOCIATE, the lower
1984 * layers are responsible for association cleanup.
1990 sctp_association_free(asoc);
1995 return sctp_error(sk, msg_flags, err);
2002 err = sock_error(sk);
2012 /* This is an extended version of skb_pull() that removes the data from the
2013 * start of a skb even when data is spread across the list of skb's in the
2014 * frag_list. len specifies the total amount of data that needs to be removed.
2015 * when 'len' bytes could be removed from the skb, it returns 0.
2016 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2017 * could not be removed.
2019 static int sctp_skb_pull(struct sk_buff *skb, int len)
2021 struct sk_buff *list;
2022 int skb_len = skb_headlen(skb);
2025 if (len <= skb_len) {
2026 __skb_pull(skb, len);
2030 __skb_pull(skb, skb_len);
2032 skb_walk_frags(skb, list) {
2033 rlen = sctp_skb_pull(list, len);
2034 skb->len -= (len-rlen);
2035 skb->data_len -= (len-rlen);
2046 /* API 3.1.3 recvmsg() - UDP Style Syntax
2048 * ssize_t recvmsg(int socket, struct msghdr *message,
2051 * socket - the socket descriptor of the endpoint.
2052 * message - pointer to the msghdr structure which contains a single
2053 * user message and possibly some ancillary data.
2055 * See Section 5 for complete description of the data
2058 * flags - flags sent or received with the user message, see Section
2059 * 5 for complete description of the flags.
2061 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2062 int noblock, int flags, int *addr_len)
2064 struct sctp_ulpevent *event = NULL;
2065 struct sctp_sock *sp = sctp_sk(sk);
2066 struct sk_buff *skb;
2071 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2072 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2077 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2082 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2086 /* Get the total length of the skb including any skb's in the
2095 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2097 event = sctp_skb2event(skb);
2102 sock_recv_ts_and_drops(msg, sk, skb);
2103 if (sctp_ulpevent_is_notification(event)) {
2104 msg->msg_flags |= MSG_NOTIFICATION;
2105 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2107 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2110 /* Check if we allow SCTP_NXTINFO. */
2111 if (sp->recvnxtinfo)
2112 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2113 /* Check if we allow SCTP_RCVINFO. */
2114 if (sp->recvrcvinfo)
2115 sctp_ulpevent_read_rcvinfo(event, msg);
2116 /* Check if we allow SCTP_SNDRCVINFO. */
2117 if (sp->subscribe.sctp_data_io_event)
2118 sctp_ulpevent_read_sndrcvinfo(event, msg);
2122 /* If skb's length exceeds the user's buffer, update the skb and
2123 * push it back to the receive_queue so that the next call to
2124 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2126 if (skb_len > copied) {
2127 msg->msg_flags &= ~MSG_EOR;
2128 if (flags & MSG_PEEK)
2130 sctp_skb_pull(skb, copied);
2131 skb_queue_head(&sk->sk_receive_queue, skb);
2133 /* When only partial message is copied to the user, increase
2134 * rwnd by that amount. If all the data in the skb is read,
2135 * rwnd is updated when the event is freed.
2137 if (!sctp_ulpevent_is_notification(event))
2138 sctp_assoc_rwnd_increase(event->asoc, copied);
2140 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2141 (event->msg_flags & MSG_EOR))
2142 msg->msg_flags |= MSG_EOR;
2144 msg->msg_flags &= ~MSG_EOR;
2147 if (flags & MSG_PEEK) {
2148 /* Release the skb reference acquired after peeking the skb in
2149 * sctp_skb_recv_datagram().
2153 /* Free the event which includes releasing the reference to
2154 * the owner of the skb, freeing the skb and updating the
2157 sctp_ulpevent_free(event);
2164 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2166 * This option is a on/off flag. If enabled no SCTP message
2167 * fragmentation will be performed. Instead if a message being sent
2168 * exceeds the current PMTU size, the message will NOT be sent and
2169 * instead a error will be indicated to the user.
2171 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2172 char __user *optval,
2173 unsigned int optlen)
2177 if (optlen < sizeof(int))
2180 if (get_user(val, (int __user *)optval))
2183 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2188 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2189 unsigned int optlen)
2191 struct sctp_association *asoc;
2192 struct sctp_ulpevent *event;
2194 if (optlen > sizeof(struct sctp_event_subscribe))
2196 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2199 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2200 * if there is no data to be sent or retransmit, the stack will
2201 * immediately send up this notification.
2203 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2204 &sctp_sk(sk)->subscribe)) {
2205 asoc = sctp_id2assoc(sk, 0);
2207 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2208 event = sctp_ulpevent_make_sender_dry_event(asoc,
2213 sctp_ulpq_tail_event(&asoc->ulpq, event);
2220 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2222 * This socket option is applicable to the UDP-style socket only. When
2223 * set it will cause associations that are idle for more than the
2224 * specified number of seconds to automatically close. An association
2225 * being idle is defined an association that has NOT sent or received
2226 * user data. The special value of '0' indicates that no automatic
2227 * close of any associations should be performed. The option expects an
2228 * integer defining the number of seconds of idle time before an
2229 * association is closed.
2231 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2232 unsigned int optlen)
2234 struct sctp_sock *sp = sctp_sk(sk);
2235 struct net *net = sock_net(sk);
2237 /* Applicable to UDP-style socket only */
2238 if (sctp_style(sk, TCP))
2240 if (optlen != sizeof(int))
2242 if (copy_from_user(&sp->autoclose, optval, optlen))
2245 if (sp->autoclose > net->sctp.max_autoclose)
2246 sp->autoclose = net->sctp.max_autoclose;
2251 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2253 * Applications can enable or disable heartbeats for any peer address of
2254 * an association, modify an address's heartbeat interval, force a
2255 * heartbeat to be sent immediately, and adjust the address's maximum
2256 * number of retransmissions sent before an address is considered
2257 * unreachable. The following structure is used to access and modify an
2258 * address's parameters:
2260 * struct sctp_paddrparams {
2261 * sctp_assoc_t spp_assoc_id;
2262 * struct sockaddr_storage spp_address;
2263 * uint32_t spp_hbinterval;
2264 * uint16_t spp_pathmaxrxt;
2265 * uint32_t spp_pathmtu;
2266 * uint32_t spp_sackdelay;
2267 * uint32_t spp_flags;
2270 * spp_assoc_id - (one-to-many style socket) This is filled in the
2271 * application, and identifies the association for
2273 * spp_address - This specifies which address is of interest.
2274 * spp_hbinterval - This contains the value of the heartbeat interval,
2275 * in milliseconds. If a value of zero
2276 * is present in this field then no changes are to
2277 * be made to this parameter.
2278 * spp_pathmaxrxt - This contains the maximum number of
2279 * retransmissions before this address shall be
2280 * considered unreachable. If a value of zero
2281 * is present in this field then no changes are to
2282 * be made to this parameter.
2283 * spp_pathmtu - When Path MTU discovery is disabled the value
2284 * specified here will be the "fixed" path mtu.
2285 * Note that if the spp_address field is empty
2286 * then all associations on this address will
2287 * have this fixed path mtu set upon them.
2289 * spp_sackdelay - When delayed sack is enabled, this value specifies
2290 * the number of milliseconds that sacks will be delayed
2291 * for. This value will apply to all addresses of an
2292 * association if the spp_address field is empty. Note
2293 * also, that if delayed sack is enabled and this
2294 * value is set to 0, no change is made to the last
2295 * recorded delayed sack timer value.
2297 * spp_flags - These flags are used to control various features
2298 * on an association. The flag field may contain
2299 * zero or more of the following options.
2301 * SPP_HB_ENABLE - Enable heartbeats on the
2302 * specified address. Note that if the address
2303 * field is empty all addresses for the association
2304 * have heartbeats enabled upon them.
2306 * SPP_HB_DISABLE - Disable heartbeats on the
2307 * speicifed address. Note that if the address
2308 * field is empty all addresses for the association
2309 * will have their heartbeats disabled. Note also
2310 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2311 * mutually exclusive, only one of these two should
2312 * be specified. Enabling both fields will have
2313 * undetermined results.
2315 * SPP_HB_DEMAND - Request a user initiated heartbeat
2316 * to be made immediately.
2318 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2319 * heartbeat delayis to be set to the value of 0
2322 * SPP_PMTUD_ENABLE - This field will enable PMTU
2323 * discovery upon the specified address. Note that
2324 * if the address feild is empty then all addresses
2325 * on the association are effected.
2327 * SPP_PMTUD_DISABLE - This field will disable PMTU
2328 * discovery upon the specified address. Note that
2329 * if the address feild is empty then all addresses
2330 * on the association are effected. Not also that
2331 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2332 * exclusive. Enabling both will have undetermined
2335 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2336 * on delayed sack. The time specified in spp_sackdelay
2337 * is used to specify the sack delay for this address. Note
2338 * that if spp_address is empty then all addresses will
2339 * enable delayed sack and take on the sack delay
2340 * value specified in spp_sackdelay.
2341 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2342 * off delayed sack. If the spp_address field is blank then
2343 * delayed sack is disabled for the entire association. Note
2344 * also that this field is mutually exclusive to
2345 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2348 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2349 struct sctp_transport *trans,
2350 struct sctp_association *asoc,
2351 struct sctp_sock *sp,
2354 int sackdelay_change)
2358 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2359 struct net *net = sock_net(trans->asoc->base.sk);
2361 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2366 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2367 * this field is ignored. Note also that a value of zero indicates
2368 * the current setting should be left unchanged.
2370 if (params->spp_flags & SPP_HB_ENABLE) {
2372 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2373 * set. This lets us use 0 value when this flag
2376 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2377 params->spp_hbinterval = 0;
2379 if (params->spp_hbinterval ||
2380 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2383 msecs_to_jiffies(params->spp_hbinterval);
2386 msecs_to_jiffies(params->spp_hbinterval);
2388 sp->hbinterval = params->spp_hbinterval;
2395 trans->param_flags =
2396 (trans->param_flags & ~SPP_HB) | hb_change;
2399 (asoc->param_flags & ~SPP_HB) | hb_change;
2402 (sp->param_flags & ~SPP_HB) | hb_change;
2406 /* When Path MTU discovery is disabled the value specified here will
2407 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2408 * include the flag SPP_PMTUD_DISABLE for this field to have any
2411 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2413 trans->pathmtu = params->spp_pathmtu;
2414 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2416 asoc->pathmtu = params->spp_pathmtu;
2417 sctp_frag_point(asoc, params->spp_pathmtu);
2419 sp->pathmtu = params->spp_pathmtu;
2425 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2426 (params->spp_flags & SPP_PMTUD_ENABLE);
2427 trans->param_flags =
2428 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2430 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2431 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2435 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2438 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2442 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2443 * value of this field is ignored. Note also that a value of zero
2444 * indicates the current setting should be left unchanged.
2446 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2449 msecs_to_jiffies(params->spp_sackdelay);
2452 msecs_to_jiffies(params->spp_sackdelay);
2454 sp->sackdelay = params->spp_sackdelay;
2458 if (sackdelay_change) {
2460 trans->param_flags =
2461 (trans->param_flags & ~SPP_SACKDELAY) |
2465 (asoc->param_flags & ~SPP_SACKDELAY) |
2469 (sp->param_flags & ~SPP_SACKDELAY) |
2474 /* Note that a value of zero indicates the current setting should be
2477 if (params->spp_pathmaxrxt) {
2479 trans->pathmaxrxt = params->spp_pathmaxrxt;
2481 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2483 sp->pathmaxrxt = params->spp_pathmaxrxt;
2490 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2491 char __user *optval,
2492 unsigned int optlen)
2494 struct sctp_paddrparams params;
2495 struct sctp_transport *trans = NULL;
2496 struct sctp_association *asoc = NULL;
2497 struct sctp_sock *sp = sctp_sk(sk);
2499 int hb_change, pmtud_change, sackdelay_change;
2501 if (optlen != sizeof(struct sctp_paddrparams))
2504 if (copy_from_user(¶ms, optval, optlen))
2507 /* Validate flags and value parameters. */
2508 hb_change = params.spp_flags & SPP_HB;
2509 pmtud_change = params.spp_flags & SPP_PMTUD;
2510 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2512 if (hb_change == SPP_HB ||
2513 pmtud_change == SPP_PMTUD ||
2514 sackdelay_change == SPP_SACKDELAY ||
2515 params.spp_sackdelay > 500 ||
2516 (params.spp_pathmtu &&
2517 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2520 /* If an address other than INADDR_ANY is specified, and
2521 * no transport is found, then the request is invalid.
2523 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
2524 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2525 params.spp_assoc_id);
2530 /* Get association, if assoc_id != 0 and the socket is a one
2531 * to many style socket, and an association was not found, then
2532 * the id was invalid.
2534 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2535 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2538 /* Heartbeat demand can only be sent on a transport or
2539 * association, but not a socket.
2541 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2544 /* Process parameters. */
2545 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2546 hb_change, pmtud_change,
2552 /* If changes are for association, also apply parameters to each
2555 if (!trans && asoc) {
2556 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2558 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2559 hb_change, pmtud_change,
2567 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2569 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2572 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2574 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2578 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2580 * This option will effect the way delayed acks are performed. This
2581 * option allows you to get or set the delayed ack time, in
2582 * milliseconds. It also allows changing the delayed ack frequency.
2583 * Changing the frequency to 1 disables the delayed sack algorithm. If
2584 * the assoc_id is 0, then this sets or gets the endpoints default
2585 * values. If the assoc_id field is non-zero, then the set or get
2586 * effects the specified association for the one to many model (the
2587 * assoc_id field is ignored by the one to one model). Note that if
2588 * sack_delay or sack_freq are 0 when setting this option, then the
2589 * current values will remain unchanged.
2591 * struct sctp_sack_info {
2592 * sctp_assoc_t sack_assoc_id;
2593 * uint32_t sack_delay;
2594 * uint32_t sack_freq;
2597 * sack_assoc_id - This parameter, indicates which association the user
2598 * is performing an action upon. Note that if this field's value is
2599 * zero then the endpoints default value is changed (effecting future
2600 * associations only).
2602 * sack_delay - This parameter contains the number of milliseconds that
2603 * the user is requesting the delayed ACK timer be set to. Note that
2604 * this value is defined in the standard to be between 200 and 500
2607 * sack_freq - This parameter contains the number of packets that must
2608 * be received before a sack is sent without waiting for the delay
2609 * timer to expire. The default value for this is 2, setting this
2610 * value to 1 will disable the delayed sack algorithm.
2613 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2614 char __user *optval, unsigned int optlen)
2616 struct sctp_sack_info params;
2617 struct sctp_transport *trans = NULL;
2618 struct sctp_association *asoc = NULL;
2619 struct sctp_sock *sp = sctp_sk(sk);
2621 if (optlen == sizeof(struct sctp_sack_info)) {
2622 if (copy_from_user(¶ms, optval, optlen))
2625 if (params.sack_delay == 0 && params.sack_freq == 0)
2627 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2628 pr_warn_ratelimited(DEPRECATED
2630 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2631 "Use struct sctp_sack_info instead\n",
2632 current->comm, task_pid_nr(current));
2633 if (copy_from_user(¶ms, optval, optlen))
2636 if (params.sack_delay == 0)
2637 params.sack_freq = 1;
2639 params.sack_freq = 0;
2643 /* Validate value parameter. */
2644 if (params.sack_delay > 500)
2647 /* Get association, if sack_assoc_id != 0 and the socket is a one
2648 * to many style socket, and an association was not found, then
2649 * the id was invalid.
2651 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2652 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2655 if (params.sack_delay) {
2658 msecs_to_jiffies(params.sack_delay);
2660 sctp_spp_sackdelay_enable(asoc->param_flags);
2662 sp->sackdelay = params.sack_delay;
2664 sctp_spp_sackdelay_enable(sp->param_flags);
2668 if (params.sack_freq == 1) {
2671 sctp_spp_sackdelay_disable(asoc->param_flags);
2674 sctp_spp_sackdelay_disable(sp->param_flags);
2676 } else if (params.sack_freq > 1) {
2678 asoc->sackfreq = params.sack_freq;
2680 sctp_spp_sackdelay_enable(asoc->param_flags);
2682 sp->sackfreq = params.sack_freq;
2684 sctp_spp_sackdelay_enable(sp->param_flags);
2688 /* If change is for association, also apply to each transport. */
2690 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2692 if (params.sack_delay) {
2694 msecs_to_jiffies(params.sack_delay);
2695 trans->param_flags =
2696 sctp_spp_sackdelay_enable(trans->param_flags);
2698 if (params.sack_freq == 1) {
2699 trans->param_flags =
2700 sctp_spp_sackdelay_disable(trans->param_flags);
2701 } else if (params.sack_freq > 1) {
2702 trans->sackfreq = params.sack_freq;
2703 trans->param_flags =
2704 sctp_spp_sackdelay_enable(trans->param_flags);
2712 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2714 * Applications can specify protocol parameters for the default association
2715 * initialization. The option name argument to setsockopt() and getsockopt()
2718 * Setting initialization parameters is effective only on an unconnected
2719 * socket (for UDP-style sockets only future associations are effected
2720 * by the change). With TCP-style sockets, this option is inherited by
2721 * sockets derived from a listener socket.
2723 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2725 struct sctp_initmsg sinit;
2726 struct sctp_sock *sp = sctp_sk(sk);
2728 if (optlen != sizeof(struct sctp_initmsg))
2730 if (copy_from_user(&sinit, optval, optlen))
2733 if (sinit.sinit_num_ostreams)
2734 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2735 if (sinit.sinit_max_instreams)
2736 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2737 if (sinit.sinit_max_attempts)
2738 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2739 if (sinit.sinit_max_init_timeo)
2740 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2746 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2748 * Applications that wish to use the sendto() system call may wish to
2749 * specify a default set of parameters that would normally be supplied
2750 * through the inclusion of ancillary data. This socket option allows
2751 * such an application to set the default sctp_sndrcvinfo structure.
2752 * The application that wishes to use this socket option simply passes
2753 * in to this call the sctp_sndrcvinfo structure defined in Section
2754 * 5.2.2) The input parameters accepted by this call include
2755 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2756 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2757 * to this call if the caller is using the UDP model.
2759 static int sctp_setsockopt_default_send_param(struct sock *sk,
2760 char __user *optval,
2761 unsigned int optlen)
2763 struct sctp_sock *sp = sctp_sk(sk);
2764 struct sctp_association *asoc;
2765 struct sctp_sndrcvinfo info;
2767 if (optlen != sizeof(info))
2769 if (copy_from_user(&info, optval, optlen))
2771 if (info.sinfo_flags &
2772 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2773 SCTP_ABORT | SCTP_EOF))
2776 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2777 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2780 asoc->default_stream = info.sinfo_stream;
2781 asoc->default_flags = info.sinfo_flags;
2782 asoc->default_ppid = info.sinfo_ppid;
2783 asoc->default_context = info.sinfo_context;
2784 asoc->default_timetolive = info.sinfo_timetolive;
2786 sp->default_stream = info.sinfo_stream;
2787 sp->default_flags = info.sinfo_flags;
2788 sp->default_ppid = info.sinfo_ppid;
2789 sp->default_context = info.sinfo_context;
2790 sp->default_timetolive = info.sinfo_timetolive;
2796 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2797 * (SCTP_DEFAULT_SNDINFO)
2799 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
2800 char __user *optval,
2801 unsigned int optlen)
2803 struct sctp_sock *sp = sctp_sk(sk);
2804 struct sctp_association *asoc;
2805 struct sctp_sndinfo info;
2807 if (optlen != sizeof(info))
2809 if (copy_from_user(&info, optval, optlen))
2811 if (info.snd_flags &
2812 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2813 SCTP_ABORT | SCTP_EOF))
2816 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
2817 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
2820 asoc->default_stream = info.snd_sid;
2821 asoc->default_flags = info.snd_flags;
2822 asoc->default_ppid = info.snd_ppid;
2823 asoc->default_context = info.snd_context;
2825 sp->default_stream = info.snd_sid;
2826 sp->default_flags = info.snd_flags;
2827 sp->default_ppid = info.snd_ppid;
2828 sp->default_context = info.snd_context;
2834 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2836 * Requests that the local SCTP stack use the enclosed peer address as
2837 * the association primary. The enclosed address must be one of the
2838 * association peer's addresses.
2840 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2841 unsigned int optlen)
2843 struct sctp_prim prim;
2844 struct sctp_transport *trans;
2846 if (optlen != sizeof(struct sctp_prim))
2849 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2852 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2856 sctp_assoc_set_primary(trans->asoc, trans);
2862 * 7.1.5 SCTP_NODELAY
2864 * Turn on/off any Nagle-like algorithm. This means that packets are
2865 * generally sent as soon as possible and no unnecessary delays are
2866 * introduced, at the cost of more packets in the network. Expects an
2867 * integer boolean flag.
2869 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2870 unsigned int optlen)
2874 if (optlen < sizeof(int))
2876 if (get_user(val, (int __user *)optval))
2879 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2885 * 7.1.1 SCTP_RTOINFO
2887 * The protocol parameters used to initialize and bound retransmission
2888 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2889 * and modify these parameters.
2890 * All parameters are time values, in milliseconds. A value of 0, when
2891 * modifying the parameters, indicates that the current value should not
2895 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2897 struct sctp_rtoinfo rtoinfo;
2898 struct sctp_association *asoc;
2899 unsigned long rto_min, rto_max;
2900 struct sctp_sock *sp = sctp_sk(sk);
2902 if (optlen != sizeof (struct sctp_rtoinfo))
2905 if (copy_from_user(&rtoinfo, optval, optlen))
2908 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2910 /* Set the values to the specific association */
2911 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2914 rto_max = rtoinfo.srto_max;
2915 rto_min = rtoinfo.srto_min;
2918 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2920 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2923 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2925 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2927 if (rto_min > rto_max)
2931 if (rtoinfo.srto_initial != 0)
2933 msecs_to_jiffies(rtoinfo.srto_initial);
2934 asoc->rto_max = rto_max;
2935 asoc->rto_min = rto_min;
2937 /* If there is no association or the association-id = 0
2938 * set the values to the endpoint.
2940 if (rtoinfo.srto_initial != 0)
2941 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2942 sp->rtoinfo.srto_max = rto_max;
2943 sp->rtoinfo.srto_min = rto_min;
2951 * 7.1.2 SCTP_ASSOCINFO
2953 * This option is used to tune the maximum retransmission attempts
2954 * of the association.
2955 * Returns an error if the new association retransmission value is
2956 * greater than the sum of the retransmission value of the peer.
2957 * See [SCTP] for more information.
2960 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2963 struct sctp_assocparams assocparams;
2964 struct sctp_association *asoc;
2966 if (optlen != sizeof(struct sctp_assocparams))
2968 if (copy_from_user(&assocparams, optval, optlen))
2971 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2973 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2976 /* Set the values to the specific association */
2978 if (assocparams.sasoc_asocmaxrxt != 0) {
2981 struct sctp_transport *peer_addr;
2983 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2985 path_sum += peer_addr->pathmaxrxt;
2989 /* Only validate asocmaxrxt if we have more than
2990 * one path/transport. We do this because path
2991 * retransmissions are only counted when we have more
2995 assocparams.sasoc_asocmaxrxt > path_sum)
2998 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3001 if (assocparams.sasoc_cookie_life != 0)
3002 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3004 /* Set the values to the endpoint */
3005 struct sctp_sock *sp = sctp_sk(sk);
3007 if (assocparams.sasoc_asocmaxrxt != 0)
3008 sp->assocparams.sasoc_asocmaxrxt =
3009 assocparams.sasoc_asocmaxrxt;
3010 if (assocparams.sasoc_cookie_life != 0)
3011 sp->assocparams.sasoc_cookie_life =
3012 assocparams.sasoc_cookie_life;
3018 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3020 * This socket option is a boolean flag which turns on or off mapped V4
3021 * addresses. If this option is turned on and the socket is type
3022 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3023 * If this option is turned off, then no mapping will be done of V4
3024 * addresses and a user will receive both PF_INET6 and PF_INET type
3025 * addresses on the socket.
3027 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3030 struct sctp_sock *sp = sctp_sk(sk);
3032 if (optlen < sizeof(int))
3034 if (get_user(val, (int __user *)optval))
3045 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3046 * This option will get or set the maximum size to put in any outgoing
3047 * SCTP DATA chunk. If a message is larger than this size it will be
3048 * fragmented by SCTP into the specified size. Note that the underlying
3049 * SCTP implementation may fragment into smaller sized chunks when the
3050 * PMTU of the underlying association is smaller than the value set by
3051 * the user. The default value for this option is '0' which indicates
3052 * the user is NOT limiting fragmentation and only the PMTU will effect
3053 * SCTP's choice of DATA chunk size. Note also that values set larger
3054 * than the maximum size of an IP datagram will effectively let SCTP
3055 * control fragmentation (i.e. the same as setting this option to 0).
3057 * The following structure is used to access and modify this parameter:
3059 * struct sctp_assoc_value {
3060 * sctp_assoc_t assoc_id;
3061 * uint32_t assoc_value;
3064 * assoc_id: This parameter is ignored for one-to-one style sockets.
3065 * For one-to-many style sockets this parameter indicates which
3066 * association the user is performing an action upon. Note that if
3067 * this field's value is zero then the endpoints default value is
3068 * changed (effecting future associations only).
3069 * assoc_value: This parameter specifies the maximum size in bytes.
3071 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3073 struct sctp_assoc_value params;
3074 struct sctp_association *asoc;
3075 struct sctp_sock *sp = sctp_sk(sk);
3078 if (optlen == sizeof(int)) {
3079 pr_warn_ratelimited(DEPRECATED
3081 "Use of int in maxseg socket option.\n"
3082 "Use struct sctp_assoc_value instead\n",
3083 current->comm, task_pid_nr(current));
3084 if (copy_from_user(&val, optval, optlen))
3086 params.assoc_id = 0;
3087 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3088 if (copy_from_user(¶ms, optval, optlen))
3090 val = params.assoc_value;
3094 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3097 asoc = sctp_id2assoc(sk, params.assoc_id);
3098 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3103 val = asoc->pathmtu;
3104 val -= sp->pf->af->net_header_len;
3105 val -= sizeof(struct sctphdr) +
3106 sizeof(struct sctp_data_chunk);
3108 asoc->user_frag = val;
3109 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3111 sp->user_frag = val;
3119 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3121 * Requests that the peer mark the enclosed address as the association
3122 * primary. The enclosed address must be one of the association's
3123 * locally bound addresses. The following structure is used to make a
3124 * set primary request:
3126 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3127 unsigned int optlen)
3129 struct net *net = sock_net(sk);
3130 struct sctp_sock *sp;
3131 struct sctp_association *asoc = NULL;
3132 struct sctp_setpeerprim prim;
3133 struct sctp_chunk *chunk;
3139 if (!net->sctp.addip_enable)
3142 if (optlen != sizeof(struct sctp_setpeerprim))
3145 if (copy_from_user(&prim, optval, optlen))
3148 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3152 if (!asoc->peer.asconf_capable)
3155 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3158 if (!sctp_state(asoc, ESTABLISHED))
3161 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3165 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3166 return -EADDRNOTAVAIL;
3168 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3169 return -EADDRNOTAVAIL;
3171 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3172 chunk = sctp_make_asconf_set_prim(asoc,
3173 (union sctp_addr *)&prim.sspp_addr);
3177 err = sctp_send_asconf(asoc, chunk);
3179 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3184 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3185 unsigned int optlen)
3187 struct sctp_setadaptation adaptation;
3189 if (optlen != sizeof(struct sctp_setadaptation))
3191 if (copy_from_user(&adaptation, optval, optlen))
3194 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3200 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3202 * The context field in the sctp_sndrcvinfo structure is normally only
3203 * used when a failed message is retrieved holding the value that was
3204 * sent down on the actual send call. This option allows the setting of
3205 * a default context on an association basis that will be received on
3206 * reading messages from the peer. This is especially helpful in the
3207 * one-2-many model for an application to keep some reference to an
3208 * internal state machine that is processing messages on the
3209 * association. Note that the setting of this value only effects
3210 * received messages from the peer and does not effect the value that is
3211 * saved with outbound messages.
3213 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3214 unsigned int optlen)
3216 struct sctp_assoc_value params;
3217 struct sctp_sock *sp;
3218 struct sctp_association *asoc;
3220 if (optlen != sizeof(struct sctp_assoc_value))
3222 if (copy_from_user(¶ms, optval, optlen))
3227 if (params.assoc_id != 0) {
3228 asoc = sctp_id2assoc(sk, params.assoc_id);
3231 asoc->default_rcv_context = params.assoc_value;
3233 sp->default_rcv_context = params.assoc_value;
3240 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3242 * This options will at a minimum specify if the implementation is doing
3243 * fragmented interleave. Fragmented interleave, for a one to many
3244 * socket, is when subsequent calls to receive a message may return
3245 * parts of messages from different associations. Some implementations
3246 * may allow you to turn this value on or off. If so, when turned off,
3247 * no fragment interleave will occur (which will cause a head of line
3248 * blocking amongst multiple associations sharing the same one to many
3249 * socket). When this option is turned on, then each receive call may
3250 * come from a different association (thus the user must receive data
3251 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3252 * association each receive belongs to.
3254 * This option takes a boolean value. A non-zero value indicates that
3255 * fragmented interleave is on. A value of zero indicates that
3256 * fragmented interleave is off.
3258 * Note that it is important that an implementation that allows this
3259 * option to be turned on, have it off by default. Otherwise an unaware
3260 * application using the one to many model may become confused and act
3263 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3264 char __user *optval,
3265 unsigned int optlen)
3269 if (optlen != sizeof(int))
3271 if (get_user(val, (int __user *)optval))
3274 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3280 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3281 * (SCTP_PARTIAL_DELIVERY_POINT)
3283 * This option will set or get the SCTP partial delivery point. This
3284 * point is the size of a message where the partial delivery API will be
3285 * invoked to help free up rwnd space for the peer. Setting this to a
3286 * lower value will cause partial deliveries to happen more often. The
3287 * calls argument is an integer that sets or gets the partial delivery
3288 * point. Note also that the call will fail if the user attempts to set
3289 * this value larger than the socket receive buffer size.
3291 * Note that any single message having a length smaller than or equal to
3292 * the SCTP partial delivery point will be delivered in one single read
3293 * call as long as the user provided buffer is large enough to hold the
3296 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3297 char __user *optval,
3298 unsigned int optlen)
3302 if (optlen != sizeof(u32))
3304 if (get_user(val, (int __user *)optval))
3307 /* Note: We double the receive buffer from what the user sets
3308 * it to be, also initial rwnd is based on rcvbuf/2.
3310 if (val > (sk->sk_rcvbuf >> 1))
3313 sctp_sk(sk)->pd_point = val;
3315 return 0; /* is this the right error code? */
3319 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3321 * This option will allow a user to change the maximum burst of packets
3322 * that can be emitted by this association. Note that the default value
3323 * is 4, and some implementations may restrict this setting so that it
3324 * can only be lowered.
3326 * NOTE: This text doesn't seem right. Do this on a socket basis with
3327 * future associations inheriting the socket value.
3329 static int sctp_setsockopt_maxburst(struct sock *sk,
3330 char __user *optval,
3331 unsigned int optlen)
3333 struct sctp_assoc_value params;
3334 struct sctp_sock *sp;
3335 struct sctp_association *asoc;
3339 if (optlen == sizeof(int)) {
3340 pr_warn_ratelimited(DEPRECATED
3342 "Use of int in max_burst socket option deprecated.\n"
3343 "Use struct sctp_assoc_value instead\n",
3344 current->comm, task_pid_nr(current));
3345 if (copy_from_user(&val, optval, optlen))
3347 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3348 if (copy_from_user(¶ms, optval, optlen))
3350 val = params.assoc_value;
3351 assoc_id = params.assoc_id;
3357 if (assoc_id != 0) {
3358 asoc = sctp_id2assoc(sk, assoc_id);
3361 asoc->max_burst = val;
3363 sp->max_burst = val;
3369 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3371 * This set option adds a chunk type that the user is requesting to be
3372 * received only in an authenticated way. Changes to the list of chunks
3373 * will only effect future associations on the socket.
3375 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3376 char __user *optval,
3377 unsigned int optlen)
3379 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3380 struct sctp_authchunk val;
3382 if (!ep->auth_enable)
3385 if (optlen != sizeof(struct sctp_authchunk))
3387 if (copy_from_user(&val, optval, optlen))
3390 switch (val.sauth_chunk) {
3392 case SCTP_CID_INIT_ACK:
3393 case SCTP_CID_SHUTDOWN_COMPLETE:
3398 /* add this chunk id to the endpoint */
3399 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3403 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3405 * This option gets or sets the list of HMAC algorithms that the local
3406 * endpoint requires the peer to use.
3408 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3409 char __user *optval,
3410 unsigned int optlen)
3412 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3413 struct sctp_hmacalgo *hmacs;
3417 if (!ep->auth_enable)
3420 if (optlen < sizeof(struct sctp_hmacalgo))
3423 hmacs = memdup_user(optval, optlen);
3425 return PTR_ERR(hmacs);
3427 idents = hmacs->shmac_num_idents;
3428 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3429 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3434 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3441 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3443 * This option will set a shared secret key which is used to build an
3444 * association shared key.
3446 static int sctp_setsockopt_auth_key(struct sock *sk,
3447 char __user *optval,
3448 unsigned int optlen)
3450 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3451 struct sctp_authkey *authkey;
3452 struct sctp_association *asoc;
3455 if (!ep->auth_enable)
3458 if (optlen <= sizeof(struct sctp_authkey))
3461 authkey = memdup_user(optval, optlen);
3462 if (IS_ERR(authkey))
3463 return PTR_ERR(authkey);
3465 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3470 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3471 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3476 ret = sctp_auth_set_key(ep, asoc, authkey);
3483 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3485 * This option will get or set the active shared key to be used to build
3486 * the association shared key.
3488 static int sctp_setsockopt_active_key(struct sock *sk,
3489 char __user *optval,
3490 unsigned int optlen)
3492 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3493 struct sctp_authkeyid val;
3494 struct sctp_association *asoc;
3496 if (!ep->auth_enable)
3499 if (optlen != sizeof(struct sctp_authkeyid))
3501 if (copy_from_user(&val, optval, optlen))
3504 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3505 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3508 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3512 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3514 * This set option will delete a shared secret key from use.
3516 static int sctp_setsockopt_del_key(struct sock *sk,
3517 char __user *optval,
3518 unsigned int optlen)
3520 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3521 struct sctp_authkeyid val;
3522 struct sctp_association *asoc;
3524 if (!ep->auth_enable)
3527 if (optlen != sizeof(struct sctp_authkeyid))
3529 if (copy_from_user(&val, optval, optlen))
3532 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3533 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3536 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3541 * 8.1.23 SCTP_AUTO_ASCONF
3543 * This option will enable or disable the use of the automatic generation of
3544 * ASCONF chunks to add and delete addresses to an existing association. Note
3545 * that this option has two caveats namely: a) it only affects sockets that
3546 * are bound to all addresses available to the SCTP stack, and b) the system
3547 * administrator may have an overriding control that turns the ASCONF feature
3548 * off no matter what setting the socket option may have.
3549 * This option expects an integer boolean flag, where a non-zero value turns on
3550 * the option, and a zero value turns off the option.
3551 * Note. In this implementation, socket operation overrides default parameter
3552 * being set by sysctl as well as FreeBSD implementation
3554 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3555 unsigned int optlen)
3558 struct sctp_sock *sp = sctp_sk(sk);
3560 if (optlen < sizeof(int))
3562 if (get_user(val, (int __user *)optval))
3564 if (!sctp_is_ep_boundall(sk) && val)
3566 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3569 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3570 if (val == 0 && sp->do_auto_asconf) {
3571 list_del(&sp->auto_asconf_list);
3572 sp->do_auto_asconf = 0;
3573 } else if (val && !sp->do_auto_asconf) {
3574 list_add_tail(&sp->auto_asconf_list,
3575 &sock_net(sk)->sctp.auto_asconf_splist);
3576 sp->do_auto_asconf = 1;
3578 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3583 * SCTP_PEER_ADDR_THLDS
3585 * This option allows us to alter the partially failed threshold for one or all
3586 * transports in an association. See Section 6.1 of:
3587 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3589 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3590 char __user *optval,
3591 unsigned int optlen)
3593 struct sctp_paddrthlds val;
3594 struct sctp_transport *trans;
3595 struct sctp_association *asoc;
3597 if (optlen < sizeof(struct sctp_paddrthlds))
3599 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3600 sizeof(struct sctp_paddrthlds)))
3604 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3605 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3608 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3610 if (val.spt_pathmaxrxt)
3611 trans->pathmaxrxt = val.spt_pathmaxrxt;
3612 trans->pf_retrans = val.spt_pathpfthld;
3615 if (val.spt_pathmaxrxt)
3616 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3617 asoc->pf_retrans = val.spt_pathpfthld;
3619 trans = sctp_addr_id2transport(sk, &val.spt_address,
3624 if (val.spt_pathmaxrxt)
3625 trans->pathmaxrxt = val.spt_pathmaxrxt;
3626 trans->pf_retrans = val.spt_pathpfthld;
3632 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3633 char __user *optval,
3634 unsigned int optlen)
3638 if (optlen < sizeof(int))
3640 if (get_user(val, (int __user *) optval))
3643 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3648 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3649 char __user *optval,
3650 unsigned int optlen)
3654 if (optlen < sizeof(int))
3656 if (get_user(val, (int __user *) optval))
3659 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3664 /* API 6.2 setsockopt(), getsockopt()
3666 * Applications use setsockopt() and getsockopt() to set or retrieve
3667 * socket options. Socket options are used to change the default
3668 * behavior of sockets calls. They are described in Section 7.
3672 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3673 * int __user *optlen);
3674 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3677 * sd - the socket descript.
3678 * level - set to IPPROTO_SCTP for all SCTP options.
3679 * optname - the option name.
3680 * optval - the buffer to store the value of the option.
3681 * optlen - the size of the buffer.
3683 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3684 char __user *optval, unsigned int optlen)
3688 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3690 /* I can hardly begin to describe how wrong this is. This is
3691 * so broken as to be worse than useless. The API draft
3692 * REALLY is NOT helpful here... I am not convinced that the
3693 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3694 * are at all well-founded.
3696 if (level != SOL_SCTP) {
3697 struct sctp_af *af = sctp_sk(sk)->pf->af;
3698 retval = af->setsockopt(sk, level, optname, optval, optlen);
3705 case SCTP_SOCKOPT_BINDX_ADD:
3706 /* 'optlen' is the size of the addresses buffer. */
3707 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3708 optlen, SCTP_BINDX_ADD_ADDR);
3711 case SCTP_SOCKOPT_BINDX_REM:
3712 /* 'optlen' is the size of the addresses buffer. */
3713 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3714 optlen, SCTP_BINDX_REM_ADDR);
3717 case SCTP_SOCKOPT_CONNECTX_OLD:
3718 /* 'optlen' is the size of the addresses buffer. */
3719 retval = sctp_setsockopt_connectx_old(sk,
3720 (struct sockaddr __user *)optval,
3724 case SCTP_SOCKOPT_CONNECTX:
3725 /* 'optlen' is the size of the addresses buffer. */
3726 retval = sctp_setsockopt_connectx(sk,
3727 (struct sockaddr __user *)optval,
3731 case SCTP_DISABLE_FRAGMENTS:
3732 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3736 retval = sctp_setsockopt_events(sk, optval, optlen);
3739 case SCTP_AUTOCLOSE:
3740 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3743 case SCTP_PEER_ADDR_PARAMS:
3744 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3747 case SCTP_DELAYED_SACK:
3748 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3750 case SCTP_PARTIAL_DELIVERY_POINT:
3751 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3755 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3757 case SCTP_DEFAULT_SEND_PARAM:
3758 retval = sctp_setsockopt_default_send_param(sk, optval,
3761 case SCTP_DEFAULT_SNDINFO:
3762 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
3764 case SCTP_PRIMARY_ADDR:
3765 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3767 case SCTP_SET_PEER_PRIMARY_ADDR:
3768 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3771 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3774 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3776 case SCTP_ASSOCINFO:
3777 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3779 case SCTP_I_WANT_MAPPED_V4_ADDR:
3780 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3783 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3785 case SCTP_ADAPTATION_LAYER:
3786 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3789 retval = sctp_setsockopt_context(sk, optval, optlen);
3791 case SCTP_FRAGMENT_INTERLEAVE:
3792 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3794 case SCTP_MAX_BURST:
3795 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3797 case SCTP_AUTH_CHUNK:
3798 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3800 case SCTP_HMAC_IDENT:
3801 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3804 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3806 case SCTP_AUTH_ACTIVE_KEY:
3807 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3809 case SCTP_AUTH_DELETE_KEY:
3810 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3812 case SCTP_AUTO_ASCONF:
3813 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3815 case SCTP_PEER_ADDR_THLDS:
3816 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3818 case SCTP_RECVRCVINFO:
3819 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
3821 case SCTP_RECVNXTINFO:
3822 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
3825 retval = -ENOPROTOOPT;
3835 /* API 3.1.6 connect() - UDP Style Syntax
3837 * An application may use the connect() call in the UDP model to initiate an
3838 * association without sending data.
3842 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3844 * sd: the socket descriptor to have a new association added to.
3846 * nam: the address structure (either struct sockaddr_in or struct
3847 * sockaddr_in6 defined in RFC2553 [7]).
3849 * len: the size of the address.
3851 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3859 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
3862 /* Validate addr_len before calling common connect/connectx routine. */
3863 af = sctp_get_af_specific(addr->sa_family);
3864 if (!af || addr_len < af->sockaddr_len) {
3867 /* Pass correct addr len to common routine (so it knows there
3868 * is only one address being passed.
3870 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3877 /* FIXME: Write comments. */
3878 static int sctp_disconnect(struct sock *sk, int flags)
3880 return -EOPNOTSUPP; /* STUB */
3883 /* 4.1.4 accept() - TCP Style Syntax
3885 * Applications use accept() call to remove an established SCTP
3886 * association from the accept queue of the endpoint. A new socket
3887 * descriptor will be returned from accept() to represent the newly
3888 * formed association.
3890 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3892 struct sctp_sock *sp;
3893 struct sctp_endpoint *ep;
3894 struct sock *newsk = NULL;
3895 struct sctp_association *asoc;
3904 if (!sctp_style(sk, TCP)) {
3905 error = -EOPNOTSUPP;
3909 if (!sctp_sstate(sk, LISTENING)) {
3914 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3916 error = sctp_wait_for_accept(sk, timeo);
3920 /* We treat the list of associations on the endpoint as the accept
3921 * queue and pick the first association on the list.
3923 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3925 newsk = sp->pf->create_accept_sk(sk, asoc);
3931 /* Populate the fields of the newsk from the oldsk and migrate the
3932 * asoc to the newsk.
3934 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3942 /* The SCTP ioctl handler. */
3943 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3950 * SEQPACKET-style sockets in LISTENING state are valid, for
3951 * SCTP, so only discard TCP-style sockets in LISTENING state.
3953 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3958 struct sk_buff *skb;
3959 unsigned int amount = 0;
3961 skb = skb_peek(&sk->sk_receive_queue);
3964 * We will only return the amount of this packet since
3965 * that is all that will be read.
3969 rc = put_user(amount, (int __user *)arg);
3981 /* This is the function which gets called during socket creation to
3982 * initialized the SCTP-specific portion of the sock.
3983 * The sock structure should already be zero-filled memory.
3985 static int sctp_init_sock(struct sock *sk)
3987 struct net *net = sock_net(sk);
3988 struct sctp_sock *sp;
3990 pr_debug("%s: sk:%p\n", __func__, sk);
3994 /* Initialize the SCTP per socket area. */
3995 switch (sk->sk_type) {
3996 case SOCK_SEQPACKET:
3997 sp->type = SCTP_SOCKET_UDP;
4000 sp->type = SCTP_SOCKET_TCP;
4003 return -ESOCKTNOSUPPORT;
4006 sk->sk_gso_type = SKB_GSO_SCTP;
4008 /* Initialize default send parameters. These parameters can be
4009 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4011 sp->default_stream = 0;
4012 sp->default_ppid = 0;
4013 sp->default_flags = 0;
4014 sp->default_context = 0;
4015 sp->default_timetolive = 0;
4017 sp->default_rcv_context = 0;
4018 sp->max_burst = net->sctp.max_burst;
4020 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4022 /* Initialize default setup parameters. These parameters
4023 * can be modified with the SCTP_INITMSG socket option or
4024 * overridden by the SCTP_INIT CMSG.
4026 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4027 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4028 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4029 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4031 /* Initialize default RTO related parameters. These parameters can
4032 * be modified for with the SCTP_RTOINFO socket option.
4034 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4035 sp->rtoinfo.srto_max = net->sctp.rto_max;
4036 sp->rtoinfo.srto_min = net->sctp.rto_min;
4038 /* Initialize default association related parameters. These parameters
4039 * can be modified with the SCTP_ASSOCINFO socket option.
4041 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4042 sp->assocparams.sasoc_number_peer_destinations = 0;
4043 sp->assocparams.sasoc_peer_rwnd = 0;
4044 sp->assocparams.sasoc_local_rwnd = 0;
4045 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4047 /* Initialize default event subscriptions. By default, all the
4050 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4052 /* Default Peer Address Parameters. These defaults can
4053 * be modified via SCTP_PEER_ADDR_PARAMS
4055 sp->hbinterval = net->sctp.hb_interval;
4056 sp->pathmaxrxt = net->sctp.max_retrans_path;
4057 sp->pathmtu = 0; /* allow default discovery */
4058 sp->sackdelay = net->sctp.sack_timeout;
4060 sp->param_flags = SPP_HB_ENABLE |
4062 SPP_SACKDELAY_ENABLE;
4064 /* If enabled no SCTP message fragmentation will be performed.
4065 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4067 sp->disable_fragments = 0;
4069 /* Enable Nagle algorithm by default. */
4072 sp->recvrcvinfo = 0;
4073 sp->recvnxtinfo = 0;
4075 /* Enable by default. */
4078 /* Auto-close idle associations after the configured
4079 * number of seconds. A value of 0 disables this
4080 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4081 * for UDP-style sockets only.
4085 /* User specified fragmentation limit. */
4088 sp->adaptation_ind = 0;
4090 sp->pf = sctp_get_pf_specific(sk->sk_family);
4092 /* Control variables for partial data delivery. */
4093 atomic_set(&sp->pd_mode, 0);
4094 skb_queue_head_init(&sp->pd_lobby);
4095 sp->frag_interleave = 0;
4097 /* Create a per socket endpoint structure. Even if we
4098 * change the data structure relationships, this may still
4099 * be useful for storing pre-connect address information.
4101 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4107 sk->sk_destruct = sctp_destruct_sock;
4109 SCTP_DBG_OBJCNT_INC(sock);
4112 percpu_counter_inc(&sctp_sockets_allocated);
4113 sock_prot_inuse_add(net, sk->sk_prot, 1);
4115 /* Nothing can fail after this block, otherwise
4116 * sctp_destroy_sock() will be called without addr_wq_lock held
4118 if (net->sctp.default_auto_asconf) {
4119 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
4120 list_add_tail(&sp->auto_asconf_list,
4121 &net->sctp.auto_asconf_splist);
4122 sp->do_auto_asconf = 1;
4123 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4125 sp->do_auto_asconf = 0;
4133 /* Cleanup any SCTP per socket resources. Must be called with
4134 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4136 static void sctp_destroy_sock(struct sock *sk)
4138 struct sctp_sock *sp;
4140 pr_debug("%s: sk:%p\n", __func__, sk);
4142 /* Release our hold on the endpoint. */
4144 /* This could happen during socket init, thus we bail out
4145 * early, since the rest of the below is not setup either.
4150 if (sp->do_auto_asconf) {
4151 sp->do_auto_asconf = 0;
4152 list_del(&sp->auto_asconf_list);
4154 sctp_endpoint_free(sp->ep);
4156 percpu_counter_dec(&sctp_sockets_allocated);
4157 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4161 /* Triggered when there are no references on the socket anymore */
4162 static void sctp_destruct_sock(struct sock *sk)
4164 struct sctp_sock *sp = sctp_sk(sk);
4166 /* Free up the HMAC transform. */
4167 crypto_free_shash(sp->hmac);
4169 inet_sock_destruct(sk);
4172 /* API 4.1.7 shutdown() - TCP Style Syntax
4173 * int shutdown(int socket, int how);
4175 * sd - the socket descriptor of the association to be closed.
4176 * how - Specifies the type of shutdown. The values are
4179 * Disables further receive operations. No SCTP
4180 * protocol action is taken.
4182 * Disables further send operations, and initiates
4183 * the SCTP shutdown sequence.
4185 * Disables further send and receive operations
4186 * and initiates the SCTP shutdown sequence.
4188 static void sctp_shutdown(struct sock *sk, int how)
4190 struct net *net = sock_net(sk);
4191 struct sctp_endpoint *ep;
4192 struct sctp_association *asoc;
4194 if (!sctp_style(sk, TCP))
4197 if (how & SEND_SHUTDOWN) {
4198 ep = sctp_sk(sk)->ep;
4199 if (!list_empty(&ep->asocs)) {
4200 asoc = list_entry(ep->asocs.next,
4201 struct sctp_association, asocs);
4202 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4207 int sctp_get_sctp_info(struct sock *sk, struct sctp_association *asoc,
4208 struct sctp_info *info)
4210 struct sctp_transport *prim;
4211 struct list_head *pos;
4214 memset(info, 0, sizeof(*info));
4216 struct sctp_sock *sp = sctp_sk(sk);
4218 info->sctpi_s_autoclose = sp->autoclose;
4219 info->sctpi_s_adaptation_ind = sp->adaptation_ind;
4220 info->sctpi_s_pd_point = sp->pd_point;
4221 info->sctpi_s_nodelay = sp->nodelay;
4222 info->sctpi_s_disable_fragments = sp->disable_fragments;
4223 info->sctpi_s_v4mapped = sp->v4mapped;
4224 info->sctpi_s_frag_interleave = sp->frag_interleave;
4225 info->sctpi_s_type = sp->type;
4230 info->sctpi_tag = asoc->c.my_vtag;
4231 info->sctpi_state = asoc->state;
4232 info->sctpi_rwnd = asoc->a_rwnd;
4233 info->sctpi_unackdata = asoc->unack_data;
4234 info->sctpi_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4235 info->sctpi_instrms = asoc->c.sinit_max_instreams;
4236 info->sctpi_outstrms = asoc->c.sinit_num_ostreams;
4237 list_for_each(pos, &asoc->base.inqueue.in_chunk_list)
4238 info->sctpi_inqueue++;
4239 list_for_each(pos, &asoc->outqueue.out_chunk_list)
4240 info->sctpi_outqueue++;
4241 info->sctpi_overall_error = asoc->overall_error_count;
4242 info->sctpi_max_burst = asoc->max_burst;
4243 info->sctpi_maxseg = asoc->frag_point;
4244 info->sctpi_peer_rwnd = asoc->peer.rwnd;
4245 info->sctpi_peer_tag = asoc->c.peer_vtag;
4247 mask = asoc->peer.ecn_capable << 1;
4248 mask = (mask | asoc->peer.ipv4_address) << 1;
4249 mask = (mask | asoc->peer.ipv6_address) << 1;
4250 mask = (mask | asoc->peer.hostname_address) << 1;
4251 mask = (mask | asoc->peer.asconf_capable) << 1;
4252 mask = (mask | asoc->peer.prsctp_capable) << 1;
4253 mask = (mask | asoc->peer.auth_capable);
4254 info->sctpi_peer_capable = mask;
4255 mask = asoc->peer.sack_needed << 1;
4256 mask = (mask | asoc->peer.sack_generation) << 1;
4257 mask = (mask | asoc->peer.zero_window_announced);
4258 info->sctpi_peer_sack = mask;
4260 info->sctpi_isacks = asoc->stats.isacks;
4261 info->sctpi_osacks = asoc->stats.osacks;
4262 info->sctpi_opackets = asoc->stats.opackets;
4263 info->sctpi_ipackets = asoc->stats.ipackets;
4264 info->sctpi_rtxchunks = asoc->stats.rtxchunks;
4265 info->sctpi_outofseqtsns = asoc->stats.outofseqtsns;
4266 info->sctpi_idupchunks = asoc->stats.idupchunks;
4267 info->sctpi_gapcnt = asoc->stats.gapcnt;
4268 info->sctpi_ouodchunks = asoc->stats.ouodchunks;
4269 info->sctpi_iuodchunks = asoc->stats.iuodchunks;
4270 info->sctpi_oodchunks = asoc->stats.oodchunks;
4271 info->sctpi_iodchunks = asoc->stats.iodchunks;
4272 info->sctpi_octrlchunks = asoc->stats.octrlchunks;
4273 info->sctpi_ictrlchunks = asoc->stats.ictrlchunks;
4275 prim = asoc->peer.primary_path;
4276 memcpy(&info->sctpi_p_address, &prim->ipaddr,
4277 sizeof(struct sockaddr_storage));
4278 info->sctpi_p_state = prim->state;
4279 info->sctpi_p_cwnd = prim->cwnd;
4280 info->sctpi_p_srtt = prim->srtt;
4281 info->sctpi_p_rto = jiffies_to_msecs(prim->rto);
4282 info->sctpi_p_hbinterval = prim->hbinterval;
4283 info->sctpi_p_pathmaxrxt = prim->pathmaxrxt;
4284 info->sctpi_p_sackdelay = jiffies_to_msecs(prim->sackdelay);
4285 info->sctpi_p_ssthresh = prim->ssthresh;
4286 info->sctpi_p_partial_bytes_acked = prim->partial_bytes_acked;
4287 info->sctpi_p_flight_size = prim->flight_size;
4288 info->sctpi_p_error = prim->error_count;
4292 EXPORT_SYMBOL_GPL(sctp_get_sctp_info);
4294 /* use callback to avoid exporting the core structure */
4295 int sctp_transport_walk_start(struct rhashtable_iter *iter)
4299 err = rhashtable_walk_init(&sctp_transport_hashtable, iter,
4304 err = rhashtable_walk_start(iter);
4305 if (err && err != -EAGAIN) {
4306 rhashtable_walk_exit(iter);
4313 void sctp_transport_walk_stop(struct rhashtable_iter *iter)
4315 rhashtable_walk_stop(iter);
4316 rhashtable_walk_exit(iter);
4319 struct sctp_transport *sctp_transport_get_next(struct net *net,
4320 struct rhashtable_iter *iter)
4322 struct sctp_transport *t;
4324 t = rhashtable_walk_next(iter);
4325 for (; t; t = rhashtable_walk_next(iter)) {
4327 if (PTR_ERR(t) == -EAGAIN)
4332 if (net_eq(sock_net(t->asoc->base.sk), net) &&
4333 t->asoc->peer.primary_path == t)
4340 struct sctp_transport *sctp_transport_get_idx(struct net *net,
4341 struct rhashtable_iter *iter,
4344 void *obj = SEQ_START_TOKEN;
4346 while (pos && (obj = sctp_transport_get_next(net, iter)) &&
4353 int sctp_for_each_endpoint(int (*cb)(struct sctp_endpoint *, void *),
4357 struct sctp_ep_common *epb;
4358 struct sctp_hashbucket *head;
4360 for (head = sctp_ep_hashtable; hash < sctp_ep_hashsize;
4362 read_lock(&head->lock);
4363 sctp_for_each_hentry(epb, &head->chain) {
4364 err = cb(sctp_ep(epb), p);
4368 read_unlock(&head->lock);
4373 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint);
4375 int sctp_transport_lookup_process(int (*cb)(struct sctp_transport *, void *),
4377 const union sctp_addr *laddr,
4378 const union sctp_addr *paddr, void *p)
4380 struct sctp_transport *transport;
4384 transport = sctp_addrs_lookup_transport(net, laddr, paddr);
4385 if (!transport || !sctp_transport_hold(transport))
4387 err = cb(transport, p);
4388 sctp_transport_put(transport);
4394 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process);
4396 int sctp_for_each_transport(int (*cb)(struct sctp_transport *, void *),
4397 struct net *net, int pos, void *p) {
4398 struct rhashtable_iter hti;
4402 err = sctp_transport_walk_start(&hti);
4406 sctp_transport_get_idx(net, &hti, pos);
4407 obj = sctp_transport_get_next(net, &hti);
4408 for (; obj && !IS_ERR(obj); obj = sctp_transport_get_next(net, &hti)) {
4409 struct sctp_transport *transport = obj;
4411 if (!sctp_transport_hold(transport))
4413 err = cb(transport, p);
4414 sctp_transport_put(transport);
4418 sctp_transport_walk_stop(&hti);
4422 EXPORT_SYMBOL_GPL(sctp_for_each_transport);
4424 /* 7.2.1 Association Status (SCTP_STATUS)
4426 * Applications can retrieve current status information about an
4427 * association, including association state, peer receiver window size,
4428 * number of unacked data chunks, and number of data chunks pending
4429 * receipt. This information is read-only.
4431 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4432 char __user *optval,
4435 struct sctp_status status;
4436 struct sctp_association *asoc = NULL;
4437 struct sctp_transport *transport;
4438 sctp_assoc_t associd;
4441 if (len < sizeof(status)) {
4446 len = sizeof(status);
4447 if (copy_from_user(&status, optval, len)) {
4452 associd = status.sstat_assoc_id;
4453 asoc = sctp_id2assoc(sk, associd);
4459 transport = asoc->peer.primary_path;
4461 status.sstat_assoc_id = sctp_assoc2id(asoc);
4462 status.sstat_state = sctp_assoc_to_state(asoc);
4463 status.sstat_rwnd = asoc->peer.rwnd;
4464 status.sstat_unackdata = asoc->unack_data;
4466 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4467 status.sstat_instrms = asoc->c.sinit_max_instreams;
4468 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4469 status.sstat_fragmentation_point = asoc->frag_point;
4470 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4471 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4472 transport->af_specific->sockaddr_len);
4473 /* Map ipv4 address into v4-mapped-on-v6 address. */
4474 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
4475 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4476 status.sstat_primary.spinfo_state = transport->state;
4477 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4478 status.sstat_primary.spinfo_srtt = transport->srtt;
4479 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4480 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4482 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4483 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4485 if (put_user(len, optlen)) {
4490 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4491 __func__, len, status.sstat_state, status.sstat_rwnd,
4492 status.sstat_assoc_id);
4494 if (copy_to_user(optval, &status, len)) {
4504 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4506 * Applications can retrieve information about a specific peer address
4507 * of an association, including its reachability state, congestion
4508 * window, and retransmission timer values. This information is
4511 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4512 char __user *optval,
4515 struct sctp_paddrinfo pinfo;
4516 struct sctp_transport *transport;
4519 if (len < sizeof(pinfo)) {
4524 len = sizeof(pinfo);
4525 if (copy_from_user(&pinfo, optval, len)) {
4530 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4531 pinfo.spinfo_assoc_id);
4535 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4536 pinfo.spinfo_state = transport->state;
4537 pinfo.spinfo_cwnd = transport->cwnd;
4538 pinfo.spinfo_srtt = transport->srtt;
4539 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4540 pinfo.spinfo_mtu = transport->pathmtu;
4542 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4543 pinfo.spinfo_state = SCTP_ACTIVE;
4545 if (put_user(len, optlen)) {
4550 if (copy_to_user(optval, &pinfo, len)) {
4559 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4561 * This option is a on/off flag. If enabled no SCTP message
4562 * fragmentation will be performed. Instead if a message being sent
4563 * exceeds the current PMTU size, the message will NOT be sent and
4564 * instead a error will be indicated to the user.
4566 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4567 char __user *optval, int __user *optlen)
4571 if (len < sizeof(int))
4575 val = (sctp_sk(sk)->disable_fragments == 1);
4576 if (put_user(len, optlen))
4578 if (copy_to_user(optval, &val, len))
4583 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4585 * This socket option is used to specify various notifications and
4586 * ancillary data the user wishes to receive.
4588 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4593 if (len > sizeof(struct sctp_event_subscribe))
4594 len = sizeof(struct sctp_event_subscribe);
4595 if (put_user(len, optlen))
4597 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4602 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4604 * This socket option is applicable to the UDP-style socket only. When
4605 * set it will cause associations that are idle for more than the
4606 * specified number of seconds to automatically close. An association
4607 * being idle is defined an association that has NOT sent or received
4608 * user data. The special value of '0' indicates that no automatic
4609 * close of any associations should be performed. The option expects an
4610 * integer defining the number of seconds of idle time before an
4611 * association is closed.
4613 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4615 /* Applicable to UDP-style socket only */
4616 if (sctp_style(sk, TCP))
4618 if (len < sizeof(int))
4621 if (put_user(len, optlen))
4623 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4628 /* Helper routine to branch off an association to a new socket. */
4629 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4631 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4632 struct sctp_sock *sp = sctp_sk(sk);
4633 struct socket *sock;
4639 /* An association cannot be branched off from an already peeled-off
4640 * socket, nor is this supported for tcp style sockets.
4642 if (!sctp_style(sk, UDP))
4645 /* Create a new socket. */
4646 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4650 sctp_copy_sock(sock->sk, sk, asoc);
4652 /* Make peeled-off sockets more like 1-1 accepted sockets.
4653 * Set the daddr and initialize id to something more random
4655 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
4657 /* Populate the fields of the newsk from the oldsk and migrate the
4658 * asoc to the newsk.
4660 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4666 EXPORT_SYMBOL(sctp_do_peeloff);
4668 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4670 sctp_peeloff_arg_t peeloff;
4671 struct socket *newsock;
4672 struct file *newfile;
4675 if (len < sizeof(sctp_peeloff_arg_t))
4677 len = sizeof(sctp_peeloff_arg_t);
4678 if (copy_from_user(&peeloff, optval, len))
4681 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4685 /* Map the socket to an unused fd that can be returned to the user. */
4686 retval = get_unused_fd_flags(0);
4688 sock_release(newsock);
4692 newfile = sock_alloc_file(newsock, 0, NULL);
4693 if (IS_ERR(newfile)) {
4694 put_unused_fd(retval);
4695 sock_release(newsock);
4696 return PTR_ERR(newfile);
4699 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4702 /* Return the fd mapped to the new socket. */
4703 if (put_user(len, optlen)) {
4705 put_unused_fd(retval);
4708 peeloff.sd = retval;
4709 if (copy_to_user(optval, &peeloff, len)) {
4711 put_unused_fd(retval);
4714 fd_install(retval, newfile);
4719 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4721 * Applications can enable or disable heartbeats for any peer address of
4722 * an association, modify an address's heartbeat interval, force a
4723 * heartbeat to be sent immediately, and adjust the address's maximum
4724 * number of retransmissions sent before an address is considered
4725 * unreachable. The following structure is used to access and modify an
4726 * address's parameters:
4728 * struct sctp_paddrparams {
4729 * sctp_assoc_t spp_assoc_id;
4730 * struct sockaddr_storage spp_address;
4731 * uint32_t spp_hbinterval;
4732 * uint16_t spp_pathmaxrxt;
4733 * uint32_t spp_pathmtu;
4734 * uint32_t spp_sackdelay;
4735 * uint32_t spp_flags;
4738 * spp_assoc_id - (one-to-many style socket) This is filled in the
4739 * application, and identifies the association for
4741 * spp_address - This specifies which address is of interest.
4742 * spp_hbinterval - This contains the value of the heartbeat interval,
4743 * in milliseconds. If a value of zero
4744 * is present in this field then no changes are to
4745 * be made to this parameter.
4746 * spp_pathmaxrxt - This contains the maximum number of
4747 * retransmissions before this address shall be
4748 * considered unreachable. If a value of zero
4749 * is present in this field then no changes are to
4750 * be made to this parameter.
4751 * spp_pathmtu - When Path MTU discovery is disabled the value
4752 * specified here will be the "fixed" path mtu.
4753 * Note that if the spp_address field is empty
4754 * then all associations on this address will
4755 * have this fixed path mtu set upon them.
4757 * spp_sackdelay - When delayed sack is enabled, this value specifies
4758 * the number of milliseconds that sacks will be delayed
4759 * for. This value will apply to all addresses of an
4760 * association if the spp_address field is empty. Note
4761 * also, that if delayed sack is enabled and this
4762 * value is set to 0, no change is made to the last
4763 * recorded delayed sack timer value.
4765 * spp_flags - These flags are used to control various features
4766 * on an association. The flag field may contain
4767 * zero or more of the following options.
4769 * SPP_HB_ENABLE - Enable heartbeats on the
4770 * specified address. Note that if the address
4771 * field is empty all addresses for the association
4772 * have heartbeats enabled upon them.
4774 * SPP_HB_DISABLE - Disable heartbeats on the
4775 * speicifed address. Note that if the address
4776 * field is empty all addresses for the association
4777 * will have their heartbeats disabled. Note also
4778 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4779 * mutually exclusive, only one of these two should
4780 * be specified. Enabling both fields will have
4781 * undetermined results.
4783 * SPP_HB_DEMAND - Request a user initiated heartbeat
4784 * to be made immediately.
4786 * SPP_PMTUD_ENABLE - This field will enable PMTU
4787 * discovery upon the specified address. Note that
4788 * if the address feild is empty then all addresses
4789 * on the association are effected.
4791 * SPP_PMTUD_DISABLE - This field will disable PMTU
4792 * discovery upon the specified address. Note that
4793 * if the address feild is empty then all addresses
4794 * on the association are effected. Not also that
4795 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4796 * exclusive. Enabling both will have undetermined
4799 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4800 * on delayed sack. The time specified in spp_sackdelay
4801 * is used to specify the sack delay for this address. Note
4802 * that if spp_address is empty then all addresses will
4803 * enable delayed sack and take on the sack delay
4804 * value specified in spp_sackdelay.
4805 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4806 * off delayed sack. If the spp_address field is blank then
4807 * delayed sack is disabled for the entire association. Note
4808 * also that this field is mutually exclusive to
4809 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4812 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4813 char __user *optval, int __user *optlen)
4815 struct sctp_paddrparams params;
4816 struct sctp_transport *trans = NULL;
4817 struct sctp_association *asoc = NULL;
4818 struct sctp_sock *sp = sctp_sk(sk);
4820 if (len < sizeof(struct sctp_paddrparams))
4822 len = sizeof(struct sctp_paddrparams);
4823 if (copy_from_user(¶ms, optval, len))
4826 /* If an address other than INADDR_ANY is specified, and
4827 * no transport is found, then the request is invalid.
4829 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
4830 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4831 params.spp_assoc_id);
4833 pr_debug("%s: failed no transport\n", __func__);
4838 /* Get association, if assoc_id != 0 and the socket is a one
4839 * to many style socket, and an association was not found, then
4840 * the id was invalid.
4842 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4843 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4844 pr_debug("%s: failed no association\n", __func__);
4849 /* Fetch transport values. */
4850 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4851 params.spp_pathmtu = trans->pathmtu;
4852 params.spp_pathmaxrxt = trans->pathmaxrxt;
4853 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4855 /*draft-11 doesn't say what to return in spp_flags*/
4856 params.spp_flags = trans->param_flags;
4858 /* Fetch association values. */
4859 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4860 params.spp_pathmtu = asoc->pathmtu;
4861 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4862 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4864 /*draft-11 doesn't say what to return in spp_flags*/
4865 params.spp_flags = asoc->param_flags;
4867 /* Fetch socket values. */
4868 params.spp_hbinterval = sp->hbinterval;
4869 params.spp_pathmtu = sp->pathmtu;
4870 params.spp_sackdelay = sp->sackdelay;
4871 params.spp_pathmaxrxt = sp->pathmaxrxt;
4873 /*draft-11 doesn't say what to return in spp_flags*/
4874 params.spp_flags = sp->param_flags;
4877 if (copy_to_user(optval, ¶ms, len))
4880 if (put_user(len, optlen))
4887 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4889 * This option will effect the way delayed acks are performed. This
4890 * option allows you to get or set the delayed ack time, in
4891 * milliseconds. It also allows changing the delayed ack frequency.
4892 * Changing the frequency to 1 disables the delayed sack algorithm. If
4893 * the assoc_id is 0, then this sets or gets the endpoints default
4894 * values. If the assoc_id field is non-zero, then the set or get
4895 * effects the specified association for the one to many model (the
4896 * assoc_id field is ignored by the one to one model). Note that if
4897 * sack_delay or sack_freq are 0 when setting this option, then the
4898 * current values will remain unchanged.
4900 * struct sctp_sack_info {
4901 * sctp_assoc_t sack_assoc_id;
4902 * uint32_t sack_delay;
4903 * uint32_t sack_freq;
4906 * sack_assoc_id - This parameter, indicates which association the user
4907 * is performing an action upon. Note that if this field's value is
4908 * zero then the endpoints default value is changed (effecting future
4909 * associations only).
4911 * sack_delay - This parameter contains the number of milliseconds that
4912 * the user is requesting the delayed ACK timer be set to. Note that
4913 * this value is defined in the standard to be between 200 and 500
4916 * sack_freq - This parameter contains the number of packets that must
4917 * be received before a sack is sent without waiting for the delay
4918 * timer to expire. The default value for this is 2, setting this
4919 * value to 1 will disable the delayed sack algorithm.
4921 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4922 char __user *optval,
4925 struct sctp_sack_info params;
4926 struct sctp_association *asoc = NULL;
4927 struct sctp_sock *sp = sctp_sk(sk);
4929 if (len >= sizeof(struct sctp_sack_info)) {
4930 len = sizeof(struct sctp_sack_info);
4932 if (copy_from_user(¶ms, optval, len))
4934 } else if (len == sizeof(struct sctp_assoc_value)) {
4935 pr_warn_ratelimited(DEPRECATED
4937 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
4938 "Use struct sctp_sack_info instead\n",
4939 current->comm, task_pid_nr(current));
4940 if (copy_from_user(¶ms, optval, len))
4945 /* Get association, if sack_assoc_id != 0 and the socket is a one
4946 * to many style socket, and an association was not found, then
4947 * the id was invalid.
4949 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4950 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4954 /* Fetch association values. */
4955 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4956 params.sack_delay = jiffies_to_msecs(
4958 params.sack_freq = asoc->sackfreq;
4961 params.sack_delay = 0;
4962 params.sack_freq = 1;
4965 /* Fetch socket values. */
4966 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4967 params.sack_delay = sp->sackdelay;
4968 params.sack_freq = sp->sackfreq;
4970 params.sack_delay = 0;
4971 params.sack_freq = 1;
4975 if (copy_to_user(optval, ¶ms, len))
4978 if (put_user(len, optlen))
4984 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4986 * Applications can specify protocol parameters for the default association
4987 * initialization. The option name argument to setsockopt() and getsockopt()
4990 * Setting initialization parameters is effective only on an unconnected
4991 * socket (for UDP-style sockets only future associations are effected
4992 * by the change). With TCP-style sockets, this option is inherited by
4993 * sockets derived from a listener socket.
4995 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4997 if (len < sizeof(struct sctp_initmsg))
4999 len = sizeof(struct sctp_initmsg);
5000 if (put_user(len, optlen))
5002 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
5008 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
5009 char __user *optval, int __user *optlen)
5011 struct sctp_association *asoc;
5013 struct sctp_getaddrs getaddrs;
5014 struct sctp_transport *from;
5016 union sctp_addr temp;
5017 struct sctp_sock *sp = sctp_sk(sk);
5022 if (len < sizeof(struct sctp_getaddrs))
5025 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5028 /* For UDP-style sockets, id specifies the association to query. */
5029 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5033 to = optval + offsetof(struct sctp_getaddrs, addrs);
5034 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5036 list_for_each_entry(from, &asoc->peer.transport_addr_list,
5038 memcpy(&temp, &from->ipaddr, sizeof(temp));
5039 addrlen = sctp_get_pf_specific(sk->sk_family)
5040 ->addr_to_user(sp, &temp);
5041 if (space_left < addrlen)
5043 if (copy_to_user(to, &temp, addrlen))
5047 space_left -= addrlen;
5050 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
5052 bytes_copied = ((char __user *)to) - optval;
5053 if (put_user(bytes_copied, optlen))
5059 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
5060 size_t space_left, int *bytes_copied)
5062 struct sctp_sockaddr_entry *addr;
5063 union sctp_addr temp;
5066 struct net *net = sock_net(sk);
5069 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
5073 if ((PF_INET == sk->sk_family) &&
5074 (AF_INET6 == addr->a.sa.sa_family))
5076 if ((PF_INET6 == sk->sk_family) &&
5077 inet_v6_ipv6only(sk) &&
5078 (AF_INET == addr->a.sa.sa_family))
5080 memcpy(&temp, &addr->a, sizeof(temp));
5081 if (!temp.v4.sin_port)
5082 temp.v4.sin_port = htons(port);
5084 addrlen = sctp_get_pf_specific(sk->sk_family)
5085 ->addr_to_user(sctp_sk(sk), &temp);
5087 if (space_left < addrlen) {
5091 memcpy(to, &temp, addrlen);
5095 space_left -= addrlen;
5096 *bytes_copied += addrlen;
5104 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
5105 char __user *optval, int __user *optlen)
5107 struct sctp_bind_addr *bp;
5108 struct sctp_association *asoc;
5110 struct sctp_getaddrs getaddrs;
5111 struct sctp_sockaddr_entry *addr;
5113 union sctp_addr temp;
5114 struct sctp_sock *sp = sctp_sk(sk);
5118 int bytes_copied = 0;
5122 if (len < sizeof(struct sctp_getaddrs))
5125 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5129 * For UDP-style sockets, id specifies the association to query.
5130 * If the id field is set to the value '0' then the locally bound
5131 * addresses are returned without regard to any particular
5134 if (0 == getaddrs.assoc_id) {
5135 bp = &sctp_sk(sk)->ep->base.bind_addr;
5137 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5140 bp = &asoc->base.bind_addr;
5143 to = optval + offsetof(struct sctp_getaddrs, addrs);
5144 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5146 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN);
5150 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5151 * addresses from the global local address list.
5153 if (sctp_list_single_entry(&bp->address_list)) {
5154 addr = list_entry(bp->address_list.next,
5155 struct sctp_sockaddr_entry, list);
5156 if (sctp_is_any(sk, &addr->a)) {
5157 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
5158 space_left, &bytes_copied);
5168 /* Protection on the bound address list is not needed since
5169 * in the socket option context we hold a socket lock and
5170 * thus the bound address list can't change.
5172 list_for_each_entry(addr, &bp->address_list, list) {
5173 memcpy(&temp, &addr->a, sizeof(temp));
5174 addrlen = sctp_get_pf_specific(sk->sk_family)
5175 ->addr_to_user(sp, &temp);
5176 if (space_left < addrlen) {
5177 err = -ENOMEM; /*fixme: right error?*/
5180 memcpy(buf, &temp, addrlen);
5182 bytes_copied += addrlen;
5184 space_left -= addrlen;
5188 if (copy_to_user(to, addrs, bytes_copied)) {
5192 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
5196 if (put_user(bytes_copied, optlen))
5203 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
5205 * Requests that the local SCTP stack use the enclosed peer address as
5206 * the association primary. The enclosed address must be one of the
5207 * association peer's addresses.
5209 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
5210 char __user *optval, int __user *optlen)
5212 struct sctp_prim prim;
5213 struct sctp_association *asoc;
5214 struct sctp_sock *sp = sctp_sk(sk);
5216 if (len < sizeof(struct sctp_prim))
5219 len = sizeof(struct sctp_prim);
5221 if (copy_from_user(&prim, optval, len))
5224 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
5228 if (!asoc->peer.primary_path)
5231 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
5232 asoc->peer.primary_path->af_specific->sockaddr_len);
5234 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
5235 (union sctp_addr *)&prim.ssp_addr);
5237 if (put_user(len, optlen))
5239 if (copy_to_user(optval, &prim, len))
5246 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5248 * Requests that the local endpoint set the specified Adaptation Layer
5249 * Indication parameter for all future INIT and INIT-ACK exchanges.
5251 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
5252 char __user *optval, int __user *optlen)
5254 struct sctp_setadaptation adaptation;
5256 if (len < sizeof(struct sctp_setadaptation))
5259 len = sizeof(struct sctp_setadaptation);
5261 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
5263 if (put_user(len, optlen))
5265 if (copy_to_user(optval, &adaptation, len))
5273 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5275 * Applications that wish to use the sendto() system call may wish to
5276 * specify a default set of parameters that would normally be supplied
5277 * through the inclusion of ancillary data. This socket option allows
5278 * such an application to set the default sctp_sndrcvinfo structure.
5281 * The application that wishes to use this socket option simply passes
5282 * in to this call the sctp_sndrcvinfo structure defined in Section
5283 * 5.2.2) The input parameters accepted by this call include
5284 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5285 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5286 * to this call if the caller is using the UDP model.
5288 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5290 static int sctp_getsockopt_default_send_param(struct sock *sk,
5291 int len, char __user *optval,
5294 struct sctp_sock *sp = sctp_sk(sk);
5295 struct sctp_association *asoc;
5296 struct sctp_sndrcvinfo info;
5298 if (len < sizeof(info))
5303 if (copy_from_user(&info, optval, len))
5306 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
5307 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
5310 info.sinfo_stream = asoc->default_stream;
5311 info.sinfo_flags = asoc->default_flags;
5312 info.sinfo_ppid = asoc->default_ppid;
5313 info.sinfo_context = asoc->default_context;
5314 info.sinfo_timetolive = asoc->default_timetolive;
5316 info.sinfo_stream = sp->default_stream;
5317 info.sinfo_flags = sp->default_flags;
5318 info.sinfo_ppid = sp->default_ppid;
5319 info.sinfo_context = sp->default_context;
5320 info.sinfo_timetolive = sp->default_timetolive;
5323 if (put_user(len, optlen))
5325 if (copy_to_user(optval, &info, len))
5331 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5332 * (SCTP_DEFAULT_SNDINFO)
5334 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
5335 char __user *optval,
5338 struct sctp_sock *sp = sctp_sk(sk);
5339 struct sctp_association *asoc;
5340 struct sctp_sndinfo info;
5342 if (len < sizeof(info))
5347 if (copy_from_user(&info, optval, len))
5350 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
5351 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
5354 info.snd_sid = asoc->default_stream;
5355 info.snd_flags = asoc->default_flags;
5356 info.snd_ppid = asoc->default_ppid;
5357 info.snd_context = asoc->default_context;
5359 info.snd_sid = sp->default_stream;
5360 info.snd_flags = sp->default_flags;
5361 info.snd_ppid = sp->default_ppid;
5362 info.snd_context = sp->default_context;
5365 if (put_user(len, optlen))
5367 if (copy_to_user(optval, &info, len))
5375 * 7.1.5 SCTP_NODELAY
5377 * Turn on/off any Nagle-like algorithm. This means that packets are
5378 * generally sent as soon as possible and no unnecessary delays are
5379 * introduced, at the cost of more packets in the network. Expects an
5380 * integer boolean flag.
5383 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5384 char __user *optval, int __user *optlen)
5388 if (len < sizeof(int))
5392 val = (sctp_sk(sk)->nodelay == 1);
5393 if (put_user(len, optlen))
5395 if (copy_to_user(optval, &val, len))
5402 * 7.1.1 SCTP_RTOINFO
5404 * The protocol parameters used to initialize and bound retransmission
5405 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5406 * and modify these parameters.
5407 * All parameters are time values, in milliseconds. A value of 0, when
5408 * modifying the parameters, indicates that the current value should not
5412 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5413 char __user *optval,
5414 int __user *optlen) {
5415 struct sctp_rtoinfo rtoinfo;
5416 struct sctp_association *asoc;
5418 if (len < sizeof (struct sctp_rtoinfo))
5421 len = sizeof(struct sctp_rtoinfo);
5423 if (copy_from_user(&rtoinfo, optval, len))
5426 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5428 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5431 /* Values corresponding to the specific association. */
5433 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5434 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5435 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5437 /* Values corresponding to the endpoint. */
5438 struct sctp_sock *sp = sctp_sk(sk);
5440 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5441 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5442 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5445 if (put_user(len, optlen))
5448 if (copy_to_user(optval, &rtoinfo, len))
5456 * 7.1.2 SCTP_ASSOCINFO
5458 * This option is used to tune the maximum retransmission attempts
5459 * of the association.
5460 * Returns an error if the new association retransmission value is
5461 * greater than the sum of the retransmission value of the peer.
5462 * See [SCTP] for more information.
5465 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5466 char __user *optval,
5470 struct sctp_assocparams assocparams;
5471 struct sctp_association *asoc;
5472 struct list_head *pos;
5475 if (len < sizeof (struct sctp_assocparams))
5478 len = sizeof(struct sctp_assocparams);
5480 if (copy_from_user(&assocparams, optval, len))
5483 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5485 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5488 /* Values correspoinding to the specific association */
5490 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5491 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5492 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5493 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5495 list_for_each(pos, &asoc->peer.transport_addr_list) {
5499 assocparams.sasoc_number_peer_destinations = cnt;
5501 /* Values corresponding to the endpoint */
5502 struct sctp_sock *sp = sctp_sk(sk);
5504 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5505 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5506 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5507 assocparams.sasoc_cookie_life =
5508 sp->assocparams.sasoc_cookie_life;
5509 assocparams.sasoc_number_peer_destinations =
5511 sasoc_number_peer_destinations;
5514 if (put_user(len, optlen))
5517 if (copy_to_user(optval, &assocparams, len))
5524 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5526 * This socket option is a boolean flag which turns on or off mapped V4
5527 * addresses. If this option is turned on and the socket is type
5528 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5529 * If this option is turned off, then no mapping will be done of V4
5530 * addresses and a user will receive both PF_INET6 and PF_INET type
5531 * addresses on the socket.
5533 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5534 char __user *optval, int __user *optlen)
5537 struct sctp_sock *sp = sctp_sk(sk);
5539 if (len < sizeof(int))
5544 if (put_user(len, optlen))
5546 if (copy_to_user(optval, &val, len))
5553 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5554 * (chapter and verse is quoted at sctp_setsockopt_context())
5556 static int sctp_getsockopt_context(struct sock *sk, int len,
5557 char __user *optval, int __user *optlen)
5559 struct sctp_assoc_value params;
5560 struct sctp_sock *sp;
5561 struct sctp_association *asoc;
5563 if (len < sizeof(struct sctp_assoc_value))
5566 len = sizeof(struct sctp_assoc_value);
5568 if (copy_from_user(¶ms, optval, len))
5573 if (params.assoc_id != 0) {
5574 asoc = sctp_id2assoc(sk, params.assoc_id);
5577 params.assoc_value = asoc->default_rcv_context;
5579 params.assoc_value = sp->default_rcv_context;
5582 if (put_user(len, optlen))
5584 if (copy_to_user(optval, ¶ms, len))
5591 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5592 * This option will get or set the maximum size to put in any outgoing
5593 * SCTP DATA chunk. If a message is larger than this size it will be
5594 * fragmented by SCTP into the specified size. Note that the underlying
5595 * SCTP implementation may fragment into smaller sized chunks when the
5596 * PMTU of the underlying association is smaller than the value set by
5597 * the user. The default value for this option is '0' which indicates
5598 * the user is NOT limiting fragmentation and only the PMTU will effect
5599 * SCTP's choice of DATA chunk size. Note also that values set larger
5600 * than the maximum size of an IP datagram will effectively let SCTP
5601 * control fragmentation (i.e. the same as setting this option to 0).
5603 * The following structure is used to access and modify this parameter:
5605 * struct sctp_assoc_value {
5606 * sctp_assoc_t assoc_id;
5607 * uint32_t assoc_value;
5610 * assoc_id: This parameter is ignored for one-to-one style sockets.
5611 * For one-to-many style sockets this parameter indicates which
5612 * association the user is performing an action upon. Note that if
5613 * this field's value is zero then the endpoints default value is
5614 * changed (effecting future associations only).
5615 * assoc_value: This parameter specifies the maximum size in bytes.
5617 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5618 char __user *optval, int __user *optlen)
5620 struct sctp_assoc_value params;
5621 struct sctp_association *asoc;
5623 if (len == sizeof(int)) {
5624 pr_warn_ratelimited(DEPRECATED
5626 "Use of int in maxseg socket option.\n"
5627 "Use struct sctp_assoc_value instead\n",
5628 current->comm, task_pid_nr(current));
5629 params.assoc_id = 0;
5630 } else if (len >= sizeof(struct sctp_assoc_value)) {
5631 len = sizeof(struct sctp_assoc_value);
5632 if (copy_from_user(¶ms, optval, sizeof(params)))
5637 asoc = sctp_id2assoc(sk, params.assoc_id);
5638 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5642 params.assoc_value = asoc->frag_point;
5644 params.assoc_value = sctp_sk(sk)->user_frag;
5646 if (put_user(len, optlen))
5648 if (len == sizeof(int)) {
5649 if (copy_to_user(optval, ¶ms.assoc_value, len))
5652 if (copy_to_user(optval, ¶ms, len))
5660 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5661 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5663 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5664 char __user *optval, int __user *optlen)
5668 if (len < sizeof(int))
5673 val = sctp_sk(sk)->frag_interleave;
5674 if (put_user(len, optlen))
5676 if (copy_to_user(optval, &val, len))
5683 * 7.1.25. Set or Get the sctp partial delivery point
5684 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5686 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5687 char __user *optval,
5692 if (len < sizeof(u32))
5697 val = sctp_sk(sk)->pd_point;
5698 if (put_user(len, optlen))
5700 if (copy_to_user(optval, &val, len))
5707 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5708 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5710 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5711 char __user *optval,
5714 struct sctp_assoc_value params;
5715 struct sctp_sock *sp;
5716 struct sctp_association *asoc;
5718 if (len == sizeof(int)) {
5719 pr_warn_ratelimited(DEPRECATED
5721 "Use of int in max_burst socket option.\n"
5722 "Use struct sctp_assoc_value instead\n",
5723 current->comm, task_pid_nr(current));
5724 params.assoc_id = 0;
5725 } else if (len >= sizeof(struct sctp_assoc_value)) {
5726 len = sizeof(struct sctp_assoc_value);
5727 if (copy_from_user(¶ms, optval, len))
5734 if (params.assoc_id != 0) {
5735 asoc = sctp_id2assoc(sk, params.assoc_id);
5738 params.assoc_value = asoc->max_burst;
5740 params.assoc_value = sp->max_burst;
5742 if (len == sizeof(int)) {
5743 if (copy_to_user(optval, ¶ms.assoc_value, len))
5746 if (copy_to_user(optval, ¶ms, len))
5754 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5755 char __user *optval, int __user *optlen)
5757 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5758 struct sctp_hmacalgo __user *p = (void __user *)optval;
5759 struct sctp_hmac_algo_param *hmacs;
5764 if (!ep->auth_enable)
5767 hmacs = ep->auth_hmacs_list;
5768 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5770 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5773 len = sizeof(struct sctp_hmacalgo) + data_len;
5774 num_idents = data_len / sizeof(u16);
5776 if (put_user(len, optlen))
5778 if (put_user(num_idents, &p->shmac_num_idents))
5780 for (i = 0; i < num_idents; i++) {
5781 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
5783 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
5789 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5790 char __user *optval, int __user *optlen)
5792 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5793 struct sctp_authkeyid val;
5794 struct sctp_association *asoc;
5796 if (!ep->auth_enable)
5799 if (len < sizeof(struct sctp_authkeyid))
5801 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5804 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5805 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5809 val.scact_keynumber = asoc->active_key_id;
5811 val.scact_keynumber = ep->active_key_id;
5813 len = sizeof(struct sctp_authkeyid);
5814 if (put_user(len, optlen))
5816 if (copy_to_user(optval, &val, len))
5822 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5823 char __user *optval, int __user *optlen)
5825 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5826 struct sctp_authchunks __user *p = (void __user *)optval;
5827 struct sctp_authchunks val;
5828 struct sctp_association *asoc;
5829 struct sctp_chunks_param *ch;
5833 if (!ep->auth_enable)
5836 if (len < sizeof(struct sctp_authchunks))
5839 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5842 to = p->gauth_chunks;
5843 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5847 ch = asoc->peer.peer_chunks;
5851 /* See if the user provided enough room for all the data */
5852 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5853 if (len < num_chunks)
5856 if (copy_to_user(to, ch->chunks, num_chunks))
5859 len = sizeof(struct sctp_authchunks) + num_chunks;
5860 if (put_user(len, optlen))
5862 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5867 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5868 char __user *optval, int __user *optlen)
5870 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5871 struct sctp_authchunks __user *p = (void __user *)optval;
5872 struct sctp_authchunks val;
5873 struct sctp_association *asoc;
5874 struct sctp_chunks_param *ch;
5878 if (!ep->auth_enable)
5881 if (len < sizeof(struct sctp_authchunks))
5884 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5887 to = p->gauth_chunks;
5888 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5889 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5893 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
5895 ch = ep->auth_chunk_list;
5900 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5901 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5904 if (copy_to_user(to, ch->chunks, num_chunks))
5907 len = sizeof(struct sctp_authchunks) + num_chunks;
5908 if (put_user(len, optlen))
5910 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5917 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5918 * This option gets the current number of associations that are attached
5919 * to a one-to-many style socket. The option value is an uint32_t.
5921 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5922 char __user *optval, int __user *optlen)
5924 struct sctp_sock *sp = sctp_sk(sk);
5925 struct sctp_association *asoc;
5928 if (sctp_style(sk, TCP))
5931 if (len < sizeof(u32))
5936 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5940 if (put_user(len, optlen))
5942 if (copy_to_user(optval, &val, len))
5949 * 8.1.23 SCTP_AUTO_ASCONF
5950 * See the corresponding setsockopt entry as description
5952 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5953 char __user *optval, int __user *optlen)
5957 if (len < sizeof(int))
5961 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5963 if (put_user(len, optlen))
5965 if (copy_to_user(optval, &val, len))
5971 * 8.2.6. Get the Current Identifiers of Associations
5972 * (SCTP_GET_ASSOC_ID_LIST)
5974 * This option gets the current list of SCTP association identifiers of
5975 * the SCTP associations handled by a one-to-many style socket.
5977 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5978 char __user *optval, int __user *optlen)
5980 struct sctp_sock *sp = sctp_sk(sk);
5981 struct sctp_association *asoc;
5982 struct sctp_assoc_ids *ids;
5985 if (sctp_style(sk, TCP))
5988 if (len < sizeof(struct sctp_assoc_ids))
5991 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5995 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5998 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
6000 ids = kmalloc(len, GFP_USER | __GFP_NOWARN);
6004 ids->gaids_number_of_ids = num;
6006 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6007 ids->gaids_assoc_id[num++] = asoc->assoc_id;
6010 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
6020 * SCTP_PEER_ADDR_THLDS
6022 * This option allows us to fetch the partially failed threshold for one or all
6023 * transports in an association. See Section 6.1 of:
6024 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6026 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
6027 char __user *optval,
6031 struct sctp_paddrthlds val;
6032 struct sctp_transport *trans;
6033 struct sctp_association *asoc;
6035 if (len < sizeof(struct sctp_paddrthlds))
6037 len = sizeof(struct sctp_paddrthlds);
6038 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
6041 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
6042 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
6046 val.spt_pathpfthld = asoc->pf_retrans;
6047 val.spt_pathmaxrxt = asoc->pathmaxrxt;
6049 trans = sctp_addr_id2transport(sk, &val.spt_address,
6054 val.spt_pathmaxrxt = trans->pathmaxrxt;
6055 val.spt_pathpfthld = trans->pf_retrans;
6058 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
6065 * SCTP_GET_ASSOC_STATS
6067 * This option retrieves local per endpoint statistics. It is modeled
6068 * after OpenSolaris' implementation
6070 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
6071 char __user *optval,
6074 struct sctp_assoc_stats sas;
6075 struct sctp_association *asoc = NULL;
6077 /* User must provide at least the assoc id */
6078 if (len < sizeof(sctp_assoc_t))
6081 /* Allow the struct to grow and fill in as much as possible */
6082 len = min_t(size_t, len, sizeof(sas));
6084 if (copy_from_user(&sas, optval, len))
6087 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
6091 sas.sas_rtxchunks = asoc->stats.rtxchunks;
6092 sas.sas_gapcnt = asoc->stats.gapcnt;
6093 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
6094 sas.sas_osacks = asoc->stats.osacks;
6095 sas.sas_isacks = asoc->stats.isacks;
6096 sas.sas_octrlchunks = asoc->stats.octrlchunks;
6097 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
6098 sas.sas_oodchunks = asoc->stats.oodchunks;
6099 sas.sas_iodchunks = asoc->stats.iodchunks;
6100 sas.sas_ouodchunks = asoc->stats.ouodchunks;
6101 sas.sas_iuodchunks = asoc->stats.iuodchunks;
6102 sas.sas_idupchunks = asoc->stats.idupchunks;
6103 sas.sas_opackets = asoc->stats.opackets;
6104 sas.sas_ipackets = asoc->stats.ipackets;
6106 /* New high max rto observed, will return 0 if not a single
6107 * RTO update took place. obs_rto_ipaddr will be bogus
6110 sas.sas_maxrto = asoc->stats.max_obs_rto;
6111 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
6112 sizeof(struct sockaddr_storage));
6114 /* Mark beginning of a new observation period */
6115 asoc->stats.max_obs_rto = asoc->rto_min;
6117 if (put_user(len, optlen))
6120 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
6122 if (copy_to_user(optval, &sas, len))
6128 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
6129 char __user *optval,
6134 if (len < sizeof(int))
6138 if (sctp_sk(sk)->recvrcvinfo)
6140 if (put_user(len, optlen))
6142 if (copy_to_user(optval, &val, len))
6148 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
6149 char __user *optval,
6154 if (len < sizeof(int))
6158 if (sctp_sk(sk)->recvnxtinfo)
6160 if (put_user(len, optlen))
6162 if (copy_to_user(optval, &val, len))
6168 static int sctp_getsockopt(struct sock *sk, int level, int optname,
6169 char __user *optval, int __user *optlen)
6174 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
6176 /* I can hardly begin to describe how wrong this is. This is
6177 * so broken as to be worse than useless. The API draft
6178 * REALLY is NOT helpful here... I am not convinced that the
6179 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
6180 * are at all well-founded.
6182 if (level != SOL_SCTP) {
6183 struct sctp_af *af = sctp_sk(sk)->pf->af;
6185 retval = af->getsockopt(sk, level, optname, optval, optlen);
6189 if (get_user(len, optlen))
6196 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
6198 case SCTP_DISABLE_FRAGMENTS:
6199 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
6203 retval = sctp_getsockopt_events(sk, len, optval, optlen);
6205 case SCTP_AUTOCLOSE:
6206 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
6208 case SCTP_SOCKOPT_PEELOFF:
6209 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
6211 case SCTP_PEER_ADDR_PARAMS:
6212 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
6215 case SCTP_DELAYED_SACK:
6216 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
6220 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
6222 case SCTP_GET_PEER_ADDRS:
6223 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
6226 case SCTP_GET_LOCAL_ADDRS:
6227 retval = sctp_getsockopt_local_addrs(sk, len, optval,
6230 case SCTP_SOCKOPT_CONNECTX3:
6231 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
6233 case SCTP_DEFAULT_SEND_PARAM:
6234 retval = sctp_getsockopt_default_send_param(sk, len,
6237 case SCTP_DEFAULT_SNDINFO:
6238 retval = sctp_getsockopt_default_sndinfo(sk, len,
6241 case SCTP_PRIMARY_ADDR:
6242 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
6245 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
6248 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
6250 case SCTP_ASSOCINFO:
6251 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
6253 case SCTP_I_WANT_MAPPED_V4_ADDR:
6254 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
6257 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
6259 case SCTP_GET_PEER_ADDR_INFO:
6260 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
6263 case SCTP_ADAPTATION_LAYER:
6264 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
6268 retval = sctp_getsockopt_context(sk, len, optval, optlen);
6270 case SCTP_FRAGMENT_INTERLEAVE:
6271 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
6274 case SCTP_PARTIAL_DELIVERY_POINT:
6275 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
6278 case SCTP_MAX_BURST:
6279 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
6282 case SCTP_AUTH_CHUNK:
6283 case SCTP_AUTH_DELETE_KEY:
6284 retval = -EOPNOTSUPP;
6286 case SCTP_HMAC_IDENT:
6287 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
6289 case SCTP_AUTH_ACTIVE_KEY:
6290 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
6292 case SCTP_PEER_AUTH_CHUNKS:
6293 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
6296 case SCTP_LOCAL_AUTH_CHUNKS:
6297 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
6300 case SCTP_GET_ASSOC_NUMBER:
6301 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
6303 case SCTP_GET_ASSOC_ID_LIST:
6304 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
6306 case SCTP_AUTO_ASCONF:
6307 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
6309 case SCTP_PEER_ADDR_THLDS:
6310 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
6312 case SCTP_GET_ASSOC_STATS:
6313 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
6315 case SCTP_RECVRCVINFO:
6316 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
6318 case SCTP_RECVNXTINFO:
6319 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
6322 retval = -ENOPROTOOPT;
6330 static int sctp_hash(struct sock *sk)
6336 static void sctp_unhash(struct sock *sk)
6341 /* Check if port is acceptable. Possibly find first available port.
6343 * The port hash table (contained in the 'global' SCTP protocol storage
6344 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
6345 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
6346 * list (the list number is the port number hashed out, so as you
6347 * would expect from a hash function, all the ports in a given list have
6348 * such a number that hashes out to the same list number; you were
6349 * expecting that, right?); so each list has a set of ports, with a
6350 * link to the socket (struct sock) that uses it, the port number and
6351 * a fastreuse flag (FIXME: NPI ipg).
6353 static struct sctp_bind_bucket *sctp_bucket_create(
6354 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
6356 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
6358 struct sctp_bind_hashbucket *head; /* hash list */
6359 struct sctp_bind_bucket *pp;
6360 unsigned short snum;
6363 snum = ntohs(addr->v4.sin_port);
6365 pr_debug("%s: begins, snum:%d\n", __func__, snum);
6370 /* Search for an available port. */
6371 int low, high, remaining, index;
6373 struct net *net = sock_net(sk);
6375 inet_get_local_port_range(net, &low, &high);
6376 remaining = (high - low) + 1;
6377 rover = prandom_u32() % remaining + low;
6381 if ((rover < low) || (rover > high))
6383 if (inet_is_local_reserved_port(net, rover))
6385 index = sctp_phashfn(sock_net(sk), rover);
6386 head = &sctp_port_hashtable[index];
6387 spin_lock(&head->lock);
6388 sctp_for_each_hentry(pp, &head->chain)
6389 if ((pp->port == rover) &&
6390 net_eq(sock_net(sk), pp->net))
6394 spin_unlock(&head->lock);
6395 } while (--remaining > 0);
6397 /* Exhausted local port range during search? */
6402 /* OK, here is the one we will use. HEAD (the port
6403 * hash table list entry) is non-NULL and we hold it's
6408 /* We are given an specific port number; we verify
6409 * that it is not being used. If it is used, we will
6410 * exahust the search in the hash list corresponding
6411 * to the port number (snum) - we detect that with the
6412 * port iterator, pp being NULL.
6414 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
6415 spin_lock(&head->lock);
6416 sctp_for_each_hentry(pp, &head->chain) {
6417 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
6424 if (!hlist_empty(&pp->owner)) {
6425 /* We had a port hash table hit - there is an
6426 * available port (pp != NULL) and it is being
6427 * used by other socket (pp->owner not empty); that other
6428 * socket is going to be sk2.
6430 int reuse = sk->sk_reuse;
6433 pr_debug("%s: found a possible match\n", __func__);
6435 if (pp->fastreuse && sk->sk_reuse &&
6436 sk->sk_state != SCTP_SS_LISTENING)
6439 /* Run through the list of sockets bound to the port
6440 * (pp->port) [via the pointers bind_next and
6441 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6442 * we get the endpoint they describe and run through
6443 * the endpoint's list of IP (v4 or v6) addresses,
6444 * comparing each of the addresses with the address of
6445 * the socket sk. If we find a match, then that means
6446 * that this port/socket (sk) combination are already
6449 sk_for_each_bound(sk2, &pp->owner) {
6450 struct sctp_endpoint *ep2;
6451 ep2 = sctp_sk(sk2)->ep;
6454 (reuse && sk2->sk_reuse &&
6455 sk2->sk_state != SCTP_SS_LISTENING))
6458 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
6459 sctp_sk(sk2), sctp_sk(sk))) {
6465 pr_debug("%s: found a match\n", __func__);
6468 /* If there was a hash table miss, create a new port. */
6470 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6473 /* In either case (hit or miss), make sure fastreuse is 1 only
6474 * if sk->sk_reuse is too (that is, if the caller requested
6475 * SO_REUSEADDR on this socket -sk-).
6477 if (hlist_empty(&pp->owner)) {
6478 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6482 } else if (pp->fastreuse &&
6483 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6486 /* We are set, so fill up all the data in the hash table
6487 * entry, tie the socket list information with the rest of the
6488 * sockets FIXME: Blurry, NPI (ipg).
6491 if (!sctp_sk(sk)->bind_hash) {
6492 inet_sk(sk)->inet_num = snum;
6493 sk_add_bind_node(sk, &pp->owner);
6494 sctp_sk(sk)->bind_hash = pp;
6499 spin_unlock(&head->lock);
6506 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6507 * port is requested.
6509 static int sctp_get_port(struct sock *sk, unsigned short snum)
6511 union sctp_addr addr;
6512 struct sctp_af *af = sctp_sk(sk)->pf->af;
6514 /* Set up a dummy address struct from the sk. */
6515 af->from_sk(&addr, sk);
6516 addr.v4.sin_port = htons(snum);
6518 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6519 return !!sctp_get_port_local(sk, &addr);
6523 * Move a socket to LISTENING state.
6525 static int sctp_listen_start(struct sock *sk, int backlog)
6527 struct sctp_sock *sp = sctp_sk(sk);
6528 struct sctp_endpoint *ep = sp->ep;
6529 struct crypto_shash *tfm = NULL;
6532 /* Allocate HMAC for generating cookie. */
6533 if (!sp->hmac && sp->sctp_hmac_alg) {
6534 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6535 tfm = crypto_alloc_shash(alg, 0, 0);
6537 net_info_ratelimited("failed to load transform for %s: %ld\n",
6538 sp->sctp_hmac_alg, PTR_ERR(tfm));
6541 sctp_sk(sk)->hmac = tfm;
6545 * If a bind() or sctp_bindx() is not called prior to a listen()
6546 * call that allows new associations to be accepted, the system
6547 * picks an ephemeral port and will choose an address set equivalent
6548 * to binding with a wildcard address.
6550 * This is not currently spelled out in the SCTP sockets
6551 * extensions draft, but follows the practice as seen in TCP
6555 sk->sk_state = SCTP_SS_LISTENING;
6556 if (!ep->base.bind_addr.port) {
6557 if (sctp_autobind(sk))
6560 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6561 sk->sk_state = SCTP_SS_CLOSED;
6566 sk->sk_max_ack_backlog = backlog;
6567 sctp_hash_endpoint(ep);
6572 * 4.1.3 / 5.1.3 listen()
6574 * By default, new associations are not accepted for UDP style sockets.
6575 * An application uses listen() to mark a socket as being able to
6576 * accept new associations.
6578 * On TCP style sockets, applications use listen() to ready the SCTP
6579 * endpoint for accepting inbound associations.
6581 * On both types of endpoints a backlog of '0' disables listening.
6583 * Move a socket to LISTENING state.
6585 int sctp_inet_listen(struct socket *sock, int backlog)
6587 struct sock *sk = sock->sk;
6588 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6591 if (unlikely(backlog < 0))
6596 /* Peeled-off sockets are not allowed to listen(). */
6597 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6600 if (sock->state != SS_UNCONNECTED)
6603 /* If backlog is zero, disable listening. */
6605 if (sctp_sstate(sk, CLOSED))
6609 sctp_unhash_endpoint(ep);
6610 sk->sk_state = SCTP_SS_CLOSED;
6612 sctp_sk(sk)->bind_hash->fastreuse = 1;
6616 /* If we are already listening, just update the backlog */
6617 if (sctp_sstate(sk, LISTENING))
6618 sk->sk_max_ack_backlog = backlog;
6620 err = sctp_listen_start(sk, backlog);
6632 * This function is done by modeling the current datagram_poll() and the
6633 * tcp_poll(). Note that, based on these implementations, we don't
6634 * lock the socket in this function, even though it seems that,
6635 * ideally, locking or some other mechanisms can be used to ensure
6636 * the integrity of the counters (sndbuf and wmem_alloc) used
6637 * in this place. We assume that we don't need locks either until proven
6640 * Another thing to note is that we include the Async I/O support
6641 * here, again, by modeling the current TCP/UDP code. We don't have
6642 * a good way to test with it yet.
6644 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6646 struct sock *sk = sock->sk;
6647 struct sctp_sock *sp = sctp_sk(sk);
6650 poll_wait(file, sk_sleep(sk), wait);
6652 sock_rps_record_flow(sk);
6654 /* A TCP-style listening socket becomes readable when the accept queue
6657 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6658 return (!list_empty(&sp->ep->asocs)) ?
6659 (POLLIN | POLLRDNORM) : 0;
6663 /* Is there any exceptional events? */
6664 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6666 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6667 if (sk->sk_shutdown & RCV_SHUTDOWN)
6668 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6669 if (sk->sk_shutdown == SHUTDOWN_MASK)
6672 /* Is it readable? Reconsider this code with TCP-style support. */
6673 if (!skb_queue_empty(&sk->sk_receive_queue))
6674 mask |= POLLIN | POLLRDNORM;
6676 /* The association is either gone or not ready. */
6677 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6680 /* Is it writable? */
6681 if (sctp_writeable(sk)) {
6682 mask |= POLLOUT | POLLWRNORM;
6684 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
6686 * Since the socket is not locked, the buffer
6687 * might be made available after the writeable check and
6688 * before the bit is set. This could cause a lost I/O
6689 * signal. tcp_poll() has a race breaker for this race
6690 * condition. Based on their implementation, we put
6691 * in the following code to cover it as well.
6693 if (sctp_writeable(sk))
6694 mask |= POLLOUT | POLLWRNORM;
6699 /********************************************************************
6700 * 2nd Level Abstractions
6701 ********************************************************************/
6703 static struct sctp_bind_bucket *sctp_bucket_create(
6704 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6706 struct sctp_bind_bucket *pp;
6708 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6710 SCTP_DBG_OBJCNT_INC(bind_bucket);
6713 INIT_HLIST_HEAD(&pp->owner);
6715 hlist_add_head(&pp->node, &head->chain);
6720 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6721 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6723 if (pp && hlist_empty(&pp->owner)) {
6724 __hlist_del(&pp->node);
6725 kmem_cache_free(sctp_bucket_cachep, pp);
6726 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6730 /* Release this socket's reference to a local port. */
6731 static inline void __sctp_put_port(struct sock *sk)
6733 struct sctp_bind_hashbucket *head =
6734 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6735 inet_sk(sk)->inet_num)];
6736 struct sctp_bind_bucket *pp;
6738 spin_lock(&head->lock);
6739 pp = sctp_sk(sk)->bind_hash;
6740 __sk_del_bind_node(sk);
6741 sctp_sk(sk)->bind_hash = NULL;
6742 inet_sk(sk)->inet_num = 0;
6743 sctp_bucket_destroy(pp);
6744 spin_unlock(&head->lock);
6747 void sctp_put_port(struct sock *sk)
6750 __sctp_put_port(sk);
6755 * The system picks an ephemeral port and choose an address set equivalent
6756 * to binding with a wildcard address.
6757 * One of those addresses will be the primary address for the association.
6758 * This automatically enables the multihoming capability of SCTP.
6760 static int sctp_autobind(struct sock *sk)
6762 union sctp_addr autoaddr;
6766 /* Initialize a local sockaddr structure to INADDR_ANY. */
6767 af = sctp_sk(sk)->pf->af;
6769 port = htons(inet_sk(sk)->inet_num);
6770 af->inaddr_any(&autoaddr, port);
6772 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6775 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6778 * 4.2 The cmsghdr Structure *
6780 * When ancillary data is sent or received, any number of ancillary data
6781 * objects can be specified by the msg_control and msg_controllen members of
6782 * the msghdr structure, because each object is preceded by
6783 * a cmsghdr structure defining the object's length (the cmsg_len member).
6784 * Historically Berkeley-derived implementations have passed only one object
6785 * at a time, but this API allows multiple objects to be
6786 * passed in a single call to sendmsg() or recvmsg(). The following example
6787 * shows two ancillary data objects in a control buffer.
6789 * |<--------------------------- msg_controllen -------------------------->|
6792 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6794 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6797 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6799 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6802 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6803 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6805 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6807 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6814 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
6816 struct cmsghdr *cmsg;
6817 struct msghdr *my_msg = (struct msghdr *)msg;
6819 for_each_cmsghdr(cmsg, my_msg) {
6820 if (!CMSG_OK(my_msg, cmsg))
6823 /* Should we parse this header or ignore? */
6824 if (cmsg->cmsg_level != IPPROTO_SCTP)
6827 /* Strictly check lengths following example in SCM code. */
6828 switch (cmsg->cmsg_type) {
6830 /* SCTP Socket API Extension
6831 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
6833 * This cmsghdr structure provides information for
6834 * initializing new SCTP associations with sendmsg().
6835 * The SCTP_INITMSG socket option uses this same data
6836 * structure. This structure is not used for
6839 * cmsg_level cmsg_type cmsg_data[]
6840 * ------------ ------------ ----------------------
6841 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6843 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
6846 cmsgs->init = CMSG_DATA(cmsg);
6850 /* SCTP Socket API Extension
6851 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
6853 * This cmsghdr structure specifies SCTP options for
6854 * sendmsg() and describes SCTP header information
6855 * about a received message through recvmsg().
6857 * cmsg_level cmsg_type cmsg_data[]
6858 * ------------ ------------ ----------------------
6859 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6861 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6864 cmsgs->srinfo = CMSG_DATA(cmsg);
6866 if (cmsgs->srinfo->sinfo_flags &
6867 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6868 SCTP_SACK_IMMEDIATELY |
6869 SCTP_ABORT | SCTP_EOF))
6874 /* SCTP Socket API Extension
6875 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
6877 * This cmsghdr structure specifies SCTP options for
6878 * sendmsg(). This structure and SCTP_RCVINFO replaces
6879 * SCTP_SNDRCV which has been deprecated.
6881 * cmsg_level cmsg_type cmsg_data[]
6882 * ------------ ------------ ---------------------
6883 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
6885 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
6888 cmsgs->sinfo = CMSG_DATA(cmsg);
6890 if (cmsgs->sinfo->snd_flags &
6891 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6892 SCTP_SACK_IMMEDIATELY |
6893 SCTP_ABORT | SCTP_EOF))
6905 * Wait for a packet..
6906 * Note: This function is the same function as in core/datagram.c
6907 * with a few modifications to make lksctp work.
6909 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
6914 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6916 /* Socket errors? */
6917 error = sock_error(sk);
6921 if (!skb_queue_empty(&sk->sk_receive_queue))
6924 /* Socket shut down? */
6925 if (sk->sk_shutdown & RCV_SHUTDOWN)
6928 /* Sequenced packets can come disconnected. If so we report the
6933 /* Is there a good reason to think that we may receive some data? */
6934 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6937 /* Handle signals. */
6938 if (signal_pending(current))
6941 /* Let another process have a go. Since we are going to sleep
6942 * anyway. Note: This may cause odd behaviors if the message
6943 * does not fit in the user's buffer, but this seems to be the
6944 * only way to honor MSG_DONTWAIT realistically.
6947 *timeo_p = schedule_timeout(*timeo_p);
6951 finish_wait(sk_sleep(sk), &wait);
6955 error = sock_intr_errno(*timeo_p);
6958 finish_wait(sk_sleep(sk), &wait);
6963 /* Receive a datagram.
6964 * Note: This is pretty much the same routine as in core/datagram.c
6965 * with a few changes to make lksctp work.
6967 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6968 int noblock, int *err)
6971 struct sk_buff *skb;
6974 timeo = sock_rcvtimeo(sk, noblock);
6976 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
6977 MAX_SCHEDULE_TIMEOUT);
6980 /* Again only user level code calls this function,
6981 * so nothing interrupt level
6982 * will suddenly eat the receive_queue.
6984 * Look at current nfs client by the way...
6985 * However, this function was correct in any case. 8)
6987 if (flags & MSG_PEEK) {
6988 skb = skb_peek(&sk->sk_receive_queue);
6990 atomic_inc(&skb->users);
6992 skb = __skb_dequeue(&sk->sk_receive_queue);
6998 /* Caller is allowed not to check sk->sk_err before calling. */
6999 error = sock_error(sk);
7003 if (sk->sk_shutdown & RCV_SHUTDOWN)
7006 if (sk_can_busy_loop(sk) &&
7007 sk_busy_loop(sk, noblock))
7010 /* User doesn't want to wait. */
7014 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
7023 /* If sndbuf has changed, wake up per association sndbuf waiters. */
7024 static void __sctp_write_space(struct sctp_association *asoc)
7026 struct sock *sk = asoc->base.sk;
7028 if (sctp_wspace(asoc) <= 0)
7031 if (waitqueue_active(&asoc->wait))
7032 wake_up_interruptible(&asoc->wait);
7034 if (sctp_writeable(sk)) {
7035 struct socket_wq *wq;
7038 wq = rcu_dereference(sk->sk_wq);
7040 if (waitqueue_active(&wq->wait))
7041 wake_up_interruptible(&wq->wait);
7043 /* Note that we try to include the Async I/O support
7044 * here by modeling from the current TCP/UDP code.
7045 * We have not tested with it yet.
7047 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
7048 sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT);
7054 static void sctp_wake_up_waiters(struct sock *sk,
7055 struct sctp_association *asoc)
7057 struct sctp_association *tmp = asoc;
7059 /* We do accounting for the sndbuf space per association,
7060 * so we only need to wake our own association.
7062 if (asoc->ep->sndbuf_policy)
7063 return __sctp_write_space(asoc);
7065 /* If association goes down and is just flushing its
7066 * outq, then just normally notify others.
7068 if (asoc->base.dead)
7069 return sctp_write_space(sk);
7071 /* Accounting for the sndbuf space is per socket, so we
7072 * need to wake up others, try to be fair and in case of
7073 * other associations, let them have a go first instead
7074 * of just doing a sctp_write_space() call.
7076 * Note that we reach sctp_wake_up_waiters() only when
7077 * associations free up queued chunks, thus we are under
7078 * lock and the list of associations on a socket is
7079 * guaranteed not to change.
7081 for (tmp = list_next_entry(tmp, asocs); 1;
7082 tmp = list_next_entry(tmp, asocs)) {
7083 /* Manually skip the head element. */
7084 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
7086 /* Wake up association. */
7087 __sctp_write_space(tmp);
7088 /* We've reached the end. */
7094 /* Do accounting for the sndbuf space.
7095 * Decrement the used sndbuf space of the corresponding association by the
7096 * data size which was just transmitted(freed).
7098 static void sctp_wfree(struct sk_buff *skb)
7100 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
7101 struct sctp_association *asoc = chunk->asoc;
7102 struct sock *sk = asoc->base.sk;
7104 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
7105 sizeof(struct sk_buff) +
7106 sizeof(struct sctp_chunk);
7108 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
7111 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
7113 sk->sk_wmem_queued -= skb->truesize;
7114 sk_mem_uncharge(sk, skb->truesize);
7117 sctp_wake_up_waiters(sk, asoc);
7119 sctp_association_put(asoc);
7122 /* Do accounting for the receive space on the socket.
7123 * Accounting for the association is done in ulpevent.c
7124 * We set this as a destructor for the cloned data skbs so that
7125 * accounting is done at the correct time.
7127 void sctp_sock_rfree(struct sk_buff *skb)
7129 struct sock *sk = skb->sk;
7130 struct sctp_ulpevent *event = sctp_skb2event(skb);
7132 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
7135 * Mimic the behavior of sock_rfree
7137 sk_mem_uncharge(sk, event->rmem_len);
7141 /* Helper function to wait for space in the sndbuf. */
7142 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
7145 struct sock *sk = asoc->base.sk;
7147 long current_timeo = *timeo_p;
7150 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
7153 /* Increment the association's refcnt. */
7154 sctp_association_hold(asoc);
7156 /* Wait on the association specific sndbuf space. */
7158 prepare_to_wait_exclusive(&asoc->wait, &wait,
7159 TASK_INTERRUPTIBLE);
7162 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7165 if (signal_pending(current))
7166 goto do_interrupted;
7167 if (msg_len <= sctp_wspace(asoc))
7170 /* Let another process have a go. Since we are going
7174 current_timeo = schedule_timeout(current_timeo);
7175 BUG_ON(sk != asoc->base.sk);
7178 *timeo_p = current_timeo;
7182 finish_wait(&asoc->wait, &wait);
7184 /* Release the association's refcnt. */
7185 sctp_association_put(asoc);
7194 err = sock_intr_errno(*timeo_p);
7202 void sctp_data_ready(struct sock *sk)
7204 struct socket_wq *wq;
7207 wq = rcu_dereference(sk->sk_wq);
7208 if (skwq_has_sleeper(wq))
7209 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
7210 POLLRDNORM | POLLRDBAND);
7211 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
7215 /* If socket sndbuf has changed, wake up all per association waiters. */
7216 void sctp_write_space(struct sock *sk)
7218 struct sctp_association *asoc;
7220 /* Wake up the tasks in each wait queue. */
7221 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
7222 __sctp_write_space(asoc);
7226 /* Is there any sndbuf space available on the socket?
7228 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
7229 * associations on the same socket. For a UDP-style socket with
7230 * multiple associations, it is possible for it to be "unwriteable"
7231 * prematurely. I assume that this is acceptable because
7232 * a premature "unwriteable" is better than an accidental "writeable" which
7233 * would cause an unwanted block under certain circumstances. For the 1-1
7234 * UDP-style sockets or TCP-style sockets, this code should work.
7237 static int sctp_writeable(struct sock *sk)
7241 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
7247 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
7248 * returns immediately with EINPROGRESS.
7250 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
7252 struct sock *sk = asoc->base.sk;
7254 long current_timeo = *timeo_p;
7257 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
7259 /* Increment the association's refcnt. */
7260 sctp_association_hold(asoc);
7263 prepare_to_wait_exclusive(&asoc->wait, &wait,
7264 TASK_INTERRUPTIBLE);
7267 if (sk->sk_shutdown & RCV_SHUTDOWN)
7269 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7272 if (signal_pending(current))
7273 goto do_interrupted;
7275 if (sctp_state(asoc, ESTABLISHED))
7278 /* Let another process have a go. Since we are going
7282 current_timeo = schedule_timeout(current_timeo);
7285 *timeo_p = current_timeo;
7289 finish_wait(&asoc->wait, &wait);
7291 /* Release the association's refcnt. */
7292 sctp_association_put(asoc);
7297 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
7300 err = -ECONNREFUSED;
7304 err = sock_intr_errno(*timeo_p);
7312 static int sctp_wait_for_accept(struct sock *sk, long timeo)
7314 struct sctp_endpoint *ep;
7318 ep = sctp_sk(sk)->ep;
7322 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
7323 TASK_INTERRUPTIBLE);
7325 if (list_empty(&ep->asocs)) {
7327 timeo = schedule_timeout(timeo);
7332 if (!sctp_sstate(sk, LISTENING))
7336 if (!list_empty(&ep->asocs))
7339 err = sock_intr_errno(timeo);
7340 if (signal_pending(current))
7348 finish_wait(sk_sleep(sk), &wait);
7353 static void sctp_wait_for_close(struct sock *sk, long timeout)
7358 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7359 if (list_empty(&sctp_sk(sk)->ep->asocs))
7362 timeout = schedule_timeout(timeout);
7364 } while (!signal_pending(current) && timeout);
7366 finish_wait(sk_sleep(sk), &wait);
7369 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
7371 struct sk_buff *frag;
7376 /* Don't forget the fragments. */
7377 skb_walk_frags(skb, frag)
7378 sctp_skb_set_owner_r_frag(frag, sk);
7381 sctp_skb_set_owner_r(skb, sk);
7384 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
7385 struct sctp_association *asoc)
7387 struct inet_sock *inet = inet_sk(sk);
7388 struct inet_sock *newinet;
7390 newsk->sk_type = sk->sk_type;
7391 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
7392 newsk->sk_flags = sk->sk_flags;
7393 newsk->sk_tsflags = sk->sk_tsflags;
7394 newsk->sk_no_check_tx = sk->sk_no_check_tx;
7395 newsk->sk_no_check_rx = sk->sk_no_check_rx;
7396 newsk->sk_reuse = sk->sk_reuse;
7398 newsk->sk_shutdown = sk->sk_shutdown;
7399 newsk->sk_destruct = sctp_destruct_sock;
7400 newsk->sk_family = sk->sk_family;
7401 newsk->sk_protocol = IPPROTO_SCTP;
7402 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
7403 newsk->sk_sndbuf = sk->sk_sndbuf;
7404 newsk->sk_rcvbuf = sk->sk_rcvbuf;
7405 newsk->sk_lingertime = sk->sk_lingertime;
7406 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
7407 newsk->sk_sndtimeo = sk->sk_sndtimeo;
7408 newsk->sk_rxhash = sk->sk_rxhash;
7410 newinet = inet_sk(newsk);
7412 /* Initialize sk's sport, dport, rcv_saddr and daddr for
7413 * getsockname() and getpeername()
7415 newinet->inet_sport = inet->inet_sport;
7416 newinet->inet_saddr = inet->inet_saddr;
7417 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
7418 newinet->inet_dport = htons(asoc->peer.port);
7419 newinet->pmtudisc = inet->pmtudisc;
7420 newinet->inet_id = asoc->next_tsn ^ jiffies;
7422 newinet->uc_ttl = inet->uc_ttl;
7423 newinet->mc_loop = 1;
7424 newinet->mc_ttl = 1;
7425 newinet->mc_index = 0;
7426 newinet->mc_list = NULL;
7428 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
7429 net_enable_timestamp();
7431 security_sk_clone(sk, newsk);
7434 static inline void sctp_copy_descendant(struct sock *sk_to,
7435 const struct sock *sk_from)
7437 int ancestor_size = sizeof(struct inet_sock) +
7438 sizeof(struct sctp_sock) -
7439 offsetof(struct sctp_sock, auto_asconf_list);
7441 if (sk_from->sk_family == PF_INET6)
7442 ancestor_size += sizeof(struct ipv6_pinfo);
7444 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
7447 /* Populate the fields of the newsk from the oldsk and migrate the assoc
7448 * and its messages to the newsk.
7450 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
7451 struct sctp_association *assoc,
7452 sctp_socket_type_t type)
7454 struct sctp_sock *oldsp = sctp_sk(oldsk);
7455 struct sctp_sock *newsp = sctp_sk(newsk);
7456 struct sctp_bind_bucket *pp; /* hash list port iterator */
7457 struct sctp_endpoint *newep = newsp->ep;
7458 struct sk_buff *skb, *tmp;
7459 struct sctp_ulpevent *event;
7460 struct sctp_bind_hashbucket *head;
7462 /* Migrate socket buffer sizes and all the socket level options to the
7465 newsk->sk_sndbuf = oldsk->sk_sndbuf;
7466 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
7467 /* Brute force copy old sctp opt. */
7468 sctp_copy_descendant(newsk, oldsk);
7470 /* Restore the ep value that was overwritten with the above structure
7476 /* Hook this new socket in to the bind_hash list. */
7477 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
7478 inet_sk(oldsk)->inet_num)];
7479 spin_lock_bh(&head->lock);
7480 pp = sctp_sk(oldsk)->bind_hash;
7481 sk_add_bind_node(newsk, &pp->owner);
7482 sctp_sk(newsk)->bind_hash = pp;
7483 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
7484 spin_unlock_bh(&head->lock);
7486 /* Copy the bind_addr list from the original endpoint to the new
7487 * endpoint so that we can handle restarts properly
7489 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
7490 &oldsp->ep->base.bind_addr, GFP_KERNEL);
7492 /* Move any messages in the old socket's receive queue that are for the
7493 * peeled off association to the new socket's receive queue.
7495 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
7496 event = sctp_skb2event(skb);
7497 if (event->asoc == assoc) {
7498 __skb_unlink(skb, &oldsk->sk_receive_queue);
7499 __skb_queue_tail(&newsk->sk_receive_queue, skb);
7500 sctp_skb_set_owner_r_frag(skb, newsk);
7504 /* Clean up any messages pending delivery due to partial
7505 * delivery. Three cases:
7506 * 1) No partial deliver; no work.
7507 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7508 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7510 skb_queue_head_init(&newsp->pd_lobby);
7511 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
7513 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
7514 struct sk_buff_head *queue;
7516 /* Decide which queue to move pd_lobby skbs to. */
7517 if (assoc->ulpq.pd_mode) {
7518 queue = &newsp->pd_lobby;
7520 queue = &newsk->sk_receive_queue;
7522 /* Walk through the pd_lobby, looking for skbs that
7523 * need moved to the new socket.
7525 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
7526 event = sctp_skb2event(skb);
7527 if (event->asoc == assoc) {
7528 __skb_unlink(skb, &oldsp->pd_lobby);
7529 __skb_queue_tail(queue, skb);
7530 sctp_skb_set_owner_r_frag(skb, newsk);
7534 /* Clear up any skbs waiting for the partial
7535 * delivery to finish.
7537 if (assoc->ulpq.pd_mode)
7538 sctp_clear_pd(oldsk, NULL);
7542 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7543 sctp_skb_set_owner_r_frag(skb, newsk);
7545 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7546 sctp_skb_set_owner_r_frag(skb, newsk);
7548 /* Set the type of socket to indicate that it is peeled off from the
7549 * original UDP-style socket or created with the accept() call on a
7550 * TCP-style socket..
7554 /* Mark the new socket "in-use" by the user so that any packets
7555 * that may arrive on the association after we've moved it are
7556 * queued to the backlog. This prevents a potential race between
7557 * backlog processing on the old socket and new-packet processing
7558 * on the new socket.
7560 * The caller has just allocated newsk so we can guarantee that other
7561 * paths won't try to lock it and then oldsk.
7563 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7564 sctp_assoc_migrate(assoc, newsk);
7566 /* If the association on the newsk is already closed before accept()
7567 * is called, set RCV_SHUTDOWN flag.
7569 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
7570 newsk->sk_shutdown |= RCV_SHUTDOWN;
7572 newsk->sk_state = SCTP_SS_ESTABLISHED;
7573 release_sock(newsk);
7577 /* This proto struct describes the ULP interface for SCTP. */
7578 struct proto sctp_prot = {
7580 .owner = THIS_MODULE,
7581 .close = sctp_close,
7582 .connect = sctp_connect,
7583 .disconnect = sctp_disconnect,
7584 .accept = sctp_accept,
7585 .ioctl = sctp_ioctl,
7586 .init = sctp_init_sock,
7587 .destroy = sctp_destroy_sock,
7588 .shutdown = sctp_shutdown,
7589 .setsockopt = sctp_setsockopt,
7590 .getsockopt = sctp_getsockopt,
7591 .sendmsg = sctp_sendmsg,
7592 .recvmsg = sctp_recvmsg,
7594 .backlog_rcv = sctp_backlog_rcv,
7596 .unhash = sctp_unhash,
7597 .get_port = sctp_get_port,
7598 .obj_size = sizeof(struct sctp_sock),
7599 .sysctl_mem = sysctl_sctp_mem,
7600 .sysctl_rmem = sysctl_sctp_rmem,
7601 .sysctl_wmem = sysctl_sctp_wmem,
7602 .memory_pressure = &sctp_memory_pressure,
7603 .enter_memory_pressure = sctp_enter_memory_pressure,
7604 .memory_allocated = &sctp_memory_allocated,
7605 .sockets_allocated = &sctp_sockets_allocated,
7608 #if IS_ENABLED(CONFIG_IPV6)
7610 #include <net/transp_v6.h>
7611 static void sctp_v6_destroy_sock(struct sock *sk)
7613 sctp_destroy_sock(sk);
7614 inet6_destroy_sock(sk);
7617 struct proto sctpv6_prot = {
7619 .owner = THIS_MODULE,
7620 .close = sctp_close,
7621 .connect = sctp_connect,
7622 .disconnect = sctp_disconnect,
7623 .accept = sctp_accept,
7624 .ioctl = sctp_ioctl,
7625 .init = sctp_init_sock,
7626 .destroy = sctp_v6_destroy_sock,
7627 .shutdown = sctp_shutdown,
7628 .setsockopt = sctp_setsockopt,
7629 .getsockopt = sctp_getsockopt,
7630 .sendmsg = sctp_sendmsg,
7631 .recvmsg = sctp_recvmsg,
7633 .backlog_rcv = sctp_backlog_rcv,
7635 .unhash = sctp_unhash,
7636 .get_port = sctp_get_port,
7637 .obj_size = sizeof(struct sctp6_sock),
7638 .sysctl_mem = sysctl_sctp_mem,
7639 .sysctl_rmem = sysctl_sctp_rmem,
7640 .sysctl_wmem = sysctl_sctp_wmem,
7641 .memory_pressure = &sctp_memory_pressure,
7642 .enter_memory_pressure = sctp_enter_memory_pressure,
7643 .memory_allocated = &sctp_memory_allocated,
7644 .sockets_allocated = &sctp_sockets_allocated,
7646 #endif /* IS_ENABLED(CONFIG_IPV6) */