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, write to
32 * the Free Software Foundation, 59 Temple Place - Suite 330,
33 * Boston, MA 02111-1307, USA.
35 * Please send any bug reports or fixes you make to the
37 * lksctp developers <lksctp-developers@lists.sourceforge.net>
39 * Or submit a bug report through the following website:
40 * http://www.sf.net/projects/lksctp
42 * Written or modified by:
43 * La Monte H.P. Yarroll <piggy@acm.org>
44 * Narasimha Budihal <narsi@refcode.org>
45 * Karl Knutson <karl@athena.chicago.il.us>
46 * Jon Grimm <jgrimm@us.ibm.com>
47 * Xingang Guo <xingang.guo@intel.com>
48 * Daisy Chang <daisyc@us.ibm.com>
49 * Sridhar Samudrala <samudrala@us.ibm.com>
50 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
51 * Ardelle Fan <ardelle.fan@intel.com>
52 * Ryan Layer <rmlayer@us.ibm.com>
53 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
54 * Kevin Gao <kevin.gao@intel.com>
56 * Any bugs reported given to us we will try to fix... any fixes shared will
57 * be incorporated into the next SCTP release.
60 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
62 #include <linux/types.h>
63 #include <linux/kernel.h>
64 #include <linux/wait.h>
65 #include <linux/time.h>
67 #include <linux/capability.h>
68 #include <linux/fcntl.h>
69 #include <linux/poll.h>
70 #include <linux/init.h>
71 #include <linux/crypto.h>
72 #include <linux/slab.h>
73 #include <linux/file.h>
77 #include <net/route.h>
79 #include <net/inet_common.h>
81 #include <linux/socket.h> /* for sa_family_t */
82 #include <linux/export.h>
84 #include <net/sctp/sctp.h>
85 #include <net/sctp/sm.h>
87 /* Forward declarations for internal helper functions. */
88 static int sctp_writeable(struct sock *sk);
89 static void sctp_wfree(struct sk_buff *skb);
90 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
92 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
93 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
94 static int sctp_wait_for_accept(struct sock *sk, long timeo);
95 static void sctp_wait_for_close(struct sock *sk, long timeo);
96 static void sctp_destruct_sock(struct sock *sk);
97 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
98 union sctp_addr *addr, int len);
99 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
100 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
101 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
102 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
103 static int sctp_send_asconf(struct sctp_association *asoc,
104 struct sctp_chunk *chunk);
105 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
106 static int sctp_autobind(struct sock *sk);
107 static void sctp_sock_migrate(struct sock *, struct sock *,
108 struct sctp_association *, sctp_socket_type_t);
110 extern struct kmem_cache *sctp_bucket_cachep;
111 extern long sysctl_sctp_mem[3];
112 extern int sysctl_sctp_rmem[3];
113 extern int sysctl_sctp_wmem[3];
115 static int sctp_memory_pressure;
116 static atomic_long_t sctp_memory_allocated;
117 struct percpu_counter sctp_sockets_allocated;
119 static void sctp_enter_memory_pressure(struct sock *sk)
121 sctp_memory_pressure = 1;
125 /* Get the sndbuf space available at the time on the association. */
126 static inline int sctp_wspace(struct sctp_association *asoc)
130 if (asoc->ep->sndbuf_policy)
131 amt = asoc->sndbuf_used;
133 amt = sk_wmem_alloc_get(asoc->base.sk);
135 if (amt >= asoc->base.sk->sk_sndbuf) {
136 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
139 amt = sk_stream_wspace(asoc->base.sk);
144 amt = asoc->base.sk->sk_sndbuf - amt;
149 /* Increment the used sndbuf space count of the corresponding association by
150 * the size of the outgoing data chunk.
151 * Also, set the skb destructor for sndbuf accounting later.
153 * Since it is always 1-1 between chunk and skb, and also a new skb is always
154 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
155 * destructor in the data chunk skb for the purpose of the sndbuf space
158 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
160 struct sctp_association *asoc = chunk->asoc;
161 struct sock *sk = asoc->base.sk;
163 /* The sndbuf space is tracked per association. */
164 sctp_association_hold(asoc);
166 skb_set_owner_w(chunk->skb, sk);
168 chunk->skb->destructor = sctp_wfree;
169 /* Save the chunk pointer in skb for sctp_wfree to use later. */
170 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
172 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
173 sizeof(struct sk_buff) +
174 sizeof(struct sctp_chunk);
176 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
177 sk->sk_wmem_queued += chunk->skb->truesize;
178 sk_mem_charge(sk, chunk->skb->truesize);
181 /* Verify that this is a valid address. */
182 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
187 /* Verify basic sockaddr. */
188 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
192 /* Is this a valid SCTP address? */
193 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
196 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
202 /* Look up the association by its id. If this is not a UDP-style
203 * socket, the ID field is always ignored.
205 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
207 struct sctp_association *asoc = NULL;
209 /* If this is not a UDP-style socket, assoc id should be ignored. */
210 if (!sctp_style(sk, UDP)) {
211 /* Return NULL if the socket state is not ESTABLISHED. It
212 * could be a TCP-style listening socket or a socket which
213 * hasn't yet called connect() to establish an association.
215 if (!sctp_sstate(sk, ESTABLISHED))
218 /* Get the first and the only association from the list. */
219 if (!list_empty(&sctp_sk(sk)->ep->asocs))
220 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
221 struct sctp_association, asocs);
225 /* Otherwise this is a UDP-style socket. */
226 if (!id || (id == (sctp_assoc_t)-1))
229 spin_lock_bh(&sctp_assocs_id_lock);
230 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
231 spin_unlock_bh(&sctp_assocs_id_lock);
233 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
239 /* Look up the transport from an address and an assoc id. If both address and
240 * id are specified, the associations matching the address and the id should be
243 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
244 struct sockaddr_storage *addr,
247 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
248 struct sctp_transport *transport;
249 union sctp_addr *laddr = (union sctp_addr *)addr;
251 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
258 id_asoc = sctp_id2assoc(sk, id);
259 if (id_asoc && (id_asoc != addr_asoc))
262 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
263 (union sctp_addr *)addr);
268 /* API 3.1.2 bind() - UDP Style Syntax
269 * The syntax of bind() is,
271 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
273 * sd - the socket descriptor returned by socket().
274 * addr - the address structure (struct sockaddr_in or struct
275 * sockaddr_in6 [RFC 2553]),
276 * addr_len - the size of the address structure.
278 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
284 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
287 /* Disallow binding twice. */
288 if (!sctp_sk(sk)->ep->base.bind_addr.port)
289 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
294 sctp_release_sock(sk);
299 static long sctp_get_port_local(struct sock *, union sctp_addr *);
301 /* Verify this is a valid sockaddr. */
302 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
303 union sctp_addr *addr, int len)
307 /* Check minimum size. */
308 if (len < sizeof (struct sockaddr))
311 /* V4 mapped address are really of AF_INET family */
312 if (addr->sa.sa_family == AF_INET6 &&
313 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
314 if (!opt->pf->af_supported(AF_INET, opt))
317 /* Does this PF support this AF? */
318 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
322 /* If we get this far, af is valid. */
323 af = sctp_get_af_specific(addr->sa.sa_family);
325 if (len < af->sockaddr_len)
331 /* Bind a local address either to an endpoint or to an association. */
332 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
334 struct net *net = sock_net(sk);
335 struct sctp_sock *sp = sctp_sk(sk);
336 struct sctp_endpoint *ep = sp->ep;
337 struct sctp_bind_addr *bp = &ep->base.bind_addr;
342 /* Common sockaddr verification. */
343 af = sctp_sockaddr_af(sp, addr, len);
345 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
350 snum = ntohs(addr->v4.sin_port);
352 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
353 ", port: %d, new port: %d, len: %d)\n",
359 /* PF specific bind() address verification. */
360 if (!sp->pf->bind_verify(sp, addr))
361 return -EADDRNOTAVAIL;
363 /* We must either be unbound, or bind to the same port.
364 * It's OK to allow 0 ports if we are already bound.
365 * We'll just inhert an already bound port in this case
370 else if (snum != bp->port) {
371 SCTP_DEBUG_PRINTK("sctp_do_bind:"
372 " New port %d does not match existing port "
373 "%d.\n", snum, bp->port);
378 if (snum && snum < PROT_SOCK &&
379 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
382 /* See if the address matches any of the addresses we may have
383 * already bound before checking against other endpoints.
385 if (sctp_bind_addr_match(bp, addr, sp))
388 /* Make sure we are allowed to bind here.
389 * The function sctp_get_port_local() does duplicate address
392 addr->v4.sin_port = htons(snum);
393 if ((ret = sctp_get_port_local(sk, addr))) {
397 /* Refresh ephemeral port. */
399 bp->port = inet_sk(sk)->inet_num;
401 /* Add the address to the bind address list.
402 * Use GFP_ATOMIC since BHs will be disabled.
404 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
406 /* Copy back into socket for getsockname() use. */
408 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
409 af->to_sk_saddr(addr, sk);
415 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
417 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
418 * at any one time. If a sender, after sending an ASCONF chunk, decides
419 * it needs to transfer another ASCONF Chunk, it MUST wait until the
420 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
421 * subsequent ASCONF. Note this restriction binds each side, so at any
422 * time two ASCONF may be in-transit on any given association (one sent
423 * from each endpoint).
425 static int sctp_send_asconf(struct sctp_association *asoc,
426 struct sctp_chunk *chunk)
428 struct net *net = sock_net(asoc->base.sk);
431 /* If there is an outstanding ASCONF chunk, queue it for later
434 if (asoc->addip_last_asconf) {
435 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
439 /* Hold the chunk until an ASCONF_ACK is received. */
440 sctp_chunk_hold(chunk);
441 retval = sctp_primitive_ASCONF(net, asoc, chunk);
443 sctp_chunk_free(chunk);
445 asoc->addip_last_asconf = chunk;
451 /* Add a list of addresses as bind addresses to local endpoint or
454 * Basically run through each address specified in the addrs/addrcnt
455 * array/length pair, determine if it is IPv6 or IPv4 and call
456 * sctp_do_bind() on it.
458 * If any of them fails, then the operation will be reversed and the
459 * ones that were added will be removed.
461 * Only sctp_setsockopt_bindx() is supposed to call this function.
463 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
468 struct sockaddr *sa_addr;
471 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
475 for (cnt = 0; cnt < addrcnt; cnt++) {
476 /* The list may contain either IPv4 or IPv6 address;
477 * determine the address length for walking thru the list.
480 af = sctp_get_af_specific(sa_addr->sa_family);
486 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
489 addr_buf += af->sockaddr_len;
493 /* Failed. Cleanup the ones that have been added */
495 sctp_bindx_rem(sk, addrs, cnt);
503 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
504 * associations that are part of the endpoint indicating that a list of local
505 * addresses are added to the endpoint.
507 * If any of the addresses is already in the bind address list of the
508 * association, we do not send the chunk for that association. But it will not
509 * affect other associations.
511 * Only sctp_setsockopt_bindx() is supposed to call this function.
513 static int sctp_send_asconf_add_ip(struct sock *sk,
514 struct sockaddr *addrs,
517 struct net *net = sock_net(sk);
518 struct sctp_sock *sp;
519 struct sctp_endpoint *ep;
520 struct sctp_association *asoc;
521 struct sctp_bind_addr *bp;
522 struct sctp_chunk *chunk;
523 struct sctp_sockaddr_entry *laddr;
524 union sctp_addr *addr;
525 union sctp_addr saveaddr;
532 if (!net->sctp.addip_enable)
538 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
539 __func__, sk, addrs, addrcnt);
541 list_for_each_entry(asoc, &ep->asocs, asocs) {
543 if (!asoc->peer.asconf_capable)
546 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
549 if (!sctp_state(asoc, ESTABLISHED))
552 /* Check if any address in the packed array of addresses is
553 * in the bind address list of the association. If so,
554 * do not send the asconf chunk to its peer, but continue with
555 * other associations.
558 for (i = 0; i < addrcnt; i++) {
560 af = sctp_get_af_specific(addr->v4.sin_family);
566 if (sctp_assoc_lookup_laddr(asoc, addr))
569 addr_buf += af->sockaddr_len;
574 /* Use the first valid address in bind addr list of
575 * association as Address Parameter of ASCONF CHUNK.
577 bp = &asoc->base.bind_addr;
578 p = bp->address_list.next;
579 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
580 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
581 addrcnt, SCTP_PARAM_ADD_IP);
587 /* Add the new addresses to the bind address list with
588 * use_as_src set to 0.
591 for (i = 0; i < addrcnt; i++) {
593 af = sctp_get_af_specific(addr->v4.sin_family);
594 memcpy(&saveaddr, addr, af->sockaddr_len);
595 retval = sctp_add_bind_addr(bp, &saveaddr,
596 SCTP_ADDR_NEW, GFP_ATOMIC);
597 addr_buf += af->sockaddr_len;
599 if (asoc->src_out_of_asoc_ok) {
600 struct sctp_transport *trans;
602 list_for_each_entry(trans,
603 &asoc->peer.transport_addr_list, transports) {
604 /* Clear the source and route cache */
605 dst_release(trans->dst);
606 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
607 2*asoc->pathmtu, 4380));
608 trans->ssthresh = asoc->peer.i.a_rwnd;
609 trans->rto = asoc->rto_initial;
610 sctp_max_rto(asoc, trans);
611 trans->rtt = trans->srtt = trans->rttvar = 0;
612 sctp_transport_route(trans, NULL,
613 sctp_sk(asoc->base.sk));
616 retval = sctp_send_asconf(asoc, chunk);
623 /* Remove a list of addresses from bind addresses list. Do not remove the
626 * Basically run through each address specified in the addrs/addrcnt
627 * array/length pair, determine if it is IPv6 or IPv4 and call
628 * sctp_del_bind() on it.
630 * If any of them fails, then the operation will be reversed and the
631 * ones that were removed will be added back.
633 * At least one address has to be left; if only one address is
634 * available, the operation will return -EBUSY.
636 * Only sctp_setsockopt_bindx() is supposed to call this function.
638 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
640 struct sctp_sock *sp = sctp_sk(sk);
641 struct sctp_endpoint *ep = sp->ep;
643 struct sctp_bind_addr *bp = &ep->base.bind_addr;
646 union sctp_addr *sa_addr;
649 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
653 for (cnt = 0; cnt < addrcnt; cnt++) {
654 /* If the bind address list is empty or if there is only one
655 * bind address, there is nothing more to be removed (we need
656 * at least one address here).
658 if (list_empty(&bp->address_list) ||
659 (sctp_list_single_entry(&bp->address_list))) {
665 af = sctp_get_af_specific(sa_addr->sa.sa_family);
671 if (!af->addr_valid(sa_addr, sp, NULL)) {
672 retval = -EADDRNOTAVAIL;
676 if (sa_addr->v4.sin_port &&
677 sa_addr->v4.sin_port != htons(bp->port)) {
682 if (!sa_addr->v4.sin_port)
683 sa_addr->v4.sin_port = htons(bp->port);
685 /* FIXME - There is probably a need to check if sk->sk_saddr and
686 * sk->sk_rcv_addr are currently set to one of the addresses to
687 * be removed. This is something which needs to be looked into
688 * when we are fixing the outstanding issues with multi-homing
689 * socket routing and failover schemes. Refer to comments in
690 * sctp_do_bind(). -daisy
692 retval = sctp_del_bind_addr(bp, sa_addr);
694 addr_buf += af->sockaddr_len;
697 /* Failed. Add the ones that has been removed back */
699 sctp_bindx_add(sk, addrs, cnt);
707 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
708 * the associations that are part of the endpoint indicating that a list of
709 * local addresses are removed from the endpoint.
711 * If any of the addresses is already in the bind address list of the
712 * association, we do not send the chunk for that association. But it will not
713 * affect other associations.
715 * Only sctp_setsockopt_bindx() is supposed to call this function.
717 static int sctp_send_asconf_del_ip(struct sock *sk,
718 struct sockaddr *addrs,
721 struct net *net = sock_net(sk);
722 struct sctp_sock *sp;
723 struct sctp_endpoint *ep;
724 struct sctp_association *asoc;
725 struct sctp_transport *transport;
726 struct sctp_bind_addr *bp;
727 struct sctp_chunk *chunk;
728 union sctp_addr *laddr;
731 struct sctp_sockaddr_entry *saddr;
737 if (!net->sctp.addip_enable)
743 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
744 __func__, sk, addrs, addrcnt);
746 list_for_each_entry(asoc, &ep->asocs, asocs) {
748 if (!asoc->peer.asconf_capable)
751 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
754 if (!sctp_state(asoc, ESTABLISHED))
757 /* Check if any address in the packed array of addresses is
758 * not present in the bind address list of the association.
759 * If so, do not send the asconf chunk to its peer, but
760 * continue with other associations.
763 for (i = 0; i < addrcnt; i++) {
765 af = sctp_get_af_specific(laddr->v4.sin_family);
771 if (!sctp_assoc_lookup_laddr(asoc, laddr))
774 addr_buf += af->sockaddr_len;
779 /* Find one address in the association's bind address list
780 * that is not in the packed array of addresses. This is to
781 * make sure that we do not delete all the addresses in the
784 bp = &asoc->base.bind_addr;
785 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
787 if ((laddr == NULL) && (addrcnt == 1)) {
788 if (asoc->asconf_addr_del_pending)
790 asoc->asconf_addr_del_pending =
791 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
792 if (asoc->asconf_addr_del_pending == NULL) {
796 asoc->asconf_addr_del_pending->sa.sa_family =
798 asoc->asconf_addr_del_pending->v4.sin_port =
800 if (addrs->sa_family == AF_INET) {
801 struct sockaddr_in *sin;
803 sin = (struct sockaddr_in *)addrs;
804 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
805 } else if (addrs->sa_family == AF_INET6) {
806 struct sockaddr_in6 *sin6;
808 sin6 = (struct sockaddr_in6 *)addrs;
809 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
811 SCTP_DEBUG_PRINTK_IPADDR("send_asconf_del_ip: keep the last address asoc: %p ",
812 " at %p\n", asoc, asoc->asconf_addr_del_pending,
813 asoc->asconf_addr_del_pending);
814 asoc->src_out_of_asoc_ok = 1;
819 /* We do not need RCU protection throughout this loop
820 * because this is done under a socket lock from the
823 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
831 /* Reset use_as_src flag for the addresses in the bind address
832 * list that are to be deleted.
835 for (i = 0; i < addrcnt; i++) {
837 af = sctp_get_af_specific(laddr->v4.sin_family);
838 list_for_each_entry(saddr, &bp->address_list, list) {
839 if (sctp_cmp_addr_exact(&saddr->a, laddr))
840 saddr->state = SCTP_ADDR_DEL;
842 addr_buf += af->sockaddr_len;
845 /* Update the route and saddr entries for all the transports
846 * as some of the addresses in the bind address list are
847 * about to be deleted and cannot be used as source addresses.
849 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
851 dst_release(transport->dst);
852 sctp_transport_route(transport, NULL,
853 sctp_sk(asoc->base.sk));
857 /* We don't need to transmit ASCONF */
859 retval = sctp_send_asconf(asoc, chunk);
865 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
866 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
868 struct sock *sk = sctp_opt2sk(sp);
869 union sctp_addr *addr;
872 /* It is safe to write port space in caller. */
874 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
875 af = sctp_get_af_specific(addr->sa.sa_family);
878 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
881 if (addrw->state == SCTP_ADDR_NEW)
882 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
884 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
887 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
890 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
893 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
894 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
897 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
898 * Section 3.1.2 for this usage.
900 * addrs is a pointer to an array of one or more socket addresses. Each
901 * address is contained in its appropriate structure (i.e. struct
902 * sockaddr_in or struct sockaddr_in6) the family of the address type
903 * must be used to distinguish the address length (note that this
904 * representation is termed a "packed array" of addresses). The caller
905 * specifies the number of addresses in the array with addrcnt.
907 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
908 * -1, and sets errno to the appropriate error code.
910 * For SCTP, the port given in each socket address must be the same, or
911 * sctp_bindx() will fail, setting errno to EINVAL.
913 * The flags parameter is formed from the bitwise OR of zero or more of
914 * the following currently defined flags:
916 * SCTP_BINDX_ADD_ADDR
918 * SCTP_BINDX_REM_ADDR
920 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
921 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
922 * addresses from the association. The two flags are mutually exclusive;
923 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
924 * not remove all addresses from an association; sctp_bindx() will
925 * reject such an attempt with EINVAL.
927 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
928 * additional addresses with an endpoint after calling bind(). Or use
929 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
930 * socket is associated with so that no new association accepted will be
931 * associated with those addresses. If the endpoint supports dynamic
932 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
933 * endpoint to send the appropriate message to the peer to change the
934 * peers address lists.
936 * Adding and removing addresses from a connected association is
937 * optional functionality. Implementations that do not support this
938 * functionality should return EOPNOTSUPP.
940 * Basically do nothing but copying the addresses from user to kernel
941 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
942 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
945 * We don't use copy_from_user() for optimization: we first do the
946 * sanity checks (buffer size -fast- and access check-healthy
947 * pointer); if all of those succeed, then we can alloc the memory
948 * (expensive operation) needed to copy the data to kernel. Then we do
949 * the copying without checking the user space area
950 * (__copy_from_user()).
952 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
955 * sk The sk of the socket
956 * addrs The pointer to the addresses in user land
957 * addrssize Size of the addrs buffer
958 * op Operation to perform (add or remove, see the flags of
961 * Returns 0 if ok, <0 errno code on error.
963 static int sctp_setsockopt_bindx(struct sock* sk,
964 struct sockaddr __user *addrs,
965 int addrs_size, int op)
967 struct sockaddr *kaddrs;
971 struct sockaddr *sa_addr;
975 SCTP_DEBUG_PRINTK("sctp_setsockopt_bindx: sk %p addrs %p"
976 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
978 if (unlikely(addrs_size <= 0))
981 /* Check the user passed a healthy pointer. */
982 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
985 /* Alloc space for the address array in kernel memory. */
986 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
987 if (unlikely(!kaddrs))
990 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
995 /* Walk through the addrs buffer and count the number of addresses. */
997 while (walk_size < addrs_size) {
998 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1004 af = sctp_get_af_specific(sa_addr->sa_family);
1006 /* If the address family is not supported or if this address
1007 * causes the address buffer to overflow return EINVAL.
1009 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1014 addr_buf += af->sockaddr_len;
1015 walk_size += af->sockaddr_len;
1020 case SCTP_BINDX_ADD_ADDR:
1021 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1024 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1027 case SCTP_BINDX_REM_ADDR:
1028 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1031 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1045 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1047 * Common routine for handling connect() and sctp_connectx().
1048 * Connect will come in with just a single address.
1050 static int __sctp_connect(struct sock* sk,
1051 struct sockaddr *kaddrs,
1053 sctp_assoc_t *assoc_id)
1055 struct net *net = sock_net(sk);
1056 struct sctp_sock *sp;
1057 struct sctp_endpoint *ep;
1058 struct sctp_association *asoc = NULL;
1059 struct sctp_association *asoc2;
1060 struct sctp_transport *transport;
1068 union sctp_addr *sa_addr = NULL;
1070 unsigned short port;
1071 unsigned int f_flags = 0;
1076 /* connect() cannot be done on a socket that is already in ESTABLISHED
1077 * state - UDP-style peeled off socket or a TCP-style socket that
1078 * is already connected.
1079 * It cannot be done even on a TCP-style listening socket.
1081 if (sctp_sstate(sk, ESTABLISHED) ||
1082 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1087 /* Walk through the addrs buffer and count the number of addresses. */
1089 while (walk_size < addrs_size) {
1090 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1096 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1098 /* If the address family is not supported or if this address
1099 * causes the address buffer to overflow return EINVAL.
1101 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1106 port = ntohs(sa_addr->v4.sin_port);
1108 /* Save current address so we can work with it */
1109 memcpy(&to, sa_addr, af->sockaddr_len);
1111 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1115 /* Make sure the destination port is correctly set
1118 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1123 /* Check if there already is a matching association on the
1124 * endpoint (other than the one created here).
1126 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1127 if (asoc2 && asoc2 != asoc) {
1128 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1135 /* If we could not find a matching association on the endpoint,
1136 * make sure that there is no peeled-off association matching
1137 * the peer address even on another socket.
1139 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1140 err = -EADDRNOTAVAIL;
1145 /* If a bind() or sctp_bindx() is not called prior to
1146 * an sctp_connectx() call, the system picks an
1147 * ephemeral port and will choose an address set
1148 * equivalent to binding with a wildcard address.
1150 if (!ep->base.bind_addr.port) {
1151 if (sctp_autobind(sk)) {
1157 * If an unprivileged user inherits a 1-many
1158 * style socket with open associations on a
1159 * privileged port, it MAY be permitted to
1160 * accept new associations, but it SHOULD NOT
1161 * be permitted to open new associations.
1163 if (ep->base.bind_addr.port < PROT_SOCK &&
1164 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1170 scope = sctp_scope(&to);
1171 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1177 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1185 /* Prime the peer's transport structures. */
1186 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1194 addr_buf += af->sockaddr_len;
1195 walk_size += af->sockaddr_len;
1198 /* In case the user of sctp_connectx() wants an association
1199 * id back, assign one now.
1202 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1207 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1212 /* Initialize sk's dport and daddr for getpeername() */
1213 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1214 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1215 af->to_sk_daddr(sa_addr, sk);
1218 /* in-kernel sockets don't generally have a file allocated to them
1219 * if all they do is call sock_create_kern().
1221 if (sk->sk_socket->file)
1222 f_flags = sk->sk_socket->file->f_flags;
1224 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1226 err = sctp_wait_for_connect(asoc, &timeo);
1227 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1228 *assoc_id = asoc->assoc_id;
1230 /* Don't free association on exit. */
1235 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1236 " kaddrs: %p err: %d\n",
1239 /* sctp_primitive_ASSOCIATE may have added this association
1240 * To the hash table, try to unhash it, just in case, its a noop
1241 * if it wasn't hashed so we're safe
1243 sctp_unhash_established(asoc);
1244 sctp_association_free(asoc);
1249 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1252 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1253 * sctp_assoc_t *asoc);
1255 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1256 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1257 * or IPv6 addresses.
1259 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1260 * Section 3.1.2 for this usage.
1262 * addrs is a pointer to an array of one or more socket addresses. Each
1263 * address is contained in its appropriate structure (i.e. struct
1264 * sockaddr_in or struct sockaddr_in6) the family of the address type
1265 * must be used to distengish the address length (note that this
1266 * representation is termed a "packed array" of addresses). The caller
1267 * specifies the number of addresses in the array with addrcnt.
1269 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1270 * the association id of the new association. On failure, sctp_connectx()
1271 * returns -1, and sets errno to the appropriate error code. The assoc_id
1272 * is not touched by the kernel.
1274 * For SCTP, the port given in each socket address must be the same, or
1275 * sctp_connectx() will fail, setting errno to EINVAL.
1277 * An application can use sctp_connectx to initiate an association with
1278 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1279 * allows a caller to specify multiple addresses at which a peer can be
1280 * reached. The way the SCTP stack uses the list of addresses to set up
1281 * the association is implementation dependent. This function only
1282 * specifies that the stack will try to make use of all the addresses in
1283 * the list when needed.
1285 * Note that the list of addresses passed in is only used for setting up
1286 * the association. It does not necessarily equal the set of addresses
1287 * the peer uses for the resulting association. If the caller wants to
1288 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1289 * retrieve them after the association has been set up.
1291 * Basically do nothing but copying the addresses from user to kernel
1292 * land and invoking either sctp_connectx(). This is used for tunneling
1293 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1295 * We don't use copy_from_user() for optimization: we first do the
1296 * sanity checks (buffer size -fast- and access check-healthy
1297 * pointer); if all of those succeed, then we can alloc the memory
1298 * (expensive operation) needed to copy the data to kernel. Then we do
1299 * the copying without checking the user space area
1300 * (__copy_from_user()).
1302 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1305 * sk The sk of the socket
1306 * addrs The pointer to the addresses in user land
1307 * addrssize Size of the addrs buffer
1309 * Returns >=0 if ok, <0 errno code on error.
1311 static int __sctp_setsockopt_connectx(struct sock* sk,
1312 struct sockaddr __user *addrs,
1314 sctp_assoc_t *assoc_id)
1317 struct sockaddr *kaddrs;
1319 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1320 __func__, sk, addrs, addrs_size);
1322 if (unlikely(addrs_size <= 0))
1325 /* Check the user passed a healthy pointer. */
1326 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1329 /* Alloc space for the address array in kernel memory. */
1330 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1331 if (unlikely(!kaddrs))
1334 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1337 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1346 * This is an older interface. It's kept for backward compatibility
1347 * to the option that doesn't provide association id.
1349 static int sctp_setsockopt_connectx_old(struct sock* sk,
1350 struct sockaddr __user *addrs,
1353 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1357 * New interface for the API. The since the API is done with a socket
1358 * option, to make it simple we feed back the association id is as a return
1359 * indication to the call. Error is always negative and association id is
1362 static int sctp_setsockopt_connectx(struct sock* sk,
1363 struct sockaddr __user *addrs,
1366 sctp_assoc_t assoc_id = 0;
1369 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1378 * New (hopefully final) interface for the API.
1379 * We use the sctp_getaddrs_old structure so that use-space library
1380 * can avoid any unnecessary allocations. The only defferent part
1381 * is that we store the actual length of the address buffer into the
1382 * addrs_num structure member. That way we can re-use the existing
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 if (len < sizeof(param))
1396 if (copy_from_user(¶m, optval, sizeof(param)))
1399 err = __sctp_setsockopt_connectx(sk,
1400 (struct sockaddr __user *)param.addrs,
1401 param.addr_num, &assoc_id);
1403 if (err == 0 || err == -EINPROGRESS) {
1404 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1406 if (put_user(sizeof(assoc_id), optlen))
1413 /* API 3.1.4 close() - UDP Style Syntax
1414 * Applications use close() to perform graceful shutdown (as described in
1415 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1416 * by a UDP-style socket.
1420 * ret = close(int sd);
1422 * sd - the socket descriptor of the associations to be closed.
1424 * To gracefully shutdown a specific association represented by the
1425 * UDP-style socket, an application should use the sendmsg() call,
1426 * passing no user data, but including the appropriate flag in the
1427 * ancillary data (see Section xxxx).
1429 * If sd in the close() call is a branched-off socket representing only
1430 * one association, the shutdown is performed on that association only.
1432 * 4.1.6 close() - TCP Style Syntax
1434 * Applications use close() to gracefully close down an association.
1438 * int close(int sd);
1440 * sd - the socket descriptor of the association to be closed.
1442 * After an application calls close() on a socket descriptor, no further
1443 * socket operations will succeed on that descriptor.
1445 * API 7.1.4 SO_LINGER
1447 * An application using the TCP-style socket can use this option to
1448 * perform the SCTP ABORT primitive. The linger option structure is:
1451 * int l_onoff; // option on/off
1452 * int l_linger; // linger time
1455 * To enable the option, set l_onoff to 1. If the l_linger value is set
1456 * to 0, calling close() is the same as the ABORT primitive. If the
1457 * value is set to a negative value, the setsockopt() call will return
1458 * an error. If the value is set to a positive value linger_time, the
1459 * close() can be blocked for at most linger_time ms. If the graceful
1460 * shutdown phase does not finish during this period, close() will
1461 * return but the graceful shutdown phase continues in the system.
1463 static void sctp_close(struct sock *sk, long timeout)
1465 struct net *net = sock_net(sk);
1466 struct sctp_endpoint *ep;
1467 struct sctp_association *asoc;
1468 struct list_head *pos, *temp;
1469 unsigned int data_was_unread;
1471 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1474 sk->sk_shutdown = SHUTDOWN_MASK;
1475 sk->sk_state = SCTP_SS_CLOSING;
1477 ep = sctp_sk(sk)->ep;
1479 /* Clean up any skbs sitting on the receive queue. */
1480 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1481 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1483 /* Walk all associations on an endpoint. */
1484 list_for_each_safe(pos, temp, &ep->asocs) {
1485 asoc = list_entry(pos, struct sctp_association, asocs);
1487 if (sctp_style(sk, TCP)) {
1488 /* A closed association can still be in the list if
1489 * it belongs to a TCP-style listening socket that is
1490 * not yet accepted. If so, free it. If not, send an
1491 * ABORT or SHUTDOWN based on the linger options.
1493 if (sctp_state(asoc, CLOSED)) {
1494 sctp_unhash_established(asoc);
1495 sctp_association_free(asoc);
1500 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1501 !skb_queue_empty(&asoc->ulpq.reasm) ||
1502 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1503 struct sctp_chunk *chunk;
1505 chunk = sctp_make_abort_user(asoc, NULL, 0);
1507 sctp_primitive_ABORT(net, asoc, chunk);
1509 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1512 /* On a TCP-style socket, block for at most linger_time if set. */
1513 if (sctp_style(sk, TCP) && timeout)
1514 sctp_wait_for_close(sk, timeout);
1516 /* This will run the backlog queue. */
1517 sctp_release_sock(sk);
1519 /* Supposedly, no process has access to the socket, but
1520 * the net layers still may.
1522 sctp_local_bh_disable();
1523 sctp_bh_lock_sock(sk);
1525 /* Hold the sock, since sk_common_release() will put sock_put()
1526 * and we have just a little more cleanup.
1529 sk_common_release(sk);
1531 sctp_bh_unlock_sock(sk);
1532 sctp_local_bh_enable();
1536 SCTP_DBG_OBJCNT_DEC(sock);
1539 /* Handle EPIPE error. */
1540 static int sctp_error(struct sock *sk, int flags, int err)
1543 err = sock_error(sk) ? : -EPIPE;
1544 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1545 send_sig(SIGPIPE, current, 0);
1549 /* API 3.1.3 sendmsg() - UDP Style Syntax
1551 * An application uses sendmsg() and recvmsg() calls to transmit data to
1552 * and receive data from its peer.
1554 * ssize_t sendmsg(int socket, const struct msghdr *message,
1557 * socket - the socket descriptor of the endpoint.
1558 * message - pointer to the msghdr structure which contains a single
1559 * user message and possibly some ancillary data.
1561 * See Section 5 for complete description of the data
1564 * flags - flags sent or received with the user message, see Section
1565 * 5 for complete description of the flags.
1567 * Note: This function could use a rewrite especially when explicit
1568 * connect support comes in.
1570 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1572 static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1574 static int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1575 struct msghdr *msg, size_t msg_len)
1577 struct net *net = sock_net(sk);
1578 struct sctp_sock *sp;
1579 struct sctp_endpoint *ep;
1580 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1581 struct sctp_transport *transport, *chunk_tp;
1582 struct sctp_chunk *chunk;
1584 struct sockaddr *msg_name = NULL;
1585 struct sctp_sndrcvinfo default_sinfo;
1586 struct sctp_sndrcvinfo *sinfo;
1587 struct sctp_initmsg *sinit;
1588 sctp_assoc_t associd = 0;
1589 sctp_cmsgs_t cmsgs = { NULL };
1593 __u16 sinfo_flags = 0;
1594 struct sctp_datamsg *datamsg;
1595 int msg_flags = msg->msg_flags;
1597 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1604 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1606 /* We cannot send a message over a TCP-style listening socket. */
1607 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1612 /* Parse out the SCTP CMSGs. */
1613 err = sctp_msghdr_parse(msg, &cmsgs);
1616 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1620 /* Fetch the destination address for this packet. This
1621 * address only selects the association--it is not necessarily
1622 * the address we will send to.
1623 * For a peeled-off socket, msg_name is ignored.
1625 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1626 int msg_namelen = msg->msg_namelen;
1628 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1633 if (msg_namelen > sizeof(to))
1634 msg_namelen = sizeof(to);
1635 memcpy(&to, msg->msg_name, msg_namelen);
1636 msg_name = msg->msg_name;
1642 /* Did the user specify SNDRCVINFO? */
1644 sinfo_flags = sinfo->sinfo_flags;
1645 associd = sinfo->sinfo_assoc_id;
1648 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1649 msg_len, sinfo_flags);
1651 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1652 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1657 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1658 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1659 * If SCTP_ABORT is set, the message length could be non zero with
1660 * the msg_iov set to the user abort reason.
1662 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1663 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1668 /* If SCTP_ADDR_OVER is set, there must be an address
1669 * specified in msg_name.
1671 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1678 SCTP_DEBUG_PRINTK("About to look up association.\n");
1682 /* If a msg_name has been specified, assume this is to be used. */
1684 /* Look for a matching association on the endpoint. */
1685 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1687 /* If we could not find a matching association on the
1688 * endpoint, make sure that it is not a TCP-style
1689 * socket that already has an association or there is
1690 * no peeled-off association on another socket.
1692 if ((sctp_style(sk, TCP) &&
1693 sctp_sstate(sk, ESTABLISHED)) ||
1694 sctp_endpoint_is_peeled_off(ep, &to)) {
1695 err = -EADDRNOTAVAIL;
1700 asoc = sctp_id2assoc(sk, associd);
1708 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1710 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1711 * socket that has an association in CLOSED state. This can
1712 * happen when an accepted socket has an association that is
1715 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1720 if (sinfo_flags & SCTP_EOF) {
1721 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1723 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1727 if (sinfo_flags & SCTP_ABORT) {
1729 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1735 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1736 sctp_primitive_ABORT(net, asoc, chunk);
1742 /* Do we need to create the association? */
1744 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1746 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1751 /* Check for invalid stream against the stream counts,
1752 * either the default or the user specified stream counts.
1755 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1756 /* Check against the defaults. */
1757 if (sinfo->sinfo_stream >=
1758 sp->initmsg.sinit_num_ostreams) {
1763 /* Check against the requested. */
1764 if (sinfo->sinfo_stream >=
1765 sinit->sinit_num_ostreams) {
1773 * API 3.1.2 bind() - UDP Style Syntax
1774 * If a bind() or sctp_bindx() is not called prior to a
1775 * sendmsg() call that initiates a new association, the
1776 * system picks an ephemeral port and will choose an address
1777 * set equivalent to binding with a wildcard address.
1779 if (!ep->base.bind_addr.port) {
1780 if (sctp_autobind(sk)) {
1786 * If an unprivileged user inherits a one-to-many
1787 * style socket with open associations on a privileged
1788 * port, it MAY be permitted to accept new associations,
1789 * but it SHOULD NOT be permitted to open new
1792 if (ep->base.bind_addr.port < PROT_SOCK &&
1793 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1799 scope = sctp_scope(&to);
1800 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1806 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1812 /* If the SCTP_INIT ancillary data is specified, set all
1813 * the association init values accordingly.
1816 if (sinit->sinit_num_ostreams) {
1817 asoc->c.sinit_num_ostreams =
1818 sinit->sinit_num_ostreams;
1820 if (sinit->sinit_max_instreams) {
1821 asoc->c.sinit_max_instreams =
1822 sinit->sinit_max_instreams;
1824 if (sinit->sinit_max_attempts) {
1825 asoc->max_init_attempts
1826 = sinit->sinit_max_attempts;
1828 if (sinit->sinit_max_init_timeo) {
1829 asoc->max_init_timeo =
1830 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1834 /* Prime the peer's transport structures. */
1835 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1842 /* ASSERT: we have a valid association at this point. */
1843 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1846 /* If the user didn't specify SNDRCVINFO, make up one with
1849 memset(&default_sinfo, 0, sizeof(default_sinfo));
1850 default_sinfo.sinfo_stream = asoc->default_stream;
1851 default_sinfo.sinfo_flags = asoc->default_flags;
1852 default_sinfo.sinfo_ppid = asoc->default_ppid;
1853 default_sinfo.sinfo_context = asoc->default_context;
1854 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1855 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1856 sinfo = &default_sinfo;
1859 /* API 7.1.7, the sndbuf size per association bounds the
1860 * maximum size of data that can be sent in a single send call.
1862 if (msg_len > sk->sk_sndbuf) {
1867 if (asoc->pmtu_pending)
1868 sctp_assoc_pending_pmtu(sk, asoc);
1870 /* If fragmentation is disabled and the message length exceeds the
1871 * association fragmentation point, return EMSGSIZE. The I-D
1872 * does not specify what this error is, but this looks like
1875 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1880 /* Check for invalid stream. */
1881 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1886 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1887 if (!sctp_wspace(asoc)) {
1888 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1893 /* If an address is passed with the sendto/sendmsg call, it is used
1894 * to override the primary destination address in the TCP model, or
1895 * when SCTP_ADDR_OVER flag is set in the UDP model.
1897 if ((sctp_style(sk, TCP) && msg_name) ||
1898 (sinfo_flags & SCTP_ADDR_OVER)) {
1899 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1907 /* Auto-connect, if we aren't connected already. */
1908 if (sctp_state(asoc, CLOSED)) {
1909 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1912 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1915 /* Break the message into multiple chunks of maximum size. */
1916 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1917 if (IS_ERR(datamsg)) {
1918 err = PTR_ERR(datamsg);
1922 /* Now send the (possibly) fragmented message. */
1923 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1924 sctp_chunk_hold(chunk);
1926 /* Do accounting for the write space. */
1927 sctp_set_owner_w(chunk);
1929 chunk->transport = chunk_tp;
1932 /* Send it to the lower layers. Note: all chunks
1933 * must either fail or succeed. The lower layer
1934 * works that way today. Keep it that way or this
1937 err = sctp_primitive_SEND(net, asoc, datamsg);
1938 /* Did the lower layer accept the chunk? */
1940 sctp_datamsg_free(datamsg);
1942 sctp_datamsg_put(datamsg);
1944 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1951 /* If we are already past ASSOCIATE, the lower
1952 * layers are responsible for association cleanup.
1958 sctp_unhash_established(asoc);
1959 sctp_association_free(asoc);
1962 sctp_release_sock(sk);
1965 return sctp_error(sk, msg_flags, err);
1972 err = sock_error(sk);
1982 /* This is an extended version of skb_pull() that removes the data from the
1983 * start of a skb even when data is spread across the list of skb's in the
1984 * frag_list. len specifies the total amount of data that needs to be removed.
1985 * when 'len' bytes could be removed from the skb, it returns 0.
1986 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1987 * could not be removed.
1989 static int sctp_skb_pull(struct sk_buff *skb, int len)
1991 struct sk_buff *list;
1992 int skb_len = skb_headlen(skb);
1995 if (len <= skb_len) {
1996 __skb_pull(skb, len);
2000 __skb_pull(skb, skb_len);
2002 skb_walk_frags(skb, list) {
2003 rlen = sctp_skb_pull(list, len);
2004 skb->len -= (len-rlen);
2005 skb->data_len -= (len-rlen);
2016 /* API 3.1.3 recvmsg() - UDP Style Syntax
2018 * ssize_t recvmsg(int socket, struct msghdr *message,
2021 * socket - the socket descriptor of the endpoint.
2022 * message - pointer to the msghdr structure which contains a single
2023 * user message and possibly some ancillary data.
2025 * See Section 5 for complete description of the data
2028 * flags - flags sent or received with the user message, see Section
2029 * 5 for complete description of the flags.
2031 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2033 static int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2034 struct msghdr *msg, size_t len, int noblock,
2035 int flags, int *addr_len)
2037 struct sctp_ulpevent *event = NULL;
2038 struct sctp_sock *sp = sctp_sk(sk);
2039 struct sk_buff *skb;
2044 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
2045 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
2046 "len", len, "knoblauch", noblock,
2047 "flags", flags, "addr_len", addr_len);
2051 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2056 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2060 /* Get the total length of the skb including any skb's in the
2069 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2071 event = sctp_skb2event(skb);
2076 sock_recv_ts_and_drops(msg, sk, skb);
2077 if (sctp_ulpevent_is_notification(event)) {
2078 msg->msg_flags |= MSG_NOTIFICATION;
2079 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2081 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2084 /* Check if we allow SCTP_SNDRCVINFO. */
2085 if (sp->subscribe.sctp_data_io_event)
2086 sctp_ulpevent_read_sndrcvinfo(event, msg);
2088 /* FIXME: we should be calling IP/IPv6 layers. */
2089 if (sk->sk_protinfo.af_inet.cmsg_flags)
2090 ip_cmsg_recv(msg, skb);
2095 /* If skb's length exceeds the user's buffer, update the skb and
2096 * push it back to the receive_queue so that the next call to
2097 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2099 if (skb_len > copied) {
2100 msg->msg_flags &= ~MSG_EOR;
2101 if (flags & MSG_PEEK)
2103 sctp_skb_pull(skb, copied);
2104 skb_queue_head(&sk->sk_receive_queue, skb);
2106 /* When only partial message is copied to the user, increase
2107 * rwnd by that amount. If all the data in the skb is read,
2108 * rwnd is updated when the event is freed.
2110 if (!sctp_ulpevent_is_notification(event))
2111 sctp_assoc_rwnd_increase(event->asoc, copied);
2113 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2114 (event->msg_flags & MSG_EOR))
2115 msg->msg_flags |= MSG_EOR;
2117 msg->msg_flags &= ~MSG_EOR;
2120 if (flags & MSG_PEEK) {
2121 /* Release the skb reference acquired after peeking the skb in
2122 * sctp_skb_recv_datagram().
2126 /* Free the event which includes releasing the reference to
2127 * the owner of the skb, freeing the skb and updating the
2130 sctp_ulpevent_free(event);
2133 sctp_release_sock(sk);
2137 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2139 * This option is a on/off flag. If enabled no SCTP message
2140 * fragmentation will be performed. Instead if a message being sent
2141 * exceeds the current PMTU size, the message will NOT be sent and
2142 * instead a error will be indicated to the user.
2144 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2145 char __user *optval,
2146 unsigned int optlen)
2150 if (optlen < sizeof(int))
2153 if (get_user(val, (int __user *)optval))
2156 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2161 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2162 unsigned int optlen)
2164 struct sctp_association *asoc;
2165 struct sctp_ulpevent *event;
2167 if (optlen > sizeof(struct sctp_event_subscribe))
2169 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2173 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2174 * if there is no data to be sent or retransmit, the stack will
2175 * immediately send up this notification.
2177 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2178 &sctp_sk(sk)->subscribe)) {
2179 asoc = sctp_id2assoc(sk, 0);
2181 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2182 event = sctp_ulpevent_make_sender_dry_event(asoc,
2187 sctp_ulpq_tail_event(&asoc->ulpq, event);
2194 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2196 * This socket option is applicable to the UDP-style socket only. When
2197 * set it will cause associations that are idle for more than the
2198 * specified number of seconds to automatically close. An association
2199 * being idle is defined an association that has NOT sent or received
2200 * user data. The special value of '0' indicates that no automatic
2201 * close of any associations should be performed. The option expects an
2202 * integer defining the number of seconds of idle time before an
2203 * association is closed.
2205 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2206 unsigned int optlen)
2208 struct sctp_sock *sp = sctp_sk(sk);
2210 /* Applicable to UDP-style socket only */
2211 if (sctp_style(sk, TCP))
2213 if (optlen != sizeof(int))
2215 if (copy_from_user(&sp->autoclose, optval, optlen))
2221 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2223 * Applications can enable or disable heartbeats for any peer address of
2224 * an association, modify an address's heartbeat interval, force a
2225 * heartbeat to be sent immediately, and adjust the address's maximum
2226 * number of retransmissions sent before an address is considered
2227 * unreachable. The following structure is used to access and modify an
2228 * address's parameters:
2230 * struct sctp_paddrparams {
2231 * sctp_assoc_t spp_assoc_id;
2232 * struct sockaddr_storage spp_address;
2233 * uint32_t spp_hbinterval;
2234 * uint16_t spp_pathmaxrxt;
2235 * uint32_t spp_pathmtu;
2236 * uint32_t spp_sackdelay;
2237 * uint32_t spp_flags;
2240 * spp_assoc_id - (one-to-many style socket) This is filled in the
2241 * application, and identifies the association for
2243 * spp_address - This specifies which address is of interest.
2244 * spp_hbinterval - This contains the value of the heartbeat interval,
2245 * in milliseconds. If a value of zero
2246 * is present in this field then no changes are to
2247 * be made to this parameter.
2248 * spp_pathmaxrxt - This contains the maximum number of
2249 * retransmissions before this address shall be
2250 * considered unreachable. If a value of zero
2251 * is present in this field then no changes are to
2252 * be made to this parameter.
2253 * spp_pathmtu - When Path MTU discovery is disabled the value
2254 * specified here will be the "fixed" path mtu.
2255 * Note that if the spp_address field is empty
2256 * then all associations on this address will
2257 * have this fixed path mtu set upon them.
2259 * spp_sackdelay - When delayed sack is enabled, this value specifies
2260 * the number of milliseconds that sacks will be delayed
2261 * for. This value will apply to all addresses of an
2262 * association if the spp_address field is empty. Note
2263 * also, that if delayed sack is enabled and this
2264 * value is set to 0, no change is made to the last
2265 * recorded delayed sack timer value.
2267 * spp_flags - These flags are used to control various features
2268 * on an association. The flag field may contain
2269 * zero or more of the following options.
2271 * SPP_HB_ENABLE - Enable heartbeats on the
2272 * specified address. Note that if the address
2273 * field is empty all addresses for the association
2274 * have heartbeats enabled upon them.
2276 * SPP_HB_DISABLE - Disable heartbeats on the
2277 * speicifed address. Note that if the address
2278 * field is empty all addresses for the association
2279 * will have their heartbeats disabled. Note also
2280 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2281 * mutually exclusive, only one of these two should
2282 * be specified. Enabling both fields will have
2283 * undetermined results.
2285 * SPP_HB_DEMAND - Request a user initiated heartbeat
2286 * to be made immediately.
2288 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2289 * heartbeat delayis to be set to the value of 0
2292 * SPP_PMTUD_ENABLE - This field will enable PMTU
2293 * discovery upon the specified address. Note that
2294 * if the address feild is empty then all addresses
2295 * on the association are effected.
2297 * SPP_PMTUD_DISABLE - This field will disable PMTU
2298 * discovery upon the specified address. Note that
2299 * if the address feild is empty then all addresses
2300 * on the association are effected. Not also that
2301 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2302 * exclusive. Enabling both will have undetermined
2305 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2306 * on delayed sack. The time specified in spp_sackdelay
2307 * is used to specify the sack delay for this address. Note
2308 * that if spp_address is empty then all addresses will
2309 * enable delayed sack and take on the sack delay
2310 * value specified in spp_sackdelay.
2311 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2312 * off delayed sack. If the spp_address field is blank then
2313 * delayed sack is disabled for the entire association. Note
2314 * also that this field is mutually exclusive to
2315 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2318 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2319 struct sctp_transport *trans,
2320 struct sctp_association *asoc,
2321 struct sctp_sock *sp,
2324 int sackdelay_change)
2328 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2329 struct net *net = sock_net(trans->asoc->base.sk);
2331 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2336 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2337 * this field is ignored. Note also that a value of zero indicates
2338 * the current setting should be left unchanged.
2340 if (params->spp_flags & SPP_HB_ENABLE) {
2342 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2343 * set. This lets us use 0 value when this flag
2346 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2347 params->spp_hbinterval = 0;
2349 if (params->spp_hbinterval ||
2350 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2353 msecs_to_jiffies(params->spp_hbinterval);
2356 msecs_to_jiffies(params->spp_hbinterval);
2358 sp->hbinterval = params->spp_hbinterval;
2365 trans->param_flags =
2366 (trans->param_flags & ~SPP_HB) | hb_change;
2369 (asoc->param_flags & ~SPP_HB) | hb_change;
2372 (sp->param_flags & ~SPP_HB) | hb_change;
2376 /* When Path MTU discovery is disabled the value specified here will
2377 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2378 * include the flag SPP_PMTUD_DISABLE for this field to have any
2381 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2383 trans->pathmtu = params->spp_pathmtu;
2384 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2386 asoc->pathmtu = params->spp_pathmtu;
2387 sctp_frag_point(asoc, params->spp_pathmtu);
2389 sp->pathmtu = params->spp_pathmtu;
2395 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2396 (params->spp_flags & SPP_PMTUD_ENABLE);
2397 trans->param_flags =
2398 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2400 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2401 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2405 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2408 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2412 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2413 * value of this field is ignored. Note also that a value of zero
2414 * indicates the current setting should be left unchanged.
2416 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2419 msecs_to_jiffies(params->spp_sackdelay);
2422 msecs_to_jiffies(params->spp_sackdelay);
2424 sp->sackdelay = params->spp_sackdelay;
2428 if (sackdelay_change) {
2430 trans->param_flags =
2431 (trans->param_flags & ~SPP_SACKDELAY) |
2435 (asoc->param_flags & ~SPP_SACKDELAY) |
2439 (sp->param_flags & ~SPP_SACKDELAY) |
2444 /* Note that a value of zero indicates the current setting should be
2447 if (params->spp_pathmaxrxt) {
2449 trans->pathmaxrxt = params->spp_pathmaxrxt;
2451 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2453 sp->pathmaxrxt = params->spp_pathmaxrxt;
2460 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2461 char __user *optval,
2462 unsigned int optlen)
2464 struct sctp_paddrparams params;
2465 struct sctp_transport *trans = NULL;
2466 struct sctp_association *asoc = NULL;
2467 struct sctp_sock *sp = sctp_sk(sk);
2469 int hb_change, pmtud_change, sackdelay_change;
2471 if (optlen != sizeof(struct sctp_paddrparams))
2474 if (copy_from_user(¶ms, optval, optlen))
2477 /* Validate flags and value parameters. */
2478 hb_change = params.spp_flags & SPP_HB;
2479 pmtud_change = params.spp_flags & SPP_PMTUD;
2480 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2482 if (hb_change == SPP_HB ||
2483 pmtud_change == SPP_PMTUD ||
2484 sackdelay_change == SPP_SACKDELAY ||
2485 params.spp_sackdelay > 500 ||
2486 (params.spp_pathmtu &&
2487 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2490 /* If an address other than INADDR_ANY is specified, and
2491 * no transport is found, then the request is invalid.
2493 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
2494 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2495 params.spp_assoc_id);
2500 /* Get association, if assoc_id != 0 and the socket is a one
2501 * to many style socket, and an association was not found, then
2502 * the id was invalid.
2504 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2505 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2508 /* Heartbeat demand can only be sent on a transport or
2509 * association, but not a socket.
2511 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2514 /* Process parameters. */
2515 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2516 hb_change, pmtud_change,
2522 /* If changes are for association, also apply parameters to each
2525 if (!trans && asoc) {
2526 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2528 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2529 hb_change, pmtud_change,
2538 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2540 * This option will effect the way delayed acks are performed. This
2541 * option allows you to get or set the delayed ack time, in
2542 * milliseconds. It also allows changing the delayed ack frequency.
2543 * Changing the frequency to 1 disables the delayed sack algorithm. If
2544 * the assoc_id is 0, then this sets or gets the endpoints default
2545 * values. If the assoc_id field is non-zero, then the set or get
2546 * effects the specified association for the one to many model (the
2547 * assoc_id field is ignored by the one to one model). Note that if
2548 * sack_delay or sack_freq are 0 when setting this option, then the
2549 * current values will remain unchanged.
2551 * struct sctp_sack_info {
2552 * sctp_assoc_t sack_assoc_id;
2553 * uint32_t sack_delay;
2554 * uint32_t sack_freq;
2557 * sack_assoc_id - This parameter, indicates which association the user
2558 * is performing an action upon. Note that if this field's value is
2559 * zero then the endpoints default value is changed (effecting future
2560 * associations only).
2562 * sack_delay - This parameter contains the number of milliseconds that
2563 * the user is requesting the delayed ACK timer be set to. Note that
2564 * this value is defined in the standard to be between 200 and 500
2567 * sack_freq - This parameter contains the number of packets that must
2568 * be received before a sack is sent without waiting for the delay
2569 * timer to expire. The default value for this is 2, setting this
2570 * value to 1 will disable the delayed sack algorithm.
2573 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2574 char __user *optval, unsigned int optlen)
2576 struct sctp_sack_info params;
2577 struct sctp_transport *trans = NULL;
2578 struct sctp_association *asoc = NULL;
2579 struct sctp_sock *sp = sctp_sk(sk);
2581 if (optlen == sizeof(struct sctp_sack_info)) {
2582 if (copy_from_user(¶ms, optval, optlen))
2585 if (params.sack_delay == 0 && params.sack_freq == 0)
2587 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2588 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2589 pr_warn("Use struct sctp_sack_info instead\n");
2590 if (copy_from_user(¶ms, optval, optlen))
2593 if (params.sack_delay == 0)
2594 params.sack_freq = 1;
2596 params.sack_freq = 0;
2600 /* Validate value parameter. */
2601 if (params.sack_delay > 500)
2604 /* Get association, if sack_assoc_id != 0 and the socket is a one
2605 * to many style socket, and an association was not found, then
2606 * the id was invalid.
2608 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2609 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2612 if (params.sack_delay) {
2615 msecs_to_jiffies(params.sack_delay);
2617 (asoc->param_flags & ~SPP_SACKDELAY) |
2618 SPP_SACKDELAY_ENABLE;
2620 sp->sackdelay = params.sack_delay;
2622 (sp->param_flags & ~SPP_SACKDELAY) |
2623 SPP_SACKDELAY_ENABLE;
2627 if (params.sack_freq == 1) {
2630 (asoc->param_flags & ~SPP_SACKDELAY) |
2631 SPP_SACKDELAY_DISABLE;
2634 (sp->param_flags & ~SPP_SACKDELAY) |
2635 SPP_SACKDELAY_DISABLE;
2637 } else if (params.sack_freq > 1) {
2639 asoc->sackfreq = params.sack_freq;
2641 (asoc->param_flags & ~SPP_SACKDELAY) |
2642 SPP_SACKDELAY_ENABLE;
2644 sp->sackfreq = params.sack_freq;
2646 (sp->param_flags & ~SPP_SACKDELAY) |
2647 SPP_SACKDELAY_ENABLE;
2651 /* If change is for association, also apply to each transport. */
2653 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2655 if (params.sack_delay) {
2657 msecs_to_jiffies(params.sack_delay);
2658 trans->param_flags =
2659 (trans->param_flags & ~SPP_SACKDELAY) |
2660 SPP_SACKDELAY_ENABLE;
2662 if (params.sack_freq == 1) {
2663 trans->param_flags =
2664 (trans->param_flags & ~SPP_SACKDELAY) |
2665 SPP_SACKDELAY_DISABLE;
2666 } else if (params.sack_freq > 1) {
2667 trans->sackfreq = params.sack_freq;
2668 trans->param_flags =
2669 (trans->param_flags & ~SPP_SACKDELAY) |
2670 SPP_SACKDELAY_ENABLE;
2678 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2680 * Applications can specify protocol parameters for the default association
2681 * initialization. The option name argument to setsockopt() and getsockopt()
2684 * Setting initialization parameters is effective only on an unconnected
2685 * socket (for UDP-style sockets only future associations are effected
2686 * by the change). With TCP-style sockets, this option is inherited by
2687 * sockets derived from a listener socket.
2689 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2691 struct sctp_initmsg sinit;
2692 struct sctp_sock *sp = sctp_sk(sk);
2694 if (optlen != sizeof(struct sctp_initmsg))
2696 if (copy_from_user(&sinit, optval, optlen))
2699 if (sinit.sinit_num_ostreams)
2700 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2701 if (sinit.sinit_max_instreams)
2702 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2703 if (sinit.sinit_max_attempts)
2704 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2705 if (sinit.sinit_max_init_timeo)
2706 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2712 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2714 * Applications that wish to use the sendto() system call may wish to
2715 * specify a default set of parameters that would normally be supplied
2716 * through the inclusion of ancillary data. This socket option allows
2717 * such an application to set the default sctp_sndrcvinfo structure.
2718 * The application that wishes to use this socket option simply passes
2719 * in to this call the sctp_sndrcvinfo structure defined in Section
2720 * 5.2.2) The input parameters accepted by this call include
2721 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2722 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2723 * to this call if the caller is using the UDP model.
2725 static int sctp_setsockopt_default_send_param(struct sock *sk,
2726 char __user *optval,
2727 unsigned int optlen)
2729 struct sctp_sndrcvinfo info;
2730 struct sctp_association *asoc;
2731 struct sctp_sock *sp = sctp_sk(sk);
2733 if (optlen != sizeof(struct sctp_sndrcvinfo))
2735 if (copy_from_user(&info, optval, optlen))
2738 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2739 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2743 asoc->default_stream = info.sinfo_stream;
2744 asoc->default_flags = info.sinfo_flags;
2745 asoc->default_ppid = info.sinfo_ppid;
2746 asoc->default_context = info.sinfo_context;
2747 asoc->default_timetolive = info.sinfo_timetolive;
2749 sp->default_stream = info.sinfo_stream;
2750 sp->default_flags = info.sinfo_flags;
2751 sp->default_ppid = info.sinfo_ppid;
2752 sp->default_context = info.sinfo_context;
2753 sp->default_timetolive = info.sinfo_timetolive;
2759 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2761 * Requests that the local SCTP stack use the enclosed peer address as
2762 * the association primary. The enclosed address must be one of the
2763 * association peer's addresses.
2765 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2766 unsigned int optlen)
2768 struct sctp_prim prim;
2769 struct sctp_transport *trans;
2771 if (optlen != sizeof(struct sctp_prim))
2774 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2777 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2781 sctp_assoc_set_primary(trans->asoc, trans);
2787 * 7.1.5 SCTP_NODELAY
2789 * Turn on/off any Nagle-like algorithm. This means that packets are
2790 * generally sent as soon as possible and no unnecessary delays are
2791 * introduced, at the cost of more packets in the network. Expects an
2792 * integer boolean flag.
2794 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2795 unsigned int optlen)
2799 if (optlen < sizeof(int))
2801 if (get_user(val, (int __user *)optval))
2804 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2810 * 7.1.1 SCTP_RTOINFO
2812 * The protocol parameters used to initialize and bound retransmission
2813 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2814 * and modify these parameters.
2815 * All parameters are time values, in milliseconds. A value of 0, when
2816 * modifying the parameters, indicates that the current value should not
2820 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2822 struct sctp_rtoinfo rtoinfo;
2823 struct sctp_association *asoc;
2825 if (optlen != sizeof (struct sctp_rtoinfo))
2828 if (copy_from_user(&rtoinfo, optval, optlen))
2831 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2833 /* Set the values to the specific association */
2834 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2838 if (rtoinfo.srto_initial != 0)
2840 msecs_to_jiffies(rtoinfo.srto_initial);
2841 if (rtoinfo.srto_max != 0)
2842 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2843 if (rtoinfo.srto_min != 0)
2844 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2846 /* If there is no association or the association-id = 0
2847 * set the values to the endpoint.
2849 struct sctp_sock *sp = sctp_sk(sk);
2851 if (rtoinfo.srto_initial != 0)
2852 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2853 if (rtoinfo.srto_max != 0)
2854 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2855 if (rtoinfo.srto_min != 0)
2856 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2864 * 7.1.2 SCTP_ASSOCINFO
2866 * This option is used to tune the maximum retransmission attempts
2867 * of the association.
2868 * Returns an error if the new association retransmission value is
2869 * greater than the sum of the retransmission value of the peer.
2870 * See [SCTP] for more information.
2873 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2876 struct sctp_assocparams assocparams;
2877 struct sctp_association *asoc;
2879 if (optlen != sizeof(struct sctp_assocparams))
2881 if (copy_from_user(&assocparams, optval, optlen))
2884 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2886 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2889 /* Set the values to the specific association */
2891 if (assocparams.sasoc_asocmaxrxt != 0) {
2894 struct sctp_transport *peer_addr;
2896 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2898 path_sum += peer_addr->pathmaxrxt;
2902 /* Only validate asocmaxrxt if we have more than
2903 * one path/transport. We do this because path
2904 * retransmissions are only counted when we have more
2908 assocparams.sasoc_asocmaxrxt > path_sum)
2911 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2914 if (assocparams.sasoc_cookie_life != 0)
2915 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
2917 /* Set the values to the endpoint */
2918 struct sctp_sock *sp = sctp_sk(sk);
2920 if (assocparams.sasoc_asocmaxrxt != 0)
2921 sp->assocparams.sasoc_asocmaxrxt =
2922 assocparams.sasoc_asocmaxrxt;
2923 if (assocparams.sasoc_cookie_life != 0)
2924 sp->assocparams.sasoc_cookie_life =
2925 assocparams.sasoc_cookie_life;
2931 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2933 * This socket option is a boolean flag which turns on or off mapped V4
2934 * addresses. If this option is turned on and the socket is type
2935 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2936 * If this option is turned off, then no mapping will be done of V4
2937 * addresses and a user will receive both PF_INET6 and PF_INET type
2938 * addresses on the socket.
2940 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2943 struct sctp_sock *sp = sctp_sk(sk);
2945 if (optlen < sizeof(int))
2947 if (get_user(val, (int __user *)optval))
2958 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2959 * This option will get or set the maximum size to put in any outgoing
2960 * SCTP DATA chunk. If a message is larger than this size it will be
2961 * fragmented by SCTP into the specified size. Note that the underlying
2962 * SCTP implementation may fragment into smaller sized chunks when the
2963 * PMTU of the underlying association is smaller than the value set by
2964 * the user. The default value for this option is '0' which indicates
2965 * the user is NOT limiting fragmentation and only the PMTU will effect
2966 * SCTP's choice of DATA chunk size. Note also that values set larger
2967 * than the maximum size of an IP datagram will effectively let SCTP
2968 * control fragmentation (i.e. the same as setting this option to 0).
2970 * The following structure is used to access and modify this parameter:
2972 * struct sctp_assoc_value {
2973 * sctp_assoc_t assoc_id;
2974 * uint32_t assoc_value;
2977 * assoc_id: This parameter is ignored for one-to-one style sockets.
2978 * For one-to-many style sockets this parameter indicates which
2979 * association the user is performing an action upon. Note that if
2980 * this field's value is zero then the endpoints default value is
2981 * changed (effecting future associations only).
2982 * assoc_value: This parameter specifies the maximum size in bytes.
2984 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2986 struct sctp_assoc_value params;
2987 struct sctp_association *asoc;
2988 struct sctp_sock *sp = sctp_sk(sk);
2991 if (optlen == sizeof(int)) {
2992 pr_warn("Use of int in maxseg socket option deprecated\n");
2993 pr_warn("Use struct sctp_assoc_value instead\n");
2994 if (copy_from_user(&val, optval, optlen))
2996 params.assoc_id = 0;
2997 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2998 if (copy_from_user(¶ms, optval, optlen))
3000 val = params.assoc_value;
3004 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3007 asoc = sctp_id2assoc(sk, params.assoc_id);
3008 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3013 val = asoc->pathmtu;
3014 val -= sp->pf->af->net_header_len;
3015 val -= sizeof(struct sctphdr) +
3016 sizeof(struct sctp_data_chunk);
3018 asoc->user_frag = val;
3019 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3021 sp->user_frag = val;
3029 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3031 * Requests that the peer mark the enclosed address as the association
3032 * primary. The enclosed address must be one of the association's
3033 * locally bound addresses. The following structure is used to make a
3034 * set primary request:
3036 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3037 unsigned int optlen)
3039 struct net *net = sock_net(sk);
3040 struct sctp_sock *sp;
3041 struct sctp_association *asoc = NULL;
3042 struct sctp_setpeerprim prim;
3043 struct sctp_chunk *chunk;
3049 if (!net->sctp.addip_enable)
3052 if (optlen != sizeof(struct sctp_setpeerprim))
3055 if (copy_from_user(&prim, optval, optlen))
3058 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3062 if (!asoc->peer.asconf_capable)
3065 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3068 if (!sctp_state(asoc, ESTABLISHED))
3071 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3075 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3076 return -EADDRNOTAVAIL;
3078 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3079 return -EADDRNOTAVAIL;
3081 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3082 chunk = sctp_make_asconf_set_prim(asoc,
3083 (union sctp_addr *)&prim.sspp_addr);
3087 err = sctp_send_asconf(asoc, chunk);
3089 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
3094 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3095 unsigned int optlen)
3097 struct sctp_setadaptation adaptation;
3099 if (optlen != sizeof(struct sctp_setadaptation))
3101 if (copy_from_user(&adaptation, optval, optlen))
3104 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3110 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3112 * The context field in the sctp_sndrcvinfo structure is normally only
3113 * used when a failed message is retrieved holding the value that was
3114 * sent down on the actual send call. This option allows the setting of
3115 * a default context on an association basis that will be received on
3116 * reading messages from the peer. This is especially helpful in the
3117 * one-2-many model for an application to keep some reference to an
3118 * internal state machine that is processing messages on the
3119 * association. Note that the setting of this value only effects
3120 * received messages from the peer and does not effect the value that is
3121 * saved with outbound messages.
3123 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3124 unsigned int optlen)
3126 struct sctp_assoc_value params;
3127 struct sctp_sock *sp;
3128 struct sctp_association *asoc;
3130 if (optlen != sizeof(struct sctp_assoc_value))
3132 if (copy_from_user(¶ms, optval, optlen))
3137 if (params.assoc_id != 0) {
3138 asoc = sctp_id2assoc(sk, params.assoc_id);
3141 asoc->default_rcv_context = params.assoc_value;
3143 sp->default_rcv_context = params.assoc_value;
3150 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3152 * This options will at a minimum specify if the implementation is doing
3153 * fragmented interleave. Fragmented interleave, for a one to many
3154 * socket, is when subsequent calls to receive a message may return
3155 * parts of messages from different associations. Some implementations
3156 * may allow you to turn this value on or off. If so, when turned off,
3157 * no fragment interleave will occur (which will cause a head of line
3158 * blocking amongst multiple associations sharing the same one to many
3159 * socket). When this option is turned on, then each receive call may
3160 * come from a different association (thus the user must receive data
3161 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3162 * association each receive belongs to.
3164 * This option takes a boolean value. A non-zero value indicates that
3165 * fragmented interleave is on. A value of zero indicates that
3166 * fragmented interleave is off.
3168 * Note that it is important that an implementation that allows this
3169 * option to be turned on, have it off by default. Otherwise an unaware
3170 * application using the one to many model may become confused and act
3173 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3174 char __user *optval,
3175 unsigned int optlen)
3179 if (optlen != sizeof(int))
3181 if (get_user(val, (int __user *)optval))
3184 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3190 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3191 * (SCTP_PARTIAL_DELIVERY_POINT)
3193 * This option will set or get the SCTP partial delivery point. This
3194 * point is the size of a message where the partial delivery API will be
3195 * invoked to help free up rwnd space for the peer. Setting this to a
3196 * lower value will cause partial deliveries to happen more often. The
3197 * calls argument is an integer that sets or gets the partial delivery
3198 * point. Note also that the call will fail if the user attempts to set
3199 * this value larger than the socket receive buffer size.
3201 * Note that any single message having a length smaller than or equal to
3202 * the SCTP partial delivery point will be delivered in one single read
3203 * call as long as the user provided buffer is large enough to hold the
3206 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3207 char __user *optval,
3208 unsigned int optlen)
3212 if (optlen != sizeof(u32))
3214 if (get_user(val, (int __user *)optval))
3217 /* Note: We double the receive buffer from what the user sets
3218 * it to be, also initial rwnd is based on rcvbuf/2.
3220 if (val > (sk->sk_rcvbuf >> 1))
3223 sctp_sk(sk)->pd_point = val;
3225 return 0; /* is this the right error code? */
3229 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3231 * This option will allow a user to change the maximum burst of packets
3232 * that can be emitted by this association. Note that the default value
3233 * is 4, and some implementations may restrict this setting so that it
3234 * can only be lowered.
3236 * NOTE: This text doesn't seem right. Do this on a socket basis with
3237 * future associations inheriting the socket value.
3239 static int sctp_setsockopt_maxburst(struct sock *sk,
3240 char __user *optval,
3241 unsigned int optlen)
3243 struct sctp_assoc_value params;
3244 struct sctp_sock *sp;
3245 struct sctp_association *asoc;
3249 if (optlen == sizeof(int)) {
3250 pr_warn("Use of int in max_burst socket option deprecated\n");
3251 pr_warn("Use struct sctp_assoc_value instead\n");
3252 if (copy_from_user(&val, optval, optlen))
3254 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3255 if (copy_from_user(¶ms, optval, optlen))
3257 val = params.assoc_value;
3258 assoc_id = params.assoc_id;
3264 if (assoc_id != 0) {
3265 asoc = sctp_id2assoc(sk, assoc_id);
3268 asoc->max_burst = val;
3270 sp->max_burst = val;
3276 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3278 * This set option adds a chunk type that the user is requesting to be
3279 * received only in an authenticated way. Changes to the list of chunks
3280 * will only effect future associations on the socket.
3282 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3283 char __user *optval,
3284 unsigned int optlen)
3286 struct net *net = sock_net(sk);
3287 struct sctp_authchunk val;
3289 if (!net->sctp.auth_enable)
3292 if (optlen != sizeof(struct sctp_authchunk))
3294 if (copy_from_user(&val, optval, optlen))
3297 switch (val.sauth_chunk) {
3299 case SCTP_CID_INIT_ACK:
3300 case SCTP_CID_SHUTDOWN_COMPLETE:
3305 /* add this chunk id to the endpoint */
3306 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3310 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3312 * This option gets or sets the list of HMAC algorithms that the local
3313 * endpoint requires the peer to use.
3315 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3316 char __user *optval,
3317 unsigned int optlen)
3319 struct net *net = sock_net(sk);
3320 struct sctp_hmacalgo *hmacs;
3324 if (!net->sctp.auth_enable)
3327 if (optlen < sizeof(struct sctp_hmacalgo))
3330 hmacs= memdup_user(optval, optlen);
3332 return PTR_ERR(hmacs);
3334 idents = hmacs->shmac_num_idents;
3335 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3336 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3341 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3348 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3350 * This option will set a shared secret key which is used to build an
3351 * association shared key.
3353 static int sctp_setsockopt_auth_key(struct sock *sk,
3354 char __user *optval,
3355 unsigned int optlen)
3357 struct net *net = sock_net(sk);
3358 struct sctp_authkey *authkey;
3359 struct sctp_association *asoc;
3362 if (!net->sctp.auth_enable)
3365 if (optlen <= sizeof(struct sctp_authkey))
3368 authkey= memdup_user(optval, optlen);
3369 if (IS_ERR(authkey))
3370 return PTR_ERR(authkey);
3372 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3377 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3378 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3383 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3390 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3392 * This option will get or set the active shared key to be used to build
3393 * the association shared key.
3395 static int sctp_setsockopt_active_key(struct sock *sk,
3396 char __user *optval,
3397 unsigned int optlen)
3399 struct net *net = sock_net(sk);
3400 struct sctp_authkeyid val;
3401 struct sctp_association *asoc;
3403 if (!net->sctp.auth_enable)
3406 if (optlen != sizeof(struct sctp_authkeyid))
3408 if (copy_from_user(&val, optval, optlen))
3411 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3412 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3415 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3416 val.scact_keynumber);
3420 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3422 * This set option will delete a shared secret key from use.
3424 static int sctp_setsockopt_del_key(struct sock *sk,
3425 char __user *optval,
3426 unsigned int optlen)
3428 struct net *net = sock_net(sk);
3429 struct sctp_authkeyid val;
3430 struct sctp_association *asoc;
3432 if (!net->sctp.auth_enable)
3435 if (optlen != sizeof(struct sctp_authkeyid))
3437 if (copy_from_user(&val, optval, optlen))
3440 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3441 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3444 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3445 val.scact_keynumber);
3450 * 8.1.23 SCTP_AUTO_ASCONF
3452 * This option will enable or disable the use of the automatic generation of
3453 * ASCONF chunks to add and delete addresses to an existing association. Note
3454 * that this option has two caveats namely: a) it only affects sockets that
3455 * are bound to all addresses available to the SCTP stack, and b) the system
3456 * administrator may have an overriding control that turns the ASCONF feature
3457 * off no matter what setting the socket option may have.
3458 * This option expects an integer boolean flag, where a non-zero value turns on
3459 * the option, and a zero value turns off the option.
3460 * Note. In this implementation, socket operation overrides default parameter
3461 * being set by sysctl as well as FreeBSD implementation
3463 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3464 unsigned int optlen)
3467 struct sctp_sock *sp = sctp_sk(sk);
3469 if (optlen < sizeof(int))
3471 if (get_user(val, (int __user *)optval))
3473 if (!sctp_is_ep_boundall(sk) && val)
3475 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3478 if (val == 0 && sp->do_auto_asconf) {
3479 list_del(&sp->auto_asconf_list);
3480 sp->do_auto_asconf = 0;
3481 } else if (val && !sp->do_auto_asconf) {
3482 list_add_tail(&sp->auto_asconf_list,
3483 &sock_net(sk)->sctp.auto_asconf_splist);
3484 sp->do_auto_asconf = 1;
3491 * SCTP_PEER_ADDR_THLDS
3493 * This option allows us to alter the partially failed threshold for one or all
3494 * transports in an association. See Section 6.1 of:
3495 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3497 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3498 char __user *optval,
3499 unsigned int optlen)
3501 struct sctp_paddrthlds val;
3502 struct sctp_transport *trans;
3503 struct sctp_association *asoc;
3505 if (optlen < sizeof(struct sctp_paddrthlds))
3507 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3508 sizeof(struct sctp_paddrthlds)))
3512 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3513 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3516 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3518 if (val.spt_pathmaxrxt)
3519 trans->pathmaxrxt = val.spt_pathmaxrxt;
3520 trans->pf_retrans = val.spt_pathpfthld;
3523 if (val.spt_pathmaxrxt)
3524 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3525 asoc->pf_retrans = val.spt_pathpfthld;
3527 trans = sctp_addr_id2transport(sk, &val.spt_address,
3532 if (val.spt_pathmaxrxt)
3533 trans->pathmaxrxt = val.spt_pathmaxrxt;
3534 trans->pf_retrans = val.spt_pathpfthld;
3540 /* API 6.2 setsockopt(), getsockopt()
3542 * Applications use setsockopt() and getsockopt() to set or retrieve
3543 * socket options. Socket options are used to change the default
3544 * behavior of sockets calls. They are described in Section 7.
3548 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3549 * int __user *optlen);
3550 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3553 * sd - the socket descript.
3554 * level - set to IPPROTO_SCTP for all SCTP options.
3555 * optname - the option name.
3556 * optval - the buffer to store the value of the option.
3557 * optlen - the size of the buffer.
3559 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3560 char __user *optval, unsigned int optlen)
3564 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3567 /* I can hardly begin to describe how wrong this is. This is
3568 * so broken as to be worse than useless. The API draft
3569 * REALLY is NOT helpful here... I am not convinced that the
3570 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3571 * are at all well-founded.
3573 if (level != SOL_SCTP) {
3574 struct sctp_af *af = sctp_sk(sk)->pf->af;
3575 retval = af->setsockopt(sk, level, optname, optval, optlen);
3582 case SCTP_SOCKOPT_BINDX_ADD:
3583 /* 'optlen' is the size of the addresses buffer. */
3584 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3585 optlen, SCTP_BINDX_ADD_ADDR);
3588 case SCTP_SOCKOPT_BINDX_REM:
3589 /* 'optlen' is the size of the addresses buffer. */
3590 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3591 optlen, SCTP_BINDX_REM_ADDR);
3594 case SCTP_SOCKOPT_CONNECTX_OLD:
3595 /* 'optlen' is the size of the addresses buffer. */
3596 retval = sctp_setsockopt_connectx_old(sk,
3597 (struct sockaddr __user *)optval,
3601 case SCTP_SOCKOPT_CONNECTX:
3602 /* 'optlen' is the size of the addresses buffer. */
3603 retval = sctp_setsockopt_connectx(sk,
3604 (struct sockaddr __user *)optval,
3608 case SCTP_DISABLE_FRAGMENTS:
3609 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3613 retval = sctp_setsockopt_events(sk, optval, optlen);
3616 case SCTP_AUTOCLOSE:
3617 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3620 case SCTP_PEER_ADDR_PARAMS:
3621 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3624 case SCTP_DELAYED_SACK:
3625 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3627 case SCTP_PARTIAL_DELIVERY_POINT:
3628 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3632 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3634 case SCTP_DEFAULT_SEND_PARAM:
3635 retval = sctp_setsockopt_default_send_param(sk, optval,
3638 case SCTP_PRIMARY_ADDR:
3639 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3641 case SCTP_SET_PEER_PRIMARY_ADDR:
3642 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3645 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3648 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3650 case SCTP_ASSOCINFO:
3651 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3653 case SCTP_I_WANT_MAPPED_V4_ADDR:
3654 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3657 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3659 case SCTP_ADAPTATION_LAYER:
3660 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3663 retval = sctp_setsockopt_context(sk, optval, optlen);
3665 case SCTP_FRAGMENT_INTERLEAVE:
3666 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3668 case SCTP_MAX_BURST:
3669 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3671 case SCTP_AUTH_CHUNK:
3672 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3674 case SCTP_HMAC_IDENT:
3675 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3678 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3680 case SCTP_AUTH_ACTIVE_KEY:
3681 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3683 case SCTP_AUTH_DELETE_KEY:
3684 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3686 case SCTP_AUTO_ASCONF:
3687 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3689 case SCTP_PEER_ADDR_THLDS:
3690 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3693 retval = -ENOPROTOOPT;
3697 sctp_release_sock(sk);
3703 /* API 3.1.6 connect() - UDP Style Syntax
3705 * An application may use the connect() call in the UDP model to initiate an
3706 * association without sending data.
3710 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3712 * sd: the socket descriptor to have a new association added to.
3714 * nam: the address structure (either struct sockaddr_in or struct
3715 * sockaddr_in6 defined in RFC2553 [7]).
3717 * len: the size of the address.
3719 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3727 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3728 __func__, sk, addr, addr_len);
3730 /* Validate addr_len before calling common connect/connectx routine. */
3731 af = sctp_get_af_specific(addr->sa_family);
3732 if (!af || addr_len < af->sockaddr_len) {
3735 /* Pass correct addr len to common routine (so it knows there
3736 * is only one address being passed.
3738 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3741 sctp_release_sock(sk);
3745 /* FIXME: Write comments. */
3746 static int sctp_disconnect(struct sock *sk, int flags)
3748 return -EOPNOTSUPP; /* STUB */
3751 /* 4.1.4 accept() - TCP Style Syntax
3753 * Applications use accept() call to remove an established SCTP
3754 * association from the accept queue of the endpoint. A new socket
3755 * descriptor will be returned from accept() to represent the newly
3756 * formed association.
3758 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3760 struct sctp_sock *sp;
3761 struct sctp_endpoint *ep;
3762 struct sock *newsk = NULL;
3763 struct sctp_association *asoc;
3772 if (!sctp_style(sk, TCP)) {
3773 error = -EOPNOTSUPP;
3777 if (!sctp_sstate(sk, LISTENING)) {
3782 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3784 error = sctp_wait_for_accept(sk, timeo);
3788 /* We treat the list of associations on the endpoint as the accept
3789 * queue and pick the first association on the list.
3791 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3793 newsk = sp->pf->create_accept_sk(sk, asoc);
3799 /* Populate the fields of the newsk from the oldsk and migrate the
3800 * asoc to the newsk.
3802 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3805 sctp_release_sock(sk);
3810 /* The SCTP ioctl handler. */
3811 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3818 * SEQPACKET-style sockets in LISTENING state are valid, for
3819 * SCTP, so only discard TCP-style sockets in LISTENING state.
3821 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3826 struct sk_buff *skb;
3827 unsigned int amount = 0;
3829 skb = skb_peek(&sk->sk_receive_queue);
3832 * We will only return the amount of this packet since
3833 * that is all that will be read.
3837 rc = put_user(amount, (int __user *)arg);
3845 sctp_release_sock(sk);
3849 /* This is the function which gets called during socket creation to
3850 * initialized the SCTP-specific portion of the sock.
3851 * The sock structure should already be zero-filled memory.
3853 static int sctp_init_sock(struct sock *sk)
3855 struct net *net = sock_net(sk);
3856 struct sctp_sock *sp;
3858 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3862 /* Initialize the SCTP per socket area. */
3863 switch (sk->sk_type) {
3864 case SOCK_SEQPACKET:
3865 sp->type = SCTP_SOCKET_UDP;
3868 sp->type = SCTP_SOCKET_TCP;
3871 return -ESOCKTNOSUPPORT;
3874 /* Initialize default send parameters. These parameters can be
3875 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3877 sp->default_stream = 0;
3878 sp->default_ppid = 0;
3879 sp->default_flags = 0;
3880 sp->default_context = 0;
3881 sp->default_timetolive = 0;
3883 sp->default_rcv_context = 0;
3884 sp->max_burst = net->sctp.max_burst;
3886 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
3888 /* Initialize default setup parameters. These parameters
3889 * can be modified with the SCTP_INITMSG socket option or
3890 * overridden by the SCTP_INIT CMSG.
3892 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3893 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3894 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
3895 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
3897 /* Initialize default RTO related parameters. These parameters can
3898 * be modified for with the SCTP_RTOINFO socket option.
3900 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
3901 sp->rtoinfo.srto_max = net->sctp.rto_max;
3902 sp->rtoinfo.srto_min = net->sctp.rto_min;
3904 /* Initialize default association related parameters. These parameters
3905 * can be modified with the SCTP_ASSOCINFO socket option.
3907 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
3908 sp->assocparams.sasoc_number_peer_destinations = 0;
3909 sp->assocparams.sasoc_peer_rwnd = 0;
3910 sp->assocparams.sasoc_local_rwnd = 0;
3911 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
3913 /* Initialize default event subscriptions. By default, all the
3916 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3918 /* Default Peer Address Parameters. These defaults can
3919 * be modified via SCTP_PEER_ADDR_PARAMS
3921 sp->hbinterval = net->sctp.hb_interval;
3922 sp->pathmaxrxt = net->sctp.max_retrans_path;
3923 sp->pathmtu = 0; // allow default discovery
3924 sp->sackdelay = net->sctp.sack_timeout;
3926 sp->param_flags = SPP_HB_ENABLE |
3928 SPP_SACKDELAY_ENABLE;
3930 /* If enabled no SCTP message fragmentation will be performed.
3931 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3933 sp->disable_fragments = 0;
3935 /* Enable Nagle algorithm by default. */
3938 /* Enable by default. */
3941 /* Auto-close idle associations after the configured
3942 * number of seconds. A value of 0 disables this
3943 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3944 * for UDP-style sockets only.
3948 /* User specified fragmentation limit. */
3951 sp->adaptation_ind = 0;
3953 sp->pf = sctp_get_pf_specific(sk->sk_family);
3955 /* Control variables for partial data delivery. */
3956 atomic_set(&sp->pd_mode, 0);
3957 skb_queue_head_init(&sp->pd_lobby);
3958 sp->frag_interleave = 0;
3960 /* Create a per socket endpoint structure. Even if we
3961 * change the data structure relationships, this may still
3962 * be useful for storing pre-connect address information.
3964 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
3970 sk->sk_destruct = sctp_destruct_sock;
3972 SCTP_DBG_OBJCNT_INC(sock);
3975 percpu_counter_inc(&sctp_sockets_allocated);
3976 sock_prot_inuse_add(net, sk->sk_prot, 1);
3977 if (net->sctp.default_auto_asconf) {
3978 list_add_tail(&sp->auto_asconf_list,
3979 &net->sctp.auto_asconf_splist);
3980 sp->do_auto_asconf = 1;
3982 sp->do_auto_asconf = 0;
3988 /* Cleanup any SCTP per socket resources. */
3989 static void sctp_destroy_sock(struct sock *sk)
3991 struct sctp_sock *sp;
3993 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3995 /* Release our hold on the endpoint. */
3997 /* This could happen during socket init, thus we bail out
3998 * early, since the rest of the below is not setup either.
4003 if (sp->do_auto_asconf) {
4004 sp->do_auto_asconf = 0;
4005 list_del(&sp->auto_asconf_list);
4007 sctp_endpoint_free(sp->ep);
4009 percpu_counter_dec(&sctp_sockets_allocated);
4010 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4014 /* Triggered when there are no references on the socket anymore */
4015 static void sctp_destruct_sock(struct sock *sk)
4017 struct sctp_sock *sp = sctp_sk(sk);
4019 /* Free up the HMAC transform. */
4020 crypto_free_hash(sp->hmac);
4022 inet_sock_destruct(sk);
4025 /* API 4.1.7 shutdown() - TCP Style Syntax
4026 * int shutdown(int socket, int how);
4028 * sd - the socket descriptor of the association to be closed.
4029 * how - Specifies the type of shutdown. The values are
4032 * Disables further receive operations. No SCTP
4033 * protocol action is taken.
4035 * Disables further send operations, and initiates
4036 * the SCTP shutdown sequence.
4038 * Disables further send and receive operations
4039 * and initiates the SCTP shutdown sequence.
4041 static void sctp_shutdown(struct sock *sk, int how)
4043 struct net *net = sock_net(sk);
4044 struct sctp_endpoint *ep;
4045 struct sctp_association *asoc;
4047 if (!sctp_style(sk, TCP))
4050 if (how & SEND_SHUTDOWN) {
4051 ep = sctp_sk(sk)->ep;
4052 if (!list_empty(&ep->asocs)) {
4053 asoc = list_entry(ep->asocs.next,
4054 struct sctp_association, asocs);
4055 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4060 /* 7.2.1 Association Status (SCTP_STATUS)
4062 * Applications can retrieve current status information about an
4063 * association, including association state, peer receiver window size,
4064 * number of unacked data chunks, and number of data chunks pending
4065 * receipt. This information is read-only.
4067 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4068 char __user *optval,
4071 struct sctp_status status;
4072 struct sctp_association *asoc = NULL;
4073 struct sctp_transport *transport;
4074 sctp_assoc_t associd;
4077 if (len < sizeof(status)) {
4082 len = sizeof(status);
4083 if (copy_from_user(&status, optval, len)) {
4088 associd = status.sstat_assoc_id;
4089 asoc = sctp_id2assoc(sk, associd);
4095 transport = asoc->peer.primary_path;
4097 status.sstat_assoc_id = sctp_assoc2id(asoc);
4098 status.sstat_state = asoc->state;
4099 status.sstat_rwnd = asoc->peer.rwnd;
4100 status.sstat_unackdata = asoc->unack_data;
4102 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4103 status.sstat_instrms = asoc->c.sinit_max_instreams;
4104 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4105 status.sstat_fragmentation_point = asoc->frag_point;
4106 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4107 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4108 transport->af_specific->sockaddr_len);
4109 /* Map ipv4 address into v4-mapped-on-v6 address. */
4110 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4111 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4112 status.sstat_primary.spinfo_state = transport->state;
4113 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4114 status.sstat_primary.spinfo_srtt = transport->srtt;
4115 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4116 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4118 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4119 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4121 if (put_user(len, optlen)) {
4126 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
4127 len, status.sstat_state, status.sstat_rwnd,
4128 status.sstat_assoc_id);
4130 if (copy_to_user(optval, &status, len)) {
4140 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4142 * Applications can retrieve information about a specific peer address
4143 * of an association, including its reachability state, congestion
4144 * window, and retransmission timer values. This information is
4147 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4148 char __user *optval,
4151 struct sctp_paddrinfo pinfo;
4152 struct sctp_transport *transport;
4155 if (len < sizeof(pinfo)) {
4160 len = sizeof(pinfo);
4161 if (copy_from_user(&pinfo, optval, len)) {
4166 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4167 pinfo.spinfo_assoc_id);
4171 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4172 pinfo.spinfo_state = transport->state;
4173 pinfo.spinfo_cwnd = transport->cwnd;
4174 pinfo.spinfo_srtt = transport->srtt;
4175 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4176 pinfo.spinfo_mtu = transport->pathmtu;
4178 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4179 pinfo.spinfo_state = SCTP_ACTIVE;
4181 if (put_user(len, optlen)) {
4186 if (copy_to_user(optval, &pinfo, len)) {
4195 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4197 * This option is a on/off flag. If enabled no SCTP message
4198 * fragmentation will be performed. Instead if a message being sent
4199 * exceeds the current PMTU size, the message will NOT be sent and
4200 * instead a error will be indicated to the user.
4202 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4203 char __user *optval, int __user *optlen)
4207 if (len < sizeof(int))
4211 val = (sctp_sk(sk)->disable_fragments == 1);
4212 if (put_user(len, optlen))
4214 if (copy_to_user(optval, &val, len))
4219 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4221 * This socket option is used to specify various notifications and
4222 * ancillary data the user wishes to receive.
4224 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4229 if (len > sizeof(struct sctp_event_subscribe))
4230 len = sizeof(struct sctp_event_subscribe);
4231 if (put_user(len, optlen))
4233 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4238 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4240 * This socket option is applicable to the UDP-style socket only. When
4241 * set it will cause associations that are idle for more than the
4242 * specified number of seconds to automatically close. An association
4243 * being idle is defined an association that has NOT sent or received
4244 * user data. The special value of '0' indicates that no automatic
4245 * close of any associations should be performed. The option expects an
4246 * integer defining the number of seconds of idle time before an
4247 * association is closed.
4249 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4251 /* Applicable to UDP-style socket only */
4252 if (sctp_style(sk, TCP))
4254 if (len < sizeof(int))
4257 if (put_user(len, optlen))
4259 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4264 /* Helper routine to branch off an association to a new socket. */
4265 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4267 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4268 struct socket *sock;
4275 /* An association cannot be branched off from an already peeled-off
4276 * socket, nor is this supported for tcp style sockets.
4278 if (!sctp_style(sk, UDP))
4281 /* Create a new socket. */
4282 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4286 sctp_copy_sock(sock->sk, sk, asoc);
4288 /* Make peeled-off sockets more like 1-1 accepted sockets.
4289 * Set the daddr and initialize id to something more random
4291 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4292 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4294 /* Populate the fields of the newsk from the oldsk and migrate the
4295 * asoc to the newsk.
4297 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4303 EXPORT_SYMBOL(sctp_do_peeloff);
4305 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4307 sctp_peeloff_arg_t peeloff;
4308 struct socket *newsock;
4309 struct file *newfile;
4312 if (len < sizeof(sctp_peeloff_arg_t))
4314 len = sizeof(sctp_peeloff_arg_t);
4315 if (copy_from_user(&peeloff, optval, len))
4318 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4322 /* Map the socket to an unused fd that can be returned to the user. */
4323 retval = get_unused_fd();
4325 sock_release(newsock);
4329 newfile = sock_alloc_file(newsock, 0, NULL);
4330 if (unlikely(IS_ERR(newfile))) {
4331 put_unused_fd(retval);
4332 sock_release(newsock);
4333 return PTR_ERR(newfile);
4336 SCTP_DEBUG_PRINTK("%s: sk: %p newsk: %p sd: %d\n",
4337 __func__, sk, newsock->sk, retval);
4339 /* Return the fd mapped to the new socket. */
4340 if (put_user(len, optlen)) {
4342 put_unused_fd(retval);
4345 peeloff.sd = retval;
4346 if (copy_to_user(optval, &peeloff, len)) {
4348 put_unused_fd(retval);
4351 fd_install(retval, newfile);
4356 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4358 * Applications can enable or disable heartbeats for any peer address of
4359 * an association, modify an address's heartbeat interval, force a
4360 * heartbeat to be sent immediately, and adjust the address's maximum
4361 * number of retransmissions sent before an address is considered
4362 * unreachable. The following structure is used to access and modify an
4363 * address's parameters:
4365 * struct sctp_paddrparams {
4366 * sctp_assoc_t spp_assoc_id;
4367 * struct sockaddr_storage spp_address;
4368 * uint32_t spp_hbinterval;
4369 * uint16_t spp_pathmaxrxt;
4370 * uint32_t spp_pathmtu;
4371 * uint32_t spp_sackdelay;
4372 * uint32_t spp_flags;
4375 * spp_assoc_id - (one-to-many style socket) This is filled in the
4376 * application, and identifies the association for
4378 * spp_address - This specifies which address is of interest.
4379 * spp_hbinterval - This contains the value of the heartbeat interval,
4380 * in milliseconds. If a value of zero
4381 * is present in this field then no changes are to
4382 * be made to this parameter.
4383 * spp_pathmaxrxt - This contains the maximum number of
4384 * retransmissions before this address shall be
4385 * considered unreachable. If a value of zero
4386 * is present in this field then no changes are to
4387 * be made to this parameter.
4388 * spp_pathmtu - When Path MTU discovery is disabled the value
4389 * specified here will be the "fixed" path mtu.
4390 * Note that if the spp_address field is empty
4391 * then all associations on this address will
4392 * have this fixed path mtu set upon them.
4394 * spp_sackdelay - When delayed sack is enabled, this value specifies
4395 * the number of milliseconds that sacks will be delayed
4396 * for. This value will apply to all addresses of an
4397 * association if the spp_address field is empty. Note
4398 * also, that if delayed sack is enabled and this
4399 * value is set to 0, no change is made to the last
4400 * recorded delayed sack timer value.
4402 * spp_flags - These flags are used to control various features
4403 * on an association. The flag field may contain
4404 * zero or more of the following options.
4406 * SPP_HB_ENABLE - Enable heartbeats on the
4407 * specified address. Note that if the address
4408 * field is empty all addresses for the association
4409 * have heartbeats enabled upon them.
4411 * SPP_HB_DISABLE - Disable heartbeats on the
4412 * speicifed address. Note that if the address
4413 * field is empty all addresses for the association
4414 * will have their heartbeats disabled. Note also
4415 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4416 * mutually exclusive, only one of these two should
4417 * be specified. Enabling both fields will have
4418 * undetermined results.
4420 * SPP_HB_DEMAND - Request a user initiated heartbeat
4421 * to be made immediately.
4423 * SPP_PMTUD_ENABLE - This field will enable PMTU
4424 * discovery upon the specified address. Note that
4425 * if the address feild is empty then all addresses
4426 * on the association are effected.
4428 * SPP_PMTUD_DISABLE - This field will disable PMTU
4429 * discovery upon the specified address. Note that
4430 * if the address feild is empty then all addresses
4431 * on the association are effected. Not also that
4432 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4433 * exclusive. Enabling both will have undetermined
4436 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4437 * on delayed sack. The time specified in spp_sackdelay
4438 * is used to specify the sack delay for this address. Note
4439 * that if spp_address is empty then all addresses will
4440 * enable delayed sack and take on the sack delay
4441 * value specified in spp_sackdelay.
4442 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4443 * off delayed sack. If the spp_address field is blank then
4444 * delayed sack is disabled for the entire association. Note
4445 * also that this field is mutually exclusive to
4446 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4449 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4450 char __user *optval, int __user *optlen)
4452 struct sctp_paddrparams params;
4453 struct sctp_transport *trans = NULL;
4454 struct sctp_association *asoc = NULL;
4455 struct sctp_sock *sp = sctp_sk(sk);
4457 if (len < sizeof(struct sctp_paddrparams))
4459 len = sizeof(struct sctp_paddrparams);
4460 if (copy_from_user(¶ms, optval, len))
4463 /* If an address other than INADDR_ANY is specified, and
4464 * no transport is found, then the request is invalid.
4466 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
4467 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4468 params.spp_assoc_id);
4470 SCTP_DEBUG_PRINTK("Failed no transport\n");
4475 /* Get association, if assoc_id != 0 and the socket is a one
4476 * to many style socket, and an association was not found, then
4477 * the id was invalid.
4479 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4480 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4481 SCTP_DEBUG_PRINTK("Failed no association\n");
4486 /* Fetch transport values. */
4487 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4488 params.spp_pathmtu = trans->pathmtu;
4489 params.spp_pathmaxrxt = trans->pathmaxrxt;
4490 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4492 /*draft-11 doesn't say what to return in spp_flags*/
4493 params.spp_flags = trans->param_flags;
4495 /* Fetch association values. */
4496 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4497 params.spp_pathmtu = asoc->pathmtu;
4498 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4499 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4501 /*draft-11 doesn't say what to return in spp_flags*/
4502 params.spp_flags = asoc->param_flags;
4504 /* Fetch socket values. */
4505 params.spp_hbinterval = sp->hbinterval;
4506 params.spp_pathmtu = sp->pathmtu;
4507 params.spp_sackdelay = sp->sackdelay;
4508 params.spp_pathmaxrxt = sp->pathmaxrxt;
4510 /*draft-11 doesn't say what to return in spp_flags*/
4511 params.spp_flags = sp->param_flags;
4514 if (copy_to_user(optval, ¶ms, len))
4517 if (put_user(len, optlen))
4524 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4526 * This option will effect the way delayed acks are performed. This
4527 * option allows you to get or set the delayed ack time, in
4528 * milliseconds. It also allows changing the delayed ack frequency.
4529 * Changing the frequency to 1 disables the delayed sack algorithm. If
4530 * the assoc_id is 0, then this sets or gets the endpoints default
4531 * values. If the assoc_id field is non-zero, then the set or get
4532 * effects the specified association for the one to many model (the
4533 * assoc_id field is ignored by the one to one model). Note that if
4534 * sack_delay or sack_freq are 0 when setting this option, then the
4535 * current values will remain unchanged.
4537 * struct sctp_sack_info {
4538 * sctp_assoc_t sack_assoc_id;
4539 * uint32_t sack_delay;
4540 * uint32_t sack_freq;
4543 * sack_assoc_id - This parameter, indicates which association the user
4544 * is performing an action upon. Note that if this field's value is
4545 * zero then the endpoints default value is changed (effecting future
4546 * associations only).
4548 * sack_delay - This parameter contains the number of milliseconds that
4549 * the user is requesting the delayed ACK timer be set to. Note that
4550 * this value is defined in the standard to be between 200 and 500
4553 * sack_freq - This parameter contains the number of packets that must
4554 * be received before a sack is sent without waiting for the delay
4555 * timer to expire. The default value for this is 2, setting this
4556 * value to 1 will disable the delayed sack algorithm.
4558 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4559 char __user *optval,
4562 struct sctp_sack_info params;
4563 struct sctp_association *asoc = NULL;
4564 struct sctp_sock *sp = sctp_sk(sk);
4566 if (len >= sizeof(struct sctp_sack_info)) {
4567 len = sizeof(struct sctp_sack_info);
4569 if (copy_from_user(¶ms, optval, len))
4571 } else if (len == sizeof(struct sctp_assoc_value)) {
4572 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4573 pr_warn("Use struct sctp_sack_info instead\n");
4574 if (copy_from_user(¶ms, optval, len))
4579 /* Get association, if sack_assoc_id != 0 and the socket is a one
4580 * to many style socket, and an association was not found, then
4581 * the id was invalid.
4583 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4584 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4588 /* Fetch association values. */
4589 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4590 params.sack_delay = jiffies_to_msecs(
4592 params.sack_freq = asoc->sackfreq;
4595 params.sack_delay = 0;
4596 params.sack_freq = 1;
4599 /* Fetch socket values. */
4600 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4601 params.sack_delay = sp->sackdelay;
4602 params.sack_freq = sp->sackfreq;
4604 params.sack_delay = 0;
4605 params.sack_freq = 1;
4609 if (copy_to_user(optval, ¶ms, len))
4612 if (put_user(len, optlen))
4618 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4620 * Applications can specify protocol parameters for the default association
4621 * initialization. The option name argument to setsockopt() and getsockopt()
4624 * Setting initialization parameters is effective only on an unconnected
4625 * socket (for UDP-style sockets only future associations are effected
4626 * by the change). With TCP-style sockets, this option is inherited by
4627 * sockets derived from a listener socket.
4629 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4631 if (len < sizeof(struct sctp_initmsg))
4633 len = sizeof(struct sctp_initmsg);
4634 if (put_user(len, optlen))
4636 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4642 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4643 char __user *optval, int __user *optlen)
4645 struct sctp_association *asoc;
4647 struct sctp_getaddrs getaddrs;
4648 struct sctp_transport *from;
4650 union sctp_addr temp;
4651 struct sctp_sock *sp = sctp_sk(sk);
4656 if (len < sizeof(struct sctp_getaddrs))
4659 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4662 /* For UDP-style sockets, id specifies the association to query. */
4663 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4667 to = optval + offsetof(struct sctp_getaddrs,addrs);
4668 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4670 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4672 memcpy(&temp, &from->ipaddr, sizeof(temp));
4673 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4674 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4675 if (space_left < addrlen)
4677 if (copy_to_user(to, &temp, addrlen))
4681 space_left -= addrlen;
4684 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4686 bytes_copied = ((char __user *)to) - optval;
4687 if (put_user(bytes_copied, optlen))
4693 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4694 size_t space_left, int *bytes_copied)
4696 struct sctp_sockaddr_entry *addr;
4697 union sctp_addr temp;
4700 struct net *net = sock_net(sk);
4703 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4707 if ((PF_INET == sk->sk_family) &&
4708 (AF_INET6 == addr->a.sa.sa_family))
4710 if ((PF_INET6 == sk->sk_family) &&
4711 inet_v6_ipv6only(sk) &&
4712 (AF_INET == addr->a.sa.sa_family))
4714 memcpy(&temp, &addr->a, sizeof(temp));
4715 if (!temp.v4.sin_port)
4716 temp.v4.sin_port = htons(port);
4718 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4720 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4721 if (space_left < addrlen) {
4725 memcpy(to, &temp, addrlen);
4729 space_left -= addrlen;
4730 *bytes_copied += addrlen;
4738 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4739 char __user *optval, int __user *optlen)
4741 struct sctp_bind_addr *bp;
4742 struct sctp_association *asoc;
4744 struct sctp_getaddrs getaddrs;
4745 struct sctp_sockaddr_entry *addr;
4747 union sctp_addr temp;
4748 struct sctp_sock *sp = sctp_sk(sk);
4752 int bytes_copied = 0;
4756 if (len < sizeof(struct sctp_getaddrs))
4759 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4763 * For UDP-style sockets, id specifies the association to query.
4764 * If the id field is set to the value '0' then the locally bound
4765 * addresses are returned without regard to any particular
4768 if (0 == getaddrs.assoc_id) {
4769 bp = &sctp_sk(sk)->ep->base.bind_addr;
4771 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4774 bp = &asoc->base.bind_addr;
4777 to = optval + offsetof(struct sctp_getaddrs,addrs);
4778 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4780 addrs = kmalloc(space_left, GFP_KERNEL);
4784 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4785 * addresses from the global local address list.
4787 if (sctp_list_single_entry(&bp->address_list)) {
4788 addr = list_entry(bp->address_list.next,
4789 struct sctp_sockaddr_entry, list);
4790 if (sctp_is_any(sk, &addr->a)) {
4791 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4792 space_left, &bytes_copied);
4802 /* Protection on the bound address list is not needed since
4803 * in the socket option context we hold a socket lock and
4804 * thus the bound address list can't change.
4806 list_for_each_entry(addr, &bp->address_list, list) {
4807 memcpy(&temp, &addr->a, sizeof(temp));
4808 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4809 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4810 if (space_left < addrlen) {
4811 err = -ENOMEM; /*fixme: right error?*/
4814 memcpy(buf, &temp, addrlen);
4816 bytes_copied += addrlen;
4818 space_left -= addrlen;
4822 if (copy_to_user(to, addrs, bytes_copied)) {
4826 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4830 if (put_user(bytes_copied, optlen))
4837 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4839 * Requests that the local SCTP stack use the enclosed peer address as
4840 * the association primary. The enclosed address must be one of the
4841 * association peer's addresses.
4843 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4844 char __user *optval, int __user *optlen)
4846 struct sctp_prim prim;
4847 struct sctp_association *asoc;
4848 struct sctp_sock *sp = sctp_sk(sk);
4850 if (len < sizeof(struct sctp_prim))
4853 len = sizeof(struct sctp_prim);
4855 if (copy_from_user(&prim, optval, len))
4858 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4862 if (!asoc->peer.primary_path)
4865 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4866 asoc->peer.primary_path->af_specific->sockaddr_len);
4868 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4869 (union sctp_addr *)&prim.ssp_addr);
4871 if (put_user(len, optlen))
4873 if (copy_to_user(optval, &prim, len))
4880 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4882 * Requests that the local endpoint set the specified Adaptation Layer
4883 * Indication parameter for all future INIT and INIT-ACK exchanges.
4885 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4886 char __user *optval, int __user *optlen)
4888 struct sctp_setadaptation adaptation;
4890 if (len < sizeof(struct sctp_setadaptation))
4893 len = sizeof(struct sctp_setadaptation);
4895 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4897 if (put_user(len, optlen))
4899 if (copy_to_user(optval, &adaptation, len))
4907 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4909 * Applications that wish to use the sendto() system call may wish to
4910 * specify a default set of parameters that would normally be supplied
4911 * through the inclusion of ancillary data. This socket option allows
4912 * such an application to set the default sctp_sndrcvinfo structure.
4915 * The application that wishes to use this socket option simply passes
4916 * in to this call the sctp_sndrcvinfo structure defined in Section
4917 * 5.2.2) The input parameters accepted by this call include
4918 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4919 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4920 * to this call if the caller is using the UDP model.
4922 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4924 static int sctp_getsockopt_default_send_param(struct sock *sk,
4925 int len, char __user *optval,
4928 struct sctp_sndrcvinfo info;
4929 struct sctp_association *asoc;
4930 struct sctp_sock *sp = sctp_sk(sk);
4932 if (len < sizeof(struct sctp_sndrcvinfo))
4935 len = sizeof(struct sctp_sndrcvinfo);
4937 if (copy_from_user(&info, optval, len))
4940 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4941 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4945 info.sinfo_stream = asoc->default_stream;
4946 info.sinfo_flags = asoc->default_flags;
4947 info.sinfo_ppid = asoc->default_ppid;
4948 info.sinfo_context = asoc->default_context;
4949 info.sinfo_timetolive = asoc->default_timetolive;
4951 info.sinfo_stream = sp->default_stream;
4952 info.sinfo_flags = sp->default_flags;
4953 info.sinfo_ppid = sp->default_ppid;
4954 info.sinfo_context = sp->default_context;
4955 info.sinfo_timetolive = sp->default_timetolive;
4958 if (put_user(len, optlen))
4960 if (copy_to_user(optval, &info, len))
4968 * 7.1.5 SCTP_NODELAY
4970 * Turn on/off any Nagle-like algorithm. This means that packets are
4971 * generally sent as soon as possible and no unnecessary delays are
4972 * introduced, at the cost of more packets in the network. Expects an
4973 * integer boolean flag.
4976 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4977 char __user *optval, int __user *optlen)
4981 if (len < sizeof(int))
4985 val = (sctp_sk(sk)->nodelay == 1);
4986 if (put_user(len, optlen))
4988 if (copy_to_user(optval, &val, len))
4995 * 7.1.1 SCTP_RTOINFO
4997 * The protocol parameters used to initialize and bound retransmission
4998 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4999 * and modify these parameters.
5000 * All parameters are time values, in milliseconds. A value of 0, when
5001 * modifying the parameters, indicates that the current value should not
5005 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5006 char __user *optval,
5007 int __user *optlen) {
5008 struct sctp_rtoinfo rtoinfo;
5009 struct sctp_association *asoc;
5011 if (len < sizeof (struct sctp_rtoinfo))
5014 len = sizeof(struct sctp_rtoinfo);
5016 if (copy_from_user(&rtoinfo, optval, len))
5019 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5021 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5024 /* Values corresponding to the specific association. */
5026 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5027 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5028 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5030 /* Values corresponding to the endpoint. */
5031 struct sctp_sock *sp = sctp_sk(sk);
5033 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5034 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5035 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5038 if (put_user(len, optlen))
5041 if (copy_to_user(optval, &rtoinfo, len))
5049 * 7.1.2 SCTP_ASSOCINFO
5051 * This option is used to tune the maximum retransmission attempts
5052 * of the association.
5053 * Returns an error if the new association retransmission value is
5054 * greater than the sum of the retransmission value of the peer.
5055 * See [SCTP] for more information.
5058 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5059 char __user *optval,
5063 struct sctp_assocparams assocparams;
5064 struct sctp_association *asoc;
5065 struct list_head *pos;
5068 if (len < sizeof (struct sctp_assocparams))
5071 len = sizeof(struct sctp_assocparams);
5073 if (copy_from_user(&assocparams, optval, len))
5076 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5078 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5081 /* Values correspoinding to the specific association */
5083 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5084 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5085 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5086 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5088 list_for_each(pos, &asoc->peer.transport_addr_list) {
5092 assocparams.sasoc_number_peer_destinations = cnt;
5094 /* Values corresponding to the endpoint */
5095 struct sctp_sock *sp = sctp_sk(sk);
5097 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5098 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5099 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5100 assocparams.sasoc_cookie_life =
5101 sp->assocparams.sasoc_cookie_life;
5102 assocparams.sasoc_number_peer_destinations =
5104 sasoc_number_peer_destinations;
5107 if (put_user(len, optlen))
5110 if (copy_to_user(optval, &assocparams, len))
5117 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5119 * This socket option is a boolean flag which turns on or off mapped V4
5120 * addresses. If this option is turned on and the socket is type
5121 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5122 * If this option is turned off, then no mapping will be done of V4
5123 * addresses and a user will receive both PF_INET6 and PF_INET type
5124 * addresses on the socket.
5126 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5127 char __user *optval, int __user *optlen)
5130 struct sctp_sock *sp = sctp_sk(sk);
5132 if (len < sizeof(int))
5137 if (put_user(len, optlen))
5139 if (copy_to_user(optval, &val, len))
5146 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5147 * (chapter and verse is quoted at sctp_setsockopt_context())
5149 static int sctp_getsockopt_context(struct sock *sk, int len,
5150 char __user *optval, int __user *optlen)
5152 struct sctp_assoc_value params;
5153 struct sctp_sock *sp;
5154 struct sctp_association *asoc;
5156 if (len < sizeof(struct sctp_assoc_value))
5159 len = sizeof(struct sctp_assoc_value);
5161 if (copy_from_user(¶ms, optval, len))
5166 if (params.assoc_id != 0) {
5167 asoc = sctp_id2assoc(sk, params.assoc_id);
5170 params.assoc_value = asoc->default_rcv_context;
5172 params.assoc_value = sp->default_rcv_context;
5175 if (put_user(len, optlen))
5177 if (copy_to_user(optval, ¶ms, len))
5184 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5185 * This option will get or set the maximum size to put in any outgoing
5186 * SCTP DATA chunk. If a message is larger than this size it will be
5187 * fragmented by SCTP into the specified size. Note that the underlying
5188 * SCTP implementation may fragment into smaller sized chunks when the
5189 * PMTU of the underlying association is smaller than the value set by
5190 * the user. The default value for this option is '0' which indicates
5191 * the user is NOT limiting fragmentation and only the PMTU will effect
5192 * SCTP's choice of DATA chunk size. Note also that values set larger
5193 * than the maximum size of an IP datagram will effectively let SCTP
5194 * control fragmentation (i.e. the same as setting this option to 0).
5196 * The following structure is used to access and modify this parameter:
5198 * struct sctp_assoc_value {
5199 * sctp_assoc_t assoc_id;
5200 * uint32_t assoc_value;
5203 * assoc_id: This parameter is ignored for one-to-one style sockets.
5204 * For one-to-many style sockets this parameter indicates which
5205 * association the user is performing an action upon. Note that if
5206 * this field's value is zero then the endpoints default value is
5207 * changed (effecting future associations only).
5208 * assoc_value: This parameter specifies the maximum size in bytes.
5210 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5211 char __user *optval, int __user *optlen)
5213 struct sctp_assoc_value params;
5214 struct sctp_association *asoc;
5216 if (len == sizeof(int)) {
5217 pr_warn("Use of int in maxseg socket option deprecated\n");
5218 pr_warn("Use struct sctp_assoc_value instead\n");
5219 params.assoc_id = 0;
5220 } else if (len >= sizeof(struct sctp_assoc_value)) {
5221 len = sizeof(struct sctp_assoc_value);
5222 if (copy_from_user(¶ms, optval, sizeof(params)))
5227 asoc = sctp_id2assoc(sk, params.assoc_id);
5228 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5232 params.assoc_value = asoc->frag_point;
5234 params.assoc_value = sctp_sk(sk)->user_frag;
5236 if (put_user(len, optlen))
5238 if (len == sizeof(int)) {
5239 if (copy_to_user(optval, ¶ms.assoc_value, len))
5242 if (copy_to_user(optval, ¶ms, len))
5250 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5251 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5253 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5254 char __user *optval, int __user *optlen)
5258 if (len < sizeof(int))
5263 val = sctp_sk(sk)->frag_interleave;
5264 if (put_user(len, optlen))
5266 if (copy_to_user(optval, &val, len))
5273 * 7.1.25. Set or Get the sctp partial delivery point
5274 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5276 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5277 char __user *optval,
5282 if (len < sizeof(u32))
5287 val = sctp_sk(sk)->pd_point;
5288 if (put_user(len, optlen))
5290 if (copy_to_user(optval, &val, len))
5297 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5298 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5300 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5301 char __user *optval,
5304 struct sctp_assoc_value params;
5305 struct sctp_sock *sp;
5306 struct sctp_association *asoc;
5308 if (len == sizeof(int)) {
5309 pr_warn("Use of int in max_burst socket option deprecated\n");
5310 pr_warn("Use struct sctp_assoc_value instead\n");
5311 params.assoc_id = 0;
5312 } else if (len >= sizeof(struct sctp_assoc_value)) {
5313 len = sizeof(struct sctp_assoc_value);
5314 if (copy_from_user(¶ms, optval, len))
5321 if (params.assoc_id != 0) {
5322 asoc = sctp_id2assoc(sk, params.assoc_id);
5325 params.assoc_value = asoc->max_burst;
5327 params.assoc_value = sp->max_burst;
5329 if (len == sizeof(int)) {
5330 if (copy_to_user(optval, ¶ms.assoc_value, len))
5333 if (copy_to_user(optval, ¶ms, len))
5341 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5342 char __user *optval, int __user *optlen)
5344 struct net *net = sock_net(sk);
5345 struct sctp_hmacalgo __user *p = (void __user *)optval;
5346 struct sctp_hmac_algo_param *hmacs;
5350 if (!net->sctp.auth_enable)
5353 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5354 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5356 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5359 len = sizeof(struct sctp_hmacalgo) + data_len;
5360 num_idents = data_len / sizeof(u16);
5362 if (put_user(len, optlen))
5364 if (put_user(num_idents, &p->shmac_num_idents))
5366 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5371 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5372 char __user *optval, int __user *optlen)
5374 struct net *net = sock_net(sk);
5375 struct sctp_authkeyid val;
5376 struct sctp_association *asoc;
5378 if (!net->sctp.auth_enable)
5381 if (len < sizeof(struct sctp_authkeyid))
5383 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5386 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5387 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5391 val.scact_keynumber = asoc->active_key_id;
5393 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5395 len = sizeof(struct sctp_authkeyid);
5396 if (put_user(len, optlen))
5398 if (copy_to_user(optval, &val, len))
5404 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5405 char __user *optval, int __user *optlen)
5407 struct net *net = sock_net(sk);
5408 struct sctp_authchunks __user *p = (void __user *)optval;
5409 struct sctp_authchunks val;
5410 struct sctp_association *asoc;
5411 struct sctp_chunks_param *ch;
5415 if (!net->sctp.auth_enable)
5418 if (len < sizeof(struct sctp_authchunks))
5421 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5424 to = p->gauth_chunks;
5425 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5429 ch = asoc->peer.peer_chunks;
5433 /* See if the user provided enough room for all the data */
5434 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5435 if (len < num_chunks)
5438 if (copy_to_user(to, ch->chunks, num_chunks))
5441 len = sizeof(struct sctp_authchunks) + num_chunks;
5442 if (put_user(len, optlen)) return -EFAULT;
5443 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5448 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5449 char __user *optval, int __user *optlen)
5451 struct net *net = sock_net(sk);
5452 struct sctp_authchunks __user *p = (void __user *)optval;
5453 struct sctp_authchunks val;
5454 struct sctp_association *asoc;
5455 struct sctp_chunks_param *ch;
5459 if (!net->sctp.auth_enable)
5462 if (len < sizeof(struct sctp_authchunks))
5465 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5468 to = p->gauth_chunks;
5469 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5470 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5474 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5476 ch = sctp_sk(sk)->ep->auth_chunk_list;
5481 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5482 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5485 if (copy_to_user(to, ch->chunks, num_chunks))
5488 len = sizeof(struct sctp_authchunks) + num_chunks;
5489 if (put_user(len, optlen))
5491 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5498 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5499 * This option gets the current number of associations that are attached
5500 * to a one-to-many style socket. The option value is an uint32_t.
5502 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5503 char __user *optval, int __user *optlen)
5505 struct sctp_sock *sp = sctp_sk(sk);
5506 struct sctp_association *asoc;
5509 if (sctp_style(sk, TCP))
5512 if (len < sizeof(u32))
5517 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5521 if (put_user(len, optlen))
5523 if (copy_to_user(optval, &val, len))
5530 * 8.1.23 SCTP_AUTO_ASCONF
5531 * See the corresponding setsockopt entry as description
5533 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5534 char __user *optval, int __user *optlen)
5538 if (len < sizeof(int))
5542 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5544 if (put_user(len, optlen))
5546 if (copy_to_user(optval, &val, len))
5552 * 8.2.6. Get the Current Identifiers of Associations
5553 * (SCTP_GET_ASSOC_ID_LIST)
5555 * This option gets the current list of SCTP association identifiers of
5556 * the SCTP associations handled by a one-to-many style socket.
5558 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5559 char __user *optval, int __user *optlen)
5561 struct sctp_sock *sp = sctp_sk(sk);
5562 struct sctp_association *asoc;
5563 struct sctp_assoc_ids *ids;
5566 if (sctp_style(sk, TCP))
5569 if (len < sizeof(struct sctp_assoc_ids))
5572 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5576 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5579 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5581 ids = kmalloc(len, GFP_KERNEL);
5585 ids->gaids_number_of_ids = num;
5587 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5588 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5591 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5601 * SCTP_PEER_ADDR_THLDS
5603 * This option allows us to fetch the partially failed threshold for one or all
5604 * transports in an association. See Section 6.1 of:
5605 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5607 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5608 char __user *optval,
5612 struct sctp_paddrthlds val;
5613 struct sctp_transport *trans;
5614 struct sctp_association *asoc;
5616 if (len < sizeof(struct sctp_paddrthlds))
5618 len = sizeof(struct sctp_paddrthlds);
5619 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5622 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5623 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5627 val.spt_pathpfthld = asoc->pf_retrans;
5628 val.spt_pathmaxrxt = asoc->pathmaxrxt;
5630 trans = sctp_addr_id2transport(sk, &val.spt_address,
5635 val.spt_pathmaxrxt = trans->pathmaxrxt;
5636 val.spt_pathpfthld = trans->pf_retrans;
5639 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5646 * SCTP_GET_ASSOC_STATS
5648 * This option retrieves local per endpoint statistics. It is modeled
5649 * after OpenSolaris' implementation
5651 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
5652 char __user *optval,
5655 struct sctp_assoc_stats sas;
5656 struct sctp_association *asoc = NULL;
5658 /* User must provide at least the assoc id */
5659 if (len < sizeof(sctp_assoc_t))
5662 /* Allow the struct to grow and fill in as much as possible */
5663 len = min_t(size_t, len, sizeof(sas));
5665 if (copy_from_user(&sas, optval, len))
5668 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
5672 sas.sas_rtxchunks = asoc->stats.rtxchunks;
5673 sas.sas_gapcnt = asoc->stats.gapcnt;
5674 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
5675 sas.sas_osacks = asoc->stats.osacks;
5676 sas.sas_isacks = asoc->stats.isacks;
5677 sas.sas_octrlchunks = asoc->stats.octrlchunks;
5678 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
5679 sas.sas_oodchunks = asoc->stats.oodchunks;
5680 sas.sas_iodchunks = asoc->stats.iodchunks;
5681 sas.sas_ouodchunks = asoc->stats.ouodchunks;
5682 sas.sas_iuodchunks = asoc->stats.iuodchunks;
5683 sas.sas_idupchunks = asoc->stats.idupchunks;
5684 sas.sas_opackets = asoc->stats.opackets;
5685 sas.sas_ipackets = asoc->stats.ipackets;
5687 /* New high max rto observed, will return 0 if not a single
5688 * RTO update took place. obs_rto_ipaddr will be bogus
5691 sas.sas_maxrto = asoc->stats.max_obs_rto;
5692 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
5693 sizeof(struct sockaddr_storage));
5695 /* Mark beginning of a new observation period */
5696 asoc->stats.max_obs_rto = asoc->rto_min;
5698 if (put_user(len, optlen))
5701 SCTP_DEBUG_PRINTK("sctp_getsockopt_assoc_stat(%d): %d\n",
5702 len, sas.sas_assoc_id);
5704 if (copy_to_user(optval, &sas, len))
5710 static int sctp_getsockopt(struct sock *sk, int level, int optname,
5711 char __user *optval, int __user *optlen)
5716 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5719 /* I can hardly begin to describe how wrong this is. This is
5720 * so broken as to be worse than useless. The API draft
5721 * REALLY is NOT helpful here... I am not convinced that the
5722 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5723 * are at all well-founded.
5725 if (level != SOL_SCTP) {
5726 struct sctp_af *af = sctp_sk(sk)->pf->af;
5728 retval = af->getsockopt(sk, level, optname, optval, optlen);
5732 if (get_user(len, optlen))
5739 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5741 case SCTP_DISABLE_FRAGMENTS:
5742 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5746 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5748 case SCTP_AUTOCLOSE:
5749 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5751 case SCTP_SOCKOPT_PEELOFF:
5752 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5754 case SCTP_PEER_ADDR_PARAMS:
5755 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5758 case SCTP_DELAYED_SACK:
5759 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5763 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5765 case SCTP_GET_PEER_ADDRS:
5766 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5769 case SCTP_GET_LOCAL_ADDRS:
5770 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5773 case SCTP_SOCKOPT_CONNECTX3:
5774 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5776 case SCTP_DEFAULT_SEND_PARAM:
5777 retval = sctp_getsockopt_default_send_param(sk, len,
5780 case SCTP_PRIMARY_ADDR:
5781 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5784 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5787 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5789 case SCTP_ASSOCINFO:
5790 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5792 case SCTP_I_WANT_MAPPED_V4_ADDR:
5793 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5796 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5798 case SCTP_GET_PEER_ADDR_INFO:
5799 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5802 case SCTP_ADAPTATION_LAYER:
5803 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5807 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5809 case SCTP_FRAGMENT_INTERLEAVE:
5810 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5813 case SCTP_PARTIAL_DELIVERY_POINT:
5814 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5817 case SCTP_MAX_BURST:
5818 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5821 case SCTP_AUTH_CHUNK:
5822 case SCTP_AUTH_DELETE_KEY:
5823 retval = -EOPNOTSUPP;
5825 case SCTP_HMAC_IDENT:
5826 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5828 case SCTP_AUTH_ACTIVE_KEY:
5829 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5831 case SCTP_PEER_AUTH_CHUNKS:
5832 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5835 case SCTP_LOCAL_AUTH_CHUNKS:
5836 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5839 case SCTP_GET_ASSOC_NUMBER:
5840 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5842 case SCTP_GET_ASSOC_ID_LIST:
5843 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5845 case SCTP_AUTO_ASCONF:
5846 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5848 case SCTP_PEER_ADDR_THLDS:
5849 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
5851 case SCTP_GET_ASSOC_STATS:
5852 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
5855 retval = -ENOPROTOOPT;
5859 sctp_release_sock(sk);
5863 static void sctp_hash(struct sock *sk)
5868 static void sctp_unhash(struct sock *sk)
5873 /* Check if port is acceptable. Possibly find first available port.
5875 * The port hash table (contained in the 'global' SCTP protocol storage
5876 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5877 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5878 * list (the list number is the port number hashed out, so as you
5879 * would expect from a hash function, all the ports in a given list have
5880 * such a number that hashes out to the same list number; you were
5881 * expecting that, right?); so each list has a set of ports, with a
5882 * link to the socket (struct sock) that uses it, the port number and
5883 * a fastreuse flag (FIXME: NPI ipg).
5885 static struct sctp_bind_bucket *sctp_bucket_create(
5886 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
5888 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5890 struct sctp_bind_hashbucket *head; /* hash list */
5891 struct sctp_bind_bucket *pp;
5892 unsigned short snum;
5895 snum = ntohs(addr->v4.sin_port);
5897 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5898 sctp_local_bh_disable();
5901 /* Search for an available port. */
5902 int low, high, remaining, index;
5905 inet_get_local_port_range(&low, &high);
5906 remaining = (high - low) + 1;
5907 rover = net_random() % remaining + low;
5911 if ((rover < low) || (rover > high))
5913 if (inet_is_reserved_local_port(rover))
5915 index = sctp_phashfn(sock_net(sk), rover);
5916 head = &sctp_port_hashtable[index];
5917 sctp_spin_lock(&head->lock);
5918 sctp_for_each_hentry(pp, &head->chain)
5919 if ((pp->port == rover) &&
5920 net_eq(sock_net(sk), pp->net))
5924 sctp_spin_unlock(&head->lock);
5925 } while (--remaining > 0);
5927 /* Exhausted local port range during search? */
5932 /* OK, here is the one we will use. HEAD (the port
5933 * hash table list entry) is non-NULL and we hold it's
5938 /* We are given an specific port number; we verify
5939 * that it is not being used. If it is used, we will
5940 * exahust the search in the hash list corresponding
5941 * to the port number (snum) - we detect that with the
5942 * port iterator, pp being NULL.
5944 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
5945 sctp_spin_lock(&head->lock);
5946 sctp_for_each_hentry(pp, &head->chain) {
5947 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
5954 if (!hlist_empty(&pp->owner)) {
5955 /* We had a port hash table hit - there is an
5956 * available port (pp != NULL) and it is being
5957 * used by other socket (pp->owner not empty); that other
5958 * socket is going to be sk2.
5960 int reuse = sk->sk_reuse;
5963 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5964 if (pp->fastreuse && sk->sk_reuse &&
5965 sk->sk_state != SCTP_SS_LISTENING)
5968 /* Run through the list of sockets bound to the port
5969 * (pp->port) [via the pointers bind_next and
5970 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5971 * we get the endpoint they describe and run through
5972 * the endpoint's list of IP (v4 or v6) addresses,
5973 * comparing each of the addresses with the address of
5974 * the socket sk. If we find a match, then that means
5975 * that this port/socket (sk) combination are already
5978 sk_for_each_bound(sk2, &pp->owner) {
5979 struct sctp_endpoint *ep2;
5980 ep2 = sctp_sk(sk2)->ep;
5983 (reuse && sk2->sk_reuse &&
5984 sk2->sk_state != SCTP_SS_LISTENING))
5987 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5988 sctp_sk(sk2), sctp_sk(sk))) {
5993 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5996 /* If there was a hash table miss, create a new port. */
5998 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6001 /* In either case (hit or miss), make sure fastreuse is 1 only
6002 * if sk->sk_reuse is too (that is, if the caller requested
6003 * SO_REUSEADDR on this socket -sk-).
6005 if (hlist_empty(&pp->owner)) {
6006 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6010 } else if (pp->fastreuse &&
6011 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6014 /* We are set, so fill up all the data in the hash table
6015 * entry, tie the socket list information with the rest of the
6016 * sockets FIXME: Blurry, NPI (ipg).
6019 if (!sctp_sk(sk)->bind_hash) {
6020 inet_sk(sk)->inet_num = snum;
6021 sk_add_bind_node(sk, &pp->owner);
6022 sctp_sk(sk)->bind_hash = pp;
6027 sctp_spin_unlock(&head->lock);
6030 sctp_local_bh_enable();
6034 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6035 * port is requested.
6037 static int sctp_get_port(struct sock *sk, unsigned short snum)
6039 union sctp_addr addr;
6040 struct sctp_af *af = sctp_sk(sk)->pf->af;
6042 /* Set up a dummy address struct from the sk. */
6043 af->from_sk(&addr, sk);
6044 addr.v4.sin_port = htons(snum);
6046 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6047 return !!sctp_get_port_local(sk, &addr);
6051 * Move a socket to LISTENING state.
6053 static int sctp_listen_start(struct sock *sk, int backlog)
6055 struct sctp_sock *sp = sctp_sk(sk);
6056 struct sctp_endpoint *ep = sp->ep;
6057 struct crypto_hash *tfm = NULL;
6060 /* Allocate HMAC for generating cookie. */
6061 if (!sp->hmac && sp->sctp_hmac_alg) {
6062 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6063 tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
6065 net_info_ratelimited("failed to load transform for %s: %ld\n",
6066 sp->sctp_hmac_alg, PTR_ERR(tfm));
6069 sctp_sk(sk)->hmac = tfm;
6073 * If a bind() or sctp_bindx() is not called prior to a listen()
6074 * call that allows new associations to be accepted, the system
6075 * picks an ephemeral port and will choose an address set equivalent
6076 * to binding with a wildcard address.
6078 * This is not currently spelled out in the SCTP sockets
6079 * extensions draft, but follows the practice as seen in TCP
6083 sk->sk_state = SCTP_SS_LISTENING;
6084 if (!ep->base.bind_addr.port) {
6085 if (sctp_autobind(sk))
6088 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6089 sk->sk_state = SCTP_SS_CLOSED;
6094 sk->sk_max_ack_backlog = backlog;
6095 sctp_hash_endpoint(ep);
6100 * 4.1.3 / 5.1.3 listen()
6102 * By default, new associations are not accepted for UDP style sockets.
6103 * An application uses listen() to mark a socket as being able to
6104 * accept new associations.
6106 * On TCP style sockets, applications use listen() to ready the SCTP
6107 * endpoint for accepting inbound associations.
6109 * On both types of endpoints a backlog of '0' disables listening.
6111 * Move a socket to LISTENING state.
6113 int sctp_inet_listen(struct socket *sock, int backlog)
6115 struct sock *sk = sock->sk;
6116 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6119 if (unlikely(backlog < 0))
6124 /* Peeled-off sockets are not allowed to listen(). */
6125 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6128 if (sock->state != SS_UNCONNECTED)
6131 /* If backlog is zero, disable listening. */
6133 if (sctp_sstate(sk, CLOSED))
6137 sctp_unhash_endpoint(ep);
6138 sk->sk_state = SCTP_SS_CLOSED;
6140 sctp_sk(sk)->bind_hash->fastreuse = 1;
6144 /* If we are already listening, just update the backlog */
6145 if (sctp_sstate(sk, LISTENING))
6146 sk->sk_max_ack_backlog = backlog;
6148 err = sctp_listen_start(sk, backlog);
6155 sctp_release_sock(sk);
6160 * This function is done by modeling the current datagram_poll() and the
6161 * tcp_poll(). Note that, based on these implementations, we don't
6162 * lock the socket in this function, even though it seems that,
6163 * ideally, locking or some other mechanisms can be used to ensure
6164 * the integrity of the counters (sndbuf and wmem_alloc) used
6165 * in this place. We assume that we don't need locks either until proven
6168 * Another thing to note is that we include the Async I/O support
6169 * here, again, by modeling the current TCP/UDP code. We don't have
6170 * a good way to test with it yet.
6172 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6174 struct sock *sk = sock->sk;
6175 struct sctp_sock *sp = sctp_sk(sk);
6178 poll_wait(file, sk_sleep(sk), wait);
6180 /* A TCP-style listening socket becomes readable when the accept queue
6183 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6184 return (!list_empty(&sp->ep->asocs)) ?
6185 (POLLIN | POLLRDNORM) : 0;
6189 /* Is there any exceptional events? */
6190 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6192 sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0;
6193 if (sk->sk_shutdown & RCV_SHUTDOWN)
6194 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6195 if (sk->sk_shutdown == SHUTDOWN_MASK)
6198 /* Is it readable? Reconsider this code with TCP-style support. */
6199 if (!skb_queue_empty(&sk->sk_receive_queue))
6200 mask |= POLLIN | POLLRDNORM;
6202 /* The association is either gone or not ready. */
6203 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6206 /* Is it writable? */
6207 if (sctp_writeable(sk)) {
6208 mask |= POLLOUT | POLLWRNORM;
6210 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6212 * Since the socket is not locked, the buffer
6213 * might be made available after the writeable check and
6214 * before the bit is set. This could cause a lost I/O
6215 * signal. tcp_poll() has a race breaker for this race
6216 * condition. Based on their implementation, we put
6217 * in the following code to cover it as well.
6219 if (sctp_writeable(sk))
6220 mask |= POLLOUT | POLLWRNORM;
6225 /********************************************************************
6226 * 2nd Level Abstractions
6227 ********************************************************************/
6229 static struct sctp_bind_bucket *sctp_bucket_create(
6230 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6232 struct sctp_bind_bucket *pp;
6234 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6236 SCTP_DBG_OBJCNT_INC(bind_bucket);
6239 INIT_HLIST_HEAD(&pp->owner);
6241 hlist_add_head(&pp->node, &head->chain);
6246 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6247 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6249 if (pp && hlist_empty(&pp->owner)) {
6250 __hlist_del(&pp->node);
6251 kmem_cache_free(sctp_bucket_cachep, pp);
6252 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6256 /* Release this socket's reference to a local port. */
6257 static inline void __sctp_put_port(struct sock *sk)
6259 struct sctp_bind_hashbucket *head =
6260 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6261 inet_sk(sk)->inet_num)];
6262 struct sctp_bind_bucket *pp;
6264 sctp_spin_lock(&head->lock);
6265 pp = sctp_sk(sk)->bind_hash;
6266 __sk_del_bind_node(sk);
6267 sctp_sk(sk)->bind_hash = NULL;
6268 inet_sk(sk)->inet_num = 0;
6269 sctp_bucket_destroy(pp);
6270 sctp_spin_unlock(&head->lock);
6273 void sctp_put_port(struct sock *sk)
6275 sctp_local_bh_disable();
6276 __sctp_put_port(sk);
6277 sctp_local_bh_enable();
6281 * The system picks an ephemeral port and choose an address set equivalent
6282 * to binding with a wildcard address.
6283 * One of those addresses will be the primary address for the association.
6284 * This automatically enables the multihoming capability of SCTP.
6286 static int sctp_autobind(struct sock *sk)
6288 union sctp_addr autoaddr;
6292 /* Initialize a local sockaddr structure to INADDR_ANY. */
6293 af = sctp_sk(sk)->pf->af;
6295 port = htons(inet_sk(sk)->inet_num);
6296 af->inaddr_any(&autoaddr, port);
6298 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6301 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6304 * 4.2 The cmsghdr Structure *
6306 * When ancillary data is sent or received, any number of ancillary data
6307 * objects can be specified by the msg_control and msg_controllen members of
6308 * the msghdr structure, because each object is preceded by
6309 * a cmsghdr structure defining the object's length (the cmsg_len member).
6310 * Historically Berkeley-derived implementations have passed only one object
6311 * at a time, but this API allows multiple objects to be
6312 * passed in a single call to sendmsg() or recvmsg(). The following example
6313 * shows two ancillary data objects in a control buffer.
6315 * |<--------------------------- msg_controllen -------------------------->|
6318 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6320 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6323 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6325 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6328 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6329 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6331 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6333 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6340 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
6342 struct cmsghdr *cmsg;
6343 struct msghdr *my_msg = (struct msghdr *)msg;
6345 for (cmsg = CMSG_FIRSTHDR(msg);
6347 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6348 if (!CMSG_OK(my_msg, cmsg))
6351 /* Should we parse this header or ignore? */
6352 if (cmsg->cmsg_level != IPPROTO_SCTP)
6355 /* Strictly check lengths following example in SCM code. */
6356 switch (cmsg->cmsg_type) {
6358 /* SCTP Socket API Extension
6359 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6361 * This cmsghdr structure provides information for
6362 * initializing new SCTP associations with sendmsg().
6363 * The SCTP_INITMSG socket option uses this same data
6364 * structure. This structure is not used for
6367 * cmsg_level cmsg_type cmsg_data[]
6368 * ------------ ------------ ----------------------
6369 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6371 if (cmsg->cmsg_len !=
6372 CMSG_LEN(sizeof(struct sctp_initmsg)))
6374 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6378 /* SCTP Socket API Extension
6379 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6381 * This cmsghdr structure specifies SCTP options for
6382 * sendmsg() and describes SCTP header information
6383 * about a received message through recvmsg().
6385 * cmsg_level cmsg_type cmsg_data[]
6386 * ------------ ------------ ----------------------
6387 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6389 if (cmsg->cmsg_len !=
6390 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6394 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6396 /* Minimally, validate the sinfo_flags. */
6397 if (cmsgs->info->sinfo_flags &
6398 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6399 SCTP_ABORT | SCTP_EOF))
6411 * Wait for a packet..
6412 * Note: This function is the same function as in core/datagram.c
6413 * with a few modifications to make lksctp work.
6415 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6420 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6422 /* Socket errors? */
6423 error = sock_error(sk);
6427 if (!skb_queue_empty(&sk->sk_receive_queue))
6430 /* Socket shut down? */
6431 if (sk->sk_shutdown & RCV_SHUTDOWN)
6434 /* Sequenced packets can come disconnected. If so we report the
6439 /* Is there a good reason to think that we may receive some data? */
6440 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6443 /* Handle signals. */
6444 if (signal_pending(current))
6447 /* Let another process have a go. Since we are going to sleep
6448 * anyway. Note: This may cause odd behaviors if the message
6449 * does not fit in the user's buffer, but this seems to be the
6450 * only way to honor MSG_DONTWAIT realistically.
6452 sctp_release_sock(sk);
6453 *timeo_p = schedule_timeout(*timeo_p);
6457 finish_wait(sk_sleep(sk), &wait);
6461 error = sock_intr_errno(*timeo_p);
6464 finish_wait(sk_sleep(sk), &wait);
6469 /* Receive a datagram.
6470 * Note: This is pretty much the same routine as in core/datagram.c
6471 * with a few changes to make lksctp work.
6473 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6474 int noblock, int *err)
6477 struct sk_buff *skb;
6480 timeo = sock_rcvtimeo(sk, noblock);
6482 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6483 timeo, MAX_SCHEDULE_TIMEOUT);
6486 /* Again only user level code calls this function,
6487 * so nothing interrupt level
6488 * will suddenly eat the receive_queue.
6490 * Look at current nfs client by the way...
6491 * However, this function was correct in any case. 8)
6493 if (flags & MSG_PEEK) {
6494 spin_lock_bh(&sk->sk_receive_queue.lock);
6495 skb = skb_peek(&sk->sk_receive_queue);
6497 atomic_inc(&skb->users);
6498 spin_unlock_bh(&sk->sk_receive_queue.lock);
6500 skb = skb_dequeue(&sk->sk_receive_queue);
6506 /* Caller is allowed not to check sk->sk_err before calling. */
6507 error = sock_error(sk);
6511 if (sk->sk_shutdown & RCV_SHUTDOWN)
6514 /* User doesn't want to wait. */
6518 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6527 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6528 static void __sctp_write_space(struct sctp_association *asoc)
6530 struct sock *sk = asoc->base.sk;
6531 struct socket *sock = sk->sk_socket;
6533 if ((sctp_wspace(asoc) > 0) && sock) {
6534 if (waitqueue_active(&asoc->wait))
6535 wake_up_interruptible(&asoc->wait);
6537 if (sctp_writeable(sk)) {
6538 wait_queue_head_t *wq = sk_sleep(sk);
6540 if (wq && waitqueue_active(wq))
6541 wake_up_interruptible(wq);
6543 /* Note that we try to include the Async I/O support
6544 * here by modeling from the current TCP/UDP code.
6545 * We have not tested with it yet.
6547 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6548 sock_wake_async(sock,
6549 SOCK_WAKE_SPACE, POLL_OUT);
6554 /* Do accounting for the sndbuf space.
6555 * Decrement the used sndbuf space of the corresponding association by the
6556 * data size which was just transmitted(freed).
6558 static void sctp_wfree(struct sk_buff *skb)
6560 struct sctp_association *asoc;
6561 struct sctp_chunk *chunk;
6564 /* Get the saved chunk pointer. */
6565 chunk = *((struct sctp_chunk **)(skb->cb));
6568 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6569 sizeof(struct sk_buff) +
6570 sizeof(struct sctp_chunk);
6572 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6575 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6577 sk->sk_wmem_queued -= skb->truesize;
6578 sk_mem_uncharge(sk, skb->truesize);
6581 __sctp_write_space(asoc);
6583 sctp_association_put(asoc);
6586 /* Do accounting for the receive space on the socket.
6587 * Accounting for the association is done in ulpevent.c
6588 * We set this as a destructor for the cloned data skbs so that
6589 * accounting is done at the correct time.
6591 void sctp_sock_rfree(struct sk_buff *skb)
6593 struct sock *sk = skb->sk;
6594 struct sctp_ulpevent *event = sctp_skb2event(skb);
6596 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6599 * Mimic the behavior of sock_rfree
6601 sk_mem_uncharge(sk, event->rmem_len);
6605 /* Helper function to wait for space in the sndbuf. */
6606 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6609 struct sock *sk = asoc->base.sk;
6611 long current_timeo = *timeo_p;
6614 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6615 asoc, (long)(*timeo_p), msg_len);
6617 /* Increment the association's refcnt. */
6618 sctp_association_hold(asoc);
6620 /* Wait on the association specific sndbuf space. */
6622 prepare_to_wait_exclusive(&asoc->wait, &wait,
6623 TASK_INTERRUPTIBLE);
6626 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6629 if (signal_pending(current))
6630 goto do_interrupted;
6631 if (msg_len <= sctp_wspace(asoc))
6634 /* Let another process have a go. Since we are going
6637 sctp_release_sock(sk);
6638 current_timeo = schedule_timeout(current_timeo);
6639 BUG_ON(sk != asoc->base.sk);
6642 *timeo_p = current_timeo;
6646 finish_wait(&asoc->wait, &wait);
6648 /* Release the association's refcnt. */
6649 sctp_association_put(asoc);
6658 err = sock_intr_errno(*timeo_p);
6666 void sctp_data_ready(struct sock *sk, int len)
6668 struct socket_wq *wq;
6671 wq = rcu_dereference(sk->sk_wq);
6672 if (wq_has_sleeper(wq))
6673 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6674 POLLRDNORM | POLLRDBAND);
6675 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6679 /* If socket sndbuf has changed, wake up all per association waiters. */
6680 void sctp_write_space(struct sock *sk)
6682 struct sctp_association *asoc;
6684 /* Wake up the tasks in each wait queue. */
6685 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6686 __sctp_write_space(asoc);
6690 /* Is there any sndbuf space available on the socket?
6692 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6693 * associations on the same socket. For a UDP-style socket with
6694 * multiple associations, it is possible for it to be "unwriteable"
6695 * prematurely. I assume that this is acceptable because
6696 * a premature "unwriteable" is better than an accidental "writeable" which
6697 * would cause an unwanted block under certain circumstances. For the 1-1
6698 * UDP-style sockets or TCP-style sockets, this code should work.
6701 static int sctp_writeable(struct sock *sk)
6705 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6711 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6712 * returns immediately with EINPROGRESS.
6714 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6716 struct sock *sk = asoc->base.sk;
6718 long current_timeo = *timeo_p;
6721 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6724 /* Increment the association's refcnt. */
6725 sctp_association_hold(asoc);
6728 prepare_to_wait_exclusive(&asoc->wait, &wait,
6729 TASK_INTERRUPTIBLE);
6732 if (sk->sk_shutdown & RCV_SHUTDOWN)
6734 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6737 if (signal_pending(current))
6738 goto do_interrupted;
6740 if (sctp_state(asoc, ESTABLISHED))
6743 /* Let another process have a go. Since we are going
6746 sctp_release_sock(sk);
6747 current_timeo = schedule_timeout(current_timeo);
6750 *timeo_p = current_timeo;
6754 finish_wait(&asoc->wait, &wait);
6756 /* Release the association's refcnt. */
6757 sctp_association_put(asoc);
6762 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6765 err = -ECONNREFUSED;
6769 err = sock_intr_errno(*timeo_p);
6777 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6779 struct sctp_endpoint *ep;
6783 ep = sctp_sk(sk)->ep;
6787 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6788 TASK_INTERRUPTIBLE);
6790 if (list_empty(&ep->asocs)) {
6791 sctp_release_sock(sk);
6792 timeo = schedule_timeout(timeo);
6797 if (!sctp_sstate(sk, LISTENING))
6801 if (!list_empty(&ep->asocs))
6804 err = sock_intr_errno(timeo);
6805 if (signal_pending(current))
6813 finish_wait(sk_sleep(sk), &wait);
6818 static void sctp_wait_for_close(struct sock *sk, long timeout)
6823 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6824 if (list_empty(&sctp_sk(sk)->ep->asocs))
6826 sctp_release_sock(sk);
6827 timeout = schedule_timeout(timeout);
6829 } while (!signal_pending(current) && timeout);
6831 finish_wait(sk_sleep(sk), &wait);
6834 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6836 struct sk_buff *frag;
6841 /* Don't forget the fragments. */
6842 skb_walk_frags(skb, frag)
6843 sctp_skb_set_owner_r_frag(frag, sk);
6846 sctp_skb_set_owner_r(skb, sk);
6849 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6850 struct sctp_association *asoc)
6852 struct inet_sock *inet = inet_sk(sk);
6853 struct inet_sock *newinet;
6855 newsk->sk_type = sk->sk_type;
6856 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6857 newsk->sk_flags = sk->sk_flags;
6858 newsk->sk_no_check = sk->sk_no_check;
6859 newsk->sk_reuse = sk->sk_reuse;
6861 newsk->sk_shutdown = sk->sk_shutdown;
6862 newsk->sk_destruct = sctp_destruct_sock;
6863 newsk->sk_family = sk->sk_family;
6864 newsk->sk_protocol = IPPROTO_SCTP;
6865 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6866 newsk->sk_sndbuf = sk->sk_sndbuf;
6867 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6868 newsk->sk_lingertime = sk->sk_lingertime;
6869 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6870 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6872 newinet = inet_sk(newsk);
6874 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6875 * getsockname() and getpeername()
6877 newinet->inet_sport = inet->inet_sport;
6878 newinet->inet_saddr = inet->inet_saddr;
6879 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6880 newinet->inet_dport = htons(asoc->peer.port);
6881 newinet->pmtudisc = inet->pmtudisc;
6882 newinet->inet_id = asoc->next_tsn ^ jiffies;
6884 newinet->uc_ttl = inet->uc_ttl;
6885 newinet->mc_loop = 1;
6886 newinet->mc_ttl = 1;
6887 newinet->mc_index = 0;
6888 newinet->mc_list = NULL;
6891 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6892 * and its messages to the newsk.
6894 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6895 struct sctp_association *assoc,
6896 sctp_socket_type_t type)
6898 struct sctp_sock *oldsp = sctp_sk(oldsk);
6899 struct sctp_sock *newsp = sctp_sk(newsk);
6900 struct sctp_bind_bucket *pp; /* hash list port iterator */
6901 struct sctp_endpoint *newep = newsp->ep;
6902 struct sk_buff *skb, *tmp;
6903 struct sctp_ulpevent *event;
6904 struct sctp_bind_hashbucket *head;
6905 struct list_head tmplist;
6907 /* Migrate socket buffer sizes and all the socket level options to the
6910 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6911 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6912 /* Brute force copy old sctp opt. */
6913 if (oldsp->do_auto_asconf) {
6914 memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
6915 inet_sk_copy_descendant(newsk, oldsk);
6916 memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
6918 inet_sk_copy_descendant(newsk, oldsk);
6920 /* Restore the ep value that was overwritten with the above structure
6926 /* Hook this new socket in to the bind_hash list. */
6927 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
6928 inet_sk(oldsk)->inet_num)];
6929 sctp_local_bh_disable();
6930 sctp_spin_lock(&head->lock);
6931 pp = sctp_sk(oldsk)->bind_hash;
6932 sk_add_bind_node(newsk, &pp->owner);
6933 sctp_sk(newsk)->bind_hash = pp;
6934 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6935 sctp_spin_unlock(&head->lock);
6936 sctp_local_bh_enable();
6938 /* Copy the bind_addr list from the original endpoint to the new
6939 * endpoint so that we can handle restarts properly
6941 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6942 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6944 /* Move any messages in the old socket's receive queue that are for the
6945 * peeled off association to the new socket's receive queue.
6947 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6948 event = sctp_skb2event(skb);
6949 if (event->asoc == assoc) {
6950 __skb_unlink(skb, &oldsk->sk_receive_queue);
6951 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6952 sctp_skb_set_owner_r_frag(skb, newsk);
6956 /* Clean up any messages pending delivery due to partial
6957 * delivery. Three cases:
6958 * 1) No partial deliver; no work.
6959 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6960 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6962 skb_queue_head_init(&newsp->pd_lobby);
6963 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6965 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6966 struct sk_buff_head *queue;
6968 /* Decide which queue to move pd_lobby skbs to. */
6969 if (assoc->ulpq.pd_mode) {
6970 queue = &newsp->pd_lobby;
6972 queue = &newsk->sk_receive_queue;
6974 /* Walk through the pd_lobby, looking for skbs that
6975 * need moved to the new socket.
6977 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6978 event = sctp_skb2event(skb);
6979 if (event->asoc == assoc) {
6980 __skb_unlink(skb, &oldsp->pd_lobby);
6981 __skb_queue_tail(queue, skb);
6982 sctp_skb_set_owner_r_frag(skb, newsk);
6986 /* Clear up any skbs waiting for the partial
6987 * delivery to finish.
6989 if (assoc->ulpq.pd_mode)
6990 sctp_clear_pd(oldsk, NULL);
6994 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6995 sctp_skb_set_owner_r_frag(skb, newsk);
6997 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6998 sctp_skb_set_owner_r_frag(skb, newsk);
7000 /* Set the type of socket to indicate that it is peeled off from the
7001 * original UDP-style socket or created with the accept() call on a
7002 * TCP-style socket..
7006 /* Mark the new socket "in-use" by the user so that any packets
7007 * that may arrive on the association after we've moved it are
7008 * queued to the backlog. This prevents a potential race between
7009 * backlog processing on the old socket and new-packet processing
7010 * on the new socket.
7012 * The caller has just allocated newsk so we can guarantee that other
7013 * paths won't try to lock it and then oldsk.
7015 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7016 sctp_assoc_migrate(assoc, newsk);
7018 /* If the association on the newsk is already closed before accept()
7019 * is called, set RCV_SHUTDOWN flag.
7021 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
7022 newsk->sk_shutdown |= RCV_SHUTDOWN;
7024 newsk->sk_state = SCTP_SS_ESTABLISHED;
7025 sctp_release_sock(newsk);
7029 /* This proto struct describes the ULP interface for SCTP. */
7030 struct proto sctp_prot = {
7032 .owner = THIS_MODULE,
7033 .close = sctp_close,
7034 .connect = sctp_connect,
7035 .disconnect = sctp_disconnect,
7036 .accept = sctp_accept,
7037 .ioctl = sctp_ioctl,
7038 .init = sctp_init_sock,
7039 .destroy = sctp_destroy_sock,
7040 .shutdown = sctp_shutdown,
7041 .setsockopt = sctp_setsockopt,
7042 .getsockopt = sctp_getsockopt,
7043 .sendmsg = sctp_sendmsg,
7044 .recvmsg = sctp_recvmsg,
7046 .backlog_rcv = sctp_backlog_rcv,
7048 .unhash = sctp_unhash,
7049 .get_port = sctp_get_port,
7050 .obj_size = sizeof(struct sctp_sock),
7051 .sysctl_mem = sysctl_sctp_mem,
7052 .sysctl_rmem = sysctl_sctp_rmem,
7053 .sysctl_wmem = sysctl_sctp_wmem,
7054 .memory_pressure = &sctp_memory_pressure,
7055 .enter_memory_pressure = sctp_enter_memory_pressure,
7056 .memory_allocated = &sctp_memory_allocated,
7057 .sockets_allocated = &sctp_sockets_allocated,
7060 #if IS_ENABLED(CONFIG_IPV6)
7062 struct proto sctpv6_prot = {
7064 .owner = THIS_MODULE,
7065 .close = sctp_close,
7066 .connect = sctp_connect,
7067 .disconnect = sctp_disconnect,
7068 .accept = sctp_accept,
7069 .ioctl = sctp_ioctl,
7070 .init = sctp_init_sock,
7071 .destroy = sctp_destroy_sock,
7072 .shutdown = sctp_shutdown,
7073 .setsockopt = sctp_setsockopt,
7074 .getsockopt = sctp_getsockopt,
7075 .sendmsg = sctp_sendmsg,
7076 .recvmsg = sctp_recvmsg,
7078 .backlog_rcv = sctp_backlog_rcv,
7080 .unhash = sctp_unhash,
7081 .get_port = sctp_get_port,
7082 .obj_size = sizeof(struct sctp6_sock),
7083 .sysctl_mem = sysctl_sctp_mem,
7084 .sysctl_rmem = sysctl_sctp_rmem,
7085 .sysctl_wmem = sysctl_sctp_wmem,
7086 .memory_pressure = &sctp_memory_pressure,
7087 .enter_memory_pressure = sctp_enter_memory_pressure,
7088 .memory_allocated = &sctp_memory_allocated,
7089 .sockets_allocated = &sctp_sockets_allocated,
7091 #endif /* IS_ENABLED(CONFIG_IPV6) */