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>
76 #include <net/route.h>
78 #include <net/inet_common.h>
80 #include <linux/socket.h> /* for sa_family_t */
81 #include <linux/export.h>
83 #include <net/sctp/sctp.h>
84 #include <net/sctp/sm.h>
86 /* WARNING: Please do not remove the SCTP_STATIC attribute to
87 * any of the functions below as they are used to export functions
88 * used by a project regression testsuite.
91 /* Forward declarations for internal helper functions. */
92 static int sctp_writeable(struct sock *sk);
93 static void sctp_wfree(struct sk_buff *skb);
94 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
96 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
97 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
98 static int sctp_wait_for_accept(struct sock *sk, long timeo);
99 static void sctp_wait_for_close(struct sock *sk, long timeo);
100 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
101 union sctp_addr *addr, int len);
102 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
103 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
104 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
105 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
106 static int sctp_send_asconf(struct sctp_association *asoc,
107 struct sctp_chunk *chunk);
108 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
109 static int sctp_autobind(struct sock *sk);
110 static void sctp_sock_migrate(struct sock *, struct sock *,
111 struct sctp_association *, sctp_socket_type_t);
112 static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
114 extern struct kmem_cache *sctp_bucket_cachep;
115 extern long sysctl_sctp_mem[3];
116 extern int sysctl_sctp_rmem[3];
117 extern int sysctl_sctp_wmem[3];
119 static int sctp_memory_pressure;
120 static atomic_long_t sctp_memory_allocated;
121 struct percpu_counter sctp_sockets_allocated;
123 static void sctp_enter_memory_pressure(struct sock *sk)
125 sctp_memory_pressure = 1;
129 /* Get the sndbuf space available at the time on the association. */
130 static inline int sctp_wspace(struct sctp_association *asoc)
134 if (asoc->ep->sndbuf_policy)
135 amt = asoc->sndbuf_used;
137 amt = sk_wmem_alloc_get(asoc->base.sk);
139 if (amt >= asoc->base.sk->sk_sndbuf) {
140 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
143 amt = sk_stream_wspace(asoc->base.sk);
148 amt = asoc->base.sk->sk_sndbuf - amt;
153 /* Increment the used sndbuf space count of the corresponding association by
154 * the size of the outgoing data chunk.
155 * Also, set the skb destructor for sndbuf accounting later.
157 * Since it is always 1-1 between chunk and skb, and also a new skb is always
158 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
159 * destructor in the data chunk skb for the purpose of the sndbuf space
162 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
164 struct sctp_association *asoc = chunk->asoc;
165 struct sock *sk = asoc->base.sk;
167 /* The sndbuf space is tracked per association. */
168 sctp_association_hold(asoc);
170 skb_set_owner_w(chunk->skb, sk);
172 chunk->skb->destructor = sctp_wfree;
173 /* Save the chunk pointer in skb for sctp_wfree to use later. */
174 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
176 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
177 sizeof(struct sk_buff) +
178 sizeof(struct sctp_chunk);
180 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
181 sk->sk_wmem_queued += chunk->skb->truesize;
182 sk_mem_charge(sk, chunk->skb->truesize);
185 /* Verify that this is a valid address. */
186 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
191 /* Verify basic sockaddr. */
192 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
196 /* Is this a valid SCTP address? */
197 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
200 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
206 /* Look up the association by its id. If this is not a UDP-style
207 * socket, the ID field is always ignored.
209 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
211 struct sctp_association *asoc = NULL;
213 /* If this is not a UDP-style socket, assoc id should be ignored. */
214 if (!sctp_style(sk, UDP)) {
215 /* Return NULL if the socket state is not ESTABLISHED. It
216 * could be a TCP-style listening socket or a socket which
217 * hasn't yet called connect() to establish an association.
219 if (!sctp_sstate(sk, ESTABLISHED))
222 /* Get the first and the only association from the list. */
223 if (!list_empty(&sctp_sk(sk)->ep->asocs))
224 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
225 struct sctp_association, asocs);
229 /* Otherwise this is a UDP-style socket. */
230 if (!id || (id == (sctp_assoc_t)-1))
233 spin_lock_bh(&sctp_assocs_id_lock);
234 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
235 spin_unlock_bh(&sctp_assocs_id_lock);
237 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
243 /* Look up the transport from an address and an assoc id. If both address and
244 * id are specified, the associations matching the address and the id should be
247 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
248 struct sockaddr_storage *addr,
251 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
252 struct sctp_transport *transport;
253 union sctp_addr *laddr = (union sctp_addr *)addr;
255 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
262 id_asoc = sctp_id2assoc(sk, id);
263 if (id_asoc && (id_asoc != addr_asoc))
266 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
267 (union sctp_addr *)addr);
272 /* API 3.1.2 bind() - UDP Style Syntax
273 * The syntax of bind() is,
275 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
277 * sd - the socket descriptor returned by socket().
278 * addr - the address structure (struct sockaddr_in or struct
279 * sockaddr_in6 [RFC 2553]),
280 * addr_len - the size of the address structure.
282 SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
288 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
291 /* Disallow binding twice. */
292 if (!sctp_sk(sk)->ep->base.bind_addr.port)
293 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
298 sctp_release_sock(sk);
303 static long sctp_get_port_local(struct sock *, union sctp_addr *);
305 /* Verify this is a valid sockaddr. */
306 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
307 union sctp_addr *addr, int len)
311 /* Check minimum size. */
312 if (len < sizeof (struct sockaddr))
315 /* V4 mapped address are really of AF_INET family */
316 if (addr->sa.sa_family == AF_INET6 &&
317 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
318 if (!opt->pf->af_supported(AF_INET, opt))
321 /* Does this PF support this AF? */
322 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
326 /* If we get this far, af is valid. */
327 af = sctp_get_af_specific(addr->sa.sa_family);
329 if (len < af->sockaddr_len)
335 /* Bind a local address either to an endpoint or to an association. */
336 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
338 struct sctp_sock *sp = sctp_sk(sk);
339 struct sctp_endpoint *ep = sp->ep;
340 struct sctp_bind_addr *bp = &ep->base.bind_addr;
345 /* Common sockaddr verification. */
346 af = sctp_sockaddr_af(sp, addr, len);
348 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
353 snum = ntohs(addr->v4.sin_port);
355 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
356 ", port: %d, new port: %d, len: %d)\n",
362 /* PF specific bind() address verification. */
363 if (!sp->pf->bind_verify(sp, addr))
364 return -EADDRNOTAVAIL;
366 /* We must either be unbound, or bind to the same port.
367 * It's OK to allow 0 ports if we are already bound.
368 * We'll just inhert an already bound port in this case
373 else if (snum != bp->port) {
374 SCTP_DEBUG_PRINTK("sctp_do_bind:"
375 " New port %d does not match existing port "
376 "%d.\n", snum, bp->port);
381 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
384 /* See if the address matches any of the addresses we may have
385 * already bound before checking against other endpoints.
387 if (sctp_bind_addr_match(bp, addr, sp))
390 /* Make sure we are allowed to bind here.
391 * The function sctp_get_port_local() does duplicate address
394 addr->v4.sin_port = htons(snum);
395 if ((ret = sctp_get_port_local(sk, addr))) {
399 /* Refresh ephemeral port. */
401 bp->port = inet_sk(sk)->inet_num;
403 /* Add the address to the bind address list.
404 * Use GFP_ATOMIC since BHs will be disabled.
406 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
408 /* Copy back into socket for getsockname() use. */
410 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
411 af->to_sk_saddr(addr, sk);
417 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
419 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
420 * at any one time. If a sender, after sending an ASCONF chunk, decides
421 * it needs to transfer another ASCONF Chunk, it MUST wait until the
422 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
423 * subsequent ASCONF. Note this restriction binds each side, so at any
424 * time two ASCONF may be in-transit on any given association (one sent
425 * from each endpoint).
427 static int sctp_send_asconf(struct sctp_association *asoc,
428 struct sctp_chunk *chunk)
432 /* If there is an outstanding ASCONF chunk, queue it for later
435 if (asoc->addip_last_asconf) {
436 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
440 /* Hold the chunk until an ASCONF_ACK is received. */
441 sctp_chunk_hold(chunk);
442 retval = sctp_primitive_ASCONF(asoc, chunk);
444 sctp_chunk_free(chunk);
446 asoc->addip_last_asconf = chunk;
452 /* Add a list of addresses as bind addresses to local endpoint or
455 * Basically run through each address specified in the addrs/addrcnt
456 * array/length pair, determine if it is IPv6 or IPv4 and call
457 * sctp_do_bind() on it.
459 * If any of them fails, then the operation will be reversed and the
460 * ones that were added will be removed.
462 * Only sctp_setsockopt_bindx() is supposed to call this function.
464 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
469 struct sockaddr *sa_addr;
472 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
476 for (cnt = 0; cnt < addrcnt; cnt++) {
477 /* The list may contain either IPv4 or IPv6 address;
478 * determine the address length for walking thru the list.
481 af = sctp_get_af_specific(sa_addr->sa_family);
487 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
490 addr_buf += af->sockaddr_len;
494 /* Failed. Cleanup the ones that have been added */
496 sctp_bindx_rem(sk, addrs, cnt);
504 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
505 * associations that are part of the endpoint indicating that a list of local
506 * addresses are added to the endpoint.
508 * If any of the addresses is already in the bind address list of the
509 * association, we do not send the chunk for that association. But it will not
510 * affect other associations.
512 * Only sctp_setsockopt_bindx() is supposed to call this function.
514 static int sctp_send_asconf_add_ip(struct sock *sk,
515 struct sockaddr *addrs,
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 (!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 trans->rtt = trans->srtt = trans->rttvar = 0;
611 sctp_transport_route(trans, NULL,
612 sctp_sk(asoc->base.sk));
615 retval = sctp_send_asconf(asoc, chunk);
622 /* Remove a list of addresses from bind addresses list. Do not remove the
625 * Basically run through each address specified in the addrs/addrcnt
626 * array/length pair, determine if it is IPv6 or IPv4 and call
627 * sctp_del_bind() on it.
629 * If any of them fails, then the operation will be reversed and the
630 * ones that were removed will be added back.
632 * At least one address has to be left; if only one address is
633 * available, the operation will return -EBUSY.
635 * Only sctp_setsockopt_bindx() is supposed to call this function.
637 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
639 struct sctp_sock *sp = sctp_sk(sk);
640 struct sctp_endpoint *ep = sp->ep;
642 struct sctp_bind_addr *bp = &ep->base.bind_addr;
645 union sctp_addr *sa_addr;
648 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
652 for (cnt = 0; cnt < addrcnt; cnt++) {
653 /* If the bind address list is empty or if there is only one
654 * bind address, there is nothing more to be removed (we need
655 * at least one address here).
657 if (list_empty(&bp->address_list) ||
658 (sctp_list_single_entry(&bp->address_list))) {
664 af = sctp_get_af_specific(sa_addr->sa.sa_family);
670 if (!af->addr_valid(sa_addr, sp, NULL)) {
671 retval = -EADDRNOTAVAIL;
675 if (sa_addr->v4.sin_port &&
676 sa_addr->v4.sin_port != htons(bp->port)) {
681 if (!sa_addr->v4.sin_port)
682 sa_addr->v4.sin_port = htons(bp->port);
684 /* FIXME - There is probably a need to check if sk->sk_saddr and
685 * sk->sk_rcv_addr are currently set to one of the addresses to
686 * be removed. This is something which needs to be looked into
687 * when we are fixing the outstanding issues with multi-homing
688 * socket routing and failover schemes. Refer to comments in
689 * sctp_do_bind(). -daisy
691 retval = sctp_del_bind_addr(bp, sa_addr);
693 addr_buf += af->sockaddr_len;
696 /* Failed. Add the ones that has been removed back */
698 sctp_bindx_add(sk, addrs, cnt);
706 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
707 * the associations that are part of the endpoint indicating that a list of
708 * local addresses are removed from the endpoint.
710 * If any of the addresses is already in the bind address list of the
711 * association, we do not send the chunk for that association. But it will not
712 * affect other associations.
714 * Only sctp_setsockopt_bindx() is supposed to call this function.
716 static int sctp_send_asconf_del_ip(struct sock *sk,
717 struct sockaddr *addrs,
720 struct sctp_sock *sp;
721 struct sctp_endpoint *ep;
722 struct sctp_association *asoc;
723 struct sctp_transport *transport;
724 struct sctp_bind_addr *bp;
725 struct sctp_chunk *chunk;
726 union sctp_addr *laddr;
729 struct sctp_sockaddr_entry *saddr;
735 if (!sctp_addip_enable)
741 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
742 __func__, sk, addrs, addrcnt);
744 list_for_each_entry(asoc, &ep->asocs, asocs) {
746 if (!asoc->peer.asconf_capable)
749 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
752 if (!sctp_state(asoc, ESTABLISHED))
755 /* Check if any address in the packed array of addresses is
756 * not present in the bind address list of the association.
757 * If so, do not send the asconf chunk to its peer, but
758 * continue with other associations.
761 for (i = 0; i < addrcnt; i++) {
763 af = sctp_get_af_specific(laddr->v4.sin_family);
769 if (!sctp_assoc_lookup_laddr(asoc, laddr))
772 addr_buf += af->sockaddr_len;
777 /* Find one address in the association's bind address list
778 * that is not in the packed array of addresses. This is to
779 * make sure that we do not delete all the addresses in the
782 bp = &asoc->base.bind_addr;
783 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
785 if ((laddr == NULL) && (addrcnt == 1)) {
786 if (asoc->asconf_addr_del_pending)
788 asoc->asconf_addr_del_pending =
789 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
790 if (asoc->asconf_addr_del_pending == NULL) {
794 asoc->asconf_addr_del_pending->sa.sa_family =
796 asoc->asconf_addr_del_pending->v4.sin_port =
798 if (addrs->sa_family == AF_INET) {
799 struct sockaddr_in *sin;
801 sin = (struct sockaddr_in *)addrs;
802 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
803 } else if (addrs->sa_family == AF_INET6) {
804 struct sockaddr_in6 *sin6;
806 sin6 = (struct sockaddr_in6 *)addrs;
807 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
809 SCTP_DEBUG_PRINTK_IPADDR("send_asconf_del_ip: keep the last address asoc: %p ",
810 " at %p\n", asoc, asoc->asconf_addr_del_pending,
811 asoc->asconf_addr_del_pending);
812 asoc->src_out_of_asoc_ok = 1;
817 /* We do not need RCU protection throughout this loop
818 * because this is done under a socket lock from the
821 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
829 /* Reset use_as_src flag for the addresses in the bind address
830 * list that are to be deleted.
833 for (i = 0; i < addrcnt; i++) {
835 af = sctp_get_af_specific(laddr->v4.sin_family);
836 list_for_each_entry(saddr, &bp->address_list, list) {
837 if (sctp_cmp_addr_exact(&saddr->a, laddr))
838 saddr->state = SCTP_ADDR_DEL;
840 addr_buf += af->sockaddr_len;
843 /* Update the route and saddr entries for all the transports
844 * as some of the addresses in the bind address list are
845 * about to be deleted and cannot be used as source addresses.
847 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
849 dst_release(transport->dst);
850 sctp_transport_route(transport, NULL,
851 sctp_sk(asoc->base.sk));
855 /* We don't need to transmit ASCONF */
857 retval = sctp_send_asconf(asoc, chunk);
863 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
864 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
866 struct sock *sk = sctp_opt2sk(sp);
867 union sctp_addr *addr;
870 /* It is safe to write port space in caller. */
872 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
873 af = sctp_get_af_specific(addr->sa.sa_family);
876 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
879 if (addrw->state == SCTP_ADDR_NEW)
880 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
882 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
885 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
888 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
891 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
892 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
895 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
896 * Section 3.1.2 for this usage.
898 * addrs is a pointer to an array of one or more socket addresses. Each
899 * address is contained in its appropriate structure (i.e. struct
900 * sockaddr_in or struct sockaddr_in6) the family of the address type
901 * must be used to distinguish the address length (note that this
902 * representation is termed a "packed array" of addresses). The caller
903 * specifies the number of addresses in the array with addrcnt.
905 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
906 * -1, and sets errno to the appropriate error code.
908 * For SCTP, the port given in each socket address must be the same, or
909 * sctp_bindx() will fail, setting errno to EINVAL.
911 * The flags parameter is formed from the bitwise OR of zero or more of
912 * the following currently defined flags:
914 * SCTP_BINDX_ADD_ADDR
916 * SCTP_BINDX_REM_ADDR
918 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
919 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
920 * addresses from the association. The two flags are mutually exclusive;
921 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
922 * not remove all addresses from an association; sctp_bindx() will
923 * reject such an attempt with EINVAL.
925 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
926 * additional addresses with an endpoint after calling bind(). Or use
927 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
928 * socket is associated with so that no new association accepted will be
929 * associated with those addresses. If the endpoint supports dynamic
930 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
931 * endpoint to send the appropriate message to the peer to change the
932 * peers address lists.
934 * Adding and removing addresses from a connected association is
935 * optional functionality. Implementations that do not support this
936 * functionality should return EOPNOTSUPP.
938 * Basically do nothing but copying the addresses from user to kernel
939 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
940 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
943 * We don't use copy_from_user() for optimization: we first do the
944 * sanity checks (buffer size -fast- and access check-healthy
945 * pointer); if all of those succeed, then we can alloc the memory
946 * (expensive operation) needed to copy the data to kernel. Then we do
947 * the copying without checking the user space area
948 * (__copy_from_user()).
950 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
953 * sk The sk of the socket
954 * addrs The pointer to the addresses in user land
955 * addrssize Size of the addrs buffer
956 * op Operation to perform (add or remove, see the flags of
959 * Returns 0 if ok, <0 errno code on error.
961 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
962 struct sockaddr __user *addrs,
963 int addrs_size, int op)
965 struct sockaddr *kaddrs;
969 struct sockaddr *sa_addr;
973 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
974 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
976 if (unlikely(addrs_size <= 0))
979 /* Check the user passed a healthy pointer. */
980 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
983 /* Alloc space for the address array in kernel memory. */
984 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
985 if (unlikely(!kaddrs))
988 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
993 /* Walk through the addrs buffer and count the number of addresses. */
995 while (walk_size < addrs_size) {
996 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1002 af = sctp_get_af_specific(sa_addr->sa_family);
1004 /* If the address family is not supported or if this address
1005 * causes the address buffer to overflow return EINVAL.
1007 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1012 addr_buf += af->sockaddr_len;
1013 walk_size += af->sockaddr_len;
1018 case SCTP_BINDX_ADD_ADDR:
1019 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1022 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1025 case SCTP_BINDX_REM_ADDR:
1026 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1029 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1043 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1045 * Common routine for handling connect() and sctp_connectx().
1046 * Connect will come in with just a single address.
1048 static int __sctp_connect(struct sock* sk,
1049 struct sockaddr *kaddrs,
1051 sctp_assoc_t *assoc_id)
1053 struct sctp_sock *sp;
1054 struct sctp_endpoint *ep;
1055 struct sctp_association *asoc = NULL;
1056 struct sctp_association *asoc2;
1057 struct sctp_transport *transport;
1065 union sctp_addr *sa_addr = NULL;
1067 unsigned short port;
1068 unsigned int f_flags = 0;
1073 /* connect() cannot be done on a socket that is already in ESTABLISHED
1074 * state - UDP-style peeled off socket or a TCP-style socket that
1075 * is already connected.
1076 * It cannot be done even on a TCP-style listening socket.
1078 if (sctp_sstate(sk, ESTABLISHED) ||
1079 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1084 /* Walk through the addrs buffer and count the number of addresses. */
1086 while (walk_size < addrs_size) {
1087 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1093 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1095 /* If the address family is not supported or if this address
1096 * causes the address buffer to overflow return EINVAL.
1098 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1103 port = ntohs(sa_addr->v4.sin_port);
1105 /* Save current address so we can work with it */
1106 memcpy(&to, sa_addr, af->sockaddr_len);
1108 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1112 /* Make sure the destination port is correctly set
1115 if (asoc && asoc->peer.port && asoc->peer.port != port)
1119 /* Check if there already is a matching association on the
1120 * endpoint (other than the one created here).
1122 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1123 if (asoc2 && asoc2 != asoc) {
1124 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1131 /* If we could not find a matching association on the endpoint,
1132 * make sure that there is no peeled-off association matching
1133 * the peer address even on another socket.
1135 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1136 err = -EADDRNOTAVAIL;
1141 /* If a bind() or sctp_bindx() is not called prior to
1142 * an sctp_connectx() call, the system picks an
1143 * ephemeral port and will choose an address set
1144 * equivalent to binding with a wildcard address.
1146 if (!ep->base.bind_addr.port) {
1147 if (sctp_autobind(sk)) {
1153 * If an unprivileged user inherits a 1-many
1154 * style socket with open associations on a
1155 * privileged port, it MAY be permitted to
1156 * accept new associations, but it SHOULD NOT
1157 * be permitted to open new associations.
1159 if (ep->base.bind_addr.port < PROT_SOCK &&
1160 !capable(CAP_NET_BIND_SERVICE)) {
1166 scope = sctp_scope(&to);
1167 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1173 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1181 /* Prime the peer's transport structures. */
1182 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1190 addr_buf += af->sockaddr_len;
1191 walk_size += af->sockaddr_len;
1194 /* In case the user of sctp_connectx() wants an association
1195 * id back, assign one now.
1198 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1203 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1208 /* Initialize sk's dport and daddr for getpeername() */
1209 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1210 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1211 af->to_sk_daddr(sa_addr, sk);
1214 /* in-kernel sockets don't generally have a file allocated to them
1215 * if all they do is call sock_create_kern().
1217 if (sk->sk_socket->file)
1218 f_flags = sk->sk_socket->file->f_flags;
1220 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1222 err = sctp_wait_for_connect(asoc, &timeo);
1223 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1224 *assoc_id = asoc->assoc_id;
1226 /* Don't free association on exit. */
1231 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1232 " kaddrs: %p err: %d\n",
1235 /* sctp_primitive_ASSOCIATE may have added this association
1236 * To the hash table, try to unhash it, just in case, its a noop
1237 * if it wasn't hashed so we're safe
1239 sctp_unhash_established(asoc);
1240 sctp_association_free(asoc);
1245 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1248 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1249 * sctp_assoc_t *asoc);
1251 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1252 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1253 * or IPv6 addresses.
1255 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1256 * Section 3.1.2 for this usage.
1258 * addrs is a pointer to an array of one or more socket addresses. Each
1259 * address is contained in its appropriate structure (i.e. struct
1260 * sockaddr_in or struct sockaddr_in6) the family of the address type
1261 * must be used to distengish the address length (note that this
1262 * representation is termed a "packed array" of addresses). The caller
1263 * specifies the number of addresses in the array with addrcnt.
1265 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1266 * the association id of the new association. On failure, sctp_connectx()
1267 * returns -1, and sets errno to the appropriate error code. The assoc_id
1268 * is not touched by the kernel.
1270 * For SCTP, the port given in each socket address must be the same, or
1271 * sctp_connectx() will fail, setting errno to EINVAL.
1273 * An application can use sctp_connectx to initiate an association with
1274 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1275 * allows a caller to specify multiple addresses at which a peer can be
1276 * reached. The way the SCTP stack uses the list of addresses to set up
1277 * the association is implementation dependent. This function only
1278 * specifies that the stack will try to make use of all the addresses in
1279 * the list when needed.
1281 * Note that the list of addresses passed in is only used for setting up
1282 * the association. It does not necessarily equal the set of addresses
1283 * the peer uses for the resulting association. If the caller wants to
1284 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1285 * retrieve them after the association has been set up.
1287 * Basically do nothing but copying the addresses from user to kernel
1288 * land and invoking either sctp_connectx(). This is used for tunneling
1289 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1291 * We don't use copy_from_user() for optimization: we first do the
1292 * sanity checks (buffer size -fast- and access check-healthy
1293 * pointer); if all of those succeed, then we can alloc the memory
1294 * (expensive operation) needed to copy the data to kernel. Then we do
1295 * the copying without checking the user space area
1296 * (__copy_from_user()).
1298 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1301 * sk The sk of the socket
1302 * addrs The pointer to the addresses in user land
1303 * addrssize Size of the addrs buffer
1305 * Returns >=0 if ok, <0 errno code on error.
1307 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1308 struct sockaddr __user *addrs,
1310 sctp_assoc_t *assoc_id)
1313 struct sockaddr *kaddrs;
1315 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1316 __func__, sk, addrs, addrs_size);
1318 if (unlikely(addrs_size <= 0))
1321 /* Check the user passed a healthy pointer. */
1322 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1325 /* Alloc space for the address array in kernel memory. */
1326 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1327 if (unlikely(!kaddrs))
1330 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1333 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1342 * This is an older interface. It's kept for backward compatibility
1343 * to the option that doesn't provide association id.
1345 SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1346 struct sockaddr __user *addrs,
1349 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1353 * New interface for the API. The since the API is done with a socket
1354 * option, to make it simple we feed back the association id is as a return
1355 * indication to the call. Error is always negative and association id is
1358 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1359 struct sockaddr __user *addrs,
1362 sctp_assoc_t assoc_id = 0;
1365 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1374 * New (hopefully final) interface for the API.
1375 * We use the sctp_getaddrs_old structure so that use-space library
1376 * can avoid any unnecessary allocations. The only defferent part
1377 * is that we store the actual length of the address buffer into the
1378 * addrs_num structure member. That way we can re-use the existing
1381 SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1382 char __user *optval,
1385 struct sctp_getaddrs_old param;
1386 sctp_assoc_t assoc_id = 0;
1389 if (len < sizeof(param))
1392 if (copy_from_user(¶m, optval, sizeof(param)))
1395 err = __sctp_setsockopt_connectx(sk,
1396 (struct sockaddr __user *)param.addrs,
1397 param.addr_num, &assoc_id);
1399 if (err == 0 || err == -EINPROGRESS) {
1400 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1402 if (put_user(sizeof(assoc_id), optlen))
1409 /* API 3.1.4 close() - UDP Style Syntax
1410 * Applications use close() to perform graceful shutdown (as described in
1411 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1412 * by a UDP-style socket.
1416 * ret = close(int sd);
1418 * sd - the socket descriptor of the associations to be closed.
1420 * To gracefully shutdown a specific association represented by the
1421 * UDP-style socket, an application should use the sendmsg() call,
1422 * passing no user data, but including the appropriate flag in the
1423 * ancillary data (see Section xxxx).
1425 * If sd in the close() call is a branched-off socket representing only
1426 * one association, the shutdown is performed on that association only.
1428 * 4.1.6 close() - TCP Style Syntax
1430 * Applications use close() to gracefully close down an association.
1434 * int close(int sd);
1436 * sd - the socket descriptor of the association to be closed.
1438 * After an application calls close() on a socket descriptor, no further
1439 * socket operations will succeed on that descriptor.
1441 * API 7.1.4 SO_LINGER
1443 * An application using the TCP-style socket can use this option to
1444 * perform the SCTP ABORT primitive. The linger option structure is:
1447 * int l_onoff; // option on/off
1448 * int l_linger; // linger time
1451 * To enable the option, set l_onoff to 1. If the l_linger value is set
1452 * to 0, calling close() is the same as the ABORT primitive. If the
1453 * value is set to a negative value, the setsockopt() call will return
1454 * an error. If the value is set to a positive value linger_time, the
1455 * close() can be blocked for at most linger_time ms. If the graceful
1456 * shutdown phase does not finish during this period, close() will
1457 * return but the graceful shutdown phase continues in the system.
1459 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1461 struct sctp_endpoint *ep;
1462 struct sctp_association *asoc;
1463 struct list_head *pos, *temp;
1464 unsigned int data_was_unread;
1466 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1469 sk->sk_shutdown = SHUTDOWN_MASK;
1470 sk->sk_state = SCTP_SS_CLOSING;
1472 ep = sctp_sk(sk)->ep;
1474 /* Clean up any skbs sitting on the receive queue. */
1475 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1476 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1478 /* Walk all associations on an endpoint. */
1479 list_for_each_safe(pos, temp, &ep->asocs) {
1480 asoc = list_entry(pos, struct sctp_association, asocs);
1482 if (sctp_style(sk, TCP)) {
1483 /* A closed association can still be in the list if
1484 * it belongs to a TCP-style listening socket that is
1485 * not yet accepted. If so, free it. If not, send an
1486 * ABORT or SHUTDOWN based on the linger options.
1488 if (sctp_state(asoc, CLOSED)) {
1489 sctp_unhash_established(asoc);
1490 sctp_association_free(asoc);
1495 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1496 !skb_queue_empty(&asoc->ulpq.reasm) ||
1497 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1498 struct sctp_chunk *chunk;
1500 chunk = sctp_make_abort_user(asoc, NULL, 0);
1502 sctp_primitive_ABORT(asoc, chunk);
1504 sctp_primitive_SHUTDOWN(asoc, NULL);
1507 /* On a TCP-style socket, block for at most linger_time if set. */
1508 if (sctp_style(sk, TCP) && timeout)
1509 sctp_wait_for_close(sk, timeout);
1511 /* This will run the backlog queue. */
1512 sctp_release_sock(sk);
1514 /* Supposedly, no process has access to the socket, but
1515 * the net layers still may.
1517 sctp_local_bh_disable();
1518 sctp_bh_lock_sock(sk);
1520 /* Hold the sock, since sk_common_release() will put sock_put()
1521 * and we have just a little more cleanup.
1524 sk_common_release(sk);
1526 sctp_bh_unlock_sock(sk);
1527 sctp_local_bh_enable();
1531 SCTP_DBG_OBJCNT_DEC(sock);
1534 /* Handle EPIPE error. */
1535 static int sctp_error(struct sock *sk, int flags, int err)
1538 err = sock_error(sk) ? : -EPIPE;
1539 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1540 send_sig(SIGPIPE, current, 0);
1544 /* API 3.1.3 sendmsg() - UDP Style Syntax
1546 * An application uses sendmsg() and recvmsg() calls to transmit data to
1547 * and receive data from its peer.
1549 * ssize_t sendmsg(int socket, const struct msghdr *message,
1552 * socket - the socket descriptor of the endpoint.
1553 * message - pointer to the msghdr structure which contains a single
1554 * user message and possibly some ancillary data.
1556 * See Section 5 for complete description of the data
1559 * flags - flags sent or received with the user message, see Section
1560 * 5 for complete description of the flags.
1562 * Note: This function could use a rewrite especially when explicit
1563 * connect support comes in.
1565 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1567 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1569 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1570 struct msghdr *msg, size_t msg_len)
1572 struct sctp_sock *sp;
1573 struct sctp_endpoint *ep;
1574 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1575 struct sctp_transport *transport, *chunk_tp;
1576 struct sctp_chunk *chunk;
1578 struct sockaddr *msg_name = NULL;
1579 struct sctp_sndrcvinfo default_sinfo;
1580 struct sctp_sndrcvinfo *sinfo;
1581 struct sctp_initmsg *sinit;
1582 sctp_assoc_t associd = 0;
1583 sctp_cmsgs_t cmsgs = { NULL };
1587 __u16 sinfo_flags = 0;
1588 struct sctp_datamsg *datamsg;
1589 int msg_flags = msg->msg_flags;
1591 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1598 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1600 /* We cannot send a message over a TCP-style listening socket. */
1601 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1606 /* Parse out the SCTP CMSGs. */
1607 err = sctp_msghdr_parse(msg, &cmsgs);
1610 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1614 /* Fetch the destination address for this packet. This
1615 * address only selects the association--it is not necessarily
1616 * the address we will send to.
1617 * For a peeled-off socket, msg_name is ignored.
1619 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1620 int msg_namelen = msg->msg_namelen;
1622 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1627 if (msg_namelen > sizeof(to))
1628 msg_namelen = sizeof(to);
1629 memcpy(&to, msg->msg_name, msg_namelen);
1630 msg_name = msg->msg_name;
1636 /* Did the user specify SNDRCVINFO? */
1638 sinfo_flags = sinfo->sinfo_flags;
1639 associd = sinfo->sinfo_assoc_id;
1642 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1643 msg_len, sinfo_flags);
1645 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1646 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1651 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1652 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1653 * If SCTP_ABORT is set, the message length could be non zero with
1654 * the msg_iov set to the user abort reason.
1656 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1657 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1662 /* If SCTP_ADDR_OVER is set, there must be an address
1663 * specified in msg_name.
1665 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1672 SCTP_DEBUG_PRINTK("About to look up association.\n");
1676 /* If a msg_name has been specified, assume this is to be used. */
1678 /* Look for a matching association on the endpoint. */
1679 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1681 /* If we could not find a matching association on the
1682 * endpoint, make sure that it is not a TCP-style
1683 * socket that already has an association or there is
1684 * no peeled-off association on another socket.
1686 if ((sctp_style(sk, TCP) &&
1687 sctp_sstate(sk, ESTABLISHED)) ||
1688 sctp_endpoint_is_peeled_off(ep, &to)) {
1689 err = -EADDRNOTAVAIL;
1694 asoc = sctp_id2assoc(sk, associd);
1702 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1704 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1705 * socket that has an association in CLOSED state. This can
1706 * happen when an accepted socket has an association that is
1709 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1714 if (sinfo_flags & SCTP_EOF) {
1715 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1717 sctp_primitive_SHUTDOWN(asoc, NULL);
1721 if (sinfo_flags & SCTP_ABORT) {
1723 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1729 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1730 sctp_primitive_ABORT(asoc, chunk);
1736 /* Do we need to create the association? */
1738 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1740 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1745 /* Check for invalid stream against the stream counts,
1746 * either the default or the user specified stream counts.
1749 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1750 /* Check against the defaults. */
1751 if (sinfo->sinfo_stream >=
1752 sp->initmsg.sinit_num_ostreams) {
1757 /* Check against the requested. */
1758 if (sinfo->sinfo_stream >=
1759 sinit->sinit_num_ostreams) {
1767 * API 3.1.2 bind() - UDP Style Syntax
1768 * If a bind() or sctp_bindx() is not called prior to a
1769 * sendmsg() call that initiates a new association, the
1770 * system picks an ephemeral port and will choose an address
1771 * set equivalent to binding with a wildcard address.
1773 if (!ep->base.bind_addr.port) {
1774 if (sctp_autobind(sk)) {
1780 * If an unprivileged user inherits a one-to-many
1781 * style socket with open associations on a privileged
1782 * port, it MAY be permitted to accept new associations,
1783 * but it SHOULD NOT be permitted to open new
1786 if (ep->base.bind_addr.port < PROT_SOCK &&
1787 !capable(CAP_NET_BIND_SERVICE)) {
1793 scope = sctp_scope(&to);
1794 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1800 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1806 /* If the SCTP_INIT ancillary data is specified, set all
1807 * the association init values accordingly.
1810 if (sinit->sinit_num_ostreams) {
1811 asoc->c.sinit_num_ostreams =
1812 sinit->sinit_num_ostreams;
1814 if (sinit->sinit_max_instreams) {
1815 asoc->c.sinit_max_instreams =
1816 sinit->sinit_max_instreams;
1818 if (sinit->sinit_max_attempts) {
1819 asoc->max_init_attempts
1820 = sinit->sinit_max_attempts;
1822 if (sinit->sinit_max_init_timeo) {
1823 asoc->max_init_timeo =
1824 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1828 /* Prime the peer's transport structures. */
1829 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1836 /* ASSERT: we have a valid association at this point. */
1837 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1840 /* If the user didn't specify SNDRCVINFO, make up one with
1843 memset(&default_sinfo, 0, sizeof(default_sinfo));
1844 default_sinfo.sinfo_stream = asoc->default_stream;
1845 default_sinfo.sinfo_flags = asoc->default_flags;
1846 default_sinfo.sinfo_ppid = asoc->default_ppid;
1847 default_sinfo.sinfo_context = asoc->default_context;
1848 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1849 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1850 sinfo = &default_sinfo;
1853 /* API 7.1.7, the sndbuf size per association bounds the
1854 * maximum size of data that can be sent in a single send call.
1856 if (msg_len > sk->sk_sndbuf) {
1861 if (asoc->pmtu_pending)
1862 sctp_assoc_pending_pmtu(sk, asoc);
1864 /* If fragmentation is disabled and the message length exceeds the
1865 * association fragmentation point, return EMSGSIZE. The I-D
1866 * does not specify what this error is, but this looks like
1869 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1874 /* Check for invalid stream. */
1875 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1880 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1881 if (!sctp_wspace(asoc)) {
1882 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1887 /* If an address is passed with the sendto/sendmsg call, it is used
1888 * to override the primary destination address in the TCP model, or
1889 * when SCTP_ADDR_OVER flag is set in the UDP model.
1891 if ((sctp_style(sk, TCP) && msg_name) ||
1892 (sinfo_flags & SCTP_ADDR_OVER)) {
1893 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1901 /* Auto-connect, if we aren't connected already. */
1902 if (sctp_state(asoc, CLOSED)) {
1903 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1906 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1909 /* Break the message into multiple chunks of maximum size. */
1910 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1916 /* Now send the (possibly) fragmented message. */
1917 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1918 sctp_chunk_hold(chunk);
1920 /* Do accounting for the write space. */
1921 sctp_set_owner_w(chunk);
1923 chunk->transport = chunk_tp;
1926 /* Send it to the lower layers. Note: all chunks
1927 * must either fail or succeed. The lower layer
1928 * works that way today. Keep it that way or this
1931 err = sctp_primitive_SEND(asoc, datamsg);
1932 /* Did the lower layer accept the chunk? */
1934 sctp_datamsg_free(datamsg);
1936 sctp_datamsg_put(datamsg);
1938 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1945 /* If we are already past ASSOCIATE, the lower
1946 * layers are responsible for association cleanup.
1952 sctp_unhash_established(asoc);
1953 sctp_association_free(asoc);
1956 sctp_release_sock(sk);
1959 return sctp_error(sk, msg_flags, err);
1966 err = sock_error(sk);
1976 /* This is an extended version of skb_pull() that removes the data from the
1977 * start of a skb even when data is spread across the list of skb's in the
1978 * frag_list. len specifies the total amount of data that needs to be removed.
1979 * when 'len' bytes could be removed from the skb, it returns 0.
1980 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1981 * could not be removed.
1983 static int sctp_skb_pull(struct sk_buff *skb, int len)
1985 struct sk_buff *list;
1986 int skb_len = skb_headlen(skb);
1989 if (len <= skb_len) {
1990 __skb_pull(skb, len);
1994 __skb_pull(skb, skb_len);
1996 skb_walk_frags(skb, list) {
1997 rlen = sctp_skb_pull(list, len);
1998 skb->len -= (len-rlen);
1999 skb->data_len -= (len-rlen);
2010 /* API 3.1.3 recvmsg() - UDP Style Syntax
2012 * ssize_t recvmsg(int socket, struct msghdr *message,
2015 * socket - the socket descriptor of the endpoint.
2016 * message - pointer to the msghdr structure which contains a single
2017 * user message and possibly some ancillary data.
2019 * See Section 5 for complete description of the data
2022 * flags - flags sent or received with the user message, see Section
2023 * 5 for complete description of the flags.
2025 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2027 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2028 struct msghdr *msg, size_t len, int noblock,
2029 int flags, int *addr_len)
2031 struct sctp_ulpevent *event = NULL;
2032 struct sctp_sock *sp = sctp_sk(sk);
2033 struct sk_buff *skb;
2038 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
2039 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
2040 "len", len, "knoblauch", noblock,
2041 "flags", flags, "addr_len", addr_len);
2045 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2050 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2054 /* Get the total length of the skb including any skb's in the
2063 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2065 event = sctp_skb2event(skb);
2070 sock_recv_ts_and_drops(msg, sk, skb);
2071 if (sctp_ulpevent_is_notification(event)) {
2072 msg->msg_flags |= MSG_NOTIFICATION;
2073 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2075 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2078 /* Check if we allow SCTP_SNDRCVINFO. */
2079 if (sp->subscribe.sctp_data_io_event)
2080 sctp_ulpevent_read_sndrcvinfo(event, msg);
2082 /* FIXME: we should be calling IP/IPv6 layers. */
2083 if (sk->sk_protinfo.af_inet.cmsg_flags)
2084 ip_cmsg_recv(msg, skb);
2089 /* If skb's length exceeds the user's buffer, update the skb and
2090 * push it back to the receive_queue so that the next call to
2091 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2093 if (skb_len > copied) {
2094 msg->msg_flags &= ~MSG_EOR;
2095 if (flags & MSG_PEEK)
2097 sctp_skb_pull(skb, copied);
2098 skb_queue_head(&sk->sk_receive_queue, skb);
2100 /* When only partial message is copied to the user, increase
2101 * rwnd by that amount. If all the data in the skb is read,
2102 * rwnd is updated when the event is freed.
2104 if (!sctp_ulpevent_is_notification(event))
2105 sctp_assoc_rwnd_increase(event->asoc, copied);
2107 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2108 (event->msg_flags & MSG_EOR))
2109 msg->msg_flags |= MSG_EOR;
2111 msg->msg_flags &= ~MSG_EOR;
2114 if (flags & MSG_PEEK) {
2115 /* Release the skb reference acquired after peeking the skb in
2116 * sctp_skb_recv_datagram().
2120 /* Free the event which includes releasing the reference to
2121 * the owner of the skb, freeing the skb and updating the
2124 sctp_ulpevent_free(event);
2127 sctp_release_sock(sk);
2131 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2133 * This option is a on/off flag. If enabled no SCTP message
2134 * fragmentation will be performed. Instead if a message being sent
2135 * exceeds the current PMTU size, the message will NOT be sent and
2136 * instead a error will be indicated to the user.
2138 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2139 char __user *optval,
2140 unsigned int optlen)
2144 if (optlen < sizeof(int))
2147 if (get_user(val, (int __user *)optval))
2150 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2155 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2156 unsigned int optlen)
2158 struct sctp_association *asoc;
2159 struct sctp_ulpevent *event;
2161 if (optlen > sizeof(struct sctp_event_subscribe))
2163 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2167 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2168 * if there is no data to be sent or retransmit, the stack will
2169 * immediately send up this notification.
2171 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2172 &sctp_sk(sk)->subscribe)) {
2173 asoc = sctp_id2assoc(sk, 0);
2175 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2176 event = sctp_ulpevent_make_sender_dry_event(asoc,
2181 sctp_ulpq_tail_event(&asoc->ulpq, event);
2188 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2190 * This socket option is applicable to the UDP-style socket only. When
2191 * set it will cause associations that are idle for more than the
2192 * specified number of seconds to automatically close. An association
2193 * being idle is defined an association that has NOT sent or received
2194 * user data. The special value of '0' indicates that no automatic
2195 * close of any associations should be performed. The option expects an
2196 * integer defining the number of seconds of idle time before an
2197 * association is closed.
2199 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2200 unsigned int optlen)
2202 struct sctp_sock *sp = sctp_sk(sk);
2204 /* Applicable to UDP-style socket only */
2205 if (sctp_style(sk, TCP))
2207 if (optlen != sizeof(int))
2209 if (copy_from_user(&sp->autoclose, optval, optlen))
2215 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2217 * Applications can enable or disable heartbeats for any peer address of
2218 * an association, modify an address's heartbeat interval, force a
2219 * heartbeat to be sent immediately, and adjust the address's maximum
2220 * number of retransmissions sent before an address is considered
2221 * unreachable. The following structure is used to access and modify an
2222 * address's parameters:
2224 * struct sctp_paddrparams {
2225 * sctp_assoc_t spp_assoc_id;
2226 * struct sockaddr_storage spp_address;
2227 * uint32_t spp_hbinterval;
2228 * uint16_t spp_pathmaxrxt;
2229 * uint32_t spp_pathmtu;
2230 * uint32_t spp_sackdelay;
2231 * uint32_t spp_flags;
2234 * spp_assoc_id - (one-to-many style socket) This is filled in the
2235 * application, and identifies the association for
2237 * spp_address - This specifies which address is of interest.
2238 * spp_hbinterval - This contains the value of the heartbeat interval,
2239 * in milliseconds. If a value of zero
2240 * is present in this field then no changes are to
2241 * be made to this parameter.
2242 * spp_pathmaxrxt - This contains the maximum number of
2243 * retransmissions before this address shall be
2244 * considered unreachable. If a value of zero
2245 * is present in this field then no changes are to
2246 * be made to this parameter.
2247 * spp_pathmtu - When Path MTU discovery is disabled the value
2248 * specified here will be the "fixed" path mtu.
2249 * Note that if the spp_address field is empty
2250 * then all associations on this address will
2251 * have this fixed path mtu set upon them.
2253 * spp_sackdelay - When delayed sack is enabled, this value specifies
2254 * the number of milliseconds that sacks will be delayed
2255 * for. This value will apply to all addresses of an
2256 * association if the spp_address field is empty. Note
2257 * also, that if delayed sack is enabled and this
2258 * value is set to 0, no change is made to the last
2259 * recorded delayed sack timer value.
2261 * spp_flags - These flags are used to control various features
2262 * on an association. The flag field may contain
2263 * zero or more of the following options.
2265 * SPP_HB_ENABLE - Enable heartbeats on the
2266 * specified address. Note that if the address
2267 * field is empty all addresses for the association
2268 * have heartbeats enabled upon them.
2270 * SPP_HB_DISABLE - Disable heartbeats on the
2271 * speicifed address. Note that if the address
2272 * field is empty all addresses for the association
2273 * will have their heartbeats disabled. Note also
2274 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2275 * mutually exclusive, only one of these two should
2276 * be specified. Enabling both fields will have
2277 * undetermined results.
2279 * SPP_HB_DEMAND - Request a user initiated heartbeat
2280 * to be made immediately.
2282 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2283 * heartbeat delayis to be set to the value of 0
2286 * SPP_PMTUD_ENABLE - This field will enable PMTU
2287 * discovery upon the specified address. Note that
2288 * if the address feild is empty then all addresses
2289 * on the association are effected.
2291 * SPP_PMTUD_DISABLE - This field will disable PMTU
2292 * discovery upon the specified address. Note that
2293 * if the address feild is empty then all addresses
2294 * on the association are effected. Not also that
2295 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2296 * exclusive. Enabling both will have undetermined
2299 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2300 * on delayed sack. The time specified in spp_sackdelay
2301 * is used to specify the sack delay for this address. Note
2302 * that if spp_address is empty then all addresses will
2303 * enable delayed sack and take on the sack delay
2304 * value specified in spp_sackdelay.
2305 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2306 * off delayed sack. If the spp_address field is blank then
2307 * delayed sack is disabled for the entire association. Note
2308 * also that this field is mutually exclusive to
2309 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2312 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2313 struct sctp_transport *trans,
2314 struct sctp_association *asoc,
2315 struct sctp_sock *sp,
2318 int sackdelay_change)
2322 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2323 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2328 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2329 * this field is ignored. Note also that a value of zero indicates
2330 * the current setting should be left unchanged.
2332 if (params->spp_flags & SPP_HB_ENABLE) {
2334 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2335 * set. This lets us use 0 value when this flag
2338 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2339 params->spp_hbinterval = 0;
2341 if (params->spp_hbinterval ||
2342 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2345 msecs_to_jiffies(params->spp_hbinterval);
2348 msecs_to_jiffies(params->spp_hbinterval);
2350 sp->hbinterval = params->spp_hbinterval;
2357 trans->param_flags =
2358 (trans->param_flags & ~SPP_HB) | hb_change;
2361 (asoc->param_flags & ~SPP_HB) | hb_change;
2364 (sp->param_flags & ~SPP_HB) | hb_change;
2368 /* When Path MTU discovery is disabled the value specified here will
2369 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2370 * include the flag SPP_PMTUD_DISABLE for this field to have any
2373 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2375 trans->pathmtu = params->spp_pathmtu;
2376 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2378 asoc->pathmtu = params->spp_pathmtu;
2379 sctp_frag_point(asoc, params->spp_pathmtu);
2381 sp->pathmtu = params->spp_pathmtu;
2387 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2388 (params->spp_flags & SPP_PMTUD_ENABLE);
2389 trans->param_flags =
2390 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2392 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2393 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2397 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2400 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2404 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2405 * value of this field is ignored. Note also that a value of zero
2406 * indicates the current setting should be left unchanged.
2408 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2411 msecs_to_jiffies(params->spp_sackdelay);
2414 msecs_to_jiffies(params->spp_sackdelay);
2416 sp->sackdelay = params->spp_sackdelay;
2420 if (sackdelay_change) {
2422 trans->param_flags =
2423 (trans->param_flags & ~SPP_SACKDELAY) |
2427 (asoc->param_flags & ~SPP_SACKDELAY) |
2431 (sp->param_flags & ~SPP_SACKDELAY) |
2436 /* Note that a value of zero indicates the current setting should be
2439 if (params->spp_pathmaxrxt) {
2441 trans->pathmaxrxt = params->spp_pathmaxrxt;
2443 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2445 sp->pathmaxrxt = params->spp_pathmaxrxt;
2452 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2453 char __user *optval,
2454 unsigned int optlen)
2456 struct sctp_paddrparams params;
2457 struct sctp_transport *trans = NULL;
2458 struct sctp_association *asoc = NULL;
2459 struct sctp_sock *sp = sctp_sk(sk);
2461 int hb_change, pmtud_change, sackdelay_change;
2463 if (optlen != sizeof(struct sctp_paddrparams))
2466 if (copy_from_user(¶ms, optval, optlen))
2469 /* Validate flags and value parameters. */
2470 hb_change = params.spp_flags & SPP_HB;
2471 pmtud_change = params.spp_flags & SPP_PMTUD;
2472 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2474 if (hb_change == SPP_HB ||
2475 pmtud_change == SPP_PMTUD ||
2476 sackdelay_change == SPP_SACKDELAY ||
2477 params.spp_sackdelay > 500 ||
2478 (params.spp_pathmtu &&
2479 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2482 /* If an address other than INADDR_ANY is specified, and
2483 * no transport is found, then the request is invalid.
2485 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
2486 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2487 params.spp_assoc_id);
2492 /* Get association, if assoc_id != 0 and the socket is a one
2493 * to many style socket, and an association was not found, then
2494 * the id was invalid.
2496 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2497 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2500 /* Heartbeat demand can only be sent on a transport or
2501 * association, but not a socket.
2503 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2506 /* Process parameters. */
2507 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2508 hb_change, pmtud_change,
2514 /* If changes are for association, also apply parameters to each
2517 if (!trans && asoc) {
2518 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2520 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2521 hb_change, pmtud_change,
2530 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2532 * This option will effect the way delayed acks are performed. This
2533 * option allows you to get or set the delayed ack time, in
2534 * milliseconds. It also allows changing the delayed ack frequency.
2535 * Changing the frequency to 1 disables the delayed sack algorithm. If
2536 * the assoc_id is 0, then this sets or gets the endpoints default
2537 * values. If the assoc_id field is non-zero, then the set or get
2538 * effects the specified association for the one to many model (the
2539 * assoc_id field is ignored by the one to one model). Note that if
2540 * sack_delay or sack_freq are 0 when setting this option, then the
2541 * current values will remain unchanged.
2543 * struct sctp_sack_info {
2544 * sctp_assoc_t sack_assoc_id;
2545 * uint32_t sack_delay;
2546 * uint32_t sack_freq;
2549 * sack_assoc_id - This parameter, indicates which association the user
2550 * is performing an action upon. Note that if this field's value is
2551 * zero then the endpoints default value is changed (effecting future
2552 * associations only).
2554 * sack_delay - This parameter contains the number of milliseconds that
2555 * the user is requesting the delayed ACK timer be set to. Note that
2556 * this value is defined in the standard to be between 200 and 500
2559 * sack_freq - This parameter contains the number of packets that must
2560 * be received before a sack is sent without waiting for the delay
2561 * timer to expire. The default value for this is 2, setting this
2562 * value to 1 will disable the delayed sack algorithm.
2565 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2566 char __user *optval, unsigned int optlen)
2568 struct sctp_sack_info params;
2569 struct sctp_transport *trans = NULL;
2570 struct sctp_association *asoc = NULL;
2571 struct sctp_sock *sp = sctp_sk(sk);
2573 if (optlen == sizeof(struct sctp_sack_info)) {
2574 if (copy_from_user(¶ms, optval, optlen))
2577 if (params.sack_delay == 0 && params.sack_freq == 0)
2579 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2580 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2581 pr_warn("Use struct sctp_sack_info instead\n");
2582 if (copy_from_user(¶ms, optval, optlen))
2585 if (params.sack_delay == 0)
2586 params.sack_freq = 1;
2588 params.sack_freq = 0;
2592 /* Validate value parameter. */
2593 if (params.sack_delay > 500)
2596 /* Get association, if sack_assoc_id != 0 and the socket is a one
2597 * to many style socket, and an association was not found, then
2598 * the id was invalid.
2600 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2601 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2604 if (params.sack_delay) {
2607 msecs_to_jiffies(params.sack_delay);
2609 (asoc->param_flags & ~SPP_SACKDELAY) |
2610 SPP_SACKDELAY_ENABLE;
2612 sp->sackdelay = params.sack_delay;
2614 (sp->param_flags & ~SPP_SACKDELAY) |
2615 SPP_SACKDELAY_ENABLE;
2619 if (params.sack_freq == 1) {
2622 (asoc->param_flags & ~SPP_SACKDELAY) |
2623 SPP_SACKDELAY_DISABLE;
2626 (sp->param_flags & ~SPP_SACKDELAY) |
2627 SPP_SACKDELAY_DISABLE;
2629 } else if (params.sack_freq > 1) {
2631 asoc->sackfreq = params.sack_freq;
2633 (asoc->param_flags & ~SPP_SACKDELAY) |
2634 SPP_SACKDELAY_ENABLE;
2636 sp->sackfreq = params.sack_freq;
2638 (sp->param_flags & ~SPP_SACKDELAY) |
2639 SPP_SACKDELAY_ENABLE;
2643 /* If change is for association, also apply to each transport. */
2645 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2647 if (params.sack_delay) {
2649 msecs_to_jiffies(params.sack_delay);
2650 trans->param_flags =
2651 (trans->param_flags & ~SPP_SACKDELAY) |
2652 SPP_SACKDELAY_ENABLE;
2654 if (params.sack_freq == 1) {
2655 trans->param_flags =
2656 (trans->param_flags & ~SPP_SACKDELAY) |
2657 SPP_SACKDELAY_DISABLE;
2658 } else if (params.sack_freq > 1) {
2659 trans->sackfreq = params.sack_freq;
2660 trans->param_flags =
2661 (trans->param_flags & ~SPP_SACKDELAY) |
2662 SPP_SACKDELAY_ENABLE;
2670 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2672 * Applications can specify protocol parameters for the default association
2673 * initialization. The option name argument to setsockopt() and getsockopt()
2676 * Setting initialization parameters is effective only on an unconnected
2677 * socket (for UDP-style sockets only future associations are effected
2678 * by the change). With TCP-style sockets, this option is inherited by
2679 * sockets derived from a listener socket.
2681 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2683 struct sctp_initmsg sinit;
2684 struct sctp_sock *sp = sctp_sk(sk);
2686 if (optlen != sizeof(struct sctp_initmsg))
2688 if (copy_from_user(&sinit, optval, optlen))
2691 if (sinit.sinit_num_ostreams)
2692 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2693 if (sinit.sinit_max_instreams)
2694 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2695 if (sinit.sinit_max_attempts)
2696 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2697 if (sinit.sinit_max_init_timeo)
2698 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2704 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2706 * Applications that wish to use the sendto() system call may wish to
2707 * specify a default set of parameters that would normally be supplied
2708 * through the inclusion of ancillary data. This socket option allows
2709 * such an application to set the default sctp_sndrcvinfo structure.
2710 * The application that wishes to use this socket option simply passes
2711 * in to this call the sctp_sndrcvinfo structure defined in Section
2712 * 5.2.2) The input parameters accepted by this call include
2713 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2714 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2715 * to this call if the caller is using the UDP model.
2717 static int sctp_setsockopt_default_send_param(struct sock *sk,
2718 char __user *optval,
2719 unsigned int optlen)
2721 struct sctp_sndrcvinfo info;
2722 struct sctp_association *asoc;
2723 struct sctp_sock *sp = sctp_sk(sk);
2725 if (optlen != sizeof(struct sctp_sndrcvinfo))
2727 if (copy_from_user(&info, optval, optlen))
2730 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2731 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2735 asoc->default_stream = info.sinfo_stream;
2736 asoc->default_flags = info.sinfo_flags;
2737 asoc->default_ppid = info.sinfo_ppid;
2738 asoc->default_context = info.sinfo_context;
2739 asoc->default_timetolive = info.sinfo_timetolive;
2741 sp->default_stream = info.sinfo_stream;
2742 sp->default_flags = info.sinfo_flags;
2743 sp->default_ppid = info.sinfo_ppid;
2744 sp->default_context = info.sinfo_context;
2745 sp->default_timetolive = info.sinfo_timetolive;
2751 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2753 * Requests that the local SCTP stack use the enclosed peer address as
2754 * the association primary. The enclosed address must be one of the
2755 * association peer's addresses.
2757 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2758 unsigned int optlen)
2760 struct sctp_prim prim;
2761 struct sctp_transport *trans;
2763 if (optlen != sizeof(struct sctp_prim))
2766 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2769 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2773 sctp_assoc_set_primary(trans->asoc, trans);
2779 * 7.1.5 SCTP_NODELAY
2781 * Turn on/off any Nagle-like algorithm. This means that packets are
2782 * generally sent as soon as possible and no unnecessary delays are
2783 * introduced, at the cost of more packets in the network. Expects an
2784 * integer boolean flag.
2786 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2787 unsigned int optlen)
2791 if (optlen < sizeof(int))
2793 if (get_user(val, (int __user *)optval))
2796 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2802 * 7.1.1 SCTP_RTOINFO
2804 * The protocol parameters used to initialize and bound retransmission
2805 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2806 * and modify these parameters.
2807 * All parameters are time values, in milliseconds. A value of 0, when
2808 * modifying the parameters, indicates that the current value should not
2812 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2814 struct sctp_rtoinfo rtoinfo;
2815 struct sctp_association *asoc;
2817 if (optlen != sizeof (struct sctp_rtoinfo))
2820 if (copy_from_user(&rtoinfo, optval, optlen))
2823 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2825 /* Set the values to the specific association */
2826 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2830 if (rtoinfo.srto_initial != 0)
2832 msecs_to_jiffies(rtoinfo.srto_initial);
2833 if (rtoinfo.srto_max != 0)
2834 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2835 if (rtoinfo.srto_min != 0)
2836 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2838 /* If there is no association or the association-id = 0
2839 * set the values to the endpoint.
2841 struct sctp_sock *sp = sctp_sk(sk);
2843 if (rtoinfo.srto_initial != 0)
2844 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2845 if (rtoinfo.srto_max != 0)
2846 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2847 if (rtoinfo.srto_min != 0)
2848 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2856 * 7.1.2 SCTP_ASSOCINFO
2858 * This option is used to tune the maximum retransmission attempts
2859 * of the association.
2860 * Returns an error if the new association retransmission value is
2861 * greater than the sum of the retransmission value of the peer.
2862 * See [SCTP] for more information.
2865 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2868 struct sctp_assocparams assocparams;
2869 struct sctp_association *asoc;
2871 if (optlen != sizeof(struct sctp_assocparams))
2873 if (copy_from_user(&assocparams, optval, optlen))
2876 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2878 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2881 /* Set the values to the specific association */
2883 if (assocparams.sasoc_asocmaxrxt != 0) {
2886 struct sctp_transport *peer_addr;
2888 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2890 path_sum += peer_addr->pathmaxrxt;
2894 /* Only validate asocmaxrxt if we have more than
2895 * one path/transport. We do this because path
2896 * retransmissions are only counted when we have more
2900 assocparams.sasoc_asocmaxrxt > path_sum)
2903 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2906 if (assocparams.sasoc_cookie_life != 0) {
2907 asoc->cookie_life.tv_sec =
2908 assocparams.sasoc_cookie_life / 1000;
2909 asoc->cookie_life.tv_usec =
2910 (assocparams.sasoc_cookie_life % 1000)
2914 /* Set the values to the endpoint */
2915 struct sctp_sock *sp = sctp_sk(sk);
2917 if (assocparams.sasoc_asocmaxrxt != 0)
2918 sp->assocparams.sasoc_asocmaxrxt =
2919 assocparams.sasoc_asocmaxrxt;
2920 if (assocparams.sasoc_cookie_life != 0)
2921 sp->assocparams.sasoc_cookie_life =
2922 assocparams.sasoc_cookie_life;
2928 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2930 * This socket option is a boolean flag which turns on or off mapped V4
2931 * addresses. If this option is turned on and the socket is type
2932 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2933 * If this option is turned off, then no mapping will be done of V4
2934 * addresses and a user will receive both PF_INET6 and PF_INET type
2935 * addresses on the socket.
2937 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2940 struct sctp_sock *sp = sctp_sk(sk);
2942 if (optlen < sizeof(int))
2944 if (get_user(val, (int __user *)optval))
2955 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2956 * This option will get or set the maximum size to put in any outgoing
2957 * SCTP DATA chunk. If a message is larger than this size it will be
2958 * fragmented by SCTP into the specified size. Note that the underlying
2959 * SCTP implementation may fragment into smaller sized chunks when the
2960 * PMTU of the underlying association is smaller than the value set by
2961 * the user. The default value for this option is '0' which indicates
2962 * the user is NOT limiting fragmentation and only the PMTU will effect
2963 * SCTP's choice of DATA chunk size. Note also that values set larger
2964 * than the maximum size of an IP datagram will effectively let SCTP
2965 * control fragmentation (i.e. the same as setting this option to 0).
2967 * The following structure is used to access and modify this parameter:
2969 * struct sctp_assoc_value {
2970 * sctp_assoc_t assoc_id;
2971 * uint32_t assoc_value;
2974 * assoc_id: This parameter is ignored for one-to-one style sockets.
2975 * For one-to-many style sockets this parameter indicates which
2976 * association the user is performing an action upon. Note that if
2977 * this field's value is zero then the endpoints default value is
2978 * changed (effecting future associations only).
2979 * assoc_value: This parameter specifies the maximum size in bytes.
2981 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2983 struct sctp_assoc_value params;
2984 struct sctp_association *asoc;
2985 struct sctp_sock *sp = sctp_sk(sk);
2988 if (optlen == sizeof(int)) {
2989 pr_warn("Use of int in maxseg socket option deprecated\n");
2990 pr_warn("Use struct sctp_assoc_value instead\n");
2991 if (copy_from_user(&val, optval, optlen))
2993 params.assoc_id = 0;
2994 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2995 if (copy_from_user(¶ms, optval, optlen))
2997 val = params.assoc_value;
3001 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3004 asoc = sctp_id2assoc(sk, params.assoc_id);
3005 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3010 val = asoc->pathmtu;
3011 val -= sp->pf->af->net_header_len;
3012 val -= sizeof(struct sctphdr) +
3013 sizeof(struct sctp_data_chunk);
3015 asoc->user_frag = val;
3016 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3018 sp->user_frag = val;
3026 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3028 * Requests that the peer mark the enclosed address as the association
3029 * primary. The enclosed address must be one of the association's
3030 * locally bound addresses. The following structure is used to make a
3031 * set primary request:
3033 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3034 unsigned int optlen)
3036 struct sctp_sock *sp;
3037 struct sctp_association *asoc = NULL;
3038 struct sctp_setpeerprim prim;
3039 struct sctp_chunk *chunk;
3045 if (!sctp_addip_enable)
3048 if (optlen != sizeof(struct sctp_setpeerprim))
3051 if (copy_from_user(&prim, optval, optlen))
3054 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3058 if (!asoc->peer.asconf_capable)
3061 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3064 if (!sctp_state(asoc, ESTABLISHED))
3067 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3071 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3072 return -EADDRNOTAVAIL;
3074 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3075 return -EADDRNOTAVAIL;
3077 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3078 chunk = sctp_make_asconf_set_prim(asoc,
3079 (union sctp_addr *)&prim.sspp_addr);
3083 err = sctp_send_asconf(asoc, chunk);
3085 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
3090 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3091 unsigned int optlen)
3093 struct sctp_setadaptation adaptation;
3095 if (optlen != sizeof(struct sctp_setadaptation))
3097 if (copy_from_user(&adaptation, optval, optlen))
3100 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3106 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3108 * The context field in the sctp_sndrcvinfo structure is normally only
3109 * used when a failed message is retrieved holding the value that was
3110 * sent down on the actual send call. This option allows the setting of
3111 * a default context on an association basis that will be received on
3112 * reading messages from the peer. This is especially helpful in the
3113 * one-2-many model for an application to keep some reference to an
3114 * internal state machine that is processing messages on the
3115 * association. Note that the setting of this value only effects
3116 * received messages from the peer and does not effect the value that is
3117 * saved with outbound messages.
3119 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3120 unsigned int optlen)
3122 struct sctp_assoc_value params;
3123 struct sctp_sock *sp;
3124 struct sctp_association *asoc;
3126 if (optlen != sizeof(struct sctp_assoc_value))
3128 if (copy_from_user(¶ms, optval, optlen))
3133 if (params.assoc_id != 0) {
3134 asoc = sctp_id2assoc(sk, params.assoc_id);
3137 asoc->default_rcv_context = params.assoc_value;
3139 sp->default_rcv_context = params.assoc_value;
3146 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3148 * This options will at a minimum specify if the implementation is doing
3149 * fragmented interleave. Fragmented interleave, for a one to many
3150 * socket, is when subsequent calls to receive a message may return
3151 * parts of messages from different associations. Some implementations
3152 * may allow you to turn this value on or off. If so, when turned off,
3153 * no fragment interleave will occur (which will cause a head of line
3154 * blocking amongst multiple associations sharing the same one to many
3155 * socket). When this option is turned on, then each receive call may
3156 * come from a different association (thus the user must receive data
3157 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3158 * association each receive belongs to.
3160 * This option takes a boolean value. A non-zero value indicates that
3161 * fragmented interleave is on. A value of zero indicates that
3162 * fragmented interleave is off.
3164 * Note that it is important that an implementation that allows this
3165 * option to be turned on, have it off by default. Otherwise an unaware
3166 * application using the one to many model may become confused and act
3169 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3170 char __user *optval,
3171 unsigned int optlen)
3175 if (optlen != sizeof(int))
3177 if (get_user(val, (int __user *)optval))
3180 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3186 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3187 * (SCTP_PARTIAL_DELIVERY_POINT)
3189 * This option will set or get the SCTP partial delivery point. This
3190 * point is the size of a message where the partial delivery API will be
3191 * invoked to help free up rwnd space for the peer. Setting this to a
3192 * lower value will cause partial deliveries to happen more often. The
3193 * calls argument is an integer that sets or gets the partial delivery
3194 * point. Note also that the call will fail if the user attempts to set
3195 * this value larger than the socket receive buffer size.
3197 * Note that any single message having a length smaller than or equal to
3198 * the SCTP partial delivery point will be delivered in one single read
3199 * call as long as the user provided buffer is large enough to hold the
3202 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3203 char __user *optval,
3204 unsigned int optlen)
3208 if (optlen != sizeof(u32))
3210 if (get_user(val, (int __user *)optval))
3213 /* Note: We double the receive buffer from what the user sets
3214 * it to be, also initial rwnd is based on rcvbuf/2.
3216 if (val > (sk->sk_rcvbuf >> 1))
3219 sctp_sk(sk)->pd_point = val;
3221 return 0; /* is this the right error code? */
3225 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3227 * This option will allow a user to change the maximum burst of packets
3228 * that can be emitted by this association. Note that the default value
3229 * is 4, and some implementations may restrict this setting so that it
3230 * can only be lowered.
3232 * NOTE: This text doesn't seem right. Do this on a socket basis with
3233 * future associations inheriting the socket value.
3235 static int sctp_setsockopt_maxburst(struct sock *sk,
3236 char __user *optval,
3237 unsigned int optlen)
3239 struct sctp_assoc_value params;
3240 struct sctp_sock *sp;
3241 struct sctp_association *asoc;
3245 if (optlen == sizeof(int)) {
3246 pr_warn("Use of int in max_burst socket option deprecated\n");
3247 pr_warn("Use struct sctp_assoc_value instead\n");
3248 if (copy_from_user(&val, optval, optlen))
3250 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3251 if (copy_from_user(¶ms, optval, optlen))
3253 val = params.assoc_value;
3254 assoc_id = params.assoc_id;
3260 if (assoc_id != 0) {
3261 asoc = sctp_id2assoc(sk, assoc_id);
3264 asoc->max_burst = val;
3266 sp->max_burst = val;
3272 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3274 * This set option adds a chunk type that the user is requesting to be
3275 * received only in an authenticated way. Changes to the list of chunks
3276 * will only effect future associations on the socket.
3278 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3279 char __user *optval,
3280 unsigned int optlen)
3282 struct sctp_authchunk val;
3284 if (!sctp_auth_enable)
3287 if (optlen != sizeof(struct sctp_authchunk))
3289 if (copy_from_user(&val, optval, optlen))
3292 switch (val.sauth_chunk) {
3294 case SCTP_CID_INIT_ACK:
3295 case SCTP_CID_SHUTDOWN_COMPLETE:
3300 /* add this chunk id to the endpoint */
3301 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3305 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3307 * This option gets or sets the list of HMAC algorithms that the local
3308 * endpoint requires the peer to use.
3310 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3311 char __user *optval,
3312 unsigned int optlen)
3314 struct sctp_hmacalgo *hmacs;
3318 if (!sctp_auth_enable)
3321 if (optlen < sizeof(struct sctp_hmacalgo))
3324 hmacs= memdup_user(optval, optlen);
3326 return PTR_ERR(hmacs);
3328 idents = hmacs->shmac_num_idents;
3329 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3330 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3335 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3342 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3344 * This option will set a shared secret key which is used to build an
3345 * association shared key.
3347 static int sctp_setsockopt_auth_key(struct sock *sk,
3348 char __user *optval,
3349 unsigned int optlen)
3351 struct sctp_authkey *authkey;
3352 struct sctp_association *asoc;
3355 if (!sctp_auth_enable)
3358 if (optlen <= sizeof(struct sctp_authkey))
3361 authkey= memdup_user(optval, optlen);
3362 if (IS_ERR(authkey))
3363 return PTR_ERR(authkey);
3365 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3370 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3371 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3376 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3383 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3385 * This option will get or set the active shared key to be used to build
3386 * the association shared key.
3388 static int sctp_setsockopt_active_key(struct sock *sk,
3389 char __user *optval,
3390 unsigned int optlen)
3392 struct sctp_authkeyid val;
3393 struct sctp_association *asoc;
3395 if (!sctp_auth_enable)
3398 if (optlen != sizeof(struct sctp_authkeyid))
3400 if (copy_from_user(&val, optval, optlen))
3403 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3404 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3407 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3408 val.scact_keynumber);
3412 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3414 * This set option will delete a shared secret key from use.
3416 static int sctp_setsockopt_del_key(struct sock *sk,
3417 char __user *optval,
3418 unsigned int optlen)
3420 struct sctp_authkeyid val;
3421 struct sctp_association *asoc;
3423 if (!sctp_auth_enable)
3426 if (optlen != sizeof(struct sctp_authkeyid))
3428 if (copy_from_user(&val, optval, optlen))
3431 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3432 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3435 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3436 val.scact_keynumber);
3441 * 8.1.23 SCTP_AUTO_ASCONF
3443 * This option will enable or disable the use of the automatic generation of
3444 * ASCONF chunks to add and delete addresses to an existing association. Note
3445 * that this option has two caveats namely: a) it only affects sockets that
3446 * are bound to all addresses available to the SCTP stack, and b) the system
3447 * administrator may have an overriding control that turns the ASCONF feature
3448 * off no matter what setting the socket option may have.
3449 * This option expects an integer boolean flag, where a non-zero value turns on
3450 * the option, and a zero value turns off the option.
3451 * Note. In this implementation, socket operation overrides default parameter
3452 * being set by sysctl as well as FreeBSD implementation
3454 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3455 unsigned int optlen)
3458 struct sctp_sock *sp = sctp_sk(sk);
3460 if (optlen < sizeof(int))
3462 if (get_user(val, (int __user *)optval))
3464 if (!sctp_is_ep_boundall(sk) && val)
3466 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3469 if (val == 0 && sp->do_auto_asconf) {
3470 list_del(&sp->auto_asconf_list);
3471 sp->do_auto_asconf = 0;
3472 } else if (val && !sp->do_auto_asconf) {
3473 list_add_tail(&sp->auto_asconf_list,
3474 &sock_net(sk)->sctp.auto_asconf_splist);
3475 sp->do_auto_asconf = 1;
3482 * SCTP_PEER_ADDR_THLDS
3484 * This option allows us to alter the partially failed threshold for one or all
3485 * transports in an association. See Section 6.1 of:
3486 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3488 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3489 char __user *optval,
3490 unsigned int optlen)
3492 struct sctp_paddrthlds val;
3493 struct sctp_transport *trans;
3494 struct sctp_association *asoc;
3496 if (optlen < sizeof(struct sctp_paddrthlds))
3498 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3499 sizeof(struct sctp_paddrthlds)))
3503 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3504 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3507 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3509 if (val.spt_pathmaxrxt)
3510 trans->pathmaxrxt = val.spt_pathmaxrxt;
3511 trans->pf_retrans = val.spt_pathpfthld;
3514 if (val.spt_pathmaxrxt)
3515 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3516 asoc->pf_retrans = val.spt_pathpfthld;
3518 trans = sctp_addr_id2transport(sk, &val.spt_address,
3523 if (val.spt_pathmaxrxt)
3524 trans->pathmaxrxt = val.spt_pathmaxrxt;
3525 trans->pf_retrans = val.spt_pathpfthld;
3531 /* API 6.2 setsockopt(), getsockopt()
3533 * Applications use setsockopt() and getsockopt() to set or retrieve
3534 * socket options. Socket options are used to change the default
3535 * behavior of sockets calls. They are described in Section 7.
3539 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3540 * int __user *optlen);
3541 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3544 * sd - the socket descript.
3545 * level - set to IPPROTO_SCTP for all SCTP options.
3546 * optname - the option name.
3547 * optval - the buffer to store the value of the option.
3548 * optlen - the size of the buffer.
3550 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3551 char __user *optval, unsigned int optlen)
3555 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3558 /* I can hardly begin to describe how wrong this is. This is
3559 * so broken as to be worse than useless. The API draft
3560 * REALLY is NOT helpful here... I am not convinced that the
3561 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3562 * are at all well-founded.
3564 if (level != SOL_SCTP) {
3565 struct sctp_af *af = sctp_sk(sk)->pf->af;
3566 retval = af->setsockopt(sk, level, optname, optval, optlen);
3573 case SCTP_SOCKOPT_BINDX_ADD:
3574 /* 'optlen' is the size of the addresses buffer. */
3575 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3576 optlen, SCTP_BINDX_ADD_ADDR);
3579 case SCTP_SOCKOPT_BINDX_REM:
3580 /* 'optlen' is the size of the addresses buffer. */
3581 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3582 optlen, SCTP_BINDX_REM_ADDR);
3585 case SCTP_SOCKOPT_CONNECTX_OLD:
3586 /* 'optlen' is the size of the addresses buffer. */
3587 retval = sctp_setsockopt_connectx_old(sk,
3588 (struct sockaddr __user *)optval,
3592 case SCTP_SOCKOPT_CONNECTX:
3593 /* 'optlen' is the size of the addresses buffer. */
3594 retval = sctp_setsockopt_connectx(sk,
3595 (struct sockaddr __user *)optval,
3599 case SCTP_DISABLE_FRAGMENTS:
3600 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3604 retval = sctp_setsockopt_events(sk, optval, optlen);
3607 case SCTP_AUTOCLOSE:
3608 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3611 case SCTP_PEER_ADDR_PARAMS:
3612 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3615 case SCTP_DELAYED_SACK:
3616 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3618 case SCTP_PARTIAL_DELIVERY_POINT:
3619 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3623 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3625 case SCTP_DEFAULT_SEND_PARAM:
3626 retval = sctp_setsockopt_default_send_param(sk, optval,
3629 case SCTP_PRIMARY_ADDR:
3630 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3632 case SCTP_SET_PEER_PRIMARY_ADDR:
3633 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3636 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3639 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3641 case SCTP_ASSOCINFO:
3642 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3644 case SCTP_I_WANT_MAPPED_V4_ADDR:
3645 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3648 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3650 case SCTP_ADAPTATION_LAYER:
3651 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3654 retval = sctp_setsockopt_context(sk, optval, optlen);
3656 case SCTP_FRAGMENT_INTERLEAVE:
3657 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3659 case SCTP_MAX_BURST:
3660 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3662 case SCTP_AUTH_CHUNK:
3663 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3665 case SCTP_HMAC_IDENT:
3666 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3669 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3671 case SCTP_AUTH_ACTIVE_KEY:
3672 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3674 case SCTP_AUTH_DELETE_KEY:
3675 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3677 case SCTP_AUTO_ASCONF:
3678 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3680 case SCTP_PEER_ADDR_THLDS:
3681 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3684 retval = -ENOPROTOOPT;
3688 sctp_release_sock(sk);
3694 /* API 3.1.6 connect() - UDP Style Syntax
3696 * An application may use the connect() call in the UDP model to initiate an
3697 * association without sending data.
3701 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3703 * sd: the socket descriptor to have a new association added to.
3705 * nam: the address structure (either struct sockaddr_in or struct
3706 * sockaddr_in6 defined in RFC2553 [7]).
3708 * len: the size of the address.
3710 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3718 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3719 __func__, sk, addr, addr_len);
3721 /* Validate addr_len before calling common connect/connectx routine. */
3722 af = sctp_get_af_specific(addr->sa_family);
3723 if (!af || addr_len < af->sockaddr_len) {
3726 /* Pass correct addr len to common routine (so it knows there
3727 * is only one address being passed.
3729 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3732 sctp_release_sock(sk);
3736 /* FIXME: Write comments. */
3737 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3739 return -EOPNOTSUPP; /* STUB */
3742 /* 4.1.4 accept() - TCP Style Syntax
3744 * Applications use accept() call to remove an established SCTP
3745 * association from the accept queue of the endpoint. A new socket
3746 * descriptor will be returned from accept() to represent the newly
3747 * formed association.
3749 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3751 struct sctp_sock *sp;
3752 struct sctp_endpoint *ep;
3753 struct sock *newsk = NULL;
3754 struct sctp_association *asoc;
3763 if (!sctp_style(sk, TCP)) {
3764 error = -EOPNOTSUPP;
3768 if (!sctp_sstate(sk, LISTENING)) {
3773 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3775 error = sctp_wait_for_accept(sk, timeo);
3779 /* We treat the list of associations on the endpoint as the accept
3780 * queue and pick the first association on the list.
3782 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3784 newsk = sp->pf->create_accept_sk(sk, asoc);
3790 /* Populate the fields of the newsk from the oldsk and migrate the
3791 * asoc to the newsk.
3793 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3796 sctp_release_sock(sk);
3801 /* The SCTP ioctl handler. */
3802 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3809 * SEQPACKET-style sockets in LISTENING state are valid, for
3810 * SCTP, so only discard TCP-style sockets in LISTENING state.
3812 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3817 struct sk_buff *skb;
3818 unsigned int amount = 0;
3820 skb = skb_peek(&sk->sk_receive_queue);
3823 * We will only return the amount of this packet since
3824 * that is all that will be read.
3828 rc = put_user(amount, (int __user *)arg);
3836 sctp_release_sock(sk);
3840 /* This is the function which gets called during socket creation to
3841 * initialized the SCTP-specific portion of the sock.
3842 * The sock structure should already be zero-filled memory.
3844 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3846 struct sctp_endpoint *ep;
3847 struct sctp_sock *sp;
3849 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3853 /* Initialize the SCTP per socket area. */
3854 switch (sk->sk_type) {
3855 case SOCK_SEQPACKET:
3856 sp->type = SCTP_SOCKET_UDP;
3859 sp->type = SCTP_SOCKET_TCP;
3862 return -ESOCKTNOSUPPORT;
3865 /* Initialize default send parameters. These parameters can be
3866 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3868 sp->default_stream = 0;
3869 sp->default_ppid = 0;
3870 sp->default_flags = 0;
3871 sp->default_context = 0;
3872 sp->default_timetolive = 0;
3874 sp->default_rcv_context = 0;
3875 sp->max_burst = sctp_max_burst;
3877 /* Initialize default setup parameters. These parameters
3878 * can be modified with the SCTP_INITMSG socket option or
3879 * overridden by the SCTP_INIT CMSG.
3881 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3882 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3883 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3884 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3886 /* Initialize default RTO related parameters. These parameters can
3887 * be modified for with the SCTP_RTOINFO socket option.
3889 sp->rtoinfo.srto_initial = sctp_rto_initial;
3890 sp->rtoinfo.srto_max = sctp_rto_max;
3891 sp->rtoinfo.srto_min = sctp_rto_min;
3893 /* Initialize default association related parameters. These parameters
3894 * can be modified with the SCTP_ASSOCINFO socket option.
3896 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3897 sp->assocparams.sasoc_number_peer_destinations = 0;
3898 sp->assocparams.sasoc_peer_rwnd = 0;
3899 sp->assocparams.sasoc_local_rwnd = 0;
3900 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3902 /* Initialize default event subscriptions. By default, all the
3905 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3907 /* Default Peer Address Parameters. These defaults can
3908 * be modified via SCTP_PEER_ADDR_PARAMS
3910 sp->hbinterval = sctp_hb_interval;
3911 sp->pathmaxrxt = sctp_max_retrans_path;
3912 sp->pathmtu = 0; // allow default discovery
3913 sp->sackdelay = sctp_sack_timeout;
3915 sp->param_flags = SPP_HB_ENABLE |
3917 SPP_SACKDELAY_ENABLE;
3919 /* If enabled no SCTP message fragmentation will be performed.
3920 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3922 sp->disable_fragments = 0;
3924 /* Enable Nagle algorithm by default. */
3927 /* Enable by default. */
3930 /* Auto-close idle associations after the configured
3931 * number of seconds. A value of 0 disables this
3932 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3933 * for UDP-style sockets only.
3937 /* User specified fragmentation limit. */
3940 sp->adaptation_ind = 0;
3942 sp->pf = sctp_get_pf_specific(sk->sk_family);
3944 /* Control variables for partial data delivery. */
3945 atomic_set(&sp->pd_mode, 0);
3946 skb_queue_head_init(&sp->pd_lobby);
3947 sp->frag_interleave = 0;
3949 /* Create a per socket endpoint structure. Even if we
3950 * change the data structure relationships, this may still
3951 * be useful for storing pre-connect address information.
3953 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3960 SCTP_DBG_OBJCNT_INC(sock);
3963 percpu_counter_inc(&sctp_sockets_allocated);
3964 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3965 if (sctp_default_auto_asconf) {
3966 list_add_tail(&sp->auto_asconf_list,
3967 &sock_net(sk)->sctp.auto_asconf_splist);
3968 sp->do_auto_asconf = 1;
3970 sp->do_auto_asconf = 0;
3976 /* Cleanup any SCTP per socket resources. */
3977 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3979 struct sctp_sock *sp;
3981 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3983 /* Release our hold on the endpoint. */
3985 if (sp->do_auto_asconf) {
3986 sp->do_auto_asconf = 0;
3987 list_del(&sp->auto_asconf_list);
3989 sctp_endpoint_free(sp->ep);
3991 percpu_counter_dec(&sctp_sockets_allocated);
3992 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
3996 /* API 4.1.7 shutdown() - TCP Style Syntax
3997 * int shutdown(int socket, int how);
3999 * sd - the socket descriptor of the association to be closed.
4000 * how - Specifies the type of shutdown. The values are
4003 * Disables further receive operations. No SCTP
4004 * protocol action is taken.
4006 * Disables further send operations, and initiates
4007 * the SCTP shutdown sequence.
4009 * Disables further send and receive operations
4010 * and initiates the SCTP shutdown sequence.
4012 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
4014 struct sctp_endpoint *ep;
4015 struct sctp_association *asoc;
4017 if (!sctp_style(sk, TCP))
4020 if (how & SEND_SHUTDOWN) {
4021 ep = sctp_sk(sk)->ep;
4022 if (!list_empty(&ep->asocs)) {
4023 asoc = list_entry(ep->asocs.next,
4024 struct sctp_association, asocs);
4025 sctp_primitive_SHUTDOWN(asoc, NULL);
4030 /* 7.2.1 Association Status (SCTP_STATUS)
4032 * Applications can retrieve current status information about an
4033 * association, including association state, peer receiver window size,
4034 * number of unacked data chunks, and number of data chunks pending
4035 * receipt. This information is read-only.
4037 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4038 char __user *optval,
4041 struct sctp_status status;
4042 struct sctp_association *asoc = NULL;
4043 struct sctp_transport *transport;
4044 sctp_assoc_t associd;
4047 if (len < sizeof(status)) {
4052 len = sizeof(status);
4053 if (copy_from_user(&status, optval, len)) {
4058 associd = status.sstat_assoc_id;
4059 asoc = sctp_id2assoc(sk, associd);
4065 transport = asoc->peer.primary_path;
4067 status.sstat_assoc_id = sctp_assoc2id(asoc);
4068 status.sstat_state = asoc->state;
4069 status.sstat_rwnd = asoc->peer.rwnd;
4070 status.sstat_unackdata = asoc->unack_data;
4072 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4073 status.sstat_instrms = asoc->c.sinit_max_instreams;
4074 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4075 status.sstat_fragmentation_point = asoc->frag_point;
4076 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4077 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4078 transport->af_specific->sockaddr_len);
4079 /* Map ipv4 address into v4-mapped-on-v6 address. */
4080 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4081 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4082 status.sstat_primary.spinfo_state = transport->state;
4083 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4084 status.sstat_primary.spinfo_srtt = transport->srtt;
4085 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4086 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4088 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4089 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4091 if (put_user(len, optlen)) {
4096 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
4097 len, status.sstat_state, status.sstat_rwnd,
4098 status.sstat_assoc_id);
4100 if (copy_to_user(optval, &status, len)) {
4110 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4112 * Applications can retrieve information about a specific peer address
4113 * of an association, including its reachability state, congestion
4114 * window, and retransmission timer values. This information is
4117 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4118 char __user *optval,
4121 struct sctp_paddrinfo pinfo;
4122 struct sctp_transport *transport;
4125 if (len < sizeof(pinfo)) {
4130 len = sizeof(pinfo);
4131 if (copy_from_user(&pinfo, optval, len)) {
4136 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4137 pinfo.spinfo_assoc_id);
4141 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4142 pinfo.spinfo_state = transport->state;
4143 pinfo.spinfo_cwnd = transport->cwnd;
4144 pinfo.spinfo_srtt = transport->srtt;
4145 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4146 pinfo.spinfo_mtu = transport->pathmtu;
4148 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4149 pinfo.spinfo_state = SCTP_ACTIVE;
4151 if (put_user(len, optlen)) {
4156 if (copy_to_user(optval, &pinfo, len)) {
4165 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4167 * This option is a on/off flag. If enabled no SCTP message
4168 * fragmentation will be performed. Instead if a message being sent
4169 * exceeds the current PMTU size, the message will NOT be sent and
4170 * instead a error will be indicated to the user.
4172 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4173 char __user *optval, int __user *optlen)
4177 if (len < sizeof(int))
4181 val = (sctp_sk(sk)->disable_fragments == 1);
4182 if (put_user(len, optlen))
4184 if (copy_to_user(optval, &val, len))
4189 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4191 * This socket option is used to specify various notifications and
4192 * ancillary data the user wishes to receive.
4194 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4199 if (len > sizeof(struct sctp_event_subscribe))
4200 len = sizeof(struct sctp_event_subscribe);
4201 if (put_user(len, optlen))
4203 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4208 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4210 * This socket option is applicable to the UDP-style socket only. When
4211 * set it will cause associations that are idle for more than the
4212 * specified number of seconds to automatically close. An association
4213 * being idle is defined an association that has NOT sent or received
4214 * user data. The special value of '0' indicates that no automatic
4215 * close of any associations should be performed. The option expects an
4216 * integer defining the number of seconds of idle time before an
4217 * association is closed.
4219 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4221 /* Applicable to UDP-style socket only */
4222 if (sctp_style(sk, TCP))
4224 if (len < sizeof(int))
4227 if (put_user(len, optlen))
4229 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4234 /* Helper routine to branch off an association to a new socket. */
4235 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4237 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4238 struct socket *sock;
4245 /* An association cannot be branched off from an already peeled-off
4246 * socket, nor is this supported for tcp style sockets.
4248 if (!sctp_style(sk, UDP))
4251 /* Create a new socket. */
4252 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4256 sctp_copy_sock(sock->sk, sk, asoc);
4258 /* Make peeled-off sockets more like 1-1 accepted sockets.
4259 * Set the daddr and initialize id to something more random
4261 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4262 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4264 /* Populate the fields of the newsk from the oldsk and migrate the
4265 * asoc to the newsk.
4267 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4273 EXPORT_SYMBOL(sctp_do_peeloff);
4275 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4277 sctp_peeloff_arg_t peeloff;
4278 struct socket *newsock;
4281 if (len < sizeof(sctp_peeloff_arg_t))
4283 len = sizeof(sctp_peeloff_arg_t);
4284 if (copy_from_user(&peeloff, optval, len))
4287 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4291 /* Map the socket to an unused fd that can be returned to the user. */
4292 retval = sock_map_fd(newsock, 0);
4294 sock_release(newsock);
4298 SCTP_DEBUG_PRINTK("%s: sk: %p newsk: %p sd: %d\n",
4299 __func__, sk, newsock->sk, retval);
4301 /* Return the fd mapped to the new socket. */
4302 peeloff.sd = retval;
4303 if (put_user(len, optlen))
4305 if (copy_to_user(optval, &peeloff, len))
4312 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4314 * Applications can enable or disable heartbeats for any peer address of
4315 * an association, modify an address's heartbeat interval, force a
4316 * heartbeat to be sent immediately, and adjust the address's maximum
4317 * number of retransmissions sent before an address is considered
4318 * unreachable. The following structure is used to access and modify an
4319 * address's parameters:
4321 * struct sctp_paddrparams {
4322 * sctp_assoc_t spp_assoc_id;
4323 * struct sockaddr_storage spp_address;
4324 * uint32_t spp_hbinterval;
4325 * uint16_t spp_pathmaxrxt;
4326 * uint32_t spp_pathmtu;
4327 * uint32_t spp_sackdelay;
4328 * uint32_t spp_flags;
4331 * spp_assoc_id - (one-to-many style socket) This is filled in the
4332 * application, and identifies the association for
4334 * spp_address - This specifies which address is of interest.
4335 * spp_hbinterval - This contains the value of the heartbeat interval,
4336 * in milliseconds. If a value of zero
4337 * is present in this field then no changes are to
4338 * be made to this parameter.
4339 * spp_pathmaxrxt - This contains the maximum number of
4340 * retransmissions before this address shall be
4341 * considered unreachable. If a value of zero
4342 * is present in this field then no changes are to
4343 * be made to this parameter.
4344 * spp_pathmtu - When Path MTU discovery is disabled the value
4345 * specified here will be the "fixed" path mtu.
4346 * Note that if the spp_address field is empty
4347 * then all associations on this address will
4348 * have this fixed path mtu set upon them.
4350 * spp_sackdelay - When delayed sack is enabled, this value specifies
4351 * the number of milliseconds that sacks will be delayed
4352 * for. This value will apply to all addresses of an
4353 * association if the spp_address field is empty. Note
4354 * also, that if delayed sack is enabled and this
4355 * value is set to 0, no change is made to the last
4356 * recorded delayed sack timer value.
4358 * spp_flags - These flags are used to control various features
4359 * on an association. The flag field may contain
4360 * zero or more of the following options.
4362 * SPP_HB_ENABLE - Enable heartbeats on the
4363 * specified address. Note that if the address
4364 * field is empty all addresses for the association
4365 * have heartbeats enabled upon them.
4367 * SPP_HB_DISABLE - Disable heartbeats on the
4368 * speicifed address. Note that if the address
4369 * field is empty all addresses for the association
4370 * will have their heartbeats disabled. Note also
4371 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4372 * mutually exclusive, only one of these two should
4373 * be specified. Enabling both fields will have
4374 * undetermined results.
4376 * SPP_HB_DEMAND - Request a user initiated heartbeat
4377 * to be made immediately.
4379 * SPP_PMTUD_ENABLE - This field will enable PMTU
4380 * discovery upon the specified address. Note that
4381 * if the address feild is empty then all addresses
4382 * on the association are effected.
4384 * SPP_PMTUD_DISABLE - This field will disable PMTU
4385 * discovery upon the specified address. Note that
4386 * if the address feild is empty then all addresses
4387 * on the association are effected. Not also that
4388 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4389 * exclusive. Enabling both will have undetermined
4392 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4393 * on delayed sack. The time specified in spp_sackdelay
4394 * is used to specify the sack delay for this address. Note
4395 * that if spp_address is empty then all addresses will
4396 * enable delayed sack and take on the sack delay
4397 * value specified in spp_sackdelay.
4398 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4399 * off delayed sack. If the spp_address field is blank then
4400 * delayed sack is disabled for the entire association. Note
4401 * also that this field is mutually exclusive to
4402 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4405 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4406 char __user *optval, int __user *optlen)
4408 struct sctp_paddrparams params;
4409 struct sctp_transport *trans = NULL;
4410 struct sctp_association *asoc = NULL;
4411 struct sctp_sock *sp = sctp_sk(sk);
4413 if (len < sizeof(struct sctp_paddrparams))
4415 len = sizeof(struct sctp_paddrparams);
4416 if (copy_from_user(¶ms, optval, len))
4419 /* If an address other than INADDR_ANY is specified, and
4420 * no transport is found, then the request is invalid.
4422 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
4423 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4424 params.spp_assoc_id);
4426 SCTP_DEBUG_PRINTK("Failed no transport\n");
4431 /* Get association, if assoc_id != 0 and the socket is a one
4432 * to many style socket, and an association was not found, then
4433 * the id was invalid.
4435 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4436 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4437 SCTP_DEBUG_PRINTK("Failed no association\n");
4442 /* Fetch transport values. */
4443 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4444 params.spp_pathmtu = trans->pathmtu;
4445 params.spp_pathmaxrxt = trans->pathmaxrxt;
4446 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4448 /*draft-11 doesn't say what to return in spp_flags*/
4449 params.spp_flags = trans->param_flags;
4451 /* Fetch association values. */
4452 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4453 params.spp_pathmtu = asoc->pathmtu;
4454 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4455 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4457 /*draft-11 doesn't say what to return in spp_flags*/
4458 params.spp_flags = asoc->param_flags;
4460 /* Fetch socket values. */
4461 params.spp_hbinterval = sp->hbinterval;
4462 params.spp_pathmtu = sp->pathmtu;
4463 params.spp_sackdelay = sp->sackdelay;
4464 params.spp_pathmaxrxt = sp->pathmaxrxt;
4466 /*draft-11 doesn't say what to return in spp_flags*/
4467 params.spp_flags = sp->param_flags;
4470 if (copy_to_user(optval, ¶ms, len))
4473 if (put_user(len, optlen))
4480 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4482 * This option will effect the way delayed acks are performed. This
4483 * option allows you to get or set the delayed ack time, in
4484 * milliseconds. It also allows changing the delayed ack frequency.
4485 * Changing the frequency to 1 disables the delayed sack algorithm. If
4486 * the assoc_id is 0, then this sets or gets the endpoints default
4487 * values. If the assoc_id field is non-zero, then the set or get
4488 * effects the specified association for the one to many model (the
4489 * assoc_id field is ignored by the one to one model). Note that if
4490 * sack_delay or sack_freq are 0 when setting this option, then the
4491 * current values will remain unchanged.
4493 * struct sctp_sack_info {
4494 * sctp_assoc_t sack_assoc_id;
4495 * uint32_t sack_delay;
4496 * uint32_t sack_freq;
4499 * sack_assoc_id - This parameter, indicates which association the user
4500 * is performing an action upon. Note that if this field's value is
4501 * zero then the endpoints default value is changed (effecting future
4502 * associations only).
4504 * sack_delay - This parameter contains the number of milliseconds that
4505 * the user is requesting the delayed ACK timer be set to. Note that
4506 * this value is defined in the standard to be between 200 and 500
4509 * sack_freq - This parameter contains the number of packets that must
4510 * be received before a sack is sent without waiting for the delay
4511 * timer to expire. The default value for this is 2, setting this
4512 * value to 1 will disable the delayed sack algorithm.
4514 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4515 char __user *optval,
4518 struct sctp_sack_info params;
4519 struct sctp_association *asoc = NULL;
4520 struct sctp_sock *sp = sctp_sk(sk);
4522 if (len >= sizeof(struct sctp_sack_info)) {
4523 len = sizeof(struct sctp_sack_info);
4525 if (copy_from_user(¶ms, optval, len))
4527 } else if (len == sizeof(struct sctp_assoc_value)) {
4528 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4529 pr_warn("Use struct sctp_sack_info instead\n");
4530 if (copy_from_user(¶ms, optval, len))
4535 /* Get association, if sack_assoc_id != 0 and the socket is a one
4536 * to many style socket, and an association was not found, then
4537 * the id was invalid.
4539 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4540 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4544 /* Fetch association values. */
4545 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4546 params.sack_delay = jiffies_to_msecs(
4548 params.sack_freq = asoc->sackfreq;
4551 params.sack_delay = 0;
4552 params.sack_freq = 1;
4555 /* Fetch socket values. */
4556 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4557 params.sack_delay = sp->sackdelay;
4558 params.sack_freq = sp->sackfreq;
4560 params.sack_delay = 0;
4561 params.sack_freq = 1;
4565 if (copy_to_user(optval, ¶ms, len))
4568 if (put_user(len, optlen))
4574 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4576 * Applications can specify protocol parameters for the default association
4577 * initialization. The option name argument to setsockopt() and getsockopt()
4580 * Setting initialization parameters is effective only on an unconnected
4581 * socket (for UDP-style sockets only future associations are effected
4582 * by the change). With TCP-style sockets, this option is inherited by
4583 * sockets derived from a listener socket.
4585 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4587 if (len < sizeof(struct sctp_initmsg))
4589 len = sizeof(struct sctp_initmsg);
4590 if (put_user(len, optlen))
4592 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4598 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4599 char __user *optval, int __user *optlen)
4601 struct sctp_association *asoc;
4603 struct sctp_getaddrs getaddrs;
4604 struct sctp_transport *from;
4606 union sctp_addr temp;
4607 struct sctp_sock *sp = sctp_sk(sk);
4612 if (len < sizeof(struct sctp_getaddrs))
4615 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4618 /* For UDP-style sockets, id specifies the association to query. */
4619 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4623 to = optval + offsetof(struct sctp_getaddrs,addrs);
4624 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4626 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4628 memcpy(&temp, &from->ipaddr, sizeof(temp));
4629 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4630 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4631 if (space_left < addrlen)
4633 if (copy_to_user(to, &temp, addrlen))
4637 space_left -= addrlen;
4640 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4642 bytes_copied = ((char __user *)to) - optval;
4643 if (put_user(bytes_copied, optlen))
4649 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4650 size_t space_left, int *bytes_copied)
4652 struct sctp_sockaddr_entry *addr;
4653 union sctp_addr temp;
4656 struct net *net = sock_net(sk);
4659 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4663 if ((PF_INET == sk->sk_family) &&
4664 (AF_INET6 == addr->a.sa.sa_family))
4666 if ((PF_INET6 == sk->sk_family) &&
4667 inet_v6_ipv6only(sk) &&
4668 (AF_INET == addr->a.sa.sa_family))
4670 memcpy(&temp, &addr->a, sizeof(temp));
4671 if (!temp.v4.sin_port)
4672 temp.v4.sin_port = htons(port);
4674 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4676 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4677 if (space_left < addrlen) {
4681 memcpy(to, &temp, addrlen);
4685 space_left -= addrlen;
4686 *bytes_copied += addrlen;
4694 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4695 char __user *optval, int __user *optlen)
4697 struct sctp_bind_addr *bp;
4698 struct sctp_association *asoc;
4700 struct sctp_getaddrs getaddrs;
4701 struct sctp_sockaddr_entry *addr;
4703 union sctp_addr temp;
4704 struct sctp_sock *sp = sctp_sk(sk);
4708 int bytes_copied = 0;
4712 if (len < sizeof(struct sctp_getaddrs))
4715 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4719 * For UDP-style sockets, id specifies the association to query.
4720 * If the id field is set to the value '0' then the locally bound
4721 * addresses are returned without regard to any particular
4724 if (0 == getaddrs.assoc_id) {
4725 bp = &sctp_sk(sk)->ep->base.bind_addr;
4727 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4730 bp = &asoc->base.bind_addr;
4733 to = optval + offsetof(struct sctp_getaddrs,addrs);
4734 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4736 addrs = kmalloc(space_left, GFP_KERNEL);
4740 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4741 * addresses from the global local address list.
4743 if (sctp_list_single_entry(&bp->address_list)) {
4744 addr = list_entry(bp->address_list.next,
4745 struct sctp_sockaddr_entry, list);
4746 if (sctp_is_any(sk, &addr->a)) {
4747 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4748 space_left, &bytes_copied);
4758 /* Protection on the bound address list is not needed since
4759 * in the socket option context we hold a socket lock and
4760 * thus the bound address list can't change.
4762 list_for_each_entry(addr, &bp->address_list, list) {
4763 memcpy(&temp, &addr->a, sizeof(temp));
4764 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4765 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4766 if (space_left < addrlen) {
4767 err = -ENOMEM; /*fixme: right error?*/
4770 memcpy(buf, &temp, addrlen);
4772 bytes_copied += addrlen;
4774 space_left -= addrlen;
4778 if (copy_to_user(to, addrs, bytes_copied)) {
4782 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4786 if (put_user(bytes_copied, optlen))
4793 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4795 * Requests that the local SCTP stack use the enclosed peer address as
4796 * the association primary. The enclosed address must be one of the
4797 * association peer's addresses.
4799 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4800 char __user *optval, int __user *optlen)
4802 struct sctp_prim prim;
4803 struct sctp_association *asoc;
4804 struct sctp_sock *sp = sctp_sk(sk);
4806 if (len < sizeof(struct sctp_prim))
4809 len = sizeof(struct sctp_prim);
4811 if (copy_from_user(&prim, optval, len))
4814 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4818 if (!asoc->peer.primary_path)
4821 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4822 asoc->peer.primary_path->af_specific->sockaddr_len);
4824 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4825 (union sctp_addr *)&prim.ssp_addr);
4827 if (put_user(len, optlen))
4829 if (copy_to_user(optval, &prim, len))
4836 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4838 * Requests that the local endpoint set the specified Adaptation Layer
4839 * Indication parameter for all future INIT and INIT-ACK exchanges.
4841 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4842 char __user *optval, int __user *optlen)
4844 struct sctp_setadaptation adaptation;
4846 if (len < sizeof(struct sctp_setadaptation))
4849 len = sizeof(struct sctp_setadaptation);
4851 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4853 if (put_user(len, optlen))
4855 if (copy_to_user(optval, &adaptation, len))
4863 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4865 * Applications that wish to use the sendto() system call may wish to
4866 * specify a default set of parameters that would normally be supplied
4867 * through the inclusion of ancillary data. This socket option allows
4868 * such an application to set the default sctp_sndrcvinfo structure.
4871 * The application that wishes to use this socket option simply passes
4872 * in to this call the sctp_sndrcvinfo structure defined in Section
4873 * 5.2.2) The input parameters accepted by this call include
4874 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4875 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4876 * to this call if the caller is using the UDP model.
4878 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4880 static int sctp_getsockopt_default_send_param(struct sock *sk,
4881 int len, char __user *optval,
4884 struct sctp_sndrcvinfo info;
4885 struct sctp_association *asoc;
4886 struct sctp_sock *sp = sctp_sk(sk);
4888 if (len < sizeof(struct sctp_sndrcvinfo))
4891 len = sizeof(struct sctp_sndrcvinfo);
4893 if (copy_from_user(&info, optval, len))
4896 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4897 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4901 info.sinfo_stream = asoc->default_stream;
4902 info.sinfo_flags = asoc->default_flags;
4903 info.sinfo_ppid = asoc->default_ppid;
4904 info.sinfo_context = asoc->default_context;
4905 info.sinfo_timetolive = asoc->default_timetolive;
4907 info.sinfo_stream = sp->default_stream;
4908 info.sinfo_flags = sp->default_flags;
4909 info.sinfo_ppid = sp->default_ppid;
4910 info.sinfo_context = sp->default_context;
4911 info.sinfo_timetolive = sp->default_timetolive;
4914 if (put_user(len, optlen))
4916 if (copy_to_user(optval, &info, len))
4924 * 7.1.5 SCTP_NODELAY
4926 * Turn on/off any Nagle-like algorithm. This means that packets are
4927 * generally sent as soon as possible and no unnecessary delays are
4928 * introduced, at the cost of more packets in the network. Expects an
4929 * integer boolean flag.
4932 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4933 char __user *optval, int __user *optlen)
4937 if (len < sizeof(int))
4941 val = (sctp_sk(sk)->nodelay == 1);
4942 if (put_user(len, optlen))
4944 if (copy_to_user(optval, &val, len))
4951 * 7.1.1 SCTP_RTOINFO
4953 * The protocol parameters used to initialize and bound retransmission
4954 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4955 * and modify these parameters.
4956 * All parameters are time values, in milliseconds. A value of 0, when
4957 * modifying the parameters, indicates that the current value should not
4961 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4962 char __user *optval,
4963 int __user *optlen) {
4964 struct sctp_rtoinfo rtoinfo;
4965 struct sctp_association *asoc;
4967 if (len < sizeof (struct sctp_rtoinfo))
4970 len = sizeof(struct sctp_rtoinfo);
4972 if (copy_from_user(&rtoinfo, optval, len))
4975 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4977 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4980 /* Values corresponding to the specific association. */
4982 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4983 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4984 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4986 /* Values corresponding to the endpoint. */
4987 struct sctp_sock *sp = sctp_sk(sk);
4989 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4990 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4991 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4994 if (put_user(len, optlen))
4997 if (copy_to_user(optval, &rtoinfo, len))
5005 * 7.1.2 SCTP_ASSOCINFO
5007 * This option is used to tune the maximum retransmission attempts
5008 * of the association.
5009 * Returns an error if the new association retransmission value is
5010 * greater than the sum of the retransmission value of the peer.
5011 * See [SCTP] for more information.
5014 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5015 char __user *optval,
5019 struct sctp_assocparams assocparams;
5020 struct sctp_association *asoc;
5021 struct list_head *pos;
5024 if (len < sizeof (struct sctp_assocparams))
5027 len = sizeof(struct sctp_assocparams);
5029 if (copy_from_user(&assocparams, optval, len))
5032 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5034 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5037 /* Values correspoinding to the specific association */
5039 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5040 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5041 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5042 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
5044 (asoc->cookie_life.tv_usec
5047 list_for_each(pos, &asoc->peer.transport_addr_list) {
5051 assocparams.sasoc_number_peer_destinations = cnt;
5053 /* Values corresponding to the endpoint */
5054 struct sctp_sock *sp = sctp_sk(sk);
5056 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5057 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5058 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5059 assocparams.sasoc_cookie_life =
5060 sp->assocparams.sasoc_cookie_life;
5061 assocparams.sasoc_number_peer_destinations =
5063 sasoc_number_peer_destinations;
5066 if (put_user(len, optlen))
5069 if (copy_to_user(optval, &assocparams, len))
5076 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5078 * This socket option is a boolean flag which turns on or off mapped V4
5079 * addresses. If this option is turned on and the socket is type
5080 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5081 * If this option is turned off, then no mapping will be done of V4
5082 * addresses and a user will receive both PF_INET6 and PF_INET type
5083 * addresses on the socket.
5085 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5086 char __user *optval, int __user *optlen)
5089 struct sctp_sock *sp = sctp_sk(sk);
5091 if (len < sizeof(int))
5096 if (put_user(len, optlen))
5098 if (copy_to_user(optval, &val, len))
5105 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5106 * (chapter and verse is quoted at sctp_setsockopt_context())
5108 static int sctp_getsockopt_context(struct sock *sk, int len,
5109 char __user *optval, int __user *optlen)
5111 struct sctp_assoc_value params;
5112 struct sctp_sock *sp;
5113 struct sctp_association *asoc;
5115 if (len < sizeof(struct sctp_assoc_value))
5118 len = sizeof(struct sctp_assoc_value);
5120 if (copy_from_user(¶ms, optval, len))
5125 if (params.assoc_id != 0) {
5126 asoc = sctp_id2assoc(sk, params.assoc_id);
5129 params.assoc_value = asoc->default_rcv_context;
5131 params.assoc_value = sp->default_rcv_context;
5134 if (put_user(len, optlen))
5136 if (copy_to_user(optval, ¶ms, len))
5143 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5144 * This option will get or set the maximum size to put in any outgoing
5145 * SCTP DATA chunk. If a message is larger than this size it will be
5146 * fragmented by SCTP into the specified size. Note that the underlying
5147 * SCTP implementation may fragment into smaller sized chunks when the
5148 * PMTU of the underlying association is smaller than the value set by
5149 * the user. The default value for this option is '0' which indicates
5150 * the user is NOT limiting fragmentation and only the PMTU will effect
5151 * SCTP's choice of DATA chunk size. Note also that values set larger
5152 * than the maximum size of an IP datagram will effectively let SCTP
5153 * control fragmentation (i.e. the same as setting this option to 0).
5155 * The following structure is used to access and modify this parameter:
5157 * struct sctp_assoc_value {
5158 * sctp_assoc_t assoc_id;
5159 * uint32_t assoc_value;
5162 * assoc_id: This parameter is ignored for one-to-one style sockets.
5163 * For one-to-many style sockets this parameter indicates which
5164 * association the user is performing an action upon. Note that if
5165 * this field's value is zero then the endpoints default value is
5166 * changed (effecting future associations only).
5167 * assoc_value: This parameter specifies the maximum size in bytes.
5169 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5170 char __user *optval, int __user *optlen)
5172 struct sctp_assoc_value params;
5173 struct sctp_association *asoc;
5175 if (len == sizeof(int)) {
5176 pr_warn("Use of int in maxseg socket option deprecated\n");
5177 pr_warn("Use struct sctp_assoc_value instead\n");
5178 params.assoc_id = 0;
5179 } else if (len >= sizeof(struct sctp_assoc_value)) {
5180 len = sizeof(struct sctp_assoc_value);
5181 if (copy_from_user(¶ms, optval, sizeof(params)))
5186 asoc = sctp_id2assoc(sk, params.assoc_id);
5187 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5191 params.assoc_value = asoc->frag_point;
5193 params.assoc_value = sctp_sk(sk)->user_frag;
5195 if (put_user(len, optlen))
5197 if (len == sizeof(int)) {
5198 if (copy_to_user(optval, ¶ms.assoc_value, len))
5201 if (copy_to_user(optval, ¶ms, len))
5209 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5210 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5212 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5213 char __user *optval, int __user *optlen)
5217 if (len < sizeof(int))
5222 val = sctp_sk(sk)->frag_interleave;
5223 if (put_user(len, optlen))
5225 if (copy_to_user(optval, &val, len))
5232 * 7.1.25. Set or Get the sctp partial delivery point
5233 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5235 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5236 char __user *optval,
5241 if (len < sizeof(u32))
5246 val = sctp_sk(sk)->pd_point;
5247 if (put_user(len, optlen))
5249 if (copy_to_user(optval, &val, len))
5256 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5257 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5259 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5260 char __user *optval,
5263 struct sctp_assoc_value params;
5264 struct sctp_sock *sp;
5265 struct sctp_association *asoc;
5267 if (len == sizeof(int)) {
5268 pr_warn("Use of int in max_burst socket option deprecated\n");
5269 pr_warn("Use struct sctp_assoc_value instead\n");
5270 params.assoc_id = 0;
5271 } else if (len >= sizeof(struct sctp_assoc_value)) {
5272 len = sizeof(struct sctp_assoc_value);
5273 if (copy_from_user(¶ms, optval, len))
5280 if (params.assoc_id != 0) {
5281 asoc = sctp_id2assoc(sk, params.assoc_id);
5284 params.assoc_value = asoc->max_burst;
5286 params.assoc_value = sp->max_burst;
5288 if (len == sizeof(int)) {
5289 if (copy_to_user(optval, ¶ms.assoc_value, len))
5292 if (copy_to_user(optval, ¶ms, len))
5300 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5301 char __user *optval, int __user *optlen)
5303 struct sctp_hmacalgo __user *p = (void __user *)optval;
5304 struct sctp_hmac_algo_param *hmacs;
5308 if (!sctp_auth_enable)
5311 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5312 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5314 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5317 len = sizeof(struct sctp_hmacalgo) + data_len;
5318 num_idents = data_len / sizeof(u16);
5320 if (put_user(len, optlen))
5322 if (put_user(num_idents, &p->shmac_num_idents))
5324 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5329 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5330 char __user *optval, int __user *optlen)
5332 struct sctp_authkeyid val;
5333 struct sctp_association *asoc;
5335 if (!sctp_auth_enable)
5338 if (len < sizeof(struct sctp_authkeyid))
5340 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5343 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5344 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5348 val.scact_keynumber = asoc->active_key_id;
5350 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5352 len = sizeof(struct sctp_authkeyid);
5353 if (put_user(len, optlen))
5355 if (copy_to_user(optval, &val, len))
5361 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5362 char __user *optval, int __user *optlen)
5364 struct sctp_authchunks __user *p = (void __user *)optval;
5365 struct sctp_authchunks val;
5366 struct sctp_association *asoc;
5367 struct sctp_chunks_param *ch;
5371 if (!sctp_auth_enable)
5374 if (len < sizeof(struct sctp_authchunks))
5377 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5380 to = p->gauth_chunks;
5381 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5385 ch = asoc->peer.peer_chunks;
5389 /* See if the user provided enough room for all the data */
5390 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5391 if (len < num_chunks)
5394 if (copy_to_user(to, ch->chunks, num_chunks))
5397 len = sizeof(struct sctp_authchunks) + num_chunks;
5398 if (put_user(len, optlen)) return -EFAULT;
5399 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5404 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5405 char __user *optval, int __user *optlen)
5407 struct sctp_authchunks __user *p = (void __user *)optval;
5408 struct sctp_authchunks val;
5409 struct sctp_association *asoc;
5410 struct sctp_chunks_param *ch;
5414 if (!sctp_auth_enable)
5417 if (len < sizeof(struct sctp_authchunks))
5420 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5423 to = p->gauth_chunks;
5424 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5425 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5429 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5431 ch = sctp_sk(sk)->ep->auth_chunk_list;
5436 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5437 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5440 if (copy_to_user(to, ch->chunks, num_chunks))
5443 len = sizeof(struct sctp_authchunks) + num_chunks;
5444 if (put_user(len, optlen))
5446 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5453 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5454 * This option gets the current number of associations that are attached
5455 * to a one-to-many style socket. The option value is an uint32_t.
5457 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5458 char __user *optval, int __user *optlen)
5460 struct sctp_sock *sp = sctp_sk(sk);
5461 struct sctp_association *asoc;
5464 if (sctp_style(sk, TCP))
5467 if (len < sizeof(u32))
5472 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5476 if (put_user(len, optlen))
5478 if (copy_to_user(optval, &val, len))
5485 * 8.1.23 SCTP_AUTO_ASCONF
5486 * See the corresponding setsockopt entry as description
5488 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5489 char __user *optval, int __user *optlen)
5493 if (len < sizeof(int))
5497 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5499 if (put_user(len, optlen))
5501 if (copy_to_user(optval, &val, len))
5507 * 8.2.6. Get the Current Identifiers of Associations
5508 * (SCTP_GET_ASSOC_ID_LIST)
5510 * This option gets the current list of SCTP association identifiers of
5511 * the SCTP associations handled by a one-to-many style socket.
5513 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5514 char __user *optval, int __user *optlen)
5516 struct sctp_sock *sp = sctp_sk(sk);
5517 struct sctp_association *asoc;
5518 struct sctp_assoc_ids *ids;
5521 if (sctp_style(sk, TCP))
5524 if (len < sizeof(struct sctp_assoc_ids))
5527 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5531 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5534 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5536 ids = kmalloc(len, GFP_KERNEL);
5540 ids->gaids_number_of_ids = num;
5542 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5543 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5546 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5556 * SCTP_PEER_ADDR_THLDS
5558 * This option allows us to fetch the partially failed threshold for one or all
5559 * transports in an association. See Section 6.1 of:
5560 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5562 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5563 char __user *optval,
5567 struct sctp_paddrthlds val;
5568 struct sctp_transport *trans;
5569 struct sctp_association *asoc;
5571 if (len < sizeof(struct sctp_paddrthlds))
5573 len = sizeof(struct sctp_paddrthlds);
5574 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5577 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5578 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5582 val.spt_pathpfthld = asoc->pf_retrans;
5583 val.spt_pathmaxrxt = asoc->pathmaxrxt;
5585 trans = sctp_addr_id2transport(sk, &val.spt_address,
5590 val.spt_pathmaxrxt = trans->pathmaxrxt;
5591 val.spt_pathpfthld = trans->pf_retrans;
5594 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5600 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5601 char __user *optval, int __user *optlen)
5606 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5609 /* I can hardly begin to describe how wrong this is. This is
5610 * so broken as to be worse than useless. The API draft
5611 * REALLY is NOT helpful here... I am not convinced that the
5612 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5613 * are at all well-founded.
5615 if (level != SOL_SCTP) {
5616 struct sctp_af *af = sctp_sk(sk)->pf->af;
5618 retval = af->getsockopt(sk, level, optname, optval, optlen);
5622 if (get_user(len, optlen))
5629 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5631 case SCTP_DISABLE_FRAGMENTS:
5632 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5636 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5638 case SCTP_AUTOCLOSE:
5639 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5641 case SCTP_SOCKOPT_PEELOFF:
5642 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5644 case SCTP_PEER_ADDR_PARAMS:
5645 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5648 case SCTP_DELAYED_SACK:
5649 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5653 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5655 case SCTP_GET_PEER_ADDRS:
5656 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5659 case SCTP_GET_LOCAL_ADDRS:
5660 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5663 case SCTP_SOCKOPT_CONNECTX3:
5664 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5666 case SCTP_DEFAULT_SEND_PARAM:
5667 retval = sctp_getsockopt_default_send_param(sk, len,
5670 case SCTP_PRIMARY_ADDR:
5671 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5674 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5677 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5679 case SCTP_ASSOCINFO:
5680 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5682 case SCTP_I_WANT_MAPPED_V4_ADDR:
5683 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5686 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5688 case SCTP_GET_PEER_ADDR_INFO:
5689 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5692 case SCTP_ADAPTATION_LAYER:
5693 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5697 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5699 case SCTP_FRAGMENT_INTERLEAVE:
5700 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5703 case SCTP_PARTIAL_DELIVERY_POINT:
5704 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5707 case SCTP_MAX_BURST:
5708 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5711 case SCTP_AUTH_CHUNK:
5712 case SCTP_AUTH_DELETE_KEY:
5713 retval = -EOPNOTSUPP;
5715 case SCTP_HMAC_IDENT:
5716 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5718 case SCTP_AUTH_ACTIVE_KEY:
5719 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5721 case SCTP_PEER_AUTH_CHUNKS:
5722 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5725 case SCTP_LOCAL_AUTH_CHUNKS:
5726 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5729 case SCTP_GET_ASSOC_NUMBER:
5730 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5732 case SCTP_GET_ASSOC_ID_LIST:
5733 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5735 case SCTP_AUTO_ASCONF:
5736 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5738 case SCTP_PEER_ADDR_THLDS:
5739 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
5742 retval = -ENOPROTOOPT;
5746 sctp_release_sock(sk);
5750 static void sctp_hash(struct sock *sk)
5755 static void sctp_unhash(struct sock *sk)
5760 /* Check if port is acceptable. Possibly find first available port.
5762 * The port hash table (contained in the 'global' SCTP protocol storage
5763 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5764 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5765 * list (the list number is the port number hashed out, so as you
5766 * would expect from a hash function, all the ports in a given list have
5767 * such a number that hashes out to the same list number; you were
5768 * expecting that, right?); so each list has a set of ports, with a
5769 * link to the socket (struct sock) that uses it, the port number and
5770 * a fastreuse flag (FIXME: NPI ipg).
5772 static struct sctp_bind_bucket *sctp_bucket_create(
5773 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
5775 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5777 struct sctp_bind_hashbucket *head; /* hash list */
5778 struct sctp_bind_bucket *pp; /* hash list port iterator */
5779 struct hlist_node *node;
5780 unsigned short snum;
5783 snum = ntohs(addr->v4.sin_port);
5785 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5786 sctp_local_bh_disable();
5789 /* Search for an available port. */
5790 int low, high, remaining, index;
5793 inet_get_local_port_range(&low, &high);
5794 remaining = (high - low) + 1;
5795 rover = net_random() % remaining + low;
5799 if ((rover < low) || (rover > high))
5801 if (inet_is_reserved_local_port(rover))
5803 index = sctp_phashfn(sock_net(sk), rover);
5804 head = &sctp_port_hashtable[index];
5805 sctp_spin_lock(&head->lock);
5806 sctp_for_each_hentry(pp, node, &head->chain)
5807 if ((pp->port == rover) &&
5808 net_eq(sock_net(sk), pp->net))
5812 sctp_spin_unlock(&head->lock);
5813 } while (--remaining > 0);
5815 /* Exhausted local port range during search? */
5820 /* OK, here is the one we will use. HEAD (the port
5821 * hash table list entry) is non-NULL and we hold it's
5826 /* We are given an specific port number; we verify
5827 * that it is not being used. If it is used, we will
5828 * exahust the search in the hash list corresponding
5829 * to the port number (snum) - we detect that with the
5830 * port iterator, pp being NULL.
5832 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
5833 sctp_spin_lock(&head->lock);
5834 sctp_for_each_hentry(pp, node, &head->chain) {
5835 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
5842 if (!hlist_empty(&pp->owner)) {
5843 /* We had a port hash table hit - there is an
5844 * available port (pp != NULL) and it is being
5845 * used by other socket (pp->owner not empty); that other
5846 * socket is going to be sk2.
5848 int reuse = sk->sk_reuse;
5851 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5852 if (pp->fastreuse && sk->sk_reuse &&
5853 sk->sk_state != SCTP_SS_LISTENING)
5856 /* Run through the list of sockets bound to the port
5857 * (pp->port) [via the pointers bind_next and
5858 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5859 * we get the endpoint they describe and run through
5860 * the endpoint's list of IP (v4 or v6) addresses,
5861 * comparing each of the addresses with the address of
5862 * the socket sk. If we find a match, then that means
5863 * that this port/socket (sk) combination are already
5866 sk_for_each_bound(sk2, node, &pp->owner) {
5867 struct sctp_endpoint *ep2;
5868 ep2 = sctp_sk(sk2)->ep;
5871 (reuse && sk2->sk_reuse &&
5872 sk2->sk_state != SCTP_SS_LISTENING))
5875 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5876 sctp_sk(sk2), sctp_sk(sk))) {
5881 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5884 /* If there was a hash table miss, create a new port. */
5886 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
5889 /* In either case (hit or miss), make sure fastreuse is 1 only
5890 * if sk->sk_reuse is too (that is, if the caller requested
5891 * SO_REUSEADDR on this socket -sk-).
5893 if (hlist_empty(&pp->owner)) {
5894 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5898 } else if (pp->fastreuse &&
5899 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5902 /* We are set, so fill up all the data in the hash table
5903 * entry, tie the socket list information with the rest of the
5904 * sockets FIXME: Blurry, NPI (ipg).
5907 if (!sctp_sk(sk)->bind_hash) {
5908 inet_sk(sk)->inet_num = snum;
5909 sk_add_bind_node(sk, &pp->owner);
5910 sctp_sk(sk)->bind_hash = pp;
5915 sctp_spin_unlock(&head->lock);
5918 sctp_local_bh_enable();
5922 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5923 * port is requested.
5925 static int sctp_get_port(struct sock *sk, unsigned short snum)
5928 union sctp_addr addr;
5929 struct sctp_af *af = sctp_sk(sk)->pf->af;
5931 /* Set up a dummy address struct from the sk. */
5932 af->from_sk(&addr, sk);
5933 addr.v4.sin_port = htons(snum);
5935 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5936 ret = sctp_get_port_local(sk, &addr);
5942 * Move a socket to LISTENING state.
5944 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5946 struct sctp_sock *sp = sctp_sk(sk);
5947 struct sctp_endpoint *ep = sp->ep;
5948 struct crypto_hash *tfm = NULL;
5950 /* Allocate HMAC for generating cookie. */
5951 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5952 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5954 net_info_ratelimited("failed to load transform for %s: %ld\n",
5955 sctp_hmac_alg, PTR_ERR(tfm));
5958 sctp_sk(sk)->hmac = tfm;
5962 * If a bind() or sctp_bindx() is not called prior to a listen()
5963 * call that allows new associations to be accepted, the system
5964 * picks an ephemeral port and will choose an address set equivalent
5965 * to binding with a wildcard address.
5967 * This is not currently spelled out in the SCTP sockets
5968 * extensions draft, but follows the practice as seen in TCP
5972 sk->sk_state = SCTP_SS_LISTENING;
5973 if (!ep->base.bind_addr.port) {
5974 if (sctp_autobind(sk))
5977 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
5978 sk->sk_state = SCTP_SS_CLOSED;
5983 sk->sk_max_ack_backlog = backlog;
5984 sctp_hash_endpoint(ep);
5989 * 4.1.3 / 5.1.3 listen()
5991 * By default, new associations are not accepted for UDP style sockets.
5992 * An application uses listen() to mark a socket as being able to
5993 * accept new associations.
5995 * On TCP style sockets, applications use listen() to ready the SCTP
5996 * endpoint for accepting inbound associations.
5998 * On both types of endpoints a backlog of '0' disables listening.
6000 * Move a socket to LISTENING state.
6002 int sctp_inet_listen(struct socket *sock, int backlog)
6004 struct sock *sk = sock->sk;
6005 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6008 if (unlikely(backlog < 0))
6013 /* Peeled-off sockets are not allowed to listen(). */
6014 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6017 if (sock->state != SS_UNCONNECTED)
6020 /* If backlog is zero, disable listening. */
6022 if (sctp_sstate(sk, CLOSED))
6026 sctp_unhash_endpoint(ep);
6027 sk->sk_state = SCTP_SS_CLOSED;
6029 sctp_sk(sk)->bind_hash->fastreuse = 1;
6033 /* If we are already listening, just update the backlog */
6034 if (sctp_sstate(sk, LISTENING))
6035 sk->sk_max_ack_backlog = backlog;
6037 err = sctp_listen_start(sk, backlog);
6044 sctp_release_sock(sk);
6049 * This function is done by modeling the current datagram_poll() and the
6050 * tcp_poll(). Note that, based on these implementations, we don't
6051 * lock the socket in this function, even though it seems that,
6052 * ideally, locking or some other mechanisms can be used to ensure
6053 * the integrity of the counters (sndbuf and wmem_alloc) used
6054 * in this place. We assume that we don't need locks either until proven
6057 * Another thing to note is that we include the Async I/O support
6058 * here, again, by modeling the current TCP/UDP code. We don't have
6059 * a good way to test with it yet.
6061 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6063 struct sock *sk = sock->sk;
6064 struct sctp_sock *sp = sctp_sk(sk);
6067 poll_wait(file, sk_sleep(sk), wait);
6069 /* A TCP-style listening socket becomes readable when the accept queue
6072 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6073 return (!list_empty(&sp->ep->asocs)) ?
6074 (POLLIN | POLLRDNORM) : 0;
6078 /* Is there any exceptional events? */
6079 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6081 if (sk->sk_shutdown & RCV_SHUTDOWN)
6082 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6083 if (sk->sk_shutdown == SHUTDOWN_MASK)
6086 /* Is it readable? Reconsider this code with TCP-style support. */
6087 if (!skb_queue_empty(&sk->sk_receive_queue))
6088 mask |= POLLIN | POLLRDNORM;
6090 /* The association is either gone or not ready. */
6091 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6094 /* Is it writable? */
6095 if (sctp_writeable(sk)) {
6096 mask |= POLLOUT | POLLWRNORM;
6098 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6100 * Since the socket is not locked, the buffer
6101 * might be made available after the writeable check and
6102 * before the bit is set. This could cause a lost I/O
6103 * signal. tcp_poll() has a race breaker for this race
6104 * condition. Based on their implementation, we put
6105 * in the following code to cover it as well.
6107 if (sctp_writeable(sk))
6108 mask |= POLLOUT | POLLWRNORM;
6113 /********************************************************************
6114 * 2nd Level Abstractions
6115 ********************************************************************/
6117 static struct sctp_bind_bucket *sctp_bucket_create(
6118 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6120 struct sctp_bind_bucket *pp;
6122 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6124 SCTP_DBG_OBJCNT_INC(bind_bucket);
6127 INIT_HLIST_HEAD(&pp->owner);
6129 hlist_add_head(&pp->node, &head->chain);
6134 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6135 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6137 if (pp && hlist_empty(&pp->owner)) {
6138 __hlist_del(&pp->node);
6139 kmem_cache_free(sctp_bucket_cachep, pp);
6140 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6144 /* Release this socket's reference to a local port. */
6145 static inline void __sctp_put_port(struct sock *sk)
6147 struct sctp_bind_hashbucket *head =
6148 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6149 inet_sk(sk)->inet_num)];
6150 struct sctp_bind_bucket *pp;
6152 sctp_spin_lock(&head->lock);
6153 pp = sctp_sk(sk)->bind_hash;
6154 __sk_del_bind_node(sk);
6155 sctp_sk(sk)->bind_hash = NULL;
6156 inet_sk(sk)->inet_num = 0;
6157 sctp_bucket_destroy(pp);
6158 sctp_spin_unlock(&head->lock);
6161 void sctp_put_port(struct sock *sk)
6163 sctp_local_bh_disable();
6164 __sctp_put_port(sk);
6165 sctp_local_bh_enable();
6169 * The system picks an ephemeral port and choose an address set equivalent
6170 * to binding with a wildcard address.
6171 * One of those addresses will be the primary address for the association.
6172 * This automatically enables the multihoming capability of SCTP.
6174 static int sctp_autobind(struct sock *sk)
6176 union sctp_addr autoaddr;
6180 /* Initialize a local sockaddr structure to INADDR_ANY. */
6181 af = sctp_sk(sk)->pf->af;
6183 port = htons(inet_sk(sk)->inet_num);
6184 af->inaddr_any(&autoaddr, port);
6186 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6189 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6192 * 4.2 The cmsghdr Structure *
6194 * When ancillary data is sent or received, any number of ancillary data
6195 * objects can be specified by the msg_control and msg_controllen members of
6196 * the msghdr structure, because each object is preceded by
6197 * a cmsghdr structure defining the object's length (the cmsg_len member).
6198 * Historically Berkeley-derived implementations have passed only one object
6199 * at a time, but this API allows multiple objects to be
6200 * passed in a single call to sendmsg() or recvmsg(). The following example
6201 * shows two ancillary data objects in a control buffer.
6203 * |<--------------------------- msg_controllen -------------------------->|
6206 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6208 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6211 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6213 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6216 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6217 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6219 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6221 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6228 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
6229 sctp_cmsgs_t *cmsgs)
6231 struct cmsghdr *cmsg;
6232 struct msghdr *my_msg = (struct msghdr *)msg;
6234 for (cmsg = CMSG_FIRSTHDR(msg);
6236 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6237 if (!CMSG_OK(my_msg, cmsg))
6240 /* Should we parse this header or ignore? */
6241 if (cmsg->cmsg_level != IPPROTO_SCTP)
6244 /* Strictly check lengths following example in SCM code. */
6245 switch (cmsg->cmsg_type) {
6247 /* SCTP Socket API Extension
6248 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6250 * This cmsghdr structure provides information for
6251 * initializing new SCTP associations with sendmsg().
6252 * The SCTP_INITMSG socket option uses this same data
6253 * structure. This structure is not used for
6256 * cmsg_level cmsg_type cmsg_data[]
6257 * ------------ ------------ ----------------------
6258 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6260 if (cmsg->cmsg_len !=
6261 CMSG_LEN(sizeof(struct sctp_initmsg)))
6263 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6267 /* SCTP Socket API Extension
6268 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6270 * This cmsghdr structure specifies SCTP options for
6271 * sendmsg() and describes SCTP header information
6272 * about a received message through recvmsg().
6274 * cmsg_level cmsg_type cmsg_data[]
6275 * ------------ ------------ ----------------------
6276 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6278 if (cmsg->cmsg_len !=
6279 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6283 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6285 /* Minimally, validate the sinfo_flags. */
6286 if (cmsgs->info->sinfo_flags &
6287 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6288 SCTP_ABORT | SCTP_EOF))
6300 * Wait for a packet..
6301 * Note: This function is the same function as in core/datagram.c
6302 * with a few modifications to make lksctp work.
6304 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6309 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6311 /* Socket errors? */
6312 error = sock_error(sk);
6316 if (!skb_queue_empty(&sk->sk_receive_queue))
6319 /* Socket shut down? */
6320 if (sk->sk_shutdown & RCV_SHUTDOWN)
6323 /* Sequenced packets can come disconnected. If so we report the
6328 /* Is there a good reason to think that we may receive some data? */
6329 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6332 /* Handle signals. */
6333 if (signal_pending(current))
6336 /* Let another process have a go. Since we are going to sleep
6337 * anyway. Note: This may cause odd behaviors if the message
6338 * does not fit in the user's buffer, but this seems to be the
6339 * only way to honor MSG_DONTWAIT realistically.
6341 sctp_release_sock(sk);
6342 *timeo_p = schedule_timeout(*timeo_p);
6346 finish_wait(sk_sleep(sk), &wait);
6350 error = sock_intr_errno(*timeo_p);
6353 finish_wait(sk_sleep(sk), &wait);
6358 /* Receive a datagram.
6359 * Note: This is pretty much the same routine as in core/datagram.c
6360 * with a few changes to make lksctp work.
6362 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6363 int noblock, int *err)
6366 struct sk_buff *skb;
6369 timeo = sock_rcvtimeo(sk, noblock);
6371 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6372 timeo, MAX_SCHEDULE_TIMEOUT);
6375 /* Again only user level code calls this function,
6376 * so nothing interrupt level
6377 * will suddenly eat the receive_queue.
6379 * Look at current nfs client by the way...
6380 * However, this function was correct in any case. 8)
6382 if (flags & MSG_PEEK) {
6383 spin_lock_bh(&sk->sk_receive_queue.lock);
6384 skb = skb_peek(&sk->sk_receive_queue);
6386 atomic_inc(&skb->users);
6387 spin_unlock_bh(&sk->sk_receive_queue.lock);
6389 skb = skb_dequeue(&sk->sk_receive_queue);
6395 /* Caller is allowed not to check sk->sk_err before calling. */
6396 error = sock_error(sk);
6400 if (sk->sk_shutdown & RCV_SHUTDOWN)
6403 /* User doesn't want to wait. */
6407 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6416 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6417 static void __sctp_write_space(struct sctp_association *asoc)
6419 struct sock *sk = asoc->base.sk;
6420 struct socket *sock = sk->sk_socket;
6422 if ((sctp_wspace(asoc) > 0) && sock) {
6423 if (waitqueue_active(&asoc->wait))
6424 wake_up_interruptible(&asoc->wait);
6426 if (sctp_writeable(sk)) {
6427 wait_queue_head_t *wq = sk_sleep(sk);
6429 if (wq && waitqueue_active(wq))
6430 wake_up_interruptible(wq);
6432 /* Note that we try to include the Async I/O support
6433 * here by modeling from the current TCP/UDP code.
6434 * We have not tested with it yet.
6436 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6437 sock_wake_async(sock,
6438 SOCK_WAKE_SPACE, POLL_OUT);
6443 /* Do accounting for the sndbuf space.
6444 * Decrement the used sndbuf space of the corresponding association by the
6445 * data size which was just transmitted(freed).
6447 static void sctp_wfree(struct sk_buff *skb)
6449 struct sctp_association *asoc;
6450 struct sctp_chunk *chunk;
6453 /* Get the saved chunk pointer. */
6454 chunk = *((struct sctp_chunk **)(skb->cb));
6457 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6458 sizeof(struct sk_buff) +
6459 sizeof(struct sctp_chunk);
6461 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6464 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6466 sk->sk_wmem_queued -= skb->truesize;
6467 sk_mem_uncharge(sk, skb->truesize);
6470 __sctp_write_space(asoc);
6472 sctp_association_put(asoc);
6475 /* Do accounting for the receive space on the socket.
6476 * Accounting for the association is done in ulpevent.c
6477 * We set this as a destructor for the cloned data skbs so that
6478 * accounting is done at the correct time.
6480 void sctp_sock_rfree(struct sk_buff *skb)
6482 struct sock *sk = skb->sk;
6483 struct sctp_ulpevent *event = sctp_skb2event(skb);
6485 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6488 * Mimic the behavior of sock_rfree
6490 sk_mem_uncharge(sk, event->rmem_len);
6494 /* Helper function to wait for space in the sndbuf. */
6495 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6498 struct sock *sk = asoc->base.sk;
6500 long current_timeo = *timeo_p;
6503 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6504 asoc, (long)(*timeo_p), msg_len);
6506 /* Increment the association's refcnt. */
6507 sctp_association_hold(asoc);
6509 /* Wait on the association specific sndbuf space. */
6511 prepare_to_wait_exclusive(&asoc->wait, &wait,
6512 TASK_INTERRUPTIBLE);
6515 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6518 if (signal_pending(current))
6519 goto do_interrupted;
6520 if (msg_len <= sctp_wspace(asoc))
6523 /* Let another process have a go. Since we are going
6526 sctp_release_sock(sk);
6527 current_timeo = schedule_timeout(current_timeo);
6528 BUG_ON(sk != asoc->base.sk);
6531 *timeo_p = current_timeo;
6535 finish_wait(&asoc->wait, &wait);
6537 /* Release the association's refcnt. */
6538 sctp_association_put(asoc);
6547 err = sock_intr_errno(*timeo_p);
6555 void sctp_data_ready(struct sock *sk, int len)
6557 struct socket_wq *wq;
6560 wq = rcu_dereference(sk->sk_wq);
6561 if (wq_has_sleeper(wq))
6562 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6563 POLLRDNORM | POLLRDBAND);
6564 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6568 /* If socket sndbuf has changed, wake up all per association waiters. */
6569 void sctp_write_space(struct sock *sk)
6571 struct sctp_association *asoc;
6573 /* Wake up the tasks in each wait queue. */
6574 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6575 __sctp_write_space(asoc);
6579 /* Is there any sndbuf space available on the socket?
6581 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6582 * associations on the same socket. For a UDP-style socket with
6583 * multiple associations, it is possible for it to be "unwriteable"
6584 * prematurely. I assume that this is acceptable because
6585 * a premature "unwriteable" is better than an accidental "writeable" which
6586 * would cause an unwanted block under certain circumstances. For the 1-1
6587 * UDP-style sockets or TCP-style sockets, this code should work.
6590 static int sctp_writeable(struct sock *sk)
6594 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6600 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6601 * returns immediately with EINPROGRESS.
6603 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6605 struct sock *sk = asoc->base.sk;
6607 long current_timeo = *timeo_p;
6610 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6613 /* Increment the association's refcnt. */
6614 sctp_association_hold(asoc);
6617 prepare_to_wait_exclusive(&asoc->wait, &wait,
6618 TASK_INTERRUPTIBLE);
6621 if (sk->sk_shutdown & RCV_SHUTDOWN)
6623 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6626 if (signal_pending(current))
6627 goto do_interrupted;
6629 if (sctp_state(asoc, ESTABLISHED))
6632 /* Let another process have a go. Since we are going
6635 sctp_release_sock(sk);
6636 current_timeo = schedule_timeout(current_timeo);
6639 *timeo_p = current_timeo;
6643 finish_wait(&asoc->wait, &wait);
6645 /* Release the association's refcnt. */
6646 sctp_association_put(asoc);
6651 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6654 err = -ECONNREFUSED;
6658 err = sock_intr_errno(*timeo_p);
6666 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6668 struct sctp_endpoint *ep;
6672 ep = sctp_sk(sk)->ep;
6676 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6677 TASK_INTERRUPTIBLE);
6679 if (list_empty(&ep->asocs)) {
6680 sctp_release_sock(sk);
6681 timeo = schedule_timeout(timeo);
6686 if (!sctp_sstate(sk, LISTENING))
6690 if (!list_empty(&ep->asocs))
6693 err = sock_intr_errno(timeo);
6694 if (signal_pending(current))
6702 finish_wait(sk_sleep(sk), &wait);
6707 static void sctp_wait_for_close(struct sock *sk, long timeout)
6712 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6713 if (list_empty(&sctp_sk(sk)->ep->asocs))
6715 sctp_release_sock(sk);
6716 timeout = schedule_timeout(timeout);
6718 } while (!signal_pending(current) && timeout);
6720 finish_wait(sk_sleep(sk), &wait);
6723 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6725 struct sk_buff *frag;
6730 /* Don't forget the fragments. */
6731 skb_walk_frags(skb, frag)
6732 sctp_skb_set_owner_r_frag(frag, sk);
6735 sctp_skb_set_owner_r(skb, sk);
6738 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6739 struct sctp_association *asoc)
6741 struct inet_sock *inet = inet_sk(sk);
6742 struct inet_sock *newinet;
6744 newsk->sk_type = sk->sk_type;
6745 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6746 newsk->sk_flags = sk->sk_flags;
6747 newsk->sk_no_check = sk->sk_no_check;
6748 newsk->sk_reuse = sk->sk_reuse;
6750 newsk->sk_shutdown = sk->sk_shutdown;
6751 newsk->sk_destruct = inet_sock_destruct;
6752 newsk->sk_family = sk->sk_family;
6753 newsk->sk_protocol = IPPROTO_SCTP;
6754 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6755 newsk->sk_sndbuf = sk->sk_sndbuf;
6756 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6757 newsk->sk_lingertime = sk->sk_lingertime;
6758 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6759 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6761 newinet = inet_sk(newsk);
6763 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6764 * getsockname() and getpeername()
6766 newinet->inet_sport = inet->inet_sport;
6767 newinet->inet_saddr = inet->inet_saddr;
6768 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6769 newinet->inet_dport = htons(asoc->peer.port);
6770 newinet->pmtudisc = inet->pmtudisc;
6771 newinet->inet_id = asoc->next_tsn ^ jiffies;
6773 newinet->uc_ttl = inet->uc_ttl;
6774 newinet->mc_loop = 1;
6775 newinet->mc_ttl = 1;
6776 newinet->mc_index = 0;
6777 newinet->mc_list = NULL;
6780 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6781 * and its messages to the newsk.
6783 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6784 struct sctp_association *assoc,
6785 sctp_socket_type_t type)
6787 struct sctp_sock *oldsp = sctp_sk(oldsk);
6788 struct sctp_sock *newsp = sctp_sk(newsk);
6789 struct sctp_bind_bucket *pp; /* hash list port iterator */
6790 struct sctp_endpoint *newep = newsp->ep;
6791 struct sk_buff *skb, *tmp;
6792 struct sctp_ulpevent *event;
6793 struct sctp_bind_hashbucket *head;
6794 struct list_head tmplist;
6796 /* Migrate socket buffer sizes and all the socket level options to the
6799 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6800 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6801 /* Brute force copy old sctp opt. */
6802 if (oldsp->do_auto_asconf) {
6803 memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
6804 inet_sk_copy_descendant(newsk, oldsk);
6805 memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
6807 inet_sk_copy_descendant(newsk, oldsk);
6809 /* Restore the ep value that was overwritten with the above structure
6815 /* Hook this new socket in to the bind_hash list. */
6816 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
6817 inet_sk(oldsk)->inet_num)];
6818 sctp_local_bh_disable();
6819 sctp_spin_lock(&head->lock);
6820 pp = sctp_sk(oldsk)->bind_hash;
6821 sk_add_bind_node(newsk, &pp->owner);
6822 sctp_sk(newsk)->bind_hash = pp;
6823 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6824 sctp_spin_unlock(&head->lock);
6825 sctp_local_bh_enable();
6827 /* Copy the bind_addr list from the original endpoint to the new
6828 * endpoint so that we can handle restarts properly
6830 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6831 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6833 /* Move any messages in the old socket's receive queue that are for the
6834 * peeled off association to the new socket's receive queue.
6836 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6837 event = sctp_skb2event(skb);
6838 if (event->asoc == assoc) {
6839 __skb_unlink(skb, &oldsk->sk_receive_queue);
6840 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6841 sctp_skb_set_owner_r_frag(skb, newsk);
6845 /* Clean up any messages pending delivery due to partial
6846 * delivery. Three cases:
6847 * 1) No partial deliver; no work.
6848 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6849 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6851 skb_queue_head_init(&newsp->pd_lobby);
6852 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6854 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6855 struct sk_buff_head *queue;
6857 /* Decide which queue to move pd_lobby skbs to. */
6858 if (assoc->ulpq.pd_mode) {
6859 queue = &newsp->pd_lobby;
6861 queue = &newsk->sk_receive_queue;
6863 /* Walk through the pd_lobby, looking for skbs that
6864 * need moved to the new socket.
6866 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6867 event = sctp_skb2event(skb);
6868 if (event->asoc == assoc) {
6869 __skb_unlink(skb, &oldsp->pd_lobby);
6870 __skb_queue_tail(queue, skb);
6871 sctp_skb_set_owner_r_frag(skb, newsk);
6875 /* Clear up any skbs waiting for the partial
6876 * delivery to finish.
6878 if (assoc->ulpq.pd_mode)
6879 sctp_clear_pd(oldsk, NULL);
6883 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6884 sctp_skb_set_owner_r_frag(skb, newsk);
6886 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6887 sctp_skb_set_owner_r_frag(skb, newsk);
6889 /* Set the type of socket to indicate that it is peeled off from the
6890 * original UDP-style socket or created with the accept() call on a
6891 * TCP-style socket..
6895 /* Mark the new socket "in-use" by the user so that any packets
6896 * that may arrive on the association after we've moved it are
6897 * queued to the backlog. This prevents a potential race between
6898 * backlog processing on the old socket and new-packet processing
6899 * on the new socket.
6901 * The caller has just allocated newsk so we can guarantee that other
6902 * paths won't try to lock it and then oldsk.
6904 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6905 sctp_assoc_migrate(assoc, newsk);
6907 /* If the association on the newsk is already closed before accept()
6908 * is called, set RCV_SHUTDOWN flag.
6910 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6911 newsk->sk_shutdown |= RCV_SHUTDOWN;
6913 newsk->sk_state = SCTP_SS_ESTABLISHED;
6914 sctp_release_sock(newsk);
6918 /* This proto struct describes the ULP interface for SCTP. */
6919 struct proto sctp_prot = {
6921 .owner = THIS_MODULE,
6922 .close = sctp_close,
6923 .connect = sctp_connect,
6924 .disconnect = sctp_disconnect,
6925 .accept = sctp_accept,
6926 .ioctl = sctp_ioctl,
6927 .init = sctp_init_sock,
6928 .destroy = sctp_destroy_sock,
6929 .shutdown = sctp_shutdown,
6930 .setsockopt = sctp_setsockopt,
6931 .getsockopt = sctp_getsockopt,
6932 .sendmsg = sctp_sendmsg,
6933 .recvmsg = sctp_recvmsg,
6935 .backlog_rcv = sctp_backlog_rcv,
6937 .unhash = sctp_unhash,
6938 .get_port = sctp_get_port,
6939 .obj_size = sizeof(struct sctp_sock),
6940 .sysctl_mem = sysctl_sctp_mem,
6941 .sysctl_rmem = sysctl_sctp_rmem,
6942 .sysctl_wmem = sysctl_sctp_wmem,
6943 .memory_pressure = &sctp_memory_pressure,
6944 .enter_memory_pressure = sctp_enter_memory_pressure,
6945 .memory_allocated = &sctp_memory_allocated,
6946 .sockets_allocated = &sctp_sockets_allocated,
6949 #if IS_ENABLED(CONFIG_IPV6)
6951 struct proto sctpv6_prot = {
6953 .owner = THIS_MODULE,
6954 .close = sctp_close,
6955 .connect = sctp_connect,
6956 .disconnect = sctp_disconnect,
6957 .accept = sctp_accept,
6958 .ioctl = sctp_ioctl,
6959 .init = sctp_init_sock,
6960 .destroy = sctp_destroy_sock,
6961 .shutdown = sctp_shutdown,
6962 .setsockopt = sctp_setsockopt,
6963 .getsockopt = sctp_getsockopt,
6964 .sendmsg = sctp_sendmsg,
6965 .recvmsg = sctp_recvmsg,
6967 .backlog_rcv = sctp_backlog_rcv,
6969 .unhash = sctp_unhash,
6970 .get_port = sctp_get_port,
6971 .obj_size = sizeof(struct sctp6_sock),
6972 .sysctl_mem = sysctl_sctp_mem,
6973 .sysctl_rmem = sysctl_sctp_rmem,
6974 .sysctl_wmem = sysctl_sctp_wmem,
6975 .memory_pressure = &sctp_memory_pressure,
6976 .enter_memory_pressure = sctp_enter_memory_pressure,
6977 .memory_allocated = &sctp_memory_allocated,
6978 .sockets_allocated = &sctp_sockets_allocated,
6980 #endif /* IS_ENABLED(CONFIG_IPV6) */