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 ipv6_addr_copy(&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;
820 /* We do not need RCU protection throughout this loop
821 * because this is done under a socket lock from the
824 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
832 /* Reset use_as_src flag for the addresses in the bind address
833 * list that are to be deleted.
836 for (i = 0; i < addrcnt; i++) {
838 af = sctp_get_af_specific(laddr->v4.sin_family);
839 list_for_each_entry(saddr, &bp->address_list, list) {
840 if (sctp_cmp_addr_exact(&saddr->a, laddr))
841 saddr->state = SCTP_ADDR_DEL;
843 addr_buf += af->sockaddr_len;
846 /* Update the route and saddr entries for all the transports
847 * as some of the addresses in the bind address list are
848 * about to be deleted and cannot be used as source addresses.
850 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
852 dst_release(transport->dst);
853 sctp_transport_route(transport, NULL,
854 sctp_sk(asoc->base.sk));
858 /* We don't need to transmit ASCONF */
860 retval = sctp_send_asconf(asoc, chunk);
866 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
867 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
869 struct sock *sk = sctp_opt2sk(sp);
870 union sctp_addr *addr;
873 /* It is safe to write port space in caller. */
875 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
876 af = sctp_get_af_specific(addr->sa.sa_family);
879 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
882 if (addrw->state == SCTP_ADDR_NEW)
883 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
885 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
888 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
891 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
894 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
895 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
898 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
899 * Section 3.1.2 for this usage.
901 * addrs is a pointer to an array of one or more socket addresses. Each
902 * address is contained in its appropriate structure (i.e. struct
903 * sockaddr_in or struct sockaddr_in6) the family of the address type
904 * must be used to distinguish the address length (note that this
905 * representation is termed a "packed array" of addresses). The caller
906 * specifies the number of addresses in the array with addrcnt.
908 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
909 * -1, and sets errno to the appropriate error code.
911 * For SCTP, the port given in each socket address must be the same, or
912 * sctp_bindx() will fail, setting errno to EINVAL.
914 * The flags parameter is formed from the bitwise OR of zero or more of
915 * the following currently defined flags:
917 * SCTP_BINDX_ADD_ADDR
919 * SCTP_BINDX_REM_ADDR
921 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
922 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
923 * addresses from the association. The two flags are mutually exclusive;
924 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
925 * not remove all addresses from an association; sctp_bindx() will
926 * reject such an attempt with EINVAL.
928 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
929 * additional addresses with an endpoint after calling bind(). Or use
930 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
931 * socket is associated with so that no new association accepted will be
932 * associated with those addresses. If the endpoint supports dynamic
933 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
934 * endpoint to send the appropriate message to the peer to change the
935 * peers address lists.
937 * Adding and removing addresses from a connected association is
938 * optional functionality. Implementations that do not support this
939 * functionality should return EOPNOTSUPP.
941 * Basically do nothing but copying the addresses from user to kernel
942 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
943 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
946 * We don't use copy_from_user() for optimization: we first do the
947 * sanity checks (buffer size -fast- and access check-healthy
948 * pointer); if all of those succeed, then we can alloc the memory
949 * (expensive operation) needed to copy the data to kernel. Then we do
950 * the copying without checking the user space area
951 * (__copy_from_user()).
953 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
956 * sk The sk of the socket
957 * addrs The pointer to the addresses in user land
958 * addrssize Size of the addrs buffer
959 * op Operation to perform (add or remove, see the flags of
962 * Returns 0 if ok, <0 errno code on error.
964 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
965 struct sockaddr __user *addrs,
966 int addrs_size, int op)
968 struct sockaddr *kaddrs;
972 struct sockaddr *sa_addr;
976 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
977 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
979 if (unlikely(addrs_size <= 0))
982 /* Check the user passed a healthy pointer. */
983 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
986 /* Alloc space for the address array in kernel memory. */
987 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
988 if (unlikely(!kaddrs))
991 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
996 /* Walk through the addrs buffer and count the number of addresses. */
998 while (walk_size < addrs_size) {
999 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1005 af = sctp_get_af_specific(sa_addr->sa_family);
1007 /* If the address family is not supported or if this address
1008 * causes the address buffer to overflow return EINVAL.
1010 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1015 addr_buf += af->sockaddr_len;
1016 walk_size += af->sockaddr_len;
1021 case SCTP_BINDX_ADD_ADDR:
1022 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1025 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1028 case SCTP_BINDX_REM_ADDR:
1029 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1032 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1046 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1048 * Common routine for handling connect() and sctp_connectx().
1049 * Connect will come in with just a single address.
1051 static int __sctp_connect(struct sock* sk,
1052 struct sockaddr *kaddrs,
1054 sctp_assoc_t *assoc_id)
1056 struct sctp_sock *sp;
1057 struct sctp_endpoint *ep;
1058 struct sctp_association *asoc = NULL;
1059 struct sctp_association *asoc2;
1060 struct sctp_transport *transport;
1068 union sctp_addr *sa_addr = NULL;
1070 unsigned short port;
1071 unsigned int f_flags = 0;
1076 /* connect() cannot be done on a socket that is already in ESTABLISHED
1077 * state - UDP-style peeled off socket or a TCP-style socket that
1078 * is already connected.
1079 * It cannot be done even on a TCP-style listening socket.
1081 if (sctp_sstate(sk, ESTABLISHED) ||
1082 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1087 /* Walk through the addrs buffer and count the number of addresses. */
1089 while (walk_size < addrs_size) {
1090 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1096 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1098 /* If the address family is not supported or if this address
1099 * causes the address buffer to overflow return EINVAL.
1101 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1106 port = ntohs(sa_addr->v4.sin_port);
1108 /* Save current address so we can work with it */
1109 memcpy(&to, sa_addr, af->sockaddr_len);
1111 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1115 /* Make sure the destination port is correctly set
1118 if (asoc && asoc->peer.port && asoc->peer.port != port)
1122 /* Check if there already is a matching association on the
1123 * endpoint (other than the one created here).
1125 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1126 if (asoc2 && asoc2 != asoc) {
1127 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1134 /* If we could not find a matching association on the endpoint,
1135 * make sure that there is no peeled-off association matching
1136 * the peer address even on another socket.
1138 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1139 err = -EADDRNOTAVAIL;
1144 /* If a bind() or sctp_bindx() is not called prior to
1145 * an sctp_connectx() call, the system picks an
1146 * ephemeral port and will choose an address set
1147 * equivalent to binding with a wildcard address.
1149 if (!ep->base.bind_addr.port) {
1150 if (sctp_autobind(sk)) {
1156 * If an unprivileged user inherits a 1-many
1157 * style socket with open associations on a
1158 * privileged port, it MAY be permitted to
1159 * accept new associations, but it SHOULD NOT
1160 * be permitted to open new associations.
1162 if (ep->base.bind_addr.port < PROT_SOCK &&
1163 !capable(CAP_NET_BIND_SERVICE)) {
1169 scope = sctp_scope(&to);
1170 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1176 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1184 /* Prime the peer's transport structures. */
1185 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1193 addr_buf += af->sockaddr_len;
1194 walk_size += af->sockaddr_len;
1197 /* In case the user of sctp_connectx() wants an association
1198 * id back, assign one now.
1201 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1206 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1211 /* Initialize sk's dport and daddr for getpeername() */
1212 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1213 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1214 af->to_sk_daddr(sa_addr, sk);
1217 /* in-kernel sockets don't generally have a file allocated to them
1218 * if all they do is call sock_create_kern().
1220 if (sk->sk_socket->file)
1221 f_flags = sk->sk_socket->file->f_flags;
1223 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1225 err = sctp_wait_for_connect(asoc, &timeo);
1226 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1227 *assoc_id = asoc->assoc_id;
1229 /* Don't free association on exit. */
1234 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1235 " kaddrs: %p err: %d\n",
1238 /* sctp_primitive_ASSOCIATE may have added this association
1239 * To the hash table, try to unhash it, just in case, its a noop
1240 * if it wasn't hashed so we're safe
1242 sctp_unhash_established(asoc);
1243 sctp_association_free(asoc);
1248 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1251 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1252 * sctp_assoc_t *asoc);
1254 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1255 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1256 * or IPv6 addresses.
1258 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1259 * Section 3.1.2 for this usage.
1261 * addrs is a pointer to an array of one or more socket addresses. Each
1262 * address is contained in its appropriate structure (i.e. struct
1263 * sockaddr_in or struct sockaddr_in6) the family of the address type
1264 * must be used to distengish the address length (note that this
1265 * representation is termed a "packed array" of addresses). The caller
1266 * specifies the number of addresses in the array with addrcnt.
1268 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1269 * the association id of the new association. On failure, sctp_connectx()
1270 * returns -1, and sets errno to the appropriate error code. The assoc_id
1271 * is not touched by the kernel.
1273 * For SCTP, the port given in each socket address must be the same, or
1274 * sctp_connectx() will fail, setting errno to EINVAL.
1276 * An application can use sctp_connectx to initiate an association with
1277 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1278 * allows a caller to specify multiple addresses at which a peer can be
1279 * reached. The way the SCTP stack uses the list of addresses to set up
1280 * the association is implementation dependent. This function only
1281 * specifies that the stack will try to make use of all the addresses in
1282 * the list when needed.
1284 * Note that the list of addresses passed in is only used for setting up
1285 * the association. It does not necessarily equal the set of addresses
1286 * the peer uses for the resulting association. If the caller wants to
1287 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1288 * retrieve them after the association has been set up.
1290 * Basically do nothing but copying the addresses from user to kernel
1291 * land and invoking either sctp_connectx(). This is used for tunneling
1292 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1294 * We don't use copy_from_user() for optimization: we first do the
1295 * sanity checks (buffer size -fast- and access check-healthy
1296 * pointer); if all of those succeed, then we can alloc the memory
1297 * (expensive operation) needed to copy the data to kernel. Then we do
1298 * the copying without checking the user space area
1299 * (__copy_from_user()).
1301 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1304 * sk The sk of the socket
1305 * addrs The pointer to the addresses in user land
1306 * addrssize Size of the addrs buffer
1308 * Returns >=0 if ok, <0 errno code on error.
1310 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1311 struct sockaddr __user *addrs,
1313 sctp_assoc_t *assoc_id)
1316 struct sockaddr *kaddrs;
1318 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1319 __func__, sk, addrs, addrs_size);
1321 if (unlikely(addrs_size <= 0))
1324 /* Check the user passed a healthy pointer. */
1325 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1328 /* Alloc space for the address array in kernel memory. */
1329 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1330 if (unlikely(!kaddrs))
1333 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1336 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1345 * This is an older interface. It's kept for backward compatibility
1346 * to the option that doesn't provide association id.
1348 SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1349 struct sockaddr __user *addrs,
1352 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1356 * New interface for the API. The since the API is done with a socket
1357 * option, to make it simple we feed back the association id is as a return
1358 * indication to the call. Error is always negative and association id is
1361 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1362 struct sockaddr __user *addrs,
1365 sctp_assoc_t assoc_id = 0;
1368 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1377 * New (hopefully final) interface for the API.
1378 * We use the sctp_getaddrs_old structure so that use-space library
1379 * can avoid any unnecessary allocations. The only defferent part
1380 * is that we store the actual length of the address buffer into the
1381 * addrs_num structure member. That way we can re-use the existing
1384 SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1385 char __user *optval,
1388 struct sctp_getaddrs_old param;
1389 sctp_assoc_t assoc_id = 0;
1392 if (len < sizeof(param))
1395 if (copy_from_user(¶m, optval, sizeof(param)))
1398 err = __sctp_setsockopt_connectx(sk,
1399 (struct sockaddr __user *)param.addrs,
1400 param.addr_num, &assoc_id);
1402 if (err == 0 || err == -EINPROGRESS) {
1403 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1405 if (put_user(sizeof(assoc_id), optlen))
1412 /* API 3.1.4 close() - UDP Style Syntax
1413 * Applications use close() to perform graceful shutdown (as described in
1414 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1415 * by a UDP-style socket.
1419 * ret = close(int sd);
1421 * sd - the socket descriptor of the associations to be closed.
1423 * To gracefully shutdown a specific association represented by the
1424 * UDP-style socket, an application should use the sendmsg() call,
1425 * passing no user data, but including the appropriate flag in the
1426 * ancillary data (see Section xxxx).
1428 * If sd in the close() call is a branched-off socket representing only
1429 * one association, the shutdown is performed on that association only.
1431 * 4.1.6 close() - TCP Style Syntax
1433 * Applications use close() to gracefully close down an association.
1437 * int close(int sd);
1439 * sd - the socket descriptor of the association to be closed.
1441 * After an application calls close() on a socket descriptor, no further
1442 * socket operations will succeed on that descriptor.
1444 * API 7.1.4 SO_LINGER
1446 * An application using the TCP-style socket can use this option to
1447 * perform the SCTP ABORT primitive. The linger option structure is:
1450 * int l_onoff; // option on/off
1451 * int l_linger; // linger time
1454 * To enable the option, set l_onoff to 1. If the l_linger value is set
1455 * to 0, calling close() is the same as the ABORT primitive. If the
1456 * value is set to a negative value, the setsockopt() call will return
1457 * an error. If the value is set to a positive value linger_time, the
1458 * close() can be blocked for at most linger_time ms. If the graceful
1459 * shutdown phase does not finish during this period, close() will
1460 * return but the graceful shutdown phase continues in the system.
1462 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1464 struct sctp_endpoint *ep;
1465 struct sctp_association *asoc;
1466 struct list_head *pos, *temp;
1467 unsigned int data_was_unread;
1469 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1472 sk->sk_shutdown = SHUTDOWN_MASK;
1473 sk->sk_state = SCTP_SS_CLOSING;
1475 ep = sctp_sk(sk)->ep;
1477 /* Clean up any skbs sitting on the receive queue. */
1478 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1479 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1481 /* Walk all associations on an endpoint. */
1482 list_for_each_safe(pos, temp, &ep->asocs) {
1483 asoc = list_entry(pos, struct sctp_association, asocs);
1485 if (sctp_style(sk, TCP)) {
1486 /* A closed association can still be in the list if
1487 * it belongs to a TCP-style listening socket that is
1488 * not yet accepted. If so, free it. If not, send an
1489 * ABORT or SHUTDOWN based on the linger options.
1491 if (sctp_state(asoc, CLOSED)) {
1492 sctp_unhash_established(asoc);
1493 sctp_association_free(asoc);
1498 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1499 !skb_queue_empty(&asoc->ulpq.reasm) ||
1500 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1501 struct sctp_chunk *chunk;
1503 chunk = sctp_make_abort_user(asoc, NULL, 0);
1505 sctp_primitive_ABORT(asoc, chunk);
1507 sctp_primitive_SHUTDOWN(asoc, NULL);
1510 /* On a TCP-style socket, block for at most linger_time if set. */
1511 if (sctp_style(sk, TCP) && timeout)
1512 sctp_wait_for_close(sk, timeout);
1514 /* This will run the backlog queue. */
1515 sctp_release_sock(sk);
1517 /* Supposedly, no process has access to the socket, but
1518 * the net layers still may.
1520 sctp_local_bh_disable();
1521 sctp_bh_lock_sock(sk);
1523 /* Hold the sock, since sk_common_release() will put sock_put()
1524 * and we have just a little more cleanup.
1527 sk_common_release(sk);
1529 sctp_bh_unlock_sock(sk);
1530 sctp_local_bh_enable();
1534 SCTP_DBG_OBJCNT_DEC(sock);
1537 /* Handle EPIPE error. */
1538 static int sctp_error(struct sock *sk, int flags, int err)
1541 err = sock_error(sk) ? : -EPIPE;
1542 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1543 send_sig(SIGPIPE, current, 0);
1547 /* API 3.1.3 sendmsg() - UDP Style Syntax
1549 * An application uses sendmsg() and recvmsg() calls to transmit data to
1550 * and receive data from its peer.
1552 * ssize_t sendmsg(int socket, const struct msghdr *message,
1555 * socket - the socket descriptor of the endpoint.
1556 * message - pointer to the msghdr structure which contains a single
1557 * user message and possibly some ancillary data.
1559 * See Section 5 for complete description of the data
1562 * flags - flags sent or received with the user message, see Section
1563 * 5 for complete description of the flags.
1565 * Note: This function could use a rewrite especially when explicit
1566 * connect support comes in.
1568 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1570 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1572 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1573 struct msghdr *msg, size_t msg_len)
1575 struct sctp_sock *sp;
1576 struct sctp_endpoint *ep;
1577 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1578 struct sctp_transport *transport, *chunk_tp;
1579 struct sctp_chunk *chunk;
1581 struct sockaddr *msg_name = NULL;
1582 struct sctp_sndrcvinfo default_sinfo;
1583 struct sctp_sndrcvinfo *sinfo;
1584 struct sctp_initmsg *sinit;
1585 sctp_assoc_t associd = 0;
1586 sctp_cmsgs_t cmsgs = { NULL };
1590 __u16 sinfo_flags = 0;
1591 struct sctp_datamsg *datamsg;
1592 int msg_flags = msg->msg_flags;
1594 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1601 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1603 /* We cannot send a message over a TCP-style listening socket. */
1604 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1609 /* Parse out the SCTP CMSGs. */
1610 err = sctp_msghdr_parse(msg, &cmsgs);
1613 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1617 /* Fetch the destination address for this packet. This
1618 * address only selects the association--it is not necessarily
1619 * the address we will send to.
1620 * For a peeled-off socket, msg_name is ignored.
1622 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1623 int msg_namelen = msg->msg_namelen;
1625 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1630 if (msg_namelen > sizeof(to))
1631 msg_namelen = sizeof(to);
1632 memcpy(&to, msg->msg_name, msg_namelen);
1633 msg_name = msg->msg_name;
1639 /* Did the user specify SNDRCVINFO? */
1641 sinfo_flags = sinfo->sinfo_flags;
1642 associd = sinfo->sinfo_assoc_id;
1645 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1646 msg_len, sinfo_flags);
1648 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1649 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1654 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1655 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1656 * If SCTP_ABORT is set, the message length could be non zero with
1657 * the msg_iov set to the user abort reason.
1659 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1660 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1665 /* If SCTP_ADDR_OVER is set, there must be an address
1666 * specified in msg_name.
1668 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1675 SCTP_DEBUG_PRINTK("About to look up association.\n");
1679 /* If a msg_name has been specified, assume this is to be used. */
1681 /* Look for a matching association on the endpoint. */
1682 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1684 /* If we could not find a matching association on the
1685 * endpoint, make sure that it is not a TCP-style
1686 * socket that already has an association or there is
1687 * no peeled-off association on another socket.
1689 if ((sctp_style(sk, TCP) &&
1690 sctp_sstate(sk, ESTABLISHED)) ||
1691 sctp_endpoint_is_peeled_off(ep, &to)) {
1692 err = -EADDRNOTAVAIL;
1697 asoc = sctp_id2assoc(sk, associd);
1705 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1707 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1708 * socket that has an association in CLOSED state. This can
1709 * happen when an accepted socket has an association that is
1712 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1717 if (sinfo_flags & SCTP_EOF) {
1718 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1720 sctp_primitive_SHUTDOWN(asoc, NULL);
1724 if (sinfo_flags & SCTP_ABORT) {
1726 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1732 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1733 sctp_primitive_ABORT(asoc, chunk);
1739 /* Do we need to create the association? */
1741 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1743 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1748 /* Check for invalid stream against the stream counts,
1749 * either the default or the user specified stream counts.
1752 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1753 /* Check against the defaults. */
1754 if (sinfo->sinfo_stream >=
1755 sp->initmsg.sinit_num_ostreams) {
1760 /* Check against the requested. */
1761 if (sinfo->sinfo_stream >=
1762 sinit->sinit_num_ostreams) {
1770 * API 3.1.2 bind() - UDP Style Syntax
1771 * If a bind() or sctp_bindx() is not called prior to a
1772 * sendmsg() call that initiates a new association, the
1773 * system picks an ephemeral port and will choose an address
1774 * set equivalent to binding with a wildcard address.
1776 if (!ep->base.bind_addr.port) {
1777 if (sctp_autobind(sk)) {
1783 * If an unprivileged user inherits a one-to-many
1784 * style socket with open associations on a privileged
1785 * port, it MAY be permitted to accept new associations,
1786 * but it SHOULD NOT be permitted to open new
1789 if (ep->base.bind_addr.port < PROT_SOCK &&
1790 !capable(CAP_NET_BIND_SERVICE)) {
1796 scope = sctp_scope(&to);
1797 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1803 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1809 /* If the SCTP_INIT ancillary data is specified, set all
1810 * the association init values accordingly.
1813 if (sinit->sinit_num_ostreams) {
1814 asoc->c.sinit_num_ostreams =
1815 sinit->sinit_num_ostreams;
1817 if (sinit->sinit_max_instreams) {
1818 asoc->c.sinit_max_instreams =
1819 sinit->sinit_max_instreams;
1821 if (sinit->sinit_max_attempts) {
1822 asoc->max_init_attempts
1823 = sinit->sinit_max_attempts;
1825 if (sinit->sinit_max_init_timeo) {
1826 asoc->max_init_timeo =
1827 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1831 /* Prime the peer's transport structures. */
1832 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1839 /* ASSERT: we have a valid association at this point. */
1840 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1843 /* If the user didn't specify SNDRCVINFO, make up one with
1846 memset(&default_sinfo, 0, sizeof(default_sinfo));
1847 default_sinfo.sinfo_stream = asoc->default_stream;
1848 default_sinfo.sinfo_flags = asoc->default_flags;
1849 default_sinfo.sinfo_ppid = asoc->default_ppid;
1850 default_sinfo.sinfo_context = asoc->default_context;
1851 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1852 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1853 sinfo = &default_sinfo;
1856 /* API 7.1.7, the sndbuf size per association bounds the
1857 * maximum size of data that can be sent in a single send call.
1859 if (msg_len > sk->sk_sndbuf) {
1864 if (asoc->pmtu_pending)
1865 sctp_assoc_pending_pmtu(asoc);
1867 /* If fragmentation is disabled and the message length exceeds the
1868 * association fragmentation point, return EMSGSIZE. The I-D
1869 * does not specify what this error is, but this looks like
1872 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1877 /* Check for invalid stream. */
1878 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1883 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1884 if (!sctp_wspace(asoc)) {
1885 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1890 /* If an address is passed with the sendto/sendmsg call, it is used
1891 * to override the primary destination address in the TCP model, or
1892 * when SCTP_ADDR_OVER flag is set in the UDP model.
1894 if ((sctp_style(sk, TCP) && msg_name) ||
1895 (sinfo_flags & SCTP_ADDR_OVER)) {
1896 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1904 /* Auto-connect, if we aren't connected already. */
1905 if (sctp_state(asoc, CLOSED)) {
1906 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1909 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1912 /* Break the message into multiple chunks of maximum size. */
1913 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1914 if (IS_ERR(datamsg)) {
1915 err = PTR_ERR(datamsg);
1919 /* Now send the (possibly) fragmented message. */
1920 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1921 sctp_chunk_hold(chunk);
1923 /* Do accounting for the write space. */
1924 sctp_set_owner_w(chunk);
1926 chunk->transport = chunk_tp;
1929 /* Send it to the lower layers. Note: all chunks
1930 * must either fail or succeed. The lower layer
1931 * works that way today. Keep it that way or this
1934 err = sctp_primitive_SEND(asoc, datamsg);
1935 /* Did the lower layer accept the chunk? */
1937 sctp_datamsg_free(datamsg);
1939 sctp_datamsg_put(datamsg);
1941 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1948 /* If we are already past ASSOCIATE, the lower
1949 * layers are responsible for association cleanup.
1955 sctp_unhash_established(asoc);
1956 sctp_association_free(asoc);
1959 sctp_release_sock(sk);
1962 return sctp_error(sk, msg_flags, err);
1969 err = sock_error(sk);
1979 /* This is an extended version of skb_pull() that removes the data from the
1980 * start of a skb even when data is spread across the list of skb's in the
1981 * frag_list. len specifies the total amount of data that needs to be removed.
1982 * when 'len' bytes could be removed from the skb, it returns 0.
1983 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1984 * could not be removed.
1986 static int sctp_skb_pull(struct sk_buff *skb, int len)
1988 struct sk_buff *list;
1989 int skb_len = skb_headlen(skb);
1992 if (len <= skb_len) {
1993 __skb_pull(skb, len);
1997 __skb_pull(skb, skb_len);
1999 skb_walk_frags(skb, list) {
2000 rlen = sctp_skb_pull(list, len);
2001 skb->len -= (len-rlen);
2002 skb->data_len -= (len-rlen);
2013 /* API 3.1.3 recvmsg() - UDP Style Syntax
2015 * ssize_t recvmsg(int socket, struct msghdr *message,
2018 * socket - the socket descriptor of the endpoint.
2019 * message - pointer to the msghdr structure which contains a single
2020 * user message and possibly some ancillary data.
2022 * See Section 5 for complete description of the data
2025 * flags - flags sent or received with the user message, see Section
2026 * 5 for complete description of the flags.
2028 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2030 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2031 struct msghdr *msg, size_t len, int noblock,
2032 int flags, int *addr_len)
2034 struct sctp_ulpevent *event = NULL;
2035 struct sctp_sock *sp = sctp_sk(sk);
2036 struct sk_buff *skb;
2041 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
2042 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
2043 "len", len, "knoblauch", noblock,
2044 "flags", flags, "addr_len", addr_len);
2048 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2053 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2057 /* Get the total length of the skb including any skb's in the
2066 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2068 event = sctp_skb2event(skb);
2073 sock_recv_ts_and_drops(msg, sk, skb);
2074 if (sctp_ulpevent_is_notification(event)) {
2075 msg->msg_flags |= MSG_NOTIFICATION;
2076 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2078 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2081 /* Check if we allow SCTP_SNDRCVINFO. */
2082 if (sp->subscribe.sctp_data_io_event)
2083 sctp_ulpevent_read_sndrcvinfo(event, msg);
2085 /* FIXME: we should be calling IP/IPv6 layers. */
2086 if (sk->sk_protinfo.af_inet.cmsg_flags)
2087 ip_cmsg_recv(msg, skb);
2092 /* If skb's length exceeds the user's buffer, update the skb and
2093 * push it back to the receive_queue so that the next call to
2094 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2096 if (skb_len > copied) {
2097 msg->msg_flags &= ~MSG_EOR;
2098 if (flags & MSG_PEEK)
2100 sctp_skb_pull(skb, copied);
2101 skb_queue_head(&sk->sk_receive_queue, skb);
2103 /* When only partial message is copied to the user, increase
2104 * rwnd by that amount. If all the data in the skb is read,
2105 * rwnd is updated when the event is freed.
2107 if (!sctp_ulpevent_is_notification(event))
2108 sctp_assoc_rwnd_increase(event->asoc, copied);
2110 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2111 (event->msg_flags & MSG_EOR))
2112 msg->msg_flags |= MSG_EOR;
2114 msg->msg_flags &= ~MSG_EOR;
2117 if (flags & MSG_PEEK) {
2118 /* Release the skb reference acquired after peeking the skb in
2119 * sctp_skb_recv_datagram().
2123 /* Free the event which includes releasing the reference to
2124 * the owner of the skb, freeing the skb and updating the
2127 sctp_ulpevent_free(event);
2130 sctp_release_sock(sk);
2134 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2136 * This option is a on/off flag. If enabled no SCTP message
2137 * fragmentation will be performed. Instead if a message being sent
2138 * exceeds the current PMTU size, the message will NOT be sent and
2139 * instead a error will be indicated to the user.
2141 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2142 char __user *optval,
2143 unsigned int optlen)
2147 if (optlen < sizeof(int))
2150 if (get_user(val, (int __user *)optval))
2153 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2158 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2159 unsigned int optlen)
2161 struct sctp_association *asoc;
2162 struct sctp_ulpevent *event;
2164 if (optlen > sizeof(struct sctp_event_subscribe))
2166 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2170 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2171 * if there is no data to be sent or retransmit, the stack will
2172 * immediately send up this notification.
2174 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2175 &sctp_sk(sk)->subscribe)) {
2176 asoc = sctp_id2assoc(sk, 0);
2178 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2179 event = sctp_ulpevent_make_sender_dry_event(asoc,
2184 sctp_ulpq_tail_event(&asoc->ulpq, event);
2191 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2193 * This socket option is applicable to the UDP-style socket only. When
2194 * set it will cause associations that are idle for more than the
2195 * specified number of seconds to automatically close. An association
2196 * being idle is defined an association that has NOT sent or received
2197 * user data. The special value of '0' indicates that no automatic
2198 * close of any associations should be performed. The option expects an
2199 * integer defining the number of seconds of idle time before an
2200 * association is closed.
2202 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2203 unsigned int optlen)
2205 struct sctp_sock *sp = sctp_sk(sk);
2207 /* Applicable to UDP-style socket only */
2208 if (sctp_style(sk, TCP))
2210 if (optlen != sizeof(int))
2212 if (copy_from_user(&sp->autoclose, optval, optlen))
2218 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2220 * Applications can enable or disable heartbeats for any peer address of
2221 * an association, modify an address's heartbeat interval, force a
2222 * heartbeat to be sent immediately, and adjust the address's maximum
2223 * number of retransmissions sent before an address is considered
2224 * unreachable. The following structure is used to access and modify an
2225 * address's parameters:
2227 * struct sctp_paddrparams {
2228 * sctp_assoc_t spp_assoc_id;
2229 * struct sockaddr_storage spp_address;
2230 * uint32_t spp_hbinterval;
2231 * uint16_t spp_pathmaxrxt;
2232 * uint32_t spp_pathmtu;
2233 * uint32_t spp_sackdelay;
2234 * uint32_t spp_flags;
2237 * spp_assoc_id - (one-to-many style socket) This is filled in the
2238 * application, and identifies the association for
2240 * spp_address - This specifies which address is of interest.
2241 * spp_hbinterval - This contains the value of the heartbeat interval,
2242 * in milliseconds. If a value of zero
2243 * is present in this field then no changes are to
2244 * be made to this parameter.
2245 * spp_pathmaxrxt - This contains the maximum number of
2246 * retransmissions before this address shall be
2247 * considered unreachable. If a value of zero
2248 * is present in this field then no changes are to
2249 * be made to this parameter.
2250 * spp_pathmtu - When Path MTU discovery is disabled the value
2251 * specified here will be the "fixed" path mtu.
2252 * Note that if the spp_address field is empty
2253 * then all associations on this address will
2254 * have this fixed path mtu set upon them.
2256 * spp_sackdelay - When delayed sack is enabled, this value specifies
2257 * the number of milliseconds that sacks will be delayed
2258 * for. This value will apply to all addresses of an
2259 * association if the spp_address field is empty. Note
2260 * also, that if delayed sack is enabled and this
2261 * value is set to 0, no change is made to the last
2262 * recorded delayed sack timer value.
2264 * spp_flags - These flags are used to control various features
2265 * on an association. The flag field may contain
2266 * zero or more of the following options.
2268 * SPP_HB_ENABLE - Enable heartbeats on the
2269 * specified address. Note that if the address
2270 * field is empty all addresses for the association
2271 * have heartbeats enabled upon them.
2273 * SPP_HB_DISABLE - Disable heartbeats on the
2274 * speicifed address. Note that if the address
2275 * field is empty all addresses for the association
2276 * will have their heartbeats disabled. Note also
2277 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2278 * mutually exclusive, only one of these two should
2279 * be specified. Enabling both fields will have
2280 * undetermined results.
2282 * SPP_HB_DEMAND - Request a user initiated heartbeat
2283 * to be made immediately.
2285 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2286 * heartbeat delayis to be set to the value of 0
2289 * SPP_PMTUD_ENABLE - This field will enable PMTU
2290 * discovery upon the specified address. Note that
2291 * if the address feild is empty then all addresses
2292 * on the association are effected.
2294 * SPP_PMTUD_DISABLE - This field will disable PMTU
2295 * discovery upon the specified address. Note that
2296 * if the address feild is empty then all addresses
2297 * on the association are effected. Not also that
2298 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2299 * exclusive. Enabling both will have undetermined
2302 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2303 * on delayed sack. The time specified in spp_sackdelay
2304 * is used to specify the sack delay for this address. Note
2305 * that if spp_address is empty then all addresses will
2306 * enable delayed sack and take on the sack delay
2307 * value specified in spp_sackdelay.
2308 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2309 * off delayed sack. If the spp_address field is blank then
2310 * delayed sack is disabled for the entire association. Note
2311 * also that this field is mutually exclusive to
2312 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2315 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2316 struct sctp_transport *trans,
2317 struct sctp_association *asoc,
2318 struct sctp_sock *sp,
2321 int sackdelay_change)
2325 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2326 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2331 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2332 * this field is ignored. Note also that a value of zero indicates
2333 * the current setting should be left unchanged.
2335 if (params->spp_flags & SPP_HB_ENABLE) {
2337 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2338 * set. This lets us use 0 value when this flag
2341 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2342 params->spp_hbinterval = 0;
2344 if (params->spp_hbinterval ||
2345 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2348 msecs_to_jiffies(params->spp_hbinterval);
2351 msecs_to_jiffies(params->spp_hbinterval);
2353 sp->hbinterval = params->spp_hbinterval;
2360 trans->param_flags =
2361 (trans->param_flags & ~SPP_HB) | hb_change;
2364 (asoc->param_flags & ~SPP_HB) | hb_change;
2367 (sp->param_flags & ~SPP_HB) | hb_change;
2371 /* When Path MTU discovery is disabled the value specified here will
2372 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2373 * include the flag SPP_PMTUD_DISABLE for this field to have any
2376 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2378 trans->pathmtu = params->spp_pathmtu;
2379 sctp_assoc_sync_pmtu(asoc);
2381 asoc->pathmtu = params->spp_pathmtu;
2382 sctp_frag_point(asoc, params->spp_pathmtu);
2384 sp->pathmtu = params->spp_pathmtu;
2390 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2391 (params->spp_flags & SPP_PMTUD_ENABLE);
2392 trans->param_flags =
2393 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2395 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2396 sctp_assoc_sync_pmtu(asoc);
2400 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2403 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2407 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2408 * value of this field is ignored. Note also that a value of zero
2409 * indicates the current setting should be left unchanged.
2411 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2414 msecs_to_jiffies(params->spp_sackdelay);
2417 msecs_to_jiffies(params->spp_sackdelay);
2419 sp->sackdelay = params->spp_sackdelay;
2423 if (sackdelay_change) {
2425 trans->param_flags =
2426 (trans->param_flags & ~SPP_SACKDELAY) |
2430 (asoc->param_flags & ~SPP_SACKDELAY) |
2434 (sp->param_flags & ~SPP_SACKDELAY) |
2439 /* Note that a value of zero indicates the current setting should be
2442 if (params->spp_pathmaxrxt) {
2444 trans->pathmaxrxt = params->spp_pathmaxrxt;
2446 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2448 sp->pathmaxrxt = params->spp_pathmaxrxt;
2455 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2456 char __user *optval,
2457 unsigned int optlen)
2459 struct sctp_paddrparams params;
2460 struct sctp_transport *trans = NULL;
2461 struct sctp_association *asoc = NULL;
2462 struct sctp_sock *sp = sctp_sk(sk);
2464 int hb_change, pmtud_change, sackdelay_change;
2466 if (optlen != sizeof(struct sctp_paddrparams))
2469 if (copy_from_user(¶ms, optval, optlen))
2472 /* Validate flags and value parameters. */
2473 hb_change = params.spp_flags & SPP_HB;
2474 pmtud_change = params.spp_flags & SPP_PMTUD;
2475 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2477 if (hb_change == SPP_HB ||
2478 pmtud_change == SPP_PMTUD ||
2479 sackdelay_change == SPP_SACKDELAY ||
2480 params.spp_sackdelay > 500 ||
2481 (params.spp_pathmtu &&
2482 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2485 /* If an address other than INADDR_ANY is specified, and
2486 * no transport is found, then the request is invalid.
2488 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
2489 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2490 params.spp_assoc_id);
2495 /* Get association, if assoc_id != 0 and the socket is a one
2496 * to many style socket, and an association was not found, then
2497 * the id was invalid.
2499 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2500 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2503 /* Heartbeat demand can only be sent on a transport or
2504 * association, but not a socket.
2506 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2509 /* Process parameters. */
2510 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2511 hb_change, pmtud_change,
2517 /* If changes are for association, also apply parameters to each
2520 if (!trans && asoc) {
2521 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2523 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2524 hb_change, pmtud_change,
2533 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2535 * This option will effect the way delayed acks are performed. This
2536 * option allows you to get or set the delayed ack time, in
2537 * milliseconds. It also allows changing the delayed ack frequency.
2538 * Changing the frequency to 1 disables the delayed sack algorithm. If
2539 * the assoc_id is 0, then this sets or gets the endpoints default
2540 * values. If the assoc_id field is non-zero, then the set or get
2541 * effects the specified association for the one to many model (the
2542 * assoc_id field is ignored by the one to one model). Note that if
2543 * sack_delay or sack_freq are 0 when setting this option, then the
2544 * current values will remain unchanged.
2546 * struct sctp_sack_info {
2547 * sctp_assoc_t sack_assoc_id;
2548 * uint32_t sack_delay;
2549 * uint32_t sack_freq;
2552 * sack_assoc_id - This parameter, indicates which association the user
2553 * is performing an action upon. Note that if this field's value is
2554 * zero then the endpoints default value is changed (effecting future
2555 * associations only).
2557 * sack_delay - This parameter contains the number of milliseconds that
2558 * the user is requesting the delayed ACK timer be set to. Note that
2559 * this value is defined in the standard to be between 200 and 500
2562 * sack_freq - This parameter contains the number of packets that must
2563 * be received before a sack is sent without waiting for the delay
2564 * timer to expire. The default value for this is 2, setting this
2565 * value to 1 will disable the delayed sack algorithm.
2568 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2569 char __user *optval, unsigned int optlen)
2571 struct sctp_sack_info params;
2572 struct sctp_transport *trans = NULL;
2573 struct sctp_association *asoc = NULL;
2574 struct sctp_sock *sp = sctp_sk(sk);
2576 if (optlen == sizeof(struct sctp_sack_info)) {
2577 if (copy_from_user(¶ms, optval, optlen))
2580 if (params.sack_delay == 0 && params.sack_freq == 0)
2582 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2583 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2584 pr_warn("Use struct sctp_sack_info instead\n");
2585 if (copy_from_user(¶ms, optval, optlen))
2588 if (params.sack_delay == 0)
2589 params.sack_freq = 1;
2591 params.sack_freq = 0;
2595 /* Validate value parameter. */
2596 if (params.sack_delay > 500)
2599 /* Get association, if sack_assoc_id != 0 and the socket is a one
2600 * to many style socket, and an association was not found, then
2601 * the id was invalid.
2603 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2604 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2607 if (params.sack_delay) {
2610 msecs_to_jiffies(params.sack_delay);
2612 (asoc->param_flags & ~SPP_SACKDELAY) |
2613 SPP_SACKDELAY_ENABLE;
2615 sp->sackdelay = params.sack_delay;
2617 (sp->param_flags & ~SPP_SACKDELAY) |
2618 SPP_SACKDELAY_ENABLE;
2622 if (params.sack_freq == 1) {
2625 (asoc->param_flags & ~SPP_SACKDELAY) |
2626 SPP_SACKDELAY_DISABLE;
2629 (sp->param_flags & ~SPP_SACKDELAY) |
2630 SPP_SACKDELAY_DISABLE;
2632 } else if (params.sack_freq > 1) {
2634 asoc->sackfreq = params.sack_freq;
2636 (asoc->param_flags & ~SPP_SACKDELAY) |
2637 SPP_SACKDELAY_ENABLE;
2639 sp->sackfreq = params.sack_freq;
2641 (sp->param_flags & ~SPP_SACKDELAY) |
2642 SPP_SACKDELAY_ENABLE;
2646 /* If change is for association, also apply to each transport. */
2648 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2650 if (params.sack_delay) {
2652 msecs_to_jiffies(params.sack_delay);
2653 trans->param_flags =
2654 (trans->param_flags & ~SPP_SACKDELAY) |
2655 SPP_SACKDELAY_ENABLE;
2657 if (params.sack_freq == 1) {
2658 trans->param_flags =
2659 (trans->param_flags & ~SPP_SACKDELAY) |
2660 SPP_SACKDELAY_DISABLE;
2661 } else if (params.sack_freq > 1) {
2662 trans->sackfreq = params.sack_freq;
2663 trans->param_flags =
2664 (trans->param_flags & ~SPP_SACKDELAY) |
2665 SPP_SACKDELAY_ENABLE;
2673 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2675 * Applications can specify protocol parameters for the default association
2676 * initialization. The option name argument to setsockopt() and getsockopt()
2679 * Setting initialization parameters is effective only on an unconnected
2680 * socket (for UDP-style sockets only future associations are effected
2681 * by the change). With TCP-style sockets, this option is inherited by
2682 * sockets derived from a listener socket.
2684 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2686 struct sctp_initmsg sinit;
2687 struct sctp_sock *sp = sctp_sk(sk);
2689 if (optlen != sizeof(struct sctp_initmsg))
2691 if (copy_from_user(&sinit, optval, optlen))
2694 if (sinit.sinit_num_ostreams)
2695 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2696 if (sinit.sinit_max_instreams)
2697 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2698 if (sinit.sinit_max_attempts)
2699 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2700 if (sinit.sinit_max_init_timeo)
2701 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2707 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2709 * Applications that wish to use the sendto() system call may wish to
2710 * specify a default set of parameters that would normally be supplied
2711 * through the inclusion of ancillary data. This socket option allows
2712 * such an application to set the default sctp_sndrcvinfo structure.
2713 * The application that wishes to use this socket option simply passes
2714 * in to this call the sctp_sndrcvinfo structure defined in Section
2715 * 5.2.2) The input parameters accepted by this call include
2716 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2717 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2718 * to this call if the caller is using the UDP model.
2720 static int sctp_setsockopt_default_send_param(struct sock *sk,
2721 char __user *optval,
2722 unsigned int optlen)
2724 struct sctp_sndrcvinfo info;
2725 struct sctp_association *asoc;
2726 struct sctp_sock *sp = sctp_sk(sk);
2728 if (optlen != sizeof(struct sctp_sndrcvinfo))
2730 if (copy_from_user(&info, optval, optlen))
2733 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2734 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2738 asoc->default_stream = info.sinfo_stream;
2739 asoc->default_flags = info.sinfo_flags;
2740 asoc->default_ppid = info.sinfo_ppid;
2741 asoc->default_context = info.sinfo_context;
2742 asoc->default_timetolive = info.sinfo_timetolive;
2744 sp->default_stream = info.sinfo_stream;
2745 sp->default_flags = info.sinfo_flags;
2746 sp->default_ppid = info.sinfo_ppid;
2747 sp->default_context = info.sinfo_context;
2748 sp->default_timetolive = info.sinfo_timetolive;
2754 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2756 * Requests that the local SCTP stack use the enclosed peer address as
2757 * the association primary. The enclosed address must be one of the
2758 * association peer's addresses.
2760 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2761 unsigned int optlen)
2763 struct sctp_prim prim;
2764 struct sctp_transport *trans;
2766 if (optlen != sizeof(struct sctp_prim))
2769 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2772 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2776 sctp_assoc_set_primary(trans->asoc, trans);
2782 * 7.1.5 SCTP_NODELAY
2784 * Turn on/off any Nagle-like algorithm. This means that packets are
2785 * generally sent as soon as possible and no unnecessary delays are
2786 * introduced, at the cost of more packets in the network. Expects an
2787 * integer boolean flag.
2789 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2790 unsigned int optlen)
2794 if (optlen < sizeof(int))
2796 if (get_user(val, (int __user *)optval))
2799 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2805 * 7.1.1 SCTP_RTOINFO
2807 * The protocol parameters used to initialize and bound retransmission
2808 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2809 * and modify these parameters.
2810 * All parameters are time values, in milliseconds. A value of 0, when
2811 * modifying the parameters, indicates that the current value should not
2815 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2817 struct sctp_rtoinfo rtoinfo;
2818 struct sctp_association *asoc;
2820 if (optlen != sizeof (struct sctp_rtoinfo))
2823 if (copy_from_user(&rtoinfo, optval, optlen))
2826 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2828 /* Set the values to the specific association */
2829 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2833 if (rtoinfo.srto_initial != 0)
2835 msecs_to_jiffies(rtoinfo.srto_initial);
2836 if (rtoinfo.srto_max != 0)
2837 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2838 if (rtoinfo.srto_min != 0)
2839 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2841 /* If there is no association or the association-id = 0
2842 * set the values to the endpoint.
2844 struct sctp_sock *sp = sctp_sk(sk);
2846 if (rtoinfo.srto_initial != 0)
2847 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2848 if (rtoinfo.srto_max != 0)
2849 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2850 if (rtoinfo.srto_min != 0)
2851 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2859 * 7.1.2 SCTP_ASSOCINFO
2861 * This option is used to tune the maximum retransmission attempts
2862 * of the association.
2863 * Returns an error if the new association retransmission value is
2864 * greater than the sum of the retransmission value of the peer.
2865 * See [SCTP] for more information.
2868 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2871 struct sctp_assocparams assocparams;
2872 struct sctp_association *asoc;
2874 if (optlen != sizeof(struct sctp_assocparams))
2876 if (copy_from_user(&assocparams, optval, optlen))
2879 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2881 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2884 /* Set the values to the specific association */
2886 if (assocparams.sasoc_asocmaxrxt != 0) {
2889 struct sctp_transport *peer_addr;
2891 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2893 path_sum += peer_addr->pathmaxrxt;
2897 /* Only validate asocmaxrxt if we have more than
2898 * one path/transport. We do this because path
2899 * retransmissions are only counted when we have more
2903 assocparams.sasoc_asocmaxrxt > path_sum)
2906 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2909 if (assocparams.sasoc_cookie_life != 0) {
2910 asoc->cookie_life.tv_sec =
2911 assocparams.sasoc_cookie_life / 1000;
2912 asoc->cookie_life.tv_usec =
2913 (assocparams.sasoc_cookie_life % 1000)
2917 /* Set the values to the endpoint */
2918 struct sctp_sock *sp = sctp_sk(sk);
2920 if (assocparams.sasoc_asocmaxrxt != 0)
2921 sp->assocparams.sasoc_asocmaxrxt =
2922 assocparams.sasoc_asocmaxrxt;
2923 if (assocparams.sasoc_cookie_life != 0)
2924 sp->assocparams.sasoc_cookie_life =
2925 assocparams.sasoc_cookie_life;
2931 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2933 * This socket option is a boolean flag which turns on or off mapped V4
2934 * addresses. If this option is turned on and the socket is type
2935 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2936 * If this option is turned off, then no mapping will be done of V4
2937 * addresses and a user will receive both PF_INET6 and PF_INET type
2938 * addresses on the socket.
2940 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2943 struct sctp_sock *sp = sctp_sk(sk);
2945 if (optlen < sizeof(int))
2947 if (get_user(val, (int __user *)optval))
2958 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2959 * This option will get or set the maximum size to put in any outgoing
2960 * SCTP DATA chunk. If a message is larger than this size it will be
2961 * fragmented by SCTP into the specified size. Note that the underlying
2962 * SCTP implementation may fragment into smaller sized chunks when the
2963 * PMTU of the underlying association is smaller than the value set by
2964 * the user. The default value for this option is '0' which indicates
2965 * the user is NOT limiting fragmentation and only the PMTU will effect
2966 * SCTP's choice of DATA chunk size. Note also that values set larger
2967 * than the maximum size of an IP datagram will effectively let SCTP
2968 * control fragmentation (i.e. the same as setting this option to 0).
2970 * The following structure is used to access and modify this parameter:
2972 * struct sctp_assoc_value {
2973 * sctp_assoc_t assoc_id;
2974 * uint32_t assoc_value;
2977 * assoc_id: This parameter is ignored for one-to-one style sockets.
2978 * For one-to-many style sockets this parameter indicates which
2979 * association the user is performing an action upon. Note that if
2980 * this field's value is zero then the endpoints default value is
2981 * changed (effecting future associations only).
2982 * assoc_value: This parameter specifies the maximum size in bytes.
2984 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2986 struct sctp_assoc_value params;
2987 struct sctp_association *asoc;
2988 struct sctp_sock *sp = sctp_sk(sk);
2991 if (optlen == sizeof(int)) {
2992 pr_warn("Use of int in maxseg socket option deprecated\n");
2993 pr_warn("Use struct sctp_assoc_value instead\n");
2994 if (copy_from_user(&val, optval, optlen))
2996 params.assoc_id = 0;
2997 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2998 if (copy_from_user(¶ms, optval, optlen))
3000 val = params.assoc_value;
3004 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3007 asoc = sctp_id2assoc(sk, params.assoc_id);
3008 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3013 val = asoc->pathmtu;
3014 val -= sp->pf->af->net_header_len;
3015 val -= sizeof(struct sctphdr) +
3016 sizeof(struct sctp_data_chunk);
3018 asoc->user_frag = val;
3019 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3021 sp->user_frag = val;
3029 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3031 * Requests that the peer mark the enclosed address as the association
3032 * primary. The enclosed address must be one of the association's
3033 * locally bound addresses. The following structure is used to make a
3034 * set primary request:
3036 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3037 unsigned int optlen)
3039 struct sctp_sock *sp;
3040 struct sctp_association *asoc = NULL;
3041 struct sctp_setpeerprim prim;
3042 struct sctp_chunk *chunk;
3048 if (!sctp_addip_enable)
3051 if (optlen != sizeof(struct sctp_setpeerprim))
3054 if (copy_from_user(&prim, optval, optlen))
3057 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3061 if (!asoc->peer.asconf_capable)
3064 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3067 if (!sctp_state(asoc, ESTABLISHED))
3070 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3074 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3075 return -EADDRNOTAVAIL;
3077 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3078 return -EADDRNOTAVAIL;
3080 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3081 chunk = sctp_make_asconf_set_prim(asoc,
3082 (union sctp_addr *)&prim.sspp_addr);
3086 err = sctp_send_asconf(asoc, chunk);
3088 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
3093 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3094 unsigned int optlen)
3096 struct sctp_setadaptation adaptation;
3098 if (optlen != sizeof(struct sctp_setadaptation))
3100 if (copy_from_user(&adaptation, optval, optlen))
3103 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3109 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3111 * The context field in the sctp_sndrcvinfo structure is normally only
3112 * used when a failed message is retrieved holding the value that was
3113 * sent down on the actual send call. This option allows the setting of
3114 * a default context on an association basis that will be received on
3115 * reading messages from the peer. This is especially helpful in the
3116 * one-2-many model for an application to keep some reference to an
3117 * internal state machine that is processing messages on the
3118 * association. Note that the setting of this value only effects
3119 * received messages from the peer and does not effect the value that is
3120 * saved with outbound messages.
3122 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3123 unsigned int optlen)
3125 struct sctp_assoc_value params;
3126 struct sctp_sock *sp;
3127 struct sctp_association *asoc;
3129 if (optlen != sizeof(struct sctp_assoc_value))
3131 if (copy_from_user(¶ms, optval, optlen))
3136 if (params.assoc_id != 0) {
3137 asoc = sctp_id2assoc(sk, params.assoc_id);
3140 asoc->default_rcv_context = params.assoc_value;
3142 sp->default_rcv_context = params.assoc_value;
3149 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3151 * This options will at a minimum specify if the implementation is doing
3152 * fragmented interleave. Fragmented interleave, for a one to many
3153 * socket, is when subsequent calls to receive a message may return
3154 * parts of messages from different associations. Some implementations
3155 * may allow you to turn this value on or off. If so, when turned off,
3156 * no fragment interleave will occur (which will cause a head of line
3157 * blocking amongst multiple associations sharing the same one to many
3158 * socket). When this option is turned on, then each receive call may
3159 * come from a different association (thus the user must receive data
3160 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3161 * association each receive belongs to.
3163 * This option takes a boolean value. A non-zero value indicates that
3164 * fragmented interleave is on. A value of zero indicates that
3165 * fragmented interleave is off.
3167 * Note that it is important that an implementation that allows this
3168 * option to be turned on, have it off by default. Otherwise an unaware
3169 * application using the one to many model may become confused and act
3172 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3173 char __user *optval,
3174 unsigned int optlen)
3178 if (optlen != sizeof(int))
3180 if (get_user(val, (int __user *)optval))
3183 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3189 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3190 * (SCTP_PARTIAL_DELIVERY_POINT)
3192 * This option will set or get the SCTP partial delivery point. This
3193 * point is the size of a message where the partial delivery API will be
3194 * invoked to help free up rwnd space for the peer. Setting this to a
3195 * lower value will cause partial deliveries to happen more often. The
3196 * calls argument is an integer that sets or gets the partial delivery
3197 * point. Note also that the call will fail if the user attempts to set
3198 * this value larger than the socket receive buffer size.
3200 * Note that any single message having a length smaller than or equal to
3201 * the SCTP partial delivery point will be delivered in one single read
3202 * call as long as the user provided buffer is large enough to hold the
3205 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3206 char __user *optval,
3207 unsigned int optlen)
3211 if (optlen != sizeof(u32))
3213 if (get_user(val, (int __user *)optval))
3216 /* Note: We double the receive buffer from what the user sets
3217 * it to be, also initial rwnd is based on rcvbuf/2.
3219 if (val > (sk->sk_rcvbuf >> 1))
3222 sctp_sk(sk)->pd_point = val;
3224 return 0; /* is this the right error code? */
3228 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3230 * This option will allow a user to change the maximum burst of packets
3231 * that can be emitted by this association. Note that the default value
3232 * is 4, and some implementations may restrict this setting so that it
3233 * can only be lowered.
3235 * NOTE: This text doesn't seem right. Do this on a socket basis with
3236 * future associations inheriting the socket value.
3238 static int sctp_setsockopt_maxburst(struct sock *sk,
3239 char __user *optval,
3240 unsigned int optlen)
3242 struct sctp_assoc_value params;
3243 struct sctp_sock *sp;
3244 struct sctp_association *asoc;
3248 if (optlen == sizeof(int)) {
3249 pr_warn("Use of int in max_burst socket option deprecated\n");
3250 pr_warn("Use struct sctp_assoc_value instead\n");
3251 if (copy_from_user(&val, optval, optlen))
3253 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3254 if (copy_from_user(¶ms, optval, optlen))
3256 val = params.assoc_value;
3257 assoc_id = params.assoc_id;
3263 if (assoc_id != 0) {
3264 asoc = sctp_id2assoc(sk, assoc_id);
3267 asoc->max_burst = val;
3269 sp->max_burst = val;
3275 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3277 * This set option adds a chunk type that the user is requesting to be
3278 * received only in an authenticated way. Changes to the list of chunks
3279 * will only effect future associations on the socket.
3281 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3282 char __user *optval,
3283 unsigned int optlen)
3285 struct sctp_authchunk val;
3287 if (!sctp_auth_enable)
3290 if (optlen != sizeof(struct sctp_authchunk))
3292 if (copy_from_user(&val, optval, optlen))
3295 switch (val.sauth_chunk) {
3297 case SCTP_CID_INIT_ACK:
3298 case SCTP_CID_SHUTDOWN_COMPLETE:
3303 /* add this chunk id to the endpoint */
3304 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3308 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3310 * This option gets or sets the list of HMAC algorithms that the local
3311 * endpoint requires the peer to use.
3313 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3314 char __user *optval,
3315 unsigned int optlen)
3317 struct sctp_hmacalgo *hmacs;
3321 if (!sctp_auth_enable)
3324 if (optlen < sizeof(struct sctp_hmacalgo))
3327 hmacs= memdup_user(optval, optlen);
3329 return PTR_ERR(hmacs);
3331 idents = hmacs->shmac_num_idents;
3332 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3333 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3338 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3345 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3347 * This option will set a shared secret key which is used to build an
3348 * association shared key.
3350 static int sctp_setsockopt_auth_key(struct sock *sk,
3351 char __user *optval,
3352 unsigned int optlen)
3354 struct sctp_authkey *authkey;
3355 struct sctp_association *asoc;
3358 if (!sctp_auth_enable)
3361 if (optlen <= sizeof(struct sctp_authkey))
3364 authkey= memdup_user(optval, optlen);
3365 if (IS_ERR(authkey))
3366 return PTR_ERR(authkey);
3368 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3373 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3374 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3379 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3386 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3388 * This option will get or set the active shared key to be used to build
3389 * the association shared key.
3391 static int sctp_setsockopt_active_key(struct sock *sk,
3392 char __user *optval,
3393 unsigned int optlen)
3395 struct sctp_authkeyid val;
3396 struct sctp_association *asoc;
3398 if (!sctp_auth_enable)
3401 if (optlen != sizeof(struct sctp_authkeyid))
3403 if (copy_from_user(&val, optval, optlen))
3406 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3407 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3410 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3411 val.scact_keynumber);
3415 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3417 * This set option will delete a shared secret key from use.
3419 static int sctp_setsockopt_del_key(struct sock *sk,
3420 char __user *optval,
3421 unsigned int optlen)
3423 struct sctp_authkeyid val;
3424 struct sctp_association *asoc;
3426 if (!sctp_auth_enable)
3429 if (optlen != sizeof(struct sctp_authkeyid))
3431 if (copy_from_user(&val, optval, optlen))
3434 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3435 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3438 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3439 val.scact_keynumber);
3444 * 8.1.23 SCTP_AUTO_ASCONF
3446 * This option will enable or disable the use of the automatic generation of
3447 * ASCONF chunks to add and delete addresses to an existing association. Note
3448 * that this option has two caveats namely: a) it only affects sockets that
3449 * are bound to all addresses available to the SCTP stack, and b) the system
3450 * administrator may have an overriding control that turns the ASCONF feature
3451 * off no matter what setting the socket option may have.
3452 * This option expects an integer boolean flag, where a non-zero value turns on
3453 * the option, and a zero value turns off the option.
3454 * Note. In this implementation, socket operation overrides default parameter
3455 * being set by sysctl as well as FreeBSD implementation
3457 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3458 unsigned int optlen)
3461 struct sctp_sock *sp = sctp_sk(sk);
3463 if (optlen < sizeof(int))
3465 if (get_user(val, (int __user *)optval))
3467 if (!sctp_is_ep_boundall(sk) && val)
3469 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3472 if (val == 0 && sp->do_auto_asconf) {
3473 list_del(&sp->auto_asconf_list);
3474 sp->do_auto_asconf = 0;
3475 } else if (val && !sp->do_auto_asconf) {
3476 list_add_tail(&sp->auto_asconf_list,
3477 &sctp_auto_asconf_splist);
3478 sp->do_auto_asconf = 1;
3484 /* API 6.2 setsockopt(), getsockopt()
3486 * Applications use setsockopt() and getsockopt() to set or retrieve
3487 * socket options. Socket options are used to change the default
3488 * behavior of sockets calls. They are described in Section 7.
3492 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3493 * int __user *optlen);
3494 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3497 * sd - the socket descript.
3498 * level - set to IPPROTO_SCTP for all SCTP options.
3499 * optname - the option name.
3500 * optval - the buffer to store the value of the option.
3501 * optlen - the size of the buffer.
3503 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3504 char __user *optval, unsigned int optlen)
3508 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3511 /* I can hardly begin to describe how wrong this is. This is
3512 * so broken as to be worse than useless. The API draft
3513 * REALLY is NOT helpful here... I am not convinced that the
3514 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3515 * are at all well-founded.
3517 if (level != SOL_SCTP) {
3518 struct sctp_af *af = sctp_sk(sk)->pf->af;
3519 retval = af->setsockopt(sk, level, optname, optval, optlen);
3526 case SCTP_SOCKOPT_BINDX_ADD:
3527 /* 'optlen' is the size of the addresses buffer. */
3528 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3529 optlen, SCTP_BINDX_ADD_ADDR);
3532 case SCTP_SOCKOPT_BINDX_REM:
3533 /* 'optlen' is the size of the addresses buffer. */
3534 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3535 optlen, SCTP_BINDX_REM_ADDR);
3538 case SCTP_SOCKOPT_CONNECTX_OLD:
3539 /* 'optlen' is the size of the addresses buffer. */
3540 retval = sctp_setsockopt_connectx_old(sk,
3541 (struct sockaddr __user *)optval,
3545 case SCTP_SOCKOPT_CONNECTX:
3546 /* 'optlen' is the size of the addresses buffer. */
3547 retval = sctp_setsockopt_connectx(sk,
3548 (struct sockaddr __user *)optval,
3552 case SCTP_DISABLE_FRAGMENTS:
3553 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3557 retval = sctp_setsockopt_events(sk, optval, optlen);
3560 case SCTP_AUTOCLOSE:
3561 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3564 case SCTP_PEER_ADDR_PARAMS:
3565 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3568 case SCTP_DELAYED_SACK:
3569 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3571 case SCTP_PARTIAL_DELIVERY_POINT:
3572 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3576 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3578 case SCTP_DEFAULT_SEND_PARAM:
3579 retval = sctp_setsockopt_default_send_param(sk, optval,
3582 case SCTP_PRIMARY_ADDR:
3583 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3585 case SCTP_SET_PEER_PRIMARY_ADDR:
3586 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3589 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3592 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3594 case SCTP_ASSOCINFO:
3595 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3597 case SCTP_I_WANT_MAPPED_V4_ADDR:
3598 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3601 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3603 case SCTP_ADAPTATION_LAYER:
3604 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3607 retval = sctp_setsockopt_context(sk, optval, optlen);
3609 case SCTP_FRAGMENT_INTERLEAVE:
3610 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3612 case SCTP_MAX_BURST:
3613 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3615 case SCTP_AUTH_CHUNK:
3616 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3618 case SCTP_HMAC_IDENT:
3619 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3622 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3624 case SCTP_AUTH_ACTIVE_KEY:
3625 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3627 case SCTP_AUTH_DELETE_KEY:
3628 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3630 case SCTP_AUTO_ASCONF:
3631 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3634 retval = -ENOPROTOOPT;
3638 sctp_release_sock(sk);
3644 /* API 3.1.6 connect() - UDP Style Syntax
3646 * An application may use the connect() call in the UDP model to initiate an
3647 * association without sending data.
3651 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3653 * sd: the socket descriptor to have a new association added to.
3655 * nam: the address structure (either struct sockaddr_in or struct
3656 * sockaddr_in6 defined in RFC2553 [7]).
3658 * len: the size of the address.
3660 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3668 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3669 __func__, sk, addr, addr_len);
3671 /* Validate addr_len before calling common connect/connectx routine. */
3672 af = sctp_get_af_specific(addr->sa_family);
3673 if (!af || addr_len < af->sockaddr_len) {
3676 /* Pass correct addr len to common routine (so it knows there
3677 * is only one address being passed.
3679 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3682 sctp_release_sock(sk);
3686 /* FIXME: Write comments. */
3687 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3689 return -EOPNOTSUPP; /* STUB */
3692 /* 4.1.4 accept() - TCP Style Syntax
3694 * Applications use accept() call to remove an established SCTP
3695 * association from the accept queue of the endpoint. A new socket
3696 * descriptor will be returned from accept() to represent the newly
3697 * formed association.
3699 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3701 struct sctp_sock *sp;
3702 struct sctp_endpoint *ep;
3703 struct sock *newsk = NULL;
3704 struct sctp_association *asoc;
3713 if (!sctp_style(sk, TCP)) {
3714 error = -EOPNOTSUPP;
3718 if (!sctp_sstate(sk, LISTENING)) {
3723 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3725 error = sctp_wait_for_accept(sk, timeo);
3729 /* We treat the list of associations on the endpoint as the accept
3730 * queue and pick the first association on the list.
3732 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3734 newsk = sp->pf->create_accept_sk(sk, asoc);
3740 /* Populate the fields of the newsk from the oldsk and migrate the
3741 * asoc to the newsk.
3743 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3746 sctp_release_sock(sk);
3751 /* The SCTP ioctl handler. */
3752 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3759 * SEQPACKET-style sockets in LISTENING state are valid, for
3760 * SCTP, so only discard TCP-style sockets in LISTENING state.
3762 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3767 struct sk_buff *skb;
3768 unsigned int amount = 0;
3770 skb = skb_peek(&sk->sk_receive_queue);
3773 * We will only return the amount of this packet since
3774 * that is all that will be read.
3778 rc = put_user(amount, (int __user *)arg);
3786 sctp_release_sock(sk);
3790 /* This is the function which gets called during socket creation to
3791 * initialized the SCTP-specific portion of the sock.
3792 * The sock structure should already be zero-filled memory.
3794 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3796 struct sctp_endpoint *ep;
3797 struct sctp_sock *sp;
3799 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3803 /* Initialize the SCTP per socket area. */
3804 switch (sk->sk_type) {
3805 case SOCK_SEQPACKET:
3806 sp->type = SCTP_SOCKET_UDP;
3809 sp->type = SCTP_SOCKET_TCP;
3812 return -ESOCKTNOSUPPORT;
3815 /* Initialize default send parameters. These parameters can be
3816 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3818 sp->default_stream = 0;
3819 sp->default_ppid = 0;
3820 sp->default_flags = 0;
3821 sp->default_context = 0;
3822 sp->default_timetolive = 0;
3824 sp->default_rcv_context = 0;
3825 sp->max_burst = sctp_max_burst;
3827 /* Initialize default setup parameters. These parameters
3828 * can be modified with the SCTP_INITMSG socket option or
3829 * overridden by the SCTP_INIT CMSG.
3831 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3832 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3833 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3834 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3836 /* Initialize default RTO related parameters. These parameters can
3837 * be modified for with the SCTP_RTOINFO socket option.
3839 sp->rtoinfo.srto_initial = sctp_rto_initial;
3840 sp->rtoinfo.srto_max = sctp_rto_max;
3841 sp->rtoinfo.srto_min = sctp_rto_min;
3843 /* Initialize default association related parameters. These parameters
3844 * can be modified with the SCTP_ASSOCINFO socket option.
3846 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3847 sp->assocparams.sasoc_number_peer_destinations = 0;
3848 sp->assocparams.sasoc_peer_rwnd = 0;
3849 sp->assocparams.sasoc_local_rwnd = 0;
3850 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3852 /* Initialize default event subscriptions. By default, all the
3855 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3857 /* Default Peer Address Parameters. These defaults can
3858 * be modified via SCTP_PEER_ADDR_PARAMS
3860 sp->hbinterval = sctp_hb_interval;
3861 sp->pathmaxrxt = sctp_max_retrans_path;
3862 sp->pathmtu = 0; // allow default discovery
3863 sp->sackdelay = sctp_sack_timeout;
3865 sp->param_flags = SPP_HB_ENABLE |
3867 SPP_SACKDELAY_ENABLE;
3869 /* If enabled no SCTP message fragmentation will be performed.
3870 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3872 sp->disable_fragments = 0;
3874 /* Enable Nagle algorithm by default. */
3877 /* Enable by default. */
3880 /* Auto-close idle associations after the configured
3881 * number of seconds. A value of 0 disables this
3882 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3883 * for UDP-style sockets only.
3887 /* User specified fragmentation limit. */
3890 sp->adaptation_ind = 0;
3892 sp->pf = sctp_get_pf_specific(sk->sk_family);
3894 /* Control variables for partial data delivery. */
3895 atomic_set(&sp->pd_mode, 0);
3896 skb_queue_head_init(&sp->pd_lobby);
3897 sp->frag_interleave = 0;
3899 /* Create a per socket endpoint structure. Even if we
3900 * change the data structure relationships, this may still
3901 * be useful for storing pre-connect address information.
3903 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3910 SCTP_DBG_OBJCNT_INC(sock);
3913 percpu_counter_inc(&sctp_sockets_allocated);
3914 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3915 if (sctp_default_auto_asconf) {
3916 list_add_tail(&sp->auto_asconf_list,
3917 &sctp_auto_asconf_splist);
3918 sp->do_auto_asconf = 1;
3920 sp->do_auto_asconf = 0;
3926 /* Cleanup any SCTP per socket resources. */
3927 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3929 struct sctp_sock *sp;
3931 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3933 /* Release our hold on the endpoint. */
3935 /* This could happen during socket init, thus we bail out
3936 * early, since the rest of the below is not setup either.
3941 if (sp->do_auto_asconf) {
3942 sp->do_auto_asconf = 0;
3943 list_del(&sp->auto_asconf_list);
3945 sctp_endpoint_free(sp->ep);
3947 percpu_counter_dec(&sctp_sockets_allocated);
3948 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
3952 /* API 4.1.7 shutdown() - TCP Style Syntax
3953 * int shutdown(int socket, int how);
3955 * sd - the socket descriptor of the association to be closed.
3956 * how - Specifies the type of shutdown. The values are
3959 * Disables further receive operations. No SCTP
3960 * protocol action is taken.
3962 * Disables further send operations, and initiates
3963 * the SCTP shutdown sequence.
3965 * Disables further send and receive operations
3966 * and initiates the SCTP shutdown sequence.
3968 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3970 struct sctp_endpoint *ep;
3971 struct sctp_association *asoc;
3973 if (!sctp_style(sk, TCP))
3976 if (how & SEND_SHUTDOWN) {
3977 ep = sctp_sk(sk)->ep;
3978 if (!list_empty(&ep->asocs)) {
3979 asoc = list_entry(ep->asocs.next,
3980 struct sctp_association, asocs);
3981 sctp_primitive_SHUTDOWN(asoc, NULL);
3986 /* 7.2.1 Association Status (SCTP_STATUS)
3988 * Applications can retrieve current status information about an
3989 * association, including association state, peer receiver window size,
3990 * number of unacked data chunks, and number of data chunks pending
3991 * receipt. This information is read-only.
3993 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3994 char __user *optval,
3997 struct sctp_status status;
3998 struct sctp_association *asoc = NULL;
3999 struct sctp_transport *transport;
4000 sctp_assoc_t associd;
4003 if (len < sizeof(status)) {
4008 len = sizeof(status);
4009 if (copy_from_user(&status, optval, len)) {
4014 associd = status.sstat_assoc_id;
4015 asoc = sctp_id2assoc(sk, associd);
4021 transport = asoc->peer.primary_path;
4023 status.sstat_assoc_id = sctp_assoc2id(asoc);
4024 status.sstat_state = asoc->state;
4025 status.sstat_rwnd = asoc->peer.rwnd;
4026 status.sstat_unackdata = asoc->unack_data;
4028 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4029 status.sstat_instrms = asoc->c.sinit_max_instreams;
4030 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4031 status.sstat_fragmentation_point = asoc->frag_point;
4032 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4033 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4034 transport->af_specific->sockaddr_len);
4035 /* Map ipv4 address into v4-mapped-on-v6 address. */
4036 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4037 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4038 status.sstat_primary.spinfo_state = transport->state;
4039 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4040 status.sstat_primary.spinfo_srtt = transport->srtt;
4041 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4042 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4044 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4045 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4047 if (put_user(len, optlen)) {
4052 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
4053 len, status.sstat_state, status.sstat_rwnd,
4054 status.sstat_assoc_id);
4056 if (copy_to_user(optval, &status, len)) {
4066 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4068 * Applications can retrieve information about a specific peer address
4069 * of an association, including its reachability state, congestion
4070 * window, and retransmission timer values. This information is
4073 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4074 char __user *optval,
4077 struct sctp_paddrinfo pinfo;
4078 struct sctp_transport *transport;
4081 if (len < sizeof(pinfo)) {
4086 len = sizeof(pinfo);
4087 if (copy_from_user(&pinfo, optval, len)) {
4092 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4093 pinfo.spinfo_assoc_id);
4097 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4098 pinfo.spinfo_state = transport->state;
4099 pinfo.spinfo_cwnd = transport->cwnd;
4100 pinfo.spinfo_srtt = transport->srtt;
4101 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4102 pinfo.spinfo_mtu = transport->pathmtu;
4104 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4105 pinfo.spinfo_state = SCTP_ACTIVE;
4107 if (put_user(len, optlen)) {
4112 if (copy_to_user(optval, &pinfo, len)) {
4121 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4123 * This option is a on/off flag. If enabled no SCTP message
4124 * fragmentation will be performed. Instead if a message being sent
4125 * exceeds the current PMTU size, the message will NOT be sent and
4126 * instead a error will be indicated to the user.
4128 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4129 char __user *optval, int __user *optlen)
4133 if (len < sizeof(int))
4137 val = (sctp_sk(sk)->disable_fragments == 1);
4138 if (put_user(len, optlen))
4140 if (copy_to_user(optval, &val, len))
4145 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4147 * This socket option is used to specify various notifications and
4148 * ancillary data the user wishes to receive.
4150 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4155 if (len > sizeof(struct sctp_event_subscribe))
4156 len = sizeof(struct sctp_event_subscribe);
4157 if (put_user(len, optlen))
4159 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4164 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4166 * This socket option is applicable to the UDP-style socket only. When
4167 * set it will cause associations that are idle for more than the
4168 * specified number of seconds to automatically close. An association
4169 * being idle is defined an association that has NOT sent or received
4170 * user data. The special value of '0' indicates that no automatic
4171 * close of any associations should be performed. The option expects an
4172 * integer defining the number of seconds of idle time before an
4173 * association is closed.
4175 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4177 /* Applicable to UDP-style socket only */
4178 if (sctp_style(sk, TCP))
4180 if (len < sizeof(int))
4183 if (put_user(len, optlen))
4185 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4190 /* Helper routine to branch off an association to a new socket. */
4191 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
4192 struct socket **sockp)
4194 struct sock *sk = asoc->base.sk;
4195 struct socket *sock;
4199 /* An association cannot be branched off from an already peeled-off
4200 * socket, nor is this supported for tcp style sockets.
4202 if (!sctp_style(sk, UDP))
4205 /* Create a new socket. */
4206 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4210 sctp_copy_sock(sock->sk, sk, asoc);
4212 /* Make peeled-off sockets more like 1-1 accepted sockets.
4213 * Set the daddr and initialize id to something more random
4215 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4216 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4218 /* Populate the fields of the newsk from the oldsk and migrate the
4219 * asoc to the newsk.
4221 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4228 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4230 sctp_peeloff_arg_t peeloff;
4231 struct socket *newsock;
4233 struct sctp_association *asoc;
4235 if (len < sizeof(sctp_peeloff_arg_t))
4237 len = sizeof(sctp_peeloff_arg_t);
4238 if (copy_from_user(&peeloff, optval, len))
4241 asoc = sctp_id2assoc(sk, peeloff.associd);
4247 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
4249 retval = sctp_do_peeloff(asoc, &newsock);
4253 /* Map the socket to an unused fd that can be returned to the user. */
4254 retval = sock_map_fd(newsock, 0);
4256 sock_release(newsock);
4260 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
4261 __func__, sk, asoc, newsock->sk, retval);
4263 /* Return the fd mapped to the new socket. */
4264 peeloff.sd = retval;
4265 if (put_user(len, optlen))
4267 if (copy_to_user(optval, &peeloff, len))
4274 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4276 * Applications can enable or disable heartbeats for any peer address of
4277 * an association, modify an address's heartbeat interval, force a
4278 * heartbeat to be sent immediately, and adjust the address's maximum
4279 * number of retransmissions sent before an address is considered
4280 * unreachable. The following structure is used to access and modify an
4281 * address's parameters:
4283 * struct sctp_paddrparams {
4284 * sctp_assoc_t spp_assoc_id;
4285 * struct sockaddr_storage spp_address;
4286 * uint32_t spp_hbinterval;
4287 * uint16_t spp_pathmaxrxt;
4288 * uint32_t spp_pathmtu;
4289 * uint32_t spp_sackdelay;
4290 * uint32_t spp_flags;
4293 * spp_assoc_id - (one-to-many style socket) This is filled in the
4294 * application, and identifies the association for
4296 * spp_address - This specifies which address is of interest.
4297 * spp_hbinterval - This contains the value of the heartbeat interval,
4298 * in milliseconds. If a value of zero
4299 * is present in this field then no changes are to
4300 * be made to this parameter.
4301 * spp_pathmaxrxt - This contains the maximum number of
4302 * retransmissions before this address shall be
4303 * considered unreachable. If a value of zero
4304 * is present in this field then no changes are to
4305 * be made to this parameter.
4306 * spp_pathmtu - When Path MTU discovery is disabled the value
4307 * specified here will be the "fixed" path mtu.
4308 * Note that if the spp_address field is empty
4309 * then all associations on this address will
4310 * have this fixed path mtu set upon them.
4312 * spp_sackdelay - When delayed sack is enabled, this value specifies
4313 * the number of milliseconds that sacks will be delayed
4314 * for. This value will apply to all addresses of an
4315 * association if the spp_address field is empty. Note
4316 * also, that if delayed sack is enabled and this
4317 * value is set to 0, no change is made to the last
4318 * recorded delayed sack timer value.
4320 * spp_flags - These flags are used to control various features
4321 * on an association. The flag field may contain
4322 * zero or more of the following options.
4324 * SPP_HB_ENABLE - Enable heartbeats on the
4325 * specified address. Note that if the address
4326 * field is empty all addresses for the association
4327 * have heartbeats enabled upon them.
4329 * SPP_HB_DISABLE - Disable heartbeats on the
4330 * speicifed address. Note that if the address
4331 * field is empty all addresses for the association
4332 * will have their heartbeats disabled. Note also
4333 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4334 * mutually exclusive, only one of these two should
4335 * be specified. Enabling both fields will have
4336 * undetermined results.
4338 * SPP_HB_DEMAND - Request a user initiated heartbeat
4339 * to be made immediately.
4341 * SPP_PMTUD_ENABLE - This field will enable PMTU
4342 * discovery upon the specified address. Note that
4343 * if the address feild is empty then all addresses
4344 * on the association are effected.
4346 * SPP_PMTUD_DISABLE - This field will disable PMTU
4347 * discovery upon the specified address. Note that
4348 * if the address feild is empty then all addresses
4349 * on the association are effected. Not also that
4350 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4351 * exclusive. Enabling both will have undetermined
4354 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4355 * on delayed sack. The time specified in spp_sackdelay
4356 * is used to specify the sack delay for this address. Note
4357 * that if spp_address is empty then all addresses will
4358 * enable delayed sack and take on the sack delay
4359 * value specified in spp_sackdelay.
4360 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4361 * off delayed sack. If the spp_address field is blank then
4362 * delayed sack is disabled for the entire association. Note
4363 * also that this field is mutually exclusive to
4364 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4367 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4368 char __user *optval, int __user *optlen)
4370 struct sctp_paddrparams params;
4371 struct sctp_transport *trans = NULL;
4372 struct sctp_association *asoc = NULL;
4373 struct sctp_sock *sp = sctp_sk(sk);
4375 if (len < sizeof(struct sctp_paddrparams))
4377 len = sizeof(struct sctp_paddrparams);
4378 if (copy_from_user(¶ms, optval, len))
4381 /* If an address other than INADDR_ANY is specified, and
4382 * no transport is found, then the request is invalid.
4384 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
4385 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4386 params.spp_assoc_id);
4388 SCTP_DEBUG_PRINTK("Failed no transport\n");
4393 /* Get association, if assoc_id != 0 and the socket is a one
4394 * to many style socket, and an association was not found, then
4395 * the id was invalid.
4397 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4398 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4399 SCTP_DEBUG_PRINTK("Failed no association\n");
4404 /* Fetch transport values. */
4405 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4406 params.spp_pathmtu = trans->pathmtu;
4407 params.spp_pathmaxrxt = trans->pathmaxrxt;
4408 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4410 /*draft-11 doesn't say what to return in spp_flags*/
4411 params.spp_flags = trans->param_flags;
4413 /* Fetch association values. */
4414 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4415 params.spp_pathmtu = asoc->pathmtu;
4416 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4417 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4419 /*draft-11 doesn't say what to return in spp_flags*/
4420 params.spp_flags = asoc->param_flags;
4422 /* Fetch socket values. */
4423 params.spp_hbinterval = sp->hbinterval;
4424 params.spp_pathmtu = sp->pathmtu;
4425 params.spp_sackdelay = sp->sackdelay;
4426 params.spp_pathmaxrxt = sp->pathmaxrxt;
4428 /*draft-11 doesn't say what to return in spp_flags*/
4429 params.spp_flags = sp->param_flags;
4432 if (copy_to_user(optval, ¶ms, len))
4435 if (put_user(len, optlen))
4442 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4444 * This option will effect the way delayed acks are performed. This
4445 * option allows you to get or set the delayed ack time, in
4446 * milliseconds. It also allows changing the delayed ack frequency.
4447 * Changing the frequency to 1 disables the delayed sack algorithm. If
4448 * the assoc_id is 0, then this sets or gets the endpoints default
4449 * values. If the assoc_id field is non-zero, then the set or get
4450 * effects the specified association for the one to many model (the
4451 * assoc_id field is ignored by the one to one model). Note that if
4452 * sack_delay or sack_freq are 0 when setting this option, then the
4453 * current values will remain unchanged.
4455 * struct sctp_sack_info {
4456 * sctp_assoc_t sack_assoc_id;
4457 * uint32_t sack_delay;
4458 * uint32_t sack_freq;
4461 * sack_assoc_id - This parameter, indicates which association the user
4462 * is performing an action upon. Note that if this field's value is
4463 * zero then the endpoints default value is changed (effecting future
4464 * associations only).
4466 * sack_delay - This parameter contains the number of milliseconds that
4467 * the user is requesting the delayed ACK timer be set to. Note that
4468 * this value is defined in the standard to be between 200 and 500
4471 * sack_freq - This parameter contains the number of packets that must
4472 * be received before a sack is sent without waiting for the delay
4473 * timer to expire. The default value for this is 2, setting this
4474 * value to 1 will disable the delayed sack algorithm.
4476 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4477 char __user *optval,
4480 struct sctp_sack_info params;
4481 struct sctp_association *asoc = NULL;
4482 struct sctp_sock *sp = sctp_sk(sk);
4484 if (len >= sizeof(struct sctp_sack_info)) {
4485 len = sizeof(struct sctp_sack_info);
4487 if (copy_from_user(¶ms, optval, len))
4489 } else if (len == sizeof(struct sctp_assoc_value)) {
4490 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4491 pr_warn("Use struct sctp_sack_info instead\n");
4492 if (copy_from_user(¶ms, optval, len))
4497 /* Get association, if sack_assoc_id != 0 and the socket is a one
4498 * to many style socket, and an association was not found, then
4499 * the id was invalid.
4501 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4502 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4506 /* Fetch association values. */
4507 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4508 params.sack_delay = jiffies_to_msecs(
4510 params.sack_freq = asoc->sackfreq;
4513 params.sack_delay = 0;
4514 params.sack_freq = 1;
4517 /* Fetch socket values. */
4518 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4519 params.sack_delay = sp->sackdelay;
4520 params.sack_freq = sp->sackfreq;
4522 params.sack_delay = 0;
4523 params.sack_freq = 1;
4527 if (copy_to_user(optval, ¶ms, len))
4530 if (put_user(len, optlen))
4536 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4538 * Applications can specify protocol parameters for the default association
4539 * initialization. The option name argument to setsockopt() and getsockopt()
4542 * Setting initialization parameters is effective only on an unconnected
4543 * socket (for UDP-style sockets only future associations are effected
4544 * by the change). With TCP-style sockets, this option is inherited by
4545 * sockets derived from a listener socket.
4547 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4549 if (len < sizeof(struct sctp_initmsg))
4551 len = sizeof(struct sctp_initmsg);
4552 if (put_user(len, optlen))
4554 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4560 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4561 char __user *optval, int __user *optlen)
4563 struct sctp_association *asoc;
4565 struct sctp_getaddrs getaddrs;
4566 struct sctp_transport *from;
4568 union sctp_addr temp;
4569 struct sctp_sock *sp = sctp_sk(sk);
4574 if (len < sizeof(struct sctp_getaddrs))
4577 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4580 /* For UDP-style sockets, id specifies the association to query. */
4581 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4585 to = optval + offsetof(struct sctp_getaddrs,addrs);
4586 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4588 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4590 memcpy(&temp, &from->ipaddr, sizeof(temp));
4591 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4592 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4593 if (space_left < addrlen)
4595 if (copy_to_user(to, &temp, addrlen))
4599 space_left -= addrlen;
4602 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4604 bytes_copied = ((char __user *)to) - optval;
4605 if (put_user(bytes_copied, optlen))
4611 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4612 size_t space_left, int *bytes_copied)
4614 struct sctp_sockaddr_entry *addr;
4615 union sctp_addr temp;
4620 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4624 if ((PF_INET == sk->sk_family) &&
4625 (AF_INET6 == addr->a.sa.sa_family))
4627 if ((PF_INET6 == sk->sk_family) &&
4628 inet_v6_ipv6only(sk) &&
4629 (AF_INET == addr->a.sa.sa_family))
4631 memcpy(&temp, &addr->a, sizeof(temp));
4632 if (!temp.v4.sin_port)
4633 temp.v4.sin_port = htons(port);
4635 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4637 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4638 if (space_left < addrlen) {
4642 memcpy(to, &temp, addrlen);
4646 space_left -= addrlen;
4647 *bytes_copied += addrlen;
4655 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4656 char __user *optval, int __user *optlen)
4658 struct sctp_bind_addr *bp;
4659 struct sctp_association *asoc;
4661 struct sctp_getaddrs getaddrs;
4662 struct sctp_sockaddr_entry *addr;
4664 union sctp_addr temp;
4665 struct sctp_sock *sp = sctp_sk(sk);
4669 int bytes_copied = 0;
4673 if (len < sizeof(struct sctp_getaddrs))
4676 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4680 * For UDP-style sockets, id specifies the association to query.
4681 * If the id field is set to the value '0' then the locally bound
4682 * addresses are returned without regard to any particular
4685 if (0 == getaddrs.assoc_id) {
4686 bp = &sctp_sk(sk)->ep->base.bind_addr;
4688 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4691 bp = &asoc->base.bind_addr;
4694 to = optval + offsetof(struct sctp_getaddrs,addrs);
4695 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4697 addrs = kmalloc(space_left, GFP_KERNEL);
4701 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4702 * addresses from the global local address list.
4704 if (sctp_list_single_entry(&bp->address_list)) {
4705 addr = list_entry(bp->address_list.next,
4706 struct sctp_sockaddr_entry, list);
4707 if (sctp_is_any(sk, &addr->a)) {
4708 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4709 space_left, &bytes_copied);
4719 /* Protection on the bound address list is not needed since
4720 * in the socket option context we hold a socket lock and
4721 * thus the bound address list can't change.
4723 list_for_each_entry(addr, &bp->address_list, list) {
4724 memcpy(&temp, &addr->a, sizeof(temp));
4725 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4726 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4727 if (space_left < addrlen) {
4728 err = -ENOMEM; /*fixme: right error?*/
4731 memcpy(buf, &temp, addrlen);
4733 bytes_copied += addrlen;
4735 space_left -= addrlen;
4739 if (copy_to_user(to, addrs, bytes_copied)) {
4743 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4747 if (put_user(bytes_copied, optlen))
4754 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4756 * Requests that the local SCTP stack use the enclosed peer address as
4757 * the association primary. The enclosed address must be one of the
4758 * association peer's addresses.
4760 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4761 char __user *optval, int __user *optlen)
4763 struct sctp_prim prim;
4764 struct sctp_association *asoc;
4765 struct sctp_sock *sp = sctp_sk(sk);
4767 if (len < sizeof(struct sctp_prim))
4770 len = sizeof(struct sctp_prim);
4772 if (copy_from_user(&prim, optval, len))
4775 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4779 if (!asoc->peer.primary_path)
4782 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4783 asoc->peer.primary_path->af_specific->sockaddr_len);
4785 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4786 (union sctp_addr *)&prim.ssp_addr);
4788 if (put_user(len, optlen))
4790 if (copy_to_user(optval, &prim, len))
4797 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4799 * Requests that the local endpoint set the specified Adaptation Layer
4800 * Indication parameter for all future INIT and INIT-ACK exchanges.
4802 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4803 char __user *optval, int __user *optlen)
4805 struct sctp_setadaptation adaptation;
4807 if (len < sizeof(struct sctp_setadaptation))
4810 len = sizeof(struct sctp_setadaptation);
4812 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4814 if (put_user(len, optlen))
4816 if (copy_to_user(optval, &adaptation, len))
4824 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4826 * Applications that wish to use the sendto() system call may wish to
4827 * specify a default set of parameters that would normally be supplied
4828 * through the inclusion of ancillary data. This socket option allows
4829 * such an application to set the default sctp_sndrcvinfo structure.
4832 * The application that wishes to use this socket option simply passes
4833 * in to this call the sctp_sndrcvinfo structure defined in Section
4834 * 5.2.2) The input parameters accepted by this call include
4835 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4836 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4837 * to this call if the caller is using the UDP model.
4839 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4841 static int sctp_getsockopt_default_send_param(struct sock *sk,
4842 int len, char __user *optval,
4845 struct sctp_sndrcvinfo info;
4846 struct sctp_association *asoc;
4847 struct sctp_sock *sp = sctp_sk(sk);
4849 if (len < sizeof(struct sctp_sndrcvinfo))
4852 len = sizeof(struct sctp_sndrcvinfo);
4854 if (copy_from_user(&info, optval, len))
4857 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4858 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4862 info.sinfo_stream = asoc->default_stream;
4863 info.sinfo_flags = asoc->default_flags;
4864 info.sinfo_ppid = asoc->default_ppid;
4865 info.sinfo_context = asoc->default_context;
4866 info.sinfo_timetolive = asoc->default_timetolive;
4868 info.sinfo_stream = sp->default_stream;
4869 info.sinfo_flags = sp->default_flags;
4870 info.sinfo_ppid = sp->default_ppid;
4871 info.sinfo_context = sp->default_context;
4872 info.sinfo_timetolive = sp->default_timetolive;
4875 if (put_user(len, optlen))
4877 if (copy_to_user(optval, &info, len))
4885 * 7.1.5 SCTP_NODELAY
4887 * Turn on/off any Nagle-like algorithm. This means that packets are
4888 * generally sent as soon as possible and no unnecessary delays are
4889 * introduced, at the cost of more packets in the network. Expects an
4890 * integer boolean flag.
4893 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4894 char __user *optval, int __user *optlen)
4898 if (len < sizeof(int))
4902 val = (sctp_sk(sk)->nodelay == 1);
4903 if (put_user(len, optlen))
4905 if (copy_to_user(optval, &val, len))
4912 * 7.1.1 SCTP_RTOINFO
4914 * The protocol parameters used to initialize and bound retransmission
4915 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4916 * and modify these parameters.
4917 * All parameters are time values, in milliseconds. A value of 0, when
4918 * modifying the parameters, indicates that the current value should not
4922 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4923 char __user *optval,
4924 int __user *optlen) {
4925 struct sctp_rtoinfo rtoinfo;
4926 struct sctp_association *asoc;
4928 if (len < sizeof (struct sctp_rtoinfo))
4931 len = sizeof(struct sctp_rtoinfo);
4933 if (copy_from_user(&rtoinfo, optval, len))
4936 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4938 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4941 /* Values corresponding to the specific association. */
4943 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4944 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4945 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4947 /* Values corresponding to the endpoint. */
4948 struct sctp_sock *sp = sctp_sk(sk);
4950 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4951 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4952 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4955 if (put_user(len, optlen))
4958 if (copy_to_user(optval, &rtoinfo, len))
4966 * 7.1.2 SCTP_ASSOCINFO
4968 * This option is used to tune the maximum retransmission attempts
4969 * of the association.
4970 * Returns an error if the new association retransmission value is
4971 * greater than the sum of the retransmission value of the peer.
4972 * See [SCTP] for more information.
4975 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4976 char __user *optval,
4980 struct sctp_assocparams assocparams;
4981 struct sctp_association *asoc;
4982 struct list_head *pos;
4985 if (len < sizeof (struct sctp_assocparams))
4988 len = sizeof(struct sctp_assocparams);
4990 if (copy_from_user(&assocparams, optval, len))
4993 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4995 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4998 /* Values correspoinding to the specific association */
5000 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5001 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5002 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5003 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
5005 (asoc->cookie_life.tv_usec
5008 list_for_each(pos, &asoc->peer.transport_addr_list) {
5012 assocparams.sasoc_number_peer_destinations = cnt;
5014 /* Values corresponding to the endpoint */
5015 struct sctp_sock *sp = sctp_sk(sk);
5017 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5018 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5019 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5020 assocparams.sasoc_cookie_life =
5021 sp->assocparams.sasoc_cookie_life;
5022 assocparams.sasoc_number_peer_destinations =
5024 sasoc_number_peer_destinations;
5027 if (put_user(len, optlen))
5030 if (copy_to_user(optval, &assocparams, len))
5037 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5039 * This socket option is a boolean flag which turns on or off mapped V4
5040 * addresses. If this option is turned on and the socket is type
5041 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5042 * If this option is turned off, then no mapping will be done of V4
5043 * addresses and a user will receive both PF_INET6 and PF_INET type
5044 * addresses on the socket.
5046 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5047 char __user *optval, int __user *optlen)
5050 struct sctp_sock *sp = sctp_sk(sk);
5052 if (len < sizeof(int))
5057 if (put_user(len, optlen))
5059 if (copy_to_user(optval, &val, len))
5066 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5067 * (chapter and verse is quoted at sctp_setsockopt_context())
5069 static int sctp_getsockopt_context(struct sock *sk, int len,
5070 char __user *optval, int __user *optlen)
5072 struct sctp_assoc_value params;
5073 struct sctp_sock *sp;
5074 struct sctp_association *asoc;
5076 if (len < sizeof(struct sctp_assoc_value))
5079 len = sizeof(struct sctp_assoc_value);
5081 if (copy_from_user(¶ms, optval, len))
5086 if (params.assoc_id != 0) {
5087 asoc = sctp_id2assoc(sk, params.assoc_id);
5090 params.assoc_value = asoc->default_rcv_context;
5092 params.assoc_value = sp->default_rcv_context;
5095 if (put_user(len, optlen))
5097 if (copy_to_user(optval, ¶ms, len))
5104 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5105 * This option will get or set the maximum size to put in any outgoing
5106 * SCTP DATA chunk. If a message is larger than this size it will be
5107 * fragmented by SCTP into the specified size. Note that the underlying
5108 * SCTP implementation may fragment into smaller sized chunks when the
5109 * PMTU of the underlying association is smaller than the value set by
5110 * the user. The default value for this option is '0' which indicates
5111 * the user is NOT limiting fragmentation and only the PMTU will effect
5112 * SCTP's choice of DATA chunk size. Note also that values set larger
5113 * than the maximum size of an IP datagram will effectively let SCTP
5114 * control fragmentation (i.e. the same as setting this option to 0).
5116 * The following structure is used to access and modify this parameter:
5118 * struct sctp_assoc_value {
5119 * sctp_assoc_t assoc_id;
5120 * uint32_t assoc_value;
5123 * assoc_id: This parameter is ignored for one-to-one style sockets.
5124 * For one-to-many style sockets this parameter indicates which
5125 * association the user is performing an action upon. Note that if
5126 * this field's value is zero then the endpoints default value is
5127 * changed (effecting future associations only).
5128 * assoc_value: This parameter specifies the maximum size in bytes.
5130 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5131 char __user *optval, int __user *optlen)
5133 struct sctp_assoc_value params;
5134 struct sctp_association *asoc;
5136 if (len == sizeof(int)) {
5137 pr_warn("Use of int in maxseg socket option deprecated\n");
5138 pr_warn("Use struct sctp_assoc_value instead\n");
5139 params.assoc_id = 0;
5140 } else if (len >= sizeof(struct sctp_assoc_value)) {
5141 len = sizeof(struct sctp_assoc_value);
5142 if (copy_from_user(¶ms, optval, sizeof(params)))
5147 asoc = sctp_id2assoc(sk, params.assoc_id);
5148 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5152 params.assoc_value = asoc->frag_point;
5154 params.assoc_value = sctp_sk(sk)->user_frag;
5156 if (put_user(len, optlen))
5158 if (len == sizeof(int)) {
5159 if (copy_to_user(optval, ¶ms.assoc_value, len))
5162 if (copy_to_user(optval, ¶ms, len))
5170 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5171 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5173 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5174 char __user *optval, int __user *optlen)
5178 if (len < sizeof(int))
5183 val = sctp_sk(sk)->frag_interleave;
5184 if (put_user(len, optlen))
5186 if (copy_to_user(optval, &val, len))
5193 * 7.1.25. Set or Get the sctp partial delivery point
5194 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5196 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5197 char __user *optval,
5202 if (len < sizeof(u32))
5207 val = sctp_sk(sk)->pd_point;
5208 if (put_user(len, optlen))
5210 if (copy_to_user(optval, &val, len))
5217 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5218 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5220 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5221 char __user *optval,
5224 struct sctp_assoc_value params;
5225 struct sctp_sock *sp;
5226 struct sctp_association *asoc;
5228 if (len == sizeof(int)) {
5229 pr_warn("Use of int in max_burst socket option deprecated\n");
5230 pr_warn("Use struct sctp_assoc_value instead\n");
5231 params.assoc_id = 0;
5232 } else if (len >= sizeof(struct sctp_assoc_value)) {
5233 len = sizeof(struct sctp_assoc_value);
5234 if (copy_from_user(¶ms, optval, len))
5241 if (params.assoc_id != 0) {
5242 asoc = sctp_id2assoc(sk, params.assoc_id);
5245 params.assoc_value = asoc->max_burst;
5247 params.assoc_value = sp->max_burst;
5249 if (len == sizeof(int)) {
5250 if (copy_to_user(optval, ¶ms.assoc_value, len))
5253 if (copy_to_user(optval, ¶ms, len))
5261 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5262 char __user *optval, int __user *optlen)
5264 struct sctp_hmacalgo __user *p = (void __user *)optval;
5265 struct sctp_hmac_algo_param *hmacs;
5269 if (!sctp_auth_enable)
5272 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5273 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5275 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5278 len = sizeof(struct sctp_hmacalgo) + data_len;
5279 num_idents = data_len / sizeof(u16);
5281 if (put_user(len, optlen))
5283 if (put_user(num_idents, &p->shmac_num_idents))
5285 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5290 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5291 char __user *optval, int __user *optlen)
5293 struct sctp_authkeyid val;
5294 struct sctp_association *asoc;
5296 if (!sctp_auth_enable)
5299 if (len < sizeof(struct sctp_authkeyid))
5301 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5304 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5305 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5309 val.scact_keynumber = asoc->active_key_id;
5311 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5313 len = sizeof(struct sctp_authkeyid);
5314 if (put_user(len, optlen))
5316 if (copy_to_user(optval, &val, len))
5322 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5323 char __user *optval, int __user *optlen)
5325 struct sctp_authchunks __user *p = (void __user *)optval;
5326 struct sctp_authchunks val;
5327 struct sctp_association *asoc;
5328 struct sctp_chunks_param *ch;
5332 if (!sctp_auth_enable)
5335 if (len < sizeof(struct sctp_authchunks))
5338 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5341 to = p->gauth_chunks;
5342 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5346 ch = asoc->peer.peer_chunks;
5350 /* See if the user provided enough room for all the data */
5351 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5352 if (len < num_chunks)
5355 if (copy_to_user(to, ch->chunks, num_chunks))
5358 len = sizeof(struct sctp_authchunks) + num_chunks;
5359 if (put_user(len, optlen)) return -EFAULT;
5360 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5365 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5366 char __user *optval, int __user *optlen)
5368 struct sctp_authchunks __user *p = (void __user *)optval;
5369 struct sctp_authchunks val;
5370 struct sctp_association *asoc;
5371 struct sctp_chunks_param *ch;
5375 if (!sctp_auth_enable)
5378 if (len < sizeof(struct sctp_authchunks))
5381 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5384 to = p->gauth_chunks;
5385 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5386 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5390 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5392 ch = sctp_sk(sk)->ep->auth_chunk_list;
5397 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5398 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5401 if (copy_to_user(to, ch->chunks, num_chunks))
5404 len = sizeof(struct sctp_authchunks) + num_chunks;
5405 if (put_user(len, optlen))
5407 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5414 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5415 * This option gets the current number of associations that are attached
5416 * to a one-to-many style socket. The option value is an uint32_t.
5418 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5419 char __user *optval, int __user *optlen)
5421 struct sctp_sock *sp = sctp_sk(sk);
5422 struct sctp_association *asoc;
5425 if (sctp_style(sk, TCP))
5428 if (len < sizeof(u32))
5433 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5437 if (put_user(len, optlen))
5439 if (copy_to_user(optval, &val, len))
5446 * 8.1.23 SCTP_AUTO_ASCONF
5447 * See the corresponding setsockopt entry as description
5449 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5450 char __user *optval, int __user *optlen)
5454 if (len < sizeof(int))
5458 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5460 if (put_user(len, optlen))
5462 if (copy_to_user(optval, &val, len))
5468 * 8.2.6. Get the Current Identifiers of Associations
5469 * (SCTP_GET_ASSOC_ID_LIST)
5471 * This option gets the current list of SCTP association identifiers of
5472 * the SCTP associations handled by a one-to-many style socket.
5474 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5475 char __user *optval, int __user *optlen)
5477 struct sctp_sock *sp = sctp_sk(sk);
5478 struct sctp_association *asoc;
5479 struct sctp_assoc_ids *ids;
5482 if (sctp_style(sk, TCP))
5485 if (len < sizeof(struct sctp_assoc_ids))
5488 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5492 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5495 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5497 ids = kmalloc(len, GFP_KERNEL);
5501 ids->gaids_number_of_ids = num;
5503 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5504 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5507 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5516 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5517 char __user *optval, int __user *optlen)
5522 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5525 /* I can hardly begin to describe how wrong this is. This is
5526 * so broken as to be worse than useless. The API draft
5527 * REALLY is NOT helpful here... I am not convinced that the
5528 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5529 * are at all well-founded.
5531 if (level != SOL_SCTP) {
5532 struct sctp_af *af = sctp_sk(sk)->pf->af;
5534 retval = af->getsockopt(sk, level, optname, optval, optlen);
5538 if (get_user(len, optlen))
5545 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5547 case SCTP_DISABLE_FRAGMENTS:
5548 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5552 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5554 case SCTP_AUTOCLOSE:
5555 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5557 case SCTP_SOCKOPT_PEELOFF:
5558 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5560 case SCTP_PEER_ADDR_PARAMS:
5561 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5564 case SCTP_DELAYED_SACK:
5565 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5569 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5571 case SCTP_GET_PEER_ADDRS:
5572 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5575 case SCTP_GET_LOCAL_ADDRS:
5576 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5579 case SCTP_SOCKOPT_CONNECTX3:
5580 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5582 case SCTP_DEFAULT_SEND_PARAM:
5583 retval = sctp_getsockopt_default_send_param(sk, len,
5586 case SCTP_PRIMARY_ADDR:
5587 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5590 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5593 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5595 case SCTP_ASSOCINFO:
5596 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5598 case SCTP_I_WANT_MAPPED_V4_ADDR:
5599 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5602 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5604 case SCTP_GET_PEER_ADDR_INFO:
5605 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5608 case SCTP_ADAPTATION_LAYER:
5609 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5613 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5615 case SCTP_FRAGMENT_INTERLEAVE:
5616 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5619 case SCTP_PARTIAL_DELIVERY_POINT:
5620 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5623 case SCTP_MAX_BURST:
5624 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5627 case SCTP_AUTH_CHUNK:
5628 case SCTP_AUTH_DELETE_KEY:
5629 retval = -EOPNOTSUPP;
5631 case SCTP_HMAC_IDENT:
5632 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5634 case SCTP_AUTH_ACTIVE_KEY:
5635 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5637 case SCTP_PEER_AUTH_CHUNKS:
5638 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5641 case SCTP_LOCAL_AUTH_CHUNKS:
5642 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5645 case SCTP_GET_ASSOC_NUMBER:
5646 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5648 case SCTP_GET_ASSOC_ID_LIST:
5649 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5651 case SCTP_AUTO_ASCONF:
5652 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5655 retval = -ENOPROTOOPT;
5659 sctp_release_sock(sk);
5663 static void sctp_hash(struct sock *sk)
5668 static void sctp_unhash(struct sock *sk)
5673 /* Check if port is acceptable. Possibly find first available port.
5675 * The port hash table (contained in the 'global' SCTP protocol storage
5676 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5677 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5678 * list (the list number is the port number hashed out, so as you
5679 * would expect from a hash function, all the ports in a given list have
5680 * such a number that hashes out to the same list number; you were
5681 * expecting that, right?); so each list has a set of ports, with a
5682 * link to the socket (struct sock) that uses it, the port number and
5683 * a fastreuse flag (FIXME: NPI ipg).
5685 static struct sctp_bind_bucket *sctp_bucket_create(
5686 struct sctp_bind_hashbucket *head, unsigned short snum);
5688 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5690 struct sctp_bind_hashbucket *head; /* hash list */
5691 struct sctp_bind_bucket *pp; /* hash list port iterator */
5692 struct hlist_node *node;
5693 unsigned short snum;
5696 snum = ntohs(addr->v4.sin_port);
5698 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5699 sctp_local_bh_disable();
5702 /* Search for an available port. */
5703 int low, high, remaining, index;
5706 inet_get_local_port_range(&low, &high);
5707 remaining = (high - low) + 1;
5708 rover = net_random() % remaining + low;
5712 if ((rover < low) || (rover > high))
5714 if (inet_is_reserved_local_port(rover))
5716 index = sctp_phashfn(rover);
5717 head = &sctp_port_hashtable[index];
5718 sctp_spin_lock(&head->lock);
5719 sctp_for_each_hentry(pp, node, &head->chain)
5720 if (pp->port == rover)
5724 sctp_spin_unlock(&head->lock);
5725 } while (--remaining > 0);
5727 /* Exhausted local port range during search? */
5732 /* OK, here is the one we will use. HEAD (the port
5733 * hash table list entry) is non-NULL and we hold it's
5738 /* We are given an specific port number; we verify
5739 * that it is not being used. If it is used, we will
5740 * exahust the search in the hash list corresponding
5741 * to the port number (snum) - we detect that with the
5742 * port iterator, pp being NULL.
5744 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5745 sctp_spin_lock(&head->lock);
5746 sctp_for_each_hentry(pp, node, &head->chain) {
5747 if (pp->port == snum)
5754 if (!hlist_empty(&pp->owner)) {
5755 /* We had a port hash table hit - there is an
5756 * available port (pp != NULL) and it is being
5757 * used by other socket (pp->owner not empty); that other
5758 * socket is going to be sk2.
5760 int reuse = sk->sk_reuse;
5763 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5764 if (pp->fastreuse && sk->sk_reuse &&
5765 sk->sk_state != SCTP_SS_LISTENING)
5768 /* Run through the list of sockets bound to the port
5769 * (pp->port) [via the pointers bind_next and
5770 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5771 * we get the endpoint they describe and run through
5772 * the endpoint's list of IP (v4 or v6) addresses,
5773 * comparing each of the addresses with the address of
5774 * the socket sk. If we find a match, then that means
5775 * that this port/socket (sk) combination are already
5778 sk_for_each_bound(sk2, node, &pp->owner) {
5779 struct sctp_endpoint *ep2;
5780 ep2 = sctp_sk(sk2)->ep;
5783 (reuse && sk2->sk_reuse &&
5784 sk2->sk_state != SCTP_SS_LISTENING))
5787 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5788 sctp_sk(sk2), sctp_sk(sk))) {
5793 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5796 /* If there was a hash table miss, create a new port. */
5798 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5801 /* In either case (hit or miss), make sure fastreuse is 1 only
5802 * if sk->sk_reuse is too (that is, if the caller requested
5803 * SO_REUSEADDR on this socket -sk-).
5805 if (hlist_empty(&pp->owner)) {
5806 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5810 } else if (pp->fastreuse &&
5811 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5814 /* We are set, so fill up all the data in the hash table
5815 * entry, tie the socket list information with the rest of the
5816 * sockets FIXME: Blurry, NPI (ipg).
5819 if (!sctp_sk(sk)->bind_hash) {
5820 inet_sk(sk)->inet_num = snum;
5821 sk_add_bind_node(sk, &pp->owner);
5822 sctp_sk(sk)->bind_hash = pp;
5827 sctp_spin_unlock(&head->lock);
5830 sctp_local_bh_enable();
5834 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5835 * port is requested.
5837 static int sctp_get_port(struct sock *sk, unsigned short snum)
5840 union sctp_addr addr;
5841 struct sctp_af *af = sctp_sk(sk)->pf->af;
5843 /* Set up a dummy address struct from the sk. */
5844 af->from_sk(&addr, sk);
5845 addr.v4.sin_port = htons(snum);
5847 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5848 ret = sctp_get_port_local(sk, &addr);
5854 * Move a socket to LISTENING state.
5856 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5858 struct sctp_sock *sp = sctp_sk(sk);
5859 struct sctp_endpoint *ep = sp->ep;
5860 struct crypto_hash *tfm = NULL;
5862 /* Allocate HMAC for generating cookie. */
5863 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5864 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5866 if (net_ratelimit()) {
5867 pr_info("failed to load transform for %s: %ld\n",
5868 sctp_hmac_alg, PTR_ERR(tfm));
5872 sctp_sk(sk)->hmac = tfm;
5876 * If a bind() or sctp_bindx() is not called prior to a listen()
5877 * call that allows new associations to be accepted, the system
5878 * picks an ephemeral port and will choose an address set equivalent
5879 * to binding with a wildcard address.
5881 * This is not currently spelled out in the SCTP sockets
5882 * extensions draft, but follows the practice as seen in TCP
5886 sk->sk_state = SCTP_SS_LISTENING;
5887 if (!ep->base.bind_addr.port) {
5888 if (sctp_autobind(sk))
5891 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
5892 sk->sk_state = SCTP_SS_CLOSED;
5897 sk->sk_max_ack_backlog = backlog;
5898 sctp_hash_endpoint(ep);
5903 * 4.1.3 / 5.1.3 listen()
5905 * By default, new associations are not accepted for UDP style sockets.
5906 * An application uses listen() to mark a socket as being able to
5907 * accept new associations.
5909 * On TCP style sockets, applications use listen() to ready the SCTP
5910 * endpoint for accepting inbound associations.
5912 * On both types of endpoints a backlog of '0' disables listening.
5914 * Move a socket to LISTENING state.
5916 int sctp_inet_listen(struct socket *sock, int backlog)
5918 struct sock *sk = sock->sk;
5919 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5922 if (unlikely(backlog < 0))
5927 /* Peeled-off sockets are not allowed to listen(). */
5928 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
5931 if (sock->state != SS_UNCONNECTED)
5934 /* If backlog is zero, disable listening. */
5936 if (sctp_sstate(sk, CLOSED))
5940 sctp_unhash_endpoint(ep);
5941 sk->sk_state = SCTP_SS_CLOSED;
5943 sctp_sk(sk)->bind_hash->fastreuse = 1;
5947 /* If we are already listening, just update the backlog */
5948 if (sctp_sstate(sk, LISTENING))
5949 sk->sk_max_ack_backlog = backlog;
5951 err = sctp_listen_start(sk, backlog);
5958 sctp_release_sock(sk);
5963 * This function is done by modeling the current datagram_poll() and the
5964 * tcp_poll(). Note that, based on these implementations, we don't
5965 * lock the socket in this function, even though it seems that,
5966 * ideally, locking or some other mechanisms can be used to ensure
5967 * the integrity of the counters (sndbuf and wmem_alloc) used
5968 * in this place. We assume that we don't need locks either until proven
5971 * Another thing to note is that we include the Async I/O support
5972 * here, again, by modeling the current TCP/UDP code. We don't have
5973 * a good way to test with it yet.
5975 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5977 struct sock *sk = sock->sk;
5978 struct sctp_sock *sp = sctp_sk(sk);
5981 poll_wait(file, sk_sleep(sk), wait);
5983 /* A TCP-style listening socket becomes readable when the accept queue
5986 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5987 return (!list_empty(&sp->ep->asocs)) ?
5988 (POLLIN | POLLRDNORM) : 0;
5992 /* Is there any exceptional events? */
5993 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5995 if (sk->sk_shutdown & RCV_SHUTDOWN)
5996 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
5997 if (sk->sk_shutdown == SHUTDOWN_MASK)
6000 /* Is it readable? Reconsider this code with TCP-style support. */
6001 if (!skb_queue_empty(&sk->sk_receive_queue))
6002 mask |= POLLIN | POLLRDNORM;
6004 /* The association is either gone or not ready. */
6005 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6008 /* Is it writable? */
6009 if (sctp_writeable(sk)) {
6010 mask |= POLLOUT | POLLWRNORM;
6012 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6014 * Since the socket is not locked, the buffer
6015 * might be made available after the writeable check and
6016 * before the bit is set. This could cause a lost I/O
6017 * signal. tcp_poll() has a race breaker for this race
6018 * condition. Based on their implementation, we put
6019 * in the following code to cover it as well.
6021 if (sctp_writeable(sk))
6022 mask |= POLLOUT | POLLWRNORM;
6027 /********************************************************************
6028 * 2nd Level Abstractions
6029 ********************************************************************/
6031 static struct sctp_bind_bucket *sctp_bucket_create(
6032 struct sctp_bind_hashbucket *head, unsigned short snum)
6034 struct sctp_bind_bucket *pp;
6036 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6038 SCTP_DBG_OBJCNT_INC(bind_bucket);
6041 INIT_HLIST_HEAD(&pp->owner);
6042 hlist_add_head(&pp->node, &head->chain);
6047 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6048 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6050 if (pp && hlist_empty(&pp->owner)) {
6051 __hlist_del(&pp->node);
6052 kmem_cache_free(sctp_bucket_cachep, pp);
6053 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6057 /* Release this socket's reference to a local port. */
6058 static inline void __sctp_put_port(struct sock *sk)
6060 struct sctp_bind_hashbucket *head =
6061 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->inet_num)];
6062 struct sctp_bind_bucket *pp;
6064 sctp_spin_lock(&head->lock);
6065 pp = sctp_sk(sk)->bind_hash;
6066 __sk_del_bind_node(sk);
6067 sctp_sk(sk)->bind_hash = NULL;
6068 inet_sk(sk)->inet_num = 0;
6069 sctp_bucket_destroy(pp);
6070 sctp_spin_unlock(&head->lock);
6073 void sctp_put_port(struct sock *sk)
6075 sctp_local_bh_disable();
6076 __sctp_put_port(sk);
6077 sctp_local_bh_enable();
6081 * The system picks an ephemeral port and choose an address set equivalent
6082 * to binding with a wildcard address.
6083 * One of those addresses will be the primary address for the association.
6084 * This automatically enables the multihoming capability of SCTP.
6086 static int sctp_autobind(struct sock *sk)
6088 union sctp_addr autoaddr;
6092 /* Initialize a local sockaddr structure to INADDR_ANY. */
6093 af = sctp_sk(sk)->pf->af;
6095 port = htons(inet_sk(sk)->inet_num);
6096 af->inaddr_any(&autoaddr, port);
6098 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6101 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6104 * 4.2 The cmsghdr Structure *
6106 * When ancillary data is sent or received, any number of ancillary data
6107 * objects can be specified by the msg_control and msg_controllen members of
6108 * the msghdr structure, because each object is preceded by
6109 * a cmsghdr structure defining the object's length (the cmsg_len member).
6110 * Historically Berkeley-derived implementations have passed only one object
6111 * at a time, but this API allows multiple objects to be
6112 * passed in a single call to sendmsg() or recvmsg(). The following example
6113 * shows two ancillary data objects in a control buffer.
6115 * |<--------------------------- msg_controllen -------------------------->|
6118 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6120 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6123 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6125 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6128 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6129 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6131 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6133 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6140 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
6141 sctp_cmsgs_t *cmsgs)
6143 struct cmsghdr *cmsg;
6144 struct msghdr *my_msg = (struct msghdr *)msg;
6146 for (cmsg = CMSG_FIRSTHDR(msg);
6148 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6149 if (!CMSG_OK(my_msg, cmsg))
6152 /* Should we parse this header or ignore? */
6153 if (cmsg->cmsg_level != IPPROTO_SCTP)
6156 /* Strictly check lengths following example in SCM code. */
6157 switch (cmsg->cmsg_type) {
6159 /* SCTP Socket API Extension
6160 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6162 * This cmsghdr structure provides information for
6163 * initializing new SCTP associations with sendmsg().
6164 * The SCTP_INITMSG socket option uses this same data
6165 * structure. This structure is not used for
6168 * cmsg_level cmsg_type cmsg_data[]
6169 * ------------ ------------ ----------------------
6170 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6172 if (cmsg->cmsg_len !=
6173 CMSG_LEN(sizeof(struct sctp_initmsg)))
6175 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6179 /* SCTP Socket API Extension
6180 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6182 * This cmsghdr structure specifies SCTP options for
6183 * sendmsg() and describes SCTP header information
6184 * about a received message through recvmsg().
6186 * cmsg_level cmsg_type cmsg_data[]
6187 * ------------ ------------ ----------------------
6188 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6190 if (cmsg->cmsg_len !=
6191 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6195 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6197 /* Minimally, validate the sinfo_flags. */
6198 if (cmsgs->info->sinfo_flags &
6199 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6200 SCTP_ABORT | SCTP_EOF))
6212 * Wait for a packet..
6213 * Note: This function is the same function as in core/datagram.c
6214 * with a few modifications to make lksctp work.
6216 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6221 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6223 /* Socket errors? */
6224 error = sock_error(sk);
6228 if (!skb_queue_empty(&sk->sk_receive_queue))
6231 /* Socket shut down? */
6232 if (sk->sk_shutdown & RCV_SHUTDOWN)
6235 /* Sequenced packets can come disconnected. If so we report the
6240 /* Is there a good reason to think that we may receive some data? */
6241 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6244 /* Handle signals. */
6245 if (signal_pending(current))
6248 /* Let another process have a go. Since we are going to sleep
6249 * anyway. Note: This may cause odd behaviors if the message
6250 * does not fit in the user's buffer, but this seems to be the
6251 * only way to honor MSG_DONTWAIT realistically.
6253 sctp_release_sock(sk);
6254 *timeo_p = schedule_timeout(*timeo_p);
6258 finish_wait(sk_sleep(sk), &wait);
6262 error = sock_intr_errno(*timeo_p);
6265 finish_wait(sk_sleep(sk), &wait);
6270 /* Receive a datagram.
6271 * Note: This is pretty much the same routine as in core/datagram.c
6272 * with a few changes to make lksctp work.
6274 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6275 int noblock, int *err)
6278 struct sk_buff *skb;
6281 timeo = sock_rcvtimeo(sk, noblock);
6283 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6284 timeo, MAX_SCHEDULE_TIMEOUT);
6287 /* Again only user level code calls this function,
6288 * so nothing interrupt level
6289 * will suddenly eat the receive_queue.
6291 * Look at current nfs client by the way...
6292 * However, this function was correct in any case. 8)
6294 if (flags & MSG_PEEK) {
6295 spin_lock_bh(&sk->sk_receive_queue.lock);
6296 skb = skb_peek(&sk->sk_receive_queue);
6298 atomic_inc(&skb->users);
6299 spin_unlock_bh(&sk->sk_receive_queue.lock);
6301 skb = skb_dequeue(&sk->sk_receive_queue);
6307 /* Caller is allowed not to check sk->sk_err before calling. */
6308 error = sock_error(sk);
6312 if (sk->sk_shutdown & RCV_SHUTDOWN)
6315 /* User doesn't want to wait. */
6319 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6328 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6329 static void __sctp_write_space(struct sctp_association *asoc)
6331 struct sock *sk = asoc->base.sk;
6332 struct socket *sock = sk->sk_socket;
6334 if ((sctp_wspace(asoc) > 0) && sock) {
6335 if (waitqueue_active(&asoc->wait))
6336 wake_up_interruptible(&asoc->wait);
6338 if (sctp_writeable(sk)) {
6339 wait_queue_head_t *wq = sk_sleep(sk);
6341 if (wq && waitqueue_active(wq))
6342 wake_up_interruptible(wq);
6344 /* Note that we try to include the Async I/O support
6345 * here by modeling from the current TCP/UDP code.
6346 * We have not tested with it yet.
6348 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6349 sock_wake_async(sock,
6350 SOCK_WAKE_SPACE, POLL_OUT);
6355 /* Do accounting for the sndbuf space.
6356 * Decrement the used sndbuf space of the corresponding association by the
6357 * data size which was just transmitted(freed).
6359 static void sctp_wfree(struct sk_buff *skb)
6361 struct sctp_association *asoc;
6362 struct sctp_chunk *chunk;
6365 /* Get the saved chunk pointer. */
6366 chunk = *((struct sctp_chunk **)(skb->cb));
6369 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6370 sizeof(struct sk_buff) +
6371 sizeof(struct sctp_chunk);
6373 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6376 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6378 sk->sk_wmem_queued -= skb->truesize;
6379 sk_mem_uncharge(sk, skb->truesize);
6382 __sctp_write_space(asoc);
6384 sctp_association_put(asoc);
6387 /* Do accounting for the receive space on the socket.
6388 * Accounting for the association is done in ulpevent.c
6389 * We set this as a destructor for the cloned data skbs so that
6390 * accounting is done at the correct time.
6392 void sctp_sock_rfree(struct sk_buff *skb)
6394 struct sock *sk = skb->sk;
6395 struct sctp_ulpevent *event = sctp_skb2event(skb);
6397 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6400 * Mimic the behavior of sock_rfree
6402 sk_mem_uncharge(sk, event->rmem_len);
6406 /* Helper function to wait for space in the sndbuf. */
6407 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6410 struct sock *sk = asoc->base.sk;
6412 long current_timeo = *timeo_p;
6415 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6416 asoc, (long)(*timeo_p), msg_len);
6418 /* Increment the association's refcnt. */
6419 sctp_association_hold(asoc);
6421 /* Wait on the association specific sndbuf space. */
6423 prepare_to_wait_exclusive(&asoc->wait, &wait,
6424 TASK_INTERRUPTIBLE);
6427 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6430 if (signal_pending(current))
6431 goto do_interrupted;
6432 if (msg_len <= sctp_wspace(asoc))
6435 /* Let another process have a go. Since we are going
6438 sctp_release_sock(sk);
6439 current_timeo = schedule_timeout(current_timeo);
6440 BUG_ON(sk != asoc->base.sk);
6443 *timeo_p = current_timeo;
6447 finish_wait(&asoc->wait, &wait);
6449 /* Release the association's refcnt. */
6450 sctp_association_put(asoc);
6459 err = sock_intr_errno(*timeo_p);
6467 void sctp_data_ready(struct sock *sk, int len)
6469 struct socket_wq *wq;
6472 wq = rcu_dereference(sk->sk_wq);
6473 if (wq_has_sleeper(wq))
6474 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6475 POLLRDNORM | POLLRDBAND);
6476 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6480 /* If socket sndbuf has changed, wake up all per association waiters. */
6481 void sctp_write_space(struct sock *sk)
6483 struct sctp_association *asoc;
6485 /* Wake up the tasks in each wait queue. */
6486 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6487 __sctp_write_space(asoc);
6491 /* Is there any sndbuf space available on the socket?
6493 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6494 * associations on the same socket. For a UDP-style socket with
6495 * multiple associations, it is possible for it to be "unwriteable"
6496 * prematurely. I assume that this is acceptable because
6497 * a premature "unwriteable" is better than an accidental "writeable" which
6498 * would cause an unwanted block under certain circumstances. For the 1-1
6499 * UDP-style sockets or TCP-style sockets, this code should work.
6502 static int sctp_writeable(struct sock *sk)
6506 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6512 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6513 * returns immediately with EINPROGRESS.
6515 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6517 struct sock *sk = asoc->base.sk;
6519 long current_timeo = *timeo_p;
6522 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6525 /* Increment the association's refcnt. */
6526 sctp_association_hold(asoc);
6529 prepare_to_wait_exclusive(&asoc->wait, &wait,
6530 TASK_INTERRUPTIBLE);
6533 if (sk->sk_shutdown & RCV_SHUTDOWN)
6535 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6538 if (signal_pending(current))
6539 goto do_interrupted;
6541 if (sctp_state(asoc, ESTABLISHED))
6544 /* Let another process have a go. Since we are going
6547 sctp_release_sock(sk);
6548 current_timeo = schedule_timeout(current_timeo);
6551 *timeo_p = current_timeo;
6555 finish_wait(&asoc->wait, &wait);
6557 /* Release the association's refcnt. */
6558 sctp_association_put(asoc);
6563 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6566 err = -ECONNREFUSED;
6570 err = sock_intr_errno(*timeo_p);
6578 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6580 struct sctp_endpoint *ep;
6584 ep = sctp_sk(sk)->ep;
6588 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6589 TASK_INTERRUPTIBLE);
6591 if (list_empty(&ep->asocs)) {
6592 sctp_release_sock(sk);
6593 timeo = schedule_timeout(timeo);
6598 if (!sctp_sstate(sk, LISTENING))
6602 if (!list_empty(&ep->asocs))
6605 err = sock_intr_errno(timeo);
6606 if (signal_pending(current))
6614 finish_wait(sk_sleep(sk), &wait);
6619 static void sctp_wait_for_close(struct sock *sk, long timeout)
6624 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6625 if (list_empty(&sctp_sk(sk)->ep->asocs))
6627 sctp_release_sock(sk);
6628 timeout = schedule_timeout(timeout);
6630 } while (!signal_pending(current) && timeout);
6632 finish_wait(sk_sleep(sk), &wait);
6635 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6637 struct sk_buff *frag;
6642 /* Don't forget the fragments. */
6643 skb_walk_frags(skb, frag)
6644 sctp_skb_set_owner_r_frag(frag, sk);
6647 sctp_skb_set_owner_r(skb, sk);
6650 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6651 struct sctp_association *asoc)
6653 struct inet_sock *inet = inet_sk(sk);
6654 struct inet_sock *newinet;
6656 newsk->sk_type = sk->sk_type;
6657 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6658 newsk->sk_flags = sk->sk_flags;
6659 newsk->sk_no_check = sk->sk_no_check;
6660 newsk->sk_reuse = sk->sk_reuse;
6662 newsk->sk_shutdown = sk->sk_shutdown;
6663 newsk->sk_destruct = inet_sock_destruct;
6664 newsk->sk_family = sk->sk_family;
6665 newsk->sk_protocol = IPPROTO_SCTP;
6666 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6667 newsk->sk_sndbuf = sk->sk_sndbuf;
6668 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6669 newsk->sk_lingertime = sk->sk_lingertime;
6670 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6671 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6673 newinet = inet_sk(newsk);
6675 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6676 * getsockname() and getpeername()
6678 newinet->inet_sport = inet->inet_sport;
6679 newinet->inet_saddr = inet->inet_saddr;
6680 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6681 newinet->inet_dport = htons(asoc->peer.port);
6682 newinet->pmtudisc = inet->pmtudisc;
6683 newinet->inet_id = asoc->next_tsn ^ jiffies;
6685 newinet->uc_ttl = inet->uc_ttl;
6686 newinet->mc_loop = 1;
6687 newinet->mc_ttl = 1;
6688 newinet->mc_index = 0;
6689 newinet->mc_list = NULL;
6692 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6693 * and its messages to the newsk.
6695 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6696 struct sctp_association *assoc,
6697 sctp_socket_type_t type)
6699 struct sctp_sock *oldsp = sctp_sk(oldsk);
6700 struct sctp_sock *newsp = sctp_sk(newsk);
6701 struct sctp_bind_bucket *pp; /* hash list port iterator */
6702 struct sctp_endpoint *newep = newsp->ep;
6703 struct sk_buff *skb, *tmp;
6704 struct sctp_ulpevent *event;
6705 struct sctp_bind_hashbucket *head;
6706 struct list_head tmplist;
6708 /* Migrate socket buffer sizes and all the socket level options to the
6711 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6712 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6713 /* Brute force copy old sctp opt. */
6714 if (oldsp->do_auto_asconf) {
6715 memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
6716 inet_sk_copy_descendant(newsk, oldsk);
6717 memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
6719 inet_sk_copy_descendant(newsk, oldsk);
6721 /* Restore the ep value that was overwritten with the above structure
6727 /* Hook this new socket in to the bind_hash list. */
6728 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->inet_num)];
6729 sctp_local_bh_disable();
6730 sctp_spin_lock(&head->lock);
6731 pp = sctp_sk(oldsk)->bind_hash;
6732 sk_add_bind_node(newsk, &pp->owner);
6733 sctp_sk(newsk)->bind_hash = pp;
6734 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6735 sctp_spin_unlock(&head->lock);
6736 sctp_local_bh_enable();
6738 /* Copy the bind_addr list from the original endpoint to the new
6739 * endpoint so that we can handle restarts properly
6741 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6742 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6744 /* Move any messages in the old socket's receive queue that are for the
6745 * peeled off association to the new socket's receive queue.
6747 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6748 event = sctp_skb2event(skb);
6749 if (event->asoc == assoc) {
6750 __skb_unlink(skb, &oldsk->sk_receive_queue);
6751 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6752 sctp_skb_set_owner_r_frag(skb, newsk);
6756 /* Clean up any messages pending delivery due to partial
6757 * delivery. Three cases:
6758 * 1) No partial deliver; no work.
6759 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6760 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6762 skb_queue_head_init(&newsp->pd_lobby);
6763 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6765 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6766 struct sk_buff_head *queue;
6768 /* Decide which queue to move pd_lobby skbs to. */
6769 if (assoc->ulpq.pd_mode) {
6770 queue = &newsp->pd_lobby;
6772 queue = &newsk->sk_receive_queue;
6774 /* Walk through the pd_lobby, looking for skbs that
6775 * need moved to the new socket.
6777 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6778 event = sctp_skb2event(skb);
6779 if (event->asoc == assoc) {
6780 __skb_unlink(skb, &oldsp->pd_lobby);
6781 __skb_queue_tail(queue, skb);
6782 sctp_skb_set_owner_r_frag(skb, newsk);
6786 /* Clear up any skbs waiting for the partial
6787 * delivery to finish.
6789 if (assoc->ulpq.pd_mode)
6790 sctp_clear_pd(oldsk, NULL);
6794 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6795 sctp_skb_set_owner_r_frag(skb, newsk);
6797 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6798 sctp_skb_set_owner_r_frag(skb, newsk);
6800 /* Set the type of socket to indicate that it is peeled off from the
6801 * original UDP-style socket or created with the accept() call on a
6802 * TCP-style socket..
6806 /* Mark the new socket "in-use" by the user so that any packets
6807 * that may arrive on the association after we've moved it are
6808 * queued to the backlog. This prevents a potential race between
6809 * backlog processing on the old socket and new-packet processing
6810 * on the new socket.
6812 * The caller has just allocated newsk so we can guarantee that other
6813 * paths won't try to lock it and then oldsk.
6815 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6816 sctp_assoc_migrate(assoc, newsk);
6818 /* If the association on the newsk is already closed before accept()
6819 * is called, set RCV_SHUTDOWN flag.
6821 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6822 newsk->sk_shutdown |= RCV_SHUTDOWN;
6824 newsk->sk_state = SCTP_SS_ESTABLISHED;
6825 sctp_release_sock(newsk);
6829 /* This proto struct describes the ULP interface for SCTP. */
6830 struct proto sctp_prot = {
6832 .owner = THIS_MODULE,
6833 .close = sctp_close,
6834 .connect = sctp_connect,
6835 .disconnect = sctp_disconnect,
6836 .accept = sctp_accept,
6837 .ioctl = sctp_ioctl,
6838 .init = sctp_init_sock,
6839 .destroy = sctp_destroy_sock,
6840 .shutdown = sctp_shutdown,
6841 .setsockopt = sctp_setsockopt,
6842 .getsockopt = sctp_getsockopt,
6843 .sendmsg = sctp_sendmsg,
6844 .recvmsg = sctp_recvmsg,
6846 .backlog_rcv = sctp_backlog_rcv,
6848 .unhash = sctp_unhash,
6849 .get_port = sctp_get_port,
6850 .obj_size = sizeof(struct sctp_sock),
6851 .sysctl_mem = sysctl_sctp_mem,
6852 .sysctl_rmem = sysctl_sctp_rmem,
6853 .sysctl_wmem = sysctl_sctp_wmem,
6854 .memory_pressure = &sctp_memory_pressure,
6855 .enter_memory_pressure = sctp_enter_memory_pressure,
6856 .memory_allocated = &sctp_memory_allocated,
6857 .sockets_allocated = &sctp_sockets_allocated,
6860 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6862 struct proto sctpv6_prot = {
6864 .owner = THIS_MODULE,
6865 .close = sctp_close,
6866 .connect = sctp_connect,
6867 .disconnect = sctp_disconnect,
6868 .accept = sctp_accept,
6869 .ioctl = sctp_ioctl,
6870 .init = sctp_init_sock,
6871 .destroy = sctp_destroy_sock,
6872 .shutdown = sctp_shutdown,
6873 .setsockopt = sctp_setsockopt,
6874 .getsockopt = sctp_getsockopt,
6875 .sendmsg = sctp_sendmsg,
6876 .recvmsg = sctp_recvmsg,
6878 .backlog_rcv = sctp_backlog_rcv,
6880 .unhash = sctp_unhash,
6881 .get_port = sctp_get_port,
6882 .obj_size = sizeof(struct sctp6_sock),
6883 .sysctl_mem = sysctl_sctp_mem,
6884 .sysctl_rmem = sysctl_sctp_rmem,
6885 .sysctl_wmem = sysctl_sctp_wmem,
6886 .memory_pressure = &sctp_memory_pressure,
6887 .enter_memory_pressure = sctp_enter_memory_pressure,
6888 .memory_allocated = &sctp_memory_allocated,
6889 .sockets_allocated = &sctp_sockets_allocated,
6891 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */