1 /* SCTP kernel implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
9 * This file is part of the SCTP kernel implementation
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
18 * This SCTP implementation is free software;
19 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
24 * This SCTP implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, write to
32 * the Free Software Foundation, 59 Temple Place - Suite 330,
33 * Boston, MA 02111-1307, USA.
35 * Please send any bug reports or fixes you make to the
37 * lksctp developers <lksctp-developers@lists.sourceforge.net>
39 * Or submit a bug report through the following website:
40 * http://www.sf.net/projects/lksctp
42 * Written or modified by:
43 * La Monte H.P. Yarroll <piggy@acm.org>
44 * Narasimha Budihal <narsi@refcode.org>
45 * Karl Knutson <karl@athena.chicago.il.us>
46 * Jon Grimm <jgrimm@us.ibm.com>
47 * Xingang Guo <xingang.guo@intel.com>
48 * Daisy Chang <daisyc@us.ibm.com>
49 * Sridhar Samudrala <samudrala@us.ibm.com>
50 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
51 * Ardelle Fan <ardelle.fan@intel.com>
52 * Ryan Layer <rmlayer@us.ibm.com>
53 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
54 * Kevin Gao <kevin.gao@intel.com>
56 * Any bugs reported given to us we will try to fix... any fixes shared will
57 * be incorporated into the next SCTP release.
60 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
62 #include <linux/types.h>
63 #include <linux/kernel.h>
64 #include <linux/wait.h>
65 #include <linux/time.h>
67 #include <linux/capability.h>
68 #include <linux/fcntl.h>
69 #include <linux/poll.h>
70 #include <linux/init.h>
71 #include <linux/crypto.h>
72 #include <linux/slab.h>
73 #include <linux/compat.h>
74 #include <linux/nsproxy.h>
78 #include <net/route.h>
80 #include <net/inet_common.h>
82 #include <linux/socket.h> /* for sa_family_t */
83 #include <linux/export.h>
85 #include <net/sctp/sctp.h>
86 #include <net/sctp/sm.h>
88 /* WARNING: Please do not remove the SCTP_STATIC attribute to
89 * any of the functions below as they are used to export functions
90 * used by a project regression testsuite.
93 /* Forward declarations for internal helper functions. */
94 static int sctp_writeable(struct sock *sk);
95 static void sctp_wfree(struct sk_buff *skb);
96 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
98 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
99 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
100 static int sctp_wait_for_accept(struct sock *sk, long timeo);
101 static void sctp_wait_for_close(struct sock *sk, long timeo);
102 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
103 union sctp_addr *addr, int len);
104 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
105 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
106 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
107 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
108 static int sctp_send_asconf(struct sctp_association *asoc,
109 struct sctp_chunk *chunk);
110 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
111 static int sctp_autobind(struct sock *sk);
112 static void sctp_sock_migrate(struct sock *, struct sock *,
113 struct sctp_association *, sctp_socket_type_t);
114 static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
116 extern struct kmem_cache *sctp_bucket_cachep;
117 extern long sysctl_sctp_mem[3];
118 extern int sysctl_sctp_rmem[3];
119 extern int sysctl_sctp_wmem[3];
121 static int sctp_memory_pressure;
122 static atomic_long_t sctp_memory_allocated;
123 struct percpu_counter sctp_sockets_allocated;
125 static void sctp_enter_memory_pressure(struct sock *sk)
127 sctp_memory_pressure = 1;
131 /* Get the sndbuf space available at the time on the association. */
132 static inline int sctp_wspace(struct sctp_association *asoc)
136 if (asoc->ep->sndbuf_policy)
137 amt = asoc->sndbuf_used;
139 amt = sk_wmem_alloc_get(asoc->base.sk);
141 if (amt >= asoc->base.sk->sk_sndbuf) {
142 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
145 amt = sk_stream_wspace(asoc->base.sk);
150 amt = asoc->base.sk->sk_sndbuf - amt;
155 /* Increment the used sndbuf space count of the corresponding association by
156 * the size of the outgoing data chunk.
157 * Also, set the skb destructor for sndbuf accounting later.
159 * Since it is always 1-1 between chunk and skb, and also a new skb is always
160 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
161 * destructor in the data chunk skb for the purpose of the sndbuf space
164 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
166 struct sctp_association *asoc = chunk->asoc;
167 struct sock *sk = asoc->base.sk;
169 /* The sndbuf space is tracked per association. */
170 sctp_association_hold(asoc);
172 skb_set_owner_w(chunk->skb, sk);
174 chunk->skb->destructor = sctp_wfree;
175 /* Save the chunk pointer in skb for sctp_wfree to use later. */
176 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
178 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
179 sizeof(struct sk_buff) +
180 sizeof(struct sctp_chunk);
182 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
183 sk->sk_wmem_queued += chunk->skb->truesize;
184 sk_mem_charge(sk, chunk->skb->truesize);
187 /* Verify that this is a valid address. */
188 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
193 /* Verify basic sockaddr. */
194 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
198 /* Is this a valid SCTP address? */
199 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
202 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
208 /* Look up the association by its id. If this is not a UDP-style
209 * socket, the ID field is always ignored.
211 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
213 struct sctp_association *asoc = NULL;
215 /* If this is not a UDP-style socket, assoc id should be ignored. */
216 if (!sctp_style(sk, UDP)) {
217 /* Return NULL if the socket state is not ESTABLISHED. It
218 * could be a TCP-style listening socket or a socket which
219 * hasn't yet called connect() to establish an association.
221 if (!sctp_sstate(sk, ESTABLISHED))
224 /* Get the first and the only association from the list. */
225 if (!list_empty(&sctp_sk(sk)->ep->asocs))
226 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
227 struct sctp_association, asocs);
231 /* Otherwise this is a UDP-style socket. */
232 if (!id || (id == (sctp_assoc_t)-1))
235 spin_lock_bh(&sctp_assocs_id_lock);
236 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
237 spin_unlock_bh(&sctp_assocs_id_lock);
239 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
245 /* Look up the transport from an address and an assoc id. If both address and
246 * id are specified, the associations matching the address and the id should be
249 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
250 struct sockaddr_storage *addr,
253 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
254 struct sctp_transport *transport;
255 union sctp_addr *laddr = (union sctp_addr *)addr;
257 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
264 id_asoc = sctp_id2assoc(sk, id);
265 if (id_asoc && (id_asoc != addr_asoc))
268 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
269 (union sctp_addr *)addr);
274 /* API 3.1.2 bind() - UDP Style Syntax
275 * The syntax of bind() is,
277 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
279 * sd - the socket descriptor returned by socket().
280 * addr - the address structure (struct sockaddr_in or struct
281 * sockaddr_in6 [RFC 2553]),
282 * addr_len - the size of the address structure.
284 SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
290 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
293 /* Disallow binding twice. */
294 if (!sctp_sk(sk)->ep->base.bind_addr.port)
295 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
300 sctp_release_sock(sk);
305 static long sctp_get_port_local(struct sock *, union sctp_addr *);
307 /* Verify this is a valid sockaddr. */
308 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
309 union sctp_addr *addr, int len)
313 /* Check minimum size. */
314 if (len < sizeof (struct sockaddr))
317 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
320 /* V4 mapped address are really of AF_INET family */
321 if (addr->sa.sa_family == AF_INET6 &&
322 ipv6_addr_v4mapped(&addr->v6.sin6_addr) &&
323 !opt->pf->af_supported(AF_INET, 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 sp->pf->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;
1067 union sctp_addr *sa_addr = NULL;
1069 unsigned short port;
1070 unsigned int f_flags = 0;
1075 /* connect() cannot be done on a socket that is already in ESTABLISHED
1076 * state - UDP-style peeled off socket or a TCP-style socket that
1077 * is already connected.
1078 * It cannot be done even on a TCP-style listening socket.
1080 if (sctp_sstate(sk, ESTABLISHED) ||
1081 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1086 /* Walk through the addrs buffer and count the number of addresses. */
1088 while (walk_size < addrs_size) {
1091 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1097 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1099 /* If the address family is not supported or if this address
1100 * causes the address buffer to overflow return EINVAL.
1102 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1107 port = ntohs(sa_addr->v4.sin_port);
1109 /* Save current address so we can work with it */
1110 memcpy(&to, sa_addr, af->sockaddr_len);
1112 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1116 /* Make sure the destination port is correctly set
1119 if (asoc && asoc->peer.port && asoc->peer.port != port)
1123 /* Check if there already is a matching association on the
1124 * endpoint (other than the one created here).
1126 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1127 if (asoc2 && asoc2 != asoc) {
1128 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1135 /* If we could not find a matching association on the endpoint,
1136 * make sure that there is no peeled-off association matching
1137 * the peer address even on another socket.
1139 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1140 err = -EADDRNOTAVAIL;
1145 /* If a bind() or sctp_bindx() is not called prior to
1146 * an sctp_connectx() call, the system picks an
1147 * ephemeral port and will choose an address set
1148 * equivalent to binding with a wildcard address.
1150 if (!ep->base.bind_addr.port) {
1151 if (sctp_autobind(sk)) {
1157 * If an unprivileged user inherits a 1-many
1158 * style socket with open associations on a
1159 * privileged port, it MAY be permitted to
1160 * accept new associations, but it SHOULD NOT
1161 * be permitted to open new associations.
1163 if (ep->base.bind_addr.port < PROT_SOCK &&
1164 !capable(CAP_NET_BIND_SERVICE)) {
1170 scope = sctp_scope(&to);
1171 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1177 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1185 /* Prime the peer's transport structures. */
1186 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1194 addr_buf += af->sockaddr_len;
1195 walk_size += af->sockaddr_len;
1198 /* In case the user of sctp_connectx() wants an association
1199 * id back, assign one now.
1202 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1207 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1212 /* Initialize sk's dport and daddr for getpeername() */
1213 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1214 sp->pf->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);
1226 *assoc_id = asoc->assoc_id;
1227 err = sctp_wait_for_connect(asoc, &timeo);
1228 /* Note: the asoc may be freed after the return of
1229 * sctp_wait_for_connect.
1232 /* Don't free association on exit. */
1237 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1238 " kaddrs: %p err: %d\n",
1241 /* sctp_primitive_ASSOCIATE may have added this association
1242 * To the hash table, try to unhash it, just in case, its a noop
1243 * if it wasn't hashed so we're safe
1245 sctp_unhash_established(asoc);
1246 sctp_association_free(asoc);
1251 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1254 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1255 * sctp_assoc_t *asoc);
1257 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1258 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1259 * or IPv6 addresses.
1261 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1262 * Section 3.1.2 for this usage.
1264 * addrs is a pointer to an array of one or more socket addresses. Each
1265 * address is contained in its appropriate structure (i.e. struct
1266 * sockaddr_in or struct sockaddr_in6) the family of the address type
1267 * must be used to distengish the address length (note that this
1268 * representation is termed a "packed array" of addresses). The caller
1269 * specifies the number of addresses in the array with addrcnt.
1271 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1272 * the association id of the new association. On failure, sctp_connectx()
1273 * returns -1, and sets errno to the appropriate error code. The assoc_id
1274 * is not touched by the kernel.
1276 * For SCTP, the port given in each socket address must be the same, or
1277 * sctp_connectx() will fail, setting errno to EINVAL.
1279 * An application can use sctp_connectx to initiate an association with
1280 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1281 * allows a caller to specify multiple addresses at which a peer can be
1282 * reached. The way the SCTP stack uses the list of addresses to set up
1283 * the association is implementation dependent. This function only
1284 * specifies that the stack will try to make use of all the addresses in
1285 * the list when needed.
1287 * Note that the list of addresses passed in is only used for setting up
1288 * the association. It does not necessarily equal the set of addresses
1289 * the peer uses for the resulting association. If the caller wants to
1290 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1291 * retrieve them after the association has been set up.
1293 * Basically do nothing but copying the addresses from user to kernel
1294 * land and invoking either sctp_connectx(). This is used for tunneling
1295 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1297 * We don't use copy_from_user() for optimization: we first do the
1298 * sanity checks (buffer size -fast- and access check-healthy
1299 * pointer); if all of those succeed, then we can alloc the memory
1300 * (expensive operation) needed to copy the data to kernel. Then we do
1301 * the copying without checking the user space area
1302 * (__copy_from_user()).
1304 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1307 * sk The sk of the socket
1308 * addrs The pointer to the addresses in user land
1309 * addrssize Size of the addrs buffer
1311 * Returns >=0 if ok, <0 errno code on error.
1313 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1314 struct sockaddr __user *addrs,
1316 sctp_assoc_t *assoc_id)
1319 struct sockaddr *kaddrs;
1321 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1322 __func__, sk, addrs, addrs_size);
1324 if (unlikely(addrs_size <= 0))
1327 /* Check the user passed a healthy pointer. */
1328 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1331 /* Alloc space for the address array in kernel memory. */
1332 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1333 if (unlikely(!kaddrs))
1336 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1339 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1348 * This is an older interface. It's kept for backward compatibility
1349 * to the option that doesn't provide association id.
1351 SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1352 struct sockaddr __user *addrs,
1355 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1359 * New interface for the API. The since the API is done with a socket
1360 * option, to make it simple we feed back the association id is as a return
1361 * indication to the call. Error is always negative and association id is
1364 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1365 struct sockaddr __user *addrs,
1368 sctp_assoc_t assoc_id = 0;
1371 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1380 * New (hopefully final) interface for the API.
1381 * We use the sctp_getaddrs_old structure so that use-space library
1382 * can avoid any unnecessary allocations. The only different part
1383 * is that we store the actual length of the address buffer into the
1384 * addrs_num structure member. That way we can re-use the existing
1387 #ifdef CONFIG_COMPAT
1388 struct compat_sctp_getaddrs_old {
1389 sctp_assoc_t assoc_id;
1391 compat_uptr_t addrs; /* struct sockaddr * */
1395 SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1396 char __user *optval,
1399 struct sctp_getaddrs_old param;
1400 sctp_assoc_t assoc_id = 0;
1403 #ifdef CONFIG_COMPAT
1404 if (is_compat_task()) {
1405 struct compat_sctp_getaddrs_old param32;
1407 if (len < sizeof(param32))
1409 if (copy_from_user(¶m32, optval, sizeof(param32)))
1412 param.assoc_id = param32.assoc_id;
1413 param.addr_num = param32.addr_num;
1414 param.addrs = compat_ptr(param32.addrs);
1418 if (len < sizeof(param))
1420 if (copy_from_user(¶m, optval, sizeof(param)))
1424 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1425 param.addrs, param.addr_num,
1427 if (err == 0 || err == -EINPROGRESS) {
1428 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1430 if (put_user(sizeof(assoc_id), optlen))
1437 /* API 3.1.4 close() - UDP Style Syntax
1438 * Applications use close() to perform graceful shutdown (as described in
1439 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1440 * by a UDP-style socket.
1444 * ret = close(int sd);
1446 * sd - the socket descriptor of the associations to be closed.
1448 * To gracefully shutdown a specific association represented by the
1449 * UDP-style socket, an application should use the sendmsg() call,
1450 * passing no user data, but including the appropriate flag in the
1451 * ancillary data (see Section xxxx).
1453 * If sd in the close() call is a branched-off socket representing only
1454 * one association, the shutdown is performed on that association only.
1456 * 4.1.6 close() - TCP Style Syntax
1458 * Applications use close() to gracefully close down an association.
1462 * int close(int sd);
1464 * sd - the socket descriptor of the association to be closed.
1466 * After an application calls close() on a socket descriptor, no further
1467 * socket operations will succeed on that descriptor.
1469 * API 7.1.4 SO_LINGER
1471 * An application using the TCP-style socket can use this option to
1472 * perform the SCTP ABORT primitive. The linger option structure is:
1475 * int l_onoff; // option on/off
1476 * int l_linger; // linger time
1479 * To enable the option, set l_onoff to 1. If the l_linger value is set
1480 * to 0, calling close() is the same as the ABORT primitive. If the
1481 * value is set to a negative value, the setsockopt() call will return
1482 * an error. If the value is set to a positive value linger_time, the
1483 * close() can be blocked for at most linger_time ms. If the graceful
1484 * shutdown phase does not finish during this period, close() will
1485 * return but the graceful shutdown phase continues in the system.
1487 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1489 struct sctp_endpoint *ep;
1490 struct sctp_association *asoc;
1491 struct list_head *pos, *temp;
1492 unsigned int data_was_unread;
1494 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1497 sk->sk_shutdown = SHUTDOWN_MASK;
1498 sk->sk_state = SCTP_SS_CLOSING;
1500 ep = sctp_sk(sk)->ep;
1502 /* Clean up any skbs sitting on the receive queue. */
1503 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1504 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1506 /* Walk all associations on an endpoint. */
1507 list_for_each_safe(pos, temp, &ep->asocs) {
1508 asoc = list_entry(pos, struct sctp_association, asocs);
1510 if (sctp_style(sk, TCP)) {
1511 /* A closed association can still be in the list if
1512 * it belongs to a TCP-style listening socket that is
1513 * not yet accepted. If so, free it. If not, send an
1514 * ABORT or SHUTDOWN based on the linger options.
1516 if (sctp_state(asoc, CLOSED)) {
1517 sctp_unhash_established(asoc);
1518 sctp_association_free(asoc);
1523 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1524 !skb_queue_empty(&asoc->ulpq.reasm) ||
1525 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1526 struct sctp_chunk *chunk;
1528 chunk = sctp_make_abort_user(asoc, NULL, 0);
1529 sctp_primitive_ABORT(asoc, chunk);
1531 sctp_primitive_SHUTDOWN(asoc, NULL);
1534 /* On a TCP-style socket, block for at most linger_time if set. */
1535 if (sctp_style(sk, TCP) && timeout)
1536 sctp_wait_for_close(sk, timeout);
1538 /* This will run the backlog queue. */
1539 sctp_release_sock(sk);
1541 /* Supposedly, no process has access to the socket, but
1542 * the net layers still may.
1543 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1544 * held and that should be grabbed before socket lock.
1546 spin_lock_bh(&sctp_globals.addr_wq_lock);
1547 sctp_bh_lock_sock(sk);
1549 /* Hold the sock, since sk_common_release() will put sock_put()
1550 * and we have just a little more cleanup.
1553 sk_common_release(sk);
1555 sctp_bh_unlock_sock(sk);
1556 spin_unlock_bh(&sctp_globals.addr_wq_lock);
1560 SCTP_DBG_OBJCNT_DEC(sock);
1563 /* Handle EPIPE error. */
1564 static int sctp_error(struct sock *sk, int flags, int err)
1567 err = sock_error(sk) ? : -EPIPE;
1568 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1569 send_sig(SIGPIPE, current, 0);
1573 /* API 3.1.3 sendmsg() - UDP Style Syntax
1575 * An application uses sendmsg() and recvmsg() calls to transmit data to
1576 * and receive data from its peer.
1578 * ssize_t sendmsg(int socket, const struct msghdr *message,
1581 * socket - the socket descriptor of the endpoint.
1582 * message - pointer to the msghdr structure which contains a single
1583 * user message and possibly some ancillary data.
1585 * See Section 5 for complete description of the data
1588 * flags - flags sent or received with the user message, see Section
1589 * 5 for complete description of the flags.
1591 * Note: This function could use a rewrite especially when explicit
1592 * connect support comes in.
1594 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1596 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1598 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1599 struct msghdr *msg, size_t msg_len)
1601 struct sctp_sock *sp;
1602 struct sctp_endpoint *ep;
1603 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1604 struct sctp_transport *transport, *chunk_tp;
1605 struct sctp_chunk *chunk;
1607 struct sockaddr *msg_name = NULL;
1608 struct sctp_sndrcvinfo default_sinfo;
1609 struct sctp_sndrcvinfo *sinfo;
1610 struct sctp_initmsg *sinit;
1611 sctp_assoc_t associd = 0;
1612 sctp_cmsgs_t cmsgs = { NULL };
1616 __u16 sinfo_flags = 0;
1617 bool wait_connect = false;
1618 struct sctp_datamsg *datamsg;
1619 int msg_flags = msg->msg_flags;
1621 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1628 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1630 /* We cannot send a message over a TCP-style listening socket. */
1631 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1636 /* Parse out the SCTP CMSGs. */
1637 err = sctp_msghdr_parse(msg, &cmsgs);
1640 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1644 /* Fetch the destination address for this packet. This
1645 * address only selects the association--it is not necessarily
1646 * the address we will send to.
1647 * For a peeled-off socket, msg_name is ignored.
1649 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1650 int msg_namelen = msg->msg_namelen;
1652 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1657 if (msg_namelen > sizeof(to))
1658 msg_namelen = sizeof(to);
1659 memcpy(&to, msg->msg_name, msg_namelen);
1660 msg_name = msg->msg_name;
1666 /* Did the user specify SNDRCVINFO? */
1668 sinfo_flags = sinfo->sinfo_flags;
1669 associd = sinfo->sinfo_assoc_id;
1672 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1673 msg_len, sinfo_flags);
1675 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1676 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1681 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1682 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1683 * If SCTP_ABORT is set, the message length could be non zero with
1684 * the msg_iov set to the user abort reason.
1686 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1687 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1692 /* If SCTP_ADDR_OVER is set, there must be an address
1693 * specified in msg_name.
1695 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1702 SCTP_DEBUG_PRINTK("About to look up association.\n");
1706 /* If a msg_name has been specified, assume this is to be used. */
1708 /* Look for a matching association on the endpoint. */
1709 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1711 /* If we could not find a matching association on the
1712 * endpoint, make sure that it is not a TCP-style
1713 * socket that already has an association or there is
1714 * no peeled-off association on another socket.
1716 if ((sctp_style(sk, TCP) &&
1717 sctp_sstate(sk, ESTABLISHED)) ||
1718 sctp_endpoint_is_peeled_off(ep, &to)) {
1719 err = -EADDRNOTAVAIL;
1724 asoc = sctp_id2assoc(sk, associd);
1732 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1734 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1735 * socket that has an association in CLOSED state. This can
1736 * happen when an accepted socket has an association that is
1739 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1744 if (sinfo_flags & SCTP_EOF) {
1745 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1747 sctp_primitive_SHUTDOWN(asoc, NULL);
1751 if (sinfo_flags & SCTP_ABORT) {
1753 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1759 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1760 sctp_primitive_ABORT(asoc, chunk);
1766 /* Do we need to create the association? */
1768 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1770 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1775 /* Check for invalid stream against the stream counts,
1776 * either the default or the user specified stream counts.
1779 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1780 /* Check against the defaults. */
1781 if (sinfo->sinfo_stream >=
1782 sp->initmsg.sinit_num_ostreams) {
1787 /* Check against the requested. */
1788 if (sinfo->sinfo_stream >=
1789 sinit->sinit_num_ostreams) {
1797 * API 3.1.2 bind() - UDP Style Syntax
1798 * If a bind() or sctp_bindx() is not called prior to a
1799 * sendmsg() call that initiates a new association, the
1800 * system picks an ephemeral port and will choose an address
1801 * set equivalent to binding with a wildcard address.
1803 if (!ep->base.bind_addr.port) {
1804 if (sctp_autobind(sk)) {
1810 * If an unprivileged user inherits a one-to-many
1811 * style socket with open associations on a privileged
1812 * port, it MAY be permitted to accept new associations,
1813 * but it SHOULD NOT be permitted to open new
1816 if (ep->base.bind_addr.port < PROT_SOCK &&
1817 !capable(CAP_NET_BIND_SERVICE)) {
1823 scope = sctp_scope(&to);
1824 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1830 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1836 /* If the SCTP_INIT ancillary data is specified, set all
1837 * the association init values accordingly.
1840 if (sinit->sinit_num_ostreams) {
1841 asoc->c.sinit_num_ostreams =
1842 sinit->sinit_num_ostreams;
1844 if (sinit->sinit_max_instreams) {
1845 asoc->c.sinit_max_instreams =
1846 sinit->sinit_max_instreams;
1848 if (sinit->sinit_max_attempts) {
1849 asoc->max_init_attempts
1850 = sinit->sinit_max_attempts;
1852 if (sinit->sinit_max_init_timeo) {
1853 asoc->max_init_timeo =
1854 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1858 /* Prime the peer's transport structures. */
1859 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1866 /* ASSERT: we have a valid association at this point. */
1867 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1870 /* If the user didn't specify SNDRCVINFO, make up one with
1873 memset(&default_sinfo, 0, sizeof(default_sinfo));
1874 default_sinfo.sinfo_stream = asoc->default_stream;
1875 default_sinfo.sinfo_flags = asoc->default_flags;
1876 default_sinfo.sinfo_ppid = asoc->default_ppid;
1877 default_sinfo.sinfo_context = asoc->default_context;
1878 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1879 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1880 sinfo = &default_sinfo;
1883 /* API 7.1.7, the sndbuf size per association bounds the
1884 * maximum size of data that can be sent in a single send call.
1886 if (msg_len > sk->sk_sndbuf) {
1891 if (asoc->pmtu_pending)
1892 sctp_assoc_pending_pmtu(asoc);
1894 /* If fragmentation is disabled and the message length exceeds the
1895 * association fragmentation point, return EMSGSIZE. The I-D
1896 * does not specify what this error is, but this looks like
1899 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1904 /* Check for invalid stream. */
1905 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1910 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1911 if (!sctp_wspace(asoc)) {
1912 /* sk can be changed by peel off when waiting for buf. */
1913 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1918 /* If an address is passed with the sendto/sendmsg call, it is used
1919 * to override the primary destination address in the TCP model, or
1920 * when SCTP_ADDR_OVER flag is set in the UDP model.
1922 if ((sctp_style(sk, TCP) && msg_name) ||
1923 (sinfo_flags & SCTP_ADDR_OVER)) {
1924 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1932 /* Auto-connect, if we aren't connected already. */
1933 if (sctp_state(asoc, CLOSED)) {
1934 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1937 wait_connect = true;
1938 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1941 /* Break the message into multiple chunks of maximum size. */
1942 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1943 if (IS_ERR(datamsg)) {
1944 err = PTR_ERR(datamsg);
1948 /* Now send the (possibly) fragmented message. */
1949 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1950 sctp_chunk_hold(chunk);
1952 /* Do accounting for the write space. */
1953 sctp_set_owner_w(chunk);
1955 chunk->transport = chunk_tp;
1958 /* Send it to the lower layers. Note: all chunks
1959 * must either fail or succeed. The lower layer
1960 * works that way today. Keep it that way or this
1963 err = sctp_primitive_SEND(asoc, datamsg);
1964 /* Did the lower layer accept the chunk? */
1966 sctp_datamsg_free(datamsg);
1968 sctp_datamsg_put(datamsg);
1970 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1977 if (unlikely(wait_connect)) {
1978 timeo = sock_sndtimeo(sk, msg_flags & MSG_DONTWAIT);
1979 sctp_wait_for_connect(asoc, &timeo);
1982 /* If we are already past ASSOCIATE, the lower
1983 * layers are responsible for association cleanup.
1989 sctp_unhash_established(asoc);
1990 sctp_association_free(asoc);
1993 sctp_release_sock(sk);
1996 return sctp_error(sk, msg_flags, err);
2003 err = sock_error(sk);
2013 /* This is an extended version of skb_pull() that removes the data from the
2014 * start of a skb even when data is spread across the list of skb's in the
2015 * frag_list. len specifies the total amount of data that needs to be removed.
2016 * when 'len' bytes could be removed from the skb, it returns 0.
2017 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2018 * could not be removed.
2020 static int sctp_skb_pull(struct sk_buff *skb, int len)
2022 struct sk_buff *list;
2023 int skb_len = skb_headlen(skb);
2026 if (len <= skb_len) {
2027 __skb_pull(skb, len);
2031 __skb_pull(skb, skb_len);
2033 skb_walk_frags(skb, list) {
2034 rlen = sctp_skb_pull(list, len);
2035 skb->len -= (len-rlen);
2036 skb->data_len -= (len-rlen);
2047 /* API 3.1.3 recvmsg() - UDP Style Syntax
2049 * ssize_t recvmsg(int socket, struct msghdr *message,
2052 * socket - the socket descriptor of the endpoint.
2053 * message - pointer to the msghdr structure which contains a single
2054 * user message and possibly some ancillary data.
2056 * See Section 5 for complete description of the data
2059 * flags - flags sent or received with the user message, see Section
2060 * 5 for complete description of the flags.
2062 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2064 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2065 struct msghdr *msg, size_t len, int noblock,
2066 int flags, int *addr_len)
2068 struct sctp_ulpevent *event = NULL;
2069 struct sctp_sock *sp = sctp_sk(sk);
2070 struct sk_buff *skb;
2075 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
2076 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
2077 "len", len, "knoblauch", noblock,
2078 "flags", flags, "addr_len", addr_len);
2082 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2087 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2091 /* Get the total length of the skb including any skb's in the
2100 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2102 event = sctp_skb2event(skb);
2107 sock_recv_ts_and_drops(msg, sk, skb);
2108 if (sctp_ulpevent_is_notification(event)) {
2109 msg->msg_flags |= MSG_NOTIFICATION;
2110 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2112 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2115 /* Check if we allow SCTP_SNDRCVINFO. */
2116 if (sp->subscribe.sctp_data_io_event)
2117 sctp_ulpevent_read_sndrcvinfo(event, msg);
2119 /* FIXME: we should be calling IP/IPv6 layers. */
2120 if (sk->sk_protinfo.af_inet.cmsg_flags)
2121 ip_cmsg_recv(msg, skb);
2126 /* If skb's length exceeds the user's buffer, update the skb and
2127 * push it back to the receive_queue so that the next call to
2128 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2130 if (skb_len > copied) {
2131 msg->msg_flags &= ~MSG_EOR;
2132 if (flags & MSG_PEEK)
2134 sctp_skb_pull(skb, copied);
2135 skb_queue_head(&sk->sk_receive_queue, skb);
2137 /* When only partial message is copied to the user, increase
2138 * rwnd by that amount. If all the data in the skb is read,
2139 * rwnd is updated when the event is freed.
2141 if (!sctp_ulpevent_is_notification(event))
2142 sctp_assoc_rwnd_increase(event->asoc, copied);
2144 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2145 (event->msg_flags & MSG_EOR))
2146 msg->msg_flags |= MSG_EOR;
2148 msg->msg_flags &= ~MSG_EOR;
2151 if (flags & MSG_PEEK) {
2152 /* Release the skb reference acquired after peeking the skb in
2153 * sctp_skb_recv_datagram().
2157 /* Free the event which includes releasing the reference to
2158 * the owner of the skb, freeing the skb and updating the
2161 sctp_ulpevent_free(event);
2164 sctp_release_sock(sk);
2168 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2170 * This option is a on/off flag. If enabled no SCTP message
2171 * fragmentation will be performed. Instead if a message being sent
2172 * exceeds the current PMTU size, the message will NOT be sent and
2173 * instead a error will be indicated to the user.
2175 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2176 char __user *optval,
2177 unsigned int optlen)
2181 if (optlen < sizeof(int))
2184 if (get_user(val, (int __user *)optval))
2187 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2192 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2193 unsigned int optlen)
2195 struct sctp_association *asoc;
2196 struct sctp_ulpevent *event;
2198 if (optlen > sizeof(struct sctp_event_subscribe))
2200 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2204 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2205 * if there is no data to be sent or retransmit, the stack will
2206 * immediately send up this notification.
2208 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2209 &sctp_sk(sk)->subscribe)) {
2210 asoc = sctp_id2assoc(sk, 0);
2212 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2213 event = sctp_ulpevent_make_sender_dry_event(asoc,
2218 sctp_ulpq_tail_event(&asoc->ulpq, event);
2225 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2227 * This socket option is applicable to the UDP-style socket only. When
2228 * set it will cause associations that are idle for more than the
2229 * specified number of seconds to automatically close. An association
2230 * being idle is defined an association that has NOT sent or received
2231 * user data. The special value of '0' indicates that no automatic
2232 * close of any associations should be performed. The option expects an
2233 * integer defining the number of seconds of idle time before an
2234 * association is closed.
2236 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2237 unsigned int optlen)
2239 struct sctp_sock *sp = sctp_sk(sk);
2241 /* Applicable to UDP-style socket only */
2242 if (sctp_style(sk, TCP))
2244 if (optlen != sizeof(int))
2246 if (copy_from_user(&sp->autoclose, optval, optlen))
2252 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2254 * Applications can enable or disable heartbeats for any peer address of
2255 * an association, modify an address's heartbeat interval, force a
2256 * heartbeat to be sent immediately, and adjust the address's maximum
2257 * number of retransmissions sent before an address is considered
2258 * unreachable. The following structure is used to access and modify an
2259 * address's parameters:
2261 * struct sctp_paddrparams {
2262 * sctp_assoc_t spp_assoc_id;
2263 * struct sockaddr_storage spp_address;
2264 * uint32_t spp_hbinterval;
2265 * uint16_t spp_pathmaxrxt;
2266 * uint32_t spp_pathmtu;
2267 * uint32_t spp_sackdelay;
2268 * uint32_t spp_flags;
2271 * spp_assoc_id - (one-to-many style socket) This is filled in the
2272 * application, and identifies the association for
2274 * spp_address - This specifies which address is of interest.
2275 * spp_hbinterval - This contains the value of the heartbeat interval,
2276 * in milliseconds. If a value of zero
2277 * is present in this field then no changes are to
2278 * be made to this parameter.
2279 * spp_pathmaxrxt - This contains the maximum number of
2280 * retransmissions before this address shall be
2281 * considered unreachable. If a value of zero
2282 * is present in this field then no changes are to
2283 * be made to this parameter.
2284 * spp_pathmtu - When Path MTU discovery is disabled the value
2285 * specified here will be the "fixed" path mtu.
2286 * Note that if the spp_address field is empty
2287 * then all associations on this address will
2288 * have this fixed path mtu set upon them.
2290 * spp_sackdelay - When delayed sack is enabled, this value specifies
2291 * the number of milliseconds that sacks will be delayed
2292 * for. This value will apply to all addresses of an
2293 * association if the spp_address field is empty. Note
2294 * also, that if delayed sack is enabled and this
2295 * value is set to 0, no change is made to the last
2296 * recorded delayed sack timer value.
2298 * spp_flags - These flags are used to control various features
2299 * on an association. The flag field may contain
2300 * zero or more of the following options.
2302 * SPP_HB_ENABLE - Enable heartbeats on the
2303 * specified address. Note that if the address
2304 * field is empty all addresses for the association
2305 * have heartbeats enabled upon them.
2307 * SPP_HB_DISABLE - Disable heartbeats on the
2308 * speicifed address. Note that if the address
2309 * field is empty all addresses for the association
2310 * will have their heartbeats disabled. Note also
2311 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2312 * mutually exclusive, only one of these two should
2313 * be specified. Enabling both fields will have
2314 * undetermined results.
2316 * SPP_HB_DEMAND - Request a user initiated heartbeat
2317 * to be made immediately.
2319 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2320 * heartbeat delayis to be set to the value of 0
2323 * SPP_PMTUD_ENABLE - This field will enable PMTU
2324 * discovery upon the specified address. Note that
2325 * if the address feild is empty then all addresses
2326 * on the association are effected.
2328 * SPP_PMTUD_DISABLE - This field will disable PMTU
2329 * discovery upon the specified address. Note that
2330 * if the address feild is empty then all addresses
2331 * on the association are effected. Not also that
2332 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2333 * exclusive. Enabling both will have undetermined
2336 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2337 * on delayed sack. The time specified in spp_sackdelay
2338 * is used to specify the sack delay for this address. Note
2339 * that if spp_address is empty then all addresses will
2340 * enable delayed sack and take on the sack delay
2341 * value specified in spp_sackdelay.
2342 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2343 * off delayed sack. If the spp_address field is blank then
2344 * delayed sack is disabled for the entire association. Note
2345 * also that this field is mutually exclusive to
2346 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2349 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2350 struct sctp_transport *trans,
2351 struct sctp_association *asoc,
2352 struct sctp_sock *sp,
2355 int sackdelay_change)
2359 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2360 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2365 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2366 * this field is ignored. Note also that a value of zero indicates
2367 * the current setting should be left unchanged.
2369 if (params->spp_flags & SPP_HB_ENABLE) {
2371 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2372 * set. This lets us use 0 value when this flag
2375 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2376 params->spp_hbinterval = 0;
2378 if (params->spp_hbinterval ||
2379 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2382 msecs_to_jiffies(params->spp_hbinterval);
2385 msecs_to_jiffies(params->spp_hbinterval);
2387 sp->hbinterval = params->spp_hbinterval;
2394 trans->param_flags =
2395 (trans->param_flags & ~SPP_HB) | hb_change;
2398 (asoc->param_flags & ~SPP_HB) | hb_change;
2401 (sp->param_flags & ~SPP_HB) | hb_change;
2405 /* When Path MTU discovery is disabled the value specified here will
2406 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2407 * include the flag SPP_PMTUD_DISABLE for this field to have any
2410 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2412 trans->pathmtu = params->spp_pathmtu;
2413 sctp_assoc_sync_pmtu(asoc);
2415 asoc->pathmtu = params->spp_pathmtu;
2416 sctp_frag_point(asoc, params->spp_pathmtu);
2418 sp->pathmtu = params->spp_pathmtu;
2424 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2425 (params->spp_flags & SPP_PMTUD_ENABLE);
2426 trans->param_flags =
2427 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2429 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2430 sctp_assoc_sync_pmtu(asoc);
2434 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2437 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2441 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2442 * value of this field is ignored. Note also that a value of zero
2443 * indicates the current setting should be left unchanged.
2445 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2448 msecs_to_jiffies(params->spp_sackdelay);
2451 msecs_to_jiffies(params->spp_sackdelay);
2453 sp->sackdelay = params->spp_sackdelay;
2457 if (sackdelay_change) {
2459 trans->param_flags =
2460 (trans->param_flags & ~SPP_SACKDELAY) |
2464 (asoc->param_flags & ~SPP_SACKDELAY) |
2468 (sp->param_flags & ~SPP_SACKDELAY) |
2473 /* Note that a value of zero indicates the current setting should be
2476 if (params->spp_pathmaxrxt) {
2478 trans->pathmaxrxt = params->spp_pathmaxrxt;
2480 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2482 sp->pathmaxrxt = params->spp_pathmaxrxt;
2489 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2490 char __user *optval,
2491 unsigned int optlen)
2493 struct sctp_paddrparams params;
2494 struct sctp_transport *trans = NULL;
2495 struct sctp_association *asoc = NULL;
2496 struct sctp_sock *sp = sctp_sk(sk);
2498 int hb_change, pmtud_change, sackdelay_change;
2500 if (optlen != sizeof(struct sctp_paddrparams))
2503 if (copy_from_user(¶ms, optval, optlen))
2506 /* Validate flags and value parameters. */
2507 hb_change = params.spp_flags & SPP_HB;
2508 pmtud_change = params.spp_flags & SPP_PMTUD;
2509 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2511 if (hb_change == SPP_HB ||
2512 pmtud_change == SPP_PMTUD ||
2513 sackdelay_change == SPP_SACKDELAY ||
2514 params.spp_sackdelay > 500 ||
2515 (params.spp_pathmtu &&
2516 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2519 /* If an address other than INADDR_ANY is specified, and
2520 * no transport is found, then the request is invalid.
2522 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
2523 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2524 params.spp_assoc_id);
2529 /* Get association, if assoc_id != 0 and the socket is a one
2530 * to many style socket, and an association was not found, then
2531 * the id was invalid.
2533 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2534 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2537 /* Heartbeat demand can only be sent on a transport or
2538 * association, but not a socket.
2540 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2543 /* Process parameters. */
2544 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2545 hb_change, pmtud_change,
2551 /* If changes are for association, also apply parameters to each
2554 if (!trans && asoc) {
2555 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2557 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2558 hb_change, pmtud_change,
2567 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2569 * This option will effect the way delayed acks are performed. This
2570 * option allows you to get or set the delayed ack time, in
2571 * milliseconds. It also allows changing the delayed ack frequency.
2572 * Changing the frequency to 1 disables the delayed sack algorithm. If
2573 * the assoc_id is 0, then this sets or gets the endpoints default
2574 * values. If the assoc_id field is non-zero, then the set or get
2575 * effects the specified association for the one to many model (the
2576 * assoc_id field is ignored by the one to one model). Note that if
2577 * sack_delay or sack_freq are 0 when setting this option, then the
2578 * current values will remain unchanged.
2580 * struct sctp_sack_info {
2581 * sctp_assoc_t sack_assoc_id;
2582 * uint32_t sack_delay;
2583 * uint32_t sack_freq;
2586 * sack_assoc_id - This parameter, indicates which association the user
2587 * is performing an action upon. Note that if this field's value is
2588 * zero then the endpoints default value is changed (effecting future
2589 * associations only).
2591 * sack_delay - This parameter contains the number of milliseconds that
2592 * the user is requesting the delayed ACK timer be set to. Note that
2593 * this value is defined in the standard to be between 200 and 500
2596 * sack_freq - This parameter contains the number of packets that must
2597 * be received before a sack is sent without waiting for the delay
2598 * timer to expire. The default value for this is 2, setting this
2599 * value to 1 will disable the delayed sack algorithm.
2602 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2603 char __user *optval, unsigned int optlen)
2605 struct sctp_sack_info params;
2606 struct sctp_transport *trans = NULL;
2607 struct sctp_association *asoc = NULL;
2608 struct sctp_sock *sp = sctp_sk(sk);
2610 if (optlen == sizeof(struct sctp_sack_info)) {
2611 if (copy_from_user(¶ms, optval, optlen))
2614 if (params.sack_delay == 0 && params.sack_freq == 0)
2616 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2617 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2618 pr_warn("Use struct sctp_sack_info instead\n");
2619 if (copy_from_user(¶ms, optval, optlen))
2622 if (params.sack_delay == 0)
2623 params.sack_freq = 1;
2625 params.sack_freq = 0;
2629 /* Validate value parameter. */
2630 if (params.sack_delay > 500)
2633 /* Get association, if sack_assoc_id != 0 and the socket is a one
2634 * to many style socket, and an association was not found, then
2635 * the id was invalid.
2637 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2638 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2641 if (params.sack_delay) {
2644 msecs_to_jiffies(params.sack_delay);
2646 (asoc->param_flags & ~SPP_SACKDELAY) |
2647 SPP_SACKDELAY_ENABLE;
2649 sp->sackdelay = params.sack_delay;
2651 (sp->param_flags & ~SPP_SACKDELAY) |
2652 SPP_SACKDELAY_ENABLE;
2656 if (params.sack_freq == 1) {
2659 (asoc->param_flags & ~SPP_SACKDELAY) |
2660 SPP_SACKDELAY_DISABLE;
2663 (sp->param_flags & ~SPP_SACKDELAY) |
2664 SPP_SACKDELAY_DISABLE;
2666 } else if (params.sack_freq > 1) {
2668 asoc->sackfreq = params.sack_freq;
2670 (asoc->param_flags & ~SPP_SACKDELAY) |
2671 SPP_SACKDELAY_ENABLE;
2673 sp->sackfreq = params.sack_freq;
2675 (sp->param_flags & ~SPP_SACKDELAY) |
2676 SPP_SACKDELAY_ENABLE;
2680 /* If change is for association, also apply to each transport. */
2682 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2684 if (params.sack_delay) {
2686 msecs_to_jiffies(params.sack_delay);
2687 trans->param_flags =
2688 (trans->param_flags & ~SPP_SACKDELAY) |
2689 SPP_SACKDELAY_ENABLE;
2691 if (params.sack_freq == 1) {
2692 trans->param_flags =
2693 (trans->param_flags & ~SPP_SACKDELAY) |
2694 SPP_SACKDELAY_DISABLE;
2695 } else if (params.sack_freq > 1) {
2696 trans->sackfreq = params.sack_freq;
2697 trans->param_flags =
2698 (trans->param_flags & ~SPP_SACKDELAY) |
2699 SPP_SACKDELAY_ENABLE;
2707 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2709 * Applications can specify protocol parameters for the default association
2710 * initialization. The option name argument to setsockopt() and getsockopt()
2713 * Setting initialization parameters is effective only on an unconnected
2714 * socket (for UDP-style sockets only future associations are effected
2715 * by the change). With TCP-style sockets, this option is inherited by
2716 * sockets derived from a listener socket.
2718 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2720 struct sctp_initmsg sinit;
2721 struct sctp_sock *sp = sctp_sk(sk);
2723 if (optlen != sizeof(struct sctp_initmsg))
2725 if (copy_from_user(&sinit, optval, optlen))
2728 if (sinit.sinit_num_ostreams)
2729 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2730 if (sinit.sinit_max_instreams)
2731 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2732 if (sinit.sinit_max_attempts)
2733 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2734 if (sinit.sinit_max_init_timeo)
2735 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2741 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2743 * Applications that wish to use the sendto() system call may wish to
2744 * specify a default set of parameters that would normally be supplied
2745 * through the inclusion of ancillary data. This socket option allows
2746 * such an application to set the default sctp_sndrcvinfo structure.
2747 * The application that wishes to use this socket option simply passes
2748 * in to this call the sctp_sndrcvinfo structure defined in Section
2749 * 5.2.2) The input parameters accepted by this call include
2750 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2751 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2752 * to this call if the caller is using the UDP model.
2754 static int sctp_setsockopt_default_send_param(struct sock *sk,
2755 char __user *optval,
2756 unsigned int optlen)
2758 struct sctp_sndrcvinfo info;
2759 struct sctp_association *asoc;
2760 struct sctp_sock *sp = sctp_sk(sk);
2762 if (optlen != sizeof(struct sctp_sndrcvinfo))
2764 if (copy_from_user(&info, optval, optlen))
2767 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2768 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2772 asoc->default_stream = info.sinfo_stream;
2773 asoc->default_flags = info.sinfo_flags;
2774 asoc->default_ppid = info.sinfo_ppid;
2775 asoc->default_context = info.sinfo_context;
2776 asoc->default_timetolive = info.sinfo_timetolive;
2778 sp->default_stream = info.sinfo_stream;
2779 sp->default_flags = info.sinfo_flags;
2780 sp->default_ppid = info.sinfo_ppid;
2781 sp->default_context = info.sinfo_context;
2782 sp->default_timetolive = info.sinfo_timetolive;
2788 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2790 * Requests that the local SCTP stack use the enclosed peer address as
2791 * the association primary. The enclosed address must be one of the
2792 * association peer's addresses.
2794 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2795 unsigned int optlen)
2797 struct sctp_prim prim;
2798 struct sctp_transport *trans;
2800 if (optlen != sizeof(struct sctp_prim))
2803 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2806 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2810 sctp_assoc_set_primary(trans->asoc, trans);
2816 * 7.1.5 SCTP_NODELAY
2818 * Turn on/off any Nagle-like algorithm. This means that packets are
2819 * generally sent as soon as possible and no unnecessary delays are
2820 * introduced, at the cost of more packets in the network. Expects an
2821 * integer boolean flag.
2823 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2824 unsigned int optlen)
2828 if (optlen < sizeof(int))
2830 if (get_user(val, (int __user *)optval))
2833 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2839 * 7.1.1 SCTP_RTOINFO
2841 * The protocol parameters used to initialize and bound retransmission
2842 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2843 * and modify these parameters.
2844 * All parameters are time values, in milliseconds. A value of 0, when
2845 * modifying the parameters, indicates that the current value should not
2849 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2851 struct sctp_rtoinfo rtoinfo;
2852 struct sctp_association *asoc;
2854 if (optlen != sizeof (struct sctp_rtoinfo))
2857 if (copy_from_user(&rtoinfo, optval, optlen))
2860 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2862 /* Set the values to the specific association */
2863 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2867 if (rtoinfo.srto_initial != 0)
2869 msecs_to_jiffies(rtoinfo.srto_initial);
2870 if (rtoinfo.srto_max != 0)
2871 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2872 if (rtoinfo.srto_min != 0)
2873 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2875 /* If there is no association or the association-id = 0
2876 * set the values to the endpoint.
2878 struct sctp_sock *sp = sctp_sk(sk);
2880 if (rtoinfo.srto_initial != 0)
2881 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2882 if (rtoinfo.srto_max != 0)
2883 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2884 if (rtoinfo.srto_min != 0)
2885 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2893 * 7.1.2 SCTP_ASSOCINFO
2895 * This option is used to tune the maximum retransmission attempts
2896 * of the association.
2897 * Returns an error if the new association retransmission value is
2898 * greater than the sum of the retransmission value of the peer.
2899 * See [SCTP] for more information.
2902 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2905 struct sctp_assocparams assocparams;
2906 struct sctp_association *asoc;
2908 if (optlen != sizeof(struct sctp_assocparams))
2910 if (copy_from_user(&assocparams, optval, optlen))
2913 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2915 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2918 /* Set the values to the specific association */
2920 if (assocparams.sasoc_asocmaxrxt != 0) {
2923 struct sctp_transport *peer_addr;
2925 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2927 path_sum += peer_addr->pathmaxrxt;
2931 /* Only validate asocmaxrxt if we have more than
2932 * one path/transport. We do this because path
2933 * retransmissions are only counted when we have more
2937 assocparams.sasoc_asocmaxrxt > path_sum)
2940 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2943 if (assocparams.sasoc_cookie_life != 0) {
2944 asoc->cookie_life.tv_sec =
2945 assocparams.sasoc_cookie_life / 1000;
2946 asoc->cookie_life.tv_usec =
2947 (assocparams.sasoc_cookie_life % 1000)
2951 /* Set the values to the endpoint */
2952 struct sctp_sock *sp = sctp_sk(sk);
2954 if (assocparams.sasoc_asocmaxrxt != 0)
2955 sp->assocparams.sasoc_asocmaxrxt =
2956 assocparams.sasoc_asocmaxrxt;
2957 if (assocparams.sasoc_cookie_life != 0)
2958 sp->assocparams.sasoc_cookie_life =
2959 assocparams.sasoc_cookie_life;
2965 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2967 * This socket option is a boolean flag which turns on or off mapped V4
2968 * addresses. If this option is turned on and the socket is type
2969 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2970 * If this option is turned off, then no mapping will be done of V4
2971 * addresses and a user will receive both PF_INET6 and PF_INET type
2972 * addresses on the socket.
2974 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2977 struct sctp_sock *sp = sctp_sk(sk);
2979 if (optlen < sizeof(int))
2981 if (get_user(val, (int __user *)optval))
2992 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2993 * This option will get or set the maximum size to put in any outgoing
2994 * SCTP DATA chunk. If a message is larger than this size it will be
2995 * fragmented by SCTP into the specified size. Note that the underlying
2996 * SCTP implementation may fragment into smaller sized chunks when the
2997 * PMTU of the underlying association is smaller than the value set by
2998 * the user. The default value for this option is '0' which indicates
2999 * the user is NOT limiting fragmentation and only the PMTU will effect
3000 * SCTP's choice of DATA chunk size. Note also that values set larger
3001 * than the maximum size of an IP datagram will effectively let SCTP
3002 * control fragmentation (i.e. the same as setting this option to 0).
3004 * The following structure is used to access and modify this parameter:
3006 * struct sctp_assoc_value {
3007 * sctp_assoc_t assoc_id;
3008 * uint32_t assoc_value;
3011 * assoc_id: This parameter is ignored for one-to-one style sockets.
3012 * For one-to-many style sockets this parameter indicates which
3013 * association the user is performing an action upon. Note that if
3014 * this field's value is zero then the endpoints default value is
3015 * changed (effecting future associations only).
3016 * assoc_value: This parameter specifies the maximum size in bytes.
3018 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3020 struct sctp_assoc_value params;
3021 struct sctp_association *asoc;
3022 struct sctp_sock *sp = sctp_sk(sk);
3025 if (optlen == sizeof(int)) {
3026 pr_warn("Use of int in maxseg socket option deprecated\n");
3027 pr_warn("Use struct sctp_assoc_value instead\n");
3028 if (copy_from_user(&val, optval, optlen))
3030 params.assoc_id = 0;
3031 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3032 if (copy_from_user(¶ms, optval, optlen))
3034 val = params.assoc_value;
3038 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3041 asoc = sctp_id2assoc(sk, params.assoc_id);
3042 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3047 val = asoc->pathmtu;
3048 val -= sp->pf->af->net_header_len;
3049 val -= sizeof(struct sctphdr) +
3050 sizeof(struct sctp_data_chunk);
3052 asoc->user_frag = val;
3053 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3055 sp->user_frag = val;
3063 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3065 * Requests that the peer mark the enclosed address as the association
3066 * primary. The enclosed address must be one of the association's
3067 * locally bound addresses. The following structure is used to make a
3068 * set primary request:
3070 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3071 unsigned int optlen)
3073 struct sctp_sock *sp;
3074 struct sctp_association *asoc = NULL;
3075 struct sctp_setpeerprim prim;
3076 struct sctp_chunk *chunk;
3082 if (!sctp_addip_enable)
3085 if (optlen != sizeof(struct sctp_setpeerprim))
3088 if (copy_from_user(&prim, optval, optlen))
3091 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3095 if (!asoc->peer.asconf_capable)
3098 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3101 if (!sctp_state(asoc, ESTABLISHED))
3104 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3108 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3109 return -EADDRNOTAVAIL;
3111 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3112 return -EADDRNOTAVAIL;
3114 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3115 chunk = sctp_make_asconf_set_prim(asoc,
3116 (union sctp_addr *)&prim.sspp_addr);
3120 err = sctp_send_asconf(asoc, chunk);
3122 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
3127 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3128 unsigned int optlen)
3130 struct sctp_setadaptation adaptation;
3132 if (optlen != sizeof(struct sctp_setadaptation))
3134 if (copy_from_user(&adaptation, optval, optlen))
3137 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3143 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3145 * The context field in the sctp_sndrcvinfo structure is normally only
3146 * used when a failed message is retrieved holding the value that was
3147 * sent down on the actual send call. This option allows the setting of
3148 * a default context on an association basis that will be received on
3149 * reading messages from the peer. This is especially helpful in the
3150 * one-2-many model for an application to keep some reference to an
3151 * internal state machine that is processing messages on the
3152 * association. Note that the setting of this value only effects
3153 * received messages from the peer and does not effect the value that is
3154 * saved with outbound messages.
3156 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3157 unsigned int optlen)
3159 struct sctp_assoc_value params;
3160 struct sctp_sock *sp;
3161 struct sctp_association *asoc;
3163 if (optlen != sizeof(struct sctp_assoc_value))
3165 if (copy_from_user(¶ms, optval, optlen))
3170 if (params.assoc_id != 0) {
3171 asoc = sctp_id2assoc(sk, params.assoc_id);
3174 asoc->default_rcv_context = params.assoc_value;
3176 sp->default_rcv_context = params.assoc_value;
3183 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3185 * This options will at a minimum specify if the implementation is doing
3186 * fragmented interleave. Fragmented interleave, for a one to many
3187 * socket, is when subsequent calls to receive a message may return
3188 * parts of messages from different associations. Some implementations
3189 * may allow you to turn this value on or off. If so, when turned off,
3190 * no fragment interleave will occur (which will cause a head of line
3191 * blocking amongst multiple associations sharing the same one to many
3192 * socket). When this option is turned on, then each receive call may
3193 * come from a different association (thus the user must receive data
3194 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3195 * association each receive belongs to.
3197 * This option takes a boolean value. A non-zero value indicates that
3198 * fragmented interleave is on. A value of zero indicates that
3199 * fragmented interleave is off.
3201 * Note that it is important that an implementation that allows this
3202 * option to be turned on, have it off by default. Otherwise an unaware
3203 * application using the one to many model may become confused and act
3206 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3207 char __user *optval,
3208 unsigned int optlen)
3212 if (optlen != sizeof(int))
3214 if (get_user(val, (int __user *)optval))
3217 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3223 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3224 * (SCTP_PARTIAL_DELIVERY_POINT)
3226 * This option will set or get the SCTP partial delivery point. This
3227 * point is the size of a message where the partial delivery API will be
3228 * invoked to help free up rwnd space for the peer. Setting this to a
3229 * lower value will cause partial deliveries to happen more often. The
3230 * calls argument is an integer that sets or gets the partial delivery
3231 * point. Note also that the call will fail if the user attempts to set
3232 * this value larger than the socket receive buffer size.
3234 * Note that any single message having a length smaller than or equal to
3235 * the SCTP partial delivery point will be delivered in one single read
3236 * call as long as the user provided buffer is large enough to hold the
3239 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3240 char __user *optval,
3241 unsigned int optlen)
3245 if (optlen != sizeof(u32))
3247 if (get_user(val, (int __user *)optval))
3250 /* Note: We double the receive buffer from what the user sets
3251 * it to be, also initial rwnd is based on rcvbuf/2.
3253 if (val > (sk->sk_rcvbuf >> 1))
3256 sctp_sk(sk)->pd_point = val;
3258 return 0; /* is this the right error code? */
3262 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3264 * This option will allow a user to change the maximum burst of packets
3265 * that can be emitted by this association. Note that the default value
3266 * is 4, and some implementations may restrict this setting so that it
3267 * can only be lowered.
3269 * NOTE: This text doesn't seem right. Do this on a socket basis with
3270 * future associations inheriting the socket value.
3272 static int sctp_setsockopt_maxburst(struct sock *sk,
3273 char __user *optval,
3274 unsigned int optlen)
3276 struct sctp_assoc_value params;
3277 struct sctp_sock *sp;
3278 struct sctp_association *asoc;
3282 if (optlen == sizeof(int)) {
3283 pr_warn("Use of int in max_burst socket option deprecated\n");
3284 pr_warn("Use struct sctp_assoc_value instead\n");
3285 if (copy_from_user(&val, optval, optlen))
3287 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3288 if (copy_from_user(¶ms, optval, optlen))
3290 val = params.assoc_value;
3291 assoc_id = params.assoc_id;
3297 if (assoc_id != 0) {
3298 asoc = sctp_id2assoc(sk, assoc_id);
3301 asoc->max_burst = val;
3303 sp->max_burst = val;
3309 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3311 * This set option adds a chunk type that the user is requesting to be
3312 * received only in an authenticated way. Changes to the list of chunks
3313 * will only effect future associations on the socket.
3315 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3316 char __user *optval,
3317 unsigned int optlen)
3319 struct sctp_authchunk val;
3321 if (!sctp_auth_enable)
3324 if (optlen != sizeof(struct sctp_authchunk))
3326 if (copy_from_user(&val, optval, optlen))
3329 switch (val.sauth_chunk) {
3331 case SCTP_CID_INIT_ACK:
3332 case SCTP_CID_SHUTDOWN_COMPLETE:
3337 /* add this chunk id to the endpoint */
3338 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3342 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3344 * This option gets or sets the list of HMAC algorithms that the local
3345 * endpoint requires the peer to use.
3347 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3348 char __user *optval,
3349 unsigned int optlen)
3351 struct sctp_hmacalgo *hmacs;
3355 if (!sctp_auth_enable)
3358 if (optlen < sizeof(struct sctp_hmacalgo))
3361 hmacs= memdup_user(optval, optlen);
3363 return PTR_ERR(hmacs);
3365 idents = hmacs->shmac_num_idents;
3366 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3367 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3372 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3379 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3381 * This option will set a shared secret key which is used to build an
3382 * association shared key.
3384 static int sctp_setsockopt_auth_key(struct sock *sk,
3385 char __user *optval,
3386 unsigned int optlen)
3388 struct sctp_authkey *authkey;
3389 struct sctp_association *asoc;
3392 if (!sctp_auth_enable)
3395 if (optlen <= sizeof(struct sctp_authkey))
3398 authkey= memdup_user(optval, optlen);
3399 if (IS_ERR(authkey))
3400 return PTR_ERR(authkey);
3402 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3407 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3408 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3413 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3420 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3422 * This option will get or set the active shared key to be used to build
3423 * the association shared key.
3425 static int sctp_setsockopt_active_key(struct sock *sk,
3426 char __user *optval,
3427 unsigned int optlen)
3429 struct sctp_authkeyid val;
3430 struct sctp_association *asoc;
3432 if (!sctp_auth_enable)
3435 if (optlen != sizeof(struct sctp_authkeyid))
3437 if (copy_from_user(&val, optval, optlen))
3440 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3441 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3444 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3445 val.scact_keynumber);
3449 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3451 * This set option will delete a shared secret key from use.
3453 static int sctp_setsockopt_del_key(struct sock *sk,
3454 char __user *optval,
3455 unsigned int optlen)
3457 struct sctp_authkeyid val;
3458 struct sctp_association *asoc;
3460 if (!sctp_auth_enable)
3463 if (optlen != sizeof(struct sctp_authkeyid))
3465 if (copy_from_user(&val, optval, optlen))
3468 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3469 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3472 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3473 val.scact_keynumber);
3478 * 8.1.23 SCTP_AUTO_ASCONF
3480 * This option will enable or disable the use of the automatic generation of
3481 * ASCONF chunks to add and delete addresses to an existing association. Note
3482 * that this option has two caveats namely: a) it only affects sockets that
3483 * are bound to all addresses available to the SCTP stack, and b) the system
3484 * administrator may have an overriding control that turns the ASCONF feature
3485 * off no matter what setting the socket option may have.
3486 * This option expects an integer boolean flag, where a non-zero value turns on
3487 * the option, and a zero value turns off the option.
3488 * Note. In this implementation, socket operation overrides default parameter
3489 * being set by sysctl as well as FreeBSD implementation
3491 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3492 unsigned int optlen)
3495 struct sctp_sock *sp = sctp_sk(sk);
3497 if (optlen < sizeof(int))
3499 if (get_user(val, (int __user *)optval))
3501 if (!sctp_is_ep_boundall(sk) && val)
3503 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3506 spin_lock_bh(&sctp_globals.addr_wq_lock);
3507 if (val == 0 && sp->do_auto_asconf) {
3508 list_del(&sp->auto_asconf_list);
3509 sp->do_auto_asconf = 0;
3510 } else if (val && !sp->do_auto_asconf) {
3511 list_add_tail(&sp->auto_asconf_list,
3512 &sctp_auto_asconf_splist);
3513 sp->do_auto_asconf = 1;
3515 spin_unlock_bh(&sctp_globals.addr_wq_lock);
3520 /* API 6.2 setsockopt(), getsockopt()
3522 * Applications use setsockopt() and getsockopt() to set or retrieve
3523 * socket options. Socket options are used to change the default
3524 * behavior of sockets calls. They are described in Section 7.
3528 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3529 * int __user *optlen);
3530 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3533 * sd - the socket descript.
3534 * level - set to IPPROTO_SCTP for all SCTP options.
3535 * optname - the option name.
3536 * optval - the buffer to store the value of the option.
3537 * optlen - the size of the buffer.
3539 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3540 char __user *optval, unsigned int optlen)
3544 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3547 /* I can hardly begin to describe how wrong this is. This is
3548 * so broken as to be worse than useless. The API draft
3549 * REALLY is NOT helpful here... I am not convinced that the
3550 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3551 * are at all well-founded.
3553 if (level != SOL_SCTP) {
3554 struct sctp_af *af = sctp_sk(sk)->pf->af;
3555 retval = af->setsockopt(sk, level, optname, optval, optlen);
3562 case SCTP_SOCKOPT_BINDX_ADD:
3563 /* 'optlen' is the size of the addresses buffer. */
3564 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3565 optlen, SCTP_BINDX_ADD_ADDR);
3568 case SCTP_SOCKOPT_BINDX_REM:
3569 /* 'optlen' is the size of the addresses buffer. */
3570 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3571 optlen, SCTP_BINDX_REM_ADDR);
3574 case SCTP_SOCKOPT_CONNECTX_OLD:
3575 /* 'optlen' is the size of the addresses buffer. */
3576 retval = sctp_setsockopt_connectx_old(sk,
3577 (struct sockaddr __user *)optval,
3581 case SCTP_SOCKOPT_CONNECTX:
3582 /* 'optlen' is the size of the addresses buffer. */
3583 retval = sctp_setsockopt_connectx(sk,
3584 (struct sockaddr __user *)optval,
3588 case SCTP_DISABLE_FRAGMENTS:
3589 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3593 retval = sctp_setsockopt_events(sk, optval, optlen);
3596 case SCTP_AUTOCLOSE:
3597 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3600 case SCTP_PEER_ADDR_PARAMS:
3601 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3604 case SCTP_DELAYED_SACK:
3605 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3607 case SCTP_PARTIAL_DELIVERY_POINT:
3608 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3612 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3614 case SCTP_DEFAULT_SEND_PARAM:
3615 retval = sctp_setsockopt_default_send_param(sk, optval,
3618 case SCTP_PRIMARY_ADDR:
3619 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3621 case SCTP_SET_PEER_PRIMARY_ADDR:
3622 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3625 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3628 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3630 case SCTP_ASSOCINFO:
3631 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3633 case SCTP_I_WANT_MAPPED_V4_ADDR:
3634 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3637 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3639 case SCTP_ADAPTATION_LAYER:
3640 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3643 retval = sctp_setsockopt_context(sk, optval, optlen);
3645 case SCTP_FRAGMENT_INTERLEAVE:
3646 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3648 case SCTP_MAX_BURST:
3649 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3651 case SCTP_AUTH_CHUNK:
3652 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3654 case SCTP_HMAC_IDENT:
3655 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3658 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3660 case SCTP_AUTH_ACTIVE_KEY:
3661 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3663 case SCTP_AUTH_DELETE_KEY:
3664 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3666 case SCTP_AUTO_ASCONF:
3667 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3670 retval = -ENOPROTOOPT;
3674 sctp_release_sock(sk);
3680 /* API 3.1.6 connect() - UDP Style Syntax
3682 * An application may use the connect() call in the UDP model to initiate an
3683 * association without sending data.
3687 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3689 * sd: the socket descriptor to have a new association added to.
3691 * nam: the address structure (either struct sockaddr_in or struct
3692 * sockaddr_in6 defined in RFC2553 [7]).
3694 * len: the size of the address.
3696 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3704 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3705 __func__, sk, addr, addr_len);
3707 /* Validate addr_len before calling common connect/connectx routine. */
3708 af = sctp_get_af_specific(addr->sa_family);
3709 if (!af || addr_len < af->sockaddr_len) {
3712 /* Pass correct addr len to common routine (so it knows there
3713 * is only one address being passed.
3715 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3718 sctp_release_sock(sk);
3722 /* FIXME: Write comments. */
3723 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3725 return -EOPNOTSUPP; /* STUB */
3728 /* 4.1.4 accept() - TCP Style Syntax
3730 * Applications use accept() call to remove an established SCTP
3731 * association from the accept queue of the endpoint. A new socket
3732 * descriptor will be returned from accept() to represent the newly
3733 * formed association.
3735 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3737 struct sctp_sock *sp;
3738 struct sctp_endpoint *ep;
3739 struct sock *newsk = NULL;
3740 struct sctp_association *asoc;
3749 if (!sctp_style(sk, TCP)) {
3750 error = -EOPNOTSUPP;
3754 if (!sctp_sstate(sk, LISTENING)) {
3759 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3761 error = sctp_wait_for_accept(sk, timeo);
3765 /* We treat the list of associations on the endpoint as the accept
3766 * queue and pick the first association on the list.
3768 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3770 newsk = sp->pf->create_accept_sk(sk, asoc);
3776 /* Populate the fields of the newsk from the oldsk and migrate the
3777 * asoc to the newsk.
3779 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3782 sctp_release_sock(sk);
3787 /* The SCTP ioctl handler. */
3788 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3795 * SEQPACKET-style sockets in LISTENING state are valid, for
3796 * SCTP, so only discard TCP-style sockets in LISTENING state.
3798 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3803 struct sk_buff *skb;
3804 unsigned int amount = 0;
3806 skb = skb_peek(&sk->sk_receive_queue);
3809 * We will only return the amount of this packet since
3810 * that is all that will be read.
3814 rc = put_user(amount, (int __user *)arg);
3822 sctp_release_sock(sk);
3826 /* This is the function which gets called during socket creation to
3827 * initialized the SCTP-specific portion of the sock.
3828 * The sock structure should already be zero-filled memory.
3830 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3832 struct sctp_endpoint *ep;
3833 struct sctp_sock *sp;
3835 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3839 /* Initialize the SCTP per socket area. */
3840 switch (sk->sk_type) {
3841 case SOCK_SEQPACKET:
3842 sp->type = SCTP_SOCKET_UDP;
3845 sp->type = SCTP_SOCKET_TCP;
3848 return -ESOCKTNOSUPPORT;
3851 /* Initialize default send parameters. These parameters can be
3852 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3854 sp->default_stream = 0;
3855 sp->default_ppid = 0;
3856 sp->default_flags = 0;
3857 sp->default_context = 0;
3858 sp->default_timetolive = 0;
3860 sp->default_rcv_context = 0;
3861 sp->max_burst = sctp_max_burst;
3863 /* Initialize default setup parameters. These parameters
3864 * can be modified with the SCTP_INITMSG socket option or
3865 * overridden by the SCTP_INIT CMSG.
3867 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3868 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3869 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3870 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3872 /* Initialize default RTO related parameters. These parameters can
3873 * be modified for with the SCTP_RTOINFO socket option.
3875 sp->rtoinfo.srto_initial = sctp_rto_initial;
3876 sp->rtoinfo.srto_max = sctp_rto_max;
3877 sp->rtoinfo.srto_min = sctp_rto_min;
3879 /* Initialize default association related parameters. These parameters
3880 * can be modified with the SCTP_ASSOCINFO socket option.
3882 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3883 sp->assocparams.sasoc_number_peer_destinations = 0;
3884 sp->assocparams.sasoc_peer_rwnd = 0;
3885 sp->assocparams.sasoc_local_rwnd = 0;
3886 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3888 /* Initialize default event subscriptions. By default, all the
3891 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3893 /* Default Peer Address Parameters. These defaults can
3894 * be modified via SCTP_PEER_ADDR_PARAMS
3896 sp->hbinterval = sctp_hb_interval;
3897 sp->pathmaxrxt = sctp_max_retrans_path;
3898 sp->pathmtu = 0; // allow default discovery
3899 sp->sackdelay = sctp_sack_timeout;
3901 sp->param_flags = SPP_HB_ENABLE |
3903 SPP_SACKDELAY_ENABLE;
3905 /* If enabled no SCTP message fragmentation will be performed.
3906 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3908 sp->disable_fragments = 0;
3910 /* Enable Nagle algorithm by default. */
3913 /* Enable by default. */
3916 /* Auto-close idle associations after the configured
3917 * number of seconds. A value of 0 disables this
3918 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3919 * for UDP-style sockets only.
3923 /* User specified fragmentation limit. */
3926 sp->adaptation_ind = 0;
3928 sp->pf = sctp_get_pf_specific(sk->sk_family);
3930 /* Control variables for partial data delivery. */
3931 atomic_set(&sp->pd_mode, 0);
3932 skb_queue_head_init(&sp->pd_lobby);
3933 sp->frag_interleave = 0;
3935 /* Create a per socket endpoint structure. Even if we
3936 * change the data structure relationships, this may still
3937 * be useful for storing pre-connect address information.
3939 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3946 SCTP_DBG_OBJCNT_INC(sock);
3949 percpu_counter_inc(&sctp_sockets_allocated);
3950 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3952 /* Nothing can fail after this block, otherwise
3953 * sctp_destroy_sock() will be called without addr_wq_lock held
3955 if (sctp_default_auto_asconf) {
3956 spin_lock(&sctp_globals.addr_wq_lock);
3957 list_add_tail(&sp->auto_asconf_list,
3958 &sctp_auto_asconf_splist);
3959 sp->do_auto_asconf = 1;
3960 spin_unlock(&sctp_globals.addr_wq_lock);
3962 sp->do_auto_asconf = 0;
3970 /* Cleanup any SCTP per socket resources. Must be called with
3971 * sctp_globals.addr_wq_lock held if sp->do_auto_asconf is true
3973 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3975 struct sctp_sock *sp;
3977 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3979 /* Release our hold on the endpoint. */
3981 /* This could happen during socket init, thus we bail out
3982 * early, since the rest of the below is not setup either.
3987 if (sp->do_auto_asconf) {
3988 sp->do_auto_asconf = 0;
3989 list_del(&sp->auto_asconf_list);
3991 sctp_endpoint_free(sp->ep);
3993 percpu_counter_dec(&sctp_sockets_allocated);
3994 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
3998 /* API 4.1.7 shutdown() - TCP Style Syntax
3999 * int shutdown(int socket, int how);
4001 * sd - the socket descriptor of the association to be closed.
4002 * how - Specifies the type of shutdown. The values are
4005 * Disables further receive operations. No SCTP
4006 * protocol action is taken.
4008 * Disables further send operations, and initiates
4009 * the SCTP shutdown sequence.
4011 * Disables further send and receive operations
4012 * and initiates the SCTP shutdown sequence.
4014 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
4016 struct sctp_endpoint *ep;
4017 struct sctp_association *asoc;
4019 if (!sctp_style(sk, TCP))
4022 if (how & SEND_SHUTDOWN) {
4023 ep = sctp_sk(sk)->ep;
4024 if (!list_empty(&ep->asocs)) {
4025 asoc = list_entry(ep->asocs.next,
4026 struct sctp_association, asocs);
4027 sctp_primitive_SHUTDOWN(asoc, NULL);
4032 /* 7.2.1 Association Status (SCTP_STATUS)
4034 * Applications can retrieve current status information about an
4035 * association, including association state, peer receiver window size,
4036 * number of unacked data chunks, and number of data chunks pending
4037 * receipt. This information is read-only.
4039 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4040 char __user *optval,
4043 struct sctp_status status;
4044 struct sctp_association *asoc = NULL;
4045 struct sctp_transport *transport;
4046 sctp_assoc_t associd;
4049 if (len < sizeof(status)) {
4054 len = sizeof(status);
4055 if (copy_from_user(&status, optval, len)) {
4060 associd = status.sstat_assoc_id;
4061 asoc = sctp_id2assoc(sk, associd);
4067 transport = asoc->peer.primary_path;
4069 status.sstat_assoc_id = sctp_assoc2id(asoc);
4070 status.sstat_state = asoc->state;
4071 status.sstat_rwnd = asoc->peer.rwnd;
4072 status.sstat_unackdata = asoc->unack_data;
4074 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4075 status.sstat_instrms = asoc->c.sinit_max_instreams;
4076 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4077 status.sstat_fragmentation_point = asoc->frag_point;
4078 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4079 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4080 transport->af_specific->sockaddr_len);
4081 /* Map ipv4 address into v4-mapped-on-v6 address. */
4082 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
4083 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4084 status.sstat_primary.spinfo_state = transport->state;
4085 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4086 status.sstat_primary.spinfo_srtt = transport->srtt;
4087 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4088 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4090 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4091 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4093 if (put_user(len, optlen)) {
4098 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
4099 len, status.sstat_state, status.sstat_rwnd,
4100 status.sstat_assoc_id);
4102 if (copy_to_user(optval, &status, len)) {
4112 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4114 * Applications can retrieve information about a specific peer address
4115 * of an association, including its reachability state, congestion
4116 * window, and retransmission timer values. This information is
4119 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4120 char __user *optval,
4123 struct sctp_paddrinfo pinfo;
4124 struct sctp_transport *transport;
4127 if (len < sizeof(pinfo)) {
4132 len = sizeof(pinfo);
4133 if (copy_from_user(&pinfo, optval, len)) {
4138 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4139 pinfo.spinfo_assoc_id);
4143 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4144 pinfo.spinfo_state = transport->state;
4145 pinfo.spinfo_cwnd = transport->cwnd;
4146 pinfo.spinfo_srtt = transport->srtt;
4147 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4148 pinfo.spinfo_mtu = transport->pathmtu;
4150 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4151 pinfo.spinfo_state = SCTP_ACTIVE;
4153 if (put_user(len, optlen)) {
4158 if (copy_to_user(optval, &pinfo, len)) {
4167 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4169 * This option is a on/off flag. If enabled no SCTP message
4170 * fragmentation will be performed. Instead if a message being sent
4171 * exceeds the current PMTU size, the message will NOT be sent and
4172 * instead a error will be indicated to the user.
4174 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4175 char __user *optval, int __user *optlen)
4179 if (len < sizeof(int))
4183 val = (sctp_sk(sk)->disable_fragments == 1);
4184 if (put_user(len, optlen))
4186 if (copy_to_user(optval, &val, len))
4191 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4193 * This socket option is used to specify various notifications and
4194 * ancillary data the user wishes to receive.
4196 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4201 if (len > sizeof(struct sctp_event_subscribe))
4202 len = sizeof(struct sctp_event_subscribe);
4203 if (put_user(len, optlen))
4205 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4210 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4212 * This socket option is applicable to the UDP-style socket only. When
4213 * set it will cause associations that are idle for more than the
4214 * specified number of seconds to automatically close. An association
4215 * being idle is defined an association that has NOT sent or received
4216 * user data. The special value of '0' indicates that no automatic
4217 * close of any associations should be performed. The option expects an
4218 * integer defining the number of seconds of idle time before an
4219 * association is closed.
4221 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4223 /* Applicable to UDP-style socket only */
4224 if (sctp_style(sk, TCP))
4226 if (len < sizeof(int))
4229 if (put_user(len, optlen))
4231 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4236 /* Helper routine to branch off an association to a new socket. */
4237 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
4238 struct socket **sockp)
4240 struct sock *sk = asoc->base.sk;
4241 struct sctp_sock *sp = sctp_sk(sk);
4242 struct socket *sock;
4245 /* Do not peel off from one netns to another one. */
4246 if (!net_eq(current->nsproxy->net_ns, sock_net(sk)))
4249 /* An association cannot be branched off from an already peeled-off
4250 * socket, nor is this supported for tcp style sockets.
4252 if (!sctp_style(sk, UDP))
4255 /* Create a new socket. */
4256 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4260 sctp_copy_sock(sock->sk, sk, asoc);
4262 /* Make peeled-off sockets more like 1-1 accepted sockets.
4263 * Set the daddr and initialize id to something more random
4265 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
4267 /* Populate the fields of the newsk from the oldsk and migrate the
4268 * asoc to the newsk.
4270 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4277 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4279 sctp_peeloff_arg_t peeloff;
4280 struct socket *newsock;
4282 struct sctp_association *asoc;
4284 if (len < sizeof(sctp_peeloff_arg_t))
4286 len = sizeof(sctp_peeloff_arg_t);
4287 if (copy_from_user(&peeloff, optval, len))
4290 asoc = sctp_id2assoc(sk, peeloff.associd);
4296 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
4298 retval = sctp_do_peeloff(asoc, &newsock);
4302 /* Map the socket to an unused fd that can be returned to the user. */
4303 retval = sock_map_fd(newsock, 0);
4305 sock_release(newsock);
4309 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
4310 __func__, sk, asoc, newsock->sk, retval);
4312 /* Return the fd mapped to the new socket. */
4313 peeloff.sd = retval;
4314 if (put_user(len, optlen))
4316 if (copy_to_user(optval, &peeloff, len))
4323 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4325 * Applications can enable or disable heartbeats for any peer address of
4326 * an association, modify an address's heartbeat interval, force a
4327 * heartbeat to be sent immediately, and adjust the address's maximum
4328 * number of retransmissions sent before an address is considered
4329 * unreachable. The following structure is used to access and modify an
4330 * address's parameters:
4332 * struct sctp_paddrparams {
4333 * sctp_assoc_t spp_assoc_id;
4334 * struct sockaddr_storage spp_address;
4335 * uint32_t spp_hbinterval;
4336 * uint16_t spp_pathmaxrxt;
4337 * uint32_t spp_pathmtu;
4338 * uint32_t spp_sackdelay;
4339 * uint32_t spp_flags;
4342 * spp_assoc_id - (one-to-many style socket) This is filled in the
4343 * application, and identifies the association for
4345 * spp_address - This specifies which address is of interest.
4346 * spp_hbinterval - This contains the value of the heartbeat interval,
4347 * in milliseconds. If a value of zero
4348 * is present in this field then no changes are to
4349 * be made to this parameter.
4350 * spp_pathmaxrxt - This contains the maximum number of
4351 * retransmissions before this address shall be
4352 * considered unreachable. If a value of zero
4353 * is present in this field then no changes are to
4354 * be made to this parameter.
4355 * spp_pathmtu - When Path MTU discovery is disabled the value
4356 * specified here will be the "fixed" path mtu.
4357 * Note that if the spp_address field is empty
4358 * then all associations on this address will
4359 * have this fixed path mtu set upon them.
4361 * spp_sackdelay - When delayed sack is enabled, this value specifies
4362 * the number of milliseconds that sacks will be delayed
4363 * for. This value will apply to all addresses of an
4364 * association if the spp_address field is empty. Note
4365 * also, that if delayed sack is enabled and this
4366 * value is set to 0, no change is made to the last
4367 * recorded delayed sack timer value.
4369 * spp_flags - These flags are used to control various features
4370 * on an association. The flag field may contain
4371 * zero or more of the following options.
4373 * SPP_HB_ENABLE - Enable heartbeats on the
4374 * specified address. Note that if the address
4375 * field is empty all addresses for the association
4376 * have heartbeats enabled upon them.
4378 * SPP_HB_DISABLE - Disable heartbeats on the
4379 * speicifed address. Note that if the address
4380 * field is empty all addresses for the association
4381 * will have their heartbeats disabled. Note also
4382 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4383 * mutually exclusive, only one of these two should
4384 * be specified. Enabling both fields will have
4385 * undetermined results.
4387 * SPP_HB_DEMAND - Request a user initiated heartbeat
4388 * to be made immediately.
4390 * SPP_PMTUD_ENABLE - This field will enable PMTU
4391 * discovery upon the specified address. Note that
4392 * if the address feild is empty then all addresses
4393 * on the association are effected.
4395 * SPP_PMTUD_DISABLE - This field will disable PMTU
4396 * discovery upon the specified address. Note that
4397 * if the address feild is empty then all addresses
4398 * on the association are effected. Not also that
4399 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4400 * exclusive. Enabling both will have undetermined
4403 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4404 * on delayed sack. The time specified in spp_sackdelay
4405 * is used to specify the sack delay for this address. Note
4406 * that if spp_address is empty then all addresses will
4407 * enable delayed sack and take on the sack delay
4408 * value specified in spp_sackdelay.
4409 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4410 * off delayed sack. If the spp_address field is blank then
4411 * delayed sack is disabled for the entire association. Note
4412 * also that this field is mutually exclusive to
4413 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4416 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4417 char __user *optval, int __user *optlen)
4419 struct sctp_paddrparams params;
4420 struct sctp_transport *trans = NULL;
4421 struct sctp_association *asoc = NULL;
4422 struct sctp_sock *sp = sctp_sk(sk);
4424 if (len < sizeof(struct sctp_paddrparams))
4426 len = sizeof(struct sctp_paddrparams);
4427 if (copy_from_user(¶ms, optval, len))
4430 /* If an address other than INADDR_ANY is specified, and
4431 * no transport is found, then the request is invalid.
4433 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
4434 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4435 params.spp_assoc_id);
4437 SCTP_DEBUG_PRINTK("Failed no transport\n");
4442 /* Get association, if assoc_id != 0 and the socket is a one
4443 * to many style socket, and an association was not found, then
4444 * the id was invalid.
4446 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4447 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4448 SCTP_DEBUG_PRINTK("Failed no association\n");
4453 /* Fetch transport values. */
4454 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4455 params.spp_pathmtu = trans->pathmtu;
4456 params.spp_pathmaxrxt = trans->pathmaxrxt;
4457 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4459 /*draft-11 doesn't say what to return in spp_flags*/
4460 params.spp_flags = trans->param_flags;
4462 /* Fetch association values. */
4463 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4464 params.spp_pathmtu = asoc->pathmtu;
4465 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4466 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4468 /*draft-11 doesn't say what to return in spp_flags*/
4469 params.spp_flags = asoc->param_flags;
4471 /* Fetch socket values. */
4472 params.spp_hbinterval = sp->hbinterval;
4473 params.spp_pathmtu = sp->pathmtu;
4474 params.spp_sackdelay = sp->sackdelay;
4475 params.spp_pathmaxrxt = sp->pathmaxrxt;
4477 /*draft-11 doesn't say what to return in spp_flags*/
4478 params.spp_flags = sp->param_flags;
4481 if (copy_to_user(optval, ¶ms, len))
4484 if (put_user(len, optlen))
4491 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4493 * This option will effect the way delayed acks are performed. This
4494 * option allows you to get or set the delayed ack time, in
4495 * milliseconds. It also allows changing the delayed ack frequency.
4496 * Changing the frequency to 1 disables the delayed sack algorithm. If
4497 * the assoc_id is 0, then this sets or gets the endpoints default
4498 * values. If the assoc_id field is non-zero, then the set or get
4499 * effects the specified association for the one to many model (the
4500 * assoc_id field is ignored by the one to one model). Note that if
4501 * sack_delay or sack_freq are 0 when setting this option, then the
4502 * current values will remain unchanged.
4504 * struct sctp_sack_info {
4505 * sctp_assoc_t sack_assoc_id;
4506 * uint32_t sack_delay;
4507 * uint32_t sack_freq;
4510 * sack_assoc_id - This parameter, indicates which association the user
4511 * is performing an action upon. Note that if this field's value is
4512 * zero then the endpoints default value is changed (effecting future
4513 * associations only).
4515 * sack_delay - This parameter contains the number of milliseconds that
4516 * the user is requesting the delayed ACK timer be set to. Note that
4517 * this value is defined in the standard to be between 200 and 500
4520 * sack_freq - This parameter contains the number of packets that must
4521 * be received before a sack is sent without waiting for the delay
4522 * timer to expire. The default value for this is 2, setting this
4523 * value to 1 will disable the delayed sack algorithm.
4525 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4526 char __user *optval,
4529 struct sctp_sack_info params;
4530 struct sctp_association *asoc = NULL;
4531 struct sctp_sock *sp = sctp_sk(sk);
4533 if (len >= sizeof(struct sctp_sack_info)) {
4534 len = sizeof(struct sctp_sack_info);
4536 if (copy_from_user(¶ms, optval, len))
4538 } else if (len == sizeof(struct sctp_assoc_value)) {
4539 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4540 pr_warn("Use struct sctp_sack_info instead\n");
4541 if (copy_from_user(¶ms, optval, len))
4546 /* Get association, if sack_assoc_id != 0 and the socket is a one
4547 * to many style socket, and an association was not found, then
4548 * the id was invalid.
4550 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4551 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4555 /* Fetch association values. */
4556 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4557 params.sack_delay = jiffies_to_msecs(
4559 params.sack_freq = asoc->sackfreq;
4562 params.sack_delay = 0;
4563 params.sack_freq = 1;
4566 /* Fetch socket values. */
4567 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4568 params.sack_delay = sp->sackdelay;
4569 params.sack_freq = sp->sackfreq;
4571 params.sack_delay = 0;
4572 params.sack_freq = 1;
4576 if (copy_to_user(optval, ¶ms, len))
4579 if (put_user(len, optlen))
4585 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4587 * Applications can specify protocol parameters for the default association
4588 * initialization. The option name argument to setsockopt() and getsockopt()
4591 * Setting initialization parameters is effective only on an unconnected
4592 * socket (for UDP-style sockets only future associations are effected
4593 * by the change). With TCP-style sockets, this option is inherited by
4594 * sockets derived from a listener socket.
4596 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4598 if (len < sizeof(struct sctp_initmsg))
4600 len = sizeof(struct sctp_initmsg);
4601 if (put_user(len, optlen))
4603 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4609 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4610 char __user *optval, int __user *optlen)
4612 struct sctp_association *asoc;
4614 struct sctp_getaddrs getaddrs;
4615 struct sctp_transport *from;
4617 union sctp_addr temp;
4618 struct sctp_sock *sp = sctp_sk(sk);
4623 if (len < sizeof(struct sctp_getaddrs))
4626 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4629 /* For UDP-style sockets, id specifies the association to query. */
4630 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4634 to = optval + offsetof(struct sctp_getaddrs,addrs);
4635 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4637 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4639 memcpy(&temp, &from->ipaddr, sizeof(temp));
4640 addrlen = sctp_get_pf_specific(sk->sk_family)
4641 ->addr_to_user(sp, &temp);
4642 if (space_left < addrlen)
4644 if (copy_to_user(to, &temp, addrlen))
4648 space_left -= addrlen;
4651 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4653 bytes_copied = ((char __user *)to) - optval;
4654 if (put_user(bytes_copied, optlen))
4660 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4661 size_t space_left, int *bytes_copied)
4663 struct sctp_sockaddr_entry *addr;
4664 union sctp_addr temp;
4669 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4673 if ((PF_INET == sk->sk_family) &&
4674 (AF_INET6 == addr->a.sa.sa_family))
4676 if ((PF_INET6 == sk->sk_family) &&
4677 inet_v6_ipv6only(sk) &&
4678 (AF_INET == addr->a.sa.sa_family))
4680 memcpy(&temp, &addr->a, sizeof(temp));
4681 if (!temp.v4.sin_port)
4682 temp.v4.sin_port = htons(port);
4684 addrlen = sctp_get_pf_specific(sk->sk_family)
4685 ->addr_to_user(sctp_sk(sk), &temp);
4687 if (space_left < addrlen) {
4691 memcpy(to, &temp, addrlen);
4695 space_left -= addrlen;
4696 *bytes_copied += addrlen;
4704 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4705 char __user *optval, int __user *optlen)
4707 struct sctp_bind_addr *bp;
4708 struct sctp_association *asoc;
4710 struct sctp_getaddrs getaddrs;
4711 struct sctp_sockaddr_entry *addr;
4713 union sctp_addr temp;
4714 struct sctp_sock *sp = sctp_sk(sk);
4718 int bytes_copied = 0;
4722 if (len < sizeof(struct sctp_getaddrs))
4725 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4729 * For UDP-style sockets, id specifies the association to query.
4730 * If the id field is set to the value '0' then the locally bound
4731 * addresses are returned without regard to any particular
4734 if (0 == getaddrs.assoc_id) {
4735 bp = &sctp_sk(sk)->ep->base.bind_addr;
4737 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4740 bp = &asoc->base.bind_addr;
4743 to = optval + offsetof(struct sctp_getaddrs,addrs);
4744 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4746 addrs = kmalloc(space_left, GFP_KERNEL);
4750 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4751 * addresses from the global local address list.
4753 if (sctp_list_single_entry(&bp->address_list)) {
4754 addr = list_entry(bp->address_list.next,
4755 struct sctp_sockaddr_entry, list);
4756 if (sctp_is_any(sk, &addr->a)) {
4757 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4758 space_left, &bytes_copied);
4768 /* Protection on the bound address list is not needed since
4769 * in the socket option context we hold a socket lock and
4770 * thus the bound address list can't change.
4772 list_for_each_entry(addr, &bp->address_list, list) {
4773 memcpy(&temp, &addr->a, sizeof(temp));
4774 addrlen = sctp_get_pf_specific(sk->sk_family)
4775 ->addr_to_user(sp, &temp);
4776 if (space_left < addrlen) {
4777 err = -ENOMEM; /*fixme: right error?*/
4780 memcpy(buf, &temp, addrlen);
4782 bytes_copied += addrlen;
4784 space_left -= addrlen;
4788 if (copy_to_user(to, addrs, bytes_copied)) {
4792 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4796 if (put_user(bytes_copied, optlen))
4803 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4805 * Requests that the local SCTP stack use the enclosed peer address as
4806 * the association primary. The enclosed address must be one of the
4807 * association peer's addresses.
4809 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4810 char __user *optval, int __user *optlen)
4812 struct sctp_prim prim;
4813 struct sctp_association *asoc;
4814 struct sctp_sock *sp = sctp_sk(sk);
4816 if (len < sizeof(struct sctp_prim))
4819 len = sizeof(struct sctp_prim);
4821 if (copy_from_user(&prim, optval, len))
4824 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4828 if (!asoc->peer.primary_path)
4831 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4832 asoc->peer.primary_path->af_specific->sockaddr_len);
4834 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
4835 (union sctp_addr *)&prim.ssp_addr);
4837 if (put_user(len, optlen))
4839 if (copy_to_user(optval, &prim, len))
4846 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4848 * Requests that the local endpoint set the specified Adaptation Layer
4849 * Indication parameter for all future INIT and INIT-ACK exchanges.
4851 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4852 char __user *optval, int __user *optlen)
4854 struct sctp_setadaptation adaptation;
4856 if (len < sizeof(struct sctp_setadaptation))
4859 len = sizeof(struct sctp_setadaptation);
4861 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4863 if (put_user(len, optlen))
4865 if (copy_to_user(optval, &adaptation, len))
4873 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4875 * Applications that wish to use the sendto() system call may wish to
4876 * specify a default set of parameters that would normally be supplied
4877 * through the inclusion of ancillary data. This socket option allows
4878 * such an application to set the default sctp_sndrcvinfo structure.
4881 * The application that wishes to use this socket option simply passes
4882 * in to this call the sctp_sndrcvinfo structure defined in Section
4883 * 5.2.2) The input parameters accepted by this call include
4884 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4885 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4886 * to this call if the caller is using the UDP model.
4888 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4890 static int sctp_getsockopt_default_send_param(struct sock *sk,
4891 int len, char __user *optval,
4894 struct sctp_sndrcvinfo info;
4895 struct sctp_association *asoc;
4896 struct sctp_sock *sp = sctp_sk(sk);
4898 if (len < sizeof(struct sctp_sndrcvinfo))
4901 len = sizeof(struct sctp_sndrcvinfo);
4903 if (copy_from_user(&info, optval, len))
4906 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4907 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4911 info.sinfo_stream = asoc->default_stream;
4912 info.sinfo_flags = asoc->default_flags;
4913 info.sinfo_ppid = asoc->default_ppid;
4914 info.sinfo_context = asoc->default_context;
4915 info.sinfo_timetolive = asoc->default_timetolive;
4917 info.sinfo_stream = sp->default_stream;
4918 info.sinfo_flags = sp->default_flags;
4919 info.sinfo_ppid = sp->default_ppid;
4920 info.sinfo_context = sp->default_context;
4921 info.sinfo_timetolive = sp->default_timetolive;
4924 if (put_user(len, optlen))
4926 if (copy_to_user(optval, &info, len))
4934 * 7.1.5 SCTP_NODELAY
4936 * Turn on/off any Nagle-like algorithm. This means that packets are
4937 * generally sent as soon as possible and no unnecessary delays are
4938 * introduced, at the cost of more packets in the network. Expects an
4939 * integer boolean flag.
4942 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4943 char __user *optval, int __user *optlen)
4947 if (len < sizeof(int))
4951 val = (sctp_sk(sk)->nodelay == 1);
4952 if (put_user(len, optlen))
4954 if (copy_to_user(optval, &val, len))
4961 * 7.1.1 SCTP_RTOINFO
4963 * The protocol parameters used to initialize and bound retransmission
4964 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4965 * and modify these parameters.
4966 * All parameters are time values, in milliseconds. A value of 0, when
4967 * modifying the parameters, indicates that the current value should not
4971 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4972 char __user *optval,
4973 int __user *optlen) {
4974 struct sctp_rtoinfo rtoinfo;
4975 struct sctp_association *asoc;
4977 if (len < sizeof (struct sctp_rtoinfo))
4980 len = sizeof(struct sctp_rtoinfo);
4982 if (copy_from_user(&rtoinfo, optval, len))
4985 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4987 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4990 /* Values corresponding to the specific association. */
4992 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4993 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4994 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4996 /* Values corresponding to the endpoint. */
4997 struct sctp_sock *sp = sctp_sk(sk);
4999 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5000 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5001 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5004 if (put_user(len, optlen))
5007 if (copy_to_user(optval, &rtoinfo, len))
5015 * 7.1.2 SCTP_ASSOCINFO
5017 * This option is used to tune the maximum retransmission attempts
5018 * of the association.
5019 * Returns an error if the new association retransmission value is
5020 * greater than the sum of the retransmission value of the peer.
5021 * See [SCTP] for more information.
5024 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5025 char __user *optval,
5029 struct sctp_assocparams assocparams;
5030 struct sctp_association *asoc;
5031 struct list_head *pos;
5034 if (len < sizeof (struct sctp_assocparams))
5037 len = sizeof(struct sctp_assocparams);
5039 if (copy_from_user(&assocparams, optval, len))
5042 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5044 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5047 /* Values correspoinding to the specific association */
5049 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5050 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5051 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5052 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
5054 (asoc->cookie_life.tv_usec
5057 list_for_each(pos, &asoc->peer.transport_addr_list) {
5061 assocparams.sasoc_number_peer_destinations = cnt;
5063 /* Values corresponding to the endpoint */
5064 struct sctp_sock *sp = sctp_sk(sk);
5066 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5067 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5068 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5069 assocparams.sasoc_cookie_life =
5070 sp->assocparams.sasoc_cookie_life;
5071 assocparams.sasoc_number_peer_destinations =
5073 sasoc_number_peer_destinations;
5076 if (put_user(len, optlen))
5079 if (copy_to_user(optval, &assocparams, len))
5086 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5088 * This socket option is a boolean flag which turns on or off mapped V4
5089 * addresses. If this option is turned on and the socket is type
5090 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5091 * If this option is turned off, then no mapping will be done of V4
5092 * addresses and a user will receive both PF_INET6 and PF_INET type
5093 * addresses on the socket.
5095 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5096 char __user *optval, int __user *optlen)
5099 struct sctp_sock *sp = sctp_sk(sk);
5101 if (len < sizeof(int))
5106 if (put_user(len, optlen))
5108 if (copy_to_user(optval, &val, len))
5115 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5116 * (chapter and verse is quoted at sctp_setsockopt_context())
5118 static int sctp_getsockopt_context(struct sock *sk, int len,
5119 char __user *optval, int __user *optlen)
5121 struct sctp_assoc_value params;
5122 struct sctp_sock *sp;
5123 struct sctp_association *asoc;
5125 if (len < sizeof(struct sctp_assoc_value))
5128 len = sizeof(struct sctp_assoc_value);
5130 if (copy_from_user(¶ms, optval, len))
5135 if (params.assoc_id != 0) {
5136 asoc = sctp_id2assoc(sk, params.assoc_id);
5139 params.assoc_value = asoc->default_rcv_context;
5141 params.assoc_value = sp->default_rcv_context;
5144 if (put_user(len, optlen))
5146 if (copy_to_user(optval, ¶ms, len))
5153 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5154 * This option will get or set the maximum size to put in any outgoing
5155 * SCTP DATA chunk. If a message is larger than this size it will be
5156 * fragmented by SCTP into the specified size. Note that the underlying
5157 * SCTP implementation may fragment into smaller sized chunks when the
5158 * PMTU of the underlying association is smaller than the value set by
5159 * the user. The default value for this option is '0' which indicates
5160 * the user is NOT limiting fragmentation and only the PMTU will effect
5161 * SCTP's choice of DATA chunk size. Note also that values set larger
5162 * than the maximum size of an IP datagram will effectively let SCTP
5163 * control fragmentation (i.e. the same as setting this option to 0).
5165 * The following structure is used to access and modify this parameter:
5167 * struct sctp_assoc_value {
5168 * sctp_assoc_t assoc_id;
5169 * uint32_t assoc_value;
5172 * assoc_id: This parameter is ignored for one-to-one style sockets.
5173 * For one-to-many style sockets this parameter indicates which
5174 * association the user is performing an action upon. Note that if
5175 * this field's value is zero then the endpoints default value is
5176 * changed (effecting future associations only).
5177 * assoc_value: This parameter specifies the maximum size in bytes.
5179 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5180 char __user *optval, int __user *optlen)
5182 struct sctp_assoc_value params;
5183 struct sctp_association *asoc;
5185 if (len == sizeof(int)) {
5186 pr_warn("Use of int in maxseg socket option deprecated\n");
5187 pr_warn("Use struct sctp_assoc_value instead\n");
5188 params.assoc_id = 0;
5189 } else if (len >= sizeof(struct sctp_assoc_value)) {
5190 len = sizeof(struct sctp_assoc_value);
5191 if (copy_from_user(¶ms, optval, sizeof(params)))
5196 asoc = sctp_id2assoc(sk, params.assoc_id);
5197 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5201 params.assoc_value = asoc->frag_point;
5203 params.assoc_value = sctp_sk(sk)->user_frag;
5205 if (put_user(len, optlen))
5207 if (len == sizeof(int)) {
5208 if (copy_to_user(optval, ¶ms.assoc_value, len))
5211 if (copy_to_user(optval, ¶ms, len))
5219 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5220 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5222 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5223 char __user *optval, int __user *optlen)
5227 if (len < sizeof(int))
5232 val = sctp_sk(sk)->frag_interleave;
5233 if (put_user(len, optlen))
5235 if (copy_to_user(optval, &val, len))
5242 * 7.1.25. Set or Get the sctp partial delivery point
5243 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5245 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5246 char __user *optval,
5251 if (len < sizeof(u32))
5256 val = sctp_sk(sk)->pd_point;
5257 if (put_user(len, optlen))
5259 if (copy_to_user(optval, &val, len))
5266 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5267 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5269 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5270 char __user *optval,
5273 struct sctp_assoc_value params;
5274 struct sctp_sock *sp;
5275 struct sctp_association *asoc;
5277 if (len == sizeof(int)) {
5278 pr_warn("Use of int in max_burst socket option deprecated\n");
5279 pr_warn("Use struct sctp_assoc_value instead\n");
5280 params.assoc_id = 0;
5281 } else if (len >= sizeof(struct sctp_assoc_value)) {
5282 len = sizeof(struct sctp_assoc_value);
5283 if (copy_from_user(¶ms, optval, len))
5290 if (params.assoc_id != 0) {
5291 asoc = sctp_id2assoc(sk, params.assoc_id);
5294 params.assoc_value = asoc->max_burst;
5296 params.assoc_value = sp->max_burst;
5298 if (len == sizeof(int)) {
5299 if (copy_to_user(optval, ¶ms.assoc_value, len))
5302 if (copy_to_user(optval, ¶ms, len))
5310 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5311 char __user *optval, int __user *optlen)
5313 struct sctp_hmacalgo __user *p = (void __user *)optval;
5314 struct sctp_hmac_algo_param *hmacs;
5319 if (!sctp_auth_enable)
5322 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5323 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5325 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5328 len = sizeof(struct sctp_hmacalgo) + data_len;
5329 num_idents = data_len / sizeof(u16);
5331 if (put_user(len, optlen))
5333 if (put_user(num_idents, &p->shmac_num_idents))
5335 for (i = 0; i < num_idents; i++) {
5336 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
5338 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
5344 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5345 char __user *optval, int __user *optlen)
5347 struct sctp_authkeyid val;
5348 struct sctp_association *asoc;
5350 if (!sctp_auth_enable)
5353 if (len < sizeof(struct sctp_authkeyid))
5355 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5358 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5359 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5363 val.scact_keynumber = asoc->active_key_id;
5365 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5367 len = sizeof(struct sctp_authkeyid);
5368 if (put_user(len, optlen))
5370 if (copy_to_user(optval, &val, len))
5376 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5377 char __user *optval, int __user *optlen)
5379 struct sctp_authchunks __user *p = (void __user *)optval;
5380 struct sctp_authchunks val;
5381 struct sctp_association *asoc;
5382 struct sctp_chunks_param *ch;
5386 if (!sctp_auth_enable)
5389 if (len < sizeof(struct sctp_authchunks))
5392 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5395 to = p->gauth_chunks;
5396 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5400 ch = asoc->peer.peer_chunks;
5404 /* See if the user provided enough room for all the data */
5405 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5406 if (len < num_chunks)
5409 if (copy_to_user(to, ch->chunks, num_chunks))
5412 len = sizeof(struct sctp_authchunks) + num_chunks;
5413 if (put_user(len, optlen)) return -EFAULT;
5414 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5419 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5420 char __user *optval, int __user *optlen)
5422 struct sctp_authchunks __user *p = (void __user *)optval;
5423 struct sctp_authchunks val;
5424 struct sctp_association *asoc;
5425 struct sctp_chunks_param *ch;
5429 if (!sctp_auth_enable)
5432 if (len < sizeof(struct sctp_authchunks))
5435 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5438 to = p->gauth_chunks;
5439 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5440 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5444 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5446 ch = sctp_sk(sk)->ep->auth_chunk_list;
5451 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5452 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5455 if (copy_to_user(to, ch->chunks, num_chunks))
5458 len = sizeof(struct sctp_authchunks) + num_chunks;
5459 if (put_user(len, optlen))
5461 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5468 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5469 * This option gets the current number of associations that are attached
5470 * to a one-to-many style socket. The option value is an uint32_t.
5472 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5473 char __user *optval, int __user *optlen)
5475 struct sctp_sock *sp = sctp_sk(sk);
5476 struct sctp_association *asoc;
5479 if (sctp_style(sk, TCP))
5482 if (len < sizeof(u32))
5487 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5491 if (put_user(len, optlen))
5493 if (copy_to_user(optval, &val, len))
5500 * 8.1.23 SCTP_AUTO_ASCONF
5501 * See the corresponding setsockopt entry as description
5503 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5504 char __user *optval, int __user *optlen)
5508 if (len < sizeof(int))
5512 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5514 if (put_user(len, optlen))
5516 if (copy_to_user(optval, &val, len))
5522 * 8.2.6. Get the Current Identifiers of Associations
5523 * (SCTP_GET_ASSOC_ID_LIST)
5525 * This option gets the current list of SCTP association identifiers of
5526 * the SCTP associations handled by a one-to-many style socket.
5528 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5529 char __user *optval, int __user *optlen)
5531 struct sctp_sock *sp = sctp_sk(sk);
5532 struct sctp_association *asoc;
5533 struct sctp_assoc_ids *ids;
5536 if (sctp_style(sk, TCP))
5539 if (len < sizeof(struct sctp_assoc_ids))
5542 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5546 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5549 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5551 ids = kmalloc(len, GFP_KERNEL);
5555 ids->gaids_number_of_ids = num;
5557 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5558 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5561 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5570 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5571 char __user *optval, int __user *optlen)
5576 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5579 /* I can hardly begin to describe how wrong this is. This is
5580 * so broken as to be worse than useless. The API draft
5581 * REALLY is NOT helpful here... I am not convinced that the
5582 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5583 * are at all well-founded.
5585 if (level != SOL_SCTP) {
5586 struct sctp_af *af = sctp_sk(sk)->pf->af;
5588 retval = af->getsockopt(sk, level, optname, optval, optlen);
5592 if (get_user(len, optlen))
5602 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5604 case SCTP_DISABLE_FRAGMENTS:
5605 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5609 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5611 case SCTP_AUTOCLOSE:
5612 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5614 case SCTP_SOCKOPT_PEELOFF:
5615 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5617 case SCTP_PEER_ADDR_PARAMS:
5618 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5621 case SCTP_DELAYED_SACK:
5622 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5626 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5628 case SCTP_GET_PEER_ADDRS:
5629 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5632 case SCTP_GET_LOCAL_ADDRS:
5633 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5636 case SCTP_SOCKOPT_CONNECTX3:
5637 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5639 case SCTP_DEFAULT_SEND_PARAM:
5640 retval = sctp_getsockopt_default_send_param(sk, len,
5643 case SCTP_PRIMARY_ADDR:
5644 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5647 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5650 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5652 case SCTP_ASSOCINFO:
5653 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5655 case SCTP_I_WANT_MAPPED_V4_ADDR:
5656 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5659 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5661 case SCTP_GET_PEER_ADDR_INFO:
5662 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5665 case SCTP_ADAPTATION_LAYER:
5666 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5670 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5672 case SCTP_FRAGMENT_INTERLEAVE:
5673 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5676 case SCTP_PARTIAL_DELIVERY_POINT:
5677 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5680 case SCTP_MAX_BURST:
5681 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5684 case SCTP_AUTH_CHUNK:
5685 case SCTP_AUTH_DELETE_KEY:
5686 retval = -EOPNOTSUPP;
5688 case SCTP_HMAC_IDENT:
5689 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5691 case SCTP_AUTH_ACTIVE_KEY:
5692 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5694 case SCTP_PEER_AUTH_CHUNKS:
5695 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5698 case SCTP_LOCAL_AUTH_CHUNKS:
5699 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5702 case SCTP_GET_ASSOC_NUMBER:
5703 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5705 case SCTP_GET_ASSOC_ID_LIST:
5706 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5708 case SCTP_AUTO_ASCONF:
5709 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5712 retval = -ENOPROTOOPT;
5716 sctp_release_sock(sk);
5720 static void sctp_hash(struct sock *sk)
5725 static void sctp_unhash(struct sock *sk)
5730 /* Check if port is acceptable. Possibly find first available port.
5732 * The port hash table (contained in the 'global' SCTP protocol storage
5733 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5734 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5735 * list (the list number is the port number hashed out, so as you
5736 * would expect from a hash function, all the ports in a given list have
5737 * such a number that hashes out to the same list number; you were
5738 * expecting that, right?); so each list has a set of ports, with a
5739 * link to the socket (struct sock) that uses it, the port number and
5740 * a fastreuse flag (FIXME: NPI ipg).
5742 static struct sctp_bind_bucket *sctp_bucket_create(
5743 struct sctp_bind_hashbucket *head, unsigned short snum);
5745 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5747 struct sctp_bind_hashbucket *head; /* hash list */
5748 struct sctp_bind_bucket *pp; /* hash list port iterator */
5749 struct hlist_node *node;
5750 unsigned short snum;
5753 snum = ntohs(addr->v4.sin_port);
5755 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5756 sctp_local_bh_disable();
5759 /* Search for an available port. */
5760 int low, high, remaining, index;
5763 inet_get_local_port_range(&low, &high);
5764 remaining = (high - low) + 1;
5765 rover = net_random() % remaining + low;
5769 if ((rover < low) || (rover > high))
5771 if (inet_is_reserved_local_port(rover))
5773 index = sctp_phashfn(rover);
5774 head = &sctp_port_hashtable[index];
5775 sctp_spin_lock(&head->lock);
5776 sctp_for_each_hentry(pp, node, &head->chain)
5777 if (pp->port == rover)
5781 sctp_spin_unlock(&head->lock);
5782 } while (--remaining > 0);
5784 /* Exhausted local port range during search? */
5789 /* OK, here is the one we will use. HEAD (the port
5790 * hash table list entry) is non-NULL and we hold it's
5795 /* We are given an specific port number; we verify
5796 * that it is not being used. If it is used, we will
5797 * exahust the search in the hash list corresponding
5798 * to the port number (snum) - we detect that with the
5799 * port iterator, pp being NULL.
5801 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5802 sctp_spin_lock(&head->lock);
5803 sctp_for_each_hentry(pp, node, &head->chain) {
5804 if (pp->port == snum)
5811 if (!hlist_empty(&pp->owner)) {
5812 /* We had a port hash table hit - there is an
5813 * available port (pp != NULL) and it is being
5814 * used by other socket (pp->owner not empty); that other
5815 * socket is going to be sk2.
5817 int reuse = sk->sk_reuse;
5820 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5821 if (pp->fastreuse && sk->sk_reuse &&
5822 sk->sk_state != SCTP_SS_LISTENING)
5825 /* Run through the list of sockets bound to the port
5826 * (pp->port) [via the pointers bind_next and
5827 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5828 * we get the endpoint they describe and run through
5829 * the endpoint's list of IP (v4 or v6) addresses,
5830 * comparing each of the addresses with the address of
5831 * the socket sk. If we find a match, then that means
5832 * that this port/socket (sk) combination are already
5835 sk_for_each_bound(sk2, node, &pp->owner) {
5836 struct sctp_endpoint *ep2;
5837 ep2 = sctp_sk(sk2)->ep;
5840 (reuse && sk2->sk_reuse &&
5841 sk2->sk_state != SCTP_SS_LISTENING))
5844 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5845 sctp_sk(sk2), sctp_sk(sk))) {
5850 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5853 /* If there was a hash table miss, create a new port. */
5855 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5858 /* In either case (hit or miss), make sure fastreuse is 1 only
5859 * if sk->sk_reuse is too (that is, if the caller requested
5860 * SO_REUSEADDR on this socket -sk-).
5862 if (hlist_empty(&pp->owner)) {
5863 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5867 } else if (pp->fastreuse &&
5868 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5871 /* We are set, so fill up all the data in the hash table
5872 * entry, tie the socket list information with the rest of the
5873 * sockets FIXME: Blurry, NPI (ipg).
5876 if (!sctp_sk(sk)->bind_hash) {
5877 inet_sk(sk)->inet_num = snum;
5878 sk_add_bind_node(sk, &pp->owner);
5879 sctp_sk(sk)->bind_hash = pp;
5884 sctp_spin_unlock(&head->lock);
5887 sctp_local_bh_enable();
5891 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5892 * port is requested.
5894 static int sctp_get_port(struct sock *sk, unsigned short snum)
5897 union sctp_addr addr;
5898 struct sctp_af *af = sctp_sk(sk)->pf->af;
5900 /* Set up a dummy address struct from the sk. */
5901 af->from_sk(&addr, sk);
5902 addr.v4.sin_port = htons(snum);
5904 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5905 ret = sctp_get_port_local(sk, &addr);
5911 * Move a socket to LISTENING state.
5913 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5915 struct sctp_sock *sp = sctp_sk(sk);
5916 struct sctp_endpoint *ep = sp->ep;
5917 struct crypto_hash *tfm = NULL;
5919 /* Allocate HMAC for generating cookie. */
5920 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5921 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5923 if (net_ratelimit()) {
5924 pr_info("failed to load transform for %s: %ld\n",
5925 sctp_hmac_alg, PTR_ERR(tfm));
5929 sctp_sk(sk)->hmac = tfm;
5933 * If a bind() or sctp_bindx() is not called prior to a listen()
5934 * call that allows new associations to be accepted, the system
5935 * picks an ephemeral port and will choose an address set equivalent
5936 * to binding with a wildcard address.
5938 * This is not currently spelled out in the SCTP sockets
5939 * extensions draft, but follows the practice as seen in TCP
5943 sk->sk_state = SCTP_SS_LISTENING;
5944 if (!ep->base.bind_addr.port) {
5945 if (sctp_autobind(sk))
5948 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
5949 sk->sk_state = SCTP_SS_CLOSED;
5954 sk->sk_max_ack_backlog = backlog;
5955 sctp_hash_endpoint(ep);
5960 * 4.1.3 / 5.1.3 listen()
5962 * By default, new associations are not accepted for UDP style sockets.
5963 * An application uses listen() to mark a socket as being able to
5964 * accept new associations.
5966 * On TCP style sockets, applications use listen() to ready the SCTP
5967 * endpoint for accepting inbound associations.
5969 * On both types of endpoints a backlog of '0' disables listening.
5971 * Move a socket to LISTENING state.
5973 int sctp_inet_listen(struct socket *sock, int backlog)
5975 struct sock *sk = sock->sk;
5976 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5979 if (unlikely(backlog < 0))
5984 /* Peeled-off sockets are not allowed to listen(). */
5985 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
5988 if (sock->state != SS_UNCONNECTED)
5991 /* If backlog is zero, disable listening. */
5993 if (sctp_sstate(sk, CLOSED))
5997 sctp_unhash_endpoint(ep);
5998 sk->sk_state = SCTP_SS_CLOSED;
6000 sctp_sk(sk)->bind_hash->fastreuse = 1;
6004 /* If we are already listening, just update the backlog */
6005 if (sctp_sstate(sk, LISTENING))
6006 sk->sk_max_ack_backlog = backlog;
6008 err = sctp_listen_start(sk, backlog);
6015 sctp_release_sock(sk);
6020 * This function is done by modeling the current datagram_poll() and the
6021 * tcp_poll(). Note that, based on these implementations, we don't
6022 * lock the socket in this function, even though it seems that,
6023 * ideally, locking or some other mechanisms can be used to ensure
6024 * the integrity of the counters (sndbuf and wmem_alloc) used
6025 * in this place. We assume that we don't need locks either until proven
6028 * Another thing to note is that we include the Async I/O support
6029 * here, again, by modeling the current TCP/UDP code. We don't have
6030 * a good way to test with it yet.
6032 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6034 struct sock *sk = sock->sk;
6035 struct sctp_sock *sp = sctp_sk(sk);
6038 poll_wait(file, sk_sleep(sk), wait);
6040 /* A TCP-style listening socket becomes readable when the accept queue
6043 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6044 return (!list_empty(&sp->ep->asocs)) ?
6045 (POLLIN | POLLRDNORM) : 0;
6049 /* Is there any exceptional events? */
6050 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6052 if (sk->sk_shutdown & RCV_SHUTDOWN)
6053 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6054 if (sk->sk_shutdown == SHUTDOWN_MASK)
6057 /* Is it readable? Reconsider this code with TCP-style support. */
6058 if (!skb_queue_empty(&sk->sk_receive_queue))
6059 mask |= POLLIN | POLLRDNORM;
6061 /* The association is either gone or not ready. */
6062 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6065 /* Is it writable? */
6066 if (sctp_writeable(sk)) {
6067 mask |= POLLOUT | POLLWRNORM;
6069 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6071 * Since the socket is not locked, the buffer
6072 * might be made available after the writeable check and
6073 * before the bit is set. This could cause a lost I/O
6074 * signal. tcp_poll() has a race breaker for this race
6075 * condition. Based on their implementation, we put
6076 * in the following code to cover it as well.
6078 if (sctp_writeable(sk))
6079 mask |= POLLOUT | POLLWRNORM;
6084 /********************************************************************
6085 * 2nd Level Abstractions
6086 ********************************************************************/
6088 static struct sctp_bind_bucket *sctp_bucket_create(
6089 struct sctp_bind_hashbucket *head, unsigned short snum)
6091 struct sctp_bind_bucket *pp;
6093 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6095 SCTP_DBG_OBJCNT_INC(bind_bucket);
6098 INIT_HLIST_HEAD(&pp->owner);
6099 hlist_add_head(&pp->node, &head->chain);
6104 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6105 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6107 if (pp && hlist_empty(&pp->owner)) {
6108 __hlist_del(&pp->node);
6109 kmem_cache_free(sctp_bucket_cachep, pp);
6110 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6114 /* Release this socket's reference to a local port. */
6115 static inline void __sctp_put_port(struct sock *sk)
6117 struct sctp_bind_hashbucket *head =
6118 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->inet_num)];
6119 struct sctp_bind_bucket *pp;
6121 sctp_spin_lock(&head->lock);
6122 pp = sctp_sk(sk)->bind_hash;
6123 __sk_del_bind_node(sk);
6124 sctp_sk(sk)->bind_hash = NULL;
6125 inet_sk(sk)->inet_num = 0;
6126 sctp_bucket_destroy(pp);
6127 sctp_spin_unlock(&head->lock);
6130 void sctp_put_port(struct sock *sk)
6132 sctp_local_bh_disable();
6133 __sctp_put_port(sk);
6134 sctp_local_bh_enable();
6138 * The system picks an ephemeral port and choose an address set equivalent
6139 * to binding with a wildcard address.
6140 * One of those addresses will be the primary address for the association.
6141 * This automatically enables the multihoming capability of SCTP.
6143 static int sctp_autobind(struct sock *sk)
6145 union sctp_addr autoaddr;
6149 /* Initialize a local sockaddr structure to INADDR_ANY. */
6150 af = sctp_sk(sk)->pf->af;
6152 port = htons(inet_sk(sk)->inet_num);
6153 af->inaddr_any(&autoaddr, port);
6155 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6158 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6161 * 4.2 The cmsghdr Structure *
6163 * When ancillary data is sent or received, any number of ancillary data
6164 * objects can be specified by the msg_control and msg_controllen members of
6165 * the msghdr structure, because each object is preceded by
6166 * a cmsghdr structure defining the object's length (the cmsg_len member).
6167 * Historically Berkeley-derived implementations have passed only one object
6168 * at a time, but this API allows multiple objects to be
6169 * passed in a single call to sendmsg() or recvmsg(). The following example
6170 * shows two ancillary data objects in a control buffer.
6172 * |<--------------------------- msg_controllen -------------------------->|
6175 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6177 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6180 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6182 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6185 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6186 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6188 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6190 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6197 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
6198 sctp_cmsgs_t *cmsgs)
6200 struct cmsghdr *cmsg;
6201 struct msghdr *my_msg = (struct msghdr *)msg;
6203 for (cmsg = CMSG_FIRSTHDR(msg);
6205 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6206 if (!CMSG_OK(my_msg, cmsg))
6209 /* Should we parse this header or ignore? */
6210 if (cmsg->cmsg_level != IPPROTO_SCTP)
6213 /* Strictly check lengths following example in SCM code. */
6214 switch (cmsg->cmsg_type) {
6216 /* SCTP Socket API Extension
6217 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6219 * This cmsghdr structure provides information for
6220 * initializing new SCTP associations with sendmsg().
6221 * The SCTP_INITMSG socket option uses this same data
6222 * structure. This structure is not used for
6225 * cmsg_level cmsg_type cmsg_data[]
6226 * ------------ ------------ ----------------------
6227 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6229 if (cmsg->cmsg_len !=
6230 CMSG_LEN(sizeof(struct sctp_initmsg)))
6232 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6236 /* SCTP Socket API Extension
6237 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6239 * This cmsghdr structure specifies SCTP options for
6240 * sendmsg() and describes SCTP header information
6241 * about a received message through recvmsg().
6243 * cmsg_level cmsg_type cmsg_data[]
6244 * ------------ ------------ ----------------------
6245 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6247 if (cmsg->cmsg_len !=
6248 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6252 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6254 /* Minimally, validate the sinfo_flags. */
6255 if (cmsgs->info->sinfo_flags &
6256 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6257 SCTP_SACK_IMMEDIATELY |
6258 SCTP_ABORT | SCTP_EOF))
6270 * Wait for a packet..
6271 * Note: This function is the same function as in core/datagram.c
6272 * with a few modifications to make lksctp work.
6274 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6279 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6281 /* Socket errors? */
6282 error = sock_error(sk);
6286 if (!skb_queue_empty(&sk->sk_receive_queue))
6289 /* Socket shut down? */
6290 if (sk->sk_shutdown & RCV_SHUTDOWN)
6293 /* Sequenced packets can come disconnected. If so we report the
6298 /* Is there a good reason to think that we may receive some data? */
6299 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6302 /* Handle signals. */
6303 if (signal_pending(current))
6306 /* Let another process have a go. Since we are going to sleep
6307 * anyway. Note: This may cause odd behaviors if the message
6308 * does not fit in the user's buffer, but this seems to be the
6309 * only way to honor MSG_DONTWAIT realistically.
6311 sctp_release_sock(sk);
6312 *timeo_p = schedule_timeout(*timeo_p);
6316 finish_wait(sk_sleep(sk), &wait);
6320 error = sock_intr_errno(*timeo_p);
6323 finish_wait(sk_sleep(sk), &wait);
6328 /* Receive a datagram.
6329 * Note: This is pretty much the same routine as in core/datagram.c
6330 * with a few changes to make lksctp work.
6332 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6333 int noblock, int *err)
6336 struct sk_buff *skb;
6339 timeo = sock_rcvtimeo(sk, noblock);
6341 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6342 timeo, MAX_SCHEDULE_TIMEOUT);
6345 /* Again only user level code calls this function,
6346 * so nothing interrupt level
6347 * will suddenly eat the receive_queue.
6349 * Look at current nfs client by the way...
6350 * However, this function was correct in any case. 8)
6352 if (flags & MSG_PEEK) {
6353 spin_lock_bh(&sk->sk_receive_queue.lock);
6354 skb = skb_peek(&sk->sk_receive_queue);
6356 atomic_inc(&skb->users);
6357 spin_unlock_bh(&sk->sk_receive_queue.lock);
6359 skb = skb_dequeue(&sk->sk_receive_queue);
6365 /* Caller is allowed not to check sk->sk_err before calling. */
6366 error = sock_error(sk);
6370 if (sk->sk_shutdown & RCV_SHUTDOWN)
6373 /* User doesn't want to wait. */
6377 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6386 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6387 static void __sctp_write_space(struct sctp_association *asoc)
6389 struct sock *sk = asoc->base.sk;
6390 struct socket *sock = sk->sk_socket;
6392 if ((sctp_wspace(asoc) > 0) && sock) {
6393 if (waitqueue_active(&asoc->wait))
6394 wake_up_interruptible(&asoc->wait);
6396 if (sctp_writeable(sk)) {
6397 wait_queue_head_t *wq = sk_sleep(sk);
6399 if (wq && waitqueue_active(wq))
6400 wake_up_interruptible(wq);
6402 /* Note that we try to include the Async I/O support
6403 * here by modeling from the current TCP/UDP code.
6404 * We have not tested with it yet.
6406 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6407 sock_wake_async(sock,
6408 SOCK_WAKE_SPACE, POLL_OUT);
6413 /* Do accounting for the sndbuf space.
6414 * Decrement the used sndbuf space of the corresponding association by the
6415 * data size which was just transmitted(freed).
6417 static void sctp_wfree(struct sk_buff *skb)
6419 struct sctp_association *asoc;
6420 struct sctp_chunk *chunk;
6423 /* Get the saved chunk pointer. */
6424 chunk = *((struct sctp_chunk **)(skb->cb));
6427 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6428 sizeof(struct sk_buff) +
6429 sizeof(struct sctp_chunk);
6431 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6434 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6436 sk->sk_wmem_queued -= skb->truesize;
6437 sk_mem_uncharge(sk, skb->truesize);
6440 __sctp_write_space(asoc);
6442 sctp_association_put(asoc);
6445 /* Do accounting for the receive space on the socket.
6446 * Accounting for the association is done in ulpevent.c
6447 * We set this as a destructor for the cloned data skbs so that
6448 * accounting is done at the correct time.
6450 void sctp_sock_rfree(struct sk_buff *skb)
6452 struct sock *sk = skb->sk;
6453 struct sctp_ulpevent *event = sctp_skb2event(skb);
6455 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6458 * Mimic the behavior of sock_rfree
6460 sk_mem_uncharge(sk, event->rmem_len);
6464 /* Helper function to wait for space in the sndbuf. */
6465 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6468 struct sock *sk = asoc->base.sk;
6469 long current_timeo = *timeo_p;
6473 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6474 asoc, (long)(*timeo_p), msg_len);
6476 /* Increment the association's refcnt. */
6477 sctp_association_hold(asoc);
6479 /* Wait on the association specific sndbuf space. */
6481 prepare_to_wait_exclusive(&asoc->wait, &wait,
6482 TASK_INTERRUPTIBLE);
6485 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6488 if (signal_pending(current))
6489 goto do_interrupted;
6490 if (msg_len <= sctp_wspace(asoc))
6493 /* Let another process have a go. Since we are going
6496 sctp_release_sock(sk);
6497 current_timeo = schedule_timeout(current_timeo);
6499 if (sk != asoc->base.sk)
6502 *timeo_p = current_timeo;
6506 finish_wait(&asoc->wait, &wait);
6508 /* Release the association's refcnt. */
6509 sctp_association_put(asoc);
6518 err = sock_intr_errno(*timeo_p);
6526 void sctp_data_ready(struct sock *sk, int len)
6528 struct socket_wq *wq;
6531 wq = rcu_dereference(sk->sk_wq);
6532 if (wq_has_sleeper(wq))
6533 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6534 POLLRDNORM | POLLRDBAND);
6535 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6539 /* If socket sndbuf has changed, wake up all per association waiters. */
6540 void sctp_write_space(struct sock *sk)
6542 struct sctp_association *asoc;
6544 /* Wake up the tasks in each wait queue. */
6545 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6546 __sctp_write_space(asoc);
6550 /* Is there any sndbuf space available on the socket?
6552 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6553 * associations on the same socket. For a UDP-style socket with
6554 * multiple associations, it is possible for it to be "unwriteable"
6555 * prematurely. I assume that this is acceptable because
6556 * a premature "unwriteable" is better than an accidental "writeable" which
6557 * would cause an unwanted block under certain circumstances. For the 1-1
6558 * UDP-style sockets or TCP-style sockets, this code should work.
6561 static int sctp_writeable(struct sock *sk)
6565 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6571 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6572 * returns immediately with EINPROGRESS.
6574 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6576 struct sock *sk = asoc->base.sk;
6578 long current_timeo = *timeo_p;
6581 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6584 /* Increment the association's refcnt. */
6585 sctp_association_hold(asoc);
6588 prepare_to_wait_exclusive(&asoc->wait, &wait,
6589 TASK_INTERRUPTIBLE);
6592 if (sk->sk_shutdown & RCV_SHUTDOWN)
6594 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6597 if (signal_pending(current))
6598 goto do_interrupted;
6600 if (sctp_state(asoc, ESTABLISHED))
6603 /* Let another process have a go. Since we are going
6606 sctp_release_sock(sk);
6607 current_timeo = schedule_timeout(current_timeo);
6610 *timeo_p = current_timeo;
6614 finish_wait(&asoc->wait, &wait);
6616 /* Release the association's refcnt. */
6617 sctp_association_put(asoc);
6622 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6625 err = -ECONNREFUSED;
6629 err = sock_intr_errno(*timeo_p);
6637 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6639 struct sctp_endpoint *ep;
6643 ep = sctp_sk(sk)->ep;
6647 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6648 TASK_INTERRUPTIBLE);
6650 if (list_empty(&ep->asocs)) {
6651 sctp_release_sock(sk);
6652 timeo = schedule_timeout(timeo);
6657 if (!sctp_sstate(sk, LISTENING))
6661 if (!list_empty(&ep->asocs))
6664 err = sock_intr_errno(timeo);
6665 if (signal_pending(current))
6673 finish_wait(sk_sleep(sk), &wait);
6678 static void sctp_wait_for_close(struct sock *sk, long timeout)
6683 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6684 if (list_empty(&sctp_sk(sk)->ep->asocs))
6686 sctp_release_sock(sk);
6687 timeout = schedule_timeout(timeout);
6689 } while (!signal_pending(current) && timeout);
6691 finish_wait(sk_sleep(sk), &wait);
6694 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6696 struct sk_buff *frag;
6701 /* Don't forget the fragments. */
6702 skb_walk_frags(skb, frag)
6703 sctp_skb_set_owner_r_frag(frag, sk);
6706 sctp_skb_set_owner_r(skb, sk);
6709 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6710 struct sctp_association *asoc)
6712 struct inet_sock *inet = inet_sk(sk);
6713 struct inet_sock *newinet;
6715 newsk->sk_type = sk->sk_type;
6716 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6717 newsk->sk_flags = sk->sk_flags;
6718 newsk->sk_no_check = sk->sk_no_check;
6719 newsk->sk_reuse = sk->sk_reuse;
6721 newsk->sk_shutdown = sk->sk_shutdown;
6722 newsk->sk_destruct = inet_sock_destruct;
6723 newsk->sk_family = sk->sk_family;
6724 newsk->sk_protocol = IPPROTO_SCTP;
6725 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6726 newsk->sk_sndbuf = sk->sk_sndbuf;
6727 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6728 newsk->sk_lingertime = sk->sk_lingertime;
6729 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6730 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6732 newinet = inet_sk(newsk);
6734 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6735 * getsockname() and getpeername()
6737 newinet->inet_sport = inet->inet_sport;
6738 newinet->inet_saddr = inet->inet_saddr;
6739 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6740 newinet->inet_dport = htons(asoc->peer.port);
6741 newinet->pmtudisc = inet->pmtudisc;
6742 newinet->inet_id = asoc->next_tsn ^ jiffies;
6744 newinet->uc_ttl = inet->uc_ttl;
6745 newinet->mc_loop = 1;
6746 newinet->mc_ttl = 1;
6747 newinet->mc_index = 0;
6748 newinet->mc_list = NULL;
6750 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
6751 net_enable_timestamp();
6754 static inline void sctp_copy_descendant(struct sock *sk_to,
6755 const struct sock *sk_from)
6757 int ancestor_size = sizeof(struct inet_sock) +
6758 sizeof(struct sctp_sock) -
6759 offsetof(struct sctp_sock, auto_asconf_list);
6761 if (sk_from->sk_family == PF_INET6)
6762 ancestor_size += sizeof(struct ipv6_pinfo);
6764 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
6767 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6768 * and its messages to the newsk.
6770 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6771 struct sctp_association *assoc,
6772 sctp_socket_type_t type)
6774 struct sctp_sock *oldsp = sctp_sk(oldsk);
6775 struct sctp_sock *newsp = sctp_sk(newsk);
6776 struct sctp_bind_bucket *pp; /* hash list port iterator */
6777 struct sctp_endpoint *newep = newsp->ep;
6778 struct sk_buff *skb, *tmp;
6779 struct sctp_ulpevent *event;
6780 struct sctp_bind_hashbucket *head;
6782 /* Migrate socket buffer sizes and all the socket level options to the
6785 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6786 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6787 /* Brute force copy old sctp opt. */
6788 sctp_copy_descendant(newsk, oldsk);
6790 /* Restore the ep value that was overwritten with the above structure
6796 /* Hook this new socket in to the bind_hash list. */
6797 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->inet_num)];
6798 sctp_local_bh_disable();
6799 sctp_spin_lock(&head->lock);
6800 pp = sctp_sk(oldsk)->bind_hash;
6801 sk_add_bind_node(newsk, &pp->owner);
6802 sctp_sk(newsk)->bind_hash = pp;
6803 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6804 sctp_spin_unlock(&head->lock);
6805 sctp_local_bh_enable();
6807 /* Copy the bind_addr list from the original endpoint to the new
6808 * endpoint so that we can handle restarts properly
6810 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6811 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6813 /* Move any messages in the old socket's receive queue that are for the
6814 * peeled off association to the new socket's receive queue.
6816 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6817 event = sctp_skb2event(skb);
6818 if (event->asoc == assoc) {
6819 __skb_unlink(skb, &oldsk->sk_receive_queue);
6820 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6821 sctp_skb_set_owner_r_frag(skb, newsk);
6825 /* Clean up any messages pending delivery due to partial
6826 * delivery. Three cases:
6827 * 1) No partial deliver; no work.
6828 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6829 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6831 skb_queue_head_init(&newsp->pd_lobby);
6832 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6834 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6835 struct sk_buff_head *queue;
6837 /* Decide which queue to move pd_lobby skbs to. */
6838 if (assoc->ulpq.pd_mode) {
6839 queue = &newsp->pd_lobby;
6841 queue = &newsk->sk_receive_queue;
6843 /* Walk through the pd_lobby, looking for skbs that
6844 * need moved to the new socket.
6846 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6847 event = sctp_skb2event(skb);
6848 if (event->asoc == assoc) {
6849 __skb_unlink(skb, &oldsp->pd_lobby);
6850 __skb_queue_tail(queue, skb);
6851 sctp_skb_set_owner_r_frag(skb, newsk);
6855 /* Clear up any skbs waiting for the partial
6856 * delivery to finish.
6858 if (assoc->ulpq.pd_mode)
6859 sctp_clear_pd(oldsk, NULL);
6863 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6864 sctp_skb_set_owner_r_frag(skb, newsk);
6866 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6867 sctp_skb_set_owner_r_frag(skb, newsk);
6869 /* Set the type of socket to indicate that it is peeled off from the
6870 * original UDP-style socket or created with the accept() call on a
6871 * TCP-style socket..
6875 /* Mark the new socket "in-use" by the user so that any packets
6876 * that may arrive on the association after we've moved it are
6877 * queued to the backlog. This prevents a potential race between
6878 * backlog processing on the old socket and new-packet processing
6879 * on the new socket.
6881 * The caller has just allocated newsk so we can guarantee that other
6882 * paths won't try to lock it and then oldsk.
6884 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6885 sctp_assoc_migrate(assoc, newsk);
6887 /* If the association on the newsk is already closed before accept()
6888 * is called, set RCV_SHUTDOWN flag.
6890 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6891 newsk->sk_shutdown |= RCV_SHUTDOWN;
6893 newsk->sk_state = SCTP_SS_ESTABLISHED;
6894 sctp_release_sock(newsk);
6898 /* This proto struct describes the ULP interface for SCTP. */
6899 struct proto sctp_prot = {
6901 .owner = THIS_MODULE,
6902 .close = sctp_close,
6903 .connect = sctp_connect,
6904 .disconnect = sctp_disconnect,
6905 .accept = sctp_accept,
6906 .ioctl = sctp_ioctl,
6907 .init = sctp_init_sock,
6908 .destroy = sctp_destroy_sock,
6909 .shutdown = sctp_shutdown,
6910 .setsockopt = sctp_setsockopt,
6911 .getsockopt = sctp_getsockopt,
6912 .sendmsg = sctp_sendmsg,
6913 .recvmsg = sctp_recvmsg,
6915 .backlog_rcv = sctp_backlog_rcv,
6917 .unhash = sctp_unhash,
6918 .get_port = sctp_get_port,
6919 .obj_size = sizeof(struct sctp_sock),
6920 .sysctl_mem = sysctl_sctp_mem,
6921 .sysctl_rmem = sysctl_sctp_rmem,
6922 .sysctl_wmem = sysctl_sctp_wmem,
6923 .memory_pressure = &sctp_memory_pressure,
6924 .enter_memory_pressure = sctp_enter_memory_pressure,
6925 .memory_allocated = &sctp_memory_allocated,
6926 .sockets_allocated = &sctp_sockets_allocated,
6929 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6931 #include <net/transp_v6.h>
6932 static void sctp_v6_destroy_sock(struct sock *sk)
6934 sctp_destroy_sock(sk);
6935 inet6_destroy_sock(sk);
6938 struct proto sctpv6_prot = {
6940 .owner = THIS_MODULE,
6941 .close = sctp_close,
6942 .connect = sctp_connect,
6943 .disconnect = sctp_disconnect,
6944 .accept = sctp_accept,
6945 .ioctl = sctp_ioctl,
6946 .init = sctp_init_sock,
6947 .destroy = sctp_v6_destroy_sock,
6948 .shutdown = sctp_shutdown,
6949 .setsockopt = sctp_setsockopt,
6950 .getsockopt = sctp_getsockopt,
6951 .sendmsg = sctp_sendmsg,
6952 .recvmsg = sctp_recvmsg,
6954 .backlog_rcv = sctp_backlog_rcv,
6956 .unhash = sctp_unhash,
6957 .get_port = sctp_get_port,
6958 .obj_size = sizeof(struct sctp6_sock),
6959 .sysctl_mem = sysctl_sctp_mem,
6960 .sysctl_rmem = sysctl_sctp_rmem,
6961 .sysctl_wmem = sysctl_sctp_wmem,
6962 .memory_pressure = &sctp_memory_pressure,
6963 .enter_memory_pressure = sctp_enter_memory_pressure,
6964 .memory_allocated = &sctp_memory_allocated,
6965 .sockets_allocated = &sctp_sockets_allocated,
6967 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */