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,
97 size_t msg_len, struct sock **orig_sk);
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 /* V4 mapped address are really of AF_INET family */
318 if (addr->sa.sa_family == AF_INET6 &&
319 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
320 if (!opt->pf->af_supported(AF_INET, opt))
323 /* Does this PF support this AF? */
324 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
328 /* If we get this far, af is valid. */
329 af = sctp_get_af_specific(addr->sa.sa_family);
331 if (len < af->sockaddr_len)
337 /* Bind a local address either to an endpoint or to an association. */
338 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
340 struct sctp_sock *sp = sctp_sk(sk);
341 struct sctp_endpoint *ep = sp->ep;
342 struct sctp_bind_addr *bp = &ep->base.bind_addr;
347 /* Common sockaddr verification. */
348 af = sctp_sockaddr_af(sp, addr, len);
350 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
355 snum = ntohs(addr->v4.sin_port);
357 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
358 ", port: %d, new port: %d, len: %d)\n",
364 /* PF specific bind() address verification. */
365 if (!sp->pf->bind_verify(sp, addr))
366 return -EADDRNOTAVAIL;
368 /* We must either be unbound, or bind to the same port.
369 * It's OK to allow 0 ports if we are already bound.
370 * We'll just inhert an already bound port in this case
375 else if (snum != bp->port) {
376 SCTP_DEBUG_PRINTK("sctp_do_bind:"
377 " New port %d does not match existing port "
378 "%d.\n", snum, bp->port);
383 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
386 /* See if the address matches any of the addresses we may have
387 * already bound before checking against other endpoints.
389 if (sctp_bind_addr_match(bp, addr, sp))
392 /* Make sure we are allowed to bind here.
393 * The function sctp_get_port_local() does duplicate address
396 addr->v4.sin_port = htons(snum);
397 if ((ret = sctp_get_port_local(sk, addr))) {
401 /* Refresh ephemeral port. */
403 bp->port = inet_sk(sk)->inet_num;
405 /* Add the address to the bind address list.
406 * Use GFP_ATOMIC since BHs will be disabled.
408 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
410 /* Copy back into socket for getsockname() use. */
412 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
413 sp->pf->to_sk_saddr(addr, sk);
419 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
421 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
422 * at any one time. If a sender, after sending an ASCONF chunk, decides
423 * it needs to transfer another ASCONF Chunk, it MUST wait until the
424 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
425 * subsequent ASCONF. Note this restriction binds each side, so at any
426 * time two ASCONF may be in-transit on any given association (one sent
427 * from each endpoint).
429 static int sctp_send_asconf(struct sctp_association *asoc,
430 struct sctp_chunk *chunk)
434 /* If there is an outstanding ASCONF chunk, queue it for later
437 if (asoc->addip_last_asconf) {
438 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
442 /* Hold the chunk until an ASCONF_ACK is received. */
443 sctp_chunk_hold(chunk);
444 retval = sctp_primitive_ASCONF(asoc, chunk);
446 sctp_chunk_free(chunk);
448 asoc->addip_last_asconf = chunk;
454 /* Add a list of addresses as bind addresses to local endpoint or
457 * Basically run through each address specified in the addrs/addrcnt
458 * array/length pair, determine if it is IPv6 or IPv4 and call
459 * sctp_do_bind() on it.
461 * If any of them fails, then the operation will be reversed and the
462 * ones that were added will be removed.
464 * Only sctp_setsockopt_bindx() is supposed to call this function.
466 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
471 struct sockaddr *sa_addr;
474 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
478 for (cnt = 0; cnt < addrcnt; cnt++) {
479 /* The list may contain either IPv4 or IPv6 address;
480 * determine the address length for walking thru the list.
483 af = sctp_get_af_specific(sa_addr->sa_family);
489 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
492 addr_buf += af->sockaddr_len;
496 /* Failed. Cleanup the ones that have been added */
498 sctp_bindx_rem(sk, addrs, cnt);
506 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
507 * associations that are part of the endpoint indicating that a list of local
508 * addresses are added to the endpoint.
510 * If any of the addresses is already in the bind address list of the
511 * association, we do not send the chunk for that association. But it will not
512 * affect other associations.
514 * Only sctp_setsockopt_bindx() is supposed to call this function.
516 static int sctp_send_asconf_add_ip(struct sock *sk,
517 struct sockaddr *addrs,
520 struct sctp_sock *sp;
521 struct sctp_endpoint *ep;
522 struct sctp_association *asoc;
523 struct sctp_bind_addr *bp;
524 struct sctp_chunk *chunk;
525 struct sctp_sockaddr_entry *laddr;
526 union sctp_addr *addr;
527 union sctp_addr saveaddr;
534 if (!sctp_addip_enable)
540 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
541 __func__, sk, addrs, addrcnt);
543 list_for_each_entry(asoc, &ep->asocs, asocs) {
545 if (!asoc->peer.asconf_capable)
548 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
551 if (!sctp_state(asoc, ESTABLISHED))
554 /* Check if any address in the packed array of addresses is
555 * in the bind address list of the association. If so,
556 * do not send the asconf chunk to its peer, but continue with
557 * other associations.
560 for (i = 0; i < addrcnt; i++) {
562 af = sctp_get_af_specific(addr->v4.sin_family);
568 if (sctp_assoc_lookup_laddr(asoc, addr))
571 addr_buf += af->sockaddr_len;
576 /* Use the first valid address in bind addr list of
577 * association as Address Parameter of ASCONF CHUNK.
579 bp = &asoc->base.bind_addr;
580 p = bp->address_list.next;
581 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
582 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
583 addrcnt, SCTP_PARAM_ADD_IP);
589 /* Add the new addresses to the bind address list with
590 * use_as_src set to 0.
593 for (i = 0; i < addrcnt; i++) {
595 af = sctp_get_af_specific(addr->v4.sin_family);
596 memcpy(&saveaddr, addr, af->sockaddr_len);
597 retval = sctp_add_bind_addr(bp, &saveaddr,
598 SCTP_ADDR_NEW, GFP_ATOMIC);
599 addr_buf += af->sockaddr_len;
601 if (asoc->src_out_of_asoc_ok) {
602 struct sctp_transport *trans;
604 list_for_each_entry(trans,
605 &asoc->peer.transport_addr_list, transports) {
606 /* Clear the source and route cache */
607 dst_release(trans->dst);
608 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
609 2*asoc->pathmtu, 4380));
610 trans->ssthresh = asoc->peer.i.a_rwnd;
611 trans->rto = asoc->rto_initial;
612 trans->rtt = trans->srtt = trans->rttvar = 0;
613 sctp_transport_route(trans, NULL,
614 sctp_sk(asoc->base.sk));
617 retval = sctp_send_asconf(asoc, chunk);
624 /* Remove a list of addresses from bind addresses list. Do not remove the
627 * Basically run through each address specified in the addrs/addrcnt
628 * array/length pair, determine if it is IPv6 or IPv4 and call
629 * sctp_del_bind() on it.
631 * If any of them fails, then the operation will be reversed and the
632 * ones that were removed will be added back.
634 * At least one address has to be left; if only one address is
635 * available, the operation will return -EBUSY.
637 * Only sctp_setsockopt_bindx() is supposed to call this function.
639 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
641 struct sctp_sock *sp = sctp_sk(sk);
642 struct sctp_endpoint *ep = sp->ep;
644 struct sctp_bind_addr *bp = &ep->base.bind_addr;
647 union sctp_addr *sa_addr;
650 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
654 for (cnt = 0; cnt < addrcnt; cnt++) {
655 /* If the bind address list is empty or if there is only one
656 * bind address, there is nothing more to be removed (we need
657 * at least one address here).
659 if (list_empty(&bp->address_list) ||
660 (sctp_list_single_entry(&bp->address_list))) {
666 af = sctp_get_af_specific(sa_addr->sa.sa_family);
672 if (!af->addr_valid(sa_addr, sp, NULL)) {
673 retval = -EADDRNOTAVAIL;
677 if (sa_addr->v4.sin_port &&
678 sa_addr->v4.sin_port != htons(bp->port)) {
683 if (!sa_addr->v4.sin_port)
684 sa_addr->v4.sin_port = htons(bp->port);
686 /* FIXME - There is probably a need to check if sk->sk_saddr and
687 * sk->sk_rcv_addr are currently set to one of the addresses to
688 * be removed. This is something which needs to be looked into
689 * when we are fixing the outstanding issues with multi-homing
690 * socket routing and failover schemes. Refer to comments in
691 * sctp_do_bind(). -daisy
693 retval = sctp_del_bind_addr(bp, sa_addr);
695 addr_buf += af->sockaddr_len;
698 /* Failed. Add the ones that has been removed back */
700 sctp_bindx_add(sk, addrs, cnt);
708 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
709 * the associations that are part of the endpoint indicating that a list of
710 * local addresses are removed from the endpoint.
712 * If any of the addresses is already in the bind address list of the
713 * association, we do not send the chunk for that association. But it will not
714 * affect other associations.
716 * Only sctp_setsockopt_bindx() is supposed to call this function.
718 static int sctp_send_asconf_del_ip(struct sock *sk,
719 struct sockaddr *addrs,
722 struct sctp_sock *sp;
723 struct sctp_endpoint *ep;
724 struct sctp_association *asoc;
725 struct sctp_transport *transport;
726 struct sctp_bind_addr *bp;
727 struct sctp_chunk *chunk;
728 union sctp_addr *laddr;
731 struct sctp_sockaddr_entry *saddr;
737 if (!sctp_addip_enable)
743 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
744 __func__, sk, addrs, addrcnt);
746 list_for_each_entry(asoc, &ep->asocs, asocs) {
748 if (!asoc->peer.asconf_capable)
751 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
754 if (!sctp_state(asoc, ESTABLISHED))
757 /* Check if any address in the packed array of addresses is
758 * not present in the bind address list of the association.
759 * If so, do not send the asconf chunk to its peer, but
760 * continue with other associations.
763 for (i = 0; i < addrcnt; i++) {
765 af = sctp_get_af_specific(laddr->v4.sin_family);
771 if (!sctp_assoc_lookup_laddr(asoc, laddr))
774 addr_buf += af->sockaddr_len;
779 /* Find one address in the association's bind address list
780 * that is not in the packed array of addresses. This is to
781 * make sure that we do not delete all the addresses in the
784 bp = &asoc->base.bind_addr;
785 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
787 if ((laddr == NULL) && (addrcnt == 1)) {
788 if (asoc->asconf_addr_del_pending)
790 asoc->asconf_addr_del_pending =
791 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
792 if (asoc->asconf_addr_del_pending == NULL) {
796 asoc->asconf_addr_del_pending->sa.sa_family =
798 asoc->asconf_addr_del_pending->v4.sin_port =
800 if (addrs->sa_family == AF_INET) {
801 struct sockaddr_in *sin;
803 sin = (struct sockaddr_in *)addrs;
804 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
805 } else if (addrs->sa_family == AF_INET6) {
806 struct sockaddr_in6 *sin6;
808 sin6 = (struct sockaddr_in6 *)addrs;
809 ipv6_addr_copy(&asoc->asconf_addr_del_pending->v6.sin6_addr, &sin6->sin6_addr);
811 SCTP_DEBUG_PRINTK_IPADDR("send_asconf_del_ip: keep the last address asoc: %p ",
812 " at %p\n", asoc, asoc->asconf_addr_del_pending,
813 asoc->asconf_addr_del_pending);
814 asoc->src_out_of_asoc_ok = 1;
822 /* We do not need RCU protection throughout this loop
823 * because this is done under a socket lock from the
826 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
834 /* Reset use_as_src flag for the addresses in the bind address
835 * list that are to be deleted.
838 for (i = 0; i < addrcnt; i++) {
840 af = sctp_get_af_specific(laddr->v4.sin_family);
841 list_for_each_entry(saddr, &bp->address_list, list) {
842 if (sctp_cmp_addr_exact(&saddr->a, laddr))
843 saddr->state = SCTP_ADDR_DEL;
845 addr_buf += af->sockaddr_len;
848 /* Update the route and saddr entries for all the transports
849 * as some of the addresses in the bind address list are
850 * about to be deleted and cannot be used as source addresses.
852 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
854 dst_release(transport->dst);
855 sctp_transport_route(transport, NULL,
856 sctp_sk(asoc->base.sk));
860 /* We don't need to transmit ASCONF */
862 retval = sctp_send_asconf(asoc, chunk);
868 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
869 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
871 struct sock *sk = sctp_opt2sk(sp);
872 union sctp_addr *addr;
875 /* It is safe to write port space in caller. */
877 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
878 af = sctp_get_af_specific(addr->sa.sa_family);
881 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
884 if (addrw->state == SCTP_ADDR_NEW)
885 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
887 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
890 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
893 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
896 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
897 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
900 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
901 * Section 3.1.2 for this usage.
903 * addrs is a pointer to an array of one or more socket addresses. Each
904 * address is contained in its appropriate structure (i.e. struct
905 * sockaddr_in or struct sockaddr_in6) the family of the address type
906 * must be used to distinguish the address length (note that this
907 * representation is termed a "packed array" of addresses). The caller
908 * specifies the number of addresses in the array with addrcnt.
910 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
911 * -1, and sets errno to the appropriate error code.
913 * For SCTP, the port given in each socket address must be the same, or
914 * sctp_bindx() will fail, setting errno to EINVAL.
916 * The flags parameter is formed from the bitwise OR of zero or more of
917 * the following currently defined flags:
919 * SCTP_BINDX_ADD_ADDR
921 * SCTP_BINDX_REM_ADDR
923 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
924 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
925 * addresses from the association. The two flags are mutually exclusive;
926 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
927 * not remove all addresses from an association; sctp_bindx() will
928 * reject such an attempt with EINVAL.
930 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
931 * additional addresses with an endpoint after calling bind(). Or use
932 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
933 * socket is associated with so that no new association accepted will be
934 * associated with those addresses. If the endpoint supports dynamic
935 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
936 * endpoint to send the appropriate message to the peer to change the
937 * peers address lists.
939 * Adding and removing addresses from a connected association is
940 * optional functionality. Implementations that do not support this
941 * functionality should return EOPNOTSUPP.
943 * Basically do nothing but copying the addresses from user to kernel
944 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
945 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
948 * We don't use copy_from_user() for optimization: we first do the
949 * sanity checks (buffer size -fast- and access check-healthy
950 * pointer); if all of those succeed, then we can alloc the memory
951 * (expensive operation) needed to copy the data to kernel. Then we do
952 * the copying without checking the user space area
953 * (__copy_from_user()).
955 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
958 * sk The sk of the socket
959 * addrs The pointer to the addresses in user land
960 * addrssize Size of the addrs buffer
961 * op Operation to perform (add or remove, see the flags of
964 * Returns 0 if ok, <0 errno code on error.
966 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
967 struct sockaddr __user *addrs,
968 int addrs_size, int op)
970 struct sockaddr *kaddrs;
974 struct sockaddr *sa_addr;
978 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
979 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
981 if (unlikely(addrs_size <= 0))
984 /* Check the user passed a healthy pointer. */
985 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
988 /* Alloc space for the address array in kernel memory. */
989 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
990 if (unlikely(!kaddrs))
993 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
998 /* Walk through the addrs buffer and count the number of addresses. */
1000 while (walk_size < addrs_size) {
1001 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1007 af = sctp_get_af_specific(sa_addr->sa_family);
1009 /* If the address family is not supported or if this address
1010 * causes the address buffer to overflow return EINVAL.
1012 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1017 addr_buf += af->sockaddr_len;
1018 walk_size += af->sockaddr_len;
1023 case SCTP_BINDX_ADD_ADDR:
1024 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1027 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1030 case SCTP_BINDX_REM_ADDR:
1031 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1034 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1048 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1050 * Common routine for handling connect() and sctp_connectx().
1051 * Connect will come in with just a single address.
1053 static int __sctp_connect(struct sock* sk,
1054 struct sockaddr *kaddrs,
1056 sctp_assoc_t *assoc_id)
1058 struct sctp_sock *sp;
1059 struct sctp_endpoint *ep;
1060 struct sctp_association *asoc = NULL;
1061 struct sctp_association *asoc2;
1062 struct sctp_transport *transport;
1069 union sctp_addr *sa_addr = NULL;
1071 unsigned short port;
1072 unsigned int f_flags = 0;
1077 /* connect() cannot be done on a socket that is already in ESTABLISHED
1078 * state - UDP-style peeled off socket or a TCP-style socket that
1079 * is already connected.
1080 * It cannot be done even on a TCP-style listening socket.
1082 if (sctp_sstate(sk, ESTABLISHED) ||
1083 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1088 /* Walk through the addrs buffer and count the number of addresses. */
1090 while (walk_size < addrs_size) {
1093 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1099 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1101 /* If the address family is not supported or if this address
1102 * causes the address buffer to overflow return EINVAL.
1104 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1109 port = ntohs(sa_addr->v4.sin_port);
1111 /* Save current address so we can work with it */
1112 memcpy(&to, sa_addr, af->sockaddr_len);
1114 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1118 /* Make sure the destination port is correctly set
1121 if (asoc && asoc->peer.port && asoc->peer.port != port)
1125 /* Check if there already is a matching association on the
1126 * endpoint (other than the one created here).
1128 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1129 if (asoc2 && asoc2 != asoc) {
1130 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1137 /* If we could not find a matching association on the endpoint,
1138 * make sure that there is no peeled-off association matching
1139 * the peer address even on another socket.
1141 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1142 err = -EADDRNOTAVAIL;
1147 /* If a bind() or sctp_bindx() is not called prior to
1148 * an sctp_connectx() call, the system picks an
1149 * ephemeral port and will choose an address set
1150 * equivalent to binding with a wildcard address.
1152 if (!ep->base.bind_addr.port) {
1153 if (sctp_autobind(sk)) {
1159 * If an unprivileged user inherits a 1-many
1160 * style socket with open associations on a
1161 * privileged port, it MAY be permitted to
1162 * accept new associations, but it SHOULD NOT
1163 * be permitted to open new associations.
1165 if (ep->base.bind_addr.port < PROT_SOCK &&
1166 !capable(CAP_NET_BIND_SERVICE)) {
1172 scope = sctp_scope(&to);
1173 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1179 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1187 /* Prime the peer's transport structures. */
1188 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1196 addr_buf += af->sockaddr_len;
1197 walk_size += af->sockaddr_len;
1200 /* In case the user of sctp_connectx() wants an association
1201 * id back, assign one now.
1204 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1209 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1214 /* Initialize sk's dport and daddr for getpeername() */
1215 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1216 sp->pf->to_sk_daddr(sa_addr, sk);
1219 /* in-kernel sockets don't generally have a file allocated to them
1220 * if all they do is call sock_create_kern().
1222 if (sk->sk_socket->file)
1223 f_flags = sk->sk_socket->file->f_flags;
1225 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1228 *assoc_id = asoc->assoc_id;
1229 err = sctp_wait_for_connect(asoc, &timeo);
1230 /* Note: the asoc may be freed after the return of
1231 * sctp_wait_for_connect.
1234 /* Don't free association on exit. */
1239 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1240 " kaddrs: %p err: %d\n",
1243 /* sctp_primitive_ASSOCIATE may have added this association
1244 * To the hash table, try to unhash it, just in case, its a noop
1245 * if it wasn't hashed so we're safe
1247 sctp_unhash_established(asoc);
1248 sctp_association_free(asoc);
1253 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1256 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1257 * sctp_assoc_t *asoc);
1259 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1260 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1261 * or IPv6 addresses.
1263 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1264 * Section 3.1.2 for this usage.
1266 * addrs is a pointer to an array of one or more socket addresses. Each
1267 * address is contained in its appropriate structure (i.e. struct
1268 * sockaddr_in or struct sockaddr_in6) the family of the address type
1269 * must be used to distengish the address length (note that this
1270 * representation is termed a "packed array" of addresses). The caller
1271 * specifies the number of addresses in the array with addrcnt.
1273 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1274 * the association id of the new association. On failure, sctp_connectx()
1275 * returns -1, and sets errno to the appropriate error code. The assoc_id
1276 * is not touched by the kernel.
1278 * For SCTP, the port given in each socket address must be the same, or
1279 * sctp_connectx() will fail, setting errno to EINVAL.
1281 * An application can use sctp_connectx to initiate an association with
1282 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1283 * allows a caller to specify multiple addresses at which a peer can be
1284 * reached. The way the SCTP stack uses the list of addresses to set up
1285 * the association is implementation dependent. This function only
1286 * specifies that the stack will try to make use of all the addresses in
1287 * the list when needed.
1289 * Note that the list of addresses passed in is only used for setting up
1290 * the association. It does not necessarily equal the set of addresses
1291 * the peer uses for the resulting association. If the caller wants to
1292 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1293 * retrieve them after the association has been set up.
1295 * Basically do nothing but copying the addresses from user to kernel
1296 * land and invoking either sctp_connectx(). This is used for tunneling
1297 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1299 * We don't use copy_from_user() for optimization: we first do the
1300 * sanity checks (buffer size -fast- and access check-healthy
1301 * pointer); if all of those succeed, then we can alloc the memory
1302 * (expensive operation) needed to copy the data to kernel. Then we do
1303 * the copying without checking the user space area
1304 * (__copy_from_user()).
1306 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1309 * sk The sk of the socket
1310 * addrs The pointer to the addresses in user land
1311 * addrssize Size of the addrs buffer
1313 * Returns >=0 if ok, <0 errno code on error.
1315 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1316 struct sockaddr __user *addrs,
1318 sctp_assoc_t *assoc_id)
1321 struct sockaddr *kaddrs;
1323 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1324 __func__, sk, addrs, addrs_size);
1326 if (unlikely(addrs_size <= 0))
1329 /* Check the user passed a healthy pointer. */
1330 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1333 /* Alloc space for the address array in kernel memory. */
1334 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1335 if (unlikely(!kaddrs))
1338 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1341 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1350 * This is an older interface. It's kept for backward compatibility
1351 * to the option that doesn't provide association id.
1353 SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1354 struct sockaddr __user *addrs,
1357 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1361 * New interface for the API. The since the API is done with a socket
1362 * option, to make it simple we feed back the association id is as a return
1363 * indication to the call. Error is always negative and association id is
1366 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1367 struct sockaddr __user *addrs,
1370 sctp_assoc_t assoc_id = 0;
1373 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1382 * New (hopefully final) interface for the API.
1383 * We use the sctp_getaddrs_old structure so that use-space library
1384 * can avoid any unnecessary allocations. The only different part
1385 * is that we store the actual length of the address buffer into the
1386 * addrs_num structure member. That way we can re-use the existing
1389 #ifdef CONFIG_COMPAT
1390 struct compat_sctp_getaddrs_old {
1391 sctp_assoc_t assoc_id;
1393 compat_uptr_t addrs; /* struct sockaddr * */
1397 SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1398 char __user *optval,
1401 struct sctp_getaddrs_old param;
1402 sctp_assoc_t assoc_id = 0;
1405 #ifdef CONFIG_COMPAT
1406 if (is_compat_task()) {
1407 struct compat_sctp_getaddrs_old param32;
1409 if (len < sizeof(param32))
1411 if (copy_from_user(¶m32, optval, sizeof(param32)))
1414 param.assoc_id = param32.assoc_id;
1415 param.addr_num = param32.addr_num;
1416 param.addrs = compat_ptr(param32.addrs);
1420 if (len < sizeof(param))
1422 if (copy_from_user(¶m, optval, sizeof(param)))
1426 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1427 param.addrs, param.addr_num,
1429 if (err == 0 || err == -EINPROGRESS) {
1430 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1432 if (put_user(sizeof(assoc_id), optlen))
1439 /* API 3.1.4 close() - UDP Style Syntax
1440 * Applications use close() to perform graceful shutdown (as described in
1441 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1442 * by a UDP-style socket.
1446 * ret = close(int sd);
1448 * sd - the socket descriptor of the associations to be closed.
1450 * To gracefully shutdown a specific association represented by the
1451 * UDP-style socket, an application should use the sendmsg() call,
1452 * passing no user data, but including the appropriate flag in the
1453 * ancillary data (see Section xxxx).
1455 * If sd in the close() call is a branched-off socket representing only
1456 * one association, the shutdown is performed on that association only.
1458 * 4.1.6 close() - TCP Style Syntax
1460 * Applications use close() to gracefully close down an association.
1464 * int close(int sd);
1466 * sd - the socket descriptor of the association to be closed.
1468 * After an application calls close() on a socket descriptor, no further
1469 * socket operations will succeed on that descriptor.
1471 * API 7.1.4 SO_LINGER
1473 * An application using the TCP-style socket can use this option to
1474 * perform the SCTP ABORT primitive. The linger option structure is:
1477 * int l_onoff; // option on/off
1478 * int l_linger; // linger time
1481 * To enable the option, set l_onoff to 1. If the l_linger value is set
1482 * to 0, calling close() is the same as the ABORT primitive. If the
1483 * value is set to a negative value, the setsockopt() call will return
1484 * an error. If the value is set to a positive value linger_time, the
1485 * close() can be blocked for at most linger_time ms. If the graceful
1486 * shutdown phase does not finish during this period, close() will
1487 * return but the graceful shutdown phase continues in the system.
1489 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1491 struct sctp_endpoint *ep;
1492 struct sctp_association *asoc;
1493 struct list_head *pos, *temp;
1494 unsigned int data_was_unread;
1496 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1499 sk->sk_shutdown = SHUTDOWN_MASK;
1500 sk->sk_state = SCTP_SS_CLOSING;
1502 ep = sctp_sk(sk)->ep;
1504 /* Clean up any skbs sitting on the receive queue. */
1505 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1506 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1508 /* Walk all associations on an endpoint. */
1509 list_for_each_safe(pos, temp, &ep->asocs) {
1510 asoc = list_entry(pos, struct sctp_association, asocs);
1512 if (sctp_style(sk, TCP)) {
1513 /* A closed association can still be in the list if
1514 * it belongs to a TCP-style listening socket that is
1515 * not yet accepted. If so, free it. If not, send an
1516 * ABORT or SHUTDOWN based on the linger options.
1518 if (sctp_state(asoc, CLOSED)) {
1519 sctp_unhash_established(asoc);
1520 sctp_association_free(asoc);
1525 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1526 !skb_queue_empty(&asoc->ulpq.reasm) ||
1527 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1528 struct sctp_chunk *chunk;
1530 chunk = sctp_make_abort_user(asoc, NULL, 0);
1531 sctp_primitive_ABORT(asoc, chunk);
1533 sctp_primitive_SHUTDOWN(asoc, NULL);
1536 /* On a TCP-style socket, block for at most linger_time if set. */
1537 if (sctp_style(sk, TCP) && timeout)
1538 sctp_wait_for_close(sk, timeout);
1540 /* This will run the backlog queue. */
1541 sctp_release_sock(sk);
1543 /* Supposedly, no process has access to the socket, but
1544 * the net layers still may.
1545 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1546 * held and that should be grabbed before socket lock.
1548 spin_lock_bh(&sctp_globals.addr_wq_lock);
1549 sctp_bh_lock_sock(sk);
1551 /* Hold the sock, since sk_common_release() will put sock_put()
1552 * and we have just a little more cleanup.
1555 sk_common_release(sk);
1557 sctp_bh_unlock_sock(sk);
1558 spin_unlock_bh(&sctp_globals.addr_wq_lock);
1562 SCTP_DBG_OBJCNT_DEC(sock);
1565 /* Handle EPIPE error. */
1566 static int sctp_error(struct sock *sk, int flags, int err)
1569 err = sock_error(sk) ? : -EPIPE;
1570 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1571 send_sig(SIGPIPE, current, 0);
1575 /* API 3.1.3 sendmsg() - UDP Style Syntax
1577 * An application uses sendmsg() and recvmsg() calls to transmit data to
1578 * and receive data from its peer.
1580 * ssize_t sendmsg(int socket, const struct msghdr *message,
1583 * socket - the socket descriptor of the endpoint.
1584 * message - pointer to the msghdr structure which contains a single
1585 * user message and possibly some ancillary data.
1587 * See Section 5 for complete description of the data
1590 * flags - flags sent or received with the user message, see Section
1591 * 5 for complete description of the flags.
1593 * Note: This function could use a rewrite especially when explicit
1594 * connect support comes in.
1596 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1598 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1600 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1601 struct msghdr *msg, size_t msg_len)
1603 struct sctp_sock *sp;
1604 struct sctp_endpoint *ep;
1605 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1606 struct sctp_transport *transport, *chunk_tp;
1607 struct sctp_chunk *chunk;
1609 struct sockaddr *msg_name = NULL;
1610 struct sctp_sndrcvinfo default_sinfo;
1611 struct sctp_sndrcvinfo *sinfo;
1612 struct sctp_initmsg *sinit;
1613 sctp_assoc_t associd = 0;
1614 sctp_cmsgs_t cmsgs = { NULL };
1618 __u16 sinfo_flags = 0;
1619 bool wait_connect = false;
1620 struct sctp_datamsg *datamsg;
1621 int msg_flags = msg->msg_flags;
1623 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1630 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1632 /* We cannot send a message over a TCP-style listening socket. */
1633 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1638 /* Parse out the SCTP CMSGs. */
1639 err = sctp_msghdr_parse(msg, &cmsgs);
1642 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1646 /* Fetch the destination address for this packet. This
1647 * address only selects the association--it is not necessarily
1648 * the address we will send to.
1649 * For a peeled-off socket, msg_name is ignored.
1651 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1652 int msg_namelen = msg->msg_namelen;
1654 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1659 if (msg_namelen > sizeof(to))
1660 msg_namelen = sizeof(to);
1661 memcpy(&to, msg->msg_name, msg_namelen);
1662 msg_name = msg->msg_name;
1668 /* Did the user specify SNDRCVINFO? */
1670 sinfo_flags = sinfo->sinfo_flags;
1671 associd = sinfo->sinfo_assoc_id;
1674 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1675 msg_len, sinfo_flags);
1677 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1678 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1683 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1684 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1685 * If SCTP_ABORT is set, the message length could be non zero with
1686 * the msg_iov set to the user abort reason.
1688 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1689 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1694 /* If SCTP_ADDR_OVER is set, there must be an address
1695 * specified in msg_name.
1697 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1704 SCTP_DEBUG_PRINTK("About to look up association.\n");
1708 /* If a msg_name has been specified, assume this is to be used. */
1710 /* Look for a matching association on the endpoint. */
1711 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1713 /* If we could not find a matching association on the
1714 * endpoint, make sure that it is not a TCP-style
1715 * socket that already has an association or there is
1716 * no peeled-off association on another socket.
1718 if ((sctp_style(sk, TCP) &&
1719 sctp_sstate(sk, ESTABLISHED)) ||
1720 sctp_endpoint_is_peeled_off(ep, &to)) {
1721 err = -EADDRNOTAVAIL;
1726 asoc = sctp_id2assoc(sk, associd);
1734 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1736 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1737 * socket that has an association in CLOSED state. This can
1738 * happen when an accepted socket has an association that is
1741 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1746 if (sinfo_flags & SCTP_EOF) {
1747 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1749 sctp_primitive_SHUTDOWN(asoc, NULL);
1753 if (sinfo_flags & SCTP_ABORT) {
1755 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1761 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1762 sctp_primitive_ABORT(asoc, chunk);
1768 /* Do we need to create the association? */
1770 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1772 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1777 /* Check for invalid stream against the stream counts,
1778 * either the default or the user specified stream counts.
1781 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1782 /* Check against the defaults. */
1783 if (sinfo->sinfo_stream >=
1784 sp->initmsg.sinit_num_ostreams) {
1789 /* Check against the requested. */
1790 if (sinfo->sinfo_stream >=
1791 sinit->sinit_num_ostreams) {
1799 * API 3.1.2 bind() - UDP Style Syntax
1800 * If a bind() or sctp_bindx() is not called prior to a
1801 * sendmsg() call that initiates a new association, the
1802 * system picks an ephemeral port and will choose an address
1803 * set equivalent to binding with a wildcard address.
1805 if (!ep->base.bind_addr.port) {
1806 if (sctp_autobind(sk)) {
1812 * If an unprivileged user inherits a one-to-many
1813 * style socket with open associations on a privileged
1814 * port, it MAY be permitted to accept new associations,
1815 * but it SHOULD NOT be permitted to open new
1818 if (ep->base.bind_addr.port < PROT_SOCK &&
1819 !capable(CAP_NET_BIND_SERVICE)) {
1825 scope = sctp_scope(&to);
1826 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1832 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1838 /* If the SCTP_INIT ancillary data is specified, set all
1839 * the association init values accordingly.
1842 if (sinit->sinit_num_ostreams) {
1843 asoc->c.sinit_num_ostreams =
1844 sinit->sinit_num_ostreams;
1846 if (sinit->sinit_max_instreams) {
1847 asoc->c.sinit_max_instreams =
1848 sinit->sinit_max_instreams;
1850 if (sinit->sinit_max_attempts) {
1851 asoc->max_init_attempts
1852 = sinit->sinit_max_attempts;
1854 if (sinit->sinit_max_init_timeo) {
1855 asoc->max_init_timeo =
1856 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1860 /* Prime the peer's transport structures. */
1861 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1868 /* ASSERT: we have a valid association at this point. */
1869 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1872 /* If the user didn't specify SNDRCVINFO, make up one with
1875 memset(&default_sinfo, 0, sizeof(default_sinfo));
1876 default_sinfo.sinfo_stream = asoc->default_stream;
1877 default_sinfo.sinfo_flags = asoc->default_flags;
1878 default_sinfo.sinfo_ppid = asoc->default_ppid;
1879 default_sinfo.sinfo_context = asoc->default_context;
1880 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1881 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1882 sinfo = &default_sinfo;
1885 /* API 7.1.7, the sndbuf size per association bounds the
1886 * maximum size of data that can be sent in a single send call.
1888 if (msg_len > sk->sk_sndbuf) {
1893 if (asoc->pmtu_pending)
1894 sctp_assoc_pending_pmtu(asoc);
1896 /* If fragmentation is disabled and the message length exceeds the
1897 * association fragmentation point, return EMSGSIZE. The I-D
1898 * does not specify what this error is, but this looks like
1901 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1906 /* Check for invalid stream. */
1907 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1912 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1913 if (!sctp_wspace(asoc)) {
1914 /* sk can be changed by peel off when waiting for buf. */
1915 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len, &sk);
1920 /* If an address is passed with the sendto/sendmsg call, it is used
1921 * to override the primary destination address in the TCP model, or
1922 * when SCTP_ADDR_OVER flag is set in the UDP model.
1924 if ((sctp_style(sk, TCP) && msg_name) ||
1925 (sinfo_flags & SCTP_ADDR_OVER)) {
1926 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1934 /* Auto-connect, if we aren't connected already. */
1935 if (sctp_state(asoc, CLOSED)) {
1936 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1939 wait_connect = true;
1940 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1943 /* Break the message into multiple chunks of maximum size. */
1944 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1945 if (IS_ERR(datamsg)) {
1946 err = PTR_ERR(datamsg);
1950 /* Now send the (possibly) fragmented message. */
1951 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1952 sctp_chunk_hold(chunk);
1954 /* Do accounting for the write space. */
1955 sctp_set_owner_w(chunk);
1957 chunk->transport = chunk_tp;
1960 /* Send it to the lower layers. Note: all chunks
1961 * must either fail or succeed. The lower layer
1962 * works that way today. Keep it that way or this
1965 err = sctp_primitive_SEND(asoc, datamsg);
1966 /* Did the lower layer accept the chunk? */
1968 sctp_datamsg_free(datamsg);
1970 sctp_datamsg_put(datamsg);
1972 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1979 if (unlikely(wait_connect)) {
1980 timeo = sock_sndtimeo(sk, msg_flags & MSG_DONTWAIT);
1981 sctp_wait_for_connect(asoc, &timeo);
1984 /* If we are already past ASSOCIATE, the lower
1985 * layers are responsible for association cleanup.
1991 sctp_unhash_established(asoc);
1992 sctp_association_free(asoc);
1995 sctp_release_sock(sk);
1998 return sctp_error(sk, msg_flags, err);
2005 err = sock_error(sk);
2015 /* This is an extended version of skb_pull() that removes the data from the
2016 * start of a skb even when data is spread across the list of skb's in the
2017 * frag_list. len specifies the total amount of data that needs to be removed.
2018 * when 'len' bytes could be removed from the skb, it returns 0.
2019 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2020 * could not be removed.
2022 static int sctp_skb_pull(struct sk_buff *skb, int len)
2024 struct sk_buff *list;
2025 int skb_len = skb_headlen(skb);
2028 if (len <= skb_len) {
2029 __skb_pull(skb, len);
2033 __skb_pull(skb, skb_len);
2035 skb_walk_frags(skb, list) {
2036 rlen = sctp_skb_pull(list, len);
2037 skb->len -= (len-rlen);
2038 skb->data_len -= (len-rlen);
2049 /* API 3.1.3 recvmsg() - UDP Style Syntax
2051 * ssize_t recvmsg(int socket, struct msghdr *message,
2054 * socket - the socket descriptor of the endpoint.
2055 * message - pointer to the msghdr structure which contains a single
2056 * user message and possibly some ancillary data.
2058 * See Section 5 for complete description of the data
2061 * flags - flags sent or received with the user message, see Section
2062 * 5 for complete description of the flags.
2064 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2066 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2067 struct msghdr *msg, size_t len, int noblock,
2068 int flags, int *addr_len)
2070 struct sctp_ulpevent *event = NULL;
2071 struct sctp_sock *sp = sctp_sk(sk);
2072 struct sk_buff *skb;
2077 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
2078 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
2079 "len", len, "knoblauch", noblock,
2080 "flags", flags, "addr_len", addr_len);
2084 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2089 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2093 /* Get the total length of the skb including any skb's in the
2102 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2104 event = sctp_skb2event(skb);
2109 sock_recv_ts_and_drops(msg, sk, skb);
2110 if (sctp_ulpevent_is_notification(event)) {
2111 msg->msg_flags |= MSG_NOTIFICATION;
2112 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2114 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2117 /* Check if we allow SCTP_SNDRCVINFO. */
2118 if (sp->subscribe.sctp_data_io_event)
2119 sctp_ulpevent_read_sndrcvinfo(event, msg);
2121 /* FIXME: we should be calling IP/IPv6 layers. */
2122 if (sk->sk_protinfo.af_inet.cmsg_flags)
2123 ip_cmsg_recv(msg, skb);
2128 /* If skb's length exceeds the user's buffer, update the skb and
2129 * push it back to the receive_queue so that the next call to
2130 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2132 if (skb_len > copied) {
2133 msg->msg_flags &= ~MSG_EOR;
2134 if (flags & MSG_PEEK)
2136 sctp_skb_pull(skb, copied);
2137 skb_queue_head(&sk->sk_receive_queue, skb);
2139 /* When only partial message is copied to the user, increase
2140 * rwnd by that amount. If all the data in the skb is read,
2141 * rwnd is updated when the event is freed.
2143 if (!sctp_ulpevent_is_notification(event))
2144 sctp_assoc_rwnd_increase(event->asoc, copied);
2146 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2147 (event->msg_flags & MSG_EOR))
2148 msg->msg_flags |= MSG_EOR;
2150 msg->msg_flags &= ~MSG_EOR;
2153 if (flags & MSG_PEEK) {
2154 /* Release the skb reference acquired after peeking the skb in
2155 * sctp_skb_recv_datagram().
2159 /* Free the event which includes releasing the reference to
2160 * the owner of the skb, freeing the skb and updating the
2163 sctp_ulpevent_free(event);
2166 sctp_release_sock(sk);
2170 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2172 * This option is a on/off flag. If enabled no SCTP message
2173 * fragmentation will be performed. Instead if a message being sent
2174 * exceeds the current PMTU size, the message will NOT be sent and
2175 * instead a error will be indicated to the user.
2177 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2178 char __user *optval,
2179 unsigned int optlen)
2183 if (optlen < sizeof(int))
2186 if (get_user(val, (int __user *)optval))
2189 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2194 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2195 unsigned int optlen)
2197 struct sctp_association *asoc;
2198 struct sctp_ulpevent *event;
2200 if (optlen > sizeof(struct sctp_event_subscribe))
2202 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2206 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2207 * if there is no data to be sent or retransmit, the stack will
2208 * immediately send up this notification.
2210 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2211 &sctp_sk(sk)->subscribe)) {
2212 asoc = sctp_id2assoc(sk, 0);
2214 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2215 event = sctp_ulpevent_make_sender_dry_event(asoc,
2220 sctp_ulpq_tail_event(&asoc->ulpq, event);
2227 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2229 * This socket option is applicable to the UDP-style socket only. When
2230 * set it will cause associations that are idle for more than the
2231 * specified number of seconds to automatically close. An association
2232 * being idle is defined an association that has NOT sent or received
2233 * user data. The special value of '0' indicates that no automatic
2234 * close of any associations should be performed. The option expects an
2235 * integer defining the number of seconds of idle time before an
2236 * association is closed.
2238 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2239 unsigned int optlen)
2241 struct sctp_sock *sp = sctp_sk(sk);
2243 /* Applicable to UDP-style socket only */
2244 if (sctp_style(sk, TCP))
2246 if (optlen != sizeof(int))
2248 if (copy_from_user(&sp->autoclose, optval, optlen))
2254 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2256 * Applications can enable or disable heartbeats for any peer address of
2257 * an association, modify an address's heartbeat interval, force a
2258 * heartbeat to be sent immediately, and adjust the address's maximum
2259 * number of retransmissions sent before an address is considered
2260 * unreachable. The following structure is used to access and modify an
2261 * address's parameters:
2263 * struct sctp_paddrparams {
2264 * sctp_assoc_t spp_assoc_id;
2265 * struct sockaddr_storage spp_address;
2266 * uint32_t spp_hbinterval;
2267 * uint16_t spp_pathmaxrxt;
2268 * uint32_t spp_pathmtu;
2269 * uint32_t spp_sackdelay;
2270 * uint32_t spp_flags;
2273 * spp_assoc_id - (one-to-many style socket) This is filled in the
2274 * application, and identifies the association for
2276 * spp_address - This specifies which address is of interest.
2277 * spp_hbinterval - This contains the value of the heartbeat interval,
2278 * in milliseconds. If a value of zero
2279 * is present in this field then no changes are to
2280 * be made to this parameter.
2281 * spp_pathmaxrxt - This contains the maximum number of
2282 * retransmissions before this address shall be
2283 * considered unreachable. If a value of zero
2284 * is present in this field then no changes are to
2285 * be made to this parameter.
2286 * spp_pathmtu - When Path MTU discovery is disabled the value
2287 * specified here will be the "fixed" path mtu.
2288 * Note that if the spp_address field is empty
2289 * then all associations on this address will
2290 * have this fixed path mtu set upon them.
2292 * spp_sackdelay - When delayed sack is enabled, this value specifies
2293 * the number of milliseconds that sacks will be delayed
2294 * for. This value will apply to all addresses of an
2295 * association if the spp_address field is empty. Note
2296 * also, that if delayed sack is enabled and this
2297 * value is set to 0, no change is made to the last
2298 * recorded delayed sack timer value.
2300 * spp_flags - These flags are used to control various features
2301 * on an association. The flag field may contain
2302 * zero or more of the following options.
2304 * SPP_HB_ENABLE - Enable heartbeats on the
2305 * specified address. Note that if the address
2306 * field is empty all addresses for the association
2307 * have heartbeats enabled upon them.
2309 * SPP_HB_DISABLE - Disable heartbeats on the
2310 * speicifed address. Note that if the address
2311 * field is empty all addresses for the association
2312 * will have their heartbeats disabled. Note also
2313 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2314 * mutually exclusive, only one of these two should
2315 * be specified. Enabling both fields will have
2316 * undetermined results.
2318 * SPP_HB_DEMAND - Request a user initiated heartbeat
2319 * to be made immediately.
2321 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2322 * heartbeat delayis to be set to the value of 0
2325 * SPP_PMTUD_ENABLE - This field will enable PMTU
2326 * discovery upon the specified address. Note that
2327 * if the address feild is empty then all addresses
2328 * on the association are effected.
2330 * SPP_PMTUD_DISABLE - This field will disable PMTU
2331 * discovery upon the specified address. Note that
2332 * if the address feild is empty then all addresses
2333 * on the association are effected. Not also that
2334 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2335 * exclusive. Enabling both will have undetermined
2338 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2339 * on delayed sack. The time specified in spp_sackdelay
2340 * is used to specify the sack delay for this address. Note
2341 * that if spp_address is empty then all addresses will
2342 * enable delayed sack and take on the sack delay
2343 * value specified in spp_sackdelay.
2344 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2345 * off delayed sack. If the spp_address field is blank then
2346 * delayed sack is disabled for the entire association. Note
2347 * also that this field is mutually exclusive to
2348 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2351 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2352 struct sctp_transport *trans,
2353 struct sctp_association *asoc,
2354 struct sctp_sock *sp,
2357 int sackdelay_change)
2361 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2362 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2367 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2368 * this field is ignored. Note also that a value of zero indicates
2369 * the current setting should be left unchanged.
2371 if (params->spp_flags & SPP_HB_ENABLE) {
2373 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2374 * set. This lets us use 0 value when this flag
2377 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2378 params->spp_hbinterval = 0;
2380 if (params->spp_hbinterval ||
2381 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2384 msecs_to_jiffies(params->spp_hbinterval);
2387 msecs_to_jiffies(params->spp_hbinterval);
2389 sp->hbinterval = params->spp_hbinterval;
2396 trans->param_flags =
2397 (trans->param_flags & ~SPP_HB) | hb_change;
2400 (asoc->param_flags & ~SPP_HB) | hb_change;
2403 (sp->param_flags & ~SPP_HB) | hb_change;
2407 /* When Path MTU discovery is disabled the value specified here will
2408 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2409 * include the flag SPP_PMTUD_DISABLE for this field to have any
2412 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2414 trans->pathmtu = params->spp_pathmtu;
2415 sctp_assoc_sync_pmtu(asoc);
2417 asoc->pathmtu = params->spp_pathmtu;
2418 sctp_frag_point(asoc, params->spp_pathmtu);
2420 sp->pathmtu = params->spp_pathmtu;
2426 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2427 (params->spp_flags & SPP_PMTUD_ENABLE);
2428 trans->param_flags =
2429 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2431 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2432 sctp_assoc_sync_pmtu(asoc);
2436 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2439 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2443 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2444 * value of this field is ignored. Note also that a value of zero
2445 * indicates the current setting should be left unchanged.
2447 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2450 msecs_to_jiffies(params->spp_sackdelay);
2453 msecs_to_jiffies(params->spp_sackdelay);
2455 sp->sackdelay = params->spp_sackdelay;
2459 if (sackdelay_change) {
2461 trans->param_flags =
2462 (trans->param_flags & ~SPP_SACKDELAY) |
2466 (asoc->param_flags & ~SPP_SACKDELAY) |
2470 (sp->param_flags & ~SPP_SACKDELAY) |
2475 /* Note that a value of zero indicates the current setting should be
2478 if (params->spp_pathmaxrxt) {
2480 trans->pathmaxrxt = params->spp_pathmaxrxt;
2482 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2484 sp->pathmaxrxt = params->spp_pathmaxrxt;
2491 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2492 char __user *optval,
2493 unsigned int optlen)
2495 struct sctp_paddrparams params;
2496 struct sctp_transport *trans = NULL;
2497 struct sctp_association *asoc = NULL;
2498 struct sctp_sock *sp = sctp_sk(sk);
2500 int hb_change, pmtud_change, sackdelay_change;
2502 if (optlen != sizeof(struct sctp_paddrparams))
2505 if (copy_from_user(¶ms, optval, optlen))
2508 /* Validate flags and value parameters. */
2509 hb_change = params.spp_flags & SPP_HB;
2510 pmtud_change = params.spp_flags & SPP_PMTUD;
2511 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2513 if (hb_change == SPP_HB ||
2514 pmtud_change == SPP_PMTUD ||
2515 sackdelay_change == SPP_SACKDELAY ||
2516 params.spp_sackdelay > 500 ||
2517 (params.spp_pathmtu &&
2518 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2521 /* If an address other than INADDR_ANY is specified, and
2522 * no transport is found, then the request is invalid.
2524 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
2525 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2526 params.spp_assoc_id);
2531 /* Get association, if assoc_id != 0 and the socket is a one
2532 * to many style socket, and an association was not found, then
2533 * the id was invalid.
2535 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2536 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2539 /* Heartbeat demand can only be sent on a transport or
2540 * association, but not a socket.
2542 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2545 /* Process parameters. */
2546 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2547 hb_change, pmtud_change,
2553 /* If changes are for association, also apply parameters to each
2556 if (!trans && asoc) {
2557 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2559 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2560 hb_change, pmtud_change,
2569 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2571 * This option will effect the way delayed acks are performed. This
2572 * option allows you to get or set the delayed ack time, in
2573 * milliseconds. It also allows changing the delayed ack frequency.
2574 * Changing the frequency to 1 disables the delayed sack algorithm. If
2575 * the assoc_id is 0, then this sets or gets the endpoints default
2576 * values. If the assoc_id field is non-zero, then the set or get
2577 * effects the specified association for the one to many model (the
2578 * assoc_id field is ignored by the one to one model). Note that if
2579 * sack_delay or sack_freq are 0 when setting this option, then the
2580 * current values will remain unchanged.
2582 * struct sctp_sack_info {
2583 * sctp_assoc_t sack_assoc_id;
2584 * uint32_t sack_delay;
2585 * uint32_t sack_freq;
2588 * sack_assoc_id - This parameter, indicates which association the user
2589 * is performing an action upon. Note that if this field's value is
2590 * zero then the endpoints default value is changed (effecting future
2591 * associations only).
2593 * sack_delay - This parameter contains the number of milliseconds that
2594 * the user is requesting the delayed ACK timer be set to. Note that
2595 * this value is defined in the standard to be between 200 and 500
2598 * sack_freq - This parameter contains the number of packets that must
2599 * be received before a sack is sent without waiting for the delay
2600 * timer to expire. The default value for this is 2, setting this
2601 * value to 1 will disable the delayed sack algorithm.
2604 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2605 char __user *optval, unsigned int optlen)
2607 struct sctp_sack_info params;
2608 struct sctp_transport *trans = NULL;
2609 struct sctp_association *asoc = NULL;
2610 struct sctp_sock *sp = sctp_sk(sk);
2612 if (optlen == sizeof(struct sctp_sack_info)) {
2613 if (copy_from_user(¶ms, optval, optlen))
2616 if (params.sack_delay == 0 && params.sack_freq == 0)
2618 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2619 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2620 pr_warn("Use struct sctp_sack_info instead\n");
2621 if (copy_from_user(¶ms, optval, optlen))
2624 if (params.sack_delay == 0)
2625 params.sack_freq = 1;
2627 params.sack_freq = 0;
2631 /* Validate value parameter. */
2632 if (params.sack_delay > 500)
2635 /* Get association, if sack_assoc_id != 0 and the socket is a one
2636 * to many style socket, and an association was not found, then
2637 * the id was invalid.
2639 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2640 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2643 if (params.sack_delay) {
2646 msecs_to_jiffies(params.sack_delay);
2648 (asoc->param_flags & ~SPP_SACKDELAY) |
2649 SPP_SACKDELAY_ENABLE;
2651 sp->sackdelay = params.sack_delay;
2653 (sp->param_flags & ~SPP_SACKDELAY) |
2654 SPP_SACKDELAY_ENABLE;
2658 if (params.sack_freq == 1) {
2661 (asoc->param_flags & ~SPP_SACKDELAY) |
2662 SPP_SACKDELAY_DISABLE;
2665 (sp->param_flags & ~SPP_SACKDELAY) |
2666 SPP_SACKDELAY_DISABLE;
2668 } else if (params.sack_freq > 1) {
2670 asoc->sackfreq = params.sack_freq;
2672 (asoc->param_flags & ~SPP_SACKDELAY) |
2673 SPP_SACKDELAY_ENABLE;
2675 sp->sackfreq = params.sack_freq;
2677 (sp->param_flags & ~SPP_SACKDELAY) |
2678 SPP_SACKDELAY_ENABLE;
2682 /* If change is for association, also apply to each transport. */
2684 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2686 if (params.sack_delay) {
2688 msecs_to_jiffies(params.sack_delay);
2689 trans->param_flags =
2690 (trans->param_flags & ~SPP_SACKDELAY) |
2691 SPP_SACKDELAY_ENABLE;
2693 if (params.sack_freq == 1) {
2694 trans->param_flags =
2695 (trans->param_flags & ~SPP_SACKDELAY) |
2696 SPP_SACKDELAY_DISABLE;
2697 } else if (params.sack_freq > 1) {
2698 trans->sackfreq = params.sack_freq;
2699 trans->param_flags =
2700 (trans->param_flags & ~SPP_SACKDELAY) |
2701 SPP_SACKDELAY_ENABLE;
2709 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2711 * Applications can specify protocol parameters for the default association
2712 * initialization. The option name argument to setsockopt() and getsockopt()
2715 * Setting initialization parameters is effective only on an unconnected
2716 * socket (for UDP-style sockets only future associations are effected
2717 * by the change). With TCP-style sockets, this option is inherited by
2718 * sockets derived from a listener socket.
2720 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2722 struct sctp_initmsg sinit;
2723 struct sctp_sock *sp = sctp_sk(sk);
2725 if (optlen != sizeof(struct sctp_initmsg))
2727 if (copy_from_user(&sinit, optval, optlen))
2730 if (sinit.sinit_num_ostreams)
2731 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2732 if (sinit.sinit_max_instreams)
2733 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2734 if (sinit.sinit_max_attempts)
2735 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2736 if (sinit.sinit_max_init_timeo)
2737 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2743 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2745 * Applications that wish to use the sendto() system call may wish to
2746 * specify a default set of parameters that would normally be supplied
2747 * through the inclusion of ancillary data. This socket option allows
2748 * such an application to set the default sctp_sndrcvinfo structure.
2749 * The application that wishes to use this socket option simply passes
2750 * in to this call the sctp_sndrcvinfo structure defined in Section
2751 * 5.2.2) The input parameters accepted by this call include
2752 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2753 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2754 * to this call if the caller is using the UDP model.
2756 static int sctp_setsockopt_default_send_param(struct sock *sk,
2757 char __user *optval,
2758 unsigned int optlen)
2760 struct sctp_sndrcvinfo info;
2761 struct sctp_association *asoc;
2762 struct sctp_sock *sp = sctp_sk(sk);
2764 if (optlen != sizeof(struct sctp_sndrcvinfo))
2766 if (copy_from_user(&info, optval, optlen))
2769 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2770 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2774 asoc->default_stream = info.sinfo_stream;
2775 asoc->default_flags = info.sinfo_flags;
2776 asoc->default_ppid = info.sinfo_ppid;
2777 asoc->default_context = info.sinfo_context;
2778 asoc->default_timetolive = info.sinfo_timetolive;
2780 sp->default_stream = info.sinfo_stream;
2781 sp->default_flags = info.sinfo_flags;
2782 sp->default_ppid = info.sinfo_ppid;
2783 sp->default_context = info.sinfo_context;
2784 sp->default_timetolive = info.sinfo_timetolive;
2790 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2792 * Requests that the local SCTP stack use the enclosed peer address as
2793 * the association primary. The enclosed address must be one of the
2794 * association peer's addresses.
2796 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2797 unsigned int optlen)
2799 struct sctp_prim prim;
2800 struct sctp_transport *trans;
2802 if (optlen != sizeof(struct sctp_prim))
2805 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2808 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2812 sctp_assoc_set_primary(trans->asoc, trans);
2818 * 7.1.5 SCTP_NODELAY
2820 * Turn on/off any Nagle-like algorithm. This means that packets are
2821 * generally sent as soon as possible and no unnecessary delays are
2822 * introduced, at the cost of more packets in the network. Expects an
2823 * integer boolean flag.
2825 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2826 unsigned int optlen)
2830 if (optlen < sizeof(int))
2832 if (get_user(val, (int __user *)optval))
2835 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2841 * 7.1.1 SCTP_RTOINFO
2843 * The protocol parameters used to initialize and bound retransmission
2844 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2845 * and modify these parameters.
2846 * All parameters are time values, in milliseconds. A value of 0, when
2847 * modifying the parameters, indicates that the current value should not
2851 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2853 struct sctp_rtoinfo rtoinfo;
2854 struct sctp_association *asoc;
2856 if (optlen != sizeof (struct sctp_rtoinfo))
2859 if (copy_from_user(&rtoinfo, optval, optlen))
2862 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2864 /* Set the values to the specific association */
2865 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2869 if (rtoinfo.srto_initial != 0)
2871 msecs_to_jiffies(rtoinfo.srto_initial);
2872 if (rtoinfo.srto_max != 0)
2873 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2874 if (rtoinfo.srto_min != 0)
2875 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2877 /* If there is no association or the association-id = 0
2878 * set the values to the endpoint.
2880 struct sctp_sock *sp = sctp_sk(sk);
2882 if (rtoinfo.srto_initial != 0)
2883 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2884 if (rtoinfo.srto_max != 0)
2885 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2886 if (rtoinfo.srto_min != 0)
2887 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2895 * 7.1.2 SCTP_ASSOCINFO
2897 * This option is used to tune the maximum retransmission attempts
2898 * of the association.
2899 * Returns an error if the new association retransmission value is
2900 * greater than the sum of the retransmission value of the peer.
2901 * See [SCTP] for more information.
2904 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2907 struct sctp_assocparams assocparams;
2908 struct sctp_association *asoc;
2910 if (optlen != sizeof(struct sctp_assocparams))
2912 if (copy_from_user(&assocparams, optval, optlen))
2915 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2917 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2920 /* Set the values to the specific association */
2922 if (assocparams.sasoc_asocmaxrxt != 0) {
2925 struct sctp_transport *peer_addr;
2927 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2929 path_sum += peer_addr->pathmaxrxt;
2933 /* Only validate asocmaxrxt if we have more than
2934 * one path/transport. We do this because path
2935 * retransmissions are only counted when we have more
2939 assocparams.sasoc_asocmaxrxt > path_sum)
2942 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2945 if (assocparams.sasoc_cookie_life != 0) {
2946 asoc->cookie_life.tv_sec =
2947 assocparams.sasoc_cookie_life / 1000;
2948 asoc->cookie_life.tv_usec =
2949 (assocparams.sasoc_cookie_life % 1000)
2953 /* Set the values to the endpoint */
2954 struct sctp_sock *sp = sctp_sk(sk);
2956 if (assocparams.sasoc_asocmaxrxt != 0)
2957 sp->assocparams.sasoc_asocmaxrxt =
2958 assocparams.sasoc_asocmaxrxt;
2959 if (assocparams.sasoc_cookie_life != 0)
2960 sp->assocparams.sasoc_cookie_life =
2961 assocparams.sasoc_cookie_life;
2967 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2969 * This socket option is a boolean flag which turns on or off mapped V4
2970 * addresses. If this option is turned on and the socket is type
2971 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2972 * If this option is turned off, then no mapping will be done of V4
2973 * addresses and a user will receive both PF_INET6 and PF_INET type
2974 * addresses on the socket.
2976 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2979 struct sctp_sock *sp = sctp_sk(sk);
2981 if (optlen < sizeof(int))
2983 if (get_user(val, (int __user *)optval))
2994 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2995 * This option will get or set the maximum size to put in any outgoing
2996 * SCTP DATA chunk. If a message is larger than this size it will be
2997 * fragmented by SCTP into the specified size. Note that the underlying
2998 * SCTP implementation may fragment into smaller sized chunks when the
2999 * PMTU of the underlying association is smaller than the value set by
3000 * the user. The default value for this option is '0' which indicates
3001 * the user is NOT limiting fragmentation and only the PMTU will effect
3002 * SCTP's choice of DATA chunk size. Note also that values set larger
3003 * than the maximum size of an IP datagram will effectively let SCTP
3004 * control fragmentation (i.e. the same as setting this option to 0).
3006 * The following structure is used to access and modify this parameter:
3008 * struct sctp_assoc_value {
3009 * sctp_assoc_t assoc_id;
3010 * uint32_t assoc_value;
3013 * assoc_id: This parameter is ignored for one-to-one style sockets.
3014 * For one-to-many style sockets this parameter indicates which
3015 * association the user is performing an action upon. Note that if
3016 * this field's value is zero then the endpoints default value is
3017 * changed (effecting future associations only).
3018 * assoc_value: This parameter specifies the maximum size in bytes.
3020 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3022 struct sctp_assoc_value params;
3023 struct sctp_association *asoc;
3024 struct sctp_sock *sp = sctp_sk(sk);
3027 if (optlen == sizeof(int)) {
3028 pr_warn("Use of int in maxseg socket option deprecated\n");
3029 pr_warn("Use struct sctp_assoc_value instead\n");
3030 if (copy_from_user(&val, optval, optlen))
3032 params.assoc_id = 0;
3033 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3034 if (copy_from_user(¶ms, optval, optlen))
3036 val = params.assoc_value;
3040 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3043 asoc = sctp_id2assoc(sk, params.assoc_id);
3044 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3049 val = asoc->pathmtu;
3050 val -= sp->pf->af->net_header_len;
3051 val -= sizeof(struct sctphdr) +
3052 sizeof(struct sctp_data_chunk);
3054 asoc->user_frag = val;
3055 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3057 sp->user_frag = val;
3065 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3067 * Requests that the peer mark the enclosed address as the association
3068 * primary. The enclosed address must be one of the association's
3069 * locally bound addresses. The following structure is used to make a
3070 * set primary request:
3072 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3073 unsigned int optlen)
3075 struct sctp_sock *sp;
3076 struct sctp_association *asoc = NULL;
3077 struct sctp_setpeerprim prim;
3078 struct sctp_chunk *chunk;
3084 if (!sctp_addip_enable)
3087 if (optlen != sizeof(struct sctp_setpeerprim))
3090 if (copy_from_user(&prim, optval, optlen))
3093 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3097 if (!asoc->peer.asconf_capable)
3100 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3103 if (!sctp_state(asoc, ESTABLISHED))
3106 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3110 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3111 return -EADDRNOTAVAIL;
3113 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3114 return -EADDRNOTAVAIL;
3116 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3117 chunk = sctp_make_asconf_set_prim(asoc,
3118 (union sctp_addr *)&prim.sspp_addr);
3122 err = sctp_send_asconf(asoc, chunk);
3124 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
3129 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3130 unsigned int optlen)
3132 struct sctp_setadaptation adaptation;
3134 if (optlen != sizeof(struct sctp_setadaptation))
3136 if (copy_from_user(&adaptation, optval, optlen))
3139 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3145 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3147 * The context field in the sctp_sndrcvinfo structure is normally only
3148 * used when a failed message is retrieved holding the value that was
3149 * sent down on the actual send call. This option allows the setting of
3150 * a default context on an association basis that will be received on
3151 * reading messages from the peer. This is especially helpful in the
3152 * one-2-many model for an application to keep some reference to an
3153 * internal state machine that is processing messages on the
3154 * association. Note that the setting of this value only effects
3155 * received messages from the peer and does not effect the value that is
3156 * saved with outbound messages.
3158 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3159 unsigned int optlen)
3161 struct sctp_assoc_value params;
3162 struct sctp_sock *sp;
3163 struct sctp_association *asoc;
3165 if (optlen != sizeof(struct sctp_assoc_value))
3167 if (copy_from_user(¶ms, optval, optlen))
3172 if (params.assoc_id != 0) {
3173 asoc = sctp_id2assoc(sk, params.assoc_id);
3176 asoc->default_rcv_context = params.assoc_value;
3178 sp->default_rcv_context = params.assoc_value;
3185 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3187 * This options will at a minimum specify if the implementation is doing
3188 * fragmented interleave. Fragmented interleave, for a one to many
3189 * socket, is when subsequent calls to receive a message may return
3190 * parts of messages from different associations. Some implementations
3191 * may allow you to turn this value on or off. If so, when turned off,
3192 * no fragment interleave will occur (which will cause a head of line
3193 * blocking amongst multiple associations sharing the same one to many
3194 * socket). When this option is turned on, then each receive call may
3195 * come from a different association (thus the user must receive data
3196 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3197 * association each receive belongs to.
3199 * This option takes a boolean value. A non-zero value indicates that
3200 * fragmented interleave is on. A value of zero indicates that
3201 * fragmented interleave is off.
3203 * Note that it is important that an implementation that allows this
3204 * option to be turned on, have it off by default. Otherwise an unaware
3205 * application using the one to many model may become confused and act
3208 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3209 char __user *optval,
3210 unsigned int optlen)
3214 if (optlen != sizeof(int))
3216 if (get_user(val, (int __user *)optval))
3219 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3225 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3226 * (SCTP_PARTIAL_DELIVERY_POINT)
3228 * This option will set or get the SCTP partial delivery point. This
3229 * point is the size of a message where the partial delivery API will be
3230 * invoked to help free up rwnd space for the peer. Setting this to a
3231 * lower value will cause partial deliveries to happen more often. The
3232 * calls argument is an integer that sets or gets the partial delivery
3233 * point. Note also that the call will fail if the user attempts to set
3234 * this value larger than the socket receive buffer size.
3236 * Note that any single message having a length smaller than or equal to
3237 * the SCTP partial delivery point will be delivered in one single read
3238 * call as long as the user provided buffer is large enough to hold the
3241 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3242 char __user *optval,
3243 unsigned int optlen)
3247 if (optlen != sizeof(u32))
3249 if (get_user(val, (int __user *)optval))
3252 /* Note: We double the receive buffer from what the user sets
3253 * it to be, also initial rwnd is based on rcvbuf/2.
3255 if (val > (sk->sk_rcvbuf >> 1))
3258 sctp_sk(sk)->pd_point = val;
3260 return 0; /* is this the right error code? */
3264 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3266 * This option will allow a user to change the maximum burst of packets
3267 * that can be emitted by this association. Note that the default value
3268 * is 4, and some implementations may restrict this setting so that it
3269 * can only be lowered.
3271 * NOTE: This text doesn't seem right. Do this on a socket basis with
3272 * future associations inheriting the socket value.
3274 static int sctp_setsockopt_maxburst(struct sock *sk,
3275 char __user *optval,
3276 unsigned int optlen)
3278 struct sctp_assoc_value params;
3279 struct sctp_sock *sp;
3280 struct sctp_association *asoc;
3284 if (optlen == sizeof(int)) {
3285 pr_warn("Use of int in max_burst socket option deprecated\n");
3286 pr_warn("Use struct sctp_assoc_value instead\n");
3287 if (copy_from_user(&val, optval, optlen))
3289 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3290 if (copy_from_user(¶ms, optval, optlen))
3292 val = params.assoc_value;
3293 assoc_id = params.assoc_id;
3299 if (assoc_id != 0) {
3300 asoc = sctp_id2assoc(sk, assoc_id);
3303 asoc->max_burst = val;
3305 sp->max_burst = val;
3311 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3313 * This set option adds a chunk type that the user is requesting to be
3314 * received only in an authenticated way. Changes to the list of chunks
3315 * will only effect future associations on the socket.
3317 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3318 char __user *optval,
3319 unsigned int optlen)
3321 struct sctp_authchunk val;
3323 if (!sctp_auth_enable)
3326 if (optlen != sizeof(struct sctp_authchunk))
3328 if (copy_from_user(&val, optval, optlen))
3331 switch (val.sauth_chunk) {
3333 case SCTP_CID_INIT_ACK:
3334 case SCTP_CID_SHUTDOWN_COMPLETE:
3339 /* add this chunk id to the endpoint */
3340 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3344 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3346 * This option gets or sets the list of HMAC algorithms that the local
3347 * endpoint requires the peer to use.
3349 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3350 char __user *optval,
3351 unsigned int optlen)
3353 struct sctp_hmacalgo *hmacs;
3357 if (!sctp_auth_enable)
3360 if (optlen < sizeof(struct sctp_hmacalgo))
3363 hmacs= memdup_user(optval, optlen);
3365 return PTR_ERR(hmacs);
3367 idents = hmacs->shmac_num_idents;
3368 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3369 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3374 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3381 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3383 * This option will set a shared secret key which is used to build an
3384 * association shared key.
3386 static int sctp_setsockopt_auth_key(struct sock *sk,
3387 char __user *optval,
3388 unsigned int optlen)
3390 struct sctp_authkey *authkey;
3391 struct sctp_association *asoc;
3394 if (!sctp_auth_enable)
3397 if (optlen <= sizeof(struct sctp_authkey))
3400 authkey= memdup_user(optval, optlen);
3401 if (IS_ERR(authkey))
3402 return PTR_ERR(authkey);
3404 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3409 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3410 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3415 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3422 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3424 * This option will get or set the active shared key to be used to build
3425 * the association shared key.
3427 static int sctp_setsockopt_active_key(struct sock *sk,
3428 char __user *optval,
3429 unsigned int optlen)
3431 struct sctp_authkeyid val;
3432 struct sctp_association *asoc;
3434 if (!sctp_auth_enable)
3437 if (optlen != sizeof(struct sctp_authkeyid))
3439 if (copy_from_user(&val, optval, optlen))
3442 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3443 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3446 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3447 val.scact_keynumber);
3451 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3453 * This set option will delete a shared secret key from use.
3455 static int sctp_setsockopt_del_key(struct sock *sk,
3456 char __user *optval,
3457 unsigned int optlen)
3459 struct sctp_authkeyid val;
3460 struct sctp_association *asoc;
3462 if (!sctp_auth_enable)
3465 if (optlen != sizeof(struct sctp_authkeyid))
3467 if (copy_from_user(&val, optval, optlen))
3470 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3471 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3474 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3475 val.scact_keynumber);
3480 * 8.1.23 SCTP_AUTO_ASCONF
3482 * This option will enable or disable the use of the automatic generation of
3483 * ASCONF chunks to add and delete addresses to an existing association. Note
3484 * that this option has two caveats namely: a) it only affects sockets that
3485 * are bound to all addresses available to the SCTP stack, and b) the system
3486 * administrator may have an overriding control that turns the ASCONF feature
3487 * off no matter what setting the socket option may have.
3488 * This option expects an integer boolean flag, where a non-zero value turns on
3489 * the option, and a zero value turns off the option.
3490 * Note. In this implementation, socket operation overrides default parameter
3491 * being set by sysctl as well as FreeBSD implementation
3493 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3494 unsigned int optlen)
3497 struct sctp_sock *sp = sctp_sk(sk);
3499 if (optlen < sizeof(int))
3501 if (get_user(val, (int __user *)optval))
3503 if (!sctp_is_ep_boundall(sk) && val)
3505 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3508 spin_lock_bh(&sctp_globals.addr_wq_lock);
3509 if (val == 0 && sp->do_auto_asconf) {
3510 list_del(&sp->auto_asconf_list);
3511 sp->do_auto_asconf = 0;
3512 } else if (val && !sp->do_auto_asconf) {
3513 list_add_tail(&sp->auto_asconf_list,
3514 &sctp_auto_asconf_splist);
3515 sp->do_auto_asconf = 1;
3517 spin_unlock_bh(&sctp_globals.addr_wq_lock);
3522 /* API 6.2 setsockopt(), getsockopt()
3524 * Applications use setsockopt() and getsockopt() to set or retrieve
3525 * socket options. Socket options are used to change the default
3526 * behavior of sockets calls. They are described in Section 7.
3530 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3531 * int __user *optlen);
3532 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3535 * sd - the socket descript.
3536 * level - set to IPPROTO_SCTP for all SCTP options.
3537 * optname - the option name.
3538 * optval - the buffer to store the value of the option.
3539 * optlen - the size of the buffer.
3541 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3542 char __user *optval, unsigned int optlen)
3546 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3549 /* I can hardly begin to describe how wrong this is. This is
3550 * so broken as to be worse than useless. The API draft
3551 * REALLY is NOT helpful here... I am not convinced that the
3552 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3553 * are at all well-founded.
3555 if (level != SOL_SCTP) {
3556 struct sctp_af *af = sctp_sk(sk)->pf->af;
3557 retval = af->setsockopt(sk, level, optname, optval, optlen);
3564 case SCTP_SOCKOPT_BINDX_ADD:
3565 /* 'optlen' is the size of the addresses buffer. */
3566 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3567 optlen, SCTP_BINDX_ADD_ADDR);
3570 case SCTP_SOCKOPT_BINDX_REM:
3571 /* 'optlen' is the size of the addresses buffer. */
3572 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3573 optlen, SCTP_BINDX_REM_ADDR);
3576 case SCTP_SOCKOPT_CONNECTX_OLD:
3577 /* 'optlen' is the size of the addresses buffer. */
3578 retval = sctp_setsockopt_connectx_old(sk,
3579 (struct sockaddr __user *)optval,
3583 case SCTP_SOCKOPT_CONNECTX:
3584 /* 'optlen' is the size of the addresses buffer. */
3585 retval = sctp_setsockopt_connectx(sk,
3586 (struct sockaddr __user *)optval,
3590 case SCTP_DISABLE_FRAGMENTS:
3591 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3595 retval = sctp_setsockopt_events(sk, optval, optlen);
3598 case SCTP_AUTOCLOSE:
3599 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3602 case SCTP_PEER_ADDR_PARAMS:
3603 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3606 case SCTP_DELAYED_SACK:
3607 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3609 case SCTP_PARTIAL_DELIVERY_POINT:
3610 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3614 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3616 case SCTP_DEFAULT_SEND_PARAM:
3617 retval = sctp_setsockopt_default_send_param(sk, optval,
3620 case SCTP_PRIMARY_ADDR:
3621 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3623 case SCTP_SET_PEER_PRIMARY_ADDR:
3624 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3627 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3630 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3632 case SCTP_ASSOCINFO:
3633 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3635 case SCTP_I_WANT_MAPPED_V4_ADDR:
3636 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3639 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3641 case SCTP_ADAPTATION_LAYER:
3642 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3645 retval = sctp_setsockopt_context(sk, optval, optlen);
3647 case SCTP_FRAGMENT_INTERLEAVE:
3648 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3650 case SCTP_MAX_BURST:
3651 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3653 case SCTP_AUTH_CHUNK:
3654 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3656 case SCTP_HMAC_IDENT:
3657 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3660 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3662 case SCTP_AUTH_ACTIVE_KEY:
3663 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3665 case SCTP_AUTH_DELETE_KEY:
3666 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3668 case SCTP_AUTO_ASCONF:
3669 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3672 retval = -ENOPROTOOPT;
3676 sctp_release_sock(sk);
3682 /* API 3.1.6 connect() - UDP Style Syntax
3684 * An application may use the connect() call in the UDP model to initiate an
3685 * association without sending data.
3689 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3691 * sd: the socket descriptor to have a new association added to.
3693 * nam: the address structure (either struct sockaddr_in or struct
3694 * sockaddr_in6 defined in RFC2553 [7]).
3696 * len: the size of the address.
3698 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3706 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3707 __func__, sk, addr, addr_len);
3709 /* Validate addr_len before calling common connect/connectx routine. */
3710 af = sctp_get_af_specific(addr->sa_family);
3711 if (!af || addr_len < af->sockaddr_len) {
3714 /* Pass correct addr len to common routine (so it knows there
3715 * is only one address being passed.
3717 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3720 sctp_release_sock(sk);
3724 /* FIXME: Write comments. */
3725 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3727 return -EOPNOTSUPP; /* STUB */
3730 /* 4.1.4 accept() - TCP Style Syntax
3732 * Applications use accept() call to remove an established SCTP
3733 * association from the accept queue of the endpoint. A new socket
3734 * descriptor will be returned from accept() to represent the newly
3735 * formed association.
3737 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3739 struct sctp_sock *sp;
3740 struct sctp_endpoint *ep;
3741 struct sock *newsk = NULL;
3742 struct sctp_association *asoc;
3751 if (!sctp_style(sk, TCP)) {
3752 error = -EOPNOTSUPP;
3756 if (!sctp_sstate(sk, LISTENING)) {
3761 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3763 error = sctp_wait_for_accept(sk, timeo);
3767 /* We treat the list of associations on the endpoint as the accept
3768 * queue and pick the first association on the list.
3770 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3772 newsk = sp->pf->create_accept_sk(sk, asoc);
3778 /* Populate the fields of the newsk from the oldsk and migrate the
3779 * asoc to the newsk.
3781 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3784 sctp_release_sock(sk);
3789 /* The SCTP ioctl handler. */
3790 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3797 * SEQPACKET-style sockets in LISTENING state are valid, for
3798 * SCTP, so only discard TCP-style sockets in LISTENING state.
3800 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3805 struct sk_buff *skb;
3806 unsigned int amount = 0;
3808 skb = skb_peek(&sk->sk_receive_queue);
3811 * We will only return the amount of this packet since
3812 * that is all that will be read.
3816 rc = put_user(amount, (int __user *)arg);
3824 sctp_release_sock(sk);
3828 /* This is the function which gets called during socket creation to
3829 * initialized the SCTP-specific portion of the sock.
3830 * The sock structure should already be zero-filled memory.
3832 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3834 struct sctp_endpoint *ep;
3835 struct sctp_sock *sp;
3837 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3841 /* Initialize the SCTP per socket area. */
3842 switch (sk->sk_type) {
3843 case SOCK_SEQPACKET:
3844 sp->type = SCTP_SOCKET_UDP;
3847 sp->type = SCTP_SOCKET_TCP;
3850 return -ESOCKTNOSUPPORT;
3853 /* Initialize default send parameters. These parameters can be
3854 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3856 sp->default_stream = 0;
3857 sp->default_ppid = 0;
3858 sp->default_flags = 0;
3859 sp->default_context = 0;
3860 sp->default_timetolive = 0;
3862 sp->default_rcv_context = 0;
3863 sp->max_burst = sctp_max_burst;
3865 /* Initialize default setup parameters. These parameters
3866 * can be modified with the SCTP_INITMSG socket option or
3867 * overridden by the SCTP_INIT CMSG.
3869 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3870 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3871 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3872 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3874 /* Initialize default RTO related parameters. These parameters can
3875 * be modified for with the SCTP_RTOINFO socket option.
3877 sp->rtoinfo.srto_initial = sctp_rto_initial;
3878 sp->rtoinfo.srto_max = sctp_rto_max;
3879 sp->rtoinfo.srto_min = sctp_rto_min;
3881 /* Initialize default association related parameters. These parameters
3882 * can be modified with the SCTP_ASSOCINFO socket option.
3884 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3885 sp->assocparams.sasoc_number_peer_destinations = 0;
3886 sp->assocparams.sasoc_peer_rwnd = 0;
3887 sp->assocparams.sasoc_local_rwnd = 0;
3888 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3890 /* Initialize default event subscriptions. By default, all the
3893 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3895 /* Default Peer Address Parameters. These defaults can
3896 * be modified via SCTP_PEER_ADDR_PARAMS
3898 sp->hbinterval = sctp_hb_interval;
3899 sp->pathmaxrxt = sctp_max_retrans_path;
3900 sp->pathmtu = 0; // allow default discovery
3901 sp->sackdelay = sctp_sack_timeout;
3903 sp->param_flags = SPP_HB_ENABLE |
3905 SPP_SACKDELAY_ENABLE;
3907 /* If enabled no SCTP message fragmentation will be performed.
3908 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3910 sp->disable_fragments = 0;
3912 /* Enable Nagle algorithm by default. */
3915 /* Enable by default. */
3918 /* Auto-close idle associations after the configured
3919 * number of seconds. A value of 0 disables this
3920 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3921 * for UDP-style sockets only.
3925 /* User specified fragmentation limit. */
3928 sp->adaptation_ind = 0;
3930 sp->pf = sctp_get_pf_specific(sk->sk_family);
3932 /* Control variables for partial data delivery. */
3933 atomic_set(&sp->pd_mode, 0);
3934 skb_queue_head_init(&sp->pd_lobby);
3935 sp->frag_interleave = 0;
3937 /* Create a per socket endpoint structure. Even if we
3938 * change the data structure relationships, this may still
3939 * be useful for storing pre-connect address information.
3941 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3948 SCTP_DBG_OBJCNT_INC(sock);
3951 percpu_counter_inc(&sctp_sockets_allocated);
3952 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3954 /* Nothing can fail after this block, otherwise
3955 * sctp_destroy_sock() will be called without addr_wq_lock held
3957 if (sctp_default_auto_asconf) {
3958 spin_lock(&sctp_globals.addr_wq_lock);
3959 list_add_tail(&sp->auto_asconf_list,
3960 &sctp_auto_asconf_splist);
3961 sp->do_auto_asconf = 1;
3962 spin_unlock(&sctp_globals.addr_wq_lock);
3964 sp->do_auto_asconf = 0;
3972 /* Cleanup any SCTP per socket resources. Must be called with
3973 * sctp_globals.addr_wq_lock held if sp->do_auto_asconf is true
3975 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3977 struct sctp_sock *sp;
3979 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3981 /* Release our hold on the endpoint. */
3983 /* This could happen during socket init, thus we bail out
3984 * early, since the rest of the below is not setup either.
3989 if (sp->do_auto_asconf) {
3990 sp->do_auto_asconf = 0;
3991 list_del(&sp->auto_asconf_list);
3993 sctp_endpoint_free(sp->ep);
3995 percpu_counter_dec(&sctp_sockets_allocated);
3996 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4000 /* API 4.1.7 shutdown() - TCP Style Syntax
4001 * int shutdown(int socket, int how);
4003 * sd - the socket descriptor of the association to be closed.
4004 * how - Specifies the type of shutdown. The values are
4007 * Disables further receive operations. No SCTP
4008 * protocol action is taken.
4010 * Disables further send operations, and initiates
4011 * the SCTP shutdown sequence.
4013 * Disables further send and receive operations
4014 * and initiates the SCTP shutdown sequence.
4016 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
4018 struct sctp_endpoint *ep;
4019 struct sctp_association *asoc;
4021 if (!sctp_style(sk, TCP))
4024 if (how & SEND_SHUTDOWN) {
4025 ep = sctp_sk(sk)->ep;
4026 if (!list_empty(&ep->asocs)) {
4027 asoc = list_entry(ep->asocs.next,
4028 struct sctp_association, asocs);
4029 sctp_primitive_SHUTDOWN(asoc, NULL);
4034 /* 7.2.1 Association Status (SCTP_STATUS)
4036 * Applications can retrieve current status information about an
4037 * association, including association state, peer receiver window size,
4038 * number of unacked data chunks, and number of data chunks pending
4039 * receipt. This information is read-only.
4041 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4042 char __user *optval,
4045 struct sctp_status status;
4046 struct sctp_association *asoc = NULL;
4047 struct sctp_transport *transport;
4048 sctp_assoc_t associd;
4051 if (len < sizeof(status)) {
4056 len = sizeof(status);
4057 if (copy_from_user(&status, optval, len)) {
4062 associd = status.sstat_assoc_id;
4063 asoc = sctp_id2assoc(sk, associd);
4069 transport = asoc->peer.primary_path;
4071 status.sstat_assoc_id = sctp_assoc2id(asoc);
4072 status.sstat_state = asoc->state;
4073 status.sstat_rwnd = asoc->peer.rwnd;
4074 status.sstat_unackdata = asoc->unack_data;
4076 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4077 status.sstat_instrms = asoc->c.sinit_max_instreams;
4078 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4079 status.sstat_fragmentation_point = asoc->frag_point;
4080 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4081 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4082 transport->af_specific->sockaddr_len);
4083 /* Map ipv4 address into v4-mapped-on-v6 address. */
4084 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
4085 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4086 status.sstat_primary.spinfo_state = transport->state;
4087 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4088 status.sstat_primary.spinfo_srtt = transport->srtt;
4089 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4090 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4092 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4093 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4095 if (put_user(len, optlen)) {
4100 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
4101 len, status.sstat_state, status.sstat_rwnd,
4102 status.sstat_assoc_id);
4104 if (copy_to_user(optval, &status, len)) {
4114 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4116 * Applications can retrieve information about a specific peer address
4117 * of an association, including its reachability state, congestion
4118 * window, and retransmission timer values. This information is
4121 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4122 char __user *optval,
4125 struct sctp_paddrinfo pinfo;
4126 struct sctp_transport *transport;
4129 if (len < sizeof(pinfo)) {
4134 len = sizeof(pinfo);
4135 if (copy_from_user(&pinfo, optval, len)) {
4140 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4141 pinfo.spinfo_assoc_id);
4145 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4146 pinfo.spinfo_state = transport->state;
4147 pinfo.spinfo_cwnd = transport->cwnd;
4148 pinfo.spinfo_srtt = transport->srtt;
4149 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4150 pinfo.spinfo_mtu = transport->pathmtu;
4152 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4153 pinfo.spinfo_state = SCTP_ACTIVE;
4155 if (put_user(len, optlen)) {
4160 if (copy_to_user(optval, &pinfo, len)) {
4169 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4171 * This option is a on/off flag. If enabled no SCTP message
4172 * fragmentation will be performed. Instead if a message being sent
4173 * exceeds the current PMTU size, the message will NOT be sent and
4174 * instead a error will be indicated to the user.
4176 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4177 char __user *optval, int __user *optlen)
4181 if (len < sizeof(int))
4185 val = (sctp_sk(sk)->disable_fragments == 1);
4186 if (put_user(len, optlen))
4188 if (copy_to_user(optval, &val, len))
4193 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4195 * This socket option is used to specify various notifications and
4196 * ancillary data the user wishes to receive.
4198 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4203 if (len > sizeof(struct sctp_event_subscribe))
4204 len = sizeof(struct sctp_event_subscribe);
4205 if (put_user(len, optlen))
4207 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4212 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4214 * This socket option is applicable to the UDP-style socket only. When
4215 * set it will cause associations that are idle for more than the
4216 * specified number of seconds to automatically close. An association
4217 * being idle is defined an association that has NOT sent or received
4218 * user data. The special value of '0' indicates that no automatic
4219 * close of any associations should be performed. The option expects an
4220 * integer defining the number of seconds of idle time before an
4221 * association is closed.
4223 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4225 /* Applicable to UDP-style socket only */
4226 if (sctp_style(sk, TCP))
4228 if (len < sizeof(int))
4231 if (put_user(len, optlen))
4233 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4238 /* Helper routine to branch off an association to a new socket. */
4239 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
4240 struct socket **sockp)
4242 struct sock *sk = asoc->base.sk;
4243 struct sctp_sock *sp = sctp_sk(sk);
4244 struct socket *sock;
4247 /* Do not peel off from one netns to another one. */
4248 if (!net_eq(current->nsproxy->net_ns, sock_net(sk)))
4251 /* An association cannot be branched off from an already peeled-off
4252 * socket, nor is this supported for tcp style sockets.
4254 if (!sctp_style(sk, UDP))
4257 /* Create a new socket. */
4258 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4262 sctp_copy_sock(sock->sk, sk, asoc);
4264 /* Make peeled-off sockets more like 1-1 accepted sockets.
4265 * Set the daddr and initialize id to something more random
4267 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
4269 /* Populate the fields of the newsk from the oldsk and migrate the
4270 * asoc to the newsk.
4272 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4279 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4281 sctp_peeloff_arg_t peeloff;
4282 struct socket *newsock;
4284 struct sctp_association *asoc;
4286 if (len < sizeof(sctp_peeloff_arg_t))
4288 len = sizeof(sctp_peeloff_arg_t);
4289 if (copy_from_user(&peeloff, optval, len))
4292 asoc = sctp_id2assoc(sk, peeloff.associd);
4298 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
4300 retval = sctp_do_peeloff(asoc, &newsock);
4304 /* Map the socket to an unused fd that can be returned to the user. */
4305 retval = sock_map_fd(newsock, 0);
4307 sock_release(newsock);
4311 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
4312 __func__, sk, asoc, newsock->sk, retval);
4314 /* Return the fd mapped to the new socket. */
4315 peeloff.sd = retval;
4316 if (put_user(len, optlen))
4318 if (copy_to_user(optval, &peeloff, len))
4325 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4327 * Applications can enable or disable heartbeats for any peer address of
4328 * an association, modify an address's heartbeat interval, force a
4329 * heartbeat to be sent immediately, and adjust the address's maximum
4330 * number of retransmissions sent before an address is considered
4331 * unreachable. The following structure is used to access and modify an
4332 * address's parameters:
4334 * struct sctp_paddrparams {
4335 * sctp_assoc_t spp_assoc_id;
4336 * struct sockaddr_storage spp_address;
4337 * uint32_t spp_hbinterval;
4338 * uint16_t spp_pathmaxrxt;
4339 * uint32_t spp_pathmtu;
4340 * uint32_t spp_sackdelay;
4341 * uint32_t spp_flags;
4344 * spp_assoc_id - (one-to-many style socket) This is filled in the
4345 * application, and identifies the association for
4347 * spp_address - This specifies which address is of interest.
4348 * spp_hbinterval - This contains the value of the heartbeat interval,
4349 * in milliseconds. If a value of zero
4350 * is present in this field then no changes are to
4351 * be made to this parameter.
4352 * spp_pathmaxrxt - This contains the maximum number of
4353 * retransmissions before this address shall be
4354 * considered unreachable. If a value of zero
4355 * is present in this field then no changes are to
4356 * be made to this parameter.
4357 * spp_pathmtu - When Path MTU discovery is disabled the value
4358 * specified here will be the "fixed" path mtu.
4359 * Note that if the spp_address field is empty
4360 * then all associations on this address will
4361 * have this fixed path mtu set upon them.
4363 * spp_sackdelay - When delayed sack is enabled, this value specifies
4364 * the number of milliseconds that sacks will be delayed
4365 * for. This value will apply to all addresses of an
4366 * association if the spp_address field is empty. Note
4367 * also, that if delayed sack is enabled and this
4368 * value is set to 0, no change is made to the last
4369 * recorded delayed sack timer value.
4371 * spp_flags - These flags are used to control various features
4372 * on an association. The flag field may contain
4373 * zero or more of the following options.
4375 * SPP_HB_ENABLE - Enable heartbeats on the
4376 * specified address. Note that if the address
4377 * field is empty all addresses for the association
4378 * have heartbeats enabled upon them.
4380 * SPP_HB_DISABLE - Disable heartbeats on the
4381 * speicifed address. Note that if the address
4382 * field is empty all addresses for the association
4383 * will have their heartbeats disabled. Note also
4384 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4385 * mutually exclusive, only one of these two should
4386 * be specified. Enabling both fields will have
4387 * undetermined results.
4389 * SPP_HB_DEMAND - Request a user initiated heartbeat
4390 * to be made immediately.
4392 * SPP_PMTUD_ENABLE - This field will enable PMTU
4393 * discovery upon the specified address. Note that
4394 * if the address feild is empty then all addresses
4395 * on the association are effected.
4397 * SPP_PMTUD_DISABLE - This field will disable PMTU
4398 * discovery upon the specified address. Note that
4399 * if the address feild is empty then all addresses
4400 * on the association are effected. Not also that
4401 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4402 * exclusive. Enabling both will have undetermined
4405 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4406 * on delayed sack. The time specified in spp_sackdelay
4407 * is used to specify the sack delay for this address. Note
4408 * that if spp_address is empty then all addresses will
4409 * enable delayed sack and take on the sack delay
4410 * value specified in spp_sackdelay.
4411 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4412 * off delayed sack. If the spp_address field is blank then
4413 * delayed sack is disabled for the entire association. Note
4414 * also that this field is mutually exclusive to
4415 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4418 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4419 char __user *optval, int __user *optlen)
4421 struct sctp_paddrparams params;
4422 struct sctp_transport *trans = NULL;
4423 struct sctp_association *asoc = NULL;
4424 struct sctp_sock *sp = sctp_sk(sk);
4426 if (len < sizeof(struct sctp_paddrparams))
4428 len = sizeof(struct sctp_paddrparams);
4429 if (copy_from_user(¶ms, optval, len))
4432 /* If an address other than INADDR_ANY is specified, and
4433 * no transport is found, then the request is invalid.
4435 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
4436 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4437 params.spp_assoc_id);
4439 SCTP_DEBUG_PRINTK("Failed no transport\n");
4444 /* Get association, if assoc_id != 0 and the socket is a one
4445 * to many style socket, and an association was not found, then
4446 * the id was invalid.
4448 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4449 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4450 SCTP_DEBUG_PRINTK("Failed no association\n");
4455 /* Fetch transport values. */
4456 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4457 params.spp_pathmtu = trans->pathmtu;
4458 params.spp_pathmaxrxt = trans->pathmaxrxt;
4459 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4461 /*draft-11 doesn't say what to return in spp_flags*/
4462 params.spp_flags = trans->param_flags;
4464 /* Fetch association values. */
4465 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4466 params.spp_pathmtu = asoc->pathmtu;
4467 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4468 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4470 /*draft-11 doesn't say what to return in spp_flags*/
4471 params.spp_flags = asoc->param_flags;
4473 /* Fetch socket values. */
4474 params.spp_hbinterval = sp->hbinterval;
4475 params.spp_pathmtu = sp->pathmtu;
4476 params.spp_sackdelay = sp->sackdelay;
4477 params.spp_pathmaxrxt = sp->pathmaxrxt;
4479 /*draft-11 doesn't say what to return in spp_flags*/
4480 params.spp_flags = sp->param_flags;
4483 if (copy_to_user(optval, ¶ms, len))
4486 if (put_user(len, optlen))
4493 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4495 * This option will effect the way delayed acks are performed. This
4496 * option allows you to get or set the delayed ack time, in
4497 * milliseconds. It also allows changing the delayed ack frequency.
4498 * Changing the frequency to 1 disables the delayed sack algorithm. If
4499 * the assoc_id is 0, then this sets or gets the endpoints default
4500 * values. If the assoc_id field is non-zero, then the set or get
4501 * effects the specified association for the one to many model (the
4502 * assoc_id field is ignored by the one to one model). Note that if
4503 * sack_delay or sack_freq are 0 when setting this option, then the
4504 * current values will remain unchanged.
4506 * struct sctp_sack_info {
4507 * sctp_assoc_t sack_assoc_id;
4508 * uint32_t sack_delay;
4509 * uint32_t sack_freq;
4512 * sack_assoc_id - This parameter, indicates which association the user
4513 * is performing an action upon. Note that if this field's value is
4514 * zero then the endpoints default value is changed (effecting future
4515 * associations only).
4517 * sack_delay - This parameter contains the number of milliseconds that
4518 * the user is requesting the delayed ACK timer be set to. Note that
4519 * this value is defined in the standard to be between 200 and 500
4522 * sack_freq - This parameter contains the number of packets that must
4523 * be received before a sack is sent without waiting for the delay
4524 * timer to expire. The default value for this is 2, setting this
4525 * value to 1 will disable the delayed sack algorithm.
4527 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4528 char __user *optval,
4531 struct sctp_sack_info params;
4532 struct sctp_association *asoc = NULL;
4533 struct sctp_sock *sp = sctp_sk(sk);
4535 if (len >= sizeof(struct sctp_sack_info)) {
4536 len = sizeof(struct sctp_sack_info);
4538 if (copy_from_user(¶ms, optval, len))
4540 } else if (len == sizeof(struct sctp_assoc_value)) {
4541 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4542 pr_warn("Use struct sctp_sack_info instead\n");
4543 if (copy_from_user(¶ms, optval, len))
4548 /* Get association, if sack_assoc_id != 0 and the socket is a one
4549 * to many style socket, and an association was not found, then
4550 * the id was invalid.
4552 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4553 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4557 /* Fetch association values. */
4558 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4559 params.sack_delay = jiffies_to_msecs(
4561 params.sack_freq = asoc->sackfreq;
4564 params.sack_delay = 0;
4565 params.sack_freq = 1;
4568 /* Fetch socket values. */
4569 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4570 params.sack_delay = sp->sackdelay;
4571 params.sack_freq = sp->sackfreq;
4573 params.sack_delay = 0;
4574 params.sack_freq = 1;
4578 if (copy_to_user(optval, ¶ms, len))
4581 if (put_user(len, optlen))
4587 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4589 * Applications can specify protocol parameters for the default association
4590 * initialization. The option name argument to setsockopt() and getsockopt()
4593 * Setting initialization parameters is effective only on an unconnected
4594 * socket (for UDP-style sockets only future associations are effected
4595 * by the change). With TCP-style sockets, this option is inherited by
4596 * sockets derived from a listener socket.
4598 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4600 if (len < sizeof(struct sctp_initmsg))
4602 len = sizeof(struct sctp_initmsg);
4603 if (put_user(len, optlen))
4605 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4611 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4612 char __user *optval, int __user *optlen)
4614 struct sctp_association *asoc;
4616 struct sctp_getaddrs getaddrs;
4617 struct sctp_transport *from;
4619 union sctp_addr temp;
4620 struct sctp_sock *sp = sctp_sk(sk);
4625 if (len < sizeof(struct sctp_getaddrs))
4628 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4631 /* For UDP-style sockets, id specifies the association to query. */
4632 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4636 to = optval + offsetof(struct sctp_getaddrs,addrs);
4637 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4639 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4641 memcpy(&temp, &from->ipaddr, sizeof(temp));
4642 addrlen = sctp_get_pf_specific(sk->sk_family)
4643 ->addr_to_user(sp, &temp);
4644 if (space_left < addrlen)
4646 if (copy_to_user(to, &temp, addrlen))
4650 space_left -= addrlen;
4653 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4655 bytes_copied = ((char __user *)to) - optval;
4656 if (put_user(bytes_copied, optlen))
4662 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4663 size_t space_left, int *bytes_copied)
4665 struct sctp_sockaddr_entry *addr;
4666 union sctp_addr temp;
4671 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4675 if ((PF_INET == sk->sk_family) &&
4676 (AF_INET6 == addr->a.sa.sa_family))
4678 if ((PF_INET6 == sk->sk_family) &&
4679 inet_v6_ipv6only(sk) &&
4680 (AF_INET == addr->a.sa.sa_family))
4682 memcpy(&temp, &addr->a, sizeof(temp));
4683 if (!temp.v4.sin_port)
4684 temp.v4.sin_port = htons(port);
4686 addrlen = sctp_get_pf_specific(sk->sk_family)
4687 ->addr_to_user(sctp_sk(sk), &temp);
4689 if (space_left < addrlen) {
4693 memcpy(to, &temp, addrlen);
4697 space_left -= addrlen;
4698 *bytes_copied += addrlen;
4706 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4707 char __user *optval, int __user *optlen)
4709 struct sctp_bind_addr *bp;
4710 struct sctp_association *asoc;
4712 struct sctp_getaddrs getaddrs;
4713 struct sctp_sockaddr_entry *addr;
4715 union sctp_addr temp;
4716 struct sctp_sock *sp = sctp_sk(sk);
4720 int bytes_copied = 0;
4724 if (len < sizeof(struct sctp_getaddrs))
4727 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4731 * For UDP-style sockets, id specifies the association to query.
4732 * If the id field is set to the value '0' then the locally bound
4733 * addresses are returned without regard to any particular
4736 if (0 == getaddrs.assoc_id) {
4737 bp = &sctp_sk(sk)->ep->base.bind_addr;
4739 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4742 bp = &asoc->base.bind_addr;
4745 to = optval + offsetof(struct sctp_getaddrs,addrs);
4746 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4748 addrs = kmalloc(space_left, GFP_KERNEL);
4752 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4753 * addresses from the global local address list.
4755 if (sctp_list_single_entry(&bp->address_list)) {
4756 addr = list_entry(bp->address_list.next,
4757 struct sctp_sockaddr_entry, list);
4758 if (sctp_is_any(sk, &addr->a)) {
4759 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4760 space_left, &bytes_copied);
4770 /* Protection on the bound address list is not needed since
4771 * in the socket option context we hold a socket lock and
4772 * thus the bound address list can't change.
4774 list_for_each_entry(addr, &bp->address_list, list) {
4775 memcpy(&temp, &addr->a, sizeof(temp));
4776 addrlen = sctp_get_pf_specific(sk->sk_family)
4777 ->addr_to_user(sp, &temp);
4778 if (space_left < addrlen) {
4779 err = -ENOMEM; /*fixme: right error?*/
4782 memcpy(buf, &temp, addrlen);
4784 bytes_copied += addrlen;
4786 space_left -= addrlen;
4790 if (copy_to_user(to, addrs, bytes_copied)) {
4794 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4798 if (put_user(bytes_copied, optlen))
4805 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4807 * Requests that the local SCTP stack use the enclosed peer address as
4808 * the association primary. The enclosed address must be one of the
4809 * association peer's addresses.
4811 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4812 char __user *optval, int __user *optlen)
4814 struct sctp_prim prim;
4815 struct sctp_association *asoc;
4816 struct sctp_sock *sp = sctp_sk(sk);
4818 if (len < sizeof(struct sctp_prim))
4821 len = sizeof(struct sctp_prim);
4823 if (copy_from_user(&prim, optval, len))
4826 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4830 if (!asoc->peer.primary_path)
4833 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4834 asoc->peer.primary_path->af_specific->sockaddr_len);
4836 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
4837 (union sctp_addr *)&prim.ssp_addr);
4839 if (put_user(len, optlen))
4841 if (copy_to_user(optval, &prim, len))
4848 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4850 * Requests that the local endpoint set the specified Adaptation Layer
4851 * Indication parameter for all future INIT and INIT-ACK exchanges.
4853 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4854 char __user *optval, int __user *optlen)
4856 struct sctp_setadaptation adaptation;
4858 if (len < sizeof(struct sctp_setadaptation))
4861 len = sizeof(struct sctp_setadaptation);
4863 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4865 if (put_user(len, optlen))
4867 if (copy_to_user(optval, &adaptation, len))
4875 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4877 * Applications that wish to use the sendto() system call may wish to
4878 * specify a default set of parameters that would normally be supplied
4879 * through the inclusion of ancillary data. This socket option allows
4880 * such an application to set the default sctp_sndrcvinfo structure.
4883 * The application that wishes to use this socket option simply passes
4884 * in to this call the sctp_sndrcvinfo structure defined in Section
4885 * 5.2.2) The input parameters accepted by this call include
4886 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4887 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4888 * to this call if the caller is using the UDP model.
4890 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4892 static int sctp_getsockopt_default_send_param(struct sock *sk,
4893 int len, char __user *optval,
4896 struct sctp_sndrcvinfo info;
4897 struct sctp_association *asoc;
4898 struct sctp_sock *sp = sctp_sk(sk);
4900 if (len < sizeof(struct sctp_sndrcvinfo))
4903 len = sizeof(struct sctp_sndrcvinfo);
4905 if (copy_from_user(&info, optval, len))
4908 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4909 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4913 info.sinfo_stream = asoc->default_stream;
4914 info.sinfo_flags = asoc->default_flags;
4915 info.sinfo_ppid = asoc->default_ppid;
4916 info.sinfo_context = asoc->default_context;
4917 info.sinfo_timetolive = asoc->default_timetolive;
4919 info.sinfo_stream = sp->default_stream;
4920 info.sinfo_flags = sp->default_flags;
4921 info.sinfo_ppid = sp->default_ppid;
4922 info.sinfo_context = sp->default_context;
4923 info.sinfo_timetolive = sp->default_timetolive;
4926 if (put_user(len, optlen))
4928 if (copy_to_user(optval, &info, len))
4936 * 7.1.5 SCTP_NODELAY
4938 * Turn on/off any Nagle-like algorithm. This means that packets are
4939 * generally sent as soon as possible and no unnecessary delays are
4940 * introduced, at the cost of more packets in the network. Expects an
4941 * integer boolean flag.
4944 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4945 char __user *optval, int __user *optlen)
4949 if (len < sizeof(int))
4953 val = (sctp_sk(sk)->nodelay == 1);
4954 if (put_user(len, optlen))
4956 if (copy_to_user(optval, &val, len))
4963 * 7.1.1 SCTP_RTOINFO
4965 * The protocol parameters used to initialize and bound retransmission
4966 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4967 * and modify these parameters.
4968 * All parameters are time values, in milliseconds. A value of 0, when
4969 * modifying the parameters, indicates that the current value should not
4973 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4974 char __user *optval,
4975 int __user *optlen) {
4976 struct sctp_rtoinfo rtoinfo;
4977 struct sctp_association *asoc;
4979 if (len < sizeof (struct sctp_rtoinfo))
4982 len = sizeof(struct sctp_rtoinfo);
4984 if (copy_from_user(&rtoinfo, optval, len))
4987 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4989 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4992 /* Values corresponding to the specific association. */
4994 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4995 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4996 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4998 /* Values corresponding to the endpoint. */
4999 struct sctp_sock *sp = sctp_sk(sk);
5001 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5002 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5003 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5006 if (put_user(len, optlen))
5009 if (copy_to_user(optval, &rtoinfo, len))
5017 * 7.1.2 SCTP_ASSOCINFO
5019 * This option is used to tune the maximum retransmission attempts
5020 * of the association.
5021 * Returns an error if the new association retransmission value is
5022 * greater than the sum of the retransmission value of the peer.
5023 * See [SCTP] for more information.
5026 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5027 char __user *optval,
5031 struct sctp_assocparams assocparams;
5032 struct sctp_association *asoc;
5033 struct list_head *pos;
5036 if (len < sizeof (struct sctp_assocparams))
5039 len = sizeof(struct sctp_assocparams);
5041 if (copy_from_user(&assocparams, optval, len))
5044 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5046 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5049 /* Values correspoinding to the specific association */
5051 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5052 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5053 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5054 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
5056 (asoc->cookie_life.tv_usec
5059 list_for_each(pos, &asoc->peer.transport_addr_list) {
5063 assocparams.sasoc_number_peer_destinations = cnt;
5065 /* Values corresponding to the endpoint */
5066 struct sctp_sock *sp = sctp_sk(sk);
5068 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5069 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5070 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5071 assocparams.sasoc_cookie_life =
5072 sp->assocparams.sasoc_cookie_life;
5073 assocparams.sasoc_number_peer_destinations =
5075 sasoc_number_peer_destinations;
5078 if (put_user(len, optlen))
5081 if (copy_to_user(optval, &assocparams, len))
5088 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5090 * This socket option is a boolean flag which turns on or off mapped V4
5091 * addresses. If this option is turned on and the socket is type
5092 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5093 * If this option is turned off, then no mapping will be done of V4
5094 * addresses and a user will receive both PF_INET6 and PF_INET type
5095 * addresses on the socket.
5097 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5098 char __user *optval, int __user *optlen)
5101 struct sctp_sock *sp = sctp_sk(sk);
5103 if (len < sizeof(int))
5108 if (put_user(len, optlen))
5110 if (copy_to_user(optval, &val, len))
5117 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5118 * (chapter and verse is quoted at sctp_setsockopt_context())
5120 static int sctp_getsockopt_context(struct sock *sk, int len,
5121 char __user *optval, int __user *optlen)
5123 struct sctp_assoc_value params;
5124 struct sctp_sock *sp;
5125 struct sctp_association *asoc;
5127 if (len < sizeof(struct sctp_assoc_value))
5130 len = sizeof(struct sctp_assoc_value);
5132 if (copy_from_user(¶ms, optval, len))
5137 if (params.assoc_id != 0) {
5138 asoc = sctp_id2assoc(sk, params.assoc_id);
5141 params.assoc_value = asoc->default_rcv_context;
5143 params.assoc_value = sp->default_rcv_context;
5146 if (put_user(len, optlen))
5148 if (copy_to_user(optval, ¶ms, len))
5155 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5156 * This option will get or set the maximum size to put in any outgoing
5157 * SCTP DATA chunk. If a message is larger than this size it will be
5158 * fragmented by SCTP into the specified size. Note that the underlying
5159 * SCTP implementation may fragment into smaller sized chunks when the
5160 * PMTU of the underlying association is smaller than the value set by
5161 * the user. The default value for this option is '0' which indicates
5162 * the user is NOT limiting fragmentation and only the PMTU will effect
5163 * SCTP's choice of DATA chunk size. Note also that values set larger
5164 * than the maximum size of an IP datagram will effectively let SCTP
5165 * control fragmentation (i.e. the same as setting this option to 0).
5167 * The following structure is used to access and modify this parameter:
5169 * struct sctp_assoc_value {
5170 * sctp_assoc_t assoc_id;
5171 * uint32_t assoc_value;
5174 * assoc_id: This parameter is ignored for one-to-one style sockets.
5175 * For one-to-many style sockets this parameter indicates which
5176 * association the user is performing an action upon. Note that if
5177 * this field's value is zero then the endpoints default value is
5178 * changed (effecting future associations only).
5179 * assoc_value: This parameter specifies the maximum size in bytes.
5181 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5182 char __user *optval, int __user *optlen)
5184 struct sctp_assoc_value params;
5185 struct sctp_association *asoc;
5187 if (len == sizeof(int)) {
5188 pr_warn("Use of int in maxseg socket option deprecated\n");
5189 pr_warn("Use struct sctp_assoc_value instead\n");
5190 params.assoc_id = 0;
5191 } else if (len >= sizeof(struct sctp_assoc_value)) {
5192 len = sizeof(struct sctp_assoc_value);
5193 if (copy_from_user(¶ms, optval, sizeof(params)))
5198 asoc = sctp_id2assoc(sk, params.assoc_id);
5199 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5203 params.assoc_value = asoc->frag_point;
5205 params.assoc_value = sctp_sk(sk)->user_frag;
5207 if (put_user(len, optlen))
5209 if (len == sizeof(int)) {
5210 if (copy_to_user(optval, ¶ms.assoc_value, len))
5213 if (copy_to_user(optval, ¶ms, len))
5221 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5222 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5224 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5225 char __user *optval, int __user *optlen)
5229 if (len < sizeof(int))
5234 val = sctp_sk(sk)->frag_interleave;
5235 if (put_user(len, optlen))
5237 if (copy_to_user(optval, &val, len))
5244 * 7.1.25. Set or Get the sctp partial delivery point
5245 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5247 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5248 char __user *optval,
5253 if (len < sizeof(u32))
5258 val = sctp_sk(sk)->pd_point;
5259 if (put_user(len, optlen))
5261 if (copy_to_user(optval, &val, len))
5268 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5269 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5271 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5272 char __user *optval,
5275 struct sctp_assoc_value params;
5276 struct sctp_sock *sp;
5277 struct sctp_association *asoc;
5279 if (len == sizeof(int)) {
5280 pr_warn("Use of int in max_burst socket option deprecated\n");
5281 pr_warn("Use struct sctp_assoc_value instead\n");
5282 params.assoc_id = 0;
5283 } else if (len >= sizeof(struct sctp_assoc_value)) {
5284 len = sizeof(struct sctp_assoc_value);
5285 if (copy_from_user(¶ms, optval, len))
5292 if (params.assoc_id != 0) {
5293 asoc = sctp_id2assoc(sk, params.assoc_id);
5296 params.assoc_value = asoc->max_burst;
5298 params.assoc_value = sp->max_burst;
5300 if (len == sizeof(int)) {
5301 if (copy_to_user(optval, ¶ms.assoc_value, len))
5304 if (copy_to_user(optval, ¶ms, len))
5312 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5313 char __user *optval, int __user *optlen)
5315 struct sctp_hmacalgo __user *p = (void __user *)optval;
5316 struct sctp_hmac_algo_param *hmacs;
5321 if (!sctp_auth_enable)
5324 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5325 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5327 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5330 len = sizeof(struct sctp_hmacalgo) + data_len;
5331 num_idents = data_len / sizeof(u16);
5333 if (put_user(len, optlen))
5335 if (put_user(num_idents, &p->shmac_num_idents))
5337 for (i = 0; i < num_idents; i++) {
5338 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
5340 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
5346 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5347 char __user *optval, int __user *optlen)
5349 struct sctp_authkeyid val;
5350 struct sctp_association *asoc;
5352 if (!sctp_auth_enable)
5355 if (len < sizeof(struct sctp_authkeyid))
5357 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5360 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5361 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5365 val.scact_keynumber = asoc->active_key_id;
5367 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5369 len = sizeof(struct sctp_authkeyid);
5370 if (put_user(len, optlen))
5372 if (copy_to_user(optval, &val, len))
5378 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5379 char __user *optval, int __user *optlen)
5381 struct sctp_authchunks __user *p = (void __user *)optval;
5382 struct sctp_authchunks val;
5383 struct sctp_association *asoc;
5384 struct sctp_chunks_param *ch;
5388 if (!sctp_auth_enable)
5391 if (len < sizeof(struct sctp_authchunks))
5394 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5397 to = p->gauth_chunks;
5398 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5402 ch = asoc->peer.peer_chunks;
5406 /* See if the user provided enough room for all the data */
5407 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5408 if (len < num_chunks)
5411 if (copy_to_user(to, ch->chunks, num_chunks))
5414 len = sizeof(struct sctp_authchunks) + num_chunks;
5415 if (put_user(len, optlen)) return -EFAULT;
5416 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5421 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5422 char __user *optval, int __user *optlen)
5424 struct sctp_authchunks __user *p = (void __user *)optval;
5425 struct sctp_authchunks val;
5426 struct sctp_association *asoc;
5427 struct sctp_chunks_param *ch;
5431 if (!sctp_auth_enable)
5434 if (len < sizeof(struct sctp_authchunks))
5437 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5440 to = p->gauth_chunks;
5441 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5442 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5446 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5448 ch = sctp_sk(sk)->ep->auth_chunk_list;
5453 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5454 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5457 if (copy_to_user(to, ch->chunks, num_chunks))
5460 len = sizeof(struct sctp_authchunks) + num_chunks;
5461 if (put_user(len, optlen))
5463 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5470 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5471 * This option gets the current number of associations that are attached
5472 * to a one-to-many style socket. The option value is an uint32_t.
5474 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5475 char __user *optval, int __user *optlen)
5477 struct sctp_sock *sp = sctp_sk(sk);
5478 struct sctp_association *asoc;
5481 if (sctp_style(sk, TCP))
5484 if (len < sizeof(u32))
5489 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5493 if (put_user(len, optlen))
5495 if (copy_to_user(optval, &val, len))
5502 * 8.1.23 SCTP_AUTO_ASCONF
5503 * See the corresponding setsockopt entry as description
5505 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5506 char __user *optval, int __user *optlen)
5510 if (len < sizeof(int))
5514 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5516 if (put_user(len, optlen))
5518 if (copy_to_user(optval, &val, len))
5524 * 8.2.6. Get the Current Identifiers of Associations
5525 * (SCTP_GET_ASSOC_ID_LIST)
5527 * This option gets the current list of SCTP association identifiers of
5528 * the SCTP associations handled by a one-to-many style socket.
5530 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5531 char __user *optval, int __user *optlen)
5533 struct sctp_sock *sp = sctp_sk(sk);
5534 struct sctp_association *asoc;
5535 struct sctp_assoc_ids *ids;
5538 if (sctp_style(sk, TCP))
5541 if (len < sizeof(struct sctp_assoc_ids))
5544 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5548 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5551 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5553 ids = kmalloc(len, GFP_KERNEL);
5557 ids->gaids_number_of_ids = num;
5559 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5560 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5563 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5572 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5573 char __user *optval, int __user *optlen)
5578 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5581 /* I can hardly begin to describe how wrong this is. This is
5582 * so broken as to be worse than useless. The API draft
5583 * REALLY is NOT helpful here... I am not convinced that the
5584 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5585 * are at all well-founded.
5587 if (level != SOL_SCTP) {
5588 struct sctp_af *af = sctp_sk(sk)->pf->af;
5590 retval = af->getsockopt(sk, level, optname, optval, optlen);
5594 if (get_user(len, optlen))
5604 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5606 case SCTP_DISABLE_FRAGMENTS:
5607 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5611 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5613 case SCTP_AUTOCLOSE:
5614 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5616 case SCTP_SOCKOPT_PEELOFF:
5617 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5619 case SCTP_PEER_ADDR_PARAMS:
5620 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5623 case SCTP_DELAYED_SACK:
5624 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5628 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5630 case SCTP_GET_PEER_ADDRS:
5631 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5634 case SCTP_GET_LOCAL_ADDRS:
5635 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5638 case SCTP_SOCKOPT_CONNECTX3:
5639 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5641 case SCTP_DEFAULT_SEND_PARAM:
5642 retval = sctp_getsockopt_default_send_param(sk, len,
5645 case SCTP_PRIMARY_ADDR:
5646 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5649 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5652 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5654 case SCTP_ASSOCINFO:
5655 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5657 case SCTP_I_WANT_MAPPED_V4_ADDR:
5658 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5661 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5663 case SCTP_GET_PEER_ADDR_INFO:
5664 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5667 case SCTP_ADAPTATION_LAYER:
5668 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5672 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5674 case SCTP_FRAGMENT_INTERLEAVE:
5675 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5678 case SCTP_PARTIAL_DELIVERY_POINT:
5679 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5682 case SCTP_MAX_BURST:
5683 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5686 case SCTP_AUTH_CHUNK:
5687 case SCTP_AUTH_DELETE_KEY:
5688 retval = -EOPNOTSUPP;
5690 case SCTP_HMAC_IDENT:
5691 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5693 case SCTP_AUTH_ACTIVE_KEY:
5694 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5696 case SCTP_PEER_AUTH_CHUNKS:
5697 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5700 case SCTP_LOCAL_AUTH_CHUNKS:
5701 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5704 case SCTP_GET_ASSOC_NUMBER:
5705 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5707 case SCTP_GET_ASSOC_ID_LIST:
5708 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5710 case SCTP_AUTO_ASCONF:
5711 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5714 retval = -ENOPROTOOPT;
5718 sctp_release_sock(sk);
5722 static void sctp_hash(struct sock *sk)
5727 static void sctp_unhash(struct sock *sk)
5732 /* Check if port is acceptable. Possibly find first available port.
5734 * The port hash table (contained in the 'global' SCTP protocol storage
5735 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5736 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5737 * list (the list number is the port number hashed out, so as you
5738 * would expect from a hash function, all the ports in a given list have
5739 * such a number that hashes out to the same list number; you were
5740 * expecting that, right?); so each list has a set of ports, with a
5741 * link to the socket (struct sock) that uses it, the port number and
5742 * a fastreuse flag (FIXME: NPI ipg).
5744 static struct sctp_bind_bucket *sctp_bucket_create(
5745 struct sctp_bind_hashbucket *head, unsigned short snum);
5747 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5749 struct sctp_bind_hashbucket *head; /* hash list */
5750 struct sctp_bind_bucket *pp; /* hash list port iterator */
5751 struct hlist_node *node;
5752 unsigned short snum;
5755 snum = ntohs(addr->v4.sin_port);
5757 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5758 sctp_local_bh_disable();
5761 /* Search for an available port. */
5762 int low, high, remaining, index;
5765 inet_get_local_port_range(&low, &high);
5766 remaining = (high - low) + 1;
5767 rover = net_random() % remaining + low;
5771 if ((rover < low) || (rover > high))
5773 if (inet_is_reserved_local_port(rover))
5775 index = sctp_phashfn(rover);
5776 head = &sctp_port_hashtable[index];
5777 sctp_spin_lock(&head->lock);
5778 sctp_for_each_hentry(pp, node, &head->chain)
5779 if (pp->port == rover)
5783 sctp_spin_unlock(&head->lock);
5784 } while (--remaining > 0);
5786 /* Exhausted local port range during search? */
5791 /* OK, here is the one we will use. HEAD (the port
5792 * hash table list entry) is non-NULL and we hold it's
5797 /* We are given an specific port number; we verify
5798 * that it is not being used. If it is used, we will
5799 * exahust the search in the hash list corresponding
5800 * to the port number (snum) - we detect that with the
5801 * port iterator, pp being NULL.
5803 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5804 sctp_spin_lock(&head->lock);
5805 sctp_for_each_hentry(pp, node, &head->chain) {
5806 if (pp->port == snum)
5813 if (!hlist_empty(&pp->owner)) {
5814 /* We had a port hash table hit - there is an
5815 * available port (pp != NULL) and it is being
5816 * used by other socket (pp->owner not empty); that other
5817 * socket is going to be sk2.
5819 int reuse = sk->sk_reuse;
5822 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5823 if (pp->fastreuse && sk->sk_reuse &&
5824 sk->sk_state != SCTP_SS_LISTENING)
5827 /* Run through the list of sockets bound to the port
5828 * (pp->port) [via the pointers bind_next and
5829 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5830 * we get the endpoint they describe and run through
5831 * the endpoint's list of IP (v4 or v6) addresses,
5832 * comparing each of the addresses with the address of
5833 * the socket sk. If we find a match, then that means
5834 * that this port/socket (sk) combination are already
5837 sk_for_each_bound(sk2, node, &pp->owner) {
5838 struct sctp_endpoint *ep2;
5839 ep2 = sctp_sk(sk2)->ep;
5842 (reuse && sk2->sk_reuse &&
5843 sk2->sk_state != SCTP_SS_LISTENING))
5846 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5847 sctp_sk(sk2), sctp_sk(sk))) {
5852 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5855 /* If there was a hash table miss, create a new port. */
5857 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5860 /* In either case (hit or miss), make sure fastreuse is 1 only
5861 * if sk->sk_reuse is too (that is, if the caller requested
5862 * SO_REUSEADDR on this socket -sk-).
5864 if (hlist_empty(&pp->owner)) {
5865 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5869 } else if (pp->fastreuse &&
5870 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5873 /* We are set, so fill up all the data in the hash table
5874 * entry, tie the socket list information with the rest of the
5875 * sockets FIXME: Blurry, NPI (ipg).
5878 if (!sctp_sk(sk)->bind_hash) {
5879 inet_sk(sk)->inet_num = snum;
5880 sk_add_bind_node(sk, &pp->owner);
5881 sctp_sk(sk)->bind_hash = pp;
5886 sctp_spin_unlock(&head->lock);
5889 sctp_local_bh_enable();
5893 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5894 * port is requested.
5896 static int sctp_get_port(struct sock *sk, unsigned short snum)
5899 union sctp_addr addr;
5900 struct sctp_af *af = sctp_sk(sk)->pf->af;
5902 /* Set up a dummy address struct from the sk. */
5903 af->from_sk(&addr, sk);
5904 addr.v4.sin_port = htons(snum);
5906 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5907 ret = sctp_get_port_local(sk, &addr);
5913 * Move a socket to LISTENING state.
5915 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5917 struct sctp_sock *sp = sctp_sk(sk);
5918 struct sctp_endpoint *ep = sp->ep;
5919 struct crypto_hash *tfm = NULL;
5921 /* Allocate HMAC for generating cookie. */
5922 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5923 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5925 if (net_ratelimit()) {
5926 pr_info("failed to load transform for %s: %ld\n",
5927 sctp_hmac_alg, PTR_ERR(tfm));
5931 sctp_sk(sk)->hmac = tfm;
5935 * If a bind() or sctp_bindx() is not called prior to a listen()
5936 * call that allows new associations to be accepted, the system
5937 * picks an ephemeral port and will choose an address set equivalent
5938 * to binding with a wildcard address.
5940 * This is not currently spelled out in the SCTP sockets
5941 * extensions draft, but follows the practice as seen in TCP
5945 sk->sk_state = SCTP_SS_LISTENING;
5946 if (!ep->base.bind_addr.port) {
5947 if (sctp_autobind(sk))
5950 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
5951 sk->sk_state = SCTP_SS_CLOSED;
5956 sk->sk_max_ack_backlog = backlog;
5957 sctp_hash_endpoint(ep);
5962 * 4.1.3 / 5.1.3 listen()
5964 * By default, new associations are not accepted for UDP style sockets.
5965 * An application uses listen() to mark a socket as being able to
5966 * accept new associations.
5968 * On TCP style sockets, applications use listen() to ready the SCTP
5969 * endpoint for accepting inbound associations.
5971 * On both types of endpoints a backlog of '0' disables listening.
5973 * Move a socket to LISTENING state.
5975 int sctp_inet_listen(struct socket *sock, int backlog)
5977 struct sock *sk = sock->sk;
5978 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5981 if (unlikely(backlog < 0))
5986 /* Peeled-off sockets are not allowed to listen(). */
5987 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
5990 if (sock->state != SS_UNCONNECTED)
5993 /* If backlog is zero, disable listening. */
5995 if (sctp_sstate(sk, CLOSED))
5999 sctp_unhash_endpoint(ep);
6000 sk->sk_state = SCTP_SS_CLOSED;
6002 sctp_sk(sk)->bind_hash->fastreuse = 1;
6006 /* If we are already listening, just update the backlog */
6007 if (sctp_sstate(sk, LISTENING))
6008 sk->sk_max_ack_backlog = backlog;
6010 err = sctp_listen_start(sk, backlog);
6017 sctp_release_sock(sk);
6022 * This function is done by modeling the current datagram_poll() and the
6023 * tcp_poll(). Note that, based on these implementations, we don't
6024 * lock the socket in this function, even though it seems that,
6025 * ideally, locking or some other mechanisms can be used to ensure
6026 * the integrity of the counters (sndbuf and wmem_alloc) used
6027 * in this place. We assume that we don't need locks either until proven
6030 * Another thing to note is that we include the Async I/O support
6031 * here, again, by modeling the current TCP/UDP code. We don't have
6032 * a good way to test with it yet.
6034 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6036 struct sock *sk = sock->sk;
6037 struct sctp_sock *sp = sctp_sk(sk);
6040 poll_wait(file, sk_sleep(sk), wait);
6042 /* A TCP-style listening socket becomes readable when the accept queue
6045 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6046 return (!list_empty(&sp->ep->asocs)) ?
6047 (POLLIN | POLLRDNORM) : 0;
6051 /* Is there any exceptional events? */
6052 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6054 if (sk->sk_shutdown & RCV_SHUTDOWN)
6055 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6056 if (sk->sk_shutdown == SHUTDOWN_MASK)
6059 /* Is it readable? Reconsider this code with TCP-style support. */
6060 if (!skb_queue_empty(&sk->sk_receive_queue))
6061 mask |= POLLIN | POLLRDNORM;
6063 /* The association is either gone or not ready. */
6064 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6067 /* Is it writable? */
6068 if (sctp_writeable(sk)) {
6069 mask |= POLLOUT | POLLWRNORM;
6071 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6073 * Since the socket is not locked, the buffer
6074 * might be made available after the writeable check and
6075 * before the bit is set. This could cause a lost I/O
6076 * signal. tcp_poll() has a race breaker for this race
6077 * condition. Based on their implementation, we put
6078 * in the following code to cover it as well.
6080 if (sctp_writeable(sk))
6081 mask |= POLLOUT | POLLWRNORM;
6086 /********************************************************************
6087 * 2nd Level Abstractions
6088 ********************************************************************/
6090 static struct sctp_bind_bucket *sctp_bucket_create(
6091 struct sctp_bind_hashbucket *head, unsigned short snum)
6093 struct sctp_bind_bucket *pp;
6095 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6097 SCTP_DBG_OBJCNT_INC(bind_bucket);
6100 INIT_HLIST_HEAD(&pp->owner);
6101 hlist_add_head(&pp->node, &head->chain);
6106 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6107 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6109 if (pp && hlist_empty(&pp->owner)) {
6110 __hlist_del(&pp->node);
6111 kmem_cache_free(sctp_bucket_cachep, pp);
6112 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6116 /* Release this socket's reference to a local port. */
6117 static inline void __sctp_put_port(struct sock *sk)
6119 struct sctp_bind_hashbucket *head =
6120 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->inet_num)];
6121 struct sctp_bind_bucket *pp;
6123 sctp_spin_lock(&head->lock);
6124 pp = sctp_sk(sk)->bind_hash;
6125 __sk_del_bind_node(sk);
6126 sctp_sk(sk)->bind_hash = NULL;
6127 inet_sk(sk)->inet_num = 0;
6128 sctp_bucket_destroy(pp);
6129 sctp_spin_unlock(&head->lock);
6132 void sctp_put_port(struct sock *sk)
6134 sctp_local_bh_disable();
6135 __sctp_put_port(sk);
6136 sctp_local_bh_enable();
6140 * The system picks an ephemeral port and choose an address set equivalent
6141 * to binding with a wildcard address.
6142 * One of those addresses will be the primary address for the association.
6143 * This automatically enables the multihoming capability of SCTP.
6145 static int sctp_autobind(struct sock *sk)
6147 union sctp_addr autoaddr;
6151 /* Initialize a local sockaddr structure to INADDR_ANY. */
6152 af = sctp_sk(sk)->pf->af;
6154 port = htons(inet_sk(sk)->inet_num);
6155 af->inaddr_any(&autoaddr, port);
6157 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6160 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6163 * 4.2 The cmsghdr Structure *
6165 * When ancillary data is sent or received, any number of ancillary data
6166 * objects can be specified by the msg_control and msg_controllen members of
6167 * the msghdr structure, because each object is preceded by
6168 * a cmsghdr structure defining the object's length (the cmsg_len member).
6169 * Historically Berkeley-derived implementations have passed only one object
6170 * at a time, but this API allows multiple objects to be
6171 * passed in a single call to sendmsg() or recvmsg(). The following example
6172 * shows two ancillary data objects in a control buffer.
6174 * |<--------------------------- msg_controllen -------------------------->|
6177 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6179 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6182 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6184 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6187 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6188 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6190 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6192 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6199 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
6200 sctp_cmsgs_t *cmsgs)
6202 struct cmsghdr *cmsg;
6203 struct msghdr *my_msg = (struct msghdr *)msg;
6205 for (cmsg = CMSG_FIRSTHDR(msg);
6207 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6208 if (!CMSG_OK(my_msg, cmsg))
6211 /* Should we parse this header or ignore? */
6212 if (cmsg->cmsg_level != IPPROTO_SCTP)
6215 /* Strictly check lengths following example in SCM code. */
6216 switch (cmsg->cmsg_type) {
6218 /* SCTP Socket API Extension
6219 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6221 * This cmsghdr structure provides information for
6222 * initializing new SCTP associations with sendmsg().
6223 * The SCTP_INITMSG socket option uses this same data
6224 * structure. This structure is not used for
6227 * cmsg_level cmsg_type cmsg_data[]
6228 * ------------ ------------ ----------------------
6229 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6231 if (cmsg->cmsg_len !=
6232 CMSG_LEN(sizeof(struct sctp_initmsg)))
6234 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6238 /* SCTP Socket API Extension
6239 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6241 * This cmsghdr structure specifies SCTP options for
6242 * sendmsg() and describes SCTP header information
6243 * about a received message through recvmsg().
6245 * cmsg_level cmsg_type cmsg_data[]
6246 * ------------ ------------ ----------------------
6247 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6249 if (cmsg->cmsg_len !=
6250 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6254 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6256 /* Minimally, validate the sinfo_flags. */
6257 if (cmsgs->info->sinfo_flags &
6258 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6259 SCTP_SACK_IMMEDIATELY |
6260 SCTP_ABORT | SCTP_EOF))
6272 * Wait for a packet..
6273 * Note: This function is the same function as in core/datagram.c
6274 * with a few modifications to make lksctp work.
6276 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6281 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6283 /* Socket errors? */
6284 error = sock_error(sk);
6288 if (!skb_queue_empty(&sk->sk_receive_queue))
6291 /* Socket shut down? */
6292 if (sk->sk_shutdown & RCV_SHUTDOWN)
6295 /* Sequenced packets can come disconnected. If so we report the
6300 /* Is there a good reason to think that we may receive some data? */
6301 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6304 /* Handle signals. */
6305 if (signal_pending(current))
6308 /* Let another process have a go. Since we are going to sleep
6309 * anyway. Note: This may cause odd behaviors if the message
6310 * does not fit in the user's buffer, but this seems to be the
6311 * only way to honor MSG_DONTWAIT realistically.
6313 sctp_release_sock(sk);
6314 *timeo_p = schedule_timeout(*timeo_p);
6318 finish_wait(sk_sleep(sk), &wait);
6322 error = sock_intr_errno(*timeo_p);
6325 finish_wait(sk_sleep(sk), &wait);
6330 /* Receive a datagram.
6331 * Note: This is pretty much the same routine as in core/datagram.c
6332 * with a few changes to make lksctp work.
6334 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6335 int noblock, int *err)
6338 struct sk_buff *skb;
6341 timeo = sock_rcvtimeo(sk, noblock);
6343 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6344 timeo, MAX_SCHEDULE_TIMEOUT);
6347 /* Again only user level code calls this function,
6348 * so nothing interrupt level
6349 * will suddenly eat the receive_queue.
6351 * Look at current nfs client by the way...
6352 * However, this function was correct in any case. 8)
6354 if (flags & MSG_PEEK) {
6355 spin_lock_bh(&sk->sk_receive_queue.lock);
6356 skb = skb_peek(&sk->sk_receive_queue);
6358 atomic_inc(&skb->users);
6359 spin_unlock_bh(&sk->sk_receive_queue.lock);
6361 skb = skb_dequeue(&sk->sk_receive_queue);
6367 /* Caller is allowed not to check sk->sk_err before calling. */
6368 error = sock_error(sk);
6372 if (sk->sk_shutdown & RCV_SHUTDOWN)
6375 /* User doesn't want to wait. */
6379 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6388 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6389 static void __sctp_write_space(struct sctp_association *asoc)
6391 struct sock *sk = asoc->base.sk;
6392 struct socket *sock = sk->sk_socket;
6394 if ((sctp_wspace(asoc) > 0) && sock) {
6395 if (waitqueue_active(&asoc->wait))
6396 wake_up_interruptible(&asoc->wait);
6398 if (sctp_writeable(sk)) {
6399 wait_queue_head_t *wq = sk_sleep(sk);
6401 if (wq && waitqueue_active(wq))
6402 wake_up_interruptible(wq);
6404 /* Note that we try to include the Async I/O support
6405 * here by modeling from the current TCP/UDP code.
6406 * We have not tested with it yet.
6408 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6409 sock_wake_async(sock,
6410 SOCK_WAKE_SPACE, POLL_OUT);
6415 /* Do accounting for the sndbuf space.
6416 * Decrement the used sndbuf space of the corresponding association by the
6417 * data size which was just transmitted(freed).
6419 static void sctp_wfree(struct sk_buff *skb)
6421 struct sctp_association *asoc;
6422 struct sctp_chunk *chunk;
6425 /* Get the saved chunk pointer. */
6426 chunk = *((struct sctp_chunk **)(skb->cb));
6429 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6430 sizeof(struct sk_buff) +
6431 sizeof(struct sctp_chunk);
6433 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6436 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6438 sk->sk_wmem_queued -= skb->truesize;
6439 sk_mem_uncharge(sk, skb->truesize);
6442 __sctp_write_space(asoc);
6444 sctp_association_put(asoc);
6447 /* Do accounting for the receive space on the socket.
6448 * Accounting for the association is done in ulpevent.c
6449 * We set this as a destructor for the cloned data skbs so that
6450 * accounting is done at the correct time.
6452 void sctp_sock_rfree(struct sk_buff *skb)
6454 struct sock *sk = skb->sk;
6455 struct sctp_ulpevent *event = sctp_skb2event(skb);
6457 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6460 * Mimic the behavior of sock_rfree
6462 sk_mem_uncharge(sk, event->rmem_len);
6466 /* Helper function to wait for space in the sndbuf. */
6467 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6468 size_t msg_len, struct sock **orig_sk)
6470 struct sock *sk = asoc->base.sk;
6472 long current_timeo = *timeo_p;
6475 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6476 asoc, (long)(*timeo_p), msg_len);
6478 /* Increment the association's refcnt. */
6479 sctp_association_hold(asoc);
6481 /* Wait on the association specific sndbuf space. */
6483 prepare_to_wait_exclusive(&asoc->wait, &wait,
6484 TASK_INTERRUPTIBLE);
6487 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6490 if (signal_pending(current))
6491 goto do_interrupted;
6492 if (msg_len <= sctp_wspace(asoc))
6495 /* Let another process have a go. Since we are going
6498 sctp_release_sock(sk);
6499 current_timeo = schedule_timeout(current_timeo);
6501 if (sk != asoc->base.sk) {
6507 *timeo_p = current_timeo;
6512 finish_wait(&asoc->wait, &wait);
6514 /* Release the association's refcnt. */
6515 sctp_association_put(asoc);
6524 err = sock_intr_errno(*timeo_p);
6532 void sctp_data_ready(struct sock *sk, int len)
6534 struct socket_wq *wq;
6537 wq = rcu_dereference(sk->sk_wq);
6538 if (wq_has_sleeper(wq))
6539 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6540 POLLRDNORM | POLLRDBAND);
6541 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6545 /* If socket sndbuf has changed, wake up all per association waiters. */
6546 void sctp_write_space(struct sock *sk)
6548 struct sctp_association *asoc;
6550 /* Wake up the tasks in each wait queue. */
6551 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6552 __sctp_write_space(asoc);
6556 /* Is there any sndbuf space available on the socket?
6558 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6559 * associations on the same socket. For a UDP-style socket with
6560 * multiple associations, it is possible for it to be "unwriteable"
6561 * prematurely. I assume that this is acceptable because
6562 * a premature "unwriteable" is better than an accidental "writeable" which
6563 * would cause an unwanted block under certain circumstances. For the 1-1
6564 * UDP-style sockets or TCP-style sockets, this code should work.
6567 static int sctp_writeable(struct sock *sk)
6571 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6577 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6578 * returns immediately with EINPROGRESS.
6580 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6582 struct sock *sk = asoc->base.sk;
6584 long current_timeo = *timeo_p;
6587 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6590 /* Increment the association's refcnt. */
6591 sctp_association_hold(asoc);
6594 prepare_to_wait_exclusive(&asoc->wait, &wait,
6595 TASK_INTERRUPTIBLE);
6598 if (sk->sk_shutdown & RCV_SHUTDOWN)
6600 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6603 if (signal_pending(current))
6604 goto do_interrupted;
6606 if (sctp_state(asoc, ESTABLISHED))
6609 /* Let another process have a go. Since we are going
6612 sctp_release_sock(sk);
6613 current_timeo = schedule_timeout(current_timeo);
6616 *timeo_p = current_timeo;
6620 finish_wait(&asoc->wait, &wait);
6622 /* Release the association's refcnt. */
6623 sctp_association_put(asoc);
6628 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6631 err = -ECONNREFUSED;
6635 err = sock_intr_errno(*timeo_p);
6643 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6645 struct sctp_endpoint *ep;
6649 ep = sctp_sk(sk)->ep;
6653 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6654 TASK_INTERRUPTIBLE);
6656 if (list_empty(&ep->asocs)) {
6657 sctp_release_sock(sk);
6658 timeo = schedule_timeout(timeo);
6663 if (!sctp_sstate(sk, LISTENING))
6667 if (!list_empty(&ep->asocs))
6670 err = sock_intr_errno(timeo);
6671 if (signal_pending(current))
6679 finish_wait(sk_sleep(sk), &wait);
6684 static void sctp_wait_for_close(struct sock *sk, long timeout)
6689 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6690 if (list_empty(&sctp_sk(sk)->ep->asocs))
6692 sctp_release_sock(sk);
6693 timeout = schedule_timeout(timeout);
6695 } while (!signal_pending(current) && timeout);
6697 finish_wait(sk_sleep(sk), &wait);
6700 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6702 struct sk_buff *frag;
6707 /* Don't forget the fragments. */
6708 skb_walk_frags(skb, frag)
6709 sctp_skb_set_owner_r_frag(frag, sk);
6712 sctp_skb_set_owner_r(skb, sk);
6715 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6716 struct sctp_association *asoc)
6718 struct inet_sock *inet = inet_sk(sk);
6719 struct inet_sock *newinet;
6721 newsk->sk_type = sk->sk_type;
6722 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6723 newsk->sk_flags = sk->sk_flags;
6724 newsk->sk_no_check = sk->sk_no_check;
6725 newsk->sk_reuse = sk->sk_reuse;
6727 newsk->sk_shutdown = sk->sk_shutdown;
6728 newsk->sk_destruct = inet_sock_destruct;
6729 newsk->sk_family = sk->sk_family;
6730 newsk->sk_protocol = IPPROTO_SCTP;
6731 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6732 newsk->sk_sndbuf = sk->sk_sndbuf;
6733 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6734 newsk->sk_lingertime = sk->sk_lingertime;
6735 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6736 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6738 newinet = inet_sk(newsk);
6740 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6741 * getsockname() and getpeername()
6743 newinet->inet_sport = inet->inet_sport;
6744 newinet->inet_saddr = inet->inet_saddr;
6745 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6746 newinet->inet_dport = htons(asoc->peer.port);
6747 newinet->pmtudisc = inet->pmtudisc;
6748 newinet->inet_id = asoc->next_tsn ^ jiffies;
6750 newinet->uc_ttl = inet->uc_ttl;
6751 newinet->mc_loop = 1;
6752 newinet->mc_ttl = 1;
6753 newinet->mc_index = 0;
6754 newinet->mc_list = NULL;
6756 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
6757 net_enable_timestamp();
6760 static inline void sctp_copy_descendant(struct sock *sk_to,
6761 const struct sock *sk_from)
6763 int ancestor_size = sizeof(struct inet_sock) +
6764 sizeof(struct sctp_sock) -
6765 offsetof(struct sctp_sock, auto_asconf_list);
6767 if (sk_from->sk_family == PF_INET6)
6768 ancestor_size += sizeof(struct ipv6_pinfo);
6770 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
6773 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6774 * and its messages to the newsk.
6776 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6777 struct sctp_association *assoc,
6778 sctp_socket_type_t type)
6780 struct sctp_sock *oldsp = sctp_sk(oldsk);
6781 struct sctp_sock *newsp = sctp_sk(newsk);
6782 struct sctp_bind_bucket *pp; /* hash list port iterator */
6783 struct sctp_endpoint *newep = newsp->ep;
6784 struct sk_buff *skb, *tmp;
6785 struct sctp_ulpevent *event;
6786 struct sctp_bind_hashbucket *head;
6788 /* Migrate socket buffer sizes and all the socket level options to the
6791 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6792 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6793 /* Brute force copy old sctp opt. */
6794 sctp_copy_descendant(newsk, oldsk);
6796 /* Restore the ep value that was overwritten with the above structure
6802 /* Hook this new socket in to the bind_hash list. */
6803 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->inet_num)];
6804 sctp_local_bh_disable();
6805 sctp_spin_lock(&head->lock);
6806 pp = sctp_sk(oldsk)->bind_hash;
6807 sk_add_bind_node(newsk, &pp->owner);
6808 sctp_sk(newsk)->bind_hash = pp;
6809 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6810 sctp_spin_unlock(&head->lock);
6811 sctp_local_bh_enable();
6813 /* Copy the bind_addr list from the original endpoint to the new
6814 * endpoint so that we can handle restarts properly
6816 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6817 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6819 /* Move any messages in the old socket's receive queue that are for the
6820 * peeled off association to the new socket's receive queue.
6822 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6823 event = sctp_skb2event(skb);
6824 if (event->asoc == assoc) {
6825 __skb_unlink(skb, &oldsk->sk_receive_queue);
6826 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6827 sctp_skb_set_owner_r_frag(skb, newsk);
6831 /* Clean up any messages pending delivery due to partial
6832 * delivery. Three cases:
6833 * 1) No partial deliver; no work.
6834 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6835 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6837 skb_queue_head_init(&newsp->pd_lobby);
6838 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6840 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6841 struct sk_buff_head *queue;
6843 /* Decide which queue to move pd_lobby skbs to. */
6844 if (assoc->ulpq.pd_mode) {
6845 queue = &newsp->pd_lobby;
6847 queue = &newsk->sk_receive_queue;
6849 /* Walk through the pd_lobby, looking for skbs that
6850 * need moved to the new socket.
6852 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6853 event = sctp_skb2event(skb);
6854 if (event->asoc == assoc) {
6855 __skb_unlink(skb, &oldsp->pd_lobby);
6856 __skb_queue_tail(queue, skb);
6857 sctp_skb_set_owner_r_frag(skb, newsk);
6861 /* Clear up any skbs waiting for the partial
6862 * delivery to finish.
6864 if (assoc->ulpq.pd_mode)
6865 sctp_clear_pd(oldsk, NULL);
6869 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6870 sctp_skb_set_owner_r_frag(skb, newsk);
6872 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6873 sctp_skb_set_owner_r_frag(skb, newsk);
6875 /* Set the type of socket to indicate that it is peeled off from the
6876 * original UDP-style socket or created with the accept() call on a
6877 * TCP-style socket..
6881 /* Mark the new socket "in-use" by the user so that any packets
6882 * that may arrive on the association after we've moved it are
6883 * queued to the backlog. This prevents a potential race between
6884 * backlog processing on the old socket and new-packet processing
6885 * on the new socket.
6887 * The caller has just allocated newsk so we can guarantee that other
6888 * paths won't try to lock it and then oldsk.
6890 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6891 sctp_assoc_migrate(assoc, newsk);
6893 /* If the association on the newsk is already closed before accept()
6894 * is called, set RCV_SHUTDOWN flag.
6896 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6897 newsk->sk_shutdown |= RCV_SHUTDOWN;
6899 newsk->sk_state = SCTP_SS_ESTABLISHED;
6900 sctp_release_sock(newsk);
6904 /* This proto struct describes the ULP interface for SCTP. */
6905 struct proto sctp_prot = {
6907 .owner = THIS_MODULE,
6908 .close = sctp_close,
6909 .connect = sctp_connect,
6910 .disconnect = sctp_disconnect,
6911 .accept = sctp_accept,
6912 .ioctl = sctp_ioctl,
6913 .init = sctp_init_sock,
6914 .destroy = sctp_destroy_sock,
6915 .shutdown = sctp_shutdown,
6916 .setsockopt = sctp_setsockopt,
6917 .getsockopt = sctp_getsockopt,
6918 .sendmsg = sctp_sendmsg,
6919 .recvmsg = sctp_recvmsg,
6921 .backlog_rcv = sctp_backlog_rcv,
6923 .unhash = sctp_unhash,
6924 .get_port = sctp_get_port,
6925 .obj_size = sizeof(struct sctp_sock),
6926 .sysctl_mem = sysctl_sctp_mem,
6927 .sysctl_rmem = sysctl_sctp_rmem,
6928 .sysctl_wmem = sysctl_sctp_wmem,
6929 .memory_pressure = &sctp_memory_pressure,
6930 .enter_memory_pressure = sctp_enter_memory_pressure,
6931 .memory_allocated = &sctp_memory_allocated,
6932 .sockets_allocated = &sctp_sockets_allocated,
6935 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6937 #include <net/transp_v6.h>
6938 static void sctp_v6_destroy_sock(struct sock *sk)
6940 sctp_destroy_sock(sk);
6941 inet6_destroy_sock(sk);
6944 struct proto sctpv6_prot = {
6946 .owner = THIS_MODULE,
6947 .close = sctp_close,
6948 .connect = sctp_connect,
6949 .disconnect = sctp_disconnect,
6950 .accept = sctp_accept,
6951 .ioctl = sctp_ioctl,
6952 .init = sctp_init_sock,
6953 .destroy = sctp_v6_destroy_sock,
6954 .shutdown = sctp_shutdown,
6955 .setsockopt = sctp_setsockopt,
6956 .getsockopt = sctp_getsockopt,
6957 .sendmsg = sctp_sendmsg,
6958 .recvmsg = sctp_recvmsg,
6960 .backlog_rcv = sctp_backlog_rcv,
6962 .unhash = sctp_unhash,
6963 .get_port = sctp_get_port,
6964 .obj_size = sizeof(struct sctp6_sock),
6965 .sysctl_mem = sysctl_sctp_mem,
6966 .sysctl_rmem = sysctl_sctp_rmem,
6967 .sysctl_wmem = sysctl_sctp_wmem,
6968 .memory_pressure = &sctp_memory_pressure,
6969 .enter_memory_pressure = sctp_enter_memory_pressure,
6970 .memory_allocated = &sctp_memory_allocated,
6971 .sockets_allocated = &sctp_sockets_allocated,
6973 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */