1 /* SCTP kernel implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
9 * This file is part of the SCTP kernel implementation
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
18 * This SCTP implementation is free software;
19 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
24 * This SCTP implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, write to
32 * the Free Software Foundation, 59 Temple Place - Suite 330,
33 * Boston, MA 02111-1307, USA.
35 * Please send any bug reports or fixes you make to the
37 * lksctp developers <lksctp-developers@lists.sourceforge.net>
39 * Or submit a bug report through the following website:
40 * http://www.sf.net/projects/lksctp
42 * Written or modified by:
43 * La Monte H.P. Yarroll <piggy@acm.org>
44 * Narasimha Budihal <narsi@refcode.org>
45 * Karl Knutson <karl@athena.chicago.il.us>
46 * Jon Grimm <jgrimm@us.ibm.com>
47 * Xingang Guo <xingang.guo@intel.com>
48 * Daisy Chang <daisyc@us.ibm.com>
49 * Sridhar Samudrala <samudrala@us.ibm.com>
50 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
51 * Ardelle Fan <ardelle.fan@intel.com>
52 * Ryan Layer <rmlayer@us.ibm.com>
53 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
54 * Kevin Gao <kevin.gao@intel.com>
56 * Any bugs reported given to us we will try to fix... any fixes shared will
57 * be incorporated into the next SCTP release.
60 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
62 #include <linux/types.h>
63 #include <linux/kernel.h>
64 #include <linux/wait.h>
65 #include <linux/time.h>
67 #include <linux/capability.h>
68 #include <linux/fcntl.h>
69 #include <linux/poll.h>
70 #include <linux/init.h>
71 #include <linux/crypto.h>
72 #include <linux/slab.h>
73 #include <linux/file.h>
74 #include <linux/compat.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 *, long *timeo_p,
98 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
99 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
100 static int sctp_wait_for_accept(struct sock *sk, long timeo);
101 static void sctp_wait_for_close(struct sock *sk, long timeo);
102 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
103 union sctp_addr *addr, int len);
104 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
105 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
106 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
107 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
108 static int sctp_send_asconf(struct sctp_association *asoc,
109 struct sctp_chunk *chunk);
110 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
111 static int sctp_autobind(struct sock *sk);
112 static void sctp_sock_migrate(struct sock *, struct sock *,
113 struct sctp_association *, sctp_socket_type_t);
115 extern struct kmem_cache *sctp_bucket_cachep;
116 extern long sysctl_sctp_mem[3];
117 extern int sysctl_sctp_rmem[3];
118 extern int sysctl_sctp_wmem[3];
120 static int sctp_memory_pressure;
121 static atomic_long_t sctp_memory_allocated;
122 struct percpu_counter sctp_sockets_allocated;
124 static void sctp_enter_memory_pressure(struct sock *sk)
126 sctp_memory_pressure = 1;
130 /* Get the sndbuf space available at the time on the association. */
131 static inline int sctp_wspace(struct sctp_association *asoc)
135 if (asoc->ep->sndbuf_policy)
136 amt = asoc->sndbuf_used;
138 amt = sk_wmem_alloc_get(asoc->base.sk);
140 if (amt >= asoc->base.sk->sk_sndbuf) {
141 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
144 amt = sk_stream_wspace(asoc->base.sk);
149 amt = asoc->base.sk->sk_sndbuf - amt;
154 /* Increment the used sndbuf space count of the corresponding association by
155 * the size of the outgoing data chunk.
156 * Also, set the skb destructor for sndbuf accounting later.
158 * Since it is always 1-1 between chunk and skb, and also a new skb is always
159 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
160 * destructor in the data chunk skb for the purpose of the sndbuf space
163 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
165 struct sctp_association *asoc = chunk->asoc;
166 struct sock *sk = asoc->base.sk;
168 /* The sndbuf space is tracked per association. */
169 sctp_association_hold(asoc);
171 skb_set_owner_w(chunk->skb, sk);
173 chunk->skb->destructor = sctp_wfree;
174 /* Save the chunk pointer in skb for sctp_wfree to use later. */
175 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
177 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
178 sizeof(struct sk_buff) +
179 sizeof(struct sctp_chunk);
181 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
182 sk->sk_wmem_queued += chunk->skb->truesize;
183 sk_mem_charge(sk, chunk->skb->truesize);
186 /* Verify that this is a valid address. */
187 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
192 /* Verify basic sockaddr. */
193 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
197 /* Is this a valid SCTP address? */
198 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
201 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
207 /* Look up the association by its id. If this is not a UDP-style
208 * socket, the ID field is always ignored.
210 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
212 struct sctp_association *asoc = NULL;
214 /* If this is not a UDP-style socket, assoc id should be ignored. */
215 if (!sctp_style(sk, UDP)) {
216 /* Return NULL if the socket state is not ESTABLISHED. It
217 * could be a TCP-style listening socket or a socket which
218 * hasn't yet called connect() to establish an association.
220 if (!sctp_sstate(sk, ESTABLISHED))
223 /* Get the first and the only association from the list. */
224 if (!list_empty(&sctp_sk(sk)->ep->asocs))
225 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
226 struct sctp_association, asocs);
230 /* Otherwise this is a UDP-style socket. */
231 if (!id || (id == (sctp_assoc_t)-1))
234 spin_lock_bh(&sctp_assocs_id_lock);
235 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
236 spin_unlock_bh(&sctp_assocs_id_lock);
238 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
244 /* Look up the transport from an address and an assoc id. If both address and
245 * id are specified, the associations matching the address and the id should be
248 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
249 struct sockaddr_storage *addr,
252 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
253 struct sctp_transport *transport;
254 union sctp_addr *laddr = (union sctp_addr *)addr;
256 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
263 id_asoc = sctp_id2assoc(sk, id);
264 if (id_asoc && (id_asoc != addr_asoc))
267 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
268 (union sctp_addr *)addr);
273 /* API 3.1.2 bind() - UDP Style Syntax
274 * The syntax of bind() is,
276 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
278 * sd - the socket descriptor returned by socket().
279 * addr - the address structure (struct sockaddr_in or struct
280 * sockaddr_in6 [RFC 2553]),
281 * addr_len - the size of the address structure.
283 SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
289 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
292 /* Disallow binding twice. */
293 if (!sctp_sk(sk)->ep->base.bind_addr.port)
294 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
299 sctp_release_sock(sk);
304 static long sctp_get_port_local(struct sock *, union sctp_addr *);
306 /* Verify this is a valid sockaddr. */
307 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
308 union sctp_addr *addr, int len)
312 /* Check minimum size. */
313 if (len < sizeof (struct sockaddr))
316 /* V4 mapped address are really of AF_INET family */
317 if (addr->sa.sa_family == AF_INET6 &&
318 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
319 if (!opt->pf->af_supported(AF_INET, opt))
322 /* Does this PF support this AF? */
323 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
327 /* If we get this far, af is valid. */
328 af = sctp_get_af_specific(addr->sa.sa_family);
330 if (len < af->sockaddr_len)
336 /* Bind a local address either to an endpoint or to an association. */
337 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
339 struct net *net = sock_net(sk);
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 &&
384 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
387 /* See if the address matches any of the addresses we may have
388 * already bound before checking against other endpoints.
390 if (sctp_bind_addr_match(bp, addr, sp))
393 /* Make sure we are allowed to bind here.
394 * The function sctp_get_port_local() does duplicate address
397 addr->v4.sin_port = htons(snum);
398 if ((ret = sctp_get_port_local(sk, addr))) {
402 /* Refresh ephemeral port. */
404 bp->port = inet_sk(sk)->inet_num;
406 /* Add the address to the bind address list.
407 * Use GFP_ATOMIC since BHs will be disabled.
409 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
411 /* Copy back into socket for getsockname() use. */
413 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
414 af->to_sk_saddr(addr, sk);
420 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
422 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
423 * at any one time. If a sender, after sending an ASCONF chunk, decides
424 * it needs to transfer another ASCONF Chunk, it MUST wait until the
425 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
426 * subsequent ASCONF. Note this restriction binds each side, so at any
427 * time two ASCONF may be in-transit on any given association (one sent
428 * from each endpoint).
430 static int sctp_send_asconf(struct sctp_association *asoc,
431 struct sctp_chunk *chunk)
433 struct net *net = sock_net(asoc->base.sk);
436 /* If there is an outstanding ASCONF chunk, queue it for later
439 if (asoc->addip_last_asconf) {
440 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
444 /* Hold the chunk until an ASCONF_ACK is received. */
445 sctp_chunk_hold(chunk);
446 retval = sctp_primitive_ASCONF(net, asoc, chunk);
448 sctp_chunk_free(chunk);
450 asoc->addip_last_asconf = chunk;
456 /* Add a list of addresses as bind addresses to local endpoint or
459 * Basically run through each address specified in the addrs/addrcnt
460 * array/length pair, determine if it is IPv6 or IPv4 and call
461 * sctp_do_bind() on it.
463 * If any of them fails, then the operation will be reversed and the
464 * ones that were added will be removed.
466 * Only sctp_setsockopt_bindx() is supposed to call this function.
468 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
473 struct sockaddr *sa_addr;
476 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
480 for (cnt = 0; cnt < addrcnt; cnt++) {
481 /* The list may contain either IPv4 or IPv6 address;
482 * determine the address length for walking thru the list.
485 af = sctp_get_af_specific(sa_addr->sa_family);
491 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
494 addr_buf += af->sockaddr_len;
498 /* Failed. Cleanup the ones that have been added */
500 sctp_bindx_rem(sk, addrs, cnt);
508 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
509 * associations that are part of the endpoint indicating that a list of local
510 * addresses are added to the endpoint.
512 * If any of the addresses is already in the bind address list of the
513 * association, we do not send the chunk for that association. But it will not
514 * affect other associations.
516 * Only sctp_setsockopt_bindx() is supposed to call this function.
518 static int sctp_send_asconf_add_ip(struct sock *sk,
519 struct sockaddr *addrs,
522 struct net *net = sock_net(sk);
523 struct sctp_sock *sp;
524 struct sctp_endpoint *ep;
525 struct sctp_association *asoc;
526 struct sctp_bind_addr *bp;
527 struct sctp_chunk *chunk;
528 struct sctp_sockaddr_entry *laddr;
529 union sctp_addr *addr;
530 union sctp_addr saveaddr;
537 if (!net->sctp.addip_enable)
543 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
544 __func__, sk, addrs, addrcnt);
546 list_for_each_entry(asoc, &ep->asocs, asocs) {
548 if (!asoc->peer.asconf_capable)
551 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
554 if (!sctp_state(asoc, ESTABLISHED))
557 /* Check if any address in the packed array of addresses is
558 * in the bind address list of the association. If so,
559 * do not send the asconf chunk to its peer, but continue with
560 * other associations.
563 for (i = 0; i < addrcnt; i++) {
565 af = sctp_get_af_specific(addr->v4.sin_family);
571 if (sctp_assoc_lookup_laddr(asoc, addr))
574 addr_buf += af->sockaddr_len;
579 /* Use the first valid address in bind addr list of
580 * association as Address Parameter of ASCONF CHUNK.
582 bp = &asoc->base.bind_addr;
583 p = bp->address_list.next;
584 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
585 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
586 addrcnt, SCTP_PARAM_ADD_IP);
592 /* Add the new addresses to the bind address list with
593 * use_as_src set to 0.
596 for (i = 0; i < addrcnt; i++) {
598 af = sctp_get_af_specific(addr->v4.sin_family);
599 memcpy(&saveaddr, addr, af->sockaddr_len);
600 retval = sctp_add_bind_addr(bp, &saveaddr,
601 SCTP_ADDR_NEW, GFP_ATOMIC);
602 addr_buf += af->sockaddr_len;
604 if (asoc->src_out_of_asoc_ok) {
605 struct sctp_transport *trans;
607 list_for_each_entry(trans,
608 &asoc->peer.transport_addr_list, transports) {
609 /* Clear the source and route cache */
610 dst_release(trans->dst);
611 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
612 2*asoc->pathmtu, 4380));
613 trans->ssthresh = asoc->peer.i.a_rwnd;
614 trans->rto = asoc->rto_initial;
615 sctp_max_rto(asoc, trans);
616 trans->rtt = trans->srtt = trans->rttvar = 0;
617 sctp_transport_route(trans, NULL,
618 sctp_sk(asoc->base.sk));
621 retval = sctp_send_asconf(asoc, chunk);
628 /* Remove a list of addresses from bind addresses list. Do not remove the
631 * Basically run through each address specified in the addrs/addrcnt
632 * array/length pair, determine if it is IPv6 or IPv4 and call
633 * sctp_del_bind() on it.
635 * If any of them fails, then the operation will be reversed and the
636 * ones that were removed will be added back.
638 * At least one address has to be left; if only one address is
639 * available, the operation will return -EBUSY.
641 * Only sctp_setsockopt_bindx() is supposed to call this function.
643 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
645 struct sctp_sock *sp = sctp_sk(sk);
646 struct sctp_endpoint *ep = sp->ep;
648 struct sctp_bind_addr *bp = &ep->base.bind_addr;
651 union sctp_addr *sa_addr;
654 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
658 for (cnt = 0; cnt < addrcnt; cnt++) {
659 /* If the bind address list is empty or if there is only one
660 * bind address, there is nothing more to be removed (we need
661 * at least one address here).
663 if (list_empty(&bp->address_list) ||
664 (sctp_list_single_entry(&bp->address_list))) {
670 af = sctp_get_af_specific(sa_addr->sa.sa_family);
676 if (!af->addr_valid(sa_addr, sp, NULL)) {
677 retval = -EADDRNOTAVAIL;
681 if (sa_addr->v4.sin_port &&
682 sa_addr->v4.sin_port != htons(bp->port)) {
687 if (!sa_addr->v4.sin_port)
688 sa_addr->v4.sin_port = htons(bp->port);
690 /* FIXME - There is probably a need to check if sk->sk_saddr and
691 * sk->sk_rcv_addr are currently set to one of the addresses to
692 * be removed. This is something which needs to be looked into
693 * when we are fixing the outstanding issues with multi-homing
694 * socket routing and failover schemes. Refer to comments in
695 * sctp_do_bind(). -daisy
697 retval = sctp_del_bind_addr(bp, sa_addr);
699 addr_buf += af->sockaddr_len;
702 /* Failed. Add the ones that has been removed back */
704 sctp_bindx_add(sk, addrs, cnt);
712 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
713 * the associations that are part of the endpoint indicating that a list of
714 * local addresses are removed from the endpoint.
716 * If any of the addresses is already in the bind address list of the
717 * association, we do not send the chunk for that association. But it will not
718 * affect other associations.
720 * Only sctp_setsockopt_bindx() is supposed to call this function.
722 static int sctp_send_asconf_del_ip(struct sock *sk,
723 struct sockaddr *addrs,
726 struct net *net = sock_net(sk);
727 struct sctp_sock *sp;
728 struct sctp_endpoint *ep;
729 struct sctp_association *asoc;
730 struct sctp_transport *transport;
731 struct sctp_bind_addr *bp;
732 struct sctp_chunk *chunk;
733 union sctp_addr *laddr;
736 struct sctp_sockaddr_entry *saddr;
742 if (!net->sctp.addip_enable)
748 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
749 __func__, sk, addrs, addrcnt);
751 list_for_each_entry(asoc, &ep->asocs, asocs) {
753 if (!asoc->peer.asconf_capable)
756 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
759 if (!sctp_state(asoc, ESTABLISHED))
762 /* Check if any address in the packed array of addresses is
763 * not present in the bind address list of the association.
764 * If so, do not send the asconf chunk to its peer, but
765 * continue with other associations.
768 for (i = 0; i < addrcnt; i++) {
770 af = sctp_get_af_specific(laddr->v4.sin_family);
776 if (!sctp_assoc_lookup_laddr(asoc, laddr))
779 addr_buf += af->sockaddr_len;
784 /* Find one address in the association's bind address list
785 * that is not in the packed array of addresses. This is to
786 * make sure that we do not delete all the addresses in the
789 bp = &asoc->base.bind_addr;
790 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
792 if ((laddr == NULL) && (addrcnt == 1)) {
793 if (asoc->asconf_addr_del_pending)
795 asoc->asconf_addr_del_pending =
796 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
797 if (asoc->asconf_addr_del_pending == NULL) {
801 asoc->asconf_addr_del_pending->sa.sa_family =
803 asoc->asconf_addr_del_pending->v4.sin_port =
805 if (addrs->sa_family == AF_INET) {
806 struct sockaddr_in *sin;
808 sin = (struct sockaddr_in *)addrs;
809 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
810 } else if (addrs->sa_family == AF_INET6) {
811 struct sockaddr_in6 *sin6;
813 sin6 = (struct sockaddr_in6 *)addrs;
814 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
816 SCTP_DEBUG_PRINTK_IPADDR("send_asconf_del_ip: keep the last address asoc: %p ",
817 " at %p\n", asoc, asoc->asconf_addr_del_pending,
818 asoc->asconf_addr_del_pending);
819 asoc->src_out_of_asoc_ok = 1;
827 /* We do not need RCU protection throughout this loop
828 * because this is done under a socket lock from the
831 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
839 /* Reset use_as_src flag for the addresses in the bind address
840 * list that are to be deleted.
843 for (i = 0; i < addrcnt; i++) {
845 af = sctp_get_af_specific(laddr->v4.sin_family);
846 list_for_each_entry(saddr, &bp->address_list, list) {
847 if (sctp_cmp_addr_exact(&saddr->a, laddr))
848 saddr->state = SCTP_ADDR_DEL;
850 addr_buf += af->sockaddr_len;
853 /* Update the route and saddr entries for all the transports
854 * as some of the addresses in the bind address list are
855 * about to be deleted and cannot be used as source addresses.
857 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
859 dst_release(transport->dst);
860 sctp_transport_route(transport, NULL,
861 sctp_sk(asoc->base.sk));
865 /* We don't need to transmit ASCONF */
867 retval = sctp_send_asconf(asoc, chunk);
873 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
874 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
876 struct sock *sk = sctp_opt2sk(sp);
877 union sctp_addr *addr;
880 /* It is safe to write port space in caller. */
882 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
883 af = sctp_get_af_specific(addr->sa.sa_family);
886 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
889 if (addrw->state == SCTP_ADDR_NEW)
890 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
892 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
895 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
898 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
901 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
902 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
905 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
906 * Section 3.1.2 for this usage.
908 * addrs is a pointer to an array of one or more socket addresses. Each
909 * address is contained in its appropriate structure (i.e. struct
910 * sockaddr_in or struct sockaddr_in6) the family of the address type
911 * must be used to distinguish the address length (note that this
912 * representation is termed a "packed array" of addresses). The caller
913 * specifies the number of addresses in the array with addrcnt.
915 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
916 * -1, and sets errno to the appropriate error code.
918 * For SCTP, the port given in each socket address must be the same, or
919 * sctp_bindx() will fail, setting errno to EINVAL.
921 * The flags parameter is formed from the bitwise OR of zero or more of
922 * the following currently defined flags:
924 * SCTP_BINDX_ADD_ADDR
926 * SCTP_BINDX_REM_ADDR
928 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
929 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
930 * addresses from the association. The two flags are mutually exclusive;
931 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
932 * not remove all addresses from an association; sctp_bindx() will
933 * reject such an attempt with EINVAL.
935 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
936 * additional addresses with an endpoint after calling bind(). Or use
937 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
938 * socket is associated with so that no new association accepted will be
939 * associated with those addresses. If the endpoint supports dynamic
940 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
941 * endpoint to send the appropriate message to the peer to change the
942 * peers address lists.
944 * Adding and removing addresses from a connected association is
945 * optional functionality. Implementations that do not support this
946 * functionality should return EOPNOTSUPP.
948 * Basically do nothing but copying the addresses from user to kernel
949 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
950 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
953 * We don't use copy_from_user() for optimization: we first do the
954 * sanity checks (buffer size -fast- and access check-healthy
955 * pointer); if all of those succeed, then we can alloc the memory
956 * (expensive operation) needed to copy the data to kernel. Then we do
957 * the copying without checking the user space area
958 * (__copy_from_user()).
960 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
963 * sk The sk of the socket
964 * addrs The pointer to the addresses in user land
965 * addrssize Size of the addrs buffer
966 * op Operation to perform (add or remove, see the flags of
969 * Returns 0 if ok, <0 errno code on error.
971 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
972 struct sockaddr __user *addrs,
973 int addrs_size, int op)
975 struct sockaddr *kaddrs;
979 struct sockaddr *sa_addr;
983 SCTP_DEBUG_PRINTK("sctp_setsockopt_bindx: sk %p addrs %p"
984 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
986 if (unlikely(addrs_size <= 0))
989 /* Check the user passed a healthy pointer. */
990 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
993 /* Alloc space for the address array in kernel memory. */
994 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
995 if (unlikely(!kaddrs))
998 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1003 /* Walk through the addrs buffer and count the number of addresses. */
1005 while (walk_size < addrs_size) {
1006 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1012 af = sctp_get_af_specific(sa_addr->sa_family);
1014 /* If the address family is not supported or if this address
1015 * causes the address buffer to overflow return EINVAL.
1017 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1022 addr_buf += af->sockaddr_len;
1023 walk_size += af->sockaddr_len;
1028 case SCTP_BINDX_ADD_ADDR:
1029 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1032 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1035 case SCTP_BINDX_REM_ADDR:
1036 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1039 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1053 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1055 * Common routine for handling connect() and sctp_connectx().
1056 * Connect will come in with just a single address.
1058 static int __sctp_connect(struct sock* sk,
1059 struct sockaddr *kaddrs,
1061 sctp_assoc_t *assoc_id)
1063 struct net *net = sock_net(sk);
1064 struct sctp_sock *sp;
1065 struct sctp_endpoint *ep;
1066 struct sctp_association *asoc = NULL;
1067 struct sctp_association *asoc2;
1068 struct sctp_transport *transport;
1076 union sctp_addr *sa_addr = NULL;
1078 unsigned short port;
1079 unsigned int f_flags = 0;
1084 /* connect() cannot be done on a socket that is already in ESTABLISHED
1085 * state - UDP-style peeled off socket or a TCP-style socket that
1086 * is already connected.
1087 * It cannot be done even on a TCP-style listening socket.
1089 if (sctp_sstate(sk, ESTABLISHED) ||
1090 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1095 /* Walk through the addrs buffer and count the number of addresses. */
1097 while (walk_size < addrs_size) {
1098 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1104 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1106 /* If the address family is not supported or if this address
1107 * causes the address buffer to overflow return EINVAL.
1109 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1114 port = ntohs(sa_addr->v4.sin_port);
1116 /* Save current address so we can work with it */
1117 memcpy(&to, sa_addr, af->sockaddr_len);
1119 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1123 /* Make sure the destination port is correctly set
1126 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1131 /* Check if there already is a matching association on the
1132 * endpoint (other than the one created here).
1134 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1135 if (asoc2 && asoc2 != asoc) {
1136 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1143 /* If we could not find a matching association on the endpoint,
1144 * make sure that there is no peeled-off association matching
1145 * the peer address even on another socket.
1147 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1148 err = -EADDRNOTAVAIL;
1153 /* If a bind() or sctp_bindx() is not called prior to
1154 * an sctp_connectx() call, the system picks an
1155 * ephemeral port and will choose an address set
1156 * equivalent to binding with a wildcard address.
1158 if (!ep->base.bind_addr.port) {
1159 if (sctp_autobind(sk)) {
1165 * If an unprivileged user inherits a 1-many
1166 * style socket with open associations on a
1167 * privileged port, it MAY be permitted to
1168 * accept new associations, but it SHOULD NOT
1169 * be permitted to open new associations.
1171 if (ep->base.bind_addr.port < PROT_SOCK &&
1172 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1178 scope = sctp_scope(&to);
1179 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1185 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1193 /* Prime the peer's transport structures. */
1194 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1202 addr_buf += af->sockaddr_len;
1203 walk_size += af->sockaddr_len;
1206 /* In case the user of sctp_connectx() wants an association
1207 * id back, assign one now.
1210 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1215 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1220 /* Initialize sk's dport and daddr for getpeername() */
1221 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1222 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1223 af->to_sk_daddr(sa_addr, sk);
1226 /* in-kernel sockets don't generally have a file allocated to them
1227 * if all they do is call sock_create_kern().
1229 if (sk->sk_socket->file)
1230 f_flags = sk->sk_socket->file->f_flags;
1232 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1234 err = sctp_wait_for_connect(asoc, &timeo);
1235 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1236 *assoc_id = asoc->assoc_id;
1238 /* Don't free association on exit. */
1243 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1244 " kaddrs: %p err: %d\n",
1247 /* sctp_primitive_ASSOCIATE may have added this association
1248 * To the hash table, try to unhash it, just in case, its a noop
1249 * if it wasn't hashed so we're safe
1251 sctp_unhash_established(asoc);
1252 sctp_association_free(asoc);
1257 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1260 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1261 * sctp_assoc_t *asoc);
1263 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1264 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1265 * or IPv6 addresses.
1267 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1268 * Section 3.1.2 for this usage.
1270 * addrs is a pointer to an array of one or more socket addresses. Each
1271 * address is contained in its appropriate structure (i.e. struct
1272 * sockaddr_in or struct sockaddr_in6) the family of the address type
1273 * must be used to distengish the address length (note that this
1274 * representation is termed a "packed array" of addresses). The caller
1275 * specifies the number of addresses in the array with addrcnt.
1277 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1278 * the association id of the new association. On failure, sctp_connectx()
1279 * returns -1, and sets errno to the appropriate error code. The assoc_id
1280 * is not touched by the kernel.
1282 * For SCTP, the port given in each socket address must be the same, or
1283 * sctp_connectx() will fail, setting errno to EINVAL.
1285 * An application can use sctp_connectx to initiate an association with
1286 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1287 * allows a caller to specify multiple addresses at which a peer can be
1288 * reached. The way the SCTP stack uses the list of addresses to set up
1289 * the association is implementation dependent. This function only
1290 * specifies that the stack will try to make use of all the addresses in
1291 * the list when needed.
1293 * Note that the list of addresses passed in is only used for setting up
1294 * the association. It does not necessarily equal the set of addresses
1295 * the peer uses for the resulting association. If the caller wants to
1296 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1297 * retrieve them after the association has been set up.
1299 * Basically do nothing but copying the addresses from user to kernel
1300 * land and invoking either sctp_connectx(). This is used for tunneling
1301 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1303 * We don't use copy_from_user() for optimization: we first do the
1304 * sanity checks (buffer size -fast- and access check-healthy
1305 * pointer); if all of those succeed, then we can alloc the memory
1306 * (expensive operation) needed to copy the data to kernel. Then we do
1307 * the copying without checking the user space area
1308 * (__copy_from_user()).
1310 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1313 * sk The sk of the socket
1314 * addrs The pointer to the addresses in user land
1315 * addrssize Size of the addrs buffer
1317 * Returns >=0 if ok, <0 errno code on error.
1319 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1320 struct sockaddr __user *addrs,
1322 sctp_assoc_t *assoc_id)
1325 struct sockaddr *kaddrs;
1327 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1328 __func__, sk, addrs, addrs_size);
1330 if (unlikely(addrs_size <= 0))
1333 /* Check the user passed a healthy pointer. */
1334 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1337 /* Alloc space for the address array in kernel memory. */
1338 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1339 if (unlikely(!kaddrs))
1342 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1345 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1354 * This is an older interface. It's kept for backward compatibility
1355 * to the option that doesn't provide association id.
1357 SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1358 struct sockaddr __user *addrs,
1361 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1365 * New interface for the API. The since the API is done with a socket
1366 * option, to make it simple we feed back the association id is as a return
1367 * indication to the call. Error is always negative and association id is
1370 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1371 struct sockaddr __user *addrs,
1374 sctp_assoc_t assoc_id = 0;
1377 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1386 * New (hopefully final) interface for the API.
1387 * We use the sctp_getaddrs_old structure so that use-space library
1388 * can avoid any unnecessary allocations. The only different part
1389 * is that we store the actual length of the address buffer into the
1390 * addrs_num structure member. That way we can re-use the existing
1393 #ifdef CONFIG_COMPAT
1394 struct compat_sctp_getaddrs_old {
1395 sctp_assoc_t assoc_id;
1397 compat_uptr_t addrs; /* struct sockaddr * */
1401 SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1402 char __user *optval,
1405 struct sctp_getaddrs_old param;
1406 sctp_assoc_t assoc_id = 0;
1409 #ifdef CONFIG_COMPAT
1410 if (is_compat_task()) {
1411 struct compat_sctp_getaddrs_old param32;
1413 if (len < sizeof(param32))
1415 if (copy_from_user(¶m32, optval, sizeof(param32)))
1418 param.assoc_id = param32.assoc_id;
1419 param.addr_num = param32.addr_num;
1420 param.addrs = compat_ptr(param32.addrs);
1424 if (len < sizeof(param))
1426 if (copy_from_user(¶m, optval, sizeof(param)))
1430 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1431 param.addrs, param.addr_num,
1433 if (err == 0 || err == -EINPROGRESS) {
1434 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1436 if (put_user(sizeof(assoc_id), optlen))
1443 /* API 3.1.4 close() - UDP Style Syntax
1444 * Applications use close() to perform graceful shutdown (as described in
1445 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1446 * by a UDP-style socket.
1450 * ret = close(int sd);
1452 * sd - the socket descriptor of the associations to be closed.
1454 * To gracefully shutdown a specific association represented by the
1455 * UDP-style socket, an application should use the sendmsg() call,
1456 * passing no user data, but including the appropriate flag in the
1457 * ancillary data (see Section xxxx).
1459 * If sd in the close() call is a branched-off socket representing only
1460 * one association, the shutdown is performed on that association only.
1462 * 4.1.6 close() - TCP Style Syntax
1464 * Applications use close() to gracefully close down an association.
1468 * int close(int sd);
1470 * sd - the socket descriptor of the association to be closed.
1472 * After an application calls close() on a socket descriptor, no further
1473 * socket operations will succeed on that descriptor.
1475 * API 7.1.4 SO_LINGER
1477 * An application using the TCP-style socket can use this option to
1478 * perform the SCTP ABORT primitive. The linger option structure is:
1481 * int l_onoff; // option on/off
1482 * int l_linger; // linger time
1485 * To enable the option, set l_onoff to 1. If the l_linger value is set
1486 * to 0, calling close() is the same as the ABORT primitive. If the
1487 * value is set to a negative value, the setsockopt() call will return
1488 * an error. If the value is set to a positive value linger_time, the
1489 * close() can be blocked for at most linger_time ms. If the graceful
1490 * shutdown phase does not finish during this period, close() will
1491 * return but the graceful shutdown phase continues in the system.
1493 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1495 struct net *net = sock_net(sk);
1496 struct sctp_endpoint *ep;
1497 struct sctp_association *asoc;
1498 struct list_head *pos, *temp;
1499 unsigned int data_was_unread;
1501 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1504 sk->sk_shutdown = SHUTDOWN_MASK;
1505 sk->sk_state = SCTP_SS_CLOSING;
1507 ep = sctp_sk(sk)->ep;
1509 /* Clean up any skbs sitting on the receive queue. */
1510 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1511 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1513 /* Walk all associations on an endpoint. */
1514 list_for_each_safe(pos, temp, &ep->asocs) {
1515 asoc = list_entry(pos, struct sctp_association, asocs);
1517 if (sctp_style(sk, TCP)) {
1518 /* A closed association can still be in the list if
1519 * it belongs to a TCP-style listening socket that is
1520 * not yet accepted. If so, free it. If not, send an
1521 * ABORT or SHUTDOWN based on the linger options.
1523 if (sctp_state(asoc, CLOSED)) {
1524 sctp_unhash_established(asoc);
1525 sctp_association_free(asoc);
1530 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1531 !skb_queue_empty(&asoc->ulpq.reasm) ||
1532 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1533 struct sctp_chunk *chunk;
1535 chunk = sctp_make_abort_user(asoc, NULL, 0);
1537 sctp_primitive_ABORT(net, asoc, chunk);
1539 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1542 /* On a TCP-style socket, block for at most linger_time if set. */
1543 if (sctp_style(sk, TCP) && timeout)
1544 sctp_wait_for_close(sk, timeout);
1546 /* This will run the backlog queue. */
1547 sctp_release_sock(sk);
1549 /* Supposedly, no process has access to the socket, but
1550 * the net layers still may.
1552 sctp_local_bh_disable();
1553 sctp_bh_lock_sock(sk);
1555 /* Hold the sock, since sk_common_release() will put sock_put()
1556 * and we have just a little more cleanup.
1559 sk_common_release(sk);
1561 sctp_bh_unlock_sock(sk);
1562 sctp_local_bh_enable();
1566 SCTP_DBG_OBJCNT_DEC(sock);
1569 /* Handle EPIPE error. */
1570 static int sctp_error(struct sock *sk, int flags, int err)
1573 err = sock_error(sk) ? : -EPIPE;
1574 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1575 send_sig(SIGPIPE, current, 0);
1579 /* API 3.1.3 sendmsg() - UDP Style Syntax
1581 * An application uses sendmsg() and recvmsg() calls to transmit data to
1582 * and receive data from its peer.
1584 * ssize_t sendmsg(int socket, const struct msghdr *message,
1587 * socket - the socket descriptor of the endpoint.
1588 * message - pointer to the msghdr structure which contains a single
1589 * user message and possibly some ancillary data.
1591 * See Section 5 for complete description of the data
1594 * flags - flags sent or received with the user message, see Section
1595 * 5 for complete description of the flags.
1597 * Note: This function could use a rewrite especially when explicit
1598 * connect support comes in.
1600 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1602 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1604 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1605 struct msghdr *msg, size_t msg_len)
1607 struct net *net = sock_net(sk);
1608 struct sctp_sock *sp;
1609 struct sctp_endpoint *ep;
1610 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1611 struct sctp_transport *transport, *chunk_tp;
1612 struct sctp_chunk *chunk;
1614 struct sockaddr *msg_name = NULL;
1615 struct sctp_sndrcvinfo default_sinfo;
1616 struct sctp_sndrcvinfo *sinfo;
1617 struct sctp_initmsg *sinit;
1618 sctp_assoc_t associd = 0;
1619 sctp_cmsgs_t cmsgs = { NULL };
1623 __u16 sinfo_flags = 0;
1624 struct sctp_datamsg *datamsg;
1625 int msg_flags = msg->msg_flags;
1627 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1634 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1636 /* We cannot send a message over a TCP-style listening socket. */
1637 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1642 /* Parse out the SCTP CMSGs. */
1643 err = sctp_msghdr_parse(msg, &cmsgs);
1646 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1650 /* Fetch the destination address for this packet. This
1651 * address only selects the association--it is not necessarily
1652 * the address we will send to.
1653 * For a peeled-off socket, msg_name is ignored.
1655 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1656 int msg_namelen = msg->msg_namelen;
1658 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1663 if (msg_namelen > sizeof(to))
1664 msg_namelen = sizeof(to);
1665 memcpy(&to, msg->msg_name, msg_namelen);
1666 msg_name = msg->msg_name;
1672 /* Did the user specify SNDRCVINFO? */
1674 sinfo_flags = sinfo->sinfo_flags;
1675 associd = sinfo->sinfo_assoc_id;
1678 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1679 msg_len, sinfo_flags);
1681 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1682 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1687 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1688 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1689 * If SCTP_ABORT is set, the message length could be non zero with
1690 * the msg_iov set to the user abort reason.
1692 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1693 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1698 /* If SCTP_ADDR_OVER is set, there must be an address
1699 * specified in msg_name.
1701 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1708 SCTP_DEBUG_PRINTK("About to look up association.\n");
1712 /* If a msg_name has been specified, assume this is to be used. */
1714 /* Look for a matching association on the endpoint. */
1715 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1717 /* If we could not find a matching association on the
1718 * endpoint, make sure that it is not a TCP-style
1719 * socket that already has an association or there is
1720 * no peeled-off association on another socket.
1722 if ((sctp_style(sk, TCP) &&
1723 sctp_sstate(sk, ESTABLISHED)) ||
1724 sctp_endpoint_is_peeled_off(ep, &to)) {
1725 err = -EADDRNOTAVAIL;
1730 asoc = sctp_id2assoc(sk, associd);
1738 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1740 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1741 * socket that has an association in CLOSED state. This can
1742 * happen when an accepted socket has an association that is
1745 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1750 if (sinfo_flags & SCTP_EOF) {
1751 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1753 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1757 if (sinfo_flags & SCTP_ABORT) {
1759 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1765 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1766 sctp_primitive_ABORT(net, asoc, chunk);
1772 /* Do we need to create the association? */
1774 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1776 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1781 /* Check for invalid stream against the stream counts,
1782 * either the default or the user specified stream counts.
1785 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1786 /* Check against the defaults. */
1787 if (sinfo->sinfo_stream >=
1788 sp->initmsg.sinit_num_ostreams) {
1793 /* Check against the requested. */
1794 if (sinfo->sinfo_stream >=
1795 sinit->sinit_num_ostreams) {
1803 * API 3.1.2 bind() - UDP Style Syntax
1804 * If a bind() or sctp_bindx() is not called prior to a
1805 * sendmsg() call that initiates a new association, the
1806 * system picks an ephemeral port and will choose an address
1807 * set equivalent to binding with a wildcard address.
1809 if (!ep->base.bind_addr.port) {
1810 if (sctp_autobind(sk)) {
1816 * If an unprivileged user inherits a one-to-many
1817 * style socket with open associations on a privileged
1818 * port, it MAY be permitted to accept new associations,
1819 * but it SHOULD NOT be permitted to open new
1822 if (ep->base.bind_addr.port < PROT_SOCK &&
1823 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1829 scope = sctp_scope(&to);
1830 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1836 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1842 /* If the SCTP_INIT ancillary data is specified, set all
1843 * the association init values accordingly.
1846 if (sinit->sinit_num_ostreams) {
1847 asoc->c.sinit_num_ostreams =
1848 sinit->sinit_num_ostreams;
1850 if (sinit->sinit_max_instreams) {
1851 asoc->c.sinit_max_instreams =
1852 sinit->sinit_max_instreams;
1854 if (sinit->sinit_max_attempts) {
1855 asoc->max_init_attempts
1856 = sinit->sinit_max_attempts;
1858 if (sinit->sinit_max_init_timeo) {
1859 asoc->max_init_timeo =
1860 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1864 /* Prime the peer's transport structures. */
1865 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1872 /* ASSERT: we have a valid association at this point. */
1873 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1876 /* If the user didn't specify SNDRCVINFO, make up one with
1879 memset(&default_sinfo, 0, sizeof(default_sinfo));
1880 default_sinfo.sinfo_stream = asoc->default_stream;
1881 default_sinfo.sinfo_flags = asoc->default_flags;
1882 default_sinfo.sinfo_ppid = asoc->default_ppid;
1883 default_sinfo.sinfo_context = asoc->default_context;
1884 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1885 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1886 sinfo = &default_sinfo;
1889 /* API 7.1.7, the sndbuf size per association bounds the
1890 * maximum size of data that can be sent in a single send call.
1892 if (msg_len > sk->sk_sndbuf) {
1897 if (asoc->pmtu_pending)
1898 sctp_assoc_pending_pmtu(sk, asoc);
1900 /* If fragmentation is disabled and the message length exceeds the
1901 * association fragmentation point, return EMSGSIZE. The I-D
1902 * does not specify what this error is, but this looks like
1905 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1910 /* Check for invalid stream. */
1911 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1916 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1917 if (!sctp_wspace(asoc)) {
1918 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1923 /* If an address is passed with the sendto/sendmsg call, it is used
1924 * to override the primary destination address in the TCP model, or
1925 * when SCTP_ADDR_OVER flag is set in the UDP model.
1927 if ((sctp_style(sk, TCP) && msg_name) ||
1928 (sinfo_flags & SCTP_ADDR_OVER)) {
1929 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1937 /* Auto-connect, if we aren't connected already. */
1938 if (sctp_state(asoc, CLOSED)) {
1939 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1942 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1945 /* Break the message into multiple chunks of maximum size. */
1946 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1947 if (IS_ERR(datamsg)) {
1948 err = PTR_ERR(datamsg);
1952 /* Now send the (possibly) fragmented message. */
1953 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1954 sctp_chunk_hold(chunk);
1956 /* Do accounting for the write space. */
1957 sctp_set_owner_w(chunk);
1959 chunk->transport = chunk_tp;
1962 /* Send it to the lower layers. Note: all chunks
1963 * must either fail or succeed. The lower layer
1964 * works that way today. Keep it that way or this
1967 err = sctp_primitive_SEND(net, asoc, datamsg);
1968 /* Did the lower layer accept the chunk? */
1970 sctp_datamsg_free(datamsg);
1972 sctp_datamsg_put(datamsg);
1974 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1981 /* If we are already past ASSOCIATE, the lower
1982 * layers are responsible for association cleanup.
1988 sctp_unhash_established(asoc);
1989 sctp_association_free(asoc);
1992 sctp_release_sock(sk);
1995 return sctp_error(sk, msg_flags, err);
2002 err = sock_error(sk);
2012 /* This is an extended version of skb_pull() that removes the data from the
2013 * start of a skb even when data is spread across the list of skb's in the
2014 * frag_list. len specifies the total amount of data that needs to be removed.
2015 * when 'len' bytes could be removed from the skb, it returns 0.
2016 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2017 * could not be removed.
2019 static int sctp_skb_pull(struct sk_buff *skb, int len)
2021 struct sk_buff *list;
2022 int skb_len = skb_headlen(skb);
2025 if (len <= skb_len) {
2026 __skb_pull(skb, len);
2030 __skb_pull(skb, skb_len);
2032 skb_walk_frags(skb, list) {
2033 rlen = sctp_skb_pull(list, len);
2034 skb->len -= (len-rlen);
2035 skb->data_len -= (len-rlen);
2046 /* API 3.1.3 recvmsg() - UDP Style Syntax
2048 * ssize_t recvmsg(int socket, struct msghdr *message,
2051 * socket - the socket descriptor of the endpoint.
2052 * message - pointer to the msghdr structure which contains a single
2053 * user message and possibly some ancillary data.
2055 * See Section 5 for complete description of the data
2058 * flags - flags sent or received with the user message, see Section
2059 * 5 for complete description of the flags.
2061 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2063 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2064 struct msghdr *msg, size_t len, int noblock,
2065 int flags, int *addr_len)
2067 struct sctp_ulpevent *event = NULL;
2068 struct sctp_sock *sp = sctp_sk(sk);
2069 struct sk_buff *skb;
2074 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
2075 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
2076 "len", len, "knoblauch", noblock,
2077 "flags", flags, "addr_len", addr_len);
2081 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2086 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2090 /* Get the total length of the skb including any skb's in the
2099 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2101 event = sctp_skb2event(skb);
2106 sock_recv_ts_and_drops(msg, sk, skb);
2107 if (sctp_ulpevent_is_notification(event)) {
2108 msg->msg_flags |= MSG_NOTIFICATION;
2109 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2111 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2114 /* Check if we allow SCTP_SNDRCVINFO. */
2115 if (sp->subscribe.sctp_data_io_event)
2116 sctp_ulpevent_read_sndrcvinfo(event, msg);
2118 /* FIXME: we should be calling IP/IPv6 layers. */
2119 if (sk->sk_protinfo.af_inet.cmsg_flags)
2120 ip_cmsg_recv(msg, skb);
2125 /* If skb's length exceeds the user's buffer, update the skb and
2126 * push it back to the receive_queue so that the next call to
2127 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2129 if (skb_len > copied) {
2130 msg->msg_flags &= ~MSG_EOR;
2131 if (flags & MSG_PEEK)
2133 sctp_skb_pull(skb, copied);
2134 skb_queue_head(&sk->sk_receive_queue, skb);
2136 /* When only partial message is copied to the user, increase
2137 * rwnd by that amount. If all the data in the skb is read,
2138 * rwnd is updated when the event is freed.
2140 if (!sctp_ulpevent_is_notification(event))
2141 sctp_assoc_rwnd_increase(event->asoc, copied);
2143 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2144 (event->msg_flags & MSG_EOR))
2145 msg->msg_flags |= MSG_EOR;
2147 msg->msg_flags &= ~MSG_EOR;
2150 if (flags & MSG_PEEK) {
2151 /* Release the skb reference acquired after peeking the skb in
2152 * sctp_skb_recv_datagram().
2156 /* Free the event which includes releasing the reference to
2157 * the owner of the skb, freeing the skb and updating the
2160 sctp_ulpevent_free(event);
2163 sctp_release_sock(sk);
2167 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2169 * This option is a on/off flag. If enabled no SCTP message
2170 * fragmentation will be performed. Instead if a message being sent
2171 * exceeds the current PMTU size, the message will NOT be sent and
2172 * instead a error will be indicated to the user.
2174 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2175 char __user *optval,
2176 unsigned int optlen)
2180 if (optlen < sizeof(int))
2183 if (get_user(val, (int __user *)optval))
2186 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2191 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2192 unsigned int optlen)
2194 struct sctp_association *asoc;
2195 struct sctp_ulpevent *event;
2197 if (optlen > sizeof(struct sctp_event_subscribe))
2199 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2203 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2204 * if there is no data to be sent or retransmit, the stack will
2205 * immediately send up this notification.
2207 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2208 &sctp_sk(sk)->subscribe)) {
2209 asoc = sctp_id2assoc(sk, 0);
2211 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2212 event = sctp_ulpevent_make_sender_dry_event(asoc,
2217 sctp_ulpq_tail_event(&asoc->ulpq, event);
2224 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2226 * This socket option is applicable to the UDP-style socket only. When
2227 * set it will cause associations that are idle for more than the
2228 * specified number of seconds to automatically close. An association
2229 * being idle is defined an association that has NOT sent or received
2230 * user data. The special value of '0' indicates that no automatic
2231 * close of any associations should be performed. The option expects an
2232 * integer defining the number of seconds of idle time before an
2233 * association is closed.
2235 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2236 unsigned int optlen)
2238 struct sctp_sock *sp = sctp_sk(sk);
2240 /* Applicable to UDP-style socket only */
2241 if (sctp_style(sk, TCP))
2243 if (optlen != sizeof(int))
2245 if (copy_from_user(&sp->autoclose, optval, optlen))
2251 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2253 * Applications can enable or disable heartbeats for any peer address of
2254 * an association, modify an address's heartbeat interval, force a
2255 * heartbeat to be sent immediately, and adjust the address's maximum
2256 * number of retransmissions sent before an address is considered
2257 * unreachable. The following structure is used to access and modify an
2258 * address's parameters:
2260 * struct sctp_paddrparams {
2261 * sctp_assoc_t spp_assoc_id;
2262 * struct sockaddr_storage spp_address;
2263 * uint32_t spp_hbinterval;
2264 * uint16_t spp_pathmaxrxt;
2265 * uint32_t spp_pathmtu;
2266 * uint32_t spp_sackdelay;
2267 * uint32_t spp_flags;
2270 * spp_assoc_id - (one-to-many style socket) This is filled in the
2271 * application, and identifies the association for
2273 * spp_address - This specifies which address is of interest.
2274 * spp_hbinterval - This contains the value of the heartbeat interval,
2275 * in milliseconds. If a value of zero
2276 * is present in this field then no changes are to
2277 * be made to this parameter.
2278 * spp_pathmaxrxt - This contains the maximum number of
2279 * retransmissions before this address shall be
2280 * considered unreachable. If a value of zero
2281 * is present in this field then no changes are to
2282 * be made to this parameter.
2283 * spp_pathmtu - When Path MTU discovery is disabled the value
2284 * specified here will be the "fixed" path mtu.
2285 * Note that if the spp_address field is empty
2286 * then all associations on this address will
2287 * have this fixed path mtu set upon them.
2289 * spp_sackdelay - When delayed sack is enabled, this value specifies
2290 * the number of milliseconds that sacks will be delayed
2291 * for. This value will apply to all addresses of an
2292 * association if the spp_address field is empty. Note
2293 * also, that if delayed sack is enabled and this
2294 * value is set to 0, no change is made to the last
2295 * recorded delayed sack timer value.
2297 * spp_flags - These flags are used to control various features
2298 * on an association. The flag field may contain
2299 * zero or more of the following options.
2301 * SPP_HB_ENABLE - Enable heartbeats on the
2302 * specified address. Note that if the address
2303 * field is empty all addresses for the association
2304 * have heartbeats enabled upon them.
2306 * SPP_HB_DISABLE - Disable heartbeats on the
2307 * speicifed address. Note that if the address
2308 * field is empty all addresses for the association
2309 * will have their heartbeats disabled. Note also
2310 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2311 * mutually exclusive, only one of these two should
2312 * be specified. Enabling both fields will have
2313 * undetermined results.
2315 * SPP_HB_DEMAND - Request a user initiated heartbeat
2316 * to be made immediately.
2318 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2319 * heartbeat delayis to be set to the value of 0
2322 * SPP_PMTUD_ENABLE - This field will enable PMTU
2323 * discovery upon the specified address. Note that
2324 * if the address feild is empty then all addresses
2325 * on the association are effected.
2327 * SPP_PMTUD_DISABLE - This field will disable PMTU
2328 * discovery upon the specified address. Note that
2329 * if the address feild is empty then all addresses
2330 * on the association are effected. Not also that
2331 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2332 * exclusive. Enabling both will have undetermined
2335 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2336 * on delayed sack. The time specified in spp_sackdelay
2337 * is used to specify the sack delay for this address. Note
2338 * that if spp_address is empty then all addresses will
2339 * enable delayed sack and take on the sack delay
2340 * value specified in spp_sackdelay.
2341 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2342 * off delayed sack. If the spp_address field is blank then
2343 * delayed sack is disabled for the entire association. Note
2344 * also that this field is mutually exclusive to
2345 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2348 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2349 struct sctp_transport *trans,
2350 struct sctp_association *asoc,
2351 struct sctp_sock *sp,
2354 int sackdelay_change)
2358 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2359 struct net *net = sock_net(trans->asoc->base.sk);
2361 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2366 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2367 * this field is ignored. Note also that a value of zero indicates
2368 * the current setting should be left unchanged.
2370 if (params->spp_flags & SPP_HB_ENABLE) {
2372 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2373 * set. This lets us use 0 value when this flag
2376 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2377 params->spp_hbinterval = 0;
2379 if (params->spp_hbinterval ||
2380 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2383 msecs_to_jiffies(params->spp_hbinterval);
2386 msecs_to_jiffies(params->spp_hbinterval);
2388 sp->hbinterval = params->spp_hbinterval;
2395 trans->param_flags =
2396 (trans->param_flags & ~SPP_HB) | hb_change;
2399 (asoc->param_flags & ~SPP_HB) | hb_change;
2402 (sp->param_flags & ~SPP_HB) | hb_change;
2406 /* When Path MTU discovery is disabled the value specified here will
2407 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2408 * include the flag SPP_PMTUD_DISABLE for this field to have any
2411 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2413 trans->pathmtu = params->spp_pathmtu;
2414 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2416 asoc->pathmtu = params->spp_pathmtu;
2417 sctp_frag_point(asoc, params->spp_pathmtu);
2419 sp->pathmtu = params->spp_pathmtu;
2425 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2426 (params->spp_flags & SPP_PMTUD_ENABLE);
2427 trans->param_flags =
2428 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2430 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2431 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2435 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2438 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2442 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2443 * value of this field is ignored. Note also that a value of zero
2444 * indicates the current setting should be left unchanged.
2446 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2449 msecs_to_jiffies(params->spp_sackdelay);
2452 msecs_to_jiffies(params->spp_sackdelay);
2454 sp->sackdelay = params->spp_sackdelay;
2458 if (sackdelay_change) {
2460 trans->param_flags =
2461 (trans->param_flags & ~SPP_SACKDELAY) |
2465 (asoc->param_flags & ~SPP_SACKDELAY) |
2469 (sp->param_flags & ~SPP_SACKDELAY) |
2474 /* Note that a value of zero indicates the current setting should be
2477 if (params->spp_pathmaxrxt) {
2479 trans->pathmaxrxt = params->spp_pathmaxrxt;
2481 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2483 sp->pathmaxrxt = params->spp_pathmaxrxt;
2490 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2491 char __user *optval,
2492 unsigned int optlen)
2494 struct sctp_paddrparams params;
2495 struct sctp_transport *trans = NULL;
2496 struct sctp_association *asoc = NULL;
2497 struct sctp_sock *sp = sctp_sk(sk);
2499 int hb_change, pmtud_change, sackdelay_change;
2501 if (optlen != sizeof(struct sctp_paddrparams))
2504 if (copy_from_user(¶ms, optval, optlen))
2507 /* Validate flags and value parameters. */
2508 hb_change = params.spp_flags & SPP_HB;
2509 pmtud_change = params.spp_flags & SPP_PMTUD;
2510 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2512 if (hb_change == SPP_HB ||
2513 pmtud_change == SPP_PMTUD ||
2514 sackdelay_change == SPP_SACKDELAY ||
2515 params.spp_sackdelay > 500 ||
2516 (params.spp_pathmtu &&
2517 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2520 /* If an address other than INADDR_ANY is specified, and
2521 * no transport is found, then the request is invalid.
2523 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
2524 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2525 params.spp_assoc_id);
2530 /* Get association, if assoc_id != 0 and the socket is a one
2531 * to many style socket, and an association was not found, then
2532 * the id was invalid.
2534 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2535 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2538 /* Heartbeat demand can only be sent on a transport or
2539 * association, but not a socket.
2541 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2544 /* Process parameters. */
2545 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2546 hb_change, pmtud_change,
2552 /* If changes are for association, also apply parameters to each
2555 if (!trans && asoc) {
2556 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2558 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2559 hb_change, pmtud_change,
2568 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2570 * This option will effect the way delayed acks are performed. This
2571 * option allows you to get or set the delayed ack time, in
2572 * milliseconds. It also allows changing the delayed ack frequency.
2573 * Changing the frequency to 1 disables the delayed sack algorithm. If
2574 * the assoc_id is 0, then this sets or gets the endpoints default
2575 * values. If the assoc_id field is non-zero, then the set or get
2576 * effects the specified association for the one to many model (the
2577 * assoc_id field is ignored by the one to one model). Note that if
2578 * sack_delay or sack_freq are 0 when setting this option, then the
2579 * current values will remain unchanged.
2581 * struct sctp_sack_info {
2582 * sctp_assoc_t sack_assoc_id;
2583 * uint32_t sack_delay;
2584 * uint32_t sack_freq;
2587 * sack_assoc_id - This parameter, indicates which association the user
2588 * is performing an action upon. Note that if this field's value is
2589 * zero then the endpoints default value is changed (effecting future
2590 * associations only).
2592 * sack_delay - This parameter contains the number of milliseconds that
2593 * the user is requesting the delayed ACK timer be set to. Note that
2594 * this value is defined in the standard to be between 200 and 500
2597 * sack_freq - This parameter contains the number of packets that must
2598 * be received before a sack is sent without waiting for the delay
2599 * timer to expire. The default value for this is 2, setting this
2600 * value to 1 will disable the delayed sack algorithm.
2603 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2604 char __user *optval, unsigned int optlen)
2606 struct sctp_sack_info params;
2607 struct sctp_transport *trans = NULL;
2608 struct sctp_association *asoc = NULL;
2609 struct sctp_sock *sp = sctp_sk(sk);
2611 if (optlen == sizeof(struct sctp_sack_info)) {
2612 if (copy_from_user(¶ms, optval, optlen))
2615 if (params.sack_delay == 0 && params.sack_freq == 0)
2617 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2618 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2619 pr_warn("Use struct sctp_sack_info instead\n");
2620 if (copy_from_user(¶ms, optval, optlen))
2623 if (params.sack_delay == 0)
2624 params.sack_freq = 1;
2626 params.sack_freq = 0;
2630 /* Validate value parameter. */
2631 if (params.sack_delay > 500)
2634 /* Get association, if sack_assoc_id != 0 and the socket is a one
2635 * to many style socket, and an association was not found, then
2636 * the id was invalid.
2638 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2639 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2642 if (params.sack_delay) {
2645 msecs_to_jiffies(params.sack_delay);
2647 (asoc->param_flags & ~SPP_SACKDELAY) |
2648 SPP_SACKDELAY_ENABLE;
2650 sp->sackdelay = params.sack_delay;
2652 (sp->param_flags & ~SPP_SACKDELAY) |
2653 SPP_SACKDELAY_ENABLE;
2657 if (params.sack_freq == 1) {
2660 (asoc->param_flags & ~SPP_SACKDELAY) |
2661 SPP_SACKDELAY_DISABLE;
2664 (sp->param_flags & ~SPP_SACKDELAY) |
2665 SPP_SACKDELAY_DISABLE;
2667 } else if (params.sack_freq > 1) {
2669 asoc->sackfreq = params.sack_freq;
2671 (asoc->param_flags & ~SPP_SACKDELAY) |
2672 SPP_SACKDELAY_ENABLE;
2674 sp->sackfreq = params.sack_freq;
2676 (sp->param_flags & ~SPP_SACKDELAY) |
2677 SPP_SACKDELAY_ENABLE;
2681 /* If change is for association, also apply to each transport. */
2683 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2685 if (params.sack_delay) {
2687 msecs_to_jiffies(params.sack_delay);
2688 trans->param_flags =
2689 (trans->param_flags & ~SPP_SACKDELAY) |
2690 SPP_SACKDELAY_ENABLE;
2692 if (params.sack_freq == 1) {
2693 trans->param_flags =
2694 (trans->param_flags & ~SPP_SACKDELAY) |
2695 SPP_SACKDELAY_DISABLE;
2696 } else if (params.sack_freq > 1) {
2697 trans->sackfreq = params.sack_freq;
2698 trans->param_flags =
2699 (trans->param_flags & ~SPP_SACKDELAY) |
2700 SPP_SACKDELAY_ENABLE;
2708 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2710 * Applications can specify protocol parameters for the default association
2711 * initialization. The option name argument to setsockopt() and getsockopt()
2714 * Setting initialization parameters is effective only on an unconnected
2715 * socket (for UDP-style sockets only future associations are effected
2716 * by the change). With TCP-style sockets, this option is inherited by
2717 * sockets derived from a listener socket.
2719 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2721 struct sctp_initmsg sinit;
2722 struct sctp_sock *sp = sctp_sk(sk);
2724 if (optlen != sizeof(struct sctp_initmsg))
2726 if (copy_from_user(&sinit, optval, optlen))
2729 if (sinit.sinit_num_ostreams)
2730 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2731 if (sinit.sinit_max_instreams)
2732 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2733 if (sinit.sinit_max_attempts)
2734 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2735 if (sinit.sinit_max_init_timeo)
2736 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2742 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2744 * Applications that wish to use the sendto() system call may wish to
2745 * specify a default set of parameters that would normally be supplied
2746 * through the inclusion of ancillary data. This socket option allows
2747 * such an application to set the default sctp_sndrcvinfo structure.
2748 * The application that wishes to use this socket option simply passes
2749 * in to this call the sctp_sndrcvinfo structure defined in Section
2750 * 5.2.2) The input parameters accepted by this call include
2751 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2752 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2753 * to this call if the caller is using the UDP model.
2755 static int sctp_setsockopt_default_send_param(struct sock *sk,
2756 char __user *optval,
2757 unsigned int optlen)
2759 struct sctp_sndrcvinfo info;
2760 struct sctp_association *asoc;
2761 struct sctp_sock *sp = sctp_sk(sk);
2763 if (optlen != sizeof(struct sctp_sndrcvinfo))
2765 if (copy_from_user(&info, optval, optlen))
2768 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2769 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2773 asoc->default_stream = info.sinfo_stream;
2774 asoc->default_flags = info.sinfo_flags;
2775 asoc->default_ppid = info.sinfo_ppid;
2776 asoc->default_context = info.sinfo_context;
2777 asoc->default_timetolive = info.sinfo_timetolive;
2779 sp->default_stream = info.sinfo_stream;
2780 sp->default_flags = info.sinfo_flags;
2781 sp->default_ppid = info.sinfo_ppid;
2782 sp->default_context = info.sinfo_context;
2783 sp->default_timetolive = info.sinfo_timetolive;
2789 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2791 * Requests that the local SCTP stack use the enclosed peer address as
2792 * the association primary. The enclosed address must be one of the
2793 * association peer's addresses.
2795 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2796 unsigned int optlen)
2798 struct sctp_prim prim;
2799 struct sctp_transport *trans;
2801 if (optlen != sizeof(struct sctp_prim))
2804 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2807 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2811 sctp_assoc_set_primary(trans->asoc, trans);
2817 * 7.1.5 SCTP_NODELAY
2819 * Turn on/off any Nagle-like algorithm. This means that packets are
2820 * generally sent as soon as possible and no unnecessary delays are
2821 * introduced, at the cost of more packets in the network. Expects an
2822 * integer boolean flag.
2824 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2825 unsigned int optlen)
2829 if (optlen < sizeof(int))
2831 if (get_user(val, (int __user *)optval))
2834 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2840 * 7.1.1 SCTP_RTOINFO
2842 * The protocol parameters used to initialize and bound retransmission
2843 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2844 * and modify these parameters.
2845 * All parameters are time values, in milliseconds. A value of 0, when
2846 * modifying the parameters, indicates that the current value should not
2850 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2852 struct sctp_rtoinfo rtoinfo;
2853 struct sctp_association *asoc;
2855 if (optlen != sizeof (struct sctp_rtoinfo))
2858 if (copy_from_user(&rtoinfo, optval, optlen))
2861 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2863 /* Set the values to the specific association */
2864 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2868 if (rtoinfo.srto_initial != 0)
2870 msecs_to_jiffies(rtoinfo.srto_initial);
2871 if (rtoinfo.srto_max != 0)
2872 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2873 if (rtoinfo.srto_min != 0)
2874 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2876 /* If there is no association or the association-id = 0
2877 * set the values to the endpoint.
2879 struct sctp_sock *sp = sctp_sk(sk);
2881 if (rtoinfo.srto_initial != 0)
2882 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2883 if (rtoinfo.srto_max != 0)
2884 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2885 if (rtoinfo.srto_min != 0)
2886 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2894 * 7.1.2 SCTP_ASSOCINFO
2896 * This option is used to tune the maximum retransmission attempts
2897 * of the association.
2898 * Returns an error if the new association retransmission value is
2899 * greater than the sum of the retransmission value of the peer.
2900 * See [SCTP] for more information.
2903 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2906 struct sctp_assocparams assocparams;
2907 struct sctp_association *asoc;
2909 if (optlen != sizeof(struct sctp_assocparams))
2911 if (copy_from_user(&assocparams, optval, optlen))
2914 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2916 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2919 /* Set the values to the specific association */
2921 if (assocparams.sasoc_asocmaxrxt != 0) {
2924 struct sctp_transport *peer_addr;
2926 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2928 path_sum += peer_addr->pathmaxrxt;
2932 /* Only validate asocmaxrxt if we have more than
2933 * one path/transport. We do this because path
2934 * retransmissions are only counted when we have more
2938 assocparams.sasoc_asocmaxrxt > path_sum)
2941 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2944 if (assocparams.sasoc_cookie_life != 0) {
2945 asoc->cookie_life.tv_sec =
2946 assocparams.sasoc_cookie_life / 1000;
2947 asoc->cookie_life.tv_usec =
2948 (assocparams.sasoc_cookie_life % 1000)
2952 /* Set the values to the endpoint */
2953 struct sctp_sock *sp = sctp_sk(sk);
2955 if (assocparams.sasoc_asocmaxrxt != 0)
2956 sp->assocparams.sasoc_asocmaxrxt =
2957 assocparams.sasoc_asocmaxrxt;
2958 if (assocparams.sasoc_cookie_life != 0)
2959 sp->assocparams.sasoc_cookie_life =
2960 assocparams.sasoc_cookie_life;
2966 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2968 * This socket option is a boolean flag which turns on or off mapped V4
2969 * addresses. If this option is turned on and the socket is type
2970 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2971 * If this option is turned off, then no mapping will be done of V4
2972 * addresses and a user will receive both PF_INET6 and PF_INET type
2973 * addresses on the socket.
2975 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2978 struct sctp_sock *sp = sctp_sk(sk);
2980 if (optlen < sizeof(int))
2982 if (get_user(val, (int __user *)optval))
2993 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2994 * This option will get or set the maximum size to put in any outgoing
2995 * SCTP DATA chunk. If a message is larger than this size it will be
2996 * fragmented by SCTP into the specified size. Note that the underlying
2997 * SCTP implementation may fragment into smaller sized chunks when the
2998 * PMTU of the underlying association is smaller than the value set by
2999 * the user. The default value for this option is '0' which indicates
3000 * the user is NOT limiting fragmentation and only the PMTU will effect
3001 * SCTP's choice of DATA chunk size. Note also that values set larger
3002 * than the maximum size of an IP datagram will effectively let SCTP
3003 * control fragmentation (i.e. the same as setting this option to 0).
3005 * The following structure is used to access and modify this parameter:
3007 * struct sctp_assoc_value {
3008 * sctp_assoc_t assoc_id;
3009 * uint32_t assoc_value;
3012 * assoc_id: This parameter is ignored for one-to-one style sockets.
3013 * For one-to-many style sockets this parameter indicates which
3014 * association the user is performing an action upon. Note that if
3015 * this field's value is zero then the endpoints default value is
3016 * changed (effecting future associations only).
3017 * assoc_value: This parameter specifies the maximum size in bytes.
3019 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3021 struct sctp_assoc_value params;
3022 struct sctp_association *asoc;
3023 struct sctp_sock *sp = sctp_sk(sk);
3026 if (optlen == sizeof(int)) {
3027 pr_warn("Use of int in maxseg socket option deprecated\n");
3028 pr_warn("Use struct sctp_assoc_value instead\n");
3029 if (copy_from_user(&val, optval, optlen))
3031 params.assoc_id = 0;
3032 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3033 if (copy_from_user(¶ms, optval, optlen))
3035 val = params.assoc_value;
3039 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3042 asoc = sctp_id2assoc(sk, params.assoc_id);
3043 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3048 val = asoc->pathmtu;
3049 val -= sp->pf->af->net_header_len;
3050 val -= sizeof(struct sctphdr) +
3051 sizeof(struct sctp_data_chunk);
3053 asoc->user_frag = val;
3054 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3056 sp->user_frag = val;
3064 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3066 * Requests that the peer mark the enclosed address as the association
3067 * primary. The enclosed address must be one of the association's
3068 * locally bound addresses. The following structure is used to make a
3069 * set primary request:
3071 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3072 unsigned int optlen)
3074 struct net *net = sock_net(sk);
3075 struct sctp_sock *sp;
3076 struct sctp_association *asoc = NULL;
3077 struct sctp_setpeerprim prim;
3078 struct sctp_chunk *chunk;
3084 if (!net->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 net *net = sock_net(sk);
3322 struct sctp_authchunk val;
3324 if (!net->sctp.auth_enable)
3327 if (optlen != sizeof(struct sctp_authchunk))
3329 if (copy_from_user(&val, optval, optlen))
3332 switch (val.sauth_chunk) {
3334 case SCTP_CID_INIT_ACK:
3335 case SCTP_CID_SHUTDOWN_COMPLETE:
3340 /* add this chunk id to the endpoint */
3341 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3345 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3347 * This option gets or sets the list of HMAC algorithms that the local
3348 * endpoint requires the peer to use.
3350 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3351 char __user *optval,
3352 unsigned int optlen)
3354 struct net *net = sock_net(sk);
3355 struct sctp_hmacalgo *hmacs;
3359 if (!net->sctp.auth_enable)
3362 if (optlen < sizeof(struct sctp_hmacalgo))
3365 hmacs= memdup_user(optval, optlen);
3367 return PTR_ERR(hmacs);
3369 idents = hmacs->shmac_num_idents;
3370 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3371 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3376 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3383 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3385 * This option will set a shared secret key which is used to build an
3386 * association shared key.
3388 static int sctp_setsockopt_auth_key(struct sock *sk,
3389 char __user *optval,
3390 unsigned int optlen)
3392 struct net *net = sock_net(sk);
3393 struct sctp_authkey *authkey;
3394 struct sctp_association *asoc;
3397 if (!net->sctp.auth_enable)
3400 if (optlen <= sizeof(struct sctp_authkey))
3403 authkey= memdup_user(optval, optlen);
3404 if (IS_ERR(authkey))
3405 return PTR_ERR(authkey);
3407 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3412 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3413 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3418 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3425 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3427 * This option will get or set the active shared key to be used to build
3428 * the association shared key.
3430 static int sctp_setsockopt_active_key(struct sock *sk,
3431 char __user *optval,
3432 unsigned int optlen)
3434 struct net *net = sock_net(sk);
3435 struct sctp_authkeyid val;
3436 struct sctp_association *asoc;
3438 if (!net->sctp.auth_enable)
3441 if (optlen != sizeof(struct sctp_authkeyid))
3443 if (copy_from_user(&val, optval, optlen))
3446 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3447 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3450 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3451 val.scact_keynumber);
3455 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3457 * This set option will delete a shared secret key from use.
3459 static int sctp_setsockopt_del_key(struct sock *sk,
3460 char __user *optval,
3461 unsigned int optlen)
3463 struct net *net = sock_net(sk);
3464 struct sctp_authkeyid val;
3465 struct sctp_association *asoc;
3467 if (!net->sctp.auth_enable)
3470 if (optlen != sizeof(struct sctp_authkeyid))
3472 if (copy_from_user(&val, optval, optlen))
3475 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3476 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3479 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3480 val.scact_keynumber);
3485 * 8.1.23 SCTP_AUTO_ASCONF
3487 * This option will enable or disable the use of the automatic generation of
3488 * ASCONF chunks to add and delete addresses to an existing association. Note
3489 * that this option has two caveats namely: a) it only affects sockets that
3490 * are bound to all addresses available to the SCTP stack, and b) the system
3491 * administrator may have an overriding control that turns the ASCONF feature
3492 * off no matter what setting the socket option may have.
3493 * This option expects an integer boolean flag, where a non-zero value turns on
3494 * the option, and a zero value turns off the option.
3495 * Note. In this implementation, socket operation overrides default parameter
3496 * being set by sysctl as well as FreeBSD implementation
3498 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3499 unsigned int optlen)
3502 struct sctp_sock *sp = sctp_sk(sk);
3504 if (optlen < sizeof(int))
3506 if (get_user(val, (int __user *)optval))
3508 if (!sctp_is_ep_boundall(sk) && val)
3510 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3513 if (val == 0 && sp->do_auto_asconf) {
3514 list_del(&sp->auto_asconf_list);
3515 sp->do_auto_asconf = 0;
3516 } else if (val && !sp->do_auto_asconf) {
3517 list_add_tail(&sp->auto_asconf_list,
3518 &sock_net(sk)->sctp.auto_asconf_splist);
3519 sp->do_auto_asconf = 1;
3526 * SCTP_PEER_ADDR_THLDS
3528 * This option allows us to alter the partially failed threshold for one or all
3529 * transports in an association. See Section 6.1 of:
3530 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3532 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3533 char __user *optval,
3534 unsigned int optlen)
3536 struct sctp_paddrthlds val;
3537 struct sctp_transport *trans;
3538 struct sctp_association *asoc;
3540 if (optlen < sizeof(struct sctp_paddrthlds))
3542 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3543 sizeof(struct sctp_paddrthlds)))
3547 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3548 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3551 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3553 if (val.spt_pathmaxrxt)
3554 trans->pathmaxrxt = val.spt_pathmaxrxt;
3555 trans->pf_retrans = val.spt_pathpfthld;
3558 if (val.spt_pathmaxrxt)
3559 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3560 asoc->pf_retrans = val.spt_pathpfthld;
3562 trans = sctp_addr_id2transport(sk, &val.spt_address,
3567 if (val.spt_pathmaxrxt)
3568 trans->pathmaxrxt = val.spt_pathmaxrxt;
3569 trans->pf_retrans = val.spt_pathpfthld;
3575 /* API 6.2 setsockopt(), getsockopt()
3577 * Applications use setsockopt() and getsockopt() to set or retrieve
3578 * socket options. Socket options are used to change the default
3579 * behavior of sockets calls. They are described in Section 7.
3583 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3584 * int __user *optlen);
3585 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3588 * sd - the socket descript.
3589 * level - set to IPPROTO_SCTP for all SCTP options.
3590 * optname - the option name.
3591 * optval - the buffer to store the value of the option.
3592 * optlen - the size of the buffer.
3594 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3595 char __user *optval, unsigned int optlen)
3599 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3602 /* I can hardly begin to describe how wrong this is. This is
3603 * so broken as to be worse than useless. The API draft
3604 * REALLY is NOT helpful here... I am not convinced that the
3605 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3606 * are at all well-founded.
3608 if (level != SOL_SCTP) {
3609 struct sctp_af *af = sctp_sk(sk)->pf->af;
3610 retval = af->setsockopt(sk, level, optname, optval, optlen);
3617 case SCTP_SOCKOPT_BINDX_ADD:
3618 /* 'optlen' is the size of the addresses buffer. */
3619 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3620 optlen, SCTP_BINDX_ADD_ADDR);
3623 case SCTP_SOCKOPT_BINDX_REM:
3624 /* 'optlen' is the size of the addresses buffer. */
3625 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3626 optlen, SCTP_BINDX_REM_ADDR);
3629 case SCTP_SOCKOPT_CONNECTX_OLD:
3630 /* 'optlen' is the size of the addresses buffer. */
3631 retval = sctp_setsockopt_connectx_old(sk,
3632 (struct sockaddr __user *)optval,
3636 case SCTP_SOCKOPT_CONNECTX:
3637 /* 'optlen' is the size of the addresses buffer. */
3638 retval = sctp_setsockopt_connectx(sk,
3639 (struct sockaddr __user *)optval,
3643 case SCTP_DISABLE_FRAGMENTS:
3644 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3648 retval = sctp_setsockopt_events(sk, optval, optlen);
3651 case SCTP_AUTOCLOSE:
3652 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3655 case SCTP_PEER_ADDR_PARAMS:
3656 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3659 case SCTP_DELAYED_SACK:
3660 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3662 case SCTP_PARTIAL_DELIVERY_POINT:
3663 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3667 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3669 case SCTP_DEFAULT_SEND_PARAM:
3670 retval = sctp_setsockopt_default_send_param(sk, optval,
3673 case SCTP_PRIMARY_ADDR:
3674 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3676 case SCTP_SET_PEER_PRIMARY_ADDR:
3677 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3680 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3683 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3685 case SCTP_ASSOCINFO:
3686 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3688 case SCTP_I_WANT_MAPPED_V4_ADDR:
3689 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3692 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3694 case SCTP_ADAPTATION_LAYER:
3695 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3698 retval = sctp_setsockopt_context(sk, optval, optlen);
3700 case SCTP_FRAGMENT_INTERLEAVE:
3701 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3703 case SCTP_MAX_BURST:
3704 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3706 case SCTP_AUTH_CHUNK:
3707 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3709 case SCTP_HMAC_IDENT:
3710 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3713 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3715 case SCTP_AUTH_ACTIVE_KEY:
3716 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3718 case SCTP_AUTH_DELETE_KEY:
3719 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3721 case SCTP_AUTO_ASCONF:
3722 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3724 case SCTP_PEER_ADDR_THLDS:
3725 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3728 retval = -ENOPROTOOPT;
3732 sctp_release_sock(sk);
3738 /* API 3.1.6 connect() - UDP Style Syntax
3740 * An application may use the connect() call in the UDP model to initiate an
3741 * association without sending data.
3745 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3747 * sd: the socket descriptor to have a new association added to.
3749 * nam: the address structure (either struct sockaddr_in or struct
3750 * sockaddr_in6 defined in RFC2553 [7]).
3752 * len: the size of the address.
3754 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3762 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3763 __func__, sk, addr, addr_len);
3765 /* Validate addr_len before calling common connect/connectx routine. */
3766 af = sctp_get_af_specific(addr->sa_family);
3767 if (!af || addr_len < af->sockaddr_len) {
3770 /* Pass correct addr len to common routine (so it knows there
3771 * is only one address being passed.
3773 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3776 sctp_release_sock(sk);
3780 /* FIXME: Write comments. */
3781 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3783 return -EOPNOTSUPP; /* STUB */
3786 /* 4.1.4 accept() - TCP Style Syntax
3788 * Applications use accept() call to remove an established SCTP
3789 * association from the accept queue of the endpoint. A new socket
3790 * descriptor will be returned from accept() to represent the newly
3791 * formed association.
3793 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3795 struct sctp_sock *sp;
3796 struct sctp_endpoint *ep;
3797 struct sock *newsk = NULL;
3798 struct sctp_association *asoc;
3807 if (!sctp_style(sk, TCP)) {
3808 error = -EOPNOTSUPP;
3812 if (!sctp_sstate(sk, LISTENING)) {
3817 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3819 error = sctp_wait_for_accept(sk, timeo);
3823 /* We treat the list of associations on the endpoint as the accept
3824 * queue and pick the first association on the list.
3826 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3828 newsk = sp->pf->create_accept_sk(sk, asoc);
3834 /* Populate the fields of the newsk from the oldsk and migrate the
3835 * asoc to the newsk.
3837 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3840 sctp_release_sock(sk);
3845 /* The SCTP ioctl handler. */
3846 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3853 * SEQPACKET-style sockets in LISTENING state are valid, for
3854 * SCTP, so only discard TCP-style sockets in LISTENING state.
3856 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3861 struct sk_buff *skb;
3862 unsigned int amount = 0;
3864 skb = skb_peek(&sk->sk_receive_queue);
3867 * We will only return the amount of this packet since
3868 * that is all that will be read.
3872 rc = put_user(amount, (int __user *)arg);
3880 sctp_release_sock(sk);
3884 /* This is the function which gets called during socket creation to
3885 * initialized the SCTP-specific portion of the sock.
3886 * The sock structure should already be zero-filled memory.
3888 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3890 struct net *net = sock_net(sk);
3891 struct sctp_endpoint *ep;
3892 struct sctp_sock *sp;
3894 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3898 /* Initialize the SCTP per socket area. */
3899 switch (sk->sk_type) {
3900 case SOCK_SEQPACKET:
3901 sp->type = SCTP_SOCKET_UDP;
3904 sp->type = SCTP_SOCKET_TCP;
3907 return -ESOCKTNOSUPPORT;
3910 /* Initialize default send parameters. These parameters can be
3911 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3913 sp->default_stream = 0;
3914 sp->default_ppid = 0;
3915 sp->default_flags = 0;
3916 sp->default_context = 0;
3917 sp->default_timetolive = 0;
3919 sp->default_rcv_context = 0;
3920 sp->max_burst = net->sctp.max_burst;
3922 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
3924 /* Initialize default setup parameters. These parameters
3925 * can be modified with the SCTP_INITMSG socket option or
3926 * overridden by the SCTP_INIT CMSG.
3928 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3929 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3930 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
3931 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
3933 /* Initialize default RTO related parameters. These parameters can
3934 * be modified for with the SCTP_RTOINFO socket option.
3936 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
3937 sp->rtoinfo.srto_max = net->sctp.rto_max;
3938 sp->rtoinfo.srto_min = net->sctp.rto_min;
3940 /* Initialize default association related parameters. These parameters
3941 * can be modified with the SCTP_ASSOCINFO socket option.
3943 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
3944 sp->assocparams.sasoc_number_peer_destinations = 0;
3945 sp->assocparams.sasoc_peer_rwnd = 0;
3946 sp->assocparams.sasoc_local_rwnd = 0;
3947 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
3949 /* Initialize default event subscriptions. By default, all the
3952 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3954 /* Default Peer Address Parameters. These defaults can
3955 * be modified via SCTP_PEER_ADDR_PARAMS
3957 sp->hbinterval = net->sctp.hb_interval;
3958 sp->pathmaxrxt = net->sctp.max_retrans_path;
3959 sp->pathmtu = 0; // allow default discovery
3960 sp->sackdelay = net->sctp.sack_timeout;
3962 sp->param_flags = SPP_HB_ENABLE |
3964 SPP_SACKDELAY_ENABLE;
3966 /* If enabled no SCTP message fragmentation will be performed.
3967 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3969 sp->disable_fragments = 0;
3971 /* Enable Nagle algorithm by default. */
3974 /* Enable by default. */
3977 /* Auto-close idle associations after the configured
3978 * number of seconds. A value of 0 disables this
3979 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3980 * for UDP-style sockets only.
3984 /* User specified fragmentation limit. */
3987 sp->adaptation_ind = 0;
3989 sp->pf = sctp_get_pf_specific(sk->sk_family);
3991 /* Control variables for partial data delivery. */
3992 atomic_set(&sp->pd_mode, 0);
3993 skb_queue_head_init(&sp->pd_lobby);
3994 sp->frag_interleave = 0;
3996 /* Create a per socket endpoint structure. Even if we
3997 * change the data structure relationships, this may still
3998 * be useful for storing pre-connect address information.
4000 ep = sctp_endpoint_new(sk, GFP_KERNEL);
4007 SCTP_DBG_OBJCNT_INC(sock);
4010 percpu_counter_inc(&sctp_sockets_allocated);
4011 sock_prot_inuse_add(net, sk->sk_prot, 1);
4012 if (net->sctp.default_auto_asconf) {
4013 list_add_tail(&sp->auto_asconf_list,
4014 &net->sctp.auto_asconf_splist);
4015 sp->do_auto_asconf = 1;
4017 sp->do_auto_asconf = 0;
4023 /* Cleanup any SCTP per socket resources. */
4024 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
4026 struct sctp_sock *sp;
4028 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
4030 /* Release our hold on the endpoint. */
4032 /* This could happen during socket init, thus we bail out
4033 * early, since the rest of the below is not setup either.
4038 if (sp->do_auto_asconf) {
4039 sp->do_auto_asconf = 0;
4040 list_del(&sp->auto_asconf_list);
4042 sctp_endpoint_free(sp->ep);
4044 percpu_counter_dec(&sctp_sockets_allocated);
4045 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4049 /* API 4.1.7 shutdown() - TCP Style Syntax
4050 * int shutdown(int socket, int how);
4052 * sd - the socket descriptor of the association to be closed.
4053 * how - Specifies the type of shutdown. The values are
4056 * Disables further receive operations. No SCTP
4057 * protocol action is taken.
4059 * Disables further send operations, and initiates
4060 * the SCTP shutdown sequence.
4062 * Disables further send and receive operations
4063 * and initiates the SCTP shutdown sequence.
4065 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
4067 struct net *net = sock_net(sk);
4068 struct sctp_endpoint *ep;
4069 struct sctp_association *asoc;
4071 if (!sctp_style(sk, TCP))
4074 if (how & SEND_SHUTDOWN) {
4075 ep = sctp_sk(sk)->ep;
4076 if (!list_empty(&ep->asocs)) {
4077 asoc = list_entry(ep->asocs.next,
4078 struct sctp_association, asocs);
4079 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4084 /* 7.2.1 Association Status (SCTP_STATUS)
4086 * Applications can retrieve current status information about an
4087 * association, including association state, peer receiver window size,
4088 * number of unacked data chunks, and number of data chunks pending
4089 * receipt. This information is read-only.
4091 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4092 char __user *optval,
4095 struct sctp_status status;
4096 struct sctp_association *asoc = NULL;
4097 struct sctp_transport *transport;
4098 sctp_assoc_t associd;
4101 if (len < sizeof(status)) {
4106 len = sizeof(status);
4107 if (copy_from_user(&status, optval, len)) {
4112 associd = status.sstat_assoc_id;
4113 asoc = sctp_id2assoc(sk, associd);
4119 transport = asoc->peer.primary_path;
4121 status.sstat_assoc_id = sctp_assoc2id(asoc);
4122 status.sstat_state = asoc->state;
4123 status.sstat_rwnd = asoc->peer.rwnd;
4124 status.sstat_unackdata = asoc->unack_data;
4126 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4127 status.sstat_instrms = asoc->c.sinit_max_instreams;
4128 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4129 status.sstat_fragmentation_point = asoc->frag_point;
4130 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4131 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4132 transport->af_specific->sockaddr_len);
4133 /* Map ipv4 address into v4-mapped-on-v6 address. */
4134 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4135 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4136 status.sstat_primary.spinfo_state = transport->state;
4137 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4138 status.sstat_primary.spinfo_srtt = transport->srtt;
4139 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4140 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4142 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4143 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4145 if (put_user(len, optlen)) {
4150 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
4151 len, status.sstat_state, status.sstat_rwnd,
4152 status.sstat_assoc_id);
4154 if (copy_to_user(optval, &status, len)) {
4164 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4166 * Applications can retrieve information about a specific peer address
4167 * of an association, including its reachability state, congestion
4168 * window, and retransmission timer values. This information is
4171 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4172 char __user *optval,
4175 struct sctp_paddrinfo pinfo;
4176 struct sctp_transport *transport;
4179 if (len < sizeof(pinfo)) {
4184 len = sizeof(pinfo);
4185 if (copy_from_user(&pinfo, optval, len)) {
4190 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4191 pinfo.spinfo_assoc_id);
4195 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4196 pinfo.spinfo_state = transport->state;
4197 pinfo.spinfo_cwnd = transport->cwnd;
4198 pinfo.spinfo_srtt = transport->srtt;
4199 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4200 pinfo.spinfo_mtu = transport->pathmtu;
4202 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4203 pinfo.spinfo_state = SCTP_ACTIVE;
4205 if (put_user(len, optlen)) {
4210 if (copy_to_user(optval, &pinfo, len)) {
4219 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4221 * This option is a on/off flag. If enabled no SCTP message
4222 * fragmentation will be performed. Instead if a message being sent
4223 * exceeds the current PMTU size, the message will NOT be sent and
4224 * instead a error will be indicated to the user.
4226 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4227 char __user *optval, int __user *optlen)
4231 if (len < sizeof(int))
4235 val = (sctp_sk(sk)->disable_fragments == 1);
4236 if (put_user(len, optlen))
4238 if (copy_to_user(optval, &val, len))
4243 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4245 * This socket option is used to specify various notifications and
4246 * ancillary data the user wishes to receive.
4248 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4253 if (len > sizeof(struct sctp_event_subscribe))
4254 len = sizeof(struct sctp_event_subscribe);
4255 if (put_user(len, optlen))
4257 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4262 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4264 * This socket option is applicable to the UDP-style socket only. When
4265 * set it will cause associations that are idle for more than the
4266 * specified number of seconds to automatically close. An association
4267 * being idle is defined an association that has NOT sent or received
4268 * user data. The special value of '0' indicates that no automatic
4269 * close of any associations should be performed. The option expects an
4270 * integer defining the number of seconds of idle time before an
4271 * association is closed.
4273 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4275 /* Applicable to UDP-style socket only */
4276 if (sctp_style(sk, TCP))
4278 if (len < sizeof(int))
4281 if (put_user(len, optlen))
4283 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4288 /* Helper routine to branch off an association to a new socket. */
4289 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4291 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4292 struct socket *sock;
4299 /* An association cannot be branched off from an already peeled-off
4300 * socket, nor is this supported for tcp style sockets.
4302 if (!sctp_style(sk, UDP))
4305 /* Create a new socket. */
4306 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4310 sctp_copy_sock(sock->sk, sk, asoc);
4312 /* Make peeled-off sockets more like 1-1 accepted sockets.
4313 * Set the daddr and initialize id to something more random
4315 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4316 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4318 /* Populate the fields of the newsk from the oldsk and migrate the
4319 * asoc to the newsk.
4321 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4327 EXPORT_SYMBOL(sctp_do_peeloff);
4329 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4331 sctp_peeloff_arg_t peeloff;
4332 struct socket *newsock;
4333 struct file *newfile;
4336 if (len < sizeof(sctp_peeloff_arg_t))
4338 len = sizeof(sctp_peeloff_arg_t);
4339 if (copy_from_user(&peeloff, optval, len))
4342 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4346 /* Map the socket to an unused fd that can be returned to the user. */
4347 retval = get_unused_fd();
4349 sock_release(newsock);
4353 newfile = sock_alloc_file(newsock, 0, NULL);
4354 if (unlikely(IS_ERR(newfile))) {
4355 put_unused_fd(retval);
4356 sock_release(newsock);
4357 return PTR_ERR(newfile);
4360 SCTP_DEBUG_PRINTK("%s: sk: %p newsk: %p sd: %d\n",
4361 __func__, sk, newsock->sk, retval);
4363 /* Return the fd mapped to the new socket. */
4364 if (put_user(len, optlen)) {
4366 put_unused_fd(retval);
4369 peeloff.sd = retval;
4370 if (copy_to_user(optval, &peeloff, len)) {
4372 put_unused_fd(retval);
4375 fd_install(retval, newfile);
4380 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4382 * Applications can enable or disable heartbeats for any peer address of
4383 * an association, modify an address's heartbeat interval, force a
4384 * heartbeat to be sent immediately, and adjust the address's maximum
4385 * number of retransmissions sent before an address is considered
4386 * unreachable. The following structure is used to access and modify an
4387 * address's parameters:
4389 * struct sctp_paddrparams {
4390 * sctp_assoc_t spp_assoc_id;
4391 * struct sockaddr_storage spp_address;
4392 * uint32_t spp_hbinterval;
4393 * uint16_t spp_pathmaxrxt;
4394 * uint32_t spp_pathmtu;
4395 * uint32_t spp_sackdelay;
4396 * uint32_t spp_flags;
4399 * spp_assoc_id - (one-to-many style socket) This is filled in the
4400 * application, and identifies the association for
4402 * spp_address - This specifies which address is of interest.
4403 * spp_hbinterval - This contains the value of the heartbeat interval,
4404 * in milliseconds. If a value of zero
4405 * is present in this field then no changes are to
4406 * be made to this parameter.
4407 * spp_pathmaxrxt - This contains the maximum number of
4408 * retransmissions before this address shall be
4409 * considered unreachable. If a value of zero
4410 * is present in this field then no changes are to
4411 * be made to this parameter.
4412 * spp_pathmtu - When Path MTU discovery is disabled the value
4413 * specified here will be the "fixed" path mtu.
4414 * Note that if the spp_address field is empty
4415 * then all associations on this address will
4416 * have this fixed path mtu set upon them.
4418 * spp_sackdelay - When delayed sack is enabled, this value specifies
4419 * the number of milliseconds that sacks will be delayed
4420 * for. This value will apply to all addresses of an
4421 * association if the spp_address field is empty. Note
4422 * also, that if delayed sack is enabled and this
4423 * value is set to 0, no change is made to the last
4424 * recorded delayed sack timer value.
4426 * spp_flags - These flags are used to control various features
4427 * on an association. The flag field may contain
4428 * zero or more of the following options.
4430 * SPP_HB_ENABLE - Enable heartbeats on the
4431 * specified address. Note that if the address
4432 * field is empty all addresses for the association
4433 * have heartbeats enabled upon them.
4435 * SPP_HB_DISABLE - Disable heartbeats on the
4436 * speicifed address. Note that if the address
4437 * field is empty all addresses for the association
4438 * will have their heartbeats disabled. Note also
4439 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4440 * mutually exclusive, only one of these two should
4441 * be specified. Enabling both fields will have
4442 * undetermined results.
4444 * SPP_HB_DEMAND - Request a user initiated heartbeat
4445 * to be made immediately.
4447 * SPP_PMTUD_ENABLE - This field will enable PMTU
4448 * discovery upon the specified address. Note that
4449 * if the address feild is empty then all addresses
4450 * on the association are effected.
4452 * SPP_PMTUD_DISABLE - This field will disable PMTU
4453 * discovery upon the specified address. Note that
4454 * if the address feild is empty then all addresses
4455 * on the association are effected. Not also that
4456 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4457 * exclusive. Enabling both will have undetermined
4460 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4461 * on delayed sack. The time specified in spp_sackdelay
4462 * is used to specify the sack delay for this address. Note
4463 * that if spp_address is empty then all addresses will
4464 * enable delayed sack and take on the sack delay
4465 * value specified in spp_sackdelay.
4466 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4467 * off delayed sack. If the spp_address field is blank then
4468 * delayed sack is disabled for the entire association. Note
4469 * also that this field is mutually exclusive to
4470 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4473 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4474 char __user *optval, int __user *optlen)
4476 struct sctp_paddrparams params;
4477 struct sctp_transport *trans = NULL;
4478 struct sctp_association *asoc = NULL;
4479 struct sctp_sock *sp = sctp_sk(sk);
4481 if (len < sizeof(struct sctp_paddrparams))
4483 len = sizeof(struct sctp_paddrparams);
4484 if (copy_from_user(¶ms, optval, len))
4487 /* If an address other than INADDR_ANY is specified, and
4488 * no transport is found, then the request is invalid.
4490 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
4491 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4492 params.spp_assoc_id);
4494 SCTP_DEBUG_PRINTK("Failed no transport\n");
4499 /* Get association, if assoc_id != 0 and the socket is a one
4500 * to many style socket, and an association was not found, then
4501 * the id was invalid.
4503 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4504 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4505 SCTP_DEBUG_PRINTK("Failed no association\n");
4510 /* Fetch transport values. */
4511 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4512 params.spp_pathmtu = trans->pathmtu;
4513 params.spp_pathmaxrxt = trans->pathmaxrxt;
4514 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4516 /*draft-11 doesn't say what to return in spp_flags*/
4517 params.spp_flags = trans->param_flags;
4519 /* Fetch association values. */
4520 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4521 params.spp_pathmtu = asoc->pathmtu;
4522 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4523 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4525 /*draft-11 doesn't say what to return in spp_flags*/
4526 params.spp_flags = asoc->param_flags;
4528 /* Fetch socket values. */
4529 params.spp_hbinterval = sp->hbinterval;
4530 params.spp_pathmtu = sp->pathmtu;
4531 params.spp_sackdelay = sp->sackdelay;
4532 params.spp_pathmaxrxt = sp->pathmaxrxt;
4534 /*draft-11 doesn't say what to return in spp_flags*/
4535 params.spp_flags = sp->param_flags;
4538 if (copy_to_user(optval, ¶ms, len))
4541 if (put_user(len, optlen))
4548 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4550 * This option will effect the way delayed acks are performed. This
4551 * option allows you to get or set the delayed ack time, in
4552 * milliseconds. It also allows changing the delayed ack frequency.
4553 * Changing the frequency to 1 disables the delayed sack algorithm. If
4554 * the assoc_id is 0, then this sets or gets the endpoints default
4555 * values. If the assoc_id field is non-zero, then the set or get
4556 * effects the specified association for the one to many model (the
4557 * assoc_id field is ignored by the one to one model). Note that if
4558 * sack_delay or sack_freq are 0 when setting this option, then the
4559 * current values will remain unchanged.
4561 * struct sctp_sack_info {
4562 * sctp_assoc_t sack_assoc_id;
4563 * uint32_t sack_delay;
4564 * uint32_t sack_freq;
4567 * sack_assoc_id - This parameter, indicates which association the user
4568 * is performing an action upon. Note that if this field's value is
4569 * zero then the endpoints default value is changed (effecting future
4570 * associations only).
4572 * sack_delay - This parameter contains the number of milliseconds that
4573 * the user is requesting the delayed ACK timer be set to. Note that
4574 * this value is defined in the standard to be between 200 and 500
4577 * sack_freq - This parameter contains the number of packets that must
4578 * be received before a sack is sent without waiting for the delay
4579 * timer to expire. The default value for this is 2, setting this
4580 * value to 1 will disable the delayed sack algorithm.
4582 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4583 char __user *optval,
4586 struct sctp_sack_info params;
4587 struct sctp_association *asoc = NULL;
4588 struct sctp_sock *sp = sctp_sk(sk);
4590 if (len >= sizeof(struct sctp_sack_info)) {
4591 len = sizeof(struct sctp_sack_info);
4593 if (copy_from_user(¶ms, optval, len))
4595 } else if (len == sizeof(struct sctp_assoc_value)) {
4596 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4597 pr_warn("Use struct sctp_sack_info instead\n");
4598 if (copy_from_user(¶ms, optval, len))
4603 /* Get association, if sack_assoc_id != 0 and the socket is a one
4604 * to many style socket, and an association was not found, then
4605 * the id was invalid.
4607 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4608 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4612 /* Fetch association values. */
4613 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4614 params.sack_delay = jiffies_to_msecs(
4616 params.sack_freq = asoc->sackfreq;
4619 params.sack_delay = 0;
4620 params.sack_freq = 1;
4623 /* Fetch socket values. */
4624 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4625 params.sack_delay = sp->sackdelay;
4626 params.sack_freq = sp->sackfreq;
4628 params.sack_delay = 0;
4629 params.sack_freq = 1;
4633 if (copy_to_user(optval, ¶ms, len))
4636 if (put_user(len, optlen))
4642 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4644 * Applications can specify protocol parameters for the default association
4645 * initialization. The option name argument to setsockopt() and getsockopt()
4648 * Setting initialization parameters is effective only on an unconnected
4649 * socket (for UDP-style sockets only future associations are effected
4650 * by the change). With TCP-style sockets, this option is inherited by
4651 * sockets derived from a listener socket.
4653 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4655 if (len < sizeof(struct sctp_initmsg))
4657 len = sizeof(struct sctp_initmsg);
4658 if (put_user(len, optlen))
4660 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4666 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4667 char __user *optval, int __user *optlen)
4669 struct sctp_association *asoc;
4671 struct sctp_getaddrs getaddrs;
4672 struct sctp_transport *from;
4674 union sctp_addr temp;
4675 struct sctp_sock *sp = sctp_sk(sk);
4680 if (len < sizeof(struct sctp_getaddrs))
4683 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4686 /* For UDP-style sockets, id specifies the association to query. */
4687 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4691 to = optval + offsetof(struct sctp_getaddrs,addrs);
4692 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4694 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4696 memcpy(&temp, &from->ipaddr, sizeof(temp));
4697 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4698 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4699 if (space_left < addrlen)
4701 if (copy_to_user(to, &temp, addrlen))
4705 space_left -= addrlen;
4708 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4710 bytes_copied = ((char __user *)to) - optval;
4711 if (put_user(bytes_copied, optlen))
4717 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4718 size_t space_left, int *bytes_copied)
4720 struct sctp_sockaddr_entry *addr;
4721 union sctp_addr temp;
4724 struct net *net = sock_net(sk);
4727 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4731 if ((PF_INET == sk->sk_family) &&
4732 (AF_INET6 == addr->a.sa.sa_family))
4734 if ((PF_INET6 == sk->sk_family) &&
4735 inet_v6_ipv6only(sk) &&
4736 (AF_INET == addr->a.sa.sa_family))
4738 memcpy(&temp, &addr->a, sizeof(temp));
4739 if (!temp.v4.sin_port)
4740 temp.v4.sin_port = htons(port);
4742 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4744 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4745 if (space_left < addrlen) {
4749 memcpy(to, &temp, addrlen);
4753 space_left -= addrlen;
4754 *bytes_copied += addrlen;
4762 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4763 char __user *optval, int __user *optlen)
4765 struct sctp_bind_addr *bp;
4766 struct sctp_association *asoc;
4768 struct sctp_getaddrs getaddrs;
4769 struct sctp_sockaddr_entry *addr;
4771 union sctp_addr temp;
4772 struct sctp_sock *sp = sctp_sk(sk);
4776 int bytes_copied = 0;
4780 if (len < sizeof(struct sctp_getaddrs))
4783 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4787 * For UDP-style sockets, id specifies the association to query.
4788 * If the id field is set to the value '0' then the locally bound
4789 * addresses are returned without regard to any particular
4792 if (0 == getaddrs.assoc_id) {
4793 bp = &sctp_sk(sk)->ep->base.bind_addr;
4795 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4798 bp = &asoc->base.bind_addr;
4801 to = optval + offsetof(struct sctp_getaddrs,addrs);
4802 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4804 addrs = kmalloc(space_left, GFP_KERNEL);
4808 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4809 * addresses from the global local address list.
4811 if (sctp_list_single_entry(&bp->address_list)) {
4812 addr = list_entry(bp->address_list.next,
4813 struct sctp_sockaddr_entry, list);
4814 if (sctp_is_any(sk, &addr->a)) {
4815 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4816 space_left, &bytes_copied);
4826 /* Protection on the bound address list is not needed since
4827 * in the socket option context we hold a socket lock and
4828 * thus the bound address list can't change.
4830 list_for_each_entry(addr, &bp->address_list, list) {
4831 memcpy(&temp, &addr->a, sizeof(temp));
4832 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4833 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4834 if (space_left < addrlen) {
4835 err = -ENOMEM; /*fixme: right error?*/
4838 memcpy(buf, &temp, addrlen);
4840 bytes_copied += addrlen;
4842 space_left -= addrlen;
4846 if (copy_to_user(to, addrs, bytes_copied)) {
4850 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4854 if (put_user(bytes_copied, optlen))
4861 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4863 * Requests that the local SCTP stack use the enclosed peer address as
4864 * the association primary. The enclosed address must be one of the
4865 * association peer's addresses.
4867 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4868 char __user *optval, int __user *optlen)
4870 struct sctp_prim prim;
4871 struct sctp_association *asoc;
4872 struct sctp_sock *sp = sctp_sk(sk);
4874 if (len < sizeof(struct sctp_prim))
4877 len = sizeof(struct sctp_prim);
4879 if (copy_from_user(&prim, optval, len))
4882 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4886 if (!asoc->peer.primary_path)
4889 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4890 asoc->peer.primary_path->af_specific->sockaddr_len);
4892 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4893 (union sctp_addr *)&prim.ssp_addr);
4895 if (put_user(len, optlen))
4897 if (copy_to_user(optval, &prim, len))
4904 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4906 * Requests that the local endpoint set the specified Adaptation Layer
4907 * Indication parameter for all future INIT and INIT-ACK exchanges.
4909 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4910 char __user *optval, int __user *optlen)
4912 struct sctp_setadaptation adaptation;
4914 if (len < sizeof(struct sctp_setadaptation))
4917 len = sizeof(struct sctp_setadaptation);
4919 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4921 if (put_user(len, optlen))
4923 if (copy_to_user(optval, &adaptation, len))
4931 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4933 * Applications that wish to use the sendto() system call may wish to
4934 * specify a default set of parameters that would normally be supplied
4935 * through the inclusion of ancillary data. This socket option allows
4936 * such an application to set the default sctp_sndrcvinfo structure.
4939 * The application that wishes to use this socket option simply passes
4940 * in to this call the sctp_sndrcvinfo structure defined in Section
4941 * 5.2.2) The input parameters accepted by this call include
4942 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4943 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4944 * to this call if the caller is using the UDP model.
4946 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4948 static int sctp_getsockopt_default_send_param(struct sock *sk,
4949 int len, char __user *optval,
4952 struct sctp_sndrcvinfo info;
4953 struct sctp_association *asoc;
4954 struct sctp_sock *sp = sctp_sk(sk);
4956 if (len < sizeof(struct sctp_sndrcvinfo))
4959 len = sizeof(struct sctp_sndrcvinfo);
4961 if (copy_from_user(&info, optval, len))
4964 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4965 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4969 info.sinfo_stream = asoc->default_stream;
4970 info.sinfo_flags = asoc->default_flags;
4971 info.sinfo_ppid = asoc->default_ppid;
4972 info.sinfo_context = asoc->default_context;
4973 info.sinfo_timetolive = asoc->default_timetolive;
4975 info.sinfo_stream = sp->default_stream;
4976 info.sinfo_flags = sp->default_flags;
4977 info.sinfo_ppid = sp->default_ppid;
4978 info.sinfo_context = sp->default_context;
4979 info.sinfo_timetolive = sp->default_timetolive;
4982 if (put_user(len, optlen))
4984 if (copy_to_user(optval, &info, len))
4992 * 7.1.5 SCTP_NODELAY
4994 * Turn on/off any Nagle-like algorithm. This means that packets are
4995 * generally sent as soon as possible and no unnecessary delays are
4996 * introduced, at the cost of more packets in the network. Expects an
4997 * integer boolean flag.
5000 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5001 char __user *optval, int __user *optlen)
5005 if (len < sizeof(int))
5009 val = (sctp_sk(sk)->nodelay == 1);
5010 if (put_user(len, optlen))
5012 if (copy_to_user(optval, &val, len))
5019 * 7.1.1 SCTP_RTOINFO
5021 * The protocol parameters used to initialize and bound retransmission
5022 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5023 * and modify these parameters.
5024 * All parameters are time values, in milliseconds. A value of 0, when
5025 * modifying the parameters, indicates that the current value should not
5029 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5030 char __user *optval,
5031 int __user *optlen) {
5032 struct sctp_rtoinfo rtoinfo;
5033 struct sctp_association *asoc;
5035 if (len < sizeof (struct sctp_rtoinfo))
5038 len = sizeof(struct sctp_rtoinfo);
5040 if (copy_from_user(&rtoinfo, optval, len))
5043 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5045 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5048 /* Values corresponding to the specific association. */
5050 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5051 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5052 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5054 /* Values corresponding to the endpoint. */
5055 struct sctp_sock *sp = sctp_sk(sk);
5057 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5058 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5059 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5062 if (put_user(len, optlen))
5065 if (copy_to_user(optval, &rtoinfo, len))
5073 * 7.1.2 SCTP_ASSOCINFO
5075 * This option is used to tune the maximum retransmission attempts
5076 * of the association.
5077 * Returns an error if the new association retransmission value is
5078 * greater than the sum of the retransmission value of the peer.
5079 * See [SCTP] for more information.
5082 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5083 char __user *optval,
5087 struct sctp_assocparams assocparams;
5088 struct sctp_association *asoc;
5089 struct list_head *pos;
5092 if (len < sizeof (struct sctp_assocparams))
5095 len = sizeof(struct sctp_assocparams);
5097 if (copy_from_user(&assocparams, optval, len))
5100 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5102 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5105 /* Values correspoinding to the specific association */
5107 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5108 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5109 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5110 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
5112 (asoc->cookie_life.tv_usec
5115 list_for_each(pos, &asoc->peer.transport_addr_list) {
5119 assocparams.sasoc_number_peer_destinations = cnt;
5121 /* Values corresponding to the endpoint */
5122 struct sctp_sock *sp = sctp_sk(sk);
5124 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5125 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5126 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5127 assocparams.sasoc_cookie_life =
5128 sp->assocparams.sasoc_cookie_life;
5129 assocparams.sasoc_number_peer_destinations =
5131 sasoc_number_peer_destinations;
5134 if (put_user(len, optlen))
5137 if (copy_to_user(optval, &assocparams, len))
5144 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5146 * This socket option is a boolean flag which turns on or off mapped V4
5147 * addresses. If this option is turned on and the socket is type
5148 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5149 * If this option is turned off, then no mapping will be done of V4
5150 * addresses and a user will receive both PF_INET6 and PF_INET type
5151 * addresses on the socket.
5153 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5154 char __user *optval, int __user *optlen)
5157 struct sctp_sock *sp = sctp_sk(sk);
5159 if (len < sizeof(int))
5164 if (put_user(len, optlen))
5166 if (copy_to_user(optval, &val, len))
5173 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5174 * (chapter and verse is quoted at sctp_setsockopt_context())
5176 static int sctp_getsockopt_context(struct sock *sk, int len,
5177 char __user *optval, int __user *optlen)
5179 struct sctp_assoc_value params;
5180 struct sctp_sock *sp;
5181 struct sctp_association *asoc;
5183 if (len < sizeof(struct sctp_assoc_value))
5186 len = sizeof(struct sctp_assoc_value);
5188 if (copy_from_user(¶ms, optval, len))
5193 if (params.assoc_id != 0) {
5194 asoc = sctp_id2assoc(sk, params.assoc_id);
5197 params.assoc_value = asoc->default_rcv_context;
5199 params.assoc_value = sp->default_rcv_context;
5202 if (put_user(len, optlen))
5204 if (copy_to_user(optval, ¶ms, len))
5211 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5212 * This option will get or set the maximum size to put in any outgoing
5213 * SCTP DATA chunk. If a message is larger than this size it will be
5214 * fragmented by SCTP into the specified size. Note that the underlying
5215 * SCTP implementation may fragment into smaller sized chunks when the
5216 * PMTU of the underlying association is smaller than the value set by
5217 * the user. The default value for this option is '0' which indicates
5218 * the user is NOT limiting fragmentation and only the PMTU will effect
5219 * SCTP's choice of DATA chunk size. Note also that values set larger
5220 * than the maximum size of an IP datagram will effectively let SCTP
5221 * control fragmentation (i.e. the same as setting this option to 0).
5223 * The following structure is used to access and modify this parameter:
5225 * struct sctp_assoc_value {
5226 * sctp_assoc_t assoc_id;
5227 * uint32_t assoc_value;
5230 * assoc_id: This parameter is ignored for one-to-one style sockets.
5231 * For one-to-many style sockets this parameter indicates which
5232 * association the user is performing an action upon. Note that if
5233 * this field's value is zero then the endpoints default value is
5234 * changed (effecting future associations only).
5235 * assoc_value: This parameter specifies the maximum size in bytes.
5237 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5238 char __user *optval, int __user *optlen)
5240 struct sctp_assoc_value params;
5241 struct sctp_association *asoc;
5243 if (len == sizeof(int)) {
5244 pr_warn("Use of int in maxseg socket option deprecated\n");
5245 pr_warn("Use struct sctp_assoc_value instead\n");
5246 params.assoc_id = 0;
5247 } else if (len >= sizeof(struct sctp_assoc_value)) {
5248 len = sizeof(struct sctp_assoc_value);
5249 if (copy_from_user(¶ms, optval, sizeof(params)))
5254 asoc = sctp_id2assoc(sk, params.assoc_id);
5255 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5259 params.assoc_value = asoc->frag_point;
5261 params.assoc_value = sctp_sk(sk)->user_frag;
5263 if (put_user(len, optlen))
5265 if (len == sizeof(int)) {
5266 if (copy_to_user(optval, ¶ms.assoc_value, len))
5269 if (copy_to_user(optval, ¶ms, len))
5277 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5278 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5280 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5281 char __user *optval, int __user *optlen)
5285 if (len < sizeof(int))
5290 val = sctp_sk(sk)->frag_interleave;
5291 if (put_user(len, optlen))
5293 if (copy_to_user(optval, &val, len))
5300 * 7.1.25. Set or Get the sctp partial delivery point
5301 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5303 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5304 char __user *optval,
5309 if (len < sizeof(u32))
5314 val = sctp_sk(sk)->pd_point;
5315 if (put_user(len, optlen))
5317 if (copy_to_user(optval, &val, len))
5324 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5325 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5327 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5328 char __user *optval,
5331 struct sctp_assoc_value params;
5332 struct sctp_sock *sp;
5333 struct sctp_association *asoc;
5335 if (len == sizeof(int)) {
5336 pr_warn("Use of int in max_burst socket option deprecated\n");
5337 pr_warn("Use struct sctp_assoc_value instead\n");
5338 params.assoc_id = 0;
5339 } else if (len >= sizeof(struct sctp_assoc_value)) {
5340 len = sizeof(struct sctp_assoc_value);
5341 if (copy_from_user(¶ms, optval, len))
5348 if (params.assoc_id != 0) {
5349 asoc = sctp_id2assoc(sk, params.assoc_id);
5352 params.assoc_value = asoc->max_burst;
5354 params.assoc_value = sp->max_burst;
5356 if (len == sizeof(int)) {
5357 if (copy_to_user(optval, ¶ms.assoc_value, len))
5360 if (copy_to_user(optval, ¶ms, len))
5368 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5369 char __user *optval, int __user *optlen)
5371 struct net *net = sock_net(sk);
5372 struct sctp_hmacalgo __user *p = (void __user *)optval;
5373 struct sctp_hmac_algo_param *hmacs;
5377 if (!net->sctp.auth_enable)
5380 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5381 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5383 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5386 len = sizeof(struct sctp_hmacalgo) + data_len;
5387 num_idents = data_len / sizeof(u16);
5389 if (put_user(len, optlen))
5391 if (put_user(num_idents, &p->shmac_num_idents))
5393 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5398 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5399 char __user *optval, int __user *optlen)
5401 struct net *net = sock_net(sk);
5402 struct sctp_authkeyid val;
5403 struct sctp_association *asoc;
5405 if (!net->sctp.auth_enable)
5408 if (len < sizeof(struct sctp_authkeyid))
5410 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5413 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5414 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5418 val.scact_keynumber = asoc->active_key_id;
5420 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5422 len = sizeof(struct sctp_authkeyid);
5423 if (put_user(len, optlen))
5425 if (copy_to_user(optval, &val, len))
5431 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5432 char __user *optval, int __user *optlen)
5434 struct net *net = sock_net(sk);
5435 struct sctp_authchunks __user *p = (void __user *)optval;
5436 struct sctp_authchunks val;
5437 struct sctp_association *asoc;
5438 struct sctp_chunks_param *ch;
5442 if (!net->sctp.auth_enable)
5445 if (len < sizeof(struct sctp_authchunks))
5448 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5451 to = p->gauth_chunks;
5452 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5456 ch = asoc->peer.peer_chunks;
5460 /* See if the user provided enough room for all the data */
5461 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5462 if (len < num_chunks)
5465 if (copy_to_user(to, ch->chunks, num_chunks))
5468 len = sizeof(struct sctp_authchunks) + num_chunks;
5469 if (put_user(len, optlen)) return -EFAULT;
5470 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5475 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5476 char __user *optval, int __user *optlen)
5478 struct net *net = sock_net(sk);
5479 struct sctp_authchunks __user *p = (void __user *)optval;
5480 struct sctp_authchunks val;
5481 struct sctp_association *asoc;
5482 struct sctp_chunks_param *ch;
5486 if (!net->sctp.auth_enable)
5489 if (len < sizeof(struct sctp_authchunks))
5492 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5495 to = p->gauth_chunks;
5496 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5497 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5501 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5503 ch = sctp_sk(sk)->ep->auth_chunk_list;
5508 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5509 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5512 if (copy_to_user(to, ch->chunks, num_chunks))
5515 len = sizeof(struct sctp_authchunks) + num_chunks;
5516 if (put_user(len, optlen))
5518 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5525 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5526 * This option gets the current number of associations that are attached
5527 * to a one-to-many style socket. The option value is an uint32_t.
5529 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5530 char __user *optval, int __user *optlen)
5532 struct sctp_sock *sp = sctp_sk(sk);
5533 struct sctp_association *asoc;
5536 if (sctp_style(sk, TCP))
5539 if (len < sizeof(u32))
5544 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5548 if (put_user(len, optlen))
5550 if (copy_to_user(optval, &val, len))
5557 * 8.1.23 SCTP_AUTO_ASCONF
5558 * See the corresponding setsockopt entry as description
5560 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5561 char __user *optval, int __user *optlen)
5565 if (len < sizeof(int))
5569 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5571 if (put_user(len, optlen))
5573 if (copy_to_user(optval, &val, len))
5579 * 8.2.6. Get the Current Identifiers of Associations
5580 * (SCTP_GET_ASSOC_ID_LIST)
5582 * This option gets the current list of SCTP association identifiers of
5583 * the SCTP associations handled by a one-to-many style socket.
5585 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5586 char __user *optval, int __user *optlen)
5588 struct sctp_sock *sp = sctp_sk(sk);
5589 struct sctp_association *asoc;
5590 struct sctp_assoc_ids *ids;
5593 if (sctp_style(sk, TCP))
5596 if (len < sizeof(struct sctp_assoc_ids))
5599 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5603 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5606 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5608 ids = kmalloc(len, GFP_KERNEL);
5612 ids->gaids_number_of_ids = num;
5614 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5615 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5618 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5628 * SCTP_PEER_ADDR_THLDS
5630 * This option allows us to fetch the partially failed threshold for one or all
5631 * transports in an association. See Section 6.1 of:
5632 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5634 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5635 char __user *optval,
5639 struct sctp_paddrthlds val;
5640 struct sctp_transport *trans;
5641 struct sctp_association *asoc;
5643 if (len < sizeof(struct sctp_paddrthlds))
5645 len = sizeof(struct sctp_paddrthlds);
5646 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5649 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5650 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5654 val.spt_pathpfthld = asoc->pf_retrans;
5655 val.spt_pathmaxrxt = asoc->pathmaxrxt;
5657 trans = sctp_addr_id2transport(sk, &val.spt_address,
5662 val.spt_pathmaxrxt = trans->pathmaxrxt;
5663 val.spt_pathpfthld = trans->pf_retrans;
5666 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5673 * SCTP_GET_ASSOC_STATS
5675 * This option retrieves local per endpoint statistics. It is modeled
5676 * after OpenSolaris' implementation
5678 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
5679 char __user *optval,
5682 struct sctp_assoc_stats sas;
5683 struct sctp_association *asoc = NULL;
5685 /* User must provide at least the assoc id */
5686 if (len < sizeof(sctp_assoc_t))
5689 /* Allow the struct to grow and fill in as much as possible */
5690 len = min_t(size_t, len, sizeof(sas));
5692 if (copy_from_user(&sas, optval, len))
5695 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
5699 sas.sas_rtxchunks = asoc->stats.rtxchunks;
5700 sas.sas_gapcnt = asoc->stats.gapcnt;
5701 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
5702 sas.sas_osacks = asoc->stats.osacks;
5703 sas.sas_isacks = asoc->stats.isacks;
5704 sas.sas_octrlchunks = asoc->stats.octrlchunks;
5705 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
5706 sas.sas_oodchunks = asoc->stats.oodchunks;
5707 sas.sas_iodchunks = asoc->stats.iodchunks;
5708 sas.sas_ouodchunks = asoc->stats.ouodchunks;
5709 sas.sas_iuodchunks = asoc->stats.iuodchunks;
5710 sas.sas_idupchunks = asoc->stats.idupchunks;
5711 sas.sas_opackets = asoc->stats.opackets;
5712 sas.sas_ipackets = asoc->stats.ipackets;
5714 /* New high max rto observed, will return 0 if not a single
5715 * RTO update took place. obs_rto_ipaddr will be bogus
5718 sas.sas_maxrto = asoc->stats.max_obs_rto;
5719 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
5720 sizeof(struct sockaddr_storage));
5722 /* Mark beginning of a new observation period */
5723 asoc->stats.max_obs_rto = asoc->rto_min;
5725 if (put_user(len, optlen))
5728 SCTP_DEBUG_PRINTK("sctp_getsockopt_assoc_stat(%d): %d\n",
5729 len, sas.sas_assoc_id);
5731 if (copy_to_user(optval, &sas, len))
5737 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5738 char __user *optval, int __user *optlen)
5743 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5746 /* I can hardly begin to describe how wrong this is. This is
5747 * so broken as to be worse than useless. The API draft
5748 * REALLY is NOT helpful here... I am not convinced that the
5749 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5750 * are at all well-founded.
5752 if (level != SOL_SCTP) {
5753 struct sctp_af *af = sctp_sk(sk)->pf->af;
5755 retval = af->getsockopt(sk, level, optname, optval, optlen);
5759 if (get_user(len, optlen))
5766 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5768 case SCTP_DISABLE_FRAGMENTS:
5769 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5773 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5775 case SCTP_AUTOCLOSE:
5776 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5778 case SCTP_SOCKOPT_PEELOFF:
5779 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5781 case SCTP_PEER_ADDR_PARAMS:
5782 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5785 case SCTP_DELAYED_SACK:
5786 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5790 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5792 case SCTP_GET_PEER_ADDRS:
5793 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5796 case SCTP_GET_LOCAL_ADDRS:
5797 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5800 case SCTP_SOCKOPT_CONNECTX3:
5801 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5803 case SCTP_DEFAULT_SEND_PARAM:
5804 retval = sctp_getsockopt_default_send_param(sk, len,
5807 case SCTP_PRIMARY_ADDR:
5808 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5811 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5814 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5816 case SCTP_ASSOCINFO:
5817 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5819 case SCTP_I_WANT_MAPPED_V4_ADDR:
5820 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5823 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5825 case SCTP_GET_PEER_ADDR_INFO:
5826 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5829 case SCTP_ADAPTATION_LAYER:
5830 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5834 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5836 case SCTP_FRAGMENT_INTERLEAVE:
5837 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5840 case SCTP_PARTIAL_DELIVERY_POINT:
5841 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5844 case SCTP_MAX_BURST:
5845 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5848 case SCTP_AUTH_CHUNK:
5849 case SCTP_AUTH_DELETE_KEY:
5850 retval = -EOPNOTSUPP;
5852 case SCTP_HMAC_IDENT:
5853 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5855 case SCTP_AUTH_ACTIVE_KEY:
5856 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5858 case SCTP_PEER_AUTH_CHUNKS:
5859 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5862 case SCTP_LOCAL_AUTH_CHUNKS:
5863 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5866 case SCTP_GET_ASSOC_NUMBER:
5867 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5869 case SCTP_GET_ASSOC_ID_LIST:
5870 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5872 case SCTP_AUTO_ASCONF:
5873 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5875 case SCTP_PEER_ADDR_THLDS:
5876 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
5878 case SCTP_GET_ASSOC_STATS:
5879 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
5882 retval = -ENOPROTOOPT;
5886 sctp_release_sock(sk);
5890 static void sctp_hash(struct sock *sk)
5895 static void sctp_unhash(struct sock *sk)
5900 /* Check if port is acceptable. Possibly find first available port.
5902 * The port hash table (contained in the 'global' SCTP protocol storage
5903 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5904 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5905 * list (the list number is the port number hashed out, so as you
5906 * would expect from a hash function, all the ports in a given list have
5907 * such a number that hashes out to the same list number; you were
5908 * expecting that, right?); so each list has a set of ports, with a
5909 * link to the socket (struct sock) that uses it, the port number and
5910 * a fastreuse flag (FIXME: NPI ipg).
5912 static struct sctp_bind_bucket *sctp_bucket_create(
5913 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
5915 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5917 struct sctp_bind_hashbucket *head; /* hash list */
5918 struct sctp_bind_bucket *pp;
5919 unsigned short snum;
5922 snum = ntohs(addr->v4.sin_port);
5924 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5925 sctp_local_bh_disable();
5928 /* Search for an available port. */
5929 int low, high, remaining, index;
5932 inet_get_local_port_range(&low, &high);
5933 remaining = (high - low) + 1;
5934 rover = net_random() % remaining + low;
5938 if ((rover < low) || (rover > high))
5940 if (inet_is_reserved_local_port(rover))
5942 index = sctp_phashfn(sock_net(sk), rover);
5943 head = &sctp_port_hashtable[index];
5944 sctp_spin_lock(&head->lock);
5945 sctp_for_each_hentry(pp, &head->chain)
5946 if ((pp->port == rover) &&
5947 net_eq(sock_net(sk), pp->net))
5951 sctp_spin_unlock(&head->lock);
5952 } while (--remaining > 0);
5954 /* Exhausted local port range during search? */
5959 /* OK, here is the one we will use. HEAD (the port
5960 * hash table list entry) is non-NULL and we hold it's
5965 /* We are given an specific port number; we verify
5966 * that it is not being used. If it is used, we will
5967 * exahust the search in the hash list corresponding
5968 * to the port number (snum) - we detect that with the
5969 * port iterator, pp being NULL.
5971 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
5972 sctp_spin_lock(&head->lock);
5973 sctp_for_each_hentry(pp, &head->chain) {
5974 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
5981 if (!hlist_empty(&pp->owner)) {
5982 /* We had a port hash table hit - there is an
5983 * available port (pp != NULL) and it is being
5984 * used by other socket (pp->owner not empty); that other
5985 * socket is going to be sk2.
5987 int reuse = sk->sk_reuse;
5990 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5991 if (pp->fastreuse && sk->sk_reuse &&
5992 sk->sk_state != SCTP_SS_LISTENING)
5995 /* Run through the list of sockets bound to the port
5996 * (pp->port) [via the pointers bind_next and
5997 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5998 * we get the endpoint they describe and run through
5999 * the endpoint's list of IP (v4 or v6) addresses,
6000 * comparing each of the addresses with the address of
6001 * the socket sk. If we find a match, then that means
6002 * that this port/socket (sk) combination are already
6005 sk_for_each_bound(sk2, &pp->owner) {
6006 struct sctp_endpoint *ep2;
6007 ep2 = sctp_sk(sk2)->ep;
6010 (reuse && sk2->sk_reuse &&
6011 sk2->sk_state != SCTP_SS_LISTENING))
6014 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
6015 sctp_sk(sk2), sctp_sk(sk))) {
6020 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
6023 /* If there was a hash table miss, create a new port. */
6025 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6028 /* In either case (hit or miss), make sure fastreuse is 1 only
6029 * if sk->sk_reuse is too (that is, if the caller requested
6030 * SO_REUSEADDR on this socket -sk-).
6032 if (hlist_empty(&pp->owner)) {
6033 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6037 } else if (pp->fastreuse &&
6038 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6041 /* We are set, so fill up all the data in the hash table
6042 * entry, tie the socket list information with the rest of the
6043 * sockets FIXME: Blurry, NPI (ipg).
6046 if (!sctp_sk(sk)->bind_hash) {
6047 inet_sk(sk)->inet_num = snum;
6048 sk_add_bind_node(sk, &pp->owner);
6049 sctp_sk(sk)->bind_hash = pp;
6054 sctp_spin_unlock(&head->lock);
6057 sctp_local_bh_enable();
6061 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6062 * port is requested.
6064 static int sctp_get_port(struct sock *sk, unsigned short snum)
6067 union sctp_addr addr;
6068 struct sctp_af *af = sctp_sk(sk)->pf->af;
6070 /* Set up a dummy address struct from the sk. */
6071 af->from_sk(&addr, sk);
6072 addr.v4.sin_port = htons(snum);
6074 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6075 ret = sctp_get_port_local(sk, &addr);
6081 * Move a socket to LISTENING state.
6083 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
6085 struct sctp_sock *sp = sctp_sk(sk);
6086 struct sctp_endpoint *ep = sp->ep;
6087 struct crypto_hash *tfm = NULL;
6090 /* Allocate HMAC for generating cookie. */
6091 if (!sp->hmac && sp->sctp_hmac_alg) {
6092 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6093 tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
6095 net_info_ratelimited("failed to load transform for %s: %ld\n",
6096 sp->sctp_hmac_alg, PTR_ERR(tfm));
6099 sctp_sk(sk)->hmac = tfm;
6103 * If a bind() or sctp_bindx() is not called prior to a listen()
6104 * call that allows new associations to be accepted, the system
6105 * picks an ephemeral port and will choose an address set equivalent
6106 * to binding with a wildcard address.
6108 * This is not currently spelled out in the SCTP sockets
6109 * extensions draft, but follows the practice as seen in TCP
6113 sk->sk_state = SCTP_SS_LISTENING;
6114 if (!ep->base.bind_addr.port) {
6115 if (sctp_autobind(sk))
6118 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6119 sk->sk_state = SCTP_SS_CLOSED;
6124 sk->sk_max_ack_backlog = backlog;
6125 sctp_hash_endpoint(ep);
6130 * 4.1.3 / 5.1.3 listen()
6132 * By default, new associations are not accepted for UDP style sockets.
6133 * An application uses listen() to mark a socket as being able to
6134 * accept new associations.
6136 * On TCP style sockets, applications use listen() to ready the SCTP
6137 * endpoint for accepting inbound associations.
6139 * On both types of endpoints a backlog of '0' disables listening.
6141 * Move a socket to LISTENING state.
6143 int sctp_inet_listen(struct socket *sock, int backlog)
6145 struct sock *sk = sock->sk;
6146 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6149 if (unlikely(backlog < 0))
6154 /* Peeled-off sockets are not allowed to listen(). */
6155 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6158 if (sock->state != SS_UNCONNECTED)
6161 /* If backlog is zero, disable listening. */
6163 if (sctp_sstate(sk, CLOSED))
6167 sctp_unhash_endpoint(ep);
6168 sk->sk_state = SCTP_SS_CLOSED;
6170 sctp_sk(sk)->bind_hash->fastreuse = 1;
6174 /* If we are already listening, just update the backlog */
6175 if (sctp_sstate(sk, LISTENING))
6176 sk->sk_max_ack_backlog = backlog;
6178 err = sctp_listen_start(sk, backlog);
6185 sctp_release_sock(sk);
6190 * This function is done by modeling the current datagram_poll() and the
6191 * tcp_poll(). Note that, based on these implementations, we don't
6192 * lock the socket in this function, even though it seems that,
6193 * ideally, locking or some other mechanisms can be used to ensure
6194 * the integrity of the counters (sndbuf and wmem_alloc) used
6195 * in this place. We assume that we don't need locks either until proven
6198 * Another thing to note is that we include the Async I/O support
6199 * here, again, by modeling the current TCP/UDP code. We don't have
6200 * a good way to test with it yet.
6202 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6204 struct sock *sk = sock->sk;
6205 struct sctp_sock *sp = sctp_sk(sk);
6208 poll_wait(file, sk_sleep(sk), wait);
6210 /* A TCP-style listening socket becomes readable when the accept queue
6213 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6214 return (!list_empty(&sp->ep->asocs)) ?
6215 (POLLIN | POLLRDNORM) : 0;
6219 /* Is there any exceptional events? */
6220 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6222 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6223 if (sk->sk_shutdown & RCV_SHUTDOWN)
6224 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6225 if (sk->sk_shutdown == SHUTDOWN_MASK)
6228 /* Is it readable? Reconsider this code with TCP-style support. */
6229 if (!skb_queue_empty(&sk->sk_receive_queue))
6230 mask |= POLLIN | POLLRDNORM;
6232 /* The association is either gone or not ready. */
6233 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6236 /* Is it writable? */
6237 if (sctp_writeable(sk)) {
6238 mask |= POLLOUT | POLLWRNORM;
6240 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6242 * Since the socket is not locked, the buffer
6243 * might be made available after the writeable check and
6244 * before the bit is set. This could cause a lost I/O
6245 * signal. tcp_poll() has a race breaker for this race
6246 * condition. Based on their implementation, we put
6247 * in the following code to cover it as well.
6249 if (sctp_writeable(sk))
6250 mask |= POLLOUT | POLLWRNORM;
6255 /********************************************************************
6256 * 2nd Level Abstractions
6257 ********************************************************************/
6259 static struct sctp_bind_bucket *sctp_bucket_create(
6260 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6262 struct sctp_bind_bucket *pp;
6264 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6266 SCTP_DBG_OBJCNT_INC(bind_bucket);
6269 INIT_HLIST_HEAD(&pp->owner);
6271 hlist_add_head(&pp->node, &head->chain);
6276 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6277 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6279 if (pp && hlist_empty(&pp->owner)) {
6280 __hlist_del(&pp->node);
6281 kmem_cache_free(sctp_bucket_cachep, pp);
6282 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6286 /* Release this socket's reference to a local port. */
6287 static inline void __sctp_put_port(struct sock *sk)
6289 struct sctp_bind_hashbucket *head =
6290 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6291 inet_sk(sk)->inet_num)];
6292 struct sctp_bind_bucket *pp;
6294 sctp_spin_lock(&head->lock);
6295 pp = sctp_sk(sk)->bind_hash;
6296 __sk_del_bind_node(sk);
6297 sctp_sk(sk)->bind_hash = NULL;
6298 inet_sk(sk)->inet_num = 0;
6299 sctp_bucket_destroy(pp);
6300 sctp_spin_unlock(&head->lock);
6303 void sctp_put_port(struct sock *sk)
6305 sctp_local_bh_disable();
6306 __sctp_put_port(sk);
6307 sctp_local_bh_enable();
6311 * The system picks an ephemeral port and choose an address set equivalent
6312 * to binding with a wildcard address.
6313 * One of those addresses will be the primary address for the association.
6314 * This automatically enables the multihoming capability of SCTP.
6316 static int sctp_autobind(struct sock *sk)
6318 union sctp_addr autoaddr;
6322 /* Initialize a local sockaddr structure to INADDR_ANY. */
6323 af = sctp_sk(sk)->pf->af;
6325 port = htons(inet_sk(sk)->inet_num);
6326 af->inaddr_any(&autoaddr, port);
6328 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6331 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6334 * 4.2 The cmsghdr Structure *
6336 * When ancillary data is sent or received, any number of ancillary data
6337 * objects can be specified by the msg_control and msg_controllen members of
6338 * the msghdr structure, because each object is preceded by
6339 * a cmsghdr structure defining the object's length (the cmsg_len member).
6340 * Historically Berkeley-derived implementations have passed only one object
6341 * at a time, but this API allows multiple objects to be
6342 * passed in a single call to sendmsg() or recvmsg(). The following example
6343 * shows two ancillary data objects in a control buffer.
6345 * |<--------------------------- msg_controllen -------------------------->|
6348 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6350 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6353 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6355 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6358 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6359 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6361 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6363 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6370 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
6371 sctp_cmsgs_t *cmsgs)
6373 struct cmsghdr *cmsg;
6374 struct msghdr *my_msg = (struct msghdr *)msg;
6376 for (cmsg = CMSG_FIRSTHDR(msg);
6378 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6379 if (!CMSG_OK(my_msg, cmsg))
6382 /* Should we parse this header or ignore? */
6383 if (cmsg->cmsg_level != IPPROTO_SCTP)
6386 /* Strictly check lengths following example in SCM code. */
6387 switch (cmsg->cmsg_type) {
6389 /* SCTP Socket API Extension
6390 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6392 * This cmsghdr structure provides information for
6393 * initializing new SCTP associations with sendmsg().
6394 * The SCTP_INITMSG socket option uses this same data
6395 * structure. This structure is not used for
6398 * cmsg_level cmsg_type cmsg_data[]
6399 * ------------ ------------ ----------------------
6400 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6402 if (cmsg->cmsg_len !=
6403 CMSG_LEN(sizeof(struct sctp_initmsg)))
6405 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6409 /* SCTP Socket API Extension
6410 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6412 * This cmsghdr structure specifies SCTP options for
6413 * sendmsg() and describes SCTP header information
6414 * about a received message through recvmsg().
6416 * cmsg_level cmsg_type cmsg_data[]
6417 * ------------ ------------ ----------------------
6418 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6420 if (cmsg->cmsg_len !=
6421 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6425 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6427 /* Minimally, validate the sinfo_flags. */
6428 if (cmsgs->info->sinfo_flags &
6429 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6430 SCTP_ABORT | SCTP_EOF))
6442 * Wait for a packet..
6443 * Note: This function is the same function as in core/datagram.c
6444 * with a few modifications to make lksctp work.
6446 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6451 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6453 /* Socket errors? */
6454 error = sock_error(sk);
6458 if (!skb_queue_empty(&sk->sk_receive_queue))
6461 /* Socket shut down? */
6462 if (sk->sk_shutdown & RCV_SHUTDOWN)
6465 /* Sequenced packets can come disconnected. If so we report the
6470 /* Is there a good reason to think that we may receive some data? */
6471 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6474 /* Handle signals. */
6475 if (signal_pending(current))
6478 /* Let another process have a go. Since we are going to sleep
6479 * anyway. Note: This may cause odd behaviors if the message
6480 * does not fit in the user's buffer, but this seems to be the
6481 * only way to honor MSG_DONTWAIT realistically.
6483 sctp_release_sock(sk);
6484 *timeo_p = schedule_timeout(*timeo_p);
6488 finish_wait(sk_sleep(sk), &wait);
6492 error = sock_intr_errno(*timeo_p);
6495 finish_wait(sk_sleep(sk), &wait);
6500 /* Receive a datagram.
6501 * Note: This is pretty much the same routine as in core/datagram.c
6502 * with a few changes to make lksctp work.
6504 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6505 int noblock, int *err)
6508 struct sk_buff *skb;
6511 timeo = sock_rcvtimeo(sk, noblock);
6513 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6514 timeo, MAX_SCHEDULE_TIMEOUT);
6517 /* Again only user level code calls this function,
6518 * so nothing interrupt level
6519 * will suddenly eat the receive_queue.
6521 * Look at current nfs client by the way...
6522 * However, this function was correct in any case. 8)
6524 if (flags & MSG_PEEK) {
6525 spin_lock_bh(&sk->sk_receive_queue.lock);
6526 skb = skb_peek(&sk->sk_receive_queue);
6528 atomic_inc(&skb->users);
6529 spin_unlock_bh(&sk->sk_receive_queue.lock);
6531 skb = skb_dequeue(&sk->sk_receive_queue);
6537 /* Caller is allowed not to check sk->sk_err before calling. */
6538 error = sock_error(sk);
6542 if (sk->sk_shutdown & RCV_SHUTDOWN)
6545 /* User doesn't want to wait. */
6549 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6558 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6559 static void __sctp_write_space(struct sctp_association *asoc)
6561 struct sock *sk = asoc->base.sk;
6562 struct socket *sock = sk->sk_socket;
6564 if ((sctp_wspace(asoc) > 0) && sock) {
6565 if (waitqueue_active(&asoc->wait))
6566 wake_up_interruptible(&asoc->wait);
6568 if (sctp_writeable(sk)) {
6569 wait_queue_head_t *wq = sk_sleep(sk);
6571 if (wq && waitqueue_active(wq))
6572 wake_up_interruptible(wq);
6574 /* Note that we try to include the Async I/O support
6575 * here by modeling from the current TCP/UDP code.
6576 * We have not tested with it yet.
6578 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6579 sock_wake_async(sock,
6580 SOCK_WAKE_SPACE, POLL_OUT);
6585 /* Do accounting for the sndbuf space.
6586 * Decrement the used sndbuf space of the corresponding association by the
6587 * data size which was just transmitted(freed).
6589 static void sctp_wfree(struct sk_buff *skb)
6591 struct sctp_association *asoc;
6592 struct sctp_chunk *chunk;
6595 /* Get the saved chunk pointer. */
6596 chunk = *((struct sctp_chunk **)(skb->cb));
6599 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6600 sizeof(struct sk_buff) +
6601 sizeof(struct sctp_chunk);
6603 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6606 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6608 sk->sk_wmem_queued -= skb->truesize;
6609 sk_mem_uncharge(sk, skb->truesize);
6612 __sctp_write_space(asoc);
6614 sctp_association_put(asoc);
6617 /* Do accounting for the receive space on the socket.
6618 * Accounting for the association is done in ulpevent.c
6619 * We set this as a destructor for the cloned data skbs so that
6620 * accounting is done at the correct time.
6622 void sctp_sock_rfree(struct sk_buff *skb)
6624 struct sock *sk = skb->sk;
6625 struct sctp_ulpevent *event = sctp_skb2event(skb);
6627 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6630 * Mimic the behavior of sock_rfree
6632 sk_mem_uncharge(sk, event->rmem_len);
6636 /* Helper function to wait for space in the sndbuf. */
6637 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6640 struct sock *sk = asoc->base.sk;
6642 long current_timeo = *timeo_p;
6645 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6646 asoc, (long)(*timeo_p), msg_len);
6648 /* Increment the association's refcnt. */
6649 sctp_association_hold(asoc);
6651 /* Wait on the association specific sndbuf space. */
6653 prepare_to_wait_exclusive(&asoc->wait, &wait,
6654 TASK_INTERRUPTIBLE);
6657 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6660 if (signal_pending(current))
6661 goto do_interrupted;
6662 if (msg_len <= sctp_wspace(asoc))
6665 /* Let another process have a go. Since we are going
6668 sctp_release_sock(sk);
6669 current_timeo = schedule_timeout(current_timeo);
6670 BUG_ON(sk != asoc->base.sk);
6673 *timeo_p = current_timeo;
6677 finish_wait(&asoc->wait, &wait);
6679 /* Release the association's refcnt. */
6680 sctp_association_put(asoc);
6689 err = sock_intr_errno(*timeo_p);
6697 void sctp_data_ready(struct sock *sk, int len)
6699 struct socket_wq *wq;
6702 wq = rcu_dereference(sk->sk_wq);
6703 if (wq_has_sleeper(wq))
6704 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6705 POLLRDNORM | POLLRDBAND);
6706 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6710 /* If socket sndbuf has changed, wake up all per association waiters. */
6711 void sctp_write_space(struct sock *sk)
6713 struct sctp_association *asoc;
6715 /* Wake up the tasks in each wait queue. */
6716 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6717 __sctp_write_space(asoc);
6721 /* Is there any sndbuf space available on the socket?
6723 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6724 * associations on the same socket. For a UDP-style socket with
6725 * multiple associations, it is possible for it to be "unwriteable"
6726 * prematurely. I assume that this is acceptable because
6727 * a premature "unwriteable" is better than an accidental "writeable" which
6728 * would cause an unwanted block under certain circumstances. For the 1-1
6729 * UDP-style sockets or TCP-style sockets, this code should work.
6732 static int sctp_writeable(struct sock *sk)
6736 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6742 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6743 * returns immediately with EINPROGRESS.
6745 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6747 struct sock *sk = asoc->base.sk;
6749 long current_timeo = *timeo_p;
6752 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6755 /* Increment the association's refcnt. */
6756 sctp_association_hold(asoc);
6759 prepare_to_wait_exclusive(&asoc->wait, &wait,
6760 TASK_INTERRUPTIBLE);
6763 if (sk->sk_shutdown & RCV_SHUTDOWN)
6765 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6768 if (signal_pending(current))
6769 goto do_interrupted;
6771 if (sctp_state(asoc, ESTABLISHED))
6774 /* Let another process have a go. Since we are going
6777 sctp_release_sock(sk);
6778 current_timeo = schedule_timeout(current_timeo);
6781 *timeo_p = current_timeo;
6785 finish_wait(&asoc->wait, &wait);
6787 /* Release the association's refcnt. */
6788 sctp_association_put(asoc);
6793 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6796 err = -ECONNREFUSED;
6800 err = sock_intr_errno(*timeo_p);
6808 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6810 struct sctp_endpoint *ep;
6814 ep = sctp_sk(sk)->ep;
6818 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6819 TASK_INTERRUPTIBLE);
6821 if (list_empty(&ep->asocs)) {
6822 sctp_release_sock(sk);
6823 timeo = schedule_timeout(timeo);
6828 if (!sctp_sstate(sk, LISTENING))
6832 if (!list_empty(&ep->asocs))
6835 err = sock_intr_errno(timeo);
6836 if (signal_pending(current))
6844 finish_wait(sk_sleep(sk), &wait);
6849 static void sctp_wait_for_close(struct sock *sk, long timeout)
6854 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6855 if (list_empty(&sctp_sk(sk)->ep->asocs))
6857 sctp_release_sock(sk);
6858 timeout = schedule_timeout(timeout);
6860 } while (!signal_pending(current) && timeout);
6862 finish_wait(sk_sleep(sk), &wait);
6865 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6867 struct sk_buff *frag;
6872 /* Don't forget the fragments. */
6873 skb_walk_frags(skb, frag)
6874 sctp_skb_set_owner_r_frag(frag, sk);
6877 sctp_skb_set_owner_r(skb, sk);
6880 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6881 struct sctp_association *asoc)
6883 struct inet_sock *inet = inet_sk(sk);
6884 struct inet_sock *newinet;
6886 newsk->sk_type = sk->sk_type;
6887 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6888 newsk->sk_flags = sk->sk_flags;
6889 newsk->sk_no_check = sk->sk_no_check;
6890 newsk->sk_reuse = sk->sk_reuse;
6892 newsk->sk_shutdown = sk->sk_shutdown;
6893 newsk->sk_destruct = inet_sock_destruct;
6894 newsk->sk_family = sk->sk_family;
6895 newsk->sk_protocol = IPPROTO_SCTP;
6896 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6897 newsk->sk_sndbuf = sk->sk_sndbuf;
6898 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6899 newsk->sk_lingertime = sk->sk_lingertime;
6900 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6901 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6903 newinet = inet_sk(newsk);
6905 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6906 * getsockname() and getpeername()
6908 newinet->inet_sport = inet->inet_sport;
6909 newinet->inet_saddr = inet->inet_saddr;
6910 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6911 newinet->inet_dport = htons(asoc->peer.port);
6912 newinet->pmtudisc = inet->pmtudisc;
6913 newinet->inet_id = asoc->next_tsn ^ jiffies;
6915 newinet->uc_ttl = inet->uc_ttl;
6916 newinet->mc_loop = 1;
6917 newinet->mc_ttl = 1;
6918 newinet->mc_index = 0;
6919 newinet->mc_list = NULL;
6922 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6923 * and its messages to the newsk.
6925 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6926 struct sctp_association *assoc,
6927 sctp_socket_type_t type)
6929 struct sctp_sock *oldsp = sctp_sk(oldsk);
6930 struct sctp_sock *newsp = sctp_sk(newsk);
6931 struct sctp_bind_bucket *pp; /* hash list port iterator */
6932 struct sctp_endpoint *newep = newsp->ep;
6933 struct sk_buff *skb, *tmp;
6934 struct sctp_ulpevent *event;
6935 struct sctp_bind_hashbucket *head;
6936 struct list_head tmplist;
6938 /* Migrate socket buffer sizes and all the socket level options to the
6941 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6942 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6943 /* Brute force copy old sctp opt. */
6944 if (oldsp->do_auto_asconf) {
6945 memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
6946 inet_sk_copy_descendant(newsk, oldsk);
6947 memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
6949 inet_sk_copy_descendant(newsk, oldsk);
6951 /* Restore the ep value that was overwritten with the above structure
6957 /* Hook this new socket in to the bind_hash list. */
6958 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
6959 inet_sk(oldsk)->inet_num)];
6960 sctp_local_bh_disable();
6961 sctp_spin_lock(&head->lock);
6962 pp = sctp_sk(oldsk)->bind_hash;
6963 sk_add_bind_node(newsk, &pp->owner);
6964 sctp_sk(newsk)->bind_hash = pp;
6965 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6966 sctp_spin_unlock(&head->lock);
6967 sctp_local_bh_enable();
6969 /* Copy the bind_addr list from the original endpoint to the new
6970 * endpoint so that we can handle restarts properly
6972 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6973 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6975 /* Move any messages in the old socket's receive queue that are for the
6976 * peeled off association to the new socket's receive queue.
6978 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6979 event = sctp_skb2event(skb);
6980 if (event->asoc == assoc) {
6981 __skb_unlink(skb, &oldsk->sk_receive_queue);
6982 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6983 sctp_skb_set_owner_r_frag(skb, newsk);
6987 /* Clean up any messages pending delivery due to partial
6988 * delivery. Three cases:
6989 * 1) No partial deliver; no work.
6990 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6991 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6993 skb_queue_head_init(&newsp->pd_lobby);
6994 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6996 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6997 struct sk_buff_head *queue;
6999 /* Decide which queue to move pd_lobby skbs to. */
7000 if (assoc->ulpq.pd_mode) {
7001 queue = &newsp->pd_lobby;
7003 queue = &newsk->sk_receive_queue;
7005 /* Walk through the pd_lobby, looking for skbs that
7006 * need moved to the new socket.
7008 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
7009 event = sctp_skb2event(skb);
7010 if (event->asoc == assoc) {
7011 __skb_unlink(skb, &oldsp->pd_lobby);
7012 __skb_queue_tail(queue, skb);
7013 sctp_skb_set_owner_r_frag(skb, newsk);
7017 /* Clear up any skbs waiting for the partial
7018 * delivery to finish.
7020 if (assoc->ulpq.pd_mode)
7021 sctp_clear_pd(oldsk, NULL);
7025 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7026 sctp_skb_set_owner_r_frag(skb, newsk);
7028 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7029 sctp_skb_set_owner_r_frag(skb, newsk);
7031 /* Set the type of socket to indicate that it is peeled off from the
7032 * original UDP-style socket or created with the accept() call on a
7033 * TCP-style socket..
7037 /* Mark the new socket "in-use" by the user so that any packets
7038 * that may arrive on the association after we've moved it are
7039 * queued to the backlog. This prevents a potential race between
7040 * backlog processing on the old socket and new-packet processing
7041 * on the new socket.
7043 * The caller has just allocated newsk so we can guarantee that other
7044 * paths won't try to lock it and then oldsk.
7046 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7047 sctp_assoc_migrate(assoc, newsk);
7049 /* If the association on the newsk is already closed before accept()
7050 * is called, set RCV_SHUTDOWN flag.
7052 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
7053 newsk->sk_shutdown |= RCV_SHUTDOWN;
7055 newsk->sk_state = SCTP_SS_ESTABLISHED;
7056 sctp_release_sock(newsk);
7060 /* This proto struct describes the ULP interface for SCTP. */
7061 struct proto sctp_prot = {
7063 .owner = THIS_MODULE,
7064 .close = sctp_close,
7065 .connect = sctp_connect,
7066 .disconnect = sctp_disconnect,
7067 .accept = sctp_accept,
7068 .ioctl = sctp_ioctl,
7069 .init = sctp_init_sock,
7070 .destroy = sctp_destroy_sock,
7071 .shutdown = sctp_shutdown,
7072 .setsockopt = sctp_setsockopt,
7073 .getsockopt = sctp_getsockopt,
7074 .sendmsg = sctp_sendmsg,
7075 .recvmsg = sctp_recvmsg,
7077 .backlog_rcv = sctp_backlog_rcv,
7079 .unhash = sctp_unhash,
7080 .get_port = sctp_get_port,
7081 .obj_size = sizeof(struct sctp_sock),
7082 .sysctl_mem = sysctl_sctp_mem,
7083 .sysctl_rmem = sysctl_sctp_rmem,
7084 .sysctl_wmem = sysctl_sctp_wmem,
7085 .memory_pressure = &sctp_memory_pressure,
7086 .enter_memory_pressure = sctp_enter_memory_pressure,
7087 .memory_allocated = &sctp_memory_allocated,
7088 .sockets_allocated = &sctp_sockets_allocated,
7091 #if IS_ENABLED(CONFIG_IPV6)
7093 struct proto sctpv6_prot = {
7095 .owner = THIS_MODULE,
7096 .close = sctp_close,
7097 .connect = sctp_connect,
7098 .disconnect = sctp_disconnect,
7099 .accept = sctp_accept,
7100 .ioctl = sctp_ioctl,
7101 .init = sctp_init_sock,
7102 .destroy = sctp_destroy_sock,
7103 .shutdown = sctp_shutdown,
7104 .setsockopt = sctp_setsockopt,
7105 .getsockopt = sctp_getsockopt,
7106 .sendmsg = sctp_sendmsg,
7107 .recvmsg = sctp_recvmsg,
7109 .backlog_rcv = sctp_backlog_rcv,
7111 .unhash = sctp_unhash,
7112 .get_port = sctp_get_port,
7113 .obj_size = sizeof(struct sctp6_sock),
7114 .sysctl_mem = sysctl_sctp_mem,
7115 .sysctl_rmem = sysctl_sctp_rmem,
7116 .sysctl_wmem = sysctl_sctp_wmem,
7117 .memory_pressure = &sctp_memory_pressure,
7118 .enter_memory_pressure = sctp_enter_memory_pressure,
7119 .memory_allocated = &sctp_memory_allocated,
7120 .sockets_allocated = &sctp_sockets_allocated,
7122 #endif /* IS_ENABLED(CONFIG_IPV6) */