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, see
32 * <http://www.gnu.org/licenses/>.
34 * Please send any bug reports or fixes you make to the
36 * lksctp developers <linux-sctp@vger.kernel.org>
38 * Written or modified by:
39 * La Monte H.P. Yarroll <piggy@acm.org>
40 * Narasimha Budihal <narsi@refcode.org>
41 * Karl Knutson <karl@athena.chicago.il.us>
42 * Jon Grimm <jgrimm@us.ibm.com>
43 * Xingang Guo <xingang.guo@intel.com>
44 * Daisy Chang <daisyc@us.ibm.com>
45 * Sridhar Samudrala <samudrala@us.ibm.com>
46 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
47 * Ardelle Fan <ardelle.fan@intel.com>
48 * Ryan Layer <rmlayer@us.ibm.com>
49 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
50 * Kevin Gao <kevin.gao@intel.com>
53 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
55 #include <linux/types.h>
56 #include <linux/kernel.h>
57 #include <linux/wait.h>
58 #include <linux/time.h>
60 #include <linux/capability.h>
61 #include <linux/fcntl.h>
62 #include <linux/poll.h>
63 #include <linux/init.h>
64 #include <linux/crypto.h>
65 #include <linux/slab.h>
66 #include <linux/file.h>
70 #include <net/route.h>
72 #include <net/inet_common.h>
74 #include <linux/socket.h> /* for sa_family_t */
75 #include <linux/export.h>
77 #include <net/sctp/sctp.h>
78 #include <net/sctp/sm.h>
80 /* Forward declarations for internal helper functions. */
81 static int sctp_writeable(struct sock *sk);
82 static void sctp_wfree(struct sk_buff *skb);
83 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
85 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
86 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
87 static int sctp_wait_for_accept(struct sock *sk, long timeo);
88 static void sctp_wait_for_close(struct sock *sk, long timeo);
89 static void sctp_destruct_sock(struct sock *sk);
90 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
91 union sctp_addr *addr, int len);
92 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
93 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
94 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
95 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
96 static int sctp_send_asconf(struct sctp_association *asoc,
97 struct sctp_chunk *chunk);
98 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
99 static int sctp_autobind(struct sock *sk);
100 static void sctp_sock_migrate(struct sock *, struct sock *,
101 struct sctp_association *, sctp_socket_type_t);
103 extern struct kmem_cache *sctp_bucket_cachep;
104 extern long sysctl_sctp_mem[3];
105 extern int sysctl_sctp_rmem[3];
106 extern int sysctl_sctp_wmem[3];
108 static int sctp_memory_pressure;
109 static atomic_long_t sctp_memory_allocated;
110 struct percpu_counter sctp_sockets_allocated;
112 static void sctp_enter_memory_pressure(struct sock *sk)
114 sctp_memory_pressure = 1;
118 /* Get the sndbuf space available at the time on the association. */
119 static inline int sctp_wspace(struct sctp_association *asoc)
123 if (asoc->ep->sndbuf_policy)
124 amt = asoc->sndbuf_used;
126 amt = sk_wmem_alloc_get(asoc->base.sk);
128 if (amt >= asoc->base.sk->sk_sndbuf) {
129 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
132 amt = sk_stream_wspace(asoc->base.sk);
137 amt = asoc->base.sk->sk_sndbuf - amt;
142 /* Increment the used sndbuf space count of the corresponding association by
143 * the size of the outgoing data chunk.
144 * Also, set the skb destructor for sndbuf accounting later.
146 * Since it is always 1-1 between chunk and skb, and also a new skb is always
147 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
148 * destructor in the data chunk skb for the purpose of the sndbuf space
151 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
153 struct sctp_association *asoc = chunk->asoc;
154 struct sock *sk = asoc->base.sk;
156 /* The sndbuf space is tracked per association. */
157 sctp_association_hold(asoc);
159 skb_set_owner_w(chunk->skb, sk);
161 chunk->skb->destructor = sctp_wfree;
162 /* Save the chunk pointer in skb for sctp_wfree to use later. */
163 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
165 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
166 sizeof(struct sk_buff) +
167 sizeof(struct sctp_chunk);
169 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
170 sk->sk_wmem_queued += chunk->skb->truesize;
171 sk_mem_charge(sk, chunk->skb->truesize);
174 /* Verify that this is a valid address. */
175 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
180 /* Verify basic sockaddr. */
181 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
185 /* Is this a valid SCTP address? */
186 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
189 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
195 /* Look up the association by its id. If this is not a UDP-style
196 * socket, the ID field is always ignored.
198 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
200 struct sctp_association *asoc = NULL;
202 /* If this is not a UDP-style socket, assoc id should be ignored. */
203 if (!sctp_style(sk, UDP)) {
204 /* Return NULL if the socket state is not ESTABLISHED. It
205 * could be a TCP-style listening socket or a socket which
206 * hasn't yet called connect() to establish an association.
208 if (!sctp_sstate(sk, ESTABLISHED))
211 /* Get the first and the only association from the list. */
212 if (!list_empty(&sctp_sk(sk)->ep->asocs))
213 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
214 struct sctp_association, asocs);
218 /* Otherwise this is a UDP-style socket. */
219 if (!id || (id == (sctp_assoc_t)-1))
222 spin_lock_bh(&sctp_assocs_id_lock);
223 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
224 spin_unlock_bh(&sctp_assocs_id_lock);
226 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
232 /* Look up the transport from an address and an assoc id. If both address and
233 * id are specified, the associations matching the address and the id should be
236 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
237 struct sockaddr_storage *addr,
240 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
241 struct sctp_transport *transport;
242 union sctp_addr *laddr = (union sctp_addr *)addr;
244 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
251 id_asoc = sctp_id2assoc(sk, id);
252 if (id_asoc && (id_asoc != addr_asoc))
255 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
256 (union sctp_addr *)addr);
261 /* API 3.1.2 bind() - UDP Style Syntax
262 * The syntax of bind() is,
264 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
266 * sd - the socket descriptor returned by socket().
267 * addr - the address structure (struct sockaddr_in or struct
268 * sockaddr_in6 [RFC 2553]),
269 * addr_len - the size of the address structure.
271 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
277 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
280 /* Disallow binding twice. */
281 if (!sctp_sk(sk)->ep->base.bind_addr.port)
282 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
287 sctp_release_sock(sk);
292 static long sctp_get_port_local(struct sock *, union sctp_addr *);
294 /* Verify this is a valid sockaddr. */
295 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
296 union sctp_addr *addr, int len)
300 /* Check minimum size. */
301 if (len < sizeof (struct sockaddr))
304 /* V4 mapped address are really of AF_INET family */
305 if (addr->sa.sa_family == AF_INET6 &&
306 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
307 if (!opt->pf->af_supported(AF_INET, opt))
310 /* Does this PF support this AF? */
311 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
315 /* If we get this far, af is valid. */
316 af = sctp_get_af_specific(addr->sa.sa_family);
318 if (len < af->sockaddr_len)
324 /* Bind a local address either to an endpoint or to an association. */
325 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
327 struct net *net = sock_net(sk);
328 struct sctp_sock *sp = sctp_sk(sk);
329 struct sctp_endpoint *ep = sp->ep;
330 struct sctp_bind_addr *bp = &ep->base.bind_addr;
335 /* Common sockaddr verification. */
336 af = sctp_sockaddr_af(sp, addr, len);
338 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
339 __func__, sk, addr, len);
343 snum = ntohs(addr->v4.sin_port);
345 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
346 __func__, sk, &addr->sa, bp->port, snum, len);
348 /* PF specific bind() address verification. */
349 if (!sp->pf->bind_verify(sp, addr))
350 return -EADDRNOTAVAIL;
352 /* We must either be unbound, or bind to the same port.
353 * It's OK to allow 0 ports if we are already bound.
354 * We'll just inhert an already bound port in this case
359 else if (snum != bp->port) {
360 pr_debug("%s: new port %d doesn't match existing port "
361 "%d\n", __func__, snum, bp->port);
366 if (snum && snum < PROT_SOCK &&
367 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
370 /* See if the address matches any of the addresses we may have
371 * already bound before checking against other endpoints.
373 if (sctp_bind_addr_match(bp, addr, sp))
376 /* Make sure we are allowed to bind here.
377 * The function sctp_get_port_local() does duplicate address
380 addr->v4.sin_port = htons(snum);
381 if ((ret = sctp_get_port_local(sk, addr))) {
385 /* Refresh ephemeral port. */
387 bp->port = inet_sk(sk)->inet_num;
389 /* Add the address to the bind address list.
390 * Use GFP_ATOMIC since BHs will be disabled.
392 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
394 /* Copy back into socket for getsockname() use. */
396 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
397 af->to_sk_saddr(addr, sk);
403 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
405 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
406 * at any one time. If a sender, after sending an ASCONF chunk, decides
407 * it needs to transfer another ASCONF Chunk, it MUST wait until the
408 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
409 * subsequent ASCONF. Note this restriction binds each side, so at any
410 * time two ASCONF may be in-transit on any given association (one sent
411 * from each endpoint).
413 static int sctp_send_asconf(struct sctp_association *asoc,
414 struct sctp_chunk *chunk)
416 struct net *net = sock_net(asoc->base.sk);
419 /* If there is an outstanding ASCONF chunk, queue it for later
422 if (asoc->addip_last_asconf) {
423 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
427 /* Hold the chunk until an ASCONF_ACK is received. */
428 sctp_chunk_hold(chunk);
429 retval = sctp_primitive_ASCONF(net, asoc, chunk);
431 sctp_chunk_free(chunk);
433 asoc->addip_last_asconf = chunk;
439 /* Add a list of addresses as bind addresses to local endpoint or
442 * Basically run through each address specified in the addrs/addrcnt
443 * array/length pair, determine if it is IPv6 or IPv4 and call
444 * sctp_do_bind() on it.
446 * If any of them fails, then the operation will be reversed and the
447 * ones that were added will be removed.
449 * Only sctp_setsockopt_bindx() is supposed to call this function.
451 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
456 struct sockaddr *sa_addr;
459 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
463 for (cnt = 0; cnt < addrcnt; cnt++) {
464 /* The list may contain either IPv4 or IPv6 address;
465 * determine the address length for walking thru the list.
468 af = sctp_get_af_specific(sa_addr->sa_family);
474 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
477 addr_buf += af->sockaddr_len;
481 /* Failed. Cleanup the ones that have been added */
483 sctp_bindx_rem(sk, addrs, cnt);
491 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
492 * associations that are part of the endpoint indicating that a list of local
493 * addresses are added to the endpoint.
495 * If any of the addresses is already in the bind address list of the
496 * association, we do not send the chunk for that association. But it will not
497 * affect other associations.
499 * Only sctp_setsockopt_bindx() is supposed to call this function.
501 static int sctp_send_asconf_add_ip(struct sock *sk,
502 struct sockaddr *addrs,
505 struct net *net = sock_net(sk);
506 struct sctp_sock *sp;
507 struct sctp_endpoint *ep;
508 struct sctp_association *asoc;
509 struct sctp_bind_addr *bp;
510 struct sctp_chunk *chunk;
511 struct sctp_sockaddr_entry *laddr;
512 union sctp_addr *addr;
513 union sctp_addr saveaddr;
520 if (!net->sctp.addip_enable)
526 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
527 __func__, sk, addrs, addrcnt);
529 list_for_each_entry(asoc, &ep->asocs, asocs) {
530 if (!asoc->peer.asconf_capable)
533 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
536 if (!sctp_state(asoc, ESTABLISHED))
539 /* Check if any address in the packed array of addresses is
540 * in the bind address list of the association. If so,
541 * do not send the asconf chunk to its peer, but continue with
542 * other associations.
545 for (i = 0; i < addrcnt; i++) {
547 af = sctp_get_af_specific(addr->v4.sin_family);
553 if (sctp_assoc_lookup_laddr(asoc, addr))
556 addr_buf += af->sockaddr_len;
561 /* Use the first valid address in bind addr list of
562 * association as Address Parameter of ASCONF CHUNK.
564 bp = &asoc->base.bind_addr;
565 p = bp->address_list.next;
566 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
567 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
568 addrcnt, SCTP_PARAM_ADD_IP);
574 /* Add the new addresses to the bind address list with
575 * use_as_src set to 0.
578 for (i = 0; i < addrcnt; i++) {
580 af = sctp_get_af_specific(addr->v4.sin_family);
581 memcpy(&saveaddr, addr, af->sockaddr_len);
582 retval = sctp_add_bind_addr(bp, &saveaddr,
583 SCTP_ADDR_NEW, GFP_ATOMIC);
584 addr_buf += af->sockaddr_len;
586 if (asoc->src_out_of_asoc_ok) {
587 struct sctp_transport *trans;
589 list_for_each_entry(trans,
590 &asoc->peer.transport_addr_list, transports) {
591 /* Clear the source and route cache */
592 dst_release(trans->dst);
593 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
594 2*asoc->pathmtu, 4380));
595 trans->ssthresh = asoc->peer.i.a_rwnd;
596 trans->rto = asoc->rto_initial;
597 sctp_max_rto(asoc, trans);
598 trans->rtt = trans->srtt = trans->rttvar = 0;
599 sctp_transport_route(trans, NULL,
600 sctp_sk(asoc->base.sk));
603 retval = sctp_send_asconf(asoc, chunk);
610 /* Remove a list of addresses from bind addresses list. Do not remove the
613 * Basically run through each address specified in the addrs/addrcnt
614 * array/length pair, determine if it is IPv6 or IPv4 and call
615 * sctp_del_bind() on it.
617 * If any of them fails, then the operation will be reversed and the
618 * ones that were removed will be added back.
620 * At least one address has to be left; if only one address is
621 * available, the operation will return -EBUSY.
623 * Only sctp_setsockopt_bindx() is supposed to call this function.
625 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
627 struct sctp_sock *sp = sctp_sk(sk);
628 struct sctp_endpoint *ep = sp->ep;
630 struct sctp_bind_addr *bp = &ep->base.bind_addr;
633 union sctp_addr *sa_addr;
636 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
637 __func__, sk, addrs, addrcnt);
640 for (cnt = 0; cnt < addrcnt; cnt++) {
641 /* If the bind address list is empty or if there is only one
642 * bind address, there is nothing more to be removed (we need
643 * at least one address here).
645 if (list_empty(&bp->address_list) ||
646 (sctp_list_single_entry(&bp->address_list))) {
652 af = sctp_get_af_specific(sa_addr->sa.sa_family);
658 if (!af->addr_valid(sa_addr, sp, NULL)) {
659 retval = -EADDRNOTAVAIL;
663 if (sa_addr->v4.sin_port &&
664 sa_addr->v4.sin_port != htons(bp->port)) {
669 if (!sa_addr->v4.sin_port)
670 sa_addr->v4.sin_port = htons(bp->port);
672 /* FIXME - There is probably a need to check if sk->sk_saddr and
673 * sk->sk_rcv_addr are currently set to one of the addresses to
674 * be removed. This is something which needs to be looked into
675 * when we are fixing the outstanding issues with multi-homing
676 * socket routing and failover schemes. Refer to comments in
677 * sctp_do_bind(). -daisy
679 retval = sctp_del_bind_addr(bp, sa_addr);
681 addr_buf += af->sockaddr_len;
684 /* Failed. Add the ones that has been removed back */
686 sctp_bindx_add(sk, addrs, cnt);
694 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
695 * the associations that are part of the endpoint indicating that a list of
696 * local addresses are removed from the endpoint.
698 * If any of the addresses is already in the bind address list of the
699 * association, we do not send the chunk for that association. But it will not
700 * affect other associations.
702 * Only sctp_setsockopt_bindx() is supposed to call this function.
704 static int sctp_send_asconf_del_ip(struct sock *sk,
705 struct sockaddr *addrs,
708 struct net *net = sock_net(sk);
709 struct sctp_sock *sp;
710 struct sctp_endpoint *ep;
711 struct sctp_association *asoc;
712 struct sctp_transport *transport;
713 struct sctp_bind_addr *bp;
714 struct sctp_chunk *chunk;
715 union sctp_addr *laddr;
718 struct sctp_sockaddr_entry *saddr;
724 if (!net->sctp.addip_enable)
730 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
731 __func__, sk, addrs, addrcnt);
733 list_for_each_entry(asoc, &ep->asocs, asocs) {
735 if (!asoc->peer.asconf_capable)
738 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
741 if (!sctp_state(asoc, ESTABLISHED))
744 /* Check if any address in the packed array of addresses is
745 * not present in the bind address list of the association.
746 * If so, do not send the asconf chunk to its peer, but
747 * continue with other associations.
750 for (i = 0; i < addrcnt; i++) {
752 af = sctp_get_af_specific(laddr->v4.sin_family);
758 if (!sctp_assoc_lookup_laddr(asoc, laddr))
761 addr_buf += af->sockaddr_len;
766 /* Find one address in the association's bind address list
767 * that is not in the packed array of addresses. This is to
768 * make sure that we do not delete all the addresses in the
771 bp = &asoc->base.bind_addr;
772 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
774 if ((laddr == NULL) && (addrcnt == 1)) {
775 if (asoc->asconf_addr_del_pending)
777 asoc->asconf_addr_del_pending =
778 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
779 if (asoc->asconf_addr_del_pending == NULL) {
783 asoc->asconf_addr_del_pending->sa.sa_family =
785 asoc->asconf_addr_del_pending->v4.sin_port =
787 if (addrs->sa_family == AF_INET) {
788 struct sockaddr_in *sin;
790 sin = (struct sockaddr_in *)addrs;
791 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
792 } else if (addrs->sa_family == AF_INET6) {
793 struct sockaddr_in6 *sin6;
795 sin6 = (struct sockaddr_in6 *)addrs;
796 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
799 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
800 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
801 asoc->asconf_addr_del_pending);
803 asoc->src_out_of_asoc_ok = 1;
811 /* We do not need RCU protection throughout this loop
812 * because this is done under a socket lock from the
815 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
823 /* Reset use_as_src flag for the addresses in the bind address
824 * list that are to be deleted.
827 for (i = 0; i < addrcnt; i++) {
829 af = sctp_get_af_specific(laddr->v4.sin_family);
830 list_for_each_entry(saddr, &bp->address_list, list) {
831 if (sctp_cmp_addr_exact(&saddr->a, laddr))
832 saddr->state = SCTP_ADDR_DEL;
834 addr_buf += af->sockaddr_len;
837 /* Update the route and saddr entries for all the transports
838 * as some of the addresses in the bind address list are
839 * about to be deleted and cannot be used as source addresses.
841 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
843 dst_release(transport->dst);
844 sctp_transport_route(transport, NULL,
845 sctp_sk(asoc->base.sk));
849 /* We don't need to transmit ASCONF */
851 retval = sctp_send_asconf(asoc, chunk);
857 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
858 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
860 struct sock *sk = sctp_opt2sk(sp);
861 union sctp_addr *addr;
864 /* It is safe to write port space in caller. */
866 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
867 af = sctp_get_af_specific(addr->sa.sa_family);
870 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
873 if (addrw->state == SCTP_ADDR_NEW)
874 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
876 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
879 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
882 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
885 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
886 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
889 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
890 * Section 3.1.2 for this usage.
892 * addrs is a pointer to an array of one or more socket addresses. Each
893 * address is contained in its appropriate structure (i.e. struct
894 * sockaddr_in or struct sockaddr_in6) the family of the address type
895 * must be used to distinguish the address length (note that this
896 * representation is termed a "packed array" of addresses). The caller
897 * specifies the number of addresses in the array with addrcnt.
899 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
900 * -1, and sets errno to the appropriate error code.
902 * For SCTP, the port given in each socket address must be the same, or
903 * sctp_bindx() will fail, setting errno to EINVAL.
905 * The flags parameter is formed from the bitwise OR of zero or more of
906 * the following currently defined flags:
908 * SCTP_BINDX_ADD_ADDR
910 * SCTP_BINDX_REM_ADDR
912 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
913 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
914 * addresses from the association. The two flags are mutually exclusive;
915 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
916 * not remove all addresses from an association; sctp_bindx() will
917 * reject such an attempt with EINVAL.
919 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
920 * additional addresses with an endpoint after calling bind(). Or use
921 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
922 * socket is associated with so that no new association accepted will be
923 * associated with those addresses. If the endpoint supports dynamic
924 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
925 * endpoint to send the appropriate message to the peer to change the
926 * peers address lists.
928 * Adding and removing addresses from a connected association is
929 * optional functionality. Implementations that do not support this
930 * functionality should return EOPNOTSUPP.
932 * Basically do nothing but copying the addresses from user to kernel
933 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
934 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
937 * We don't use copy_from_user() for optimization: we first do the
938 * sanity checks (buffer size -fast- and access check-healthy
939 * pointer); if all of those succeed, then we can alloc the memory
940 * (expensive operation) needed to copy the data to kernel. Then we do
941 * the copying without checking the user space area
942 * (__copy_from_user()).
944 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
947 * sk The sk of the socket
948 * addrs The pointer to the addresses in user land
949 * addrssize Size of the addrs buffer
950 * op Operation to perform (add or remove, see the flags of
953 * Returns 0 if ok, <0 errno code on error.
955 static int sctp_setsockopt_bindx(struct sock* sk,
956 struct sockaddr __user *addrs,
957 int addrs_size, int op)
959 struct sockaddr *kaddrs;
963 struct sockaddr *sa_addr;
967 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
968 __func__, sk, addrs, addrs_size, op);
970 if (unlikely(addrs_size <= 0))
973 /* Check the user passed a healthy pointer. */
974 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
977 /* Alloc space for the address array in kernel memory. */
978 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
979 if (unlikely(!kaddrs))
982 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
987 /* Walk through the addrs buffer and count the number of addresses. */
989 while (walk_size < addrs_size) {
990 if (walk_size + sizeof(sa_family_t) > addrs_size) {
996 af = sctp_get_af_specific(sa_addr->sa_family);
998 /* If the address family is not supported or if this address
999 * causes the address buffer to overflow return EINVAL.
1001 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1006 addr_buf += af->sockaddr_len;
1007 walk_size += af->sockaddr_len;
1012 case SCTP_BINDX_ADD_ADDR:
1013 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1016 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1019 case SCTP_BINDX_REM_ADDR:
1020 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1023 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1037 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1039 * Common routine for handling connect() and sctp_connectx().
1040 * Connect will come in with just a single address.
1042 static int __sctp_connect(struct sock* sk,
1043 struct sockaddr *kaddrs,
1045 sctp_assoc_t *assoc_id)
1047 struct net *net = sock_net(sk);
1048 struct sctp_sock *sp;
1049 struct sctp_endpoint *ep;
1050 struct sctp_association *asoc = NULL;
1051 struct sctp_association *asoc2;
1052 struct sctp_transport *transport;
1060 union sctp_addr *sa_addr = NULL;
1062 unsigned short port;
1063 unsigned int f_flags = 0;
1068 /* connect() cannot be done on a socket that is already in ESTABLISHED
1069 * state - UDP-style peeled off socket or a TCP-style socket that
1070 * is already connected.
1071 * It cannot be done even on a TCP-style listening socket.
1073 if (sctp_sstate(sk, ESTABLISHED) ||
1074 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1079 /* Walk through the addrs buffer and count the number of addresses. */
1081 while (walk_size < addrs_size) {
1082 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1088 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1090 /* If the address family is not supported or if this address
1091 * causes the address buffer to overflow return EINVAL.
1093 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1098 port = ntohs(sa_addr->v4.sin_port);
1100 /* Save current address so we can work with it */
1101 memcpy(&to, sa_addr, af->sockaddr_len);
1103 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1107 /* Make sure the destination port is correctly set
1110 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1115 /* Check if there already is a matching association on the
1116 * endpoint (other than the one created here).
1118 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1119 if (asoc2 && asoc2 != asoc) {
1120 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1127 /* If we could not find a matching association on the endpoint,
1128 * make sure that there is no peeled-off association matching
1129 * the peer address even on another socket.
1131 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1132 err = -EADDRNOTAVAIL;
1137 /* If a bind() or sctp_bindx() is not called prior to
1138 * an sctp_connectx() call, the system picks an
1139 * ephemeral port and will choose an address set
1140 * equivalent to binding with a wildcard address.
1142 if (!ep->base.bind_addr.port) {
1143 if (sctp_autobind(sk)) {
1149 * If an unprivileged user inherits a 1-many
1150 * style socket with open associations on a
1151 * privileged port, it MAY be permitted to
1152 * accept new associations, but it SHOULD NOT
1153 * be permitted to open new associations.
1155 if (ep->base.bind_addr.port < PROT_SOCK &&
1156 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1162 scope = sctp_scope(&to);
1163 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1169 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1177 /* Prime the peer's transport structures. */
1178 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1186 addr_buf += af->sockaddr_len;
1187 walk_size += af->sockaddr_len;
1190 /* In case the user of sctp_connectx() wants an association
1191 * id back, assign one now.
1194 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1199 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1204 /* Initialize sk's dport and daddr for getpeername() */
1205 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1206 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1207 af->to_sk_daddr(sa_addr, sk);
1210 /* in-kernel sockets don't generally have a file allocated to them
1211 * if all they do is call sock_create_kern().
1213 if (sk->sk_socket->file)
1214 f_flags = sk->sk_socket->file->f_flags;
1216 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1218 err = sctp_wait_for_connect(asoc, &timeo);
1219 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1220 *assoc_id = asoc->assoc_id;
1222 /* Don't free association on exit. */
1226 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1227 __func__, asoc, kaddrs, err);
1230 /* sctp_primitive_ASSOCIATE may have added this association
1231 * To the hash table, try to unhash it, just in case, its a noop
1232 * if it wasn't hashed so we're safe
1234 sctp_unhash_established(asoc);
1235 sctp_association_free(asoc);
1240 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1243 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1244 * sctp_assoc_t *asoc);
1246 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1247 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1248 * or IPv6 addresses.
1250 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1251 * Section 3.1.2 for this usage.
1253 * addrs is a pointer to an array of one or more socket addresses. Each
1254 * address is contained in its appropriate structure (i.e. struct
1255 * sockaddr_in or struct sockaddr_in6) the family of the address type
1256 * must be used to distengish the address length (note that this
1257 * representation is termed a "packed array" of addresses). The caller
1258 * specifies the number of addresses in the array with addrcnt.
1260 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1261 * the association id of the new association. On failure, sctp_connectx()
1262 * returns -1, and sets errno to the appropriate error code. The assoc_id
1263 * is not touched by the kernel.
1265 * For SCTP, the port given in each socket address must be the same, or
1266 * sctp_connectx() will fail, setting errno to EINVAL.
1268 * An application can use sctp_connectx to initiate an association with
1269 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1270 * allows a caller to specify multiple addresses at which a peer can be
1271 * reached. The way the SCTP stack uses the list of addresses to set up
1272 * the association is implementation dependent. This function only
1273 * specifies that the stack will try to make use of all the addresses in
1274 * the list when needed.
1276 * Note that the list of addresses passed in is only used for setting up
1277 * the association. It does not necessarily equal the set of addresses
1278 * the peer uses for the resulting association. If the caller wants to
1279 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1280 * retrieve them after the association has been set up.
1282 * Basically do nothing but copying the addresses from user to kernel
1283 * land and invoking either sctp_connectx(). This is used for tunneling
1284 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1286 * We don't use copy_from_user() for optimization: we first do the
1287 * sanity checks (buffer size -fast- and access check-healthy
1288 * pointer); if all of those succeed, then we can alloc the memory
1289 * (expensive operation) needed to copy the data to kernel. Then we do
1290 * the copying without checking the user space area
1291 * (__copy_from_user()).
1293 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1296 * sk The sk of the socket
1297 * addrs The pointer to the addresses in user land
1298 * addrssize Size of the addrs buffer
1300 * Returns >=0 if ok, <0 errno code on error.
1302 static int __sctp_setsockopt_connectx(struct sock* sk,
1303 struct sockaddr __user *addrs,
1305 sctp_assoc_t *assoc_id)
1308 struct sockaddr *kaddrs;
1310 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1311 __func__, sk, addrs, addrs_size);
1313 if (unlikely(addrs_size <= 0))
1316 /* Check the user passed a healthy pointer. */
1317 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1320 /* Alloc space for the address array in kernel memory. */
1321 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1322 if (unlikely(!kaddrs))
1325 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1328 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1337 * This is an older interface. It's kept for backward compatibility
1338 * to the option that doesn't provide association id.
1340 static int sctp_setsockopt_connectx_old(struct sock* sk,
1341 struct sockaddr __user *addrs,
1344 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1348 * New interface for the API. The since the API is done with a socket
1349 * option, to make it simple we feed back the association id is as a return
1350 * indication to the call. Error is always negative and association id is
1353 static int sctp_setsockopt_connectx(struct sock* sk,
1354 struct sockaddr __user *addrs,
1357 sctp_assoc_t assoc_id = 0;
1360 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1369 * New (hopefully final) interface for the API.
1370 * We use the sctp_getaddrs_old structure so that use-space library
1371 * can avoid any unnecessary allocations. The only defferent part
1372 * is that we store the actual length of the address buffer into the
1373 * addrs_num structure member. That way we can re-use the existing
1376 static int sctp_getsockopt_connectx3(struct sock* sk, int len,
1377 char __user *optval,
1380 struct sctp_getaddrs_old param;
1381 sctp_assoc_t assoc_id = 0;
1384 if (len < sizeof(param))
1387 if (copy_from_user(¶m, optval, sizeof(param)))
1390 err = __sctp_setsockopt_connectx(sk,
1391 (struct sockaddr __user *)param.addrs,
1392 param.addr_num, &assoc_id);
1394 if (err == 0 || err == -EINPROGRESS) {
1395 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1397 if (put_user(sizeof(assoc_id), optlen))
1404 /* API 3.1.4 close() - UDP Style Syntax
1405 * Applications use close() to perform graceful shutdown (as described in
1406 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1407 * by a UDP-style socket.
1411 * ret = close(int sd);
1413 * sd - the socket descriptor of the associations to be closed.
1415 * To gracefully shutdown a specific association represented by the
1416 * UDP-style socket, an application should use the sendmsg() call,
1417 * passing no user data, but including the appropriate flag in the
1418 * ancillary data (see Section xxxx).
1420 * If sd in the close() call is a branched-off socket representing only
1421 * one association, the shutdown is performed on that association only.
1423 * 4.1.6 close() - TCP Style Syntax
1425 * Applications use close() to gracefully close down an association.
1429 * int close(int sd);
1431 * sd - the socket descriptor of the association to be closed.
1433 * After an application calls close() on a socket descriptor, no further
1434 * socket operations will succeed on that descriptor.
1436 * API 7.1.4 SO_LINGER
1438 * An application using the TCP-style socket can use this option to
1439 * perform the SCTP ABORT primitive. The linger option structure is:
1442 * int l_onoff; // option on/off
1443 * int l_linger; // linger time
1446 * To enable the option, set l_onoff to 1. If the l_linger value is set
1447 * to 0, calling close() is the same as the ABORT primitive. If the
1448 * value is set to a negative value, the setsockopt() call will return
1449 * an error. If the value is set to a positive value linger_time, the
1450 * close() can be blocked for at most linger_time ms. If the graceful
1451 * shutdown phase does not finish during this period, close() will
1452 * return but the graceful shutdown phase continues in the system.
1454 static void sctp_close(struct sock *sk, long timeout)
1456 struct net *net = sock_net(sk);
1457 struct sctp_endpoint *ep;
1458 struct sctp_association *asoc;
1459 struct list_head *pos, *temp;
1460 unsigned int data_was_unread;
1462 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1465 sk->sk_shutdown = SHUTDOWN_MASK;
1466 sk->sk_state = SCTP_SS_CLOSING;
1468 ep = sctp_sk(sk)->ep;
1470 /* Clean up any skbs sitting on the receive queue. */
1471 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1472 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1474 /* Walk all associations on an endpoint. */
1475 list_for_each_safe(pos, temp, &ep->asocs) {
1476 asoc = list_entry(pos, struct sctp_association, asocs);
1478 if (sctp_style(sk, TCP)) {
1479 /* A closed association can still be in the list if
1480 * it belongs to a TCP-style listening socket that is
1481 * not yet accepted. If so, free it. If not, send an
1482 * ABORT or SHUTDOWN based on the linger options.
1484 if (sctp_state(asoc, CLOSED)) {
1485 sctp_unhash_established(asoc);
1486 sctp_association_free(asoc);
1491 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1492 !skb_queue_empty(&asoc->ulpq.reasm) ||
1493 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1494 struct sctp_chunk *chunk;
1496 chunk = sctp_make_abort_user(asoc, NULL, 0);
1498 sctp_primitive_ABORT(net, asoc, chunk);
1500 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1503 /* On a TCP-style socket, block for at most linger_time if set. */
1504 if (sctp_style(sk, TCP) && timeout)
1505 sctp_wait_for_close(sk, timeout);
1507 /* This will run the backlog queue. */
1508 sctp_release_sock(sk);
1510 /* Supposedly, no process has access to the socket, but
1511 * the net layers still may.
1513 sctp_local_bh_disable();
1514 sctp_bh_lock_sock(sk);
1516 /* Hold the sock, since sk_common_release() will put sock_put()
1517 * and we have just a little more cleanup.
1520 sk_common_release(sk);
1522 sctp_bh_unlock_sock(sk);
1523 sctp_local_bh_enable();
1527 SCTP_DBG_OBJCNT_DEC(sock);
1530 /* Handle EPIPE error. */
1531 static int sctp_error(struct sock *sk, int flags, int err)
1534 err = sock_error(sk) ? : -EPIPE;
1535 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1536 send_sig(SIGPIPE, current, 0);
1540 /* API 3.1.3 sendmsg() - UDP Style Syntax
1542 * An application uses sendmsg() and recvmsg() calls to transmit data to
1543 * and receive data from its peer.
1545 * ssize_t sendmsg(int socket, const struct msghdr *message,
1548 * socket - the socket descriptor of the endpoint.
1549 * message - pointer to the msghdr structure which contains a single
1550 * user message and possibly some ancillary data.
1552 * See Section 5 for complete description of the data
1555 * flags - flags sent or received with the user message, see Section
1556 * 5 for complete description of the flags.
1558 * Note: This function could use a rewrite especially when explicit
1559 * connect support comes in.
1561 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1563 static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1565 static int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1566 struct msghdr *msg, size_t msg_len)
1568 struct net *net = sock_net(sk);
1569 struct sctp_sock *sp;
1570 struct sctp_endpoint *ep;
1571 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1572 struct sctp_transport *transport, *chunk_tp;
1573 struct sctp_chunk *chunk;
1575 struct sockaddr *msg_name = NULL;
1576 struct sctp_sndrcvinfo default_sinfo;
1577 struct sctp_sndrcvinfo *sinfo;
1578 struct sctp_initmsg *sinit;
1579 sctp_assoc_t associd = 0;
1580 sctp_cmsgs_t cmsgs = { NULL };
1584 __u16 sinfo_flags = 0;
1585 struct sctp_datamsg *datamsg;
1586 int msg_flags = msg->msg_flags;
1592 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1595 /* We cannot send a message over a TCP-style listening socket. */
1596 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1601 /* Parse out the SCTP CMSGs. */
1602 err = sctp_msghdr_parse(msg, &cmsgs);
1604 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1608 /* Fetch the destination address for this packet. This
1609 * address only selects the association--it is not necessarily
1610 * the address we will send to.
1611 * For a peeled-off socket, msg_name is ignored.
1613 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1614 int msg_namelen = msg->msg_namelen;
1616 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1621 if (msg_namelen > sizeof(to))
1622 msg_namelen = sizeof(to);
1623 memcpy(&to, msg->msg_name, msg_namelen);
1624 msg_name = msg->msg_name;
1630 /* Did the user specify SNDRCVINFO? */
1632 sinfo_flags = sinfo->sinfo_flags;
1633 associd = sinfo->sinfo_assoc_id;
1636 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1637 msg_len, sinfo_flags);
1639 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1640 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1645 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1646 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1647 * If SCTP_ABORT is set, the message length could be non zero with
1648 * the msg_iov set to the user abort reason.
1650 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1651 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1656 /* If SCTP_ADDR_OVER is set, there must be an address
1657 * specified in msg_name.
1659 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1666 pr_debug("%s: about to look up association\n", __func__);
1670 /* If a msg_name has been specified, assume this is to be used. */
1672 /* Look for a matching association on the endpoint. */
1673 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1675 /* If we could not find a matching association on the
1676 * endpoint, make sure that it is not a TCP-style
1677 * socket that already has an association or there is
1678 * no peeled-off association on another socket.
1680 if ((sctp_style(sk, TCP) &&
1681 sctp_sstate(sk, ESTABLISHED)) ||
1682 sctp_endpoint_is_peeled_off(ep, &to)) {
1683 err = -EADDRNOTAVAIL;
1688 asoc = sctp_id2assoc(sk, associd);
1696 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1698 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1699 * socket that has an association in CLOSED state. This can
1700 * happen when an accepted socket has an association that is
1703 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1708 if (sinfo_flags & SCTP_EOF) {
1709 pr_debug("%s: shutting down association:%p\n",
1712 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1716 if (sinfo_flags & SCTP_ABORT) {
1718 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1724 pr_debug("%s: aborting association:%p\n",
1727 sctp_primitive_ABORT(net, asoc, chunk);
1733 /* Do we need to create the association? */
1735 pr_debug("%s: there is no association yet\n", __func__);
1737 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1742 /* Check for invalid stream against the stream counts,
1743 * either the default or the user specified stream counts.
1746 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1747 /* Check against the defaults. */
1748 if (sinfo->sinfo_stream >=
1749 sp->initmsg.sinit_num_ostreams) {
1754 /* Check against the requested. */
1755 if (sinfo->sinfo_stream >=
1756 sinit->sinit_num_ostreams) {
1764 * API 3.1.2 bind() - UDP Style Syntax
1765 * If a bind() or sctp_bindx() is not called prior to a
1766 * sendmsg() call that initiates a new association, the
1767 * system picks an ephemeral port and will choose an address
1768 * set equivalent to binding with a wildcard address.
1770 if (!ep->base.bind_addr.port) {
1771 if (sctp_autobind(sk)) {
1777 * If an unprivileged user inherits a one-to-many
1778 * style socket with open associations on a privileged
1779 * port, it MAY be permitted to accept new associations,
1780 * but it SHOULD NOT be permitted to open new
1783 if (ep->base.bind_addr.port < PROT_SOCK &&
1784 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1790 scope = sctp_scope(&to);
1791 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1797 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1803 /* If the SCTP_INIT ancillary data is specified, set all
1804 * the association init values accordingly.
1807 if (sinit->sinit_num_ostreams) {
1808 asoc->c.sinit_num_ostreams =
1809 sinit->sinit_num_ostreams;
1811 if (sinit->sinit_max_instreams) {
1812 asoc->c.sinit_max_instreams =
1813 sinit->sinit_max_instreams;
1815 if (sinit->sinit_max_attempts) {
1816 asoc->max_init_attempts
1817 = sinit->sinit_max_attempts;
1819 if (sinit->sinit_max_init_timeo) {
1820 asoc->max_init_timeo =
1821 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1825 /* Prime the peer's transport structures. */
1826 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1833 /* ASSERT: we have a valid association at this point. */
1834 pr_debug("%s: we have a valid association\n", __func__);
1837 /* If the user didn't specify SNDRCVINFO, make up one with
1840 memset(&default_sinfo, 0, sizeof(default_sinfo));
1841 default_sinfo.sinfo_stream = asoc->default_stream;
1842 default_sinfo.sinfo_flags = asoc->default_flags;
1843 default_sinfo.sinfo_ppid = asoc->default_ppid;
1844 default_sinfo.sinfo_context = asoc->default_context;
1845 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1846 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1847 sinfo = &default_sinfo;
1850 /* API 7.1.7, the sndbuf size per association bounds the
1851 * maximum size of data that can be sent in a single send call.
1853 if (msg_len > sk->sk_sndbuf) {
1858 if (asoc->pmtu_pending)
1859 sctp_assoc_pending_pmtu(sk, asoc);
1861 /* If fragmentation is disabled and the message length exceeds the
1862 * association fragmentation point, return EMSGSIZE. The I-D
1863 * does not specify what this error is, but this looks like
1866 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1871 /* Check for invalid stream. */
1872 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1877 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1878 if (!sctp_wspace(asoc)) {
1879 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1884 /* If an address is passed with the sendto/sendmsg call, it is used
1885 * to override the primary destination address in the TCP model, or
1886 * when SCTP_ADDR_OVER flag is set in the UDP model.
1888 if ((sctp_style(sk, TCP) && msg_name) ||
1889 (sinfo_flags & SCTP_ADDR_OVER)) {
1890 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1898 /* Auto-connect, if we aren't connected already. */
1899 if (sctp_state(asoc, CLOSED)) {
1900 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1904 pr_debug("%s: we associated primitively\n", __func__);
1907 /* Break the message into multiple chunks of maximum size. */
1908 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1909 if (IS_ERR(datamsg)) {
1910 err = PTR_ERR(datamsg);
1914 /* Now send the (possibly) fragmented message. */
1915 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1916 sctp_chunk_hold(chunk);
1918 /* Do accounting for the write space. */
1919 sctp_set_owner_w(chunk);
1921 chunk->transport = chunk_tp;
1924 /* Send it to the lower layers. Note: all chunks
1925 * must either fail or succeed. The lower layer
1926 * works that way today. Keep it that way or this
1929 err = sctp_primitive_SEND(net, asoc, datamsg);
1930 /* Did the lower layer accept the chunk? */
1932 sctp_datamsg_free(datamsg);
1936 pr_debug("%s: we sent primitively\n", __func__);
1938 sctp_datamsg_put(datamsg);
1941 /* If we are already past ASSOCIATE, the lower
1942 * layers are responsible for association cleanup.
1948 sctp_unhash_established(asoc);
1949 sctp_association_free(asoc);
1952 sctp_release_sock(sk);
1955 return sctp_error(sk, msg_flags, err);
1962 err = sock_error(sk);
1972 /* This is an extended version of skb_pull() that removes the data from the
1973 * start of a skb even when data is spread across the list of skb's in the
1974 * frag_list. len specifies the total amount of data that needs to be removed.
1975 * when 'len' bytes could be removed from the skb, it returns 0.
1976 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1977 * could not be removed.
1979 static int sctp_skb_pull(struct sk_buff *skb, int len)
1981 struct sk_buff *list;
1982 int skb_len = skb_headlen(skb);
1985 if (len <= skb_len) {
1986 __skb_pull(skb, len);
1990 __skb_pull(skb, skb_len);
1992 skb_walk_frags(skb, list) {
1993 rlen = sctp_skb_pull(list, len);
1994 skb->len -= (len-rlen);
1995 skb->data_len -= (len-rlen);
2006 /* API 3.1.3 recvmsg() - UDP Style Syntax
2008 * ssize_t recvmsg(int socket, struct msghdr *message,
2011 * socket - the socket descriptor of the endpoint.
2012 * message - pointer to the msghdr structure which contains a single
2013 * user message and possibly some ancillary data.
2015 * See Section 5 for complete description of the data
2018 * flags - flags sent or received with the user message, see Section
2019 * 5 for complete description of the flags.
2021 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2023 static int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2024 struct msghdr *msg, size_t len, int noblock,
2025 int flags, int *addr_len)
2027 struct sctp_ulpevent *event = NULL;
2028 struct sctp_sock *sp = sctp_sk(sk);
2029 struct sk_buff *skb;
2034 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2035 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2040 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2045 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2049 /* Get the total length of the skb including any skb's in the
2058 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2060 event = sctp_skb2event(skb);
2065 sock_recv_ts_and_drops(msg, sk, skb);
2066 if (sctp_ulpevent_is_notification(event)) {
2067 msg->msg_flags |= MSG_NOTIFICATION;
2068 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2070 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2073 /* Check if we allow SCTP_SNDRCVINFO. */
2074 if (sp->subscribe.sctp_data_io_event)
2075 sctp_ulpevent_read_sndrcvinfo(event, msg);
2077 /* FIXME: we should be calling IP/IPv6 layers. */
2078 if (sk->sk_protinfo.af_inet.cmsg_flags)
2079 ip_cmsg_recv(msg, skb);
2084 /* If skb's length exceeds the user's buffer, update the skb and
2085 * push it back to the receive_queue so that the next call to
2086 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2088 if (skb_len > copied) {
2089 msg->msg_flags &= ~MSG_EOR;
2090 if (flags & MSG_PEEK)
2092 sctp_skb_pull(skb, copied);
2093 skb_queue_head(&sk->sk_receive_queue, skb);
2095 /* When only partial message is copied to the user, increase
2096 * rwnd by that amount. If all the data in the skb is read,
2097 * rwnd is updated when the event is freed.
2099 if (!sctp_ulpevent_is_notification(event))
2100 sctp_assoc_rwnd_increase(event->asoc, copied);
2102 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2103 (event->msg_flags & MSG_EOR))
2104 msg->msg_flags |= MSG_EOR;
2106 msg->msg_flags &= ~MSG_EOR;
2109 if (flags & MSG_PEEK) {
2110 /* Release the skb reference acquired after peeking the skb in
2111 * sctp_skb_recv_datagram().
2115 /* Free the event which includes releasing the reference to
2116 * the owner of the skb, freeing the skb and updating the
2119 sctp_ulpevent_free(event);
2122 sctp_release_sock(sk);
2126 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2128 * This option is a on/off flag. If enabled no SCTP message
2129 * fragmentation will be performed. Instead if a message being sent
2130 * exceeds the current PMTU size, the message will NOT be sent and
2131 * instead a error will be indicated to the user.
2133 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2134 char __user *optval,
2135 unsigned int optlen)
2139 if (optlen < sizeof(int))
2142 if (get_user(val, (int __user *)optval))
2145 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2150 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2151 unsigned int optlen)
2153 struct sctp_association *asoc;
2154 struct sctp_ulpevent *event;
2156 if (optlen > sizeof(struct sctp_event_subscribe))
2158 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2162 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2163 * if there is no data to be sent or retransmit, the stack will
2164 * immediately send up this notification.
2166 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2167 &sctp_sk(sk)->subscribe)) {
2168 asoc = sctp_id2assoc(sk, 0);
2170 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2171 event = sctp_ulpevent_make_sender_dry_event(asoc,
2176 sctp_ulpq_tail_event(&asoc->ulpq, event);
2183 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2185 * This socket option is applicable to the UDP-style socket only. When
2186 * set it will cause associations that are idle for more than the
2187 * specified number of seconds to automatically close. An association
2188 * being idle is defined an association that has NOT sent or received
2189 * user data. The special value of '0' indicates that no automatic
2190 * close of any associations should be performed. The option expects an
2191 * integer defining the number of seconds of idle time before an
2192 * association is closed.
2194 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2195 unsigned int optlen)
2197 struct sctp_sock *sp = sctp_sk(sk);
2199 /* Applicable to UDP-style socket only */
2200 if (sctp_style(sk, TCP))
2202 if (optlen != sizeof(int))
2204 if (copy_from_user(&sp->autoclose, optval, optlen))
2210 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2212 * Applications can enable or disable heartbeats for any peer address of
2213 * an association, modify an address's heartbeat interval, force a
2214 * heartbeat to be sent immediately, and adjust the address's maximum
2215 * number of retransmissions sent before an address is considered
2216 * unreachable. The following structure is used to access and modify an
2217 * address's parameters:
2219 * struct sctp_paddrparams {
2220 * sctp_assoc_t spp_assoc_id;
2221 * struct sockaddr_storage spp_address;
2222 * uint32_t spp_hbinterval;
2223 * uint16_t spp_pathmaxrxt;
2224 * uint32_t spp_pathmtu;
2225 * uint32_t spp_sackdelay;
2226 * uint32_t spp_flags;
2229 * spp_assoc_id - (one-to-many style socket) This is filled in the
2230 * application, and identifies the association for
2232 * spp_address - This specifies which address is of interest.
2233 * spp_hbinterval - This contains the value of the heartbeat interval,
2234 * in milliseconds. If a value of zero
2235 * is present in this field then no changes are to
2236 * be made to this parameter.
2237 * spp_pathmaxrxt - This contains the maximum number of
2238 * retransmissions before this address shall be
2239 * considered unreachable. If a value of zero
2240 * is present in this field then no changes are to
2241 * be made to this parameter.
2242 * spp_pathmtu - When Path MTU discovery is disabled the value
2243 * specified here will be the "fixed" path mtu.
2244 * Note that if the spp_address field is empty
2245 * then all associations on this address will
2246 * have this fixed path mtu set upon them.
2248 * spp_sackdelay - When delayed sack is enabled, this value specifies
2249 * the number of milliseconds that sacks will be delayed
2250 * for. This value will apply to all addresses of an
2251 * association if the spp_address field is empty. Note
2252 * also, that if delayed sack is enabled and this
2253 * value is set to 0, no change is made to the last
2254 * recorded delayed sack timer value.
2256 * spp_flags - These flags are used to control various features
2257 * on an association. The flag field may contain
2258 * zero or more of the following options.
2260 * SPP_HB_ENABLE - Enable heartbeats on the
2261 * specified address. Note that if the address
2262 * field is empty all addresses for the association
2263 * have heartbeats enabled upon them.
2265 * SPP_HB_DISABLE - Disable heartbeats on the
2266 * speicifed address. Note that if the address
2267 * field is empty all addresses for the association
2268 * will have their heartbeats disabled. Note also
2269 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2270 * mutually exclusive, only one of these two should
2271 * be specified. Enabling both fields will have
2272 * undetermined results.
2274 * SPP_HB_DEMAND - Request a user initiated heartbeat
2275 * to be made immediately.
2277 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2278 * heartbeat delayis to be set to the value of 0
2281 * SPP_PMTUD_ENABLE - This field will enable PMTU
2282 * discovery upon the specified address. Note that
2283 * if the address feild is empty then all addresses
2284 * on the association are effected.
2286 * SPP_PMTUD_DISABLE - This field will disable PMTU
2287 * discovery upon the specified address. Note that
2288 * if the address feild is empty then all addresses
2289 * on the association are effected. Not also that
2290 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2291 * exclusive. Enabling both will have undetermined
2294 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2295 * on delayed sack. The time specified in spp_sackdelay
2296 * is used to specify the sack delay for this address. Note
2297 * that if spp_address is empty then all addresses will
2298 * enable delayed sack and take on the sack delay
2299 * value specified in spp_sackdelay.
2300 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2301 * off delayed sack. If the spp_address field is blank then
2302 * delayed sack is disabled for the entire association. Note
2303 * also that this field is mutually exclusive to
2304 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2307 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2308 struct sctp_transport *trans,
2309 struct sctp_association *asoc,
2310 struct sctp_sock *sp,
2313 int sackdelay_change)
2317 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2318 struct net *net = sock_net(trans->asoc->base.sk);
2320 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2325 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2326 * this field is ignored. Note also that a value of zero indicates
2327 * the current setting should be left unchanged.
2329 if (params->spp_flags & SPP_HB_ENABLE) {
2331 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2332 * set. This lets us use 0 value when this flag
2335 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2336 params->spp_hbinterval = 0;
2338 if (params->spp_hbinterval ||
2339 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2342 msecs_to_jiffies(params->spp_hbinterval);
2345 msecs_to_jiffies(params->spp_hbinterval);
2347 sp->hbinterval = params->spp_hbinterval;
2354 trans->param_flags =
2355 (trans->param_flags & ~SPP_HB) | hb_change;
2358 (asoc->param_flags & ~SPP_HB) | hb_change;
2361 (sp->param_flags & ~SPP_HB) | hb_change;
2365 /* When Path MTU discovery is disabled the value specified here will
2366 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2367 * include the flag SPP_PMTUD_DISABLE for this field to have any
2370 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2372 trans->pathmtu = params->spp_pathmtu;
2373 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2375 asoc->pathmtu = params->spp_pathmtu;
2376 sctp_frag_point(asoc, params->spp_pathmtu);
2378 sp->pathmtu = params->spp_pathmtu;
2384 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2385 (params->spp_flags & SPP_PMTUD_ENABLE);
2386 trans->param_flags =
2387 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2389 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2390 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2394 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2397 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2401 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2402 * value of this field is ignored. Note also that a value of zero
2403 * indicates the current setting should be left unchanged.
2405 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2408 msecs_to_jiffies(params->spp_sackdelay);
2411 msecs_to_jiffies(params->spp_sackdelay);
2413 sp->sackdelay = params->spp_sackdelay;
2417 if (sackdelay_change) {
2419 trans->param_flags =
2420 (trans->param_flags & ~SPP_SACKDELAY) |
2424 (asoc->param_flags & ~SPP_SACKDELAY) |
2428 (sp->param_flags & ~SPP_SACKDELAY) |
2433 /* Note that a value of zero indicates the current setting should be
2436 if (params->spp_pathmaxrxt) {
2438 trans->pathmaxrxt = params->spp_pathmaxrxt;
2440 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2442 sp->pathmaxrxt = params->spp_pathmaxrxt;
2449 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2450 char __user *optval,
2451 unsigned int optlen)
2453 struct sctp_paddrparams params;
2454 struct sctp_transport *trans = NULL;
2455 struct sctp_association *asoc = NULL;
2456 struct sctp_sock *sp = sctp_sk(sk);
2458 int hb_change, pmtud_change, sackdelay_change;
2460 if (optlen != sizeof(struct sctp_paddrparams))
2463 if (copy_from_user(¶ms, optval, optlen))
2466 /* Validate flags and value parameters. */
2467 hb_change = params.spp_flags & SPP_HB;
2468 pmtud_change = params.spp_flags & SPP_PMTUD;
2469 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2471 if (hb_change == SPP_HB ||
2472 pmtud_change == SPP_PMTUD ||
2473 sackdelay_change == SPP_SACKDELAY ||
2474 params.spp_sackdelay > 500 ||
2475 (params.spp_pathmtu &&
2476 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2479 /* If an address other than INADDR_ANY is specified, and
2480 * no transport is found, then the request is invalid.
2482 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
2483 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2484 params.spp_assoc_id);
2489 /* Get association, if assoc_id != 0 and the socket is a one
2490 * to many style socket, and an association was not found, then
2491 * the id was invalid.
2493 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2494 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2497 /* Heartbeat demand can only be sent on a transport or
2498 * association, but not a socket.
2500 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2503 /* Process parameters. */
2504 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2505 hb_change, pmtud_change,
2511 /* If changes are for association, also apply parameters to each
2514 if (!trans && asoc) {
2515 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2517 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2518 hb_change, pmtud_change,
2527 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2529 * This option will effect the way delayed acks are performed. This
2530 * option allows you to get or set the delayed ack time, in
2531 * milliseconds. It also allows changing the delayed ack frequency.
2532 * Changing the frequency to 1 disables the delayed sack algorithm. If
2533 * the assoc_id is 0, then this sets or gets the endpoints default
2534 * values. If the assoc_id field is non-zero, then the set or get
2535 * effects the specified association for the one to many model (the
2536 * assoc_id field is ignored by the one to one model). Note that if
2537 * sack_delay or sack_freq are 0 when setting this option, then the
2538 * current values will remain unchanged.
2540 * struct sctp_sack_info {
2541 * sctp_assoc_t sack_assoc_id;
2542 * uint32_t sack_delay;
2543 * uint32_t sack_freq;
2546 * sack_assoc_id - This parameter, indicates which association the user
2547 * is performing an action upon. Note that if this field's value is
2548 * zero then the endpoints default value is changed (effecting future
2549 * associations only).
2551 * sack_delay - This parameter contains the number of milliseconds that
2552 * the user is requesting the delayed ACK timer be set to. Note that
2553 * this value is defined in the standard to be between 200 and 500
2556 * sack_freq - This parameter contains the number of packets that must
2557 * be received before a sack is sent without waiting for the delay
2558 * timer to expire. The default value for this is 2, setting this
2559 * value to 1 will disable the delayed sack algorithm.
2562 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2563 char __user *optval, unsigned int optlen)
2565 struct sctp_sack_info params;
2566 struct sctp_transport *trans = NULL;
2567 struct sctp_association *asoc = NULL;
2568 struct sctp_sock *sp = sctp_sk(sk);
2570 if (optlen == sizeof(struct sctp_sack_info)) {
2571 if (copy_from_user(¶ms, optval, optlen))
2574 if (params.sack_delay == 0 && params.sack_freq == 0)
2576 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2577 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2578 pr_warn("Use struct sctp_sack_info instead\n");
2579 if (copy_from_user(¶ms, optval, optlen))
2582 if (params.sack_delay == 0)
2583 params.sack_freq = 1;
2585 params.sack_freq = 0;
2589 /* Validate value parameter. */
2590 if (params.sack_delay > 500)
2593 /* Get association, if sack_assoc_id != 0 and the socket is a one
2594 * to many style socket, and an association was not found, then
2595 * the id was invalid.
2597 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2598 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2601 if (params.sack_delay) {
2604 msecs_to_jiffies(params.sack_delay);
2606 (asoc->param_flags & ~SPP_SACKDELAY) |
2607 SPP_SACKDELAY_ENABLE;
2609 sp->sackdelay = params.sack_delay;
2611 (sp->param_flags & ~SPP_SACKDELAY) |
2612 SPP_SACKDELAY_ENABLE;
2616 if (params.sack_freq == 1) {
2619 (asoc->param_flags & ~SPP_SACKDELAY) |
2620 SPP_SACKDELAY_DISABLE;
2623 (sp->param_flags & ~SPP_SACKDELAY) |
2624 SPP_SACKDELAY_DISABLE;
2626 } else if (params.sack_freq > 1) {
2628 asoc->sackfreq = params.sack_freq;
2630 (asoc->param_flags & ~SPP_SACKDELAY) |
2631 SPP_SACKDELAY_ENABLE;
2633 sp->sackfreq = params.sack_freq;
2635 (sp->param_flags & ~SPP_SACKDELAY) |
2636 SPP_SACKDELAY_ENABLE;
2640 /* If change is for association, also apply to each transport. */
2642 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2644 if (params.sack_delay) {
2646 msecs_to_jiffies(params.sack_delay);
2647 trans->param_flags =
2648 (trans->param_flags & ~SPP_SACKDELAY) |
2649 SPP_SACKDELAY_ENABLE;
2651 if (params.sack_freq == 1) {
2652 trans->param_flags =
2653 (trans->param_flags & ~SPP_SACKDELAY) |
2654 SPP_SACKDELAY_DISABLE;
2655 } else if (params.sack_freq > 1) {
2656 trans->sackfreq = params.sack_freq;
2657 trans->param_flags =
2658 (trans->param_flags & ~SPP_SACKDELAY) |
2659 SPP_SACKDELAY_ENABLE;
2667 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2669 * Applications can specify protocol parameters for the default association
2670 * initialization. The option name argument to setsockopt() and getsockopt()
2673 * Setting initialization parameters is effective only on an unconnected
2674 * socket (for UDP-style sockets only future associations are effected
2675 * by the change). With TCP-style sockets, this option is inherited by
2676 * sockets derived from a listener socket.
2678 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2680 struct sctp_initmsg sinit;
2681 struct sctp_sock *sp = sctp_sk(sk);
2683 if (optlen != sizeof(struct sctp_initmsg))
2685 if (copy_from_user(&sinit, optval, optlen))
2688 if (sinit.sinit_num_ostreams)
2689 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2690 if (sinit.sinit_max_instreams)
2691 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2692 if (sinit.sinit_max_attempts)
2693 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2694 if (sinit.sinit_max_init_timeo)
2695 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2701 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2703 * Applications that wish to use the sendto() system call may wish to
2704 * specify a default set of parameters that would normally be supplied
2705 * through the inclusion of ancillary data. This socket option allows
2706 * such an application to set the default sctp_sndrcvinfo structure.
2707 * The application that wishes to use this socket option simply passes
2708 * in to this call the sctp_sndrcvinfo structure defined in Section
2709 * 5.2.2) The input parameters accepted by this call include
2710 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2711 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2712 * to this call if the caller is using the UDP model.
2714 static int sctp_setsockopt_default_send_param(struct sock *sk,
2715 char __user *optval,
2716 unsigned int optlen)
2718 struct sctp_sndrcvinfo info;
2719 struct sctp_association *asoc;
2720 struct sctp_sock *sp = sctp_sk(sk);
2722 if (optlen != sizeof(struct sctp_sndrcvinfo))
2724 if (copy_from_user(&info, optval, optlen))
2727 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2728 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2732 asoc->default_stream = info.sinfo_stream;
2733 asoc->default_flags = info.sinfo_flags;
2734 asoc->default_ppid = info.sinfo_ppid;
2735 asoc->default_context = info.sinfo_context;
2736 asoc->default_timetolive = info.sinfo_timetolive;
2738 sp->default_stream = info.sinfo_stream;
2739 sp->default_flags = info.sinfo_flags;
2740 sp->default_ppid = info.sinfo_ppid;
2741 sp->default_context = info.sinfo_context;
2742 sp->default_timetolive = info.sinfo_timetolive;
2748 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2750 * Requests that the local SCTP stack use the enclosed peer address as
2751 * the association primary. The enclosed address must be one of the
2752 * association peer's addresses.
2754 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2755 unsigned int optlen)
2757 struct sctp_prim prim;
2758 struct sctp_transport *trans;
2760 if (optlen != sizeof(struct sctp_prim))
2763 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2766 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2770 sctp_assoc_set_primary(trans->asoc, trans);
2776 * 7.1.5 SCTP_NODELAY
2778 * Turn on/off any Nagle-like algorithm. This means that packets are
2779 * generally sent as soon as possible and no unnecessary delays are
2780 * introduced, at the cost of more packets in the network. Expects an
2781 * integer boolean flag.
2783 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2784 unsigned int optlen)
2788 if (optlen < sizeof(int))
2790 if (get_user(val, (int __user *)optval))
2793 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2799 * 7.1.1 SCTP_RTOINFO
2801 * The protocol parameters used to initialize and bound retransmission
2802 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2803 * and modify these parameters.
2804 * All parameters are time values, in milliseconds. A value of 0, when
2805 * modifying the parameters, indicates that the current value should not
2809 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2811 struct sctp_rtoinfo rtoinfo;
2812 struct sctp_association *asoc;
2814 if (optlen != sizeof (struct sctp_rtoinfo))
2817 if (copy_from_user(&rtoinfo, optval, optlen))
2820 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2822 /* Set the values to the specific association */
2823 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2827 if (rtoinfo.srto_initial != 0)
2829 msecs_to_jiffies(rtoinfo.srto_initial);
2830 if (rtoinfo.srto_max != 0)
2831 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2832 if (rtoinfo.srto_min != 0)
2833 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2835 /* If there is no association or the association-id = 0
2836 * set the values to the endpoint.
2838 struct sctp_sock *sp = sctp_sk(sk);
2840 if (rtoinfo.srto_initial != 0)
2841 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2842 if (rtoinfo.srto_max != 0)
2843 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2844 if (rtoinfo.srto_min != 0)
2845 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2853 * 7.1.2 SCTP_ASSOCINFO
2855 * This option is used to tune the maximum retransmission attempts
2856 * of the association.
2857 * Returns an error if the new association retransmission value is
2858 * greater than the sum of the retransmission value of the peer.
2859 * See [SCTP] for more information.
2862 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2865 struct sctp_assocparams assocparams;
2866 struct sctp_association *asoc;
2868 if (optlen != sizeof(struct sctp_assocparams))
2870 if (copy_from_user(&assocparams, optval, optlen))
2873 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2875 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2878 /* Set the values to the specific association */
2880 if (assocparams.sasoc_asocmaxrxt != 0) {
2883 struct sctp_transport *peer_addr;
2885 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2887 path_sum += peer_addr->pathmaxrxt;
2891 /* Only validate asocmaxrxt if we have more than
2892 * one path/transport. We do this because path
2893 * retransmissions are only counted when we have more
2897 assocparams.sasoc_asocmaxrxt > path_sum)
2900 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2903 if (assocparams.sasoc_cookie_life != 0)
2904 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
2906 /* Set the values to the endpoint */
2907 struct sctp_sock *sp = sctp_sk(sk);
2909 if (assocparams.sasoc_asocmaxrxt != 0)
2910 sp->assocparams.sasoc_asocmaxrxt =
2911 assocparams.sasoc_asocmaxrxt;
2912 if (assocparams.sasoc_cookie_life != 0)
2913 sp->assocparams.sasoc_cookie_life =
2914 assocparams.sasoc_cookie_life;
2920 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2922 * This socket option is a boolean flag which turns on or off mapped V4
2923 * addresses. If this option is turned on and the socket is type
2924 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2925 * If this option is turned off, then no mapping will be done of V4
2926 * addresses and a user will receive both PF_INET6 and PF_INET type
2927 * addresses on the socket.
2929 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2932 struct sctp_sock *sp = sctp_sk(sk);
2934 if (optlen < sizeof(int))
2936 if (get_user(val, (int __user *)optval))
2947 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2948 * This option will get or set the maximum size to put in any outgoing
2949 * SCTP DATA chunk. If a message is larger than this size it will be
2950 * fragmented by SCTP into the specified size. Note that the underlying
2951 * SCTP implementation may fragment into smaller sized chunks when the
2952 * PMTU of the underlying association is smaller than the value set by
2953 * the user. The default value for this option is '0' which indicates
2954 * the user is NOT limiting fragmentation and only the PMTU will effect
2955 * SCTP's choice of DATA chunk size. Note also that values set larger
2956 * than the maximum size of an IP datagram will effectively let SCTP
2957 * control fragmentation (i.e. the same as setting this option to 0).
2959 * The following structure is used to access and modify this parameter:
2961 * struct sctp_assoc_value {
2962 * sctp_assoc_t assoc_id;
2963 * uint32_t assoc_value;
2966 * assoc_id: This parameter is ignored for one-to-one style sockets.
2967 * For one-to-many style sockets this parameter indicates which
2968 * association the user is performing an action upon. Note that if
2969 * this field's value is zero then the endpoints default value is
2970 * changed (effecting future associations only).
2971 * assoc_value: This parameter specifies the maximum size in bytes.
2973 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2975 struct sctp_assoc_value params;
2976 struct sctp_association *asoc;
2977 struct sctp_sock *sp = sctp_sk(sk);
2980 if (optlen == sizeof(int)) {
2981 pr_warn("Use of int in maxseg socket option deprecated\n");
2982 pr_warn("Use struct sctp_assoc_value instead\n");
2983 if (copy_from_user(&val, optval, optlen))
2985 params.assoc_id = 0;
2986 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2987 if (copy_from_user(¶ms, optval, optlen))
2989 val = params.assoc_value;
2993 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2996 asoc = sctp_id2assoc(sk, params.assoc_id);
2997 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3002 val = asoc->pathmtu;
3003 val -= sp->pf->af->net_header_len;
3004 val -= sizeof(struct sctphdr) +
3005 sizeof(struct sctp_data_chunk);
3007 asoc->user_frag = val;
3008 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3010 sp->user_frag = val;
3018 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3020 * Requests that the peer mark the enclosed address as the association
3021 * primary. The enclosed address must be one of the association's
3022 * locally bound addresses. The following structure is used to make a
3023 * set primary request:
3025 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3026 unsigned int optlen)
3028 struct net *net = sock_net(sk);
3029 struct sctp_sock *sp;
3030 struct sctp_association *asoc = NULL;
3031 struct sctp_setpeerprim prim;
3032 struct sctp_chunk *chunk;
3038 if (!net->sctp.addip_enable)
3041 if (optlen != sizeof(struct sctp_setpeerprim))
3044 if (copy_from_user(&prim, optval, optlen))
3047 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3051 if (!asoc->peer.asconf_capable)
3054 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3057 if (!sctp_state(asoc, ESTABLISHED))
3060 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3064 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3065 return -EADDRNOTAVAIL;
3067 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3068 return -EADDRNOTAVAIL;
3070 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3071 chunk = sctp_make_asconf_set_prim(asoc,
3072 (union sctp_addr *)&prim.sspp_addr);
3076 err = sctp_send_asconf(asoc, chunk);
3078 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3083 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3084 unsigned int optlen)
3086 struct sctp_setadaptation adaptation;
3088 if (optlen != sizeof(struct sctp_setadaptation))
3090 if (copy_from_user(&adaptation, optval, optlen))
3093 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3099 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3101 * The context field in the sctp_sndrcvinfo structure is normally only
3102 * used when a failed message is retrieved holding the value that was
3103 * sent down on the actual send call. This option allows the setting of
3104 * a default context on an association basis that will be received on
3105 * reading messages from the peer. This is especially helpful in the
3106 * one-2-many model for an application to keep some reference to an
3107 * internal state machine that is processing messages on the
3108 * association. Note that the setting of this value only effects
3109 * received messages from the peer and does not effect the value that is
3110 * saved with outbound messages.
3112 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3113 unsigned int optlen)
3115 struct sctp_assoc_value params;
3116 struct sctp_sock *sp;
3117 struct sctp_association *asoc;
3119 if (optlen != sizeof(struct sctp_assoc_value))
3121 if (copy_from_user(¶ms, optval, optlen))
3126 if (params.assoc_id != 0) {
3127 asoc = sctp_id2assoc(sk, params.assoc_id);
3130 asoc->default_rcv_context = params.assoc_value;
3132 sp->default_rcv_context = params.assoc_value;
3139 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3141 * This options will at a minimum specify if the implementation is doing
3142 * fragmented interleave. Fragmented interleave, for a one to many
3143 * socket, is when subsequent calls to receive a message may return
3144 * parts of messages from different associations. Some implementations
3145 * may allow you to turn this value on or off. If so, when turned off,
3146 * no fragment interleave will occur (which will cause a head of line
3147 * blocking amongst multiple associations sharing the same one to many
3148 * socket). When this option is turned on, then each receive call may
3149 * come from a different association (thus the user must receive data
3150 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3151 * association each receive belongs to.
3153 * This option takes a boolean value. A non-zero value indicates that
3154 * fragmented interleave is on. A value of zero indicates that
3155 * fragmented interleave is off.
3157 * Note that it is important that an implementation that allows this
3158 * option to be turned on, have it off by default. Otherwise an unaware
3159 * application using the one to many model may become confused and act
3162 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3163 char __user *optval,
3164 unsigned int optlen)
3168 if (optlen != sizeof(int))
3170 if (get_user(val, (int __user *)optval))
3173 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3179 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3180 * (SCTP_PARTIAL_DELIVERY_POINT)
3182 * This option will set or get the SCTP partial delivery point. This
3183 * point is the size of a message where the partial delivery API will be
3184 * invoked to help free up rwnd space for the peer. Setting this to a
3185 * lower value will cause partial deliveries to happen more often. The
3186 * calls argument is an integer that sets or gets the partial delivery
3187 * point. Note also that the call will fail if the user attempts to set
3188 * this value larger than the socket receive buffer size.
3190 * Note that any single message having a length smaller than or equal to
3191 * the SCTP partial delivery point will be delivered in one single read
3192 * call as long as the user provided buffer is large enough to hold the
3195 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3196 char __user *optval,
3197 unsigned int optlen)
3201 if (optlen != sizeof(u32))
3203 if (get_user(val, (int __user *)optval))
3206 /* Note: We double the receive buffer from what the user sets
3207 * it to be, also initial rwnd is based on rcvbuf/2.
3209 if (val > (sk->sk_rcvbuf >> 1))
3212 sctp_sk(sk)->pd_point = val;
3214 return 0; /* is this the right error code? */
3218 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3220 * This option will allow a user to change the maximum burst of packets
3221 * that can be emitted by this association. Note that the default value
3222 * is 4, and some implementations may restrict this setting so that it
3223 * can only be lowered.
3225 * NOTE: This text doesn't seem right. Do this on a socket basis with
3226 * future associations inheriting the socket value.
3228 static int sctp_setsockopt_maxburst(struct sock *sk,
3229 char __user *optval,
3230 unsigned int optlen)
3232 struct sctp_assoc_value params;
3233 struct sctp_sock *sp;
3234 struct sctp_association *asoc;
3238 if (optlen == sizeof(int)) {
3239 pr_warn("Use of int in max_burst socket option deprecated\n");
3240 pr_warn("Use struct sctp_assoc_value instead\n");
3241 if (copy_from_user(&val, optval, optlen))
3243 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3244 if (copy_from_user(¶ms, optval, optlen))
3246 val = params.assoc_value;
3247 assoc_id = params.assoc_id;
3253 if (assoc_id != 0) {
3254 asoc = sctp_id2assoc(sk, assoc_id);
3257 asoc->max_burst = val;
3259 sp->max_burst = val;
3265 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3267 * This set option adds a chunk type that the user is requesting to be
3268 * received only in an authenticated way. Changes to the list of chunks
3269 * will only effect future associations on the socket.
3271 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3272 char __user *optval,
3273 unsigned int optlen)
3275 struct net *net = sock_net(sk);
3276 struct sctp_authchunk val;
3278 if (!net->sctp.auth_enable)
3281 if (optlen != sizeof(struct sctp_authchunk))
3283 if (copy_from_user(&val, optval, optlen))
3286 switch (val.sauth_chunk) {
3288 case SCTP_CID_INIT_ACK:
3289 case SCTP_CID_SHUTDOWN_COMPLETE:
3294 /* add this chunk id to the endpoint */
3295 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3299 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3301 * This option gets or sets the list of HMAC algorithms that the local
3302 * endpoint requires the peer to use.
3304 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3305 char __user *optval,
3306 unsigned int optlen)
3308 struct net *net = sock_net(sk);
3309 struct sctp_hmacalgo *hmacs;
3313 if (!net->sctp.auth_enable)
3316 if (optlen < sizeof(struct sctp_hmacalgo))
3319 hmacs= memdup_user(optval, optlen);
3321 return PTR_ERR(hmacs);
3323 idents = hmacs->shmac_num_idents;
3324 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3325 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3330 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3337 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3339 * This option will set a shared secret key which is used to build an
3340 * association shared key.
3342 static int sctp_setsockopt_auth_key(struct sock *sk,
3343 char __user *optval,
3344 unsigned int optlen)
3346 struct net *net = sock_net(sk);
3347 struct sctp_authkey *authkey;
3348 struct sctp_association *asoc;
3351 if (!net->sctp.auth_enable)
3354 if (optlen <= sizeof(struct sctp_authkey))
3357 authkey= memdup_user(optval, optlen);
3358 if (IS_ERR(authkey))
3359 return PTR_ERR(authkey);
3361 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3366 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3367 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3372 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3379 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3381 * This option will get or set the active shared key to be used to build
3382 * the association shared key.
3384 static int sctp_setsockopt_active_key(struct sock *sk,
3385 char __user *optval,
3386 unsigned int optlen)
3388 struct net *net = sock_net(sk);
3389 struct sctp_authkeyid val;
3390 struct sctp_association *asoc;
3392 if (!net->sctp.auth_enable)
3395 if (optlen != sizeof(struct sctp_authkeyid))
3397 if (copy_from_user(&val, optval, optlen))
3400 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3401 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3404 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3405 val.scact_keynumber);
3409 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3411 * This set option will delete a shared secret key from use.
3413 static int sctp_setsockopt_del_key(struct sock *sk,
3414 char __user *optval,
3415 unsigned int optlen)
3417 struct net *net = sock_net(sk);
3418 struct sctp_authkeyid val;
3419 struct sctp_association *asoc;
3421 if (!net->sctp.auth_enable)
3424 if (optlen != sizeof(struct sctp_authkeyid))
3426 if (copy_from_user(&val, optval, optlen))
3429 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3430 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3433 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3434 val.scact_keynumber);
3439 * 8.1.23 SCTP_AUTO_ASCONF
3441 * This option will enable or disable the use of the automatic generation of
3442 * ASCONF chunks to add and delete addresses to an existing association. Note
3443 * that this option has two caveats namely: a) it only affects sockets that
3444 * are bound to all addresses available to the SCTP stack, and b) the system
3445 * administrator may have an overriding control that turns the ASCONF feature
3446 * off no matter what setting the socket option may have.
3447 * This option expects an integer boolean flag, where a non-zero value turns on
3448 * the option, and a zero value turns off the option.
3449 * Note. In this implementation, socket operation overrides default parameter
3450 * being set by sysctl as well as FreeBSD implementation
3452 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3453 unsigned int optlen)
3456 struct sctp_sock *sp = sctp_sk(sk);
3458 if (optlen < sizeof(int))
3460 if (get_user(val, (int __user *)optval))
3462 if (!sctp_is_ep_boundall(sk) && val)
3464 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3467 if (val == 0 && sp->do_auto_asconf) {
3468 list_del(&sp->auto_asconf_list);
3469 sp->do_auto_asconf = 0;
3470 } else if (val && !sp->do_auto_asconf) {
3471 list_add_tail(&sp->auto_asconf_list,
3472 &sock_net(sk)->sctp.auto_asconf_splist);
3473 sp->do_auto_asconf = 1;
3480 * SCTP_PEER_ADDR_THLDS
3482 * This option allows us to alter the partially failed threshold for one or all
3483 * transports in an association. See Section 6.1 of:
3484 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3486 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3487 char __user *optval,
3488 unsigned int optlen)
3490 struct sctp_paddrthlds val;
3491 struct sctp_transport *trans;
3492 struct sctp_association *asoc;
3494 if (optlen < sizeof(struct sctp_paddrthlds))
3496 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3497 sizeof(struct sctp_paddrthlds)))
3501 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3502 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3505 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3507 if (val.spt_pathmaxrxt)
3508 trans->pathmaxrxt = val.spt_pathmaxrxt;
3509 trans->pf_retrans = val.spt_pathpfthld;
3512 if (val.spt_pathmaxrxt)
3513 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3514 asoc->pf_retrans = val.spt_pathpfthld;
3516 trans = sctp_addr_id2transport(sk, &val.spt_address,
3521 if (val.spt_pathmaxrxt)
3522 trans->pathmaxrxt = val.spt_pathmaxrxt;
3523 trans->pf_retrans = val.spt_pathpfthld;
3529 /* API 6.2 setsockopt(), getsockopt()
3531 * Applications use setsockopt() and getsockopt() to set or retrieve
3532 * socket options. Socket options are used to change the default
3533 * behavior of sockets calls. They are described in Section 7.
3537 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3538 * int __user *optlen);
3539 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3542 * sd - the socket descript.
3543 * level - set to IPPROTO_SCTP for all SCTP options.
3544 * optname - the option name.
3545 * optval - the buffer to store the value of the option.
3546 * optlen - the size of the buffer.
3548 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3549 char __user *optval, unsigned int optlen)
3553 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3555 /* I can hardly begin to describe how wrong this is. This is
3556 * so broken as to be worse than useless. The API draft
3557 * REALLY is NOT helpful here... I am not convinced that the
3558 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3559 * are at all well-founded.
3561 if (level != SOL_SCTP) {
3562 struct sctp_af *af = sctp_sk(sk)->pf->af;
3563 retval = af->setsockopt(sk, level, optname, optval, optlen);
3570 case SCTP_SOCKOPT_BINDX_ADD:
3571 /* 'optlen' is the size of the addresses buffer. */
3572 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3573 optlen, SCTP_BINDX_ADD_ADDR);
3576 case SCTP_SOCKOPT_BINDX_REM:
3577 /* 'optlen' is the size of the addresses buffer. */
3578 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3579 optlen, SCTP_BINDX_REM_ADDR);
3582 case SCTP_SOCKOPT_CONNECTX_OLD:
3583 /* 'optlen' is the size of the addresses buffer. */
3584 retval = sctp_setsockopt_connectx_old(sk,
3585 (struct sockaddr __user *)optval,
3589 case SCTP_SOCKOPT_CONNECTX:
3590 /* 'optlen' is the size of the addresses buffer. */
3591 retval = sctp_setsockopt_connectx(sk,
3592 (struct sockaddr __user *)optval,
3596 case SCTP_DISABLE_FRAGMENTS:
3597 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3601 retval = sctp_setsockopt_events(sk, optval, optlen);
3604 case SCTP_AUTOCLOSE:
3605 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3608 case SCTP_PEER_ADDR_PARAMS:
3609 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3612 case SCTP_DELAYED_SACK:
3613 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3615 case SCTP_PARTIAL_DELIVERY_POINT:
3616 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3620 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3622 case SCTP_DEFAULT_SEND_PARAM:
3623 retval = sctp_setsockopt_default_send_param(sk, optval,
3626 case SCTP_PRIMARY_ADDR:
3627 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3629 case SCTP_SET_PEER_PRIMARY_ADDR:
3630 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3633 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3636 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3638 case SCTP_ASSOCINFO:
3639 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3641 case SCTP_I_WANT_MAPPED_V4_ADDR:
3642 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3645 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3647 case SCTP_ADAPTATION_LAYER:
3648 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3651 retval = sctp_setsockopt_context(sk, optval, optlen);
3653 case SCTP_FRAGMENT_INTERLEAVE:
3654 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3656 case SCTP_MAX_BURST:
3657 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3659 case SCTP_AUTH_CHUNK:
3660 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3662 case SCTP_HMAC_IDENT:
3663 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3666 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3668 case SCTP_AUTH_ACTIVE_KEY:
3669 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3671 case SCTP_AUTH_DELETE_KEY:
3672 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3674 case SCTP_AUTO_ASCONF:
3675 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3677 case SCTP_PEER_ADDR_THLDS:
3678 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3681 retval = -ENOPROTOOPT;
3685 sctp_release_sock(sk);
3691 /* API 3.1.6 connect() - UDP Style Syntax
3693 * An application may use the connect() call in the UDP model to initiate an
3694 * association without sending data.
3698 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3700 * sd: the socket descriptor to have a new association added to.
3702 * nam: the address structure (either struct sockaddr_in or struct
3703 * sockaddr_in6 defined in RFC2553 [7]).
3705 * len: the size of the address.
3707 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3715 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
3718 /* Validate addr_len before calling common connect/connectx routine. */
3719 af = sctp_get_af_specific(addr->sa_family);
3720 if (!af || addr_len < af->sockaddr_len) {
3723 /* Pass correct addr len to common routine (so it knows there
3724 * is only one address being passed.
3726 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3729 sctp_release_sock(sk);
3733 /* FIXME: Write comments. */
3734 static int sctp_disconnect(struct sock *sk, int flags)
3736 return -EOPNOTSUPP; /* STUB */
3739 /* 4.1.4 accept() - TCP Style Syntax
3741 * Applications use accept() call to remove an established SCTP
3742 * association from the accept queue of the endpoint. A new socket
3743 * descriptor will be returned from accept() to represent the newly
3744 * formed association.
3746 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3748 struct sctp_sock *sp;
3749 struct sctp_endpoint *ep;
3750 struct sock *newsk = NULL;
3751 struct sctp_association *asoc;
3760 if (!sctp_style(sk, TCP)) {
3761 error = -EOPNOTSUPP;
3765 if (!sctp_sstate(sk, LISTENING)) {
3770 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3772 error = sctp_wait_for_accept(sk, timeo);
3776 /* We treat the list of associations on the endpoint as the accept
3777 * queue and pick the first association on the list.
3779 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3781 newsk = sp->pf->create_accept_sk(sk, asoc);
3787 /* Populate the fields of the newsk from the oldsk and migrate the
3788 * asoc to the newsk.
3790 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3793 sctp_release_sock(sk);
3798 /* The SCTP ioctl handler. */
3799 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3806 * SEQPACKET-style sockets in LISTENING state are valid, for
3807 * SCTP, so only discard TCP-style sockets in LISTENING state.
3809 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3814 struct sk_buff *skb;
3815 unsigned int amount = 0;
3817 skb = skb_peek(&sk->sk_receive_queue);
3820 * We will only return the amount of this packet since
3821 * that is all that will be read.
3825 rc = put_user(amount, (int __user *)arg);
3833 sctp_release_sock(sk);
3837 /* This is the function which gets called during socket creation to
3838 * initialized the SCTP-specific portion of the sock.
3839 * The sock structure should already be zero-filled memory.
3841 static int sctp_init_sock(struct sock *sk)
3843 struct net *net = sock_net(sk);
3844 struct sctp_sock *sp;
3846 pr_debug("%s: sk:%p\n", __func__, sk);
3850 /* Initialize the SCTP per socket area. */
3851 switch (sk->sk_type) {
3852 case SOCK_SEQPACKET:
3853 sp->type = SCTP_SOCKET_UDP;
3856 sp->type = SCTP_SOCKET_TCP;
3859 return -ESOCKTNOSUPPORT;
3862 /* Initialize default send parameters. These parameters can be
3863 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3865 sp->default_stream = 0;
3866 sp->default_ppid = 0;
3867 sp->default_flags = 0;
3868 sp->default_context = 0;
3869 sp->default_timetolive = 0;
3871 sp->default_rcv_context = 0;
3872 sp->max_burst = net->sctp.max_burst;
3874 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
3876 /* Initialize default setup parameters. These parameters
3877 * can be modified with the SCTP_INITMSG socket option or
3878 * overridden by the SCTP_INIT CMSG.
3880 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3881 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3882 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
3883 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
3885 /* Initialize default RTO related parameters. These parameters can
3886 * be modified for with the SCTP_RTOINFO socket option.
3888 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
3889 sp->rtoinfo.srto_max = net->sctp.rto_max;
3890 sp->rtoinfo.srto_min = net->sctp.rto_min;
3892 /* Initialize default association related parameters. These parameters
3893 * can be modified with the SCTP_ASSOCINFO socket option.
3895 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
3896 sp->assocparams.sasoc_number_peer_destinations = 0;
3897 sp->assocparams.sasoc_peer_rwnd = 0;
3898 sp->assocparams.sasoc_local_rwnd = 0;
3899 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
3901 /* Initialize default event subscriptions. By default, all the
3904 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3906 /* Default Peer Address Parameters. These defaults can
3907 * be modified via SCTP_PEER_ADDR_PARAMS
3909 sp->hbinterval = net->sctp.hb_interval;
3910 sp->pathmaxrxt = net->sctp.max_retrans_path;
3911 sp->pathmtu = 0; // allow default discovery
3912 sp->sackdelay = net->sctp.sack_timeout;
3914 sp->param_flags = SPP_HB_ENABLE |
3916 SPP_SACKDELAY_ENABLE;
3918 /* If enabled no SCTP message fragmentation will be performed.
3919 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3921 sp->disable_fragments = 0;
3923 /* Enable Nagle algorithm by default. */
3926 /* Enable by default. */
3929 /* Auto-close idle associations after the configured
3930 * number of seconds. A value of 0 disables this
3931 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3932 * for UDP-style sockets only.
3936 /* User specified fragmentation limit. */
3939 sp->adaptation_ind = 0;
3941 sp->pf = sctp_get_pf_specific(sk->sk_family);
3943 /* Control variables for partial data delivery. */
3944 atomic_set(&sp->pd_mode, 0);
3945 skb_queue_head_init(&sp->pd_lobby);
3946 sp->frag_interleave = 0;
3948 /* Create a per socket endpoint structure. Even if we
3949 * change the data structure relationships, this may still
3950 * be useful for storing pre-connect address information.
3952 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
3958 sk->sk_destruct = sctp_destruct_sock;
3960 SCTP_DBG_OBJCNT_INC(sock);
3963 percpu_counter_inc(&sctp_sockets_allocated);
3964 sock_prot_inuse_add(net, sk->sk_prot, 1);
3965 if (net->sctp.default_auto_asconf) {
3966 list_add_tail(&sp->auto_asconf_list,
3967 &net->sctp.auto_asconf_splist);
3968 sp->do_auto_asconf = 1;
3970 sp->do_auto_asconf = 0;
3976 /* Cleanup any SCTP per socket resources. */
3977 static void sctp_destroy_sock(struct sock *sk)
3979 struct sctp_sock *sp;
3981 pr_debug("%s: sk:%p\n", __func__, sk);
3983 /* Release our hold on the endpoint. */
3985 /* This could happen during socket init, thus we bail out
3986 * early, since the rest of the below is not setup either.
3991 if (sp->do_auto_asconf) {
3992 sp->do_auto_asconf = 0;
3993 list_del(&sp->auto_asconf_list);
3995 sctp_endpoint_free(sp->ep);
3997 percpu_counter_dec(&sctp_sockets_allocated);
3998 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4002 /* Triggered when there are no references on the socket anymore */
4003 static void sctp_destruct_sock(struct sock *sk)
4005 struct sctp_sock *sp = sctp_sk(sk);
4007 /* Free up the HMAC transform. */
4008 crypto_free_hash(sp->hmac);
4010 inet_sock_destruct(sk);
4013 /* API 4.1.7 shutdown() - TCP Style Syntax
4014 * int shutdown(int socket, int how);
4016 * sd - the socket descriptor of the association to be closed.
4017 * how - Specifies the type of shutdown. The values are
4020 * Disables further receive operations. No SCTP
4021 * protocol action is taken.
4023 * Disables further send operations, and initiates
4024 * the SCTP shutdown sequence.
4026 * Disables further send and receive operations
4027 * and initiates the SCTP shutdown sequence.
4029 static void sctp_shutdown(struct sock *sk, int how)
4031 struct net *net = sock_net(sk);
4032 struct sctp_endpoint *ep;
4033 struct sctp_association *asoc;
4035 if (!sctp_style(sk, TCP))
4038 if (how & SEND_SHUTDOWN) {
4039 ep = sctp_sk(sk)->ep;
4040 if (!list_empty(&ep->asocs)) {
4041 asoc = list_entry(ep->asocs.next,
4042 struct sctp_association, asocs);
4043 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4048 /* 7.2.1 Association Status (SCTP_STATUS)
4050 * Applications can retrieve current status information about an
4051 * association, including association state, peer receiver window size,
4052 * number of unacked data chunks, and number of data chunks pending
4053 * receipt. This information is read-only.
4055 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4056 char __user *optval,
4059 struct sctp_status status;
4060 struct sctp_association *asoc = NULL;
4061 struct sctp_transport *transport;
4062 sctp_assoc_t associd;
4065 if (len < sizeof(status)) {
4070 len = sizeof(status);
4071 if (copy_from_user(&status, optval, len)) {
4076 associd = status.sstat_assoc_id;
4077 asoc = sctp_id2assoc(sk, associd);
4083 transport = asoc->peer.primary_path;
4085 status.sstat_assoc_id = sctp_assoc2id(asoc);
4086 status.sstat_state = asoc->state;
4087 status.sstat_rwnd = asoc->peer.rwnd;
4088 status.sstat_unackdata = asoc->unack_data;
4090 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4091 status.sstat_instrms = asoc->c.sinit_max_instreams;
4092 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4093 status.sstat_fragmentation_point = asoc->frag_point;
4094 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4095 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4096 transport->af_specific->sockaddr_len);
4097 /* Map ipv4 address into v4-mapped-on-v6 address. */
4098 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4099 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4100 status.sstat_primary.spinfo_state = transport->state;
4101 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4102 status.sstat_primary.spinfo_srtt = transport->srtt;
4103 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4104 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4106 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4107 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4109 if (put_user(len, optlen)) {
4114 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4115 __func__, len, status.sstat_state, status.sstat_rwnd,
4116 status.sstat_assoc_id);
4118 if (copy_to_user(optval, &status, len)) {
4128 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4130 * Applications can retrieve information about a specific peer address
4131 * of an association, including its reachability state, congestion
4132 * window, and retransmission timer values. This information is
4135 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4136 char __user *optval,
4139 struct sctp_paddrinfo pinfo;
4140 struct sctp_transport *transport;
4143 if (len < sizeof(pinfo)) {
4148 len = sizeof(pinfo);
4149 if (copy_from_user(&pinfo, optval, len)) {
4154 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4155 pinfo.spinfo_assoc_id);
4159 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4160 pinfo.spinfo_state = transport->state;
4161 pinfo.spinfo_cwnd = transport->cwnd;
4162 pinfo.spinfo_srtt = transport->srtt;
4163 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4164 pinfo.spinfo_mtu = transport->pathmtu;
4166 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4167 pinfo.spinfo_state = SCTP_ACTIVE;
4169 if (put_user(len, optlen)) {
4174 if (copy_to_user(optval, &pinfo, len)) {
4183 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4185 * This option is a on/off flag. If enabled no SCTP message
4186 * fragmentation will be performed. Instead if a message being sent
4187 * exceeds the current PMTU size, the message will NOT be sent and
4188 * instead a error will be indicated to the user.
4190 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4191 char __user *optval, int __user *optlen)
4195 if (len < sizeof(int))
4199 val = (sctp_sk(sk)->disable_fragments == 1);
4200 if (put_user(len, optlen))
4202 if (copy_to_user(optval, &val, len))
4207 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4209 * This socket option is used to specify various notifications and
4210 * ancillary data the user wishes to receive.
4212 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4217 if (len > sizeof(struct sctp_event_subscribe))
4218 len = sizeof(struct sctp_event_subscribe);
4219 if (put_user(len, optlen))
4221 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4226 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4228 * This socket option is applicable to the UDP-style socket only. When
4229 * set it will cause associations that are idle for more than the
4230 * specified number of seconds to automatically close. An association
4231 * being idle is defined an association that has NOT sent or received
4232 * user data. The special value of '0' indicates that no automatic
4233 * close of any associations should be performed. The option expects an
4234 * integer defining the number of seconds of idle time before an
4235 * association is closed.
4237 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4239 /* Applicable to UDP-style socket only */
4240 if (sctp_style(sk, TCP))
4242 if (len < sizeof(int))
4245 if (put_user(len, optlen))
4247 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4252 /* Helper routine to branch off an association to a new socket. */
4253 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4255 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4256 struct socket *sock;
4263 /* An association cannot be branched off from an already peeled-off
4264 * socket, nor is this supported for tcp style sockets.
4266 if (!sctp_style(sk, UDP))
4269 /* Create a new socket. */
4270 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4274 sctp_copy_sock(sock->sk, sk, asoc);
4276 /* Make peeled-off sockets more like 1-1 accepted sockets.
4277 * Set the daddr and initialize id to something more random
4279 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4280 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4282 /* Populate the fields of the newsk from the oldsk and migrate the
4283 * asoc to the newsk.
4285 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4291 EXPORT_SYMBOL(sctp_do_peeloff);
4293 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4295 sctp_peeloff_arg_t peeloff;
4296 struct socket *newsock;
4297 struct file *newfile;
4300 if (len < sizeof(sctp_peeloff_arg_t))
4302 len = sizeof(sctp_peeloff_arg_t);
4303 if (copy_from_user(&peeloff, optval, len))
4306 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4310 /* Map the socket to an unused fd that can be returned to the user. */
4311 retval = get_unused_fd_flags(0);
4313 sock_release(newsock);
4317 newfile = sock_alloc_file(newsock, 0, NULL);
4318 if (unlikely(IS_ERR(newfile))) {
4319 put_unused_fd(retval);
4320 sock_release(newsock);
4321 return PTR_ERR(newfile);
4324 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4327 /* Return the fd mapped to the new socket. */
4328 if (put_user(len, optlen)) {
4330 put_unused_fd(retval);
4333 peeloff.sd = retval;
4334 if (copy_to_user(optval, &peeloff, len)) {
4336 put_unused_fd(retval);
4339 fd_install(retval, newfile);
4344 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4346 * Applications can enable or disable heartbeats for any peer address of
4347 * an association, modify an address's heartbeat interval, force a
4348 * heartbeat to be sent immediately, and adjust the address's maximum
4349 * number of retransmissions sent before an address is considered
4350 * unreachable. The following structure is used to access and modify an
4351 * address's parameters:
4353 * struct sctp_paddrparams {
4354 * sctp_assoc_t spp_assoc_id;
4355 * struct sockaddr_storage spp_address;
4356 * uint32_t spp_hbinterval;
4357 * uint16_t spp_pathmaxrxt;
4358 * uint32_t spp_pathmtu;
4359 * uint32_t spp_sackdelay;
4360 * uint32_t spp_flags;
4363 * spp_assoc_id - (one-to-many style socket) This is filled in the
4364 * application, and identifies the association for
4366 * spp_address - This specifies which address is of interest.
4367 * spp_hbinterval - This contains the value of the heartbeat interval,
4368 * in milliseconds. If a value of zero
4369 * is present in this field then no changes are to
4370 * be made to this parameter.
4371 * spp_pathmaxrxt - This contains the maximum number of
4372 * retransmissions before this address shall be
4373 * considered unreachable. If a value of zero
4374 * is present in this field then no changes are to
4375 * be made to this parameter.
4376 * spp_pathmtu - When Path MTU discovery is disabled the value
4377 * specified here will be the "fixed" path mtu.
4378 * Note that if the spp_address field is empty
4379 * then all associations on this address will
4380 * have this fixed path mtu set upon them.
4382 * spp_sackdelay - When delayed sack is enabled, this value specifies
4383 * the number of milliseconds that sacks will be delayed
4384 * for. This value will apply to all addresses of an
4385 * association if the spp_address field is empty. Note
4386 * also, that if delayed sack is enabled and this
4387 * value is set to 0, no change is made to the last
4388 * recorded delayed sack timer value.
4390 * spp_flags - These flags are used to control various features
4391 * on an association. The flag field may contain
4392 * zero or more of the following options.
4394 * SPP_HB_ENABLE - Enable heartbeats on the
4395 * specified address. Note that if the address
4396 * field is empty all addresses for the association
4397 * have heartbeats enabled upon them.
4399 * SPP_HB_DISABLE - Disable heartbeats on the
4400 * speicifed address. Note that if the address
4401 * field is empty all addresses for the association
4402 * will have their heartbeats disabled. Note also
4403 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4404 * mutually exclusive, only one of these two should
4405 * be specified. Enabling both fields will have
4406 * undetermined results.
4408 * SPP_HB_DEMAND - Request a user initiated heartbeat
4409 * to be made immediately.
4411 * SPP_PMTUD_ENABLE - This field will enable PMTU
4412 * discovery upon the specified address. Note that
4413 * if the address feild is empty then all addresses
4414 * on the association are effected.
4416 * SPP_PMTUD_DISABLE - This field will disable PMTU
4417 * discovery upon the specified address. Note that
4418 * if the address feild is empty then all addresses
4419 * on the association are effected. Not also that
4420 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4421 * exclusive. Enabling both will have undetermined
4424 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4425 * on delayed sack. The time specified in spp_sackdelay
4426 * is used to specify the sack delay for this address. Note
4427 * that if spp_address is empty then all addresses will
4428 * enable delayed sack and take on the sack delay
4429 * value specified in spp_sackdelay.
4430 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4431 * off delayed sack. If the spp_address field is blank then
4432 * delayed sack is disabled for the entire association. Note
4433 * also that this field is mutually exclusive to
4434 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4437 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4438 char __user *optval, int __user *optlen)
4440 struct sctp_paddrparams params;
4441 struct sctp_transport *trans = NULL;
4442 struct sctp_association *asoc = NULL;
4443 struct sctp_sock *sp = sctp_sk(sk);
4445 if (len < sizeof(struct sctp_paddrparams))
4447 len = sizeof(struct sctp_paddrparams);
4448 if (copy_from_user(¶ms, optval, len))
4451 /* If an address other than INADDR_ANY is specified, and
4452 * no transport is found, then the request is invalid.
4454 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
4455 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4456 params.spp_assoc_id);
4458 pr_debug("%s: failed no transport\n", __func__);
4463 /* Get association, if assoc_id != 0 and the socket is a one
4464 * to many style socket, and an association was not found, then
4465 * the id was invalid.
4467 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4468 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4469 pr_debug("%s: failed no association\n", __func__);
4474 /* Fetch transport values. */
4475 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4476 params.spp_pathmtu = trans->pathmtu;
4477 params.spp_pathmaxrxt = trans->pathmaxrxt;
4478 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4480 /*draft-11 doesn't say what to return in spp_flags*/
4481 params.spp_flags = trans->param_flags;
4483 /* Fetch association values. */
4484 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4485 params.spp_pathmtu = asoc->pathmtu;
4486 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4487 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4489 /*draft-11 doesn't say what to return in spp_flags*/
4490 params.spp_flags = asoc->param_flags;
4492 /* Fetch socket values. */
4493 params.spp_hbinterval = sp->hbinterval;
4494 params.spp_pathmtu = sp->pathmtu;
4495 params.spp_sackdelay = sp->sackdelay;
4496 params.spp_pathmaxrxt = sp->pathmaxrxt;
4498 /*draft-11 doesn't say what to return in spp_flags*/
4499 params.spp_flags = sp->param_flags;
4502 if (copy_to_user(optval, ¶ms, len))
4505 if (put_user(len, optlen))
4512 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4514 * This option will effect the way delayed acks are performed. This
4515 * option allows you to get or set the delayed ack time, in
4516 * milliseconds. It also allows changing the delayed ack frequency.
4517 * Changing the frequency to 1 disables the delayed sack algorithm. If
4518 * the assoc_id is 0, then this sets or gets the endpoints default
4519 * values. If the assoc_id field is non-zero, then the set or get
4520 * effects the specified association for the one to many model (the
4521 * assoc_id field is ignored by the one to one model). Note that if
4522 * sack_delay or sack_freq are 0 when setting this option, then the
4523 * current values will remain unchanged.
4525 * struct sctp_sack_info {
4526 * sctp_assoc_t sack_assoc_id;
4527 * uint32_t sack_delay;
4528 * uint32_t sack_freq;
4531 * sack_assoc_id - This parameter, indicates which association the user
4532 * is performing an action upon. Note that if this field's value is
4533 * zero then the endpoints default value is changed (effecting future
4534 * associations only).
4536 * sack_delay - This parameter contains the number of milliseconds that
4537 * the user is requesting the delayed ACK timer be set to. Note that
4538 * this value is defined in the standard to be between 200 and 500
4541 * sack_freq - This parameter contains the number of packets that must
4542 * be received before a sack is sent without waiting for the delay
4543 * timer to expire. The default value for this is 2, setting this
4544 * value to 1 will disable the delayed sack algorithm.
4546 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4547 char __user *optval,
4550 struct sctp_sack_info params;
4551 struct sctp_association *asoc = NULL;
4552 struct sctp_sock *sp = sctp_sk(sk);
4554 if (len >= sizeof(struct sctp_sack_info)) {
4555 len = sizeof(struct sctp_sack_info);
4557 if (copy_from_user(¶ms, optval, len))
4559 } else if (len == sizeof(struct sctp_assoc_value)) {
4560 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4561 pr_warn("Use struct sctp_sack_info instead\n");
4562 if (copy_from_user(¶ms, optval, len))
4567 /* Get association, if sack_assoc_id != 0 and the socket is a one
4568 * to many style socket, and an association was not found, then
4569 * the id was invalid.
4571 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4572 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4576 /* Fetch association values. */
4577 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4578 params.sack_delay = jiffies_to_msecs(
4580 params.sack_freq = asoc->sackfreq;
4583 params.sack_delay = 0;
4584 params.sack_freq = 1;
4587 /* Fetch socket values. */
4588 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4589 params.sack_delay = sp->sackdelay;
4590 params.sack_freq = sp->sackfreq;
4592 params.sack_delay = 0;
4593 params.sack_freq = 1;
4597 if (copy_to_user(optval, ¶ms, len))
4600 if (put_user(len, optlen))
4606 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4608 * Applications can specify protocol parameters for the default association
4609 * initialization. The option name argument to setsockopt() and getsockopt()
4612 * Setting initialization parameters is effective only on an unconnected
4613 * socket (for UDP-style sockets only future associations are effected
4614 * by the change). With TCP-style sockets, this option is inherited by
4615 * sockets derived from a listener socket.
4617 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4619 if (len < sizeof(struct sctp_initmsg))
4621 len = sizeof(struct sctp_initmsg);
4622 if (put_user(len, optlen))
4624 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4630 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4631 char __user *optval, int __user *optlen)
4633 struct sctp_association *asoc;
4635 struct sctp_getaddrs getaddrs;
4636 struct sctp_transport *from;
4638 union sctp_addr temp;
4639 struct sctp_sock *sp = sctp_sk(sk);
4644 if (len < sizeof(struct sctp_getaddrs))
4647 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4650 /* For UDP-style sockets, id specifies the association to query. */
4651 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4655 to = optval + offsetof(struct sctp_getaddrs,addrs);
4656 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4658 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4660 memcpy(&temp, &from->ipaddr, sizeof(temp));
4661 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4662 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4663 if (space_left < addrlen)
4665 if (copy_to_user(to, &temp, addrlen))
4669 space_left -= addrlen;
4672 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4674 bytes_copied = ((char __user *)to) - optval;
4675 if (put_user(bytes_copied, optlen))
4681 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4682 size_t space_left, int *bytes_copied)
4684 struct sctp_sockaddr_entry *addr;
4685 union sctp_addr temp;
4688 struct net *net = sock_net(sk);
4691 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4695 if ((PF_INET == sk->sk_family) &&
4696 (AF_INET6 == addr->a.sa.sa_family))
4698 if ((PF_INET6 == sk->sk_family) &&
4699 inet_v6_ipv6only(sk) &&
4700 (AF_INET == addr->a.sa.sa_family))
4702 memcpy(&temp, &addr->a, sizeof(temp));
4703 if (!temp.v4.sin_port)
4704 temp.v4.sin_port = htons(port);
4706 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4708 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4709 if (space_left < addrlen) {
4713 memcpy(to, &temp, addrlen);
4717 space_left -= addrlen;
4718 *bytes_copied += addrlen;
4726 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4727 char __user *optval, int __user *optlen)
4729 struct sctp_bind_addr *bp;
4730 struct sctp_association *asoc;
4732 struct sctp_getaddrs getaddrs;
4733 struct sctp_sockaddr_entry *addr;
4735 union sctp_addr temp;
4736 struct sctp_sock *sp = sctp_sk(sk);
4740 int bytes_copied = 0;
4744 if (len < sizeof(struct sctp_getaddrs))
4747 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4751 * For UDP-style sockets, id specifies the association to query.
4752 * If the id field is set to the value '0' then the locally bound
4753 * addresses are returned without regard to any particular
4756 if (0 == getaddrs.assoc_id) {
4757 bp = &sctp_sk(sk)->ep->base.bind_addr;
4759 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4762 bp = &asoc->base.bind_addr;
4765 to = optval + offsetof(struct sctp_getaddrs,addrs);
4766 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4768 addrs = kmalloc(space_left, GFP_KERNEL);
4772 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4773 * addresses from the global local address list.
4775 if (sctp_list_single_entry(&bp->address_list)) {
4776 addr = list_entry(bp->address_list.next,
4777 struct sctp_sockaddr_entry, list);
4778 if (sctp_is_any(sk, &addr->a)) {
4779 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4780 space_left, &bytes_copied);
4790 /* Protection on the bound address list is not needed since
4791 * in the socket option context we hold a socket lock and
4792 * thus the bound address list can't change.
4794 list_for_each_entry(addr, &bp->address_list, list) {
4795 memcpy(&temp, &addr->a, sizeof(temp));
4796 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4797 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4798 if (space_left < addrlen) {
4799 err = -ENOMEM; /*fixme: right error?*/
4802 memcpy(buf, &temp, addrlen);
4804 bytes_copied += addrlen;
4806 space_left -= addrlen;
4810 if (copy_to_user(to, addrs, bytes_copied)) {
4814 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4818 if (put_user(bytes_copied, optlen))
4825 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4827 * Requests that the local SCTP stack use the enclosed peer address as
4828 * the association primary. The enclosed address must be one of the
4829 * association peer's addresses.
4831 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4832 char __user *optval, int __user *optlen)
4834 struct sctp_prim prim;
4835 struct sctp_association *asoc;
4836 struct sctp_sock *sp = sctp_sk(sk);
4838 if (len < sizeof(struct sctp_prim))
4841 len = sizeof(struct sctp_prim);
4843 if (copy_from_user(&prim, optval, len))
4846 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4850 if (!asoc->peer.primary_path)
4853 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4854 asoc->peer.primary_path->af_specific->sockaddr_len);
4856 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4857 (union sctp_addr *)&prim.ssp_addr);
4859 if (put_user(len, optlen))
4861 if (copy_to_user(optval, &prim, len))
4868 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4870 * Requests that the local endpoint set the specified Adaptation Layer
4871 * Indication parameter for all future INIT and INIT-ACK exchanges.
4873 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4874 char __user *optval, int __user *optlen)
4876 struct sctp_setadaptation adaptation;
4878 if (len < sizeof(struct sctp_setadaptation))
4881 len = sizeof(struct sctp_setadaptation);
4883 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4885 if (put_user(len, optlen))
4887 if (copy_to_user(optval, &adaptation, len))
4895 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4897 * Applications that wish to use the sendto() system call may wish to
4898 * specify a default set of parameters that would normally be supplied
4899 * through the inclusion of ancillary data. This socket option allows
4900 * such an application to set the default sctp_sndrcvinfo structure.
4903 * The application that wishes to use this socket option simply passes
4904 * in to this call the sctp_sndrcvinfo structure defined in Section
4905 * 5.2.2) The input parameters accepted by this call include
4906 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4907 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4908 * to this call if the caller is using the UDP model.
4910 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4912 static int sctp_getsockopt_default_send_param(struct sock *sk,
4913 int len, char __user *optval,
4916 struct sctp_sndrcvinfo info;
4917 struct sctp_association *asoc;
4918 struct sctp_sock *sp = sctp_sk(sk);
4920 if (len < sizeof(struct sctp_sndrcvinfo))
4923 len = sizeof(struct sctp_sndrcvinfo);
4925 if (copy_from_user(&info, optval, len))
4928 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4929 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4933 info.sinfo_stream = asoc->default_stream;
4934 info.sinfo_flags = asoc->default_flags;
4935 info.sinfo_ppid = asoc->default_ppid;
4936 info.sinfo_context = asoc->default_context;
4937 info.sinfo_timetolive = asoc->default_timetolive;
4939 info.sinfo_stream = sp->default_stream;
4940 info.sinfo_flags = sp->default_flags;
4941 info.sinfo_ppid = sp->default_ppid;
4942 info.sinfo_context = sp->default_context;
4943 info.sinfo_timetolive = sp->default_timetolive;
4946 if (put_user(len, optlen))
4948 if (copy_to_user(optval, &info, len))
4956 * 7.1.5 SCTP_NODELAY
4958 * Turn on/off any Nagle-like algorithm. This means that packets are
4959 * generally sent as soon as possible and no unnecessary delays are
4960 * introduced, at the cost of more packets in the network. Expects an
4961 * integer boolean flag.
4964 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4965 char __user *optval, int __user *optlen)
4969 if (len < sizeof(int))
4973 val = (sctp_sk(sk)->nodelay == 1);
4974 if (put_user(len, optlen))
4976 if (copy_to_user(optval, &val, len))
4983 * 7.1.1 SCTP_RTOINFO
4985 * The protocol parameters used to initialize and bound retransmission
4986 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4987 * and modify these parameters.
4988 * All parameters are time values, in milliseconds. A value of 0, when
4989 * modifying the parameters, indicates that the current value should not
4993 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4994 char __user *optval,
4995 int __user *optlen) {
4996 struct sctp_rtoinfo rtoinfo;
4997 struct sctp_association *asoc;
4999 if (len < sizeof (struct sctp_rtoinfo))
5002 len = sizeof(struct sctp_rtoinfo);
5004 if (copy_from_user(&rtoinfo, optval, len))
5007 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5009 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5012 /* Values corresponding to the specific association. */
5014 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5015 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5016 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5018 /* Values corresponding to the endpoint. */
5019 struct sctp_sock *sp = sctp_sk(sk);
5021 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5022 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5023 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5026 if (put_user(len, optlen))
5029 if (copy_to_user(optval, &rtoinfo, len))
5037 * 7.1.2 SCTP_ASSOCINFO
5039 * This option is used to tune the maximum retransmission attempts
5040 * of the association.
5041 * Returns an error if the new association retransmission value is
5042 * greater than the sum of the retransmission value of the peer.
5043 * See [SCTP] for more information.
5046 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5047 char __user *optval,
5051 struct sctp_assocparams assocparams;
5052 struct sctp_association *asoc;
5053 struct list_head *pos;
5056 if (len < sizeof (struct sctp_assocparams))
5059 len = sizeof(struct sctp_assocparams);
5061 if (copy_from_user(&assocparams, optval, len))
5064 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5066 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5069 /* Values correspoinding to the specific association */
5071 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5072 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5073 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5074 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5076 list_for_each(pos, &asoc->peer.transport_addr_list) {
5080 assocparams.sasoc_number_peer_destinations = cnt;
5082 /* Values corresponding to the endpoint */
5083 struct sctp_sock *sp = sctp_sk(sk);
5085 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5086 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5087 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5088 assocparams.sasoc_cookie_life =
5089 sp->assocparams.sasoc_cookie_life;
5090 assocparams.sasoc_number_peer_destinations =
5092 sasoc_number_peer_destinations;
5095 if (put_user(len, optlen))
5098 if (copy_to_user(optval, &assocparams, len))
5105 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5107 * This socket option is a boolean flag which turns on or off mapped V4
5108 * addresses. If this option is turned on and the socket is type
5109 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5110 * If this option is turned off, then no mapping will be done of V4
5111 * addresses and a user will receive both PF_INET6 and PF_INET type
5112 * addresses on the socket.
5114 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5115 char __user *optval, int __user *optlen)
5118 struct sctp_sock *sp = sctp_sk(sk);
5120 if (len < sizeof(int))
5125 if (put_user(len, optlen))
5127 if (copy_to_user(optval, &val, len))
5134 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5135 * (chapter and verse is quoted at sctp_setsockopt_context())
5137 static int sctp_getsockopt_context(struct sock *sk, int len,
5138 char __user *optval, int __user *optlen)
5140 struct sctp_assoc_value params;
5141 struct sctp_sock *sp;
5142 struct sctp_association *asoc;
5144 if (len < sizeof(struct sctp_assoc_value))
5147 len = sizeof(struct sctp_assoc_value);
5149 if (copy_from_user(¶ms, optval, len))
5154 if (params.assoc_id != 0) {
5155 asoc = sctp_id2assoc(sk, params.assoc_id);
5158 params.assoc_value = asoc->default_rcv_context;
5160 params.assoc_value = sp->default_rcv_context;
5163 if (put_user(len, optlen))
5165 if (copy_to_user(optval, ¶ms, len))
5172 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5173 * This option will get or set the maximum size to put in any outgoing
5174 * SCTP DATA chunk. If a message is larger than this size it will be
5175 * fragmented by SCTP into the specified size. Note that the underlying
5176 * SCTP implementation may fragment into smaller sized chunks when the
5177 * PMTU of the underlying association is smaller than the value set by
5178 * the user. The default value for this option is '0' which indicates
5179 * the user is NOT limiting fragmentation and only the PMTU will effect
5180 * SCTP's choice of DATA chunk size. Note also that values set larger
5181 * than the maximum size of an IP datagram will effectively let SCTP
5182 * control fragmentation (i.e. the same as setting this option to 0).
5184 * The following structure is used to access and modify this parameter:
5186 * struct sctp_assoc_value {
5187 * sctp_assoc_t assoc_id;
5188 * uint32_t assoc_value;
5191 * assoc_id: This parameter is ignored for one-to-one style sockets.
5192 * For one-to-many style sockets this parameter indicates which
5193 * association the user is performing an action upon. Note that if
5194 * this field's value is zero then the endpoints default value is
5195 * changed (effecting future associations only).
5196 * assoc_value: This parameter specifies the maximum size in bytes.
5198 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5199 char __user *optval, int __user *optlen)
5201 struct sctp_assoc_value params;
5202 struct sctp_association *asoc;
5204 if (len == sizeof(int)) {
5205 pr_warn("Use of int in maxseg socket option deprecated\n");
5206 pr_warn("Use struct sctp_assoc_value instead\n");
5207 params.assoc_id = 0;
5208 } else if (len >= sizeof(struct sctp_assoc_value)) {
5209 len = sizeof(struct sctp_assoc_value);
5210 if (copy_from_user(¶ms, optval, sizeof(params)))
5215 asoc = sctp_id2assoc(sk, params.assoc_id);
5216 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5220 params.assoc_value = asoc->frag_point;
5222 params.assoc_value = sctp_sk(sk)->user_frag;
5224 if (put_user(len, optlen))
5226 if (len == sizeof(int)) {
5227 if (copy_to_user(optval, ¶ms.assoc_value, len))
5230 if (copy_to_user(optval, ¶ms, len))
5238 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5239 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5241 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5242 char __user *optval, int __user *optlen)
5246 if (len < sizeof(int))
5251 val = sctp_sk(sk)->frag_interleave;
5252 if (put_user(len, optlen))
5254 if (copy_to_user(optval, &val, len))
5261 * 7.1.25. Set or Get the sctp partial delivery point
5262 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5264 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5265 char __user *optval,
5270 if (len < sizeof(u32))
5275 val = sctp_sk(sk)->pd_point;
5276 if (put_user(len, optlen))
5278 if (copy_to_user(optval, &val, len))
5285 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5286 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5288 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5289 char __user *optval,
5292 struct sctp_assoc_value params;
5293 struct sctp_sock *sp;
5294 struct sctp_association *asoc;
5296 if (len == sizeof(int)) {
5297 pr_warn("Use of int in max_burst socket option deprecated\n");
5298 pr_warn("Use struct sctp_assoc_value instead\n");
5299 params.assoc_id = 0;
5300 } else if (len >= sizeof(struct sctp_assoc_value)) {
5301 len = sizeof(struct sctp_assoc_value);
5302 if (copy_from_user(¶ms, optval, len))
5309 if (params.assoc_id != 0) {
5310 asoc = sctp_id2assoc(sk, params.assoc_id);
5313 params.assoc_value = asoc->max_burst;
5315 params.assoc_value = sp->max_burst;
5317 if (len == sizeof(int)) {
5318 if (copy_to_user(optval, ¶ms.assoc_value, len))
5321 if (copy_to_user(optval, ¶ms, len))
5329 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5330 char __user *optval, int __user *optlen)
5332 struct net *net = sock_net(sk);
5333 struct sctp_hmacalgo __user *p = (void __user *)optval;
5334 struct sctp_hmac_algo_param *hmacs;
5338 if (!net->sctp.auth_enable)
5341 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5342 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5344 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5347 len = sizeof(struct sctp_hmacalgo) + data_len;
5348 num_idents = data_len / sizeof(u16);
5350 if (put_user(len, optlen))
5352 if (put_user(num_idents, &p->shmac_num_idents))
5354 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5359 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5360 char __user *optval, int __user *optlen)
5362 struct net *net = sock_net(sk);
5363 struct sctp_authkeyid val;
5364 struct sctp_association *asoc;
5366 if (!net->sctp.auth_enable)
5369 if (len < sizeof(struct sctp_authkeyid))
5371 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5374 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5375 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5379 val.scact_keynumber = asoc->active_key_id;
5381 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5383 len = sizeof(struct sctp_authkeyid);
5384 if (put_user(len, optlen))
5386 if (copy_to_user(optval, &val, len))
5392 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5393 char __user *optval, int __user *optlen)
5395 struct net *net = sock_net(sk);
5396 struct sctp_authchunks __user *p = (void __user *)optval;
5397 struct sctp_authchunks val;
5398 struct sctp_association *asoc;
5399 struct sctp_chunks_param *ch;
5403 if (!net->sctp.auth_enable)
5406 if (len < sizeof(struct sctp_authchunks))
5409 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5412 to = p->gauth_chunks;
5413 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5417 ch = asoc->peer.peer_chunks;
5421 /* See if the user provided enough room for all the data */
5422 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5423 if (len < num_chunks)
5426 if (copy_to_user(to, ch->chunks, num_chunks))
5429 len = sizeof(struct sctp_authchunks) + num_chunks;
5430 if (put_user(len, optlen)) return -EFAULT;
5431 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5436 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5437 char __user *optval, int __user *optlen)
5439 struct net *net = sock_net(sk);
5440 struct sctp_authchunks __user *p = (void __user *)optval;
5441 struct sctp_authchunks val;
5442 struct sctp_association *asoc;
5443 struct sctp_chunks_param *ch;
5447 if (!net->sctp.auth_enable)
5450 if (len < sizeof(struct sctp_authchunks))
5453 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5456 to = p->gauth_chunks;
5457 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5458 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5462 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5464 ch = sctp_sk(sk)->ep->auth_chunk_list;
5469 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5470 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5473 if (copy_to_user(to, ch->chunks, num_chunks))
5476 len = sizeof(struct sctp_authchunks) + num_chunks;
5477 if (put_user(len, optlen))
5479 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5486 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5487 * This option gets the current number of associations that are attached
5488 * to a one-to-many style socket. The option value is an uint32_t.
5490 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5491 char __user *optval, int __user *optlen)
5493 struct sctp_sock *sp = sctp_sk(sk);
5494 struct sctp_association *asoc;
5497 if (sctp_style(sk, TCP))
5500 if (len < sizeof(u32))
5505 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5509 if (put_user(len, optlen))
5511 if (copy_to_user(optval, &val, len))
5518 * 8.1.23 SCTP_AUTO_ASCONF
5519 * See the corresponding setsockopt entry as description
5521 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5522 char __user *optval, int __user *optlen)
5526 if (len < sizeof(int))
5530 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5532 if (put_user(len, optlen))
5534 if (copy_to_user(optval, &val, len))
5540 * 8.2.6. Get the Current Identifiers of Associations
5541 * (SCTP_GET_ASSOC_ID_LIST)
5543 * This option gets the current list of SCTP association identifiers of
5544 * the SCTP associations handled by a one-to-many style socket.
5546 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5547 char __user *optval, int __user *optlen)
5549 struct sctp_sock *sp = sctp_sk(sk);
5550 struct sctp_association *asoc;
5551 struct sctp_assoc_ids *ids;
5554 if (sctp_style(sk, TCP))
5557 if (len < sizeof(struct sctp_assoc_ids))
5560 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5564 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5567 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5569 ids = kmalloc(len, GFP_KERNEL);
5573 ids->gaids_number_of_ids = num;
5575 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5576 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5579 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5589 * SCTP_PEER_ADDR_THLDS
5591 * This option allows us to fetch the partially failed threshold for one or all
5592 * transports in an association. See Section 6.1 of:
5593 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5595 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5596 char __user *optval,
5600 struct sctp_paddrthlds val;
5601 struct sctp_transport *trans;
5602 struct sctp_association *asoc;
5604 if (len < sizeof(struct sctp_paddrthlds))
5606 len = sizeof(struct sctp_paddrthlds);
5607 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5610 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5611 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5615 val.spt_pathpfthld = asoc->pf_retrans;
5616 val.spt_pathmaxrxt = asoc->pathmaxrxt;
5618 trans = sctp_addr_id2transport(sk, &val.spt_address,
5623 val.spt_pathmaxrxt = trans->pathmaxrxt;
5624 val.spt_pathpfthld = trans->pf_retrans;
5627 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5634 * SCTP_GET_ASSOC_STATS
5636 * This option retrieves local per endpoint statistics. It is modeled
5637 * after OpenSolaris' implementation
5639 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
5640 char __user *optval,
5643 struct sctp_assoc_stats sas;
5644 struct sctp_association *asoc = NULL;
5646 /* User must provide at least the assoc id */
5647 if (len < sizeof(sctp_assoc_t))
5650 /* Allow the struct to grow and fill in as much as possible */
5651 len = min_t(size_t, len, sizeof(sas));
5653 if (copy_from_user(&sas, optval, len))
5656 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
5660 sas.sas_rtxchunks = asoc->stats.rtxchunks;
5661 sas.sas_gapcnt = asoc->stats.gapcnt;
5662 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
5663 sas.sas_osacks = asoc->stats.osacks;
5664 sas.sas_isacks = asoc->stats.isacks;
5665 sas.sas_octrlchunks = asoc->stats.octrlchunks;
5666 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
5667 sas.sas_oodchunks = asoc->stats.oodchunks;
5668 sas.sas_iodchunks = asoc->stats.iodchunks;
5669 sas.sas_ouodchunks = asoc->stats.ouodchunks;
5670 sas.sas_iuodchunks = asoc->stats.iuodchunks;
5671 sas.sas_idupchunks = asoc->stats.idupchunks;
5672 sas.sas_opackets = asoc->stats.opackets;
5673 sas.sas_ipackets = asoc->stats.ipackets;
5675 /* New high max rto observed, will return 0 if not a single
5676 * RTO update took place. obs_rto_ipaddr will be bogus
5679 sas.sas_maxrto = asoc->stats.max_obs_rto;
5680 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
5681 sizeof(struct sockaddr_storage));
5683 /* Mark beginning of a new observation period */
5684 asoc->stats.max_obs_rto = asoc->rto_min;
5686 if (put_user(len, optlen))
5689 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
5691 if (copy_to_user(optval, &sas, len))
5697 static int sctp_getsockopt(struct sock *sk, int level, int optname,
5698 char __user *optval, int __user *optlen)
5703 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
5705 /* I can hardly begin to describe how wrong this is. This is
5706 * so broken as to be worse than useless. The API draft
5707 * REALLY is NOT helpful here... I am not convinced that the
5708 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5709 * are at all well-founded.
5711 if (level != SOL_SCTP) {
5712 struct sctp_af *af = sctp_sk(sk)->pf->af;
5714 retval = af->getsockopt(sk, level, optname, optval, optlen);
5718 if (get_user(len, optlen))
5725 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5727 case SCTP_DISABLE_FRAGMENTS:
5728 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5732 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5734 case SCTP_AUTOCLOSE:
5735 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5737 case SCTP_SOCKOPT_PEELOFF:
5738 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5740 case SCTP_PEER_ADDR_PARAMS:
5741 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5744 case SCTP_DELAYED_SACK:
5745 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5749 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5751 case SCTP_GET_PEER_ADDRS:
5752 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5755 case SCTP_GET_LOCAL_ADDRS:
5756 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5759 case SCTP_SOCKOPT_CONNECTX3:
5760 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5762 case SCTP_DEFAULT_SEND_PARAM:
5763 retval = sctp_getsockopt_default_send_param(sk, len,
5766 case SCTP_PRIMARY_ADDR:
5767 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5770 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5773 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5775 case SCTP_ASSOCINFO:
5776 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5778 case SCTP_I_WANT_MAPPED_V4_ADDR:
5779 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5782 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5784 case SCTP_GET_PEER_ADDR_INFO:
5785 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5788 case SCTP_ADAPTATION_LAYER:
5789 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5793 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5795 case SCTP_FRAGMENT_INTERLEAVE:
5796 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5799 case SCTP_PARTIAL_DELIVERY_POINT:
5800 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5803 case SCTP_MAX_BURST:
5804 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5807 case SCTP_AUTH_CHUNK:
5808 case SCTP_AUTH_DELETE_KEY:
5809 retval = -EOPNOTSUPP;
5811 case SCTP_HMAC_IDENT:
5812 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5814 case SCTP_AUTH_ACTIVE_KEY:
5815 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5817 case SCTP_PEER_AUTH_CHUNKS:
5818 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5821 case SCTP_LOCAL_AUTH_CHUNKS:
5822 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5825 case SCTP_GET_ASSOC_NUMBER:
5826 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5828 case SCTP_GET_ASSOC_ID_LIST:
5829 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5831 case SCTP_AUTO_ASCONF:
5832 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5834 case SCTP_PEER_ADDR_THLDS:
5835 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
5837 case SCTP_GET_ASSOC_STATS:
5838 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
5841 retval = -ENOPROTOOPT;
5845 sctp_release_sock(sk);
5849 static void sctp_hash(struct sock *sk)
5854 static void sctp_unhash(struct sock *sk)
5859 /* Check if port is acceptable. Possibly find first available port.
5861 * The port hash table (contained in the 'global' SCTP protocol storage
5862 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5863 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5864 * list (the list number is the port number hashed out, so as you
5865 * would expect from a hash function, all the ports in a given list have
5866 * such a number that hashes out to the same list number; you were
5867 * expecting that, right?); so each list has a set of ports, with a
5868 * link to the socket (struct sock) that uses it, the port number and
5869 * a fastreuse flag (FIXME: NPI ipg).
5871 static struct sctp_bind_bucket *sctp_bucket_create(
5872 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
5874 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5876 struct sctp_bind_hashbucket *head; /* hash list */
5877 struct sctp_bind_bucket *pp;
5878 unsigned short snum;
5881 snum = ntohs(addr->v4.sin_port);
5883 pr_debug("%s: begins, snum:%d\n", __func__, snum);
5885 sctp_local_bh_disable();
5888 /* Search for an available port. */
5889 int low, high, remaining, index;
5892 inet_get_local_port_range(sock_net(sk), &low, &high);
5893 remaining = (high - low) + 1;
5894 rover = net_random() % remaining + low;
5898 if ((rover < low) || (rover > high))
5900 if (inet_is_reserved_local_port(rover))
5902 index = sctp_phashfn(sock_net(sk), rover);
5903 head = &sctp_port_hashtable[index];
5904 sctp_spin_lock(&head->lock);
5905 sctp_for_each_hentry(pp, &head->chain)
5906 if ((pp->port == rover) &&
5907 net_eq(sock_net(sk), pp->net))
5911 sctp_spin_unlock(&head->lock);
5912 } while (--remaining > 0);
5914 /* Exhausted local port range during search? */
5919 /* OK, here is the one we will use. HEAD (the port
5920 * hash table list entry) is non-NULL and we hold it's
5925 /* We are given an specific port number; we verify
5926 * that it is not being used. If it is used, we will
5927 * exahust the search in the hash list corresponding
5928 * to the port number (snum) - we detect that with the
5929 * port iterator, pp being NULL.
5931 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
5932 sctp_spin_lock(&head->lock);
5933 sctp_for_each_hentry(pp, &head->chain) {
5934 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
5941 if (!hlist_empty(&pp->owner)) {
5942 /* We had a port hash table hit - there is an
5943 * available port (pp != NULL) and it is being
5944 * used by other socket (pp->owner not empty); that other
5945 * socket is going to be sk2.
5947 int reuse = sk->sk_reuse;
5950 pr_debug("%s: found a possible match\n", __func__);
5952 if (pp->fastreuse && sk->sk_reuse &&
5953 sk->sk_state != SCTP_SS_LISTENING)
5956 /* Run through the list of sockets bound to the port
5957 * (pp->port) [via the pointers bind_next and
5958 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5959 * we get the endpoint they describe and run through
5960 * the endpoint's list of IP (v4 or v6) addresses,
5961 * comparing each of the addresses with the address of
5962 * the socket sk. If we find a match, then that means
5963 * that this port/socket (sk) combination are already
5966 sk_for_each_bound(sk2, &pp->owner) {
5967 struct sctp_endpoint *ep2;
5968 ep2 = sctp_sk(sk2)->ep;
5971 (reuse && sk2->sk_reuse &&
5972 sk2->sk_state != SCTP_SS_LISTENING))
5975 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5976 sctp_sk(sk2), sctp_sk(sk))) {
5982 pr_debug("%s: found a match\n", __func__);
5985 /* If there was a hash table miss, create a new port. */
5987 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
5990 /* In either case (hit or miss), make sure fastreuse is 1 only
5991 * if sk->sk_reuse is too (that is, if the caller requested
5992 * SO_REUSEADDR on this socket -sk-).
5994 if (hlist_empty(&pp->owner)) {
5995 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5999 } else if (pp->fastreuse &&
6000 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6003 /* We are set, so fill up all the data in the hash table
6004 * entry, tie the socket list information with the rest of the
6005 * sockets FIXME: Blurry, NPI (ipg).
6008 if (!sctp_sk(sk)->bind_hash) {
6009 inet_sk(sk)->inet_num = snum;
6010 sk_add_bind_node(sk, &pp->owner);
6011 sctp_sk(sk)->bind_hash = pp;
6016 sctp_spin_unlock(&head->lock);
6019 sctp_local_bh_enable();
6023 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6024 * port is requested.
6026 static int sctp_get_port(struct sock *sk, unsigned short snum)
6028 union sctp_addr addr;
6029 struct sctp_af *af = sctp_sk(sk)->pf->af;
6031 /* Set up a dummy address struct from the sk. */
6032 af->from_sk(&addr, sk);
6033 addr.v4.sin_port = htons(snum);
6035 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6036 return !!sctp_get_port_local(sk, &addr);
6040 * Move a socket to LISTENING state.
6042 static int sctp_listen_start(struct sock *sk, int backlog)
6044 struct sctp_sock *sp = sctp_sk(sk);
6045 struct sctp_endpoint *ep = sp->ep;
6046 struct crypto_hash *tfm = NULL;
6049 /* Allocate HMAC for generating cookie. */
6050 if (!sp->hmac && sp->sctp_hmac_alg) {
6051 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6052 tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
6054 net_info_ratelimited("failed to load transform for %s: %ld\n",
6055 sp->sctp_hmac_alg, PTR_ERR(tfm));
6058 sctp_sk(sk)->hmac = tfm;
6062 * If a bind() or sctp_bindx() is not called prior to a listen()
6063 * call that allows new associations to be accepted, the system
6064 * picks an ephemeral port and will choose an address set equivalent
6065 * to binding with a wildcard address.
6067 * This is not currently spelled out in the SCTP sockets
6068 * extensions draft, but follows the practice as seen in TCP
6072 sk->sk_state = SCTP_SS_LISTENING;
6073 if (!ep->base.bind_addr.port) {
6074 if (sctp_autobind(sk))
6077 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6078 sk->sk_state = SCTP_SS_CLOSED;
6083 sk->sk_max_ack_backlog = backlog;
6084 sctp_hash_endpoint(ep);
6089 * 4.1.3 / 5.1.3 listen()
6091 * By default, new associations are not accepted for UDP style sockets.
6092 * An application uses listen() to mark a socket as being able to
6093 * accept new associations.
6095 * On TCP style sockets, applications use listen() to ready the SCTP
6096 * endpoint for accepting inbound associations.
6098 * On both types of endpoints a backlog of '0' disables listening.
6100 * Move a socket to LISTENING state.
6102 int sctp_inet_listen(struct socket *sock, int backlog)
6104 struct sock *sk = sock->sk;
6105 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6108 if (unlikely(backlog < 0))
6113 /* Peeled-off sockets are not allowed to listen(). */
6114 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6117 if (sock->state != SS_UNCONNECTED)
6120 /* If backlog is zero, disable listening. */
6122 if (sctp_sstate(sk, CLOSED))
6126 sctp_unhash_endpoint(ep);
6127 sk->sk_state = SCTP_SS_CLOSED;
6129 sctp_sk(sk)->bind_hash->fastreuse = 1;
6133 /* If we are already listening, just update the backlog */
6134 if (sctp_sstate(sk, LISTENING))
6135 sk->sk_max_ack_backlog = backlog;
6137 err = sctp_listen_start(sk, backlog);
6144 sctp_release_sock(sk);
6149 * This function is done by modeling the current datagram_poll() and the
6150 * tcp_poll(). Note that, based on these implementations, we don't
6151 * lock the socket in this function, even though it seems that,
6152 * ideally, locking or some other mechanisms can be used to ensure
6153 * the integrity of the counters (sndbuf and wmem_alloc) used
6154 * in this place. We assume that we don't need locks either until proven
6157 * Another thing to note is that we include the Async I/O support
6158 * here, again, by modeling the current TCP/UDP code. We don't have
6159 * a good way to test with it yet.
6161 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6163 struct sock *sk = sock->sk;
6164 struct sctp_sock *sp = sctp_sk(sk);
6167 poll_wait(file, sk_sleep(sk), wait);
6169 /* A TCP-style listening socket becomes readable when the accept queue
6172 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6173 return (!list_empty(&sp->ep->asocs)) ?
6174 (POLLIN | POLLRDNORM) : 0;
6178 /* Is there any exceptional events? */
6179 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6181 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6182 if (sk->sk_shutdown & RCV_SHUTDOWN)
6183 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6184 if (sk->sk_shutdown == SHUTDOWN_MASK)
6187 /* Is it readable? Reconsider this code with TCP-style support. */
6188 if (!skb_queue_empty(&sk->sk_receive_queue))
6189 mask |= POLLIN | POLLRDNORM;
6191 /* The association is either gone or not ready. */
6192 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6195 /* Is it writable? */
6196 if (sctp_writeable(sk)) {
6197 mask |= POLLOUT | POLLWRNORM;
6199 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6201 * Since the socket is not locked, the buffer
6202 * might be made available after the writeable check and
6203 * before the bit is set. This could cause a lost I/O
6204 * signal. tcp_poll() has a race breaker for this race
6205 * condition. Based on their implementation, we put
6206 * in the following code to cover it as well.
6208 if (sctp_writeable(sk))
6209 mask |= POLLOUT | POLLWRNORM;
6214 /********************************************************************
6215 * 2nd Level Abstractions
6216 ********************************************************************/
6218 static struct sctp_bind_bucket *sctp_bucket_create(
6219 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6221 struct sctp_bind_bucket *pp;
6223 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6225 SCTP_DBG_OBJCNT_INC(bind_bucket);
6228 INIT_HLIST_HEAD(&pp->owner);
6230 hlist_add_head(&pp->node, &head->chain);
6235 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6236 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6238 if (pp && hlist_empty(&pp->owner)) {
6239 __hlist_del(&pp->node);
6240 kmem_cache_free(sctp_bucket_cachep, pp);
6241 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6245 /* Release this socket's reference to a local port. */
6246 static inline void __sctp_put_port(struct sock *sk)
6248 struct sctp_bind_hashbucket *head =
6249 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6250 inet_sk(sk)->inet_num)];
6251 struct sctp_bind_bucket *pp;
6253 sctp_spin_lock(&head->lock);
6254 pp = sctp_sk(sk)->bind_hash;
6255 __sk_del_bind_node(sk);
6256 sctp_sk(sk)->bind_hash = NULL;
6257 inet_sk(sk)->inet_num = 0;
6258 sctp_bucket_destroy(pp);
6259 sctp_spin_unlock(&head->lock);
6262 void sctp_put_port(struct sock *sk)
6264 sctp_local_bh_disable();
6265 __sctp_put_port(sk);
6266 sctp_local_bh_enable();
6270 * The system picks an ephemeral port and choose an address set equivalent
6271 * to binding with a wildcard address.
6272 * One of those addresses will be the primary address for the association.
6273 * This automatically enables the multihoming capability of SCTP.
6275 static int sctp_autobind(struct sock *sk)
6277 union sctp_addr autoaddr;
6281 /* Initialize a local sockaddr structure to INADDR_ANY. */
6282 af = sctp_sk(sk)->pf->af;
6284 port = htons(inet_sk(sk)->inet_num);
6285 af->inaddr_any(&autoaddr, port);
6287 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6290 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6293 * 4.2 The cmsghdr Structure *
6295 * When ancillary data is sent or received, any number of ancillary data
6296 * objects can be specified by the msg_control and msg_controllen members of
6297 * the msghdr structure, because each object is preceded by
6298 * a cmsghdr structure defining the object's length (the cmsg_len member).
6299 * Historically Berkeley-derived implementations have passed only one object
6300 * at a time, but this API allows multiple objects to be
6301 * passed in a single call to sendmsg() or recvmsg(). The following example
6302 * shows two ancillary data objects in a control buffer.
6304 * |<--------------------------- msg_controllen -------------------------->|
6307 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6309 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6312 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6314 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6317 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6318 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6320 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6322 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6329 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
6331 struct cmsghdr *cmsg;
6332 struct msghdr *my_msg = (struct msghdr *)msg;
6334 for (cmsg = CMSG_FIRSTHDR(msg);
6336 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6337 if (!CMSG_OK(my_msg, cmsg))
6340 /* Should we parse this header or ignore? */
6341 if (cmsg->cmsg_level != IPPROTO_SCTP)
6344 /* Strictly check lengths following example in SCM code. */
6345 switch (cmsg->cmsg_type) {
6347 /* SCTP Socket API Extension
6348 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6350 * This cmsghdr structure provides information for
6351 * initializing new SCTP associations with sendmsg().
6352 * The SCTP_INITMSG socket option uses this same data
6353 * structure. This structure is not used for
6356 * cmsg_level cmsg_type cmsg_data[]
6357 * ------------ ------------ ----------------------
6358 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6360 if (cmsg->cmsg_len !=
6361 CMSG_LEN(sizeof(struct sctp_initmsg)))
6363 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6367 /* SCTP Socket API Extension
6368 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6370 * This cmsghdr structure specifies SCTP options for
6371 * sendmsg() and describes SCTP header information
6372 * about a received message through recvmsg().
6374 * cmsg_level cmsg_type cmsg_data[]
6375 * ------------ ------------ ----------------------
6376 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6378 if (cmsg->cmsg_len !=
6379 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6383 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6385 /* Minimally, validate the sinfo_flags. */
6386 if (cmsgs->info->sinfo_flags &
6387 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6388 SCTP_ABORT | SCTP_EOF))
6400 * Wait for a packet..
6401 * Note: This function is the same function as in core/datagram.c
6402 * with a few modifications to make lksctp work.
6404 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6409 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6411 /* Socket errors? */
6412 error = sock_error(sk);
6416 if (!skb_queue_empty(&sk->sk_receive_queue))
6419 /* Socket shut down? */
6420 if (sk->sk_shutdown & RCV_SHUTDOWN)
6423 /* Sequenced packets can come disconnected. If so we report the
6428 /* Is there a good reason to think that we may receive some data? */
6429 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6432 /* Handle signals. */
6433 if (signal_pending(current))
6436 /* Let another process have a go. Since we are going to sleep
6437 * anyway. Note: This may cause odd behaviors if the message
6438 * does not fit in the user's buffer, but this seems to be the
6439 * only way to honor MSG_DONTWAIT realistically.
6441 sctp_release_sock(sk);
6442 *timeo_p = schedule_timeout(*timeo_p);
6446 finish_wait(sk_sleep(sk), &wait);
6450 error = sock_intr_errno(*timeo_p);
6453 finish_wait(sk_sleep(sk), &wait);
6458 /* Receive a datagram.
6459 * Note: This is pretty much the same routine as in core/datagram.c
6460 * with a few changes to make lksctp work.
6462 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6463 int noblock, int *err)
6466 struct sk_buff *skb;
6469 timeo = sock_rcvtimeo(sk, noblock);
6471 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
6472 MAX_SCHEDULE_TIMEOUT);
6475 /* Again only user level code calls this function,
6476 * so nothing interrupt level
6477 * will suddenly eat the receive_queue.
6479 * Look at current nfs client by the way...
6480 * However, this function was correct in any case. 8)
6482 if (flags & MSG_PEEK) {
6483 spin_lock_bh(&sk->sk_receive_queue.lock);
6484 skb = skb_peek(&sk->sk_receive_queue);
6486 atomic_inc(&skb->users);
6487 spin_unlock_bh(&sk->sk_receive_queue.lock);
6489 skb = skb_dequeue(&sk->sk_receive_queue);
6495 /* Caller is allowed not to check sk->sk_err before calling. */
6496 error = sock_error(sk);
6500 if (sk->sk_shutdown & RCV_SHUTDOWN)
6503 /* User doesn't want to wait. */
6507 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6516 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6517 static void __sctp_write_space(struct sctp_association *asoc)
6519 struct sock *sk = asoc->base.sk;
6520 struct socket *sock = sk->sk_socket;
6522 if ((sctp_wspace(asoc) > 0) && sock) {
6523 if (waitqueue_active(&asoc->wait))
6524 wake_up_interruptible(&asoc->wait);
6526 if (sctp_writeable(sk)) {
6527 wait_queue_head_t *wq = sk_sleep(sk);
6529 if (wq && waitqueue_active(wq))
6530 wake_up_interruptible(wq);
6532 /* Note that we try to include the Async I/O support
6533 * here by modeling from the current TCP/UDP code.
6534 * We have not tested with it yet.
6536 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6537 sock_wake_async(sock,
6538 SOCK_WAKE_SPACE, POLL_OUT);
6543 /* Do accounting for the sndbuf space.
6544 * Decrement the used sndbuf space of the corresponding association by the
6545 * data size which was just transmitted(freed).
6547 static void sctp_wfree(struct sk_buff *skb)
6549 struct sctp_association *asoc;
6550 struct sctp_chunk *chunk;
6553 /* Get the saved chunk pointer. */
6554 chunk = *((struct sctp_chunk **)(skb->cb));
6557 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6558 sizeof(struct sk_buff) +
6559 sizeof(struct sctp_chunk);
6561 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6564 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6566 sk->sk_wmem_queued -= skb->truesize;
6567 sk_mem_uncharge(sk, skb->truesize);
6570 __sctp_write_space(asoc);
6572 sctp_association_put(asoc);
6575 /* Do accounting for the receive space on the socket.
6576 * Accounting for the association is done in ulpevent.c
6577 * We set this as a destructor for the cloned data skbs so that
6578 * accounting is done at the correct time.
6580 void sctp_sock_rfree(struct sk_buff *skb)
6582 struct sock *sk = skb->sk;
6583 struct sctp_ulpevent *event = sctp_skb2event(skb);
6585 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6588 * Mimic the behavior of sock_rfree
6590 sk_mem_uncharge(sk, event->rmem_len);
6594 /* Helper function to wait for space in the sndbuf. */
6595 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6598 struct sock *sk = asoc->base.sk;
6600 long current_timeo = *timeo_p;
6603 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
6606 /* Increment the association's refcnt. */
6607 sctp_association_hold(asoc);
6609 /* Wait on the association specific sndbuf space. */
6611 prepare_to_wait_exclusive(&asoc->wait, &wait,
6612 TASK_INTERRUPTIBLE);
6615 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6618 if (signal_pending(current))
6619 goto do_interrupted;
6620 if (msg_len <= sctp_wspace(asoc))
6623 /* Let another process have a go. Since we are going
6626 sctp_release_sock(sk);
6627 current_timeo = schedule_timeout(current_timeo);
6628 BUG_ON(sk != asoc->base.sk);
6631 *timeo_p = current_timeo;
6635 finish_wait(&asoc->wait, &wait);
6637 /* Release the association's refcnt. */
6638 sctp_association_put(asoc);
6647 err = sock_intr_errno(*timeo_p);
6655 void sctp_data_ready(struct sock *sk, int len)
6657 struct socket_wq *wq;
6660 wq = rcu_dereference(sk->sk_wq);
6661 if (wq_has_sleeper(wq))
6662 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6663 POLLRDNORM | POLLRDBAND);
6664 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6668 /* If socket sndbuf has changed, wake up all per association waiters. */
6669 void sctp_write_space(struct sock *sk)
6671 struct sctp_association *asoc;
6673 /* Wake up the tasks in each wait queue. */
6674 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6675 __sctp_write_space(asoc);
6679 /* Is there any sndbuf space available on the socket?
6681 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6682 * associations on the same socket. For a UDP-style socket with
6683 * multiple associations, it is possible for it to be "unwriteable"
6684 * prematurely. I assume that this is acceptable because
6685 * a premature "unwriteable" is better than an accidental "writeable" which
6686 * would cause an unwanted block under certain circumstances. For the 1-1
6687 * UDP-style sockets or TCP-style sockets, this code should work.
6690 static int sctp_writeable(struct sock *sk)
6694 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6700 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6701 * returns immediately with EINPROGRESS.
6703 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6705 struct sock *sk = asoc->base.sk;
6707 long current_timeo = *timeo_p;
6710 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
6712 /* Increment the association's refcnt. */
6713 sctp_association_hold(asoc);
6716 prepare_to_wait_exclusive(&asoc->wait, &wait,
6717 TASK_INTERRUPTIBLE);
6720 if (sk->sk_shutdown & RCV_SHUTDOWN)
6722 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6725 if (signal_pending(current))
6726 goto do_interrupted;
6728 if (sctp_state(asoc, ESTABLISHED))
6731 /* Let another process have a go. Since we are going
6734 sctp_release_sock(sk);
6735 current_timeo = schedule_timeout(current_timeo);
6738 *timeo_p = current_timeo;
6742 finish_wait(&asoc->wait, &wait);
6744 /* Release the association's refcnt. */
6745 sctp_association_put(asoc);
6750 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6753 err = -ECONNREFUSED;
6757 err = sock_intr_errno(*timeo_p);
6765 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6767 struct sctp_endpoint *ep;
6771 ep = sctp_sk(sk)->ep;
6775 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6776 TASK_INTERRUPTIBLE);
6778 if (list_empty(&ep->asocs)) {
6779 sctp_release_sock(sk);
6780 timeo = schedule_timeout(timeo);
6785 if (!sctp_sstate(sk, LISTENING))
6789 if (!list_empty(&ep->asocs))
6792 err = sock_intr_errno(timeo);
6793 if (signal_pending(current))
6801 finish_wait(sk_sleep(sk), &wait);
6806 static void sctp_wait_for_close(struct sock *sk, long timeout)
6811 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6812 if (list_empty(&sctp_sk(sk)->ep->asocs))
6814 sctp_release_sock(sk);
6815 timeout = schedule_timeout(timeout);
6817 } while (!signal_pending(current) && timeout);
6819 finish_wait(sk_sleep(sk), &wait);
6822 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6824 struct sk_buff *frag;
6829 /* Don't forget the fragments. */
6830 skb_walk_frags(skb, frag)
6831 sctp_skb_set_owner_r_frag(frag, sk);
6834 sctp_skb_set_owner_r(skb, sk);
6837 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6838 struct sctp_association *asoc)
6840 struct inet_sock *inet = inet_sk(sk);
6841 struct inet_sock *newinet;
6843 newsk->sk_type = sk->sk_type;
6844 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6845 newsk->sk_flags = sk->sk_flags;
6846 newsk->sk_no_check = sk->sk_no_check;
6847 newsk->sk_reuse = sk->sk_reuse;
6849 newsk->sk_shutdown = sk->sk_shutdown;
6850 newsk->sk_destruct = sctp_destruct_sock;
6851 newsk->sk_family = sk->sk_family;
6852 newsk->sk_protocol = IPPROTO_SCTP;
6853 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6854 newsk->sk_sndbuf = sk->sk_sndbuf;
6855 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6856 newsk->sk_lingertime = sk->sk_lingertime;
6857 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6858 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6860 newinet = inet_sk(newsk);
6862 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6863 * getsockname() and getpeername()
6865 newinet->inet_sport = inet->inet_sport;
6866 newinet->inet_saddr = inet->inet_saddr;
6867 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6868 newinet->inet_dport = htons(asoc->peer.port);
6869 newinet->pmtudisc = inet->pmtudisc;
6870 newinet->inet_id = asoc->next_tsn ^ jiffies;
6872 newinet->uc_ttl = inet->uc_ttl;
6873 newinet->mc_loop = 1;
6874 newinet->mc_ttl = 1;
6875 newinet->mc_index = 0;
6876 newinet->mc_list = NULL;
6879 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6880 * and its messages to the newsk.
6882 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6883 struct sctp_association *assoc,
6884 sctp_socket_type_t type)
6886 struct sctp_sock *oldsp = sctp_sk(oldsk);
6887 struct sctp_sock *newsp = sctp_sk(newsk);
6888 struct sctp_bind_bucket *pp; /* hash list port iterator */
6889 struct sctp_endpoint *newep = newsp->ep;
6890 struct sk_buff *skb, *tmp;
6891 struct sctp_ulpevent *event;
6892 struct sctp_bind_hashbucket *head;
6893 struct list_head tmplist;
6895 /* Migrate socket buffer sizes and all the socket level options to the
6898 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6899 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6900 /* Brute force copy old sctp opt. */
6901 if (oldsp->do_auto_asconf) {
6902 memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
6903 inet_sk_copy_descendant(newsk, oldsk);
6904 memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
6906 inet_sk_copy_descendant(newsk, oldsk);
6908 /* Restore the ep value that was overwritten with the above structure
6914 /* Hook this new socket in to the bind_hash list. */
6915 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
6916 inet_sk(oldsk)->inet_num)];
6917 sctp_local_bh_disable();
6918 sctp_spin_lock(&head->lock);
6919 pp = sctp_sk(oldsk)->bind_hash;
6920 sk_add_bind_node(newsk, &pp->owner);
6921 sctp_sk(newsk)->bind_hash = pp;
6922 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6923 sctp_spin_unlock(&head->lock);
6924 sctp_local_bh_enable();
6926 /* Copy the bind_addr list from the original endpoint to the new
6927 * endpoint so that we can handle restarts properly
6929 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6930 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6932 /* Move any messages in the old socket's receive queue that are for the
6933 * peeled off association to the new socket's receive queue.
6935 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6936 event = sctp_skb2event(skb);
6937 if (event->asoc == assoc) {
6938 __skb_unlink(skb, &oldsk->sk_receive_queue);
6939 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6940 sctp_skb_set_owner_r_frag(skb, newsk);
6944 /* Clean up any messages pending delivery due to partial
6945 * delivery. Three cases:
6946 * 1) No partial deliver; no work.
6947 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6948 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6950 skb_queue_head_init(&newsp->pd_lobby);
6951 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6953 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6954 struct sk_buff_head *queue;
6956 /* Decide which queue to move pd_lobby skbs to. */
6957 if (assoc->ulpq.pd_mode) {
6958 queue = &newsp->pd_lobby;
6960 queue = &newsk->sk_receive_queue;
6962 /* Walk through the pd_lobby, looking for skbs that
6963 * need moved to the new socket.
6965 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6966 event = sctp_skb2event(skb);
6967 if (event->asoc == assoc) {
6968 __skb_unlink(skb, &oldsp->pd_lobby);
6969 __skb_queue_tail(queue, skb);
6970 sctp_skb_set_owner_r_frag(skb, newsk);
6974 /* Clear up any skbs waiting for the partial
6975 * delivery to finish.
6977 if (assoc->ulpq.pd_mode)
6978 sctp_clear_pd(oldsk, NULL);
6982 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6983 sctp_skb_set_owner_r_frag(skb, newsk);
6985 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6986 sctp_skb_set_owner_r_frag(skb, newsk);
6988 /* Set the type of socket to indicate that it is peeled off from the
6989 * original UDP-style socket or created with the accept() call on a
6990 * TCP-style socket..
6994 /* Mark the new socket "in-use" by the user so that any packets
6995 * that may arrive on the association after we've moved it are
6996 * queued to the backlog. This prevents a potential race between
6997 * backlog processing on the old socket and new-packet processing
6998 * on the new socket.
7000 * The caller has just allocated newsk so we can guarantee that other
7001 * paths won't try to lock it and then oldsk.
7003 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7004 sctp_assoc_migrate(assoc, newsk);
7006 /* If the association on the newsk is already closed before accept()
7007 * is called, set RCV_SHUTDOWN flag.
7009 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
7010 newsk->sk_shutdown |= RCV_SHUTDOWN;
7012 newsk->sk_state = SCTP_SS_ESTABLISHED;
7013 sctp_release_sock(newsk);
7017 /* This proto struct describes the ULP interface for SCTP. */
7018 struct proto sctp_prot = {
7020 .owner = THIS_MODULE,
7021 .close = sctp_close,
7022 .connect = sctp_connect,
7023 .disconnect = sctp_disconnect,
7024 .accept = sctp_accept,
7025 .ioctl = sctp_ioctl,
7026 .init = sctp_init_sock,
7027 .destroy = sctp_destroy_sock,
7028 .shutdown = sctp_shutdown,
7029 .setsockopt = sctp_setsockopt,
7030 .getsockopt = sctp_getsockopt,
7031 .sendmsg = sctp_sendmsg,
7032 .recvmsg = sctp_recvmsg,
7034 .backlog_rcv = sctp_backlog_rcv,
7036 .unhash = sctp_unhash,
7037 .get_port = sctp_get_port,
7038 .obj_size = sizeof(struct sctp_sock),
7039 .sysctl_mem = sysctl_sctp_mem,
7040 .sysctl_rmem = sysctl_sctp_rmem,
7041 .sysctl_wmem = sysctl_sctp_wmem,
7042 .memory_pressure = &sctp_memory_pressure,
7043 .enter_memory_pressure = sctp_enter_memory_pressure,
7044 .memory_allocated = &sctp_memory_allocated,
7045 .sockets_allocated = &sctp_sockets_allocated,
7048 #if IS_ENABLED(CONFIG_IPV6)
7050 struct proto sctpv6_prot = {
7052 .owner = THIS_MODULE,
7053 .close = sctp_close,
7054 .connect = sctp_connect,
7055 .disconnect = sctp_disconnect,
7056 .accept = sctp_accept,
7057 .ioctl = sctp_ioctl,
7058 .init = sctp_init_sock,
7059 .destroy = sctp_destroy_sock,
7060 .shutdown = sctp_shutdown,
7061 .setsockopt = sctp_setsockopt,
7062 .getsockopt = sctp_getsockopt,
7063 .sendmsg = sctp_sendmsg,
7064 .recvmsg = sctp_recvmsg,
7066 .backlog_rcv = sctp_backlog_rcv,
7068 .unhash = sctp_unhash,
7069 .get_port = sctp_get_port,
7070 .obj_size = sizeof(struct sctp6_sock),
7071 .sysctl_mem = sysctl_sctp_mem,
7072 .sysctl_rmem = sysctl_sctp_rmem,
7073 .sysctl_wmem = sysctl_sctp_wmem,
7074 .memory_pressure = &sctp_memory_pressure,
7075 .enter_memory_pressure = sctp_enter_memory_pressure,
7076 .memory_allocated = &sctp_memory_allocated,
7077 .sockets_allocated = &sctp_sockets_allocated,
7079 #endif /* IS_ENABLED(CONFIG_IPV6) */