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,
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. */
1510 /* Supposedly, no process has access to the socket, but
1511 * the net layers still may.
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);
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_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);
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);
2096 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2097 (event->msg_flags & MSG_EOR))
2098 msg->msg_flags |= MSG_EOR;
2100 msg->msg_flags &= ~MSG_EOR;
2103 if (flags & MSG_PEEK) {
2104 /* Release the skb reference acquired after peeking the skb in
2105 * sctp_skb_recv_datagram().
2109 /* Free the event which includes releasing the reference to
2110 * the owner of the skb, freeing the skb and updating the
2113 sctp_ulpevent_free(event);
2120 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2122 * This option is a on/off flag. If enabled no SCTP message
2123 * fragmentation will be performed. Instead if a message being sent
2124 * exceeds the current PMTU size, the message will NOT be sent and
2125 * instead a error will be indicated to the user.
2127 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2128 char __user *optval,
2129 unsigned int optlen)
2133 if (optlen < sizeof(int))
2136 if (get_user(val, (int __user *)optval))
2139 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2144 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2145 unsigned int optlen)
2147 struct sctp_association *asoc;
2148 struct sctp_ulpevent *event;
2150 if (optlen > sizeof(struct sctp_event_subscribe))
2152 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2156 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2157 * if there is no data to be sent or retransmit, the stack will
2158 * immediately send up this notification.
2160 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2161 &sctp_sk(sk)->subscribe)) {
2162 asoc = sctp_id2assoc(sk, 0);
2164 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2165 event = sctp_ulpevent_make_sender_dry_event(asoc,
2170 sctp_ulpq_tail_event(&asoc->ulpq, event);
2177 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2179 * This socket option is applicable to the UDP-style socket only. When
2180 * set it will cause associations that are idle for more than the
2181 * specified number of seconds to automatically close. An association
2182 * being idle is defined an association that has NOT sent or received
2183 * user data. The special value of '0' indicates that no automatic
2184 * close of any associations should be performed. The option expects an
2185 * integer defining the number of seconds of idle time before an
2186 * association is closed.
2188 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2189 unsigned int optlen)
2191 struct sctp_sock *sp = sctp_sk(sk);
2192 struct net *net = sock_net(sk);
2194 /* Applicable to UDP-style socket only */
2195 if (sctp_style(sk, TCP))
2197 if (optlen != sizeof(int))
2199 if (copy_from_user(&sp->autoclose, optval, optlen))
2202 if (sp->autoclose > net->sctp.max_autoclose)
2203 sp->autoclose = net->sctp.max_autoclose;
2208 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2210 * Applications can enable or disable heartbeats for any peer address of
2211 * an association, modify an address's heartbeat interval, force a
2212 * heartbeat to be sent immediately, and adjust the address's maximum
2213 * number of retransmissions sent before an address is considered
2214 * unreachable. The following structure is used to access and modify an
2215 * address's parameters:
2217 * struct sctp_paddrparams {
2218 * sctp_assoc_t spp_assoc_id;
2219 * struct sockaddr_storage spp_address;
2220 * uint32_t spp_hbinterval;
2221 * uint16_t spp_pathmaxrxt;
2222 * uint32_t spp_pathmtu;
2223 * uint32_t spp_sackdelay;
2224 * uint32_t spp_flags;
2227 * spp_assoc_id - (one-to-many style socket) This is filled in the
2228 * application, and identifies the association for
2230 * spp_address - This specifies which address is of interest.
2231 * spp_hbinterval - This contains the value of the heartbeat interval,
2232 * in milliseconds. If a value of zero
2233 * is present in this field then no changes are to
2234 * be made to this parameter.
2235 * spp_pathmaxrxt - This contains the maximum number of
2236 * retransmissions before this address shall be
2237 * considered unreachable. If a value of zero
2238 * is present in this field then no changes are to
2239 * be made to this parameter.
2240 * spp_pathmtu - When Path MTU discovery is disabled the value
2241 * specified here will be the "fixed" path mtu.
2242 * Note that if the spp_address field is empty
2243 * then all associations on this address will
2244 * have this fixed path mtu set upon them.
2246 * spp_sackdelay - When delayed sack is enabled, this value specifies
2247 * the number of milliseconds that sacks will be delayed
2248 * for. This value will apply to all addresses of an
2249 * association if the spp_address field is empty. Note
2250 * also, that if delayed sack is enabled and this
2251 * value is set to 0, no change is made to the last
2252 * recorded delayed sack timer value.
2254 * spp_flags - These flags are used to control various features
2255 * on an association. The flag field may contain
2256 * zero or more of the following options.
2258 * SPP_HB_ENABLE - Enable heartbeats on the
2259 * specified address. Note that if the address
2260 * field is empty all addresses for the association
2261 * have heartbeats enabled upon them.
2263 * SPP_HB_DISABLE - Disable heartbeats on the
2264 * speicifed address. Note that if the address
2265 * field is empty all addresses for the association
2266 * will have their heartbeats disabled. Note also
2267 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2268 * mutually exclusive, only one of these two should
2269 * be specified. Enabling both fields will have
2270 * undetermined results.
2272 * SPP_HB_DEMAND - Request a user initiated heartbeat
2273 * to be made immediately.
2275 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2276 * heartbeat delayis to be set to the value of 0
2279 * SPP_PMTUD_ENABLE - This field will enable PMTU
2280 * discovery upon the specified address. Note that
2281 * if the address feild is empty then all addresses
2282 * on the association are effected.
2284 * SPP_PMTUD_DISABLE - This field will disable PMTU
2285 * discovery upon the specified address. Note that
2286 * if the address feild is empty then all addresses
2287 * on the association are effected. Not also that
2288 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2289 * exclusive. Enabling both will have undetermined
2292 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2293 * on delayed sack. The time specified in spp_sackdelay
2294 * is used to specify the sack delay for this address. Note
2295 * that if spp_address is empty then all addresses will
2296 * enable delayed sack and take on the sack delay
2297 * value specified in spp_sackdelay.
2298 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2299 * off delayed sack. If the spp_address field is blank then
2300 * delayed sack is disabled for the entire association. Note
2301 * also that this field is mutually exclusive to
2302 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2305 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2306 struct sctp_transport *trans,
2307 struct sctp_association *asoc,
2308 struct sctp_sock *sp,
2311 int sackdelay_change)
2315 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2316 struct net *net = sock_net(trans->asoc->base.sk);
2318 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2323 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2324 * this field is ignored. Note also that a value of zero indicates
2325 * the current setting should be left unchanged.
2327 if (params->spp_flags & SPP_HB_ENABLE) {
2329 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2330 * set. This lets us use 0 value when this flag
2333 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2334 params->spp_hbinterval = 0;
2336 if (params->spp_hbinterval ||
2337 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2340 msecs_to_jiffies(params->spp_hbinterval);
2343 msecs_to_jiffies(params->spp_hbinterval);
2345 sp->hbinterval = params->spp_hbinterval;
2352 trans->param_flags =
2353 (trans->param_flags & ~SPP_HB) | hb_change;
2356 (asoc->param_flags & ~SPP_HB) | hb_change;
2359 (sp->param_flags & ~SPP_HB) | hb_change;
2363 /* When Path MTU discovery is disabled the value specified here will
2364 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2365 * include the flag SPP_PMTUD_DISABLE for this field to have any
2368 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2370 trans->pathmtu = params->spp_pathmtu;
2371 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2373 asoc->pathmtu = params->spp_pathmtu;
2374 sctp_frag_point(asoc, params->spp_pathmtu);
2376 sp->pathmtu = params->spp_pathmtu;
2382 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2383 (params->spp_flags & SPP_PMTUD_ENABLE);
2384 trans->param_flags =
2385 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2387 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2388 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2392 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2395 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2399 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2400 * value of this field is ignored. Note also that a value of zero
2401 * indicates the current setting should be left unchanged.
2403 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2406 msecs_to_jiffies(params->spp_sackdelay);
2409 msecs_to_jiffies(params->spp_sackdelay);
2411 sp->sackdelay = params->spp_sackdelay;
2415 if (sackdelay_change) {
2417 trans->param_flags =
2418 (trans->param_flags & ~SPP_SACKDELAY) |
2422 (asoc->param_flags & ~SPP_SACKDELAY) |
2426 (sp->param_flags & ~SPP_SACKDELAY) |
2431 /* Note that a value of zero indicates the current setting should be
2434 if (params->spp_pathmaxrxt) {
2436 trans->pathmaxrxt = params->spp_pathmaxrxt;
2438 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2440 sp->pathmaxrxt = params->spp_pathmaxrxt;
2447 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2448 char __user *optval,
2449 unsigned int optlen)
2451 struct sctp_paddrparams params;
2452 struct sctp_transport *trans = NULL;
2453 struct sctp_association *asoc = NULL;
2454 struct sctp_sock *sp = sctp_sk(sk);
2456 int hb_change, pmtud_change, sackdelay_change;
2458 if (optlen != sizeof(struct sctp_paddrparams))
2461 if (copy_from_user(¶ms, optval, optlen))
2464 /* Validate flags and value parameters. */
2465 hb_change = params.spp_flags & SPP_HB;
2466 pmtud_change = params.spp_flags & SPP_PMTUD;
2467 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2469 if (hb_change == SPP_HB ||
2470 pmtud_change == SPP_PMTUD ||
2471 sackdelay_change == SPP_SACKDELAY ||
2472 params.spp_sackdelay > 500 ||
2473 (params.spp_pathmtu &&
2474 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2477 /* If an address other than INADDR_ANY is specified, and
2478 * no transport is found, then the request is invalid.
2480 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
2481 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2482 params.spp_assoc_id);
2487 /* Get association, if assoc_id != 0 and the socket is a one
2488 * to many style socket, and an association was not found, then
2489 * the id was invalid.
2491 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2492 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2495 /* Heartbeat demand can only be sent on a transport or
2496 * association, but not a socket.
2498 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2501 /* Process parameters. */
2502 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2503 hb_change, pmtud_change,
2509 /* If changes are for association, also apply parameters to each
2512 if (!trans && asoc) {
2513 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2515 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2516 hb_change, pmtud_change,
2524 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2526 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2529 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2531 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2535 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2537 * This option will effect the way delayed acks are performed. This
2538 * option allows you to get or set the delayed ack time, in
2539 * milliseconds. It also allows changing the delayed ack frequency.
2540 * Changing the frequency to 1 disables the delayed sack algorithm. If
2541 * the assoc_id is 0, then this sets or gets the endpoints default
2542 * values. If the assoc_id field is non-zero, then the set or get
2543 * effects the specified association for the one to many model (the
2544 * assoc_id field is ignored by the one to one model). Note that if
2545 * sack_delay or sack_freq are 0 when setting this option, then the
2546 * current values will remain unchanged.
2548 * struct sctp_sack_info {
2549 * sctp_assoc_t sack_assoc_id;
2550 * uint32_t sack_delay;
2551 * uint32_t sack_freq;
2554 * sack_assoc_id - This parameter, indicates which association the user
2555 * is performing an action upon. Note that if this field's value is
2556 * zero then the endpoints default value is changed (effecting future
2557 * associations only).
2559 * sack_delay - This parameter contains the number of milliseconds that
2560 * the user is requesting the delayed ACK timer be set to. Note that
2561 * this value is defined in the standard to be between 200 and 500
2564 * sack_freq - This parameter contains the number of packets that must
2565 * be received before a sack is sent without waiting for the delay
2566 * timer to expire. The default value for this is 2, setting this
2567 * value to 1 will disable the delayed sack algorithm.
2570 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2571 char __user *optval, unsigned int optlen)
2573 struct sctp_sack_info params;
2574 struct sctp_transport *trans = NULL;
2575 struct sctp_association *asoc = NULL;
2576 struct sctp_sock *sp = sctp_sk(sk);
2578 if (optlen == sizeof(struct sctp_sack_info)) {
2579 if (copy_from_user(¶ms, optval, optlen))
2582 if (params.sack_delay == 0 && params.sack_freq == 0)
2584 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2585 pr_warn_ratelimited(DEPRECATED
2587 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2588 "Use struct sctp_sack_info instead\n",
2589 current->comm, task_pid_nr(current));
2590 if (copy_from_user(¶ms, optval, optlen))
2593 if (params.sack_delay == 0)
2594 params.sack_freq = 1;
2596 params.sack_freq = 0;
2600 /* Validate value parameter. */
2601 if (params.sack_delay > 500)
2604 /* Get association, if sack_assoc_id != 0 and the socket is a one
2605 * to many style socket, and an association was not found, then
2606 * the id was invalid.
2608 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2609 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2612 if (params.sack_delay) {
2615 msecs_to_jiffies(params.sack_delay);
2617 sctp_spp_sackdelay_enable(asoc->param_flags);
2619 sp->sackdelay = params.sack_delay;
2621 sctp_spp_sackdelay_enable(sp->param_flags);
2625 if (params.sack_freq == 1) {
2628 sctp_spp_sackdelay_disable(asoc->param_flags);
2631 sctp_spp_sackdelay_disable(sp->param_flags);
2633 } else if (params.sack_freq > 1) {
2635 asoc->sackfreq = params.sack_freq;
2637 sctp_spp_sackdelay_enable(asoc->param_flags);
2639 sp->sackfreq = params.sack_freq;
2641 sctp_spp_sackdelay_enable(sp->param_flags);
2645 /* If change is for association, also apply to each transport. */
2647 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2649 if (params.sack_delay) {
2651 msecs_to_jiffies(params.sack_delay);
2652 trans->param_flags =
2653 sctp_spp_sackdelay_enable(trans->param_flags);
2655 if (params.sack_freq == 1) {
2656 trans->param_flags =
2657 sctp_spp_sackdelay_disable(trans->param_flags);
2658 } else if (params.sack_freq > 1) {
2659 trans->sackfreq = params.sack_freq;
2660 trans->param_flags =
2661 sctp_spp_sackdelay_enable(trans->param_flags);
2669 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2671 * Applications can specify protocol parameters for the default association
2672 * initialization. The option name argument to setsockopt() and getsockopt()
2675 * Setting initialization parameters is effective only on an unconnected
2676 * socket (for UDP-style sockets only future associations are effected
2677 * by the change). With TCP-style sockets, this option is inherited by
2678 * sockets derived from a listener socket.
2680 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2682 struct sctp_initmsg sinit;
2683 struct sctp_sock *sp = sctp_sk(sk);
2685 if (optlen != sizeof(struct sctp_initmsg))
2687 if (copy_from_user(&sinit, optval, optlen))
2690 if (sinit.sinit_num_ostreams)
2691 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2692 if (sinit.sinit_max_instreams)
2693 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2694 if (sinit.sinit_max_attempts)
2695 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2696 if (sinit.sinit_max_init_timeo)
2697 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2703 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2705 * Applications that wish to use the sendto() system call may wish to
2706 * specify a default set of parameters that would normally be supplied
2707 * through the inclusion of ancillary data. This socket option allows
2708 * such an application to set the default sctp_sndrcvinfo structure.
2709 * The application that wishes to use this socket option simply passes
2710 * in to this call the sctp_sndrcvinfo structure defined in Section
2711 * 5.2.2) The input parameters accepted by this call include
2712 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2713 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2714 * to this call if the caller is using the UDP model.
2716 static int sctp_setsockopt_default_send_param(struct sock *sk,
2717 char __user *optval,
2718 unsigned int optlen)
2720 struct sctp_sndrcvinfo info;
2721 struct sctp_association *asoc;
2722 struct sctp_sock *sp = sctp_sk(sk);
2724 if (optlen != sizeof(struct sctp_sndrcvinfo))
2726 if (copy_from_user(&info, optval, optlen))
2729 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2730 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2734 asoc->default_stream = info.sinfo_stream;
2735 asoc->default_flags = info.sinfo_flags;
2736 asoc->default_ppid = info.sinfo_ppid;
2737 asoc->default_context = info.sinfo_context;
2738 asoc->default_timetolive = info.sinfo_timetolive;
2740 sp->default_stream = info.sinfo_stream;
2741 sp->default_flags = info.sinfo_flags;
2742 sp->default_ppid = info.sinfo_ppid;
2743 sp->default_context = info.sinfo_context;
2744 sp->default_timetolive = info.sinfo_timetolive;
2750 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2752 * Requests that the local SCTP stack use the enclosed peer address as
2753 * the association primary. The enclosed address must be one of the
2754 * association peer's addresses.
2756 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2757 unsigned int optlen)
2759 struct sctp_prim prim;
2760 struct sctp_transport *trans;
2762 if (optlen != sizeof(struct sctp_prim))
2765 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2768 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2772 sctp_assoc_set_primary(trans->asoc, trans);
2778 * 7.1.5 SCTP_NODELAY
2780 * Turn on/off any Nagle-like algorithm. This means that packets are
2781 * generally sent as soon as possible and no unnecessary delays are
2782 * introduced, at the cost of more packets in the network. Expects an
2783 * integer boolean flag.
2785 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2786 unsigned int optlen)
2790 if (optlen < sizeof(int))
2792 if (get_user(val, (int __user *)optval))
2795 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2801 * 7.1.1 SCTP_RTOINFO
2803 * The protocol parameters used to initialize and bound retransmission
2804 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2805 * and modify these parameters.
2806 * All parameters are time values, in milliseconds. A value of 0, when
2807 * modifying the parameters, indicates that the current value should not
2811 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2813 struct sctp_rtoinfo rtoinfo;
2814 struct sctp_association *asoc;
2815 unsigned long rto_min, rto_max;
2816 struct sctp_sock *sp = sctp_sk(sk);
2818 if (optlen != sizeof (struct sctp_rtoinfo))
2821 if (copy_from_user(&rtoinfo, optval, optlen))
2824 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2826 /* Set the values to the specific association */
2827 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2830 rto_max = rtoinfo.srto_max;
2831 rto_min = rtoinfo.srto_min;
2834 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2836 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2839 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2841 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2843 if (rto_min > rto_max)
2847 if (rtoinfo.srto_initial != 0)
2849 msecs_to_jiffies(rtoinfo.srto_initial);
2850 asoc->rto_max = rto_max;
2851 asoc->rto_min = rto_min;
2853 /* If there is no association or the association-id = 0
2854 * set the values to the endpoint.
2856 if (rtoinfo.srto_initial != 0)
2857 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2858 sp->rtoinfo.srto_max = rto_max;
2859 sp->rtoinfo.srto_min = rto_min;
2867 * 7.1.2 SCTP_ASSOCINFO
2869 * This option is used to tune the maximum retransmission attempts
2870 * of the association.
2871 * Returns an error if the new association retransmission value is
2872 * greater than the sum of the retransmission value of the peer.
2873 * See [SCTP] for more information.
2876 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2879 struct sctp_assocparams assocparams;
2880 struct sctp_association *asoc;
2882 if (optlen != sizeof(struct sctp_assocparams))
2884 if (copy_from_user(&assocparams, optval, optlen))
2887 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2889 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2892 /* Set the values to the specific association */
2894 if (assocparams.sasoc_asocmaxrxt != 0) {
2897 struct sctp_transport *peer_addr;
2899 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2901 path_sum += peer_addr->pathmaxrxt;
2905 /* Only validate asocmaxrxt if we have more than
2906 * one path/transport. We do this because path
2907 * retransmissions are only counted when we have more
2911 assocparams.sasoc_asocmaxrxt > path_sum)
2914 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2917 if (assocparams.sasoc_cookie_life != 0)
2918 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
2920 /* Set the values to the endpoint */
2921 struct sctp_sock *sp = sctp_sk(sk);
2923 if (assocparams.sasoc_asocmaxrxt != 0)
2924 sp->assocparams.sasoc_asocmaxrxt =
2925 assocparams.sasoc_asocmaxrxt;
2926 if (assocparams.sasoc_cookie_life != 0)
2927 sp->assocparams.sasoc_cookie_life =
2928 assocparams.sasoc_cookie_life;
2934 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2936 * This socket option is a boolean flag which turns on or off mapped V4
2937 * addresses. If this option is turned on and the socket is type
2938 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2939 * If this option is turned off, then no mapping will be done of V4
2940 * addresses and a user will receive both PF_INET6 and PF_INET type
2941 * addresses on the socket.
2943 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2946 struct sctp_sock *sp = sctp_sk(sk);
2948 if (optlen < sizeof(int))
2950 if (get_user(val, (int __user *)optval))
2961 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2962 * This option will get or set the maximum size to put in any outgoing
2963 * SCTP DATA chunk. If a message is larger than this size it will be
2964 * fragmented by SCTP into the specified size. Note that the underlying
2965 * SCTP implementation may fragment into smaller sized chunks when the
2966 * PMTU of the underlying association is smaller than the value set by
2967 * the user. The default value for this option is '0' which indicates
2968 * the user is NOT limiting fragmentation and only the PMTU will effect
2969 * SCTP's choice of DATA chunk size. Note also that values set larger
2970 * than the maximum size of an IP datagram will effectively let SCTP
2971 * control fragmentation (i.e. the same as setting this option to 0).
2973 * The following structure is used to access and modify this parameter:
2975 * struct sctp_assoc_value {
2976 * sctp_assoc_t assoc_id;
2977 * uint32_t assoc_value;
2980 * assoc_id: This parameter is ignored for one-to-one style sockets.
2981 * For one-to-many style sockets this parameter indicates which
2982 * association the user is performing an action upon. Note that if
2983 * this field's value is zero then the endpoints default value is
2984 * changed (effecting future associations only).
2985 * assoc_value: This parameter specifies the maximum size in bytes.
2987 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2989 struct sctp_assoc_value params;
2990 struct sctp_association *asoc;
2991 struct sctp_sock *sp = sctp_sk(sk);
2994 if (optlen == sizeof(int)) {
2995 pr_warn_ratelimited(DEPRECATED
2997 "Use of int in maxseg socket option.\n"
2998 "Use struct sctp_assoc_value instead\n",
2999 current->comm, task_pid_nr(current));
3000 if (copy_from_user(&val, optval, optlen))
3002 params.assoc_id = 0;
3003 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3004 if (copy_from_user(¶ms, optval, optlen))
3006 val = params.assoc_value;
3010 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3013 asoc = sctp_id2assoc(sk, params.assoc_id);
3014 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3019 val = asoc->pathmtu;
3020 val -= sp->pf->af->net_header_len;
3021 val -= sizeof(struct sctphdr) +
3022 sizeof(struct sctp_data_chunk);
3024 asoc->user_frag = val;
3025 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3027 sp->user_frag = val;
3035 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3037 * Requests that the peer mark the enclosed address as the association
3038 * primary. The enclosed address must be one of the association's
3039 * locally bound addresses. The following structure is used to make a
3040 * set primary request:
3042 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3043 unsigned int optlen)
3045 struct net *net = sock_net(sk);
3046 struct sctp_sock *sp;
3047 struct sctp_association *asoc = NULL;
3048 struct sctp_setpeerprim prim;
3049 struct sctp_chunk *chunk;
3055 if (!net->sctp.addip_enable)
3058 if (optlen != sizeof(struct sctp_setpeerprim))
3061 if (copy_from_user(&prim, optval, optlen))
3064 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3068 if (!asoc->peer.asconf_capable)
3071 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3074 if (!sctp_state(asoc, ESTABLISHED))
3077 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3081 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3082 return -EADDRNOTAVAIL;
3084 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3085 return -EADDRNOTAVAIL;
3087 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3088 chunk = sctp_make_asconf_set_prim(asoc,
3089 (union sctp_addr *)&prim.sspp_addr);
3093 err = sctp_send_asconf(asoc, chunk);
3095 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3100 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3101 unsigned int optlen)
3103 struct sctp_setadaptation adaptation;
3105 if (optlen != sizeof(struct sctp_setadaptation))
3107 if (copy_from_user(&adaptation, optval, optlen))
3110 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3116 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3118 * The context field in the sctp_sndrcvinfo structure is normally only
3119 * used when a failed message is retrieved holding the value that was
3120 * sent down on the actual send call. This option allows the setting of
3121 * a default context on an association basis that will be received on
3122 * reading messages from the peer. This is especially helpful in the
3123 * one-2-many model for an application to keep some reference to an
3124 * internal state machine that is processing messages on the
3125 * association. Note that the setting of this value only effects
3126 * received messages from the peer and does not effect the value that is
3127 * saved with outbound messages.
3129 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3130 unsigned int optlen)
3132 struct sctp_assoc_value params;
3133 struct sctp_sock *sp;
3134 struct sctp_association *asoc;
3136 if (optlen != sizeof(struct sctp_assoc_value))
3138 if (copy_from_user(¶ms, optval, optlen))
3143 if (params.assoc_id != 0) {
3144 asoc = sctp_id2assoc(sk, params.assoc_id);
3147 asoc->default_rcv_context = params.assoc_value;
3149 sp->default_rcv_context = params.assoc_value;
3156 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3158 * This options will at a minimum specify if the implementation is doing
3159 * fragmented interleave. Fragmented interleave, for a one to many
3160 * socket, is when subsequent calls to receive a message may return
3161 * parts of messages from different associations. Some implementations
3162 * may allow you to turn this value on or off. If so, when turned off,
3163 * no fragment interleave will occur (which will cause a head of line
3164 * blocking amongst multiple associations sharing the same one to many
3165 * socket). When this option is turned on, then each receive call may
3166 * come from a different association (thus the user must receive data
3167 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3168 * association each receive belongs to.
3170 * This option takes a boolean value. A non-zero value indicates that
3171 * fragmented interleave is on. A value of zero indicates that
3172 * fragmented interleave is off.
3174 * Note that it is important that an implementation that allows this
3175 * option to be turned on, have it off by default. Otherwise an unaware
3176 * application using the one to many model may become confused and act
3179 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3180 char __user *optval,
3181 unsigned int optlen)
3185 if (optlen != sizeof(int))
3187 if (get_user(val, (int __user *)optval))
3190 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3196 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3197 * (SCTP_PARTIAL_DELIVERY_POINT)
3199 * This option will set or get the SCTP partial delivery point. This
3200 * point is the size of a message where the partial delivery API will be
3201 * invoked to help free up rwnd space for the peer. Setting this to a
3202 * lower value will cause partial deliveries to happen more often. The
3203 * calls argument is an integer that sets or gets the partial delivery
3204 * point. Note also that the call will fail if the user attempts to set
3205 * this value larger than the socket receive buffer size.
3207 * Note that any single message having a length smaller than or equal to
3208 * the SCTP partial delivery point will be delivered in one single read
3209 * call as long as the user provided buffer is large enough to hold the
3212 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3213 char __user *optval,
3214 unsigned int optlen)
3218 if (optlen != sizeof(u32))
3220 if (get_user(val, (int __user *)optval))
3223 /* Note: We double the receive buffer from what the user sets
3224 * it to be, also initial rwnd is based on rcvbuf/2.
3226 if (val > (sk->sk_rcvbuf >> 1))
3229 sctp_sk(sk)->pd_point = val;
3231 return 0; /* is this the right error code? */
3235 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3237 * This option will allow a user to change the maximum burst of packets
3238 * that can be emitted by this association. Note that the default value
3239 * is 4, and some implementations may restrict this setting so that it
3240 * can only be lowered.
3242 * NOTE: This text doesn't seem right. Do this on a socket basis with
3243 * future associations inheriting the socket value.
3245 static int sctp_setsockopt_maxburst(struct sock *sk,
3246 char __user *optval,
3247 unsigned int optlen)
3249 struct sctp_assoc_value params;
3250 struct sctp_sock *sp;
3251 struct sctp_association *asoc;
3255 if (optlen == sizeof(int)) {
3256 pr_warn_ratelimited(DEPRECATED
3258 "Use of int in max_burst socket option deprecated.\n"
3259 "Use struct sctp_assoc_value instead\n",
3260 current->comm, task_pid_nr(current));
3261 if (copy_from_user(&val, optval, optlen))
3263 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3264 if (copy_from_user(¶ms, optval, optlen))
3266 val = params.assoc_value;
3267 assoc_id = params.assoc_id;
3273 if (assoc_id != 0) {
3274 asoc = sctp_id2assoc(sk, assoc_id);
3277 asoc->max_burst = val;
3279 sp->max_burst = val;
3285 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3287 * This set option adds a chunk type that the user is requesting to be
3288 * received only in an authenticated way. Changes to the list of chunks
3289 * will only effect future associations on the socket.
3291 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3292 char __user *optval,
3293 unsigned int optlen)
3295 struct net *net = sock_net(sk);
3296 struct sctp_authchunk val;
3298 if (!net->sctp.auth_enable)
3301 if (optlen != sizeof(struct sctp_authchunk))
3303 if (copy_from_user(&val, optval, optlen))
3306 switch (val.sauth_chunk) {
3308 case SCTP_CID_INIT_ACK:
3309 case SCTP_CID_SHUTDOWN_COMPLETE:
3314 /* add this chunk id to the endpoint */
3315 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3319 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3321 * This option gets or sets the list of HMAC algorithms that the local
3322 * endpoint requires the peer to use.
3324 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3325 char __user *optval,
3326 unsigned int optlen)
3328 struct net *net = sock_net(sk);
3329 struct sctp_hmacalgo *hmacs;
3333 if (!net->sctp.auth_enable)
3336 if (optlen < sizeof(struct sctp_hmacalgo))
3339 hmacs = memdup_user(optval, optlen);
3341 return PTR_ERR(hmacs);
3343 idents = hmacs->shmac_num_idents;
3344 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3345 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3350 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3357 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3359 * This option will set a shared secret key which is used to build an
3360 * association shared key.
3362 static int sctp_setsockopt_auth_key(struct sock *sk,
3363 char __user *optval,
3364 unsigned int optlen)
3366 struct net *net = sock_net(sk);
3367 struct sctp_authkey *authkey;
3368 struct sctp_association *asoc;
3371 if (!net->sctp.auth_enable)
3374 if (optlen <= sizeof(struct sctp_authkey))
3377 authkey = memdup_user(optval, optlen);
3378 if (IS_ERR(authkey))
3379 return PTR_ERR(authkey);
3381 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3386 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3387 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3392 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3399 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3401 * This option will get or set the active shared key to be used to build
3402 * the association shared key.
3404 static int sctp_setsockopt_active_key(struct sock *sk,
3405 char __user *optval,
3406 unsigned int optlen)
3408 struct net *net = sock_net(sk);
3409 struct sctp_authkeyid val;
3410 struct sctp_association *asoc;
3412 if (!net->sctp.auth_enable)
3415 if (optlen != sizeof(struct sctp_authkeyid))
3417 if (copy_from_user(&val, optval, optlen))
3420 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3421 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3424 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3425 val.scact_keynumber);
3429 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3431 * This set option will delete a shared secret key from use.
3433 static int sctp_setsockopt_del_key(struct sock *sk,
3434 char __user *optval,
3435 unsigned int optlen)
3437 struct net *net = sock_net(sk);
3438 struct sctp_authkeyid val;
3439 struct sctp_association *asoc;
3441 if (!net->sctp.auth_enable)
3444 if (optlen != sizeof(struct sctp_authkeyid))
3446 if (copy_from_user(&val, optval, optlen))
3449 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3450 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3453 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3454 val.scact_keynumber);
3459 * 8.1.23 SCTP_AUTO_ASCONF
3461 * This option will enable or disable the use of the automatic generation of
3462 * ASCONF chunks to add and delete addresses to an existing association. Note
3463 * that this option has two caveats namely: a) it only affects sockets that
3464 * are bound to all addresses available to the SCTP stack, and b) the system
3465 * administrator may have an overriding control that turns the ASCONF feature
3466 * off no matter what setting the socket option may have.
3467 * This option expects an integer boolean flag, where a non-zero value turns on
3468 * the option, and a zero value turns off the option.
3469 * Note. In this implementation, socket operation overrides default parameter
3470 * being set by sysctl as well as FreeBSD implementation
3472 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3473 unsigned int optlen)
3476 struct sctp_sock *sp = sctp_sk(sk);
3478 if (optlen < sizeof(int))
3480 if (get_user(val, (int __user *)optval))
3482 if (!sctp_is_ep_boundall(sk) && val)
3484 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3487 if (val == 0 && sp->do_auto_asconf) {
3488 list_del(&sp->auto_asconf_list);
3489 sp->do_auto_asconf = 0;
3490 } else if (val && !sp->do_auto_asconf) {
3491 list_add_tail(&sp->auto_asconf_list,
3492 &sock_net(sk)->sctp.auto_asconf_splist);
3493 sp->do_auto_asconf = 1;
3500 * SCTP_PEER_ADDR_THLDS
3502 * This option allows us to alter the partially failed threshold for one or all
3503 * transports in an association. See Section 6.1 of:
3504 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3506 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3507 char __user *optval,
3508 unsigned int optlen)
3510 struct sctp_paddrthlds val;
3511 struct sctp_transport *trans;
3512 struct sctp_association *asoc;
3514 if (optlen < sizeof(struct sctp_paddrthlds))
3516 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3517 sizeof(struct sctp_paddrthlds)))
3521 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3522 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3525 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3527 if (val.spt_pathmaxrxt)
3528 trans->pathmaxrxt = val.spt_pathmaxrxt;
3529 trans->pf_retrans = val.spt_pathpfthld;
3532 if (val.spt_pathmaxrxt)
3533 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3534 asoc->pf_retrans = val.spt_pathpfthld;
3536 trans = sctp_addr_id2transport(sk, &val.spt_address,
3541 if (val.spt_pathmaxrxt)
3542 trans->pathmaxrxt = val.spt_pathmaxrxt;
3543 trans->pf_retrans = val.spt_pathpfthld;
3549 /* API 6.2 setsockopt(), getsockopt()
3551 * Applications use setsockopt() and getsockopt() to set or retrieve
3552 * socket options. Socket options are used to change the default
3553 * behavior of sockets calls. They are described in Section 7.
3557 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3558 * int __user *optlen);
3559 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3562 * sd - the socket descript.
3563 * level - set to IPPROTO_SCTP for all SCTP options.
3564 * optname - the option name.
3565 * optval - the buffer to store the value of the option.
3566 * optlen - the size of the buffer.
3568 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3569 char __user *optval, unsigned int optlen)
3573 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3575 /* I can hardly begin to describe how wrong this is. This is
3576 * so broken as to be worse than useless. The API draft
3577 * REALLY is NOT helpful here... I am not convinced that the
3578 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3579 * are at all well-founded.
3581 if (level != SOL_SCTP) {
3582 struct sctp_af *af = sctp_sk(sk)->pf->af;
3583 retval = af->setsockopt(sk, level, optname, optval, optlen);
3590 case SCTP_SOCKOPT_BINDX_ADD:
3591 /* 'optlen' is the size of the addresses buffer. */
3592 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3593 optlen, SCTP_BINDX_ADD_ADDR);
3596 case SCTP_SOCKOPT_BINDX_REM:
3597 /* 'optlen' is the size of the addresses buffer. */
3598 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3599 optlen, SCTP_BINDX_REM_ADDR);
3602 case SCTP_SOCKOPT_CONNECTX_OLD:
3603 /* 'optlen' is the size of the addresses buffer. */
3604 retval = sctp_setsockopt_connectx_old(sk,
3605 (struct sockaddr __user *)optval,
3609 case SCTP_SOCKOPT_CONNECTX:
3610 /* 'optlen' is the size of the addresses buffer. */
3611 retval = sctp_setsockopt_connectx(sk,
3612 (struct sockaddr __user *)optval,
3616 case SCTP_DISABLE_FRAGMENTS:
3617 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3621 retval = sctp_setsockopt_events(sk, optval, optlen);
3624 case SCTP_AUTOCLOSE:
3625 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3628 case SCTP_PEER_ADDR_PARAMS:
3629 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3632 case SCTP_DELAYED_SACK:
3633 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3635 case SCTP_PARTIAL_DELIVERY_POINT:
3636 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3640 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3642 case SCTP_DEFAULT_SEND_PARAM:
3643 retval = sctp_setsockopt_default_send_param(sk, optval,
3646 case SCTP_PRIMARY_ADDR:
3647 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3649 case SCTP_SET_PEER_PRIMARY_ADDR:
3650 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3653 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3656 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3658 case SCTP_ASSOCINFO:
3659 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3661 case SCTP_I_WANT_MAPPED_V4_ADDR:
3662 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3665 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3667 case SCTP_ADAPTATION_LAYER:
3668 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3671 retval = sctp_setsockopt_context(sk, optval, optlen);
3673 case SCTP_FRAGMENT_INTERLEAVE:
3674 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3676 case SCTP_MAX_BURST:
3677 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3679 case SCTP_AUTH_CHUNK:
3680 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3682 case SCTP_HMAC_IDENT:
3683 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3686 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3688 case SCTP_AUTH_ACTIVE_KEY:
3689 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3691 case SCTP_AUTH_DELETE_KEY:
3692 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3694 case SCTP_AUTO_ASCONF:
3695 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3697 case SCTP_PEER_ADDR_THLDS:
3698 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3701 retval = -ENOPROTOOPT;
3711 /* API 3.1.6 connect() - UDP Style Syntax
3713 * An application may use the connect() call in the UDP model to initiate an
3714 * association without sending data.
3718 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3720 * sd: the socket descriptor to have a new association added to.
3722 * nam: the address structure (either struct sockaddr_in or struct
3723 * sockaddr_in6 defined in RFC2553 [7]).
3725 * len: the size of the address.
3727 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3735 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
3738 /* Validate addr_len before calling common connect/connectx routine. */
3739 af = sctp_get_af_specific(addr->sa_family);
3740 if (!af || addr_len < af->sockaddr_len) {
3743 /* Pass correct addr len to common routine (so it knows there
3744 * is only one address being passed.
3746 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3753 /* FIXME: Write comments. */
3754 static int sctp_disconnect(struct sock *sk, int flags)
3756 return -EOPNOTSUPP; /* STUB */
3759 /* 4.1.4 accept() - TCP Style Syntax
3761 * Applications use accept() call to remove an established SCTP
3762 * association from the accept queue of the endpoint. A new socket
3763 * descriptor will be returned from accept() to represent the newly
3764 * formed association.
3766 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3768 struct sctp_sock *sp;
3769 struct sctp_endpoint *ep;
3770 struct sock *newsk = NULL;
3771 struct sctp_association *asoc;
3780 if (!sctp_style(sk, TCP)) {
3781 error = -EOPNOTSUPP;
3785 if (!sctp_sstate(sk, LISTENING)) {
3790 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3792 error = sctp_wait_for_accept(sk, timeo);
3796 /* We treat the list of associations on the endpoint as the accept
3797 * queue and pick the first association on the list.
3799 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3801 newsk = sp->pf->create_accept_sk(sk, asoc);
3807 /* Populate the fields of the newsk from the oldsk and migrate the
3808 * asoc to the newsk.
3810 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3818 /* The SCTP ioctl handler. */
3819 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3826 * SEQPACKET-style sockets in LISTENING state are valid, for
3827 * SCTP, so only discard TCP-style sockets in LISTENING state.
3829 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3834 struct sk_buff *skb;
3835 unsigned int amount = 0;
3837 skb = skb_peek(&sk->sk_receive_queue);
3840 * We will only return the amount of this packet since
3841 * that is all that will be read.
3845 rc = put_user(amount, (int __user *)arg);
3857 /* This is the function which gets called during socket creation to
3858 * initialized the SCTP-specific portion of the sock.
3859 * The sock structure should already be zero-filled memory.
3861 static int sctp_init_sock(struct sock *sk)
3863 struct net *net = sock_net(sk);
3864 struct sctp_sock *sp;
3866 pr_debug("%s: sk:%p\n", __func__, sk);
3870 /* Initialize the SCTP per socket area. */
3871 switch (sk->sk_type) {
3872 case SOCK_SEQPACKET:
3873 sp->type = SCTP_SOCKET_UDP;
3876 sp->type = SCTP_SOCKET_TCP;
3879 return -ESOCKTNOSUPPORT;
3882 /* Initialize default send parameters. These parameters can be
3883 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3885 sp->default_stream = 0;
3886 sp->default_ppid = 0;
3887 sp->default_flags = 0;
3888 sp->default_context = 0;
3889 sp->default_timetolive = 0;
3891 sp->default_rcv_context = 0;
3892 sp->max_burst = net->sctp.max_burst;
3894 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
3896 /* Initialize default setup parameters. These parameters
3897 * can be modified with the SCTP_INITMSG socket option or
3898 * overridden by the SCTP_INIT CMSG.
3900 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3901 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3902 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
3903 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
3905 /* Initialize default RTO related parameters. These parameters can
3906 * be modified for with the SCTP_RTOINFO socket option.
3908 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
3909 sp->rtoinfo.srto_max = net->sctp.rto_max;
3910 sp->rtoinfo.srto_min = net->sctp.rto_min;
3912 /* Initialize default association related parameters. These parameters
3913 * can be modified with the SCTP_ASSOCINFO socket option.
3915 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
3916 sp->assocparams.sasoc_number_peer_destinations = 0;
3917 sp->assocparams.sasoc_peer_rwnd = 0;
3918 sp->assocparams.sasoc_local_rwnd = 0;
3919 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
3921 /* Initialize default event subscriptions. By default, all the
3924 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3926 /* Default Peer Address Parameters. These defaults can
3927 * be modified via SCTP_PEER_ADDR_PARAMS
3929 sp->hbinterval = net->sctp.hb_interval;
3930 sp->pathmaxrxt = net->sctp.max_retrans_path;
3931 sp->pathmtu = 0; /* allow default discovery */
3932 sp->sackdelay = net->sctp.sack_timeout;
3934 sp->param_flags = SPP_HB_ENABLE |
3936 SPP_SACKDELAY_ENABLE;
3938 /* If enabled no SCTP message fragmentation will be performed.
3939 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3941 sp->disable_fragments = 0;
3943 /* Enable Nagle algorithm by default. */
3946 /* Enable by default. */
3949 /* Auto-close idle associations after the configured
3950 * number of seconds. A value of 0 disables this
3951 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3952 * for UDP-style sockets only.
3956 /* User specified fragmentation limit. */
3959 sp->adaptation_ind = 0;
3961 sp->pf = sctp_get_pf_specific(sk->sk_family);
3963 /* Control variables for partial data delivery. */
3964 atomic_set(&sp->pd_mode, 0);
3965 skb_queue_head_init(&sp->pd_lobby);
3966 sp->frag_interleave = 0;
3968 /* Create a per socket endpoint structure. Even if we
3969 * change the data structure relationships, this may still
3970 * be useful for storing pre-connect address information.
3972 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
3978 sk->sk_destruct = sctp_destruct_sock;
3980 SCTP_DBG_OBJCNT_INC(sock);
3983 percpu_counter_inc(&sctp_sockets_allocated);
3984 sock_prot_inuse_add(net, sk->sk_prot, 1);
3985 if (net->sctp.default_auto_asconf) {
3986 list_add_tail(&sp->auto_asconf_list,
3987 &net->sctp.auto_asconf_splist);
3988 sp->do_auto_asconf = 1;
3990 sp->do_auto_asconf = 0;
3996 /* Cleanup any SCTP per socket resources. */
3997 static void sctp_destroy_sock(struct sock *sk)
3999 struct sctp_sock *sp;
4001 pr_debug("%s: sk:%p\n", __func__, sk);
4003 /* Release our hold on the endpoint. */
4005 /* This could happen during socket init, thus we bail out
4006 * early, since the rest of the below is not setup either.
4011 if (sp->do_auto_asconf) {
4012 sp->do_auto_asconf = 0;
4013 list_del(&sp->auto_asconf_list);
4015 sctp_endpoint_free(sp->ep);
4017 percpu_counter_dec(&sctp_sockets_allocated);
4018 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4022 /* Triggered when there are no references on the socket anymore */
4023 static void sctp_destruct_sock(struct sock *sk)
4025 struct sctp_sock *sp = sctp_sk(sk);
4027 /* Free up the HMAC transform. */
4028 crypto_free_hash(sp->hmac);
4030 inet_sock_destruct(sk);
4033 /* API 4.1.7 shutdown() - TCP Style Syntax
4034 * int shutdown(int socket, int how);
4036 * sd - the socket descriptor of the association to be closed.
4037 * how - Specifies the type of shutdown. The values are
4040 * Disables further receive operations. No SCTP
4041 * protocol action is taken.
4043 * Disables further send operations, and initiates
4044 * the SCTP shutdown sequence.
4046 * Disables further send and receive operations
4047 * and initiates the SCTP shutdown sequence.
4049 static void sctp_shutdown(struct sock *sk, int how)
4051 struct net *net = sock_net(sk);
4052 struct sctp_endpoint *ep;
4053 struct sctp_association *asoc;
4055 if (!sctp_style(sk, TCP))
4058 if (how & SEND_SHUTDOWN) {
4059 ep = sctp_sk(sk)->ep;
4060 if (!list_empty(&ep->asocs)) {
4061 asoc = list_entry(ep->asocs.next,
4062 struct sctp_association, asocs);
4063 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4068 /* 7.2.1 Association Status (SCTP_STATUS)
4070 * Applications can retrieve current status information about an
4071 * association, including association state, peer receiver window size,
4072 * number of unacked data chunks, and number of data chunks pending
4073 * receipt. This information is read-only.
4075 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4076 char __user *optval,
4079 struct sctp_status status;
4080 struct sctp_association *asoc = NULL;
4081 struct sctp_transport *transport;
4082 sctp_assoc_t associd;
4085 if (len < sizeof(status)) {
4090 len = sizeof(status);
4091 if (copy_from_user(&status, optval, len)) {
4096 associd = status.sstat_assoc_id;
4097 asoc = sctp_id2assoc(sk, associd);
4103 transport = asoc->peer.primary_path;
4105 status.sstat_assoc_id = sctp_assoc2id(asoc);
4106 status.sstat_state = asoc->state;
4107 status.sstat_rwnd = asoc->peer.rwnd;
4108 status.sstat_unackdata = asoc->unack_data;
4110 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4111 status.sstat_instrms = asoc->c.sinit_max_instreams;
4112 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4113 status.sstat_fragmentation_point = asoc->frag_point;
4114 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4115 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4116 transport->af_specific->sockaddr_len);
4117 /* Map ipv4 address into v4-mapped-on-v6 address. */
4118 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4119 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4120 status.sstat_primary.spinfo_state = transport->state;
4121 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4122 status.sstat_primary.spinfo_srtt = transport->srtt;
4123 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4124 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4126 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4127 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4129 if (put_user(len, optlen)) {
4134 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4135 __func__, len, status.sstat_state, status.sstat_rwnd,
4136 status.sstat_assoc_id);
4138 if (copy_to_user(optval, &status, len)) {
4148 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4150 * Applications can retrieve information about a specific peer address
4151 * of an association, including its reachability state, congestion
4152 * window, and retransmission timer values. This information is
4155 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4156 char __user *optval,
4159 struct sctp_paddrinfo pinfo;
4160 struct sctp_transport *transport;
4163 if (len < sizeof(pinfo)) {
4168 len = sizeof(pinfo);
4169 if (copy_from_user(&pinfo, optval, len)) {
4174 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4175 pinfo.spinfo_assoc_id);
4179 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4180 pinfo.spinfo_state = transport->state;
4181 pinfo.spinfo_cwnd = transport->cwnd;
4182 pinfo.spinfo_srtt = transport->srtt;
4183 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4184 pinfo.spinfo_mtu = transport->pathmtu;
4186 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4187 pinfo.spinfo_state = SCTP_ACTIVE;
4189 if (put_user(len, optlen)) {
4194 if (copy_to_user(optval, &pinfo, len)) {
4203 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4205 * This option is a on/off flag. If enabled no SCTP message
4206 * fragmentation will be performed. Instead if a message being sent
4207 * exceeds the current PMTU size, the message will NOT be sent and
4208 * instead a error will be indicated to the user.
4210 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4211 char __user *optval, int __user *optlen)
4215 if (len < sizeof(int))
4219 val = (sctp_sk(sk)->disable_fragments == 1);
4220 if (put_user(len, optlen))
4222 if (copy_to_user(optval, &val, len))
4227 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4229 * This socket option is used to specify various notifications and
4230 * ancillary data the user wishes to receive.
4232 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4237 if (len > sizeof(struct sctp_event_subscribe))
4238 len = sizeof(struct sctp_event_subscribe);
4239 if (put_user(len, optlen))
4241 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4246 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4248 * This socket option is applicable to the UDP-style socket only. When
4249 * set it will cause associations that are idle for more than the
4250 * specified number of seconds to automatically close. An association
4251 * being idle is defined an association that has NOT sent or received
4252 * user data. The special value of '0' indicates that no automatic
4253 * close of any associations should be performed. The option expects an
4254 * integer defining the number of seconds of idle time before an
4255 * association is closed.
4257 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4259 /* Applicable to UDP-style socket only */
4260 if (sctp_style(sk, TCP))
4262 if (len < sizeof(int))
4265 if (put_user(len, optlen))
4267 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4272 /* Helper routine to branch off an association to a new socket. */
4273 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4275 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4276 struct socket *sock;
4283 /* An association cannot be branched off from an already peeled-off
4284 * socket, nor is this supported for tcp style sockets.
4286 if (!sctp_style(sk, UDP))
4289 /* Create a new socket. */
4290 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4294 sctp_copy_sock(sock->sk, sk, asoc);
4296 /* Make peeled-off sockets more like 1-1 accepted sockets.
4297 * Set the daddr and initialize id to something more random
4299 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4300 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4302 /* Populate the fields of the newsk from the oldsk and migrate the
4303 * asoc to the newsk.
4305 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4311 EXPORT_SYMBOL(sctp_do_peeloff);
4313 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4315 sctp_peeloff_arg_t peeloff;
4316 struct socket *newsock;
4317 struct file *newfile;
4320 if (len < sizeof(sctp_peeloff_arg_t))
4322 len = sizeof(sctp_peeloff_arg_t);
4323 if (copy_from_user(&peeloff, optval, len))
4326 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4330 /* Map the socket to an unused fd that can be returned to the user. */
4331 retval = get_unused_fd_flags(0);
4333 sock_release(newsock);
4337 newfile = sock_alloc_file(newsock, 0, NULL);
4338 if (unlikely(IS_ERR(newfile))) {
4339 put_unused_fd(retval);
4340 sock_release(newsock);
4341 return PTR_ERR(newfile);
4344 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4347 /* Return the fd mapped to the new socket. */
4348 if (put_user(len, optlen)) {
4350 put_unused_fd(retval);
4353 peeloff.sd = retval;
4354 if (copy_to_user(optval, &peeloff, len)) {
4356 put_unused_fd(retval);
4359 fd_install(retval, newfile);
4364 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4366 * Applications can enable or disable heartbeats for any peer address of
4367 * an association, modify an address's heartbeat interval, force a
4368 * heartbeat to be sent immediately, and adjust the address's maximum
4369 * number of retransmissions sent before an address is considered
4370 * unreachable. The following structure is used to access and modify an
4371 * address's parameters:
4373 * struct sctp_paddrparams {
4374 * sctp_assoc_t spp_assoc_id;
4375 * struct sockaddr_storage spp_address;
4376 * uint32_t spp_hbinterval;
4377 * uint16_t spp_pathmaxrxt;
4378 * uint32_t spp_pathmtu;
4379 * uint32_t spp_sackdelay;
4380 * uint32_t spp_flags;
4383 * spp_assoc_id - (one-to-many style socket) This is filled in the
4384 * application, and identifies the association for
4386 * spp_address - This specifies which address is of interest.
4387 * spp_hbinterval - This contains the value of the heartbeat interval,
4388 * in milliseconds. If a value of zero
4389 * is present in this field then no changes are to
4390 * be made to this parameter.
4391 * spp_pathmaxrxt - This contains the maximum number of
4392 * retransmissions before this address shall be
4393 * considered unreachable. If a value of zero
4394 * is present in this field then no changes are to
4395 * be made to this parameter.
4396 * spp_pathmtu - When Path MTU discovery is disabled the value
4397 * specified here will be the "fixed" path mtu.
4398 * Note that if the spp_address field is empty
4399 * then all associations on this address will
4400 * have this fixed path mtu set upon them.
4402 * spp_sackdelay - When delayed sack is enabled, this value specifies
4403 * the number of milliseconds that sacks will be delayed
4404 * for. This value will apply to all addresses of an
4405 * association if the spp_address field is empty. Note
4406 * also, that if delayed sack is enabled and this
4407 * value is set to 0, no change is made to the last
4408 * recorded delayed sack timer value.
4410 * spp_flags - These flags are used to control various features
4411 * on an association. The flag field may contain
4412 * zero or more of the following options.
4414 * SPP_HB_ENABLE - Enable heartbeats on the
4415 * specified address. Note that if the address
4416 * field is empty all addresses for the association
4417 * have heartbeats enabled upon them.
4419 * SPP_HB_DISABLE - Disable heartbeats on the
4420 * speicifed address. Note that if the address
4421 * field is empty all addresses for the association
4422 * will have their heartbeats disabled. Note also
4423 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4424 * mutually exclusive, only one of these two should
4425 * be specified. Enabling both fields will have
4426 * undetermined results.
4428 * SPP_HB_DEMAND - Request a user initiated heartbeat
4429 * to be made immediately.
4431 * SPP_PMTUD_ENABLE - This field will enable PMTU
4432 * discovery upon the specified address. Note that
4433 * if the address feild is empty then all addresses
4434 * on the association are effected.
4436 * SPP_PMTUD_DISABLE - This field will disable PMTU
4437 * discovery upon the specified address. Note that
4438 * if the address feild is empty then all addresses
4439 * on the association are effected. Not also that
4440 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4441 * exclusive. Enabling both will have undetermined
4444 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4445 * on delayed sack. The time specified in spp_sackdelay
4446 * is used to specify the sack delay for this address. Note
4447 * that if spp_address is empty then all addresses will
4448 * enable delayed sack and take on the sack delay
4449 * value specified in spp_sackdelay.
4450 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4451 * off delayed sack. If the spp_address field is blank then
4452 * delayed sack is disabled for the entire association. Note
4453 * also that this field is mutually exclusive to
4454 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4457 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4458 char __user *optval, int __user *optlen)
4460 struct sctp_paddrparams params;
4461 struct sctp_transport *trans = NULL;
4462 struct sctp_association *asoc = NULL;
4463 struct sctp_sock *sp = sctp_sk(sk);
4465 if (len < sizeof(struct sctp_paddrparams))
4467 len = sizeof(struct sctp_paddrparams);
4468 if (copy_from_user(¶ms, optval, len))
4471 /* If an address other than INADDR_ANY is specified, and
4472 * no transport is found, then the request is invalid.
4474 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
4475 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4476 params.spp_assoc_id);
4478 pr_debug("%s: failed no transport\n", __func__);
4483 /* Get association, if assoc_id != 0 and the socket is a one
4484 * to many style socket, and an association was not found, then
4485 * the id was invalid.
4487 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4488 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4489 pr_debug("%s: failed no association\n", __func__);
4494 /* Fetch transport values. */
4495 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4496 params.spp_pathmtu = trans->pathmtu;
4497 params.spp_pathmaxrxt = trans->pathmaxrxt;
4498 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4500 /*draft-11 doesn't say what to return in spp_flags*/
4501 params.spp_flags = trans->param_flags;
4503 /* Fetch association values. */
4504 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4505 params.spp_pathmtu = asoc->pathmtu;
4506 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4507 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4509 /*draft-11 doesn't say what to return in spp_flags*/
4510 params.spp_flags = asoc->param_flags;
4512 /* Fetch socket values. */
4513 params.spp_hbinterval = sp->hbinterval;
4514 params.spp_pathmtu = sp->pathmtu;
4515 params.spp_sackdelay = sp->sackdelay;
4516 params.spp_pathmaxrxt = sp->pathmaxrxt;
4518 /*draft-11 doesn't say what to return in spp_flags*/
4519 params.spp_flags = sp->param_flags;
4522 if (copy_to_user(optval, ¶ms, len))
4525 if (put_user(len, optlen))
4532 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4534 * This option will effect the way delayed acks are performed. This
4535 * option allows you to get or set the delayed ack time, in
4536 * milliseconds. It also allows changing the delayed ack frequency.
4537 * Changing the frequency to 1 disables the delayed sack algorithm. If
4538 * the assoc_id is 0, then this sets or gets the endpoints default
4539 * values. If the assoc_id field is non-zero, then the set or get
4540 * effects the specified association for the one to many model (the
4541 * assoc_id field is ignored by the one to one model). Note that if
4542 * sack_delay or sack_freq are 0 when setting this option, then the
4543 * current values will remain unchanged.
4545 * struct sctp_sack_info {
4546 * sctp_assoc_t sack_assoc_id;
4547 * uint32_t sack_delay;
4548 * uint32_t sack_freq;
4551 * sack_assoc_id - This parameter, indicates which association the user
4552 * is performing an action upon. Note that if this field's value is
4553 * zero then the endpoints default value is changed (effecting future
4554 * associations only).
4556 * sack_delay - This parameter contains the number of milliseconds that
4557 * the user is requesting the delayed ACK timer be set to. Note that
4558 * this value is defined in the standard to be between 200 and 500
4561 * sack_freq - This parameter contains the number of packets that must
4562 * be received before a sack is sent without waiting for the delay
4563 * timer to expire. The default value for this is 2, setting this
4564 * value to 1 will disable the delayed sack algorithm.
4566 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4567 char __user *optval,
4570 struct sctp_sack_info params;
4571 struct sctp_association *asoc = NULL;
4572 struct sctp_sock *sp = sctp_sk(sk);
4574 if (len >= sizeof(struct sctp_sack_info)) {
4575 len = sizeof(struct sctp_sack_info);
4577 if (copy_from_user(¶ms, optval, len))
4579 } else if (len == sizeof(struct sctp_assoc_value)) {
4580 pr_warn_ratelimited(DEPRECATED
4582 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
4583 "Use struct sctp_sack_info instead\n",
4584 current->comm, task_pid_nr(current));
4585 if (copy_from_user(¶ms, optval, len))
4590 /* Get association, if sack_assoc_id != 0 and the socket is a one
4591 * to many style socket, and an association was not found, then
4592 * the id was invalid.
4594 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4595 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4599 /* Fetch association values. */
4600 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4601 params.sack_delay = jiffies_to_msecs(
4603 params.sack_freq = asoc->sackfreq;
4606 params.sack_delay = 0;
4607 params.sack_freq = 1;
4610 /* Fetch socket values. */
4611 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4612 params.sack_delay = sp->sackdelay;
4613 params.sack_freq = sp->sackfreq;
4615 params.sack_delay = 0;
4616 params.sack_freq = 1;
4620 if (copy_to_user(optval, ¶ms, len))
4623 if (put_user(len, optlen))
4629 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4631 * Applications can specify protocol parameters for the default association
4632 * initialization. The option name argument to setsockopt() and getsockopt()
4635 * Setting initialization parameters is effective only on an unconnected
4636 * socket (for UDP-style sockets only future associations are effected
4637 * by the change). With TCP-style sockets, this option is inherited by
4638 * sockets derived from a listener socket.
4640 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4642 if (len < sizeof(struct sctp_initmsg))
4644 len = sizeof(struct sctp_initmsg);
4645 if (put_user(len, optlen))
4647 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4653 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4654 char __user *optval, int __user *optlen)
4656 struct sctp_association *asoc;
4658 struct sctp_getaddrs getaddrs;
4659 struct sctp_transport *from;
4661 union sctp_addr temp;
4662 struct sctp_sock *sp = sctp_sk(sk);
4667 if (len < sizeof(struct sctp_getaddrs))
4670 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4673 /* For UDP-style sockets, id specifies the association to query. */
4674 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4678 to = optval + offsetof(struct sctp_getaddrs, addrs);
4679 space_left = len - offsetof(struct sctp_getaddrs, addrs);
4681 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4683 memcpy(&temp, &from->ipaddr, sizeof(temp));
4684 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4685 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4686 if (space_left < addrlen)
4688 if (copy_to_user(to, &temp, addrlen))
4692 space_left -= addrlen;
4695 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4697 bytes_copied = ((char __user *)to) - optval;
4698 if (put_user(bytes_copied, optlen))
4704 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4705 size_t space_left, int *bytes_copied)
4707 struct sctp_sockaddr_entry *addr;
4708 union sctp_addr temp;
4711 struct net *net = sock_net(sk);
4714 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4718 if ((PF_INET == sk->sk_family) &&
4719 (AF_INET6 == addr->a.sa.sa_family))
4721 if ((PF_INET6 == sk->sk_family) &&
4722 inet_v6_ipv6only(sk) &&
4723 (AF_INET == addr->a.sa.sa_family))
4725 memcpy(&temp, &addr->a, sizeof(temp));
4726 if (!temp.v4.sin_port)
4727 temp.v4.sin_port = htons(port);
4729 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4731 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4732 if (space_left < addrlen) {
4736 memcpy(to, &temp, addrlen);
4740 space_left -= addrlen;
4741 *bytes_copied += addrlen;
4749 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4750 char __user *optval, int __user *optlen)
4752 struct sctp_bind_addr *bp;
4753 struct sctp_association *asoc;
4755 struct sctp_getaddrs getaddrs;
4756 struct sctp_sockaddr_entry *addr;
4758 union sctp_addr temp;
4759 struct sctp_sock *sp = sctp_sk(sk);
4763 int bytes_copied = 0;
4767 if (len < sizeof(struct sctp_getaddrs))
4770 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4774 * For UDP-style sockets, id specifies the association to query.
4775 * If the id field is set to the value '0' then the locally bound
4776 * addresses are returned without regard to any particular
4779 if (0 == getaddrs.assoc_id) {
4780 bp = &sctp_sk(sk)->ep->base.bind_addr;
4782 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4785 bp = &asoc->base.bind_addr;
4788 to = optval + offsetof(struct sctp_getaddrs, addrs);
4789 space_left = len - offsetof(struct sctp_getaddrs, addrs);
4791 addrs = kmalloc(space_left, GFP_KERNEL);
4795 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4796 * addresses from the global local address list.
4798 if (sctp_list_single_entry(&bp->address_list)) {
4799 addr = list_entry(bp->address_list.next,
4800 struct sctp_sockaddr_entry, list);
4801 if (sctp_is_any(sk, &addr->a)) {
4802 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4803 space_left, &bytes_copied);
4813 /* Protection on the bound address list is not needed since
4814 * in the socket option context we hold a socket lock and
4815 * thus the bound address list can't change.
4817 list_for_each_entry(addr, &bp->address_list, list) {
4818 memcpy(&temp, &addr->a, sizeof(temp));
4819 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4820 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4821 if (space_left < addrlen) {
4822 err = -ENOMEM; /*fixme: right error?*/
4825 memcpy(buf, &temp, addrlen);
4827 bytes_copied += addrlen;
4829 space_left -= addrlen;
4833 if (copy_to_user(to, addrs, bytes_copied)) {
4837 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4841 if (put_user(bytes_copied, optlen))
4848 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4850 * Requests that the local SCTP stack use the enclosed peer address as
4851 * the association primary. The enclosed address must be one of the
4852 * association peer's addresses.
4854 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4855 char __user *optval, int __user *optlen)
4857 struct sctp_prim prim;
4858 struct sctp_association *asoc;
4859 struct sctp_sock *sp = sctp_sk(sk);
4861 if (len < sizeof(struct sctp_prim))
4864 len = sizeof(struct sctp_prim);
4866 if (copy_from_user(&prim, optval, len))
4869 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4873 if (!asoc->peer.primary_path)
4876 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4877 asoc->peer.primary_path->af_specific->sockaddr_len);
4879 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4880 (union sctp_addr *)&prim.ssp_addr);
4882 if (put_user(len, optlen))
4884 if (copy_to_user(optval, &prim, len))
4891 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4893 * Requests that the local endpoint set the specified Adaptation Layer
4894 * Indication parameter for all future INIT and INIT-ACK exchanges.
4896 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4897 char __user *optval, int __user *optlen)
4899 struct sctp_setadaptation adaptation;
4901 if (len < sizeof(struct sctp_setadaptation))
4904 len = sizeof(struct sctp_setadaptation);
4906 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4908 if (put_user(len, optlen))
4910 if (copy_to_user(optval, &adaptation, len))
4918 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4920 * Applications that wish to use the sendto() system call may wish to
4921 * specify a default set of parameters that would normally be supplied
4922 * through the inclusion of ancillary data. This socket option allows
4923 * such an application to set the default sctp_sndrcvinfo structure.
4926 * The application that wishes to use this socket option simply passes
4927 * in to this call the sctp_sndrcvinfo structure defined in Section
4928 * 5.2.2) The input parameters accepted by this call include
4929 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4930 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4931 * to this call if the caller is using the UDP model.
4933 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4935 static int sctp_getsockopt_default_send_param(struct sock *sk,
4936 int len, char __user *optval,
4939 struct sctp_sndrcvinfo info;
4940 struct sctp_association *asoc;
4941 struct sctp_sock *sp = sctp_sk(sk);
4943 if (len < sizeof(struct sctp_sndrcvinfo))
4946 len = sizeof(struct sctp_sndrcvinfo);
4948 if (copy_from_user(&info, optval, len))
4951 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4952 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4956 info.sinfo_stream = asoc->default_stream;
4957 info.sinfo_flags = asoc->default_flags;
4958 info.sinfo_ppid = asoc->default_ppid;
4959 info.sinfo_context = asoc->default_context;
4960 info.sinfo_timetolive = asoc->default_timetolive;
4962 info.sinfo_stream = sp->default_stream;
4963 info.sinfo_flags = sp->default_flags;
4964 info.sinfo_ppid = sp->default_ppid;
4965 info.sinfo_context = sp->default_context;
4966 info.sinfo_timetolive = sp->default_timetolive;
4969 if (put_user(len, optlen))
4971 if (copy_to_user(optval, &info, len))
4979 * 7.1.5 SCTP_NODELAY
4981 * Turn on/off any Nagle-like algorithm. This means that packets are
4982 * generally sent as soon as possible and no unnecessary delays are
4983 * introduced, at the cost of more packets in the network. Expects an
4984 * integer boolean flag.
4987 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4988 char __user *optval, int __user *optlen)
4992 if (len < sizeof(int))
4996 val = (sctp_sk(sk)->nodelay == 1);
4997 if (put_user(len, optlen))
4999 if (copy_to_user(optval, &val, len))
5006 * 7.1.1 SCTP_RTOINFO
5008 * The protocol parameters used to initialize and bound retransmission
5009 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5010 * and modify these parameters.
5011 * All parameters are time values, in milliseconds. A value of 0, when
5012 * modifying the parameters, indicates that the current value should not
5016 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5017 char __user *optval,
5018 int __user *optlen) {
5019 struct sctp_rtoinfo rtoinfo;
5020 struct sctp_association *asoc;
5022 if (len < sizeof (struct sctp_rtoinfo))
5025 len = sizeof(struct sctp_rtoinfo);
5027 if (copy_from_user(&rtoinfo, optval, len))
5030 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5032 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5035 /* Values corresponding to the specific association. */
5037 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5038 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5039 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5041 /* Values corresponding to the endpoint. */
5042 struct sctp_sock *sp = sctp_sk(sk);
5044 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5045 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5046 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5049 if (put_user(len, optlen))
5052 if (copy_to_user(optval, &rtoinfo, len))
5060 * 7.1.2 SCTP_ASSOCINFO
5062 * This option is used to tune the maximum retransmission attempts
5063 * of the association.
5064 * Returns an error if the new association retransmission value is
5065 * greater than the sum of the retransmission value of the peer.
5066 * See [SCTP] for more information.
5069 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5070 char __user *optval,
5074 struct sctp_assocparams assocparams;
5075 struct sctp_association *asoc;
5076 struct list_head *pos;
5079 if (len < sizeof (struct sctp_assocparams))
5082 len = sizeof(struct sctp_assocparams);
5084 if (copy_from_user(&assocparams, optval, len))
5087 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5089 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5092 /* Values correspoinding to the specific association */
5094 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5095 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5096 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5097 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5099 list_for_each(pos, &asoc->peer.transport_addr_list) {
5103 assocparams.sasoc_number_peer_destinations = cnt;
5105 /* Values corresponding to the endpoint */
5106 struct sctp_sock *sp = sctp_sk(sk);
5108 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5109 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5110 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5111 assocparams.sasoc_cookie_life =
5112 sp->assocparams.sasoc_cookie_life;
5113 assocparams.sasoc_number_peer_destinations =
5115 sasoc_number_peer_destinations;
5118 if (put_user(len, optlen))
5121 if (copy_to_user(optval, &assocparams, len))
5128 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5130 * This socket option is a boolean flag which turns on or off mapped V4
5131 * addresses. If this option is turned on and the socket is type
5132 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5133 * If this option is turned off, then no mapping will be done of V4
5134 * addresses and a user will receive both PF_INET6 and PF_INET type
5135 * addresses on the socket.
5137 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5138 char __user *optval, int __user *optlen)
5141 struct sctp_sock *sp = sctp_sk(sk);
5143 if (len < sizeof(int))
5148 if (put_user(len, optlen))
5150 if (copy_to_user(optval, &val, len))
5157 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5158 * (chapter and verse is quoted at sctp_setsockopt_context())
5160 static int sctp_getsockopt_context(struct sock *sk, int len,
5161 char __user *optval, int __user *optlen)
5163 struct sctp_assoc_value params;
5164 struct sctp_sock *sp;
5165 struct sctp_association *asoc;
5167 if (len < sizeof(struct sctp_assoc_value))
5170 len = sizeof(struct sctp_assoc_value);
5172 if (copy_from_user(¶ms, optval, len))
5177 if (params.assoc_id != 0) {
5178 asoc = sctp_id2assoc(sk, params.assoc_id);
5181 params.assoc_value = asoc->default_rcv_context;
5183 params.assoc_value = sp->default_rcv_context;
5186 if (put_user(len, optlen))
5188 if (copy_to_user(optval, ¶ms, len))
5195 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5196 * This option will get or set the maximum size to put in any outgoing
5197 * SCTP DATA chunk. If a message is larger than this size it will be
5198 * fragmented by SCTP into the specified size. Note that the underlying
5199 * SCTP implementation may fragment into smaller sized chunks when the
5200 * PMTU of the underlying association is smaller than the value set by
5201 * the user. The default value for this option is '0' which indicates
5202 * the user is NOT limiting fragmentation and only the PMTU will effect
5203 * SCTP's choice of DATA chunk size. Note also that values set larger
5204 * than the maximum size of an IP datagram will effectively let SCTP
5205 * control fragmentation (i.e. the same as setting this option to 0).
5207 * The following structure is used to access and modify this parameter:
5209 * struct sctp_assoc_value {
5210 * sctp_assoc_t assoc_id;
5211 * uint32_t assoc_value;
5214 * assoc_id: This parameter is ignored for one-to-one style sockets.
5215 * For one-to-many style sockets this parameter indicates which
5216 * association the user is performing an action upon. Note that if
5217 * this field's value is zero then the endpoints default value is
5218 * changed (effecting future associations only).
5219 * assoc_value: This parameter specifies the maximum size in bytes.
5221 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5222 char __user *optval, int __user *optlen)
5224 struct sctp_assoc_value params;
5225 struct sctp_association *asoc;
5227 if (len == sizeof(int)) {
5228 pr_warn_ratelimited(DEPRECATED
5230 "Use of int in maxseg socket option.\n"
5231 "Use struct sctp_assoc_value instead\n",
5232 current->comm, task_pid_nr(current));
5233 params.assoc_id = 0;
5234 } else if (len >= sizeof(struct sctp_assoc_value)) {
5235 len = sizeof(struct sctp_assoc_value);
5236 if (copy_from_user(¶ms, optval, sizeof(params)))
5241 asoc = sctp_id2assoc(sk, params.assoc_id);
5242 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5246 params.assoc_value = asoc->frag_point;
5248 params.assoc_value = sctp_sk(sk)->user_frag;
5250 if (put_user(len, optlen))
5252 if (len == sizeof(int)) {
5253 if (copy_to_user(optval, ¶ms.assoc_value, len))
5256 if (copy_to_user(optval, ¶ms, len))
5264 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5265 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5267 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5268 char __user *optval, int __user *optlen)
5272 if (len < sizeof(int))
5277 val = sctp_sk(sk)->frag_interleave;
5278 if (put_user(len, optlen))
5280 if (copy_to_user(optval, &val, len))
5287 * 7.1.25. Set or Get the sctp partial delivery point
5288 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5290 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5291 char __user *optval,
5296 if (len < sizeof(u32))
5301 val = sctp_sk(sk)->pd_point;
5302 if (put_user(len, optlen))
5304 if (copy_to_user(optval, &val, len))
5311 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5312 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5314 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5315 char __user *optval,
5318 struct sctp_assoc_value params;
5319 struct sctp_sock *sp;
5320 struct sctp_association *asoc;
5322 if (len == sizeof(int)) {
5323 pr_warn_ratelimited(DEPRECATED
5325 "Use of int in max_burst socket option.\n"
5326 "Use struct sctp_assoc_value instead\n",
5327 current->comm, task_pid_nr(current));
5328 params.assoc_id = 0;
5329 } else if (len >= sizeof(struct sctp_assoc_value)) {
5330 len = sizeof(struct sctp_assoc_value);
5331 if (copy_from_user(¶ms, optval, len))
5338 if (params.assoc_id != 0) {
5339 asoc = sctp_id2assoc(sk, params.assoc_id);
5342 params.assoc_value = asoc->max_burst;
5344 params.assoc_value = sp->max_burst;
5346 if (len == sizeof(int)) {
5347 if (copy_to_user(optval, ¶ms.assoc_value, len))
5350 if (copy_to_user(optval, ¶ms, len))
5358 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5359 char __user *optval, int __user *optlen)
5361 struct net *net = sock_net(sk);
5362 struct sctp_hmacalgo __user *p = (void __user *)optval;
5363 struct sctp_hmac_algo_param *hmacs;
5367 if (!net->sctp.auth_enable)
5370 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5371 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5373 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5376 len = sizeof(struct sctp_hmacalgo) + data_len;
5377 num_idents = data_len / sizeof(u16);
5379 if (put_user(len, optlen))
5381 if (put_user(num_idents, &p->shmac_num_idents))
5383 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5388 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5389 char __user *optval, int __user *optlen)
5391 struct net *net = sock_net(sk);
5392 struct sctp_authkeyid val;
5393 struct sctp_association *asoc;
5395 if (!net->sctp.auth_enable)
5398 if (len < sizeof(struct sctp_authkeyid))
5400 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5403 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5404 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5408 val.scact_keynumber = asoc->active_key_id;
5410 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5412 len = sizeof(struct sctp_authkeyid);
5413 if (put_user(len, optlen))
5415 if (copy_to_user(optval, &val, len))
5421 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5422 char __user *optval, int __user *optlen)
5424 struct net *net = sock_net(sk);
5425 struct sctp_authchunks __user *p = (void __user *)optval;
5426 struct sctp_authchunks val;
5427 struct sctp_association *asoc;
5428 struct sctp_chunks_param *ch;
5432 if (!net->sctp.auth_enable)
5435 if (len < sizeof(struct sctp_authchunks))
5438 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5441 to = p->gauth_chunks;
5442 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5446 ch = asoc->peer.peer_chunks;
5450 /* See if the user provided enough room for all the data */
5451 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5452 if (len < num_chunks)
5455 if (copy_to_user(to, ch->chunks, num_chunks))
5458 len = sizeof(struct sctp_authchunks) + num_chunks;
5459 if (put_user(len, optlen))
5461 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5466 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5467 char __user *optval, int __user *optlen)
5469 struct net *net = sock_net(sk);
5470 struct sctp_authchunks __user *p = (void __user *)optval;
5471 struct sctp_authchunks val;
5472 struct sctp_association *asoc;
5473 struct sctp_chunks_param *ch;
5477 if (!net->sctp.auth_enable)
5480 if (len < sizeof(struct sctp_authchunks))
5483 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5486 to = p->gauth_chunks;
5487 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5488 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5492 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
5494 ch = sctp_sk(sk)->ep->auth_chunk_list;
5499 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5500 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5503 if (copy_to_user(to, ch->chunks, num_chunks))
5506 len = sizeof(struct sctp_authchunks) + num_chunks;
5507 if (put_user(len, optlen))
5509 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5516 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5517 * This option gets the current number of associations that are attached
5518 * to a one-to-many style socket. The option value is an uint32_t.
5520 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5521 char __user *optval, int __user *optlen)
5523 struct sctp_sock *sp = sctp_sk(sk);
5524 struct sctp_association *asoc;
5527 if (sctp_style(sk, TCP))
5530 if (len < sizeof(u32))
5535 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5539 if (put_user(len, optlen))
5541 if (copy_to_user(optval, &val, len))
5548 * 8.1.23 SCTP_AUTO_ASCONF
5549 * See the corresponding setsockopt entry as description
5551 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5552 char __user *optval, int __user *optlen)
5556 if (len < sizeof(int))
5560 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5562 if (put_user(len, optlen))
5564 if (copy_to_user(optval, &val, len))
5570 * 8.2.6. Get the Current Identifiers of Associations
5571 * (SCTP_GET_ASSOC_ID_LIST)
5573 * This option gets the current list of SCTP association identifiers of
5574 * the SCTP associations handled by a one-to-many style socket.
5576 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5577 char __user *optval, int __user *optlen)
5579 struct sctp_sock *sp = sctp_sk(sk);
5580 struct sctp_association *asoc;
5581 struct sctp_assoc_ids *ids;
5584 if (sctp_style(sk, TCP))
5587 if (len < sizeof(struct sctp_assoc_ids))
5590 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5594 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5597 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5599 ids = kmalloc(len, GFP_KERNEL);
5603 ids->gaids_number_of_ids = num;
5605 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5606 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5609 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5619 * SCTP_PEER_ADDR_THLDS
5621 * This option allows us to fetch the partially failed threshold for one or all
5622 * transports in an association. See Section 6.1 of:
5623 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5625 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5626 char __user *optval,
5630 struct sctp_paddrthlds val;
5631 struct sctp_transport *trans;
5632 struct sctp_association *asoc;
5634 if (len < sizeof(struct sctp_paddrthlds))
5636 len = sizeof(struct sctp_paddrthlds);
5637 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5640 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5641 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5645 val.spt_pathpfthld = asoc->pf_retrans;
5646 val.spt_pathmaxrxt = asoc->pathmaxrxt;
5648 trans = sctp_addr_id2transport(sk, &val.spt_address,
5653 val.spt_pathmaxrxt = trans->pathmaxrxt;
5654 val.spt_pathpfthld = trans->pf_retrans;
5657 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5664 * SCTP_GET_ASSOC_STATS
5666 * This option retrieves local per endpoint statistics. It is modeled
5667 * after OpenSolaris' implementation
5669 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
5670 char __user *optval,
5673 struct sctp_assoc_stats sas;
5674 struct sctp_association *asoc = NULL;
5676 /* User must provide at least the assoc id */
5677 if (len < sizeof(sctp_assoc_t))
5680 /* Allow the struct to grow and fill in as much as possible */
5681 len = min_t(size_t, len, sizeof(sas));
5683 if (copy_from_user(&sas, optval, len))
5686 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
5690 sas.sas_rtxchunks = asoc->stats.rtxchunks;
5691 sas.sas_gapcnt = asoc->stats.gapcnt;
5692 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
5693 sas.sas_osacks = asoc->stats.osacks;
5694 sas.sas_isacks = asoc->stats.isacks;
5695 sas.sas_octrlchunks = asoc->stats.octrlchunks;
5696 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
5697 sas.sas_oodchunks = asoc->stats.oodchunks;
5698 sas.sas_iodchunks = asoc->stats.iodchunks;
5699 sas.sas_ouodchunks = asoc->stats.ouodchunks;
5700 sas.sas_iuodchunks = asoc->stats.iuodchunks;
5701 sas.sas_idupchunks = asoc->stats.idupchunks;
5702 sas.sas_opackets = asoc->stats.opackets;
5703 sas.sas_ipackets = asoc->stats.ipackets;
5705 /* New high max rto observed, will return 0 if not a single
5706 * RTO update took place. obs_rto_ipaddr will be bogus
5709 sas.sas_maxrto = asoc->stats.max_obs_rto;
5710 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
5711 sizeof(struct sockaddr_storage));
5713 /* Mark beginning of a new observation period */
5714 asoc->stats.max_obs_rto = asoc->rto_min;
5716 if (put_user(len, optlen))
5719 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
5721 if (copy_to_user(optval, &sas, len))
5727 static int sctp_getsockopt(struct sock *sk, int level, int optname,
5728 char __user *optval, int __user *optlen)
5733 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
5735 /* I can hardly begin to describe how wrong this is. This is
5736 * so broken as to be worse than useless. The API draft
5737 * REALLY is NOT helpful here... I am not convinced that the
5738 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5739 * are at all well-founded.
5741 if (level != SOL_SCTP) {
5742 struct sctp_af *af = sctp_sk(sk)->pf->af;
5744 retval = af->getsockopt(sk, level, optname, optval, optlen);
5748 if (get_user(len, optlen))
5755 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5757 case SCTP_DISABLE_FRAGMENTS:
5758 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5762 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5764 case SCTP_AUTOCLOSE:
5765 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5767 case SCTP_SOCKOPT_PEELOFF:
5768 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5770 case SCTP_PEER_ADDR_PARAMS:
5771 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5774 case SCTP_DELAYED_SACK:
5775 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5779 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5781 case SCTP_GET_PEER_ADDRS:
5782 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5785 case SCTP_GET_LOCAL_ADDRS:
5786 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5789 case SCTP_SOCKOPT_CONNECTX3:
5790 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5792 case SCTP_DEFAULT_SEND_PARAM:
5793 retval = sctp_getsockopt_default_send_param(sk, len,
5796 case SCTP_PRIMARY_ADDR:
5797 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5800 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5803 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5805 case SCTP_ASSOCINFO:
5806 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5808 case SCTP_I_WANT_MAPPED_V4_ADDR:
5809 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5812 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5814 case SCTP_GET_PEER_ADDR_INFO:
5815 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5818 case SCTP_ADAPTATION_LAYER:
5819 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5823 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5825 case SCTP_FRAGMENT_INTERLEAVE:
5826 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5829 case SCTP_PARTIAL_DELIVERY_POINT:
5830 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5833 case SCTP_MAX_BURST:
5834 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5837 case SCTP_AUTH_CHUNK:
5838 case SCTP_AUTH_DELETE_KEY:
5839 retval = -EOPNOTSUPP;
5841 case SCTP_HMAC_IDENT:
5842 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5844 case SCTP_AUTH_ACTIVE_KEY:
5845 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5847 case SCTP_PEER_AUTH_CHUNKS:
5848 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5851 case SCTP_LOCAL_AUTH_CHUNKS:
5852 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5855 case SCTP_GET_ASSOC_NUMBER:
5856 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5858 case SCTP_GET_ASSOC_ID_LIST:
5859 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5861 case SCTP_AUTO_ASCONF:
5862 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5864 case SCTP_PEER_ADDR_THLDS:
5865 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
5867 case SCTP_GET_ASSOC_STATS:
5868 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
5871 retval = -ENOPROTOOPT;
5879 static void sctp_hash(struct sock *sk)
5884 static void sctp_unhash(struct sock *sk)
5889 /* Check if port is acceptable. Possibly find first available port.
5891 * The port hash table (contained in the 'global' SCTP protocol storage
5892 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5893 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5894 * list (the list number is the port number hashed out, so as you
5895 * would expect from a hash function, all the ports in a given list have
5896 * such a number that hashes out to the same list number; you were
5897 * expecting that, right?); so each list has a set of ports, with a
5898 * link to the socket (struct sock) that uses it, the port number and
5899 * a fastreuse flag (FIXME: NPI ipg).
5901 static struct sctp_bind_bucket *sctp_bucket_create(
5902 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
5904 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5906 struct sctp_bind_hashbucket *head; /* hash list */
5907 struct sctp_bind_bucket *pp;
5908 unsigned short snum;
5911 snum = ntohs(addr->v4.sin_port);
5913 pr_debug("%s: begins, snum:%d\n", __func__, snum);
5918 /* Search for an available port. */
5919 int low, high, remaining, index;
5922 inet_get_local_port_range(sock_net(sk), &low, &high);
5923 remaining = (high - low) + 1;
5924 rover = prandom_u32() % remaining + low;
5928 if ((rover < low) || (rover > high))
5930 if (inet_is_reserved_local_port(rover))
5932 index = sctp_phashfn(sock_net(sk), rover);
5933 head = &sctp_port_hashtable[index];
5934 spin_lock(&head->lock);
5935 sctp_for_each_hentry(pp, &head->chain)
5936 if ((pp->port == rover) &&
5937 net_eq(sock_net(sk), pp->net))
5941 spin_unlock(&head->lock);
5942 } while (--remaining > 0);
5944 /* Exhausted local port range during search? */
5949 /* OK, here is the one we will use. HEAD (the port
5950 * hash table list entry) is non-NULL and we hold it's
5955 /* We are given an specific port number; we verify
5956 * that it is not being used. If it is used, we will
5957 * exahust the search in the hash list corresponding
5958 * to the port number (snum) - we detect that with the
5959 * port iterator, pp being NULL.
5961 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
5962 spin_lock(&head->lock);
5963 sctp_for_each_hentry(pp, &head->chain) {
5964 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
5971 if (!hlist_empty(&pp->owner)) {
5972 /* We had a port hash table hit - there is an
5973 * available port (pp != NULL) and it is being
5974 * used by other socket (pp->owner not empty); that other
5975 * socket is going to be sk2.
5977 int reuse = sk->sk_reuse;
5980 pr_debug("%s: found a possible match\n", __func__);
5982 if (pp->fastreuse && sk->sk_reuse &&
5983 sk->sk_state != SCTP_SS_LISTENING)
5986 /* Run through the list of sockets bound to the port
5987 * (pp->port) [via the pointers bind_next and
5988 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5989 * we get the endpoint they describe and run through
5990 * the endpoint's list of IP (v4 or v6) addresses,
5991 * comparing each of the addresses with the address of
5992 * the socket sk. If we find a match, then that means
5993 * that this port/socket (sk) combination are already
5996 sk_for_each_bound(sk2, &pp->owner) {
5997 struct sctp_endpoint *ep2;
5998 ep2 = sctp_sk(sk2)->ep;
6001 (reuse && sk2->sk_reuse &&
6002 sk2->sk_state != SCTP_SS_LISTENING))
6005 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
6006 sctp_sk(sk2), sctp_sk(sk))) {
6012 pr_debug("%s: found a match\n", __func__);
6015 /* If there was a hash table miss, create a new port. */
6017 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6020 /* In either case (hit or miss), make sure fastreuse is 1 only
6021 * if sk->sk_reuse is too (that is, if the caller requested
6022 * SO_REUSEADDR on this socket -sk-).
6024 if (hlist_empty(&pp->owner)) {
6025 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6029 } else if (pp->fastreuse &&
6030 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6033 /* We are set, so fill up all the data in the hash table
6034 * entry, tie the socket list information with the rest of the
6035 * sockets FIXME: Blurry, NPI (ipg).
6038 if (!sctp_sk(sk)->bind_hash) {
6039 inet_sk(sk)->inet_num = snum;
6040 sk_add_bind_node(sk, &pp->owner);
6041 sctp_sk(sk)->bind_hash = pp;
6046 spin_unlock(&head->lock);
6053 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6054 * port is requested.
6056 static int sctp_get_port(struct sock *sk, unsigned short snum)
6058 union sctp_addr addr;
6059 struct sctp_af *af = sctp_sk(sk)->pf->af;
6061 /* Set up a dummy address struct from the sk. */
6062 af->from_sk(&addr, sk);
6063 addr.v4.sin_port = htons(snum);
6065 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6066 return !!sctp_get_port_local(sk, &addr);
6070 * Move a socket to LISTENING state.
6072 static int sctp_listen_start(struct sock *sk, int backlog)
6074 struct sctp_sock *sp = sctp_sk(sk);
6075 struct sctp_endpoint *ep = sp->ep;
6076 struct crypto_hash *tfm = NULL;
6079 /* Allocate HMAC for generating cookie. */
6080 if (!sp->hmac && sp->sctp_hmac_alg) {
6081 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6082 tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
6084 net_info_ratelimited("failed to load transform for %s: %ld\n",
6085 sp->sctp_hmac_alg, PTR_ERR(tfm));
6088 sctp_sk(sk)->hmac = tfm;
6092 * If a bind() or sctp_bindx() is not called prior to a listen()
6093 * call that allows new associations to be accepted, the system
6094 * picks an ephemeral port and will choose an address set equivalent
6095 * to binding with a wildcard address.
6097 * This is not currently spelled out in the SCTP sockets
6098 * extensions draft, but follows the practice as seen in TCP
6102 sk->sk_state = SCTP_SS_LISTENING;
6103 if (!ep->base.bind_addr.port) {
6104 if (sctp_autobind(sk))
6107 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6108 sk->sk_state = SCTP_SS_CLOSED;
6113 sk->sk_max_ack_backlog = backlog;
6114 sctp_hash_endpoint(ep);
6119 * 4.1.3 / 5.1.3 listen()
6121 * By default, new associations are not accepted for UDP style sockets.
6122 * An application uses listen() to mark a socket as being able to
6123 * accept new associations.
6125 * On TCP style sockets, applications use listen() to ready the SCTP
6126 * endpoint for accepting inbound associations.
6128 * On both types of endpoints a backlog of '0' disables listening.
6130 * Move a socket to LISTENING state.
6132 int sctp_inet_listen(struct socket *sock, int backlog)
6134 struct sock *sk = sock->sk;
6135 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6138 if (unlikely(backlog < 0))
6143 /* Peeled-off sockets are not allowed to listen(). */
6144 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6147 if (sock->state != SS_UNCONNECTED)
6150 /* If backlog is zero, disable listening. */
6152 if (sctp_sstate(sk, CLOSED))
6156 sctp_unhash_endpoint(ep);
6157 sk->sk_state = SCTP_SS_CLOSED;
6159 sctp_sk(sk)->bind_hash->fastreuse = 1;
6163 /* If we are already listening, just update the backlog */
6164 if (sctp_sstate(sk, LISTENING))
6165 sk->sk_max_ack_backlog = backlog;
6167 err = sctp_listen_start(sk, backlog);
6179 * This function is done by modeling the current datagram_poll() and the
6180 * tcp_poll(). Note that, based on these implementations, we don't
6181 * lock the socket in this function, even though it seems that,
6182 * ideally, locking or some other mechanisms can be used to ensure
6183 * the integrity of the counters (sndbuf and wmem_alloc) used
6184 * in this place. We assume that we don't need locks either until proven
6187 * Another thing to note is that we include the Async I/O support
6188 * here, again, by modeling the current TCP/UDP code. We don't have
6189 * a good way to test with it yet.
6191 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6193 struct sock *sk = sock->sk;
6194 struct sctp_sock *sp = sctp_sk(sk);
6197 poll_wait(file, sk_sleep(sk), wait);
6199 /* A TCP-style listening socket becomes readable when the accept queue
6202 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6203 return (!list_empty(&sp->ep->asocs)) ?
6204 (POLLIN | POLLRDNORM) : 0;
6208 /* Is there any exceptional events? */
6209 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6211 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6212 if (sk->sk_shutdown & RCV_SHUTDOWN)
6213 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6214 if (sk->sk_shutdown == SHUTDOWN_MASK)
6217 /* Is it readable? Reconsider this code with TCP-style support. */
6218 if (!skb_queue_empty(&sk->sk_receive_queue))
6219 mask |= POLLIN | POLLRDNORM;
6221 /* The association is either gone or not ready. */
6222 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6225 /* Is it writable? */
6226 if (sctp_writeable(sk)) {
6227 mask |= POLLOUT | POLLWRNORM;
6229 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6231 * Since the socket is not locked, the buffer
6232 * might be made available after the writeable check and
6233 * before the bit is set. This could cause a lost I/O
6234 * signal. tcp_poll() has a race breaker for this race
6235 * condition. Based on their implementation, we put
6236 * in the following code to cover it as well.
6238 if (sctp_writeable(sk))
6239 mask |= POLLOUT | POLLWRNORM;
6244 /********************************************************************
6245 * 2nd Level Abstractions
6246 ********************************************************************/
6248 static struct sctp_bind_bucket *sctp_bucket_create(
6249 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6251 struct sctp_bind_bucket *pp;
6253 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6255 SCTP_DBG_OBJCNT_INC(bind_bucket);
6258 INIT_HLIST_HEAD(&pp->owner);
6260 hlist_add_head(&pp->node, &head->chain);
6265 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6266 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6268 if (pp && hlist_empty(&pp->owner)) {
6269 __hlist_del(&pp->node);
6270 kmem_cache_free(sctp_bucket_cachep, pp);
6271 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6275 /* Release this socket's reference to a local port. */
6276 static inline void __sctp_put_port(struct sock *sk)
6278 struct sctp_bind_hashbucket *head =
6279 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6280 inet_sk(sk)->inet_num)];
6281 struct sctp_bind_bucket *pp;
6283 spin_lock(&head->lock);
6284 pp = sctp_sk(sk)->bind_hash;
6285 __sk_del_bind_node(sk);
6286 sctp_sk(sk)->bind_hash = NULL;
6287 inet_sk(sk)->inet_num = 0;
6288 sctp_bucket_destroy(pp);
6289 spin_unlock(&head->lock);
6292 void sctp_put_port(struct sock *sk)
6295 __sctp_put_port(sk);
6300 * The system picks an ephemeral port and choose an address set equivalent
6301 * to binding with a wildcard address.
6302 * One of those addresses will be the primary address for the association.
6303 * This automatically enables the multihoming capability of SCTP.
6305 static int sctp_autobind(struct sock *sk)
6307 union sctp_addr autoaddr;
6311 /* Initialize a local sockaddr structure to INADDR_ANY. */
6312 af = sctp_sk(sk)->pf->af;
6314 port = htons(inet_sk(sk)->inet_num);
6315 af->inaddr_any(&autoaddr, port);
6317 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6320 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6323 * 4.2 The cmsghdr Structure *
6325 * When ancillary data is sent or received, any number of ancillary data
6326 * objects can be specified by the msg_control and msg_controllen members of
6327 * the msghdr structure, because each object is preceded by
6328 * a cmsghdr structure defining the object's length (the cmsg_len member).
6329 * Historically Berkeley-derived implementations have passed only one object
6330 * at a time, but this API allows multiple objects to be
6331 * passed in a single call to sendmsg() or recvmsg(). The following example
6332 * shows two ancillary data objects in a control buffer.
6334 * |<--------------------------- msg_controllen -------------------------->|
6337 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6339 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6342 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6344 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6347 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6348 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6350 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6352 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6359 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
6361 struct cmsghdr *cmsg;
6362 struct msghdr *my_msg = (struct msghdr *)msg;
6364 for (cmsg = CMSG_FIRSTHDR(msg);
6366 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6367 if (!CMSG_OK(my_msg, cmsg))
6370 /* Should we parse this header or ignore? */
6371 if (cmsg->cmsg_level != IPPROTO_SCTP)
6374 /* Strictly check lengths following example in SCM code. */
6375 switch (cmsg->cmsg_type) {
6377 /* SCTP Socket API Extension
6378 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6380 * This cmsghdr structure provides information for
6381 * initializing new SCTP associations with sendmsg().
6382 * The SCTP_INITMSG socket option uses this same data
6383 * structure. This structure is not used for
6386 * cmsg_level cmsg_type cmsg_data[]
6387 * ------------ ------------ ----------------------
6388 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6390 if (cmsg->cmsg_len !=
6391 CMSG_LEN(sizeof(struct sctp_initmsg)))
6393 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6397 /* SCTP Socket API Extension
6398 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6400 * This cmsghdr structure specifies SCTP options for
6401 * sendmsg() and describes SCTP header information
6402 * about a received message through recvmsg().
6404 * cmsg_level cmsg_type cmsg_data[]
6405 * ------------ ------------ ----------------------
6406 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6408 if (cmsg->cmsg_len !=
6409 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6413 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6415 /* Minimally, validate the sinfo_flags. */
6416 if (cmsgs->info->sinfo_flags &
6417 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6418 SCTP_ABORT | SCTP_EOF))
6430 * Wait for a packet..
6431 * Note: This function is the same function as in core/datagram.c
6432 * with a few modifications to make lksctp work.
6434 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
6439 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6441 /* Socket errors? */
6442 error = sock_error(sk);
6446 if (!skb_queue_empty(&sk->sk_receive_queue))
6449 /* Socket shut down? */
6450 if (sk->sk_shutdown & RCV_SHUTDOWN)
6453 /* Sequenced packets can come disconnected. If so we report the
6458 /* Is there a good reason to think that we may receive some data? */
6459 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6462 /* Handle signals. */
6463 if (signal_pending(current))
6466 /* Let another process have a go. Since we are going to sleep
6467 * anyway. Note: This may cause odd behaviors if the message
6468 * does not fit in the user's buffer, but this seems to be the
6469 * only way to honor MSG_DONTWAIT realistically.
6472 *timeo_p = schedule_timeout(*timeo_p);
6476 finish_wait(sk_sleep(sk), &wait);
6480 error = sock_intr_errno(*timeo_p);
6483 finish_wait(sk_sleep(sk), &wait);
6488 /* Receive a datagram.
6489 * Note: This is pretty much the same routine as in core/datagram.c
6490 * with a few changes to make lksctp work.
6492 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6493 int noblock, int *err)
6496 struct sk_buff *skb;
6499 timeo = sock_rcvtimeo(sk, noblock);
6501 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
6502 MAX_SCHEDULE_TIMEOUT);
6505 /* Again only user level code calls this function,
6506 * so nothing interrupt level
6507 * will suddenly eat the receive_queue.
6509 * Look at current nfs client by the way...
6510 * However, this function was correct in any case. 8)
6512 if (flags & MSG_PEEK) {
6513 spin_lock_bh(&sk->sk_receive_queue.lock);
6514 skb = skb_peek(&sk->sk_receive_queue);
6516 atomic_inc(&skb->users);
6517 spin_unlock_bh(&sk->sk_receive_queue.lock);
6519 skb = skb_dequeue(&sk->sk_receive_queue);
6525 /* Caller is allowed not to check sk->sk_err before calling. */
6526 error = sock_error(sk);
6530 if (sk->sk_shutdown & RCV_SHUTDOWN)
6533 /* User doesn't want to wait. */
6537 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6546 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6547 static void __sctp_write_space(struct sctp_association *asoc)
6549 struct sock *sk = asoc->base.sk;
6550 struct socket *sock = sk->sk_socket;
6552 if ((sctp_wspace(asoc) > 0) && sock) {
6553 if (waitqueue_active(&asoc->wait))
6554 wake_up_interruptible(&asoc->wait);
6556 if (sctp_writeable(sk)) {
6557 wait_queue_head_t *wq = sk_sleep(sk);
6559 if (wq && waitqueue_active(wq))
6560 wake_up_interruptible(wq);
6562 /* Note that we try to include the Async I/O support
6563 * here by modeling from the current TCP/UDP code.
6564 * We have not tested with it yet.
6566 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6567 sock_wake_async(sock,
6568 SOCK_WAKE_SPACE, POLL_OUT);
6573 /* Do accounting for the sndbuf space.
6574 * Decrement the used sndbuf space of the corresponding association by the
6575 * data size which was just transmitted(freed).
6577 static void sctp_wfree(struct sk_buff *skb)
6579 struct sctp_association *asoc;
6580 struct sctp_chunk *chunk;
6583 /* Get the saved chunk pointer. */
6584 chunk = *((struct sctp_chunk **)(skb->cb));
6587 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6588 sizeof(struct sk_buff) +
6589 sizeof(struct sctp_chunk);
6591 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6594 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6596 sk->sk_wmem_queued -= skb->truesize;
6597 sk_mem_uncharge(sk, skb->truesize);
6600 __sctp_write_space(asoc);
6602 sctp_association_put(asoc);
6605 /* Do accounting for the receive space on the socket.
6606 * Accounting for the association is done in ulpevent.c
6607 * We set this as a destructor for the cloned data skbs so that
6608 * accounting is done at the correct time.
6610 void sctp_sock_rfree(struct sk_buff *skb)
6612 struct sock *sk = skb->sk;
6613 struct sctp_ulpevent *event = sctp_skb2event(skb);
6615 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6618 * Mimic the behavior of sock_rfree
6620 sk_mem_uncharge(sk, event->rmem_len);
6624 /* Helper function to wait for space in the sndbuf. */
6625 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6628 struct sock *sk = asoc->base.sk;
6630 long current_timeo = *timeo_p;
6633 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
6636 /* Increment the association's refcnt. */
6637 sctp_association_hold(asoc);
6639 /* Wait on the association specific sndbuf space. */
6641 prepare_to_wait_exclusive(&asoc->wait, &wait,
6642 TASK_INTERRUPTIBLE);
6645 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6648 if (signal_pending(current))
6649 goto do_interrupted;
6650 if (msg_len <= sctp_wspace(asoc))
6653 /* Let another process have a go. Since we are going
6657 current_timeo = schedule_timeout(current_timeo);
6658 BUG_ON(sk != asoc->base.sk);
6661 *timeo_p = current_timeo;
6665 finish_wait(&asoc->wait, &wait);
6667 /* Release the association's refcnt. */
6668 sctp_association_put(asoc);
6677 err = sock_intr_errno(*timeo_p);
6685 void sctp_data_ready(struct sock *sk, int len)
6687 struct socket_wq *wq;
6690 wq = rcu_dereference(sk->sk_wq);
6691 if (wq_has_sleeper(wq))
6692 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6693 POLLRDNORM | POLLRDBAND);
6694 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6698 /* If socket sndbuf has changed, wake up all per association waiters. */
6699 void sctp_write_space(struct sock *sk)
6701 struct sctp_association *asoc;
6703 /* Wake up the tasks in each wait queue. */
6704 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6705 __sctp_write_space(asoc);
6709 /* Is there any sndbuf space available on the socket?
6711 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6712 * associations on the same socket. For a UDP-style socket with
6713 * multiple associations, it is possible for it to be "unwriteable"
6714 * prematurely. I assume that this is acceptable because
6715 * a premature "unwriteable" is better than an accidental "writeable" which
6716 * would cause an unwanted block under certain circumstances. For the 1-1
6717 * UDP-style sockets or TCP-style sockets, this code should work.
6720 static int sctp_writeable(struct sock *sk)
6724 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6730 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6731 * returns immediately with EINPROGRESS.
6733 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6735 struct sock *sk = asoc->base.sk;
6737 long current_timeo = *timeo_p;
6740 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
6742 /* Increment the association's refcnt. */
6743 sctp_association_hold(asoc);
6746 prepare_to_wait_exclusive(&asoc->wait, &wait,
6747 TASK_INTERRUPTIBLE);
6750 if (sk->sk_shutdown & RCV_SHUTDOWN)
6752 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6755 if (signal_pending(current))
6756 goto do_interrupted;
6758 if (sctp_state(asoc, ESTABLISHED))
6761 /* Let another process have a go. Since we are going
6765 current_timeo = schedule_timeout(current_timeo);
6768 *timeo_p = current_timeo;
6772 finish_wait(&asoc->wait, &wait);
6774 /* Release the association's refcnt. */
6775 sctp_association_put(asoc);
6780 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6783 err = -ECONNREFUSED;
6787 err = sock_intr_errno(*timeo_p);
6795 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6797 struct sctp_endpoint *ep;
6801 ep = sctp_sk(sk)->ep;
6805 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6806 TASK_INTERRUPTIBLE);
6808 if (list_empty(&ep->asocs)) {
6810 timeo = schedule_timeout(timeo);
6815 if (!sctp_sstate(sk, LISTENING))
6819 if (!list_empty(&ep->asocs))
6822 err = sock_intr_errno(timeo);
6823 if (signal_pending(current))
6831 finish_wait(sk_sleep(sk), &wait);
6836 static void sctp_wait_for_close(struct sock *sk, long timeout)
6841 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6842 if (list_empty(&sctp_sk(sk)->ep->asocs))
6845 timeout = schedule_timeout(timeout);
6847 } while (!signal_pending(current) && timeout);
6849 finish_wait(sk_sleep(sk), &wait);
6852 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6854 struct sk_buff *frag;
6859 /* Don't forget the fragments. */
6860 skb_walk_frags(skb, frag)
6861 sctp_skb_set_owner_r_frag(frag, sk);
6864 sctp_skb_set_owner_r(skb, sk);
6867 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6868 struct sctp_association *asoc)
6870 struct inet_sock *inet = inet_sk(sk);
6871 struct inet_sock *newinet;
6873 newsk->sk_type = sk->sk_type;
6874 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6875 newsk->sk_flags = sk->sk_flags;
6876 newsk->sk_no_check = sk->sk_no_check;
6877 newsk->sk_reuse = sk->sk_reuse;
6879 newsk->sk_shutdown = sk->sk_shutdown;
6880 newsk->sk_destruct = sctp_destruct_sock;
6881 newsk->sk_family = sk->sk_family;
6882 newsk->sk_protocol = IPPROTO_SCTP;
6883 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6884 newsk->sk_sndbuf = sk->sk_sndbuf;
6885 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6886 newsk->sk_lingertime = sk->sk_lingertime;
6887 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6888 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6890 newinet = inet_sk(newsk);
6892 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6893 * getsockname() and getpeername()
6895 newinet->inet_sport = inet->inet_sport;
6896 newinet->inet_saddr = inet->inet_saddr;
6897 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6898 newinet->inet_dport = htons(asoc->peer.port);
6899 newinet->pmtudisc = inet->pmtudisc;
6900 newinet->inet_id = asoc->next_tsn ^ jiffies;
6902 newinet->uc_ttl = inet->uc_ttl;
6903 newinet->mc_loop = 1;
6904 newinet->mc_ttl = 1;
6905 newinet->mc_index = 0;
6906 newinet->mc_list = NULL;
6909 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6910 * and its messages to the newsk.
6912 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6913 struct sctp_association *assoc,
6914 sctp_socket_type_t type)
6916 struct sctp_sock *oldsp = sctp_sk(oldsk);
6917 struct sctp_sock *newsp = sctp_sk(newsk);
6918 struct sctp_bind_bucket *pp; /* hash list port iterator */
6919 struct sctp_endpoint *newep = newsp->ep;
6920 struct sk_buff *skb, *tmp;
6921 struct sctp_ulpevent *event;
6922 struct sctp_bind_hashbucket *head;
6923 struct list_head tmplist;
6925 /* Migrate socket buffer sizes and all the socket level options to the
6928 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6929 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6930 /* Brute force copy old sctp opt. */
6931 if (oldsp->do_auto_asconf) {
6932 memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
6933 inet_sk_copy_descendant(newsk, oldsk);
6934 memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
6936 inet_sk_copy_descendant(newsk, oldsk);
6938 /* Restore the ep value that was overwritten with the above structure
6944 /* Hook this new socket in to the bind_hash list. */
6945 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
6946 inet_sk(oldsk)->inet_num)];
6948 spin_lock(&head->lock);
6949 pp = sctp_sk(oldsk)->bind_hash;
6950 sk_add_bind_node(newsk, &pp->owner);
6951 sctp_sk(newsk)->bind_hash = pp;
6952 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6953 spin_unlock(&head->lock);
6956 /* Copy the bind_addr list from the original endpoint to the new
6957 * endpoint so that we can handle restarts properly
6959 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6960 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6962 /* Move any messages in the old socket's receive queue that are for the
6963 * peeled off association to the new socket's receive queue.
6965 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6966 event = sctp_skb2event(skb);
6967 if (event->asoc == assoc) {
6968 __skb_unlink(skb, &oldsk->sk_receive_queue);
6969 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6970 sctp_skb_set_owner_r_frag(skb, newsk);
6974 /* Clean up any messages pending delivery due to partial
6975 * delivery. Three cases:
6976 * 1) No partial deliver; no work.
6977 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6978 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6980 skb_queue_head_init(&newsp->pd_lobby);
6981 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6983 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6984 struct sk_buff_head *queue;
6986 /* Decide which queue to move pd_lobby skbs to. */
6987 if (assoc->ulpq.pd_mode) {
6988 queue = &newsp->pd_lobby;
6990 queue = &newsk->sk_receive_queue;
6992 /* Walk through the pd_lobby, looking for skbs that
6993 * need moved to the new socket.
6995 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6996 event = sctp_skb2event(skb);
6997 if (event->asoc == assoc) {
6998 __skb_unlink(skb, &oldsp->pd_lobby);
6999 __skb_queue_tail(queue, skb);
7000 sctp_skb_set_owner_r_frag(skb, newsk);
7004 /* Clear up any skbs waiting for the partial
7005 * delivery to finish.
7007 if (assoc->ulpq.pd_mode)
7008 sctp_clear_pd(oldsk, NULL);
7012 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7013 sctp_skb_set_owner_r_frag(skb, newsk);
7015 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7016 sctp_skb_set_owner_r_frag(skb, newsk);
7018 /* Set the type of socket to indicate that it is peeled off from the
7019 * original UDP-style socket or created with the accept() call on a
7020 * TCP-style socket..
7024 /* Mark the new socket "in-use" by the user so that any packets
7025 * that may arrive on the association after we've moved it are
7026 * queued to the backlog. This prevents a potential race between
7027 * backlog processing on the old socket and new-packet processing
7028 * on the new socket.
7030 * The caller has just allocated newsk so we can guarantee that other
7031 * paths won't try to lock it and then oldsk.
7033 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7034 sctp_assoc_migrate(assoc, newsk);
7036 /* If the association on the newsk is already closed before accept()
7037 * is called, set RCV_SHUTDOWN flag.
7039 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
7040 newsk->sk_shutdown |= RCV_SHUTDOWN;
7042 newsk->sk_state = SCTP_SS_ESTABLISHED;
7043 release_sock(newsk);
7047 /* This proto struct describes the ULP interface for SCTP. */
7048 struct proto sctp_prot = {
7050 .owner = THIS_MODULE,
7051 .close = sctp_close,
7052 .connect = sctp_connect,
7053 .disconnect = sctp_disconnect,
7054 .accept = sctp_accept,
7055 .ioctl = sctp_ioctl,
7056 .init = sctp_init_sock,
7057 .destroy = sctp_destroy_sock,
7058 .shutdown = sctp_shutdown,
7059 .setsockopt = sctp_setsockopt,
7060 .getsockopt = sctp_getsockopt,
7061 .sendmsg = sctp_sendmsg,
7062 .recvmsg = sctp_recvmsg,
7064 .backlog_rcv = sctp_backlog_rcv,
7066 .unhash = sctp_unhash,
7067 .get_port = sctp_get_port,
7068 .obj_size = sizeof(struct sctp_sock),
7069 .sysctl_mem = sysctl_sctp_mem,
7070 .sysctl_rmem = sysctl_sctp_rmem,
7071 .sysctl_wmem = sysctl_sctp_wmem,
7072 .memory_pressure = &sctp_memory_pressure,
7073 .enter_memory_pressure = sctp_enter_memory_pressure,
7074 .memory_allocated = &sctp_memory_allocated,
7075 .sockets_allocated = &sctp_sockets_allocated,
7078 #if IS_ENABLED(CONFIG_IPV6)
7080 struct proto sctpv6_prot = {
7082 .owner = THIS_MODULE,
7083 .close = sctp_close,
7084 .connect = sctp_connect,
7085 .disconnect = sctp_disconnect,
7086 .accept = sctp_accept,
7087 .ioctl = sctp_ioctl,
7088 .init = sctp_init_sock,
7089 .destroy = sctp_destroy_sock,
7090 .shutdown = sctp_shutdown,
7091 .setsockopt = sctp_setsockopt,
7092 .getsockopt = sctp_getsockopt,
7093 .sendmsg = sctp_sendmsg,
7094 .recvmsg = sctp_recvmsg,
7096 .backlog_rcv = sctp_backlog_rcv,
7098 .unhash = sctp_unhash,
7099 .get_port = sctp_get_port,
7100 .obj_size = sizeof(struct sctp6_sock),
7101 .sysctl_mem = sysctl_sctp_mem,
7102 .sysctl_rmem = sysctl_sctp_rmem,
7103 .sysctl_wmem = sysctl_sctp_wmem,
7104 .memory_pressure = &sctp_memory_pressure,
7105 .enter_memory_pressure = sctp_enter_memory_pressure,
7106 .memory_allocated = &sctp_memory_allocated,
7107 .sockets_allocated = &sctp_sockets_allocated,
7109 #endif /* IS_ENABLED(CONFIG_IPV6) */