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 <crypto/hash.h>
56 #include <linux/types.h>
57 #include <linux/kernel.h>
58 #include <linux/wait.h>
59 #include <linux/time.h>
61 #include <linux/capability.h>
62 #include <linux/fcntl.h>
63 #include <linux/poll.h>
64 #include <linux/init.h>
65 #include <linux/slab.h>
66 #include <linux/file.h>
67 #include <linux/compat.h>
71 #include <net/route.h>
73 #include <net/inet_common.h>
74 #include <net/busy_poll.h>
76 #include <linux/socket.h> /* for sa_family_t */
77 #include <linux/export.h>
79 #include <net/sctp/sctp.h>
80 #include <net/sctp/sm.h>
82 /* Forward declarations for internal helper functions. */
83 static int sctp_writeable(struct sock *sk);
84 static void sctp_wfree(struct sk_buff *skb);
85 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
87 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p);
88 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
89 static int sctp_wait_for_accept(struct sock *sk, long timeo);
90 static void sctp_wait_for_close(struct sock *sk, long timeo);
91 static void sctp_destruct_sock(struct sock *sk);
92 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
93 union sctp_addr *addr, int len);
94 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
95 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
96 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
97 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
98 static int sctp_send_asconf(struct sctp_association *asoc,
99 struct sctp_chunk *chunk);
100 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
101 static int sctp_autobind(struct sock *sk);
102 static void sctp_sock_migrate(struct sock *, struct sock *,
103 struct sctp_association *, sctp_socket_type_t);
105 static int sctp_memory_pressure;
106 static atomic_long_t sctp_memory_allocated;
107 struct percpu_counter sctp_sockets_allocated;
109 static void sctp_enter_memory_pressure(struct sock *sk)
111 sctp_memory_pressure = 1;
115 /* Get the sndbuf space available at the time on the association. */
116 static inline int sctp_wspace(struct sctp_association *asoc)
120 if (asoc->ep->sndbuf_policy)
121 amt = asoc->sndbuf_used;
123 amt = sk_wmem_alloc_get(asoc->base.sk);
125 if (amt >= asoc->base.sk->sk_sndbuf) {
126 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
129 amt = sk_stream_wspace(asoc->base.sk);
134 amt = asoc->base.sk->sk_sndbuf - amt;
139 /* Increment the used sndbuf space count of the corresponding association by
140 * the size of the outgoing data chunk.
141 * Also, set the skb destructor for sndbuf accounting later.
143 * Since it is always 1-1 between chunk and skb, and also a new skb is always
144 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
145 * destructor in the data chunk skb for the purpose of the sndbuf space
148 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
150 struct sctp_association *asoc = chunk->asoc;
151 struct sock *sk = asoc->base.sk;
153 /* The sndbuf space is tracked per association. */
154 sctp_association_hold(asoc);
156 skb_set_owner_w(chunk->skb, sk);
158 chunk->skb->destructor = sctp_wfree;
159 /* Save the chunk pointer in skb for sctp_wfree to use later. */
160 skb_shinfo(chunk->skb)->destructor_arg = chunk;
162 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
163 sizeof(struct sk_buff) +
164 sizeof(struct sctp_chunk);
166 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
167 sk->sk_wmem_queued += chunk->skb->truesize;
168 sk_mem_charge(sk, chunk->skb->truesize);
171 static void sctp_clear_owner_w(struct sctp_chunk *chunk)
173 skb_orphan(chunk->skb);
176 #define traverse_and_process() \
179 if (msg == prev_msg) \
181 list_for_each_entry(c, &msg->chunks, frag_list) { \
182 if ((clear && asoc->base.sk == c->skb->sk) || \
183 (!clear && asoc->base.sk != c->skb->sk)) \
189 static void sctp_for_each_tx_datachunk(struct sctp_association *asoc,
191 void (*cb)(struct sctp_chunk *))
194 struct sctp_datamsg *msg, *prev_msg = NULL;
195 struct sctp_outq *q = &asoc->outqueue;
196 struct sctp_chunk *chunk, *c;
197 struct sctp_transport *t;
199 list_for_each_entry(t, &asoc->peer.transport_addr_list, transports)
200 list_for_each_entry(chunk, &t->transmitted, transmitted_list)
201 traverse_and_process();
203 list_for_each_entry(chunk, &q->retransmit, transmitted_list)
204 traverse_and_process();
206 list_for_each_entry(chunk, &q->sacked, transmitted_list)
207 traverse_and_process();
209 list_for_each_entry(chunk, &q->abandoned, transmitted_list)
210 traverse_and_process();
212 list_for_each_entry(chunk, &q->out_chunk_list, list)
213 traverse_and_process();
216 /* Verify that this is a valid address. */
217 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
222 /* Verify basic sockaddr. */
223 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
227 /* Is this a valid SCTP address? */
228 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
231 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
237 /* Look up the association by its id. If this is not a UDP-style
238 * socket, the ID field is always ignored.
240 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
242 struct sctp_association *asoc = NULL;
244 /* If this is not a UDP-style socket, assoc id should be ignored. */
245 if (!sctp_style(sk, UDP)) {
246 /* Return NULL if the socket state is not ESTABLISHED. It
247 * could be a TCP-style listening socket or a socket which
248 * hasn't yet called connect() to establish an association.
250 if (!sctp_sstate(sk, ESTABLISHED) && !sctp_sstate(sk, CLOSING))
253 /* Get the first and the only association from the list. */
254 if (!list_empty(&sctp_sk(sk)->ep->asocs))
255 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
256 struct sctp_association, asocs);
260 /* Otherwise this is a UDP-style socket. */
261 if (!id || (id == (sctp_assoc_t)-1))
264 spin_lock_bh(&sctp_assocs_id_lock);
265 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
266 if (asoc && (asoc->base.sk != sk || asoc->base.dead))
268 spin_unlock_bh(&sctp_assocs_id_lock);
273 /* Look up the transport from an address and an assoc id. If both address and
274 * id are specified, the associations matching the address and the id should be
277 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
278 struct sockaddr_storage *addr,
281 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
282 struct sctp_af *af = sctp_get_af_specific(addr->ss_family);
283 union sctp_addr *laddr = (union sctp_addr *)addr;
284 struct sctp_transport *transport;
286 if (!af || sctp_verify_addr(sk, laddr, af->sockaddr_len))
289 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
296 id_asoc = sctp_id2assoc(sk, id);
297 if (id_asoc && (id_asoc != addr_asoc))
300 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
301 (union sctp_addr *)addr);
306 /* API 3.1.2 bind() - UDP Style Syntax
307 * The syntax of bind() is,
309 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
311 * sd - the socket descriptor returned by socket().
312 * addr - the address structure (struct sockaddr_in or struct
313 * sockaddr_in6 [RFC 2553]),
314 * addr_len - the size of the address structure.
316 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
322 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
325 /* Disallow binding twice. */
326 if (!sctp_sk(sk)->ep->base.bind_addr.port)
327 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
337 static long sctp_get_port_local(struct sock *, union sctp_addr *);
339 /* Verify this is a valid sockaddr. */
340 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
341 union sctp_addr *addr, int len)
345 /* Check minimum size. */
346 if (len < sizeof (struct sockaddr))
349 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
352 if (addr->sa.sa_family == AF_INET6) {
353 if (len < SIN6_LEN_RFC2133)
355 /* V4 mapped address are really of AF_INET family */
356 if (ipv6_addr_v4mapped(&addr->v6.sin6_addr) &&
357 !opt->pf->af_supported(AF_INET, opt))
361 /* If we get this far, af is valid. */
362 af = sctp_get_af_specific(addr->sa.sa_family);
364 if (len < af->sockaddr_len)
370 static void sctp_auto_asconf_init(struct sctp_sock *sp)
372 struct net *net = sock_net(&sp->inet.sk);
374 if (net->sctp.default_auto_asconf) {
375 spin_lock(&net->sctp.addr_wq_lock);
376 list_add_tail(&sp->auto_asconf_list, &net->sctp.auto_asconf_splist);
377 spin_unlock(&net->sctp.addr_wq_lock);
378 sp->do_auto_asconf = 1;
382 /* Bind a local address either to an endpoint or to an association. */
383 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
385 struct net *net = sock_net(sk);
386 struct sctp_sock *sp = sctp_sk(sk);
387 struct sctp_endpoint *ep = sp->ep;
388 struct sctp_bind_addr *bp = &ep->base.bind_addr;
393 /* Common sockaddr verification. */
394 af = sctp_sockaddr_af(sp, addr, len);
396 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
397 __func__, sk, addr, len);
401 snum = ntohs(addr->v4.sin_port);
403 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
404 __func__, sk, &addr->sa, bp->port, snum, len);
406 /* PF specific bind() address verification. */
407 if (!sp->pf->bind_verify(sp, addr))
408 return -EADDRNOTAVAIL;
410 /* We must either be unbound, or bind to the same port.
411 * It's OK to allow 0 ports if we are already bound.
412 * We'll just inhert an already bound port in this case
417 else if (snum != bp->port) {
418 pr_debug("%s: new port %d doesn't match existing port "
419 "%d\n", __func__, snum, bp->port);
424 if (snum && snum < PROT_SOCK &&
425 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
428 /* See if the address matches any of the addresses we may have
429 * already bound before checking against other endpoints.
431 if (sctp_bind_addr_match(bp, addr, sp))
434 /* Make sure we are allowed to bind here.
435 * The function sctp_get_port_local() does duplicate address
438 addr->v4.sin_port = htons(snum);
439 if ((ret = sctp_get_port_local(sk, addr))) {
443 /* Refresh ephemeral port. */
445 bp->port = inet_sk(sk)->inet_num;
446 sctp_auto_asconf_init(sp);
449 /* Add the address to the bind address list.
450 * Use GFP_ATOMIC since BHs will be disabled.
452 ret = sctp_add_bind_addr(bp, addr, af->sockaddr_len,
453 SCTP_ADDR_SRC, GFP_ATOMIC);
455 /* Copy back into socket for getsockname() use. */
457 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
458 sp->pf->to_sk_saddr(addr, sk);
464 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
466 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
467 * at any one time. If a sender, after sending an ASCONF chunk, decides
468 * it needs to transfer another ASCONF Chunk, it MUST wait until the
469 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
470 * subsequent ASCONF. Note this restriction binds each side, so at any
471 * time two ASCONF may be in-transit on any given association (one sent
472 * from each endpoint).
474 static int sctp_send_asconf(struct sctp_association *asoc,
475 struct sctp_chunk *chunk)
477 struct net *net = sock_net(asoc->base.sk);
480 /* If there is an outstanding ASCONF chunk, queue it for later
483 if (asoc->addip_last_asconf) {
484 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
488 /* Hold the chunk until an ASCONF_ACK is received. */
489 sctp_chunk_hold(chunk);
490 retval = sctp_primitive_ASCONF(net, asoc, chunk);
492 sctp_chunk_free(chunk);
494 asoc->addip_last_asconf = chunk;
500 /* Add a list of addresses as bind addresses to local endpoint or
503 * Basically run through each address specified in the addrs/addrcnt
504 * array/length pair, determine if it is IPv6 or IPv4 and call
505 * sctp_do_bind() on it.
507 * If any of them fails, then the operation will be reversed and the
508 * ones that were added will be removed.
510 * Only sctp_setsockopt_bindx() is supposed to call this function.
512 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
517 struct sockaddr *sa_addr;
520 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
524 for (cnt = 0; cnt < addrcnt; cnt++) {
525 /* The list may contain either IPv4 or IPv6 address;
526 * determine the address length for walking thru the list.
529 af = sctp_get_af_specific(sa_addr->sa_family);
535 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
538 addr_buf += af->sockaddr_len;
542 /* Failed. Cleanup the ones that have been added */
544 sctp_bindx_rem(sk, addrs, cnt);
552 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
553 * associations that are part of the endpoint indicating that a list of local
554 * addresses are added to the endpoint.
556 * If any of the addresses is already in the bind address list of the
557 * association, we do not send the chunk for that association. But it will not
558 * affect other associations.
560 * Only sctp_setsockopt_bindx() is supposed to call this function.
562 static int sctp_send_asconf_add_ip(struct sock *sk,
563 struct sockaddr *addrs,
566 struct net *net = sock_net(sk);
567 struct sctp_sock *sp;
568 struct sctp_endpoint *ep;
569 struct sctp_association *asoc;
570 struct sctp_bind_addr *bp;
571 struct sctp_chunk *chunk;
572 struct sctp_sockaddr_entry *laddr;
573 union sctp_addr *addr;
574 union sctp_addr saveaddr;
581 if (!net->sctp.addip_enable)
587 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
588 __func__, sk, addrs, addrcnt);
590 list_for_each_entry(asoc, &ep->asocs, asocs) {
591 if (!asoc->peer.asconf_capable)
594 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
597 if (!sctp_state(asoc, ESTABLISHED))
600 /* Check if any address in the packed array of addresses is
601 * in the bind address list of the association. If so,
602 * do not send the asconf chunk to its peer, but continue with
603 * other associations.
606 for (i = 0; i < addrcnt; i++) {
608 af = sctp_get_af_specific(addr->v4.sin_family);
614 if (sctp_assoc_lookup_laddr(asoc, addr))
617 addr_buf += af->sockaddr_len;
622 /* Use the first valid address in bind addr list of
623 * association as Address Parameter of ASCONF CHUNK.
625 bp = &asoc->base.bind_addr;
626 p = bp->address_list.next;
627 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
628 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
629 addrcnt, SCTP_PARAM_ADD_IP);
635 /* Add the new addresses to the bind address list with
636 * use_as_src set to 0.
639 for (i = 0; i < addrcnt; i++) {
641 af = sctp_get_af_specific(addr->v4.sin_family);
642 memcpy(&saveaddr, addr, af->sockaddr_len);
643 retval = sctp_add_bind_addr(bp, &saveaddr,
645 SCTP_ADDR_NEW, GFP_ATOMIC);
646 addr_buf += af->sockaddr_len;
648 if (asoc->src_out_of_asoc_ok) {
649 struct sctp_transport *trans;
651 list_for_each_entry(trans,
652 &asoc->peer.transport_addr_list, transports) {
653 /* Clear the source and route cache */
654 dst_release(trans->dst);
655 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
656 2*asoc->pathmtu, 4380));
657 trans->ssthresh = asoc->peer.i.a_rwnd;
658 trans->rto = asoc->rto_initial;
659 sctp_max_rto(asoc, trans);
660 trans->rtt = trans->srtt = trans->rttvar = 0;
661 sctp_transport_route(trans, NULL,
662 sctp_sk(asoc->base.sk));
665 retval = sctp_send_asconf(asoc, chunk);
672 /* Remove a list of addresses from bind addresses list. Do not remove the
675 * Basically run through each address specified in the addrs/addrcnt
676 * array/length pair, determine if it is IPv6 or IPv4 and call
677 * sctp_del_bind() on it.
679 * If any of them fails, then the operation will be reversed and the
680 * ones that were removed will be added back.
682 * At least one address has to be left; if only one address is
683 * available, the operation will return -EBUSY.
685 * Only sctp_setsockopt_bindx() is supposed to call this function.
687 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
689 struct sctp_sock *sp = sctp_sk(sk);
690 struct sctp_endpoint *ep = sp->ep;
692 struct sctp_bind_addr *bp = &ep->base.bind_addr;
695 union sctp_addr *sa_addr;
698 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
699 __func__, sk, addrs, addrcnt);
702 for (cnt = 0; cnt < addrcnt; cnt++) {
703 /* If the bind address list is empty or if there is only one
704 * bind address, there is nothing more to be removed (we need
705 * at least one address here).
707 if (list_empty(&bp->address_list) ||
708 (sctp_list_single_entry(&bp->address_list))) {
714 af = sctp_get_af_specific(sa_addr->sa.sa_family);
720 if (!af->addr_valid(sa_addr, sp, NULL)) {
721 retval = -EADDRNOTAVAIL;
725 if (sa_addr->v4.sin_port &&
726 sa_addr->v4.sin_port != htons(bp->port)) {
731 if (!sa_addr->v4.sin_port)
732 sa_addr->v4.sin_port = htons(bp->port);
734 /* FIXME - There is probably a need to check if sk->sk_saddr and
735 * sk->sk_rcv_addr are currently set to one of the addresses to
736 * be removed. This is something which needs to be looked into
737 * when we are fixing the outstanding issues with multi-homing
738 * socket routing and failover schemes. Refer to comments in
739 * sctp_do_bind(). -daisy
741 retval = sctp_del_bind_addr(bp, sa_addr);
743 addr_buf += af->sockaddr_len;
746 /* Failed. Add the ones that has been removed back */
748 sctp_bindx_add(sk, addrs, cnt);
756 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
757 * the associations that are part of the endpoint indicating that a list of
758 * local addresses are removed from the endpoint.
760 * If any of the addresses is already in the bind address list of the
761 * association, we do not send the chunk for that association. But it will not
762 * affect other associations.
764 * Only sctp_setsockopt_bindx() is supposed to call this function.
766 static int sctp_send_asconf_del_ip(struct sock *sk,
767 struct sockaddr *addrs,
770 struct net *net = sock_net(sk);
771 struct sctp_sock *sp;
772 struct sctp_endpoint *ep;
773 struct sctp_association *asoc;
774 struct sctp_transport *transport;
775 struct sctp_bind_addr *bp;
776 struct sctp_chunk *chunk;
777 union sctp_addr *laddr;
780 struct sctp_sockaddr_entry *saddr;
786 if (!net->sctp.addip_enable)
792 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
793 __func__, sk, addrs, addrcnt);
795 list_for_each_entry(asoc, &ep->asocs, asocs) {
797 if (!asoc->peer.asconf_capable)
800 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
803 if (!sctp_state(asoc, ESTABLISHED))
806 /* Check if any address in the packed array of addresses is
807 * not present in the bind address list of the association.
808 * If so, do not send the asconf chunk to its peer, but
809 * continue with other associations.
812 for (i = 0; i < addrcnt; i++) {
814 af = sctp_get_af_specific(laddr->v4.sin_family);
820 if (!sctp_assoc_lookup_laddr(asoc, laddr))
823 addr_buf += af->sockaddr_len;
828 /* Find one address in the association's bind address list
829 * that is not in the packed array of addresses. This is to
830 * make sure that we do not delete all the addresses in the
833 bp = &asoc->base.bind_addr;
834 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
836 if ((laddr == NULL) && (addrcnt == 1)) {
837 if (asoc->asconf_addr_del_pending)
839 asoc->asconf_addr_del_pending =
840 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
841 if (asoc->asconf_addr_del_pending == NULL) {
845 asoc->asconf_addr_del_pending->sa.sa_family =
847 asoc->asconf_addr_del_pending->v4.sin_port =
849 if (addrs->sa_family == AF_INET) {
850 struct sockaddr_in *sin;
852 sin = (struct sockaddr_in *)addrs;
853 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
854 } else if (addrs->sa_family == AF_INET6) {
855 struct sockaddr_in6 *sin6;
857 sin6 = (struct sockaddr_in6 *)addrs;
858 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
861 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
862 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
863 asoc->asconf_addr_del_pending);
865 asoc->src_out_of_asoc_ok = 1;
873 /* We do not need RCU protection throughout this loop
874 * because this is done under a socket lock from the
877 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
885 /* Reset use_as_src flag for the addresses in the bind address
886 * list that are to be deleted.
889 for (i = 0; i < addrcnt; i++) {
891 af = sctp_get_af_specific(laddr->v4.sin_family);
892 list_for_each_entry(saddr, &bp->address_list, list) {
893 if (sctp_cmp_addr_exact(&saddr->a, laddr))
894 saddr->state = SCTP_ADDR_DEL;
896 addr_buf += af->sockaddr_len;
899 /* Update the route and saddr entries for all the transports
900 * as some of the addresses in the bind address list are
901 * about to be deleted and cannot be used as source addresses.
903 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
905 dst_release(transport->dst);
906 sctp_transport_route(transport, NULL,
907 sctp_sk(asoc->base.sk));
911 /* We don't need to transmit ASCONF */
913 retval = sctp_send_asconf(asoc, chunk);
919 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
920 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
922 struct sock *sk = sctp_opt2sk(sp);
923 union sctp_addr *addr;
926 /* It is safe to write port space in caller. */
928 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
929 af = sctp_get_af_specific(addr->sa.sa_family);
932 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
935 if (addrw->state == SCTP_ADDR_NEW)
936 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
938 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
941 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
944 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
947 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
948 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
951 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
952 * Section 3.1.2 for this usage.
954 * addrs is a pointer to an array of one or more socket addresses. Each
955 * address is contained in its appropriate structure (i.e. struct
956 * sockaddr_in or struct sockaddr_in6) the family of the address type
957 * must be used to distinguish the address length (note that this
958 * representation is termed a "packed array" of addresses). The caller
959 * specifies the number of addresses in the array with addrcnt.
961 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
962 * -1, and sets errno to the appropriate error code.
964 * For SCTP, the port given in each socket address must be the same, or
965 * sctp_bindx() will fail, setting errno to EINVAL.
967 * The flags parameter is formed from the bitwise OR of zero or more of
968 * the following currently defined flags:
970 * SCTP_BINDX_ADD_ADDR
972 * SCTP_BINDX_REM_ADDR
974 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
975 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
976 * addresses from the association. The two flags are mutually exclusive;
977 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
978 * not remove all addresses from an association; sctp_bindx() will
979 * reject such an attempt with EINVAL.
981 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
982 * additional addresses with an endpoint after calling bind(). Or use
983 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
984 * socket is associated with so that no new association accepted will be
985 * associated with those addresses. If the endpoint supports dynamic
986 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
987 * endpoint to send the appropriate message to the peer to change the
988 * peers address lists.
990 * Adding and removing addresses from a connected association is
991 * optional functionality. Implementations that do not support this
992 * functionality should return EOPNOTSUPP.
994 * Basically do nothing but copying the addresses from user to kernel
995 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
996 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
999 * We don't use copy_from_user() for optimization: we first do the
1000 * sanity checks (buffer size -fast- and access check-healthy
1001 * pointer); if all of those succeed, then we can alloc the memory
1002 * (expensive operation) needed to copy the data to kernel. Then we do
1003 * the copying without checking the user space area
1004 * (__copy_from_user()).
1006 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1009 * sk The sk of the socket
1010 * addrs The pointer to the addresses in user land
1011 * addrssize Size of the addrs buffer
1012 * op Operation to perform (add or remove, see the flags of
1015 * Returns 0 if ok, <0 errno code on error.
1017 static int sctp_setsockopt_bindx(struct sock *sk,
1018 struct sockaddr __user *addrs,
1019 int addrs_size, int op)
1021 struct sockaddr *kaddrs;
1025 struct sockaddr *sa_addr;
1029 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
1030 __func__, sk, addrs, addrs_size, op);
1032 if (unlikely(addrs_size <= 0))
1035 /* Check the user passed a healthy pointer. */
1036 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1039 /* Alloc space for the address array in kernel memory. */
1040 kaddrs = kmalloc(addrs_size, GFP_USER | __GFP_NOWARN);
1041 if (unlikely(!kaddrs))
1044 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1049 /* Walk through the addrs buffer and count the number of addresses. */
1051 while (walk_size < addrs_size) {
1052 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1058 af = sctp_get_af_specific(sa_addr->sa_family);
1060 /* If the address family is not supported or if this address
1061 * causes the address buffer to overflow return EINVAL.
1063 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1068 addr_buf += af->sockaddr_len;
1069 walk_size += af->sockaddr_len;
1074 case SCTP_BINDX_ADD_ADDR:
1075 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1078 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1081 case SCTP_BINDX_REM_ADDR:
1082 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1085 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1099 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1101 * Common routine for handling connect() and sctp_connectx().
1102 * Connect will come in with just a single address.
1104 static int __sctp_connect(struct sock *sk,
1105 struct sockaddr *kaddrs,
1106 int addrs_size, int flags,
1107 sctp_assoc_t *assoc_id)
1109 struct net *net = sock_net(sk);
1110 struct sctp_sock *sp;
1111 struct sctp_endpoint *ep;
1112 struct sctp_association *asoc = NULL;
1113 struct sctp_association *asoc2;
1114 struct sctp_transport *transport;
1121 union sctp_addr *sa_addr = NULL;
1123 unsigned short port;
1128 /* connect() cannot be done on a socket that is already in ESTABLISHED
1129 * state - UDP-style peeled off socket or a TCP-style socket that
1130 * is already connected.
1131 * It cannot be done even on a TCP-style listening socket.
1133 if (sctp_sstate(sk, ESTABLISHED) || sctp_sstate(sk, CLOSING) ||
1134 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1139 /* Walk through the addrs buffer and count the number of addresses. */
1141 while (walk_size < addrs_size) {
1144 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1150 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1152 /* If the address family is not supported or if this address
1153 * causes the address buffer to overflow return EINVAL.
1155 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1160 port = ntohs(sa_addr->v4.sin_port);
1162 /* Save current address so we can work with it */
1163 memcpy(&to, sa_addr, af->sockaddr_len);
1165 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1169 /* Make sure the destination port is correctly set
1172 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1177 /* Check if there already is a matching association on the
1178 * endpoint (other than the one created here).
1180 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1181 if (asoc2 && asoc2 != asoc) {
1182 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1189 /* If we could not find a matching association on the endpoint,
1190 * make sure that there is no peeled-off association matching
1191 * the peer address even on another socket.
1193 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1194 err = -EADDRNOTAVAIL;
1199 /* If a bind() or sctp_bindx() is not called prior to
1200 * an sctp_connectx() call, the system picks an
1201 * ephemeral port and will choose an address set
1202 * equivalent to binding with a wildcard address.
1204 if (!ep->base.bind_addr.port) {
1205 if (sctp_autobind(sk)) {
1211 * If an unprivileged user inherits a 1-many
1212 * style socket with open associations on a
1213 * privileged port, it MAY be permitted to
1214 * accept new associations, but it SHOULD NOT
1215 * be permitted to open new associations.
1217 if (ep->base.bind_addr.port < PROT_SOCK &&
1218 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1224 scope = sctp_scope(&to);
1225 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1231 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1239 /* Prime the peer's transport structures. */
1240 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1248 addr_buf += af->sockaddr_len;
1249 walk_size += af->sockaddr_len;
1252 /* In case the user of sctp_connectx() wants an association
1253 * id back, assign one now.
1256 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1261 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1266 /* Initialize sk's dport and daddr for getpeername() */
1267 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1268 sp->pf->to_sk_daddr(sa_addr, sk);
1271 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
1274 *assoc_id = asoc->assoc_id;
1275 err = sctp_wait_for_connect(asoc, &timeo);
1276 /* Note: the asoc may be freed after the return of
1277 * sctp_wait_for_connect.
1280 /* Don't free association on exit. */
1284 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1285 __func__, asoc, kaddrs, err);
1288 /* sctp_primitive_ASSOCIATE may have added this association
1289 * To the hash table, try to unhash it, just in case, its a noop
1290 * if it wasn't hashed so we're safe
1292 sctp_association_free(asoc);
1297 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1300 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1301 * sctp_assoc_t *asoc);
1303 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1304 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1305 * or IPv6 addresses.
1307 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1308 * Section 3.1.2 for this usage.
1310 * addrs is a pointer to an array of one or more socket addresses. Each
1311 * address is contained in its appropriate structure (i.e. struct
1312 * sockaddr_in or struct sockaddr_in6) the family of the address type
1313 * must be used to distengish the address length (note that this
1314 * representation is termed a "packed array" of addresses). The caller
1315 * specifies the number of addresses in the array with addrcnt.
1317 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1318 * the association id of the new association. On failure, sctp_connectx()
1319 * returns -1, and sets errno to the appropriate error code. The assoc_id
1320 * is not touched by the kernel.
1322 * For SCTP, the port given in each socket address must be the same, or
1323 * sctp_connectx() will fail, setting errno to EINVAL.
1325 * An application can use sctp_connectx to initiate an association with
1326 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1327 * allows a caller to specify multiple addresses at which a peer can be
1328 * reached. The way the SCTP stack uses the list of addresses to set up
1329 * the association is implementation dependent. This function only
1330 * specifies that the stack will try to make use of all the addresses in
1331 * the list when needed.
1333 * Note that the list of addresses passed in is only used for setting up
1334 * the association. It does not necessarily equal the set of addresses
1335 * the peer uses for the resulting association. If the caller wants to
1336 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1337 * retrieve them after the association has been set up.
1339 * Basically do nothing but copying the addresses from user to kernel
1340 * land and invoking either sctp_connectx(). This is used for tunneling
1341 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1343 * We don't use copy_from_user() for optimization: we first do the
1344 * sanity checks (buffer size -fast- and access check-healthy
1345 * pointer); if all of those succeed, then we can alloc the memory
1346 * (expensive operation) needed to copy the data to kernel. Then we do
1347 * the copying without checking the user space area
1348 * (__copy_from_user()).
1350 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1353 * sk The sk of the socket
1354 * addrs The pointer to the addresses in user land
1355 * addrssize Size of the addrs buffer
1357 * Returns >=0 if ok, <0 errno code on error.
1359 static int __sctp_setsockopt_connectx(struct sock *sk,
1360 struct sockaddr __user *addrs,
1362 sctp_assoc_t *assoc_id)
1364 struct sockaddr *kaddrs;
1365 gfp_t gfp = GFP_KERNEL;
1366 int err = 0, flags = 0;
1368 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1369 __func__, sk, addrs, addrs_size);
1371 if (unlikely(addrs_size <= 0))
1374 /* Check the user passed a healthy pointer. */
1375 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1378 /* Alloc space for the address array in kernel memory. */
1379 if (sk->sk_socket->file)
1380 gfp = GFP_USER | __GFP_NOWARN;
1381 kaddrs = kmalloc(addrs_size, gfp);
1382 if (unlikely(!kaddrs))
1385 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1390 /* in-kernel sockets don't generally have a file allocated to them
1391 * if all they do is call sock_create_kern().
1393 if (sk->sk_socket->file)
1394 flags = sk->sk_socket->file->f_flags;
1396 err = __sctp_connect(sk, kaddrs, addrs_size, flags, assoc_id);
1404 * This is an older interface. It's kept for backward compatibility
1405 * to the option that doesn't provide association id.
1407 static int sctp_setsockopt_connectx_old(struct sock *sk,
1408 struct sockaddr __user *addrs,
1411 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1415 * New interface for the API. The since the API is done with a socket
1416 * option, to make it simple we feed back the association id is as a return
1417 * indication to the call. Error is always negative and association id is
1420 static int sctp_setsockopt_connectx(struct sock *sk,
1421 struct sockaddr __user *addrs,
1424 sctp_assoc_t assoc_id = 0;
1427 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1436 * New (hopefully final) interface for the API.
1437 * We use the sctp_getaddrs_old structure so that use-space library
1438 * can avoid any unnecessary allocations. The only different part
1439 * is that we store the actual length of the address buffer into the
1440 * addrs_num structure member. That way we can re-use the existing
1443 #ifdef CONFIG_COMPAT
1444 struct compat_sctp_getaddrs_old {
1445 sctp_assoc_t assoc_id;
1447 compat_uptr_t addrs; /* struct sockaddr * */
1451 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1452 char __user *optval,
1455 struct sctp_getaddrs_old param;
1456 sctp_assoc_t assoc_id = 0;
1459 #ifdef CONFIG_COMPAT
1460 if (in_compat_syscall()) {
1461 struct compat_sctp_getaddrs_old param32;
1463 if (len < sizeof(param32))
1465 if (copy_from_user(¶m32, optval, sizeof(param32)))
1468 param.assoc_id = param32.assoc_id;
1469 param.addr_num = param32.addr_num;
1470 param.addrs = compat_ptr(param32.addrs);
1474 if (len < sizeof(param))
1476 if (copy_from_user(¶m, optval, sizeof(param)))
1480 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1481 param.addrs, param.addr_num,
1483 if (err == 0 || err == -EINPROGRESS) {
1484 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1486 if (put_user(sizeof(assoc_id), optlen))
1493 /* API 3.1.4 close() - UDP Style Syntax
1494 * Applications use close() to perform graceful shutdown (as described in
1495 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1496 * by a UDP-style socket.
1500 * ret = close(int sd);
1502 * sd - the socket descriptor of the associations to be closed.
1504 * To gracefully shutdown a specific association represented by the
1505 * UDP-style socket, an application should use the sendmsg() call,
1506 * passing no user data, but including the appropriate flag in the
1507 * ancillary data (see Section xxxx).
1509 * If sd in the close() call is a branched-off socket representing only
1510 * one association, the shutdown is performed on that association only.
1512 * 4.1.6 close() - TCP Style Syntax
1514 * Applications use close() to gracefully close down an association.
1518 * int close(int sd);
1520 * sd - the socket descriptor of the association to be closed.
1522 * After an application calls close() on a socket descriptor, no further
1523 * socket operations will succeed on that descriptor.
1525 * API 7.1.4 SO_LINGER
1527 * An application using the TCP-style socket can use this option to
1528 * perform the SCTP ABORT primitive. The linger option structure is:
1531 * int l_onoff; // option on/off
1532 * int l_linger; // linger time
1535 * To enable the option, set l_onoff to 1. If the l_linger value is set
1536 * to 0, calling close() is the same as the ABORT primitive. If the
1537 * value is set to a negative value, the setsockopt() call will return
1538 * an error. If the value is set to a positive value linger_time, the
1539 * close() can be blocked for at most linger_time ms. If the graceful
1540 * shutdown phase does not finish during this period, close() will
1541 * return but the graceful shutdown phase continues in the system.
1543 static void sctp_close(struct sock *sk, long timeout)
1545 struct net *net = sock_net(sk);
1546 struct sctp_endpoint *ep;
1547 struct sctp_association *asoc;
1548 struct list_head *pos, *temp;
1549 unsigned int data_was_unread;
1551 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1553 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
1554 sk->sk_shutdown = SHUTDOWN_MASK;
1555 sk->sk_state = SCTP_SS_CLOSING;
1557 ep = sctp_sk(sk)->ep;
1559 /* Clean up any skbs sitting on the receive queue. */
1560 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1561 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1563 /* Walk all associations on an endpoint. */
1564 list_for_each_safe(pos, temp, &ep->asocs) {
1565 asoc = list_entry(pos, struct sctp_association, asocs);
1567 if (sctp_style(sk, TCP)) {
1568 /* A closed association can still be in the list if
1569 * it belongs to a TCP-style listening socket that is
1570 * not yet accepted. If so, free it. If not, send an
1571 * ABORT or SHUTDOWN based on the linger options.
1573 if (sctp_state(asoc, CLOSED)) {
1574 sctp_association_free(asoc);
1579 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1580 !skb_queue_empty(&asoc->ulpq.reasm) ||
1581 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1582 struct sctp_chunk *chunk;
1584 chunk = sctp_make_abort_user(asoc, NULL, 0);
1585 sctp_primitive_ABORT(net, asoc, chunk);
1587 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1590 /* On a TCP-style socket, block for at most linger_time if set. */
1591 if (sctp_style(sk, TCP) && timeout)
1592 sctp_wait_for_close(sk, timeout);
1594 /* This will run the backlog queue. */
1597 /* Supposedly, no process has access to the socket, but
1598 * the net layers still may.
1599 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1600 * held and that should be grabbed before socket lock.
1602 spin_lock_bh(&net->sctp.addr_wq_lock);
1603 bh_lock_sock_nested(sk);
1605 /* Hold the sock, since sk_common_release() will put sock_put()
1606 * and we have just a little more cleanup.
1609 sk_common_release(sk);
1612 spin_unlock_bh(&net->sctp.addr_wq_lock);
1616 SCTP_DBG_OBJCNT_DEC(sock);
1619 /* Handle EPIPE error. */
1620 static int sctp_error(struct sock *sk, int flags, int err)
1623 err = sock_error(sk) ? : -EPIPE;
1624 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1625 send_sig(SIGPIPE, current, 0);
1629 /* API 3.1.3 sendmsg() - UDP Style Syntax
1631 * An application uses sendmsg() and recvmsg() calls to transmit data to
1632 * and receive data from its peer.
1634 * ssize_t sendmsg(int socket, const struct msghdr *message,
1637 * socket - the socket descriptor of the endpoint.
1638 * message - pointer to the msghdr structure which contains a single
1639 * user message and possibly some ancillary data.
1641 * See Section 5 for complete description of the data
1644 * flags - flags sent or received with the user message, see Section
1645 * 5 for complete description of the flags.
1647 * Note: This function could use a rewrite especially when explicit
1648 * connect support comes in.
1650 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1652 static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1654 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
1656 struct net *net = sock_net(sk);
1657 struct sctp_sock *sp;
1658 struct sctp_endpoint *ep;
1659 struct sctp_association *new_asoc = NULL, *asoc = NULL;
1660 struct sctp_transport *transport, *chunk_tp;
1661 struct sctp_chunk *chunk;
1663 struct sockaddr *msg_name = NULL;
1664 struct sctp_sndrcvinfo default_sinfo;
1665 struct sctp_sndrcvinfo *sinfo;
1666 struct sctp_initmsg *sinit;
1667 sctp_assoc_t associd = 0;
1668 sctp_cmsgs_t cmsgs = { NULL };
1670 bool fill_sinfo_ttl = false, wait_connect = false;
1671 struct sctp_datamsg *datamsg;
1672 int msg_flags = msg->msg_flags;
1673 __u16 sinfo_flags = 0;
1681 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1684 /* We cannot send a message over a TCP-style listening socket. */
1685 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1690 /* Parse out the SCTP CMSGs. */
1691 err = sctp_msghdr_parse(msg, &cmsgs);
1693 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1697 /* Fetch the destination address for this packet. This
1698 * address only selects the association--it is not necessarily
1699 * the address we will send to.
1700 * For a peeled-off socket, msg_name is ignored.
1702 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1703 int msg_namelen = msg->msg_namelen;
1705 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1710 if (msg_namelen > sizeof(to))
1711 msg_namelen = sizeof(to);
1712 memcpy(&to, msg->msg_name, msg_namelen);
1713 msg_name = msg->msg_name;
1717 if (cmsgs.sinfo != NULL) {
1718 memset(&default_sinfo, 0, sizeof(default_sinfo));
1719 default_sinfo.sinfo_stream = cmsgs.sinfo->snd_sid;
1720 default_sinfo.sinfo_flags = cmsgs.sinfo->snd_flags;
1721 default_sinfo.sinfo_ppid = cmsgs.sinfo->snd_ppid;
1722 default_sinfo.sinfo_context = cmsgs.sinfo->snd_context;
1723 default_sinfo.sinfo_assoc_id = cmsgs.sinfo->snd_assoc_id;
1725 sinfo = &default_sinfo;
1726 fill_sinfo_ttl = true;
1728 sinfo = cmsgs.srinfo;
1730 /* Did the user specify SNDINFO/SNDRCVINFO? */
1732 sinfo_flags = sinfo->sinfo_flags;
1733 associd = sinfo->sinfo_assoc_id;
1736 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1737 msg_len, sinfo_flags);
1739 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1740 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1745 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1746 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1747 * If SCTP_ABORT is set, the message length could be non zero with
1748 * the msg_iov set to the user abort reason.
1750 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1751 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1756 /* If SCTP_ADDR_OVER is set, there must be an address
1757 * specified in msg_name.
1759 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1766 pr_debug("%s: about to look up association\n", __func__);
1770 /* If a msg_name has been specified, assume this is to be used. */
1772 /* Look for a matching association on the endpoint. */
1773 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1775 /* If we could not find a matching association on the
1776 * endpoint, make sure that it is not a TCP-style
1777 * socket that already has an association or there is
1778 * no peeled-off association on another socket.
1781 ((sctp_style(sk, TCP) &&
1782 (sctp_sstate(sk, ESTABLISHED) ||
1783 sctp_sstate(sk, CLOSING))) ||
1784 sctp_endpoint_is_peeled_off(ep, &to))) {
1785 err = -EADDRNOTAVAIL;
1789 asoc = sctp_id2assoc(sk, associd);
1797 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1799 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1800 * socket that has an association in CLOSED state. This can
1801 * happen when an accepted socket has an association that is
1804 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1809 if (sinfo_flags & SCTP_EOF) {
1810 pr_debug("%s: shutting down association:%p\n",
1813 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1817 if (sinfo_flags & SCTP_ABORT) {
1819 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1825 pr_debug("%s: aborting association:%p\n",
1828 sctp_primitive_ABORT(net, asoc, chunk);
1834 /* Do we need to create the association? */
1836 pr_debug("%s: there is no association yet\n", __func__);
1838 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1843 /* Check for invalid stream against the stream counts,
1844 * either the default or the user specified stream counts.
1847 if (!sinit || !sinit->sinit_num_ostreams) {
1848 /* Check against the defaults. */
1849 if (sinfo->sinfo_stream >=
1850 sp->initmsg.sinit_num_ostreams) {
1855 /* Check against the requested. */
1856 if (sinfo->sinfo_stream >=
1857 sinit->sinit_num_ostreams) {
1865 * API 3.1.2 bind() - UDP Style Syntax
1866 * If a bind() or sctp_bindx() is not called prior to a
1867 * sendmsg() call that initiates a new association, the
1868 * system picks an ephemeral port and will choose an address
1869 * set equivalent to binding with a wildcard address.
1871 if (!ep->base.bind_addr.port) {
1872 if (sctp_autobind(sk)) {
1878 * If an unprivileged user inherits a one-to-many
1879 * style socket with open associations on a privileged
1880 * port, it MAY be permitted to accept new associations,
1881 * but it SHOULD NOT be permitted to open new
1884 if (ep->base.bind_addr.port < PROT_SOCK &&
1885 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1891 scope = sctp_scope(&to);
1892 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1898 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1904 /* If the SCTP_INIT ancillary data is specified, set all
1905 * the association init values accordingly.
1908 if (sinit->sinit_num_ostreams) {
1909 asoc->c.sinit_num_ostreams =
1910 sinit->sinit_num_ostreams;
1912 if (sinit->sinit_max_instreams) {
1913 asoc->c.sinit_max_instreams =
1914 sinit->sinit_max_instreams;
1916 if (sinit->sinit_max_attempts) {
1917 asoc->max_init_attempts
1918 = sinit->sinit_max_attempts;
1920 if (sinit->sinit_max_init_timeo) {
1921 asoc->max_init_timeo =
1922 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1926 /* Prime the peer's transport structures. */
1927 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1934 /* ASSERT: we have a valid association at this point. */
1935 pr_debug("%s: we have a valid association\n", __func__);
1938 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1939 * one with some defaults.
1941 memset(&default_sinfo, 0, sizeof(default_sinfo));
1942 default_sinfo.sinfo_stream = asoc->default_stream;
1943 default_sinfo.sinfo_flags = asoc->default_flags;
1944 default_sinfo.sinfo_ppid = asoc->default_ppid;
1945 default_sinfo.sinfo_context = asoc->default_context;
1946 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1947 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1949 sinfo = &default_sinfo;
1950 } else if (fill_sinfo_ttl) {
1951 /* In case SNDINFO was specified, we still need to fill
1952 * it with a default ttl from the assoc here.
1954 sinfo->sinfo_timetolive = asoc->default_timetolive;
1957 /* API 7.1.7, the sndbuf size per association bounds the
1958 * maximum size of data that can be sent in a single send call.
1960 if (msg_len > sk->sk_sndbuf) {
1965 if (asoc->pmtu_pending)
1966 sctp_assoc_pending_pmtu(sk, asoc);
1968 /* If fragmentation is disabled and the message length exceeds the
1969 * association fragmentation point, return EMSGSIZE. The I-D
1970 * does not specify what this error is, but this looks like
1973 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1978 /* Check for invalid stream. */
1979 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1984 if (sctp_wspace(asoc) < msg_len)
1985 sctp_prsctp_prune(asoc, sinfo, msg_len - sctp_wspace(asoc));
1987 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1988 if (!sctp_wspace(asoc)) {
1989 /* sk can be changed by peel off when waiting for buf. */
1990 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1992 if (err == -ESRCH) {
1993 /* asoc is already dead. */
2001 /* If an address is passed with the sendto/sendmsg call, it is used
2002 * to override the primary destination address in the TCP model, or
2003 * when SCTP_ADDR_OVER flag is set in the UDP model.
2005 if ((sctp_style(sk, TCP) && msg_name) ||
2006 (sinfo_flags & SCTP_ADDR_OVER)) {
2007 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
2015 /* Auto-connect, if we aren't connected already. */
2016 if (sctp_state(asoc, CLOSED)) {
2017 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
2021 wait_connect = true;
2022 pr_debug("%s: we associated primitively\n", __func__);
2025 /* Break the message into multiple chunks of maximum size. */
2026 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
2027 if (IS_ERR(datamsg)) {
2028 err = PTR_ERR(datamsg);
2032 /* Now send the (possibly) fragmented message. */
2033 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
2034 sctp_chunk_hold(chunk);
2036 /* Do accounting for the write space. */
2037 sctp_set_owner_w(chunk);
2039 chunk->transport = chunk_tp;
2042 /* Send it to the lower layers. Note: all chunks
2043 * must either fail or succeed. The lower layer
2044 * works that way today. Keep it that way or this
2047 err = sctp_primitive_SEND(net, asoc, datamsg);
2048 /* Did the lower layer accept the chunk? */
2050 sctp_datamsg_free(datamsg);
2054 pr_debug("%s: we sent primitively\n", __func__);
2056 sctp_datamsg_put(datamsg);
2059 if (unlikely(wait_connect)) {
2060 timeo = sock_sndtimeo(sk, msg_flags & MSG_DONTWAIT);
2061 sctp_wait_for_connect(asoc, &timeo);
2064 /* If we are already past ASSOCIATE, the lower
2065 * layers are responsible for association cleanup.
2071 sctp_association_free(asoc);
2076 return sctp_error(sk, msg_flags, err);
2083 err = sock_error(sk);
2093 /* This is an extended version of skb_pull() that removes the data from the
2094 * start of a skb even when data is spread across the list of skb's in the
2095 * frag_list. len specifies the total amount of data that needs to be removed.
2096 * when 'len' bytes could be removed from the skb, it returns 0.
2097 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2098 * could not be removed.
2100 static int sctp_skb_pull(struct sk_buff *skb, int len)
2102 struct sk_buff *list;
2103 int skb_len = skb_headlen(skb);
2106 if (len <= skb_len) {
2107 __skb_pull(skb, len);
2111 __skb_pull(skb, skb_len);
2113 skb_walk_frags(skb, list) {
2114 rlen = sctp_skb_pull(list, len);
2115 skb->len -= (len-rlen);
2116 skb->data_len -= (len-rlen);
2127 /* API 3.1.3 recvmsg() - UDP Style Syntax
2129 * ssize_t recvmsg(int socket, struct msghdr *message,
2132 * socket - the socket descriptor of the endpoint.
2133 * message - pointer to the msghdr structure which contains a single
2134 * user message and possibly some ancillary data.
2136 * See Section 5 for complete description of the data
2139 * flags - flags sent or received with the user message, see Section
2140 * 5 for complete description of the flags.
2142 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2143 int noblock, int flags, int *addr_len)
2145 struct sctp_ulpevent *event = NULL;
2146 struct sctp_sock *sp = sctp_sk(sk);
2147 struct sk_buff *skb, *head_skb;
2152 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2153 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2158 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED) &&
2159 !sctp_sstate(sk, CLOSING) && !sctp_sstate(sk, CLOSED)) {
2164 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2168 /* Get the total length of the skb including any skb's in the
2177 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2179 event = sctp_skb2event(skb);
2184 if (event->chunk && event->chunk->head_skb)
2185 head_skb = event->chunk->head_skb;
2188 sock_recv_ts_and_drops(msg, sk, head_skb);
2189 if (sctp_ulpevent_is_notification(event)) {
2190 msg->msg_flags |= MSG_NOTIFICATION;
2191 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2193 sp->pf->skb_msgname(head_skb, msg->msg_name, addr_len);
2196 /* Check if we allow SCTP_NXTINFO. */
2197 if (sp->recvnxtinfo)
2198 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2199 /* Check if we allow SCTP_RCVINFO. */
2200 if (sp->recvrcvinfo)
2201 sctp_ulpevent_read_rcvinfo(event, msg);
2202 /* Check if we allow SCTP_SNDRCVINFO. */
2203 if (sp->subscribe.sctp_data_io_event)
2204 sctp_ulpevent_read_sndrcvinfo(event, msg);
2208 /* If skb's length exceeds the user's buffer, update the skb and
2209 * push it back to the receive_queue so that the next call to
2210 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2212 if (skb_len > copied) {
2213 msg->msg_flags &= ~MSG_EOR;
2214 if (flags & MSG_PEEK)
2216 sctp_skb_pull(skb, copied);
2217 skb_queue_head(&sk->sk_receive_queue, skb);
2219 /* When only partial message is copied to the user, increase
2220 * rwnd by that amount. If all the data in the skb is read,
2221 * rwnd is updated when the event is freed.
2223 if (!sctp_ulpevent_is_notification(event))
2224 sctp_assoc_rwnd_increase(event->asoc, copied);
2226 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2227 (event->msg_flags & MSG_EOR))
2228 msg->msg_flags |= MSG_EOR;
2230 msg->msg_flags &= ~MSG_EOR;
2233 if (flags & MSG_PEEK) {
2234 /* Release the skb reference acquired after peeking the skb in
2235 * sctp_skb_recv_datagram().
2239 /* Free the event which includes releasing the reference to
2240 * the owner of the skb, freeing the skb and updating the
2243 sctp_ulpevent_free(event);
2250 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2252 * This option is a on/off flag. If enabled no SCTP message
2253 * fragmentation will be performed. Instead if a message being sent
2254 * exceeds the current PMTU size, the message will NOT be sent and
2255 * instead a error will be indicated to the user.
2257 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2258 char __user *optval,
2259 unsigned int optlen)
2263 if (optlen < sizeof(int))
2266 if (get_user(val, (int __user *)optval))
2269 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2274 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2275 unsigned int optlen)
2277 struct sctp_association *asoc;
2278 struct sctp_ulpevent *event;
2280 if (optlen > sizeof(struct sctp_event_subscribe))
2282 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2285 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2286 * if there is no data to be sent or retransmit, the stack will
2287 * immediately send up this notification.
2289 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2290 &sctp_sk(sk)->subscribe)) {
2291 asoc = sctp_id2assoc(sk, 0);
2293 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2294 event = sctp_ulpevent_make_sender_dry_event(asoc,
2299 sctp_ulpq_tail_event(&asoc->ulpq, event);
2306 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2308 * This socket option is applicable to the UDP-style socket only. When
2309 * set it will cause associations that are idle for more than the
2310 * specified number of seconds to automatically close. An association
2311 * being idle is defined an association that has NOT sent or received
2312 * user data. The special value of '0' indicates that no automatic
2313 * close of any associations should be performed. The option expects an
2314 * integer defining the number of seconds of idle time before an
2315 * association is closed.
2317 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2318 unsigned int optlen)
2320 struct sctp_sock *sp = sctp_sk(sk);
2321 struct net *net = sock_net(sk);
2323 /* Applicable to UDP-style socket only */
2324 if (sctp_style(sk, TCP))
2326 if (optlen != sizeof(int))
2328 if (copy_from_user(&sp->autoclose, optval, optlen))
2331 if (sp->autoclose > net->sctp.max_autoclose)
2332 sp->autoclose = net->sctp.max_autoclose;
2337 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2339 * Applications can enable or disable heartbeats for any peer address of
2340 * an association, modify an address's heartbeat interval, force a
2341 * heartbeat to be sent immediately, and adjust the address's maximum
2342 * number of retransmissions sent before an address is considered
2343 * unreachable. The following structure is used to access and modify an
2344 * address's parameters:
2346 * struct sctp_paddrparams {
2347 * sctp_assoc_t spp_assoc_id;
2348 * struct sockaddr_storage spp_address;
2349 * uint32_t spp_hbinterval;
2350 * uint16_t spp_pathmaxrxt;
2351 * uint32_t spp_pathmtu;
2352 * uint32_t spp_sackdelay;
2353 * uint32_t spp_flags;
2356 * spp_assoc_id - (one-to-many style socket) This is filled in the
2357 * application, and identifies the association for
2359 * spp_address - This specifies which address is of interest.
2360 * spp_hbinterval - This contains the value of the heartbeat interval,
2361 * in milliseconds. If a value of zero
2362 * is present in this field then no changes are to
2363 * be made to this parameter.
2364 * spp_pathmaxrxt - This contains the maximum number of
2365 * retransmissions before this address shall be
2366 * considered unreachable. If a value of zero
2367 * is present in this field then no changes are to
2368 * be made to this parameter.
2369 * spp_pathmtu - When Path MTU discovery is disabled the value
2370 * specified here will be the "fixed" path mtu.
2371 * Note that if the spp_address field is empty
2372 * then all associations on this address will
2373 * have this fixed path mtu set upon them.
2375 * spp_sackdelay - When delayed sack is enabled, this value specifies
2376 * the number of milliseconds that sacks will be delayed
2377 * for. This value will apply to all addresses of an
2378 * association if the spp_address field is empty. Note
2379 * also, that if delayed sack is enabled and this
2380 * value is set to 0, no change is made to the last
2381 * recorded delayed sack timer value.
2383 * spp_flags - These flags are used to control various features
2384 * on an association. The flag field may contain
2385 * zero or more of the following options.
2387 * SPP_HB_ENABLE - Enable heartbeats on the
2388 * specified address. Note that if the address
2389 * field is empty all addresses for the association
2390 * have heartbeats enabled upon them.
2392 * SPP_HB_DISABLE - Disable heartbeats on the
2393 * speicifed address. Note that if the address
2394 * field is empty all addresses for the association
2395 * will have their heartbeats disabled. Note also
2396 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2397 * mutually exclusive, only one of these two should
2398 * be specified. Enabling both fields will have
2399 * undetermined results.
2401 * SPP_HB_DEMAND - Request a user initiated heartbeat
2402 * to be made immediately.
2404 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2405 * heartbeat delayis to be set to the value of 0
2408 * SPP_PMTUD_ENABLE - This field will enable PMTU
2409 * discovery upon the specified address. Note that
2410 * if the address feild is empty then all addresses
2411 * on the association are effected.
2413 * SPP_PMTUD_DISABLE - This field will disable PMTU
2414 * discovery upon the specified address. Note that
2415 * if the address feild is empty then all addresses
2416 * on the association are effected. Not also that
2417 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2418 * exclusive. Enabling both will have undetermined
2421 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2422 * on delayed sack. The time specified in spp_sackdelay
2423 * is used to specify the sack delay for this address. Note
2424 * that if spp_address is empty then all addresses will
2425 * enable delayed sack and take on the sack delay
2426 * value specified in spp_sackdelay.
2427 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2428 * off delayed sack. If the spp_address field is blank then
2429 * delayed sack is disabled for the entire association. Note
2430 * also that this field is mutually exclusive to
2431 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2434 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2435 struct sctp_transport *trans,
2436 struct sctp_association *asoc,
2437 struct sctp_sock *sp,
2440 int sackdelay_change)
2444 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2445 struct net *net = sock_net(trans->asoc->base.sk);
2447 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2452 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2453 * this field is ignored. Note also that a value of zero indicates
2454 * the current setting should be left unchanged.
2456 if (params->spp_flags & SPP_HB_ENABLE) {
2458 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2459 * set. This lets us use 0 value when this flag
2462 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2463 params->spp_hbinterval = 0;
2465 if (params->spp_hbinterval ||
2466 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2469 msecs_to_jiffies(params->spp_hbinterval);
2472 msecs_to_jiffies(params->spp_hbinterval);
2474 sp->hbinterval = params->spp_hbinterval;
2481 trans->param_flags =
2482 (trans->param_flags & ~SPP_HB) | hb_change;
2485 (asoc->param_flags & ~SPP_HB) | hb_change;
2488 (sp->param_flags & ~SPP_HB) | hb_change;
2492 /* When Path MTU discovery is disabled the value specified here will
2493 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2494 * include the flag SPP_PMTUD_DISABLE for this field to have any
2497 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2499 trans->pathmtu = params->spp_pathmtu;
2500 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2502 asoc->pathmtu = params->spp_pathmtu;
2503 sctp_frag_point(asoc, params->spp_pathmtu);
2505 sp->pathmtu = params->spp_pathmtu;
2511 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2512 (params->spp_flags & SPP_PMTUD_ENABLE);
2513 trans->param_flags =
2514 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2516 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2517 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2521 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2524 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2528 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2529 * value of this field is ignored. Note also that a value of zero
2530 * indicates the current setting should be left unchanged.
2532 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2535 msecs_to_jiffies(params->spp_sackdelay);
2538 msecs_to_jiffies(params->spp_sackdelay);
2540 sp->sackdelay = params->spp_sackdelay;
2544 if (sackdelay_change) {
2546 trans->param_flags =
2547 (trans->param_flags & ~SPP_SACKDELAY) |
2551 (asoc->param_flags & ~SPP_SACKDELAY) |
2555 (sp->param_flags & ~SPP_SACKDELAY) |
2560 /* Note that a value of zero indicates the current setting should be
2563 if (params->spp_pathmaxrxt) {
2565 trans->pathmaxrxt = params->spp_pathmaxrxt;
2567 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2569 sp->pathmaxrxt = params->spp_pathmaxrxt;
2576 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2577 char __user *optval,
2578 unsigned int optlen)
2580 struct sctp_paddrparams params;
2581 struct sctp_transport *trans = NULL;
2582 struct sctp_association *asoc = NULL;
2583 struct sctp_sock *sp = sctp_sk(sk);
2585 int hb_change, pmtud_change, sackdelay_change;
2587 if (optlen != sizeof(struct sctp_paddrparams))
2590 if (copy_from_user(¶ms, optval, optlen))
2593 /* Validate flags and value parameters. */
2594 hb_change = params.spp_flags & SPP_HB;
2595 pmtud_change = params.spp_flags & SPP_PMTUD;
2596 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2598 if (hb_change == SPP_HB ||
2599 pmtud_change == SPP_PMTUD ||
2600 sackdelay_change == SPP_SACKDELAY ||
2601 params.spp_sackdelay > 500 ||
2602 (params.spp_pathmtu &&
2603 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2606 /* If an address other than INADDR_ANY is specified, and
2607 * no transport is found, then the request is invalid.
2609 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
2610 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2611 params.spp_assoc_id);
2616 /* Get association, if assoc_id != 0 and the socket is a one
2617 * to many style socket, and an association was not found, then
2618 * the id was invalid.
2620 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2621 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2624 /* Heartbeat demand can only be sent on a transport or
2625 * association, but not a socket.
2627 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2630 /* Process parameters. */
2631 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2632 hb_change, pmtud_change,
2638 /* If changes are for association, also apply parameters to each
2641 if (!trans && asoc) {
2642 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2644 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2645 hb_change, pmtud_change,
2653 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2655 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2658 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2660 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2664 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2666 * This option will effect the way delayed acks are performed. This
2667 * option allows you to get or set the delayed ack time, in
2668 * milliseconds. It also allows changing the delayed ack frequency.
2669 * Changing the frequency to 1 disables the delayed sack algorithm. If
2670 * the assoc_id is 0, then this sets or gets the endpoints default
2671 * values. If the assoc_id field is non-zero, then the set or get
2672 * effects the specified association for the one to many model (the
2673 * assoc_id field is ignored by the one to one model). Note that if
2674 * sack_delay or sack_freq are 0 when setting this option, then the
2675 * current values will remain unchanged.
2677 * struct sctp_sack_info {
2678 * sctp_assoc_t sack_assoc_id;
2679 * uint32_t sack_delay;
2680 * uint32_t sack_freq;
2683 * sack_assoc_id - This parameter, indicates which association the user
2684 * is performing an action upon. Note that if this field's value is
2685 * zero then the endpoints default value is changed (effecting future
2686 * associations only).
2688 * sack_delay - This parameter contains the number of milliseconds that
2689 * the user is requesting the delayed ACK timer be set to. Note that
2690 * this value is defined in the standard to be between 200 and 500
2693 * sack_freq - This parameter contains the number of packets that must
2694 * be received before a sack is sent without waiting for the delay
2695 * timer to expire. The default value for this is 2, setting this
2696 * value to 1 will disable the delayed sack algorithm.
2699 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2700 char __user *optval, unsigned int optlen)
2702 struct sctp_sack_info params;
2703 struct sctp_transport *trans = NULL;
2704 struct sctp_association *asoc = NULL;
2705 struct sctp_sock *sp = sctp_sk(sk);
2707 if (optlen == sizeof(struct sctp_sack_info)) {
2708 if (copy_from_user(¶ms, optval, optlen))
2711 if (params.sack_delay == 0 && params.sack_freq == 0)
2713 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2714 pr_warn_ratelimited(DEPRECATED
2716 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2717 "Use struct sctp_sack_info instead\n",
2718 current->comm, task_pid_nr(current));
2719 if (copy_from_user(¶ms, optval, optlen))
2722 if (params.sack_delay == 0)
2723 params.sack_freq = 1;
2725 params.sack_freq = 0;
2729 /* Validate value parameter. */
2730 if (params.sack_delay > 500)
2733 /* Get association, if sack_assoc_id != 0 and the socket is a one
2734 * to many style socket, and an association was not found, then
2735 * the id was invalid.
2737 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2738 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2741 if (params.sack_delay) {
2744 msecs_to_jiffies(params.sack_delay);
2746 sctp_spp_sackdelay_enable(asoc->param_flags);
2748 sp->sackdelay = params.sack_delay;
2750 sctp_spp_sackdelay_enable(sp->param_flags);
2754 if (params.sack_freq == 1) {
2757 sctp_spp_sackdelay_disable(asoc->param_flags);
2760 sctp_spp_sackdelay_disable(sp->param_flags);
2762 } else if (params.sack_freq > 1) {
2764 asoc->sackfreq = params.sack_freq;
2766 sctp_spp_sackdelay_enable(asoc->param_flags);
2768 sp->sackfreq = params.sack_freq;
2770 sctp_spp_sackdelay_enable(sp->param_flags);
2774 /* If change is for association, also apply to each transport. */
2776 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2778 if (params.sack_delay) {
2780 msecs_to_jiffies(params.sack_delay);
2781 trans->param_flags =
2782 sctp_spp_sackdelay_enable(trans->param_flags);
2784 if (params.sack_freq == 1) {
2785 trans->param_flags =
2786 sctp_spp_sackdelay_disable(trans->param_flags);
2787 } else if (params.sack_freq > 1) {
2788 trans->sackfreq = params.sack_freq;
2789 trans->param_flags =
2790 sctp_spp_sackdelay_enable(trans->param_flags);
2798 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2800 * Applications can specify protocol parameters for the default association
2801 * initialization. The option name argument to setsockopt() and getsockopt()
2804 * Setting initialization parameters is effective only on an unconnected
2805 * socket (for UDP-style sockets only future associations are effected
2806 * by the change). With TCP-style sockets, this option is inherited by
2807 * sockets derived from a listener socket.
2809 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2811 struct sctp_initmsg sinit;
2812 struct sctp_sock *sp = sctp_sk(sk);
2814 if (optlen != sizeof(struct sctp_initmsg))
2816 if (copy_from_user(&sinit, optval, optlen))
2819 if (sinit.sinit_num_ostreams)
2820 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2821 if (sinit.sinit_max_instreams)
2822 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2823 if (sinit.sinit_max_attempts)
2824 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2825 if (sinit.sinit_max_init_timeo)
2826 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2832 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2834 * Applications that wish to use the sendto() system call may wish to
2835 * specify a default set of parameters that would normally be supplied
2836 * through the inclusion of ancillary data. This socket option allows
2837 * such an application to set the default sctp_sndrcvinfo structure.
2838 * The application that wishes to use this socket option simply passes
2839 * in to this call the sctp_sndrcvinfo structure defined in Section
2840 * 5.2.2) The input parameters accepted by this call include
2841 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2842 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2843 * to this call if the caller is using the UDP model.
2845 static int sctp_setsockopt_default_send_param(struct sock *sk,
2846 char __user *optval,
2847 unsigned int optlen)
2849 struct sctp_sock *sp = sctp_sk(sk);
2850 struct sctp_association *asoc;
2851 struct sctp_sndrcvinfo info;
2853 if (optlen != sizeof(info))
2855 if (copy_from_user(&info, optval, optlen))
2857 if (info.sinfo_flags &
2858 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2859 SCTP_ABORT | SCTP_EOF))
2862 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2863 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2866 asoc->default_stream = info.sinfo_stream;
2867 asoc->default_flags = info.sinfo_flags;
2868 asoc->default_ppid = info.sinfo_ppid;
2869 asoc->default_context = info.sinfo_context;
2870 asoc->default_timetolive = info.sinfo_timetolive;
2872 sp->default_stream = info.sinfo_stream;
2873 sp->default_flags = info.sinfo_flags;
2874 sp->default_ppid = info.sinfo_ppid;
2875 sp->default_context = info.sinfo_context;
2876 sp->default_timetolive = info.sinfo_timetolive;
2882 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2883 * (SCTP_DEFAULT_SNDINFO)
2885 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
2886 char __user *optval,
2887 unsigned int optlen)
2889 struct sctp_sock *sp = sctp_sk(sk);
2890 struct sctp_association *asoc;
2891 struct sctp_sndinfo info;
2893 if (optlen != sizeof(info))
2895 if (copy_from_user(&info, optval, optlen))
2897 if (info.snd_flags &
2898 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2899 SCTP_ABORT | SCTP_EOF))
2902 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
2903 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
2906 asoc->default_stream = info.snd_sid;
2907 asoc->default_flags = info.snd_flags;
2908 asoc->default_ppid = info.snd_ppid;
2909 asoc->default_context = info.snd_context;
2911 sp->default_stream = info.snd_sid;
2912 sp->default_flags = info.snd_flags;
2913 sp->default_ppid = info.snd_ppid;
2914 sp->default_context = info.snd_context;
2920 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2922 * Requests that the local SCTP stack use the enclosed peer address as
2923 * the association primary. The enclosed address must be one of the
2924 * association peer's addresses.
2926 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2927 unsigned int optlen)
2929 struct sctp_prim prim;
2930 struct sctp_transport *trans;
2932 if (optlen != sizeof(struct sctp_prim))
2935 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2938 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2942 sctp_assoc_set_primary(trans->asoc, trans);
2948 * 7.1.5 SCTP_NODELAY
2950 * Turn on/off any Nagle-like algorithm. This means that packets are
2951 * generally sent as soon as possible and no unnecessary delays are
2952 * introduced, at the cost of more packets in the network. Expects an
2953 * integer boolean flag.
2955 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2956 unsigned int optlen)
2960 if (optlen < sizeof(int))
2962 if (get_user(val, (int __user *)optval))
2965 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2971 * 7.1.1 SCTP_RTOINFO
2973 * The protocol parameters used to initialize and bound retransmission
2974 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2975 * and modify these parameters.
2976 * All parameters are time values, in milliseconds. A value of 0, when
2977 * modifying the parameters, indicates that the current value should not
2981 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2983 struct sctp_rtoinfo rtoinfo;
2984 struct sctp_association *asoc;
2985 unsigned long rto_min, rto_max;
2986 struct sctp_sock *sp = sctp_sk(sk);
2988 if (optlen != sizeof (struct sctp_rtoinfo))
2991 if (copy_from_user(&rtoinfo, optval, optlen))
2994 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2996 /* Set the values to the specific association */
2997 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
3000 rto_max = rtoinfo.srto_max;
3001 rto_min = rtoinfo.srto_min;
3004 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
3006 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
3009 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
3011 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
3013 if (rto_min > rto_max)
3017 if (rtoinfo.srto_initial != 0)
3019 msecs_to_jiffies(rtoinfo.srto_initial);
3020 asoc->rto_max = rto_max;
3021 asoc->rto_min = rto_min;
3023 /* If there is no association or the association-id = 0
3024 * set the values to the endpoint.
3026 if (rtoinfo.srto_initial != 0)
3027 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
3028 sp->rtoinfo.srto_max = rto_max;
3029 sp->rtoinfo.srto_min = rto_min;
3037 * 7.1.2 SCTP_ASSOCINFO
3039 * This option is used to tune the maximum retransmission attempts
3040 * of the association.
3041 * Returns an error if the new association retransmission value is
3042 * greater than the sum of the retransmission value of the peer.
3043 * See [SCTP] for more information.
3046 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
3049 struct sctp_assocparams assocparams;
3050 struct sctp_association *asoc;
3052 if (optlen != sizeof(struct sctp_assocparams))
3054 if (copy_from_user(&assocparams, optval, optlen))
3057 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
3059 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
3062 /* Set the values to the specific association */
3064 if (assocparams.sasoc_asocmaxrxt != 0) {
3067 struct sctp_transport *peer_addr;
3069 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
3071 path_sum += peer_addr->pathmaxrxt;
3075 /* Only validate asocmaxrxt if we have more than
3076 * one path/transport. We do this because path
3077 * retransmissions are only counted when we have more
3081 assocparams.sasoc_asocmaxrxt > path_sum)
3084 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3087 if (assocparams.sasoc_cookie_life != 0)
3088 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3090 /* Set the values to the endpoint */
3091 struct sctp_sock *sp = sctp_sk(sk);
3093 if (assocparams.sasoc_asocmaxrxt != 0)
3094 sp->assocparams.sasoc_asocmaxrxt =
3095 assocparams.sasoc_asocmaxrxt;
3096 if (assocparams.sasoc_cookie_life != 0)
3097 sp->assocparams.sasoc_cookie_life =
3098 assocparams.sasoc_cookie_life;
3104 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3106 * This socket option is a boolean flag which turns on or off mapped V4
3107 * addresses. If this option is turned on and the socket is type
3108 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3109 * If this option is turned off, then no mapping will be done of V4
3110 * addresses and a user will receive both PF_INET6 and PF_INET type
3111 * addresses on the socket.
3113 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3116 struct sctp_sock *sp = sctp_sk(sk);
3118 if (optlen < sizeof(int))
3120 if (get_user(val, (int __user *)optval))
3131 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3132 * This option will get or set the maximum size to put in any outgoing
3133 * SCTP DATA chunk. If a message is larger than this size it will be
3134 * fragmented by SCTP into the specified size. Note that the underlying
3135 * SCTP implementation may fragment into smaller sized chunks when the
3136 * PMTU of the underlying association is smaller than the value set by
3137 * the user. The default value for this option is '0' which indicates
3138 * the user is NOT limiting fragmentation and only the PMTU will effect
3139 * SCTP's choice of DATA chunk size. Note also that values set larger
3140 * than the maximum size of an IP datagram will effectively let SCTP
3141 * control fragmentation (i.e. the same as setting this option to 0).
3143 * The following structure is used to access and modify this parameter:
3145 * struct sctp_assoc_value {
3146 * sctp_assoc_t assoc_id;
3147 * uint32_t assoc_value;
3150 * assoc_id: This parameter is ignored for one-to-one style sockets.
3151 * For one-to-many style sockets this parameter indicates which
3152 * association the user is performing an action upon. Note that if
3153 * this field's value is zero then the endpoints default value is
3154 * changed (effecting future associations only).
3155 * assoc_value: This parameter specifies the maximum size in bytes.
3157 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3159 struct sctp_sock *sp = sctp_sk(sk);
3160 struct sctp_assoc_value params;
3161 struct sctp_association *asoc;
3164 if (optlen == sizeof(int)) {
3165 pr_warn_ratelimited(DEPRECATED
3167 "Use of int in maxseg socket option.\n"
3168 "Use struct sctp_assoc_value instead\n",
3169 current->comm, task_pid_nr(current));
3170 if (copy_from_user(&val, optval, optlen))
3172 params.assoc_id = 0;
3173 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3174 if (copy_from_user(¶ms, optval, optlen))
3176 val = params.assoc_value;
3182 int min_len, max_len;
3184 min_len = SCTP_DEFAULT_MINSEGMENT - sp->pf->af->net_header_len;
3185 min_len -= sizeof(struct sctphdr) +
3186 sizeof(struct sctp_data_chunk);
3188 max_len = SCTP_MAX_CHUNK_LEN - sizeof(struct sctp_data_chunk);
3190 if (val < min_len || val > max_len)
3194 asoc = sctp_id2assoc(sk, params.assoc_id);
3197 val = asoc->pathmtu - sp->pf->af->net_header_len;
3198 val -= sizeof(struct sctphdr) +
3199 sizeof(struct sctp_data_chunk);
3201 asoc->user_frag = val;
3202 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3204 if (params.assoc_id && sctp_style(sk, UDP))
3206 sp->user_frag = val;
3214 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3216 * Requests that the peer mark the enclosed address as the association
3217 * primary. The enclosed address must be one of the association's
3218 * locally bound addresses. The following structure is used to make a
3219 * set primary request:
3221 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3222 unsigned int optlen)
3224 struct net *net = sock_net(sk);
3225 struct sctp_sock *sp;
3226 struct sctp_association *asoc = NULL;
3227 struct sctp_setpeerprim prim;
3228 struct sctp_chunk *chunk;
3234 if (!net->sctp.addip_enable)
3237 if (optlen != sizeof(struct sctp_setpeerprim))
3240 if (copy_from_user(&prim, optval, optlen))
3243 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3247 if (!asoc->peer.asconf_capable)
3250 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3253 if (!sctp_state(asoc, ESTABLISHED))
3256 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3260 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3261 return -EADDRNOTAVAIL;
3263 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3264 return -EADDRNOTAVAIL;
3266 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3267 chunk = sctp_make_asconf_set_prim(asoc,
3268 (union sctp_addr *)&prim.sspp_addr);
3272 err = sctp_send_asconf(asoc, chunk);
3274 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3279 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3280 unsigned int optlen)
3282 struct sctp_setadaptation adaptation;
3284 if (optlen != sizeof(struct sctp_setadaptation))
3286 if (copy_from_user(&adaptation, optval, optlen))
3289 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3295 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3297 * The context field in the sctp_sndrcvinfo structure is normally only
3298 * used when a failed message is retrieved holding the value that was
3299 * sent down on the actual send call. This option allows the setting of
3300 * a default context on an association basis that will be received on
3301 * reading messages from the peer. This is especially helpful in the
3302 * one-2-many model for an application to keep some reference to an
3303 * internal state machine that is processing messages on the
3304 * association. Note that the setting of this value only effects
3305 * received messages from the peer and does not effect the value that is
3306 * saved with outbound messages.
3308 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3309 unsigned int optlen)
3311 struct sctp_assoc_value params;
3312 struct sctp_sock *sp;
3313 struct sctp_association *asoc;
3315 if (optlen != sizeof(struct sctp_assoc_value))
3317 if (copy_from_user(¶ms, optval, optlen))
3322 if (params.assoc_id != 0) {
3323 asoc = sctp_id2assoc(sk, params.assoc_id);
3326 asoc->default_rcv_context = params.assoc_value;
3328 sp->default_rcv_context = params.assoc_value;
3335 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3337 * This options will at a minimum specify if the implementation is doing
3338 * fragmented interleave. Fragmented interleave, for a one to many
3339 * socket, is when subsequent calls to receive a message may return
3340 * parts of messages from different associations. Some implementations
3341 * may allow you to turn this value on or off. If so, when turned off,
3342 * no fragment interleave will occur (which will cause a head of line
3343 * blocking amongst multiple associations sharing the same one to many
3344 * socket). When this option is turned on, then each receive call may
3345 * come from a different association (thus the user must receive data
3346 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3347 * association each receive belongs to.
3349 * This option takes a boolean value. A non-zero value indicates that
3350 * fragmented interleave is on. A value of zero indicates that
3351 * fragmented interleave is off.
3353 * Note that it is important that an implementation that allows this
3354 * option to be turned on, have it off by default. Otherwise an unaware
3355 * application using the one to many model may become confused and act
3358 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3359 char __user *optval,
3360 unsigned int optlen)
3364 if (optlen != sizeof(int))
3366 if (get_user(val, (int __user *)optval))
3369 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3375 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3376 * (SCTP_PARTIAL_DELIVERY_POINT)
3378 * This option will set or get the SCTP partial delivery point. This
3379 * point is the size of a message where the partial delivery API will be
3380 * invoked to help free up rwnd space for the peer. Setting this to a
3381 * lower value will cause partial deliveries to happen more often. The
3382 * calls argument is an integer that sets or gets the partial delivery
3383 * point. Note also that the call will fail if the user attempts to set
3384 * this value larger than the socket receive buffer size.
3386 * Note that any single message having a length smaller than or equal to
3387 * the SCTP partial delivery point will be delivered in one single read
3388 * call as long as the user provided buffer is large enough to hold the
3391 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3392 char __user *optval,
3393 unsigned int optlen)
3397 if (optlen != sizeof(u32))
3399 if (get_user(val, (int __user *)optval))
3402 /* Note: We double the receive buffer from what the user sets
3403 * it to be, also initial rwnd is based on rcvbuf/2.
3405 if (val > (sk->sk_rcvbuf >> 1))
3408 sctp_sk(sk)->pd_point = val;
3410 return 0; /* is this the right error code? */
3414 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3416 * This option will allow a user to change the maximum burst of packets
3417 * that can be emitted by this association. Note that the default value
3418 * is 4, and some implementations may restrict this setting so that it
3419 * can only be lowered.
3421 * NOTE: This text doesn't seem right. Do this on a socket basis with
3422 * future associations inheriting the socket value.
3424 static int sctp_setsockopt_maxburst(struct sock *sk,
3425 char __user *optval,
3426 unsigned int optlen)
3428 struct sctp_assoc_value params;
3429 struct sctp_sock *sp;
3430 struct sctp_association *asoc;
3434 if (optlen == sizeof(int)) {
3435 pr_warn_ratelimited(DEPRECATED
3437 "Use of int in max_burst socket option deprecated.\n"
3438 "Use struct sctp_assoc_value instead\n",
3439 current->comm, task_pid_nr(current));
3440 if (copy_from_user(&val, optval, optlen))
3442 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3443 if (copy_from_user(¶ms, optval, optlen))
3445 val = params.assoc_value;
3446 assoc_id = params.assoc_id;
3452 if (assoc_id != 0) {
3453 asoc = sctp_id2assoc(sk, assoc_id);
3456 asoc->max_burst = val;
3458 sp->max_burst = val;
3464 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3466 * This set option adds a chunk type that the user is requesting to be
3467 * received only in an authenticated way. Changes to the list of chunks
3468 * will only effect future associations on the socket.
3470 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3471 char __user *optval,
3472 unsigned int optlen)
3474 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3475 struct sctp_authchunk val;
3477 if (!ep->auth_enable)
3480 if (optlen != sizeof(struct sctp_authchunk))
3482 if (copy_from_user(&val, optval, optlen))
3485 switch (val.sauth_chunk) {
3487 case SCTP_CID_INIT_ACK:
3488 case SCTP_CID_SHUTDOWN_COMPLETE:
3493 /* add this chunk id to the endpoint */
3494 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3498 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3500 * This option gets or sets the list of HMAC algorithms that the local
3501 * endpoint requires the peer to use.
3503 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3504 char __user *optval,
3505 unsigned int optlen)
3507 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3508 struct sctp_hmacalgo *hmacs;
3512 if (!ep->auth_enable)
3515 if (optlen < sizeof(struct sctp_hmacalgo))
3518 hmacs = memdup_user(optval, optlen);
3520 return PTR_ERR(hmacs);
3522 idents = hmacs->shmac_num_idents;
3523 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3524 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3529 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3536 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3538 * This option will set a shared secret key which is used to build an
3539 * association shared key.
3541 static int sctp_setsockopt_auth_key(struct sock *sk,
3542 char __user *optval,
3543 unsigned int optlen)
3545 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3546 struct sctp_authkey *authkey;
3547 struct sctp_association *asoc;
3550 if (!ep->auth_enable)
3553 if (optlen <= sizeof(struct sctp_authkey))
3556 authkey = memdup_user(optval, optlen);
3557 if (IS_ERR(authkey))
3558 return PTR_ERR(authkey);
3560 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3565 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3566 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3571 ret = sctp_auth_set_key(ep, asoc, authkey);
3578 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3580 * This option will get or set the active shared key to be used to build
3581 * the association shared key.
3583 static int sctp_setsockopt_active_key(struct sock *sk,
3584 char __user *optval,
3585 unsigned int optlen)
3587 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3588 struct sctp_authkeyid val;
3589 struct sctp_association *asoc;
3591 if (!ep->auth_enable)
3594 if (optlen != sizeof(struct sctp_authkeyid))
3596 if (copy_from_user(&val, optval, optlen))
3599 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3600 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3603 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3607 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3609 * This set option will delete a shared secret key from use.
3611 static int sctp_setsockopt_del_key(struct sock *sk,
3612 char __user *optval,
3613 unsigned int optlen)
3615 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3616 struct sctp_authkeyid val;
3617 struct sctp_association *asoc;
3619 if (!ep->auth_enable)
3622 if (optlen != sizeof(struct sctp_authkeyid))
3624 if (copy_from_user(&val, optval, optlen))
3627 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3628 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3631 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3636 * 8.1.23 SCTP_AUTO_ASCONF
3638 * This option will enable or disable the use of the automatic generation of
3639 * ASCONF chunks to add and delete addresses to an existing association. Note
3640 * that this option has two caveats namely: a) it only affects sockets that
3641 * are bound to all addresses available to the SCTP stack, and b) the system
3642 * administrator may have an overriding control that turns the ASCONF feature
3643 * off no matter what setting the socket option may have.
3644 * This option expects an integer boolean flag, where a non-zero value turns on
3645 * the option, and a zero value turns off the option.
3646 * Note. In this implementation, socket operation overrides default parameter
3647 * being set by sysctl as well as FreeBSD implementation
3649 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3650 unsigned int optlen)
3653 struct sctp_sock *sp = sctp_sk(sk);
3655 if (optlen < sizeof(int))
3657 if (get_user(val, (int __user *)optval))
3659 if (!sctp_is_ep_boundall(sk) && val)
3661 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3664 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3665 if (val == 0 && sp->do_auto_asconf) {
3666 list_del(&sp->auto_asconf_list);
3667 sp->do_auto_asconf = 0;
3668 } else if (val && !sp->do_auto_asconf) {
3669 list_add_tail(&sp->auto_asconf_list,
3670 &sock_net(sk)->sctp.auto_asconf_splist);
3671 sp->do_auto_asconf = 1;
3673 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3678 * SCTP_PEER_ADDR_THLDS
3680 * This option allows us to alter the partially failed threshold for one or all
3681 * transports in an association. See Section 6.1 of:
3682 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3684 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3685 char __user *optval,
3686 unsigned int optlen)
3688 struct sctp_paddrthlds val;
3689 struct sctp_transport *trans;
3690 struct sctp_association *asoc;
3692 if (optlen < sizeof(struct sctp_paddrthlds))
3694 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3695 sizeof(struct sctp_paddrthlds)))
3699 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3700 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3703 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3705 if (val.spt_pathmaxrxt)
3706 trans->pathmaxrxt = val.spt_pathmaxrxt;
3707 trans->pf_retrans = val.spt_pathpfthld;
3710 if (val.spt_pathmaxrxt)
3711 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3712 asoc->pf_retrans = val.spt_pathpfthld;
3714 trans = sctp_addr_id2transport(sk, &val.spt_address,
3719 if (val.spt_pathmaxrxt)
3720 trans->pathmaxrxt = val.spt_pathmaxrxt;
3721 trans->pf_retrans = val.spt_pathpfthld;
3727 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3728 char __user *optval,
3729 unsigned int optlen)
3733 if (optlen < sizeof(int))
3735 if (get_user(val, (int __user *) optval))
3738 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3743 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3744 char __user *optval,
3745 unsigned int optlen)
3749 if (optlen < sizeof(int))
3751 if (get_user(val, (int __user *) optval))
3754 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3759 static int sctp_setsockopt_pr_supported(struct sock *sk,
3760 char __user *optval,
3761 unsigned int optlen)
3763 struct sctp_assoc_value params;
3765 if (optlen != sizeof(params))
3768 if (copy_from_user(¶ms, optval, optlen))
3771 sctp_sk(sk)->ep->prsctp_enable = !!params.assoc_value;
3776 static int sctp_setsockopt_default_prinfo(struct sock *sk,
3777 char __user *optval,
3778 unsigned int optlen)
3780 struct sctp_default_prinfo info;
3781 struct sctp_association *asoc;
3782 int retval = -EINVAL;
3784 if (optlen != sizeof(info))
3787 if (copy_from_user(&info, optval, sizeof(info))) {
3792 if (info.pr_policy & ~SCTP_PR_SCTP_MASK)
3795 if (info.pr_policy == SCTP_PR_SCTP_NONE)
3798 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
3800 SCTP_PR_SET_POLICY(asoc->default_flags, info.pr_policy);
3801 asoc->default_timetolive = info.pr_value;
3802 } else if (!info.pr_assoc_id) {
3803 struct sctp_sock *sp = sctp_sk(sk);
3805 SCTP_PR_SET_POLICY(sp->default_flags, info.pr_policy);
3806 sp->default_timetolive = info.pr_value;
3817 /* API 6.2 setsockopt(), getsockopt()
3819 * Applications use setsockopt() and getsockopt() to set or retrieve
3820 * socket options. Socket options are used to change the default
3821 * behavior of sockets calls. They are described in Section 7.
3825 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3826 * int __user *optlen);
3827 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3830 * sd - the socket descript.
3831 * level - set to IPPROTO_SCTP for all SCTP options.
3832 * optname - the option name.
3833 * optval - the buffer to store the value of the option.
3834 * optlen - the size of the buffer.
3836 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3837 char __user *optval, unsigned int optlen)
3841 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3843 /* I can hardly begin to describe how wrong this is. This is
3844 * so broken as to be worse than useless. The API draft
3845 * REALLY is NOT helpful here... I am not convinced that the
3846 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3847 * are at all well-founded.
3849 if (level != SOL_SCTP) {
3850 struct sctp_af *af = sctp_sk(sk)->pf->af;
3851 retval = af->setsockopt(sk, level, optname, optval, optlen);
3858 case SCTP_SOCKOPT_BINDX_ADD:
3859 /* 'optlen' is the size of the addresses buffer. */
3860 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3861 optlen, SCTP_BINDX_ADD_ADDR);
3864 case SCTP_SOCKOPT_BINDX_REM:
3865 /* 'optlen' is the size of the addresses buffer. */
3866 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3867 optlen, SCTP_BINDX_REM_ADDR);
3870 case SCTP_SOCKOPT_CONNECTX_OLD:
3871 /* 'optlen' is the size of the addresses buffer. */
3872 retval = sctp_setsockopt_connectx_old(sk,
3873 (struct sockaddr __user *)optval,
3877 case SCTP_SOCKOPT_CONNECTX:
3878 /* 'optlen' is the size of the addresses buffer. */
3879 retval = sctp_setsockopt_connectx(sk,
3880 (struct sockaddr __user *)optval,
3884 case SCTP_DISABLE_FRAGMENTS:
3885 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3889 retval = sctp_setsockopt_events(sk, optval, optlen);
3892 case SCTP_AUTOCLOSE:
3893 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3896 case SCTP_PEER_ADDR_PARAMS:
3897 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3900 case SCTP_DELAYED_SACK:
3901 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3903 case SCTP_PARTIAL_DELIVERY_POINT:
3904 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3908 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3910 case SCTP_DEFAULT_SEND_PARAM:
3911 retval = sctp_setsockopt_default_send_param(sk, optval,
3914 case SCTP_DEFAULT_SNDINFO:
3915 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
3917 case SCTP_PRIMARY_ADDR:
3918 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3920 case SCTP_SET_PEER_PRIMARY_ADDR:
3921 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3924 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3927 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3929 case SCTP_ASSOCINFO:
3930 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3932 case SCTP_I_WANT_MAPPED_V4_ADDR:
3933 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3936 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3938 case SCTP_ADAPTATION_LAYER:
3939 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3942 retval = sctp_setsockopt_context(sk, optval, optlen);
3944 case SCTP_FRAGMENT_INTERLEAVE:
3945 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3947 case SCTP_MAX_BURST:
3948 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3950 case SCTP_AUTH_CHUNK:
3951 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3953 case SCTP_HMAC_IDENT:
3954 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3957 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3959 case SCTP_AUTH_ACTIVE_KEY:
3960 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3962 case SCTP_AUTH_DELETE_KEY:
3963 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3965 case SCTP_AUTO_ASCONF:
3966 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3968 case SCTP_PEER_ADDR_THLDS:
3969 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3971 case SCTP_RECVRCVINFO:
3972 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
3974 case SCTP_RECVNXTINFO:
3975 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
3977 case SCTP_PR_SUPPORTED:
3978 retval = sctp_setsockopt_pr_supported(sk, optval, optlen);
3980 case SCTP_DEFAULT_PRINFO:
3981 retval = sctp_setsockopt_default_prinfo(sk, optval, optlen);
3984 retval = -ENOPROTOOPT;
3994 /* API 3.1.6 connect() - UDP Style Syntax
3996 * An application may use the connect() call in the UDP model to initiate an
3997 * association without sending data.
4001 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
4003 * sd: the socket descriptor to have a new association added to.
4005 * nam: the address structure (either struct sockaddr_in or struct
4006 * sockaddr_in6 defined in RFC2553 [7]).
4008 * len: the size of the address.
4010 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
4011 int addr_len, int flags)
4017 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
4020 /* Validate addr_len before calling common connect/connectx routine. */
4021 af = sctp_get_af_specific(addr->sa_family);
4022 if (af && addr_len >= af->sockaddr_len)
4023 err = __sctp_connect(sk, addr, af->sockaddr_len, flags, NULL);
4029 int sctp_inet_connect(struct socket *sock, struct sockaddr *uaddr,
4030 int addr_len, int flags)
4032 if (addr_len < sizeof(uaddr->sa_family))
4035 if (uaddr->sa_family == AF_UNSPEC)
4038 return sctp_connect(sock->sk, uaddr, addr_len, flags);
4041 /* FIXME: Write comments. */
4042 static int sctp_disconnect(struct sock *sk, int flags)
4044 return -EOPNOTSUPP; /* STUB */
4047 /* 4.1.4 accept() - TCP Style Syntax
4049 * Applications use accept() call to remove an established SCTP
4050 * association from the accept queue of the endpoint. A new socket
4051 * descriptor will be returned from accept() to represent the newly
4052 * formed association.
4054 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
4056 struct sctp_sock *sp;
4057 struct sctp_endpoint *ep;
4058 struct sock *newsk = NULL;
4059 struct sctp_association *asoc;
4068 if (!sctp_style(sk, TCP)) {
4069 error = -EOPNOTSUPP;
4073 if (!sctp_sstate(sk, LISTENING)) {
4078 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
4080 error = sctp_wait_for_accept(sk, timeo);
4084 /* We treat the list of associations on the endpoint as the accept
4085 * queue and pick the first association on the list.
4087 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
4089 newsk = sp->pf->create_accept_sk(sk, asoc);
4095 /* Populate the fields of the newsk from the oldsk and migrate the
4096 * asoc to the newsk.
4098 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
4106 /* The SCTP ioctl handler. */
4107 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
4114 * SEQPACKET-style sockets in LISTENING state are valid, for
4115 * SCTP, so only discard TCP-style sockets in LISTENING state.
4117 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4122 struct sk_buff *skb;
4123 unsigned int amount = 0;
4125 skb = skb_peek(&sk->sk_receive_queue);
4128 * We will only return the amount of this packet since
4129 * that is all that will be read.
4133 rc = put_user(amount, (int __user *)arg);
4145 /* This is the function which gets called during socket creation to
4146 * initialized the SCTP-specific portion of the sock.
4147 * The sock structure should already be zero-filled memory.
4149 static int sctp_init_sock(struct sock *sk)
4151 struct net *net = sock_net(sk);
4152 struct sctp_sock *sp;
4154 pr_debug("%s: sk:%p\n", __func__, sk);
4158 /* Initialize the SCTP per socket area. */
4159 switch (sk->sk_type) {
4160 case SOCK_SEQPACKET:
4161 sp->type = SCTP_SOCKET_UDP;
4164 sp->type = SCTP_SOCKET_TCP;
4167 return -ESOCKTNOSUPPORT;
4170 sk->sk_gso_type = SKB_GSO_SCTP;
4172 /* Initialize default send parameters. These parameters can be
4173 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4175 sp->default_stream = 0;
4176 sp->default_ppid = 0;
4177 sp->default_flags = 0;
4178 sp->default_context = 0;
4179 sp->default_timetolive = 0;
4181 sp->default_rcv_context = 0;
4182 sp->max_burst = net->sctp.max_burst;
4184 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4186 /* Initialize default setup parameters. These parameters
4187 * can be modified with the SCTP_INITMSG socket option or
4188 * overridden by the SCTP_INIT CMSG.
4190 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4191 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4192 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4193 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4195 /* Initialize default RTO related parameters. These parameters can
4196 * be modified for with the SCTP_RTOINFO socket option.
4198 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4199 sp->rtoinfo.srto_max = net->sctp.rto_max;
4200 sp->rtoinfo.srto_min = net->sctp.rto_min;
4202 /* Initialize default association related parameters. These parameters
4203 * can be modified with the SCTP_ASSOCINFO socket option.
4205 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4206 sp->assocparams.sasoc_number_peer_destinations = 0;
4207 sp->assocparams.sasoc_peer_rwnd = 0;
4208 sp->assocparams.sasoc_local_rwnd = 0;
4209 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4211 /* Initialize default event subscriptions. By default, all the
4214 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4216 /* Default Peer Address Parameters. These defaults can
4217 * be modified via SCTP_PEER_ADDR_PARAMS
4219 sp->hbinterval = net->sctp.hb_interval;
4220 sp->pathmaxrxt = net->sctp.max_retrans_path;
4221 sp->pathmtu = 0; /* allow default discovery */
4222 sp->sackdelay = net->sctp.sack_timeout;
4224 sp->param_flags = SPP_HB_ENABLE |
4226 SPP_SACKDELAY_ENABLE;
4228 /* If enabled no SCTP message fragmentation will be performed.
4229 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4231 sp->disable_fragments = 0;
4233 /* Enable Nagle algorithm by default. */
4236 sp->recvrcvinfo = 0;
4237 sp->recvnxtinfo = 0;
4239 /* Enable by default. */
4242 /* Auto-close idle associations after the configured
4243 * number of seconds. A value of 0 disables this
4244 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4245 * for UDP-style sockets only.
4249 /* User specified fragmentation limit. */
4252 sp->adaptation_ind = 0;
4254 sp->pf = sctp_get_pf_specific(sk->sk_family);
4256 /* Control variables for partial data delivery. */
4257 atomic_set(&sp->pd_mode, 0);
4258 skb_queue_head_init(&sp->pd_lobby);
4259 sp->frag_interleave = 0;
4261 /* Create a per socket endpoint structure. Even if we
4262 * change the data structure relationships, this may still
4263 * be useful for storing pre-connect address information.
4265 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4271 sk->sk_destruct = sctp_destruct_sock;
4273 SCTP_DBG_OBJCNT_INC(sock);
4276 sk_sockets_allocated_inc(sk);
4277 sock_prot_inuse_add(net, sk->sk_prot, 1);
4284 /* Cleanup any SCTP per socket resources. Must be called with
4285 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4287 static void sctp_destroy_sock(struct sock *sk)
4289 struct sctp_sock *sp;
4291 pr_debug("%s: sk:%p\n", __func__, sk);
4293 /* Release our hold on the endpoint. */
4295 /* This could happen during socket init, thus we bail out
4296 * early, since the rest of the below is not setup either.
4301 if (sp->do_auto_asconf) {
4302 sp->do_auto_asconf = 0;
4303 list_del(&sp->auto_asconf_list);
4305 sctp_endpoint_free(sp->ep);
4307 sk_sockets_allocated_dec(sk);
4308 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4312 /* Triggered when there are no references on the socket anymore */
4313 static void sctp_destruct_sock(struct sock *sk)
4315 struct sctp_sock *sp = sctp_sk(sk);
4317 /* Free up the HMAC transform. */
4318 crypto_free_shash(sp->hmac);
4320 inet_sock_destruct(sk);
4323 /* API 4.1.7 shutdown() - TCP Style Syntax
4324 * int shutdown(int socket, int how);
4326 * sd - the socket descriptor of the association to be closed.
4327 * how - Specifies the type of shutdown. The values are
4330 * Disables further receive operations. No SCTP
4331 * protocol action is taken.
4333 * Disables further send operations, and initiates
4334 * the SCTP shutdown sequence.
4336 * Disables further send and receive operations
4337 * and initiates the SCTP shutdown sequence.
4339 static void sctp_shutdown(struct sock *sk, int how)
4341 struct net *net = sock_net(sk);
4342 struct sctp_endpoint *ep;
4344 if (!sctp_style(sk, TCP))
4347 ep = sctp_sk(sk)->ep;
4348 if (how & SEND_SHUTDOWN && !list_empty(&ep->asocs)) {
4349 struct sctp_association *asoc;
4351 sk->sk_state = SCTP_SS_CLOSING;
4352 asoc = list_entry(ep->asocs.next,
4353 struct sctp_association, asocs);
4354 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4358 int sctp_get_sctp_info(struct sock *sk, struct sctp_association *asoc,
4359 struct sctp_info *info)
4361 struct sctp_transport *prim;
4362 struct list_head *pos;
4365 memset(info, 0, sizeof(*info));
4367 struct sctp_sock *sp = sctp_sk(sk);
4369 info->sctpi_s_autoclose = sp->autoclose;
4370 info->sctpi_s_adaptation_ind = sp->adaptation_ind;
4371 info->sctpi_s_pd_point = sp->pd_point;
4372 info->sctpi_s_nodelay = sp->nodelay;
4373 info->sctpi_s_disable_fragments = sp->disable_fragments;
4374 info->sctpi_s_v4mapped = sp->v4mapped;
4375 info->sctpi_s_frag_interleave = sp->frag_interleave;
4376 info->sctpi_s_type = sp->type;
4381 info->sctpi_tag = asoc->c.my_vtag;
4382 info->sctpi_state = asoc->state;
4383 info->sctpi_rwnd = asoc->a_rwnd;
4384 info->sctpi_unackdata = asoc->unack_data;
4385 info->sctpi_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4386 info->sctpi_instrms = asoc->c.sinit_max_instreams;
4387 info->sctpi_outstrms = asoc->c.sinit_num_ostreams;
4388 list_for_each(pos, &asoc->base.inqueue.in_chunk_list)
4389 info->sctpi_inqueue++;
4390 list_for_each(pos, &asoc->outqueue.out_chunk_list)
4391 info->sctpi_outqueue++;
4392 info->sctpi_overall_error = asoc->overall_error_count;
4393 info->sctpi_max_burst = asoc->max_burst;
4394 info->sctpi_maxseg = asoc->frag_point;
4395 info->sctpi_peer_rwnd = asoc->peer.rwnd;
4396 info->sctpi_peer_tag = asoc->c.peer_vtag;
4398 mask = asoc->peer.ecn_capable << 1;
4399 mask = (mask | asoc->peer.ipv4_address) << 1;
4400 mask = (mask | asoc->peer.ipv6_address) << 1;
4401 mask = (mask | asoc->peer.hostname_address) << 1;
4402 mask = (mask | asoc->peer.asconf_capable) << 1;
4403 mask = (mask | asoc->peer.prsctp_capable) << 1;
4404 mask = (mask | asoc->peer.auth_capable);
4405 info->sctpi_peer_capable = mask;
4406 mask = asoc->peer.sack_needed << 1;
4407 mask = (mask | asoc->peer.sack_generation) << 1;
4408 mask = (mask | asoc->peer.zero_window_announced);
4409 info->sctpi_peer_sack = mask;
4411 info->sctpi_isacks = asoc->stats.isacks;
4412 info->sctpi_osacks = asoc->stats.osacks;
4413 info->sctpi_opackets = asoc->stats.opackets;
4414 info->sctpi_ipackets = asoc->stats.ipackets;
4415 info->sctpi_rtxchunks = asoc->stats.rtxchunks;
4416 info->sctpi_outofseqtsns = asoc->stats.outofseqtsns;
4417 info->sctpi_idupchunks = asoc->stats.idupchunks;
4418 info->sctpi_gapcnt = asoc->stats.gapcnt;
4419 info->sctpi_ouodchunks = asoc->stats.ouodchunks;
4420 info->sctpi_iuodchunks = asoc->stats.iuodchunks;
4421 info->sctpi_oodchunks = asoc->stats.oodchunks;
4422 info->sctpi_iodchunks = asoc->stats.iodchunks;
4423 info->sctpi_octrlchunks = asoc->stats.octrlchunks;
4424 info->sctpi_ictrlchunks = asoc->stats.ictrlchunks;
4426 prim = asoc->peer.primary_path;
4427 memcpy(&info->sctpi_p_address, &prim->ipaddr, sizeof(prim->ipaddr));
4428 info->sctpi_p_state = prim->state;
4429 info->sctpi_p_cwnd = prim->cwnd;
4430 info->sctpi_p_srtt = prim->srtt;
4431 info->sctpi_p_rto = jiffies_to_msecs(prim->rto);
4432 info->sctpi_p_hbinterval = prim->hbinterval;
4433 info->sctpi_p_pathmaxrxt = prim->pathmaxrxt;
4434 info->sctpi_p_sackdelay = jiffies_to_msecs(prim->sackdelay);
4435 info->sctpi_p_ssthresh = prim->ssthresh;
4436 info->sctpi_p_partial_bytes_acked = prim->partial_bytes_acked;
4437 info->sctpi_p_flight_size = prim->flight_size;
4438 info->sctpi_p_error = prim->error_count;
4442 EXPORT_SYMBOL_GPL(sctp_get_sctp_info);
4444 /* use callback to avoid exporting the core structure */
4445 int sctp_transport_walk_start(struct rhashtable_iter *iter)
4449 err = rhashtable_walk_init(&sctp_transport_hashtable, iter,
4454 err = rhashtable_walk_start(iter);
4455 if (err && err != -EAGAIN) {
4456 rhashtable_walk_stop(iter);
4457 rhashtable_walk_exit(iter);
4464 void sctp_transport_walk_stop(struct rhashtable_iter *iter)
4466 rhashtable_walk_stop(iter);
4467 rhashtable_walk_exit(iter);
4470 struct sctp_transport *sctp_transport_get_next(struct net *net,
4471 struct rhashtable_iter *iter)
4473 struct sctp_transport *t;
4475 t = rhashtable_walk_next(iter);
4476 for (; t; t = rhashtable_walk_next(iter)) {
4478 if (PTR_ERR(t) == -EAGAIN)
4483 if (!sctp_transport_hold(t))
4486 if (net_eq(sock_net(t->asoc->base.sk), net) &&
4487 t->asoc->peer.primary_path == t)
4490 sctp_transport_put(t);
4496 struct sctp_transport *sctp_transport_get_idx(struct net *net,
4497 struct rhashtable_iter *iter,
4500 struct sctp_transport *t;
4503 return SEQ_START_TOKEN;
4505 while ((t = sctp_transport_get_next(net, iter)) && !IS_ERR(t)) {
4508 sctp_transport_put(t);
4514 int sctp_for_each_endpoint(int (*cb)(struct sctp_endpoint *, void *),
4518 struct sctp_ep_common *epb;
4519 struct sctp_hashbucket *head;
4521 for (head = sctp_ep_hashtable; hash < sctp_ep_hashsize;
4523 read_lock_bh(&head->lock);
4524 sctp_for_each_hentry(epb, &head->chain) {
4525 err = cb(sctp_ep(epb), p);
4529 read_unlock_bh(&head->lock);
4534 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint);
4536 int sctp_transport_lookup_process(int (*cb)(struct sctp_transport *, void *),
4538 const union sctp_addr *laddr,
4539 const union sctp_addr *paddr, void *p)
4541 struct sctp_transport *transport;
4545 transport = sctp_addrs_lookup_transport(net, laddr, paddr);
4546 if (!transport || !sctp_transport_hold(transport)) {
4551 err = cb(transport, p);
4552 sctp_transport_put(transport);
4557 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process);
4559 int sctp_for_each_transport(int (*cb)(struct sctp_transport *, void *),
4560 struct net *net, int pos, void *p) {
4561 struct rhashtable_iter hti;
4565 err = sctp_transport_walk_start(&hti);
4569 obj = sctp_transport_get_idx(net, &hti, pos + 1);
4570 for (; !IS_ERR_OR_NULL(obj); obj = sctp_transport_get_next(net, &hti)) {
4571 struct sctp_transport *transport = obj;
4573 err = cb(transport, p);
4574 sctp_transport_put(transport);
4578 sctp_transport_walk_stop(&hti);
4582 EXPORT_SYMBOL_GPL(sctp_for_each_transport);
4584 /* 7.2.1 Association Status (SCTP_STATUS)
4586 * Applications can retrieve current status information about an
4587 * association, including association state, peer receiver window size,
4588 * number of unacked data chunks, and number of data chunks pending
4589 * receipt. This information is read-only.
4591 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4592 char __user *optval,
4595 struct sctp_status status;
4596 struct sctp_association *asoc = NULL;
4597 struct sctp_transport *transport;
4598 sctp_assoc_t associd;
4601 if (len < sizeof(status)) {
4606 len = sizeof(status);
4607 if (copy_from_user(&status, optval, len)) {
4612 associd = status.sstat_assoc_id;
4613 asoc = sctp_id2assoc(sk, associd);
4619 transport = asoc->peer.primary_path;
4621 status.sstat_assoc_id = sctp_assoc2id(asoc);
4622 status.sstat_state = sctp_assoc_to_state(asoc);
4623 status.sstat_rwnd = asoc->peer.rwnd;
4624 status.sstat_unackdata = asoc->unack_data;
4626 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4627 status.sstat_instrms = asoc->c.sinit_max_instreams;
4628 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4629 status.sstat_fragmentation_point = asoc->frag_point;
4630 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4631 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4632 transport->af_specific->sockaddr_len);
4633 /* Map ipv4 address into v4-mapped-on-v6 address. */
4634 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
4635 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4636 status.sstat_primary.spinfo_state = transport->state;
4637 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4638 status.sstat_primary.spinfo_srtt = transport->srtt;
4639 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4640 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4642 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4643 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4645 if (put_user(len, optlen)) {
4650 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4651 __func__, len, status.sstat_state, status.sstat_rwnd,
4652 status.sstat_assoc_id);
4654 if (copy_to_user(optval, &status, len)) {
4664 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4666 * Applications can retrieve information about a specific peer address
4667 * of an association, including its reachability state, congestion
4668 * window, and retransmission timer values. This information is
4671 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4672 char __user *optval,
4675 struct sctp_paddrinfo pinfo;
4676 struct sctp_transport *transport;
4679 if (len < sizeof(pinfo)) {
4684 len = sizeof(pinfo);
4685 if (copy_from_user(&pinfo, optval, len)) {
4690 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4691 pinfo.spinfo_assoc_id);
4695 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4696 pinfo.spinfo_state = transport->state;
4697 pinfo.spinfo_cwnd = transport->cwnd;
4698 pinfo.spinfo_srtt = transport->srtt;
4699 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4700 pinfo.spinfo_mtu = transport->pathmtu;
4702 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4703 pinfo.spinfo_state = SCTP_ACTIVE;
4705 if (put_user(len, optlen)) {
4710 if (copy_to_user(optval, &pinfo, len)) {
4719 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4721 * This option is a on/off flag. If enabled no SCTP message
4722 * fragmentation will be performed. Instead if a message being sent
4723 * exceeds the current PMTU size, the message will NOT be sent and
4724 * instead a error will be indicated to the user.
4726 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4727 char __user *optval, int __user *optlen)
4731 if (len < sizeof(int))
4735 val = (sctp_sk(sk)->disable_fragments == 1);
4736 if (put_user(len, optlen))
4738 if (copy_to_user(optval, &val, len))
4743 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4745 * This socket option is used to specify various notifications and
4746 * ancillary data the user wishes to receive.
4748 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4753 if (len > sizeof(struct sctp_event_subscribe))
4754 len = sizeof(struct sctp_event_subscribe);
4755 if (put_user(len, optlen))
4757 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4762 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4764 * This socket option is applicable to the UDP-style socket only. When
4765 * set it will cause associations that are idle for more than the
4766 * specified number of seconds to automatically close. An association
4767 * being idle is defined an association that has NOT sent or received
4768 * user data. The special value of '0' indicates that no automatic
4769 * close of any associations should be performed. The option expects an
4770 * integer defining the number of seconds of idle time before an
4771 * association is closed.
4773 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4775 /* Applicable to UDP-style socket only */
4776 if (sctp_style(sk, TCP))
4778 if (len < sizeof(int))
4781 if (put_user(len, optlen))
4783 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, len))
4788 /* Helper routine to branch off an association to a new socket. */
4789 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4791 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4792 struct sctp_sock *sp = sctp_sk(sk);
4793 struct socket *sock;
4796 /* Do not peel off from one netns to another one. */
4797 if (!net_eq(current->nsproxy->net_ns, sock_net(sk)))
4803 /* An association cannot be branched off from an already peeled-off
4804 * socket, nor is this supported for tcp style sockets.
4806 if (!sctp_style(sk, UDP))
4809 /* Create a new socket. */
4810 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4814 sctp_copy_sock(sock->sk, sk, asoc);
4816 /* Make peeled-off sockets more like 1-1 accepted sockets.
4817 * Set the daddr and initialize id to something more random
4819 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
4821 /* Populate the fields of the newsk from the oldsk and migrate the
4822 * asoc to the newsk.
4824 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4830 EXPORT_SYMBOL(sctp_do_peeloff);
4832 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4834 sctp_peeloff_arg_t peeloff;
4835 struct socket *newsock;
4836 struct file *newfile;
4839 if (len < sizeof(sctp_peeloff_arg_t))
4841 len = sizeof(sctp_peeloff_arg_t);
4842 if (copy_from_user(&peeloff, optval, len))
4845 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4849 /* Map the socket to an unused fd that can be returned to the user. */
4850 retval = get_unused_fd_flags(0);
4852 sock_release(newsock);
4856 newfile = sock_alloc_file(newsock, 0, NULL);
4857 if (IS_ERR(newfile)) {
4858 put_unused_fd(retval);
4859 sock_release(newsock);
4860 return PTR_ERR(newfile);
4863 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4866 /* Return the fd mapped to the new socket. */
4867 if (put_user(len, optlen)) {
4869 put_unused_fd(retval);
4872 peeloff.sd = retval;
4873 if (copy_to_user(optval, &peeloff, len)) {
4875 put_unused_fd(retval);
4878 fd_install(retval, newfile);
4883 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4885 * Applications can enable or disable heartbeats for any peer address of
4886 * an association, modify an address's heartbeat interval, force a
4887 * heartbeat to be sent immediately, and adjust the address's maximum
4888 * number of retransmissions sent before an address is considered
4889 * unreachable. The following structure is used to access and modify an
4890 * address's parameters:
4892 * struct sctp_paddrparams {
4893 * sctp_assoc_t spp_assoc_id;
4894 * struct sockaddr_storage spp_address;
4895 * uint32_t spp_hbinterval;
4896 * uint16_t spp_pathmaxrxt;
4897 * uint32_t spp_pathmtu;
4898 * uint32_t spp_sackdelay;
4899 * uint32_t spp_flags;
4902 * spp_assoc_id - (one-to-many style socket) This is filled in the
4903 * application, and identifies the association for
4905 * spp_address - This specifies which address is of interest.
4906 * spp_hbinterval - This contains the value of the heartbeat interval,
4907 * in milliseconds. If a value of zero
4908 * is present in this field then no changes are to
4909 * be made to this parameter.
4910 * spp_pathmaxrxt - This contains the maximum number of
4911 * retransmissions before this address shall be
4912 * considered unreachable. If a value of zero
4913 * is present in this field then no changes are to
4914 * be made to this parameter.
4915 * spp_pathmtu - When Path MTU discovery is disabled the value
4916 * specified here will be the "fixed" path mtu.
4917 * Note that if the spp_address field is empty
4918 * then all associations on this address will
4919 * have this fixed path mtu set upon them.
4921 * spp_sackdelay - When delayed sack is enabled, this value specifies
4922 * the number of milliseconds that sacks will be delayed
4923 * for. This value will apply to all addresses of an
4924 * association if the spp_address field is empty. Note
4925 * also, that if delayed sack is enabled and this
4926 * value is set to 0, no change is made to the last
4927 * recorded delayed sack timer value.
4929 * spp_flags - These flags are used to control various features
4930 * on an association. The flag field may contain
4931 * zero or more of the following options.
4933 * SPP_HB_ENABLE - Enable heartbeats on the
4934 * specified address. Note that if the address
4935 * field is empty all addresses for the association
4936 * have heartbeats enabled upon them.
4938 * SPP_HB_DISABLE - Disable heartbeats on the
4939 * speicifed address. Note that if the address
4940 * field is empty all addresses for the association
4941 * will have their heartbeats disabled. Note also
4942 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4943 * mutually exclusive, only one of these two should
4944 * be specified. Enabling both fields will have
4945 * undetermined results.
4947 * SPP_HB_DEMAND - Request a user initiated heartbeat
4948 * to be made immediately.
4950 * SPP_PMTUD_ENABLE - This field will enable PMTU
4951 * discovery upon the specified address. Note that
4952 * if the address feild is empty then all addresses
4953 * on the association are effected.
4955 * SPP_PMTUD_DISABLE - This field will disable PMTU
4956 * discovery upon the specified address. Note that
4957 * if the address feild is empty then all addresses
4958 * on the association are effected. Not also that
4959 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4960 * exclusive. Enabling both will have undetermined
4963 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4964 * on delayed sack. The time specified in spp_sackdelay
4965 * is used to specify the sack delay for this address. Note
4966 * that if spp_address is empty then all addresses will
4967 * enable delayed sack and take on the sack delay
4968 * value specified in spp_sackdelay.
4969 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4970 * off delayed sack. If the spp_address field is blank then
4971 * delayed sack is disabled for the entire association. Note
4972 * also that this field is mutually exclusive to
4973 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4976 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4977 char __user *optval, int __user *optlen)
4979 struct sctp_paddrparams params;
4980 struct sctp_transport *trans = NULL;
4981 struct sctp_association *asoc = NULL;
4982 struct sctp_sock *sp = sctp_sk(sk);
4984 if (len < sizeof(struct sctp_paddrparams))
4986 len = sizeof(struct sctp_paddrparams);
4987 if (copy_from_user(¶ms, optval, len))
4990 /* If an address other than INADDR_ANY is specified, and
4991 * no transport is found, then the request is invalid.
4993 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
4994 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4995 params.spp_assoc_id);
4997 pr_debug("%s: failed no transport\n", __func__);
5002 /* Get association, if assoc_id != 0 and the socket is a one
5003 * to many style socket, and an association was not found, then
5004 * the id was invalid.
5006 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
5007 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
5008 pr_debug("%s: failed no association\n", __func__);
5013 /* Fetch transport values. */
5014 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
5015 params.spp_pathmtu = trans->pathmtu;
5016 params.spp_pathmaxrxt = trans->pathmaxrxt;
5017 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
5019 /*draft-11 doesn't say what to return in spp_flags*/
5020 params.spp_flags = trans->param_flags;
5022 /* Fetch association values. */
5023 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
5024 params.spp_pathmtu = asoc->pathmtu;
5025 params.spp_pathmaxrxt = asoc->pathmaxrxt;
5026 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
5028 /*draft-11 doesn't say what to return in spp_flags*/
5029 params.spp_flags = asoc->param_flags;
5031 /* Fetch socket values. */
5032 params.spp_hbinterval = sp->hbinterval;
5033 params.spp_pathmtu = sp->pathmtu;
5034 params.spp_sackdelay = sp->sackdelay;
5035 params.spp_pathmaxrxt = sp->pathmaxrxt;
5037 /*draft-11 doesn't say what to return in spp_flags*/
5038 params.spp_flags = sp->param_flags;
5041 if (copy_to_user(optval, ¶ms, len))
5044 if (put_user(len, optlen))
5051 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
5053 * This option will effect the way delayed acks are performed. This
5054 * option allows you to get or set the delayed ack time, in
5055 * milliseconds. It also allows changing the delayed ack frequency.
5056 * Changing the frequency to 1 disables the delayed sack algorithm. If
5057 * the assoc_id is 0, then this sets or gets the endpoints default
5058 * values. If the assoc_id field is non-zero, then the set or get
5059 * effects the specified association for the one to many model (the
5060 * assoc_id field is ignored by the one to one model). Note that if
5061 * sack_delay or sack_freq are 0 when setting this option, then the
5062 * current values will remain unchanged.
5064 * struct sctp_sack_info {
5065 * sctp_assoc_t sack_assoc_id;
5066 * uint32_t sack_delay;
5067 * uint32_t sack_freq;
5070 * sack_assoc_id - This parameter, indicates which association the user
5071 * is performing an action upon. Note that if this field's value is
5072 * zero then the endpoints default value is changed (effecting future
5073 * associations only).
5075 * sack_delay - This parameter contains the number of milliseconds that
5076 * the user is requesting the delayed ACK timer be set to. Note that
5077 * this value is defined in the standard to be between 200 and 500
5080 * sack_freq - This parameter contains the number of packets that must
5081 * be received before a sack is sent without waiting for the delay
5082 * timer to expire. The default value for this is 2, setting this
5083 * value to 1 will disable the delayed sack algorithm.
5085 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
5086 char __user *optval,
5089 struct sctp_sack_info params;
5090 struct sctp_association *asoc = NULL;
5091 struct sctp_sock *sp = sctp_sk(sk);
5093 if (len >= sizeof(struct sctp_sack_info)) {
5094 len = sizeof(struct sctp_sack_info);
5096 if (copy_from_user(¶ms, optval, len))
5098 } else if (len == sizeof(struct sctp_assoc_value)) {
5099 pr_warn_ratelimited(DEPRECATED
5101 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
5102 "Use struct sctp_sack_info instead\n",
5103 current->comm, task_pid_nr(current));
5104 if (copy_from_user(¶ms, optval, len))
5109 /* Get association, if sack_assoc_id != 0 and the socket is a one
5110 * to many style socket, and an association was not found, then
5111 * the id was invalid.
5113 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
5114 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
5118 /* Fetch association values. */
5119 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
5120 params.sack_delay = jiffies_to_msecs(
5122 params.sack_freq = asoc->sackfreq;
5125 params.sack_delay = 0;
5126 params.sack_freq = 1;
5129 /* Fetch socket values. */
5130 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
5131 params.sack_delay = sp->sackdelay;
5132 params.sack_freq = sp->sackfreq;
5134 params.sack_delay = 0;
5135 params.sack_freq = 1;
5139 if (copy_to_user(optval, ¶ms, len))
5142 if (put_user(len, optlen))
5148 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5150 * Applications can specify protocol parameters for the default association
5151 * initialization. The option name argument to setsockopt() and getsockopt()
5154 * Setting initialization parameters is effective only on an unconnected
5155 * socket (for UDP-style sockets only future associations are effected
5156 * by the change). With TCP-style sockets, this option is inherited by
5157 * sockets derived from a listener socket.
5159 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
5161 if (len < sizeof(struct sctp_initmsg))
5163 len = sizeof(struct sctp_initmsg);
5164 if (put_user(len, optlen))
5166 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
5172 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
5173 char __user *optval, int __user *optlen)
5175 struct sctp_association *asoc;
5177 struct sctp_getaddrs getaddrs;
5178 struct sctp_transport *from;
5180 union sctp_addr temp;
5181 struct sctp_sock *sp = sctp_sk(sk);
5186 if (len < sizeof(struct sctp_getaddrs))
5189 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5192 /* For UDP-style sockets, id specifies the association to query. */
5193 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5197 to = optval + offsetof(struct sctp_getaddrs, addrs);
5198 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5200 list_for_each_entry(from, &asoc->peer.transport_addr_list,
5202 memcpy(&temp, &from->ipaddr, sizeof(temp));
5203 addrlen = sctp_get_pf_specific(sk->sk_family)
5204 ->addr_to_user(sp, &temp);
5205 if (space_left < addrlen)
5207 if (copy_to_user(to, &temp, addrlen))
5211 space_left -= addrlen;
5214 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
5216 bytes_copied = ((char __user *)to) - optval;
5217 if (put_user(bytes_copied, optlen))
5223 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
5224 size_t space_left, int *bytes_copied)
5226 struct sctp_sockaddr_entry *addr;
5227 union sctp_addr temp;
5230 struct net *net = sock_net(sk);
5233 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
5237 if ((PF_INET == sk->sk_family) &&
5238 (AF_INET6 == addr->a.sa.sa_family))
5240 if ((PF_INET6 == sk->sk_family) &&
5241 inet_v6_ipv6only(sk) &&
5242 (AF_INET == addr->a.sa.sa_family))
5244 memcpy(&temp, &addr->a, sizeof(temp));
5245 if (!temp.v4.sin_port)
5246 temp.v4.sin_port = htons(port);
5248 addrlen = sctp_get_pf_specific(sk->sk_family)
5249 ->addr_to_user(sctp_sk(sk), &temp);
5251 if (space_left < addrlen) {
5255 memcpy(to, &temp, addrlen);
5259 space_left -= addrlen;
5260 *bytes_copied += addrlen;
5268 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
5269 char __user *optval, int __user *optlen)
5271 struct sctp_bind_addr *bp;
5272 struct sctp_association *asoc;
5274 struct sctp_getaddrs getaddrs;
5275 struct sctp_sockaddr_entry *addr;
5277 union sctp_addr temp;
5278 struct sctp_sock *sp = sctp_sk(sk);
5282 int bytes_copied = 0;
5286 if (len < sizeof(struct sctp_getaddrs))
5289 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5293 * For UDP-style sockets, id specifies the association to query.
5294 * If the id field is set to the value '0' then the locally bound
5295 * addresses are returned without regard to any particular
5298 if (0 == getaddrs.assoc_id) {
5299 bp = &sctp_sk(sk)->ep->base.bind_addr;
5301 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5304 bp = &asoc->base.bind_addr;
5307 to = optval + offsetof(struct sctp_getaddrs, addrs);
5308 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5310 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN);
5314 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5315 * addresses from the global local address list.
5317 if (sctp_list_single_entry(&bp->address_list)) {
5318 addr = list_entry(bp->address_list.next,
5319 struct sctp_sockaddr_entry, list);
5320 if (sctp_is_any(sk, &addr->a)) {
5321 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
5322 space_left, &bytes_copied);
5332 /* Protection on the bound address list is not needed since
5333 * in the socket option context we hold a socket lock and
5334 * thus the bound address list can't change.
5336 list_for_each_entry(addr, &bp->address_list, list) {
5337 memcpy(&temp, &addr->a, sizeof(temp));
5338 addrlen = sctp_get_pf_specific(sk->sk_family)
5339 ->addr_to_user(sp, &temp);
5340 if (space_left < addrlen) {
5341 err = -ENOMEM; /*fixme: right error?*/
5344 memcpy(buf, &temp, addrlen);
5346 bytes_copied += addrlen;
5348 space_left -= addrlen;
5352 if (copy_to_user(to, addrs, bytes_copied)) {
5356 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
5360 /* XXX: We should have accounted for sizeof(struct sctp_getaddrs) too,
5361 * but we can't change it anymore.
5363 if (put_user(bytes_copied, optlen))
5370 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
5372 * Requests that the local SCTP stack use the enclosed peer address as
5373 * the association primary. The enclosed address must be one of the
5374 * association peer's addresses.
5376 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
5377 char __user *optval, int __user *optlen)
5379 struct sctp_prim prim;
5380 struct sctp_association *asoc;
5381 struct sctp_sock *sp = sctp_sk(sk);
5383 if (len < sizeof(struct sctp_prim))
5386 len = sizeof(struct sctp_prim);
5388 if (copy_from_user(&prim, optval, len))
5391 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
5395 if (!asoc->peer.primary_path)
5398 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
5399 asoc->peer.primary_path->af_specific->sockaddr_len);
5401 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
5402 (union sctp_addr *)&prim.ssp_addr);
5404 if (put_user(len, optlen))
5406 if (copy_to_user(optval, &prim, len))
5413 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5415 * Requests that the local endpoint set the specified Adaptation Layer
5416 * Indication parameter for all future INIT and INIT-ACK exchanges.
5418 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
5419 char __user *optval, int __user *optlen)
5421 struct sctp_setadaptation adaptation;
5423 if (len < sizeof(struct sctp_setadaptation))
5426 len = sizeof(struct sctp_setadaptation);
5428 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
5430 if (put_user(len, optlen))
5432 if (copy_to_user(optval, &adaptation, len))
5440 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5442 * Applications that wish to use the sendto() system call may wish to
5443 * specify a default set of parameters that would normally be supplied
5444 * through the inclusion of ancillary data. This socket option allows
5445 * such an application to set the default sctp_sndrcvinfo structure.
5448 * The application that wishes to use this socket option simply passes
5449 * in to this call the sctp_sndrcvinfo structure defined in Section
5450 * 5.2.2) The input parameters accepted by this call include
5451 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5452 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5453 * to this call if the caller is using the UDP model.
5455 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5457 static int sctp_getsockopt_default_send_param(struct sock *sk,
5458 int len, char __user *optval,
5461 struct sctp_sock *sp = sctp_sk(sk);
5462 struct sctp_association *asoc;
5463 struct sctp_sndrcvinfo info;
5465 if (len < sizeof(info))
5470 if (copy_from_user(&info, optval, len))
5473 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
5474 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
5477 info.sinfo_stream = asoc->default_stream;
5478 info.sinfo_flags = asoc->default_flags;
5479 info.sinfo_ppid = asoc->default_ppid;
5480 info.sinfo_context = asoc->default_context;
5481 info.sinfo_timetolive = asoc->default_timetolive;
5483 info.sinfo_stream = sp->default_stream;
5484 info.sinfo_flags = sp->default_flags;
5485 info.sinfo_ppid = sp->default_ppid;
5486 info.sinfo_context = sp->default_context;
5487 info.sinfo_timetolive = sp->default_timetolive;
5490 if (put_user(len, optlen))
5492 if (copy_to_user(optval, &info, len))
5498 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5499 * (SCTP_DEFAULT_SNDINFO)
5501 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
5502 char __user *optval,
5505 struct sctp_sock *sp = sctp_sk(sk);
5506 struct sctp_association *asoc;
5507 struct sctp_sndinfo info;
5509 if (len < sizeof(info))
5514 if (copy_from_user(&info, optval, len))
5517 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
5518 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
5521 info.snd_sid = asoc->default_stream;
5522 info.snd_flags = asoc->default_flags;
5523 info.snd_ppid = asoc->default_ppid;
5524 info.snd_context = asoc->default_context;
5526 info.snd_sid = sp->default_stream;
5527 info.snd_flags = sp->default_flags;
5528 info.snd_ppid = sp->default_ppid;
5529 info.snd_context = sp->default_context;
5532 if (put_user(len, optlen))
5534 if (copy_to_user(optval, &info, len))
5542 * 7.1.5 SCTP_NODELAY
5544 * Turn on/off any Nagle-like algorithm. This means that packets are
5545 * generally sent as soon as possible and no unnecessary delays are
5546 * introduced, at the cost of more packets in the network. Expects an
5547 * integer boolean flag.
5550 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5551 char __user *optval, int __user *optlen)
5555 if (len < sizeof(int))
5559 val = (sctp_sk(sk)->nodelay == 1);
5560 if (put_user(len, optlen))
5562 if (copy_to_user(optval, &val, len))
5569 * 7.1.1 SCTP_RTOINFO
5571 * The protocol parameters used to initialize and bound retransmission
5572 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5573 * and modify these parameters.
5574 * All parameters are time values, in milliseconds. A value of 0, when
5575 * modifying the parameters, indicates that the current value should not
5579 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5580 char __user *optval,
5581 int __user *optlen) {
5582 struct sctp_rtoinfo rtoinfo;
5583 struct sctp_association *asoc;
5585 if (len < sizeof (struct sctp_rtoinfo))
5588 len = sizeof(struct sctp_rtoinfo);
5590 if (copy_from_user(&rtoinfo, optval, len))
5593 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5595 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5598 /* Values corresponding to the specific association. */
5600 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5601 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5602 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5604 /* Values corresponding to the endpoint. */
5605 struct sctp_sock *sp = sctp_sk(sk);
5607 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5608 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5609 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5612 if (put_user(len, optlen))
5615 if (copy_to_user(optval, &rtoinfo, len))
5623 * 7.1.2 SCTP_ASSOCINFO
5625 * This option is used to tune the maximum retransmission attempts
5626 * of the association.
5627 * Returns an error if the new association retransmission value is
5628 * greater than the sum of the retransmission value of the peer.
5629 * See [SCTP] for more information.
5632 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5633 char __user *optval,
5637 struct sctp_assocparams assocparams;
5638 struct sctp_association *asoc;
5639 struct list_head *pos;
5642 if (len < sizeof (struct sctp_assocparams))
5645 len = sizeof(struct sctp_assocparams);
5647 if (copy_from_user(&assocparams, optval, len))
5650 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5652 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5655 /* Values correspoinding to the specific association */
5657 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5658 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5659 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5660 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5662 list_for_each(pos, &asoc->peer.transport_addr_list) {
5666 assocparams.sasoc_number_peer_destinations = cnt;
5668 /* Values corresponding to the endpoint */
5669 struct sctp_sock *sp = sctp_sk(sk);
5671 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5672 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5673 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5674 assocparams.sasoc_cookie_life =
5675 sp->assocparams.sasoc_cookie_life;
5676 assocparams.sasoc_number_peer_destinations =
5678 sasoc_number_peer_destinations;
5681 if (put_user(len, optlen))
5684 if (copy_to_user(optval, &assocparams, len))
5691 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5693 * This socket option is a boolean flag which turns on or off mapped V4
5694 * addresses. If this option is turned on and the socket is type
5695 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5696 * If this option is turned off, then no mapping will be done of V4
5697 * addresses and a user will receive both PF_INET6 and PF_INET type
5698 * addresses on the socket.
5700 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5701 char __user *optval, int __user *optlen)
5704 struct sctp_sock *sp = sctp_sk(sk);
5706 if (len < sizeof(int))
5711 if (put_user(len, optlen))
5713 if (copy_to_user(optval, &val, len))
5720 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5721 * (chapter and verse is quoted at sctp_setsockopt_context())
5723 static int sctp_getsockopt_context(struct sock *sk, int len,
5724 char __user *optval, int __user *optlen)
5726 struct sctp_assoc_value params;
5727 struct sctp_sock *sp;
5728 struct sctp_association *asoc;
5730 if (len < sizeof(struct sctp_assoc_value))
5733 len = sizeof(struct sctp_assoc_value);
5735 if (copy_from_user(¶ms, optval, len))
5740 if (params.assoc_id != 0) {
5741 asoc = sctp_id2assoc(sk, params.assoc_id);
5744 params.assoc_value = asoc->default_rcv_context;
5746 params.assoc_value = sp->default_rcv_context;
5749 if (put_user(len, optlen))
5751 if (copy_to_user(optval, ¶ms, len))
5758 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5759 * This option will get or set the maximum size to put in any outgoing
5760 * SCTP DATA chunk. If a message is larger than this size it will be
5761 * fragmented by SCTP into the specified size. Note that the underlying
5762 * SCTP implementation may fragment into smaller sized chunks when the
5763 * PMTU of the underlying association is smaller than the value set by
5764 * the user. The default value for this option is '0' which indicates
5765 * the user is NOT limiting fragmentation and only the PMTU will effect
5766 * SCTP's choice of DATA chunk size. Note also that values set larger
5767 * than the maximum size of an IP datagram will effectively let SCTP
5768 * control fragmentation (i.e. the same as setting this option to 0).
5770 * The following structure is used to access and modify this parameter:
5772 * struct sctp_assoc_value {
5773 * sctp_assoc_t assoc_id;
5774 * uint32_t assoc_value;
5777 * assoc_id: This parameter is ignored for one-to-one style sockets.
5778 * For one-to-many style sockets this parameter indicates which
5779 * association the user is performing an action upon. Note that if
5780 * this field's value is zero then the endpoints default value is
5781 * changed (effecting future associations only).
5782 * assoc_value: This parameter specifies the maximum size in bytes.
5784 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5785 char __user *optval, int __user *optlen)
5787 struct sctp_assoc_value params;
5788 struct sctp_association *asoc;
5790 if (len == sizeof(int)) {
5791 pr_warn_ratelimited(DEPRECATED
5793 "Use of int in maxseg socket option.\n"
5794 "Use struct sctp_assoc_value instead\n",
5795 current->comm, task_pid_nr(current));
5796 params.assoc_id = 0;
5797 } else if (len >= sizeof(struct sctp_assoc_value)) {
5798 len = sizeof(struct sctp_assoc_value);
5799 if (copy_from_user(¶ms, optval, len))
5804 asoc = sctp_id2assoc(sk, params.assoc_id);
5805 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5809 params.assoc_value = asoc->frag_point;
5811 params.assoc_value = sctp_sk(sk)->user_frag;
5813 if (put_user(len, optlen))
5815 if (len == sizeof(int)) {
5816 if (copy_to_user(optval, ¶ms.assoc_value, len))
5819 if (copy_to_user(optval, ¶ms, len))
5827 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5828 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5830 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5831 char __user *optval, int __user *optlen)
5835 if (len < sizeof(int))
5840 val = sctp_sk(sk)->frag_interleave;
5841 if (put_user(len, optlen))
5843 if (copy_to_user(optval, &val, len))
5850 * 7.1.25. Set or Get the sctp partial delivery point
5851 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5853 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5854 char __user *optval,
5859 if (len < sizeof(u32))
5864 val = sctp_sk(sk)->pd_point;
5865 if (put_user(len, optlen))
5867 if (copy_to_user(optval, &val, len))
5874 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5875 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5877 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5878 char __user *optval,
5881 struct sctp_assoc_value params;
5882 struct sctp_sock *sp;
5883 struct sctp_association *asoc;
5885 if (len == sizeof(int)) {
5886 pr_warn_ratelimited(DEPRECATED
5888 "Use of int in max_burst socket option.\n"
5889 "Use struct sctp_assoc_value instead\n",
5890 current->comm, task_pid_nr(current));
5891 params.assoc_id = 0;
5892 } else if (len >= sizeof(struct sctp_assoc_value)) {
5893 len = sizeof(struct sctp_assoc_value);
5894 if (copy_from_user(¶ms, optval, len))
5901 if (params.assoc_id != 0) {
5902 asoc = sctp_id2assoc(sk, params.assoc_id);
5905 params.assoc_value = asoc->max_burst;
5907 params.assoc_value = sp->max_burst;
5909 if (len == sizeof(int)) {
5910 if (copy_to_user(optval, ¶ms.assoc_value, len))
5913 if (copy_to_user(optval, ¶ms, len))
5921 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5922 char __user *optval, int __user *optlen)
5924 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5925 struct sctp_hmacalgo __user *p = (void __user *)optval;
5926 struct sctp_hmac_algo_param *hmacs;
5931 if (!ep->auth_enable)
5934 hmacs = ep->auth_hmacs_list;
5935 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5937 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5940 len = sizeof(struct sctp_hmacalgo) + data_len;
5941 num_idents = data_len / sizeof(u16);
5943 if (put_user(len, optlen))
5945 if (put_user(num_idents, &p->shmac_num_idents))
5947 for (i = 0; i < num_idents; i++) {
5948 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
5950 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
5956 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5957 char __user *optval, int __user *optlen)
5959 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5960 struct sctp_authkeyid val;
5961 struct sctp_association *asoc;
5963 if (!ep->auth_enable)
5966 if (len < sizeof(struct sctp_authkeyid))
5969 len = sizeof(struct sctp_authkeyid);
5970 if (copy_from_user(&val, optval, len))
5973 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5974 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5978 val.scact_keynumber = asoc->active_key_id;
5980 val.scact_keynumber = ep->active_key_id;
5982 if (put_user(len, optlen))
5984 if (copy_to_user(optval, &val, len))
5990 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5991 char __user *optval, int __user *optlen)
5993 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5994 struct sctp_authchunks __user *p = (void __user *)optval;
5995 struct sctp_authchunks val;
5996 struct sctp_association *asoc;
5997 struct sctp_chunks_param *ch;
6001 if (!ep->auth_enable)
6004 if (len < sizeof(struct sctp_authchunks))
6007 if (copy_from_user(&val, optval, sizeof(val)))
6010 to = p->gauth_chunks;
6011 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6015 ch = asoc->peer.peer_chunks;
6019 /* See if the user provided enough room for all the data */
6020 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
6021 if (len < num_chunks)
6024 if (copy_to_user(to, ch->chunks, num_chunks))
6027 len = sizeof(struct sctp_authchunks) + num_chunks;
6028 if (put_user(len, optlen))
6030 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6035 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
6036 char __user *optval, int __user *optlen)
6038 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6039 struct sctp_authchunks __user *p = (void __user *)optval;
6040 struct sctp_authchunks val;
6041 struct sctp_association *asoc;
6042 struct sctp_chunks_param *ch;
6046 if (!ep->auth_enable)
6049 if (len < sizeof(struct sctp_authchunks))
6052 if (copy_from_user(&val, optval, sizeof(val)))
6055 to = p->gauth_chunks;
6056 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6057 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
6061 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
6063 ch = ep->auth_chunk_list;
6068 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
6069 if (len < sizeof(struct sctp_authchunks) + num_chunks)
6072 if (copy_to_user(to, ch->chunks, num_chunks))
6075 len = sizeof(struct sctp_authchunks) + num_chunks;
6076 if (put_user(len, optlen))
6078 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6085 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
6086 * This option gets the current number of associations that are attached
6087 * to a one-to-many style socket. The option value is an uint32_t.
6089 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
6090 char __user *optval, int __user *optlen)
6092 struct sctp_sock *sp = sctp_sk(sk);
6093 struct sctp_association *asoc;
6096 if (sctp_style(sk, TCP))
6099 if (len < sizeof(u32))
6104 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6108 if (put_user(len, optlen))
6110 if (copy_to_user(optval, &val, len))
6117 * 8.1.23 SCTP_AUTO_ASCONF
6118 * See the corresponding setsockopt entry as description
6120 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
6121 char __user *optval, int __user *optlen)
6125 if (len < sizeof(int))
6129 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
6131 if (put_user(len, optlen))
6133 if (copy_to_user(optval, &val, len))
6139 * 8.2.6. Get the Current Identifiers of Associations
6140 * (SCTP_GET_ASSOC_ID_LIST)
6142 * This option gets the current list of SCTP association identifiers of
6143 * the SCTP associations handled by a one-to-many style socket.
6145 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
6146 char __user *optval, int __user *optlen)
6148 struct sctp_sock *sp = sctp_sk(sk);
6149 struct sctp_association *asoc;
6150 struct sctp_assoc_ids *ids;
6153 if (sctp_style(sk, TCP))
6156 if (len < sizeof(struct sctp_assoc_ids))
6159 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6163 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
6166 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
6168 ids = kmalloc(len, GFP_USER | __GFP_NOWARN);
6172 ids->gaids_number_of_ids = num;
6174 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6175 ids->gaids_assoc_id[num++] = asoc->assoc_id;
6178 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
6188 * SCTP_PEER_ADDR_THLDS
6190 * This option allows us to fetch the partially failed threshold for one or all
6191 * transports in an association. See Section 6.1 of:
6192 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6194 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
6195 char __user *optval,
6199 struct sctp_paddrthlds val;
6200 struct sctp_transport *trans;
6201 struct sctp_association *asoc;
6203 if (len < sizeof(struct sctp_paddrthlds))
6205 len = sizeof(struct sctp_paddrthlds);
6206 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
6209 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
6210 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
6214 val.spt_pathpfthld = asoc->pf_retrans;
6215 val.spt_pathmaxrxt = asoc->pathmaxrxt;
6217 trans = sctp_addr_id2transport(sk, &val.spt_address,
6222 val.spt_pathmaxrxt = trans->pathmaxrxt;
6223 val.spt_pathpfthld = trans->pf_retrans;
6226 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
6233 * SCTP_GET_ASSOC_STATS
6235 * This option retrieves local per endpoint statistics. It is modeled
6236 * after OpenSolaris' implementation
6238 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
6239 char __user *optval,
6242 struct sctp_assoc_stats sas;
6243 struct sctp_association *asoc = NULL;
6245 /* User must provide at least the assoc id */
6246 if (len < sizeof(sctp_assoc_t))
6249 /* Allow the struct to grow and fill in as much as possible */
6250 len = min_t(size_t, len, sizeof(sas));
6252 if (copy_from_user(&sas, optval, len))
6255 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
6259 sas.sas_rtxchunks = asoc->stats.rtxchunks;
6260 sas.sas_gapcnt = asoc->stats.gapcnt;
6261 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
6262 sas.sas_osacks = asoc->stats.osacks;
6263 sas.sas_isacks = asoc->stats.isacks;
6264 sas.sas_octrlchunks = asoc->stats.octrlchunks;
6265 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
6266 sas.sas_oodchunks = asoc->stats.oodchunks;
6267 sas.sas_iodchunks = asoc->stats.iodchunks;
6268 sas.sas_ouodchunks = asoc->stats.ouodchunks;
6269 sas.sas_iuodchunks = asoc->stats.iuodchunks;
6270 sas.sas_idupchunks = asoc->stats.idupchunks;
6271 sas.sas_opackets = asoc->stats.opackets;
6272 sas.sas_ipackets = asoc->stats.ipackets;
6274 /* New high max rto observed, will return 0 if not a single
6275 * RTO update took place. obs_rto_ipaddr will be bogus
6278 sas.sas_maxrto = asoc->stats.max_obs_rto;
6279 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
6280 sizeof(struct sockaddr_storage));
6282 /* Mark beginning of a new observation period */
6283 asoc->stats.max_obs_rto = asoc->rto_min;
6285 if (put_user(len, optlen))
6288 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
6290 if (copy_to_user(optval, &sas, len))
6296 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
6297 char __user *optval,
6302 if (len < sizeof(int))
6306 if (sctp_sk(sk)->recvrcvinfo)
6308 if (put_user(len, optlen))
6310 if (copy_to_user(optval, &val, len))
6316 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
6317 char __user *optval,
6322 if (len < sizeof(int))
6326 if (sctp_sk(sk)->recvnxtinfo)
6328 if (put_user(len, optlen))
6330 if (copy_to_user(optval, &val, len))
6336 static int sctp_getsockopt_pr_supported(struct sock *sk, int len,
6337 char __user *optval,
6340 struct sctp_assoc_value params;
6341 struct sctp_association *asoc;
6342 int retval = -EFAULT;
6344 if (len < sizeof(params)) {
6349 len = sizeof(params);
6350 if (copy_from_user(¶ms, optval, len))
6353 asoc = sctp_id2assoc(sk, params.assoc_id);
6355 params.assoc_value = asoc->prsctp_enable;
6356 } else if (!params.assoc_id) {
6357 struct sctp_sock *sp = sctp_sk(sk);
6359 params.assoc_value = sp->ep->prsctp_enable;
6365 if (put_user(len, optlen))
6368 if (copy_to_user(optval, ¶ms, len))
6377 static int sctp_getsockopt_default_prinfo(struct sock *sk, int len,
6378 char __user *optval,
6381 struct sctp_default_prinfo info;
6382 struct sctp_association *asoc;
6383 int retval = -EFAULT;
6385 if (len < sizeof(info)) {
6391 if (copy_from_user(&info, optval, len))
6394 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
6396 info.pr_policy = SCTP_PR_POLICY(asoc->default_flags);
6397 info.pr_value = asoc->default_timetolive;
6398 } else if (!info.pr_assoc_id) {
6399 struct sctp_sock *sp = sctp_sk(sk);
6401 info.pr_policy = SCTP_PR_POLICY(sp->default_flags);
6402 info.pr_value = sp->default_timetolive;
6408 if (put_user(len, optlen))
6411 if (copy_to_user(optval, &info, len))
6420 static int sctp_getsockopt_pr_assocstatus(struct sock *sk, int len,
6421 char __user *optval,
6424 struct sctp_prstatus params;
6425 struct sctp_association *asoc;
6427 int retval = -EINVAL;
6429 if (len < sizeof(params))
6432 len = sizeof(params);
6433 if (copy_from_user(¶ms, optval, len)) {
6438 policy = params.sprstat_policy;
6439 if (policy & ~SCTP_PR_SCTP_MASK)
6442 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
6446 if (policy == SCTP_PR_SCTP_NONE) {
6447 params.sprstat_abandoned_unsent = 0;
6448 params.sprstat_abandoned_sent = 0;
6449 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
6450 params.sprstat_abandoned_unsent +=
6451 asoc->abandoned_unsent[policy];
6452 params.sprstat_abandoned_sent +=
6453 asoc->abandoned_sent[policy];
6456 params.sprstat_abandoned_unsent =
6457 asoc->abandoned_unsent[__SCTP_PR_INDEX(policy)];
6458 params.sprstat_abandoned_sent =
6459 asoc->abandoned_sent[__SCTP_PR_INDEX(policy)];
6462 if (put_user(len, optlen)) {
6467 if (copy_to_user(optval, ¶ms, len)) {
6478 static int sctp_getsockopt(struct sock *sk, int level, int optname,
6479 char __user *optval, int __user *optlen)
6484 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
6486 /* I can hardly begin to describe how wrong this is. This is
6487 * so broken as to be worse than useless. The API draft
6488 * REALLY is NOT helpful here... I am not convinced that the
6489 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
6490 * are at all well-founded.
6492 if (level != SOL_SCTP) {
6493 struct sctp_af *af = sctp_sk(sk)->pf->af;
6495 retval = af->getsockopt(sk, level, optname, optval, optlen);
6499 if (get_user(len, optlen))
6509 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
6511 case SCTP_DISABLE_FRAGMENTS:
6512 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
6516 retval = sctp_getsockopt_events(sk, len, optval, optlen);
6518 case SCTP_AUTOCLOSE:
6519 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
6521 case SCTP_SOCKOPT_PEELOFF:
6522 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
6524 case SCTP_PEER_ADDR_PARAMS:
6525 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
6528 case SCTP_DELAYED_SACK:
6529 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
6533 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
6535 case SCTP_GET_PEER_ADDRS:
6536 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
6539 case SCTP_GET_LOCAL_ADDRS:
6540 retval = sctp_getsockopt_local_addrs(sk, len, optval,
6543 case SCTP_SOCKOPT_CONNECTX3:
6544 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
6546 case SCTP_DEFAULT_SEND_PARAM:
6547 retval = sctp_getsockopt_default_send_param(sk, len,
6550 case SCTP_DEFAULT_SNDINFO:
6551 retval = sctp_getsockopt_default_sndinfo(sk, len,
6554 case SCTP_PRIMARY_ADDR:
6555 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
6558 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
6561 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
6563 case SCTP_ASSOCINFO:
6564 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
6566 case SCTP_I_WANT_MAPPED_V4_ADDR:
6567 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
6570 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
6572 case SCTP_GET_PEER_ADDR_INFO:
6573 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
6576 case SCTP_ADAPTATION_LAYER:
6577 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
6581 retval = sctp_getsockopt_context(sk, len, optval, optlen);
6583 case SCTP_FRAGMENT_INTERLEAVE:
6584 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
6587 case SCTP_PARTIAL_DELIVERY_POINT:
6588 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
6591 case SCTP_MAX_BURST:
6592 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
6595 case SCTP_AUTH_CHUNK:
6596 case SCTP_AUTH_DELETE_KEY:
6597 retval = -EOPNOTSUPP;
6599 case SCTP_HMAC_IDENT:
6600 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
6602 case SCTP_AUTH_ACTIVE_KEY:
6603 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
6605 case SCTP_PEER_AUTH_CHUNKS:
6606 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
6609 case SCTP_LOCAL_AUTH_CHUNKS:
6610 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
6613 case SCTP_GET_ASSOC_NUMBER:
6614 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
6616 case SCTP_GET_ASSOC_ID_LIST:
6617 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
6619 case SCTP_AUTO_ASCONF:
6620 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
6622 case SCTP_PEER_ADDR_THLDS:
6623 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
6625 case SCTP_GET_ASSOC_STATS:
6626 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
6628 case SCTP_RECVRCVINFO:
6629 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
6631 case SCTP_RECVNXTINFO:
6632 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
6634 case SCTP_PR_SUPPORTED:
6635 retval = sctp_getsockopt_pr_supported(sk, len, optval, optlen);
6637 case SCTP_DEFAULT_PRINFO:
6638 retval = sctp_getsockopt_default_prinfo(sk, len, optval,
6641 case SCTP_PR_ASSOC_STATUS:
6642 retval = sctp_getsockopt_pr_assocstatus(sk, len, optval,
6646 retval = -ENOPROTOOPT;
6654 static int sctp_hash(struct sock *sk)
6660 static void sctp_unhash(struct sock *sk)
6665 /* Check if port is acceptable. Possibly find first available port.
6667 * The port hash table (contained in the 'global' SCTP protocol storage
6668 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
6669 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
6670 * list (the list number is the port number hashed out, so as you
6671 * would expect from a hash function, all the ports in a given list have
6672 * such a number that hashes out to the same list number; you were
6673 * expecting that, right?); so each list has a set of ports, with a
6674 * link to the socket (struct sock) that uses it, the port number and
6675 * a fastreuse flag (FIXME: NPI ipg).
6677 static struct sctp_bind_bucket *sctp_bucket_create(
6678 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
6680 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
6682 struct sctp_bind_hashbucket *head; /* hash list */
6683 struct sctp_bind_bucket *pp;
6684 unsigned short snum;
6687 snum = ntohs(addr->v4.sin_port);
6689 pr_debug("%s: begins, snum:%d\n", __func__, snum);
6692 /* Search for an available port. */
6693 int low, high, remaining, index;
6695 struct net *net = sock_net(sk);
6697 inet_get_local_port_range(net, &low, &high);
6698 remaining = (high - low) + 1;
6699 rover = prandom_u32() % remaining + low;
6703 if ((rover < low) || (rover > high))
6705 if (inet_is_local_reserved_port(net, rover))
6707 index = sctp_phashfn(sock_net(sk), rover);
6708 head = &sctp_port_hashtable[index];
6709 spin_lock_bh(&head->lock);
6710 sctp_for_each_hentry(pp, &head->chain)
6711 if ((pp->port == rover) &&
6712 net_eq(sock_net(sk), pp->net))
6716 spin_unlock_bh(&head->lock);
6718 } while (--remaining > 0);
6720 /* Exhausted local port range during search? */
6725 /* OK, here is the one we will use. HEAD (the port
6726 * hash table list entry) is non-NULL and we hold it's
6731 /* We are given an specific port number; we verify
6732 * that it is not being used. If it is used, we will
6733 * exahust the search in the hash list corresponding
6734 * to the port number (snum) - we detect that with the
6735 * port iterator, pp being NULL.
6737 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
6738 spin_lock_bh(&head->lock);
6739 sctp_for_each_hentry(pp, &head->chain) {
6740 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
6747 if (!hlist_empty(&pp->owner)) {
6748 /* We had a port hash table hit - there is an
6749 * available port (pp != NULL) and it is being
6750 * used by other socket (pp->owner not empty); that other
6751 * socket is going to be sk2.
6753 int reuse = sk->sk_reuse;
6756 pr_debug("%s: found a possible match\n", __func__);
6758 if (pp->fastreuse && sk->sk_reuse &&
6759 sk->sk_state != SCTP_SS_LISTENING)
6762 /* Run through the list of sockets bound to the port
6763 * (pp->port) [via the pointers bind_next and
6764 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6765 * we get the endpoint they describe and run through
6766 * the endpoint's list of IP (v4 or v6) addresses,
6767 * comparing each of the addresses with the address of
6768 * the socket sk. If we find a match, then that means
6769 * that this port/socket (sk) combination are already
6772 sk_for_each_bound(sk2, &pp->owner) {
6773 struct sctp_endpoint *ep2;
6774 ep2 = sctp_sk(sk2)->ep;
6777 (reuse && sk2->sk_reuse &&
6778 sk2->sk_state != SCTP_SS_LISTENING))
6781 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
6782 sctp_sk(sk2), sctp_sk(sk))) {
6788 pr_debug("%s: found a match\n", __func__);
6791 /* If there was a hash table miss, create a new port. */
6793 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6796 /* In either case (hit or miss), make sure fastreuse is 1 only
6797 * if sk->sk_reuse is too (that is, if the caller requested
6798 * SO_REUSEADDR on this socket -sk-).
6800 if (hlist_empty(&pp->owner)) {
6801 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6805 } else if (pp->fastreuse &&
6806 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6809 /* We are set, so fill up all the data in the hash table
6810 * entry, tie the socket list information with the rest of the
6811 * sockets FIXME: Blurry, NPI (ipg).
6814 if (!sctp_sk(sk)->bind_hash) {
6815 inet_sk(sk)->inet_num = snum;
6816 sk_add_bind_node(sk, &pp->owner);
6817 sctp_sk(sk)->bind_hash = pp;
6822 spin_unlock_bh(&head->lock);
6826 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6827 * port is requested.
6829 static int sctp_get_port(struct sock *sk, unsigned short snum)
6831 union sctp_addr addr;
6832 struct sctp_af *af = sctp_sk(sk)->pf->af;
6834 /* Set up a dummy address struct from the sk. */
6835 af->from_sk(&addr, sk);
6836 addr.v4.sin_port = htons(snum);
6838 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6839 return !!sctp_get_port_local(sk, &addr);
6843 * Move a socket to LISTENING state.
6845 static int sctp_listen_start(struct sock *sk, int backlog)
6847 struct sctp_sock *sp = sctp_sk(sk);
6848 struct sctp_endpoint *ep = sp->ep;
6849 struct crypto_shash *tfm = NULL;
6852 /* Allocate HMAC for generating cookie. */
6853 if (!sp->hmac && sp->sctp_hmac_alg) {
6854 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6855 tfm = crypto_alloc_shash(alg, 0, 0);
6857 net_info_ratelimited("failed to load transform for %s: %ld\n",
6858 sp->sctp_hmac_alg, PTR_ERR(tfm));
6861 sctp_sk(sk)->hmac = tfm;
6865 * If a bind() or sctp_bindx() is not called prior to a listen()
6866 * call that allows new associations to be accepted, the system
6867 * picks an ephemeral port and will choose an address set equivalent
6868 * to binding with a wildcard address.
6870 * This is not currently spelled out in the SCTP sockets
6871 * extensions draft, but follows the practice as seen in TCP
6875 sk->sk_state = SCTP_SS_LISTENING;
6876 if (!ep->base.bind_addr.port) {
6877 if (sctp_autobind(sk))
6880 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6881 sk->sk_state = SCTP_SS_CLOSED;
6886 sk->sk_max_ack_backlog = backlog;
6887 sctp_hash_endpoint(ep);
6892 * 4.1.3 / 5.1.3 listen()
6894 * By default, new associations are not accepted for UDP style sockets.
6895 * An application uses listen() to mark a socket as being able to
6896 * accept new associations.
6898 * On TCP style sockets, applications use listen() to ready the SCTP
6899 * endpoint for accepting inbound associations.
6901 * On both types of endpoints a backlog of '0' disables listening.
6903 * Move a socket to LISTENING state.
6905 int sctp_inet_listen(struct socket *sock, int backlog)
6907 struct sock *sk = sock->sk;
6908 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6911 if (unlikely(backlog < 0))
6916 /* Peeled-off sockets are not allowed to listen(). */
6917 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6920 if (sock->state != SS_UNCONNECTED)
6923 if (!sctp_sstate(sk, LISTENING) && !sctp_sstate(sk, CLOSED))
6926 /* If backlog is zero, disable listening. */
6928 if (sctp_sstate(sk, CLOSED))
6932 sctp_unhash_endpoint(ep);
6933 sk->sk_state = SCTP_SS_CLOSED;
6935 sctp_sk(sk)->bind_hash->fastreuse = 1;
6939 /* If we are already listening, just update the backlog */
6940 if (sctp_sstate(sk, LISTENING))
6941 sk->sk_max_ack_backlog = backlog;
6943 err = sctp_listen_start(sk, backlog);
6955 * This function is done by modeling the current datagram_poll() and the
6956 * tcp_poll(). Note that, based on these implementations, we don't
6957 * lock the socket in this function, even though it seems that,
6958 * ideally, locking or some other mechanisms can be used to ensure
6959 * the integrity of the counters (sndbuf and wmem_alloc) used
6960 * in this place. We assume that we don't need locks either until proven
6963 * Another thing to note is that we include the Async I/O support
6964 * here, again, by modeling the current TCP/UDP code. We don't have
6965 * a good way to test with it yet.
6967 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6969 struct sock *sk = sock->sk;
6970 struct sctp_sock *sp = sctp_sk(sk);
6973 poll_wait(file, sk_sleep(sk), wait);
6975 sock_rps_record_flow(sk);
6977 /* A TCP-style listening socket becomes readable when the accept queue
6980 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6981 return (!list_empty(&sp->ep->asocs)) ?
6982 (POLLIN | POLLRDNORM) : 0;
6986 /* Is there any exceptional events? */
6987 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6989 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6990 if (sk->sk_shutdown & RCV_SHUTDOWN)
6991 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6992 if (sk->sk_shutdown == SHUTDOWN_MASK)
6995 /* Is it readable? Reconsider this code with TCP-style support. */
6996 if (!skb_queue_empty(&sk->sk_receive_queue))
6997 mask |= POLLIN | POLLRDNORM;
6999 /* The association is either gone or not ready. */
7000 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
7003 /* Is it writable? */
7004 if (sctp_writeable(sk)) {
7005 mask |= POLLOUT | POLLWRNORM;
7007 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
7009 * Since the socket is not locked, the buffer
7010 * might be made available after the writeable check and
7011 * before the bit is set. This could cause a lost I/O
7012 * signal. tcp_poll() has a race breaker for this race
7013 * condition. Based on their implementation, we put
7014 * in the following code to cover it as well.
7016 if (sctp_writeable(sk))
7017 mask |= POLLOUT | POLLWRNORM;
7022 /********************************************************************
7023 * 2nd Level Abstractions
7024 ********************************************************************/
7026 static struct sctp_bind_bucket *sctp_bucket_create(
7027 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
7029 struct sctp_bind_bucket *pp;
7031 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
7033 SCTP_DBG_OBJCNT_INC(bind_bucket);
7036 INIT_HLIST_HEAD(&pp->owner);
7038 hlist_add_head(&pp->node, &head->chain);
7043 /* Caller must hold hashbucket lock for this tb with local BH disabled */
7044 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
7046 if (pp && hlist_empty(&pp->owner)) {
7047 __hlist_del(&pp->node);
7048 kmem_cache_free(sctp_bucket_cachep, pp);
7049 SCTP_DBG_OBJCNT_DEC(bind_bucket);
7053 /* Release this socket's reference to a local port. */
7054 static inline void __sctp_put_port(struct sock *sk)
7056 struct sctp_bind_hashbucket *head =
7057 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
7058 inet_sk(sk)->inet_num)];
7059 struct sctp_bind_bucket *pp;
7061 spin_lock(&head->lock);
7062 pp = sctp_sk(sk)->bind_hash;
7063 __sk_del_bind_node(sk);
7064 sctp_sk(sk)->bind_hash = NULL;
7065 inet_sk(sk)->inet_num = 0;
7066 sctp_bucket_destroy(pp);
7067 spin_unlock(&head->lock);
7070 void sctp_put_port(struct sock *sk)
7073 __sctp_put_port(sk);
7078 * The system picks an ephemeral port and choose an address set equivalent
7079 * to binding with a wildcard address.
7080 * One of those addresses will be the primary address for the association.
7081 * This automatically enables the multihoming capability of SCTP.
7083 static int sctp_autobind(struct sock *sk)
7085 union sctp_addr autoaddr;
7089 /* Initialize a local sockaddr structure to INADDR_ANY. */
7090 af = sctp_sk(sk)->pf->af;
7092 port = htons(inet_sk(sk)->inet_num);
7093 af->inaddr_any(&autoaddr, port);
7095 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
7098 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
7101 * 4.2 The cmsghdr Structure *
7103 * When ancillary data is sent or received, any number of ancillary data
7104 * objects can be specified by the msg_control and msg_controllen members of
7105 * the msghdr structure, because each object is preceded by
7106 * a cmsghdr structure defining the object's length (the cmsg_len member).
7107 * Historically Berkeley-derived implementations have passed only one object
7108 * at a time, but this API allows multiple objects to be
7109 * passed in a single call to sendmsg() or recvmsg(). The following example
7110 * shows two ancillary data objects in a control buffer.
7112 * |<--------------------------- msg_controllen -------------------------->|
7115 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
7117 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
7120 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
7122 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
7125 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7126 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
7128 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
7130 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7137 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
7139 struct cmsghdr *cmsg;
7140 struct msghdr *my_msg = (struct msghdr *)msg;
7142 for_each_cmsghdr(cmsg, my_msg) {
7143 if (!CMSG_OK(my_msg, cmsg))
7146 /* Should we parse this header or ignore? */
7147 if (cmsg->cmsg_level != IPPROTO_SCTP)
7150 /* Strictly check lengths following example in SCM code. */
7151 switch (cmsg->cmsg_type) {
7153 /* SCTP Socket API Extension
7154 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
7156 * This cmsghdr structure provides information for
7157 * initializing new SCTP associations with sendmsg().
7158 * The SCTP_INITMSG socket option uses this same data
7159 * structure. This structure is not used for
7162 * cmsg_level cmsg_type cmsg_data[]
7163 * ------------ ------------ ----------------------
7164 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
7166 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
7169 cmsgs->init = CMSG_DATA(cmsg);
7173 /* SCTP Socket API Extension
7174 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
7176 * This cmsghdr structure specifies SCTP options for
7177 * sendmsg() and describes SCTP header information
7178 * about a received message through recvmsg().
7180 * cmsg_level cmsg_type cmsg_data[]
7181 * ------------ ------------ ----------------------
7182 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
7184 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
7187 cmsgs->srinfo = CMSG_DATA(cmsg);
7189 if (cmsgs->srinfo->sinfo_flags &
7190 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7191 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7192 SCTP_ABORT | SCTP_EOF))
7197 /* SCTP Socket API Extension
7198 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
7200 * This cmsghdr structure specifies SCTP options for
7201 * sendmsg(). This structure and SCTP_RCVINFO replaces
7202 * SCTP_SNDRCV which has been deprecated.
7204 * cmsg_level cmsg_type cmsg_data[]
7205 * ------------ ------------ ---------------------
7206 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
7208 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
7211 cmsgs->sinfo = CMSG_DATA(cmsg);
7213 if (cmsgs->sinfo->snd_flags &
7214 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7215 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7216 SCTP_ABORT | SCTP_EOF))
7228 * Wait for a packet..
7229 * Note: This function is the same function as in core/datagram.c
7230 * with a few modifications to make lksctp work.
7232 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
7237 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7239 /* Socket errors? */
7240 error = sock_error(sk);
7244 if (!skb_queue_empty(&sk->sk_receive_queue))
7247 /* Socket shut down? */
7248 if (sk->sk_shutdown & RCV_SHUTDOWN)
7251 /* Sequenced packets can come disconnected. If so we report the
7256 /* Is there a good reason to think that we may receive some data? */
7257 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
7260 /* Handle signals. */
7261 if (signal_pending(current))
7264 /* Let another process have a go. Since we are going to sleep
7265 * anyway. Note: This may cause odd behaviors if the message
7266 * does not fit in the user's buffer, but this seems to be the
7267 * only way to honor MSG_DONTWAIT realistically.
7270 *timeo_p = schedule_timeout(*timeo_p);
7274 finish_wait(sk_sleep(sk), &wait);
7278 error = sock_intr_errno(*timeo_p);
7281 finish_wait(sk_sleep(sk), &wait);
7286 /* Receive a datagram.
7287 * Note: This is pretty much the same routine as in core/datagram.c
7288 * with a few changes to make lksctp work.
7290 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
7291 int noblock, int *err)
7294 struct sk_buff *skb;
7297 timeo = sock_rcvtimeo(sk, noblock);
7299 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
7300 MAX_SCHEDULE_TIMEOUT);
7303 /* Again only user level code calls this function,
7304 * so nothing interrupt level
7305 * will suddenly eat the receive_queue.
7307 * Look at current nfs client by the way...
7308 * However, this function was correct in any case. 8)
7310 if (flags & MSG_PEEK) {
7311 skb = skb_peek(&sk->sk_receive_queue);
7313 atomic_inc(&skb->users);
7315 skb = __skb_dequeue(&sk->sk_receive_queue);
7321 /* Caller is allowed not to check sk->sk_err before calling. */
7322 error = sock_error(sk);
7326 if (sk->sk_shutdown & RCV_SHUTDOWN)
7329 if (sk_can_busy_loop(sk) &&
7330 sk_busy_loop(sk, noblock))
7333 /* User doesn't want to wait. */
7337 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
7346 /* If sndbuf has changed, wake up per association sndbuf waiters. */
7347 static void __sctp_write_space(struct sctp_association *asoc)
7349 struct sock *sk = asoc->base.sk;
7351 if (sctp_wspace(asoc) <= 0)
7354 if (waitqueue_active(&asoc->wait))
7355 wake_up_interruptible(&asoc->wait);
7357 if (sctp_writeable(sk)) {
7358 struct socket_wq *wq;
7361 wq = rcu_dereference(sk->sk_wq);
7363 if (waitqueue_active(&wq->wait))
7364 wake_up_interruptible(&wq->wait);
7366 /* Note that we try to include the Async I/O support
7367 * here by modeling from the current TCP/UDP code.
7368 * We have not tested with it yet.
7370 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
7371 sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT);
7377 static void sctp_wake_up_waiters(struct sock *sk,
7378 struct sctp_association *asoc)
7380 struct sctp_association *tmp = asoc;
7382 /* We do accounting for the sndbuf space per association,
7383 * so we only need to wake our own association.
7385 if (asoc->ep->sndbuf_policy)
7386 return __sctp_write_space(asoc);
7388 /* If association goes down and is just flushing its
7389 * outq, then just normally notify others.
7391 if (asoc->base.dead)
7392 return sctp_write_space(sk);
7394 /* Accounting for the sndbuf space is per socket, so we
7395 * need to wake up others, try to be fair and in case of
7396 * other associations, let them have a go first instead
7397 * of just doing a sctp_write_space() call.
7399 * Note that we reach sctp_wake_up_waiters() only when
7400 * associations free up queued chunks, thus we are under
7401 * lock and the list of associations on a socket is
7402 * guaranteed not to change.
7404 for (tmp = list_next_entry(tmp, asocs); 1;
7405 tmp = list_next_entry(tmp, asocs)) {
7406 /* Manually skip the head element. */
7407 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
7409 /* Wake up association. */
7410 __sctp_write_space(tmp);
7411 /* We've reached the end. */
7417 /* Do accounting for the sndbuf space.
7418 * Decrement the used sndbuf space of the corresponding association by the
7419 * data size which was just transmitted(freed).
7421 static void sctp_wfree(struct sk_buff *skb)
7423 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
7424 struct sctp_association *asoc = chunk->asoc;
7425 struct sock *sk = asoc->base.sk;
7427 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
7428 sizeof(struct sk_buff) +
7429 sizeof(struct sctp_chunk);
7431 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
7434 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
7436 sk->sk_wmem_queued -= skb->truesize;
7437 sk_mem_uncharge(sk, skb->truesize);
7440 sctp_wake_up_waiters(sk, asoc);
7442 sctp_association_put(asoc);
7445 /* Do accounting for the receive space on the socket.
7446 * Accounting for the association is done in ulpevent.c
7447 * We set this as a destructor for the cloned data skbs so that
7448 * accounting is done at the correct time.
7450 void sctp_sock_rfree(struct sk_buff *skb)
7452 struct sock *sk = skb->sk;
7453 struct sctp_ulpevent *event = sctp_skb2event(skb);
7455 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
7458 * Mimic the behavior of sock_rfree
7460 sk_mem_uncharge(sk, event->rmem_len);
7464 /* Helper function to wait for space in the sndbuf. */
7465 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
7468 struct sock *sk = asoc->base.sk;
7469 long current_timeo = *timeo_p;
7473 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
7476 /* Increment the association's refcnt. */
7477 sctp_association_hold(asoc);
7479 /* Wait on the association specific sndbuf space. */
7481 prepare_to_wait_exclusive(&asoc->wait, &wait,
7482 TASK_INTERRUPTIBLE);
7483 if (asoc->base.dead)
7487 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING)
7489 if (signal_pending(current))
7490 goto do_interrupted;
7491 if (msg_len <= sctp_wspace(asoc))
7494 /* Let another process have a go. Since we are going
7498 current_timeo = schedule_timeout(current_timeo);
7500 if (sk != asoc->base.sk)
7503 *timeo_p = current_timeo;
7507 finish_wait(&asoc->wait, &wait);
7509 /* Release the association's refcnt. */
7510 sctp_association_put(asoc);
7523 err = sock_intr_errno(*timeo_p);
7531 void sctp_data_ready(struct sock *sk)
7533 struct socket_wq *wq;
7536 wq = rcu_dereference(sk->sk_wq);
7537 if (skwq_has_sleeper(wq))
7538 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
7539 POLLRDNORM | POLLRDBAND);
7540 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
7544 /* If socket sndbuf has changed, wake up all per association waiters. */
7545 void sctp_write_space(struct sock *sk)
7547 struct sctp_association *asoc;
7549 /* Wake up the tasks in each wait queue. */
7550 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
7551 __sctp_write_space(asoc);
7555 /* Is there any sndbuf space available on the socket?
7557 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
7558 * associations on the same socket. For a UDP-style socket with
7559 * multiple associations, it is possible for it to be "unwriteable"
7560 * prematurely. I assume that this is acceptable because
7561 * a premature "unwriteable" is better than an accidental "writeable" which
7562 * would cause an unwanted block under certain circumstances. For the 1-1
7563 * UDP-style sockets or TCP-style sockets, this code should work.
7566 static int sctp_writeable(struct sock *sk)
7570 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
7576 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
7577 * returns immediately with EINPROGRESS.
7579 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
7581 struct sock *sk = asoc->base.sk;
7583 long current_timeo = *timeo_p;
7586 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
7588 /* Increment the association's refcnt. */
7589 sctp_association_hold(asoc);
7592 prepare_to_wait_exclusive(&asoc->wait, &wait,
7593 TASK_INTERRUPTIBLE);
7596 if (sk->sk_shutdown & RCV_SHUTDOWN)
7598 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7601 if (signal_pending(current))
7602 goto do_interrupted;
7604 if (sctp_state(asoc, ESTABLISHED))
7607 /* Let another process have a go. Since we are going
7611 current_timeo = schedule_timeout(current_timeo);
7614 *timeo_p = current_timeo;
7618 finish_wait(&asoc->wait, &wait);
7620 /* Release the association's refcnt. */
7621 sctp_association_put(asoc);
7626 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
7629 err = -ECONNREFUSED;
7633 err = sock_intr_errno(*timeo_p);
7641 static int sctp_wait_for_accept(struct sock *sk, long timeo)
7643 struct sctp_endpoint *ep;
7647 ep = sctp_sk(sk)->ep;
7651 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
7652 TASK_INTERRUPTIBLE);
7654 if (list_empty(&ep->asocs)) {
7656 timeo = schedule_timeout(timeo);
7661 if (!sctp_sstate(sk, LISTENING))
7665 if (!list_empty(&ep->asocs))
7668 err = sock_intr_errno(timeo);
7669 if (signal_pending(current))
7677 finish_wait(sk_sleep(sk), &wait);
7682 static void sctp_wait_for_close(struct sock *sk, long timeout)
7687 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7688 if (list_empty(&sctp_sk(sk)->ep->asocs))
7691 timeout = schedule_timeout(timeout);
7693 } while (!signal_pending(current) && timeout);
7695 finish_wait(sk_sleep(sk), &wait);
7698 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
7700 struct sk_buff *frag;
7705 /* Don't forget the fragments. */
7706 skb_walk_frags(skb, frag)
7707 sctp_skb_set_owner_r_frag(frag, sk);
7710 sctp_skb_set_owner_r(skb, sk);
7713 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
7714 struct sctp_association *asoc)
7716 struct inet_sock *inet = inet_sk(sk);
7717 struct inet_sock *newinet;
7719 newsk->sk_type = sk->sk_type;
7720 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
7721 newsk->sk_flags = sk->sk_flags;
7722 newsk->sk_tsflags = sk->sk_tsflags;
7723 newsk->sk_no_check_tx = sk->sk_no_check_tx;
7724 newsk->sk_no_check_rx = sk->sk_no_check_rx;
7725 newsk->sk_reuse = sk->sk_reuse;
7727 newsk->sk_shutdown = sk->sk_shutdown;
7728 newsk->sk_destruct = sctp_destruct_sock;
7729 newsk->sk_family = sk->sk_family;
7730 newsk->sk_protocol = IPPROTO_SCTP;
7731 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
7732 newsk->sk_sndbuf = sk->sk_sndbuf;
7733 newsk->sk_rcvbuf = sk->sk_rcvbuf;
7734 newsk->sk_lingertime = sk->sk_lingertime;
7735 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
7736 newsk->sk_sndtimeo = sk->sk_sndtimeo;
7737 newsk->sk_rxhash = sk->sk_rxhash;
7739 newinet = inet_sk(newsk);
7741 /* Initialize sk's sport, dport, rcv_saddr and daddr for
7742 * getsockname() and getpeername()
7744 newinet->inet_sport = inet->inet_sport;
7745 newinet->inet_saddr = inet->inet_saddr;
7746 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
7747 newinet->inet_dport = htons(asoc->peer.port);
7748 newinet->pmtudisc = inet->pmtudisc;
7749 newinet->inet_id = prandom_u32();
7751 newinet->uc_ttl = inet->uc_ttl;
7752 newinet->mc_loop = 1;
7753 newinet->mc_ttl = 1;
7754 newinet->mc_index = 0;
7755 newinet->mc_list = NULL;
7757 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
7758 net_enable_timestamp();
7760 security_sk_clone(sk, newsk);
7763 static inline void sctp_copy_descendant(struct sock *sk_to,
7764 const struct sock *sk_from)
7766 int ancestor_size = sizeof(struct inet_sock) +
7767 sizeof(struct sctp_sock) -
7768 offsetof(struct sctp_sock, auto_asconf_list);
7770 if (sk_from->sk_family == PF_INET6)
7771 ancestor_size += sizeof(struct ipv6_pinfo);
7773 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
7776 /* Populate the fields of the newsk from the oldsk and migrate the assoc
7777 * and its messages to the newsk.
7779 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
7780 struct sctp_association *assoc,
7781 sctp_socket_type_t type)
7783 struct sctp_sock *oldsp = sctp_sk(oldsk);
7784 struct sctp_sock *newsp = sctp_sk(newsk);
7785 struct sctp_bind_bucket *pp; /* hash list port iterator */
7786 struct sctp_endpoint *newep = newsp->ep;
7787 struct sk_buff *skb, *tmp;
7788 struct sctp_ulpevent *event;
7789 struct sctp_bind_hashbucket *head;
7791 /* Migrate socket buffer sizes and all the socket level options to the
7794 newsk->sk_sndbuf = oldsk->sk_sndbuf;
7795 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
7796 /* Brute force copy old sctp opt. */
7797 sctp_copy_descendant(newsk, oldsk);
7799 /* Restore the ep value that was overwritten with the above structure
7805 /* Hook this new socket in to the bind_hash list. */
7806 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
7807 inet_sk(oldsk)->inet_num)];
7808 spin_lock_bh(&head->lock);
7809 pp = sctp_sk(oldsk)->bind_hash;
7810 sk_add_bind_node(newsk, &pp->owner);
7811 sctp_sk(newsk)->bind_hash = pp;
7812 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
7813 spin_unlock_bh(&head->lock);
7815 /* Copy the bind_addr list from the original endpoint to the new
7816 * endpoint so that we can handle restarts properly
7818 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
7819 &oldsp->ep->base.bind_addr, GFP_KERNEL);
7821 sctp_auto_asconf_init(newsp);
7823 /* Move any messages in the old socket's receive queue that are for the
7824 * peeled off association to the new socket's receive queue.
7826 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
7827 event = sctp_skb2event(skb);
7828 if (event->asoc == assoc) {
7829 __skb_unlink(skb, &oldsk->sk_receive_queue);
7830 __skb_queue_tail(&newsk->sk_receive_queue, skb);
7831 sctp_skb_set_owner_r_frag(skb, newsk);
7835 /* Clean up any messages pending delivery due to partial
7836 * delivery. Three cases:
7837 * 1) No partial deliver; no work.
7838 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7839 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7841 skb_queue_head_init(&newsp->pd_lobby);
7842 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
7844 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
7845 struct sk_buff_head *queue;
7847 /* Decide which queue to move pd_lobby skbs to. */
7848 if (assoc->ulpq.pd_mode) {
7849 queue = &newsp->pd_lobby;
7851 queue = &newsk->sk_receive_queue;
7853 /* Walk through the pd_lobby, looking for skbs that
7854 * need moved to the new socket.
7856 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
7857 event = sctp_skb2event(skb);
7858 if (event->asoc == assoc) {
7859 __skb_unlink(skb, &oldsp->pd_lobby);
7860 __skb_queue_tail(queue, skb);
7861 sctp_skb_set_owner_r_frag(skb, newsk);
7865 /* Clear up any skbs waiting for the partial
7866 * delivery to finish.
7868 if (assoc->ulpq.pd_mode)
7869 sctp_clear_pd(oldsk, NULL);
7873 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7874 sctp_skb_set_owner_r_frag(skb, newsk);
7876 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7877 sctp_skb_set_owner_r_frag(skb, newsk);
7879 /* Set the type of socket to indicate that it is peeled off from the
7880 * original UDP-style socket or created with the accept() call on a
7881 * TCP-style socket..
7885 /* Mark the new socket "in-use" by the user so that any packets
7886 * that may arrive on the association after we've moved it are
7887 * queued to the backlog. This prevents a potential race between
7888 * backlog processing on the old socket and new-packet processing
7889 * on the new socket.
7891 * The caller has just allocated newsk so we can guarantee that other
7892 * paths won't try to lock it and then oldsk.
7894 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7895 sctp_for_each_tx_datachunk(assoc, true, sctp_clear_owner_w);
7896 sctp_assoc_migrate(assoc, newsk);
7897 sctp_for_each_tx_datachunk(assoc, false, sctp_set_owner_w);
7899 /* If the association on the newsk is already closed before accept()
7900 * is called, set RCV_SHUTDOWN flag.
7902 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP)) {
7903 newsk->sk_state = SCTP_SS_CLOSED;
7904 newsk->sk_shutdown |= RCV_SHUTDOWN;
7906 newsk->sk_state = SCTP_SS_ESTABLISHED;
7909 release_sock(newsk);
7913 /* This proto struct describes the ULP interface for SCTP. */
7914 struct proto sctp_prot = {
7916 .owner = THIS_MODULE,
7917 .close = sctp_close,
7918 .disconnect = sctp_disconnect,
7919 .accept = sctp_accept,
7920 .ioctl = sctp_ioctl,
7921 .init = sctp_init_sock,
7922 .destroy = sctp_destroy_sock,
7923 .shutdown = sctp_shutdown,
7924 .setsockopt = sctp_setsockopt,
7925 .getsockopt = sctp_getsockopt,
7926 .sendmsg = sctp_sendmsg,
7927 .recvmsg = sctp_recvmsg,
7929 .backlog_rcv = sctp_backlog_rcv,
7931 .unhash = sctp_unhash,
7932 .no_autobind = true,
7933 .obj_size = sizeof(struct sctp_sock),
7934 .sysctl_mem = sysctl_sctp_mem,
7935 .sysctl_rmem = sysctl_sctp_rmem,
7936 .sysctl_wmem = sysctl_sctp_wmem,
7937 .memory_pressure = &sctp_memory_pressure,
7938 .enter_memory_pressure = sctp_enter_memory_pressure,
7939 .memory_allocated = &sctp_memory_allocated,
7940 .sockets_allocated = &sctp_sockets_allocated,
7943 #if IS_ENABLED(CONFIG_IPV6)
7945 #include <net/transp_v6.h>
7946 static void sctp_v6_destroy_sock(struct sock *sk)
7948 sctp_destroy_sock(sk);
7949 inet6_destroy_sock(sk);
7952 struct proto sctpv6_prot = {
7954 .owner = THIS_MODULE,
7955 .close = sctp_close,
7956 .disconnect = sctp_disconnect,
7957 .accept = sctp_accept,
7958 .ioctl = sctp_ioctl,
7959 .init = sctp_init_sock,
7960 .destroy = sctp_v6_destroy_sock,
7961 .shutdown = sctp_shutdown,
7962 .setsockopt = sctp_setsockopt,
7963 .getsockopt = sctp_getsockopt,
7964 .sendmsg = sctp_sendmsg,
7965 .recvmsg = sctp_recvmsg,
7967 .backlog_rcv = sctp_backlog_rcv,
7969 .unhash = sctp_unhash,
7970 .no_autobind = true,
7971 .obj_size = sizeof(struct sctp6_sock),
7972 .sysctl_mem = sysctl_sctp_mem,
7973 .sysctl_rmem = sysctl_sctp_rmem,
7974 .sysctl_wmem = sysctl_sctp_wmem,
7975 .memory_pressure = &sctp_memory_pressure,
7976 .enter_memory_pressure = sctp_enter_memory_pressure,
7977 .memory_allocated = &sctp_memory_allocated,
7978 .sockets_allocated = &sctp_sockets_allocated,
7980 #endif /* IS_ENABLED(CONFIG_IPV6) */