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 *, 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 /* Verify that this is a valid address. */
172 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
177 /* Verify basic sockaddr. */
178 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
182 /* Is this a valid SCTP address? */
183 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
186 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
192 /* Look up the association by its id. If this is not a UDP-style
193 * socket, the ID field is always ignored.
195 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
197 struct sctp_association *asoc = NULL;
199 /* If this is not a UDP-style socket, assoc id should be ignored. */
200 if (!sctp_style(sk, UDP)) {
201 /* Return NULL if the socket state is not ESTABLISHED. It
202 * could be a TCP-style listening socket or a socket which
203 * hasn't yet called connect() to establish an association.
205 if (!sctp_sstate(sk, ESTABLISHED) && !sctp_sstate(sk, CLOSING))
208 /* Get the first and the only association from the list. */
209 if (!list_empty(&sctp_sk(sk)->ep->asocs))
210 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
211 struct sctp_association, asocs);
215 /* Otherwise this is a UDP-style socket. */
216 if (!id || (id == (sctp_assoc_t)-1))
219 spin_lock_bh(&sctp_assocs_id_lock);
220 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
221 spin_unlock_bh(&sctp_assocs_id_lock);
223 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
229 /* Look up the transport from an address and an assoc id. If both address and
230 * id are specified, the associations matching the address and the id should be
233 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
234 struct sockaddr_storage *addr,
237 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
238 struct sctp_af *af = sctp_get_af_specific(addr->ss_family);
239 union sctp_addr *laddr = (union sctp_addr *)addr;
240 struct sctp_transport *transport;
242 if (!af || sctp_verify_addr(sk, laddr, af->sockaddr_len))
245 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
252 id_asoc = sctp_id2assoc(sk, id);
253 if (id_asoc && (id_asoc != addr_asoc))
256 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
257 (union sctp_addr *)addr);
262 /* API 3.1.2 bind() - UDP Style Syntax
263 * The syntax of bind() is,
265 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
267 * sd - the socket descriptor returned by socket().
268 * addr - the address structure (struct sockaddr_in or struct
269 * sockaddr_in6 [RFC 2553]),
270 * addr_len - the size of the address structure.
272 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
278 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
281 /* Disallow binding twice. */
282 if (!sctp_sk(sk)->ep->base.bind_addr.port)
283 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
293 static long sctp_get_port_local(struct sock *, union sctp_addr *);
295 /* Verify this is a valid sockaddr. */
296 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
297 union sctp_addr *addr, int len)
301 /* Check minimum size. */
302 if (len < sizeof (struct sockaddr))
305 /* V4 mapped address are really of AF_INET family */
306 if (addr->sa.sa_family == AF_INET6 &&
307 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
308 if (!opt->pf->af_supported(AF_INET, opt))
311 /* Does this PF support this AF? */
312 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
316 /* If we get this far, af is valid. */
317 af = sctp_get_af_specific(addr->sa.sa_family);
319 if (len < af->sockaddr_len)
325 /* Bind a local address either to an endpoint or to an association. */
326 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
328 struct net *net = sock_net(sk);
329 struct sctp_sock *sp = sctp_sk(sk);
330 struct sctp_endpoint *ep = sp->ep;
331 struct sctp_bind_addr *bp = &ep->base.bind_addr;
336 /* Common sockaddr verification. */
337 af = sctp_sockaddr_af(sp, addr, len);
339 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
340 __func__, sk, addr, len);
344 snum = ntohs(addr->v4.sin_port);
346 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
347 __func__, sk, &addr->sa, bp->port, snum, len);
349 /* PF specific bind() address verification. */
350 if (!sp->pf->bind_verify(sp, addr))
351 return -EADDRNOTAVAIL;
353 /* We must either be unbound, or bind to the same port.
354 * It's OK to allow 0 ports if we are already bound.
355 * We'll just inhert an already bound port in this case
360 else if (snum != bp->port) {
361 pr_debug("%s: new port %d doesn't match existing port "
362 "%d\n", __func__, snum, bp->port);
367 if (snum && snum < inet_prot_sock(net) &&
368 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
371 /* See if the address matches any of the addresses we may have
372 * already bound before checking against other endpoints.
374 if (sctp_bind_addr_match(bp, addr, sp))
377 /* Make sure we are allowed to bind here.
378 * The function sctp_get_port_local() does duplicate address
381 addr->v4.sin_port = htons(snum);
382 if ((ret = sctp_get_port_local(sk, addr))) {
386 /* Refresh ephemeral port. */
388 bp->port = inet_sk(sk)->inet_num;
390 /* Add the address to the bind address list.
391 * Use GFP_ATOMIC since BHs will be disabled.
393 ret = sctp_add_bind_addr(bp, addr, af->sockaddr_len,
394 SCTP_ADDR_SRC, GFP_ATOMIC);
396 /* Copy back into socket for getsockname() use. */
398 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
399 sp->pf->to_sk_saddr(addr, sk);
405 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
407 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
408 * at any one time. If a sender, after sending an ASCONF chunk, decides
409 * it needs to transfer another ASCONF Chunk, it MUST wait until the
410 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
411 * subsequent ASCONF. Note this restriction binds each side, so at any
412 * time two ASCONF may be in-transit on any given association (one sent
413 * from each endpoint).
415 static int sctp_send_asconf(struct sctp_association *asoc,
416 struct sctp_chunk *chunk)
418 struct net *net = sock_net(asoc->base.sk);
421 /* If there is an outstanding ASCONF chunk, queue it for later
424 if (asoc->addip_last_asconf) {
425 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
429 /* Hold the chunk until an ASCONF_ACK is received. */
430 sctp_chunk_hold(chunk);
431 retval = sctp_primitive_ASCONF(net, asoc, chunk);
433 sctp_chunk_free(chunk);
435 asoc->addip_last_asconf = chunk;
441 /* Add a list of addresses as bind addresses to local endpoint or
444 * Basically run through each address specified in the addrs/addrcnt
445 * array/length pair, determine if it is IPv6 or IPv4 and call
446 * sctp_do_bind() on it.
448 * If any of them fails, then the operation will be reversed and the
449 * ones that were added will be removed.
451 * Only sctp_setsockopt_bindx() is supposed to call this function.
453 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
458 struct sockaddr *sa_addr;
461 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
465 for (cnt = 0; cnt < addrcnt; cnt++) {
466 /* The list may contain either IPv4 or IPv6 address;
467 * determine the address length for walking thru the list.
470 af = sctp_get_af_specific(sa_addr->sa_family);
476 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
479 addr_buf += af->sockaddr_len;
483 /* Failed. Cleanup the ones that have been added */
485 sctp_bindx_rem(sk, addrs, cnt);
493 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
494 * associations that are part of the endpoint indicating that a list of local
495 * addresses are added to the endpoint.
497 * If any of the addresses is already in the bind address list of the
498 * association, we do not send the chunk for that association. But it will not
499 * affect other associations.
501 * Only sctp_setsockopt_bindx() is supposed to call this function.
503 static int sctp_send_asconf_add_ip(struct sock *sk,
504 struct sockaddr *addrs,
507 struct net *net = sock_net(sk);
508 struct sctp_sock *sp;
509 struct sctp_endpoint *ep;
510 struct sctp_association *asoc;
511 struct sctp_bind_addr *bp;
512 struct sctp_chunk *chunk;
513 struct sctp_sockaddr_entry *laddr;
514 union sctp_addr *addr;
515 union sctp_addr saveaddr;
522 if (!net->sctp.addip_enable)
528 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
529 __func__, sk, addrs, addrcnt);
531 list_for_each_entry(asoc, &ep->asocs, asocs) {
532 if (!asoc->peer.asconf_capable)
535 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
538 if (!sctp_state(asoc, ESTABLISHED))
541 /* Check if any address in the packed array of addresses is
542 * in the bind address list of the association. If so,
543 * do not send the asconf chunk to its peer, but continue with
544 * other associations.
547 for (i = 0; i < addrcnt; i++) {
549 af = sctp_get_af_specific(addr->v4.sin_family);
555 if (sctp_assoc_lookup_laddr(asoc, addr))
558 addr_buf += af->sockaddr_len;
563 /* Use the first valid address in bind addr list of
564 * association as Address Parameter of ASCONF CHUNK.
566 bp = &asoc->base.bind_addr;
567 p = bp->address_list.next;
568 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
569 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
570 addrcnt, SCTP_PARAM_ADD_IP);
576 /* Add the new addresses to the bind address list with
577 * use_as_src set to 0.
580 for (i = 0; i < addrcnt; i++) {
582 af = sctp_get_af_specific(addr->v4.sin_family);
583 memcpy(&saveaddr, addr, af->sockaddr_len);
584 retval = sctp_add_bind_addr(bp, &saveaddr,
586 SCTP_ADDR_NEW, GFP_ATOMIC);
587 addr_buf += af->sockaddr_len;
589 if (asoc->src_out_of_asoc_ok) {
590 struct sctp_transport *trans;
592 list_for_each_entry(trans,
593 &asoc->peer.transport_addr_list, transports) {
594 /* Clear the source and route cache */
595 sctp_transport_dst_release(trans);
596 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
597 2*asoc->pathmtu, 4380));
598 trans->ssthresh = asoc->peer.i.a_rwnd;
599 trans->rto = asoc->rto_initial;
600 sctp_max_rto(asoc, trans);
601 trans->rtt = trans->srtt = trans->rttvar = 0;
602 sctp_transport_route(trans, NULL,
603 sctp_sk(asoc->base.sk));
606 retval = sctp_send_asconf(asoc, chunk);
613 /* Remove a list of addresses from bind addresses list. Do not remove the
616 * Basically run through each address specified in the addrs/addrcnt
617 * array/length pair, determine if it is IPv6 or IPv4 and call
618 * sctp_del_bind() on it.
620 * If any of them fails, then the operation will be reversed and the
621 * ones that were removed will be added back.
623 * At least one address has to be left; if only one address is
624 * available, the operation will return -EBUSY.
626 * Only sctp_setsockopt_bindx() is supposed to call this function.
628 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
630 struct sctp_sock *sp = sctp_sk(sk);
631 struct sctp_endpoint *ep = sp->ep;
633 struct sctp_bind_addr *bp = &ep->base.bind_addr;
636 union sctp_addr *sa_addr;
639 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
640 __func__, sk, addrs, addrcnt);
643 for (cnt = 0; cnt < addrcnt; cnt++) {
644 /* If the bind address list is empty or if there is only one
645 * bind address, there is nothing more to be removed (we need
646 * at least one address here).
648 if (list_empty(&bp->address_list) ||
649 (sctp_list_single_entry(&bp->address_list))) {
655 af = sctp_get_af_specific(sa_addr->sa.sa_family);
661 if (!af->addr_valid(sa_addr, sp, NULL)) {
662 retval = -EADDRNOTAVAIL;
666 if (sa_addr->v4.sin_port &&
667 sa_addr->v4.sin_port != htons(bp->port)) {
672 if (!sa_addr->v4.sin_port)
673 sa_addr->v4.sin_port = htons(bp->port);
675 /* FIXME - There is probably a need to check if sk->sk_saddr and
676 * sk->sk_rcv_addr are currently set to one of the addresses to
677 * be removed. This is something which needs to be looked into
678 * when we are fixing the outstanding issues with multi-homing
679 * socket routing and failover schemes. Refer to comments in
680 * sctp_do_bind(). -daisy
682 retval = sctp_del_bind_addr(bp, sa_addr);
684 addr_buf += af->sockaddr_len;
687 /* Failed. Add the ones that has been removed back */
689 sctp_bindx_add(sk, addrs, cnt);
697 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
698 * the associations that are part of the endpoint indicating that a list of
699 * local addresses are removed from the endpoint.
701 * If any of the addresses is already in the bind address list of the
702 * association, we do not send the chunk for that association. But it will not
703 * affect other associations.
705 * Only sctp_setsockopt_bindx() is supposed to call this function.
707 static int sctp_send_asconf_del_ip(struct sock *sk,
708 struct sockaddr *addrs,
711 struct net *net = sock_net(sk);
712 struct sctp_sock *sp;
713 struct sctp_endpoint *ep;
714 struct sctp_association *asoc;
715 struct sctp_transport *transport;
716 struct sctp_bind_addr *bp;
717 struct sctp_chunk *chunk;
718 union sctp_addr *laddr;
721 struct sctp_sockaddr_entry *saddr;
727 if (!net->sctp.addip_enable)
733 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
734 __func__, sk, addrs, addrcnt);
736 list_for_each_entry(asoc, &ep->asocs, asocs) {
738 if (!asoc->peer.asconf_capable)
741 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
744 if (!sctp_state(asoc, ESTABLISHED))
747 /* Check if any address in the packed array of addresses is
748 * not present in the bind address list of the association.
749 * If so, do not send the asconf chunk to its peer, but
750 * continue with other associations.
753 for (i = 0; i < addrcnt; i++) {
755 af = sctp_get_af_specific(laddr->v4.sin_family);
761 if (!sctp_assoc_lookup_laddr(asoc, laddr))
764 addr_buf += af->sockaddr_len;
769 /* Find one address in the association's bind address list
770 * that is not in the packed array of addresses. This is to
771 * make sure that we do not delete all the addresses in the
774 bp = &asoc->base.bind_addr;
775 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
777 if ((laddr == NULL) && (addrcnt == 1)) {
778 if (asoc->asconf_addr_del_pending)
780 asoc->asconf_addr_del_pending =
781 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
782 if (asoc->asconf_addr_del_pending == NULL) {
786 asoc->asconf_addr_del_pending->sa.sa_family =
788 asoc->asconf_addr_del_pending->v4.sin_port =
790 if (addrs->sa_family == AF_INET) {
791 struct sockaddr_in *sin;
793 sin = (struct sockaddr_in *)addrs;
794 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
795 } else if (addrs->sa_family == AF_INET6) {
796 struct sockaddr_in6 *sin6;
798 sin6 = (struct sockaddr_in6 *)addrs;
799 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
802 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
803 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
804 asoc->asconf_addr_del_pending);
806 asoc->src_out_of_asoc_ok = 1;
814 /* We do not need RCU protection throughout this loop
815 * because this is done under a socket lock from the
818 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
826 /* Reset use_as_src flag for the addresses in the bind address
827 * list that are to be deleted.
830 for (i = 0; i < addrcnt; i++) {
832 af = sctp_get_af_specific(laddr->v4.sin_family);
833 list_for_each_entry(saddr, &bp->address_list, list) {
834 if (sctp_cmp_addr_exact(&saddr->a, laddr))
835 saddr->state = SCTP_ADDR_DEL;
837 addr_buf += af->sockaddr_len;
840 /* Update the route and saddr entries for all the transports
841 * as some of the addresses in the bind address list are
842 * about to be deleted and cannot be used as source addresses.
844 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
846 sctp_transport_dst_release(transport);
847 sctp_transport_route(transport, NULL,
848 sctp_sk(asoc->base.sk));
852 /* We don't need to transmit ASCONF */
854 retval = sctp_send_asconf(asoc, chunk);
860 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
861 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
863 struct sock *sk = sctp_opt2sk(sp);
864 union sctp_addr *addr;
867 /* It is safe to write port space in caller. */
869 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
870 af = sctp_get_af_specific(addr->sa.sa_family);
873 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
876 if (addrw->state == SCTP_ADDR_NEW)
877 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
879 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
882 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
885 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
888 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
889 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
892 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
893 * Section 3.1.2 for this usage.
895 * addrs is a pointer to an array of one or more socket addresses. Each
896 * address is contained in its appropriate structure (i.e. struct
897 * sockaddr_in or struct sockaddr_in6) the family of the address type
898 * must be used to distinguish the address length (note that this
899 * representation is termed a "packed array" of addresses). The caller
900 * specifies the number of addresses in the array with addrcnt.
902 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
903 * -1, and sets errno to the appropriate error code.
905 * For SCTP, the port given in each socket address must be the same, or
906 * sctp_bindx() will fail, setting errno to EINVAL.
908 * The flags parameter is formed from the bitwise OR of zero or more of
909 * the following currently defined flags:
911 * SCTP_BINDX_ADD_ADDR
913 * SCTP_BINDX_REM_ADDR
915 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
916 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
917 * addresses from the association. The two flags are mutually exclusive;
918 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
919 * not remove all addresses from an association; sctp_bindx() will
920 * reject such an attempt with EINVAL.
922 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
923 * additional addresses with an endpoint after calling bind(). Or use
924 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
925 * socket is associated with so that no new association accepted will be
926 * associated with those addresses. If the endpoint supports dynamic
927 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
928 * endpoint to send the appropriate message to the peer to change the
929 * peers address lists.
931 * Adding and removing addresses from a connected association is
932 * optional functionality. Implementations that do not support this
933 * functionality should return EOPNOTSUPP.
935 * Basically do nothing but copying the addresses from user to kernel
936 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
937 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
940 * We don't use copy_from_user() for optimization: we first do the
941 * sanity checks (buffer size -fast- and access check-healthy
942 * pointer); if all of those succeed, then we can alloc the memory
943 * (expensive operation) needed to copy the data to kernel. Then we do
944 * the copying without checking the user space area
945 * (__copy_from_user()).
947 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
950 * sk The sk of the socket
951 * addrs The pointer to the addresses in user land
952 * addrssize Size of the addrs buffer
953 * op Operation to perform (add or remove, see the flags of
956 * Returns 0 if ok, <0 errno code on error.
958 static int sctp_setsockopt_bindx(struct sock *sk,
959 struct sockaddr __user *addrs,
960 int addrs_size, int op)
962 struct sockaddr *kaddrs;
966 struct sockaddr *sa_addr;
970 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
971 __func__, sk, addrs, addrs_size, op);
973 if (unlikely(addrs_size <= 0))
976 /* Check the user passed a healthy pointer. */
977 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
980 /* Alloc space for the address array in kernel memory. */
981 kaddrs = kmalloc(addrs_size, GFP_USER | __GFP_NOWARN);
982 if (unlikely(!kaddrs))
985 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
990 /* Walk through the addrs buffer and count the number of addresses. */
992 while (walk_size < addrs_size) {
993 if (walk_size + sizeof(sa_family_t) > addrs_size) {
999 af = sctp_get_af_specific(sa_addr->sa_family);
1001 /* If the address family is not supported or if this address
1002 * causes the address buffer to overflow return EINVAL.
1004 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1009 addr_buf += af->sockaddr_len;
1010 walk_size += af->sockaddr_len;
1015 case SCTP_BINDX_ADD_ADDR:
1016 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1019 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1022 case SCTP_BINDX_REM_ADDR:
1023 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1026 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1040 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1042 * Common routine for handling connect() and sctp_connectx().
1043 * Connect will come in with just a single address.
1045 static int __sctp_connect(struct sock *sk,
1046 struct sockaddr *kaddrs,
1048 sctp_assoc_t *assoc_id)
1050 struct net *net = sock_net(sk);
1051 struct sctp_sock *sp;
1052 struct sctp_endpoint *ep;
1053 struct sctp_association *asoc = NULL;
1054 struct sctp_association *asoc2;
1055 struct sctp_transport *transport;
1062 union sctp_addr *sa_addr = NULL;
1064 unsigned short port;
1065 unsigned int f_flags = 0;
1070 /* connect() cannot be done on a socket that is already in ESTABLISHED
1071 * state - UDP-style peeled off socket or a TCP-style socket that
1072 * is already connected.
1073 * It cannot be done even on a TCP-style listening socket.
1075 if (sctp_sstate(sk, ESTABLISHED) || sctp_sstate(sk, CLOSING) ||
1076 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1081 /* Walk through the addrs buffer and count the number of addresses. */
1083 while (walk_size < addrs_size) {
1086 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1092 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1094 /* If the address family is not supported or if this address
1095 * causes the address buffer to overflow return EINVAL.
1097 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1102 port = ntohs(sa_addr->v4.sin_port);
1104 /* Save current address so we can work with it */
1105 memcpy(&to, sa_addr, af->sockaddr_len);
1107 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1111 /* Make sure the destination port is correctly set
1114 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1119 /* Check if there already is a matching association on the
1120 * endpoint (other than the one created here).
1122 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1123 if (asoc2 && asoc2 != asoc) {
1124 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1131 /* If we could not find a matching association on the endpoint,
1132 * make sure that there is no peeled-off association matching
1133 * the peer address even on another socket.
1135 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1136 err = -EADDRNOTAVAIL;
1141 /* If a bind() or sctp_bindx() is not called prior to
1142 * an sctp_connectx() call, the system picks an
1143 * ephemeral port and will choose an address set
1144 * equivalent to binding with a wildcard address.
1146 if (!ep->base.bind_addr.port) {
1147 if (sctp_autobind(sk)) {
1153 * If an unprivileged user inherits a 1-many
1154 * style socket with open associations on a
1155 * privileged port, it MAY be permitted to
1156 * accept new associations, but it SHOULD NOT
1157 * be permitted to open new associations.
1159 if (ep->base.bind_addr.port <
1160 inet_prot_sock(net) &&
1161 !ns_capable(net->user_ns,
1162 CAP_NET_BIND_SERVICE)) {
1168 scope = sctp_scope(&to);
1169 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1175 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1183 /* Prime the peer's transport structures. */
1184 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1192 addr_buf += af->sockaddr_len;
1193 walk_size += af->sockaddr_len;
1196 /* In case the user of sctp_connectx() wants an association
1197 * id back, assign one now.
1200 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1205 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1210 /* Initialize sk's dport and daddr for getpeername() */
1211 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1212 sp->pf->to_sk_daddr(sa_addr, sk);
1215 /* in-kernel sockets don't generally have a file allocated to them
1216 * if all they do is call sock_create_kern().
1218 if (sk->sk_socket->file)
1219 f_flags = sk->sk_socket->file->f_flags;
1221 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1224 *assoc_id = asoc->assoc_id;
1225 err = sctp_wait_for_connect(asoc, &timeo);
1226 /* Note: the asoc may be freed after the return of
1227 * sctp_wait_for_connect.
1230 /* Don't free association on exit. */
1234 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1235 __func__, asoc, kaddrs, err);
1238 /* sctp_primitive_ASSOCIATE may have added this association
1239 * To the hash table, try to unhash it, just in case, its a noop
1240 * if it wasn't hashed so we're safe
1242 sctp_association_free(asoc);
1247 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1250 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1251 * sctp_assoc_t *asoc);
1253 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1254 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1255 * or IPv6 addresses.
1257 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1258 * Section 3.1.2 for this usage.
1260 * addrs is a pointer to an array of one or more socket addresses. Each
1261 * address is contained in its appropriate structure (i.e. struct
1262 * sockaddr_in or struct sockaddr_in6) the family of the address type
1263 * must be used to distengish the address length (note that this
1264 * representation is termed a "packed array" of addresses). The caller
1265 * specifies the number of addresses in the array with addrcnt.
1267 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1268 * the association id of the new association. On failure, sctp_connectx()
1269 * returns -1, and sets errno to the appropriate error code. The assoc_id
1270 * is not touched by the kernel.
1272 * For SCTP, the port given in each socket address must be the same, or
1273 * sctp_connectx() will fail, setting errno to EINVAL.
1275 * An application can use sctp_connectx to initiate an association with
1276 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1277 * allows a caller to specify multiple addresses at which a peer can be
1278 * reached. The way the SCTP stack uses the list of addresses to set up
1279 * the association is implementation dependent. This function only
1280 * specifies that the stack will try to make use of all the addresses in
1281 * the list when needed.
1283 * Note that the list of addresses passed in is only used for setting up
1284 * the association. It does not necessarily equal the set of addresses
1285 * the peer uses for the resulting association. If the caller wants to
1286 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1287 * retrieve them after the association has been set up.
1289 * Basically do nothing but copying the addresses from user to kernel
1290 * land and invoking either sctp_connectx(). This is used for tunneling
1291 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1293 * We don't use copy_from_user() for optimization: we first do the
1294 * sanity checks (buffer size -fast- and access check-healthy
1295 * pointer); if all of those succeed, then we can alloc the memory
1296 * (expensive operation) needed to copy the data to kernel. Then we do
1297 * the copying without checking the user space area
1298 * (__copy_from_user()).
1300 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1303 * sk The sk of the socket
1304 * addrs The pointer to the addresses in user land
1305 * addrssize Size of the addrs buffer
1307 * Returns >=0 if ok, <0 errno code on error.
1309 static int __sctp_setsockopt_connectx(struct sock *sk,
1310 struct sockaddr __user *addrs,
1312 sctp_assoc_t *assoc_id)
1314 struct sockaddr *kaddrs;
1315 gfp_t gfp = GFP_KERNEL;
1318 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1319 __func__, sk, addrs, addrs_size);
1321 if (unlikely(addrs_size <= 0))
1324 /* Check the user passed a healthy pointer. */
1325 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1328 /* Alloc space for the address array in kernel memory. */
1329 if (sk->sk_socket->file)
1330 gfp = GFP_USER | __GFP_NOWARN;
1331 kaddrs = kmalloc(addrs_size, gfp);
1332 if (unlikely(!kaddrs))
1335 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1338 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1347 * This is an older interface. It's kept for backward compatibility
1348 * to the option that doesn't provide association id.
1350 static int sctp_setsockopt_connectx_old(struct sock *sk,
1351 struct sockaddr __user *addrs,
1354 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1358 * New interface for the API. The since the API is done with a socket
1359 * option, to make it simple we feed back the association id is as a return
1360 * indication to the call. Error is always negative and association id is
1363 static int sctp_setsockopt_connectx(struct sock *sk,
1364 struct sockaddr __user *addrs,
1367 sctp_assoc_t assoc_id = 0;
1370 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1379 * New (hopefully final) interface for the API.
1380 * We use the sctp_getaddrs_old structure so that use-space library
1381 * can avoid any unnecessary allocations. The only different part
1382 * is that we store the actual length of the address buffer into the
1383 * addrs_num structure member. That way we can re-use the existing
1386 #ifdef CONFIG_COMPAT
1387 struct compat_sctp_getaddrs_old {
1388 sctp_assoc_t assoc_id;
1390 compat_uptr_t addrs; /* struct sockaddr * */
1394 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1395 char __user *optval,
1398 struct sctp_getaddrs_old param;
1399 sctp_assoc_t assoc_id = 0;
1402 #ifdef CONFIG_COMPAT
1403 if (in_compat_syscall()) {
1404 struct compat_sctp_getaddrs_old param32;
1406 if (len < sizeof(param32))
1408 if (copy_from_user(¶m32, optval, sizeof(param32)))
1411 param.assoc_id = param32.assoc_id;
1412 param.addr_num = param32.addr_num;
1413 param.addrs = compat_ptr(param32.addrs);
1417 if (len < sizeof(param))
1419 if (copy_from_user(¶m, optval, sizeof(param)))
1423 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1424 param.addrs, param.addr_num,
1426 if (err == 0 || err == -EINPROGRESS) {
1427 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1429 if (put_user(sizeof(assoc_id), optlen))
1436 /* API 3.1.4 close() - UDP Style Syntax
1437 * Applications use close() to perform graceful shutdown (as described in
1438 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1439 * by a UDP-style socket.
1443 * ret = close(int sd);
1445 * sd - the socket descriptor of the associations to be closed.
1447 * To gracefully shutdown a specific association represented by the
1448 * UDP-style socket, an application should use the sendmsg() call,
1449 * passing no user data, but including the appropriate flag in the
1450 * ancillary data (see Section xxxx).
1452 * If sd in the close() call is a branched-off socket representing only
1453 * one association, the shutdown is performed on that association only.
1455 * 4.1.6 close() - TCP Style Syntax
1457 * Applications use close() to gracefully close down an association.
1461 * int close(int sd);
1463 * sd - the socket descriptor of the association to be closed.
1465 * After an application calls close() on a socket descriptor, no further
1466 * socket operations will succeed on that descriptor.
1468 * API 7.1.4 SO_LINGER
1470 * An application using the TCP-style socket can use this option to
1471 * perform the SCTP ABORT primitive. The linger option structure is:
1474 * int l_onoff; // option on/off
1475 * int l_linger; // linger time
1478 * To enable the option, set l_onoff to 1. If the l_linger value is set
1479 * to 0, calling close() is the same as the ABORT primitive. If the
1480 * value is set to a negative value, the setsockopt() call will return
1481 * an error. If the value is set to a positive value linger_time, the
1482 * close() can be blocked for at most linger_time ms. If the graceful
1483 * shutdown phase does not finish during this period, close() will
1484 * return but the graceful shutdown phase continues in the system.
1486 static void sctp_close(struct sock *sk, long timeout)
1488 struct net *net = sock_net(sk);
1489 struct sctp_endpoint *ep;
1490 struct sctp_association *asoc;
1491 struct list_head *pos, *temp;
1492 unsigned int data_was_unread;
1494 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1497 sk->sk_shutdown = SHUTDOWN_MASK;
1498 sk->sk_state = SCTP_SS_CLOSING;
1500 ep = sctp_sk(sk)->ep;
1502 /* Clean up any skbs sitting on the receive queue. */
1503 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1504 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1506 /* Walk all associations on an endpoint. */
1507 list_for_each_safe(pos, temp, &ep->asocs) {
1508 asoc = list_entry(pos, struct sctp_association, asocs);
1510 if (sctp_style(sk, TCP)) {
1511 /* A closed association can still be in the list if
1512 * it belongs to a TCP-style listening socket that is
1513 * not yet accepted. If so, free it. If not, send an
1514 * ABORT or SHUTDOWN based on the linger options.
1516 if (sctp_state(asoc, CLOSED)) {
1517 sctp_association_free(asoc);
1522 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1523 !skb_queue_empty(&asoc->ulpq.reasm) ||
1524 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1525 struct sctp_chunk *chunk;
1527 chunk = sctp_make_abort_user(asoc, NULL, 0);
1528 sctp_primitive_ABORT(net, asoc, chunk);
1530 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1533 /* On a TCP-style socket, block for at most linger_time if set. */
1534 if (sctp_style(sk, TCP) && timeout)
1535 sctp_wait_for_close(sk, timeout);
1537 /* This will run the backlog queue. */
1540 /* Supposedly, no process has access to the socket, but
1541 * the net layers still may.
1542 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1543 * held and that should be grabbed before socket lock.
1545 spin_lock_bh(&net->sctp.addr_wq_lock);
1548 /* Hold the sock, since sk_common_release() will put sock_put()
1549 * and we have just a little more cleanup.
1552 sk_common_release(sk);
1555 spin_unlock_bh(&net->sctp.addr_wq_lock);
1559 SCTP_DBG_OBJCNT_DEC(sock);
1562 /* Handle EPIPE error. */
1563 static int sctp_error(struct sock *sk, int flags, int err)
1566 err = sock_error(sk) ? : -EPIPE;
1567 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1568 send_sig(SIGPIPE, current, 0);
1572 /* API 3.1.3 sendmsg() - UDP Style Syntax
1574 * An application uses sendmsg() and recvmsg() calls to transmit data to
1575 * and receive data from its peer.
1577 * ssize_t sendmsg(int socket, const struct msghdr *message,
1580 * socket - the socket descriptor of the endpoint.
1581 * message - pointer to the msghdr structure which contains a single
1582 * user message and possibly some ancillary data.
1584 * See Section 5 for complete description of the data
1587 * flags - flags sent or received with the user message, see Section
1588 * 5 for complete description of the flags.
1590 * Note: This function could use a rewrite especially when explicit
1591 * connect support comes in.
1593 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1595 static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1597 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
1599 struct net *net = sock_net(sk);
1600 struct sctp_sock *sp;
1601 struct sctp_endpoint *ep;
1602 struct sctp_association *new_asoc = NULL, *asoc = NULL;
1603 struct sctp_transport *transport, *chunk_tp;
1604 struct sctp_chunk *chunk;
1606 struct sockaddr *msg_name = NULL;
1607 struct sctp_sndrcvinfo default_sinfo;
1608 struct sctp_sndrcvinfo *sinfo;
1609 struct sctp_initmsg *sinit;
1610 sctp_assoc_t associd = 0;
1611 sctp_cmsgs_t cmsgs = { NULL };
1613 bool fill_sinfo_ttl = false, wait_connect = false;
1614 struct sctp_datamsg *datamsg;
1615 int msg_flags = msg->msg_flags;
1616 __u16 sinfo_flags = 0;
1624 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1627 /* We cannot send a message over a TCP-style listening socket. */
1628 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1633 /* Parse out the SCTP CMSGs. */
1634 err = sctp_msghdr_parse(msg, &cmsgs);
1636 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1640 /* Fetch the destination address for this packet. This
1641 * address only selects the association--it is not necessarily
1642 * the address we will send to.
1643 * For a peeled-off socket, msg_name is ignored.
1645 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1646 int msg_namelen = msg->msg_namelen;
1648 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1653 if (msg_namelen > sizeof(to))
1654 msg_namelen = sizeof(to);
1655 memcpy(&to, msg->msg_name, msg_namelen);
1656 msg_name = msg->msg_name;
1660 if (cmsgs.sinfo != NULL) {
1661 memset(&default_sinfo, 0, sizeof(default_sinfo));
1662 default_sinfo.sinfo_stream = cmsgs.sinfo->snd_sid;
1663 default_sinfo.sinfo_flags = cmsgs.sinfo->snd_flags;
1664 default_sinfo.sinfo_ppid = cmsgs.sinfo->snd_ppid;
1665 default_sinfo.sinfo_context = cmsgs.sinfo->snd_context;
1666 default_sinfo.sinfo_assoc_id = cmsgs.sinfo->snd_assoc_id;
1668 sinfo = &default_sinfo;
1669 fill_sinfo_ttl = true;
1671 sinfo = cmsgs.srinfo;
1673 /* Did the user specify SNDINFO/SNDRCVINFO? */
1675 sinfo_flags = sinfo->sinfo_flags;
1676 associd = sinfo->sinfo_assoc_id;
1679 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1680 msg_len, sinfo_flags);
1682 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1683 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1688 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1689 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1690 * If SCTP_ABORT is set, the message length could be non zero with
1691 * the msg_iov set to the user abort reason.
1693 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1694 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1699 /* If SCTP_ADDR_OVER is set, there must be an address
1700 * specified in msg_name.
1702 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1709 pr_debug("%s: about to look up association\n", __func__);
1713 /* If a msg_name has been specified, assume this is to be used. */
1715 /* Look for a matching association on the endpoint. */
1716 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1718 /* If we could not find a matching association on the
1719 * endpoint, make sure that it is not a TCP-style
1720 * socket that already has an association or there is
1721 * no peeled-off association on another socket.
1724 ((sctp_style(sk, TCP) &&
1725 (sctp_sstate(sk, ESTABLISHED) ||
1726 sctp_sstate(sk, CLOSING))) ||
1727 sctp_endpoint_is_peeled_off(ep, &to))) {
1728 err = -EADDRNOTAVAIL;
1732 asoc = sctp_id2assoc(sk, associd);
1740 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1742 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1743 * socket that has an association in CLOSED state. This can
1744 * happen when an accepted socket has an association that is
1747 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1752 if (sinfo_flags & SCTP_EOF) {
1753 pr_debug("%s: shutting down association:%p\n",
1756 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1760 if (sinfo_flags & SCTP_ABORT) {
1762 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1768 pr_debug("%s: aborting association:%p\n",
1771 sctp_primitive_ABORT(net, asoc, chunk);
1777 /* Do we need to create the association? */
1779 pr_debug("%s: there is no association yet\n", __func__);
1781 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1786 /* Check for invalid stream against the stream counts,
1787 * either the default or the user specified stream counts.
1790 if (!sinit || !sinit->sinit_num_ostreams) {
1791 /* Check against the defaults. */
1792 if (sinfo->sinfo_stream >=
1793 sp->initmsg.sinit_num_ostreams) {
1798 /* Check against the requested. */
1799 if (sinfo->sinfo_stream >=
1800 sinit->sinit_num_ostreams) {
1808 * API 3.1.2 bind() - UDP Style Syntax
1809 * If a bind() or sctp_bindx() is not called prior to a
1810 * sendmsg() call that initiates a new association, the
1811 * system picks an ephemeral port and will choose an address
1812 * set equivalent to binding with a wildcard address.
1814 if (!ep->base.bind_addr.port) {
1815 if (sctp_autobind(sk)) {
1821 * If an unprivileged user inherits a one-to-many
1822 * style socket with open associations on a privileged
1823 * port, it MAY be permitted to accept new associations,
1824 * but it SHOULD NOT be permitted to open new
1827 if (ep->base.bind_addr.port < inet_prot_sock(net) &&
1828 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1834 scope = sctp_scope(&to);
1835 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1841 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1847 /* If the SCTP_INIT ancillary data is specified, set all
1848 * the association init values accordingly.
1851 if (sinit->sinit_num_ostreams) {
1852 asoc->c.sinit_num_ostreams =
1853 sinit->sinit_num_ostreams;
1855 if (sinit->sinit_max_instreams) {
1856 asoc->c.sinit_max_instreams =
1857 sinit->sinit_max_instreams;
1859 if (sinit->sinit_max_attempts) {
1860 asoc->max_init_attempts
1861 = sinit->sinit_max_attempts;
1863 if (sinit->sinit_max_init_timeo) {
1864 asoc->max_init_timeo =
1865 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1869 /* Prime the peer's transport structures. */
1870 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1877 /* ASSERT: we have a valid association at this point. */
1878 pr_debug("%s: we have a valid association\n", __func__);
1881 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1882 * one with some defaults.
1884 memset(&default_sinfo, 0, sizeof(default_sinfo));
1885 default_sinfo.sinfo_stream = asoc->default_stream;
1886 default_sinfo.sinfo_flags = asoc->default_flags;
1887 default_sinfo.sinfo_ppid = asoc->default_ppid;
1888 default_sinfo.sinfo_context = asoc->default_context;
1889 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1890 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1892 sinfo = &default_sinfo;
1893 } else if (fill_sinfo_ttl) {
1894 /* In case SNDINFO was specified, we still need to fill
1895 * it with a default ttl from the assoc here.
1897 sinfo->sinfo_timetolive = asoc->default_timetolive;
1900 /* API 7.1.7, the sndbuf size per association bounds the
1901 * maximum size of data that can be sent in a single send call.
1903 if (msg_len > sk->sk_sndbuf) {
1908 if (asoc->pmtu_pending)
1909 sctp_assoc_pending_pmtu(sk, asoc);
1911 /* If fragmentation is disabled and the message length exceeds the
1912 * association fragmentation point, return EMSGSIZE. The I-D
1913 * does not specify what this error is, but this looks like
1916 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1921 /* Check for invalid stream. */
1922 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1927 if (sctp_wspace(asoc) < msg_len)
1928 sctp_prsctp_prune(asoc, sinfo, msg_len - sctp_wspace(asoc));
1930 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1931 if (!sctp_wspace(asoc)) {
1932 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1937 /* If an address is passed with the sendto/sendmsg call, it is used
1938 * to override the primary destination address in the TCP model, or
1939 * when SCTP_ADDR_OVER flag is set in the UDP model.
1941 if ((sctp_style(sk, TCP) && msg_name) ||
1942 (sinfo_flags & SCTP_ADDR_OVER)) {
1943 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1951 /* Auto-connect, if we aren't connected already. */
1952 if (sctp_state(asoc, CLOSED)) {
1953 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1957 wait_connect = true;
1958 pr_debug("%s: we associated primitively\n", __func__);
1961 /* Break the message into multiple chunks of maximum size. */
1962 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
1963 if (IS_ERR(datamsg)) {
1964 err = PTR_ERR(datamsg);
1968 /* Now send the (possibly) fragmented message. */
1969 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1970 sctp_chunk_hold(chunk);
1972 /* Do accounting for the write space. */
1973 sctp_set_owner_w(chunk);
1975 chunk->transport = chunk_tp;
1978 /* Send it to the lower layers. Note: all chunks
1979 * must either fail or succeed. The lower layer
1980 * works that way today. Keep it that way or this
1983 err = sctp_primitive_SEND(net, asoc, datamsg);
1984 /* Did the lower layer accept the chunk? */
1986 sctp_datamsg_free(datamsg);
1990 pr_debug("%s: we sent primitively\n", __func__);
1992 sctp_datamsg_put(datamsg);
1995 if (unlikely(wait_connect)) {
1996 timeo = sock_sndtimeo(sk, msg_flags & MSG_DONTWAIT);
1997 sctp_wait_for_connect(asoc, &timeo);
2000 /* If we are already past ASSOCIATE, the lower
2001 * layers are responsible for association cleanup.
2007 sctp_association_free(asoc);
2012 return sctp_error(sk, msg_flags, err);
2019 err = sock_error(sk);
2029 /* This is an extended version of skb_pull() that removes the data from the
2030 * start of a skb even when data is spread across the list of skb's in the
2031 * frag_list. len specifies the total amount of data that needs to be removed.
2032 * when 'len' bytes could be removed from the skb, it returns 0.
2033 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2034 * could not be removed.
2036 static int sctp_skb_pull(struct sk_buff *skb, int len)
2038 struct sk_buff *list;
2039 int skb_len = skb_headlen(skb);
2042 if (len <= skb_len) {
2043 __skb_pull(skb, len);
2047 __skb_pull(skb, skb_len);
2049 skb_walk_frags(skb, list) {
2050 rlen = sctp_skb_pull(list, len);
2051 skb->len -= (len-rlen);
2052 skb->data_len -= (len-rlen);
2063 /* API 3.1.3 recvmsg() - UDP Style Syntax
2065 * ssize_t recvmsg(int socket, struct msghdr *message,
2068 * socket - the socket descriptor of the endpoint.
2069 * message - pointer to the msghdr structure which contains a single
2070 * user message and possibly some ancillary data.
2072 * See Section 5 for complete description of the data
2075 * flags - flags sent or received with the user message, see Section
2076 * 5 for complete description of the flags.
2078 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2079 int noblock, int flags, int *addr_len)
2081 struct sctp_ulpevent *event = NULL;
2082 struct sctp_sock *sp = sctp_sk(sk);
2083 struct sk_buff *skb, *head_skb;
2088 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2089 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2094 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED) &&
2095 !sctp_sstate(sk, CLOSING) && !sctp_sstate(sk, CLOSED)) {
2100 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2104 /* Get the total length of the skb including any skb's in the
2113 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2115 event = sctp_skb2event(skb);
2120 if (event->chunk && event->chunk->head_skb)
2121 head_skb = event->chunk->head_skb;
2124 sock_recv_ts_and_drops(msg, sk, head_skb);
2125 if (sctp_ulpevent_is_notification(event)) {
2126 msg->msg_flags |= MSG_NOTIFICATION;
2127 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2129 sp->pf->skb_msgname(head_skb, msg->msg_name, addr_len);
2132 /* Check if we allow SCTP_NXTINFO. */
2133 if (sp->recvnxtinfo)
2134 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2135 /* Check if we allow SCTP_RCVINFO. */
2136 if (sp->recvrcvinfo)
2137 sctp_ulpevent_read_rcvinfo(event, msg);
2138 /* Check if we allow SCTP_SNDRCVINFO. */
2139 if (sp->subscribe.sctp_data_io_event)
2140 sctp_ulpevent_read_sndrcvinfo(event, msg);
2144 /* If skb's length exceeds the user's buffer, update the skb and
2145 * push it back to the receive_queue so that the next call to
2146 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2148 if (skb_len > copied) {
2149 msg->msg_flags &= ~MSG_EOR;
2150 if (flags & MSG_PEEK)
2152 sctp_skb_pull(skb, copied);
2153 skb_queue_head(&sk->sk_receive_queue, skb);
2155 /* When only partial message is copied to the user, increase
2156 * rwnd by that amount. If all the data in the skb is read,
2157 * rwnd is updated when the event is freed.
2159 if (!sctp_ulpevent_is_notification(event))
2160 sctp_assoc_rwnd_increase(event->asoc, copied);
2162 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2163 (event->msg_flags & MSG_EOR))
2164 msg->msg_flags |= MSG_EOR;
2166 msg->msg_flags &= ~MSG_EOR;
2169 if (flags & MSG_PEEK) {
2170 /* Release the skb reference acquired after peeking the skb in
2171 * sctp_skb_recv_datagram().
2175 /* Free the event which includes releasing the reference to
2176 * the owner of the skb, freeing the skb and updating the
2179 sctp_ulpevent_free(event);
2186 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2188 * This option is a on/off flag. If enabled no SCTP message
2189 * fragmentation will be performed. Instead if a message being sent
2190 * exceeds the current PMTU size, the message will NOT be sent and
2191 * instead a error will be indicated to the user.
2193 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2194 char __user *optval,
2195 unsigned int optlen)
2199 if (optlen < sizeof(int))
2202 if (get_user(val, (int __user *)optval))
2205 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2210 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2211 unsigned int optlen)
2213 struct sctp_association *asoc;
2214 struct sctp_ulpevent *event;
2216 if (optlen > sizeof(struct sctp_event_subscribe))
2218 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2221 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2222 * if there is no data to be sent or retransmit, the stack will
2223 * immediately send up this notification.
2225 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2226 &sctp_sk(sk)->subscribe)) {
2227 asoc = sctp_id2assoc(sk, 0);
2229 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2230 event = sctp_ulpevent_make_sender_dry_event(asoc,
2235 sctp_ulpq_tail_event(&asoc->ulpq, event);
2242 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2244 * This socket option is applicable to the UDP-style socket only. When
2245 * set it will cause associations that are idle for more than the
2246 * specified number of seconds to automatically close. An association
2247 * being idle is defined an association that has NOT sent or received
2248 * user data. The special value of '0' indicates that no automatic
2249 * close of any associations should be performed. The option expects an
2250 * integer defining the number of seconds of idle time before an
2251 * association is closed.
2253 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2254 unsigned int optlen)
2256 struct sctp_sock *sp = sctp_sk(sk);
2257 struct net *net = sock_net(sk);
2259 /* Applicable to UDP-style socket only */
2260 if (sctp_style(sk, TCP))
2262 if (optlen != sizeof(int))
2264 if (copy_from_user(&sp->autoclose, optval, optlen))
2267 if (sp->autoclose > net->sctp.max_autoclose)
2268 sp->autoclose = net->sctp.max_autoclose;
2273 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2275 * Applications can enable or disable heartbeats for any peer address of
2276 * an association, modify an address's heartbeat interval, force a
2277 * heartbeat to be sent immediately, and adjust the address's maximum
2278 * number of retransmissions sent before an address is considered
2279 * unreachable. The following structure is used to access and modify an
2280 * address's parameters:
2282 * struct sctp_paddrparams {
2283 * sctp_assoc_t spp_assoc_id;
2284 * struct sockaddr_storage spp_address;
2285 * uint32_t spp_hbinterval;
2286 * uint16_t spp_pathmaxrxt;
2287 * uint32_t spp_pathmtu;
2288 * uint32_t spp_sackdelay;
2289 * uint32_t spp_flags;
2292 * spp_assoc_id - (one-to-many style socket) This is filled in the
2293 * application, and identifies the association for
2295 * spp_address - This specifies which address is of interest.
2296 * spp_hbinterval - This contains the value of the heartbeat interval,
2297 * in milliseconds. If a value of zero
2298 * is present in this field then no changes are to
2299 * be made to this parameter.
2300 * spp_pathmaxrxt - This contains the maximum number of
2301 * retransmissions before this address shall be
2302 * considered unreachable. If a value of zero
2303 * is present in this field then no changes are to
2304 * be made to this parameter.
2305 * spp_pathmtu - When Path MTU discovery is disabled the value
2306 * specified here will be the "fixed" path mtu.
2307 * Note that if the spp_address field is empty
2308 * then all associations on this address will
2309 * have this fixed path mtu set upon them.
2311 * spp_sackdelay - When delayed sack is enabled, this value specifies
2312 * the number of milliseconds that sacks will be delayed
2313 * for. This value will apply to all addresses of an
2314 * association if the spp_address field is empty. Note
2315 * also, that if delayed sack is enabled and this
2316 * value is set to 0, no change is made to the last
2317 * recorded delayed sack timer value.
2319 * spp_flags - These flags are used to control various features
2320 * on an association. The flag field may contain
2321 * zero or more of the following options.
2323 * SPP_HB_ENABLE - Enable heartbeats on the
2324 * specified address. Note that if the address
2325 * field is empty all addresses for the association
2326 * have heartbeats enabled upon them.
2328 * SPP_HB_DISABLE - Disable heartbeats on the
2329 * speicifed address. Note that if the address
2330 * field is empty all addresses for the association
2331 * will have their heartbeats disabled. Note also
2332 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2333 * mutually exclusive, only one of these two should
2334 * be specified. Enabling both fields will have
2335 * undetermined results.
2337 * SPP_HB_DEMAND - Request a user initiated heartbeat
2338 * to be made immediately.
2340 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2341 * heartbeat delayis to be set to the value of 0
2344 * SPP_PMTUD_ENABLE - This field will enable PMTU
2345 * discovery upon the specified address. Note that
2346 * if the address feild is empty then all addresses
2347 * on the association are effected.
2349 * SPP_PMTUD_DISABLE - This field will disable PMTU
2350 * discovery upon the specified address. Note that
2351 * if the address feild is empty then all addresses
2352 * on the association are effected. Not also that
2353 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2354 * exclusive. Enabling both will have undetermined
2357 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2358 * on delayed sack. The time specified in spp_sackdelay
2359 * is used to specify the sack delay for this address. Note
2360 * that if spp_address is empty then all addresses will
2361 * enable delayed sack and take on the sack delay
2362 * value specified in spp_sackdelay.
2363 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2364 * off delayed sack. If the spp_address field is blank then
2365 * delayed sack is disabled for the entire association. Note
2366 * also that this field is mutually exclusive to
2367 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2370 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2371 struct sctp_transport *trans,
2372 struct sctp_association *asoc,
2373 struct sctp_sock *sp,
2376 int sackdelay_change)
2380 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2381 struct net *net = sock_net(trans->asoc->base.sk);
2383 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2388 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2389 * this field is ignored. Note also that a value of zero indicates
2390 * the current setting should be left unchanged.
2392 if (params->spp_flags & SPP_HB_ENABLE) {
2394 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2395 * set. This lets us use 0 value when this flag
2398 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2399 params->spp_hbinterval = 0;
2401 if (params->spp_hbinterval ||
2402 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2405 msecs_to_jiffies(params->spp_hbinterval);
2408 msecs_to_jiffies(params->spp_hbinterval);
2410 sp->hbinterval = params->spp_hbinterval;
2417 trans->param_flags =
2418 (trans->param_flags & ~SPP_HB) | hb_change;
2421 (asoc->param_flags & ~SPP_HB) | hb_change;
2424 (sp->param_flags & ~SPP_HB) | hb_change;
2428 /* When Path MTU discovery is disabled the value specified here will
2429 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2430 * include the flag SPP_PMTUD_DISABLE for this field to have any
2433 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2435 trans->pathmtu = params->spp_pathmtu;
2436 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2438 asoc->pathmtu = params->spp_pathmtu;
2440 sp->pathmtu = params->spp_pathmtu;
2446 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2447 (params->spp_flags & SPP_PMTUD_ENABLE);
2448 trans->param_flags =
2449 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2451 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2452 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2456 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2459 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2463 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2464 * value of this field is ignored. Note also that a value of zero
2465 * indicates the current setting should be left unchanged.
2467 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2470 msecs_to_jiffies(params->spp_sackdelay);
2473 msecs_to_jiffies(params->spp_sackdelay);
2475 sp->sackdelay = params->spp_sackdelay;
2479 if (sackdelay_change) {
2481 trans->param_flags =
2482 (trans->param_flags & ~SPP_SACKDELAY) |
2486 (asoc->param_flags & ~SPP_SACKDELAY) |
2490 (sp->param_flags & ~SPP_SACKDELAY) |
2495 /* Note that a value of zero indicates the current setting should be
2498 if (params->spp_pathmaxrxt) {
2500 trans->pathmaxrxt = params->spp_pathmaxrxt;
2502 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2504 sp->pathmaxrxt = params->spp_pathmaxrxt;
2511 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2512 char __user *optval,
2513 unsigned int optlen)
2515 struct sctp_paddrparams params;
2516 struct sctp_transport *trans = NULL;
2517 struct sctp_association *asoc = NULL;
2518 struct sctp_sock *sp = sctp_sk(sk);
2520 int hb_change, pmtud_change, sackdelay_change;
2522 if (optlen != sizeof(struct sctp_paddrparams))
2525 if (copy_from_user(¶ms, optval, optlen))
2528 /* Validate flags and value parameters. */
2529 hb_change = params.spp_flags & SPP_HB;
2530 pmtud_change = params.spp_flags & SPP_PMTUD;
2531 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2533 if (hb_change == SPP_HB ||
2534 pmtud_change == SPP_PMTUD ||
2535 sackdelay_change == SPP_SACKDELAY ||
2536 params.spp_sackdelay > 500 ||
2537 (params.spp_pathmtu &&
2538 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2541 /* If an address other than INADDR_ANY is specified, and
2542 * no transport is found, then the request is invalid.
2544 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
2545 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2546 params.spp_assoc_id);
2551 /* Get association, if assoc_id != 0 and the socket is a one
2552 * to many style socket, and an association was not found, then
2553 * the id was invalid.
2555 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2556 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2559 /* Heartbeat demand can only be sent on a transport or
2560 * association, but not a socket.
2562 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2565 /* Process parameters. */
2566 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2567 hb_change, pmtud_change,
2573 /* If changes are for association, also apply parameters to each
2576 if (!trans && asoc) {
2577 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2579 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2580 hb_change, pmtud_change,
2588 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2590 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2593 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2595 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2599 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2601 * This option will effect the way delayed acks are performed. This
2602 * option allows you to get or set the delayed ack time, in
2603 * milliseconds. It also allows changing the delayed ack frequency.
2604 * Changing the frequency to 1 disables the delayed sack algorithm. If
2605 * the assoc_id is 0, then this sets or gets the endpoints default
2606 * values. If the assoc_id field is non-zero, then the set or get
2607 * effects the specified association for the one to many model (the
2608 * assoc_id field is ignored by the one to one model). Note that if
2609 * sack_delay or sack_freq are 0 when setting this option, then the
2610 * current values will remain unchanged.
2612 * struct sctp_sack_info {
2613 * sctp_assoc_t sack_assoc_id;
2614 * uint32_t sack_delay;
2615 * uint32_t sack_freq;
2618 * sack_assoc_id - This parameter, indicates which association the user
2619 * is performing an action upon. Note that if this field's value is
2620 * zero then the endpoints default value is changed (effecting future
2621 * associations only).
2623 * sack_delay - This parameter contains the number of milliseconds that
2624 * the user is requesting the delayed ACK timer be set to. Note that
2625 * this value is defined in the standard to be between 200 and 500
2628 * sack_freq - This parameter contains the number of packets that must
2629 * be received before a sack is sent without waiting for the delay
2630 * timer to expire. The default value for this is 2, setting this
2631 * value to 1 will disable the delayed sack algorithm.
2634 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2635 char __user *optval, unsigned int optlen)
2637 struct sctp_sack_info params;
2638 struct sctp_transport *trans = NULL;
2639 struct sctp_association *asoc = NULL;
2640 struct sctp_sock *sp = sctp_sk(sk);
2642 if (optlen == sizeof(struct sctp_sack_info)) {
2643 if (copy_from_user(¶ms, optval, optlen))
2646 if (params.sack_delay == 0 && params.sack_freq == 0)
2648 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2649 pr_warn_ratelimited(DEPRECATED
2651 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2652 "Use struct sctp_sack_info instead\n",
2653 current->comm, task_pid_nr(current));
2654 if (copy_from_user(¶ms, optval, optlen))
2657 if (params.sack_delay == 0)
2658 params.sack_freq = 1;
2660 params.sack_freq = 0;
2664 /* Validate value parameter. */
2665 if (params.sack_delay > 500)
2668 /* Get association, if sack_assoc_id != 0 and the socket is a one
2669 * to many style socket, and an association was not found, then
2670 * the id was invalid.
2672 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2673 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2676 if (params.sack_delay) {
2679 msecs_to_jiffies(params.sack_delay);
2681 sctp_spp_sackdelay_enable(asoc->param_flags);
2683 sp->sackdelay = params.sack_delay;
2685 sctp_spp_sackdelay_enable(sp->param_flags);
2689 if (params.sack_freq == 1) {
2692 sctp_spp_sackdelay_disable(asoc->param_flags);
2695 sctp_spp_sackdelay_disable(sp->param_flags);
2697 } else if (params.sack_freq > 1) {
2699 asoc->sackfreq = params.sack_freq;
2701 sctp_spp_sackdelay_enable(asoc->param_flags);
2703 sp->sackfreq = params.sack_freq;
2705 sctp_spp_sackdelay_enable(sp->param_flags);
2709 /* If change is for association, also apply to each transport. */
2711 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2713 if (params.sack_delay) {
2715 msecs_to_jiffies(params.sack_delay);
2716 trans->param_flags =
2717 sctp_spp_sackdelay_enable(trans->param_flags);
2719 if (params.sack_freq == 1) {
2720 trans->param_flags =
2721 sctp_spp_sackdelay_disable(trans->param_flags);
2722 } else if (params.sack_freq > 1) {
2723 trans->sackfreq = params.sack_freq;
2724 trans->param_flags =
2725 sctp_spp_sackdelay_enable(trans->param_flags);
2733 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2735 * Applications can specify protocol parameters for the default association
2736 * initialization. The option name argument to setsockopt() and getsockopt()
2739 * Setting initialization parameters is effective only on an unconnected
2740 * socket (for UDP-style sockets only future associations are effected
2741 * by the change). With TCP-style sockets, this option is inherited by
2742 * sockets derived from a listener socket.
2744 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2746 struct sctp_initmsg sinit;
2747 struct sctp_sock *sp = sctp_sk(sk);
2749 if (optlen != sizeof(struct sctp_initmsg))
2751 if (copy_from_user(&sinit, optval, optlen))
2754 if (sinit.sinit_num_ostreams)
2755 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2756 if (sinit.sinit_max_instreams)
2757 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2758 if (sinit.sinit_max_attempts)
2759 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2760 if (sinit.sinit_max_init_timeo)
2761 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2767 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2769 * Applications that wish to use the sendto() system call may wish to
2770 * specify a default set of parameters that would normally be supplied
2771 * through the inclusion of ancillary data. This socket option allows
2772 * such an application to set the default sctp_sndrcvinfo structure.
2773 * The application that wishes to use this socket option simply passes
2774 * in to this call the sctp_sndrcvinfo structure defined in Section
2775 * 5.2.2) The input parameters accepted by this call include
2776 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2777 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2778 * to this call if the caller is using the UDP model.
2780 static int sctp_setsockopt_default_send_param(struct sock *sk,
2781 char __user *optval,
2782 unsigned int optlen)
2784 struct sctp_sock *sp = sctp_sk(sk);
2785 struct sctp_association *asoc;
2786 struct sctp_sndrcvinfo info;
2788 if (optlen != sizeof(info))
2790 if (copy_from_user(&info, optval, optlen))
2792 if (info.sinfo_flags &
2793 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2794 SCTP_ABORT | SCTP_EOF))
2797 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2798 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2801 asoc->default_stream = info.sinfo_stream;
2802 asoc->default_flags = info.sinfo_flags;
2803 asoc->default_ppid = info.sinfo_ppid;
2804 asoc->default_context = info.sinfo_context;
2805 asoc->default_timetolive = info.sinfo_timetolive;
2807 sp->default_stream = info.sinfo_stream;
2808 sp->default_flags = info.sinfo_flags;
2809 sp->default_ppid = info.sinfo_ppid;
2810 sp->default_context = info.sinfo_context;
2811 sp->default_timetolive = info.sinfo_timetolive;
2817 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2818 * (SCTP_DEFAULT_SNDINFO)
2820 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
2821 char __user *optval,
2822 unsigned int optlen)
2824 struct sctp_sock *sp = sctp_sk(sk);
2825 struct sctp_association *asoc;
2826 struct sctp_sndinfo info;
2828 if (optlen != sizeof(info))
2830 if (copy_from_user(&info, optval, optlen))
2832 if (info.snd_flags &
2833 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2834 SCTP_ABORT | SCTP_EOF))
2837 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
2838 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
2841 asoc->default_stream = info.snd_sid;
2842 asoc->default_flags = info.snd_flags;
2843 asoc->default_ppid = info.snd_ppid;
2844 asoc->default_context = info.snd_context;
2846 sp->default_stream = info.snd_sid;
2847 sp->default_flags = info.snd_flags;
2848 sp->default_ppid = info.snd_ppid;
2849 sp->default_context = info.snd_context;
2855 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2857 * Requests that the local SCTP stack use the enclosed peer address as
2858 * the association primary. The enclosed address must be one of the
2859 * association peer's addresses.
2861 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2862 unsigned int optlen)
2864 struct sctp_prim prim;
2865 struct sctp_transport *trans;
2867 if (optlen != sizeof(struct sctp_prim))
2870 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2873 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2877 sctp_assoc_set_primary(trans->asoc, trans);
2883 * 7.1.5 SCTP_NODELAY
2885 * Turn on/off any Nagle-like algorithm. This means that packets are
2886 * generally sent as soon as possible and no unnecessary delays are
2887 * introduced, at the cost of more packets in the network. Expects an
2888 * integer boolean flag.
2890 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2891 unsigned int optlen)
2895 if (optlen < sizeof(int))
2897 if (get_user(val, (int __user *)optval))
2900 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2906 * 7.1.1 SCTP_RTOINFO
2908 * The protocol parameters used to initialize and bound retransmission
2909 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2910 * and modify these parameters.
2911 * All parameters are time values, in milliseconds. A value of 0, when
2912 * modifying the parameters, indicates that the current value should not
2916 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2918 struct sctp_rtoinfo rtoinfo;
2919 struct sctp_association *asoc;
2920 unsigned long rto_min, rto_max;
2921 struct sctp_sock *sp = sctp_sk(sk);
2923 if (optlen != sizeof (struct sctp_rtoinfo))
2926 if (copy_from_user(&rtoinfo, optval, optlen))
2929 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2931 /* Set the values to the specific association */
2932 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2935 rto_max = rtoinfo.srto_max;
2936 rto_min = rtoinfo.srto_min;
2939 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2941 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2944 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2946 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2948 if (rto_min > rto_max)
2952 if (rtoinfo.srto_initial != 0)
2954 msecs_to_jiffies(rtoinfo.srto_initial);
2955 asoc->rto_max = rto_max;
2956 asoc->rto_min = rto_min;
2958 /* If there is no association or the association-id = 0
2959 * set the values to the endpoint.
2961 if (rtoinfo.srto_initial != 0)
2962 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2963 sp->rtoinfo.srto_max = rto_max;
2964 sp->rtoinfo.srto_min = rto_min;
2972 * 7.1.2 SCTP_ASSOCINFO
2974 * This option is used to tune the maximum retransmission attempts
2975 * of the association.
2976 * Returns an error if the new association retransmission value is
2977 * greater than the sum of the retransmission value of the peer.
2978 * See [SCTP] for more information.
2981 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2984 struct sctp_assocparams assocparams;
2985 struct sctp_association *asoc;
2987 if (optlen != sizeof(struct sctp_assocparams))
2989 if (copy_from_user(&assocparams, optval, optlen))
2992 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2994 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2997 /* Set the values to the specific association */
2999 if (assocparams.sasoc_asocmaxrxt != 0) {
3002 struct sctp_transport *peer_addr;
3004 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
3006 path_sum += peer_addr->pathmaxrxt;
3010 /* Only validate asocmaxrxt if we have more than
3011 * one path/transport. We do this because path
3012 * retransmissions are only counted when we have more
3016 assocparams.sasoc_asocmaxrxt > path_sum)
3019 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3022 if (assocparams.sasoc_cookie_life != 0)
3023 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3025 /* Set the values to the endpoint */
3026 struct sctp_sock *sp = sctp_sk(sk);
3028 if (assocparams.sasoc_asocmaxrxt != 0)
3029 sp->assocparams.sasoc_asocmaxrxt =
3030 assocparams.sasoc_asocmaxrxt;
3031 if (assocparams.sasoc_cookie_life != 0)
3032 sp->assocparams.sasoc_cookie_life =
3033 assocparams.sasoc_cookie_life;
3039 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3041 * This socket option is a boolean flag which turns on or off mapped V4
3042 * addresses. If this option is turned on and the socket is type
3043 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3044 * If this option is turned off, then no mapping will be done of V4
3045 * addresses and a user will receive both PF_INET6 and PF_INET type
3046 * addresses on the socket.
3048 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3051 struct sctp_sock *sp = sctp_sk(sk);
3053 if (optlen < sizeof(int))
3055 if (get_user(val, (int __user *)optval))
3066 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3067 * This option will get or set the maximum size to put in any outgoing
3068 * SCTP DATA chunk. If a message is larger than this size it will be
3069 * fragmented by SCTP into the specified size. Note that the underlying
3070 * SCTP implementation may fragment into smaller sized chunks when the
3071 * PMTU of the underlying association is smaller than the value set by
3072 * the user. The default value for this option is '0' which indicates
3073 * the user is NOT limiting fragmentation and only the PMTU will effect
3074 * SCTP's choice of DATA chunk size. Note also that values set larger
3075 * than the maximum size of an IP datagram will effectively let SCTP
3076 * control fragmentation (i.e. the same as setting this option to 0).
3078 * The following structure is used to access and modify this parameter:
3080 * struct sctp_assoc_value {
3081 * sctp_assoc_t assoc_id;
3082 * uint32_t assoc_value;
3085 * assoc_id: This parameter is ignored for one-to-one style sockets.
3086 * For one-to-many style sockets this parameter indicates which
3087 * association the user is performing an action upon. Note that if
3088 * this field's value is zero then the endpoints default value is
3089 * changed (effecting future associations only).
3090 * assoc_value: This parameter specifies the maximum size in bytes.
3092 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3094 struct sctp_assoc_value params;
3095 struct sctp_association *asoc;
3096 struct sctp_sock *sp = sctp_sk(sk);
3099 if (optlen == sizeof(int)) {
3100 pr_warn_ratelimited(DEPRECATED
3102 "Use of int in maxseg socket option.\n"
3103 "Use struct sctp_assoc_value instead\n",
3104 current->comm, task_pid_nr(current));
3105 if (copy_from_user(&val, optval, optlen))
3107 params.assoc_id = 0;
3108 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3109 if (copy_from_user(¶ms, optval, optlen))
3111 val = params.assoc_value;
3115 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3118 asoc = sctp_id2assoc(sk, params.assoc_id);
3119 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3124 val = asoc->pathmtu;
3125 val -= sp->pf->af->net_header_len;
3126 val -= sizeof(struct sctphdr) +
3127 sizeof(struct sctp_data_chunk);
3129 asoc->user_frag = val;
3130 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3132 sp->user_frag = val;
3140 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3142 * Requests that the peer mark the enclosed address as the association
3143 * primary. The enclosed address must be one of the association's
3144 * locally bound addresses. The following structure is used to make a
3145 * set primary request:
3147 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3148 unsigned int optlen)
3150 struct net *net = sock_net(sk);
3151 struct sctp_sock *sp;
3152 struct sctp_association *asoc = NULL;
3153 struct sctp_setpeerprim prim;
3154 struct sctp_chunk *chunk;
3160 if (!net->sctp.addip_enable)
3163 if (optlen != sizeof(struct sctp_setpeerprim))
3166 if (copy_from_user(&prim, optval, optlen))
3169 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3173 if (!asoc->peer.asconf_capable)
3176 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3179 if (!sctp_state(asoc, ESTABLISHED))
3182 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3186 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3187 return -EADDRNOTAVAIL;
3189 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3190 return -EADDRNOTAVAIL;
3192 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3193 chunk = sctp_make_asconf_set_prim(asoc,
3194 (union sctp_addr *)&prim.sspp_addr);
3198 err = sctp_send_asconf(asoc, chunk);
3200 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3205 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3206 unsigned int optlen)
3208 struct sctp_setadaptation adaptation;
3210 if (optlen != sizeof(struct sctp_setadaptation))
3212 if (copy_from_user(&adaptation, optval, optlen))
3215 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3221 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3223 * The context field in the sctp_sndrcvinfo structure is normally only
3224 * used when a failed message is retrieved holding the value that was
3225 * sent down on the actual send call. This option allows the setting of
3226 * a default context on an association basis that will be received on
3227 * reading messages from the peer. This is especially helpful in the
3228 * one-2-many model for an application to keep some reference to an
3229 * internal state machine that is processing messages on the
3230 * association. Note that the setting of this value only effects
3231 * received messages from the peer and does not effect the value that is
3232 * saved with outbound messages.
3234 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3235 unsigned int optlen)
3237 struct sctp_assoc_value params;
3238 struct sctp_sock *sp;
3239 struct sctp_association *asoc;
3241 if (optlen != sizeof(struct sctp_assoc_value))
3243 if (copy_from_user(¶ms, optval, optlen))
3248 if (params.assoc_id != 0) {
3249 asoc = sctp_id2assoc(sk, params.assoc_id);
3252 asoc->default_rcv_context = params.assoc_value;
3254 sp->default_rcv_context = params.assoc_value;
3261 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3263 * This options will at a minimum specify if the implementation is doing
3264 * fragmented interleave. Fragmented interleave, for a one to many
3265 * socket, is when subsequent calls to receive a message may return
3266 * parts of messages from different associations. Some implementations
3267 * may allow you to turn this value on or off. If so, when turned off,
3268 * no fragment interleave will occur (which will cause a head of line
3269 * blocking amongst multiple associations sharing the same one to many
3270 * socket). When this option is turned on, then each receive call may
3271 * come from a different association (thus the user must receive data
3272 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3273 * association each receive belongs to.
3275 * This option takes a boolean value. A non-zero value indicates that
3276 * fragmented interleave is on. A value of zero indicates that
3277 * fragmented interleave is off.
3279 * Note that it is important that an implementation that allows this
3280 * option to be turned on, have it off by default. Otherwise an unaware
3281 * application using the one to many model may become confused and act
3284 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3285 char __user *optval,
3286 unsigned int optlen)
3290 if (optlen != sizeof(int))
3292 if (get_user(val, (int __user *)optval))
3295 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3301 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3302 * (SCTP_PARTIAL_DELIVERY_POINT)
3304 * This option will set or get the SCTP partial delivery point. This
3305 * point is the size of a message where the partial delivery API will be
3306 * invoked to help free up rwnd space for the peer. Setting this to a
3307 * lower value will cause partial deliveries to happen more often. The
3308 * calls argument is an integer that sets or gets the partial delivery
3309 * point. Note also that the call will fail if the user attempts to set
3310 * this value larger than the socket receive buffer size.
3312 * Note that any single message having a length smaller than or equal to
3313 * the SCTP partial delivery point will be delivered in one single read
3314 * call as long as the user provided buffer is large enough to hold the
3317 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3318 char __user *optval,
3319 unsigned int optlen)
3323 if (optlen != sizeof(u32))
3325 if (get_user(val, (int __user *)optval))
3328 /* Note: We double the receive buffer from what the user sets
3329 * it to be, also initial rwnd is based on rcvbuf/2.
3331 if (val > (sk->sk_rcvbuf >> 1))
3334 sctp_sk(sk)->pd_point = val;
3336 return 0; /* is this the right error code? */
3340 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3342 * This option will allow a user to change the maximum burst of packets
3343 * that can be emitted by this association. Note that the default value
3344 * is 4, and some implementations may restrict this setting so that it
3345 * can only be lowered.
3347 * NOTE: This text doesn't seem right. Do this on a socket basis with
3348 * future associations inheriting the socket value.
3350 static int sctp_setsockopt_maxburst(struct sock *sk,
3351 char __user *optval,
3352 unsigned int optlen)
3354 struct sctp_assoc_value params;
3355 struct sctp_sock *sp;
3356 struct sctp_association *asoc;
3360 if (optlen == sizeof(int)) {
3361 pr_warn_ratelimited(DEPRECATED
3363 "Use of int in max_burst socket option deprecated.\n"
3364 "Use struct sctp_assoc_value instead\n",
3365 current->comm, task_pid_nr(current));
3366 if (copy_from_user(&val, optval, optlen))
3368 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3369 if (copy_from_user(¶ms, optval, optlen))
3371 val = params.assoc_value;
3372 assoc_id = params.assoc_id;
3378 if (assoc_id != 0) {
3379 asoc = sctp_id2assoc(sk, assoc_id);
3382 asoc->max_burst = val;
3384 sp->max_burst = val;
3390 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3392 * This set option adds a chunk type that the user is requesting to be
3393 * received only in an authenticated way. Changes to the list of chunks
3394 * will only effect future associations on the socket.
3396 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3397 char __user *optval,
3398 unsigned int optlen)
3400 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3401 struct sctp_authchunk val;
3403 if (!ep->auth_enable)
3406 if (optlen != sizeof(struct sctp_authchunk))
3408 if (copy_from_user(&val, optval, optlen))
3411 switch (val.sauth_chunk) {
3413 case SCTP_CID_INIT_ACK:
3414 case SCTP_CID_SHUTDOWN_COMPLETE:
3419 /* add this chunk id to the endpoint */
3420 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3424 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3426 * This option gets or sets the list of HMAC algorithms that the local
3427 * endpoint requires the peer to use.
3429 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3430 char __user *optval,
3431 unsigned int optlen)
3433 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3434 struct sctp_hmacalgo *hmacs;
3438 if (!ep->auth_enable)
3441 if (optlen < sizeof(struct sctp_hmacalgo))
3444 hmacs = memdup_user(optval, optlen);
3446 return PTR_ERR(hmacs);
3448 idents = hmacs->shmac_num_idents;
3449 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3450 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3455 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3462 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3464 * This option will set a shared secret key which is used to build an
3465 * association shared key.
3467 static int sctp_setsockopt_auth_key(struct sock *sk,
3468 char __user *optval,
3469 unsigned int optlen)
3471 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3472 struct sctp_authkey *authkey;
3473 struct sctp_association *asoc;
3476 if (!ep->auth_enable)
3479 if (optlen <= sizeof(struct sctp_authkey))
3482 authkey = memdup_user(optval, optlen);
3483 if (IS_ERR(authkey))
3484 return PTR_ERR(authkey);
3486 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3491 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3492 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3497 ret = sctp_auth_set_key(ep, asoc, authkey);
3504 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3506 * This option will get or set the active shared key to be used to build
3507 * the association shared key.
3509 static int sctp_setsockopt_active_key(struct sock *sk,
3510 char __user *optval,
3511 unsigned int optlen)
3513 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3514 struct sctp_authkeyid val;
3515 struct sctp_association *asoc;
3517 if (!ep->auth_enable)
3520 if (optlen != sizeof(struct sctp_authkeyid))
3522 if (copy_from_user(&val, optval, optlen))
3525 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3526 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3529 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3533 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3535 * This set option will delete a shared secret key from use.
3537 static int sctp_setsockopt_del_key(struct sock *sk,
3538 char __user *optval,
3539 unsigned int optlen)
3541 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3542 struct sctp_authkeyid val;
3543 struct sctp_association *asoc;
3545 if (!ep->auth_enable)
3548 if (optlen != sizeof(struct sctp_authkeyid))
3550 if (copy_from_user(&val, optval, optlen))
3553 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3554 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3557 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3562 * 8.1.23 SCTP_AUTO_ASCONF
3564 * This option will enable or disable the use of the automatic generation of
3565 * ASCONF chunks to add and delete addresses to an existing association. Note
3566 * that this option has two caveats namely: a) it only affects sockets that
3567 * are bound to all addresses available to the SCTP stack, and b) the system
3568 * administrator may have an overriding control that turns the ASCONF feature
3569 * off no matter what setting the socket option may have.
3570 * This option expects an integer boolean flag, where a non-zero value turns on
3571 * the option, and a zero value turns off the option.
3572 * Note. In this implementation, socket operation overrides default parameter
3573 * being set by sysctl as well as FreeBSD implementation
3575 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3576 unsigned int optlen)
3579 struct sctp_sock *sp = sctp_sk(sk);
3581 if (optlen < sizeof(int))
3583 if (get_user(val, (int __user *)optval))
3585 if (!sctp_is_ep_boundall(sk) && val)
3587 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3590 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3591 if (val == 0 && sp->do_auto_asconf) {
3592 list_del(&sp->auto_asconf_list);
3593 sp->do_auto_asconf = 0;
3594 } else if (val && !sp->do_auto_asconf) {
3595 list_add_tail(&sp->auto_asconf_list,
3596 &sock_net(sk)->sctp.auto_asconf_splist);
3597 sp->do_auto_asconf = 1;
3599 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3604 * SCTP_PEER_ADDR_THLDS
3606 * This option allows us to alter the partially failed threshold for one or all
3607 * transports in an association. See Section 6.1 of:
3608 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3610 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3611 char __user *optval,
3612 unsigned int optlen)
3614 struct sctp_paddrthlds val;
3615 struct sctp_transport *trans;
3616 struct sctp_association *asoc;
3618 if (optlen < sizeof(struct sctp_paddrthlds))
3620 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3621 sizeof(struct sctp_paddrthlds)))
3625 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3626 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3629 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3631 if (val.spt_pathmaxrxt)
3632 trans->pathmaxrxt = val.spt_pathmaxrxt;
3633 trans->pf_retrans = val.spt_pathpfthld;
3636 if (val.spt_pathmaxrxt)
3637 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3638 asoc->pf_retrans = val.spt_pathpfthld;
3640 trans = sctp_addr_id2transport(sk, &val.spt_address,
3645 if (val.spt_pathmaxrxt)
3646 trans->pathmaxrxt = val.spt_pathmaxrxt;
3647 trans->pf_retrans = val.spt_pathpfthld;
3653 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3654 char __user *optval,
3655 unsigned int optlen)
3659 if (optlen < sizeof(int))
3661 if (get_user(val, (int __user *) optval))
3664 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3669 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3670 char __user *optval,
3671 unsigned int optlen)
3675 if (optlen < sizeof(int))
3677 if (get_user(val, (int __user *) optval))
3680 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3685 static int sctp_setsockopt_pr_supported(struct sock *sk,
3686 char __user *optval,
3687 unsigned int optlen)
3689 struct sctp_assoc_value params;
3690 struct sctp_association *asoc;
3691 int retval = -EINVAL;
3693 if (optlen != sizeof(params))
3696 if (copy_from_user(¶ms, optval, optlen)) {
3701 asoc = sctp_id2assoc(sk, params.assoc_id);
3703 asoc->prsctp_enable = !!params.assoc_value;
3704 } else if (!params.assoc_id) {
3705 struct sctp_sock *sp = sctp_sk(sk);
3707 sp->ep->prsctp_enable = !!params.assoc_value;
3718 static int sctp_setsockopt_default_prinfo(struct sock *sk,
3719 char __user *optval,
3720 unsigned int optlen)
3722 struct sctp_default_prinfo info;
3723 struct sctp_association *asoc;
3724 int retval = -EINVAL;
3726 if (optlen != sizeof(info))
3729 if (copy_from_user(&info, optval, sizeof(info))) {
3734 if (info.pr_policy & ~SCTP_PR_SCTP_MASK)
3737 if (info.pr_policy == SCTP_PR_SCTP_NONE)
3740 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
3742 SCTP_PR_SET_POLICY(asoc->default_flags, info.pr_policy);
3743 asoc->default_timetolive = info.pr_value;
3744 } else if (!info.pr_assoc_id) {
3745 struct sctp_sock *sp = sctp_sk(sk);
3747 SCTP_PR_SET_POLICY(sp->default_flags, info.pr_policy);
3748 sp->default_timetolive = info.pr_value;
3759 static int sctp_setsockopt_enable_strreset(struct sock *sk,
3760 char __user *optval,
3761 unsigned int optlen)
3763 struct sctp_assoc_value params;
3764 struct sctp_association *asoc;
3765 int retval = -EINVAL;
3767 if (optlen != sizeof(params))
3770 if (copy_from_user(¶ms, optval, optlen)) {
3775 if (params.assoc_value & (~SCTP_ENABLE_STRRESET_MASK))
3778 asoc = sctp_id2assoc(sk, params.assoc_id);
3780 asoc->strreset_enable = params.assoc_value;
3781 } else if (!params.assoc_id) {
3782 struct sctp_sock *sp = sctp_sk(sk);
3784 sp->ep->strreset_enable = params.assoc_value;
3795 static int sctp_setsockopt_reset_streams(struct sock *sk,
3796 char __user *optval,
3797 unsigned int optlen)
3799 struct sctp_reset_streams *params;
3800 struct sctp_association *asoc;
3801 int retval = -EINVAL;
3803 if (optlen < sizeof(struct sctp_reset_streams))
3806 params = memdup_user(optval, optlen);
3808 return PTR_ERR(params);
3810 asoc = sctp_id2assoc(sk, params->srs_assoc_id);
3814 retval = sctp_send_reset_streams(asoc, params);
3821 /* API 6.2 setsockopt(), getsockopt()
3823 * Applications use setsockopt() and getsockopt() to set or retrieve
3824 * socket options. Socket options are used to change the default
3825 * behavior of sockets calls. They are described in Section 7.
3829 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3830 * int __user *optlen);
3831 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3834 * sd - the socket descript.
3835 * level - set to IPPROTO_SCTP for all SCTP options.
3836 * optname - the option name.
3837 * optval - the buffer to store the value of the option.
3838 * optlen - the size of the buffer.
3840 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3841 char __user *optval, unsigned int optlen)
3845 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3847 /* I can hardly begin to describe how wrong this is. This is
3848 * so broken as to be worse than useless. The API draft
3849 * REALLY is NOT helpful here... I am not convinced that the
3850 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3851 * are at all well-founded.
3853 if (level != SOL_SCTP) {
3854 struct sctp_af *af = sctp_sk(sk)->pf->af;
3855 retval = af->setsockopt(sk, level, optname, optval, optlen);
3862 case SCTP_SOCKOPT_BINDX_ADD:
3863 /* 'optlen' is the size of the addresses buffer. */
3864 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3865 optlen, SCTP_BINDX_ADD_ADDR);
3868 case SCTP_SOCKOPT_BINDX_REM:
3869 /* 'optlen' is the size of the addresses buffer. */
3870 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3871 optlen, SCTP_BINDX_REM_ADDR);
3874 case SCTP_SOCKOPT_CONNECTX_OLD:
3875 /* 'optlen' is the size of the addresses buffer. */
3876 retval = sctp_setsockopt_connectx_old(sk,
3877 (struct sockaddr __user *)optval,
3881 case SCTP_SOCKOPT_CONNECTX:
3882 /* 'optlen' is the size of the addresses buffer. */
3883 retval = sctp_setsockopt_connectx(sk,
3884 (struct sockaddr __user *)optval,
3888 case SCTP_DISABLE_FRAGMENTS:
3889 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3893 retval = sctp_setsockopt_events(sk, optval, optlen);
3896 case SCTP_AUTOCLOSE:
3897 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3900 case SCTP_PEER_ADDR_PARAMS:
3901 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3904 case SCTP_DELAYED_SACK:
3905 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3907 case SCTP_PARTIAL_DELIVERY_POINT:
3908 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3912 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3914 case SCTP_DEFAULT_SEND_PARAM:
3915 retval = sctp_setsockopt_default_send_param(sk, optval,
3918 case SCTP_DEFAULT_SNDINFO:
3919 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
3921 case SCTP_PRIMARY_ADDR:
3922 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3924 case SCTP_SET_PEER_PRIMARY_ADDR:
3925 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3928 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3931 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3933 case SCTP_ASSOCINFO:
3934 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3936 case SCTP_I_WANT_MAPPED_V4_ADDR:
3937 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3940 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3942 case SCTP_ADAPTATION_LAYER:
3943 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3946 retval = sctp_setsockopt_context(sk, optval, optlen);
3948 case SCTP_FRAGMENT_INTERLEAVE:
3949 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3951 case SCTP_MAX_BURST:
3952 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3954 case SCTP_AUTH_CHUNK:
3955 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3957 case SCTP_HMAC_IDENT:
3958 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3961 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3963 case SCTP_AUTH_ACTIVE_KEY:
3964 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3966 case SCTP_AUTH_DELETE_KEY:
3967 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3969 case SCTP_AUTO_ASCONF:
3970 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3972 case SCTP_PEER_ADDR_THLDS:
3973 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3975 case SCTP_RECVRCVINFO:
3976 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
3978 case SCTP_RECVNXTINFO:
3979 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
3981 case SCTP_PR_SUPPORTED:
3982 retval = sctp_setsockopt_pr_supported(sk, optval, optlen);
3984 case SCTP_DEFAULT_PRINFO:
3985 retval = sctp_setsockopt_default_prinfo(sk, optval, optlen);
3987 case SCTP_ENABLE_STREAM_RESET:
3988 retval = sctp_setsockopt_enable_strreset(sk, optval, optlen);
3990 case SCTP_RESET_STREAMS:
3991 retval = sctp_setsockopt_reset_streams(sk, optval, optlen);
3994 retval = -ENOPROTOOPT;
4004 /* API 3.1.6 connect() - UDP Style Syntax
4006 * An application may use the connect() call in the UDP model to initiate an
4007 * association without sending data.
4011 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
4013 * sd: the socket descriptor to have a new association added to.
4015 * nam: the address structure (either struct sockaddr_in or struct
4016 * sockaddr_in6 defined in RFC2553 [7]).
4018 * len: the size of the address.
4020 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
4028 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
4031 /* Validate addr_len before calling common connect/connectx routine. */
4032 af = sctp_get_af_specific(addr->sa_family);
4033 if (!af || addr_len < af->sockaddr_len) {
4036 /* Pass correct addr len to common routine (so it knows there
4037 * is only one address being passed.
4039 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
4046 /* FIXME: Write comments. */
4047 static int sctp_disconnect(struct sock *sk, int flags)
4049 return -EOPNOTSUPP; /* STUB */
4052 /* 4.1.4 accept() - TCP Style Syntax
4054 * Applications use accept() call to remove an established SCTP
4055 * association from the accept queue of the endpoint. A new socket
4056 * descriptor will be returned from accept() to represent the newly
4057 * formed association.
4059 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
4061 struct sctp_sock *sp;
4062 struct sctp_endpoint *ep;
4063 struct sock *newsk = NULL;
4064 struct sctp_association *asoc;
4073 if (!sctp_style(sk, TCP)) {
4074 error = -EOPNOTSUPP;
4078 if (!sctp_sstate(sk, LISTENING)) {
4083 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
4085 error = sctp_wait_for_accept(sk, timeo);
4089 /* We treat the list of associations on the endpoint as the accept
4090 * queue and pick the first association on the list.
4092 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
4094 newsk = sp->pf->create_accept_sk(sk, asoc);
4100 /* Populate the fields of the newsk from the oldsk and migrate the
4101 * asoc to the newsk.
4103 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
4111 /* The SCTP ioctl handler. */
4112 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
4119 * SEQPACKET-style sockets in LISTENING state are valid, for
4120 * SCTP, so only discard TCP-style sockets in LISTENING state.
4122 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4127 struct sk_buff *skb;
4128 unsigned int amount = 0;
4130 skb = skb_peek(&sk->sk_receive_queue);
4133 * We will only return the amount of this packet since
4134 * that is all that will be read.
4138 rc = put_user(amount, (int __user *)arg);
4150 /* This is the function which gets called during socket creation to
4151 * initialized the SCTP-specific portion of the sock.
4152 * The sock structure should already be zero-filled memory.
4154 static int sctp_init_sock(struct sock *sk)
4156 struct net *net = sock_net(sk);
4157 struct sctp_sock *sp;
4159 pr_debug("%s: sk:%p\n", __func__, sk);
4163 /* Initialize the SCTP per socket area. */
4164 switch (sk->sk_type) {
4165 case SOCK_SEQPACKET:
4166 sp->type = SCTP_SOCKET_UDP;
4169 sp->type = SCTP_SOCKET_TCP;
4172 return -ESOCKTNOSUPPORT;
4175 sk->sk_gso_type = SKB_GSO_SCTP;
4177 /* Initialize default send parameters. These parameters can be
4178 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4180 sp->default_stream = 0;
4181 sp->default_ppid = 0;
4182 sp->default_flags = 0;
4183 sp->default_context = 0;
4184 sp->default_timetolive = 0;
4186 sp->default_rcv_context = 0;
4187 sp->max_burst = net->sctp.max_burst;
4189 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4191 /* Initialize default setup parameters. These parameters
4192 * can be modified with the SCTP_INITMSG socket option or
4193 * overridden by the SCTP_INIT CMSG.
4195 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4196 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4197 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4198 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4200 /* Initialize default RTO related parameters. These parameters can
4201 * be modified for with the SCTP_RTOINFO socket option.
4203 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4204 sp->rtoinfo.srto_max = net->sctp.rto_max;
4205 sp->rtoinfo.srto_min = net->sctp.rto_min;
4207 /* Initialize default association related parameters. These parameters
4208 * can be modified with the SCTP_ASSOCINFO socket option.
4210 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4211 sp->assocparams.sasoc_number_peer_destinations = 0;
4212 sp->assocparams.sasoc_peer_rwnd = 0;
4213 sp->assocparams.sasoc_local_rwnd = 0;
4214 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4216 /* Initialize default event subscriptions. By default, all the
4219 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4221 /* Default Peer Address Parameters. These defaults can
4222 * be modified via SCTP_PEER_ADDR_PARAMS
4224 sp->hbinterval = net->sctp.hb_interval;
4225 sp->pathmaxrxt = net->sctp.max_retrans_path;
4226 sp->pathmtu = 0; /* allow default discovery */
4227 sp->sackdelay = net->sctp.sack_timeout;
4229 sp->param_flags = SPP_HB_ENABLE |
4231 SPP_SACKDELAY_ENABLE;
4233 /* If enabled no SCTP message fragmentation will be performed.
4234 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4236 sp->disable_fragments = 0;
4238 /* Enable Nagle algorithm by default. */
4241 sp->recvrcvinfo = 0;
4242 sp->recvnxtinfo = 0;
4244 /* Enable by default. */
4247 /* Auto-close idle associations after the configured
4248 * number of seconds. A value of 0 disables this
4249 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4250 * for UDP-style sockets only.
4254 /* User specified fragmentation limit. */
4257 sp->adaptation_ind = 0;
4259 sp->pf = sctp_get_pf_specific(sk->sk_family);
4261 /* Control variables for partial data delivery. */
4262 atomic_set(&sp->pd_mode, 0);
4263 skb_queue_head_init(&sp->pd_lobby);
4264 sp->frag_interleave = 0;
4266 /* Create a per socket endpoint structure. Even if we
4267 * change the data structure relationships, this may still
4268 * be useful for storing pre-connect address information.
4270 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4276 sk->sk_destruct = sctp_destruct_sock;
4278 SCTP_DBG_OBJCNT_INC(sock);
4281 percpu_counter_inc(&sctp_sockets_allocated);
4282 sock_prot_inuse_add(net, sk->sk_prot, 1);
4284 /* Nothing can fail after this block, otherwise
4285 * sctp_destroy_sock() will be called without addr_wq_lock held
4287 if (net->sctp.default_auto_asconf) {
4288 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
4289 list_add_tail(&sp->auto_asconf_list,
4290 &net->sctp.auto_asconf_splist);
4291 sp->do_auto_asconf = 1;
4292 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4294 sp->do_auto_asconf = 0;
4302 /* Cleanup any SCTP per socket resources. Must be called with
4303 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4305 static void sctp_destroy_sock(struct sock *sk)
4307 struct sctp_sock *sp;
4309 pr_debug("%s: sk:%p\n", __func__, sk);
4311 /* Release our hold on the endpoint. */
4313 /* This could happen during socket init, thus we bail out
4314 * early, since the rest of the below is not setup either.
4319 if (sp->do_auto_asconf) {
4320 sp->do_auto_asconf = 0;
4321 list_del(&sp->auto_asconf_list);
4323 sctp_endpoint_free(sp->ep);
4325 percpu_counter_dec(&sctp_sockets_allocated);
4326 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4330 /* Triggered when there are no references on the socket anymore */
4331 static void sctp_destruct_sock(struct sock *sk)
4333 struct sctp_sock *sp = sctp_sk(sk);
4335 /* Free up the HMAC transform. */
4336 crypto_free_shash(sp->hmac);
4338 inet_sock_destruct(sk);
4341 /* API 4.1.7 shutdown() - TCP Style Syntax
4342 * int shutdown(int socket, int how);
4344 * sd - the socket descriptor of the association to be closed.
4345 * how - Specifies the type of shutdown. The values are
4348 * Disables further receive operations. No SCTP
4349 * protocol action is taken.
4351 * Disables further send operations, and initiates
4352 * the SCTP shutdown sequence.
4354 * Disables further send and receive operations
4355 * and initiates the SCTP shutdown sequence.
4357 static void sctp_shutdown(struct sock *sk, int how)
4359 struct net *net = sock_net(sk);
4360 struct sctp_endpoint *ep;
4362 if (!sctp_style(sk, TCP))
4365 ep = sctp_sk(sk)->ep;
4366 if (how & SEND_SHUTDOWN && !list_empty(&ep->asocs)) {
4367 struct sctp_association *asoc;
4369 sk->sk_state = SCTP_SS_CLOSING;
4370 asoc = list_entry(ep->asocs.next,
4371 struct sctp_association, asocs);
4372 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4376 int sctp_get_sctp_info(struct sock *sk, struct sctp_association *asoc,
4377 struct sctp_info *info)
4379 struct sctp_transport *prim;
4380 struct list_head *pos;
4383 memset(info, 0, sizeof(*info));
4385 struct sctp_sock *sp = sctp_sk(sk);
4387 info->sctpi_s_autoclose = sp->autoclose;
4388 info->sctpi_s_adaptation_ind = sp->adaptation_ind;
4389 info->sctpi_s_pd_point = sp->pd_point;
4390 info->sctpi_s_nodelay = sp->nodelay;
4391 info->sctpi_s_disable_fragments = sp->disable_fragments;
4392 info->sctpi_s_v4mapped = sp->v4mapped;
4393 info->sctpi_s_frag_interleave = sp->frag_interleave;
4394 info->sctpi_s_type = sp->type;
4399 info->sctpi_tag = asoc->c.my_vtag;
4400 info->sctpi_state = asoc->state;
4401 info->sctpi_rwnd = asoc->a_rwnd;
4402 info->sctpi_unackdata = asoc->unack_data;
4403 info->sctpi_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4404 info->sctpi_instrms = asoc->c.sinit_max_instreams;
4405 info->sctpi_outstrms = asoc->c.sinit_num_ostreams;
4406 list_for_each(pos, &asoc->base.inqueue.in_chunk_list)
4407 info->sctpi_inqueue++;
4408 list_for_each(pos, &asoc->outqueue.out_chunk_list)
4409 info->sctpi_outqueue++;
4410 info->sctpi_overall_error = asoc->overall_error_count;
4411 info->sctpi_max_burst = asoc->max_burst;
4412 info->sctpi_maxseg = asoc->frag_point;
4413 info->sctpi_peer_rwnd = asoc->peer.rwnd;
4414 info->sctpi_peer_tag = asoc->c.peer_vtag;
4416 mask = asoc->peer.ecn_capable << 1;
4417 mask = (mask | asoc->peer.ipv4_address) << 1;
4418 mask = (mask | asoc->peer.ipv6_address) << 1;
4419 mask = (mask | asoc->peer.hostname_address) << 1;
4420 mask = (mask | asoc->peer.asconf_capable) << 1;
4421 mask = (mask | asoc->peer.prsctp_capable) << 1;
4422 mask = (mask | asoc->peer.auth_capable);
4423 info->sctpi_peer_capable = mask;
4424 mask = asoc->peer.sack_needed << 1;
4425 mask = (mask | asoc->peer.sack_generation) << 1;
4426 mask = (mask | asoc->peer.zero_window_announced);
4427 info->sctpi_peer_sack = mask;
4429 info->sctpi_isacks = asoc->stats.isacks;
4430 info->sctpi_osacks = asoc->stats.osacks;
4431 info->sctpi_opackets = asoc->stats.opackets;
4432 info->sctpi_ipackets = asoc->stats.ipackets;
4433 info->sctpi_rtxchunks = asoc->stats.rtxchunks;
4434 info->sctpi_outofseqtsns = asoc->stats.outofseqtsns;
4435 info->sctpi_idupchunks = asoc->stats.idupchunks;
4436 info->sctpi_gapcnt = asoc->stats.gapcnt;
4437 info->sctpi_ouodchunks = asoc->stats.ouodchunks;
4438 info->sctpi_iuodchunks = asoc->stats.iuodchunks;
4439 info->sctpi_oodchunks = asoc->stats.oodchunks;
4440 info->sctpi_iodchunks = asoc->stats.iodchunks;
4441 info->sctpi_octrlchunks = asoc->stats.octrlchunks;
4442 info->sctpi_ictrlchunks = asoc->stats.ictrlchunks;
4444 prim = asoc->peer.primary_path;
4445 memcpy(&info->sctpi_p_address, &prim->ipaddr,
4446 sizeof(struct sockaddr_storage));
4447 info->sctpi_p_state = prim->state;
4448 info->sctpi_p_cwnd = prim->cwnd;
4449 info->sctpi_p_srtt = prim->srtt;
4450 info->sctpi_p_rto = jiffies_to_msecs(prim->rto);
4451 info->sctpi_p_hbinterval = prim->hbinterval;
4452 info->sctpi_p_pathmaxrxt = prim->pathmaxrxt;
4453 info->sctpi_p_sackdelay = jiffies_to_msecs(prim->sackdelay);
4454 info->sctpi_p_ssthresh = prim->ssthresh;
4455 info->sctpi_p_partial_bytes_acked = prim->partial_bytes_acked;
4456 info->sctpi_p_flight_size = prim->flight_size;
4457 info->sctpi_p_error = prim->error_count;
4461 EXPORT_SYMBOL_GPL(sctp_get_sctp_info);
4463 /* use callback to avoid exporting the core structure */
4464 int sctp_transport_walk_start(struct rhashtable_iter *iter)
4468 rhltable_walk_enter(&sctp_transport_hashtable, iter);
4470 err = rhashtable_walk_start(iter);
4471 if (err && err != -EAGAIN) {
4472 rhashtable_walk_stop(iter);
4473 rhashtable_walk_exit(iter);
4480 void sctp_transport_walk_stop(struct rhashtable_iter *iter)
4482 rhashtable_walk_stop(iter);
4483 rhashtable_walk_exit(iter);
4486 struct sctp_transport *sctp_transport_get_next(struct net *net,
4487 struct rhashtable_iter *iter)
4489 struct sctp_transport *t;
4491 t = rhashtable_walk_next(iter);
4492 for (; t; t = rhashtable_walk_next(iter)) {
4494 if (PTR_ERR(t) == -EAGAIN)
4499 if (net_eq(sock_net(t->asoc->base.sk), net) &&
4500 t->asoc->peer.primary_path == t)
4507 struct sctp_transport *sctp_transport_get_idx(struct net *net,
4508 struct rhashtable_iter *iter,
4511 void *obj = SEQ_START_TOKEN;
4513 while (pos && (obj = sctp_transport_get_next(net, iter)) &&
4520 int sctp_for_each_endpoint(int (*cb)(struct sctp_endpoint *, void *),
4524 struct sctp_ep_common *epb;
4525 struct sctp_hashbucket *head;
4527 for (head = sctp_ep_hashtable; hash < sctp_ep_hashsize;
4529 read_lock(&head->lock);
4530 sctp_for_each_hentry(epb, &head->chain) {
4531 err = cb(sctp_ep(epb), p);
4535 read_unlock(&head->lock);
4540 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint);
4542 int sctp_transport_lookup_process(int (*cb)(struct sctp_transport *, void *),
4544 const union sctp_addr *laddr,
4545 const union sctp_addr *paddr, void *p)
4547 struct sctp_transport *transport;
4551 transport = sctp_addrs_lookup_transport(net, laddr, paddr);
4556 err = cb(transport, p);
4557 sctp_transport_put(transport);
4561 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process);
4563 int sctp_for_each_transport(int (*cb)(struct sctp_transport *, void *),
4564 struct net *net, int pos, void *p) {
4565 struct rhashtable_iter hti;
4569 err = sctp_transport_walk_start(&hti);
4573 sctp_transport_get_idx(net, &hti, pos);
4574 obj = sctp_transport_get_next(net, &hti);
4575 for (; obj && !IS_ERR(obj); obj = sctp_transport_get_next(net, &hti)) {
4576 struct sctp_transport *transport = obj;
4578 if (!sctp_transport_hold(transport))
4580 err = cb(transport, p);
4581 sctp_transport_put(transport);
4585 sctp_transport_walk_stop(&hti);
4589 EXPORT_SYMBOL_GPL(sctp_for_each_transport);
4591 /* 7.2.1 Association Status (SCTP_STATUS)
4593 * Applications can retrieve current status information about an
4594 * association, including association state, peer receiver window size,
4595 * number of unacked data chunks, and number of data chunks pending
4596 * receipt. This information is read-only.
4598 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4599 char __user *optval,
4602 struct sctp_status status;
4603 struct sctp_association *asoc = NULL;
4604 struct sctp_transport *transport;
4605 sctp_assoc_t associd;
4608 if (len < sizeof(status)) {
4613 len = sizeof(status);
4614 if (copy_from_user(&status, optval, len)) {
4619 associd = status.sstat_assoc_id;
4620 asoc = sctp_id2assoc(sk, associd);
4626 transport = asoc->peer.primary_path;
4628 status.sstat_assoc_id = sctp_assoc2id(asoc);
4629 status.sstat_state = sctp_assoc_to_state(asoc);
4630 status.sstat_rwnd = asoc->peer.rwnd;
4631 status.sstat_unackdata = asoc->unack_data;
4633 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4634 status.sstat_instrms = asoc->c.sinit_max_instreams;
4635 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4636 status.sstat_fragmentation_point = asoc->frag_point;
4637 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4638 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4639 transport->af_specific->sockaddr_len);
4640 /* Map ipv4 address into v4-mapped-on-v6 address. */
4641 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
4642 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4643 status.sstat_primary.spinfo_state = transport->state;
4644 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4645 status.sstat_primary.spinfo_srtt = transport->srtt;
4646 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4647 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4649 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4650 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4652 if (put_user(len, optlen)) {
4657 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4658 __func__, len, status.sstat_state, status.sstat_rwnd,
4659 status.sstat_assoc_id);
4661 if (copy_to_user(optval, &status, len)) {
4671 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4673 * Applications can retrieve information about a specific peer address
4674 * of an association, including its reachability state, congestion
4675 * window, and retransmission timer values. This information is
4678 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4679 char __user *optval,
4682 struct sctp_paddrinfo pinfo;
4683 struct sctp_transport *transport;
4686 if (len < sizeof(pinfo)) {
4691 len = sizeof(pinfo);
4692 if (copy_from_user(&pinfo, optval, len)) {
4697 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4698 pinfo.spinfo_assoc_id);
4702 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4703 pinfo.spinfo_state = transport->state;
4704 pinfo.spinfo_cwnd = transport->cwnd;
4705 pinfo.spinfo_srtt = transport->srtt;
4706 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4707 pinfo.spinfo_mtu = transport->pathmtu;
4709 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4710 pinfo.spinfo_state = SCTP_ACTIVE;
4712 if (put_user(len, optlen)) {
4717 if (copy_to_user(optval, &pinfo, len)) {
4726 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4728 * This option is a on/off flag. If enabled no SCTP message
4729 * fragmentation will be performed. Instead if a message being sent
4730 * exceeds the current PMTU size, the message will NOT be sent and
4731 * instead a error will be indicated to the user.
4733 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4734 char __user *optval, int __user *optlen)
4738 if (len < sizeof(int))
4742 val = (sctp_sk(sk)->disable_fragments == 1);
4743 if (put_user(len, optlen))
4745 if (copy_to_user(optval, &val, len))
4750 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4752 * This socket option is used to specify various notifications and
4753 * ancillary data the user wishes to receive.
4755 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4760 if (len > sizeof(struct sctp_event_subscribe))
4761 len = sizeof(struct sctp_event_subscribe);
4762 if (put_user(len, optlen))
4764 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4769 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4771 * This socket option is applicable to the UDP-style socket only. When
4772 * set it will cause associations that are idle for more than the
4773 * specified number of seconds to automatically close. An association
4774 * being idle is defined an association that has NOT sent or received
4775 * user data. The special value of '0' indicates that no automatic
4776 * close of any associations should be performed. The option expects an
4777 * integer defining the number of seconds of idle time before an
4778 * association is closed.
4780 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4782 /* Applicable to UDP-style socket only */
4783 if (sctp_style(sk, TCP))
4785 if (len < sizeof(int))
4788 if (put_user(len, optlen))
4790 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4795 /* Helper routine to branch off an association to a new socket. */
4796 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4798 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4799 struct sctp_sock *sp = sctp_sk(sk);
4800 struct socket *sock;
4806 /* An association cannot be branched off from an already peeled-off
4807 * socket, nor is this supported for tcp style sockets.
4809 if (!sctp_style(sk, UDP))
4812 /* Create a new socket. */
4813 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4817 sctp_copy_sock(sock->sk, sk, asoc);
4819 /* Make peeled-off sockets more like 1-1 accepted sockets.
4820 * Set the daddr and initialize id to something more random
4822 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
4824 /* Populate the fields of the newsk from the oldsk and migrate the
4825 * asoc to the newsk.
4827 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4833 EXPORT_SYMBOL(sctp_do_peeloff);
4835 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4837 sctp_peeloff_arg_t peeloff;
4838 struct socket *newsock;
4839 struct file *newfile;
4842 if (len < sizeof(sctp_peeloff_arg_t))
4844 len = sizeof(sctp_peeloff_arg_t);
4845 if (copy_from_user(&peeloff, optval, len))
4848 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4852 /* Map the socket to an unused fd that can be returned to the user. */
4853 retval = get_unused_fd_flags(0);
4855 sock_release(newsock);
4859 newfile = sock_alloc_file(newsock, 0, NULL);
4860 if (IS_ERR(newfile)) {
4861 put_unused_fd(retval);
4862 sock_release(newsock);
4863 return PTR_ERR(newfile);
4866 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4869 /* Return the fd mapped to the new socket. */
4870 if (put_user(len, optlen)) {
4872 put_unused_fd(retval);
4875 peeloff.sd = retval;
4876 if (copy_to_user(optval, &peeloff, len)) {
4878 put_unused_fd(retval);
4881 fd_install(retval, newfile);
4886 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4888 * Applications can enable or disable heartbeats for any peer address of
4889 * an association, modify an address's heartbeat interval, force a
4890 * heartbeat to be sent immediately, and adjust the address's maximum
4891 * number of retransmissions sent before an address is considered
4892 * unreachable. The following structure is used to access and modify an
4893 * address's parameters:
4895 * struct sctp_paddrparams {
4896 * sctp_assoc_t spp_assoc_id;
4897 * struct sockaddr_storage spp_address;
4898 * uint32_t spp_hbinterval;
4899 * uint16_t spp_pathmaxrxt;
4900 * uint32_t spp_pathmtu;
4901 * uint32_t spp_sackdelay;
4902 * uint32_t spp_flags;
4905 * spp_assoc_id - (one-to-many style socket) This is filled in the
4906 * application, and identifies the association for
4908 * spp_address - This specifies which address is of interest.
4909 * spp_hbinterval - This contains the value of the heartbeat interval,
4910 * in milliseconds. If a value of zero
4911 * is present in this field then no changes are to
4912 * be made to this parameter.
4913 * spp_pathmaxrxt - This contains the maximum number of
4914 * retransmissions before this address shall be
4915 * considered unreachable. If a value of zero
4916 * is present in this field then no changes are to
4917 * be made to this parameter.
4918 * spp_pathmtu - When Path MTU discovery is disabled the value
4919 * specified here will be the "fixed" path mtu.
4920 * Note that if the spp_address field is empty
4921 * then all associations on this address will
4922 * have this fixed path mtu set upon them.
4924 * spp_sackdelay - When delayed sack is enabled, this value specifies
4925 * the number of milliseconds that sacks will be delayed
4926 * for. This value will apply to all addresses of an
4927 * association if the spp_address field is empty. Note
4928 * also, that if delayed sack is enabled and this
4929 * value is set to 0, no change is made to the last
4930 * recorded delayed sack timer value.
4932 * spp_flags - These flags are used to control various features
4933 * on an association. The flag field may contain
4934 * zero or more of the following options.
4936 * SPP_HB_ENABLE - Enable heartbeats on the
4937 * specified address. Note that if the address
4938 * field is empty all addresses for the association
4939 * have heartbeats enabled upon them.
4941 * SPP_HB_DISABLE - Disable heartbeats on the
4942 * speicifed address. Note that if the address
4943 * field is empty all addresses for the association
4944 * will have their heartbeats disabled. Note also
4945 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4946 * mutually exclusive, only one of these two should
4947 * be specified. Enabling both fields will have
4948 * undetermined results.
4950 * SPP_HB_DEMAND - Request a user initiated heartbeat
4951 * to be made immediately.
4953 * SPP_PMTUD_ENABLE - This field will enable PMTU
4954 * discovery upon the specified address. Note that
4955 * if the address feild is empty then all addresses
4956 * on the association are effected.
4958 * SPP_PMTUD_DISABLE - This field will disable PMTU
4959 * discovery upon the specified address. Note that
4960 * if the address feild is empty then all addresses
4961 * on the association are effected. Not also that
4962 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4963 * exclusive. Enabling both will have undetermined
4966 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4967 * on delayed sack. The time specified in spp_sackdelay
4968 * is used to specify the sack delay for this address. Note
4969 * that if spp_address is empty then all addresses will
4970 * enable delayed sack and take on the sack delay
4971 * value specified in spp_sackdelay.
4972 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4973 * off delayed sack. If the spp_address field is blank then
4974 * delayed sack is disabled for the entire association. Note
4975 * also that this field is mutually exclusive to
4976 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4979 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4980 char __user *optval, int __user *optlen)
4982 struct sctp_paddrparams params;
4983 struct sctp_transport *trans = NULL;
4984 struct sctp_association *asoc = NULL;
4985 struct sctp_sock *sp = sctp_sk(sk);
4987 if (len < sizeof(struct sctp_paddrparams))
4989 len = sizeof(struct sctp_paddrparams);
4990 if (copy_from_user(¶ms, optval, len))
4993 /* If an address other than INADDR_ANY is specified, and
4994 * no transport is found, then the request is invalid.
4996 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
4997 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4998 params.spp_assoc_id);
5000 pr_debug("%s: failed no transport\n", __func__);
5005 /* Get association, if assoc_id != 0 and the socket is a one
5006 * to many style socket, and an association was not found, then
5007 * the id was invalid.
5009 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
5010 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
5011 pr_debug("%s: failed no association\n", __func__);
5016 /* Fetch transport values. */
5017 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
5018 params.spp_pathmtu = trans->pathmtu;
5019 params.spp_pathmaxrxt = trans->pathmaxrxt;
5020 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
5022 /*draft-11 doesn't say what to return in spp_flags*/
5023 params.spp_flags = trans->param_flags;
5025 /* Fetch association values. */
5026 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
5027 params.spp_pathmtu = asoc->pathmtu;
5028 params.spp_pathmaxrxt = asoc->pathmaxrxt;
5029 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
5031 /*draft-11 doesn't say what to return in spp_flags*/
5032 params.spp_flags = asoc->param_flags;
5034 /* Fetch socket values. */
5035 params.spp_hbinterval = sp->hbinterval;
5036 params.spp_pathmtu = sp->pathmtu;
5037 params.spp_sackdelay = sp->sackdelay;
5038 params.spp_pathmaxrxt = sp->pathmaxrxt;
5040 /*draft-11 doesn't say what to return in spp_flags*/
5041 params.spp_flags = sp->param_flags;
5044 if (copy_to_user(optval, ¶ms, len))
5047 if (put_user(len, optlen))
5054 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
5056 * This option will effect the way delayed acks are performed. This
5057 * option allows you to get or set the delayed ack time, in
5058 * milliseconds. It also allows changing the delayed ack frequency.
5059 * Changing the frequency to 1 disables the delayed sack algorithm. If
5060 * the assoc_id is 0, then this sets or gets the endpoints default
5061 * values. If the assoc_id field is non-zero, then the set or get
5062 * effects the specified association for the one to many model (the
5063 * assoc_id field is ignored by the one to one model). Note that if
5064 * sack_delay or sack_freq are 0 when setting this option, then the
5065 * current values will remain unchanged.
5067 * struct sctp_sack_info {
5068 * sctp_assoc_t sack_assoc_id;
5069 * uint32_t sack_delay;
5070 * uint32_t sack_freq;
5073 * sack_assoc_id - This parameter, indicates which association the user
5074 * is performing an action upon. Note that if this field's value is
5075 * zero then the endpoints default value is changed (effecting future
5076 * associations only).
5078 * sack_delay - This parameter contains the number of milliseconds that
5079 * the user is requesting the delayed ACK timer be set to. Note that
5080 * this value is defined in the standard to be between 200 and 500
5083 * sack_freq - This parameter contains the number of packets that must
5084 * be received before a sack is sent without waiting for the delay
5085 * timer to expire. The default value for this is 2, setting this
5086 * value to 1 will disable the delayed sack algorithm.
5088 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
5089 char __user *optval,
5092 struct sctp_sack_info params;
5093 struct sctp_association *asoc = NULL;
5094 struct sctp_sock *sp = sctp_sk(sk);
5096 if (len >= sizeof(struct sctp_sack_info)) {
5097 len = sizeof(struct sctp_sack_info);
5099 if (copy_from_user(¶ms, optval, len))
5101 } else if (len == sizeof(struct sctp_assoc_value)) {
5102 pr_warn_ratelimited(DEPRECATED
5104 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
5105 "Use struct sctp_sack_info instead\n",
5106 current->comm, task_pid_nr(current));
5107 if (copy_from_user(¶ms, optval, len))
5112 /* Get association, if sack_assoc_id != 0 and the socket is a one
5113 * to many style socket, and an association was not found, then
5114 * the id was invalid.
5116 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
5117 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
5121 /* Fetch association values. */
5122 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
5123 params.sack_delay = jiffies_to_msecs(
5125 params.sack_freq = asoc->sackfreq;
5128 params.sack_delay = 0;
5129 params.sack_freq = 1;
5132 /* Fetch socket values. */
5133 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
5134 params.sack_delay = sp->sackdelay;
5135 params.sack_freq = sp->sackfreq;
5137 params.sack_delay = 0;
5138 params.sack_freq = 1;
5142 if (copy_to_user(optval, ¶ms, len))
5145 if (put_user(len, optlen))
5151 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5153 * Applications can specify protocol parameters for the default association
5154 * initialization. The option name argument to setsockopt() and getsockopt()
5157 * Setting initialization parameters is effective only on an unconnected
5158 * socket (for UDP-style sockets only future associations are effected
5159 * by the change). With TCP-style sockets, this option is inherited by
5160 * sockets derived from a listener socket.
5162 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
5164 if (len < sizeof(struct sctp_initmsg))
5166 len = sizeof(struct sctp_initmsg);
5167 if (put_user(len, optlen))
5169 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
5175 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
5176 char __user *optval, int __user *optlen)
5178 struct sctp_association *asoc;
5180 struct sctp_getaddrs getaddrs;
5181 struct sctp_transport *from;
5183 union sctp_addr temp;
5184 struct sctp_sock *sp = sctp_sk(sk);
5189 if (len < sizeof(struct sctp_getaddrs))
5192 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5195 /* For UDP-style sockets, id specifies the association to query. */
5196 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5200 to = optval + offsetof(struct sctp_getaddrs, addrs);
5201 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5203 list_for_each_entry(from, &asoc->peer.transport_addr_list,
5205 memcpy(&temp, &from->ipaddr, sizeof(temp));
5206 addrlen = sctp_get_pf_specific(sk->sk_family)
5207 ->addr_to_user(sp, &temp);
5208 if (space_left < addrlen)
5210 if (copy_to_user(to, &temp, addrlen))
5214 space_left -= addrlen;
5217 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
5219 bytes_copied = ((char __user *)to) - optval;
5220 if (put_user(bytes_copied, optlen))
5226 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
5227 size_t space_left, int *bytes_copied)
5229 struct sctp_sockaddr_entry *addr;
5230 union sctp_addr temp;
5233 struct net *net = sock_net(sk);
5236 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
5240 if ((PF_INET == sk->sk_family) &&
5241 (AF_INET6 == addr->a.sa.sa_family))
5243 if ((PF_INET6 == sk->sk_family) &&
5244 inet_v6_ipv6only(sk) &&
5245 (AF_INET == addr->a.sa.sa_family))
5247 memcpy(&temp, &addr->a, sizeof(temp));
5248 if (!temp.v4.sin_port)
5249 temp.v4.sin_port = htons(port);
5251 addrlen = sctp_get_pf_specific(sk->sk_family)
5252 ->addr_to_user(sctp_sk(sk), &temp);
5254 if (space_left < addrlen) {
5258 memcpy(to, &temp, addrlen);
5262 space_left -= addrlen;
5263 *bytes_copied += addrlen;
5271 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
5272 char __user *optval, int __user *optlen)
5274 struct sctp_bind_addr *bp;
5275 struct sctp_association *asoc;
5277 struct sctp_getaddrs getaddrs;
5278 struct sctp_sockaddr_entry *addr;
5280 union sctp_addr temp;
5281 struct sctp_sock *sp = sctp_sk(sk);
5285 int bytes_copied = 0;
5289 if (len < sizeof(struct sctp_getaddrs))
5292 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5296 * For UDP-style sockets, id specifies the association to query.
5297 * If the id field is set to the value '0' then the locally bound
5298 * addresses are returned without regard to any particular
5301 if (0 == getaddrs.assoc_id) {
5302 bp = &sctp_sk(sk)->ep->base.bind_addr;
5304 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5307 bp = &asoc->base.bind_addr;
5310 to = optval + offsetof(struct sctp_getaddrs, addrs);
5311 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5313 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN);
5317 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5318 * addresses from the global local address list.
5320 if (sctp_list_single_entry(&bp->address_list)) {
5321 addr = list_entry(bp->address_list.next,
5322 struct sctp_sockaddr_entry, list);
5323 if (sctp_is_any(sk, &addr->a)) {
5324 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
5325 space_left, &bytes_copied);
5335 /* Protection on the bound address list is not needed since
5336 * in the socket option context we hold a socket lock and
5337 * thus the bound address list can't change.
5339 list_for_each_entry(addr, &bp->address_list, list) {
5340 memcpy(&temp, &addr->a, sizeof(temp));
5341 addrlen = sctp_get_pf_specific(sk->sk_family)
5342 ->addr_to_user(sp, &temp);
5343 if (space_left < addrlen) {
5344 err = -ENOMEM; /*fixme: right error?*/
5347 memcpy(buf, &temp, addrlen);
5349 bytes_copied += addrlen;
5351 space_left -= addrlen;
5355 if (copy_to_user(to, addrs, bytes_copied)) {
5359 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
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, sizeof(params)))
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))
5968 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5971 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5972 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5976 val.scact_keynumber = asoc->active_key_id;
5978 val.scact_keynumber = ep->active_key_id;
5980 len = sizeof(struct sctp_authkeyid);
5981 if (put_user(len, optlen))
5983 if (copy_to_user(optval, &val, len))
5989 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5990 char __user *optval, int __user *optlen)
5992 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5993 struct sctp_authchunks __user *p = (void __user *)optval;
5994 struct sctp_authchunks val;
5995 struct sctp_association *asoc;
5996 struct sctp_chunks_param *ch;
6000 if (!ep->auth_enable)
6003 if (len < sizeof(struct sctp_authchunks))
6006 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
6009 to = p->gauth_chunks;
6010 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6014 ch = asoc->peer.peer_chunks;
6018 /* See if the user provided enough room for all the data */
6019 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
6020 if (len < num_chunks)
6023 if (copy_to_user(to, ch->chunks, num_chunks))
6026 len = sizeof(struct sctp_authchunks) + num_chunks;
6027 if (put_user(len, optlen))
6029 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6034 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
6035 char __user *optval, int __user *optlen)
6037 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6038 struct sctp_authchunks __user *p = (void __user *)optval;
6039 struct sctp_authchunks val;
6040 struct sctp_association *asoc;
6041 struct sctp_chunks_param *ch;
6045 if (!ep->auth_enable)
6048 if (len < sizeof(struct sctp_authchunks))
6051 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
6054 to = p->gauth_chunks;
6055 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6056 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
6060 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
6062 ch = ep->auth_chunk_list;
6067 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
6068 if (len < sizeof(struct sctp_authchunks) + num_chunks)
6071 if (copy_to_user(to, ch->chunks, num_chunks))
6074 len = sizeof(struct sctp_authchunks) + num_chunks;
6075 if (put_user(len, optlen))
6077 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6084 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
6085 * This option gets the current number of associations that are attached
6086 * to a one-to-many style socket. The option value is an uint32_t.
6088 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
6089 char __user *optval, int __user *optlen)
6091 struct sctp_sock *sp = sctp_sk(sk);
6092 struct sctp_association *asoc;
6095 if (sctp_style(sk, TCP))
6098 if (len < sizeof(u32))
6103 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6107 if (put_user(len, optlen))
6109 if (copy_to_user(optval, &val, len))
6116 * 8.1.23 SCTP_AUTO_ASCONF
6117 * See the corresponding setsockopt entry as description
6119 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
6120 char __user *optval, int __user *optlen)
6124 if (len < sizeof(int))
6128 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
6130 if (put_user(len, optlen))
6132 if (copy_to_user(optval, &val, len))
6138 * 8.2.6. Get the Current Identifiers of Associations
6139 * (SCTP_GET_ASSOC_ID_LIST)
6141 * This option gets the current list of SCTP association identifiers of
6142 * the SCTP associations handled by a one-to-many style socket.
6144 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
6145 char __user *optval, int __user *optlen)
6147 struct sctp_sock *sp = sctp_sk(sk);
6148 struct sctp_association *asoc;
6149 struct sctp_assoc_ids *ids;
6152 if (sctp_style(sk, TCP))
6155 if (len < sizeof(struct sctp_assoc_ids))
6158 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6162 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
6165 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
6167 ids = kmalloc(len, GFP_USER | __GFP_NOWARN);
6171 ids->gaids_number_of_ids = num;
6173 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6174 ids->gaids_assoc_id[num++] = asoc->assoc_id;
6177 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
6187 * SCTP_PEER_ADDR_THLDS
6189 * This option allows us to fetch the partially failed threshold for one or all
6190 * transports in an association. See Section 6.1 of:
6191 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6193 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
6194 char __user *optval,
6198 struct sctp_paddrthlds val;
6199 struct sctp_transport *trans;
6200 struct sctp_association *asoc;
6202 if (len < sizeof(struct sctp_paddrthlds))
6204 len = sizeof(struct sctp_paddrthlds);
6205 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
6208 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
6209 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
6213 val.spt_pathpfthld = asoc->pf_retrans;
6214 val.spt_pathmaxrxt = asoc->pathmaxrxt;
6216 trans = sctp_addr_id2transport(sk, &val.spt_address,
6221 val.spt_pathmaxrxt = trans->pathmaxrxt;
6222 val.spt_pathpfthld = trans->pf_retrans;
6225 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
6232 * SCTP_GET_ASSOC_STATS
6234 * This option retrieves local per endpoint statistics. It is modeled
6235 * after OpenSolaris' implementation
6237 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
6238 char __user *optval,
6241 struct sctp_assoc_stats sas;
6242 struct sctp_association *asoc = NULL;
6244 /* User must provide at least the assoc id */
6245 if (len < sizeof(sctp_assoc_t))
6248 /* Allow the struct to grow and fill in as much as possible */
6249 len = min_t(size_t, len, sizeof(sas));
6251 if (copy_from_user(&sas, optval, len))
6254 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
6258 sas.sas_rtxchunks = asoc->stats.rtxchunks;
6259 sas.sas_gapcnt = asoc->stats.gapcnt;
6260 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
6261 sas.sas_osacks = asoc->stats.osacks;
6262 sas.sas_isacks = asoc->stats.isacks;
6263 sas.sas_octrlchunks = asoc->stats.octrlchunks;
6264 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
6265 sas.sas_oodchunks = asoc->stats.oodchunks;
6266 sas.sas_iodchunks = asoc->stats.iodchunks;
6267 sas.sas_ouodchunks = asoc->stats.ouodchunks;
6268 sas.sas_iuodchunks = asoc->stats.iuodchunks;
6269 sas.sas_idupchunks = asoc->stats.idupchunks;
6270 sas.sas_opackets = asoc->stats.opackets;
6271 sas.sas_ipackets = asoc->stats.ipackets;
6273 /* New high max rto observed, will return 0 if not a single
6274 * RTO update took place. obs_rto_ipaddr will be bogus
6277 sas.sas_maxrto = asoc->stats.max_obs_rto;
6278 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
6279 sizeof(struct sockaddr_storage));
6281 /* Mark beginning of a new observation period */
6282 asoc->stats.max_obs_rto = asoc->rto_min;
6284 if (put_user(len, optlen))
6287 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
6289 if (copy_to_user(optval, &sas, len))
6295 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
6296 char __user *optval,
6301 if (len < sizeof(int))
6305 if (sctp_sk(sk)->recvrcvinfo)
6307 if (put_user(len, optlen))
6309 if (copy_to_user(optval, &val, len))
6315 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
6316 char __user *optval,
6321 if (len < sizeof(int))
6325 if (sctp_sk(sk)->recvnxtinfo)
6327 if (put_user(len, optlen))
6329 if (copy_to_user(optval, &val, len))
6335 static int sctp_getsockopt_pr_supported(struct sock *sk, int len,
6336 char __user *optval,
6339 struct sctp_assoc_value params;
6340 struct sctp_association *asoc;
6341 int retval = -EFAULT;
6343 if (len < sizeof(params)) {
6348 len = sizeof(params);
6349 if (copy_from_user(¶ms, optval, len))
6352 asoc = sctp_id2assoc(sk, params.assoc_id);
6354 params.assoc_value = asoc->prsctp_enable;
6355 } else if (!params.assoc_id) {
6356 struct sctp_sock *sp = sctp_sk(sk);
6358 params.assoc_value = sp->ep->prsctp_enable;
6364 if (put_user(len, optlen))
6367 if (copy_to_user(optval, ¶ms, len))
6376 static int sctp_getsockopt_default_prinfo(struct sock *sk, int len,
6377 char __user *optval,
6380 struct sctp_default_prinfo info;
6381 struct sctp_association *asoc;
6382 int retval = -EFAULT;
6384 if (len < sizeof(info)) {
6390 if (copy_from_user(&info, optval, len))
6393 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
6395 info.pr_policy = SCTP_PR_POLICY(asoc->default_flags);
6396 info.pr_value = asoc->default_timetolive;
6397 } else if (!info.pr_assoc_id) {
6398 struct sctp_sock *sp = sctp_sk(sk);
6400 info.pr_policy = SCTP_PR_POLICY(sp->default_flags);
6401 info.pr_value = sp->default_timetolive;
6407 if (put_user(len, optlen))
6410 if (copy_to_user(optval, &info, len))
6419 static int sctp_getsockopt_pr_assocstatus(struct sock *sk, int len,
6420 char __user *optval,
6423 struct sctp_prstatus params;
6424 struct sctp_association *asoc;
6426 int retval = -EINVAL;
6428 if (len < sizeof(params))
6431 len = sizeof(params);
6432 if (copy_from_user(¶ms, optval, len)) {
6437 policy = params.sprstat_policy;
6438 if (policy & ~SCTP_PR_SCTP_MASK)
6441 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
6445 if (policy == SCTP_PR_SCTP_NONE) {
6446 params.sprstat_abandoned_unsent = 0;
6447 params.sprstat_abandoned_sent = 0;
6448 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
6449 params.sprstat_abandoned_unsent +=
6450 asoc->abandoned_unsent[policy];
6451 params.sprstat_abandoned_sent +=
6452 asoc->abandoned_sent[policy];
6455 params.sprstat_abandoned_unsent =
6456 asoc->abandoned_unsent[__SCTP_PR_INDEX(policy)];
6457 params.sprstat_abandoned_sent =
6458 asoc->abandoned_sent[__SCTP_PR_INDEX(policy)];
6461 if (put_user(len, optlen)) {
6466 if (copy_to_user(optval, ¶ms, len)) {
6477 static int sctp_getsockopt_enable_strreset(struct sock *sk, int len,
6478 char __user *optval,
6481 struct sctp_assoc_value params;
6482 struct sctp_association *asoc;
6483 int retval = -EFAULT;
6485 if (len < sizeof(params)) {
6490 len = sizeof(params);
6491 if (copy_from_user(¶ms, optval, len))
6494 asoc = sctp_id2assoc(sk, params.assoc_id);
6496 params.assoc_value = asoc->strreset_enable;
6497 } else if (!params.assoc_id) {
6498 struct sctp_sock *sp = sctp_sk(sk);
6500 params.assoc_value = sp->ep->strreset_enable;
6506 if (put_user(len, optlen))
6509 if (copy_to_user(optval, ¶ms, len))
6518 static int sctp_getsockopt(struct sock *sk, int level, int optname,
6519 char __user *optval, int __user *optlen)
6524 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
6526 /* I can hardly begin to describe how wrong this is. This is
6527 * so broken as to be worse than useless. The API draft
6528 * REALLY is NOT helpful here... I am not convinced that the
6529 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
6530 * are at all well-founded.
6532 if (level != SOL_SCTP) {
6533 struct sctp_af *af = sctp_sk(sk)->pf->af;
6535 retval = af->getsockopt(sk, level, optname, optval, optlen);
6539 if (get_user(len, optlen))
6549 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
6551 case SCTP_DISABLE_FRAGMENTS:
6552 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
6556 retval = sctp_getsockopt_events(sk, len, optval, optlen);
6558 case SCTP_AUTOCLOSE:
6559 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
6561 case SCTP_SOCKOPT_PEELOFF:
6562 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
6564 case SCTP_PEER_ADDR_PARAMS:
6565 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
6568 case SCTP_DELAYED_SACK:
6569 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
6573 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
6575 case SCTP_GET_PEER_ADDRS:
6576 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
6579 case SCTP_GET_LOCAL_ADDRS:
6580 retval = sctp_getsockopt_local_addrs(sk, len, optval,
6583 case SCTP_SOCKOPT_CONNECTX3:
6584 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
6586 case SCTP_DEFAULT_SEND_PARAM:
6587 retval = sctp_getsockopt_default_send_param(sk, len,
6590 case SCTP_DEFAULT_SNDINFO:
6591 retval = sctp_getsockopt_default_sndinfo(sk, len,
6594 case SCTP_PRIMARY_ADDR:
6595 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
6598 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
6601 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
6603 case SCTP_ASSOCINFO:
6604 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
6606 case SCTP_I_WANT_MAPPED_V4_ADDR:
6607 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
6610 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
6612 case SCTP_GET_PEER_ADDR_INFO:
6613 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
6616 case SCTP_ADAPTATION_LAYER:
6617 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
6621 retval = sctp_getsockopt_context(sk, len, optval, optlen);
6623 case SCTP_FRAGMENT_INTERLEAVE:
6624 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
6627 case SCTP_PARTIAL_DELIVERY_POINT:
6628 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
6631 case SCTP_MAX_BURST:
6632 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
6635 case SCTP_AUTH_CHUNK:
6636 case SCTP_AUTH_DELETE_KEY:
6637 retval = -EOPNOTSUPP;
6639 case SCTP_HMAC_IDENT:
6640 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
6642 case SCTP_AUTH_ACTIVE_KEY:
6643 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
6645 case SCTP_PEER_AUTH_CHUNKS:
6646 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
6649 case SCTP_LOCAL_AUTH_CHUNKS:
6650 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
6653 case SCTP_GET_ASSOC_NUMBER:
6654 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
6656 case SCTP_GET_ASSOC_ID_LIST:
6657 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
6659 case SCTP_AUTO_ASCONF:
6660 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
6662 case SCTP_PEER_ADDR_THLDS:
6663 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
6665 case SCTP_GET_ASSOC_STATS:
6666 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
6668 case SCTP_RECVRCVINFO:
6669 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
6671 case SCTP_RECVNXTINFO:
6672 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
6674 case SCTP_PR_SUPPORTED:
6675 retval = sctp_getsockopt_pr_supported(sk, len, optval, optlen);
6677 case SCTP_DEFAULT_PRINFO:
6678 retval = sctp_getsockopt_default_prinfo(sk, len, optval,
6681 case SCTP_PR_ASSOC_STATUS:
6682 retval = sctp_getsockopt_pr_assocstatus(sk, len, optval,
6685 case SCTP_ENABLE_STREAM_RESET:
6686 retval = sctp_getsockopt_enable_strreset(sk, len, optval,
6690 retval = -ENOPROTOOPT;
6698 static int sctp_hash(struct sock *sk)
6704 static void sctp_unhash(struct sock *sk)
6709 /* Check if port is acceptable. Possibly find first available port.
6711 * The port hash table (contained in the 'global' SCTP protocol storage
6712 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
6713 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
6714 * list (the list number is the port number hashed out, so as you
6715 * would expect from a hash function, all the ports in a given list have
6716 * such a number that hashes out to the same list number; you were
6717 * expecting that, right?); so each list has a set of ports, with a
6718 * link to the socket (struct sock) that uses it, the port number and
6719 * a fastreuse flag (FIXME: NPI ipg).
6721 static struct sctp_bind_bucket *sctp_bucket_create(
6722 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
6724 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
6726 struct sctp_bind_hashbucket *head; /* hash list */
6727 struct sctp_bind_bucket *pp;
6728 unsigned short snum;
6731 snum = ntohs(addr->v4.sin_port);
6733 pr_debug("%s: begins, snum:%d\n", __func__, snum);
6738 /* Search for an available port. */
6739 int low, high, remaining, index;
6741 struct net *net = sock_net(sk);
6743 inet_get_local_port_range(net, &low, &high);
6744 remaining = (high - low) + 1;
6745 rover = prandom_u32() % remaining + low;
6749 if ((rover < low) || (rover > high))
6751 if (inet_is_local_reserved_port(net, rover))
6753 index = sctp_phashfn(sock_net(sk), rover);
6754 head = &sctp_port_hashtable[index];
6755 spin_lock(&head->lock);
6756 sctp_for_each_hentry(pp, &head->chain)
6757 if ((pp->port == rover) &&
6758 net_eq(sock_net(sk), pp->net))
6762 spin_unlock(&head->lock);
6763 } while (--remaining > 0);
6765 /* Exhausted local port range during search? */
6770 /* OK, here is the one we will use. HEAD (the port
6771 * hash table list entry) is non-NULL and we hold it's
6776 /* We are given an specific port number; we verify
6777 * that it is not being used. If it is used, we will
6778 * exahust the search in the hash list corresponding
6779 * to the port number (snum) - we detect that with the
6780 * port iterator, pp being NULL.
6782 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
6783 spin_lock(&head->lock);
6784 sctp_for_each_hentry(pp, &head->chain) {
6785 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
6792 if (!hlist_empty(&pp->owner)) {
6793 /* We had a port hash table hit - there is an
6794 * available port (pp != NULL) and it is being
6795 * used by other socket (pp->owner not empty); that other
6796 * socket is going to be sk2.
6798 int reuse = sk->sk_reuse;
6801 pr_debug("%s: found a possible match\n", __func__);
6803 if (pp->fastreuse && sk->sk_reuse &&
6804 sk->sk_state != SCTP_SS_LISTENING)
6807 /* Run through the list of sockets bound to the port
6808 * (pp->port) [via the pointers bind_next and
6809 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6810 * we get the endpoint they describe and run through
6811 * the endpoint's list of IP (v4 or v6) addresses,
6812 * comparing each of the addresses with the address of
6813 * the socket sk. If we find a match, then that means
6814 * that this port/socket (sk) combination are already
6817 sk_for_each_bound(sk2, &pp->owner) {
6818 struct sctp_endpoint *ep2;
6819 ep2 = sctp_sk(sk2)->ep;
6822 (reuse && sk2->sk_reuse &&
6823 sk2->sk_state != SCTP_SS_LISTENING))
6826 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
6827 sctp_sk(sk2), sctp_sk(sk))) {
6833 pr_debug("%s: found a match\n", __func__);
6836 /* If there was a hash table miss, create a new port. */
6838 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6841 /* In either case (hit or miss), make sure fastreuse is 1 only
6842 * if sk->sk_reuse is too (that is, if the caller requested
6843 * SO_REUSEADDR on this socket -sk-).
6845 if (hlist_empty(&pp->owner)) {
6846 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6850 } else if (pp->fastreuse &&
6851 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6854 /* We are set, so fill up all the data in the hash table
6855 * entry, tie the socket list information with the rest of the
6856 * sockets FIXME: Blurry, NPI (ipg).
6859 if (!sctp_sk(sk)->bind_hash) {
6860 inet_sk(sk)->inet_num = snum;
6861 sk_add_bind_node(sk, &pp->owner);
6862 sctp_sk(sk)->bind_hash = pp;
6867 spin_unlock(&head->lock);
6874 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6875 * port is requested.
6877 static int sctp_get_port(struct sock *sk, unsigned short snum)
6879 union sctp_addr addr;
6880 struct sctp_af *af = sctp_sk(sk)->pf->af;
6882 /* Set up a dummy address struct from the sk. */
6883 af->from_sk(&addr, sk);
6884 addr.v4.sin_port = htons(snum);
6886 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6887 return !!sctp_get_port_local(sk, &addr);
6891 * Move a socket to LISTENING state.
6893 static int sctp_listen_start(struct sock *sk, int backlog)
6895 struct sctp_sock *sp = sctp_sk(sk);
6896 struct sctp_endpoint *ep = sp->ep;
6897 struct crypto_shash *tfm = NULL;
6900 /* Allocate HMAC for generating cookie. */
6901 if (!sp->hmac && sp->sctp_hmac_alg) {
6902 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6903 tfm = crypto_alloc_shash(alg, 0, 0);
6905 net_info_ratelimited("failed to load transform for %s: %ld\n",
6906 sp->sctp_hmac_alg, PTR_ERR(tfm));
6909 sctp_sk(sk)->hmac = tfm;
6913 * If a bind() or sctp_bindx() is not called prior to a listen()
6914 * call that allows new associations to be accepted, the system
6915 * picks an ephemeral port and will choose an address set equivalent
6916 * to binding with a wildcard address.
6918 * This is not currently spelled out in the SCTP sockets
6919 * extensions draft, but follows the practice as seen in TCP
6923 sk->sk_state = SCTP_SS_LISTENING;
6924 if (!ep->base.bind_addr.port) {
6925 if (sctp_autobind(sk))
6928 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6929 sk->sk_state = SCTP_SS_CLOSED;
6934 sk->sk_max_ack_backlog = backlog;
6935 sctp_hash_endpoint(ep);
6940 * 4.1.3 / 5.1.3 listen()
6942 * By default, new associations are not accepted for UDP style sockets.
6943 * An application uses listen() to mark a socket as being able to
6944 * accept new associations.
6946 * On TCP style sockets, applications use listen() to ready the SCTP
6947 * endpoint for accepting inbound associations.
6949 * On both types of endpoints a backlog of '0' disables listening.
6951 * Move a socket to LISTENING state.
6953 int sctp_inet_listen(struct socket *sock, int backlog)
6955 struct sock *sk = sock->sk;
6956 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6959 if (unlikely(backlog < 0))
6964 /* Peeled-off sockets are not allowed to listen(). */
6965 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6968 if (sock->state != SS_UNCONNECTED)
6971 /* If backlog is zero, disable listening. */
6973 if (sctp_sstate(sk, CLOSED))
6977 sctp_unhash_endpoint(ep);
6978 sk->sk_state = SCTP_SS_CLOSED;
6980 sctp_sk(sk)->bind_hash->fastreuse = 1;
6984 /* If we are already listening, just update the backlog */
6985 if (sctp_sstate(sk, LISTENING))
6986 sk->sk_max_ack_backlog = backlog;
6988 err = sctp_listen_start(sk, backlog);
7000 * This function is done by modeling the current datagram_poll() and the
7001 * tcp_poll(). Note that, based on these implementations, we don't
7002 * lock the socket in this function, even though it seems that,
7003 * ideally, locking or some other mechanisms can be used to ensure
7004 * the integrity of the counters (sndbuf and wmem_alloc) used
7005 * in this place. We assume that we don't need locks either until proven
7008 * Another thing to note is that we include the Async I/O support
7009 * here, again, by modeling the current TCP/UDP code. We don't have
7010 * a good way to test with it yet.
7012 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
7014 struct sock *sk = sock->sk;
7015 struct sctp_sock *sp = sctp_sk(sk);
7018 poll_wait(file, sk_sleep(sk), wait);
7020 sock_rps_record_flow(sk);
7022 /* A TCP-style listening socket becomes readable when the accept queue
7025 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
7026 return (!list_empty(&sp->ep->asocs)) ?
7027 (POLLIN | POLLRDNORM) : 0;
7031 /* Is there any exceptional events? */
7032 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
7034 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
7035 if (sk->sk_shutdown & RCV_SHUTDOWN)
7036 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
7037 if (sk->sk_shutdown == SHUTDOWN_MASK)
7040 /* Is it readable? Reconsider this code with TCP-style support. */
7041 if (!skb_queue_empty(&sk->sk_receive_queue))
7042 mask |= POLLIN | POLLRDNORM;
7044 /* The association is either gone or not ready. */
7045 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
7048 /* Is it writable? */
7049 if (sctp_writeable(sk)) {
7050 mask |= POLLOUT | POLLWRNORM;
7052 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
7054 * Since the socket is not locked, the buffer
7055 * might be made available after the writeable check and
7056 * before the bit is set. This could cause a lost I/O
7057 * signal. tcp_poll() has a race breaker for this race
7058 * condition. Based on their implementation, we put
7059 * in the following code to cover it as well.
7061 if (sctp_writeable(sk))
7062 mask |= POLLOUT | POLLWRNORM;
7067 /********************************************************************
7068 * 2nd Level Abstractions
7069 ********************************************************************/
7071 static struct sctp_bind_bucket *sctp_bucket_create(
7072 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
7074 struct sctp_bind_bucket *pp;
7076 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
7078 SCTP_DBG_OBJCNT_INC(bind_bucket);
7081 INIT_HLIST_HEAD(&pp->owner);
7083 hlist_add_head(&pp->node, &head->chain);
7088 /* Caller must hold hashbucket lock for this tb with local BH disabled */
7089 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
7091 if (pp && hlist_empty(&pp->owner)) {
7092 __hlist_del(&pp->node);
7093 kmem_cache_free(sctp_bucket_cachep, pp);
7094 SCTP_DBG_OBJCNT_DEC(bind_bucket);
7098 /* Release this socket's reference to a local port. */
7099 static inline void __sctp_put_port(struct sock *sk)
7101 struct sctp_bind_hashbucket *head =
7102 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
7103 inet_sk(sk)->inet_num)];
7104 struct sctp_bind_bucket *pp;
7106 spin_lock(&head->lock);
7107 pp = sctp_sk(sk)->bind_hash;
7108 __sk_del_bind_node(sk);
7109 sctp_sk(sk)->bind_hash = NULL;
7110 inet_sk(sk)->inet_num = 0;
7111 sctp_bucket_destroy(pp);
7112 spin_unlock(&head->lock);
7115 void sctp_put_port(struct sock *sk)
7118 __sctp_put_port(sk);
7123 * The system picks an ephemeral port and choose an address set equivalent
7124 * to binding with a wildcard address.
7125 * One of those addresses will be the primary address for the association.
7126 * This automatically enables the multihoming capability of SCTP.
7128 static int sctp_autobind(struct sock *sk)
7130 union sctp_addr autoaddr;
7134 /* Initialize a local sockaddr structure to INADDR_ANY. */
7135 af = sctp_sk(sk)->pf->af;
7137 port = htons(inet_sk(sk)->inet_num);
7138 af->inaddr_any(&autoaddr, port);
7140 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
7143 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
7146 * 4.2 The cmsghdr Structure *
7148 * When ancillary data is sent or received, any number of ancillary data
7149 * objects can be specified by the msg_control and msg_controllen members of
7150 * the msghdr structure, because each object is preceded by
7151 * a cmsghdr structure defining the object's length (the cmsg_len member).
7152 * Historically Berkeley-derived implementations have passed only one object
7153 * at a time, but this API allows multiple objects to be
7154 * passed in a single call to sendmsg() or recvmsg(). The following example
7155 * shows two ancillary data objects in a control buffer.
7157 * |<--------------------------- msg_controllen -------------------------->|
7160 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
7162 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
7165 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
7167 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
7170 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7171 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
7173 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
7175 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7182 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
7184 struct cmsghdr *cmsg;
7185 struct msghdr *my_msg = (struct msghdr *)msg;
7187 for_each_cmsghdr(cmsg, my_msg) {
7188 if (!CMSG_OK(my_msg, cmsg))
7191 /* Should we parse this header or ignore? */
7192 if (cmsg->cmsg_level != IPPROTO_SCTP)
7195 /* Strictly check lengths following example in SCM code. */
7196 switch (cmsg->cmsg_type) {
7198 /* SCTP Socket API Extension
7199 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
7201 * This cmsghdr structure provides information for
7202 * initializing new SCTP associations with sendmsg().
7203 * The SCTP_INITMSG socket option uses this same data
7204 * structure. This structure is not used for
7207 * cmsg_level cmsg_type cmsg_data[]
7208 * ------------ ------------ ----------------------
7209 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
7211 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
7214 cmsgs->init = CMSG_DATA(cmsg);
7218 /* SCTP Socket API Extension
7219 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
7221 * This cmsghdr structure specifies SCTP options for
7222 * sendmsg() and describes SCTP header information
7223 * about a received message through recvmsg().
7225 * cmsg_level cmsg_type cmsg_data[]
7226 * ------------ ------------ ----------------------
7227 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
7229 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
7232 cmsgs->srinfo = CMSG_DATA(cmsg);
7234 if (cmsgs->srinfo->sinfo_flags &
7235 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7236 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7237 SCTP_ABORT | SCTP_EOF))
7242 /* SCTP Socket API Extension
7243 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
7245 * This cmsghdr structure specifies SCTP options for
7246 * sendmsg(). This structure and SCTP_RCVINFO replaces
7247 * SCTP_SNDRCV which has been deprecated.
7249 * cmsg_level cmsg_type cmsg_data[]
7250 * ------------ ------------ ---------------------
7251 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
7253 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
7256 cmsgs->sinfo = CMSG_DATA(cmsg);
7258 if (cmsgs->sinfo->snd_flags &
7259 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7260 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7261 SCTP_ABORT | SCTP_EOF))
7273 * Wait for a packet..
7274 * Note: This function is the same function as in core/datagram.c
7275 * with a few modifications to make lksctp work.
7277 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
7282 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7284 /* Socket errors? */
7285 error = sock_error(sk);
7289 if (!skb_queue_empty(&sk->sk_receive_queue))
7292 /* Socket shut down? */
7293 if (sk->sk_shutdown & RCV_SHUTDOWN)
7296 /* Sequenced packets can come disconnected. If so we report the
7301 /* Is there a good reason to think that we may receive some data? */
7302 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
7305 /* Handle signals. */
7306 if (signal_pending(current))
7309 /* Let another process have a go. Since we are going to sleep
7310 * anyway. Note: This may cause odd behaviors if the message
7311 * does not fit in the user's buffer, but this seems to be the
7312 * only way to honor MSG_DONTWAIT realistically.
7315 *timeo_p = schedule_timeout(*timeo_p);
7319 finish_wait(sk_sleep(sk), &wait);
7323 error = sock_intr_errno(*timeo_p);
7326 finish_wait(sk_sleep(sk), &wait);
7331 /* Receive a datagram.
7332 * Note: This is pretty much the same routine as in core/datagram.c
7333 * with a few changes to make lksctp work.
7335 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
7336 int noblock, int *err)
7339 struct sk_buff *skb;
7342 timeo = sock_rcvtimeo(sk, noblock);
7344 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
7345 MAX_SCHEDULE_TIMEOUT);
7348 /* Again only user level code calls this function,
7349 * so nothing interrupt level
7350 * will suddenly eat the receive_queue.
7352 * Look at current nfs client by the way...
7353 * However, this function was correct in any case. 8)
7355 if (flags & MSG_PEEK) {
7356 skb = skb_peek(&sk->sk_receive_queue);
7358 atomic_inc(&skb->users);
7360 skb = __skb_dequeue(&sk->sk_receive_queue);
7366 /* Caller is allowed not to check sk->sk_err before calling. */
7367 error = sock_error(sk);
7371 if (sk->sk_shutdown & RCV_SHUTDOWN)
7374 if (sk_can_busy_loop(sk) &&
7375 sk_busy_loop(sk, noblock))
7378 /* User doesn't want to wait. */
7382 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
7391 /* If sndbuf has changed, wake up per association sndbuf waiters. */
7392 static void __sctp_write_space(struct sctp_association *asoc)
7394 struct sock *sk = asoc->base.sk;
7396 if (sctp_wspace(asoc) <= 0)
7399 if (waitqueue_active(&asoc->wait))
7400 wake_up_interruptible(&asoc->wait);
7402 if (sctp_writeable(sk)) {
7403 struct socket_wq *wq;
7406 wq = rcu_dereference(sk->sk_wq);
7408 if (waitqueue_active(&wq->wait))
7409 wake_up_interruptible(&wq->wait);
7411 /* Note that we try to include the Async I/O support
7412 * here by modeling from the current TCP/UDP code.
7413 * We have not tested with it yet.
7415 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
7416 sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT);
7422 static void sctp_wake_up_waiters(struct sock *sk,
7423 struct sctp_association *asoc)
7425 struct sctp_association *tmp = asoc;
7427 /* We do accounting for the sndbuf space per association,
7428 * so we only need to wake our own association.
7430 if (asoc->ep->sndbuf_policy)
7431 return __sctp_write_space(asoc);
7433 /* If association goes down and is just flushing its
7434 * outq, then just normally notify others.
7436 if (asoc->base.dead)
7437 return sctp_write_space(sk);
7439 /* Accounting for the sndbuf space is per socket, so we
7440 * need to wake up others, try to be fair and in case of
7441 * other associations, let them have a go first instead
7442 * of just doing a sctp_write_space() call.
7444 * Note that we reach sctp_wake_up_waiters() only when
7445 * associations free up queued chunks, thus we are under
7446 * lock and the list of associations on a socket is
7447 * guaranteed not to change.
7449 for (tmp = list_next_entry(tmp, asocs); 1;
7450 tmp = list_next_entry(tmp, asocs)) {
7451 /* Manually skip the head element. */
7452 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
7454 /* Wake up association. */
7455 __sctp_write_space(tmp);
7456 /* We've reached the end. */
7462 /* Do accounting for the sndbuf space.
7463 * Decrement the used sndbuf space of the corresponding association by the
7464 * data size which was just transmitted(freed).
7466 static void sctp_wfree(struct sk_buff *skb)
7468 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
7469 struct sctp_association *asoc = chunk->asoc;
7470 struct sock *sk = asoc->base.sk;
7472 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
7473 sizeof(struct sk_buff) +
7474 sizeof(struct sctp_chunk);
7476 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
7479 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
7481 sk->sk_wmem_queued -= skb->truesize;
7482 sk_mem_uncharge(sk, skb->truesize);
7485 sctp_wake_up_waiters(sk, asoc);
7487 sctp_association_put(asoc);
7490 /* Do accounting for the receive space on the socket.
7491 * Accounting for the association is done in ulpevent.c
7492 * We set this as a destructor for the cloned data skbs so that
7493 * accounting is done at the correct time.
7495 void sctp_sock_rfree(struct sk_buff *skb)
7497 struct sock *sk = skb->sk;
7498 struct sctp_ulpevent *event = sctp_skb2event(skb);
7500 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
7503 * Mimic the behavior of sock_rfree
7505 sk_mem_uncharge(sk, event->rmem_len);
7509 /* Helper function to wait for space in the sndbuf. */
7510 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
7513 struct sock *sk = asoc->base.sk;
7515 long current_timeo = *timeo_p;
7518 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
7521 /* Increment the association's refcnt. */
7522 sctp_association_hold(asoc);
7524 /* Wait on the association specific sndbuf space. */
7526 prepare_to_wait_exclusive(&asoc->wait, &wait,
7527 TASK_INTERRUPTIBLE);
7530 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7533 if (signal_pending(current))
7534 goto do_interrupted;
7535 if (msg_len <= sctp_wspace(asoc))
7538 /* Let another process have a go. Since we are going
7542 current_timeo = schedule_timeout(current_timeo);
7543 if (sk != asoc->base.sk)
7547 *timeo_p = current_timeo;
7551 finish_wait(&asoc->wait, &wait);
7553 /* Release the association's refcnt. */
7554 sctp_association_put(asoc);
7563 err = sock_intr_errno(*timeo_p);
7571 void sctp_data_ready(struct sock *sk)
7573 struct socket_wq *wq;
7576 wq = rcu_dereference(sk->sk_wq);
7577 if (skwq_has_sleeper(wq))
7578 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
7579 POLLRDNORM | POLLRDBAND);
7580 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
7584 /* If socket sndbuf has changed, wake up all per association waiters. */
7585 void sctp_write_space(struct sock *sk)
7587 struct sctp_association *asoc;
7589 /* Wake up the tasks in each wait queue. */
7590 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
7591 __sctp_write_space(asoc);
7595 /* Is there any sndbuf space available on the socket?
7597 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
7598 * associations on the same socket. For a UDP-style socket with
7599 * multiple associations, it is possible for it to be "unwriteable"
7600 * prematurely. I assume that this is acceptable because
7601 * a premature "unwriteable" is better than an accidental "writeable" which
7602 * would cause an unwanted block under certain circumstances. For the 1-1
7603 * UDP-style sockets or TCP-style sockets, this code should work.
7606 static int sctp_writeable(struct sock *sk)
7610 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
7616 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
7617 * returns immediately with EINPROGRESS.
7619 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
7621 struct sock *sk = asoc->base.sk;
7623 long current_timeo = *timeo_p;
7626 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
7628 /* Increment the association's refcnt. */
7629 sctp_association_hold(asoc);
7632 prepare_to_wait_exclusive(&asoc->wait, &wait,
7633 TASK_INTERRUPTIBLE);
7636 if (sk->sk_shutdown & RCV_SHUTDOWN)
7638 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7641 if (signal_pending(current))
7642 goto do_interrupted;
7644 if (sctp_state(asoc, ESTABLISHED))
7647 /* Let another process have a go. Since we are going
7651 current_timeo = schedule_timeout(current_timeo);
7654 *timeo_p = current_timeo;
7658 finish_wait(&asoc->wait, &wait);
7660 /* Release the association's refcnt. */
7661 sctp_association_put(asoc);
7666 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
7669 err = -ECONNREFUSED;
7673 err = sock_intr_errno(*timeo_p);
7681 static int sctp_wait_for_accept(struct sock *sk, long timeo)
7683 struct sctp_endpoint *ep;
7687 ep = sctp_sk(sk)->ep;
7691 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
7692 TASK_INTERRUPTIBLE);
7694 if (list_empty(&ep->asocs)) {
7696 timeo = schedule_timeout(timeo);
7701 if (!sctp_sstate(sk, LISTENING))
7705 if (!list_empty(&ep->asocs))
7708 err = sock_intr_errno(timeo);
7709 if (signal_pending(current))
7717 finish_wait(sk_sleep(sk), &wait);
7722 static void sctp_wait_for_close(struct sock *sk, long timeout)
7727 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7728 if (list_empty(&sctp_sk(sk)->ep->asocs))
7731 timeout = schedule_timeout(timeout);
7733 } while (!signal_pending(current) && timeout);
7735 finish_wait(sk_sleep(sk), &wait);
7738 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
7740 struct sk_buff *frag;
7745 /* Don't forget the fragments. */
7746 skb_walk_frags(skb, frag)
7747 sctp_skb_set_owner_r_frag(frag, sk);
7750 sctp_skb_set_owner_r(skb, sk);
7753 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
7754 struct sctp_association *asoc)
7756 struct inet_sock *inet = inet_sk(sk);
7757 struct inet_sock *newinet;
7759 newsk->sk_type = sk->sk_type;
7760 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
7761 newsk->sk_flags = sk->sk_flags;
7762 newsk->sk_tsflags = sk->sk_tsflags;
7763 newsk->sk_no_check_tx = sk->sk_no_check_tx;
7764 newsk->sk_no_check_rx = sk->sk_no_check_rx;
7765 newsk->sk_reuse = sk->sk_reuse;
7767 newsk->sk_shutdown = sk->sk_shutdown;
7768 newsk->sk_destruct = sctp_destruct_sock;
7769 newsk->sk_family = sk->sk_family;
7770 newsk->sk_protocol = IPPROTO_SCTP;
7771 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
7772 newsk->sk_sndbuf = sk->sk_sndbuf;
7773 newsk->sk_rcvbuf = sk->sk_rcvbuf;
7774 newsk->sk_lingertime = sk->sk_lingertime;
7775 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
7776 newsk->sk_sndtimeo = sk->sk_sndtimeo;
7777 newsk->sk_rxhash = sk->sk_rxhash;
7779 newinet = inet_sk(newsk);
7781 /* Initialize sk's sport, dport, rcv_saddr and daddr for
7782 * getsockname() and getpeername()
7784 newinet->inet_sport = inet->inet_sport;
7785 newinet->inet_saddr = inet->inet_saddr;
7786 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
7787 newinet->inet_dport = htons(asoc->peer.port);
7788 newinet->pmtudisc = inet->pmtudisc;
7789 newinet->inet_id = asoc->next_tsn ^ jiffies;
7791 newinet->uc_ttl = inet->uc_ttl;
7792 newinet->mc_loop = 1;
7793 newinet->mc_ttl = 1;
7794 newinet->mc_index = 0;
7795 newinet->mc_list = NULL;
7797 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
7798 net_enable_timestamp();
7800 security_sk_clone(sk, newsk);
7803 static inline void sctp_copy_descendant(struct sock *sk_to,
7804 const struct sock *sk_from)
7806 int ancestor_size = sizeof(struct inet_sock) +
7807 sizeof(struct sctp_sock) -
7808 offsetof(struct sctp_sock, auto_asconf_list);
7810 if (sk_from->sk_family == PF_INET6)
7811 ancestor_size += sizeof(struct ipv6_pinfo);
7813 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
7816 /* Populate the fields of the newsk from the oldsk and migrate the assoc
7817 * and its messages to the newsk.
7819 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
7820 struct sctp_association *assoc,
7821 sctp_socket_type_t type)
7823 struct sctp_sock *oldsp = sctp_sk(oldsk);
7824 struct sctp_sock *newsp = sctp_sk(newsk);
7825 struct sctp_bind_bucket *pp; /* hash list port iterator */
7826 struct sctp_endpoint *newep = newsp->ep;
7827 struct sk_buff *skb, *tmp;
7828 struct sctp_ulpevent *event;
7829 struct sctp_bind_hashbucket *head;
7831 /* Migrate socket buffer sizes and all the socket level options to the
7834 newsk->sk_sndbuf = oldsk->sk_sndbuf;
7835 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
7836 /* Brute force copy old sctp opt. */
7837 sctp_copy_descendant(newsk, oldsk);
7839 /* Restore the ep value that was overwritten with the above structure
7845 /* Hook this new socket in to the bind_hash list. */
7846 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
7847 inet_sk(oldsk)->inet_num)];
7848 spin_lock_bh(&head->lock);
7849 pp = sctp_sk(oldsk)->bind_hash;
7850 sk_add_bind_node(newsk, &pp->owner);
7851 sctp_sk(newsk)->bind_hash = pp;
7852 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
7853 spin_unlock_bh(&head->lock);
7855 /* Copy the bind_addr list from the original endpoint to the new
7856 * endpoint so that we can handle restarts properly
7858 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
7859 &oldsp->ep->base.bind_addr, GFP_KERNEL);
7861 /* Move any messages in the old socket's receive queue that are for the
7862 * peeled off association to the new socket's receive queue.
7864 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
7865 event = sctp_skb2event(skb);
7866 if (event->asoc == assoc) {
7867 __skb_unlink(skb, &oldsk->sk_receive_queue);
7868 __skb_queue_tail(&newsk->sk_receive_queue, skb);
7869 sctp_skb_set_owner_r_frag(skb, newsk);
7873 /* Clean up any messages pending delivery due to partial
7874 * delivery. Three cases:
7875 * 1) No partial deliver; no work.
7876 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7877 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7879 skb_queue_head_init(&newsp->pd_lobby);
7880 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
7882 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
7883 struct sk_buff_head *queue;
7885 /* Decide which queue to move pd_lobby skbs to. */
7886 if (assoc->ulpq.pd_mode) {
7887 queue = &newsp->pd_lobby;
7889 queue = &newsk->sk_receive_queue;
7891 /* Walk through the pd_lobby, looking for skbs that
7892 * need moved to the new socket.
7894 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
7895 event = sctp_skb2event(skb);
7896 if (event->asoc == assoc) {
7897 __skb_unlink(skb, &oldsp->pd_lobby);
7898 __skb_queue_tail(queue, skb);
7899 sctp_skb_set_owner_r_frag(skb, newsk);
7903 /* Clear up any skbs waiting for the partial
7904 * delivery to finish.
7906 if (assoc->ulpq.pd_mode)
7907 sctp_clear_pd(oldsk, NULL);
7911 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7912 sctp_skb_set_owner_r_frag(skb, newsk);
7914 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7915 sctp_skb_set_owner_r_frag(skb, newsk);
7917 /* Set the type of socket to indicate that it is peeled off from the
7918 * original UDP-style socket or created with the accept() call on a
7919 * TCP-style socket..
7923 /* Mark the new socket "in-use" by the user so that any packets
7924 * that may arrive on the association after we've moved it are
7925 * queued to the backlog. This prevents a potential race between
7926 * backlog processing on the old socket and new-packet processing
7927 * on the new socket.
7929 * The caller has just allocated newsk so we can guarantee that other
7930 * paths won't try to lock it and then oldsk.
7932 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7933 sctp_assoc_migrate(assoc, newsk);
7935 /* If the association on the newsk is already closed before accept()
7936 * is called, set RCV_SHUTDOWN flag.
7938 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP)) {
7939 newsk->sk_state = SCTP_SS_CLOSED;
7940 newsk->sk_shutdown |= RCV_SHUTDOWN;
7942 newsk->sk_state = SCTP_SS_ESTABLISHED;
7945 release_sock(newsk);
7949 /* This proto struct describes the ULP interface for SCTP. */
7950 struct proto sctp_prot = {
7952 .owner = THIS_MODULE,
7953 .close = sctp_close,
7954 .connect = sctp_connect,
7955 .disconnect = sctp_disconnect,
7956 .accept = sctp_accept,
7957 .ioctl = sctp_ioctl,
7958 .init = sctp_init_sock,
7959 .destroy = sctp_destroy_sock,
7960 .shutdown = sctp_shutdown,
7961 .setsockopt = sctp_setsockopt,
7962 .getsockopt = sctp_getsockopt,
7963 .sendmsg = sctp_sendmsg,
7964 .recvmsg = sctp_recvmsg,
7966 .backlog_rcv = sctp_backlog_rcv,
7968 .unhash = sctp_unhash,
7969 .get_port = sctp_get_port,
7970 .obj_size = sizeof(struct sctp_sock),
7971 .sysctl_mem = sysctl_sctp_mem,
7972 .sysctl_rmem = sysctl_sctp_rmem,
7973 .sysctl_wmem = sysctl_sctp_wmem,
7974 .memory_pressure = &sctp_memory_pressure,
7975 .enter_memory_pressure = sctp_enter_memory_pressure,
7976 .memory_allocated = &sctp_memory_allocated,
7977 .sockets_allocated = &sctp_sockets_allocated,
7980 #if IS_ENABLED(CONFIG_IPV6)
7982 #include <net/transp_v6.h>
7983 static void sctp_v6_destroy_sock(struct sock *sk)
7985 sctp_destroy_sock(sk);
7986 inet6_destroy_sock(sk);
7989 struct proto sctpv6_prot = {
7991 .owner = THIS_MODULE,
7992 .close = sctp_close,
7993 .connect = sctp_connect,
7994 .disconnect = sctp_disconnect,
7995 .accept = sctp_accept,
7996 .ioctl = sctp_ioctl,
7997 .init = sctp_init_sock,
7998 .destroy = sctp_v6_destroy_sock,
7999 .shutdown = sctp_shutdown,
8000 .setsockopt = sctp_setsockopt,
8001 .getsockopt = sctp_getsockopt,
8002 .sendmsg = sctp_sendmsg,
8003 .recvmsg = sctp_recvmsg,
8005 .backlog_rcv = sctp_backlog_rcv,
8007 .unhash = sctp_unhash,
8008 .get_port = sctp_get_port,
8009 .obj_size = sizeof(struct sctp6_sock),
8010 .sysctl_mem = sysctl_sctp_mem,
8011 .sysctl_rmem = sysctl_sctp_rmem,
8012 .sysctl_wmem = sysctl_sctp_wmem,
8013 .memory_pressure = &sctp_memory_pressure,
8014 .enter_memory_pressure = sctp_enter_memory_pressure,
8015 .memory_allocated = &sctp_memory_allocated,
8016 .sockets_allocated = &sctp_sockets_allocated,
8018 #endif /* IS_ENABLED(CONFIG_IPV6) */