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);
1967 datamsg->force_delay = !!(msg->msg_flags & MSG_MORE);
1969 /* Now send the (possibly) fragmented message. */
1970 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1971 sctp_chunk_hold(chunk);
1973 /* Do accounting for the write space. */
1974 sctp_set_owner_w(chunk);
1976 chunk->transport = chunk_tp;
1979 /* Send it to the lower layers. Note: all chunks
1980 * must either fail or succeed. The lower layer
1981 * works that way today. Keep it that way or this
1984 err = sctp_primitive_SEND(net, asoc, datamsg);
1985 /* Did the lower layer accept the chunk? */
1987 sctp_datamsg_free(datamsg);
1991 pr_debug("%s: we sent primitively\n", __func__);
1993 sctp_datamsg_put(datamsg);
1996 if (unlikely(wait_connect)) {
1997 timeo = sock_sndtimeo(sk, msg_flags & MSG_DONTWAIT);
1998 sctp_wait_for_connect(asoc, &timeo);
2001 /* If we are already past ASSOCIATE, the lower
2002 * layers are responsible for association cleanup.
2008 sctp_association_free(asoc);
2013 return sctp_error(sk, msg_flags, err);
2020 err = sock_error(sk);
2030 /* This is an extended version of skb_pull() that removes the data from the
2031 * start of a skb even when data is spread across the list of skb's in the
2032 * frag_list. len specifies the total amount of data that needs to be removed.
2033 * when 'len' bytes could be removed from the skb, it returns 0.
2034 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2035 * could not be removed.
2037 static int sctp_skb_pull(struct sk_buff *skb, int len)
2039 struct sk_buff *list;
2040 int skb_len = skb_headlen(skb);
2043 if (len <= skb_len) {
2044 __skb_pull(skb, len);
2048 __skb_pull(skb, skb_len);
2050 skb_walk_frags(skb, list) {
2051 rlen = sctp_skb_pull(list, len);
2052 skb->len -= (len-rlen);
2053 skb->data_len -= (len-rlen);
2064 /* API 3.1.3 recvmsg() - UDP Style Syntax
2066 * ssize_t recvmsg(int socket, struct msghdr *message,
2069 * socket - the socket descriptor of the endpoint.
2070 * message - pointer to the msghdr structure which contains a single
2071 * user message and possibly some ancillary data.
2073 * See Section 5 for complete description of the data
2076 * flags - flags sent or received with the user message, see Section
2077 * 5 for complete description of the flags.
2079 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2080 int noblock, int flags, int *addr_len)
2082 struct sctp_ulpevent *event = NULL;
2083 struct sctp_sock *sp = sctp_sk(sk);
2084 struct sk_buff *skb, *head_skb;
2089 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2090 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2095 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED) &&
2096 !sctp_sstate(sk, CLOSING) && !sctp_sstate(sk, CLOSED)) {
2101 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2105 /* Get the total length of the skb including any skb's in the
2114 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2116 event = sctp_skb2event(skb);
2121 if (event->chunk && event->chunk->head_skb)
2122 head_skb = event->chunk->head_skb;
2125 sock_recv_ts_and_drops(msg, sk, head_skb);
2126 if (sctp_ulpevent_is_notification(event)) {
2127 msg->msg_flags |= MSG_NOTIFICATION;
2128 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2130 sp->pf->skb_msgname(head_skb, msg->msg_name, addr_len);
2133 /* Check if we allow SCTP_NXTINFO. */
2134 if (sp->recvnxtinfo)
2135 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2136 /* Check if we allow SCTP_RCVINFO. */
2137 if (sp->recvrcvinfo)
2138 sctp_ulpevent_read_rcvinfo(event, msg);
2139 /* Check if we allow SCTP_SNDRCVINFO. */
2140 if (sp->subscribe.sctp_data_io_event)
2141 sctp_ulpevent_read_sndrcvinfo(event, msg);
2145 /* If skb's length exceeds the user's buffer, update the skb and
2146 * push it back to the receive_queue so that the next call to
2147 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2149 if (skb_len > copied) {
2150 msg->msg_flags &= ~MSG_EOR;
2151 if (flags & MSG_PEEK)
2153 sctp_skb_pull(skb, copied);
2154 skb_queue_head(&sk->sk_receive_queue, skb);
2156 /* When only partial message is copied to the user, increase
2157 * rwnd by that amount. If all the data in the skb is read,
2158 * rwnd is updated when the event is freed.
2160 if (!sctp_ulpevent_is_notification(event))
2161 sctp_assoc_rwnd_increase(event->asoc, copied);
2163 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2164 (event->msg_flags & MSG_EOR))
2165 msg->msg_flags |= MSG_EOR;
2167 msg->msg_flags &= ~MSG_EOR;
2170 if (flags & MSG_PEEK) {
2171 /* Release the skb reference acquired after peeking the skb in
2172 * sctp_skb_recv_datagram().
2176 /* Free the event which includes releasing the reference to
2177 * the owner of the skb, freeing the skb and updating the
2180 sctp_ulpevent_free(event);
2187 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2189 * This option is a on/off flag. If enabled no SCTP message
2190 * fragmentation will be performed. Instead if a message being sent
2191 * exceeds the current PMTU size, the message will NOT be sent and
2192 * instead a error will be indicated to the user.
2194 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2195 char __user *optval,
2196 unsigned int optlen)
2200 if (optlen < sizeof(int))
2203 if (get_user(val, (int __user *)optval))
2206 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2211 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2212 unsigned int optlen)
2214 struct sctp_association *asoc;
2215 struct sctp_ulpevent *event;
2217 if (optlen > sizeof(struct sctp_event_subscribe))
2219 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2222 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2223 * if there is no data to be sent or retransmit, the stack will
2224 * immediately send up this notification.
2226 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2227 &sctp_sk(sk)->subscribe)) {
2228 asoc = sctp_id2assoc(sk, 0);
2230 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2231 event = sctp_ulpevent_make_sender_dry_event(asoc,
2236 sctp_ulpq_tail_event(&asoc->ulpq, event);
2243 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2245 * This socket option is applicable to the UDP-style socket only. When
2246 * set it will cause associations that are idle for more than the
2247 * specified number of seconds to automatically close. An association
2248 * being idle is defined an association that has NOT sent or received
2249 * user data. The special value of '0' indicates that no automatic
2250 * close of any associations should be performed. The option expects an
2251 * integer defining the number of seconds of idle time before an
2252 * association is closed.
2254 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2255 unsigned int optlen)
2257 struct sctp_sock *sp = sctp_sk(sk);
2258 struct net *net = sock_net(sk);
2260 /* Applicable to UDP-style socket only */
2261 if (sctp_style(sk, TCP))
2263 if (optlen != sizeof(int))
2265 if (copy_from_user(&sp->autoclose, optval, optlen))
2268 if (sp->autoclose > net->sctp.max_autoclose)
2269 sp->autoclose = net->sctp.max_autoclose;
2274 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2276 * Applications can enable or disable heartbeats for any peer address of
2277 * an association, modify an address's heartbeat interval, force a
2278 * heartbeat to be sent immediately, and adjust the address's maximum
2279 * number of retransmissions sent before an address is considered
2280 * unreachable. The following structure is used to access and modify an
2281 * address's parameters:
2283 * struct sctp_paddrparams {
2284 * sctp_assoc_t spp_assoc_id;
2285 * struct sockaddr_storage spp_address;
2286 * uint32_t spp_hbinterval;
2287 * uint16_t spp_pathmaxrxt;
2288 * uint32_t spp_pathmtu;
2289 * uint32_t spp_sackdelay;
2290 * uint32_t spp_flags;
2293 * spp_assoc_id - (one-to-many style socket) This is filled in the
2294 * application, and identifies the association for
2296 * spp_address - This specifies which address is of interest.
2297 * spp_hbinterval - This contains the value of the heartbeat interval,
2298 * in milliseconds. If a value of zero
2299 * is present in this field then no changes are to
2300 * be made to this parameter.
2301 * spp_pathmaxrxt - This contains the maximum number of
2302 * retransmissions before this address shall be
2303 * considered unreachable. If a value of zero
2304 * is present in this field then no changes are to
2305 * be made to this parameter.
2306 * spp_pathmtu - When Path MTU discovery is disabled the value
2307 * specified here will be the "fixed" path mtu.
2308 * Note that if the spp_address field is empty
2309 * then all associations on this address will
2310 * have this fixed path mtu set upon them.
2312 * spp_sackdelay - When delayed sack is enabled, this value specifies
2313 * the number of milliseconds that sacks will be delayed
2314 * for. This value will apply to all addresses of an
2315 * association if the spp_address field is empty. Note
2316 * also, that if delayed sack is enabled and this
2317 * value is set to 0, no change is made to the last
2318 * recorded delayed sack timer value.
2320 * spp_flags - These flags are used to control various features
2321 * on an association. The flag field may contain
2322 * zero or more of the following options.
2324 * SPP_HB_ENABLE - Enable heartbeats on the
2325 * specified address. Note that if the address
2326 * field is empty all addresses for the association
2327 * have heartbeats enabled upon them.
2329 * SPP_HB_DISABLE - Disable heartbeats on the
2330 * speicifed address. Note that if the address
2331 * field is empty all addresses for the association
2332 * will have their heartbeats disabled. Note also
2333 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2334 * mutually exclusive, only one of these two should
2335 * be specified. Enabling both fields will have
2336 * undetermined results.
2338 * SPP_HB_DEMAND - Request a user initiated heartbeat
2339 * to be made immediately.
2341 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2342 * heartbeat delayis to be set to the value of 0
2345 * SPP_PMTUD_ENABLE - This field will enable PMTU
2346 * discovery upon the specified address. Note that
2347 * if the address feild is empty then all addresses
2348 * on the association are effected.
2350 * SPP_PMTUD_DISABLE - This field will disable PMTU
2351 * discovery upon the specified address. Note that
2352 * if the address feild is empty then all addresses
2353 * on the association are effected. Not also that
2354 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2355 * exclusive. Enabling both will have undetermined
2358 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2359 * on delayed sack. The time specified in spp_sackdelay
2360 * is used to specify the sack delay for this address. Note
2361 * that if spp_address is empty then all addresses will
2362 * enable delayed sack and take on the sack delay
2363 * value specified in spp_sackdelay.
2364 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2365 * off delayed sack. If the spp_address field is blank then
2366 * delayed sack is disabled for the entire association. Note
2367 * also that this field is mutually exclusive to
2368 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2371 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2372 struct sctp_transport *trans,
2373 struct sctp_association *asoc,
2374 struct sctp_sock *sp,
2377 int sackdelay_change)
2381 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2382 struct net *net = sock_net(trans->asoc->base.sk);
2384 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2389 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2390 * this field is ignored. Note also that a value of zero indicates
2391 * the current setting should be left unchanged.
2393 if (params->spp_flags & SPP_HB_ENABLE) {
2395 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2396 * set. This lets us use 0 value when this flag
2399 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2400 params->spp_hbinterval = 0;
2402 if (params->spp_hbinterval ||
2403 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2406 msecs_to_jiffies(params->spp_hbinterval);
2409 msecs_to_jiffies(params->spp_hbinterval);
2411 sp->hbinterval = params->spp_hbinterval;
2418 trans->param_flags =
2419 (trans->param_flags & ~SPP_HB) | hb_change;
2422 (asoc->param_flags & ~SPP_HB) | hb_change;
2425 (sp->param_flags & ~SPP_HB) | hb_change;
2429 /* When Path MTU discovery is disabled the value specified here will
2430 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2431 * include the flag SPP_PMTUD_DISABLE for this field to have any
2434 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2436 trans->pathmtu = params->spp_pathmtu;
2437 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2439 asoc->pathmtu = params->spp_pathmtu;
2441 sp->pathmtu = params->spp_pathmtu;
2447 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2448 (params->spp_flags & SPP_PMTUD_ENABLE);
2449 trans->param_flags =
2450 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2452 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2453 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2457 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2460 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2464 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2465 * value of this field is ignored. Note also that a value of zero
2466 * indicates the current setting should be left unchanged.
2468 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2471 msecs_to_jiffies(params->spp_sackdelay);
2474 msecs_to_jiffies(params->spp_sackdelay);
2476 sp->sackdelay = params->spp_sackdelay;
2480 if (sackdelay_change) {
2482 trans->param_flags =
2483 (trans->param_flags & ~SPP_SACKDELAY) |
2487 (asoc->param_flags & ~SPP_SACKDELAY) |
2491 (sp->param_flags & ~SPP_SACKDELAY) |
2496 /* Note that a value of zero indicates the current setting should be
2499 if (params->spp_pathmaxrxt) {
2501 trans->pathmaxrxt = params->spp_pathmaxrxt;
2503 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2505 sp->pathmaxrxt = params->spp_pathmaxrxt;
2512 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2513 char __user *optval,
2514 unsigned int optlen)
2516 struct sctp_paddrparams params;
2517 struct sctp_transport *trans = NULL;
2518 struct sctp_association *asoc = NULL;
2519 struct sctp_sock *sp = sctp_sk(sk);
2521 int hb_change, pmtud_change, sackdelay_change;
2523 if (optlen != sizeof(struct sctp_paddrparams))
2526 if (copy_from_user(¶ms, optval, optlen))
2529 /* Validate flags and value parameters. */
2530 hb_change = params.spp_flags & SPP_HB;
2531 pmtud_change = params.spp_flags & SPP_PMTUD;
2532 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2534 if (hb_change == SPP_HB ||
2535 pmtud_change == SPP_PMTUD ||
2536 sackdelay_change == SPP_SACKDELAY ||
2537 params.spp_sackdelay > 500 ||
2538 (params.spp_pathmtu &&
2539 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2542 /* If an address other than INADDR_ANY is specified, and
2543 * no transport is found, then the request is invalid.
2545 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
2546 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2547 params.spp_assoc_id);
2552 /* Get association, if assoc_id != 0 and the socket is a one
2553 * to many style socket, and an association was not found, then
2554 * the id was invalid.
2556 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2557 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2560 /* Heartbeat demand can only be sent on a transport or
2561 * association, but not a socket.
2563 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2566 /* Process parameters. */
2567 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2568 hb_change, pmtud_change,
2574 /* If changes are for association, also apply parameters to each
2577 if (!trans && asoc) {
2578 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2580 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2581 hb_change, pmtud_change,
2589 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2591 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2594 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2596 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2600 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2602 * This option will effect the way delayed acks are performed. This
2603 * option allows you to get or set the delayed ack time, in
2604 * milliseconds. It also allows changing the delayed ack frequency.
2605 * Changing the frequency to 1 disables the delayed sack algorithm. If
2606 * the assoc_id is 0, then this sets or gets the endpoints default
2607 * values. If the assoc_id field is non-zero, then the set or get
2608 * effects the specified association for the one to many model (the
2609 * assoc_id field is ignored by the one to one model). Note that if
2610 * sack_delay or sack_freq are 0 when setting this option, then the
2611 * current values will remain unchanged.
2613 * struct sctp_sack_info {
2614 * sctp_assoc_t sack_assoc_id;
2615 * uint32_t sack_delay;
2616 * uint32_t sack_freq;
2619 * sack_assoc_id - This parameter, indicates which association the user
2620 * is performing an action upon. Note that if this field's value is
2621 * zero then the endpoints default value is changed (effecting future
2622 * associations only).
2624 * sack_delay - This parameter contains the number of milliseconds that
2625 * the user is requesting the delayed ACK timer be set to. Note that
2626 * this value is defined in the standard to be between 200 and 500
2629 * sack_freq - This parameter contains the number of packets that must
2630 * be received before a sack is sent without waiting for the delay
2631 * timer to expire. The default value for this is 2, setting this
2632 * value to 1 will disable the delayed sack algorithm.
2635 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2636 char __user *optval, unsigned int optlen)
2638 struct sctp_sack_info params;
2639 struct sctp_transport *trans = NULL;
2640 struct sctp_association *asoc = NULL;
2641 struct sctp_sock *sp = sctp_sk(sk);
2643 if (optlen == sizeof(struct sctp_sack_info)) {
2644 if (copy_from_user(¶ms, optval, optlen))
2647 if (params.sack_delay == 0 && params.sack_freq == 0)
2649 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2650 pr_warn_ratelimited(DEPRECATED
2652 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2653 "Use struct sctp_sack_info instead\n",
2654 current->comm, task_pid_nr(current));
2655 if (copy_from_user(¶ms, optval, optlen))
2658 if (params.sack_delay == 0)
2659 params.sack_freq = 1;
2661 params.sack_freq = 0;
2665 /* Validate value parameter. */
2666 if (params.sack_delay > 500)
2669 /* Get association, if sack_assoc_id != 0 and the socket is a one
2670 * to many style socket, and an association was not found, then
2671 * the id was invalid.
2673 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2674 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2677 if (params.sack_delay) {
2680 msecs_to_jiffies(params.sack_delay);
2682 sctp_spp_sackdelay_enable(asoc->param_flags);
2684 sp->sackdelay = params.sack_delay;
2686 sctp_spp_sackdelay_enable(sp->param_flags);
2690 if (params.sack_freq == 1) {
2693 sctp_spp_sackdelay_disable(asoc->param_flags);
2696 sctp_spp_sackdelay_disable(sp->param_flags);
2698 } else if (params.sack_freq > 1) {
2700 asoc->sackfreq = params.sack_freq;
2702 sctp_spp_sackdelay_enable(asoc->param_flags);
2704 sp->sackfreq = params.sack_freq;
2706 sctp_spp_sackdelay_enable(sp->param_flags);
2710 /* If change is for association, also apply to each transport. */
2712 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2714 if (params.sack_delay) {
2716 msecs_to_jiffies(params.sack_delay);
2717 trans->param_flags =
2718 sctp_spp_sackdelay_enable(trans->param_flags);
2720 if (params.sack_freq == 1) {
2721 trans->param_flags =
2722 sctp_spp_sackdelay_disable(trans->param_flags);
2723 } else if (params.sack_freq > 1) {
2724 trans->sackfreq = params.sack_freq;
2725 trans->param_flags =
2726 sctp_spp_sackdelay_enable(trans->param_flags);
2734 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2736 * Applications can specify protocol parameters for the default association
2737 * initialization. The option name argument to setsockopt() and getsockopt()
2740 * Setting initialization parameters is effective only on an unconnected
2741 * socket (for UDP-style sockets only future associations are effected
2742 * by the change). With TCP-style sockets, this option is inherited by
2743 * sockets derived from a listener socket.
2745 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2747 struct sctp_initmsg sinit;
2748 struct sctp_sock *sp = sctp_sk(sk);
2750 if (optlen != sizeof(struct sctp_initmsg))
2752 if (copy_from_user(&sinit, optval, optlen))
2755 if (sinit.sinit_num_ostreams)
2756 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2757 if (sinit.sinit_max_instreams)
2758 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2759 if (sinit.sinit_max_attempts)
2760 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2761 if (sinit.sinit_max_init_timeo)
2762 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2768 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2770 * Applications that wish to use the sendto() system call may wish to
2771 * specify a default set of parameters that would normally be supplied
2772 * through the inclusion of ancillary data. This socket option allows
2773 * such an application to set the default sctp_sndrcvinfo structure.
2774 * The application that wishes to use this socket option simply passes
2775 * in to this call the sctp_sndrcvinfo structure defined in Section
2776 * 5.2.2) The input parameters accepted by this call include
2777 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2778 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2779 * to this call if the caller is using the UDP model.
2781 static int sctp_setsockopt_default_send_param(struct sock *sk,
2782 char __user *optval,
2783 unsigned int optlen)
2785 struct sctp_sock *sp = sctp_sk(sk);
2786 struct sctp_association *asoc;
2787 struct sctp_sndrcvinfo info;
2789 if (optlen != sizeof(info))
2791 if (copy_from_user(&info, optval, optlen))
2793 if (info.sinfo_flags &
2794 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2795 SCTP_ABORT | SCTP_EOF))
2798 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2799 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2802 asoc->default_stream = info.sinfo_stream;
2803 asoc->default_flags = info.sinfo_flags;
2804 asoc->default_ppid = info.sinfo_ppid;
2805 asoc->default_context = info.sinfo_context;
2806 asoc->default_timetolive = info.sinfo_timetolive;
2808 sp->default_stream = info.sinfo_stream;
2809 sp->default_flags = info.sinfo_flags;
2810 sp->default_ppid = info.sinfo_ppid;
2811 sp->default_context = info.sinfo_context;
2812 sp->default_timetolive = info.sinfo_timetolive;
2818 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2819 * (SCTP_DEFAULT_SNDINFO)
2821 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
2822 char __user *optval,
2823 unsigned int optlen)
2825 struct sctp_sock *sp = sctp_sk(sk);
2826 struct sctp_association *asoc;
2827 struct sctp_sndinfo info;
2829 if (optlen != sizeof(info))
2831 if (copy_from_user(&info, optval, optlen))
2833 if (info.snd_flags &
2834 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2835 SCTP_ABORT | SCTP_EOF))
2838 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
2839 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
2842 asoc->default_stream = info.snd_sid;
2843 asoc->default_flags = info.snd_flags;
2844 asoc->default_ppid = info.snd_ppid;
2845 asoc->default_context = info.snd_context;
2847 sp->default_stream = info.snd_sid;
2848 sp->default_flags = info.snd_flags;
2849 sp->default_ppid = info.snd_ppid;
2850 sp->default_context = info.snd_context;
2856 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2858 * Requests that the local SCTP stack use the enclosed peer address as
2859 * the association primary. The enclosed address must be one of the
2860 * association peer's addresses.
2862 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2863 unsigned int optlen)
2865 struct sctp_prim prim;
2866 struct sctp_transport *trans;
2868 if (optlen != sizeof(struct sctp_prim))
2871 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2874 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2878 sctp_assoc_set_primary(trans->asoc, trans);
2884 * 7.1.5 SCTP_NODELAY
2886 * Turn on/off any Nagle-like algorithm. This means that packets are
2887 * generally sent as soon as possible and no unnecessary delays are
2888 * introduced, at the cost of more packets in the network. Expects an
2889 * integer boolean flag.
2891 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2892 unsigned int optlen)
2896 if (optlen < sizeof(int))
2898 if (get_user(val, (int __user *)optval))
2901 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2907 * 7.1.1 SCTP_RTOINFO
2909 * The protocol parameters used to initialize and bound retransmission
2910 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2911 * and modify these parameters.
2912 * All parameters are time values, in milliseconds. A value of 0, when
2913 * modifying the parameters, indicates that the current value should not
2917 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2919 struct sctp_rtoinfo rtoinfo;
2920 struct sctp_association *asoc;
2921 unsigned long rto_min, rto_max;
2922 struct sctp_sock *sp = sctp_sk(sk);
2924 if (optlen != sizeof (struct sctp_rtoinfo))
2927 if (copy_from_user(&rtoinfo, optval, optlen))
2930 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2932 /* Set the values to the specific association */
2933 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2936 rto_max = rtoinfo.srto_max;
2937 rto_min = rtoinfo.srto_min;
2940 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2942 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2945 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2947 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2949 if (rto_min > rto_max)
2953 if (rtoinfo.srto_initial != 0)
2955 msecs_to_jiffies(rtoinfo.srto_initial);
2956 asoc->rto_max = rto_max;
2957 asoc->rto_min = rto_min;
2959 /* If there is no association or the association-id = 0
2960 * set the values to the endpoint.
2962 if (rtoinfo.srto_initial != 0)
2963 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2964 sp->rtoinfo.srto_max = rto_max;
2965 sp->rtoinfo.srto_min = rto_min;
2973 * 7.1.2 SCTP_ASSOCINFO
2975 * This option is used to tune the maximum retransmission attempts
2976 * of the association.
2977 * Returns an error if the new association retransmission value is
2978 * greater than the sum of the retransmission value of the peer.
2979 * See [SCTP] for more information.
2982 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2985 struct sctp_assocparams assocparams;
2986 struct sctp_association *asoc;
2988 if (optlen != sizeof(struct sctp_assocparams))
2990 if (copy_from_user(&assocparams, optval, optlen))
2993 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2995 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2998 /* Set the values to the specific association */
3000 if (assocparams.sasoc_asocmaxrxt != 0) {
3003 struct sctp_transport *peer_addr;
3005 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
3007 path_sum += peer_addr->pathmaxrxt;
3011 /* Only validate asocmaxrxt if we have more than
3012 * one path/transport. We do this because path
3013 * retransmissions are only counted when we have more
3017 assocparams.sasoc_asocmaxrxt > path_sum)
3020 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3023 if (assocparams.sasoc_cookie_life != 0)
3024 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3026 /* Set the values to the endpoint */
3027 struct sctp_sock *sp = sctp_sk(sk);
3029 if (assocparams.sasoc_asocmaxrxt != 0)
3030 sp->assocparams.sasoc_asocmaxrxt =
3031 assocparams.sasoc_asocmaxrxt;
3032 if (assocparams.sasoc_cookie_life != 0)
3033 sp->assocparams.sasoc_cookie_life =
3034 assocparams.sasoc_cookie_life;
3040 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3042 * This socket option is a boolean flag which turns on or off mapped V4
3043 * addresses. If this option is turned on and the socket is type
3044 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3045 * If this option is turned off, then no mapping will be done of V4
3046 * addresses and a user will receive both PF_INET6 and PF_INET type
3047 * addresses on the socket.
3049 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3052 struct sctp_sock *sp = sctp_sk(sk);
3054 if (optlen < sizeof(int))
3056 if (get_user(val, (int __user *)optval))
3067 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3068 * This option will get or set the maximum size to put in any outgoing
3069 * SCTP DATA chunk. If a message is larger than this size it will be
3070 * fragmented by SCTP into the specified size. Note that the underlying
3071 * SCTP implementation may fragment into smaller sized chunks when the
3072 * PMTU of the underlying association is smaller than the value set by
3073 * the user. The default value for this option is '0' which indicates
3074 * the user is NOT limiting fragmentation and only the PMTU will effect
3075 * SCTP's choice of DATA chunk size. Note also that values set larger
3076 * than the maximum size of an IP datagram will effectively let SCTP
3077 * control fragmentation (i.e. the same as setting this option to 0).
3079 * The following structure is used to access and modify this parameter:
3081 * struct sctp_assoc_value {
3082 * sctp_assoc_t assoc_id;
3083 * uint32_t assoc_value;
3086 * assoc_id: This parameter is ignored for one-to-one style sockets.
3087 * For one-to-many style sockets this parameter indicates which
3088 * association the user is performing an action upon. Note that if
3089 * this field's value is zero then the endpoints default value is
3090 * changed (effecting future associations only).
3091 * assoc_value: This parameter specifies the maximum size in bytes.
3093 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3095 struct sctp_assoc_value params;
3096 struct sctp_association *asoc;
3097 struct sctp_sock *sp = sctp_sk(sk);
3100 if (optlen == sizeof(int)) {
3101 pr_warn_ratelimited(DEPRECATED
3103 "Use of int in maxseg socket option.\n"
3104 "Use struct sctp_assoc_value instead\n",
3105 current->comm, task_pid_nr(current));
3106 if (copy_from_user(&val, optval, optlen))
3108 params.assoc_id = 0;
3109 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3110 if (copy_from_user(¶ms, optval, optlen))
3112 val = params.assoc_value;
3116 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3119 asoc = sctp_id2assoc(sk, params.assoc_id);
3120 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3125 val = asoc->pathmtu;
3126 val -= sp->pf->af->net_header_len;
3127 val -= sizeof(struct sctphdr) +
3128 sizeof(struct sctp_data_chunk);
3130 asoc->user_frag = val;
3131 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3133 sp->user_frag = val;
3141 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3143 * Requests that the peer mark the enclosed address as the association
3144 * primary. The enclosed address must be one of the association's
3145 * locally bound addresses. The following structure is used to make a
3146 * set primary request:
3148 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3149 unsigned int optlen)
3151 struct net *net = sock_net(sk);
3152 struct sctp_sock *sp;
3153 struct sctp_association *asoc = NULL;
3154 struct sctp_setpeerprim prim;
3155 struct sctp_chunk *chunk;
3161 if (!net->sctp.addip_enable)
3164 if (optlen != sizeof(struct sctp_setpeerprim))
3167 if (copy_from_user(&prim, optval, optlen))
3170 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3174 if (!asoc->peer.asconf_capable)
3177 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3180 if (!sctp_state(asoc, ESTABLISHED))
3183 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3187 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3188 return -EADDRNOTAVAIL;
3190 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3191 return -EADDRNOTAVAIL;
3193 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3194 chunk = sctp_make_asconf_set_prim(asoc,
3195 (union sctp_addr *)&prim.sspp_addr);
3199 err = sctp_send_asconf(asoc, chunk);
3201 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3206 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3207 unsigned int optlen)
3209 struct sctp_setadaptation adaptation;
3211 if (optlen != sizeof(struct sctp_setadaptation))
3213 if (copy_from_user(&adaptation, optval, optlen))
3216 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3222 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3224 * The context field in the sctp_sndrcvinfo structure is normally only
3225 * used when a failed message is retrieved holding the value that was
3226 * sent down on the actual send call. This option allows the setting of
3227 * a default context on an association basis that will be received on
3228 * reading messages from the peer. This is especially helpful in the
3229 * one-2-many model for an application to keep some reference to an
3230 * internal state machine that is processing messages on the
3231 * association. Note that the setting of this value only effects
3232 * received messages from the peer and does not effect the value that is
3233 * saved with outbound messages.
3235 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3236 unsigned int optlen)
3238 struct sctp_assoc_value params;
3239 struct sctp_sock *sp;
3240 struct sctp_association *asoc;
3242 if (optlen != sizeof(struct sctp_assoc_value))
3244 if (copy_from_user(¶ms, optval, optlen))
3249 if (params.assoc_id != 0) {
3250 asoc = sctp_id2assoc(sk, params.assoc_id);
3253 asoc->default_rcv_context = params.assoc_value;
3255 sp->default_rcv_context = params.assoc_value;
3262 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3264 * This options will at a minimum specify if the implementation is doing
3265 * fragmented interleave. Fragmented interleave, for a one to many
3266 * socket, is when subsequent calls to receive a message may return
3267 * parts of messages from different associations. Some implementations
3268 * may allow you to turn this value on or off. If so, when turned off,
3269 * no fragment interleave will occur (which will cause a head of line
3270 * blocking amongst multiple associations sharing the same one to many
3271 * socket). When this option is turned on, then each receive call may
3272 * come from a different association (thus the user must receive data
3273 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3274 * association each receive belongs to.
3276 * This option takes a boolean value. A non-zero value indicates that
3277 * fragmented interleave is on. A value of zero indicates that
3278 * fragmented interleave is off.
3280 * Note that it is important that an implementation that allows this
3281 * option to be turned on, have it off by default. Otherwise an unaware
3282 * application using the one to many model may become confused and act
3285 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3286 char __user *optval,
3287 unsigned int optlen)
3291 if (optlen != sizeof(int))
3293 if (get_user(val, (int __user *)optval))
3296 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3302 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3303 * (SCTP_PARTIAL_DELIVERY_POINT)
3305 * This option will set or get the SCTP partial delivery point. This
3306 * point is the size of a message where the partial delivery API will be
3307 * invoked to help free up rwnd space for the peer. Setting this to a
3308 * lower value will cause partial deliveries to happen more often. The
3309 * calls argument is an integer that sets or gets the partial delivery
3310 * point. Note also that the call will fail if the user attempts to set
3311 * this value larger than the socket receive buffer size.
3313 * Note that any single message having a length smaller than or equal to
3314 * the SCTP partial delivery point will be delivered in one single read
3315 * call as long as the user provided buffer is large enough to hold the
3318 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3319 char __user *optval,
3320 unsigned int optlen)
3324 if (optlen != sizeof(u32))
3326 if (get_user(val, (int __user *)optval))
3329 /* Note: We double the receive buffer from what the user sets
3330 * it to be, also initial rwnd is based on rcvbuf/2.
3332 if (val > (sk->sk_rcvbuf >> 1))
3335 sctp_sk(sk)->pd_point = val;
3337 return 0; /* is this the right error code? */
3341 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3343 * This option will allow a user to change the maximum burst of packets
3344 * that can be emitted by this association. Note that the default value
3345 * is 4, and some implementations may restrict this setting so that it
3346 * can only be lowered.
3348 * NOTE: This text doesn't seem right. Do this on a socket basis with
3349 * future associations inheriting the socket value.
3351 static int sctp_setsockopt_maxburst(struct sock *sk,
3352 char __user *optval,
3353 unsigned int optlen)
3355 struct sctp_assoc_value params;
3356 struct sctp_sock *sp;
3357 struct sctp_association *asoc;
3361 if (optlen == sizeof(int)) {
3362 pr_warn_ratelimited(DEPRECATED
3364 "Use of int in max_burst socket option deprecated.\n"
3365 "Use struct sctp_assoc_value instead\n",
3366 current->comm, task_pid_nr(current));
3367 if (copy_from_user(&val, optval, optlen))
3369 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3370 if (copy_from_user(¶ms, optval, optlen))
3372 val = params.assoc_value;
3373 assoc_id = params.assoc_id;
3379 if (assoc_id != 0) {
3380 asoc = sctp_id2assoc(sk, assoc_id);
3383 asoc->max_burst = val;
3385 sp->max_burst = val;
3391 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3393 * This set option adds a chunk type that the user is requesting to be
3394 * received only in an authenticated way. Changes to the list of chunks
3395 * will only effect future associations on the socket.
3397 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3398 char __user *optval,
3399 unsigned int optlen)
3401 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3402 struct sctp_authchunk val;
3404 if (!ep->auth_enable)
3407 if (optlen != sizeof(struct sctp_authchunk))
3409 if (copy_from_user(&val, optval, optlen))
3412 switch (val.sauth_chunk) {
3414 case SCTP_CID_INIT_ACK:
3415 case SCTP_CID_SHUTDOWN_COMPLETE:
3420 /* add this chunk id to the endpoint */
3421 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3425 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3427 * This option gets or sets the list of HMAC algorithms that the local
3428 * endpoint requires the peer to use.
3430 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3431 char __user *optval,
3432 unsigned int optlen)
3434 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3435 struct sctp_hmacalgo *hmacs;
3439 if (!ep->auth_enable)
3442 if (optlen < sizeof(struct sctp_hmacalgo))
3445 hmacs = memdup_user(optval, optlen);
3447 return PTR_ERR(hmacs);
3449 idents = hmacs->shmac_num_idents;
3450 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3451 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3456 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3463 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3465 * This option will set a shared secret key which is used to build an
3466 * association shared key.
3468 static int sctp_setsockopt_auth_key(struct sock *sk,
3469 char __user *optval,
3470 unsigned int optlen)
3472 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3473 struct sctp_authkey *authkey;
3474 struct sctp_association *asoc;
3477 if (!ep->auth_enable)
3480 if (optlen <= sizeof(struct sctp_authkey))
3483 authkey = memdup_user(optval, optlen);
3484 if (IS_ERR(authkey))
3485 return PTR_ERR(authkey);
3487 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3492 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3493 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3498 ret = sctp_auth_set_key(ep, asoc, authkey);
3505 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3507 * This option will get or set the active shared key to be used to build
3508 * the association shared key.
3510 static int sctp_setsockopt_active_key(struct sock *sk,
3511 char __user *optval,
3512 unsigned int optlen)
3514 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3515 struct sctp_authkeyid val;
3516 struct sctp_association *asoc;
3518 if (!ep->auth_enable)
3521 if (optlen != sizeof(struct sctp_authkeyid))
3523 if (copy_from_user(&val, optval, optlen))
3526 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3527 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3530 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3534 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3536 * This set option will delete a shared secret key from use.
3538 static int sctp_setsockopt_del_key(struct sock *sk,
3539 char __user *optval,
3540 unsigned int optlen)
3542 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3543 struct sctp_authkeyid val;
3544 struct sctp_association *asoc;
3546 if (!ep->auth_enable)
3549 if (optlen != sizeof(struct sctp_authkeyid))
3551 if (copy_from_user(&val, optval, optlen))
3554 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3555 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3558 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3563 * 8.1.23 SCTP_AUTO_ASCONF
3565 * This option will enable or disable the use of the automatic generation of
3566 * ASCONF chunks to add and delete addresses to an existing association. Note
3567 * that this option has two caveats namely: a) it only affects sockets that
3568 * are bound to all addresses available to the SCTP stack, and b) the system
3569 * administrator may have an overriding control that turns the ASCONF feature
3570 * off no matter what setting the socket option may have.
3571 * This option expects an integer boolean flag, where a non-zero value turns on
3572 * the option, and a zero value turns off the option.
3573 * Note. In this implementation, socket operation overrides default parameter
3574 * being set by sysctl as well as FreeBSD implementation
3576 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3577 unsigned int optlen)
3580 struct sctp_sock *sp = sctp_sk(sk);
3582 if (optlen < sizeof(int))
3584 if (get_user(val, (int __user *)optval))
3586 if (!sctp_is_ep_boundall(sk) && val)
3588 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3591 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3592 if (val == 0 && sp->do_auto_asconf) {
3593 list_del(&sp->auto_asconf_list);
3594 sp->do_auto_asconf = 0;
3595 } else if (val && !sp->do_auto_asconf) {
3596 list_add_tail(&sp->auto_asconf_list,
3597 &sock_net(sk)->sctp.auto_asconf_splist);
3598 sp->do_auto_asconf = 1;
3600 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3605 * SCTP_PEER_ADDR_THLDS
3607 * This option allows us to alter the partially failed threshold for one or all
3608 * transports in an association. See Section 6.1 of:
3609 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3611 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3612 char __user *optval,
3613 unsigned int optlen)
3615 struct sctp_paddrthlds val;
3616 struct sctp_transport *trans;
3617 struct sctp_association *asoc;
3619 if (optlen < sizeof(struct sctp_paddrthlds))
3621 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3622 sizeof(struct sctp_paddrthlds)))
3626 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3627 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3630 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3632 if (val.spt_pathmaxrxt)
3633 trans->pathmaxrxt = val.spt_pathmaxrxt;
3634 trans->pf_retrans = val.spt_pathpfthld;
3637 if (val.spt_pathmaxrxt)
3638 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3639 asoc->pf_retrans = val.spt_pathpfthld;
3641 trans = sctp_addr_id2transport(sk, &val.spt_address,
3646 if (val.spt_pathmaxrxt)
3647 trans->pathmaxrxt = val.spt_pathmaxrxt;
3648 trans->pf_retrans = val.spt_pathpfthld;
3654 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3655 char __user *optval,
3656 unsigned int optlen)
3660 if (optlen < sizeof(int))
3662 if (get_user(val, (int __user *) optval))
3665 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3670 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3671 char __user *optval,
3672 unsigned int optlen)
3676 if (optlen < sizeof(int))
3678 if (get_user(val, (int __user *) optval))
3681 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3686 static int sctp_setsockopt_pr_supported(struct sock *sk,
3687 char __user *optval,
3688 unsigned int optlen)
3690 struct sctp_assoc_value params;
3691 struct sctp_association *asoc;
3692 int retval = -EINVAL;
3694 if (optlen != sizeof(params))
3697 if (copy_from_user(¶ms, optval, optlen)) {
3702 asoc = sctp_id2assoc(sk, params.assoc_id);
3704 asoc->prsctp_enable = !!params.assoc_value;
3705 } else if (!params.assoc_id) {
3706 struct sctp_sock *sp = sctp_sk(sk);
3708 sp->ep->prsctp_enable = !!params.assoc_value;
3719 static int sctp_setsockopt_default_prinfo(struct sock *sk,
3720 char __user *optval,
3721 unsigned int optlen)
3723 struct sctp_default_prinfo info;
3724 struct sctp_association *asoc;
3725 int retval = -EINVAL;
3727 if (optlen != sizeof(info))
3730 if (copy_from_user(&info, optval, sizeof(info))) {
3735 if (info.pr_policy & ~SCTP_PR_SCTP_MASK)
3738 if (info.pr_policy == SCTP_PR_SCTP_NONE)
3741 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
3743 SCTP_PR_SET_POLICY(asoc->default_flags, info.pr_policy);
3744 asoc->default_timetolive = info.pr_value;
3745 } else if (!info.pr_assoc_id) {
3746 struct sctp_sock *sp = sctp_sk(sk);
3748 SCTP_PR_SET_POLICY(sp->default_flags, info.pr_policy);
3749 sp->default_timetolive = info.pr_value;
3760 static int sctp_setsockopt_enable_strreset(struct sock *sk,
3761 char __user *optval,
3762 unsigned int optlen)
3764 struct sctp_assoc_value params;
3765 struct sctp_association *asoc;
3766 int retval = -EINVAL;
3768 if (optlen != sizeof(params))
3771 if (copy_from_user(¶ms, optval, optlen)) {
3776 if (params.assoc_value & (~SCTP_ENABLE_STRRESET_MASK))
3779 asoc = sctp_id2assoc(sk, params.assoc_id);
3781 asoc->strreset_enable = params.assoc_value;
3782 } else if (!params.assoc_id) {
3783 struct sctp_sock *sp = sctp_sk(sk);
3785 sp->ep->strreset_enable = params.assoc_value;
3796 static int sctp_setsockopt_reset_streams(struct sock *sk,
3797 char __user *optval,
3798 unsigned int optlen)
3800 struct sctp_reset_streams *params;
3801 struct sctp_association *asoc;
3802 int retval = -EINVAL;
3804 if (optlen < sizeof(struct sctp_reset_streams))
3807 params = memdup_user(optval, optlen);
3809 return PTR_ERR(params);
3811 asoc = sctp_id2assoc(sk, params->srs_assoc_id);
3815 retval = sctp_send_reset_streams(asoc, params);
3822 static int sctp_setsockopt_reset_assoc(struct sock *sk,
3823 char __user *optval,
3824 unsigned int optlen)
3826 struct sctp_association *asoc;
3827 sctp_assoc_t associd;
3828 int retval = -EINVAL;
3830 if (optlen != sizeof(associd))
3833 if (copy_from_user(&associd, optval, optlen)) {
3838 asoc = sctp_id2assoc(sk, associd);
3842 retval = sctp_send_reset_assoc(asoc);
3848 static int sctp_setsockopt_add_streams(struct sock *sk,
3849 char __user *optval,
3850 unsigned int optlen)
3852 struct sctp_association *asoc;
3853 struct sctp_add_streams params;
3854 int retval = -EINVAL;
3856 if (optlen != sizeof(params))
3859 if (copy_from_user(¶ms, optval, optlen)) {
3864 asoc = sctp_id2assoc(sk, params.sas_assoc_id);
3868 retval = sctp_send_add_streams(asoc, ¶ms);
3874 /* API 6.2 setsockopt(), getsockopt()
3876 * Applications use setsockopt() and getsockopt() to set or retrieve
3877 * socket options. Socket options are used to change the default
3878 * behavior of sockets calls. They are described in Section 7.
3882 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3883 * int __user *optlen);
3884 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3887 * sd - the socket descript.
3888 * level - set to IPPROTO_SCTP for all SCTP options.
3889 * optname - the option name.
3890 * optval - the buffer to store the value of the option.
3891 * optlen - the size of the buffer.
3893 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3894 char __user *optval, unsigned int optlen)
3898 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3900 /* I can hardly begin to describe how wrong this is. This is
3901 * so broken as to be worse than useless. The API draft
3902 * REALLY is NOT helpful here... I am not convinced that the
3903 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3904 * are at all well-founded.
3906 if (level != SOL_SCTP) {
3907 struct sctp_af *af = sctp_sk(sk)->pf->af;
3908 retval = af->setsockopt(sk, level, optname, optval, optlen);
3915 case SCTP_SOCKOPT_BINDX_ADD:
3916 /* 'optlen' is the size of the addresses buffer. */
3917 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3918 optlen, SCTP_BINDX_ADD_ADDR);
3921 case SCTP_SOCKOPT_BINDX_REM:
3922 /* 'optlen' is the size of the addresses buffer. */
3923 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3924 optlen, SCTP_BINDX_REM_ADDR);
3927 case SCTP_SOCKOPT_CONNECTX_OLD:
3928 /* 'optlen' is the size of the addresses buffer. */
3929 retval = sctp_setsockopt_connectx_old(sk,
3930 (struct sockaddr __user *)optval,
3934 case SCTP_SOCKOPT_CONNECTX:
3935 /* 'optlen' is the size of the addresses buffer. */
3936 retval = sctp_setsockopt_connectx(sk,
3937 (struct sockaddr __user *)optval,
3941 case SCTP_DISABLE_FRAGMENTS:
3942 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3946 retval = sctp_setsockopt_events(sk, optval, optlen);
3949 case SCTP_AUTOCLOSE:
3950 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3953 case SCTP_PEER_ADDR_PARAMS:
3954 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3957 case SCTP_DELAYED_SACK:
3958 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3960 case SCTP_PARTIAL_DELIVERY_POINT:
3961 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3965 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3967 case SCTP_DEFAULT_SEND_PARAM:
3968 retval = sctp_setsockopt_default_send_param(sk, optval,
3971 case SCTP_DEFAULT_SNDINFO:
3972 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
3974 case SCTP_PRIMARY_ADDR:
3975 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3977 case SCTP_SET_PEER_PRIMARY_ADDR:
3978 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3981 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3984 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3986 case SCTP_ASSOCINFO:
3987 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3989 case SCTP_I_WANT_MAPPED_V4_ADDR:
3990 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3993 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3995 case SCTP_ADAPTATION_LAYER:
3996 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3999 retval = sctp_setsockopt_context(sk, optval, optlen);
4001 case SCTP_FRAGMENT_INTERLEAVE:
4002 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
4004 case SCTP_MAX_BURST:
4005 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
4007 case SCTP_AUTH_CHUNK:
4008 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
4010 case SCTP_HMAC_IDENT:
4011 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
4014 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
4016 case SCTP_AUTH_ACTIVE_KEY:
4017 retval = sctp_setsockopt_active_key(sk, optval, optlen);
4019 case SCTP_AUTH_DELETE_KEY:
4020 retval = sctp_setsockopt_del_key(sk, optval, optlen);
4022 case SCTP_AUTO_ASCONF:
4023 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
4025 case SCTP_PEER_ADDR_THLDS:
4026 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
4028 case SCTP_RECVRCVINFO:
4029 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
4031 case SCTP_RECVNXTINFO:
4032 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
4034 case SCTP_PR_SUPPORTED:
4035 retval = sctp_setsockopt_pr_supported(sk, optval, optlen);
4037 case SCTP_DEFAULT_PRINFO:
4038 retval = sctp_setsockopt_default_prinfo(sk, optval, optlen);
4040 case SCTP_ENABLE_STREAM_RESET:
4041 retval = sctp_setsockopt_enable_strreset(sk, optval, optlen);
4043 case SCTP_RESET_STREAMS:
4044 retval = sctp_setsockopt_reset_streams(sk, optval, optlen);
4046 case SCTP_RESET_ASSOC:
4047 retval = sctp_setsockopt_reset_assoc(sk, optval, optlen);
4049 case SCTP_ADD_STREAMS:
4050 retval = sctp_setsockopt_add_streams(sk, optval, optlen);
4053 retval = -ENOPROTOOPT;
4063 /* API 3.1.6 connect() - UDP Style Syntax
4065 * An application may use the connect() call in the UDP model to initiate an
4066 * association without sending data.
4070 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
4072 * sd: the socket descriptor to have a new association added to.
4074 * nam: the address structure (either struct sockaddr_in or struct
4075 * sockaddr_in6 defined in RFC2553 [7]).
4077 * len: the size of the address.
4079 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
4087 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
4090 /* Validate addr_len before calling common connect/connectx routine. */
4091 af = sctp_get_af_specific(addr->sa_family);
4092 if (!af || addr_len < af->sockaddr_len) {
4095 /* Pass correct addr len to common routine (so it knows there
4096 * is only one address being passed.
4098 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
4105 /* FIXME: Write comments. */
4106 static int sctp_disconnect(struct sock *sk, int flags)
4108 return -EOPNOTSUPP; /* STUB */
4111 /* 4.1.4 accept() - TCP Style Syntax
4113 * Applications use accept() call to remove an established SCTP
4114 * association from the accept queue of the endpoint. A new socket
4115 * descriptor will be returned from accept() to represent the newly
4116 * formed association.
4118 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
4120 struct sctp_sock *sp;
4121 struct sctp_endpoint *ep;
4122 struct sock *newsk = NULL;
4123 struct sctp_association *asoc;
4132 if (!sctp_style(sk, TCP)) {
4133 error = -EOPNOTSUPP;
4137 if (!sctp_sstate(sk, LISTENING)) {
4142 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
4144 error = sctp_wait_for_accept(sk, timeo);
4148 /* We treat the list of associations on the endpoint as the accept
4149 * queue and pick the first association on the list.
4151 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
4153 newsk = sp->pf->create_accept_sk(sk, asoc);
4159 /* Populate the fields of the newsk from the oldsk and migrate the
4160 * asoc to the newsk.
4162 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
4170 /* The SCTP ioctl handler. */
4171 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
4178 * SEQPACKET-style sockets in LISTENING state are valid, for
4179 * SCTP, so only discard TCP-style sockets in LISTENING state.
4181 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4186 struct sk_buff *skb;
4187 unsigned int amount = 0;
4189 skb = skb_peek(&sk->sk_receive_queue);
4192 * We will only return the amount of this packet since
4193 * that is all that will be read.
4197 rc = put_user(amount, (int __user *)arg);
4209 /* This is the function which gets called during socket creation to
4210 * initialized the SCTP-specific portion of the sock.
4211 * The sock structure should already be zero-filled memory.
4213 static int sctp_init_sock(struct sock *sk)
4215 struct net *net = sock_net(sk);
4216 struct sctp_sock *sp;
4218 pr_debug("%s: sk:%p\n", __func__, sk);
4222 /* Initialize the SCTP per socket area. */
4223 switch (sk->sk_type) {
4224 case SOCK_SEQPACKET:
4225 sp->type = SCTP_SOCKET_UDP;
4228 sp->type = SCTP_SOCKET_TCP;
4231 return -ESOCKTNOSUPPORT;
4234 sk->sk_gso_type = SKB_GSO_SCTP;
4236 /* Initialize default send parameters. These parameters can be
4237 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4239 sp->default_stream = 0;
4240 sp->default_ppid = 0;
4241 sp->default_flags = 0;
4242 sp->default_context = 0;
4243 sp->default_timetolive = 0;
4245 sp->default_rcv_context = 0;
4246 sp->max_burst = net->sctp.max_burst;
4248 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4250 /* Initialize default setup parameters. These parameters
4251 * can be modified with the SCTP_INITMSG socket option or
4252 * overridden by the SCTP_INIT CMSG.
4254 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4255 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4256 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4257 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4259 /* Initialize default RTO related parameters. These parameters can
4260 * be modified for with the SCTP_RTOINFO socket option.
4262 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4263 sp->rtoinfo.srto_max = net->sctp.rto_max;
4264 sp->rtoinfo.srto_min = net->sctp.rto_min;
4266 /* Initialize default association related parameters. These parameters
4267 * can be modified with the SCTP_ASSOCINFO socket option.
4269 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4270 sp->assocparams.sasoc_number_peer_destinations = 0;
4271 sp->assocparams.sasoc_peer_rwnd = 0;
4272 sp->assocparams.sasoc_local_rwnd = 0;
4273 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4275 /* Initialize default event subscriptions. By default, all the
4278 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4280 /* Default Peer Address Parameters. These defaults can
4281 * be modified via SCTP_PEER_ADDR_PARAMS
4283 sp->hbinterval = net->sctp.hb_interval;
4284 sp->pathmaxrxt = net->sctp.max_retrans_path;
4285 sp->pathmtu = 0; /* allow default discovery */
4286 sp->sackdelay = net->sctp.sack_timeout;
4288 sp->param_flags = SPP_HB_ENABLE |
4290 SPP_SACKDELAY_ENABLE;
4292 /* If enabled no SCTP message fragmentation will be performed.
4293 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4295 sp->disable_fragments = 0;
4297 /* Enable Nagle algorithm by default. */
4300 sp->recvrcvinfo = 0;
4301 sp->recvnxtinfo = 0;
4303 /* Enable by default. */
4306 /* Auto-close idle associations after the configured
4307 * number of seconds. A value of 0 disables this
4308 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4309 * for UDP-style sockets only.
4313 /* User specified fragmentation limit. */
4316 sp->adaptation_ind = 0;
4318 sp->pf = sctp_get_pf_specific(sk->sk_family);
4320 /* Control variables for partial data delivery. */
4321 atomic_set(&sp->pd_mode, 0);
4322 skb_queue_head_init(&sp->pd_lobby);
4323 sp->frag_interleave = 0;
4325 /* Create a per socket endpoint structure. Even if we
4326 * change the data structure relationships, this may still
4327 * be useful for storing pre-connect address information.
4329 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4335 sk->sk_destruct = sctp_destruct_sock;
4337 SCTP_DBG_OBJCNT_INC(sock);
4340 percpu_counter_inc(&sctp_sockets_allocated);
4341 sock_prot_inuse_add(net, sk->sk_prot, 1);
4343 /* Nothing can fail after this block, otherwise
4344 * sctp_destroy_sock() will be called without addr_wq_lock held
4346 if (net->sctp.default_auto_asconf) {
4347 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
4348 list_add_tail(&sp->auto_asconf_list,
4349 &net->sctp.auto_asconf_splist);
4350 sp->do_auto_asconf = 1;
4351 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4353 sp->do_auto_asconf = 0;
4361 /* Cleanup any SCTP per socket resources. Must be called with
4362 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4364 static void sctp_destroy_sock(struct sock *sk)
4366 struct sctp_sock *sp;
4368 pr_debug("%s: sk:%p\n", __func__, sk);
4370 /* Release our hold on the endpoint. */
4372 /* This could happen during socket init, thus we bail out
4373 * early, since the rest of the below is not setup either.
4378 if (sp->do_auto_asconf) {
4379 sp->do_auto_asconf = 0;
4380 list_del(&sp->auto_asconf_list);
4382 sctp_endpoint_free(sp->ep);
4384 percpu_counter_dec(&sctp_sockets_allocated);
4385 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4389 /* Triggered when there are no references on the socket anymore */
4390 static void sctp_destruct_sock(struct sock *sk)
4392 struct sctp_sock *sp = sctp_sk(sk);
4394 /* Free up the HMAC transform. */
4395 crypto_free_shash(sp->hmac);
4397 inet_sock_destruct(sk);
4400 /* API 4.1.7 shutdown() - TCP Style Syntax
4401 * int shutdown(int socket, int how);
4403 * sd - the socket descriptor of the association to be closed.
4404 * how - Specifies the type of shutdown. The values are
4407 * Disables further receive operations. No SCTP
4408 * protocol action is taken.
4410 * Disables further send operations, and initiates
4411 * the SCTP shutdown sequence.
4413 * Disables further send and receive operations
4414 * and initiates the SCTP shutdown sequence.
4416 static void sctp_shutdown(struct sock *sk, int how)
4418 struct net *net = sock_net(sk);
4419 struct sctp_endpoint *ep;
4421 if (!sctp_style(sk, TCP))
4424 ep = sctp_sk(sk)->ep;
4425 if (how & SEND_SHUTDOWN && !list_empty(&ep->asocs)) {
4426 struct sctp_association *asoc;
4428 sk->sk_state = SCTP_SS_CLOSING;
4429 asoc = list_entry(ep->asocs.next,
4430 struct sctp_association, asocs);
4431 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4435 int sctp_get_sctp_info(struct sock *sk, struct sctp_association *asoc,
4436 struct sctp_info *info)
4438 struct sctp_transport *prim;
4439 struct list_head *pos;
4442 memset(info, 0, sizeof(*info));
4444 struct sctp_sock *sp = sctp_sk(sk);
4446 info->sctpi_s_autoclose = sp->autoclose;
4447 info->sctpi_s_adaptation_ind = sp->adaptation_ind;
4448 info->sctpi_s_pd_point = sp->pd_point;
4449 info->sctpi_s_nodelay = sp->nodelay;
4450 info->sctpi_s_disable_fragments = sp->disable_fragments;
4451 info->sctpi_s_v4mapped = sp->v4mapped;
4452 info->sctpi_s_frag_interleave = sp->frag_interleave;
4453 info->sctpi_s_type = sp->type;
4458 info->sctpi_tag = asoc->c.my_vtag;
4459 info->sctpi_state = asoc->state;
4460 info->sctpi_rwnd = asoc->a_rwnd;
4461 info->sctpi_unackdata = asoc->unack_data;
4462 info->sctpi_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4463 info->sctpi_instrms = asoc->c.sinit_max_instreams;
4464 info->sctpi_outstrms = asoc->c.sinit_num_ostreams;
4465 list_for_each(pos, &asoc->base.inqueue.in_chunk_list)
4466 info->sctpi_inqueue++;
4467 list_for_each(pos, &asoc->outqueue.out_chunk_list)
4468 info->sctpi_outqueue++;
4469 info->sctpi_overall_error = asoc->overall_error_count;
4470 info->sctpi_max_burst = asoc->max_burst;
4471 info->sctpi_maxseg = asoc->frag_point;
4472 info->sctpi_peer_rwnd = asoc->peer.rwnd;
4473 info->sctpi_peer_tag = asoc->c.peer_vtag;
4475 mask = asoc->peer.ecn_capable << 1;
4476 mask = (mask | asoc->peer.ipv4_address) << 1;
4477 mask = (mask | asoc->peer.ipv6_address) << 1;
4478 mask = (mask | asoc->peer.hostname_address) << 1;
4479 mask = (mask | asoc->peer.asconf_capable) << 1;
4480 mask = (mask | asoc->peer.prsctp_capable) << 1;
4481 mask = (mask | asoc->peer.auth_capable);
4482 info->sctpi_peer_capable = mask;
4483 mask = asoc->peer.sack_needed << 1;
4484 mask = (mask | asoc->peer.sack_generation) << 1;
4485 mask = (mask | asoc->peer.zero_window_announced);
4486 info->sctpi_peer_sack = mask;
4488 info->sctpi_isacks = asoc->stats.isacks;
4489 info->sctpi_osacks = asoc->stats.osacks;
4490 info->sctpi_opackets = asoc->stats.opackets;
4491 info->sctpi_ipackets = asoc->stats.ipackets;
4492 info->sctpi_rtxchunks = asoc->stats.rtxchunks;
4493 info->sctpi_outofseqtsns = asoc->stats.outofseqtsns;
4494 info->sctpi_idupchunks = asoc->stats.idupchunks;
4495 info->sctpi_gapcnt = asoc->stats.gapcnt;
4496 info->sctpi_ouodchunks = asoc->stats.ouodchunks;
4497 info->sctpi_iuodchunks = asoc->stats.iuodchunks;
4498 info->sctpi_oodchunks = asoc->stats.oodchunks;
4499 info->sctpi_iodchunks = asoc->stats.iodchunks;
4500 info->sctpi_octrlchunks = asoc->stats.octrlchunks;
4501 info->sctpi_ictrlchunks = asoc->stats.ictrlchunks;
4503 prim = asoc->peer.primary_path;
4504 memcpy(&info->sctpi_p_address, &prim->ipaddr,
4505 sizeof(struct sockaddr_storage));
4506 info->sctpi_p_state = prim->state;
4507 info->sctpi_p_cwnd = prim->cwnd;
4508 info->sctpi_p_srtt = prim->srtt;
4509 info->sctpi_p_rto = jiffies_to_msecs(prim->rto);
4510 info->sctpi_p_hbinterval = prim->hbinterval;
4511 info->sctpi_p_pathmaxrxt = prim->pathmaxrxt;
4512 info->sctpi_p_sackdelay = jiffies_to_msecs(prim->sackdelay);
4513 info->sctpi_p_ssthresh = prim->ssthresh;
4514 info->sctpi_p_partial_bytes_acked = prim->partial_bytes_acked;
4515 info->sctpi_p_flight_size = prim->flight_size;
4516 info->sctpi_p_error = prim->error_count;
4520 EXPORT_SYMBOL_GPL(sctp_get_sctp_info);
4522 /* use callback to avoid exporting the core structure */
4523 int sctp_transport_walk_start(struct rhashtable_iter *iter)
4527 rhltable_walk_enter(&sctp_transport_hashtable, iter);
4529 err = rhashtable_walk_start(iter);
4530 if (err && err != -EAGAIN) {
4531 rhashtable_walk_stop(iter);
4532 rhashtable_walk_exit(iter);
4539 void sctp_transport_walk_stop(struct rhashtable_iter *iter)
4541 rhashtable_walk_stop(iter);
4542 rhashtable_walk_exit(iter);
4545 struct sctp_transport *sctp_transport_get_next(struct net *net,
4546 struct rhashtable_iter *iter)
4548 struct sctp_transport *t;
4550 t = rhashtable_walk_next(iter);
4551 for (; t; t = rhashtable_walk_next(iter)) {
4553 if (PTR_ERR(t) == -EAGAIN)
4558 if (net_eq(sock_net(t->asoc->base.sk), net) &&
4559 t->asoc->peer.primary_path == t)
4566 struct sctp_transport *sctp_transport_get_idx(struct net *net,
4567 struct rhashtable_iter *iter,
4570 void *obj = SEQ_START_TOKEN;
4572 while (pos && (obj = sctp_transport_get_next(net, iter)) &&
4579 int sctp_for_each_endpoint(int (*cb)(struct sctp_endpoint *, void *),
4583 struct sctp_ep_common *epb;
4584 struct sctp_hashbucket *head;
4586 for (head = sctp_ep_hashtable; hash < sctp_ep_hashsize;
4588 read_lock(&head->lock);
4589 sctp_for_each_hentry(epb, &head->chain) {
4590 err = cb(sctp_ep(epb), p);
4594 read_unlock(&head->lock);
4599 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint);
4601 int sctp_transport_lookup_process(int (*cb)(struct sctp_transport *, void *),
4603 const union sctp_addr *laddr,
4604 const union sctp_addr *paddr, void *p)
4606 struct sctp_transport *transport;
4610 transport = sctp_addrs_lookup_transport(net, laddr, paddr);
4615 err = cb(transport, p);
4616 sctp_transport_put(transport);
4620 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process);
4622 int sctp_for_each_transport(int (*cb)(struct sctp_transport *, void *),
4623 struct net *net, int pos, void *p) {
4624 struct rhashtable_iter hti;
4628 err = sctp_transport_walk_start(&hti);
4632 sctp_transport_get_idx(net, &hti, pos);
4633 obj = sctp_transport_get_next(net, &hti);
4634 for (; obj && !IS_ERR(obj); obj = sctp_transport_get_next(net, &hti)) {
4635 struct sctp_transport *transport = obj;
4637 if (!sctp_transport_hold(transport))
4639 err = cb(transport, p);
4640 sctp_transport_put(transport);
4644 sctp_transport_walk_stop(&hti);
4648 EXPORT_SYMBOL_GPL(sctp_for_each_transport);
4650 /* 7.2.1 Association Status (SCTP_STATUS)
4652 * Applications can retrieve current status information about an
4653 * association, including association state, peer receiver window size,
4654 * number of unacked data chunks, and number of data chunks pending
4655 * receipt. This information is read-only.
4657 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4658 char __user *optval,
4661 struct sctp_status status;
4662 struct sctp_association *asoc = NULL;
4663 struct sctp_transport *transport;
4664 sctp_assoc_t associd;
4667 if (len < sizeof(status)) {
4672 len = sizeof(status);
4673 if (copy_from_user(&status, optval, len)) {
4678 associd = status.sstat_assoc_id;
4679 asoc = sctp_id2assoc(sk, associd);
4685 transport = asoc->peer.primary_path;
4687 status.sstat_assoc_id = sctp_assoc2id(asoc);
4688 status.sstat_state = sctp_assoc_to_state(asoc);
4689 status.sstat_rwnd = asoc->peer.rwnd;
4690 status.sstat_unackdata = asoc->unack_data;
4692 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4693 status.sstat_instrms = asoc->c.sinit_max_instreams;
4694 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4695 status.sstat_fragmentation_point = asoc->frag_point;
4696 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4697 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4698 transport->af_specific->sockaddr_len);
4699 /* Map ipv4 address into v4-mapped-on-v6 address. */
4700 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
4701 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4702 status.sstat_primary.spinfo_state = transport->state;
4703 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4704 status.sstat_primary.spinfo_srtt = transport->srtt;
4705 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4706 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4708 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4709 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4711 if (put_user(len, optlen)) {
4716 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4717 __func__, len, status.sstat_state, status.sstat_rwnd,
4718 status.sstat_assoc_id);
4720 if (copy_to_user(optval, &status, len)) {
4730 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4732 * Applications can retrieve information about a specific peer address
4733 * of an association, including its reachability state, congestion
4734 * window, and retransmission timer values. This information is
4737 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4738 char __user *optval,
4741 struct sctp_paddrinfo pinfo;
4742 struct sctp_transport *transport;
4745 if (len < sizeof(pinfo)) {
4750 len = sizeof(pinfo);
4751 if (copy_from_user(&pinfo, optval, len)) {
4756 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4757 pinfo.spinfo_assoc_id);
4761 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4762 pinfo.spinfo_state = transport->state;
4763 pinfo.spinfo_cwnd = transport->cwnd;
4764 pinfo.spinfo_srtt = transport->srtt;
4765 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4766 pinfo.spinfo_mtu = transport->pathmtu;
4768 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4769 pinfo.spinfo_state = SCTP_ACTIVE;
4771 if (put_user(len, optlen)) {
4776 if (copy_to_user(optval, &pinfo, len)) {
4785 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4787 * This option is a on/off flag. If enabled no SCTP message
4788 * fragmentation will be performed. Instead if a message being sent
4789 * exceeds the current PMTU size, the message will NOT be sent and
4790 * instead a error will be indicated to the user.
4792 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4793 char __user *optval, int __user *optlen)
4797 if (len < sizeof(int))
4801 val = (sctp_sk(sk)->disable_fragments == 1);
4802 if (put_user(len, optlen))
4804 if (copy_to_user(optval, &val, len))
4809 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4811 * This socket option is used to specify various notifications and
4812 * ancillary data the user wishes to receive.
4814 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4819 if (len > sizeof(struct sctp_event_subscribe))
4820 len = sizeof(struct sctp_event_subscribe);
4821 if (put_user(len, optlen))
4823 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4828 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4830 * This socket option is applicable to the UDP-style socket only. When
4831 * set it will cause associations that are idle for more than the
4832 * specified number of seconds to automatically close. An association
4833 * being idle is defined an association that has NOT sent or received
4834 * user data. The special value of '0' indicates that no automatic
4835 * close of any associations should be performed. The option expects an
4836 * integer defining the number of seconds of idle time before an
4837 * association is closed.
4839 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4841 /* Applicable to UDP-style socket only */
4842 if (sctp_style(sk, TCP))
4844 if (len < sizeof(int))
4847 if (put_user(len, optlen))
4849 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4854 /* Helper routine to branch off an association to a new socket. */
4855 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4857 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4858 struct sctp_sock *sp = sctp_sk(sk);
4859 struct socket *sock;
4865 /* If there is a thread waiting on more sndbuf space for
4866 * sending on this asoc, it cannot be peeled.
4868 if (waitqueue_active(&asoc->wait))
4871 /* An association cannot be branched off from an already peeled-off
4872 * socket, nor is this supported for tcp style sockets.
4874 if (!sctp_style(sk, UDP))
4877 /* Create a new socket. */
4878 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4882 sctp_copy_sock(sock->sk, sk, asoc);
4884 /* Make peeled-off sockets more like 1-1 accepted sockets.
4885 * Set the daddr and initialize id to something more random
4887 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
4889 /* Populate the fields of the newsk from the oldsk and migrate the
4890 * asoc to the newsk.
4892 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4898 EXPORT_SYMBOL(sctp_do_peeloff);
4900 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4902 sctp_peeloff_arg_t peeloff;
4903 struct socket *newsock;
4904 struct file *newfile;
4907 if (len < sizeof(sctp_peeloff_arg_t))
4909 len = sizeof(sctp_peeloff_arg_t);
4910 if (copy_from_user(&peeloff, optval, len))
4913 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4917 /* Map the socket to an unused fd that can be returned to the user. */
4918 retval = get_unused_fd_flags(0);
4920 sock_release(newsock);
4924 newfile = sock_alloc_file(newsock, 0, NULL);
4925 if (IS_ERR(newfile)) {
4926 put_unused_fd(retval);
4927 sock_release(newsock);
4928 return PTR_ERR(newfile);
4931 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4934 /* Return the fd mapped to the new socket. */
4935 if (put_user(len, optlen)) {
4937 put_unused_fd(retval);
4940 peeloff.sd = retval;
4941 if (copy_to_user(optval, &peeloff, len)) {
4943 put_unused_fd(retval);
4946 fd_install(retval, newfile);
4951 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4953 * Applications can enable or disable heartbeats for any peer address of
4954 * an association, modify an address's heartbeat interval, force a
4955 * heartbeat to be sent immediately, and adjust the address's maximum
4956 * number of retransmissions sent before an address is considered
4957 * unreachable. The following structure is used to access and modify an
4958 * address's parameters:
4960 * struct sctp_paddrparams {
4961 * sctp_assoc_t spp_assoc_id;
4962 * struct sockaddr_storage spp_address;
4963 * uint32_t spp_hbinterval;
4964 * uint16_t spp_pathmaxrxt;
4965 * uint32_t spp_pathmtu;
4966 * uint32_t spp_sackdelay;
4967 * uint32_t spp_flags;
4970 * spp_assoc_id - (one-to-many style socket) This is filled in the
4971 * application, and identifies the association for
4973 * spp_address - This specifies which address is of interest.
4974 * spp_hbinterval - This contains the value of the heartbeat interval,
4975 * in milliseconds. If a value of zero
4976 * is present in this field then no changes are to
4977 * be made to this parameter.
4978 * spp_pathmaxrxt - This contains the maximum number of
4979 * retransmissions before this address shall be
4980 * considered unreachable. If a value of zero
4981 * is present in this field then no changes are to
4982 * be made to this parameter.
4983 * spp_pathmtu - When Path MTU discovery is disabled the value
4984 * specified here will be the "fixed" path mtu.
4985 * Note that if the spp_address field is empty
4986 * then all associations on this address will
4987 * have this fixed path mtu set upon them.
4989 * spp_sackdelay - When delayed sack is enabled, this value specifies
4990 * the number of milliseconds that sacks will be delayed
4991 * for. This value will apply to all addresses of an
4992 * association if the spp_address field is empty. Note
4993 * also, that if delayed sack is enabled and this
4994 * value is set to 0, no change is made to the last
4995 * recorded delayed sack timer value.
4997 * spp_flags - These flags are used to control various features
4998 * on an association. The flag field may contain
4999 * zero or more of the following options.
5001 * SPP_HB_ENABLE - Enable heartbeats on the
5002 * specified address. Note that if the address
5003 * field is empty all addresses for the association
5004 * have heartbeats enabled upon them.
5006 * SPP_HB_DISABLE - Disable heartbeats on the
5007 * speicifed address. Note that if the address
5008 * field is empty all addresses for the association
5009 * will have their heartbeats disabled. Note also
5010 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
5011 * mutually exclusive, only one of these two should
5012 * be specified. Enabling both fields will have
5013 * undetermined results.
5015 * SPP_HB_DEMAND - Request a user initiated heartbeat
5016 * to be made immediately.
5018 * SPP_PMTUD_ENABLE - This field will enable PMTU
5019 * discovery upon the specified address. Note that
5020 * if the address feild is empty then all addresses
5021 * on the association are effected.
5023 * SPP_PMTUD_DISABLE - This field will disable PMTU
5024 * discovery upon the specified address. Note that
5025 * if the address feild is empty then all addresses
5026 * on the association are effected. Not also that
5027 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
5028 * exclusive. Enabling both will have undetermined
5031 * SPP_SACKDELAY_ENABLE - Setting this flag turns
5032 * on delayed sack. The time specified in spp_sackdelay
5033 * is used to specify the sack delay for this address. Note
5034 * that if spp_address is empty then all addresses will
5035 * enable delayed sack and take on the sack delay
5036 * value specified in spp_sackdelay.
5037 * SPP_SACKDELAY_DISABLE - Setting this flag turns
5038 * off delayed sack. If the spp_address field is blank then
5039 * delayed sack is disabled for the entire association. Note
5040 * also that this field is mutually exclusive to
5041 * SPP_SACKDELAY_ENABLE, setting both will have undefined
5044 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
5045 char __user *optval, int __user *optlen)
5047 struct sctp_paddrparams params;
5048 struct sctp_transport *trans = NULL;
5049 struct sctp_association *asoc = NULL;
5050 struct sctp_sock *sp = sctp_sk(sk);
5052 if (len < sizeof(struct sctp_paddrparams))
5054 len = sizeof(struct sctp_paddrparams);
5055 if (copy_from_user(¶ms, optval, len))
5058 /* If an address other than INADDR_ANY is specified, and
5059 * no transport is found, then the request is invalid.
5061 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
5062 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
5063 params.spp_assoc_id);
5065 pr_debug("%s: failed no transport\n", __func__);
5070 /* Get association, if assoc_id != 0 and the socket is a one
5071 * to many style socket, and an association was not found, then
5072 * the id was invalid.
5074 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
5075 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
5076 pr_debug("%s: failed no association\n", __func__);
5081 /* Fetch transport values. */
5082 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
5083 params.spp_pathmtu = trans->pathmtu;
5084 params.spp_pathmaxrxt = trans->pathmaxrxt;
5085 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
5087 /*draft-11 doesn't say what to return in spp_flags*/
5088 params.spp_flags = trans->param_flags;
5090 /* Fetch association values. */
5091 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
5092 params.spp_pathmtu = asoc->pathmtu;
5093 params.spp_pathmaxrxt = asoc->pathmaxrxt;
5094 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
5096 /*draft-11 doesn't say what to return in spp_flags*/
5097 params.spp_flags = asoc->param_flags;
5099 /* Fetch socket values. */
5100 params.spp_hbinterval = sp->hbinterval;
5101 params.spp_pathmtu = sp->pathmtu;
5102 params.spp_sackdelay = sp->sackdelay;
5103 params.spp_pathmaxrxt = sp->pathmaxrxt;
5105 /*draft-11 doesn't say what to return in spp_flags*/
5106 params.spp_flags = sp->param_flags;
5109 if (copy_to_user(optval, ¶ms, len))
5112 if (put_user(len, optlen))
5119 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
5121 * This option will effect the way delayed acks are performed. This
5122 * option allows you to get or set the delayed ack time, in
5123 * milliseconds. It also allows changing the delayed ack frequency.
5124 * Changing the frequency to 1 disables the delayed sack algorithm. If
5125 * the assoc_id is 0, then this sets or gets the endpoints default
5126 * values. If the assoc_id field is non-zero, then the set or get
5127 * effects the specified association for the one to many model (the
5128 * assoc_id field is ignored by the one to one model). Note that if
5129 * sack_delay or sack_freq are 0 when setting this option, then the
5130 * current values will remain unchanged.
5132 * struct sctp_sack_info {
5133 * sctp_assoc_t sack_assoc_id;
5134 * uint32_t sack_delay;
5135 * uint32_t sack_freq;
5138 * sack_assoc_id - This parameter, indicates which association the user
5139 * is performing an action upon. Note that if this field's value is
5140 * zero then the endpoints default value is changed (effecting future
5141 * associations only).
5143 * sack_delay - This parameter contains the number of milliseconds that
5144 * the user is requesting the delayed ACK timer be set to. Note that
5145 * this value is defined in the standard to be between 200 and 500
5148 * sack_freq - This parameter contains the number of packets that must
5149 * be received before a sack is sent without waiting for the delay
5150 * timer to expire. The default value for this is 2, setting this
5151 * value to 1 will disable the delayed sack algorithm.
5153 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
5154 char __user *optval,
5157 struct sctp_sack_info params;
5158 struct sctp_association *asoc = NULL;
5159 struct sctp_sock *sp = sctp_sk(sk);
5161 if (len >= sizeof(struct sctp_sack_info)) {
5162 len = sizeof(struct sctp_sack_info);
5164 if (copy_from_user(¶ms, optval, len))
5166 } else if (len == sizeof(struct sctp_assoc_value)) {
5167 pr_warn_ratelimited(DEPRECATED
5169 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
5170 "Use struct sctp_sack_info instead\n",
5171 current->comm, task_pid_nr(current));
5172 if (copy_from_user(¶ms, optval, len))
5177 /* Get association, if sack_assoc_id != 0 and the socket is a one
5178 * to many style socket, and an association was not found, then
5179 * the id was invalid.
5181 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
5182 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
5186 /* Fetch association values. */
5187 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
5188 params.sack_delay = jiffies_to_msecs(
5190 params.sack_freq = asoc->sackfreq;
5193 params.sack_delay = 0;
5194 params.sack_freq = 1;
5197 /* Fetch socket values. */
5198 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
5199 params.sack_delay = sp->sackdelay;
5200 params.sack_freq = sp->sackfreq;
5202 params.sack_delay = 0;
5203 params.sack_freq = 1;
5207 if (copy_to_user(optval, ¶ms, len))
5210 if (put_user(len, optlen))
5216 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5218 * Applications can specify protocol parameters for the default association
5219 * initialization. The option name argument to setsockopt() and getsockopt()
5222 * Setting initialization parameters is effective only on an unconnected
5223 * socket (for UDP-style sockets only future associations are effected
5224 * by the change). With TCP-style sockets, this option is inherited by
5225 * sockets derived from a listener socket.
5227 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
5229 if (len < sizeof(struct sctp_initmsg))
5231 len = sizeof(struct sctp_initmsg);
5232 if (put_user(len, optlen))
5234 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
5240 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
5241 char __user *optval, int __user *optlen)
5243 struct sctp_association *asoc;
5245 struct sctp_getaddrs getaddrs;
5246 struct sctp_transport *from;
5248 union sctp_addr temp;
5249 struct sctp_sock *sp = sctp_sk(sk);
5254 if (len < sizeof(struct sctp_getaddrs))
5257 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5260 /* For UDP-style sockets, id specifies the association to query. */
5261 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5265 to = optval + offsetof(struct sctp_getaddrs, addrs);
5266 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5268 list_for_each_entry(from, &asoc->peer.transport_addr_list,
5270 memcpy(&temp, &from->ipaddr, sizeof(temp));
5271 addrlen = sctp_get_pf_specific(sk->sk_family)
5272 ->addr_to_user(sp, &temp);
5273 if (space_left < addrlen)
5275 if (copy_to_user(to, &temp, addrlen))
5279 space_left -= addrlen;
5282 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
5284 bytes_copied = ((char __user *)to) - optval;
5285 if (put_user(bytes_copied, optlen))
5291 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
5292 size_t space_left, int *bytes_copied)
5294 struct sctp_sockaddr_entry *addr;
5295 union sctp_addr temp;
5298 struct net *net = sock_net(sk);
5301 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
5305 if ((PF_INET == sk->sk_family) &&
5306 (AF_INET6 == addr->a.sa.sa_family))
5308 if ((PF_INET6 == sk->sk_family) &&
5309 inet_v6_ipv6only(sk) &&
5310 (AF_INET == addr->a.sa.sa_family))
5312 memcpy(&temp, &addr->a, sizeof(temp));
5313 if (!temp.v4.sin_port)
5314 temp.v4.sin_port = htons(port);
5316 addrlen = sctp_get_pf_specific(sk->sk_family)
5317 ->addr_to_user(sctp_sk(sk), &temp);
5319 if (space_left < addrlen) {
5323 memcpy(to, &temp, addrlen);
5327 space_left -= addrlen;
5328 *bytes_copied += addrlen;
5336 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
5337 char __user *optval, int __user *optlen)
5339 struct sctp_bind_addr *bp;
5340 struct sctp_association *asoc;
5342 struct sctp_getaddrs getaddrs;
5343 struct sctp_sockaddr_entry *addr;
5345 union sctp_addr temp;
5346 struct sctp_sock *sp = sctp_sk(sk);
5350 int bytes_copied = 0;
5354 if (len < sizeof(struct sctp_getaddrs))
5357 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5361 * For UDP-style sockets, id specifies the association to query.
5362 * If the id field is set to the value '0' then the locally bound
5363 * addresses are returned without regard to any particular
5366 if (0 == getaddrs.assoc_id) {
5367 bp = &sctp_sk(sk)->ep->base.bind_addr;
5369 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5372 bp = &asoc->base.bind_addr;
5375 to = optval + offsetof(struct sctp_getaddrs, addrs);
5376 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5378 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN);
5382 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5383 * addresses from the global local address list.
5385 if (sctp_list_single_entry(&bp->address_list)) {
5386 addr = list_entry(bp->address_list.next,
5387 struct sctp_sockaddr_entry, list);
5388 if (sctp_is_any(sk, &addr->a)) {
5389 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
5390 space_left, &bytes_copied);
5400 /* Protection on the bound address list is not needed since
5401 * in the socket option context we hold a socket lock and
5402 * thus the bound address list can't change.
5404 list_for_each_entry(addr, &bp->address_list, list) {
5405 memcpy(&temp, &addr->a, sizeof(temp));
5406 addrlen = sctp_get_pf_specific(sk->sk_family)
5407 ->addr_to_user(sp, &temp);
5408 if (space_left < addrlen) {
5409 err = -ENOMEM; /*fixme: right error?*/
5412 memcpy(buf, &temp, addrlen);
5414 bytes_copied += addrlen;
5416 space_left -= addrlen;
5420 if (copy_to_user(to, addrs, bytes_copied)) {
5424 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
5428 if (put_user(bytes_copied, optlen))
5435 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
5437 * Requests that the local SCTP stack use the enclosed peer address as
5438 * the association primary. The enclosed address must be one of the
5439 * association peer's addresses.
5441 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
5442 char __user *optval, int __user *optlen)
5444 struct sctp_prim prim;
5445 struct sctp_association *asoc;
5446 struct sctp_sock *sp = sctp_sk(sk);
5448 if (len < sizeof(struct sctp_prim))
5451 len = sizeof(struct sctp_prim);
5453 if (copy_from_user(&prim, optval, len))
5456 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
5460 if (!asoc->peer.primary_path)
5463 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
5464 asoc->peer.primary_path->af_specific->sockaddr_len);
5466 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
5467 (union sctp_addr *)&prim.ssp_addr);
5469 if (put_user(len, optlen))
5471 if (copy_to_user(optval, &prim, len))
5478 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5480 * Requests that the local endpoint set the specified Adaptation Layer
5481 * Indication parameter for all future INIT and INIT-ACK exchanges.
5483 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
5484 char __user *optval, int __user *optlen)
5486 struct sctp_setadaptation adaptation;
5488 if (len < sizeof(struct sctp_setadaptation))
5491 len = sizeof(struct sctp_setadaptation);
5493 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
5495 if (put_user(len, optlen))
5497 if (copy_to_user(optval, &adaptation, len))
5505 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5507 * Applications that wish to use the sendto() system call may wish to
5508 * specify a default set of parameters that would normally be supplied
5509 * through the inclusion of ancillary data. This socket option allows
5510 * such an application to set the default sctp_sndrcvinfo structure.
5513 * The application that wishes to use this socket option simply passes
5514 * in to this call the sctp_sndrcvinfo structure defined in Section
5515 * 5.2.2) The input parameters accepted by this call include
5516 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5517 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5518 * to this call if the caller is using the UDP model.
5520 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5522 static int sctp_getsockopt_default_send_param(struct sock *sk,
5523 int len, char __user *optval,
5526 struct sctp_sock *sp = sctp_sk(sk);
5527 struct sctp_association *asoc;
5528 struct sctp_sndrcvinfo info;
5530 if (len < sizeof(info))
5535 if (copy_from_user(&info, optval, len))
5538 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
5539 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
5542 info.sinfo_stream = asoc->default_stream;
5543 info.sinfo_flags = asoc->default_flags;
5544 info.sinfo_ppid = asoc->default_ppid;
5545 info.sinfo_context = asoc->default_context;
5546 info.sinfo_timetolive = asoc->default_timetolive;
5548 info.sinfo_stream = sp->default_stream;
5549 info.sinfo_flags = sp->default_flags;
5550 info.sinfo_ppid = sp->default_ppid;
5551 info.sinfo_context = sp->default_context;
5552 info.sinfo_timetolive = sp->default_timetolive;
5555 if (put_user(len, optlen))
5557 if (copy_to_user(optval, &info, len))
5563 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5564 * (SCTP_DEFAULT_SNDINFO)
5566 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
5567 char __user *optval,
5570 struct sctp_sock *sp = sctp_sk(sk);
5571 struct sctp_association *asoc;
5572 struct sctp_sndinfo info;
5574 if (len < sizeof(info))
5579 if (copy_from_user(&info, optval, len))
5582 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
5583 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
5586 info.snd_sid = asoc->default_stream;
5587 info.snd_flags = asoc->default_flags;
5588 info.snd_ppid = asoc->default_ppid;
5589 info.snd_context = asoc->default_context;
5591 info.snd_sid = sp->default_stream;
5592 info.snd_flags = sp->default_flags;
5593 info.snd_ppid = sp->default_ppid;
5594 info.snd_context = sp->default_context;
5597 if (put_user(len, optlen))
5599 if (copy_to_user(optval, &info, len))
5607 * 7.1.5 SCTP_NODELAY
5609 * Turn on/off any Nagle-like algorithm. This means that packets are
5610 * generally sent as soon as possible and no unnecessary delays are
5611 * introduced, at the cost of more packets in the network. Expects an
5612 * integer boolean flag.
5615 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5616 char __user *optval, int __user *optlen)
5620 if (len < sizeof(int))
5624 val = (sctp_sk(sk)->nodelay == 1);
5625 if (put_user(len, optlen))
5627 if (copy_to_user(optval, &val, len))
5634 * 7.1.1 SCTP_RTOINFO
5636 * The protocol parameters used to initialize and bound retransmission
5637 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5638 * and modify these parameters.
5639 * All parameters are time values, in milliseconds. A value of 0, when
5640 * modifying the parameters, indicates that the current value should not
5644 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5645 char __user *optval,
5646 int __user *optlen) {
5647 struct sctp_rtoinfo rtoinfo;
5648 struct sctp_association *asoc;
5650 if (len < sizeof (struct sctp_rtoinfo))
5653 len = sizeof(struct sctp_rtoinfo);
5655 if (copy_from_user(&rtoinfo, optval, len))
5658 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5660 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5663 /* Values corresponding to the specific association. */
5665 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5666 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5667 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5669 /* Values corresponding to the endpoint. */
5670 struct sctp_sock *sp = sctp_sk(sk);
5672 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5673 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5674 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5677 if (put_user(len, optlen))
5680 if (copy_to_user(optval, &rtoinfo, len))
5688 * 7.1.2 SCTP_ASSOCINFO
5690 * This option is used to tune the maximum retransmission attempts
5691 * of the association.
5692 * Returns an error if the new association retransmission value is
5693 * greater than the sum of the retransmission value of the peer.
5694 * See [SCTP] for more information.
5697 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5698 char __user *optval,
5702 struct sctp_assocparams assocparams;
5703 struct sctp_association *asoc;
5704 struct list_head *pos;
5707 if (len < sizeof (struct sctp_assocparams))
5710 len = sizeof(struct sctp_assocparams);
5712 if (copy_from_user(&assocparams, optval, len))
5715 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5717 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5720 /* Values correspoinding to the specific association */
5722 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5723 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5724 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5725 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5727 list_for_each(pos, &asoc->peer.transport_addr_list) {
5731 assocparams.sasoc_number_peer_destinations = cnt;
5733 /* Values corresponding to the endpoint */
5734 struct sctp_sock *sp = sctp_sk(sk);
5736 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5737 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5738 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5739 assocparams.sasoc_cookie_life =
5740 sp->assocparams.sasoc_cookie_life;
5741 assocparams.sasoc_number_peer_destinations =
5743 sasoc_number_peer_destinations;
5746 if (put_user(len, optlen))
5749 if (copy_to_user(optval, &assocparams, len))
5756 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5758 * This socket option is a boolean flag which turns on or off mapped V4
5759 * addresses. If this option is turned on and the socket is type
5760 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5761 * If this option is turned off, then no mapping will be done of V4
5762 * addresses and a user will receive both PF_INET6 and PF_INET type
5763 * addresses on the socket.
5765 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5766 char __user *optval, int __user *optlen)
5769 struct sctp_sock *sp = sctp_sk(sk);
5771 if (len < sizeof(int))
5776 if (put_user(len, optlen))
5778 if (copy_to_user(optval, &val, len))
5785 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5786 * (chapter and verse is quoted at sctp_setsockopt_context())
5788 static int sctp_getsockopt_context(struct sock *sk, int len,
5789 char __user *optval, int __user *optlen)
5791 struct sctp_assoc_value params;
5792 struct sctp_sock *sp;
5793 struct sctp_association *asoc;
5795 if (len < sizeof(struct sctp_assoc_value))
5798 len = sizeof(struct sctp_assoc_value);
5800 if (copy_from_user(¶ms, optval, len))
5805 if (params.assoc_id != 0) {
5806 asoc = sctp_id2assoc(sk, params.assoc_id);
5809 params.assoc_value = asoc->default_rcv_context;
5811 params.assoc_value = sp->default_rcv_context;
5814 if (put_user(len, optlen))
5816 if (copy_to_user(optval, ¶ms, len))
5823 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5824 * This option will get or set the maximum size to put in any outgoing
5825 * SCTP DATA chunk. If a message is larger than this size it will be
5826 * fragmented by SCTP into the specified size. Note that the underlying
5827 * SCTP implementation may fragment into smaller sized chunks when the
5828 * PMTU of the underlying association is smaller than the value set by
5829 * the user. The default value for this option is '0' which indicates
5830 * the user is NOT limiting fragmentation and only the PMTU will effect
5831 * SCTP's choice of DATA chunk size. Note also that values set larger
5832 * than the maximum size of an IP datagram will effectively let SCTP
5833 * control fragmentation (i.e. the same as setting this option to 0).
5835 * The following structure is used to access and modify this parameter:
5837 * struct sctp_assoc_value {
5838 * sctp_assoc_t assoc_id;
5839 * uint32_t assoc_value;
5842 * assoc_id: This parameter is ignored for one-to-one style sockets.
5843 * For one-to-many style sockets this parameter indicates which
5844 * association the user is performing an action upon. Note that if
5845 * this field's value is zero then the endpoints default value is
5846 * changed (effecting future associations only).
5847 * assoc_value: This parameter specifies the maximum size in bytes.
5849 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5850 char __user *optval, int __user *optlen)
5852 struct sctp_assoc_value params;
5853 struct sctp_association *asoc;
5855 if (len == sizeof(int)) {
5856 pr_warn_ratelimited(DEPRECATED
5858 "Use of int in maxseg socket option.\n"
5859 "Use struct sctp_assoc_value instead\n",
5860 current->comm, task_pid_nr(current));
5861 params.assoc_id = 0;
5862 } else if (len >= sizeof(struct sctp_assoc_value)) {
5863 len = sizeof(struct sctp_assoc_value);
5864 if (copy_from_user(¶ms, optval, sizeof(params)))
5869 asoc = sctp_id2assoc(sk, params.assoc_id);
5870 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5874 params.assoc_value = asoc->frag_point;
5876 params.assoc_value = sctp_sk(sk)->user_frag;
5878 if (put_user(len, optlen))
5880 if (len == sizeof(int)) {
5881 if (copy_to_user(optval, ¶ms.assoc_value, len))
5884 if (copy_to_user(optval, ¶ms, len))
5892 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5893 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5895 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5896 char __user *optval, int __user *optlen)
5900 if (len < sizeof(int))
5905 val = sctp_sk(sk)->frag_interleave;
5906 if (put_user(len, optlen))
5908 if (copy_to_user(optval, &val, len))
5915 * 7.1.25. Set or Get the sctp partial delivery point
5916 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5918 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5919 char __user *optval,
5924 if (len < sizeof(u32))
5929 val = sctp_sk(sk)->pd_point;
5930 if (put_user(len, optlen))
5932 if (copy_to_user(optval, &val, len))
5939 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5940 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5942 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5943 char __user *optval,
5946 struct sctp_assoc_value params;
5947 struct sctp_sock *sp;
5948 struct sctp_association *asoc;
5950 if (len == sizeof(int)) {
5951 pr_warn_ratelimited(DEPRECATED
5953 "Use of int in max_burst socket option.\n"
5954 "Use struct sctp_assoc_value instead\n",
5955 current->comm, task_pid_nr(current));
5956 params.assoc_id = 0;
5957 } else if (len >= sizeof(struct sctp_assoc_value)) {
5958 len = sizeof(struct sctp_assoc_value);
5959 if (copy_from_user(¶ms, optval, len))
5966 if (params.assoc_id != 0) {
5967 asoc = sctp_id2assoc(sk, params.assoc_id);
5970 params.assoc_value = asoc->max_burst;
5972 params.assoc_value = sp->max_burst;
5974 if (len == sizeof(int)) {
5975 if (copy_to_user(optval, ¶ms.assoc_value, len))
5978 if (copy_to_user(optval, ¶ms, len))
5986 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5987 char __user *optval, int __user *optlen)
5989 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5990 struct sctp_hmacalgo __user *p = (void __user *)optval;
5991 struct sctp_hmac_algo_param *hmacs;
5996 if (!ep->auth_enable)
5999 hmacs = ep->auth_hmacs_list;
6000 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
6002 if (len < sizeof(struct sctp_hmacalgo) + data_len)
6005 len = sizeof(struct sctp_hmacalgo) + data_len;
6006 num_idents = data_len / sizeof(u16);
6008 if (put_user(len, optlen))
6010 if (put_user(num_idents, &p->shmac_num_idents))
6012 for (i = 0; i < num_idents; i++) {
6013 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
6015 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
6021 static int sctp_getsockopt_active_key(struct sock *sk, int len,
6022 char __user *optval, int __user *optlen)
6024 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6025 struct sctp_authkeyid val;
6026 struct sctp_association *asoc;
6028 if (!ep->auth_enable)
6031 if (len < sizeof(struct sctp_authkeyid))
6033 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
6036 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
6037 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
6041 val.scact_keynumber = asoc->active_key_id;
6043 val.scact_keynumber = ep->active_key_id;
6045 len = sizeof(struct sctp_authkeyid);
6046 if (put_user(len, optlen))
6048 if (copy_to_user(optval, &val, len))
6054 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
6055 char __user *optval, int __user *optlen)
6057 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6058 struct sctp_authchunks __user *p = (void __user *)optval;
6059 struct sctp_authchunks val;
6060 struct sctp_association *asoc;
6061 struct sctp_chunks_param *ch;
6065 if (!ep->auth_enable)
6068 if (len < sizeof(struct sctp_authchunks))
6071 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
6074 to = p->gauth_chunks;
6075 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6079 ch = asoc->peer.peer_chunks;
6083 /* See if the user provided enough room for all the data */
6084 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
6085 if (len < num_chunks)
6088 if (copy_to_user(to, ch->chunks, num_chunks))
6091 len = sizeof(struct sctp_authchunks) + num_chunks;
6092 if (put_user(len, optlen))
6094 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6099 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
6100 char __user *optval, int __user *optlen)
6102 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6103 struct sctp_authchunks __user *p = (void __user *)optval;
6104 struct sctp_authchunks val;
6105 struct sctp_association *asoc;
6106 struct sctp_chunks_param *ch;
6110 if (!ep->auth_enable)
6113 if (len < sizeof(struct sctp_authchunks))
6116 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
6119 to = p->gauth_chunks;
6120 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6121 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
6125 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
6127 ch = ep->auth_chunk_list;
6132 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
6133 if (len < sizeof(struct sctp_authchunks) + num_chunks)
6136 if (copy_to_user(to, ch->chunks, num_chunks))
6139 len = sizeof(struct sctp_authchunks) + num_chunks;
6140 if (put_user(len, optlen))
6142 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6149 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
6150 * This option gets the current number of associations that are attached
6151 * to a one-to-many style socket. The option value is an uint32_t.
6153 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
6154 char __user *optval, int __user *optlen)
6156 struct sctp_sock *sp = sctp_sk(sk);
6157 struct sctp_association *asoc;
6160 if (sctp_style(sk, TCP))
6163 if (len < sizeof(u32))
6168 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6172 if (put_user(len, optlen))
6174 if (copy_to_user(optval, &val, len))
6181 * 8.1.23 SCTP_AUTO_ASCONF
6182 * See the corresponding setsockopt entry as description
6184 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
6185 char __user *optval, int __user *optlen)
6189 if (len < sizeof(int))
6193 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
6195 if (put_user(len, optlen))
6197 if (copy_to_user(optval, &val, len))
6203 * 8.2.6. Get the Current Identifiers of Associations
6204 * (SCTP_GET_ASSOC_ID_LIST)
6206 * This option gets the current list of SCTP association identifiers of
6207 * the SCTP associations handled by a one-to-many style socket.
6209 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
6210 char __user *optval, int __user *optlen)
6212 struct sctp_sock *sp = sctp_sk(sk);
6213 struct sctp_association *asoc;
6214 struct sctp_assoc_ids *ids;
6217 if (sctp_style(sk, TCP))
6220 if (len < sizeof(struct sctp_assoc_ids))
6223 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6227 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
6230 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
6232 ids = kmalloc(len, GFP_USER | __GFP_NOWARN);
6236 ids->gaids_number_of_ids = num;
6238 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6239 ids->gaids_assoc_id[num++] = asoc->assoc_id;
6242 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
6252 * SCTP_PEER_ADDR_THLDS
6254 * This option allows us to fetch the partially failed threshold for one or all
6255 * transports in an association. See Section 6.1 of:
6256 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6258 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
6259 char __user *optval,
6263 struct sctp_paddrthlds val;
6264 struct sctp_transport *trans;
6265 struct sctp_association *asoc;
6267 if (len < sizeof(struct sctp_paddrthlds))
6269 len = sizeof(struct sctp_paddrthlds);
6270 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
6273 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
6274 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
6278 val.spt_pathpfthld = asoc->pf_retrans;
6279 val.spt_pathmaxrxt = asoc->pathmaxrxt;
6281 trans = sctp_addr_id2transport(sk, &val.spt_address,
6286 val.spt_pathmaxrxt = trans->pathmaxrxt;
6287 val.spt_pathpfthld = trans->pf_retrans;
6290 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
6297 * SCTP_GET_ASSOC_STATS
6299 * This option retrieves local per endpoint statistics. It is modeled
6300 * after OpenSolaris' implementation
6302 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
6303 char __user *optval,
6306 struct sctp_assoc_stats sas;
6307 struct sctp_association *asoc = NULL;
6309 /* User must provide at least the assoc id */
6310 if (len < sizeof(sctp_assoc_t))
6313 /* Allow the struct to grow and fill in as much as possible */
6314 len = min_t(size_t, len, sizeof(sas));
6316 if (copy_from_user(&sas, optval, len))
6319 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
6323 sas.sas_rtxchunks = asoc->stats.rtxchunks;
6324 sas.sas_gapcnt = asoc->stats.gapcnt;
6325 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
6326 sas.sas_osacks = asoc->stats.osacks;
6327 sas.sas_isacks = asoc->stats.isacks;
6328 sas.sas_octrlchunks = asoc->stats.octrlchunks;
6329 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
6330 sas.sas_oodchunks = asoc->stats.oodchunks;
6331 sas.sas_iodchunks = asoc->stats.iodchunks;
6332 sas.sas_ouodchunks = asoc->stats.ouodchunks;
6333 sas.sas_iuodchunks = asoc->stats.iuodchunks;
6334 sas.sas_idupchunks = asoc->stats.idupchunks;
6335 sas.sas_opackets = asoc->stats.opackets;
6336 sas.sas_ipackets = asoc->stats.ipackets;
6338 /* New high max rto observed, will return 0 if not a single
6339 * RTO update took place. obs_rto_ipaddr will be bogus
6342 sas.sas_maxrto = asoc->stats.max_obs_rto;
6343 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
6344 sizeof(struct sockaddr_storage));
6346 /* Mark beginning of a new observation period */
6347 asoc->stats.max_obs_rto = asoc->rto_min;
6349 if (put_user(len, optlen))
6352 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
6354 if (copy_to_user(optval, &sas, len))
6360 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
6361 char __user *optval,
6366 if (len < sizeof(int))
6370 if (sctp_sk(sk)->recvrcvinfo)
6372 if (put_user(len, optlen))
6374 if (copy_to_user(optval, &val, len))
6380 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
6381 char __user *optval,
6386 if (len < sizeof(int))
6390 if (sctp_sk(sk)->recvnxtinfo)
6392 if (put_user(len, optlen))
6394 if (copy_to_user(optval, &val, len))
6400 static int sctp_getsockopt_pr_supported(struct sock *sk, int len,
6401 char __user *optval,
6404 struct sctp_assoc_value params;
6405 struct sctp_association *asoc;
6406 int retval = -EFAULT;
6408 if (len < sizeof(params)) {
6413 len = sizeof(params);
6414 if (copy_from_user(¶ms, optval, len))
6417 asoc = sctp_id2assoc(sk, params.assoc_id);
6419 params.assoc_value = asoc->prsctp_enable;
6420 } else if (!params.assoc_id) {
6421 struct sctp_sock *sp = sctp_sk(sk);
6423 params.assoc_value = sp->ep->prsctp_enable;
6429 if (put_user(len, optlen))
6432 if (copy_to_user(optval, ¶ms, len))
6441 static int sctp_getsockopt_default_prinfo(struct sock *sk, int len,
6442 char __user *optval,
6445 struct sctp_default_prinfo info;
6446 struct sctp_association *asoc;
6447 int retval = -EFAULT;
6449 if (len < sizeof(info)) {
6455 if (copy_from_user(&info, optval, len))
6458 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
6460 info.pr_policy = SCTP_PR_POLICY(asoc->default_flags);
6461 info.pr_value = asoc->default_timetolive;
6462 } else if (!info.pr_assoc_id) {
6463 struct sctp_sock *sp = sctp_sk(sk);
6465 info.pr_policy = SCTP_PR_POLICY(sp->default_flags);
6466 info.pr_value = sp->default_timetolive;
6472 if (put_user(len, optlen))
6475 if (copy_to_user(optval, &info, len))
6484 static int sctp_getsockopt_pr_assocstatus(struct sock *sk, int len,
6485 char __user *optval,
6488 struct sctp_prstatus params;
6489 struct sctp_association *asoc;
6491 int retval = -EINVAL;
6493 if (len < sizeof(params))
6496 len = sizeof(params);
6497 if (copy_from_user(¶ms, optval, len)) {
6502 policy = params.sprstat_policy;
6503 if (policy & ~SCTP_PR_SCTP_MASK)
6506 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
6510 if (policy == SCTP_PR_SCTP_NONE) {
6511 params.sprstat_abandoned_unsent = 0;
6512 params.sprstat_abandoned_sent = 0;
6513 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
6514 params.sprstat_abandoned_unsent +=
6515 asoc->abandoned_unsent[policy];
6516 params.sprstat_abandoned_sent +=
6517 asoc->abandoned_sent[policy];
6520 params.sprstat_abandoned_unsent =
6521 asoc->abandoned_unsent[__SCTP_PR_INDEX(policy)];
6522 params.sprstat_abandoned_sent =
6523 asoc->abandoned_sent[__SCTP_PR_INDEX(policy)];
6526 if (put_user(len, optlen)) {
6531 if (copy_to_user(optval, ¶ms, len)) {
6542 static int sctp_getsockopt_enable_strreset(struct sock *sk, int len,
6543 char __user *optval,
6546 struct sctp_assoc_value params;
6547 struct sctp_association *asoc;
6548 int retval = -EFAULT;
6550 if (len < sizeof(params)) {
6555 len = sizeof(params);
6556 if (copy_from_user(¶ms, optval, len))
6559 asoc = sctp_id2assoc(sk, params.assoc_id);
6561 params.assoc_value = asoc->strreset_enable;
6562 } else if (!params.assoc_id) {
6563 struct sctp_sock *sp = sctp_sk(sk);
6565 params.assoc_value = sp->ep->strreset_enable;
6571 if (put_user(len, optlen))
6574 if (copy_to_user(optval, ¶ms, len))
6583 static int sctp_getsockopt(struct sock *sk, int level, int optname,
6584 char __user *optval, int __user *optlen)
6589 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
6591 /* I can hardly begin to describe how wrong this is. This is
6592 * so broken as to be worse than useless. The API draft
6593 * REALLY is NOT helpful here... I am not convinced that the
6594 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
6595 * are at all well-founded.
6597 if (level != SOL_SCTP) {
6598 struct sctp_af *af = sctp_sk(sk)->pf->af;
6600 retval = af->getsockopt(sk, level, optname, optval, optlen);
6604 if (get_user(len, optlen))
6614 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
6616 case SCTP_DISABLE_FRAGMENTS:
6617 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
6621 retval = sctp_getsockopt_events(sk, len, optval, optlen);
6623 case SCTP_AUTOCLOSE:
6624 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
6626 case SCTP_SOCKOPT_PEELOFF:
6627 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
6629 case SCTP_PEER_ADDR_PARAMS:
6630 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
6633 case SCTP_DELAYED_SACK:
6634 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
6638 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
6640 case SCTP_GET_PEER_ADDRS:
6641 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
6644 case SCTP_GET_LOCAL_ADDRS:
6645 retval = sctp_getsockopt_local_addrs(sk, len, optval,
6648 case SCTP_SOCKOPT_CONNECTX3:
6649 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
6651 case SCTP_DEFAULT_SEND_PARAM:
6652 retval = sctp_getsockopt_default_send_param(sk, len,
6655 case SCTP_DEFAULT_SNDINFO:
6656 retval = sctp_getsockopt_default_sndinfo(sk, len,
6659 case SCTP_PRIMARY_ADDR:
6660 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
6663 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
6666 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
6668 case SCTP_ASSOCINFO:
6669 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
6671 case SCTP_I_WANT_MAPPED_V4_ADDR:
6672 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
6675 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
6677 case SCTP_GET_PEER_ADDR_INFO:
6678 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
6681 case SCTP_ADAPTATION_LAYER:
6682 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
6686 retval = sctp_getsockopt_context(sk, len, optval, optlen);
6688 case SCTP_FRAGMENT_INTERLEAVE:
6689 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
6692 case SCTP_PARTIAL_DELIVERY_POINT:
6693 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
6696 case SCTP_MAX_BURST:
6697 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
6700 case SCTP_AUTH_CHUNK:
6701 case SCTP_AUTH_DELETE_KEY:
6702 retval = -EOPNOTSUPP;
6704 case SCTP_HMAC_IDENT:
6705 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
6707 case SCTP_AUTH_ACTIVE_KEY:
6708 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
6710 case SCTP_PEER_AUTH_CHUNKS:
6711 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
6714 case SCTP_LOCAL_AUTH_CHUNKS:
6715 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
6718 case SCTP_GET_ASSOC_NUMBER:
6719 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
6721 case SCTP_GET_ASSOC_ID_LIST:
6722 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
6724 case SCTP_AUTO_ASCONF:
6725 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
6727 case SCTP_PEER_ADDR_THLDS:
6728 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
6730 case SCTP_GET_ASSOC_STATS:
6731 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
6733 case SCTP_RECVRCVINFO:
6734 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
6736 case SCTP_RECVNXTINFO:
6737 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
6739 case SCTP_PR_SUPPORTED:
6740 retval = sctp_getsockopt_pr_supported(sk, len, optval, optlen);
6742 case SCTP_DEFAULT_PRINFO:
6743 retval = sctp_getsockopt_default_prinfo(sk, len, optval,
6746 case SCTP_PR_ASSOC_STATUS:
6747 retval = sctp_getsockopt_pr_assocstatus(sk, len, optval,
6750 case SCTP_ENABLE_STREAM_RESET:
6751 retval = sctp_getsockopt_enable_strreset(sk, len, optval,
6755 retval = -ENOPROTOOPT;
6763 static int sctp_hash(struct sock *sk)
6769 static void sctp_unhash(struct sock *sk)
6774 /* Check if port is acceptable. Possibly find first available port.
6776 * The port hash table (contained in the 'global' SCTP protocol storage
6777 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
6778 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
6779 * list (the list number is the port number hashed out, so as you
6780 * would expect from a hash function, all the ports in a given list have
6781 * such a number that hashes out to the same list number; you were
6782 * expecting that, right?); so each list has a set of ports, with a
6783 * link to the socket (struct sock) that uses it, the port number and
6784 * a fastreuse flag (FIXME: NPI ipg).
6786 static struct sctp_bind_bucket *sctp_bucket_create(
6787 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
6789 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
6791 struct sctp_bind_hashbucket *head; /* hash list */
6792 struct sctp_bind_bucket *pp;
6793 unsigned short snum;
6796 snum = ntohs(addr->v4.sin_port);
6798 pr_debug("%s: begins, snum:%d\n", __func__, snum);
6803 /* Search for an available port. */
6804 int low, high, remaining, index;
6806 struct net *net = sock_net(sk);
6808 inet_get_local_port_range(net, &low, &high);
6809 remaining = (high - low) + 1;
6810 rover = prandom_u32() % remaining + low;
6814 if ((rover < low) || (rover > high))
6816 if (inet_is_local_reserved_port(net, rover))
6818 index = sctp_phashfn(sock_net(sk), rover);
6819 head = &sctp_port_hashtable[index];
6820 spin_lock(&head->lock);
6821 sctp_for_each_hentry(pp, &head->chain)
6822 if ((pp->port == rover) &&
6823 net_eq(sock_net(sk), pp->net))
6827 spin_unlock(&head->lock);
6828 } while (--remaining > 0);
6830 /* Exhausted local port range during search? */
6835 /* OK, here is the one we will use. HEAD (the port
6836 * hash table list entry) is non-NULL and we hold it's
6841 /* We are given an specific port number; we verify
6842 * that it is not being used. If it is used, we will
6843 * exahust the search in the hash list corresponding
6844 * to the port number (snum) - we detect that with the
6845 * port iterator, pp being NULL.
6847 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
6848 spin_lock(&head->lock);
6849 sctp_for_each_hentry(pp, &head->chain) {
6850 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
6857 if (!hlist_empty(&pp->owner)) {
6858 /* We had a port hash table hit - there is an
6859 * available port (pp != NULL) and it is being
6860 * used by other socket (pp->owner not empty); that other
6861 * socket is going to be sk2.
6863 int reuse = sk->sk_reuse;
6866 pr_debug("%s: found a possible match\n", __func__);
6868 if (pp->fastreuse && sk->sk_reuse &&
6869 sk->sk_state != SCTP_SS_LISTENING)
6872 /* Run through the list of sockets bound to the port
6873 * (pp->port) [via the pointers bind_next and
6874 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6875 * we get the endpoint they describe and run through
6876 * the endpoint's list of IP (v4 or v6) addresses,
6877 * comparing each of the addresses with the address of
6878 * the socket sk. If we find a match, then that means
6879 * that this port/socket (sk) combination are already
6882 sk_for_each_bound(sk2, &pp->owner) {
6883 struct sctp_endpoint *ep2;
6884 ep2 = sctp_sk(sk2)->ep;
6887 (reuse && sk2->sk_reuse &&
6888 sk2->sk_state != SCTP_SS_LISTENING))
6891 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
6892 sctp_sk(sk2), sctp_sk(sk))) {
6898 pr_debug("%s: found a match\n", __func__);
6901 /* If there was a hash table miss, create a new port. */
6903 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6906 /* In either case (hit or miss), make sure fastreuse is 1 only
6907 * if sk->sk_reuse is too (that is, if the caller requested
6908 * SO_REUSEADDR on this socket -sk-).
6910 if (hlist_empty(&pp->owner)) {
6911 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6915 } else if (pp->fastreuse &&
6916 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6919 /* We are set, so fill up all the data in the hash table
6920 * entry, tie the socket list information with the rest of the
6921 * sockets FIXME: Blurry, NPI (ipg).
6924 if (!sctp_sk(sk)->bind_hash) {
6925 inet_sk(sk)->inet_num = snum;
6926 sk_add_bind_node(sk, &pp->owner);
6927 sctp_sk(sk)->bind_hash = pp;
6932 spin_unlock(&head->lock);
6939 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6940 * port is requested.
6942 static int sctp_get_port(struct sock *sk, unsigned short snum)
6944 union sctp_addr addr;
6945 struct sctp_af *af = sctp_sk(sk)->pf->af;
6947 /* Set up a dummy address struct from the sk. */
6948 af->from_sk(&addr, sk);
6949 addr.v4.sin_port = htons(snum);
6951 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6952 return !!sctp_get_port_local(sk, &addr);
6956 * Move a socket to LISTENING state.
6958 static int sctp_listen_start(struct sock *sk, int backlog)
6960 struct sctp_sock *sp = sctp_sk(sk);
6961 struct sctp_endpoint *ep = sp->ep;
6962 struct crypto_shash *tfm = NULL;
6965 /* Allocate HMAC for generating cookie. */
6966 if (!sp->hmac && sp->sctp_hmac_alg) {
6967 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6968 tfm = crypto_alloc_shash(alg, 0, 0);
6970 net_info_ratelimited("failed to load transform for %s: %ld\n",
6971 sp->sctp_hmac_alg, PTR_ERR(tfm));
6974 sctp_sk(sk)->hmac = tfm;
6978 * If a bind() or sctp_bindx() is not called prior to a listen()
6979 * call that allows new associations to be accepted, the system
6980 * picks an ephemeral port and will choose an address set equivalent
6981 * to binding with a wildcard address.
6983 * This is not currently spelled out in the SCTP sockets
6984 * extensions draft, but follows the practice as seen in TCP
6988 sk->sk_state = SCTP_SS_LISTENING;
6989 if (!ep->base.bind_addr.port) {
6990 if (sctp_autobind(sk))
6993 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6994 sk->sk_state = SCTP_SS_CLOSED;
6999 sk->sk_max_ack_backlog = backlog;
7000 sctp_hash_endpoint(ep);
7005 * 4.1.3 / 5.1.3 listen()
7007 * By default, new associations are not accepted for UDP style sockets.
7008 * An application uses listen() to mark a socket as being able to
7009 * accept new associations.
7011 * On TCP style sockets, applications use listen() to ready the SCTP
7012 * endpoint for accepting inbound associations.
7014 * On both types of endpoints a backlog of '0' disables listening.
7016 * Move a socket to LISTENING state.
7018 int sctp_inet_listen(struct socket *sock, int backlog)
7020 struct sock *sk = sock->sk;
7021 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
7024 if (unlikely(backlog < 0))
7029 /* Peeled-off sockets are not allowed to listen(). */
7030 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
7033 if (sock->state != SS_UNCONNECTED)
7036 /* If backlog is zero, disable listening. */
7038 if (sctp_sstate(sk, CLOSED))
7042 sctp_unhash_endpoint(ep);
7043 sk->sk_state = SCTP_SS_CLOSED;
7045 sctp_sk(sk)->bind_hash->fastreuse = 1;
7049 /* If we are already listening, just update the backlog */
7050 if (sctp_sstate(sk, LISTENING))
7051 sk->sk_max_ack_backlog = backlog;
7053 err = sctp_listen_start(sk, backlog);
7065 * This function is done by modeling the current datagram_poll() and the
7066 * tcp_poll(). Note that, based on these implementations, we don't
7067 * lock the socket in this function, even though it seems that,
7068 * ideally, locking or some other mechanisms can be used to ensure
7069 * the integrity of the counters (sndbuf and wmem_alloc) used
7070 * in this place. We assume that we don't need locks either until proven
7073 * Another thing to note is that we include the Async I/O support
7074 * here, again, by modeling the current TCP/UDP code. We don't have
7075 * a good way to test with it yet.
7077 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
7079 struct sock *sk = sock->sk;
7080 struct sctp_sock *sp = sctp_sk(sk);
7083 poll_wait(file, sk_sleep(sk), wait);
7085 sock_rps_record_flow(sk);
7087 /* A TCP-style listening socket becomes readable when the accept queue
7090 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
7091 return (!list_empty(&sp->ep->asocs)) ?
7092 (POLLIN | POLLRDNORM) : 0;
7096 /* Is there any exceptional events? */
7097 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
7099 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
7100 if (sk->sk_shutdown & RCV_SHUTDOWN)
7101 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
7102 if (sk->sk_shutdown == SHUTDOWN_MASK)
7105 /* Is it readable? Reconsider this code with TCP-style support. */
7106 if (!skb_queue_empty(&sk->sk_receive_queue))
7107 mask |= POLLIN | POLLRDNORM;
7109 /* The association is either gone or not ready. */
7110 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
7113 /* Is it writable? */
7114 if (sctp_writeable(sk)) {
7115 mask |= POLLOUT | POLLWRNORM;
7117 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
7119 * Since the socket is not locked, the buffer
7120 * might be made available after the writeable check and
7121 * before the bit is set. This could cause a lost I/O
7122 * signal. tcp_poll() has a race breaker for this race
7123 * condition. Based on their implementation, we put
7124 * in the following code to cover it as well.
7126 if (sctp_writeable(sk))
7127 mask |= POLLOUT | POLLWRNORM;
7132 /********************************************************************
7133 * 2nd Level Abstractions
7134 ********************************************************************/
7136 static struct sctp_bind_bucket *sctp_bucket_create(
7137 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
7139 struct sctp_bind_bucket *pp;
7141 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
7143 SCTP_DBG_OBJCNT_INC(bind_bucket);
7146 INIT_HLIST_HEAD(&pp->owner);
7148 hlist_add_head(&pp->node, &head->chain);
7153 /* Caller must hold hashbucket lock for this tb with local BH disabled */
7154 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
7156 if (pp && hlist_empty(&pp->owner)) {
7157 __hlist_del(&pp->node);
7158 kmem_cache_free(sctp_bucket_cachep, pp);
7159 SCTP_DBG_OBJCNT_DEC(bind_bucket);
7163 /* Release this socket's reference to a local port. */
7164 static inline void __sctp_put_port(struct sock *sk)
7166 struct sctp_bind_hashbucket *head =
7167 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
7168 inet_sk(sk)->inet_num)];
7169 struct sctp_bind_bucket *pp;
7171 spin_lock(&head->lock);
7172 pp = sctp_sk(sk)->bind_hash;
7173 __sk_del_bind_node(sk);
7174 sctp_sk(sk)->bind_hash = NULL;
7175 inet_sk(sk)->inet_num = 0;
7176 sctp_bucket_destroy(pp);
7177 spin_unlock(&head->lock);
7180 void sctp_put_port(struct sock *sk)
7183 __sctp_put_port(sk);
7188 * The system picks an ephemeral port and choose an address set equivalent
7189 * to binding with a wildcard address.
7190 * One of those addresses will be the primary address for the association.
7191 * This automatically enables the multihoming capability of SCTP.
7193 static int sctp_autobind(struct sock *sk)
7195 union sctp_addr autoaddr;
7199 /* Initialize a local sockaddr structure to INADDR_ANY. */
7200 af = sctp_sk(sk)->pf->af;
7202 port = htons(inet_sk(sk)->inet_num);
7203 af->inaddr_any(&autoaddr, port);
7205 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
7208 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
7211 * 4.2 The cmsghdr Structure *
7213 * When ancillary data is sent or received, any number of ancillary data
7214 * objects can be specified by the msg_control and msg_controllen members of
7215 * the msghdr structure, because each object is preceded by
7216 * a cmsghdr structure defining the object's length (the cmsg_len member).
7217 * Historically Berkeley-derived implementations have passed only one object
7218 * at a time, but this API allows multiple objects to be
7219 * passed in a single call to sendmsg() or recvmsg(). The following example
7220 * shows two ancillary data objects in a control buffer.
7222 * |<--------------------------- msg_controllen -------------------------->|
7225 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
7227 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
7230 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
7232 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
7235 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7236 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
7238 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
7240 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7247 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
7249 struct cmsghdr *cmsg;
7250 struct msghdr *my_msg = (struct msghdr *)msg;
7252 for_each_cmsghdr(cmsg, my_msg) {
7253 if (!CMSG_OK(my_msg, cmsg))
7256 /* Should we parse this header or ignore? */
7257 if (cmsg->cmsg_level != IPPROTO_SCTP)
7260 /* Strictly check lengths following example in SCM code. */
7261 switch (cmsg->cmsg_type) {
7263 /* SCTP Socket API Extension
7264 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
7266 * This cmsghdr structure provides information for
7267 * initializing new SCTP associations with sendmsg().
7268 * The SCTP_INITMSG socket option uses this same data
7269 * structure. This structure is not used for
7272 * cmsg_level cmsg_type cmsg_data[]
7273 * ------------ ------------ ----------------------
7274 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
7276 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
7279 cmsgs->init = CMSG_DATA(cmsg);
7283 /* SCTP Socket API Extension
7284 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
7286 * This cmsghdr structure specifies SCTP options for
7287 * sendmsg() and describes SCTP header information
7288 * about a received message through recvmsg().
7290 * cmsg_level cmsg_type cmsg_data[]
7291 * ------------ ------------ ----------------------
7292 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
7294 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
7297 cmsgs->srinfo = CMSG_DATA(cmsg);
7299 if (cmsgs->srinfo->sinfo_flags &
7300 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7301 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7302 SCTP_ABORT | SCTP_EOF))
7307 /* SCTP Socket API Extension
7308 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
7310 * This cmsghdr structure specifies SCTP options for
7311 * sendmsg(). This structure and SCTP_RCVINFO replaces
7312 * SCTP_SNDRCV which has been deprecated.
7314 * cmsg_level cmsg_type cmsg_data[]
7315 * ------------ ------------ ---------------------
7316 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
7318 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
7321 cmsgs->sinfo = CMSG_DATA(cmsg);
7323 if (cmsgs->sinfo->snd_flags &
7324 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7325 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7326 SCTP_ABORT | SCTP_EOF))
7338 * Wait for a packet..
7339 * Note: This function is the same function as in core/datagram.c
7340 * with a few modifications to make lksctp work.
7342 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
7347 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7349 /* Socket errors? */
7350 error = sock_error(sk);
7354 if (!skb_queue_empty(&sk->sk_receive_queue))
7357 /* Socket shut down? */
7358 if (sk->sk_shutdown & RCV_SHUTDOWN)
7361 /* Sequenced packets can come disconnected. If so we report the
7366 /* Is there a good reason to think that we may receive some data? */
7367 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
7370 /* Handle signals. */
7371 if (signal_pending(current))
7374 /* Let another process have a go. Since we are going to sleep
7375 * anyway. Note: This may cause odd behaviors if the message
7376 * does not fit in the user's buffer, but this seems to be the
7377 * only way to honor MSG_DONTWAIT realistically.
7380 *timeo_p = schedule_timeout(*timeo_p);
7384 finish_wait(sk_sleep(sk), &wait);
7388 error = sock_intr_errno(*timeo_p);
7391 finish_wait(sk_sleep(sk), &wait);
7396 /* Receive a datagram.
7397 * Note: This is pretty much the same routine as in core/datagram.c
7398 * with a few changes to make lksctp work.
7400 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
7401 int noblock, int *err)
7404 struct sk_buff *skb;
7407 timeo = sock_rcvtimeo(sk, noblock);
7409 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
7410 MAX_SCHEDULE_TIMEOUT);
7413 /* Again only user level code calls this function,
7414 * so nothing interrupt level
7415 * will suddenly eat the receive_queue.
7417 * Look at current nfs client by the way...
7418 * However, this function was correct in any case. 8)
7420 if (flags & MSG_PEEK) {
7421 skb = skb_peek(&sk->sk_receive_queue);
7423 atomic_inc(&skb->users);
7425 skb = __skb_dequeue(&sk->sk_receive_queue);
7431 /* Caller is allowed not to check sk->sk_err before calling. */
7432 error = sock_error(sk);
7436 if (sk->sk_shutdown & RCV_SHUTDOWN)
7439 if (sk_can_busy_loop(sk) &&
7440 sk_busy_loop(sk, noblock))
7443 /* User doesn't want to wait. */
7447 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
7456 /* If sndbuf has changed, wake up per association sndbuf waiters. */
7457 static void __sctp_write_space(struct sctp_association *asoc)
7459 struct sock *sk = asoc->base.sk;
7461 if (sctp_wspace(asoc) <= 0)
7464 if (waitqueue_active(&asoc->wait))
7465 wake_up_interruptible(&asoc->wait);
7467 if (sctp_writeable(sk)) {
7468 struct socket_wq *wq;
7471 wq = rcu_dereference(sk->sk_wq);
7473 if (waitqueue_active(&wq->wait))
7474 wake_up_interruptible(&wq->wait);
7476 /* Note that we try to include the Async I/O support
7477 * here by modeling from the current TCP/UDP code.
7478 * We have not tested with it yet.
7480 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
7481 sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT);
7487 static void sctp_wake_up_waiters(struct sock *sk,
7488 struct sctp_association *asoc)
7490 struct sctp_association *tmp = asoc;
7492 /* We do accounting for the sndbuf space per association,
7493 * so we only need to wake our own association.
7495 if (asoc->ep->sndbuf_policy)
7496 return __sctp_write_space(asoc);
7498 /* If association goes down and is just flushing its
7499 * outq, then just normally notify others.
7501 if (asoc->base.dead)
7502 return sctp_write_space(sk);
7504 /* Accounting for the sndbuf space is per socket, so we
7505 * need to wake up others, try to be fair and in case of
7506 * other associations, let them have a go first instead
7507 * of just doing a sctp_write_space() call.
7509 * Note that we reach sctp_wake_up_waiters() only when
7510 * associations free up queued chunks, thus we are under
7511 * lock and the list of associations on a socket is
7512 * guaranteed not to change.
7514 for (tmp = list_next_entry(tmp, asocs); 1;
7515 tmp = list_next_entry(tmp, asocs)) {
7516 /* Manually skip the head element. */
7517 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
7519 /* Wake up association. */
7520 __sctp_write_space(tmp);
7521 /* We've reached the end. */
7527 /* Do accounting for the sndbuf space.
7528 * Decrement the used sndbuf space of the corresponding association by the
7529 * data size which was just transmitted(freed).
7531 static void sctp_wfree(struct sk_buff *skb)
7533 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
7534 struct sctp_association *asoc = chunk->asoc;
7535 struct sock *sk = asoc->base.sk;
7537 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
7538 sizeof(struct sk_buff) +
7539 sizeof(struct sctp_chunk);
7541 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
7544 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
7546 sk->sk_wmem_queued -= skb->truesize;
7547 sk_mem_uncharge(sk, skb->truesize);
7550 sctp_wake_up_waiters(sk, asoc);
7552 sctp_association_put(asoc);
7555 /* Do accounting for the receive space on the socket.
7556 * Accounting for the association is done in ulpevent.c
7557 * We set this as a destructor for the cloned data skbs so that
7558 * accounting is done at the correct time.
7560 void sctp_sock_rfree(struct sk_buff *skb)
7562 struct sock *sk = skb->sk;
7563 struct sctp_ulpevent *event = sctp_skb2event(skb);
7565 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
7568 * Mimic the behavior of sock_rfree
7570 sk_mem_uncharge(sk, event->rmem_len);
7574 /* Helper function to wait for space in the sndbuf. */
7575 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
7578 struct sock *sk = asoc->base.sk;
7580 long current_timeo = *timeo_p;
7583 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
7586 /* Increment the association's refcnt. */
7587 sctp_association_hold(asoc);
7589 /* Wait on the association specific sndbuf space. */
7591 prepare_to_wait_exclusive(&asoc->wait, &wait,
7592 TASK_INTERRUPTIBLE);
7595 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7598 if (signal_pending(current))
7599 goto do_interrupted;
7600 if (msg_len <= sctp_wspace(asoc))
7603 /* Let another process have a go. Since we are going
7607 current_timeo = schedule_timeout(current_timeo);
7610 *timeo_p = current_timeo;
7614 finish_wait(&asoc->wait, &wait);
7616 /* Release the association's refcnt. */
7617 sctp_association_put(asoc);
7626 err = sock_intr_errno(*timeo_p);
7634 void sctp_data_ready(struct sock *sk)
7636 struct socket_wq *wq;
7639 wq = rcu_dereference(sk->sk_wq);
7640 if (skwq_has_sleeper(wq))
7641 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
7642 POLLRDNORM | POLLRDBAND);
7643 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
7647 /* If socket sndbuf has changed, wake up all per association waiters. */
7648 void sctp_write_space(struct sock *sk)
7650 struct sctp_association *asoc;
7652 /* Wake up the tasks in each wait queue. */
7653 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
7654 __sctp_write_space(asoc);
7658 /* Is there any sndbuf space available on the socket?
7660 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
7661 * associations on the same socket. For a UDP-style socket with
7662 * multiple associations, it is possible for it to be "unwriteable"
7663 * prematurely. I assume that this is acceptable because
7664 * a premature "unwriteable" is better than an accidental "writeable" which
7665 * would cause an unwanted block under certain circumstances. For the 1-1
7666 * UDP-style sockets or TCP-style sockets, this code should work.
7669 static int sctp_writeable(struct sock *sk)
7673 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
7679 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
7680 * returns immediately with EINPROGRESS.
7682 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
7684 struct sock *sk = asoc->base.sk;
7686 long current_timeo = *timeo_p;
7689 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
7691 /* Increment the association's refcnt. */
7692 sctp_association_hold(asoc);
7695 prepare_to_wait_exclusive(&asoc->wait, &wait,
7696 TASK_INTERRUPTIBLE);
7699 if (sk->sk_shutdown & RCV_SHUTDOWN)
7701 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7704 if (signal_pending(current))
7705 goto do_interrupted;
7707 if (sctp_state(asoc, ESTABLISHED))
7710 /* Let another process have a go. Since we are going
7714 current_timeo = schedule_timeout(current_timeo);
7717 *timeo_p = current_timeo;
7721 finish_wait(&asoc->wait, &wait);
7723 /* Release the association's refcnt. */
7724 sctp_association_put(asoc);
7729 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
7732 err = -ECONNREFUSED;
7736 err = sock_intr_errno(*timeo_p);
7744 static int sctp_wait_for_accept(struct sock *sk, long timeo)
7746 struct sctp_endpoint *ep;
7750 ep = sctp_sk(sk)->ep;
7754 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
7755 TASK_INTERRUPTIBLE);
7757 if (list_empty(&ep->asocs)) {
7759 timeo = schedule_timeout(timeo);
7764 if (!sctp_sstate(sk, LISTENING))
7768 if (!list_empty(&ep->asocs))
7771 err = sock_intr_errno(timeo);
7772 if (signal_pending(current))
7780 finish_wait(sk_sleep(sk), &wait);
7785 static void sctp_wait_for_close(struct sock *sk, long timeout)
7790 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7791 if (list_empty(&sctp_sk(sk)->ep->asocs))
7794 timeout = schedule_timeout(timeout);
7796 } while (!signal_pending(current) && timeout);
7798 finish_wait(sk_sleep(sk), &wait);
7801 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
7803 struct sk_buff *frag;
7808 /* Don't forget the fragments. */
7809 skb_walk_frags(skb, frag)
7810 sctp_skb_set_owner_r_frag(frag, sk);
7813 sctp_skb_set_owner_r(skb, sk);
7816 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
7817 struct sctp_association *asoc)
7819 struct inet_sock *inet = inet_sk(sk);
7820 struct inet_sock *newinet;
7822 newsk->sk_type = sk->sk_type;
7823 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
7824 newsk->sk_flags = sk->sk_flags;
7825 newsk->sk_tsflags = sk->sk_tsflags;
7826 newsk->sk_no_check_tx = sk->sk_no_check_tx;
7827 newsk->sk_no_check_rx = sk->sk_no_check_rx;
7828 newsk->sk_reuse = sk->sk_reuse;
7830 newsk->sk_shutdown = sk->sk_shutdown;
7831 newsk->sk_destruct = sctp_destruct_sock;
7832 newsk->sk_family = sk->sk_family;
7833 newsk->sk_protocol = IPPROTO_SCTP;
7834 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
7835 newsk->sk_sndbuf = sk->sk_sndbuf;
7836 newsk->sk_rcvbuf = sk->sk_rcvbuf;
7837 newsk->sk_lingertime = sk->sk_lingertime;
7838 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
7839 newsk->sk_sndtimeo = sk->sk_sndtimeo;
7840 newsk->sk_rxhash = sk->sk_rxhash;
7842 newinet = inet_sk(newsk);
7844 /* Initialize sk's sport, dport, rcv_saddr and daddr for
7845 * getsockname() and getpeername()
7847 newinet->inet_sport = inet->inet_sport;
7848 newinet->inet_saddr = inet->inet_saddr;
7849 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
7850 newinet->inet_dport = htons(asoc->peer.port);
7851 newinet->pmtudisc = inet->pmtudisc;
7852 newinet->inet_id = asoc->next_tsn ^ jiffies;
7854 newinet->uc_ttl = inet->uc_ttl;
7855 newinet->mc_loop = 1;
7856 newinet->mc_ttl = 1;
7857 newinet->mc_index = 0;
7858 newinet->mc_list = NULL;
7860 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
7861 net_enable_timestamp();
7863 security_sk_clone(sk, newsk);
7866 static inline void sctp_copy_descendant(struct sock *sk_to,
7867 const struct sock *sk_from)
7869 int ancestor_size = sizeof(struct inet_sock) +
7870 sizeof(struct sctp_sock) -
7871 offsetof(struct sctp_sock, auto_asconf_list);
7873 if (sk_from->sk_family == PF_INET6)
7874 ancestor_size += sizeof(struct ipv6_pinfo);
7876 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
7879 /* Populate the fields of the newsk from the oldsk and migrate the assoc
7880 * and its messages to the newsk.
7882 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
7883 struct sctp_association *assoc,
7884 sctp_socket_type_t type)
7886 struct sctp_sock *oldsp = sctp_sk(oldsk);
7887 struct sctp_sock *newsp = sctp_sk(newsk);
7888 struct sctp_bind_bucket *pp; /* hash list port iterator */
7889 struct sctp_endpoint *newep = newsp->ep;
7890 struct sk_buff *skb, *tmp;
7891 struct sctp_ulpevent *event;
7892 struct sctp_bind_hashbucket *head;
7894 /* Migrate socket buffer sizes and all the socket level options to the
7897 newsk->sk_sndbuf = oldsk->sk_sndbuf;
7898 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
7899 /* Brute force copy old sctp opt. */
7900 sctp_copy_descendant(newsk, oldsk);
7902 /* Restore the ep value that was overwritten with the above structure
7908 /* Hook this new socket in to the bind_hash list. */
7909 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
7910 inet_sk(oldsk)->inet_num)];
7911 spin_lock_bh(&head->lock);
7912 pp = sctp_sk(oldsk)->bind_hash;
7913 sk_add_bind_node(newsk, &pp->owner);
7914 sctp_sk(newsk)->bind_hash = pp;
7915 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
7916 spin_unlock_bh(&head->lock);
7918 /* Copy the bind_addr list from the original endpoint to the new
7919 * endpoint so that we can handle restarts properly
7921 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
7922 &oldsp->ep->base.bind_addr, GFP_KERNEL);
7924 /* Move any messages in the old socket's receive queue that are for the
7925 * peeled off association to the new socket's receive queue.
7927 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
7928 event = sctp_skb2event(skb);
7929 if (event->asoc == assoc) {
7930 __skb_unlink(skb, &oldsk->sk_receive_queue);
7931 __skb_queue_tail(&newsk->sk_receive_queue, skb);
7932 sctp_skb_set_owner_r_frag(skb, newsk);
7936 /* Clean up any messages pending delivery due to partial
7937 * delivery. Three cases:
7938 * 1) No partial deliver; no work.
7939 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7940 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7942 skb_queue_head_init(&newsp->pd_lobby);
7943 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
7945 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
7946 struct sk_buff_head *queue;
7948 /* Decide which queue to move pd_lobby skbs to. */
7949 if (assoc->ulpq.pd_mode) {
7950 queue = &newsp->pd_lobby;
7952 queue = &newsk->sk_receive_queue;
7954 /* Walk through the pd_lobby, looking for skbs that
7955 * need moved to the new socket.
7957 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
7958 event = sctp_skb2event(skb);
7959 if (event->asoc == assoc) {
7960 __skb_unlink(skb, &oldsp->pd_lobby);
7961 __skb_queue_tail(queue, skb);
7962 sctp_skb_set_owner_r_frag(skb, newsk);
7966 /* Clear up any skbs waiting for the partial
7967 * delivery to finish.
7969 if (assoc->ulpq.pd_mode)
7970 sctp_clear_pd(oldsk, NULL);
7974 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7975 sctp_skb_set_owner_r_frag(skb, newsk);
7977 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7978 sctp_skb_set_owner_r_frag(skb, newsk);
7980 /* Set the type of socket to indicate that it is peeled off from the
7981 * original UDP-style socket or created with the accept() call on a
7982 * TCP-style socket..
7986 /* Mark the new socket "in-use" by the user so that any packets
7987 * that may arrive on the association after we've moved it are
7988 * queued to the backlog. This prevents a potential race between
7989 * backlog processing on the old socket and new-packet processing
7990 * on the new socket.
7992 * The caller has just allocated newsk so we can guarantee that other
7993 * paths won't try to lock it and then oldsk.
7995 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7996 sctp_assoc_migrate(assoc, newsk);
7998 /* If the association on the newsk is already closed before accept()
7999 * is called, set RCV_SHUTDOWN flag.
8001 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP)) {
8002 newsk->sk_state = SCTP_SS_CLOSED;
8003 newsk->sk_shutdown |= RCV_SHUTDOWN;
8005 newsk->sk_state = SCTP_SS_ESTABLISHED;
8008 release_sock(newsk);
8012 /* This proto struct describes the ULP interface for SCTP. */
8013 struct proto sctp_prot = {
8015 .owner = THIS_MODULE,
8016 .close = sctp_close,
8017 .connect = sctp_connect,
8018 .disconnect = sctp_disconnect,
8019 .accept = sctp_accept,
8020 .ioctl = sctp_ioctl,
8021 .init = sctp_init_sock,
8022 .destroy = sctp_destroy_sock,
8023 .shutdown = sctp_shutdown,
8024 .setsockopt = sctp_setsockopt,
8025 .getsockopt = sctp_getsockopt,
8026 .sendmsg = sctp_sendmsg,
8027 .recvmsg = sctp_recvmsg,
8029 .backlog_rcv = sctp_backlog_rcv,
8031 .unhash = sctp_unhash,
8032 .get_port = sctp_get_port,
8033 .obj_size = sizeof(struct sctp_sock),
8034 .sysctl_mem = sysctl_sctp_mem,
8035 .sysctl_rmem = sysctl_sctp_rmem,
8036 .sysctl_wmem = sysctl_sctp_wmem,
8037 .memory_pressure = &sctp_memory_pressure,
8038 .enter_memory_pressure = sctp_enter_memory_pressure,
8039 .memory_allocated = &sctp_memory_allocated,
8040 .sockets_allocated = &sctp_sockets_allocated,
8043 #if IS_ENABLED(CONFIG_IPV6)
8045 #include <net/transp_v6.h>
8046 static void sctp_v6_destroy_sock(struct sock *sk)
8048 sctp_destroy_sock(sk);
8049 inet6_destroy_sock(sk);
8052 struct proto sctpv6_prot = {
8054 .owner = THIS_MODULE,
8055 .close = sctp_close,
8056 .connect = sctp_connect,
8057 .disconnect = sctp_disconnect,
8058 .accept = sctp_accept,
8059 .ioctl = sctp_ioctl,
8060 .init = sctp_init_sock,
8061 .destroy = sctp_v6_destroy_sock,
8062 .shutdown = sctp_shutdown,
8063 .setsockopt = sctp_setsockopt,
8064 .getsockopt = sctp_getsockopt,
8065 .sendmsg = sctp_sendmsg,
8066 .recvmsg = sctp_recvmsg,
8068 .backlog_rcv = sctp_backlog_rcv,
8070 .unhash = sctp_unhash,
8071 .get_port = sctp_get_port,
8072 .obj_size = sizeof(struct sctp6_sock),
8073 .sysctl_mem = sysctl_sctp_mem,
8074 .sysctl_rmem = sysctl_sctp_rmem,
8075 .sysctl_wmem = sysctl_sctp_wmem,
8076 .memory_pressure = &sctp_memory_pressure,
8077 .enter_memory_pressure = sctp_enter_memory_pressure,
8078 .memory_allocated = &sctp_memory_allocated,
8079 .sockets_allocated = &sctp_sockets_allocated,
8081 #endif /* IS_ENABLED(CONFIG_IPV6) */