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
38 * Written or modified by:
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 < PROT_SOCK &&
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 dst_release(trans->dst);
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 dst_release(transport->dst);
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 < PROT_SOCK &&
1160 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1166 scope = sctp_scope(&to);
1167 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1173 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1181 /* Prime the peer's transport structures. */
1182 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1190 addr_buf += af->sockaddr_len;
1191 walk_size += af->sockaddr_len;
1194 /* In case the user of sctp_connectx() wants an association
1195 * id back, assign one now.
1198 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1203 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1208 /* Initialize sk's dport and daddr for getpeername() */
1209 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1210 sp->pf->to_sk_daddr(sa_addr, sk);
1213 /* in-kernel sockets don't generally have a file allocated to them
1214 * if all they do is call sock_create_kern().
1216 if (sk->sk_socket->file)
1217 f_flags = sk->sk_socket->file->f_flags;
1219 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1222 *assoc_id = asoc->assoc_id;
1223 err = sctp_wait_for_connect(asoc, &timeo);
1224 /* Note: the asoc may be freed after the return of
1225 * sctp_wait_for_connect.
1228 /* Don't free association on exit. */
1232 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1233 __func__, asoc, kaddrs, err);
1236 /* sctp_primitive_ASSOCIATE may have added this association
1237 * To the hash table, try to unhash it, just in case, its a noop
1238 * if it wasn't hashed so we're safe
1240 sctp_association_free(asoc);
1245 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1248 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1249 * sctp_assoc_t *asoc);
1251 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1252 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1253 * or IPv6 addresses.
1255 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1256 * Section 3.1.2 for this usage.
1258 * addrs is a pointer to an array of one or more socket addresses. Each
1259 * address is contained in its appropriate structure (i.e. struct
1260 * sockaddr_in or struct sockaddr_in6) the family of the address type
1261 * must be used to distengish the address length (note that this
1262 * representation is termed a "packed array" of addresses). The caller
1263 * specifies the number of addresses in the array with addrcnt.
1265 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1266 * the association id of the new association. On failure, sctp_connectx()
1267 * returns -1, and sets errno to the appropriate error code. The assoc_id
1268 * is not touched by the kernel.
1270 * For SCTP, the port given in each socket address must be the same, or
1271 * sctp_connectx() will fail, setting errno to EINVAL.
1273 * An application can use sctp_connectx to initiate an association with
1274 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1275 * allows a caller to specify multiple addresses at which a peer can be
1276 * reached. The way the SCTP stack uses the list of addresses to set up
1277 * the association is implementation dependent. This function only
1278 * specifies that the stack will try to make use of all the addresses in
1279 * the list when needed.
1281 * Note that the list of addresses passed in is only used for setting up
1282 * the association. It does not necessarily equal the set of addresses
1283 * the peer uses for the resulting association. If the caller wants to
1284 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1285 * retrieve them after the association has been set up.
1287 * Basically do nothing but copying the addresses from user to kernel
1288 * land and invoking either sctp_connectx(). This is used for tunneling
1289 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1291 * We don't use copy_from_user() for optimization: we first do the
1292 * sanity checks (buffer size -fast- and access check-healthy
1293 * pointer); if all of those succeed, then we can alloc the memory
1294 * (expensive operation) needed to copy the data to kernel. Then we do
1295 * the copying without checking the user space area
1296 * (__copy_from_user()).
1298 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1301 * sk The sk of the socket
1302 * addrs The pointer to the addresses in user land
1303 * addrssize Size of the addrs buffer
1305 * Returns >=0 if ok, <0 errno code on error.
1307 static int __sctp_setsockopt_connectx(struct sock *sk,
1308 struct sockaddr __user *addrs,
1310 sctp_assoc_t *assoc_id)
1312 struct sockaddr *kaddrs;
1313 gfp_t gfp = GFP_KERNEL;
1316 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1317 __func__, sk, addrs, addrs_size);
1319 if (unlikely(addrs_size <= 0))
1322 /* Check the user passed a healthy pointer. */
1323 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1326 /* Alloc space for the address array in kernel memory. */
1327 if (sk->sk_socket->file)
1328 gfp = GFP_USER | __GFP_NOWARN;
1329 kaddrs = kmalloc(addrs_size, gfp);
1330 if (unlikely(!kaddrs))
1333 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1336 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1345 * This is an older interface. It's kept for backward compatibility
1346 * to the option that doesn't provide association id.
1348 static int sctp_setsockopt_connectx_old(struct sock *sk,
1349 struct sockaddr __user *addrs,
1352 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1356 * New interface for the API. The since the API is done with a socket
1357 * option, to make it simple we feed back the association id is as a return
1358 * indication to the call. Error is always negative and association id is
1361 static int sctp_setsockopt_connectx(struct sock *sk,
1362 struct sockaddr __user *addrs,
1365 sctp_assoc_t assoc_id = 0;
1368 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1377 * New (hopefully final) interface for the API.
1378 * We use the sctp_getaddrs_old structure so that use-space library
1379 * can avoid any unnecessary allocations. The only different part
1380 * is that we store the actual length of the address buffer into the
1381 * addrs_num structure member. That way we can re-use the existing
1384 #ifdef CONFIG_COMPAT
1385 struct compat_sctp_getaddrs_old {
1386 sctp_assoc_t assoc_id;
1388 compat_uptr_t addrs; /* struct sockaddr * */
1392 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1393 char __user *optval,
1396 struct sctp_getaddrs_old param;
1397 sctp_assoc_t assoc_id = 0;
1400 #ifdef CONFIG_COMPAT
1401 if (in_compat_syscall()) {
1402 struct compat_sctp_getaddrs_old param32;
1404 if (len < sizeof(param32))
1406 if (copy_from_user(¶m32, optval, sizeof(param32)))
1409 param.assoc_id = param32.assoc_id;
1410 param.addr_num = param32.addr_num;
1411 param.addrs = compat_ptr(param32.addrs);
1415 if (len < sizeof(param))
1417 if (copy_from_user(¶m, optval, sizeof(param)))
1421 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1422 param.addrs, param.addr_num,
1424 if (err == 0 || err == -EINPROGRESS) {
1425 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1427 if (put_user(sizeof(assoc_id), optlen))
1434 /* API 3.1.4 close() - UDP Style Syntax
1435 * Applications use close() to perform graceful shutdown (as described in
1436 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1437 * by a UDP-style socket.
1441 * ret = close(int sd);
1443 * sd - the socket descriptor of the associations to be closed.
1445 * To gracefully shutdown a specific association represented by the
1446 * UDP-style socket, an application should use the sendmsg() call,
1447 * passing no user data, but including the appropriate flag in the
1448 * ancillary data (see Section xxxx).
1450 * If sd in the close() call is a branched-off socket representing only
1451 * one association, the shutdown is performed on that association only.
1453 * 4.1.6 close() - TCP Style Syntax
1455 * Applications use close() to gracefully close down an association.
1459 * int close(int sd);
1461 * sd - the socket descriptor of the association to be closed.
1463 * After an application calls close() on a socket descriptor, no further
1464 * socket operations will succeed on that descriptor.
1466 * API 7.1.4 SO_LINGER
1468 * An application using the TCP-style socket can use this option to
1469 * perform the SCTP ABORT primitive. The linger option structure is:
1472 * int l_onoff; // option on/off
1473 * int l_linger; // linger time
1476 * To enable the option, set l_onoff to 1. If the l_linger value is set
1477 * to 0, calling close() is the same as the ABORT primitive. If the
1478 * value is set to a negative value, the setsockopt() call will return
1479 * an error. If the value is set to a positive value linger_time, the
1480 * close() can be blocked for at most linger_time ms. If the graceful
1481 * shutdown phase does not finish during this period, close() will
1482 * return but the graceful shutdown phase continues in the system.
1484 static void sctp_close(struct sock *sk, long timeout)
1486 struct net *net = sock_net(sk);
1487 struct sctp_endpoint *ep;
1488 struct sctp_association *asoc;
1489 struct list_head *pos, *temp;
1490 unsigned int data_was_unread;
1492 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1495 sk->sk_shutdown = SHUTDOWN_MASK;
1496 sk->sk_state = SCTP_SS_CLOSING;
1498 ep = sctp_sk(sk)->ep;
1500 /* Clean up any skbs sitting on the receive queue. */
1501 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1502 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1504 /* Walk all associations on an endpoint. */
1505 list_for_each_safe(pos, temp, &ep->asocs) {
1506 asoc = list_entry(pos, struct sctp_association, asocs);
1508 if (sctp_style(sk, TCP)) {
1509 /* A closed association can still be in the list if
1510 * it belongs to a TCP-style listening socket that is
1511 * not yet accepted. If so, free it. If not, send an
1512 * ABORT or SHUTDOWN based on the linger options.
1514 if (sctp_state(asoc, CLOSED)) {
1515 sctp_association_free(asoc);
1520 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1521 !skb_queue_empty(&asoc->ulpq.reasm) ||
1522 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1523 struct sctp_chunk *chunk;
1525 chunk = sctp_make_abort_user(asoc, NULL, 0);
1526 sctp_primitive_ABORT(net, asoc, chunk);
1528 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1531 /* On a TCP-style socket, block for at most linger_time if set. */
1532 if (sctp_style(sk, TCP) && timeout)
1533 sctp_wait_for_close(sk, timeout);
1535 /* This will run the backlog queue. */
1538 /* Supposedly, no process has access to the socket, but
1539 * the net layers still may.
1540 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1541 * held and that should be grabbed before socket lock.
1543 spin_lock_bh(&net->sctp.addr_wq_lock);
1546 /* Hold the sock, since sk_common_release() will put sock_put()
1547 * and we have just a little more cleanup.
1550 sk_common_release(sk);
1553 spin_unlock_bh(&net->sctp.addr_wq_lock);
1557 SCTP_DBG_OBJCNT_DEC(sock);
1560 /* Handle EPIPE error. */
1561 static int sctp_error(struct sock *sk, int flags, int err)
1564 err = sock_error(sk) ? : -EPIPE;
1565 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1566 send_sig(SIGPIPE, current, 0);
1570 /* API 3.1.3 sendmsg() - UDP Style Syntax
1572 * An application uses sendmsg() and recvmsg() calls to transmit data to
1573 * and receive data from its peer.
1575 * ssize_t sendmsg(int socket, const struct msghdr *message,
1578 * socket - the socket descriptor of the endpoint.
1579 * message - pointer to the msghdr structure which contains a single
1580 * user message and possibly some ancillary data.
1582 * See Section 5 for complete description of the data
1585 * flags - flags sent or received with the user message, see Section
1586 * 5 for complete description of the flags.
1588 * Note: This function could use a rewrite especially when explicit
1589 * connect support comes in.
1591 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1593 static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1595 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
1597 struct net *net = sock_net(sk);
1598 struct sctp_sock *sp;
1599 struct sctp_endpoint *ep;
1600 struct sctp_association *new_asoc = NULL, *asoc = NULL;
1601 struct sctp_transport *transport, *chunk_tp;
1602 struct sctp_chunk *chunk;
1604 struct sockaddr *msg_name = NULL;
1605 struct sctp_sndrcvinfo default_sinfo;
1606 struct sctp_sndrcvinfo *sinfo;
1607 struct sctp_initmsg *sinit;
1608 sctp_assoc_t associd = 0;
1609 sctp_cmsgs_t cmsgs = { NULL };
1611 bool fill_sinfo_ttl = false, wait_connect = false;
1612 struct sctp_datamsg *datamsg;
1613 int msg_flags = msg->msg_flags;
1614 __u16 sinfo_flags = 0;
1622 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1625 /* We cannot send a message over a TCP-style listening socket. */
1626 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1631 /* Parse out the SCTP CMSGs. */
1632 err = sctp_msghdr_parse(msg, &cmsgs);
1634 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1638 /* Fetch the destination address for this packet. This
1639 * address only selects the association--it is not necessarily
1640 * the address we will send to.
1641 * For a peeled-off socket, msg_name is ignored.
1643 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1644 int msg_namelen = msg->msg_namelen;
1646 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1651 if (msg_namelen > sizeof(to))
1652 msg_namelen = sizeof(to);
1653 memcpy(&to, msg->msg_name, msg_namelen);
1654 msg_name = msg->msg_name;
1658 if (cmsgs.sinfo != NULL) {
1659 memset(&default_sinfo, 0, sizeof(default_sinfo));
1660 default_sinfo.sinfo_stream = cmsgs.sinfo->snd_sid;
1661 default_sinfo.sinfo_flags = cmsgs.sinfo->snd_flags;
1662 default_sinfo.sinfo_ppid = cmsgs.sinfo->snd_ppid;
1663 default_sinfo.sinfo_context = cmsgs.sinfo->snd_context;
1664 default_sinfo.sinfo_assoc_id = cmsgs.sinfo->snd_assoc_id;
1666 sinfo = &default_sinfo;
1667 fill_sinfo_ttl = true;
1669 sinfo = cmsgs.srinfo;
1671 /* Did the user specify SNDINFO/SNDRCVINFO? */
1673 sinfo_flags = sinfo->sinfo_flags;
1674 associd = sinfo->sinfo_assoc_id;
1677 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1678 msg_len, sinfo_flags);
1680 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1681 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1686 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1687 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1688 * If SCTP_ABORT is set, the message length could be non zero with
1689 * the msg_iov set to the user abort reason.
1691 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1692 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1697 /* If SCTP_ADDR_OVER is set, there must be an address
1698 * specified in msg_name.
1700 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1707 pr_debug("%s: about to look up association\n", __func__);
1711 /* If a msg_name has been specified, assume this is to be used. */
1713 /* Look for a matching association on the endpoint. */
1714 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1716 /* If we could not find a matching association on the
1717 * endpoint, make sure that it is not a TCP-style
1718 * socket that already has an association or there is
1719 * no peeled-off association on another socket.
1722 ((sctp_style(sk, TCP) &&
1723 (sctp_sstate(sk, ESTABLISHED) ||
1724 sctp_sstate(sk, CLOSING))) ||
1725 sctp_endpoint_is_peeled_off(ep, &to))) {
1726 err = -EADDRNOTAVAIL;
1730 asoc = sctp_id2assoc(sk, associd);
1738 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1740 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1741 * socket that has an association in CLOSED state. This can
1742 * happen when an accepted socket has an association that is
1745 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1750 if (sinfo_flags & SCTP_EOF) {
1751 pr_debug("%s: shutting down association:%p\n",
1754 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1758 if (sinfo_flags & SCTP_ABORT) {
1760 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1766 pr_debug("%s: aborting association:%p\n",
1769 sctp_primitive_ABORT(net, asoc, chunk);
1775 /* Do we need to create the association? */
1777 pr_debug("%s: there is no association yet\n", __func__);
1779 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1784 /* Check for invalid stream against the stream counts,
1785 * either the default or the user specified stream counts.
1788 if (!sinit || !sinit->sinit_num_ostreams) {
1789 /* Check against the defaults. */
1790 if (sinfo->sinfo_stream >=
1791 sp->initmsg.sinit_num_ostreams) {
1796 /* Check against the requested. */
1797 if (sinfo->sinfo_stream >=
1798 sinit->sinit_num_ostreams) {
1806 * API 3.1.2 bind() - UDP Style Syntax
1807 * If a bind() or sctp_bindx() is not called prior to a
1808 * sendmsg() call that initiates a new association, the
1809 * system picks an ephemeral port and will choose an address
1810 * set equivalent to binding with a wildcard address.
1812 if (!ep->base.bind_addr.port) {
1813 if (sctp_autobind(sk)) {
1819 * If an unprivileged user inherits a one-to-many
1820 * style socket with open associations on a privileged
1821 * port, it MAY be permitted to accept new associations,
1822 * but it SHOULD NOT be permitted to open new
1825 if (ep->base.bind_addr.port < PROT_SOCK &&
1826 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1832 scope = sctp_scope(&to);
1833 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1839 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1845 /* If the SCTP_INIT ancillary data is specified, set all
1846 * the association init values accordingly.
1849 if (sinit->sinit_num_ostreams) {
1850 asoc->c.sinit_num_ostreams =
1851 sinit->sinit_num_ostreams;
1853 if (sinit->sinit_max_instreams) {
1854 asoc->c.sinit_max_instreams =
1855 sinit->sinit_max_instreams;
1857 if (sinit->sinit_max_attempts) {
1858 asoc->max_init_attempts
1859 = sinit->sinit_max_attempts;
1861 if (sinit->sinit_max_init_timeo) {
1862 asoc->max_init_timeo =
1863 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1867 /* Prime the peer's transport structures. */
1868 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1875 /* ASSERT: we have a valid association at this point. */
1876 pr_debug("%s: we have a valid association\n", __func__);
1879 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1880 * one with some defaults.
1882 memset(&default_sinfo, 0, sizeof(default_sinfo));
1883 default_sinfo.sinfo_stream = asoc->default_stream;
1884 default_sinfo.sinfo_flags = asoc->default_flags;
1885 default_sinfo.sinfo_ppid = asoc->default_ppid;
1886 default_sinfo.sinfo_context = asoc->default_context;
1887 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1888 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1890 sinfo = &default_sinfo;
1891 } else if (fill_sinfo_ttl) {
1892 /* In case SNDINFO was specified, we still need to fill
1893 * it with a default ttl from the assoc here.
1895 sinfo->sinfo_timetolive = asoc->default_timetolive;
1898 /* API 7.1.7, the sndbuf size per association bounds the
1899 * maximum size of data that can be sent in a single send call.
1901 if (msg_len > sk->sk_sndbuf) {
1906 if (asoc->pmtu_pending)
1907 sctp_assoc_pending_pmtu(sk, asoc);
1909 /* If fragmentation is disabled and the message length exceeds the
1910 * association fragmentation point, return EMSGSIZE. The I-D
1911 * does not specify what this error is, but this looks like
1914 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1919 /* Check for invalid stream. */
1920 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1925 if (sctp_wspace(asoc) < msg_len)
1926 sctp_prsctp_prune(asoc, sinfo, msg_len - sctp_wspace(asoc));
1928 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1929 if (!sctp_wspace(asoc)) {
1930 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1935 /* If an address is passed with the sendto/sendmsg call, it is used
1936 * to override the primary destination address in the TCP model, or
1937 * when SCTP_ADDR_OVER flag is set in the UDP model.
1939 if ((sctp_style(sk, TCP) && msg_name) ||
1940 (sinfo_flags & SCTP_ADDR_OVER)) {
1941 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1949 /* Auto-connect, if we aren't connected already. */
1950 if (sctp_state(asoc, CLOSED)) {
1951 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1955 wait_connect = true;
1956 pr_debug("%s: we associated primitively\n", __func__);
1959 /* Break the message into multiple chunks of maximum size. */
1960 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
1961 if (IS_ERR(datamsg)) {
1962 err = PTR_ERR(datamsg);
1966 /* Now send the (possibly) fragmented message. */
1967 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1968 sctp_chunk_hold(chunk);
1970 /* Do accounting for the write space. */
1971 sctp_set_owner_w(chunk);
1973 chunk->transport = chunk_tp;
1976 /* Send it to the lower layers. Note: all chunks
1977 * must either fail or succeed. The lower layer
1978 * works that way today. Keep it that way or this
1981 err = sctp_primitive_SEND(net, asoc, datamsg);
1982 /* Did the lower layer accept the chunk? */
1984 sctp_datamsg_free(datamsg);
1988 pr_debug("%s: we sent primitively\n", __func__);
1990 sctp_datamsg_put(datamsg);
1993 if (unlikely(wait_connect)) {
1994 timeo = sock_sndtimeo(sk, msg_flags & MSG_DONTWAIT);
1995 sctp_wait_for_connect(asoc, &timeo);
1998 /* If we are already past ASSOCIATE, the lower
1999 * layers are responsible for association cleanup.
2005 sctp_association_free(asoc);
2010 return sctp_error(sk, msg_flags, err);
2017 err = sock_error(sk);
2027 /* This is an extended version of skb_pull() that removes the data from the
2028 * start of a skb even when data is spread across the list of skb's in the
2029 * frag_list. len specifies the total amount of data that needs to be removed.
2030 * when 'len' bytes could be removed from the skb, it returns 0.
2031 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2032 * could not be removed.
2034 static int sctp_skb_pull(struct sk_buff *skb, int len)
2036 struct sk_buff *list;
2037 int skb_len = skb_headlen(skb);
2040 if (len <= skb_len) {
2041 __skb_pull(skb, len);
2045 __skb_pull(skb, skb_len);
2047 skb_walk_frags(skb, list) {
2048 rlen = sctp_skb_pull(list, len);
2049 skb->len -= (len-rlen);
2050 skb->data_len -= (len-rlen);
2061 /* API 3.1.3 recvmsg() - UDP Style Syntax
2063 * ssize_t recvmsg(int socket, struct msghdr *message,
2066 * socket - the socket descriptor of the endpoint.
2067 * message - pointer to the msghdr structure which contains a single
2068 * user message and possibly some ancillary data.
2070 * See Section 5 for complete description of the data
2073 * flags - flags sent or received with the user message, see Section
2074 * 5 for complete description of the flags.
2076 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2077 int noblock, int flags, int *addr_len)
2079 struct sctp_ulpevent *event = NULL;
2080 struct sctp_sock *sp = sctp_sk(sk);
2081 struct sk_buff *skb, *head_skb;
2086 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2087 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2092 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED) &&
2093 !sctp_sstate(sk, CLOSING) && !sctp_sstate(sk, CLOSED)) {
2098 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2102 /* Get the total length of the skb including any skb's in the
2111 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2113 event = sctp_skb2event(skb);
2118 if (event->chunk && event->chunk->head_skb)
2119 head_skb = event->chunk->head_skb;
2122 sock_recv_ts_and_drops(msg, sk, head_skb);
2123 if (sctp_ulpevent_is_notification(event)) {
2124 msg->msg_flags |= MSG_NOTIFICATION;
2125 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2127 sp->pf->skb_msgname(head_skb, msg->msg_name, addr_len);
2130 /* Check if we allow SCTP_NXTINFO. */
2131 if (sp->recvnxtinfo)
2132 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2133 /* Check if we allow SCTP_RCVINFO. */
2134 if (sp->recvrcvinfo)
2135 sctp_ulpevent_read_rcvinfo(event, msg);
2136 /* Check if we allow SCTP_SNDRCVINFO. */
2137 if (sp->subscribe.sctp_data_io_event)
2138 sctp_ulpevent_read_sndrcvinfo(event, msg);
2142 /* If skb's length exceeds the user's buffer, update the skb and
2143 * push it back to the receive_queue so that the next call to
2144 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2146 if (skb_len > copied) {
2147 msg->msg_flags &= ~MSG_EOR;
2148 if (flags & MSG_PEEK)
2150 sctp_skb_pull(skb, copied);
2151 skb_queue_head(&sk->sk_receive_queue, skb);
2153 /* When only partial message is copied to the user, increase
2154 * rwnd by that amount. If all the data in the skb is read,
2155 * rwnd is updated when the event is freed.
2157 if (!sctp_ulpevent_is_notification(event))
2158 sctp_assoc_rwnd_increase(event->asoc, copied);
2160 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2161 (event->msg_flags & MSG_EOR))
2162 msg->msg_flags |= MSG_EOR;
2164 msg->msg_flags &= ~MSG_EOR;
2167 if (flags & MSG_PEEK) {
2168 /* Release the skb reference acquired after peeking the skb in
2169 * sctp_skb_recv_datagram().
2173 /* Free the event which includes releasing the reference to
2174 * the owner of the skb, freeing the skb and updating the
2177 sctp_ulpevent_free(event);
2184 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2186 * This option is a on/off flag. If enabled no SCTP message
2187 * fragmentation will be performed. Instead if a message being sent
2188 * exceeds the current PMTU size, the message will NOT be sent and
2189 * instead a error will be indicated to the user.
2191 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2192 char __user *optval,
2193 unsigned int optlen)
2197 if (optlen < sizeof(int))
2200 if (get_user(val, (int __user *)optval))
2203 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2208 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2209 unsigned int optlen)
2211 struct sctp_association *asoc;
2212 struct sctp_ulpevent *event;
2214 if (optlen > sizeof(struct sctp_event_subscribe))
2216 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2219 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2220 * if there is no data to be sent or retransmit, the stack will
2221 * immediately send up this notification.
2223 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2224 &sctp_sk(sk)->subscribe)) {
2225 asoc = sctp_id2assoc(sk, 0);
2227 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2228 event = sctp_ulpevent_make_sender_dry_event(asoc,
2233 sctp_ulpq_tail_event(&asoc->ulpq, event);
2240 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2242 * This socket option is applicable to the UDP-style socket only. When
2243 * set it will cause associations that are idle for more than the
2244 * specified number of seconds to automatically close. An association
2245 * being idle is defined an association that has NOT sent or received
2246 * user data. The special value of '0' indicates that no automatic
2247 * close of any associations should be performed. The option expects an
2248 * integer defining the number of seconds of idle time before an
2249 * association is closed.
2251 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2252 unsigned int optlen)
2254 struct sctp_sock *sp = sctp_sk(sk);
2255 struct net *net = sock_net(sk);
2257 /* Applicable to UDP-style socket only */
2258 if (sctp_style(sk, TCP))
2260 if (optlen != sizeof(int))
2262 if (copy_from_user(&sp->autoclose, optval, optlen))
2265 if (sp->autoclose > net->sctp.max_autoclose)
2266 sp->autoclose = net->sctp.max_autoclose;
2271 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2273 * Applications can enable or disable heartbeats for any peer address of
2274 * an association, modify an address's heartbeat interval, force a
2275 * heartbeat to be sent immediately, and adjust the address's maximum
2276 * number of retransmissions sent before an address is considered
2277 * unreachable. The following structure is used to access and modify an
2278 * address's parameters:
2280 * struct sctp_paddrparams {
2281 * sctp_assoc_t spp_assoc_id;
2282 * struct sockaddr_storage spp_address;
2283 * uint32_t spp_hbinterval;
2284 * uint16_t spp_pathmaxrxt;
2285 * uint32_t spp_pathmtu;
2286 * uint32_t spp_sackdelay;
2287 * uint32_t spp_flags;
2290 * spp_assoc_id - (one-to-many style socket) This is filled in the
2291 * application, and identifies the association for
2293 * spp_address - This specifies which address is of interest.
2294 * spp_hbinterval - This contains the value of the heartbeat interval,
2295 * in milliseconds. If a value of zero
2296 * is present in this field then no changes are to
2297 * be made to this parameter.
2298 * spp_pathmaxrxt - This contains the maximum number of
2299 * retransmissions before this address shall be
2300 * considered unreachable. If a value of zero
2301 * is present in this field then no changes are to
2302 * be made to this parameter.
2303 * spp_pathmtu - When Path MTU discovery is disabled the value
2304 * specified here will be the "fixed" path mtu.
2305 * Note that if the spp_address field is empty
2306 * then all associations on this address will
2307 * have this fixed path mtu set upon them.
2309 * spp_sackdelay - When delayed sack is enabled, this value specifies
2310 * the number of milliseconds that sacks will be delayed
2311 * for. This value will apply to all addresses of an
2312 * association if the spp_address field is empty. Note
2313 * also, that if delayed sack is enabled and this
2314 * value is set to 0, no change is made to the last
2315 * recorded delayed sack timer value.
2317 * spp_flags - These flags are used to control various features
2318 * on an association. The flag field may contain
2319 * zero or more of the following options.
2321 * SPP_HB_ENABLE - Enable heartbeats on the
2322 * specified address. Note that if the address
2323 * field is empty all addresses for the association
2324 * have heartbeats enabled upon them.
2326 * SPP_HB_DISABLE - Disable heartbeats on the
2327 * speicifed address. Note that if the address
2328 * field is empty all addresses for the association
2329 * will have their heartbeats disabled. Note also
2330 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2331 * mutually exclusive, only one of these two should
2332 * be specified. Enabling both fields will have
2333 * undetermined results.
2335 * SPP_HB_DEMAND - Request a user initiated heartbeat
2336 * to be made immediately.
2338 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2339 * heartbeat delayis to be set to the value of 0
2342 * SPP_PMTUD_ENABLE - This field will enable PMTU
2343 * discovery upon the specified address. Note that
2344 * if the address feild is empty then all addresses
2345 * on the association are effected.
2347 * SPP_PMTUD_DISABLE - This field will disable PMTU
2348 * discovery upon the specified address. Note that
2349 * if the address feild is empty then all addresses
2350 * on the association are effected. Not also that
2351 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2352 * exclusive. Enabling both will have undetermined
2355 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2356 * on delayed sack. The time specified in spp_sackdelay
2357 * is used to specify the sack delay for this address. Note
2358 * that if spp_address is empty then all addresses will
2359 * enable delayed sack and take on the sack delay
2360 * value specified in spp_sackdelay.
2361 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2362 * off delayed sack. If the spp_address field is blank then
2363 * delayed sack is disabled for the entire association. Note
2364 * also that this field is mutually exclusive to
2365 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2368 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2369 struct sctp_transport *trans,
2370 struct sctp_association *asoc,
2371 struct sctp_sock *sp,
2374 int sackdelay_change)
2378 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2379 struct net *net = sock_net(trans->asoc->base.sk);
2381 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2386 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2387 * this field is ignored. Note also that a value of zero indicates
2388 * the current setting should be left unchanged.
2390 if (params->spp_flags & SPP_HB_ENABLE) {
2392 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2393 * set. This lets us use 0 value when this flag
2396 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2397 params->spp_hbinterval = 0;
2399 if (params->spp_hbinterval ||
2400 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2403 msecs_to_jiffies(params->spp_hbinterval);
2406 msecs_to_jiffies(params->spp_hbinterval);
2408 sp->hbinterval = params->spp_hbinterval;
2415 trans->param_flags =
2416 (trans->param_flags & ~SPP_HB) | hb_change;
2419 (asoc->param_flags & ~SPP_HB) | hb_change;
2422 (sp->param_flags & ~SPP_HB) | hb_change;
2426 /* When Path MTU discovery is disabled the value specified here will
2427 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2428 * include the flag SPP_PMTUD_DISABLE for this field to have any
2431 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2433 trans->pathmtu = params->spp_pathmtu;
2434 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2436 asoc->pathmtu = params->spp_pathmtu;
2437 sctp_frag_point(asoc, params->spp_pathmtu);
2439 sp->pathmtu = params->spp_pathmtu;
2445 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2446 (params->spp_flags & SPP_PMTUD_ENABLE);
2447 trans->param_flags =
2448 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2450 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2451 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2455 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2458 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2462 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2463 * value of this field is ignored. Note also that a value of zero
2464 * indicates the current setting should be left unchanged.
2466 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2469 msecs_to_jiffies(params->spp_sackdelay);
2472 msecs_to_jiffies(params->spp_sackdelay);
2474 sp->sackdelay = params->spp_sackdelay;
2478 if (sackdelay_change) {
2480 trans->param_flags =
2481 (trans->param_flags & ~SPP_SACKDELAY) |
2485 (asoc->param_flags & ~SPP_SACKDELAY) |
2489 (sp->param_flags & ~SPP_SACKDELAY) |
2494 /* Note that a value of zero indicates the current setting should be
2497 if (params->spp_pathmaxrxt) {
2499 trans->pathmaxrxt = params->spp_pathmaxrxt;
2501 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2503 sp->pathmaxrxt = params->spp_pathmaxrxt;
2510 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2511 char __user *optval,
2512 unsigned int optlen)
2514 struct sctp_paddrparams params;
2515 struct sctp_transport *trans = NULL;
2516 struct sctp_association *asoc = NULL;
2517 struct sctp_sock *sp = sctp_sk(sk);
2519 int hb_change, pmtud_change, sackdelay_change;
2521 if (optlen != sizeof(struct sctp_paddrparams))
2524 if (copy_from_user(¶ms, optval, optlen))
2527 /* Validate flags and value parameters. */
2528 hb_change = params.spp_flags & SPP_HB;
2529 pmtud_change = params.spp_flags & SPP_PMTUD;
2530 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2532 if (hb_change == SPP_HB ||
2533 pmtud_change == SPP_PMTUD ||
2534 sackdelay_change == SPP_SACKDELAY ||
2535 params.spp_sackdelay > 500 ||
2536 (params.spp_pathmtu &&
2537 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2540 /* If an address other than INADDR_ANY is specified, and
2541 * no transport is found, then the request is invalid.
2543 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
2544 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2545 params.spp_assoc_id);
2550 /* Get association, if assoc_id != 0 and the socket is a one
2551 * to many style socket, and an association was not found, then
2552 * the id was invalid.
2554 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2555 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2558 /* Heartbeat demand can only be sent on a transport or
2559 * association, but not a socket.
2561 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2564 /* Process parameters. */
2565 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2566 hb_change, pmtud_change,
2572 /* If changes are for association, also apply parameters to each
2575 if (!trans && asoc) {
2576 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2578 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2579 hb_change, pmtud_change,
2587 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2589 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2592 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2594 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2598 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2600 * This option will effect the way delayed acks are performed. This
2601 * option allows you to get or set the delayed ack time, in
2602 * milliseconds. It also allows changing the delayed ack frequency.
2603 * Changing the frequency to 1 disables the delayed sack algorithm. If
2604 * the assoc_id is 0, then this sets or gets the endpoints default
2605 * values. If the assoc_id field is non-zero, then the set or get
2606 * effects the specified association for the one to many model (the
2607 * assoc_id field is ignored by the one to one model). Note that if
2608 * sack_delay or sack_freq are 0 when setting this option, then the
2609 * current values will remain unchanged.
2611 * struct sctp_sack_info {
2612 * sctp_assoc_t sack_assoc_id;
2613 * uint32_t sack_delay;
2614 * uint32_t sack_freq;
2617 * sack_assoc_id - This parameter, indicates which association the user
2618 * is performing an action upon. Note that if this field's value is
2619 * zero then the endpoints default value is changed (effecting future
2620 * associations only).
2622 * sack_delay - This parameter contains the number of milliseconds that
2623 * the user is requesting the delayed ACK timer be set to. Note that
2624 * this value is defined in the standard to be between 200 and 500
2627 * sack_freq - This parameter contains the number of packets that must
2628 * be received before a sack is sent without waiting for the delay
2629 * timer to expire. The default value for this is 2, setting this
2630 * value to 1 will disable the delayed sack algorithm.
2633 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2634 char __user *optval, unsigned int optlen)
2636 struct sctp_sack_info params;
2637 struct sctp_transport *trans = NULL;
2638 struct sctp_association *asoc = NULL;
2639 struct sctp_sock *sp = sctp_sk(sk);
2641 if (optlen == sizeof(struct sctp_sack_info)) {
2642 if (copy_from_user(¶ms, optval, optlen))
2645 if (params.sack_delay == 0 && params.sack_freq == 0)
2647 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2648 pr_warn_ratelimited(DEPRECATED
2650 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2651 "Use struct sctp_sack_info instead\n",
2652 current->comm, task_pid_nr(current));
2653 if (copy_from_user(¶ms, optval, optlen))
2656 if (params.sack_delay == 0)
2657 params.sack_freq = 1;
2659 params.sack_freq = 0;
2663 /* Validate value parameter. */
2664 if (params.sack_delay > 500)
2667 /* Get association, if sack_assoc_id != 0 and the socket is a one
2668 * to many style socket, and an association was not found, then
2669 * the id was invalid.
2671 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2672 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2675 if (params.sack_delay) {
2678 msecs_to_jiffies(params.sack_delay);
2680 sctp_spp_sackdelay_enable(asoc->param_flags);
2682 sp->sackdelay = params.sack_delay;
2684 sctp_spp_sackdelay_enable(sp->param_flags);
2688 if (params.sack_freq == 1) {
2691 sctp_spp_sackdelay_disable(asoc->param_flags);
2694 sctp_spp_sackdelay_disable(sp->param_flags);
2696 } else if (params.sack_freq > 1) {
2698 asoc->sackfreq = params.sack_freq;
2700 sctp_spp_sackdelay_enable(asoc->param_flags);
2702 sp->sackfreq = params.sack_freq;
2704 sctp_spp_sackdelay_enable(sp->param_flags);
2708 /* If change is for association, also apply to each transport. */
2710 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2712 if (params.sack_delay) {
2714 msecs_to_jiffies(params.sack_delay);
2715 trans->param_flags =
2716 sctp_spp_sackdelay_enable(trans->param_flags);
2718 if (params.sack_freq == 1) {
2719 trans->param_flags =
2720 sctp_spp_sackdelay_disable(trans->param_flags);
2721 } else if (params.sack_freq > 1) {
2722 trans->sackfreq = params.sack_freq;
2723 trans->param_flags =
2724 sctp_spp_sackdelay_enable(trans->param_flags);
2732 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2734 * Applications can specify protocol parameters for the default association
2735 * initialization. The option name argument to setsockopt() and getsockopt()
2738 * Setting initialization parameters is effective only on an unconnected
2739 * socket (for UDP-style sockets only future associations are effected
2740 * by the change). With TCP-style sockets, this option is inherited by
2741 * sockets derived from a listener socket.
2743 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2745 struct sctp_initmsg sinit;
2746 struct sctp_sock *sp = sctp_sk(sk);
2748 if (optlen != sizeof(struct sctp_initmsg))
2750 if (copy_from_user(&sinit, optval, optlen))
2753 if (sinit.sinit_num_ostreams)
2754 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2755 if (sinit.sinit_max_instreams)
2756 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2757 if (sinit.sinit_max_attempts)
2758 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2759 if (sinit.sinit_max_init_timeo)
2760 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2766 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2768 * Applications that wish to use the sendto() system call may wish to
2769 * specify a default set of parameters that would normally be supplied
2770 * through the inclusion of ancillary data. This socket option allows
2771 * such an application to set the default sctp_sndrcvinfo structure.
2772 * The application that wishes to use this socket option simply passes
2773 * in to this call the sctp_sndrcvinfo structure defined in Section
2774 * 5.2.2) The input parameters accepted by this call include
2775 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2776 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2777 * to this call if the caller is using the UDP model.
2779 static int sctp_setsockopt_default_send_param(struct sock *sk,
2780 char __user *optval,
2781 unsigned int optlen)
2783 struct sctp_sock *sp = sctp_sk(sk);
2784 struct sctp_association *asoc;
2785 struct sctp_sndrcvinfo info;
2787 if (optlen != sizeof(info))
2789 if (copy_from_user(&info, optval, optlen))
2791 if (info.sinfo_flags &
2792 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2793 SCTP_ABORT | SCTP_EOF))
2796 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2797 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2800 asoc->default_stream = info.sinfo_stream;
2801 asoc->default_flags = info.sinfo_flags;
2802 asoc->default_ppid = info.sinfo_ppid;
2803 asoc->default_context = info.sinfo_context;
2804 asoc->default_timetolive = info.sinfo_timetolive;
2806 sp->default_stream = info.sinfo_stream;
2807 sp->default_flags = info.sinfo_flags;
2808 sp->default_ppid = info.sinfo_ppid;
2809 sp->default_context = info.sinfo_context;
2810 sp->default_timetolive = info.sinfo_timetolive;
2816 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2817 * (SCTP_DEFAULT_SNDINFO)
2819 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
2820 char __user *optval,
2821 unsigned int optlen)
2823 struct sctp_sock *sp = sctp_sk(sk);
2824 struct sctp_association *asoc;
2825 struct sctp_sndinfo info;
2827 if (optlen != sizeof(info))
2829 if (copy_from_user(&info, optval, optlen))
2831 if (info.snd_flags &
2832 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2833 SCTP_ABORT | SCTP_EOF))
2836 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
2837 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
2840 asoc->default_stream = info.snd_sid;
2841 asoc->default_flags = info.snd_flags;
2842 asoc->default_ppid = info.snd_ppid;
2843 asoc->default_context = info.snd_context;
2845 sp->default_stream = info.snd_sid;
2846 sp->default_flags = info.snd_flags;
2847 sp->default_ppid = info.snd_ppid;
2848 sp->default_context = info.snd_context;
2854 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2856 * Requests that the local SCTP stack use the enclosed peer address as
2857 * the association primary. The enclosed address must be one of the
2858 * association peer's addresses.
2860 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2861 unsigned int optlen)
2863 struct sctp_prim prim;
2864 struct sctp_transport *trans;
2866 if (optlen != sizeof(struct sctp_prim))
2869 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2872 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2876 sctp_assoc_set_primary(trans->asoc, trans);
2882 * 7.1.5 SCTP_NODELAY
2884 * Turn on/off any Nagle-like algorithm. This means that packets are
2885 * generally sent as soon as possible and no unnecessary delays are
2886 * introduced, at the cost of more packets in the network. Expects an
2887 * integer boolean flag.
2889 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2890 unsigned int optlen)
2894 if (optlen < sizeof(int))
2896 if (get_user(val, (int __user *)optval))
2899 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2905 * 7.1.1 SCTP_RTOINFO
2907 * The protocol parameters used to initialize and bound retransmission
2908 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2909 * and modify these parameters.
2910 * All parameters are time values, in milliseconds. A value of 0, when
2911 * modifying the parameters, indicates that the current value should not
2915 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2917 struct sctp_rtoinfo rtoinfo;
2918 struct sctp_association *asoc;
2919 unsigned long rto_min, rto_max;
2920 struct sctp_sock *sp = sctp_sk(sk);
2922 if (optlen != sizeof (struct sctp_rtoinfo))
2925 if (copy_from_user(&rtoinfo, optval, optlen))
2928 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2930 /* Set the values to the specific association */
2931 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2934 rto_max = rtoinfo.srto_max;
2935 rto_min = rtoinfo.srto_min;
2938 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2940 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2943 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2945 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2947 if (rto_min > rto_max)
2951 if (rtoinfo.srto_initial != 0)
2953 msecs_to_jiffies(rtoinfo.srto_initial);
2954 asoc->rto_max = rto_max;
2955 asoc->rto_min = rto_min;
2957 /* If there is no association or the association-id = 0
2958 * set the values to the endpoint.
2960 if (rtoinfo.srto_initial != 0)
2961 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2962 sp->rtoinfo.srto_max = rto_max;
2963 sp->rtoinfo.srto_min = rto_min;
2971 * 7.1.2 SCTP_ASSOCINFO
2973 * This option is used to tune the maximum retransmission attempts
2974 * of the association.
2975 * Returns an error if the new association retransmission value is
2976 * greater than the sum of the retransmission value of the peer.
2977 * See [SCTP] for more information.
2980 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2983 struct sctp_assocparams assocparams;
2984 struct sctp_association *asoc;
2986 if (optlen != sizeof(struct sctp_assocparams))
2988 if (copy_from_user(&assocparams, optval, optlen))
2991 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2993 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2996 /* Set the values to the specific association */
2998 if (assocparams.sasoc_asocmaxrxt != 0) {
3001 struct sctp_transport *peer_addr;
3003 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
3005 path_sum += peer_addr->pathmaxrxt;
3009 /* Only validate asocmaxrxt if we have more than
3010 * one path/transport. We do this because path
3011 * retransmissions are only counted when we have more
3015 assocparams.sasoc_asocmaxrxt > path_sum)
3018 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3021 if (assocparams.sasoc_cookie_life != 0)
3022 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3024 /* Set the values to the endpoint */
3025 struct sctp_sock *sp = sctp_sk(sk);
3027 if (assocparams.sasoc_asocmaxrxt != 0)
3028 sp->assocparams.sasoc_asocmaxrxt =
3029 assocparams.sasoc_asocmaxrxt;
3030 if (assocparams.sasoc_cookie_life != 0)
3031 sp->assocparams.sasoc_cookie_life =
3032 assocparams.sasoc_cookie_life;
3038 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3040 * This socket option is a boolean flag which turns on or off mapped V4
3041 * addresses. If this option is turned on and the socket is type
3042 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3043 * If this option is turned off, then no mapping will be done of V4
3044 * addresses and a user will receive both PF_INET6 and PF_INET type
3045 * addresses on the socket.
3047 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3050 struct sctp_sock *sp = sctp_sk(sk);
3052 if (optlen < sizeof(int))
3054 if (get_user(val, (int __user *)optval))
3065 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3066 * This option will get or set the maximum size to put in any outgoing
3067 * SCTP DATA chunk. If a message is larger than this size it will be
3068 * fragmented by SCTP into the specified size. Note that the underlying
3069 * SCTP implementation may fragment into smaller sized chunks when the
3070 * PMTU of the underlying association is smaller than the value set by
3071 * the user. The default value for this option is '0' which indicates
3072 * the user is NOT limiting fragmentation and only the PMTU will effect
3073 * SCTP's choice of DATA chunk size. Note also that values set larger
3074 * than the maximum size of an IP datagram will effectively let SCTP
3075 * control fragmentation (i.e. the same as setting this option to 0).
3077 * The following structure is used to access and modify this parameter:
3079 * struct sctp_assoc_value {
3080 * sctp_assoc_t assoc_id;
3081 * uint32_t assoc_value;
3084 * assoc_id: This parameter is ignored for one-to-one style sockets.
3085 * For one-to-many style sockets this parameter indicates which
3086 * association the user is performing an action upon. Note that if
3087 * this field's value is zero then the endpoints default value is
3088 * changed (effecting future associations only).
3089 * assoc_value: This parameter specifies the maximum size in bytes.
3091 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3093 struct sctp_assoc_value params;
3094 struct sctp_association *asoc;
3095 struct sctp_sock *sp = sctp_sk(sk);
3098 if (optlen == sizeof(int)) {
3099 pr_warn_ratelimited(DEPRECATED
3101 "Use of int in maxseg socket option.\n"
3102 "Use struct sctp_assoc_value instead\n",
3103 current->comm, task_pid_nr(current));
3104 if (copy_from_user(&val, optval, optlen))
3106 params.assoc_id = 0;
3107 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3108 if (copy_from_user(¶ms, optval, optlen))
3110 val = params.assoc_value;
3114 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3117 asoc = sctp_id2assoc(sk, params.assoc_id);
3118 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3123 val = asoc->pathmtu;
3124 val -= sp->pf->af->net_header_len;
3125 val -= sizeof(struct sctphdr) +
3126 sizeof(struct sctp_data_chunk);
3128 asoc->user_frag = val;
3129 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3131 sp->user_frag = val;
3139 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3141 * Requests that the peer mark the enclosed address as the association
3142 * primary. The enclosed address must be one of the association's
3143 * locally bound addresses. The following structure is used to make a
3144 * set primary request:
3146 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3147 unsigned int optlen)
3149 struct net *net = sock_net(sk);
3150 struct sctp_sock *sp;
3151 struct sctp_association *asoc = NULL;
3152 struct sctp_setpeerprim prim;
3153 struct sctp_chunk *chunk;
3159 if (!net->sctp.addip_enable)
3162 if (optlen != sizeof(struct sctp_setpeerprim))
3165 if (copy_from_user(&prim, optval, optlen))
3168 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3172 if (!asoc->peer.asconf_capable)
3175 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3178 if (!sctp_state(asoc, ESTABLISHED))
3181 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3185 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3186 return -EADDRNOTAVAIL;
3188 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3189 return -EADDRNOTAVAIL;
3191 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3192 chunk = sctp_make_asconf_set_prim(asoc,
3193 (union sctp_addr *)&prim.sspp_addr);
3197 err = sctp_send_asconf(asoc, chunk);
3199 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3204 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3205 unsigned int optlen)
3207 struct sctp_setadaptation adaptation;
3209 if (optlen != sizeof(struct sctp_setadaptation))
3211 if (copy_from_user(&adaptation, optval, optlen))
3214 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3220 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3222 * The context field in the sctp_sndrcvinfo structure is normally only
3223 * used when a failed message is retrieved holding the value that was
3224 * sent down on the actual send call. This option allows the setting of
3225 * a default context on an association basis that will be received on
3226 * reading messages from the peer. This is especially helpful in the
3227 * one-2-many model for an application to keep some reference to an
3228 * internal state machine that is processing messages on the
3229 * association. Note that the setting of this value only effects
3230 * received messages from the peer and does not effect the value that is
3231 * saved with outbound messages.
3233 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3234 unsigned int optlen)
3236 struct sctp_assoc_value params;
3237 struct sctp_sock *sp;
3238 struct sctp_association *asoc;
3240 if (optlen != sizeof(struct sctp_assoc_value))
3242 if (copy_from_user(¶ms, optval, optlen))
3247 if (params.assoc_id != 0) {
3248 asoc = sctp_id2assoc(sk, params.assoc_id);
3251 asoc->default_rcv_context = params.assoc_value;
3253 sp->default_rcv_context = params.assoc_value;
3260 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3262 * This options will at a minimum specify if the implementation is doing
3263 * fragmented interleave. Fragmented interleave, for a one to many
3264 * socket, is when subsequent calls to receive a message may return
3265 * parts of messages from different associations. Some implementations
3266 * may allow you to turn this value on or off. If so, when turned off,
3267 * no fragment interleave will occur (which will cause a head of line
3268 * blocking amongst multiple associations sharing the same one to many
3269 * socket). When this option is turned on, then each receive call may
3270 * come from a different association (thus the user must receive data
3271 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3272 * association each receive belongs to.
3274 * This option takes a boolean value. A non-zero value indicates that
3275 * fragmented interleave is on. A value of zero indicates that
3276 * fragmented interleave is off.
3278 * Note that it is important that an implementation that allows this
3279 * option to be turned on, have it off by default. Otherwise an unaware
3280 * application using the one to many model may become confused and act
3283 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3284 char __user *optval,
3285 unsigned int optlen)
3289 if (optlen != sizeof(int))
3291 if (get_user(val, (int __user *)optval))
3294 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3300 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3301 * (SCTP_PARTIAL_DELIVERY_POINT)
3303 * This option will set or get the SCTP partial delivery point. This
3304 * point is the size of a message where the partial delivery API will be
3305 * invoked to help free up rwnd space for the peer. Setting this to a
3306 * lower value will cause partial deliveries to happen more often. The
3307 * calls argument is an integer that sets or gets the partial delivery
3308 * point. Note also that the call will fail if the user attempts to set
3309 * this value larger than the socket receive buffer size.
3311 * Note that any single message having a length smaller than or equal to
3312 * the SCTP partial delivery point will be delivered in one single read
3313 * call as long as the user provided buffer is large enough to hold the
3316 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3317 char __user *optval,
3318 unsigned int optlen)
3322 if (optlen != sizeof(u32))
3324 if (get_user(val, (int __user *)optval))
3327 /* Note: We double the receive buffer from what the user sets
3328 * it to be, also initial rwnd is based on rcvbuf/2.
3330 if (val > (sk->sk_rcvbuf >> 1))
3333 sctp_sk(sk)->pd_point = val;
3335 return 0; /* is this the right error code? */
3339 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3341 * This option will allow a user to change the maximum burst of packets
3342 * that can be emitted by this association. Note that the default value
3343 * is 4, and some implementations may restrict this setting so that it
3344 * can only be lowered.
3346 * NOTE: This text doesn't seem right. Do this on a socket basis with
3347 * future associations inheriting the socket value.
3349 static int sctp_setsockopt_maxburst(struct sock *sk,
3350 char __user *optval,
3351 unsigned int optlen)
3353 struct sctp_assoc_value params;
3354 struct sctp_sock *sp;
3355 struct sctp_association *asoc;
3359 if (optlen == sizeof(int)) {
3360 pr_warn_ratelimited(DEPRECATED
3362 "Use of int in max_burst socket option deprecated.\n"
3363 "Use struct sctp_assoc_value instead\n",
3364 current->comm, task_pid_nr(current));
3365 if (copy_from_user(&val, optval, optlen))
3367 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3368 if (copy_from_user(¶ms, optval, optlen))
3370 val = params.assoc_value;
3371 assoc_id = params.assoc_id;
3377 if (assoc_id != 0) {
3378 asoc = sctp_id2assoc(sk, assoc_id);
3381 asoc->max_burst = val;
3383 sp->max_burst = val;
3389 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3391 * This set option adds a chunk type that the user is requesting to be
3392 * received only in an authenticated way. Changes to the list of chunks
3393 * will only effect future associations on the socket.
3395 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3396 char __user *optval,
3397 unsigned int optlen)
3399 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3400 struct sctp_authchunk val;
3402 if (!ep->auth_enable)
3405 if (optlen != sizeof(struct sctp_authchunk))
3407 if (copy_from_user(&val, optval, optlen))
3410 switch (val.sauth_chunk) {
3412 case SCTP_CID_INIT_ACK:
3413 case SCTP_CID_SHUTDOWN_COMPLETE:
3418 /* add this chunk id to the endpoint */
3419 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3423 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3425 * This option gets or sets the list of HMAC algorithms that the local
3426 * endpoint requires the peer to use.
3428 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3429 char __user *optval,
3430 unsigned int optlen)
3432 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3433 struct sctp_hmacalgo *hmacs;
3437 if (!ep->auth_enable)
3440 if (optlen < sizeof(struct sctp_hmacalgo))
3443 hmacs = memdup_user(optval, optlen);
3445 return PTR_ERR(hmacs);
3447 idents = hmacs->shmac_num_idents;
3448 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3449 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3454 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3461 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3463 * This option will set a shared secret key which is used to build an
3464 * association shared key.
3466 static int sctp_setsockopt_auth_key(struct sock *sk,
3467 char __user *optval,
3468 unsigned int optlen)
3470 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3471 struct sctp_authkey *authkey;
3472 struct sctp_association *asoc;
3475 if (!ep->auth_enable)
3478 if (optlen <= sizeof(struct sctp_authkey))
3481 authkey = memdup_user(optval, optlen);
3482 if (IS_ERR(authkey))
3483 return PTR_ERR(authkey);
3485 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3490 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3491 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3496 ret = sctp_auth_set_key(ep, asoc, authkey);
3503 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3505 * This option will get or set the active shared key to be used to build
3506 * the association shared key.
3508 static int sctp_setsockopt_active_key(struct sock *sk,
3509 char __user *optval,
3510 unsigned int optlen)
3512 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3513 struct sctp_authkeyid val;
3514 struct sctp_association *asoc;
3516 if (!ep->auth_enable)
3519 if (optlen != sizeof(struct sctp_authkeyid))
3521 if (copy_from_user(&val, optval, optlen))
3524 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3525 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3528 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3532 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3534 * This set option will delete a shared secret key from use.
3536 static int sctp_setsockopt_del_key(struct sock *sk,
3537 char __user *optval,
3538 unsigned int optlen)
3540 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3541 struct sctp_authkeyid val;
3542 struct sctp_association *asoc;
3544 if (!ep->auth_enable)
3547 if (optlen != sizeof(struct sctp_authkeyid))
3549 if (copy_from_user(&val, optval, optlen))
3552 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3553 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3556 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3561 * 8.1.23 SCTP_AUTO_ASCONF
3563 * This option will enable or disable the use of the automatic generation of
3564 * ASCONF chunks to add and delete addresses to an existing association. Note
3565 * that this option has two caveats namely: a) it only affects sockets that
3566 * are bound to all addresses available to the SCTP stack, and b) the system
3567 * administrator may have an overriding control that turns the ASCONF feature
3568 * off no matter what setting the socket option may have.
3569 * This option expects an integer boolean flag, where a non-zero value turns on
3570 * the option, and a zero value turns off the option.
3571 * Note. In this implementation, socket operation overrides default parameter
3572 * being set by sysctl as well as FreeBSD implementation
3574 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3575 unsigned int optlen)
3578 struct sctp_sock *sp = sctp_sk(sk);
3580 if (optlen < sizeof(int))
3582 if (get_user(val, (int __user *)optval))
3584 if (!sctp_is_ep_boundall(sk) && val)
3586 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3589 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3590 if (val == 0 && sp->do_auto_asconf) {
3591 list_del(&sp->auto_asconf_list);
3592 sp->do_auto_asconf = 0;
3593 } else if (val && !sp->do_auto_asconf) {
3594 list_add_tail(&sp->auto_asconf_list,
3595 &sock_net(sk)->sctp.auto_asconf_splist);
3596 sp->do_auto_asconf = 1;
3598 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3603 * SCTP_PEER_ADDR_THLDS
3605 * This option allows us to alter the partially failed threshold for one or all
3606 * transports in an association. See Section 6.1 of:
3607 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3609 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3610 char __user *optval,
3611 unsigned int optlen)
3613 struct sctp_paddrthlds val;
3614 struct sctp_transport *trans;
3615 struct sctp_association *asoc;
3617 if (optlen < sizeof(struct sctp_paddrthlds))
3619 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3620 sizeof(struct sctp_paddrthlds)))
3624 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3625 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3628 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3630 if (val.spt_pathmaxrxt)
3631 trans->pathmaxrxt = val.spt_pathmaxrxt;
3632 trans->pf_retrans = val.spt_pathpfthld;
3635 if (val.spt_pathmaxrxt)
3636 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3637 asoc->pf_retrans = val.spt_pathpfthld;
3639 trans = sctp_addr_id2transport(sk, &val.spt_address,
3644 if (val.spt_pathmaxrxt)
3645 trans->pathmaxrxt = val.spt_pathmaxrxt;
3646 trans->pf_retrans = val.spt_pathpfthld;
3652 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3653 char __user *optval,
3654 unsigned int optlen)
3658 if (optlen < sizeof(int))
3660 if (get_user(val, (int __user *) optval))
3663 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3668 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3669 char __user *optval,
3670 unsigned int optlen)
3674 if (optlen < sizeof(int))
3676 if (get_user(val, (int __user *) optval))
3679 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3684 static int sctp_setsockopt_pr_supported(struct sock *sk,
3685 char __user *optval,
3686 unsigned int optlen)
3688 struct sctp_assoc_value params;
3689 struct sctp_association *asoc;
3690 int retval = -EINVAL;
3692 if (optlen != sizeof(params))
3695 if (copy_from_user(¶ms, optval, optlen)) {
3700 asoc = sctp_id2assoc(sk, params.assoc_id);
3702 asoc->prsctp_enable = !!params.assoc_value;
3703 } else if (!params.assoc_id) {
3704 struct sctp_sock *sp = sctp_sk(sk);
3706 sp->ep->prsctp_enable = !!params.assoc_value;
3717 static int sctp_setsockopt_default_prinfo(struct sock *sk,
3718 char __user *optval,
3719 unsigned int optlen)
3721 struct sctp_default_prinfo info;
3722 struct sctp_association *asoc;
3723 int retval = -EINVAL;
3725 if (optlen != sizeof(info))
3728 if (copy_from_user(&info, optval, sizeof(info))) {
3733 if (info.pr_policy & ~SCTP_PR_SCTP_MASK)
3736 if (info.pr_policy == SCTP_PR_SCTP_NONE)
3739 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
3741 SCTP_PR_SET_POLICY(asoc->default_flags, info.pr_policy);
3742 asoc->default_timetolive = info.pr_value;
3743 } else if (!info.pr_assoc_id) {
3744 struct sctp_sock *sp = sctp_sk(sk);
3746 SCTP_PR_SET_POLICY(sp->default_flags, info.pr_policy);
3747 sp->default_timetolive = info.pr_value;
3758 /* API 6.2 setsockopt(), getsockopt()
3760 * Applications use setsockopt() and getsockopt() to set or retrieve
3761 * socket options. Socket options are used to change the default
3762 * behavior of sockets calls. They are described in Section 7.
3766 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3767 * int __user *optlen);
3768 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3771 * sd - the socket descript.
3772 * level - set to IPPROTO_SCTP for all SCTP options.
3773 * optname - the option name.
3774 * optval - the buffer to store the value of the option.
3775 * optlen - the size of the buffer.
3777 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3778 char __user *optval, unsigned int optlen)
3782 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3784 /* I can hardly begin to describe how wrong this is. This is
3785 * so broken as to be worse than useless. The API draft
3786 * REALLY is NOT helpful here... I am not convinced that the
3787 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3788 * are at all well-founded.
3790 if (level != SOL_SCTP) {
3791 struct sctp_af *af = sctp_sk(sk)->pf->af;
3792 retval = af->setsockopt(sk, level, optname, optval, optlen);
3799 case SCTP_SOCKOPT_BINDX_ADD:
3800 /* 'optlen' is the size of the addresses buffer. */
3801 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3802 optlen, SCTP_BINDX_ADD_ADDR);
3805 case SCTP_SOCKOPT_BINDX_REM:
3806 /* 'optlen' is the size of the addresses buffer. */
3807 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3808 optlen, SCTP_BINDX_REM_ADDR);
3811 case SCTP_SOCKOPT_CONNECTX_OLD:
3812 /* 'optlen' is the size of the addresses buffer. */
3813 retval = sctp_setsockopt_connectx_old(sk,
3814 (struct sockaddr __user *)optval,
3818 case SCTP_SOCKOPT_CONNECTX:
3819 /* 'optlen' is the size of the addresses buffer. */
3820 retval = sctp_setsockopt_connectx(sk,
3821 (struct sockaddr __user *)optval,
3825 case SCTP_DISABLE_FRAGMENTS:
3826 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3830 retval = sctp_setsockopt_events(sk, optval, optlen);
3833 case SCTP_AUTOCLOSE:
3834 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3837 case SCTP_PEER_ADDR_PARAMS:
3838 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3841 case SCTP_DELAYED_SACK:
3842 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3844 case SCTP_PARTIAL_DELIVERY_POINT:
3845 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3849 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3851 case SCTP_DEFAULT_SEND_PARAM:
3852 retval = sctp_setsockopt_default_send_param(sk, optval,
3855 case SCTP_DEFAULT_SNDINFO:
3856 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
3858 case SCTP_PRIMARY_ADDR:
3859 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3861 case SCTP_SET_PEER_PRIMARY_ADDR:
3862 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3865 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3868 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3870 case SCTP_ASSOCINFO:
3871 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3873 case SCTP_I_WANT_MAPPED_V4_ADDR:
3874 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3877 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3879 case SCTP_ADAPTATION_LAYER:
3880 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3883 retval = sctp_setsockopt_context(sk, optval, optlen);
3885 case SCTP_FRAGMENT_INTERLEAVE:
3886 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3888 case SCTP_MAX_BURST:
3889 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3891 case SCTP_AUTH_CHUNK:
3892 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3894 case SCTP_HMAC_IDENT:
3895 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3898 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3900 case SCTP_AUTH_ACTIVE_KEY:
3901 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3903 case SCTP_AUTH_DELETE_KEY:
3904 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3906 case SCTP_AUTO_ASCONF:
3907 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3909 case SCTP_PEER_ADDR_THLDS:
3910 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3912 case SCTP_RECVRCVINFO:
3913 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
3915 case SCTP_RECVNXTINFO:
3916 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
3918 case SCTP_PR_SUPPORTED:
3919 retval = sctp_setsockopt_pr_supported(sk, optval, optlen);
3921 case SCTP_DEFAULT_PRINFO:
3922 retval = sctp_setsockopt_default_prinfo(sk, optval, optlen);
3925 retval = -ENOPROTOOPT;
3935 /* API 3.1.6 connect() - UDP Style Syntax
3937 * An application may use the connect() call in the UDP model to initiate an
3938 * association without sending data.
3942 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3944 * sd: the socket descriptor to have a new association added to.
3946 * nam: the address structure (either struct sockaddr_in or struct
3947 * sockaddr_in6 defined in RFC2553 [7]).
3949 * len: the size of the address.
3951 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3959 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
3962 /* Validate addr_len before calling common connect/connectx routine. */
3963 af = sctp_get_af_specific(addr->sa_family);
3964 if (!af || addr_len < af->sockaddr_len) {
3967 /* Pass correct addr len to common routine (so it knows there
3968 * is only one address being passed.
3970 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3977 /* FIXME: Write comments. */
3978 static int sctp_disconnect(struct sock *sk, int flags)
3980 return -EOPNOTSUPP; /* STUB */
3983 /* 4.1.4 accept() - TCP Style Syntax
3985 * Applications use accept() call to remove an established SCTP
3986 * association from the accept queue of the endpoint. A new socket
3987 * descriptor will be returned from accept() to represent the newly
3988 * formed association.
3990 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3992 struct sctp_sock *sp;
3993 struct sctp_endpoint *ep;
3994 struct sock *newsk = NULL;
3995 struct sctp_association *asoc;
4004 if (!sctp_style(sk, TCP)) {
4005 error = -EOPNOTSUPP;
4009 if (!sctp_sstate(sk, LISTENING)) {
4014 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
4016 error = sctp_wait_for_accept(sk, timeo);
4020 /* We treat the list of associations on the endpoint as the accept
4021 * queue and pick the first association on the list.
4023 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
4025 newsk = sp->pf->create_accept_sk(sk, asoc);
4031 /* Populate the fields of the newsk from the oldsk and migrate the
4032 * asoc to the newsk.
4034 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
4042 /* The SCTP ioctl handler. */
4043 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
4050 * SEQPACKET-style sockets in LISTENING state are valid, for
4051 * SCTP, so only discard TCP-style sockets in LISTENING state.
4053 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4058 struct sk_buff *skb;
4059 unsigned int amount = 0;
4061 skb = skb_peek(&sk->sk_receive_queue);
4064 * We will only return the amount of this packet since
4065 * that is all that will be read.
4069 rc = put_user(amount, (int __user *)arg);
4081 /* This is the function which gets called during socket creation to
4082 * initialized the SCTP-specific portion of the sock.
4083 * The sock structure should already be zero-filled memory.
4085 static int sctp_init_sock(struct sock *sk)
4087 struct net *net = sock_net(sk);
4088 struct sctp_sock *sp;
4090 pr_debug("%s: sk:%p\n", __func__, sk);
4094 /* Initialize the SCTP per socket area. */
4095 switch (sk->sk_type) {
4096 case SOCK_SEQPACKET:
4097 sp->type = SCTP_SOCKET_UDP;
4100 sp->type = SCTP_SOCKET_TCP;
4103 return -ESOCKTNOSUPPORT;
4106 sk->sk_gso_type = SKB_GSO_SCTP;
4108 /* Initialize default send parameters. These parameters can be
4109 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4111 sp->default_stream = 0;
4112 sp->default_ppid = 0;
4113 sp->default_flags = 0;
4114 sp->default_context = 0;
4115 sp->default_timetolive = 0;
4117 sp->default_rcv_context = 0;
4118 sp->max_burst = net->sctp.max_burst;
4120 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4122 /* Initialize default setup parameters. These parameters
4123 * can be modified with the SCTP_INITMSG socket option or
4124 * overridden by the SCTP_INIT CMSG.
4126 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4127 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4128 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4129 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4131 /* Initialize default RTO related parameters. These parameters can
4132 * be modified for with the SCTP_RTOINFO socket option.
4134 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4135 sp->rtoinfo.srto_max = net->sctp.rto_max;
4136 sp->rtoinfo.srto_min = net->sctp.rto_min;
4138 /* Initialize default association related parameters. These parameters
4139 * can be modified with the SCTP_ASSOCINFO socket option.
4141 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4142 sp->assocparams.sasoc_number_peer_destinations = 0;
4143 sp->assocparams.sasoc_peer_rwnd = 0;
4144 sp->assocparams.sasoc_local_rwnd = 0;
4145 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4147 /* Initialize default event subscriptions. By default, all the
4150 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4152 /* Default Peer Address Parameters. These defaults can
4153 * be modified via SCTP_PEER_ADDR_PARAMS
4155 sp->hbinterval = net->sctp.hb_interval;
4156 sp->pathmaxrxt = net->sctp.max_retrans_path;
4157 sp->pathmtu = 0; /* allow default discovery */
4158 sp->sackdelay = net->sctp.sack_timeout;
4160 sp->param_flags = SPP_HB_ENABLE |
4162 SPP_SACKDELAY_ENABLE;
4164 /* If enabled no SCTP message fragmentation will be performed.
4165 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4167 sp->disable_fragments = 0;
4169 /* Enable Nagle algorithm by default. */
4172 sp->recvrcvinfo = 0;
4173 sp->recvnxtinfo = 0;
4175 /* Enable by default. */
4178 /* Auto-close idle associations after the configured
4179 * number of seconds. A value of 0 disables this
4180 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4181 * for UDP-style sockets only.
4185 /* User specified fragmentation limit. */
4188 sp->adaptation_ind = 0;
4190 sp->pf = sctp_get_pf_specific(sk->sk_family);
4192 /* Control variables for partial data delivery. */
4193 atomic_set(&sp->pd_mode, 0);
4194 skb_queue_head_init(&sp->pd_lobby);
4195 sp->frag_interleave = 0;
4197 /* Create a per socket endpoint structure. Even if we
4198 * change the data structure relationships, this may still
4199 * be useful for storing pre-connect address information.
4201 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4207 sk->sk_destruct = sctp_destruct_sock;
4209 SCTP_DBG_OBJCNT_INC(sock);
4212 percpu_counter_inc(&sctp_sockets_allocated);
4213 sock_prot_inuse_add(net, sk->sk_prot, 1);
4215 /* Nothing can fail after this block, otherwise
4216 * sctp_destroy_sock() will be called without addr_wq_lock held
4218 if (net->sctp.default_auto_asconf) {
4219 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
4220 list_add_tail(&sp->auto_asconf_list,
4221 &net->sctp.auto_asconf_splist);
4222 sp->do_auto_asconf = 1;
4223 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4225 sp->do_auto_asconf = 0;
4233 /* Cleanup any SCTP per socket resources. Must be called with
4234 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4236 static void sctp_destroy_sock(struct sock *sk)
4238 struct sctp_sock *sp;
4240 pr_debug("%s: sk:%p\n", __func__, sk);
4242 /* Release our hold on the endpoint. */
4244 /* This could happen during socket init, thus we bail out
4245 * early, since the rest of the below is not setup either.
4250 if (sp->do_auto_asconf) {
4251 sp->do_auto_asconf = 0;
4252 list_del(&sp->auto_asconf_list);
4254 sctp_endpoint_free(sp->ep);
4256 percpu_counter_dec(&sctp_sockets_allocated);
4257 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4261 /* Triggered when there are no references on the socket anymore */
4262 static void sctp_destruct_sock(struct sock *sk)
4264 struct sctp_sock *sp = sctp_sk(sk);
4266 /* Free up the HMAC transform. */
4267 crypto_free_shash(sp->hmac);
4269 inet_sock_destruct(sk);
4272 /* API 4.1.7 shutdown() - TCP Style Syntax
4273 * int shutdown(int socket, int how);
4275 * sd - the socket descriptor of the association to be closed.
4276 * how - Specifies the type of shutdown. The values are
4279 * Disables further receive operations. No SCTP
4280 * protocol action is taken.
4282 * Disables further send operations, and initiates
4283 * the SCTP shutdown sequence.
4285 * Disables further send and receive operations
4286 * and initiates the SCTP shutdown sequence.
4288 static void sctp_shutdown(struct sock *sk, int how)
4290 struct net *net = sock_net(sk);
4291 struct sctp_endpoint *ep;
4293 if (!sctp_style(sk, TCP))
4296 ep = sctp_sk(sk)->ep;
4297 if (how & SEND_SHUTDOWN && !list_empty(&ep->asocs)) {
4298 struct sctp_association *asoc;
4300 sk->sk_state = SCTP_SS_CLOSING;
4301 asoc = list_entry(ep->asocs.next,
4302 struct sctp_association, asocs);
4303 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4307 int sctp_get_sctp_info(struct sock *sk, struct sctp_association *asoc,
4308 struct sctp_info *info)
4310 struct sctp_transport *prim;
4311 struct list_head *pos;
4314 memset(info, 0, sizeof(*info));
4316 struct sctp_sock *sp = sctp_sk(sk);
4318 info->sctpi_s_autoclose = sp->autoclose;
4319 info->sctpi_s_adaptation_ind = sp->adaptation_ind;
4320 info->sctpi_s_pd_point = sp->pd_point;
4321 info->sctpi_s_nodelay = sp->nodelay;
4322 info->sctpi_s_disable_fragments = sp->disable_fragments;
4323 info->sctpi_s_v4mapped = sp->v4mapped;
4324 info->sctpi_s_frag_interleave = sp->frag_interleave;
4325 info->sctpi_s_type = sp->type;
4330 info->sctpi_tag = asoc->c.my_vtag;
4331 info->sctpi_state = asoc->state;
4332 info->sctpi_rwnd = asoc->a_rwnd;
4333 info->sctpi_unackdata = asoc->unack_data;
4334 info->sctpi_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4335 info->sctpi_instrms = asoc->c.sinit_max_instreams;
4336 info->sctpi_outstrms = asoc->c.sinit_num_ostreams;
4337 list_for_each(pos, &asoc->base.inqueue.in_chunk_list)
4338 info->sctpi_inqueue++;
4339 list_for_each(pos, &asoc->outqueue.out_chunk_list)
4340 info->sctpi_outqueue++;
4341 info->sctpi_overall_error = asoc->overall_error_count;
4342 info->sctpi_max_burst = asoc->max_burst;
4343 info->sctpi_maxseg = asoc->frag_point;
4344 info->sctpi_peer_rwnd = asoc->peer.rwnd;
4345 info->sctpi_peer_tag = asoc->c.peer_vtag;
4347 mask = asoc->peer.ecn_capable << 1;
4348 mask = (mask | asoc->peer.ipv4_address) << 1;
4349 mask = (mask | asoc->peer.ipv6_address) << 1;
4350 mask = (mask | asoc->peer.hostname_address) << 1;
4351 mask = (mask | asoc->peer.asconf_capable) << 1;
4352 mask = (mask | asoc->peer.prsctp_capable) << 1;
4353 mask = (mask | asoc->peer.auth_capable);
4354 info->sctpi_peer_capable = mask;
4355 mask = asoc->peer.sack_needed << 1;
4356 mask = (mask | asoc->peer.sack_generation) << 1;
4357 mask = (mask | asoc->peer.zero_window_announced);
4358 info->sctpi_peer_sack = mask;
4360 info->sctpi_isacks = asoc->stats.isacks;
4361 info->sctpi_osacks = asoc->stats.osacks;
4362 info->sctpi_opackets = asoc->stats.opackets;
4363 info->sctpi_ipackets = asoc->stats.ipackets;
4364 info->sctpi_rtxchunks = asoc->stats.rtxchunks;
4365 info->sctpi_outofseqtsns = asoc->stats.outofseqtsns;
4366 info->sctpi_idupchunks = asoc->stats.idupchunks;
4367 info->sctpi_gapcnt = asoc->stats.gapcnt;
4368 info->sctpi_ouodchunks = asoc->stats.ouodchunks;
4369 info->sctpi_iuodchunks = asoc->stats.iuodchunks;
4370 info->sctpi_oodchunks = asoc->stats.oodchunks;
4371 info->sctpi_iodchunks = asoc->stats.iodchunks;
4372 info->sctpi_octrlchunks = asoc->stats.octrlchunks;
4373 info->sctpi_ictrlchunks = asoc->stats.ictrlchunks;
4375 prim = asoc->peer.primary_path;
4376 memcpy(&info->sctpi_p_address, &prim->ipaddr,
4377 sizeof(struct sockaddr_storage));
4378 info->sctpi_p_state = prim->state;
4379 info->sctpi_p_cwnd = prim->cwnd;
4380 info->sctpi_p_srtt = prim->srtt;
4381 info->sctpi_p_rto = jiffies_to_msecs(prim->rto);
4382 info->sctpi_p_hbinterval = prim->hbinterval;
4383 info->sctpi_p_pathmaxrxt = prim->pathmaxrxt;
4384 info->sctpi_p_sackdelay = jiffies_to_msecs(prim->sackdelay);
4385 info->sctpi_p_ssthresh = prim->ssthresh;
4386 info->sctpi_p_partial_bytes_acked = prim->partial_bytes_acked;
4387 info->sctpi_p_flight_size = prim->flight_size;
4388 info->sctpi_p_error = prim->error_count;
4392 EXPORT_SYMBOL_GPL(sctp_get_sctp_info);
4394 /* use callback to avoid exporting the core structure */
4395 int sctp_transport_walk_start(struct rhashtable_iter *iter)
4399 rhltable_walk_enter(&sctp_transport_hashtable, iter);
4401 err = rhashtable_walk_start(iter);
4402 if (err && err != -EAGAIN) {
4403 rhashtable_walk_stop(iter);
4404 rhashtable_walk_exit(iter);
4411 void sctp_transport_walk_stop(struct rhashtable_iter *iter)
4413 rhashtable_walk_stop(iter);
4414 rhashtable_walk_exit(iter);
4417 struct sctp_transport *sctp_transport_get_next(struct net *net,
4418 struct rhashtable_iter *iter)
4420 struct sctp_transport *t;
4422 t = rhashtable_walk_next(iter);
4423 for (; t; t = rhashtable_walk_next(iter)) {
4425 if (PTR_ERR(t) == -EAGAIN)
4430 if (net_eq(sock_net(t->asoc->base.sk), net) &&
4431 t->asoc->peer.primary_path == t)
4438 struct sctp_transport *sctp_transport_get_idx(struct net *net,
4439 struct rhashtable_iter *iter,
4442 void *obj = SEQ_START_TOKEN;
4444 while (pos && (obj = sctp_transport_get_next(net, iter)) &&
4451 int sctp_for_each_endpoint(int (*cb)(struct sctp_endpoint *, void *),
4455 struct sctp_ep_common *epb;
4456 struct sctp_hashbucket *head;
4458 for (head = sctp_ep_hashtable; hash < sctp_ep_hashsize;
4460 read_lock(&head->lock);
4461 sctp_for_each_hentry(epb, &head->chain) {
4462 err = cb(sctp_ep(epb), p);
4466 read_unlock(&head->lock);
4471 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint);
4473 int sctp_transport_lookup_process(int (*cb)(struct sctp_transport *, void *),
4475 const union sctp_addr *laddr,
4476 const union sctp_addr *paddr, void *p)
4478 struct sctp_transport *transport;
4482 transport = sctp_addrs_lookup_transport(net, laddr, paddr);
4487 err = cb(transport, p);
4488 sctp_transport_put(transport);
4492 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process);
4494 int sctp_for_each_transport(int (*cb)(struct sctp_transport *, void *),
4495 struct net *net, int pos, void *p) {
4496 struct rhashtable_iter hti;
4500 err = sctp_transport_walk_start(&hti);
4504 sctp_transport_get_idx(net, &hti, pos);
4505 obj = sctp_transport_get_next(net, &hti);
4506 for (; obj && !IS_ERR(obj); obj = sctp_transport_get_next(net, &hti)) {
4507 struct sctp_transport *transport = obj;
4509 if (!sctp_transport_hold(transport))
4511 err = cb(transport, p);
4512 sctp_transport_put(transport);
4516 sctp_transport_walk_stop(&hti);
4520 EXPORT_SYMBOL_GPL(sctp_for_each_transport);
4522 /* 7.2.1 Association Status (SCTP_STATUS)
4524 * Applications can retrieve current status information about an
4525 * association, including association state, peer receiver window size,
4526 * number of unacked data chunks, and number of data chunks pending
4527 * receipt. This information is read-only.
4529 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4530 char __user *optval,
4533 struct sctp_status status;
4534 struct sctp_association *asoc = NULL;
4535 struct sctp_transport *transport;
4536 sctp_assoc_t associd;
4539 if (len < sizeof(status)) {
4544 len = sizeof(status);
4545 if (copy_from_user(&status, optval, len)) {
4550 associd = status.sstat_assoc_id;
4551 asoc = sctp_id2assoc(sk, associd);
4557 transport = asoc->peer.primary_path;
4559 status.sstat_assoc_id = sctp_assoc2id(asoc);
4560 status.sstat_state = sctp_assoc_to_state(asoc);
4561 status.sstat_rwnd = asoc->peer.rwnd;
4562 status.sstat_unackdata = asoc->unack_data;
4564 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4565 status.sstat_instrms = asoc->c.sinit_max_instreams;
4566 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4567 status.sstat_fragmentation_point = asoc->frag_point;
4568 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4569 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4570 transport->af_specific->sockaddr_len);
4571 /* Map ipv4 address into v4-mapped-on-v6 address. */
4572 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
4573 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4574 status.sstat_primary.spinfo_state = transport->state;
4575 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4576 status.sstat_primary.spinfo_srtt = transport->srtt;
4577 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4578 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4580 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4581 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4583 if (put_user(len, optlen)) {
4588 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4589 __func__, len, status.sstat_state, status.sstat_rwnd,
4590 status.sstat_assoc_id);
4592 if (copy_to_user(optval, &status, len)) {
4602 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4604 * Applications can retrieve information about a specific peer address
4605 * of an association, including its reachability state, congestion
4606 * window, and retransmission timer values. This information is
4609 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4610 char __user *optval,
4613 struct sctp_paddrinfo pinfo;
4614 struct sctp_transport *transport;
4617 if (len < sizeof(pinfo)) {
4622 len = sizeof(pinfo);
4623 if (copy_from_user(&pinfo, optval, len)) {
4628 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4629 pinfo.spinfo_assoc_id);
4633 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4634 pinfo.spinfo_state = transport->state;
4635 pinfo.spinfo_cwnd = transport->cwnd;
4636 pinfo.spinfo_srtt = transport->srtt;
4637 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4638 pinfo.spinfo_mtu = transport->pathmtu;
4640 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4641 pinfo.spinfo_state = SCTP_ACTIVE;
4643 if (put_user(len, optlen)) {
4648 if (copy_to_user(optval, &pinfo, len)) {
4657 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4659 * This option is a on/off flag. If enabled no SCTP message
4660 * fragmentation will be performed. Instead if a message being sent
4661 * exceeds the current PMTU size, the message will NOT be sent and
4662 * instead a error will be indicated to the user.
4664 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4665 char __user *optval, int __user *optlen)
4669 if (len < sizeof(int))
4673 val = (sctp_sk(sk)->disable_fragments == 1);
4674 if (put_user(len, optlen))
4676 if (copy_to_user(optval, &val, len))
4681 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4683 * This socket option is used to specify various notifications and
4684 * ancillary data the user wishes to receive.
4686 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4691 if (len > sizeof(struct sctp_event_subscribe))
4692 len = sizeof(struct sctp_event_subscribe);
4693 if (put_user(len, optlen))
4695 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4700 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4702 * This socket option is applicable to the UDP-style socket only. When
4703 * set it will cause associations that are idle for more than the
4704 * specified number of seconds to automatically close. An association
4705 * being idle is defined an association that has NOT sent or received
4706 * user data. The special value of '0' indicates that no automatic
4707 * close of any associations should be performed. The option expects an
4708 * integer defining the number of seconds of idle time before an
4709 * association is closed.
4711 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4713 /* Applicable to UDP-style socket only */
4714 if (sctp_style(sk, TCP))
4716 if (len < sizeof(int))
4719 if (put_user(len, optlen))
4721 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4726 /* Helper routine to branch off an association to a new socket. */
4727 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4729 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4730 struct sctp_sock *sp = sctp_sk(sk);
4731 struct socket *sock;
4737 /* An association cannot be branched off from an already peeled-off
4738 * socket, nor is this supported for tcp style sockets.
4740 if (!sctp_style(sk, UDP))
4743 /* Create a new socket. */
4744 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4748 sctp_copy_sock(sock->sk, sk, asoc);
4750 /* Make peeled-off sockets more like 1-1 accepted sockets.
4751 * Set the daddr and initialize id to something more random
4753 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
4755 /* Populate the fields of the newsk from the oldsk and migrate the
4756 * asoc to the newsk.
4758 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4764 EXPORT_SYMBOL(sctp_do_peeloff);
4766 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4768 sctp_peeloff_arg_t peeloff;
4769 struct socket *newsock;
4770 struct file *newfile;
4773 if (len < sizeof(sctp_peeloff_arg_t))
4775 len = sizeof(sctp_peeloff_arg_t);
4776 if (copy_from_user(&peeloff, optval, len))
4779 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4783 /* Map the socket to an unused fd that can be returned to the user. */
4784 retval = get_unused_fd_flags(0);
4786 sock_release(newsock);
4790 newfile = sock_alloc_file(newsock, 0, NULL);
4791 if (IS_ERR(newfile)) {
4792 put_unused_fd(retval);
4793 sock_release(newsock);
4794 return PTR_ERR(newfile);
4797 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4800 /* Return the fd mapped to the new socket. */
4801 if (put_user(len, optlen)) {
4803 put_unused_fd(retval);
4806 peeloff.sd = retval;
4807 if (copy_to_user(optval, &peeloff, len)) {
4809 put_unused_fd(retval);
4812 fd_install(retval, newfile);
4817 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4819 * Applications can enable or disable heartbeats for any peer address of
4820 * an association, modify an address's heartbeat interval, force a
4821 * heartbeat to be sent immediately, and adjust the address's maximum
4822 * number of retransmissions sent before an address is considered
4823 * unreachable. The following structure is used to access and modify an
4824 * address's parameters:
4826 * struct sctp_paddrparams {
4827 * sctp_assoc_t spp_assoc_id;
4828 * struct sockaddr_storage spp_address;
4829 * uint32_t spp_hbinterval;
4830 * uint16_t spp_pathmaxrxt;
4831 * uint32_t spp_pathmtu;
4832 * uint32_t spp_sackdelay;
4833 * uint32_t spp_flags;
4836 * spp_assoc_id - (one-to-many style socket) This is filled in the
4837 * application, and identifies the association for
4839 * spp_address - This specifies which address is of interest.
4840 * spp_hbinterval - This contains the value of the heartbeat interval,
4841 * in milliseconds. If a value of zero
4842 * is present in this field then no changes are to
4843 * be made to this parameter.
4844 * spp_pathmaxrxt - This contains the maximum number of
4845 * retransmissions before this address shall be
4846 * considered unreachable. If a value of zero
4847 * is present in this field then no changes are to
4848 * be made to this parameter.
4849 * spp_pathmtu - When Path MTU discovery is disabled the value
4850 * specified here will be the "fixed" path mtu.
4851 * Note that if the spp_address field is empty
4852 * then all associations on this address will
4853 * have this fixed path mtu set upon them.
4855 * spp_sackdelay - When delayed sack is enabled, this value specifies
4856 * the number of milliseconds that sacks will be delayed
4857 * for. This value will apply to all addresses of an
4858 * association if the spp_address field is empty. Note
4859 * also, that if delayed sack is enabled and this
4860 * value is set to 0, no change is made to the last
4861 * recorded delayed sack timer value.
4863 * spp_flags - These flags are used to control various features
4864 * on an association. The flag field may contain
4865 * zero or more of the following options.
4867 * SPP_HB_ENABLE - Enable heartbeats on the
4868 * specified address. Note that if the address
4869 * field is empty all addresses for the association
4870 * have heartbeats enabled upon them.
4872 * SPP_HB_DISABLE - Disable heartbeats on the
4873 * speicifed address. Note that if the address
4874 * field is empty all addresses for the association
4875 * will have their heartbeats disabled. Note also
4876 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4877 * mutually exclusive, only one of these two should
4878 * be specified. Enabling both fields will have
4879 * undetermined results.
4881 * SPP_HB_DEMAND - Request a user initiated heartbeat
4882 * to be made immediately.
4884 * SPP_PMTUD_ENABLE - This field will enable PMTU
4885 * discovery upon the specified address. Note that
4886 * if the address feild is empty then all addresses
4887 * on the association are effected.
4889 * SPP_PMTUD_DISABLE - This field will disable PMTU
4890 * discovery upon the specified address. Note that
4891 * if the address feild is empty then all addresses
4892 * on the association are effected. Not also that
4893 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4894 * exclusive. Enabling both will have undetermined
4897 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4898 * on delayed sack. The time specified in spp_sackdelay
4899 * is used to specify the sack delay for this address. Note
4900 * that if spp_address is empty then all addresses will
4901 * enable delayed sack and take on the sack delay
4902 * value specified in spp_sackdelay.
4903 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4904 * off delayed sack. If the spp_address field is blank then
4905 * delayed sack is disabled for the entire association. Note
4906 * also that this field is mutually exclusive to
4907 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4910 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4911 char __user *optval, int __user *optlen)
4913 struct sctp_paddrparams params;
4914 struct sctp_transport *trans = NULL;
4915 struct sctp_association *asoc = NULL;
4916 struct sctp_sock *sp = sctp_sk(sk);
4918 if (len < sizeof(struct sctp_paddrparams))
4920 len = sizeof(struct sctp_paddrparams);
4921 if (copy_from_user(¶ms, optval, len))
4924 /* If an address other than INADDR_ANY is specified, and
4925 * no transport is found, then the request is invalid.
4927 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
4928 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4929 params.spp_assoc_id);
4931 pr_debug("%s: failed no transport\n", __func__);
4936 /* Get association, if assoc_id != 0 and the socket is a one
4937 * to many style socket, and an association was not found, then
4938 * the id was invalid.
4940 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4941 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4942 pr_debug("%s: failed no association\n", __func__);
4947 /* Fetch transport values. */
4948 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4949 params.spp_pathmtu = trans->pathmtu;
4950 params.spp_pathmaxrxt = trans->pathmaxrxt;
4951 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4953 /*draft-11 doesn't say what to return in spp_flags*/
4954 params.spp_flags = trans->param_flags;
4956 /* Fetch association values. */
4957 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4958 params.spp_pathmtu = asoc->pathmtu;
4959 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4960 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4962 /*draft-11 doesn't say what to return in spp_flags*/
4963 params.spp_flags = asoc->param_flags;
4965 /* Fetch socket values. */
4966 params.spp_hbinterval = sp->hbinterval;
4967 params.spp_pathmtu = sp->pathmtu;
4968 params.spp_sackdelay = sp->sackdelay;
4969 params.spp_pathmaxrxt = sp->pathmaxrxt;
4971 /*draft-11 doesn't say what to return in spp_flags*/
4972 params.spp_flags = sp->param_flags;
4975 if (copy_to_user(optval, ¶ms, len))
4978 if (put_user(len, optlen))
4985 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4987 * This option will effect the way delayed acks are performed. This
4988 * option allows you to get or set the delayed ack time, in
4989 * milliseconds. It also allows changing the delayed ack frequency.
4990 * Changing the frequency to 1 disables the delayed sack algorithm. If
4991 * the assoc_id is 0, then this sets or gets the endpoints default
4992 * values. If the assoc_id field is non-zero, then the set or get
4993 * effects the specified association for the one to many model (the
4994 * assoc_id field is ignored by the one to one model). Note that if
4995 * sack_delay or sack_freq are 0 when setting this option, then the
4996 * current values will remain unchanged.
4998 * struct sctp_sack_info {
4999 * sctp_assoc_t sack_assoc_id;
5000 * uint32_t sack_delay;
5001 * uint32_t sack_freq;
5004 * sack_assoc_id - This parameter, indicates which association the user
5005 * is performing an action upon. Note that if this field's value is
5006 * zero then the endpoints default value is changed (effecting future
5007 * associations only).
5009 * sack_delay - This parameter contains the number of milliseconds that
5010 * the user is requesting the delayed ACK timer be set to. Note that
5011 * this value is defined in the standard to be between 200 and 500
5014 * sack_freq - This parameter contains the number of packets that must
5015 * be received before a sack is sent without waiting for the delay
5016 * timer to expire. The default value for this is 2, setting this
5017 * value to 1 will disable the delayed sack algorithm.
5019 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
5020 char __user *optval,
5023 struct sctp_sack_info params;
5024 struct sctp_association *asoc = NULL;
5025 struct sctp_sock *sp = sctp_sk(sk);
5027 if (len >= sizeof(struct sctp_sack_info)) {
5028 len = sizeof(struct sctp_sack_info);
5030 if (copy_from_user(¶ms, optval, len))
5032 } else if (len == sizeof(struct sctp_assoc_value)) {
5033 pr_warn_ratelimited(DEPRECATED
5035 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
5036 "Use struct sctp_sack_info instead\n",
5037 current->comm, task_pid_nr(current));
5038 if (copy_from_user(¶ms, optval, len))
5043 /* Get association, if sack_assoc_id != 0 and the socket is a one
5044 * to many style socket, and an association was not found, then
5045 * the id was invalid.
5047 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
5048 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
5052 /* Fetch association values. */
5053 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
5054 params.sack_delay = jiffies_to_msecs(
5056 params.sack_freq = asoc->sackfreq;
5059 params.sack_delay = 0;
5060 params.sack_freq = 1;
5063 /* Fetch socket values. */
5064 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
5065 params.sack_delay = sp->sackdelay;
5066 params.sack_freq = sp->sackfreq;
5068 params.sack_delay = 0;
5069 params.sack_freq = 1;
5073 if (copy_to_user(optval, ¶ms, len))
5076 if (put_user(len, optlen))
5082 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5084 * Applications can specify protocol parameters for the default association
5085 * initialization. The option name argument to setsockopt() and getsockopt()
5088 * Setting initialization parameters is effective only on an unconnected
5089 * socket (for UDP-style sockets only future associations are effected
5090 * by the change). With TCP-style sockets, this option is inherited by
5091 * sockets derived from a listener socket.
5093 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
5095 if (len < sizeof(struct sctp_initmsg))
5097 len = sizeof(struct sctp_initmsg);
5098 if (put_user(len, optlen))
5100 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
5106 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
5107 char __user *optval, int __user *optlen)
5109 struct sctp_association *asoc;
5111 struct sctp_getaddrs getaddrs;
5112 struct sctp_transport *from;
5114 union sctp_addr temp;
5115 struct sctp_sock *sp = sctp_sk(sk);
5120 if (len < sizeof(struct sctp_getaddrs))
5123 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5126 /* For UDP-style sockets, id specifies the association to query. */
5127 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5131 to = optval + offsetof(struct sctp_getaddrs, addrs);
5132 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5134 list_for_each_entry(from, &asoc->peer.transport_addr_list,
5136 memcpy(&temp, &from->ipaddr, sizeof(temp));
5137 addrlen = sctp_get_pf_specific(sk->sk_family)
5138 ->addr_to_user(sp, &temp);
5139 if (space_left < addrlen)
5141 if (copy_to_user(to, &temp, addrlen))
5145 space_left -= addrlen;
5148 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
5150 bytes_copied = ((char __user *)to) - optval;
5151 if (put_user(bytes_copied, optlen))
5157 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
5158 size_t space_left, int *bytes_copied)
5160 struct sctp_sockaddr_entry *addr;
5161 union sctp_addr temp;
5164 struct net *net = sock_net(sk);
5167 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
5171 if ((PF_INET == sk->sk_family) &&
5172 (AF_INET6 == addr->a.sa.sa_family))
5174 if ((PF_INET6 == sk->sk_family) &&
5175 inet_v6_ipv6only(sk) &&
5176 (AF_INET == addr->a.sa.sa_family))
5178 memcpy(&temp, &addr->a, sizeof(temp));
5179 if (!temp.v4.sin_port)
5180 temp.v4.sin_port = htons(port);
5182 addrlen = sctp_get_pf_specific(sk->sk_family)
5183 ->addr_to_user(sctp_sk(sk), &temp);
5185 if (space_left < addrlen) {
5189 memcpy(to, &temp, addrlen);
5193 space_left -= addrlen;
5194 *bytes_copied += addrlen;
5202 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
5203 char __user *optval, int __user *optlen)
5205 struct sctp_bind_addr *bp;
5206 struct sctp_association *asoc;
5208 struct sctp_getaddrs getaddrs;
5209 struct sctp_sockaddr_entry *addr;
5211 union sctp_addr temp;
5212 struct sctp_sock *sp = sctp_sk(sk);
5216 int bytes_copied = 0;
5220 if (len < sizeof(struct sctp_getaddrs))
5223 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5227 * For UDP-style sockets, id specifies the association to query.
5228 * If the id field is set to the value '0' then the locally bound
5229 * addresses are returned without regard to any particular
5232 if (0 == getaddrs.assoc_id) {
5233 bp = &sctp_sk(sk)->ep->base.bind_addr;
5235 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5238 bp = &asoc->base.bind_addr;
5241 to = optval + offsetof(struct sctp_getaddrs, addrs);
5242 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5244 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN);
5248 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5249 * addresses from the global local address list.
5251 if (sctp_list_single_entry(&bp->address_list)) {
5252 addr = list_entry(bp->address_list.next,
5253 struct sctp_sockaddr_entry, list);
5254 if (sctp_is_any(sk, &addr->a)) {
5255 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
5256 space_left, &bytes_copied);
5266 /* Protection on the bound address list is not needed since
5267 * in the socket option context we hold a socket lock and
5268 * thus the bound address list can't change.
5270 list_for_each_entry(addr, &bp->address_list, list) {
5271 memcpy(&temp, &addr->a, sizeof(temp));
5272 addrlen = sctp_get_pf_specific(sk->sk_family)
5273 ->addr_to_user(sp, &temp);
5274 if (space_left < addrlen) {
5275 err = -ENOMEM; /*fixme: right error?*/
5278 memcpy(buf, &temp, addrlen);
5280 bytes_copied += addrlen;
5282 space_left -= addrlen;
5286 if (copy_to_user(to, addrs, bytes_copied)) {
5290 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
5294 if (put_user(bytes_copied, optlen))
5301 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
5303 * Requests that the local SCTP stack use the enclosed peer address as
5304 * the association primary. The enclosed address must be one of the
5305 * association peer's addresses.
5307 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
5308 char __user *optval, int __user *optlen)
5310 struct sctp_prim prim;
5311 struct sctp_association *asoc;
5312 struct sctp_sock *sp = sctp_sk(sk);
5314 if (len < sizeof(struct sctp_prim))
5317 len = sizeof(struct sctp_prim);
5319 if (copy_from_user(&prim, optval, len))
5322 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
5326 if (!asoc->peer.primary_path)
5329 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
5330 asoc->peer.primary_path->af_specific->sockaddr_len);
5332 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
5333 (union sctp_addr *)&prim.ssp_addr);
5335 if (put_user(len, optlen))
5337 if (copy_to_user(optval, &prim, len))
5344 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5346 * Requests that the local endpoint set the specified Adaptation Layer
5347 * Indication parameter for all future INIT and INIT-ACK exchanges.
5349 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
5350 char __user *optval, int __user *optlen)
5352 struct sctp_setadaptation adaptation;
5354 if (len < sizeof(struct sctp_setadaptation))
5357 len = sizeof(struct sctp_setadaptation);
5359 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
5361 if (put_user(len, optlen))
5363 if (copy_to_user(optval, &adaptation, len))
5371 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5373 * Applications that wish to use the sendto() system call may wish to
5374 * specify a default set of parameters that would normally be supplied
5375 * through the inclusion of ancillary data. This socket option allows
5376 * such an application to set the default sctp_sndrcvinfo structure.
5379 * The application that wishes to use this socket option simply passes
5380 * in to this call the sctp_sndrcvinfo structure defined in Section
5381 * 5.2.2) The input parameters accepted by this call include
5382 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5383 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5384 * to this call if the caller is using the UDP model.
5386 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5388 static int sctp_getsockopt_default_send_param(struct sock *sk,
5389 int len, char __user *optval,
5392 struct sctp_sock *sp = sctp_sk(sk);
5393 struct sctp_association *asoc;
5394 struct sctp_sndrcvinfo info;
5396 if (len < sizeof(info))
5401 if (copy_from_user(&info, optval, len))
5404 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
5405 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
5408 info.sinfo_stream = asoc->default_stream;
5409 info.sinfo_flags = asoc->default_flags;
5410 info.sinfo_ppid = asoc->default_ppid;
5411 info.sinfo_context = asoc->default_context;
5412 info.sinfo_timetolive = asoc->default_timetolive;
5414 info.sinfo_stream = sp->default_stream;
5415 info.sinfo_flags = sp->default_flags;
5416 info.sinfo_ppid = sp->default_ppid;
5417 info.sinfo_context = sp->default_context;
5418 info.sinfo_timetolive = sp->default_timetolive;
5421 if (put_user(len, optlen))
5423 if (copy_to_user(optval, &info, len))
5429 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5430 * (SCTP_DEFAULT_SNDINFO)
5432 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
5433 char __user *optval,
5436 struct sctp_sock *sp = sctp_sk(sk);
5437 struct sctp_association *asoc;
5438 struct sctp_sndinfo info;
5440 if (len < sizeof(info))
5445 if (copy_from_user(&info, optval, len))
5448 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
5449 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
5452 info.snd_sid = asoc->default_stream;
5453 info.snd_flags = asoc->default_flags;
5454 info.snd_ppid = asoc->default_ppid;
5455 info.snd_context = asoc->default_context;
5457 info.snd_sid = sp->default_stream;
5458 info.snd_flags = sp->default_flags;
5459 info.snd_ppid = sp->default_ppid;
5460 info.snd_context = sp->default_context;
5463 if (put_user(len, optlen))
5465 if (copy_to_user(optval, &info, len))
5473 * 7.1.5 SCTP_NODELAY
5475 * Turn on/off any Nagle-like algorithm. This means that packets are
5476 * generally sent as soon as possible and no unnecessary delays are
5477 * introduced, at the cost of more packets in the network. Expects an
5478 * integer boolean flag.
5481 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5482 char __user *optval, int __user *optlen)
5486 if (len < sizeof(int))
5490 val = (sctp_sk(sk)->nodelay == 1);
5491 if (put_user(len, optlen))
5493 if (copy_to_user(optval, &val, len))
5500 * 7.1.1 SCTP_RTOINFO
5502 * The protocol parameters used to initialize and bound retransmission
5503 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5504 * and modify these parameters.
5505 * All parameters are time values, in milliseconds. A value of 0, when
5506 * modifying the parameters, indicates that the current value should not
5510 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5511 char __user *optval,
5512 int __user *optlen) {
5513 struct sctp_rtoinfo rtoinfo;
5514 struct sctp_association *asoc;
5516 if (len < sizeof (struct sctp_rtoinfo))
5519 len = sizeof(struct sctp_rtoinfo);
5521 if (copy_from_user(&rtoinfo, optval, len))
5524 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5526 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5529 /* Values corresponding to the specific association. */
5531 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5532 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5533 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5535 /* Values corresponding to the endpoint. */
5536 struct sctp_sock *sp = sctp_sk(sk);
5538 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5539 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5540 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5543 if (put_user(len, optlen))
5546 if (copy_to_user(optval, &rtoinfo, len))
5554 * 7.1.2 SCTP_ASSOCINFO
5556 * This option is used to tune the maximum retransmission attempts
5557 * of the association.
5558 * Returns an error if the new association retransmission value is
5559 * greater than the sum of the retransmission value of the peer.
5560 * See [SCTP] for more information.
5563 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5564 char __user *optval,
5568 struct sctp_assocparams assocparams;
5569 struct sctp_association *asoc;
5570 struct list_head *pos;
5573 if (len < sizeof (struct sctp_assocparams))
5576 len = sizeof(struct sctp_assocparams);
5578 if (copy_from_user(&assocparams, optval, len))
5581 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5583 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5586 /* Values correspoinding to the specific association */
5588 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5589 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5590 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5591 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5593 list_for_each(pos, &asoc->peer.transport_addr_list) {
5597 assocparams.sasoc_number_peer_destinations = cnt;
5599 /* Values corresponding to the endpoint */
5600 struct sctp_sock *sp = sctp_sk(sk);
5602 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5603 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5604 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5605 assocparams.sasoc_cookie_life =
5606 sp->assocparams.sasoc_cookie_life;
5607 assocparams.sasoc_number_peer_destinations =
5609 sasoc_number_peer_destinations;
5612 if (put_user(len, optlen))
5615 if (copy_to_user(optval, &assocparams, len))
5622 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5624 * This socket option is a boolean flag which turns on or off mapped V4
5625 * addresses. If this option is turned on and the socket is type
5626 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5627 * If this option is turned off, then no mapping will be done of V4
5628 * addresses and a user will receive both PF_INET6 and PF_INET type
5629 * addresses on the socket.
5631 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5632 char __user *optval, int __user *optlen)
5635 struct sctp_sock *sp = sctp_sk(sk);
5637 if (len < sizeof(int))
5642 if (put_user(len, optlen))
5644 if (copy_to_user(optval, &val, len))
5651 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5652 * (chapter and verse is quoted at sctp_setsockopt_context())
5654 static int sctp_getsockopt_context(struct sock *sk, int len,
5655 char __user *optval, int __user *optlen)
5657 struct sctp_assoc_value params;
5658 struct sctp_sock *sp;
5659 struct sctp_association *asoc;
5661 if (len < sizeof(struct sctp_assoc_value))
5664 len = sizeof(struct sctp_assoc_value);
5666 if (copy_from_user(¶ms, optval, len))
5671 if (params.assoc_id != 0) {
5672 asoc = sctp_id2assoc(sk, params.assoc_id);
5675 params.assoc_value = asoc->default_rcv_context;
5677 params.assoc_value = sp->default_rcv_context;
5680 if (put_user(len, optlen))
5682 if (copy_to_user(optval, ¶ms, len))
5689 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5690 * This option will get or set the maximum size to put in any outgoing
5691 * SCTP DATA chunk. If a message is larger than this size it will be
5692 * fragmented by SCTP into the specified size. Note that the underlying
5693 * SCTP implementation may fragment into smaller sized chunks when the
5694 * PMTU of the underlying association is smaller than the value set by
5695 * the user. The default value for this option is '0' which indicates
5696 * the user is NOT limiting fragmentation and only the PMTU will effect
5697 * SCTP's choice of DATA chunk size. Note also that values set larger
5698 * than the maximum size of an IP datagram will effectively let SCTP
5699 * control fragmentation (i.e. the same as setting this option to 0).
5701 * The following structure is used to access and modify this parameter:
5703 * struct sctp_assoc_value {
5704 * sctp_assoc_t assoc_id;
5705 * uint32_t assoc_value;
5708 * assoc_id: This parameter is ignored for one-to-one style sockets.
5709 * For one-to-many style sockets this parameter indicates which
5710 * association the user is performing an action upon. Note that if
5711 * this field's value is zero then the endpoints default value is
5712 * changed (effecting future associations only).
5713 * assoc_value: This parameter specifies the maximum size in bytes.
5715 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5716 char __user *optval, int __user *optlen)
5718 struct sctp_assoc_value params;
5719 struct sctp_association *asoc;
5721 if (len == sizeof(int)) {
5722 pr_warn_ratelimited(DEPRECATED
5724 "Use of int in maxseg socket option.\n"
5725 "Use struct sctp_assoc_value instead\n",
5726 current->comm, task_pid_nr(current));
5727 params.assoc_id = 0;
5728 } else if (len >= sizeof(struct sctp_assoc_value)) {
5729 len = sizeof(struct sctp_assoc_value);
5730 if (copy_from_user(¶ms, optval, sizeof(params)))
5735 asoc = sctp_id2assoc(sk, params.assoc_id);
5736 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5740 params.assoc_value = asoc->frag_point;
5742 params.assoc_value = sctp_sk(sk)->user_frag;
5744 if (put_user(len, optlen))
5746 if (len == sizeof(int)) {
5747 if (copy_to_user(optval, ¶ms.assoc_value, len))
5750 if (copy_to_user(optval, ¶ms, len))
5758 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5759 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5761 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5762 char __user *optval, int __user *optlen)
5766 if (len < sizeof(int))
5771 val = sctp_sk(sk)->frag_interleave;
5772 if (put_user(len, optlen))
5774 if (copy_to_user(optval, &val, len))
5781 * 7.1.25. Set or Get the sctp partial delivery point
5782 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5784 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5785 char __user *optval,
5790 if (len < sizeof(u32))
5795 val = sctp_sk(sk)->pd_point;
5796 if (put_user(len, optlen))
5798 if (copy_to_user(optval, &val, len))
5805 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5806 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5808 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5809 char __user *optval,
5812 struct sctp_assoc_value params;
5813 struct sctp_sock *sp;
5814 struct sctp_association *asoc;
5816 if (len == sizeof(int)) {
5817 pr_warn_ratelimited(DEPRECATED
5819 "Use of int in max_burst socket option.\n"
5820 "Use struct sctp_assoc_value instead\n",
5821 current->comm, task_pid_nr(current));
5822 params.assoc_id = 0;
5823 } else if (len >= sizeof(struct sctp_assoc_value)) {
5824 len = sizeof(struct sctp_assoc_value);
5825 if (copy_from_user(¶ms, optval, len))
5832 if (params.assoc_id != 0) {
5833 asoc = sctp_id2assoc(sk, params.assoc_id);
5836 params.assoc_value = asoc->max_burst;
5838 params.assoc_value = sp->max_burst;
5840 if (len == sizeof(int)) {
5841 if (copy_to_user(optval, ¶ms.assoc_value, len))
5844 if (copy_to_user(optval, ¶ms, len))
5852 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5853 char __user *optval, int __user *optlen)
5855 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5856 struct sctp_hmacalgo __user *p = (void __user *)optval;
5857 struct sctp_hmac_algo_param *hmacs;
5862 if (!ep->auth_enable)
5865 hmacs = ep->auth_hmacs_list;
5866 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5868 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5871 len = sizeof(struct sctp_hmacalgo) + data_len;
5872 num_idents = data_len / sizeof(u16);
5874 if (put_user(len, optlen))
5876 if (put_user(num_idents, &p->shmac_num_idents))
5878 for (i = 0; i < num_idents; i++) {
5879 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
5881 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
5887 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5888 char __user *optval, int __user *optlen)
5890 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5891 struct sctp_authkeyid val;
5892 struct sctp_association *asoc;
5894 if (!ep->auth_enable)
5897 if (len < sizeof(struct sctp_authkeyid))
5899 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5902 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5903 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5907 val.scact_keynumber = asoc->active_key_id;
5909 val.scact_keynumber = ep->active_key_id;
5911 len = sizeof(struct sctp_authkeyid);
5912 if (put_user(len, optlen))
5914 if (copy_to_user(optval, &val, len))
5920 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5921 char __user *optval, int __user *optlen)
5923 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5924 struct sctp_authchunks __user *p = (void __user *)optval;
5925 struct sctp_authchunks val;
5926 struct sctp_association *asoc;
5927 struct sctp_chunks_param *ch;
5931 if (!ep->auth_enable)
5934 if (len < sizeof(struct sctp_authchunks))
5937 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5940 to = p->gauth_chunks;
5941 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5945 ch = asoc->peer.peer_chunks;
5949 /* See if the user provided enough room for all the data */
5950 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5951 if (len < num_chunks)
5954 if (copy_to_user(to, ch->chunks, num_chunks))
5957 len = sizeof(struct sctp_authchunks) + num_chunks;
5958 if (put_user(len, optlen))
5960 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5965 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5966 char __user *optval, int __user *optlen)
5968 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5969 struct sctp_authchunks __user *p = (void __user *)optval;
5970 struct sctp_authchunks val;
5971 struct sctp_association *asoc;
5972 struct sctp_chunks_param *ch;
5976 if (!ep->auth_enable)
5979 if (len < sizeof(struct sctp_authchunks))
5982 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5985 to = p->gauth_chunks;
5986 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5987 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5991 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
5993 ch = ep->auth_chunk_list;
5998 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5999 if (len < sizeof(struct sctp_authchunks) + num_chunks)
6002 if (copy_to_user(to, ch->chunks, num_chunks))
6005 len = sizeof(struct sctp_authchunks) + num_chunks;
6006 if (put_user(len, optlen))
6008 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6015 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
6016 * This option gets the current number of associations that are attached
6017 * to a one-to-many style socket. The option value is an uint32_t.
6019 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
6020 char __user *optval, int __user *optlen)
6022 struct sctp_sock *sp = sctp_sk(sk);
6023 struct sctp_association *asoc;
6026 if (sctp_style(sk, TCP))
6029 if (len < sizeof(u32))
6034 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6038 if (put_user(len, optlen))
6040 if (copy_to_user(optval, &val, len))
6047 * 8.1.23 SCTP_AUTO_ASCONF
6048 * See the corresponding setsockopt entry as description
6050 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
6051 char __user *optval, int __user *optlen)
6055 if (len < sizeof(int))
6059 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
6061 if (put_user(len, optlen))
6063 if (copy_to_user(optval, &val, len))
6069 * 8.2.6. Get the Current Identifiers of Associations
6070 * (SCTP_GET_ASSOC_ID_LIST)
6072 * This option gets the current list of SCTP association identifiers of
6073 * the SCTP associations handled by a one-to-many style socket.
6075 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
6076 char __user *optval, int __user *optlen)
6078 struct sctp_sock *sp = sctp_sk(sk);
6079 struct sctp_association *asoc;
6080 struct sctp_assoc_ids *ids;
6083 if (sctp_style(sk, TCP))
6086 if (len < sizeof(struct sctp_assoc_ids))
6089 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6093 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
6096 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
6098 ids = kmalloc(len, GFP_USER | __GFP_NOWARN);
6102 ids->gaids_number_of_ids = num;
6104 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6105 ids->gaids_assoc_id[num++] = asoc->assoc_id;
6108 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
6118 * SCTP_PEER_ADDR_THLDS
6120 * This option allows us to fetch the partially failed threshold for one or all
6121 * transports in an association. See Section 6.1 of:
6122 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6124 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
6125 char __user *optval,
6129 struct sctp_paddrthlds val;
6130 struct sctp_transport *trans;
6131 struct sctp_association *asoc;
6133 if (len < sizeof(struct sctp_paddrthlds))
6135 len = sizeof(struct sctp_paddrthlds);
6136 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
6139 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
6140 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
6144 val.spt_pathpfthld = asoc->pf_retrans;
6145 val.spt_pathmaxrxt = asoc->pathmaxrxt;
6147 trans = sctp_addr_id2transport(sk, &val.spt_address,
6152 val.spt_pathmaxrxt = trans->pathmaxrxt;
6153 val.spt_pathpfthld = trans->pf_retrans;
6156 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
6163 * SCTP_GET_ASSOC_STATS
6165 * This option retrieves local per endpoint statistics. It is modeled
6166 * after OpenSolaris' implementation
6168 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
6169 char __user *optval,
6172 struct sctp_assoc_stats sas;
6173 struct sctp_association *asoc = NULL;
6175 /* User must provide at least the assoc id */
6176 if (len < sizeof(sctp_assoc_t))
6179 /* Allow the struct to grow and fill in as much as possible */
6180 len = min_t(size_t, len, sizeof(sas));
6182 if (copy_from_user(&sas, optval, len))
6185 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
6189 sas.sas_rtxchunks = asoc->stats.rtxchunks;
6190 sas.sas_gapcnt = asoc->stats.gapcnt;
6191 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
6192 sas.sas_osacks = asoc->stats.osacks;
6193 sas.sas_isacks = asoc->stats.isacks;
6194 sas.sas_octrlchunks = asoc->stats.octrlchunks;
6195 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
6196 sas.sas_oodchunks = asoc->stats.oodchunks;
6197 sas.sas_iodchunks = asoc->stats.iodchunks;
6198 sas.sas_ouodchunks = asoc->stats.ouodchunks;
6199 sas.sas_iuodchunks = asoc->stats.iuodchunks;
6200 sas.sas_idupchunks = asoc->stats.idupchunks;
6201 sas.sas_opackets = asoc->stats.opackets;
6202 sas.sas_ipackets = asoc->stats.ipackets;
6204 /* New high max rto observed, will return 0 if not a single
6205 * RTO update took place. obs_rto_ipaddr will be bogus
6208 sas.sas_maxrto = asoc->stats.max_obs_rto;
6209 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
6210 sizeof(struct sockaddr_storage));
6212 /* Mark beginning of a new observation period */
6213 asoc->stats.max_obs_rto = asoc->rto_min;
6215 if (put_user(len, optlen))
6218 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
6220 if (copy_to_user(optval, &sas, len))
6226 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
6227 char __user *optval,
6232 if (len < sizeof(int))
6236 if (sctp_sk(sk)->recvrcvinfo)
6238 if (put_user(len, optlen))
6240 if (copy_to_user(optval, &val, len))
6246 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
6247 char __user *optval,
6252 if (len < sizeof(int))
6256 if (sctp_sk(sk)->recvnxtinfo)
6258 if (put_user(len, optlen))
6260 if (copy_to_user(optval, &val, len))
6266 static int sctp_getsockopt_pr_supported(struct sock *sk, int len,
6267 char __user *optval,
6270 struct sctp_assoc_value params;
6271 struct sctp_association *asoc;
6272 int retval = -EFAULT;
6274 if (len < sizeof(params)) {
6279 len = sizeof(params);
6280 if (copy_from_user(¶ms, optval, len))
6283 asoc = sctp_id2assoc(sk, params.assoc_id);
6285 params.assoc_value = asoc->prsctp_enable;
6286 } else if (!params.assoc_id) {
6287 struct sctp_sock *sp = sctp_sk(sk);
6289 params.assoc_value = sp->ep->prsctp_enable;
6295 if (put_user(len, optlen))
6298 if (copy_to_user(optval, ¶ms, len))
6307 static int sctp_getsockopt_default_prinfo(struct sock *sk, int len,
6308 char __user *optval,
6311 struct sctp_default_prinfo info;
6312 struct sctp_association *asoc;
6313 int retval = -EFAULT;
6315 if (len < sizeof(info)) {
6321 if (copy_from_user(&info, optval, len))
6324 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
6326 info.pr_policy = SCTP_PR_POLICY(asoc->default_flags);
6327 info.pr_value = asoc->default_timetolive;
6328 } else if (!info.pr_assoc_id) {
6329 struct sctp_sock *sp = sctp_sk(sk);
6331 info.pr_policy = SCTP_PR_POLICY(sp->default_flags);
6332 info.pr_value = sp->default_timetolive;
6338 if (put_user(len, optlen))
6341 if (copy_to_user(optval, &info, len))
6350 static int sctp_getsockopt_pr_assocstatus(struct sock *sk, int len,
6351 char __user *optval,
6354 struct sctp_prstatus params;
6355 struct sctp_association *asoc;
6357 int retval = -EINVAL;
6359 if (len < sizeof(params))
6362 len = sizeof(params);
6363 if (copy_from_user(¶ms, optval, len)) {
6368 policy = params.sprstat_policy;
6369 if (policy & ~SCTP_PR_SCTP_MASK)
6372 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
6376 if (policy == SCTP_PR_SCTP_NONE) {
6377 params.sprstat_abandoned_unsent = 0;
6378 params.sprstat_abandoned_sent = 0;
6379 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
6380 params.sprstat_abandoned_unsent +=
6381 asoc->abandoned_unsent[policy];
6382 params.sprstat_abandoned_sent +=
6383 asoc->abandoned_sent[policy];
6386 params.sprstat_abandoned_unsent =
6387 asoc->abandoned_unsent[__SCTP_PR_INDEX(policy)];
6388 params.sprstat_abandoned_sent =
6389 asoc->abandoned_sent[__SCTP_PR_INDEX(policy)];
6392 if (put_user(len, optlen)) {
6397 if (copy_to_user(optval, ¶ms, len)) {
6408 static int sctp_getsockopt(struct sock *sk, int level, int optname,
6409 char __user *optval, int __user *optlen)
6414 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
6416 /* I can hardly begin to describe how wrong this is. This is
6417 * so broken as to be worse than useless. The API draft
6418 * REALLY is NOT helpful here... I am not convinced that the
6419 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
6420 * are at all well-founded.
6422 if (level != SOL_SCTP) {
6423 struct sctp_af *af = sctp_sk(sk)->pf->af;
6425 retval = af->getsockopt(sk, level, optname, optval, optlen);
6429 if (get_user(len, optlen))
6439 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
6441 case SCTP_DISABLE_FRAGMENTS:
6442 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
6446 retval = sctp_getsockopt_events(sk, len, optval, optlen);
6448 case SCTP_AUTOCLOSE:
6449 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
6451 case SCTP_SOCKOPT_PEELOFF:
6452 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
6454 case SCTP_PEER_ADDR_PARAMS:
6455 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
6458 case SCTP_DELAYED_SACK:
6459 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
6463 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
6465 case SCTP_GET_PEER_ADDRS:
6466 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
6469 case SCTP_GET_LOCAL_ADDRS:
6470 retval = sctp_getsockopt_local_addrs(sk, len, optval,
6473 case SCTP_SOCKOPT_CONNECTX3:
6474 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
6476 case SCTP_DEFAULT_SEND_PARAM:
6477 retval = sctp_getsockopt_default_send_param(sk, len,
6480 case SCTP_DEFAULT_SNDINFO:
6481 retval = sctp_getsockopt_default_sndinfo(sk, len,
6484 case SCTP_PRIMARY_ADDR:
6485 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
6488 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
6491 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
6493 case SCTP_ASSOCINFO:
6494 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
6496 case SCTP_I_WANT_MAPPED_V4_ADDR:
6497 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
6500 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
6502 case SCTP_GET_PEER_ADDR_INFO:
6503 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
6506 case SCTP_ADAPTATION_LAYER:
6507 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
6511 retval = sctp_getsockopt_context(sk, len, optval, optlen);
6513 case SCTP_FRAGMENT_INTERLEAVE:
6514 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
6517 case SCTP_PARTIAL_DELIVERY_POINT:
6518 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
6521 case SCTP_MAX_BURST:
6522 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
6525 case SCTP_AUTH_CHUNK:
6526 case SCTP_AUTH_DELETE_KEY:
6527 retval = -EOPNOTSUPP;
6529 case SCTP_HMAC_IDENT:
6530 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
6532 case SCTP_AUTH_ACTIVE_KEY:
6533 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
6535 case SCTP_PEER_AUTH_CHUNKS:
6536 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
6539 case SCTP_LOCAL_AUTH_CHUNKS:
6540 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
6543 case SCTP_GET_ASSOC_NUMBER:
6544 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
6546 case SCTP_GET_ASSOC_ID_LIST:
6547 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
6549 case SCTP_AUTO_ASCONF:
6550 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
6552 case SCTP_PEER_ADDR_THLDS:
6553 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
6555 case SCTP_GET_ASSOC_STATS:
6556 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
6558 case SCTP_RECVRCVINFO:
6559 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
6561 case SCTP_RECVNXTINFO:
6562 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
6564 case SCTP_PR_SUPPORTED:
6565 retval = sctp_getsockopt_pr_supported(sk, len, optval, optlen);
6567 case SCTP_DEFAULT_PRINFO:
6568 retval = sctp_getsockopt_default_prinfo(sk, len, optval,
6571 case SCTP_PR_ASSOC_STATUS:
6572 retval = sctp_getsockopt_pr_assocstatus(sk, len, optval,
6576 retval = -ENOPROTOOPT;
6584 static int sctp_hash(struct sock *sk)
6590 static void sctp_unhash(struct sock *sk)
6595 /* Check if port is acceptable. Possibly find first available port.
6597 * The port hash table (contained in the 'global' SCTP protocol storage
6598 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
6599 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
6600 * list (the list number is the port number hashed out, so as you
6601 * would expect from a hash function, all the ports in a given list have
6602 * such a number that hashes out to the same list number; you were
6603 * expecting that, right?); so each list has a set of ports, with a
6604 * link to the socket (struct sock) that uses it, the port number and
6605 * a fastreuse flag (FIXME: NPI ipg).
6607 static struct sctp_bind_bucket *sctp_bucket_create(
6608 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
6610 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
6612 struct sctp_bind_hashbucket *head; /* hash list */
6613 struct sctp_bind_bucket *pp;
6614 unsigned short snum;
6617 snum = ntohs(addr->v4.sin_port);
6619 pr_debug("%s: begins, snum:%d\n", __func__, snum);
6624 /* Search for an available port. */
6625 int low, high, remaining, index;
6627 struct net *net = sock_net(sk);
6629 inet_get_local_port_range(net, &low, &high);
6630 remaining = (high - low) + 1;
6631 rover = prandom_u32() % remaining + low;
6635 if ((rover < low) || (rover > high))
6637 if (inet_is_local_reserved_port(net, rover))
6639 index = sctp_phashfn(sock_net(sk), rover);
6640 head = &sctp_port_hashtable[index];
6641 spin_lock(&head->lock);
6642 sctp_for_each_hentry(pp, &head->chain)
6643 if ((pp->port == rover) &&
6644 net_eq(sock_net(sk), pp->net))
6648 spin_unlock(&head->lock);
6649 } while (--remaining > 0);
6651 /* Exhausted local port range during search? */
6656 /* OK, here is the one we will use. HEAD (the port
6657 * hash table list entry) is non-NULL and we hold it's
6662 /* We are given an specific port number; we verify
6663 * that it is not being used. If it is used, we will
6664 * exahust the search in the hash list corresponding
6665 * to the port number (snum) - we detect that with the
6666 * port iterator, pp being NULL.
6668 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
6669 spin_lock(&head->lock);
6670 sctp_for_each_hentry(pp, &head->chain) {
6671 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
6678 if (!hlist_empty(&pp->owner)) {
6679 /* We had a port hash table hit - there is an
6680 * available port (pp != NULL) and it is being
6681 * used by other socket (pp->owner not empty); that other
6682 * socket is going to be sk2.
6684 int reuse = sk->sk_reuse;
6687 pr_debug("%s: found a possible match\n", __func__);
6689 if (pp->fastreuse && sk->sk_reuse &&
6690 sk->sk_state != SCTP_SS_LISTENING)
6693 /* Run through the list of sockets bound to the port
6694 * (pp->port) [via the pointers bind_next and
6695 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6696 * we get the endpoint they describe and run through
6697 * the endpoint's list of IP (v4 or v6) addresses,
6698 * comparing each of the addresses with the address of
6699 * the socket sk. If we find a match, then that means
6700 * that this port/socket (sk) combination are already
6703 sk_for_each_bound(sk2, &pp->owner) {
6704 struct sctp_endpoint *ep2;
6705 ep2 = sctp_sk(sk2)->ep;
6708 (reuse && sk2->sk_reuse &&
6709 sk2->sk_state != SCTP_SS_LISTENING))
6712 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
6713 sctp_sk(sk2), sctp_sk(sk))) {
6719 pr_debug("%s: found a match\n", __func__);
6722 /* If there was a hash table miss, create a new port. */
6724 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6727 /* In either case (hit or miss), make sure fastreuse is 1 only
6728 * if sk->sk_reuse is too (that is, if the caller requested
6729 * SO_REUSEADDR on this socket -sk-).
6731 if (hlist_empty(&pp->owner)) {
6732 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6736 } else if (pp->fastreuse &&
6737 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6740 /* We are set, so fill up all the data in the hash table
6741 * entry, tie the socket list information with the rest of the
6742 * sockets FIXME: Blurry, NPI (ipg).
6745 if (!sctp_sk(sk)->bind_hash) {
6746 inet_sk(sk)->inet_num = snum;
6747 sk_add_bind_node(sk, &pp->owner);
6748 sctp_sk(sk)->bind_hash = pp;
6753 spin_unlock(&head->lock);
6760 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6761 * port is requested.
6763 static int sctp_get_port(struct sock *sk, unsigned short snum)
6765 union sctp_addr addr;
6766 struct sctp_af *af = sctp_sk(sk)->pf->af;
6768 /* Set up a dummy address struct from the sk. */
6769 af->from_sk(&addr, sk);
6770 addr.v4.sin_port = htons(snum);
6772 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6773 return !!sctp_get_port_local(sk, &addr);
6777 * Move a socket to LISTENING state.
6779 static int sctp_listen_start(struct sock *sk, int backlog)
6781 struct sctp_sock *sp = sctp_sk(sk);
6782 struct sctp_endpoint *ep = sp->ep;
6783 struct crypto_shash *tfm = NULL;
6786 /* Allocate HMAC for generating cookie. */
6787 if (!sp->hmac && sp->sctp_hmac_alg) {
6788 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6789 tfm = crypto_alloc_shash(alg, 0, 0);
6791 net_info_ratelimited("failed to load transform for %s: %ld\n",
6792 sp->sctp_hmac_alg, PTR_ERR(tfm));
6795 sctp_sk(sk)->hmac = tfm;
6799 * If a bind() or sctp_bindx() is not called prior to a listen()
6800 * call that allows new associations to be accepted, the system
6801 * picks an ephemeral port and will choose an address set equivalent
6802 * to binding with a wildcard address.
6804 * This is not currently spelled out in the SCTP sockets
6805 * extensions draft, but follows the practice as seen in TCP
6809 sk->sk_state = SCTP_SS_LISTENING;
6810 if (!ep->base.bind_addr.port) {
6811 if (sctp_autobind(sk))
6814 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6815 sk->sk_state = SCTP_SS_CLOSED;
6820 sk->sk_max_ack_backlog = backlog;
6821 sctp_hash_endpoint(ep);
6826 * 4.1.3 / 5.1.3 listen()
6828 * By default, new associations are not accepted for UDP style sockets.
6829 * An application uses listen() to mark a socket as being able to
6830 * accept new associations.
6832 * On TCP style sockets, applications use listen() to ready the SCTP
6833 * endpoint for accepting inbound associations.
6835 * On both types of endpoints a backlog of '0' disables listening.
6837 * Move a socket to LISTENING state.
6839 int sctp_inet_listen(struct socket *sock, int backlog)
6841 struct sock *sk = sock->sk;
6842 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6845 if (unlikely(backlog < 0))
6850 /* Peeled-off sockets are not allowed to listen(). */
6851 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6854 if (sock->state != SS_UNCONNECTED)
6857 /* If backlog is zero, disable listening. */
6859 if (sctp_sstate(sk, CLOSED))
6863 sctp_unhash_endpoint(ep);
6864 sk->sk_state = SCTP_SS_CLOSED;
6866 sctp_sk(sk)->bind_hash->fastreuse = 1;
6870 /* If we are already listening, just update the backlog */
6871 if (sctp_sstate(sk, LISTENING))
6872 sk->sk_max_ack_backlog = backlog;
6874 err = sctp_listen_start(sk, backlog);
6886 * This function is done by modeling the current datagram_poll() and the
6887 * tcp_poll(). Note that, based on these implementations, we don't
6888 * lock the socket in this function, even though it seems that,
6889 * ideally, locking or some other mechanisms can be used to ensure
6890 * the integrity of the counters (sndbuf and wmem_alloc) used
6891 * in this place. We assume that we don't need locks either until proven
6894 * Another thing to note is that we include the Async I/O support
6895 * here, again, by modeling the current TCP/UDP code. We don't have
6896 * a good way to test with it yet.
6898 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6900 struct sock *sk = sock->sk;
6901 struct sctp_sock *sp = sctp_sk(sk);
6904 poll_wait(file, sk_sleep(sk), wait);
6906 sock_rps_record_flow(sk);
6908 /* A TCP-style listening socket becomes readable when the accept queue
6911 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6912 return (!list_empty(&sp->ep->asocs)) ?
6913 (POLLIN | POLLRDNORM) : 0;
6917 /* Is there any exceptional events? */
6918 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6920 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6921 if (sk->sk_shutdown & RCV_SHUTDOWN)
6922 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6923 if (sk->sk_shutdown == SHUTDOWN_MASK)
6926 /* Is it readable? Reconsider this code with TCP-style support. */
6927 if (!skb_queue_empty(&sk->sk_receive_queue))
6928 mask |= POLLIN | POLLRDNORM;
6930 /* The association is either gone or not ready. */
6931 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6934 /* Is it writable? */
6935 if (sctp_writeable(sk)) {
6936 mask |= POLLOUT | POLLWRNORM;
6938 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
6940 * Since the socket is not locked, the buffer
6941 * might be made available after the writeable check and
6942 * before the bit is set. This could cause a lost I/O
6943 * signal. tcp_poll() has a race breaker for this race
6944 * condition. Based on their implementation, we put
6945 * in the following code to cover it as well.
6947 if (sctp_writeable(sk))
6948 mask |= POLLOUT | POLLWRNORM;
6953 /********************************************************************
6954 * 2nd Level Abstractions
6955 ********************************************************************/
6957 static struct sctp_bind_bucket *sctp_bucket_create(
6958 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6960 struct sctp_bind_bucket *pp;
6962 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6964 SCTP_DBG_OBJCNT_INC(bind_bucket);
6967 INIT_HLIST_HEAD(&pp->owner);
6969 hlist_add_head(&pp->node, &head->chain);
6974 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6975 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6977 if (pp && hlist_empty(&pp->owner)) {
6978 __hlist_del(&pp->node);
6979 kmem_cache_free(sctp_bucket_cachep, pp);
6980 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6984 /* Release this socket's reference to a local port. */
6985 static inline void __sctp_put_port(struct sock *sk)
6987 struct sctp_bind_hashbucket *head =
6988 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6989 inet_sk(sk)->inet_num)];
6990 struct sctp_bind_bucket *pp;
6992 spin_lock(&head->lock);
6993 pp = sctp_sk(sk)->bind_hash;
6994 __sk_del_bind_node(sk);
6995 sctp_sk(sk)->bind_hash = NULL;
6996 inet_sk(sk)->inet_num = 0;
6997 sctp_bucket_destroy(pp);
6998 spin_unlock(&head->lock);
7001 void sctp_put_port(struct sock *sk)
7004 __sctp_put_port(sk);
7009 * The system picks an ephemeral port and choose an address set equivalent
7010 * to binding with a wildcard address.
7011 * One of those addresses will be the primary address for the association.
7012 * This automatically enables the multihoming capability of SCTP.
7014 static int sctp_autobind(struct sock *sk)
7016 union sctp_addr autoaddr;
7020 /* Initialize a local sockaddr structure to INADDR_ANY. */
7021 af = sctp_sk(sk)->pf->af;
7023 port = htons(inet_sk(sk)->inet_num);
7024 af->inaddr_any(&autoaddr, port);
7026 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
7029 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
7032 * 4.2 The cmsghdr Structure *
7034 * When ancillary data is sent or received, any number of ancillary data
7035 * objects can be specified by the msg_control and msg_controllen members of
7036 * the msghdr structure, because each object is preceded by
7037 * a cmsghdr structure defining the object's length (the cmsg_len member).
7038 * Historically Berkeley-derived implementations have passed only one object
7039 * at a time, but this API allows multiple objects to be
7040 * passed in a single call to sendmsg() or recvmsg(). The following example
7041 * shows two ancillary data objects in a control buffer.
7043 * |<--------------------------- msg_controllen -------------------------->|
7046 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
7048 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
7051 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
7053 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
7056 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7057 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
7059 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
7061 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7068 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
7070 struct cmsghdr *cmsg;
7071 struct msghdr *my_msg = (struct msghdr *)msg;
7073 for_each_cmsghdr(cmsg, my_msg) {
7074 if (!CMSG_OK(my_msg, cmsg))
7077 /* Should we parse this header or ignore? */
7078 if (cmsg->cmsg_level != IPPROTO_SCTP)
7081 /* Strictly check lengths following example in SCM code. */
7082 switch (cmsg->cmsg_type) {
7084 /* SCTP Socket API Extension
7085 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
7087 * This cmsghdr structure provides information for
7088 * initializing new SCTP associations with sendmsg().
7089 * The SCTP_INITMSG socket option uses this same data
7090 * structure. This structure is not used for
7093 * cmsg_level cmsg_type cmsg_data[]
7094 * ------------ ------------ ----------------------
7095 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
7097 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
7100 cmsgs->init = CMSG_DATA(cmsg);
7104 /* SCTP Socket API Extension
7105 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
7107 * This cmsghdr structure specifies SCTP options for
7108 * sendmsg() and describes SCTP header information
7109 * about a received message through recvmsg().
7111 * cmsg_level cmsg_type cmsg_data[]
7112 * ------------ ------------ ----------------------
7113 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
7115 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
7118 cmsgs->srinfo = CMSG_DATA(cmsg);
7120 if (cmsgs->srinfo->sinfo_flags &
7121 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7122 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7123 SCTP_ABORT | SCTP_EOF))
7128 /* SCTP Socket API Extension
7129 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
7131 * This cmsghdr structure specifies SCTP options for
7132 * sendmsg(). This structure and SCTP_RCVINFO replaces
7133 * SCTP_SNDRCV which has been deprecated.
7135 * cmsg_level cmsg_type cmsg_data[]
7136 * ------------ ------------ ---------------------
7137 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
7139 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
7142 cmsgs->sinfo = CMSG_DATA(cmsg);
7144 if (cmsgs->sinfo->snd_flags &
7145 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7146 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7147 SCTP_ABORT | SCTP_EOF))
7159 * Wait for a packet..
7160 * Note: This function is the same function as in core/datagram.c
7161 * with a few modifications to make lksctp work.
7163 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
7168 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7170 /* Socket errors? */
7171 error = sock_error(sk);
7175 if (!skb_queue_empty(&sk->sk_receive_queue))
7178 /* Socket shut down? */
7179 if (sk->sk_shutdown & RCV_SHUTDOWN)
7182 /* Sequenced packets can come disconnected. If so we report the
7187 /* Is there a good reason to think that we may receive some data? */
7188 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
7191 /* Handle signals. */
7192 if (signal_pending(current))
7195 /* Let another process have a go. Since we are going to sleep
7196 * anyway. Note: This may cause odd behaviors if the message
7197 * does not fit in the user's buffer, but this seems to be the
7198 * only way to honor MSG_DONTWAIT realistically.
7201 *timeo_p = schedule_timeout(*timeo_p);
7205 finish_wait(sk_sleep(sk), &wait);
7209 error = sock_intr_errno(*timeo_p);
7212 finish_wait(sk_sleep(sk), &wait);
7217 /* Receive a datagram.
7218 * Note: This is pretty much the same routine as in core/datagram.c
7219 * with a few changes to make lksctp work.
7221 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
7222 int noblock, int *err)
7225 struct sk_buff *skb;
7228 timeo = sock_rcvtimeo(sk, noblock);
7230 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
7231 MAX_SCHEDULE_TIMEOUT);
7234 /* Again only user level code calls this function,
7235 * so nothing interrupt level
7236 * will suddenly eat the receive_queue.
7238 * Look at current nfs client by the way...
7239 * However, this function was correct in any case. 8)
7241 if (flags & MSG_PEEK) {
7242 skb = skb_peek(&sk->sk_receive_queue);
7244 atomic_inc(&skb->users);
7246 skb = __skb_dequeue(&sk->sk_receive_queue);
7252 /* Caller is allowed not to check sk->sk_err before calling. */
7253 error = sock_error(sk);
7257 if (sk->sk_shutdown & RCV_SHUTDOWN)
7260 if (sk_can_busy_loop(sk) &&
7261 sk_busy_loop(sk, noblock))
7264 /* User doesn't want to wait. */
7268 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
7277 /* If sndbuf has changed, wake up per association sndbuf waiters. */
7278 static void __sctp_write_space(struct sctp_association *asoc)
7280 struct sock *sk = asoc->base.sk;
7282 if (sctp_wspace(asoc) <= 0)
7285 if (waitqueue_active(&asoc->wait))
7286 wake_up_interruptible(&asoc->wait);
7288 if (sctp_writeable(sk)) {
7289 struct socket_wq *wq;
7292 wq = rcu_dereference(sk->sk_wq);
7294 if (waitqueue_active(&wq->wait))
7295 wake_up_interruptible(&wq->wait);
7297 /* Note that we try to include the Async I/O support
7298 * here by modeling from the current TCP/UDP code.
7299 * We have not tested with it yet.
7301 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
7302 sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT);
7308 static void sctp_wake_up_waiters(struct sock *sk,
7309 struct sctp_association *asoc)
7311 struct sctp_association *tmp = asoc;
7313 /* We do accounting for the sndbuf space per association,
7314 * so we only need to wake our own association.
7316 if (asoc->ep->sndbuf_policy)
7317 return __sctp_write_space(asoc);
7319 /* If association goes down and is just flushing its
7320 * outq, then just normally notify others.
7322 if (asoc->base.dead)
7323 return sctp_write_space(sk);
7325 /* Accounting for the sndbuf space is per socket, so we
7326 * need to wake up others, try to be fair and in case of
7327 * other associations, let them have a go first instead
7328 * of just doing a sctp_write_space() call.
7330 * Note that we reach sctp_wake_up_waiters() only when
7331 * associations free up queued chunks, thus we are under
7332 * lock and the list of associations on a socket is
7333 * guaranteed not to change.
7335 for (tmp = list_next_entry(tmp, asocs); 1;
7336 tmp = list_next_entry(tmp, asocs)) {
7337 /* Manually skip the head element. */
7338 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
7340 /* Wake up association. */
7341 __sctp_write_space(tmp);
7342 /* We've reached the end. */
7348 /* Do accounting for the sndbuf space.
7349 * Decrement the used sndbuf space of the corresponding association by the
7350 * data size which was just transmitted(freed).
7352 static void sctp_wfree(struct sk_buff *skb)
7354 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
7355 struct sctp_association *asoc = chunk->asoc;
7356 struct sock *sk = asoc->base.sk;
7358 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
7359 sizeof(struct sk_buff) +
7360 sizeof(struct sctp_chunk);
7362 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
7365 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
7367 sk->sk_wmem_queued -= skb->truesize;
7368 sk_mem_uncharge(sk, skb->truesize);
7371 sctp_wake_up_waiters(sk, asoc);
7373 sctp_association_put(asoc);
7376 /* Do accounting for the receive space on the socket.
7377 * Accounting for the association is done in ulpevent.c
7378 * We set this as a destructor for the cloned data skbs so that
7379 * accounting is done at the correct time.
7381 void sctp_sock_rfree(struct sk_buff *skb)
7383 struct sock *sk = skb->sk;
7384 struct sctp_ulpevent *event = sctp_skb2event(skb);
7386 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
7389 * Mimic the behavior of sock_rfree
7391 sk_mem_uncharge(sk, event->rmem_len);
7395 /* Helper function to wait for space in the sndbuf. */
7396 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
7399 struct sock *sk = asoc->base.sk;
7401 long current_timeo = *timeo_p;
7404 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
7407 /* Increment the association's refcnt. */
7408 sctp_association_hold(asoc);
7410 /* Wait on the association specific sndbuf space. */
7412 prepare_to_wait_exclusive(&asoc->wait, &wait,
7413 TASK_INTERRUPTIBLE);
7416 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7419 if (signal_pending(current))
7420 goto do_interrupted;
7421 if (msg_len <= sctp_wspace(asoc))
7424 /* Let another process have a go. Since we are going
7428 current_timeo = schedule_timeout(current_timeo);
7429 if (sk != asoc->base.sk)
7433 *timeo_p = current_timeo;
7437 finish_wait(&asoc->wait, &wait);
7439 /* Release the association's refcnt. */
7440 sctp_association_put(asoc);
7449 err = sock_intr_errno(*timeo_p);
7457 void sctp_data_ready(struct sock *sk)
7459 struct socket_wq *wq;
7462 wq = rcu_dereference(sk->sk_wq);
7463 if (skwq_has_sleeper(wq))
7464 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
7465 POLLRDNORM | POLLRDBAND);
7466 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
7470 /* If socket sndbuf has changed, wake up all per association waiters. */
7471 void sctp_write_space(struct sock *sk)
7473 struct sctp_association *asoc;
7475 /* Wake up the tasks in each wait queue. */
7476 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
7477 __sctp_write_space(asoc);
7481 /* Is there any sndbuf space available on the socket?
7483 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
7484 * associations on the same socket. For a UDP-style socket with
7485 * multiple associations, it is possible for it to be "unwriteable"
7486 * prematurely. I assume that this is acceptable because
7487 * a premature "unwriteable" is better than an accidental "writeable" which
7488 * would cause an unwanted block under certain circumstances. For the 1-1
7489 * UDP-style sockets or TCP-style sockets, this code should work.
7492 static int sctp_writeable(struct sock *sk)
7496 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
7502 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
7503 * returns immediately with EINPROGRESS.
7505 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
7507 struct sock *sk = asoc->base.sk;
7509 long current_timeo = *timeo_p;
7512 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
7514 /* Increment the association's refcnt. */
7515 sctp_association_hold(asoc);
7518 prepare_to_wait_exclusive(&asoc->wait, &wait,
7519 TASK_INTERRUPTIBLE);
7522 if (sk->sk_shutdown & RCV_SHUTDOWN)
7524 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7527 if (signal_pending(current))
7528 goto do_interrupted;
7530 if (sctp_state(asoc, ESTABLISHED))
7533 /* Let another process have a go. Since we are going
7537 current_timeo = schedule_timeout(current_timeo);
7540 *timeo_p = current_timeo;
7544 finish_wait(&asoc->wait, &wait);
7546 /* Release the association's refcnt. */
7547 sctp_association_put(asoc);
7552 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
7555 err = -ECONNREFUSED;
7559 err = sock_intr_errno(*timeo_p);
7567 static int sctp_wait_for_accept(struct sock *sk, long timeo)
7569 struct sctp_endpoint *ep;
7573 ep = sctp_sk(sk)->ep;
7577 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
7578 TASK_INTERRUPTIBLE);
7580 if (list_empty(&ep->asocs)) {
7582 timeo = schedule_timeout(timeo);
7587 if (!sctp_sstate(sk, LISTENING))
7591 if (!list_empty(&ep->asocs))
7594 err = sock_intr_errno(timeo);
7595 if (signal_pending(current))
7603 finish_wait(sk_sleep(sk), &wait);
7608 static void sctp_wait_for_close(struct sock *sk, long timeout)
7613 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7614 if (list_empty(&sctp_sk(sk)->ep->asocs))
7617 timeout = schedule_timeout(timeout);
7619 } while (!signal_pending(current) && timeout);
7621 finish_wait(sk_sleep(sk), &wait);
7624 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
7626 struct sk_buff *frag;
7631 /* Don't forget the fragments. */
7632 skb_walk_frags(skb, frag)
7633 sctp_skb_set_owner_r_frag(frag, sk);
7636 sctp_skb_set_owner_r(skb, sk);
7639 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
7640 struct sctp_association *asoc)
7642 struct inet_sock *inet = inet_sk(sk);
7643 struct inet_sock *newinet;
7645 newsk->sk_type = sk->sk_type;
7646 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
7647 newsk->sk_flags = sk->sk_flags;
7648 newsk->sk_tsflags = sk->sk_tsflags;
7649 newsk->sk_no_check_tx = sk->sk_no_check_tx;
7650 newsk->sk_no_check_rx = sk->sk_no_check_rx;
7651 newsk->sk_reuse = sk->sk_reuse;
7653 newsk->sk_shutdown = sk->sk_shutdown;
7654 newsk->sk_destruct = sctp_destruct_sock;
7655 newsk->sk_family = sk->sk_family;
7656 newsk->sk_protocol = IPPROTO_SCTP;
7657 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
7658 newsk->sk_sndbuf = sk->sk_sndbuf;
7659 newsk->sk_rcvbuf = sk->sk_rcvbuf;
7660 newsk->sk_lingertime = sk->sk_lingertime;
7661 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
7662 newsk->sk_sndtimeo = sk->sk_sndtimeo;
7663 newsk->sk_rxhash = sk->sk_rxhash;
7665 newinet = inet_sk(newsk);
7667 /* Initialize sk's sport, dport, rcv_saddr and daddr for
7668 * getsockname() and getpeername()
7670 newinet->inet_sport = inet->inet_sport;
7671 newinet->inet_saddr = inet->inet_saddr;
7672 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
7673 newinet->inet_dport = htons(asoc->peer.port);
7674 newinet->pmtudisc = inet->pmtudisc;
7675 newinet->inet_id = asoc->next_tsn ^ jiffies;
7677 newinet->uc_ttl = inet->uc_ttl;
7678 newinet->mc_loop = 1;
7679 newinet->mc_ttl = 1;
7680 newinet->mc_index = 0;
7681 newinet->mc_list = NULL;
7683 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
7684 net_enable_timestamp();
7686 security_sk_clone(sk, newsk);
7689 static inline void sctp_copy_descendant(struct sock *sk_to,
7690 const struct sock *sk_from)
7692 int ancestor_size = sizeof(struct inet_sock) +
7693 sizeof(struct sctp_sock) -
7694 offsetof(struct sctp_sock, auto_asconf_list);
7696 if (sk_from->sk_family == PF_INET6)
7697 ancestor_size += sizeof(struct ipv6_pinfo);
7699 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
7702 /* Populate the fields of the newsk from the oldsk and migrate the assoc
7703 * and its messages to the newsk.
7705 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
7706 struct sctp_association *assoc,
7707 sctp_socket_type_t type)
7709 struct sctp_sock *oldsp = sctp_sk(oldsk);
7710 struct sctp_sock *newsp = sctp_sk(newsk);
7711 struct sctp_bind_bucket *pp; /* hash list port iterator */
7712 struct sctp_endpoint *newep = newsp->ep;
7713 struct sk_buff *skb, *tmp;
7714 struct sctp_ulpevent *event;
7715 struct sctp_bind_hashbucket *head;
7717 /* Migrate socket buffer sizes and all the socket level options to the
7720 newsk->sk_sndbuf = oldsk->sk_sndbuf;
7721 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
7722 /* Brute force copy old sctp opt. */
7723 sctp_copy_descendant(newsk, oldsk);
7725 /* Restore the ep value that was overwritten with the above structure
7731 /* Hook this new socket in to the bind_hash list. */
7732 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
7733 inet_sk(oldsk)->inet_num)];
7734 spin_lock_bh(&head->lock);
7735 pp = sctp_sk(oldsk)->bind_hash;
7736 sk_add_bind_node(newsk, &pp->owner);
7737 sctp_sk(newsk)->bind_hash = pp;
7738 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
7739 spin_unlock_bh(&head->lock);
7741 /* Copy the bind_addr list from the original endpoint to the new
7742 * endpoint so that we can handle restarts properly
7744 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
7745 &oldsp->ep->base.bind_addr, GFP_KERNEL);
7747 /* Move any messages in the old socket's receive queue that are for the
7748 * peeled off association to the new socket's receive queue.
7750 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
7751 event = sctp_skb2event(skb);
7752 if (event->asoc == assoc) {
7753 __skb_unlink(skb, &oldsk->sk_receive_queue);
7754 __skb_queue_tail(&newsk->sk_receive_queue, skb);
7755 sctp_skb_set_owner_r_frag(skb, newsk);
7759 /* Clean up any messages pending delivery due to partial
7760 * delivery. Three cases:
7761 * 1) No partial deliver; no work.
7762 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7763 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7765 skb_queue_head_init(&newsp->pd_lobby);
7766 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
7768 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
7769 struct sk_buff_head *queue;
7771 /* Decide which queue to move pd_lobby skbs to. */
7772 if (assoc->ulpq.pd_mode) {
7773 queue = &newsp->pd_lobby;
7775 queue = &newsk->sk_receive_queue;
7777 /* Walk through the pd_lobby, looking for skbs that
7778 * need moved to the new socket.
7780 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
7781 event = sctp_skb2event(skb);
7782 if (event->asoc == assoc) {
7783 __skb_unlink(skb, &oldsp->pd_lobby);
7784 __skb_queue_tail(queue, skb);
7785 sctp_skb_set_owner_r_frag(skb, newsk);
7789 /* Clear up any skbs waiting for the partial
7790 * delivery to finish.
7792 if (assoc->ulpq.pd_mode)
7793 sctp_clear_pd(oldsk, NULL);
7797 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7798 sctp_skb_set_owner_r_frag(skb, newsk);
7800 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7801 sctp_skb_set_owner_r_frag(skb, newsk);
7803 /* Set the type of socket to indicate that it is peeled off from the
7804 * original UDP-style socket or created with the accept() call on a
7805 * TCP-style socket..
7809 /* Mark the new socket "in-use" by the user so that any packets
7810 * that may arrive on the association after we've moved it are
7811 * queued to the backlog. This prevents a potential race between
7812 * backlog processing on the old socket and new-packet processing
7813 * on the new socket.
7815 * The caller has just allocated newsk so we can guarantee that other
7816 * paths won't try to lock it and then oldsk.
7818 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7819 sctp_assoc_migrate(assoc, newsk);
7821 /* If the association on the newsk is already closed before accept()
7822 * is called, set RCV_SHUTDOWN flag.
7824 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP)) {
7825 newsk->sk_state = SCTP_SS_CLOSED;
7826 newsk->sk_shutdown |= RCV_SHUTDOWN;
7828 newsk->sk_state = SCTP_SS_ESTABLISHED;
7831 release_sock(newsk);
7835 /* This proto struct describes the ULP interface for SCTP. */
7836 struct proto sctp_prot = {
7838 .owner = THIS_MODULE,
7839 .close = sctp_close,
7840 .connect = sctp_connect,
7841 .disconnect = sctp_disconnect,
7842 .accept = sctp_accept,
7843 .ioctl = sctp_ioctl,
7844 .init = sctp_init_sock,
7845 .destroy = sctp_destroy_sock,
7846 .shutdown = sctp_shutdown,
7847 .setsockopt = sctp_setsockopt,
7848 .getsockopt = sctp_getsockopt,
7849 .sendmsg = sctp_sendmsg,
7850 .recvmsg = sctp_recvmsg,
7852 .backlog_rcv = sctp_backlog_rcv,
7854 .unhash = sctp_unhash,
7855 .get_port = sctp_get_port,
7856 .obj_size = sizeof(struct sctp_sock),
7857 .sysctl_mem = sysctl_sctp_mem,
7858 .sysctl_rmem = sysctl_sctp_rmem,
7859 .sysctl_wmem = sysctl_sctp_wmem,
7860 .memory_pressure = &sctp_memory_pressure,
7861 .enter_memory_pressure = sctp_enter_memory_pressure,
7862 .memory_allocated = &sctp_memory_allocated,
7863 .sockets_allocated = &sctp_sockets_allocated,
7866 #if IS_ENABLED(CONFIG_IPV6)
7868 #include <net/transp_v6.h>
7869 static void sctp_v6_destroy_sock(struct sock *sk)
7871 sctp_destroy_sock(sk);
7872 inet6_destroy_sock(sk);
7875 struct proto sctpv6_prot = {
7877 .owner = THIS_MODULE,
7878 .close = sctp_close,
7879 .connect = sctp_connect,
7880 .disconnect = sctp_disconnect,
7881 .accept = sctp_accept,
7882 .ioctl = sctp_ioctl,
7883 .init = sctp_init_sock,
7884 .destroy = sctp_v6_destroy_sock,
7885 .shutdown = sctp_shutdown,
7886 .setsockopt = sctp_setsockopt,
7887 .getsockopt = sctp_getsockopt,
7888 .sendmsg = sctp_sendmsg,
7889 .recvmsg = sctp_recvmsg,
7891 .backlog_rcv = sctp_backlog_rcv,
7893 .unhash = sctp_unhash,
7894 .get_port = sctp_get_port,
7895 .obj_size = sizeof(struct sctp6_sock),
7896 .sysctl_mem = sysctl_sctp_mem,
7897 .sysctl_rmem = sysctl_sctp_rmem,
7898 .sysctl_wmem = sysctl_sctp_wmem,
7899 .memory_pressure = &sctp_memory_pressure,
7900 .enter_memory_pressure = sctp_enter_memory_pressure,
7901 .memory_allocated = &sctp_memory_allocated,
7902 .sockets_allocated = &sctp_sockets_allocated,
7904 #endif /* IS_ENABLED(CONFIG_IPV6) */