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>
60 #include <linux/sched/signal.h>
62 #include <linux/capability.h>
63 #include <linux/fcntl.h>
64 #include <linux/poll.h>
65 #include <linux/init.h>
66 #include <linux/slab.h>
67 #include <linux/file.h>
68 #include <linux/compat.h>
69 #include <linux/rhashtable.h>
73 #include <net/route.h>
75 #include <net/inet_common.h>
76 #include <net/busy_poll.h>
78 #include <linux/socket.h> /* for sa_family_t */
79 #include <linux/export.h>
81 #include <net/sctp/sctp.h>
82 #include <net/sctp/sm.h>
83 #include <net/sctp/stream_sched.h>
85 /* Forward declarations for internal helper functions. */
86 static bool sctp_writeable(struct sock *sk);
87 static void sctp_wfree(struct sk_buff *skb);
88 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
90 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p);
91 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
92 static int sctp_wait_for_accept(struct sock *sk, long timeo);
93 static void sctp_wait_for_close(struct sock *sk, long timeo);
94 static void sctp_destruct_sock(struct sock *sk);
95 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
96 union sctp_addr *addr, int len);
97 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
98 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
99 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
100 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
101 static int sctp_send_asconf(struct sctp_association *asoc,
102 struct sctp_chunk *chunk);
103 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
104 static int sctp_autobind(struct sock *sk);
105 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
106 struct sctp_association *assoc,
107 enum sctp_socket_type type);
109 static unsigned long sctp_memory_pressure;
110 static atomic_long_t sctp_memory_allocated;
111 struct percpu_counter sctp_sockets_allocated;
113 static void sctp_enter_memory_pressure(struct sock *sk)
115 sctp_memory_pressure = 1;
119 /* Get the sndbuf space available at the time on the association. */
120 static inline int sctp_wspace(struct sctp_association *asoc)
122 struct sock *sk = asoc->base.sk;
124 return asoc->ep->sndbuf_policy ? sk->sk_sndbuf - asoc->sndbuf_used
125 : sk_stream_wspace(sk);
128 /* Increment the used sndbuf space count of the corresponding association by
129 * the size of the outgoing data chunk.
130 * Also, set the skb destructor for sndbuf accounting later.
132 * Since it is always 1-1 between chunk and skb, and also a new skb is always
133 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
134 * destructor in the data chunk skb for the purpose of the sndbuf space
137 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
139 struct sctp_association *asoc = chunk->asoc;
140 struct sock *sk = asoc->base.sk;
142 /* The sndbuf space is tracked per association. */
143 sctp_association_hold(asoc);
146 sctp_auth_shkey_hold(chunk->shkey);
148 skb_set_owner_w(chunk->skb, sk);
150 chunk->skb->destructor = sctp_wfree;
151 /* Save the chunk pointer in skb for sctp_wfree to use later. */
152 skb_shinfo(chunk->skb)->destructor_arg = chunk;
154 refcount_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
155 asoc->sndbuf_used += chunk->skb->truesize + sizeof(struct sctp_chunk);
156 sk->sk_wmem_queued += chunk->skb->truesize + sizeof(struct sctp_chunk);
157 sk_mem_charge(sk, chunk->skb->truesize);
160 static void sctp_clear_owner_w(struct sctp_chunk *chunk)
162 skb_orphan(chunk->skb);
165 static void sctp_for_each_tx_datachunk(struct sctp_association *asoc,
166 void (*cb)(struct sctp_chunk *))
169 struct sctp_outq *q = &asoc->outqueue;
170 struct sctp_transport *t;
171 struct sctp_chunk *chunk;
173 list_for_each_entry(t, &asoc->peer.transport_addr_list, transports)
174 list_for_each_entry(chunk, &t->transmitted, transmitted_list)
177 list_for_each_entry(chunk, &q->retransmit, transmitted_list)
180 list_for_each_entry(chunk, &q->sacked, transmitted_list)
183 list_for_each_entry(chunk, &q->abandoned, transmitted_list)
186 list_for_each_entry(chunk, &q->out_chunk_list, list)
190 static void sctp_for_each_rx_skb(struct sctp_association *asoc, struct sock *sk,
191 void (*cb)(struct sk_buff *, struct sock *))
194 struct sk_buff *skb, *tmp;
196 sctp_skb_for_each(skb, &asoc->ulpq.lobby, tmp)
199 sctp_skb_for_each(skb, &asoc->ulpq.reasm, tmp)
202 sctp_skb_for_each(skb, &asoc->ulpq.reasm_uo, tmp)
206 /* Verify that this is a valid address. */
207 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
212 /* Verify basic sockaddr. */
213 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
217 /* Is this a valid SCTP address? */
218 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
221 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
227 /* Look up the association by its id. If this is not a UDP-style
228 * socket, the ID field is always ignored.
230 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
232 struct sctp_association *asoc = NULL;
234 /* If this is not a UDP-style socket, assoc id should be ignored. */
235 if (!sctp_style(sk, UDP)) {
236 /* Return NULL if the socket state is not ESTABLISHED. It
237 * could be a TCP-style listening socket or a socket which
238 * hasn't yet called connect() to establish an association.
240 if (!sctp_sstate(sk, ESTABLISHED) && !sctp_sstate(sk, CLOSING))
243 /* Get the first and the only association from the list. */
244 if (!list_empty(&sctp_sk(sk)->ep->asocs))
245 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
246 struct sctp_association, asocs);
250 /* Otherwise this is a UDP-style socket. */
251 if (!id || (id == (sctp_assoc_t)-1))
254 spin_lock_bh(&sctp_assocs_id_lock);
255 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
256 if (asoc && (asoc->base.sk != sk || asoc->base.dead))
258 spin_unlock_bh(&sctp_assocs_id_lock);
263 /* Look up the transport from an address and an assoc id. If both address and
264 * id are specified, the associations matching the address and the id should be
267 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
268 struct sockaddr_storage *addr,
271 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
272 struct sctp_af *af = sctp_get_af_specific(addr->ss_family);
273 union sctp_addr *laddr = (union sctp_addr *)addr;
274 struct sctp_transport *transport;
276 if (!af || sctp_verify_addr(sk, laddr, af->sockaddr_len))
279 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
286 id_asoc = sctp_id2assoc(sk, id);
287 if (id_asoc && (id_asoc != addr_asoc))
290 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
291 (union sctp_addr *)addr);
296 /* API 3.1.2 bind() - UDP Style Syntax
297 * The syntax of bind() is,
299 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
301 * sd - the socket descriptor returned by socket().
302 * addr - the address structure (struct sockaddr_in or struct
303 * sockaddr_in6 [RFC 2553]),
304 * addr_len - the size of the address structure.
306 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
312 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
315 /* Disallow binding twice. */
316 if (!sctp_sk(sk)->ep->base.bind_addr.port)
317 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
327 static long sctp_get_port_local(struct sock *, union sctp_addr *);
329 /* Verify this is a valid sockaddr. */
330 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
331 union sctp_addr *addr, int len)
335 /* Check minimum size. */
336 if (len < sizeof (struct sockaddr))
339 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
342 if (addr->sa.sa_family == AF_INET6) {
343 if (len < SIN6_LEN_RFC2133)
345 /* V4 mapped address are really of AF_INET family */
346 if (ipv6_addr_v4mapped(&addr->v6.sin6_addr) &&
347 !opt->pf->af_supported(AF_INET, opt))
351 /* If we get this far, af is valid. */
352 af = sctp_get_af_specific(addr->sa.sa_family);
354 if (len < af->sockaddr_len)
360 /* Bind a local address either to an endpoint or to an association. */
361 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
363 struct net *net = sock_net(sk);
364 struct sctp_sock *sp = sctp_sk(sk);
365 struct sctp_endpoint *ep = sp->ep;
366 struct sctp_bind_addr *bp = &ep->base.bind_addr;
371 /* Common sockaddr verification. */
372 af = sctp_sockaddr_af(sp, addr, len);
374 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
375 __func__, sk, addr, len);
379 snum = ntohs(addr->v4.sin_port);
381 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
382 __func__, sk, &addr->sa, bp->port, snum, len);
384 /* PF specific bind() address verification. */
385 if (!sp->pf->bind_verify(sp, addr))
386 return -EADDRNOTAVAIL;
388 /* We must either be unbound, or bind to the same port.
389 * It's OK to allow 0 ports if we are already bound.
390 * We'll just inhert an already bound port in this case
395 else if (snum != bp->port) {
396 pr_debug("%s: new port %d doesn't match existing port "
397 "%d\n", __func__, snum, bp->port);
402 if (snum && snum < inet_prot_sock(net) &&
403 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
406 /* See if the address matches any of the addresses we may have
407 * already bound before checking against other endpoints.
409 if (sctp_bind_addr_match(bp, addr, sp))
412 /* Make sure we are allowed to bind here.
413 * The function sctp_get_port_local() does duplicate address
416 addr->v4.sin_port = htons(snum);
417 if ((ret = sctp_get_port_local(sk, addr))) {
421 /* Refresh ephemeral port. */
423 bp->port = inet_sk(sk)->inet_num;
425 /* Add the address to the bind address list.
426 * Use GFP_ATOMIC since BHs will be disabled.
428 ret = sctp_add_bind_addr(bp, addr, af->sockaddr_len,
429 SCTP_ADDR_SRC, GFP_ATOMIC);
431 /* Copy back into socket for getsockname() use. */
433 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
434 sp->pf->to_sk_saddr(addr, sk);
440 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
442 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
443 * at any one time. If a sender, after sending an ASCONF chunk, decides
444 * it needs to transfer another ASCONF Chunk, it MUST wait until the
445 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
446 * subsequent ASCONF. Note this restriction binds each side, so at any
447 * time two ASCONF may be in-transit on any given association (one sent
448 * from each endpoint).
450 static int sctp_send_asconf(struct sctp_association *asoc,
451 struct sctp_chunk *chunk)
453 struct net *net = sock_net(asoc->base.sk);
456 /* If there is an outstanding ASCONF chunk, queue it for later
459 if (asoc->addip_last_asconf) {
460 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
464 /* Hold the chunk until an ASCONF_ACK is received. */
465 sctp_chunk_hold(chunk);
466 retval = sctp_primitive_ASCONF(net, asoc, chunk);
468 sctp_chunk_free(chunk);
470 asoc->addip_last_asconf = chunk;
476 /* Add a list of addresses as bind addresses to local endpoint or
479 * Basically run through each address specified in the addrs/addrcnt
480 * array/length pair, determine if it is IPv6 or IPv4 and call
481 * sctp_do_bind() on it.
483 * If any of them fails, then the operation will be reversed and the
484 * ones that were added will be removed.
486 * Only sctp_setsockopt_bindx() is supposed to call this function.
488 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
493 struct sockaddr *sa_addr;
496 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
500 for (cnt = 0; cnt < addrcnt; cnt++) {
501 /* The list may contain either IPv4 or IPv6 address;
502 * determine the address length for walking thru the list.
505 af = sctp_get_af_specific(sa_addr->sa_family);
511 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
514 addr_buf += af->sockaddr_len;
518 /* Failed. Cleanup the ones that have been added */
520 sctp_bindx_rem(sk, addrs, cnt);
528 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
529 * associations that are part of the endpoint indicating that a list of local
530 * addresses are added to the endpoint.
532 * If any of the addresses is already in the bind address list of the
533 * association, we do not send the chunk for that association. But it will not
534 * affect other associations.
536 * Only sctp_setsockopt_bindx() is supposed to call this function.
538 static int sctp_send_asconf_add_ip(struct sock *sk,
539 struct sockaddr *addrs,
542 struct net *net = sock_net(sk);
543 struct sctp_sock *sp;
544 struct sctp_endpoint *ep;
545 struct sctp_association *asoc;
546 struct sctp_bind_addr *bp;
547 struct sctp_chunk *chunk;
548 struct sctp_sockaddr_entry *laddr;
549 union sctp_addr *addr;
550 union sctp_addr saveaddr;
557 if (!net->sctp.addip_enable)
563 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
564 __func__, sk, addrs, addrcnt);
566 list_for_each_entry(asoc, &ep->asocs, asocs) {
567 if (!asoc->peer.asconf_capable)
570 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
573 if (!sctp_state(asoc, ESTABLISHED))
576 /* Check if any address in the packed array of addresses is
577 * in the bind address list of the association. If so,
578 * do not send the asconf chunk to its peer, but continue with
579 * other associations.
582 for (i = 0; i < addrcnt; i++) {
584 af = sctp_get_af_specific(addr->v4.sin_family);
590 if (sctp_assoc_lookup_laddr(asoc, addr))
593 addr_buf += af->sockaddr_len;
598 /* Use the first valid address in bind addr list of
599 * association as Address Parameter of ASCONF CHUNK.
601 bp = &asoc->base.bind_addr;
602 p = bp->address_list.next;
603 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
604 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
605 addrcnt, SCTP_PARAM_ADD_IP);
611 /* Add the new addresses to the bind address list with
612 * use_as_src set to 0.
615 for (i = 0; i < addrcnt; i++) {
617 af = sctp_get_af_specific(addr->v4.sin_family);
618 memcpy(&saveaddr, addr, af->sockaddr_len);
619 retval = sctp_add_bind_addr(bp, &saveaddr,
621 SCTP_ADDR_NEW, GFP_ATOMIC);
622 addr_buf += af->sockaddr_len;
624 if (asoc->src_out_of_asoc_ok) {
625 struct sctp_transport *trans;
627 list_for_each_entry(trans,
628 &asoc->peer.transport_addr_list, transports) {
629 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
630 2*asoc->pathmtu, 4380));
631 trans->ssthresh = asoc->peer.i.a_rwnd;
632 trans->rto = asoc->rto_initial;
633 sctp_max_rto(asoc, trans);
634 trans->rtt = trans->srtt = trans->rttvar = 0;
635 /* Clear the source and route cache */
636 sctp_transport_route(trans, NULL,
637 sctp_sk(asoc->base.sk));
640 retval = sctp_send_asconf(asoc, chunk);
647 /* Remove a list of addresses from bind addresses list. Do not remove the
650 * Basically run through each address specified in the addrs/addrcnt
651 * array/length pair, determine if it is IPv6 or IPv4 and call
652 * sctp_del_bind() on it.
654 * If any of them fails, then the operation will be reversed and the
655 * ones that were removed will be added back.
657 * At least one address has to be left; if only one address is
658 * available, the operation will return -EBUSY.
660 * Only sctp_setsockopt_bindx() is supposed to call this function.
662 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
664 struct sctp_sock *sp = sctp_sk(sk);
665 struct sctp_endpoint *ep = sp->ep;
667 struct sctp_bind_addr *bp = &ep->base.bind_addr;
670 union sctp_addr *sa_addr;
673 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
674 __func__, sk, addrs, addrcnt);
677 for (cnt = 0; cnt < addrcnt; cnt++) {
678 /* If the bind address list is empty or if there is only one
679 * bind address, there is nothing more to be removed (we need
680 * at least one address here).
682 if (list_empty(&bp->address_list) ||
683 (sctp_list_single_entry(&bp->address_list))) {
689 af = sctp_get_af_specific(sa_addr->sa.sa_family);
695 if (!af->addr_valid(sa_addr, sp, NULL)) {
696 retval = -EADDRNOTAVAIL;
700 if (sa_addr->v4.sin_port &&
701 sa_addr->v4.sin_port != htons(bp->port)) {
706 if (!sa_addr->v4.sin_port)
707 sa_addr->v4.sin_port = htons(bp->port);
709 /* FIXME - There is probably a need to check if sk->sk_saddr and
710 * sk->sk_rcv_addr are currently set to one of the addresses to
711 * be removed. This is something which needs to be looked into
712 * when we are fixing the outstanding issues with multi-homing
713 * socket routing and failover schemes. Refer to comments in
714 * sctp_do_bind(). -daisy
716 retval = sctp_del_bind_addr(bp, sa_addr);
718 addr_buf += af->sockaddr_len;
721 /* Failed. Add the ones that has been removed back */
723 sctp_bindx_add(sk, addrs, cnt);
731 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
732 * the associations that are part of the endpoint indicating that a list of
733 * local addresses are removed from the endpoint.
735 * If any of the addresses is already in the bind address list of the
736 * association, we do not send the chunk for that association. But it will not
737 * affect other associations.
739 * Only sctp_setsockopt_bindx() is supposed to call this function.
741 static int sctp_send_asconf_del_ip(struct sock *sk,
742 struct sockaddr *addrs,
745 struct net *net = sock_net(sk);
746 struct sctp_sock *sp;
747 struct sctp_endpoint *ep;
748 struct sctp_association *asoc;
749 struct sctp_transport *transport;
750 struct sctp_bind_addr *bp;
751 struct sctp_chunk *chunk;
752 union sctp_addr *laddr;
755 struct sctp_sockaddr_entry *saddr;
761 if (!net->sctp.addip_enable)
767 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
768 __func__, sk, addrs, addrcnt);
770 list_for_each_entry(asoc, &ep->asocs, asocs) {
772 if (!asoc->peer.asconf_capable)
775 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
778 if (!sctp_state(asoc, ESTABLISHED))
781 /* Check if any address in the packed array of addresses is
782 * not present in the bind address list of the association.
783 * If so, do not send the asconf chunk to its peer, but
784 * continue with other associations.
787 for (i = 0; i < addrcnt; i++) {
789 af = sctp_get_af_specific(laddr->v4.sin_family);
795 if (!sctp_assoc_lookup_laddr(asoc, laddr))
798 addr_buf += af->sockaddr_len;
803 /* Find one address in the association's bind address list
804 * that is not in the packed array of addresses. This is to
805 * make sure that we do not delete all the addresses in the
808 bp = &asoc->base.bind_addr;
809 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
811 if ((laddr == NULL) && (addrcnt == 1)) {
812 if (asoc->asconf_addr_del_pending)
814 asoc->asconf_addr_del_pending =
815 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
816 if (asoc->asconf_addr_del_pending == NULL) {
820 asoc->asconf_addr_del_pending->sa.sa_family =
822 asoc->asconf_addr_del_pending->v4.sin_port =
824 if (addrs->sa_family == AF_INET) {
825 struct sockaddr_in *sin;
827 sin = (struct sockaddr_in *)addrs;
828 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
829 } else if (addrs->sa_family == AF_INET6) {
830 struct sockaddr_in6 *sin6;
832 sin6 = (struct sockaddr_in6 *)addrs;
833 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
836 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
837 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
838 asoc->asconf_addr_del_pending);
840 asoc->src_out_of_asoc_ok = 1;
848 /* We do not need RCU protection throughout this loop
849 * because this is done under a socket lock from the
852 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
860 /* Reset use_as_src flag for the addresses in the bind address
861 * list that are to be deleted.
864 for (i = 0; i < addrcnt; i++) {
866 af = sctp_get_af_specific(laddr->v4.sin_family);
867 list_for_each_entry(saddr, &bp->address_list, list) {
868 if (sctp_cmp_addr_exact(&saddr->a, laddr))
869 saddr->state = SCTP_ADDR_DEL;
871 addr_buf += af->sockaddr_len;
874 /* Update the route and saddr entries for all the transports
875 * as some of the addresses in the bind address list are
876 * about to be deleted and cannot be used as source addresses.
878 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
880 sctp_transport_route(transport, NULL,
881 sctp_sk(asoc->base.sk));
885 /* We don't need to transmit ASCONF */
887 retval = sctp_send_asconf(asoc, chunk);
893 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
894 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
896 struct sock *sk = sctp_opt2sk(sp);
897 union sctp_addr *addr;
900 /* It is safe to write port space in caller. */
902 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
903 af = sctp_get_af_specific(addr->sa.sa_family);
906 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
909 if (addrw->state == SCTP_ADDR_NEW)
910 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
912 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
915 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
918 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
921 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
922 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
925 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
926 * Section 3.1.2 for this usage.
928 * addrs is a pointer to an array of one or more socket addresses. Each
929 * address is contained in its appropriate structure (i.e. struct
930 * sockaddr_in or struct sockaddr_in6) the family of the address type
931 * must be used to distinguish the address length (note that this
932 * representation is termed a "packed array" of addresses). The caller
933 * specifies the number of addresses in the array with addrcnt.
935 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
936 * -1, and sets errno to the appropriate error code.
938 * For SCTP, the port given in each socket address must be the same, or
939 * sctp_bindx() will fail, setting errno to EINVAL.
941 * The flags parameter is formed from the bitwise OR of zero or more of
942 * the following currently defined flags:
944 * SCTP_BINDX_ADD_ADDR
946 * SCTP_BINDX_REM_ADDR
948 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
949 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
950 * addresses from the association. The two flags are mutually exclusive;
951 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
952 * not remove all addresses from an association; sctp_bindx() will
953 * reject such an attempt with EINVAL.
955 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
956 * additional addresses with an endpoint after calling bind(). Or use
957 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
958 * socket is associated with so that no new association accepted will be
959 * associated with those addresses. If the endpoint supports dynamic
960 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
961 * endpoint to send the appropriate message to the peer to change the
962 * peers address lists.
964 * Adding and removing addresses from a connected association is
965 * optional functionality. Implementations that do not support this
966 * functionality should return EOPNOTSUPP.
968 * Basically do nothing but copying the addresses from user to kernel
969 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
970 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
973 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
976 * sk The sk of the socket
977 * addrs The pointer to the addresses in user land
978 * addrssize Size of the addrs buffer
979 * op Operation to perform (add or remove, see the flags of
982 * Returns 0 if ok, <0 errno code on error.
984 static int sctp_setsockopt_bindx(struct sock *sk,
985 struct sockaddr __user *addrs,
986 int addrs_size, int op)
988 struct sockaddr *kaddrs;
992 struct sockaddr *sa_addr;
996 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
997 __func__, sk, addrs, addrs_size, op);
999 if (unlikely(addrs_size <= 0))
1002 kaddrs = vmemdup_user(addrs, addrs_size);
1003 if (unlikely(IS_ERR(kaddrs)))
1004 return PTR_ERR(kaddrs);
1006 /* Walk through the addrs buffer and count the number of addresses. */
1008 while (walk_size < addrs_size) {
1009 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1015 af = sctp_get_af_specific(sa_addr->sa_family);
1017 /* If the address family is not supported or if this address
1018 * causes the address buffer to overflow return EINVAL.
1020 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1025 addr_buf += af->sockaddr_len;
1026 walk_size += af->sockaddr_len;
1031 case SCTP_BINDX_ADD_ADDR:
1032 /* Allow security module to validate bindx addresses. */
1033 err = security_sctp_bind_connect(sk, SCTP_SOCKOPT_BINDX_ADD,
1034 (struct sockaddr *)kaddrs,
1038 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1041 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1044 case SCTP_BINDX_REM_ADDR:
1045 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1048 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1062 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1064 * Common routine for handling connect() and sctp_connectx().
1065 * Connect will come in with just a single address.
1067 static int __sctp_connect(struct sock *sk,
1068 struct sockaddr *kaddrs,
1069 int addrs_size, int flags,
1070 sctp_assoc_t *assoc_id)
1072 struct net *net = sock_net(sk);
1073 struct sctp_sock *sp;
1074 struct sctp_endpoint *ep;
1075 struct sctp_association *asoc = NULL;
1076 struct sctp_association *asoc2;
1077 struct sctp_transport *transport;
1079 enum sctp_scope scope;
1084 union sctp_addr *sa_addr = NULL;
1086 unsigned short port;
1091 /* connect() cannot be done on a socket that is already in ESTABLISHED
1092 * state - UDP-style peeled off socket or a TCP-style socket that
1093 * is already connected.
1094 * It cannot be done even on a TCP-style listening socket.
1096 if (sctp_sstate(sk, ESTABLISHED) || sctp_sstate(sk, CLOSING) ||
1097 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1102 /* Walk through the addrs buffer and count the number of addresses. */
1104 while (walk_size < addrs_size) {
1107 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1113 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1115 /* If the address family is not supported or if this address
1116 * causes the address buffer to overflow return EINVAL.
1118 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1123 port = ntohs(sa_addr->v4.sin_port);
1125 /* Save current address so we can work with it */
1126 memcpy(&to, sa_addr, af->sockaddr_len);
1128 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1132 /* Make sure the destination port is correctly set
1135 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1140 /* Check if there already is a matching association on the
1141 * endpoint (other than the one created here).
1143 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1144 if (asoc2 && asoc2 != asoc) {
1145 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1152 /* If we could not find a matching association on the endpoint,
1153 * make sure that there is no peeled-off association matching
1154 * the peer address even on another socket.
1156 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1157 err = -EADDRNOTAVAIL;
1162 /* If a bind() or sctp_bindx() is not called prior to
1163 * an sctp_connectx() call, the system picks an
1164 * ephemeral port and will choose an address set
1165 * equivalent to binding with a wildcard address.
1167 if (!ep->base.bind_addr.port) {
1168 if (sctp_autobind(sk)) {
1174 * If an unprivileged user inherits a 1-many
1175 * style socket with open associations on a
1176 * privileged port, it MAY be permitted to
1177 * accept new associations, but it SHOULD NOT
1178 * be permitted to open new associations.
1180 if (ep->base.bind_addr.port <
1181 inet_prot_sock(net) &&
1182 !ns_capable(net->user_ns,
1183 CAP_NET_BIND_SERVICE)) {
1189 scope = sctp_scope(&to);
1190 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1196 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1204 /* Prime the peer's transport structures. */
1205 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1213 addr_buf += af->sockaddr_len;
1214 walk_size += af->sockaddr_len;
1217 /* In case the user of sctp_connectx() wants an association
1218 * id back, assign one now.
1221 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1226 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1231 /* Initialize sk's dport and daddr for getpeername() */
1232 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1233 sp->pf->to_sk_daddr(sa_addr, sk);
1236 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
1239 *assoc_id = asoc->assoc_id;
1241 err = sctp_wait_for_connect(asoc, &timeo);
1242 /* Note: the asoc may be freed after the return of
1243 * sctp_wait_for_connect.
1246 /* Don't free association on exit. */
1250 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1251 __func__, asoc, kaddrs, err);
1254 /* sctp_primitive_ASSOCIATE may have added this association
1255 * To the hash table, try to unhash it, just in case, its a noop
1256 * if it wasn't hashed so we're safe
1258 sctp_association_free(asoc);
1263 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1266 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1267 * sctp_assoc_t *asoc);
1269 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1270 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1271 * or IPv6 addresses.
1273 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1274 * Section 3.1.2 for this usage.
1276 * addrs is a pointer to an array of one or more socket addresses. Each
1277 * address is contained in its appropriate structure (i.e. struct
1278 * sockaddr_in or struct sockaddr_in6) the family of the address type
1279 * must be used to distengish the address length (note that this
1280 * representation is termed a "packed array" of addresses). The caller
1281 * specifies the number of addresses in the array with addrcnt.
1283 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1284 * the association id of the new association. On failure, sctp_connectx()
1285 * returns -1, and sets errno to the appropriate error code. The assoc_id
1286 * is not touched by the kernel.
1288 * For SCTP, the port given in each socket address must be the same, or
1289 * sctp_connectx() will fail, setting errno to EINVAL.
1291 * An application can use sctp_connectx to initiate an association with
1292 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1293 * allows a caller to specify multiple addresses at which a peer can be
1294 * reached. The way the SCTP stack uses the list of addresses to set up
1295 * the association is implementation dependent. This function only
1296 * specifies that the stack will try to make use of all the addresses in
1297 * the list when needed.
1299 * Note that the list of addresses passed in is only used for setting up
1300 * the association. It does not necessarily equal the set of addresses
1301 * the peer uses for the resulting association. If the caller wants to
1302 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1303 * retrieve them after the association has been set up.
1305 * Basically do nothing but copying the addresses from user to kernel
1306 * land and invoking either sctp_connectx(). This is used for tunneling
1307 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1309 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1312 * sk The sk of the socket
1313 * addrs The pointer to the addresses in user land
1314 * addrssize Size of the addrs buffer
1316 * Returns >=0 if ok, <0 errno code on error.
1318 static int __sctp_setsockopt_connectx(struct sock *sk,
1319 struct sockaddr __user *addrs,
1321 sctp_assoc_t *assoc_id)
1323 struct sockaddr *kaddrs;
1324 int err = 0, flags = 0;
1326 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1327 __func__, sk, addrs, addrs_size);
1329 if (unlikely(addrs_size <= 0))
1332 kaddrs = vmemdup_user(addrs, addrs_size);
1333 if (unlikely(IS_ERR(kaddrs)))
1334 return PTR_ERR(kaddrs);
1336 /* Allow security module to validate connectx addresses. */
1337 err = security_sctp_bind_connect(sk, SCTP_SOCKOPT_CONNECTX,
1338 (struct sockaddr *)kaddrs,
1343 /* in-kernel sockets don't generally have a file allocated to them
1344 * if all they do is call sock_create_kern().
1346 if (sk->sk_socket->file)
1347 flags = sk->sk_socket->file->f_flags;
1349 err = __sctp_connect(sk, kaddrs, addrs_size, flags, assoc_id);
1358 * This is an older interface. It's kept for backward compatibility
1359 * to the option that doesn't provide association id.
1361 static int sctp_setsockopt_connectx_old(struct sock *sk,
1362 struct sockaddr __user *addrs,
1365 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1369 * New interface for the API. The since the API is done with a socket
1370 * option, to make it simple we feed back the association id is as a return
1371 * indication to the call. Error is always negative and association id is
1374 static int sctp_setsockopt_connectx(struct sock *sk,
1375 struct sockaddr __user *addrs,
1378 sctp_assoc_t assoc_id = 0;
1381 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1390 * New (hopefully final) interface for the API.
1391 * We use the sctp_getaddrs_old structure so that use-space library
1392 * can avoid any unnecessary allocations. The only different part
1393 * is that we store the actual length of the address buffer into the
1394 * addrs_num structure member. That way we can re-use the existing
1397 #ifdef CONFIG_COMPAT
1398 struct compat_sctp_getaddrs_old {
1399 sctp_assoc_t assoc_id;
1401 compat_uptr_t addrs; /* struct sockaddr * */
1405 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1406 char __user *optval,
1409 struct sctp_getaddrs_old param;
1410 sctp_assoc_t assoc_id = 0;
1413 #ifdef CONFIG_COMPAT
1414 if (in_compat_syscall()) {
1415 struct compat_sctp_getaddrs_old param32;
1417 if (len < sizeof(param32))
1419 if (copy_from_user(¶m32, optval, sizeof(param32)))
1422 param.assoc_id = param32.assoc_id;
1423 param.addr_num = param32.addr_num;
1424 param.addrs = compat_ptr(param32.addrs);
1428 if (len < sizeof(param))
1430 if (copy_from_user(¶m, optval, sizeof(param)))
1434 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1435 param.addrs, param.addr_num,
1437 if (err == 0 || err == -EINPROGRESS) {
1438 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1440 if (put_user(sizeof(assoc_id), optlen))
1447 /* API 3.1.4 close() - UDP Style Syntax
1448 * Applications use close() to perform graceful shutdown (as described in
1449 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1450 * by a UDP-style socket.
1454 * ret = close(int sd);
1456 * sd - the socket descriptor of the associations to be closed.
1458 * To gracefully shutdown a specific association represented by the
1459 * UDP-style socket, an application should use the sendmsg() call,
1460 * passing no user data, but including the appropriate flag in the
1461 * ancillary data (see Section xxxx).
1463 * If sd in the close() call is a branched-off socket representing only
1464 * one association, the shutdown is performed on that association only.
1466 * 4.1.6 close() - TCP Style Syntax
1468 * Applications use close() to gracefully close down an association.
1472 * int close(int sd);
1474 * sd - the socket descriptor of the association to be closed.
1476 * After an application calls close() on a socket descriptor, no further
1477 * socket operations will succeed on that descriptor.
1479 * API 7.1.4 SO_LINGER
1481 * An application using the TCP-style socket can use this option to
1482 * perform the SCTP ABORT primitive. The linger option structure is:
1485 * int l_onoff; // option on/off
1486 * int l_linger; // linger time
1489 * To enable the option, set l_onoff to 1. If the l_linger value is set
1490 * to 0, calling close() is the same as the ABORT primitive. If the
1491 * value is set to a negative value, the setsockopt() call will return
1492 * an error. If the value is set to a positive value linger_time, the
1493 * close() can be blocked for at most linger_time ms. If the graceful
1494 * shutdown phase does not finish during this period, close() will
1495 * return but the graceful shutdown phase continues in the system.
1497 static void sctp_close(struct sock *sk, long timeout)
1499 struct net *net = sock_net(sk);
1500 struct sctp_endpoint *ep;
1501 struct sctp_association *asoc;
1502 struct list_head *pos, *temp;
1503 unsigned int data_was_unread;
1505 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1507 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
1508 sk->sk_shutdown = SHUTDOWN_MASK;
1509 inet_sk_set_state(sk, SCTP_SS_CLOSING);
1511 ep = sctp_sk(sk)->ep;
1513 /* Clean up any skbs sitting on the receive queue. */
1514 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1515 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1517 /* Walk all associations on an endpoint. */
1518 list_for_each_safe(pos, temp, &ep->asocs) {
1519 asoc = list_entry(pos, struct sctp_association, asocs);
1521 if (sctp_style(sk, TCP)) {
1522 /* A closed association can still be in the list if
1523 * it belongs to a TCP-style listening socket that is
1524 * not yet accepted. If so, free it. If not, send an
1525 * ABORT or SHUTDOWN based on the linger options.
1527 if (sctp_state(asoc, CLOSED)) {
1528 sctp_association_free(asoc);
1533 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1534 !skb_queue_empty(&asoc->ulpq.reasm) ||
1535 !skb_queue_empty(&asoc->ulpq.reasm_uo) ||
1536 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1537 struct sctp_chunk *chunk;
1539 chunk = sctp_make_abort_user(asoc, NULL, 0);
1540 sctp_primitive_ABORT(net, asoc, chunk);
1542 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1545 /* On a TCP-style socket, block for at most linger_time if set. */
1546 if (sctp_style(sk, TCP) && timeout)
1547 sctp_wait_for_close(sk, timeout);
1549 /* This will run the backlog queue. */
1552 /* Supposedly, no process has access to the socket, but
1553 * the net layers still may.
1554 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1555 * held and that should be grabbed before socket lock.
1557 spin_lock_bh(&net->sctp.addr_wq_lock);
1558 bh_lock_sock_nested(sk);
1560 /* Hold the sock, since sk_common_release() will put sock_put()
1561 * and we have just a little more cleanup.
1564 sk_common_release(sk);
1567 spin_unlock_bh(&net->sctp.addr_wq_lock);
1571 SCTP_DBG_OBJCNT_DEC(sock);
1574 /* Handle EPIPE error. */
1575 static int sctp_error(struct sock *sk, int flags, int err)
1578 err = sock_error(sk) ? : -EPIPE;
1579 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1580 send_sig(SIGPIPE, current, 0);
1584 /* API 3.1.3 sendmsg() - UDP Style Syntax
1586 * An application uses sendmsg() and recvmsg() calls to transmit data to
1587 * and receive data from its peer.
1589 * ssize_t sendmsg(int socket, const struct msghdr *message,
1592 * socket - the socket descriptor of the endpoint.
1593 * message - pointer to the msghdr structure which contains a single
1594 * user message and possibly some ancillary data.
1596 * See Section 5 for complete description of the data
1599 * flags - flags sent or received with the user message, see Section
1600 * 5 for complete description of the flags.
1602 * Note: This function could use a rewrite especially when explicit
1603 * connect support comes in.
1605 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1607 static int sctp_msghdr_parse(const struct msghdr *msg,
1608 struct sctp_cmsgs *cmsgs);
1610 static int sctp_sendmsg_parse(struct sock *sk, struct sctp_cmsgs *cmsgs,
1611 struct sctp_sndrcvinfo *srinfo,
1612 const struct msghdr *msg, size_t msg_len)
1617 if (sctp_sstate(sk, LISTENING) && sctp_style(sk, TCP))
1620 if (msg_len > sk->sk_sndbuf)
1623 memset(cmsgs, 0, sizeof(*cmsgs));
1624 err = sctp_msghdr_parse(msg, cmsgs);
1626 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1630 memset(srinfo, 0, sizeof(*srinfo));
1631 if (cmsgs->srinfo) {
1632 srinfo->sinfo_stream = cmsgs->srinfo->sinfo_stream;
1633 srinfo->sinfo_flags = cmsgs->srinfo->sinfo_flags;
1634 srinfo->sinfo_ppid = cmsgs->srinfo->sinfo_ppid;
1635 srinfo->sinfo_context = cmsgs->srinfo->sinfo_context;
1636 srinfo->sinfo_assoc_id = cmsgs->srinfo->sinfo_assoc_id;
1637 srinfo->sinfo_timetolive = cmsgs->srinfo->sinfo_timetolive;
1641 srinfo->sinfo_stream = cmsgs->sinfo->snd_sid;
1642 srinfo->sinfo_flags = cmsgs->sinfo->snd_flags;
1643 srinfo->sinfo_ppid = cmsgs->sinfo->snd_ppid;
1644 srinfo->sinfo_context = cmsgs->sinfo->snd_context;
1645 srinfo->sinfo_assoc_id = cmsgs->sinfo->snd_assoc_id;
1648 if (cmsgs->prinfo) {
1649 srinfo->sinfo_timetolive = cmsgs->prinfo->pr_value;
1650 SCTP_PR_SET_POLICY(srinfo->sinfo_flags,
1651 cmsgs->prinfo->pr_policy);
1654 sflags = srinfo->sinfo_flags;
1655 if (!sflags && msg_len)
1658 if (sctp_style(sk, TCP) && (sflags & (SCTP_EOF | SCTP_ABORT)))
1661 if (((sflags & SCTP_EOF) && msg_len > 0) ||
1662 (!(sflags & (SCTP_EOF | SCTP_ABORT)) && msg_len == 0))
1665 if ((sflags & SCTP_ADDR_OVER) && !msg->msg_name)
1671 static int sctp_sendmsg_new_asoc(struct sock *sk, __u16 sflags,
1672 struct sctp_cmsgs *cmsgs,
1673 union sctp_addr *daddr,
1674 struct sctp_transport **tp)
1676 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
1677 struct net *net = sock_net(sk);
1678 struct sctp_association *asoc;
1679 enum sctp_scope scope;
1680 struct cmsghdr *cmsg;
1681 __be32 flowinfo = 0;
1687 if (sflags & (SCTP_EOF | SCTP_ABORT))
1690 if (sctp_style(sk, TCP) && (sctp_sstate(sk, ESTABLISHED) ||
1691 sctp_sstate(sk, CLOSING)))
1692 return -EADDRNOTAVAIL;
1694 if (sctp_endpoint_is_peeled_off(ep, daddr))
1695 return -EADDRNOTAVAIL;
1697 if (!ep->base.bind_addr.port) {
1698 if (sctp_autobind(sk))
1701 if (ep->base.bind_addr.port < inet_prot_sock(net) &&
1702 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
1706 scope = sctp_scope(daddr);
1708 /* Label connection socket for first association 1-to-many
1709 * style for client sequence socket()->sendmsg(). This
1710 * needs to be done before sctp_assoc_add_peer() as that will
1711 * set up the initial packet that needs to account for any
1712 * security ip options (CIPSO/CALIPSO) added to the packet.
1714 af = sctp_get_af_specific(daddr->sa.sa_family);
1717 err = security_sctp_bind_connect(sk, SCTP_SENDMSG_CONNECT,
1718 (struct sockaddr *)daddr,
1723 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1727 if (sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL) < 0) {
1733 struct sctp_initmsg *init = cmsgs->init;
1735 if (init->sinit_num_ostreams) {
1736 __u16 outcnt = init->sinit_num_ostreams;
1738 asoc->c.sinit_num_ostreams = outcnt;
1739 /* outcnt has been changed, need to re-init stream */
1740 err = sctp_stream_init(&asoc->stream, outcnt, 0,
1746 if (init->sinit_max_instreams)
1747 asoc->c.sinit_max_instreams = init->sinit_max_instreams;
1749 if (init->sinit_max_attempts)
1750 asoc->max_init_attempts = init->sinit_max_attempts;
1752 if (init->sinit_max_init_timeo)
1753 asoc->max_init_timeo =
1754 msecs_to_jiffies(init->sinit_max_init_timeo);
1757 *tp = sctp_assoc_add_peer(asoc, daddr, GFP_KERNEL, SCTP_UNKNOWN);
1763 if (!cmsgs->addrs_msg)
1766 if (daddr->sa.sa_family == AF_INET6)
1767 flowinfo = daddr->v6.sin6_flowinfo;
1769 /* sendv addr list parse */
1770 for_each_cmsghdr(cmsg, cmsgs->addrs_msg) {
1771 struct sctp_transport *transport;
1772 struct sctp_association *old;
1773 union sctp_addr _daddr;
1776 if (cmsg->cmsg_level != IPPROTO_SCTP ||
1777 (cmsg->cmsg_type != SCTP_DSTADDRV4 &&
1778 cmsg->cmsg_type != SCTP_DSTADDRV6))
1782 memset(daddr, 0, sizeof(*daddr));
1783 dlen = cmsg->cmsg_len - sizeof(struct cmsghdr);
1784 if (cmsg->cmsg_type == SCTP_DSTADDRV4) {
1785 if (dlen < sizeof(struct in_addr)) {
1790 dlen = sizeof(struct in_addr);
1791 daddr->v4.sin_family = AF_INET;
1792 daddr->v4.sin_port = htons(asoc->peer.port);
1793 memcpy(&daddr->v4.sin_addr, CMSG_DATA(cmsg), dlen);
1795 if (dlen < sizeof(struct in6_addr)) {
1800 dlen = sizeof(struct in6_addr);
1801 daddr->v6.sin6_flowinfo = flowinfo;
1802 daddr->v6.sin6_family = AF_INET6;
1803 daddr->v6.sin6_port = htons(asoc->peer.port);
1804 memcpy(&daddr->v6.sin6_addr, CMSG_DATA(cmsg), dlen);
1806 err = sctp_verify_addr(sk, daddr, sizeof(*daddr));
1810 old = sctp_endpoint_lookup_assoc(ep, daddr, &transport);
1811 if (old && old != asoc) {
1812 if (old->state >= SCTP_STATE_ESTABLISHED)
1819 if (sctp_endpoint_is_peeled_off(ep, daddr)) {
1820 err = -EADDRNOTAVAIL;
1824 transport = sctp_assoc_add_peer(asoc, daddr, GFP_KERNEL,
1835 sctp_association_free(asoc);
1839 static int sctp_sendmsg_check_sflags(struct sctp_association *asoc,
1840 __u16 sflags, struct msghdr *msg,
1843 struct sock *sk = asoc->base.sk;
1844 struct net *net = sock_net(sk);
1846 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP))
1849 if ((sflags & SCTP_SENDALL) && sctp_style(sk, UDP) &&
1850 !sctp_state(asoc, ESTABLISHED))
1853 if (sflags & SCTP_EOF) {
1854 pr_debug("%s: shutting down association:%p\n", __func__, asoc);
1855 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1860 if (sflags & SCTP_ABORT) {
1861 struct sctp_chunk *chunk;
1863 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1867 pr_debug("%s: aborting association:%p\n", __func__, asoc);
1868 sctp_primitive_ABORT(net, asoc, chunk);
1876 static int sctp_sendmsg_to_asoc(struct sctp_association *asoc,
1877 struct msghdr *msg, size_t msg_len,
1878 struct sctp_transport *transport,
1879 struct sctp_sndrcvinfo *sinfo)
1881 struct sock *sk = asoc->base.sk;
1882 struct sctp_sock *sp = sctp_sk(sk);
1883 struct net *net = sock_net(sk);
1884 struct sctp_datamsg *datamsg;
1885 bool wait_connect = false;
1886 struct sctp_chunk *chunk;
1890 if (sinfo->sinfo_stream >= asoc->stream.outcnt) {
1895 if (unlikely(!SCTP_SO(&asoc->stream, sinfo->sinfo_stream)->ext)) {
1896 err = sctp_stream_init_ext(&asoc->stream, sinfo->sinfo_stream);
1901 if (sp->disable_fragments && msg_len > asoc->frag_point) {
1906 if (asoc->pmtu_pending) {
1907 if (sp->param_flags & SPP_PMTUD_ENABLE)
1908 sctp_assoc_sync_pmtu(asoc);
1909 asoc->pmtu_pending = 0;
1912 if (sctp_wspace(asoc) < (int)msg_len)
1913 sctp_prsctp_prune(asoc, sinfo, msg_len - sctp_wspace(asoc));
1915 if (sctp_wspace(asoc) <= 0) {
1916 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1917 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1922 if (sctp_state(asoc, CLOSED)) {
1923 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1927 if (sp->strm_interleave) {
1928 timeo = sock_sndtimeo(sk, 0);
1929 err = sctp_wait_for_connect(asoc, &timeo);
1935 wait_connect = true;
1938 pr_debug("%s: we associated primitively\n", __func__);
1941 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
1942 if (IS_ERR(datamsg)) {
1943 err = PTR_ERR(datamsg);
1947 asoc->force_delay = !!(msg->msg_flags & MSG_MORE);
1949 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1950 sctp_chunk_hold(chunk);
1951 sctp_set_owner_w(chunk);
1952 chunk->transport = transport;
1955 err = sctp_primitive_SEND(net, asoc, datamsg);
1957 sctp_datamsg_free(datamsg);
1961 pr_debug("%s: we sent primitively\n", __func__);
1963 sctp_datamsg_put(datamsg);
1965 if (unlikely(wait_connect)) {
1966 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1967 sctp_wait_for_connect(asoc, &timeo);
1976 static union sctp_addr *sctp_sendmsg_get_daddr(struct sock *sk,
1977 const struct msghdr *msg,
1978 struct sctp_cmsgs *cmsgs)
1980 union sctp_addr *daddr = NULL;
1983 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1984 int len = msg->msg_namelen;
1986 if (len > sizeof(*daddr))
1987 len = sizeof(*daddr);
1989 daddr = (union sctp_addr *)msg->msg_name;
1991 err = sctp_verify_addr(sk, daddr, len);
1993 return ERR_PTR(err);
1999 static void sctp_sendmsg_update_sinfo(struct sctp_association *asoc,
2000 struct sctp_sndrcvinfo *sinfo,
2001 struct sctp_cmsgs *cmsgs)
2003 if (!cmsgs->srinfo && !cmsgs->sinfo) {
2004 sinfo->sinfo_stream = asoc->default_stream;
2005 sinfo->sinfo_ppid = asoc->default_ppid;
2006 sinfo->sinfo_context = asoc->default_context;
2007 sinfo->sinfo_assoc_id = sctp_assoc2id(asoc);
2010 sinfo->sinfo_flags = asoc->default_flags;
2013 if (!cmsgs->srinfo && !cmsgs->prinfo)
2014 sinfo->sinfo_timetolive = asoc->default_timetolive;
2016 if (cmsgs->authinfo) {
2017 /* Reuse sinfo_tsn to indicate that authinfo was set and
2018 * sinfo_ssn to save the keyid on tx path.
2020 sinfo->sinfo_tsn = 1;
2021 sinfo->sinfo_ssn = cmsgs->authinfo->auth_keynumber;
2025 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
2027 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
2028 struct sctp_transport *transport = NULL;
2029 struct sctp_sndrcvinfo _sinfo, *sinfo;
2030 struct sctp_association *asoc;
2031 struct sctp_cmsgs cmsgs;
2032 union sctp_addr *daddr;
2037 /* Parse and get snd_info */
2038 err = sctp_sendmsg_parse(sk, &cmsgs, &_sinfo, msg, msg_len);
2043 sflags = sinfo->sinfo_flags;
2045 /* Get daddr from msg */
2046 daddr = sctp_sendmsg_get_daddr(sk, msg, &cmsgs);
2047 if (IS_ERR(daddr)) {
2048 err = PTR_ERR(daddr);
2054 /* SCTP_SENDALL process */
2055 if ((sflags & SCTP_SENDALL) && sctp_style(sk, UDP)) {
2056 list_for_each_entry(asoc, &ep->asocs, asocs) {
2057 err = sctp_sendmsg_check_sflags(asoc, sflags, msg,
2064 sctp_sendmsg_update_sinfo(asoc, sinfo, &cmsgs);
2066 err = sctp_sendmsg_to_asoc(asoc, msg, msg_len,
2071 iov_iter_revert(&msg->msg_iter, err);
2077 /* Get and check or create asoc */
2079 asoc = sctp_endpoint_lookup_assoc(ep, daddr, &transport);
2081 err = sctp_sendmsg_check_sflags(asoc, sflags, msg,
2086 err = sctp_sendmsg_new_asoc(sk, sflags, &cmsgs, daddr,
2091 asoc = transport->asoc;
2095 if (!sctp_style(sk, TCP) && !(sflags & SCTP_ADDR_OVER))
2098 asoc = sctp_id2assoc(sk, sinfo->sinfo_assoc_id);
2104 err = sctp_sendmsg_check_sflags(asoc, sflags, msg, msg_len);
2109 /* Update snd_info with the asoc */
2110 sctp_sendmsg_update_sinfo(asoc, sinfo, &cmsgs);
2112 /* Send msg to the asoc */
2113 err = sctp_sendmsg_to_asoc(asoc, msg, msg_len, transport, sinfo);
2114 if (err < 0 && err != -ESRCH && new)
2115 sctp_association_free(asoc);
2120 return sctp_error(sk, msg->msg_flags, err);
2123 /* This is an extended version of skb_pull() that removes the data from the
2124 * start of a skb even when data is spread across the list of skb's in the
2125 * frag_list. len specifies the total amount of data that needs to be removed.
2126 * when 'len' bytes could be removed from the skb, it returns 0.
2127 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2128 * could not be removed.
2130 static int sctp_skb_pull(struct sk_buff *skb, int len)
2132 struct sk_buff *list;
2133 int skb_len = skb_headlen(skb);
2136 if (len <= skb_len) {
2137 __skb_pull(skb, len);
2141 __skb_pull(skb, skb_len);
2143 skb_walk_frags(skb, list) {
2144 rlen = sctp_skb_pull(list, len);
2145 skb->len -= (len-rlen);
2146 skb->data_len -= (len-rlen);
2157 /* API 3.1.3 recvmsg() - UDP Style Syntax
2159 * ssize_t recvmsg(int socket, struct msghdr *message,
2162 * socket - the socket descriptor of the endpoint.
2163 * message - pointer to the msghdr structure which contains a single
2164 * user message and possibly some ancillary data.
2166 * See Section 5 for complete description of the data
2169 * flags - flags sent or received with the user message, see Section
2170 * 5 for complete description of the flags.
2172 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2173 int noblock, int flags, int *addr_len)
2175 struct sctp_ulpevent *event = NULL;
2176 struct sctp_sock *sp = sctp_sk(sk);
2177 struct sk_buff *skb, *head_skb;
2182 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2183 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2188 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED) &&
2189 !sctp_sstate(sk, CLOSING) && !sctp_sstate(sk, CLOSED)) {
2194 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2198 /* Get the total length of the skb including any skb's in the
2207 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2209 event = sctp_skb2event(skb);
2214 if (event->chunk && event->chunk->head_skb)
2215 head_skb = event->chunk->head_skb;
2218 sock_recv_ts_and_drops(msg, sk, head_skb);
2219 if (sctp_ulpevent_is_notification(event)) {
2220 msg->msg_flags |= MSG_NOTIFICATION;
2221 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2223 sp->pf->skb_msgname(head_skb, msg->msg_name, addr_len);
2226 /* Check if we allow SCTP_NXTINFO. */
2227 if (sp->recvnxtinfo)
2228 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2229 /* Check if we allow SCTP_RCVINFO. */
2230 if (sp->recvrcvinfo)
2231 sctp_ulpevent_read_rcvinfo(event, msg);
2232 /* Check if we allow SCTP_SNDRCVINFO. */
2233 if (sp->subscribe.sctp_data_io_event)
2234 sctp_ulpevent_read_sndrcvinfo(event, msg);
2238 /* If skb's length exceeds the user's buffer, update the skb and
2239 * push it back to the receive_queue so that the next call to
2240 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2242 if (skb_len > copied) {
2243 msg->msg_flags &= ~MSG_EOR;
2244 if (flags & MSG_PEEK)
2246 sctp_skb_pull(skb, copied);
2247 skb_queue_head(&sk->sk_receive_queue, skb);
2249 /* When only partial message is copied to the user, increase
2250 * rwnd by that amount. If all the data in the skb is read,
2251 * rwnd is updated when the event is freed.
2253 if (!sctp_ulpevent_is_notification(event))
2254 sctp_assoc_rwnd_increase(event->asoc, copied);
2256 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2257 (event->msg_flags & MSG_EOR))
2258 msg->msg_flags |= MSG_EOR;
2260 msg->msg_flags &= ~MSG_EOR;
2263 if (flags & MSG_PEEK) {
2264 /* Release the skb reference acquired after peeking the skb in
2265 * sctp_skb_recv_datagram().
2269 /* Free the event which includes releasing the reference to
2270 * the owner of the skb, freeing the skb and updating the
2273 sctp_ulpevent_free(event);
2280 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2282 * This option is a on/off flag. If enabled no SCTP message
2283 * fragmentation will be performed. Instead if a message being sent
2284 * exceeds the current PMTU size, the message will NOT be sent and
2285 * instead a error will be indicated to the user.
2287 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2288 char __user *optval,
2289 unsigned int optlen)
2293 if (optlen < sizeof(int))
2296 if (get_user(val, (int __user *)optval))
2299 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2304 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2305 unsigned int optlen)
2307 struct sctp_association *asoc;
2308 struct sctp_ulpevent *event;
2310 if (optlen > sizeof(struct sctp_event_subscribe))
2312 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2315 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2316 * if there is no data to be sent or retransmit, the stack will
2317 * immediately send up this notification.
2319 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2320 &sctp_sk(sk)->subscribe)) {
2321 asoc = sctp_id2assoc(sk, 0);
2323 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2324 event = sctp_ulpevent_make_sender_dry_event(asoc,
2325 GFP_USER | __GFP_NOWARN);
2329 asoc->stream.si->enqueue_event(&asoc->ulpq, event);
2336 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2338 * This socket option is applicable to the UDP-style socket only. When
2339 * set it will cause associations that are idle for more than the
2340 * specified number of seconds to automatically close. An association
2341 * being idle is defined an association that has NOT sent or received
2342 * user data. The special value of '0' indicates that no automatic
2343 * close of any associations should be performed. The option expects an
2344 * integer defining the number of seconds of idle time before an
2345 * association is closed.
2347 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2348 unsigned int optlen)
2350 struct sctp_sock *sp = sctp_sk(sk);
2351 struct net *net = sock_net(sk);
2353 /* Applicable to UDP-style socket only */
2354 if (sctp_style(sk, TCP))
2356 if (optlen != sizeof(int))
2358 if (copy_from_user(&sp->autoclose, optval, optlen))
2361 if (sp->autoclose > net->sctp.max_autoclose)
2362 sp->autoclose = net->sctp.max_autoclose;
2367 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2369 * Applications can enable or disable heartbeats for any peer address of
2370 * an association, modify an address's heartbeat interval, force a
2371 * heartbeat to be sent immediately, and adjust the address's maximum
2372 * number of retransmissions sent before an address is considered
2373 * unreachable. The following structure is used to access and modify an
2374 * address's parameters:
2376 * struct sctp_paddrparams {
2377 * sctp_assoc_t spp_assoc_id;
2378 * struct sockaddr_storage spp_address;
2379 * uint32_t spp_hbinterval;
2380 * uint16_t spp_pathmaxrxt;
2381 * uint32_t spp_pathmtu;
2382 * uint32_t spp_sackdelay;
2383 * uint32_t spp_flags;
2384 * uint32_t spp_ipv6_flowlabel;
2388 * spp_assoc_id - (one-to-many style socket) This is filled in the
2389 * application, and identifies the association for
2391 * spp_address - This specifies which address is of interest.
2392 * spp_hbinterval - This contains the value of the heartbeat interval,
2393 * in milliseconds. If a value of zero
2394 * is present in this field then no changes are to
2395 * be made to this parameter.
2396 * spp_pathmaxrxt - This contains the maximum number of
2397 * retransmissions before this address shall be
2398 * considered unreachable. If a value of zero
2399 * is present in this field then no changes are to
2400 * be made to this parameter.
2401 * spp_pathmtu - When Path MTU discovery is disabled the value
2402 * specified here will be the "fixed" path mtu.
2403 * Note that if the spp_address field is empty
2404 * then all associations on this address will
2405 * have this fixed path mtu set upon them.
2407 * spp_sackdelay - When delayed sack is enabled, this value specifies
2408 * the number of milliseconds that sacks will be delayed
2409 * for. This value will apply to all addresses of an
2410 * association if the spp_address field is empty. Note
2411 * also, that if delayed sack is enabled and this
2412 * value is set to 0, no change is made to the last
2413 * recorded delayed sack timer value.
2415 * spp_flags - These flags are used to control various features
2416 * on an association. The flag field may contain
2417 * zero or more of the following options.
2419 * SPP_HB_ENABLE - Enable heartbeats on the
2420 * specified address. Note that if the address
2421 * field is empty all addresses for the association
2422 * have heartbeats enabled upon them.
2424 * SPP_HB_DISABLE - Disable heartbeats on the
2425 * speicifed address. Note that if the address
2426 * field is empty all addresses for the association
2427 * will have their heartbeats disabled. Note also
2428 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2429 * mutually exclusive, only one of these two should
2430 * be specified. Enabling both fields will have
2431 * undetermined results.
2433 * SPP_HB_DEMAND - Request a user initiated heartbeat
2434 * to be made immediately.
2436 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2437 * heartbeat delayis to be set to the value of 0
2440 * SPP_PMTUD_ENABLE - This field will enable PMTU
2441 * discovery upon the specified address. Note that
2442 * if the address feild is empty then all addresses
2443 * on the association are effected.
2445 * SPP_PMTUD_DISABLE - This field will disable PMTU
2446 * discovery upon the specified address. Note that
2447 * if the address feild is empty then all addresses
2448 * on the association are effected. Not also that
2449 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2450 * exclusive. Enabling both will have undetermined
2453 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2454 * on delayed sack. The time specified in spp_sackdelay
2455 * is used to specify the sack delay for this address. Note
2456 * that if spp_address is empty then all addresses will
2457 * enable delayed sack and take on the sack delay
2458 * value specified in spp_sackdelay.
2459 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2460 * off delayed sack. If the spp_address field is blank then
2461 * delayed sack is disabled for the entire association. Note
2462 * also that this field is mutually exclusive to
2463 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2466 * SPP_IPV6_FLOWLABEL: Setting this flag enables the
2467 * setting of the IPV6 flow label value. The value is
2468 * contained in the spp_ipv6_flowlabel field.
2469 * Upon retrieval, this flag will be set to indicate that
2470 * the spp_ipv6_flowlabel field has a valid value returned.
2471 * If a specific destination address is set (in the
2472 * spp_address field), then the value returned is that of
2473 * the address. If just an association is specified (and
2474 * no address), then the association's default flow label
2475 * is returned. If neither an association nor a destination
2476 * is specified, then the socket's default flow label is
2477 * returned. For non-IPv6 sockets, this flag will be left
2480 * SPP_DSCP: Setting this flag enables the setting of the
2481 * Differentiated Services Code Point (DSCP) value
2482 * associated with either the association or a specific
2483 * address. The value is obtained in the spp_dscp field.
2484 * Upon retrieval, this flag will be set to indicate that
2485 * the spp_dscp field has a valid value returned. If a
2486 * specific destination address is set when called (in the
2487 * spp_address field), then that specific destination
2488 * address's DSCP value is returned. If just an association
2489 * is specified, then the association's default DSCP is
2490 * returned. If neither an association nor a destination is
2491 * specified, then the socket's default DSCP is returned.
2493 * spp_ipv6_flowlabel
2494 * - This field is used in conjunction with the
2495 * SPP_IPV6_FLOWLABEL flag and contains the IPv6 flow label.
2496 * The 20 least significant bits are used for the flow
2497 * label. This setting has precedence over any IPv6-layer
2500 * spp_dscp - This field is used in conjunction with the SPP_DSCP flag
2501 * and contains the DSCP. The 6 most significant bits are
2502 * used for the DSCP. This setting has precedence over any
2503 * IPv4- or IPv6- layer setting.
2505 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2506 struct sctp_transport *trans,
2507 struct sctp_association *asoc,
2508 struct sctp_sock *sp,
2511 int sackdelay_change)
2515 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2516 struct net *net = sock_net(trans->asoc->base.sk);
2518 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2523 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2524 * this field is ignored. Note also that a value of zero indicates
2525 * the current setting should be left unchanged.
2527 if (params->spp_flags & SPP_HB_ENABLE) {
2529 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2530 * set. This lets us use 0 value when this flag
2533 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2534 params->spp_hbinterval = 0;
2536 if (params->spp_hbinterval ||
2537 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2540 msecs_to_jiffies(params->spp_hbinterval);
2543 msecs_to_jiffies(params->spp_hbinterval);
2545 sp->hbinterval = params->spp_hbinterval;
2552 trans->param_flags =
2553 (trans->param_flags & ~SPP_HB) | hb_change;
2556 (asoc->param_flags & ~SPP_HB) | hb_change;
2559 (sp->param_flags & ~SPP_HB) | hb_change;
2563 /* When Path MTU discovery is disabled the value specified here will
2564 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2565 * include the flag SPP_PMTUD_DISABLE for this field to have any
2568 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2570 trans->pathmtu = params->spp_pathmtu;
2571 sctp_assoc_sync_pmtu(asoc);
2573 sctp_assoc_set_pmtu(asoc, params->spp_pathmtu);
2575 sp->pathmtu = params->spp_pathmtu;
2581 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2582 (params->spp_flags & SPP_PMTUD_ENABLE);
2583 trans->param_flags =
2584 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2586 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2587 sctp_assoc_sync_pmtu(asoc);
2591 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2594 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2598 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2599 * value of this field is ignored. Note also that a value of zero
2600 * indicates the current setting should be left unchanged.
2602 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2605 msecs_to_jiffies(params->spp_sackdelay);
2608 msecs_to_jiffies(params->spp_sackdelay);
2610 sp->sackdelay = params->spp_sackdelay;
2614 if (sackdelay_change) {
2616 trans->param_flags =
2617 (trans->param_flags & ~SPP_SACKDELAY) |
2621 (asoc->param_flags & ~SPP_SACKDELAY) |
2625 (sp->param_flags & ~SPP_SACKDELAY) |
2630 /* Note that a value of zero indicates the current setting should be
2633 if (params->spp_pathmaxrxt) {
2635 trans->pathmaxrxt = params->spp_pathmaxrxt;
2637 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2639 sp->pathmaxrxt = params->spp_pathmaxrxt;
2643 if (params->spp_flags & SPP_IPV6_FLOWLABEL) {
2645 if (trans->ipaddr.sa.sa_family == AF_INET6) {
2646 trans->flowlabel = params->spp_ipv6_flowlabel &
2647 SCTP_FLOWLABEL_VAL_MASK;
2648 trans->flowlabel |= SCTP_FLOWLABEL_SET_MASK;
2651 struct sctp_transport *t;
2653 list_for_each_entry(t, &asoc->peer.transport_addr_list,
2655 if (t->ipaddr.sa.sa_family != AF_INET6)
2657 t->flowlabel = params->spp_ipv6_flowlabel &
2658 SCTP_FLOWLABEL_VAL_MASK;
2659 t->flowlabel |= SCTP_FLOWLABEL_SET_MASK;
2661 asoc->flowlabel = params->spp_ipv6_flowlabel &
2662 SCTP_FLOWLABEL_VAL_MASK;
2663 asoc->flowlabel |= SCTP_FLOWLABEL_SET_MASK;
2664 } else if (sctp_opt2sk(sp)->sk_family == AF_INET6) {
2665 sp->flowlabel = params->spp_ipv6_flowlabel &
2666 SCTP_FLOWLABEL_VAL_MASK;
2667 sp->flowlabel |= SCTP_FLOWLABEL_SET_MASK;
2671 if (params->spp_flags & SPP_DSCP) {
2673 trans->dscp = params->spp_dscp & SCTP_DSCP_VAL_MASK;
2674 trans->dscp |= SCTP_DSCP_SET_MASK;
2676 struct sctp_transport *t;
2678 list_for_each_entry(t, &asoc->peer.transport_addr_list,
2680 t->dscp = params->spp_dscp &
2682 t->dscp |= SCTP_DSCP_SET_MASK;
2684 asoc->dscp = params->spp_dscp & SCTP_DSCP_VAL_MASK;
2685 asoc->dscp |= SCTP_DSCP_SET_MASK;
2687 sp->dscp = params->spp_dscp & SCTP_DSCP_VAL_MASK;
2688 sp->dscp |= SCTP_DSCP_SET_MASK;
2695 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2696 char __user *optval,
2697 unsigned int optlen)
2699 struct sctp_paddrparams params;
2700 struct sctp_transport *trans = NULL;
2701 struct sctp_association *asoc = NULL;
2702 struct sctp_sock *sp = sctp_sk(sk);
2704 int hb_change, pmtud_change, sackdelay_change;
2706 if (optlen == sizeof(params)) {
2707 if (copy_from_user(¶ms, optval, optlen))
2709 } else if (optlen == ALIGN(offsetof(struct sctp_paddrparams,
2710 spp_ipv6_flowlabel), 4)) {
2711 if (copy_from_user(¶ms, optval, optlen))
2713 if (params.spp_flags & (SPP_DSCP | SPP_IPV6_FLOWLABEL))
2719 /* Validate flags and value parameters. */
2720 hb_change = params.spp_flags & SPP_HB;
2721 pmtud_change = params.spp_flags & SPP_PMTUD;
2722 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2724 if (hb_change == SPP_HB ||
2725 pmtud_change == SPP_PMTUD ||
2726 sackdelay_change == SPP_SACKDELAY ||
2727 params.spp_sackdelay > 500 ||
2728 (params.spp_pathmtu &&
2729 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2732 /* If an address other than INADDR_ANY is specified, and
2733 * no transport is found, then the request is invalid.
2735 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
2736 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2737 params.spp_assoc_id);
2742 /* Get association, if assoc_id != 0 and the socket is a one
2743 * to many style socket, and an association was not found, then
2744 * the id was invalid.
2746 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2747 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2750 /* Heartbeat demand can only be sent on a transport or
2751 * association, but not a socket.
2753 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2756 /* Process parameters. */
2757 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2758 hb_change, pmtud_change,
2764 /* If changes are for association, also apply parameters to each
2767 if (!trans && asoc) {
2768 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2770 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2771 hb_change, pmtud_change,
2779 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2781 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2784 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2786 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2790 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2792 * This option will effect the way delayed acks are performed. This
2793 * option allows you to get or set the delayed ack time, in
2794 * milliseconds. It also allows changing the delayed ack frequency.
2795 * Changing the frequency to 1 disables the delayed sack algorithm. If
2796 * the assoc_id is 0, then this sets or gets the endpoints default
2797 * values. If the assoc_id field is non-zero, then the set or get
2798 * effects the specified association for the one to many model (the
2799 * assoc_id field is ignored by the one to one model). Note that if
2800 * sack_delay or sack_freq are 0 when setting this option, then the
2801 * current values will remain unchanged.
2803 * struct sctp_sack_info {
2804 * sctp_assoc_t sack_assoc_id;
2805 * uint32_t sack_delay;
2806 * uint32_t sack_freq;
2809 * sack_assoc_id - This parameter, indicates which association the user
2810 * is performing an action upon. Note that if this field's value is
2811 * zero then the endpoints default value is changed (effecting future
2812 * associations only).
2814 * sack_delay - This parameter contains the number of milliseconds that
2815 * the user is requesting the delayed ACK timer be set to. Note that
2816 * this value is defined in the standard to be between 200 and 500
2819 * sack_freq - This parameter contains the number of packets that must
2820 * be received before a sack is sent without waiting for the delay
2821 * timer to expire. The default value for this is 2, setting this
2822 * value to 1 will disable the delayed sack algorithm.
2825 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2826 char __user *optval, unsigned int optlen)
2828 struct sctp_sack_info params;
2829 struct sctp_transport *trans = NULL;
2830 struct sctp_association *asoc = NULL;
2831 struct sctp_sock *sp = sctp_sk(sk);
2833 if (optlen == sizeof(struct sctp_sack_info)) {
2834 if (copy_from_user(¶ms, optval, optlen))
2837 if (params.sack_delay == 0 && params.sack_freq == 0)
2839 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2840 pr_warn_ratelimited(DEPRECATED
2842 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2843 "Use struct sctp_sack_info instead\n",
2844 current->comm, task_pid_nr(current));
2845 if (copy_from_user(¶ms, optval, optlen))
2848 if (params.sack_delay == 0)
2849 params.sack_freq = 1;
2851 params.sack_freq = 0;
2855 /* Validate value parameter. */
2856 if (params.sack_delay > 500)
2859 /* Get association, if sack_assoc_id != 0 and the socket is a one
2860 * to many style socket, and an association was not found, then
2861 * the id was invalid.
2863 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2864 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2867 if (params.sack_delay) {
2870 msecs_to_jiffies(params.sack_delay);
2872 sctp_spp_sackdelay_enable(asoc->param_flags);
2874 sp->sackdelay = params.sack_delay;
2876 sctp_spp_sackdelay_enable(sp->param_flags);
2880 if (params.sack_freq == 1) {
2883 sctp_spp_sackdelay_disable(asoc->param_flags);
2886 sctp_spp_sackdelay_disable(sp->param_flags);
2888 } else if (params.sack_freq > 1) {
2890 asoc->sackfreq = params.sack_freq;
2892 sctp_spp_sackdelay_enable(asoc->param_flags);
2894 sp->sackfreq = params.sack_freq;
2896 sctp_spp_sackdelay_enable(sp->param_flags);
2900 /* If change is for association, also apply to each transport. */
2902 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2904 if (params.sack_delay) {
2906 msecs_to_jiffies(params.sack_delay);
2907 trans->param_flags =
2908 sctp_spp_sackdelay_enable(trans->param_flags);
2910 if (params.sack_freq == 1) {
2911 trans->param_flags =
2912 sctp_spp_sackdelay_disable(trans->param_flags);
2913 } else if (params.sack_freq > 1) {
2914 trans->sackfreq = params.sack_freq;
2915 trans->param_flags =
2916 sctp_spp_sackdelay_enable(trans->param_flags);
2924 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2926 * Applications can specify protocol parameters for the default association
2927 * initialization. The option name argument to setsockopt() and getsockopt()
2930 * Setting initialization parameters is effective only on an unconnected
2931 * socket (for UDP-style sockets only future associations are effected
2932 * by the change). With TCP-style sockets, this option is inherited by
2933 * sockets derived from a listener socket.
2935 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2937 struct sctp_initmsg sinit;
2938 struct sctp_sock *sp = sctp_sk(sk);
2940 if (optlen != sizeof(struct sctp_initmsg))
2942 if (copy_from_user(&sinit, optval, optlen))
2945 if (sinit.sinit_num_ostreams)
2946 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2947 if (sinit.sinit_max_instreams)
2948 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2949 if (sinit.sinit_max_attempts)
2950 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2951 if (sinit.sinit_max_init_timeo)
2952 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2958 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2960 * Applications that wish to use the sendto() system call may wish to
2961 * specify a default set of parameters that would normally be supplied
2962 * through the inclusion of ancillary data. This socket option allows
2963 * such an application to set the default sctp_sndrcvinfo structure.
2964 * The application that wishes to use this socket option simply passes
2965 * in to this call the sctp_sndrcvinfo structure defined in Section
2966 * 5.2.2) The input parameters accepted by this call include
2967 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2968 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2969 * to this call if the caller is using the UDP model.
2971 static int sctp_setsockopt_default_send_param(struct sock *sk,
2972 char __user *optval,
2973 unsigned int optlen)
2975 struct sctp_sock *sp = sctp_sk(sk);
2976 struct sctp_association *asoc;
2977 struct sctp_sndrcvinfo info;
2979 if (optlen != sizeof(info))
2981 if (copy_from_user(&info, optval, optlen))
2983 if (info.sinfo_flags &
2984 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2985 SCTP_ABORT | SCTP_EOF))
2988 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2989 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2992 asoc->default_stream = info.sinfo_stream;
2993 asoc->default_flags = info.sinfo_flags;
2994 asoc->default_ppid = info.sinfo_ppid;
2995 asoc->default_context = info.sinfo_context;
2996 asoc->default_timetolive = info.sinfo_timetolive;
2998 sp->default_stream = info.sinfo_stream;
2999 sp->default_flags = info.sinfo_flags;
3000 sp->default_ppid = info.sinfo_ppid;
3001 sp->default_context = info.sinfo_context;
3002 sp->default_timetolive = info.sinfo_timetolive;
3008 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
3009 * (SCTP_DEFAULT_SNDINFO)
3011 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
3012 char __user *optval,
3013 unsigned int optlen)
3015 struct sctp_sock *sp = sctp_sk(sk);
3016 struct sctp_association *asoc;
3017 struct sctp_sndinfo info;
3019 if (optlen != sizeof(info))
3021 if (copy_from_user(&info, optval, optlen))
3023 if (info.snd_flags &
3024 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
3025 SCTP_ABORT | SCTP_EOF))
3028 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
3029 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
3032 asoc->default_stream = info.snd_sid;
3033 asoc->default_flags = info.snd_flags;
3034 asoc->default_ppid = info.snd_ppid;
3035 asoc->default_context = info.snd_context;
3037 sp->default_stream = info.snd_sid;
3038 sp->default_flags = info.snd_flags;
3039 sp->default_ppid = info.snd_ppid;
3040 sp->default_context = info.snd_context;
3046 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
3048 * Requests that the local SCTP stack use the enclosed peer address as
3049 * the association primary. The enclosed address must be one of the
3050 * association peer's addresses.
3052 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
3053 unsigned int optlen)
3055 struct sctp_prim prim;
3056 struct sctp_transport *trans;
3060 if (optlen != sizeof(struct sctp_prim))
3063 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
3066 /* Allow security module to validate address but need address len. */
3067 af = sctp_get_af_specific(prim.ssp_addr.ss_family);
3071 err = security_sctp_bind_connect(sk, SCTP_PRIMARY_ADDR,
3072 (struct sockaddr *)&prim.ssp_addr,
3077 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
3081 sctp_assoc_set_primary(trans->asoc, trans);
3087 * 7.1.5 SCTP_NODELAY
3089 * Turn on/off any Nagle-like algorithm. This means that packets are
3090 * generally sent as soon as possible and no unnecessary delays are
3091 * introduced, at the cost of more packets in the network. Expects an
3092 * integer boolean flag.
3094 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
3095 unsigned int optlen)
3099 if (optlen < sizeof(int))
3101 if (get_user(val, (int __user *)optval))
3104 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
3110 * 7.1.1 SCTP_RTOINFO
3112 * The protocol parameters used to initialize and bound retransmission
3113 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
3114 * and modify these parameters.
3115 * All parameters are time values, in milliseconds. A value of 0, when
3116 * modifying the parameters, indicates that the current value should not
3120 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
3122 struct sctp_rtoinfo rtoinfo;
3123 struct sctp_association *asoc;
3124 unsigned long rto_min, rto_max;
3125 struct sctp_sock *sp = sctp_sk(sk);
3127 if (optlen != sizeof (struct sctp_rtoinfo))
3130 if (copy_from_user(&rtoinfo, optval, optlen))
3133 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
3135 /* Set the values to the specific association */
3136 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
3139 rto_max = rtoinfo.srto_max;
3140 rto_min = rtoinfo.srto_min;
3143 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
3145 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
3148 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
3150 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
3152 if (rto_min > rto_max)
3156 if (rtoinfo.srto_initial != 0)
3158 msecs_to_jiffies(rtoinfo.srto_initial);
3159 asoc->rto_max = rto_max;
3160 asoc->rto_min = rto_min;
3162 /* If there is no association or the association-id = 0
3163 * set the values to the endpoint.
3165 if (rtoinfo.srto_initial != 0)
3166 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
3167 sp->rtoinfo.srto_max = rto_max;
3168 sp->rtoinfo.srto_min = rto_min;
3176 * 7.1.2 SCTP_ASSOCINFO
3178 * This option is used to tune the maximum retransmission attempts
3179 * of the association.
3180 * Returns an error if the new association retransmission value is
3181 * greater than the sum of the retransmission value of the peer.
3182 * See [SCTP] for more information.
3185 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
3188 struct sctp_assocparams assocparams;
3189 struct sctp_association *asoc;
3191 if (optlen != sizeof(struct sctp_assocparams))
3193 if (copy_from_user(&assocparams, optval, optlen))
3196 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
3198 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
3201 /* Set the values to the specific association */
3203 if (assocparams.sasoc_asocmaxrxt != 0) {
3206 struct sctp_transport *peer_addr;
3208 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
3210 path_sum += peer_addr->pathmaxrxt;
3214 /* Only validate asocmaxrxt if we have more than
3215 * one path/transport. We do this because path
3216 * retransmissions are only counted when we have more
3220 assocparams.sasoc_asocmaxrxt > path_sum)
3223 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3226 if (assocparams.sasoc_cookie_life != 0)
3227 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3229 /* Set the values to the endpoint */
3230 struct sctp_sock *sp = sctp_sk(sk);
3232 if (assocparams.sasoc_asocmaxrxt != 0)
3233 sp->assocparams.sasoc_asocmaxrxt =
3234 assocparams.sasoc_asocmaxrxt;
3235 if (assocparams.sasoc_cookie_life != 0)
3236 sp->assocparams.sasoc_cookie_life =
3237 assocparams.sasoc_cookie_life;
3243 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3245 * This socket option is a boolean flag which turns on or off mapped V4
3246 * addresses. If this option is turned on and the socket is type
3247 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3248 * If this option is turned off, then no mapping will be done of V4
3249 * addresses and a user will receive both PF_INET6 and PF_INET type
3250 * addresses on the socket.
3252 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3255 struct sctp_sock *sp = sctp_sk(sk);
3257 if (optlen < sizeof(int))
3259 if (get_user(val, (int __user *)optval))
3270 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3271 * This option will get or set the maximum size to put in any outgoing
3272 * SCTP DATA chunk. If a message is larger than this size it will be
3273 * fragmented by SCTP into the specified size. Note that the underlying
3274 * SCTP implementation may fragment into smaller sized chunks when the
3275 * PMTU of the underlying association is smaller than the value set by
3276 * the user. The default value for this option is '0' which indicates
3277 * the user is NOT limiting fragmentation and only the PMTU will effect
3278 * SCTP's choice of DATA chunk size. Note also that values set larger
3279 * than the maximum size of an IP datagram will effectively let SCTP
3280 * control fragmentation (i.e. the same as setting this option to 0).
3282 * The following structure is used to access and modify this parameter:
3284 * struct sctp_assoc_value {
3285 * sctp_assoc_t assoc_id;
3286 * uint32_t assoc_value;
3289 * assoc_id: This parameter is ignored for one-to-one style sockets.
3290 * For one-to-many style sockets this parameter indicates which
3291 * association the user is performing an action upon. Note that if
3292 * this field's value is zero then the endpoints default value is
3293 * changed (effecting future associations only).
3294 * assoc_value: This parameter specifies the maximum size in bytes.
3296 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3298 struct sctp_sock *sp = sctp_sk(sk);
3299 struct sctp_assoc_value params;
3300 struct sctp_association *asoc;
3303 if (optlen == sizeof(int)) {
3304 pr_warn_ratelimited(DEPRECATED
3306 "Use of int in maxseg socket option.\n"
3307 "Use struct sctp_assoc_value instead\n",
3308 current->comm, task_pid_nr(current));
3309 if (copy_from_user(&val, optval, optlen))
3311 params.assoc_id = 0;
3312 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3313 if (copy_from_user(¶ms, optval, optlen))
3315 val = params.assoc_value;
3320 asoc = sctp_id2assoc(sk, params.assoc_id);
3323 int min_len, max_len;
3324 __u16 datasize = asoc ? sctp_datachk_len(&asoc->stream) :
3325 sizeof(struct sctp_data_chunk);
3327 min_len = sctp_mtu_payload(sp, SCTP_DEFAULT_MINSEGMENT,
3329 max_len = SCTP_MAX_CHUNK_LEN - datasize;
3331 if (val < min_len || val > max_len)
3336 asoc->user_frag = val;
3337 sctp_assoc_update_frag_point(asoc);
3339 if (params.assoc_id && sctp_style(sk, UDP))
3341 sp->user_frag = val;
3349 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3351 * Requests that the peer mark the enclosed address as the association
3352 * primary. The enclosed address must be one of the association's
3353 * locally bound addresses. The following structure is used to make a
3354 * set primary request:
3356 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3357 unsigned int optlen)
3359 struct net *net = sock_net(sk);
3360 struct sctp_sock *sp;
3361 struct sctp_association *asoc = NULL;
3362 struct sctp_setpeerprim prim;
3363 struct sctp_chunk *chunk;
3369 if (!net->sctp.addip_enable)
3372 if (optlen != sizeof(struct sctp_setpeerprim))
3375 if (copy_from_user(&prim, optval, optlen))
3378 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3382 if (!asoc->peer.asconf_capable)
3385 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3388 if (!sctp_state(asoc, ESTABLISHED))
3391 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3395 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3396 return -EADDRNOTAVAIL;
3398 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3399 return -EADDRNOTAVAIL;
3401 /* Allow security module to validate address. */
3402 err = security_sctp_bind_connect(sk, SCTP_SET_PEER_PRIMARY_ADDR,
3403 (struct sockaddr *)&prim.sspp_addr,
3408 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3409 chunk = sctp_make_asconf_set_prim(asoc,
3410 (union sctp_addr *)&prim.sspp_addr);
3414 err = sctp_send_asconf(asoc, chunk);
3416 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3421 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3422 unsigned int optlen)
3424 struct sctp_setadaptation adaptation;
3426 if (optlen != sizeof(struct sctp_setadaptation))
3428 if (copy_from_user(&adaptation, optval, optlen))
3431 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3437 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3439 * The context field in the sctp_sndrcvinfo structure is normally only
3440 * used when a failed message is retrieved holding the value that was
3441 * sent down on the actual send call. This option allows the setting of
3442 * a default context on an association basis that will be received on
3443 * reading messages from the peer. This is especially helpful in the
3444 * one-2-many model for an application to keep some reference to an
3445 * internal state machine that is processing messages on the
3446 * association. Note that the setting of this value only effects
3447 * received messages from the peer and does not effect the value that is
3448 * saved with outbound messages.
3450 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3451 unsigned int optlen)
3453 struct sctp_assoc_value params;
3454 struct sctp_sock *sp;
3455 struct sctp_association *asoc;
3457 if (optlen != sizeof(struct sctp_assoc_value))
3459 if (copy_from_user(¶ms, optval, optlen))
3464 if (params.assoc_id != 0) {
3465 asoc = sctp_id2assoc(sk, params.assoc_id);
3468 asoc->default_rcv_context = params.assoc_value;
3470 sp->default_rcv_context = params.assoc_value;
3477 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3479 * This options will at a minimum specify if the implementation is doing
3480 * fragmented interleave. Fragmented interleave, for a one to many
3481 * socket, is when subsequent calls to receive a message may return
3482 * parts of messages from different associations. Some implementations
3483 * may allow you to turn this value on or off. If so, when turned off,
3484 * no fragment interleave will occur (which will cause a head of line
3485 * blocking amongst multiple associations sharing the same one to many
3486 * socket). When this option is turned on, then each receive call may
3487 * come from a different association (thus the user must receive data
3488 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3489 * association each receive belongs to.
3491 * This option takes a boolean value. A non-zero value indicates that
3492 * fragmented interleave is on. A value of zero indicates that
3493 * fragmented interleave is off.
3495 * Note that it is important that an implementation that allows this
3496 * option to be turned on, have it off by default. Otherwise an unaware
3497 * application using the one to many model may become confused and act
3500 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3501 char __user *optval,
3502 unsigned int optlen)
3506 if (optlen != sizeof(int))
3508 if (get_user(val, (int __user *)optval))
3511 sctp_sk(sk)->frag_interleave = !!val;
3513 if (!sctp_sk(sk)->frag_interleave)
3514 sctp_sk(sk)->strm_interleave = 0;
3520 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3521 * (SCTP_PARTIAL_DELIVERY_POINT)
3523 * This option will set or get the SCTP partial delivery point. This
3524 * point is the size of a message where the partial delivery API will be
3525 * invoked to help free up rwnd space for the peer. Setting this to a
3526 * lower value will cause partial deliveries to happen more often. The
3527 * calls argument is an integer that sets or gets the partial delivery
3528 * point. Note also that the call will fail if the user attempts to set
3529 * this value larger than the socket receive buffer size.
3531 * Note that any single message having a length smaller than or equal to
3532 * the SCTP partial delivery point will be delivered in one single read
3533 * call as long as the user provided buffer is large enough to hold the
3536 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3537 char __user *optval,
3538 unsigned int optlen)
3542 if (optlen != sizeof(u32))
3544 if (get_user(val, (int __user *)optval))
3547 /* Note: We double the receive buffer from what the user sets
3548 * it to be, also initial rwnd is based on rcvbuf/2.
3550 if (val > (sk->sk_rcvbuf >> 1))
3553 sctp_sk(sk)->pd_point = val;
3555 return 0; /* is this the right error code? */
3559 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3561 * This option will allow a user to change the maximum burst of packets
3562 * that can be emitted by this association. Note that the default value
3563 * is 4, and some implementations may restrict this setting so that it
3564 * can only be lowered.
3566 * NOTE: This text doesn't seem right. Do this on a socket basis with
3567 * future associations inheriting the socket value.
3569 static int sctp_setsockopt_maxburst(struct sock *sk,
3570 char __user *optval,
3571 unsigned int optlen)
3573 struct sctp_assoc_value params;
3574 struct sctp_sock *sp;
3575 struct sctp_association *asoc;
3579 if (optlen == sizeof(int)) {
3580 pr_warn_ratelimited(DEPRECATED
3582 "Use of int in max_burst socket option deprecated.\n"
3583 "Use struct sctp_assoc_value instead\n",
3584 current->comm, task_pid_nr(current));
3585 if (copy_from_user(&val, optval, optlen))
3587 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3588 if (copy_from_user(¶ms, optval, optlen))
3590 val = params.assoc_value;
3591 assoc_id = params.assoc_id;
3597 if (assoc_id != 0) {
3598 asoc = sctp_id2assoc(sk, assoc_id);
3601 asoc->max_burst = val;
3603 sp->max_burst = val;
3609 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3611 * This set option adds a chunk type that the user is requesting to be
3612 * received only in an authenticated way. Changes to the list of chunks
3613 * will only effect future associations on the socket.
3615 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3616 char __user *optval,
3617 unsigned int optlen)
3619 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3620 struct sctp_authchunk val;
3622 if (!ep->auth_enable)
3625 if (optlen != sizeof(struct sctp_authchunk))
3627 if (copy_from_user(&val, optval, optlen))
3630 switch (val.sauth_chunk) {
3632 case SCTP_CID_INIT_ACK:
3633 case SCTP_CID_SHUTDOWN_COMPLETE:
3638 /* add this chunk id to the endpoint */
3639 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3643 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3645 * This option gets or sets the list of HMAC algorithms that the local
3646 * endpoint requires the peer to use.
3648 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3649 char __user *optval,
3650 unsigned int optlen)
3652 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3653 struct sctp_hmacalgo *hmacs;
3657 if (!ep->auth_enable)
3660 if (optlen < sizeof(struct sctp_hmacalgo))
3662 optlen = min_t(unsigned int, optlen, sizeof(struct sctp_hmacalgo) +
3663 SCTP_AUTH_NUM_HMACS * sizeof(u16));
3665 hmacs = memdup_user(optval, optlen);
3667 return PTR_ERR(hmacs);
3669 idents = hmacs->shmac_num_idents;
3670 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3671 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3676 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3683 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3685 * This option will set a shared secret key which is used to build an
3686 * association shared key.
3688 static int sctp_setsockopt_auth_key(struct sock *sk,
3689 char __user *optval,
3690 unsigned int optlen)
3692 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3693 struct sctp_authkey *authkey;
3694 struct sctp_association *asoc;
3697 if (!ep->auth_enable)
3700 if (optlen <= sizeof(struct sctp_authkey))
3702 /* authkey->sca_keylength is u16, so optlen can't be bigger than
3705 optlen = min_t(unsigned int, optlen, USHRT_MAX +
3706 sizeof(struct sctp_authkey));
3708 authkey = memdup_user(optval, optlen);
3709 if (IS_ERR(authkey))
3710 return PTR_ERR(authkey);
3712 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3717 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3718 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3723 ret = sctp_auth_set_key(ep, asoc, authkey);
3730 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3732 * This option will get or set the active shared key to be used to build
3733 * the association shared key.
3735 static int sctp_setsockopt_active_key(struct sock *sk,
3736 char __user *optval,
3737 unsigned int optlen)
3739 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3740 struct sctp_authkeyid val;
3741 struct sctp_association *asoc;
3743 if (!ep->auth_enable)
3746 if (optlen != sizeof(struct sctp_authkeyid))
3748 if (copy_from_user(&val, optval, optlen))
3751 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3752 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3755 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3759 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3761 * This set option will delete a shared secret key from use.
3763 static int sctp_setsockopt_del_key(struct sock *sk,
3764 char __user *optval,
3765 unsigned int optlen)
3767 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3768 struct sctp_authkeyid val;
3769 struct sctp_association *asoc;
3771 if (!ep->auth_enable)
3774 if (optlen != sizeof(struct sctp_authkeyid))
3776 if (copy_from_user(&val, optval, optlen))
3779 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3780 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3783 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3788 * 8.3.4 Deactivate a Shared Key (SCTP_AUTH_DEACTIVATE_KEY)
3790 * This set option will deactivate a shared secret key.
3792 static int sctp_setsockopt_deactivate_key(struct sock *sk, char __user *optval,
3793 unsigned int optlen)
3795 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3796 struct sctp_authkeyid val;
3797 struct sctp_association *asoc;
3799 if (!ep->auth_enable)
3802 if (optlen != sizeof(struct sctp_authkeyid))
3804 if (copy_from_user(&val, optval, optlen))
3807 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3808 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3811 return sctp_auth_deact_key_id(ep, asoc, val.scact_keynumber);
3815 * 8.1.23 SCTP_AUTO_ASCONF
3817 * This option will enable or disable the use of the automatic generation of
3818 * ASCONF chunks to add and delete addresses to an existing association. Note
3819 * that this option has two caveats namely: a) it only affects sockets that
3820 * are bound to all addresses available to the SCTP stack, and b) the system
3821 * administrator may have an overriding control that turns the ASCONF feature
3822 * off no matter what setting the socket option may have.
3823 * This option expects an integer boolean flag, where a non-zero value turns on
3824 * the option, and a zero value turns off the option.
3825 * Note. In this implementation, socket operation overrides default parameter
3826 * being set by sysctl as well as FreeBSD implementation
3828 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3829 unsigned int optlen)
3832 struct sctp_sock *sp = sctp_sk(sk);
3834 if (optlen < sizeof(int))
3836 if (get_user(val, (int __user *)optval))
3838 if (!sctp_is_ep_boundall(sk) && val)
3840 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3843 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3844 if (val == 0 && sp->do_auto_asconf) {
3845 list_del(&sp->auto_asconf_list);
3846 sp->do_auto_asconf = 0;
3847 } else if (val && !sp->do_auto_asconf) {
3848 list_add_tail(&sp->auto_asconf_list,
3849 &sock_net(sk)->sctp.auto_asconf_splist);
3850 sp->do_auto_asconf = 1;
3852 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3857 * SCTP_PEER_ADDR_THLDS
3859 * This option allows us to alter the partially failed threshold for one or all
3860 * transports in an association. See Section 6.1 of:
3861 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3863 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3864 char __user *optval,
3865 unsigned int optlen)
3867 struct sctp_paddrthlds val;
3868 struct sctp_transport *trans;
3869 struct sctp_association *asoc;
3871 if (optlen < sizeof(struct sctp_paddrthlds))
3873 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3874 sizeof(struct sctp_paddrthlds)))
3878 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3879 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3882 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3884 if (val.spt_pathmaxrxt)
3885 trans->pathmaxrxt = val.spt_pathmaxrxt;
3886 trans->pf_retrans = val.spt_pathpfthld;
3889 if (val.spt_pathmaxrxt)
3890 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3891 asoc->pf_retrans = val.spt_pathpfthld;
3893 trans = sctp_addr_id2transport(sk, &val.spt_address,
3898 if (val.spt_pathmaxrxt)
3899 trans->pathmaxrxt = val.spt_pathmaxrxt;
3900 trans->pf_retrans = val.spt_pathpfthld;
3906 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3907 char __user *optval,
3908 unsigned int optlen)
3912 if (optlen < sizeof(int))
3914 if (get_user(val, (int __user *) optval))
3917 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3922 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3923 char __user *optval,
3924 unsigned int optlen)
3928 if (optlen < sizeof(int))
3930 if (get_user(val, (int __user *) optval))
3933 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3938 static int sctp_setsockopt_pr_supported(struct sock *sk,
3939 char __user *optval,
3940 unsigned int optlen)
3942 struct sctp_assoc_value params;
3944 if (optlen != sizeof(params))
3947 if (copy_from_user(¶ms, optval, optlen))
3950 sctp_sk(sk)->ep->prsctp_enable = !!params.assoc_value;
3955 static int sctp_setsockopt_default_prinfo(struct sock *sk,
3956 char __user *optval,
3957 unsigned int optlen)
3959 struct sctp_default_prinfo info;
3960 struct sctp_association *asoc;
3961 int retval = -EINVAL;
3963 if (optlen != sizeof(info))
3966 if (copy_from_user(&info, optval, sizeof(info))) {
3971 if (info.pr_policy & ~SCTP_PR_SCTP_MASK)
3974 if (info.pr_policy == SCTP_PR_SCTP_NONE)
3977 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
3979 SCTP_PR_SET_POLICY(asoc->default_flags, info.pr_policy);
3980 asoc->default_timetolive = info.pr_value;
3981 } else if (!info.pr_assoc_id) {
3982 struct sctp_sock *sp = sctp_sk(sk);
3984 SCTP_PR_SET_POLICY(sp->default_flags, info.pr_policy);
3985 sp->default_timetolive = info.pr_value;
3996 static int sctp_setsockopt_reconfig_supported(struct sock *sk,
3997 char __user *optval,
3998 unsigned int optlen)
4000 struct sctp_assoc_value params;
4001 struct sctp_association *asoc;
4002 int retval = -EINVAL;
4004 if (optlen != sizeof(params))
4007 if (copy_from_user(¶ms, optval, optlen)) {
4012 asoc = sctp_id2assoc(sk, params.assoc_id);
4014 asoc->reconf_enable = !!params.assoc_value;
4015 } else if (!params.assoc_id) {
4016 struct sctp_sock *sp = sctp_sk(sk);
4018 sp->ep->reconf_enable = !!params.assoc_value;
4029 static int sctp_setsockopt_enable_strreset(struct sock *sk,
4030 char __user *optval,
4031 unsigned int optlen)
4033 struct sctp_assoc_value params;
4034 struct sctp_association *asoc;
4035 int retval = -EINVAL;
4037 if (optlen != sizeof(params))
4040 if (copy_from_user(¶ms, optval, optlen)) {
4045 if (params.assoc_value & (~SCTP_ENABLE_STRRESET_MASK))
4048 asoc = sctp_id2assoc(sk, params.assoc_id);
4050 asoc->strreset_enable = params.assoc_value;
4051 } else if (!params.assoc_id) {
4052 struct sctp_sock *sp = sctp_sk(sk);
4054 sp->ep->strreset_enable = params.assoc_value;
4065 static int sctp_setsockopt_reset_streams(struct sock *sk,
4066 char __user *optval,
4067 unsigned int optlen)
4069 struct sctp_reset_streams *params;
4070 struct sctp_association *asoc;
4071 int retval = -EINVAL;
4073 if (optlen < sizeof(*params))
4075 /* srs_number_streams is u16, so optlen can't be bigger than this. */
4076 optlen = min_t(unsigned int, optlen, USHRT_MAX +
4077 sizeof(__u16) * sizeof(*params));
4079 params = memdup_user(optval, optlen);
4081 return PTR_ERR(params);
4083 if (params->srs_number_streams * sizeof(__u16) >
4084 optlen - sizeof(*params))
4087 asoc = sctp_id2assoc(sk, params->srs_assoc_id);
4091 retval = sctp_send_reset_streams(asoc, params);
4098 static int sctp_setsockopt_reset_assoc(struct sock *sk,
4099 char __user *optval,
4100 unsigned int optlen)
4102 struct sctp_association *asoc;
4103 sctp_assoc_t associd;
4104 int retval = -EINVAL;
4106 if (optlen != sizeof(associd))
4109 if (copy_from_user(&associd, optval, optlen)) {
4114 asoc = sctp_id2assoc(sk, associd);
4118 retval = sctp_send_reset_assoc(asoc);
4124 static int sctp_setsockopt_add_streams(struct sock *sk,
4125 char __user *optval,
4126 unsigned int optlen)
4128 struct sctp_association *asoc;
4129 struct sctp_add_streams params;
4130 int retval = -EINVAL;
4132 if (optlen != sizeof(params))
4135 if (copy_from_user(¶ms, optval, optlen)) {
4140 asoc = sctp_id2assoc(sk, params.sas_assoc_id);
4144 retval = sctp_send_add_streams(asoc, ¶ms);
4150 static int sctp_setsockopt_scheduler(struct sock *sk,
4151 char __user *optval,
4152 unsigned int optlen)
4154 struct sctp_association *asoc;
4155 struct sctp_assoc_value params;
4156 int retval = -EINVAL;
4158 if (optlen < sizeof(params))
4161 optlen = sizeof(params);
4162 if (copy_from_user(¶ms, optval, optlen)) {
4167 if (params.assoc_value > SCTP_SS_MAX)
4170 asoc = sctp_id2assoc(sk, params.assoc_id);
4174 retval = sctp_sched_set_sched(asoc, params.assoc_value);
4180 static int sctp_setsockopt_scheduler_value(struct sock *sk,
4181 char __user *optval,
4182 unsigned int optlen)
4184 struct sctp_association *asoc;
4185 struct sctp_stream_value params;
4186 int retval = -EINVAL;
4188 if (optlen < sizeof(params))
4191 optlen = sizeof(params);
4192 if (copy_from_user(¶ms, optval, optlen)) {
4197 asoc = sctp_id2assoc(sk, params.assoc_id);
4201 retval = sctp_sched_set_value(asoc, params.stream_id,
4202 params.stream_value, GFP_KERNEL);
4208 static int sctp_setsockopt_interleaving_supported(struct sock *sk,
4209 char __user *optval,
4210 unsigned int optlen)
4212 struct sctp_sock *sp = sctp_sk(sk);
4213 struct net *net = sock_net(sk);
4214 struct sctp_assoc_value params;
4215 int retval = -EINVAL;
4217 if (optlen < sizeof(params))
4220 optlen = sizeof(params);
4221 if (copy_from_user(¶ms, optval, optlen)) {
4226 if (params.assoc_id)
4229 if (!net->sctp.intl_enable || !sp->frag_interleave) {
4234 sp->strm_interleave = !!params.assoc_value;
4242 static int sctp_setsockopt_reuse_port(struct sock *sk, char __user *optval,
4243 unsigned int optlen)
4247 if (!sctp_style(sk, TCP))
4250 if (sctp_sk(sk)->ep->base.bind_addr.port)
4253 if (optlen < sizeof(int))
4256 if (get_user(val, (int __user *)optval))
4259 sctp_sk(sk)->reuse = !!val;
4264 /* API 6.2 setsockopt(), getsockopt()
4266 * Applications use setsockopt() and getsockopt() to set or retrieve
4267 * socket options. Socket options are used to change the default
4268 * behavior of sockets calls. They are described in Section 7.
4272 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
4273 * int __user *optlen);
4274 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
4277 * sd - the socket descript.
4278 * level - set to IPPROTO_SCTP for all SCTP options.
4279 * optname - the option name.
4280 * optval - the buffer to store the value of the option.
4281 * optlen - the size of the buffer.
4283 static int sctp_setsockopt(struct sock *sk, int level, int optname,
4284 char __user *optval, unsigned int optlen)
4288 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
4290 /* I can hardly begin to describe how wrong this is. This is
4291 * so broken as to be worse than useless. The API draft
4292 * REALLY is NOT helpful here... I am not convinced that the
4293 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
4294 * are at all well-founded.
4296 if (level != SOL_SCTP) {
4297 struct sctp_af *af = sctp_sk(sk)->pf->af;
4298 retval = af->setsockopt(sk, level, optname, optval, optlen);
4305 case SCTP_SOCKOPT_BINDX_ADD:
4306 /* 'optlen' is the size of the addresses buffer. */
4307 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
4308 optlen, SCTP_BINDX_ADD_ADDR);
4311 case SCTP_SOCKOPT_BINDX_REM:
4312 /* 'optlen' is the size of the addresses buffer. */
4313 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
4314 optlen, SCTP_BINDX_REM_ADDR);
4317 case SCTP_SOCKOPT_CONNECTX_OLD:
4318 /* 'optlen' is the size of the addresses buffer. */
4319 retval = sctp_setsockopt_connectx_old(sk,
4320 (struct sockaddr __user *)optval,
4324 case SCTP_SOCKOPT_CONNECTX:
4325 /* 'optlen' is the size of the addresses buffer. */
4326 retval = sctp_setsockopt_connectx(sk,
4327 (struct sockaddr __user *)optval,
4331 case SCTP_DISABLE_FRAGMENTS:
4332 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
4336 retval = sctp_setsockopt_events(sk, optval, optlen);
4339 case SCTP_AUTOCLOSE:
4340 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
4343 case SCTP_PEER_ADDR_PARAMS:
4344 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
4347 case SCTP_DELAYED_SACK:
4348 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
4350 case SCTP_PARTIAL_DELIVERY_POINT:
4351 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
4355 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
4357 case SCTP_DEFAULT_SEND_PARAM:
4358 retval = sctp_setsockopt_default_send_param(sk, optval,
4361 case SCTP_DEFAULT_SNDINFO:
4362 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
4364 case SCTP_PRIMARY_ADDR:
4365 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
4367 case SCTP_SET_PEER_PRIMARY_ADDR:
4368 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
4371 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
4374 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
4376 case SCTP_ASSOCINFO:
4377 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
4379 case SCTP_I_WANT_MAPPED_V4_ADDR:
4380 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
4383 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
4385 case SCTP_ADAPTATION_LAYER:
4386 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
4389 retval = sctp_setsockopt_context(sk, optval, optlen);
4391 case SCTP_FRAGMENT_INTERLEAVE:
4392 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
4394 case SCTP_MAX_BURST:
4395 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
4397 case SCTP_AUTH_CHUNK:
4398 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
4400 case SCTP_HMAC_IDENT:
4401 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
4404 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
4406 case SCTP_AUTH_ACTIVE_KEY:
4407 retval = sctp_setsockopt_active_key(sk, optval, optlen);
4409 case SCTP_AUTH_DELETE_KEY:
4410 retval = sctp_setsockopt_del_key(sk, optval, optlen);
4412 case SCTP_AUTH_DEACTIVATE_KEY:
4413 retval = sctp_setsockopt_deactivate_key(sk, optval, optlen);
4415 case SCTP_AUTO_ASCONF:
4416 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
4418 case SCTP_PEER_ADDR_THLDS:
4419 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
4421 case SCTP_RECVRCVINFO:
4422 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
4424 case SCTP_RECVNXTINFO:
4425 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
4427 case SCTP_PR_SUPPORTED:
4428 retval = sctp_setsockopt_pr_supported(sk, optval, optlen);
4430 case SCTP_DEFAULT_PRINFO:
4431 retval = sctp_setsockopt_default_prinfo(sk, optval, optlen);
4433 case SCTP_RECONFIG_SUPPORTED:
4434 retval = sctp_setsockopt_reconfig_supported(sk, optval, optlen);
4436 case SCTP_ENABLE_STREAM_RESET:
4437 retval = sctp_setsockopt_enable_strreset(sk, optval, optlen);
4439 case SCTP_RESET_STREAMS:
4440 retval = sctp_setsockopt_reset_streams(sk, optval, optlen);
4442 case SCTP_RESET_ASSOC:
4443 retval = sctp_setsockopt_reset_assoc(sk, optval, optlen);
4445 case SCTP_ADD_STREAMS:
4446 retval = sctp_setsockopt_add_streams(sk, optval, optlen);
4448 case SCTP_STREAM_SCHEDULER:
4449 retval = sctp_setsockopt_scheduler(sk, optval, optlen);
4451 case SCTP_STREAM_SCHEDULER_VALUE:
4452 retval = sctp_setsockopt_scheduler_value(sk, optval, optlen);
4454 case SCTP_INTERLEAVING_SUPPORTED:
4455 retval = sctp_setsockopt_interleaving_supported(sk, optval,
4458 case SCTP_REUSE_PORT:
4459 retval = sctp_setsockopt_reuse_port(sk, optval, optlen);
4462 retval = -ENOPROTOOPT;
4472 /* API 3.1.6 connect() - UDP Style Syntax
4474 * An application may use the connect() call in the UDP model to initiate an
4475 * association without sending data.
4479 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
4481 * sd: the socket descriptor to have a new association added to.
4483 * nam: the address structure (either struct sockaddr_in or struct
4484 * sockaddr_in6 defined in RFC2553 [7]).
4486 * len: the size of the address.
4488 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
4489 int addr_len, int flags)
4491 struct inet_sock *inet = inet_sk(sk);
4497 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
4500 /* We may need to bind the socket. */
4501 if (!inet->inet_num) {
4502 if (sk->sk_prot->get_port(sk, 0)) {
4506 inet->inet_sport = htons(inet->inet_num);
4509 /* Validate addr_len before calling common connect/connectx routine. */
4510 af = sctp_get_af_specific(addr->sa_family);
4511 if (!af || addr_len < af->sockaddr_len) {
4514 /* Pass correct addr len to common routine (so it knows there
4515 * is only one address being passed.
4517 err = __sctp_connect(sk, addr, af->sockaddr_len, flags, NULL);
4524 int sctp_inet_connect(struct socket *sock, struct sockaddr *uaddr,
4525 int addr_len, int flags)
4527 if (addr_len < sizeof(uaddr->sa_family))
4530 if (uaddr->sa_family == AF_UNSPEC)
4533 return sctp_connect(sock->sk, uaddr, addr_len, flags);
4536 /* FIXME: Write comments. */
4537 static int sctp_disconnect(struct sock *sk, int flags)
4539 return -EOPNOTSUPP; /* STUB */
4542 /* 4.1.4 accept() - TCP Style Syntax
4544 * Applications use accept() call to remove an established SCTP
4545 * association from the accept queue of the endpoint. A new socket
4546 * descriptor will be returned from accept() to represent the newly
4547 * formed association.
4549 static struct sock *sctp_accept(struct sock *sk, int flags, int *err, bool kern)
4551 struct sctp_sock *sp;
4552 struct sctp_endpoint *ep;
4553 struct sock *newsk = NULL;
4554 struct sctp_association *asoc;
4563 if (!sctp_style(sk, TCP)) {
4564 error = -EOPNOTSUPP;
4568 if (!sctp_sstate(sk, LISTENING)) {
4573 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
4575 error = sctp_wait_for_accept(sk, timeo);
4579 /* We treat the list of associations on the endpoint as the accept
4580 * queue and pick the first association on the list.
4582 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
4584 newsk = sp->pf->create_accept_sk(sk, asoc, kern);
4590 /* Populate the fields of the newsk from the oldsk and migrate the
4591 * asoc to the newsk.
4593 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
4601 /* The SCTP ioctl handler. */
4602 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
4609 * SEQPACKET-style sockets in LISTENING state are valid, for
4610 * SCTP, so only discard TCP-style sockets in LISTENING state.
4612 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4617 struct sk_buff *skb;
4618 unsigned int amount = 0;
4620 skb = skb_peek(&sk->sk_receive_queue);
4623 * We will only return the amount of this packet since
4624 * that is all that will be read.
4628 rc = put_user(amount, (int __user *)arg);
4640 /* This is the function which gets called during socket creation to
4641 * initialized the SCTP-specific portion of the sock.
4642 * The sock structure should already be zero-filled memory.
4644 static int sctp_init_sock(struct sock *sk)
4646 struct net *net = sock_net(sk);
4647 struct sctp_sock *sp;
4649 pr_debug("%s: sk:%p\n", __func__, sk);
4653 /* Initialize the SCTP per socket area. */
4654 switch (sk->sk_type) {
4655 case SOCK_SEQPACKET:
4656 sp->type = SCTP_SOCKET_UDP;
4659 sp->type = SCTP_SOCKET_TCP;
4662 return -ESOCKTNOSUPPORT;
4665 sk->sk_gso_type = SKB_GSO_SCTP;
4667 /* Initialize default send parameters. These parameters can be
4668 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4670 sp->default_stream = 0;
4671 sp->default_ppid = 0;
4672 sp->default_flags = 0;
4673 sp->default_context = 0;
4674 sp->default_timetolive = 0;
4676 sp->default_rcv_context = 0;
4677 sp->max_burst = net->sctp.max_burst;
4679 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4681 /* Initialize default setup parameters. These parameters
4682 * can be modified with the SCTP_INITMSG socket option or
4683 * overridden by the SCTP_INIT CMSG.
4685 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4686 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4687 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4688 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4690 /* Initialize default RTO related parameters. These parameters can
4691 * be modified for with the SCTP_RTOINFO socket option.
4693 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4694 sp->rtoinfo.srto_max = net->sctp.rto_max;
4695 sp->rtoinfo.srto_min = net->sctp.rto_min;
4697 /* Initialize default association related parameters. These parameters
4698 * can be modified with the SCTP_ASSOCINFO socket option.
4700 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4701 sp->assocparams.sasoc_number_peer_destinations = 0;
4702 sp->assocparams.sasoc_peer_rwnd = 0;
4703 sp->assocparams.sasoc_local_rwnd = 0;
4704 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4706 /* Initialize default event subscriptions. By default, all the
4709 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4711 /* Default Peer Address Parameters. These defaults can
4712 * be modified via SCTP_PEER_ADDR_PARAMS
4714 sp->hbinterval = net->sctp.hb_interval;
4715 sp->pathmaxrxt = net->sctp.max_retrans_path;
4716 sp->pathmtu = 0; /* allow default discovery */
4717 sp->sackdelay = net->sctp.sack_timeout;
4719 sp->param_flags = SPP_HB_ENABLE |
4721 SPP_SACKDELAY_ENABLE;
4723 /* If enabled no SCTP message fragmentation will be performed.
4724 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4726 sp->disable_fragments = 0;
4728 /* Enable Nagle algorithm by default. */
4731 sp->recvrcvinfo = 0;
4732 sp->recvnxtinfo = 0;
4734 /* Enable by default. */
4737 /* Auto-close idle associations after the configured
4738 * number of seconds. A value of 0 disables this
4739 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4740 * for UDP-style sockets only.
4744 /* User specified fragmentation limit. */
4747 sp->adaptation_ind = 0;
4749 sp->pf = sctp_get_pf_specific(sk->sk_family);
4751 /* Control variables for partial data delivery. */
4752 atomic_set(&sp->pd_mode, 0);
4753 skb_queue_head_init(&sp->pd_lobby);
4754 sp->frag_interleave = 0;
4756 /* Create a per socket endpoint structure. Even if we
4757 * change the data structure relationships, this may still
4758 * be useful for storing pre-connect address information.
4760 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4766 sk->sk_destruct = sctp_destruct_sock;
4768 SCTP_DBG_OBJCNT_INC(sock);
4771 sk_sockets_allocated_inc(sk);
4772 sock_prot_inuse_add(net, sk->sk_prot, 1);
4774 /* Nothing can fail after this block, otherwise
4775 * sctp_destroy_sock() will be called without addr_wq_lock held
4777 if (net->sctp.default_auto_asconf) {
4778 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
4779 list_add_tail(&sp->auto_asconf_list,
4780 &net->sctp.auto_asconf_splist);
4781 sp->do_auto_asconf = 1;
4782 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4784 sp->do_auto_asconf = 0;
4792 /* Cleanup any SCTP per socket resources. Must be called with
4793 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4795 static void sctp_destroy_sock(struct sock *sk)
4797 struct sctp_sock *sp;
4799 pr_debug("%s: sk:%p\n", __func__, sk);
4801 /* Release our hold on the endpoint. */
4803 /* This could happen during socket init, thus we bail out
4804 * early, since the rest of the below is not setup either.
4809 if (sp->do_auto_asconf) {
4810 sp->do_auto_asconf = 0;
4811 list_del(&sp->auto_asconf_list);
4813 sctp_endpoint_free(sp->ep);
4815 sk_sockets_allocated_dec(sk);
4816 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4820 /* Triggered when there are no references on the socket anymore */
4821 static void sctp_destruct_sock(struct sock *sk)
4823 struct sctp_sock *sp = sctp_sk(sk);
4825 /* Free up the HMAC transform. */
4826 crypto_free_shash(sp->hmac);
4828 inet_sock_destruct(sk);
4831 /* API 4.1.7 shutdown() - TCP Style Syntax
4832 * int shutdown(int socket, int how);
4834 * sd - the socket descriptor of the association to be closed.
4835 * how - Specifies the type of shutdown. The values are
4838 * Disables further receive operations. No SCTP
4839 * protocol action is taken.
4841 * Disables further send operations, and initiates
4842 * the SCTP shutdown sequence.
4844 * Disables further send and receive operations
4845 * and initiates the SCTP shutdown sequence.
4847 static void sctp_shutdown(struct sock *sk, int how)
4849 struct net *net = sock_net(sk);
4850 struct sctp_endpoint *ep;
4852 if (!sctp_style(sk, TCP))
4855 ep = sctp_sk(sk)->ep;
4856 if (how & SEND_SHUTDOWN && !list_empty(&ep->asocs)) {
4857 struct sctp_association *asoc;
4859 inet_sk_set_state(sk, SCTP_SS_CLOSING);
4860 asoc = list_entry(ep->asocs.next,
4861 struct sctp_association, asocs);
4862 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4866 int sctp_get_sctp_info(struct sock *sk, struct sctp_association *asoc,
4867 struct sctp_info *info)
4869 struct sctp_transport *prim;
4870 struct list_head *pos;
4873 memset(info, 0, sizeof(*info));
4875 struct sctp_sock *sp = sctp_sk(sk);
4877 info->sctpi_s_autoclose = sp->autoclose;
4878 info->sctpi_s_adaptation_ind = sp->adaptation_ind;
4879 info->sctpi_s_pd_point = sp->pd_point;
4880 info->sctpi_s_nodelay = sp->nodelay;
4881 info->sctpi_s_disable_fragments = sp->disable_fragments;
4882 info->sctpi_s_v4mapped = sp->v4mapped;
4883 info->sctpi_s_frag_interleave = sp->frag_interleave;
4884 info->sctpi_s_type = sp->type;
4889 info->sctpi_tag = asoc->c.my_vtag;
4890 info->sctpi_state = asoc->state;
4891 info->sctpi_rwnd = asoc->a_rwnd;
4892 info->sctpi_unackdata = asoc->unack_data;
4893 info->sctpi_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4894 info->sctpi_instrms = asoc->stream.incnt;
4895 info->sctpi_outstrms = asoc->stream.outcnt;
4896 list_for_each(pos, &asoc->base.inqueue.in_chunk_list)
4897 info->sctpi_inqueue++;
4898 list_for_each(pos, &asoc->outqueue.out_chunk_list)
4899 info->sctpi_outqueue++;
4900 info->sctpi_overall_error = asoc->overall_error_count;
4901 info->sctpi_max_burst = asoc->max_burst;
4902 info->sctpi_maxseg = asoc->frag_point;
4903 info->sctpi_peer_rwnd = asoc->peer.rwnd;
4904 info->sctpi_peer_tag = asoc->c.peer_vtag;
4906 mask = asoc->peer.ecn_capable << 1;
4907 mask = (mask | asoc->peer.ipv4_address) << 1;
4908 mask = (mask | asoc->peer.ipv6_address) << 1;
4909 mask = (mask | asoc->peer.hostname_address) << 1;
4910 mask = (mask | asoc->peer.asconf_capable) << 1;
4911 mask = (mask | asoc->peer.prsctp_capable) << 1;
4912 mask = (mask | asoc->peer.auth_capable);
4913 info->sctpi_peer_capable = mask;
4914 mask = asoc->peer.sack_needed << 1;
4915 mask = (mask | asoc->peer.sack_generation) << 1;
4916 mask = (mask | asoc->peer.zero_window_announced);
4917 info->sctpi_peer_sack = mask;
4919 info->sctpi_isacks = asoc->stats.isacks;
4920 info->sctpi_osacks = asoc->stats.osacks;
4921 info->sctpi_opackets = asoc->stats.opackets;
4922 info->sctpi_ipackets = asoc->stats.ipackets;
4923 info->sctpi_rtxchunks = asoc->stats.rtxchunks;
4924 info->sctpi_outofseqtsns = asoc->stats.outofseqtsns;
4925 info->sctpi_idupchunks = asoc->stats.idupchunks;
4926 info->sctpi_gapcnt = asoc->stats.gapcnt;
4927 info->sctpi_ouodchunks = asoc->stats.ouodchunks;
4928 info->sctpi_iuodchunks = asoc->stats.iuodchunks;
4929 info->sctpi_oodchunks = asoc->stats.oodchunks;
4930 info->sctpi_iodchunks = asoc->stats.iodchunks;
4931 info->sctpi_octrlchunks = asoc->stats.octrlchunks;
4932 info->sctpi_ictrlchunks = asoc->stats.ictrlchunks;
4934 prim = asoc->peer.primary_path;
4935 memcpy(&info->sctpi_p_address, &prim->ipaddr, sizeof(prim->ipaddr));
4936 info->sctpi_p_state = prim->state;
4937 info->sctpi_p_cwnd = prim->cwnd;
4938 info->sctpi_p_srtt = prim->srtt;
4939 info->sctpi_p_rto = jiffies_to_msecs(prim->rto);
4940 info->sctpi_p_hbinterval = prim->hbinterval;
4941 info->sctpi_p_pathmaxrxt = prim->pathmaxrxt;
4942 info->sctpi_p_sackdelay = jiffies_to_msecs(prim->sackdelay);
4943 info->sctpi_p_ssthresh = prim->ssthresh;
4944 info->sctpi_p_partial_bytes_acked = prim->partial_bytes_acked;
4945 info->sctpi_p_flight_size = prim->flight_size;
4946 info->sctpi_p_error = prim->error_count;
4950 EXPORT_SYMBOL_GPL(sctp_get_sctp_info);
4952 /* use callback to avoid exporting the core structure */
4953 void sctp_transport_walk_start(struct rhashtable_iter *iter)
4955 rhltable_walk_enter(&sctp_transport_hashtable, iter);
4957 rhashtable_walk_start(iter);
4960 void sctp_transport_walk_stop(struct rhashtable_iter *iter)
4962 rhashtable_walk_stop(iter);
4963 rhashtable_walk_exit(iter);
4966 struct sctp_transport *sctp_transport_get_next(struct net *net,
4967 struct rhashtable_iter *iter)
4969 struct sctp_transport *t;
4971 t = rhashtable_walk_next(iter);
4972 for (; t; t = rhashtable_walk_next(iter)) {
4974 if (PTR_ERR(t) == -EAGAIN)
4979 if (!sctp_transport_hold(t))
4982 if (net_eq(sock_net(t->asoc->base.sk), net) &&
4983 t->asoc->peer.primary_path == t)
4986 sctp_transport_put(t);
4992 struct sctp_transport *sctp_transport_get_idx(struct net *net,
4993 struct rhashtable_iter *iter,
4996 struct sctp_transport *t;
4999 return SEQ_START_TOKEN;
5001 while ((t = sctp_transport_get_next(net, iter)) && !IS_ERR(t)) {
5004 sctp_transport_put(t);
5010 int sctp_for_each_endpoint(int (*cb)(struct sctp_endpoint *, void *),
5014 struct sctp_ep_common *epb;
5015 struct sctp_hashbucket *head;
5017 for (head = sctp_ep_hashtable; hash < sctp_ep_hashsize;
5019 read_lock_bh(&head->lock);
5020 sctp_for_each_hentry(epb, &head->chain) {
5021 err = cb(sctp_ep(epb), p);
5025 read_unlock_bh(&head->lock);
5030 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint);
5032 int sctp_transport_lookup_process(int (*cb)(struct sctp_transport *, void *),
5034 const union sctp_addr *laddr,
5035 const union sctp_addr *paddr, void *p)
5037 struct sctp_transport *transport;
5041 transport = sctp_addrs_lookup_transport(net, laddr, paddr);
5046 err = cb(transport, p);
5047 sctp_transport_put(transport);
5051 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process);
5053 int sctp_for_each_transport(int (*cb)(struct sctp_transport *, void *),
5054 int (*cb_done)(struct sctp_transport *, void *),
5055 struct net *net, int *pos, void *p) {
5056 struct rhashtable_iter hti;
5057 struct sctp_transport *tsp;
5062 sctp_transport_walk_start(&hti);
5064 tsp = sctp_transport_get_idx(net, &hti, *pos + 1);
5065 for (; !IS_ERR_OR_NULL(tsp); tsp = sctp_transport_get_next(net, &hti)) {
5070 sctp_transport_put(tsp);
5072 sctp_transport_walk_stop(&hti);
5075 if (cb_done && !cb_done(tsp, p)) {
5077 sctp_transport_put(tsp);
5080 sctp_transport_put(tsp);
5085 EXPORT_SYMBOL_GPL(sctp_for_each_transport);
5087 /* 7.2.1 Association Status (SCTP_STATUS)
5089 * Applications can retrieve current status information about an
5090 * association, including association state, peer receiver window size,
5091 * number of unacked data chunks, and number of data chunks pending
5092 * receipt. This information is read-only.
5094 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
5095 char __user *optval,
5098 struct sctp_status status;
5099 struct sctp_association *asoc = NULL;
5100 struct sctp_transport *transport;
5101 sctp_assoc_t associd;
5104 if (len < sizeof(status)) {
5109 len = sizeof(status);
5110 if (copy_from_user(&status, optval, len)) {
5115 associd = status.sstat_assoc_id;
5116 asoc = sctp_id2assoc(sk, associd);
5122 transport = asoc->peer.primary_path;
5124 status.sstat_assoc_id = sctp_assoc2id(asoc);
5125 status.sstat_state = sctp_assoc_to_state(asoc);
5126 status.sstat_rwnd = asoc->peer.rwnd;
5127 status.sstat_unackdata = asoc->unack_data;
5129 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
5130 status.sstat_instrms = asoc->stream.incnt;
5131 status.sstat_outstrms = asoc->stream.outcnt;
5132 status.sstat_fragmentation_point = asoc->frag_point;
5133 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
5134 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
5135 transport->af_specific->sockaddr_len);
5136 /* Map ipv4 address into v4-mapped-on-v6 address. */
5137 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
5138 (union sctp_addr *)&status.sstat_primary.spinfo_address);
5139 status.sstat_primary.spinfo_state = transport->state;
5140 status.sstat_primary.spinfo_cwnd = transport->cwnd;
5141 status.sstat_primary.spinfo_srtt = transport->srtt;
5142 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
5143 status.sstat_primary.spinfo_mtu = transport->pathmtu;
5145 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
5146 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
5148 if (put_user(len, optlen)) {
5153 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
5154 __func__, len, status.sstat_state, status.sstat_rwnd,
5155 status.sstat_assoc_id);
5157 if (copy_to_user(optval, &status, len)) {
5167 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
5169 * Applications can retrieve information about a specific peer address
5170 * of an association, including its reachability state, congestion
5171 * window, and retransmission timer values. This information is
5174 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
5175 char __user *optval,
5178 struct sctp_paddrinfo pinfo;
5179 struct sctp_transport *transport;
5182 if (len < sizeof(pinfo)) {
5187 len = sizeof(pinfo);
5188 if (copy_from_user(&pinfo, optval, len)) {
5193 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
5194 pinfo.spinfo_assoc_id);
5198 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
5199 pinfo.spinfo_state = transport->state;
5200 pinfo.spinfo_cwnd = transport->cwnd;
5201 pinfo.spinfo_srtt = transport->srtt;
5202 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
5203 pinfo.spinfo_mtu = transport->pathmtu;
5205 if (pinfo.spinfo_state == SCTP_UNKNOWN)
5206 pinfo.spinfo_state = SCTP_ACTIVE;
5208 if (put_user(len, optlen)) {
5213 if (copy_to_user(optval, &pinfo, len)) {
5222 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
5224 * This option is a on/off flag. If enabled no SCTP message
5225 * fragmentation will be performed. Instead if a message being sent
5226 * exceeds the current PMTU size, the message will NOT be sent and
5227 * instead a error will be indicated to the user.
5229 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
5230 char __user *optval, int __user *optlen)
5234 if (len < sizeof(int))
5238 val = (sctp_sk(sk)->disable_fragments == 1);
5239 if (put_user(len, optlen))
5241 if (copy_to_user(optval, &val, len))
5246 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
5248 * This socket option is used to specify various notifications and
5249 * ancillary data the user wishes to receive.
5251 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
5256 if (len > sizeof(struct sctp_event_subscribe))
5257 len = sizeof(struct sctp_event_subscribe);
5258 if (put_user(len, optlen))
5260 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
5265 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
5267 * This socket option is applicable to the UDP-style socket only. When
5268 * set it will cause associations that are idle for more than the
5269 * specified number of seconds to automatically close. An association
5270 * being idle is defined an association that has NOT sent or received
5271 * user data. The special value of '0' indicates that no automatic
5272 * close of any associations should be performed. The option expects an
5273 * integer defining the number of seconds of idle time before an
5274 * association is closed.
5276 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
5278 /* Applicable to UDP-style socket only */
5279 if (sctp_style(sk, TCP))
5281 if (len < sizeof(int))
5284 if (put_user(len, optlen))
5286 if (put_user(sctp_sk(sk)->autoclose, (int __user *)optval))
5291 /* Helper routine to branch off an association to a new socket. */
5292 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
5294 struct sctp_association *asoc = sctp_id2assoc(sk, id);
5295 struct sctp_sock *sp = sctp_sk(sk);
5296 struct socket *sock;
5299 /* Do not peel off from one netns to another one. */
5300 if (!net_eq(current->nsproxy->net_ns, sock_net(sk)))
5306 /* An association cannot be branched off from an already peeled-off
5307 * socket, nor is this supported for tcp style sockets.
5309 if (!sctp_style(sk, UDP))
5312 /* Create a new socket. */
5313 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
5317 sctp_copy_sock(sock->sk, sk, asoc);
5319 /* Make peeled-off sockets more like 1-1 accepted sockets.
5320 * Set the daddr and initialize id to something more random and also
5321 * copy over any ip options.
5323 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
5324 sp->pf->copy_ip_options(sk, sock->sk);
5326 /* Populate the fields of the newsk from the oldsk and migrate the
5327 * asoc to the newsk.
5329 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
5335 EXPORT_SYMBOL(sctp_do_peeloff);
5337 static int sctp_getsockopt_peeloff_common(struct sock *sk, sctp_peeloff_arg_t *peeloff,
5338 struct file **newfile, unsigned flags)
5340 struct socket *newsock;
5343 retval = sctp_do_peeloff(sk, peeloff->associd, &newsock);
5347 /* Map the socket to an unused fd that can be returned to the user. */
5348 retval = get_unused_fd_flags(flags & SOCK_CLOEXEC);
5350 sock_release(newsock);
5354 *newfile = sock_alloc_file(newsock, 0, NULL);
5355 if (IS_ERR(*newfile)) {
5356 put_unused_fd(retval);
5357 retval = PTR_ERR(*newfile);
5362 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
5365 peeloff->sd = retval;
5367 if (flags & SOCK_NONBLOCK)
5368 (*newfile)->f_flags |= O_NONBLOCK;
5373 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
5375 sctp_peeloff_arg_t peeloff;
5376 struct file *newfile = NULL;
5379 if (len < sizeof(sctp_peeloff_arg_t))
5381 len = sizeof(sctp_peeloff_arg_t);
5382 if (copy_from_user(&peeloff, optval, len))
5385 retval = sctp_getsockopt_peeloff_common(sk, &peeloff, &newfile, 0);
5389 /* Return the fd mapped to the new socket. */
5390 if (put_user(len, optlen)) {
5392 put_unused_fd(retval);
5396 if (copy_to_user(optval, &peeloff, len)) {
5398 put_unused_fd(retval);
5401 fd_install(retval, newfile);
5406 static int sctp_getsockopt_peeloff_flags(struct sock *sk, int len,
5407 char __user *optval, int __user *optlen)
5409 sctp_peeloff_flags_arg_t peeloff;
5410 struct file *newfile = NULL;
5413 if (len < sizeof(sctp_peeloff_flags_arg_t))
5415 len = sizeof(sctp_peeloff_flags_arg_t);
5416 if (copy_from_user(&peeloff, optval, len))
5419 retval = sctp_getsockopt_peeloff_common(sk, &peeloff.p_arg,
5420 &newfile, peeloff.flags);
5424 /* Return the fd mapped to the new socket. */
5425 if (put_user(len, optlen)) {
5427 put_unused_fd(retval);
5431 if (copy_to_user(optval, &peeloff, len)) {
5433 put_unused_fd(retval);
5436 fd_install(retval, newfile);
5441 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
5443 * Applications can enable or disable heartbeats for any peer address of
5444 * an association, modify an address's heartbeat interval, force a
5445 * heartbeat to be sent immediately, and adjust the address's maximum
5446 * number of retransmissions sent before an address is considered
5447 * unreachable. The following structure is used to access and modify an
5448 * address's parameters:
5450 * struct sctp_paddrparams {
5451 * sctp_assoc_t spp_assoc_id;
5452 * struct sockaddr_storage spp_address;
5453 * uint32_t spp_hbinterval;
5454 * uint16_t spp_pathmaxrxt;
5455 * uint32_t spp_pathmtu;
5456 * uint32_t spp_sackdelay;
5457 * uint32_t spp_flags;
5460 * spp_assoc_id - (one-to-many style socket) This is filled in the
5461 * application, and identifies the association for
5463 * spp_address - This specifies which address is of interest.
5464 * spp_hbinterval - This contains the value of the heartbeat interval,
5465 * in milliseconds. If a value of zero
5466 * is present in this field then no changes are to
5467 * be made to this parameter.
5468 * spp_pathmaxrxt - This contains the maximum number of
5469 * retransmissions before this address shall be
5470 * considered unreachable. If a value of zero
5471 * is present in this field then no changes are to
5472 * be made to this parameter.
5473 * spp_pathmtu - When Path MTU discovery is disabled the value
5474 * specified here will be the "fixed" path mtu.
5475 * Note that if the spp_address field is empty
5476 * then all associations on this address will
5477 * have this fixed path mtu set upon them.
5479 * spp_sackdelay - When delayed sack is enabled, this value specifies
5480 * the number of milliseconds that sacks will be delayed
5481 * for. This value will apply to all addresses of an
5482 * association if the spp_address field is empty. Note
5483 * also, that if delayed sack is enabled and this
5484 * value is set to 0, no change is made to the last
5485 * recorded delayed sack timer value.
5487 * spp_flags - These flags are used to control various features
5488 * on an association. The flag field may contain
5489 * zero or more of the following options.
5491 * SPP_HB_ENABLE - Enable heartbeats on the
5492 * specified address. Note that if the address
5493 * field is empty all addresses for the association
5494 * have heartbeats enabled upon them.
5496 * SPP_HB_DISABLE - Disable heartbeats on the
5497 * speicifed address. Note that if the address
5498 * field is empty all addresses for the association
5499 * will have their heartbeats disabled. Note also
5500 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
5501 * mutually exclusive, only one of these two should
5502 * be specified. Enabling both fields will have
5503 * undetermined results.
5505 * SPP_HB_DEMAND - Request a user initiated heartbeat
5506 * to be made immediately.
5508 * SPP_PMTUD_ENABLE - This field will enable PMTU
5509 * discovery upon the specified address. Note that
5510 * if the address feild is empty then all addresses
5511 * on the association are effected.
5513 * SPP_PMTUD_DISABLE - This field will disable PMTU
5514 * discovery upon the specified address. Note that
5515 * if the address feild is empty then all addresses
5516 * on the association are effected. Not also that
5517 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
5518 * exclusive. Enabling both will have undetermined
5521 * SPP_SACKDELAY_ENABLE - Setting this flag turns
5522 * on delayed sack. The time specified in spp_sackdelay
5523 * is used to specify the sack delay for this address. Note
5524 * that if spp_address is empty then all addresses will
5525 * enable delayed sack and take on the sack delay
5526 * value specified in spp_sackdelay.
5527 * SPP_SACKDELAY_DISABLE - Setting this flag turns
5528 * off delayed sack. If the spp_address field is blank then
5529 * delayed sack is disabled for the entire association. Note
5530 * also that this field is mutually exclusive to
5531 * SPP_SACKDELAY_ENABLE, setting both will have undefined
5534 * SPP_IPV6_FLOWLABEL: Setting this flag enables the
5535 * setting of the IPV6 flow label value. The value is
5536 * contained in the spp_ipv6_flowlabel field.
5537 * Upon retrieval, this flag will be set to indicate that
5538 * the spp_ipv6_flowlabel field has a valid value returned.
5539 * If a specific destination address is set (in the
5540 * spp_address field), then the value returned is that of
5541 * the address. If just an association is specified (and
5542 * no address), then the association's default flow label
5543 * is returned. If neither an association nor a destination
5544 * is specified, then the socket's default flow label is
5545 * returned. For non-IPv6 sockets, this flag will be left
5548 * SPP_DSCP: Setting this flag enables the setting of the
5549 * Differentiated Services Code Point (DSCP) value
5550 * associated with either the association or a specific
5551 * address. The value is obtained in the spp_dscp field.
5552 * Upon retrieval, this flag will be set to indicate that
5553 * the spp_dscp field has a valid value returned. If a
5554 * specific destination address is set when called (in the
5555 * spp_address field), then that specific destination
5556 * address's DSCP value is returned. If just an association
5557 * is specified, then the association's default DSCP is
5558 * returned. If neither an association nor a destination is
5559 * specified, then the socket's default DSCP is returned.
5561 * spp_ipv6_flowlabel
5562 * - This field is used in conjunction with the
5563 * SPP_IPV6_FLOWLABEL flag and contains the IPv6 flow label.
5564 * The 20 least significant bits are used for the flow
5565 * label. This setting has precedence over any IPv6-layer
5568 * spp_dscp - This field is used in conjunction with the SPP_DSCP flag
5569 * and contains the DSCP. The 6 most significant bits are
5570 * used for the DSCP. This setting has precedence over any
5571 * IPv4- or IPv6- layer setting.
5573 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
5574 char __user *optval, int __user *optlen)
5576 struct sctp_paddrparams params;
5577 struct sctp_transport *trans = NULL;
5578 struct sctp_association *asoc = NULL;
5579 struct sctp_sock *sp = sctp_sk(sk);
5581 if (len >= sizeof(params))
5582 len = sizeof(params);
5583 else if (len >= ALIGN(offsetof(struct sctp_paddrparams,
5584 spp_ipv6_flowlabel), 4))
5585 len = ALIGN(offsetof(struct sctp_paddrparams,
5586 spp_ipv6_flowlabel), 4);
5590 if (copy_from_user(¶ms, optval, len))
5593 /* If an address other than INADDR_ANY is specified, and
5594 * no transport is found, then the request is invalid.
5596 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
5597 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
5598 params.spp_assoc_id);
5600 pr_debug("%s: failed no transport\n", __func__);
5605 /* Get association, if assoc_id != 0 and the socket is a one
5606 * to many style socket, and an association was not found, then
5607 * the id was invalid.
5609 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
5610 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
5611 pr_debug("%s: failed no association\n", __func__);
5616 /* Fetch transport values. */
5617 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
5618 params.spp_pathmtu = trans->pathmtu;
5619 params.spp_pathmaxrxt = trans->pathmaxrxt;
5620 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
5622 /*draft-11 doesn't say what to return in spp_flags*/
5623 params.spp_flags = trans->param_flags;
5624 if (trans->flowlabel & SCTP_FLOWLABEL_SET_MASK) {
5625 params.spp_ipv6_flowlabel = trans->flowlabel &
5626 SCTP_FLOWLABEL_VAL_MASK;
5627 params.spp_flags |= SPP_IPV6_FLOWLABEL;
5629 if (trans->dscp & SCTP_DSCP_SET_MASK) {
5630 params.spp_dscp = trans->dscp & SCTP_DSCP_VAL_MASK;
5631 params.spp_flags |= SPP_DSCP;
5634 /* Fetch association values. */
5635 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
5636 params.spp_pathmtu = asoc->pathmtu;
5637 params.spp_pathmaxrxt = asoc->pathmaxrxt;
5638 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
5640 /*draft-11 doesn't say what to return in spp_flags*/
5641 params.spp_flags = asoc->param_flags;
5642 if (asoc->flowlabel & SCTP_FLOWLABEL_SET_MASK) {
5643 params.spp_ipv6_flowlabel = asoc->flowlabel &
5644 SCTP_FLOWLABEL_VAL_MASK;
5645 params.spp_flags |= SPP_IPV6_FLOWLABEL;
5647 if (asoc->dscp & SCTP_DSCP_SET_MASK) {
5648 params.spp_dscp = asoc->dscp & SCTP_DSCP_VAL_MASK;
5649 params.spp_flags |= SPP_DSCP;
5652 /* Fetch socket values. */
5653 params.spp_hbinterval = sp->hbinterval;
5654 params.spp_pathmtu = sp->pathmtu;
5655 params.spp_sackdelay = sp->sackdelay;
5656 params.spp_pathmaxrxt = sp->pathmaxrxt;
5658 /*draft-11 doesn't say what to return in spp_flags*/
5659 params.spp_flags = sp->param_flags;
5660 if (sp->flowlabel & SCTP_FLOWLABEL_SET_MASK) {
5661 params.spp_ipv6_flowlabel = sp->flowlabel &
5662 SCTP_FLOWLABEL_VAL_MASK;
5663 params.spp_flags |= SPP_IPV6_FLOWLABEL;
5665 if (sp->dscp & SCTP_DSCP_SET_MASK) {
5666 params.spp_dscp = sp->dscp & SCTP_DSCP_VAL_MASK;
5667 params.spp_flags |= SPP_DSCP;
5671 if (copy_to_user(optval, ¶ms, len))
5674 if (put_user(len, optlen))
5681 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
5683 * This option will effect the way delayed acks are performed. This
5684 * option allows you to get or set the delayed ack time, in
5685 * milliseconds. It also allows changing the delayed ack frequency.
5686 * Changing the frequency to 1 disables the delayed sack algorithm. If
5687 * the assoc_id is 0, then this sets or gets the endpoints default
5688 * values. If the assoc_id field is non-zero, then the set or get
5689 * effects the specified association for the one to many model (the
5690 * assoc_id field is ignored by the one to one model). Note that if
5691 * sack_delay or sack_freq are 0 when setting this option, then the
5692 * current values will remain unchanged.
5694 * struct sctp_sack_info {
5695 * sctp_assoc_t sack_assoc_id;
5696 * uint32_t sack_delay;
5697 * uint32_t sack_freq;
5700 * sack_assoc_id - This parameter, indicates which association the user
5701 * is performing an action upon. Note that if this field's value is
5702 * zero then the endpoints default value is changed (effecting future
5703 * associations only).
5705 * sack_delay - This parameter contains the number of milliseconds that
5706 * the user is requesting the delayed ACK timer be set to. Note that
5707 * this value is defined in the standard to be between 200 and 500
5710 * sack_freq - This parameter contains the number of packets that must
5711 * be received before a sack is sent without waiting for the delay
5712 * timer to expire. The default value for this is 2, setting this
5713 * value to 1 will disable the delayed sack algorithm.
5715 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
5716 char __user *optval,
5719 struct sctp_sack_info params;
5720 struct sctp_association *asoc = NULL;
5721 struct sctp_sock *sp = sctp_sk(sk);
5723 if (len >= sizeof(struct sctp_sack_info)) {
5724 len = sizeof(struct sctp_sack_info);
5726 if (copy_from_user(¶ms, optval, len))
5728 } else if (len == sizeof(struct sctp_assoc_value)) {
5729 pr_warn_ratelimited(DEPRECATED
5731 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
5732 "Use struct sctp_sack_info instead\n",
5733 current->comm, task_pid_nr(current));
5734 if (copy_from_user(¶ms, optval, len))
5739 /* Get association, if sack_assoc_id != 0 and the socket is a one
5740 * to many style socket, and an association was not found, then
5741 * the id was invalid.
5743 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
5744 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
5748 /* Fetch association values. */
5749 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
5750 params.sack_delay = jiffies_to_msecs(
5752 params.sack_freq = asoc->sackfreq;
5755 params.sack_delay = 0;
5756 params.sack_freq = 1;
5759 /* Fetch socket values. */
5760 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
5761 params.sack_delay = sp->sackdelay;
5762 params.sack_freq = sp->sackfreq;
5764 params.sack_delay = 0;
5765 params.sack_freq = 1;
5769 if (copy_to_user(optval, ¶ms, len))
5772 if (put_user(len, optlen))
5778 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5780 * Applications can specify protocol parameters for the default association
5781 * initialization. The option name argument to setsockopt() and getsockopt()
5784 * Setting initialization parameters is effective only on an unconnected
5785 * socket (for UDP-style sockets only future associations are effected
5786 * by the change). With TCP-style sockets, this option is inherited by
5787 * sockets derived from a listener socket.
5789 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
5791 if (len < sizeof(struct sctp_initmsg))
5793 len = sizeof(struct sctp_initmsg);
5794 if (put_user(len, optlen))
5796 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
5802 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
5803 char __user *optval, int __user *optlen)
5805 struct sctp_association *asoc;
5807 struct sctp_getaddrs getaddrs;
5808 struct sctp_transport *from;
5810 union sctp_addr temp;
5811 struct sctp_sock *sp = sctp_sk(sk);
5816 if (len < sizeof(struct sctp_getaddrs))
5819 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5822 /* For UDP-style sockets, id specifies the association to query. */
5823 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5827 to = optval + offsetof(struct sctp_getaddrs, addrs);
5828 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5830 list_for_each_entry(from, &asoc->peer.transport_addr_list,
5832 memcpy(&temp, &from->ipaddr, sizeof(temp));
5833 addrlen = sctp_get_pf_specific(sk->sk_family)
5834 ->addr_to_user(sp, &temp);
5835 if (space_left < addrlen)
5837 if (copy_to_user(to, &temp, addrlen))
5841 space_left -= addrlen;
5844 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
5846 bytes_copied = ((char __user *)to) - optval;
5847 if (put_user(bytes_copied, optlen))
5853 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
5854 size_t space_left, int *bytes_copied)
5856 struct sctp_sockaddr_entry *addr;
5857 union sctp_addr temp;
5860 struct net *net = sock_net(sk);
5863 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
5867 if ((PF_INET == sk->sk_family) &&
5868 (AF_INET6 == addr->a.sa.sa_family))
5870 if ((PF_INET6 == sk->sk_family) &&
5871 inet_v6_ipv6only(sk) &&
5872 (AF_INET == addr->a.sa.sa_family))
5874 memcpy(&temp, &addr->a, sizeof(temp));
5875 if (!temp.v4.sin_port)
5876 temp.v4.sin_port = htons(port);
5878 addrlen = sctp_get_pf_specific(sk->sk_family)
5879 ->addr_to_user(sctp_sk(sk), &temp);
5881 if (space_left < addrlen) {
5885 memcpy(to, &temp, addrlen);
5889 space_left -= addrlen;
5890 *bytes_copied += addrlen;
5898 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
5899 char __user *optval, int __user *optlen)
5901 struct sctp_bind_addr *bp;
5902 struct sctp_association *asoc;
5904 struct sctp_getaddrs getaddrs;
5905 struct sctp_sockaddr_entry *addr;
5907 union sctp_addr temp;
5908 struct sctp_sock *sp = sctp_sk(sk);
5912 int bytes_copied = 0;
5916 if (len < sizeof(struct sctp_getaddrs))
5919 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5923 * For UDP-style sockets, id specifies the association to query.
5924 * If the id field is set to the value '0' then the locally bound
5925 * addresses are returned without regard to any particular
5928 if (0 == getaddrs.assoc_id) {
5929 bp = &sctp_sk(sk)->ep->base.bind_addr;
5931 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5934 bp = &asoc->base.bind_addr;
5937 to = optval + offsetof(struct sctp_getaddrs, addrs);
5938 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5940 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN);
5944 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5945 * addresses from the global local address list.
5947 if (sctp_list_single_entry(&bp->address_list)) {
5948 addr = list_entry(bp->address_list.next,
5949 struct sctp_sockaddr_entry, list);
5950 if (sctp_is_any(sk, &addr->a)) {
5951 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
5952 space_left, &bytes_copied);
5962 /* Protection on the bound address list is not needed since
5963 * in the socket option context we hold a socket lock and
5964 * thus the bound address list can't change.
5966 list_for_each_entry(addr, &bp->address_list, list) {
5967 memcpy(&temp, &addr->a, sizeof(temp));
5968 addrlen = sctp_get_pf_specific(sk->sk_family)
5969 ->addr_to_user(sp, &temp);
5970 if (space_left < addrlen) {
5971 err = -ENOMEM; /*fixme: right error?*/
5974 memcpy(buf, &temp, addrlen);
5976 bytes_copied += addrlen;
5978 space_left -= addrlen;
5982 if (copy_to_user(to, addrs, bytes_copied)) {
5986 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
5990 /* XXX: We should have accounted for sizeof(struct sctp_getaddrs) too,
5991 * but we can't change it anymore.
5993 if (put_user(bytes_copied, optlen))
6000 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
6002 * Requests that the local SCTP stack use the enclosed peer address as
6003 * the association primary. The enclosed address must be one of the
6004 * association peer's addresses.
6006 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
6007 char __user *optval, int __user *optlen)
6009 struct sctp_prim prim;
6010 struct sctp_association *asoc;
6011 struct sctp_sock *sp = sctp_sk(sk);
6013 if (len < sizeof(struct sctp_prim))
6016 len = sizeof(struct sctp_prim);
6018 if (copy_from_user(&prim, optval, len))
6021 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
6025 if (!asoc->peer.primary_path)
6028 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
6029 asoc->peer.primary_path->af_specific->sockaddr_len);
6031 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
6032 (union sctp_addr *)&prim.ssp_addr);
6034 if (put_user(len, optlen))
6036 if (copy_to_user(optval, &prim, len))
6043 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
6045 * Requests that the local endpoint set the specified Adaptation Layer
6046 * Indication parameter for all future INIT and INIT-ACK exchanges.
6048 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
6049 char __user *optval, int __user *optlen)
6051 struct sctp_setadaptation adaptation;
6053 if (len < sizeof(struct sctp_setadaptation))
6056 len = sizeof(struct sctp_setadaptation);
6058 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
6060 if (put_user(len, optlen))
6062 if (copy_to_user(optval, &adaptation, len))
6070 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
6072 * Applications that wish to use the sendto() system call may wish to
6073 * specify a default set of parameters that would normally be supplied
6074 * through the inclusion of ancillary data. This socket option allows
6075 * such an application to set the default sctp_sndrcvinfo structure.
6078 * The application that wishes to use this socket option simply passes
6079 * in to this call the sctp_sndrcvinfo structure defined in Section
6080 * 5.2.2) The input parameters accepted by this call include
6081 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
6082 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
6083 * to this call if the caller is using the UDP model.
6085 * For getsockopt, it get the default sctp_sndrcvinfo structure.
6087 static int sctp_getsockopt_default_send_param(struct sock *sk,
6088 int len, char __user *optval,
6091 struct sctp_sock *sp = sctp_sk(sk);
6092 struct sctp_association *asoc;
6093 struct sctp_sndrcvinfo info;
6095 if (len < sizeof(info))
6100 if (copy_from_user(&info, optval, len))
6103 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
6104 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
6107 info.sinfo_stream = asoc->default_stream;
6108 info.sinfo_flags = asoc->default_flags;
6109 info.sinfo_ppid = asoc->default_ppid;
6110 info.sinfo_context = asoc->default_context;
6111 info.sinfo_timetolive = asoc->default_timetolive;
6113 info.sinfo_stream = sp->default_stream;
6114 info.sinfo_flags = sp->default_flags;
6115 info.sinfo_ppid = sp->default_ppid;
6116 info.sinfo_context = sp->default_context;
6117 info.sinfo_timetolive = sp->default_timetolive;
6120 if (put_user(len, optlen))
6122 if (copy_to_user(optval, &info, len))
6128 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
6129 * (SCTP_DEFAULT_SNDINFO)
6131 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
6132 char __user *optval,
6135 struct sctp_sock *sp = sctp_sk(sk);
6136 struct sctp_association *asoc;
6137 struct sctp_sndinfo info;
6139 if (len < sizeof(info))
6144 if (copy_from_user(&info, optval, len))
6147 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
6148 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
6151 info.snd_sid = asoc->default_stream;
6152 info.snd_flags = asoc->default_flags;
6153 info.snd_ppid = asoc->default_ppid;
6154 info.snd_context = asoc->default_context;
6156 info.snd_sid = sp->default_stream;
6157 info.snd_flags = sp->default_flags;
6158 info.snd_ppid = sp->default_ppid;
6159 info.snd_context = sp->default_context;
6162 if (put_user(len, optlen))
6164 if (copy_to_user(optval, &info, len))
6172 * 7.1.5 SCTP_NODELAY
6174 * Turn on/off any Nagle-like algorithm. This means that packets are
6175 * generally sent as soon as possible and no unnecessary delays are
6176 * introduced, at the cost of more packets in the network. Expects an
6177 * integer boolean flag.
6180 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
6181 char __user *optval, int __user *optlen)
6185 if (len < sizeof(int))
6189 val = (sctp_sk(sk)->nodelay == 1);
6190 if (put_user(len, optlen))
6192 if (copy_to_user(optval, &val, len))
6199 * 7.1.1 SCTP_RTOINFO
6201 * The protocol parameters used to initialize and bound retransmission
6202 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
6203 * and modify these parameters.
6204 * All parameters are time values, in milliseconds. A value of 0, when
6205 * modifying the parameters, indicates that the current value should not
6209 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
6210 char __user *optval,
6211 int __user *optlen) {
6212 struct sctp_rtoinfo rtoinfo;
6213 struct sctp_association *asoc;
6215 if (len < sizeof (struct sctp_rtoinfo))
6218 len = sizeof(struct sctp_rtoinfo);
6220 if (copy_from_user(&rtoinfo, optval, len))
6223 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
6225 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
6228 /* Values corresponding to the specific association. */
6230 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
6231 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
6232 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
6234 /* Values corresponding to the endpoint. */
6235 struct sctp_sock *sp = sctp_sk(sk);
6237 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
6238 rtoinfo.srto_max = sp->rtoinfo.srto_max;
6239 rtoinfo.srto_min = sp->rtoinfo.srto_min;
6242 if (put_user(len, optlen))
6245 if (copy_to_user(optval, &rtoinfo, len))
6253 * 7.1.2 SCTP_ASSOCINFO
6255 * This option is used to tune the maximum retransmission attempts
6256 * of the association.
6257 * Returns an error if the new association retransmission value is
6258 * greater than the sum of the retransmission value of the peer.
6259 * See [SCTP] for more information.
6262 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
6263 char __user *optval,
6267 struct sctp_assocparams assocparams;
6268 struct sctp_association *asoc;
6269 struct list_head *pos;
6272 if (len < sizeof (struct sctp_assocparams))
6275 len = sizeof(struct sctp_assocparams);
6277 if (copy_from_user(&assocparams, optval, len))
6280 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
6282 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
6285 /* Values correspoinding to the specific association */
6287 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
6288 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
6289 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
6290 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
6292 list_for_each(pos, &asoc->peer.transport_addr_list) {
6296 assocparams.sasoc_number_peer_destinations = cnt;
6298 /* Values corresponding to the endpoint */
6299 struct sctp_sock *sp = sctp_sk(sk);
6301 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
6302 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
6303 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
6304 assocparams.sasoc_cookie_life =
6305 sp->assocparams.sasoc_cookie_life;
6306 assocparams.sasoc_number_peer_destinations =
6308 sasoc_number_peer_destinations;
6311 if (put_user(len, optlen))
6314 if (copy_to_user(optval, &assocparams, len))
6321 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
6323 * This socket option is a boolean flag which turns on or off mapped V4
6324 * addresses. If this option is turned on and the socket is type
6325 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
6326 * If this option is turned off, then no mapping will be done of V4
6327 * addresses and a user will receive both PF_INET6 and PF_INET type
6328 * addresses on the socket.
6330 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
6331 char __user *optval, int __user *optlen)
6334 struct sctp_sock *sp = sctp_sk(sk);
6336 if (len < sizeof(int))
6341 if (put_user(len, optlen))
6343 if (copy_to_user(optval, &val, len))
6350 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
6351 * (chapter and verse is quoted at sctp_setsockopt_context())
6353 static int sctp_getsockopt_context(struct sock *sk, int len,
6354 char __user *optval, int __user *optlen)
6356 struct sctp_assoc_value params;
6357 struct sctp_sock *sp;
6358 struct sctp_association *asoc;
6360 if (len < sizeof(struct sctp_assoc_value))
6363 len = sizeof(struct sctp_assoc_value);
6365 if (copy_from_user(¶ms, optval, len))
6370 if (params.assoc_id != 0) {
6371 asoc = sctp_id2assoc(sk, params.assoc_id);
6374 params.assoc_value = asoc->default_rcv_context;
6376 params.assoc_value = sp->default_rcv_context;
6379 if (put_user(len, optlen))
6381 if (copy_to_user(optval, ¶ms, len))
6388 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
6389 * This option will get or set the maximum size to put in any outgoing
6390 * SCTP DATA chunk. If a message is larger than this size it will be
6391 * fragmented by SCTP into the specified size. Note that the underlying
6392 * SCTP implementation may fragment into smaller sized chunks when the
6393 * PMTU of the underlying association is smaller than the value set by
6394 * the user. The default value for this option is '0' which indicates
6395 * the user is NOT limiting fragmentation and only the PMTU will effect
6396 * SCTP's choice of DATA chunk size. Note also that values set larger
6397 * than the maximum size of an IP datagram will effectively let SCTP
6398 * control fragmentation (i.e. the same as setting this option to 0).
6400 * The following structure is used to access and modify this parameter:
6402 * struct sctp_assoc_value {
6403 * sctp_assoc_t assoc_id;
6404 * uint32_t assoc_value;
6407 * assoc_id: This parameter is ignored for one-to-one style sockets.
6408 * For one-to-many style sockets this parameter indicates which
6409 * association the user is performing an action upon. Note that if
6410 * this field's value is zero then the endpoints default value is
6411 * changed (effecting future associations only).
6412 * assoc_value: This parameter specifies the maximum size in bytes.
6414 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
6415 char __user *optval, int __user *optlen)
6417 struct sctp_assoc_value params;
6418 struct sctp_association *asoc;
6420 if (len == sizeof(int)) {
6421 pr_warn_ratelimited(DEPRECATED
6423 "Use of int in maxseg socket option.\n"
6424 "Use struct sctp_assoc_value instead\n",
6425 current->comm, task_pid_nr(current));
6426 params.assoc_id = 0;
6427 } else if (len >= sizeof(struct sctp_assoc_value)) {
6428 len = sizeof(struct sctp_assoc_value);
6429 if (copy_from_user(¶ms, optval, len))
6434 asoc = sctp_id2assoc(sk, params.assoc_id);
6435 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
6439 params.assoc_value = asoc->frag_point;
6441 params.assoc_value = sctp_sk(sk)->user_frag;
6443 if (put_user(len, optlen))
6445 if (len == sizeof(int)) {
6446 if (copy_to_user(optval, ¶ms.assoc_value, len))
6449 if (copy_to_user(optval, ¶ms, len))
6457 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
6458 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
6460 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
6461 char __user *optval, int __user *optlen)
6465 if (len < sizeof(int))
6470 val = sctp_sk(sk)->frag_interleave;
6471 if (put_user(len, optlen))
6473 if (copy_to_user(optval, &val, len))
6480 * 7.1.25. Set or Get the sctp partial delivery point
6481 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
6483 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
6484 char __user *optval,
6489 if (len < sizeof(u32))
6494 val = sctp_sk(sk)->pd_point;
6495 if (put_user(len, optlen))
6497 if (copy_to_user(optval, &val, len))
6504 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
6505 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
6507 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
6508 char __user *optval,
6511 struct sctp_assoc_value params;
6512 struct sctp_sock *sp;
6513 struct sctp_association *asoc;
6515 if (len == sizeof(int)) {
6516 pr_warn_ratelimited(DEPRECATED
6518 "Use of int in max_burst socket option.\n"
6519 "Use struct sctp_assoc_value instead\n",
6520 current->comm, task_pid_nr(current));
6521 params.assoc_id = 0;
6522 } else if (len >= sizeof(struct sctp_assoc_value)) {
6523 len = sizeof(struct sctp_assoc_value);
6524 if (copy_from_user(¶ms, optval, len))
6531 if (params.assoc_id != 0) {
6532 asoc = sctp_id2assoc(sk, params.assoc_id);
6535 params.assoc_value = asoc->max_burst;
6537 params.assoc_value = sp->max_burst;
6539 if (len == sizeof(int)) {
6540 if (copy_to_user(optval, ¶ms.assoc_value, len))
6543 if (copy_to_user(optval, ¶ms, len))
6551 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
6552 char __user *optval, int __user *optlen)
6554 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6555 struct sctp_hmacalgo __user *p = (void __user *)optval;
6556 struct sctp_hmac_algo_param *hmacs;
6561 if (!ep->auth_enable)
6564 hmacs = ep->auth_hmacs_list;
6565 data_len = ntohs(hmacs->param_hdr.length) -
6566 sizeof(struct sctp_paramhdr);
6568 if (len < sizeof(struct sctp_hmacalgo) + data_len)
6571 len = sizeof(struct sctp_hmacalgo) + data_len;
6572 num_idents = data_len / sizeof(u16);
6574 if (put_user(len, optlen))
6576 if (put_user(num_idents, &p->shmac_num_idents))
6578 for (i = 0; i < num_idents; i++) {
6579 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
6581 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
6587 static int sctp_getsockopt_active_key(struct sock *sk, int len,
6588 char __user *optval, int __user *optlen)
6590 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6591 struct sctp_authkeyid val;
6592 struct sctp_association *asoc;
6594 if (!ep->auth_enable)
6597 if (len < sizeof(struct sctp_authkeyid))
6600 len = sizeof(struct sctp_authkeyid);
6601 if (copy_from_user(&val, optval, len))
6604 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
6605 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
6609 val.scact_keynumber = asoc->active_key_id;
6611 val.scact_keynumber = ep->active_key_id;
6613 if (put_user(len, optlen))
6615 if (copy_to_user(optval, &val, len))
6621 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
6622 char __user *optval, int __user *optlen)
6624 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6625 struct sctp_authchunks __user *p = (void __user *)optval;
6626 struct sctp_authchunks val;
6627 struct sctp_association *asoc;
6628 struct sctp_chunks_param *ch;
6632 if (!ep->auth_enable)
6635 if (len < sizeof(struct sctp_authchunks))
6638 if (copy_from_user(&val, optval, sizeof(val)))
6641 to = p->gauth_chunks;
6642 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6646 ch = asoc->peer.peer_chunks;
6650 /* See if the user provided enough room for all the data */
6651 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6652 if (len < num_chunks)
6655 if (copy_to_user(to, ch->chunks, num_chunks))
6658 len = sizeof(struct sctp_authchunks) + num_chunks;
6659 if (put_user(len, optlen))
6661 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6666 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
6667 char __user *optval, int __user *optlen)
6669 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6670 struct sctp_authchunks __user *p = (void __user *)optval;
6671 struct sctp_authchunks val;
6672 struct sctp_association *asoc;
6673 struct sctp_chunks_param *ch;
6677 if (!ep->auth_enable)
6680 if (len < sizeof(struct sctp_authchunks))
6683 if (copy_from_user(&val, optval, sizeof(val)))
6686 to = p->gauth_chunks;
6687 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6688 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
6692 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
6694 ch = ep->auth_chunk_list;
6699 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6700 if (len < sizeof(struct sctp_authchunks) + num_chunks)
6703 if (copy_to_user(to, ch->chunks, num_chunks))
6706 len = sizeof(struct sctp_authchunks) + num_chunks;
6707 if (put_user(len, optlen))
6709 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6716 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
6717 * This option gets the current number of associations that are attached
6718 * to a one-to-many style socket. The option value is an uint32_t.
6720 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
6721 char __user *optval, int __user *optlen)
6723 struct sctp_sock *sp = sctp_sk(sk);
6724 struct sctp_association *asoc;
6727 if (sctp_style(sk, TCP))
6730 if (len < sizeof(u32))
6735 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6739 if (put_user(len, optlen))
6741 if (copy_to_user(optval, &val, len))
6748 * 8.1.23 SCTP_AUTO_ASCONF
6749 * See the corresponding setsockopt entry as description
6751 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
6752 char __user *optval, int __user *optlen)
6756 if (len < sizeof(int))
6760 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
6762 if (put_user(len, optlen))
6764 if (copy_to_user(optval, &val, len))
6770 * 8.2.6. Get the Current Identifiers of Associations
6771 * (SCTP_GET_ASSOC_ID_LIST)
6773 * This option gets the current list of SCTP association identifiers of
6774 * the SCTP associations handled by a one-to-many style socket.
6776 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
6777 char __user *optval, int __user *optlen)
6779 struct sctp_sock *sp = sctp_sk(sk);
6780 struct sctp_association *asoc;
6781 struct sctp_assoc_ids *ids;
6784 if (sctp_style(sk, TCP))
6787 if (len < sizeof(struct sctp_assoc_ids))
6790 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6794 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
6797 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
6799 ids = kmalloc(len, GFP_USER | __GFP_NOWARN);
6803 ids->gaids_number_of_ids = num;
6805 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6806 ids->gaids_assoc_id[num++] = asoc->assoc_id;
6809 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
6819 * SCTP_PEER_ADDR_THLDS
6821 * This option allows us to fetch the partially failed threshold for one or all
6822 * transports in an association. See Section 6.1 of:
6823 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6825 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
6826 char __user *optval,
6830 struct sctp_paddrthlds val;
6831 struct sctp_transport *trans;
6832 struct sctp_association *asoc;
6834 if (len < sizeof(struct sctp_paddrthlds))
6836 len = sizeof(struct sctp_paddrthlds);
6837 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
6840 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
6841 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
6845 val.spt_pathpfthld = asoc->pf_retrans;
6846 val.spt_pathmaxrxt = asoc->pathmaxrxt;
6848 trans = sctp_addr_id2transport(sk, &val.spt_address,
6853 val.spt_pathmaxrxt = trans->pathmaxrxt;
6854 val.spt_pathpfthld = trans->pf_retrans;
6857 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
6864 * SCTP_GET_ASSOC_STATS
6866 * This option retrieves local per endpoint statistics. It is modeled
6867 * after OpenSolaris' implementation
6869 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
6870 char __user *optval,
6873 struct sctp_assoc_stats sas;
6874 struct sctp_association *asoc = NULL;
6876 /* User must provide at least the assoc id */
6877 if (len < sizeof(sctp_assoc_t))
6880 /* Allow the struct to grow and fill in as much as possible */
6881 len = min_t(size_t, len, sizeof(sas));
6883 if (copy_from_user(&sas, optval, len))
6886 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
6890 sas.sas_rtxchunks = asoc->stats.rtxchunks;
6891 sas.sas_gapcnt = asoc->stats.gapcnt;
6892 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
6893 sas.sas_osacks = asoc->stats.osacks;
6894 sas.sas_isacks = asoc->stats.isacks;
6895 sas.sas_octrlchunks = asoc->stats.octrlchunks;
6896 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
6897 sas.sas_oodchunks = asoc->stats.oodchunks;
6898 sas.sas_iodchunks = asoc->stats.iodchunks;
6899 sas.sas_ouodchunks = asoc->stats.ouodchunks;
6900 sas.sas_iuodchunks = asoc->stats.iuodchunks;
6901 sas.sas_idupchunks = asoc->stats.idupchunks;
6902 sas.sas_opackets = asoc->stats.opackets;
6903 sas.sas_ipackets = asoc->stats.ipackets;
6905 /* New high max rto observed, will return 0 if not a single
6906 * RTO update took place. obs_rto_ipaddr will be bogus
6909 sas.sas_maxrto = asoc->stats.max_obs_rto;
6910 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
6911 sizeof(struct sockaddr_storage));
6913 /* Mark beginning of a new observation period */
6914 asoc->stats.max_obs_rto = asoc->rto_min;
6916 if (put_user(len, optlen))
6919 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
6921 if (copy_to_user(optval, &sas, len))
6927 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
6928 char __user *optval,
6933 if (len < sizeof(int))
6937 if (sctp_sk(sk)->recvrcvinfo)
6939 if (put_user(len, optlen))
6941 if (copy_to_user(optval, &val, len))
6947 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
6948 char __user *optval,
6953 if (len < sizeof(int))
6957 if (sctp_sk(sk)->recvnxtinfo)
6959 if (put_user(len, optlen))
6961 if (copy_to_user(optval, &val, len))
6967 static int sctp_getsockopt_pr_supported(struct sock *sk, int len,
6968 char __user *optval,
6971 struct sctp_assoc_value params;
6972 struct sctp_association *asoc;
6973 int retval = -EFAULT;
6975 if (len < sizeof(params)) {
6980 len = sizeof(params);
6981 if (copy_from_user(¶ms, optval, len))
6984 asoc = sctp_id2assoc(sk, params.assoc_id);
6986 params.assoc_value = asoc->prsctp_enable;
6987 } else if (!params.assoc_id) {
6988 struct sctp_sock *sp = sctp_sk(sk);
6990 params.assoc_value = sp->ep->prsctp_enable;
6996 if (put_user(len, optlen))
6999 if (copy_to_user(optval, ¶ms, len))
7008 static int sctp_getsockopt_default_prinfo(struct sock *sk, int len,
7009 char __user *optval,
7012 struct sctp_default_prinfo info;
7013 struct sctp_association *asoc;
7014 int retval = -EFAULT;
7016 if (len < sizeof(info)) {
7022 if (copy_from_user(&info, optval, len))
7025 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
7027 info.pr_policy = SCTP_PR_POLICY(asoc->default_flags);
7028 info.pr_value = asoc->default_timetolive;
7029 } else if (!info.pr_assoc_id) {
7030 struct sctp_sock *sp = sctp_sk(sk);
7032 info.pr_policy = SCTP_PR_POLICY(sp->default_flags);
7033 info.pr_value = sp->default_timetolive;
7039 if (put_user(len, optlen))
7042 if (copy_to_user(optval, &info, len))
7051 static int sctp_getsockopt_pr_assocstatus(struct sock *sk, int len,
7052 char __user *optval,
7055 struct sctp_prstatus params;
7056 struct sctp_association *asoc;
7058 int retval = -EINVAL;
7060 if (len < sizeof(params))
7063 len = sizeof(params);
7064 if (copy_from_user(¶ms, optval, len)) {
7069 policy = params.sprstat_policy;
7070 if (!policy || (policy & ~(SCTP_PR_SCTP_MASK | SCTP_PR_SCTP_ALL)) ||
7071 ((policy & SCTP_PR_SCTP_ALL) && (policy & SCTP_PR_SCTP_MASK)))
7074 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
7078 if (policy == SCTP_PR_SCTP_ALL) {
7079 params.sprstat_abandoned_unsent = 0;
7080 params.sprstat_abandoned_sent = 0;
7081 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
7082 params.sprstat_abandoned_unsent +=
7083 asoc->abandoned_unsent[policy];
7084 params.sprstat_abandoned_sent +=
7085 asoc->abandoned_sent[policy];
7088 params.sprstat_abandoned_unsent =
7089 asoc->abandoned_unsent[__SCTP_PR_INDEX(policy)];
7090 params.sprstat_abandoned_sent =
7091 asoc->abandoned_sent[__SCTP_PR_INDEX(policy)];
7094 if (put_user(len, optlen)) {
7099 if (copy_to_user(optval, ¶ms, len)) {
7110 static int sctp_getsockopt_pr_streamstatus(struct sock *sk, int len,
7111 char __user *optval,
7114 struct sctp_stream_out_ext *streamoute;
7115 struct sctp_association *asoc;
7116 struct sctp_prstatus params;
7117 int retval = -EINVAL;
7120 if (len < sizeof(params))
7123 len = sizeof(params);
7124 if (copy_from_user(¶ms, optval, len)) {
7129 policy = params.sprstat_policy;
7130 if (!policy || (policy & ~(SCTP_PR_SCTP_MASK | SCTP_PR_SCTP_ALL)) ||
7131 ((policy & SCTP_PR_SCTP_ALL) && (policy & SCTP_PR_SCTP_MASK)))
7134 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
7135 if (!asoc || params.sprstat_sid >= asoc->stream.outcnt)
7138 streamoute = SCTP_SO(&asoc->stream, params.sprstat_sid)->ext;
7140 /* Not allocated yet, means all stats are 0 */
7141 params.sprstat_abandoned_unsent = 0;
7142 params.sprstat_abandoned_sent = 0;
7147 if (policy == SCTP_PR_SCTP_ALL) {
7148 params.sprstat_abandoned_unsent = 0;
7149 params.sprstat_abandoned_sent = 0;
7150 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
7151 params.sprstat_abandoned_unsent +=
7152 streamoute->abandoned_unsent[policy];
7153 params.sprstat_abandoned_sent +=
7154 streamoute->abandoned_sent[policy];
7157 params.sprstat_abandoned_unsent =
7158 streamoute->abandoned_unsent[__SCTP_PR_INDEX(policy)];
7159 params.sprstat_abandoned_sent =
7160 streamoute->abandoned_sent[__SCTP_PR_INDEX(policy)];
7163 if (put_user(len, optlen) || copy_to_user(optval, ¶ms, len)) {
7174 static int sctp_getsockopt_reconfig_supported(struct sock *sk, int len,
7175 char __user *optval,
7178 struct sctp_assoc_value params;
7179 struct sctp_association *asoc;
7180 int retval = -EFAULT;
7182 if (len < sizeof(params)) {
7187 len = sizeof(params);
7188 if (copy_from_user(¶ms, optval, len))
7191 asoc = sctp_id2assoc(sk, params.assoc_id);
7193 params.assoc_value = asoc->reconf_enable;
7194 } else if (!params.assoc_id) {
7195 struct sctp_sock *sp = sctp_sk(sk);
7197 params.assoc_value = sp->ep->reconf_enable;
7203 if (put_user(len, optlen))
7206 if (copy_to_user(optval, ¶ms, len))
7215 static int sctp_getsockopt_enable_strreset(struct sock *sk, int len,
7216 char __user *optval,
7219 struct sctp_assoc_value params;
7220 struct sctp_association *asoc;
7221 int retval = -EFAULT;
7223 if (len < sizeof(params)) {
7228 len = sizeof(params);
7229 if (copy_from_user(¶ms, optval, len))
7232 asoc = sctp_id2assoc(sk, params.assoc_id);
7234 params.assoc_value = asoc->strreset_enable;
7235 } else if (!params.assoc_id) {
7236 struct sctp_sock *sp = sctp_sk(sk);
7238 params.assoc_value = sp->ep->strreset_enable;
7244 if (put_user(len, optlen))
7247 if (copy_to_user(optval, ¶ms, len))
7256 static int sctp_getsockopt_scheduler(struct sock *sk, int len,
7257 char __user *optval,
7260 struct sctp_assoc_value params;
7261 struct sctp_association *asoc;
7262 int retval = -EFAULT;
7264 if (len < sizeof(params)) {
7269 len = sizeof(params);
7270 if (copy_from_user(¶ms, optval, len))
7273 asoc = sctp_id2assoc(sk, params.assoc_id);
7279 params.assoc_value = sctp_sched_get_sched(asoc);
7281 if (put_user(len, optlen))
7284 if (copy_to_user(optval, ¶ms, len))
7293 static int sctp_getsockopt_scheduler_value(struct sock *sk, int len,
7294 char __user *optval,
7297 struct sctp_stream_value params;
7298 struct sctp_association *asoc;
7299 int retval = -EFAULT;
7301 if (len < sizeof(params)) {
7306 len = sizeof(params);
7307 if (copy_from_user(¶ms, optval, len))
7310 asoc = sctp_id2assoc(sk, params.assoc_id);
7316 retval = sctp_sched_get_value(asoc, params.stream_id,
7317 ¶ms.stream_value);
7321 if (put_user(len, optlen)) {
7326 if (copy_to_user(optval, ¶ms, len)) {
7335 static int sctp_getsockopt_interleaving_supported(struct sock *sk, int len,
7336 char __user *optval,
7339 struct sctp_assoc_value params;
7340 struct sctp_association *asoc;
7341 int retval = -EFAULT;
7343 if (len < sizeof(params)) {
7348 len = sizeof(params);
7349 if (copy_from_user(¶ms, optval, len))
7352 asoc = sctp_id2assoc(sk, params.assoc_id);
7354 params.assoc_value = asoc->intl_enable;
7355 } else if (!params.assoc_id) {
7356 struct sctp_sock *sp = sctp_sk(sk);
7358 params.assoc_value = sp->strm_interleave;
7364 if (put_user(len, optlen))
7367 if (copy_to_user(optval, ¶ms, len))
7376 static int sctp_getsockopt_reuse_port(struct sock *sk, int len,
7377 char __user *optval,
7382 if (len < sizeof(int))
7386 val = sctp_sk(sk)->reuse;
7387 if (put_user(len, optlen))
7390 if (copy_to_user(optval, &val, len))
7396 static int sctp_getsockopt(struct sock *sk, int level, int optname,
7397 char __user *optval, int __user *optlen)
7402 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
7404 /* I can hardly begin to describe how wrong this is. This is
7405 * so broken as to be worse than useless. The API draft
7406 * REALLY is NOT helpful here... I am not convinced that the
7407 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
7408 * are at all well-founded.
7410 if (level != SOL_SCTP) {
7411 struct sctp_af *af = sctp_sk(sk)->pf->af;
7413 retval = af->getsockopt(sk, level, optname, optval, optlen);
7417 if (get_user(len, optlen))
7427 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
7429 case SCTP_DISABLE_FRAGMENTS:
7430 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
7434 retval = sctp_getsockopt_events(sk, len, optval, optlen);
7436 case SCTP_AUTOCLOSE:
7437 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
7439 case SCTP_SOCKOPT_PEELOFF:
7440 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
7442 case SCTP_SOCKOPT_PEELOFF_FLAGS:
7443 retval = sctp_getsockopt_peeloff_flags(sk, len, optval, optlen);
7445 case SCTP_PEER_ADDR_PARAMS:
7446 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
7449 case SCTP_DELAYED_SACK:
7450 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
7454 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
7456 case SCTP_GET_PEER_ADDRS:
7457 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
7460 case SCTP_GET_LOCAL_ADDRS:
7461 retval = sctp_getsockopt_local_addrs(sk, len, optval,
7464 case SCTP_SOCKOPT_CONNECTX3:
7465 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
7467 case SCTP_DEFAULT_SEND_PARAM:
7468 retval = sctp_getsockopt_default_send_param(sk, len,
7471 case SCTP_DEFAULT_SNDINFO:
7472 retval = sctp_getsockopt_default_sndinfo(sk, len,
7475 case SCTP_PRIMARY_ADDR:
7476 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
7479 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
7482 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
7484 case SCTP_ASSOCINFO:
7485 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
7487 case SCTP_I_WANT_MAPPED_V4_ADDR:
7488 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
7491 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
7493 case SCTP_GET_PEER_ADDR_INFO:
7494 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
7497 case SCTP_ADAPTATION_LAYER:
7498 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
7502 retval = sctp_getsockopt_context(sk, len, optval, optlen);
7504 case SCTP_FRAGMENT_INTERLEAVE:
7505 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
7508 case SCTP_PARTIAL_DELIVERY_POINT:
7509 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
7512 case SCTP_MAX_BURST:
7513 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
7516 case SCTP_AUTH_CHUNK:
7517 case SCTP_AUTH_DELETE_KEY:
7518 case SCTP_AUTH_DEACTIVATE_KEY:
7519 retval = -EOPNOTSUPP;
7521 case SCTP_HMAC_IDENT:
7522 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
7524 case SCTP_AUTH_ACTIVE_KEY:
7525 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
7527 case SCTP_PEER_AUTH_CHUNKS:
7528 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
7531 case SCTP_LOCAL_AUTH_CHUNKS:
7532 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
7535 case SCTP_GET_ASSOC_NUMBER:
7536 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
7538 case SCTP_GET_ASSOC_ID_LIST:
7539 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
7541 case SCTP_AUTO_ASCONF:
7542 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
7544 case SCTP_PEER_ADDR_THLDS:
7545 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
7547 case SCTP_GET_ASSOC_STATS:
7548 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
7550 case SCTP_RECVRCVINFO:
7551 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
7553 case SCTP_RECVNXTINFO:
7554 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
7556 case SCTP_PR_SUPPORTED:
7557 retval = sctp_getsockopt_pr_supported(sk, len, optval, optlen);
7559 case SCTP_DEFAULT_PRINFO:
7560 retval = sctp_getsockopt_default_prinfo(sk, len, optval,
7563 case SCTP_PR_ASSOC_STATUS:
7564 retval = sctp_getsockopt_pr_assocstatus(sk, len, optval,
7567 case SCTP_PR_STREAM_STATUS:
7568 retval = sctp_getsockopt_pr_streamstatus(sk, len, optval,
7571 case SCTP_RECONFIG_SUPPORTED:
7572 retval = sctp_getsockopt_reconfig_supported(sk, len, optval,
7575 case SCTP_ENABLE_STREAM_RESET:
7576 retval = sctp_getsockopt_enable_strreset(sk, len, optval,
7579 case SCTP_STREAM_SCHEDULER:
7580 retval = sctp_getsockopt_scheduler(sk, len, optval,
7583 case SCTP_STREAM_SCHEDULER_VALUE:
7584 retval = sctp_getsockopt_scheduler_value(sk, len, optval,
7587 case SCTP_INTERLEAVING_SUPPORTED:
7588 retval = sctp_getsockopt_interleaving_supported(sk, len, optval,
7591 case SCTP_REUSE_PORT:
7592 retval = sctp_getsockopt_reuse_port(sk, len, optval, optlen);
7595 retval = -ENOPROTOOPT;
7603 static int sctp_hash(struct sock *sk)
7609 static void sctp_unhash(struct sock *sk)
7614 /* Check if port is acceptable. Possibly find first available port.
7616 * The port hash table (contained in the 'global' SCTP protocol storage
7617 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
7618 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
7619 * list (the list number is the port number hashed out, so as you
7620 * would expect from a hash function, all the ports in a given list have
7621 * such a number that hashes out to the same list number; you were
7622 * expecting that, right?); so each list has a set of ports, with a
7623 * link to the socket (struct sock) that uses it, the port number and
7624 * a fastreuse flag (FIXME: NPI ipg).
7626 static struct sctp_bind_bucket *sctp_bucket_create(
7627 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
7629 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
7631 bool reuse = (sk->sk_reuse || sctp_sk(sk)->reuse);
7632 struct sctp_bind_hashbucket *head; /* hash list */
7633 struct sctp_bind_bucket *pp;
7634 unsigned short snum;
7637 snum = ntohs(addr->v4.sin_port);
7639 pr_debug("%s: begins, snum:%d\n", __func__, snum);
7644 /* Search for an available port. */
7645 int low, high, remaining, index;
7647 struct net *net = sock_net(sk);
7649 inet_get_local_port_range(net, &low, &high);
7650 remaining = (high - low) + 1;
7651 rover = prandom_u32() % remaining + low;
7655 if ((rover < low) || (rover > high))
7657 if (inet_is_local_reserved_port(net, rover))
7659 index = sctp_phashfn(sock_net(sk), rover);
7660 head = &sctp_port_hashtable[index];
7661 spin_lock(&head->lock);
7662 sctp_for_each_hentry(pp, &head->chain)
7663 if ((pp->port == rover) &&
7664 net_eq(sock_net(sk), pp->net))
7668 spin_unlock(&head->lock);
7669 } while (--remaining > 0);
7671 /* Exhausted local port range during search? */
7676 /* OK, here is the one we will use. HEAD (the port
7677 * hash table list entry) is non-NULL and we hold it's
7682 /* We are given an specific port number; we verify
7683 * that it is not being used. If it is used, we will
7684 * exahust the search in the hash list corresponding
7685 * to the port number (snum) - we detect that with the
7686 * port iterator, pp being NULL.
7688 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
7689 spin_lock(&head->lock);
7690 sctp_for_each_hentry(pp, &head->chain) {
7691 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
7698 if (!hlist_empty(&pp->owner)) {
7699 /* We had a port hash table hit - there is an
7700 * available port (pp != NULL) and it is being
7701 * used by other socket (pp->owner not empty); that other
7702 * socket is going to be sk2.
7706 pr_debug("%s: found a possible match\n", __func__);
7708 if (pp->fastreuse && reuse && sk->sk_state != SCTP_SS_LISTENING)
7711 /* Run through the list of sockets bound to the port
7712 * (pp->port) [via the pointers bind_next and
7713 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
7714 * we get the endpoint they describe and run through
7715 * the endpoint's list of IP (v4 or v6) addresses,
7716 * comparing each of the addresses with the address of
7717 * the socket sk. If we find a match, then that means
7718 * that this port/socket (sk) combination are already
7721 sk_for_each_bound(sk2, &pp->owner) {
7722 struct sctp_endpoint *ep2;
7723 ep2 = sctp_sk(sk2)->ep;
7726 (reuse && (sk2->sk_reuse || sctp_sk(sk2)->reuse) &&
7727 sk2->sk_state != SCTP_SS_LISTENING))
7730 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
7731 sctp_sk(sk2), sctp_sk(sk))) {
7737 pr_debug("%s: found a match\n", __func__);
7740 /* If there was a hash table miss, create a new port. */
7742 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
7745 /* In either case (hit or miss), make sure fastreuse is 1 only
7746 * if sk->sk_reuse is too (that is, if the caller requested
7747 * SO_REUSEADDR on this socket -sk-).
7749 if (hlist_empty(&pp->owner)) {
7750 if (reuse && sk->sk_state != SCTP_SS_LISTENING)
7754 } else if (pp->fastreuse &&
7755 (!reuse || sk->sk_state == SCTP_SS_LISTENING))
7758 /* We are set, so fill up all the data in the hash table
7759 * entry, tie the socket list information with the rest of the
7760 * sockets FIXME: Blurry, NPI (ipg).
7763 if (!sctp_sk(sk)->bind_hash) {
7764 inet_sk(sk)->inet_num = snum;
7765 sk_add_bind_node(sk, &pp->owner);
7766 sctp_sk(sk)->bind_hash = pp;
7771 spin_unlock(&head->lock);
7778 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
7779 * port is requested.
7781 static int sctp_get_port(struct sock *sk, unsigned short snum)
7783 union sctp_addr addr;
7784 struct sctp_af *af = sctp_sk(sk)->pf->af;
7786 /* Set up a dummy address struct from the sk. */
7787 af->from_sk(&addr, sk);
7788 addr.v4.sin_port = htons(snum);
7790 /* Note: sk->sk_num gets filled in if ephemeral port request. */
7791 return !!sctp_get_port_local(sk, &addr);
7795 * Move a socket to LISTENING state.
7797 static int sctp_listen_start(struct sock *sk, int backlog)
7799 struct sctp_sock *sp = sctp_sk(sk);
7800 struct sctp_endpoint *ep = sp->ep;
7801 struct crypto_shash *tfm = NULL;
7804 /* Allocate HMAC for generating cookie. */
7805 if (!sp->hmac && sp->sctp_hmac_alg) {
7806 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
7807 tfm = crypto_alloc_shash(alg, 0, 0);
7809 net_info_ratelimited("failed to load transform for %s: %ld\n",
7810 sp->sctp_hmac_alg, PTR_ERR(tfm));
7813 sctp_sk(sk)->hmac = tfm;
7817 * If a bind() or sctp_bindx() is not called prior to a listen()
7818 * call that allows new associations to be accepted, the system
7819 * picks an ephemeral port and will choose an address set equivalent
7820 * to binding with a wildcard address.
7822 * This is not currently spelled out in the SCTP sockets
7823 * extensions draft, but follows the practice as seen in TCP
7827 inet_sk_set_state(sk, SCTP_SS_LISTENING);
7828 if (!ep->base.bind_addr.port) {
7829 if (sctp_autobind(sk))
7832 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
7833 inet_sk_set_state(sk, SCTP_SS_CLOSED);
7838 sk->sk_max_ack_backlog = backlog;
7839 sctp_hash_endpoint(ep);
7844 * 4.1.3 / 5.1.3 listen()
7846 * By default, new associations are not accepted for UDP style sockets.
7847 * An application uses listen() to mark a socket as being able to
7848 * accept new associations.
7850 * On TCP style sockets, applications use listen() to ready the SCTP
7851 * endpoint for accepting inbound associations.
7853 * On both types of endpoints a backlog of '0' disables listening.
7855 * Move a socket to LISTENING state.
7857 int sctp_inet_listen(struct socket *sock, int backlog)
7859 struct sock *sk = sock->sk;
7860 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
7863 if (unlikely(backlog < 0))
7868 /* Peeled-off sockets are not allowed to listen(). */
7869 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
7872 if (sock->state != SS_UNCONNECTED)
7875 if (!sctp_sstate(sk, LISTENING) && !sctp_sstate(sk, CLOSED))
7878 /* If backlog is zero, disable listening. */
7880 if (sctp_sstate(sk, CLOSED))
7884 sctp_unhash_endpoint(ep);
7885 sk->sk_state = SCTP_SS_CLOSED;
7886 if (sk->sk_reuse || sctp_sk(sk)->reuse)
7887 sctp_sk(sk)->bind_hash->fastreuse = 1;
7891 /* If we are already listening, just update the backlog */
7892 if (sctp_sstate(sk, LISTENING))
7893 sk->sk_max_ack_backlog = backlog;
7895 err = sctp_listen_start(sk, backlog);
7907 * This function is done by modeling the current datagram_poll() and the
7908 * tcp_poll(). Note that, based on these implementations, we don't
7909 * lock the socket in this function, even though it seems that,
7910 * ideally, locking or some other mechanisms can be used to ensure
7911 * the integrity of the counters (sndbuf and wmem_alloc) used
7912 * in this place. We assume that we don't need locks either until proven
7915 * Another thing to note is that we include the Async I/O support
7916 * here, again, by modeling the current TCP/UDP code. We don't have
7917 * a good way to test with it yet.
7919 __poll_t sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
7921 struct sock *sk = sock->sk;
7922 struct sctp_sock *sp = sctp_sk(sk);
7925 poll_wait(file, sk_sleep(sk), wait);
7927 sock_rps_record_flow(sk);
7929 /* A TCP-style listening socket becomes readable when the accept queue
7932 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
7933 return (!list_empty(&sp->ep->asocs)) ?
7934 (EPOLLIN | EPOLLRDNORM) : 0;
7938 /* Is there any exceptional events? */
7939 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
7941 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? EPOLLPRI : 0);
7942 if (sk->sk_shutdown & RCV_SHUTDOWN)
7943 mask |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
7944 if (sk->sk_shutdown == SHUTDOWN_MASK)
7947 /* Is it readable? Reconsider this code with TCP-style support. */
7948 if (!skb_queue_empty(&sk->sk_receive_queue))
7949 mask |= EPOLLIN | EPOLLRDNORM;
7951 /* The association is either gone or not ready. */
7952 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
7955 /* Is it writable? */
7956 if (sctp_writeable(sk)) {
7957 mask |= EPOLLOUT | EPOLLWRNORM;
7959 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
7961 * Since the socket is not locked, the buffer
7962 * might be made available after the writeable check and
7963 * before the bit is set. This could cause a lost I/O
7964 * signal. tcp_poll() has a race breaker for this race
7965 * condition. Based on their implementation, we put
7966 * in the following code to cover it as well.
7968 if (sctp_writeable(sk))
7969 mask |= EPOLLOUT | EPOLLWRNORM;
7974 /********************************************************************
7975 * 2nd Level Abstractions
7976 ********************************************************************/
7978 static struct sctp_bind_bucket *sctp_bucket_create(
7979 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
7981 struct sctp_bind_bucket *pp;
7983 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
7985 SCTP_DBG_OBJCNT_INC(bind_bucket);
7988 INIT_HLIST_HEAD(&pp->owner);
7990 hlist_add_head(&pp->node, &head->chain);
7995 /* Caller must hold hashbucket lock for this tb with local BH disabled */
7996 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
7998 if (pp && hlist_empty(&pp->owner)) {
7999 __hlist_del(&pp->node);
8000 kmem_cache_free(sctp_bucket_cachep, pp);
8001 SCTP_DBG_OBJCNT_DEC(bind_bucket);
8005 /* Release this socket's reference to a local port. */
8006 static inline void __sctp_put_port(struct sock *sk)
8008 struct sctp_bind_hashbucket *head =
8009 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
8010 inet_sk(sk)->inet_num)];
8011 struct sctp_bind_bucket *pp;
8013 spin_lock(&head->lock);
8014 pp = sctp_sk(sk)->bind_hash;
8015 __sk_del_bind_node(sk);
8016 sctp_sk(sk)->bind_hash = NULL;
8017 inet_sk(sk)->inet_num = 0;
8018 sctp_bucket_destroy(pp);
8019 spin_unlock(&head->lock);
8022 void sctp_put_port(struct sock *sk)
8025 __sctp_put_port(sk);
8030 * The system picks an ephemeral port and choose an address set equivalent
8031 * to binding with a wildcard address.
8032 * One of those addresses will be the primary address for the association.
8033 * This automatically enables the multihoming capability of SCTP.
8035 static int sctp_autobind(struct sock *sk)
8037 union sctp_addr autoaddr;
8041 /* Initialize a local sockaddr structure to INADDR_ANY. */
8042 af = sctp_sk(sk)->pf->af;
8044 port = htons(inet_sk(sk)->inet_num);
8045 af->inaddr_any(&autoaddr, port);
8047 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
8050 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
8053 * 4.2 The cmsghdr Structure *
8055 * When ancillary data is sent or received, any number of ancillary data
8056 * objects can be specified by the msg_control and msg_controllen members of
8057 * the msghdr structure, because each object is preceded by
8058 * a cmsghdr structure defining the object's length (the cmsg_len member).
8059 * Historically Berkeley-derived implementations have passed only one object
8060 * at a time, but this API allows multiple objects to be
8061 * passed in a single call to sendmsg() or recvmsg(). The following example
8062 * shows two ancillary data objects in a control buffer.
8064 * |<--------------------------- msg_controllen -------------------------->|
8067 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
8069 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
8072 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
8074 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
8077 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
8078 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
8080 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
8082 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
8089 static int sctp_msghdr_parse(const struct msghdr *msg, struct sctp_cmsgs *cmsgs)
8091 struct msghdr *my_msg = (struct msghdr *)msg;
8092 struct cmsghdr *cmsg;
8094 for_each_cmsghdr(cmsg, my_msg) {
8095 if (!CMSG_OK(my_msg, cmsg))
8098 /* Should we parse this header or ignore? */
8099 if (cmsg->cmsg_level != IPPROTO_SCTP)
8102 /* Strictly check lengths following example in SCM code. */
8103 switch (cmsg->cmsg_type) {
8105 /* SCTP Socket API Extension
8106 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
8108 * This cmsghdr structure provides information for
8109 * initializing new SCTP associations with sendmsg().
8110 * The SCTP_INITMSG socket option uses this same data
8111 * structure. This structure is not used for
8114 * cmsg_level cmsg_type cmsg_data[]
8115 * ------------ ------------ ----------------------
8116 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
8118 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
8121 cmsgs->init = CMSG_DATA(cmsg);
8125 /* SCTP Socket API Extension
8126 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
8128 * This cmsghdr structure specifies SCTP options for
8129 * sendmsg() and describes SCTP header information
8130 * about a received message through recvmsg().
8132 * cmsg_level cmsg_type cmsg_data[]
8133 * ------------ ------------ ----------------------
8134 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
8136 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
8139 cmsgs->srinfo = CMSG_DATA(cmsg);
8141 if (cmsgs->srinfo->sinfo_flags &
8142 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
8143 SCTP_SACK_IMMEDIATELY | SCTP_SENDALL |
8144 SCTP_PR_SCTP_MASK | SCTP_ABORT | SCTP_EOF))
8149 /* SCTP Socket API Extension
8150 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
8152 * This cmsghdr structure specifies SCTP options for
8153 * sendmsg(). This structure and SCTP_RCVINFO replaces
8154 * SCTP_SNDRCV which has been deprecated.
8156 * cmsg_level cmsg_type cmsg_data[]
8157 * ------------ ------------ ---------------------
8158 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
8160 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
8163 cmsgs->sinfo = CMSG_DATA(cmsg);
8165 if (cmsgs->sinfo->snd_flags &
8166 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
8167 SCTP_SACK_IMMEDIATELY | SCTP_SENDALL |
8168 SCTP_PR_SCTP_MASK | SCTP_ABORT | SCTP_EOF))
8172 /* SCTP Socket API Extension
8173 * 5.3.7 SCTP PR-SCTP Information Structure (SCTP_PRINFO)
8175 * This cmsghdr structure specifies SCTP options for sendmsg().
8177 * cmsg_level cmsg_type cmsg_data[]
8178 * ------------ ------------ ---------------------
8179 * IPPROTO_SCTP SCTP_PRINFO struct sctp_prinfo
8181 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_prinfo)))
8184 cmsgs->prinfo = CMSG_DATA(cmsg);
8185 if (cmsgs->prinfo->pr_policy & ~SCTP_PR_SCTP_MASK)
8188 if (cmsgs->prinfo->pr_policy == SCTP_PR_SCTP_NONE)
8189 cmsgs->prinfo->pr_value = 0;
8192 /* SCTP Socket API Extension
8193 * 5.3.8 SCTP AUTH Information Structure (SCTP_AUTHINFO)
8195 * This cmsghdr structure specifies SCTP options for sendmsg().
8197 * cmsg_level cmsg_type cmsg_data[]
8198 * ------------ ------------ ---------------------
8199 * IPPROTO_SCTP SCTP_AUTHINFO struct sctp_authinfo
8201 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_authinfo)))
8204 cmsgs->authinfo = CMSG_DATA(cmsg);
8206 case SCTP_DSTADDRV4:
8207 case SCTP_DSTADDRV6:
8208 /* SCTP Socket API Extension
8209 * 5.3.9/10 SCTP Destination IPv4/6 Address Structure (SCTP_DSTADDRV4/6)
8211 * This cmsghdr structure specifies SCTP options for sendmsg().
8213 * cmsg_level cmsg_type cmsg_data[]
8214 * ------------ ------------ ---------------------
8215 * IPPROTO_SCTP SCTP_DSTADDRV4 struct in_addr
8216 * ------------ ------------ ---------------------
8217 * IPPROTO_SCTP SCTP_DSTADDRV6 struct in6_addr
8219 cmsgs->addrs_msg = my_msg;
8230 * Wait for a packet..
8231 * Note: This function is the same function as in core/datagram.c
8232 * with a few modifications to make lksctp work.
8234 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
8239 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
8241 /* Socket errors? */
8242 error = sock_error(sk);
8246 if (!skb_queue_empty(&sk->sk_receive_queue))
8249 /* Socket shut down? */
8250 if (sk->sk_shutdown & RCV_SHUTDOWN)
8253 /* Sequenced packets can come disconnected. If so we report the
8258 /* Is there a good reason to think that we may receive some data? */
8259 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
8262 /* Handle signals. */
8263 if (signal_pending(current))
8266 /* Let another process have a go. Since we are going to sleep
8267 * anyway. Note: This may cause odd behaviors if the message
8268 * does not fit in the user's buffer, but this seems to be the
8269 * only way to honor MSG_DONTWAIT realistically.
8272 *timeo_p = schedule_timeout(*timeo_p);
8276 finish_wait(sk_sleep(sk), &wait);
8280 error = sock_intr_errno(*timeo_p);
8283 finish_wait(sk_sleep(sk), &wait);
8288 /* Receive a datagram.
8289 * Note: This is pretty much the same routine as in core/datagram.c
8290 * with a few changes to make lksctp work.
8292 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
8293 int noblock, int *err)
8296 struct sk_buff *skb;
8299 timeo = sock_rcvtimeo(sk, noblock);
8301 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
8302 MAX_SCHEDULE_TIMEOUT);
8305 /* Again only user level code calls this function,
8306 * so nothing interrupt level
8307 * will suddenly eat the receive_queue.
8309 * Look at current nfs client by the way...
8310 * However, this function was correct in any case. 8)
8312 if (flags & MSG_PEEK) {
8313 skb = skb_peek(&sk->sk_receive_queue);
8315 refcount_inc(&skb->users);
8317 skb = __skb_dequeue(&sk->sk_receive_queue);
8323 /* Caller is allowed not to check sk->sk_err before calling. */
8324 error = sock_error(sk);
8328 if (sk->sk_shutdown & RCV_SHUTDOWN)
8331 if (sk_can_busy_loop(sk)) {
8332 sk_busy_loop(sk, noblock);
8334 if (!skb_queue_empty(&sk->sk_receive_queue))
8338 /* User doesn't want to wait. */
8342 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
8351 /* If sndbuf has changed, wake up per association sndbuf waiters. */
8352 static void __sctp_write_space(struct sctp_association *asoc)
8354 struct sock *sk = asoc->base.sk;
8356 if (sctp_wspace(asoc) <= 0)
8359 if (waitqueue_active(&asoc->wait))
8360 wake_up_interruptible(&asoc->wait);
8362 if (sctp_writeable(sk)) {
8363 struct socket_wq *wq;
8366 wq = rcu_dereference(sk->sk_wq);
8368 if (waitqueue_active(&wq->wait))
8369 wake_up_interruptible(&wq->wait);
8371 /* Note that we try to include the Async I/O support
8372 * here by modeling from the current TCP/UDP code.
8373 * We have not tested with it yet.
8375 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
8376 sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT);
8382 static void sctp_wake_up_waiters(struct sock *sk,
8383 struct sctp_association *asoc)
8385 struct sctp_association *tmp = asoc;
8387 /* We do accounting for the sndbuf space per association,
8388 * so we only need to wake our own association.
8390 if (asoc->ep->sndbuf_policy)
8391 return __sctp_write_space(asoc);
8393 /* If association goes down and is just flushing its
8394 * outq, then just normally notify others.
8396 if (asoc->base.dead)
8397 return sctp_write_space(sk);
8399 /* Accounting for the sndbuf space is per socket, so we
8400 * need to wake up others, try to be fair and in case of
8401 * other associations, let them have a go first instead
8402 * of just doing a sctp_write_space() call.
8404 * Note that we reach sctp_wake_up_waiters() only when
8405 * associations free up queued chunks, thus we are under
8406 * lock and the list of associations on a socket is
8407 * guaranteed not to change.
8409 for (tmp = list_next_entry(tmp, asocs); 1;
8410 tmp = list_next_entry(tmp, asocs)) {
8411 /* Manually skip the head element. */
8412 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
8414 /* Wake up association. */
8415 __sctp_write_space(tmp);
8416 /* We've reached the end. */
8422 /* Do accounting for the sndbuf space.
8423 * Decrement the used sndbuf space of the corresponding association by the
8424 * data size which was just transmitted(freed).
8426 static void sctp_wfree(struct sk_buff *skb)
8428 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
8429 struct sctp_association *asoc = chunk->asoc;
8430 struct sock *sk = asoc->base.sk;
8432 sk_mem_uncharge(sk, skb->truesize);
8433 sk->sk_wmem_queued -= skb->truesize + sizeof(struct sctp_chunk);
8434 asoc->sndbuf_used -= skb->truesize + sizeof(struct sctp_chunk);
8435 WARN_ON(refcount_sub_and_test(sizeof(struct sctp_chunk),
8436 &sk->sk_wmem_alloc));
8439 struct sctp_shared_key *shkey = chunk->shkey;
8441 /* refcnt == 2 and !list_empty mean after this release, it's
8442 * not being used anywhere, and it's time to notify userland
8443 * that this shkey can be freed if it's been deactivated.
8445 if (shkey->deactivated && !list_empty(&shkey->key_list) &&
8446 refcount_read(&shkey->refcnt) == 2) {
8447 struct sctp_ulpevent *ev;
8449 ev = sctp_ulpevent_make_authkey(asoc, shkey->key_id,
8453 asoc->stream.si->enqueue_event(&asoc->ulpq, ev);
8455 sctp_auth_shkey_release(chunk->shkey);
8459 sctp_wake_up_waiters(sk, asoc);
8461 sctp_association_put(asoc);
8464 /* Do accounting for the receive space on the socket.
8465 * Accounting for the association is done in ulpevent.c
8466 * We set this as a destructor for the cloned data skbs so that
8467 * accounting is done at the correct time.
8469 void sctp_sock_rfree(struct sk_buff *skb)
8471 struct sock *sk = skb->sk;
8472 struct sctp_ulpevent *event = sctp_skb2event(skb);
8474 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
8477 * Mimic the behavior of sock_rfree
8479 sk_mem_uncharge(sk, event->rmem_len);
8483 /* Helper function to wait for space in the sndbuf. */
8484 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
8487 struct sock *sk = asoc->base.sk;
8488 long current_timeo = *timeo_p;
8492 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
8495 /* Increment the association's refcnt. */
8496 sctp_association_hold(asoc);
8498 /* Wait on the association specific sndbuf space. */
8500 prepare_to_wait_exclusive(&asoc->wait, &wait,
8501 TASK_INTERRUPTIBLE);
8502 if (asoc->base.dead)
8506 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING)
8508 if (signal_pending(current))
8509 goto do_interrupted;
8510 if ((int)msg_len <= sctp_wspace(asoc))
8513 /* Let another process have a go. Since we are going
8517 current_timeo = schedule_timeout(current_timeo);
8519 if (sk != asoc->base.sk)
8522 *timeo_p = current_timeo;
8526 finish_wait(&asoc->wait, &wait);
8528 /* Release the association's refcnt. */
8529 sctp_association_put(asoc);
8542 err = sock_intr_errno(*timeo_p);
8550 void sctp_data_ready(struct sock *sk)
8552 struct socket_wq *wq;
8555 wq = rcu_dereference(sk->sk_wq);
8556 if (skwq_has_sleeper(wq))
8557 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
8558 EPOLLRDNORM | EPOLLRDBAND);
8559 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
8563 /* If socket sndbuf has changed, wake up all per association waiters. */
8564 void sctp_write_space(struct sock *sk)
8566 struct sctp_association *asoc;
8568 /* Wake up the tasks in each wait queue. */
8569 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
8570 __sctp_write_space(asoc);
8574 /* Is there any sndbuf space available on the socket?
8576 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
8577 * associations on the same socket. For a UDP-style socket with
8578 * multiple associations, it is possible for it to be "unwriteable"
8579 * prematurely. I assume that this is acceptable because
8580 * a premature "unwriteable" is better than an accidental "writeable" which
8581 * would cause an unwanted block under certain circumstances. For the 1-1
8582 * UDP-style sockets or TCP-style sockets, this code should work.
8585 static bool sctp_writeable(struct sock *sk)
8587 return sk->sk_sndbuf > sk->sk_wmem_queued;
8590 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
8591 * returns immediately with EINPROGRESS.
8593 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
8595 struct sock *sk = asoc->base.sk;
8597 long current_timeo = *timeo_p;
8600 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
8602 /* Increment the association's refcnt. */
8603 sctp_association_hold(asoc);
8606 prepare_to_wait_exclusive(&asoc->wait, &wait,
8607 TASK_INTERRUPTIBLE);
8610 if (sk->sk_shutdown & RCV_SHUTDOWN)
8612 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
8615 if (signal_pending(current))
8616 goto do_interrupted;
8618 if (sctp_state(asoc, ESTABLISHED))
8621 /* Let another process have a go. Since we are going
8625 current_timeo = schedule_timeout(current_timeo);
8628 *timeo_p = current_timeo;
8632 finish_wait(&asoc->wait, &wait);
8634 /* Release the association's refcnt. */
8635 sctp_association_put(asoc);
8640 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
8643 err = -ECONNREFUSED;
8647 err = sock_intr_errno(*timeo_p);
8655 static int sctp_wait_for_accept(struct sock *sk, long timeo)
8657 struct sctp_endpoint *ep;
8661 ep = sctp_sk(sk)->ep;
8665 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
8666 TASK_INTERRUPTIBLE);
8668 if (list_empty(&ep->asocs)) {
8670 timeo = schedule_timeout(timeo);
8675 if (!sctp_sstate(sk, LISTENING))
8679 if (!list_empty(&ep->asocs))
8682 err = sock_intr_errno(timeo);
8683 if (signal_pending(current))
8691 finish_wait(sk_sleep(sk), &wait);
8696 static void sctp_wait_for_close(struct sock *sk, long timeout)
8701 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
8702 if (list_empty(&sctp_sk(sk)->ep->asocs))
8705 timeout = schedule_timeout(timeout);
8707 } while (!signal_pending(current) && timeout);
8709 finish_wait(sk_sleep(sk), &wait);
8712 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
8714 struct sk_buff *frag;
8719 /* Don't forget the fragments. */
8720 skb_walk_frags(skb, frag)
8721 sctp_skb_set_owner_r_frag(frag, sk);
8724 sctp_skb_set_owner_r(skb, sk);
8727 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
8728 struct sctp_association *asoc)
8730 struct inet_sock *inet = inet_sk(sk);
8731 struct inet_sock *newinet;
8732 struct sctp_sock *sp = sctp_sk(sk);
8733 struct sctp_endpoint *ep = sp->ep;
8735 newsk->sk_type = sk->sk_type;
8736 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
8737 newsk->sk_flags = sk->sk_flags;
8738 newsk->sk_tsflags = sk->sk_tsflags;
8739 newsk->sk_no_check_tx = sk->sk_no_check_tx;
8740 newsk->sk_no_check_rx = sk->sk_no_check_rx;
8741 newsk->sk_reuse = sk->sk_reuse;
8742 sctp_sk(newsk)->reuse = sp->reuse;
8744 newsk->sk_shutdown = sk->sk_shutdown;
8745 newsk->sk_destruct = sctp_destruct_sock;
8746 newsk->sk_family = sk->sk_family;
8747 newsk->sk_protocol = IPPROTO_SCTP;
8748 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
8749 newsk->sk_sndbuf = sk->sk_sndbuf;
8750 newsk->sk_rcvbuf = sk->sk_rcvbuf;
8751 newsk->sk_lingertime = sk->sk_lingertime;
8752 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
8753 newsk->sk_sndtimeo = sk->sk_sndtimeo;
8754 newsk->sk_rxhash = sk->sk_rxhash;
8756 newinet = inet_sk(newsk);
8758 /* Initialize sk's sport, dport, rcv_saddr and daddr for
8759 * getsockname() and getpeername()
8761 newinet->inet_sport = inet->inet_sport;
8762 newinet->inet_saddr = inet->inet_saddr;
8763 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
8764 newinet->inet_dport = htons(asoc->peer.port);
8765 newinet->pmtudisc = inet->pmtudisc;
8766 newinet->inet_id = asoc->next_tsn ^ jiffies;
8768 newinet->uc_ttl = inet->uc_ttl;
8769 newinet->mc_loop = 1;
8770 newinet->mc_ttl = 1;
8771 newinet->mc_index = 0;
8772 newinet->mc_list = NULL;
8774 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
8775 net_enable_timestamp();
8777 /* Set newsk security attributes from orginal sk and connection
8778 * security attribute from ep.
8780 security_sctp_sk_clone(ep, sk, newsk);
8783 static inline void sctp_copy_descendant(struct sock *sk_to,
8784 const struct sock *sk_from)
8786 int ancestor_size = sizeof(struct inet_sock) +
8787 sizeof(struct sctp_sock) -
8788 offsetof(struct sctp_sock, auto_asconf_list);
8790 if (sk_from->sk_family == PF_INET6)
8791 ancestor_size += sizeof(struct ipv6_pinfo);
8793 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
8796 /* Populate the fields of the newsk from the oldsk and migrate the assoc
8797 * and its messages to the newsk.
8799 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
8800 struct sctp_association *assoc,
8801 enum sctp_socket_type type)
8803 struct sctp_sock *oldsp = sctp_sk(oldsk);
8804 struct sctp_sock *newsp = sctp_sk(newsk);
8805 struct sctp_bind_bucket *pp; /* hash list port iterator */
8806 struct sctp_endpoint *newep = newsp->ep;
8807 struct sk_buff *skb, *tmp;
8808 struct sctp_ulpevent *event;
8809 struct sctp_bind_hashbucket *head;
8811 /* Migrate socket buffer sizes and all the socket level options to the
8814 newsk->sk_sndbuf = oldsk->sk_sndbuf;
8815 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
8816 /* Brute force copy old sctp opt. */
8817 sctp_copy_descendant(newsk, oldsk);
8819 /* Restore the ep value that was overwritten with the above structure
8825 /* Hook this new socket in to the bind_hash list. */
8826 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
8827 inet_sk(oldsk)->inet_num)];
8828 spin_lock_bh(&head->lock);
8829 pp = sctp_sk(oldsk)->bind_hash;
8830 sk_add_bind_node(newsk, &pp->owner);
8831 sctp_sk(newsk)->bind_hash = pp;
8832 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
8833 spin_unlock_bh(&head->lock);
8835 /* Copy the bind_addr list from the original endpoint to the new
8836 * endpoint so that we can handle restarts properly
8838 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
8839 &oldsp->ep->base.bind_addr, GFP_KERNEL);
8841 /* Move any messages in the old socket's receive queue that are for the
8842 * peeled off association to the new socket's receive queue.
8844 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
8845 event = sctp_skb2event(skb);
8846 if (event->asoc == assoc) {
8847 __skb_unlink(skb, &oldsk->sk_receive_queue);
8848 __skb_queue_tail(&newsk->sk_receive_queue, skb);
8849 sctp_skb_set_owner_r_frag(skb, newsk);
8853 /* Clean up any messages pending delivery due to partial
8854 * delivery. Three cases:
8855 * 1) No partial deliver; no work.
8856 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
8857 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
8859 skb_queue_head_init(&newsp->pd_lobby);
8860 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
8862 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
8863 struct sk_buff_head *queue;
8865 /* Decide which queue to move pd_lobby skbs to. */
8866 if (assoc->ulpq.pd_mode) {
8867 queue = &newsp->pd_lobby;
8869 queue = &newsk->sk_receive_queue;
8871 /* Walk through the pd_lobby, looking for skbs that
8872 * need moved to the new socket.
8874 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
8875 event = sctp_skb2event(skb);
8876 if (event->asoc == assoc) {
8877 __skb_unlink(skb, &oldsp->pd_lobby);
8878 __skb_queue_tail(queue, skb);
8879 sctp_skb_set_owner_r_frag(skb, newsk);
8883 /* Clear up any skbs waiting for the partial
8884 * delivery to finish.
8886 if (assoc->ulpq.pd_mode)
8887 sctp_clear_pd(oldsk, NULL);
8891 sctp_for_each_rx_skb(assoc, newsk, sctp_skb_set_owner_r_frag);
8893 /* Set the type of socket to indicate that it is peeled off from the
8894 * original UDP-style socket or created with the accept() call on a
8895 * TCP-style socket..
8899 /* Mark the new socket "in-use" by the user so that any packets
8900 * that may arrive on the association after we've moved it are
8901 * queued to the backlog. This prevents a potential race between
8902 * backlog processing on the old socket and new-packet processing
8903 * on the new socket.
8905 * The caller has just allocated newsk so we can guarantee that other
8906 * paths won't try to lock it and then oldsk.
8908 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
8909 sctp_for_each_tx_datachunk(assoc, sctp_clear_owner_w);
8910 sctp_assoc_migrate(assoc, newsk);
8911 sctp_for_each_tx_datachunk(assoc, sctp_set_owner_w);
8913 /* If the association on the newsk is already closed before accept()
8914 * is called, set RCV_SHUTDOWN flag.
8916 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP)) {
8917 inet_sk_set_state(newsk, SCTP_SS_CLOSED);
8918 newsk->sk_shutdown |= RCV_SHUTDOWN;
8920 inet_sk_set_state(newsk, SCTP_SS_ESTABLISHED);
8923 release_sock(newsk);
8927 /* This proto struct describes the ULP interface for SCTP. */
8928 struct proto sctp_prot = {
8930 .owner = THIS_MODULE,
8931 .close = sctp_close,
8932 .disconnect = sctp_disconnect,
8933 .accept = sctp_accept,
8934 .ioctl = sctp_ioctl,
8935 .init = sctp_init_sock,
8936 .destroy = sctp_destroy_sock,
8937 .shutdown = sctp_shutdown,
8938 .setsockopt = sctp_setsockopt,
8939 .getsockopt = sctp_getsockopt,
8940 .sendmsg = sctp_sendmsg,
8941 .recvmsg = sctp_recvmsg,
8943 .backlog_rcv = sctp_backlog_rcv,
8945 .unhash = sctp_unhash,
8946 .get_port = sctp_get_port,
8947 .obj_size = sizeof(struct sctp_sock),
8948 .useroffset = offsetof(struct sctp_sock, subscribe),
8949 .usersize = offsetof(struct sctp_sock, initmsg) -
8950 offsetof(struct sctp_sock, subscribe) +
8951 sizeof_field(struct sctp_sock, initmsg),
8952 .sysctl_mem = sysctl_sctp_mem,
8953 .sysctl_rmem = sysctl_sctp_rmem,
8954 .sysctl_wmem = sysctl_sctp_wmem,
8955 .memory_pressure = &sctp_memory_pressure,
8956 .enter_memory_pressure = sctp_enter_memory_pressure,
8957 .memory_allocated = &sctp_memory_allocated,
8958 .sockets_allocated = &sctp_sockets_allocated,
8961 #if IS_ENABLED(CONFIG_IPV6)
8963 #include <net/transp_v6.h>
8964 static void sctp_v6_destroy_sock(struct sock *sk)
8966 sctp_destroy_sock(sk);
8967 inet6_destroy_sock(sk);
8970 struct proto sctpv6_prot = {
8972 .owner = THIS_MODULE,
8973 .close = sctp_close,
8974 .disconnect = sctp_disconnect,
8975 .accept = sctp_accept,
8976 .ioctl = sctp_ioctl,
8977 .init = sctp_init_sock,
8978 .destroy = sctp_v6_destroy_sock,
8979 .shutdown = sctp_shutdown,
8980 .setsockopt = sctp_setsockopt,
8981 .getsockopt = sctp_getsockopt,
8982 .sendmsg = sctp_sendmsg,
8983 .recvmsg = sctp_recvmsg,
8985 .backlog_rcv = sctp_backlog_rcv,
8987 .unhash = sctp_unhash,
8988 .get_port = sctp_get_port,
8989 .obj_size = sizeof(struct sctp6_sock),
8990 .useroffset = offsetof(struct sctp6_sock, sctp.subscribe),
8991 .usersize = offsetof(struct sctp6_sock, sctp.initmsg) -
8992 offsetof(struct sctp6_sock, sctp.subscribe) +
8993 sizeof_field(struct sctp6_sock, sctp.initmsg),
8994 .sysctl_mem = sysctl_sctp_mem,
8995 .sysctl_rmem = sysctl_sctp_rmem,
8996 .sysctl_wmem = sysctl_sctp_wmem,
8997 .memory_pressure = &sctp_memory_pressure,
8998 .enter_memory_pressure = sctp_enter_memory_pressure,
8999 .memory_allocated = &sctp_memory_allocated,
9000 .sockets_allocated = &sctp_sockets_allocated,
9002 #endif /* IS_ENABLED(CONFIG_IPV6) */