1 /* SCTP kernel implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
9 * This file is part of the SCTP kernel implementation
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
18 * This SCTP implementation is free software;
19 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
24 * This SCTP implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, see
32 * <http://www.gnu.org/licenses/>.
34 * Please send any bug reports or fixes you make to the
36 * lksctp developers <linux-sctp@vger.kernel.org>
38 * Written or modified by:
39 * La Monte H.P. Yarroll <piggy@acm.org>
40 * Narasimha Budihal <narsi@refcode.org>
41 * Karl Knutson <karl@athena.chicago.il.us>
42 * Jon Grimm <jgrimm@us.ibm.com>
43 * Xingang Guo <xingang.guo@intel.com>
44 * Daisy Chang <daisyc@us.ibm.com>
45 * Sridhar Samudrala <samudrala@us.ibm.com>
46 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
47 * Ardelle Fan <ardelle.fan@intel.com>
48 * Ryan Layer <rmlayer@us.ibm.com>
49 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
50 * Kevin Gao <kevin.gao@intel.com>
53 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
55 #include <crypto/hash.h>
56 #include <linux/types.h>
57 #include <linux/kernel.h>
58 #include <linux/wait.h>
59 #include <linux/time.h>
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 <= SCTP_ALL_ASSOC)
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);
1869 iov_iter_revert(&msg->msg_iter, msg_len);
1877 static int sctp_sendmsg_to_asoc(struct sctp_association *asoc,
1878 struct msghdr *msg, size_t msg_len,
1879 struct sctp_transport *transport,
1880 struct sctp_sndrcvinfo *sinfo)
1882 struct sock *sk = asoc->base.sk;
1883 struct sctp_sock *sp = sctp_sk(sk);
1884 struct net *net = sock_net(sk);
1885 struct sctp_datamsg *datamsg;
1886 bool wait_connect = false;
1887 struct sctp_chunk *chunk;
1891 if (sinfo->sinfo_stream >= asoc->stream.outcnt) {
1896 if (unlikely(!SCTP_SO(&asoc->stream, sinfo->sinfo_stream)->ext)) {
1897 err = sctp_stream_init_ext(&asoc->stream, sinfo->sinfo_stream);
1902 if (sp->disable_fragments && msg_len > asoc->frag_point) {
1907 if (asoc->pmtu_pending) {
1908 if (sp->param_flags & SPP_PMTUD_ENABLE)
1909 sctp_assoc_sync_pmtu(asoc);
1910 asoc->pmtu_pending = 0;
1913 if (sctp_wspace(asoc) < (int)msg_len)
1914 sctp_prsctp_prune(asoc, sinfo, msg_len - sctp_wspace(asoc));
1916 if (sctp_wspace(asoc) <= 0) {
1917 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1918 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1923 if (sctp_state(asoc, CLOSED)) {
1924 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1928 if (sp->strm_interleave) {
1929 timeo = sock_sndtimeo(sk, 0);
1930 err = sctp_wait_for_connect(asoc, &timeo);
1936 wait_connect = true;
1939 pr_debug("%s: we associated primitively\n", __func__);
1942 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
1943 if (IS_ERR(datamsg)) {
1944 err = PTR_ERR(datamsg);
1948 asoc->force_delay = !!(msg->msg_flags & MSG_MORE);
1950 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1951 sctp_chunk_hold(chunk);
1952 sctp_set_owner_w(chunk);
1953 chunk->transport = transport;
1956 err = sctp_primitive_SEND(net, asoc, datamsg);
1958 sctp_datamsg_free(datamsg);
1962 pr_debug("%s: we sent primitively\n", __func__);
1964 sctp_datamsg_put(datamsg);
1966 if (unlikely(wait_connect)) {
1967 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1968 sctp_wait_for_connect(asoc, &timeo);
1977 static union sctp_addr *sctp_sendmsg_get_daddr(struct sock *sk,
1978 const struct msghdr *msg,
1979 struct sctp_cmsgs *cmsgs)
1981 union sctp_addr *daddr = NULL;
1984 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1985 int len = msg->msg_namelen;
1987 if (len > sizeof(*daddr))
1988 len = sizeof(*daddr);
1990 daddr = (union sctp_addr *)msg->msg_name;
1992 err = sctp_verify_addr(sk, daddr, len);
1994 return ERR_PTR(err);
2000 static void sctp_sendmsg_update_sinfo(struct sctp_association *asoc,
2001 struct sctp_sndrcvinfo *sinfo,
2002 struct sctp_cmsgs *cmsgs)
2004 if (!cmsgs->srinfo && !cmsgs->sinfo) {
2005 sinfo->sinfo_stream = asoc->default_stream;
2006 sinfo->sinfo_ppid = asoc->default_ppid;
2007 sinfo->sinfo_context = asoc->default_context;
2008 sinfo->sinfo_assoc_id = sctp_assoc2id(asoc);
2011 sinfo->sinfo_flags = asoc->default_flags;
2014 if (!cmsgs->srinfo && !cmsgs->prinfo)
2015 sinfo->sinfo_timetolive = asoc->default_timetolive;
2017 if (cmsgs->authinfo) {
2018 /* Reuse sinfo_tsn to indicate that authinfo was set and
2019 * sinfo_ssn to save the keyid on tx path.
2021 sinfo->sinfo_tsn = 1;
2022 sinfo->sinfo_ssn = cmsgs->authinfo->auth_keynumber;
2026 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
2028 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
2029 struct sctp_transport *transport = NULL;
2030 struct sctp_sndrcvinfo _sinfo, *sinfo;
2031 struct sctp_association *asoc, *tmp;
2032 struct sctp_cmsgs cmsgs;
2033 union sctp_addr *daddr;
2038 /* Parse and get snd_info */
2039 err = sctp_sendmsg_parse(sk, &cmsgs, &_sinfo, msg, msg_len);
2044 sflags = sinfo->sinfo_flags;
2046 /* Get daddr from msg */
2047 daddr = sctp_sendmsg_get_daddr(sk, msg, &cmsgs);
2048 if (IS_ERR(daddr)) {
2049 err = PTR_ERR(daddr);
2055 /* SCTP_SENDALL process */
2056 if ((sflags & SCTP_SENDALL) && sctp_style(sk, UDP)) {
2057 list_for_each_entry_safe(asoc, tmp, &ep->asocs, asocs) {
2058 err = sctp_sendmsg_check_sflags(asoc, sflags, msg,
2065 sctp_sendmsg_update_sinfo(asoc, sinfo, &cmsgs);
2067 err = sctp_sendmsg_to_asoc(asoc, msg, msg_len,
2072 iov_iter_revert(&msg->msg_iter, err);
2078 /* Get and check or create asoc */
2080 asoc = sctp_endpoint_lookup_assoc(ep, daddr, &transport);
2082 err = sctp_sendmsg_check_sflags(asoc, sflags, msg,
2087 err = sctp_sendmsg_new_asoc(sk, sflags, &cmsgs, daddr,
2092 asoc = transport->asoc;
2096 if (!sctp_style(sk, TCP) && !(sflags & SCTP_ADDR_OVER))
2099 asoc = sctp_id2assoc(sk, sinfo->sinfo_assoc_id);
2105 err = sctp_sendmsg_check_sflags(asoc, sflags, msg, msg_len);
2110 /* Update snd_info with the asoc */
2111 sctp_sendmsg_update_sinfo(asoc, sinfo, &cmsgs);
2113 /* Send msg to the asoc */
2114 err = sctp_sendmsg_to_asoc(asoc, msg, msg_len, transport, sinfo);
2115 if (err < 0 && err != -ESRCH && new)
2116 sctp_association_free(asoc);
2121 return sctp_error(sk, msg->msg_flags, err);
2124 /* This is an extended version of skb_pull() that removes the data from the
2125 * start of a skb even when data is spread across the list of skb's in the
2126 * frag_list. len specifies the total amount of data that needs to be removed.
2127 * when 'len' bytes could be removed from the skb, it returns 0.
2128 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2129 * could not be removed.
2131 static int sctp_skb_pull(struct sk_buff *skb, int len)
2133 struct sk_buff *list;
2134 int skb_len = skb_headlen(skb);
2137 if (len <= skb_len) {
2138 __skb_pull(skb, len);
2142 __skb_pull(skb, skb_len);
2144 skb_walk_frags(skb, list) {
2145 rlen = sctp_skb_pull(list, len);
2146 skb->len -= (len-rlen);
2147 skb->data_len -= (len-rlen);
2158 /* API 3.1.3 recvmsg() - UDP Style Syntax
2160 * ssize_t recvmsg(int socket, struct msghdr *message,
2163 * socket - the socket descriptor of the endpoint.
2164 * message - pointer to the msghdr structure which contains a single
2165 * user message and possibly some ancillary data.
2167 * See Section 5 for complete description of the data
2170 * flags - flags sent or received with the user message, see Section
2171 * 5 for complete description of the flags.
2173 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2174 int noblock, int flags, int *addr_len)
2176 struct sctp_ulpevent *event = NULL;
2177 struct sctp_sock *sp = sctp_sk(sk);
2178 struct sk_buff *skb, *head_skb;
2183 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2184 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2189 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED) &&
2190 !sctp_sstate(sk, CLOSING) && !sctp_sstate(sk, CLOSED)) {
2195 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2199 /* Get the total length of the skb including any skb's in the
2208 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2210 event = sctp_skb2event(skb);
2215 if (event->chunk && event->chunk->head_skb)
2216 head_skb = event->chunk->head_skb;
2219 sock_recv_ts_and_drops(msg, sk, head_skb);
2220 if (sctp_ulpevent_is_notification(event)) {
2221 msg->msg_flags |= MSG_NOTIFICATION;
2222 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2224 sp->pf->skb_msgname(head_skb, msg->msg_name, addr_len);
2227 /* Check if we allow SCTP_NXTINFO. */
2228 if (sp->recvnxtinfo)
2229 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2230 /* Check if we allow SCTP_RCVINFO. */
2231 if (sp->recvrcvinfo)
2232 sctp_ulpevent_read_rcvinfo(event, msg);
2233 /* Check if we allow SCTP_SNDRCVINFO. */
2234 if (sctp_ulpevent_type_enabled(sp->subscribe, SCTP_DATA_IO_EVENT))
2235 sctp_ulpevent_read_sndrcvinfo(event, msg);
2239 /* If skb's length exceeds the user's buffer, update the skb and
2240 * push it back to the receive_queue so that the next call to
2241 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2243 if (skb_len > copied) {
2244 msg->msg_flags &= ~MSG_EOR;
2245 if (flags & MSG_PEEK)
2247 sctp_skb_pull(skb, copied);
2248 skb_queue_head(&sk->sk_receive_queue, skb);
2250 /* When only partial message is copied to the user, increase
2251 * rwnd by that amount. If all the data in the skb is read,
2252 * rwnd is updated when the event is freed.
2254 if (!sctp_ulpevent_is_notification(event))
2255 sctp_assoc_rwnd_increase(event->asoc, copied);
2257 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2258 (event->msg_flags & MSG_EOR))
2259 msg->msg_flags |= MSG_EOR;
2261 msg->msg_flags &= ~MSG_EOR;
2264 if (flags & MSG_PEEK) {
2265 /* Release the skb reference acquired after peeking the skb in
2266 * sctp_skb_recv_datagram().
2270 /* Free the event which includes releasing the reference to
2271 * the owner of the skb, freeing the skb and updating the
2274 sctp_ulpevent_free(event);
2281 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2283 * This option is a on/off flag. If enabled no SCTP message
2284 * fragmentation will be performed. Instead if a message being sent
2285 * exceeds the current PMTU size, the message will NOT be sent and
2286 * instead a error will be indicated to the user.
2288 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2289 char __user *optval,
2290 unsigned int optlen)
2294 if (optlen < sizeof(int))
2297 if (get_user(val, (int __user *)optval))
2300 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2305 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2306 unsigned int optlen)
2308 struct sctp_event_subscribe subscribe;
2309 __u8 *sn_type = (__u8 *)&subscribe;
2310 struct sctp_sock *sp = sctp_sk(sk);
2311 struct sctp_association *asoc;
2314 if (optlen > sizeof(struct sctp_event_subscribe))
2317 if (copy_from_user(&subscribe, optval, optlen))
2320 for (i = 0; i < optlen; i++)
2321 sctp_ulpevent_type_set(&sp->subscribe, SCTP_SN_TYPE_BASE + i,
2324 list_for_each_entry(asoc, &sp->ep->asocs, asocs)
2325 asoc->subscribe = sctp_sk(sk)->subscribe;
2327 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2328 * if there is no data to be sent or retransmit, the stack will
2329 * immediately send up this notification.
2331 if (sctp_ulpevent_type_enabled(sp->subscribe, SCTP_SENDER_DRY_EVENT)) {
2332 struct sctp_ulpevent *event;
2334 asoc = sctp_id2assoc(sk, 0);
2335 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2336 event = sctp_ulpevent_make_sender_dry_event(asoc,
2337 GFP_USER | __GFP_NOWARN);
2341 asoc->stream.si->enqueue_event(&asoc->ulpq, event);
2348 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2350 * This socket option is applicable to the UDP-style socket only. When
2351 * set it will cause associations that are idle for more than the
2352 * specified number of seconds to automatically close. An association
2353 * being idle is defined an association that has NOT sent or received
2354 * user data. The special value of '0' indicates that no automatic
2355 * close of any associations should be performed. The option expects an
2356 * integer defining the number of seconds of idle time before an
2357 * association is closed.
2359 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2360 unsigned int optlen)
2362 struct sctp_sock *sp = sctp_sk(sk);
2363 struct net *net = sock_net(sk);
2365 /* Applicable to UDP-style socket only */
2366 if (sctp_style(sk, TCP))
2368 if (optlen != sizeof(int))
2370 if (copy_from_user(&sp->autoclose, optval, optlen))
2373 if (sp->autoclose > net->sctp.max_autoclose)
2374 sp->autoclose = net->sctp.max_autoclose;
2379 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2381 * Applications can enable or disable heartbeats for any peer address of
2382 * an association, modify an address's heartbeat interval, force a
2383 * heartbeat to be sent immediately, and adjust the address's maximum
2384 * number of retransmissions sent before an address is considered
2385 * unreachable. The following structure is used to access and modify an
2386 * address's parameters:
2388 * struct sctp_paddrparams {
2389 * sctp_assoc_t spp_assoc_id;
2390 * struct sockaddr_storage spp_address;
2391 * uint32_t spp_hbinterval;
2392 * uint16_t spp_pathmaxrxt;
2393 * uint32_t spp_pathmtu;
2394 * uint32_t spp_sackdelay;
2395 * uint32_t spp_flags;
2396 * uint32_t spp_ipv6_flowlabel;
2400 * spp_assoc_id - (one-to-many style socket) This is filled in the
2401 * application, and identifies the association for
2403 * spp_address - This specifies which address is of interest.
2404 * spp_hbinterval - This contains the value of the heartbeat interval,
2405 * in milliseconds. If a value of zero
2406 * is present in this field then no changes are to
2407 * be made to this parameter.
2408 * spp_pathmaxrxt - This contains the maximum number of
2409 * retransmissions before this address shall be
2410 * considered unreachable. If a value of zero
2411 * is present in this field then no changes are to
2412 * be made to this parameter.
2413 * spp_pathmtu - When Path MTU discovery is disabled the value
2414 * specified here will be the "fixed" path mtu.
2415 * Note that if the spp_address field is empty
2416 * then all associations on this address will
2417 * have this fixed path mtu set upon them.
2419 * spp_sackdelay - When delayed sack is enabled, this value specifies
2420 * the number of milliseconds that sacks will be delayed
2421 * for. This value will apply to all addresses of an
2422 * association if the spp_address field is empty. Note
2423 * also, that if delayed sack is enabled and this
2424 * value is set to 0, no change is made to the last
2425 * recorded delayed sack timer value.
2427 * spp_flags - These flags are used to control various features
2428 * on an association. The flag field may contain
2429 * zero or more of the following options.
2431 * SPP_HB_ENABLE - Enable heartbeats on the
2432 * specified address. Note that if the address
2433 * field is empty all addresses for the association
2434 * have heartbeats enabled upon them.
2436 * SPP_HB_DISABLE - Disable heartbeats on the
2437 * speicifed address. Note that if the address
2438 * field is empty all addresses for the association
2439 * will have their heartbeats disabled. Note also
2440 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2441 * mutually exclusive, only one of these two should
2442 * be specified. Enabling both fields will have
2443 * undetermined results.
2445 * SPP_HB_DEMAND - Request a user initiated heartbeat
2446 * to be made immediately.
2448 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2449 * heartbeat delayis to be set to the value of 0
2452 * SPP_PMTUD_ENABLE - This field will enable PMTU
2453 * discovery upon the specified address. Note that
2454 * if the address feild is empty then all addresses
2455 * on the association are effected.
2457 * SPP_PMTUD_DISABLE - This field will disable PMTU
2458 * discovery upon the specified address. Note that
2459 * if the address feild is empty then all addresses
2460 * on the association are effected. Not also that
2461 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2462 * exclusive. Enabling both will have undetermined
2465 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2466 * on delayed sack. The time specified in spp_sackdelay
2467 * is used to specify the sack delay for this address. Note
2468 * that if spp_address is empty then all addresses will
2469 * enable delayed sack and take on the sack delay
2470 * value specified in spp_sackdelay.
2471 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2472 * off delayed sack. If the spp_address field is blank then
2473 * delayed sack is disabled for the entire association. Note
2474 * also that this field is mutually exclusive to
2475 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2478 * SPP_IPV6_FLOWLABEL: Setting this flag enables the
2479 * setting of the IPV6 flow label value. The value is
2480 * contained in the spp_ipv6_flowlabel field.
2481 * Upon retrieval, this flag will be set to indicate that
2482 * the spp_ipv6_flowlabel field has a valid value returned.
2483 * If a specific destination address is set (in the
2484 * spp_address field), then the value returned is that of
2485 * the address. If just an association is specified (and
2486 * no address), then the association's default flow label
2487 * is returned. If neither an association nor a destination
2488 * is specified, then the socket's default flow label is
2489 * returned. For non-IPv6 sockets, this flag will be left
2492 * SPP_DSCP: Setting this flag enables the setting of the
2493 * Differentiated Services Code Point (DSCP) value
2494 * associated with either the association or a specific
2495 * address. The value is obtained in the spp_dscp field.
2496 * Upon retrieval, this flag will be set to indicate that
2497 * the spp_dscp field has a valid value returned. If a
2498 * specific destination address is set when called (in the
2499 * spp_address field), then that specific destination
2500 * address's DSCP value is returned. If just an association
2501 * is specified, then the association's default DSCP is
2502 * returned. If neither an association nor a destination is
2503 * specified, then the socket's default DSCP is returned.
2505 * spp_ipv6_flowlabel
2506 * - This field is used in conjunction with the
2507 * SPP_IPV6_FLOWLABEL flag and contains the IPv6 flow label.
2508 * The 20 least significant bits are used for the flow
2509 * label. This setting has precedence over any IPv6-layer
2512 * spp_dscp - This field is used in conjunction with the SPP_DSCP flag
2513 * and contains the DSCP. The 6 most significant bits are
2514 * used for the DSCP. This setting has precedence over any
2515 * IPv4- or IPv6- layer setting.
2517 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2518 struct sctp_transport *trans,
2519 struct sctp_association *asoc,
2520 struct sctp_sock *sp,
2523 int sackdelay_change)
2527 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2528 struct net *net = sock_net(trans->asoc->base.sk);
2530 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2535 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2536 * this field is ignored. Note also that a value of zero indicates
2537 * the current setting should be left unchanged.
2539 if (params->spp_flags & SPP_HB_ENABLE) {
2541 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2542 * set. This lets us use 0 value when this flag
2545 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2546 params->spp_hbinterval = 0;
2548 if (params->spp_hbinterval ||
2549 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2552 msecs_to_jiffies(params->spp_hbinterval);
2555 msecs_to_jiffies(params->spp_hbinterval);
2557 sp->hbinterval = params->spp_hbinterval;
2564 trans->param_flags =
2565 (trans->param_flags & ~SPP_HB) | hb_change;
2568 (asoc->param_flags & ~SPP_HB) | hb_change;
2571 (sp->param_flags & ~SPP_HB) | hb_change;
2575 /* When Path MTU discovery is disabled the value specified here will
2576 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2577 * include the flag SPP_PMTUD_DISABLE for this field to have any
2580 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2582 trans->pathmtu = params->spp_pathmtu;
2583 sctp_assoc_sync_pmtu(asoc);
2585 sctp_assoc_set_pmtu(asoc, params->spp_pathmtu);
2587 sp->pathmtu = params->spp_pathmtu;
2593 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2594 (params->spp_flags & SPP_PMTUD_ENABLE);
2595 trans->param_flags =
2596 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2598 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2599 sctp_assoc_sync_pmtu(asoc);
2603 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2606 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2610 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2611 * value of this field is ignored. Note also that a value of zero
2612 * indicates the current setting should be left unchanged.
2614 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2617 msecs_to_jiffies(params->spp_sackdelay);
2620 msecs_to_jiffies(params->spp_sackdelay);
2622 sp->sackdelay = params->spp_sackdelay;
2626 if (sackdelay_change) {
2628 trans->param_flags =
2629 (trans->param_flags & ~SPP_SACKDELAY) |
2633 (asoc->param_flags & ~SPP_SACKDELAY) |
2637 (sp->param_flags & ~SPP_SACKDELAY) |
2642 /* Note that a value of zero indicates the current setting should be
2645 if (params->spp_pathmaxrxt) {
2647 trans->pathmaxrxt = params->spp_pathmaxrxt;
2649 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2651 sp->pathmaxrxt = params->spp_pathmaxrxt;
2655 if (params->spp_flags & SPP_IPV6_FLOWLABEL) {
2657 if (trans->ipaddr.sa.sa_family == AF_INET6) {
2658 trans->flowlabel = params->spp_ipv6_flowlabel &
2659 SCTP_FLOWLABEL_VAL_MASK;
2660 trans->flowlabel |= SCTP_FLOWLABEL_SET_MASK;
2663 struct sctp_transport *t;
2665 list_for_each_entry(t, &asoc->peer.transport_addr_list,
2667 if (t->ipaddr.sa.sa_family != AF_INET6)
2669 t->flowlabel = params->spp_ipv6_flowlabel &
2670 SCTP_FLOWLABEL_VAL_MASK;
2671 t->flowlabel |= SCTP_FLOWLABEL_SET_MASK;
2673 asoc->flowlabel = params->spp_ipv6_flowlabel &
2674 SCTP_FLOWLABEL_VAL_MASK;
2675 asoc->flowlabel |= SCTP_FLOWLABEL_SET_MASK;
2676 } else if (sctp_opt2sk(sp)->sk_family == AF_INET6) {
2677 sp->flowlabel = params->spp_ipv6_flowlabel &
2678 SCTP_FLOWLABEL_VAL_MASK;
2679 sp->flowlabel |= SCTP_FLOWLABEL_SET_MASK;
2683 if (params->spp_flags & SPP_DSCP) {
2685 trans->dscp = params->spp_dscp & SCTP_DSCP_VAL_MASK;
2686 trans->dscp |= SCTP_DSCP_SET_MASK;
2688 struct sctp_transport *t;
2690 list_for_each_entry(t, &asoc->peer.transport_addr_list,
2692 t->dscp = params->spp_dscp &
2694 t->dscp |= SCTP_DSCP_SET_MASK;
2696 asoc->dscp = params->spp_dscp & SCTP_DSCP_VAL_MASK;
2697 asoc->dscp |= SCTP_DSCP_SET_MASK;
2699 sp->dscp = params->spp_dscp & SCTP_DSCP_VAL_MASK;
2700 sp->dscp |= SCTP_DSCP_SET_MASK;
2707 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2708 char __user *optval,
2709 unsigned int optlen)
2711 struct sctp_paddrparams params;
2712 struct sctp_transport *trans = NULL;
2713 struct sctp_association *asoc = NULL;
2714 struct sctp_sock *sp = sctp_sk(sk);
2716 int hb_change, pmtud_change, sackdelay_change;
2718 if (optlen == sizeof(params)) {
2719 if (copy_from_user(¶ms, optval, optlen))
2721 } else if (optlen == ALIGN(offsetof(struct sctp_paddrparams,
2722 spp_ipv6_flowlabel), 4)) {
2723 if (copy_from_user(¶ms, optval, optlen))
2725 if (params.spp_flags & (SPP_DSCP | SPP_IPV6_FLOWLABEL))
2731 /* Validate flags and value parameters. */
2732 hb_change = params.spp_flags & SPP_HB;
2733 pmtud_change = params.spp_flags & SPP_PMTUD;
2734 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2736 if (hb_change == SPP_HB ||
2737 pmtud_change == SPP_PMTUD ||
2738 sackdelay_change == SPP_SACKDELAY ||
2739 params.spp_sackdelay > 500 ||
2740 (params.spp_pathmtu &&
2741 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2744 /* If an address other than INADDR_ANY is specified, and
2745 * no transport is found, then the request is invalid.
2747 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
2748 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2749 params.spp_assoc_id);
2754 /* Get association, if assoc_id != SCTP_FUTURE_ASSOC and the
2755 * socket is a one to many style socket, and an association
2756 * was not found, then the id was invalid.
2758 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2759 if (!asoc && params.spp_assoc_id != SCTP_FUTURE_ASSOC &&
2760 sctp_style(sk, UDP))
2763 /* Heartbeat demand can only be sent on a transport or
2764 * association, but not a socket.
2766 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2769 /* Process parameters. */
2770 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2771 hb_change, pmtud_change,
2777 /* If changes are for association, also apply parameters to each
2780 if (!trans && asoc) {
2781 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2783 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2784 hb_change, pmtud_change,
2792 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2794 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2797 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2799 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2802 static void sctp_apply_asoc_delayed_ack(struct sctp_sack_info *params,
2803 struct sctp_association *asoc)
2805 struct sctp_transport *trans;
2807 if (params->sack_delay) {
2808 asoc->sackdelay = msecs_to_jiffies(params->sack_delay);
2810 sctp_spp_sackdelay_enable(asoc->param_flags);
2812 if (params->sack_freq == 1) {
2814 sctp_spp_sackdelay_disable(asoc->param_flags);
2815 } else if (params->sack_freq > 1) {
2816 asoc->sackfreq = params->sack_freq;
2818 sctp_spp_sackdelay_enable(asoc->param_flags);
2821 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2823 if (params->sack_delay) {
2824 trans->sackdelay = msecs_to_jiffies(params->sack_delay);
2825 trans->param_flags =
2826 sctp_spp_sackdelay_enable(trans->param_flags);
2828 if (params->sack_freq == 1) {
2829 trans->param_flags =
2830 sctp_spp_sackdelay_disable(trans->param_flags);
2831 } else if (params->sack_freq > 1) {
2832 trans->sackfreq = params->sack_freq;
2833 trans->param_flags =
2834 sctp_spp_sackdelay_enable(trans->param_flags);
2840 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2842 * This option will effect the way delayed acks are performed. This
2843 * option allows you to get or set the delayed ack time, in
2844 * milliseconds. It also allows changing the delayed ack frequency.
2845 * Changing the frequency to 1 disables the delayed sack algorithm. If
2846 * the assoc_id is 0, then this sets or gets the endpoints default
2847 * values. If the assoc_id field is non-zero, then the set or get
2848 * effects the specified association for the one to many model (the
2849 * assoc_id field is ignored by the one to one model). Note that if
2850 * sack_delay or sack_freq are 0 when setting this option, then the
2851 * current values will remain unchanged.
2853 * struct sctp_sack_info {
2854 * sctp_assoc_t sack_assoc_id;
2855 * uint32_t sack_delay;
2856 * uint32_t sack_freq;
2859 * sack_assoc_id - This parameter, indicates which association the user
2860 * is performing an action upon. Note that if this field's value is
2861 * zero then the endpoints default value is changed (effecting future
2862 * associations only).
2864 * sack_delay - This parameter contains the number of milliseconds that
2865 * the user is requesting the delayed ACK timer be set to. Note that
2866 * this value is defined in the standard to be between 200 and 500
2869 * sack_freq - This parameter contains the number of packets that must
2870 * be received before a sack is sent without waiting for the delay
2871 * timer to expire. The default value for this is 2, setting this
2872 * value to 1 will disable the delayed sack algorithm.
2875 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2876 char __user *optval, unsigned int optlen)
2878 struct sctp_sock *sp = sctp_sk(sk);
2879 struct sctp_association *asoc;
2880 struct sctp_sack_info params;
2882 if (optlen == sizeof(struct sctp_sack_info)) {
2883 if (copy_from_user(¶ms, optval, optlen))
2886 if (params.sack_delay == 0 && params.sack_freq == 0)
2888 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2889 pr_warn_ratelimited(DEPRECATED
2891 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2892 "Use struct sctp_sack_info instead\n",
2893 current->comm, task_pid_nr(current));
2894 if (copy_from_user(¶ms, optval, optlen))
2897 if (params.sack_delay == 0)
2898 params.sack_freq = 1;
2900 params.sack_freq = 0;
2904 /* Validate value parameter. */
2905 if (params.sack_delay > 500)
2908 /* Get association, if sack_assoc_id != SCTP_FUTURE_ASSOC and the
2909 * socket is a one to many style socket, and an association
2910 * was not found, then the id was invalid.
2912 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2913 if (!asoc && params.sack_assoc_id > SCTP_ALL_ASSOC &&
2914 sctp_style(sk, UDP))
2918 sctp_apply_asoc_delayed_ack(¶ms, asoc);
2923 if (params.sack_assoc_id == SCTP_FUTURE_ASSOC ||
2924 params.sack_assoc_id == SCTP_ALL_ASSOC) {
2925 if (params.sack_delay) {
2926 sp->sackdelay = params.sack_delay;
2928 sctp_spp_sackdelay_enable(sp->param_flags);
2930 if (params.sack_freq == 1) {
2932 sctp_spp_sackdelay_disable(sp->param_flags);
2933 } else if (params.sack_freq > 1) {
2934 sp->sackfreq = params.sack_freq;
2936 sctp_spp_sackdelay_enable(sp->param_flags);
2940 if (params.sack_assoc_id == SCTP_CURRENT_ASSOC ||
2941 params.sack_assoc_id == SCTP_ALL_ASSOC)
2942 list_for_each_entry(asoc, &sp->ep->asocs, asocs)
2943 sctp_apply_asoc_delayed_ack(¶ms, asoc);
2948 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2950 * Applications can specify protocol parameters for the default association
2951 * initialization. The option name argument to setsockopt() and getsockopt()
2954 * Setting initialization parameters is effective only on an unconnected
2955 * socket (for UDP-style sockets only future associations are effected
2956 * by the change). With TCP-style sockets, this option is inherited by
2957 * sockets derived from a listener socket.
2959 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2961 struct sctp_initmsg sinit;
2962 struct sctp_sock *sp = sctp_sk(sk);
2964 if (optlen != sizeof(struct sctp_initmsg))
2966 if (copy_from_user(&sinit, optval, optlen))
2969 if (sinit.sinit_num_ostreams)
2970 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2971 if (sinit.sinit_max_instreams)
2972 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2973 if (sinit.sinit_max_attempts)
2974 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2975 if (sinit.sinit_max_init_timeo)
2976 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2982 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2984 * Applications that wish to use the sendto() system call may wish to
2985 * specify a default set of parameters that would normally be supplied
2986 * through the inclusion of ancillary data. This socket option allows
2987 * such an application to set the default sctp_sndrcvinfo structure.
2988 * The application that wishes to use this socket option simply passes
2989 * in to this call the sctp_sndrcvinfo structure defined in Section
2990 * 5.2.2) The input parameters accepted by this call include
2991 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2992 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2993 * to this call if the caller is using the UDP model.
2995 static int sctp_setsockopt_default_send_param(struct sock *sk,
2996 char __user *optval,
2997 unsigned int optlen)
2999 struct sctp_sock *sp = sctp_sk(sk);
3000 struct sctp_association *asoc;
3001 struct sctp_sndrcvinfo info;
3003 if (optlen != sizeof(info))
3005 if (copy_from_user(&info, optval, optlen))
3007 if (info.sinfo_flags &
3008 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
3009 SCTP_ABORT | SCTP_EOF))
3012 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
3013 if (!asoc && info.sinfo_assoc_id > SCTP_ALL_ASSOC &&
3014 sctp_style(sk, UDP))
3018 asoc->default_stream = info.sinfo_stream;
3019 asoc->default_flags = info.sinfo_flags;
3020 asoc->default_ppid = info.sinfo_ppid;
3021 asoc->default_context = info.sinfo_context;
3022 asoc->default_timetolive = info.sinfo_timetolive;
3027 if (info.sinfo_assoc_id == SCTP_FUTURE_ASSOC ||
3028 info.sinfo_assoc_id == SCTP_ALL_ASSOC) {
3029 sp->default_stream = info.sinfo_stream;
3030 sp->default_flags = info.sinfo_flags;
3031 sp->default_ppid = info.sinfo_ppid;
3032 sp->default_context = info.sinfo_context;
3033 sp->default_timetolive = info.sinfo_timetolive;
3036 if (info.sinfo_assoc_id == SCTP_CURRENT_ASSOC ||
3037 info.sinfo_assoc_id == SCTP_ALL_ASSOC) {
3038 list_for_each_entry(asoc, &sp->ep->asocs, asocs) {
3039 asoc->default_stream = info.sinfo_stream;
3040 asoc->default_flags = info.sinfo_flags;
3041 asoc->default_ppid = info.sinfo_ppid;
3042 asoc->default_context = info.sinfo_context;
3043 asoc->default_timetolive = info.sinfo_timetolive;
3050 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
3051 * (SCTP_DEFAULT_SNDINFO)
3053 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
3054 char __user *optval,
3055 unsigned int optlen)
3057 struct sctp_sock *sp = sctp_sk(sk);
3058 struct sctp_association *asoc;
3059 struct sctp_sndinfo info;
3061 if (optlen != sizeof(info))
3063 if (copy_from_user(&info, optval, optlen))
3065 if (info.snd_flags &
3066 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
3067 SCTP_ABORT | SCTP_EOF))
3070 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
3071 if (!asoc && info.snd_assoc_id > SCTP_ALL_ASSOC &&
3072 sctp_style(sk, UDP))
3076 asoc->default_stream = info.snd_sid;
3077 asoc->default_flags = info.snd_flags;
3078 asoc->default_ppid = info.snd_ppid;
3079 asoc->default_context = info.snd_context;
3084 if (info.snd_assoc_id == SCTP_FUTURE_ASSOC ||
3085 info.snd_assoc_id == SCTP_ALL_ASSOC) {
3086 sp->default_stream = info.snd_sid;
3087 sp->default_flags = info.snd_flags;
3088 sp->default_ppid = info.snd_ppid;
3089 sp->default_context = info.snd_context;
3092 if (info.snd_assoc_id == SCTP_CURRENT_ASSOC ||
3093 info.snd_assoc_id == SCTP_ALL_ASSOC) {
3094 list_for_each_entry(asoc, &sp->ep->asocs, asocs) {
3095 asoc->default_stream = info.snd_sid;
3096 asoc->default_flags = info.snd_flags;
3097 asoc->default_ppid = info.snd_ppid;
3098 asoc->default_context = info.snd_context;
3105 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
3107 * Requests that the local SCTP stack use the enclosed peer address as
3108 * the association primary. The enclosed address must be one of the
3109 * association peer's addresses.
3111 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
3112 unsigned int optlen)
3114 struct sctp_prim prim;
3115 struct sctp_transport *trans;
3119 if (optlen != sizeof(struct sctp_prim))
3122 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
3125 /* Allow security module to validate address but need address len. */
3126 af = sctp_get_af_specific(prim.ssp_addr.ss_family);
3130 err = security_sctp_bind_connect(sk, SCTP_PRIMARY_ADDR,
3131 (struct sockaddr *)&prim.ssp_addr,
3136 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
3140 sctp_assoc_set_primary(trans->asoc, trans);
3146 * 7.1.5 SCTP_NODELAY
3148 * Turn on/off any Nagle-like algorithm. This means that packets are
3149 * generally sent as soon as possible and no unnecessary delays are
3150 * introduced, at the cost of more packets in the network. Expects an
3151 * integer boolean flag.
3153 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
3154 unsigned int optlen)
3158 if (optlen < sizeof(int))
3160 if (get_user(val, (int __user *)optval))
3163 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
3169 * 7.1.1 SCTP_RTOINFO
3171 * The protocol parameters used to initialize and bound retransmission
3172 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
3173 * and modify these parameters.
3174 * All parameters are time values, in milliseconds. A value of 0, when
3175 * modifying the parameters, indicates that the current value should not
3179 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
3181 struct sctp_rtoinfo rtoinfo;
3182 struct sctp_association *asoc;
3183 unsigned long rto_min, rto_max;
3184 struct sctp_sock *sp = sctp_sk(sk);
3186 if (optlen != sizeof (struct sctp_rtoinfo))
3189 if (copy_from_user(&rtoinfo, optval, optlen))
3192 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
3194 /* Set the values to the specific association */
3195 if (!asoc && rtoinfo.srto_assoc_id != SCTP_FUTURE_ASSOC &&
3196 sctp_style(sk, UDP))
3199 rto_max = rtoinfo.srto_max;
3200 rto_min = rtoinfo.srto_min;
3203 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
3205 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
3208 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
3210 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
3212 if (rto_min > rto_max)
3216 if (rtoinfo.srto_initial != 0)
3218 msecs_to_jiffies(rtoinfo.srto_initial);
3219 asoc->rto_max = rto_max;
3220 asoc->rto_min = rto_min;
3222 /* If there is no association or the association-id = 0
3223 * set the values to the endpoint.
3225 if (rtoinfo.srto_initial != 0)
3226 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
3227 sp->rtoinfo.srto_max = rto_max;
3228 sp->rtoinfo.srto_min = rto_min;
3236 * 7.1.2 SCTP_ASSOCINFO
3238 * This option is used to tune the maximum retransmission attempts
3239 * of the association.
3240 * Returns an error if the new association retransmission value is
3241 * greater than the sum of the retransmission value of the peer.
3242 * See [SCTP] for more information.
3245 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
3248 struct sctp_assocparams assocparams;
3249 struct sctp_association *asoc;
3251 if (optlen != sizeof(struct sctp_assocparams))
3253 if (copy_from_user(&assocparams, optval, optlen))
3256 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
3258 if (!asoc && assocparams.sasoc_assoc_id != SCTP_FUTURE_ASSOC &&
3259 sctp_style(sk, UDP))
3262 /* Set the values to the specific association */
3264 if (assocparams.sasoc_asocmaxrxt != 0) {
3267 struct sctp_transport *peer_addr;
3269 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
3271 path_sum += peer_addr->pathmaxrxt;
3275 /* Only validate asocmaxrxt if we have more than
3276 * one path/transport. We do this because path
3277 * retransmissions are only counted when we have more
3281 assocparams.sasoc_asocmaxrxt > path_sum)
3284 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3287 if (assocparams.sasoc_cookie_life != 0)
3288 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3290 /* Set the values to the endpoint */
3291 struct sctp_sock *sp = sctp_sk(sk);
3293 if (assocparams.sasoc_asocmaxrxt != 0)
3294 sp->assocparams.sasoc_asocmaxrxt =
3295 assocparams.sasoc_asocmaxrxt;
3296 if (assocparams.sasoc_cookie_life != 0)
3297 sp->assocparams.sasoc_cookie_life =
3298 assocparams.sasoc_cookie_life;
3304 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3306 * This socket option is a boolean flag which turns on or off mapped V4
3307 * addresses. If this option is turned on and the socket is type
3308 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3309 * If this option is turned off, then no mapping will be done of V4
3310 * addresses and a user will receive both PF_INET6 and PF_INET type
3311 * addresses on the socket.
3313 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3316 struct sctp_sock *sp = sctp_sk(sk);
3318 if (optlen < sizeof(int))
3320 if (get_user(val, (int __user *)optval))
3331 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3332 * This option will get or set the maximum size to put in any outgoing
3333 * SCTP DATA chunk. If a message is larger than this size it will be
3334 * fragmented by SCTP into the specified size. Note that the underlying
3335 * SCTP implementation may fragment into smaller sized chunks when the
3336 * PMTU of the underlying association is smaller than the value set by
3337 * the user. The default value for this option is '0' which indicates
3338 * the user is NOT limiting fragmentation and only the PMTU will effect
3339 * SCTP's choice of DATA chunk size. Note also that values set larger
3340 * than the maximum size of an IP datagram will effectively let SCTP
3341 * control fragmentation (i.e. the same as setting this option to 0).
3343 * The following structure is used to access and modify this parameter:
3345 * struct sctp_assoc_value {
3346 * sctp_assoc_t assoc_id;
3347 * uint32_t assoc_value;
3350 * assoc_id: This parameter is ignored for one-to-one style sockets.
3351 * For one-to-many style sockets this parameter indicates which
3352 * association the user is performing an action upon. Note that if
3353 * this field's value is zero then the endpoints default value is
3354 * changed (effecting future associations only).
3355 * assoc_value: This parameter specifies the maximum size in bytes.
3357 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3359 struct sctp_sock *sp = sctp_sk(sk);
3360 struct sctp_assoc_value params;
3361 struct sctp_association *asoc;
3364 if (optlen == sizeof(int)) {
3365 pr_warn_ratelimited(DEPRECATED
3367 "Use of int in maxseg socket option.\n"
3368 "Use struct sctp_assoc_value instead\n",
3369 current->comm, task_pid_nr(current));
3370 if (copy_from_user(&val, optval, optlen))
3372 params.assoc_id = SCTP_FUTURE_ASSOC;
3373 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3374 if (copy_from_user(¶ms, optval, optlen))
3376 val = params.assoc_value;
3381 asoc = sctp_id2assoc(sk, params.assoc_id);
3382 if (!asoc && params.assoc_id != SCTP_FUTURE_ASSOC &&
3383 sctp_style(sk, UDP))
3387 int min_len, max_len;
3388 __u16 datasize = asoc ? sctp_datachk_len(&asoc->stream) :
3389 sizeof(struct sctp_data_chunk);
3391 min_len = sctp_min_frag_point(sp, datasize);
3392 max_len = SCTP_MAX_CHUNK_LEN - datasize;
3394 if (val < min_len || val > max_len)
3399 asoc->user_frag = val;
3400 sctp_assoc_update_frag_point(asoc);
3402 sp->user_frag = val;
3410 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3412 * Requests that the peer mark the enclosed address as the association
3413 * primary. The enclosed address must be one of the association's
3414 * locally bound addresses. The following structure is used to make a
3415 * set primary request:
3417 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3418 unsigned int optlen)
3420 struct net *net = sock_net(sk);
3421 struct sctp_sock *sp;
3422 struct sctp_association *asoc = NULL;
3423 struct sctp_setpeerprim prim;
3424 struct sctp_chunk *chunk;
3430 if (!net->sctp.addip_enable)
3433 if (optlen != sizeof(struct sctp_setpeerprim))
3436 if (copy_from_user(&prim, optval, optlen))
3439 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3443 if (!asoc->peer.asconf_capable)
3446 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3449 if (!sctp_state(asoc, ESTABLISHED))
3452 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3456 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3457 return -EADDRNOTAVAIL;
3459 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3460 return -EADDRNOTAVAIL;
3462 /* Allow security module to validate address. */
3463 err = security_sctp_bind_connect(sk, SCTP_SET_PEER_PRIMARY_ADDR,
3464 (struct sockaddr *)&prim.sspp_addr,
3469 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3470 chunk = sctp_make_asconf_set_prim(asoc,
3471 (union sctp_addr *)&prim.sspp_addr);
3475 err = sctp_send_asconf(asoc, chunk);
3477 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3482 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3483 unsigned int optlen)
3485 struct sctp_setadaptation adaptation;
3487 if (optlen != sizeof(struct sctp_setadaptation))
3489 if (copy_from_user(&adaptation, optval, optlen))
3492 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3498 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3500 * The context field in the sctp_sndrcvinfo structure is normally only
3501 * used when a failed message is retrieved holding the value that was
3502 * sent down on the actual send call. This option allows the setting of
3503 * a default context on an association basis that will be received on
3504 * reading messages from the peer. This is especially helpful in the
3505 * one-2-many model for an application to keep some reference to an
3506 * internal state machine that is processing messages on the
3507 * association. Note that the setting of this value only effects
3508 * received messages from the peer and does not effect the value that is
3509 * saved with outbound messages.
3511 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3512 unsigned int optlen)
3514 struct sctp_sock *sp = sctp_sk(sk);
3515 struct sctp_assoc_value params;
3516 struct sctp_association *asoc;
3518 if (optlen != sizeof(struct sctp_assoc_value))
3520 if (copy_from_user(¶ms, optval, optlen))
3523 asoc = sctp_id2assoc(sk, params.assoc_id);
3524 if (!asoc && params.assoc_id > SCTP_ALL_ASSOC &&
3525 sctp_style(sk, UDP))
3529 asoc->default_rcv_context = params.assoc_value;
3534 if (params.assoc_id == SCTP_FUTURE_ASSOC ||
3535 params.assoc_id == SCTP_ALL_ASSOC)
3536 sp->default_rcv_context = params.assoc_value;
3538 if (params.assoc_id == SCTP_CURRENT_ASSOC ||
3539 params.assoc_id == SCTP_ALL_ASSOC)
3540 list_for_each_entry(asoc, &sp->ep->asocs, asocs)
3541 asoc->default_rcv_context = params.assoc_value;
3547 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3549 * This options will at a minimum specify if the implementation is doing
3550 * fragmented interleave. Fragmented interleave, for a one to many
3551 * socket, is when subsequent calls to receive a message may return
3552 * parts of messages from different associations. Some implementations
3553 * may allow you to turn this value on or off. If so, when turned off,
3554 * no fragment interleave will occur (which will cause a head of line
3555 * blocking amongst multiple associations sharing the same one to many
3556 * socket). When this option is turned on, then each receive call may
3557 * come from a different association (thus the user must receive data
3558 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3559 * association each receive belongs to.
3561 * This option takes a boolean value. A non-zero value indicates that
3562 * fragmented interleave is on. A value of zero indicates that
3563 * fragmented interleave is off.
3565 * Note that it is important that an implementation that allows this
3566 * option to be turned on, have it off by default. Otherwise an unaware
3567 * application using the one to many model may become confused and act
3570 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3571 char __user *optval,
3572 unsigned int optlen)
3576 if (optlen != sizeof(int))
3578 if (get_user(val, (int __user *)optval))
3581 sctp_sk(sk)->frag_interleave = !!val;
3583 if (!sctp_sk(sk)->frag_interleave)
3584 sctp_sk(sk)->strm_interleave = 0;
3590 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3591 * (SCTP_PARTIAL_DELIVERY_POINT)
3593 * This option will set or get the SCTP partial delivery point. This
3594 * point is the size of a message where the partial delivery API will be
3595 * invoked to help free up rwnd space for the peer. Setting this to a
3596 * lower value will cause partial deliveries to happen more often. The
3597 * calls argument is an integer that sets or gets the partial delivery
3598 * point. Note also that the call will fail if the user attempts to set
3599 * this value larger than the socket receive buffer size.
3601 * Note that any single message having a length smaller than or equal to
3602 * the SCTP partial delivery point will be delivered in one single read
3603 * call as long as the user provided buffer is large enough to hold the
3606 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3607 char __user *optval,
3608 unsigned int optlen)
3612 if (optlen != sizeof(u32))
3614 if (get_user(val, (int __user *)optval))
3617 /* Note: We double the receive buffer from what the user sets
3618 * it to be, also initial rwnd is based on rcvbuf/2.
3620 if (val > (sk->sk_rcvbuf >> 1))
3623 sctp_sk(sk)->pd_point = val;
3625 return 0; /* is this the right error code? */
3629 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3631 * This option will allow a user to change the maximum burst of packets
3632 * that can be emitted by this association. Note that the default value
3633 * is 4, and some implementations may restrict this setting so that it
3634 * can only be lowered.
3636 * NOTE: This text doesn't seem right. Do this on a socket basis with
3637 * future associations inheriting the socket value.
3639 static int sctp_setsockopt_maxburst(struct sock *sk,
3640 char __user *optval,
3641 unsigned int optlen)
3643 struct sctp_sock *sp = sctp_sk(sk);
3644 struct sctp_assoc_value params;
3645 struct sctp_association *asoc;
3647 if (optlen == sizeof(int)) {
3648 pr_warn_ratelimited(DEPRECATED
3650 "Use of int in max_burst socket option deprecated.\n"
3651 "Use struct sctp_assoc_value instead\n",
3652 current->comm, task_pid_nr(current));
3653 if (copy_from_user(¶ms.assoc_value, optval, optlen))
3655 params.assoc_id = SCTP_FUTURE_ASSOC;
3656 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3657 if (copy_from_user(¶ms, optval, optlen))
3662 asoc = sctp_id2assoc(sk, params.assoc_id);
3663 if (!asoc && params.assoc_id > SCTP_ALL_ASSOC &&
3664 sctp_style(sk, UDP))
3668 asoc->max_burst = params.assoc_value;
3673 if (params.assoc_id == SCTP_FUTURE_ASSOC ||
3674 params.assoc_id == SCTP_ALL_ASSOC)
3675 sp->max_burst = params.assoc_value;
3677 if (params.assoc_id == SCTP_CURRENT_ASSOC ||
3678 params.assoc_id == SCTP_ALL_ASSOC)
3679 list_for_each_entry(asoc, &sp->ep->asocs, asocs)
3680 asoc->max_burst = params.assoc_value;
3686 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3688 * This set option adds a chunk type that the user is requesting to be
3689 * received only in an authenticated way. Changes to the list of chunks
3690 * will only effect future associations on the socket.
3692 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3693 char __user *optval,
3694 unsigned int optlen)
3696 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3697 struct sctp_authchunk val;
3699 if (!ep->auth_enable)
3702 if (optlen != sizeof(struct sctp_authchunk))
3704 if (copy_from_user(&val, optval, optlen))
3707 switch (val.sauth_chunk) {
3709 case SCTP_CID_INIT_ACK:
3710 case SCTP_CID_SHUTDOWN_COMPLETE:
3715 /* add this chunk id to the endpoint */
3716 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3720 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3722 * This option gets or sets the list of HMAC algorithms that the local
3723 * endpoint requires the peer to use.
3725 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3726 char __user *optval,
3727 unsigned int optlen)
3729 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3730 struct sctp_hmacalgo *hmacs;
3734 if (!ep->auth_enable)
3737 if (optlen < sizeof(struct sctp_hmacalgo))
3739 optlen = min_t(unsigned int, optlen, sizeof(struct sctp_hmacalgo) +
3740 SCTP_AUTH_NUM_HMACS * sizeof(u16));
3742 hmacs = memdup_user(optval, optlen);
3744 return PTR_ERR(hmacs);
3746 idents = hmacs->shmac_num_idents;
3747 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3748 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3753 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3760 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3762 * This option will set a shared secret key which is used to build an
3763 * association shared key.
3765 static int sctp_setsockopt_auth_key(struct sock *sk,
3766 char __user *optval,
3767 unsigned int optlen)
3769 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3770 struct sctp_authkey *authkey;
3771 struct sctp_association *asoc;
3774 if (!ep->auth_enable)
3777 if (optlen <= sizeof(struct sctp_authkey))
3779 /* authkey->sca_keylength is u16, so optlen can't be bigger than
3782 optlen = min_t(unsigned int, optlen, USHRT_MAX + sizeof(*authkey));
3784 authkey = memdup_user(optval, optlen);
3785 if (IS_ERR(authkey))
3786 return PTR_ERR(authkey);
3788 if (authkey->sca_keylength > optlen - sizeof(*authkey))
3791 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3792 if (!asoc && authkey->sca_assoc_id > SCTP_ALL_ASSOC &&
3793 sctp_style(sk, UDP))
3797 ret = sctp_auth_set_key(ep, asoc, authkey);
3801 if (authkey->sca_assoc_id == SCTP_FUTURE_ASSOC ||
3802 authkey->sca_assoc_id == SCTP_ALL_ASSOC) {
3803 ret = sctp_auth_set_key(ep, asoc, authkey);
3810 if (authkey->sca_assoc_id == SCTP_CURRENT_ASSOC ||
3811 authkey->sca_assoc_id == SCTP_ALL_ASSOC) {
3812 list_for_each_entry(asoc, &ep->asocs, asocs) {
3813 int res = sctp_auth_set_key(ep, asoc, authkey);
3826 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3828 * This option will get or set the active shared key to be used to build
3829 * the association shared key.
3831 static int sctp_setsockopt_active_key(struct sock *sk,
3832 char __user *optval,
3833 unsigned int optlen)
3835 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3836 struct sctp_association *asoc;
3837 struct sctp_authkeyid val;
3840 if (!ep->auth_enable)
3843 if (optlen != sizeof(struct sctp_authkeyid))
3845 if (copy_from_user(&val, optval, optlen))
3848 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3849 if (!asoc && val.scact_assoc_id > SCTP_ALL_ASSOC &&
3850 sctp_style(sk, UDP))
3854 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3856 if (val.scact_assoc_id == SCTP_FUTURE_ASSOC ||
3857 val.scact_assoc_id == SCTP_ALL_ASSOC) {
3858 ret = sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3863 if (val.scact_assoc_id == SCTP_CURRENT_ASSOC ||
3864 val.scact_assoc_id == SCTP_ALL_ASSOC) {
3865 list_for_each_entry(asoc, &ep->asocs, asocs) {
3866 int res = sctp_auth_set_active_key(ep, asoc,
3867 val.scact_keynumber);
3878 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3880 * This set option will delete a shared secret key from use.
3882 static int sctp_setsockopt_del_key(struct sock *sk,
3883 char __user *optval,
3884 unsigned int optlen)
3886 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3887 struct sctp_association *asoc;
3888 struct sctp_authkeyid val;
3891 if (!ep->auth_enable)
3894 if (optlen != sizeof(struct sctp_authkeyid))
3896 if (copy_from_user(&val, optval, optlen))
3899 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3900 if (!asoc && val.scact_assoc_id > SCTP_ALL_ASSOC &&
3901 sctp_style(sk, UDP))
3905 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3907 if (val.scact_assoc_id == SCTP_FUTURE_ASSOC ||
3908 val.scact_assoc_id == SCTP_ALL_ASSOC) {
3909 ret = sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3914 if (val.scact_assoc_id == SCTP_CURRENT_ASSOC ||
3915 val.scact_assoc_id == SCTP_ALL_ASSOC) {
3916 list_for_each_entry(asoc, &ep->asocs, asocs) {
3917 int res = sctp_auth_del_key_id(ep, asoc,
3918 val.scact_keynumber);
3929 * 8.3.4 Deactivate a Shared Key (SCTP_AUTH_DEACTIVATE_KEY)
3931 * This set option will deactivate a shared secret key.
3933 static int sctp_setsockopt_deactivate_key(struct sock *sk, char __user *optval,
3934 unsigned int optlen)
3936 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3937 struct sctp_association *asoc;
3938 struct sctp_authkeyid val;
3941 if (!ep->auth_enable)
3944 if (optlen != sizeof(struct sctp_authkeyid))
3946 if (copy_from_user(&val, optval, optlen))
3949 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3950 if (!asoc && val.scact_assoc_id > SCTP_ALL_ASSOC &&
3951 sctp_style(sk, UDP))
3955 return sctp_auth_deact_key_id(ep, asoc, val.scact_keynumber);
3957 if (val.scact_assoc_id == SCTP_FUTURE_ASSOC ||
3958 val.scact_assoc_id == SCTP_ALL_ASSOC) {
3959 ret = sctp_auth_deact_key_id(ep, asoc, val.scact_keynumber);
3964 if (val.scact_assoc_id == SCTP_CURRENT_ASSOC ||
3965 val.scact_assoc_id == SCTP_ALL_ASSOC) {
3966 list_for_each_entry(asoc, &ep->asocs, asocs) {
3967 int res = sctp_auth_deact_key_id(ep, asoc,
3968 val.scact_keynumber);
3979 * 8.1.23 SCTP_AUTO_ASCONF
3981 * This option will enable or disable the use of the automatic generation of
3982 * ASCONF chunks to add and delete addresses to an existing association. Note
3983 * that this option has two caveats namely: a) it only affects sockets that
3984 * are bound to all addresses available to the SCTP stack, and b) the system
3985 * administrator may have an overriding control that turns the ASCONF feature
3986 * off no matter what setting the socket option may have.
3987 * This option expects an integer boolean flag, where a non-zero value turns on
3988 * the option, and a zero value turns off the option.
3989 * Note. In this implementation, socket operation overrides default parameter
3990 * being set by sysctl as well as FreeBSD implementation
3992 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3993 unsigned int optlen)
3996 struct sctp_sock *sp = sctp_sk(sk);
3998 if (optlen < sizeof(int))
4000 if (get_user(val, (int __user *)optval))
4002 if (!sctp_is_ep_boundall(sk) && val)
4004 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
4007 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
4008 if (val == 0 && sp->do_auto_asconf) {
4009 list_del(&sp->auto_asconf_list);
4010 sp->do_auto_asconf = 0;
4011 } else if (val && !sp->do_auto_asconf) {
4012 list_add_tail(&sp->auto_asconf_list,
4013 &sock_net(sk)->sctp.auto_asconf_splist);
4014 sp->do_auto_asconf = 1;
4016 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
4021 * SCTP_PEER_ADDR_THLDS
4023 * This option allows us to alter the partially failed threshold for one or all
4024 * transports in an association. See Section 6.1 of:
4025 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
4027 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
4028 char __user *optval,
4029 unsigned int optlen)
4031 struct sctp_paddrthlds val;
4032 struct sctp_transport *trans;
4033 struct sctp_association *asoc;
4035 if (optlen < sizeof(struct sctp_paddrthlds))
4037 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
4038 sizeof(struct sctp_paddrthlds)))
4041 if (!sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
4042 trans = sctp_addr_id2transport(sk, &val.spt_address,
4047 if (val.spt_pathmaxrxt)
4048 trans->pathmaxrxt = val.spt_pathmaxrxt;
4049 trans->pf_retrans = val.spt_pathpfthld;
4054 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
4055 if (!asoc && val.spt_assoc_id != SCTP_FUTURE_ASSOC &&
4056 sctp_style(sk, UDP))
4060 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
4062 if (val.spt_pathmaxrxt)
4063 trans->pathmaxrxt = val.spt_pathmaxrxt;
4064 trans->pf_retrans = val.spt_pathpfthld;
4067 if (val.spt_pathmaxrxt)
4068 asoc->pathmaxrxt = val.spt_pathmaxrxt;
4069 asoc->pf_retrans = val.spt_pathpfthld;
4071 struct sctp_sock *sp = sctp_sk(sk);
4073 if (val.spt_pathmaxrxt)
4074 sp->pathmaxrxt = val.spt_pathmaxrxt;
4075 sp->pf_retrans = val.spt_pathpfthld;
4081 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
4082 char __user *optval,
4083 unsigned int optlen)
4087 if (optlen < sizeof(int))
4089 if (get_user(val, (int __user *) optval))
4092 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
4097 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
4098 char __user *optval,
4099 unsigned int optlen)
4103 if (optlen < sizeof(int))
4105 if (get_user(val, (int __user *) optval))
4108 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
4113 static int sctp_setsockopt_pr_supported(struct sock *sk,
4114 char __user *optval,
4115 unsigned int optlen)
4117 struct sctp_assoc_value params;
4118 struct sctp_association *asoc;
4120 if (optlen != sizeof(params))
4123 if (copy_from_user(¶ms, optval, optlen))
4126 asoc = sctp_id2assoc(sk, params.assoc_id);
4127 if (!asoc && params.assoc_id != SCTP_FUTURE_ASSOC &&
4128 sctp_style(sk, UDP))
4131 sctp_sk(sk)->ep->prsctp_enable = !!params.assoc_value;
4136 static int sctp_setsockopt_default_prinfo(struct sock *sk,
4137 char __user *optval,
4138 unsigned int optlen)
4140 struct sctp_sock *sp = sctp_sk(sk);
4141 struct sctp_default_prinfo info;
4142 struct sctp_association *asoc;
4143 int retval = -EINVAL;
4145 if (optlen != sizeof(info))
4148 if (copy_from_user(&info, optval, sizeof(info))) {
4153 if (info.pr_policy & ~SCTP_PR_SCTP_MASK)
4156 if (info.pr_policy == SCTP_PR_SCTP_NONE)
4159 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
4160 if (!asoc && info.pr_assoc_id > SCTP_ALL_ASSOC &&
4161 sctp_style(sk, UDP))
4167 SCTP_PR_SET_POLICY(asoc->default_flags, info.pr_policy);
4168 asoc->default_timetolive = info.pr_value;
4172 if (info.pr_assoc_id == SCTP_FUTURE_ASSOC ||
4173 info.pr_assoc_id == SCTP_ALL_ASSOC) {
4174 SCTP_PR_SET_POLICY(sp->default_flags, info.pr_policy);
4175 sp->default_timetolive = info.pr_value;
4178 if (info.pr_assoc_id == SCTP_CURRENT_ASSOC ||
4179 info.pr_assoc_id == SCTP_ALL_ASSOC) {
4180 list_for_each_entry(asoc, &sp->ep->asocs, asocs) {
4181 SCTP_PR_SET_POLICY(asoc->default_flags, info.pr_policy);
4182 asoc->default_timetolive = info.pr_value;
4190 static int sctp_setsockopt_reconfig_supported(struct sock *sk,
4191 char __user *optval,
4192 unsigned int optlen)
4194 struct sctp_assoc_value params;
4195 struct sctp_association *asoc;
4196 int retval = -EINVAL;
4198 if (optlen != sizeof(params))
4201 if (copy_from_user(¶ms, optval, optlen)) {
4206 asoc = sctp_id2assoc(sk, params.assoc_id);
4207 if (!asoc && params.assoc_id != SCTP_FUTURE_ASSOC &&
4208 sctp_style(sk, UDP))
4212 asoc->reconf_enable = !!params.assoc_value;
4214 sctp_sk(sk)->ep->reconf_enable = !!params.assoc_value;
4222 static int sctp_setsockopt_enable_strreset(struct sock *sk,
4223 char __user *optval,
4224 unsigned int optlen)
4226 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
4227 struct sctp_assoc_value params;
4228 struct sctp_association *asoc;
4229 int retval = -EINVAL;
4231 if (optlen != sizeof(params))
4234 if (copy_from_user(¶ms, optval, optlen)) {
4239 if (params.assoc_value & (~SCTP_ENABLE_STRRESET_MASK))
4242 asoc = sctp_id2assoc(sk, params.assoc_id);
4243 if (!asoc && params.assoc_id > SCTP_ALL_ASSOC &&
4244 sctp_style(sk, UDP))
4250 asoc->strreset_enable = params.assoc_value;
4254 if (params.assoc_id == SCTP_FUTURE_ASSOC ||
4255 params.assoc_id == SCTP_ALL_ASSOC)
4256 ep->strreset_enable = params.assoc_value;
4258 if (params.assoc_id == SCTP_CURRENT_ASSOC ||
4259 params.assoc_id == SCTP_ALL_ASSOC)
4260 list_for_each_entry(asoc, &ep->asocs, asocs)
4261 asoc->strreset_enable = params.assoc_value;
4267 static int sctp_setsockopt_reset_streams(struct sock *sk,
4268 char __user *optval,
4269 unsigned int optlen)
4271 struct sctp_reset_streams *params;
4272 struct sctp_association *asoc;
4273 int retval = -EINVAL;
4275 if (optlen < sizeof(*params))
4277 /* srs_number_streams is u16, so optlen can't be bigger than this. */
4278 optlen = min_t(unsigned int, optlen, USHRT_MAX +
4279 sizeof(__u16) * sizeof(*params));
4281 params = memdup_user(optval, optlen);
4283 return PTR_ERR(params);
4285 if (params->srs_number_streams * sizeof(__u16) >
4286 optlen - sizeof(*params))
4289 asoc = sctp_id2assoc(sk, params->srs_assoc_id);
4293 retval = sctp_send_reset_streams(asoc, params);
4300 static int sctp_setsockopt_reset_assoc(struct sock *sk,
4301 char __user *optval,
4302 unsigned int optlen)
4304 struct sctp_association *asoc;
4305 sctp_assoc_t associd;
4306 int retval = -EINVAL;
4308 if (optlen != sizeof(associd))
4311 if (copy_from_user(&associd, optval, optlen)) {
4316 asoc = sctp_id2assoc(sk, associd);
4320 retval = sctp_send_reset_assoc(asoc);
4326 static int sctp_setsockopt_add_streams(struct sock *sk,
4327 char __user *optval,
4328 unsigned int optlen)
4330 struct sctp_association *asoc;
4331 struct sctp_add_streams params;
4332 int retval = -EINVAL;
4334 if (optlen != sizeof(params))
4337 if (copy_from_user(¶ms, optval, optlen)) {
4342 asoc = sctp_id2assoc(sk, params.sas_assoc_id);
4346 retval = sctp_send_add_streams(asoc, ¶ms);
4352 static int sctp_setsockopt_scheduler(struct sock *sk,
4353 char __user *optval,
4354 unsigned int optlen)
4356 struct sctp_sock *sp = sctp_sk(sk);
4357 struct sctp_association *asoc;
4358 struct sctp_assoc_value params;
4361 if (optlen < sizeof(params))
4364 optlen = sizeof(params);
4365 if (copy_from_user(¶ms, optval, optlen))
4368 if (params.assoc_value > SCTP_SS_MAX)
4371 asoc = sctp_id2assoc(sk, params.assoc_id);
4372 if (!asoc && params.assoc_id > SCTP_ALL_ASSOC &&
4373 sctp_style(sk, UDP))
4377 return sctp_sched_set_sched(asoc, params.assoc_value);
4379 if (params.assoc_id == SCTP_FUTURE_ASSOC ||
4380 params.assoc_id == SCTP_ALL_ASSOC)
4381 sp->default_ss = params.assoc_value;
4383 if (params.assoc_id == SCTP_CURRENT_ASSOC ||
4384 params.assoc_id == SCTP_ALL_ASSOC) {
4385 list_for_each_entry(asoc, &sp->ep->asocs, asocs) {
4386 int ret = sctp_sched_set_sched(asoc,
4387 params.assoc_value);
4397 static int sctp_setsockopt_scheduler_value(struct sock *sk,
4398 char __user *optval,
4399 unsigned int optlen)
4401 struct sctp_stream_value params;
4402 struct sctp_association *asoc;
4403 int retval = -EINVAL;
4405 if (optlen < sizeof(params))
4408 optlen = sizeof(params);
4409 if (copy_from_user(¶ms, optval, optlen)) {
4414 asoc = sctp_id2assoc(sk, params.assoc_id);
4415 if (!asoc && params.assoc_id != SCTP_CURRENT_ASSOC &&
4416 sctp_style(sk, UDP))
4420 retval = sctp_sched_set_value(asoc, params.stream_id,
4421 params.stream_value, GFP_KERNEL);
4427 list_for_each_entry(asoc, &sctp_sk(sk)->ep->asocs, asocs) {
4428 int ret = sctp_sched_set_value(asoc, params.stream_id,
4429 params.stream_value, GFP_KERNEL);
4430 if (ret && !retval) /* try to return the 1st error. */
4438 static int sctp_setsockopt_interleaving_supported(struct sock *sk,
4439 char __user *optval,
4440 unsigned int optlen)
4442 struct sctp_sock *sp = sctp_sk(sk);
4443 struct sctp_assoc_value params;
4444 struct sctp_association *asoc;
4445 int retval = -EINVAL;
4447 if (optlen < sizeof(params))
4450 optlen = sizeof(params);
4451 if (copy_from_user(¶ms, optval, optlen)) {
4456 asoc = sctp_id2assoc(sk, params.assoc_id);
4457 if (!asoc && params.assoc_id != SCTP_FUTURE_ASSOC &&
4458 sctp_style(sk, UDP))
4461 if (!sock_net(sk)->sctp.intl_enable || !sp->frag_interleave) {
4466 sp->strm_interleave = !!params.assoc_value;
4474 static int sctp_setsockopt_reuse_port(struct sock *sk, char __user *optval,
4475 unsigned int optlen)
4479 if (!sctp_style(sk, TCP))
4482 if (sctp_sk(sk)->ep->base.bind_addr.port)
4485 if (optlen < sizeof(int))
4488 if (get_user(val, (int __user *)optval))
4491 sctp_sk(sk)->reuse = !!val;
4496 static int sctp_assoc_ulpevent_type_set(struct sctp_event *param,
4497 struct sctp_association *asoc)
4499 struct sctp_ulpevent *event;
4501 sctp_ulpevent_type_set(&asoc->subscribe, param->se_type, param->se_on);
4503 if (param->se_type == SCTP_SENDER_DRY_EVENT && param->se_on) {
4504 if (sctp_outq_is_empty(&asoc->outqueue)) {
4505 event = sctp_ulpevent_make_sender_dry_event(asoc,
4506 GFP_USER | __GFP_NOWARN);
4510 asoc->stream.si->enqueue_event(&asoc->ulpq, event);
4517 static int sctp_setsockopt_event(struct sock *sk, char __user *optval,
4518 unsigned int optlen)
4520 struct sctp_sock *sp = sctp_sk(sk);
4521 struct sctp_association *asoc;
4522 struct sctp_event param;
4525 if (optlen < sizeof(param))
4528 optlen = sizeof(param);
4529 if (copy_from_user(¶m, optval, optlen))
4532 if (param.se_type < SCTP_SN_TYPE_BASE ||
4533 param.se_type > SCTP_SN_TYPE_MAX)
4536 asoc = sctp_id2assoc(sk, param.se_assoc_id);
4537 if (!asoc && param.se_assoc_id > SCTP_ALL_ASSOC &&
4538 sctp_style(sk, UDP))
4542 return sctp_assoc_ulpevent_type_set(¶m, asoc);
4544 if (param.se_assoc_id == SCTP_FUTURE_ASSOC ||
4545 param.se_assoc_id == SCTP_ALL_ASSOC)
4546 sctp_ulpevent_type_set(&sp->subscribe,
4547 param.se_type, param.se_on);
4549 if (param.se_assoc_id == SCTP_CURRENT_ASSOC ||
4550 param.se_assoc_id == SCTP_ALL_ASSOC) {
4551 list_for_each_entry(asoc, &sp->ep->asocs, asocs) {
4552 int ret = sctp_assoc_ulpevent_type_set(¶m, asoc);
4562 /* API 6.2 setsockopt(), getsockopt()
4564 * Applications use setsockopt() and getsockopt() to set or retrieve
4565 * socket options. Socket options are used to change the default
4566 * behavior of sockets calls. They are described in Section 7.
4570 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
4571 * int __user *optlen);
4572 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
4575 * sd - the socket descript.
4576 * level - set to IPPROTO_SCTP for all SCTP options.
4577 * optname - the option name.
4578 * optval - the buffer to store the value of the option.
4579 * optlen - the size of the buffer.
4581 static int sctp_setsockopt(struct sock *sk, int level, int optname,
4582 char __user *optval, unsigned int optlen)
4586 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
4588 /* I can hardly begin to describe how wrong this is. This is
4589 * so broken as to be worse than useless. The API draft
4590 * REALLY is NOT helpful here... I am not convinced that the
4591 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
4592 * are at all well-founded.
4594 if (level != SOL_SCTP) {
4595 struct sctp_af *af = sctp_sk(sk)->pf->af;
4596 retval = af->setsockopt(sk, level, optname, optval, optlen);
4603 case SCTP_SOCKOPT_BINDX_ADD:
4604 /* 'optlen' is the size of the addresses buffer. */
4605 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
4606 optlen, SCTP_BINDX_ADD_ADDR);
4609 case SCTP_SOCKOPT_BINDX_REM:
4610 /* 'optlen' is the size of the addresses buffer. */
4611 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
4612 optlen, SCTP_BINDX_REM_ADDR);
4615 case SCTP_SOCKOPT_CONNECTX_OLD:
4616 /* 'optlen' is the size of the addresses buffer. */
4617 retval = sctp_setsockopt_connectx_old(sk,
4618 (struct sockaddr __user *)optval,
4622 case SCTP_SOCKOPT_CONNECTX:
4623 /* 'optlen' is the size of the addresses buffer. */
4624 retval = sctp_setsockopt_connectx(sk,
4625 (struct sockaddr __user *)optval,
4629 case SCTP_DISABLE_FRAGMENTS:
4630 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
4634 retval = sctp_setsockopt_events(sk, optval, optlen);
4637 case SCTP_AUTOCLOSE:
4638 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
4641 case SCTP_PEER_ADDR_PARAMS:
4642 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
4645 case SCTP_DELAYED_SACK:
4646 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
4648 case SCTP_PARTIAL_DELIVERY_POINT:
4649 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
4653 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
4655 case SCTP_DEFAULT_SEND_PARAM:
4656 retval = sctp_setsockopt_default_send_param(sk, optval,
4659 case SCTP_DEFAULT_SNDINFO:
4660 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
4662 case SCTP_PRIMARY_ADDR:
4663 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
4665 case SCTP_SET_PEER_PRIMARY_ADDR:
4666 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
4669 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
4672 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
4674 case SCTP_ASSOCINFO:
4675 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
4677 case SCTP_I_WANT_MAPPED_V4_ADDR:
4678 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
4681 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
4683 case SCTP_ADAPTATION_LAYER:
4684 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
4687 retval = sctp_setsockopt_context(sk, optval, optlen);
4689 case SCTP_FRAGMENT_INTERLEAVE:
4690 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
4692 case SCTP_MAX_BURST:
4693 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
4695 case SCTP_AUTH_CHUNK:
4696 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
4698 case SCTP_HMAC_IDENT:
4699 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
4702 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
4704 case SCTP_AUTH_ACTIVE_KEY:
4705 retval = sctp_setsockopt_active_key(sk, optval, optlen);
4707 case SCTP_AUTH_DELETE_KEY:
4708 retval = sctp_setsockopt_del_key(sk, optval, optlen);
4710 case SCTP_AUTH_DEACTIVATE_KEY:
4711 retval = sctp_setsockopt_deactivate_key(sk, optval, optlen);
4713 case SCTP_AUTO_ASCONF:
4714 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
4716 case SCTP_PEER_ADDR_THLDS:
4717 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
4719 case SCTP_RECVRCVINFO:
4720 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
4722 case SCTP_RECVNXTINFO:
4723 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
4725 case SCTP_PR_SUPPORTED:
4726 retval = sctp_setsockopt_pr_supported(sk, optval, optlen);
4728 case SCTP_DEFAULT_PRINFO:
4729 retval = sctp_setsockopt_default_prinfo(sk, optval, optlen);
4731 case SCTP_RECONFIG_SUPPORTED:
4732 retval = sctp_setsockopt_reconfig_supported(sk, optval, optlen);
4734 case SCTP_ENABLE_STREAM_RESET:
4735 retval = sctp_setsockopt_enable_strreset(sk, optval, optlen);
4737 case SCTP_RESET_STREAMS:
4738 retval = sctp_setsockopt_reset_streams(sk, optval, optlen);
4740 case SCTP_RESET_ASSOC:
4741 retval = sctp_setsockopt_reset_assoc(sk, optval, optlen);
4743 case SCTP_ADD_STREAMS:
4744 retval = sctp_setsockopt_add_streams(sk, optval, optlen);
4746 case SCTP_STREAM_SCHEDULER:
4747 retval = sctp_setsockopt_scheduler(sk, optval, optlen);
4749 case SCTP_STREAM_SCHEDULER_VALUE:
4750 retval = sctp_setsockopt_scheduler_value(sk, optval, optlen);
4752 case SCTP_INTERLEAVING_SUPPORTED:
4753 retval = sctp_setsockopt_interleaving_supported(sk, optval,
4756 case SCTP_REUSE_PORT:
4757 retval = sctp_setsockopt_reuse_port(sk, optval, optlen);
4760 retval = sctp_setsockopt_event(sk, optval, optlen);
4763 retval = -ENOPROTOOPT;
4773 /* API 3.1.6 connect() - UDP Style Syntax
4775 * An application may use the connect() call in the UDP model to initiate an
4776 * association without sending data.
4780 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
4782 * sd: the socket descriptor to have a new association added to.
4784 * nam: the address structure (either struct sockaddr_in or struct
4785 * sockaddr_in6 defined in RFC2553 [7]).
4787 * len: the size of the address.
4789 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
4790 int addr_len, int flags)
4792 struct inet_sock *inet = inet_sk(sk);
4798 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
4801 /* We may need to bind the socket. */
4802 if (!inet->inet_num) {
4803 if (sk->sk_prot->get_port(sk, 0)) {
4807 inet->inet_sport = htons(inet->inet_num);
4810 /* Validate addr_len before calling common connect/connectx routine. */
4811 af = sctp_get_af_specific(addr->sa_family);
4812 if (!af || addr_len < af->sockaddr_len) {
4815 /* Pass correct addr len to common routine (so it knows there
4816 * is only one address being passed.
4818 err = __sctp_connect(sk, addr, af->sockaddr_len, flags, NULL);
4825 int sctp_inet_connect(struct socket *sock, struct sockaddr *uaddr,
4826 int addr_len, int flags)
4828 if (addr_len < sizeof(uaddr->sa_family))
4831 if (uaddr->sa_family == AF_UNSPEC)
4834 return sctp_connect(sock->sk, uaddr, addr_len, flags);
4837 /* FIXME: Write comments. */
4838 static int sctp_disconnect(struct sock *sk, int flags)
4840 return -EOPNOTSUPP; /* STUB */
4843 /* 4.1.4 accept() - TCP Style Syntax
4845 * Applications use accept() call to remove an established SCTP
4846 * association from the accept queue of the endpoint. A new socket
4847 * descriptor will be returned from accept() to represent the newly
4848 * formed association.
4850 static struct sock *sctp_accept(struct sock *sk, int flags, int *err, bool kern)
4852 struct sctp_sock *sp;
4853 struct sctp_endpoint *ep;
4854 struct sock *newsk = NULL;
4855 struct sctp_association *asoc;
4864 if (!sctp_style(sk, TCP)) {
4865 error = -EOPNOTSUPP;
4869 if (!sctp_sstate(sk, LISTENING)) {
4874 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
4876 error = sctp_wait_for_accept(sk, timeo);
4880 /* We treat the list of associations on the endpoint as the accept
4881 * queue and pick the first association on the list.
4883 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
4885 newsk = sp->pf->create_accept_sk(sk, asoc, kern);
4891 /* Populate the fields of the newsk from the oldsk and migrate the
4892 * asoc to the newsk.
4894 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
4902 /* The SCTP ioctl handler. */
4903 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
4910 * SEQPACKET-style sockets in LISTENING state are valid, for
4911 * SCTP, so only discard TCP-style sockets in LISTENING state.
4913 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4918 struct sk_buff *skb;
4919 unsigned int amount = 0;
4921 skb = skb_peek(&sk->sk_receive_queue);
4924 * We will only return the amount of this packet since
4925 * that is all that will be read.
4929 rc = put_user(amount, (int __user *)arg);
4941 /* This is the function which gets called during socket creation to
4942 * initialized the SCTP-specific portion of the sock.
4943 * The sock structure should already be zero-filled memory.
4945 static int sctp_init_sock(struct sock *sk)
4947 struct net *net = sock_net(sk);
4948 struct sctp_sock *sp;
4950 pr_debug("%s: sk:%p\n", __func__, sk);
4954 /* Initialize the SCTP per socket area. */
4955 switch (sk->sk_type) {
4956 case SOCK_SEQPACKET:
4957 sp->type = SCTP_SOCKET_UDP;
4960 sp->type = SCTP_SOCKET_TCP;
4963 return -ESOCKTNOSUPPORT;
4966 sk->sk_gso_type = SKB_GSO_SCTP;
4968 /* Initialize default send parameters. These parameters can be
4969 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4971 sp->default_stream = 0;
4972 sp->default_ppid = 0;
4973 sp->default_flags = 0;
4974 sp->default_context = 0;
4975 sp->default_timetolive = 0;
4977 sp->default_rcv_context = 0;
4978 sp->max_burst = net->sctp.max_burst;
4980 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4982 /* Initialize default setup parameters. These parameters
4983 * can be modified with the SCTP_INITMSG socket option or
4984 * overridden by the SCTP_INIT CMSG.
4986 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4987 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4988 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4989 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4991 /* Initialize default RTO related parameters. These parameters can
4992 * be modified for with the SCTP_RTOINFO socket option.
4994 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4995 sp->rtoinfo.srto_max = net->sctp.rto_max;
4996 sp->rtoinfo.srto_min = net->sctp.rto_min;
4998 /* Initialize default association related parameters. These parameters
4999 * can be modified with the SCTP_ASSOCINFO socket option.
5001 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
5002 sp->assocparams.sasoc_number_peer_destinations = 0;
5003 sp->assocparams.sasoc_peer_rwnd = 0;
5004 sp->assocparams.sasoc_local_rwnd = 0;
5005 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
5007 /* Initialize default event subscriptions. By default, all the
5012 /* Default Peer Address Parameters. These defaults can
5013 * be modified via SCTP_PEER_ADDR_PARAMS
5015 sp->hbinterval = net->sctp.hb_interval;
5016 sp->pathmaxrxt = net->sctp.max_retrans_path;
5017 sp->pf_retrans = net->sctp.pf_retrans;
5018 sp->pathmtu = 0; /* allow default discovery */
5019 sp->sackdelay = net->sctp.sack_timeout;
5021 sp->param_flags = SPP_HB_ENABLE |
5023 SPP_SACKDELAY_ENABLE;
5024 sp->default_ss = SCTP_SS_DEFAULT;
5026 /* If enabled no SCTP message fragmentation will be performed.
5027 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
5029 sp->disable_fragments = 0;
5031 /* Enable Nagle algorithm by default. */
5034 sp->recvrcvinfo = 0;
5035 sp->recvnxtinfo = 0;
5037 /* Enable by default. */
5040 /* Auto-close idle associations after the configured
5041 * number of seconds. A value of 0 disables this
5042 * feature. Configure through the SCTP_AUTOCLOSE socket option,
5043 * for UDP-style sockets only.
5047 /* User specified fragmentation limit. */
5050 sp->adaptation_ind = 0;
5052 sp->pf = sctp_get_pf_specific(sk->sk_family);
5054 /* Control variables for partial data delivery. */
5055 atomic_set(&sp->pd_mode, 0);
5056 skb_queue_head_init(&sp->pd_lobby);
5057 sp->frag_interleave = 0;
5059 /* Create a per socket endpoint structure. Even if we
5060 * change the data structure relationships, this may still
5061 * be useful for storing pre-connect address information.
5063 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
5069 sk->sk_destruct = sctp_destruct_sock;
5071 SCTP_DBG_OBJCNT_INC(sock);
5074 sk_sockets_allocated_inc(sk);
5075 sock_prot_inuse_add(net, sk->sk_prot, 1);
5077 /* Nothing can fail after this block, otherwise
5078 * sctp_destroy_sock() will be called without addr_wq_lock held
5080 if (net->sctp.default_auto_asconf) {
5081 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
5082 list_add_tail(&sp->auto_asconf_list,
5083 &net->sctp.auto_asconf_splist);
5084 sp->do_auto_asconf = 1;
5085 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
5087 sp->do_auto_asconf = 0;
5095 /* Cleanup any SCTP per socket resources. Must be called with
5096 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
5098 static void sctp_destroy_sock(struct sock *sk)
5100 struct sctp_sock *sp;
5102 pr_debug("%s: sk:%p\n", __func__, sk);
5104 /* Release our hold on the endpoint. */
5106 /* This could happen during socket init, thus we bail out
5107 * early, since the rest of the below is not setup either.
5112 if (sp->do_auto_asconf) {
5113 sp->do_auto_asconf = 0;
5114 list_del(&sp->auto_asconf_list);
5116 sctp_endpoint_free(sp->ep);
5118 sk_sockets_allocated_dec(sk);
5119 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
5123 /* Triggered when there are no references on the socket anymore */
5124 static void sctp_destruct_sock(struct sock *sk)
5126 struct sctp_sock *sp = sctp_sk(sk);
5128 /* Free up the HMAC transform. */
5129 crypto_free_shash(sp->hmac);
5131 inet_sock_destruct(sk);
5134 /* API 4.1.7 shutdown() - TCP Style Syntax
5135 * int shutdown(int socket, int how);
5137 * sd - the socket descriptor of the association to be closed.
5138 * how - Specifies the type of shutdown. The values are
5141 * Disables further receive operations. No SCTP
5142 * protocol action is taken.
5144 * Disables further send operations, and initiates
5145 * the SCTP shutdown sequence.
5147 * Disables further send and receive operations
5148 * and initiates the SCTP shutdown sequence.
5150 static void sctp_shutdown(struct sock *sk, int how)
5152 struct net *net = sock_net(sk);
5153 struct sctp_endpoint *ep;
5155 if (!sctp_style(sk, TCP))
5158 ep = sctp_sk(sk)->ep;
5159 if (how & SEND_SHUTDOWN && !list_empty(&ep->asocs)) {
5160 struct sctp_association *asoc;
5162 inet_sk_set_state(sk, SCTP_SS_CLOSING);
5163 asoc = list_entry(ep->asocs.next,
5164 struct sctp_association, asocs);
5165 sctp_primitive_SHUTDOWN(net, asoc, NULL);
5169 int sctp_get_sctp_info(struct sock *sk, struct sctp_association *asoc,
5170 struct sctp_info *info)
5172 struct sctp_transport *prim;
5173 struct list_head *pos;
5176 memset(info, 0, sizeof(*info));
5178 struct sctp_sock *sp = sctp_sk(sk);
5180 info->sctpi_s_autoclose = sp->autoclose;
5181 info->sctpi_s_adaptation_ind = sp->adaptation_ind;
5182 info->sctpi_s_pd_point = sp->pd_point;
5183 info->sctpi_s_nodelay = sp->nodelay;
5184 info->sctpi_s_disable_fragments = sp->disable_fragments;
5185 info->sctpi_s_v4mapped = sp->v4mapped;
5186 info->sctpi_s_frag_interleave = sp->frag_interleave;
5187 info->sctpi_s_type = sp->type;
5192 info->sctpi_tag = asoc->c.my_vtag;
5193 info->sctpi_state = asoc->state;
5194 info->sctpi_rwnd = asoc->a_rwnd;
5195 info->sctpi_unackdata = asoc->unack_data;
5196 info->sctpi_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
5197 info->sctpi_instrms = asoc->stream.incnt;
5198 info->sctpi_outstrms = asoc->stream.outcnt;
5199 list_for_each(pos, &asoc->base.inqueue.in_chunk_list)
5200 info->sctpi_inqueue++;
5201 list_for_each(pos, &asoc->outqueue.out_chunk_list)
5202 info->sctpi_outqueue++;
5203 info->sctpi_overall_error = asoc->overall_error_count;
5204 info->sctpi_max_burst = asoc->max_burst;
5205 info->sctpi_maxseg = asoc->frag_point;
5206 info->sctpi_peer_rwnd = asoc->peer.rwnd;
5207 info->sctpi_peer_tag = asoc->c.peer_vtag;
5209 mask = asoc->peer.ecn_capable << 1;
5210 mask = (mask | asoc->peer.ipv4_address) << 1;
5211 mask = (mask | asoc->peer.ipv6_address) << 1;
5212 mask = (mask | asoc->peer.hostname_address) << 1;
5213 mask = (mask | asoc->peer.asconf_capable) << 1;
5214 mask = (mask | asoc->peer.prsctp_capable) << 1;
5215 mask = (mask | asoc->peer.auth_capable);
5216 info->sctpi_peer_capable = mask;
5217 mask = asoc->peer.sack_needed << 1;
5218 mask = (mask | asoc->peer.sack_generation) << 1;
5219 mask = (mask | asoc->peer.zero_window_announced);
5220 info->sctpi_peer_sack = mask;
5222 info->sctpi_isacks = asoc->stats.isacks;
5223 info->sctpi_osacks = asoc->stats.osacks;
5224 info->sctpi_opackets = asoc->stats.opackets;
5225 info->sctpi_ipackets = asoc->stats.ipackets;
5226 info->sctpi_rtxchunks = asoc->stats.rtxchunks;
5227 info->sctpi_outofseqtsns = asoc->stats.outofseqtsns;
5228 info->sctpi_idupchunks = asoc->stats.idupchunks;
5229 info->sctpi_gapcnt = asoc->stats.gapcnt;
5230 info->sctpi_ouodchunks = asoc->stats.ouodchunks;
5231 info->sctpi_iuodchunks = asoc->stats.iuodchunks;
5232 info->sctpi_oodchunks = asoc->stats.oodchunks;
5233 info->sctpi_iodchunks = asoc->stats.iodchunks;
5234 info->sctpi_octrlchunks = asoc->stats.octrlchunks;
5235 info->sctpi_ictrlchunks = asoc->stats.ictrlchunks;
5237 prim = asoc->peer.primary_path;
5238 memcpy(&info->sctpi_p_address, &prim->ipaddr, sizeof(prim->ipaddr));
5239 info->sctpi_p_state = prim->state;
5240 info->sctpi_p_cwnd = prim->cwnd;
5241 info->sctpi_p_srtt = prim->srtt;
5242 info->sctpi_p_rto = jiffies_to_msecs(prim->rto);
5243 info->sctpi_p_hbinterval = prim->hbinterval;
5244 info->sctpi_p_pathmaxrxt = prim->pathmaxrxt;
5245 info->sctpi_p_sackdelay = jiffies_to_msecs(prim->sackdelay);
5246 info->sctpi_p_ssthresh = prim->ssthresh;
5247 info->sctpi_p_partial_bytes_acked = prim->partial_bytes_acked;
5248 info->sctpi_p_flight_size = prim->flight_size;
5249 info->sctpi_p_error = prim->error_count;
5253 EXPORT_SYMBOL_GPL(sctp_get_sctp_info);
5255 /* use callback to avoid exporting the core structure */
5256 void sctp_transport_walk_start(struct rhashtable_iter *iter)
5258 rhltable_walk_enter(&sctp_transport_hashtable, iter);
5260 rhashtable_walk_start(iter);
5263 void sctp_transport_walk_stop(struct rhashtable_iter *iter)
5265 rhashtable_walk_stop(iter);
5266 rhashtable_walk_exit(iter);
5269 struct sctp_transport *sctp_transport_get_next(struct net *net,
5270 struct rhashtable_iter *iter)
5272 struct sctp_transport *t;
5274 t = rhashtable_walk_next(iter);
5275 for (; t; t = rhashtable_walk_next(iter)) {
5277 if (PTR_ERR(t) == -EAGAIN)
5282 if (!sctp_transport_hold(t))
5285 if (net_eq(sock_net(t->asoc->base.sk), net) &&
5286 t->asoc->peer.primary_path == t)
5289 sctp_transport_put(t);
5295 struct sctp_transport *sctp_transport_get_idx(struct net *net,
5296 struct rhashtable_iter *iter,
5299 struct sctp_transport *t;
5302 return SEQ_START_TOKEN;
5304 while ((t = sctp_transport_get_next(net, iter)) && !IS_ERR(t)) {
5307 sctp_transport_put(t);
5313 int sctp_for_each_endpoint(int (*cb)(struct sctp_endpoint *, void *),
5317 struct sctp_ep_common *epb;
5318 struct sctp_hashbucket *head;
5320 for (head = sctp_ep_hashtable; hash < sctp_ep_hashsize;
5322 read_lock_bh(&head->lock);
5323 sctp_for_each_hentry(epb, &head->chain) {
5324 err = cb(sctp_ep(epb), p);
5328 read_unlock_bh(&head->lock);
5333 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint);
5335 int sctp_transport_lookup_process(int (*cb)(struct sctp_transport *, void *),
5337 const union sctp_addr *laddr,
5338 const union sctp_addr *paddr, void *p)
5340 struct sctp_transport *transport;
5344 transport = sctp_addrs_lookup_transport(net, laddr, paddr);
5349 err = cb(transport, p);
5350 sctp_transport_put(transport);
5354 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process);
5356 int sctp_for_each_transport(int (*cb)(struct sctp_transport *, void *),
5357 int (*cb_done)(struct sctp_transport *, void *),
5358 struct net *net, int *pos, void *p) {
5359 struct rhashtable_iter hti;
5360 struct sctp_transport *tsp;
5365 sctp_transport_walk_start(&hti);
5367 tsp = sctp_transport_get_idx(net, &hti, *pos + 1);
5368 for (; !IS_ERR_OR_NULL(tsp); tsp = sctp_transport_get_next(net, &hti)) {
5373 sctp_transport_put(tsp);
5375 sctp_transport_walk_stop(&hti);
5378 if (cb_done && !cb_done(tsp, p)) {
5380 sctp_transport_put(tsp);
5383 sctp_transport_put(tsp);
5388 EXPORT_SYMBOL_GPL(sctp_for_each_transport);
5390 /* 7.2.1 Association Status (SCTP_STATUS)
5392 * Applications can retrieve current status information about an
5393 * association, including association state, peer receiver window size,
5394 * number of unacked data chunks, and number of data chunks pending
5395 * receipt. This information is read-only.
5397 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
5398 char __user *optval,
5401 struct sctp_status status;
5402 struct sctp_association *asoc = NULL;
5403 struct sctp_transport *transport;
5404 sctp_assoc_t associd;
5407 if (len < sizeof(status)) {
5412 len = sizeof(status);
5413 if (copy_from_user(&status, optval, len)) {
5418 associd = status.sstat_assoc_id;
5419 asoc = sctp_id2assoc(sk, associd);
5425 transport = asoc->peer.primary_path;
5427 status.sstat_assoc_id = sctp_assoc2id(asoc);
5428 status.sstat_state = sctp_assoc_to_state(asoc);
5429 status.sstat_rwnd = asoc->peer.rwnd;
5430 status.sstat_unackdata = asoc->unack_data;
5432 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
5433 status.sstat_instrms = asoc->stream.incnt;
5434 status.sstat_outstrms = asoc->stream.outcnt;
5435 status.sstat_fragmentation_point = asoc->frag_point;
5436 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
5437 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
5438 transport->af_specific->sockaddr_len);
5439 /* Map ipv4 address into v4-mapped-on-v6 address. */
5440 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
5441 (union sctp_addr *)&status.sstat_primary.spinfo_address);
5442 status.sstat_primary.spinfo_state = transport->state;
5443 status.sstat_primary.spinfo_cwnd = transport->cwnd;
5444 status.sstat_primary.spinfo_srtt = transport->srtt;
5445 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
5446 status.sstat_primary.spinfo_mtu = transport->pathmtu;
5448 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
5449 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
5451 if (put_user(len, optlen)) {
5456 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
5457 __func__, len, status.sstat_state, status.sstat_rwnd,
5458 status.sstat_assoc_id);
5460 if (copy_to_user(optval, &status, len)) {
5470 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
5472 * Applications can retrieve information about a specific peer address
5473 * of an association, including its reachability state, congestion
5474 * window, and retransmission timer values. This information is
5477 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
5478 char __user *optval,
5481 struct sctp_paddrinfo pinfo;
5482 struct sctp_transport *transport;
5485 if (len < sizeof(pinfo)) {
5490 len = sizeof(pinfo);
5491 if (copy_from_user(&pinfo, optval, len)) {
5496 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
5497 pinfo.spinfo_assoc_id);
5501 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
5502 pinfo.spinfo_state = transport->state;
5503 pinfo.spinfo_cwnd = transport->cwnd;
5504 pinfo.spinfo_srtt = transport->srtt;
5505 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
5506 pinfo.spinfo_mtu = transport->pathmtu;
5508 if (pinfo.spinfo_state == SCTP_UNKNOWN)
5509 pinfo.spinfo_state = SCTP_ACTIVE;
5511 if (put_user(len, optlen)) {
5516 if (copy_to_user(optval, &pinfo, len)) {
5525 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
5527 * This option is a on/off flag. If enabled no SCTP message
5528 * fragmentation will be performed. Instead if a message being sent
5529 * exceeds the current PMTU size, the message will NOT be sent and
5530 * instead a error will be indicated to the user.
5532 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
5533 char __user *optval, int __user *optlen)
5537 if (len < sizeof(int))
5541 val = (sctp_sk(sk)->disable_fragments == 1);
5542 if (put_user(len, optlen))
5544 if (copy_to_user(optval, &val, len))
5549 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
5551 * This socket option is used to specify various notifications and
5552 * ancillary data the user wishes to receive.
5554 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
5557 struct sctp_event_subscribe subscribe;
5558 __u8 *sn_type = (__u8 *)&subscribe;
5563 if (len > sizeof(struct sctp_event_subscribe))
5564 len = sizeof(struct sctp_event_subscribe);
5565 if (put_user(len, optlen))
5568 for (i = 0; i < len; i++)
5569 sn_type[i] = sctp_ulpevent_type_enabled(sctp_sk(sk)->subscribe,
5570 SCTP_SN_TYPE_BASE + i);
5572 if (copy_to_user(optval, &subscribe, len))
5578 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
5580 * This socket option is applicable to the UDP-style socket only. When
5581 * set it will cause associations that are idle for more than the
5582 * specified number of seconds to automatically close. An association
5583 * being idle is defined an association that has NOT sent or received
5584 * user data. The special value of '0' indicates that no automatic
5585 * close of any associations should be performed. The option expects an
5586 * integer defining the number of seconds of idle time before an
5587 * association is closed.
5589 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
5591 /* Applicable to UDP-style socket only */
5592 if (sctp_style(sk, TCP))
5594 if (len < sizeof(int))
5597 if (put_user(len, optlen))
5599 if (put_user(sctp_sk(sk)->autoclose, (int __user *)optval))
5604 /* Helper routine to branch off an association to a new socket. */
5605 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
5607 struct sctp_association *asoc = sctp_id2assoc(sk, id);
5608 struct sctp_sock *sp = sctp_sk(sk);
5609 struct socket *sock;
5612 /* Do not peel off from one netns to another one. */
5613 if (!net_eq(current->nsproxy->net_ns, sock_net(sk)))
5619 /* An association cannot be branched off from an already peeled-off
5620 * socket, nor is this supported for tcp style sockets.
5622 if (!sctp_style(sk, UDP))
5625 /* Create a new socket. */
5626 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
5630 sctp_copy_sock(sock->sk, sk, asoc);
5632 /* Make peeled-off sockets more like 1-1 accepted sockets.
5633 * Set the daddr and initialize id to something more random and also
5634 * copy over any ip options.
5636 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
5637 sp->pf->copy_ip_options(sk, sock->sk);
5639 /* Populate the fields of the newsk from the oldsk and migrate the
5640 * asoc to the newsk.
5642 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
5648 EXPORT_SYMBOL(sctp_do_peeloff);
5650 static int sctp_getsockopt_peeloff_common(struct sock *sk, sctp_peeloff_arg_t *peeloff,
5651 struct file **newfile, unsigned flags)
5653 struct socket *newsock;
5656 retval = sctp_do_peeloff(sk, peeloff->associd, &newsock);
5660 /* Map the socket to an unused fd that can be returned to the user. */
5661 retval = get_unused_fd_flags(flags & SOCK_CLOEXEC);
5663 sock_release(newsock);
5667 *newfile = sock_alloc_file(newsock, 0, NULL);
5668 if (IS_ERR(*newfile)) {
5669 put_unused_fd(retval);
5670 retval = PTR_ERR(*newfile);
5675 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
5678 peeloff->sd = retval;
5680 if (flags & SOCK_NONBLOCK)
5681 (*newfile)->f_flags |= O_NONBLOCK;
5686 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
5688 sctp_peeloff_arg_t peeloff;
5689 struct file *newfile = NULL;
5692 if (len < sizeof(sctp_peeloff_arg_t))
5694 len = sizeof(sctp_peeloff_arg_t);
5695 if (copy_from_user(&peeloff, optval, len))
5698 retval = sctp_getsockopt_peeloff_common(sk, &peeloff, &newfile, 0);
5702 /* Return the fd mapped to the new socket. */
5703 if (put_user(len, optlen)) {
5705 put_unused_fd(retval);
5709 if (copy_to_user(optval, &peeloff, len)) {
5711 put_unused_fd(retval);
5714 fd_install(retval, newfile);
5719 static int sctp_getsockopt_peeloff_flags(struct sock *sk, int len,
5720 char __user *optval, int __user *optlen)
5722 sctp_peeloff_flags_arg_t peeloff;
5723 struct file *newfile = NULL;
5726 if (len < sizeof(sctp_peeloff_flags_arg_t))
5728 len = sizeof(sctp_peeloff_flags_arg_t);
5729 if (copy_from_user(&peeloff, optval, len))
5732 retval = sctp_getsockopt_peeloff_common(sk, &peeloff.p_arg,
5733 &newfile, peeloff.flags);
5737 /* Return the fd mapped to the new socket. */
5738 if (put_user(len, optlen)) {
5740 put_unused_fd(retval);
5744 if (copy_to_user(optval, &peeloff, len)) {
5746 put_unused_fd(retval);
5749 fd_install(retval, newfile);
5754 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
5756 * Applications can enable or disable heartbeats for any peer address of
5757 * an association, modify an address's heartbeat interval, force a
5758 * heartbeat to be sent immediately, and adjust the address's maximum
5759 * number of retransmissions sent before an address is considered
5760 * unreachable. The following structure is used to access and modify an
5761 * address's parameters:
5763 * struct sctp_paddrparams {
5764 * sctp_assoc_t spp_assoc_id;
5765 * struct sockaddr_storage spp_address;
5766 * uint32_t spp_hbinterval;
5767 * uint16_t spp_pathmaxrxt;
5768 * uint32_t spp_pathmtu;
5769 * uint32_t spp_sackdelay;
5770 * uint32_t spp_flags;
5773 * spp_assoc_id - (one-to-many style socket) This is filled in the
5774 * application, and identifies the association for
5776 * spp_address - This specifies which address is of interest.
5777 * spp_hbinterval - This contains the value of the heartbeat interval,
5778 * in milliseconds. If a value of zero
5779 * is present in this field then no changes are to
5780 * be made to this parameter.
5781 * spp_pathmaxrxt - This contains the maximum number of
5782 * retransmissions before this address shall be
5783 * considered unreachable. If a value of zero
5784 * is present in this field then no changes are to
5785 * be made to this parameter.
5786 * spp_pathmtu - When Path MTU discovery is disabled the value
5787 * specified here will be the "fixed" path mtu.
5788 * Note that if the spp_address field is empty
5789 * then all associations on this address will
5790 * have this fixed path mtu set upon them.
5792 * spp_sackdelay - When delayed sack is enabled, this value specifies
5793 * the number of milliseconds that sacks will be delayed
5794 * for. This value will apply to all addresses of an
5795 * association if the spp_address field is empty. Note
5796 * also, that if delayed sack is enabled and this
5797 * value is set to 0, no change is made to the last
5798 * recorded delayed sack timer value.
5800 * spp_flags - These flags are used to control various features
5801 * on an association. The flag field may contain
5802 * zero or more of the following options.
5804 * SPP_HB_ENABLE - Enable heartbeats on the
5805 * specified address. Note that if the address
5806 * field is empty all addresses for the association
5807 * have heartbeats enabled upon them.
5809 * SPP_HB_DISABLE - Disable heartbeats on the
5810 * speicifed address. Note that if the address
5811 * field is empty all addresses for the association
5812 * will have their heartbeats disabled. Note also
5813 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
5814 * mutually exclusive, only one of these two should
5815 * be specified. Enabling both fields will have
5816 * undetermined results.
5818 * SPP_HB_DEMAND - Request a user initiated heartbeat
5819 * to be made immediately.
5821 * SPP_PMTUD_ENABLE - This field will enable PMTU
5822 * discovery upon the specified address. Note that
5823 * if the address feild is empty then all addresses
5824 * on the association are effected.
5826 * SPP_PMTUD_DISABLE - This field will disable PMTU
5827 * discovery upon the specified address. Note that
5828 * if the address feild is empty then all addresses
5829 * on the association are effected. Not also that
5830 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
5831 * exclusive. Enabling both will have undetermined
5834 * SPP_SACKDELAY_ENABLE - Setting this flag turns
5835 * on delayed sack. The time specified in spp_sackdelay
5836 * is used to specify the sack delay for this address. Note
5837 * that if spp_address is empty then all addresses will
5838 * enable delayed sack and take on the sack delay
5839 * value specified in spp_sackdelay.
5840 * SPP_SACKDELAY_DISABLE - Setting this flag turns
5841 * off delayed sack. If the spp_address field is blank then
5842 * delayed sack is disabled for the entire association. Note
5843 * also that this field is mutually exclusive to
5844 * SPP_SACKDELAY_ENABLE, setting both will have undefined
5847 * SPP_IPV6_FLOWLABEL: Setting this flag enables the
5848 * setting of the IPV6 flow label value. The value is
5849 * contained in the spp_ipv6_flowlabel field.
5850 * Upon retrieval, this flag will be set to indicate that
5851 * the spp_ipv6_flowlabel field has a valid value returned.
5852 * If a specific destination address is set (in the
5853 * spp_address field), then the value returned is that of
5854 * the address. If just an association is specified (and
5855 * no address), then the association's default flow label
5856 * is returned. If neither an association nor a destination
5857 * is specified, then the socket's default flow label is
5858 * returned. For non-IPv6 sockets, this flag will be left
5861 * SPP_DSCP: Setting this flag enables the setting of the
5862 * Differentiated Services Code Point (DSCP) value
5863 * associated with either the association or a specific
5864 * address. The value is obtained in the spp_dscp field.
5865 * Upon retrieval, this flag will be set to indicate that
5866 * the spp_dscp field has a valid value returned. If a
5867 * specific destination address is set when called (in the
5868 * spp_address field), then that specific destination
5869 * address's DSCP value is returned. If just an association
5870 * is specified, then the association's default DSCP is
5871 * returned. If neither an association nor a destination is
5872 * specified, then the socket's default DSCP is returned.
5874 * spp_ipv6_flowlabel
5875 * - This field is used in conjunction with the
5876 * SPP_IPV6_FLOWLABEL flag and contains the IPv6 flow label.
5877 * The 20 least significant bits are used for the flow
5878 * label. This setting has precedence over any IPv6-layer
5881 * spp_dscp - This field is used in conjunction with the SPP_DSCP flag
5882 * and contains the DSCP. The 6 most significant bits are
5883 * used for the DSCP. This setting has precedence over any
5884 * IPv4- or IPv6- layer setting.
5886 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
5887 char __user *optval, int __user *optlen)
5889 struct sctp_paddrparams params;
5890 struct sctp_transport *trans = NULL;
5891 struct sctp_association *asoc = NULL;
5892 struct sctp_sock *sp = sctp_sk(sk);
5894 if (len >= sizeof(params))
5895 len = sizeof(params);
5896 else if (len >= ALIGN(offsetof(struct sctp_paddrparams,
5897 spp_ipv6_flowlabel), 4))
5898 len = ALIGN(offsetof(struct sctp_paddrparams,
5899 spp_ipv6_flowlabel), 4);
5903 if (copy_from_user(¶ms, optval, len))
5906 /* If an address other than INADDR_ANY is specified, and
5907 * no transport is found, then the request is invalid.
5909 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
5910 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
5911 params.spp_assoc_id);
5913 pr_debug("%s: failed no transport\n", __func__);
5918 /* Get association, if assoc_id != SCTP_FUTURE_ASSOC and the
5919 * socket is a one to many style socket, and an association
5920 * was not found, then the id was invalid.
5922 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
5923 if (!asoc && params.spp_assoc_id != SCTP_FUTURE_ASSOC &&
5924 sctp_style(sk, UDP)) {
5925 pr_debug("%s: failed no association\n", __func__);
5930 /* Fetch transport values. */
5931 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
5932 params.spp_pathmtu = trans->pathmtu;
5933 params.spp_pathmaxrxt = trans->pathmaxrxt;
5934 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
5936 /*draft-11 doesn't say what to return in spp_flags*/
5937 params.spp_flags = trans->param_flags;
5938 if (trans->flowlabel & SCTP_FLOWLABEL_SET_MASK) {
5939 params.spp_ipv6_flowlabel = trans->flowlabel &
5940 SCTP_FLOWLABEL_VAL_MASK;
5941 params.spp_flags |= SPP_IPV6_FLOWLABEL;
5943 if (trans->dscp & SCTP_DSCP_SET_MASK) {
5944 params.spp_dscp = trans->dscp & SCTP_DSCP_VAL_MASK;
5945 params.spp_flags |= SPP_DSCP;
5948 /* Fetch association values. */
5949 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
5950 params.spp_pathmtu = asoc->pathmtu;
5951 params.spp_pathmaxrxt = asoc->pathmaxrxt;
5952 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
5954 /*draft-11 doesn't say what to return in spp_flags*/
5955 params.spp_flags = asoc->param_flags;
5956 if (asoc->flowlabel & SCTP_FLOWLABEL_SET_MASK) {
5957 params.spp_ipv6_flowlabel = asoc->flowlabel &
5958 SCTP_FLOWLABEL_VAL_MASK;
5959 params.spp_flags |= SPP_IPV6_FLOWLABEL;
5961 if (asoc->dscp & SCTP_DSCP_SET_MASK) {
5962 params.spp_dscp = asoc->dscp & SCTP_DSCP_VAL_MASK;
5963 params.spp_flags |= SPP_DSCP;
5966 /* Fetch socket values. */
5967 params.spp_hbinterval = sp->hbinterval;
5968 params.spp_pathmtu = sp->pathmtu;
5969 params.spp_sackdelay = sp->sackdelay;
5970 params.spp_pathmaxrxt = sp->pathmaxrxt;
5972 /*draft-11 doesn't say what to return in spp_flags*/
5973 params.spp_flags = sp->param_flags;
5974 if (sp->flowlabel & SCTP_FLOWLABEL_SET_MASK) {
5975 params.spp_ipv6_flowlabel = sp->flowlabel &
5976 SCTP_FLOWLABEL_VAL_MASK;
5977 params.spp_flags |= SPP_IPV6_FLOWLABEL;
5979 if (sp->dscp & SCTP_DSCP_SET_MASK) {
5980 params.spp_dscp = sp->dscp & SCTP_DSCP_VAL_MASK;
5981 params.spp_flags |= SPP_DSCP;
5985 if (copy_to_user(optval, ¶ms, len))
5988 if (put_user(len, optlen))
5995 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
5997 * This option will effect the way delayed acks are performed. This
5998 * option allows you to get or set the delayed ack time, in
5999 * milliseconds. It also allows changing the delayed ack frequency.
6000 * Changing the frequency to 1 disables the delayed sack algorithm. If
6001 * the assoc_id is 0, then this sets or gets the endpoints default
6002 * values. If the assoc_id field is non-zero, then the set or get
6003 * effects the specified association for the one to many model (the
6004 * assoc_id field is ignored by the one to one model). Note that if
6005 * sack_delay or sack_freq are 0 when setting this option, then the
6006 * current values will remain unchanged.
6008 * struct sctp_sack_info {
6009 * sctp_assoc_t sack_assoc_id;
6010 * uint32_t sack_delay;
6011 * uint32_t sack_freq;
6014 * sack_assoc_id - This parameter, indicates which association the user
6015 * is performing an action upon. Note that if this field's value is
6016 * zero then the endpoints default value is changed (effecting future
6017 * associations only).
6019 * sack_delay - This parameter contains the number of milliseconds that
6020 * the user is requesting the delayed ACK timer be set to. Note that
6021 * this value is defined in the standard to be between 200 and 500
6024 * sack_freq - This parameter contains the number of packets that must
6025 * be received before a sack is sent without waiting for the delay
6026 * timer to expire. The default value for this is 2, setting this
6027 * value to 1 will disable the delayed sack algorithm.
6029 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
6030 char __user *optval,
6033 struct sctp_sack_info params;
6034 struct sctp_association *asoc = NULL;
6035 struct sctp_sock *sp = sctp_sk(sk);
6037 if (len >= sizeof(struct sctp_sack_info)) {
6038 len = sizeof(struct sctp_sack_info);
6040 if (copy_from_user(¶ms, optval, len))
6042 } else if (len == sizeof(struct sctp_assoc_value)) {
6043 pr_warn_ratelimited(DEPRECATED
6045 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
6046 "Use struct sctp_sack_info instead\n",
6047 current->comm, task_pid_nr(current));
6048 if (copy_from_user(¶ms, optval, len))
6053 /* Get association, if sack_assoc_id != SCTP_FUTURE_ASSOC and the
6054 * socket is a one to many style socket, and an association
6055 * was not found, then the id was invalid.
6057 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
6058 if (!asoc && params.sack_assoc_id != SCTP_FUTURE_ASSOC &&
6059 sctp_style(sk, UDP))
6063 /* Fetch association values. */
6064 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
6065 params.sack_delay = jiffies_to_msecs(asoc->sackdelay);
6066 params.sack_freq = asoc->sackfreq;
6069 params.sack_delay = 0;
6070 params.sack_freq = 1;
6073 /* Fetch socket values. */
6074 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
6075 params.sack_delay = sp->sackdelay;
6076 params.sack_freq = sp->sackfreq;
6078 params.sack_delay = 0;
6079 params.sack_freq = 1;
6083 if (copy_to_user(optval, ¶ms, len))
6086 if (put_user(len, optlen))
6092 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
6094 * Applications can specify protocol parameters for the default association
6095 * initialization. The option name argument to setsockopt() and getsockopt()
6098 * Setting initialization parameters is effective only on an unconnected
6099 * socket (for UDP-style sockets only future associations are effected
6100 * by the change). With TCP-style sockets, this option is inherited by
6101 * sockets derived from a listener socket.
6103 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
6105 if (len < sizeof(struct sctp_initmsg))
6107 len = sizeof(struct sctp_initmsg);
6108 if (put_user(len, optlen))
6110 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
6116 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
6117 char __user *optval, int __user *optlen)
6119 struct sctp_association *asoc;
6121 struct sctp_getaddrs getaddrs;
6122 struct sctp_transport *from;
6124 union sctp_addr temp;
6125 struct sctp_sock *sp = sctp_sk(sk);
6130 if (len < sizeof(struct sctp_getaddrs))
6133 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
6136 /* For UDP-style sockets, id specifies the association to query. */
6137 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
6141 to = optval + offsetof(struct sctp_getaddrs, addrs);
6142 space_left = len - offsetof(struct sctp_getaddrs, addrs);
6144 list_for_each_entry(from, &asoc->peer.transport_addr_list,
6146 memcpy(&temp, &from->ipaddr, sizeof(temp));
6147 addrlen = sctp_get_pf_specific(sk->sk_family)
6148 ->addr_to_user(sp, &temp);
6149 if (space_left < addrlen)
6151 if (copy_to_user(to, &temp, addrlen))
6155 space_left -= addrlen;
6158 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
6160 bytes_copied = ((char __user *)to) - optval;
6161 if (put_user(bytes_copied, optlen))
6167 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
6168 size_t space_left, int *bytes_copied)
6170 struct sctp_sockaddr_entry *addr;
6171 union sctp_addr temp;
6174 struct net *net = sock_net(sk);
6177 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
6181 if ((PF_INET == sk->sk_family) &&
6182 (AF_INET6 == addr->a.sa.sa_family))
6184 if ((PF_INET6 == sk->sk_family) &&
6185 inet_v6_ipv6only(sk) &&
6186 (AF_INET == addr->a.sa.sa_family))
6188 memcpy(&temp, &addr->a, sizeof(temp));
6189 if (!temp.v4.sin_port)
6190 temp.v4.sin_port = htons(port);
6192 addrlen = sctp_get_pf_specific(sk->sk_family)
6193 ->addr_to_user(sctp_sk(sk), &temp);
6195 if (space_left < addrlen) {
6199 memcpy(to, &temp, addrlen);
6203 space_left -= addrlen;
6204 *bytes_copied += addrlen;
6212 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
6213 char __user *optval, int __user *optlen)
6215 struct sctp_bind_addr *bp;
6216 struct sctp_association *asoc;
6218 struct sctp_getaddrs getaddrs;
6219 struct sctp_sockaddr_entry *addr;
6221 union sctp_addr temp;
6222 struct sctp_sock *sp = sctp_sk(sk);
6226 int bytes_copied = 0;
6230 if (len < sizeof(struct sctp_getaddrs))
6233 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
6237 * For UDP-style sockets, id specifies the association to query.
6238 * If the id field is set to the value '0' then the locally bound
6239 * addresses are returned without regard to any particular
6242 if (0 == getaddrs.assoc_id) {
6243 bp = &sctp_sk(sk)->ep->base.bind_addr;
6245 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
6248 bp = &asoc->base.bind_addr;
6251 to = optval + offsetof(struct sctp_getaddrs, addrs);
6252 space_left = len - offsetof(struct sctp_getaddrs, addrs);
6254 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN);
6258 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
6259 * addresses from the global local address list.
6261 if (sctp_list_single_entry(&bp->address_list)) {
6262 addr = list_entry(bp->address_list.next,
6263 struct sctp_sockaddr_entry, list);
6264 if (sctp_is_any(sk, &addr->a)) {
6265 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
6266 space_left, &bytes_copied);
6276 /* Protection on the bound address list is not needed since
6277 * in the socket option context we hold a socket lock and
6278 * thus the bound address list can't change.
6280 list_for_each_entry(addr, &bp->address_list, list) {
6281 memcpy(&temp, &addr->a, sizeof(temp));
6282 addrlen = sctp_get_pf_specific(sk->sk_family)
6283 ->addr_to_user(sp, &temp);
6284 if (space_left < addrlen) {
6285 err = -ENOMEM; /*fixme: right error?*/
6288 memcpy(buf, &temp, addrlen);
6290 bytes_copied += addrlen;
6292 space_left -= addrlen;
6296 if (copy_to_user(to, addrs, bytes_copied)) {
6300 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
6304 /* XXX: We should have accounted for sizeof(struct sctp_getaddrs) too,
6305 * but we can't change it anymore.
6307 if (put_user(bytes_copied, optlen))
6314 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
6316 * Requests that the local SCTP stack use the enclosed peer address as
6317 * the association primary. The enclosed address must be one of the
6318 * association peer's addresses.
6320 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
6321 char __user *optval, int __user *optlen)
6323 struct sctp_prim prim;
6324 struct sctp_association *asoc;
6325 struct sctp_sock *sp = sctp_sk(sk);
6327 if (len < sizeof(struct sctp_prim))
6330 len = sizeof(struct sctp_prim);
6332 if (copy_from_user(&prim, optval, len))
6335 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
6339 if (!asoc->peer.primary_path)
6342 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
6343 asoc->peer.primary_path->af_specific->sockaddr_len);
6345 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
6346 (union sctp_addr *)&prim.ssp_addr);
6348 if (put_user(len, optlen))
6350 if (copy_to_user(optval, &prim, len))
6357 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
6359 * Requests that the local endpoint set the specified Adaptation Layer
6360 * Indication parameter for all future INIT and INIT-ACK exchanges.
6362 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
6363 char __user *optval, int __user *optlen)
6365 struct sctp_setadaptation adaptation;
6367 if (len < sizeof(struct sctp_setadaptation))
6370 len = sizeof(struct sctp_setadaptation);
6372 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
6374 if (put_user(len, optlen))
6376 if (copy_to_user(optval, &adaptation, len))
6384 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
6386 * Applications that wish to use the sendto() system call may wish to
6387 * specify a default set of parameters that would normally be supplied
6388 * through the inclusion of ancillary data. This socket option allows
6389 * such an application to set the default sctp_sndrcvinfo structure.
6392 * The application that wishes to use this socket option simply passes
6393 * in to this call the sctp_sndrcvinfo structure defined in Section
6394 * 5.2.2) The input parameters accepted by this call include
6395 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
6396 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
6397 * to this call if the caller is using the UDP model.
6399 * For getsockopt, it get the default sctp_sndrcvinfo structure.
6401 static int sctp_getsockopt_default_send_param(struct sock *sk,
6402 int len, char __user *optval,
6405 struct sctp_sock *sp = sctp_sk(sk);
6406 struct sctp_association *asoc;
6407 struct sctp_sndrcvinfo info;
6409 if (len < sizeof(info))
6414 if (copy_from_user(&info, optval, len))
6417 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
6418 if (!asoc && info.sinfo_assoc_id != SCTP_FUTURE_ASSOC &&
6419 sctp_style(sk, UDP))
6423 info.sinfo_stream = asoc->default_stream;
6424 info.sinfo_flags = asoc->default_flags;
6425 info.sinfo_ppid = asoc->default_ppid;
6426 info.sinfo_context = asoc->default_context;
6427 info.sinfo_timetolive = asoc->default_timetolive;
6429 info.sinfo_stream = sp->default_stream;
6430 info.sinfo_flags = sp->default_flags;
6431 info.sinfo_ppid = sp->default_ppid;
6432 info.sinfo_context = sp->default_context;
6433 info.sinfo_timetolive = sp->default_timetolive;
6436 if (put_user(len, optlen))
6438 if (copy_to_user(optval, &info, len))
6444 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
6445 * (SCTP_DEFAULT_SNDINFO)
6447 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
6448 char __user *optval,
6451 struct sctp_sock *sp = sctp_sk(sk);
6452 struct sctp_association *asoc;
6453 struct sctp_sndinfo info;
6455 if (len < sizeof(info))
6460 if (copy_from_user(&info, optval, len))
6463 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
6464 if (!asoc && info.snd_assoc_id != SCTP_FUTURE_ASSOC &&
6465 sctp_style(sk, UDP))
6469 info.snd_sid = asoc->default_stream;
6470 info.snd_flags = asoc->default_flags;
6471 info.snd_ppid = asoc->default_ppid;
6472 info.snd_context = asoc->default_context;
6474 info.snd_sid = sp->default_stream;
6475 info.snd_flags = sp->default_flags;
6476 info.snd_ppid = sp->default_ppid;
6477 info.snd_context = sp->default_context;
6480 if (put_user(len, optlen))
6482 if (copy_to_user(optval, &info, len))
6490 * 7.1.5 SCTP_NODELAY
6492 * Turn on/off any Nagle-like algorithm. This means that packets are
6493 * generally sent as soon as possible and no unnecessary delays are
6494 * introduced, at the cost of more packets in the network. Expects an
6495 * integer boolean flag.
6498 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
6499 char __user *optval, int __user *optlen)
6503 if (len < sizeof(int))
6507 val = (sctp_sk(sk)->nodelay == 1);
6508 if (put_user(len, optlen))
6510 if (copy_to_user(optval, &val, len))
6517 * 7.1.1 SCTP_RTOINFO
6519 * The protocol parameters used to initialize and bound retransmission
6520 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
6521 * and modify these parameters.
6522 * All parameters are time values, in milliseconds. A value of 0, when
6523 * modifying the parameters, indicates that the current value should not
6527 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
6528 char __user *optval,
6529 int __user *optlen) {
6530 struct sctp_rtoinfo rtoinfo;
6531 struct sctp_association *asoc;
6533 if (len < sizeof (struct sctp_rtoinfo))
6536 len = sizeof(struct sctp_rtoinfo);
6538 if (copy_from_user(&rtoinfo, optval, len))
6541 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
6543 if (!asoc && rtoinfo.srto_assoc_id != SCTP_FUTURE_ASSOC &&
6544 sctp_style(sk, UDP))
6547 /* Values corresponding to the specific association. */
6549 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
6550 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
6551 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
6553 /* Values corresponding to the endpoint. */
6554 struct sctp_sock *sp = sctp_sk(sk);
6556 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
6557 rtoinfo.srto_max = sp->rtoinfo.srto_max;
6558 rtoinfo.srto_min = sp->rtoinfo.srto_min;
6561 if (put_user(len, optlen))
6564 if (copy_to_user(optval, &rtoinfo, len))
6572 * 7.1.2 SCTP_ASSOCINFO
6574 * This option is used to tune the maximum retransmission attempts
6575 * of the association.
6576 * Returns an error if the new association retransmission value is
6577 * greater than the sum of the retransmission value of the peer.
6578 * See [SCTP] for more information.
6581 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
6582 char __user *optval,
6586 struct sctp_assocparams assocparams;
6587 struct sctp_association *asoc;
6588 struct list_head *pos;
6591 if (len < sizeof (struct sctp_assocparams))
6594 len = sizeof(struct sctp_assocparams);
6596 if (copy_from_user(&assocparams, optval, len))
6599 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
6601 if (!asoc && assocparams.sasoc_assoc_id != SCTP_FUTURE_ASSOC &&
6602 sctp_style(sk, UDP))
6605 /* Values correspoinding to the specific association */
6607 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
6608 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
6609 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
6610 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
6612 list_for_each(pos, &asoc->peer.transport_addr_list) {
6616 assocparams.sasoc_number_peer_destinations = cnt;
6618 /* Values corresponding to the endpoint */
6619 struct sctp_sock *sp = sctp_sk(sk);
6621 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
6622 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
6623 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
6624 assocparams.sasoc_cookie_life =
6625 sp->assocparams.sasoc_cookie_life;
6626 assocparams.sasoc_number_peer_destinations =
6628 sasoc_number_peer_destinations;
6631 if (put_user(len, optlen))
6634 if (copy_to_user(optval, &assocparams, len))
6641 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
6643 * This socket option is a boolean flag which turns on or off mapped V4
6644 * addresses. If this option is turned on and the socket is type
6645 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
6646 * If this option is turned off, then no mapping will be done of V4
6647 * addresses and a user will receive both PF_INET6 and PF_INET type
6648 * addresses on the socket.
6650 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
6651 char __user *optval, int __user *optlen)
6654 struct sctp_sock *sp = sctp_sk(sk);
6656 if (len < sizeof(int))
6661 if (put_user(len, optlen))
6663 if (copy_to_user(optval, &val, len))
6670 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
6671 * (chapter and verse is quoted at sctp_setsockopt_context())
6673 static int sctp_getsockopt_context(struct sock *sk, int len,
6674 char __user *optval, int __user *optlen)
6676 struct sctp_assoc_value params;
6677 struct sctp_association *asoc;
6679 if (len < sizeof(struct sctp_assoc_value))
6682 len = sizeof(struct sctp_assoc_value);
6684 if (copy_from_user(¶ms, optval, len))
6687 asoc = sctp_id2assoc(sk, params.assoc_id);
6688 if (!asoc && params.assoc_id != SCTP_FUTURE_ASSOC &&
6689 sctp_style(sk, UDP))
6692 params.assoc_value = asoc ? asoc->default_rcv_context
6693 : sctp_sk(sk)->default_rcv_context;
6695 if (put_user(len, optlen))
6697 if (copy_to_user(optval, ¶ms, len))
6704 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
6705 * This option will get or set the maximum size to put in any outgoing
6706 * SCTP DATA chunk. If a message is larger than this size it will be
6707 * fragmented by SCTP into the specified size. Note that the underlying
6708 * SCTP implementation may fragment into smaller sized chunks when the
6709 * PMTU of the underlying association is smaller than the value set by
6710 * the user. The default value for this option is '0' which indicates
6711 * the user is NOT limiting fragmentation and only the PMTU will effect
6712 * SCTP's choice of DATA chunk size. Note also that values set larger
6713 * than the maximum size of an IP datagram will effectively let SCTP
6714 * control fragmentation (i.e. the same as setting this option to 0).
6716 * The following structure is used to access and modify this parameter:
6718 * struct sctp_assoc_value {
6719 * sctp_assoc_t assoc_id;
6720 * uint32_t assoc_value;
6723 * assoc_id: This parameter is ignored for one-to-one style sockets.
6724 * For one-to-many style sockets this parameter indicates which
6725 * association the user is performing an action upon. Note that if
6726 * this field's value is zero then the endpoints default value is
6727 * changed (effecting future associations only).
6728 * assoc_value: This parameter specifies the maximum size in bytes.
6730 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
6731 char __user *optval, int __user *optlen)
6733 struct sctp_assoc_value params;
6734 struct sctp_association *asoc;
6736 if (len == sizeof(int)) {
6737 pr_warn_ratelimited(DEPRECATED
6739 "Use of int in maxseg socket option.\n"
6740 "Use struct sctp_assoc_value instead\n",
6741 current->comm, task_pid_nr(current));
6742 params.assoc_id = SCTP_FUTURE_ASSOC;
6743 } else if (len >= sizeof(struct sctp_assoc_value)) {
6744 len = sizeof(struct sctp_assoc_value);
6745 if (copy_from_user(¶ms, optval, len))
6750 asoc = sctp_id2assoc(sk, params.assoc_id);
6751 if (!asoc && params.assoc_id != SCTP_FUTURE_ASSOC &&
6752 sctp_style(sk, UDP))
6756 params.assoc_value = asoc->frag_point;
6758 params.assoc_value = sctp_sk(sk)->user_frag;
6760 if (put_user(len, optlen))
6762 if (len == sizeof(int)) {
6763 if (copy_to_user(optval, ¶ms.assoc_value, len))
6766 if (copy_to_user(optval, ¶ms, len))
6774 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
6775 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
6777 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
6778 char __user *optval, int __user *optlen)
6782 if (len < sizeof(int))
6787 val = sctp_sk(sk)->frag_interleave;
6788 if (put_user(len, optlen))
6790 if (copy_to_user(optval, &val, len))
6797 * 7.1.25. Set or Get the sctp partial delivery point
6798 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
6800 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
6801 char __user *optval,
6806 if (len < sizeof(u32))
6811 val = sctp_sk(sk)->pd_point;
6812 if (put_user(len, optlen))
6814 if (copy_to_user(optval, &val, len))
6821 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
6822 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
6824 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
6825 char __user *optval,
6828 struct sctp_assoc_value params;
6829 struct sctp_association *asoc;
6831 if (len == sizeof(int)) {
6832 pr_warn_ratelimited(DEPRECATED
6834 "Use of int in max_burst socket option.\n"
6835 "Use struct sctp_assoc_value instead\n",
6836 current->comm, task_pid_nr(current));
6837 params.assoc_id = SCTP_FUTURE_ASSOC;
6838 } else if (len >= sizeof(struct sctp_assoc_value)) {
6839 len = sizeof(struct sctp_assoc_value);
6840 if (copy_from_user(¶ms, optval, len))
6845 asoc = sctp_id2assoc(sk, params.assoc_id);
6846 if (!asoc && params.assoc_id != SCTP_FUTURE_ASSOC &&
6847 sctp_style(sk, UDP))
6850 params.assoc_value = asoc ? asoc->max_burst : sctp_sk(sk)->max_burst;
6852 if (len == sizeof(int)) {
6853 if (copy_to_user(optval, ¶ms.assoc_value, len))
6856 if (copy_to_user(optval, ¶ms, len))
6864 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
6865 char __user *optval, int __user *optlen)
6867 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6868 struct sctp_hmacalgo __user *p = (void __user *)optval;
6869 struct sctp_hmac_algo_param *hmacs;
6874 if (!ep->auth_enable)
6877 hmacs = ep->auth_hmacs_list;
6878 data_len = ntohs(hmacs->param_hdr.length) -
6879 sizeof(struct sctp_paramhdr);
6881 if (len < sizeof(struct sctp_hmacalgo) + data_len)
6884 len = sizeof(struct sctp_hmacalgo) + data_len;
6885 num_idents = data_len / sizeof(u16);
6887 if (put_user(len, optlen))
6889 if (put_user(num_idents, &p->shmac_num_idents))
6891 for (i = 0; i < num_idents; i++) {
6892 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
6894 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
6900 static int sctp_getsockopt_active_key(struct sock *sk, int len,
6901 char __user *optval, int __user *optlen)
6903 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6904 struct sctp_authkeyid val;
6905 struct sctp_association *asoc;
6907 if (!ep->auth_enable)
6910 if (len < sizeof(struct sctp_authkeyid))
6913 len = sizeof(struct sctp_authkeyid);
6914 if (copy_from_user(&val, optval, len))
6917 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
6918 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
6922 val.scact_keynumber = asoc->active_key_id;
6924 val.scact_keynumber = ep->active_key_id;
6926 if (put_user(len, optlen))
6928 if (copy_to_user(optval, &val, len))
6934 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
6935 char __user *optval, int __user *optlen)
6937 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6938 struct sctp_authchunks __user *p = (void __user *)optval;
6939 struct sctp_authchunks val;
6940 struct sctp_association *asoc;
6941 struct sctp_chunks_param *ch;
6945 if (!ep->auth_enable)
6948 if (len < sizeof(struct sctp_authchunks))
6951 if (copy_from_user(&val, optval, sizeof(val)))
6954 to = p->gauth_chunks;
6955 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6959 ch = asoc->peer.peer_chunks;
6963 /* See if the user provided enough room for all the data */
6964 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6965 if (len < num_chunks)
6968 if (copy_to_user(to, ch->chunks, num_chunks))
6971 len = sizeof(struct sctp_authchunks) + num_chunks;
6972 if (put_user(len, optlen))
6974 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6979 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
6980 char __user *optval, int __user *optlen)
6982 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6983 struct sctp_authchunks __user *p = (void __user *)optval;
6984 struct sctp_authchunks val;
6985 struct sctp_association *asoc;
6986 struct sctp_chunks_param *ch;
6990 if (!ep->auth_enable)
6993 if (len < sizeof(struct sctp_authchunks))
6996 if (copy_from_user(&val, optval, sizeof(val)))
6999 to = p->gauth_chunks;
7000 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
7001 if (!asoc && val.gauth_assoc_id != SCTP_FUTURE_ASSOC &&
7002 sctp_style(sk, UDP))
7005 ch = asoc ? (struct sctp_chunks_param *)asoc->c.auth_chunks
7006 : ep->auth_chunk_list;
7010 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
7011 if (len < sizeof(struct sctp_authchunks) + num_chunks)
7014 if (copy_to_user(to, ch->chunks, num_chunks))
7017 len = sizeof(struct sctp_authchunks) + num_chunks;
7018 if (put_user(len, optlen))
7020 if (put_user(num_chunks, &p->gauth_number_of_chunks))
7027 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
7028 * This option gets the current number of associations that are attached
7029 * to a one-to-many style socket. The option value is an uint32_t.
7031 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
7032 char __user *optval, int __user *optlen)
7034 struct sctp_sock *sp = sctp_sk(sk);
7035 struct sctp_association *asoc;
7038 if (sctp_style(sk, TCP))
7041 if (len < sizeof(u32))
7046 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
7050 if (put_user(len, optlen))
7052 if (copy_to_user(optval, &val, len))
7059 * 8.1.23 SCTP_AUTO_ASCONF
7060 * See the corresponding setsockopt entry as description
7062 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
7063 char __user *optval, int __user *optlen)
7067 if (len < sizeof(int))
7071 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
7073 if (put_user(len, optlen))
7075 if (copy_to_user(optval, &val, len))
7081 * 8.2.6. Get the Current Identifiers of Associations
7082 * (SCTP_GET_ASSOC_ID_LIST)
7084 * This option gets the current list of SCTP association identifiers of
7085 * the SCTP associations handled by a one-to-many style socket.
7087 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
7088 char __user *optval, int __user *optlen)
7090 struct sctp_sock *sp = sctp_sk(sk);
7091 struct sctp_association *asoc;
7092 struct sctp_assoc_ids *ids;
7095 if (sctp_style(sk, TCP))
7098 if (len < sizeof(struct sctp_assoc_ids))
7101 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
7105 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
7108 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
7110 ids = kmalloc(len, GFP_USER | __GFP_NOWARN);
7114 ids->gaids_number_of_ids = num;
7116 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
7117 ids->gaids_assoc_id[num++] = asoc->assoc_id;
7120 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
7130 * SCTP_PEER_ADDR_THLDS
7132 * This option allows us to fetch the partially failed threshold for one or all
7133 * transports in an association. See Section 6.1 of:
7134 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
7136 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
7137 char __user *optval,
7141 struct sctp_paddrthlds val;
7142 struct sctp_transport *trans;
7143 struct sctp_association *asoc;
7145 if (len < sizeof(struct sctp_paddrthlds))
7147 len = sizeof(struct sctp_paddrthlds);
7148 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
7151 if (!sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
7152 trans = sctp_addr_id2transport(sk, &val.spt_address,
7157 val.spt_pathmaxrxt = trans->pathmaxrxt;
7158 val.spt_pathpfthld = trans->pf_retrans;
7163 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
7164 if (!asoc && val.spt_assoc_id != SCTP_FUTURE_ASSOC &&
7165 sctp_style(sk, UDP))
7169 val.spt_pathpfthld = asoc->pf_retrans;
7170 val.spt_pathmaxrxt = asoc->pathmaxrxt;
7172 struct sctp_sock *sp = sctp_sk(sk);
7174 val.spt_pathpfthld = sp->pf_retrans;
7175 val.spt_pathmaxrxt = sp->pathmaxrxt;
7178 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
7185 * SCTP_GET_ASSOC_STATS
7187 * This option retrieves local per endpoint statistics. It is modeled
7188 * after OpenSolaris' implementation
7190 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
7191 char __user *optval,
7194 struct sctp_assoc_stats sas;
7195 struct sctp_association *asoc = NULL;
7197 /* User must provide at least the assoc id */
7198 if (len < sizeof(sctp_assoc_t))
7201 /* Allow the struct to grow and fill in as much as possible */
7202 len = min_t(size_t, len, sizeof(sas));
7204 if (copy_from_user(&sas, optval, len))
7207 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
7211 sas.sas_rtxchunks = asoc->stats.rtxchunks;
7212 sas.sas_gapcnt = asoc->stats.gapcnt;
7213 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
7214 sas.sas_osacks = asoc->stats.osacks;
7215 sas.sas_isacks = asoc->stats.isacks;
7216 sas.sas_octrlchunks = asoc->stats.octrlchunks;
7217 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
7218 sas.sas_oodchunks = asoc->stats.oodchunks;
7219 sas.sas_iodchunks = asoc->stats.iodchunks;
7220 sas.sas_ouodchunks = asoc->stats.ouodchunks;
7221 sas.sas_iuodchunks = asoc->stats.iuodchunks;
7222 sas.sas_idupchunks = asoc->stats.idupchunks;
7223 sas.sas_opackets = asoc->stats.opackets;
7224 sas.sas_ipackets = asoc->stats.ipackets;
7226 /* New high max rto observed, will return 0 if not a single
7227 * RTO update took place. obs_rto_ipaddr will be bogus
7230 sas.sas_maxrto = asoc->stats.max_obs_rto;
7231 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
7232 sizeof(struct sockaddr_storage));
7234 /* Mark beginning of a new observation period */
7235 asoc->stats.max_obs_rto = asoc->rto_min;
7237 if (put_user(len, optlen))
7240 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
7242 if (copy_to_user(optval, &sas, len))
7248 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
7249 char __user *optval,
7254 if (len < sizeof(int))
7258 if (sctp_sk(sk)->recvrcvinfo)
7260 if (put_user(len, optlen))
7262 if (copy_to_user(optval, &val, len))
7268 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
7269 char __user *optval,
7274 if (len < sizeof(int))
7278 if (sctp_sk(sk)->recvnxtinfo)
7280 if (put_user(len, optlen))
7282 if (copy_to_user(optval, &val, len))
7288 static int sctp_getsockopt_pr_supported(struct sock *sk, int len,
7289 char __user *optval,
7292 struct sctp_assoc_value params;
7293 struct sctp_association *asoc;
7294 int retval = -EFAULT;
7296 if (len < sizeof(params)) {
7301 len = sizeof(params);
7302 if (copy_from_user(¶ms, optval, len))
7305 asoc = sctp_id2assoc(sk, params.assoc_id);
7306 if (!asoc && params.assoc_id != SCTP_FUTURE_ASSOC &&
7307 sctp_style(sk, UDP)) {
7312 params.assoc_value = asoc ? asoc->prsctp_enable
7313 : sctp_sk(sk)->ep->prsctp_enable;
7315 if (put_user(len, optlen))
7318 if (copy_to_user(optval, ¶ms, len))
7327 static int sctp_getsockopt_default_prinfo(struct sock *sk, int len,
7328 char __user *optval,
7331 struct sctp_default_prinfo info;
7332 struct sctp_association *asoc;
7333 int retval = -EFAULT;
7335 if (len < sizeof(info)) {
7341 if (copy_from_user(&info, optval, len))
7344 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
7345 if (!asoc && info.pr_assoc_id != SCTP_FUTURE_ASSOC &&
7346 sctp_style(sk, UDP)) {
7352 info.pr_policy = SCTP_PR_POLICY(asoc->default_flags);
7353 info.pr_value = asoc->default_timetolive;
7355 struct sctp_sock *sp = sctp_sk(sk);
7357 info.pr_policy = SCTP_PR_POLICY(sp->default_flags);
7358 info.pr_value = sp->default_timetolive;
7361 if (put_user(len, optlen))
7364 if (copy_to_user(optval, &info, len))
7373 static int sctp_getsockopt_pr_assocstatus(struct sock *sk, int len,
7374 char __user *optval,
7377 struct sctp_prstatus params;
7378 struct sctp_association *asoc;
7380 int retval = -EINVAL;
7382 if (len < sizeof(params))
7385 len = sizeof(params);
7386 if (copy_from_user(¶ms, optval, len)) {
7391 policy = params.sprstat_policy;
7392 if (!policy || (policy & ~(SCTP_PR_SCTP_MASK | SCTP_PR_SCTP_ALL)) ||
7393 ((policy & SCTP_PR_SCTP_ALL) && (policy & SCTP_PR_SCTP_MASK)))
7396 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
7400 if (policy == SCTP_PR_SCTP_ALL) {
7401 params.sprstat_abandoned_unsent = 0;
7402 params.sprstat_abandoned_sent = 0;
7403 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
7404 params.sprstat_abandoned_unsent +=
7405 asoc->abandoned_unsent[policy];
7406 params.sprstat_abandoned_sent +=
7407 asoc->abandoned_sent[policy];
7410 params.sprstat_abandoned_unsent =
7411 asoc->abandoned_unsent[__SCTP_PR_INDEX(policy)];
7412 params.sprstat_abandoned_sent =
7413 asoc->abandoned_sent[__SCTP_PR_INDEX(policy)];
7416 if (put_user(len, optlen)) {
7421 if (copy_to_user(optval, ¶ms, len)) {
7432 static int sctp_getsockopt_pr_streamstatus(struct sock *sk, int len,
7433 char __user *optval,
7436 struct sctp_stream_out_ext *streamoute;
7437 struct sctp_association *asoc;
7438 struct sctp_prstatus params;
7439 int retval = -EINVAL;
7442 if (len < sizeof(params))
7445 len = sizeof(params);
7446 if (copy_from_user(¶ms, optval, len)) {
7451 policy = params.sprstat_policy;
7452 if (!policy || (policy & ~(SCTP_PR_SCTP_MASK | SCTP_PR_SCTP_ALL)) ||
7453 ((policy & SCTP_PR_SCTP_ALL) && (policy & SCTP_PR_SCTP_MASK)))
7456 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
7457 if (!asoc || params.sprstat_sid >= asoc->stream.outcnt)
7460 streamoute = SCTP_SO(&asoc->stream, params.sprstat_sid)->ext;
7462 /* Not allocated yet, means all stats are 0 */
7463 params.sprstat_abandoned_unsent = 0;
7464 params.sprstat_abandoned_sent = 0;
7469 if (policy == SCTP_PR_SCTP_ALL) {
7470 params.sprstat_abandoned_unsent = 0;
7471 params.sprstat_abandoned_sent = 0;
7472 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
7473 params.sprstat_abandoned_unsent +=
7474 streamoute->abandoned_unsent[policy];
7475 params.sprstat_abandoned_sent +=
7476 streamoute->abandoned_sent[policy];
7479 params.sprstat_abandoned_unsent =
7480 streamoute->abandoned_unsent[__SCTP_PR_INDEX(policy)];
7481 params.sprstat_abandoned_sent =
7482 streamoute->abandoned_sent[__SCTP_PR_INDEX(policy)];
7485 if (put_user(len, optlen) || copy_to_user(optval, ¶ms, len)) {
7496 static int sctp_getsockopt_reconfig_supported(struct sock *sk, int len,
7497 char __user *optval,
7500 struct sctp_assoc_value params;
7501 struct sctp_association *asoc;
7502 int retval = -EFAULT;
7504 if (len < sizeof(params)) {
7509 len = sizeof(params);
7510 if (copy_from_user(¶ms, optval, len))
7513 asoc = sctp_id2assoc(sk, params.assoc_id);
7514 if (!asoc && params.assoc_id != SCTP_FUTURE_ASSOC &&
7515 sctp_style(sk, UDP)) {
7520 params.assoc_value = asoc ? asoc->reconf_enable
7521 : sctp_sk(sk)->ep->reconf_enable;
7523 if (put_user(len, optlen))
7526 if (copy_to_user(optval, ¶ms, len))
7535 static int sctp_getsockopt_enable_strreset(struct sock *sk, int len,
7536 char __user *optval,
7539 struct sctp_assoc_value params;
7540 struct sctp_association *asoc;
7541 int retval = -EFAULT;
7543 if (len < sizeof(params)) {
7548 len = sizeof(params);
7549 if (copy_from_user(¶ms, optval, len))
7552 asoc = sctp_id2assoc(sk, params.assoc_id);
7553 if (!asoc && params.assoc_id != SCTP_FUTURE_ASSOC &&
7554 sctp_style(sk, UDP)) {
7559 params.assoc_value = asoc ? asoc->strreset_enable
7560 : sctp_sk(sk)->ep->strreset_enable;
7562 if (put_user(len, optlen))
7565 if (copy_to_user(optval, ¶ms, len))
7574 static int sctp_getsockopt_scheduler(struct sock *sk, int len,
7575 char __user *optval,
7578 struct sctp_assoc_value params;
7579 struct sctp_association *asoc;
7580 int retval = -EFAULT;
7582 if (len < sizeof(params)) {
7587 len = sizeof(params);
7588 if (copy_from_user(¶ms, optval, len))
7591 asoc = sctp_id2assoc(sk, params.assoc_id);
7592 if (!asoc && params.assoc_id != SCTP_FUTURE_ASSOC &&
7593 sctp_style(sk, UDP)) {
7598 params.assoc_value = asoc ? sctp_sched_get_sched(asoc)
7599 : sctp_sk(sk)->default_ss;
7601 if (put_user(len, optlen))
7604 if (copy_to_user(optval, ¶ms, len))
7613 static int sctp_getsockopt_scheduler_value(struct sock *sk, int len,
7614 char __user *optval,
7617 struct sctp_stream_value params;
7618 struct sctp_association *asoc;
7619 int retval = -EFAULT;
7621 if (len < sizeof(params)) {
7626 len = sizeof(params);
7627 if (copy_from_user(¶ms, optval, len))
7630 asoc = sctp_id2assoc(sk, params.assoc_id);
7636 retval = sctp_sched_get_value(asoc, params.stream_id,
7637 ¶ms.stream_value);
7641 if (put_user(len, optlen)) {
7646 if (copy_to_user(optval, ¶ms, len)) {
7655 static int sctp_getsockopt_interleaving_supported(struct sock *sk, int len,
7656 char __user *optval,
7659 struct sctp_assoc_value params;
7660 struct sctp_association *asoc;
7661 int retval = -EFAULT;
7663 if (len < sizeof(params)) {
7668 len = sizeof(params);
7669 if (copy_from_user(¶ms, optval, len))
7672 asoc = sctp_id2assoc(sk, params.assoc_id);
7673 if (!asoc && params.assoc_id != SCTP_FUTURE_ASSOC &&
7674 sctp_style(sk, UDP)) {
7679 params.assoc_value = asoc ? asoc->intl_enable
7680 : sctp_sk(sk)->strm_interleave;
7682 if (put_user(len, optlen))
7685 if (copy_to_user(optval, ¶ms, len))
7694 static int sctp_getsockopt_reuse_port(struct sock *sk, int len,
7695 char __user *optval,
7700 if (len < sizeof(int))
7704 val = sctp_sk(sk)->reuse;
7705 if (put_user(len, optlen))
7708 if (copy_to_user(optval, &val, len))
7714 static int sctp_getsockopt_event(struct sock *sk, int len, char __user *optval,
7717 struct sctp_association *asoc;
7718 struct sctp_event param;
7721 if (len < sizeof(param))
7724 len = sizeof(param);
7725 if (copy_from_user(¶m, optval, len))
7728 if (param.se_type < SCTP_SN_TYPE_BASE ||
7729 param.se_type > SCTP_SN_TYPE_MAX)
7732 asoc = sctp_id2assoc(sk, param.se_assoc_id);
7733 if (!asoc && param.se_assoc_id != SCTP_FUTURE_ASSOC &&
7734 sctp_style(sk, UDP))
7737 subscribe = asoc ? asoc->subscribe : sctp_sk(sk)->subscribe;
7738 param.se_on = sctp_ulpevent_type_enabled(subscribe, param.se_type);
7740 if (put_user(len, optlen))
7743 if (copy_to_user(optval, ¶m, len))
7749 static int sctp_getsockopt(struct sock *sk, int level, int optname,
7750 char __user *optval, int __user *optlen)
7755 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
7757 /* I can hardly begin to describe how wrong this is. This is
7758 * so broken as to be worse than useless. The API draft
7759 * REALLY is NOT helpful here... I am not convinced that the
7760 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
7761 * are at all well-founded.
7763 if (level != SOL_SCTP) {
7764 struct sctp_af *af = sctp_sk(sk)->pf->af;
7766 retval = af->getsockopt(sk, level, optname, optval, optlen);
7770 if (get_user(len, optlen))
7780 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
7782 case SCTP_DISABLE_FRAGMENTS:
7783 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
7787 retval = sctp_getsockopt_events(sk, len, optval, optlen);
7789 case SCTP_AUTOCLOSE:
7790 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
7792 case SCTP_SOCKOPT_PEELOFF:
7793 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
7795 case SCTP_SOCKOPT_PEELOFF_FLAGS:
7796 retval = sctp_getsockopt_peeloff_flags(sk, len, optval, optlen);
7798 case SCTP_PEER_ADDR_PARAMS:
7799 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
7802 case SCTP_DELAYED_SACK:
7803 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
7807 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
7809 case SCTP_GET_PEER_ADDRS:
7810 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
7813 case SCTP_GET_LOCAL_ADDRS:
7814 retval = sctp_getsockopt_local_addrs(sk, len, optval,
7817 case SCTP_SOCKOPT_CONNECTX3:
7818 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
7820 case SCTP_DEFAULT_SEND_PARAM:
7821 retval = sctp_getsockopt_default_send_param(sk, len,
7824 case SCTP_DEFAULT_SNDINFO:
7825 retval = sctp_getsockopt_default_sndinfo(sk, len,
7828 case SCTP_PRIMARY_ADDR:
7829 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
7832 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
7835 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
7837 case SCTP_ASSOCINFO:
7838 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
7840 case SCTP_I_WANT_MAPPED_V4_ADDR:
7841 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
7844 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
7846 case SCTP_GET_PEER_ADDR_INFO:
7847 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
7850 case SCTP_ADAPTATION_LAYER:
7851 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
7855 retval = sctp_getsockopt_context(sk, len, optval, optlen);
7857 case SCTP_FRAGMENT_INTERLEAVE:
7858 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
7861 case SCTP_PARTIAL_DELIVERY_POINT:
7862 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
7865 case SCTP_MAX_BURST:
7866 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
7869 case SCTP_AUTH_CHUNK:
7870 case SCTP_AUTH_DELETE_KEY:
7871 case SCTP_AUTH_DEACTIVATE_KEY:
7872 retval = -EOPNOTSUPP;
7874 case SCTP_HMAC_IDENT:
7875 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
7877 case SCTP_AUTH_ACTIVE_KEY:
7878 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
7880 case SCTP_PEER_AUTH_CHUNKS:
7881 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
7884 case SCTP_LOCAL_AUTH_CHUNKS:
7885 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
7888 case SCTP_GET_ASSOC_NUMBER:
7889 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
7891 case SCTP_GET_ASSOC_ID_LIST:
7892 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
7894 case SCTP_AUTO_ASCONF:
7895 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
7897 case SCTP_PEER_ADDR_THLDS:
7898 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
7900 case SCTP_GET_ASSOC_STATS:
7901 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
7903 case SCTP_RECVRCVINFO:
7904 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
7906 case SCTP_RECVNXTINFO:
7907 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
7909 case SCTP_PR_SUPPORTED:
7910 retval = sctp_getsockopt_pr_supported(sk, len, optval, optlen);
7912 case SCTP_DEFAULT_PRINFO:
7913 retval = sctp_getsockopt_default_prinfo(sk, len, optval,
7916 case SCTP_PR_ASSOC_STATUS:
7917 retval = sctp_getsockopt_pr_assocstatus(sk, len, optval,
7920 case SCTP_PR_STREAM_STATUS:
7921 retval = sctp_getsockopt_pr_streamstatus(sk, len, optval,
7924 case SCTP_RECONFIG_SUPPORTED:
7925 retval = sctp_getsockopt_reconfig_supported(sk, len, optval,
7928 case SCTP_ENABLE_STREAM_RESET:
7929 retval = sctp_getsockopt_enable_strreset(sk, len, optval,
7932 case SCTP_STREAM_SCHEDULER:
7933 retval = sctp_getsockopt_scheduler(sk, len, optval,
7936 case SCTP_STREAM_SCHEDULER_VALUE:
7937 retval = sctp_getsockopt_scheduler_value(sk, len, optval,
7940 case SCTP_INTERLEAVING_SUPPORTED:
7941 retval = sctp_getsockopt_interleaving_supported(sk, len, optval,
7944 case SCTP_REUSE_PORT:
7945 retval = sctp_getsockopt_reuse_port(sk, len, optval, optlen);
7948 retval = sctp_getsockopt_event(sk, len, optval, optlen);
7951 retval = -ENOPROTOOPT;
7959 static int sctp_hash(struct sock *sk)
7965 static void sctp_unhash(struct sock *sk)
7970 /* Check if port is acceptable. Possibly find first available port.
7972 * The port hash table (contained in the 'global' SCTP protocol storage
7973 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
7974 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
7975 * list (the list number is the port number hashed out, so as you
7976 * would expect from a hash function, all the ports in a given list have
7977 * such a number that hashes out to the same list number; you were
7978 * expecting that, right?); so each list has a set of ports, with a
7979 * link to the socket (struct sock) that uses it, the port number and
7980 * a fastreuse flag (FIXME: NPI ipg).
7982 static struct sctp_bind_bucket *sctp_bucket_create(
7983 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
7985 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
7987 struct sctp_sock *sp = sctp_sk(sk);
7988 bool reuse = (sk->sk_reuse || sp->reuse);
7989 struct sctp_bind_hashbucket *head; /* hash list */
7990 kuid_t uid = sock_i_uid(sk);
7991 struct sctp_bind_bucket *pp;
7992 unsigned short snum;
7995 snum = ntohs(addr->v4.sin_port);
7997 pr_debug("%s: begins, snum:%d\n", __func__, snum);
8002 /* Search for an available port. */
8003 int low, high, remaining, index;
8005 struct net *net = sock_net(sk);
8007 inet_get_local_port_range(net, &low, &high);
8008 remaining = (high - low) + 1;
8009 rover = prandom_u32() % remaining + low;
8013 if ((rover < low) || (rover > high))
8015 if (inet_is_local_reserved_port(net, rover))
8017 index = sctp_phashfn(sock_net(sk), rover);
8018 head = &sctp_port_hashtable[index];
8019 spin_lock(&head->lock);
8020 sctp_for_each_hentry(pp, &head->chain)
8021 if ((pp->port == rover) &&
8022 net_eq(sock_net(sk), pp->net))
8026 spin_unlock(&head->lock);
8027 } while (--remaining > 0);
8029 /* Exhausted local port range during search? */
8034 /* OK, here is the one we will use. HEAD (the port
8035 * hash table list entry) is non-NULL and we hold it's
8040 /* We are given an specific port number; we verify
8041 * that it is not being used. If it is used, we will
8042 * exahust the search in the hash list corresponding
8043 * to the port number (snum) - we detect that with the
8044 * port iterator, pp being NULL.
8046 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
8047 spin_lock(&head->lock);
8048 sctp_for_each_hentry(pp, &head->chain) {
8049 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
8056 if (!hlist_empty(&pp->owner)) {
8057 /* We had a port hash table hit - there is an
8058 * available port (pp != NULL) and it is being
8059 * used by other socket (pp->owner not empty); that other
8060 * socket is going to be sk2.
8064 pr_debug("%s: found a possible match\n", __func__);
8066 if ((pp->fastreuse && reuse &&
8067 sk->sk_state != SCTP_SS_LISTENING) ||
8068 (pp->fastreuseport && sk->sk_reuseport &&
8069 uid_eq(pp->fastuid, uid)))
8072 /* Run through the list of sockets bound to the port
8073 * (pp->port) [via the pointers bind_next and
8074 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
8075 * we get the endpoint they describe and run through
8076 * the endpoint's list of IP (v4 or v6) addresses,
8077 * comparing each of the addresses with the address of
8078 * the socket sk. If we find a match, then that means
8079 * that this port/socket (sk) combination are already
8082 sk_for_each_bound(sk2, &pp->owner) {
8083 struct sctp_sock *sp2 = sctp_sk(sk2);
8084 struct sctp_endpoint *ep2 = sp2->ep;
8087 (reuse && (sk2->sk_reuse || sp2->reuse) &&
8088 sk2->sk_state != SCTP_SS_LISTENING) ||
8089 (sk->sk_reuseport && sk2->sk_reuseport &&
8090 uid_eq(uid, sock_i_uid(sk2))))
8093 if (sctp_bind_addr_conflict(&ep2->base.bind_addr,
8100 pr_debug("%s: found a match\n", __func__);
8103 /* If there was a hash table miss, create a new port. */
8105 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
8108 /* In either case (hit or miss), make sure fastreuse is 1 only
8109 * if sk->sk_reuse is too (that is, if the caller requested
8110 * SO_REUSEADDR on this socket -sk-).
8112 if (hlist_empty(&pp->owner)) {
8113 if (reuse && sk->sk_state != SCTP_SS_LISTENING)
8118 if (sk->sk_reuseport) {
8119 pp->fastreuseport = 1;
8122 pp->fastreuseport = 0;
8125 if (pp->fastreuse &&
8126 (!reuse || sk->sk_state == SCTP_SS_LISTENING))
8129 if (pp->fastreuseport &&
8130 (!sk->sk_reuseport || !uid_eq(pp->fastuid, uid)))
8131 pp->fastreuseport = 0;
8134 /* We are set, so fill up all the data in the hash table
8135 * entry, tie the socket list information with the rest of the
8136 * sockets FIXME: Blurry, NPI (ipg).
8139 if (!sp->bind_hash) {
8140 inet_sk(sk)->inet_num = snum;
8141 sk_add_bind_node(sk, &pp->owner);
8147 spin_unlock(&head->lock);
8154 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
8155 * port is requested.
8157 static int sctp_get_port(struct sock *sk, unsigned short snum)
8159 union sctp_addr addr;
8160 struct sctp_af *af = sctp_sk(sk)->pf->af;
8162 /* Set up a dummy address struct from the sk. */
8163 af->from_sk(&addr, sk);
8164 addr.v4.sin_port = htons(snum);
8166 /* Note: sk->sk_num gets filled in if ephemeral port request. */
8167 return !!sctp_get_port_local(sk, &addr);
8171 * Move a socket to LISTENING state.
8173 static int sctp_listen_start(struct sock *sk, int backlog)
8175 struct sctp_sock *sp = sctp_sk(sk);
8176 struct sctp_endpoint *ep = sp->ep;
8177 struct crypto_shash *tfm = NULL;
8180 /* Allocate HMAC for generating cookie. */
8181 if (!sp->hmac && sp->sctp_hmac_alg) {
8182 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
8183 tfm = crypto_alloc_shash(alg, 0, 0);
8185 net_info_ratelimited("failed to load transform for %s: %ld\n",
8186 sp->sctp_hmac_alg, PTR_ERR(tfm));
8189 sctp_sk(sk)->hmac = tfm;
8193 * If a bind() or sctp_bindx() is not called prior to a listen()
8194 * call that allows new associations to be accepted, the system
8195 * picks an ephemeral port and will choose an address set equivalent
8196 * to binding with a wildcard address.
8198 * This is not currently spelled out in the SCTP sockets
8199 * extensions draft, but follows the practice as seen in TCP
8203 inet_sk_set_state(sk, SCTP_SS_LISTENING);
8204 if (!ep->base.bind_addr.port) {
8205 if (sctp_autobind(sk))
8208 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
8209 inet_sk_set_state(sk, SCTP_SS_CLOSED);
8214 sk->sk_max_ack_backlog = backlog;
8215 return sctp_hash_endpoint(ep);
8219 * 4.1.3 / 5.1.3 listen()
8221 * By default, new associations are not accepted for UDP style sockets.
8222 * An application uses listen() to mark a socket as being able to
8223 * accept new associations.
8225 * On TCP style sockets, applications use listen() to ready the SCTP
8226 * endpoint for accepting inbound associations.
8228 * On both types of endpoints a backlog of '0' disables listening.
8230 * Move a socket to LISTENING state.
8232 int sctp_inet_listen(struct socket *sock, int backlog)
8234 struct sock *sk = sock->sk;
8235 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
8238 if (unlikely(backlog < 0))
8243 /* Peeled-off sockets are not allowed to listen(). */
8244 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
8247 if (sock->state != SS_UNCONNECTED)
8250 if (!sctp_sstate(sk, LISTENING) && !sctp_sstate(sk, CLOSED))
8253 /* If backlog is zero, disable listening. */
8255 if (sctp_sstate(sk, CLOSED))
8259 sctp_unhash_endpoint(ep);
8260 sk->sk_state = SCTP_SS_CLOSED;
8261 if (sk->sk_reuse || sctp_sk(sk)->reuse)
8262 sctp_sk(sk)->bind_hash->fastreuse = 1;
8266 /* If we are already listening, just update the backlog */
8267 if (sctp_sstate(sk, LISTENING))
8268 sk->sk_max_ack_backlog = backlog;
8270 err = sctp_listen_start(sk, backlog);
8282 * This function is done by modeling the current datagram_poll() and the
8283 * tcp_poll(). Note that, based on these implementations, we don't
8284 * lock the socket in this function, even though it seems that,
8285 * ideally, locking or some other mechanisms can be used to ensure
8286 * the integrity of the counters (sndbuf and wmem_alloc) used
8287 * in this place. We assume that we don't need locks either until proven
8290 * Another thing to note is that we include the Async I/O support
8291 * here, again, by modeling the current TCP/UDP code. We don't have
8292 * a good way to test with it yet.
8294 __poll_t sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
8296 struct sock *sk = sock->sk;
8297 struct sctp_sock *sp = sctp_sk(sk);
8300 poll_wait(file, sk_sleep(sk), wait);
8302 sock_rps_record_flow(sk);
8304 /* A TCP-style listening socket becomes readable when the accept queue
8307 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
8308 return (!list_empty(&sp->ep->asocs)) ?
8309 (EPOLLIN | EPOLLRDNORM) : 0;
8313 /* Is there any exceptional events? */
8314 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
8316 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? EPOLLPRI : 0);
8317 if (sk->sk_shutdown & RCV_SHUTDOWN)
8318 mask |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
8319 if (sk->sk_shutdown == SHUTDOWN_MASK)
8322 /* Is it readable? Reconsider this code with TCP-style support. */
8323 if (!skb_queue_empty(&sk->sk_receive_queue))
8324 mask |= EPOLLIN | EPOLLRDNORM;
8326 /* The association is either gone or not ready. */
8327 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
8330 /* Is it writable? */
8331 if (sctp_writeable(sk)) {
8332 mask |= EPOLLOUT | EPOLLWRNORM;
8334 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
8336 * Since the socket is not locked, the buffer
8337 * might be made available after the writeable check and
8338 * before the bit is set. This could cause a lost I/O
8339 * signal. tcp_poll() has a race breaker for this race
8340 * condition. Based on their implementation, we put
8341 * in the following code to cover it as well.
8343 if (sctp_writeable(sk))
8344 mask |= EPOLLOUT | EPOLLWRNORM;
8349 /********************************************************************
8350 * 2nd Level Abstractions
8351 ********************************************************************/
8353 static struct sctp_bind_bucket *sctp_bucket_create(
8354 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
8356 struct sctp_bind_bucket *pp;
8358 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
8360 SCTP_DBG_OBJCNT_INC(bind_bucket);
8363 INIT_HLIST_HEAD(&pp->owner);
8365 hlist_add_head(&pp->node, &head->chain);
8370 /* Caller must hold hashbucket lock for this tb with local BH disabled */
8371 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
8373 if (pp && hlist_empty(&pp->owner)) {
8374 __hlist_del(&pp->node);
8375 kmem_cache_free(sctp_bucket_cachep, pp);
8376 SCTP_DBG_OBJCNT_DEC(bind_bucket);
8380 /* Release this socket's reference to a local port. */
8381 static inline void __sctp_put_port(struct sock *sk)
8383 struct sctp_bind_hashbucket *head =
8384 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
8385 inet_sk(sk)->inet_num)];
8386 struct sctp_bind_bucket *pp;
8388 spin_lock(&head->lock);
8389 pp = sctp_sk(sk)->bind_hash;
8390 __sk_del_bind_node(sk);
8391 sctp_sk(sk)->bind_hash = NULL;
8392 inet_sk(sk)->inet_num = 0;
8393 sctp_bucket_destroy(pp);
8394 spin_unlock(&head->lock);
8397 void sctp_put_port(struct sock *sk)
8400 __sctp_put_port(sk);
8405 * The system picks an ephemeral port and choose an address set equivalent
8406 * to binding with a wildcard address.
8407 * One of those addresses will be the primary address for the association.
8408 * This automatically enables the multihoming capability of SCTP.
8410 static int sctp_autobind(struct sock *sk)
8412 union sctp_addr autoaddr;
8416 /* Initialize a local sockaddr structure to INADDR_ANY. */
8417 af = sctp_sk(sk)->pf->af;
8419 port = htons(inet_sk(sk)->inet_num);
8420 af->inaddr_any(&autoaddr, port);
8422 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
8425 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
8428 * 4.2 The cmsghdr Structure *
8430 * When ancillary data is sent or received, any number of ancillary data
8431 * objects can be specified by the msg_control and msg_controllen members of
8432 * the msghdr structure, because each object is preceded by
8433 * a cmsghdr structure defining the object's length (the cmsg_len member).
8434 * Historically Berkeley-derived implementations have passed only one object
8435 * at a time, but this API allows multiple objects to be
8436 * passed in a single call to sendmsg() or recvmsg(). The following example
8437 * shows two ancillary data objects in a control buffer.
8439 * |<--------------------------- msg_controllen -------------------------->|
8442 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
8444 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
8447 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
8449 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
8452 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
8453 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
8455 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
8457 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
8464 static int sctp_msghdr_parse(const struct msghdr *msg, struct sctp_cmsgs *cmsgs)
8466 struct msghdr *my_msg = (struct msghdr *)msg;
8467 struct cmsghdr *cmsg;
8469 for_each_cmsghdr(cmsg, my_msg) {
8470 if (!CMSG_OK(my_msg, cmsg))
8473 /* Should we parse this header or ignore? */
8474 if (cmsg->cmsg_level != IPPROTO_SCTP)
8477 /* Strictly check lengths following example in SCM code. */
8478 switch (cmsg->cmsg_type) {
8480 /* SCTP Socket API Extension
8481 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
8483 * This cmsghdr structure provides information for
8484 * initializing new SCTP associations with sendmsg().
8485 * The SCTP_INITMSG socket option uses this same data
8486 * structure. This structure is not used for
8489 * cmsg_level cmsg_type cmsg_data[]
8490 * ------------ ------------ ----------------------
8491 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
8493 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
8496 cmsgs->init = CMSG_DATA(cmsg);
8500 /* SCTP Socket API Extension
8501 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
8503 * This cmsghdr structure specifies SCTP options for
8504 * sendmsg() and describes SCTP header information
8505 * about a received message through recvmsg().
8507 * cmsg_level cmsg_type cmsg_data[]
8508 * ------------ ------------ ----------------------
8509 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
8511 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
8514 cmsgs->srinfo = CMSG_DATA(cmsg);
8516 if (cmsgs->srinfo->sinfo_flags &
8517 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
8518 SCTP_SACK_IMMEDIATELY | SCTP_SENDALL |
8519 SCTP_PR_SCTP_MASK | SCTP_ABORT | SCTP_EOF))
8524 /* SCTP Socket API Extension
8525 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
8527 * This cmsghdr structure specifies SCTP options for
8528 * sendmsg(). This structure and SCTP_RCVINFO replaces
8529 * SCTP_SNDRCV which has been deprecated.
8531 * cmsg_level cmsg_type cmsg_data[]
8532 * ------------ ------------ ---------------------
8533 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
8535 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
8538 cmsgs->sinfo = CMSG_DATA(cmsg);
8540 if (cmsgs->sinfo->snd_flags &
8541 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
8542 SCTP_SACK_IMMEDIATELY | SCTP_SENDALL |
8543 SCTP_PR_SCTP_MASK | SCTP_ABORT | SCTP_EOF))
8547 /* SCTP Socket API Extension
8548 * 5.3.7 SCTP PR-SCTP Information Structure (SCTP_PRINFO)
8550 * This cmsghdr structure specifies SCTP options for sendmsg().
8552 * cmsg_level cmsg_type cmsg_data[]
8553 * ------------ ------------ ---------------------
8554 * IPPROTO_SCTP SCTP_PRINFO struct sctp_prinfo
8556 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_prinfo)))
8559 cmsgs->prinfo = CMSG_DATA(cmsg);
8560 if (cmsgs->prinfo->pr_policy & ~SCTP_PR_SCTP_MASK)
8563 if (cmsgs->prinfo->pr_policy == SCTP_PR_SCTP_NONE)
8564 cmsgs->prinfo->pr_value = 0;
8567 /* SCTP Socket API Extension
8568 * 5.3.8 SCTP AUTH Information Structure (SCTP_AUTHINFO)
8570 * This cmsghdr structure specifies SCTP options for sendmsg().
8572 * cmsg_level cmsg_type cmsg_data[]
8573 * ------------ ------------ ---------------------
8574 * IPPROTO_SCTP SCTP_AUTHINFO struct sctp_authinfo
8576 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_authinfo)))
8579 cmsgs->authinfo = CMSG_DATA(cmsg);
8581 case SCTP_DSTADDRV4:
8582 case SCTP_DSTADDRV6:
8583 /* SCTP Socket API Extension
8584 * 5.3.9/10 SCTP Destination IPv4/6 Address Structure (SCTP_DSTADDRV4/6)
8586 * This cmsghdr structure specifies SCTP options for sendmsg().
8588 * cmsg_level cmsg_type cmsg_data[]
8589 * ------------ ------------ ---------------------
8590 * IPPROTO_SCTP SCTP_DSTADDRV4 struct in_addr
8591 * ------------ ------------ ---------------------
8592 * IPPROTO_SCTP SCTP_DSTADDRV6 struct in6_addr
8594 cmsgs->addrs_msg = my_msg;
8605 * Wait for a packet..
8606 * Note: This function is the same function as in core/datagram.c
8607 * with a few modifications to make lksctp work.
8609 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
8614 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
8616 /* Socket errors? */
8617 error = sock_error(sk);
8621 if (!skb_queue_empty(&sk->sk_receive_queue))
8624 /* Socket shut down? */
8625 if (sk->sk_shutdown & RCV_SHUTDOWN)
8628 /* Sequenced packets can come disconnected. If so we report the
8633 /* Is there a good reason to think that we may receive some data? */
8634 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
8637 /* Handle signals. */
8638 if (signal_pending(current))
8641 /* Let another process have a go. Since we are going to sleep
8642 * anyway. Note: This may cause odd behaviors if the message
8643 * does not fit in the user's buffer, but this seems to be the
8644 * only way to honor MSG_DONTWAIT realistically.
8647 *timeo_p = schedule_timeout(*timeo_p);
8651 finish_wait(sk_sleep(sk), &wait);
8655 error = sock_intr_errno(*timeo_p);
8658 finish_wait(sk_sleep(sk), &wait);
8663 /* Receive a datagram.
8664 * Note: This is pretty much the same routine as in core/datagram.c
8665 * with a few changes to make lksctp work.
8667 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
8668 int noblock, int *err)
8671 struct sk_buff *skb;
8674 timeo = sock_rcvtimeo(sk, noblock);
8676 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
8677 MAX_SCHEDULE_TIMEOUT);
8680 /* Again only user level code calls this function,
8681 * so nothing interrupt level
8682 * will suddenly eat the receive_queue.
8684 * Look at current nfs client by the way...
8685 * However, this function was correct in any case. 8)
8687 if (flags & MSG_PEEK) {
8688 skb = skb_peek(&sk->sk_receive_queue);
8690 refcount_inc(&skb->users);
8692 skb = __skb_dequeue(&sk->sk_receive_queue);
8698 /* Caller is allowed not to check sk->sk_err before calling. */
8699 error = sock_error(sk);
8703 if (sk->sk_shutdown & RCV_SHUTDOWN)
8706 if (sk_can_busy_loop(sk)) {
8707 sk_busy_loop(sk, noblock);
8709 if (!skb_queue_empty(&sk->sk_receive_queue))
8713 /* User doesn't want to wait. */
8717 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
8726 /* If sndbuf has changed, wake up per association sndbuf waiters. */
8727 static void __sctp_write_space(struct sctp_association *asoc)
8729 struct sock *sk = asoc->base.sk;
8731 if (sctp_wspace(asoc) <= 0)
8734 if (waitqueue_active(&asoc->wait))
8735 wake_up_interruptible(&asoc->wait);
8737 if (sctp_writeable(sk)) {
8738 struct socket_wq *wq;
8741 wq = rcu_dereference(sk->sk_wq);
8743 if (waitqueue_active(&wq->wait))
8744 wake_up_interruptible(&wq->wait);
8746 /* Note that we try to include the Async I/O support
8747 * here by modeling from the current TCP/UDP code.
8748 * We have not tested with it yet.
8750 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
8751 sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT);
8757 static void sctp_wake_up_waiters(struct sock *sk,
8758 struct sctp_association *asoc)
8760 struct sctp_association *tmp = asoc;
8762 /* We do accounting for the sndbuf space per association,
8763 * so we only need to wake our own association.
8765 if (asoc->ep->sndbuf_policy)
8766 return __sctp_write_space(asoc);
8768 /* If association goes down and is just flushing its
8769 * outq, then just normally notify others.
8771 if (asoc->base.dead)
8772 return sctp_write_space(sk);
8774 /* Accounting for the sndbuf space is per socket, so we
8775 * need to wake up others, try to be fair and in case of
8776 * other associations, let them have a go first instead
8777 * of just doing a sctp_write_space() call.
8779 * Note that we reach sctp_wake_up_waiters() only when
8780 * associations free up queued chunks, thus we are under
8781 * lock and the list of associations on a socket is
8782 * guaranteed not to change.
8784 for (tmp = list_next_entry(tmp, asocs); 1;
8785 tmp = list_next_entry(tmp, asocs)) {
8786 /* Manually skip the head element. */
8787 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
8789 /* Wake up association. */
8790 __sctp_write_space(tmp);
8791 /* We've reached the end. */
8797 /* Do accounting for the sndbuf space.
8798 * Decrement the used sndbuf space of the corresponding association by the
8799 * data size which was just transmitted(freed).
8801 static void sctp_wfree(struct sk_buff *skb)
8803 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
8804 struct sctp_association *asoc = chunk->asoc;
8805 struct sock *sk = asoc->base.sk;
8807 sk_mem_uncharge(sk, skb->truesize);
8808 sk->sk_wmem_queued -= skb->truesize + sizeof(struct sctp_chunk);
8809 asoc->sndbuf_used -= skb->truesize + sizeof(struct sctp_chunk);
8810 WARN_ON(refcount_sub_and_test(sizeof(struct sctp_chunk),
8811 &sk->sk_wmem_alloc));
8814 struct sctp_shared_key *shkey = chunk->shkey;
8816 /* refcnt == 2 and !list_empty mean after this release, it's
8817 * not being used anywhere, and it's time to notify userland
8818 * that this shkey can be freed if it's been deactivated.
8820 if (shkey->deactivated && !list_empty(&shkey->key_list) &&
8821 refcount_read(&shkey->refcnt) == 2) {
8822 struct sctp_ulpevent *ev;
8824 ev = sctp_ulpevent_make_authkey(asoc, shkey->key_id,
8828 asoc->stream.si->enqueue_event(&asoc->ulpq, ev);
8830 sctp_auth_shkey_release(chunk->shkey);
8834 sctp_wake_up_waiters(sk, asoc);
8836 sctp_association_put(asoc);
8839 /* Do accounting for the receive space on the socket.
8840 * Accounting for the association is done in ulpevent.c
8841 * We set this as a destructor for the cloned data skbs so that
8842 * accounting is done at the correct time.
8844 void sctp_sock_rfree(struct sk_buff *skb)
8846 struct sock *sk = skb->sk;
8847 struct sctp_ulpevent *event = sctp_skb2event(skb);
8849 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
8852 * Mimic the behavior of sock_rfree
8854 sk_mem_uncharge(sk, event->rmem_len);
8858 /* Helper function to wait for space in the sndbuf. */
8859 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
8862 struct sock *sk = asoc->base.sk;
8863 long current_timeo = *timeo_p;
8867 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
8870 /* Increment the association's refcnt. */
8871 sctp_association_hold(asoc);
8873 /* Wait on the association specific sndbuf space. */
8875 prepare_to_wait_exclusive(&asoc->wait, &wait,
8876 TASK_INTERRUPTIBLE);
8877 if (asoc->base.dead)
8881 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING)
8883 if (signal_pending(current))
8884 goto do_interrupted;
8885 if ((int)msg_len <= sctp_wspace(asoc))
8888 /* Let another process have a go. Since we are going
8892 current_timeo = schedule_timeout(current_timeo);
8894 if (sk != asoc->base.sk)
8897 *timeo_p = current_timeo;
8901 finish_wait(&asoc->wait, &wait);
8903 /* Release the association's refcnt. */
8904 sctp_association_put(asoc);
8917 err = sock_intr_errno(*timeo_p);
8925 void sctp_data_ready(struct sock *sk)
8927 struct socket_wq *wq;
8930 wq = rcu_dereference(sk->sk_wq);
8931 if (skwq_has_sleeper(wq))
8932 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
8933 EPOLLRDNORM | EPOLLRDBAND);
8934 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
8938 /* If socket sndbuf has changed, wake up all per association waiters. */
8939 void sctp_write_space(struct sock *sk)
8941 struct sctp_association *asoc;
8943 /* Wake up the tasks in each wait queue. */
8944 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
8945 __sctp_write_space(asoc);
8949 /* Is there any sndbuf space available on the socket?
8951 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
8952 * associations on the same socket. For a UDP-style socket with
8953 * multiple associations, it is possible for it to be "unwriteable"
8954 * prematurely. I assume that this is acceptable because
8955 * a premature "unwriteable" is better than an accidental "writeable" which
8956 * would cause an unwanted block under certain circumstances. For the 1-1
8957 * UDP-style sockets or TCP-style sockets, this code should work.
8960 static bool sctp_writeable(struct sock *sk)
8962 return sk->sk_sndbuf > sk->sk_wmem_queued;
8965 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
8966 * returns immediately with EINPROGRESS.
8968 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
8970 struct sock *sk = asoc->base.sk;
8972 long current_timeo = *timeo_p;
8975 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
8977 /* Increment the association's refcnt. */
8978 sctp_association_hold(asoc);
8981 prepare_to_wait_exclusive(&asoc->wait, &wait,
8982 TASK_INTERRUPTIBLE);
8985 if (sk->sk_shutdown & RCV_SHUTDOWN)
8987 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
8990 if (signal_pending(current))
8991 goto do_interrupted;
8993 if (sctp_state(asoc, ESTABLISHED))
8996 /* Let another process have a go. Since we are going
9000 current_timeo = schedule_timeout(current_timeo);
9003 *timeo_p = current_timeo;
9007 finish_wait(&asoc->wait, &wait);
9009 /* Release the association's refcnt. */
9010 sctp_association_put(asoc);
9015 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
9018 err = -ECONNREFUSED;
9022 err = sock_intr_errno(*timeo_p);
9030 static int sctp_wait_for_accept(struct sock *sk, long timeo)
9032 struct sctp_endpoint *ep;
9036 ep = sctp_sk(sk)->ep;
9040 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
9041 TASK_INTERRUPTIBLE);
9043 if (list_empty(&ep->asocs)) {
9045 timeo = schedule_timeout(timeo);
9050 if (!sctp_sstate(sk, LISTENING))
9054 if (!list_empty(&ep->asocs))
9057 err = sock_intr_errno(timeo);
9058 if (signal_pending(current))
9066 finish_wait(sk_sleep(sk), &wait);
9071 static void sctp_wait_for_close(struct sock *sk, long timeout)
9076 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
9077 if (list_empty(&sctp_sk(sk)->ep->asocs))
9080 timeout = schedule_timeout(timeout);
9082 } while (!signal_pending(current) && timeout);
9084 finish_wait(sk_sleep(sk), &wait);
9087 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
9089 struct sk_buff *frag;
9094 /* Don't forget the fragments. */
9095 skb_walk_frags(skb, frag)
9096 sctp_skb_set_owner_r_frag(frag, sk);
9099 sctp_skb_set_owner_r(skb, sk);
9102 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
9103 struct sctp_association *asoc)
9105 struct inet_sock *inet = inet_sk(sk);
9106 struct inet_sock *newinet;
9107 struct sctp_sock *sp = sctp_sk(sk);
9108 struct sctp_endpoint *ep = sp->ep;
9110 newsk->sk_type = sk->sk_type;
9111 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
9112 newsk->sk_flags = sk->sk_flags;
9113 newsk->sk_tsflags = sk->sk_tsflags;
9114 newsk->sk_no_check_tx = sk->sk_no_check_tx;
9115 newsk->sk_no_check_rx = sk->sk_no_check_rx;
9116 newsk->sk_reuse = sk->sk_reuse;
9117 sctp_sk(newsk)->reuse = sp->reuse;
9119 newsk->sk_shutdown = sk->sk_shutdown;
9120 newsk->sk_destruct = sctp_destruct_sock;
9121 newsk->sk_family = sk->sk_family;
9122 newsk->sk_protocol = IPPROTO_SCTP;
9123 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
9124 newsk->sk_sndbuf = sk->sk_sndbuf;
9125 newsk->sk_rcvbuf = sk->sk_rcvbuf;
9126 newsk->sk_lingertime = sk->sk_lingertime;
9127 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
9128 newsk->sk_sndtimeo = sk->sk_sndtimeo;
9129 newsk->sk_rxhash = sk->sk_rxhash;
9131 newinet = inet_sk(newsk);
9133 /* Initialize sk's sport, dport, rcv_saddr and daddr for
9134 * getsockname() and getpeername()
9136 newinet->inet_sport = inet->inet_sport;
9137 newinet->inet_saddr = inet->inet_saddr;
9138 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
9139 newinet->inet_dport = htons(asoc->peer.port);
9140 newinet->pmtudisc = inet->pmtudisc;
9141 newinet->inet_id = asoc->next_tsn ^ jiffies;
9143 newinet->uc_ttl = inet->uc_ttl;
9144 newinet->mc_loop = 1;
9145 newinet->mc_ttl = 1;
9146 newinet->mc_index = 0;
9147 newinet->mc_list = NULL;
9149 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
9150 net_enable_timestamp();
9152 /* Set newsk security attributes from orginal sk and connection
9153 * security attribute from ep.
9155 security_sctp_sk_clone(ep, sk, newsk);
9158 static inline void sctp_copy_descendant(struct sock *sk_to,
9159 const struct sock *sk_from)
9161 int ancestor_size = sizeof(struct inet_sock) +
9162 sizeof(struct sctp_sock) -
9163 offsetof(struct sctp_sock, auto_asconf_list);
9165 if (sk_from->sk_family == PF_INET6)
9166 ancestor_size += sizeof(struct ipv6_pinfo);
9168 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
9171 /* Populate the fields of the newsk from the oldsk and migrate the assoc
9172 * and its messages to the newsk.
9174 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
9175 struct sctp_association *assoc,
9176 enum sctp_socket_type type)
9178 struct sctp_sock *oldsp = sctp_sk(oldsk);
9179 struct sctp_sock *newsp = sctp_sk(newsk);
9180 struct sctp_bind_bucket *pp; /* hash list port iterator */
9181 struct sctp_endpoint *newep = newsp->ep;
9182 struct sk_buff *skb, *tmp;
9183 struct sctp_ulpevent *event;
9184 struct sctp_bind_hashbucket *head;
9186 /* Migrate socket buffer sizes and all the socket level options to the
9189 newsk->sk_sndbuf = oldsk->sk_sndbuf;
9190 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
9191 /* Brute force copy old sctp opt. */
9192 sctp_copy_descendant(newsk, oldsk);
9194 /* Restore the ep value that was overwritten with the above structure
9200 /* Hook this new socket in to the bind_hash list. */
9201 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
9202 inet_sk(oldsk)->inet_num)];
9203 spin_lock_bh(&head->lock);
9204 pp = sctp_sk(oldsk)->bind_hash;
9205 sk_add_bind_node(newsk, &pp->owner);
9206 sctp_sk(newsk)->bind_hash = pp;
9207 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
9208 spin_unlock_bh(&head->lock);
9210 /* Copy the bind_addr list from the original endpoint to the new
9211 * endpoint so that we can handle restarts properly
9213 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
9214 &oldsp->ep->base.bind_addr, GFP_KERNEL);
9216 /* Move any messages in the old socket's receive queue that are for the
9217 * peeled off association to the new socket's receive queue.
9219 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
9220 event = sctp_skb2event(skb);
9221 if (event->asoc == assoc) {
9222 __skb_unlink(skb, &oldsk->sk_receive_queue);
9223 __skb_queue_tail(&newsk->sk_receive_queue, skb);
9224 sctp_skb_set_owner_r_frag(skb, newsk);
9228 /* Clean up any messages pending delivery due to partial
9229 * delivery. Three cases:
9230 * 1) No partial deliver; no work.
9231 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
9232 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
9234 skb_queue_head_init(&newsp->pd_lobby);
9235 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
9237 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
9238 struct sk_buff_head *queue;
9240 /* Decide which queue to move pd_lobby skbs to. */
9241 if (assoc->ulpq.pd_mode) {
9242 queue = &newsp->pd_lobby;
9244 queue = &newsk->sk_receive_queue;
9246 /* Walk through the pd_lobby, looking for skbs that
9247 * need moved to the new socket.
9249 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
9250 event = sctp_skb2event(skb);
9251 if (event->asoc == assoc) {
9252 __skb_unlink(skb, &oldsp->pd_lobby);
9253 __skb_queue_tail(queue, skb);
9254 sctp_skb_set_owner_r_frag(skb, newsk);
9258 /* Clear up any skbs waiting for the partial
9259 * delivery to finish.
9261 if (assoc->ulpq.pd_mode)
9262 sctp_clear_pd(oldsk, NULL);
9266 sctp_for_each_rx_skb(assoc, newsk, sctp_skb_set_owner_r_frag);
9268 /* Set the type of socket to indicate that it is peeled off from the
9269 * original UDP-style socket or created with the accept() call on a
9270 * TCP-style socket..
9274 /* Mark the new socket "in-use" by the user so that any packets
9275 * that may arrive on the association after we've moved it are
9276 * queued to the backlog. This prevents a potential race between
9277 * backlog processing on the old socket and new-packet processing
9278 * on the new socket.
9280 * The caller has just allocated newsk so we can guarantee that other
9281 * paths won't try to lock it and then oldsk.
9283 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
9284 sctp_for_each_tx_datachunk(assoc, sctp_clear_owner_w);
9285 sctp_assoc_migrate(assoc, newsk);
9286 sctp_for_each_tx_datachunk(assoc, sctp_set_owner_w);
9288 /* If the association on the newsk is already closed before accept()
9289 * is called, set RCV_SHUTDOWN flag.
9291 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP)) {
9292 inet_sk_set_state(newsk, SCTP_SS_CLOSED);
9293 newsk->sk_shutdown |= RCV_SHUTDOWN;
9295 inet_sk_set_state(newsk, SCTP_SS_ESTABLISHED);
9298 release_sock(newsk);
9302 /* This proto struct describes the ULP interface for SCTP. */
9303 struct proto sctp_prot = {
9305 .owner = THIS_MODULE,
9306 .close = sctp_close,
9307 .disconnect = sctp_disconnect,
9308 .accept = sctp_accept,
9309 .ioctl = sctp_ioctl,
9310 .init = sctp_init_sock,
9311 .destroy = sctp_destroy_sock,
9312 .shutdown = sctp_shutdown,
9313 .setsockopt = sctp_setsockopt,
9314 .getsockopt = sctp_getsockopt,
9315 .sendmsg = sctp_sendmsg,
9316 .recvmsg = sctp_recvmsg,
9318 .backlog_rcv = sctp_backlog_rcv,
9320 .unhash = sctp_unhash,
9321 .get_port = sctp_get_port,
9322 .obj_size = sizeof(struct sctp_sock),
9323 .useroffset = offsetof(struct sctp_sock, subscribe),
9324 .usersize = offsetof(struct sctp_sock, initmsg) -
9325 offsetof(struct sctp_sock, subscribe) +
9326 sizeof_field(struct sctp_sock, initmsg),
9327 .sysctl_mem = sysctl_sctp_mem,
9328 .sysctl_rmem = sysctl_sctp_rmem,
9329 .sysctl_wmem = sysctl_sctp_wmem,
9330 .memory_pressure = &sctp_memory_pressure,
9331 .enter_memory_pressure = sctp_enter_memory_pressure,
9332 .memory_allocated = &sctp_memory_allocated,
9333 .sockets_allocated = &sctp_sockets_allocated,
9336 #if IS_ENABLED(CONFIG_IPV6)
9338 #include <net/transp_v6.h>
9339 static void sctp_v6_destroy_sock(struct sock *sk)
9341 sctp_destroy_sock(sk);
9342 inet6_destroy_sock(sk);
9345 struct proto sctpv6_prot = {
9347 .owner = THIS_MODULE,
9348 .close = sctp_close,
9349 .disconnect = sctp_disconnect,
9350 .accept = sctp_accept,
9351 .ioctl = sctp_ioctl,
9352 .init = sctp_init_sock,
9353 .destroy = sctp_v6_destroy_sock,
9354 .shutdown = sctp_shutdown,
9355 .setsockopt = sctp_setsockopt,
9356 .getsockopt = sctp_getsockopt,
9357 .sendmsg = sctp_sendmsg,
9358 .recvmsg = sctp_recvmsg,
9360 .backlog_rcv = sctp_backlog_rcv,
9362 .unhash = sctp_unhash,
9363 .get_port = sctp_get_port,
9364 .obj_size = sizeof(struct sctp6_sock),
9365 .useroffset = offsetof(struct sctp6_sock, sctp.subscribe),
9366 .usersize = offsetof(struct sctp6_sock, sctp.initmsg) -
9367 offsetof(struct sctp6_sock, sctp.subscribe) +
9368 sizeof_field(struct sctp6_sock, sctp.initmsg),
9369 .sysctl_mem = sysctl_sctp_mem,
9370 .sysctl_rmem = sysctl_sctp_rmem,
9371 .sysctl_wmem = sysctl_sctp_wmem,
9372 .memory_pressure = &sctp_memory_pressure,
9373 .enter_memory_pressure = sctp_enter_memory_pressure,
9374 .memory_allocated = &sctp_memory_allocated,
9375 .sockets_allocated = &sctp_sockets_allocated,
9377 #endif /* IS_ENABLED(CONFIG_IPV6) */