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>
72 #include <net/route.h>
74 #include <net/inet_common.h>
75 #include <net/busy_poll.h>
77 #include <linux/socket.h> /* for sa_family_t */
78 #include <linux/export.h>
80 #include <net/sctp/sctp.h>
81 #include <net/sctp/sm.h>
83 /* Forward declarations for internal helper functions. */
84 static int sctp_writeable(struct sock *sk);
85 static void sctp_wfree(struct sk_buff *skb);
86 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
88 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p);
89 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
90 static int sctp_wait_for_accept(struct sock *sk, long timeo);
91 static void sctp_wait_for_close(struct sock *sk, long timeo);
92 static void sctp_destruct_sock(struct sock *sk);
93 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
94 union sctp_addr *addr, int len);
95 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
96 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
97 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
98 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
99 static int sctp_send_asconf(struct sctp_association *asoc,
100 struct sctp_chunk *chunk);
101 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
102 static int sctp_autobind(struct sock *sk);
103 static void sctp_sock_migrate(struct sock *, struct sock *,
104 struct sctp_association *, sctp_socket_type_t);
106 static unsigned long sctp_memory_pressure;
107 static atomic_long_t sctp_memory_allocated;
108 struct percpu_counter sctp_sockets_allocated;
110 static void sctp_enter_memory_pressure(struct sock *sk)
112 sctp_memory_pressure = 1;
116 /* Get the sndbuf space available at the time on the association. */
117 static inline int sctp_wspace(struct sctp_association *asoc)
121 if (asoc->ep->sndbuf_policy)
122 amt = asoc->sndbuf_used;
124 amt = sk_wmem_alloc_get(asoc->base.sk);
126 if (amt >= asoc->base.sk->sk_sndbuf) {
127 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
130 amt = sk_stream_wspace(asoc->base.sk);
135 amt = asoc->base.sk->sk_sndbuf - amt;
140 /* Increment the used sndbuf space count of the corresponding association by
141 * the size of the outgoing data chunk.
142 * Also, set the skb destructor for sndbuf accounting later.
144 * Since it is always 1-1 between chunk and skb, and also a new skb is always
145 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
146 * destructor in the data chunk skb for the purpose of the sndbuf space
149 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
151 struct sctp_association *asoc = chunk->asoc;
152 struct sock *sk = asoc->base.sk;
154 /* The sndbuf space is tracked per association. */
155 sctp_association_hold(asoc);
157 skb_set_owner_w(chunk->skb, sk);
159 chunk->skb->destructor = sctp_wfree;
160 /* Save the chunk pointer in skb for sctp_wfree to use later. */
161 skb_shinfo(chunk->skb)->destructor_arg = chunk;
163 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
164 sizeof(struct sk_buff) +
165 sizeof(struct sctp_chunk);
167 refcount_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
168 sk->sk_wmem_queued += chunk->skb->truesize;
169 sk_mem_charge(sk, chunk->skb->truesize);
172 /* Verify that this is a valid address. */
173 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
178 /* Verify basic sockaddr. */
179 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
183 /* Is this a valid SCTP address? */
184 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
187 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
193 /* Look up the association by its id. If this is not a UDP-style
194 * socket, the ID field is always ignored.
196 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
198 struct sctp_association *asoc = NULL;
200 /* If this is not a UDP-style socket, assoc id should be ignored. */
201 if (!sctp_style(sk, UDP)) {
202 /* Return NULL if the socket state is not ESTABLISHED. It
203 * could be a TCP-style listening socket or a socket which
204 * hasn't yet called connect() to establish an association.
206 if (!sctp_sstate(sk, ESTABLISHED) && !sctp_sstate(sk, CLOSING))
209 /* Get the first and the only association from the list. */
210 if (!list_empty(&sctp_sk(sk)->ep->asocs))
211 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
212 struct sctp_association, asocs);
216 /* Otherwise this is a UDP-style socket. */
217 if (!id || (id == (sctp_assoc_t)-1))
220 spin_lock_bh(&sctp_assocs_id_lock);
221 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
222 spin_unlock_bh(&sctp_assocs_id_lock);
224 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
230 /* Look up the transport from an address and an assoc id. If both address and
231 * id are specified, the associations matching the address and the id should be
234 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
235 struct sockaddr_storage *addr,
238 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
239 struct sctp_af *af = sctp_get_af_specific(addr->ss_family);
240 union sctp_addr *laddr = (union sctp_addr *)addr;
241 struct sctp_transport *transport;
243 if (!af || sctp_verify_addr(sk, laddr, af->sockaddr_len))
246 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
253 id_asoc = sctp_id2assoc(sk, id);
254 if (id_asoc && (id_asoc != addr_asoc))
257 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
258 (union sctp_addr *)addr);
263 /* API 3.1.2 bind() - UDP Style Syntax
264 * The syntax of bind() is,
266 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
268 * sd - the socket descriptor returned by socket().
269 * addr - the address structure (struct sockaddr_in or struct
270 * sockaddr_in6 [RFC 2553]),
271 * addr_len - the size of the address structure.
273 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
279 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
282 /* Disallow binding twice. */
283 if (!sctp_sk(sk)->ep->base.bind_addr.port)
284 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
294 static long sctp_get_port_local(struct sock *, union sctp_addr *);
296 /* Verify this is a valid sockaddr. */
297 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
298 union sctp_addr *addr, int len)
302 /* Check minimum size. */
303 if (len < sizeof (struct sockaddr))
306 /* V4 mapped address are really of AF_INET family */
307 if (addr->sa.sa_family == AF_INET6 &&
308 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
309 if (!opt->pf->af_supported(AF_INET, opt))
312 /* Does this PF support this AF? */
313 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
317 /* If we get this far, af is valid. */
318 af = sctp_get_af_specific(addr->sa.sa_family);
320 if (len < af->sockaddr_len)
326 /* Bind a local address either to an endpoint or to an association. */
327 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
329 struct net *net = sock_net(sk);
330 struct sctp_sock *sp = sctp_sk(sk);
331 struct sctp_endpoint *ep = sp->ep;
332 struct sctp_bind_addr *bp = &ep->base.bind_addr;
337 /* Common sockaddr verification. */
338 af = sctp_sockaddr_af(sp, addr, len);
340 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
341 __func__, sk, addr, len);
345 snum = ntohs(addr->v4.sin_port);
347 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
348 __func__, sk, &addr->sa, bp->port, snum, len);
350 /* PF specific bind() address verification. */
351 if (!sp->pf->bind_verify(sp, addr))
352 return -EADDRNOTAVAIL;
354 /* We must either be unbound, or bind to the same port.
355 * It's OK to allow 0 ports if we are already bound.
356 * We'll just inhert an already bound port in this case
361 else if (snum != bp->port) {
362 pr_debug("%s: new port %d doesn't match existing port "
363 "%d\n", __func__, snum, bp->port);
368 if (snum && snum < inet_prot_sock(net) &&
369 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
372 /* See if the address matches any of the addresses we may have
373 * already bound before checking against other endpoints.
375 if (sctp_bind_addr_match(bp, addr, sp))
378 /* Make sure we are allowed to bind here.
379 * The function sctp_get_port_local() does duplicate address
382 addr->v4.sin_port = htons(snum);
383 if ((ret = sctp_get_port_local(sk, addr))) {
387 /* Refresh ephemeral port. */
389 bp->port = inet_sk(sk)->inet_num;
391 /* Add the address to the bind address list.
392 * Use GFP_ATOMIC since BHs will be disabled.
394 ret = sctp_add_bind_addr(bp, addr, af->sockaddr_len,
395 SCTP_ADDR_SRC, GFP_ATOMIC);
397 /* Copy back into socket for getsockname() use. */
399 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
400 sp->pf->to_sk_saddr(addr, sk);
406 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
408 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
409 * at any one time. If a sender, after sending an ASCONF chunk, decides
410 * it needs to transfer another ASCONF Chunk, it MUST wait until the
411 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
412 * subsequent ASCONF. Note this restriction binds each side, so at any
413 * time two ASCONF may be in-transit on any given association (one sent
414 * from each endpoint).
416 static int sctp_send_asconf(struct sctp_association *asoc,
417 struct sctp_chunk *chunk)
419 struct net *net = sock_net(asoc->base.sk);
422 /* If there is an outstanding ASCONF chunk, queue it for later
425 if (asoc->addip_last_asconf) {
426 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
430 /* Hold the chunk until an ASCONF_ACK is received. */
431 sctp_chunk_hold(chunk);
432 retval = sctp_primitive_ASCONF(net, asoc, chunk);
434 sctp_chunk_free(chunk);
436 asoc->addip_last_asconf = chunk;
442 /* Add a list of addresses as bind addresses to local endpoint or
445 * Basically run through each address specified in the addrs/addrcnt
446 * array/length pair, determine if it is IPv6 or IPv4 and call
447 * sctp_do_bind() on it.
449 * If any of them fails, then the operation will be reversed and the
450 * ones that were added will be removed.
452 * Only sctp_setsockopt_bindx() is supposed to call this function.
454 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
459 struct sockaddr *sa_addr;
462 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
466 for (cnt = 0; cnt < addrcnt; cnt++) {
467 /* The list may contain either IPv4 or IPv6 address;
468 * determine the address length for walking thru the list.
471 af = sctp_get_af_specific(sa_addr->sa_family);
477 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
480 addr_buf += af->sockaddr_len;
484 /* Failed. Cleanup the ones that have been added */
486 sctp_bindx_rem(sk, addrs, cnt);
494 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
495 * associations that are part of the endpoint indicating that a list of local
496 * addresses are added to the endpoint.
498 * If any of the addresses is already in the bind address list of the
499 * association, we do not send the chunk for that association. But it will not
500 * affect other associations.
502 * Only sctp_setsockopt_bindx() is supposed to call this function.
504 static int sctp_send_asconf_add_ip(struct sock *sk,
505 struct sockaddr *addrs,
508 struct net *net = sock_net(sk);
509 struct sctp_sock *sp;
510 struct sctp_endpoint *ep;
511 struct sctp_association *asoc;
512 struct sctp_bind_addr *bp;
513 struct sctp_chunk *chunk;
514 struct sctp_sockaddr_entry *laddr;
515 union sctp_addr *addr;
516 union sctp_addr saveaddr;
523 if (!net->sctp.addip_enable)
529 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
530 __func__, sk, addrs, addrcnt);
532 list_for_each_entry(asoc, &ep->asocs, asocs) {
533 if (!asoc->peer.asconf_capable)
536 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
539 if (!sctp_state(asoc, ESTABLISHED))
542 /* Check if any address in the packed array of addresses is
543 * in the bind address list of the association. If so,
544 * do not send the asconf chunk to its peer, but continue with
545 * other associations.
548 for (i = 0; i < addrcnt; i++) {
550 af = sctp_get_af_specific(addr->v4.sin_family);
556 if (sctp_assoc_lookup_laddr(asoc, addr))
559 addr_buf += af->sockaddr_len;
564 /* Use the first valid address in bind addr list of
565 * association as Address Parameter of ASCONF CHUNK.
567 bp = &asoc->base.bind_addr;
568 p = bp->address_list.next;
569 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
570 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
571 addrcnt, SCTP_PARAM_ADD_IP);
577 /* Add the new addresses to the bind address list with
578 * use_as_src set to 0.
581 for (i = 0; i < addrcnt; i++) {
583 af = sctp_get_af_specific(addr->v4.sin_family);
584 memcpy(&saveaddr, addr, af->sockaddr_len);
585 retval = sctp_add_bind_addr(bp, &saveaddr,
587 SCTP_ADDR_NEW, GFP_ATOMIC);
588 addr_buf += af->sockaddr_len;
590 if (asoc->src_out_of_asoc_ok) {
591 struct sctp_transport *trans;
593 list_for_each_entry(trans,
594 &asoc->peer.transport_addr_list, transports) {
595 /* Clear the source and route cache */
596 sctp_transport_dst_release(trans);
597 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
598 2*asoc->pathmtu, 4380));
599 trans->ssthresh = asoc->peer.i.a_rwnd;
600 trans->rto = asoc->rto_initial;
601 sctp_max_rto(asoc, trans);
602 trans->rtt = trans->srtt = trans->rttvar = 0;
603 sctp_transport_route(trans, NULL,
604 sctp_sk(asoc->base.sk));
607 retval = sctp_send_asconf(asoc, chunk);
614 /* Remove a list of addresses from bind addresses list. Do not remove the
617 * Basically run through each address specified in the addrs/addrcnt
618 * array/length pair, determine if it is IPv6 or IPv4 and call
619 * sctp_del_bind() on it.
621 * If any of them fails, then the operation will be reversed and the
622 * ones that were removed will be added back.
624 * At least one address has to be left; if only one address is
625 * available, the operation will return -EBUSY.
627 * Only sctp_setsockopt_bindx() is supposed to call this function.
629 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
631 struct sctp_sock *sp = sctp_sk(sk);
632 struct sctp_endpoint *ep = sp->ep;
634 struct sctp_bind_addr *bp = &ep->base.bind_addr;
637 union sctp_addr *sa_addr;
640 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
641 __func__, sk, addrs, addrcnt);
644 for (cnt = 0; cnt < addrcnt; cnt++) {
645 /* If the bind address list is empty or if there is only one
646 * bind address, there is nothing more to be removed (we need
647 * at least one address here).
649 if (list_empty(&bp->address_list) ||
650 (sctp_list_single_entry(&bp->address_list))) {
656 af = sctp_get_af_specific(sa_addr->sa.sa_family);
662 if (!af->addr_valid(sa_addr, sp, NULL)) {
663 retval = -EADDRNOTAVAIL;
667 if (sa_addr->v4.sin_port &&
668 sa_addr->v4.sin_port != htons(bp->port)) {
673 if (!sa_addr->v4.sin_port)
674 sa_addr->v4.sin_port = htons(bp->port);
676 /* FIXME - There is probably a need to check if sk->sk_saddr and
677 * sk->sk_rcv_addr are currently set to one of the addresses to
678 * be removed. This is something which needs to be looked into
679 * when we are fixing the outstanding issues with multi-homing
680 * socket routing and failover schemes. Refer to comments in
681 * sctp_do_bind(). -daisy
683 retval = sctp_del_bind_addr(bp, sa_addr);
685 addr_buf += af->sockaddr_len;
688 /* Failed. Add the ones that has been removed back */
690 sctp_bindx_add(sk, addrs, cnt);
698 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
699 * the associations that are part of the endpoint indicating that a list of
700 * local addresses are removed from the endpoint.
702 * If any of the addresses is already in the bind address list of the
703 * association, we do not send the chunk for that association. But it will not
704 * affect other associations.
706 * Only sctp_setsockopt_bindx() is supposed to call this function.
708 static int sctp_send_asconf_del_ip(struct sock *sk,
709 struct sockaddr *addrs,
712 struct net *net = sock_net(sk);
713 struct sctp_sock *sp;
714 struct sctp_endpoint *ep;
715 struct sctp_association *asoc;
716 struct sctp_transport *transport;
717 struct sctp_bind_addr *bp;
718 struct sctp_chunk *chunk;
719 union sctp_addr *laddr;
722 struct sctp_sockaddr_entry *saddr;
728 if (!net->sctp.addip_enable)
734 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
735 __func__, sk, addrs, addrcnt);
737 list_for_each_entry(asoc, &ep->asocs, asocs) {
739 if (!asoc->peer.asconf_capable)
742 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
745 if (!sctp_state(asoc, ESTABLISHED))
748 /* Check if any address in the packed array of addresses is
749 * not present in the bind address list of the association.
750 * If so, do not send the asconf chunk to its peer, but
751 * continue with other associations.
754 for (i = 0; i < addrcnt; i++) {
756 af = sctp_get_af_specific(laddr->v4.sin_family);
762 if (!sctp_assoc_lookup_laddr(asoc, laddr))
765 addr_buf += af->sockaddr_len;
770 /* Find one address in the association's bind address list
771 * that is not in the packed array of addresses. This is to
772 * make sure that we do not delete all the addresses in the
775 bp = &asoc->base.bind_addr;
776 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
778 if ((laddr == NULL) && (addrcnt == 1)) {
779 if (asoc->asconf_addr_del_pending)
781 asoc->asconf_addr_del_pending =
782 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
783 if (asoc->asconf_addr_del_pending == NULL) {
787 asoc->asconf_addr_del_pending->sa.sa_family =
789 asoc->asconf_addr_del_pending->v4.sin_port =
791 if (addrs->sa_family == AF_INET) {
792 struct sockaddr_in *sin;
794 sin = (struct sockaddr_in *)addrs;
795 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
796 } else if (addrs->sa_family == AF_INET6) {
797 struct sockaddr_in6 *sin6;
799 sin6 = (struct sockaddr_in6 *)addrs;
800 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
803 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
804 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
805 asoc->asconf_addr_del_pending);
807 asoc->src_out_of_asoc_ok = 1;
815 /* We do not need RCU protection throughout this loop
816 * because this is done under a socket lock from the
819 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
827 /* Reset use_as_src flag for the addresses in the bind address
828 * list that are to be deleted.
831 for (i = 0; i < addrcnt; i++) {
833 af = sctp_get_af_specific(laddr->v4.sin_family);
834 list_for_each_entry(saddr, &bp->address_list, list) {
835 if (sctp_cmp_addr_exact(&saddr->a, laddr))
836 saddr->state = SCTP_ADDR_DEL;
838 addr_buf += af->sockaddr_len;
841 /* Update the route and saddr entries for all the transports
842 * as some of the addresses in the bind address list are
843 * about to be deleted and cannot be used as source addresses.
845 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
847 sctp_transport_dst_release(transport);
848 sctp_transport_route(transport, NULL,
849 sctp_sk(asoc->base.sk));
853 /* We don't need to transmit ASCONF */
855 retval = sctp_send_asconf(asoc, chunk);
861 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
862 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
864 struct sock *sk = sctp_opt2sk(sp);
865 union sctp_addr *addr;
868 /* It is safe to write port space in caller. */
870 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
871 af = sctp_get_af_specific(addr->sa.sa_family);
874 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
877 if (addrw->state == SCTP_ADDR_NEW)
878 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
880 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
883 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
886 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
889 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
890 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
893 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
894 * Section 3.1.2 for this usage.
896 * addrs is a pointer to an array of one or more socket addresses. Each
897 * address is contained in its appropriate structure (i.e. struct
898 * sockaddr_in or struct sockaddr_in6) the family of the address type
899 * must be used to distinguish the address length (note that this
900 * representation is termed a "packed array" of addresses). The caller
901 * specifies the number of addresses in the array with addrcnt.
903 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
904 * -1, and sets errno to the appropriate error code.
906 * For SCTP, the port given in each socket address must be the same, or
907 * sctp_bindx() will fail, setting errno to EINVAL.
909 * The flags parameter is formed from the bitwise OR of zero or more of
910 * the following currently defined flags:
912 * SCTP_BINDX_ADD_ADDR
914 * SCTP_BINDX_REM_ADDR
916 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
917 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
918 * addresses from the association. The two flags are mutually exclusive;
919 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
920 * not remove all addresses from an association; sctp_bindx() will
921 * reject such an attempt with EINVAL.
923 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
924 * additional addresses with an endpoint after calling bind(). Or use
925 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
926 * socket is associated with so that no new association accepted will be
927 * associated with those addresses. If the endpoint supports dynamic
928 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
929 * endpoint to send the appropriate message to the peer to change the
930 * peers address lists.
932 * Adding and removing addresses from a connected association is
933 * optional functionality. Implementations that do not support this
934 * functionality should return EOPNOTSUPP.
936 * Basically do nothing but copying the addresses from user to kernel
937 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
938 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
941 * We don't use copy_from_user() for optimization: we first do the
942 * sanity checks (buffer size -fast- and access check-healthy
943 * pointer); if all of those succeed, then we can alloc the memory
944 * (expensive operation) needed to copy the data to kernel. Then we do
945 * the copying without checking the user space area
946 * (__copy_from_user()).
948 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
951 * sk The sk of the socket
952 * addrs The pointer to the addresses in user land
953 * addrssize Size of the addrs buffer
954 * op Operation to perform (add or remove, see the flags of
957 * Returns 0 if ok, <0 errno code on error.
959 static int sctp_setsockopt_bindx(struct sock *sk,
960 struct sockaddr __user *addrs,
961 int addrs_size, int op)
963 struct sockaddr *kaddrs;
967 struct sockaddr *sa_addr;
971 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
972 __func__, sk, addrs, addrs_size, op);
974 if (unlikely(addrs_size <= 0))
977 /* Check the user passed a healthy pointer. */
978 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
981 /* Alloc space for the address array in kernel memory. */
982 kaddrs = kmalloc(addrs_size, GFP_USER | __GFP_NOWARN);
983 if (unlikely(!kaddrs))
986 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
991 /* Walk through the addrs buffer and count the number of addresses. */
993 while (walk_size < addrs_size) {
994 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1000 af = sctp_get_af_specific(sa_addr->sa_family);
1002 /* If the address family is not supported or if this address
1003 * causes the address buffer to overflow return EINVAL.
1005 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1010 addr_buf += af->sockaddr_len;
1011 walk_size += af->sockaddr_len;
1016 case SCTP_BINDX_ADD_ADDR:
1017 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1020 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1023 case SCTP_BINDX_REM_ADDR:
1024 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1027 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1041 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1043 * Common routine for handling connect() and sctp_connectx().
1044 * Connect will come in with just a single address.
1046 static int __sctp_connect(struct sock *sk,
1047 struct sockaddr *kaddrs,
1049 sctp_assoc_t *assoc_id)
1051 struct net *net = sock_net(sk);
1052 struct sctp_sock *sp;
1053 struct sctp_endpoint *ep;
1054 struct sctp_association *asoc = NULL;
1055 struct sctp_association *asoc2;
1056 struct sctp_transport *transport;
1058 enum sctp_scope scope;
1063 union sctp_addr *sa_addr = NULL;
1065 unsigned short port;
1066 unsigned int f_flags = 0;
1071 /* connect() cannot be done on a socket that is already in ESTABLISHED
1072 * state - UDP-style peeled off socket or a TCP-style socket that
1073 * is already connected.
1074 * It cannot be done even on a TCP-style listening socket.
1076 if (sctp_sstate(sk, ESTABLISHED) || sctp_sstate(sk, CLOSING) ||
1077 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1082 /* Walk through the addrs buffer and count the number of addresses. */
1084 while (walk_size < addrs_size) {
1087 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1093 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1095 /* If the address family is not supported or if this address
1096 * causes the address buffer to overflow return EINVAL.
1098 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1103 port = ntohs(sa_addr->v4.sin_port);
1105 /* Save current address so we can work with it */
1106 memcpy(&to, sa_addr, af->sockaddr_len);
1108 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1112 /* Make sure the destination port is correctly set
1115 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1120 /* Check if there already is a matching association on the
1121 * endpoint (other than the one created here).
1123 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1124 if (asoc2 && asoc2 != asoc) {
1125 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1132 /* If we could not find a matching association on the endpoint,
1133 * make sure that there is no peeled-off association matching
1134 * the peer address even on another socket.
1136 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1137 err = -EADDRNOTAVAIL;
1142 /* If a bind() or sctp_bindx() is not called prior to
1143 * an sctp_connectx() call, the system picks an
1144 * ephemeral port and will choose an address set
1145 * equivalent to binding with a wildcard address.
1147 if (!ep->base.bind_addr.port) {
1148 if (sctp_autobind(sk)) {
1154 * If an unprivileged user inherits a 1-many
1155 * style socket with open associations on a
1156 * privileged port, it MAY be permitted to
1157 * accept new associations, but it SHOULD NOT
1158 * be permitted to open new associations.
1160 if (ep->base.bind_addr.port <
1161 inet_prot_sock(net) &&
1162 !ns_capable(net->user_ns,
1163 CAP_NET_BIND_SERVICE)) {
1169 scope = sctp_scope(&to);
1170 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1176 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1184 /* Prime the peer's transport structures. */
1185 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1193 addr_buf += af->sockaddr_len;
1194 walk_size += af->sockaddr_len;
1197 /* In case the user of sctp_connectx() wants an association
1198 * id back, assign one now.
1201 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1206 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1211 /* Initialize sk's dport and daddr for getpeername() */
1212 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1213 sp->pf->to_sk_daddr(sa_addr, sk);
1216 /* in-kernel sockets don't generally have a file allocated to them
1217 * if all they do is call sock_create_kern().
1219 if (sk->sk_socket->file)
1220 f_flags = sk->sk_socket->file->f_flags;
1222 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1225 *assoc_id = asoc->assoc_id;
1226 err = sctp_wait_for_connect(asoc, &timeo);
1227 /* Note: the asoc may be freed after the return of
1228 * sctp_wait_for_connect.
1231 /* Don't free association on exit. */
1235 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1236 __func__, asoc, kaddrs, err);
1239 /* sctp_primitive_ASSOCIATE may have added this association
1240 * To the hash table, try to unhash it, just in case, its a noop
1241 * if it wasn't hashed so we're safe
1243 sctp_association_free(asoc);
1248 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1251 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1252 * sctp_assoc_t *asoc);
1254 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1255 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1256 * or IPv6 addresses.
1258 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1259 * Section 3.1.2 for this usage.
1261 * addrs is a pointer to an array of one or more socket addresses. Each
1262 * address is contained in its appropriate structure (i.e. struct
1263 * sockaddr_in or struct sockaddr_in6) the family of the address type
1264 * must be used to distengish the address length (note that this
1265 * representation is termed a "packed array" of addresses). The caller
1266 * specifies the number of addresses in the array with addrcnt.
1268 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1269 * the association id of the new association. On failure, sctp_connectx()
1270 * returns -1, and sets errno to the appropriate error code. The assoc_id
1271 * is not touched by the kernel.
1273 * For SCTP, the port given in each socket address must be the same, or
1274 * sctp_connectx() will fail, setting errno to EINVAL.
1276 * An application can use sctp_connectx to initiate an association with
1277 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1278 * allows a caller to specify multiple addresses at which a peer can be
1279 * reached. The way the SCTP stack uses the list of addresses to set up
1280 * the association is implementation dependent. This function only
1281 * specifies that the stack will try to make use of all the addresses in
1282 * the list when needed.
1284 * Note that the list of addresses passed in is only used for setting up
1285 * the association. It does not necessarily equal the set of addresses
1286 * the peer uses for the resulting association. If the caller wants to
1287 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1288 * retrieve them after the association has been set up.
1290 * Basically do nothing but copying the addresses from user to kernel
1291 * land and invoking either sctp_connectx(). This is used for tunneling
1292 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1294 * We don't use copy_from_user() for optimization: we first do the
1295 * sanity checks (buffer size -fast- and access check-healthy
1296 * pointer); if all of those succeed, then we can alloc the memory
1297 * (expensive operation) needed to copy the data to kernel. Then we do
1298 * the copying without checking the user space area
1299 * (__copy_from_user()).
1301 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1304 * sk The sk of the socket
1305 * addrs The pointer to the addresses in user land
1306 * addrssize Size of the addrs buffer
1308 * Returns >=0 if ok, <0 errno code on error.
1310 static int __sctp_setsockopt_connectx(struct sock *sk,
1311 struct sockaddr __user *addrs,
1313 sctp_assoc_t *assoc_id)
1315 struct sockaddr *kaddrs;
1316 gfp_t gfp = GFP_KERNEL;
1319 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1320 __func__, sk, addrs, addrs_size);
1322 if (unlikely(addrs_size <= 0))
1325 /* Check the user passed a healthy pointer. */
1326 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1329 /* Alloc space for the address array in kernel memory. */
1330 if (sk->sk_socket->file)
1331 gfp = GFP_USER | __GFP_NOWARN;
1332 kaddrs = kmalloc(addrs_size, gfp);
1333 if (unlikely(!kaddrs))
1336 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1339 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1348 * This is an older interface. It's kept for backward compatibility
1349 * to the option that doesn't provide association id.
1351 static int sctp_setsockopt_connectx_old(struct sock *sk,
1352 struct sockaddr __user *addrs,
1355 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1359 * New interface for the API. The since the API is done with a socket
1360 * option, to make it simple we feed back the association id is as a return
1361 * indication to the call. Error is always negative and association id is
1364 static int sctp_setsockopt_connectx(struct sock *sk,
1365 struct sockaddr __user *addrs,
1368 sctp_assoc_t assoc_id = 0;
1371 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1380 * New (hopefully final) interface for the API.
1381 * We use the sctp_getaddrs_old structure so that use-space library
1382 * can avoid any unnecessary allocations. The only different part
1383 * is that we store the actual length of the address buffer into the
1384 * addrs_num structure member. That way we can re-use the existing
1387 #ifdef CONFIG_COMPAT
1388 struct compat_sctp_getaddrs_old {
1389 sctp_assoc_t assoc_id;
1391 compat_uptr_t addrs; /* struct sockaddr * */
1395 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1396 char __user *optval,
1399 struct sctp_getaddrs_old param;
1400 sctp_assoc_t assoc_id = 0;
1403 #ifdef CONFIG_COMPAT
1404 if (in_compat_syscall()) {
1405 struct compat_sctp_getaddrs_old param32;
1407 if (len < sizeof(param32))
1409 if (copy_from_user(¶m32, optval, sizeof(param32)))
1412 param.assoc_id = param32.assoc_id;
1413 param.addr_num = param32.addr_num;
1414 param.addrs = compat_ptr(param32.addrs);
1418 if (len < sizeof(param))
1420 if (copy_from_user(¶m, optval, sizeof(param)))
1424 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1425 param.addrs, param.addr_num,
1427 if (err == 0 || err == -EINPROGRESS) {
1428 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1430 if (put_user(sizeof(assoc_id), optlen))
1437 /* API 3.1.4 close() - UDP Style Syntax
1438 * Applications use close() to perform graceful shutdown (as described in
1439 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1440 * by a UDP-style socket.
1444 * ret = close(int sd);
1446 * sd - the socket descriptor of the associations to be closed.
1448 * To gracefully shutdown a specific association represented by the
1449 * UDP-style socket, an application should use the sendmsg() call,
1450 * passing no user data, but including the appropriate flag in the
1451 * ancillary data (see Section xxxx).
1453 * If sd in the close() call is a branched-off socket representing only
1454 * one association, the shutdown is performed on that association only.
1456 * 4.1.6 close() - TCP Style Syntax
1458 * Applications use close() to gracefully close down an association.
1462 * int close(int sd);
1464 * sd - the socket descriptor of the association to be closed.
1466 * After an application calls close() on a socket descriptor, no further
1467 * socket operations will succeed on that descriptor.
1469 * API 7.1.4 SO_LINGER
1471 * An application using the TCP-style socket can use this option to
1472 * perform the SCTP ABORT primitive. The linger option structure is:
1475 * int l_onoff; // option on/off
1476 * int l_linger; // linger time
1479 * To enable the option, set l_onoff to 1. If the l_linger value is set
1480 * to 0, calling close() is the same as the ABORT primitive. If the
1481 * value is set to a negative value, the setsockopt() call will return
1482 * an error. If the value is set to a positive value linger_time, the
1483 * close() can be blocked for at most linger_time ms. If the graceful
1484 * shutdown phase does not finish during this period, close() will
1485 * return but the graceful shutdown phase continues in the system.
1487 static void sctp_close(struct sock *sk, long timeout)
1489 struct net *net = sock_net(sk);
1490 struct sctp_endpoint *ep;
1491 struct sctp_association *asoc;
1492 struct list_head *pos, *temp;
1493 unsigned int data_was_unread;
1495 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1497 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
1498 sk->sk_shutdown = SHUTDOWN_MASK;
1499 sk->sk_state = SCTP_SS_CLOSING;
1501 ep = sctp_sk(sk)->ep;
1503 /* Clean up any skbs sitting on the receive queue. */
1504 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1505 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1507 /* Walk all associations on an endpoint. */
1508 list_for_each_safe(pos, temp, &ep->asocs) {
1509 asoc = list_entry(pos, struct sctp_association, asocs);
1511 if (sctp_style(sk, TCP)) {
1512 /* A closed association can still be in the list if
1513 * it belongs to a TCP-style listening socket that is
1514 * not yet accepted. If so, free it. If not, send an
1515 * ABORT or SHUTDOWN based on the linger options.
1517 if (sctp_state(asoc, CLOSED)) {
1518 sctp_association_free(asoc);
1523 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1524 !skb_queue_empty(&asoc->ulpq.reasm) ||
1525 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1526 struct sctp_chunk *chunk;
1528 chunk = sctp_make_abort_user(asoc, NULL, 0);
1529 sctp_primitive_ABORT(net, asoc, chunk);
1531 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1534 /* On a TCP-style socket, block for at most linger_time if set. */
1535 if (sctp_style(sk, TCP) && timeout)
1536 sctp_wait_for_close(sk, timeout);
1538 /* This will run the backlog queue. */
1541 /* Supposedly, no process has access to the socket, but
1542 * the net layers still may.
1543 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1544 * held and that should be grabbed before socket lock.
1546 spin_lock_bh(&net->sctp.addr_wq_lock);
1547 bh_lock_sock_nested(sk);
1549 /* Hold the sock, since sk_common_release() will put sock_put()
1550 * and we have just a little more cleanup.
1553 sk_common_release(sk);
1556 spin_unlock_bh(&net->sctp.addr_wq_lock);
1560 SCTP_DBG_OBJCNT_DEC(sock);
1563 /* Handle EPIPE error. */
1564 static int sctp_error(struct sock *sk, int flags, int err)
1567 err = sock_error(sk) ? : -EPIPE;
1568 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1569 send_sig(SIGPIPE, current, 0);
1573 /* API 3.1.3 sendmsg() - UDP Style Syntax
1575 * An application uses sendmsg() and recvmsg() calls to transmit data to
1576 * and receive data from its peer.
1578 * ssize_t sendmsg(int socket, const struct msghdr *message,
1581 * socket - the socket descriptor of the endpoint.
1582 * message - pointer to the msghdr structure which contains a single
1583 * user message and possibly some ancillary data.
1585 * See Section 5 for complete description of the data
1588 * flags - flags sent or received with the user message, see Section
1589 * 5 for complete description of the flags.
1591 * Note: This function could use a rewrite especially when explicit
1592 * connect support comes in.
1594 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1596 static int sctp_msghdr_parse(const struct msghdr *msg,
1597 struct sctp_cmsgs *cmsgs);
1599 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
1601 struct net *net = sock_net(sk);
1602 struct sctp_sock *sp;
1603 struct sctp_endpoint *ep;
1604 struct sctp_association *new_asoc = NULL, *asoc = NULL;
1605 struct sctp_transport *transport, *chunk_tp;
1606 struct sctp_chunk *chunk;
1608 struct sockaddr *msg_name = NULL;
1609 struct sctp_sndrcvinfo default_sinfo;
1610 struct sctp_sndrcvinfo *sinfo;
1611 struct sctp_initmsg *sinit;
1612 sctp_assoc_t associd = 0;
1613 struct sctp_cmsgs cmsgs = { NULL };
1614 enum sctp_scope scope;
1615 bool fill_sinfo_ttl = false, wait_connect = false;
1616 struct sctp_datamsg *datamsg;
1617 int msg_flags = msg->msg_flags;
1618 __u16 sinfo_flags = 0;
1626 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1629 /* We cannot send a message over a TCP-style listening socket. */
1630 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1635 /* Parse out the SCTP CMSGs. */
1636 err = sctp_msghdr_parse(msg, &cmsgs);
1638 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1642 /* Fetch the destination address for this packet. This
1643 * address only selects the association--it is not necessarily
1644 * the address we will send to.
1645 * For a peeled-off socket, msg_name is ignored.
1647 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1648 int msg_namelen = msg->msg_namelen;
1650 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1655 if (msg_namelen > sizeof(to))
1656 msg_namelen = sizeof(to);
1657 memcpy(&to, msg->msg_name, msg_namelen);
1658 msg_name = msg->msg_name;
1662 if (cmsgs.sinfo != NULL) {
1663 memset(&default_sinfo, 0, sizeof(default_sinfo));
1664 default_sinfo.sinfo_stream = cmsgs.sinfo->snd_sid;
1665 default_sinfo.sinfo_flags = cmsgs.sinfo->snd_flags;
1666 default_sinfo.sinfo_ppid = cmsgs.sinfo->snd_ppid;
1667 default_sinfo.sinfo_context = cmsgs.sinfo->snd_context;
1668 default_sinfo.sinfo_assoc_id = cmsgs.sinfo->snd_assoc_id;
1670 sinfo = &default_sinfo;
1671 fill_sinfo_ttl = true;
1673 sinfo = cmsgs.srinfo;
1675 /* Did the user specify SNDINFO/SNDRCVINFO? */
1677 sinfo_flags = sinfo->sinfo_flags;
1678 associd = sinfo->sinfo_assoc_id;
1681 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1682 msg_len, sinfo_flags);
1684 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1685 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1690 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1691 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1692 * If SCTP_ABORT is set, the message length could be non zero with
1693 * the msg_iov set to the user abort reason.
1695 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1696 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1701 /* If SCTP_ADDR_OVER is set, there must be an address
1702 * specified in msg_name.
1704 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1711 pr_debug("%s: about to look up association\n", __func__);
1715 /* If a msg_name has been specified, assume this is to be used. */
1717 /* Look for a matching association on the endpoint. */
1718 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1720 /* If we could not find a matching association on the
1721 * endpoint, make sure that it is not a TCP-style
1722 * socket that already has an association or there is
1723 * no peeled-off association on another socket.
1726 ((sctp_style(sk, TCP) &&
1727 (sctp_sstate(sk, ESTABLISHED) ||
1728 sctp_sstate(sk, CLOSING))) ||
1729 sctp_endpoint_is_peeled_off(ep, &to))) {
1730 err = -EADDRNOTAVAIL;
1734 asoc = sctp_id2assoc(sk, associd);
1742 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1744 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1745 * socket that has an association in CLOSED state. This can
1746 * happen when an accepted socket has an association that is
1749 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1754 if (sinfo_flags & SCTP_EOF) {
1755 pr_debug("%s: shutting down association:%p\n",
1758 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1762 if (sinfo_flags & SCTP_ABORT) {
1764 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1770 pr_debug("%s: aborting association:%p\n",
1773 sctp_primitive_ABORT(net, asoc, chunk);
1779 /* Do we need to create the association? */
1781 pr_debug("%s: there is no association yet\n", __func__);
1783 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1788 /* Check for invalid stream against the stream counts,
1789 * either the default or the user specified stream counts.
1792 if (!sinit || !sinit->sinit_num_ostreams) {
1793 /* Check against the defaults. */
1794 if (sinfo->sinfo_stream >=
1795 sp->initmsg.sinit_num_ostreams) {
1800 /* Check against the requested. */
1801 if (sinfo->sinfo_stream >=
1802 sinit->sinit_num_ostreams) {
1810 * API 3.1.2 bind() - UDP Style Syntax
1811 * If a bind() or sctp_bindx() is not called prior to a
1812 * sendmsg() call that initiates a new association, the
1813 * system picks an ephemeral port and will choose an address
1814 * set equivalent to binding with a wildcard address.
1816 if (!ep->base.bind_addr.port) {
1817 if (sctp_autobind(sk)) {
1823 * If an unprivileged user inherits a one-to-many
1824 * style socket with open associations on a privileged
1825 * port, it MAY be permitted to accept new associations,
1826 * but it SHOULD NOT be permitted to open new
1829 if (ep->base.bind_addr.port < inet_prot_sock(net) &&
1830 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1836 scope = sctp_scope(&to);
1837 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1843 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1849 /* If the SCTP_INIT ancillary data is specified, set all
1850 * the association init values accordingly.
1853 if (sinit->sinit_num_ostreams) {
1854 asoc->c.sinit_num_ostreams =
1855 sinit->sinit_num_ostreams;
1857 if (sinit->sinit_max_instreams) {
1858 asoc->c.sinit_max_instreams =
1859 sinit->sinit_max_instreams;
1861 if (sinit->sinit_max_attempts) {
1862 asoc->max_init_attempts
1863 = sinit->sinit_max_attempts;
1865 if (sinit->sinit_max_init_timeo) {
1866 asoc->max_init_timeo =
1867 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1871 /* Prime the peer's transport structures. */
1872 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1879 /* ASSERT: we have a valid association at this point. */
1880 pr_debug("%s: we have a valid association\n", __func__);
1883 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1884 * one with some defaults.
1886 memset(&default_sinfo, 0, sizeof(default_sinfo));
1887 default_sinfo.sinfo_stream = asoc->default_stream;
1888 default_sinfo.sinfo_flags = asoc->default_flags;
1889 default_sinfo.sinfo_ppid = asoc->default_ppid;
1890 default_sinfo.sinfo_context = asoc->default_context;
1891 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1892 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1894 sinfo = &default_sinfo;
1895 } else if (fill_sinfo_ttl) {
1896 /* In case SNDINFO was specified, we still need to fill
1897 * it with a default ttl from the assoc here.
1899 sinfo->sinfo_timetolive = asoc->default_timetolive;
1902 /* API 7.1.7, the sndbuf size per association bounds the
1903 * maximum size of data that can be sent in a single send call.
1905 if (msg_len > sk->sk_sndbuf) {
1910 if (asoc->pmtu_pending)
1911 sctp_assoc_pending_pmtu(asoc);
1913 /* If fragmentation is disabled and the message length exceeds the
1914 * association fragmentation point, return EMSGSIZE. The I-D
1915 * does not specify what this error is, but this looks like
1918 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1923 /* Check for invalid stream. */
1924 if (sinfo->sinfo_stream >= asoc->stream.outcnt) {
1929 if (sctp_wspace(asoc) < msg_len)
1930 sctp_prsctp_prune(asoc, sinfo, msg_len - sctp_wspace(asoc));
1932 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1933 if (!sctp_wspace(asoc)) {
1934 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1939 /* If an address is passed with the sendto/sendmsg call, it is used
1940 * to override the primary destination address in the TCP model, or
1941 * when SCTP_ADDR_OVER flag is set in the UDP model.
1943 if ((sctp_style(sk, TCP) && msg_name) ||
1944 (sinfo_flags & SCTP_ADDR_OVER)) {
1945 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1953 /* Auto-connect, if we aren't connected already. */
1954 if (sctp_state(asoc, CLOSED)) {
1955 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1959 wait_connect = true;
1960 pr_debug("%s: we associated primitively\n", __func__);
1963 /* Break the message into multiple chunks of maximum size. */
1964 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
1965 if (IS_ERR(datamsg)) {
1966 err = PTR_ERR(datamsg);
1969 asoc->force_delay = !!(msg->msg_flags & MSG_MORE);
1971 /* Now send the (possibly) fragmented message. */
1972 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1973 sctp_chunk_hold(chunk);
1975 /* Do accounting for the write space. */
1976 sctp_set_owner_w(chunk);
1978 chunk->transport = chunk_tp;
1981 /* Send it to the lower layers. Note: all chunks
1982 * must either fail or succeed. The lower layer
1983 * works that way today. Keep it that way or this
1986 err = sctp_primitive_SEND(net, asoc, datamsg);
1987 /* Did the lower layer accept the chunk? */
1989 sctp_datamsg_free(datamsg);
1993 pr_debug("%s: we sent primitively\n", __func__);
1995 sctp_datamsg_put(datamsg);
1998 if (unlikely(wait_connect)) {
1999 timeo = sock_sndtimeo(sk, msg_flags & MSG_DONTWAIT);
2000 sctp_wait_for_connect(asoc, &timeo);
2003 /* If we are already past ASSOCIATE, the lower
2004 * layers are responsible for association cleanup.
2010 sctp_association_free(asoc);
2015 return sctp_error(sk, msg_flags, err);
2022 err = sock_error(sk);
2032 /* This is an extended version of skb_pull() that removes the data from the
2033 * start of a skb even when data is spread across the list of skb's in the
2034 * frag_list. len specifies the total amount of data that needs to be removed.
2035 * when 'len' bytes could be removed from the skb, it returns 0.
2036 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2037 * could not be removed.
2039 static int sctp_skb_pull(struct sk_buff *skb, int len)
2041 struct sk_buff *list;
2042 int skb_len = skb_headlen(skb);
2045 if (len <= skb_len) {
2046 __skb_pull(skb, len);
2050 __skb_pull(skb, skb_len);
2052 skb_walk_frags(skb, list) {
2053 rlen = sctp_skb_pull(list, len);
2054 skb->len -= (len-rlen);
2055 skb->data_len -= (len-rlen);
2066 /* API 3.1.3 recvmsg() - UDP Style Syntax
2068 * ssize_t recvmsg(int socket, struct msghdr *message,
2071 * socket - the socket descriptor of the endpoint.
2072 * message - pointer to the msghdr structure which contains a single
2073 * user message and possibly some ancillary data.
2075 * See Section 5 for complete description of the data
2078 * flags - flags sent or received with the user message, see Section
2079 * 5 for complete description of the flags.
2081 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2082 int noblock, int flags, int *addr_len)
2084 struct sctp_ulpevent *event = NULL;
2085 struct sctp_sock *sp = sctp_sk(sk);
2086 struct sk_buff *skb, *head_skb;
2091 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2092 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2097 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED) &&
2098 !sctp_sstate(sk, CLOSING) && !sctp_sstate(sk, CLOSED)) {
2103 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2107 /* Get the total length of the skb including any skb's in the
2116 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2118 event = sctp_skb2event(skb);
2123 if (event->chunk && event->chunk->head_skb)
2124 head_skb = event->chunk->head_skb;
2127 sock_recv_ts_and_drops(msg, sk, head_skb);
2128 if (sctp_ulpevent_is_notification(event)) {
2129 msg->msg_flags |= MSG_NOTIFICATION;
2130 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2132 sp->pf->skb_msgname(head_skb, msg->msg_name, addr_len);
2135 /* Check if we allow SCTP_NXTINFO. */
2136 if (sp->recvnxtinfo)
2137 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2138 /* Check if we allow SCTP_RCVINFO. */
2139 if (sp->recvrcvinfo)
2140 sctp_ulpevent_read_rcvinfo(event, msg);
2141 /* Check if we allow SCTP_SNDRCVINFO. */
2142 if (sp->subscribe.sctp_data_io_event)
2143 sctp_ulpevent_read_sndrcvinfo(event, msg);
2147 /* If skb's length exceeds the user's buffer, update the skb and
2148 * push it back to the receive_queue so that the next call to
2149 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2151 if (skb_len > copied) {
2152 msg->msg_flags &= ~MSG_EOR;
2153 if (flags & MSG_PEEK)
2155 sctp_skb_pull(skb, copied);
2156 skb_queue_head(&sk->sk_receive_queue, skb);
2158 /* When only partial message is copied to the user, increase
2159 * rwnd by that amount. If all the data in the skb is read,
2160 * rwnd is updated when the event is freed.
2162 if (!sctp_ulpevent_is_notification(event))
2163 sctp_assoc_rwnd_increase(event->asoc, copied);
2165 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2166 (event->msg_flags & MSG_EOR))
2167 msg->msg_flags |= MSG_EOR;
2169 msg->msg_flags &= ~MSG_EOR;
2172 if (flags & MSG_PEEK) {
2173 /* Release the skb reference acquired after peeking the skb in
2174 * sctp_skb_recv_datagram().
2178 /* Free the event which includes releasing the reference to
2179 * the owner of the skb, freeing the skb and updating the
2182 sctp_ulpevent_free(event);
2189 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2191 * This option is a on/off flag. If enabled no SCTP message
2192 * fragmentation will be performed. Instead if a message being sent
2193 * exceeds the current PMTU size, the message will NOT be sent and
2194 * instead a error will be indicated to the user.
2196 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2197 char __user *optval,
2198 unsigned int optlen)
2202 if (optlen < sizeof(int))
2205 if (get_user(val, (int __user *)optval))
2208 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2213 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2214 unsigned int optlen)
2216 struct sctp_association *asoc;
2217 struct sctp_ulpevent *event;
2219 if (optlen > sizeof(struct sctp_event_subscribe))
2221 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2224 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2225 * if there is no data to be sent or retransmit, the stack will
2226 * immediately send up this notification.
2228 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2229 &sctp_sk(sk)->subscribe)) {
2230 asoc = sctp_id2assoc(sk, 0);
2232 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2233 event = sctp_ulpevent_make_sender_dry_event(asoc,
2238 sctp_ulpq_tail_event(&asoc->ulpq, event);
2245 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2247 * This socket option is applicable to the UDP-style socket only. When
2248 * set it will cause associations that are idle for more than the
2249 * specified number of seconds to automatically close. An association
2250 * being idle is defined an association that has NOT sent or received
2251 * user data. The special value of '0' indicates that no automatic
2252 * close of any associations should be performed. The option expects an
2253 * integer defining the number of seconds of idle time before an
2254 * association is closed.
2256 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2257 unsigned int optlen)
2259 struct sctp_sock *sp = sctp_sk(sk);
2260 struct net *net = sock_net(sk);
2262 /* Applicable to UDP-style socket only */
2263 if (sctp_style(sk, TCP))
2265 if (optlen != sizeof(int))
2267 if (copy_from_user(&sp->autoclose, optval, optlen))
2270 if (sp->autoclose > net->sctp.max_autoclose)
2271 sp->autoclose = net->sctp.max_autoclose;
2276 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2278 * Applications can enable or disable heartbeats for any peer address of
2279 * an association, modify an address's heartbeat interval, force a
2280 * heartbeat to be sent immediately, and adjust the address's maximum
2281 * number of retransmissions sent before an address is considered
2282 * unreachable. The following structure is used to access and modify an
2283 * address's parameters:
2285 * struct sctp_paddrparams {
2286 * sctp_assoc_t spp_assoc_id;
2287 * struct sockaddr_storage spp_address;
2288 * uint32_t spp_hbinterval;
2289 * uint16_t spp_pathmaxrxt;
2290 * uint32_t spp_pathmtu;
2291 * uint32_t spp_sackdelay;
2292 * uint32_t spp_flags;
2295 * spp_assoc_id - (one-to-many style socket) This is filled in the
2296 * application, and identifies the association for
2298 * spp_address - This specifies which address is of interest.
2299 * spp_hbinterval - This contains the value of the heartbeat interval,
2300 * in milliseconds. If a value of zero
2301 * is present in this field then no changes are to
2302 * be made to this parameter.
2303 * spp_pathmaxrxt - This contains the maximum number of
2304 * retransmissions before this address shall be
2305 * considered unreachable. If a value of zero
2306 * is present in this field then no changes are to
2307 * be made to this parameter.
2308 * spp_pathmtu - When Path MTU discovery is disabled the value
2309 * specified here will be the "fixed" path mtu.
2310 * Note that if the spp_address field is empty
2311 * then all associations on this address will
2312 * have this fixed path mtu set upon them.
2314 * spp_sackdelay - When delayed sack is enabled, this value specifies
2315 * the number of milliseconds that sacks will be delayed
2316 * for. This value will apply to all addresses of an
2317 * association if the spp_address field is empty. Note
2318 * also, that if delayed sack is enabled and this
2319 * value is set to 0, no change is made to the last
2320 * recorded delayed sack timer value.
2322 * spp_flags - These flags are used to control various features
2323 * on an association. The flag field may contain
2324 * zero or more of the following options.
2326 * SPP_HB_ENABLE - Enable heartbeats on the
2327 * specified address. Note that if the address
2328 * field is empty all addresses for the association
2329 * have heartbeats enabled upon them.
2331 * SPP_HB_DISABLE - Disable heartbeats on the
2332 * speicifed address. Note that if the address
2333 * field is empty all addresses for the association
2334 * will have their heartbeats disabled. Note also
2335 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2336 * mutually exclusive, only one of these two should
2337 * be specified. Enabling both fields will have
2338 * undetermined results.
2340 * SPP_HB_DEMAND - Request a user initiated heartbeat
2341 * to be made immediately.
2343 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2344 * heartbeat delayis to be set to the value of 0
2347 * SPP_PMTUD_ENABLE - This field will enable PMTU
2348 * discovery upon the specified address. Note that
2349 * if the address feild is empty then all addresses
2350 * on the association are effected.
2352 * SPP_PMTUD_DISABLE - This field will disable PMTU
2353 * discovery upon the specified address. Note that
2354 * if the address feild is empty then all addresses
2355 * on the association are effected. Not also that
2356 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2357 * exclusive. Enabling both will have undetermined
2360 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2361 * on delayed sack. The time specified in spp_sackdelay
2362 * is used to specify the sack delay for this address. Note
2363 * that if spp_address is empty then all addresses will
2364 * enable delayed sack and take on the sack delay
2365 * value specified in spp_sackdelay.
2366 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2367 * off delayed sack. If the spp_address field is blank then
2368 * delayed sack is disabled for the entire association. Note
2369 * also that this field is mutually exclusive to
2370 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2373 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2374 struct sctp_transport *trans,
2375 struct sctp_association *asoc,
2376 struct sctp_sock *sp,
2379 int sackdelay_change)
2383 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2384 struct net *net = sock_net(trans->asoc->base.sk);
2386 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2391 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2392 * this field is ignored. Note also that a value of zero indicates
2393 * the current setting should be left unchanged.
2395 if (params->spp_flags & SPP_HB_ENABLE) {
2397 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2398 * set. This lets us use 0 value when this flag
2401 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2402 params->spp_hbinterval = 0;
2404 if (params->spp_hbinterval ||
2405 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2408 msecs_to_jiffies(params->spp_hbinterval);
2411 msecs_to_jiffies(params->spp_hbinterval);
2413 sp->hbinterval = params->spp_hbinterval;
2420 trans->param_flags =
2421 (trans->param_flags & ~SPP_HB) | hb_change;
2424 (asoc->param_flags & ~SPP_HB) | hb_change;
2427 (sp->param_flags & ~SPP_HB) | hb_change;
2431 /* When Path MTU discovery is disabled the value specified here will
2432 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2433 * include the flag SPP_PMTUD_DISABLE for this field to have any
2436 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2438 trans->pathmtu = params->spp_pathmtu;
2439 sctp_assoc_sync_pmtu(asoc);
2441 asoc->pathmtu = params->spp_pathmtu;
2443 sp->pathmtu = params->spp_pathmtu;
2449 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2450 (params->spp_flags & SPP_PMTUD_ENABLE);
2451 trans->param_flags =
2452 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2454 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2455 sctp_assoc_sync_pmtu(asoc);
2459 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2462 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2466 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2467 * value of this field is ignored. Note also that a value of zero
2468 * indicates the current setting should be left unchanged.
2470 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2473 msecs_to_jiffies(params->spp_sackdelay);
2476 msecs_to_jiffies(params->spp_sackdelay);
2478 sp->sackdelay = params->spp_sackdelay;
2482 if (sackdelay_change) {
2484 trans->param_flags =
2485 (trans->param_flags & ~SPP_SACKDELAY) |
2489 (asoc->param_flags & ~SPP_SACKDELAY) |
2493 (sp->param_flags & ~SPP_SACKDELAY) |
2498 /* Note that a value of zero indicates the current setting should be
2501 if (params->spp_pathmaxrxt) {
2503 trans->pathmaxrxt = params->spp_pathmaxrxt;
2505 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2507 sp->pathmaxrxt = params->spp_pathmaxrxt;
2514 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2515 char __user *optval,
2516 unsigned int optlen)
2518 struct sctp_paddrparams params;
2519 struct sctp_transport *trans = NULL;
2520 struct sctp_association *asoc = NULL;
2521 struct sctp_sock *sp = sctp_sk(sk);
2523 int hb_change, pmtud_change, sackdelay_change;
2525 if (optlen != sizeof(struct sctp_paddrparams))
2528 if (copy_from_user(¶ms, optval, optlen))
2531 /* Validate flags and value parameters. */
2532 hb_change = params.spp_flags & SPP_HB;
2533 pmtud_change = params.spp_flags & SPP_PMTUD;
2534 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2536 if (hb_change == SPP_HB ||
2537 pmtud_change == SPP_PMTUD ||
2538 sackdelay_change == SPP_SACKDELAY ||
2539 params.spp_sackdelay > 500 ||
2540 (params.spp_pathmtu &&
2541 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2544 /* If an address other than INADDR_ANY is specified, and
2545 * no transport is found, then the request is invalid.
2547 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
2548 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2549 params.spp_assoc_id);
2554 /* Get association, if assoc_id != 0 and the socket is a one
2555 * to many style socket, and an association was not found, then
2556 * the id was invalid.
2558 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2559 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2562 /* Heartbeat demand can only be sent on a transport or
2563 * association, but not a socket.
2565 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2568 /* Process parameters. */
2569 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2570 hb_change, pmtud_change,
2576 /* If changes are for association, also apply parameters to each
2579 if (!trans && asoc) {
2580 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2582 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2583 hb_change, pmtud_change,
2591 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2593 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2596 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2598 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2602 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2604 * This option will effect the way delayed acks are performed. This
2605 * option allows you to get or set the delayed ack time, in
2606 * milliseconds. It also allows changing the delayed ack frequency.
2607 * Changing the frequency to 1 disables the delayed sack algorithm. If
2608 * the assoc_id is 0, then this sets or gets the endpoints default
2609 * values. If the assoc_id field is non-zero, then the set or get
2610 * effects the specified association for the one to many model (the
2611 * assoc_id field is ignored by the one to one model). Note that if
2612 * sack_delay or sack_freq are 0 when setting this option, then the
2613 * current values will remain unchanged.
2615 * struct sctp_sack_info {
2616 * sctp_assoc_t sack_assoc_id;
2617 * uint32_t sack_delay;
2618 * uint32_t sack_freq;
2621 * sack_assoc_id - This parameter, indicates which association the user
2622 * is performing an action upon. Note that if this field's value is
2623 * zero then the endpoints default value is changed (effecting future
2624 * associations only).
2626 * sack_delay - This parameter contains the number of milliseconds that
2627 * the user is requesting the delayed ACK timer be set to. Note that
2628 * this value is defined in the standard to be between 200 and 500
2631 * sack_freq - This parameter contains the number of packets that must
2632 * be received before a sack is sent without waiting for the delay
2633 * timer to expire. The default value for this is 2, setting this
2634 * value to 1 will disable the delayed sack algorithm.
2637 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2638 char __user *optval, unsigned int optlen)
2640 struct sctp_sack_info params;
2641 struct sctp_transport *trans = NULL;
2642 struct sctp_association *asoc = NULL;
2643 struct sctp_sock *sp = sctp_sk(sk);
2645 if (optlen == sizeof(struct sctp_sack_info)) {
2646 if (copy_from_user(¶ms, optval, optlen))
2649 if (params.sack_delay == 0 && params.sack_freq == 0)
2651 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2652 pr_warn_ratelimited(DEPRECATED
2654 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2655 "Use struct sctp_sack_info instead\n",
2656 current->comm, task_pid_nr(current));
2657 if (copy_from_user(¶ms, optval, optlen))
2660 if (params.sack_delay == 0)
2661 params.sack_freq = 1;
2663 params.sack_freq = 0;
2667 /* Validate value parameter. */
2668 if (params.sack_delay > 500)
2671 /* Get association, if sack_assoc_id != 0 and the socket is a one
2672 * to many style socket, and an association was not found, then
2673 * the id was invalid.
2675 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2676 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2679 if (params.sack_delay) {
2682 msecs_to_jiffies(params.sack_delay);
2684 sctp_spp_sackdelay_enable(asoc->param_flags);
2686 sp->sackdelay = params.sack_delay;
2688 sctp_spp_sackdelay_enable(sp->param_flags);
2692 if (params.sack_freq == 1) {
2695 sctp_spp_sackdelay_disable(asoc->param_flags);
2698 sctp_spp_sackdelay_disable(sp->param_flags);
2700 } else if (params.sack_freq > 1) {
2702 asoc->sackfreq = params.sack_freq;
2704 sctp_spp_sackdelay_enable(asoc->param_flags);
2706 sp->sackfreq = params.sack_freq;
2708 sctp_spp_sackdelay_enable(sp->param_flags);
2712 /* If change is for association, also apply to each transport. */
2714 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2716 if (params.sack_delay) {
2718 msecs_to_jiffies(params.sack_delay);
2719 trans->param_flags =
2720 sctp_spp_sackdelay_enable(trans->param_flags);
2722 if (params.sack_freq == 1) {
2723 trans->param_flags =
2724 sctp_spp_sackdelay_disable(trans->param_flags);
2725 } else if (params.sack_freq > 1) {
2726 trans->sackfreq = params.sack_freq;
2727 trans->param_flags =
2728 sctp_spp_sackdelay_enable(trans->param_flags);
2736 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2738 * Applications can specify protocol parameters for the default association
2739 * initialization. The option name argument to setsockopt() and getsockopt()
2742 * Setting initialization parameters is effective only on an unconnected
2743 * socket (for UDP-style sockets only future associations are effected
2744 * by the change). With TCP-style sockets, this option is inherited by
2745 * sockets derived from a listener socket.
2747 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2749 struct sctp_initmsg sinit;
2750 struct sctp_sock *sp = sctp_sk(sk);
2752 if (optlen != sizeof(struct sctp_initmsg))
2754 if (copy_from_user(&sinit, optval, optlen))
2757 if (sinit.sinit_num_ostreams)
2758 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2759 if (sinit.sinit_max_instreams)
2760 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2761 if (sinit.sinit_max_attempts)
2762 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2763 if (sinit.sinit_max_init_timeo)
2764 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2770 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2772 * Applications that wish to use the sendto() system call may wish to
2773 * specify a default set of parameters that would normally be supplied
2774 * through the inclusion of ancillary data. This socket option allows
2775 * such an application to set the default sctp_sndrcvinfo structure.
2776 * The application that wishes to use this socket option simply passes
2777 * in to this call the sctp_sndrcvinfo structure defined in Section
2778 * 5.2.2) The input parameters accepted by this call include
2779 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2780 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2781 * to this call if the caller is using the UDP model.
2783 static int sctp_setsockopt_default_send_param(struct sock *sk,
2784 char __user *optval,
2785 unsigned int optlen)
2787 struct sctp_sock *sp = sctp_sk(sk);
2788 struct sctp_association *asoc;
2789 struct sctp_sndrcvinfo info;
2791 if (optlen != sizeof(info))
2793 if (copy_from_user(&info, optval, optlen))
2795 if (info.sinfo_flags &
2796 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2797 SCTP_ABORT | SCTP_EOF))
2800 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2801 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2804 asoc->default_stream = info.sinfo_stream;
2805 asoc->default_flags = info.sinfo_flags;
2806 asoc->default_ppid = info.sinfo_ppid;
2807 asoc->default_context = info.sinfo_context;
2808 asoc->default_timetolive = info.sinfo_timetolive;
2810 sp->default_stream = info.sinfo_stream;
2811 sp->default_flags = info.sinfo_flags;
2812 sp->default_ppid = info.sinfo_ppid;
2813 sp->default_context = info.sinfo_context;
2814 sp->default_timetolive = info.sinfo_timetolive;
2820 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2821 * (SCTP_DEFAULT_SNDINFO)
2823 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
2824 char __user *optval,
2825 unsigned int optlen)
2827 struct sctp_sock *sp = sctp_sk(sk);
2828 struct sctp_association *asoc;
2829 struct sctp_sndinfo info;
2831 if (optlen != sizeof(info))
2833 if (copy_from_user(&info, optval, optlen))
2835 if (info.snd_flags &
2836 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2837 SCTP_ABORT | SCTP_EOF))
2840 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
2841 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
2844 asoc->default_stream = info.snd_sid;
2845 asoc->default_flags = info.snd_flags;
2846 asoc->default_ppid = info.snd_ppid;
2847 asoc->default_context = info.snd_context;
2849 sp->default_stream = info.snd_sid;
2850 sp->default_flags = info.snd_flags;
2851 sp->default_ppid = info.snd_ppid;
2852 sp->default_context = info.snd_context;
2858 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2860 * Requests that the local SCTP stack use the enclosed peer address as
2861 * the association primary. The enclosed address must be one of the
2862 * association peer's addresses.
2864 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2865 unsigned int optlen)
2867 struct sctp_prim prim;
2868 struct sctp_transport *trans;
2870 if (optlen != sizeof(struct sctp_prim))
2873 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2876 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2880 sctp_assoc_set_primary(trans->asoc, trans);
2886 * 7.1.5 SCTP_NODELAY
2888 * Turn on/off any Nagle-like algorithm. This means that packets are
2889 * generally sent as soon as possible and no unnecessary delays are
2890 * introduced, at the cost of more packets in the network. Expects an
2891 * integer boolean flag.
2893 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2894 unsigned int optlen)
2898 if (optlen < sizeof(int))
2900 if (get_user(val, (int __user *)optval))
2903 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2909 * 7.1.1 SCTP_RTOINFO
2911 * The protocol parameters used to initialize and bound retransmission
2912 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2913 * and modify these parameters.
2914 * All parameters are time values, in milliseconds. A value of 0, when
2915 * modifying the parameters, indicates that the current value should not
2919 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2921 struct sctp_rtoinfo rtoinfo;
2922 struct sctp_association *asoc;
2923 unsigned long rto_min, rto_max;
2924 struct sctp_sock *sp = sctp_sk(sk);
2926 if (optlen != sizeof (struct sctp_rtoinfo))
2929 if (copy_from_user(&rtoinfo, optval, optlen))
2932 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2934 /* Set the values to the specific association */
2935 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2938 rto_max = rtoinfo.srto_max;
2939 rto_min = rtoinfo.srto_min;
2942 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2944 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2947 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2949 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2951 if (rto_min > rto_max)
2955 if (rtoinfo.srto_initial != 0)
2957 msecs_to_jiffies(rtoinfo.srto_initial);
2958 asoc->rto_max = rto_max;
2959 asoc->rto_min = rto_min;
2961 /* If there is no association or the association-id = 0
2962 * set the values to the endpoint.
2964 if (rtoinfo.srto_initial != 0)
2965 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2966 sp->rtoinfo.srto_max = rto_max;
2967 sp->rtoinfo.srto_min = rto_min;
2975 * 7.1.2 SCTP_ASSOCINFO
2977 * This option is used to tune the maximum retransmission attempts
2978 * of the association.
2979 * Returns an error if the new association retransmission value is
2980 * greater than the sum of the retransmission value of the peer.
2981 * See [SCTP] for more information.
2984 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2987 struct sctp_assocparams assocparams;
2988 struct sctp_association *asoc;
2990 if (optlen != sizeof(struct sctp_assocparams))
2992 if (copy_from_user(&assocparams, optval, optlen))
2995 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2997 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
3000 /* Set the values to the specific association */
3002 if (assocparams.sasoc_asocmaxrxt != 0) {
3005 struct sctp_transport *peer_addr;
3007 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
3009 path_sum += peer_addr->pathmaxrxt;
3013 /* Only validate asocmaxrxt if we have more than
3014 * one path/transport. We do this because path
3015 * retransmissions are only counted when we have more
3019 assocparams.sasoc_asocmaxrxt > path_sum)
3022 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3025 if (assocparams.sasoc_cookie_life != 0)
3026 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3028 /* Set the values to the endpoint */
3029 struct sctp_sock *sp = sctp_sk(sk);
3031 if (assocparams.sasoc_asocmaxrxt != 0)
3032 sp->assocparams.sasoc_asocmaxrxt =
3033 assocparams.sasoc_asocmaxrxt;
3034 if (assocparams.sasoc_cookie_life != 0)
3035 sp->assocparams.sasoc_cookie_life =
3036 assocparams.sasoc_cookie_life;
3042 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3044 * This socket option is a boolean flag which turns on or off mapped V4
3045 * addresses. If this option is turned on and the socket is type
3046 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3047 * If this option is turned off, then no mapping will be done of V4
3048 * addresses and a user will receive both PF_INET6 and PF_INET type
3049 * addresses on the socket.
3051 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3054 struct sctp_sock *sp = sctp_sk(sk);
3056 if (optlen < sizeof(int))
3058 if (get_user(val, (int __user *)optval))
3069 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3070 * This option will get or set the maximum size to put in any outgoing
3071 * SCTP DATA chunk. If a message is larger than this size it will be
3072 * fragmented by SCTP into the specified size. Note that the underlying
3073 * SCTP implementation may fragment into smaller sized chunks when the
3074 * PMTU of the underlying association is smaller than the value set by
3075 * the user. The default value for this option is '0' which indicates
3076 * the user is NOT limiting fragmentation and only the PMTU will effect
3077 * SCTP's choice of DATA chunk size. Note also that values set larger
3078 * than the maximum size of an IP datagram will effectively let SCTP
3079 * control fragmentation (i.e. the same as setting this option to 0).
3081 * The following structure is used to access and modify this parameter:
3083 * struct sctp_assoc_value {
3084 * sctp_assoc_t assoc_id;
3085 * uint32_t assoc_value;
3088 * assoc_id: This parameter is ignored for one-to-one style sockets.
3089 * For one-to-many style sockets this parameter indicates which
3090 * association the user is performing an action upon. Note that if
3091 * this field's value is zero then the endpoints default value is
3092 * changed (effecting future associations only).
3093 * assoc_value: This parameter specifies the maximum size in bytes.
3095 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3097 struct sctp_assoc_value params;
3098 struct sctp_association *asoc;
3099 struct sctp_sock *sp = sctp_sk(sk);
3102 if (optlen == sizeof(int)) {
3103 pr_warn_ratelimited(DEPRECATED
3105 "Use of int in maxseg socket option.\n"
3106 "Use struct sctp_assoc_value instead\n",
3107 current->comm, task_pid_nr(current));
3108 if (copy_from_user(&val, optval, optlen))
3110 params.assoc_id = 0;
3111 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3112 if (copy_from_user(¶ms, optval, optlen))
3114 val = params.assoc_value;
3118 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3121 asoc = sctp_id2assoc(sk, params.assoc_id);
3122 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3127 val = asoc->pathmtu;
3128 val -= sp->pf->af->net_header_len;
3129 val -= sizeof(struct sctphdr) +
3130 sizeof(struct sctp_data_chunk);
3132 asoc->user_frag = val;
3133 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3135 sp->user_frag = val;
3143 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3145 * Requests that the peer mark the enclosed address as the association
3146 * primary. The enclosed address must be one of the association's
3147 * locally bound addresses. The following structure is used to make a
3148 * set primary request:
3150 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3151 unsigned int optlen)
3153 struct net *net = sock_net(sk);
3154 struct sctp_sock *sp;
3155 struct sctp_association *asoc = NULL;
3156 struct sctp_setpeerprim prim;
3157 struct sctp_chunk *chunk;
3163 if (!net->sctp.addip_enable)
3166 if (optlen != sizeof(struct sctp_setpeerprim))
3169 if (copy_from_user(&prim, optval, optlen))
3172 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3176 if (!asoc->peer.asconf_capable)
3179 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3182 if (!sctp_state(asoc, ESTABLISHED))
3185 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3189 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3190 return -EADDRNOTAVAIL;
3192 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3193 return -EADDRNOTAVAIL;
3195 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3196 chunk = sctp_make_asconf_set_prim(asoc,
3197 (union sctp_addr *)&prim.sspp_addr);
3201 err = sctp_send_asconf(asoc, chunk);
3203 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3208 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3209 unsigned int optlen)
3211 struct sctp_setadaptation adaptation;
3213 if (optlen != sizeof(struct sctp_setadaptation))
3215 if (copy_from_user(&adaptation, optval, optlen))
3218 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3224 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3226 * The context field in the sctp_sndrcvinfo structure is normally only
3227 * used when a failed message is retrieved holding the value that was
3228 * sent down on the actual send call. This option allows the setting of
3229 * a default context on an association basis that will be received on
3230 * reading messages from the peer. This is especially helpful in the
3231 * one-2-many model for an application to keep some reference to an
3232 * internal state machine that is processing messages on the
3233 * association. Note that the setting of this value only effects
3234 * received messages from the peer and does not effect the value that is
3235 * saved with outbound messages.
3237 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3238 unsigned int optlen)
3240 struct sctp_assoc_value params;
3241 struct sctp_sock *sp;
3242 struct sctp_association *asoc;
3244 if (optlen != sizeof(struct sctp_assoc_value))
3246 if (copy_from_user(¶ms, optval, optlen))
3251 if (params.assoc_id != 0) {
3252 asoc = sctp_id2assoc(sk, params.assoc_id);
3255 asoc->default_rcv_context = params.assoc_value;
3257 sp->default_rcv_context = params.assoc_value;
3264 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3266 * This options will at a minimum specify if the implementation is doing
3267 * fragmented interleave. Fragmented interleave, for a one to many
3268 * socket, is when subsequent calls to receive a message may return
3269 * parts of messages from different associations. Some implementations
3270 * may allow you to turn this value on or off. If so, when turned off,
3271 * no fragment interleave will occur (which will cause a head of line
3272 * blocking amongst multiple associations sharing the same one to many
3273 * socket). When this option is turned on, then each receive call may
3274 * come from a different association (thus the user must receive data
3275 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3276 * association each receive belongs to.
3278 * This option takes a boolean value. A non-zero value indicates that
3279 * fragmented interleave is on. A value of zero indicates that
3280 * fragmented interleave is off.
3282 * Note that it is important that an implementation that allows this
3283 * option to be turned on, have it off by default. Otherwise an unaware
3284 * application using the one to many model may become confused and act
3287 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3288 char __user *optval,
3289 unsigned int optlen)
3293 if (optlen != sizeof(int))
3295 if (get_user(val, (int __user *)optval))
3298 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3304 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3305 * (SCTP_PARTIAL_DELIVERY_POINT)
3307 * This option will set or get the SCTP partial delivery point. This
3308 * point is the size of a message where the partial delivery API will be
3309 * invoked to help free up rwnd space for the peer. Setting this to a
3310 * lower value will cause partial deliveries to happen more often. The
3311 * calls argument is an integer that sets or gets the partial delivery
3312 * point. Note also that the call will fail if the user attempts to set
3313 * this value larger than the socket receive buffer size.
3315 * Note that any single message having a length smaller than or equal to
3316 * the SCTP partial delivery point will be delivered in one single read
3317 * call as long as the user provided buffer is large enough to hold the
3320 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3321 char __user *optval,
3322 unsigned int optlen)
3326 if (optlen != sizeof(u32))
3328 if (get_user(val, (int __user *)optval))
3331 /* Note: We double the receive buffer from what the user sets
3332 * it to be, also initial rwnd is based on rcvbuf/2.
3334 if (val > (sk->sk_rcvbuf >> 1))
3337 sctp_sk(sk)->pd_point = val;
3339 return 0; /* is this the right error code? */
3343 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3345 * This option will allow a user to change the maximum burst of packets
3346 * that can be emitted by this association. Note that the default value
3347 * is 4, and some implementations may restrict this setting so that it
3348 * can only be lowered.
3350 * NOTE: This text doesn't seem right. Do this on a socket basis with
3351 * future associations inheriting the socket value.
3353 static int sctp_setsockopt_maxburst(struct sock *sk,
3354 char __user *optval,
3355 unsigned int optlen)
3357 struct sctp_assoc_value params;
3358 struct sctp_sock *sp;
3359 struct sctp_association *asoc;
3363 if (optlen == sizeof(int)) {
3364 pr_warn_ratelimited(DEPRECATED
3366 "Use of int in max_burst socket option deprecated.\n"
3367 "Use struct sctp_assoc_value instead\n",
3368 current->comm, task_pid_nr(current));
3369 if (copy_from_user(&val, optval, optlen))
3371 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3372 if (copy_from_user(¶ms, optval, optlen))
3374 val = params.assoc_value;
3375 assoc_id = params.assoc_id;
3381 if (assoc_id != 0) {
3382 asoc = sctp_id2assoc(sk, assoc_id);
3385 asoc->max_burst = val;
3387 sp->max_burst = val;
3393 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3395 * This set option adds a chunk type that the user is requesting to be
3396 * received only in an authenticated way. Changes to the list of chunks
3397 * will only effect future associations on the socket.
3399 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3400 char __user *optval,
3401 unsigned int optlen)
3403 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3404 struct sctp_authchunk val;
3406 if (!ep->auth_enable)
3409 if (optlen != sizeof(struct sctp_authchunk))
3411 if (copy_from_user(&val, optval, optlen))
3414 switch (val.sauth_chunk) {
3416 case SCTP_CID_INIT_ACK:
3417 case SCTP_CID_SHUTDOWN_COMPLETE:
3422 /* add this chunk id to the endpoint */
3423 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3427 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3429 * This option gets or sets the list of HMAC algorithms that the local
3430 * endpoint requires the peer to use.
3432 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3433 char __user *optval,
3434 unsigned int optlen)
3436 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3437 struct sctp_hmacalgo *hmacs;
3441 if (!ep->auth_enable)
3444 if (optlen < sizeof(struct sctp_hmacalgo))
3447 hmacs = memdup_user(optval, optlen);
3449 return PTR_ERR(hmacs);
3451 idents = hmacs->shmac_num_idents;
3452 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3453 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3458 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3465 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3467 * This option will set a shared secret key which is used to build an
3468 * association shared key.
3470 static int sctp_setsockopt_auth_key(struct sock *sk,
3471 char __user *optval,
3472 unsigned int optlen)
3474 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3475 struct sctp_authkey *authkey;
3476 struct sctp_association *asoc;
3479 if (!ep->auth_enable)
3482 if (optlen <= sizeof(struct sctp_authkey))
3485 authkey = memdup_user(optval, optlen);
3486 if (IS_ERR(authkey))
3487 return PTR_ERR(authkey);
3489 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3494 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3495 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3500 ret = sctp_auth_set_key(ep, asoc, authkey);
3507 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3509 * This option will get or set the active shared key to be used to build
3510 * the association shared key.
3512 static int sctp_setsockopt_active_key(struct sock *sk,
3513 char __user *optval,
3514 unsigned int optlen)
3516 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3517 struct sctp_authkeyid val;
3518 struct sctp_association *asoc;
3520 if (!ep->auth_enable)
3523 if (optlen != sizeof(struct sctp_authkeyid))
3525 if (copy_from_user(&val, optval, optlen))
3528 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3529 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3532 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3536 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3538 * This set option will delete a shared secret key from use.
3540 static int sctp_setsockopt_del_key(struct sock *sk,
3541 char __user *optval,
3542 unsigned int optlen)
3544 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3545 struct sctp_authkeyid val;
3546 struct sctp_association *asoc;
3548 if (!ep->auth_enable)
3551 if (optlen != sizeof(struct sctp_authkeyid))
3553 if (copy_from_user(&val, optval, optlen))
3556 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3557 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3560 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3565 * 8.1.23 SCTP_AUTO_ASCONF
3567 * This option will enable or disable the use of the automatic generation of
3568 * ASCONF chunks to add and delete addresses to an existing association. Note
3569 * that this option has two caveats namely: a) it only affects sockets that
3570 * are bound to all addresses available to the SCTP stack, and b) the system
3571 * administrator may have an overriding control that turns the ASCONF feature
3572 * off no matter what setting the socket option may have.
3573 * This option expects an integer boolean flag, where a non-zero value turns on
3574 * the option, and a zero value turns off the option.
3575 * Note. In this implementation, socket operation overrides default parameter
3576 * being set by sysctl as well as FreeBSD implementation
3578 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3579 unsigned int optlen)
3582 struct sctp_sock *sp = sctp_sk(sk);
3584 if (optlen < sizeof(int))
3586 if (get_user(val, (int __user *)optval))
3588 if (!sctp_is_ep_boundall(sk) && val)
3590 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3593 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3594 if (val == 0 && sp->do_auto_asconf) {
3595 list_del(&sp->auto_asconf_list);
3596 sp->do_auto_asconf = 0;
3597 } else if (val && !sp->do_auto_asconf) {
3598 list_add_tail(&sp->auto_asconf_list,
3599 &sock_net(sk)->sctp.auto_asconf_splist);
3600 sp->do_auto_asconf = 1;
3602 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3607 * SCTP_PEER_ADDR_THLDS
3609 * This option allows us to alter the partially failed threshold for one or all
3610 * transports in an association. See Section 6.1 of:
3611 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3613 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3614 char __user *optval,
3615 unsigned int optlen)
3617 struct sctp_paddrthlds val;
3618 struct sctp_transport *trans;
3619 struct sctp_association *asoc;
3621 if (optlen < sizeof(struct sctp_paddrthlds))
3623 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3624 sizeof(struct sctp_paddrthlds)))
3628 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3629 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3632 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3634 if (val.spt_pathmaxrxt)
3635 trans->pathmaxrxt = val.spt_pathmaxrxt;
3636 trans->pf_retrans = val.spt_pathpfthld;
3639 if (val.spt_pathmaxrxt)
3640 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3641 asoc->pf_retrans = val.spt_pathpfthld;
3643 trans = sctp_addr_id2transport(sk, &val.spt_address,
3648 if (val.spt_pathmaxrxt)
3649 trans->pathmaxrxt = val.spt_pathmaxrxt;
3650 trans->pf_retrans = val.spt_pathpfthld;
3656 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3657 char __user *optval,
3658 unsigned int optlen)
3662 if (optlen < sizeof(int))
3664 if (get_user(val, (int __user *) optval))
3667 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3672 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3673 char __user *optval,
3674 unsigned int optlen)
3678 if (optlen < sizeof(int))
3680 if (get_user(val, (int __user *) optval))
3683 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3688 static int sctp_setsockopt_pr_supported(struct sock *sk,
3689 char __user *optval,
3690 unsigned int optlen)
3692 struct sctp_assoc_value params;
3693 struct sctp_association *asoc;
3694 int retval = -EINVAL;
3696 if (optlen != sizeof(params))
3699 if (copy_from_user(¶ms, optval, optlen)) {
3704 asoc = sctp_id2assoc(sk, params.assoc_id);
3706 asoc->prsctp_enable = !!params.assoc_value;
3707 } else if (!params.assoc_id) {
3708 struct sctp_sock *sp = sctp_sk(sk);
3710 sp->ep->prsctp_enable = !!params.assoc_value;
3721 static int sctp_setsockopt_default_prinfo(struct sock *sk,
3722 char __user *optval,
3723 unsigned int optlen)
3725 struct sctp_default_prinfo info;
3726 struct sctp_association *asoc;
3727 int retval = -EINVAL;
3729 if (optlen != sizeof(info))
3732 if (copy_from_user(&info, optval, sizeof(info))) {
3737 if (info.pr_policy & ~SCTP_PR_SCTP_MASK)
3740 if (info.pr_policy == SCTP_PR_SCTP_NONE)
3743 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
3745 SCTP_PR_SET_POLICY(asoc->default_flags, info.pr_policy);
3746 asoc->default_timetolive = info.pr_value;
3747 } else if (!info.pr_assoc_id) {
3748 struct sctp_sock *sp = sctp_sk(sk);
3750 SCTP_PR_SET_POLICY(sp->default_flags, info.pr_policy);
3751 sp->default_timetolive = info.pr_value;
3762 static int sctp_setsockopt_reconfig_supported(struct sock *sk,
3763 char __user *optval,
3764 unsigned int optlen)
3766 struct sctp_assoc_value params;
3767 struct sctp_association *asoc;
3768 int retval = -EINVAL;
3770 if (optlen != sizeof(params))
3773 if (copy_from_user(¶ms, optval, optlen)) {
3778 asoc = sctp_id2assoc(sk, params.assoc_id);
3780 asoc->reconf_enable = !!params.assoc_value;
3781 } else if (!params.assoc_id) {
3782 struct sctp_sock *sp = sctp_sk(sk);
3784 sp->ep->reconf_enable = !!params.assoc_value;
3795 static int sctp_setsockopt_enable_strreset(struct sock *sk,
3796 char __user *optval,
3797 unsigned int optlen)
3799 struct sctp_assoc_value params;
3800 struct sctp_association *asoc;
3801 int retval = -EINVAL;
3803 if (optlen != sizeof(params))
3806 if (copy_from_user(¶ms, optval, optlen)) {
3811 if (params.assoc_value & (~SCTP_ENABLE_STRRESET_MASK))
3814 asoc = sctp_id2assoc(sk, params.assoc_id);
3816 asoc->strreset_enable = params.assoc_value;
3817 } else if (!params.assoc_id) {
3818 struct sctp_sock *sp = sctp_sk(sk);
3820 sp->ep->strreset_enable = params.assoc_value;
3831 static int sctp_setsockopt_reset_streams(struct sock *sk,
3832 char __user *optval,
3833 unsigned int optlen)
3835 struct sctp_reset_streams *params;
3836 struct sctp_association *asoc;
3837 int retval = -EINVAL;
3839 if (optlen < sizeof(struct sctp_reset_streams))
3842 params = memdup_user(optval, optlen);
3844 return PTR_ERR(params);
3846 asoc = sctp_id2assoc(sk, params->srs_assoc_id);
3850 retval = sctp_send_reset_streams(asoc, params);
3857 static int sctp_setsockopt_reset_assoc(struct sock *sk,
3858 char __user *optval,
3859 unsigned int optlen)
3861 struct sctp_association *asoc;
3862 sctp_assoc_t associd;
3863 int retval = -EINVAL;
3865 if (optlen != sizeof(associd))
3868 if (copy_from_user(&associd, optval, optlen)) {
3873 asoc = sctp_id2assoc(sk, associd);
3877 retval = sctp_send_reset_assoc(asoc);
3883 static int sctp_setsockopt_add_streams(struct sock *sk,
3884 char __user *optval,
3885 unsigned int optlen)
3887 struct sctp_association *asoc;
3888 struct sctp_add_streams params;
3889 int retval = -EINVAL;
3891 if (optlen != sizeof(params))
3894 if (copy_from_user(¶ms, optval, optlen)) {
3899 asoc = sctp_id2assoc(sk, params.sas_assoc_id);
3903 retval = sctp_send_add_streams(asoc, ¶ms);
3909 /* API 6.2 setsockopt(), getsockopt()
3911 * Applications use setsockopt() and getsockopt() to set or retrieve
3912 * socket options. Socket options are used to change the default
3913 * behavior of sockets calls. They are described in Section 7.
3917 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3918 * int __user *optlen);
3919 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3922 * sd - the socket descript.
3923 * level - set to IPPROTO_SCTP for all SCTP options.
3924 * optname - the option name.
3925 * optval - the buffer to store the value of the option.
3926 * optlen - the size of the buffer.
3928 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3929 char __user *optval, unsigned int optlen)
3933 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3935 /* I can hardly begin to describe how wrong this is. This is
3936 * so broken as to be worse than useless. The API draft
3937 * REALLY is NOT helpful here... I am not convinced that the
3938 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3939 * are at all well-founded.
3941 if (level != SOL_SCTP) {
3942 struct sctp_af *af = sctp_sk(sk)->pf->af;
3943 retval = af->setsockopt(sk, level, optname, optval, optlen);
3950 case SCTP_SOCKOPT_BINDX_ADD:
3951 /* 'optlen' is the size of the addresses buffer. */
3952 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3953 optlen, SCTP_BINDX_ADD_ADDR);
3956 case SCTP_SOCKOPT_BINDX_REM:
3957 /* 'optlen' is the size of the addresses buffer. */
3958 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3959 optlen, SCTP_BINDX_REM_ADDR);
3962 case SCTP_SOCKOPT_CONNECTX_OLD:
3963 /* 'optlen' is the size of the addresses buffer. */
3964 retval = sctp_setsockopt_connectx_old(sk,
3965 (struct sockaddr __user *)optval,
3969 case SCTP_SOCKOPT_CONNECTX:
3970 /* 'optlen' is the size of the addresses buffer. */
3971 retval = sctp_setsockopt_connectx(sk,
3972 (struct sockaddr __user *)optval,
3976 case SCTP_DISABLE_FRAGMENTS:
3977 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3981 retval = sctp_setsockopt_events(sk, optval, optlen);
3984 case SCTP_AUTOCLOSE:
3985 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3988 case SCTP_PEER_ADDR_PARAMS:
3989 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3992 case SCTP_DELAYED_SACK:
3993 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3995 case SCTP_PARTIAL_DELIVERY_POINT:
3996 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
4000 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
4002 case SCTP_DEFAULT_SEND_PARAM:
4003 retval = sctp_setsockopt_default_send_param(sk, optval,
4006 case SCTP_DEFAULT_SNDINFO:
4007 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
4009 case SCTP_PRIMARY_ADDR:
4010 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
4012 case SCTP_SET_PEER_PRIMARY_ADDR:
4013 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
4016 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
4019 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
4021 case SCTP_ASSOCINFO:
4022 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
4024 case SCTP_I_WANT_MAPPED_V4_ADDR:
4025 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
4028 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
4030 case SCTP_ADAPTATION_LAYER:
4031 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
4034 retval = sctp_setsockopt_context(sk, optval, optlen);
4036 case SCTP_FRAGMENT_INTERLEAVE:
4037 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
4039 case SCTP_MAX_BURST:
4040 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
4042 case SCTP_AUTH_CHUNK:
4043 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
4045 case SCTP_HMAC_IDENT:
4046 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
4049 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
4051 case SCTP_AUTH_ACTIVE_KEY:
4052 retval = sctp_setsockopt_active_key(sk, optval, optlen);
4054 case SCTP_AUTH_DELETE_KEY:
4055 retval = sctp_setsockopt_del_key(sk, optval, optlen);
4057 case SCTP_AUTO_ASCONF:
4058 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
4060 case SCTP_PEER_ADDR_THLDS:
4061 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
4063 case SCTP_RECVRCVINFO:
4064 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
4066 case SCTP_RECVNXTINFO:
4067 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
4069 case SCTP_PR_SUPPORTED:
4070 retval = sctp_setsockopt_pr_supported(sk, optval, optlen);
4072 case SCTP_DEFAULT_PRINFO:
4073 retval = sctp_setsockopt_default_prinfo(sk, optval, optlen);
4075 case SCTP_RECONFIG_SUPPORTED:
4076 retval = sctp_setsockopt_reconfig_supported(sk, optval, optlen);
4078 case SCTP_ENABLE_STREAM_RESET:
4079 retval = sctp_setsockopt_enable_strreset(sk, optval, optlen);
4081 case SCTP_RESET_STREAMS:
4082 retval = sctp_setsockopt_reset_streams(sk, optval, optlen);
4084 case SCTP_RESET_ASSOC:
4085 retval = sctp_setsockopt_reset_assoc(sk, optval, optlen);
4087 case SCTP_ADD_STREAMS:
4088 retval = sctp_setsockopt_add_streams(sk, optval, optlen);
4091 retval = -ENOPROTOOPT;
4101 /* API 3.1.6 connect() - UDP Style Syntax
4103 * An application may use the connect() call in the UDP model to initiate an
4104 * association without sending data.
4108 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
4110 * sd: the socket descriptor to have a new association added to.
4112 * nam: the address structure (either struct sockaddr_in or struct
4113 * sockaddr_in6 defined in RFC2553 [7]).
4115 * len: the size of the address.
4117 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
4125 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
4128 /* Validate addr_len before calling common connect/connectx routine. */
4129 af = sctp_get_af_specific(addr->sa_family);
4130 if (!af || addr_len < af->sockaddr_len) {
4133 /* Pass correct addr len to common routine (so it knows there
4134 * is only one address being passed.
4136 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
4143 /* FIXME: Write comments. */
4144 static int sctp_disconnect(struct sock *sk, int flags)
4146 return -EOPNOTSUPP; /* STUB */
4149 /* 4.1.4 accept() - TCP Style Syntax
4151 * Applications use accept() call to remove an established SCTP
4152 * association from the accept queue of the endpoint. A new socket
4153 * descriptor will be returned from accept() to represent the newly
4154 * formed association.
4156 static struct sock *sctp_accept(struct sock *sk, int flags, int *err, bool kern)
4158 struct sctp_sock *sp;
4159 struct sctp_endpoint *ep;
4160 struct sock *newsk = NULL;
4161 struct sctp_association *asoc;
4170 if (!sctp_style(sk, TCP)) {
4171 error = -EOPNOTSUPP;
4175 if (!sctp_sstate(sk, LISTENING)) {
4180 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
4182 error = sctp_wait_for_accept(sk, timeo);
4186 /* We treat the list of associations on the endpoint as the accept
4187 * queue and pick the first association on the list.
4189 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
4191 newsk = sp->pf->create_accept_sk(sk, asoc, kern);
4197 /* Populate the fields of the newsk from the oldsk and migrate the
4198 * asoc to the newsk.
4200 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
4208 /* The SCTP ioctl handler. */
4209 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
4216 * SEQPACKET-style sockets in LISTENING state are valid, for
4217 * SCTP, so only discard TCP-style sockets in LISTENING state.
4219 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4224 struct sk_buff *skb;
4225 unsigned int amount = 0;
4227 skb = skb_peek(&sk->sk_receive_queue);
4230 * We will only return the amount of this packet since
4231 * that is all that will be read.
4235 rc = put_user(amount, (int __user *)arg);
4247 /* This is the function which gets called during socket creation to
4248 * initialized the SCTP-specific portion of the sock.
4249 * The sock structure should already be zero-filled memory.
4251 static int sctp_init_sock(struct sock *sk)
4253 struct net *net = sock_net(sk);
4254 struct sctp_sock *sp;
4256 pr_debug("%s: sk:%p\n", __func__, sk);
4260 /* Initialize the SCTP per socket area. */
4261 switch (sk->sk_type) {
4262 case SOCK_SEQPACKET:
4263 sp->type = SCTP_SOCKET_UDP;
4266 sp->type = SCTP_SOCKET_TCP;
4269 return -ESOCKTNOSUPPORT;
4272 sk->sk_gso_type = SKB_GSO_SCTP;
4274 /* Initialize default send parameters. These parameters can be
4275 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4277 sp->default_stream = 0;
4278 sp->default_ppid = 0;
4279 sp->default_flags = 0;
4280 sp->default_context = 0;
4281 sp->default_timetolive = 0;
4283 sp->default_rcv_context = 0;
4284 sp->max_burst = net->sctp.max_burst;
4286 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4288 /* Initialize default setup parameters. These parameters
4289 * can be modified with the SCTP_INITMSG socket option or
4290 * overridden by the SCTP_INIT CMSG.
4292 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4293 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4294 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4295 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4297 /* Initialize default RTO related parameters. These parameters can
4298 * be modified for with the SCTP_RTOINFO socket option.
4300 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4301 sp->rtoinfo.srto_max = net->sctp.rto_max;
4302 sp->rtoinfo.srto_min = net->sctp.rto_min;
4304 /* Initialize default association related parameters. These parameters
4305 * can be modified with the SCTP_ASSOCINFO socket option.
4307 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4308 sp->assocparams.sasoc_number_peer_destinations = 0;
4309 sp->assocparams.sasoc_peer_rwnd = 0;
4310 sp->assocparams.sasoc_local_rwnd = 0;
4311 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4313 /* Initialize default event subscriptions. By default, all the
4316 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4318 /* Default Peer Address Parameters. These defaults can
4319 * be modified via SCTP_PEER_ADDR_PARAMS
4321 sp->hbinterval = net->sctp.hb_interval;
4322 sp->pathmaxrxt = net->sctp.max_retrans_path;
4323 sp->pathmtu = 0; /* allow default discovery */
4324 sp->sackdelay = net->sctp.sack_timeout;
4326 sp->param_flags = SPP_HB_ENABLE |
4328 SPP_SACKDELAY_ENABLE;
4330 /* If enabled no SCTP message fragmentation will be performed.
4331 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4333 sp->disable_fragments = 0;
4335 /* Enable Nagle algorithm by default. */
4338 sp->recvrcvinfo = 0;
4339 sp->recvnxtinfo = 0;
4341 /* Enable by default. */
4344 /* Auto-close idle associations after the configured
4345 * number of seconds. A value of 0 disables this
4346 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4347 * for UDP-style sockets only.
4351 /* User specified fragmentation limit. */
4354 sp->adaptation_ind = 0;
4356 sp->pf = sctp_get_pf_specific(sk->sk_family);
4358 /* Control variables for partial data delivery. */
4359 atomic_set(&sp->pd_mode, 0);
4360 skb_queue_head_init(&sp->pd_lobby);
4361 sp->frag_interleave = 0;
4363 /* Create a per socket endpoint structure. Even if we
4364 * change the data structure relationships, this may still
4365 * be useful for storing pre-connect address information.
4367 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4373 sk->sk_destruct = sctp_destruct_sock;
4375 SCTP_DBG_OBJCNT_INC(sock);
4378 percpu_counter_inc(&sctp_sockets_allocated);
4379 sock_prot_inuse_add(net, sk->sk_prot, 1);
4381 /* Nothing can fail after this block, otherwise
4382 * sctp_destroy_sock() will be called without addr_wq_lock held
4384 if (net->sctp.default_auto_asconf) {
4385 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
4386 list_add_tail(&sp->auto_asconf_list,
4387 &net->sctp.auto_asconf_splist);
4388 sp->do_auto_asconf = 1;
4389 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4391 sp->do_auto_asconf = 0;
4399 /* Cleanup any SCTP per socket resources. Must be called with
4400 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4402 static void sctp_destroy_sock(struct sock *sk)
4404 struct sctp_sock *sp;
4406 pr_debug("%s: sk:%p\n", __func__, sk);
4408 /* Release our hold on the endpoint. */
4410 /* This could happen during socket init, thus we bail out
4411 * early, since the rest of the below is not setup either.
4416 if (sp->do_auto_asconf) {
4417 sp->do_auto_asconf = 0;
4418 list_del(&sp->auto_asconf_list);
4420 sctp_endpoint_free(sp->ep);
4422 percpu_counter_dec(&sctp_sockets_allocated);
4423 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4427 /* Triggered when there are no references on the socket anymore */
4428 static void sctp_destruct_sock(struct sock *sk)
4430 struct sctp_sock *sp = sctp_sk(sk);
4432 /* Free up the HMAC transform. */
4433 crypto_free_shash(sp->hmac);
4435 inet_sock_destruct(sk);
4438 /* API 4.1.7 shutdown() - TCP Style Syntax
4439 * int shutdown(int socket, int how);
4441 * sd - the socket descriptor of the association to be closed.
4442 * how - Specifies the type of shutdown. The values are
4445 * Disables further receive operations. No SCTP
4446 * protocol action is taken.
4448 * Disables further send operations, and initiates
4449 * the SCTP shutdown sequence.
4451 * Disables further send and receive operations
4452 * and initiates the SCTP shutdown sequence.
4454 static void sctp_shutdown(struct sock *sk, int how)
4456 struct net *net = sock_net(sk);
4457 struct sctp_endpoint *ep;
4459 if (!sctp_style(sk, TCP))
4462 ep = sctp_sk(sk)->ep;
4463 if (how & SEND_SHUTDOWN && !list_empty(&ep->asocs)) {
4464 struct sctp_association *asoc;
4466 sk->sk_state = SCTP_SS_CLOSING;
4467 asoc = list_entry(ep->asocs.next,
4468 struct sctp_association, asocs);
4469 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4473 int sctp_get_sctp_info(struct sock *sk, struct sctp_association *asoc,
4474 struct sctp_info *info)
4476 struct sctp_transport *prim;
4477 struct list_head *pos;
4480 memset(info, 0, sizeof(*info));
4482 struct sctp_sock *sp = sctp_sk(sk);
4484 info->sctpi_s_autoclose = sp->autoclose;
4485 info->sctpi_s_adaptation_ind = sp->adaptation_ind;
4486 info->sctpi_s_pd_point = sp->pd_point;
4487 info->sctpi_s_nodelay = sp->nodelay;
4488 info->sctpi_s_disable_fragments = sp->disable_fragments;
4489 info->sctpi_s_v4mapped = sp->v4mapped;
4490 info->sctpi_s_frag_interleave = sp->frag_interleave;
4491 info->sctpi_s_type = sp->type;
4496 info->sctpi_tag = asoc->c.my_vtag;
4497 info->sctpi_state = asoc->state;
4498 info->sctpi_rwnd = asoc->a_rwnd;
4499 info->sctpi_unackdata = asoc->unack_data;
4500 info->sctpi_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4501 info->sctpi_instrms = asoc->stream.incnt;
4502 info->sctpi_outstrms = asoc->stream.outcnt;
4503 list_for_each(pos, &asoc->base.inqueue.in_chunk_list)
4504 info->sctpi_inqueue++;
4505 list_for_each(pos, &asoc->outqueue.out_chunk_list)
4506 info->sctpi_outqueue++;
4507 info->sctpi_overall_error = asoc->overall_error_count;
4508 info->sctpi_max_burst = asoc->max_burst;
4509 info->sctpi_maxseg = asoc->frag_point;
4510 info->sctpi_peer_rwnd = asoc->peer.rwnd;
4511 info->sctpi_peer_tag = asoc->c.peer_vtag;
4513 mask = asoc->peer.ecn_capable << 1;
4514 mask = (mask | asoc->peer.ipv4_address) << 1;
4515 mask = (mask | asoc->peer.ipv6_address) << 1;
4516 mask = (mask | asoc->peer.hostname_address) << 1;
4517 mask = (mask | asoc->peer.asconf_capable) << 1;
4518 mask = (mask | asoc->peer.prsctp_capable) << 1;
4519 mask = (mask | asoc->peer.auth_capable);
4520 info->sctpi_peer_capable = mask;
4521 mask = asoc->peer.sack_needed << 1;
4522 mask = (mask | asoc->peer.sack_generation) << 1;
4523 mask = (mask | asoc->peer.zero_window_announced);
4524 info->sctpi_peer_sack = mask;
4526 info->sctpi_isacks = asoc->stats.isacks;
4527 info->sctpi_osacks = asoc->stats.osacks;
4528 info->sctpi_opackets = asoc->stats.opackets;
4529 info->sctpi_ipackets = asoc->stats.ipackets;
4530 info->sctpi_rtxchunks = asoc->stats.rtxchunks;
4531 info->sctpi_outofseqtsns = asoc->stats.outofseqtsns;
4532 info->sctpi_idupchunks = asoc->stats.idupchunks;
4533 info->sctpi_gapcnt = asoc->stats.gapcnt;
4534 info->sctpi_ouodchunks = asoc->stats.ouodchunks;
4535 info->sctpi_iuodchunks = asoc->stats.iuodchunks;
4536 info->sctpi_oodchunks = asoc->stats.oodchunks;
4537 info->sctpi_iodchunks = asoc->stats.iodchunks;
4538 info->sctpi_octrlchunks = asoc->stats.octrlchunks;
4539 info->sctpi_ictrlchunks = asoc->stats.ictrlchunks;
4541 prim = asoc->peer.primary_path;
4542 memcpy(&info->sctpi_p_address, &prim->ipaddr,
4543 sizeof(struct sockaddr_storage));
4544 info->sctpi_p_state = prim->state;
4545 info->sctpi_p_cwnd = prim->cwnd;
4546 info->sctpi_p_srtt = prim->srtt;
4547 info->sctpi_p_rto = jiffies_to_msecs(prim->rto);
4548 info->sctpi_p_hbinterval = prim->hbinterval;
4549 info->sctpi_p_pathmaxrxt = prim->pathmaxrxt;
4550 info->sctpi_p_sackdelay = jiffies_to_msecs(prim->sackdelay);
4551 info->sctpi_p_ssthresh = prim->ssthresh;
4552 info->sctpi_p_partial_bytes_acked = prim->partial_bytes_acked;
4553 info->sctpi_p_flight_size = prim->flight_size;
4554 info->sctpi_p_error = prim->error_count;
4558 EXPORT_SYMBOL_GPL(sctp_get_sctp_info);
4560 /* use callback to avoid exporting the core structure */
4561 int sctp_transport_walk_start(struct rhashtable_iter *iter)
4565 rhltable_walk_enter(&sctp_transport_hashtable, iter);
4567 err = rhashtable_walk_start(iter);
4568 if (err && err != -EAGAIN) {
4569 rhashtable_walk_stop(iter);
4570 rhashtable_walk_exit(iter);
4577 void sctp_transport_walk_stop(struct rhashtable_iter *iter)
4579 rhashtable_walk_stop(iter);
4580 rhashtable_walk_exit(iter);
4583 struct sctp_transport *sctp_transport_get_next(struct net *net,
4584 struct rhashtable_iter *iter)
4586 struct sctp_transport *t;
4588 t = rhashtable_walk_next(iter);
4589 for (; t; t = rhashtable_walk_next(iter)) {
4591 if (PTR_ERR(t) == -EAGAIN)
4596 if (net_eq(sock_net(t->asoc->base.sk), net) &&
4597 t->asoc->peer.primary_path == t)
4604 struct sctp_transport *sctp_transport_get_idx(struct net *net,
4605 struct rhashtable_iter *iter,
4608 void *obj = SEQ_START_TOKEN;
4610 while (pos && (obj = sctp_transport_get_next(net, iter)) &&
4617 int sctp_for_each_endpoint(int (*cb)(struct sctp_endpoint *, void *),
4621 struct sctp_ep_common *epb;
4622 struct sctp_hashbucket *head;
4624 for (head = sctp_ep_hashtable; hash < sctp_ep_hashsize;
4626 read_lock_bh(&head->lock);
4627 sctp_for_each_hentry(epb, &head->chain) {
4628 err = cb(sctp_ep(epb), p);
4632 read_unlock_bh(&head->lock);
4637 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint);
4639 int sctp_transport_lookup_process(int (*cb)(struct sctp_transport *, void *),
4641 const union sctp_addr *laddr,
4642 const union sctp_addr *paddr, void *p)
4644 struct sctp_transport *transport;
4648 transport = sctp_addrs_lookup_transport(net, laddr, paddr);
4653 err = cb(transport, p);
4654 sctp_transport_put(transport);
4658 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process);
4660 int sctp_for_each_transport(int (*cb)(struct sctp_transport *, void *),
4661 struct net *net, int pos, void *p) {
4662 struct rhashtable_iter hti;
4666 err = sctp_transport_walk_start(&hti);
4670 obj = sctp_transport_get_idx(net, &hti, pos + 1);
4671 for (; !IS_ERR_OR_NULL(obj); obj = sctp_transport_get_next(net, &hti)) {
4672 struct sctp_transport *transport = obj;
4674 if (!sctp_transport_hold(transport))
4676 err = cb(transport, p);
4677 sctp_transport_put(transport);
4681 sctp_transport_walk_stop(&hti);
4685 EXPORT_SYMBOL_GPL(sctp_for_each_transport);
4687 /* 7.2.1 Association Status (SCTP_STATUS)
4689 * Applications can retrieve current status information about an
4690 * association, including association state, peer receiver window size,
4691 * number of unacked data chunks, and number of data chunks pending
4692 * receipt. This information is read-only.
4694 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4695 char __user *optval,
4698 struct sctp_status status;
4699 struct sctp_association *asoc = NULL;
4700 struct sctp_transport *transport;
4701 sctp_assoc_t associd;
4704 if (len < sizeof(status)) {
4709 len = sizeof(status);
4710 if (copy_from_user(&status, optval, len)) {
4715 associd = status.sstat_assoc_id;
4716 asoc = sctp_id2assoc(sk, associd);
4722 transport = asoc->peer.primary_path;
4724 status.sstat_assoc_id = sctp_assoc2id(asoc);
4725 status.sstat_state = sctp_assoc_to_state(asoc);
4726 status.sstat_rwnd = asoc->peer.rwnd;
4727 status.sstat_unackdata = asoc->unack_data;
4729 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4730 status.sstat_instrms = asoc->stream.incnt;
4731 status.sstat_outstrms = asoc->stream.outcnt;
4732 status.sstat_fragmentation_point = asoc->frag_point;
4733 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4734 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4735 transport->af_specific->sockaddr_len);
4736 /* Map ipv4 address into v4-mapped-on-v6 address. */
4737 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
4738 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4739 status.sstat_primary.spinfo_state = transport->state;
4740 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4741 status.sstat_primary.spinfo_srtt = transport->srtt;
4742 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4743 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4745 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4746 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4748 if (put_user(len, optlen)) {
4753 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4754 __func__, len, status.sstat_state, status.sstat_rwnd,
4755 status.sstat_assoc_id);
4757 if (copy_to_user(optval, &status, len)) {
4767 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4769 * Applications can retrieve information about a specific peer address
4770 * of an association, including its reachability state, congestion
4771 * window, and retransmission timer values. This information is
4774 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4775 char __user *optval,
4778 struct sctp_paddrinfo pinfo;
4779 struct sctp_transport *transport;
4782 if (len < sizeof(pinfo)) {
4787 len = sizeof(pinfo);
4788 if (copy_from_user(&pinfo, optval, len)) {
4793 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4794 pinfo.spinfo_assoc_id);
4798 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4799 pinfo.spinfo_state = transport->state;
4800 pinfo.spinfo_cwnd = transport->cwnd;
4801 pinfo.spinfo_srtt = transport->srtt;
4802 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4803 pinfo.spinfo_mtu = transport->pathmtu;
4805 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4806 pinfo.spinfo_state = SCTP_ACTIVE;
4808 if (put_user(len, optlen)) {
4813 if (copy_to_user(optval, &pinfo, len)) {
4822 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4824 * This option is a on/off flag. If enabled no SCTP message
4825 * fragmentation will be performed. Instead if a message being sent
4826 * exceeds the current PMTU size, the message will NOT be sent and
4827 * instead a error will be indicated to the user.
4829 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4830 char __user *optval, int __user *optlen)
4834 if (len < sizeof(int))
4838 val = (sctp_sk(sk)->disable_fragments == 1);
4839 if (put_user(len, optlen))
4841 if (copy_to_user(optval, &val, len))
4846 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4848 * This socket option is used to specify various notifications and
4849 * ancillary data the user wishes to receive.
4851 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4856 if (len > sizeof(struct sctp_event_subscribe))
4857 len = sizeof(struct sctp_event_subscribe);
4858 if (put_user(len, optlen))
4860 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4865 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4867 * This socket option is applicable to the UDP-style socket only. When
4868 * set it will cause associations that are idle for more than the
4869 * specified number of seconds to automatically close. An association
4870 * being idle is defined an association that has NOT sent or received
4871 * user data. The special value of '0' indicates that no automatic
4872 * close of any associations should be performed. The option expects an
4873 * integer defining the number of seconds of idle time before an
4874 * association is closed.
4876 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4878 /* Applicable to UDP-style socket only */
4879 if (sctp_style(sk, TCP))
4881 if (len < sizeof(int))
4884 if (put_user(len, optlen))
4886 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4891 /* Helper routine to branch off an association to a new socket. */
4892 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4894 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4895 struct sctp_sock *sp = sctp_sk(sk);
4896 struct socket *sock;
4902 /* If there is a thread waiting on more sndbuf space for
4903 * sending on this asoc, it cannot be peeled.
4905 if (waitqueue_active(&asoc->wait))
4908 /* An association cannot be branched off from an already peeled-off
4909 * socket, nor is this supported for tcp style sockets.
4911 if (!sctp_style(sk, UDP))
4914 /* Create a new socket. */
4915 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4919 sctp_copy_sock(sock->sk, sk, asoc);
4921 /* Make peeled-off sockets more like 1-1 accepted sockets.
4922 * Set the daddr and initialize id to something more random
4924 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
4926 /* Populate the fields of the newsk from the oldsk and migrate the
4927 * asoc to the newsk.
4929 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4935 EXPORT_SYMBOL(sctp_do_peeloff);
4937 static int sctp_getsockopt_peeloff_common(struct sock *sk, sctp_peeloff_arg_t *peeloff,
4938 struct file **newfile, unsigned flags)
4940 struct socket *newsock;
4943 retval = sctp_do_peeloff(sk, peeloff->associd, &newsock);
4947 /* Map the socket to an unused fd that can be returned to the user. */
4948 retval = get_unused_fd_flags(flags & SOCK_CLOEXEC);
4950 sock_release(newsock);
4954 *newfile = sock_alloc_file(newsock, 0, NULL);
4955 if (IS_ERR(*newfile)) {
4956 put_unused_fd(retval);
4957 sock_release(newsock);
4958 retval = PTR_ERR(*newfile);
4963 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4966 peeloff->sd = retval;
4968 if (flags & SOCK_NONBLOCK)
4969 (*newfile)->f_flags |= O_NONBLOCK;
4974 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4976 sctp_peeloff_arg_t peeloff;
4977 struct file *newfile = NULL;
4980 if (len < sizeof(sctp_peeloff_arg_t))
4982 len = sizeof(sctp_peeloff_arg_t);
4983 if (copy_from_user(&peeloff, optval, len))
4986 retval = sctp_getsockopt_peeloff_common(sk, &peeloff, &newfile, 0);
4990 /* Return the fd mapped to the new socket. */
4991 if (put_user(len, optlen)) {
4993 put_unused_fd(retval);
4997 if (copy_to_user(optval, &peeloff, len)) {
4999 put_unused_fd(retval);
5002 fd_install(retval, newfile);
5007 static int sctp_getsockopt_peeloff_flags(struct sock *sk, int len,
5008 char __user *optval, int __user *optlen)
5010 sctp_peeloff_flags_arg_t peeloff;
5011 struct file *newfile = NULL;
5014 if (len < sizeof(sctp_peeloff_flags_arg_t))
5016 len = sizeof(sctp_peeloff_flags_arg_t);
5017 if (copy_from_user(&peeloff, optval, len))
5020 retval = sctp_getsockopt_peeloff_common(sk, &peeloff.p_arg,
5021 &newfile, peeloff.flags);
5025 /* Return the fd mapped to the new socket. */
5026 if (put_user(len, optlen)) {
5028 put_unused_fd(retval);
5032 if (copy_to_user(optval, &peeloff, len)) {
5034 put_unused_fd(retval);
5037 fd_install(retval, newfile);
5042 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
5044 * Applications can enable or disable heartbeats for any peer address of
5045 * an association, modify an address's heartbeat interval, force a
5046 * heartbeat to be sent immediately, and adjust the address's maximum
5047 * number of retransmissions sent before an address is considered
5048 * unreachable. The following structure is used to access and modify an
5049 * address's parameters:
5051 * struct sctp_paddrparams {
5052 * sctp_assoc_t spp_assoc_id;
5053 * struct sockaddr_storage spp_address;
5054 * uint32_t spp_hbinterval;
5055 * uint16_t spp_pathmaxrxt;
5056 * uint32_t spp_pathmtu;
5057 * uint32_t spp_sackdelay;
5058 * uint32_t spp_flags;
5061 * spp_assoc_id - (one-to-many style socket) This is filled in the
5062 * application, and identifies the association for
5064 * spp_address - This specifies which address is of interest.
5065 * spp_hbinterval - This contains the value of the heartbeat interval,
5066 * in milliseconds. If a value of zero
5067 * is present in this field then no changes are to
5068 * be made to this parameter.
5069 * spp_pathmaxrxt - This contains the maximum number of
5070 * retransmissions before this address shall be
5071 * considered unreachable. If a value of zero
5072 * is present in this field then no changes are to
5073 * be made to this parameter.
5074 * spp_pathmtu - When Path MTU discovery is disabled the value
5075 * specified here will be the "fixed" path mtu.
5076 * Note that if the spp_address field is empty
5077 * then all associations on this address will
5078 * have this fixed path mtu set upon them.
5080 * spp_sackdelay - When delayed sack is enabled, this value specifies
5081 * the number of milliseconds that sacks will be delayed
5082 * for. This value will apply to all addresses of an
5083 * association if the spp_address field is empty. Note
5084 * also, that if delayed sack is enabled and this
5085 * value is set to 0, no change is made to the last
5086 * recorded delayed sack timer value.
5088 * spp_flags - These flags are used to control various features
5089 * on an association. The flag field may contain
5090 * zero or more of the following options.
5092 * SPP_HB_ENABLE - Enable heartbeats on the
5093 * specified address. Note that if the address
5094 * field is empty all addresses for the association
5095 * have heartbeats enabled upon them.
5097 * SPP_HB_DISABLE - Disable heartbeats on the
5098 * speicifed address. Note that if the address
5099 * field is empty all addresses for the association
5100 * will have their heartbeats disabled. Note also
5101 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
5102 * mutually exclusive, only one of these two should
5103 * be specified. Enabling both fields will have
5104 * undetermined results.
5106 * SPP_HB_DEMAND - Request a user initiated heartbeat
5107 * to be made immediately.
5109 * SPP_PMTUD_ENABLE - This field will enable PMTU
5110 * discovery upon the specified address. Note that
5111 * if the address feild is empty then all addresses
5112 * on the association are effected.
5114 * SPP_PMTUD_DISABLE - This field will disable PMTU
5115 * discovery upon the specified address. Note that
5116 * if the address feild is empty then all addresses
5117 * on the association are effected. Not also that
5118 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
5119 * exclusive. Enabling both will have undetermined
5122 * SPP_SACKDELAY_ENABLE - Setting this flag turns
5123 * on delayed sack. The time specified in spp_sackdelay
5124 * is used to specify the sack delay for this address. Note
5125 * that if spp_address is empty then all addresses will
5126 * enable delayed sack and take on the sack delay
5127 * value specified in spp_sackdelay.
5128 * SPP_SACKDELAY_DISABLE - Setting this flag turns
5129 * off delayed sack. If the spp_address field is blank then
5130 * delayed sack is disabled for the entire association. Note
5131 * also that this field is mutually exclusive to
5132 * SPP_SACKDELAY_ENABLE, setting both will have undefined
5135 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
5136 char __user *optval, int __user *optlen)
5138 struct sctp_paddrparams params;
5139 struct sctp_transport *trans = NULL;
5140 struct sctp_association *asoc = NULL;
5141 struct sctp_sock *sp = sctp_sk(sk);
5143 if (len < sizeof(struct sctp_paddrparams))
5145 len = sizeof(struct sctp_paddrparams);
5146 if (copy_from_user(¶ms, optval, len))
5149 /* If an address other than INADDR_ANY is specified, and
5150 * no transport is found, then the request is invalid.
5152 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
5153 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
5154 params.spp_assoc_id);
5156 pr_debug("%s: failed no transport\n", __func__);
5161 /* Get association, if assoc_id != 0 and the socket is a one
5162 * to many style socket, and an association was not found, then
5163 * the id was invalid.
5165 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
5166 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
5167 pr_debug("%s: failed no association\n", __func__);
5172 /* Fetch transport values. */
5173 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
5174 params.spp_pathmtu = trans->pathmtu;
5175 params.spp_pathmaxrxt = trans->pathmaxrxt;
5176 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
5178 /*draft-11 doesn't say what to return in spp_flags*/
5179 params.spp_flags = trans->param_flags;
5181 /* Fetch association values. */
5182 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
5183 params.spp_pathmtu = asoc->pathmtu;
5184 params.spp_pathmaxrxt = asoc->pathmaxrxt;
5185 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
5187 /*draft-11 doesn't say what to return in spp_flags*/
5188 params.spp_flags = asoc->param_flags;
5190 /* Fetch socket values. */
5191 params.spp_hbinterval = sp->hbinterval;
5192 params.spp_pathmtu = sp->pathmtu;
5193 params.spp_sackdelay = sp->sackdelay;
5194 params.spp_pathmaxrxt = sp->pathmaxrxt;
5196 /*draft-11 doesn't say what to return in spp_flags*/
5197 params.spp_flags = sp->param_flags;
5200 if (copy_to_user(optval, ¶ms, len))
5203 if (put_user(len, optlen))
5210 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
5212 * This option will effect the way delayed acks are performed. This
5213 * option allows you to get or set the delayed ack time, in
5214 * milliseconds. It also allows changing the delayed ack frequency.
5215 * Changing the frequency to 1 disables the delayed sack algorithm. If
5216 * the assoc_id is 0, then this sets or gets the endpoints default
5217 * values. If the assoc_id field is non-zero, then the set or get
5218 * effects the specified association for the one to many model (the
5219 * assoc_id field is ignored by the one to one model). Note that if
5220 * sack_delay or sack_freq are 0 when setting this option, then the
5221 * current values will remain unchanged.
5223 * struct sctp_sack_info {
5224 * sctp_assoc_t sack_assoc_id;
5225 * uint32_t sack_delay;
5226 * uint32_t sack_freq;
5229 * sack_assoc_id - This parameter, indicates which association the user
5230 * is performing an action upon. Note that if this field's value is
5231 * zero then the endpoints default value is changed (effecting future
5232 * associations only).
5234 * sack_delay - This parameter contains the number of milliseconds that
5235 * the user is requesting the delayed ACK timer be set to. Note that
5236 * this value is defined in the standard to be between 200 and 500
5239 * sack_freq - This parameter contains the number of packets that must
5240 * be received before a sack is sent without waiting for the delay
5241 * timer to expire. The default value for this is 2, setting this
5242 * value to 1 will disable the delayed sack algorithm.
5244 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
5245 char __user *optval,
5248 struct sctp_sack_info params;
5249 struct sctp_association *asoc = NULL;
5250 struct sctp_sock *sp = sctp_sk(sk);
5252 if (len >= sizeof(struct sctp_sack_info)) {
5253 len = sizeof(struct sctp_sack_info);
5255 if (copy_from_user(¶ms, optval, len))
5257 } else if (len == sizeof(struct sctp_assoc_value)) {
5258 pr_warn_ratelimited(DEPRECATED
5260 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
5261 "Use struct sctp_sack_info instead\n",
5262 current->comm, task_pid_nr(current));
5263 if (copy_from_user(¶ms, optval, len))
5268 /* Get association, if sack_assoc_id != 0 and the socket is a one
5269 * to many style socket, and an association was not found, then
5270 * the id was invalid.
5272 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
5273 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
5277 /* Fetch association values. */
5278 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
5279 params.sack_delay = jiffies_to_msecs(
5281 params.sack_freq = asoc->sackfreq;
5284 params.sack_delay = 0;
5285 params.sack_freq = 1;
5288 /* Fetch socket values. */
5289 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
5290 params.sack_delay = sp->sackdelay;
5291 params.sack_freq = sp->sackfreq;
5293 params.sack_delay = 0;
5294 params.sack_freq = 1;
5298 if (copy_to_user(optval, ¶ms, len))
5301 if (put_user(len, optlen))
5307 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5309 * Applications can specify protocol parameters for the default association
5310 * initialization. The option name argument to setsockopt() and getsockopt()
5313 * Setting initialization parameters is effective only on an unconnected
5314 * socket (for UDP-style sockets only future associations are effected
5315 * by the change). With TCP-style sockets, this option is inherited by
5316 * sockets derived from a listener socket.
5318 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
5320 if (len < sizeof(struct sctp_initmsg))
5322 len = sizeof(struct sctp_initmsg);
5323 if (put_user(len, optlen))
5325 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
5331 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
5332 char __user *optval, int __user *optlen)
5334 struct sctp_association *asoc;
5336 struct sctp_getaddrs getaddrs;
5337 struct sctp_transport *from;
5339 union sctp_addr temp;
5340 struct sctp_sock *sp = sctp_sk(sk);
5345 if (len < sizeof(struct sctp_getaddrs))
5348 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5351 /* For UDP-style sockets, id specifies the association to query. */
5352 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5356 to = optval + offsetof(struct sctp_getaddrs, addrs);
5357 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5359 list_for_each_entry(from, &asoc->peer.transport_addr_list,
5361 memcpy(&temp, &from->ipaddr, sizeof(temp));
5362 addrlen = sctp_get_pf_specific(sk->sk_family)
5363 ->addr_to_user(sp, &temp);
5364 if (space_left < addrlen)
5366 if (copy_to_user(to, &temp, addrlen))
5370 space_left -= addrlen;
5373 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
5375 bytes_copied = ((char __user *)to) - optval;
5376 if (put_user(bytes_copied, optlen))
5382 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
5383 size_t space_left, int *bytes_copied)
5385 struct sctp_sockaddr_entry *addr;
5386 union sctp_addr temp;
5389 struct net *net = sock_net(sk);
5392 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
5396 if ((PF_INET == sk->sk_family) &&
5397 (AF_INET6 == addr->a.sa.sa_family))
5399 if ((PF_INET6 == sk->sk_family) &&
5400 inet_v6_ipv6only(sk) &&
5401 (AF_INET == addr->a.sa.sa_family))
5403 memcpy(&temp, &addr->a, sizeof(temp));
5404 if (!temp.v4.sin_port)
5405 temp.v4.sin_port = htons(port);
5407 addrlen = sctp_get_pf_specific(sk->sk_family)
5408 ->addr_to_user(sctp_sk(sk), &temp);
5410 if (space_left < addrlen) {
5414 memcpy(to, &temp, addrlen);
5418 space_left -= addrlen;
5419 *bytes_copied += addrlen;
5427 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
5428 char __user *optval, int __user *optlen)
5430 struct sctp_bind_addr *bp;
5431 struct sctp_association *asoc;
5433 struct sctp_getaddrs getaddrs;
5434 struct sctp_sockaddr_entry *addr;
5436 union sctp_addr temp;
5437 struct sctp_sock *sp = sctp_sk(sk);
5441 int bytes_copied = 0;
5445 if (len < sizeof(struct sctp_getaddrs))
5448 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5452 * For UDP-style sockets, id specifies the association to query.
5453 * If the id field is set to the value '0' then the locally bound
5454 * addresses are returned without regard to any particular
5457 if (0 == getaddrs.assoc_id) {
5458 bp = &sctp_sk(sk)->ep->base.bind_addr;
5460 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5463 bp = &asoc->base.bind_addr;
5466 to = optval + offsetof(struct sctp_getaddrs, addrs);
5467 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5469 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN);
5473 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5474 * addresses from the global local address list.
5476 if (sctp_list_single_entry(&bp->address_list)) {
5477 addr = list_entry(bp->address_list.next,
5478 struct sctp_sockaddr_entry, list);
5479 if (sctp_is_any(sk, &addr->a)) {
5480 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
5481 space_left, &bytes_copied);
5491 /* Protection on the bound address list is not needed since
5492 * in the socket option context we hold a socket lock and
5493 * thus the bound address list can't change.
5495 list_for_each_entry(addr, &bp->address_list, list) {
5496 memcpy(&temp, &addr->a, sizeof(temp));
5497 addrlen = sctp_get_pf_specific(sk->sk_family)
5498 ->addr_to_user(sp, &temp);
5499 if (space_left < addrlen) {
5500 err = -ENOMEM; /*fixme: right error?*/
5503 memcpy(buf, &temp, addrlen);
5505 bytes_copied += addrlen;
5507 space_left -= addrlen;
5511 if (copy_to_user(to, addrs, bytes_copied)) {
5515 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
5519 if (put_user(bytes_copied, optlen))
5526 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
5528 * Requests that the local SCTP stack use the enclosed peer address as
5529 * the association primary. The enclosed address must be one of the
5530 * association peer's addresses.
5532 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
5533 char __user *optval, int __user *optlen)
5535 struct sctp_prim prim;
5536 struct sctp_association *asoc;
5537 struct sctp_sock *sp = sctp_sk(sk);
5539 if (len < sizeof(struct sctp_prim))
5542 len = sizeof(struct sctp_prim);
5544 if (copy_from_user(&prim, optval, len))
5547 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
5551 if (!asoc->peer.primary_path)
5554 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
5555 asoc->peer.primary_path->af_specific->sockaddr_len);
5557 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
5558 (union sctp_addr *)&prim.ssp_addr);
5560 if (put_user(len, optlen))
5562 if (copy_to_user(optval, &prim, len))
5569 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5571 * Requests that the local endpoint set the specified Adaptation Layer
5572 * Indication parameter for all future INIT and INIT-ACK exchanges.
5574 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
5575 char __user *optval, int __user *optlen)
5577 struct sctp_setadaptation adaptation;
5579 if (len < sizeof(struct sctp_setadaptation))
5582 len = sizeof(struct sctp_setadaptation);
5584 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
5586 if (put_user(len, optlen))
5588 if (copy_to_user(optval, &adaptation, len))
5596 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5598 * Applications that wish to use the sendto() system call may wish to
5599 * specify a default set of parameters that would normally be supplied
5600 * through the inclusion of ancillary data. This socket option allows
5601 * such an application to set the default sctp_sndrcvinfo structure.
5604 * The application that wishes to use this socket option simply passes
5605 * in to this call the sctp_sndrcvinfo structure defined in Section
5606 * 5.2.2) The input parameters accepted by this call include
5607 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5608 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5609 * to this call if the caller is using the UDP model.
5611 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5613 static int sctp_getsockopt_default_send_param(struct sock *sk,
5614 int len, char __user *optval,
5617 struct sctp_sock *sp = sctp_sk(sk);
5618 struct sctp_association *asoc;
5619 struct sctp_sndrcvinfo info;
5621 if (len < sizeof(info))
5626 if (copy_from_user(&info, optval, len))
5629 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
5630 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
5633 info.sinfo_stream = asoc->default_stream;
5634 info.sinfo_flags = asoc->default_flags;
5635 info.sinfo_ppid = asoc->default_ppid;
5636 info.sinfo_context = asoc->default_context;
5637 info.sinfo_timetolive = asoc->default_timetolive;
5639 info.sinfo_stream = sp->default_stream;
5640 info.sinfo_flags = sp->default_flags;
5641 info.sinfo_ppid = sp->default_ppid;
5642 info.sinfo_context = sp->default_context;
5643 info.sinfo_timetolive = sp->default_timetolive;
5646 if (put_user(len, optlen))
5648 if (copy_to_user(optval, &info, len))
5654 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5655 * (SCTP_DEFAULT_SNDINFO)
5657 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
5658 char __user *optval,
5661 struct sctp_sock *sp = sctp_sk(sk);
5662 struct sctp_association *asoc;
5663 struct sctp_sndinfo info;
5665 if (len < sizeof(info))
5670 if (copy_from_user(&info, optval, len))
5673 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
5674 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
5677 info.snd_sid = asoc->default_stream;
5678 info.snd_flags = asoc->default_flags;
5679 info.snd_ppid = asoc->default_ppid;
5680 info.snd_context = asoc->default_context;
5682 info.snd_sid = sp->default_stream;
5683 info.snd_flags = sp->default_flags;
5684 info.snd_ppid = sp->default_ppid;
5685 info.snd_context = sp->default_context;
5688 if (put_user(len, optlen))
5690 if (copy_to_user(optval, &info, len))
5698 * 7.1.5 SCTP_NODELAY
5700 * Turn on/off any Nagle-like algorithm. This means that packets are
5701 * generally sent as soon as possible and no unnecessary delays are
5702 * introduced, at the cost of more packets in the network. Expects an
5703 * integer boolean flag.
5706 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5707 char __user *optval, int __user *optlen)
5711 if (len < sizeof(int))
5715 val = (sctp_sk(sk)->nodelay == 1);
5716 if (put_user(len, optlen))
5718 if (copy_to_user(optval, &val, len))
5725 * 7.1.1 SCTP_RTOINFO
5727 * The protocol parameters used to initialize and bound retransmission
5728 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5729 * and modify these parameters.
5730 * All parameters are time values, in milliseconds. A value of 0, when
5731 * modifying the parameters, indicates that the current value should not
5735 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5736 char __user *optval,
5737 int __user *optlen) {
5738 struct sctp_rtoinfo rtoinfo;
5739 struct sctp_association *asoc;
5741 if (len < sizeof (struct sctp_rtoinfo))
5744 len = sizeof(struct sctp_rtoinfo);
5746 if (copy_from_user(&rtoinfo, optval, len))
5749 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5751 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5754 /* Values corresponding to the specific association. */
5756 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5757 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5758 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5760 /* Values corresponding to the endpoint. */
5761 struct sctp_sock *sp = sctp_sk(sk);
5763 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5764 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5765 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5768 if (put_user(len, optlen))
5771 if (copy_to_user(optval, &rtoinfo, len))
5779 * 7.1.2 SCTP_ASSOCINFO
5781 * This option is used to tune the maximum retransmission attempts
5782 * of the association.
5783 * Returns an error if the new association retransmission value is
5784 * greater than the sum of the retransmission value of the peer.
5785 * See [SCTP] for more information.
5788 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5789 char __user *optval,
5793 struct sctp_assocparams assocparams;
5794 struct sctp_association *asoc;
5795 struct list_head *pos;
5798 if (len < sizeof (struct sctp_assocparams))
5801 len = sizeof(struct sctp_assocparams);
5803 if (copy_from_user(&assocparams, optval, len))
5806 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5808 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5811 /* Values correspoinding to the specific association */
5813 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5814 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5815 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5816 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5818 list_for_each(pos, &asoc->peer.transport_addr_list) {
5822 assocparams.sasoc_number_peer_destinations = cnt;
5824 /* Values corresponding to the endpoint */
5825 struct sctp_sock *sp = sctp_sk(sk);
5827 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5828 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5829 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5830 assocparams.sasoc_cookie_life =
5831 sp->assocparams.sasoc_cookie_life;
5832 assocparams.sasoc_number_peer_destinations =
5834 sasoc_number_peer_destinations;
5837 if (put_user(len, optlen))
5840 if (copy_to_user(optval, &assocparams, len))
5847 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5849 * This socket option is a boolean flag which turns on or off mapped V4
5850 * addresses. If this option is turned on and the socket is type
5851 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5852 * If this option is turned off, then no mapping will be done of V4
5853 * addresses and a user will receive both PF_INET6 and PF_INET type
5854 * addresses on the socket.
5856 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5857 char __user *optval, int __user *optlen)
5860 struct sctp_sock *sp = sctp_sk(sk);
5862 if (len < sizeof(int))
5867 if (put_user(len, optlen))
5869 if (copy_to_user(optval, &val, len))
5876 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5877 * (chapter and verse is quoted at sctp_setsockopt_context())
5879 static int sctp_getsockopt_context(struct sock *sk, int len,
5880 char __user *optval, int __user *optlen)
5882 struct sctp_assoc_value params;
5883 struct sctp_sock *sp;
5884 struct sctp_association *asoc;
5886 if (len < sizeof(struct sctp_assoc_value))
5889 len = sizeof(struct sctp_assoc_value);
5891 if (copy_from_user(¶ms, optval, len))
5896 if (params.assoc_id != 0) {
5897 asoc = sctp_id2assoc(sk, params.assoc_id);
5900 params.assoc_value = asoc->default_rcv_context;
5902 params.assoc_value = sp->default_rcv_context;
5905 if (put_user(len, optlen))
5907 if (copy_to_user(optval, ¶ms, len))
5914 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5915 * This option will get or set the maximum size to put in any outgoing
5916 * SCTP DATA chunk. If a message is larger than this size it will be
5917 * fragmented by SCTP into the specified size. Note that the underlying
5918 * SCTP implementation may fragment into smaller sized chunks when the
5919 * PMTU of the underlying association is smaller than the value set by
5920 * the user. The default value for this option is '0' which indicates
5921 * the user is NOT limiting fragmentation and only the PMTU will effect
5922 * SCTP's choice of DATA chunk size. Note also that values set larger
5923 * than the maximum size of an IP datagram will effectively let SCTP
5924 * control fragmentation (i.e. the same as setting this option to 0).
5926 * The following structure is used to access and modify this parameter:
5928 * struct sctp_assoc_value {
5929 * sctp_assoc_t assoc_id;
5930 * uint32_t assoc_value;
5933 * assoc_id: This parameter is ignored for one-to-one style sockets.
5934 * For one-to-many style sockets this parameter indicates which
5935 * association the user is performing an action upon. Note that if
5936 * this field's value is zero then the endpoints default value is
5937 * changed (effecting future associations only).
5938 * assoc_value: This parameter specifies the maximum size in bytes.
5940 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5941 char __user *optval, int __user *optlen)
5943 struct sctp_assoc_value params;
5944 struct sctp_association *asoc;
5946 if (len == sizeof(int)) {
5947 pr_warn_ratelimited(DEPRECATED
5949 "Use of int in maxseg socket option.\n"
5950 "Use struct sctp_assoc_value instead\n",
5951 current->comm, task_pid_nr(current));
5952 params.assoc_id = 0;
5953 } else if (len >= sizeof(struct sctp_assoc_value)) {
5954 len = sizeof(struct sctp_assoc_value);
5955 if (copy_from_user(¶ms, optval, sizeof(params)))
5960 asoc = sctp_id2assoc(sk, params.assoc_id);
5961 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5965 params.assoc_value = asoc->frag_point;
5967 params.assoc_value = sctp_sk(sk)->user_frag;
5969 if (put_user(len, optlen))
5971 if (len == sizeof(int)) {
5972 if (copy_to_user(optval, ¶ms.assoc_value, len))
5975 if (copy_to_user(optval, ¶ms, len))
5983 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5984 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5986 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5987 char __user *optval, int __user *optlen)
5991 if (len < sizeof(int))
5996 val = sctp_sk(sk)->frag_interleave;
5997 if (put_user(len, optlen))
5999 if (copy_to_user(optval, &val, len))
6006 * 7.1.25. Set or Get the sctp partial delivery point
6007 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
6009 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
6010 char __user *optval,
6015 if (len < sizeof(u32))
6020 val = sctp_sk(sk)->pd_point;
6021 if (put_user(len, optlen))
6023 if (copy_to_user(optval, &val, len))
6030 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
6031 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
6033 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
6034 char __user *optval,
6037 struct sctp_assoc_value params;
6038 struct sctp_sock *sp;
6039 struct sctp_association *asoc;
6041 if (len == sizeof(int)) {
6042 pr_warn_ratelimited(DEPRECATED
6044 "Use of int in max_burst socket option.\n"
6045 "Use struct sctp_assoc_value instead\n",
6046 current->comm, task_pid_nr(current));
6047 params.assoc_id = 0;
6048 } else if (len >= sizeof(struct sctp_assoc_value)) {
6049 len = sizeof(struct sctp_assoc_value);
6050 if (copy_from_user(¶ms, optval, len))
6057 if (params.assoc_id != 0) {
6058 asoc = sctp_id2assoc(sk, params.assoc_id);
6061 params.assoc_value = asoc->max_burst;
6063 params.assoc_value = sp->max_burst;
6065 if (len == sizeof(int)) {
6066 if (copy_to_user(optval, ¶ms.assoc_value, len))
6069 if (copy_to_user(optval, ¶ms, len))
6077 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
6078 char __user *optval, int __user *optlen)
6080 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6081 struct sctp_hmacalgo __user *p = (void __user *)optval;
6082 struct sctp_hmac_algo_param *hmacs;
6087 if (!ep->auth_enable)
6090 hmacs = ep->auth_hmacs_list;
6091 data_len = ntohs(hmacs->param_hdr.length) -
6092 sizeof(struct sctp_paramhdr);
6094 if (len < sizeof(struct sctp_hmacalgo) + data_len)
6097 len = sizeof(struct sctp_hmacalgo) + data_len;
6098 num_idents = data_len / sizeof(u16);
6100 if (put_user(len, optlen))
6102 if (put_user(num_idents, &p->shmac_num_idents))
6104 for (i = 0; i < num_idents; i++) {
6105 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
6107 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
6113 static int sctp_getsockopt_active_key(struct sock *sk, int len,
6114 char __user *optval, int __user *optlen)
6116 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6117 struct sctp_authkeyid val;
6118 struct sctp_association *asoc;
6120 if (!ep->auth_enable)
6123 if (len < sizeof(struct sctp_authkeyid))
6125 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
6128 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
6129 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
6133 val.scact_keynumber = asoc->active_key_id;
6135 val.scact_keynumber = ep->active_key_id;
6137 len = sizeof(struct sctp_authkeyid);
6138 if (put_user(len, optlen))
6140 if (copy_to_user(optval, &val, len))
6146 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
6147 char __user *optval, int __user *optlen)
6149 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6150 struct sctp_authchunks __user *p = (void __user *)optval;
6151 struct sctp_authchunks val;
6152 struct sctp_association *asoc;
6153 struct sctp_chunks_param *ch;
6157 if (!ep->auth_enable)
6160 if (len < sizeof(struct sctp_authchunks))
6163 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
6166 to = p->gauth_chunks;
6167 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6171 ch = asoc->peer.peer_chunks;
6175 /* See if the user provided enough room for all the data */
6176 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6177 if (len < num_chunks)
6180 if (copy_to_user(to, ch->chunks, num_chunks))
6183 len = sizeof(struct sctp_authchunks) + num_chunks;
6184 if (put_user(len, optlen))
6186 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6191 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
6192 char __user *optval, int __user *optlen)
6194 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6195 struct sctp_authchunks __user *p = (void __user *)optval;
6196 struct sctp_authchunks val;
6197 struct sctp_association *asoc;
6198 struct sctp_chunks_param *ch;
6202 if (!ep->auth_enable)
6205 if (len < sizeof(struct sctp_authchunks))
6208 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
6211 to = p->gauth_chunks;
6212 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6213 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
6217 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
6219 ch = ep->auth_chunk_list;
6224 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6225 if (len < sizeof(struct sctp_authchunks) + num_chunks)
6228 if (copy_to_user(to, ch->chunks, num_chunks))
6231 len = sizeof(struct sctp_authchunks) + num_chunks;
6232 if (put_user(len, optlen))
6234 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6241 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
6242 * This option gets the current number of associations that are attached
6243 * to a one-to-many style socket. The option value is an uint32_t.
6245 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
6246 char __user *optval, int __user *optlen)
6248 struct sctp_sock *sp = sctp_sk(sk);
6249 struct sctp_association *asoc;
6252 if (sctp_style(sk, TCP))
6255 if (len < sizeof(u32))
6260 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6264 if (put_user(len, optlen))
6266 if (copy_to_user(optval, &val, len))
6273 * 8.1.23 SCTP_AUTO_ASCONF
6274 * See the corresponding setsockopt entry as description
6276 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
6277 char __user *optval, int __user *optlen)
6281 if (len < sizeof(int))
6285 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
6287 if (put_user(len, optlen))
6289 if (copy_to_user(optval, &val, len))
6295 * 8.2.6. Get the Current Identifiers of Associations
6296 * (SCTP_GET_ASSOC_ID_LIST)
6298 * This option gets the current list of SCTP association identifiers of
6299 * the SCTP associations handled by a one-to-many style socket.
6301 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
6302 char __user *optval, int __user *optlen)
6304 struct sctp_sock *sp = sctp_sk(sk);
6305 struct sctp_association *asoc;
6306 struct sctp_assoc_ids *ids;
6309 if (sctp_style(sk, TCP))
6312 if (len < sizeof(struct sctp_assoc_ids))
6315 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6319 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
6322 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
6324 ids = kmalloc(len, GFP_USER | __GFP_NOWARN);
6328 ids->gaids_number_of_ids = num;
6330 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6331 ids->gaids_assoc_id[num++] = asoc->assoc_id;
6334 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
6344 * SCTP_PEER_ADDR_THLDS
6346 * This option allows us to fetch the partially failed threshold for one or all
6347 * transports in an association. See Section 6.1 of:
6348 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6350 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
6351 char __user *optval,
6355 struct sctp_paddrthlds val;
6356 struct sctp_transport *trans;
6357 struct sctp_association *asoc;
6359 if (len < sizeof(struct sctp_paddrthlds))
6361 len = sizeof(struct sctp_paddrthlds);
6362 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
6365 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
6366 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
6370 val.spt_pathpfthld = asoc->pf_retrans;
6371 val.spt_pathmaxrxt = asoc->pathmaxrxt;
6373 trans = sctp_addr_id2transport(sk, &val.spt_address,
6378 val.spt_pathmaxrxt = trans->pathmaxrxt;
6379 val.spt_pathpfthld = trans->pf_retrans;
6382 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
6389 * SCTP_GET_ASSOC_STATS
6391 * This option retrieves local per endpoint statistics. It is modeled
6392 * after OpenSolaris' implementation
6394 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
6395 char __user *optval,
6398 struct sctp_assoc_stats sas;
6399 struct sctp_association *asoc = NULL;
6401 /* User must provide at least the assoc id */
6402 if (len < sizeof(sctp_assoc_t))
6405 /* Allow the struct to grow and fill in as much as possible */
6406 len = min_t(size_t, len, sizeof(sas));
6408 if (copy_from_user(&sas, optval, len))
6411 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
6415 sas.sas_rtxchunks = asoc->stats.rtxchunks;
6416 sas.sas_gapcnt = asoc->stats.gapcnt;
6417 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
6418 sas.sas_osacks = asoc->stats.osacks;
6419 sas.sas_isacks = asoc->stats.isacks;
6420 sas.sas_octrlchunks = asoc->stats.octrlchunks;
6421 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
6422 sas.sas_oodchunks = asoc->stats.oodchunks;
6423 sas.sas_iodchunks = asoc->stats.iodchunks;
6424 sas.sas_ouodchunks = asoc->stats.ouodchunks;
6425 sas.sas_iuodchunks = asoc->stats.iuodchunks;
6426 sas.sas_idupchunks = asoc->stats.idupchunks;
6427 sas.sas_opackets = asoc->stats.opackets;
6428 sas.sas_ipackets = asoc->stats.ipackets;
6430 /* New high max rto observed, will return 0 if not a single
6431 * RTO update took place. obs_rto_ipaddr will be bogus
6434 sas.sas_maxrto = asoc->stats.max_obs_rto;
6435 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
6436 sizeof(struct sockaddr_storage));
6438 /* Mark beginning of a new observation period */
6439 asoc->stats.max_obs_rto = asoc->rto_min;
6441 if (put_user(len, optlen))
6444 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
6446 if (copy_to_user(optval, &sas, len))
6452 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
6453 char __user *optval,
6458 if (len < sizeof(int))
6462 if (sctp_sk(sk)->recvrcvinfo)
6464 if (put_user(len, optlen))
6466 if (copy_to_user(optval, &val, len))
6472 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
6473 char __user *optval,
6478 if (len < sizeof(int))
6482 if (sctp_sk(sk)->recvnxtinfo)
6484 if (put_user(len, optlen))
6486 if (copy_to_user(optval, &val, len))
6492 static int sctp_getsockopt_pr_supported(struct sock *sk, int len,
6493 char __user *optval,
6496 struct sctp_assoc_value params;
6497 struct sctp_association *asoc;
6498 int retval = -EFAULT;
6500 if (len < sizeof(params)) {
6505 len = sizeof(params);
6506 if (copy_from_user(¶ms, optval, len))
6509 asoc = sctp_id2assoc(sk, params.assoc_id);
6511 params.assoc_value = asoc->prsctp_enable;
6512 } else if (!params.assoc_id) {
6513 struct sctp_sock *sp = sctp_sk(sk);
6515 params.assoc_value = sp->ep->prsctp_enable;
6521 if (put_user(len, optlen))
6524 if (copy_to_user(optval, ¶ms, len))
6533 static int sctp_getsockopt_default_prinfo(struct sock *sk, int len,
6534 char __user *optval,
6537 struct sctp_default_prinfo info;
6538 struct sctp_association *asoc;
6539 int retval = -EFAULT;
6541 if (len < sizeof(info)) {
6547 if (copy_from_user(&info, optval, len))
6550 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
6552 info.pr_policy = SCTP_PR_POLICY(asoc->default_flags);
6553 info.pr_value = asoc->default_timetolive;
6554 } else if (!info.pr_assoc_id) {
6555 struct sctp_sock *sp = sctp_sk(sk);
6557 info.pr_policy = SCTP_PR_POLICY(sp->default_flags);
6558 info.pr_value = sp->default_timetolive;
6564 if (put_user(len, optlen))
6567 if (copy_to_user(optval, &info, len))
6576 static int sctp_getsockopt_pr_assocstatus(struct sock *sk, int len,
6577 char __user *optval,
6580 struct sctp_prstatus params;
6581 struct sctp_association *asoc;
6583 int retval = -EINVAL;
6585 if (len < sizeof(params))
6588 len = sizeof(params);
6589 if (copy_from_user(¶ms, optval, len)) {
6594 policy = params.sprstat_policy;
6595 if (policy & ~SCTP_PR_SCTP_MASK)
6598 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
6602 if (policy == SCTP_PR_SCTP_NONE) {
6603 params.sprstat_abandoned_unsent = 0;
6604 params.sprstat_abandoned_sent = 0;
6605 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
6606 params.sprstat_abandoned_unsent +=
6607 asoc->abandoned_unsent[policy];
6608 params.sprstat_abandoned_sent +=
6609 asoc->abandoned_sent[policy];
6612 params.sprstat_abandoned_unsent =
6613 asoc->abandoned_unsent[__SCTP_PR_INDEX(policy)];
6614 params.sprstat_abandoned_sent =
6615 asoc->abandoned_sent[__SCTP_PR_INDEX(policy)];
6618 if (put_user(len, optlen)) {
6623 if (copy_to_user(optval, ¶ms, len)) {
6634 static int sctp_getsockopt_pr_streamstatus(struct sock *sk, int len,
6635 char __user *optval,
6638 struct sctp_stream_out *streamout;
6639 struct sctp_association *asoc;
6640 struct sctp_prstatus params;
6641 int retval = -EINVAL;
6644 if (len < sizeof(params))
6647 len = sizeof(params);
6648 if (copy_from_user(¶ms, optval, len)) {
6653 policy = params.sprstat_policy;
6654 if (policy & ~SCTP_PR_SCTP_MASK)
6657 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
6658 if (!asoc || params.sprstat_sid >= asoc->stream.outcnt)
6661 streamout = &asoc->stream.out[params.sprstat_sid];
6662 if (policy == SCTP_PR_SCTP_NONE) {
6663 params.sprstat_abandoned_unsent = 0;
6664 params.sprstat_abandoned_sent = 0;
6665 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
6666 params.sprstat_abandoned_unsent +=
6667 streamout->abandoned_unsent[policy];
6668 params.sprstat_abandoned_sent +=
6669 streamout->abandoned_sent[policy];
6672 params.sprstat_abandoned_unsent =
6673 streamout->abandoned_unsent[__SCTP_PR_INDEX(policy)];
6674 params.sprstat_abandoned_sent =
6675 streamout->abandoned_sent[__SCTP_PR_INDEX(policy)];
6678 if (put_user(len, optlen) || copy_to_user(optval, ¶ms, len)) {
6689 static int sctp_getsockopt_reconfig_supported(struct sock *sk, int len,
6690 char __user *optval,
6693 struct sctp_assoc_value params;
6694 struct sctp_association *asoc;
6695 int retval = -EFAULT;
6697 if (len < sizeof(params)) {
6702 len = sizeof(params);
6703 if (copy_from_user(¶ms, optval, len))
6706 asoc = sctp_id2assoc(sk, params.assoc_id);
6708 params.assoc_value = asoc->reconf_enable;
6709 } else if (!params.assoc_id) {
6710 struct sctp_sock *sp = sctp_sk(sk);
6712 params.assoc_value = sp->ep->reconf_enable;
6718 if (put_user(len, optlen))
6721 if (copy_to_user(optval, ¶ms, len))
6730 static int sctp_getsockopt_enable_strreset(struct sock *sk, int len,
6731 char __user *optval,
6734 struct sctp_assoc_value params;
6735 struct sctp_association *asoc;
6736 int retval = -EFAULT;
6738 if (len < sizeof(params)) {
6743 len = sizeof(params);
6744 if (copy_from_user(¶ms, optval, len))
6747 asoc = sctp_id2assoc(sk, params.assoc_id);
6749 params.assoc_value = asoc->strreset_enable;
6750 } else if (!params.assoc_id) {
6751 struct sctp_sock *sp = sctp_sk(sk);
6753 params.assoc_value = sp->ep->strreset_enable;
6759 if (put_user(len, optlen))
6762 if (copy_to_user(optval, ¶ms, len))
6771 static int sctp_getsockopt(struct sock *sk, int level, int optname,
6772 char __user *optval, int __user *optlen)
6777 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
6779 /* I can hardly begin to describe how wrong this is. This is
6780 * so broken as to be worse than useless. The API draft
6781 * REALLY is NOT helpful here... I am not convinced that the
6782 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
6783 * are at all well-founded.
6785 if (level != SOL_SCTP) {
6786 struct sctp_af *af = sctp_sk(sk)->pf->af;
6788 retval = af->getsockopt(sk, level, optname, optval, optlen);
6792 if (get_user(len, optlen))
6802 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
6804 case SCTP_DISABLE_FRAGMENTS:
6805 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
6809 retval = sctp_getsockopt_events(sk, len, optval, optlen);
6811 case SCTP_AUTOCLOSE:
6812 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
6814 case SCTP_SOCKOPT_PEELOFF:
6815 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
6817 case SCTP_SOCKOPT_PEELOFF_FLAGS:
6818 retval = sctp_getsockopt_peeloff_flags(sk, len, optval, optlen);
6820 case SCTP_PEER_ADDR_PARAMS:
6821 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
6824 case SCTP_DELAYED_SACK:
6825 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
6829 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
6831 case SCTP_GET_PEER_ADDRS:
6832 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
6835 case SCTP_GET_LOCAL_ADDRS:
6836 retval = sctp_getsockopt_local_addrs(sk, len, optval,
6839 case SCTP_SOCKOPT_CONNECTX3:
6840 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
6842 case SCTP_DEFAULT_SEND_PARAM:
6843 retval = sctp_getsockopt_default_send_param(sk, len,
6846 case SCTP_DEFAULT_SNDINFO:
6847 retval = sctp_getsockopt_default_sndinfo(sk, len,
6850 case SCTP_PRIMARY_ADDR:
6851 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
6854 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
6857 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
6859 case SCTP_ASSOCINFO:
6860 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
6862 case SCTP_I_WANT_MAPPED_V4_ADDR:
6863 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
6866 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
6868 case SCTP_GET_PEER_ADDR_INFO:
6869 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
6872 case SCTP_ADAPTATION_LAYER:
6873 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
6877 retval = sctp_getsockopt_context(sk, len, optval, optlen);
6879 case SCTP_FRAGMENT_INTERLEAVE:
6880 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
6883 case SCTP_PARTIAL_DELIVERY_POINT:
6884 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
6887 case SCTP_MAX_BURST:
6888 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
6891 case SCTP_AUTH_CHUNK:
6892 case SCTP_AUTH_DELETE_KEY:
6893 retval = -EOPNOTSUPP;
6895 case SCTP_HMAC_IDENT:
6896 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
6898 case SCTP_AUTH_ACTIVE_KEY:
6899 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
6901 case SCTP_PEER_AUTH_CHUNKS:
6902 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
6905 case SCTP_LOCAL_AUTH_CHUNKS:
6906 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
6909 case SCTP_GET_ASSOC_NUMBER:
6910 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
6912 case SCTP_GET_ASSOC_ID_LIST:
6913 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
6915 case SCTP_AUTO_ASCONF:
6916 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
6918 case SCTP_PEER_ADDR_THLDS:
6919 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
6921 case SCTP_GET_ASSOC_STATS:
6922 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
6924 case SCTP_RECVRCVINFO:
6925 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
6927 case SCTP_RECVNXTINFO:
6928 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
6930 case SCTP_PR_SUPPORTED:
6931 retval = sctp_getsockopt_pr_supported(sk, len, optval, optlen);
6933 case SCTP_DEFAULT_PRINFO:
6934 retval = sctp_getsockopt_default_prinfo(sk, len, optval,
6937 case SCTP_PR_ASSOC_STATUS:
6938 retval = sctp_getsockopt_pr_assocstatus(sk, len, optval,
6941 case SCTP_PR_STREAM_STATUS:
6942 retval = sctp_getsockopt_pr_streamstatus(sk, len, optval,
6945 case SCTP_RECONFIG_SUPPORTED:
6946 retval = sctp_getsockopt_reconfig_supported(sk, len, optval,
6949 case SCTP_ENABLE_STREAM_RESET:
6950 retval = sctp_getsockopt_enable_strreset(sk, len, optval,
6954 retval = -ENOPROTOOPT;
6962 static int sctp_hash(struct sock *sk)
6968 static void sctp_unhash(struct sock *sk)
6973 /* Check if port is acceptable. Possibly find first available port.
6975 * The port hash table (contained in the 'global' SCTP protocol storage
6976 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
6977 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
6978 * list (the list number is the port number hashed out, so as you
6979 * would expect from a hash function, all the ports in a given list have
6980 * such a number that hashes out to the same list number; you were
6981 * expecting that, right?); so each list has a set of ports, with a
6982 * link to the socket (struct sock) that uses it, the port number and
6983 * a fastreuse flag (FIXME: NPI ipg).
6985 static struct sctp_bind_bucket *sctp_bucket_create(
6986 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
6988 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
6990 struct sctp_bind_hashbucket *head; /* hash list */
6991 struct sctp_bind_bucket *pp;
6992 unsigned short snum;
6995 snum = ntohs(addr->v4.sin_port);
6997 pr_debug("%s: begins, snum:%d\n", __func__, snum);
7002 /* Search for an available port. */
7003 int low, high, remaining, index;
7005 struct net *net = sock_net(sk);
7007 inet_get_local_port_range(net, &low, &high);
7008 remaining = (high - low) + 1;
7009 rover = prandom_u32() % remaining + low;
7013 if ((rover < low) || (rover > high))
7015 if (inet_is_local_reserved_port(net, rover))
7017 index = sctp_phashfn(sock_net(sk), rover);
7018 head = &sctp_port_hashtable[index];
7019 spin_lock(&head->lock);
7020 sctp_for_each_hentry(pp, &head->chain)
7021 if ((pp->port == rover) &&
7022 net_eq(sock_net(sk), pp->net))
7026 spin_unlock(&head->lock);
7027 } while (--remaining > 0);
7029 /* Exhausted local port range during search? */
7034 /* OK, here is the one we will use. HEAD (the port
7035 * hash table list entry) is non-NULL and we hold it's
7040 /* We are given an specific port number; we verify
7041 * that it is not being used. If it is used, we will
7042 * exahust the search in the hash list corresponding
7043 * to the port number (snum) - we detect that with the
7044 * port iterator, pp being NULL.
7046 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
7047 spin_lock(&head->lock);
7048 sctp_for_each_hentry(pp, &head->chain) {
7049 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
7056 if (!hlist_empty(&pp->owner)) {
7057 /* We had a port hash table hit - there is an
7058 * available port (pp != NULL) and it is being
7059 * used by other socket (pp->owner not empty); that other
7060 * socket is going to be sk2.
7062 int reuse = sk->sk_reuse;
7065 pr_debug("%s: found a possible match\n", __func__);
7067 if (pp->fastreuse && sk->sk_reuse &&
7068 sk->sk_state != SCTP_SS_LISTENING)
7071 /* Run through the list of sockets bound to the port
7072 * (pp->port) [via the pointers bind_next and
7073 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
7074 * we get the endpoint they describe and run through
7075 * the endpoint's list of IP (v4 or v6) addresses,
7076 * comparing each of the addresses with the address of
7077 * the socket sk. If we find a match, then that means
7078 * that this port/socket (sk) combination are already
7081 sk_for_each_bound(sk2, &pp->owner) {
7082 struct sctp_endpoint *ep2;
7083 ep2 = sctp_sk(sk2)->ep;
7086 (reuse && sk2->sk_reuse &&
7087 sk2->sk_state != SCTP_SS_LISTENING))
7090 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
7091 sctp_sk(sk2), sctp_sk(sk))) {
7097 pr_debug("%s: found a match\n", __func__);
7100 /* If there was a hash table miss, create a new port. */
7102 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
7105 /* In either case (hit or miss), make sure fastreuse is 1 only
7106 * if sk->sk_reuse is too (that is, if the caller requested
7107 * SO_REUSEADDR on this socket -sk-).
7109 if (hlist_empty(&pp->owner)) {
7110 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
7114 } else if (pp->fastreuse &&
7115 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
7118 /* We are set, so fill up all the data in the hash table
7119 * entry, tie the socket list information with the rest of the
7120 * sockets FIXME: Blurry, NPI (ipg).
7123 if (!sctp_sk(sk)->bind_hash) {
7124 inet_sk(sk)->inet_num = snum;
7125 sk_add_bind_node(sk, &pp->owner);
7126 sctp_sk(sk)->bind_hash = pp;
7131 spin_unlock(&head->lock);
7138 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
7139 * port is requested.
7141 static int sctp_get_port(struct sock *sk, unsigned short snum)
7143 union sctp_addr addr;
7144 struct sctp_af *af = sctp_sk(sk)->pf->af;
7146 /* Set up a dummy address struct from the sk. */
7147 af->from_sk(&addr, sk);
7148 addr.v4.sin_port = htons(snum);
7150 /* Note: sk->sk_num gets filled in if ephemeral port request. */
7151 return !!sctp_get_port_local(sk, &addr);
7155 * Move a socket to LISTENING state.
7157 static int sctp_listen_start(struct sock *sk, int backlog)
7159 struct sctp_sock *sp = sctp_sk(sk);
7160 struct sctp_endpoint *ep = sp->ep;
7161 struct crypto_shash *tfm = NULL;
7164 /* Allocate HMAC for generating cookie. */
7165 if (!sp->hmac && sp->sctp_hmac_alg) {
7166 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
7167 tfm = crypto_alloc_shash(alg, 0, 0);
7169 net_info_ratelimited("failed to load transform for %s: %ld\n",
7170 sp->sctp_hmac_alg, PTR_ERR(tfm));
7173 sctp_sk(sk)->hmac = tfm;
7177 * If a bind() or sctp_bindx() is not called prior to a listen()
7178 * call that allows new associations to be accepted, the system
7179 * picks an ephemeral port and will choose an address set equivalent
7180 * to binding with a wildcard address.
7182 * This is not currently spelled out in the SCTP sockets
7183 * extensions draft, but follows the practice as seen in TCP
7187 sk->sk_state = SCTP_SS_LISTENING;
7188 if (!ep->base.bind_addr.port) {
7189 if (sctp_autobind(sk))
7192 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
7193 sk->sk_state = SCTP_SS_CLOSED;
7198 sk->sk_max_ack_backlog = backlog;
7199 sctp_hash_endpoint(ep);
7204 * 4.1.3 / 5.1.3 listen()
7206 * By default, new associations are not accepted for UDP style sockets.
7207 * An application uses listen() to mark a socket as being able to
7208 * accept new associations.
7210 * On TCP style sockets, applications use listen() to ready the SCTP
7211 * endpoint for accepting inbound associations.
7213 * On both types of endpoints a backlog of '0' disables listening.
7215 * Move a socket to LISTENING state.
7217 int sctp_inet_listen(struct socket *sock, int backlog)
7219 struct sock *sk = sock->sk;
7220 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
7223 if (unlikely(backlog < 0))
7228 /* Peeled-off sockets are not allowed to listen(). */
7229 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
7232 if (sock->state != SS_UNCONNECTED)
7235 if (!sctp_sstate(sk, LISTENING) && !sctp_sstate(sk, CLOSED))
7238 /* If backlog is zero, disable listening. */
7240 if (sctp_sstate(sk, CLOSED))
7244 sctp_unhash_endpoint(ep);
7245 sk->sk_state = SCTP_SS_CLOSED;
7247 sctp_sk(sk)->bind_hash->fastreuse = 1;
7251 /* If we are already listening, just update the backlog */
7252 if (sctp_sstate(sk, LISTENING))
7253 sk->sk_max_ack_backlog = backlog;
7255 err = sctp_listen_start(sk, backlog);
7267 * This function is done by modeling the current datagram_poll() and the
7268 * tcp_poll(). Note that, based on these implementations, we don't
7269 * lock the socket in this function, even though it seems that,
7270 * ideally, locking or some other mechanisms can be used to ensure
7271 * the integrity of the counters (sndbuf and wmem_alloc) used
7272 * in this place. We assume that we don't need locks either until proven
7275 * Another thing to note is that we include the Async I/O support
7276 * here, again, by modeling the current TCP/UDP code. We don't have
7277 * a good way to test with it yet.
7279 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
7281 struct sock *sk = sock->sk;
7282 struct sctp_sock *sp = sctp_sk(sk);
7285 poll_wait(file, sk_sleep(sk), wait);
7287 sock_rps_record_flow(sk);
7289 /* A TCP-style listening socket becomes readable when the accept queue
7292 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
7293 return (!list_empty(&sp->ep->asocs)) ?
7294 (POLLIN | POLLRDNORM) : 0;
7298 /* Is there any exceptional events? */
7299 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
7301 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
7302 if (sk->sk_shutdown & RCV_SHUTDOWN)
7303 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
7304 if (sk->sk_shutdown == SHUTDOWN_MASK)
7307 /* Is it readable? Reconsider this code with TCP-style support. */
7308 if (!skb_queue_empty(&sk->sk_receive_queue))
7309 mask |= POLLIN | POLLRDNORM;
7311 /* The association is either gone or not ready. */
7312 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
7315 /* Is it writable? */
7316 if (sctp_writeable(sk)) {
7317 mask |= POLLOUT | POLLWRNORM;
7319 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
7321 * Since the socket is not locked, the buffer
7322 * might be made available after the writeable check and
7323 * before the bit is set. This could cause a lost I/O
7324 * signal. tcp_poll() has a race breaker for this race
7325 * condition. Based on their implementation, we put
7326 * in the following code to cover it as well.
7328 if (sctp_writeable(sk))
7329 mask |= POLLOUT | POLLWRNORM;
7334 /********************************************************************
7335 * 2nd Level Abstractions
7336 ********************************************************************/
7338 static struct sctp_bind_bucket *sctp_bucket_create(
7339 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
7341 struct sctp_bind_bucket *pp;
7343 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
7345 SCTP_DBG_OBJCNT_INC(bind_bucket);
7348 INIT_HLIST_HEAD(&pp->owner);
7350 hlist_add_head(&pp->node, &head->chain);
7355 /* Caller must hold hashbucket lock for this tb with local BH disabled */
7356 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
7358 if (pp && hlist_empty(&pp->owner)) {
7359 __hlist_del(&pp->node);
7360 kmem_cache_free(sctp_bucket_cachep, pp);
7361 SCTP_DBG_OBJCNT_DEC(bind_bucket);
7365 /* Release this socket's reference to a local port. */
7366 static inline void __sctp_put_port(struct sock *sk)
7368 struct sctp_bind_hashbucket *head =
7369 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
7370 inet_sk(sk)->inet_num)];
7371 struct sctp_bind_bucket *pp;
7373 spin_lock(&head->lock);
7374 pp = sctp_sk(sk)->bind_hash;
7375 __sk_del_bind_node(sk);
7376 sctp_sk(sk)->bind_hash = NULL;
7377 inet_sk(sk)->inet_num = 0;
7378 sctp_bucket_destroy(pp);
7379 spin_unlock(&head->lock);
7382 void sctp_put_port(struct sock *sk)
7385 __sctp_put_port(sk);
7390 * The system picks an ephemeral port and choose an address set equivalent
7391 * to binding with a wildcard address.
7392 * One of those addresses will be the primary address for the association.
7393 * This automatically enables the multihoming capability of SCTP.
7395 static int sctp_autobind(struct sock *sk)
7397 union sctp_addr autoaddr;
7401 /* Initialize a local sockaddr structure to INADDR_ANY. */
7402 af = sctp_sk(sk)->pf->af;
7404 port = htons(inet_sk(sk)->inet_num);
7405 af->inaddr_any(&autoaddr, port);
7407 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
7410 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
7413 * 4.2 The cmsghdr Structure *
7415 * When ancillary data is sent or received, any number of ancillary data
7416 * objects can be specified by the msg_control and msg_controllen members of
7417 * the msghdr structure, because each object is preceded by
7418 * a cmsghdr structure defining the object's length (the cmsg_len member).
7419 * Historically Berkeley-derived implementations have passed only one object
7420 * at a time, but this API allows multiple objects to be
7421 * passed in a single call to sendmsg() or recvmsg(). The following example
7422 * shows two ancillary data objects in a control buffer.
7424 * |<--------------------------- msg_controllen -------------------------->|
7427 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
7429 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
7432 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
7434 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
7437 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7438 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
7440 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
7442 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7449 static int sctp_msghdr_parse(const struct msghdr *msg, struct sctp_cmsgs *cmsgs)
7451 struct msghdr *my_msg = (struct msghdr *)msg;
7452 struct cmsghdr *cmsg;
7454 for_each_cmsghdr(cmsg, my_msg) {
7455 if (!CMSG_OK(my_msg, cmsg))
7458 /* Should we parse this header or ignore? */
7459 if (cmsg->cmsg_level != IPPROTO_SCTP)
7462 /* Strictly check lengths following example in SCM code. */
7463 switch (cmsg->cmsg_type) {
7465 /* SCTP Socket API Extension
7466 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
7468 * This cmsghdr structure provides information for
7469 * initializing new SCTP associations with sendmsg().
7470 * The SCTP_INITMSG socket option uses this same data
7471 * structure. This structure is not used for
7474 * cmsg_level cmsg_type cmsg_data[]
7475 * ------------ ------------ ----------------------
7476 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
7478 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
7481 cmsgs->init = CMSG_DATA(cmsg);
7485 /* SCTP Socket API Extension
7486 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
7488 * This cmsghdr structure specifies SCTP options for
7489 * sendmsg() and describes SCTP header information
7490 * about a received message through recvmsg().
7492 * cmsg_level cmsg_type cmsg_data[]
7493 * ------------ ------------ ----------------------
7494 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
7496 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
7499 cmsgs->srinfo = CMSG_DATA(cmsg);
7501 if (cmsgs->srinfo->sinfo_flags &
7502 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7503 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7504 SCTP_ABORT | SCTP_EOF))
7509 /* SCTP Socket API Extension
7510 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
7512 * This cmsghdr structure specifies SCTP options for
7513 * sendmsg(). This structure and SCTP_RCVINFO replaces
7514 * SCTP_SNDRCV which has been deprecated.
7516 * cmsg_level cmsg_type cmsg_data[]
7517 * ------------ ------------ ---------------------
7518 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
7520 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
7523 cmsgs->sinfo = CMSG_DATA(cmsg);
7525 if (cmsgs->sinfo->snd_flags &
7526 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7527 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7528 SCTP_ABORT | SCTP_EOF))
7540 * Wait for a packet..
7541 * Note: This function is the same function as in core/datagram.c
7542 * with a few modifications to make lksctp work.
7544 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
7549 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7551 /* Socket errors? */
7552 error = sock_error(sk);
7556 if (!skb_queue_empty(&sk->sk_receive_queue))
7559 /* Socket shut down? */
7560 if (sk->sk_shutdown & RCV_SHUTDOWN)
7563 /* Sequenced packets can come disconnected. If so we report the
7568 /* Is there a good reason to think that we may receive some data? */
7569 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
7572 /* Handle signals. */
7573 if (signal_pending(current))
7576 /* Let another process have a go. Since we are going to sleep
7577 * anyway. Note: This may cause odd behaviors if the message
7578 * does not fit in the user's buffer, but this seems to be the
7579 * only way to honor MSG_DONTWAIT realistically.
7582 *timeo_p = schedule_timeout(*timeo_p);
7586 finish_wait(sk_sleep(sk), &wait);
7590 error = sock_intr_errno(*timeo_p);
7593 finish_wait(sk_sleep(sk), &wait);
7598 /* Receive a datagram.
7599 * Note: This is pretty much the same routine as in core/datagram.c
7600 * with a few changes to make lksctp work.
7602 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
7603 int noblock, int *err)
7606 struct sk_buff *skb;
7609 timeo = sock_rcvtimeo(sk, noblock);
7611 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
7612 MAX_SCHEDULE_TIMEOUT);
7615 /* Again only user level code calls this function,
7616 * so nothing interrupt level
7617 * will suddenly eat the receive_queue.
7619 * Look at current nfs client by the way...
7620 * However, this function was correct in any case. 8)
7622 if (flags & MSG_PEEK) {
7623 skb = skb_peek(&sk->sk_receive_queue);
7625 refcount_inc(&skb->users);
7627 skb = __skb_dequeue(&sk->sk_receive_queue);
7633 /* Caller is allowed not to check sk->sk_err before calling. */
7634 error = sock_error(sk);
7638 if (sk->sk_shutdown & RCV_SHUTDOWN)
7641 if (sk_can_busy_loop(sk)) {
7642 sk_busy_loop(sk, noblock);
7644 if (!skb_queue_empty(&sk->sk_receive_queue))
7648 /* User doesn't want to wait. */
7652 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
7661 /* If sndbuf has changed, wake up per association sndbuf waiters. */
7662 static void __sctp_write_space(struct sctp_association *asoc)
7664 struct sock *sk = asoc->base.sk;
7666 if (sctp_wspace(asoc) <= 0)
7669 if (waitqueue_active(&asoc->wait))
7670 wake_up_interruptible(&asoc->wait);
7672 if (sctp_writeable(sk)) {
7673 struct socket_wq *wq;
7676 wq = rcu_dereference(sk->sk_wq);
7678 if (waitqueue_active(&wq->wait))
7679 wake_up_interruptible(&wq->wait);
7681 /* Note that we try to include the Async I/O support
7682 * here by modeling from the current TCP/UDP code.
7683 * We have not tested with it yet.
7685 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
7686 sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT);
7692 static void sctp_wake_up_waiters(struct sock *sk,
7693 struct sctp_association *asoc)
7695 struct sctp_association *tmp = asoc;
7697 /* We do accounting for the sndbuf space per association,
7698 * so we only need to wake our own association.
7700 if (asoc->ep->sndbuf_policy)
7701 return __sctp_write_space(asoc);
7703 /* If association goes down and is just flushing its
7704 * outq, then just normally notify others.
7706 if (asoc->base.dead)
7707 return sctp_write_space(sk);
7709 /* Accounting for the sndbuf space is per socket, so we
7710 * need to wake up others, try to be fair and in case of
7711 * other associations, let them have a go first instead
7712 * of just doing a sctp_write_space() call.
7714 * Note that we reach sctp_wake_up_waiters() only when
7715 * associations free up queued chunks, thus we are under
7716 * lock and the list of associations on a socket is
7717 * guaranteed not to change.
7719 for (tmp = list_next_entry(tmp, asocs); 1;
7720 tmp = list_next_entry(tmp, asocs)) {
7721 /* Manually skip the head element. */
7722 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
7724 /* Wake up association. */
7725 __sctp_write_space(tmp);
7726 /* We've reached the end. */
7732 /* Do accounting for the sndbuf space.
7733 * Decrement the used sndbuf space of the corresponding association by the
7734 * data size which was just transmitted(freed).
7736 static void sctp_wfree(struct sk_buff *skb)
7738 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
7739 struct sctp_association *asoc = chunk->asoc;
7740 struct sock *sk = asoc->base.sk;
7742 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
7743 sizeof(struct sk_buff) +
7744 sizeof(struct sctp_chunk);
7746 WARN_ON(refcount_sub_and_test(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc));
7749 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
7751 sk->sk_wmem_queued -= skb->truesize;
7752 sk_mem_uncharge(sk, skb->truesize);
7755 sctp_wake_up_waiters(sk, asoc);
7757 sctp_association_put(asoc);
7760 /* Do accounting for the receive space on the socket.
7761 * Accounting for the association is done in ulpevent.c
7762 * We set this as a destructor for the cloned data skbs so that
7763 * accounting is done at the correct time.
7765 void sctp_sock_rfree(struct sk_buff *skb)
7767 struct sock *sk = skb->sk;
7768 struct sctp_ulpevent *event = sctp_skb2event(skb);
7770 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
7773 * Mimic the behavior of sock_rfree
7775 sk_mem_uncharge(sk, event->rmem_len);
7779 /* Helper function to wait for space in the sndbuf. */
7780 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
7783 struct sock *sk = asoc->base.sk;
7785 long current_timeo = *timeo_p;
7788 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
7791 /* Increment the association's refcnt. */
7792 sctp_association_hold(asoc);
7794 /* Wait on the association specific sndbuf space. */
7796 prepare_to_wait_exclusive(&asoc->wait, &wait,
7797 TASK_INTERRUPTIBLE);
7800 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7803 if (signal_pending(current))
7804 goto do_interrupted;
7805 if (msg_len <= sctp_wspace(asoc))
7808 /* Let another process have a go. Since we are going
7812 current_timeo = schedule_timeout(current_timeo);
7815 *timeo_p = current_timeo;
7819 finish_wait(&asoc->wait, &wait);
7821 /* Release the association's refcnt. */
7822 sctp_association_put(asoc);
7831 err = sock_intr_errno(*timeo_p);
7839 void sctp_data_ready(struct sock *sk)
7841 struct socket_wq *wq;
7844 wq = rcu_dereference(sk->sk_wq);
7845 if (skwq_has_sleeper(wq))
7846 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
7847 POLLRDNORM | POLLRDBAND);
7848 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
7852 /* If socket sndbuf has changed, wake up all per association waiters. */
7853 void sctp_write_space(struct sock *sk)
7855 struct sctp_association *asoc;
7857 /* Wake up the tasks in each wait queue. */
7858 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
7859 __sctp_write_space(asoc);
7863 /* Is there any sndbuf space available on the socket?
7865 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
7866 * associations on the same socket. For a UDP-style socket with
7867 * multiple associations, it is possible for it to be "unwriteable"
7868 * prematurely. I assume that this is acceptable because
7869 * a premature "unwriteable" is better than an accidental "writeable" which
7870 * would cause an unwanted block under certain circumstances. For the 1-1
7871 * UDP-style sockets or TCP-style sockets, this code should work.
7874 static int sctp_writeable(struct sock *sk)
7878 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
7884 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
7885 * returns immediately with EINPROGRESS.
7887 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
7889 struct sock *sk = asoc->base.sk;
7891 long current_timeo = *timeo_p;
7894 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
7896 /* Increment the association's refcnt. */
7897 sctp_association_hold(asoc);
7900 prepare_to_wait_exclusive(&asoc->wait, &wait,
7901 TASK_INTERRUPTIBLE);
7904 if (sk->sk_shutdown & RCV_SHUTDOWN)
7906 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7909 if (signal_pending(current))
7910 goto do_interrupted;
7912 if (sctp_state(asoc, ESTABLISHED))
7915 /* Let another process have a go. Since we are going
7919 current_timeo = schedule_timeout(current_timeo);
7922 *timeo_p = current_timeo;
7926 finish_wait(&asoc->wait, &wait);
7928 /* Release the association's refcnt. */
7929 sctp_association_put(asoc);
7934 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
7937 err = -ECONNREFUSED;
7941 err = sock_intr_errno(*timeo_p);
7949 static int sctp_wait_for_accept(struct sock *sk, long timeo)
7951 struct sctp_endpoint *ep;
7955 ep = sctp_sk(sk)->ep;
7959 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
7960 TASK_INTERRUPTIBLE);
7962 if (list_empty(&ep->asocs)) {
7964 timeo = schedule_timeout(timeo);
7969 if (!sctp_sstate(sk, LISTENING))
7973 if (!list_empty(&ep->asocs))
7976 err = sock_intr_errno(timeo);
7977 if (signal_pending(current))
7985 finish_wait(sk_sleep(sk), &wait);
7990 static void sctp_wait_for_close(struct sock *sk, long timeout)
7995 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7996 if (list_empty(&sctp_sk(sk)->ep->asocs))
7999 timeout = schedule_timeout(timeout);
8001 } while (!signal_pending(current) && timeout);
8003 finish_wait(sk_sleep(sk), &wait);
8006 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
8008 struct sk_buff *frag;
8013 /* Don't forget the fragments. */
8014 skb_walk_frags(skb, frag)
8015 sctp_skb_set_owner_r_frag(frag, sk);
8018 sctp_skb_set_owner_r(skb, sk);
8021 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
8022 struct sctp_association *asoc)
8024 struct inet_sock *inet = inet_sk(sk);
8025 struct inet_sock *newinet;
8027 newsk->sk_type = sk->sk_type;
8028 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
8029 newsk->sk_flags = sk->sk_flags;
8030 newsk->sk_tsflags = sk->sk_tsflags;
8031 newsk->sk_no_check_tx = sk->sk_no_check_tx;
8032 newsk->sk_no_check_rx = sk->sk_no_check_rx;
8033 newsk->sk_reuse = sk->sk_reuse;
8035 newsk->sk_shutdown = sk->sk_shutdown;
8036 newsk->sk_destruct = sctp_destruct_sock;
8037 newsk->sk_family = sk->sk_family;
8038 newsk->sk_protocol = IPPROTO_SCTP;
8039 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
8040 newsk->sk_sndbuf = sk->sk_sndbuf;
8041 newsk->sk_rcvbuf = sk->sk_rcvbuf;
8042 newsk->sk_lingertime = sk->sk_lingertime;
8043 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
8044 newsk->sk_sndtimeo = sk->sk_sndtimeo;
8045 newsk->sk_rxhash = sk->sk_rxhash;
8047 newinet = inet_sk(newsk);
8049 /* Initialize sk's sport, dport, rcv_saddr and daddr for
8050 * getsockname() and getpeername()
8052 newinet->inet_sport = inet->inet_sport;
8053 newinet->inet_saddr = inet->inet_saddr;
8054 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
8055 newinet->inet_dport = htons(asoc->peer.port);
8056 newinet->pmtudisc = inet->pmtudisc;
8057 newinet->inet_id = asoc->next_tsn ^ jiffies;
8059 newinet->uc_ttl = inet->uc_ttl;
8060 newinet->mc_loop = 1;
8061 newinet->mc_ttl = 1;
8062 newinet->mc_index = 0;
8063 newinet->mc_list = NULL;
8065 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
8066 net_enable_timestamp();
8068 security_sk_clone(sk, newsk);
8071 static inline void sctp_copy_descendant(struct sock *sk_to,
8072 const struct sock *sk_from)
8074 int ancestor_size = sizeof(struct inet_sock) +
8075 sizeof(struct sctp_sock) -
8076 offsetof(struct sctp_sock, auto_asconf_list);
8078 if (sk_from->sk_family == PF_INET6)
8079 ancestor_size += sizeof(struct ipv6_pinfo);
8081 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
8084 /* Populate the fields of the newsk from the oldsk and migrate the assoc
8085 * and its messages to the newsk.
8087 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
8088 struct sctp_association *assoc,
8089 sctp_socket_type_t type)
8091 struct sctp_sock *oldsp = sctp_sk(oldsk);
8092 struct sctp_sock *newsp = sctp_sk(newsk);
8093 struct sctp_bind_bucket *pp; /* hash list port iterator */
8094 struct sctp_endpoint *newep = newsp->ep;
8095 struct sk_buff *skb, *tmp;
8096 struct sctp_ulpevent *event;
8097 struct sctp_bind_hashbucket *head;
8099 /* Migrate socket buffer sizes and all the socket level options to the
8102 newsk->sk_sndbuf = oldsk->sk_sndbuf;
8103 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
8104 /* Brute force copy old sctp opt. */
8105 sctp_copy_descendant(newsk, oldsk);
8107 /* Restore the ep value that was overwritten with the above structure
8113 /* Hook this new socket in to the bind_hash list. */
8114 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
8115 inet_sk(oldsk)->inet_num)];
8116 spin_lock_bh(&head->lock);
8117 pp = sctp_sk(oldsk)->bind_hash;
8118 sk_add_bind_node(newsk, &pp->owner);
8119 sctp_sk(newsk)->bind_hash = pp;
8120 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
8121 spin_unlock_bh(&head->lock);
8123 /* Copy the bind_addr list from the original endpoint to the new
8124 * endpoint so that we can handle restarts properly
8126 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
8127 &oldsp->ep->base.bind_addr, GFP_KERNEL);
8129 /* Move any messages in the old socket's receive queue that are for the
8130 * peeled off association to the new socket's receive queue.
8132 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
8133 event = sctp_skb2event(skb);
8134 if (event->asoc == assoc) {
8135 __skb_unlink(skb, &oldsk->sk_receive_queue);
8136 __skb_queue_tail(&newsk->sk_receive_queue, skb);
8137 sctp_skb_set_owner_r_frag(skb, newsk);
8141 /* Clean up any messages pending delivery due to partial
8142 * delivery. Three cases:
8143 * 1) No partial deliver; no work.
8144 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
8145 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
8147 skb_queue_head_init(&newsp->pd_lobby);
8148 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
8150 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
8151 struct sk_buff_head *queue;
8153 /* Decide which queue to move pd_lobby skbs to. */
8154 if (assoc->ulpq.pd_mode) {
8155 queue = &newsp->pd_lobby;
8157 queue = &newsk->sk_receive_queue;
8159 /* Walk through the pd_lobby, looking for skbs that
8160 * need moved to the new socket.
8162 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
8163 event = sctp_skb2event(skb);
8164 if (event->asoc == assoc) {
8165 __skb_unlink(skb, &oldsp->pd_lobby);
8166 __skb_queue_tail(queue, skb);
8167 sctp_skb_set_owner_r_frag(skb, newsk);
8171 /* Clear up any skbs waiting for the partial
8172 * delivery to finish.
8174 if (assoc->ulpq.pd_mode)
8175 sctp_clear_pd(oldsk, NULL);
8179 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
8180 sctp_skb_set_owner_r_frag(skb, newsk);
8182 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
8183 sctp_skb_set_owner_r_frag(skb, newsk);
8185 /* Set the type of socket to indicate that it is peeled off from the
8186 * original UDP-style socket or created with the accept() call on a
8187 * TCP-style socket..
8191 /* Mark the new socket "in-use" by the user so that any packets
8192 * that may arrive on the association after we've moved it are
8193 * queued to the backlog. This prevents a potential race between
8194 * backlog processing on the old socket and new-packet processing
8195 * on the new socket.
8197 * The caller has just allocated newsk so we can guarantee that other
8198 * paths won't try to lock it and then oldsk.
8200 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
8201 sctp_assoc_migrate(assoc, newsk);
8203 /* If the association on the newsk is already closed before accept()
8204 * is called, set RCV_SHUTDOWN flag.
8206 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP)) {
8207 newsk->sk_state = SCTP_SS_CLOSED;
8208 newsk->sk_shutdown |= RCV_SHUTDOWN;
8210 newsk->sk_state = SCTP_SS_ESTABLISHED;
8213 release_sock(newsk);
8217 /* This proto struct describes the ULP interface for SCTP. */
8218 struct proto sctp_prot = {
8220 .owner = THIS_MODULE,
8221 .close = sctp_close,
8222 .connect = sctp_connect,
8223 .disconnect = sctp_disconnect,
8224 .accept = sctp_accept,
8225 .ioctl = sctp_ioctl,
8226 .init = sctp_init_sock,
8227 .destroy = sctp_destroy_sock,
8228 .shutdown = sctp_shutdown,
8229 .setsockopt = sctp_setsockopt,
8230 .getsockopt = sctp_getsockopt,
8231 .sendmsg = sctp_sendmsg,
8232 .recvmsg = sctp_recvmsg,
8234 .backlog_rcv = sctp_backlog_rcv,
8236 .unhash = sctp_unhash,
8237 .get_port = sctp_get_port,
8238 .obj_size = sizeof(struct sctp_sock),
8239 .sysctl_mem = sysctl_sctp_mem,
8240 .sysctl_rmem = sysctl_sctp_rmem,
8241 .sysctl_wmem = sysctl_sctp_wmem,
8242 .memory_pressure = &sctp_memory_pressure,
8243 .enter_memory_pressure = sctp_enter_memory_pressure,
8244 .memory_allocated = &sctp_memory_allocated,
8245 .sockets_allocated = &sctp_sockets_allocated,
8248 #if IS_ENABLED(CONFIG_IPV6)
8250 #include <net/transp_v6.h>
8251 static void sctp_v6_destroy_sock(struct sock *sk)
8253 sctp_destroy_sock(sk);
8254 inet6_destroy_sock(sk);
8257 struct proto sctpv6_prot = {
8259 .owner = THIS_MODULE,
8260 .close = sctp_close,
8261 .connect = sctp_connect,
8262 .disconnect = sctp_disconnect,
8263 .accept = sctp_accept,
8264 .ioctl = sctp_ioctl,
8265 .init = sctp_init_sock,
8266 .destroy = sctp_v6_destroy_sock,
8267 .shutdown = sctp_shutdown,
8268 .setsockopt = sctp_setsockopt,
8269 .getsockopt = sctp_getsockopt,
8270 .sendmsg = sctp_sendmsg,
8271 .recvmsg = sctp_recvmsg,
8273 .backlog_rcv = sctp_backlog_rcv,
8275 .unhash = sctp_unhash,
8276 .get_port = sctp_get_port,
8277 .obj_size = sizeof(struct sctp6_sock),
8278 .sysctl_mem = sysctl_sctp_mem,
8279 .sysctl_rmem = sysctl_sctp_rmem,
8280 .sysctl_wmem = sysctl_sctp_wmem,
8281 .memory_pressure = &sctp_memory_pressure,
8282 .enter_memory_pressure = sctp_enter_memory_pressure,
8283 .memory_allocated = &sctp_memory_allocated,
8284 .sockets_allocated = &sctp_sockets_allocated,
8286 #endif /* IS_ENABLED(CONFIG_IPV6) */