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;
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 *, sctp_cmsgs_t *);
1598 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
1600 struct net *net = sock_net(sk);
1601 struct sctp_sock *sp;
1602 struct sctp_endpoint *ep;
1603 struct sctp_association *new_asoc = NULL, *asoc = NULL;
1604 struct sctp_transport *transport, *chunk_tp;
1605 struct sctp_chunk *chunk;
1607 struct sockaddr *msg_name = NULL;
1608 struct sctp_sndrcvinfo default_sinfo;
1609 struct sctp_sndrcvinfo *sinfo;
1610 struct sctp_initmsg *sinit;
1611 sctp_assoc_t associd = 0;
1612 sctp_cmsgs_t cmsgs = { NULL };
1614 bool fill_sinfo_ttl = false, wait_connect = false;
1615 struct sctp_datamsg *datamsg;
1616 int msg_flags = msg->msg_flags;
1617 __u16 sinfo_flags = 0;
1625 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1628 /* We cannot send a message over a TCP-style listening socket. */
1629 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1634 /* Parse out the SCTP CMSGs. */
1635 err = sctp_msghdr_parse(msg, &cmsgs);
1637 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1641 /* Fetch the destination address for this packet. This
1642 * address only selects the association--it is not necessarily
1643 * the address we will send to.
1644 * For a peeled-off socket, msg_name is ignored.
1646 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1647 int msg_namelen = msg->msg_namelen;
1649 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1654 if (msg_namelen > sizeof(to))
1655 msg_namelen = sizeof(to);
1656 memcpy(&to, msg->msg_name, msg_namelen);
1657 msg_name = msg->msg_name;
1661 if (cmsgs.sinfo != NULL) {
1662 memset(&default_sinfo, 0, sizeof(default_sinfo));
1663 default_sinfo.sinfo_stream = cmsgs.sinfo->snd_sid;
1664 default_sinfo.sinfo_flags = cmsgs.sinfo->snd_flags;
1665 default_sinfo.sinfo_ppid = cmsgs.sinfo->snd_ppid;
1666 default_sinfo.sinfo_context = cmsgs.sinfo->snd_context;
1667 default_sinfo.sinfo_assoc_id = cmsgs.sinfo->snd_assoc_id;
1669 sinfo = &default_sinfo;
1670 fill_sinfo_ttl = true;
1672 sinfo = cmsgs.srinfo;
1674 /* Did the user specify SNDINFO/SNDRCVINFO? */
1676 sinfo_flags = sinfo->sinfo_flags;
1677 associd = sinfo->sinfo_assoc_id;
1680 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1681 msg_len, sinfo_flags);
1683 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1684 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1689 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1690 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1691 * If SCTP_ABORT is set, the message length could be non zero with
1692 * the msg_iov set to the user abort reason.
1694 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1695 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1700 /* If SCTP_ADDR_OVER is set, there must be an address
1701 * specified in msg_name.
1703 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1710 pr_debug("%s: about to look up association\n", __func__);
1714 /* If a msg_name has been specified, assume this is to be used. */
1716 /* Look for a matching association on the endpoint. */
1717 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1719 /* If we could not find a matching association on the
1720 * endpoint, make sure that it is not a TCP-style
1721 * socket that already has an association or there is
1722 * no peeled-off association on another socket.
1725 ((sctp_style(sk, TCP) &&
1726 (sctp_sstate(sk, ESTABLISHED) ||
1727 sctp_sstate(sk, CLOSING))) ||
1728 sctp_endpoint_is_peeled_off(ep, &to))) {
1729 err = -EADDRNOTAVAIL;
1733 asoc = sctp_id2assoc(sk, associd);
1741 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1743 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1744 * socket that has an association in CLOSED state. This can
1745 * happen when an accepted socket has an association that is
1748 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1753 if (sinfo_flags & SCTP_EOF) {
1754 pr_debug("%s: shutting down association:%p\n",
1757 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1761 if (sinfo_flags & SCTP_ABORT) {
1763 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1769 pr_debug("%s: aborting association:%p\n",
1772 sctp_primitive_ABORT(net, asoc, chunk);
1778 /* Do we need to create the association? */
1780 pr_debug("%s: there is no association yet\n", __func__);
1782 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1787 /* Check for invalid stream against the stream counts,
1788 * either the default or the user specified stream counts.
1791 if (!sinit || !sinit->sinit_num_ostreams) {
1792 /* Check against the defaults. */
1793 if (sinfo->sinfo_stream >=
1794 sp->initmsg.sinit_num_ostreams) {
1799 /* Check against the requested. */
1800 if (sinfo->sinfo_stream >=
1801 sinit->sinit_num_ostreams) {
1809 * API 3.1.2 bind() - UDP Style Syntax
1810 * If a bind() or sctp_bindx() is not called prior to a
1811 * sendmsg() call that initiates a new association, the
1812 * system picks an ephemeral port and will choose an address
1813 * set equivalent to binding with a wildcard address.
1815 if (!ep->base.bind_addr.port) {
1816 if (sctp_autobind(sk)) {
1822 * If an unprivileged user inherits a one-to-many
1823 * style socket with open associations on a privileged
1824 * port, it MAY be permitted to accept new associations,
1825 * but it SHOULD NOT be permitted to open new
1828 if (ep->base.bind_addr.port < inet_prot_sock(net) &&
1829 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1835 scope = sctp_scope(&to);
1836 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1842 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1848 /* If the SCTP_INIT ancillary data is specified, set all
1849 * the association init values accordingly.
1852 if (sinit->sinit_num_ostreams) {
1853 asoc->c.sinit_num_ostreams =
1854 sinit->sinit_num_ostreams;
1856 if (sinit->sinit_max_instreams) {
1857 asoc->c.sinit_max_instreams =
1858 sinit->sinit_max_instreams;
1860 if (sinit->sinit_max_attempts) {
1861 asoc->max_init_attempts
1862 = sinit->sinit_max_attempts;
1864 if (sinit->sinit_max_init_timeo) {
1865 asoc->max_init_timeo =
1866 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1870 /* Prime the peer's transport structures. */
1871 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1878 /* ASSERT: we have a valid association at this point. */
1879 pr_debug("%s: we have a valid association\n", __func__);
1882 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1883 * one with some defaults.
1885 memset(&default_sinfo, 0, sizeof(default_sinfo));
1886 default_sinfo.sinfo_stream = asoc->default_stream;
1887 default_sinfo.sinfo_flags = asoc->default_flags;
1888 default_sinfo.sinfo_ppid = asoc->default_ppid;
1889 default_sinfo.sinfo_context = asoc->default_context;
1890 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1891 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1893 sinfo = &default_sinfo;
1894 } else if (fill_sinfo_ttl) {
1895 /* In case SNDINFO was specified, we still need to fill
1896 * it with a default ttl from the assoc here.
1898 sinfo->sinfo_timetolive = asoc->default_timetolive;
1901 /* API 7.1.7, the sndbuf size per association bounds the
1902 * maximum size of data that can be sent in a single send call.
1904 if (msg_len > sk->sk_sndbuf) {
1909 if (asoc->pmtu_pending)
1910 sctp_assoc_pending_pmtu(asoc);
1912 /* If fragmentation is disabled and the message length exceeds the
1913 * association fragmentation point, return EMSGSIZE. The I-D
1914 * does not specify what this error is, but this looks like
1917 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1922 /* Check for invalid stream. */
1923 if (sinfo->sinfo_stream >= asoc->stream.outcnt) {
1928 if (sctp_wspace(asoc) < msg_len)
1929 sctp_prsctp_prune(asoc, sinfo, msg_len - sctp_wspace(asoc));
1931 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1932 if (!sctp_wspace(asoc)) {
1933 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1938 /* If an address is passed with the sendto/sendmsg call, it is used
1939 * to override the primary destination address in the TCP model, or
1940 * when SCTP_ADDR_OVER flag is set in the UDP model.
1942 if ((sctp_style(sk, TCP) && msg_name) ||
1943 (sinfo_flags & SCTP_ADDR_OVER)) {
1944 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1952 /* Auto-connect, if we aren't connected already. */
1953 if (sctp_state(asoc, CLOSED)) {
1954 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1958 wait_connect = true;
1959 pr_debug("%s: we associated primitively\n", __func__);
1962 /* Break the message into multiple chunks of maximum size. */
1963 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
1964 if (IS_ERR(datamsg)) {
1965 err = PTR_ERR(datamsg);
1968 asoc->force_delay = !!(msg->msg_flags & MSG_MORE);
1970 /* Now send the (possibly) fragmented message. */
1971 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1972 sctp_chunk_hold(chunk);
1974 /* Do accounting for the write space. */
1975 sctp_set_owner_w(chunk);
1977 chunk->transport = chunk_tp;
1980 /* Send it to the lower layers. Note: all chunks
1981 * must either fail or succeed. The lower layer
1982 * works that way today. Keep it that way or this
1985 err = sctp_primitive_SEND(net, asoc, datamsg);
1986 /* Did the lower layer accept the chunk? */
1988 sctp_datamsg_free(datamsg);
1992 pr_debug("%s: we sent primitively\n", __func__);
1994 sctp_datamsg_put(datamsg);
1997 if (unlikely(wait_connect)) {
1998 timeo = sock_sndtimeo(sk, msg_flags & MSG_DONTWAIT);
1999 sctp_wait_for_connect(asoc, &timeo);
2002 /* If we are already past ASSOCIATE, the lower
2003 * layers are responsible for association cleanup.
2009 sctp_association_free(asoc);
2014 return sctp_error(sk, msg_flags, err);
2021 err = sock_error(sk);
2031 /* This is an extended version of skb_pull() that removes the data from the
2032 * start of a skb even when data is spread across the list of skb's in the
2033 * frag_list. len specifies the total amount of data that needs to be removed.
2034 * when 'len' bytes could be removed from the skb, it returns 0.
2035 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2036 * could not be removed.
2038 static int sctp_skb_pull(struct sk_buff *skb, int len)
2040 struct sk_buff *list;
2041 int skb_len = skb_headlen(skb);
2044 if (len <= skb_len) {
2045 __skb_pull(skb, len);
2049 __skb_pull(skb, skb_len);
2051 skb_walk_frags(skb, list) {
2052 rlen = sctp_skb_pull(list, len);
2053 skb->len -= (len-rlen);
2054 skb->data_len -= (len-rlen);
2065 /* API 3.1.3 recvmsg() - UDP Style Syntax
2067 * ssize_t recvmsg(int socket, struct msghdr *message,
2070 * socket - the socket descriptor of the endpoint.
2071 * message - pointer to the msghdr structure which contains a single
2072 * user message and possibly some ancillary data.
2074 * See Section 5 for complete description of the data
2077 * flags - flags sent or received with the user message, see Section
2078 * 5 for complete description of the flags.
2080 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2081 int noblock, int flags, int *addr_len)
2083 struct sctp_ulpevent *event = NULL;
2084 struct sctp_sock *sp = sctp_sk(sk);
2085 struct sk_buff *skb, *head_skb;
2090 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2091 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2096 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED) &&
2097 !sctp_sstate(sk, CLOSING) && !sctp_sstate(sk, CLOSED)) {
2102 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2106 /* Get the total length of the skb including any skb's in the
2115 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2117 event = sctp_skb2event(skb);
2122 if (event->chunk && event->chunk->head_skb)
2123 head_skb = event->chunk->head_skb;
2126 sock_recv_ts_and_drops(msg, sk, head_skb);
2127 if (sctp_ulpevent_is_notification(event)) {
2128 msg->msg_flags |= MSG_NOTIFICATION;
2129 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2131 sp->pf->skb_msgname(head_skb, msg->msg_name, addr_len);
2134 /* Check if we allow SCTP_NXTINFO. */
2135 if (sp->recvnxtinfo)
2136 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2137 /* Check if we allow SCTP_RCVINFO. */
2138 if (sp->recvrcvinfo)
2139 sctp_ulpevent_read_rcvinfo(event, msg);
2140 /* Check if we allow SCTP_SNDRCVINFO. */
2141 if (sp->subscribe.sctp_data_io_event)
2142 sctp_ulpevent_read_sndrcvinfo(event, msg);
2146 /* If skb's length exceeds the user's buffer, update the skb and
2147 * push it back to the receive_queue so that the next call to
2148 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2150 if (skb_len > copied) {
2151 msg->msg_flags &= ~MSG_EOR;
2152 if (flags & MSG_PEEK)
2154 sctp_skb_pull(skb, copied);
2155 skb_queue_head(&sk->sk_receive_queue, skb);
2157 /* When only partial message is copied to the user, increase
2158 * rwnd by that amount. If all the data in the skb is read,
2159 * rwnd is updated when the event is freed.
2161 if (!sctp_ulpevent_is_notification(event))
2162 sctp_assoc_rwnd_increase(event->asoc, copied);
2164 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2165 (event->msg_flags & MSG_EOR))
2166 msg->msg_flags |= MSG_EOR;
2168 msg->msg_flags &= ~MSG_EOR;
2171 if (flags & MSG_PEEK) {
2172 /* Release the skb reference acquired after peeking the skb in
2173 * sctp_skb_recv_datagram().
2177 /* Free the event which includes releasing the reference to
2178 * the owner of the skb, freeing the skb and updating the
2181 sctp_ulpevent_free(event);
2188 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2190 * This option is a on/off flag. If enabled no SCTP message
2191 * fragmentation will be performed. Instead if a message being sent
2192 * exceeds the current PMTU size, the message will NOT be sent and
2193 * instead a error will be indicated to the user.
2195 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2196 char __user *optval,
2197 unsigned int optlen)
2201 if (optlen < sizeof(int))
2204 if (get_user(val, (int __user *)optval))
2207 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2212 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2213 unsigned int optlen)
2215 struct sctp_association *asoc;
2216 struct sctp_ulpevent *event;
2218 if (optlen > sizeof(struct sctp_event_subscribe))
2220 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2223 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2224 * if there is no data to be sent or retransmit, the stack will
2225 * immediately send up this notification.
2227 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2228 &sctp_sk(sk)->subscribe)) {
2229 asoc = sctp_id2assoc(sk, 0);
2231 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2232 event = sctp_ulpevent_make_sender_dry_event(asoc,
2237 sctp_ulpq_tail_event(&asoc->ulpq, event);
2244 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2246 * This socket option is applicable to the UDP-style socket only. When
2247 * set it will cause associations that are idle for more than the
2248 * specified number of seconds to automatically close. An association
2249 * being idle is defined an association that has NOT sent or received
2250 * user data. The special value of '0' indicates that no automatic
2251 * close of any associations should be performed. The option expects an
2252 * integer defining the number of seconds of idle time before an
2253 * association is closed.
2255 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2256 unsigned int optlen)
2258 struct sctp_sock *sp = sctp_sk(sk);
2259 struct net *net = sock_net(sk);
2261 /* Applicable to UDP-style socket only */
2262 if (sctp_style(sk, TCP))
2264 if (optlen != sizeof(int))
2266 if (copy_from_user(&sp->autoclose, optval, optlen))
2269 if (sp->autoclose > net->sctp.max_autoclose)
2270 sp->autoclose = net->sctp.max_autoclose;
2275 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2277 * Applications can enable or disable heartbeats for any peer address of
2278 * an association, modify an address's heartbeat interval, force a
2279 * heartbeat to be sent immediately, and adjust the address's maximum
2280 * number of retransmissions sent before an address is considered
2281 * unreachable. The following structure is used to access and modify an
2282 * address's parameters:
2284 * struct sctp_paddrparams {
2285 * sctp_assoc_t spp_assoc_id;
2286 * struct sockaddr_storage spp_address;
2287 * uint32_t spp_hbinterval;
2288 * uint16_t spp_pathmaxrxt;
2289 * uint32_t spp_pathmtu;
2290 * uint32_t spp_sackdelay;
2291 * uint32_t spp_flags;
2294 * spp_assoc_id - (one-to-many style socket) This is filled in the
2295 * application, and identifies the association for
2297 * spp_address - This specifies which address is of interest.
2298 * spp_hbinterval - This contains the value of the heartbeat interval,
2299 * in milliseconds. If a value of zero
2300 * is present in this field then no changes are to
2301 * be made to this parameter.
2302 * spp_pathmaxrxt - This contains the maximum number of
2303 * retransmissions before this address shall be
2304 * considered unreachable. If a value of zero
2305 * is present in this field then no changes are to
2306 * be made to this parameter.
2307 * spp_pathmtu - When Path MTU discovery is disabled the value
2308 * specified here will be the "fixed" path mtu.
2309 * Note that if the spp_address field is empty
2310 * then all associations on this address will
2311 * have this fixed path mtu set upon them.
2313 * spp_sackdelay - When delayed sack is enabled, this value specifies
2314 * the number of milliseconds that sacks will be delayed
2315 * for. This value will apply to all addresses of an
2316 * association if the spp_address field is empty. Note
2317 * also, that if delayed sack is enabled and this
2318 * value is set to 0, no change is made to the last
2319 * recorded delayed sack timer value.
2321 * spp_flags - These flags are used to control various features
2322 * on an association. The flag field may contain
2323 * zero or more of the following options.
2325 * SPP_HB_ENABLE - Enable heartbeats on the
2326 * specified address. Note that if the address
2327 * field is empty all addresses for the association
2328 * have heartbeats enabled upon them.
2330 * SPP_HB_DISABLE - Disable heartbeats on the
2331 * speicifed address. Note that if the address
2332 * field is empty all addresses for the association
2333 * will have their heartbeats disabled. Note also
2334 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2335 * mutually exclusive, only one of these two should
2336 * be specified. Enabling both fields will have
2337 * undetermined results.
2339 * SPP_HB_DEMAND - Request a user initiated heartbeat
2340 * to be made immediately.
2342 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2343 * heartbeat delayis to be set to the value of 0
2346 * SPP_PMTUD_ENABLE - This field will enable PMTU
2347 * discovery upon the specified address. Note that
2348 * if the address feild is empty then all addresses
2349 * on the association are effected.
2351 * SPP_PMTUD_DISABLE - This field will disable PMTU
2352 * discovery upon the specified address. Note that
2353 * if the address feild is empty then all addresses
2354 * on the association are effected. Not also that
2355 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2356 * exclusive. Enabling both will have undetermined
2359 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2360 * on delayed sack. The time specified in spp_sackdelay
2361 * is used to specify the sack delay for this address. Note
2362 * that if spp_address is empty then all addresses will
2363 * enable delayed sack and take on the sack delay
2364 * value specified in spp_sackdelay.
2365 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2366 * off delayed sack. If the spp_address field is blank then
2367 * delayed sack is disabled for the entire association. Note
2368 * also that this field is mutually exclusive to
2369 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2372 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2373 struct sctp_transport *trans,
2374 struct sctp_association *asoc,
2375 struct sctp_sock *sp,
2378 int sackdelay_change)
2382 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2383 struct net *net = sock_net(trans->asoc->base.sk);
2385 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2390 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2391 * this field is ignored. Note also that a value of zero indicates
2392 * the current setting should be left unchanged.
2394 if (params->spp_flags & SPP_HB_ENABLE) {
2396 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2397 * set. This lets us use 0 value when this flag
2400 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2401 params->spp_hbinterval = 0;
2403 if (params->spp_hbinterval ||
2404 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2407 msecs_to_jiffies(params->spp_hbinterval);
2410 msecs_to_jiffies(params->spp_hbinterval);
2412 sp->hbinterval = params->spp_hbinterval;
2419 trans->param_flags =
2420 (trans->param_flags & ~SPP_HB) | hb_change;
2423 (asoc->param_flags & ~SPP_HB) | hb_change;
2426 (sp->param_flags & ~SPP_HB) | hb_change;
2430 /* When Path MTU discovery is disabled the value specified here will
2431 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2432 * include the flag SPP_PMTUD_DISABLE for this field to have any
2435 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2437 trans->pathmtu = params->spp_pathmtu;
2438 sctp_assoc_sync_pmtu(asoc);
2440 asoc->pathmtu = params->spp_pathmtu;
2442 sp->pathmtu = params->spp_pathmtu;
2448 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2449 (params->spp_flags & SPP_PMTUD_ENABLE);
2450 trans->param_flags =
2451 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2453 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2454 sctp_assoc_sync_pmtu(asoc);
2458 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2461 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2465 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2466 * value of this field is ignored. Note also that a value of zero
2467 * indicates the current setting should be left unchanged.
2469 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2472 msecs_to_jiffies(params->spp_sackdelay);
2475 msecs_to_jiffies(params->spp_sackdelay);
2477 sp->sackdelay = params->spp_sackdelay;
2481 if (sackdelay_change) {
2483 trans->param_flags =
2484 (trans->param_flags & ~SPP_SACKDELAY) |
2488 (asoc->param_flags & ~SPP_SACKDELAY) |
2492 (sp->param_flags & ~SPP_SACKDELAY) |
2497 /* Note that a value of zero indicates the current setting should be
2500 if (params->spp_pathmaxrxt) {
2502 trans->pathmaxrxt = params->spp_pathmaxrxt;
2504 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2506 sp->pathmaxrxt = params->spp_pathmaxrxt;
2513 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2514 char __user *optval,
2515 unsigned int optlen)
2517 struct sctp_paddrparams params;
2518 struct sctp_transport *trans = NULL;
2519 struct sctp_association *asoc = NULL;
2520 struct sctp_sock *sp = sctp_sk(sk);
2522 int hb_change, pmtud_change, sackdelay_change;
2524 if (optlen != sizeof(struct sctp_paddrparams))
2527 if (copy_from_user(¶ms, optval, optlen))
2530 /* Validate flags and value parameters. */
2531 hb_change = params.spp_flags & SPP_HB;
2532 pmtud_change = params.spp_flags & SPP_PMTUD;
2533 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2535 if (hb_change == SPP_HB ||
2536 pmtud_change == SPP_PMTUD ||
2537 sackdelay_change == SPP_SACKDELAY ||
2538 params.spp_sackdelay > 500 ||
2539 (params.spp_pathmtu &&
2540 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2543 /* If an address other than INADDR_ANY is specified, and
2544 * no transport is found, then the request is invalid.
2546 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
2547 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2548 params.spp_assoc_id);
2553 /* Get association, if assoc_id != 0 and the socket is a one
2554 * to many style socket, and an association was not found, then
2555 * the id was invalid.
2557 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2558 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2561 /* Heartbeat demand can only be sent on a transport or
2562 * association, but not a socket.
2564 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2567 /* Process parameters. */
2568 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2569 hb_change, pmtud_change,
2575 /* If changes are for association, also apply parameters to each
2578 if (!trans && asoc) {
2579 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2581 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2582 hb_change, pmtud_change,
2590 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2592 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2595 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2597 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2601 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2603 * This option will effect the way delayed acks are performed. This
2604 * option allows you to get or set the delayed ack time, in
2605 * milliseconds. It also allows changing the delayed ack frequency.
2606 * Changing the frequency to 1 disables the delayed sack algorithm. If
2607 * the assoc_id is 0, then this sets or gets the endpoints default
2608 * values. If the assoc_id field is non-zero, then the set or get
2609 * effects the specified association for the one to many model (the
2610 * assoc_id field is ignored by the one to one model). Note that if
2611 * sack_delay or sack_freq are 0 when setting this option, then the
2612 * current values will remain unchanged.
2614 * struct sctp_sack_info {
2615 * sctp_assoc_t sack_assoc_id;
2616 * uint32_t sack_delay;
2617 * uint32_t sack_freq;
2620 * sack_assoc_id - This parameter, indicates which association the user
2621 * is performing an action upon. Note that if this field's value is
2622 * zero then the endpoints default value is changed (effecting future
2623 * associations only).
2625 * sack_delay - This parameter contains the number of milliseconds that
2626 * the user is requesting the delayed ACK timer be set to. Note that
2627 * this value is defined in the standard to be between 200 and 500
2630 * sack_freq - This parameter contains the number of packets that must
2631 * be received before a sack is sent without waiting for the delay
2632 * timer to expire. The default value for this is 2, setting this
2633 * value to 1 will disable the delayed sack algorithm.
2636 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2637 char __user *optval, unsigned int optlen)
2639 struct sctp_sack_info params;
2640 struct sctp_transport *trans = NULL;
2641 struct sctp_association *asoc = NULL;
2642 struct sctp_sock *sp = sctp_sk(sk);
2644 if (optlen == sizeof(struct sctp_sack_info)) {
2645 if (copy_from_user(¶ms, optval, optlen))
2648 if (params.sack_delay == 0 && params.sack_freq == 0)
2650 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2651 pr_warn_ratelimited(DEPRECATED
2653 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2654 "Use struct sctp_sack_info instead\n",
2655 current->comm, task_pid_nr(current));
2656 if (copy_from_user(¶ms, optval, optlen))
2659 if (params.sack_delay == 0)
2660 params.sack_freq = 1;
2662 params.sack_freq = 0;
2666 /* Validate value parameter. */
2667 if (params.sack_delay > 500)
2670 /* Get association, if sack_assoc_id != 0 and the socket is a one
2671 * to many style socket, and an association was not found, then
2672 * the id was invalid.
2674 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2675 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2678 if (params.sack_delay) {
2681 msecs_to_jiffies(params.sack_delay);
2683 sctp_spp_sackdelay_enable(asoc->param_flags);
2685 sp->sackdelay = params.sack_delay;
2687 sctp_spp_sackdelay_enable(sp->param_flags);
2691 if (params.sack_freq == 1) {
2694 sctp_spp_sackdelay_disable(asoc->param_flags);
2697 sctp_spp_sackdelay_disable(sp->param_flags);
2699 } else if (params.sack_freq > 1) {
2701 asoc->sackfreq = params.sack_freq;
2703 sctp_spp_sackdelay_enable(asoc->param_flags);
2705 sp->sackfreq = params.sack_freq;
2707 sctp_spp_sackdelay_enable(sp->param_flags);
2711 /* If change is for association, also apply to each transport. */
2713 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2715 if (params.sack_delay) {
2717 msecs_to_jiffies(params.sack_delay);
2718 trans->param_flags =
2719 sctp_spp_sackdelay_enable(trans->param_flags);
2721 if (params.sack_freq == 1) {
2722 trans->param_flags =
2723 sctp_spp_sackdelay_disable(trans->param_flags);
2724 } else if (params.sack_freq > 1) {
2725 trans->sackfreq = params.sack_freq;
2726 trans->param_flags =
2727 sctp_spp_sackdelay_enable(trans->param_flags);
2735 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2737 * Applications can specify protocol parameters for the default association
2738 * initialization. The option name argument to setsockopt() and getsockopt()
2741 * Setting initialization parameters is effective only on an unconnected
2742 * socket (for UDP-style sockets only future associations are effected
2743 * by the change). With TCP-style sockets, this option is inherited by
2744 * sockets derived from a listener socket.
2746 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2748 struct sctp_initmsg sinit;
2749 struct sctp_sock *sp = sctp_sk(sk);
2751 if (optlen != sizeof(struct sctp_initmsg))
2753 if (copy_from_user(&sinit, optval, optlen))
2756 if (sinit.sinit_num_ostreams)
2757 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2758 if (sinit.sinit_max_instreams)
2759 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2760 if (sinit.sinit_max_attempts)
2761 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2762 if (sinit.sinit_max_init_timeo)
2763 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2769 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2771 * Applications that wish to use the sendto() system call may wish to
2772 * specify a default set of parameters that would normally be supplied
2773 * through the inclusion of ancillary data. This socket option allows
2774 * such an application to set the default sctp_sndrcvinfo structure.
2775 * The application that wishes to use this socket option simply passes
2776 * in to this call the sctp_sndrcvinfo structure defined in Section
2777 * 5.2.2) The input parameters accepted by this call include
2778 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2779 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2780 * to this call if the caller is using the UDP model.
2782 static int sctp_setsockopt_default_send_param(struct sock *sk,
2783 char __user *optval,
2784 unsigned int optlen)
2786 struct sctp_sock *sp = sctp_sk(sk);
2787 struct sctp_association *asoc;
2788 struct sctp_sndrcvinfo info;
2790 if (optlen != sizeof(info))
2792 if (copy_from_user(&info, optval, optlen))
2794 if (info.sinfo_flags &
2795 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2796 SCTP_ABORT | SCTP_EOF))
2799 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2800 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2803 asoc->default_stream = info.sinfo_stream;
2804 asoc->default_flags = info.sinfo_flags;
2805 asoc->default_ppid = info.sinfo_ppid;
2806 asoc->default_context = info.sinfo_context;
2807 asoc->default_timetolive = info.sinfo_timetolive;
2809 sp->default_stream = info.sinfo_stream;
2810 sp->default_flags = info.sinfo_flags;
2811 sp->default_ppid = info.sinfo_ppid;
2812 sp->default_context = info.sinfo_context;
2813 sp->default_timetolive = info.sinfo_timetolive;
2819 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2820 * (SCTP_DEFAULT_SNDINFO)
2822 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
2823 char __user *optval,
2824 unsigned int optlen)
2826 struct sctp_sock *sp = sctp_sk(sk);
2827 struct sctp_association *asoc;
2828 struct sctp_sndinfo info;
2830 if (optlen != sizeof(info))
2832 if (copy_from_user(&info, optval, optlen))
2834 if (info.snd_flags &
2835 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2836 SCTP_ABORT | SCTP_EOF))
2839 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
2840 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
2843 asoc->default_stream = info.snd_sid;
2844 asoc->default_flags = info.snd_flags;
2845 asoc->default_ppid = info.snd_ppid;
2846 asoc->default_context = info.snd_context;
2848 sp->default_stream = info.snd_sid;
2849 sp->default_flags = info.snd_flags;
2850 sp->default_ppid = info.snd_ppid;
2851 sp->default_context = info.snd_context;
2857 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2859 * Requests that the local SCTP stack use the enclosed peer address as
2860 * the association primary. The enclosed address must be one of the
2861 * association peer's addresses.
2863 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2864 unsigned int optlen)
2866 struct sctp_prim prim;
2867 struct sctp_transport *trans;
2869 if (optlen != sizeof(struct sctp_prim))
2872 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2875 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2879 sctp_assoc_set_primary(trans->asoc, trans);
2885 * 7.1.5 SCTP_NODELAY
2887 * Turn on/off any Nagle-like algorithm. This means that packets are
2888 * generally sent as soon as possible and no unnecessary delays are
2889 * introduced, at the cost of more packets in the network. Expects an
2890 * integer boolean flag.
2892 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2893 unsigned int optlen)
2897 if (optlen < sizeof(int))
2899 if (get_user(val, (int __user *)optval))
2902 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2908 * 7.1.1 SCTP_RTOINFO
2910 * The protocol parameters used to initialize and bound retransmission
2911 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2912 * and modify these parameters.
2913 * All parameters are time values, in milliseconds. A value of 0, when
2914 * modifying the parameters, indicates that the current value should not
2918 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2920 struct sctp_rtoinfo rtoinfo;
2921 struct sctp_association *asoc;
2922 unsigned long rto_min, rto_max;
2923 struct sctp_sock *sp = sctp_sk(sk);
2925 if (optlen != sizeof (struct sctp_rtoinfo))
2928 if (copy_from_user(&rtoinfo, optval, optlen))
2931 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2933 /* Set the values to the specific association */
2934 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2937 rto_max = rtoinfo.srto_max;
2938 rto_min = rtoinfo.srto_min;
2941 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2943 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2946 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2948 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2950 if (rto_min > rto_max)
2954 if (rtoinfo.srto_initial != 0)
2956 msecs_to_jiffies(rtoinfo.srto_initial);
2957 asoc->rto_max = rto_max;
2958 asoc->rto_min = rto_min;
2960 /* If there is no association or the association-id = 0
2961 * set the values to the endpoint.
2963 if (rtoinfo.srto_initial != 0)
2964 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2965 sp->rtoinfo.srto_max = rto_max;
2966 sp->rtoinfo.srto_min = rto_min;
2974 * 7.1.2 SCTP_ASSOCINFO
2976 * This option is used to tune the maximum retransmission attempts
2977 * of the association.
2978 * Returns an error if the new association retransmission value is
2979 * greater than the sum of the retransmission value of the peer.
2980 * See [SCTP] for more information.
2983 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2986 struct sctp_assocparams assocparams;
2987 struct sctp_association *asoc;
2989 if (optlen != sizeof(struct sctp_assocparams))
2991 if (copy_from_user(&assocparams, optval, optlen))
2994 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2996 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2999 /* Set the values to the specific association */
3001 if (assocparams.sasoc_asocmaxrxt != 0) {
3004 struct sctp_transport *peer_addr;
3006 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
3008 path_sum += peer_addr->pathmaxrxt;
3012 /* Only validate asocmaxrxt if we have more than
3013 * one path/transport. We do this because path
3014 * retransmissions are only counted when we have more
3018 assocparams.sasoc_asocmaxrxt > path_sum)
3021 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3024 if (assocparams.sasoc_cookie_life != 0)
3025 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3027 /* Set the values to the endpoint */
3028 struct sctp_sock *sp = sctp_sk(sk);
3030 if (assocparams.sasoc_asocmaxrxt != 0)
3031 sp->assocparams.sasoc_asocmaxrxt =
3032 assocparams.sasoc_asocmaxrxt;
3033 if (assocparams.sasoc_cookie_life != 0)
3034 sp->assocparams.sasoc_cookie_life =
3035 assocparams.sasoc_cookie_life;
3041 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3043 * This socket option is a boolean flag which turns on or off mapped V4
3044 * addresses. If this option is turned on and the socket is type
3045 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3046 * If this option is turned off, then no mapping will be done of V4
3047 * addresses and a user will receive both PF_INET6 and PF_INET type
3048 * addresses on the socket.
3050 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3053 struct sctp_sock *sp = sctp_sk(sk);
3055 if (optlen < sizeof(int))
3057 if (get_user(val, (int __user *)optval))
3068 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3069 * This option will get or set the maximum size to put in any outgoing
3070 * SCTP DATA chunk. If a message is larger than this size it will be
3071 * fragmented by SCTP into the specified size. Note that the underlying
3072 * SCTP implementation may fragment into smaller sized chunks when the
3073 * PMTU of the underlying association is smaller than the value set by
3074 * the user. The default value for this option is '0' which indicates
3075 * the user is NOT limiting fragmentation and only the PMTU will effect
3076 * SCTP's choice of DATA chunk size. Note also that values set larger
3077 * than the maximum size of an IP datagram will effectively let SCTP
3078 * control fragmentation (i.e. the same as setting this option to 0).
3080 * The following structure is used to access and modify this parameter:
3082 * struct sctp_assoc_value {
3083 * sctp_assoc_t assoc_id;
3084 * uint32_t assoc_value;
3087 * assoc_id: This parameter is ignored for one-to-one style sockets.
3088 * For one-to-many style sockets this parameter indicates which
3089 * association the user is performing an action upon. Note that if
3090 * this field's value is zero then the endpoints default value is
3091 * changed (effecting future associations only).
3092 * assoc_value: This parameter specifies the maximum size in bytes.
3094 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3096 struct sctp_assoc_value params;
3097 struct sctp_association *asoc;
3098 struct sctp_sock *sp = sctp_sk(sk);
3101 if (optlen == sizeof(int)) {
3102 pr_warn_ratelimited(DEPRECATED
3104 "Use of int in maxseg socket option.\n"
3105 "Use struct sctp_assoc_value instead\n",
3106 current->comm, task_pid_nr(current));
3107 if (copy_from_user(&val, optval, optlen))
3109 params.assoc_id = 0;
3110 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3111 if (copy_from_user(¶ms, optval, optlen))
3113 val = params.assoc_value;
3117 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3120 asoc = sctp_id2assoc(sk, params.assoc_id);
3121 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3126 val = asoc->pathmtu;
3127 val -= sp->pf->af->net_header_len;
3128 val -= sizeof(struct sctphdr) +
3129 sizeof(struct sctp_data_chunk);
3131 asoc->user_frag = val;
3132 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3134 sp->user_frag = val;
3142 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3144 * Requests that the peer mark the enclosed address as the association
3145 * primary. The enclosed address must be one of the association's
3146 * locally bound addresses. The following structure is used to make a
3147 * set primary request:
3149 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3150 unsigned int optlen)
3152 struct net *net = sock_net(sk);
3153 struct sctp_sock *sp;
3154 struct sctp_association *asoc = NULL;
3155 struct sctp_setpeerprim prim;
3156 struct sctp_chunk *chunk;
3162 if (!net->sctp.addip_enable)
3165 if (optlen != sizeof(struct sctp_setpeerprim))
3168 if (copy_from_user(&prim, optval, optlen))
3171 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3175 if (!asoc->peer.asconf_capable)
3178 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3181 if (!sctp_state(asoc, ESTABLISHED))
3184 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3188 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3189 return -EADDRNOTAVAIL;
3191 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3192 return -EADDRNOTAVAIL;
3194 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3195 chunk = sctp_make_asconf_set_prim(asoc,
3196 (union sctp_addr *)&prim.sspp_addr);
3200 err = sctp_send_asconf(asoc, chunk);
3202 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3207 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3208 unsigned int optlen)
3210 struct sctp_setadaptation adaptation;
3212 if (optlen != sizeof(struct sctp_setadaptation))
3214 if (copy_from_user(&adaptation, optval, optlen))
3217 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3223 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3225 * The context field in the sctp_sndrcvinfo structure is normally only
3226 * used when a failed message is retrieved holding the value that was
3227 * sent down on the actual send call. This option allows the setting of
3228 * a default context on an association basis that will be received on
3229 * reading messages from the peer. This is especially helpful in the
3230 * one-2-many model for an application to keep some reference to an
3231 * internal state machine that is processing messages on the
3232 * association. Note that the setting of this value only effects
3233 * received messages from the peer and does not effect the value that is
3234 * saved with outbound messages.
3236 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3237 unsigned int optlen)
3239 struct sctp_assoc_value params;
3240 struct sctp_sock *sp;
3241 struct sctp_association *asoc;
3243 if (optlen != sizeof(struct sctp_assoc_value))
3245 if (copy_from_user(¶ms, optval, optlen))
3250 if (params.assoc_id != 0) {
3251 asoc = sctp_id2assoc(sk, params.assoc_id);
3254 asoc->default_rcv_context = params.assoc_value;
3256 sp->default_rcv_context = params.assoc_value;
3263 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3265 * This options will at a minimum specify if the implementation is doing
3266 * fragmented interleave. Fragmented interleave, for a one to many
3267 * socket, is when subsequent calls to receive a message may return
3268 * parts of messages from different associations. Some implementations
3269 * may allow you to turn this value on or off. If so, when turned off,
3270 * no fragment interleave will occur (which will cause a head of line
3271 * blocking amongst multiple associations sharing the same one to many
3272 * socket). When this option is turned on, then each receive call may
3273 * come from a different association (thus the user must receive data
3274 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3275 * association each receive belongs to.
3277 * This option takes a boolean value. A non-zero value indicates that
3278 * fragmented interleave is on. A value of zero indicates that
3279 * fragmented interleave is off.
3281 * Note that it is important that an implementation that allows this
3282 * option to be turned on, have it off by default. Otherwise an unaware
3283 * application using the one to many model may become confused and act
3286 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3287 char __user *optval,
3288 unsigned int optlen)
3292 if (optlen != sizeof(int))
3294 if (get_user(val, (int __user *)optval))
3297 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3303 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3304 * (SCTP_PARTIAL_DELIVERY_POINT)
3306 * This option will set or get the SCTP partial delivery point. This
3307 * point is the size of a message where the partial delivery API will be
3308 * invoked to help free up rwnd space for the peer. Setting this to a
3309 * lower value will cause partial deliveries to happen more often. The
3310 * calls argument is an integer that sets or gets the partial delivery
3311 * point. Note also that the call will fail if the user attempts to set
3312 * this value larger than the socket receive buffer size.
3314 * Note that any single message having a length smaller than or equal to
3315 * the SCTP partial delivery point will be delivered in one single read
3316 * call as long as the user provided buffer is large enough to hold the
3319 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3320 char __user *optval,
3321 unsigned int optlen)
3325 if (optlen != sizeof(u32))
3327 if (get_user(val, (int __user *)optval))
3330 /* Note: We double the receive buffer from what the user sets
3331 * it to be, also initial rwnd is based on rcvbuf/2.
3333 if (val > (sk->sk_rcvbuf >> 1))
3336 sctp_sk(sk)->pd_point = val;
3338 return 0; /* is this the right error code? */
3342 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3344 * This option will allow a user to change the maximum burst of packets
3345 * that can be emitted by this association. Note that the default value
3346 * is 4, and some implementations may restrict this setting so that it
3347 * can only be lowered.
3349 * NOTE: This text doesn't seem right. Do this on a socket basis with
3350 * future associations inheriting the socket value.
3352 static int sctp_setsockopt_maxburst(struct sock *sk,
3353 char __user *optval,
3354 unsigned int optlen)
3356 struct sctp_assoc_value params;
3357 struct sctp_sock *sp;
3358 struct sctp_association *asoc;
3362 if (optlen == sizeof(int)) {
3363 pr_warn_ratelimited(DEPRECATED
3365 "Use of int in max_burst socket option deprecated.\n"
3366 "Use struct sctp_assoc_value instead\n",
3367 current->comm, task_pid_nr(current));
3368 if (copy_from_user(&val, optval, optlen))
3370 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3371 if (copy_from_user(¶ms, optval, optlen))
3373 val = params.assoc_value;
3374 assoc_id = params.assoc_id;
3380 if (assoc_id != 0) {
3381 asoc = sctp_id2assoc(sk, assoc_id);
3384 asoc->max_burst = val;
3386 sp->max_burst = val;
3392 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3394 * This set option adds a chunk type that the user is requesting to be
3395 * received only in an authenticated way. Changes to the list of chunks
3396 * will only effect future associations on the socket.
3398 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3399 char __user *optval,
3400 unsigned int optlen)
3402 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3403 struct sctp_authchunk val;
3405 if (!ep->auth_enable)
3408 if (optlen != sizeof(struct sctp_authchunk))
3410 if (copy_from_user(&val, optval, optlen))
3413 switch (val.sauth_chunk) {
3415 case SCTP_CID_INIT_ACK:
3416 case SCTP_CID_SHUTDOWN_COMPLETE:
3421 /* add this chunk id to the endpoint */
3422 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3426 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3428 * This option gets or sets the list of HMAC algorithms that the local
3429 * endpoint requires the peer to use.
3431 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3432 char __user *optval,
3433 unsigned int optlen)
3435 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3436 struct sctp_hmacalgo *hmacs;
3440 if (!ep->auth_enable)
3443 if (optlen < sizeof(struct sctp_hmacalgo))
3446 hmacs = memdup_user(optval, optlen);
3448 return PTR_ERR(hmacs);
3450 idents = hmacs->shmac_num_idents;
3451 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3452 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3457 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3464 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3466 * This option will set a shared secret key which is used to build an
3467 * association shared key.
3469 static int sctp_setsockopt_auth_key(struct sock *sk,
3470 char __user *optval,
3471 unsigned int optlen)
3473 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3474 struct sctp_authkey *authkey;
3475 struct sctp_association *asoc;
3478 if (!ep->auth_enable)
3481 if (optlen <= sizeof(struct sctp_authkey))
3484 authkey = memdup_user(optval, optlen);
3485 if (IS_ERR(authkey))
3486 return PTR_ERR(authkey);
3488 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3493 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3494 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3499 ret = sctp_auth_set_key(ep, asoc, authkey);
3506 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3508 * This option will get or set the active shared key to be used to build
3509 * the association shared key.
3511 static int sctp_setsockopt_active_key(struct sock *sk,
3512 char __user *optval,
3513 unsigned int optlen)
3515 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3516 struct sctp_authkeyid val;
3517 struct sctp_association *asoc;
3519 if (!ep->auth_enable)
3522 if (optlen != sizeof(struct sctp_authkeyid))
3524 if (copy_from_user(&val, optval, optlen))
3527 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3528 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3531 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3535 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3537 * This set option will delete a shared secret key from use.
3539 static int sctp_setsockopt_del_key(struct sock *sk,
3540 char __user *optval,
3541 unsigned int optlen)
3543 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3544 struct sctp_authkeyid val;
3545 struct sctp_association *asoc;
3547 if (!ep->auth_enable)
3550 if (optlen != sizeof(struct sctp_authkeyid))
3552 if (copy_from_user(&val, optval, optlen))
3555 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3556 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3559 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3564 * 8.1.23 SCTP_AUTO_ASCONF
3566 * This option will enable or disable the use of the automatic generation of
3567 * ASCONF chunks to add and delete addresses to an existing association. Note
3568 * that this option has two caveats namely: a) it only affects sockets that
3569 * are bound to all addresses available to the SCTP stack, and b) the system
3570 * administrator may have an overriding control that turns the ASCONF feature
3571 * off no matter what setting the socket option may have.
3572 * This option expects an integer boolean flag, where a non-zero value turns on
3573 * the option, and a zero value turns off the option.
3574 * Note. In this implementation, socket operation overrides default parameter
3575 * being set by sysctl as well as FreeBSD implementation
3577 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3578 unsigned int optlen)
3581 struct sctp_sock *sp = sctp_sk(sk);
3583 if (optlen < sizeof(int))
3585 if (get_user(val, (int __user *)optval))
3587 if (!sctp_is_ep_boundall(sk) && val)
3589 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3592 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3593 if (val == 0 && sp->do_auto_asconf) {
3594 list_del(&sp->auto_asconf_list);
3595 sp->do_auto_asconf = 0;
3596 } else if (val && !sp->do_auto_asconf) {
3597 list_add_tail(&sp->auto_asconf_list,
3598 &sock_net(sk)->sctp.auto_asconf_splist);
3599 sp->do_auto_asconf = 1;
3601 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3606 * SCTP_PEER_ADDR_THLDS
3608 * This option allows us to alter the partially failed threshold for one or all
3609 * transports in an association. See Section 6.1 of:
3610 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3612 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3613 char __user *optval,
3614 unsigned int optlen)
3616 struct sctp_paddrthlds val;
3617 struct sctp_transport *trans;
3618 struct sctp_association *asoc;
3620 if (optlen < sizeof(struct sctp_paddrthlds))
3622 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3623 sizeof(struct sctp_paddrthlds)))
3627 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3628 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3631 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3633 if (val.spt_pathmaxrxt)
3634 trans->pathmaxrxt = val.spt_pathmaxrxt;
3635 trans->pf_retrans = val.spt_pathpfthld;
3638 if (val.spt_pathmaxrxt)
3639 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3640 asoc->pf_retrans = val.spt_pathpfthld;
3642 trans = sctp_addr_id2transport(sk, &val.spt_address,
3647 if (val.spt_pathmaxrxt)
3648 trans->pathmaxrxt = val.spt_pathmaxrxt;
3649 trans->pf_retrans = val.spt_pathpfthld;
3655 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3656 char __user *optval,
3657 unsigned int optlen)
3661 if (optlen < sizeof(int))
3663 if (get_user(val, (int __user *) optval))
3666 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3671 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3672 char __user *optval,
3673 unsigned int optlen)
3677 if (optlen < sizeof(int))
3679 if (get_user(val, (int __user *) optval))
3682 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3687 static int sctp_setsockopt_pr_supported(struct sock *sk,
3688 char __user *optval,
3689 unsigned int optlen)
3691 struct sctp_assoc_value params;
3692 struct sctp_association *asoc;
3693 int retval = -EINVAL;
3695 if (optlen != sizeof(params))
3698 if (copy_from_user(¶ms, optval, optlen)) {
3703 asoc = sctp_id2assoc(sk, params.assoc_id);
3705 asoc->prsctp_enable = !!params.assoc_value;
3706 } else if (!params.assoc_id) {
3707 struct sctp_sock *sp = sctp_sk(sk);
3709 sp->ep->prsctp_enable = !!params.assoc_value;
3720 static int sctp_setsockopt_default_prinfo(struct sock *sk,
3721 char __user *optval,
3722 unsigned int optlen)
3724 struct sctp_default_prinfo info;
3725 struct sctp_association *asoc;
3726 int retval = -EINVAL;
3728 if (optlen != sizeof(info))
3731 if (copy_from_user(&info, optval, sizeof(info))) {
3736 if (info.pr_policy & ~SCTP_PR_SCTP_MASK)
3739 if (info.pr_policy == SCTP_PR_SCTP_NONE)
3742 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
3744 SCTP_PR_SET_POLICY(asoc->default_flags, info.pr_policy);
3745 asoc->default_timetolive = info.pr_value;
3746 } else if (!info.pr_assoc_id) {
3747 struct sctp_sock *sp = sctp_sk(sk);
3749 SCTP_PR_SET_POLICY(sp->default_flags, info.pr_policy);
3750 sp->default_timetolive = info.pr_value;
3761 static int sctp_setsockopt_reconfig_supported(struct sock *sk,
3762 char __user *optval,
3763 unsigned int optlen)
3765 struct sctp_assoc_value params;
3766 struct sctp_association *asoc;
3767 int retval = -EINVAL;
3769 if (optlen != sizeof(params))
3772 if (copy_from_user(¶ms, optval, optlen)) {
3777 asoc = sctp_id2assoc(sk, params.assoc_id);
3779 asoc->reconf_enable = !!params.assoc_value;
3780 } else if (!params.assoc_id) {
3781 struct sctp_sock *sp = sctp_sk(sk);
3783 sp->ep->reconf_enable = !!params.assoc_value;
3794 static int sctp_setsockopt_enable_strreset(struct sock *sk,
3795 char __user *optval,
3796 unsigned int optlen)
3798 struct sctp_assoc_value params;
3799 struct sctp_association *asoc;
3800 int retval = -EINVAL;
3802 if (optlen != sizeof(params))
3805 if (copy_from_user(¶ms, optval, optlen)) {
3810 if (params.assoc_value & (~SCTP_ENABLE_STRRESET_MASK))
3813 asoc = sctp_id2assoc(sk, params.assoc_id);
3815 asoc->strreset_enable = params.assoc_value;
3816 } else if (!params.assoc_id) {
3817 struct sctp_sock *sp = sctp_sk(sk);
3819 sp->ep->strreset_enable = params.assoc_value;
3830 static int sctp_setsockopt_reset_streams(struct sock *sk,
3831 char __user *optval,
3832 unsigned int optlen)
3834 struct sctp_reset_streams *params;
3835 struct sctp_association *asoc;
3836 int retval = -EINVAL;
3838 if (optlen < sizeof(struct sctp_reset_streams))
3841 params = memdup_user(optval, optlen);
3843 return PTR_ERR(params);
3845 asoc = sctp_id2assoc(sk, params->srs_assoc_id);
3849 retval = sctp_send_reset_streams(asoc, params);
3856 static int sctp_setsockopt_reset_assoc(struct sock *sk,
3857 char __user *optval,
3858 unsigned int optlen)
3860 struct sctp_association *asoc;
3861 sctp_assoc_t associd;
3862 int retval = -EINVAL;
3864 if (optlen != sizeof(associd))
3867 if (copy_from_user(&associd, optval, optlen)) {
3872 asoc = sctp_id2assoc(sk, associd);
3876 retval = sctp_send_reset_assoc(asoc);
3882 static int sctp_setsockopt_add_streams(struct sock *sk,
3883 char __user *optval,
3884 unsigned int optlen)
3886 struct sctp_association *asoc;
3887 struct sctp_add_streams params;
3888 int retval = -EINVAL;
3890 if (optlen != sizeof(params))
3893 if (copy_from_user(¶ms, optval, optlen)) {
3898 asoc = sctp_id2assoc(sk, params.sas_assoc_id);
3902 retval = sctp_send_add_streams(asoc, ¶ms);
3908 /* API 6.2 setsockopt(), getsockopt()
3910 * Applications use setsockopt() and getsockopt() to set or retrieve
3911 * socket options. Socket options are used to change the default
3912 * behavior of sockets calls. They are described in Section 7.
3916 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3917 * int __user *optlen);
3918 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3921 * sd - the socket descript.
3922 * level - set to IPPROTO_SCTP for all SCTP options.
3923 * optname - the option name.
3924 * optval - the buffer to store the value of the option.
3925 * optlen - the size of the buffer.
3927 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3928 char __user *optval, unsigned int optlen)
3932 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3934 /* I can hardly begin to describe how wrong this is. This is
3935 * so broken as to be worse than useless. The API draft
3936 * REALLY is NOT helpful here... I am not convinced that the
3937 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3938 * are at all well-founded.
3940 if (level != SOL_SCTP) {
3941 struct sctp_af *af = sctp_sk(sk)->pf->af;
3942 retval = af->setsockopt(sk, level, optname, optval, optlen);
3949 case SCTP_SOCKOPT_BINDX_ADD:
3950 /* 'optlen' is the size of the addresses buffer. */
3951 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3952 optlen, SCTP_BINDX_ADD_ADDR);
3955 case SCTP_SOCKOPT_BINDX_REM:
3956 /* 'optlen' is the size of the addresses buffer. */
3957 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3958 optlen, SCTP_BINDX_REM_ADDR);
3961 case SCTP_SOCKOPT_CONNECTX_OLD:
3962 /* 'optlen' is the size of the addresses buffer. */
3963 retval = sctp_setsockopt_connectx_old(sk,
3964 (struct sockaddr __user *)optval,
3968 case SCTP_SOCKOPT_CONNECTX:
3969 /* 'optlen' is the size of the addresses buffer. */
3970 retval = sctp_setsockopt_connectx(sk,
3971 (struct sockaddr __user *)optval,
3975 case SCTP_DISABLE_FRAGMENTS:
3976 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3980 retval = sctp_setsockopt_events(sk, optval, optlen);
3983 case SCTP_AUTOCLOSE:
3984 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3987 case SCTP_PEER_ADDR_PARAMS:
3988 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3991 case SCTP_DELAYED_SACK:
3992 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3994 case SCTP_PARTIAL_DELIVERY_POINT:
3995 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3999 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
4001 case SCTP_DEFAULT_SEND_PARAM:
4002 retval = sctp_setsockopt_default_send_param(sk, optval,
4005 case SCTP_DEFAULT_SNDINFO:
4006 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
4008 case SCTP_PRIMARY_ADDR:
4009 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
4011 case SCTP_SET_PEER_PRIMARY_ADDR:
4012 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
4015 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
4018 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
4020 case SCTP_ASSOCINFO:
4021 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
4023 case SCTP_I_WANT_MAPPED_V4_ADDR:
4024 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
4027 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
4029 case SCTP_ADAPTATION_LAYER:
4030 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
4033 retval = sctp_setsockopt_context(sk, optval, optlen);
4035 case SCTP_FRAGMENT_INTERLEAVE:
4036 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
4038 case SCTP_MAX_BURST:
4039 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
4041 case SCTP_AUTH_CHUNK:
4042 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
4044 case SCTP_HMAC_IDENT:
4045 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
4048 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
4050 case SCTP_AUTH_ACTIVE_KEY:
4051 retval = sctp_setsockopt_active_key(sk, optval, optlen);
4053 case SCTP_AUTH_DELETE_KEY:
4054 retval = sctp_setsockopt_del_key(sk, optval, optlen);
4056 case SCTP_AUTO_ASCONF:
4057 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
4059 case SCTP_PEER_ADDR_THLDS:
4060 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
4062 case SCTP_RECVRCVINFO:
4063 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
4065 case SCTP_RECVNXTINFO:
4066 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
4068 case SCTP_PR_SUPPORTED:
4069 retval = sctp_setsockopt_pr_supported(sk, optval, optlen);
4071 case SCTP_DEFAULT_PRINFO:
4072 retval = sctp_setsockopt_default_prinfo(sk, optval, optlen);
4074 case SCTP_RECONFIG_SUPPORTED:
4075 retval = sctp_setsockopt_reconfig_supported(sk, optval, optlen);
4077 case SCTP_ENABLE_STREAM_RESET:
4078 retval = sctp_setsockopt_enable_strreset(sk, optval, optlen);
4080 case SCTP_RESET_STREAMS:
4081 retval = sctp_setsockopt_reset_streams(sk, optval, optlen);
4083 case SCTP_RESET_ASSOC:
4084 retval = sctp_setsockopt_reset_assoc(sk, optval, optlen);
4086 case SCTP_ADD_STREAMS:
4087 retval = sctp_setsockopt_add_streams(sk, optval, optlen);
4090 retval = -ENOPROTOOPT;
4100 /* API 3.1.6 connect() - UDP Style Syntax
4102 * An application may use the connect() call in the UDP model to initiate an
4103 * association without sending data.
4107 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
4109 * sd: the socket descriptor to have a new association added to.
4111 * nam: the address structure (either struct sockaddr_in or struct
4112 * sockaddr_in6 defined in RFC2553 [7]).
4114 * len: the size of the address.
4116 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
4124 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
4127 /* Validate addr_len before calling common connect/connectx routine. */
4128 af = sctp_get_af_specific(addr->sa_family);
4129 if (!af || addr_len < af->sockaddr_len) {
4132 /* Pass correct addr len to common routine (so it knows there
4133 * is only one address being passed.
4135 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
4142 /* FIXME: Write comments. */
4143 static int sctp_disconnect(struct sock *sk, int flags)
4145 return -EOPNOTSUPP; /* STUB */
4148 /* 4.1.4 accept() - TCP Style Syntax
4150 * Applications use accept() call to remove an established SCTP
4151 * association from the accept queue of the endpoint. A new socket
4152 * descriptor will be returned from accept() to represent the newly
4153 * formed association.
4155 static struct sock *sctp_accept(struct sock *sk, int flags, int *err, bool kern)
4157 struct sctp_sock *sp;
4158 struct sctp_endpoint *ep;
4159 struct sock *newsk = NULL;
4160 struct sctp_association *asoc;
4169 if (!sctp_style(sk, TCP)) {
4170 error = -EOPNOTSUPP;
4174 if (!sctp_sstate(sk, LISTENING)) {
4179 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
4181 error = sctp_wait_for_accept(sk, timeo);
4185 /* We treat the list of associations on the endpoint as the accept
4186 * queue and pick the first association on the list.
4188 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
4190 newsk = sp->pf->create_accept_sk(sk, asoc, kern);
4196 /* Populate the fields of the newsk from the oldsk and migrate the
4197 * asoc to the newsk.
4199 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
4207 /* The SCTP ioctl handler. */
4208 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
4215 * SEQPACKET-style sockets in LISTENING state are valid, for
4216 * SCTP, so only discard TCP-style sockets in LISTENING state.
4218 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4223 struct sk_buff *skb;
4224 unsigned int amount = 0;
4226 skb = skb_peek(&sk->sk_receive_queue);
4229 * We will only return the amount of this packet since
4230 * that is all that will be read.
4234 rc = put_user(amount, (int __user *)arg);
4246 /* This is the function which gets called during socket creation to
4247 * initialized the SCTP-specific portion of the sock.
4248 * The sock structure should already be zero-filled memory.
4250 static int sctp_init_sock(struct sock *sk)
4252 struct net *net = sock_net(sk);
4253 struct sctp_sock *sp;
4255 pr_debug("%s: sk:%p\n", __func__, sk);
4259 /* Initialize the SCTP per socket area. */
4260 switch (sk->sk_type) {
4261 case SOCK_SEQPACKET:
4262 sp->type = SCTP_SOCKET_UDP;
4265 sp->type = SCTP_SOCKET_TCP;
4268 return -ESOCKTNOSUPPORT;
4271 sk->sk_gso_type = SKB_GSO_SCTP;
4273 /* Initialize default send parameters. These parameters can be
4274 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4276 sp->default_stream = 0;
4277 sp->default_ppid = 0;
4278 sp->default_flags = 0;
4279 sp->default_context = 0;
4280 sp->default_timetolive = 0;
4282 sp->default_rcv_context = 0;
4283 sp->max_burst = net->sctp.max_burst;
4285 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4287 /* Initialize default setup parameters. These parameters
4288 * can be modified with the SCTP_INITMSG socket option or
4289 * overridden by the SCTP_INIT CMSG.
4291 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4292 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4293 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4294 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4296 /* Initialize default RTO related parameters. These parameters can
4297 * be modified for with the SCTP_RTOINFO socket option.
4299 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4300 sp->rtoinfo.srto_max = net->sctp.rto_max;
4301 sp->rtoinfo.srto_min = net->sctp.rto_min;
4303 /* Initialize default association related parameters. These parameters
4304 * can be modified with the SCTP_ASSOCINFO socket option.
4306 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4307 sp->assocparams.sasoc_number_peer_destinations = 0;
4308 sp->assocparams.sasoc_peer_rwnd = 0;
4309 sp->assocparams.sasoc_local_rwnd = 0;
4310 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4312 /* Initialize default event subscriptions. By default, all the
4315 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4317 /* Default Peer Address Parameters. These defaults can
4318 * be modified via SCTP_PEER_ADDR_PARAMS
4320 sp->hbinterval = net->sctp.hb_interval;
4321 sp->pathmaxrxt = net->sctp.max_retrans_path;
4322 sp->pathmtu = 0; /* allow default discovery */
4323 sp->sackdelay = net->sctp.sack_timeout;
4325 sp->param_flags = SPP_HB_ENABLE |
4327 SPP_SACKDELAY_ENABLE;
4329 /* If enabled no SCTP message fragmentation will be performed.
4330 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4332 sp->disable_fragments = 0;
4334 /* Enable Nagle algorithm by default. */
4337 sp->recvrcvinfo = 0;
4338 sp->recvnxtinfo = 0;
4340 /* Enable by default. */
4343 /* Auto-close idle associations after the configured
4344 * number of seconds. A value of 0 disables this
4345 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4346 * for UDP-style sockets only.
4350 /* User specified fragmentation limit. */
4353 sp->adaptation_ind = 0;
4355 sp->pf = sctp_get_pf_specific(sk->sk_family);
4357 /* Control variables for partial data delivery. */
4358 atomic_set(&sp->pd_mode, 0);
4359 skb_queue_head_init(&sp->pd_lobby);
4360 sp->frag_interleave = 0;
4362 /* Create a per socket endpoint structure. Even if we
4363 * change the data structure relationships, this may still
4364 * be useful for storing pre-connect address information.
4366 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4372 sk->sk_destruct = sctp_destruct_sock;
4374 SCTP_DBG_OBJCNT_INC(sock);
4377 percpu_counter_inc(&sctp_sockets_allocated);
4378 sock_prot_inuse_add(net, sk->sk_prot, 1);
4380 /* Nothing can fail after this block, otherwise
4381 * sctp_destroy_sock() will be called without addr_wq_lock held
4383 if (net->sctp.default_auto_asconf) {
4384 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
4385 list_add_tail(&sp->auto_asconf_list,
4386 &net->sctp.auto_asconf_splist);
4387 sp->do_auto_asconf = 1;
4388 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4390 sp->do_auto_asconf = 0;
4398 /* Cleanup any SCTP per socket resources. Must be called with
4399 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4401 static void sctp_destroy_sock(struct sock *sk)
4403 struct sctp_sock *sp;
4405 pr_debug("%s: sk:%p\n", __func__, sk);
4407 /* Release our hold on the endpoint. */
4409 /* This could happen during socket init, thus we bail out
4410 * early, since the rest of the below is not setup either.
4415 if (sp->do_auto_asconf) {
4416 sp->do_auto_asconf = 0;
4417 list_del(&sp->auto_asconf_list);
4419 sctp_endpoint_free(sp->ep);
4421 percpu_counter_dec(&sctp_sockets_allocated);
4422 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4426 /* Triggered when there are no references on the socket anymore */
4427 static void sctp_destruct_sock(struct sock *sk)
4429 struct sctp_sock *sp = sctp_sk(sk);
4431 /* Free up the HMAC transform. */
4432 crypto_free_shash(sp->hmac);
4434 inet_sock_destruct(sk);
4437 /* API 4.1.7 shutdown() - TCP Style Syntax
4438 * int shutdown(int socket, int how);
4440 * sd - the socket descriptor of the association to be closed.
4441 * how - Specifies the type of shutdown. The values are
4444 * Disables further receive operations. No SCTP
4445 * protocol action is taken.
4447 * Disables further send operations, and initiates
4448 * the SCTP shutdown sequence.
4450 * Disables further send and receive operations
4451 * and initiates the SCTP shutdown sequence.
4453 static void sctp_shutdown(struct sock *sk, int how)
4455 struct net *net = sock_net(sk);
4456 struct sctp_endpoint *ep;
4458 if (!sctp_style(sk, TCP))
4461 ep = sctp_sk(sk)->ep;
4462 if (how & SEND_SHUTDOWN && !list_empty(&ep->asocs)) {
4463 struct sctp_association *asoc;
4465 sk->sk_state = SCTP_SS_CLOSING;
4466 asoc = list_entry(ep->asocs.next,
4467 struct sctp_association, asocs);
4468 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4472 int sctp_get_sctp_info(struct sock *sk, struct sctp_association *asoc,
4473 struct sctp_info *info)
4475 struct sctp_transport *prim;
4476 struct list_head *pos;
4479 memset(info, 0, sizeof(*info));
4481 struct sctp_sock *sp = sctp_sk(sk);
4483 info->sctpi_s_autoclose = sp->autoclose;
4484 info->sctpi_s_adaptation_ind = sp->adaptation_ind;
4485 info->sctpi_s_pd_point = sp->pd_point;
4486 info->sctpi_s_nodelay = sp->nodelay;
4487 info->sctpi_s_disable_fragments = sp->disable_fragments;
4488 info->sctpi_s_v4mapped = sp->v4mapped;
4489 info->sctpi_s_frag_interleave = sp->frag_interleave;
4490 info->sctpi_s_type = sp->type;
4495 info->sctpi_tag = asoc->c.my_vtag;
4496 info->sctpi_state = asoc->state;
4497 info->sctpi_rwnd = asoc->a_rwnd;
4498 info->sctpi_unackdata = asoc->unack_data;
4499 info->sctpi_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4500 info->sctpi_instrms = asoc->stream.incnt;
4501 info->sctpi_outstrms = asoc->stream.outcnt;
4502 list_for_each(pos, &asoc->base.inqueue.in_chunk_list)
4503 info->sctpi_inqueue++;
4504 list_for_each(pos, &asoc->outqueue.out_chunk_list)
4505 info->sctpi_outqueue++;
4506 info->sctpi_overall_error = asoc->overall_error_count;
4507 info->sctpi_max_burst = asoc->max_burst;
4508 info->sctpi_maxseg = asoc->frag_point;
4509 info->sctpi_peer_rwnd = asoc->peer.rwnd;
4510 info->sctpi_peer_tag = asoc->c.peer_vtag;
4512 mask = asoc->peer.ecn_capable << 1;
4513 mask = (mask | asoc->peer.ipv4_address) << 1;
4514 mask = (mask | asoc->peer.ipv6_address) << 1;
4515 mask = (mask | asoc->peer.hostname_address) << 1;
4516 mask = (mask | asoc->peer.asconf_capable) << 1;
4517 mask = (mask | asoc->peer.prsctp_capable) << 1;
4518 mask = (mask | asoc->peer.auth_capable);
4519 info->sctpi_peer_capable = mask;
4520 mask = asoc->peer.sack_needed << 1;
4521 mask = (mask | asoc->peer.sack_generation) << 1;
4522 mask = (mask | asoc->peer.zero_window_announced);
4523 info->sctpi_peer_sack = mask;
4525 info->sctpi_isacks = asoc->stats.isacks;
4526 info->sctpi_osacks = asoc->stats.osacks;
4527 info->sctpi_opackets = asoc->stats.opackets;
4528 info->sctpi_ipackets = asoc->stats.ipackets;
4529 info->sctpi_rtxchunks = asoc->stats.rtxchunks;
4530 info->sctpi_outofseqtsns = asoc->stats.outofseqtsns;
4531 info->sctpi_idupchunks = asoc->stats.idupchunks;
4532 info->sctpi_gapcnt = asoc->stats.gapcnt;
4533 info->sctpi_ouodchunks = asoc->stats.ouodchunks;
4534 info->sctpi_iuodchunks = asoc->stats.iuodchunks;
4535 info->sctpi_oodchunks = asoc->stats.oodchunks;
4536 info->sctpi_iodchunks = asoc->stats.iodchunks;
4537 info->sctpi_octrlchunks = asoc->stats.octrlchunks;
4538 info->sctpi_ictrlchunks = asoc->stats.ictrlchunks;
4540 prim = asoc->peer.primary_path;
4541 memcpy(&info->sctpi_p_address, &prim->ipaddr, sizeof(prim->ipaddr));
4542 info->sctpi_p_state = prim->state;
4543 info->sctpi_p_cwnd = prim->cwnd;
4544 info->sctpi_p_srtt = prim->srtt;
4545 info->sctpi_p_rto = jiffies_to_msecs(prim->rto);
4546 info->sctpi_p_hbinterval = prim->hbinterval;
4547 info->sctpi_p_pathmaxrxt = prim->pathmaxrxt;
4548 info->sctpi_p_sackdelay = jiffies_to_msecs(prim->sackdelay);
4549 info->sctpi_p_ssthresh = prim->ssthresh;
4550 info->sctpi_p_partial_bytes_acked = prim->partial_bytes_acked;
4551 info->sctpi_p_flight_size = prim->flight_size;
4552 info->sctpi_p_error = prim->error_count;
4556 EXPORT_SYMBOL_GPL(sctp_get_sctp_info);
4558 /* use callback to avoid exporting the core structure */
4559 int sctp_transport_walk_start(struct rhashtable_iter *iter)
4563 rhltable_walk_enter(&sctp_transport_hashtable, iter);
4565 err = rhashtable_walk_start(iter);
4566 if (err && err != -EAGAIN) {
4567 rhashtable_walk_stop(iter);
4568 rhashtable_walk_exit(iter);
4575 void sctp_transport_walk_stop(struct rhashtable_iter *iter)
4577 rhashtable_walk_stop(iter);
4578 rhashtable_walk_exit(iter);
4581 struct sctp_transport *sctp_transport_get_next(struct net *net,
4582 struct rhashtable_iter *iter)
4584 struct sctp_transport *t;
4586 t = rhashtable_walk_next(iter);
4587 for (; t; t = rhashtable_walk_next(iter)) {
4589 if (PTR_ERR(t) == -EAGAIN)
4594 if (net_eq(sock_net(t->asoc->base.sk), net) &&
4595 t->asoc->peer.primary_path == t)
4602 struct sctp_transport *sctp_transport_get_idx(struct net *net,
4603 struct rhashtable_iter *iter,
4606 void *obj = SEQ_START_TOKEN;
4608 while (pos && (obj = sctp_transport_get_next(net, iter)) &&
4615 int sctp_for_each_endpoint(int (*cb)(struct sctp_endpoint *, void *),
4619 struct sctp_ep_common *epb;
4620 struct sctp_hashbucket *head;
4622 for (head = sctp_ep_hashtable; hash < sctp_ep_hashsize;
4624 read_lock_bh(&head->lock);
4625 sctp_for_each_hentry(epb, &head->chain) {
4626 err = cb(sctp_ep(epb), p);
4630 read_unlock_bh(&head->lock);
4635 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint);
4637 int sctp_transport_lookup_process(int (*cb)(struct sctp_transport *, void *),
4639 const union sctp_addr *laddr,
4640 const union sctp_addr *paddr, void *p)
4642 struct sctp_transport *transport;
4646 transport = sctp_addrs_lookup_transport(net, laddr, paddr);
4651 err = cb(transport, p);
4652 sctp_transport_put(transport);
4656 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process);
4658 int sctp_for_each_transport(int (*cb)(struct sctp_transport *, void *),
4659 struct net *net, int pos, void *p) {
4660 struct rhashtable_iter hti;
4664 err = sctp_transport_walk_start(&hti);
4668 obj = sctp_transport_get_idx(net, &hti, pos + 1);
4669 for (; !IS_ERR_OR_NULL(obj); obj = sctp_transport_get_next(net, &hti)) {
4670 struct sctp_transport *transport = obj;
4672 if (!sctp_transport_hold(transport))
4674 err = cb(transport, p);
4675 sctp_transport_put(transport);
4679 sctp_transport_walk_stop(&hti);
4683 EXPORT_SYMBOL_GPL(sctp_for_each_transport);
4685 /* 7.2.1 Association Status (SCTP_STATUS)
4687 * Applications can retrieve current status information about an
4688 * association, including association state, peer receiver window size,
4689 * number of unacked data chunks, and number of data chunks pending
4690 * receipt. This information is read-only.
4692 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4693 char __user *optval,
4696 struct sctp_status status;
4697 struct sctp_association *asoc = NULL;
4698 struct sctp_transport *transport;
4699 sctp_assoc_t associd;
4702 if (len < sizeof(status)) {
4707 len = sizeof(status);
4708 if (copy_from_user(&status, optval, len)) {
4713 associd = status.sstat_assoc_id;
4714 asoc = sctp_id2assoc(sk, associd);
4720 transport = asoc->peer.primary_path;
4722 status.sstat_assoc_id = sctp_assoc2id(asoc);
4723 status.sstat_state = sctp_assoc_to_state(asoc);
4724 status.sstat_rwnd = asoc->peer.rwnd;
4725 status.sstat_unackdata = asoc->unack_data;
4727 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4728 status.sstat_instrms = asoc->stream.incnt;
4729 status.sstat_outstrms = asoc->stream.outcnt;
4730 status.sstat_fragmentation_point = asoc->frag_point;
4731 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4732 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4733 transport->af_specific->sockaddr_len);
4734 /* Map ipv4 address into v4-mapped-on-v6 address. */
4735 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
4736 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4737 status.sstat_primary.spinfo_state = transport->state;
4738 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4739 status.sstat_primary.spinfo_srtt = transport->srtt;
4740 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4741 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4743 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4744 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4746 if (put_user(len, optlen)) {
4751 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4752 __func__, len, status.sstat_state, status.sstat_rwnd,
4753 status.sstat_assoc_id);
4755 if (copy_to_user(optval, &status, len)) {
4765 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4767 * Applications can retrieve information about a specific peer address
4768 * of an association, including its reachability state, congestion
4769 * window, and retransmission timer values. This information is
4772 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4773 char __user *optval,
4776 struct sctp_paddrinfo pinfo;
4777 struct sctp_transport *transport;
4780 if (len < sizeof(pinfo)) {
4785 len = sizeof(pinfo);
4786 if (copy_from_user(&pinfo, optval, len)) {
4791 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4792 pinfo.spinfo_assoc_id);
4796 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4797 pinfo.spinfo_state = transport->state;
4798 pinfo.spinfo_cwnd = transport->cwnd;
4799 pinfo.spinfo_srtt = transport->srtt;
4800 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4801 pinfo.spinfo_mtu = transport->pathmtu;
4803 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4804 pinfo.spinfo_state = SCTP_ACTIVE;
4806 if (put_user(len, optlen)) {
4811 if (copy_to_user(optval, &pinfo, len)) {
4820 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4822 * This option is a on/off flag. If enabled no SCTP message
4823 * fragmentation will be performed. Instead if a message being sent
4824 * exceeds the current PMTU size, the message will NOT be sent and
4825 * instead a error will be indicated to the user.
4827 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4828 char __user *optval, int __user *optlen)
4832 if (len < sizeof(int))
4836 val = (sctp_sk(sk)->disable_fragments == 1);
4837 if (put_user(len, optlen))
4839 if (copy_to_user(optval, &val, len))
4844 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4846 * This socket option is used to specify various notifications and
4847 * ancillary data the user wishes to receive.
4849 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4854 if (len > sizeof(struct sctp_event_subscribe))
4855 len = sizeof(struct sctp_event_subscribe);
4856 if (put_user(len, optlen))
4858 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4863 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4865 * This socket option is applicable to the UDP-style socket only. When
4866 * set it will cause associations that are idle for more than the
4867 * specified number of seconds to automatically close. An association
4868 * being idle is defined an association that has NOT sent or received
4869 * user data. The special value of '0' indicates that no automatic
4870 * close of any associations should be performed. The option expects an
4871 * integer defining the number of seconds of idle time before an
4872 * association is closed.
4874 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4876 /* Applicable to UDP-style socket only */
4877 if (sctp_style(sk, TCP))
4879 if (len < sizeof(int))
4882 if (put_user(len, optlen))
4884 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4889 /* Helper routine to branch off an association to a new socket. */
4890 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4892 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4893 struct sctp_sock *sp = sctp_sk(sk);
4894 struct socket *sock;
4900 /* If there is a thread waiting on more sndbuf space for
4901 * sending on this asoc, it cannot be peeled.
4903 if (waitqueue_active(&asoc->wait))
4906 /* An association cannot be branched off from an already peeled-off
4907 * socket, nor is this supported for tcp style sockets.
4909 if (!sctp_style(sk, UDP))
4912 /* Create a new socket. */
4913 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4917 sctp_copy_sock(sock->sk, sk, asoc);
4919 /* Make peeled-off sockets more like 1-1 accepted sockets.
4920 * Set the daddr and initialize id to something more random
4922 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
4924 /* Populate the fields of the newsk from the oldsk and migrate the
4925 * asoc to the newsk.
4927 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4933 EXPORT_SYMBOL(sctp_do_peeloff);
4935 static int sctp_getsockopt_peeloff_common(struct sock *sk, sctp_peeloff_arg_t *peeloff,
4936 struct file **newfile, unsigned flags)
4938 struct socket *newsock;
4941 retval = sctp_do_peeloff(sk, peeloff->associd, &newsock);
4945 /* Map the socket to an unused fd that can be returned to the user. */
4946 retval = get_unused_fd_flags(flags & SOCK_CLOEXEC);
4948 sock_release(newsock);
4952 *newfile = sock_alloc_file(newsock, 0, NULL);
4953 if (IS_ERR(*newfile)) {
4954 put_unused_fd(retval);
4955 sock_release(newsock);
4956 retval = PTR_ERR(*newfile);
4961 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4964 peeloff->sd = retval;
4966 if (flags & SOCK_NONBLOCK)
4967 (*newfile)->f_flags |= O_NONBLOCK;
4972 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4974 sctp_peeloff_arg_t peeloff;
4975 struct file *newfile = NULL;
4978 if (len < sizeof(sctp_peeloff_arg_t))
4980 len = sizeof(sctp_peeloff_arg_t);
4981 if (copy_from_user(&peeloff, optval, len))
4984 retval = sctp_getsockopt_peeloff_common(sk, &peeloff, &newfile, 0);
4988 /* Return the fd mapped to the new socket. */
4989 if (put_user(len, optlen)) {
4991 put_unused_fd(retval);
4995 if (copy_to_user(optval, &peeloff, len)) {
4997 put_unused_fd(retval);
5000 fd_install(retval, newfile);
5005 static int sctp_getsockopt_peeloff_flags(struct sock *sk, int len,
5006 char __user *optval, int __user *optlen)
5008 sctp_peeloff_flags_arg_t peeloff;
5009 struct file *newfile = NULL;
5012 if (len < sizeof(sctp_peeloff_flags_arg_t))
5014 len = sizeof(sctp_peeloff_flags_arg_t);
5015 if (copy_from_user(&peeloff, optval, len))
5018 retval = sctp_getsockopt_peeloff_common(sk, &peeloff.p_arg,
5019 &newfile, peeloff.flags);
5023 /* Return the fd mapped to the new socket. */
5024 if (put_user(len, optlen)) {
5026 put_unused_fd(retval);
5030 if (copy_to_user(optval, &peeloff, len)) {
5032 put_unused_fd(retval);
5035 fd_install(retval, newfile);
5040 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
5042 * Applications can enable or disable heartbeats for any peer address of
5043 * an association, modify an address's heartbeat interval, force a
5044 * heartbeat to be sent immediately, and adjust the address's maximum
5045 * number of retransmissions sent before an address is considered
5046 * unreachable. The following structure is used to access and modify an
5047 * address's parameters:
5049 * struct sctp_paddrparams {
5050 * sctp_assoc_t spp_assoc_id;
5051 * struct sockaddr_storage spp_address;
5052 * uint32_t spp_hbinterval;
5053 * uint16_t spp_pathmaxrxt;
5054 * uint32_t spp_pathmtu;
5055 * uint32_t spp_sackdelay;
5056 * uint32_t spp_flags;
5059 * spp_assoc_id - (one-to-many style socket) This is filled in the
5060 * application, and identifies the association for
5062 * spp_address - This specifies which address is of interest.
5063 * spp_hbinterval - This contains the value of the heartbeat interval,
5064 * in milliseconds. If a value of zero
5065 * is present in this field then no changes are to
5066 * be made to this parameter.
5067 * spp_pathmaxrxt - This contains the maximum number of
5068 * retransmissions before this address shall be
5069 * considered unreachable. If a value of zero
5070 * is present in this field then no changes are to
5071 * be made to this parameter.
5072 * spp_pathmtu - When Path MTU discovery is disabled the value
5073 * specified here will be the "fixed" path mtu.
5074 * Note that if the spp_address field is empty
5075 * then all associations on this address will
5076 * have this fixed path mtu set upon them.
5078 * spp_sackdelay - When delayed sack is enabled, this value specifies
5079 * the number of milliseconds that sacks will be delayed
5080 * for. This value will apply to all addresses of an
5081 * association if the spp_address field is empty. Note
5082 * also, that if delayed sack is enabled and this
5083 * value is set to 0, no change is made to the last
5084 * recorded delayed sack timer value.
5086 * spp_flags - These flags are used to control various features
5087 * on an association. The flag field may contain
5088 * zero or more of the following options.
5090 * SPP_HB_ENABLE - Enable heartbeats on the
5091 * specified address. Note that if the address
5092 * field is empty all addresses for the association
5093 * have heartbeats enabled upon them.
5095 * SPP_HB_DISABLE - Disable heartbeats on the
5096 * speicifed address. Note that if the address
5097 * field is empty all addresses for the association
5098 * will have their heartbeats disabled. Note also
5099 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
5100 * mutually exclusive, only one of these two should
5101 * be specified. Enabling both fields will have
5102 * undetermined results.
5104 * SPP_HB_DEMAND - Request a user initiated heartbeat
5105 * to be made immediately.
5107 * SPP_PMTUD_ENABLE - This field will enable PMTU
5108 * discovery upon the specified address. Note that
5109 * if the address feild is empty then all addresses
5110 * on the association are effected.
5112 * SPP_PMTUD_DISABLE - This field will disable PMTU
5113 * discovery upon the specified address. Note that
5114 * if the address feild is empty then all addresses
5115 * on the association are effected. Not also that
5116 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
5117 * exclusive. Enabling both will have undetermined
5120 * SPP_SACKDELAY_ENABLE - Setting this flag turns
5121 * on delayed sack. The time specified in spp_sackdelay
5122 * is used to specify the sack delay for this address. Note
5123 * that if spp_address is empty then all addresses will
5124 * enable delayed sack and take on the sack delay
5125 * value specified in spp_sackdelay.
5126 * SPP_SACKDELAY_DISABLE - Setting this flag turns
5127 * off delayed sack. If the spp_address field is blank then
5128 * delayed sack is disabled for the entire association. Note
5129 * also that this field is mutually exclusive to
5130 * SPP_SACKDELAY_ENABLE, setting both will have undefined
5133 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
5134 char __user *optval, int __user *optlen)
5136 struct sctp_paddrparams params;
5137 struct sctp_transport *trans = NULL;
5138 struct sctp_association *asoc = NULL;
5139 struct sctp_sock *sp = sctp_sk(sk);
5141 if (len < sizeof(struct sctp_paddrparams))
5143 len = sizeof(struct sctp_paddrparams);
5144 if (copy_from_user(¶ms, optval, len))
5147 /* If an address other than INADDR_ANY is specified, and
5148 * no transport is found, then the request is invalid.
5150 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
5151 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
5152 params.spp_assoc_id);
5154 pr_debug("%s: failed no transport\n", __func__);
5159 /* Get association, if assoc_id != 0 and the socket is a one
5160 * to many style socket, and an association was not found, then
5161 * the id was invalid.
5163 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
5164 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
5165 pr_debug("%s: failed no association\n", __func__);
5170 /* Fetch transport values. */
5171 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
5172 params.spp_pathmtu = trans->pathmtu;
5173 params.spp_pathmaxrxt = trans->pathmaxrxt;
5174 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
5176 /*draft-11 doesn't say what to return in spp_flags*/
5177 params.spp_flags = trans->param_flags;
5179 /* Fetch association values. */
5180 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
5181 params.spp_pathmtu = asoc->pathmtu;
5182 params.spp_pathmaxrxt = asoc->pathmaxrxt;
5183 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
5185 /*draft-11 doesn't say what to return in spp_flags*/
5186 params.spp_flags = asoc->param_flags;
5188 /* Fetch socket values. */
5189 params.spp_hbinterval = sp->hbinterval;
5190 params.spp_pathmtu = sp->pathmtu;
5191 params.spp_sackdelay = sp->sackdelay;
5192 params.spp_pathmaxrxt = sp->pathmaxrxt;
5194 /*draft-11 doesn't say what to return in spp_flags*/
5195 params.spp_flags = sp->param_flags;
5198 if (copy_to_user(optval, ¶ms, len))
5201 if (put_user(len, optlen))
5208 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
5210 * This option will effect the way delayed acks are performed. This
5211 * option allows you to get or set the delayed ack time, in
5212 * milliseconds. It also allows changing the delayed ack frequency.
5213 * Changing the frequency to 1 disables the delayed sack algorithm. If
5214 * the assoc_id is 0, then this sets or gets the endpoints default
5215 * values. If the assoc_id field is non-zero, then the set or get
5216 * effects the specified association for the one to many model (the
5217 * assoc_id field is ignored by the one to one model). Note that if
5218 * sack_delay or sack_freq are 0 when setting this option, then the
5219 * current values will remain unchanged.
5221 * struct sctp_sack_info {
5222 * sctp_assoc_t sack_assoc_id;
5223 * uint32_t sack_delay;
5224 * uint32_t sack_freq;
5227 * sack_assoc_id - This parameter, indicates which association the user
5228 * is performing an action upon. Note that if this field's value is
5229 * zero then the endpoints default value is changed (effecting future
5230 * associations only).
5232 * sack_delay - This parameter contains the number of milliseconds that
5233 * the user is requesting the delayed ACK timer be set to. Note that
5234 * this value is defined in the standard to be between 200 and 500
5237 * sack_freq - This parameter contains the number of packets that must
5238 * be received before a sack is sent without waiting for the delay
5239 * timer to expire. The default value for this is 2, setting this
5240 * value to 1 will disable the delayed sack algorithm.
5242 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
5243 char __user *optval,
5246 struct sctp_sack_info params;
5247 struct sctp_association *asoc = NULL;
5248 struct sctp_sock *sp = sctp_sk(sk);
5250 if (len >= sizeof(struct sctp_sack_info)) {
5251 len = sizeof(struct sctp_sack_info);
5253 if (copy_from_user(¶ms, optval, len))
5255 } else if (len == sizeof(struct sctp_assoc_value)) {
5256 pr_warn_ratelimited(DEPRECATED
5258 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
5259 "Use struct sctp_sack_info instead\n",
5260 current->comm, task_pid_nr(current));
5261 if (copy_from_user(¶ms, optval, len))
5266 /* Get association, if sack_assoc_id != 0 and the socket is a one
5267 * to many style socket, and an association was not found, then
5268 * the id was invalid.
5270 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
5271 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
5275 /* Fetch association values. */
5276 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
5277 params.sack_delay = jiffies_to_msecs(
5279 params.sack_freq = asoc->sackfreq;
5282 params.sack_delay = 0;
5283 params.sack_freq = 1;
5286 /* Fetch socket values. */
5287 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
5288 params.sack_delay = sp->sackdelay;
5289 params.sack_freq = sp->sackfreq;
5291 params.sack_delay = 0;
5292 params.sack_freq = 1;
5296 if (copy_to_user(optval, ¶ms, len))
5299 if (put_user(len, optlen))
5305 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5307 * Applications can specify protocol parameters for the default association
5308 * initialization. The option name argument to setsockopt() and getsockopt()
5311 * Setting initialization parameters is effective only on an unconnected
5312 * socket (for UDP-style sockets only future associations are effected
5313 * by the change). With TCP-style sockets, this option is inherited by
5314 * sockets derived from a listener socket.
5316 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
5318 if (len < sizeof(struct sctp_initmsg))
5320 len = sizeof(struct sctp_initmsg);
5321 if (put_user(len, optlen))
5323 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
5329 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
5330 char __user *optval, int __user *optlen)
5332 struct sctp_association *asoc;
5334 struct sctp_getaddrs getaddrs;
5335 struct sctp_transport *from;
5337 union sctp_addr temp;
5338 struct sctp_sock *sp = sctp_sk(sk);
5343 if (len < sizeof(struct sctp_getaddrs))
5346 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5349 /* For UDP-style sockets, id specifies the association to query. */
5350 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5354 to = optval + offsetof(struct sctp_getaddrs, addrs);
5355 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5357 list_for_each_entry(from, &asoc->peer.transport_addr_list,
5359 memcpy(&temp, &from->ipaddr, sizeof(temp));
5360 addrlen = sctp_get_pf_specific(sk->sk_family)
5361 ->addr_to_user(sp, &temp);
5362 if (space_left < addrlen)
5364 if (copy_to_user(to, &temp, addrlen))
5368 space_left -= addrlen;
5371 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
5373 bytes_copied = ((char __user *)to) - optval;
5374 if (put_user(bytes_copied, optlen))
5380 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
5381 size_t space_left, int *bytes_copied)
5383 struct sctp_sockaddr_entry *addr;
5384 union sctp_addr temp;
5387 struct net *net = sock_net(sk);
5390 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
5394 if ((PF_INET == sk->sk_family) &&
5395 (AF_INET6 == addr->a.sa.sa_family))
5397 if ((PF_INET6 == sk->sk_family) &&
5398 inet_v6_ipv6only(sk) &&
5399 (AF_INET == addr->a.sa.sa_family))
5401 memcpy(&temp, &addr->a, sizeof(temp));
5402 if (!temp.v4.sin_port)
5403 temp.v4.sin_port = htons(port);
5405 addrlen = sctp_get_pf_specific(sk->sk_family)
5406 ->addr_to_user(sctp_sk(sk), &temp);
5408 if (space_left < addrlen) {
5412 memcpy(to, &temp, addrlen);
5416 space_left -= addrlen;
5417 *bytes_copied += addrlen;
5425 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
5426 char __user *optval, int __user *optlen)
5428 struct sctp_bind_addr *bp;
5429 struct sctp_association *asoc;
5431 struct sctp_getaddrs getaddrs;
5432 struct sctp_sockaddr_entry *addr;
5434 union sctp_addr temp;
5435 struct sctp_sock *sp = sctp_sk(sk);
5439 int bytes_copied = 0;
5443 if (len < sizeof(struct sctp_getaddrs))
5446 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5450 * For UDP-style sockets, id specifies the association to query.
5451 * If the id field is set to the value '0' then the locally bound
5452 * addresses are returned without regard to any particular
5455 if (0 == getaddrs.assoc_id) {
5456 bp = &sctp_sk(sk)->ep->base.bind_addr;
5458 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5461 bp = &asoc->base.bind_addr;
5464 to = optval + offsetof(struct sctp_getaddrs, addrs);
5465 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5467 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN);
5471 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5472 * addresses from the global local address list.
5474 if (sctp_list_single_entry(&bp->address_list)) {
5475 addr = list_entry(bp->address_list.next,
5476 struct sctp_sockaddr_entry, list);
5477 if (sctp_is_any(sk, &addr->a)) {
5478 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
5479 space_left, &bytes_copied);
5489 /* Protection on the bound address list is not needed since
5490 * in the socket option context we hold a socket lock and
5491 * thus the bound address list can't change.
5493 list_for_each_entry(addr, &bp->address_list, list) {
5494 memcpy(&temp, &addr->a, sizeof(temp));
5495 addrlen = sctp_get_pf_specific(sk->sk_family)
5496 ->addr_to_user(sp, &temp);
5497 if (space_left < addrlen) {
5498 err = -ENOMEM; /*fixme: right error?*/
5501 memcpy(buf, &temp, addrlen);
5503 bytes_copied += addrlen;
5505 space_left -= addrlen;
5509 if (copy_to_user(to, addrs, bytes_copied)) {
5513 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
5517 if (put_user(bytes_copied, optlen))
5524 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
5526 * Requests that the local SCTP stack use the enclosed peer address as
5527 * the association primary. The enclosed address must be one of the
5528 * association peer's addresses.
5530 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
5531 char __user *optval, int __user *optlen)
5533 struct sctp_prim prim;
5534 struct sctp_association *asoc;
5535 struct sctp_sock *sp = sctp_sk(sk);
5537 if (len < sizeof(struct sctp_prim))
5540 len = sizeof(struct sctp_prim);
5542 if (copy_from_user(&prim, optval, len))
5545 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
5549 if (!asoc->peer.primary_path)
5552 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
5553 asoc->peer.primary_path->af_specific->sockaddr_len);
5555 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
5556 (union sctp_addr *)&prim.ssp_addr);
5558 if (put_user(len, optlen))
5560 if (copy_to_user(optval, &prim, len))
5567 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5569 * Requests that the local endpoint set the specified Adaptation Layer
5570 * Indication parameter for all future INIT and INIT-ACK exchanges.
5572 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
5573 char __user *optval, int __user *optlen)
5575 struct sctp_setadaptation adaptation;
5577 if (len < sizeof(struct sctp_setadaptation))
5580 len = sizeof(struct sctp_setadaptation);
5582 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
5584 if (put_user(len, optlen))
5586 if (copy_to_user(optval, &adaptation, len))
5594 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5596 * Applications that wish to use the sendto() system call may wish to
5597 * specify a default set of parameters that would normally be supplied
5598 * through the inclusion of ancillary data. This socket option allows
5599 * such an application to set the default sctp_sndrcvinfo structure.
5602 * The application that wishes to use this socket option simply passes
5603 * in to this call the sctp_sndrcvinfo structure defined in Section
5604 * 5.2.2) The input parameters accepted by this call include
5605 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5606 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5607 * to this call if the caller is using the UDP model.
5609 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5611 static int sctp_getsockopt_default_send_param(struct sock *sk,
5612 int len, char __user *optval,
5615 struct sctp_sock *sp = sctp_sk(sk);
5616 struct sctp_association *asoc;
5617 struct sctp_sndrcvinfo info;
5619 if (len < sizeof(info))
5624 if (copy_from_user(&info, optval, len))
5627 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
5628 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
5631 info.sinfo_stream = asoc->default_stream;
5632 info.sinfo_flags = asoc->default_flags;
5633 info.sinfo_ppid = asoc->default_ppid;
5634 info.sinfo_context = asoc->default_context;
5635 info.sinfo_timetolive = asoc->default_timetolive;
5637 info.sinfo_stream = sp->default_stream;
5638 info.sinfo_flags = sp->default_flags;
5639 info.sinfo_ppid = sp->default_ppid;
5640 info.sinfo_context = sp->default_context;
5641 info.sinfo_timetolive = sp->default_timetolive;
5644 if (put_user(len, optlen))
5646 if (copy_to_user(optval, &info, len))
5652 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5653 * (SCTP_DEFAULT_SNDINFO)
5655 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
5656 char __user *optval,
5659 struct sctp_sock *sp = sctp_sk(sk);
5660 struct sctp_association *asoc;
5661 struct sctp_sndinfo info;
5663 if (len < sizeof(info))
5668 if (copy_from_user(&info, optval, len))
5671 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
5672 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
5675 info.snd_sid = asoc->default_stream;
5676 info.snd_flags = asoc->default_flags;
5677 info.snd_ppid = asoc->default_ppid;
5678 info.snd_context = asoc->default_context;
5680 info.snd_sid = sp->default_stream;
5681 info.snd_flags = sp->default_flags;
5682 info.snd_ppid = sp->default_ppid;
5683 info.snd_context = sp->default_context;
5686 if (put_user(len, optlen))
5688 if (copy_to_user(optval, &info, len))
5696 * 7.1.5 SCTP_NODELAY
5698 * Turn on/off any Nagle-like algorithm. This means that packets are
5699 * generally sent as soon as possible and no unnecessary delays are
5700 * introduced, at the cost of more packets in the network. Expects an
5701 * integer boolean flag.
5704 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5705 char __user *optval, int __user *optlen)
5709 if (len < sizeof(int))
5713 val = (sctp_sk(sk)->nodelay == 1);
5714 if (put_user(len, optlen))
5716 if (copy_to_user(optval, &val, len))
5723 * 7.1.1 SCTP_RTOINFO
5725 * The protocol parameters used to initialize and bound retransmission
5726 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5727 * and modify these parameters.
5728 * All parameters are time values, in milliseconds. A value of 0, when
5729 * modifying the parameters, indicates that the current value should not
5733 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5734 char __user *optval,
5735 int __user *optlen) {
5736 struct sctp_rtoinfo rtoinfo;
5737 struct sctp_association *asoc;
5739 if (len < sizeof (struct sctp_rtoinfo))
5742 len = sizeof(struct sctp_rtoinfo);
5744 if (copy_from_user(&rtoinfo, optval, len))
5747 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5749 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5752 /* Values corresponding to the specific association. */
5754 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5755 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5756 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5758 /* Values corresponding to the endpoint. */
5759 struct sctp_sock *sp = sctp_sk(sk);
5761 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5762 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5763 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5766 if (put_user(len, optlen))
5769 if (copy_to_user(optval, &rtoinfo, len))
5777 * 7.1.2 SCTP_ASSOCINFO
5779 * This option is used to tune the maximum retransmission attempts
5780 * of the association.
5781 * Returns an error if the new association retransmission value is
5782 * greater than the sum of the retransmission value of the peer.
5783 * See [SCTP] for more information.
5786 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5787 char __user *optval,
5791 struct sctp_assocparams assocparams;
5792 struct sctp_association *asoc;
5793 struct list_head *pos;
5796 if (len < sizeof (struct sctp_assocparams))
5799 len = sizeof(struct sctp_assocparams);
5801 if (copy_from_user(&assocparams, optval, len))
5804 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5806 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5809 /* Values correspoinding to the specific association */
5811 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5812 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5813 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5814 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5816 list_for_each(pos, &asoc->peer.transport_addr_list) {
5820 assocparams.sasoc_number_peer_destinations = cnt;
5822 /* Values corresponding to the endpoint */
5823 struct sctp_sock *sp = sctp_sk(sk);
5825 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5826 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5827 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5828 assocparams.sasoc_cookie_life =
5829 sp->assocparams.sasoc_cookie_life;
5830 assocparams.sasoc_number_peer_destinations =
5832 sasoc_number_peer_destinations;
5835 if (put_user(len, optlen))
5838 if (copy_to_user(optval, &assocparams, len))
5845 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5847 * This socket option is a boolean flag which turns on or off mapped V4
5848 * addresses. If this option is turned on and the socket is type
5849 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5850 * If this option is turned off, then no mapping will be done of V4
5851 * addresses and a user will receive both PF_INET6 and PF_INET type
5852 * addresses on the socket.
5854 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5855 char __user *optval, int __user *optlen)
5858 struct sctp_sock *sp = sctp_sk(sk);
5860 if (len < sizeof(int))
5865 if (put_user(len, optlen))
5867 if (copy_to_user(optval, &val, len))
5874 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5875 * (chapter and verse is quoted at sctp_setsockopt_context())
5877 static int sctp_getsockopt_context(struct sock *sk, int len,
5878 char __user *optval, int __user *optlen)
5880 struct sctp_assoc_value params;
5881 struct sctp_sock *sp;
5882 struct sctp_association *asoc;
5884 if (len < sizeof(struct sctp_assoc_value))
5887 len = sizeof(struct sctp_assoc_value);
5889 if (copy_from_user(¶ms, optval, len))
5894 if (params.assoc_id != 0) {
5895 asoc = sctp_id2assoc(sk, params.assoc_id);
5898 params.assoc_value = asoc->default_rcv_context;
5900 params.assoc_value = sp->default_rcv_context;
5903 if (put_user(len, optlen))
5905 if (copy_to_user(optval, ¶ms, len))
5912 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5913 * This option will get or set the maximum size to put in any outgoing
5914 * SCTP DATA chunk. If a message is larger than this size it will be
5915 * fragmented by SCTP into the specified size. Note that the underlying
5916 * SCTP implementation may fragment into smaller sized chunks when the
5917 * PMTU of the underlying association is smaller than the value set by
5918 * the user. The default value for this option is '0' which indicates
5919 * the user is NOT limiting fragmentation and only the PMTU will effect
5920 * SCTP's choice of DATA chunk size. Note also that values set larger
5921 * than the maximum size of an IP datagram will effectively let SCTP
5922 * control fragmentation (i.e. the same as setting this option to 0).
5924 * The following structure is used to access and modify this parameter:
5926 * struct sctp_assoc_value {
5927 * sctp_assoc_t assoc_id;
5928 * uint32_t assoc_value;
5931 * assoc_id: This parameter is ignored for one-to-one style sockets.
5932 * For one-to-many style sockets this parameter indicates which
5933 * association the user is performing an action upon. Note that if
5934 * this field's value is zero then the endpoints default value is
5935 * changed (effecting future associations only).
5936 * assoc_value: This parameter specifies the maximum size in bytes.
5938 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5939 char __user *optval, int __user *optlen)
5941 struct sctp_assoc_value params;
5942 struct sctp_association *asoc;
5944 if (len == sizeof(int)) {
5945 pr_warn_ratelimited(DEPRECATED
5947 "Use of int in maxseg socket option.\n"
5948 "Use struct sctp_assoc_value instead\n",
5949 current->comm, task_pid_nr(current));
5950 params.assoc_id = 0;
5951 } else if (len >= sizeof(struct sctp_assoc_value)) {
5952 len = sizeof(struct sctp_assoc_value);
5953 if (copy_from_user(¶ms, optval, sizeof(params)))
5958 asoc = sctp_id2assoc(sk, params.assoc_id);
5959 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5963 params.assoc_value = asoc->frag_point;
5965 params.assoc_value = sctp_sk(sk)->user_frag;
5967 if (put_user(len, optlen))
5969 if (len == sizeof(int)) {
5970 if (copy_to_user(optval, ¶ms.assoc_value, len))
5973 if (copy_to_user(optval, ¶ms, len))
5981 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5982 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5984 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5985 char __user *optval, int __user *optlen)
5989 if (len < sizeof(int))
5994 val = sctp_sk(sk)->frag_interleave;
5995 if (put_user(len, optlen))
5997 if (copy_to_user(optval, &val, len))
6004 * 7.1.25. Set or Get the sctp partial delivery point
6005 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
6007 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
6008 char __user *optval,
6013 if (len < sizeof(u32))
6018 val = sctp_sk(sk)->pd_point;
6019 if (put_user(len, optlen))
6021 if (copy_to_user(optval, &val, len))
6028 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
6029 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
6031 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
6032 char __user *optval,
6035 struct sctp_assoc_value params;
6036 struct sctp_sock *sp;
6037 struct sctp_association *asoc;
6039 if (len == sizeof(int)) {
6040 pr_warn_ratelimited(DEPRECATED
6042 "Use of int in max_burst socket option.\n"
6043 "Use struct sctp_assoc_value instead\n",
6044 current->comm, task_pid_nr(current));
6045 params.assoc_id = 0;
6046 } else if (len >= sizeof(struct sctp_assoc_value)) {
6047 len = sizeof(struct sctp_assoc_value);
6048 if (copy_from_user(¶ms, optval, len))
6055 if (params.assoc_id != 0) {
6056 asoc = sctp_id2assoc(sk, params.assoc_id);
6059 params.assoc_value = asoc->max_burst;
6061 params.assoc_value = sp->max_burst;
6063 if (len == sizeof(int)) {
6064 if (copy_to_user(optval, ¶ms.assoc_value, len))
6067 if (copy_to_user(optval, ¶ms, len))
6075 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
6076 char __user *optval, int __user *optlen)
6078 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6079 struct sctp_hmacalgo __user *p = (void __user *)optval;
6080 struct sctp_hmac_algo_param *hmacs;
6085 if (!ep->auth_enable)
6088 hmacs = ep->auth_hmacs_list;
6089 data_len = ntohs(hmacs->param_hdr.length) -
6090 sizeof(struct sctp_paramhdr);
6092 if (len < sizeof(struct sctp_hmacalgo) + data_len)
6095 len = sizeof(struct sctp_hmacalgo) + data_len;
6096 num_idents = data_len / sizeof(u16);
6098 if (put_user(len, optlen))
6100 if (put_user(num_idents, &p->shmac_num_idents))
6102 for (i = 0; i < num_idents; i++) {
6103 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
6105 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
6111 static int sctp_getsockopt_active_key(struct sock *sk, int len,
6112 char __user *optval, int __user *optlen)
6114 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6115 struct sctp_authkeyid val;
6116 struct sctp_association *asoc;
6118 if (!ep->auth_enable)
6121 if (len < sizeof(struct sctp_authkeyid))
6123 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
6126 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
6127 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
6131 val.scact_keynumber = asoc->active_key_id;
6133 val.scact_keynumber = ep->active_key_id;
6135 len = sizeof(struct sctp_authkeyid);
6136 if (put_user(len, optlen))
6138 if (copy_to_user(optval, &val, len))
6144 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
6145 char __user *optval, int __user *optlen)
6147 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6148 struct sctp_authchunks __user *p = (void __user *)optval;
6149 struct sctp_authchunks val;
6150 struct sctp_association *asoc;
6151 struct sctp_chunks_param *ch;
6155 if (!ep->auth_enable)
6158 if (len < sizeof(struct sctp_authchunks))
6161 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
6164 to = p->gauth_chunks;
6165 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6169 ch = asoc->peer.peer_chunks;
6173 /* See if the user provided enough room for all the data */
6174 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6175 if (len < num_chunks)
6178 if (copy_to_user(to, ch->chunks, num_chunks))
6181 len = sizeof(struct sctp_authchunks) + num_chunks;
6182 if (put_user(len, optlen))
6184 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6189 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
6190 char __user *optval, int __user *optlen)
6192 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6193 struct sctp_authchunks __user *p = (void __user *)optval;
6194 struct sctp_authchunks val;
6195 struct sctp_association *asoc;
6196 struct sctp_chunks_param *ch;
6200 if (!ep->auth_enable)
6203 if (len < sizeof(struct sctp_authchunks))
6206 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
6209 to = p->gauth_chunks;
6210 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6211 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
6215 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
6217 ch = ep->auth_chunk_list;
6222 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6223 if (len < sizeof(struct sctp_authchunks) + num_chunks)
6226 if (copy_to_user(to, ch->chunks, num_chunks))
6229 len = sizeof(struct sctp_authchunks) + num_chunks;
6230 if (put_user(len, optlen))
6232 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6239 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
6240 * This option gets the current number of associations that are attached
6241 * to a one-to-many style socket. The option value is an uint32_t.
6243 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
6244 char __user *optval, int __user *optlen)
6246 struct sctp_sock *sp = sctp_sk(sk);
6247 struct sctp_association *asoc;
6250 if (sctp_style(sk, TCP))
6253 if (len < sizeof(u32))
6258 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6262 if (put_user(len, optlen))
6264 if (copy_to_user(optval, &val, len))
6271 * 8.1.23 SCTP_AUTO_ASCONF
6272 * See the corresponding setsockopt entry as description
6274 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
6275 char __user *optval, int __user *optlen)
6279 if (len < sizeof(int))
6283 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
6285 if (put_user(len, optlen))
6287 if (copy_to_user(optval, &val, len))
6293 * 8.2.6. Get the Current Identifiers of Associations
6294 * (SCTP_GET_ASSOC_ID_LIST)
6296 * This option gets the current list of SCTP association identifiers of
6297 * the SCTP associations handled by a one-to-many style socket.
6299 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
6300 char __user *optval, int __user *optlen)
6302 struct sctp_sock *sp = sctp_sk(sk);
6303 struct sctp_association *asoc;
6304 struct sctp_assoc_ids *ids;
6307 if (sctp_style(sk, TCP))
6310 if (len < sizeof(struct sctp_assoc_ids))
6313 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6317 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
6320 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
6322 ids = kmalloc(len, GFP_USER | __GFP_NOWARN);
6326 ids->gaids_number_of_ids = num;
6328 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6329 ids->gaids_assoc_id[num++] = asoc->assoc_id;
6332 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
6342 * SCTP_PEER_ADDR_THLDS
6344 * This option allows us to fetch the partially failed threshold for one or all
6345 * transports in an association. See Section 6.1 of:
6346 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6348 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
6349 char __user *optval,
6353 struct sctp_paddrthlds val;
6354 struct sctp_transport *trans;
6355 struct sctp_association *asoc;
6357 if (len < sizeof(struct sctp_paddrthlds))
6359 len = sizeof(struct sctp_paddrthlds);
6360 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
6363 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
6364 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
6368 val.spt_pathpfthld = asoc->pf_retrans;
6369 val.spt_pathmaxrxt = asoc->pathmaxrxt;
6371 trans = sctp_addr_id2transport(sk, &val.spt_address,
6376 val.spt_pathmaxrxt = trans->pathmaxrxt;
6377 val.spt_pathpfthld = trans->pf_retrans;
6380 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
6387 * SCTP_GET_ASSOC_STATS
6389 * This option retrieves local per endpoint statistics. It is modeled
6390 * after OpenSolaris' implementation
6392 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
6393 char __user *optval,
6396 struct sctp_assoc_stats sas;
6397 struct sctp_association *asoc = NULL;
6399 /* User must provide at least the assoc id */
6400 if (len < sizeof(sctp_assoc_t))
6403 /* Allow the struct to grow and fill in as much as possible */
6404 len = min_t(size_t, len, sizeof(sas));
6406 if (copy_from_user(&sas, optval, len))
6409 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
6413 sas.sas_rtxchunks = asoc->stats.rtxchunks;
6414 sas.sas_gapcnt = asoc->stats.gapcnt;
6415 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
6416 sas.sas_osacks = asoc->stats.osacks;
6417 sas.sas_isacks = asoc->stats.isacks;
6418 sas.sas_octrlchunks = asoc->stats.octrlchunks;
6419 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
6420 sas.sas_oodchunks = asoc->stats.oodchunks;
6421 sas.sas_iodchunks = asoc->stats.iodchunks;
6422 sas.sas_ouodchunks = asoc->stats.ouodchunks;
6423 sas.sas_iuodchunks = asoc->stats.iuodchunks;
6424 sas.sas_idupchunks = asoc->stats.idupchunks;
6425 sas.sas_opackets = asoc->stats.opackets;
6426 sas.sas_ipackets = asoc->stats.ipackets;
6428 /* New high max rto observed, will return 0 if not a single
6429 * RTO update took place. obs_rto_ipaddr will be bogus
6432 sas.sas_maxrto = asoc->stats.max_obs_rto;
6433 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
6434 sizeof(struct sockaddr_storage));
6436 /* Mark beginning of a new observation period */
6437 asoc->stats.max_obs_rto = asoc->rto_min;
6439 if (put_user(len, optlen))
6442 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
6444 if (copy_to_user(optval, &sas, len))
6450 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
6451 char __user *optval,
6456 if (len < sizeof(int))
6460 if (sctp_sk(sk)->recvrcvinfo)
6462 if (put_user(len, optlen))
6464 if (copy_to_user(optval, &val, len))
6470 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
6471 char __user *optval,
6476 if (len < sizeof(int))
6480 if (sctp_sk(sk)->recvnxtinfo)
6482 if (put_user(len, optlen))
6484 if (copy_to_user(optval, &val, len))
6490 static int sctp_getsockopt_pr_supported(struct sock *sk, int len,
6491 char __user *optval,
6494 struct sctp_assoc_value params;
6495 struct sctp_association *asoc;
6496 int retval = -EFAULT;
6498 if (len < sizeof(params)) {
6503 len = sizeof(params);
6504 if (copy_from_user(¶ms, optval, len))
6507 asoc = sctp_id2assoc(sk, params.assoc_id);
6509 params.assoc_value = asoc->prsctp_enable;
6510 } else if (!params.assoc_id) {
6511 struct sctp_sock *sp = sctp_sk(sk);
6513 params.assoc_value = sp->ep->prsctp_enable;
6519 if (put_user(len, optlen))
6522 if (copy_to_user(optval, ¶ms, len))
6531 static int sctp_getsockopt_default_prinfo(struct sock *sk, int len,
6532 char __user *optval,
6535 struct sctp_default_prinfo info;
6536 struct sctp_association *asoc;
6537 int retval = -EFAULT;
6539 if (len < sizeof(info)) {
6545 if (copy_from_user(&info, optval, len))
6548 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
6550 info.pr_policy = SCTP_PR_POLICY(asoc->default_flags);
6551 info.pr_value = asoc->default_timetolive;
6552 } else if (!info.pr_assoc_id) {
6553 struct sctp_sock *sp = sctp_sk(sk);
6555 info.pr_policy = SCTP_PR_POLICY(sp->default_flags);
6556 info.pr_value = sp->default_timetolive;
6562 if (put_user(len, optlen))
6565 if (copy_to_user(optval, &info, len))
6574 static int sctp_getsockopt_pr_assocstatus(struct sock *sk, int len,
6575 char __user *optval,
6578 struct sctp_prstatus params;
6579 struct sctp_association *asoc;
6581 int retval = -EINVAL;
6583 if (len < sizeof(params))
6586 len = sizeof(params);
6587 if (copy_from_user(¶ms, optval, len)) {
6592 policy = params.sprstat_policy;
6593 if (policy & ~SCTP_PR_SCTP_MASK)
6596 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
6600 if (policy == SCTP_PR_SCTP_NONE) {
6601 params.sprstat_abandoned_unsent = 0;
6602 params.sprstat_abandoned_sent = 0;
6603 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
6604 params.sprstat_abandoned_unsent +=
6605 asoc->abandoned_unsent[policy];
6606 params.sprstat_abandoned_sent +=
6607 asoc->abandoned_sent[policy];
6610 params.sprstat_abandoned_unsent =
6611 asoc->abandoned_unsent[__SCTP_PR_INDEX(policy)];
6612 params.sprstat_abandoned_sent =
6613 asoc->abandoned_sent[__SCTP_PR_INDEX(policy)];
6616 if (put_user(len, optlen)) {
6621 if (copy_to_user(optval, ¶ms, len)) {
6632 static int sctp_getsockopt_pr_streamstatus(struct sock *sk, int len,
6633 char __user *optval,
6636 struct sctp_stream_out *streamout;
6637 struct sctp_association *asoc;
6638 struct sctp_prstatus params;
6639 int retval = -EINVAL;
6642 if (len < sizeof(params))
6645 len = sizeof(params);
6646 if (copy_from_user(¶ms, optval, len)) {
6651 policy = params.sprstat_policy;
6652 if (policy & ~SCTP_PR_SCTP_MASK)
6655 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
6656 if (!asoc || params.sprstat_sid >= asoc->stream.outcnt)
6659 streamout = &asoc->stream.out[params.sprstat_sid];
6660 if (policy == SCTP_PR_SCTP_NONE) {
6661 params.sprstat_abandoned_unsent = 0;
6662 params.sprstat_abandoned_sent = 0;
6663 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
6664 params.sprstat_abandoned_unsent +=
6665 streamout->abandoned_unsent[policy];
6666 params.sprstat_abandoned_sent +=
6667 streamout->abandoned_sent[policy];
6670 params.sprstat_abandoned_unsent =
6671 streamout->abandoned_unsent[__SCTP_PR_INDEX(policy)];
6672 params.sprstat_abandoned_sent =
6673 streamout->abandoned_sent[__SCTP_PR_INDEX(policy)];
6676 if (put_user(len, optlen) || copy_to_user(optval, ¶ms, len)) {
6687 static int sctp_getsockopt_reconfig_supported(struct sock *sk, int len,
6688 char __user *optval,
6691 struct sctp_assoc_value params;
6692 struct sctp_association *asoc;
6693 int retval = -EFAULT;
6695 if (len < sizeof(params)) {
6700 len = sizeof(params);
6701 if (copy_from_user(¶ms, optval, len))
6704 asoc = sctp_id2assoc(sk, params.assoc_id);
6706 params.assoc_value = asoc->reconf_enable;
6707 } else if (!params.assoc_id) {
6708 struct sctp_sock *sp = sctp_sk(sk);
6710 params.assoc_value = sp->ep->reconf_enable;
6716 if (put_user(len, optlen))
6719 if (copy_to_user(optval, ¶ms, len))
6728 static int sctp_getsockopt_enable_strreset(struct sock *sk, int len,
6729 char __user *optval,
6732 struct sctp_assoc_value params;
6733 struct sctp_association *asoc;
6734 int retval = -EFAULT;
6736 if (len < sizeof(params)) {
6741 len = sizeof(params);
6742 if (copy_from_user(¶ms, optval, len))
6745 asoc = sctp_id2assoc(sk, params.assoc_id);
6747 params.assoc_value = asoc->strreset_enable;
6748 } else if (!params.assoc_id) {
6749 struct sctp_sock *sp = sctp_sk(sk);
6751 params.assoc_value = sp->ep->strreset_enable;
6757 if (put_user(len, optlen))
6760 if (copy_to_user(optval, ¶ms, len))
6769 static int sctp_getsockopt(struct sock *sk, int level, int optname,
6770 char __user *optval, int __user *optlen)
6775 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
6777 /* I can hardly begin to describe how wrong this is. This is
6778 * so broken as to be worse than useless. The API draft
6779 * REALLY is NOT helpful here... I am not convinced that the
6780 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
6781 * are at all well-founded.
6783 if (level != SOL_SCTP) {
6784 struct sctp_af *af = sctp_sk(sk)->pf->af;
6786 retval = af->getsockopt(sk, level, optname, optval, optlen);
6790 if (get_user(len, optlen))
6800 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
6802 case SCTP_DISABLE_FRAGMENTS:
6803 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
6807 retval = sctp_getsockopt_events(sk, len, optval, optlen);
6809 case SCTP_AUTOCLOSE:
6810 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
6812 case SCTP_SOCKOPT_PEELOFF:
6813 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
6815 case SCTP_SOCKOPT_PEELOFF_FLAGS:
6816 retval = sctp_getsockopt_peeloff_flags(sk, len, optval, optlen);
6818 case SCTP_PEER_ADDR_PARAMS:
6819 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
6822 case SCTP_DELAYED_SACK:
6823 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
6827 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
6829 case SCTP_GET_PEER_ADDRS:
6830 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
6833 case SCTP_GET_LOCAL_ADDRS:
6834 retval = sctp_getsockopt_local_addrs(sk, len, optval,
6837 case SCTP_SOCKOPT_CONNECTX3:
6838 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
6840 case SCTP_DEFAULT_SEND_PARAM:
6841 retval = sctp_getsockopt_default_send_param(sk, len,
6844 case SCTP_DEFAULT_SNDINFO:
6845 retval = sctp_getsockopt_default_sndinfo(sk, len,
6848 case SCTP_PRIMARY_ADDR:
6849 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
6852 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
6855 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
6857 case SCTP_ASSOCINFO:
6858 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
6860 case SCTP_I_WANT_MAPPED_V4_ADDR:
6861 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
6864 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
6866 case SCTP_GET_PEER_ADDR_INFO:
6867 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
6870 case SCTP_ADAPTATION_LAYER:
6871 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
6875 retval = sctp_getsockopt_context(sk, len, optval, optlen);
6877 case SCTP_FRAGMENT_INTERLEAVE:
6878 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
6881 case SCTP_PARTIAL_DELIVERY_POINT:
6882 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
6885 case SCTP_MAX_BURST:
6886 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
6889 case SCTP_AUTH_CHUNK:
6890 case SCTP_AUTH_DELETE_KEY:
6891 retval = -EOPNOTSUPP;
6893 case SCTP_HMAC_IDENT:
6894 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
6896 case SCTP_AUTH_ACTIVE_KEY:
6897 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
6899 case SCTP_PEER_AUTH_CHUNKS:
6900 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
6903 case SCTP_LOCAL_AUTH_CHUNKS:
6904 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
6907 case SCTP_GET_ASSOC_NUMBER:
6908 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
6910 case SCTP_GET_ASSOC_ID_LIST:
6911 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
6913 case SCTP_AUTO_ASCONF:
6914 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
6916 case SCTP_PEER_ADDR_THLDS:
6917 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
6919 case SCTP_GET_ASSOC_STATS:
6920 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
6922 case SCTP_RECVRCVINFO:
6923 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
6925 case SCTP_RECVNXTINFO:
6926 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
6928 case SCTP_PR_SUPPORTED:
6929 retval = sctp_getsockopt_pr_supported(sk, len, optval, optlen);
6931 case SCTP_DEFAULT_PRINFO:
6932 retval = sctp_getsockopt_default_prinfo(sk, len, optval,
6935 case SCTP_PR_ASSOC_STATUS:
6936 retval = sctp_getsockopt_pr_assocstatus(sk, len, optval,
6939 case SCTP_PR_STREAM_STATUS:
6940 retval = sctp_getsockopt_pr_streamstatus(sk, len, optval,
6943 case SCTP_RECONFIG_SUPPORTED:
6944 retval = sctp_getsockopt_reconfig_supported(sk, len, optval,
6947 case SCTP_ENABLE_STREAM_RESET:
6948 retval = sctp_getsockopt_enable_strreset(sk, len, optval,
6952 retval = -ENOPROTOOPT;
6960 static int sctp_hash(struct sock *sk)
6966 static void sctp_unhash(struct sock *sk)
6971 /* Check if port is acceptable. Possibly find first available port.
6973 * The port hash table (contained in the 'global' SCTP protocol storage
6974 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
6975 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
6976 * list (the list number is the port number hashed out, so as you
6977 * would expect from a hash function, all the ports in a given list have
6978 * such a number that hashes out to the same list number; you were
6979 * expecting that, right?); so each list has a set of ports, with a
6980 * link to the socket (struct sock) that uses it, the port number and
6981 * a fastreuse flag (FIXME: NPI ipg).
6983 static struct sctp_bind_bucket *sctp_bucket_create(
6984 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
6986 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
6988 struct sctp_bind_hashbucket *head; /* hash list */
6989 struct sctp_bind_bucket *pp;
6990 unsigned short snum;
6993 snum = ntohs(addr->v4.sin_port);
6995 pr_debug("%s: begins, snum:%d\n", __func__, snum);
7000 /* Search for an available port. */
7001 int low, high, remaining, index;
7003 struct net *net = sock_net(sk);
7005 inet_get_local_port_range(net, &low, &high);
7006 remaining = (high - low) + 1;
7007 rover = prandom_u32() % remaining + low;
7011 if ((rover < low) || (rover > high))
7013 if (inet_is_local_reserved_port(net, rover))
7015 index = sctp_phashfn(sock_net(sk), rover);
7016 head = &sctp_port_hashtable[index];
7017 spin_lock(&head->lock);
7018 sctp_for_each_hentry(pp, &head->chain)
7019 if ((pp->port == rover) &&
7020 net_eq(sock_net(sk), pp->net))
7024 spin_unlock(&head->lock);
7025 } while (--remaining > 0);
7027 /* Exhausted local port range during search? */
7032 /* OK, here is the one we will use. HEAD (the port
7033 * hash table list entry) is non-NULL and we hold it's
7038 /* We are given an specific port number; we verify
7039 * that it is not being used. If it is used, we will
7040 * exahust the search in the hash list corresponding
7041 * to the port number (snum) - we detect that with the
7042 * port iterator, pp being NULL.
7044 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
7045 spin_lock(&head->lock);
7046 sctp_for_each_hentry(pp, &head->chain) {
7047 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
7054 if (!hlist_empty(&pp->owner)) {
7055 /* We had a port hash table hit - there is an
7056 * available port (pp != NULL) and it is being
7057 * used by other socket (pp->owner not empty); that other
7058 * socket is going to be sk2.
7060 int reuse = sk->sk_reuse;
7063 pr_debug("%s: found a possible match\n", __func__);
7065 if (pp->fastreuse && sk->sk_reuse &&
7066 sk->sk_state != SCTP_SS_LISTENING)
7069 /* Run through the list of sockets bound to the port
7070 * (pp->port) [via the pointers bind_next and
7071 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
7072 * we get the endpoint they describe and run through
7073 * the endpoint's list of IP (v4 or v6) addresses,
7074 * comparing each of the addresses with the address of
7075 * the socket sk. If we find a match, then that means
7076 * that this port/socket (sk) combination are already
7079 sk_for_each_bound(sk2, &pp->owner) {
7080 struct sctp_endpoint *ep2;
7081 ep2 = sctp_sk(sk2)->ep;
7084 (reuse && sk2->sk_reuse &&
7085 sk2->sk_state != SCTP_SS_LISTENING))
7088 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
7089 sctp_sk(sk2), sctp_sk(sk))) {
7095 pr_debug("%s: found a match\n", __func__);
7098 /* If there was a hash table miss, create a new port. */
7100 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
7103 /* In either case (hit or miss), make sure fastreuse is 1 only
7104 * if sk->sk_reuse is too (that is, if the caller requested
7105 * SO_REUSEADDR on this socket -sk-).
7107 if (hlist_empty(&pp->owner)) {
7108 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
7112 } else if (pp->fastreuse &&
7113 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
7116 /* We are set, so fill up all the data in the hash table
7117 * entry, tie the socket list information with the rest of the
7118 * sockets FIXME: Blurry, NPI (ipg).
7121 if (!sctp_sk(sk)->bind_hash) {
7122 inet_sk(sk)->inet_num = snum;
7123 sk_add_bind_node(sk, &pp->owner);
7124 sctp_sk(sk)->bind_hash = pp;
7129 spin_unlock(&head->lock);
7136 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
7137 * port is requested.
7139 static int sctp_get_port(struct sock *sk, unsigned short snum)
7141 union sctp_addr addr;
7142 struct sctp_af *af = sctp_sk(sk)->pf->af;
7144 /* Set up a dummy address struct from the sk. */
7145 af->from_sk(&addr, sk);
7146 addr.v4.sin_port = htons(snum);
7148 /* Note: sk->sk_num gets filled in if ephemeral port request. */
7149 return !!sctp_get_port_local(sk, &addr);
7153 * Move a socket to LISTENING state.
7155 static int sctp_listen_start(struct sock *sk, int backlog)
7157 struct sctp_sock *sp = sctp_sk(sk);
7158 struct sctp_endpoint *ep = sp->ep;
7159 struct crypto_shash *tfm = NULL;
7162 /* Allocate HMAC for generating cookie. */
7163 if (!sp->hmac && sp->sctp_hmac_alg) {
7164 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
7165 tfm = crypto_alloc_shash(alg, 0, 0);
7167 net_info_ratelimited("failed to load transform for %s: %ld\n",
7168 sp->sctp_hmac_alg, PTR_ERR(tfm));
7171 sctp_sk(sk)->hmac = tfm;
7175 * If a bind() or sctp_bindx() is not called prior to a listen()
7176 * call that allows new associations to be accepted, the system
7177 * picks an ephemeral port and will choose an address set equivalent
7178 * to binding with a wildcard address.
7180 * This is not currently spelled out in the SCTP sockets
7181 * extensions draft, but follows the practice as seen in TCP
7185 sk->sk_state = SCTP_SS_LISTENING;
7186 if (!ep->base.bind_addr.port) {
7187 if (sctp_autobind(sk))
7190 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
7191 sk->sk_state = SCTP_SS_CLOSED;
7196 sk->sk_max_ack_backlog = backlog;
7197 sctp_hash_endpoint(ep);
7202 * 4.1.3 / 5.1.3 listen()
7204 * By default, new associations are not accepted for UDP style sockets.
7205 * An application uses listen() to mark a socket as being able to
7206 * accept new associations.
7208 * On TCP style sockets, applications use listen() to ready the SCTP
7209 * endpoint for accepting inbound associations.
7211 * On both types of endpoints a backlog of '0' disables listening.
7213 * Move a socket to LISTENING state.
7215 int sctp_inet_listen(struct socket *sock, int backlog)
7217 struct sock *sk = sock->sk;
7218 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
7221 if (unlikely(backlog < 0))
7226 /* Peeled-off sockets are not allowed to listen(). */
7227 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
7230 if (sock->state != SS_UNCONNECTED)
7233 if (!sctp_sstate(sk, LISTENING) && !sctp_sstate(sk, CLOSED))
7236 /* If backlog is zero, disable listening. */
7238 if (sctp_sstate(sk, CLOSED))
7242 sctp_unhash_endpoint(ep);
7243 sk->sk_state = SCTP_SS_CLOSED;
7245 sctp_sk(sk)->bind_hash->fastreuse = 1;
7249 /* If we are already listening, just update the backlog */
7250 if (sctp_sstate(sk, LISTENING))
7251 sk->sk_max_ack_backlog = backlog;
7253 err = sctp_listen_start(sk, backlog);
7265 * This function is done by modeling the current datagram_poll() and the
7266 * tcp_poll(). Note that, based on these implementations, we don't
7267 * lock the socket in this function, even though it seems that,
7268 * ideally, locking or some other mechanisms can be used to ensure
7269 * the integrity of the counters (sndbuf and wmem_alloc) used
7270 * in this place. We assume that we don't need locks either until proven
7273 * Another thing to note is that we include the Async I/O support
7274 * here, again, by modeling the current TCP/UDP code. We don't have
7275 * a good way to test with it yet.
7277 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
7279 struct sock *sk = sock->sk;
7280 struct sctp_sock *sp = sctp_sk(sk);
7283 poll_wait(file, sk_sleep(sk), wait);
7285 sock_rps_record_flow(sk);
7287 /* A TCP-style listening socket becomes readable when the accept queue
7290 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
7291 return (!list_empty(&sp->ep->asocs)) ?
7292 (POLLIN | POLLRDNORM) : 0;
7296 /* Is there any exceptional events? */
7297 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
7299 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
7300 if (sk->sk_shutdown & RCV_SHUTDOWN)
7301 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
7302 if (sk->sk_shutdown == SHUTDOWN_MASK)
7305 /* Is it readable? Reconsider this code with TCP-style support. */
7306 if (!skb_queue_empty(&sk->sk_receive_queue))
7307 mask |= POLLIN | POLLRDNORM;
7309 /* The association is either gone or not ready. */
7310 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
7313 /* Is it writable? */
7314 if (sctp_writeable(sk)) {
7315 mask |= POLLOUT | POLLWRNORM;
7317 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
7319 * Since the socket is not locked, the buffer
7320 * might be made available after the writeable check and
7321 * before the bit is set. This could cause a lost I/O
7322 * signal. tcp_poll() has a race breaker for this race
7323 * condition. Based on their implementation, we put
7324 * in the following code to cover it as well.
7326 if (sctp_writeable(sk))
7327 mask |= POLLOUT | POLLWRNORM;
7332 /********************************************************************
7333 * 2nd Level Abstractions
7334 ********************************************************************/
7336 static struct sctp_bind_bucket *sctp_bucket_create(
7337 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
7339 struct sctp_bind_bucket *pp;
7341 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
7343 SCTP_DBG_OBJCNT_INC(bind_bucket);
7346 INIT_HLIST_HEAD(&pp->owner);
7348 hlist_add_head(&pp->node, &head->chain);
7353 /* Caller must hold hashbucket lock for this tb with local BH disabled */
7354 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
7356 if (pp && hlist_empty(&pp->owner)) {
7357 __hlist_del(&pp->node);
7358 kmem_cache_free(sctp_bucket_cachep, pp);
7359 SCTP_DBG_OBJCNT_DEC(bind_bucket);
7363 /* Release this socket's reference to a local port. */
7364 static inline void __sctp_put_port(struct sock *sk)
7366 struct sctp_bind_hashbucket *head =
7367 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
7368 inet_sk(sk)->inet_num)];
7369 struct sctp_bind_bucket *pp;
7371 spin_lock(&head->lock);
7372 pp = sctp_sk(sk)->bind_hash;
7373 __sk_del_bind_node(sk);
7374 sctp_sk(sk)->bind_hash = NULL;
7375 inet_sk(sk)->inet_num = 0;
7376 sctp_bucket_destroy(pp);
7377 spin_unlock(&head->lock);
7380 void sctp_put_port(struct sock *sk)
7383 __sctp_put_port(sk);
7388 * The system picks an ephemeral port and choose an address set equivalent
7389 * to binding with a wildcard address.
7390 * One of those addresses will be the primary address for the association.
7391 * This automatically enables the multihoming capability of SCTP.
7393 static int sctp_autobind(struct sock *sk)
7395 union sctp_addr autoaddr;
7399 /* Initialize a local sockaddr structure to INADDR_ANY. */
7400 af = sctp_sk(sk)->pf->af;
7402 port = htons(inet_sk(sk)->inet_num);
7403 af->inaddr_any(&autoaddr, port);
7405 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
7408 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
7411 * 4.2 The cmsghdr Structure *
7413 * When ancillary data is sent or received, any number of ancillary data
7414 * objects can be specified by the msg_control and msg_controllen members of
7415 * the msghdr structure, because each object is preceded by
7416 * a cmsghdr structure defining the object's length (the cmsg_len member).
7417 * Historically Berkeley-derived implementations have passed only one object
7418 * at a time, but this API allows multiple objects to be
7419 * passed in a single call to sendmsg() or recvmsg(). The following example
7420 * shows two ancillary data objects in a control buffer.
7422 * |<--------------------------- msg_controllen -------------------------->|
7425 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
7427 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
7430 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
7432 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
7435 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7436 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
7438 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
7440 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7447 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
7449 struct cmsghdr *cmsg;
7450 struct msghdr *my_msg = (struct msghdr *)msg;
7452 for_each_cmsghdr(cmsg, my_msg) {
7453 if (!CMSG_OK(my_msg, cmsg))
7456 /* Should we parse this header or ignore? */
7457 if (cmsg->cmsg_level != IPPROTO_SCTP)
7460 /* Strictly check lengths following example in SCM code. */
7461 switch (cmsg->cmsg_type) {
7463 /* SCTP Socket API Extension
7464 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
7466 * This cmsghdr structure provides information for
7467 * initializing new SCTP associations with sendmsg().
7468 * The SCTP_INITMSG socket option uses this same data
7469 * structure. This structure is not used for
7472 * cmsg_level cmsg_type cmsg_data[]
7473 * ------------ ------------ ----------------------
7474 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
7476 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
7479 cmsgs->init = CMSG_DATA(cmsg);
7483 /* SCTP Socket API Extension
7484 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
7486 * This cmsghdr structure specifies SCTP options for
7487 * sendmsg() and describes SCTP header information
7488 * about a received message through recvmsg().
7490 * cmsg_level cmsg_type cmsg_data[]
7491 * ------------ ------------ ----------------------
7492 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
7494 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
7497 cmsgs->srinfo = CMSG_DATA(cmsg);
7499 if (cmsgs->srinfo->sinfo_flags &
7500 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7501 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7502 SCTP_ABORT | SCTP_EOF))
7507 /* SCTP Socket API Extension
7508 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
7510 * This cmsghdr structure specifies SCTP options for
7511 * sendmsg(). This structure and SCTP_RCVINFO replaces
7512 * SCTP_SNDRCV which has been deprecated.
7514 * cmsg_level cmsg_type cmsg_data[]
7515 * ------------ ------------ ---------------------
7516 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
7518 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
7521 cmsgs->sinfo = CMSG_DATA(cmsg);
7523 if (cmsgs->sinfo->snd_flags &
7524 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7525 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7526 SCTP_ABORT | SCTP_EOF))
7538 * Wait for a packet..
7539 * Note: This function is the same function as in core/datagram.c
7540 * with a few modifications to make lksctp work.
7542 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
7547 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7549 /* Socket errors? */
7550 error = sock_error(sk);
7554 if (!skb_queue_empty(&sk->sk_receive_queue))
7557 /* Socket shut down? */
7558 if (sk->sk_shutdown & RCV_SHUTDOWN)
7561 /* Sequenced packets can come disconnected. If so we report the
7566 /* Is there a good reason to think that we may receive some data? */
7567 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
7570 /* Handle signals. */
7571 if (signal_pending(current))
7574 /* Let another process have a go. Since we are going to sleep
7575 * anyway. Note: This may cause odd behaviors if the message
7576 * does not fit in the user's buffer, but this seems to be the
7577 * only way to honor MSG_DONTWAIT realistically.
7580 *timeo_p = schedule_timeout(*timeo_p);
7584 finish_wait(sk_sleep(sk), &wait);
7588 error = sock_intr_errno(*timeo_p);
7591 finish_wait(sk_sleep(sk), &wait);
7596 /* Receive a datagram.
7597 * Note: This is pretty much the same routine as in core/datagram.c
7598 * with a few changes to make lksctp work.
7600 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
7601 int noblock, int *err)
7604 struct sk_buff *skb;
7607 timeo = sock_rcvtimeo(sk, noblock);
7609 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
7610 MAX_SCHEDULE_TIMEOUT);
7613 /* Again only user level code calls this function,
7614 * so nothing interrupt level
7615 * will suddenly eat the receive_queue.
7617 * Look at current nfs client by the way...
7618 * However, this function was correct in any case. 8)
7620 if (flags & MSG_PEEK) {
7621 skb = skb_peek(&sk->sk_receive_queue);
7623 refcount_inc(&skb->users);
7625 skb = __skb_dequeue(&sk->sk_receive_queue);
7631 /* Caller is allowed not to check sk->sk_err before calling. */
7632 error = sock_error(sk);
7636 if (sk->sk_shutdown & RCV_SHUTDOWN)
7639 if (sk_can_busy_loop(sk)) {
7640 sk_busy_loop(sk, noblock);
7642 if (!skb_queue_empty(&sk->sk_receive_queue))
7646 /* User doesn't want to wait. */
7650 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
7659 /* If sndbuf has changed, wake up per association sndbuf waiters. */
7660 static void __sctp_write_space(struct sctp_association *asoc)
7662 struct sock *sk = asoc->base.sk;
7664 if (sctp_wspace(asoc) <= 0)
7667 if (waitqueue_active(&asoc->wait))
7668 wake_up_interruptible(&asoc->wait);
7670 if (sctp_writeable(sk)) {
7671 struct socket_wq *wq;
7674 wq = rcu_dereference(sk->sk_wq);
7676 if (waitqueue_active(&wq->wait))
7677 wake_up_interruptible(&wq->wait);
7679 /* Note that we try to include the Async I/O support
7680 * here by modeling from the current TCP/UDP code.
7681 * We have not tested with it yet.
7683 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
7684 sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT);
7690 static void sctp_wake_up_waiters(struct sock *sk,
7691 struct sctp_association *asoc)
7693 struct sctp_association *tmp = asoc;
7695 /* We do accounting for the sndbuf space per association,
7696 * so we only need to wake our own association.
7698 if (asoc->ep->sndbuf_policy)
7699 return __sctp_write_space(asoc);
7701 /* If association goes down and is just flushing its
7702 * outq, then just normally notify others.
7704 if (asoc->base.dead)
7705 return sctp_write_space(sk);
7707 /* Accounting for the sndbuf space is per socket, so we
7708 * need to wake up others, try to be fair and in case of
7709 * other associations, let them have a go first instead
7710 * of just doing a sctp_write_space() call.
7712 * Note that we reach sctp_wake_up_waiters() only when
7713 * associations free up queued chunks, thus we are under
7714 * lock and the list of associations on a socket is
7715 * guaranteed not to change.
7717 for (tmp = list_next_entry(tmp, asocs); 1;
7718 tmp = list_next_entry(tmp, asocs)) {
7719 /* Manually skip the head element. */
7720 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
7722 /* Wake up association. */
7723 __sctp_write_space(tmp);
7724 /* We've reached the end. */
7730 /* Do accounting for the sndbuf space.
7731 * Decrement the used sndbuf space of the corresponding association by the
7732 * data size which was just transmitted(freed).
7734 static void sctp_wfree(struct sk_buff *skb)
7736 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
7737 struct sctp_association *asoc = chunk->asoc;
7738 struct sock *sk = asoc->base.sk;
7740 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
7741 sizeof(struct sk_buff) +
7742 sizeof(struct sctp_chunk);
7744 WARN_ON(refcount_sub_and_test(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc));
7747 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
7749 sk->sk_wmem_queued -= skb->truesize;
7750 sk_mem_uncharge(sk, skb->truesize);
7753 sctp_wake_up_waiters(sk, asoc);
7755 sctp_association_put(asoc);
7758 /* Do accounting for the receive space on the socket.
7759 * Accounting for the association is done in ulpevent.c
7760 * We set this as a destructor for the cloned data skbs so that
7761 * accounting is done at the correct time.
7763 void sctp_sock_rfree(struct sk_buff *skb)
7765 struct sock *sk = skb->sk;
7766 struct sctp_ulpevent *event = sctp_skb2event(skb);
7768 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
7771 * Mimic the behavior of sock_rfree
7773 sk_mem_uncharge(sk, event->rmem_len);
7777 /* Helper function to wait for space in the sndbuf. */
7778 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
7781 struct sock *sk = asoc->base.sk;
7783 long current_timeo = *timeo_p;
7786 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
7789 /* Increment the association's refcnt. */
7790 sctp_association_hold(asoc);
7792 /* Wait on the association specific sndbuf space. */
7794 prepare_to_wait_exclusive(&asoc->wait, &wait,
7795 TASK_INTERRUPTIBLE);
7798 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7801 if (signal_pending(current))
7802 goto do_interrupted;
7803 if (msg_len <= sctp_wspace(asoc))
7806 /* Let another process have a go. Since we are going
7810 current_timeo = schedule_timeout(current_timeo);
7813 *timeo_p = current_timeo;
7817 finish_wait(&asoc->wait, &wait);
7819 /* Release the association's refcnt. */
7820 sctp_association_put(asoc);
7829 err = sock_intr_errno(*timeo_p);
7837 void sctp_data_ready(struct sock *sk)
7839 struct socket_wq *wq;
7842 wq = rcu_dereference(sk->sk_wq);
7843 if (skwq_has_sleeper(wq))
7844 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
7845 POLLRDNORM | POLLRDBAND);
7846 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
7850 /* If socket sndbuf has changed, wake up all per association waiters. */
7851 void sctp_write_space(struct sock *sk)
7853 struct sctp_association *asoc;
7855 /* Wake up the tasks in each wait queue. */
7856 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
7857 __sctp_write_space(asoc);
7861 /* Is there any sndbuf space available on the socket?
7863 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
7864 * associations on the same socket. For a UDP-style socket with
7865 * multiple associations, it is possible for it to be "unwriteable"
7866 * prematurely. I assume that this is acceptable because
7867 * a premature "unwriteable" is better than an accidental "writeable" which
7868 * would cause an unwanted block under certain circumstances. For the 1-1
7869 * UDP-style sockets or TCP-style sockets, this code should work.
7872 static int sctp_writeable(struct sock *sk)
7876 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
7882 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
7883 * returns immediately with EINPROGRESS.
7885 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
7887 struct sock *sk = asoc->base.sk;
7889 long current_timeo = *timeo_p;
7892 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
7894 /* Increment the association's refcnt. */
7895 sctp_association_hold(asoc);
7898 prepare_to_wait_exclusive(&asoc->wait, &wait,
7899 TASK_INTERRUPTIBLE);
7902 if (sk->sk_shutdown & RCV_SHUTDOWN)
7904 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7907 if (signal_pending(current))
7908 goto do_interrupted;
7910 if (sctp_state(asoc, ESTABLISHED))
7913 /* Let another process have a go. Since we are going
7917 current_timeo = schedule_timeout(current_timeo);
7920 *timeo_p = current_timeo;
7924 finish_wait(&asoc->wait, &wait);
7926 /* Release the association's refcnt. */
7927 sctp_association_put(asoc);
7932 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
7935 err = -ECONNREFUSED;
7939 err = sock_intr_errno(*timeo_p);
7947 static int sctp_wait_for_accept(struct sock *sk, long timeo)
7949 struct sctp_endpoint *ep;
7953 ep = sctp_sk(sk)->ep;
7957 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
7958 TASK_INTERRUPTIBLE);
7960 if (list_empty(&ep->asocs)) {
7962 timeo = schedule_timeout(timeo);
7967 if (!sctp_sstate(sk, LISTENING))
7971 if (!list_empty(&ep->asocs))
7974 err = sock_intr_errno(timeo);
7975 if (signal_pending(current))
7983 finish_wait(sk_sleep(sk), &wait);
7988 static void sctp_wait_for_close(struct sock *sk, long timeout)
7993 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7994 if (list_empty(&sctp_sk(sk)->ep->asocs))
7997 timeout = schedule_timeout(timeout);
7999 } while (!signal_pending(current) && timeout);
8001 finish_wait(sk_sleep(sk), &wait);
8004 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
8006 struct sk_buff *frag;
8011 /* Don't forget the fragments. */
8012 skb_walk_frags(skb, frag)
8013 sctp_skb_set_owner_r_frag(frag, sk);
8016 sctp_skb_set_owner_r(skb, sk);
8019 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
8020 struct sctp_association *asoc)
8022 struct inet_sock *inet = inet_sk(sk);
8023 struct inet_sock *newinet;
8025 newsk->sk_type = sk->sk_type;
8026 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
8027 newsk->sk_flags = sk->sk_flags;
8028 newsk->sk_tsflags = sk->sk_tsflags;
8029 newsk->sk_no_check_tx = sk->sk_no_check_tx;
8030 newsk->sk_no_check_rx = sk->sk_no_check_rx;
8031 newsk->sk_reuse = sk->sk_reuse;
8033 newsk->sk_shutdown = sk->sk_shutdown;
8034 newsk->sk_destruct = sctp_destruct_sock;
8035 newsk->sk_family = sk->sk_family;
8036 newsk->sk_protocol = IPPROTO_SCTP;
8037 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
8038 newsk->sk_sndbuf = sk->sk_sndbuf;
8039 newsk->sk_rcvbuf = sk->sk_rcvbuf;
8040 newsk->sk_lingertime = sk->sk_lingertime;
8041 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
8042 newsk->sk_sndtimeo = sk->sk_sndtimeo;
8043 newsk->sk_rxhash = sk->sk_rxhash;
8045 newinet = inet_sk(newsk);
8047 /* Initialize sk's sport, dport, rcv_saddr and daddr for
8048 * getsockname() and getpeername()
8050 newinet->inet_sport = inet->inet_sport;
8051 newinet->inet_saddr = inet->inet_saddr;
8052 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
8053 newinet->inet_dport = htons(asoc->peer.port);
8054 newinet->pmtudisc = inet->pmtudisc;
8055 newinet->inet_id = asoc->next_tsn ^ jiffies;
8057 newinet->uc_ttl = inet->uc_ttl;
8058 newinet->mc_loop = 1;
8059 newinet->mc_ttl = 1;
8060 newinet->mc_index = 0;
8061 newinet->mc_list = NULL;
8063 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
8064 net_enable_timestamp();
8066 security_sk_clone(sk, newsk);
8069 static inline void sctp_copy_descendant(struct sock *sk_to,
8070 const struct sock *sk_from)
8072 int ancestor_size = sizeof(struct inet_sock) +
8073 sizeof(struct sctp_sock) -
8074 offsetof(struct sctp_sock, auto_asconf_list);
8076 if (sk_from->sk_family == PF_INET6)
8077 ancestor_size += sizeof(struct ipv6_pinfo);
8079 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
8082 /* Populate the fields of the newsk from the oldsk and migrate the assoc
8083 * and its messages to the newsk.
8085 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
8086 struct sctp_association *assoc,
8087 sctp_socket_type_t type)
8089 struct sctp_sock *oldsp = sctp_sk(oldsk);
8090 struct sctp_sock *newsp = sctp_sk(newsk);
8091 struct sctp_bind_bucket *pp; /* hash list port iterator */
8092 struct sctp_endpoint *newep = newsp->ep;
8093 struct sk_buff *skb, *tmp;
8094 struct sctp_ulpevent *event;
8095 struct sctp_bind_hashbucket *head;
8097 /* Migrate socket buffer sizes and all the socket level options to the
8100 newsk->sk_sndbuf = oldsk->sk_sndbuf;
8101 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
8102 /* Brute force copy old sctp opt. */
8103 sctp_copy_descendant(newsk, oldsk);
8105 /* Restore the ep value that was overwritten with the above structure
8111 /* Hook this new socket in to the bind_hash list. */
8112 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
8113 inet_sk(oldsk)->inet_num)];
8114 spin_lock_bh(&head->lock);
8115 pp = sctp_sk(oldsk)->bind_hash;
8116 sk_add_bind_node(newsk, &pp->owner);
8117 sctp_sk(newsk)->bind_hash = pp;
8118 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
8119 spin_unlock_bh(&head->lock);
8121 /* Copy the bind_addr list from the original endpoint to the new
8122 * endpoint so that we can handle restarts properly
8124 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
8125 &oldsp->ep->base.bind_addr, GFP_KERNEL);
8127 /* Move any messages in the old socket's receive queue that are for the
8128 * peeled off association to the new socket's receive queue.
8130 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
8131 event = sctp_skb2event(skb);
8132 if (event->asoc == assoc) {
8133 __skb_unlink(skb, &oldsk->sk_receive_queue);
8134 __skb_queue_tail(&newsk->sk_receive_queue, skb);
8135 sctp_skb_set_owner_r_frag(skb, newsk);
8139 /* Clean up any messages pending delivery due to partial
8140 * delivery. Three cases:
8141 * 1) No partial deliver; no work.
8142 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
8143 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
8145 skb_queue_head_init(&newsp->pd_lobby);
8146 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
8148 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
8149 struct sk_buff_head *queue;
8151 /* Decide which queue to move pd_lobby skbs to. */
8152 if (assoc->ulpq.pd_mode) {
8153 queue = &newsp->pd_lobby;
8155 queue = &newsk->sk_receive_queue;
8157 /* Walk through the pd_lobby, looking for skbs that
8158 * need moved to the new socket.
8160 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
8161 event = sctp_skb2event(skb);
8162 if (event->asoc == assoc) {
8163 __skb_unlink(skb, &oldsp->pd_lobby);
8164 __skb_queue_tail(queue, skb);
8165 sctp_skb_set_owner_r_frag(skb, newsk);
8169 /* Clear up any skbs waiting for the partial
8170 * delivery to finish.
8172 if (assoc->ulpq.pd_mode)
8173 sctp_clear_pd(oldsk, NULL);
8177 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
8178 sctp_skb_set_owner_r_frag(skb, newsk);
8180 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
8181 sctp_skb_set_owner_r_frag(skb, newsk);
8183 /* Set the type of socket to indicate that it is peeled off from the
8184 * original UDP-style socket or created with the accept() call on a
8185 * TCP-style socket..
8189 /* Mark the new socket "in-use" by the user so that any packets
8190 * that may arrive on the association after we've moved it are
8191 * queued to the backlog. This prevents a potential race between
8192 * backlog processing on the old socket and new-packet processing
8193 * on the new socket.
8195 * The caller has just allocated newsk so we can guarantee that other
8196 * paths won't try to lock it and then oldsk.
8198 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
8199 sctp_assoc_migrate(assoc, newsk);
8201 /* If the association on the newsk is already closed before accept()
8202 * is called, set RCV_SHUTDOWN flag.
8204 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP)) {
8205 newsk->sk_state = SCTP_SS_CLOSED;
8206 newsk->sk_shutdown |= RCV_SHUTDOWN;
8208 newsk->sk_state = SCTP_SS_ESTABLISHED;
8211 release_sock(newsk);
8215 /* This proto struct describes the ULP interface for SCTP. */
8216 struct proto sctp_prot = {
8218 .owner = THIS_MODULE,
8219 .close = sctp_close,
8220 .connect = sctp_connect,
8221 .disconnect = sctp_disconnect,
8222 .accept = sctp_accept,
8223 .ioctl = sctp_ioctl,
8224 .init = sctp_init_sock,
8225 .destroy = sctp_destroy_sock,
8226 .shutdown = sctp_shutdown,
8227 .setsockopt = sctp_setsockopt,
8228 .getsockopt = sctp_getsockopt,
8229 .sendmsg = sctp_sendmsg,
8230 .recvmsg = sctp_recvmsg,
8232 .backlog_rcv = sctp_backlog_rcv,
8234 .unhash = sctp_unhash,
8235 .get_port = sctp_get_port,
8236 .obj_size = sizeof(struct sctp_sock),
8237 .sysctl_mem = sysctl_sctp_mem,
8238 .sysctl_rmem = sysctl_sctp_rmem,
8239 .sysctl_wmem = sysctl_sctp_wmem,
8240 .memory_pressure = &sctp_memory_pressure,
8241 .enter_memory_pressure = sctp_enter_memory_pressure,
8242 .memory_allocated = &sctp_memory_allocated,
8243 .sockets_allocated = &sctp_sockets_allocated,
8246 #if IS_ENABLED(CONFIG_IPV6)
8248 #include <net/transp_v6.h>
8249 static void sctp_v6_destroy_sock(struct sock *sk)
8251 sctp_destroy_sock(sk);
8252 inet6_destroy_sock(sk);
8255 struct proto sctpv6_prot = {
8257 .owner = THIS_MODULE,
8258 .close = sctp_close,
8259 .connect = sctp_connect,
8260 .disconnect = sctp_disconnect,
8261 .accept = sctp_accept,
8262 .ioctl = sctp_ioctl,
8263 .init = sctp_init_sock,
8264 .destroy = sctp_v6_destroy_sock,
8265 .shutdown = sctp_shutdown,
8266 .setsockopt = sctp_setsockopt,
8267 .getsockopt = sctp_getsockopt,
8268 .sendmsg = sctp_sendmsg,
8269 .recvmsg = sctp_recvmsg,
8271 .backlog_rcv = sctp_backlog_rcv,
8273 .unhash = sctp_unhash,
8274 .get_port = sctp_get_port,
8275 .obj_size = sizeof(struct sctp6_sock),
8276 .sysctl_mem = sysctl_sctp_mem,
8277 .sysctl_rmem = sysctl_sctp_rmem,
8278 .sysctl_wmem = sysctl_sctp_wmem,
8279 .memory_pressure = &sctp_memory_pressure,
8280 .enter_memory_pressure = sctp_enter_memory_pressure,
8281 .memory_allocated = &sctp_memory_allocated,
8282 .sockets_allocated = &sctp_sockets_allocated,
8284 #endif /* IS_ENABLED(CONFIG_IPV6) */