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>
82 #include <net/sctp/stream_sched.h>
84 /* Forward declarations for internal helper functions. */
85 static int sctp_writeable(struct sock *sk);
86 static void sctp_wfree(struct sk_buff *skb);
87 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
89 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p);
90 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
91 static int sctp_wait_for_accept(struct sock *sk, long timeo);
92 static void sctp_wait_for_close(struct sock *sk, long timeo);
93 static void sctp_destruct_sock(struct sock *sk);
94 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
95 union sctp_addr *addr, int len);
96 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
97 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
98 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
99 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
100 static int sctp_send_asconf(struct sctp_association *asoc,
101 struct sctp_chunk *chunk);
102 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
103 static int sctp_autobind(struct sock *sk);
104 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
105 struct sctp_association *assoc,
106 enum sctp_socket_type type);
108 static unsigned long sctp_memory_pressure;
109 static atomic_long_t sctp_memory_allocated;
110 struct percpu_counter sctp_sockets_allocated;
112 static void sctp_enter_memory_pressure(struct sock *sk)
114 sctp_memory_pressure = 1;
118 /* Get the sndbuf space available at the time on the association. */
119 static inline int sctp_wspace(struct sctp_association *asoc)
123 if (asoc->ep->sndbuf_policy)
124 amt = asoc->sndbuf_used;
126 amt = sk_wmem_alloc_get(asoc->base.sk);
128 if (amt >= asoc->base.sk->sk_sndbuf) {
129 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
132 amt = sk_stream_wspace(asoc->base.sk);
137 amt = asoc->base.sk->sk_sndbuf - amt;
142 /* Increment the used sndbuf space count of the corresponding association by
143 * the size of the outgoing data chunk.
144 * Also, set the skb destructor for sndbuf accounting later.
146 * Since it is always 1-1 between chunk and skb, and also a new skb is always
147 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
148 * destructor in the data chunk skb for the purpose of the sndbuf space
151 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
153 struct sctp_association *asoc = chunk->asoc;
154 struct sock *sk = asoc->base.sk;
156 /* The sndbuf space is tracked per association. */
157 sctp_association_hold(asoc);
160 sctp_auth_shkey_hold(chunk->shkey);
162 skb_set_owner_w(chunk->skb, sk);
164 chunk->skb->destructor = sctp_wfree;
165 /* Save the chunk pointer in skb for sctp_wfree to use later. */
166 skb_shinfo(chunk->skb)->destructor_arg = chunk;
168 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
169 sizeof(struct sk_buff) +
170 sizeof(struct sctp_chunk);
172 refcount_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
173 sk->sk_wmem_queued += chunk->skb->truesize;
174 sk_mem_charge(sk, chunk->skb->truesize);
177 static void sctp_clear_owner_w(struct sctp_chunk *chunk)
179 skb_orphan(chunk->skb);
182 static void sctp_for_each_tx_datachunk(struct sctp_association *asoc,
183 void (*cb)(struct sctp_chunk *))
186 struct sctp_outq *q = &asoc->outqueue;
187 struct sctp_transport *t;
188 struct sctp_chunk *chunk;
190 list_for_each_entry(t, &asoc->peer.transport_addr_list, transports)
191 list_for_each_entry(chunk, &t->transmitted, transmitted_list)
194 list_for_each_entry(chunk, &q->retransmit, transmitted_list)
197 list_for_each_entry(chunk, &q->sacked, transmitted_list)
200 list_for_each_entry(chunk, &q->abandoned, transmitted_list)
203 list_for_each_entry(chunk, &q->out_chunk_list, list)
207 static void sctp_for_each_rx_skb(struct sctp_association *asoc, struct sock *sk,
208 void (*cb)(struct sk_buff *, struct sock *))
211 struct sk_buff *skb, *tmp;
213 sctp_skb_for_each(skb, &asoc->ulpq.lobby, tmp)
216 sctp_skb_for_each(skb, &asoc->ulpq.reasm, tmp)
219 sctp_skb_for_each(skb, &asoc->ulpq.reasm_uo, tmp)
223 /* Verify that this is a valid address. */
224 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
229 /* Verify basic sockaddr. */
230 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
234 /* Is this a valid SCTP address? */
235 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
238 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
244 /* Look up the association by its id. If this is not a UDP-style
245 * socket, the ID field is always ignored.
247 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
249 struct sctp_association *asoc = NULL;
251 /* If this is not a UDP-style socket, assoc id should be ignored. */
252 if (!sctp_style(sk, UDP)) {
253 /* Return NULL if the socket state is not ESTABLISHED. It
254 * could be a TCP-style listening socket or a socket which
255 * hasn't yet called connect() to establish an association.
257 if (!sctp_sstate(sk, ESTABLISHED) && !sctp_sstate(sk, CLOSING))
260 /* Get the first and the only association from the list. */
261 if (!list_empty(&sctp_sk(sk)->ep->asocs))
262 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
263 struct sctp_association, asocs);
267 /* Otherwise this is a UDP-style socket. */
268 if (!id || (id == (sctp_assoc_t)-1))
271 spin_lock_bh(&sctp_assocs_id_lock);
272 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
273 spin_unlock_bh(&sctp_assocs_id_lock);
275 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
281 /* Look up the transport from an address and an assoc id. If both address and
282 * id are specified, the associations matching the address and the id should be
285 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
286 struct sockaddr_storage *addr,
289 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
290 struct sctp_af *af = sctp_get_af_specific(addr->ss_family);
291 union sctp_addr *laddr = (union sctp_addr *)addr;
292 struct sctp_transport *transport;
294 if (!af || sctp_verify_addr(sk, laddr, af->sockaddr_len))
297 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
304 id_asoc = sctp_id2assoc(sk, id);
305 if (id_asoc && (id_asoc != addr_asoc))
308 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
309 (union sctp_addr *)addr);
314 /* API 3.1.2 bind() - UDP Style Syntax
315 * The syntax of bind() is,
317 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
319 * sd - the socket descriptor returned by socket().
320 * addr - the address structure (struct sockaddr_in or struct
321 * sockaddr_in6 [RFC 2553]),
322 * addr_len - the size of the address structure.
324 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
330 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
333 /* Disallow binding twice. */
334 if (!sctp_sk(sk)->ep->base.bind_addr.port)
335 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
345 static long sctp_get_port_local(struct sock *, union sctp_addr *);
347 /* Verify this is a valid sockaddr. */
348 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
349 union sctp_addr *addr, int len)
353 /* Check minimum size. */
354 if (len < sizeof (struct sockaddr))
357 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
360 if (addr->sa.sa_family == AF_INET6) {
361 if (len < SIN6_LEN_RFC2133)
363 /* V4 mapped address are really of AF_INET family */
364 if (ipv6_addr_v4mapped(&addr->v6.sin6_addr) &&
365 !opt->pf->af_supported(AF_INET, opt))
369 /* If we get this far, af is valid. */
370 af = sctp_get_af_specific(addr->sa.sa_family);
372 if (len < af->sockaddr_len)
378 /* Bind a local address either to an endpoint or to an association. */
379 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
381 struct net *net = sock_net(sk);
382 struct sctp_sock *sp = sctp_sk(sk);
383 struct sctp_endpoint *ep = sp->ep;
384 struct sctp_bind_addr *bp = &ep->base.bind_addr;
389 /* Common sockaddr verification. */
390 af = sctp_sockaddr_af(sp, addr, len);
392 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
393 __func__, sk, addr, len);
397 snum = ntohs(addr->v4.sin_port);
399 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
400 __func__, sk, &addr->sa, bp->port, snum, len);
402 /* PF specific bind() address verification. */
403 if (!sp->pf->bind_verify(sp, addr))
404 return -EADDRNOTAVAIL;
406 /* We must either be unbound, or bind to the same port.
407 * It's OK to allow 0 ports if we are already bound.
408 * We'll just inhert an already bound port in this case
413 else if (snum != bp->port) {
414 pr_debug("%s: new port %d doesn't match existing port "
415 "%d\n", __func__, snum, bp->port);
420 if (snum && snum < inet_prot_sock(net) &&
421 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
424 /* See if the address matches any of the addresses we may have
425 * already bound before checking against other endpoints.
427 if (sctp_bind_addr_match(bp, addr, sp))
430 /* Make sure we are allowed to bind here.
431 * The function sctp_get_port_local() does duplicate address
434 addr->v4.sin_port = htons(snum);
435 if ((ret = sctp_get_port_local(sk, addr))) {
439 /* Refresh ephemeral port. */
441 bp->port = inet_sk(sk)->inet_num;
443 /* Add the address to the bind address list.
444 * Use GFP_ATOMIC since BHs will be disabled.
446 ret = sctp_add_bind_addr(bp, addr, af->sockaddr_len,
447 SCTP_ADDR_SRC, GFP_ATOMIC);
449 /* Copy back into socket for getsockname() use. */
451 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
452 sp->pf->to_sk_saddr(addr, sk);
458 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
460 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
461 * at any one time. If a sender, after sending an ASCONF chunk, decides
462 * it needs to transfer another ASCONF Chunk, it MUST wait until the
463 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
464 * subsequent ASCONF. Note this restriction binds each side, so at any
465 * time two ASCONF may be in-transit on any given association (one sent
466 * from each endpoint).
468 static int sctp_send_asconf(struct sctp_association *asoc,
469 struct sctp_chunk *chunk)
471 struct net *net = sock_net(asoc->base.sk);
474 /* If there is an outstanding ASCONF chunk, queue it for later
477 if (asoc->addip_last_asconf) {
478 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
482 /* Hold the chunk until an ASCONF_ACK is received. */
483 sctp_chunk_hold(chunk);
484 retval = sctp_primitive_ASCONF(net, asoc, chunk);
486 sctp_chunk_free(chunk);
488 asoc->addip_last_asconf = chunk;
494 /* Add a list of addresses as bind addresses to local endpoint or
497 * Basically run through each address specified in the addrs/addrcnt
498 * array/length pair, determine if it is IPv6 or IPv4 and call
499 * sctp_do_bind() on it.
501 * If any of them fails, then the operation will be reversed and the
502 * ones that were added will be removed.
504 * Only sctp_setsockopt_bindx() is supposed to call this function.
506 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
511 struct sockaddr *sa_addr;
514 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
518 for (cnt = 0; cnt < addrcnt; cnt++) {
519 /* The list may contain either IPv4 or IPv6 address;
520 * determine the address length for walking thru the list.
523 af = sctp_get_af_specific(sa_addr->sa_family);
529 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
532 addr_buf += af->sockaddr_len;
536 /* Failed. Cleanup the ones that have been added */
538 sctp_bindx_rem(sk, addrs, cnt);
546 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
547 * associations that are part of the endpoint indicating that a list of local
548 * addresses are added to the endpoint.
550 * If any of the addresses is already in the bind address list of the
551 * association, we do not send the chunk for that association. But it will not
552 * affect other associations.
554 * Only sctp_setsockopt_bindx() is supposed to call this function.
556 static int sctp_send_asconf_add_ip(struct sock *sk,
557 struct sockaddr *addrs,
560 struct net *net = sock_net(sk);
561 struct sctp_sock *sp;
562 struct sctp_endpoint *ep;
563 struct sctp_association *asoc;
564 struct sctp_bind_addr *bp;
565 struct sctp_chunk *chunk;
566 struct sctp_sockaddr_entry *laddr;
567 union sctp_addr *addr;
568 union sctp_addr saveaddr;
575 if (!net->sctp.addip_enable)
581 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
582 __func__, sk, addrs, addrcnt);
584 list_for_each_entry(asoc, &ep->asocs, asocs) {
585 if (!asoc->peer.asconf_capable)
588 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
591 if (!sctp_state(asoc, ESTABLISHED))
594 /* Check if any address in the packed array of addresses is
595 * in the bind address list of the association. If so,
596 * do not send the asconf chunk to its peer, but continue with
597 * other associations.
600 for (i = 0; i < addrcnt; i++) {
602 af = sctp_get_af_specific(addr->v4.sin_family);
608 if (sctp_assoc_lookup_laddr(asoc, addr))
611 addr_buf += af->sockaddr_len;
616 /* Use the first valid address in bind addr list of
617 * association as Address Parameter of ASCONF CHUNK.
619 bp = &asoc->base.bind_addr;
620 p = bp->address_list.next;
621 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
622 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
623 addrcnt, SCTP_PARAM_ADD_IP);
629 /* Add the new addresses to the bind address list with
630 * use_as_src set to 0.
633 for (i = 0; i < addrcnt; i++) {
635 af = sctp_get_af_specific(addr->v4.sin_family);
636 memcpy(&saveaddr, addr, af->sockaddr_len);
637 retval = sctp_add_bind_addr(bp, &saveaddr,
639 SCTP_ADDR_NEW, GFP_ATOMIC);
640 addr_buf += af->sockaddr_len;
642 if (asoc->src_out_of_asoc_ok) {
643 struct sctp_transport *trans;
645 list_for_each_entry(trans,
646 &asoc->peer.transport_addr_list, transports) {
647 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
648 2*asoc->pathmtu, 4380));
649 trans->ssthresh = asoc->peer.i.a_rwnd;
650 trans->rto = asoc->rto_initial;
651 sctp_max_rto(asoc, trans);
652 trans->rtt = trans->srtt = trans->rttvar = 0;
653 /* Clear the source and route cache */
654 sctp_transport_route(trans, NULL,
655 sctp_sk(asoc->base.sk));
658 retval = sctp_send_asconf(asoc, chunk);
665 /* Remove a list of addresses from bind addresses list. Do not remove the
668 * Basically run through each address specified in the addrs/addrcnt
669 * array/length pair, determine if it is IPv6 or IPv4 and call
670 * sctp_del_bind() on it.
672 * If any of them fails, then the operation will be reversed and the
673 * ones that were removed will be added back.
675 * At least one address has to be left; if only one address is
676 * available, the operation will return -EBUSY.
678 * Only sctp_setsockopt_bindx() is supposed to call this function.
680 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
682 struct sctp_sock *sp = sctp_sk(sk);
683 struct sctp_endpoint *ep = sp->ep;
685 struct sctp_bind_addr *bp = &ep->base.bind_addr;
688 union sctp_addr *sa_addr;
691 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
692 __func__, sk, addrs, addrcnt);
695 for (cnt = 0; cnt < addrcnt; cnt++) {
696 /* If the bind address list is empty or if there is only one
697 * bind address, there is nothing more to be removed (we need
698 * at least one address here).
700 if (list_empty(&bp->address_list) ||
701 (sctp_list_single_entry(&bp->address_list))) {
707 af = sctp_get_af_specific(sa_addr->sa.sa_family);
713 if (!af->addr_valid(sa_addr, sp, NULL)) {
714 retval = -EADDRNOTAVAIL;
718 if (sa_addr->v4.sin_port &&
719 sa_addr->v4.sin_port != htons(bp->port)) {
724 if (!sa_addr->v4.sin_port)
725 sa_addr->v4.sin_port = htons(bp->port);
727 /* FIXME - There is probably a need to check if sk->sk_saddr and
728 * sk->sk_rcv_addr are currently set to one of the addresses to
729 * be removed. This is something which needs to be looked into
730 * when we are fixing the outstanding issues with multi-homing
731 * socket routing and failover schemes. Refer to comments in
732 * sctp_do_bind(). -daisy
734 retval = sctp_del_bind_addr(bp, sa_addr);
736 addr_buf += af->sockaddr_len;
739 /* Failed. Add the ones that has been removed back */
741 sctp_bindx_add(sk, addrs, cnt);
749 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
750 * the associations that are part of the endpoint indicating that a list of
751 * local addresses are removed from the endpoint.
753 * If any of the addresses is already in the bind address list of the
754 * association, we do not send the chunk for that association. But it will not
755 * affect other associations.
757 * Only sctp_setsockopt_bindx() is supposed to call this function.
759 static int sctp_send_asconf_del_ip(struct sock *sk,
760 struct sockaddr *addrs,
763 struct net *net = sock_net(sk);
764 struct sctp_sock *sp;
765 struct sctp_endpoint *ep;
766 struct sctp_association *asoc;
767 struct sctp_transport *transport;
768 struct sctp_bind_addr *bp;
769 struct sctp_chunk *chunk;
770 union sctp_addr *laddr;
773 struct sctp_sockaddr_entry *saddr;
779 if (!net->sctp.addip_enable)
785 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
786 __func__, sk, addrs, addrcnt);
788 list_for_each_entry(asoc, &ep->asocs, asocs) {
790 if (!asoc->peer.asconf_capable)
793 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
796 if (!sctp_state(asoc, ESTABLISHED))
799 /* Check if any address in the packed array of addresses is
800 * not present in the bind address list of the association.
801 * If so, do not send the asconf chunk to its peer, but
802 * continue with other associations.
805 for (i = 0; i < addrcnt; i++) {
807 af = sctp_get_af_specific(laddr->v4.sin_family);
813 if (!sctp_assoc_lookup_laddr(asoc, laddr))
816 addr_buf += af->sockaddr_len;
821 /* Find one address in the association's bind address list
822 * that is not in the packed array of addresses. This is to
823 * make sure that we do not delete all the addresses in the
826 bp = &asoc->base.bind_addr;
827 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
829 if ((laddr == NULL) && (addrcnt == 1)) {
830 if (asoc->asconf_addr_del_pending)
832 asoc->asconf_addr_del_pending =
833 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
834 if (asoc->asconf_addr_del_pending == NULL) {
838 asoc->asconf_addr_del_pending->sa.sa_family =
840 asoc->asconf_addr_del_pending->v4.sin_port =
842 if (addrs->sa_family == AF_INET) {
843 struct sockaddr_in *sin;
845 sin = (struct sockaddr_in *)addrs;
846 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
847 } else if (addrs->sa_family == AF_INET6) {
848 struct sockaddr_in6 *sin6;
850 sin6 = (struct sockaddr_in6 *)addrs;
851 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
854 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
855 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
856 asoc->asconf_addr_del_pending);
858 asoc->src_out_of_asoc_ok = 1;
866 /* We do not need RCU protection throughout this loop
867 * because this is done under a socket lock from the
870 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
878 /* Reset use_as_src flag for the addresses in the bind address
879 * list that are to be deleted.
882 for (i = 0; i < addrcnt; i++) {
884 af = sctp_get_af_specific(laddr->v4.sin_family);
885 list_for_each_entry(saddr, &bp->address_list, list) {
886 if (sctp_cmp_addr_exact(&saddr->a, laddr))
887 saddr->state = SCTP_ADDR_DEL;
889 addr_buf += af->sockaddr_len;
892 /* Update the route and saddr entries for all the transports
893 * as some of the addresses in the bind address list are
894 * about to be deleted and cannot be used as source addresses.
896 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
898 sctp_transport_route(transport, NULL,
899 sctp_sk(asoc->base.sk));
903 /* We don't need to transmit ASCONF */
905 retval = sctp_send_asconf(asoc, chunk);
911 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
912 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
914 struct sock *sk = sctp_opt2sk(sp);
915 union sctp_addr *addr;
918 /* It is safe to write port space in caller. */
920 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
921 af = sctp_get_af_specific(addr->sa.sa_family);
924 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
927 if (addrw->state == SCTP_ADDR_NEW)
928 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
930 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
933 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
936 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
939 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
940 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
943 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
944 * Section 3.1.2 for this usage.
946 * addrs is a pointer to an array of one or more socket addresses. Each
947 * address is contained in its appropriate structure (i.e. struct
948 * sockaddr_in or struct sockaddr_in6) the family of the address type
949 * must be used to distinguish the address length (note that this
950 * representation is termed a "packed array" of addresses). The caller
951 * specifies the number of addresses in the array with addrcnt.
953 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
954 * -1, and sets errno to the appropriate error code.
956 * For SCTP, the port given in each socket address must be the same, or
957 * sctp_bindx() will fail, setting errno to EINVAL.
959 * The flags parameter is formed from the bitwise OR of zero or more of
960 * the following currently defined flags:
962 * SCTP_BINDX_ADD_ADDR
964 * SCTP_BINDX_REM_ADDR
966 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
967 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
968 * addresses from the association. The two flags are mutually exclusive;
969 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
970 * not remove all addresses from an association; sctp_bindx() will
971 * reject such an attempt with EINVAL.
973 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
974 * additional addresses with an endpoint after calling bind(). Or use
975 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
976 * socket is associated with so that no new association accepted will be
977 * associated with those addresses. If the endpoint supports dynamic
978 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
979 * endpoint to send the appropriate message to the peer to change the
980 * peers address lists.
982 * Adding and removing addresses from a connected association is
983 * optional functionality. Implementations that do not support this
984 * functionality should return EOPNOTSUPP.
986 * Basically do nothing but copying the addresses from user to kernel
987 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
988 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
991 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
994 * sk The sk of the socket
995 * addrs The pointer to the addresses in user land
996 * addrssize Size of the addrs buffer
997 * op Operation to perform (add or remove, see the flags of
1000 * Returns 0 if ok, <0 errno code on error.
1002 static int sctp_setsockopt_bindx(struct sock *sk,
1003 struct sockaddr __user *addrs,
1004 int addrs_size, int op)
1006 struct sockaddr *kaddrs;
1010 struct sockaddr *sa_addr;
1014 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
1015 __func__, sk, addrs, addrs_size, op);
1017 if (unlikely(addrs_size <= 0))
1020 kaddrs = vmemdup_user(addrs, addrs_size);
1021 if (unlikely(IS_ERR(kaddrs)))
1022 return PTR_ERR(kaddrs);
1024 /* Walk through the addrs buffer and count the number of addresses. */
1026 while (walk_size < addrs_size) {
1027 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1033 af = sctp_get_af_specific(sa_addr->sa_family);
1035 /* If the address family is not supported or if this address
1036 * causes the address buffer to overflow return EINVAL.
1038 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1043 addr_buf += af->sockaddr_len;
1044 walk_size += af->sockaddr_len;
1049 case SCTP_BINDX_ADD_ADDR:
1050 /* Allow security module to validate bindx addresses. */
1051 err = security_sctp_bind_connect(sk, SCTP_SOCKOPT_BINDX_ADD,
1052 (struct sockaddr *)kaddrs,
1056 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1059 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1062 case SCTP_BINDX_REM_ADDR:
1063 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1066 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1080 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1082 * Common routine for handling connect() and sctp_connectx().
1083 * Connect will come in with just a single address.
1085 static int __sctp_connect(struct sock *sk,
1086 struct sockaddr *kaddrs,
1088 sctp_assoc_t *assoc_id)
1090 struct net *net = sock_net(sk);
1091 struct sctp_sock *sp;
1092 struct sctp_endpoint *ep;
1093 struct sctp_association *asoc = NULL;
1094 struct sctp_association *asoc2;
1095 struct sctp_transport *transport;
1097 enum sctp_scope scope;
1102 union sctp_addr *sa_addr = NULL;
1104 unsigned short port;
1105 unsigned int f_flags = 0;
1110 /* connect() cannot be done on a socket that is already in ESTABLISHED
1111 * state - UDP-style peeled off socket or a TCP-style socket that
1112 * is already connected.
1113 * It cannot be done even on a TCP-style listening socket.
1115 if (sctp_sstate(sk, ESTABLISHED) || sctp_sstate(sk, CLOSING) ||
1116 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1121 /* Walk through the addrs buffer and count the number of addresses. */
1123 while (walk_size < addrs_size) {
1126 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1132 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1134 /* If the address family is not supported or if this address
1135 * causes the address buffer to overflow return EINVAL.
1137 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1142 port = ntohs(sa_addr->v4.sin_port);
1144 /* Save current address so we can work with it */
1145 memcpy(&to, sa_addr, af->sockaddr_len);
1147 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1151 /* Make sure the destination port is correctly set
1154 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1159 /* Check if there already is a matching association on the
1160 * endpoint (other than the one created here).
1162 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1163 if (asoc2 && asoc2 != asoc) {
1164 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1171 /* If we could not find a matching association on the endpoint,
1172 * make sure that there is no peeled-off association matching
1173 * the peer address even on another socket.
1175 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1176 err = -EADDRNOTAVAIL;
1181 /* If a bind() or sctp_bindx() is not called prior to
1182 * an sctp_connectx() call, the system picks an
1183 * ephemeral port and will choose an address set
1184 * equivalent to binding with a wildcard address.
1186 if (!ep->base.bind_addr.port) {
1187 if (sctp_autobind(sk)) {
1193 * If an unprivileged user inherits a 1-many
1194 * style socket with open associations on a
1195 * privileged port, it MAY be permitted to
1196 * accept new associations, but it SHOULD NOT
1197 * be permitted to open new associations.
1199 if (ep->base.bind_addr.port <
1200 inet_prot_sock(net) &&
1201 !ns_capable(net->user_ns,
1202 CAP_NET_BIND_SERVICE)) {
1208 scope = sctp_scope(&to);
1209 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1215 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1223 /* Prime the peer's transport structures. */
1224 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1232 addr_buf += af->sockaddr_len;
1233 walk_size += af->sockaddr_len;
1236 /* In case the user of sctp_connectx() wants an association
1237 * id back, assign one now.
1240 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1245 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1250 /* Initialize sk's dport and daddr for getpeername() */
1251 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1252 sp->pf->to_sk_daddr(sa_addr, sk);
1255 /* in-kernel sockets don't generally have a file allocated to them
1256 * if all they do is call sock_create_kern().
1258 if (sk->sk_socket->file)
1259 f_flags = sk->sk_socket->file->f_flags;
1261 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1264 *assoc_id = asoc->assoc_id;
1266 err = sctp_wait_for_connect(asoc, &timeo);
1267 /* Note: the asoc may be freed after the return of
1268 * sctp_wait_for_connect.
1271 /* Don't free association on exit. */
1275 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1276 __func__, asoc, kaddrs, err);
1279 /* sctp_primitive_ASSOCIATE may have added this association
1280 * To the hash table, try to unhash it, just in case, its a noop
1281 * if it wasn't hashed so we're safe
1283 sctp_association_free(asoc);
1288 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1291 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1292 * sctp_assoc_t *asoc);
1294 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1295 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1296 * or IPv6 addresses.
1298 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1299 * Section 3.1.2 for this usage.
1301 * addrs is a pointer to an array of one or more socket addresses. Each
1302 * address is contained in its appropriate structure (i.e. struct
1303 * sockaddr_in or struct sockaddr_in6) the family of the address type
1304 * must be used to distengish the address length (note that this
1305 * representation is termed a "packed array" of addresses). The caller
1306 * specifies the number of addresses in the array with addrcnt.
1308 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1309 * the association id of the new association. On failure, sctp_connectx()
1310 * returns -1, and sets errno to the appropriate error code. The assoc_id
1311 * is not touched by the kernel.
1313 * For SCTP, the port given in each socket address must be the same, or
1314 * sctp_connectx() will fail, setting errno to EINVAL.
1316 * An application can use sctp_connectx to initiate an association with
1317 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1318 * allows a caller to specify multiple addresses at which a peer can be
1319 * reached. The way the SCTP stack uses the list of addresses to set up
1320 * the association is implementation dependent. This function only
1321 * specifies that the stack will try to make use of all the addresses in
1322 * the list when needed.
1324 * Note that the list of addresses passed in is only used for setting up
1325 * the association. It does not necessarily equal the set of addresses
1326 * the peer uses for the resulting association. If the caller wants to
1327 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1328 * retrieve them after the association has been set up.
1330 * Basically do nothing but copying the addresses from user to kernel
1331 * land and invoking either sctp_connectx(). This is used for tunneling
1332 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1334 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1337 * sk The sk of the socket
1338 * addrs The pointer to the addresses in user land
1339 * addrssize Size of the addrs buffer
1341 * Returns >=0 if ok, <0 errno code on error.
1343 static int __sctp_setsockopt_connectx(struct sock *sk,
1344 struct sockaddr __user *addrs,
1346 sctp_assoc_t *assoc_id)
1348 struct sockaddr *kaddrs;
1351 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1352 __func__, sk, addrs, addrs_size);
1354 if (unlikely(addrs_size <= 0))
1357 kaddrs = vmemdup_user(addrs, addrs_size);
1358 if (unlikely(IS_ERR(kaddrs)))
1359 return PTR_ERR(kaddrs);
1361 /* Allow security module to validate connectx addresses. */
1362 err = security_sctp_bind_connect(sk, SCTP_SOCKOPT_CONNECTX,
1363 (struct sockaddr *)kaddrs,
1368 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1377 * This is an older interface. It's kept for backward compatibility
1378 * to the option that doesn't provide association id.
1380 static int sctp_setsockopt_connectx_old(struct sock *sk,
1381 struct sockaddr __user *addrs,
1384 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1388 * New interface for the API. The since the API is done with a socket
1389 * option, to make it simple we feed back the association id is as a return
1390 * indication to the call. Error is always negative and association id is
1393 static int sctp_setsockopt_connectx(struct sock *sk,
1394 struct sockaddr __user *addrs,
1397 sctp_assoc_t assoc_id = 0;
1400 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1409 * New (hopefully final) interface for the API.
1410 * We use the sctp_getaddrs_old structure so that use-space library
1411 * can avoid any unnecessary allocations. The only different part
1412 * is that we store the actual length of the address buffer into the
1413 * addrs_num structure member. That way we can re-use the existing
1416 #ifdef CONFIG_COMPAT
1417 struct compat_sctp_getaddrs_old {
1418 sctp_assoc_t assoc_id;
1420 compat_uptr_t addrs; /* struct sockaddr * */
1424 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1425 char __user *optval,
1428 struct sctp_getaddrs_old param;
1429 sctp_assoc_t assoc_id = 0;
1432 #ifdef CONFIG_COMPAT
1433 if (in_compat_syscall()) {
1434 struct compat_sctp_getaddrs_old param32;
1436 if (len < sizeof(param32))
1438 if (copy_from_user(¶m32, optval, sizeof(param32)))
1441 param.assoc_id = param32.assoc_id;
1442 param.addr_num = param32.addr_num;
1443 param.addrs = compat_ptr(param32.addrs);
1447 if (len < sizeof(param))
1449 if (copy_from_user(¶m, optval, sizeof(param)))
1453 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1454 param.addrs, param.addr_num,
1456 if (err == 0 || err == -EINPROGRESS) {
1457 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1459 if (put_user(sizeof(assoc_id), optlen))
1466 /* API 3.1.4 close() - UDP Style Syntax
1467 * Applications use close() to perform graceful shutdown (as described in
1468 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1469 * by a UDP-style socket.
1473 * ret = close(int sd);
1475 * sd - the socket descriptor of the associations to be closed.
1477 * To gracefully shutdown a specific association represented by the
1478 * UDP-style socket, an application should use the sendmsg() call,
1479 * passing no user data, but including the appropriate flag in the
1480 * ancillary data (see Section xxxx).
1482 * If sd in the close() call is a branched-off socket representing only
1483 * one association, the shutdown is performed on that association only.
1485 * 4.1.6 close() - TCP Style Syntax
1487 * Applications use close() to gracefully close down an association.
1491 * int close(int sd);
1493 * sd - the socket descriptor of the association to be closed.
1495 * After an application calls close() on a socket descriptor, no further
1496 * socket operations will succeed on that descriptor.
1498 * API 7.1.4 SO_LINGER
1500 * An application using the TCP-style socket can use this option to
1501 * perform the SCTP ABORT primitive. The linger option structure is:
1504 * int l_onoff; // option on/off
1505 * int l_linger; // linger time
1508 * To enable the option, set l_onoff to 1. If the l_linger value is set
1509 * to 0, calling close() is the same as the ABORT primitive. If the
1510 * value is set to a negative value, the setsockopt() call will return
1511 * an error. If the value is set to a positive value linger_time, the
1512 * close() can be blocked for at most linger_time ms. If the graceful
1513 * shutdown phase does not finish during this period, close() will
1514 * return but the graceful shutdown phase continues in the system.
1516 static void sctp_close(struct sock *sk, long timeout)
1518 struct net *net = sock_net(sk);
1519 struct sctp_endpoint *ep;
1520 struct sctp_association *asoc;
1521 struct list_head *pos, *temp;
1522 unsigned int data_was_unread;
1524 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1526 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
1527 sk->sk_shutdown = SHUTDOWN_MASK;
1528 inet_sk_set_state(sk, SCTP_SS_CLOSING);
1530 ep = sctp_sk(sk)->ep;
1532 /* Clean up any skbs sitting on the receive queue. */
1533 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1534 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1536 /* Walk all associations on an endpoint. */
1537 list_for_each_safe(pos, temp, &ep->asocs) {
1538 asoc = list_entry(pos, struct sctp_association, asocs);
1540 if (sctp_style(sk, TCP)) {
1541 /* A closed association can still be in the list if
1542 * it belongs to a TCP-style listening socket that is
1543 * not yet accepted. If so, free it. If not, send an
1544 * ABORT or SHUTDOWN based on the linger options.
1546 if (sctp_state(asoc, CLOSED)) {
1547 sctp_association_free(asoc);
1552 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1553 !skb_queue_empty(&asoc->ulpq.reasm) ||
1554 !skb_queue_empty(&asoc->ulpq.reasm_uo) ||
1555 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1556 struct sctp_chunk *chunk;
1558 chunk = sctp_make_abort_user(asoc, NULL, 0);
1559 sctp_primitive_ABORT(net, asoc, chunk);
1561 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1564 /* On a TCP-style socket, block for at most linger_time if set. */
1565 if (sctp_style(sk, TCP) && timeout)
1566 sctp_wait_for_close(sk, timeout);
1568 /* This will run the backlog queue. */
1571 /* Supposedly, no process has access to the socket, but
1572 * the net layers still may.
1573 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1574 * held and that should be grabbed before socket lock.
1576 spin_lock_bh(&net->sctp.addr_wq_lock);
1577 bh_lock_sock_nested(sk);
1579 /* Hold the sock, since sk_common_release() will put sock_put()
1580 * and we have just a little more cleanup.
1583 sk_common_release(sk);
1586 spin_unlock_bh(&net->sctp.addr_wq_lock);
1590 SCTP_DBG_OBJCNT_DEC(sock);
1593 /* Handle EPIPE error. */
1594 static int sctp_error(struct sock *sk, int flags, int err)
1597 err = sock_error(sk) ? : -EPIPE;
1598 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1599 send_sig(SIGPIPE, current, 0);
1603 /* API 3.1.3 sendmsg() - UDP Style Syntax
1605 * An application uses sendmsg() and recvmsg() calls to transmit data to
1606 * and receive data from its peer.
1608 * ssize_t sendmsg(int socket, const struct msghdr *message,
1611 * socket - the socket descriptor of the endpoint.
1612 * message - pointer to the msghdr structure which contains a single
1613 * user message and possibly some ancillary data.
1615 * See Section 5 for complete description of the data
1618 * flags - flags sent or received with the user message, see Section
1619 * 5 for complete description of the flags.
1621 * Note: This function could use a rewrite especially when explicit
1622 * connect support comes in.
1624 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1626 static int sctp_msghdr_parse(const struct msghdr *msg,
1627 struct sctp_cmsgs *cmsgs);
1629 static int sctp_sendmsg_parse(struct sock *sk, struct sctp_cmsgs *cmsgs,
1630 struct sctp_sndrcvinfo *srinfo,
1631 const struct msghdr *msg, size_t msg_len)
1636 if (sctp_sstate(sk, LISTENING) && sctp_style(sk, TCP))
1639 if (msg_len > sk->sk_sndbuf)
1642 memset(cmsgs, 0, sizeof(*cmsgs));
1643 err = sctp_msghdr_parse(msg, cmsgs);
1645 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1649 memset(srinfo, 0, sizeof(*srinfo));
1650 if (cmsgs->srinfo) {
1651 srinfo->sinfo_stream = cmsgs->srinfo->sinfo_stream;
1652 srinfo->sinfo_flags = cmsgs->srinfo->sinfo_flags;
1653 srinfo->sinfo_ppid = cmsgs->srinfo->sinfo_ppid;
1654 srinfo->sinfo_context = cmsgs->srinfo->sinfo_context;
1655 srinfo->sinfo_assoc_id = cmsgs->srinfo->sinfo_assoc_id;
1656 srinfo->sinfo_timetolive = cmsgs->srinfo->sinfo_timetolive;
1660 srinfo->sinfo_stream = cmsgs->sinfo->snd_sid;
1661 srinfo->sinfo_flags = cmsgs->sinfo->snd_flags;
1662 srinfo->sinfo_ppid = cmsgs->sinfo->snd_ppid;
1663 srinfo->sinfo_context = cmsgs->sinfo->snd_context;
1664 srinfo->sinfo_assoc_id = cmsgs->sinfo->snd_assoc_id;
1667 if (cmsgs->prinfo) {
1668 srinfo->sinfo_timetolive = cmsgs->prinfo->pr_value;
1669 SCTP_PR_SET_POLICY(srinfo->sinfo_flags,
1670 cmsgs->prinfo->pr_policy);
1673 sflags = srinfo->sinfo_flags;
1674 if (!sflags && msg_len)
1677 if (sctp_style(sk, TCP) && (sflags & (SCTP_EOF | SCTP_ABORT)))
1680 if (((sflags & SCTP_EOF) && msg_len > 0) ||
1681 (!(sflags & (SCTP_EOF | SCTP_ABORT)) && msg_len == 0))
1684 if ((sflags & SCTP_ADDR_OVER) && !msg->msg_name)
1690 static int sctp_sendmsg_new_asoc(struct sock *sk, __u16 sflags,
1691 struct sctp_cmsgs *cmsgs,
1692 union sctp_addr *daddr,
1693 struct sctp_transport **tp)
1695 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
1696 struct net *net = sock_net(sk);
1697 struct sctp_association *asoc;
1698 enum sctp_scope scope;
1699 struct cmsghdr *cmsg;
1705 if (sflags & (SCTP_EOF | SCTP_ABORT))
1708 if (sctp_style(sk, TCP) && (sctp_sstate(sk, ESTABLISHED) ||
1709 sctp_sstate(sk, CLOSING)))
1710 return -EADDRNOTAVAIL;
1712 if (sctp_endpoint_is_peeled_off(ep, daddr))
1713 return -EADDRNOTAVAIL;
1715 if (!ep->base.bind_addr.port) {
1716 if (sctp_autobind(sk))
1719 if (ep->base.bind_addr.port < inet_prot_sock(net) &&
1720 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
1724 scope = sctp_scope(daddr);
1726 /* Label connection socket for first association 1-to-many
1727 * style for client sequence socket()->sendmsg(). This
1728 * needs to be done before sctp_assoc_add_peer() as that will
1729 * set up the initial packet that needs to account for any
1730 * security ip options (CIPSO/CALIPSO) added to the packet.
1732 af = sctp_get_af_specific(daddr->sa.sa_family);
1735 err = security_sctp_bind_connect(sk, SCTP_SENDMSG_CONNECT,
1736 (struct sockaddr *)daddr,
1741 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1745 if (sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL) < 0) {
1751 struct sctp_initmsg *init = cmsgs->init;
1753 if (init->sinit_num_ostreams) {
1754 __u16 outcnt = init->sinit_num_ostreams;
1756 asoc->c.sinit_num_ostreams = outcnt;
1757 /* outcnt has been changed, need to re-init stream */
1758 err = sctp_stream_init(&asoc->stream, outcnt, 0,
1764 if (init->sinit_max_instreams)
1765 asoc->c.sinit_max_instreams = init->sinit_max_instreams;
1767 if (init->sinit_max_attempts)
1768 asoc->max_init_attempts = init->sinit_max_attempts;
1770 if (init->sinit_max_init_timeo)
1771 asoc->max_init_timeo =
1772 msecs_to_jiffies(init->sinit_max_init_timeo);
1775 *tp = sctp_assoc_add_peer(asoc, daddr, GFP_KERNEL, SCTP_UNKNOWN);
1781 if (!cmsgs->addrs_msg)
1784 /* sendv addr list parse */
1785 for_each_cmsghdr(cmsg, cmsgs->addrs_msg) {
1786 struct sctp_transport *transport;
1787 struct sctp_association *old;
1788 union sctp_addr _daddr;
1791 if (cmsg->cmsg_level != IPPROTO_SCTP ||
1792 (cmsg->cmsg_type != SCTP_DSTADDRV4 &&
1793 cmsg->cmsg_type != SCTP_DSTADDRV6))
1797 memset(daddr, 0, sizeof(*daddr));
1798 dlen = cmsg->cmsg_len - sizeof(struct cmsghdr);
1799 if (cmsg->cmsg_type == SCTP_DSTADDRV4) {
1800 if (dlen < sizeof(struct in_addr)) {
1805 dlen = sizeof(struct in_addr);
1806 daddr->v4.sin_family = AF_INET;
1807 daddr->v4.sin_port = htons(asoc->peer.port);
1808 memcpy(&daddr->v4.sin_addr, CMSG_DATA(cmsg), dlen);
1810 if (dlen < sizeof(struct in6_addr)) {
1815 dlen = sizeof(struct in6_addr);
1816 daddr->v6.sin6_family = AF_INET6;
1817 daddr->v6.sin6_port = htons(asoc->peer.port);
1818 memcpy(&daddr->v6.sin6_addr, CMSG_DATA(cmsg), dlen);
1820 err = sctp_verify_addr(sk, daddr, sizeof(*daddr));
1824 old = sctp_endpoint_lookup_assoc(ep, daddr, &transport);
1825 if (old && old != asoc) {
1826 if (old->state >= SCTP_STATE_ESTABLISHED)
1833 if (sctp_endpoint_is_peeled_off(ep, daddr)) {
1834 err = -EADDRNOTAVAIL;
1838 transport = sctp_assoc_add_peer(asoc, daddr, GFP_KERNEL,
1849 sctp_association_free(asoc);
1853 static int sctp_sendmsg_check_sflags(struct sctp_association *asoc,
1854 __u16 sflags, struct msghdr *msg,
1857 struct sock *sk = asoc->base.sk;
1858 struct net *net = sock_net(sk);
1860 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP))
1863 if ((sflags & SCTP_SENDALL) && sctp_style(sk, UDP) &&
1864 !sctp_state(asoc, ESTABLISHED))
1867 if (sflags & SCTP_EOF) {
1868 pr_debug("%s: shutting down association:%p\n", __func__, asoc);
1869 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1874 if (sflags & SCTP_ABORT) {
1875 struct sctp_chunk *chunk;
1877 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1881 pr_debug("%s: aborting association:%p\n", __func__, asoc);
1882 sctp_primitive_ABORT(net, asoc, chunk);
1890 static int sctp_sendmsg_to_asoc(struct sctp_association *asoc,
1891 struct msghdr *msg, size_t msg_len,
1892 struct sctp_transport *transport,
1893 struct sctp_sndrcvinfo *sinfo)
1895 struct sock *sk = asoc->base.sk;
1896 struct sctp_sock *sp = sctp_sk(sk);
1897 struct net *net = sock_net(sk);
1898 struct sctp_datamsg *datamsg;
1899 bool wait_connect = false;
1900 struct sctp_chunk *chunk;
1904 if (sinfo->sinfo_stream >= asoc->stream.outcnt) {
1909 if (unlikely(!asoc->stream.out[sinfo->sinfo_stream].ext)) {
1910 err = sctp_stream_init_ext(&asoc->stream, sinfo->sinfo_stream);
1915 if (sp->disable_fragments && msg_len > asoc->frag_point) {
1920 if (asoc->pmtu_pending) {
1921 if (sp->param_flags & SPP_PMTUD_ENABLE)
1922 sctp_assoc_sync_pmtu(asoc);
1923 asoc->pmtu_pending = 0;
1926 if (sctp_wspace(asoc) < msg_len)
1927 sctp_prsctp_prune(asoc, sinfo, msg_len - sctp_wspace(asoc));
1929 if (!sctp_wspace(asoc)) {
1930 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1931 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1936 if (sctp_state(asoc, CLOSED)) {
1937 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1941 if (sp->strm_interleave) {
1942 timeo = sock_sndtimeo(sk, 0);
1943 err = sctp_wait_for_connect(asoc, &timeo);
1947 wait_connect = true;
1950 pr_debug("%s: we associated primitively\n", __func__);
1953 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
1954 if (IS_ERR(datamsg)) {
1955 err = PTR_ERR(datamsg);
1959 asoc->force_delay = !!(msg->msg_flags & MSG_MORE);
1961 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1962 sctp_chunk_hold(chunk);
1963 sctp_set_owner_w(chunk);
1964 chunk->transport = transport;
1967 err = sctp_primitive_SEND(net, asoc, datamsg);
1969 sctp_datamsg_free(datamsg);
1973 pr_debug("%s: we sent primitively\n", __func__);
1975 sctp_datamsg_put(datamsg);
1977 if (unlikely(wait_connect)) {
1978 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1979 sctp_wait_for_connect(asoc, &timeo);
1988 static union sctp_addr *sctp_sendmsg_get_daddr(struct sock *sk,
1989 const struct msghdr *msg,
1990 struct sctp_cmsgs *cmsgs)
1992 union sctp_addr *daddr = NULL;
1995 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1996 int len = msg->msg_namelen;
1998 if (len > sizeof(*daddr))
1999 len = sizeof(*daddr);
2001 daddr = (union sctp_addr *)msg->msg_name;
2003 err = sctp_verify_addr(sk, daddr, len);
2005 return ERR_PTR(err);
2011 static void sctp_sendmsg_update_sinfo(struct sctp_association *asoc,
2012 struct sctp_sndrcvinfo *sinfo,
2013 struct sctp_cmsgs *cmsgs)
2015 if (!cmsgs->srinfo && !cmsgs->sinfo) {
2016 sinfo->sinfo_stream = asoc->default_stream;
2017 sinfo->sinfo_ppid = asoc->default_ppid;
2018 sinfo->sinfo_context = asoc->default_context;
2019 sinfo->sinfo_assoc_id = sctp_assoc2id(asoc);
2022 sinfo->sinfo_flags = asoc->default_flags;
2025 if (!cmsgs->srinfo && !cmsgs->prinfo)
2026 sinfo->sinfo_timetolive = asoc->default_timetolive;
2028 if (cmsgs->authinfo) {
2029 /* Reuse sinfo_tsn to indicate that authinfo was set and
2030 * sinfo_ssn to save the keyid on tx path.
2032 sinfo->sinfo_tsn = 1;
2033 sinfo->sinfo_ssn = cmsgs->authinfo->auth_keynumber;
2037 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
2039 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
2040 struct sctp_transport *transport = NULL;
2041 struct sctp_sndrcvinfo _sinfo, *sinfo;
2042 struct sctp_association *asoc;
2043 struct sctp_cmsgs cmsgs;
2044 union sctp_addr *daddr;
2049 /* Parse and get snd_info */
2050 err = sctp_sendmsg_parse(sk, &cmsgs, &_sinfo, msg, msg_len);
2055 sflags = sinfo->sinfo_flags;
2057 /* Get daddr from msg */
2058 daddr = sctp_sendmsg_get_daddr(sk, msg, &cmsgs);
2059 if (IS_ERR(daddr)) {
2060 err = PTR_ERR(daddr);
2066 /* SCTP_SENDALL process */
2067 if ((sflags & SCTP_SENDALL) && sctp_style(sk, UDP)) {
2068 list_for_each_entry(asoc, &ep->asocs, asocs) {
2069 err = sctp_sendmsg_check_sflags(asoc, sflags, msg,
2076 sctp_sendmsg_update_sinfo(asoc, sinfo, &cmsgs);
2078 err = sctp_sendmsg_to_asoc(asoc, msg, msg_len,
2083 iov_iter_revert(&msg->msg_iter, err);
2089 /* Get and check or create asoc */
2091 asoc = sctp_endpoint_lookup_assoc(ep, daddr, &transport);
2093 err = sctp_sendmsg_check_sflags(asoc, sflags, msg,
2098 err = sctp_sendmsg_new_asoc(sk, sflags, &cmsgs, daddr,
2103 asoc = transport->asoc;
2107 if (!sctp_style(sk, TCP) && !(sflags & SCTP_ADDR_OVER))
2110 asoc = sctp_id2assoc(sk, sinfo->sinfo_assoc_id);
2116 err = sctp_sendmsg_check_sflags(asoc, sflags, msg, msg_len);
2121 /* Update snd_info with the asoc */
2122 sctp_sendmsg_update_sinfo(asoc, sinfo, &cmsgs);
2124 /* Send msg to the asoc */
2125 err = sctp_sendmsg_to_asoc(asoc, msg, msg_len, transport, sinfo);
2126 if (err < 0 && err != -ESRCH && new)
2127 sctp_association_free(asoc);
2132 return sctp_error(sk, msg->msg_flags, err);
2135 /* This is an extended version of skb_pull() that removes the data from the
2136 * start of a skb even when data is spread across the list of skb's in the
2137 * frag_list. len specifies the total amount of data that needs to be removed.
2138 * when 'len' bytes could be removed from the skb, it returns 0.
2139 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2140 * could not be removed.
2142 static int sctp_skb_pull(struct sk_buff *skb, int len)
2144 struct sk_buff *list;
2145 int skb_len = skb_headlen(skb);
2148 if (len <= skb_len) {
2149 __skb_pull(skb, len);
2153 __skb_pull(skb, skb_len);
2155 skb_walk_frags(skb, list) {
2156 rlen = sctp_skb_pull(list, len);
2157 skb->len -= (len-rlen);
2158 skb->data_len -= (len-rlen);
2169 /* API 3.1.3 recvmsg() - UDP Style Syntax
2171 * ssize_t recvmsg(int socket, struct msghdr *message,
2174 * socket - the socket descriptor of the endpoint.
2175 * message - pointer to the msghdr structure which contains a single
2176 * user message and possibly some ancillary data.
2178 * See Section 5 for complete description of the data
2181 * flags - flags sent or received with the user message, see Section
2182 * 5 for complete description of the flags.
2184 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2185 int noblock, int flags, int *addr_len)
2187 struct sctp_ulpevent *event = NULL;
2188 struct sctp_sock *sp = sctp_sk(sk);
2189 struct sk_buff *skb, *head_skb;
2194 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2195 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2200 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED) &&
2201 !sctp_sstate(sk, CLOSING) && !sctp_sstate(sk, CLOSED)) {
2206 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2210 /* Get the total length of the skb including any skb's in the
2219 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2221 event = sctp_skb2event(skb);
2226 if (event->chunk && event->chunk->head_skb)
2227 head_skb = event->chunk->head_skb;
2230 sock_recv_ts_and_drops(msg, sk, head_skb);
2231 if (sctp_ulpevent_is_notification(event)) {
2232 msg->msg_flags |= MSG_NOTIFICATION;
2233 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2235 sp->pf->skb_msgname(head_skb, msg->msg_name, addr_len);
2238 /* Check if we allow SCTP_NXTINFO. */
2239 if (sp->recvnxtinfo)
2240 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2241 /* Check if we allow SCTP_RCVINFO. */
2242 if (sp->recvrcvinfo)
2243 sctp_ulpevent_read_rcvinfo(event, msg);
2244 /* Check if we allow SCTP_SNDRCVINFO. */
2245 if (sp->subscribe.sctp_data_io_event)
2246 sctp_ulpevent_read_sndrcvinfo(event, msg);
2250 /* If skb's length exceeds the user's buffer, update the skb and
2251 * push it back to the receive_queue so that the next call to
2252 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2254 if (skb_len > copied) {
2255 msg->msg_flags &= ~MSG_EOR;
2256 if (flags & MSG_PEEK)
2258 sctp_skb_pull(skb, copied);
2259 skb_queue_head(&sk->sk_receive_queue, skb);
2261 /* When only partial message is copied to the user, increase
2262 * rwnd by that amount. If all the data in the skb is read,
2263 * rwnd is updated when the event is freed.
2265 if (!sctp_ulpevent_is_notification(event))
2266 sctp_assoc_rwnd_increase(event->asoc, copied);
2268 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2269 (event->msg_flags & MSG_EOR))
2270 msg->msg_flags |= MSG_EOR;
2272 msg->msg_flags &= ~MSG_EOR;
2275 if (flags & MSG_PEEK) {
2276 /* Release the skb reference acquired after peeking the skb in
2277 * sctp_skb_recv_datagram().
2281 /* Free the event which includes releasing the reference to
2282 * the owner of the skb, freeing the skb and updating the
2285 sctp_ulpevent_free(event);
2292 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2294 * This option is a on/off flag. If enabled no SCTP message
2295 * fragmentation will be performed. Instead if a message being sent
2296 * exceeds the current PMTU size, the message will NOT be sent and
2297 * instead a error will be indicated to the user.
2299 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2300 char __user *optval,
2301 unsigned int optlen)
2305 if (optlen < sizeof(int))
2308 if (get_user(val, (int __user *)optval))
2311 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2316 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2317 unsigned int optlen)
2319 struct sctp_association *asoc;
2320 struct sctp_ulpevent *event;
2322 if (optlen > sizeof(struct sctp_event_subscribe))
2324 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2327 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2328 * if there is no data to be sent or retransmit, the stack will
2329 * immediately send up this notification.
2331 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2332 &sctp_sk(sk)->subscribe)) {
2333 asoc = sctp_id2assoc(sk, 0);
2335 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2336 event = sctp_ulpevent_make_sender_dry_event(asoc,
2337 GFP_USER | __GFP_NOWARN);
2341 asoc->stream.si->enqueue_event(&asoc->ulpq, event);
2348 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2350 * This socket option is applicable to the UDP-style socket only. When
2351 * set it will cause associations that are idle for more than the
2352 * specified number of seconds to automatically close. An association
2353 * being idle is defined an association that has NOT sent or received
2354 * user data. The special value of '0' indicates that no automatic
2355 * close of any associations should be performed. The option expects an
2356 * integer defining the number of seconds of idle time before an
2357 * association is closed.
2359 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2360 unsigned int optlen)
2362 struct sctp_sock *sp = sctp_sk(sk);
2363 struct net *net = sock_net(sk);
2365 /* Applicable to UDP-style socket only */
2366 if (sctp_style(sk, TCP))
2368 if (optlen != sizeof(int))
2370 if (copy_from_user(&sp->autoclose, optval, optlen))
2373 if (sp->autoclose > net->sctp.max_autoclose)
2374 sp->autoclose = net->sctp.max_autoclose;
2379 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2381 * Applications can enable or disable heartbeats for any peer address of
2382 * an association, modify an address's heartbeat interval, force a
2383 * heartbeat to be sent immediately, and adjust the address's maximum
2384 * number of retransmissions sent before an address is considered
2385 * unreachable. The following structure is used to access and modify an
2386 * address's parameters:
2388 * struct sctp_paddrparams {
2389 * sctp_assoc_t spp_assoc_id;
2390 * struct sockaddr_storage spp_address;
2391 * uint32_t spp_hbinterval;
2392 * uint16_t spp_pathmaxrxt;
2393 * uint32_t spp_pathmtu;
2394 * uint32_t spp_sackdelay;
2395 * uint32_t spp_flags;
2398 * spp_assoc_id - (one-to-many style socket) This is filled in the
2399 * application, and identifies the association for
2401 * spp_address - This specifies which address is of interest.
2402 * spp_hbinterval - This contains the value of the heartbeat interval,
2403 * in milliseconds. If a value of zero
2404 * is present in this field then no changes are to
2405 * be made to this parameter.
2406 * spp_pathmaxrxt - This contains the maximum number of
2407 * retransmissions before this address shall be
2408 * considered unreachable. If a value of zero
2409 * is present in this field then no changes are to
2410 * be made to this parameter.
2411 * spp_pathmtu - When Path MTU discovery is disabled the value
2412 * specified here will be the "fixed" path mtu.
2413 * Note that if the spp_address field is empty
2414 * then all associations on this address will
2415 * have this fixed path mtu set upon them.
2417 * spp_sackdelay - When delayed sack is enabled, this value specifies
2418 * the number of milliseconds that sacks will be delayed
2419 * for. This value will apply to all addresses of an
2420 * association if the spp_address field is empty. Note
2421 * also, that if delayed sack is enabled and this
2422 * value is set to 0, no change is made to the last
2423 * recorded delayed sack timer value.
2425 * spp_flags - These flags are used to control various features
2426 * on an association. The flag field may contain
2427 * zero or more of the following options.
2429 * SPP_HB_ENABLE - Enable heartbeats on the
2430 * specified address. Note that if the address
2431 * field is empty all addresses for the association
2432 * have heartbeats enabled upon them.
2434 * SPP_HB_DISABLE - Disable heartbeats on the
2435 * speicifed address. Note that if the address
2436 * field is empty all addresses for the association
2437 * will have their heartbeats disabled. Note also
2438 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2439 * mutually exclusive, only one of these two should
2440 * be specified. Enabling both fields will have
2441 * undetermined results.
2443 * SPP_HB_DEMAND - Request a user initiated heartbeat
2444 * to be made immediately.
2446 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2447 * heartbeat delayis to be set to the value of 0
2450 * SPP_PMTUD_ENABLE - This field will enable PMTU
2451 * discovery upon the specified address. Note that
2452 * if the address feild is empty then all addresses
2453 * on the association are effected.
2455 * SPP_PMTUD_DISABLE - This field will disable PMTU
2456 * discovery upon the specified address. Note that
2457 * if the address feild is empty then all addresses
2458 * on the association are effected. Not also that
2459 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2460 * exclusive. Enabling both will have undetermined
2463 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2464 * on delayed sack. The time specified in spp_sackdelay
2465 * is used to specify the sack delay for this address. Note
2466 * that if spp_address is empty then all addresses will
2467 * enable delayed sack and take on the sack delay
2468 * value specified in spp_sackdelay.
2469 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2470 * off delayed sack. If the spp_address field is blank then
2471 * delayed sack is disabled for the entire association. Note
2472 * also that this field is mutually exclusive to
2473 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2476 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2477 struct sctp_transport *trans,
2478 struct sctp_association *asoc,
2479 struct sctp_sock *sp,
2482 int sackdelay_change)
2486 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2487 struct net *net = sock_net(trans->asoc->base.sk);
2489 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2494 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2495 * this field is ignored. Note also that a value of zero indicates
2496 * the current setting should be left unchanged.
2498 if (params->spp_flags & SPP_HB_ENABLE) {
2500 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2501 * set. This lets us use 0 value when this flag
2504 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2505 params->spp_hbinterval = 0;
2507 if (params->spp_hbinterval ||
2508 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2511 msecs_to_jiffies(params->spp_hbinterval);
2514 msecs_to_jiffies(params->spp_hbinterval);
2516 sp->hbinterval = params->spp_hbinterval;
2523 trans->param_flags =
2524 (trans->param_flags & ~SPP_HB) | hb_change;
2527 (asoc->param_flags & ~SPP_HB) | hb_change;
2530 (sp->param_flags & ~SPP_HB) | hb_change;
2534 /* When Path MTU discovery is disabled the value specified here will
2535 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2536 * include the flag SPP_PMTUD_DISABLE for this field to have any
2539 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2541 trans->pathmtu = params->spp_pathmtu;
2542 sctp_assoc_sync_pmtu(asoc);
2544 sctp_assoc_set_pmtu(asoc, params->spp_pathmtu);
2546 sp->pathmtu = params->spp_pathmtu;
2552 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2553 (params->spp_flags & SPP_PMTUD_ENABLE);
2554 trans->param_flags =
2555 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2557 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2558 sctp_assoc_sync_pmtu(asoc);
2562 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2565 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2569 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2570 * value of this field is ignored. Note also that a value of zero
2571 * indicates the current setting should be left unchanged.
2573 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2576 msecs_to_jiffies(params->spp_sackdelay);
2579 msecs_to_jiffies(params->spp_sackdelay);
2581 sp->sackdelay = params->spp_sackdelay;
2585 if (sackdelay_change) {
2587 trans->param_flags =
2588 (trans->param_flags & ~SPP_SACKDELAY) |
2592 (asoc->param_flags & ~SPP_SACKDELAY) |
2596 (sp->param_flags & ~SPP_SACKDELAY) |
2601 /* Note that a value of zero indicates the current setting should be
2604 if (params->spp_pathmaxrxt) {
2606 trans->pathmaxrxt = params->spp_pathmaxrxt;
2608 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2610 sp->pathmaxrxt = params->spp_pathmaxrxt;
2617 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2618 char __user *optval,
2619 unsigned int optlen)
2621 struct sctp_paddrparams params;
2622 struct sctp_transport *trans = NULL;
2623 struct sctp_association *asoc = NULL;
2624 struct sctp_sock *sp = sctp_sk(sk);
2626 int hb_change, pmtud_change, sackdelay_change;
2628 if (optlen != sizeof(struct sctp_paddrparams))
2631 if (copy_from_user(¶ms, optval, optlen))
2634 /* Validate flags and value parameters. */
2635 hb_change = params.spp_flags & SPP_HB;
2636 pmtud_change = params.spp_flags & SPP_PMTUD;
2637 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2639 if (hb_change == SPP_HB ||
2640 pmtud_change == SPP_PMTUD ||
2641 sackdelay_change == SPP_SACKDELAY ||
2642 params.spp_sackdelay > 500 ||
2643 (params.spp_pathmtu &&
2644 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2647 /* If an address other than INADDR_ANY is specified, and
2648 * no transport is found, then the request is invalid.
2650 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
2651 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2652 params.spp_assoc_id);
2657 /* Get association, if assoc_id != 0 and the socket is a one
2658 * to many style socket, and an association was not found, then
2659 * the id was invalid.
2661 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2662 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2665 /* Heartbeat demand can only be sent on a transport or
2666 * association, but not a socket.
2668 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2671 /* Process parameters. */
2672 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2673 hb_change, pmtud_change,
2679 /* If changes are for association, also apply parameters to each
2682 if (!trans && asoc) {
2683 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2685 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2686 hb_change, pmtud_change,
2694 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2696 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2699 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2701 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2705 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2707 * This option will effect the way delayed acks are performed. This
2708 * option allows you to get or set the delayed ack time, in
2709 * milliseconds. It also allows changing the delayed ack frequency.
2710 * Changing the frequency to 1 disables the delayed sack algorithm. If
2711 * the assoc_id is 0, then this sets or gets the endpoints default
2712 * values. If the assoc_id field is non-zero, then the set or get
2713 * effects the specified association for the one to many model (the
2714 * assoc_id field is ignored by the one to one model). Note that if
2715 * sack_delay or sack_freq are 0 when setting this option, then the
2716 * current values will remain unchanged.
2718 * struct sctp_sack_info {
2719 * sctp_assoc_t sack_assoc_id;
2720 * uint32_t sack_delay;
2721 * uint32_t sack_freq;
2724 * sack_assoc_id - This parameter, indicates which association the user
2725 * is performing an action upon. Note that if this field's value is
2726 * zero then the endpoints default value is changed (effecting future
2727 * associations only).
2729 * sack_delay - This parameter contains the number of milliseconds that
2730 * the user is requesting the delayed ACK timer be set to. Note that
2731 * this value is defined in the standard to be between 200 and 500
2734 * sack_freq - This parameter contains the number of packets that must
2735 * be received before a sack is sent without waiting for the delay
2736 * timer to expire. The default value for this is 2, setting this
2737 * value to 1 will disable the delayed sack algorithm.
2740 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2741 char __user *optval, unsigned int optlen)
2743 struct sctp_sack_info params;
2744 struct sctp_transport *trans = NULL;
2745 struct sctp_association *asoc = NULL;
2746 struct sctp_sock *sp = sctp_sk(sk);
2748 if (optlen == sizeof(struct sctp_sack_info)) {
2749 if (copy_from_user(¶ms, optval, optlen))
2752 if (params.sack_delay == 0 && params.sack_freq == 0)
2754 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2755 pr_warn_ratelimited(DEPRECATED
2757 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2758 "Use struct sctp_sack_info instead\n",
2759 current->comm, task_pid_nr(current));
2760 if (copy_from_user(¶ms, optval, optlen))
2763 if (params.sack_delay == 0)
2764 params.sack_freq = 1;
2766 params.sack_freq = 0;
2770 /* Validate value parameter. */
2771 if (params.sack_delay > 500)
2774 /* Get association, if sack_assoc_id != 0 and the socket is a one
2775 * to many style socket, and an association was not found, then
2776 * the id was invalid.
2778 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2779 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2782 if (params.sack_delay) {
2785 msecs_to_jiffies(params.sack_delay);
2787 sctp_spp_sackdelay_enable(asoc->param_flags);
2789 sp->sackdelay = params.sack_delay;
2791 sctp_spp_sackdelay_enable(sp->param_flags);
2795 if (params.sack_freq == 1) {
2798 sctp_spp_sackdelay_disable(asoc->param_flags);
2801 sctp_spp_sackdelay_disable(sp->param_flags);
2803 } else if (params.sack_freq > 1) {
2805 asoc->sackfreq = params.sack_freq;
2807 sctp_spp_sackdelay_enable(asoc->param_flags);
2809 sp->sackfreq = params.sack_freq;
2811 sctp_spp_sackdelay_enable(sp->param_flags);
2815 /* If change is for association, also apply to each transport. */
2817 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2819 if (params.sack_delay) {
2821 msecs_to_jiffies(params.sack_delay);
2822 trans->param_flags =
2823 sctp_spp_sackdelay_enable(trans->param_flags);
2825 if (params.sack_freq == 1) {
2826 trans->param_flags =
2827 sctp_spp_sackdelay_disable(trans->param_flags);
2828 } else if (params.sack_freq > 1) {
2829 trans->sackfreq = params.sack_freq;
2830 trans->param_flags =
2831 sctp_spp_sackdelay_enable(trans->param_flags);
2839 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2841 * Applications can specify protocol parameters for the default association
2842 * initialization. The option name argument to setsockopt() and getsockopt()
2845 * Setting initialization parameters is effective only on an unconnected
2846 * socket (for UDP-style sockets only future associations are effected
2847 * by the change). With TCP-style sockets, this option is inherited by
2848 * sockets derived from a listener socket.
2850 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2852 struct sctp_initmsg sinit;
2853 struct sctp_sock *sp = sctp_sk(sk);
2855 if (optlen != sizeof(struct sctp_initmsg))
2857 if (copy_from_user(&sinit, optval, optlen))
2860 if (sinit.sinit_num_ostreams)
2861 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2862 if (sinit.sinit_max_instreams)
2863 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2864 if (sinit.sinit_max_attempts)
2865 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2866 if (sinit.sinit_max_init_timeo)
2867 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2873 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2875 * Applications that wish to use the sendto() system call may wish to
2876 * specify a default set of parameters that would normally be supplied
2877 * through the inclusion of ancillary data. This socket option allows
2878 * such an application to set the default sctp_sndrcvinfo structure.
2879 * The application that wishes to use this socket option simply passes
2880 * in to this call the sctp_sndrcvinfo structure defined in Section
2881 * 5.2.2) The input parameters accepted by this call include
2882 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2883 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2884 * to this call if the caller is using the UDP model.
2886 static int sctp_setsockopt_default_send_param(struct sock *sk,
2887 char __user *optval,
2888 unsigned int optlen)
2890 struct sctp_sock *sp = sctp_sk(sk);
2891 struct sctp_association *asoc;
2892 struct sctp_sndrcvinfo info;
2894 if (optlen != sizeof(info))
2896 if (copy_from_user(&info, optval, optlen))
2898 if (info.sinfo_flags &
2899 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2900 SCTP_ABORT | SCTP_EOF))
2903 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2904 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2907 asoc->default_stream = info.sinfo_stream;
2908 asoc->default_flags = info.sinfo_flags;
2909 asoc->default_ppid = info.sinfo_ppid;
2910 asoc->default_context = info.sinfo_context;
2911 asoc->default_timetolive = info.sinfo_timetolive;
2913 sp->default_stream = info.sinfo_stream;
2914 sp->default_flags = info.sinfo_flags;
2915 sp->default_ppid = info.sinfo_ppid;
2916 sp->default_context = info.sinfo_context;
2917 sp->default_timetolive = info.sinfo_timetolive;
2923 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2924 * (SCTP_DEFAULT_SNDINFO)
2926 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
2927 char __user *optval,
2928 unsigned int optlen)
2930 struct sctp_sock *sp = sctp_sk(sk);
2931 struct sctp_association *asoc;
2932 struct sctp_sndinfo info;
2934 if (optlen != sizeof(info))
2936 if (copy_from_user(&info, optval, optlen))
2938 if (info.snd_flags &
2939 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2940 SCTP_ABORT | SCTP_EOF))
2943 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
2944 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
2947 asoc->default_stream = info.snd_sid;
2948 asoc->default_flags = info.snd_flags;
2949 asoc->default_ppid = info.snd_ppid;
2950 asoc->default_context = info.snd_context;
2952 sp->default_stream = info.snd_sid;
2953 sp->default_flags = info.snd_flags;
2954 sp->default_ppid = info.snd_ppid;
2955 sp->default_context = info.snd_context;
2961 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2963 * Requests that the local SCTP stack use the enclosed peer address as
2964 * the association primary. The enclosed address must be one of the
2965 * association peer's addresses.
2967 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2968 unsigned int optlen)
2970 struct sctp_prim prim;
2971 struct sctp_transport *trans;
2975 if (optlen != sizeof(struct sctp_prim))
2978 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2981 /* Allow security module to validate address but need address len. */
2982 af = sctp_get_af_specific(prim.ssp_addr.ss_family);
2986 err = security_sctp_bind_connect(sk, SCTP_PRIMARY_ADDR,
2987 (struct sockaddr *)&prim.ssp_addr,
2992 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2996 sctp_assoc_set_primary(trans->asoc, trans);
3002 * 7.1.5 SCTP_NODELAY
3004 * Turn on/off any Nagle-like algorithm. This means that packets are
3005 * generally sent as soon as possible and no unnecessary delays are
3006 * introduced, at the cost of more packets in the network. Expects an
3007 * integer boolean flag.
3009 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
3010 unsigned int optlen)
3014 if (optlen < sizeof(int))
3016 if (get_user(val, (int __user *)optval))
3019 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
3025 * 7.1.1 SCTP_RTOINFO
3027 * The protocol parameters used to initialize and bound retransmission
3028 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
3029 * and modify these parameters.
3030 * All parameters are time values, in milliseconds. A value of 0, when
3031 * modifying the parameters, indicates that the current value should not
3035 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
3037 struct sctp_rtoinfo rtoinfo;
3038 struct sctp_association *asoc;
3039 unsigned long rto_min, rto_max;
3040 struct sctp_sock *sp = sctp_sk(sk);
3042 if (optlen != sizeof (struct sctp_rtoinfo))
3045 if (copy_from_user(&rtoinfo, optval, optlen))
3048 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
3050 /* Set the values to the specific association */
3051 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
3054 rto_max = rtoinfo.srto_max;
3055 rto_min = rtoinfo.srto_min;
3058 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
3060 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
3063 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
3065 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
3067 if (rto_min > rto_max)
3071 if (rtoinfo.srto_initial != 0)
3073 msecs_to_jiffies(rtoinfo.srto_initial);
3074 asoc->rto_max = rto_max;
3075 asoc->rto_min = rto_min;
3077 /* If there is no association or the association-id = 0
3078 * set the values to the endpoint.
3080 if (rtoinfo.srto_initial != 0)
3081 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
3082 sp->rtoinfo.srto_max = rto_max;
3083 sp->rtoinfo.srto_min = rto_min;
3091 * 7.1.2 SCTP_ASSOCINFO
3093 * This option is used to tune the maximum retransmission attempts
3094 * of the association.
3095 * Returns an error if the new association retransmission value is
3096 * greater than the sum of the retransmission value of the peer.
3097 * See [SCTP] for more information.
3100 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
3103 struct sctp_assocparams assocparams;
3104 struct sctp_association *asoc;
3106 if (optlen != sizeof(struct sctp_assocparams))
3108 if (copy_from_user(&assocparams, optval, optlen))
3111 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
3113 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
3116 /* Set the values to the specific association */
3118 if (assocparams.sasoc_asocmaxrxt != 0) {
3121 struct sctp_transport *peer_addr;
3123 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
3125 path_sum += peer_addr->pathmaxrxt;
3129 /* Only validate asocmaxrxt if we have more than
3130 * one path/transport. We do this because path
3131 * retransmissions are only counted when we have more
3135 assocparams.sasoc_asocmaxrxt > path_sum)
3138 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3141 if (assocparams.sasoc_cookie_life != 0)
3142 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3144 /* Set the values to the endpoint */
3145 struct sctp_sock *sp = sctp_sk(sk);
3147 if (assocparams.sasoc_asocmaxrxt != 0)
3148 sp->assocparams.sasoc_asocmaxrxt =
3149 assocparams.sasoc_asocmaxrxt;
3150 if (assocparams.sasoc_cookie_life != 0)
3151 sp->assocparams.sasoc_cookie_life =
3152 assocparams.sasoc_cookie_life;
3158 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3160 * This socket option is a boolean flag which turns on or off mapped V4
3161 * addresses. If this option is turned on and the socket is type
3162 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3163 * If this option is turned off, then no mapping will be done of V4
3164 * addresses and a user will receive both PF_INET6 and PF_INET type
3165 * addresses on the socket.
3167 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3170 struct sctp_sock *sp = sctp_sk(sk);
3172 if (optlen < sizeof(int))
3174 if (get_user(val, (int __user *)optval))
3185 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3186 * This option will get or set the maximum size to put in any outgoing
3187 * SCTP DATA chunk. If a message is larger than this size it will be
3188 * fragmented by SCTP into the specified size. Note that the underlying
3189 * SCTP implementation may fragment into smaller sized chunks when the
3190 * PMTU of the underlying association is smaller than the value set by
3191 * the user. The default value for this option is '0' which indicates
3192 * the user is NOT limiting fragmentation and only the PMTU will effect
3193 * SCTP's choice of DATA chunk size. Note also that values set larger
3194 * than the maximum size of an IP datagram will effectively let SCTP
3195 * control fragmentation (i.e. the same as setting this option to 0).
3197 * The following structure is used to access and modify this parameter:
3199 * struct sctp_assoc_value {
3200 * sctp_assoc_t assoc_id;
3201 * uint32_t assoc_value;
3204 * assoc_id: This parameter is ignored for one-to-one style sockets.
3205 * For one-to-many style sockets this parameter indicates which
3206 * association the user is performing an action upon. Note that if
3207 * this field's value is zero then the endpoints default value is
3208 * changed (effecting future associations only).
3209 * assoc_value: This parameter specifies the maximum size in bytes.
3211 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3213 struct sctp_sock *sp = sctp_sk(sk);
3214 struct sctp_af *af = sp->pf->af;
3215 struct sctp_assoc_value params;
3216 struct sctp_association *asoc;
3219 if (optlen == sizeof(int)) {
3220 pr_warn_ratelimited(DEPRECATED
3222 "Use of int in maxseg socket option.\n"
3223 "Use struct sctp_assoc_value instead\n",
3224 current->comm, task_pid_nr(current));
3225 if (copy_from_user(&val, optval, optlen))
3227 params.assoc_id = 0;
3228 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3229 if (copy_from_user(¶ms, optval, optlen))
3231 val = params.assoc_value;
3236 asoc = sctp_id2assoc(sk, params.assoc_id);
3239 int min_len, max_len;
3240 __u16 datasize = asoc ? sctp_datachk_len(&asoc->stream) :
3241 sizeof(struct sctp_data_chunk);
3243 min_len = sctp_mtu_payload(sp, SCTP_DEFAULT_MINSEGMENT,
3245 max_len = SCTP_MAX_CHUNK_LEN - datasize;
3247 if (val < min_len || val > max_len)
3253 val = asoc->pathmtu - af->net_header_len;
3254 val -= af->ip_options_len(sk);
3255 val -= sizeof(struct sctphdr) +
3256 sctp_datachk_len(&asoc->stream);
3258 asoc->user_frag = val;
3259 sctp_assoc_update_frag_point(asoc);
3261 if (params.assoc_id && sctp_style(sk, UDP))
3263 sp->user_frag = val;
3271 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3273 * Requests that the peer mark the enclosed address as the association
3274 * primary. The enclosed address must be one of the association's
3275 * locally bound addresses. The following structure is used to make a
3276 * set primary request:
3278 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3279 unsigned int optlen)
3281 struct net *net = sock_net(sk);
3282 struct sctp_sock *sp;
3283 struct sctp_association *asoc = NULL;
3284 struct sctp_setpeerprim prim;
3285 struct sctp_chunk *chunk;
3291 if (!net->sctp.addip_enable)
3294 if (optlen != sizeof(struct sctp_setpeerprim))
3297 if (copy_from_user(&prim, optval, optlen))
3300 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3304 if (!asoc->peer.asconf_capable)
3307 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3310 if (!sctp_state(asoc, ESTABLISHED))
3313 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3317 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3318 return -EADDRNOTAVAIL;
3320 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3321 return -EADDRNOTAVAIL;
3323 /* Allow security module to validate address. */
3324 err = security_sctp_bind_connect(sk, SCTP_SET_PEER_PRIMARY_ADDR,
3325 (struct sockaddr *)&prim.sspp_addr,
3330 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3331 chunk = sctp_make_asconf_set_prim(asoc,
3332 (union sctp_addr *)&prim.sspp_addr);
3336 err = sctp_send_asconf(asoc, chunk);
3338 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3343 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3344 unsigned int optlen)
3346 struct sctp_setadaptation adaptation;
3348 if (optlen != sizeof(struct sctp_setadaptation))
3350 if (copy_from_user(&adaptation, optval, optlen))
3353 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3359 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3361 * The context field in the sctp_sndrcvinfo structure is normally only
3362 * used when a failed message is retrieved holding the value that was
3363 * sent down on the actual send call. This option allows the setting of
3364 * a default context on an association basis that will be received on
3365 * reading messages from the peer. This is especially helpful in the
3366 * one-2-many model for an application to keep some reference to an
3367 * internal state machine that is processing messages on the
3368 * association. Note that the setting of this value only effects
3369 * received messages from the peer and does not effect the value that is
3370 * saved with outbound messages.
3372 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3373 unsigned int optlen)
3375 struct sctp_assoc_value params;
3376 struct sctp_sock *sp;
3377 struct sctp_association *asoc;
3379 if (optlen != sizeof(struct sctp_assoc_value))
3381 if (copy_from_user(¶ms, optval, optlen))
3386 if (params.assoc_id != 0) {
3387 asoc = sctp_id2assoc(sk, params.assoc_id);
3390 asoc->default_rcv_context = params.assoc_value;
3392 sp->default_rcv_context = params.assoc_value;
3399 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3401 * This options will at a minimum specify if the implementation is doing
3402 * fragmented interleave. Fragmented interleave, for a one to many
3403 * socket, is when subsequent calls to receive a message may return
3404 * parts of messages from different associations. Some implementations
3405 * may allow you to turn this value on or off. If so, when turned off,
3406 * no fragment interleave will occur (which will cause a head of line
3407 * blocking amongst multiple associations sharing the same one to many
3408 * socket). When this option is turned on, then each receive call may
3409 * come from a different association (thus the user must receive data
3410 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3411 * association each receive belongs to.
3413 * This option takes a boolean value. A non-zero value indicates that
3414 * fragmented interleave is on. A value of zero indicates that
3415 * fragmented interleave is off.
3417 * Note that it is important that an implementation that allows this
3418 * option to be turned on, have it off by default. Otherwise an unaware
3419 * application using the one to many model may become confused and act
3422 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3423 char __user *optval,
3424 unsigned int optlen)
3428 if (optlen != sizeof(int))
3430 if (get_user(val, (int __user *)optval))
3433 sctp_sk(sk)->frag_interleave = !!val;
3435 if (!sctp_sk(sk)->frag_interleave)
3436 sctp_sk(sk)->strm_interleave = 0;
3442 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3443 * (SCTP_PARTIAL_DELIVERY_POINT)
3445 * This option will set or get the SCTP partial delivery point. This
3446 * point is the size of a message where the partial delivery API will be
3447 * invoked to help free up rwnd space for the peer. Setting this to a
3448 * lower value will cause partial deliveries to happen more often. The
3449 * calls argument is an integer that sets or gets the partial delivery
3450 * point. Note also that the call will fail if the user attempts to set
3451 * this value larger than the socket receive buffer size.
3453 * Note that any single message having a length smaller than or equal to
3454 * the SCTP partial delivery point will be delivered in one single read
3455 * call as long as the user provided buffer is large enough to hold the
3458 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3459 char __user *optval,
3460 unsigned int optlen)
3464 if (optlen != sizeof(u32))
3466 if (get_user(val, (int __user *)optval))
3469 /* Note: We double the receive buffer from what the user sets
3470 * it to be, also initial rwnd is based on rcvbuf/2.
3472 if (val > (sk->sk_rcvbuf >> 1))
3475 sctp_sk(sk)->pd_point = val;
3477 return 0; /* is this the right error code? */
3481 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3483 * This option will allow a user to change the maximum burst of packets
3484 * that can be emitted by this association. Note that the default value
3485 * is 4, and some implementations may restrict this setting so that it
3486 * can only be lowered.
3488 * NOTE: This text doesn't seem right. Do this on a socket basis with
3489 * future associations inheriting the socket value.
3491 static int sctp_setsockopt_maxburst(struct sock *sk,
3492 char __user *optval,
3493 unsigned int optlen)
3495 struct sctp_assoc_value params;
3496 struct sctp_sock *sp;
3497 struct sctp_association *asoc;
3501 if (optlen == sizeof(int)) {
3502 pr_warn_ratelimited(DEPRECATED
3504 "Use of int in max_burst socket option deprecated.\n"
3505 "Use struct sctp_assoc_value instead\n",
3506 current->comm, task_pid_nr(current));
3507 if (copy_from_user(&val, optval, optlen))
3509 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3510 if (copy_from_user(¶ms, optval, optlen))
3512 val = params.assoc_value;
3513 assoc_id = params.assoc_id;
3519 if (assoc_id != 0) {
3520 asoc = sctp_id2assoc(sk, assoc_id);
3523 asoc->max_burst = val;
3525 sp->max_burst = val;
3531 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3533 * This set option adds a chunk type that the user is requesting to be
3534 * received only in an authenticated way. Changes to the list of chunks
3535 * will only effect future associations on the socket.
3537 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3538 char __user *optval,
3539 unsigned int optlen)
3541 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3542 struct sctp_authchunk val;
3544 if (!ep->auth_enable)
3547 if (optlen != sizeof(struct sctp_authchunk))
3549 if (copy_from_user(&val, optval, optlen))
3552 switch (val.sauth_chunk) {
3554 case SCTP_CID_INIT_ACK:
3555 case SCTP_CID_SHUTDOWN_COMPLETE:
3560 /* add this chunk id to the endpoint */
3561 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3565 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3567 * This option gets or sets the list of HMAC algorithms that the local
3568 * endpoint requires the peer to use.
3570 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3571 char __user *optval,
3572 unsigned int optlen)
3574 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3575 struct sctp_hmacalgo *hmacs;
3579 if (!ep->auth_enable)
3582 if (optlen < sizeof(struct sctp_hmacalgo))
3584 optlen = min_t(unsigned int, optlen, sizeof(struct sctp_hmacalgo) +
3585 SCTP_AUTH_NUM_HMACS * sizeof(u16));
3587 hmacs = memdup_user(optval, optlen);
3589 return PTR_ERR(hmacs);
3591 idents = hmacs->shmac_num_idents;
3592 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3593 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3598 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3605 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3607 * This option will set a shared secret key which is used to build an
3608 * association shared key.
3610 static int sctp_setsockopt_auth_key(struct sock *sk,
3611 char __user *optval,
3612 unsigned int optlen)
3614 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3615 struct sctp_authkey *authkey;
3616 struct sctp_association *asoc;
3619 if (!ep->auth_enable)
3622 if (optlen <= sizeof(struct sctp_authkey))
3624 /* authkey->sca_keylength is u16, so optlen can't be bigger than
3627 optlen = min_t(unsigned int, optlen, USHRT_MAX +
3628 sizeof(struct sctp_authkey));
3630 authkey = memdup_user(optval, optlen);
3631 if (IS_ERR(authkey))
3632 return PTR_ERR(authkey);
3634 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3639 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3640 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3645 ret = sctp_auth_set_key(ep, asoc, authkey);
3652 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3654 * This option will get or set the active shared key to be used to build
3655 * the association shared key.
3657 static int sctp_setsockopt_active_key(struct sock *sk,
3658 char __user *optval,
3659 unsigned int optlen)
3661 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3662 struct sctp_authkeyid val;
3663 struct sctp_association *asoc;
3665 if (!ep->auth_enable)
3668 if (optlen != sizeof(struct sctp_authkeyid))
3670 if (copy_from_user(&val, optval, optlen))
3673 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3674 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3677 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3681 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3683 * This set option will delete a shared secret key from use.
3685 static int sctp_setsockopt_del_key(struct sock *sk,
3686 char __user *optval,
3687 unsigned int optlen)
3689 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3690 struct sctp_authkeyid val;
3691 struct sctp_association *asoc;
3693 if (!ep->auth_enable)
3696 if (optlen != sizeof(struct sctp_authkeyid))
3698 if (copy_from_user(&val, optval, optlen))
3701 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3702 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3705 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3710 * 8.3.4 Deactivate a Shared Key (SCTP_AUTH_DEACTIVATE_KEY)
3712 * This set option will deactivate a shared secret key.
3714 static int sctp_setsockopt_deactivate_key(struct sock *sk, char __user *optval,
3715 unsigned int optlen)
3717 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3718 struct sctp_authkeyid val;
3719 struct sctp_association *asoc;
3721 if (!ep->auth_enable)
3724 if (optlen != sizeof(struct sctp_authkeyid))
3726 if (copy_from_user(&val, optval, optlen))
3729 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3730 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3733 return sctp_auth_deact_key_id(ep, asoc, val.scact_keynumber);
3737 * 8.1.23 SCTP_AUTO_ASCONF
3739 * This option will enable or disable the use of the automatic generation of
3740 * ASCONF chunks to add and delete addresses to an existing association. Note
3741 * that this option has two caveats namely: a) it only affects sockets that
3742 * are bound to all addresses available to the SCTP stack, and b) the system
3743 * administrator may have an overriding control that turns the ASCONF feature
3744 * off no matter what setting the socket option may have.
3745 * This option expects an integer boolean flag, where a non-zero value turns on
3746 * the option, and a zero value turns off the option.
3747 * Note. In this implementation, socket operation overrides default parameter
3748 * being set by sysctl as well as FreeBSD implementation
3750 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3751 unsigned int optlen)
3754 struct sctp_sock *sp = sctp_sk(sk);
3756 if (optlen < sizeof(int))
3758 if (get_user(val, (int __user *)optval))
3760 if (!sctp_is_ep_boundall(sk) && val)
3762 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3765 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3766 if (val == 0 && sp->do_auto_asconf) {
3767 list_del(&sp->auto_asconf_list);
3768 sp->do_auto_asconf = 0;
3769 } else if (val && !sp->do_auto_asconf) {
3770 list_add_tail(&sp->auto_asconf_list,
3771 &sock_net(sk)->sctp.auto_asconf_splist);
3772 sp->do_auto_asconf = 1;
3774 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3779 * SCTP_PEER_ADDR_THLDS
3781 * This option allows us to alter the partially failed threshold for one or all
3782 * transports in an association. See Section 6.1 of:
3783 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3785 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3786 char __user *optval,
3787 unsigned int optlen)
3789 struct sctp_paddrthlds val;
3790 struct sctp_transport *trans;
3791 struct sctp_association *asoc;
3793 if (optlen < sizeof(struct sctp_paddrthlds))
3795 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3796 sizeof(struct sctp_paddrthlds)))
3800 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3801 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3804 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3806 if (val.spt_pathmaxrxt)
3807 trans->pathmaxrxt = val.spt_pathmaxrxt;
3808 trans->pf_retrans = val.spt_pathpfthld;
3811 if (val.spt_pathmaxrxt)
3812 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3813 asoc->pf_retrans = val.spt_pathpfthld;
3815 trans = sctp_addr_id2transport(sk, &val.spt_address,
3820 if (val.spt_pathmaxrxt)
3821 trans->pathmaxrxt = val.spt_pathmaxrxt;
3822 trans->pf_retrans = val.spt_pathpfthld;
3828 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3829 char __user *optval,
3830 unsigned int optlen)
3834 if (optlen < sizeof(int))
3836 if (get_user(val, (int __user *) optval))
3839 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3844 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3845 char __user *optval,
3846 unsigned int optlen)
3850 if (optlen < sizeof(int))
3852 if (get_user(val, (int __user *) optval))
3855 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3860 static int sctp_setsockopt_pr_supported(struct sock *sk,
3861 char __user *optval,
3862 unsigned int optlen)
3864 struct sctp_assoc_value params;
3865 struct sctp_association *asoc;
3866 int retval = -EINVAL;
3868 if (optlen != sizeof(params))
3871 if (copy_from_user(¶ms, optval, optlen)) {
3876 asoc = sctp_id2assoc(sk, params.assoc_id);
3878 asoc->prsctp_enable = !!params.assoc_value;
3879 } else if (!params.assoc_id) {
3880 struct sctp_sock *sp = sctp_sk(sk);
3882 sp->ep->prsctp_enable = !!params.assoc_value;
3893 static int sctp_setsockopt_default_prinfo(struct sock *sk,
3894 char __user *optval,
3895 unsigned int optlen)
3897 struct sctp_default_prinfo info;
3898 struct sctp_association *asoc;
3899 int retval = -EINVAL;
3901 if (optlen != sizeof(info))
3904 if (copy_from_user(&info, optval, sizeof(info))) {
3909 if (info.pr_policy & ~SCTP_PR_SCTP_MASK)
3912 if (info.pr_policy == SCTP_PR_SCTP_NONE)
3915 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
3917 SCTP_PR_SET_POLICY(asoc->default_flags, info.pr_policy);
3918 asoc->default_timetolive = info.pr_value;
3919 } else if (!info.pr_assoc_id) {
3920 struct sctp_sock *sp = sctp_sk(sk);
3922 SCTP_PR_SET_POLICY(sp->default_flags, info.pr_policy);
3923 sp->default_timetolive = info.pr_value;
3934 static int sctp_setsockopt_reconfig_supported(struct sock *sk,
3935 char __user *optval,
3936 unsigned int optlen)
3938 struct sctp_assoc_value params;
3939 struct sctp_association *asoc;
3940 int retval = -EINVAL;
3942 if (optlen != sizeof(params))
3945 if (copy_from_user(¶ms, optval, optlen)) {
3950 asoc = sctp_id2assoc(sk, params.assoc_id);
3952 asoc->reconf_enable = !!params.assoc_value;
3953 } else if (!params.assoc_id) {
3954 struct sctp_sock *sp = sctp_sk(sk);
3956 sp->ep->reconf_enable = !!params.assoc_value;
3967 static int sctp_setsockopt_enable_strreset(struct sock *sk,
3968 char __user *optval,
3969 unsigned int optlen)
3971 struct sctp_assoc_value params;
3972 struct sctp_association *asoc;
3973 int retval = -EINVAL;
3975 if (optlen != sizeof(params))
3978 if (copy_from_user(¶ms, optval, optlen)) {
3983 if (params.assoc_value & (~SCTP_ENABLE_STRRESET_MASK))
3986 asoc = sctp_id2assoc(sk, params.assoc_id);
3988 asoc->strreset_enable = params.assoc_value;
3989 } else if (!params.assoc_id) {
3990 struct sctp_sock *sp = sctp_sk(sk);
3992 sp->ep->strreset_enable = params.assoc_value;
4003 static int sctp_setsockopt_reset_streams(struct sock *sk,
4004 char __user *optval,
4005 unsigned int optlen)
4007 struct sctp_reset_streams *params;
4008 struct sctp_association *asoc;
4009 int retval = -EINVAL;
4011 if (optlen < sizeof(*params))
4013 /* srs_number_streams is u16, so optlen can't be bigger than this. */
4014 optlen = min_t(unsigned int, optlen, USHRT_MAX +
4015 sizeof(__u16) * sizeof(*params));
4017 params = memdup_user(optval, optlen);
4019 return PTR_ERR(params);
4021 if (params->srs_number_streams * sizeof(__u16) >
4022 optlen - sizeof(*params))
4025 asoc = sctp_id2assoc(sk, params->srs_assoc_id);
4029 retval = sctp_send_reset_streams(asoc, params);
4036 static int sctp_setsockopt_reset_assoc(struct sock *sk,
4037 char __user *optval,
4038 unsigned int optlen)
4040 struct sctp_association *asoc;
4041 sctp_assoc_t associd;
4042 int retval = -EINVAL;
4044 if (optlen != sizeof(associd))
4047 if (copy_from_user(&associd, optval, optlen)) {
4052 asoc = sctp_id2assoc(sk, associd);
4056 retval = sctp_send_reset_assoc(asoc);
4062 static int sctp_setsockopt_add_streams(struct sock *sk,
4063 char __user *optval,
4064 unsigned int optlen)
4066 struct sctp_association *asoc;
4067 struct sctp_add_streams params;
4068 int retval = -EINVAL;
4070 if (optlen != sizeof(params))
4073 if (copy_from_user(¶ms, optval, optlen)) {
4078 asoc = sctp_id2assoc(sk, params.sas_assoc_id);
4082 retval = sctp_send_add_streams(asoc, ¶ms);
4088 static int sctp_setsockopt_scheduler(struct sock *sk,
4089 char __user *optval,
4090 unsigned int optlen)
4092 struct sctp_association *asoc;
4093 struct sctp_assoc_value params;
4094 int retval = -EINVAL;
4096 if (optlen < sizeof(params))
4099 optlen = sizeof(params);
4100 if (copy_from_user(¶ms, optval, optlen)) {
4105 if (params.assoc_value > SCTP_SS_MAX)
4108 asoc = sctp_id2assoc(sk, params.assoc_id);
4112 retval = sctp_sched_set_sched(asoc, params.assoc_value);
4118 static int sctp_setsockopt_scheduler_value(struct sock *sk,
4119 char __user *optval,
4120 unsigned int optlen)
4122 struct sctp_association *asoc;
4123 struct sctp_stream_value params;
4124 int retval = -EINVAL;
4126 if (optlen < sizeof(params))
4129 optlen = sizeof(params);
4130 if (copy_from_user(¶ms, optval, optlen)) {
4135 asoc = sctp_id2assoc(sk, params.assoc_id);
4139 retval = sctp_sched_set_value(asoc, params.stream_id,
4140 params.stream_value, GFP_KERNEL);
4146 static int sctp_setsockopt_interleaving_supported(struct sock *sk,
4147 char __user *optval,
4148 unsigned int optlen)
4150 struct sctp_sock *sp = sctp_sk(sk);
4151 struct net *net = sock_net(sk);
4152 struct sctp_assoc_value params;
4153 int retval = -EINVAL;
4155 if (optlen < sizeof(params))
4158 optlen = sizeof(params);
4159 if (copy_from_user(¶ms, optval, optlen)) {
4164 if (params.assoc_id)
4167 if (!net->sctp.intl_enable || !sp->frag_interleave) {
4172 sp->strm_interleave = !!params.assoc_value;
4180 /* API 6.2 setsockopt(), getsockopt()
4182 * Applications use setsockopt() and getsockopt() to set or retrieve
4183 * socket options. Socket options are used to change the default
4184 * behavior of sockets calls. They are described in Section 7.
4188 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
4189 * int __user *optlen);
4190 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
4193 * sd - the socket descript.
4194 * level - set to IPPROTO_SCTP for all SCTP options.
4195 * optname - the option name.
4196 * optval - the buffer to store the value of the option.
4197 * optlen - the size of the buffer.
4199 static int sctp_setsockopt(struct sock *sk, int level, int optname,
4200 char __user *optval, unsigned int optlen)
4204 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
4206 /* I can hardly begin to describe how wrong this is. This is
4207 * so broken as to be worse than useless. The API draft
4208 * REALLY is NOT helpful here... I am not convinced that the
4209 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
4210 * are at all well-founded.
4212 if (level != SOL_SCTP) {
4213 struct sctp_af *af = sctp_sk(sk)->pf->af;
4214 retval = af->setsockopt(sk, level, optname, optval, optlen);
4221 case SCTP_SOCKOPT_BINDX_ADD:
4222 /* 'optlen' is the size of the addresses buffer. */
4223 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
4224 optlen, SCTP_BINDX_ADD_ADDR);
4227 case SCTP_SOCKOPT_BINDX_REM:
4228 /* 'optlen' is the size of the addresses buffer. */
4229 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
4230 optlen, SCTP_BINDX_REM_ADDR);
4233 case SCTP_SOCKOPT_CONNECTX_OLD:
4234 /* 'optlen' is the size of the addresses buffer. */
4235 retval = sctp_setsockopt_connectx_old(sk,
4236 (struct sockaddr __user *)optval,
4240 case SCTP_SOCKOPT_CONNECTX:
4241 /* 'optlen' is the size of the addresses buffer. */
4242 retval = sctp_setsockopt_connectx(sk,
4243 (struct sockaddr __user *)optval,
4247 case SCTP_DISABLE_FRAGMENTS:
4248 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
4252 retval = sctp_setsockopt_events(sk, optval, optlen);
4255 case SCTP_AUTOCLOSE:
4256 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
4259 case SCTP_PEER_ADDR_PARAMS:
4260 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
4263 case SCTP_DELAYED_SACK:
4264 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
4266 case SCTP_PARTIAL_DELIVERY_POINT:
4267 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
4271 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
4273 case SCTP_DEFAULT_SEND_PARAM:
4274 retval = sctp_setsockopt_default_send_param(sk, optval,
4277 case SCTP_DEFAULT_SNDINFO:
4278 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
4280 case SCTP_PRIMARY_ADDR:
4281 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
4283 case SCTP_SET_PEER_PRIMARY_ADDR:
4284 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
4287 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
4290 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
4292 case SCTP_ASSOCINFO:
4293 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
4295 case SCTP_I_WANT_MAPPED_V4_ADDR:
4296 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
4299 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
4301 case SCTP_ADAPTATION_LAYER:
4302 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
4305 retval = sctp_setsockopt_context(sk, optval, optlen);
4307 case SCTP_FRAGMENT_INTERLEAVE:
4308 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
4310 case SCTP_MAX_BURST:
4311 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
4313 case SCTP_AUTH_CHUNK:
4314 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
4316 case SCTP_HMAC_IDENT:
4317 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
4320 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
4322 case SCTP_AUTH_ACTIVE_KEY:
4323 retval = sctp_setsockopt_active_key(sk, optval, optlen);
4325 case SCTP_AUTH_DELETE_KEY:
4326 retval = sctp_setsockopt_del_key(sk, optval, optlen);
4328 case SCTP_AUTH_DEACTIVATE_KEY:
4329 retval = sctp_setsockopt_deactivate_key(sk, optval, optlen);
4331 case SCTP_AUTO_ASCONF:
4332 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
4334 case SCTP_PEER_ADDR_THLDS:
4335 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
4337 case SCTP_RECVRCVINFO:
4338 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
4340 case SCTP_RECVNXTINFO:
4341 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
4343 case SCTP_PR_SUPPORTED:
4344 retval = sctp_setsockopt_pr_supported(sk, optval, optlen);
4346 case SCTP_DEFAULT_PRINFO:
4347 retval = sctp_setsockopt_default_prinfo(sk, optval, optlen);
4349 case SCTP_RECONFIG_SUPPORTED:
4350 retval = sctp_setsockopt_reconfig_supported(sk, optval, optlen);
4352 case SCTP_ENABLE_STREAM_RESET:
4353 retval = sctp_setsockopt_enable_strreset(sk, optval, optlen);
4355 case SCTP_RESET_STREAMS:
4356 retval = sctp_setsockopt_reset_streams(sk, optval, optlen);
4358 case SCTP_RESET_ASSOC:
4359 retval = sctp_setsockopt_reset_assoc(sk, optval, optlen);
4361 case SCTP_ADD_STREAMS:
4362 retval = sctp_setsockopt_add_streams(sk, optval, optlen);
4364 case SCTP_STREAM_SCHEDULER:
4365 retval = sctp_setsockopt_scheduler(sk, optval, optlen);
4367 case SCTP_STREAM_SCHEDULER_VALUE:
4368 retval = sctp_setsockopt_scheduler_value(sk, optval, optlen);
4370 case SCTP_INTERLEAVING_SUPPORTED:
4371 retval = sctp_setsockopt_interleaving_supported(sk, optval,
4375 retval = -ENOPROTOOPT;
4385 /* API 3.1.6 connect() - UDP Style Syntax
4387 * An application may use the connect() call in the UDP model to initiate an
4388 * association without sending data.
4392 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
4394 * sd: the socket descriptor to have a new association added to.
4396 * nam: the address structure (either struct sockaddr_in or struct
4397 * sockaddr_in6 defined in RFC2553 [7]).
4399 * len: the size of the address.
4401 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
4409 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
4412 /* Validate addr_len before calling common connect/connectx routine. */
4413 af = sctp_get_af_specific(addr->sa_family);
4414 if (!af || addr_len < af->sockaddr_len) {
4417 /* Pass correct addr len to common routine (so it knows there
4418 * is only one address being passed.
4420 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
4427 /* FIXME: Write comments. */
4428 static int sctp_disconnect(struct sock *sk, int flags)
4430 return -EOPNOTSUPP; /* STUB */
4433 /* 4.1.4 accept() - TCP Style Syntax
4435 * Applications use accept() call to remove an established SCTP
4436 * association from the accept queue of the endpoint. A new socket
4437 * descriptor will be returned from accept() to represent the newly
4438 * formed association.
4440 static struct sock *sctp_accept(struct sock *sk, int flags, int *err, bool kern)
4442 struct sctp_sock *sp;
4443 struct sctp_endpoint *ep;
4444 struct sock *newsk = NULL;
4445 struct sctp_association *asoc;
4454 if (!sctp_style(sk, TCP)) {
4455 error = -EOPNOTSUPP;
4459 if (!sctp_sstate(sk, LISTENING)) {
4464 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
4466 error = sctp_wait_for_accept(sk, timeo);
4470 /* We treat the list of associations on the endpoint as the accept
4471 * queue and pick the first association on the list.
4473 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
4475 newsk = sp->pf->create_accept_sk(sk, asoc, kern);
4481 /* Populate the fields of the newsk from the oldsk and migrate the
4482 * asoc to the newsk.
4484 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
4492 /* The SCTP ioctl handler. */
4493 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
4500 * SEQPACKET-style sockets in LISTENING state are valid, for
4501 * SCTP, so only discard TCP-style sockets in LISTENING state.
4503 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4508 struct sk_buff *skb;
4509 unsigned int amount = 0;
4511 skb = skb_peek(&sk->sk_receive_queue);
4514 * We will only return the amount of this packet since
4515 * that is all that will be read.
4519 rc = put_user(amount, (int __user *)arg);
4531 /* This is the function which gets called during socket creation to
4532 * initialized the SCTP-specific portion of the sock.
4533 * The sock structure should already be zero-filled memory.
4535 static int sctp_init_sock(struct sock *sk)
4537 struct net *net = sock_net(sk);
4538 struct sctp_sock *sp;
4540 pr_debug("%s: sk:%p\n", __func__, sk);
4544 /* Initialize the SCTP per socket area. */
4545 switch (sk->sk_type) {
4546 case SOCK_SEQPACKET:
4547 sp->type = SCTP_SOCKET_UDP;
4550 sp->type = SCTP_SOCKET_TCP;
4553 return -ESOCKTNOSUPPORT;
4556 sk->sk_gso_type = SKB_GSO_SCTP;
4558 /* Initialize default send parameters. These parameters can be
4559 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4561 sp->default_stream = 0;
4562 sp->default_ppid = 0;
4563 sp->default_flags = 0;
4564 sp->default_context = 0;
4565 sp->default_timetolive = 0;
4567 sp->default_rcv_context = 0;
4568 sp->max_burst = net->sctp.max_burst;
4570 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4572 /* Initialize default setup parameters. These parameters
4573 * can be modified with the SCTP_INITMSG socket option or
4574 * overridden by the SCTP_INIT CMSG.
4576 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4577 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4578 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4579 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4581 /* Initialize default RTO related parameters. These parameters can
4582 * be modified for with the SCTP_RTOINFO socket option.
4584 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4585 sp->rtoinfo.srto_max = net->sctp.rto_max;
4586 sp->rtoinfo.srto_min = net->sctp.rto_min;
4588 /* Initialize default association related parameters. These parameters
4589 * can be modified with the SCTP_ASSOCINFO socket option.
4591 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4592 sp->assocparams.sasoc_number_peer_destinations = 0;
4593 sp->assocparams.sasoc_peer_rwnd = 0;
4594 sp->assocparams.sasoc_local_rwnd = 0;
4595 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4597 /* Initialize default event subscriptions. By default, all the
4600 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4602 /* Default Peer Address Parameters. These defaults can
4603 * be modified via SCTP_PEER_ADDR_PARAMS
4605 sp->hbinterval = net->sctp.hb_interval;
4606 sp->pathmaxrxt = net->sctp.max_retrans_path;
4607 sp->pathmtu = 0; /* allow default discovery */
4608 sp->sackdelay = net->sctp.sack_timeout;
4610 sp->param_flags = SPP_HB_ENABLE |
4612 SPP_SACKDELAY_ENABLE;
4614 /* If enabled no SCTP message fragmentation will be performed.
4615 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4617 sp->disable_fragments = 0;
4619 /* Enable Nagle algorithm by default. */
4622 sp->recvrcvinfo = 0;
4623 sp->recvnxtinfo = 0;
4625 /* Enable by default. */
4628 /* Auto-close idle associations after the configured
4629 * number of seconds. A value of 0 disables this
4630 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4631 * for UDP-style sockets only.
4635 /* User specified fragmentation limit. */
4638 sp->adaptation_ind = 0;
4640 sp->pf = sctp_get_pf_specific(sk->sk_family);
4642 /* Control variables for partial data delivery. */
4643 atomic_set(&sp->pd_mode, 0);
4644 skb_queue_head_init(&sp->pd_lobby);
4645 sp->frag_interleave = 0;
4647 /* Create a per socket endpoint structure. Even if we
4648 * change the data structure relationships, this may still
4649 * be useful for storing pre-connect address information.
4651 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4657 sk->sk_destruct = sctp_destruct_sock;
4659 SCTP_DBG_OBJCNT_INC(sock);
4662 sk_sockets_allocated_inc(sk);
4663 sock_prot_inuse_add(net, sk->sk_prot, 1);
4665 /* Nothing can fail after this block, otherwise
4666 * sctp_destroy_sock() will be called without addr_wq_lock held
4668 if (net->sctp.default_auto_asconf) {
4669 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
4670 list_add_tail(&sp->auto_asconf_list,
4671 &net->sctp.auto_asconf_splist);
4672 sp->do_auto_asconf = 1;
4673 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4675 sp->do_auto_asconf = 0;
4683 /* Cleanup any SCTP per socket resources. Must be called with
4684 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4686 static void sctp_destroy_sock(struct sock *sk)
4688 struct sctp_sock *sp;
4690 pr_debug("%s: sk:%p\n", __func__, sk);
4692 /* Release our hold on the endpoint. */
4694 /* This could happen during socket init, thus we bail out
4695 * early, since the rest of the below is not setup either.
4700 if (sp->do_auto_asconf) {
4701 sp->do_auto_asconf = 0;
4702 list_del(&sp->auto_asconf_list);
4704 sctp_endpoint_free(sp->ep);
4706 sk_sockets_allocated_dec(sk);
4707 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4711 /* Triggered when there are no references on the socket anymore */
4712 static void sctp_destruct_sock(struct sock *sk)
4714 struct sctp_sock *sp = sctp_sk(sk);
4716 /* Free up the HMAC transform. */
4717 crypto_free_shash(sp->hmac);
4719 inet_sock_destruct(sk);
4722 /* API 4.1.7 shutdown() - TCP Style Syntax
4723 * int shutdown(int socket, int how);
4725 * sd - the socket descriptor of the association to be closed.
4726 * how - Specifies the type of shutdown. The values are
4729 * Disables further receive operations. No SCTP
4730 * protocol action is taken.
4732 * Disables further send operations, and initiates
4733 * the SCTP shutdown sequence.
4735 * Disables further send and receive operations
4736 * and initiates the SCTP shutdown sequence.
4738 static void sctp_shutdown(struct sock *sk, int how)
4740 struct net *net = sock_net(sk);
4741 struct sctp_endpoint *ep;
4743 if (!sctp_style(sk, TCP))
4746 ep = sctp_sk(sk)->ep;
4747 if (how & SEND_SHUTDOWN && !list_empty(&ep->asocs)) {
4748 struct sctp_association *asoc;
4750 inet_sk_set_state(sk, SCTP_SS_CLOSING);
4751 asoc = list_entry(ep->asocs.next,
4752 struct sctp_association, asocs);
4753 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4757 int sctp_get_sctp_info(struct sock *sk, struct sctp_association *asoc,
4758 struct sctp_info *info)
4760 struct sctp_transport *prim;
4761 struct list_head *pos;
4764 memset(info, 0, sizeof(*info));
4766 struct sctp_sock *sp = sctp_sk(sk);
4768 info->sctpi_s_autoclose = sp->autoclose;
4769 info->sctpi_s_adaptation_ind = sp->adaptation_ind;
4770 info->sctpi_s_pd_point = sp->pd_point;
4771 info->sctpi_s_nodelay = sp->nodelay;
4772 info->sctpi_s_disable_fragments = sp->disable_fragments;
4773 info->sctpi_s_v4mapped = sp->v4mapped;
4774 info->sctpi_s_frag_interleave = sp->frag_interleave;
4775 info->sctpi_s_type = sp->type;
4780 info->sctpi_tag = asoc->c.my_vtag;
4781 info->sctpi_state = asoc->state;
4782 info->sctpi_rwnd = asoc->a_rwnd;
4783 info->sctpi_unackdata = asoc->unack_data;
4784 info->sctpi_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4785 info->sctpi_instrms = asoc->stream.incnt;
4786 info->sctpi_outstrms = asoc->stream.outcnt;
4787 list_for_each(pos, &asoc->base.inqueue.in_chunk_list)
4788 info->sctpi_inqueue++;
4789 list_for_each(pos, &asoc->outqueue.out_chunk_list)
4790 info->sctpi_outqueue++;
4791 info->sctpi_overall_error = asoc->overall_error_count;
4792 info->sctpi_max_burst = asoc->max_burst;
4793 info->sctpi_maxseg = asoc->frag_point;
4794 info->sctpi_peer_rwnd = asoc->peer.rwnd;
4795 info->sctpi_peer_tag = asoc->c.peer_vtag;
4797 mask = asoc->peer.ecn_capable << 1;
4798 mask = (mask | asoc->peer.ipv4_address) << 1;
4799 mask = (mask | asoc->peer.ipv6_address) << 1;
4800 mask = (mask | asoc->peer.hostname_address) << 1;
4801 mask = (mask | asoc->peer.asconf_capable) << 1;
4802 mask = (mask | asoc->peer.prsctp_capable) << 1;
4803 mask = (mask | asoc->peer.auth_capable);
4804 info->sctpi_peer_capable = mask;
4805 mask = asoc->peer.sack_needed << 1;
4806 mask = (mask | asoc->peer.sack_generation) << 1;
4807 mask = (mask | asoc->peer.zero_window_announced);
4808 info->sctpi_peer_sack = mask;
4810 info->sctpi_isacks = asoc->stats.isacks;
4811 info->sctpi_osacks = asoc->stats.osacks;
4812 info->sctpi_opackets = asoc->stats.opackets;
4813 info->sctpi_ipackets = asoc->stats.ipackets;
4814 info->sctpi_rtxchunks = asoc->stats.rtxchunks;
4815 info->sctpi_outofseqtsns = asoc->stats.outofseqtsns;
4816 info->sctpi_idupchunks = asoc->stats.idupchunks;
4817 info->sctpi_gapcnt = asoc->stats.gapcnt;
4818 info->sctpi_ouodchunks = asoc->stats.ouodchunks;
4819 info->sctpi_iuodchunks = asoc->stats.iuodchunks;
4820 info->sctpi_oodchunks = asoc->stats.oodchunks;
4821 info->sctpi_iodchunks = asoc->stats.iodchunks;
4822 info->sctpi_octrlchunks = asoc->stats.octrlchunks;
4823 info->sctpi_ictrlchunks = asoc->stats.ictrlchunks;
4825 prim = asoc->peer.primary_path;
4826 memcpy(&info->sctpi_p_address, &prim->ipaddr, sizeof(prim->ipaddr));
4827 info->sctpi_p_state = prim->state;
4828 info->sctpi_p_cwnd = prim->cwnd;
4829 info->sctpi_p_srtt = prim->srtt;
4830 info->sctpi_p_rto = jiffies_to_msecs(prim->rto);
4831 info->sctpi_p_hbinterval = prim->hbinterval;
4832 info->sctpi_p_pathmaxrxt = prim->pathmaxrxt;
4833 info->sctpi_p_sackdelay = jiffies_to_msecs(prim->sackdelay);
4834 info->sctpi_p_ssthresh = prim->ssthresh;
4835 info->sctpi_p_partial_bytes_acked = prim->partial_bytes_acked;
4836 info->sctpi_p_flight_size = prim->flight_size;
4837 info->sctpi_p_error = prim->error_count;
4841 EXPORT_SYMBOL_GPL(sctp_get_sctp_info);
4843 /* use callback to avoid exporting the core structure */
4844 void sctp_transport_walk_start(struct rhashtable_iter *iter)
4846 rhltable_walk_enter(&sctp_transport_hashtable, iter);
4848 rhashtable_walk_start(iter);
4851 void sctp_transport_walk_stop(struct rhashtable_iter *iter)
4853 rhashtable_walk_stop(iter);
4854 rhashtable_walk_exit(iter);
4857 struct sctp_transport *sctp_transport_get_next(struct net *net,
4858 struct rhashtable_iter *iter)
4860 struct sctp_transport *t;
4862 t = rhashtable_walk_next(iter);
4863 for (; t; t = rhashtable_walk_next(iter)) {
4865 if (PTR_ERR(t) == -EAGAIN)
4870 if (net_eq(sock_net(t->asoc->base.sk), net) &&
4871 t->asoc->peer.primary_path == t)
4878 struct sctp_transport *sctp_transport_get_idx(struct net *net,
4879 struct rhashtable_iter *iter,
4882 void *obj = SEQ_START_TOKEN;
4884 while (pos && (obj = sctp_transport_get_next(net, iter)) &&
4891 int sctp_for_each_endpoint(int (*cb)(struct sctp_endpoint *, void *),
4895 struct sctp_ep_common *epb;
4896 struct sctp_hashbucket *head;
4898 for (head = sctp_ep_hashtable; hash < sctp_ep_hashsize;
4900 read_lock_bh(&head->lock);
4901 sctp_for_each_hentry(epb, &head->chain) {
4902 err = cb(sctp_ep(epb), p);
4906 read_unlock_bh(&head->lock);
4911 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint);
4913 int sctp_transport_lookup_process(int (*cb)(struct sctp_transport *, void *),
4915 const union sctp_addr *laddr,
4916 const union sctp_addr *paddr, void *p)
4918 struct sctp_transport *transport;
4922 transport = sctp_addrs_lookup_transport(net, laddr, paddr);
4927 err = cb(transport, p);
4928 sctp_transport_put(transport);
4932 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process);
4934 int sctp_for_each_transport(int (*cb)(struct sctp_transport *, void *),
4935 int (*cb_done)(struct sctp_transport *, void *),
4936 struct net *net, int *pos, void *p) {
4937 struct rhashtable_iter hti;
4938 struct sctp_transport *tsp;
4943 sctp_transport_walk_start(&hti);
4945 tsp = sctp_transport_get_idx(net, &hti, *pos + 1);
4946 for (; !IS_ERR_OR_NULL(tsp); tsp = sctp_transport_get_next(net, &hti)) {
4947 if (!sctp_transport_hold(tsp))
4953 sctp_transport_put(tsp);
4955 sctp_transport_walk_stop(&hti);
4958 if (cb_done && !cb_done(tsp, p)) {
4960 sctp_transport_put(tsp);
4963 sctp_transport_put(tsp);
4968 EXPORT_SYMBOL_GPL(sctp_for_each_transport);
4970 /* 7.2.1 Association Status (SCTP_STATUS)
4972 * Applications can retrieve current status information about an
4973 * association, including association state, peer receiver window size,
4974 * number of unacked data chunks, and number of data chunks pending
4975 * receipt. This information is read-only.
4977 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4978 char __user *optval,
4981 struct sctp_status status;
4982 struct sctp_association *asoc = NULL;
4983 struct sctp_transport *transport;
4984 sctp_assoc_t associd;
4987 if (len < sizeof(status)) {
4992 len = sizeof(status);
4993 if (copy_from_user(&status, optval, len)) {
4998 associd = status.sstat_assoc_id;
4999 asoc = sctp_id2assoc(sk, associd);
5005 transport = asoc->peer.primary_path;
5007 status.sstat_assoc_id = sctp_assoc2id(asoc);
5008 status.sstat_state = sctp_assoc_to_state(asoc);
5009 status.sstat_rwnd = asoc->peer.rwnd;
5010 status.sstat_unackdata = asoc->unack_data;
5012 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
5013 status.sstat_instrms = asoc->stream.incnt;
5014 status.sstat_outstrms = asoc->stream.outcnt;
5015 status.sstat_fragmentation_point = asoc->frag_point;
5016 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
5017 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
5018 transport->af_specific->sockaddr_len);
5019 /* Map ipv4 address into v4-mapped-on-v6 address. */
5020 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
5021 (union sctp_addr *)&status.sstat_primary.spinfo_address);
5022 status.sstat_primary.spinfo_state = transport->state;
5023 status.sstat_primary.spinfo_cwnd = transport->cwnd;
5024 status.sstat_primary.spinfo_srtt = transport->srtt;
5025 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
5026 status.sstat_primary.spinfo_mtu = transport->pathmtu;
5028 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
5029 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
5031 if (put_user(len, optlen)) {
5036 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
5037 __func__, len, status.sstat_state, status.sstat_rwnd,
5038 status.sstat_assoc_id);
5040 if (copy_to_user(optval, &status, len)) {
5050 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
5052 * Applications can retrieve information about a specific peer address
5053 * of an association, including its reachability state, congestion
5054 * window, and retransmission timer values. This information is
5057 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
5058 char __user *optval,
5061 struct sctp_paddrinfo pinfo;
5062 struct sctp_transport *transport;
5065 if (len < sizeof(pinfo)) {
5070 len = sizeof(pinfo);
5071 if (copy_from_user(&pinfo, optval, len)) {
5076 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
5077 pinfo.spinfo_assoc_id);
5081 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
5082 pinfo.spinfo_state = transport->state;
5083 pinfo.spinfo_cwnd = transport->cwnd;
5084 pinfo.spinfo_srtt = transport->srtt;
5085 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
5086 pinfo.spinfo_mtu = transport->pathmtu;
5088 if (pinfo.spinfo_state == SCTP_UNKNOWN)
5089 pinfo.spinfo_state = SCTP_ACTIVE;
5091 if (put_user(len, optlen)) {
5096 if (copy_to_user(optval, &pinfo, len)) {
5105 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
5107 * This option is a on/off flag. If enabled no SCTP message
5108 * fragmentation will be performed. Instead if a message being sent
5109 * exceeds the current PMTU size, the message will NOT be sent and
5110 * instead a error will be indicated to the user.
5112 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
5113 char __user *optval, int __user *optlen)
5117 if (len < sizeof(int))
5121 val = (sctp_sk(sk)->disable_fragments == 1);
5122 if (put_user(len, optlen))
5124 if (copy_to_user(optval, &val, len))
5129 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
5131 * This socket option is used to specify various notifications and
5132 * ancillary data the user wishes to receive.
5134 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
5139 if (len > sizeof(struct sctp_event_subscribe))
5140 len = sizeof(struct sctp_event_subscribe);
5141 if (put_user(len, optlen))
5143 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
5148 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
5150 * This socket option is applicable to the UDP-style socket only. When
5151 * set it will cause associations that are idle for more than the
5152 * specified number of seconds to automatically close. An association
5153 * being idle is defined an association that has NOT sent or received
5154 * user data. The special value of '0' indicates that no automatic
5155 * close of any associations should be performed. The option expects an
5156 * integer defining the number of seconds of idle time before an
5157 * association is closed.
5159 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
5161 /* Applicable to UDP-style socket only */
5162 if (sctp_style(sk, TCP))
5164 if (len < sizeof(int))
5167 if (put_user(len, optlen))
5169 if (put_user(sctp_sk(sk)->autoclose, (int __user *)optval))
5174 /* Helper routine to branch off an association to a new socket. */
5175 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
5177 struct sctp_association *asoc = sctp_id2assoc(sk, id);
5178 struct sctp_sock *sp = sctp_sk(sk);
5179 struct socket *sock;
5182 /* Do not peel off from one netns to another one. */
5183 if (!net_eq(current->nsproxy->net_ns, sock_net(sk)))
5189 /* An association cannot be branched off from an already peeled-off
5190 * socket, nor is this supported for tcp style sockets.
5192 if (!sctp_style(sk, UDP))
5195 /* Create a new socket. */
5196 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
5200 sctp_copy_sock(sock->sk, sk, asoc);
5202 /* Make peeled-off sockets more like 1-1 accepted sockets.
5203 * Set the daddr and initialize id to something more random and also
5204 * copy over any ip options.
5206 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
5207 sp->pf->copy_ip_options(sk, sock->sk);
5209 /* Populate the fields of the newsk from the oldsk and migrate the
5210 * asoc to the newsk.
5212 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
5218 EXPORT_SYMBOL(sctp_do_peeloff);
5220 static int sctp_getsockopt_peeloff_common(struct sock *sk, sctp_peeloff_arg_t *peeloff,
5221 struct file **newfile, unsigned flags)
5223 struct socket *newsock;
5226 retval = sctp_do_peeloff(sk, peeloff->associd, &newsock);
5230 /* Map the socket to an unused fd that can be returned to the user. */
5231 retval = get_unused_fd_flags(flags & SOCK_CLOEXEC);
5233 sock_release(newsock);
5237 *newfile = sock_alloc_file(newsock, 0, NULL);
5238 if (IS_ERR(*newfile)) {
5239 put_unused_fd(retval);
5240 retval = PTR_ERR(*newfile);
5245 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
5248 peeloff->sd = retval;
5250 if (flags & SOCK_NONBLOCK)
5251 (*newfile)->f_flags |= O_NONBLOCK;
5256 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
5258 sctp_peeloff_arg_t peeloff;
5259 struct file *newfile = NULL;
5262 if (len < sizeof(sctp_peeloff_arg_t))
5264 len = sizeof(sctp_peeloff_arg_t);
5265 if (copy_from_user(&peeloff, optval, len))
5268 retval = sctp_getsockopt_peeloff_common(sk, &peeloff, &newfile, 0);
5272 /* Return the fd mapped to the new socket. */
5273 if (put_user(len, optlen)) {
5275 put_unused_fd(retval);
5279 if (copy_to_user(optval, &peeloff, len)) {
5281 put_unused_fd(retval);
5284 fd_install(retval, newfile);
5289 static int sctp_getsockopt_peeloff_flags(struct sock *sk, int len,
5290 char __user *optval, int __user *optlen)
5292 sctp_peeloff_flags_arg_t peeloff;
5293 struct file *newfile = NULL;
5296 if (len < sizeof(sctp_peeloff_flags_arg_t))
5298 len = sizeof(sctp_peeloff_flags_arg_t);
5299 if (copy_from_user(&peeloff, optval, len))
5302 retval = sctp_getsockopt_peeloff_common(sk, &peeloff.p_arg,
5303 &newfile, peeloff.flags);
5307 /* Return the fd mapped to the new socket. */
5308 if (put_user(len, optlen)) {
5310 put_unused_fd(retval);
5314 if (copy_to_user(optval, &peeloff, len)) {
5316 put_unused_fd(retval);
5319 fd_install(retval, newfile);
5324 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
5326 * Applications can enable or disable heartbeats for any peer address of
5327 * an association, modify an address's heartbeat interval, force a
5328 * heartbeat to be sent immediately, and adjust the address's maximum
5329 * number of retransmissions sent before an address is considered
5330 * unreachable. The following structure is used to access and modify an
5331 * address's parameters:
5333 * struct sctp_paddrparams {
5334 * sctp_assoc_t spp_assoc_id;
5335 * struct sockaddr_storage spp_address;
5336 * uint32_t spp_hbinterval;
5337 * uint16_t spp_pathmaxrxt;
5338 * uint32_t spp_pathmtu;
5339 * uint32_t spp_sackdelay;
5340 * uint32_t spp_flags;
5343 * spp_assoc_id - (one-to-many style socket) This is filled in the
5344 * application, and identifies the association for
5346 * spp_address - This specifies which address is of interest.
5347 * spp_hbinterval - This contains the value of the heartbeat interval,
5348 * in milliseconds. If a value of zero
5349 * is present in this field then no changes are to
5350 * be made to this parameter.
5351 * spp_pathmaxrxt - This contains the maximum number of
5352 * retransmissions before this address shall be
5353 * considered unreachable. If a value of zero
5354 * is present in this field then no changes are to
5355 * be made to this parameter.
5356 * spp_pathmtu - When Path MTU discovery is disabled the value
5357 * specified here will be the "fixed" path mtu.
5358 * Note that if the spp_address field is empty
5359 * then all associations on this address will
5360 * have this fixed path mtu set upon them.
5362 * spp_sackdelay - When delayed sack is enabled, this value specifies
5363 * the number of milliseconds that sacks will be delayed
5364 * for. This value will apply to all addresses of an
5365 * association if the spp_address field is empty. Note
5366 * also, that if delayed sack is enabled and this
5367 * value is set to 0, no change is made to the last
5368 * recorded delayed sack timer value.
5370 * spp_flags - These flags are used to control various features
5371 * on an association. The flag field may contain
5372 * zero or more of the following options.
5374 * SPP_HB_ENABLE - Enable heartbeats on the
5375 * specified address. Note that if the address
5376 * field is empty all addresses for the association
5377 * have heartbeats enabled upon them.
5379 * SPP_HB_DISABLE - Disable heartbeats on the
5380 * speicifed address. Note that if the address
5381 * field is empty all addresses for the association
5382 * will have their heartbeats disabled. Note also
5383 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
5384 * mutually exclusive, only one of these two should
5385 * be specified. Enabling both fields will have
5386 * undetermined results.
5388 * SPP_HB_DEMAND - Request a user initiated heartbeat
5389 * to be made immediately.
5391 * SPP_PMTUD_ENABLE - This field will enable PMTU
5392 * discovery upon the specified address. Note that
5393 * if the address feild is empty then all addresses
5394 * on the association are effected.
5396 * SPP_PMTUD_DISABLE - This field will disable PMTU
5397 * discovery upon the specified address. Note that
5398 * if the address feild is empty then all addresses
5399 * on the association are effected. Not also that
5400 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
5401 * exclusive. Enabling both will have undetermined
5404 * SPP_SACKDELAY_ENABLE - Setting this flag turns
5405 * on delayed sack. The time specified in spp_sackdelay
5406 * is used to specify the sack delay for this address. Note
5407 * that if spp_address is empty then all addresses will
5408 * enable delayed sack and take on the sack delay
5409 * value specified in spp_sackdelay.
5410 * SPP_SACKDELAY_DISABLE - Setting this flag turns
5411 * off delayed sack. If the spp_address field is blank then
5412 * delayed sack is disabled for the entire association. Note
5413 * also that this field is mutually exclusive to
5414 * SPP_SACKDELAY_ENABLE, setting both will have undefined
5417 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
5418 char __user *optval, int __user *optlen)
5420 struct sctp_paddrparams params;
5421 struct sctp_transport *trans = NULL;
5422 struct sctp_association *asoc = NULL;
5423 struct sctp_sock *sp = sctp_sk(sk);
5425 if (len < sizeof(struct sctp_paddrparams))
5427 len = sizeof(struct sctp_paddrparams);
5428 if (copy_from_user(¶ms, optval, len))
5431 /* If an address other than INADDR_ANY is specified, and
5432 * no transport is found, then the request is invalid.
5434 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
5435 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
5436 params.spp_assoc_id);
5438 pr_debug("%s: failed no transport\n", __func__);
5443 /* Get association, if assoc_id != 0 and the socket is a one
5444 * to many style socket, and an association was not found, then
5445 * the id was invalid.
5447 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
5448 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
5449 pr_debug("%s: failed no association\n", __func__);
5454 /* Fetch transport values. */
5455 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
5456 params.spp_pathmtu = trans->pathmtu;
5457 params.spp_pathmaxrxt = trans->pathmaxrxt;
5458 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
5460 /*draft-11 doesn't say what to return in spp_flags*/
5461 params.spp_flags = trans->param_flags;
5463 /* Fetch association values. */
5464 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
5465 params.spp_pathmtu = asoc->pathmtu;
5466 params.spp_pathmaxrxt = asoc->pathmaxrxt;
5467 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
5469 /*draft-11 doesn't say what to return in spp_flags*/
5470 params.spp_flags = asoc->param_flags;
5472 /* Fetch socket values. */
5473 params.spp_hbinterval = sp->hbinterval;
5474 params.spp_pathmtu = sp->pathmtu;
5475 params.spp_sackdelay = sp->sackdelay;
5476 params.spp_pathmaxrxt = sp->pathmaxrxt;
5478 /*draft-11 doesn't say what to return in spp_flags*/
5479 params.spp_flags = sp->param_flags;
5482 if (copy_to_user(optval, ¶ms, len))
5485 if (put_user(len, optlen))
5492 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
5494 * This option will effect the way delayed acks are performed. This
5495 * option allows you to get or set the delayed ack time, in
5496 * milliseconds. It also allows changing the delayed ack frequency.
5497 * Changing the frequency to 1 disables the delayed sack algorithm. If
5498 * the assoc_id is 0, then this sets or gets the endpoints default
5499 * values. If the assoc_id field is non-zero, then the set or get
5500 * effects the specified association for the one to many model (the
5501 * assoc_id field is ignored by the one to one model). Note that if
5502 * sack_delay or sack_freq are 0 when setting this option, then the
5503 * current values will remain unchanged.
5505 * struct sctp_sack_info {
5506 * sctp_assoc_t sack_assoc_id;
5507 * uint32_t sack_delay;
5508 * uint32_t sack_freq;
5511 * sack_assoc_id - This parameter, indicates which association the user
5512 * is performing an action upon. Note that if this field's value is
5513 * zero then the endpoints default value is changed (effecting future
5514 * associations only).
5516 * sack_delay - This parameter contains the number of milliseconds that
5517 * the user is requesting the delayed ACK timer be set to. Note that
5518 * this value is defined in the standard to be between 200 and 500
5521 * sack_freq - This parameter contains the number of packets that must
5522 * be received before a sack is sent without waiting for the delay
5523 * timer to expire. The default value for this is 2, setting this
5524 * value to 1 will disable the delayed sack algorithm.
5526 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
5527 char __user *optval,
5530 struct sctp_sack_info params;
5531 struct sctp_association *asoc = NULL;
5532 struct sctp_sock *sp = sctp_sk(sk);
5534 if (len >= sizeof(struct sctp_sack_info)) {
5535 len = sizeof(struct sctp_sack_info);
5537 if (copy_from_user(¶ms, optval, len))
5539 } else if (len == sizeof(struct sctp_assoc_value)) {
5540 pr_warn_ratelimited(DEPRECATED
5542 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
5543 "Use struct sctp_sack_info instead\n",
5544 current->comm, task_pid_nr(current));
5545 if (copy_from_user(¶ms, optval, len))
5550 /* Get association, if sack_assoc_id != 0 and the socket is a one
5551 * to many style socket, and an association was not found, then
5552 * the id was invalid.
5554 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
5555 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
5559 /* Fetch association values. */
5560 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
5561 params.sack_delay = jiffies_to_msecs(
5563 params.sack_freq = asoc->sackfreq;
5566 params.sack_delay = 0;
5567 params.sack_freq = 1;
5570 /* Fetch socket values. */
5571 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
5572 params.sack_delay = sp->sackdelay;
5573 params.sack_freq = sp->sackfreq;
5575 params.sack_delay = 0;
5576 params.sack_freq = 1;
5580 if (copy_to_user(optval, ¶ms, len))
5583 if (put_user(len, optlen))
5589 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5591 * Applications can specify protocol parameters for the default association
5592 * initialization. The option name argument to setsockopt() and getsockopt()
5595 * Setting initialization parameters is effective only on an unconnected
5596 * socket (for UDP-style sockets only future associations are effected
5597 * by the change). With TCP-style sockets, this option is inherited by
5598 * sockets derived from a listener socket.
5600 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
5602 if (len < sizeof(struct sctp_initmsg))
5604 len = sizeof(struct sctp_initmsg);
5605 if (put_user(len, optlen))
5607 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
5613 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
5614 char __user *optval, int __user *optlen)
5616 struct sctp_association *asoc;
5618 struct sctp_getaddrs getaddrs;
5619 struct sctp_transport *from;
5621 union sctp_addr temp;
5622 struct sctp_sock *sp = sctp_sk(sk);
5627 if (len < sizeof(struct sctp_getaddrs))
5630 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5633 /* For UDP-style sockets, id specifies the association to query. */
5634 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5638 to = optval + offsetof(struct sctp_getaddrs, addrs);
5639 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5641 list_for_each_entry(from, &asoc->peer.transport_addr_list,
5643 memcpy(&temp, &from->ipaddr, sizeof(temp));
5644 addrlen = sctp_get_pf_specific(sk->sk_family)
5645 ->addr_to_user(sp, &temp);
5646 if (space_left < addrlen)
5648 if (copy_to_user(to, &temp, addrlen))
5652 space_left -= addrlen;
5655 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
5657 bytes_copied = ((char __user *)to) - optval;
5658 if (put_user(bytes_copied, optlen))
5664 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
5665 size_t space_left, int *bytes_copied)
5667 struct sctp_sockaddr_entry *addr;
5668 union sctp_addr temp;
5671 struct net *net = sock_net(sk);
5674 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
5678 if ((PF_INET == sk->sk_family) &&
5679 (AF_INET6 == addr->a.sa.sa_family))
5681 if ((PF_INET6 == sk->sk_family) &&
5682 inet_v6_ipv6only(sk) &&
5683 (AF_INET == addr->a.sa.sa_family))
5685 memcpy(&temp, &addr->a, sizeof(temp));
5686 if (!temp.v4.sin_port)
5687 temp.v4.sin_port = htons(port);
5689 addrlen = sctp_get_pf_specific(sk->sk_family)
5690 ->addr_to_user(sctp_sk(sk), &temp);
5692 if (space_left < addrlen) {
5696 memcpy(to, &temp, addrlen);
5700 space_left -= addrlen;
5701 *bytes_copied += addrlen;
5709 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
5710 char __user *optval, int __user *optlen)
5712 struct sctp_bind_addr *bp;
5713 struct sctp_association *asoc;
5715 struct sctp_getaddrs getaddrs;
5716 struct sctp_sockaddr_entry *addr;
5718 union sctp_addr temp;
5719 struct sctp_sock *sp = sctp_sk(sk);
5723 int bytes_copied = 0;
5727 if (len < sizeof(struct sctp_getaddrs))
5730 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5734 * For UDP-style sockets, id specifies the association to query.
5735 * If the id field is set to the value '0' then the locally bound
5736 * addresses are returned without regard to any particular
5739 if (0 == getaddrs.assoc_id) {
5740 bp = &sctp_sk(sk)->ep->base.bind_addr;
5742 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5745 bp = &asoc->base.bind_addr;
5748 to = optval + offsetof(struct sctp_getaddrs, addrs);
5749 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5751 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN);
5755 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5756 * addresses from the global local address list.
5758 if (sctp_list_single_entry(&bp->address_list)) {
5759 addr = list_entry(bp->address_list.next,
5760 struct sctp_sockaddr_entry, list);
5761 if (sctp_is_any(sk, &addr->a)) {
5762 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
5763 space_left, &bytes_copied);
5773 /* Protection on the bound address list is not needed since
5774 * in the socket option context we hold a socket lock and
5775 * thus the bound address list can't change.
5777 list_for_each_entry(addr, &bp->address_list, list) {
5778 memcpy(&temp, &addr->a, sizeof(temp));
5779 addrlen = sctp_get_pf_specific(sk->sk_family)
5780 ->addr_to_user(sp, &temp);
5781 if (space_left < addrlen) {
5782 err = -ENOMEM; /*fixme: right error?*/
5785 memcpy(buf, &temp, addrlen);
5787 bytes_copied += addrlen;
5789 space_left -= addrlen;
5793 if (copy_to_user(to, addrs, bytes_copied)) {
5797 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
5801 /* XXX: We should have accounted for sizeof(struct sctp_getaddrs) too,
5802 * but we can't change it anymore.
5804 if (put_user(bytes_copied, optlen))
5811 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
5813 * Requests that the local SCTP stack use the enclosed peer address as
5814 * the association primary. The enclosed address must be one of the
5815 * association peer's addresses.
5817 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
5818 char __user *optval, int __user *optlen)
5820 struct sctp_prim prim;
5821 struct sctp_association *asoc;
5822 struct sctp_sock *sp = sctp_sk(sk);
5824 if (len < sizeof(struct sctp_prim))
5827 len = sizeof(struct sctp_prim);
5829 if (copy_from_user(&prim, optval, len))
5832 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
5836 if (!asoc->peer.primary_path)
5839 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
5840 asoc->peer.primary_path->af_specific->sockaddr_len);
5842 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
5843 (union sctp_addr *)&prim.ssp_addr);
5845 if (put_user(len, optlen))
5847 if (copy_to_user(optval, &prim, len))
5854 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5856 * Requests that the local endpoint set the specified Adaptation Layer
5857 * Indication parameter for all future INIT and INIT-ACK exchanges.
5859 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
5860 char __user *optval, int __user *optlen)
5862 struct sctp_setadaptation adaptation;
5864 if (len < sizeof(struct sctp_setadaptation))
5867 len = sizeof(struct sctp_setadaptation);
5869 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
5871 if (put_user(len, optlen))
5873 if (copy_to_user(optval, &adaptation, len))
5881 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5883 * Applications that wish to use the sendto() system call may wish to
5884 * specify a default set of parameters that would normally be supplied
5885 * through the inclusion of ancillary data. This socket option allows
5886 * such an application to set the default sctp_sndrcvinfo structure.
5889 * The application that wishes to use this socket option simply passes
5890 * in to this call the sctp_sndrcvinfo structure defined in Section
5891 * 5.2.2) The input parameters accepted by this call include
5892 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5893 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5894 * to this call if the caller is using the UDP model.
5896 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5898 static int sctp_getsockopt_default_send_param(struct sock *sk,
5899 int len, char __user *optval,
5902 struct sctp_sock *sp = sctp_sk(sk);
5903 struct sctp_association *asoc;
5904 struct sctp_sndrcvinfo info;
5906 if (len < sizeof(info))
5911 if (copy_from_user(&info, optval, len))
5914 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
5915 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
5918 info.sinfo_stream = asoc->default_stream;
5919 info.sinfo_flags = asoc->default_flags;
5920 info.sinfo_ppid = asoc->default_ppid;
5921 info.sinfo_context = asoc->default_context;
5922 info.sinfo_timetolive = asoc->default_timetolive;
5924 info.sinfo_stream = sp->default_stream;
5925 info.sinfo_flags = sp->default_flags;
5926 info.sinfo_ppid = sp->default_ppid;
5927 info.sinfo_context = sp->default_context;
5928 info.sinfo_timetolive = sp->default_timetolive;
5931 if (put_user(len, optlen))
5933 if (copy_to_user(optval, &info, len))
5939 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5940 * (SCTP_DEFAULT_SNDINFO)
5942 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
5943 char __user *optval,
5946 struct sctp_sock *sp = sctp_sk(sk);
5947 struct sctp_association *asoc;
5948 struct sctp_sndinfo info;
5950 if (len < sizeof(info))
5955 if (copy_from_user(&info, optval, len))
5958 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
5959 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
5962 info.snd_sid = asoc->default_stream;
5963 info.snd_flags = asoc->default_flags;
5964 info.snd_ppid = asoc->default_ppid;
5965 info.snd_context = asoc->default_context;
5967 info.snd_sid = sp->default_stream;
5968 info.snd_flags = sp->default_flags;
5969 info.snd_ppid = sp->default_ppid;
5970 info.snd_context = sp->default_context;
5973 if (put_user(len, optlen))
5975 if (copy_to_user(optval, &info, len))
5983 * 7.1.5 SCTP_NODELAY
5985 * Turn on/off any Nagle-like algorithm. This means that packets are
5986 * generally sent as soon as possible and no unnecessary delays are
5987 * introduced, at the cost of more packets in the network. Expects an
5988 * integer boolean flag.
5991 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5992 char __user *optval, int __user *optlen)
5996 if (len < sizeof(int))
6000 val = (sctp_sk(sk)->nodelay == 1);
6001 if (put_user(len, optlen))
6003 if (copy_to_user(optval, &val, len))
6010 * 7.1.1 SCTP_RTOINFO
6012 * The protocol parameters used to initialize and bound retransmission
6013 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
6014 * and modify these parameters.
6015 * All parameters are time values, in milliseconds. A value of 0, when
6016 * modifying the parameters, indicates that the current value should not
6020 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
6021 char __user *optval,
6022 int __user *optlen) {
6023 struct sctp_rtoinfo rtoinfo;
6024 struct sctp_association *asoc;
6026 if (len < sizeof (struct sctp_rtoinfo))
6029 len = sizeof(struct sctp_rtoinfo);
6031 if (copy_from_user(&rtoinfo, optval, len))
6034 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
6036 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
6039 /* Values corresponding to the specific association. */
6041 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
6042 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
6043 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
6045 /* Values corresponding to the endpoint. */
6046 struct sctp_sock *sp = sctp_sk(sk);
6048 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
6049 rtoinfo.srto_max = sp->rtoinfo.srto_max;
6050 rtoinfo.srto_min = sp->rtoinfo.srto_min;
6053 if (put_user(len, optlen))
6056 if (copy_to_user(optval, &rtoinfo, len))
6064 * 7.1.2 SCTP_ASSOCINFO
6066 * This option is used to tune the maximum retransmission attempts
6067 * of the association.
6068 * Returns an error if the new association retransmission value is
6069 * greater than the sum of the retransmission value of the peer.
6070 * See [SCTP] for more information.
6073 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
6074 char __user *optval,
6078 struct sctp_assocparams assocparams;
6079 struct sctp_association *asoc;
6080 struct list_head *pos;
6083 if (len < sizeof (struct sctp_assocparams))
6086 len = sizeof(struct sctp_assocparams);
6088 if (copy_from_user(&assocparams, optval, len))
6091 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
6093 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
6096 /* Values correspoinding to the specific association */
6098 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
6099 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
6100 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
6101 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
6103 list_for_each(pos, &asoc->peer.transport_addr_list) {
6107 assocparams.sasoc_number_peer_destinations = cnt;
6109 /* Values corresponding to the endpoint */
6110 struct sctp_sock *sp = sctp_sk(sk);
6112 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
6113 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
6114 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
6115 assocparams.sasoc_cookie_life =
6116 sp->assocparams.sasoc_cookie_life;
6117 assocparams.sasoc_number_peer_destinations =
6119 sasoc_number_peer_destinations;
6122 if (put_user(len, optlen))
6125 if (copy_to_user(optval, &assocparams, len))
6132 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
6134 * This socket option is a boolean flag which turns on or off mapped V4
6135 * addresses. If this option is turned on and the socket is type
6136 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
6137 * If this option is turned off, then no mapping will be done of V4
6138 * addresses and a user will receive both PF_INET6 and PF_INET type
6139 * addresses on the socket.
6141 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
6142 char __user *optval, int __user *optlen)
6145 struct sctp_sock *sp = sctp_sk(sk);
6147 if (len < sizeof(int))
6152 if (put_user(len, optlen))
6154 if (copy_to_user(optval, &val, len))
6161 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
6162 * (chapter and verse is quoted at sctp_setsockopt_context())
6164 static int sctp_getsockopt_context(struct sock *sk, int len,
6165 char __user *optval, int __user *optlen)
6167 struct sctp_assoc_value params;
6168 struct sctp_sock *sp;
6169 struct sctp_association *asoc;
6171 if (len < sizeof(struct sctp_assoc_value))
6174 len = sizeof(struct sctp_assoc_value);
6176 if (copy_from_user(¶ms, optval, len))
6181 if (params.assoc_id != 0) {
6182 asoc = sctp_id2assoc(sk, params.assoc_id);
6185 params.assoc_value = asoc->default_rcv_context;
6187 params.assoc_value = sp->default_rcv_context;
6190 if (put_user(len, optlen))
6192 if (copy_to_user(optval, ¶ms, len))
6199 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
6200 * This option will get or set the maximum size to put in any outgoing
6201 * SCTP DATA chunk. If a message is larger than this size it will be
6202 * fragmented by SCTP into the specified size. Note that the underlying
6203 * SCTP implementation may fragment into smaller sized chunks when the
6204 * PMTU of the underlying association is smaller than the value set by
6205 * the user. The default value for this option is '0' which indicates
6206 * the user is NOT limiting fragmentation and only the PMTU will effect
6207 * SCTP's choice of DATA chunk size. Note also that values set larger
6208 * than the maximum size of an IP datagram will effectively let SCTP
6209 * control fragmentation (i.e. the same as setting this option to 0).
6211 * The following structure is used to access and modify this parameter:
6213 * struct sctp_assoc_value {
6214 * sctp_assoc_t assoc_id;
6215 * uint32_t assoc_value;
6218 * assoc_id: This parameter is ignored for one-to-one style sockets.
6219 * For one-to-many style sockets this parameter indicates which
6220 * association the user is performing an action upon. Note that if
6221 * this field's value is zero then the endpoints default value is
6222 * changed (effecting future associations only).
6223 * assoc_value: This parameter specifies the maximum size in bytes.
6225 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
6226 char __user *optval, int __user *optlen)
6228 struct sctp_assoc_value params;
6229 struct sctp_association *asoc;
6231 if (len == sizeof(int)) {
6232 pr_warn_ratelimited(DEPRECATED
6234 "Use of int in maxseg socket option.\n"
6235 "Use struct sctp_assoc_value instead\n",
6236 current->comm, task_pid_nr(current));
6237 params.assoc_id = 0;
6238 } else if (len >= sizeof(struct sctp_assoc_value)) {
6239 len = sizeof(struct sctp_assoc_value);
6240 if (copy_from_user(¶ms, optval, len))
6245 asoc = sctp_id2assoc(sk, params.assoc_id);
6246 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
6250 params.assoc_value = asoc->frag_point;
6252 params.assoc_value = sctp_sk(sk)->user_frag;
6254 if (put_user(len, optlen))
6256 if (len == sizeof(int)) {
6257 if (copy_to_user(optval, ¶ms.assoc_value, len))
6260 if (copy_to_user(optval, ¶ms, len))
6268 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
6269 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
6271 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
6272 char __user *optval, int __user *optlen)
6276 if (len < sizeof(int))
6281 val = sctp_sk(sk)->frag_interleave;
6282 if (put_user(len, optlen))
6284 if (copy_to_user(optval, &val, len))
6291 * 7.1.25. Set or Get the sctp partial delivery point
6292 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
6294 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
6295 char __user *optval,
6300 if (len < sizeof(u32))
6305 val = sctp_sk(sk)->pd_point;
6306 if (put_user(len, optlen))
6308 if (copy_to_user(optval, &val, len))
6315 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
6316 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
6318 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
6319 char __user *optval,
6322 struct sctp_assoc_value params;
6323 struct sctp_sock *sp;
6324 struct sctp_association *asoc;
6326 if (len == sizeof(int)) {
6327 pr_warn_ratelimited(DEPRECATED
6329 "Use of int in max_burst socket option.\n"
6330 "Use struct sctp_assoc_value instead\n",
6331 current->comm, task_pid_nr(current));
6332 params.assoc_id = 0;
6333 } else if (len >= sizeof(struct sctp_assoc_value)) {
6334 len = sizeof(struct sctp_assoc_value);
6335 if (copy_from_user(¶ms, optval, len))
6342 if (params.assoc_id != 0) {
6343 asoc = sctp_id2assoc(sk, params.assoc_id);
6346 params.assoc_value = asoc->max_burst;
6348 params.assoc_value = sp->max_burst;
6350 if (len == sizeof(int)) {
6351 if (copy_to_user(optval, ¶ms.assoc_value, len))
6354 if (copy_to_user(optval, ¶ms, len))
6362 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
6363 char __user *optval, int __user *optlen)
6365 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6366 struct sctp_hmacalgo __user *p = (void __user *)optval;
6367 struct sctp_hmac_algo_param *hmacs;
6372 if (!ep->auth_enable)
6375 hmacs = ep->auth_hmacs_list;
6376 data_len = ntohs(hmacs->param_hdr.length) -
6377 sizeof(struct sctp_paramhdr);
6379 if (len < sizeof(struct sctp_hmacalgo) + data_len)
6382 len = sizeof(struct sctp_hmacalgo) + data_len;
6383 num_idents = data_len / sizeof(u16);
6385 if (put_user(len, optlen))
6387 if (put_user(num_idents, &p->shmac_num_idents))
6389 for (i = 0; i < num_idents; i++) {
6390 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
6392 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
6398 static int sctp_getsockopt_active_key(struct sock *sk, int len,
6399 char __user *optval, int __user *optlen)
6401 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6402 struct sctp_authkeyid val;
6403 struct sctp_association *asoc;
6405 if (!ep->auth_enable)
6408 if (len < sizeof(struct sctp_authkeyid))
6411 len = sizeof(struct sctp_authkeyid);
6412 if (copy_from_user(&val, optval, len))
6415 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
6416 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
6420 val.scact_keynumber = asoc->active_key_id;
6422 val.scact_keynumber = ep->active_key_id;
6424 if (put_user(len, optlen))
6426 if (copy_to_user(optval, &val, len))
6432 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
6433 char __user *optval, int __user *optlen)
6435 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6436 struct sctp_authchunks __user *p = (void __user *)optval;
6437 struct sctp_authchunks val;
6438 struct sctp_association *asoc;
6439 struct sctp_chunks_param *ch;
6443 if (!ep->auth_enable)
6446 if (len < sizeof(struct sctp_authchunks))
6449 if (copy_from_user(&val, optval, sizeof(val)))
6452 to = p->gauth_chunks;
6453 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6457 ch = asoc->peer.peer_chunks;
6461 /* See if the user provided enough room for all the data */
6462 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6463 if (len < num_chunks)
6466 if (copy_to_user(to, ch->chunks, num_chunks))
6469 len = sizeof(struct sctp_authchunks) + num_chunks;
6470 if (put_user(len, optlen))
6472 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6477 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
6478 char __user *optval, int __user *optlen)
6480 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6481 struct sctp_authchunks __user *p = (void __user *)optval;
6482 struct sctp_authchunks val;
6483 struct sctp_association *asoc;
6484 struct sctp_chunks_param *ch;
6488 if (!ep->auth_enable)
6491 if (len < sizeof(struct sctp_authchunks))
6494 if (copy_from_user(&val, optval, sizeof(val)))
6497 to = p->gauth_chunks;
6498 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6499 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
6503 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
6505 ch = ep->auth_chunk_list;
6510 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6511 if (len < sizeof(struct sctp_authchunks) + num_chunks)
6514 if (copy_to_user(to, ch->chunks, num_chunks))
6517 len = sizeof(struct sctp_authchunks) + num_chunks;
6518 if (put_user(len, optlen))
6520 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6527 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
6528 * This option gets the current number of associations that are attached
6529 * to a one-to-many style socket. The option value is an uint32_t.
6531 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
6532 char __user *optval, int __user *optlen)
6534 struct sctp_sock *sp = sctp_sk(sk);
6535 struct sctp_association *asoc;
6538 if (sctp_style(sk, TCP))
6541 if (len < sizeof(u32))
6546 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6550 if (put_user(len, optlen))
6552 if (copy_to_user(optval, &val, len))
6559 * 8.1.23 SCTP_AUTO_ASCONF
6560 * See the corresponding setsockopt entry as description
6562 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
6563 char __user *optval, int __user *optlen)
6567 if (len < sizeof(int))
6571 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
6573 if (put_user(len, optlen))
6575 if (copy_to_user(optval, &val, len))
6581 * 8.2.6. Get the Current Identifiers of Associations
6582 * (SCTP_GET_ASSOC_ID_LIST)
6584 * This option gets the current list of SCTP association identifiers of
6585 * the SCTP associations handled by a one-to-many style socket.
6587 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
6588 char __user *optval, int __user *optlen)
6590 struct sctp_sock *sp = sctp_sk(sk);
6591 struct sctp_association *asoc;
6592 struct sctp_assoc_ids *ids;
6595 if (sctp_style(sk, TCP))
6598 if (len < sizeof(struct sctp_assoc_ids))
6601 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6605 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
6608 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
6610 ids = kmalloc(len, GFP_USER | __GFP_NOWARN);
6614 ids->gaids_number_of_ids = num;
6616 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6617 ids->gaids_assoc_id[num++] = asoc->assoc_id;
6620 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
6630 * SCTP_PEER_ADDR_THLDS
6632 * This option allows us to fetch the partially failed threshold for one or all
6633 * transports in an association. See Section 6.1 of:
6634 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6636 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
6637 char __user *optval,
6641 struct sctp_paddrthlds val;
6642 struct sctp_transport *trans;
6643 struct sctp_association *asoc;
6645 if (len < sizeof(struct sctp_paddrthlds))
6647 len = sizeof(struct sctp_paddrthlds);
6648 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
6651 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
6652 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
6656 val.spt_pathpfthld = asoc->pf_retrans;
6657 val.spt_pathmaxrxt = asoc->pathmaxrxt;
6659 trans = sctp_addr_id2transport(sk, &val.spt_address,
6664 val.spt_pathmaxrxt = trans->pathmaxrxt;
6665 val.spt_pathpfthld = trans->pf_retrans;
6668 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
6675 * SCTP_GET_ASSOC_STATS
6677 * This option retrieves local per endpoint statistics. It is modeled
6678 * after OpenSolaris' implementation
6680 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
6681 char __user *optval,
6684 struct sctp_assoc_stats sas;
6685 struct sctp_association *asoc = NULL;
6687 /* User must provide at least the assoc id */
6688 if (len < sizeof(sctp_assoc_t))
6691 /* Allow the struct to grow and fill in as much as possible */
6692 len = min_t(size_t, len, sizeof(sas));
6694 if (copy_from_user(&sas, optval, len))
6697 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
6701 sas.sas_rtxchunks = asoc->stats.rtxchunks;
6702 sas.sas_gapcnt = asoc->stats.gapcnt;
6703 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
6704 sas.sas_osacks = asoc->stats.osacks;
6705 sas.sas_isacks = asoc->stats.isacks;
6706 sas.sas_octrlchunks = asoc->stats.octrlchunks;
6707 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
6708 sas.sas_oodchunks = asoc->stats.oodchunks;
6709 sas.sas_iodchunks = asoc->stats.iodchunks;
6710 sas.sas_ouodchunks = asoc->stats.ouodchunks;
6711 sas.sas_iuodchunks = asoc->stats.iuodchunks;
6712 sas.sas_idupchunks = asoc->stats.idupchunks;
6713 sas.sas_opackets = asoc->stats.opackets;
6714 sas.sas_ipackets = asoc->stats.ipackets;
6716 /* New high max rto observed, will return 0 if not a single
6717 * RTO update took place. obs_rto_ipaddr will be bogus
6720 sas.sas_maxrto = asoc->stats.max_obs_rto;
6721 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
6722 sizeof(struct sockaddr_storage));
6724 /* Mark beginning of a new observation period */
6725 asoc->stats.max_obs_rto = asoc->rto_min;
6727 if (put_user(len, optlen))
6730 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
6732 if (copy_to_user(optval, &sas, len))
6738 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
6739 char __user *optval,
6744 if (len < sizeof(int))
6748 if (sctp_sk(sk)->recvrcvinfo)
6750 if (put_user(len, optlen))
6752 if (copy_to_user(optval, &val, len))
6758 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
6759 char __user *optval,
6764 if (len < sizeof(int))
6768 if (sctp_sk(sk)->recvnxtinfo)
6770 if (put_user(len, optlen))
6772 if (copy_to_user(optval, &val, len))
6778 static int sctp_getsockopt_pr_supported(struct sock *sk, int len,
6779 char __user *optval,
6782 struct sctp_assoc_value params;
6783 struct sctp_association *asoc;
6784 int retval = -EFAULT;
6786 if (len < sizeof(params)) {
6791 len = sizeof(params);
6792 if (copy_from_user(¶ms, optval, len))
6795 asoc = sctp_id2assoc(sk, params.assoc_id);
6797 params.assoc_value = asoc->prsctp_enable;
6798 } else if (!params.assoc_id) {
6799 struct sctp_sock *sp = sctp_sk(sk);
6801 params.assoc_value = sp->ep->prsctp_enable;
6807 if (put_user(len, optlen))
6810 if (copy_to_user(optval, ¶ms, len))
6819 static int sctp_getsockopt_default_prinfo(struct sock *sk, int len,
6820 char __user *optval,
6823 struct sctp_default_prinfo info;
6824 struct sctp_association *asoc;
6825 int retval = -EFAULT;
6827 if (len < sizeof(info)) {
6833 if (copy_from_user(&info, optval, len))
6836 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
6838 info.pr_policy = SCTP_PR_POLICY(asoc->default_flags);
6839 info.pr_value = asoc->default_timetolive;
6840 } else if (!info.pr_assoc_id) {
6841 struct sctp_sock *sp = sctp_sk(sk);
6843 info.pr_policy = SCTP_PR_POLICY(sp->default_flags);
6844 info.pr_value = sp->default_timetolive;
6850 if (put_user(len, optlen))
6853 if (copy_to_user(optval, &info, len))
6862 static int sctp_getsockopt_pr_assocstatus(struct sock *sk, int len,
6863 char __user *optval,
6866 struct sctp_prstatus params;
6867 struct sctp_association *asoc;
6869 int retval = -EINVAL;
6871 if (len < sizeof(params))
6874 len = sizeof(params);
6875 if (copy_from_user(¶ms, optval, len)) {
6880 policy = params.sprstat_policy;
6881 if (policy & ~SCTP_PR_SCTP_MASK)
6884 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
6888 if (policy == SCTP_PR_SCTP_NONE) {
6889 params.sprstat_abandoned_unsent = 0;
6890 params.sprstat_abandoned_sent = 0;
6891 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
6892 params.sprstat_abandoned_unsent +=
6893 asoc->abandoned_unsent[policy];
6894 params.sprstat_abandoned_sent +=
6895 asoc->abandoned_sent[policy];
6898 params.sprstat_abandoned_unsent =
6899 asoc->abandoned_unsent[__SCTP_PR_INDEX(policy)];
6900 params.sprstat_abandoned_sent =
6901 asoc->abandoned_sent[__SCTP_PR_INDEX(policy)];
6904 if (put_user(len, optlen)) {
6909 if (copy_to_user(optval, ¶ms, len)) {
6920 static int sctp_getsockopt_pr_streamstatus(struct sock *sk, int len,
6921 char __user *optval,
6924 struct sctp_stream_out_ext *streamoute;
6925 struct sctp_association *asoc;
6926 struct sctp_prstatus params;
6927 int retval = -EINVAL;
6930 if (len < sizeof(params))
6933 len = sizeof(params);
6934 if (copy_from_user(¶ms, optval, len)) {
6939 policy = params.sprstat_policy;
6940 if (policy & ~SCTP_PR_SCTP_MASK)
6943 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
6944 if (!asoc || params.sprstat_sid >= asoc->stream.outcnt)
6947 streamoute = asoc->stream.out[params.sprstat_sid].ext;
6949 /* Not allocated yet, means all stats are 0 */
6950 params.sprstat_abandoned_unsent = 0;
6951 params.sprstat_abandoned_sent = 0;
6956 if (policy == SCTP_PR_SCTP_NONE) {
6957 params.sprstat_abandoned_unsent = 0;
6958 params.sprstat_abandoned_sent = 0;
6959 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
6960 params.sprstat_abandoned_unsent +=
6961 streamoute->abandoned_unsent[policy];
6962 params.sprstat_abandoned_sent +=
6963 streamoute->abandoned_sent[policy];
6966 params.sprstat_abandoned_unsent =
6967 streamoute->abandoned_unsent[__SCTP_PR_INDEX(policy)];
6968 params.sprstat_abandoned_sent =
6969 streamoute->abandoned_sent[__SCTP_PR_INDEX(policy)];
6972 if (put_user(len, optlen) || copy_to_user(optval, ¶ms, len)) {
6983 static int sctp_getsockopt_reconfig_supported(struct sock *sk, int len,
6984 char __user *optval,
6987 struct sctp_assoc_value params;
6988 struct sctp_association *asoc;
6989 int retval = -EFAULT;
6991 if (len < sizeof(params)) {
6996 len = sizeof(params);
6997 if (copy_from_user(¶ms, optval, len))
7000 asoc = sctp_id2assoc(sk, params.assoc_id);
7002 params.assoc_value = asoc->reconf_enable;
7003 } else if (!params.assoc_id) {
7004 struct sctp_sock *sp = sctp_sk(sk);
7006 params.assoc_value = sp->ep->reconf_enable;
7012 if (put_user(len, optlen))
7015 if (copy_to_user(optval, ¶ms, len))
7024 static int sctp_getsockopt_enable_strreset(struct sock *sk, int len,
7025 char __user *optval,
7028 struct sctp_assoc_value params;
7029 struct sctp_association *asoc;
7030 int retval = -EFAULT;
7032 if (len < sizeof(params)) {
7037 len = sizeof(params);
7038 if (copy_from_user(¶ms, optval, len))
7041 asoc = sctp_id2assoc(sk, params.assoc_id);
7043 params.assoc_value = asoc->strreset_enable;
7044 } else if (!params.assoc_id) {
7045 struct sctp_sock *sp = sctp_sk(sk);
7047 params.assoc_value = sp->ep->strreset_enable;
7053 if (put_user(len, optlen))
7056 if (copy_to_user(optval, ¶ms, len))
7065 static int sctp_getsockopt_scheduler(struct sock *sk, int len,
7066 char __user *optval,
7069 struct sctp_assoc_value params;
7070 struct sctp_association *asoc;
7071 int retval = -EFAULT;
7073 if (len < sizeof(params)) {
7078 len = sizeof(params);
7079 if (copy_from_user(¶ms, optval, len))
7082 asoc = sctp_id2assoc(sk, params.assoc_id);
7088 params.assoc_value = sctp_sched_get_sched(asoc);
7090 if (put_user(len, optlen))
7093 if (copy_to_user(optval, ¶ms, len))
7102 static int sctp_getsockopt_scheduler_value(struct sock *sk, int len,
7103 char __user *optval,
7106 struct sctp_stream_value params;
7107 struct sctp_association *asoc;
7108 int retval = -EFAULT;
7110 if (len < sizeof(params)) {
7115 len = sizeof(params);
7116 if (copy_from_user(¶ms, optval, len))
7119 asoc = sctp_id2assoc(sk, params.assoc_id);
7125 retval = sctp_sched_get_value(asoc, params.stream_id,
7126 ¶ms.stream_value);
7130 if (put_user(len, optlen)) {
7135 if (copy_to_user(optval, ¶ms, len)) {
7144 static int sctp_getsockopt_interleaving_supported(struct sock *sk, int len,
7145 char __user *optval,
7148 struct sctp_assoc_value params;
7149 struct sctp_association *asoc;
7150 int retval = -EFAULT;
7152 if (len < sizeof(params)) {
7157 len = sizeof(params);
7158 if (copy_from_user(¶ms, optval, len))
7161 asoc = sctp_id2assoc(sk, params.assoc_id);
7163 params.assoc_value = asoc->intl_enable;
7164 } else if (!params.assoc_id) {
7165 struct sctp_sock *sp = sctp_sk(sk);
7167 params.assoc_value = sp->strm_interleave;
7173 if (put_user(len, optlen))
7176 if (copy_to_user(optval, ¶ms, len))
7185 static int sctp_getsockopt(struct sock *sk, int level, int optname,
7186 char __user *optval, int __user *optlen)
7191 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
7193 /* I can hardly begin to describe how wrong this is. This is
7194 * so broken as to be worse than useless. The API draft
7195 * REALLY is NOT helpful here... I am not convinced that the
7196 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
7197 * are at all well-founded.
7199 if (level != SOL_SCTP) {
7200 struct sctp_af *af = sctp_sk(sk)->pf->af;
7202 retval = af->getsockopt(sk, level, optname, optval, optlen);
7206 if (get_user(len, optlen))
7216 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
7218 case SCTP_DISABLE_FRAGMENTS:
7219 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
7223 retval = sctp_getsockopt_events(sk, len, optval, optlen);
7225 case SCTP_AUTOCLOSE:
7226 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
7228 case SCTP_SOCKOPT_PEELOFF:
7229 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
7231 case SCTP_SOCKOPT_PEELOFF_FLAGS:
7232 retval = sctp_getsockopt_peeloff_flags(sk, len, optval, optlen);
7234 case SCTP_PEER_ADDR_PARAMS:
7235 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
7238 case SCTP_DELAYED_SACK:
7239 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
7243 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
7245 case SCTP_GET_PEER_ADDRS:
7246 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
7249 case SCTP_GET_LOCAL_ADDRS:
7250 retval = sctp_getsockopt_local_addrs(sk, len, optval,
7253 case SCTP_SOCKOPT_CONNECTX3:
7254 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
7256 case SCTP_DEFAULT_SEND_PARAM:
7257 retval = sctp_getsockopt_default_send_param(sk, len,
7260 case SCTP_DEFAULT_SNDINFO:
7261 retval = sctp_getsockopt_default_sndinfo(sk, len,
7264 case SCTP_PRIMARY_ADDR:
7265 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
7268 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
7271 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
7273 case SCTP_ASSOCINFO:
7274 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
7276 case SCTP_I_WANT_MAPPED_V4_ADDR:
7277 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
7280 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
7282 case SCTP_GET_PEER_ADDR_INFO:
7283 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
7286 case SCTP_ADAPTATION_LAYER:
7287 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
7291 retval = sctp_getsockopt_context(sk, len, optval, optlen);
7293 case SCTP_FRAGMENT_INTERLEAVE:
7294 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
7297 case SCTP_PARTIAL_DELIVERY_POINT:
7298 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
7301 case SCTP_MAX_BURST:
7302 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
7305 case SCTP_AUTH_CHUNK:
7306 case SCTP_AUTH_DELETE_KEY:
7307 case SCTP_AUTH_DEACTIVATE_KEY:
7308 retval = -EOPNOTSUPP;
7310 case SCTP_HMAC_IDENT:
7311 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
7313 case SCTP_AUTH_ACTIVE_KEY:
7314 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
7316 case SCTP_PEER_AUTH_CHUNKS:
7317 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
7320 case SCTP_LOCAL_AUTH_CHUNKS:
7321 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
7324 case SCTP_GET_ASSOC_NUMBER:
7325 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
7327 case SCTP_GET_ASSOC_ID_LIST:
7328 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
7330 case SCTP_AUTO_ASCONF:
7331 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
7333 case SCTP_PEER_ADDR_THLDS:
7334 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
7336 case SCTP_GET_ASSOC_STATS:
7337 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
7339 case SCTP_RECVRCVINFO:
7340 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
7342 case SCTP_RECVNXTINFO:
7343 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
7345 case SCTP_PR_SUPPORTED:
7346 retval = sctp_getsockopt_pr_supported(sk, len, optval, optlen);
7348 case SCTP_DEFAULT_PRINFO:
7349 retval = sctp_getsockopt_default_prinfo(sk, len, optval,
7352 case SCTP_PR_ASSOC_STATUS:
7353 retval = sctp_getsockopt_pr_assocstatus(sk, len, optval,
7356 case SCTP_PR_STREAM_STATUS:
7357 retval = sctp_getsockopt_pr_streamstatus(sk, len, optval,
7360 case SCTP_RECONFIG_SUPPORTED:
7361 retval = sctp_getsockopt_reconfig_supported(sk, len, optval,
7364 case SCTP_ENABLE_STREAM_RESET:
7365 retval = sctp_getsockopt_enable_strreset(sk, len, optval,
7368 case SCTP_STREAM_SCHEDULER:
7369 retval = sctp_getsockopt_scheduler(sk, len, optval,
7372 case SCTP_STREAM_SCHEDULER_VALUE:
7373 retval = sctp_getsockopt_scheduler_value(sk, len, optval,
7376 case SCTP_INTERLEAVING_SUPPORTED:
7377 retval = sctp_getsockopt_interleaving_supported(sk, len, optval,
7381 retval = -ENOPROTOOPT;
7389 static int sctp_hash(struct sock *sk)
7395 static void sctp_unhash(struct sock *sk)
7400 /* Check if port is acceptable. Possibly find first available port.
7402 * The port hash table (contained in the 'global' SCTP protocol storage
7403 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
7404 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
7405 * list (the list number is the port number hashed out, so as you
7406 * would expect from a hash function, all the ports in a given list have
7407 * such a number that hashes out to the same list number; you were
7408 * expecting that, right?); so each list has a set of ports, with a
7409 * link to the socket (struct sock) that uses it, the port number and
7410 * a fastreuse flag (FIXME: NPI ipg).
7412 static struct sctp_bind_bucket *sctp_bucket_create(
7413 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
7415 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
7417 struct sctp_bind_hashbucket *head; /* hash list */
7418 struct sctp_bind_bucket *pp;
7419 unsigned short snum;
7422 snum = ntohs(addr->v4.sin_port);
7424 pr_debug("%s: begins, snum:%d\n", __func__, snum);
7429 /* Search for an available port. */
7430 int low, high, remaining, index;
7432 struct net *net = sock_net(sk);
7434 inet_get_local_port_range(net, &low, &high);
7435 remaining = (high - low) + 1;
7436 rover = prandom_u32() % remaining + low;
7440 if ((rover < low) || (rover > high))
7442 if (inet_is_local_reserved_port(net, rover))
7444 index = sctp_phashfn(sock_net(sk), rover);
7445 head = &sctp_port_hashtable[index];
7446 spin_lock(&head->lock);
7447 sctp_for_each_hentry(pp, &head->chain)
7448 if ((pp->port == rover) &&
7449 net_eq(sock_net(sk), pp->net))
7453 spin_unlock(&head->lock);
7454 } while (--remaining > 0);
7456 /* Exhausted local port range during search? */
7461 /* OK, here is the one we will use. HEAD (the port
7462 * hash table list entry) is non-NULL and we hold it's
7467 /* We are given an specific port number; we verify
7468 * that it is not being used. If it is used, we will
7469 * exahust the search in the hash list corresponding
7470 * to the port number (snum) - we detect that with the
7471 * port iterator, pp being NULL.
7473 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
7474 spin_lock(&head->lock);
7475 sctp_for_each_hentry(pp, &head->chain) {
7476 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
7483 if (!hlist_empty(&pp->owner)) {
7484 /* We had a port hash table hit - there is an
7485 * available port (pp != NULL) and it is being
7486 * used by other socket (pp->owner not empty); that other
7487 * socket is going to be sk2.
7489 int reuse = sk->sk_reuse;
7492 pr_debug("%s: found a possible match\n", __func__);
7494 if (pp->fastreuse && sk->sk_reuse &&
7495 sk->sk_state != SCTP_SS_LISTENING)
7498 /* Run through the list of sockets bound to the port
7499 * (pp->port) [via the pointers bind_next and
7500 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
7501 * we get the endpoint they describe and run through
7502 * the endpoint's list of IP (v4 or v6) addresses,
7503 * comparing each of the addresses with the address of
7504 * the socket sk. If we find a match, then that means
7505 * that this port/socket (sk) combination are already
7508 sk_for_each_bound(sk2, &pp->owner) {
7509 struct sctp_endpoint *ep2;
7510 ep2 = sctp_sk(sk2)->ep;
7513 (reuse && sk2->sk_reuse &&
7514 sk2->sk_state != SCTP_SS_LISTENING))
7517 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
7518 sctp_sk(sk2), sctp_sk(sk))) {
7524 pr_debug("%s: found a match\n", __func__);
7527 /* If there was a hash table miss, create a new port. */
7529 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
7532 /* In either case (hit or miss), make sure fastreuse is 1 only
7533 * if sk->sk_reuse is too (that is, if the caller requested
7534 * SO_REUSEADDR on this socket -sk-).
7536 if (hlist_empty(&pp->owner)) {
7537 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
7541 } else if (pp->fastreuse &&
7542 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
7545 /* We are set, so fill up all the data in the hash table
7546 * entry, tie the socket list information with the rest of the
7547 * sockets FIXME: Blurry, NPI (ipg).
7550 if (!sctp_sk(sk)->bind_hash) {
7551 inet_sk(sk)->inet_num = snum;
7552 sk_add_bind_node(sk, &pp->owner);
7553 sctp_sk(sk)->bind_hash = pp;
7558 spin_unlock(&head->lock);
7565 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
7566 * port is requested.
7568 static int sctp_get_port(struct sock *sk, unsigned short snum)
7570 union sctp_addr addr;
7571 struct sctp_af *af = sctp_sk(sk)->pf->af;
7573 /* Set up a dummy address struct from the sk. */
7574 af->from_sk(&addr, sk);
7575 addr.v4.sin_port = htons(snum);
7577 /* Note: sk->sk_num gets filled in if ephemeral port request. */
7578 return !!sctp_get_port_local(sk, &addr);
7582 * Move a socket to LISTENING state.
7584 static int sctp_listen_start(struct sock *sk, int backlog)
7586 struct sctp_sock *sp = sctp_sk(sk);
7587 struct sctp_endpoint *ep = sp->ep;
7588 struct crypto_shash *tfm = NULL;
7591 /* Allocate HMAC for generating cookie. */
7592 if (!sp->hmac && sp->sctp_hmac_alg) {
7593 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
7594 tfm = crypto_alloc_shash(alg, 0, 0);
7596 net_info_ratelimited("failed to load transform for %s: %ld\n",
7597 sp->sctp_hmac_alg, PTR_ERR(tfm));
7600 sctp_sk(sk)->hmac = tfm;
7604 * If a bind() or sctp_bindx() is not called prior to a listen()
7605 * call that allows new associations to be accepted, the system
7606 * picks an ephemeral port and will choose an address set equivalent
7607 * to binding with a wildcard address.
7609 * This is not currently spelled out in the SCTP sockets
7610 * extensions draft, but follows the practice as seen in TCP
7614 inet_sk_set_state(sk, SCTP_SS_LISTENING);
7615 if (!ep->base.bind_addr.port) {
7616 if (sctp_autobind(sk))
7619 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
7620 inet_sk_set_state(sk, SCTP_SS_CLOSED);
7625 sk->sk_max_ack_backlog = backlog;
7626 sctp_hash_endpoint(ep);
7631 * 4.1.3 / 5.1.3 listen()
7633 * By default, new associations are not accepted for UDP style sockets.
7634 * An application uses listen() to mark a socket as being able to
7635 * accept new associations.
7637 * On TCP style sockets, applications use listen() to ready the SCTP
7638 * endpoint for accepting inbound associations.
7640 * On both types of endpoints a backlog of '0' disables listening.
7642 * Move a socket to LISTENING state.
7644 int sctp_inet_listen(struct socket *sock, int backlog)
7646 struct sock *sk = sock->sk;
7647 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
7650 if (unlikely(backlog < 0))
7655 /* Peeled-off sockets are not allowed to listen(). */
7656 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
7659 if (sock->state != SS_UNCONNECTED)
7662 if (!sctp_sstate(sk, LISTENING) && !sctp_sstate(sk, CLOSED))
7665 /* If backlog is zero, disable listening. */
7667 if (sctp_sstate(sk, CLOSED))
7671 sctp_unhash_endpoint(ep);
7672 sk->sk_state = SCTP_SS_CLOSED;
7674 sctp_sk(sk)->bind_hash->fastreuse = 1;
7678 /* If we are already listening, just update the backlog */
7679 if (sctp_sstate(sk, LISTENING))
7680 sk->sk_max_ack_backlog = backlog;
7682 err = sctp_listen_start(sk, backlog);
7694 * This function is done by modeling the current datagram_poll() and the
7695 * tcp_poll(). Note that, based on these implementations, we don't
7696 * lock the socket in this function, even though it seems that,
7697 * ideally, locking or some other mechanisms can be used to ensure
7698 * the integrity of the counters (sndbuf and wmem_alloc) used
7699 * in this place. We assume that we don't need locks either until proven
7702 * Another thing to note is that we include the Async I/O support
7703 * here, again, by modeling the current TCP/UDP code. We don't have
7704 * a good way to test with it yet.
7706 __poll_t sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
7708 struct sock *sk = sock->sk;
7709 struct sctp_sock *sp = sctp_sk(sk);
7712 poll_wait(file, sk_sleep(sk), wait);
7714 sock_rps_record_flow(sk);
7716 /* A TCP-style listening socket becomes readable when the accept queue
7719 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
7720 return (!list_empty(&sp->ep->asocs)) ?
7721 (EPOLLIN | EPOLLRDNORM) : 0;
7725 /* Is there any exceptional events? */
7726 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
7728 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? EPOLLPRI : 0);
7729 if (sk->sk_shutdown & RCV_SHUTDOWN)
7730 mask |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
7731 if (sk->sk_shutdown == SHUTDOWN_MASK)
7734 /* Is it readable? Reconsider this code with TCP-style support. */
7735 if (!skb_queue_empty(&sk->sk_receive_queue))
7736 mask |= EPOLLIN | EPOLLRDNORM;
7738 /* The association is either gone or not ready. */
7739 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
7742 /* Is it writable? */
7743 if (sctp_writeable(sk)) {
7744 mask |= EPOLLOUT | EPOLLWRNORM;
7746 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
7748 * Since the socket is not locked, the buffer
7749 * might be made available after the writeable check and
7750 * before the bit is set. This could cause a lost I/O
7751 * signal. tcp_poll() has a race breaker for this race
7752 * condition. Based on their implementation, we put
7753 * in the following code to cover it as well.
7755 if (sctp_writeable(sk))
7756 mask |= EPOLLOUT | EPOLLWRNORM;
7761 /********************************************************************
7762 * 2nd Level Abstractions
7763 ********************************************************************/
7765 static struct sctp_bind_bucket *sctp_bucket_create(
7766 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
7768 struct sctp_bind_bucket *pp;
7770 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
7772 SCTP_DBG_OBJCNT_INC(bind_bucket);
7775 INIT_HLIST_HEAD(&pp->owner);
7777 hlist_add_head(&pp->node, &head->chain);
7782 /* Caller must hold hashbucket lock for this tb with local BH disabled */
7783 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
7785 if (pp && hlist_empty(&pp->owner)) {
7786 __hlist_del(&pp->node);
7787 kmem_cache_free(sctp_bucket_cachep, pp);
7788 SCTP_DBG_OBJCNT_DEC(bind_bucket);
7792 /* Release this socket's reference to a local port. */
7793 static inline void __sctp_put_port(struct sock *sk)
7795 struct sctp_bind_hashbucket *head =
7796 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
7797 inet_sk(sk)->inet_num)];
7798 struct sctp_bind_bucket *pp;
7800 spin_lock(&head->lock);
7801 pp = sctp_sk(sk)->bind_hash;
7802 __sk_del_bind_node(sk);
7803 sctp_sk(sk)->bind_hash = NULL;
7804 inet_sk(sk)->inet_num = 0;
7805 sctp_bucket_destroy(pp);
7806 spin_unlock(&head->lock);
7809 void sctp_put_port(struct sock *sk)
7812 __sctp_put_port(sk);
7817 * The system picks an ephemeral port and choose an address set equivalent
7818 * to binding with a wildcard address.
7819 * One of those addresses will be the primary address for the association.
7820 * This automatically enables the multihoming capability of SCTP.
7822 static int sctp_autobind(struct sock *sk)
7824 union sctp_addr autoaddr;
7828 /* Initialize a local sockaddr structure to INADDR_ANY. */
7829 af = sctp_sk(sk)->pf->af;
7831 port = htons(inet_sk(sk)->inet_num);
7832 af->inaddr_any(&autoaddr, port);
7834 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
7837 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
7840 * 4.2 The cmsghdr Structure *
7842 * When ancillary data is sent or received, any number of ancillary data
7843 * objects can be specified by the msg_control and msg_controllen members of
7844 * the msghdr structure, because each object is preceded by
7845 * a cmsghdr structure defining the object's length (the cmsg_len member).
7846 * Historically Berkeley-derived implementations have passed only one object
7847 * at a time, but this API allows multiple objects to be
7848 * passed in a single call to sendmsg() or recvmsg(). The following example
7849 * shows two ancillary data objects in a control buffer.
7851 * |<--------------------------- msg_controllen -------------------------->|
7854 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
7856 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
7859 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
7861 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
7864 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7865 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
7867 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
7869 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7876 static int sctp_msghdr_parse(const struct msghdr *msg, struct sctp_cmsgs *cmsgs)
7878 struct msghdr *my_msg = (struct msghdr *)msg;
7879 struct cmsghdr *cmsg;
7881 for_each_cmsghdr(cmsg, my_msg) {
7882 if (!CMSG_OK(my_msg, cmsg))
7885 /* Should we parse this header or ignore? */
7886 if (cmsg->cmsg_level != IPPROTO_SCTP)
7889 /* Strictly check lengths following example in SCM code. */
7890 switch (cmsg->cmsg_type) {
7892 /* SCTP Socket API Extension
7893 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
7895 * This cmsghdr structure provides information for
7896 * initializing new SCTP associations with sendmsg().
7897 * The SCTP_INITMSG socket option uses this same data
7898 * structure. This structure is not used for
7901 * cmsg_level cmsg_type cmsg_data[]
7902 * ------------ ------------ ----------------------
7903 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
7905 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
7908 cmsgs->init = CMSG_DATA(cmsg);
7912 /* SCTP Socket API Extension
7913 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
7915 * This cmsghdr structure specifies SCTP options for
7916 * sendmsg() and describes SCTP header information
7917 * about a received message through recvmsg().
7919 * cmsg_level cmsg_type cmsg_data[]
7920 * ------------ ------------ ----------------------
7921 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
7923 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
7926 cmsgs->srinfo = CMSG_DATA(cmsg);
7928 if (cmsgs->srinfo->sinfo_flags &
7929 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7930 SCTP_SACK_IMMEDIATELY | SCTP_SENDALL |
7931 SCTP_PR_SCTP_MASK | SCTP_ABORT | SCTP_EOF))
7936 /* SCTP Socket API Extension
7937 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
7939 * This cmsghdr structure specifies SCTP options for
7940 * sendmsg(). This structure and SCTP_RCVINFO replaces
7941 * SCTP_SNDRCV which has been deprecated.
7943 * cmsg_level cmsg_type cmsg_data[]
7944 * ------------ ------------ ---------------------
7945 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
7947 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
7950 cmsgs->sinfo = CMSG_DATA(cmsg);
7952 if (cmsgs->sinfo->snd_flags &
7953 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7954 SCTP_SACK_IMMEDIATELY | SCTP_SENDALL |
7955 SCTP_PR_SCTP_MASK | SCTP_ABORT | SCTP_EOF))
7959 /* SCTP Socket API Extension
7960 * 5.3.7 SCTP PR-SCTP Information Structure (SCTP_PRINFO)
7962 * This cmsghdr structure specifies SCTP options for sendmsg().
7964 * cmsg_level cmsg_type cmsg_data[]
7965 * ------------ ------------ ---------------------
7966 * IPPROTO_SCTP SCTP_PRINFO struct sctp_prinfo
7968 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_prinfo)))
7971 cmsgs->prinfo = CMSG_DATA(cmsg);
7972 if (cmsgs->prinfo->pr_policy & ~SCTP_PR_SCTP_MASK)
7975 if (cmsgs->prinfo->pr_policy == SCTP_PR_SCTP_NONE)
7976 cmsgs->prinfo->pr_value = 0;
7979 /* SCTP Socket API Extension
7980 * 5.3.8 SCTP AUTH Information Structure (SCTP_AUTHINFO)
7982 * This cmsghdr structure specifies SCTP options for sendmsg().
7984 * cmsg_level cmsg_type cmsg_data[]
7985 * ------------ ------------ ---------------------
7986 * IPPROTO_SCTP SCTP_AUTHINFO struct sctp_authinfo
7988 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_authinfo)))
7991 cmsgs->authinfo = CMSG_DATA(cmsg);
7993 case SCTP_DSTADDRV4:
7994 case SCTP_DSTADDRV6:
7995 /* SCTP Socket API Extension
7996 * 5.3.9/10 SCTP Destination IPv4/6 Address Structure (SCTP_DSTADDRV4/6)
7998 * This cmsghdr structure specifies SCTP options for sendmsg().
8000 * cmsg_level cmsg_type cmsg_data[]
8001 * ------------ ------------ ---------------------
8002 * IPPROTO_SCTP SCTP_DSTADDRV4 struct in_addr
8003 * ------------ ------------ ---------------------
8004 * IPPROTO_SCTP SCTP_DSTADDRV6 struct in6_addr
8006 cmsgs->addrs_msg = my_msg;
8017 * Wait for a packet..
8018 * Note: This function is the same function as in core/datagram.c
8019 * with a few modifications to make lksctp work.
8021 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
8026 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
8028 /* Socket errors? */
8029 error = sock_error(sk);
8033 if (!skb_queue_empty(&sk->sk_receive_queue))
8036 /* Socket shut down? */
8037 if (sk->sk_shutdown & RCV_SHUTDOWN)
8040 /* Sequenced packets can come disconnected. If so we report the
8045 /* Is there a good reason to think that we may receive some data? */
8046 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
8049 /* Handle signals. */
8050 if (signal_pending(current))
8053 /* Let another process have a go. Since we are going to sleep
8054 * anyway. Note: This may cause odd behaviors if the message
8055 * does not fit in the user's buffer, but this seems to be the
8056 * only way to honor MSG_DONTWAIT realistically.
8059 *timeo_p = schedule_timeout(*timeo_p);
8063 finish_wait(sk_sleep(sk), &wait);
8067 error = sock_intr_errno(*timeo_p);
8070 finish_wait(sk_sleep(sk), &wait);
8075 /* Receive a datagram.
8076 * Note: This is pretty much the same routine as in core/datagram.c
8077 * with a few changes to make lksctp work.
8079 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
8080 int noblock, int *err)
8083 struct sk_buff *skb;
8086 timeo = sock_rcvtimeo(sk, noblock);
8088 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
8089 MAX_SCHEDULE_TIMEOUT);
8092 /* Again only user level code calls this function,
8093 * so nothing interrupt level
8094 * will suddenly eat the receive_queue.
8096 * Look at current nfs client by the way...
8097 * However, this function was correct in any case. 8)
8099 if (flags & MSG_PEEK) {
8100 skb = skb_peek(&sk->sk_receive_queue);
8102 refcount_inc(&skb->users);
8104 skb = __skb_dequeue(&sk->sk_receive_queue);
8110 /* Caller is allowed not to check sk->sk_err before calling. */
8111 error = sock_error(sk);
8115 if (sk->sk_shutdown & RCV_SHUTDOWN)
8118 if (sk_can_busy_loop(sk)) {
8119 sk_busy_loop(sk, noblock);
8121 if (!skb_queue_empty(&sk->sk_receive_queue))
8125 /* User doesn't want to wait. */
8129 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
8138 /* If sndbuf has changed, wake up per association sndbuf waiters. */
8139 static void __sctp_write_space(struct sctp_association *asoc)
8141 struct sock *sk = asoc->base.sk;
8143 if (sctp_wspace(asoc) <= 0)
8146 if (waitqueue_active(&asoc->wait))
8147 wake_up_interruptible(&asoc->wait);
8149 if (sctp_writeable(sk)) {
8150 struct socket_wq *wq;
8153 wq = rcu_dereference(sk->sk_wq);
8155 if (waitqueue_active(&wq->wait))
8156 wake_up_interruptible(&wq->wait);
8158 /* Note that we try to include the Async I/O support
8159 * here by modeling from the current TCP/UDP code.
8160 * We have not tested with it yet.
8162 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
8163 sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT);
8169 static void sctp_wake_up_waiters(struct sock *sk,
8170 struct sctp_association *asoc)
8172 struct sctp_association *tmp = asoc;
8174 /* We do accounting for the sndbuf space per association,
8175 * so we only need to wake our own association.
8177 if (asoc->ep->sndbuf_policy)
8178 return __sctp_write_space(asoc);
8180 /* If association goes down and is just flushing its
8181 * outq, then just normally notify others.
8183 if (asoc->base.dead)
8184 return sctp_write_space(sk);
8186 /* Accounting for the sndbuf space is per socket, so we
8187 * need to wake up others, try to be fair and in case of
8188 * other associations, let them have a go first instead
8189 * of just doing a sctp_write_space() call.
8191 * Note that we reach sctp_wake_up_waiters() only when
8192 * associations free up queued chunks, thus we are under
8193 * lock and the list of associations on a socket is
8194 * guaranteed not to change.
8196 for (tmp = list_next_entry(tmp, asocs); 1;
8197 tmp = list_next_entry(tmp, asocs)) {
8198 /* Manually skip the head element. */
8199 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
8201 /* Wake up association. */
8202 __sctp_write_space(tmp);
8203 /* We've reached the end. */
8209 /* Do accounting for the sndbuf space.
8210 * Decrement the used sndbuf space of the corresponding association by the
8211 * data size which was just transmitted(freed).
8213 static void sctp_wfree(struct sk_buff *skb)
8215 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
8216 struct sctp_association *asoc = chunk->asoc;
8217 struct sock *sk = asoc->base.sk;
8219 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
8220 sizeof(struct sk_buff) +
8221 sizeof(struct sctp_chunk);
8223 WARN_ON(refcount_sub_and_test(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc));
8226 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
8228 sk->sk_wmem_queued -= skb->truesize;
8229 sk_mem_uncharge(sk, skb->truesize);
8232 struct sctp_shared_key *shkey = chunk->shkey;
8234 /* refcnt == 2 and !list_empty mean after this release, it's
8235 * not being used anywhere, and it's time to notify userland
8236 * that this shkey can be freed if it's been deactivated.
8238 if (shkey->deactivated && !list_empty(&shkey->key_list) &&
8239 refcount_read(&shkey->refcnt) == 2) {
8240 struct sctp_ulpevent *ev;
8242 ev = sctp_ulpevent_make_authkey(asoc, shkey->key_id,
8246 asoc->stream.si->enqueue_event(&asoc->ulpq, ev);
8248 sctp_auth_shkey_release(chunk->shkey);
8252 sctp_wake_up_waiters(sk, asoc);
8254 sctp_association_put(asoc);
8257 /* Do accounting for the receive space on the socket.
8258 * Accounting for the association is done in ulpevent.c
8259 * We set this as a destructor for the cloned data skbs so that
8260 * accounting is done at the correct time.
8262 void sctp_sock_rfree(struct sk_buff *skb)
8264 struct sock *sk = skb->sk;
8265 struct sctp_ulpevent *event = sctp_skb2event(skb);
8267 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
8270 * Mimic the behavior of sock_rfree
8272 sk_mem_uncharge(sk, event->rmem_len);
8276 /* Helper function to wait for space in the sndbuf. */
8277 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
8280 struct sock *sk = asoc->base.sk;
8281 long current_timeo = *timeo_p;
8285 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
8288 /* Increment the association's refcnt. */
8289 sctp_association_hold(asoc);
8291 /* Wait on the association specific sndbuf space. */
8293 prepare_to_wait_exclusive(&asoc->wait, &wait,
8294 TASK_INTERRUPTIBLE);
8295 if (asoc->base.dead)
8299 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING)
8301 if (signal_pending(current))
8302 goto do_interrupted;
8303 if (msg_len <= sctp_wspace(asoc))
8306 /* Let another process have a go. Since we are going
8310 current_timeo = schedule_timeout(current_timeo);
8312 if (sk != asoc->base.sk)
8315 *timeo_p = current_timeo;
8319 finish_wait(&asoc->wait, &wait);
8321 /* Release the association's refcnt. */
8322 sctp_association_put(asoc);
8335 err = sock_intr_errno(*timeo_p);
8343 void sctp_data_ready(struct sock *sk)
8345 struct socket_wq *wq;
8348 wq = rcu_dereference(sk->sk_wq);
8349 if (skwq_has_sleeper(wq))
8350 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
8351 EPOLLRDNORM | EPOLLRDBAND);
8352 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
8356 /* If socket sndbuf has changed, wake up all per association waiters. */
8357 void sctp_write_space(struct sock *sk)
8359 struct sctp_association *asoc;
8361 /* Wake up the tasks in each wait queue. */
8362 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
8363 __sctp_write_space(asoc);
8367 /* Is there any sndbuf space available on the socket?
8369 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
8370 * associations on the same socket. For a UDP-style socket with
8371 * multiple associations, it is possible for it to be "unwriteable"
8372 * prematurely. I assume that this is acceptable because
8373 * a premature "unwriteable" is better than an accidental "writeable" which
8374 * would cause an unwanted block under certain circumstances. For the 1-1
8375 * UDP-style sockets or TCP-style sockets, this code should work.
8378 static int sctp_writeable(struct sock *sk)
8382 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
8388 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
8389 * returns immediately with EINPROGRESS.
8391 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
8393 struct sock *sk = asoc->base.sk;
8395 long current_timeo = *timeo_p;
8398 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
8400 /* Increment the association's refcnt. */
8401 sctp_association_hold(asoc);
8404 prepare_to_wait_exclusive(&asoc->wait, &wait,
8405 TASK_INTERRUPTIBLE);
8408 if (sk->sk_shutdown & RCV_SHUTDOWN)
8410 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
8413 if (signal_pending(current))
8414 goto do_interrupted;
8416 if (sctp_state(asoc, ESTABLISHED))
8419 /* Let another process have a go. Since we are going
8423 current_timeo = schedule_timeout(current_timeo);
8426 *timeo_p = current_timeo;
8430 finish_wait(&asoc->wait, &wait);
8432 /* Release the association's refcnt. */
8433 sctp_association_put(asoc);
8438 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
8441 err = -ECONNREFUSED;
8445 err = sock_intr_errno(*timeo_p);
8453 static int sctp_wait_for_accept(struct sock *sk, long timeo)
8455 struct sctp_endpoint *ep;
8459 ep = sctp_sk(sk)->ep;
8463 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
8464 TASK_INTERRUPTIBLE);
8466 if (list_empty(&ep->asocs)) {
8468 timeo = schedule_timeout(timeo);
8473 if (!sctp_sstate(sk, LISTENING))
8477 if (!list_empty(&ep->asocs))
8480 err = sock_intr_errno(timeo);
8481 if (signal_pending(current))
8489 finish_wait(sk_sleep(sk), &wait);
8494 static void sctp_wait_for_close(struct sock *sk, long timeout)
8499 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
8500 if (list_empty(&sctp_sk(sk)->ep->asocs))
8503 timeout = schedule_timeout(timeout);
8505 } while (!signal_pending(current) && timeout);
8507 finish_wait(sk_sleep(sk), &wait);
8510 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
8512 struct sk_buff *frag;
8517 /* Don't forget the fragments. */
8518 skb_walk_frags(skb, frag)
8519 sctp_skb_set_owner_r_frag(frag, sk);
8522 sctp_skb_set_owner_r(skb, sk);
8525 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
8526 struct sctp_association *asoc)
8528 struct inet_sock *inet = inet_sk(sk);
8529 struct inet_sock *newinet;
8530 struct sctp_sock *sp = sctp_sk(sk);
8531 struct sctp_endpoint *ep = sp->ep;
8533 newsk->sk_type = sk->sk_type;
8534 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
8535 newsk->sk_flags = sk->sk_flags;
8536 newsk->sk_tsflags = sk->sk_tsflags;
8537 newsk->sk_no_check_tx = sk->sk_no_check_tx;
8538 newsk->sk_no_check_rx = sk->sk_no_check_rx;
8539 newsk->sk_reuse = sk->sk_reuse;
8541 newsk->sk_shutdown = sk->sk_shutdown;
8542 newsk->sk_destruct = sctp_destruct_sock;
8543 newsk->sk_family = sk->sk_family;
8544 newsk->sk_protocol = IPPROTO_SCTP;
8545 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
8546 newsk->sk_sndbuf = sk->sk_sndbuf;
8547 newsk->sk_rcvbuf = sk->sk_rcvbuf;
8548 newsk->sk_lingertime = sk->sk_lingertime;
8549 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
8550 newsk->sk_sndtimeo = sk->sk_sndtimeo;
8551 newsk->sk_rxhash = sk->sk_rxhash;
8553 newinet = inet_sk(newsk);
8555 /* Initialize sk's sport, dport, rcv_saddr and daddr for
8556 * getsockname() and getpeername()
8558 newinet->inet_sport = inet->inet_sport;
8559 newinet->inet_saddr = inet->inet_saddr;
8560 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
8561 newinet->inet_dport = htons(asoc->peer.port);
8562 newinet->pmtudisc = inet->pmtudisc;
8563 newinet->inet_id = asoc->next_tsn ^ jiffies;
8565 newinet->uc_ttl = inet->uc_ttl;
8566 newinet->mc_loop = 1;
8567 newinet->mc_ttl = 1;
8568 newinet->mc_index = 0;
8569 newinet->mc_list = NULL;
8571 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
8572 net_enable_timestamp();
8574 /* Set newsk security attributes from orginal sk and connection
8575 * security attribute from ep.
8577 security_sctp_sk_clone(ep, sk, newsk);
8580 static inline void sctp_copy_descendant(struct sock *sk_to,
8581 const struct sock *sk_from)
8583 int ancestor_size = sizeof(struct inet_sock) +
8584 sizeof(struct sctp_sock) -
8585 offsetof(struct sctp_sock, auto_asconf_list);
8587 if (sk_from->sk_family == PF_INET6)
8588 ancestor_size += sizeof(struct ipv6_pinfo);
8590 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
8593 /* Populate the fields of the newsk from the oldsk and migrate the assoc
8594 * and its messages to the newsk.
8596 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
8597 struct sctp_association *assoc,
8598 enum sctp_socket_type type)
8600 struct sctp_sock *oldsp = sctp_sk(oldsk);
8601 struct sctp_sock *newsp = sctp_sk(newsk);
8602 struct sctp_bind_bucket *pp; /* hash list port iterator */
8603 struct sctp_endpoint *newep = newsp->ep;
8604 struct sk_buff *skb, *tmp;
8605 struct sctp_ulpevent *event;
8606 struct sctp_bind_hashbucket *head;
8608 /* Migrate socket buffer sizes and all the socket level options to the
8611 newsk->sk_sndbuf = oldsk->sk_sndbuf;
8612 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
8613 /* Brute force copy old sctp opt. */
8614 sctp_copy_descendant(newsk, oldsk);
8616 /* Restore the ep value that was overwritten with the above structure
8622 /* Hook this new socket in to the bind_hash list. */
8623 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
8624 inet_sk(oldsk)->inet_num)];
8625 spin_lock_bh(&head->lock);
8626 pp = sctp_sk(oldsk)->bind_hash;
8627 sk_add_bind_node(newsk, &pp->owner);
8628 sctp_sk(newsk)->bind_hash = pp;
8629 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
8630 spin_unlock_bh(&head->lock);
8632 /* Copy the bind_addr list from the original endpoint to the new
8633 * endpoint so that we can handle restarts properly
8635 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
8636 &oldsp->ep->base.bind_addr, GFP_KERNEL);
8638 /* Move any messages in the old socket's receive queue that are for the
8639 * peeled off association to the new socket's receive queue.
8641 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
8642 event = sctp_skb2event(skb);
8643 if (event->asoc == assoc) {
8644 __skb_unlink(skb, &oldsk->sk_receive_queue);
8645 __skb_queue_tail(&newsk->sk_receive_queue, skb);
8646 sctp_skb_set_owner_r_frag(skb, newsk);
8650 /* Clean up any messages pending delivery due to partial
8651 * delivery. Three cases:
8652 * 1) No partial deliver; no work.
8653 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
8654 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
8656 skb_queue_head_init(&newsp->pd_lobby);
8657 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
8659 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
8660 struct sk_buff_head *queue;
8662 /* Decide which queue to move pd_lobby skbs to. */
8663 if (assoc->ulpq.pd_mode) {
8664 queue = &newsp->pd_lobby;
8666 queue = &newsk->sk_receive_queue;
8668 /* Walk through the pd_lobby, looking for skbs that
8669 * need moved to the new socket.
8671 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
8672 event = sctp_skb2event(skb);
8673 if (event->asoc == assoc) {
8674 __skb_unlink(skb, &oldsp->pd_lobby);
8675 __skb_queue_tail(queue, skb);
8676 sctp_skb_set_owner_r_frag(skb, newsk);
8680 /* Clear up any skbs waiting for the partial
8681 * delivery to finish.
8683 if (assoc->ulpq.pd_mode)
8684 sctp_clear_pd(oldsk, NULL);
8688 sctp_for_each_rx_skb(assoc, newsk, sctp_skb_set_owner_r_frag);
8690 /* Set the type of socket to indicate that it is peeled off from the
8691 * original UDP-style socket or created with the accept() call on a
8692 * TCP-style socket..
8696 /* Mark the new socket "in-use" by the user so that any packets
8697 * that may arrive on the association after we've moved it are
8698 * queued to the backlog. This prevents a potential race between
8699 * backlog processing on the old socket and new-packet processing
8700 * on the new socket.
8702 * The caller has just allocated newsk so we can guarantee that other
8703 * paths won't try to lock it and then oldsk.
8705 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
8706 sctp_for_each_tx_datachunk(assoc, sctp_clear_owner_w);
8707 sctp_assoc_migrate(assoc, newsk);
8708 sctp_for_each_tx_datachunk(assoc, sctp_set_owner_w);
8710 /* If the association on the newsk is already closed before accept()
8711 * is called, set RCV_SHUTDOWN flag.
8713 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP)) {
8714 inet_sk_set_state(newsk, SCTP_SS_CLOSED);
8715 newsk->sk_shutdown |= RCV_SHUTDOWN;
8717 inet_sk_set_state(newsk, SCTP_SS_ESTABLISHED);
8720 release_sock(newsk);
8724 /* This proto struct describes the ULP interface for SCTP. */
8725 struct proto sctp_prot = {
8727 .owner = THIS_MODULE,
8728 .close = sctp_close,
8729 .connect = sctp_connect,
8730 .disconnect = sctp_disconnect,
8731 .accept = sctp_accept,
8732 .ioctl = sctp_ioctl,
8733 .init = sctp_init_sock,
8734 .destroy = sctp_destroy_sock,
8735 .shutdown = sctp_shutdown,
8736 .setsockopt = sctp_setsockopt,
8737 .getsockopt = sctp_getsockopt,
8738 .sendmsg = sctp_sendmsg,
8739 .recvmsg = sctp_recvmsg,
8741 .backlog_rcv = sctp_backlog_rcv,
8743 .unhash = sctp_unhash,
8744 .get_port = sctp_get_port,
8745 .obj_size = sizeof(struct sctp_sock),
8746 .useroffset = offsetof(struct sctp_sock, subscribe),
8747 .usersize = offsetof(struct sctp_sock, initmsg) -
8748 offsetof(struct sctp_sock, subscribe) +
8749 sizeof_field(struct sctp_sock, initmsg),
8750 .sysctl_mem = sysctl_sctp_mem,
8751 .sysctl_rmem = sysctl_sctp_rmem,
8752 .sysctl_wmem = sysctl_sctp_wmem,
8753 .memory_pressure = &sctp_memory_pressure,
8754 .enter_memory_pressure = sctp_enter_memory_pressure,
8755 .memory_allocated = &sctp_memory_allocated,
8756 .sockets_allocated = &sctp_sockets_allocated,
8759 #if IS_ENABLED(CONFIG_IPV6)
8761 #include <net/transp_v6.h>
8762 static void sctp_v6_destroy_sock(struct sock *sk)
8764 sctp_destroy_sock(sk);
8765 inet6_destroy_sock(sk);
8768 struct proto sctpv6_prot = {
8770 .owner = THIS_MODULE,
8771 .close = sctp_close,
8772 .connect = sctp_connect,
8773 .disconnect = sctp_disconnect,
8774 .accept = sctp_accept,
8775 .ioctl = sctp_ioctl,
8776 .init = sctp_init_sock,
8777 .destroy = sctp_v6_destroy_sock,
8778 .shutdown = sctp_shutdown,
8779 .setsockopt = sctp_setsockopt,
8780 .getsockopt = sctp_getsockopt,
8781 .sendmsg = sctp_sendmsg,
8782 .recvmsg = sctp_recvmsg,
8784 .backlog_rcv = sctp_backlog_rcv,
8786 .unhash = sctp_unhash,
8787 .get_port = sctp_get_port,
8788 .obj_size = sizeof(struct sctp6_sock),
8789 .useroffset = offsetof(struct sctp6_sock, sctp.subscribe),
8790 .usersize = offsetof(struct sctp6_sock, sctp.initmsg) -
8791 offsetof(struct sctp6_sock, sctp.subscribe) +
8792 sizeof_field(struct sctp6_sock, sctp.initmsg),
8793 .sysctl_mem = sysctl_sctp_mem,
8794 .sysctl_rmem = sysctl_sctp_rmem,
8795 .sysctl_wmem = sysctl_sctp_wmem,
8796 .memory_pressure = &sctp_memory_pressure,
8797 .enter_memory_pressure = sctp_enter_memory_pressure,
8798 .memory_allocated = &sctp_memory_allocated,
8799 .sockets_allocated = &sctp_sockets_allocated,
8801 #endif /* IS_ENABLED(CONFIG_IPV6) */