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,
1087 int addrs_size, int flags,
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;
1109 /* connect() cannot be done on a socket that is already in ESTABLISHED
1110 * state - UDP-style peeled off socket or a TCP-style socket that
1111 * is already connected.
1112 * It cannot be done even on a TCP-style listening socket.
1114 if (sctp_sstate(sk, ESTABLISHED) || sctp_sstate(sk, CLOSING) ||
1115 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1120 /* Walk through the addrs buffer and count the number of addresses. */
1122 while (walk_size < addrs_size) {
1125 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1131 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1133 /* If the address family is not supported or if this address
1134 * causes the address buffer to overflow return EINVAL.
1136 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1141 port = ntohs(sa_addr->v4.sin_port);
1143 /* Save current address so we can work with it */
1144 memcpy(&to, sa_addr, af->sockaddr_len);
1146 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1150 /* Make sure the destination port is correctly set
1153 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1158 /* Check if there already is a matching association on the
1159 * endpoint (other than the one created here).
1161 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1162 if (asoc2 && asoc2 != asoc) {
1163 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1170 /* If we could not find a matching association on the endpoint,
1171 * make sure that there is no peeled-off association matching
1172 * the peer address even on another socket.
1174 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1175 err = -EADDRNOTAVAIL;
1180 /* If a bind() or sctp_bindx() is not called prior to
1181 * an sctp_connectx() call, the system picks an
1182 * ephemeral port and will choose an address set
1183 * equivalent to binding with a wildcard address.
1185 if (!ep->base.bind_addr.port) {
1186 if (sctp_autobind(sk)) {
1192 * If an unprivileged user inherits a 1-many
1193 * style socket with open associations on a
1194 * privileged port, it MAY be permitted to
1195 * accept new associations, but it SHOULD NOT
1196 * be permitted to open new associations.
1198 if (ep->base.bind_addr.port <
1199 inet_prot_sock(net) &&
1200 !ns_capable(net->user_ns,
1201 CAP_NET_BIND_SERVICE)) {
1207 scope = sctp_scope(&to);
1208 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1214 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1222 /* Prime the peer's transport structures. */
1223 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1231 addr_buf += af->sockaddr_len;
1232 walk_size += af->sockaddr_len;
1235 /* In case the user of sctp_connectx() wants an association
1236 * id back, assign one now.
1239 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1244 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1249 /* Initialize sk's dport and daddr for getpeername() */
1250 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1251 sp->pf->to_sk_daddr(sa_addr, sk);
1254 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
1257 *assoc_id = asoc->assoc_id;
1259 err = sctp_wait_for_connect(asoc, &timeo);
1260 /* Note: the asoc may be freed after the return of
1261 * sctp_wait_for_connect.
1264 /* Don't free association on exit. */
1268 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1269 __func__, asoc, kaddrs, err);
1272 /* sctp_primitive_ASSOCIATE may have added this association
1273 * To the hash table, try to unhash it, just in case, its a noop
1274 * if it wasn't hashed so we're safe
1276 sctp_association_free(asoc);
1281 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1284 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1285 * sctp_assoc_t *asoc);
1287 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1288 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1289 * or IPv6 addresses.
1291 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1292 * Section 3.1.2 for this usage.
1294 * addrs is a pointer to an array of one or more socket addresses. Each
1295 * address is contained in its appropriate structure (i.e. struct
1296 * sockaddr_in or struct sockaddr_in6) the family of the address type
1297 * must be used to distengish the address length (note that this
1298 * representation is termed a "packed array" of addresses). The caller
1299 * specifies the number of addresses in the array with addrcnt.
1301 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1302 * the association id of the new association. On failure, sctp_connectx()
1303 * returns -1, and sets errno to the appropriate error code. The assoc_id
1304 * is not touched by the kernel.
1306 * For SCTP, the port given in each socket address must be the same, or
1307 * sctp_connectx() will fail, setting errno to EINVAL.
1309 * An application can use sctp_connectx to initiate an association with
1310 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1311 * allows a caller to specify multiple addresses at which a peer can be
1312 * reached. The way the SCTP stack uses the list of addresses to set up
1313 * the association is implementation dependent. This function only
1314 * specifies that the stack will try to make use of all the addresses in
1315 * the list when needed.
1317 * Note that the list of addresses passed in is only used for setting up
1318 * the association. It does not necessarily equal the set of addresses
1319 * the peer uses for the resulting association. If the caller wants to
1320 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1321 * retrieve them after the association has been set up.
1323 * Basically do nothing but copying the addresses from user to kernel
1324 * land and invoking either sctp_connectx(). This is used for tunneling
1325 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1327 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1330 * sk The sk of the socket
1331 * addrs The pointer to the addresses in user land
1332 * addrssize Size of the addrs buffer
1334 * Returns >=0 if ok, <0 errno code on error.
1336 static int __sctp_setsockopt_connectx(struct sock *sk,
1337 struct sockaddr __user *addrs,
1339 sctp_assoc_t *assoc_id)
1341 struct sockaddr *kaddrs;
1342 int err = 0, flags = 0;
1344 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1345 __func__, sk, addrs, addrs_size);
1347 if (unlikely(addrs_size <= 0))
1350 kaddrs = vmemdup_user(addrs, addrs_size);
1351 if (unlikely(IS_ERR(kaddrs)))
1352 return PTR_ERR(kaddrs);
1354 /* Allow security module to validate connectx addresses. */
1355 err = security_sctp_bind_connect(sk, SCTP_SOCKOPT_CONNECTX,
1356 (struct sockaddr *)kaddrs,
1361 /* in-kernel sockets don't generally have a file allocated to them
1362 * if all they do is call sock_create_kern().
1364 if (sk->sk_socket->file)
1365 flags = sk->sk_socket->file->f_flags;
1367 err = __sctp_connect(sk, kaddrs, addrs_size, flags, assoc_id);
1376 * This is an older interface. It's kept for backward compatibility
1377 * to the option that doesn't provide association id.
1379 static int sctp_setsockopt_connectx_old(struct sock *sk,
1380 struct sockaddr __user *addrs,
1383 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1387 * New interface for the API. The since the API is done with a socket
1388 * option, to make it simple we feed back the association id is as a return
1389 * indication to the call. Error is always negative and association id is
1392 static int sctp_setsockopt_connectx(struct sock *sk,
1393 struct sockaddr __user *addrs,
1396 sctp_assoc_t assoc_id = 0;
1399 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1408 * New (hopefully final) interface for the API.
1409 * We use the sctp_getaddrs_old structure so that use-space library
1410 * can avoid any unnecessary allocations. The only different part
1411 * is that we store the actual length of the address buffer into the
1412 * addrs_num structure member. That way we can re-use the existing
1415 #ifdef CONFIG_COMPAT
1416 struct compat_sctp_getaddrs_old {
1417 sctp_assoc_t assoc_id;
1419 compat_uptr_t addrs; /* struct sockaddr * */
1423 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1424 char __user *optval,
1427 struct sctp_getaddrs_old param;
1428 sctp_assoc_t assoc_id = 0;
1431 #ifdef CONFIG_COMPAT
1432 if (in_compat_syscall()) {
1433 struct compat_sctp_getaddrs_old param32;
1435 if (len < sizeof(param32))
1437 if (copy_from_user(¶m32, optval, sizeof(param32)))
1440 param.assoc_id = param32.assoc_id;
1441 param.addr_num = param32.addr_num;
1442 param.addrs = compat_ptr(param32.addrs);
1446 if (len < sizeof(param))
1448 if (copy_from_user(¶m, optval, sizeof(param)))
1452 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1453 param.addrs, param.addr_num,
1455 if (err == 0 || err == -EINPROGRESS) {
1456 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1458 if (put_user(sizeof(assoc_id), optlen))
1465 /* API 3.1.4 close() - UDP Style Syntax
1466 * Applications use close() to perform graceful shutdown (as described in
1467 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1468 * by a UDP-style socket.
1472 * ret = close(int sd);
1474 * sd - the socket descriptor of the associations to be closed.
1476 * To gracefully shutdown a specific association represented by the
1477 * UDP-style socket, an application should use the sendmsg() call,
1478 * passing no user data, but including the appropriate flag in the
1479 * ancillary data (see Section xxxx).
1481 * If sd in the close() call is a branched-off socket representing only
1482 * one association, the shutdown is performed on that association only.
1484 * 4.1.6 close() - TCP Style Syntax
1486 * Applications use close() to gracefully close down an association.
1490 * int close(int sd);
1492 * sd - the socket descriptor of the association to be closed.
1494 * After an application calls close() on a socket descriptor, no further
1495 * socket operations will succeed on that descriptor.
1497 * API 7.1.4 SO_LINGER
1499 * An application using the TCP-style socket can use this option to
1500 * perform the SCTP ABORT primitive. The linger option structure is:
1503 * int l_onoff; // option on/off
1504 * int l_linger; // linger time
1507 * To enable the option, set l_onoff to 1. If the l_linger value is set
1508 * to 0, calling close() is the same as the ABORT primitive. If the
1509 * value is set to a negative value, the setsockopt() call will return
1510 * an error. If the value is set to a positive value linger_time, the
1511 * close() can be blocked for at most linger_time ms. If the graceful
1512 * shutdown phase does not finish during this period, close() will
1513 * return but the graceful shutdown phase continues in the system.
1515 static void sctp_close(struct sock *sk, long timeout)
1517 struct net *net = sock_net(sk);
1518 struct sctp_endpoint *ep;
1519 struct sctp_association *asoc;
1520 struct list_head *pos, *temp;
1521 unsigned int data_was_unread;
1523 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1525 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
1526 sk->sk_shutdown = SHUTDOWN_MASK;
1527 inet_sk_set_state(sk, SCTP_SS_CLOSING);
1529 ep = sctp_sk(sk)->ep;
1531 /* Clean up any skbs sitting on the receive queue. */
1532 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1533 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1535 /* Walk all associations on an endpoint. */
1536 list_for_each_safe(pos, temp, &ep->asocs) {
1537 asoc = list_entry(pos, struct sctp_association, asocs);
1539 if (sctp_style(sk, TCP)) {
1540 /* A closed association can still be in the list if
1541 * it belongs to a TCP-style listening socket that is
1542 * not yet accepted. If so, free it. If not, send an
1543 * ABORT or SHUTDOWN based on the linger options.
1545 if (sctp_state(asoc, CLOSED)) {
1546 sctp_association_free(asoc);
1551 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1552 !skb_queue_empty(&asoc->ulpq.reasm) ||
1553 !skb_queue_empty(&asoc->ulpq.reasm_uo) ||
1554 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1555 struct sctp_chunk *chunk;
1557 chunk = sctp_make_abort_user(asoc, NULL, 0);
1558 sctp_primitive_ABORT(net, asoc, chunk);
1560 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1563 /* On a TCP-style socket, block for at most linger_time if set. */
1564 if (sctp_style(sk, TCP) && timeout)
1565 sctp_wait_for_close(sk, timeout);
1567 /* This will run the backlog queue. */
1570 /* Supposedly, no process has access to the socket, but
1571 * the net layers still may.
1572 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1573 * held and that should be grabbed before socket lock.
1575 spin_lock_bh(&net->sctp.addr_wq_lock);
1576 bh_lock_sock_nested(sk);
1578 /* Hold the sock, since sk_common_release() will put sock_put()
1579 * and we have just a little more cleanup.
1582 sk_common_release(sk);
1585 spin_unlock_bh(&net->sctp.addr_wq_lock);
1589 SCTP_DBG_OBJCNT_DEC(sock);
1592 /* Handle EPIPE error. */
1593 static int sctp_error(struct sock *sk, int flags, int err)
1596 err = sock_error(sk) ? : -EPIPE;
1597 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1598 send_sig(SIGPIPE, current, 0);
1602 /* API 3.1.3 sendmsg() - UDP Style Syntax
1604 * An application uses sendmsg() and recvmsg() calls to transmit data to
1605 * and receive data from its peer.
1607 * ssize_t sendmsg(int socket, const struct msghdr *message,
1610 * socket - the socket descriptor of the endpoint.
1611 * message - pointer to the msghdr structure which contains a single
1612 * user message and possibly some ancillary data.
1614 * See Section 5 for complete description of the data
1617 * flags - flags sent or received with the user message, see Section
1618 * 5 for complete description of the flags.
1620 * Note: This function could use a rewrite especially when explicit
1621 * connect support comes in.
1623 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1625 static int sctp_msghdr_parse(const struct msghdr *msg,
1626 struct sctp_cmsgs *cmsgs);
1628 static int sctp_sendmsg_parse(struct sock *sk, struct sctp_cmsgs *cmsgs,
1629 struct sctp_sndrcvinfo *srinfo,
1630 const struct msghdr *msg, size_t msg_len)
1635 if (sctp_sstate(sk, LISTENING) && sctp_style(sk, TCP))
1638 if (msg_len > sk->sk_sndbuf)
1641 memset(cmsgs, 0, sizeof(*cmsgs));
1642 err = sctp_msghdr_parse(msg, cmsgs);
1644 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1648 memset(srinfo, 0, sizeof(*srinfo));
1649 if (cmsgs->srinfo) {
1650 srinfo->sinfo_stream = cmsgs->srinfo->sinfo_stream;
1651 srinfo->sinfo_flags = cmsgs->srinfo->sinfo_flags;
1652 srinfo->sinfo_ppid = cmsgs->srinfo->sinfo_ppid;
1653 srinfo->sinfo_context = cmsgs->srinfo->sinfo_context;
1654 srinfo->sinfo_assoc_id = cmsgs->srinfo->sinfo_assoc_id;
1655 srinfo->sinfo_timetolive = cmsgs->srinfo->sinfo_timetolive;
1659 srinfo->sinfo_stream = cmsgs->sinfo->snd_sid;
1660 srinfo->sinfo_flags = cmsgs->sinfo->snd_flags;
1661 srinfo->sinfo_ppid = cmsgs->sinfo->snd_ppid;
1662 srinfo->sinfo_context = cmsgs->sinfo->snd_context;
1663 srinfo->sinfo_assoc_id = cmsgs->sinfo->snd_assoc_id;
1666 if (cmsgs->prinfo) {
1667 srinfo->sinfo_timetolive = cmsgs->prinfo->pr_value;
1668 SCTP_PR_SET_POLICY(srinfo->sinfo_flags,
1669 cmsgs->prinfo->pr_policy);
1672 sflags = srinfo->sinfo_flags;
1673 if (!sflags && msg_len)
1676 if (sctp_style(sk, TCP) && (sflags & (SCTP_EOF | SCTP_ABORT)))
1679 if (((sflags & SCTP_EOF) && msg_len > 0) ||
1680 (!(sflags & (SCTP_EOF | SCTP_ABORT)) && msg_len == 0))
1683 if ((sflags & SCTP_ADDR_OVER) && !msg->msg_name)
1689 static int sctp_sendmsg_new_asoc(struct sock *sk, __u16 sflags,
1690 struct sctp_cmsgs *cmsgs,
1691 union sctp_addr *daddr,
1692 struct sctp_transport **tp)
1694 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
1695 struct net *net = sock_net(sk);
1696 struct sctp_association *asoc;
1697 enum sctp_scope scope;
1698 struct cmsghdr *cmsg;
1704 if (sflags & (SCTP_EOF | SCTP_ABORT))
1707 if (sctp_style(sk, TCP) && (sctp_sstate(sk, ESTABLISHED) ||
1708 sctp_sstate(sk, CLOSING)))
1709 return -EADDRNOTAVAIL;
1711 if (sctp_endpoint_is_peeled_off(ep, daddr))
1712 return -EADDRNOTAVAIL;
1714 if (!ep->base.bind_addr.port) {
1715 if (sctp_autobind(sk))
1718 if (ep->base.bind_addr.port < inet_prot_sock(net) &&
1719 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
1723 scope = sctp_scope(daddr);
1725 /* Label connection socket for first association 1-to-many
1726 * style for client sequence socket()->sendmsg(). This
1727 * needs to be done before sctp_assoc_add_peer() as that will
1728 * set up the initial packet that needs to account for any
1729 * security ip options (CIPSO/CALIPSO) added to the packet.
1731 af = sctp_get_af_specific(daddr->sa.sa_family);
1734 err = security_sctp_bind_connect(sk, SCTP_SENDMSG_CONNECT,
1735 (struct sockaddr *)daddr,
1740 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1744 if (sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL) < 0) {
1750 struct sctp_initmsg *init = cmsgs->init;
1752 if (init->sinit_num_ostreams) {
1753 __u16 outcnt = init->sinit_num_ostreams;
1755 asoc->c.sinit_num_ostreams = outcnt;
1756 /* outcnt has been changed, need to re-init stream */
1757 err = sctp_stream_init(&asoc->stream, outcnt, 0,
1763 if (init->sinit_max_instreams)
1764 asoc->c.sinit_max_instreams = init->sinit_max_instreams;
1766 if (init->sinit_max_attempts)
1767 asoc->max_init_attempts = init->sinit_max_attempts;
1769 if (init->sinit_max_init_timeo)
1770 asoc->max_init_timeo =
1771 msecs_to_jiffies(init->sinit_max_init_timeo);
1774 *tp = sctp_assoc_add_peer(asoc, daddr, GFP_KERNEL, SCTP_UNKNOWN);
1780 if (!cmsgs->addrs_msg)
1783 /* sendv addr list parse */
1784 for_each_cmsghdr(cmsg, cmsgs->addrs_msg) {
1785 struct sctp_transport *transport;
1786 struct sctp_association *old;
1787 union sctp_addr _daddr;
1790 if (cmsg->cmsg_level != IPPROTO_SCTP ||
1791 (cmsg->cmsg_type != SCTP_DSTADDRV4 &&
1792 cmsg->cmsg_type != SCTP_DSTADDRV6))
1796 memset(daddr, 0, sizeof(*daddr));
1797 dlen = cmsg->cmsg_len - sizeof(struct cmsghdr);
1798 if (cmsg->cmsg_type == SCTP_DSTADDRV4) {
1799 if (dlen < sizeof(struct in_addr)) {
1804 dlen = sizeof(struct in_addr);
1805 daddr->v4.sin_family = AF_INET;
1806 daddr->v4.sin_port = htons(asoc->peer.port);
1807 memcpy(&daddr->v4.sin_addr, CMSG_DATA(cmsg), dlen);
1809 if (dlen < sizeof(struct in6_addr)) {
1814 dlen = sizeof(struct in6_addr);
1815 daddr->v6.sin6_family = AF_INET6;
1816 daddr->v6.sin6_port = htons(asoc->peer.port);
1817 memcpy(&daddr->v6.sin6_addr, CMSG_DATA(cmsg), dlen);
1819 err = sctp_verify_addr(sk, daddr, sizeof(*daddr));
1823 old = sctp_endpoint_lookup_assoc(ep, daddr, &transport);
1824 if (old && old != asoc) {
1825 if (old->state >= SCTP_STATE_ESTABLISHED)
1832 if (sctp_endpoint_is_peeled_off(ep, daddr)) {
1833 err = -EADDRNOTAVAIL;
1837 transport = sctp_assoc_add_peer(asoc, daddr, GFP_KERNEL,
1848 sctp_association_free(asoc);
1852 static int sctp_sendmsg_check_sflags(struct sctp_association *asoc,
1853 __u16 sflags, struct msghdr *msg,
1856 struct sock *sk = asoc->base.sk;
1857 struct net *net = sock_net(sk);
1859 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP))
1862 if ((sflags & SCTP_SENDALL) && sctp_style(sk, UDP) &&
1863 !sctp_state(asoc, ESTABLISHED))
1866 if (sflags & SCTP_EOF) {
1867 pr_debug("%s: shutting down association:%p\n", __func__, asoc);
1868 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1873 if (sflags & SCTP_ABORT) {
1874 struct sctp_chunk *chunk;
1876 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1880 pr_debug("%s: aborting association:%p\n", __func__, asoc);
1881 sctp_primitive_ABORT(net, asoc, chunk);
1889 static int sctp_sendmsg_to_asoc(struct sctp_association *asoc,
1890 struct msghdr *msg, size_t msg_len,
1891 struct sctp_transport *transport,
1892 struct sctp_sndrcvinfo *sinfo)
1894 struct sock *sk = asoc->base.sk;
1895 struct sctp_sock *sp = sctp_sk(sk);
1896 struct net *net = sock_net(sk);
1897 struct sctp_datamsg *datamsg;
1898 bool wait_connect = false;
1899 struct sctp_chunk *chunk;
1903 if (sinfo->sinfo_stream >= asoc->stream.outcnt) {
1908 if (unlikely(!asoc->stream.out[sinfo->sinfo_stream].ext)) {
1909 err = sctp_stream_init_ext(&asoc->stream, sinfo->sinfo_stream);
1914 if (sp->disable_fragments && msg_len > asoc->frag_point) {
1919 if (asoc->pmtu_pending) {
1920 if (sp->param_flags & SPP_PMTUD_ENABLE)
1921 sctp_assoc_sync_pmtu(asoc);
1922 asoc->pmtu_pending = 0;
1925 if (sctp_wspace(asoc) < msg_len)
1926 sctp_prsctp_prune(asoc, sinfo, msg_len - sctp_wspace(asoc));
1928 if (!sctp_wspace(asoc)) {
1929 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1930 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1935 if (sctp_state(asoc, CLOSED)) {
1936 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1940 if (sp->strm_interleave) {
1941 timeo = sock_sndtimeo(sk, 0);
1942 err = sctp_wait_for_connect(asoc, &timeo);
1946 wait_connect = true;
1949 pr_debug("%s: we associated primitively\n", __func__);
1952 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
1953 if (IS_ERR(datamsg)) {
1954 err = PTR_ERR(datamsg);
1958 asoc->force_delay = !!(msg->msg_flags & MSG_MORE);
1960 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1961 sctp_chunk_hold(chunk);
1962 sctp_set_owner_w(chunk);
1963 chunk->transport = transport;
1966 err = sctp_primitive_SEND(net, asoc, datamsg);
1968 sctp_datamsg_free(datamsg);
1972 pr_debug("%s: we sent primitively\n", __func__);
1974 sctp_datamsg_put(datamsg);
1976 if (unlikely(wait_connect)) {
1977 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1978 sctp_wait_for_connect(asoc, &timeo);
1987 static union sctp_addr *sctp_sendmsg_get_daddr(struct sock *sk,
1988 const struct msghdr *msg,
1989 struct sctp_cmsgs *cmsgs)
1991 union sctp_addr *daddr = NULL;
1994 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1995 int len = msg->msg_namelen;
1997 if (len > sizeof(*daddr))
1998 len = sizeof(*daddr);
2000 daddr = (union sctp_addr *)msg->msg_name;
2002 err = sctp_verify_addr(sk, daddr, len);
2004 return ERR_PTR(err);
2010 static void sctp_sendmsg_update_sinfo(struct sctp_association *asoc,
2011 struct sctp_sndrcvinfo *sinfo,
2012 struct sctp_cmsgs *cmsgs)
2014 if (!cmsgs->srinfo && !cmsgs->sinfo) {
2015 sinfo->sinfo_stream = asoc->default_stream;
2016 sinfo->sinfo_ppid = asoc->default_ppid;
2017 sinfo->sinfo_context = asoc->default_context;
2018 sinfo->sinfo_assoc_id = sctp_assoc2id(asoc);
2021 sinfo->sinfo_flags = asoc->default_flags;
2024 if (!cmsgs->srinfo && !cmsgs->prinfo)
2025 sinfo->sinfo_timetolive = asoc->default_timetolive;
2027 if (cmsgs->authinfo) {
2028 /* Reuse sinfo_tsn to indicate that authinfo was set and
2029 * sinfo_ssn to save the keyid on tx path.
2031 sinfo->sinfo_tsn = 1;
2032 sinfo->sinfo_ssn = cmsgs->authinfo->auth_keynumber;
2036 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
2038 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
2039 struct sctp_transport *transport = NULL;
2040 struct sctp_sndrcvinfo _sinfo, *sinfo;
2041 struct sctp_association *asoc;
2042 struct sctp_cmsgs cmsgs;
2043 union sctp_addr *daddr;
2048 /* Parse and get snd_info */
2049 err = sctp_sendmsg_parse(sk, &cmsgs, &_sinfo, msg, msg_len);
2054 sflags = sinfo->sinfo_flags;
2056 /* Get daddr from msg */
2057 daddr = sctp_sendmsg_get_daddr(sk, msg, &cmsgs);
2058 if (IS_ERR(daddr)) {
2059 err = PTR_ERR(daddr);
2065 /* SCTP_SENDALL process */
2066 if ((sflags & SCTP_SENDALL) && sctp_style(sk, UDP)) {
2067 list_for_each_entry(asoc, &ep->asocs, asocs) {
2068 err = sctp_sendmsg_check_sflags(asoc, sflags, msg,
2075 sctp_sendmsg_update_sinfo(asoc, sinfo, &cmsgs);
2077 err = sctp_sendmsg_to_asoc(asoc, msg, msg_len,
2082 iov_iter_revert(&msg->msg_iter, err);
2088 /* Get and check or create asoc */
2090 asoc = sctp_endpoint_lookup_assoc(ep, daddr, &transport);
2092 err = sctp_sendmsg_check_sflags(asoc, sflags, msg,
2097 err = sctp_sendmsg_new_asoc(sk, sflags, &cmsgs, daddr,
2102 asoc = transport->asoc;
2106 if (!sctp_style(sk, TCP) && !(sflags & SCTP_ADDR_OVER))
2109 asoc = sctp_id2assoc(sk, sinfo->sinfo_assoc_id);
2115 err = sctp_sendmsg_check_sflags(asoc, sflags, msg, msg_len);
2120 /* Update snd_info with the asoc */
2121 sctp_sendmsg_update_sinfo(asoc, sinfo, &cmsgs);
2123 /* Send msg to the asoc */
2124 err = sctp_sendmsg_to_asoc(asoc, msg, msg_len, transport, sinfo);
2125 if (err < 0 && err != -ESRCH && new)
2126 sctp_association_free(asoc);
2131 return sctp_error(sk, msg->msg_flags, err);
2134 /* This is an extended version of skb_pull() that removes the data from the
2135 * start of a skb even when data is spread across the list of skb's in the
2136 * frag_list. len specifies the total amount of data that needs to be removed.
2137 * when 'len' bytes could be removed from the skb, it returns 0.
2138 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2139 * could not be removed.
2141 static int sctp_skb_pull(struct sk_buff *skb, int len)
2143 struct sk_buff *list;
2144 int skb_len = skb_headlen(skb);
2147 if (len <= skb_len) {
2148 __skb_pull(skb, len);
2152 __skb_pull(skb, skb_len);
2154 skb_walk_frags(skb, list) {
2155 rlen = sctp_skb_pull(list, len);
2156 skb->len -= (len-rlen);
2157 skb->data_len -= (len-rlen);
2168 /* API 3.1.3 recvmsg() - UDP Style Syntax
2170 * ssize_t recvmsg(int socket, struct msghdr *message,
2173 * socket - the socket descriptor of the endpoint.
2174 * message - pointer to the msghdr structure which contains a single
2175 * user message and possibly some ancillary data.
2177 * See Section 5 for complete description of the data
2180 * flags - flags sent or received with the user message, see Section
2181 * 5 for complete description of the flags.
2183 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2184 int noblock, int flags, int *addr_len)
2186 struct sctp_ulpevent *event = NULL;
2187 struct sctp_sock *sp = sctp_sk(sk);
2188 struct sk_buff *skb, *head_skb;
2193 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2194 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2199 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED) &&
2200 !sctp_sstate(sk, CLOSING) && !sctp_sstate(sk, CLOSED)) {
2205 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2209 /* Get the total length of the skb including any skb's in the
2218 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2220 event = sctp_skb2event(skb);
2225 if (event->chunk && event->chunk->head_skb)
2226 head_skb = event->chunk->head_skb;
2229 sock_recv_ts_and_drops(msg, sk, head_skb);
2230 if (sctp_ulpevent_is_notification(event)) {
2231 msg->msg_flags |= MSG_NOTIFICATION;
2232 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2234 sp->pf->skb_msgname(head_skb, msg->msg_name, addr_len);
2237 /* Check if we allow SCTP_NXTINFO. */
2238 if (sp->recvnxtinfo)
2239 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2240 /* Check if we allow SCTP_RCVINFO. */
2241 if (sp->recvrcvinfo)
2242 sctp_ulpevent_read_rcvinfo(event, msg);
2243 /* Check if we allow SCTP_SNDRCVINFO. */
2244 if (sp->subscribe.sctp_data_io_event)
2245 sctp_ulpevent_read_sndrcvinfo(event, msg);
2249 /* If skb's length exceeds the user's buffer, update the skb and
2250 * push it back to the receive_queue so that the next call to
2251 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2253 if (skb_len > copied) {
2254 msg->msg_flags &= ~MSG_EOR;
2255 if (flags & MSG_PEEK)
2257 sctp_skb_pull(skb, copied);
2258 skb_queue_head(&sk->sk_receive_queue, skb);
2260 /* When only partial message is copied to the user, increase
2261 * rwnd by that amount. If all the data in the skb is read,
2262 * rwnd is updated when the event is freed.
2264 if (!sctp_ulpevent_is_notification(event))
2265 sctp_assoc_rwnd_increase(event->asoc, copied);
2267 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2268 (event->msg_flags & MSG_EOR))
2269 msg->msg_flags |= MSG_EOR;
2271 msg->msg_flags &= ~MSG_EOR;
2274 if (flags & MSG_PEEK) {
2275 /* Release the skb reference acquired after peeking the skb in
2276 * sctp_skb_recv_datagram().
2280 /* Free the event which includes releasing the reference to
2281 * the owner of the skb, freeing the skb and updating the
2284 sctp_ulpevent_free(event);
2291 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2293 * This option is a on/off flag. If enabled no SCTP message
2294 * fragmentation will be performed. Instead if a message being sent
2295 * exceeds the current PMTU size, the message will NOT be sent and
2296 * instead a error will be indicated to the user.
2298 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2299 char __user *optval,
2300 unsigned int optlen)
2304 if (optlen < sizeof(int))
2307 if (get_user(val, (int __user *)optval))
2310 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2315 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2316 unsigned int optlen)
2318 struct sctp_association *asoc;
2319 struct sctp_ulpevent *event;
2321 if (optlen > sizeof(struct sctp_event_subscribe))
2323 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2326 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2327 * if there is no data to be sent or retransmit, the stack will
2328 * immediately send up this notification.
2330 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2331 &sctp_sk(sk)->subscribe)) {
2332 asoc = sctp_id2assoc(sk, 0);
2334 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2335 event = sctp_ulpevent_make_sender_dry_event(asoc,
2336 GFP_USER | __GFP_NOWARN);
2340 asoc->stream.si->enqueue_event(&asoc->ulpq, event);
2347 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2349 * This socket option is applicable to the UDP-style socket only. When
2350 * set it will cause associations that are idle for more than the
2351 * specified number of seconds to automatically close. An association
2352 * being idle is defined an association that has NOT sent or received
2353 * user data. The special value of '0' indicates that no automatic
2354 * close of any associations should be performed. The option expects an
2355 * integer defining the number of seconds of idle time before an
2356 * association is closed.
2358 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2359 unsigned int optlen)
2361 struct sctp_sock *sp = sctp_sk(sk);
2362 struct net *net = sock_net(sk);
2364 /* Applicable to UDP-style socket only */
2365 if (sctp_style(sk, TCP))
2367 if (optlen != sizeof(int))
2369 if (copy_from_user(&sp->autoclose, optval, optlen))
2372 if (sp->autoclose > net->sctp.max_autoclose)
2373 sp->autoclose = net->sctp.max_autoclose;
2378 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2380 * Applications can enable or disable heartbeats for any peer address of
2381 * an association, modify an address's heartbeat interval, force a
2382 * heartbeat to be sent immediately, and adjust the address's maximum
2383 * number of retransmissions sent before an address is considered
2384 * unreachable. The following structure is used to access and modify an
2385 * address's parameters:
2387 * struct sctp_paddrparams {
2388 * sctp_assoc_t spp_assoc_id;
2389 * struct sockaddr_storage spp_address;
2390 * uint32_t spp_hbinterval;
2391 * uint16_t spp_pathmaxrxt;
2392 * uint32_t spp_pathmtu;
2393 * uint32_t spp_sackdelay;
2394 * uint32_t spp_flags;
2397 * spp_assoc_id - (one-to-many style socket) This is filled in the
2398 * application, and identifies the association for
2400 * spp_address - This specifies which address is of interest.
2401 * spp_hbinterval - This contains the value of the heartbeat interval,
2402 * in milliseconds. If a value of zero
2403 * is present in this field then no changes are to
2404 * be made to this parameter.
2405 * spp_pathmaxrxt - This contains the maximum number of
2406 * retransmissions before this address shall be
2407 * considered unreachable. If a value of zero
2408 * is present in this field then no changes are to
2409 * be made to this parameter.
2410 * spp_pathmtu - When Path MTU discovery is disabled the value
2411 * specified here will be the "fixed" path mtu.
2412 * Note that if the spp_address field is empty
2413 * then all associations on this address will
2414 * have this fixed path mtu set upon them.
2416 * spp_sackdelay - When delayed sack is enabled, this value specifies
2417 * the number of milliseconds that sacks will be delayed
2418 * for. This value will apply to all addresses of an
2419 * association if the spp_address field is empty. Note
2420 * also, that if delayed sack is enabled and this
2421 * value is set to 0, no change is made to the last
2422 * recorded delayed sack timer value.
2424 * spp_flags - These flags are used to control various features
2425 * on an association. The flag field may contain
2426 * zero or more of the following options.
2428 * SPP_HB_ENABLE - Enable heartbeats on the
2429 * specified address. Note that if the address
2430 * field is empty all addresses for the association
2431 * have heartbeats enabled upon them.
2433 * SPP_HB_DISABLE - Disable heartbeats on the
2434 * speicifed address. Note that if the address
2435 * field is empty all addresses for the association
2436 * will have their heartbeats disabled. Note also
2437 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2438 * mutually exclusive, only one of these two should
2439 * be specified. Enabling both fields will have
2440 * undetermined results.
2442 * SPP_HB_DEMAND - Request a user initiated heartbeat
2443 * to be made immediately.
2445 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2446 * heartbeat delayis to be set to the value of 0
2449 * SPP_PMTUD_ENABLE - This field will enable PMTU
2450 * discovery upon the specified address. Note that
2451 * if the address feild is empty then all addresses
2452 * on the association are effected.
2454 * SPP_PMTUD_DISABLE - This field will disable PMTU
2455 * discovery upon the specified address. Note that
2456 * if the address feild is empty then all addresses
2457 * on the association are effected. Not also that
2458 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2459 * exclusive. Enabling both will have undetermined
2462 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2463 * on delayed sack. The time specified in spp_sackdelay
2464 * is used to specify the sack delay for this address. Note
2465 * that if spp_address is empty then all addresses will
2466 * enable delayed sack and take on the sack delay
2467 * value specified in spp_sackdelay.
2468 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2469 * off delayed sack. If the spp_address field is blank then
2470 * delayed sack is disabled for the entire association. Note
2471 * also that this field is mutually exclusive to
2472 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2475 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2476 struct sctp_transport *trans,
2477 struct sctp_association *asoc,
2478 struct sctp_sock *sp,
2481 int sackdelay_change)
2485 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2486 struct net *net = sock_net(trans->asoc->base.sk);
2488 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2493 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2494 * this field is ignored. Note also that a value of zero indicates
2495 * the current setting should be left unchanged.
2497 if (params->spp_flags & SPP_HB_ENABLE) {
2499 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2500 * set. This lets us use 0 value when this flag
2503 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2504 params->spp_hbinterval = 0;
2506 if (params->spp_hbinterval ||
2507 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2510 msecs_to_jiffies(params->spp_hbinterval);
2513 msecs_to_jiffies(params->spp_hbinterval);
2515 sp->hbinterval = params->spp_hbinterval;
2522 trans->param_flags =
2523 (trans->param_flags & ~SPP_HB) | hb_change;
2526 (asoc->param_flags & ~SPP_HB) | hb_change;
2529 (sp->param_flags & ~SPP_HB) | hb_change;
2533 /* When Path MTU discovery is disabled the value specified here will
2534 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2535 * include the flag SPP_PMTUD_DISABLE for this field to have any
2538 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2540 trans->pathmtu = params->spp_pathmtu;
2541 sctp_assoc_sync_pmtu(asoc);
2543 sctp_assoc_set_pmtu(asoc, params->spp_pathmtu);
2545 sp->pathmtu = params->spp_pathmtu;
2551 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2552 (params->spp_flags & SPP_PMTUD_ENABLE);
2553 trans->param_flags =
2554 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2556 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2557 sctp_assoc_sync_pmtu(asoc);
2561 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2564 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2568 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2569 * value of this field is ignored. Note also that a value of zero
2570 * indicates the current setting should be left unchanged.
2572 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2575 msecs_to_jiffies(params->spp_sackdelay);
2578 msecs_to_jiffies(params->spp_sackdelay);
2580 sp->sackdelay = params->spp_sackdelay;
2584 if (sackdelay_change) {
2586 trans->param_flags =
2587 (trans->param_flags & ~SPP_SACKDELAY) |
2591 (asoc->param_flags & ~SPP_SACKDELAY) |
2595 (sp->param_flags & ~SPP_SACKDELAY) |
2600 /* Note that a value of zero indicates the current setting should be
2603 if (params->spp_pathmaxrxt) {
2605 trans->pathmaxrxt = params->spp_pathmaxrxt;
2607 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2609 sp->pathmaxrxt = params->spp_pathmaxrxt;
2616 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2617 char __user *optval,
2618 unsigned int optlen)
2620 struct sctp_paddrparams params;
2621 struct sctp_transport *trans = NULL;
2622 struct sctp_association *asoc = NULL;
2623 struct sctp_sock *sp = sctp_sk(sk);
2625 int hb_change, pmtud_change, sackdelay_change;
2627 if (optlen != sizeof(struct sctp_paddrparams))
2630 if (copy_from_user(¶ms, optval, optlen))
2633 /* Validate flags and value parameters. */
2634 hb_change = params.spp_flags & SPP_HB;
2635 pmtud_change = params.spp_flags & SPP_PMTUD;
2636 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2638 if (hb_change == SPP_HB ||
2639 pmtud_change == SPP_PMTUD ||
2640 sackdelay_change == SPP_SACKDELAY ||
2641 params.spp_sackdelay > 500 ||
2642 (params.spp_pathmtu &&
2643 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2646 /* If an address other than INADDR_ANY is specified, and
2647 * no transport is found, then the request is invalid.
2649 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
2650 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2651 params.spp_assoc_id);
2656 /* Get association, if assoc_id != 0 and the socket is a one
2657 * to many style socket, and an association was not found, then
2658 * the id was invalid.
2660 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2661 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2664 /* Heartbeat demand can only be sent on a transport or
2665 * association, but not a socket.
2667 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2670 /* Process parameters. */
2671 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2672 hb_change, pmtud_change,
2678 /* If changes are for association, also apply parameters to each
2681 if (!trans && asoc) {
2682 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2684 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2685 hb_change, pmtud_change,
2693 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2695 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2698 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2700 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2704 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2706 * This option will effect the way delayed acks are performed. This
2707 * option allows you to get or set the delayed ack time, in
2708 * milliseconds. It also allows changing the delayed ack frequency.
2709 * Changing the frequency to 1 disables the delayed sack algorithm. If
2710 * the assoc_id is 0, then this sets or gets the endpoints default
2711 * values. If the assoc_id field is non-zero, then the set or get
2712 * effects the specified association for the one to many model (the
2713 * assoc_id field is ignored by the one to one model). Note that if
2714 * sack_delay or sack_freq are 0 when setting this option, then the
2715 * current values will remain unchanged.
2717 * struct sctp_sack_info {
2718 * sctp_assoc_t sack_assoc_id;
2719 * uint32_t sack_delay;
2720 * uint32_t sack_freq;
2723 * sack_assoc_id - This parameter, indicates which association the user
2724 * is performing an action upon. Note that if this field's value is
2725 * zero then the endpoints default value is changed (effecting future
2726 * associations only).
2728 * sack_delay - This parameter contains the number of milliseconds that
2729 * the user is requesting the delayed ACK timer be set to. Note that
2730 * this value is defined in the standard to be between 200 and 500
2733 * sack_freq - This parameter contains the number of packets that must
2734 * be received before a sack is sent without waiting for the delay
2735 * timer to expire. The default value for this is 2, setting this
2736 * value to 1 will disable the delayed sack algorithm.
2739 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2740 char __user *optval, unsigned int optlen)
2742 struct sctp_sack_info params;
2743 struct sctp_transport *trans = NULL;
2744 struct sctp_association *asoc = NULL;
2745 struct sctp_sock *sp = sctp_sk(sk);
2747 if (optlen == sizeof(struct sctp_sack_info)) {
2748 if (copy_from_user(¶ms, optval, optlen))
2751 if (params.sack_delay == 0 && params.sack_freq == 0)
2753 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2754 pr_warn_ratelimited(DEPRECATED
2756 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2757 "Use struct sctp_sack_info instead\n",
2758 current->comm, task_pid_nr(current));
2759 if (copy_from_user(¶ms, optval, optlen))
2762 if (params.sack_delay == 0)
2763 params.sack_freq = 1;
2765 params.sack_freq = 0;
2769 /* Validate value parameter. */
2770 if (params.sack_delay > 500)
2773 /* Get association, if sack_assoc_id != 0 and the socket is a one
2774 * to many style socket, and an association was not found, then
2775 * the id was invalid.
2777 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2778 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2781 if (params.sack_delay) {
2784 msecs_to_jiffies(params.sack_delay);
2786 sctp_spp_sackdelay_enable(asoc->param_flags);
2788 sp->sackdelay = params.sack_delay;
2790 sctp_spp_sackdelay_enable(sp->param_flags);
2794 if (params.sack_freq == 1) {
2797 sctp_spp_sackdelay_disable(asoc->param_flags);
2800 sctp_spp_sackdelay_disable(sp->param_flags);
2802 } else if (params.sack_freq > 1) {
2804 asoc->sackfreq = params.sack_freq;
2806 sctp_spp_sackdelay_enable(asoc->param_flags);
2808 sp->sackfreq = params.sack_freq;
2810 sctp_spp_sackdelay_enable(sp->param_flags);
2814 /* If change is for association, also apply to each transport. */
2816 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2818 if (params.sack_delay) {
2820 msecs_to_jiffies(params.sack_delay);
2821 trans->param_flags =
2822 sctp_spp_sackdelay_enable(trans->param_flags);
2824 if (params.sack_freq == 1) {
2825 trans->param_flags =
2826 sctp_spp_sackdelay_disable(trans->param_flags);
2827 } else if (params.sack_freq > 1) {
2828 trans->sackfreq = params.sack_freq;
2829 trans->param_flags =
2830 sctp_spp_sackdelay_enable(trans->param_flags);
2838 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2840 * Applications can specify protocol parameters for the default association
2841 * initialization. The option name argument to setsockopt() and getsockopt()
2844 * Setting initialization parameters is effective only on an unconnected
2845 * socket (for UDP-style sockets only future associations are effected
2846 * by the change). With TCP-style sockets, this option is inherited by
2847 * sockets derived from a listener socket.
2849 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2851 struct sctp_initmsg sinit;
2852 struct sctp_sock *sp = sctp_sk(sk);
2854 if (optlen != sizeof(struct sctp_initmsg))
2856 if (copy_from_user(&sinit, optval, optlen))
2859 if (sinit.sinit_num_ostreams)
2860 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2861 if (sinit.sinit_max_instreams)
2862 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2863 if (sinit.sinit_max_attempts)
2864 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2865 if (sinit.sinit_max_init_timeo)
2866 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2872 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2874 * Applications that wish to use the sendto() system call may wish to
2875 * specify a default set of parameters that would normally be supplied
2876 * through the inclusion of ancillary data. This socket option allows
2877 * such an application to set the default sctp_sndrcvinfo structure.
2878 * The application that wishes to use this socket option simply passes
2879 * in to this call the sctp_sndrcvinfo structure defined in Section
2880 * 5.2.2) The input parameters accepted by this call include
2881 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2882 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2883 * to this call if the caller is using the UDP model.
2885 static int sctp_setsockopt_default_send_param(struct sock *sk,
2886 char __user *optval,
2887 unsigned int optlen)
2889 struct sctp_sock *sp = sctp_sk(sk);
2890 struct sctp_association *asoc;
2891 struct sctp_sndrcvinfo info;
2893 if (optlen != sizeof(info))
2895 if (copy_from_user(&info, optval, optlen))
2897 if (info.sinfo_flags &
2898 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2899 SCTP_ABORT | SCTP_EOF))
2902 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2903 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2906 asoc->default_stream = info.sinfo_stream;
2907 asoc->default_flags = info.sinfo_flags;
2908 asoc->default_ppid = info.sinfo_ppid;
2909 asoc->default_context = info.sinfo_context;
2910 asoc->default_timetolive = info.sinfo_timetolive;
2912 sp->default_stream = info.sinfo_stream;
2913 sp->default_flags = info.sinfo_flags;
2914 sp->default_ppid = info.sinfo_ppid;
2915 sp->default_context = info.sinfo_context;
2916 sp->default_timetolive = info.sinfo_timetolive;
2922 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2923 * (SCTP_DEFAULT_SNDINFO)
2925 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
2926 char __user *optval,
2927 unsigned int optlen)
2929 struct sctp_sock *sp = sctp_sk(sk);
2930 struct sctp_association *asoc;
2931 struct sctp_sndinfo info;
2933 if (optlen != sizeof(info))
2935 if (copy_from_user(&info, optval, optlen))
2937 if (info.snd_flags &
2938 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2939 SCTP_ABORT | SCTP_EOF))
2942 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
2943 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
2946 asoc->default_stream = info.snd_sid;
2947 asoc->default_flags = info.snd_flags;
2948 asoc->default_ppid = info.snd_ppid;
2949 asoc->default_context = info.snd_context;
2951 sp->default_stream = info.snd_sid;
2952 sp->default_flags = info.snd_flags;
2953 sp->default_ppid = info.snd_ppid;
2954 sp->default_context = info.snd_context;
2960 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2962 * Requests that the local SCTP stack use the enclosed peer address as
2963 * the association primary. The enclosed address must be one of the
2964 * association peer's addresses.
2966 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2967 unsigned int optlen)
2969 struct sctp_prim prim;
2970 struct sctp_transport *trans;
2974 if (optlen != sizeof(struct sctp_prim))
2977 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2980 /* Allow security module to validate address but need address len. */
2981 af = sctp_get_af_specific(prim.ssp_addr.ss_family);
2985 err = security_sctp_bind_connect(sk, SCTP_PRIMARY_ADDR,
2986 (struct sockaddr *)&prim.ssp_addr,
2991 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2995 sctp_assoc_set_primary(trans->asoc, trans);
3001 * 7.1.5 SCTP_NODELAY
3003 * Turn on/off any Nagle-like algorithm. This means that packets are
3004 * generally sent as soon as possible and no unnecessary delays are
3005 * introduced, at the cost of more packets in the network. Expects an
3006 * integer boolean flag.
3008 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
3009 unsigned int optlen)
3013 if (optlen < sizeof(int))
3015 if (get_user(val, (int __user *)optval))
3018 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
3024 * 7.1.1 SCTP_RTOINFO
3026 * The protocol parameters used to initialize and bound retransmission
3027 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
3028 * and modify these parameters.
3029 * All parameters are time values, in milliseconds. A value of 0, when
3030 * modifying the parameters, indicates that the current value should not
3034 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
3036 struct sctp_rtoinfo rtoinfo;
3037 struct sctp_association *asoc;
3038 unsigned long rto_min, rto_max;
3039 struct sctp_sock *sp = sctp_sk(sk);
3041 if (optlen != sizeof (struct sctp_rtoinfo))
3044 if (copy_from_user(&rtoinfo, optval, optlen))
3047 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
3049 /* Set the values to the specific association */
3050 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
3053 rto_max = rtoinfo.srto_max;
3054 rto_min = rtoinfo.srto_min;
3057 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
3059 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
3062 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
3064 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
3066 if (rto_min > rto_max)
3070 if (rtoinfo.srto_initial != 0)
3072 msecs_to_jiffies(rtoinfo.srto_initial);
3073 asoc->rto_max = rto_max;
3074 asoc->rto_min = rto_min;
3076 /* If there is no association or the association-id = 0
3077 * set the values to the endpoint.
3079 if (rtoinfo.srto_initial != 0)
3080 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
3081 sp->rtoinfo.srto_max = rto_max;
3082 sp->rtoinfo.srto_min = rto_min;
3090 * 7.1.2 SCTP_ASSOCINFO
3092 * This option is used to tune the maximum retransmission attempts
3093 * of the association.
3094 * Returns an error if the new association retransmission value is
3095 * greater than the sum of the retransmission value of the peer.
3096 * See [SCTP] for more information.
3099 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
3102 struct sctp_assocparams assocparams;
3103 struct sctp_association *asoc;
3105 if (optlen != sizeof(struct sctp_assocparams))
3107 if (copy_from_user(&assocparams, optval, optlen))
3110 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
3112 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
3115 /* Set the values to the specific association */
3117 if (assocparams.sasoc_asocmaxrxt != 0) {
3120 struct sctp_transport *peer_addr;
3122 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
3124 path_sum += peer_addr->pathmaxrxt;
3128 /* Only validate asocmaxrxt if we have more than
3129 * one path/transport. We do this because path
3130 * retransmissions are only counted when we have more
3134 assocparams.sasoc_asocmaxrxt > path_sum)
3137 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3140 if (assocparams.sasoc_cookie_life != 0)
3141 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3143 /* Set the values to the endpoint */
3144 struct sctp_sock *sp = sctp_sk(sk);
3146 if (assocparams.sasoc_asocmaxrxt != 0)
3147 sp->assocparams.sasoc_asocmaxrxt =
3148 assocparams.sasoc_asocmaxrxt;
3149 if (assocparams.sasoc_cookie_life != 0)
3150 sp->assocparams.sasoc_cookie_life =
3151 assocparams.sasoc_cookie_life;
3157 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3159 * This socket option is a boolean flag which turns on or off mapped V4
3160 * addresses. If this option is turned on and the socket is type
3161 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3162 * If this option is turned off, then no mapping will be done of V4
3163 * addresses and a user will receive both PF_INET6 and PF_INET type
3164 * addresses on the socket.
3166 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3169 struct sctp_sock *sp = sctp_sk(sk);
3171 if (optlen < sizeof(int))
3173 if (get_user(val, (int __user *)optval))
3184 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3185 * This option will get or set the maximum size to put in any outgoing
3186 * SCTP DATA chunk. If a message is larger than this size it will be
3187 * fragmented by SCTP into the specified size. Note that the underlying
3188 * SCTP implementation may fragment into smaller sized chunks when the
3189 * PMTU of the underlying association is smaller than the value set by
3190 * the user. The default value for this option is '0' which indicates
3191 * the user is NOT limiting fragmentation and only the PMTU will effect
3192 * SCTP's choice of DATA chunk size. Note also that values set larger
3193 * than the maximum size of an IP datagram will effectively let SCTP
3194 * control fragmentation (i.e. the same as setting this option to 0).
3196 * The following structure is used to access and modify this parameter:
3198 * struct sctp_assoc_value {
3199 * sctp_assoc_t assoc_id;
3200 * uint32_t assoc_value;
3203 * assoc_id: This parameter is ignored for one-to-one style sockets.
3204 * For one-to-many style sockets this parameter indicates which
3205 * association the user is performing an action upon. Note that if
3206 * this field's value is zero then the endpoints default value is
3207 * changed (effecting future associations only).
3208 * assoc_value: This parameter specifies the maximum size in bytes.
3210 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3212 struct sctp_sock *sp = sctp_sk(sk);
3213 struct sctp_assoc_value params;
3214 struct sctp_association *asoc;
3217 if (optlen == sizeof(int)) {
3218 pr_warn_ratelimited(DEPRECATED
3220 "Use of int in maxseg socket option.\n"
3221 "Use struct sctp_assoc_value instead\n",
3222 current->comm, task_pid_nr(current));
3223 if (copy_from_user(&val, optval, optlen))
3225 params.assoc_id = 0;
3226 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3227 if (copy_from_user(¶ms, optval, optlen))
3229 val = params.assoc_value;
3234 asoc = sctp_id2assoc(sk, params.assoc_id);
3237 int min_len, max_len;
3238 __u16 datasize = asoc ? sctp_datachk_len(&asoc->stream) :
3239 sizeof(struct sctp_data_chunk);
3241 min_len = sctp_mtu_payload(sp, SCTP_DEFAULT_MINSEGMENT,
3243 max_len = SCTP_MAX_CHUNK_LEN - datasize;
3245 if (val < min_len || val > max_len)
3250 asoc->user_frag = val;
3251 sctp_assoc_update_frag_point(asoc);
3253 if (params.assoc_id && sctp_style(sk, UDP))
3255 sp->user_frag = val;
3263 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3265 * Requests that the peer mark the enclosed address as the association
3266 * primary. The enclosed address must be one of the association's
3267 * locally bound addresses. The following structure is used to make a
3268 * set primary request:
3270 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3271 unsigned int optlen)
3273 struct net *net = sock_net(sk);
3274 struct sctp_sock *sp;
3275 struct sctp_association *asoc = NULL;
3276 struct sctp_setpeerprim prim;
3277 struct sctp_chunk *chunk;
3283 if (!net->sctp.addip_enable)
3286 if (optlen != sizeof(struct sctp_setpeerprim))
3289 if (copy_from_user(&prim, optval, optlen))
3292 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3296 if (!asoc->peer.asconf_capable)
3299 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3302 if (!sctp_state(asoc, ESTABLISHED))
3305 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3309 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3310 return -EADDRNOTAVAIL;
3312 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3313 return -EADDRNOTAVAIL;
3315 /* Allow security module to validate address. */
3316 err = security_sctp_bind_connect(sk, SCTP_SET_PEER_PRIMARY_ADDR,
3317 (struct sockaddr *)&prim.sspp_addr,
3322 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3323 chunk = sctp_make_asconf_set_prim(asoc,
3324 (union sctp_addr *)&prim.sspp_addr);
3328 err = sctp_send_asconf(asoc, chunk);
3330 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3335 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3336 unsigned int optlen)
3338 struct sctp_setadaptation adaptation;
3340 if (optlen != sizeof(struct sctp_setadaptation))
3342 if (copy_from_user(&adaptation, optval, optlen))
3345 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3351 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3353 * The context field in the sctp_sndrcvinfo structure is normally only
3354 * used when a failed message is retrieved holding the value that was
3355 * sent down on the actual send call. This option allows the setting of
3356 * a default context on an association basis that will be received on
3357 * reading messages from the peer. This is especially helpful in the
3358 * one-2-many model for an application to keep some reference to an
3359 * internal state machine that is processing messages on the
3360 * association. Note that the setting of this value only effects
3361 * received messages from the peer and does not effect the value that is
3362 * saved with outbound messages.
3364 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3365 unsigned int optlen)
3367 struct sctp_assoc_value params;
3368 struct sctp_sock *sp;
3369 struct sctp_association *asoc;
3371 if (optlen != sizeof(struct sctp_assoc_value))
3373 if (copy_from_user(¶ms, optval, optlen))
3378 if (params.assoc_id != 0) {
3379 asoc = sctp_id2assoc(sk, params.assoc_id);
3382 asoc->default_rcv_context = params.assoc_value;
3384 sp->default_rcv_context = params.assoc_value;
3391 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3393 * This options will at a minimum specify if the implementation is doing
3394 * fragmented interleave. Fragmented interleave, for a one to many
3395 * socket, is when subsequent calls to receive a message may return
3396 * parts of messages from different associations. Some implementations
3397 * may allow you to turn this value on or off. If so, when turned off,
3398 * no fragment interleave will occur (which will cause a head of line
3399 * blocking amongst multiple associations sharing the same one to many
3400 * socket). When this option is turned on, then each receive call may
3401 * come from a different association (thus the user must receive data
3402 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3403 * association each receive belongs to.
3405 * This option takes a boolean value. A non-zero value indicates that
3406 * fragmented interleave is on. A value of zero indicates that
3407 * fragmented interleave is off.
3409 * Note that it is important that an implementation that allows this
3410 * option to be turned on, have it off by default. Otherwise an unaware
3411 * application using the one to many model may become confused and act
3414 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3415 char __user *optval,
3416 unsigned int optlen)
3420 if (optlen != sizeof(int))
3422 if (get_user(val, (int __user *)optval))
3425 sctp_sk(sk)->frag_interleave = !!val;
3427 if (!sctp_sk(sk)->frag_interleave)
3428 sctp_sk(sk)->strm_interleave = 0;
3434 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3435 * (SCTP_PARTIAL_DELIVERY_POINT)
3437 * This option will set or get the SCTP partial delivery point. This
3438 * point is the size of a message where the partial delivery API will be
3439 * invoked to help free up rwnd space for the peer. Setting this to a
3440 * lower value will cause partial deliveries to happen more often. The
3441 * calls argument is an integer that sets or gets the partial delivery
3442 * point. Note also that the call will fail if the user attempts to set
3443 * this value larger than the socket receive buffer size.
3445 * Note that any single message having a length smaller than or equal to
3446 * the SCTP partial delivery point will be delivered in one single read
3447 * call as long as the user provided buffer is large enough to hold the
3450 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3451 char __user *optval,
3452 unsigned int optlen)
3456 if (optlen != sizeof(u32))
3458 if (get_user(val, (int __user *)optval))
3461 /* Note: We double the receive buffer from what the user sets
3462 * it to be, also initial rwnd is based on rcvbuf/2.
3464 if (val > (sk->sk_rcvbuf >> 1))
3467 sctp_sk(sk)->pd_point = val;
3469 return 0; /* is this the right error code? */
3473 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3475 * This option will allow a user to change the maximum burst of packets
3476 * that can be emitted by this association. Note that the default value
3477 * is 4, and some implementations may restrict this setting so that it
3478 * can only be lowered.
3480 * NOTE: This text doesn't seem right. Do this on a socket basis with
3481 * future associations inheriting the socket value.
3483 static int sctp_setsockopt_maxburst(struct sock *sk,
3484 char __user *optval,
3485 unsigned int optlen)
3487 struct sctp_assoc_value params;
3488 struct sctp_sock *sp;
3489 struct sctp_association *asoc;
3493 if (optlen == sizeof(int)) {
3494 pr_warn_ratelimited(DEPRECATED
3496 "Use of int in max_burst socket option deprecated.\n"
3497 "Use struct sctp_assoc_value instead\n",
3498 current->comm, task_pid_nr(current));
3499 if (copy_from_user(&val, optval, optlen))
3501 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3502 if (copy_from_user(¶ms, optval, optlen))
3504 val = params.assoc_value;
3505 assoc_id = params.assoc_id;
3511 if (assoc_id != 0) {
3512 asoc = sctp_id2assoc(sk, assoc_id);
3515 asoc->max_burst = val;
3517 sp->max_burst = val;
3523 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3525 * This set option adds a chunk type that the user is requesting to be
3526 * received only in an authenticated way. Changes to the list of chunks
3527 * will only effect future associations on the socket.
3529 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3530 char __user *optval,
3531 unsigned int optlen)
3533 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3534 struct sctp_authchunk val;
3536 if (!ep->auth_enable)
3539 if (optlen != sizeof(struct sctp_authchunk))
3541 if (copy_from_user(&val, optval, optlen))
3544 switch (val.sauth_chunk) {
3546 case SCTP_CID_INIT_ACK:
3547 case SCTP_CID_SHUTDOWN_COMPLETE:
3552 /* add this chunk id to the endpoint */
3553 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3557 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3559 * This option gets or sets the list of HMAC algorithms that the local
3560 * endpoint requires the peer to use.
3562 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3563 char __user *optval,
3564 unsigned int optlen)
3566 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3567 struct sctp_hmacalgo *hmacs;
3571 if (!ep->auth_enable)
3574 if (optlen < sizeof(struct sctp_hmacalgo))
3576 optlen = min_t(unsigned int, optlen, sizeof(struct sctp_hmacalgo) +
3577 SCTP_AUTH_NUM_HMACS * sizeof(u16));
3579 hmacs = memdup_user(optval, optlen);
3581 return PTR_ERR(hmacs);
3583 idents = hmacs->shmac_num_idents;
3584 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3585 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3590 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3597 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3599 * This option will set a shared secret key which is used to build an
3600 * association shared key.
3602 static int sctp_setsockopt_auth_key(struct sock *sk,
3603 char __user *optval,
3604 unsigned int optlen)
3606 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3607 struct sctp_authkey *authkey;
3608 struct sctp_association *asoc;
3611 if (!ep->auth_enable)
3614 if (optlen <= sizeof(struct sctp_authkey))
3616 /* authkey->sca_keylength is u16, so optlen can't be bigger than
3619 optlen = min_t(unsigned int, optlen, USHRT_MAX +
3620 sizeof(struct sctp_authkey));
3622 authkey = memdup_user(optval, optlen);
3623 if (IS_ERR(authkey))
3624 return PTR_ERR(authkey);
3626 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3631 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3632 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3637 ret = sctp_auth_set_key(ep, asoc, authkey);
3644 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3646 * This option will get or set the active shared key to be used to build
3647 * the association shared key.
3649 static int sctp_setsockopt_active_key(struct sock *sk,
3650 char __user *optval,
3651 unsigned int optlen)
3653 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3654 struct sctp_authkeyid val;
3655 struct sctp_association *asoc;
3657 if (!ep->auth_enable)
3660 if (optlen != sizeof(struct sctp_authkeyid))
3662 if (copy_from_user(&val, optval, optlen))
3665 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3666 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3669 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3673 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3675 * This set option will delete a shared secret key from use.
3677 static int sctp_setsockopt_del_key(struct sock *sk,
3678 char __user *optval,
3679 unsigned int optlen)
3681 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3682 struct sctp_authkeyid val;
3683 struct sctp_association *asoc;
3685 if (!ep->auth_enable)
3688 if (optlen != sizeof(struct sctp_authkeyid))
3690 if (copy_from_user(&val, optval, optlen))
3693 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3694 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3697 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3702 * 8.3.4 Deactivate a Shared Key (SCTP_AUTH_DEACTIVATE_KEY)
3704 * This set option will deactivate a shared secret key.
3706 static int sctp_setsockopt_deactivate_key(struct sock *sk, char __user *optval,
3707 unsigned int optlen)
3709 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3710 struct sctp_authkeyid val;
3711 struct sctp_association *asoc;
3713 if (!ep->auth_enable)
3716 if (optlen != sizeof(struct sctp_authkeyid))
3718 if (copy_from_user(&val, optval, optlen))
3721 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3722 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3725 return sctp_auth_deact_key_id(ep, asoc, val.scact_keynumber);
3729 * 8.1.23 SCTP_AUTO_ASCONF
3731 * This option will enable or disable the use of the automatic generation of
3732 * ASCONF chunks to add and delete addresses to an existing association. Note
3733 * that this option has two caveats namely: a) it only affects sockets that
3734 * are bound to all addresses available to the SCTP stack, and b) the system
3735 * administrator may have an overriding control that turns the ASCONF feature
3736 * off no matter what setting the socket option may have.
3737 * This option expects an integer boolean flag, where a non-zero value turns on
3738 * the option, and a zero value turns off the option.
3739 * Note. In this implementation, socket operation overrides default parameter
3740 * being set by sysctl as well as FreeBSD implementation
3742 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3743 unsigned int optlen)
3746 struct sctp_sock *sp = sctp_sk(sk);
3748 if (optlen < sizeof(int))
3750 if (get_user(val, (int __user *)optval))
3752 if (!sctp_is_ep_boundall(sk) && val)
3754 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3757 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3758 if (val == 0 && sp->do_auto_asconf) {
3759 list_del(&sp->auto_asconf_list);
3760 sp->do_auto_asconf = 0;
3761 } else if (val && !sp->do_auto_asconf) {
3762 list_add_tail(&sp->auto_asconf_list,
3763 &sock_net(sk)->sctp.auto_asconf_splist);
3764 sp->do_auto_asconf = 1;
3766 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3771 * SCTP_PEER_ADDR_THLDS
3773 * This option allows us to alter the partially failed threshold for one or all
3774 * transports in an association. See Section 6.1 of:
3775 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3777 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3778 char __user *optval,
3779 unsigned int optlen)
3781 struct sctp_paddrthlds val;
3782 struct sctp_transport *trans;
3783 struct sctp_association *asoc;
3785 if (optlen < sizeof(struct sctp_paddrthlds))
3787 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3788 sizeof(struct sctp_paddrthlds)))
3792 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3793 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3796 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3798 if (val.spt_pathmaxrxt)
3799 trans->pathmaxrxt = val.spt_pathmaxrxt;
3800 trans->pf_retrans = val.spt_pathpfthld;
3803 if (val.spt_pathmaxrxt)
3804 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3805 asoc->pf_retrans = val.spt_pathpfthld;
3807 trans = sctp_addr_id2transport(sk, &val.spt_address,
3812 if (val.spt_pathmaxrxt)
3813 trans->pathmaxrxt = val.spt_pathmaxrxt;
3814 trans->pf_retrans = val.spt_pathpfthld;
3820 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3821 char __user *optval,
3822 unsigned int optlen)
3826 if (optlen < sizeof(int))
3828 if (get_user(val, (int __user *) optval))
3831 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3836 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3837 char __user *optval,
3838 unsigned int optlen)
3842 if (optlen < sizeof(int))
3844 if (get_user(val, (int __user *) optval))
3847 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3852 static int sctp_setsockopt_pr_supported(struct sock *sk,
3853 char __user *optval,
3854 unsigned int optlen)
3856 struct sctp_assoc_value params;
3857 struct sctp_association *asoc;
3858 int retval = -EINVAL;
3860 if (optlen != sizeof(params))
3863 if (copy_from_user(¶ms, optval, optlen)) {
3868 asoc = sctp_id2assoc(sk, params.assoc_id);
3870 asoc->prsctp_enable = !!params.assoc_value;
3871 } else if (!params.assoc_id) {
3872 struct sctp_sock *sp = sctp_sk(sk);
3874 sp->ep->prsctp_enable = !!params.assoc_value;
3885 static int sctp_setsockopt_default_prinfo(struct sock *sk,
3886 char __user *optval,
3887 unsigned int optlen)
3889 struct sctp_default_prinfo info;
3890 struct sctp_association *asoc;
3891 int retval = -EINVAL;
3893 if (optlen != sizeof(info))
3896 if (copy_from_user(&info, optval, sizeof(info))) {
3901 if (info.pr_policy & ~SCTP_PR_SCTP_MASK)
3904 if (info.pr_policy == SCTP_PR_SCTP_NONE)
3907 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
3909 SCTP_PR_SET_POLICY(asoc->default_flags, info.pr_policy);
3910 asoc->default_timetolive = info.pr_value;
3911 } else if (!info.pr_assoc_id) {
3912 struct sctp_sock *sp = sctp_sk(sk);
3914 SCTP_PR_SET_POLICY(sp->default_flags, info.pr_policy);
3915 sp->default_timetolive = info.pr_value;
3926 static int sctp_setsockopt_reconfig_supported(struct sock *sk,
3927 char __user *optval,
3928 unsigned int optlen)
3930 struct sctp_assoc_value params;
3931 struct sctp_association *asoc;
3932 int retval = -EINVAL;
3934 if (optlen != sizeof(params))
3937 if (copy_from_user(¶ms, optval, optlen)) {
3942 asoc = sctp_id2assoc(sk, params.assoc_id);
3944 asoc->reconf_enable = !!params.assoc_value;
3945 } else if (!params.assoc_id) {
3946 struct sctp_sock *sp = sctp_sk(sk);
3948 sp->ep->reconf_enable = !!params.assoc_value;
3959 static int sctp_setsockopt_enable_strreset(struct sock *sk,
3960 char __user *optval,
3961 unsigned int optlen)
3963 struct sctp_assoc_value params;
3964 struct sctp_association *asoc;
3965 int retval = -EINVAL;
3967 if (optlen != sizeof(params))
3970 if (copy_from_user(¶ms, optval, optlen)) {
3975 if (params.assoc_value & (~SCTP_ENABLE_STRRESET_MASK))
3978 asoc = sctp_id2assoc(sk, params.assoc_id);
3980 asoc->strreset_enable = params.assoc_value;
3981 } else if (!params.assoc_id) {
3982 struct sctp_sock *sp = sctp_sk(sk);
3984 sp->ep->strreset_enable = params.assoc_value;
3995 static int sctp_setsockopt_reset_streams(struct sock *sk,
3996 char __user *optval,
3997 unsigned int optlen)
3999 struct sctp_reset_streams *params;
4000 struct sctp_association *asoc;
4001 int retval = -EINVAL;
4003 if (optlen < sizeof(*params))
4005 /* srs_number_streams is u16, so optlen can't be bigger than this. */
4006 optlen = min_t(unsigned int, optlen, USHRT_MAX +
4007 sizeof(__u16) * sizeof(*params));
4009 params = memdup_user(optval, optlen);
4011 return PTR_ERR(params);
4013 if (params->srs_number_streams * sizeof(__u16) >
4014 optlen - sizeof(*params))
4017 asoc = sctp_id2assoc(sk, params->srs_assoc_id);
4021 retval = sctp_send_reset_streams(asoc, params);
4028 static int sctp_setsockopt_reset_assoc(struct sock *sk,
4029 char __user *optval,
4030 unsigned int optlen)
4032 struct sctp_association *asoc;
4033 sctp_assoc_t associd;
4034 int retval = -EINVAL;
4036 if (optlen != sizeof(associd))
4039 if (copy_from_user(&associd, optval, optlen)) {
4044 asoc = sctp_id2assoc(sk, associd);
4048 retval = sctp_send_reset_assoc(asoc);
4054 static int sctp_setsockopt_add_streams(struct sock *sk,
4055 char __user *optval,
4056 unsigned int optlen)
4058 struct sctp_association *asoc;
4059 struct sctp_add_streams params;
4060 int retval = -EINVAL;
4062 if (optlen != sizeof(params))
4065 if (copy_from_user(¶ms, optval, optlen)) {
4070 asoc = sctp_id2assoc(sk, params.sas_assoc_id);
4074 retval = sctp_send_add_streams(asoc, ¶ms);
4080 static int sctp_setsockopt_scheduler(struct sock *sk,
4081 char __user *optval,
4082 unsigned int optlen)
4084 struct sctp_association *asoc;
4085 struct sctp_assoc_value params;
4086 int retval = -EINVAL;
4088 if (optlen < sizeof(params))
4091 optlen = sizeof(params);
4092 if (copy_from_user(¶ms, optval, optlen)) {
4097 if (params.assoc_value > SCTP_SS_MAX)
4100 asoc = sctp_id2assoc(sk, params.assoc_id);
4104 retval = sctp_sched_set_sched(asoc, params.assoc_value);
4110 static int sctp_setsockopt_scheduler_value(struct sock *sk,
4111 char __user *optval,
4112 unsigned int optlen)
4114 struct sctp_association *asoc;
4115 struct sctp_stream_value params;
4116 int retval = -EINVAL;
4118 if (optlen < sizeof(params))
4121 optlen = sizeof(params);
4122 if (copy_from_user(¶ms, optval, optlen)) {
4127 asoc = sctp_id2assoc(sk, params.assoc_id);
4131 retval = sctp_sched_set_value(asoc, params.stream_id,
4132 params.stream_value, GFP_KERNEL);
4138 static int sctp_setsockopt_interleaving_supported(struct sock *sk,
4139 char __user *optval,
4140 unsigned int optlen)
4142 struct sctp_sock *sp = sctp_sk(sk);
4143 struct net *net = sock_net(sk);
4144 struct sctp_assoc_value params;
4145 int retval = -EINVAL;
4147 if (optlen < sizeof(params))
4150 optlen = sizeof(params);
4151 if (copy_from_user(¶ms, optval, optlen)) {
4156 if (params.assoc_id)
4159 if (!net->sctp.intl_enable || !sp->frag_interleave) {
4164 sp->strm_interleave = !!params.assoc_value;
4172 /* API 6.2 setsockopt(), getsockopt()
4174 * Applications use setsockopt() and getsockopt() to set or retrieve
4175 * socket options. Socket options are used to change the default
4176 * behavior of sockets calls. They are described in Section 7.
4180 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
4181 * int __user *optlen);
4182 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
4185 * sd - the socket descript.
4186 * level - set to IPPROTO_SCTP for all SCTP options.
4187 * optname - the option name.
4188 * optval - the buffer to store the value of the option.
4189 * optlen - the size of the buffer.
4191 static int sctp_setsockopt(struct sock *sk, int level, int optname,
4192 char __user *optval, unsigned int optlen)
4196 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
4198 /* I can hardly begin to describe how wrong this is. This is
4199 * so broken as to be worse than useless. The API draft
4200 * REALLY is NOT helpful here... I am not convinced that the
4201 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
4202 * are at all well-founded.
4204 if (level != SOL_SCTP) {
4205 struct sctp_af *af = sctp_sk(sk)->pf->af;
4206 retval = af->setsockopt(sk, level, optname, optval, optlen);
4213 case SCTP_SOCKOPT_BINDX_ADD:
4214 /* 'optlen' is the size of the addresses buffer. */
4215 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
4216 optlen, SCTP_BINDX_ADD_ADDR);
4219 case SCTP_SOCKOPT_BINDX_REM:
4220 /* 'optlen' is the size of the addresses buffer. */
4221 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
4222 optlen, SCTP_BINDX_REM_ADDR);
4225 case SCTP_SOCKOPT_CONNECTX_OLD:
4226 /* 'optlen' is the size of the addresses buffer. */
4227 retval = sctp_setsockopt_connectx_old(sk,
4228 (struct sockaddr __user *)optval,
4232 case SCTP_SOCKOPT_CONNECTX:
4233 /* 'optlen' is the size of the addresses buffer. */
4234 retval = sctp_setsockopt_connectx(sk,
4235 (struct sockaddr __user *)optval,
4239 case SCTP_DISABLE_FRAGMENTS:
4240 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
4244 retval = sctp_setsockopt_events(sk, optval, optlen);
4247 case SCTP_AUTOCLOSE:
4248 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
4251 case SCTP_PEER_ADDR_PARAMS:
4252 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
4255 case SCTP_DELAYED_SACK:
4256 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
4258 case SCTP_PARTIAL_DELIVERY_POINT:
4259 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
4263 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
4265 case SCTP_DEFAULT_SEND_PARAM:
4266 retval = sctp_setsockopt_default_send_param(sk, optval,
4269 case SCTP_DEFAULT_SNDINFO:
4270 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
4272 case SCTP_PRIMARY_ADDR:
4273 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
4275 case SCTP_SET_PEER_PRIMARY_ADDR:
4276 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
4279 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
4282 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
4284 case SCTP_ASSOCINFO:
4285 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
4287 case SCTP_I_WANT_MAPPED_V4_ADDR:
4288 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
4291 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
4293 case SCTP_ADAPTATION_LAYER:
4294 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
4297 retval = sctp_setsockopt_context(sk, optval, optlen);
4299 case SCTP_FRAGMENT_INTERLEAVE:
4300 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
4302 case SCTP_MAX_BURST:
4303 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
4305 case SCTP_AUTH_CHUNK:
4306 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
4308 case SCTP_HMAC_IDENT:
4309 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
4312 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
4314 case SCTP_AUTH_ACTIVE_KEY:
4315 retval = sctp_setsockopt_active_key(sk, optval, optlen);
4317 case SCTP_AUTH_DELETE_KEY:
4318 retval = sctp_setsockopt_del_key(sk, optval, optlen);
4320 case SCTP_AUTH_DEACTIVATE_KEY:
4321 retval = sctp_setsockopt_deactivate_key(sk, optval, optlen);
4323 case SCTP_AUTO_ASCONF:
4324 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
4326 case SCTP_PEER_ADDR_THLDS:
4327 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
4329 case SCTP_RECVRCVINFO:
4330 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
4332 case SCTP_RECVNXTINFO:
4333 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
4335 case SCTP_PR_SUPPORTED:
4336 retval = sctp_setsockopt_pr_supported(sk, optval, optlen);
4338 case SCTP_DEFAULT_PRINFO:
4339 retval = sctp_setsockopt_default_prinfo(sk, optval, optlen);
4341 case SCTP_RECONFIG_SUPPORTED:
4342 retval = sctp_setsockopt_reconfig_supported(sk, optval, optlen);
4344 case SCTP_ENABLE_STREAM_RESET:
4345 retval = sctp_setsockopt_enable_strreset(sk, optval, optlen);
4347 case SCTP_RESET_STREAMS:
4348 retval = sctp_setsockopt_reset_streams(sk, optval, optlen);
4350 case SCTP_RESET_ASSOC:
4351 retval = sctp_setsockopt_reset_assoc(sk, optval, optlen);
4353 case SCTP_ADD_STREAMS:
4354 retval = sctp_setsockopt_add_streams(sk, optval, optlen);
4356 case SCTP_STREAM_SCHEDULER:
4357 retval = sctp_setsockopt_scheduler(sk, optval, optlen);
4359 case SCTP_STREAM_SCHEDULER_VALUE:
4360 retval = sctp_setsockopt_scheduler_value(sk, optval, optlen);
4362 case SCTP_INTERLEAVING_SUPPORTED:
4363 retval = sctp_setsockopt_interleaving_supported(sk, optval,
4367 retval = -ENOPROTOOPT;
4377 /* API 3.1.6 connect() - UDP Style Syntax
4379 * An application may use the connect() call in the UDP model to initiate an
4380 * association without sending data.
4384 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
4386 * sd: the socket descriptor to have a new association added to.
4388 * nam: the address structure (either struct sockaddr_in or struct
4389 * sockaddr_in6 defined in RFC2553 [7]).
4391 * len: the size of the address.
4393 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
4394 int addr_len, int flags)
4396 struct inet_sock *inet = inet_sk(sk);
4402 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
4405 /* We may need to bind the socket. */
4406 if (!inet->inet_num) {
4407 if (sk->sk_prot->get_port(sk, 0)) {
4411 inet->inet_sport = htons(inet->inet_num);
4414 /* Validate addr_len before calling common connect/connectx routine. */
4415 af = sctp_get_af_specific(addr->sa_family);
4416 if (!af || addr_len < af->sockaddr_len) {
4419 /* Pass correct addr len to common routine (so it knows there
4420 * is only one address being passed.
4422 err = __sctp_connect(sk, addr, af->sockaddr_len, flags, NULL);
4429 int sctp_inet_connect(struct socket *sock, struct sockaddr *uaddr,
4430 int addr_len, int flags)
4432 if (addr_len < sizeof(uaddr->sa_family))
4435 if (uaddr->sa_family == AF_UNSPEC)
4438 return sctp_connect(sock->sk, uaddr, addr_len, flags);
4441 /* FIXME: Write comments. */
4442 static int sctp_disconnect(struct sock *sk, int flags)
4444 return -EOPNOTSUPP; /* STUB */
4447 /* 4.1.4 accept() - TCP Style Syntax
4449 * Applications use accept() call to remove an established SCTP
4450 * association from the accept queue of the endpoint. A new socket
4451 * descriptor will be returned from accept() to represent the newly
4452 * formed association.
4454 static struct sock *sctp_accept(struct sock *sk, int flags, int *err, bool kern)
4456 struct sctp_sock *sp;
4457 struct sctp_endpoint *ep;
4458 struct sock *newsk = NULL;
4459 struct sctp_association *asoc;
4468 if (!sctp_style(sk, TCP)) {
4469 error = -EOPNOTSUPP;
4473 if (!sctp_sstate(sk, LISTENING)) {
4478 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
4480 error = sctp_wait_for_accept(sk, timeo);
4484 /* We treat the list of associations on the endpoint as the accept
4485 * queue and pick the first association on the list.
4487 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
4489 newsk = sp->pf->create_accept_sk(sk, asoc, kern);
4495 /* Populate the fields of the newsk from the oldsk and migrate the
4496 * asoc to the newsk.
4498 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
4506 /* The SCTP ioctl handler. */
4507 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
4514 * SEQPACKET-style sockets in LISTENING state are valid, for
4515 * SCTP, so only discard TCP-style sockets in LISTENING state.
4517 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4522 struct sk_buff *skb;
4523 unsigned int amount = 0;
4525 skb = skb_peek(&sk->sk_receive_queue);
4528 * We will only return the amount of this packet since
4529 * that is all that will be read.
4533 rc = put_user(amount, (int __user *)arg);
4545 /* This is the function which gets called during socket creation to
4546 * initialized the SCTP-specific portion of the sock.
4547 * The sock structure should already be zero-filled memory.
4549 static int sctp_init_sock(struct sock *sk)
4551 struct net *net = sock_net(sk);
4552 struct sctp_sock *sp;
4554 pr_debug("%s: sk:%p\n", __func__, sk);
4558 /* Initialize the SCTP per socket area. */
4559 switch (sk->sk_type) {
4560 case SOCK_SEQPACKET:
4561 sp->type = SCTP_SOCKET_UDP;
4564 sp->type = SCTP_SOCKET_TCP;
4567 return -ESOCKTNOSUPPORT;
4570 sk->sk_gso_type = SKB_GSO_SCTP;
4572 /* Initialize default send parameters. These parameters can be
4573 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4575 sp->default_stream = 0;
4576 sp->default_ppid = 0;
4577 sp->default_flags = 0;
4578 sp->default_context = 0;
4579 sp->default_timetolive = 0;
4581 sp->default_rcv_context = 0;
4582 sp->max_burst = net->sctp.max_burst;
4584 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4586 /* Initialize default setup parameters. These parameters
4587 * can be modified with the SCTP_INITMSG socket option or
4588 * overridden by the SCTP_INIT CMSG.
4590 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4591 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4592 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4593 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4595 /* Initialize default RTO related parameters. These parameters can
4596 * be modified for with the SCTP_RTOINFO socket option.
4598 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4599 sp->rtoinfo.srto_max = net->sctp.rto_max;
4600 sp->rtoinfo.srto_min = net->sctp.rto_min;
4602 /* Initialize default association related parameters. These parameters
4603 * can be modified with the SCTP_ASSOCINFO socket option.
4605 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4606 sp->assocparams.sasoc_number_peer_destinations = 0;
4607 sp->assocparams.sasoc_peer_rwnd = 0;
4608 sp->assocparams.sasoc_local_rwnd = 0;
4609 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4611 /* Initialize default event subscriptions. By default, all the
4614 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4616 /* Default Peer Address Parameters. These defaults can
4617 * be modified via SCTP_PEER_ADDR_PARAMS
4619 sp->hbinterval = net->sctp.hb_interval;
4620 sp->pathmaxrxt = net->sctp.max_retrans_path;
4621 sp->pathmtu = 0; /* allow default discovery */
4622 sp->sackdelay = net->sctp.sack_timeout;
4624 sp->param_flags = SPP_HB_ENABLE |
4626 SPP_SACKDELAY_ENABLE;
4628 /* If enabled no SCTP message fragmentation will be performed.
4629 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4631 sp->disable_fragments = 0;
4633 /* Enable Nagle algorithm by default. */
4636 sp->recvrcvinfo = 0;
4637 sp->recvnxtinfo = 0;
4639 /* Enable by default. */
4642 /* Auto-close idle associations after the configured
4643 * number of seconds. A value of 0 disables this
4644 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4645 * for UDP-style sockets only.
4649 /* User specified fragmentation limit. */
4652 sp->adaptation_ind = 0;
4654 sp->pf = sctp_get_pf_specific(sk->sk_family);
4656 /* Control variables for partial data delivery. */
4657 atomic_set(&sp->pd_mode, 0);
4658 skb_queue_head_init(&sp->pd_lobby);
4659 sp->frag_interleave = 0;
4661 /* Create a per socket endpoint structure. Even if we
4662 * change the data structure relationships, this may still
4663 * be useful for storing pre-connect address information.
4665 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4671 sk->sk_destruct = sctp_destruct_sock;
4673 SCTP_DBG_OBJCNT_INC(sock);
4676 sk_sockets_allocated_inc(sk);
4677 sock_prot_inuse_add(net, sk->sk_prot, 1);
4679 /* Nothing can fail after this block, otherwise
4680 * sctp_destroy_sock() will be called without addr_wq_lock held
4682 if (net->sctp.default_auto_asconf) {
4683 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
4684 list_add_tail(&sp->auto_asconf_list,
4685 &net->sctp.auto_asconf_splist);
4686 sp->do_auto_asconf = 1;
4687 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4689 sp->do_auto_asconf = 0;
4697 /* Cleanup any SCTP per socket resources. Must be called with
4698 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4700 static void sctp_destroy_sock(struct sock *sk)
4702 struct sctp_sock *sp;
4704 pr_debug("%s: sk:%p\n", __func__, sk);
4706 /* Release our hold on the endpoint. */
4708 /* This could happen during socket init, thus we bail out
4709 * early, since the rest of the below is not setup either.
4714 if (sp->do_auto_asconf) {
4715 sp->do_auto_asconf = 0;
4716 list_del(&sp->auto_asconf_list);
4718 sctp_endpoint_free(sp->ep);
4720 sk_sockets_allocated_dec(sk);
4721 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4725 /* Triggered when there are no references on the socket anymore */
4726 static void sctp_destruct_sock(struct sock *sk)
4728 struct sctp_sock *sp = sctp_sk(sk);
4730 /* Free up the HMAC transform. */
4731 crypto_free_shash(sp->hmac);
4733 inet_sock_destruct(sk);
4736 /* API 4.1.7 shutdown() - TCP Style Syntax
4737 * int shutdown(int socket, int how);
4739 * sd - the socket descriptor of the association to be closed.
4740 * how - Specifies the type of shutdown. The values are
4743 * Disables further receive operations. No SCTP
4744 * protocol action is taken.
4746 * Disables further send operations, and initiates
4747 * the SCTP shutdown sequence.
4749 * Disables further send and receive operations
4750 * and initiates the SCTP shutdown sequence.
4752 static void sctp_shutdown(struct sock *sk, int how)
4754 struct net *net = sock_net(sk);
4755 struct sctp_endpoint *ep;
4757 if (!sctp_style(sk, TCP))
4760 ep = sctp_sk(sk)->ep;
4761 if (how & SEND_SHUTDOWN && !list_empty(&ep->asocs)) {
4762 struct sctp_association *asoc;
4764 inet_sk_set_state(sk, SCTP_SS_CLOSING);
4765 asoc = list_entry(ep->asocs.next,
4766 struct sctp_association, asocs);
4767 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4771 int sctp_get_sctp_info(struct sock *sk, struct sctp_association *asoc,
4772 struct sctp_info *info)
4774 struct sctp_transport *prim;
4775 struct list_head *pos;
4778 memset(info, 0, sizeof(*info));
4780 struct sctp_sock *sp = sctp_sk(sk);
4782 info->sctpi_s_autoclose = sp->autoclose;
4783 info->sctpi_s_adaptation_ind = sp->adaptation_ind;
4784 info->sctpi_s_pd_point = sp->pd_point;
4785 info->sctpi_s_nodelay = sp->nodelay;
4786 info->sctpi_s_disable_fragments = sp->disable_fragments;
4787 info->sctpi_s_v4mapped = sp->v4mapped;
4788 info->sctpi_s_frag_interleave = sp->frag_interleave;
4789 info->sctpi_s_type = sp->type;
4794 info->sctpi_tag = asoc->c.my_vtag;
4795 info->sctpi_state = asoc->state;
4796 info->sctpi_rwnd = asoc->a_rwnd;
4797 info->sctpi_unackdata = asoc->unack_data;
4798 info->sctpi_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4799 info->sctpi_instrms = asoc->stream.incnt;
4800 info->sctpi_outstrms = asoc->stream.outcnt;
4801 list_for_each(pos, &asoc->base.inqueue.in_chunk_list)
4802 info->sctpi_inqueue++;
4803 list_for_each(pos, &asoc->outqueue.out_chunk_list)
4804 info->sctpi_outqueue++;
4805 info->sctpi_overall_error = asoc->overall_error_count;
4806 info->sctpi_max_burst = asoc->max_burst;
4807 info->sctpi_maxseg = asoc->frag_point;
4808 info->sctpi_peer_rwnd = asoc->peer.rwnd;
4809 info->sctpi_peer_tag = asoc->c.peer_vtag;
4811 mask = asoc->peer.ecn_capable << 1;
4812 mask = (mask | asoc->peer.ipv4_address) << 1;
4813 mask = (mask | asoc->peer.ipv6_address) << 1;
4814 mask = (mask | asoc->peer.hostname_address) << 1;
4815 mask = (mask | asoc->peer.asconf_capable) << 1;
4816 mask = (mask | asoc->peer.prsctp_capable) << 1;
4817 mask = (mask | asoc->peer.auth_capable);
4818 info->sctpi_peer_capable = mask;
4819 mask = asoc->peer.sack_needed << 1;
4820 mask = (mask | asoc->peer.sack_generation) << 1;
4821 mask = (mask | asoc->peer.zero_window_announced);
4822 info->sctpi_peer_sack = mask;
4824 info->sctpi_isacks = asoc->stats.isacks;
4825 info->sctpi_osacks = asoc->stats.osacks;
4826 info->sctpi_opackets = asoc->stats.opackets;
4827 info->sctpi_ipackets = asoc->stats.ipackets;
4828 info->sctpi_rtxchunks = asoc->stats.rtxchunks;
4829 info->sctpi_outofseqtsns = asoc->stats.outofseqtsns;
4830 info->sctpi_idupchunks = asoc->stats.idupchunks;
4831 info->sctpi_gapcnt = asoc->stats.gapcnt;
4832 info->sctpi_ouodchunks = asoc->stats.ouodchunks;
4833 info->sctpi_iuodchunks = asoc->stats.iuodchunks;
4834 info->sctpi_oodchunks = asoc->stats.oodchunks;
4835 info->sctpi_iodchunks = asoc->stats.iodchunks;
4836 info->sctpi_octrlchunks = asoc->stats.octrlchunks;
4837 info->sctpi_ictrlchunks = asoc->stats.ictrlchunks;
4839 prim = asoc->peer.primary_path;
4840 memcpy(&info->sctpi_p_address, &prim->ipaddr, sizeof(prim->ipaddr));
4841 info->sctpi_p_state = prim->state;
4842 info->sctpi_p_cwnd = prim->cwnd;
4843 info->sctpi_p_srtt = prim->srtt;
4844 info->sctpi_p_rto = jiffies_to_msecs(prim->rto);
4845 info->sctpi_p_hbinterval = prim->hbinterval;
4846 info->sctpi_p_pathmaxrxt = prim->pathmaxrxt;
4847 info->sctpi_p_sackdelay = jiffies_to_msecs(prim->sackdelay);
4848 info->sctpi_p_ssthresh = prim->ssthresh;
4849 info->sctpi_p_partial_bytes_acked = prim->partial_bytes_acked;
4850 info->sctpi_p_flight_size = prim->flight_size;
4851 info->sctpi_p_error = prim->error_count;
4855 EXPORT_SYMBOL_GPL(sctp_get_sctp_info);
4857 /* use callback to avoid exporting the core structure */
4858 void sctp_transport_walk_start(struct rhashtable_iter *iter)
4860 rhltable_walk_enter(&sctp_transport_hashtable, iter);
4862 rhashtable_walk_start(iter);
4865 void sctp_transport_walk_stop(struct rhashtable_iter *iter)
4867 rhashtable_walk_stop(iter);
4868 rhashtable_walk_exit(iter);
4871 struct sctp_transport *sctp_transport_get_next(struct net *net,
4872 struct rhashtable_iter *iter)
4874 struct sctp_transport *t;
4876 t = rhashtable_walk_next(iter);
4877 for (; t; t = rhashtable_walk_next(iter)) {
4879 if (PTR_ERR(t) == -EAGAIN)
4884 if (net_eq(sock_net(t->asoc->base.sk), net) &&
4885 t->asoc->peer.primary_path == t)
4892 struct sctp_transport *sctp_transport_get_idx(struct net *net,
4893 struct rhashtable_iter *iter,
4896 void *obj = SEQ_START_TOKEN;
4898 while (pos && (obj = sctp_transport_get_next(net, iter)) &&
4905 int sctp_for_each_endpoint(int (*cb)(struct sctp_endpoint *, void *),
4909 struct sctp_ep_common *epb;
4910 struct sctp_hashbucket *head;
4912 for (head = sctp_ep_hashtable; hash < sctp_ep_hashsize;
4914 read_lock_bh(&head->lock);
4915 sctp_for_each_hentry(epb, &head->chain) {
4916 err = cb(sctp_ep(epb), p);
4920 read_unlock_bh(&head->lock);
4925 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint);
4927 int sctp_transport_lookup_process(int (*cb)(struct sctp_transport *, void *),
4929 const union sctp_addr *laddr,
4930 const union sctp_addr *paddr, void *p)
4932 struct sctp_transport *transport;
4936 transport = sctp_addrs_lookup_transport(net, laddr, paddr);
4941 err = cb(transport, p);
4942 sctp_transport_put(transport);
4946 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process);
4948 int sctp_for_each_transport(int (*cb)(struct sctp_transport *, void *),
4949 int (*cb_done)(struct sctp_transport *, void *),
4950 struct net *net, int *pos, void *p) {
4951 struct rhashtable_iter hti;
4952 struct sctp_transport *tsp;
4957 sctp_transport_walk_start(&hti);
4959 tsp = sctp_transport_get_idx(net, &hti, *pos + 1);
4960 for (; !IS_ERR_OR_NULL(tsp); tsp = sctp_transport_get_next(net, &hti)) {
4961 if (!sctp_transport_hold(tsp))
4967 sctp_transport_put(tsp);
4969 sctp_transport_walk_stop(&hti);
4972 if (cb_done && !cb_done(tsp, p)) {
4974 sctp_transport_put(tsp);
4977 sctp_transport_put(tsp);
4982 EXPORT_SYMBOL_GPL(sctp_for_each_transport);
4984 /* 7.2.1 Association Status (SCTP_STATUS)
4986 * Applications can retrieve current status information about an
4987 * association, including association state, peer receiver window size,
4988 * number of unacked data chunks, and number of data chunks pending
4989 * receipt. This information is read-only.
4991 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4992 char __user *optval,
4995 struct sctp_status status;
4996 struct sctp_association *asoc = NULL;
4997 struct sctp_transport *transport;
4998 sctp_assoc_t associd;
5001 if (len < sizeof(status)) {
5006 len = sizeof(status);
5007 if (copy_from_user(&status, optval, len)) {
5012 associd = status.sstat_assoc_id;
5013 asoc = sctp_id2assoc(sk, associd);
5019 transport = asoc->peer.primary_path;
5021 status.sstat_assoc_id = sctp_assoc2id(asoc);
5022 status.sstat_state = sctp_assoc_to_state(asoc);
5023 status.sstat_rwnd = asoc->peer.rwnd;
5024 status.sstat_unackdata = asoc->unack_data;
5026 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
5027 status.sstat_instrms = asoc->stream.incnt;
5028 status.sstat_outstrms = asoc->stream.outcnt;
5029 status.sstat_fragmentation_point = asoc->frag_point;
5030 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
5031 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
5032 transport->af_specific->sockaddr_len);
5033 /* Map ipv4 address into v4-mapped-on-v6 address. */
5034 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
5035 (union sctp_addr *)&status.sstat_primary.spinfo_address);
5036 status.sstat_primary.spinfo_state = transport->state;
5037 status.sstat_primary.spinfo_cwnd = transport->cwnd;
5038 status.sstat_primary.spinfo_srtt = transport->srtt;
5039 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
5040 status.sstat_primary.spinfo_mtu = transport->pathmtu;
5042 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
5043 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
5045 if (put_user(len, optlen)) {
5050 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
5051 __func__, len, status.sstat_state, status.sstat_rwnd,
5052 status.sstat_assoc_id);
5054 if (copy_to_user(optval, &status, len)) {
5064 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
5066 * Applications can retrieve information about a specific peer address
5067 * of an association, including its reachability state, congestion
5068 * window, and retransmission timer values. This information is
5071 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
5072 char __user *optval,
5075 struct sctp_paddrinfo pinfo;
5076 struct sctp_transport *transport;
5079 if (len < sizeof(pinfo)) {
5084 len = sizeof(pinfo);
5085 if (copy_from_user(&pinfo, optval, len)) {
5090 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
5091 pinfo.spinfo_assoc_id);
5095 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
5096 pinfo.spinfo_state = transport->state;
5097 pinfo.spinfo_cwnd = transport->cwnd;
5098 pinfo.spinfo_srtt = transport->srtt;
5099 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
5100 pinfo.spinfo_mtu = transport->pathmtu;
5102 if (pinfo.spinfo_state == SCTP_UNKNOWN)
5103 pinfo.spinfo_state = SCTP_ACTIVE;
5105 if (put_user(len, optlen)) {
5110 if (copy_to_user(optval, &pinfo, len)) {
5119 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
5121 * This option is a on/off flag. If enabled no SCTP message
5122 * fragmentation will be performed. Instead if a message being sent
5123 * exceeds the current PMTU size, the message will NOT be sent and
5124 * instead a error will be indicated to the user.
5126 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
5127 char __user *optval, int __user *optlen)
5131 if (len < sizeof(int))
5135 val = (sctp_sk(sk)->disable_fragments == 1);
5136 if (put_user(len, optlen))
5138 if (copy_to_user(optval, &val, len))
5143 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
5145 * This socket option is used to specify various notifications and
5146 * ancillary data the user wishes to receive.
5148 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
5153 if (len > sizeof(struct sctp_event_subscribe))
5154 len = sizeof(struct sctp_event_subscribe);
5155 if (put_user(len, optlen))
5157 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
5162 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
5164 * This socket option is applicable to the UDP-style socket only. When
5165 * set it will cause associations that are idle for more than the
5166 * specified number of seconds to automatically close. An association
5167 * being idle is defined an association that has NOT sent or received
5168 * user data. The special value of '0' indicates that no automatic
5169 * close of any associations should be performed. The option expects an
5170 * integer defining the number of seconds of idle time before an
5171 * association is closed.
5173 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
5175 /* Applicable to UDP-style socket only */
5176 if (sctp_style(sk, TCP))
5178 if (len < sizeof(int))
5181 if (put_user(len, optlen))
5183 if (put_user(sctp_sk(sk)->autoclose, (int __user *)optval))
5188 /* Helper routine to branch off an association to a new socket. */
5189 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
5191 struct sctp_association *asoc = sctp_id2assoc(sk, id);
5192 struct sctp_sock *sp = sctp_sk(sk);
5193 struct socket *sock;
5196 /* Do not peel off from one netns to another one. */
5197 if (!net_eq(current->nsproxy->net_ns, sock_net(sk)))
5203 /* An association cannot be branched off from an already peeled-off
5204 * socket, nor is this supported for tcp style sockets.
5206 if (!sctp_style(sk, UDP))
5209 /* Create a new socket. */
5210 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
5214 sctp_copy_sock(sock->sk, sk, asoc);
5216 /* Make peeled-off sockets more like 1-1 accepted sockets.
5217 * Set the daddr and initialize id to something more random and also
5218 * copy over any ip options.
5220 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
5221 sp->pf->copy_ip_options(sk, sock->sk);
5223 /* Populate the fields of the newsk from the oldsk and migrate the
5224 * asoc to the newsk.
5226 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
5232 EXPORT_SYMBOL(sctp_do_peeloff);
5234 static int sctp_getsockopt_peeloff_common(struct sock *sk, sctp_peeloff_arg_t *peeloff,
5235 struct file **newfile, unsigned flags)
5237 struct socket *newsock;
5240 retval = sctp_do_peeloff(sk, peeloff->associd, &newsock);
5244 /* Map the socket to an unused fd that can be returned to the user. */
5245 retval = get_unused_fd_flags(flags & SOCK_CLOEXEC);
5247 sock_release(newsock);
5251 *newfile = sock_alloc_file(newsock, 0, NULL);
5252 if (IS_ERR(*newfile)) {
5253 put_unused_fd(retval);
5254 retval = PTR_ERR(*newfile);
5259 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
5262 peeloff->sd = retval;
5264 if (flags & SOCK_NONBLOCK)
5265 (*newfile)->f_flags |= O_NONBLOCK;
5270 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
5272 sctp_peeloff_arg_t peeloff;
5273 struct file *newfile = NULL;
5276 if (len < sizeof(sctp_peeloff_arg_t))
5278 len = sizeof(sctp_peeloff_arg_t);
5279 if (copy_from_user(&peeloff, optval, len))
5282 retval = sctp_getsockopt_peeloff_common(sk, &peeloff, &newfile, 0);
5286 /* Return the fd mapped to the new socket. */
5287 if (put_user(len, optlen)) {
5289 put_unused_fd(retval);
5293 if (copy_to_user(optval, &peeloff, len)) {
5295 put_unused_fd(retval);
5298 fd_install(retval, newfile);
5303 static int sctp_getsockopt_peeloff_flags(struct sock *sk, int len,
5304 char __user *optval, int __user *optlen)
5306 sctp_peeloff_flags_arg_t peeloff;
5307 struct file *newfile = NULL;
5310 if (len < sizeof(sctp_peeloff_flags_arg_t))
5312 len = sizeof(sctp_peeloff_flags_arg_t);
5313 if (copy_from_user(&peeloff, optval, len))
5316 retval = sctp_getsockopt_peeloff_common(sk, &peeloff.p_arg,
5317 &newfile, peeloff.flags);
5321 /* Return the fd mapped to the new socket. */
5322 if (put_user(len, optlen)) {
5324 put_unused_fd(retval);
5328 if (copy_to_user(optval, &peeloff, len)) {
5330 put_unused_fd(retval);
5333 fd_install(retval, newfile);
5338 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
5340 * Applications can enable or disable heartbeats for any peer address of
5341 * an association, modify an address's heartbeat interval, force a
5342 * heartbeat to be sent immediately, and adjust the address's maximum
5343 * number of retransmissions sent before an address is considered
5344 * unreachable. The following structure is used to access and modify an
5345 * address's parameters:
5347 * struct sctp_paddrparams {
5348 * sctp_assoc_t spp_assoc_id;
5349 * struct sockaddr_storage spp_address;
5350 * uint32_t spp_hbinterval;
5351 * uint16_t spp_pathmaxrxt;
5352 * uint32_t spp_pathmtu;
5353 * uint32_t spp_sackdelay;
5354 * uint32_t spp_flags;
5357 * spp_assoc_id - (one-to-many style socket) This is filled in the
5358 * application, and identifies the association for
5360 * spp_address - This specifies which address is of interest.
5361 * spp_hbinterval - This contains the value of the heartbeat interval,
5362 * in milliseconds. If a value of zero
5363 * is present in this field then no changes are to
5364 * be made to this parameter.
5365 * spp_pathmaxrxt - This contains the maximum number of
5366 * retransmissions before this address shall be
5367 * considered unreachable. If a value of zero
5368 * is present in this field then no changes are to
5369 * be made to this parameter.
5370 * spp_pathmtu - When Path MTU discovery is disabled the value
5371 * specified here will be the "fixed" path mtu.
5372 * Note that if the spp_address field is empty
5373 * then all associations on this address will
5374 * have this fixed path mtu set upon them.
5376 * spp_sackdelay - When delayed sack is enabled, this value specifies
5377 * the number of milliseconds that sacks will be delayed
5378 * for. This value will apply to all addresses of an
5379 * association if the spp_address field is empty. Note
5380 * also, that if delayed sack is enabled and this
5381 * value is set to 0, no change is made to the last
5382 * recorded delayed sack timer value.
5384 * spp_flags - These flags are used to control various features
5385 * on an association. The flag field may contain
5386 * zero or more of the following options.
5388 * SPP_HB_ENABLE - Enable heartbeats on the
5389 * specified address. Note that if the address
5390 * field is empty all addresses for the association
5391 * have heartbeats enabled upon them.
5393 * SPP_HB_DISABLE - Disable heartbeats on the
5394 * speicifed address. Note that if the address
5395 * field is empty all addresses for the association
5396 * will have their heartbeats disabled. Note also
5397 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
5398 * mutually exclusive, only one of these two should
5399 * be specified. Enabling both fields will have
5400 * undetermined results.
5402 * SPP_HB_DEMAND - Request a user initiated heartbeat
5403 * to be made immediately.
5405 * SPP_PMTUD_ENABLE - This field will enable PMTU
5406 * discovery upon the specified address. Note that
5407 * if the address feild is empty then all addresses
5408 * on the association are effected.
5410 * SPP_PMTUD_DISABLE - This field will disable PMTU
5411 * discovery upon the specified address. Note that
5412 * if the address feild is empty then all addresses
5413 * on the association are effected. Not also that
5414 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
5415 * exclusive. Enabling both will have undetermined
5418 * SPP_SACKDELAY_ENABLE - Setting this flag turns
5419 * on delayed sack. The time specified in spp_sackdelay
5420 * is used to specify the sack delay for this address. Note
5421 * that if spp_address is empty then all addresses will
5422 * enable delayed sack and take on the sack delay
5423 * value specified in spp_sackdelay.
5424 * SPP_SACKDELAY_DISABLE - Setting this flag turns
5425 * off delayed sack. If the spp_address field is blank then
5426 * delayed sack is disabled for the entire association. Note
5427 * also that this field is mutually exclusive to
5428 * SPP_SACKDELAY_ENABLE, setting both will have undefined
5431 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
5432 char __user *optval, int __user *optlen)
5434 struct sctp_paddrparams params;
5435 struct sctp_transport *trans = NULL;
5436 struct sctp_association *asoc = NULL;
5437 struct sctp_sock *sp = sctp_sk(sk);
5439 if (len < sizeof(struct sctp_paddrparams))
5441 len = sizeof(struct sctp_paddrparams);
5442 if (copy_from_user(¶ms, optval, len))
5445 /* If an address other than INADDR_ANY is specified, and
5446 * no transport is found, then the request is invalid.
5448 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
5449 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
5450 params.spp_assoc_id);
5452 pr_debug("%s: failed no transport\n", __func__);
5457 /* Get association, if assoc_id != 0 and the socket is a one
5458 * to many style socket, and an association was not found, then
5459 * the id was invalid.
5461 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
5462 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
5463 pr_debug("%s: failed no association\n", __func__);
5468 /* Fetch transport values. */
5469 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
5470 params.spp_pathmtu = trans->pathmtu;
5471 params.spp_pathmaxrxt = trans->pathmaxrxt;
5472 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
5474 /*draft-11 doesn't say what to return in spp_flags*/
5475 params.spp_flags = trans->param_flags;
5477 /* Fetch association values. */
5478 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
5479 params.spp_pathmtu = asoc->pathmtu;
5480 params.spp_pathmaxrxt = asoc->pathmaxrxt;
5481 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
5483 /*draft-11 doesn't say what to return in spp_flags*/
5484 params.spp_flags = asoc->param_flags;
5486 /* Fetch socket values. */
5487 params.spp_hbinterval = sp->hbinterval;
5488 params.spp_pathmtu = sp->pathmtu;
5489 params.spp_sackdelay = sp->sackdelay;
5490 params.spp_pathmaxrxt = sp->pathmaxrxt;
5492 /*draft-11 doesn't say what to return in spp_flags*/
5493 params.spp_flags = sp->param_flags;
5496 if (copy_to_user(optval, ¶ms, len))
5499 if (put_user(len, optlen))
5506 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
5508 * This option will effect the way delayed acks are performed. This
5509 * option allows you to get or set the delayed ack time, in
5510 * milliseconds. It also allows changing the delayed ack frequency.
5511 * Changing the frequency to 1 disables the delayed sack algorithm. If
5512 * the assoc_id is 0, then this sets or gets the endpoints default
5513 * values. If the assoc_id field is non-zero, then the set or get
5514 * effects the specified association for the one to many model (the
5515 * assoc_id field is ignored by the one to one model). Note that if
5516 * sack_delay or sack_freq are 0 when setting this option, then the
5517 * current values will remain unchanged.
5519 * struct sctp_sack_info {
5520 * sctp_assoc_t sack_assoc_id;
5521 * uint32_t sack_delay;
5522 * uint32_t sack_freq;
5525 * sack_assoc_id - This parameter, indicates which association the user
5526 * is performing an action upon. Note that if this field's value is
5527 * zero then the endpoints default value is changed (effecting future
5528 * associations only).
5530 * sack_delay - This parameter contains the number of milliseconds that
5531 * the user is requesting the delayed ACK timer be set to. Note that
5532 * this value is defined in the standard to be between 200 and 500
5535 * sack_freq - This parameter contains the number of packets that must
5536 * be received before a sack is sent without waiting for the delay
5537 * timer to expire. The default value for this is 2, setting this
5538 * value to 1 will disable the delayed sack algorithm.
5540 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
5541 char __user *optval,
5544 struct sctp_sack_info params;
5545 struct sctp_association *asoc = NULL;
5546 struct sctp_sock *sp = sctp_sk(sk);
5548 if (len >= sizeof(struct sctp_sack_info)) {
5549 len = sizeof(struct sctp_sack_info);
5551 if (copy_from_user(¶ms, optval, len))
5553 } else if (len == sizeof(struct sctp_assoc_value)) {
5554 pr_warn_ratelimited(DEPRECATED
5556 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
5557 "Use struct sctp_sack_info instead\n",
5558 current->comm, task_pid_nr(current));
5559 if (copy_from_user(¶ms, optval, len))
5564 /* Get association, if sack_assoc_id != 0 and the socket is a one
5565 * to many style socket, and an association was not found, then
5566 * the id was invalid.
5568 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
5569 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
5573 /* Fetch association values. */
5574 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
5575 params.sack_delay = jiffies_to_msecs(
5577 params.sack_freq = asoc->sackfreq;
5580 params.sack_delay = 0;
5581 params.sack_freq = 1;
5584 /* Fetch socket values. */
5585 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
5586 params.sack_delay = sp->sackdelay;
5587 params.sack_freq = sp->sackfreq;
5589 params.sack_delay = 0;
5590 params.sack_freq = 1;
5594 if (copy_to_user(optval, ¶ms, len))
5597 if (put_user(len, optlen))
5603 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5605 * Applications can specify protocol parameters for the default association
5606 * initialization. The option name argument to setsockopt() and getsockopt()
5609 * Setting initialization parameters is effective only on an unconnected
5610 * socket (for UDP-style sockets only future associations are effected
5611 * by the change). With TCP-style sockets, this option is inherited by
5612 * sockets derived from a listener socket.
5614 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
5616 if (len < sizeof(struct sctp_initmsg))
5618 len = sizeof(struct sctp_initmsg);
5619 if (put_user(len, optlen))
5621 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
5627 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
5628 char __user *optval, int __user *optlen)
5630 struct sctp_association *asoc;
5632 struct sctp_getaddrs getaddrs;
5633 struct sctp_transport *from;
5635 union sctp_addr temp;
5636 struct sctp_sock *sp = sctp_sk(sk);
5641 if (len < sizeof(struct sctp_getaddrs))
5644 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5647 /* For UDP-style sockets, id specifies the association to query. */
5648 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5652 to = optval + offsetof(struct sctp_getaddrs, addrs);
5653 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5655 list_for_each_entry(from, &asoc->peer.transport_addr_list,
5657 memcpy(&temp, &from->ipaddr, sizeof(temp));
5658 addrlen = sctp_get_pf_specific(sk->sk_family)
5659 ->addr_to_user(sp, &temp);
5660 if (space_left < addrlen)
5662 if (copy_to_user(to, &temp, addrlen))
5666 space_left -= addrlen;
5669 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
5671 bytes_copied = ((char __user *)to) - optval;
5672 if (put_user(bytes_copied, optlen))
5678 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
5679 size_t space_left, int *bytes_copied)
5681 struct sctp_sockaddr_entry *addr;
5682 union sctp_addr temp;
5685 struct net *net = sock_net(sk);
5688 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
5692 if ((PF_INET == sk->sk_family) &&
5693 (AF_INET6 == addr->a.sa.sa_family))
5695 if ((PF_INET6 == sk->sk_family) &&
5696 inet_v6_ipv6only(sk) &&
5697 (AF_INET == addr->a.sa.sa_family))
5699 memcpy(&temp, &addr->a, sizeof(temp));
5700 if (!temp.v4.sin_port)
5701 temp.v4.sin_port = htons(port);
5703 addrlen = sctp_get_pf_specific(sk->sk_family)
5704 ->addr_to_user(sctp_sk(sk), &temp);
5706 if (space_left < addrlen) {
5710 memcpy(to, &temp, addrlen);
5714 space_left -= addrlen;
5715 *bytes_copied += addrlen;
5723 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
5724 char __user *optval, int __user *optlen)
5726 struct sctp_bind_addr *bp;
5727 struct sctp_association *asoc;
5729 struct sctp_getaddrs getaddrs;
5730 struct sctp_sockaddr_entry *addr;
5732 union sctp_addr temp;
5733 struct sctp_sock *sp = sctp_sk(sk);
5737 int bytes_copied = 0;
5741 if (len < sizeof(struct sctp_getaddrs))
5744 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5748 * For UDP-style sockets, id specifies the association to query.
5749 * If the id field is set to the value '0' then the locally bound
5750 * addresses are returned without regard to any particular
5753 if (0 == getaddrs.assoc_id) {
5754 bp = &sctp_sk(sk)->ep->base.bind_addr;
5756 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5759 bp = &asoc->base.bind_addr;
5762 to = optval + offsetof(struct sctp_getaddrs, addrs);
5763 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5765 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN);
5769 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5770 * addresses from the global local address list.
5772 if (sctp_list_single_entry(&bp->address_list)) {
5773 addr = list_entry(bp->address_list.next,
5774 struct sctp_sockaddr_entry, list);
5775 if (sctp_is_any(sk, &addr->a)) {
5776 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
5777 space_left, &bytes_copied);
5787 /* Protection on the bound address list is not needed since
5788 * in the socket option context we hold a socket lock and
5789 * thus the bound address list can't change.
5791 list_for_each_entry(addr, &bp->address_list, list) {
5792 memcpy(&temp, &addr->a, sizeof(temp));
5793 addrlen = sctp_get_pf_specific(sk->sk_family)
5794 ->addr_to_user(sp, &temp);
5795 if (space_left < addrlen) {
5796 err = -ENOMEM; /*fixme: right error?*/
5799 memcpy(buf, &temp, addrlen);
5801 bytes_copied += addrlen;
5803 space_left -= addrlen;
5807 if (copy_to_user(to, addrs, bytes_copied)) {
5811 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
5815 /* XXX: We should have accounted for sizeof(struct sctp_getaddrs) too,
5816 * but we can't change it anymore.
5818 if (put_user(bytes_copied, optlen))
5825 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
5827 * Requests that the local SCTP stack use the enclosed peer address as
5828 * the association primary. The enclosed address must be one of the
5829 * association peer's addresses.
5831 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
5832 char __user *optval, int __user *optlen)
5834 struct sctp_prim prim;
5835 struct sctp_association *asoc;
5836 struct sctp_sock *sp = sctp_sk(sk);
5838 if (len < sizeof(struct sctp_prim))
5841 len = sizeof(struct sctp_prim);
5843 if (copy_from_user(&prim, optval, len))
5846 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
5850 if (!asoc->peer.primary_path)
5853 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
5854 asoc->peer.primary_path->af_specific->sockaddr_len);
5856 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
5857 (union sctp_addr *)&prim.ssp_addr);
5859 if (put_user(len, optlen))
5861 if (copy_to_user(optval, &prim, len))
5868 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5870 * Requests that the local endpoint set the specified Adaptation Layer
5871 * Indication parameter for all future INIT and INIT-ACK exchanges.
5873 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
5874 char __user *optval, int __user *optlen)
5876 struct sctp_setadaptation adaptation;
5878 if (len < sizeof(struct sctp_setadaptation))
5881 len = sizeof(struct sctp_setadaptation);
5883 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
5885 if (put_user(len, optlen))
5887 if (copy_to_user(optval, &adaptation, len))
5895 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5897 * Applications that wish to use the sendto() system call may wish to
5898 * specify a default set of parameters that would normally be supplied
5899 * through the inclusion of ancillary data. This socket option allows
5900 * such an application to set the default sctp_sndrcvinfo structure.
5903 * The application that wishes to use this socket option simply passes
5904 * in to this call the sctp_sndrcvinfo structure defined in Section
5905 * 5.2.2) The input parameters accepted by this call include
5906 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5907 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5908 * to this call if the caller is using the UDP model.
5910 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5912 static int sctp_getsockopt_default_send_param(struct sock *sk,
5913 int len, char __user *optval,
5916 struct sctp_sock *sp = sctp_sk(sk);
5917 struct sctp_association *asoc;
5918 struct sctp_sndrcvinfo info;
5920 if (len < sizeof(info))
5925 if (copy_from_user(&info, optval, len))
5928 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
5929 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
5932 info.sinfo_stream = asoc->default_stream;
5933 info.sinfo_flags = asoc->default_flags;
5934 info.sinfo_ppid = asoc->default_ppid;
5935 info.sinfo_context = asoc->default_context;
5936 info.sinfo_timetolive = asoc->default_timetolive;
5938 info.sinfo_stream = sp->default_stream;
5939 info.sinfo_flags = sp->default_flags;
5940 info.sinfo_ppid = sp->default_ppid;
5941 info.sinfo_context = sp->default_context;
5942 info.sinfo_timetolive = sp->default_timetolive;
5945 if (put_user(len, optlen))
5947 if (copy_to_user(optval, &info, len))
5953 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5954 * (SCTP_DEFAULT_SNDINFO)
5956 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
5957 char __user *optval,
5960 struct sctp_sock *sp = sctp_sk(sk);
5961 struct sctp_association *asoc;
5962 struct sctp_sndinfo info;
5964 if (len < sizeof(info))
5969 if (copy_from_user(&info, optval, len))
5972 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
5973 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
5976 info.snd_sid = asoc->default_stream;
5977 info.snd_flags = asoc->default_flags;
5978 info.snd_ppid = asoc->default_ppid;
5979 info.snd_context = asoc->default_context;
5981 info.snd_sid = sp->default_stream;
5982 info.snd_flags = sp->default_flags;
5983 info.snd_ppid = sp->default_ppid;
5984 info.snd_context = sp->default_context;
5987 if (put_user(len, optlen))
5989 if (copy_to_user(optval, &info, len))
5997 * 7.1.5 SCTP_NODELAY
5999 * Turn on/off any Nagle-like algorithm. This means that packets are
6000 * generally sent as soon as possible and no unnecessary delays are
6001 * introduced, at the cost of more packets in the network. Expects an
6002 * integer boolean flag.
6005 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
6006 char __user *optval, int __user *optlen)
6010 if (len < sizeof(int))
6014 val = (sctp_sk(sk)->nodelay == 1);
6015 if (put_user(len, optlen))
6017 if (copy_to_user(optval, &val, len))
6024 * 7.1.1 SCTP_RTOINFO
6026 * The protocol parameters used to initialize and bound retransmission
6027 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
6028 * and modify these parameters.
6029 * All parameters are time values, in milliseconds. A value of 0, when
6030 * modifying the parameters, indicates that the current value should not
6034 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
6035 char __user *optval,
6036 int __user *optlen) {
6037 struct sctp_rtoinfo rtoinfo;
6038 struct sctp_association *asoc;
6040 if (len < sizeof (struct sctp_rtoinfo))
6043 len = sizeof(struct sctp_rtoinfo);
6045 if (copy_from_user(&rtoinfo, optval, len))
6048 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
6050 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
6053 /* Values corresponding to the specific association. */
6055 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
6056 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
6057 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
6059 /* Values corresponding to the endpoint. */
6060 struct sctp_sock *sp = sctp_sk(sk);
6062 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
6063 rtoinfo.srto_max = sp->rtoinfo.srto_max;
6064 rtoinfo.srto_min = sp->rtoinfo.srto_min;
6067 if (put_user(len, optlen))
6070 if (copy_to_user(optval, &rtoinfo, len))
6078 * 7.1.2 SCTP_ASSOCINFO
6080 * This option is used to tune the maximum retransmission attempts
6081 * of the association.
6082 * Returns an error if the new association retransmission value is
6083 * greater than the sum of the retransmission value of the peer.
6084 * See [SCTP] for more information.
6087 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
6088 char __user *optval,
6092 struct sctp_assocparams assocparams;
6093 struct sctp_association *asoc;
6094 struct list_head *pos;
6097 if (len < sizeof (struct sctp_assocparams))
6100 len = sizeof(struct sctp_assocparams);
6102 if (copy_from_user(&assocparams, optval, len))
6105 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
6107 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
6110 /* Values correspoinding to the specific association */
6112 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
6113 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
6114 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
6115 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
6117 list_for_each(pos, &asoc->peer.transport_addr_list) {
6121 assocparams.sasoc_number_peer_destinations = cnt;
6123 /* Values corresponding to the endpoint */
6124 struct sctp_sock *sp = sctp_sk(sk);
6126 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
6127 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
6128 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
6129 assocparams.sasoc_cookie_life =
6130 sp->assocparams.sasoc_cookie_life;
6131 assocparams.sasoc_number_peer_destinations =
6133 sasoc_number_peer_destinations;
6136 if (put_user(len, optlen))
6139 if (copy_to_user(optval, &assocparams, len))
6146 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
6148 * This socket option is a boolean flag which turns on or off mapped V4
6149 * addresses. If this option is turned on and the socket is type
6150 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
6151 * If this option is turned off, then no mapping will be done of V4
6152 * addresses and a user will receive both PF_INET6 and PF_INET type
6153 * addresses on the socket.
6155 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
6156 char __user *optval, int __user *optlen)
6159 struct sctp_sock *sp = sctp_sk(sk);
6161 if (len < sizeof(int))
6166 if (put_user(len, optlen))
6168 if (copy_to_user(optval, &val, len))
6175 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
6176 * (chapter and verse is quoted at sctp_setsockopt_context())
6178 static int sctp_getsockopt_context(struct sock *sk, int len,
6179 char __user *optval, int __user *optlen)
6181 struct sctp_assoc_value params;
6182 struct sctp_sock *sp;
6183 struct sctp_association *asoc;
6185 if (len < sizeof(struct sctp_assoc_value))
6188 len = sizeof(struct sctp_assoc_value);
6190 if (copy_from_user(¶ms, optval, len))
6195 if (params.assoc_id != 0) {
6196 asoc = sctp_id2assoc(sk, params.assoc_id);
6199 params.assoc_value = asoc->default_rcv_context;
6201 params.assoc_value = sp->default_rcv_context;
6204 if (put_user(len, optlen))
6206 if (copy_to_user(optval, ¶ms, len))
6213 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
6214 * This option will get or set the maximum size to put in any outgoing
6215 * SCTP DATA chunk. If a message is larger than this size it will be
6216 * fragmented by SCTP into the specified size. Note that the underlying
6217 * SCTP implementation may fragment into smaller sized chunks when the
6218 * PMTU of the underlying association is smaller than the value set by
6219 * the user. The default value for this option is '0' which indicates
6220 * the user is NOT limiting fragmentation and only the PMTU will effect
6221 * SCTP's choice of DATA chunk size. Note also that values set larger
6222 * than the maximum size of an IP datagram will effectively let SCTP
6223 * control fragmentation (i.e. the same as setting this option to 0).
6225 * The following structure is used to access and modify this parameter:
6227 * struct sctp_assoc_value {
6228 * sctp_assoc_t assoc_id;
6229 * uint32_t assoc_value;
6232 * assoc_id: This parameter is ignored for one-to-one style sockets.
6233 * For one-to-many style sockets this parameter indicates which
6234 * association the user is performing an action upon. Note that if
6235 * this field's value is zero then the endpoints default value is
6236 * changed (effecting future associations only).
6237 * assoc_value: This parameter specifies the maximum size in bytes.
6239 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
6240 char __user *optval, int __user *optlen)
6242 struct sctp_assoc_value params;
6243 struct sctp_association *asoc;
6245 if (len == sizeof(int)) {
6246 pr_warn_ratelimited(DEPRECATED
6248 "Use of int in maxseg socket option.\n"
6249 "Use struct sctp_assoc_value instead\n",
6250 current->comm, task_pid_nr(current));
6251 params.assoc_id = 0;
6252 } else if (len >= sizeof(struct sctp_assoc_value)) {
6253 len = sizeof(struct sctp_assoc_value);
6254 if (copy_from_user(¶ms, optval, len))
6259 asoc = sctp_id2assoc(sk, params.assoc_id);
6260 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
6264 params.assoc_value = asoc->frag_point;
6266 params.assoc_value = sctp_sk(sk)->user_frag;
6268 if (put_user(len, optlen))
6270 if (len == sizeof(int)) {
6271 if (copy_to_user(optval, ¶ms.assoc_value, len))
6274 if (copy_to_user(optval, ¶ms, len))
6282 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
6283 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
6285 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
6286 char __user *optval, int __user *optlen)
6290 if (len < sizeof(int))
6295 val = sctp_sk(sk)->frag_interleave;
6296 if (put_user(len, optlen))
6298 if (copy_to_user(optval, &val, len))
6305 * 7.1.25. Set or Get the sctp partial delivery point
6306 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
6308 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
6309 char __user *optval,
6314 if (len < sizeof(u32))
6319 val = sctp_sk(sk)->pd_point;
6320 if (put_user(len, optlen))
6322 if (copy_to_user(optval, &val, len))
6329 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
6330 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
6332 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
6333 char __user *optval,
6336 struct sctp_assoc_value params;
6337 struct sctp_sock *sp;
6338 struct sctp_association *asoc;
6340 if (len == sizeof(int)) {
6341 pr_warn_ratelimited(DEPRECATED
6343 "Use of int in max_burst socket option.\n"
6344 "Use struct sctp_assoc_value instead\n",
6345 current->comm, task_pid_nr(current));
6346 params.assoc_id = 0;
6347 } else if (len >= sizeof(struct sctp_assoc_value)) {
6348 len = sizeof(struct sctp_assoc_value);
6349 if (copy_from_user(¶ms, optval, len))
6356 if (params.assoc_id != 0) {
6357 asoc = sctp_id2assoc(sk, params.assoc_id);
6360 params.assoc_value = asoc->max_burst;
6362 params.assoc_value = sp->max_burst;
6364 if (len == sizeof(int)) {
6365 if (copy_to_user(optval, ¶ms.assoc_value, len))
6368 if (copy_to_user(optval, ¶ms, len))
6376 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
6377 char __user *optval, int __user *optlen)
6379 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6380 struct sctp_hmacalgo __user *p = (void __user *)optval;
6381 struct sctp_hmac_algo_param *hmacs;
6386 if (!ep->auth_enable)
6389 hmacs = ep->auth_hmacs_list;
6390 data_len = ntohs(hmacs->param_hdr.length) -
6391 sizeof(struct sctp_paramhdr);
6393 if (len < sizeof(struct sctp_hmacalgo) + data_len)
6396 len = sizeof(struct sctp_hmacalgo) + data_len;
6397 num_idents = data_len / sizeof(u16);
6399 if (put_user(len, optlen))
6401 if (put_user(num_idents, &p->shmac_num_idents))
6403 for (i = 0; i < num_idents; i++) {
6404 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
6406 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
6412 static int sctp_getsockopt_active_key(struct sock *sk, int len,
6413 char __user *optval, int __user *optlen)
6415 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6416 struct sctp_authkeyid val;
6417 struct sctp_association *asoc;
6419 if (!ep->auth_enable)
6422 if (len < sizeof(struct sctp_authkeyid))
6425 len = sizeof(struct sctp_authkeyid);
6426 if (copy_from_user(&val, optval, len))
6429 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
6430 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
6434 val.scact_keynumber = asoc->active_key_id;
6436 val.scact_keynumber = ep->active_key_id;
6438 if (put_user(len, optlen))
6440 if (copy_to_user(optval, &val, len))
6446 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
6447 char __user *optval, int __user *optlen)
6449 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6450 struct sctp_authchunks __user *p = (void __user *)optval;
6451 struct sctp_authchunks val;
6452 struct sctp_association *asoc;
6453 struct sctp_chunks_param *ch;
6457 if (!ep->auth_enable)
6460 if (len < sizeof(struct sctp_authchunks))
6463 if (copy_from_user(&val, optval, sizeof(val)))
6466 to = p->gauth_chunks;
6467 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6471 ch = asoc->peer.peer_chunks;
6475 /* See if the user provided enough room for all the data */
6476 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6477 if (len < num_chunks)
6480 if (copy_to_user(to, ch->chunks, num_chunks))
6483 len = sizeof(struct sctp_authchunks) + num_chunks;
6484 if (put_user(len, optlen))
6486 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6491 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
6492 char __user *optval, int __user *optlen)
6494 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6495 struct sctp_authchunks __user *p = (void __user *)optval;
6496 struct sctp_authchunks val;
6497 struct sctp_association *asoc;
6498 struct sctp_chunks_param *ch;
6502 if (!ep->auth_enable)
6505 if (len < sizeof(struct sctp_authchunks))
6508 if (copy_from_user(&val, optval, sizeof(val)))
6511 to = p->gauth_chunks;
6512 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6513 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
6517 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
6519 ch = ep->auth_chunk_list;
6524 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6525 if (len < sizeof(struct sctp_authchunks) + num_chunks)
6528 if (copy_to_user(to, ch->chunks, num_chunks))
6531 len = sizeof(struct sctp_authchunks) + num_chunks;
6532 if (put_user(len, optlen))
6534 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6541 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
6542 * This option gets the current number of associations that are attached
6543 * to a one-to-many style socket. The option value is an uint32_t.
6545 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
6546 char __user *optval, int __user *optlen)
6548 struct sctp_sock *sp = sctp_sk(sk);
6549 struct sctp_association *asoc;
6552 if (sctp_style(sk, TCP))
6555 if (len < sizeof(u32))
6560 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6564 if (put_user(len, optlen))
6566 if (copy_to_user(optval, &val, len))
6573 * 8.1.23 SCTP_AUTO_ASCONF
6574 * See the corresponding setsockopt entry as description
6576 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
6577 char __user *optval, int __user *optlen)
6581 if (len < sizeof(int))
6585 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
6587 if (put_user(len, optlen))
6589 if (copy_to_user(optval, &val, len))
6595 * 8.2.6. Get the Current Identifiers of Associations
6596 * (SCTP_GET_ASSOC_ID_LIST)
6598 * This option gets the current list of SCTP association identifiers of
6599 * the SCTP associations handled by a one-to-many style socket.
6601 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
6602 char __user *optval, int __user *optlen)
6604 struct sctp_sock *sp = sctp_sk(sk);
6605 struct sctp_association *asoc;
6606 struct sctp_assoc_ids *ids;
6609 if (sctp_style(sk, TCP))
6612 if (len < sizeof(struct sctp_assoc_ids))
6615 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6619 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
6622 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
6624 ids = kmalloc(len, GFP_USER | __GFP_NOWARN);
6628 ids->gaids_number_of_ids = num;
6630 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6631 ids->gaids_assoc_id[num++] = asoc->assoc_id;
6634 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
6644 * SCTP_PEER_ADDR_THLDS
6646 * This option allows us to fetch the partially failed threshold for one or all
6647 * transports in an association. See Section 6.1 of:
6648 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6650 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
6651 char __user *optval,
6655 struct sctp_paddrthlds val;
6656 struct sctp_transport *trans;
6657 struct sctp_association *asoc;
6659 if (len < sizeof(struct sctp_paddrthlds))
6661 len = sizeof(struct sctp_paddrthlds);
6662 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
6665 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
6666 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
6670 val.spt_pathpfthld = asoc->pf_retrans;
6671 val.spt_pathmaxrxt = asoc->pathmaxrxt;
6673 trans = sctp_addr_id2transport(sk, &val.spt_address,
6678 val.spt_pathmaxrxt = trans->pathmaxrxt;
6679 val.spt_pathpfthld = trans->pf_retrans;
6682 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
6689 * SCTP_GET_ASSOC_STATS
6691 * This option retrieves local per endpoint statistics. It is modeled
6692 * after OpenSolaris' implementation
6694 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
6695 char __user *optval,
6698 struct sctp_assoc_stats sas;
6699 struct sctp_association *asoc = NULL;
6701 /* User must provide at least the assoc id */
6702 if (len < sizeof(sctp_assoc_t))
6705 /* Allow the struct to grow and fill in as much as possible */
6706 len = min_t(size_t, len, sizeof(sas));
6708 if (copy_from_user(&sas, optval, len))
6711 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
6715 sas.sas_rtxchunks = asoc->stats.rtxchunks;
6716 sas.sas_gapcnt = asoc->stats.gapcnt;
6717 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
6718 sas.sas_osacks = asoc->stats.osacks;
6719 sas.sas_isacks = asoc->stats.isacks;
6720 sas.sas_octrlchunks = asoc->stats.octrlchunks;
6721 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
6722 sas.sas_oodchunks = asoc->stats.oodchunks;
6723 sas.sas_iodchunks = asoc->stats.iodchunks;
6724 sas.sas_ouodchunks = asoc->stats.ouodchunks;
6725 sas.sas_iuodchunks = asoc->stats.iuodchunks;
6726 sas.sas_idupchunks = asoc->stats.idupchunks;
6727 sas.sas_opackets = asoc->stats.opackets;
6728 sas.sas_ipackets = asoc->stats.ipackets;
6730 /* New high max rto observed, will return 0 if not a single
6731 * RTO update took place. obs_rto_ipaddr will be bogus
6734 sas.sas_maxrto = asoc->stats.max_obs_rto;
6735 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
6736 sizeof(struct sockaddr_storage));
6738 /* Mark beginning of a new observation period */
6739 asoc->stats.max_obs_rto = asoc->rto_min;
6741 if (put_user(len, optlen))
6744 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
6746 if (copy_to_user(optval, &sas, len))
6752 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
6753 char __user *optval,
6758 if (len < sizeof(int))
6762 if (sctp_sk(sk)->recvrcvinfo)
6764 if (put_user(len, optlen))
6766 if (copy_to_user(optval, &val, len))
6772 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
6773 char __user *optval,
6778 if (len < sizeof(int))
6782 if (sctp_sk(sk)->recvnxtinfo)
6784 if (put_user(len, optlen))
6786 if (copy_to_user(optval, &val, len))
6792 static int sctp_getsockopt_pr_supported(struct sock *sk, int len,
6793 char __user *optval,
6796 struct sctp_assoc_value params;
6797 struct sctp_association *asoc;
6798 int retval = -EFAULT;
6800 if (len < sizeof(params)) {
6805 len = sizeof(params);
6806 if (copy_from_user(¶ms, optval, len))
6809 asoc = sctp_id2assoc(sk, params.assoc_id);
6811 params.assoc_value = asoc->prsctp_enable;
6812 } else if (!params.assoc_id) {
6813 struct sctp_sock *sp = sctp_sk(sk);
6815 params.assoc_value = sp->ep->prsctp_enable;
6821 if (put_user(len, optlen))
6824 if (copy_to_user(optval, ¶ms, len))
6833 static int sctp_getsockopt_default_prinfo(struct sock *sk, int len,
6834 char __user *optval,
6837 struct sctp_default_prinfo info;
6838 struct sctp_association *asoc;
6839 int retval = -EFAULT;
6841 if (len < sizeof(info)) {
6847 if (copy_from_user(&info, optval, len))
6850 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
6852 info.pr_policy = SCTP_PR_POLICY(asoc->default_flags);
6853 info.pr_value = asoc->default_timetolive;
6854 } else if (!info.pr_assoc_id) {
6855 struct sctp_sock *sp = sctp_sk(sk);
6857 info.pr_policy = SCTP_PR_POLICY(sp->default_flags);
6858 info.pr_value = sp->default_timetolive;
6864 if (put_user(len, optlen))
6867 if (copy_to_user(optval, &info, len))
6876 static int sctp_getsockopt_pr_assocstatus(struct sock *sk, int len,
6877 char __user *optval,
6880 struct sctp_prstatus params;
6881 struct sctp_association *asoc;
6883 int retval = -EINVAL;
6885 if (len < sizeof(params))
6888 len = sizeof(params);
6889 if (copy_from_user(¶ms, optval, len)) {
6894 policy = params.sprstat_policy;
6895 if (policy & ~SCTP_PR_SCTP_MASK)
6898 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
6902 if (policy == SCTP_PR_SCTP_NONE) {
6903 params.sprstat_abandoned_unsent = 0;
6904 params.sprstat_abandoned_sent = 0;
6905 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
6906 params.sprstat_abandoned_unsent +=
6907 asoc->abandoned_unsent[policy];
6908 params.sprstat_abandoned_sent +=
6909 asoc->abandoned_sent[policy];
6912 params.sprstat_abandoned_unsent =
6913 asoc->abandoned_unsent[__SCTP_PR_INDEX(policy)];
6914 params.sprstat_abandoned_sent =
6915 asoc->abandoned_sent[__SCTP_PR_INDEX(policy)];
6918 if (put_user(len, optlen)) {
6923 if (copy_to_user(optval, ¶ms, len)) {
6934 static int sctp_getsockopt_pr_streamstatus(struct sock *sk, int len,
6935 char __user *optval,
6938 struct sctp_stream_out_ext *streamoute;
6939 struct sctp_association *asoc;
6940 struct sctp_prstatus params;
6941 int retval = -EINVAL;
6944 if (len < sizeof(params))
6947 len = sizeof(params);
6948 if (copy_from_user(¶ms, optval, len)) {
6953 policy = params.sprstat_policy;
6954 if (policy & ~SCTP_PR_SCTP_MASK)
6957 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
6958 if (!asoc || params.sprstat_sid >= asoc->stream.outcnt)
6961 streamoute = asoc->stream.out[params.sprstat_sid].ext;
6963 /* Not allocated yet, means all stats are 0 */
6964 params.sprstat_abandoned_unsent = 0;
6965 params.sprstat_abandoned_sent = 0;
6970 if (policy == SCTP_PR_SCTP_NONE) {
6971 params.sprstat_abandoned_unsent = 0;
6972 params.sprstat_abandoned_sent = 0;
6973 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
6974 params.sprstat_abandoned_unsent +=
6975 streamoute->abandoned_unsent[policy];
6976 params.sprstat_abandoned_sent +=
6977 streamoute->abandoned_sent[policy];
6980 params.sprstat_abandoned_unsent =
6981 streamoute->abandoned_unsent[__SCTP_PR_INDEX(policy)];
6982 params.sprstat_abandoned_sent =
6983 streamoute->abandoned_sent[__SCTP_PR_INDEX(policy)];
6986 if (put_user(len, optlen) || copy_to_user(optval, ¶ms, len)) {
6997 static int sctp_getsockopt_reconfig_supported(struct sock *sk, int len,
6998 char __user *optval,
7001 struct sctp_assoc_value params;
7002 struct sctp_association *asoc;
7003 int retval = -EFAULT;
7005 if (len < sizeof(params)) {
7010 len = sizeof(params);
7011 if (copy_from_user(¶ms, optval, len))
7014 asoc = sctp_id2assoc(sk, params.assoc_id);
7016 params.assoc_value = asoc->reconf_enable;
7017 } else if (!params.assoc_id) {
7018 struct sctp_sock *sp = sctp_sk(sk);
7020 params.assoc_value = sp->ep->reconf_enable;
7026 if (put_user(len, optlen))
7029 if (copy_to_user(optval, ¶ms, len))
7038 static int sctp_getsockopt_enable_strreset(struct sock *sk, int len,
7039 char __user *optval,
7042 struct sctp_assoc_value params;
7043 struct sctp_association *asoc;
7044 int retval = -EFAULT;
7046 if (len < sizeof(params)) {
7051 len = sizeof(params);
7052 if (copy_from_user(¶ms, optval, len))
7055 asoc = sctp_id2assoc(sk, params.assoc_id);
7057 params.assoc_value = asoc->strreset_enable;
7058 } else if (!params.assoc_id) {
7059 struct sctp_sock *sp = sctp_sk(sk);
7061 params.assoc_value = sp->ep->strreset_enable;
7067 if (put_user(len, optlen))
7070 if (copy_to_user(optval, ¶ms, len))
7079 static int sctp_getsockopt_scheduler(struct sock *sk, int len,
7080 char __user *optval,
7083 struct sctp_assoc_value params;
7084 struct sctp_association *asoc;
7085 int retval = -EFAULT;
7087 if (len < sizeof(params)) {
7092 len = sizeof(params);
7093 if (copy_from_user(¶ms, optval, len))
7096 asoc = sctp_id2assoc(sk, params.assoc_id);
7102 params.assoc_value = sctp_sched_get_sched(asoc);
7104 if (put_user(len, optlen))
7107 if (copy_to_user(optval, ¶ms, len))
7116 static int sctp_getsockopt_scheduler_value(struct sock *sk, int len,
7117 char __user *optval,
7120 struct sctp_stream_value params;
7121 struct sctp_association *asoc;
7122 int retval = -EFAULT;
7124 if (len < sizeof(params)) {
7129 len = sizeof(params);
7130 if (copy_from_user(¶ms, optval, len))
7133 asoc = sctp_id2assoc(sk, params.assoc_id);
7139 retval = sctp_sched_get_value(asoc, params.stream_id,
7140 ¶ms.stream_value);
7144 if (put_user(len, optlen)) {
7149 if (copy_to_user(optval, ¶ms, len)) {
7158 static int sctp_getsockopt_interleaving_supported(struct sock *sk, int len,
7159 char __user *optval,
7162 struct sctp_assoc_value params;
7163 struct sctp_association *asoc;
7164 int retval = -EFAULT;
7166 if (len < sizeof(params)) {
7171 len = sizeof(params);
7172 if (copy_from_user(¶ms, optval, len))
7175 asoc = sctp_id2assoc(sk, params.assoc_id);
7177 params.assoc_value = asoc->intl_enable;
7178 } else if (!params.assoc_id) {
7179 struct sctp_sock *sp = sctp_sk(sk);
7181 params.assoc_value = sp->strm_interleave;
7187 if (put_user(len, optlen))
7190 if (copy_to_user(optval, ¶ms, len))
7199 static int sctp_getsockopt(struct sock *sk, int level, int optname,
7200 char __user *optval, int __user *optlen)
7205 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
7207 /* I can hardly begin to describe how wrong this is. This is
7208 * so broken as to be worse than useless. The API draft
7209 * REALLY is NOT helpful here... I am not convinced that the
7210 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
7211 * are at all well-founded.
7213 if (level != SOL_SCTP) {
7214 struct sctp_af *af = sctp_sk(sk)->pf->af;
7216 retval = af->getsockopt(sk, level, optname, optval, optlen);
7220 if (get_user(len, optlen))
7230 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
7232 case SCTP_DISABLE_FRAGMENTS:
7233 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
7237 retval = sctp_getsockopt_events(sk, len, optval, optlen);
7239 case SCTP_AUTOCLOSE:
7240 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
7242 case SCTP_SOCKOPT_PEELOFF:
7243 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
7245 case SCTP_SOCKOPT_PEELOFF_FLAGS:
7246 retval = sctp_getsockopt_peeloff_flags(sk, len, optval, optlen);
7248 case SCTP_PEER_ADDR_PARAMS:
7249 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
7252 case SCTP_DELAYED_SACK:
7253 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
7257 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
7259 case SCTP_GET_PEER_ADDRS:
7260 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
7263 case SCTP_GET_LOCAL_ADDRS:
7264 retval = sctp_getsockopt_local_addrs(sk, len, optval,
7267 case SCTP_SOCKOPT_CONNECTX3:
7268 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
7270 case SCTP_DEFAULT_SEND_PARAM:
7271 retval = sctp_getsockopt_default_send_param(sk, len,
7274 case SCTP_DEFAULT_SNDINFO:
7275 retval = sctp_getsockopt_default_sndinfo(sk, len,
7278 case SCTP_PRIMARY_ADDR:
7279 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
7282 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
7285 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
7287 case SCTP_ASSOCINFO:
7288 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
7290 case SCTP_I_WANT_MAPPED_V4_ADDR:
7291 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
7294 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
7296 case SCTP_GET_PEER_ADDR_INFO:
7297 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
7300 case SCTP_ADAPTATION_LAYER:
7301 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
7305 retval = sctp_getsockopt_context(sk, len, optval, optlen);
7307 case SCTP_FRAGMENT_INTERLEAVE:
7308 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
7311 case SCTP_PARTIAL_DELIVERY_POINT:
7312 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
7315 case SCTP_MAX_BURST:
7316 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
7319 case SCTP_AUTH_CHUNK:
7320 case SCTP_AUTH_DELETE_KEY:
7321 case SCTP_AUTH_DEACTIVATE_KEY:
7322 retval = -EOPNOTSUPP;
7324 case SCTP_HMAC_IDENT:
7325 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
7327 case SCTP_AUTH_ACTIVE_KEY:
7328 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
7330 case SCTP_PEER_AUTH_CHUNKS:
7331 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
7334 case SCTP_LOCAL_AUTH_CHUNKS:
7335 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
7338 case SCTP_GET_ASSOC_NUMBER:
7339 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
7341 case SCTP_GET_ASSOC_ID_LIST:
7342 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
7344 case SCTP_AUTO_ASCONF:
7345 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
7347 case SCTP_PEER_ADDR_THLDS:
7348 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
7350 case SCTP_GET_ASSOC_STATS:
7351 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
7353 case SCTP_RECVRCVINFO:
7354 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
7356 case SCTP_RECVNXTINFO:
7357 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
7359 case SCTP_PR_SUPPORTED:
7360 retval = sctp_getsockopt_pr_supported(sk, len, optval, optlen);
7362 case SCTP_DEFAULT_PRINFO:
7363 retval = sctp_getsockopt_default_prinfo(sk, len, optval,
7366 case SCTP_PR_ASSOC_STATUS:
7367 retval = sctp_getsockopt_pr_assocstatus(sk, len, optval,
7370 case SCTP_PR_STREAM_STATUS:
7371 retval = sctp_getsockopt_pr_streamstatus(sk, len, optval,
7374 case SCTP_RECONFIG_SUPPORTED:
7375 retval = sctp_getsockopt_reconfig_supported(sk, len, optval,
7378 case SCTP_ENABLE_STREAM_RESET:
7379 retval = sctp_getsockopt_enable_strreset(sk, len, optval,
7382 case SCTP_STREAM_SCHEDULER:
7383 retval = sctp_getsockopt_scheduler(sk, len, optval,
7386 case SCTP_STREAM_SCHEDULER_VALUE:
7387 retval = sctp_getsockopt_scheduler_value(sk, len, optval,
7390 case SCTP_INTERLEAVING_SUPPORTED:
7391 retval = sctp_getsockopt_interleaving_supported(sk, len, optval,
7395 retval = -ENOPROTOOPT;
7403 static int sctp_hash(struct sock *sk)
7409 static void sctp_unhash(struct sock *sk)
7414 /* Check if port is acceptable. Possibly find first available port.
7416 * The port hash table (contained in the 'global' SCTP protocol storage
7417 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
7418 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
7419 * list (the list number is the port number hashed out, so as you
7420 * would expect from a hash function, all the ports in a given list have
7421 * such a number that hashes out to the same list number; you were
7422 * expecting that, right?); so each list has a set of ports, with a
7423 * link to the socket (struct sock) that uses it, the port number and
7424 * a fastreuse flag (FIXME: NPI ipg).
7426 static struct sctp_bind_bucket *sctp_bucket_create(
7427 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
7429 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
7431 struct sctp_bind_hashbucket *head; /* hash list */
7432 struct sctp_bind_bucket *pp;
7433 unsigned short snum;
7436 snum = ntohs(addr->v4.sin_port);
7438 pr_debug("%s: begins, snum:%d\n", __func__, snum);
7443 /* Search for an available port. */
7444 int low, high, remaining, index;
7446 struct net *net = sock_net(sk);
7448 inet_get_local_port_range(net, &low, &high);
7449 remaining = (high - low) + 1;
7450 rover = prandom_u32() % remaining + low;
7454 if ((rover < low) || (rover > high))
7456 if (inet_is_local_reserved_port(net, rover))
7458 index = sctp_phashfn(sock_net(sk), rover);
7459 head = &sctp_port_hashtable[index];
7460 spin_lock(&head->lock);
7461 sctp_for_each_hentry(pp, &head->chain)
7462 if ((pp->port == rover) &&
7463 net_eq(sock_net(sk), pp->net))
7467 spin_unlock(&head->lock);
7468 } while (--remaining > 0);
7470 /* Exhausted local port range during search? */
7475 /* OK, here is the one we will use. HEAD (the port
7476 * hash table list entry) is non-NULL and we hold it's
7481 /* We are given an specific port number; we verify
7482 * that it is not being used. If it is used, we will
7483 * exahust the search in the hash list corresponding
7484 * to the port number (snum) - we detect that with the
7485 * port iterator, pp being NULL.
7487 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
7488 spin_lock(&head->lock);
7489 sctp_for_each_hentry(pp, &head->chain) {
7490 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
7497 if (!hlist_empty(&pp->owner)) {
7498 /* We had a port hash table hit - there is an
7499 * available port (pp != NULL) and it is being
7500 * used by other socket (pp->owner not empty); that other
7501 * socket is going to be sk2.
7503 int reuse = sk->sk_reuse;
7506 pr_debug("%s: found a possible match\n", __func__);
7508 if (pp->fastreuse && sk->sk_reuse &&
7509 sk->sk_state != SCTP_SS_LISTENING)
7512 /* Run through the list of sockets bound to the port
7513 * (pp->port) [via the pointers bind_next and
7514 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
7515 * we get the endpoint they describe and run through
7516 * the endpoint's list of IP (v4 or v6) addresses,
7517 * comparing each of the addresses with the address of
7518 * the socket sk. If we find a match, then that means
7519 * that this port/socket (sk) combination are already
7522 sk_for_each_bound(sk2, &pp->owner) {
7523 struct sctp_endpoint *ep2;
7524 ep2 = sctp_sk(sk2)->ep;
7527 (reuse && sk2->sk_reuse &&
7528 sk2->sk_state != SCTP_SS_LISTENING))
7531 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
7532 sctp_sk(sk2), sctp_sk(sk))) {
7538 pr_debug("%s: found a match\n", __func__);
7541 /* If there was a hash table miss, create a new port. */
7543 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
7546 /* In either case (hit or miss), make sure fastreuse is 1 only
7547 * if sk->sk_reuse is too (that is, if the caller requested
7548 * SO_REUSEADDR on this socket -sk-).
7550 if (hlist_empty(&pp->owner)) {
7551 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
7555 } else if (pp->fastreuse &&
7556 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
7559 /* We are set, so fill up all the data in the hash table
7560 * entry, tie the socket list information with the rest of the
7561 * sockets FIXME: Blurry, NPI (ipg).
7564 if (!sctp_sk(sk)->bind_hash) {
7565 inet_sk(sk)->inet_num = snum;
7566 sk_add_bind_node(sk, &pp->owner);
7567 sctp_sk(sk)->bind_hash = pp;
7572 spin_unlock(&head->lock);
7579 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
7580 * port is requested.
7582 static int sctp_get_port(struct sock *sk, unsigned short snum)
7584 union sctp_addr addr;
7585 struct sctp_af *af = sctp_sk(sk)->pf->af;
7587 /* Set up a dummy address struct from the sk. */
7588 af->from_sk(&addr, sk);
7589 addr.v4.sin_port = htons(snum);
7591 /* Note: sk->sk_num gets filled in if ephemeral port request. */
7592 return !!sctp_get_port_local(sk, &addr);
7596 * Move a socket to LISTENING state.
7598 static int sctp_listen_start(struct sock *sk, int backlog)
7600 struct sctp_sock *sp = sctp_sk(sk);
7601 struct sctp_endpoint *ep = sp->ep;
7602 struct crypto_shash *tfm = NULL;
7605 /* Allocate HMAC for generating cookie. */
7606 if (!sp->hmac && sp->sctp_hmac_alg) {
7607 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
7608 tfm = crypto_alloc_shash(alg, 0, 0);
7610 net_info_ratelimited("failed to load transform for %s: %ld\n",
7611 sp->sctp_hmac_alg, PTR_ERR(tfm));
7614 sctp_sk(sk)->hmac = tfm;
7618 * If a bind() or sctp_bindx() is not called prior to a listen()
7619 * call that allows new associations to be accepted, the system
7620 * picks an ephemeral port and will choose an address set equivalent
7621 * to binding with a wildcard address.
7623 * This is not currently spelled out in the SCTP sockets
7624 * extensions draft, but follows the practice as seen in TCP
7628 inet_sk_set_state(sk, SCTP_SS_LISTENING);
7629 if (!ep->base.bind_addr.port) {
7630 if (sctp_autobind(sk))
7633 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
7634 inet_sk_set_state(sk, SCTP_SS_CLOSED);
7639 sk->sk_max_ack_backlog = backlog;
7640 sctp_hash_endpoint(ep);
7645 * 4.1.3 / 5.1.3 listen()
7647 * By default, new associations are not accepted for UDP style sockets.
7648 * An application uses listen() to mark a socket as being able to
7649 * accept new associations.
7651 * On TCP style sockets, applications use listen() to ready the SCTP
7652 * endpoint for accepting inbound associations.
7654 * On both types of endpoints a backlog of '0' disables listening.
7656 * Move a socket to LISTENING state.
7658 int sctp_inet_listen(struct socket *sock, int backlog)
7660 struct sock *sk = sock->sk;
7661 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
7664 if (unlikely(backlog < 0))
7669 /* Peeled-off sockets are not allowed to listen(). */
7670 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
7673 if (sock->state != SS_UNCONNECTED)
7676 if (!sctp_sstate(sk, LISTENING) && !sctp_sstate(sk, CLOSED))
7679 /* If backlog is zero, disable listening. */
7681 if (sctp_sstate(sk, CLOSED))
7685 sctp_unhash_endpoint(ep);
7686 sk->sk_state = SCTP_SS_CLOSED;
7688 sctp_sk(sk)->bind_hash->fastreuse = 1;
7692 /* If we are already listening, just update the backlog */
7693 if (sctp_sstate(sk, LISTENING))
7694 sk->sk_max_ack_backlog = backlog;
7696 err = sctp_listen_start(sk, backlog);
7708 * This function is done by modeling the current datagram_poll() and the
7709 * tcp_poll(). Note that, based on these implementations, we don't
7710 * lock the socket in this function, even though it seems that,
7711 * ideally, locking or some other mechanisms can be used to ensure
7712 * the integrity of the counters (sndbuf and wmem_alloc) used
7713 * in this place. We assume that we don't need locks either until proven
7716 * Another thing to note is that we include the Async I/O support
7717 * here, again, by modeling the current TCP/UDP code. We don't have
7718 * a good way to test with it yet.
7720 __poll_t sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
7722 struct sock *sk = sock->sk;
7723 struct sctp_sock *sp = sctp_sk(sk);
7726 poll_wait(file, sk_sleep(sk), wait);
7728 sock_rps_record_flow(sk);
7730 /* A TCP-style listening socket becomes readable when the accept queue
7733 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
7734 return (!list_empty(&sp->ep->asocs)) ?
7735 (EPOLLIN | EPOLLRDNORM) : 0;
7739 /* Is there any exceptional events? */
7740 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
7742 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? EPOLLPRI : 0);
7743 if (sk->sk_shutdown & RCV_SHUTDOWN)
7744 mask |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
7745 if (sk->sk_shutdown == SHUTDOWN_MASK)
7748 /* Is it readable? Reconsider this code with TCP-style support. */
7749 if (!skb_queue_empty(&sk->sk_receive_queue))
7750 mask |= EPOLLIN | EPOLLRDNORM;
7752 /* The association is either gone or not ready. */
7753 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
7756 /* Is it writable? */
7757 if (sctp_writeable(sk)) {
7758 mask |= EPOLLOUT | EPOLLWRNORM;
7760 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
7762 * Since the socket is not locked, the buffer
7763 * might be made available after the writeable check and
7764 * before the bit is set. This could cause a lost I/O
7765 * signal. tcp_poll() has a race breaker for this race
7766 * condition. Based on their implementation, we put
7767 * in the following code to cover it as well.
7769 if (sctp_writeable(sk))
7770 mask |= EPOLLOUT | EPOLLWRNORM;
7775 /********************************************************************
7776 * 2nd Level Abstractions
7777 ********************************************************************/
7779 static struct sctp_bind_bucket *sctp_bucket_create(
7780 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
7782 struct sctp_bind_bucket *pp;
7784 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
7786 SCTP_DBG_OBJCNT_INC(bind_bucket);
7789 INIT_HLIST_HEAD(&pp->owner);
7791 hlist_add_head(&pp->node, &head->chain);
7796 /* Caller must hold hashbucket lock for this tb with local BH disabled */
7797 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
7799 if (pp && hlist_empty(&pp->owner)) {
7800 __hlist_del(&pp->node);
7801 kmem_cache_free(sctp_bucket_cachep, pp);
7802 SCTP_DBG_OBJCNT_DEC(bind_bucket);
7806 /* Release this socket's reference to a local port. */
7807 static inline void __sctp_put_port(struct sock *sk)
7809 struct sctp_bind_hashbucket *head =
7810 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
7811 inet_sk(sk)->inet_num)];
7812 struct sctp_bind_bucket *pp;
7814 spin_lock(&head->lock);
7815 pp = sctp_sk(sk)->bind_hash;
7816 __sk_del_bind_node(sk);
7817 sctp_sk(sk)->bind_hash = NULL;
7818 inet_sk(sk)->inet_num = 0;
7819 sctp_bucket_destroy(pp);
7820 spin_unlock(&head->lock);
7823 void sctp_put_port(struct sock *sk)
7826 __sctp_put_port(sk);
7831 * The system picks an ephemeral port and choose an address set equivalent
7832 * to binding with a wildcard address.
7833 * One of those addresses will be the primary address for the association.
7834 * This automatically enables the multihoming capability of SCTP.
7836 static int sctp_autobind(struct sock *sk)
7838 union sctp_addr autoaddr;
7842 /* Initialize a local sockaddr structure to INADDR_ANY. */
7843 af = sctp_sk(sk)->pf->af;
7845 port = htons(inet_sk(sk)->inet_num);
7846 af->inaddr_any(&autoaddr, port);
7848 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
7851 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
7854 * 4.2 The cmsghdr Structure *
7856 * When ancillary data is sent or received, any number of ancillary data
7857 * objects can be specified by the msg_control and msg_controllen members of
7858 * the msghdr structure, because each object is preceded by
7859 * a cmsghdr structure defining the object's length (the cmsg_len member).
7860 * Historically Berkeley-derived implementations have passed only one object
7861 * at a time, but this API allows multiple objects to be
7862 * passed in a single call to sendmsg() or recvmsg(). The following example
7863 * shows two ancillary data objects in a control buffer.
7865 * |<--------------------------- msg_controllen -------------------------->|
7868 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
7870 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
7873 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
7875 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
7878 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7879 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
7881 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
7883 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7890 static int sctp_msghdr_parse(const struct msghdr *msg, struct sctp_cmsgs *cmsgs)
7892 struct msghdr *my_msg = (struct msghdr *)msg;
7893 struct cmsghdr *cmsg;
7895 for_each_cmsghdr(cmsg, my_msg) {
7896 if (!CMSG_OK(my_msg, cmsg))
7899 /* Should we parse this header or ignore? */
7900 if (cmsg->cmsg_level != IPPROTO_SCTP)
7903 /* Strictly check lengths following example in SCM code. */
7904 switch (cmsg->cmsg_type) {
7906 /* SCTP Socket API Extension
7907 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
7909 * This cmsghdr structure provides information for
7910 * initializing new SCTP associations with sendmsg().
7911 * The SCTP_INITMSG socket option uses this same data
7912 * structure. This structure is not used for
7915 * cmsg_level cmsg_type cmsg_data[]
7916 * ------------ ------------ ----------------------
7917 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
7919 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
7922 cmsgs->init = CMSG_DATA(cmsg);
7926 /* SCTP Socket API Extension
7927 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
7929 * This cmsghdr structure specifies SCTP options for
7930 * sendmsg() and describes SCTP header information
7931 * about a received message through recvmsg().
7933 * cmsg_level cmsg_type cmsg_data[]
7934 * ------------ ------------ ----------------------
7935 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
7937 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
7940 cmsgs->srinfo = CMSG_DATA(cmsg);
7942 if (cmsgs->srinfo->sinfo_flags &
7943 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7944 SCTP_SACK_IMMEDIATELY | SCTP_SENDALL |
7945 SCTP_PR_SCTP_MASK | SCTP_ABORT | SCTP_EOF))
7950 /* SCTP Socket API Extension
7951 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
7953 * This cmsghdr structure specifies SCTP options for
7954 * sendmsg(). This structure and SCTP_RCVINFO replaces
7955 * SCTP_SNDRCV which has been deprecated.
7957 * cmsg_level cmsg_type cmsg_data[]
7958 * ------------ ------------ ---------------------
7959 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
7961 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
7964 cmsgs->sinfo = CMSG_DATA(cmsg);
7966 if (cmsgs->sinfo->snd_flags &
7967 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7968 SCTP_SACK_IMMEDIATELY | SCTP_SENDALL |
7969 SCTP_PR_SCTP_MASK | SCTP_ABORT | SCTP_EOF))
7973 /* SCTP Socket API Extension
7974 * 5.3.7 SCTP PR-SCTP Information Structure (SCTP_PRINFO)
7976 * This cmsghdr structure specifies SCTP options for sendmsg().
7978 * cmsg_level cmsg_type cmsg_data[]
7979 * ------------ ------------ ---------------------
7980 * IPPROTO_SCTP SCTP_PRINFO struct sctp_prinfo
7982 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_prinfo)))
7985 cmsgs->prinfo = CMSG_DATA(cmsg);
7986 if (cmsgs->prinfo->pr_policy & ~SCTP_PR_SCTP_MASK)
7989 if (cmsgs->prinfo->pr_policy == SCTP_PR_SCTP_NONE)
7990 cmsgs->prinfo->pr_value = 0;
7993 /* SCTP Socket API Extension
7994 * 5.3.8 SCTP AUTH Information Structure (SCTP_AUTHINFO)
7996 * This cmsghdr structure specifies SCTP options for sendmsg().
7998 * cmsg_level cmsg_type cmsg_data[]
7999 * ------------ ------------ ---------------------
8000 * IPPROTO_SCTP SCTP_AUTHINFO struct sctp_authinfo
8002 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_authinfo)))
8005 cmsgs->authinfo = CMSG_DATA(cmsg);
8007 case SCTP_DSTADDRV4:
8008 case SCTP_DSTADDRV6:
8009 /* SCTP Socket API Extension
8010 * 5.3.9/10 SCTP Destination IPv4/6 Address Structure (SCTP_DSTADDRV4/6)
8012 * This cmsghdr structure specifies SCTP options for sendmsg().
8014 * cmsg_level cmsg_type cmsg_data[]
8015 * ------------ ------------ ---------------------
8016 * IPPROTO_SCTP SCTP_DSTADDRV4 struct in_addr
8017 * ------------ ------------ ---------------------
8018 * IPPROTO_SCTP SCTP_DSTADDRV6 struct in6_addr
8020 cmsgs->addrs_msg = my_msg;
8031 * Wait for a packet..
8032 * Note: This function is the same function as in core/datagram.c
8033 * with a few modifications to make lksctp work.
8035 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
8040 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
8042 /* Socket errors? */
8043 error = sock_error(sk);
8047 if (!skb_queue_empty(&sk->sk_receive_queue))
8050 /* Socket shut down? */
8051 if (sk->sk_shutdown & RCV_SHUTDOWN)
8054 /* Sequenced packets can come disconnected. If so we report the
8059 /* Is there a good reason to think that we may receive some data? */
8060 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
8063 /* Handle signals. */
8064 if (signal_pending(current))
8067 /* Let another process have a go. Since we are going to sleep
8068 * anyway. Note: This may cause odd behaviors if the message
8069 * does not fit in the user's buffer, but this seems to be the
8070 * only way to honor MSG_DONTWAIT realistically.
8073 *timeo_p = schedule_timeout(*timeo_p);
8077 finish_wait(sk_sleep(sk), &wait);
8081 error = sock_intr_errno(*timeo_p);
8084 finish_wait(sk_sleep(sk), &wait);
8089 /* Receive a datagram.
8090 * Note: This is pretty much the same routine as in core/datagram.c
8091 * with a few changes to make lksctp work.
8093 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
8094 int noblock, int *err)
8097 struct sk_buff *skb;
8100 timeo = sock_rcvtimeo(sk, noblock);
8102 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
8103 MAX_SCHEDULE_TIMEOUT);
8106 /* Again only user level code calls this function,
8107 * so nothing interrupt level
8108 * will suddenly eat the receive_queue.
8110 * Look at current nfs client by the way...
8111 * However, this function was correct in any case. 8)
8113 if (flags & MSG_PEEK) {
8114 skb = skb_peek(&sk->sk_receive_queue);
8116 refcount_inc(&skb->users);
8118 skb = __skb_dequeue(&sk->sk_receive_queue);
8124 /* Caller is allowed not to check sk->sk_err before calling. */
8125 error = sock_error(sk);
8129 if (sk->sk_shutdown & RCV_SHUTDOWN)
8132 if (sk_can_busy_loop(sk)) {
8133 sk_busy_loop(sk, noblock);
8135 if (!skb_queue_empty(&sk->sk_receive_queue))
8139 /* User doesn't want to wait. */
8143 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
8152 /* If sndbuf has changed, wake up per association sndbuf waiters. */
8153 static void __sctp_write_space(struct sctp_association *asoc)
8155 struct sock *sk = asoc->base.sk;
8157 if (sctp_wspace(asoc) <= 0)
8160 if (waitqueue_active(&asoc->wait))
8161 wake_up_interruptible(&asoc->wait);
8163 if (sctp_writeable(sk)) {
8164 struct socket_wq *wq;
8167 wq = rcu_dereference(sk->sk_wq);
8169 if (waitqueue_active(&wq->wait))
8170 wake_up_interruptible(&wq->wait);
8172 /* Note that we try to include the Async I/O support
8173 * here by modeling from the current TCP/UDP code.
8174 * We have not tested with it yet.
8176 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
8177 sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT);
8183 static void sctp_wake_up_waiters(struct sock *sk,
8184 struct sctp_association *asoc)
8186 struct sctp_association *tmp = asoc;
8188 /* We do accounting for the sndbuf space per association,
8189 * so we only need to wake our own association.
8191 if (asoc->ep->sndbuf_policy)
8192 return __sctp_write_space(asoc);
8194 /* If association goes down and is just flushing its
8195 * outq, then just normally notify others.
8197 if (asoc->base.dead)
8198 return sctp_write_space(sk);
8200 /* Accounting for the sndbuf space is per socket, so we
8201 * need to wake up others, try to be fair and in case of
8202 * other associations, let them have a go first instead
8203 * of just doing a sctp_write_space() call.
8205 * Note that we reach sctp_wake_up_waiters() only when
8206 * associations free up queued chunks, thus we are under
8207 * lock and the list of associations on a socket is
8208 * guaranteed not to change.
8210 for (tmp = list_next_entry(tmp, asocs); 1;
8211 tmp = list_next_entry(tmp, asocs)) {
8212 /* Manually skip the head element. */
8213 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
8215 /* Wake up association. */
8216 __sctp_write_space(tmp);
8217 /* We've reached the end. */
8223 /* Do accounting for the sndbuf space.
8224 * Decrement the used sndbuf space of the corresponding association by the
8225 * data size which was just transmitted(freed).
8227 static void sctp_wfree(struct sk_buff *skb)
8229 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
8230 struct sctp_association *asoc = chunk->asoc;
8231 struct sock *sk = asoc->base.sk;
8233 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
8234 sizeof(struct sk_buff) +
8235 sizeof(struct sctp_chunk);
8237 WARN_ON(refcount_sub_and_test(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc));
8240 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
8242 sk->sk_wmem_queued -= skb->truesize;
8243 sk_mem_uncharge(sk, skb->truesize);
8246 struct sctp_shared_key *shkey = chunk->shkey;
8248 /* refcnt == 2 and !list_empty mean after this release, it's
8249 * not being used anywhere, and it's time to notify userland
8250 * that this shkey can be freed if it's been deactivated.
8252 if (shkey->deactivated && !list_empty(&shkey->key_list) &&
8253 refcount_read(&shkey->refcnt) == 2) {
8254 struct sctp_ulpevent *ev;
8256 ev = sctp_ulpevent_make_authkey(asoc, shkey->key_id,
8260 asoc->stream.si->enqueue_event(&asoc->ulpq, ev);
8262 sctp_auth_shkey_release(chunk->shkey);
8266 sctp_wake_up_waiters(sk, asoc);
8268 sctp_association_put(asoc);
8271 /* Do accounting for the receive space on the socket.
8272 * Accounting for the association is done in ulpevent.c
8273 * We set this as a destructor for the cloned data skbs so that
8274 * accounting is done at the correct time.
8276 void sctp_sock_rfree(struct sk_buff *skb)
8278 struct sock *sk = skb->sk;
8279 struct sctp_ulpevent *event = sctp_skb2event(skb);
8281 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
8284 * Mimic the behavior of sock_rfree
8286 sk_mem_uncharge(sk, event->rmem_len);
8290 /* Helper function to wait for space in the sndbuf. */
8291 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
8294 struct sock *sk = asoc->base.sk;
8295 long current_timeo = *timeo_p;
8299 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
8302 /* Increment the association's refcnt. */
8303 sctp_association_hold(asoc);
8305 /* Wait on the association specific sndbuf space. */
8307 prepare_to_wait_exclusive(&asoc->wait, &wait,
8308 TASK_INTERRUPTIBLE);
8309 if (asoc->base.dead)
8313 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING)
8315 if (signal_pending(current))
8316 goto do_interrupted;
8317 if (msg_len <= sctp_wspace(asoc))
8320 /* Let another process have a go. Since we are going
8324 current_timeo = schedule_timeout(current_timeo);
8326 if (sk != asoc->base.sk)
8329 *timeo_p = current_timeo;
8333 finish_wait(&asoc->wait, &wait);
8335 /* Release the association's refcnt. */
8336 sctp_association_put(asoc);
8349 err = sock_intr_errno(*timeo_p);
8357 void sctp_data_ready(struct sock *sk)
8359 struct socket_wq *wq;
8362 wq = rcu_dereference(sk->sk_wq);
8363 if (skwq_has_sleeper(wq))
8364 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
8365 EPOLLRDNORM | EPOLLRDBAND);
8366 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
8370 /* If socket sndbuf has changed, wake up all per association waiters. */
8371 void sctp_write_space(struct sock *sk)
8373 struct sctp_association *asoc;
8375 /* Wake up the tasks in each wait queue. */
8376 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
8377 __sctp_write_space(asoc);
8381 /* Is there any sndbuf space available on the socket?
8383 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
8384 * associations on the same socket. For a UDP-style socket with
8385 * multiple associations, it is possible for it to be "unwriteable"
8386 * prematurely. I assume that this is acceptable because
8387 * a premature "unwriteable" is better than an accidental "writeable" which
8388 * would cause an unwanted block under certain circumstances. For the 1-1
8389 * UDP-style sockets or TCP-style sockets, this code should work.
8392 static int sctp_writeable(struct sock *sk)
8396 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
8402 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
8403 * returns immediately with EINPROGRESS.
8405 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
8407 struct sock *sk = asoc->base.sk;
8409 long current_timeo = *timeo_p;
8412 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
8414 /* Increment the association's refcnt. */
8415 sctp_association_hold(asoc);
8418 prepare_to_wait_exclusive(&asoc->wait, &wait,
8419 TASK_INTERRUPTIBLE);
8422 if (sk->sk_shutdown & RCV_SHUTDOWN)
8424 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
8427 if (signal_pending(current))
8428 goto do_interrupted;
8430 if (sctp_state(asoc, ESTABLISHED))
8433 /* Let another process have a go. Since we are going
8437 current_timeo = schedule_timeout(current_timeo);
8440 *timeo_p = current_timeo;
8444 finish_wait(&asoc->wait, &wait);
8446 /* Release the association's refcnt. */
8447 sctp_association_put(asoc);
8452 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
8455 err = -ECONNREFUSED;
8459 err = sock_intr_errno(*timeo_p);
8467 static int sctp_wait_for_accept(struct sock *sk, long timeo)
8469 struct sctp_endpoint *ep;
8473 ep = sctp_sk(sk)->ep;
8477 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
8478 TASK_INTERRUPTIBLE);
8480 if (list_empty(&ep->asocs)) {
8482 timeo = schedule_timeout(timeo);
8487 if (!sctp_sstate(sk, LISTENING))
8491 if (!list_empty(&ep->asocs))
8494 err = sock_intr_errno(timeo);
8495 if (signal_pending(current))
8503 finish_wait(sk_sleep(sk), &wait);
8508 static void sctp_wait_for_close(struct sock *sk, long timeout)
8513 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
8514 if (list_empty(&sctp_sk(sk)->ep->asocs))
8517 timeout = schedule_timeout(timeout);
8519 } while (!signal_pending(current) && timeout);
8521 finish_wait(sk_sleep(sk), &wait);
8524 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
8526 struct sk_buff *frag;
8531 /* Don't forget the fragments. */
8532 skb_walk_frags(skb, frag)
8533 sctp_skb_set_owner_r_frag(frag, sk);
8536 sctp_skb_set_owner_r(skb, sk);
8539 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
8540 struct sctp_association *asoc)
8542 struct inet_sock *inet = inet_sk(sk);
8543 struct inet_sock *newinet;
8544 struct sctp_sock *sp = sctp_sk(sk);
8545 struct sctp_endpoint *ep = sp->ep;
8547 newsk->sk_type = sk->sk_type;
8548 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
8549 newsk->sk_flags = sk->sk_flags;
8550 newsk->sk_tsflags = sk->sk_tsflags;
8551 newsk->sk_no_check_tx = sk->sk_no_check_tx;
8552 newsk->sk_no_check_rx = sk->sk_no_check_rx;
8553 newsk->sk_reuse = sk->sk_reuse;
8555 newsk->sk_shutdown = sk->sk_shutdown;
8556 newsk->sk_destruct = sctp_destruct_sock;
8557 newsk->sk_family = sk->sk_family;
8558 newsk->sk_protocol = IPPROTO_SCTP;
8559 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
8560 newsk->sk_sndbuf = sk->sk_sndbuf;
8561 newsk->sk_rcvbuf = sk->sk_rcvbuf;
8562 newsk->sk_lingertime = sk->sk_lingertime;
8563 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
8564 newsk->sk_sndtimeo = sk->sk_sndtimeo;
8565 newsk->sk_rxhash = sk->sk_rxhash;
8567 newinet = inet_sk(newsk);
8569 /* Initialize sk's sport, dport, rcv_saddr and daddr for
8570 * getsockname() and getpeername()
8572 newinet->inet_sport = inet->inet_sport;
8573 newinet->inet_saddr = inet->inet_saddr;
8574 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
8575 newinet->inet_dport = htons(asoc->peer.port);
8576 newinet->pmtudisc = inet->pmtudisc;
8577 newinet->inet_id = asoc->next_tsn ^ jiffies;
8579 newinet->uc_ttl = inet->uc_ttl;
8580 newinet->mc_loop = 1;
8581 newinet->mc_ttl = 1;
8582 newinet->mc_index = 0;
8583 newinet->mc_list = NULL;
8585 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
8586 net_enable_timestamp();
8588 /* Set newsk security attributes from orginal sk and connection
8589 * security attribute from ep.
8591 security_sctp_sk_clone(ep, sk, newsk);
8594 static inline void sctp_copy_descendant(struct sock *sk_to,
8595 const struct sock *sk_from)
8597 int ancestor_size = sizeof(struct inet_sock) +
8598 sizeof(struct sctp_sock) -
8599 offsetof(struct sctp_sock, auto_asconf_list);
8601 if (sk_from->sk_family == PF_INET6)
8602 ancestor_size += sizeof(struct ipv6_pinfo);
8604 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
8607 /* Populate the fields of the newsk from the oldsk and migrate the assoc
8608 * and its messages to the newsk.
8610 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
8611 struct sctp_association *assoc,
8612 enum sctp_socket_type type)
8614 struct sctp_sock *oldsp = sctp_sk(oldsk);
8615 struct sctp_sock *newsp = sctp_sk(newsk);
8616 struct sctp_bind_bucket *pp; /* hash list port iterator */
8617 struct sctp_endpoint *newep = newsp->ep;
8618 struct sk_buff *skb, *tmp;
8619 struct sctp_ulpevent *event;
8620 struct sctp_bind_hashbucket *head;
8622 /* Migrate socket buffer sizes and all the socket level options to the
8625 newsk->sk_sndbuf = oldsk->sk_sndbuf;
8626 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
8627 /* Brute force copy old sctp opt. */
8628 sctp_copy_descendant(newsk, oldsk);
8630 /* Restore the ep value that was overwritten with the above structure
8636 /* Hook this new socket in to the bind_hash list. */
8637 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
8638 inet_sk(oldsk)->inet_num)];
8639 spin_lock_bh(&head->lock);
8640 pp = sctp_sk(oldsk)->bind_hash;
8641 sk_add_bind_node(newsk, &pp->owner);
8642 sctp_sk(newsk)->bind_hash = pp;
8643 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
8644 spin_unlock_bh(&head->lock);
8646 /* Copy the bind_addr list from the original endpoint to the new
8647 * endpoint so that we can handle restarts properly
8649 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
8650 &oldsp->ep->base.bind_addr, GFP_KERNEL);
8652 /* Move any messages in the old socket's receive queue that are for the
8653 * peeled off association to the new socket's receive queue.
8655 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
8656 event = sctp_skb2event(skb);
8657 if (event->asoc == assoc) {
8658 __skb_unlink(skb, &oldsk->sk_receive_queue);
8659 __skb_queue_tail(&newsk->sk_receive_queue, skb);
8660 sctp_skb_set_owner_r_frag(skb, newsk);
8664 /* Clean up any messages pending delivery due to partial
8665 * delivery. Three cases:
8666 * 1) No partial deliver; no work.
8667 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
8668 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
8670 skb_queue_head_init(&newsp->pd_lobby);
8671 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
8673 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
8674 struct sk_buff_head *queue;
8676 /* Decide which queue to move pd_lobby skbs to. */
8677 if (assoc->ulpq.pd_mode) {
8678 queue = &newsp->pd_lobby;
8680 queue = &newsk->sk_receive_queue;
8682 /* Walk through the pd_lobby, looking for skbs that
8683 * need moved to the new socket.
8685 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
8686 event = sctp_skb2event(skb);
8687 if (event->asoc == assoc) {
8688 __skb_unlink(skb, &oldsp->pd_lobby);
8689 __skb_queue_tail(queue, skb);
8690 sctp_skb_set_owner_r_frag(skb, newsk);
8694 /* Clear up any skbs waiting for the partial
8695 * delivery to finish.
8697 if (assoc->ulpq.pd_mode)
8698 sctp_clear_pd(oldsk, NULL);
8702 sctp_for_each_rx_skb(assoc, newsk, sctp_skb_set_owner_r_frag);
8704 /* Set the type of socket to indicate that it is peeled off from the
8705 * original UDP-style socket or created with the accept() call on a
8706 * TCP-style socket..
8710 /* Mark the new socket "in-use" by the user so that any packets
8711 * that may arrive on the association after we've moved it are
8712 * queued to the backlog. This prevents a potential race between
8713 * backlog processing on the old socket and new-packet processing
8714 * on the new socket.
8716 * The caller has just allocated newsk so we can guarantee that other
8717 * paths won't try to lock it and then oldsk.
8719 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
8720 sctp_for_each_tx_datachunk(assoc, sctp_clear_owner_w);
8721 sctp_assoc_migrate(assoc, newsk);
8722 sctp_for_each_tx_datachunk(assoc, sctp_set_owner_w);
8724 /* If the association on the newsk is already closed before accept()
8725 * is called, set RCV_SHUTDOWN flag.
8727 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP)) {
8728 inet_sk_set_state(newsk, SCTP_SS_CLOSED);
8729 newsk->sk_shutdown |= RCV_SHUTDOWN;
8731 inet_sk_set_state(newsk, SCTP_SS_ESTABLISHED);
8734 release_sock(newsk);
8738 /* This proto struct describes the ULP interface for SCTP. */
8739 struct proto sctp_prot = {
8741 .owner = THIS_MODULE,
8742 .close = sctp_close,
8743 .disconnect = sctp_disconnect,
8744 .accept = sctp_accept,
8745 .ioctl = sctp_ioctl,
8746 .init = sctp_init_sock,
8747 .destroy = sctp_destroy_sock,
8748 .shutdown = sctp_shutdown,
8749 .setsockopt = sctp_setsockopt,
8750 .getsockopt = sctp_getsockopt,
8751 .sendmsg = sctp_sendmsg,
8752 .recvmsg = sctp_recvmsg,
8754 .backlog_rcv = sctp_backlog_rcv,
8756 .unhash = sctp_unhash,
8757 .get_port = sctp_get_port,
8758 .obj_size = sizeof(struct sctp_sock),
8759 .useroffset = offsetof(struct sctp_sock, subscribe),
8760 .usersize = offsetof(struct sctp_sock, initmsg) -
8761 offsetof(struct sctp_sock, subscribe) +
8762 sizeof_field(struct sctp_sock, initmsg),
8763 .sysctl_mem = sysctl_sctp_mem,
8764 .sysctl_rmem = sysctl_sctp_rmem,
8765 .sysctl_wmem = sysctl_sctp_wmem,
8766 .memory_pressure = &sctp_memory_pressure,
8767 .enter_memory_pressure = sctp_enter_memory_pressure,
8768 .memory_allocated = &sctp_memory_allocated,
8769 .sockets_allocated = &sctp_sockets_allocated,
8772 #if IS_ENABLED(CONFIG_IPV6)
8774 #include <net/transp_v6.h>
8775 static void sctp_v6_destroy_sock(struct sock *sk)
8777 sctp_destroy_sock(sk);
8778 inet6_destroy_sock(sk);
8781 struct proto sctpv6_prot = {
8783 .owner = THIS_MODULE,
8784 .close = sctp_close,
8785 .disconnect = sctp_disconnect,
8786 .accept = sctp_accept,
8787 .ioctl = sctp_ioctl,
8788 .init = sctp_init_sock,
8789 .destroy = sctp_v6_destroy_sock,
8790 .shutdown = sctp_shutdown,
8791 .setsockopt = sctp_setsockopt,
8792 .getsockopt = sctp_getsockopt,
8793 .sendmsg = sctp_sendmsg,
8794 .recvmsg = sctp_recvmsg,
8796 .backlog_rcv = sctp_backlog_rcv,
8798 .unhash = sctp_unhash,
8799 .get_port = sctp_get_port,
8800 .obj_size = sizeof(struct sctp6_sock),
8801 .useroffset = offsetof(struct sctp6_sock, sctp.subscribe),
8802 .usersize = offsetof(struct sctp6_sock, sctp.initmsg) -
8803 offsetof(struct sctp6_sock, sctp.subscribe) +
8804 sizeof_field(struct sctp6_sock, sctp.initmsg),
8805 .sysctl_mem = sysctl_sctp_mem,
8806 .sysctl_rmem = sysctl_sctp_rmem,
8807 .sysctl_wmem = sysctl_sctp_wmem,
8808 .memory_pressure = &sctp_memory_pressure,
8809 .enter_memory_pressure = sctp_enter_memory_pressure,
8810 .memory_allocated = &sctp_memory_allocated,
8811 .sockets_allocated = &sctp_sockets_allocated,
8813 #endif /* IS_ENABLED(CONFIG_IPV6) */