1 /* SCTP kernel implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
9 * This file is part of the SCTP kernel implementation
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
18 * This SCTP implementation is free software;
19 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
24 * This SCTP implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, see
32 * <http://www.gnu.org/licenses/>.
34 * Please send any bug reports or fixes you make to the
36 * lksctp developers <linux-sctp@vger.kernel.org>
38 * Written or modified by:
39 * La Monte H.P. Yarroll <piggy@acm.org>
40 * Narasimha Budihal <narsi@refcode.org>
41 * Karl Knutson <karl@athena.chicago.il.us>
42 * Jon Grimm <jgrimm@us.ibm.com>
43 * Xingang Guo <xingang.guo@intel.com>
44 * Daisy Chang <daisyc@us.ibm.com>
45 * Sridhar Samudrala <samudrala@us.ibm.com>
46 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
47 * Ardelle Fan <ardelle.fan@intel.com>
48 * Ryan Layer <rmlayer@us.ibm.com>
49 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
50 * Kevin Gao <kevin.gao@intel.com>
53 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
55 #include <crypto/hash.h>
56 #include <linux/types.h>
57 #include <linux/kernel.h>
58 #include <linux/wait.h>
59 #include <linux/time.h>
60 #include <linux/sched/signal.h>
62 #include <linux/capability.h>
63 #include <linux/fcntl.h>
64 #include <linux/poll.h>
65 #include <linux/init.h>
66 #include <linux/slab.h>
67 #include <linux/file.h>
68 #include <linux/compat.h>
69 #include <linux/rhashtable.h>
73 #include <net/route.h>
75 #include <net/inet_common.h>
76 #include <net/busy_poll.h>
78 #include <linux/socket.h> /* for sa_family_t */
79 #include <linux/export.h>
81 #include <net/sctp/sctp.h>
82 #include <net/sctp/sm.h>
83 #include <net/sctp/stream_sched.h>
85 /* Forward declarations for internal helper functions. */
86 static int sctp_writeable(struct sock *sk);
87 static void sctp_wfree(struct sk_buff *skb);
88 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
90 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p);
91 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
92 static int sctp_wait_for_accept(struct sock *sk, long timeo);
93 static void sctp_wait_for_close(struct sock *sk, long timeo);
94 static void sctp_destruct_sock(struct sock *sk);
95 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
96 union sctp_addr *addr, int len);
97 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
98 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
99 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
100 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
101 static int sctp_send_asconf(struct sctp_association *asoc,
102 struct sctp_chunk *chunk);
103 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
104 static int sctp_autobind(struct sock *sk);
105 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
106 struct sctp_association *assoc,
107 enum sctp_socket_type type);
109 static unsigned long sctp_memory_pressure;
110 static atomic_long_t sctp_memory_allocated;
111 struct percpu_counter sctp_sockets_allocated;
113 static void sctp_enter_memory_pressure(struct sock *sk)
115 sctp_memory_pressure = 1;
119 /* Get the sndbuf space available at the time on the association. */
120 static inline int sctp_wspace(struct sctp_association *asoc)
124 if (asoc->ep->sndbuf_policy)
125 amt = asoc->sndbuf_used;
127 amt = sk_wmem_alloc_get(asoc->base.sk);
129 if (amt >= asoc->base.sk->sk_sndbuf) {
130 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
133 amt = sk_stream_wspace(asoc->base.sk);
138 amt = asoc->base.sk->sk_sndbuf - amt;
143 /* Increment the used sndbuf space count of the corresponding association by
144 * the size of the outgoing data chunk.
145 * Also, set the skb destructor for sndbuf accounting later.
147 * Since it is always 1-1 between chunk and skb, and also a new skb is always
148 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
149 * destructor in the data chunk skb for the purpose of the sndbuf space
152 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
154 struct sctp_association *asoc = chunk->asoc;
155 struct sock *sk = asoc->base.sk;
157 /* The sndbuf space is tracked per association. */
158 sctp_association_hold(asoc);
161 sctp_auth_shkey_hold(chunk->shkey);
163 skb_set_owner_w(chunk->skb, sk);
165 chunk->skb->destructor = sctp_wfree;
166 /* Save the chunk pointer in skb for sctp_wfree to use later. */
167 skb_shinfo(chunk->skb)->destructor_arg = chunk;
169 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
170 sizeof(struct sk_buff) +
171 sizeof(struct sctp_chunk);
173 refcount_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
174 sk->sk_wmem_queued += chunk->skb->truesize;
175 sk_mem_charge(sk, chunk->skb->truesize);
178 static void sctp_clear_owner_w(struct sctp_chunk *chunk)
180 skb_orphan(chunk->skb);
183 static void sctp_for_each_tx_datachunk(struct sctp_association *asoc,
184 void (*cb)(struct sctp_chunk *))
187 struct sctp_outq *q = &asoc->outqueue;
188 struct sctp_transport *t;
189 struct sctp_chunk *chunk;
191 list_for_each_entry(t, &asoc->peer.transport_addr_list, transports)
192 list_for_each_entry(chunk, &t->transmitted, transmitted_list)
195 list_for_each_entry(chunk, &q->retransmit, transmitted_list)
198 list_for_each_entry(chunk, &q->sacked, transmitted_list)
201 list_for_each_entry(chunk, &q->abandoned, transmitted_list)
204 list_for_each_entry(chunk, &q->out_chunk_list, list)
208 static void sctp_for_each_rx_skb(struct sctp_association *asoc, struct sock *sk,
209 void (*cb)(struct sk_buff *, struct sock *))
212 struct sk_buff *skb, *tmp;
214 sctp_skb_for_each(skb, &asoc->ulpq.lobby, tmp)
217 sctp_skb_for_each(skb, &asoc->ulpq.reasm, tmp)
220 sctp_skb_for_each(skb, &asoc->ulpq.reasm_uo, tmp)
224 /* Verify that this is a valid address. */
225 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
230 /* Verify basic sockaddr. */
231 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
235 /* Is this a valid SCTP address? */
236 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
239 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
245 /* Look up the association by its id. If this is not a UDP-style
246 * socket, the ID field is always ignored.
248 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
250 struct sctp_association *asoc = NULL;
252 /* If this is not a UDP-style socket, assoc id should be ignored. */
253 if (!sctp_style(sk, UDP)) {
254 /* Return NULL if the socket state is not ESTABLISHED. It
255 * could be a TCP-style listening socket or a socket which
256 * hasn't yet called connect() to establish an association.
258 if (!sctp_sstate(sk, ESTABLISHED) && !sctp_sstate(sk, CLOSING))
261 /* Get the first and the only association from the list. */
262 if (!list_empty(&sctp_sk(sk)->ep->asocs))
263 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
264 struct sctp_association, asocs);
268 /* Otherwise this is a UDP-style socket. */
269 if (!id || (id == (sctp_assoc_t)-1))
272 spin_lock_bh(&sctp_assocs_id_lock);
273 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
274 spin_unlock_bh(&sctp_assocs_id_lock);
276 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
282 /* Look up the transport from an address and an assoc id. If both address and
283 * id are specified, the associations matching the address and the id should be
286 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
287 struct sockaddr_storage *addr,
290 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
291 struct sctp_af *af = sctp_get_af_specific(addr->ss_family);
292 union sctp_addr *laddr = (union sctp_addr *)addr;
293 struct sctp_transport *transport;
295 if (!af || sctp_verify_addr(sk, laddr, af->sockaddr_len))
298 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
305 id_asoc = sctp_id2assoc(sk, id);
306 if (id_asoc && (id_asoc != addr_asoc))
309 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
310 (union sctp_addr *)addr);
315 /* API 3.1.2 bind() - UDP Style Syntax
316 * The syntax of bind() is,
318 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
320 * sd - the socket descriptor returned by socket().
321 * addr - the address structure (struct sockaddr_in or struct
322 * sockaddr_in6 [RFC 2553]),
323 * addr_len - the size of the address structure.
325 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
331 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
334 /* Disallow binding twice. */
335 if (!sctp_sk(sk)->ep->base.bind_addr.port)
336 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
346 static long sctp_get_port_local(struct sock *, union sctp_addr *);
348 /* Verify this is a valid sockaddr. */
349 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
350 union sctp_addr *addr, int len)
354 /* Check minimum size. */
355 if (len < sizeof (struct sockaddr))
358 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
361 if (addr->sa.sa_family == AF_INET6) {
362 if (len < SIN6_LEN_RFC2133)
364 /* V4 mapped address are really of AF_INET family */
365 if (ipv6_addr_v4mapped(&addr->v6.sin6_addr) &&
366 !opt->pf->af_supported(AF_INET, opt))
370 /* If we get this far, af is valid. */
371 af = sctp_get_af_specific(addr->sa.sa_family);
373 if (len < af->sockaddr_len)
379 /* Bind a local address either to an endpoint or to an association. */
380 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
382 struct net *net = sock_net(sk);
383 struct sctp_sock *sp = sctp_sk(sk);
384 struct sctp_endpoint *ep = sp->ep;
385 struct sctp_bind_addr *bp = &ep->base.bind_addr;
390 /* Common sockaddr verification. */
391 af = sctp_sockaddr_af(sp, addr, len);
393 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
394 __func__, sk, addr, len);
398 snum = ntohs(addr->v4.sin_port);
400 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
401 __func__, sk, &addr->sa, bp->port, snum, len);
403 /* PF specific bind() address verification. */
404 if (!sp->pf->bind_verify(sp, addr))
405 return -EADDRNOTAVAIL;
407 /* We must either be unbound, or bind to the same port.
408 * It's OK to allow 0 ports if we are already bound.
409 * We'll just inhert an already bound port in this case
414 else if (snum != bp->port) {
415 pr_debug("%s: new port %d doesn't match existing port "
416 "%d\n", __func__, snum, bp->port);
421 if (snum && snum < inet_prot_sock(net) &&
422 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
425 /* See if the address matches any of the addresses we may have
426 * already bound before checking against other endpoints.
428 if (sctp_bind_addr_match(bp, addr, sp))
431 /* Make sure we are allowed to bind here.
432 * The function sctp_get_port_local() does duplicate address
435 addr->v4.sin_port = htons(snum);
436 if ((ret = sctp_get_port_local(sk, addr))) {
440 /* Refresh ephemeral port. */
442 bp->port = inet_sk(sk)->inet_num;
444 /* Add the address to the bind address list.
445 * Use GFP_ATOMIC since BHs will be disabled.
447 ret = sctp_add_bind_addr(bp, addr, af->sockaddr_len,
448 SCTP_ADDR_SRC, GFP_ATOMIC);
450 /* Copy back into socket for getsockname() use. */
452 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
453 sp->pf->to_sk_saddr(addr, sk);
459 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
461 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
462 * at any one time. If a sender, after sending an ASCONF chunk, decides
463 * it needs to transfer another ASCONF Chunk, it MUST wait until the
464 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
465 * subsequent ASCONF. Note this restriction binds each side, so at any
466 * time two ASCONF may be in-transit on any given association (one sent
467 * from each endpoint).
469 static int sctp_send_asconf(struct sctp_association *asoc,
470 struct sctp_chunk *chunk)
472 struct net *net = sock_net(asoc->base.sk);
475 /* If there is an outstanding ASCONF chunk, queue it for later
478 if (asoc->addip_last_asconf) {
479 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
483 /* Hold the chunk until an ASCONF_ACK is received. */
484 sctp_chunk_hold(chunk);
485 retval = sctp_primitive_ASCONF(net, asoc, chunk);
487 sctp_chunk_free(chunk);
489 asoc->addip_last_asconf = chunk;
495 /* Add a list of addresses as bind addresses to local endpoint or
498 * Basically run through each address specified in the addrs/addrcnt
499 * array/length pair, determine if it is IPv6 or IPv4 and call
500 * sctp_do_bind() on it.
502 * If any of them fails, then the operation will be reversed and the
503 * ones that were added will be removed.
505 * Only sctp_setsockopt_bindx() is supposed to call this function.
507 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
512 struct sockaddr *sa_addr;
515 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
519 for (cnt = 0; cnt < addrcnt; cnt++) {
520 /* The list may contain either IPv4 or IPv6 address;
521 * determine the address length for walking thru the list.
524 af = sctp_get_af_specific(sa_addr->sa_family);
530 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
533 addr_buf += af->sockaddr_len;
537 /* Failed. Cleanup the ones that have been added */
539 sctp_bindx_rem(sk, addrs, cnt);
547 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
548 * associations that are part of the endpoint indicating that a list of local
549 * addresses are added to the endpoint.
551 * If any of the addresses is already in the bind address list of the
552 * association, we do not send the chunk for that association. But it will not
553 * affect other associations.
555 * Only sctp_setsockopt_bindx() is supposed to call this function.
557 static int sctp_send_asconf_add_ip(struct sock *sk,
558 struct sockaddr *addrs,
561 struct net *net = sock_net(sk);
562 struct sctp_sock *sp;
563 struct sctp_endpoint *ep;
564 struct sctp_association *asoc;
565 struct sctp_bind_addr *bp;
566 struct sctp_chunk *chunk;
567 struct sctp_sockaddr_entry *laddr;
568 union sctp_addr *addr;
569 union sctp_addr saveaddr;
576 if (!net->sctp.addip_enable)
582 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
583 __func__, sk, addrs, addrcnt);
585 list_for_each_entry(asoc, &ep->asocs, asocs) {
586 if (!asoc->peer.asconf_capable)
589 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
592 if (!sctp_state(asoc, ESTABLISHED))
595 /* Check if any address in the packed array of addresses is
596 * in the bind address list of the association. If so,
597 * do not send the asconf chunk to its peer, but continue with
598 * other associations.
601 for (i = 0; i < addrcnt; i++) {
603 af = sctp_get_af_specific(addr->v4.sin_family);
609 if (sctp_assoc_lookup_laddr(asoc, addr))
612 addr_buf += af->sockaddr_len;
617 /* Use the first valid address in bind addr list of
618 * association as Address Parameter of ASCONF CHUNK.
620 bp = &asoc->base.bind_addr;
621 p = bp->address_list.next;
622 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
623 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
624 addrcnt, SCTP_PARAM_ADD_IP);
630 /* Add the new addresses to the bind address list with
631 * use_as_src set to 0.
634 for (i = 0; i < addrcnt; i++) {
636 af = sctp_get_af_specific(addr->v4.sin_family);
637 memcpy(&saveaddr, addr, af->sockaddr_len);
638 retval = sctp_add_bind_addr(bp, &saveaddr,
640 SCTP_ADDR_NEW, GFP_ATOMIC);
641 addr_buf += af->sockaddr_len;
643 if (asoc->src_out_of_asoc_ok) {
644 struct sctp_transport *trans;
646 list_for_each_entry(trans,
647 &asoc->peer.transport_addr_list, transports) {
648 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
649 2*asoc->pathmtu, 4380));
650 trans->ssthresh = asoc->peer.i.a_rwnd;
651 trans->rto = asoc->rto_initial;
652 sctp_max_rto(asoc, trans);
653 trans->rtt = trans->srtt = trans->rttvar = 0;
654 /* Clear the source and route cache */
655 sctp_transport_route(trans, NULL,
656 sctp_sk(asoc->base.sk));
659 retval = sctp_send_asconf(asoc, chunk);
666 /* Remove a list of addresses from bind addresses list. Do not remove the
669 * Basically run through each address specified in the addrs/addrcnt
670 * array/length pair, determine if it is IPv6 or IPv4 and call
671 * sctp_del_bind() on it.
673 * If any of them fails, then the operation will be reversed and the
674 * ones that were removed will be added back.
676 * At least one address has to be left; if only one address is
677 * available, the operation will return -EBUSY.
679 * Only sctp_setsockopt_bindx() is supposed to call this function.
681 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
683 struct sctp_sock *sp = sctp_sk(sk);
684 struct sctp_endpoint *ep = sp->ep;
686 struct sctp_bind_addr *bp = &ep->base.bind_addr;
689 union sctp_addr *sa_addr;
692 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
693 __func__, sk, addrs, addrcnt);
696 for (cnt = 0; cnt < addrcnt; cnt++) {
697 /* If the bind address list is empty or if there is only one
698 * bind address, there is nothing more to be removed (we need
699 * at least one address here).
701 if (list_empty(&bp->address_list) ||
702 (sctp_list_single_entry(&bp->address_list))) {
708 af = sctp_get_af_specific(sa_addr->sa.sa_family);
714 if (!af->addr_valid(sa_addr, sp, NULL)) {
715 retval = -EADDRNOTAVAIL;
719 if (sa_addr->v4.sin_port &&
720 sa_addr->v4.sin_port != htons(bp->port)) {
725 if (!sa_addr->v4.sin_port)
726 sa_addr->v4.sin_port = htons(bp->port);
728 /* FIXME - There is probably a need to check if sk->sk_saddr and
729 * sk->sk_rcv_addr are currently set to one of the addresses to
730 * be removed. This is something which needs to be looked into
731 * when we are fixing the outstanding issues with multi-homing
732 * socket routing and failover schemes. Refer to comments in
733 * sctp_do_bind(). -daisy
735 retval = sctp_del_bind_addr(bp, sa_addr);
737 addr_buf += af->sockaddr_len;
740 /* Failed. Add the ones that has been removed back */
742 sctp_bindx_add(sk, addrs, cnt);
750 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
751 * the associations that are part of the endpoint indicating that a list of
752 * local addresses are removed from the endpoint.
754 * If any of the addresses is already in the bind address list of the
755 * association, we do not send the chunk for that association. But it will not
756 * affect other associations.
758 * Only sctp_setsockopt_bindx() is supposed to call this function.
760 static int sctp_send_asconf_del_ip(struct sock *sk,
761 struct sockaddr *addrs,
764 struct net *net = sock_net(sk);
765 struct sctp_sock *sp;
766 struct sctp_endpoint *ep;
767 struct sctp_association *asoc;
768 struct sctp_transport *transport;
769 struct sctp_bind_addr *bp;
770 struct sctp_chunk *chunk;
771 union sctp_addr *laddr;
774 struct sctp_sockaddr_entry *saddr;
780 if (!net->sctp.addip_enable)
786 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
787 __func__, sk, addrs, addrcnt);
789 list_for_each_entry(asoc, &ep->asocs, asocs) {
791 if (!asoc->peer.asconf_capable)
794 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
797 if (!sctp_state(asoc, ESTABLISHED))
800 /* Check if any address in the packed array of addresses is
801 * not present in the bind address list of the association.
802 * If so, do not send the asconf chunk to its peer, but
803 * continue with other associations.
806 for (i = 0; i < addrcnt; i++) {
808 af = sctp_get_af_specific(laddr->v4.sin_family);
814 if (!sctp_assoc_lookup_laddr(asoc, laddr))
817 addr_buf += af->sockaddr_len;
822 /* Find one address in the association's bind address list
823 * that is not in the packed array of addresses. This is to
824 * make sure that we do not delete all the addresses in the
827 bp = &asoc->base.bind_addr;
828 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
830 if ((laddr == NULL) && (addrcnt == 1)) {
831 if (asoc->asconf_addr_del_pending)
833 asoc->asconf_addr_del_pending =
834 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
835 if (asoc->asconf_addr_del_pending == NULL) {
839 asoc->asconf_addr_del_pending->sa.sa_family =
841 asoc->asconf_addr_del_pending->v4.sin_port =
843 if (addrs->sa_family == AF_INET) {
844 struct sockaddr_in *sin;
846 sin = (struct sockaddr_in *)addrs;
847 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
848 } else if (addrs->sa_family == AF_INET6) {
849 struct sockaddr_in6 *sin6;
851 sin6 = (struct sockaddr_in6 *)addrs;
852 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
855 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
856 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
857 asoc->asconf_addr_del_pending);
859 asoc->src_out_of_asoc_ok = 1;
867 /* We do not need RCU protection throughout this loop
868 * because this is done under a socket lock from the
871 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
879 /* Reset use_as_src flag for the addresses in the bind address
880 * list that are to be deleted.
883 for (i = 0; i < addrcnt; i++) {
885 af = sctp_get_af_specific(laddr->v4.sin_family);
886 list_for_each_entry(saddr, &bp->address_list, list) {
887 if (sctp_cmp_addr_exact(&saddr->a, laddr))
888 saddr->state = SCTP_ADDR_DEL;
890 addr_buf += af->sockaddr_len;
893 /* Update the route and saddr entries for all the transports
894 * as some of the addresses in the bind address list are
895 * about to be deleted and cannot be used as source addresses.
897 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
899 sctp_transport_route(transport, NULL,
900 sctp_sk(asoc->base.sk));
904 /* We don't need to transmit ASCONF */
906 retval = sctp_send_asconf(asoc, chunk);
912 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
913 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
915 struct sock *sk = sctp_opt2sk(sp);
916 union sctp_addr *addr;
919 /* It is safe to write port space in caller. */
921 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
922 af = sctp_get_af_specific(addr->sa.sa_family);
925 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
928 if (addrw->state == SCTP_ADDR_NEW)
929 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
931 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
934 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
937 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
940 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
941 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
944 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
945 * Section 3.1.2 for this usage.
947 * addrs is a pointer to an array of one or more socket addresses. Each
948 * address is contained in its appropriate structure (i.e. struct
949 * sockaddr_in or struct sockaddr_in6) the family of the address type
950 * must be used to distinguish the address length (note that this
951 * representation is termed a "packed array" of addresses). The caller
952 * specifies the number of addresses in the array with addrcnt.
954 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
955 * -1, and sets errno to the appropriate error code.
957 * For SCTP, the port given in each socket address must be the same, or
958 * sctp_bindx() will fail, setting errno to EINVAL.
960 * The flags parameter is formed from the bitwise OR of zero or more of
961 * the following currently defined flags:
963 * SCTP_BINDX_ADD_ADDR
965 * SCTP_BINDX_REM_ADDR
967 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
968 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
969 * addresses from the association. The two flags are mutually exclusive;
970 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
971 * not remove all addresses from an association; sctp_bindx() will
972 * reject such an attempt with EINVAL.
974 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
975 * additional addresses with an endpoint after calling bind(). Or use
976 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
977 * socket is associated with so that no new association accepted will be
978 * associated with those addresses. If the endpoint supports dynamic
979 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
980 * endpoint to send the appropriate message to the peer to change the
981 * peers address lists.
983 * Adding and removing addresses from a connected association is
984 * optional functionality. Implementations that do not support this
985 * functionality should return EOPNOTSUPP.
987 * Basically do nothing but copying the addresses from user to kernel
988 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
989 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
992 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
995 * sk The sk of the socket
996 * addrs The pointer to the addresses in user land
997 * addrssize Size of the addrs buffer
998 * op Operation to perform (add or remove, see the flags of
1001 * Returns 0 if ok, <0 errno code on error.
1003 static int sctp_setsockopt_bindx(struct sock *sk,
1004 struct sockaddr __user *addrs,
1005 int addrs_size, int op)
1007 struct sockaddr *kaddrs;
1011 struct sockaddr *sa_addr;
1015 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
1016 __func__, sk, addrs, addrs_size, op);
1018 if (unlikely(addrs_size <= 0))
1021 kaddrs = vmemdup_user(addrs, addrs_size);
1022 if (unlikely(IS_ERR(kaddrs)))
1023 return PTR_ERR(kaddrs);
1025 /* Walk through the addrs buffer and count the number of addresses. */
1027 while (walk_size < addrs_size) {
1028 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1034 af = sctp_get_af_specific(sa_addr->sa_family);
1036 /* If the address family is not supported or if this address
1037 * causes the address buffer to overflow return EINVAL.
1039 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1044 addr_buf += af->sockaddr_len;
1045 walk_size += af->sockaddr_len;
1050 case SCTP_BINDX_ADD_ADDR:
1051 /* Allow security module to validate bindx addresses. */
1052 err = security_sctp_bind_connect(sk, SCTP_SOCKOPT_BINDX_ADD,
1053 (struct sockaddr *)kaddrs,
1057 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1060 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1063 case SCTP_BINDX_REM_ADDR:
1064 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1067 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1081 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1083 * Common routine for handling connect() and sctp_connectx().
1084 * Connect will come in with just a single address.
1086 static int __sctp_connect(struct sock *sk,
1087 struct sockaddr *kaddrs,
1088 int addrs_size, int flags,
1089 sctp_assoc_t *assoc_id)
1091 struct net *net = sock_net(sk);
1092 struct sctp_sock *sp;
1093 struct sctp_endpoint *ep;
1094 struct sctp_association *asoc = NULL;
1095 struct sctp_association *asoc2;
1096 struct sctp_transport *transport;
1098 enum sctp_scope scope;
1103 union sctp_addr *sa_addr = NULL;
1105 unsigned short port;
1110 /* connect() cannot be done on a socket that is already in ESTABLISHED
1111 * state - UDP-style peeled off socket or a TCP-style socket that
1112 * is already connected.
1113 * It cannot be done even on a TCP-style listening socket.
1115 if (sctp_sstate(sk, ESTABLISHED) || sctp_sstate(sk, CLOSING) ||
1116 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1121 /* Walk through the addrs buffer and count the number of addresses. */
1123 while (walk_size < addrs_size) {
1126 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1132 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1134 /* If the address family is not supported or if this address
1135 * causes the address buffer to overflow return EINVAL.
1137 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1142 port = ntohs(sa_addr->v4.sin_port);
1144 /* Save current address so we can work with it */
1145 memcpy(&to, sa_addr, af->sockaddr_len);
1147 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1151 /* Make sure the destination port is correctly set
1154 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1159 /* Check if there already is a matching association on the
1160 * endpoint (other than the one created here).
1162 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1163 if (asoc2 && asoc2 != asoc) {
1164 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1171 /* If we could not find a matching association on the endpoint,
1172 * make sure that there is no peeled-off association matching
1173 * the peer address even on another socket.
1175 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1176 err = -EADDRNOTAVAIL;
1181 /* If a bind() or sctp_bindx() is not called prior to
1182 * an sctp_connectx() call, the system picks an
1183 * ephemeral port and will choose an address set
1184 * equivalent to binding with a wildcard address.
1186 if (!ep->base.bind_addr.port) {
1187 if (sctp_autobind(sk)) {
1193 * If an unprivileged user inherits a 1-many
1194 * style socket with open associations on a
1195 * privileged port, it MAY be permitted to
1196 * accept new associations, but it SHOULD NOT
1197 * be permitted to open new associations.
1199 if (ep->base.bind_addr.port <
1200 inet_prot_sock(net) &&
1201 !ns_capable(net->user_ns,
1202 CAP_NET_BIND_SERVICE)) {
1208 scope = sctp_scope(&to);
1209 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1215 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1223 /* Prime the peer's transport structures. */
1224 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1232 addr_buf += af->sockaddr_len;
1233 walk_size += af->sockaddr_len;
1236 /* In case the user of sctp_connectx() wants an association
1237 * id back, assign one now.
1240 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1245 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1250 /* Initialize sk's dport and daddr for getpeername() */
1251 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1252 sp->pf->to_sk_daddr(sa_addr, sk);
1255 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
1258 *assoc_id = asoc->assoc_id;
1260 err = sctp_wait_for_connect(asoc, &timeo);
1261 /* Note: the asoc may be freed after the return of
1262 * sctp_wait_for_connect.
1265 /* Don't free association on exit. */
1269 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1270 __func__, asoc, kaddrs, err);
1273 /* sctp_primitive_ASSOCIATE may have added this association
1274 * To the hash table, try to unhash it, just in case, its a noop
1275 * if it wasn't hashed so we're safe
1277 sctp_association_free(asoc);
1282 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1285 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1286 * sctp_assoc_t *asoc);
1288 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1289 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1290 * or IPv6 addresses.
1292 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1293 * Section 3.1.2 for this usage.
1295 * addrs is a pointer to an array of one or more socket addresses. Each
1296 * address is contained in its appropriate structure (i.e. struct
1297 * sockaddr_in or struct sockaddr_in6) the family of the address type
1298 * must be used to distengish the address length (note that this
1299 * representation is termed a "packed array" of addresses). The caller
1300 * specifies the number of addresses in the array with addrcnt.
1302 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1303 * the association id of the new association. On failure, sctp_connectx()
1304 * returns -1, and sets errno to the appropriate error code. The assoc_id
1305 * is not touched by the kernel.
1307 * For SCTP, the port given in each socket address must be the same, or
1308 * sctp_connectx() will fail, setting errno to EINVAL.
1310 * An application can use sctp_connectx to initiate an association with
1311 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1312 * allows a caller to specify multiple addresses at which a peer can be
1313 * reached. The way the SCTP stack uses the list of addresses to set up
1314 * the association is implementation dependent. This function only
1315 * specifies that the stack will try to make use of all the addresses in
1316 * the list when needed.
1318 * Note that the list of addresses passed in is only used for setting up
1319 * the association. It does not necessarily equal the set of addresses
1320 * the peer uses for the resulting association. If the caller wants to
1321 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1322 * retrieve them after the association has been set up.
1324 * Basically do nothing but copying the addresses from user to kernel
1325 * land and invoking either sctp_connectx(). This is used for tunneling
1326 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1328 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1331 * sk The sk of the socket
1332 * addrs The pointer to the addresses in user land
1333 * addrssize Size of the addrs buffer
1335 * Returns >=0 if ok, <0 errno code on error.
1337 static int __sctp_setsockopt_connectx(struct sock *sk,
1338 struct sockaddr __user *addrs,
1340 sctp_assoc_t *assoc_id)
1342 struct sockaddr *kaddrs;
1343 int err = 0, flags = 0;
1345 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1346 __func__, sk, addrs, addrs_size);
1348 if (unlikely(addrs_size <= 0))
1351 kaddrs = vmemdup_user(addrs, addrs_size);
1352 if (unlikely(IS_ERR(kaddrs)))
1353 return PTR_ERR(kaddrs);
1355 /* Allow security module to validate connectx addresses. */
1356 err = security_sctp_bind_connect(sk, SCTP_SOCKOPT_CONNECTX,
1357 (struct sockaddr *)kaddrs,
1362 /* in-kernel sockets don't generally have a file allocated to them
1363 * if all they do is call sock_create_kern().
1365 if (sk->sk_socket->file)
1366 flags = sk->sk_socket->file->f_flags;
1368 err = __sctp_connect(sk, kaddrs, addrs_size, flags, assoc_id);
1377 * This is an older interface. It's kept for backward compatibility
1378 * to the option that doesn't provide association id.
1380 static int sctp_setsockopt_connectx_old(struct sock *sk,
1381 struct sockaddr __user *addrs,
1384 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1388 * New interface for the API. The since the API is done with a socket
1389 * option, to make it simple we feed back the association id is as a return
1390 * indication to the call. Error is always negative and association id is
1393 static int sctp_setsockopt_connectx(struct sock *sk,
1394 struct sockaddr __user *addrs,
1397 sctp_assoc_t assoc_id = 0;
1400 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1409 * New (hopefully final) interface for the API.
1410 * We use the sctp_getaddrs_old structure so that use-space library
1411 * can avoid any unnecessary allocations. The only different part
1412 * is that we store the actual length of the address buffer into the
1413 * addrs_num structure member. That way we can re-use the existing
1416 #ifdef CONFIG_COMPAT
1417 struct compat_sctp_getaddrs_old {
1418 sctp_assoc_t assoc_id;
1420 compat_uptr_t addrs; /* struct sockaddr * */
1424 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1425 char __user *optval,
1428 struct sctp_getaddrs_old param;
1429 sctp_assoc_t assoc_id = 0;
1432 #ifdef CONFIG_COMPAT
1433 if (in_compat_syscall()) {
1434 struct compat_sctp_getaddrs_old param32;
1436 if (len < sizeof(param32))
1438 if (copy_from_user(¶m32, optval, sizeof(param32)))
1441 param.assoc_id = param32.assoc_id;
1442 param.addr_num = param32.addr_num;
1443 param.addrs = compat_ptr(param32.addrs);
1447 if (len < sizeof(param))
1449 if (copy_from_user(¶m, optval, sizeof(param)))
1453 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1454 param.addrs, param.addr_num,
1456 if (err == 0 || err == -EINPROGRESS) {
1457 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1459 if (put_user(sizeof(assoc_id), optlen))
1466 /* API 3.1.4 close() - UDP Style Syntax
1467 * Applications use close() to perform graceful shutdown (as described in
1468 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1469 * by a UDP-style socket.
1473 * ret = close(int sd);
1475 * sd - the socket descriptor of the associations to be closed.
1477 * To gracefully shutdown a specific association represented by the
1478 * UDP-style socket, an application should use the sendmsg() call,
1479 * passing no user data, but including the appropriate flag in the
1480 * ancillary data (see Section xxxx).
1482 * If sd in the close() call is a branched-off socket representing only
1483 * one association, the shutdown is performed on that association only.
1485 * 4.1.6 close() - TCP Style Syntax
1487 * Applications use close() to gracefully close down an association.
1491 * int close(int sd);
1493 * sd - the socket descriptor of the association to be closed.
1495 * After an application calls close() on a socket descriptor, no further
1496 * socket operations will succeed on that descriptor.
1498 * API 7.1.4 SO_LINGER
1500 * An application using the TCP-style socket can use this option to
1501 * perform the SCTP ABORT primitive. The linger option structure is:
1504 * int l_onoff; // option on/off
1505 * int l_linger; // linger time
1508 * To enable the option, set l_onoff to 1. If the l_linger value is set
1509 * to 0, calling close() is the same as the ABORT primitive. If the
1510 * value is set to a negative value, the setsockopt() call will return
1511 * an error. If the value is set to a positive value linger_time, the
1512 * close() can be blocked for at most linger_time ms. If the graceful
1513 * shutdown phase does not finish during this period, close() will
1514 * return but the graceful shutdown phase continues in the system.
1516 static void sctp_close(struct sock *sk, long timeout)
1518 struct net *net = sock_net(sk);
1519 struct sctp_endpoint *ep;
1520 struct sctp_association *asoc;
1521 struct list_head *pos, *temp;
1522 unsigned int data_was_unread;
1524 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1526 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
1527 sk->sk_shutdown = SHUTDOWN_MASK;
1528 inet_sk_set_state(sk, SCTP_SS_CLOSING);
1530 ep = sctp_sk(sk)->ep;
1532 /* Clean up any skbs sitting on the receive queue. */
1533 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1534 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1536 /* Walk all associations on an endpoint. */
1537 list_for_each_safe(pos, temp, &ep->asocs) {
1538 asoc = list_entry(pos, struct sctp_association, asocs);
1540 if (sctp_style(sk, TCP)) {
1541 /* A closed association can still be in the list if
1542 * it belongs to a TCP-style listening socket that is
1543 * not yet accepted. If so, free it. If not, send an
1544 * ABORT or SHUTDOWN based on the linger options.
1546 if (sctp_state(asoc, CLOSED)) {
1547 sctp_association_free(asoc);
1552 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1553 !skb_queue_empty(&asoc->ulpq.reasm) ||
1554 !skb_queue_empty(&asoc->ulpq.reasm_uo) ||
1555 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1556 struct sctp_chunk *chunk;
1558 chunk = sctp_make_abort_user(asoc, NULL, 0);
1559 sctp_primitive_ABORT(net, asoc, chunk);
1561 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1564 /* On a TCP-style socket, block for at most linger_time if set. */
1565 if (sctp_style(sk, TCP) && timeout)
1566 sctp_wait_for_close(sk, timeout);
1568 /* This will run the backlog queue. */
1571 /* Supposedly, no process has access to the socket, but
1572 * the net layers still may.
1573 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1574 * held and that should be grabbed before socket lock.
1576 spin_lock_bh(&net->sctp.addr_wq_lock);
1577 bh_lock_sock_nested(sk);
1579 /* Hold the sock, since sk_common_release() will put sock_put()
1580 * and we have just a little more cleanup.
1583 sk_common_release(sk);
1586 spin_unlock_bh(&net->sctp.addr_wq_lock);
1590 SCTP_DBG_OBJCNT_DEC(sock);
1593 /* Handle EPIPE error. */
1594 static int sctp_error(struct sock *sk, int flags, int err)
1597 err = sock_error(sk) ? : -EPIPE;
1598 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1599 send_sig(SIGPIPE, current, 0);
1603 /* API 3.1.3 sendmsg() - UDP Style Syntax
1605 * An application uses sendmsg() and recvmsg() calls to transmit data to
1606 * and receive data from its peer.
1608 * ssize_t sendmsg(int socket, const struct msghdr *message,
1611 * socket - the socket descriptor of the endpoint.
1612 * message - pointer to the msghdr structure which contains a single
1613 * user message and possibly some ancillary data.
1615 * See Section 5 for complete description of the data
1618 * flags - flags sent or received with the user message, see Section
1619 * 5 for complete description of the flags.
1621 * Note: This function could use a rewrite especially when explicit
1622 * connect support comes in.
1624 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1626 static int sctp_msghdr_parse(const struct msghdr *msg,
1627 struct sctp_cmsgs *cmsgs);
1629 static int sctp_sendmsg_parse(struct sock *sk, struct sctp_cmsgs *cmsgs,
1630 struct sctp_sndrcvinfo *srinfo,
1631 const struct msghdr *msg, size_t msg_len)
1636 if (sctp_sstate(sk, LISTENING) && sctp_style(sk, TCP))
1639 if (msg_len > sk->sk_sndbuf)
1642 memset(cmsgs, 0, sizeof(*cmsgs));
1643 err = sctp_msghdr_parse(msg, cmsgs);
1645 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1649 memset(srinfo, 0, sizeof(*srinfo));
1650 if (cmsgs->srinfo) {
1651 srinfo->sinfo_stream = cmsgs->srinfo->sinfo_stream;
1652 srinfo->sinfo_flags = cmsgs->srinfo->sinfo_flags;
1653 srinfo->sinfo_ppid = cmsgs->srinfo->sinfo_ppid;
1654 srinfo->sinfo_context = cmsgs->srinfo->sinfo_context;
1655 srinfo->sinfo_assoc_id = cmsgs->srinfo->sinfo_assoc_id;
1656 srinfo->sinfo_timetolive = cmsgs->srinfo->sinfo_timetolive;
1660 srinfo->sinfo_stream = cmsgs->sinfo->snd_sid;
1661 srinfo->sinfo_flags = cmsgs->sinfo->snd_flags;
1662 srinfo->sinfo_ppid = cmsgs->sinfo->snd_ppid;
1663 srinfo->sinfo_context = cmsgs->sinfo->snd_context;
1664 srinfo->sinfo_assoc_id = cmsgs->sinfo->snd_assoc_id;
1667 if (cmsgs->prinfo) {
1668 srinfo->sinfo_timetolive = cmsgs->prinfo->pr_value;
1669 SCTP_PR_SET_POLICY(srinfo->sinfo_flags,
1670 cmsgs->prinfo->pr_policy);
1673 sflags = srinfo->sinfo_flags;
1674 if (!sflags && msg_len)
1677 if (sctp_style(sk, TCP) && (sflags & (SCTP_EOF | SCTP_ABORT)))
1680 if (((sflags & SCTP_EOF) && msg_len > 0) ||
1681 (!(sflags & (SCTP_EOF | SCTP_ABORT)) && msg_len == 0))
1684 if ((sflags & SCTP_ADDR_OVER) && !msg->msg_name)
1690 static int sctp_sendmsg_new_asoc(struct sock *sk, __u16 sflags,
1691 struct sctp_cmsgs *cmsgs,
1692 union sctp_addr *daddr,
1693 struct sctp_transport **tp)
1695 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
1696 struct net *net = sock_net(sk);
1697 struct sctp_association *asoc;
1698 enum sctp_scope scope;
1699 struct cmsghdr *cmsg;
1700 __be32 flowinfo = 0;
1706 if (sflags & (SCTP_EOF | SCTP_ABORT))
1709 if (sctp_style(sk, TCP) && (sctp_sstate(sk, ESTABLISHED) ||
1710 sctp_sstate(sk, CLOSING)))
1711 return -EADDRNOTAVAIL;
1713 if (sctp_endpoint_is_peeled_off(ep, daddr))
1714 return -EADDRNOTAVAIL;
1716 if (!ep->base.bind_addr.port) {
1717 if (sctp_autobind(sk))
1720 if (ep->base.bind_addr.port < inet_prot_sock(net) &&
1721 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
1725 scope = sctp_scope(daddr);
1727 /* Label connection socket for first association 1-to-many
1728 * style for client sequence socket()->sendmsg(). This
1729 * needs to be done before sctp_assoc_add_peer() as that will
1730 * set up the initial packet that needs to account for any
1731 * security ip options (CIPSO/CALIPSO) added to the packet.
1733 af = sctp_get_af_specific(daddr->sa.sa_family);
1736 err = security_sctp_bind_connect(sk, SCTP_SENDMSG_CONNECT,
1737 (struct sockaddr *)daddr,
1742 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1746 if (sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL) < 0) {
1752 struct sctp_initmsg *init = cmsgs->init;
1754 if (init->sinit_num_ostreams) {
1755 __u16 outcnt = init->sinit_num_ostreams;
1757 asoc->c.sinit_num_ostreams = outcnt;
1758 /* outcnt has been changed, need to re-init stream */
1759 err = sctp_stream_init(&asoc->stream, outcnt, 0,
1765 if (init->sinit_max_instreams)
1766 asoc->c.sinit_max_instreams = init->sinit_max_instreams;
1768 if (init->sinit_max_attempts)
1769 asoc->max_init_attempts = init->sinit_max_attempts;
1771 if (init->sinit_max_init_timeo)
1772 asoc->max_init_timeo =
1773 msecs_to_jiffies(init->sinit_max_init_timeo);
1776 *tp = sctp_assoc_add_peer(asoc, daddr, GFP_KERNEL, SCTP_UNKNOWN);
1782 if (!cmsgs->addrs_msg)
1785 if (daddr->sa.sa_family == AF_INET6)
1786 flowinfo = daddr->v6.sin6_flowinfo;
1788 /* sendv addr list parse */
1789 for_each_cmsghdr(cmsg, cmsgs->addrs_msg) {
1790 struct sctp_transport *transport;
1791 struct sctp_association *old;
1792 union sctp_addr _daddr;
1795 if (cmsg->cmsg_level != IPPROTO_SCTP ||
1796 (cmsg->cmsg_type != SCTP_DSTADDRV4 &&
1797 cmsg->cmsg_type != SCTP_DSTADDRV6))
1801 memset(daddr, 0, sizeof(*daddr));
1802 dlen = cmsg->cmsg_len - sizeof(struct cmsghdr);
1803 if (cmsg->cmsg_type == SCTP_DSTADDRV4) {
1804 if (dlen < sizeof(struct in_addr)) {
1809 dlen = sizeof(struct in_addr);
1810 daddr->v4.sin_family = AF_INET;
1811 daddr->v4.sin_port = htons(asoc->peer.port);
1812 memcpy(&daddr->v4.sin_addr, CMSG_DATA(cmsg), dlen);
1814 if (dlen < sizeof(struct in6_addr)) {
1819 dlen = sizeof(struct in6_addr);
1820 daddr->v6.sin6_flowinfo = flowinfo;
1821 daddr->v6.sin6_family = AF_INET6;
1822 daddr->v6.sin6_port = htons(asoc->peer.port);
1823 memcpy(&daddr->v6.sin6_addr, CMSG_DATA(cmsg), dlen);
1825 err = sctp_verify_addr(sk, daddr, sizeof(*daddr));
1829 old = sctp_endpoint_lookup_assoc(ep, daddr, &transport);
1830 if (old && old != asoc) {
1831 if (old->state >= SCTP_STATE_ESTABLISHED)
1838 if (sctp_endpoint_is_peeled_off(ep, daddr)) {
1839 err = -EADDRNOTAVAIL;
1843 transport = sctp_assoc_add_peer(asoc, daddr, GFP_KERNEL,
1854 sctp_association_free(asoc);
1858 static int sctp_sendmsg_check_sflags(struct sctp_association *asoc,
1859 __u16 sflags, struct msghdr *msg,
1862 struct sock *sk = asoc->base.sk;
1863 struct net *net = sock_net(sk);
1865 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP))
1868 if ((sflags & SCTP_SENDALL) && sctp_style(sk, UDP) &&
1869 !sctp_state(asoc, ESTABLISHED))
1872 if (sflags & SCTP_EOF) {
1873 pr_debug("%s: shutting down association:%p\n", __func__, asoc);
1874 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1879 if (sflags & SCTP_ABORT) {
1880 struct sctp_chunk *chunk;
1882 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1886 pr_debug("%s: aborting association:%p\n", __func__, asoc);
1887 sctp_primitive_ABORT(net, asoc, chunk);
1895 static int sctp_sendmsg_to_asoc(struct sctp_association *asoc,
1896 struct msghdr *msg, size_t msg_len,
1897 struct sctp_transport *transport,
1898 struct sctp_sndrcvinfo *sinfo)
1900 struct sock *sk = asoc->base.sk;
1901 struct sctp_sock *sp = sctp_sk(sk);
1902 struct net *net = sock_net(sk);
1903 struct sctp_datamsg *datamsg;
1904 bool wait_connect = false;
1905 struct sctp_chunk *chunk;
1909 if (sinfo->sinfo_stream >= asoc->stream.outcnt) {
1914 if (unlikely(!SCTP_SO(&asoc->stream, sinfo->sinfo_stream)->ext)) {
1915 err = sctp_stream_init_ext(&asoc->stream, sinfo->sinfo_stream);
1920 if (sp->disable_fragments && msg_len > asoc->frag_point) {
1925 if (asoc->pmtu_pending) {
1926 if (sp->param_flags & SPP_PMTUD_ENABLE)
1927 sctp_assoc_sync_pmtu(asoc);
1928 asoc->pmtu_pending = 0;
1931 if (sctp_wspace(asoc) < msg_len)
1932 sctp_prsctp_prune(asoc, sinfo, msg_len - sctp_wspace(asoc));
1934 if (!sctp_wspace(asoc)) {
1935 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1936 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1941 if (sctp_state(asoc, CLOSED)) {
1942 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1946 if (sp->strm_interleave) {
1947 timeo = sock_sndtimeo(sk, 0);
1948 err = sctp_wait_for_connect(asoc, &timeo);
1952 wait_connect = true;
1955 pr_debug("%s: we associated primitively\n", __func__);
1958 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
1959 if (IS_ERR(datamsg)) {
1960 err = PTR_ERR(datamsg);
1964 asoc->force_delay = !!(msg->msg_flags & MSG_MORE);
1966 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1967 sctp_chunk_hold(chunk);
1968 sctp_set_owner_w(chunk);
1969 chunk->transport = transport;
1972 err = sctp_primitive_SEND(net, asoc, datamsg);
1974 sctp_datamsg_free(datamsg);
1978 pr_debug("%s: we sent primitively\n", __func__);
1980 sctp_datamsg_put(datamsg);
1982 if (unlikely(wait_connect)) {
1983 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1984 sctp_wait_for_connect(asoc, &timeo);
1993 static union sctp_addr *sctp_sendmsg_get_daddr(struct sock *sk,
1994 const struct msghdr *msg,
1995 struct sctp_cmsgs *cmsgs)
1997 union sctp_addr *daddr = NULL;
2000 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
2001 int len = msg->msg_namelen;
2003 if (len > sizeof(*daddr))
2004 len = sizeof(*daddr);
2006 daddr = (union sctp_addr *)msg->msg_name;
2008 err = sctp_verify_addr(sk, daddr, len);
2010 return ERR_PTR(err);
2016 static void sctp_sendmsg_update_sinfo(struct sctp_association *asoc,
2017 struct sctp_sndrcvinfo *sinfo,
2018 struct sctp_cmsgs *cmsgs)
2020 if (!cmsgs->srinfo && !cmsgs->sinfo) {
2021 sinfo->sinfo_stream = asoc->default_stream;
2022 sinfo->sinfo_ppid = asoc->default_ppid;
2023 sinfo->sinfo_context = asoc->default_context;
2024 sinfo->sinfo_assoc_id = sctp_assoc2id(asoc);
2027 sinfo->sinfo_flags = asoc->default_flags;
2030 if (!cmsgs->srinfo && !cmsgs->prinfo)
2031 sinfo->sinfo_timetolive = asoc->default_timetolive;
2033 if (cmsgs->authinfo) {
2034 /* Reuse sinfo_tsn to indicate that authinfo was set and
2035 * sinfo_ssn to save the keyid on tx path.
2037 sinfo->sinfo_tsn = 1;
2038 sinfo->sinfo_ssn = cmsgs->authinfo->auth_keynumber;
2042 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
2044 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
2045 struct sctp_transport *transport = NULL;
2046 struct sctp_sndrcvinfo _sinfo, *sinfo;
2047 struct sctp_association *asoc;
2048 struct sctp_cmsgs cmsgs;
2049 union sctp_addr *daddr;
2054 /* Parse and get snd_info */
2055 err = sctp_sendmsg_parse(sk, &cmsgs, &_sinfo, msg, msg_len);
2060 sflags = sinfo->sinfo_flags;
2062 /* Get daddr from msg */
2063 daddr = sctp_sendmsg_get_daddr(sk, msg, &cmsgs);
2064 if (IS_ERR(daddr)) {
2065 err = PTR_ERR(daddr);
2071 /* SCTP_SENDALL process */
2072 if ((sflags & SCTP_SENDALL) && sctp_style(sk, UDP)) {
2073 list_for_each_entry(asoc, &ep->asocs, asocs) {
2074 err = sctp_sendmsg_check_sflags(asoc, sflags, msg,
2081 sctp_sendmsg_update_sinfo(asoc, sinfo, &cmsgs);
2083 err = sctp_sendmsg_to_asoc(asoc, msg, msg_len,
2088 iov_iter_revert(&msg->msg_iter, err);
2094 /* Get and check or create asoc */
2096 asoc = sctp_endpoint_lookup_assoc(ep, daddr, &transport);
2098 err = sctp_sendmsg_check_sflags(asoc, sflags, msg,
2103 err = sctp_sendmsg_new_asoc(sk, sflags, &cmsgs, daddr,
2108 asoc = transport->asoc;
2112 if (!sctp_style(sk, TCP) && !(sflags & SCTP_ADDR_OVER))
2115 asoc = sctp_id2assoc(sk, sinfo->sinfo_assoc_id);
2121 err = sctp_sendmsg_check_sflags(asoc, sflags, msg, msg_len);
2126 /* Update snd_info with the asoc */
2127 sctp_sendmsg_update_sinfo(asoc, sinfo, &cmsgs);
2129 /* Send msg to the asoc */
2130 err = sctp_sendmsg_to_asoc(asoc, msg, msg_len, transport, sinfo);
2131 if (err < 0 && err != -ESRCH && new)
2132 sctp_association_free(asoc);
2137 return sctp_error(sk, msg->msg_flags, err);
2140 /* This is an extended version of skb_pull() that removes the data from the
2141 * start of a skb even when data is spread across the list of skb's in the
2142 * frag_list. len specifies the total amount of data that needs to be removed.
2143 * when 'len' bytes could be removed from the skb, it returns 0.
2144 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2145 * could not be removed.
2147 static int sctp_skb_pull(struct sk_buff *skb, int len)
2149 struct sk_buff *list;
2150 int skb_len = skb_headlen(skb);
2153 if (len <= skb_len) {
2154 __skb_pull(skb, len);
2158 __skb_pull(skb, skb_len);
2160 skb_walk_frags(skb, list) {
2161 rlen = sctp_skb_pull(list, len);
2162 skb->len -= (len-rlen);
2163 skb->data_len -= (len-rlen);
2174 /* API 3.1.3 recvmsg() - UDP Style Syntax
2176 * ssize_t recvmsg(int socket, struct msghdr *message,
2179 * socket - the socket descriptor of the endpoint.
2180 * message - pointer to the msghdr structure which contains a single
2181 * user message and possibly some ancillary data.
2183 * See Section 5 for complete description of the data
2186 * flags - flags sent or received with the user message, see Section
2187 * 5 for complete description of the flags.
2189 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2190 int noblock, int flags, int *addr_len)
2192 struct sctp_ulpevent *event = NULL;
2193 struct sctp_sock *sp = sctp_sk(sk);
2194 struct sk_buff *skb, *head_skb;
2199 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2200 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2205 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED) &&
2206 !sctp_sstate(sk, CLOSING) && !sctp_sstate(sk, CLOSED)) {
2211 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2215 /* Get the total length of the skb including any skb's in the
2224 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2226 event = sctp_skb2event(skb);
2231 if (event->chunk && event->chunk->head_skb)
2232 head_skb = event->chunk->head_skb;
2235 sock_recv_ts_and_drops(msg, sk, head_skb);
2236 if (sctp_ulpevent_is_notification(event)) {
2237 msg->msg_flags |= MSG_NOTIFICATION;
2238 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2240 sp->pf->skb_msgname(head_skb, msg->msg_name, addr_len);
2243 /* Check if we allow SCTP_NXTINFO. */
2244 if (sp->recvnxtinfo)
2245 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2246 /* Check if we allow SCTP_RCVINFO. */
2247 if (sp->recvrcvinfo)
2248 sctp_ulpevent_read_rcvinfo(event, msg);
2249 /* Check if we allow SCTP_SNDRCVINFO. */
2250 if (sp->subscribe.sctp_data_io_event)
2251 sctp_ulpevent_read_sndrcvinfo(event, msg);
2255 /* If skb's length exceeds the user's buffer, update the skb and
2256 * push it back to the receive_queue so that the next call to
2257 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2259 if (skb_len > copied) {
2260 msg->msg_flags &= ~MSG_EOR;
2261 if (flags & MSG_PEEK)
2263 sctp_skb_pull(skb, copied);
2264 skb_queue_head(&sk->sk_receive_queue, skb);
2266 /* When only partial message is copied to the user, increase
2267 * rwnd by that amount. If all the data in the skb is read,
2268 * rwnd is updated when the event is freed.
2270 if (!sctp_ulpevent_is_notification(event))
2271 sctp_assoc_rwnd_increase(event->asoc, copied);
2273 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2274 (event->msg_flags & MSG_EOR))
2275 msg->msg_flags |= MSG_EOR;
2277 msg->msg_flags &= ~MSG_EOR;
2280 if (flags & MSG_PEEK) {
2281 /* Release the skb reference acquired after peeking the skb in
2282 * sctp_skb_recv_datagram().
2286 /* Free the event which includes releasing the reference to
2287 * the owner of the skb, freeing the skb and updating the
2290 sctp_ulpevent_free(event);
2297 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2299 * This option is a on/off flag. If enabled no SCTP message
2300 * fragmentation will be performed. Instead if a message being sent
2301 * exceeds the current PMTU size, the message will NOT be sent and
2302 * instead a error will be indicated to the user.
2304 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2305 char __user *optval,
2306 unsigned int optlen)
2310 if (optlen < sizeof(int))
2313 if (get_user(val, (int __user *)optval))
2316 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2321 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2322 unsigned int optlen)
2324 struct sctp_association *asoc;
2325 struct sctp_ulpevent *event;
2327 if (optlen > sizeof(struct sctp_event_subscribe))
2329 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2332 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2333 * if there is no data to be sent or retransmit, the stack will
2334 * immediately send up this notification.
2336 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2337 &sctp_sk(sk)->subscribe)) {
2338 asoc = sctp_id2assoc(sk, 0);
2340 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2341 event = sctp_ulpevent_make_sender_dry_event(asoc,
2342 GFP_USER | __GFP_NOWARN);
2346 asoc->stream.si->enqueue_event(&asoc->ulpq, event);
2353 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2355 * This socket option is applicable to the UDP-style socket only. When
2356 * set it will cause associations that are idle for more than the
2357 * specified number of seconds to automatically close. An association
2358 * being idle is defined an association that has NOT sent or received
2359 * user data. The special value of '0' indicates that no automatic
2360 * close of any associations should be performed. The option expects an
2361 * integer defining the number of seconds of idle time before an
2362 * association is closed.
2364 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2365 unsigned int optlen)
2367 struct sctp_sock *sp = sctp_sk(sk);
2368 struct net *net = sock_net(sk);
2370 /* Applicable to UDP-style socket only */
2371 if (sctp_style(sk, TCP))
2373 if (optlen != sizeof(int))
2375 if (copy_from_user(&sp->autoclose, optval, optlen))
2378 if (sp->autoclose > net->sctp.max_autoclose)
2379 sp->autoclose = net->sctp.max_autoclose;
2384 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2386 * Applications can enable or disable heartbeats for any peer address of
2387 * an association, modify an address's heartbeat interval, force a
2388 * heartbeat to be sent immediately, and adjust the address's maximum
2389 * number of retransmissions sent before an address is considered
2390 * unreachable. The following structure is used to access and modify an
2391 * address's parameters:
2393 * struct sctp_paddrparams {
2394 * sctp_assoc_t spp_assoc_id;
2395 * struct sockaddr_storage spp_address;
2396 * uint32_t spp_hbinterval;
2397 * uint16_t spp_pathmaxrxt;
2398 * uint32_t spp_pathmtu;
2399 * uint32_t spp_sackdelay;
2400 * uint32_t spp_flags;
2401 * uint32_t spp_ipv6_flowlabel;
2405 * spp_assoc_id - (one-to-many style socket) This is filled in the
2406 * application, and identifies the association for
2408 * spp_address - This specifies which address is of interest.
2409 * spp_hbinterval - This contains the value of the heartbeat interval,
2410 * in milliseconds. If a value of zero
2411 * is present in this field then no changes are to
2412 * be made to this parameter.
2413 * spp_pathmaxrxt - This contains the maximum number of
2414 * retransmissions before this address shall be
2415 * considered unreachable. If a value of zero
2416 * is present in this field then no changes are to
2417 * be made to this parameter.
2418 * spp_pathmtu - When Path MTU discovery is disabled the value
2419 * specified here will be the "fixed" path mtu.
2420 * Note that if the spp_address field is empty
2421 * then all associations on this address will
2422 * have this fixed path mtu set upon them.
2424 * spp_sackdelay - When delayed sack is enabled, this value specifies
2425 * the number of milliseconds that sacks will be delayed
2426 * for. This value will apply to all addresses of an
2427 * association if the spp_address field is empty. Note
2428 * also, that if delayed sack is enabled and this
2429 * value is set to 0, no change is made to the last
2430 * recorded delayed sack timer value.
2432 * spp_flags - These flags are used to control various features
2433 * on an association. The flag field may contain
2434 * zero or more of the following options.
2436 * SPP_HB_ENABLE - Enable heartbeats on the
2437 * specified address. Note that if the address
2438 * field is empty all addresses for the association
2439 * have heartbeats enabled upon them.
2441 * SPP_HB_DISABLE - Disable heartbeats on the
2442 * speicifed address. Note that if the address
2443 * field is empty all addresses for the association
2444 * will have their heartbeats disabled. Note also
2445 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2446 * mutually exclusive, only one of these two should
2447 * be specified. Enabling both fields will have
2448 * undetermined results.
2450 * SPP_HB_DEMAND - Request a user initiated heartbeat
2451 * to be made immediately.
2453 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2454 * heartbeat delayis to be set to the value of 0
2457 * SPP_PMTUD_ENABLE - This field will enable PMTU
2458 * discovery upon the specified address. Note that
2459 * if the address feild is empty then all addresses
2460 * on the association are effected.
2462 * SPP_PMTUD_DISABLE - This field will disable PMTU
2463 * discovery upon the specified address. Note that
2464 * if the address feild is empty then all addresses
2465 * on the association are effected. Not also that
2466 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2467 * exclusive. Enabling both will have undetermined
2470 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2471 * on delayed sack. The time specified in spp_sackdelay
2472 * is used to specify the sack delay for this address. Note
2473 * that if spp_address is empty then all addresses will
2474 * enable delayed sack and take on the sack delay
2475 * value specified in spp_sackdelay.
2476 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2477 * off delayed sack. If the spp_address field is blank then
2478 * delayed sack is disabled for the entire association. Note
2479 * also that this field is mutually exclusive to
2480 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2483 * SPP_IPV6_FLOWLABEL: Setting this flag enables the
2484 * setting of the IPV6 flow label value. The value is
2485 * contained in the spp_ipv6_flowlabel field.
2486 * Upon retrieval, this flag will be set to indicate that
2487 * the spp_ipv6_flowlabel field has a valid value returned.
2488 * If a specific destination address is set (in the
2489 * spp_address field), then the value returned is that of
2490 * the address. If just an association is specified (and
2491 * no address), then the association's default flow label
2492 * is returned. If neither an association nor a destination
2493 * is specified, then the socket's default flow label is
2494 * returned. For non-IPv6 sockets, this flag will be left
2497 * SPP_DSCP: Setting this flag enables the setting of the
2498 * Differentiated Services Code Point (DSCP) value
2499 * associated with either the association or a specific
2500 * address. The value is obtained in the spp_dscp field.
2501 * Upon retrieval, this flag will be set to indicate that
2502 * the spp_dscp field has a valid value returned. If a
2503 * specific destination address is set when called (in the
2504 * spp_address field), then that specific destination
2505 * address's DSCP value is returned. If just an association
2506 * is specified, then the association's default DSCP is
2507 * returned. If neither an association nor a destination is
2508 * specified, then the socket's default DSCP is returned.
2510 * spp_ipv6_flowlabel
2511 * - This field is used in conjunction with the
2512 * SPP_IPV6_FLOWLABEL flag and contains the IPv6 flow label.
2513 * The 20 least significant bits are used for the flow
2514 * label. This setting has precedence over any IPv6-layer
2517 * spp_dscp - This field is used in conjunction with the SPP_DSCP flag
2518 * and contains the DSCP. The 6 most significant bits are
2519 * used for the DSCP. This setting has precedence over any
2520 * IPv4- or IPv6- layer setting.
2522 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2523 struct sctp_transport *trans,
2524 struct sctp_association *asoc,
2525 struct sctp_sock *sp,
2528 int sackdelay_change)
2532 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2533 struct net *net = sock_net(trans->asoc->base.sk);
2535 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2540 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2541 * this field is ignored. Note also that a value of zero indicates
2542 * the current setting should be left unchanged.
2544 if (params->spp_flags & SPP_HB_ENABLE) {
2546 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2547 * set. This lets us use 0 value when this flag
2550 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2551 params->spp_hbinterval = 0;
2553 if (params->spp_hbinterval ||
2554 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2557 msecs_to_jiffies(params->spp_hbinterval);
2560 msecs_to_jiffies(params->spp_hbinterval);
2562 sp->hbinterval = params->spp_hbinterval;
2569 trans->param_flags =
2570 (trans->param_flags & ~SPP_HB) | hb_change;
2573 (asoc->param_flags & ~SPP_HB) | hb_change;
2576 (sp->param_flags & ~SPP_HB) | hb_change;
2580 /* When Path MTU discovery is disabled the value specified here will
2581 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2582 * include the flag SPP_PMTUD_DISABLE for this field to have any
2585 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2587 trans->pathmtu = params->spp_pathmtu;
2588 sctp_assoc_sync_pmtu(asoc);
2590 sctp_assoc_set_pmtu(asoc, params->spp_pathmtu);
2592 sp->pathmtu = params->spp_pathmtu;
2598 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2599 (params->spp_flags & SPP_PMTUD_ENABLE);
2600 trans->param_flags =
2601 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2603 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2604 sctp_assoc_sync_pmtu(asoc);
2608 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2611 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2615 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2616 * value of this field is ignored. Note also that a value of zero
2617 * indicates the current setting should be left unchanged.
2619 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2622 msecs_to_jiffies(params->spp_sackdelay);
2625 msecs_to_jiffies(params->spp_sackdelay);
2627 sp->sackdelay = params->spp_sackdelay;
2631 if (sackdelay_change) {
2633 trans->param_flags =
2634 (trans->param_flags & ~SPP_SACKDELAY) |
2638 (asoc->param_flags & ~SPP_SACKDELAY) |
2642 (sp->param_flags & ~SPP_SACKDELAY) |
2647 /* Note that a value of zero indicates the current setting should be
2650 if (params->spp_pathmaxrxt) {
2652 trans->pathmaxrxt = params->spp_pathmaxrxt;
2654 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2656 sp->pathmaxrxt = params->spp_pathmaxrxt;
2660 if (params->spp_flags & SPP_IPV6_FLOWLABEL) {
2661 if (trans && trans->ipaddr.sa.sa_family == AF_INET6) {
2662 trans->flowlabel = params->spp_ipv6_flowlabel &
2663 SCTP_FLOWLABEL_VAL_MASK;
2664 trans->flowlabel |= SCTP_FLOWLABEL_SET_MASK;
2666 list_for_each_entry(trans,
2667 &asoc->peer.transport_addr_list,
2669 if (trans->ipaddr.sa.sa_family != AF_INET6)
2671 trans->flowlabel = params->spp_ipv6_flowlabel &
2672 SCTP_FLOWLABEL_VAL_MASK;
2673 trans->flowlabel |= SCTP_FLOWLABEL_SET_MASK;
2675 asoc->flowlabel = params->spp_ipv6_flowlabel &
2676 SCTP_FLOWLABEL_VAL_MASK;
2677 asoc->flowlabel |= SCTP_FLOWLABEL_SET_MASK;
2678 } else if (sctp_opt2sk(sp)->sk_family == AF_INET6) {
2679 sp->flowlabel = params->spp_ipv6_flowlabel &
2680 SCTP_FLOWLABEL_VAL_MASK;
2681 sp->flowlabel |= SCTP_FLOWLABEL_SET_MASK;
2685 if (params->spp_flags & SPP_DSCP) {
2687 trans->dscp = params->spp_dscp & SCTP_DSCP_VAL_MASK;
2688 trans->dscp |= SCTP_DSCP_SET_MASK;
2690 list_for_each_entry(trans,
2691 &asoc->peer.transport_addr_list,
2693 trans->dscp = params->spp_dscp &
2695 trans->dscp |= SCTP_DSCP_SET_MASK;
2697 asoc->dscp = params->spp_dscp & SCTP_DSCP_VAL_MASK;
2698 asoc->dscp |= SCTP_DSCP_SET_MASK;
2700 sp->dscp = params->spp_dscp & SCTP_DSCP_VAL_MASK;
2701 sp->dscp |= SCTP_DSCP_SET_MASK;
2708 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2709 char __user *optval,
2710 unsigned int optlen)
2712 struct sctp_paddrparams params;
2713 struct sctp_transport *trans = NULL;
2714 struct sctp_association *asoc = NULL;
2715 struct sctp_sock *sp = sctp_sk(sk);
2717 int hb_change, pmtud_change, sackdelay_change;
2719 if (optlen == sizeof(params)) {
2720 if (copy_from_user(¶ms, optval, optlen))
2722 } else if (optlen == ALIGN(offsetof(struct sctp_paddrparams,
2723 spp_ipv6_flowlabel), 4)) {
2724 if (copy_from_user(¶ms, optval, optlen))
2726 if (params.spp_flags & (SPP_DSCP | SPP_IPV6_FLOWLABEL))
2732 /* Validate flags and value parameters. */
2733 hb_change = params.spp_flags & SPP_HB;
2734 pmtud_change = params.spp_flags & SPP_PMTUD;
2735 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2737 if (hb_change == SPP_HB ||
2738 pmtud_change == SPP_PMTUD ||
2739 sackdelay_change == SPP_SACKDELAY ||
2740 params.spp_sackdelay > 500 ||
2741 (params.spp_pathmtu &&
2742 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2745 /* If an address other than INADDR_ANY is specified, and
2746 * no transport is found, then the request is invalid.
2748 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
2749 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2750 params.spp_assoc_id);
2755 /* Get association, if assoc_id != 0 and the socket is a one
2756 * to many style socket, and an association was not found, then
2757 * the id was invalid.
2759 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2760 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2763 /* Heartbeat demand can only be sent on a transport or
2764 * association, but not a socket.
2766 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2769 /* Process parameters. */
2770 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2771 hb_change, pmtud_change,
2777 /* If changes are for association, also apply parameters to each
2780 if (!trans && asoc) {
2781 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2783 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2784 hb_change, pmtud_change,
2792 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2794 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2797 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2799 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2803 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2805 * This option will effect the way delayed acks are performed. This
2806 * option allows you to get or set the delayed ack time, in
2807 * milliseconds. It also allows changing the delayed ack frequency.
2808 * Changing the frequency to 1 disables the delayed sack algorithm. If
2809 * the assoc_id is 0, then this sets or gets the endpoints default
2810 * values. If the assoc_id field is non-zero, then the set or get
2811 * effects the specified association for the one to many model (the
2812 * assoc_id field is ignored by the one to one model). Note that if
2813 * sack_delay or sack_freq are 0 when setting this option, then the
2814 * current values will remain unchanged.
2816 * struct sctp_sack_info {
2817 * sctp_assoc_t sack_assoc_id;
2818 * uint32_t sack_delay;
2819 * uint32_t sack_freq;
2822 * sack_assoc_id - This parameter, indicates which association the user
2823 * is performing an action upon. Note that if this field's value is
2824 * zero then the endpoints default value is changed (effecting future
2825 * associations only).
2827 * sack_delay - This parameter contains the number of milliseconds that
2828 * the user is requesting the delayed ACK timer be set to. Note that
2829 * this value is defined in the standard to be between 200 and 500
2832 * sack_freq - This parameter contains the number of packets that must
2833 * be received before a sack is sent without waiting for the delay
2834 * timer to expire. The default value for this is 2, setting this
2835 * value to 1 will disable the delayed sack algorithm.
2838 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2839 char __user *optval, unsigned int optlen)
2841 struct sctp_sack_info params;
2842 struct sctp_transport *trans = NULL;
2843 struct sctp_association *asoc = NULL;
2844 struct sctp_sock *sp = sctp_sk(sk);
2846 if (optlen == sizeof(struct sctp_sack_info)) {
2847 if (copy_from_user(¶ms, optval, optlen))
2850 if (params.sack_delay == 0 && params.sack_freq == 0)
2852 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2853 pr_warn_ratelimited(DEPRECATED
2855 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2856 "Use struct sctp_sack_info instead\n",
2857 current->comm, task_pid_nr(current));
2858 if (copy_from_user(¶ms, optval, optlen))
2861 if (params.sack_delay == 0)
2862 params.sack_freq = 1;
2864 params.sack_freq = 0;
2868 /* Validate value parameter. */
2869 if (params.sack_delay > 500)
2872 /* Get association, if sack_assoc_id != 0 and the socket is a one
2873 * to many style socket, and an association was not found, then
2874 * the id was invalid.
2876 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2877 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2880 if (params.sack_delay) {
2883 msecs_to_jiffies(params.sack_delay);
2885 sctp_spp_sackdelay_enable(asoc->param_flags);
2887 sp->sackdelay = params.sack_delay;
2889 sctp_spp_sackdelay_enable(sp->param_flags);
2893 if (params.sack_freq == 1) {
2896 sctp_spp_sackdelay_disable(asoc->param_flags);
2899 sctp_spp_sackdelay_disable(sp->param_flags);
2901 } else if (params.sack_freq > 1) {
2903 asoc->sackfreq = params.sack_freq;
2905 sctp_spp_sackdelay_enable(asoc->param_flags);
2907 sp->sackfreq = params.sack_freq;
2909 sctp_spp_sackdelay_enable(sp->param_flags);
2913 /* If change is for association, also apply to each transport. */
2915 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2917 if (params.sack_delay) {
2919 msecs_to_jiffies(params.sack_delay);
2920 trans->param_flags =
2921 sctp_spp_sackdelay_enable(trans->param_flags);
2923 if (params.sack_freq == 1) {
2924 trans->param_flags =
2925 sctp_spp_sackdelay_disable(trans->param_flags);
2926 } else if (params.sack_freq > 1) {
2927 trans->sackfreq = params.sack_freq;
2928 trans->param_flags =
2929 sctp_spp_sackdelay_enable(trans->param_flags);
2937 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2939 * Applications can specify protocol parameters for the default association
2940 * initialization. The option name argument to setsockopt() and getsockopt()
2943 * Setting initialization parameters is effective only on an unconnected
2944 * socket (for UDP-style sockets only future associations are effected
2945 * by the change). With TCP-style sockets, this option is inherited by
2946 * sockets derived from a listener socket.
2948 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2950 struct sctp_initmsg sinit;
2951 struct sctp_sock *sp = sctp_sk(sk);
2953 if (optlen != sizeof(struct sctp_initmsg))
2955 if (copy_from_user(&sinit, optval, optlen))
2958 if (sinit.sinit_num_ostreams)
2959 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2960 if (sinit.sinit_max_instreams)
2961 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2962 if (sinit.sinit_max_attempts)
2963 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2964 if (sinit.sinit_max_init_timeo)
2965 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2971 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2973 * Applications that wish to use the sendto() system call may wish to
2974 * specify a default set of parameters that would normally be supplied
2975 * through the inclusion of ancillary data. This socket option allows
2976 * such an application to set the default sctp_sndrcvinfo structure.
2977 * The application that wishes to use this socket option simply passes
2978 * in to this call the sctp_sndrcvinfo structure defined in Section
2979 * 5.2.2) The input parameters accepted by this call include
2980 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2981 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2982 * to this call if the caller is using the UDP model.
2984 static int sctp_setsockopt_default_send_param(struct sock *sk,
2985 char __user *optval,
2986 unsigned int optlen)
2988 struct sctp_sock *sp = sctp_sk(sk);
2989 struct sctp_association *asoc;
2990 struct sctp_sndrcvinfo info;
2992 if (optlen != sizeof(info))
2994 if (copy_from_user(&info, optval, optlen))
2996 if (info.sinfo_flags &
2997 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2998 SCTP_ABORT | SCTP_EOF))
3001 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
3002 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
3005 asoc->default_stream = info.sinfo_stream;
3006 asoc->default_flags = info.sinfo_flags;
3007 asoc->default_ppid = info.sinfo_ppid;
3008 asoc->default_context = info.sinfo_context;
3009 asoc->default_timetolive = info.sinfo_timetolive;
3011 sp->default_stream = info.sinfo_stream;
3012 sp->default_flags = info.sinfo_flags;
3013 sp->default_ppid = info.sinfo_ppid;
3014 sp->default_context = info.sinfo_context;
3015 sp->default_timetolive = info.sinfo_timetolive;
3021 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
3022 * (SCTP_DEFAULT_SNDINFO)
3024 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
3025 char __user *optval,
3026 unsigned int optlen)
3028 struct sctp_sock *sp = sctp_sk(sk);
3029 struct sctp_association *asoc;
3030 struct sctp_sndinfo info;
3032 if (optlen != sizeof(info))
3034 if (copy_from_user(&info, optval, optlen))
3036 if (info.snd_flags &
3037 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
3038 SCTP_ABORT | SCTP_EOF))
3041 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
3042 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
3045 asoc->default_stream = info.snd_sid;
3046 asoc->default_flags = info.snd_flags;
3047 asoc->default_ppid = info.snd_ppid;
3048 asoc->default_context = info.snd_context;
3050 sp->default_stream = info.snd_sid;
3051 sp->default_flags = info.snd_flags;
3052 sp->default_ppid = info.snd_ppid;
3053 sp->default_context = info.snd_context;
3059 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
3061 * Requests that the local SCTP stack use the enclosed peer address as
3062 * the association primary. The enclosed address must be one of the
3063 * association peer's addresses.
3065 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
3066 unsigned int optlen)
3068 struct sctp_prim prim;
3069 struct sctp_transport *trans;
3073 if (optlen != sizeof(struct sctp_prim))
3076 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
3079 /* Allow security module to validate address but need address len. */
3080 af = sctp_get_af_specific(prim.ssp_addr.ss_family);
3084 err = security_sctp_bind_connect(sk, SCTP_PRIMARY_ADDR,
3085 (struct sockaddr *)&prim.ssp_addr,
3090 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
3094 sctp_assoc_set_primary(trans->asoc, trans);
3100 * 7.1.5 SCTP_NODELAY
3102 * Turn on/off any Nagle-like algorithm. This means that packets are
3103 * generally sent as soon as possible and no unnecessary delays are
3104 * introduced, at the cost of more packets in the network. Expects an
3105 * integer boolean flag.
3107 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
3108 unsigned int optlen)
3112 if (optlen < sizeof(int))
3114 if (get_user(val, (int __user *)optval))
3117 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
3123 * 7.1.1 SCTP_RTOINFO
3125 * The protocol parameters used to initialize and bound retransmission
3126 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
3127 * and modify these parameters.
3128 * All parameters are time values, in milliseconds. A value of 0, when
3129 * modifying the parameters, indicates that the current value should not
3133 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
3135 struct sctp_rtoinfo rtoinfo;
3136 struct sctp_association *asoc;
3137 unsigned long rto_min, rto_max;
3138 struct sctp_sock *sp = sctp_sk(sk);
3140 if (optlen != sizeof (struct sctp_rtoinfo))
3143 if (copy_from_user(&rtoinfo, optval, optlen))
3146 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
3148 /* Set the values to the specific association */
3149 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
3152 rto_max = rtoinfo.srto_max;
3153 rto_min = rtoinfo.srto_min;
3156 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
3158 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
3161 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
3163 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
3165 if (rto_min > rto_max)
3169 if (rtoinfo.srto_initial != 0)
3171 msecs_to_jiffies(rtoinfo.srto_initial);
3172 asoc->rto_max = rto_max;
3173 asoc->rto_min = rto_min;
3175 /* If there is no association or the association-id = 0
3176 * set the values to the endpoint.
3178 if (rtoinfo.srto_initial != 0)
3179 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
3180 sp->rtoinfo.srto_max = rto_max;
3181 sp->rtoinfo.srto_min = rto_min;
3189 * 7.1.2 SCTP_ASSOCINFO
3191 * This option is used to tune the maximum retransmission attempts
3192 * of the association.
3193 * Returns an error if the new association retransmission value is
3194 * greater than the sum of the retransmission value of the peer.
3195 * See [SCTP] for more information.
3198 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
3201 struct sctp_assocparams assocparams;
3202 struct sctp_association *asoc;
3204 if (optlen != sizeof(struct sctp_assocparams))
3206 if (copy_from_user(&assocparams, optval, optlen))
3209 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
3211 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
3214 /* Set the values to the specific association */
3216 if (assocparams.sasoc_asocmaxrxt != 0) {
3219 struct sctp_transport *peer_addr;
3221 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
3223 path_sum += peer_addr->pathmaxrxt;
3227 /* Only validate asocmaxrxt if we have more than
3228 * one path/transport. We do this because path
3229 * retransmissions are only counted when we have more
3233 assocparams.sasoc_asocmaxrxt > path_sum)
3236 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3239 if (assocparams.sasoc_cookie_life != 0)
3240 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3242 /* Set the values to the endpoint */
3243 struct sctp_sock *sp = sctp_sk(sk);
3245 if (assocparams.sasoc_asocmaxrxt != 0)
3246 sp->assocparams.sasoc_asocmaxrxt =
3247 assocparams.sasoc_asocmaxrxt;
3248 if (assocparams.sasoc_cookie_life != 0)
3249 sp->assocparams.sasoc_cookie_life =
3250 assocparams.sasoc_cookie_life;
3256 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3258 * This socket option is a boolean flag which turns on or off mapped V4
3259 * addresses. If this option is turned on and the socket is type
3260 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3261 * If this option is turned off, then no mapping will be done of V4
3262 * addresses and a user will receive both PF_INET6 and PF_INET type
3263 * addresses on the socket.
3265 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3268 struct sctp_sock *sp = sctp_sk(sk);
3270 if (optlen < sizeof(int))
3272 if (get_user(val, (int __user *)optval))
3283 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3284 * This option will get or set the maximum size to put in any outgoing
3285 * SCTP DATA chunk. If a message is larger than this size it will be
3286 * fragmented by SCTP into the specified size. Note that the underlying
3287 * SCTP implementation may fragment into smaller sized chunks when the
3288 * PMTU of the underlying association is smaller than the value set by
3289 * the user. The default value for this option is '0' which indicates
3290 * the user is NOT limiting fragmentation and only the PMTU will effect
3291 * SCTP's choice of DATA chunk size. Note also that values set larger
3292 * than the maximum size of an IP datagram will effectively let SCTP
3293 * control fragmentation (i.e. the same as setting this option to 0).
3295 * The following structure is used to access and modify this parameter:
3297 * struct sctp_assoc_value {
3298 * sctp_assoc_t assoc_id;
3299 * uint32_t assoc_value;
3302 * assoc_id: This parameter is ignored for one-to-one style sockets.
3303 * For one-to-many style sockets this parameter indicates which
3304 * association the user is performing an action upon. Note that if
3305 * this field's value is zero then the endpoints default value is
3306 * changed (effecting future associations only).
3307 * assoc_value: This parameter specifies the maximum size in bytes.
3309 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3311 struct sctp_sock *sp = sctp_sk(sk);
3312 struct sctp_assoc_value params;
3313 struct sctp_association *asoc;
3316 if (optlen == sizeof(int)) {
3317 pr_warn_ratelimited(DEPRECATED
3319 "Use of int in maxseg socket option.\n"
3320 "Use struct sctp_assoc_value instead\n",
3321 current->comm, task_pid_nr(current));
3322 if (copy_from_user(&val, optval, optlen))
3324 params.assoc_id = 0;
3325 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3326 if (copy_from_user(¶ms, optval, optlen))
3328 val = params.assoc_value;
3333 asoc = sctp_id2assoc(sk, params.assoc_id);
3336 int min_len, max_len;
3337 __u16 datasize = asoc ? sctp_datachk_len(&asoc->stream) :
3338 sizeof(struct sctp_data_chunk);
3340 min_len = sctp_mtu_payload(sp, SCTP_DEFAULT_MINSEGMENT,
3342 max_len = SCTP_MAX_CHUNK_LEN - datasize;
3344 if (val < min_len || val > max_len)
3349 asoc->user_frag = val;
3350 sctp_assoc_update_frag_point(asoc);
3352 if (params.assoc_id && sctp_style(sk, UDP))
3354 sp->user_frag = val;
3362 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3364 * Requests that the peer mark the enclosed address as the association
3365 * primary. The enclosed address must be one of the association's
3366 * locally bound addresses. The following structure is used to make a
3367 * set primary request:
3369 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3370 unsigned int optlen)
3372 struct net *net = sock_net(sk);
3373 struct sctp_sock *sp;
3374 struct sctp_association *asoc = NULL;
3375 struct sctp_setpeerprim prim;
3376 struct sctp_chunk *chunk;
3382 if (!net->sctp.addip_enable)
3385 if (optlen != sizeof(struct sctp_setpeerprim))
3388 if (copy_from_user(&prim, optval, optlen))
3391 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3395 if (!asoc->peer.asconf_capable)
3398 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3401 if (!sctp_state(asoc, ESTABLISHED))
3404 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3408 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3409 return -EADDRNOTAVAIL;
3411 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3412 return -EADDRNOTAVAIL;
3414 /* Allow security module to validate address. */
3415 err = security_sctp_bind_connect(sk, SCTP_SET_PEER_PRIMARY_ADDR,
3416 (struct sockaddr *)&prim.sspp_addr,
3421 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3422 chunk = sctp_make_asconf_set_prim(asoc,
3423 (union sctp_addr *)&prim.sspp_addr);
3427 err = sctp_send_asconf(asoc, chunk);
3429 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3434 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3435 unsigned int optlen)
3437 struct sctp_setadaptation adaptation;
3439 if (optlen != sizeof(struct sctp_setadaptation))
3441 if (copy_from_user(&adaptation, optval, optlen))
3444 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3450 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3452 * The context field in the sctp_sndrcvinfo structure is normally only
3453 * used when a failed message is retrieved holding the value that was
3454 * sent down on the actual send call. This option allows the setting of
3455 * a default context on an association basis that will be received on
3456 * reading messages from the peer. This is especially helpful in the
3457 * one-2-many model for an application to keep some reference to an
3458 * internal state machine that is processing messages on the
3459 * association. Note that the setting of this value only effects
3460 * received messages from the peer and does not effect the value that is
3461 * saved with outbound messages.
3463 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3464 unsigned int optlen)
3466 struct sctp_assoc_value params;
3467 struct sctp_sock *sp;
3468 struct sctp_association *asoc;
3470 if (optlen != sizeof(struct sctp_assoc_value))
3472 if (copy_from_user(¶ms, optval, optlen))
3477 if (params.assoc_id != 0) {
3478 asoc = sctp_id2assoc(sk, params.assoc_id);
3481 asoc->default_rcv_context = params.assoc_value;
3483 sp->default_rcv_context = params.assoc_value;
3490 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3492 * This options will at a minimum specify if the implementation is doing
3493 * fragmented interleave. Fragmented interleave, for a one to many
3494 * socket, is when subsequent calls to receive a message may return
3495 * parts of messages from different associations. Some implementations
3496 * may allow you to turn this value on or off. If so, when turned off,
3497 * no fragment interleave will occur (which will cause a head of line
3498 * blocking amongst multiple associations sharing the same one to many
3499 * socket). When this option is turned on, then each receive call may
3500 * come from a different association (thus the user must receive data
3501 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3502 * association each receive belongs to.
3504 * This option takes a boolean value. A non-zero value indicates that
3505 * fragmented interleave is on. A value of zero indicates that
3506 * fragmented interleave is off.
3508 * Note that it is important that an implementation that allows this
3509 * option to be turned on, have it off by default. Otherwise an unaware
3510 * application using the one to many model may become confused and act
3513 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3514 char __user *optval,
3515 unsigned int optlen)
3519 if (optlen != sizeof(int))
3521 if (get_user(val, (int __user *)optval))
3524 sctp_sk(sk)->frag_interleave = !!val;
3526 if (!sctp_sk(sk)->frag_interleave)
3527 sctp_sk(sk)->strm_interleave = 0;
3533 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3534 * (SCTP_PARTIAL_DELIVERY_POINT)
3536 * This option will set or get the SCTP partial delivery point. This
3537 * point is the size of a message where the partial delivery API will be
3538 * invoked to help free up rwnd space for the peer. Setting this to a
3539 * lower value will cause partial deliveries to happen more often. The
3540 * calls argument is an integer that sets or gets the partial delivery
3541 * point. Note also that the call will fail if the user attempts to set
3542 * this value larger than the socket receive buffer size.
3544 * Note that any single message having a length smaller than or equal to
3545 * the SCTP partial delivery point will be delivered in one single read
3546 * call as long as the user provided buffer is large enough to hold the
3549 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3550 char __user *optval,
3551 unsigned int optlen)
3555 if (optlen != sizeof(u32))
3557 if (get_user(val, (int __user *)optval))
3560 /* Note: We double the receive buffer from what the user sets
3561 * it to be, also initial rwnd is based on rcvbuf/2.
3563 if (val > (sk->sk_rcvbuf >> 1))
3566 sctp_sk(sk)->pd_point = val;
3568 return 0; /* is this the right error code? */
3572 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3574 * This option will allow a user to change the maximum burst of packets
3575 * that can be emitted by this association. Note that the default value
3576 * is 4, and some implementations may restrict this setting so that it
3577 * can only be lowered.
3579 * NOTE: This text doesn't seem right. Do this on a socket basis with
3580 * future associations inheriting the socket value.
3582 static int sctp_setsockopt_maxburst(struct sock *sk,
3583 char __user *optval,
3584 unsigned int optlen)
3586 struct sctp_assoc_value params;
3587 struct sctp_sock *sp;
3588 struct sctp_association *asoc;
3592 if (optlen == sizeof(int)) {
3593 pr_warn_ratelimited(DEPRECATED
3595 "Use of int in max_burst socket option deprecated.\n"
3596 "Use struct sctp_assoc_value instead\n",
3597 current->comm, task_pid_nr(current));
3598 if (copy_from_user(&val, optval, optlen))
3600 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3601 if (copy_from_user(¶ms, optval, optlen))
3603 val = params.assoc_value;
3604 assoc_id = params.assoc_id;
3610 if (assoc_id != 0) {
3611 asoc = sctp_id2assoc(sk, assoc_id);
3614 asoc->max_burst = val;
3616 sp->max_burst = val;
3622 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3624 * This set option adds a chunk type that the user is requesting to be
3625 * received only in an authenticated way. Changes to the list of chunks
3626 * will only effect future associations on the socket.
3628 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3629 char __user *optval,
3630 unsigned int optlen)
3632 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3633 struct sctp_authchunk val;
3635 if (!ep->auth_enable)
3638 if (optlen != sizeof(struct sctp_authchunk))
3640 if (copy_from_user(&val, optval, optlen))
3643 switch (val.sauth_chunk) {
3645 case SCTP_CID_INIT_ACK:
3646 case SCTP_CID_SHUTDOWN_COMPLETE:
3651 /* add this chunk id to the endpoint */
3652 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3656 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3658 * This option gets or sets the list of HMAC algorithms that the local
3659 * endpoint requires the peer to use.
3661 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3662 char __user *optval,
3663 unsigned int optlen)
3665 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3666 struct sctp_hmacalgo *hmacs;
3670 if (!ep->auth_enable)
3673 if (optlen < sizeof(struct sctp_hmacalgo))
3675 optlen = min_t(unsigned int, optlen, sizeof(struct sctp_hmacalgo) +
3676 SCTP_AUTH_NUM_HMACS * sizeof(u16));
3678 hmacs = memdup_user(optval, optlen);
3680 return PTR_ERR(hmacs);
3682 idents = hmacs->shmac_num_idents;
3683 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3684 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3689 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3696 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3698 * This option will set a shared secret key which is used to build an
3699 * association shared key.
3701 static int sctp_setsockopt_auth_key(struct sock *sk,
3702 char __user *optval,
3703 unsigned int optlen)
3705 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3706 struct sctp_authkey *authkey;
3707 struct sctp_association *asoc;
3710 if (!ep->auth_enable)
3713 if (optlen <= sizeof(struct sctp_authkey))
3715 /* authkey->sca_keylength is u16, so optlen can't be bigger than
3718 optlen = min_t(unsigned int, optlen, USHRT_MAX +
3719 sizeof(struct sctp_authkey));
3721 authkey = memdup_user(optval, optlen);
3722 if (IS_ERR(authkey))
3723 return PTR_ERR(authkey);
3725 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3730 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3731 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3736 ret = sctp_auth_set_key(ep, asoc, authkey);
3743 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3745 * This option will get or set the active shared key to be used to build
3746 * the association shared key.
3748 static int sctp_setsockopt_active_key(struct sock *sk,
3749 char __user *optval,
3750 unsigned int optlen)
3752 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3753 struct sctp_authkeyid val;
3754 struct sctp_association *asoc;
3756 if (!ep->auth_enable)
3759 if (optlen != sizeof(struct sctp_authkeyid))
3761 if (copy_from_user(&val, optval, optlen))
3764 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3765 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3768 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3772 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3774 * This set option will delete a shared secret key from use.
3776 static int sctp_setsockopt_del_key(struct sock *sk,
3777 char __user *optval,
3778 unsigned int optlen)
3780 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3781 struct sctp_authkeyid val;
3782 struct sctp_association *asoc;
3784 if (!ep->auth_enable)
3787 if (optlen != sizeof(struct sctp_authkeyid))
3789 if (copy_from_user(&val, optval, optlen))
3792 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3793 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3796 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3801 * 8.3.4 Deactivate a Shared Key (SCTP_AUTH_DEACTIVATE_KEY)
3803 * This set option will deactivate a shared secret key.
3805 static int sctp_setsockopt_deactivate_key(struct sock *sk, char __user *optval,
3806 unsigned int optlen)
3808 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3809 struct sctp_authkeyid val;
3810 struct sctp_association *asoc;
3812 if (!ep->auth_enable)
3815 if (optlen != sizeof(struct sctp_authkeyid))
3817 if (copy_from_user(&val, optval, optlen))
3820 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3821 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3824 return sctp_auth_deact_key_id(ep, asoc, val.scact_keynumber);
3828 * 8.1.23 SCTP_AUTO_ASCONF
3830 * This option will enable or disable the use of the automatic generation of
3831 * ASCONF chunks to add and delete addresses to an existing association. Note
3832 * that this option has two caveats namely: a) it only affects sockets that
3833 * are bound to all addresses available to the SCTP stack, and b) the system
3834 * administrator may have an overriding control that turns the ASCONF feature
3835 * off no matter what setting the socket option may have.
3836 * This option expects an integer boolean flag, where a non-zero value turns on
3837 * the option, and a zero value turns off the option.
3838 * Note. In this implementation, socket operation overrides default parameter
3839 * being set by sysctl as well as FreeBSD implementation
3841 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3842 unsigned int optlen)
3845 struct sctp_sock *sp = sctp_sk(sk);
3847 if (optlen < sizeof(int))
3849 if (get_user(val, (int __user *)optval))
3851 if (!sctp_is_ep_boundall(sk) && val)
3853 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3856 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3857 if (val == 0 && sp->do_auto_asconf) {
3858 list_del(&sp->auto_asconf_list);
3859 sp->do_auto_asconf = 0;
3860 } else if (val && !sp->do_auto_asconf) {
3861 list_add_tail(&sp->auto_asconf_list,
3862 &sock_net(sk)->sctp.auto_asconf_splist);
3863 sp->do_auto_asconf = 1;
3865 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3870 * SCTP_PEER_ADDR_THLDS
3872 * This option allows us to alter the partially failed threshold for one or all
3873 * transports in an association. See Section 6.1 of:
3874 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3876 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3877 char __user *optval,
3878 unsigned int optlen)
3880 struct sctp_paddrthlds val;
3881 struct sctp_transport *trans;
3882 struct sctp_association *asoc;
3884 if (optlen < sizeof(struct sctp_paddrthlds))
3886 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3887 sizeof(struct sctp_paddrthlds)))
3891 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3892 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3895 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3897 if (val.spt_pathmaxrxt)
3898 trans->pathmaxrxt = val.spt_pathmaxrxt;
3899 trans->pf_retrans = val.spt_pathpfthld;
3902 if (val.spt_pathmaxrxt)
3903 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3904 asoc->pf_retrans = val.spt_pathpfthld;
3906 trans = sctp_addr_id2transport(sk, &val.spt_address,
3911 if (val.spt_pathmaxrxt)
3912 trans->pathmaxrxt = val.spt_pathmaxrxt;
3913 trans->pf_retrans = val.spt_pathpfthld;
3919 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3920 char __user *optval,
3921 unsigned int optlen)
3925 if (optlen < sizeof(int))
3927 if (get_user(val, (int __user *) optval))
3930 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3935 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3936 char __user *optval,
3937 unsigned int optlen)
3941 if (optlen < sizeof(int))
3943 if (get_user(val, (int __user *) optval))
3946 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3951 static int sctp_setsockopt_pr_supported(struct sock *sk,
3952 char __user *optval,
3953 unsigned int optlen)
3955 struct sctp_assoc_value params;
3956 struct sctp_association *asoc;
3957 int retval = -EINVAL;
3959 if (optlen != sizeof(params))
3962 if (copy_from_user(¶ms, optval, optlen)) {
3967 asoc = sctp_id2assoc(sk, params.assoc_id);
3969 asoc->prsctp_enable = !!params.assoc_value;
3970 } else if (!params.assoc_id) {
3971 struct sctp_sock *sp = sctp_sk(sk);
3973 sp->ep->prsctp_enable = !!params.assoc_value;
3984 static int sctp_setsockopt_default_prinfo(struct sock *sk,
3985 char __user *optval,
3986 unsigned int optlen)
3988 struct sctp_default_prinfo info;
3989 struct sctp_association *asoc;
3990 int retval = -EINVAL;
3992 if (optlen != sizeof(info))
3995 if (copy_from_user(&info, optval, sizeof(info))) {
4000 if (info.pr_policy & ~SCTP_PR_SCTP_MASK)
4003 if (info.pr_policy == SCTP_PR_SCTP_NONE)
4006 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
4008 SCTP_PR_SET_POLICY(asoc->default_flags, info.pr_policy);
4009 asoc->default_timetolive = info.pr_value;
4010 } else if (!info.pr_assoc_id) {
4011 struct sctp_sock *sp = sctp_sk(sk);
4013 SCTP_PR_SET_POLICY(sp->default_flags, info.pr_policy);
4014 sp->default_timetolive = info.pr_value;
4025 static int sctp_setsockopt_reconfig_supported(struct sock *sk,
4026 char __user *optval,
4027 unsigned int optlen)
4029 struct sctp_assoc_value params;
4030 struct sctp_association *asoc;
4031 int retval = -EINVAL;
4033 if (optlen != sizeof(params))
4036 if (copy_from_user(¶ms, optval, optlen)) {
4041 asoc = sctp_id2assoc(sk, params.assoc_id);
4043 asoc->reconf_enable = !!params.assoc_value;
4044 } else if (!params.assoc_id) {
4045 struct sctp_sock *sp = sctp_sk(sk);
4047 sp->ep->reconf_enable = !!params.assoc_value;
4058 static int sctp_setsockopt_enable_strreset(struct sock *sk,
4059 char __user *optval,
4060 unsigned int optlen)
4062 struct sctp_assoc_value params;
4063 struct sctp_association *asoc;
4064 int retval = -EINVAL;
4066 if (optlen != sizeof(params))
4069 if (copy_from_user(¶ms, optval, optlen)) {
4074 if (params.assoc_value & (~SCTP_ENABLE_STRRESET_MASK))
4077 asoc = sctp_id2assoc(sk, params.assoc_id);
4079 asoc->strreset_enable = params.assoc_value;
4080 } else if (!params.assoc_id) {
4081 struct sctp_sock *sp = sctp_sk(sk);
4083 sp->ep->strreset_enable = params.assoc_value;
4094 static int sctp_setsockopt_reset_streams(struct sock *sk,
4095 char __user *optval,
4096 unsigned int optlen)
4098 struct sctp_reset_streams *params;
4099 struct sctp_association *asoc;
4100 int retval = -EINVAL;
4102 if (optlen < sizeof(*params))
4104 /* srs_number_streams is u16, so optlen can't be bigger than this. */
4105 optlen = min_t(unsigned int, optlen, USHRT_MAX +
4106 sizeof(__u16) * sizeof(*params));
4108 params = memdup_user(optval, optlen);
4110 return PTR_ERR(params);
4112 if (params->srs_number_streams * sizeof(__u16) >
4113 optlen - sizeof(*params))
4116 asoc = sctp_id2assoc(sk, params->srs_assoc_id);
4120 retval = sctp_send_reset_streams(asoc, params);
4127 static int sctp_setsockopt_reset_assoc(struct sock *sk,
4128 char __user *optval,
4129 unsigned int optlen)
4131 struct sctp_association *asoc;
4132 sctp_assoc_t associd;
4133 int retval = -EINVAL;
4135 if (optlen != sizeof(associd))
4138 if (copy_from_user(&associd, optval, optlen)) {
4143 asoc = sctp_id2assoc(sk, associd);
4147 retval = sctp_send_reset_assoc(asoc);
4153 static int sctp_setsockopt_add_streams(struct sock *sk,
4154 char __user *optval,
4155 unsigned int optlen)
4157 struct sctp_association *asoc;
4158 struct sctp_add_streams params;
4159 int retval = -EINVAL;
4161 if (optlen != sizeof(params))
4164 if (copy_from_user(¶ms, optval, optlen)) {
4169 asoc = sctp_id2assoc(sk, params.sas_assoc_id);
4173 retval = sctp_send_add_streams(asoc, ¶ms);
4179 static int sctp_setsockopt_scheduler(struct sock *sk,
4180 char __user *optval,
4181 unsigned int optlen)
4183 struct sctp_association *asoc;
4184 struct sctp_assoc_value params;
4185 int retval = -EINVAL;
4187 if (optlen < sizeof(params))
4190 optlen = sizeof(params);
4191 if (copy_from_user(¶ms, optval, optlen)) {
4196 if (params.assoc_value > SCTP_SS_MAX)
4199 asoc = sctp_id2assoc(sk, params.assoc_id);
4203 retval = sctp_sched_set_sched(asoc, params.assoc_value);
4209 static int sctp_setsockopt_scheduler_value(struct sock *sk,
4210 char __user *optval,
4211 unsigned int optlen)
4213 struct sctp_association *asoc;
4214 struct sctp_stream_value params;
4215 int retval = -EINVAL;
4217 if (optlen < sizeof(params))
4220 optlen = sizeof(params);
4221 if (copy_from_user(¶ms, optval, optlen)) {
4226 asoc = sctp_id2assoc(sk, params.assoc_id);
4230 retval = sctp_sched_set_value(asoc, params.stream_id,
4231 params.stream_value, GFP_KERNEL);
4237 static int sctp_setsockopt_interleaving_supported(struct sock *sk,
4238 char __user *optval,
4239 unsigned int optlen)
4241 struct sctp_sock *sp = sctp_sk(sk);
4242 struct net *net = sock_net(sk);
4243 struct sctp_assoc_value params;
4244 int retval = -EINVAL;
4246 if (optlen < sizeof(params))
4249 optlen = sizeof(params);
4250 if (copy_from_user(¶ms, optval, optlen)) {
4255 if (params.assoc_id)
4258 if (!net->sctp.intl_enable || !sp->frag_interleave) {
4263 sp->strm_interleave = !!params.assoc_value;
4271 static int sctp_setsockopt_reuse_port(struct sock *sk, char __user *optval,
4272 unsigned int optlen)
4276 if (!sctp_style(sk, TCP))
4279 if (sctp_sk(sk)->ep->base.bind_addr.port)
4282 if (optlen < sizeof(int))
4285 if (get_user(val, (int __user *)optval))
4288 sctp_sk(sk)->reuse = !!val;
4293 /* API 6.2 setsockopt(), getsockopt()
4295 * Applications use setsockopt() and getsockopt() to set or retrieve
4296 * socket options. Socket options are used to change the default
4297 * behavior of sockets calls. They are described in Section 7.
4301 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
4302 * int __user *optlen);
4303 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
4306 * sd - the socket descript.
4307 * level - set to IPPROTO_SCTP for all SCTP options.
4308 * optname - the option name.
4309 * optval - the buffer to store the value of the option.
4310 * optlen - the size of the buffer.
4312 static int sctp_setsockopt(struct sock *sk, int level, int optname,
4313 char __user *optval, unsigned int optlen)
4317 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
4319 /* I can hardly begin to describe how wrong this is. This is
4320 * so broken as to be worse than useless. The API draft
4321 * REALLY is NOT helpful here... I am not convinced that the
4322 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
4323 * are at all well-founded.
4325 if (level != SOL_SCTP) {
4326 struct sctp_af *af = sctp_sk(sk)->pf->af;
4327 retval = af->setsockopt(sk, level, optname, optval, optlen);
4334 case SCTP_SOCKOPT_BINDX_ADD:
4335 /* 'optlen' is the size of the addresses buffer. */
4336 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
4337 optlen, SCTP_BINDX_ADD_ADDR);
4340 case SCTP_SOCKOPT_BINDX_REM:
4341 /* 'optlen' is the size of the addresses buffer. */
4342 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
4343 optlen, SCTP_BINDX_REM_ADDR);
4346 case SCTP_SOCKOPT_CONNECTX_OLD:
4347 /* 'optlen' is the size of the addresses buffer. */
4348 retval = sctp_setsockopt_connectx_old(sk,
4349 (struct sockaddr __user *)optval,
4353 case SCTP_SOCKOPT_CONNECTX:
4354 /* 'optlen' is the size of the addresses buffer. */
4355 retval = sctp_setsockopt_connectx(sk,
4356 (struct sockaddr __user *)optval,
4360 case SCTP_DISABLE_FRAGMENTS:
4361 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
4365 retval = sctp_setsockopt_events(sk, optval, optlen);
4368 case SCTP_AUTOCLOSE:
4369 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
4372 case SCTP_PEER_ADDR_PARAMS:
4373 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
4376 case SCTP_DELAYED_SACK:
4377 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
4379 case SCTP_PARTIAL_DELIVERY_POINT:
4380 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
4384 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
4386 case SCTP_DEFAULT_SEND_PARAM:
4387 retval = sctp_setsockopt_default_send_param(sk, optval,
4390 case SCTP_DEFAULT_SNDINFO:
4391 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
4393 case SCTP_PRIMARY_ADDR:
4394 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
4396 case SCTP_SET_PEER_PRIMARY_ADDR:
4397 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
4400 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
4403 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
4405 case SCTP_ASSOCINFO:
4406 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
4408 case SCTP_I_WANT_MAPPED_V4_ADDR:
4409 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
4412 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
4414 case SCTP_ADAPTATION_LAYER:
4415 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
4418 retval = sctp_setsockopt_context(sk, optval, optlen);
4420 case SCTP_FRAGMENT_INTERLEAVE:
4421 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
4423 case SCTP_MAX_BURST:
4424 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
4426 case SCTP_AUTH_CHUNK:
4427 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
4429 case SCTP_HMAC_IDENT:
4430 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
4433 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
4435 case SCTP_AUTH_ACTIVE_KEY:
4436 retval = sctp_setsockopt_active_key(sk, optval, optlen);
4438 case SCTP_AUTH_DELETE_KEY:
4439 retval = sctp_setsockopt_del_key(sk, optval, optlen);
4441 case SCTP_AUTH_DEACTIVATE_KEY:
4442 retval = sctp_setsockopt_deactivate_key(sk, optval, optlen);
4444 case SCTP_AUTO_ASCONF:
4445 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
4447 case SCTP_PEER_ADDR_THLDS:
4448 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
4450 case SCTP_RECVRCVINFO:
4451 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
4453 case SCTP_RECVNXTINFO:
4454 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
4456 case SCTP_PR_SUPPORTED:
4457 retval = sctp_setsockopt_pr_supported(sk, optval, optlen);
4459 case SCTP_DEFAULT_PRINFO:
4460 retval = sctp_setsockopt_default_prinfo(sk, optval, optlen);
4462 case SCTP_RECONFIG_SUPPORTED:
4463 retval = sctp_setsockopt_reconfig_supported(sk, optval, optlen);
4465 case SCTP_ENABLE_STREAM_RESET:
4466 retval = sctp_setsockopt_enable_strreset(sk, optval, optlen);
4468 case SCTP_RESET_STREAMS:
4469 retval = sctp_setsockopt_reset_streams(sk, optval, optlen);
4471 case SCTP_RESET_ASSOC:
4472 retval = sctp_setsockopt_reset_assoc(sk, optval, optlen);
4474 case SCTP_ADD_STREAMS:
4475 retval = sctp_setsockopt_add_streams(sk, optval, optlen);
4477 case SCTP_STREAM_SCHEDULER:
4478 retval = sctp_setsockopt_scheduler(sk, optval, optlen);
4480 case SCTP_STREAM_SCHEDULER_VALUE:
4481 retval = sctp_setsockopt_scheduler_value(sk, optval, optlen);
4483 case SCTP_INTERLEAVING_SUPPORTED:
4484 retval = sctp_setsockopt_interleaving_supported(sk, optval,
4487 case SCTP_REUSE_PORT:
4488 retval = sctp_setsockopt_reuse_port(sk, optval, optlen);
4491 retval = -ENOPROTOOPT;
4501 /* API 3.1.6 connect() - UDP Style Syntax
4503 * An application may use the connect() call in the UDP model to initiate an
4504 * association without sending data.
4508 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
4510 * sd: the socket descriptor to have a new association added to.
4512 * nam: the address structure (either struct sockaddr_in or struct
4513 * sockaddr_in6 defined in RFC2553 [7]).
4515 * len: the size of the address.
4517 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
4518 int addr_len, int flags)
4520 struct inet_sock *inet = inet_sk(sk);
4526 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
4529 /* We may need to bind the socket. */
4530 if (!inet->inet_num) {
4531 if (sk->sk_prot->get_port(sk, 0)) {
4535 inet->inet_sport = htons(inet->inet_num);
4538 /* Validate addr_len before calling common connect/connectx routine. */
4539 af = sctp_get_af_specific(addr->sa_family);
4540 if (!af || addr_len < af->sockaddr_len) {
4543 /* Pass correct addr len to common routine (so it knows there
4544 * is only one address being passed.
4546 err = __sctp_connect(sk, addr, af->sockaddr_len, flags, NULL);
4553 int sctp_inet_connect(struct socket *sock, struct sockaddr *uaddr,
4554 int addr_len, int flags)
4556 if (addr_len < sizeof(uaddr->sa_family))
4559 if (uaddr->sa_family == AF_UNSPEC)
4562 return sctp_connect(sock->sk, uaddr, addr_len, flags);
4565 /* FIXME: Write comments. */
4566 static int sctp_disconnect(struct sock *sk, int flags)
4568 return -EOPNOTSUPP; /* STUB */
4571 /* 4.1.4 accept() - TCP Style Syntax
4573 * Applications use accept() call to remove an established SCTP
4574 * association from the accept queue of the endpoint. A new socket
4575 * descriptor will be returned from accept() to represent the newly
4576 * formed association.
4578 static struct sock *sctp_accept(struct sock *sk, int flags, int *err, bool kern)
4580 struct sctp_sock *sp;
4581 struct sctp_endpoint *ep;
4582 struct sock *newsk = NULL;
4583 struct sctp_association *asoc;
4592 if (!sctp_style(sk, TCP)) {
4593 error = -EOPNOTSUPP;
4597 if (!sctp_sstate(sk, LISTENING)) {
4602 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
4604 error = sctp_wait_for_accept(sk, timeo);
4608 /* We treat the list of associations on the endpoint as the accept
4609 * queue and pick the first association on the list.
4611 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
4613 newsk = sp->pf->create_accept_sk(sk, asoc, kern);
4619 /* Populate the fields of the newsk from the oldsk and migrate the
4620 * asoc to the newsk.
4622 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
4630 /* The SCTP ioctl handler. */
4631 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
4638 * SEQPACKET-style sockets in LISTENING state are valid, for
4639 * SCTP, so only discard TCP-style sockets in LISTENING state.
4641 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4646 struct sk_buff *skb;
4647 unsigned int amount = 0;
4649 skb = skb_peek(&sk->sk_receive_queue);
4652 * We will only return the amount of this packet since
4653 * that is all that will be read.
4657 rc = put_user(amount, (int __user *)arg);
4669 /* This is the function which gets called during socket creation to
4670 * initialized the SCTP-specific portion of the sock.
4671 * The sock structure should already be zero-filled memory.
4673 static int sctp_init_sock(struct sock *sk)
4675 struct net *net = sock_net(sk);
4676 struct sctp_sock *sp;
4678 pr_debug("%s: sk:%p\n", __func__, sk);
4682 /* Initialize the SCTP per socket area. */
4683 switch (sk->sk_type) {
4684 case SOCK_SEQPACKET:
4685 sp->type = SCTP_SOCKET_UDP;
4688 sp->type = SCTP_SOCKET_TCP;
4691 return -ESOCKTNOSUPPORT;
4694 sk->sk_gso_type = SKB_GSO_SCTP;
4696 /* Initialize default send parameters. These parameters can be
4697 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4699 sp->default_stream = 0;
4700 sp->default_ppid = 0;
4701 sp->default_flags = 0;
4702 sp->default_context = 0;
4703 sp->default_timetolive = 0;
4705 sp->default_rcv_context = 0;
4706 sp->max_burst = net->sctp.max_burst;
4708 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4710 /* Initialize default setup parameters. These parameters
4711 * can be modified with the SCTP_INITMSG socket option or
4712 * overridden by the SCTP_INIT CMSG.
4714 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4715 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4716 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4717 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4719 /* Initialize default RTO related parameters. These parameters can
4720 * be modified for with the SCTP_RTOINFO socket option.
4722 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4723 sp->rtoinfo.srto_max = net->sctp.rto_max;
4724 sp->rtoinfo.srto_min = net->sctp.rto_min;
4726 /* Initialize default association related parameters. These parameters
4727 * can be modified with the SCTP_ASSOCINFO socket option.
4729 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4730 sp->assocparams.sasoc_number_peer_destinations = 0;
4731 sp->assocparams.sasoc_peer_rwnd = 0;
4732 sp->assocparams.sasoc_local_rwnd = 0;
4733 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4735 /* Initialize default event subscriptions. By default, all the
4738 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4740 /* Default Peer Address Parameters. These defaults can
4741 * be modified via SCTP_PEER_ADDR_PARAMS
4743 sp->hbinterval = net->sctp.hb_interval;
4744 sp->pathmaxrxt = net->sctp.max_retrans_path;
4745 sp->pathmtu = 0; /* allow default discovery */
4746 sp->sackdelay = net->sctp.sack_timeout;
4748 sp->param_flags = SPP_HB_ENABLE |
4750 SPP_SACKDELAY_ENABLE;
4752 /* If enabled no SCTP message fragmentation will be performed.
4753 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4755 sp->disable_fragments = 0;
4757 /* Enable Nagle algorithm by default. */
4760 sp->recvrcvinfo = 0;
4761 sp->recvnxtinfo = 0;
4763 /* Enable by default. */
4766 /* Auto-close idle associations after the configured
4767 * number of seconds. A value of 0 disables this
4768 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4769 * for UDP-style sockets only.
4773 /* User specified fragmentation limit. */
4776 sp->adaptation_ind = 0;
4778 sp->pf = sctp_get_pf_specific(sk->sk_family);
4780 /* Control variables for partial data delivery. */
4781 atomic_set(&sp->pd_mode, 0);
4782 skb_queue_head_init(&sp->pd_lobby);
4783 sp->frag_interleave = 0;
4785 /* Create a per socket endpoint structure. Even if we
4786 * change the data structure relationships, this may still
4787 * be useful for storing pre-connect address information.
4789 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4795 sk->sk_destruct = sctp_destruct_sock;
4797 SCTP_DBG_OBJCNT_INC(sock);
4800 sk_sockets_allocated_inc(sk);
4801 sock_prot_inuse_add(net, sk->sk_prot, 1);
4803 /* Nothing can fail after this block, otherwise
4804 * sctp_destroy_sock() will be called without addr_wq_lock held
4806 if (net->sctp.default_auto_asconf) {
4807 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
4808 list_add_tail(&sp->auto_asconf_list,
4809 &net->sctp.auto_asconf_splist);
4810 sp->do_auto_asconf = 1;
4811 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4813 sp->do_auto_asconf = 0;
4821 /* Cleanup any SCTP per socket resources. Must be called with
4822 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4824 static void sctp_destroy_sock(struct sock *sk)
4826 struct sctp_sock *sp;
4828 pr_debug("%s: sk:%p\n", __func__, sk);
4830 /* Release our hold on the endpoint. */
4832 /* This could happen during socket init, thus we bail out
4833 * early, since the rest of the below is not setup either.
4838 if (sp->do_auto_asconf) {
4839 sp->do_auto_asconf = 0;
4840 list_del(&sp->auto_asconf_list);
4842 sctp_endpoint_free(sp->ep);
4844 sk_sockets_allocated_dec(sk);
4845 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4849 /* Triggered when there are no references on the socket anymore */
4850 static void sctp_destruct_sock(struct sock *sk)
4852 struct sctp_sock *sp = sctp_sk(sk);
4854 /* Free up the HMAC transform. */
4855 crypto_free_shash(sp->hmac);
4857 inet_sock_destruct(sk);
4860 /* API 4.1.7 shutdown() - TCP Style Syntax
4861 * int shutdown(int socket, int how);
4863 * sd - the socket descriptor of the association to be closed.
4864 * how - Specifies the type of shutdown. The values are
4867 * Disables further receive operations. No SCTP
4868 * protocol action is taken.
4870 * Disables further send operations, and initiates
4871 * the SCTP shutdown sequence.
4873 * Disables further send and receive operations
4874 * and initiates the SCTP shutdown sequence.
4876 static void sctp_shutdown(struct sock *sk, int how)
4878 struct net *net = sock_net(sk);
4879 struct sctp_endpoint *ep;
4881 if (!sctp_style(sk, TCP))
4884 ep = sctp_sk(sk)->ep;
4885 if (how & SEND_SHUTDOWN && !list_empty(&ep->asocs)) {
4886 struct sctp_association *asoc;
4888 inet_sk_set_state(sk, SCTP_SS_CLOSING);
4889 asoc = list_entry(ep->asocs.next,
4890 struct sctp_association, asocs);
4891 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4895 int sctp_get_sctp_info(struct sock *sk, struct sctp_association *asoc,
4896 struct sctp_info *info)
4898 struct sctp_transport *prim;
4899 struct list_head *pos;
4902 memset(info, 0, sizeof(*info));
4904 struct sctp_sock *sp = sctp_sk(sk);
4906 info->sctpi_s_autoclose = sp->autoclose;
4907 info->sctpi_s_adaptation_ind = sp->adaptation_ind;
4908 info->sctpi_s_pd_point = sp->pd_point;
4909 info->sctpi_s_nodelay = sp->nodelay;
4910 info->sctpi_s_disable_fragments = sp->disable_fragments;
4911 info->sctpi_s_v4mapped = sp->v4mapped;
4912 info->sctpi_s_frag_interleave = sp->frag_interleave;
4913 info->sctpi_s_type = sp->type;
4918 info->sctpi_tag = asoc->c.my_vtag;
4919 info->sctpi_state = asoc->state;
4920 info->sctpi_rwnd = asoc->a_rwnd;
4921 info->sctpi_unackdata = asoc->unack_data;
4922 info->sctpi_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4923 info->sctpi_instrms = asoc->stream.incnt;
4924 info->sctpi_outstrms = asoc->stream.outcnt;
4925 list_for_each(pos, &asoc->base.inqueue.in_chunk_list)
4926 info->sctpi_inqueue++;
4927 list_for_each(pos, &asoc->outqueue.out_chunk_list)
4928 info->sctpi_outqueue++;
4929 info->sctpi_overall_error = asoc->overall_error_count;
4930 info->sctpi_max_burst = asoc->max_burst;
4931 info->sctpi_maxseg = asoc->frag_point;
4932 info->sctpi_peer_rwnd = asoc->peer.rwnd;
4933 info->sctpi_peer_tag = asoc->c.peer_vtag;
4935 mask = asoc->peer.ecn_capable << 1;
4936 mask = (mask | asoc->peer.ipv4_address) << 1;
4937 mask = (mask | asoc->peer.ipv6_address) << 1;
4938 mask = (mask | asoc->peer.hostname_address) << 1;
4939 mask = (mask | asoc->peer.asconf_capable) << 1;
4940 mask = (mask | asoc->peer.prsctp_capable) << 1;
4941 mask = (mask | asoc->peer.auth_capable);
4942 info->sctpi_peer_capable = mask;
4943 mask = asoc->peer.sack_needed << 1;
4944 mask = (mask | asoc->peer.sack_generation) << 1;
4945 mask = (mask | asoc->peer.zero_window_announced);
4946 info->sctpi_peer_sack = mask;
4948 info->sctpi_isacks = asoc->stats.isacks;
4949 info->sctpi_osacks = asoc->stats.osacks;
4950 info->sctpi_opackets = asoc->stats.opackets;
4951 info->sctpi_ipackets = asoc->stats.ipackets;
4952 info->sctpi_rtxchunks = asoc->stats.rtxchunks;
4953 info->sctpi_outofseqtsns = asoc->stats.outofseqtsns;
4954 info->sctpi_idupchunks = asoc->stats.idupchunks;
4955 info->sctpi_gapcnt = asoc->stats.gapcnt;
4956 info->sctpi_ouodchunks = asoc->stats.ouodchunks;
4957 info->sctpi_iuodchunks = asoc->stats.iuodchunks;
4958 info->sctpi_oodchunks = asoc->stats.oodchunks;
4959 info->sctpi_iodchunks = asoc->stats.iodchunks;
4960 info->sctpi_octrlchunks = asoc->stats.octrlchunks;
4961 info->sctpi_ictrlchunks = asoc->stats.ictrlchunks;
4963 prim = asoc->peer.primary_path;
4964 memcpy(&info->sctpi_p_address, &prim->ipaddr, sizeof(prim->ipaddr));
4965 info->sctpi_p_state = prim->state;
4966 info->sctpi_p_cwnd = prim->cwnd;
4967 info->sctpi_p_srtt = prim->srtt;
4968 info->sctpi_p_rto = jiffies_to_msecs(prim->rto);
4969 info->sctpi_p_hbinterval = prim->hbinterval;
4970 info->sctpi_p_pathmaxrxt = prim->pathmaxrxt;
4971 info->sctpi_p_sackdelay = jiffies_to_msecs(prim->sackdelay);
4972 info->sctpi_p_ssthresh = prim->ssthresh;
4973 info->sctpi_p_partial_bytes_acked = prim->partial_bytes_acked;
4974 info->sctpi_p_flight_size = prim->flight_size;
4975 info->sctpi_p_error = prim->error_count;
4979 EXPORT_SYMBOL_GPL(sctp_get_sctp_info);
4981 /* use callback to avoid exporting the core structure */
4982 void sctp_transport_walk_start(struct rhashtable_iter *iter)
4984 rhltable_walk_enter(&sctp_transport_hashtable, iter);
4986 rhashtable_walk_start(iter);
4989 void sctp_transport_walk_stop(struct rhashtable_iter *iter)
4991 rhashtable_walk_stop(iter);
4992 rhashtable_walk_exit(iter);
4995 struct sctp_transport *sctp_transport_get_next(struct net *net,
4996 struct rhashtable_iter *iter)
4998 struct sctp_transport *t;
5000 t = rhashtable_walk_next(iter);
5001 for (; t; t = rhashtable_walk_next(iter)) {
5003 if (PTR_ERR(t) == -EAGAIN)
5008 if (net_eq(sock_net(t->asoc->base.sk), net) &&
5009 t->asoc->peer.primary_path == t)
5016 struct sctp_transport *sctp_transport_get_idx(struct net *net,
5017 struct rhashtable_iter *iter,
5020 void *obj = SEQ_START_TOKEN;
5022 while (pos && (obj = sctp_transport_get_next(net, iter)) &&
5029 int sctp_for_each_endpoint(int (*cb)(struct sctp_endpoint *, void *),
5033 struct sctp_ep_common *epb;
5034 struct sctp_hashbucket *head;
5036 for (head = sctp_ep_hashtable; hash < sctp_ep_hashsize;
5038 read_lock_bh(&head->lock);
5039 sctp_for_each_hentry(epb, &head->chain) {
5040 err = cb(sctp_ep(epb), p);
5044 read_unlock_bh(&head->lock);
5049 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint);
5051 int sctp_transport_lookup_process(int (*cb)(struct sctp_transport *, void *),
5053 const union sctp_addr *laddr,
5054 const union sctp_addr *paddr, void *p)
5056 struct sctp_transport *transport;
5060 transport = sctp_addrs_lookup_transport(net, laddr, paddr);
5065 err = cb(transport, p);
5066 sctp_transport_put(transport);
5070 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process);
5072 int sctp_for_each_transport(int (*cb)(struct sctp_transport *, void *),
5073 int (*cb_done)(struct sctp_transport *, void *),
5074 struct net *net, int *pos, void *p) {
5075 struct rhashtable_iter hti;
5076 struct sctp_transport *tsp;
5081 sctp_transport_walk_start(&hti);
5083 tsp = sctp_transport_get_idx(net, &hti, *pos + 1);
5084 for (; !IS_ERR_OR_NULL(tsp); tsp = sctp_transport_get_next(net, &hti)) {
5085 if (!sctp_transport_hold(tsp))
5091 sctp_transport_put(tsp);
5093 sctp_transport_walk_stop(&hti);
5096 if (cb_done && !cb_done(tsp, p)) {
5098 sctp_transport_put(tsp);
5101 sctp_transport_put(tsp);
5106 EXPORT_SYMBOL_GPL(sctp_for_each_transport);
5108 /* 7.2.1 Association Status (SCTP_STATUS)
5110 * Applications can retrieve current status information about an
5111 * association, including association state, peer receiver window size,
5112 * number of unacked data chunks, and number of data chunks pending
5113 * receipt. This information is read-only.
5115 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
5116 char __user *optval,
5119 struct sctp_status status;
5120 struct sctp_association *asoc = NULL;
5121 struct sctp_transport *transport;
5122 sctp_assoc_t associd;
5125 if (len < sizeof(status)) {
5130 len = sizeof(status);
5131 if (copy_from_user(&status, optval, len)) {
5136 associd = status.sstat_assoc_id;
5137 asoc = sctp_id2assoc(sk, associd);
5143 transport = asoc->peer.primary_path;
5145 status.sstat_assoc_id = sctp_assoc2id(asoc);
5146 status.sstat_state = sctp_assoc_to_state(asoc);
5147 status.sstat_rwnd = asoc->peer.rwnd;
5148 status.sstat_unackdata = asoc->unack_data;
5150 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
5151 status.sstat_instrms = asoc->stream.incnt;
5152 status.sstat_outstrms = asoc->stream.outcnt;
5153 status.sstat_fragmentation_point = asoc->frag_point;
5154 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
5155 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
5156 transport->af_specific->sockaddr_len);
5157 /* Map ipv4 address into v4-mapped-on-v6 address. */
5158 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
5159 (union sctp_addr *)&status.sstat_primary.spinfo_address);
5160 status.sstat_primary.spinfo_state = transport->state;
5161 status.sstat_primary.spinfo_cwnd = transport->cwnd;
5162 status.sstat_primary.spinfo_srtt = transport->srtt;
5163 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
5164 status.sstat_primary.spinfo_mtu = transport->pathmtu;
5166 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
5167 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
5169 if (put_user(len, optlen)) {
5174 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
5175 __func__, len, status.sstat_state, status.sstat_rwnd,
5176 status.sstat_assoc_id);
5178 if (copy_to_user(optval, &status, len)) {
5188 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
5190 * Applications can retrieve information about a specific peer address
5191 * of an association, including its reachability state, congestion
5192 * window, and retransmission timer values. This information is
5195 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
5196 char __user *optval,
5199 struct sctp_paddrinfo pinfo;
5200 struct sctp_transport *transport;
5203 if (len < sizeof(pinfo)) {
5208 len = sizeof(pinfo);
5209 if (copy_from_user(&pinfo, optval, len)) {
5214 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
5215 pinfo.spinfo_assoc_id);
5219 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
5220 pinfo.spinfo_state = transport->state;
5221 pinfo.spinfo_cwnd = transport->cwnd;
5222 pinfo.spinfo_srtt = transport->srtt;
5223 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
5224 pinfo.spinfo_mtu = transport->pathmtu;
5226 if (pinfo.spinfo_state == SCTP_UNKNOWN)
5227 pinfo.spinfo_state = SCTP_ACTIVE;
5229 if (put_user(len, optlen)) {
5234 if (copy_to_user(optval, &pinfo, len)) {
5243 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
5245 * This option is a on/off flag. If enabled no SCTP message
5246 * fragmentation will be performed. Instead if a message being sent
5247 * exceeds the current PMTU size, the message will NOT be sent and
5248 * instead a error will be indicated to the user.
5250 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
5251 char __user *optval, int __user *optlen)
5255 if (len < sizeof(int))
5259 val = (sctp_sk(sk)->disable_fragments == 1);
5260 if (put_user(len, optlen))
5262 if (copy_to_user(optval, &val, len))
5267 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
5269 * This socket option is used to specify various notifications and
5270 * ancillary data the user wishes to receive.
5272 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
5277 if (len > sizeof(struct sctp_event_subscribe))
5278 len = sizeof(struct sctp_event_subscribe);
5279 if (put_user(len, optlen))
5281 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
5286 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
5288 * This socket option is applicable to the UDP-style socket only. When
5289 * set it will cause associations that are idle for more than the
5290 * specified number of seconds to automatically close. An association
5291 * being idle is defined an association that has NOT sent or received
5292 * user data. The special value of '0' indicates that no automatic
5293 * close of any associations should be performed. The option expects an
5294 * integer defining the number of seconds of idle time before an
5295 * association is closed.
5297 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
5299 /* Applicable to UDP-style socket only */
5300 if (sctp_style(sk, TCP))
5302 if (len < sizeof(int))
5305 if (put_user(len, optlen))
5307 if (put_user(sctp_sk(sk)->autoclose, (int __user *)optval))
5312 /* Helper routine to branch off an association to a new socket. */
5313 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
5315 struct sctp_association *asoc = sctp_id2assoc(sk, id);
5316 struct sctp_sock *sp = sctp_sk(sk);
5317 struct socket *sock;
5320 /* Do not peel off from one netns to another one. */
5321 if (!net_eq(current->nsproxy->net_ns, sock_net(sk)))
5327 /* An association cannot be branched off from an already peeled-off
5328 * socket, nor is this supported for tcp style sockets.
5330 if (!sctp_style(sk, UDP))
5333 /* Create a new socket. */
5334 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
5338 sctp_copy_sock(sock->sk, sk, asoc);
5340 /* Make peeled-off sockets more like 1-1 accepted sockets.
5341 * Set the daddr and initialize id to something more random and also
5342 * copy over any ip options.
5344 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
5345 sp->pf->copy_ip_options(sk, sock->sk);
5347 /* Populate the fields of the newsk from the oldsk and migrate the
5348 * asoc to the newsk.
5350 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
5356 EXPORT_SYMBOL(sctp_do_peeloff);
5358 static int sctp_getsockopt_peeloff_common(struct sock *sk, sctp_peeloff_arg_t *peeloff,
5359 struct file **newfile, unsigned flags)
5361 struct socket *newsock;
5364 retval = sctp_do_peeloff(sk, peeloff->associd, &newsock);
5368 /* Map the socket to an unused fd that can be returned to the user. */
5369 retval = get_unused_fd_flags(flags & SOCK_CLOEXEC);
5371 sock_release(newsock);
5375 *newfile = sock_alloc_file(newsock, 0, NULL);
5376 if (IS_ERR(*newfile)) {
5377 put_unused_fd(retval);
5378 retval = PTR_ERR(*newfile);
5383 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
5386 peeloff->sd = retval;
5388 if (flags & SOCK_NONBLOCK)
5389 (*newfile)->f_flags |= O_NONBLOCK;
5394 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
5396 sctp_peeloff_arg_t peeloff;
5397 struct file *newfile = NULL;
5400 if (len < sizeof(sctp_peeloff_arg_t))
5402 len = sizeof(sctp_peeloff_arg_t);
5403 if (copy_from_user(&peeloff, optval, len))
5406 retval = sctp_getsockopt_peeloff_common(sk, &peeloff, &newfile, 0);
5410 /* Return the fd mapped to the new socket. */
5411 if (put_user(len, optlen)) {
5413 put_unused_fd(retval);
5417 if (copy_to_user(optval, &peeloff, len)) {
5419 put_unused_fd(retval);
5422 fd_install(retval, newfile);
5427 static int sctp_getsockopt_peeloff_flags(struct sock *sk, int len,
5428 char __user *optval, int __user *optlen)
5430 sctp_peeloff_flags_arg_t peeloff;
5431 struct file *newfile = NULL;
5434 if (len < sizeof(sctp_peeloff_flags_arg_t))
5436 len = sizeof(sctp_peeloff_flags_arg_t);
5437 if (copy_from_user(&peeloff, optval, len))
5440 retval = sctp_getsockopt_peeloff_common(sk, &peeloff.p_arg,
5441 &newfile, peeloff.flags);
5445 /* Return the fd mapped to the new socket. */
5446 if (put_user(len, optlen)) {
5448 put_unused_fd(retval);
5452 if (copy_to_user(optval, &peeloff, len)) {
5454 put_unused_fd(retval);
5457 fd_install(retval, newfile);
5462 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
5464 * Applications can enable or disable heartbeats for any peer address of
5465 * an association, modify an address's heartbeat interval, force a
5466 * heartbeat to be sent immediately, and adjust the address's maximum
5467 * number of retransmissions sent before an address is considered
5468 * unreachable. The following structure is used to access and modify an
5469 * address's parameters:
5471 * struct sctp_paddrparams {
5472 * sctp_assoc_t spp_assoc_id;
5473 * struct sockaddr_storage spp_address;
5474 * uint32_t spp_hbinterval;
5475 * uint16_t spp_pathmaxrxt;
5476 * uint32_t spp_pathmtu;
5477 * uint32_t spp_sackdelay;
5478 * uint32_t spp_flags;
5481 * spp_assoc_id - (one-to-many style socket) This is filled in the
5482 * application, and identifies the association for
5484 * spp_address - This specifies which address is of interest.
5485 * spp_hbinterval - This contains the value of the heartbeat interval,
5486 * in milliseconds. If a value of zero
5487 * is present in this field then no changes are to
5488 * be made to this parameter.
5489 * spp_pathmaxrxt - This contains the maximum number of
5490 * retransmissions before this address shall be
5491 * considered unreachable. If a value of zero
5492 * is present in this field then no changes are to
5493 * be made to this parameter.
5494 * spp_pathmtu - When Path MTU discovery is disabled the value
5495 * specified here will be the "fixed" path mtu.
5496 * Note that if the spp_address field is empty
5497 * then all associations on this address will
5498 * have this fixed path mtu set upon them.
5500 * spp_sackdelay - When delayed sack is enabled, this value specifies
5501 * the number of milliseconds that sacks will be delayed
5502 * for. This value will apply to all addresses of an
5503 * association if the spp_address field is empty. Note
5504 * also, that if delayed sack is enabled and this
5505 * value is set to 0, no change is made to the last
5506 * recorded delayed sack timer value.
5508 * spp_flags - These flags are used to control various features
5509 * on an association. The flag field may contain
5510 * zero or more of the following options.
5512 * SPP_HB_ENABLE - Enable heartbeats on the
5513 * specified address. Note that if the address
5514 * field is empty all addresses for the association
5515 * have heartbeats enabled upon them.
5517 * SPP_HB_DISABLE - Disable heartbeats on the
5518 * speicifed address. Note that if the address
5519 * field is empty all addresses for the association
5520 * will have their heartbeats disabled. Note also
5521 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
5522 * mutually exclusive, only one of these two should
5523 * be specified. Enabling both fields will have
5524 * undetermined results.
5526 * SPP_HB_DEMAND - Request a user initiated heartbeat
5527 * to be made immediately.
5529 * SPP_PMTUD_ENABLE - This field will enable PMTU
5530 * discovery upon the specified address. Note that
5531 * if the address feild is empty then all addresses
5532 * on the association are effected.
5534 * SPP_PMTUD_DISABLE - This field will disable PMTU
5535 * discovery upon the specified address. Note that
5536 * if the address feild is empty then all addresses
5537 * on the association are effected. Not also that
5538 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
5539 * exclusive. Enabling both will have undetermined
5542 * SPP_SACKDELAY_ENABLE - Setting this flag turns
5543 * on delayed sack. The time specified in spp_sackdelay
5544 * is used to specify the sack delay for this address. Note
5545 * that if spp_address is empty then all addresses will
5546 * enable delayed sack and take on the sack delay
5547 * value specified in spp_sackdelay.
5548 * SPP_SACKDELAY_DISABLE - Setting this flag turns
5549 * off delayed sack. If the spp_address field is blank then
5550 * delayed sack is disabled for the entire association. Note
5551 * also that this field is mutually exclusive to
5552 * SPP_SACKDELAY_ENABLE, setting both will have undefined
5555 * SPP_IPV6_FLOWLABEL: Setting this flag enables the
5556 * setting of the IPV6 flow label value. The value is
5557 * contained in the spp_ipv6_flowlabel field.
5558 * Upon retrieval, this flag will be set to indicate that
5559 * the spp_ipv6_flowlabel field has a valid value returned.
5560 * If a specific destination address is set (in the
5561 * spp_address field), then the value returned is that of
5562 * the address. If just an association is specified (and
5563 * no address), then the association's default flow label
5564 * is returned. If neither an association nor a destination
5565 * is specified, then the socket's default flow label is
5566 * returned. For non-IPv6 sockets, this flag will be left
5569 * SPP_DSCP: Setting this flag enables the setting of the
5570 * Differentiated Services Code Point (DSCP) value
5571 * associated with either the association or a specific
5572 * address. The value is obtained in the spp_dscp field.
5573 * Upon retrieval, this flag will be set to indicate that
5574 * the spp_dscp field has a valid value returned. If a
5575 * specific destination address is set when called (in the
5576 * spp_address field), then that specific destination
5577 * address's DSCP value is returned. If just an association
5578 * is specified, then the association's default DSCP is
5579 * returned. If neither an association nor a destination is
5580 * specified, then the socket's default DSCP is returned.
5582 * spp_ipv6_flowlabel
5583 * - This field is used in conjunction with the
5584 * SPP_IPV6_FLOWLABEL flag and contains the IPv6 flow label.
5585 * The 20 least significant bits are used for the flow
5586 * label. This setting has precedence over any IPv6-layer
5589 * spp_dscp - This field is used in conjunction with the SPP_DSCP flag
5590 * and contains the DSCP. The 6 most significant bits are
5591 * used for the DSCP. This setting has precedence over any
5592 * IPv4- or IPv6- layer setting.
5594 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
5595 char __user *optval, int __user *optlen)
5597 struct sctp_paddrparams params;
5598 struct sctp_transport *trans = NULL;
5599 struct sctp_association *asoc = NULL;
5600 struct sctp_sock *sp = sctp_sk(sk);
5602 if (len >= sizeof(params))
5603 len = sizeof(params);
5604 else if (len >= ALIGN(offsetof(struct sctp_paddrparams,
5605 spp_ipv6_flowlabel), 4))
5606 len = ALIGN(offsetof(struct sctp_paddrparams,
5607 spp_ipv6_flowlabel), 4);
5611 if (copy_from_user(¶ms, optval, len))
5614 /* If an address other than INADDR_ANY is specified, and
5615 * no transport is found, then the request is invalid.
5617 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
5618 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
5619 params.spp_assoc_id);
5621 pr_debug("%s: failed no transport\n", __func__);
5626 /* Get association, if assoc_id != 0 and the socket is a one
5627 * to many style socket, and an association was not found, then
5628 * the id was invalid.
5630 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
5631 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
5632 pr_debug("%s: failed no association\n", __func__);
5637 /* Fetch transport values. */
5638 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
5639 params.spp_pathmtu = trans->pathmtu;
5640 params.spp_pathmaxrxt = trans->pathmaxrxt;
5641 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
5643 /*draft-11 doesn't say what to return in spp_flags*/
5644 params.spp_flags = trans->param_flags;
5645 if (trans->flowlabel & SCTP_FLOWLABEL_SET_MASK) {
5646 params.spp_ipv6_flowlabel = trans->flowlabel &
5647 SCTP_FLOWLABEL_VAL_MASK;
5648 params.spp_flags |= SPP_IPV6_FLOWLABEL;
5650 if (trans->dscp & SCTP_DSCP_SET_MASK) {
5651 params.spp_dscp = trans->dscp & SCTP_DSCP_VAL_MASK;
5652 params.spp_flags |= SPP_DSCP;
5655 /* Fetch association values. */
5656 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
5657 params.spp_pathmtu = asoc->pathmtu;
5658 params.spp_pathmaxrxt = asoc->pathmaxrxt;
5659 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
5661 /*draft-11 doesn't say what to return in spp_flags*/
5662 params.spp_flags = asoc->param_flags;
5663 if (asoc->flowlabel & SCTP_FLOWLABEL_SET_MASK) {
5664 params.spp_ipv6_flowlabel = asoc->flowlabel &
5665 SCTP_FLOWLABEL_VAL_MASK;
5666 params.spp_flags |= SPP_IPV6_FLOWLABEL;
5668 if (asoc->dscp & SCTP_DSCP_SET_MASK) {
5669 params.spp_dscp = asoc->dscp & SCTP_DSCP_VAL_MASK;
5670 params.spp_flags |= SPP_DSCP;
5673 /* Fetch socket values. */
5674 params.spp_hbinterval = sp->hbinterval;
5675 params.spp_pathmtu = sp->pathmtu;
5676 params.spp_sackdelay = sp->sackdelay;
5677 params.spp_pathmaxrxt = sp->pathmaxrxt;
5679 /*draft-11 doesn't say what to return in spp_flags*/
5680 params.spp_flags = sp->param_flags;
5681 if (sp->flowlabel & SCTP_FLOWLABEL_SET_MASK) {
5682 params.spp_ipv6_flowlabel = sp->flowlabel &
5683 SCTP_FLOWLABEL_VAL_MASK;
5684 params.spp_flags |= SPP_IPV6_FLOWLABEL;
5686 if (sp->dscp & SCTP_DSCP_SET_MASK) {
5687 params.spp_dscp = sp->dscp & SCTP_DSCP_VAL_MASK;
5688 params.spp_flags |= SPP_DSCP;
5692 if (copy_to_user(optval, ¶ms, len))
5695 if (put_user(len, optlen))
5702 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
5704 * This option will effect the way delayed acks are performed. This
5705 * option allows you to get or set the delayed ack time, in
5706 * milliseconds. It also allows changing the delayed ack frequency.
5707 * Changing the frequency to 1 disables the delayed sack algorithm. If
5708 * the assoc_id is 0, then this sets or gets the endpoints default
5709 * values. If the assoc_id field is non-zero, then the set or get
5710 * effects the specified association for the one to many model (the
5711 * assoc_id field is ignored by the one to one model). Note that if
5712 * sack_delay or sack_freq are 0 when setting this option, then the
5713 * current values will remain unchanged.
5715 * struct sctp_sack_info {
5716 * sctp_assoc_t sack_assoc_id;
5717 * uint32_t sack_delay;
5718 * uint32_t sack_freq;
5721 * sack_assoc_id - This parameter, indicates which association the user
5722 * is performing an action upon. Note that if this field's value is
5723 * zero then the endpoints default value is changed (effecting future
5724 * associations only).
5726 * sack_delay - This parameter contains the number of milliseconds that
5727 * the user is requesting the delayed ACK timer be set to. Note that
5728 * this value is defined in the standard to be between 200 and 500
5731 * sack_freq - This parameter contains the number of packets that must
5732 * be received before a sack is sent without waiting for the delay
5733 * timer to expire. The default value for this is 2, setting this
5734 * value to 1 will disable the delayed sack algorithm.
5736 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
5737 char __user *optval,
5740 struct sctp_sack_info params;
5741 struct sctp_association *asoc = NULL;
5742 struct sctp_sock *sp = sctp_sk(sk);
5744 if (len >= sizeof(struct sctp_sack_info)) {
5745 len = sizeof(struct sctp_sack_info);
5747 if (copy_from_user(¶ms, optval, len))
5749 } else if (len == sizeof(struct sctp_assoc_value)) {
5750 pr_warn_ratelimited(DEPRECATED
5752 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
5753 "Use struct sctp_sack_info instead\n",
5754 current->comm, task_pid_nr(current));
5755 if (copy_from_user(¶ms, optval, len))
5760 /* Get association, if sack_assoc_id != 0 and the socket is a one
5761 * to many style socket, and an association was not found, then
5762 * the id was invalid.
5764 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
5765 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
5769 /* Fetch association values. */
5770 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
5771 params.sack_delay = jiffies_to_msecs(
5773 params.sack_freq = asoc->sackfreq;
5776 params.sack_delay = 0;
5777 params.sack_freq = 1;
5780 /* Fetch socket values. */
5781 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
5782 params.sack_delay = sp->sackdelay;
5783 params.sack_freq = sp->sackfreq;
5785 params.sack_delay = 0;
5786 params.sack_freq = 1;
5790 if (copy_to_user(optval, ¶ms, len))
5793 if (put_user(len, optlen))
5799 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5801 * Applications can specify protocol parameters for the default association
5802 * initialization. The option name argument to setsockopt() and getsockopt()
5805 * Setting initialization parameters is effective only on an unconnected
5806 * socket (for UDP-style sockets only future associations are effected
5807 * by the change). With TCP-style sockets, this option is inherited by
5808 * sockets derived from a listener socket.
5810 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
5812 if (len < sizeof(struct sctp_initmsg))
5814 len = sizeof(struct sctp_initmsg);
5815 if (put_user(len, optlen))
5817 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
5823 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
5824 char __user *optval, int __user *optlen)
5826 struct sctp_association *asoc;
5828 struct sctp_getaddrs getaddrs;
5829 struct sctp_transport *from;
5831 union sctp_addr temp;
5832 struct sctp_sock *sp = sctp_sk(sk);
5837 if (len < sizeof(struct sctp_getaddrs))
5840 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5843 /* For UDP-style sockets, id specifies the association to query. */
5844 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5848 to = optval + offsetof(struct sctp_getaddrs, addrs);
5849 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5851 list_for_each_entry(from, &asoc->peer.transport_addr_list,
5853 memcpy(&temp, &from->ipaddr, sizeof(temp));
5854 addrlen = sctp_get_pf_specific(sk->sk_family)
5855 ->addr_to_user(sp, &temp);
5856 if (space_left < addrlen)
5858 if (copy_to_user(to, &temp, addrlen))
5862 space_left -= addrlen;
5865 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
5867 bytes_copied = ((char __user *)to) - optval;
5868 if (put_user(bytes_copied, optlen))
5874 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
5875 size_t space_left, int *bytes_copied)
5877 struct sctp_sockaddr_entry *addr;
5878 union sctp_addr temp;
5881 struct net *net = sock_net(sk);
5884 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
5888 if ((PF_INET == sk->sk_family) &&
5889 (AF_INET6 == addr->a.sa.sa_family))
5891 if ((PF_INET6 == sk->sk_family) &&
5892 inet_v6_ipv6only(sk) &&
5893 (AF_INET == addr->a.sa.sa_family))
5895 memcpy(&temp, &addr->a, sizeof(temp));
5896 if (!temp.v4.sin_port)
5897 temp.v4.sin_port = htons(port);
5899 addrlen = sctp_get_pf_specific(sk->sk_family)
5900 ->addr_to_user(sctp_sk(sk), &temp);
5902 if (space_left < addrlen) {
5906 memcpy(to, &temp, addrlen);
5910 space_left -= addrlen;
5911 *bytes_copied += addrlen;
5919 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
5920 char __user *optval, int __user *optlen)
5922 struct sctp_bind_addr *bp;
5923 struct sctp_association *asoc;
5925 struct sctp_getaddrs getaddrs;
5926 struct sctp_sockaddr_entry *addr;
5928 union sctp_addr temp;
5929 struct sctp_sock *sp = sctp_sk(sk);
5933 int bytes_copied = 0;
5937 if (len < sizeof(struct sctp_getaddrs))
5940 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5944 * For UDP-style sockets, id specifies the association to query.
5945 * If the id field is set to the value '0' then the locally bound
5946 * addresses are returned without regard to any particular
5949 if (0 == getaddrs.assoc_id) {
5950 bp = &sctp_sk(sk)->ep->base.bind_addr;
5952 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5955 bp = &asoc->base.bind_addr;
5958 to = optval + offsetof(struct sctp_getaddrs, addrs);
5959 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5961 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN);
5965 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5966 * addresses from the global local address list.
5968 if (sctp_list_single_entry(&bp->address_list)) {
5969 addr = list_entry(bp->address_list.next,
5970 struct sctp_sockaddr_entry, list);
5971 if (sctp_is_any(sk, &addr->a)) {
5972 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
5973 space_left, &bytes_copied);
5983 /* Protection on the bound address list is not needed since
5984 * in the socket option context we hold a socket lock and
5985 * thus the bound address list can't change.
5987 list_for_each_entry(addr, &bp->address_list, list) {
5988 memcpy(&temp, &addr->a, sizeof(temp));
5989 addrlen = sctp_get_pf_specific(sk->sk_family)
5990 ->addr_to_user(sp, &temp);
5991 if (space_left < addrlen) {
5992 err = -ENOMEM; /*fixme: right error?*/
5995 memcpy(buf, &temp, addrlen);
5997 bytes_copied += addrlen;
5999 space_left -= addrlen;
6003 if (copy_to_user(to, addrs, bytes_copied)) {
6007 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
6011 /* XXX: We should have accounted for sizeof(struct sctp_getaddrs) too,
6012 * but we can't change it anymore.
6014 if (put_user(bytes_copied, optlen))
6021 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
6023 * Requests that the local SCTP stack use the enclosed peer address as
6024 * the association primary. The enclosed address must be one of the
6025 * association peer's addresses.
6027 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
6028 char __user *optval, int __user *optlen)
6030 struct sctp_prim prim;
6031 struct sctp_association *asoc;
6032 struct sctp_sock *sp = sctp_sk(sk);
6034 if (len < sizeof(struct sctp_prim))
6037 len = sizeof(struct sctp_prim);
6039 if (copy_from_user(&prim, optval, len))
6042 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
6046 if (!asoc->peer.primary_path)
6049 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
6050 asoc->peer.primary_path->af_specific->sockaddr_len);
6052 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
6053 (union sctp_addr *)&prim.ssp_addr);
6055 if (put_user(len, optlen))
6057 if (copy_to_user(optval, &prim, len))
6064 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
6066 * Requests that the local endpoint set the specified Adaptation Layer
6067 * Indication parameter for all future INIT and INIT-ACK exchanges.
6069 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
6070 char __user *optval, int __user *optlen)
6072 struct sctp_setadaptation adaptation;
6074 if (len < sizeof(struct sctp_setadaptation))
6077 len = sizeof(struct sctp_setadaptation);
6079 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
6081 if (put_user(len, optlen))
6083 if (copy_to_user(optval, &adaptation, len))
6091 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
6093 * Applications that wish to use the sendto() system call may wish to
6094 * specify a default set of parameters that would normally be supplied
6095 * through the inclusion of ancillary data. This socket option allows
6096 * such an application to set the default sctp_sndrcvinfo structure.
6099 * The application that wishes to use this socket option simply passes
6100 * in to this call the sctp_sndrcvinfo structure defined in Section
6101 * 5.2.2) The input parameters accepted by this call include
6102 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
6103 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
6104 * to this call if the caller is using the UDP model.
6106 * For getsockopt, it get the default sctp_sndrcvinfo structure.
6108 static int sctp_getsockopt_default_send_param(struct sock *sk,
6109 int len, char __user *optval,
6112 struct sctp_sock *sp = sctp_sk(sk);
6113 struct sctp_association *asoc;
6114 struct sctp_sndrcvinfo info;
6116 if (len < sizeof(info))
6121 if (copy_from_user(&info, optval, len))
6124 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
6125 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
6128 info.sinfo_stream = asoc->default_stream;
6129 info.sinfo_flags = asoc->default_flags;
6130 info.sinfo_ppid = asoc->default_ppid;
6131 info.sinfo_context = asoc->default_context;
6132 info.sinfo_timetolive = asoc->default_timetolive;
6134 info.sinfo_stream = sp->default_stream;
6135 info.sinfo_flags = sp->default_flags;
6136 info.sinfo_ppid = sp->default_ppid;
6137 info.sinfo_context = sp->default_context;
6138 info.sinfo_timetolive = sp->default_timetolive;
6141 if (put_user(len, optlen))
6143 if (copy_to_user(optval, &info, len))
6149 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
6150 * (SCTP_DEFAULT_SNDINFO)
6152 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
6153 char __user *optval,
6156 struct sctp_sock *sp = sctp_sk(sk);
6157 struct sctp_association *asoc;
6158 struct sctp_sndinfo info;
6160 if (len < sizeof(info))
6165 if (copy_from_user(&info, optval, len))
6168 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
6169 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
6172 info.snd_sid = asoc->default_stream;
6173 info.snd_flags = asoc->default_flags;
6174 info.snd_ppid = asoc->default_ppid;
6175 info.snd_context = asoc->default_context;
6177 info.snd_sid = sp->default_stream;
6178 info.snd_flags = sp->default_flags;
6179 info.snd_ppid = sp->default_ppid;
6180 info.snd_context = sp->default_context;
6183 if (put_user(len, optlen))
6185 if (copy_to_user(optval, &info, len))
6193 * 7.1.5 SCTP_NODELAY
6195 * Turn on/off any Nagle-like algorithm. This means that packets are
6196 * generally sent as soon as possible and no unnecessary delays are
6197 * introduced, at the cost of more packets in the network. Expects an
6198 * integer boolean flag.
6201 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
6202 char __user *optval, int __user *optlen)
6206 if (len < sizeof(int))
6210 val = (sctp_sk(sk)->nodelay == 1);
6211 if (put_user(len, optlen))
6213 if (copy_to_user(optval, &val, len))
6220 * 7.1.1 SCTP_RTOINFO
6222 * The protocol parameters used to initialize and bound retransmission
6223 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
6224 * and modify these parameters.
6225 * All parameters are time values, in milliseconds. A value of 0, when
6226 * modifying the parameters, indicates that the current value should not
6230 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
6231 char __user *optval,
6232 int __user *optlen) {
6233 struct sctp_rtoinfo rtoinfo;
6234 struct sctp_association *asoc;
6236 if (len < sizeof (struct sctp_rtoinfo))
6239 len = sizeof(struct sctp_rtoinfo);
6241 if (copy_from_user(&rtoinfo, optval, len))
6244 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
6246 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
6249 /* Values corresponding to the specific association. */
6251 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
6252 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
6253 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
6255 /* Values corresponding to the endpoint. */
6256 struct sctp_sock *sp = sctp_sk(sk);
6258 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
6259 rtoinfo.srto_max = sp->rtoinfo.srto_max;
6260 rtoinfo.srto_min = sp->rtoinfo.srto_min;
6263 if (put_user(len, optlen))
6266 if (copy_to_user(optval, &rtoinfo, len))
6274 * 7.1.2 SCTP_ASSOCINFO
6276 * This option is used to tune the maximum retransmission attempts
6277 * of the association.
6278 * Returns an error if the new association retransmission value is
6279 * greater than the sum of the retransmission value of the peer.
6280 * See [SCTP] for more information.
6283 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
6284 char __user *optval,
6288 struct sctp_assocparams assocparams;
6289 struct sctp_association *asoc;
6290 struct list_head *pos;
6293 if (len < sizeof (struct sctp_assocparams))
6296 len = sizeof(struct sctp_assocparams);
6298 if (copy_from_user(&assocparams, optval, len))
6301 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
6303 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
6306 /* Values correspoinding to the specific association */
6308 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
6309 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
6310 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
6311 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
6313 list_for_each(pos, &asoc->peer.transport_addr_list) {
6317 assocparams.sasoc_number_peer_destinations = cnt;
6319 /* Values corresponding to the endpoint */
6320 struct sctp_sock *sp = sctp_sk(sk);
6322 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
6323 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
6324 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
6325 assocparams.sasoc_cookie_life =
6326 sp->assocparams.sasoc_cookie_life;
6327 assocparams.sasoc_number_peer_destinations =
6329 sasoc_number_peer_destinations;
6332 if (put_user(len, optlen))
6335 if (copy_to_user(optval, &assocparams, len))
6342 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
6344 * This socket option is a boolean flag which turns on or off mapped V4
6345 * addresses. If this option is turned on and the socket is type
6346 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
6347 * If this option is turned off, then no mapping will be done of V4
6348 * addresses and a user will receive both PF_INET6 and PF_INET type
6349 * addresses on the socket.
6351 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
6352 char __user *optval, int __user *optlen)
6355 struct sctp_sock *sp = sctp_sk(sk);
6357 if (len < sizeof(int))
6362 if (put_user(len, optlen))
6364 if (copy_to_user(optval, &val, len))
6371 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
6372 * (chapter and verse is quoted at sctp_setsockopt_context())
6374 static int sctp_getsockopt_context(struct sock *sk, int len,
6375 char __user *optval, int __user *optlen)
6377 struct sctp_assoc_value params;
6378 struct sctp_sock *sp;
6379 struct sctp_association *asoc;
6381 if (len < sizeof(struct sctp_assoc_value))
6384 len = sizeof(struct sctp_assoc_value);
6386 if (copy_from_user(¶ms, optval, len))
6391 if (params.assoc_id != 0) {
6392 asoc = sctp_id2assoc(sk, params.assoc_id);
6395 params.assoc_value = asoc->default_rcv_context;
6397 params.assoc_value = sp->default_rcv_context;
6400 if (put_user(len, optlen))
6402 if (copy_to_user(optval, ¶ms, len))
6409 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
6410 * This option will get or set the maximum size to put in any outgoing
6411 * SCTP DATA chunk. If a message is larger than this size it will be
6412 * fragmented by SCTP into the specified size. Note that the underlying
6413 * SCTP implementation may fragment into smaller sized chunks when the
6414 * PMTU of the underlying association is smaller than the value set by
6415 * the user. The default value for this option is '0' which indicates
6416 * the user is NOT limiting fragmentation and only the PMTU will effect
6417 * SCTP's choice of DATA chunk size. Note also that values set larger
6418 * than the maximum size of an IP datagram will effectively let SCTP
6419 * control fragmentation (i.e. the same as setting this option to 0).
6421 * The following structure is used to access and modify this parameter:
6423 * struct sctp_assoc_value {
6424 * sctp_assoc_t assoc_id;
6425 * uint32_t assoc_value;
6428 * assoc_id: This parameter is ignored for one-to-one style sockets.
6429 * For one-to-many style sockets this parameter indicates which
6430 * association the user is performing an action upon. Note that if
6431 * this field's value is zero then the endpoints default value is
6432 * changed (effecting future associations only).
6433 * assoc_value: This parameter specifies the maximum size in bytes.
6435 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
6436 char __user *optval, int __user *optlen)
6438 struct sctp_assoc_value params;
6439 struct sctp_association *asoc;
6441 if (len == sizeof(int)) {
6442 pr_warn_ratelimited(DEPRECATED
6444 "Use of int in maxseg socket option.\n"
6445 "Use struct sctp_assoc_value instead\n",
6446 current->comm, task_pid_nr(current));
6447 params.assoc_id = 0;
6448 } else if (len >= sizeof(struct sctp_assoc_value)) {
6449 len = sizeof(struct sctp_assoc_value);
6450 if (copy_from_user(¶ms, optval, len))
6455 asoc = sctp_id2assoc(sk, params.assoc_id);
6456 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
6460 params.assoc_value = asoc->frag_point;
6462 params.assoc_value = sctp_sk(sk)->user_frag;
6464 if (put_user(len, optlen))
6466 if (len == sizeof(int)) {
6467 if (copy_to_user(optval, ¶ms.assoc_value, len))
6470 if (copy_to_user(optval, ¶ms, len))
6478 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
6479 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
6481 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
6482 char __user *optval, int __user *optlen)
6486 if (len < sizeof(int))
6491 val = sctp_sk(sk)->frag_interleave;
6492 if (put_user(len, optlen))
6494 if (copy_to_user(optval, &val, len))
6501 * 7.1.25. Set or Get the sctp partial delivery point
6502 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
6504 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
6505 char __user *optval,
6510 if (len < sizeof(u32))
6515 val = sctp_sk(sk)->pd_point;
6516 if (put_user(len, optlen))
6518 if (copy_to_user(optval, &val, len))
6525 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
6526 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
6528 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
6529 char __user *optval,
6532 struct sctp_assoc_value params;
6533 struct sctp_sock *sp;
6534 struct sctp_association *asoc;
6536 if (len == sizeof(int)) {
6537 pr_warn_ratelimited(DEPRECATED
6539 "Use of int in max_burst socket option.\n"
6540 "Use struct sctp_assoc_value instead\n",
6541 current->comm, task_pid_nr(current));
6542 params.assoc_id = 0;
6543 } else if (len >= sizeof(struct sctp_assoc_value)) {
6544 len = sizeof(struct sctp_assoc_value);
6545 if (copy_from_user(¶ms, optval, len))
6552 if (params.assoc_id != 0) {
6553 asoc = sctp_id2assoc(sk, params.assoc_id);
6556 params.assoc_value = asoc->max_burst;
6558 params.assoc_value = sp->max_burst;
6560 if (len == sizeof(int)) {
6561 if (copy_to_user(optval, ¶ms.assoc_value, len))
6564 if (copy_to_user(optval, ¶ms, len))
6572 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
6573 char __user *optval, int __user *optlen)
6575 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6576 struct sctp_hmacalgo __user *p = (void __user *)optval;
6577 struct sctp_hmac_algo_param *hmacs;
6582 if (!ep->auth_enable)
6585 hmacs = ep->auth_hmacs_list;
6586 data_len = ntohs(hmacs->param_hdr.length) -
6587 sizeof(struct sctp_paramhdr);
6589 if (len < sizeof(struct sctp_hmacalgo) + data_len)
6592 len = sizeof(struct sctp_hmacalgo) + data_len;
6593 num_idents = data_len / sizeof(u16);
6595 if (put_user(len, optlen))
6597 if (put_user(num_idents, &p->shmac_num_idents))
6599 for (i = 0; i < num_idents; i++) {
6600 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
6602 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
6608 static int sctp_getsockopt_active_key(struct sock *sk, int len,
6609 char __user *optval, int __user *optlen)
6611 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6612 struct sctp_authkeyid val;
6613 struct sctp_association *asoc;
6615 if (!ep->auth_enable)
6618 if (len < sizeof(struct sctp_authkeyid))
6621 len = sizeof(struct sctp_authkeyid);
6622 if (copy_from_user(&val, optval, len))
6625 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
6626 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
6630 val.scact_keynumber = asoc->active_key_id;
6632 val.scact_keynumber = ep->active_key_id;
6634 if (put_user(len, optlen))
6636 if (copy_to_user(optval, &val, len))
6642 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
6643 char __user *optval, int __user *optlen)
6645 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6646 struct sctp_authchunks __user *p = (void __user *)optval;
6647 struct sctp_authchunks val;
6648 struct sctp_association *asoc;
6649 struct sctp_chunks_param *ch;
6653 if (!ep->auth_enable)
6656 if (len < sizeof(struct sctp_authchunks))
6659 if (copy_from_user(&val, optval, sizeof(val)))
6662 to = p->gauth_chunks;
6663 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6667 ch = asoc->peer.peer_chunks;
6671 /* See if the user provided enough room for all the data */
6672 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6673 if (len < num_chunks)
6676 if (copy_to_user(to, ch->chunks, num_chunks))
6679 len = sizeof(struct sctp_authchunks) + num_chunks;
6680 if (put_user(len, optlen))
6682 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6687 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
6688 char __user *optval, int __user *optlen)
6690 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6691 struct sctp_authchunks __user *p = (void __user *)optval;
6692 struct sctp_authchunks val;
6693 struct sctp_association *asoc;
6694 struct sctp_chunks_param *ch;
6698 if (!ep->auth_enable)
6701 if (len < sizeof(struct sctp_authchunks))
6704 if (copy_from_user(&val, optval, sizeof(val)))
6707 to = p->gauth_chunks;
6708 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6709 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
6713 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
6715 ch = ep->auth_chunk_list;
6720 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6721 if (len < sizeof(struct sctp_authchunks) + num_chunks)
6724 if (copy_to_user(to, ch->chunks, num_chunks))
6727 len = sizeof(struct sctp_authchunks) + num_chunks;
6728 if (put_user(len, optlen))
6730 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6737 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
6738 * This option gets the current number of associations that are attached
6739 * to a one-to-many style socket. The option value is an uint32_t.
6741 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
6742 char __user *optval, int __user *optlen)
6744 struct sctp_sock *sp = sctp_sk(sk);
6745 struct sctp_association *asoc;
6748 if (sctp_style(sk, TCP))
6751 if (len < sizeof(u32))
6756 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6760 if (put_user(len, optlen))
6762 if (copy_to_user(optval, &val, len))
6769 * 8.1.23 SCTP_AUTO_ASCONF
6770 * See the corresponding setsockopt entry as description
6772 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
6773 char __user *optval, int __user *optlen)
6777 if (len < sizeof(int))
6781 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
6783 if (put_user(len, optlen))
6785 if (copy_to_user(optval, &val, len))
6791 * 8.2.6. Get the Current Identifiers of Associations
6792 * (SCTP_GET_ASSOC_ID_LIST)
6794 * This option gets the current list of SCTP association identifiers of
6795 * the SCTP associations handled by a one-to-many style socket.
6797 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
6798 char __user *optval, int __user *optlen)
6800 struct sctp_sock *sp = sctp_sk(sk);
6801 struct sctp_association *asoc;
6802 struct sctp_assoc_ids *ids;
6805 if (sctp_style(sk, TCP))
6808 if (len < sizeof(struct sctp_assoc_ids))
6811 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6815 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
6818 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
6820 ids = kmalloc(len, GFP_USER | __GFP_NOWARN);
6824 ids->gaids_number_of_ids = num;
6826 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6827 ids->gaids_assoc_id[num++] = asoc->assoc_id;
6830 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
6840 * SCTP_PEER_ADDR_THLDS
6842 * This option allows us to fetch the partially failed threshold for one or all
6843 * transports in an association. See Section 6.1 of:
6844 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6846 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
6847 char __user *optval,
6851 struct sctp_paddrthlds val;
6852 struct sctp_transport *trans;
6853 struct sctp_association *asoc;
6855 if (len < sizeof(struct sctp_paddrthlds))
6857 len = sizeof(struct sctp_paddrthlds);
6858 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
6861 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
6862 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
6866 val.spt_pathpfthld = asoc->pf_retrans;
6867 val.spt_pathmaxrxt = asoc->pathmaxrxt;
6869 trans = sctp_addr_id2transport(sk, &val.spt_address,
6874 val.spt_pathmaxrxt = trans->pathmaxrxt;
6875 val.spt_pathpfthld = trans->pf_retrans;
6878 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
6885 * SCTP_GET_ASSOC_STATS
6887 * This option retrieves local per endpoint statistics. It is modeled
6888 * after OpenSolaris' implementation
6890 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
6891 char __user *optval,
6894 struct sctp_assoc_stats sas;
6895 struct sctp_association *asoc = NULL;
6897 /* User must provide at least the assoc id */
6898 if (len < sizeof(sctp_assoc_t))
6901 /* Allow the struct to grow and fill in as much as possible */
6902 len = min_t(size_t, len, sizeof(sas));
6904 if (copy_from_user(&sas, optval, len))
6907 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
6911 sas.sas_rtxchunks = asoc->stats.rtxchunks;
6912 sas.sas_gapcnt = asoc->stats.gapcnt;
6913 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
6914 sas.sas_osacks = asoc->stats.osacks;
6915 sas.sas_isacks = asoc->stats.isacks;
6916 sas.sas_octrlchunks = asoc->stats.octrlchunks;
6917 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
6918 sas.sas_oodchunks = asoc->stats.oodchunks;
6919 sas.sas_iodchunks = asoc->stats.iodchunks;
6920 sas.sas_ouodchunks = asoc->stats.ouodchunks;
6921 sas.sas_iuodchunks = asoc->stats.iuodchunks;
6922 sas.sas_idupchunks = asoc->stats.idupchunks;
6923 sas.sas_opackets = asoc->stats.opackets;
6924 sas.sas_ipackets = asoc->stats.ipackets;
6926 /* New high max rto observed, will return 0 if not a single
6927 * RTO update took place. obs_rto_ipaddr will be bogus
6930 sas.sas_maxrto = asoc->stats.max_obs_rto;
6931 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
6932 sizeof(struct sockaddr_storage));
6934 /* Mark beginning of a new observation period */
6935 asoc->stats.max_obs_rto = asoc->rto_min;
6937 if (put_user(len, optlen))
6940 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
6942 if (copy_to_user(optval, &sas, len))
6948 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
6949 char __user *optval,
6954 if (len < sizeof(int))
6958 if (sctp_sk(sk)->recvrcvinfo)
6960 if (put_user(len, optlen))
6962 if (copy_to_user(optval, &val, len))
6968 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
6969 char __user *optval,
6974 if (len < sizeof(int))
6978 if (sctp_sk(sk)->recvnxtinfo)
6980 if (put_user(len, optlen))
6982 if (copy_to_user(optval, &val, len))
6988 static int sctp_getsockopt_pr_supported(struct sock *sk, int len,
6989 char __user *optval,
6992 struct sctp_assoc_value params;
6993 struct sctp_association *asoc;
6994 int retval = -EFAULT;
6996 if (len < sizeof(params)) {
7001 len = sizeof(params);
7002 if (copy_from_user(¶ms, optval, len))
7005 asoc = sctp_id2assoc(sk, params.assoc_id);
7007 params.assoc_value = asoc->prsctp_enable;
7008 } else if (!params.assoc_id) {
7009 struct sctp_sock *sp = sctp_sk(sk);
7011 params.assoc_value = sp->ep->prsctp_enable;
7017 if (put_user(len, optlen))
7020 if (copy_to_user(optval, ¶ms, len))
7029 static int sctp_getsockopt_default_prinfo(struct sock *sk, int len,
7030 char __user *optval,
7033 struct sctp_default_prinfo info;
7034 struct sctp_association *asoc;
7035 int retval = -EFAULT;
7037 if (len < sizeof(info)) {
7043 if (copy_from_user(&info, optval, len))
7046 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
7048 info.pr_policy = SCTP_PR_POLICY(asoc->default_flags);
7049 info.pr_value = asoc->default_timetolive;
7050 } else if (!info.pr_assoc_id) {
7051 struct sctp_sock *sp = sctp_sk(sk);
7053 info.pr_policy = SCTP_PR_POLICY(sp->default_flags);
7054 info.pr_value = sp->default_timetolive;
7060 if (put_user(len, optlen))
7063 if (copy_to_user(optval, &info, len))
7072 static int sctp_getsockopt_pr_assocstatus(struct sock *sk, int len,
7073 char __user *optval,
7076 struct sctp_prstatus params;
7077 struct sctp_association *asoc;
7079 int retval = -EINVAL;
7081 if (len < sizeof(params))
7084 len = sizeof(params);
7085 if (copy_from_user(¶ms, optval, len)) {
7090 policy = params.sprstat_policy;
7091 if (policy & ~SCTP_PR_SCTP_MASK)
7094 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
7098 if (policy == SCTP_PR_SCTP_NONE) {
7099 params.sprstat_abandoned_unsent = 0;
7100 params.sprstat_abandoned_sent = 0;
7101 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
7102 params.sprstat_abandoned_unsent +=
7103 asoc->abandoned_unsent[policy];
7104 params.sprstat_abandoned_sent +=
7105 asoc->abandoned_sent[policy];
7108 params.sprstat_abandoned_unsent =
7109 asoc->abandoned_unsent[__SCTP_PR_INDEX(policy)];
7110 params.sprstat_abandoned_sent =
7111 asoc->abandoned_sent[__SCTP_PR_INDEX(policy)];
7114 if (put_user(len, optlen)) {
7119 if (copy_to_user(optval, ¶ms, len)) {
7130 static int sctp_getsockopt_pr_streamstatus(struct sock *sk, int len,
7131 char __user *optval,
7134 struct sctp_stream_out_ext *streamoute;
7135 struct sctp_association *asoc;
7136 struct sctp_prstatus params;
7137 int retval = -EINVAL;
7140 if (len < sizeof(params))
7143 len = sizeof(params);
7144 if (copy_from_user(¶ms, optval, len)) {
7149 policy = params.sprstat_policy;
7150 if (policy & ~SCTP_PR_SCTP_MASK)
7153 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
7154 if (!asoc || params.sprstat_sid >= asoc->stream.outcnt)
7157 streamoute = SCTP_SO(&asoc->stream, params.sprstat_sid)->ext;
7159 /* Not allocated yet, means all stats are 0 */
7160 params.sprstat_abandoned_unsent = 0;
7161 params.sprstat_abandoned_sent = 0;
7166 if (policy == SCTP_PR_SCTP_NONE) {
7167 params.sprstat_abandoned_unsent = 0;
7168 params.sprstat_abandoned_sent = 0;
7169 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
7170 params.sprstat_abandoned_unsent +=
7171 streamoute->abandoned_unsent[policy];
7172 params.sprstat_abandoned_sent +=
7173 streamoute->abandoned_sent[policy];
7176 params.sprstat_abandoned_unsent =
7177 streamoute->abandoned_unsent[__SCTP_PR_INDEX(policy)];
7178 params.sprstat_abandoned_sent =
7179 streamoute->abandoned_sent[__SCTP_PR_INDEX(policy)];
7182 if (put_user(len, optlen) || copy_to_user(optval, ¶ms, len)) {
7193 static int sctp_getsockopt_reconfig_supported(struct sock *sk, int len,
7194 char __user *optval,
7197 struct sctp_assoc_value params;
7198 struct sctp_association *asoc;
7199 int retval = -EFAULT;
7201 if (len < sizeof(params)) {
7206 len = sizeof(params);
7207 if (copy_from_user(¶ms, optval, len))
7210 asoc = sctp_id2assoc(sk, params.assoc_id);
7212 params.assoc_value = asoc->reconf_enable;
7213 } else if (!params.assoc_id) {
7214 struct sctp_sock *sp = sctp_sk(sk);
7216 params.assoc_value = sp->ep->reconf_enable;
7222 if (put_user(len, optlen))
7225 if (copy_to_user(optval, ¶ms, len))
7234 static int sctp_getsockopt_enable_strreset(struct sock *sk, int len,
7235 char __user *optval,
7238 struct sctp_assoc_value params;
7239 struct sctp_association *asoc;
7240 int retval = -EFAULT;
7242 if (len < sizeof(params)) {
7247 len = sizeof(params);
7248 if (copy_from_user(¶ms, optval, len))
7251 asoc = sctp_id2assoc(sk, params.assoc_id);
7253 params.assoc_value = asoc->strreset_enable;
7254 } else if (!params.assoc_id) {
7255 struct sctp_sock *sp = sctp_sk(sk);
7257 params.assoc_value = sp->ep->strreset_enable;
7263 if (put_user(len, optlen))
7266 if (copy_to_user(optval, ¶ms, len))
7275 static int sctp_getsockopt_scheduler(struct sock *sk, int len,
7276 char __user *optval,
7279 struct sctp_assoc_value params;
7280 struct sctp_association *asoc;
7281 int retval = -EFAULT;
7283 if (len < sizeof(params)) {
7288 len = sizeof(params);
7289 if (copy_from_user(¶ms, optval, len))
7292 asoc = sctp_id2assoc(sk, params.assoc_id);
7298 params.assoc_value = sctp_sched_get_sched(asoc);
7300 if (put_user(len, optlen))
7303 if (copy_to_user(optval, ¶ms, len))
7312 static int sctp_getsockopt_scheduler_value(struct sock *sk, int len,
7313 char __user *optval,
7316 struct sctp_stream_value params;
7317 struct sctp_association *asoc;
7318 int retval = -EFAULT;
7320 if (len < sizeof(params)) {
7325 len = sizeof(params);
7326 if (copy_from_user(¶ms, optval, len))
7329 asoc = sctp_id2assoc(sk, params.assoc_id);
7335 retval = sctp_sched_get_value(asoc, params.stream_id,
7336 ¶ms.stream_value);
7340 if (put_user(len, optlen)) {
7345 if (copy_to_user(optval, ¶ms, len)) {
7354 static int sctp_getsockopt_interleaving_supported(struct sock *sk, int len,
7355 char __user *optval,
7358 struct sctp_assoc_value params;
7359 struct sctp_association *asoc;
7360 int retval = -EFAULT;
7362 if (len < sizeof(params)) {
7367 len = sizeof(params);
7368 if (copy_from_user(¶ms, optval, len))
7371 asoc = sctp_id2assoc(sk, params.assoc_id);
7373 params.assoc_value = asoc->intl_enable;
7374 } else if (!params.assoc_id) {
7375 struct sctp_sock *sp = sctp_sk(sk);
7377 params.assoc_value = sp->strm_interleave;
7383 if (put_user(len, optlen))
7386 if (copy_to_user(optval, ¶ms, len))
7395 static int sctp_getsockopt_reuse_port(struct sock *sk, int len,
7396 char __user *optval,
7401 if (len < sizeof(int))
7405 val = sctp_sk(sk)->reuse;
7406 if (put_user(len, optlen))
7409 if (copy_to_user(optval, &val, len))
7415 static int sctp_getsockopt(struct sock *sk, int level, int optname,
7416 char __user *optval, int __user *optlen)
7421 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
7423 /* I can hardly begin to describe how wrong this is. This is
7424 * so broken as to be worse than useless. The API draft
7425 * REALLY is NOT helpful here... I am not convinced that the
7426 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
7427 * are at all well-founded.
7429 if (level != SOL_SCTP) {
7430 struct sctp_af *af = sctp_sk(sk)->pf->af;
7432 retval = af->getsockopt(sk, level, optname, optval, optlen);
7436 if (get_user(len, optlen))
7446 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
7448 case SCTP_DISABLE_FRAGMENTS:
7449 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
7453 retval = sctp_getsockopt_events(sk, len, optval, optlen);
7455 case SCTP_AUTOCLOSE:
7456 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
7458 case SCTP_SOCKOPT_PEELOFF:
7459 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
7461 case SCTP_SOCKOPT_PEELOFF_FLAGS:
7462 retval = sctp_getsockopt_peeloff_flags(sk, len, optval, optlen);
7464 case SCTP_PEER_ADDR_PARAMS:
7465 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
7468 case SCTP_DELAYED_SACK:
7469 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
7473 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
7475 case SCTP_GET_PEER_ADDRS:
7476 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
7479 case SCTP_GET_LOCAL_ADDRS:
7480 retval = sctp_getsockopt_local_addrs(sk, len, optval,
7483 case SCTP_SOCKOPT_CONNECTX3:
7484 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
7486 case SCTP_DEFAULT_SEND_PARAM:
7487 retval = sctp_getsockopt_default_send_param(sk, len,
7490 case SCTP_DEFAULT_SNDINFO:
7491 retval = sctp_getsockopt_default_sndinfo(sk, len,
7494 case SCTP_PRIMARY_ADDR:
7495 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
7498 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
7501 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
7503 case SCTP_ASSOCINFO:
7504 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
7506 case SCTP_I_WANT_MAPPED_V4_ADDR:
7507 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
7510 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
7512 case SCTP_GET_PEER_ADDR_INFO:
7513 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
7516 case SCTP_ADAPTATION_LAYER:
7517 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
7521 retval = sctp_getsockopt_context(sk, len, optval, optlen);
7523 case SCTP_FRAGMENT_INTERLEAVE:
7524 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
7527 case SCTP_PARTIAL_DELIVERY_POINT:
7528 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
7531 case SCTP_MAX_BURST:
7532 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
7535 case SCTP_AUTH_CHUNK:
7536 case SCTP_AUTH_DELETE_KEY:
7537 case SCTP_AUTH_DEACTIVATE_KEY:
7538 retval = -EOPNOTSUPP;
7540 case SCTP_HMAC_IDENT:
7541 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
7543 case SCTP_AUTH_ACTIVE_KEY:
7544 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
7546 case SCTP_PEER_AUTH_CHUNKS:
7547 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
7550 case SCTP_LOCAL_AUTH_CHUNKS:
7551 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
7554 case SCTP_GET_ASSOC_NUMBER:
7555 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
7557 case SCTP_GET_ASSOC_ID_LIST:
7558 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
7560 case SCTP_AUTO_ASCONF:
7561 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
7563 case SCTP_PEER_ADDR_THLDS:
7564 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
7566 case SCTP_GET_ASSOC_STATS:
7567 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
7569 case SCTP_RECVRCVINFO:
7570 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
7572 case SCTP_RECVNXTINFO:
7573 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
7575 case SCTP_PR_SUPPORTED:
7576 retval = sctp_getsockopt_pr_supported(sk, len, optval, optlen);
7578 case SCTP_DEFAULT_PRINFO:
7579 retval = sctp_getsockopt_default_prinfo(sk, len, optval,
7582 case SCTP_PR_ASSOC_STATUS:
7583 retval = sctp_getsockopt_pr_assocstatus(sk, len, optval,
7586 case SCTP_PR_STREAM_STATUS:
7587 retval = sctp_getsockopt_pr_streamstatus(sk, len, optval,
7590 case SCTP_RECONFIG_SUPPORTED:
7591 retval = sctp_getsockopt_reconfig_supported(sk, len, optval,
7594 case SCTP_ENABLE_STREAM_RESET:
7595 retval = sctp_getsockopt_enable_strreset(sk, len, optval,
7598 case SCTP_STREAM_SCHEDULER:
7599 retval = sctp_getsockopt_scheduler(sk, len, optval,
7602 case SCTP_STREAM_SCHEDULER_VALUE:
7603 retval = sctp_getsockopt_scheduler_value(sk, len, optval,
7606 case SCTP_INTERLEAVING_SUPPORTED:
7607 retval = sctp_getsockopt_interleaving_supported(sk, len, optval,
7610 case SCTP_REUSE_PORT:
7611 retval = sctp_getsockopt_reuse_port(sk, len, optval, optlen);
7614 retval = -ENOPROTOOPT;
7622 static int sctp_hash(struct sock *sk)
7628 static void sctp_unhash(struct sock *sk)
7633 /* Check if port is acceptable. Possibly find first available port.
7635 * The port hash table (contained in the 'global' SCTP protocol storage
7636 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
7637 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
7638 * list (the list number is the port number hashed out, so as you
7639 * would expect from a hash function, all the ports in a given list have
7640 * such a number that hashes out to the same list number; you were
7641 * expecting that, right?); so each list has a set of ports, with a
7642 * link to the socket (struct sock) that uses it, the port number and
7643 * a fastreuse flag (FIXME: NPI ipg).
7645 static struct sctp_bind_bucket *sctp_bucket_create(
7646 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
7648 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
7650 bool reuse = (sk->sk_reuse || sctp_sk(sk)->reuse);
7651 struct sctp_bind_hashbucket *head; /* hash list */
7652 struct sctp_bind_bucket *pp;
7653 unsigned short snum;
7656 snum = ntohs(addr->v4.sin_port);
7658 pr_debug("%s: begins, snum:%d\n", __func__, snum);
7663 /* Search for an available port. */
7664 int low, high, remaining, index;
7666 struct net *net = sock_net(sk);
7668 inet_get_local_port_range(net, &low, &high);
7669 remaining = (high - low) + 1;
7670 rover = prandom_u32() % remaining + low;
7674 if ((rover < low) || (rover > high))
7676 if (inet_is_local_reserved_port(net, rover))
7678 index = sctp_phashfn(sock_net(sk), rover);
7679 head = &sctp_port_hashtable[index];
7680 spin_lock(&head->lock);
7681 sctp_for_each_hentry(pp, &head->chain)
7682 if ((pp->port == rover) &&
7683 net_eq(sock_net(sk), pp->net))
7687 spin_unlock(&head->lock);
7688 } while (--remaining > 0);
7690 /* Exhausted local port range during search? */
7695 /* OK, here is the one we will use. HEAD (the port
7696 * hash table list entry) is non-NULL and we hold it's
7701 /* We are given an specific port number; we verify
7702 * that it is not being used. If it is used, we will
7703 * exahust the search in the hash list corresponding
7704 * to the port number (snum) - we detect that with the
7705 * port iterator, pp being NULL.
7707 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
7708 spin_lock(&head->lock);
7709 sctp_for_each_hentry(pp, &head->chain) {
7710 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
7717 if (!hlist_empty(&pp->owner)) {
7718 /* We had a port hash table hit - there is an
7719 * available port (pp != NULL) and it is being
7720 * used by other socket (pp->owner not empty); that other
7721 * socket is going to be sk2.
7725 pr_debug("%s: found a possible match\n", __func__);
7727 if (pp->fastreuse && reuse && sk->sk_state != SCTP_SS_LISTENING)
7730 /* Run through the list of sockets bound to the port
7731 * (pp->port) [via the pointers bind_next and
7732 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
7733 * we get the endpoint they describe and run through
7734 * the endpoint's list of IP (v4 or v6) addresses,
7735 * comparing each of the addresses with the address of
7736 * the socket sk. If we find a match, then that means
7737 * that this port/socket (sk) combination are already
7740 sk_for_each_bound(sk2, &pp->owner) {
7741 struct sctp_endpoint *ep2;
7742 ep2 = sctp_sk(sk2)->ep;
7745 (reuse && (sk2->sk_reuse || sctp_sk(sk2)->reuse) &&
7746 sk2->sk_state != SCTP_SS_LISTENING))
7749 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
7750 sctp_sk(sk2), sctp_sk(sk))) {
7756 pr_debug("%s: found a match\n", __func__);
7759 /* If there was a hash table miss, create a new port. */
7761 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
7764 /* In either case (hit or miss), make sure fastreuse is 1 only
7765 * if sk->sk_reuse is too (that is, if the caller requested
7766 * SO_REUSEADDR on this socket -sk-).
7768 if (hlist_empty(&pp->owner)) {
7769 if (reuse && sk->sk_state != SCTP_SS_LISTENING)
7773 } else if (pp->fastreuse &&
7774 (!reuse || sk->sk_state == SCTP_SS_LISTENING))
7777 /* We are set, so fill up all the data in the hash table
7778 * entry, tie the socket list information with the rest of the
7779 * sockets FIXME: Blurry, NPI (ipg).
7782 if (!sctp_sk(sk)->bind_hash) {
7783 inet_sk(sk)->inet_num = snum;
7784 sk_add_bind_node(sk, &pp->owner);
7785 sctp_sk(sk)->bind_hash = pp;
7790 spin_unlock(&head->lock);
7797 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
7798 * port is requested.
7800 static int sctp_get_port(struct sock *sk, unsigned short snum)
7802 union sctp_addr addr;
7803 struct sctp_af *af = sctp_sk(sk)->pf->af;
7805 /* Set up a dummy address struct from the sk. */
7806 af->from_sk(&addr, sk);
7807 addr.v4.sin_port = htons(snum);
7809 /* Note: sk->sk_num gets filled in if ephemeral port request. */
7810 return !!sctp_get_port_local(sk, &addr);
7814 * Move a socket to LISTENING state.
7816 static int sctp_listen_start(struct sock *sk, int backlog)
7818 struct sctp_sock *sp = sctp_sk(sk);
7819 struct sctp_endpoint *ep = sp->ep;
7820 struct crypto_shash *tfm = NULL;
7823 /* Allocate HMAC for generating cookie. */
7824 if (!sp->hmac && sp->sctp_hmac_alg) {
7825 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
7826 tfm = crypto_alloc_shash(alg, 0, 0);
7828 net_info_ratelimited("failed to load transform for %s: %ld\n",
7829 sp->sctp_hmac_alg, PTR_ERR(tfm));
7832 sctp_sk(sk)->hmac = tfm;
7836 * If a bind() or sctp_bindx() is not called prior to a listen()
7837 * call that allows new associations to be accepted, the system
7838 * picks an ephemeral port and will choose an address set equivalent
7839 * to binding with a wildcard address.
7841 * This is not currently spelled out in the SCTP sockets
7842 * extensions draft, but follows the practice as seen in TCP
7846 inet_sk_set_state(sk, SCTP_SS_LISTENING);
7847 if (!ep->base.bind_addr.port) {
7848 if (sctp_autobind(sk))
7851 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
7852 inet_sk_set_state(sk, SCTP_SS_CLOSED);
7857 sk->sk_max_ack_backlog = backlog;
7858 sctp_hash_endpoint(ep);
7863 * 4.1.3 / 5.1.3 listen()
7865 * By default, new associations are not accepted for UDP style sockets.
7866 * An application uses listen() to mark a socket as being able to
7867 * accept new associations.
7869 * On TCP style sockets, applications use listen() to ready the SCTP
7870 * endpoint for accepting inbound associations.
7872 * On both types of endpoints a backlog of '0' disables listening.
7874 * Move a socket to LISTENING state.
7876 int sctp_inet_listen(struct socket *sock, int backlog)
7878 struct sock *sk = sock->sk;
7879 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
7882 if (unlikely(backlog < 0))
7887 /* Peeled-off sockets are not allowed to listen(). */
7888 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
7891 if (sock->state != SS_UNCONNECTED)
7894 if (!sctp_sstate(sk, LISTENING) && !sctp_sstate(sk, CLOSED))
7897 /* If backlog is zero, disable listening. */
7899 if (sctp_sstate(sk, CLOSED))
7903 sctp_unhash_endpoint(ep);
7904 sk->sk_state = SCTP_SS_CLOSED;
7905 if (sk->sk_reuse || sctp_sk(sk)->reuse)
7906 sctp_sk(sk)->bind_hash->fastreuse = 1;
7910 /* If we are already listening, just update the backlog */
7911 if (sctp_sstate(sk, LISTENING))
7912 sk->sk_max_ack_backlog = backlog;
7914 err = sctp_listen_start(sk, backlog);
7926 * This function is done by modeling the current datagram_poll() and the
7927 * tcp_poll(). Note that, based on these implementations, we don't
7928 * lock the socket in this function, even though it seems that,
7929 * ideally, locking or some other mechanisms can be used to ensure
7930 * the integrity of the counters (sndbuf and wmem_alloc) used
7931 * in this place. We assume that we don't need locks either until proven
7934 * Another thing to note is that we include the Async I/O support
7935 * here, again, by modeling the current TCP/UDP code. We don't have
7936 * a good way to test with it yet.
7938 __poll_t sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
7940 struct sock *sk = sock->sk;
7941 struct sctp_sock *sp = sctp_sk(sk);
7944 poll_wait(file, sk_sleep(sk), wait);
7946 sock_rps_record_flow(sk);
7948 /* A TCP-style listening socket becomes readable when the accept queue
7951 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
7952 return (!list_empty(&sp->ep->asocs)) ?
7953 (EPOLLIN | EPOLLRDNORM) : 0;
7957 /* Is there any exceptional events? */
7958 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
7960 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? EPOLLPRI : 0);
7961 if (sk->sk_shutdown & RCV_SHUTDOWN)
7962 mask |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
7963 if (sk->sk_shutdown == SHUTDOWN_MASK)
7966 /* Is it readable? Reconsider this code with TCP-style support. */
7967 if (!skb_queue_empty(&sk->sk_receive_queue))
7968 mask |= EPOLLIN | EPOLLRDNORM;
7970 /* The association is either gone or not ready. */
7971 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
7974 /* Is it writable? */
7975 if (sctp_writeable(sk)) {
7976 mask |= EPOLLOUT | EPOLLWRNORM;
7978 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
7980 * Since the socket is not locked, the buffer
7981 * might be made available after the writeable check and
7982 * before the bit is set. This could cause a lost I/O
7983 * signal. tcp_poll() has a race breaker for this race
7984 * condition. Based on their implementation, we put
7985 * in the following code to cover it as well.
7987 if (sctp_writeable(sk))
7988 mask |= EPOLLOUT | EPOLLWRNORM;
7993 /********************************************************************
7994 * 2nd Level Abstractions
7995 ********************************************************************/
7997 static struct sctp_bind_bucket *sctp_bucket_create(
7998 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
8000 struct sctp_bind_bucket *pp;
8002 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
8004 SCTP_DBG_OBJCNT_INC(bind_bucket);
8007 INIT_HLIST_HEAD(&pp->owner);
8009 hlist_add_head(&pp->node, &head->chain);
8014 /* Caller must hold hashbucket lock for this tb with local BH disabled */
8015 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
8017 if (pp && hlist_empty(&pp->owner)) {
8018 __hlist_del(&pp->node);
8019 kmem_cache_free(sctp_bucket_cachep, pp);
8020 SCTP_DBG_OBJCNT_DEC(bind_bucket);
8024 /* Release this socket's reference to a local port. */
8025 static inline void __sctp_put_port(struct sock *sk)
8027 struct sctp_bind_hashbucket *head =
8028 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
8029 inet_sk(sk)->inet_num)];
8030 struct sctp_bind_bucket *pp;
8032 spin_lock(&head->lock);
8033 pp = sctp_sk(sk)->bind_hash;
8034 __sk_del_bind_node(sk);
8035 sctp_sk(sk)->bind_hash = NULL;
8036 inet_sk(sk)->inet_num = 0;
8037 sctp_bucket_destroy(pp);
8038 spin_unlock(&head->lock);
8041 void sctp_put_port(struct sock *sk)
8044 __sctp_put_port(sk);
8049 * The system picks an ephemeral port and choose an address set equivalent
8050 * to binding with a wildcard address.
8051 * One of those addresses will be the primary address for the association.
8052 * This automatically enables the multihoming capability of SCTP.
8054 static int sctp_autobind(struct sock *sk)
8056 union sctp_addr autoaddr;
8060 /* Initialize a local sockaddr structure to INADDR_ANY. */
8061 af = sctp_sk(sk)->pf->af;
8063 port = htons(inet_sk(sk)->inet_num);
8064 af->inaddr_any(&autoaddr, port);
8066 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
8069 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
8072 * 4.2 The cmsghdr Structure *
8074 * When ancillary data is sent or received, any number of ancillary data
8075 * objects can be specified by the msg_control and msg_controllen members of
8076 * the msghdr structure, because each object is preceded by
8077 * a cmsghdr structure defining the object's length (the cmsg_len member).
8078 * Historically Berkeley-derived implementations have passed only one object
8079 * at a time, but this API allows multiple objects to be
8080 * passed in a single call to sendmsg() or recvmsg(). The following example
8081 * shows two ancillary data objects in a control buffer.
8083 * |<--------------------------- msg_controllen -------------------------->|
8086 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
8088 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
8091 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
8093 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
8096 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
8097 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
8099 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
8101 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
8108 static int sctp_msghdr_parse(const struct msghdr *msg, struct sctp_cmsgs *cmsgs)
8110 struct msghdr *my_msg = (struct msghdr *)msg;
8111 struct cmsghdr *cmsg;
8113 for_each_cmsghdr(cmsg, my_msg) {
8114 if (!CMSG_OK(my_msg, cmsg))
8117 /* Should we parse this header or ignore? */
8118 if (cmsg->cmsg_level != IPPROTO_SCTP)
8121 /* Strictly check lengths following example in SCM code. */
8122 switch (cmsg->cmsg_type) {
8124 /* SCTP Socket API Extension
8125 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
8127 * This cmsghdr structure provides information for
8128 * initializing new SCTP associations with sendmsg().
8129 * The SCTP_INITMSG socket option uses this same data
8130 * structure. This structure is not used for
8133 * cmsg_level cmsg_type cmsg_data[]
8134 * ------------ ------------ ----------------------
8135 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
8137 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
8140 cmsgs->init = CMSG_DATA(cmsg);
8144 /* SCTP Socket API Extension
8145 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
8147 * This cmsghdr structure specifies SCTP options for
8148 * sendmsg() and describes SCTP header information
8149 * about a received message through recvmsg().
8151 * cmsg_level cmsg_type cmsg_data[]
8152 * ------------ ------------ ----------------------
8153 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
8155 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
8158 cmsgs->srinfo = CMSG_DATA(cmsg);
8160 if (cmsgs->srinfo->sinfo_flags &
8161 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
8162 SCTP_SACK_IMMEDIATELY | SCTP_SENDALL |
8163 SCTP_PR_SCTP_MASK | SCTP_ABORT | SCTP_EOF))
8168 /* SCTP Socket API Extension
8169 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
8171 * This cmsghdr structure specifies SCTP options for
8172 * sendmsg(). This structure and SCTP_RCVINFO replaces
8173 * SCTP_SNDRCV which has been deprecated.
8175 * cmsg_level cmsg_type cmsg_data[]
8176 * ------------ ------------ ---------------------
8177 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
8179 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
8182 cmsgs->sinfo = CMSG_DATA(cmsg);
8184 if (cmsgs->sinfo->snd_flags &
8185 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
8186 SCTP_SACK_IMMEDIATELY | SCTP_SENDALL |
8187 SCTP_PR_SCTP_MASK | SCTP_ABORT | SCTP_EOF))
8191 /* SCTP Socket API Extension
8192 * 5.3.7 SCTP PR-SCTP Information Structure (SCTP_PRINFO)
8194 * This cmsghdr structure specifies SCTP options for sendmsg().
8196 * cmsg_level cmsg_type cmsg_data[]
8197 * ------------ ------------ ---------------------
8198 * IPPROTO_SCTP SCTP_PRINFO struct sctp_prinfo
8200 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_prinfo)))
8203 cmsgs->prinfo = CMSG_DATA(cmsg);
8204 if (cmsgs->prinfo->pr_policy & ~SCTP_PR_SCTP_MASK)
8207 if (cmsgs->prinfo->pr_policy == SCTP_PR_SCTP_NONE)
8208 cmsgs->prinfo->pr_value = 0;
8211 /* SCTP Socket API Extension
8212 * 5.3.8 SCTP AUTH Information Structure (SCTP_AUTHINFO)
8214 * This cmsghdr structure specifies SCTP options for sendmsg().
8216 * cmsg_level cmsg_type cmsg_data[]
8217 * ------------ ------------ ---------------------
8218 * IPPROTO_SCTP SCTP_AUTHINFO struct sctp_authinfo
8220 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_authinfo)))
8223 cmsgs->authinfo = CMSG_DATA(cmsg);
8225 case SCTP_DSTADDRV4:
8226 case SCTP_DSTADDRV6:
8227 /* SCTP Socket API Extension
8228 * 5.3.9/10 SCTP Destination IPv4/6 Address Structure (SCTP_DSTADDRV4/6)
8230 * This cmsghdr structure specifies SCTP options for sendmsg().
8232 * cmsg_level cmsg_type cmsg_data[]
8233 * ------------ ------------ ---------------------
8234 * IPPROTO_SCTP SCTP_DSTADDRV4 struct in_addr
8235 * ------------ ------------ ---------------------
8236 * IPPROTO_SCTP SCTP_DSTADDRV6 struct in6_addr
8238 cmsgs->addrs_msg = my_msg;
8249 * Wait for a packet..
8250 * Note: This function is the same function as in core/datagram.c
8251 * with a few modifications to make lksctp work.
8253 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
8258 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
8260 /* Socket errors? */
8261 error = sock_error(sk);
8265 if (!skb_queue_empty(&sk->sk_receive_queue))
8268 /* Socket shut down? */
8269 if (sk->sk_shutdown & RCV_SHUTDOWN)
8272 /* Sequenced packets can come disconnected. If so we report the
8277 /* Is there a good reason to think that we may receive some data? */
8278 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
8281 /* Handle signals. */
8282 if (signal_pending(current))
8285 /* Let another process have a go. Since we are going to sleep
8286 * anyway. Note: This may cause odd behaviors if the message
8287 * does not fit in the user's buffer, but this seems to be the
8288 * only way to honor MSG_DONTWAIT realistically.
8291 *timeo_p = schedule_timeout(*timeo_p);
8295 finish_wait(sk_sleep(sk), &wait);
8299 error = sock_intr_errno(*timeo_p);
8302 finish_wait(sk_sleep(sk), &wait);
8307 /* Receive a datagram.
8308 * Note: This is pretty much the same routine as in core/datagram.c
8309 * with a few changes to make lksctp work.
8311 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
8312 int noblock, int *err)
8315 struct sk_buff *skb;
8318 timeo = sock_rcvtimeo(sk, noblock);
8320 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
8321 MAX_SCHEDULE_TIMEOUT);
8324 /* Again only user level code calls this function,
8325 * so nothing interrupt level
8326 * will suddenly eat the receive_queue.
8328 * Look at current nfs client by the way...
8329 * However, this function was correct in any case. 8)
8331 if (flags & MSG_PEEK) {
8332 skb = skb_peek(&sk->sk_receive_queue);
8334 refcount_inc(&skb->users);
8336 skb = __skb_dequeue(&sk->sk_receive_queue);
8342 /* Caller is allowed not to check sk->sk_err before calling. */
8343 error = sock_error(sk);
8347 if (sk->sk_shutdown & RCV_SHUTDOWN)
8350 if (sk_can_busy_loop(sk)) {
8351 sk_busy_loop(sk, noblock);
8353 if (!skb_queue_empty(&sk->sk_receive_queue))
8357 /* User doesn't want to wait. */
8361 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
8370 /* If sndbuf has changed, wake up per association sndbuf waiters. */
8371 static void __sctp_write_space(struct sctp_association *asoc)
8373 struct sock *sk = asoc->base.sk;
8375 if (sctp_wspace(asoc) <= 0)
8378 if (waitqueue_active(&asoc->wait))
8379 wake_up_interruptible(&asoc->wait);
8381 if (sctp_writeable(sk)) {
8382 struct socket_wq *wq;
8385 wq = rcu_dereference(sk->sk_wq);
8387 if (waitqueue_active(&wq->wait))
8388 wake_up_interruptible(&wq->wait);
8390 /* Note that we try to include the Async I/O support
8391 * here by modeling from the current TCP/UDP code.
8392 * We have not tested with it yet.
8394 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
8395 sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT);
8401 static void sctp_wake_up_waiters(struct sock *sk,
8402 struct sctp_association *asoc)
8404 struct sctp_association *tmp = asoc;
8406 /* We do accounting for the sndbuf space per association,
8407 * so we only need to wake our own association.
8409 if (asoc->ep->sndbuf_policy)
8410 return __sctp_write_space(asoc);
8412 /* If association goes down and is just flushing its
8413 * outq, then just normally notify others.
8415 if (asoc->base.dead)
8416 return sctp_write_space(sk);
8418 /* Accounting for the sndbuf space is per socket, so we
8419 * need to wake up others, try to be fair and in case of
8420 * other associations, let them have a go first instead
8421 * of just doing a sctp_write_space() call.
8423 * Note that we reach sctp_wake_up_waiters() only when
8424 * associations free up queued chunks, thus we are under
8425 * lock and the list of associations on a socket is
8426 * guaranteed not to change.
8428 for (tmp = list_next_entry(tmp, asocs); 1;
8429 tmp = list_next_entry(tmp, asocs)) {
8430 /* Manually skip the head element. */
8431 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
8433 /* Wake up association. */
8434 __sctp_write_space(tmp);
8435 /* We've reached the end. */
8441 /* Do accounting for the sndbuf space.
8442 * Decrement the used sndbuf space of the corresponding association by the
8443 * data size which was just transmitted(freed).
8445 static void sctp_wfree(struct sk_buff *skb)
8447 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
8448 struct sctp_association *asoc = chunk->asoc;
8449 struct sock *sk = asoc->base.sk;
8451 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
8452 sizeof(struct sk_buff) +
8453 sizeof(struct sctp_chunk);
8455 WARN_ON(refcount_sub_and_test(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc));
8458 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
8460 sk->sk_wmem_queued -= skb->truesize;
8461 sk_mem_uncharge(sk, skb->truesize);
8464 struct sctp_shared_key *shkey = chunk->shkey;
8466 /* refcnt == 2 and !list_empty mean after this release, it's
8467 * not being used anywhere, and it's time to notify userland
8468 * that this shkey can be freed if it's been deactivated.
8470 if (shkey->deactivated && !list_empty(&shkey->key_list) &&
8471 refcount_read(&shkey->refcnt) == 2) {
8472 struct sctp_ulpevent *ev;
8474 ev = sctp_ulpevent_make_authkey(asoc, shkey->key_id,
8478 asoc->stream.si->enqueue_event(&asoc->ulpq, ev);
8480 sctp_auth_shkey_release(chunk->shkey);
8484 sctp_wake_up_waiters(sk, asoc);
8486 sctp_association_put(asoc);
8489 /* Do accounting for the receive space on the socket.
8490 * Accounting for the association is done in ulpevent.c
8491 * We set this as a destructor for the cloned data skbs so that
8492 * accounting is done at the correct time.
8494 void sctp_sock_rfree(struct sk_buff *skb)
8496 struct sock *sk = skb->sk;
8497 struct sctp_ulpevent *event = sctp_skb2event(skb);
8499 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
8502 * Mimic the behavior of sock_rfree
8504 sk_mem_uncharge(sk, event->rmem_len);
8508 /* Helper function to wait for space in the sndbuf. */
8509 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
8512 struct sock *sk = asoc->base.sk;
8513 long current_timeo = *timeo_p;
8517 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
8520 /* Increment the association's refcnt. */
8521 sctp_association_hold(asoc);
8523 /* Wait on the association specific sndbuf space. */
8525 prepare_to_wait_exclusive(&asoc->wait, &wait,
8526 TASK_INTERRUPTIBLE);
8527 if (asoc->base.dead)
8531 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING)
8533 if (signal_pending(current))
8534 goto do_interrupted;
8535 if (msg_len <= sctp_wspace(asoc))
8538 /* Let another process have a go. Since we are going
8542 current_timeo = schedule_timeout(current_timeo);
8544 if (sk != asoc->base.sk)
8547 *timeo_p = current_timeo;
8551 finish_wait(&asoc->wait, &wait);
8553 /* Release the association's refcnt. */
8554 sctp_association_put(asoc);
8567 err = sock_intr_errno(*timeo_p);
8575 void sctp_data_ready(struct sock *sk)
8577 struct socket_wq *wq;
8580 wq = rcu_dereference(sk->sk_wq);
8581 if (skwq_has_sleeper(wq))
8582 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
8583 EPOLLRDNORM | EPOLLRDBAND);
8584 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
8588 /* If socket sndbuf has changed, wake up all per association waiters. */
8589 void sctp_write_space(struct sock *sk)
8591 struct sctp_association *asoc;
8593 /* Wake up the tasks in each wait queue. */
8594 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
8595 __sctp_write_space(asoc);
8599 /* Is there any sndbuf space available on the socket?
8601 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
8602 * associations on the same socket. For a UDP-style socket with
8603 * multiple associations, it is possible for it to be "unwriteable"
8604 * prematurely. I assume that this is acceptable because
8605 * a premature "unwriteable" is better than an accidental "writeable" which
8606 * would cause an unwanted block under certain circumstances. For the 1-1
8607 * UDP-style sockets or TCP-style sockets, this code should work.
8610 static int sctp_writeable(struct sock *sk)
8614 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
8620 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
8621 * returns immediately with EINPROGRESS.
8623 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
8625 struct sock *sk = asoc->base.sk;
8627 long current_timeo = *timeo_p;
8630 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
8632 /* Increment the association's refcnt. */
8633 sctp_association_hold(asoc);
8636 prepare_to_wait_exclusive(&asoc->wait, &wait,
8637 TASK_INTERRUPTIBLE);
8640 if (sk->sk_shutdown & RCV_SHUTDOWN)
8642 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
8645 if (signal_pending(current))
8646 goto do_interrupted;
8648 if (sctp_state(asoc, ESTABLISHED))
8651 /* Let another process have a go. Since we are going
8655 current_timeo = schedule_timeout(current_timeo);
8658 *timeo_p = current_timeo;
8662 finish_wait(&asoc->wait, &wait);
8664 /* Release the association's refcnt. */
8665 sctp_association_put(asoc);
8670 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
8673 err = -ECONNREFUSED;
8677 err = sock_intr_errno(*timeo_p);
8685 static int sctp_wait_for_accept(struct sock *sk, long timeo)
8687 struct sctp_endpoint *ep;
8691 ep = sctp_sk(sk)->ep;
8695 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
8696 TASK_INTERRUPTIBLE);
8698 if (list_empty(&ep->asocs)) {
8700 timeo = schedule_timeout(timeo);
8705 if (!sctp_sstate(sk, LISTENING))
8709 if (!list_empty(&ep->asocs))
8712 err = sock_intr_errno(timeo);
8713 if (signal_pending(current))
8721 finish_wait(sk_sleep(sk), &wait);
8726 static void sctp_wait_for_close(struct sock *sk, long timeout)
8731 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
8732 if (list_empty(&sctp_sk(sk)->ep->asocs))
8735 timeout = schedule_timeout(timeout);
8737 } while (!signal_pending(current) && timeout);
8739 finish_wait(sk_sleep(sk), &wait);
8742 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
8744 struct sk_buff *frag;
8749 /* Don't forget the fragments. */
8750 skb_walk_frags(skb, frag)
8751 sctp_skb_set_owner_r_frag(frag, sk);
8754 sctp_skb_set_owner_r(skb, sk);
8757 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
8758 struct sctp_association *asoc)
8760 struct inet_sock *inet = inet_sk(sk);
8761 struct inet_sock *newinet;
8762 struct sctp_sock *sp = sctp_sk(sk);
8763 struct sctp_endpoint *ep = sp->ep;
8765 newsk->sk_type = sk->sk_type;
8766 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
8767 newsk->sk_flags = sk->sk_flags;
8768 newsk->sk_tsflags = sk->sk_tsflags;
8769 newsk->sk_no_check_tx = sk->sk_no_check_tx;
8770 newsk->sk_no_check_rx = sk->sk_no_check_rx;
8771 newsk->sk_reuse = sk->sk_reuse;
8772 sctp_sk(newsk)->reuse = sp->reuse;
8774 newsk->sk_shutdown = sk->sk_shutdown;
8775 newsk->sk_destruct = sctp_destruct_sock;
8776 newsk->sk_family = sk->sk_family;
8777 newsk->sk_protocol = IPPROTO_SCTP;
8778 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
8779 newsk->sk_sndbuf = sk->sk_sndbuf;
8780 newsk->sk_rcvbuf = sk->sk_rcvbuf;
8781 newsk->sk_lingertime = sk->sk_lingertime;
8782 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
8783 newsk->sk_sndtimeo = sk->sk_sndtimeo;
8784 newsk->sk_rxhash = sk->sk_rxhash;
8786 newinet = inet_sk(newsk);
8788 /* Initialize sk's sport, dport, rcv_saddr and daddr for
8789 * getsockname() and getpeername()
8791 newinet->inet_sport = inet->inet_sport;
8792 newinet->inet_saddr = inet->inet_saddr;
8793 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
8794 newinet->inet_dport = htons(asoc->peer.port);
8795 newinet->pmtudisc = inet->pmtudisc;
8796 newinet->inet_id = asoc->next_tsn ^ jiffies;
8798 newinet->uc_ttl = inet->uc_ttl;
8799 newinet->mc_loop = 1;
8800 newinet->mc_ttl = 1;
8801 newinet->mc_index = 0;
8802 newinet->mc_list = NULL;
8804 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
8805 net_enable_timestamp();
8807 /* Set newsk security attributes from orginal sk and connection
8808 * security attribute from ep.
8810 security_sctp_sk_clone(ep, sk, newsk);
8813 static inline void sctp_copy_descendant(struct sock *sk_to,
8814 const struct sock *sk_from)
8816 int ancestor_size = sizeof(struct inet_sock) +
8817 sizeof(struct sctp_sock) -
8818 offsetof(struct sctp_sock, auto_asconf_list);
8820 if (sk_from->sk_family == PF_INET6)
8821 ancestor_size += sizeof(struct ipv6_pinfo);
8823 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
8826 /* Populate the fields of the newsk from the oldsk and migrate the assoc
8827 * and its messages to the newsk.
8829 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
8830 struct sctp_association *assoc,
8831 enum sctp_socket_type type)
8833 struct sctp_sock *oldsp = sctp_sk(oldsk);
8834 struct sctp_sock *newsp = sctp_sk(newsk);
8835 struct sctp_bind_bucket *pp; /* hash list port iterator */
8836 struct sctp_endpoint *newep = newsp->ep;
8837 struct sk_buff *skb, *tmp;
8838 struct sctp_ulpevent *event;
8839 struct sctp_bind_hashbucket *head;
8841 /* Migrate socket buffer sizes and all the socket level options to the
8844 newsk->sk_sndbuf = oldsk->sk_sndbuf;
8845 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
8846 /* Brute force copy old sctp opt. */
8847 sctp_copy_descendant(newsk, oldsk);
8849 /* Restore the ep value that was overwritten with the above structure
8855 /* Hook this new socket in to the bind_hash list. */
8856 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
8857 inet_sk(oldsk)->inet_num)];
8858 spin_lock_bh(&head->lock);
8859 pp = sctp_sk(oldsk)->bind_hash;
8860 sk_add_bind_node(newsk, &pp->owner);
8861 sctp_sk(newsk)->bind_hash = pp;
8862 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
8863 spin_unlock_bh(&head->lock);
8865 /* Copy the bind_addr list from the original endpoint to the new
8866 * endpoint so that we can handle restarts properly
8868 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
8869 &oldsp->ep->base.bind_addr, GFP_KERNEL);
8871 /* Move any messages in the old socket's receive queue that are for the
8872 * peeled off association to the new socket's receive queue.
8874 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
8875 event = sctp_skb2event(skb);
8876 if (event->asoc == assoc) {
8877 __skb_unlink(skb, &oldsk->sk_receive_queue);
8878 __skb_queue_tail(&newsk->sk_receive_queue, skb);
8879 sctp_skb_set_owner_r_frag(skb, newsk);
8883 /* Clean up any messages pending delivery due to partial
8884 * delivery. Three cases:
8885 * 1) No partial deliver; no work.
8886 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
8887 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
8889 skb_queue_head_init(&newsp->pd_lobby);
8890 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
8892 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
8893 struct sk_buff_head *queue;
8895 /* Decide which queue to move pd_lobby skbs to. */
8896 if (assoc->ulpq.pd_mode) {
8897 queue = &newsp->pd_lobby;
8899 queue = &newsk->sk_receive_queue;
8901 /* Walk through the pd_lobby, looking for skbs that
8902 * need moved to the new socket.
8904 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
8905 event = sctp_skb2event(skb);
8906 if (event->asoc == assoc) {
8907 __skb_unlink(skb, &oldsp->pd_lobby);
8908 __skb_queue_tail(queue, skb);
8909 sctp_skb_set_owner_r_frag(skb, newsk);
8913 /* Clear up any skbs waiting for the partial
8914 * delivery to finish.
8916 if (assoc->ulpq.pd_mode)
8917 sctp_clear_pd(oldsk, NULL);
8921 sctp_for_each_rx_skb(assoc, newsk, sctp_skb_set_owner_r_frag);
8923 /* Set the type of socket to indicate that it is peeled off from the
8924 * original UDP-style socket or created with the accept() call on a
8925 * TCP-style socket..
8929 /* Mark the new socket "in-use" by the user so that any packets
8930 * that may arrive on the association after we've moved it are
8931 * queued to the backlog. This prevents a potential race between
8932 * backlog processing on the old socket and new-packet processing
8933 * on the new socket.
8935 * The caller has just allocated newsk so we can guarantee that other
8936 * paths won't try to lock it and then oldsk.
8938 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
8939 sctp_for_each_tx_datachunk(assoc, sctp_clear_owner_w);
8940 sctp_assoc_migrate(assoc, newsk);
8941 sctp_for_each_tx_datachunk(assoc, sctp_set_owner_w);
8943 /* If the association on the newsk is already closed before accept()
8944 * is called, set RCV_SHUTDOWN flag.
8946 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP)) {
8947 inet_sk_set_state(newsk, SCTP_SS_CLOSED);
8948 newsk->sk_shutdown |= RCV_SHUTDOWN;
8950 inet_sk_set_state(newsk, SCTP_SS_ESTABLISHED);
8953 release_sock(newsk);
8957 /* This proto struct describes the ULP interface for SCTP. */
8958 struct proto sctp_prot = {
8960 .owner = THIS_MODULE,
8961 .close = sctp_close,
8962 .disconnect = sctp_disconnect,
8963 .accept = sctp_accept,
8964 .ioctl = sctp_ioctl,
8965 .init = sctp_init_sock,
8966 .destroy = sctp_destroy_sock,
8967 .shutdown = sctp_shutdown,
8968 .setsockopt = sctp_setsockopt,
8969 .getsockopt = sctp_getsockopt,
8970 .sendmsg = sctp_sendmsg,
8971 .recvmsg = sctp_recvmsg,
8973 .backlog_rcv = sctp_backlog_rcv,
8975 .unhash = sctp_unhash,
8976 .get_port = sctp_get_port,
8977 .obj_size = sizeof(struct sctp_sock),
8978 .useroffset = offsetof(struct sctp_sock, subscribe),
8979 .usersize = offsetof(struct sctp_sock, initmsg) -
8980 offsetof(struct sctp_sock, subscribe) +
8981 sizeof_field(struct sctp_sock, initmsg),
8982 .sysctl_mem = sysctl_sctp_mem,
8983 .sysctl_rmem = sysctl_sctp_rmem,
8984 .sysctl_wmem = sysctl_sctp_wmem,
8985 .memory_pressure = &sctp_memory_pressure,
8986 .enter_memory_pressure = sctp_enter_memory_pressure,
8987 .memory_allocated = &sctp_memory_allocated,
8988 .sockets_allocated = &sctp_sockets_allocated,
8991 #if IS_ENABLED(CONFIG_IPV6)
8993 #include <net/transp_v6.h>
8994 static void sctp_v6_destroy_sock(struct sock *sk)
8996 sctp_destroy_sock(sk);
8997 inet6_destroy_sock(sk);
9000 struct proto sctpv6_prot = {
9002 .owner = THIS_MODULE,
9003 .close = sctp_close,
9004 .disconnect = sctp_disconnect,
9005 .accept = sctp_accept,
9006 .ioctl = sctp_ioctl,
9007 .init = sctp_init_sock,
9008 .destroy = sctp_v6_destroy_sock,
9009 .shutdown = sctp_shutdown,
9010 .setsockopt = sctp_setsockopt,
9011 .getsockopt = sctp_getsockopt,
9012 .sendmsg = sctp_sendmsg,
9013 .recvmsg = sctp_recvmsg,
9015 .backlog_rcv = sctp_backlog_rcv,
9017 .unhash = sctp_unhash,
9018 .get_port = sctp_get_port,
9019 .obj_size = sizeof(struct sctp6_sock),
9020 .useroffset = offsetof(struct sctp6_sock, sctp.subscribe),
9021 .usersize = offsetof(struct sctp6_sock, sctp.initmsg) -
9022 offsetof(struct sctp6_sock, sctp.subscribe) +
9023 sizeof_field(struct sctp6_sock, sctp.initmsg),
9024 .sysctl_mem = sysctl_sctp_mem,
9025 .sysctl_rmem = sysctl_sctp_rmem,
9026 .sysctl_wmem = sysctl_sctp_wmem,
9027 .memory_pressure = &sctp_memory_pressure,
9028 .enter_memory_pressure = sctp_enter_memory_pressure,
9029 .memory_allocated = &sctp_memory_allocated,
9030 .sockets_allocated = &sctp_sockets_allocated,
9032 #endif /* IS_ENABLED(CONFIG_IPV6) */