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 if (asoc && (asoc->base.sk != sk || asoc->base.dead))
276 spin_unlock_bh(&sctp_assocs_id_lock);
281 /* Look up the transport from an address and an assoc id. If both address and
282 * id are specified, the associations matching the address and the id should be
285 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
286 struct sockaddr_storage *addr,
289 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
290 struct sctp_af *af = sctp_get_af_specific(addr->ss_family);
291 union sctp_addr *laddr = (union sctp_addr *)addr;
292 struct sctp_transport *transport;
294 if (!af || sctp_verify_addr(sk, laddr, af->sockaddr_len))
297 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
304 id_asoc = sctp_id2assoc(sk, id);
305 if (id_asoc && (id_asoc != addr_asoc))
308 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
309 (union sctp_addr *)addr);
314 /* API 3.1.2 bind() - UDP Style Syntax
315 * The syntax of bind() is,
317 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
319 * sd - the socket descriptor returned by socket().
320 * addr - the address structure (struct sockaddr_in or struct
321 * sockaddr_in6 [RFC 2553]),
322 * addr_len - the size of the address structure.
324 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
330 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
333 /* Disallow binding twice. */
334 if (!sctp_sk(sk)->ep->base.bind_addr.port)
335 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
345 static long sctp_get_port_local(struct sock *, union sctp_addr *);
347 /* Verify this is a valid sockaddr. */
348 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
349 union sctp_addr *addr, int len)
353 /* Check minimum size. */
354 if (len < sizeof (struct sockaddr))
357 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
360 if (addr->sa.sa_family == AF_INET6) {
361 if (len < SIN6_LEN_RFC2133)
363 /* V4 mapped address are really of AF_INET family */
364 if (ipv6_addr_v4mapped(&addr->v6.sin6_addr) &&
365 !opt->pf->af_supported(AF_INET, opt))
369 /* If we get this far, af is valid. */
370 af = sctp_get_af_specific(addr->sa.sa_family);
372 if (len < af->sockaddr_len)
378 /* Bind a local address either to an endpoint or to an association. */
379 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
381 struct net *net = sock_net(sk);
382 struct sctp_sock *sp = sctp_sk(sk);
383 struct sctp_endpoint *ep = sp->ep;
384 struct sctp_bind_addr *bp = &ep->base.bind_addr;
389 /* Common sockaddr verification. */
390 af = sctp_sockaddr_af(sp, addr, len);
392 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
393 __func__, sk, addr, len);
397 snum = ntohs(addr->v4.sin_port);
399 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
400 __func__, sk, &addr->sa, bp->port, snum, len);
402 /* PF specific bind() address verification. */
403 if (!sp->pf->bind_verify(sp, addr))
404 return -EADDRNOTAVAIL;
406 /* We must either be unbound, or bind to the same port.
407 * It's OK to allow 0 ports if we are already bound.
408 * We'll just inhert an already bound port in this case
413 else if (snum != bp->port) {
414 pr_debug("%s: new port %d doesn't match existing port "
415 "%d\n", __func__, snum, bp->port);
420 if (snum && snum < inet_prot_sock(net) &&
421 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
424 /* See if the address matches any of the addresses we may have
425 * already bound before checking against other endpoints.
427 if (sctp_bind_addr_match(bp, addr, sp))
430 /* Make sure we are allowed to bind here.
431 * The function sctp_get_port_local() does duplicate address
434 addr->v4.sin_port = htons(snum);
435 if ((ret = sctp_get_port_local(sk, addr))) {
439 /* Refresh ephemeral port. */
441 bp->port = inet_sk(sk)->inet_num;
443 /* Add the address to the bind address list.
444 * Use GFP_ATOMIC since BHs will be disabled.
446 ret = sctp_add_bind_addr(bp, addr, af->sockaddr_len,
447 SCTP_ADDR_SRC, GFP_ATOMIC);
449 /* Copy back into socket for getsockname() use. */
451 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
452 sp->pf->to_sk_saddr(addr, sk);
458 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
460 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
461 * at any one time. If a sender, after sending an ASCONF chunk, decides
462 * it needs to transfer another ASCONF Chunk, it MUST wait until the
463 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
464 * subsequent ASCONF. Note this restriction binds each side, so at any
465 * time two ASCONF may be in-transit on any given association (one sent
466 * from each endpoint).
468 static int sctp_send_asconf(struct sctp_association *asoc,
469 struct sctp_chunk *chunk)
471 struct net *net = sock_net(asoc->base.sk);
474 /* If there is an outstanding ASCONF chunk, queue it for later
477 if (asoc->addip_last_asconf) {
478 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
482 /* Hold the chunk until an ASCONF_ACK is received. */
483 sctp_chunk_hold(chunk);
484 retval = sctp_primitive_ASCONF(net, asoc, chunk);
486 sctp_chunk_free(chunk);
488 asoc->addip_last_asconf = chunk;
494 /* Add a list of addresses as bind addresses to local endpoint or
497 * Basically run through each address specified in the addrs/addrcnt
498 * array/length pair, determine if it is IPv6 or IPv4 and call
499 * sctp_do_bind() on it.
501 * If any of them fails, then the operation will be reversed and the
502 * ones that were added will be removed.
504 * Only sctp_setsockopt_bindx() is supposed to call this function.
506 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
511 struct sockaddr *sa_addr;
514 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
518 for (cnt = 0; cnt < addrcnt; cnt++) {
519 /* The list may contain either IPv4 or IPv6 address;
520 * determine the address length for walking thru the list.
523 af = sctp_get_af_specific(sa_addr->sa_family);
529 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
532 addr_buf += af->sockaddr_len;
536 /* Failed. Cleanup the ones that have been added */
538 sctp_bindx_rem(sk, addrs, cnt);
546 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
547 * associations that are part of the endpoint indicating that a list of local
548 * addresses are added to the endpoint.
550 * If any of the addresses is already in the bind address list of the
551 * association, we do not send the chunk for that association. But it will not
552 * affect other associations.
554 * Only sctp_setsockopt_bindx() is supposed to call this function.
556 static int sctp_send_asconf_add_ip(struct sock *sk,
557 struct sockaddr *addrs,
560 struct net *net = sock_net(sk);
561 struct sctp_sock *sp;
562 struct sctp_endpoint *ep;
563 struct sctp_association *asoc;
564 struct sctp_bind_addr *bp;
565 struct sctp_chunk *chunk;
566 struct sctp_sockaddr_entry *laddr;
567 union sctp_addr *addr;
568 union sctp_addr saveaddr;
575 if (!net->sctp.addip_enable)
581 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
582 __func__, sk, addrs, addrcnt);
584 list_for_each_entry(asoc, &ep->asocs, asocs) {
585 if (!asoc->peer.asconf_capable)
588 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
591 if (!sctp_state(asoc, ESTABLISHED))
594 /* Check if any address in the packed array of addresses is
595 * in the bind address list of the association. If so,
596 * do not send the asconf chunk to its peer, but continue with
597 * other associations.
600 for (i = 0; i < addrcnt; i++) {
602 af = sctp_get_af_specific(addr->v4.sin_family);
608 if (sctp_assoc_lookup_laddr(asoc, addr))
611 addr_buf += af->sockaddr_len;
616 /* Use the first valid address in bind addr list of
617 * association as Address Parameter of ASCONF CHUNK.
619 bp = &asoc->base.bind_addr;
620 p = bp->address_list.next;
621 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
622 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
623 addrcnt, SCTP_PARAM_ADD_IP);
629 /* Add the new addresses to the bind address list with
630 * use_as_src set to 0.
633 for (i = 0; i < addrcnt; i++) {
635 af = sctp_get_af_specific(addr->v4.sin_family);
636 memcpy(&saveaddr, addr, af->sockaddr_len);
637 retval = sctp_add_bind_addr(bp, &saveaddr,
639 SCTP_ADDR_NEW, GFP_ATOMIC);
640 addr_buf += af->sockaddr_len;
642 if (asoc->src_out_of_asoc_ok) {
643 struct sctp_transport *trans;
645 list_for_each_entry(trans,
646 &asoc->peer.transport_addr_list, transports) {
647 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
648 2*asoc->pathmtu, 4380));
649 trans->ssthresh = asoc->peer.i.a_rwnd;
650 trans->rto = asoc->rto_initial;
651 sctp_max_rto(asoc, trans);
652 trans->rtt = trans->srtt = trans->rttvar = 0;
653 /* Clear the source and route cache */
654 sctp_transport_route(trans, NULL,
655 sctp_sk(asoc->base.sk));
658 retval = sctp_send_asconf(asoc, chunk);
665 /* Remove a list of addresses from bind addresses list. Do not remove the
668 * Basically run through each address specified in the addrs/addrcnt
669 * array/length pair, determine if it is IPv6 or IPv4 and call
670 * sctp_del_bind() on it.
672 * If any of them fails, then the operation will be reversed and the
673 * ones that were removed will be added back.
675 * At least one address has to be left; if only one address is
676 * available, the operation will return -EBUSY.
678 * Only sctp_setsockopt_bindx() is supposed to call this function.
680 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
682 struct sctp_sock *sp = sctp_sk(sk);
683 struct sctp_endpoint *ep = sp->ep;
685 struct sctp_bind_addr *bp = &ep->base.bind_addr;
688 union sctp_addr *sa_addr;
691 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
692 __func__, sk, addrs, addrcnt);
695 for (cnt = 0; cnt < addrcnt; cnt++) {
696 /* If the bind address list is empty or if there is only one
697 * bind address, there is nothing more to be removed (we need
698 * at least one address here).
700 if (list_empty(&bp->address_list) ||
701 (sctp_list_single_entry(&bp->address_list))) {
707 af = sctp_get_af_specific(sa_addr->sa.sa_family);
713 if (!af->addr_valid(sa_addr, sp, NULL)) {
714 retval = -EADDRNOTAVAIL;
718 if (sa_addr->v4.sin_port &&
719 sa_addr->v4.sin_port != htons(bp->port)) {
724 if (!sa_addr->v4.sin_port)
725 sa_addr->v4.sin_port = htons(bp->port);
727 /* FIXME - There is probably a need to check if sk->sk_saddr and
728 * sk->sk_rcv_addr are currently set to one of the addresses to
729 * be removed. This is something which needs to be looked into
730 * when we are fixing the outstanding issues with multi-homing
731 * socket routing and failover schemes. Refer to comments in
732 * sctp_do_bind(). -daisy
734 retval = sctp_del_bind_addr(bp, sa_addr);
736 addr_buf += af->sockaddr_len;
739 /* Failed. Add the ones that has been removed back */
741 sctp_bindx_add(sk, addrs, cnt);
749 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
750 * the associations that are part of the endpoint indicating that a list of
751 * local addresses are removed from the endpoint.
753 * If any of the addresses is already in the bind address list of the
754 * association, we do not send the chunk for that association. But it will not
755 * affect other associations.
757 * Only sctp_setsockopt_bindx() is supposed to call this function.
759 static int sctp_send_asconf_del_ip(struct sock *sk,
760 struct sockaddr *addrs,
763 struct net *net = sock_net(sk);
764 struct sctp_sock *sp;
765 struct sctp_endpoint *ep;
766 struct sctp_association *asoc;
767 struct sctp_transport *transport;
768 struct sctp_bind_addr *bp;
769 struct sctp_chunk *chunk;
770 union sctp_addr *laddr;
773 struct sctp_sockaddr_entry *saddr;
779 if (!net->sctp.addip_enable)
785 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
786 __func__, sk, addrs, addrcnt);
788 list_for_each_entry(asoc, &ep->asocs, asocs) {
790 if (!asoc->peer.asconf_capable)
793 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
796 if (!sctp_state(asoc, ESTABLISHED))
799 /* Check if any address in the packed array of addresses is
800 * not present in the bind address list of the association.
801 * If so, do not send the asconf chunk to its peer, but
802 * continue with other associations.
805 for (i = 0; i < addrcnt; i++) {
807 af = sctp_get_af_specific(laddr->v4.sin_family);
813 if (!sctp_assoc_lookup_laddr(asoc, laddr))
816 addr_buf += af->sockaddr_len;
821 /* Find one address in the association's bind address list
822 * that is not in the packed array of addresses. This is to
823 * make sure that we do not delete all the addresses in the
826 bp = &asoc->base.bind_addr;
827 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
829 if ((laddr == NULL) && (addrcnt == 1)) {
830 if (asoc->asconf_addr_del_pending)
832 asoc->asconf_addr_del_pending =
833 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
834 if (asoc->asconf_addr_del_pending == NULL) {
838 asoc->asconf_addr_del_pending->sa.sa_family =
840 asoc->asconf_addr_del_pending->v4.sin_port =
842 if (addrs->sa_family == AF_INET) {
843 struct sockaddr_in *sin;
845 sin = (struct sockaddr_in *)addrs;
846 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
847 } else if (addrs->sa_family == AF_INET6) {
848 struct sockaddr_in6 *sin6;
850 sin6 = (struct sockaddr_in6 *)addrs;
851 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
854 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
855 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
856 asoc->asconf_addr_del_pending);
858 asoc->src_out_of_asoc_ok = 1;
866 /* We do not need RCU protection throughout this loop
867 * because this is done under a socket lock from the
870 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
878 /* Reset use_as_src flag for the addresses in the bind address
879 * list that are to be deleted.
882 for (i = 0; i < addrcnt; i++) {
884 af = sctp_get_af_specific(laddr->v4.sin_family);
885 list_for_each_entry(saddr, &bp->address_list, list) {
886 if (sctp_cmp_addr_exact(&saddr->a, laddr))
887 saddr->state = SCTP_ADDR_DEL;
889 addr_buf += af->sockaddr_len;
892 /* Update the route and saddr entries for all the transports
893 * as some of the addresses in the bind address list are
894 * about to be deleted and cannot be used as source addresses.
896 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
898 sctp_transport_route(transport, NULL,
899 sctp_sk(asoc->base.sk));
903 /* We don't need to transmit ASCONF */
905 retval = sctp_send_asconf(asoc, chunk);
911 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
912 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
914 struct sock *sk = sctp_opt2sk(sp);
915 union sctp_addr *addr;
918 /* It is safe to write port space in caller. */
920 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
921 af = sctp_get_af_specific(addr->sa.sa_family);
924 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
927 if (addrw->state == SCTP_ADDR_NEW)
928 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
930 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
933 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
936 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
939 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
940 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
943 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
944 * Section 3.1.2 for this usage.
946 * addrs is a pointer to an array of one or more socket addresses. Each
947 * address is contained in its appropriate structure (i.e. struct
948 * sockaddr_in or struct sockaddr_in6) the family of the address type
949 * must be used to distinguish the address length (note that this
950 * representation is termed a "packed array" of addresses). The caller
951 * specifies the number of addresses in the array with addrcnt.
953 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
954 * -1, and sets errno to the appropriate error code.
956 * For SCTP, the port given in each socket address must be the same, or
957 * sctp_bindx() will fail, setting errno to EINVAL.
959 * The flags parameter is formed from the bitwise OR of zero or more of
960 * the following currently defined flags:
962 * SCTP_BINDX_ADD_ADDR
964 * SCTP_BINDX_REM_ADDR
966 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
967 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
968 * addresses from the association. The two flags are mutually exclusive;
969 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
970 * not remove all addresses from an association; sctp_bindx() will
971 * reject such an attempt with EINVAL.
973 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
974 * additional addresses with an endpoint after calling bind(). Or use
975 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
976 * socket is associated with so that no new association accepted will be
977 * associated with those addresses. If the endpoint supports dynamic
978 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
979 * endpoint to send the appropriate message to the peer to change the
980 * peers address lists.
982 * Adding and removing addresses from a connected association is
983 * optional functionality. Implementations that do not support this
984 * functionality should return EOPNOTSUPP.
986 * Basically do nothing but copying the addresses from user to kernel
987 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
988 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
991 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
994 * sk The sk of the socket
995 * addrs The pointer to the addresses in user land
996 * addrssize Size of the addrs buffer
997 * op Operation to perform (add or remove, see the flags of
1000 * Returns 0 if ok, <0 errno code on error.
1002 static int sctp_setsockopt_bindx(struct sock *sk,
1003 struct sockaddr __user *addrs,
1004 int addrs_size, int op)
1006 struct sockaddr *kaddrs;
1010 struct sockaddr *sa_addr;
1014 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
1015 __func__, sk, addrs, addrs_size, op);
1017 if (unlikely(addrs_size <= 0))
1020 kaddrs = vmemdup_user(addrs, addrs_size);
1021 if (unlikely(IS_ERR(kaddrs)))
1022 return PTR_ERR(kaddrs);
1024 /* Walk through the addrs buffer and count the number of addresses. */
1026 while (walk_size < addrs_size) {
1027 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1033 af = sctp_get_af_specific(sa_addr->sa_family);
1035 /* If the address family is not supported or if this address
1036 * causes the address buffer to overflow return EINVAL.
1038 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1043 addr_buf += af->sockaddr_len;
1044 walk_size += af->sockaddr_len;
1049 case SCTP_BINDX_ADD_ADDR:
1050 /* Allow security module to validate bindx addresses. */
1051 err = security_sctp_bind_connect(sk, SCTP_SOCKOPT_BINDX_ADD,
1052 (struct sockaddr *)kaddrs,
1056 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1059 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1062 case SCTP_BINDX_REM_ADDR:
1063 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1066 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1080 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1082 * Common routine for handling connect() and sctp_connectx().
1083 * Connect will come in with just a single address.
1085 static int __sctp_connect(struct sock *sk,
1086 struct sockaddr *kaddrs,
1087 int addrs_size, int flags,
1088 sctp_assoc_t *assoc_id)
1090 struct net *net = sock_net(sk);
1091 struct sctp_sock *sp;
1092 struct sctp_endpoint *ep;
1093 struct sctp_association *asoc = NULL;
1094 struct sctp_association *asoc2;
1095 struct sctp_transport *transport;
1097 enum sctp_scope scope;
1102 union sctp_addr *sa_addr = NULL;
1104 unsigned short port;
1109 /* connect() cannot be done on a socket that is already in ESTABLISHED
1110 * state - UDP-style peeled off socket or a TCP-style socket that
1111 * is already connected.
1112 * It cannot be done even on a TCP-style listening socket.
1114 if (sctp_sstate(sk, ESTABLISHED) || sctp_sstate(sk, CLOSING) ||
1115 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1120 /* Walk through the addrs buffer and count the number of addresses. */
1122 while (walk_size < addrs_size) {
1125 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1131 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1133 /* If the address family is not supported or if this address
1134 * causes the address buffer to overflow return EINVAL.
1136 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1141 port = ntohs(sa_addr->v4.sin_port);
1143 /* Save current address so we can work with it */
1144 memcpy(&to, sa_addr, af->sockaddr_len);
1146 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1150 /* Make sure the destination port is correctly set
1153 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1158 /* Check if there already is a matching association on the
1159 * endpoint (other than the one created here).
1161 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1162 if (asoc2 && asoc2 != asoc) {
1163 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1170 /* If we could not find a matching association on the endpoint,
1171 * make sure that there is no peeled-off association matching
1172 * the peer address even on another socket.
1174 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1175 err = -EADDRNOTAVAIL;
1180 /* If a bind() or sctp_bindx() is not called prior to
1181 * an sctp_connectx() call, the system picks an
1182 * ephemeral port and will choose an address set
1183 * equivalent to binding with a wildcard address.
1185 if (!ep->base.bind_addr.port) {
1186 if (sctp_autobind(sk)) {
1192 * If an unprivileged user inherits a 1-many
1193 * style socket with open associations on a
1194 * privileged port, it MAY be permitted to
1195 * accept new associations, but it SHOULD NOT
1196 * be permitted to open new associations.
1198 if (ep->base.bind_addr.port <
1199 inet_prot_sock(net) &&
1200 !ns_capable(net->user_ns,
1201 CAP_NET_BIND_SERVICE)) {
1207 scope = sctp_scope(&to);
1208 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1214 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1222 /* Prime the peer's transport structures. */
1223 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1231 addr_buf += af->sockaddr_len;
1232 walk_size += af->sockaddr_len;
1235 /* In case the user of sctp_connectx() wants an association
1236 * id back, assign one now.
1239 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1244 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1249 /* Initialize sk's dport and daddr for getpeername() */
1250 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1251 sp->pf->to_sk_daddr(sa_addr, sk);
1254 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
1257 *assoc_id = asoc->assoc_id;
1259 err = sctp_wait_for_connect(asoc, &timeo);
1260 /* Note: the asoc may be freed after the return of
1261 * sctp_wait_for_connect.
1264 /* Don't free association on exit. */
1268 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1269 __func__, asoc, kaddrs, err);
1272 /* sctp_primitive_ASSOCIATE may have added this association
1273 * To the hash table, try to unhash it, just in case, its a noop
1274 * if it wasn't hashed so we're safe
1276 sctp_association_free(asoc);
1281 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1284 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1285 * sctp_assoc_t *asoc);
1287 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1288 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1289 * or IPv6 addresses.
1291 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1292 * Section 3.1.2 for this usage.
1294 * addrs is a pointer to an array of one or more socket addresses. Each
1295 * address is contained in its appropriate structure (i.e. struct
1296 * sockaddr_in or struct sockaddr_in6) the family of the address type
1297 * must be used to distengish the address length (note that this
1298 * representation is termed a "packed array" of addresses). The caller
1299 * specifies the number of addresses in the array with addrcnt.
1301 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1302 * the association id of the new association. On failure, sctp_connectx()
1303 * returns -1, and sets errno to the appropriate error code. The assoc_id
1304 * is not touched by the kernel.
1306 * For SCTP, the port given in each socket address must be the same, or
1307 * sctp_connectx() will fail, setting errno to EINVAL.
1309 * An application can use sctp_connectx to initiate an association with
1310 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1311 * allows a caller to specify multiple addresses at which a peer can be
1312 * reached. The way the SCTP stack uses the list of addresses to set up
1313 * the association is implementation dependent. This function only
1314 * specifies that the stack will try to make use of all the addresses in
1315 * the list when needed.
1317 * Note that the list of addresses passed in is only used for setting up
1318 * the association. It does not necessarily equal the set of addresses
1319 * the peer uses for the resulting association. If the caller wants to
1320 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1321 * retrieve them after the association has been set up.
1323 * Basically do nothing but copying the addresses from user to kernel
1324 * land and invoking either sctp_connectx(). This is used for tunneling
1325 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1327 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1330 * sk The sk of the socket
1331 * addrs The pointer to the addresses in user land
1332 * addrssize Size of the addrs buffer
1334 * Returns >=0 if ok, <0 errno code on error.
1336 static int __sctp_setsockopt_connectx(struct sock *sk,
1337 struct sockaddr __user *addrs,
1339 sctp_assoc_t *assoc_id)
1341 struct sockaddr *kaddrs;
1342 int err = 0, flags = 0;
1344 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1345 __func__, sk, addrs, addrs_size);
1347 if (unlikely(addrs_size <= 0))
1350 kaddrs = vmemdup_user(addrs, addrs_size);
1351 if (unlikely(IS_ERR(kaddrs)))
1352 return PTR_ERR(kaddrs);
1354 /* Allow security module to validate connectx addresses. */
1355 err = security_sctp_bind_connect(sk, SCTP_SOCKOPT_CONNECTX,
1356 (struct sockaddr *)kaddrs,
1361 /* in-kernel sockets don't generally have a file allocated to them
1362 * if all they do is call sock_create_kern().
1364 if (sk->sk_socket->file)
1365 flags = sk->sk_socket->file->f_flags;
1367 err = __sctp_connect(sk, kaddrs, addrs_size, flags, assoc_id);
1376 * This is an older interface. It's kept for backward compatibility
1377 * to the option that doesn't provide association id.
1379 static int sctp_setsockopt_connectx_old(struct sock *sk,
1380 struct sockaddr __user *addrs,
1383 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1387 * New interface for the API. The since the API is done with a socket
1388 * option, to make it simple we feed back the association id is as a return
1389 * indication to the call. Error is always negative and association id is
1392 static int sctp_setsockopt_connectx(struct sock *sk,
1393 struct sockaddr __user *addrs,
1396 sctp_assoc_t assoc_id = 0;
1399 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1408 * New (hopefully final) interface for the API.
1409 * We use the sctp_getaddrs_old structure so that use-space library
1410 * can avoid any unnecessary allocations. The only different part
1411 * is that we store the actual length of the address buffer into the
1412 * addrs_num structure member. That way we can re-use the existing
1415 #ifdef CONFIG_COMPAT
1416 struct compat_sctp_getaddrs_old {
1417 sctp_assoc_t assoc_id;
1419 compat_uptr_t addrs; /* struct sockaddr * */
1423 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1424 char __user *optval,
1427 struct sctp_getaddrs_old param;
1428 sctp_assoc_t assoc_id = 0;
1431 #ifdef CONFIG_COMPAT
1432 if (in_compat_syscall()) {
1433 struct compat_sctp_getaddrs_old param32;
1435 if (len < sizeof(param32))
1437 if (copy_from_user(¶m32, optval, sizeof(param32)))
1440 param.assoc_id = param32.assoc_id;
1441 param.addr_num = param32.addr_num;
1442 param.addrs = compat_ptr(param32.addrs);
1446 if (len < sizeof(param))
1448 if (copy_from_user(¶m, optval, sizeof(param)))
1452 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1453 param.addrs, param.addr_num,
1455 if (err == 0 || err == -EINPROGRESS) {
1456 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1458 if (put_user(sizeof(assoc_id), optlen))
1465 /* API 3.1.4 close() - UDP Style Syntax
1466 * Applications use close() to perform graceful shutdown (as described in
1467 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1468 * by a UDP-style socket.
1472 * ret = close(int sd);
1474 * sd - the socket descriptor of the associations to be closed.
1476 * To gracefully shutdown a specific association represented by the
1477 * UDP-style socket, an application should use the sendmsg() call,
1478 * passing no user data, but including the appropriate flag in the
1479 * ancillary data (see Section xxxx).
1481 * If sd in the close() call is a branched-off socket representing only
1482 * one association, the shutdown is performed on that association only.
1484 * 4.1.6 close() - TCP Style Syntax
1486 * Applications use close() to gracefully close down an association.
1490 * int close(int sd);
1492 * sd - the socket descriptor of the association to be closed.
1494 * After an application calls close() on a socket descriptor, no further
1495 * socket operations will succeed on that descriptor.
1497 * API 7.1.4 SO_LINGER
1499 * An application using the TCP-style socket can use this option to
1500 * perform the SCTP ABORT primitive. The linger option structure is:
1503 * int l_onoff; // option on/off
1504 * int l_linger; // linger time
1507 * To enable the option, set l_onoff to 1. If the l_linger value is set
1508 * to 0, calling close() is the same as the ABORT primitive. If the
1509 * value is set to a negative value, the setsockopt() call will return
1510 * an error. If the value is set to a positive value linger_time, the
1511 * close() can be blocked for at most linger_time ms. If the graceful
1512 * shutdown phase does not finish during this period, close() will
1513 * return but the graceful shutdown phase continues in the system.
1515 static void sctp_close(struct sock *sk, long timeout)
1517 struct net *net = sock_net(sk);
1518 struct sctp_endpoint *ep;
1519 struct sctp_association *asoc;
1520 struct list_head *pos, *temp;
1521 unsigned int data_was_unread;
1523 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1525 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
1526 sk->sk_shutdown = SHUTDOWN_MASK;
1527 inet_sk_set_state(sk, SCTP_SS_CLOSING);
1529 ep = sctp_sk(sk)->ep;
1531 /* Clean up any skbs sitting on the receive queue. */
1532 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1533 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1535 /* Walk all associations on an endpoint. */
1536 list_for_each_safe(pos, temp, &ep->asocs) {
1537 asoc = list_entry(pos, struct sctp_association, asocs);
1539 if (sctp_style(sk, TCP)) {
1540 /* A closed association can still be in the list if
1541 * it belongs to a TCP-style listening socket that is
1542 * not yet accepted. If so, free it. If not, send an
1543 * ABORT or SHUTDOWN based on the linger options.
1545 if (sctp_state(asoc, CLOSED)) {
1546 sctp_association_free(asoc);
1551 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1552 !skb_queue_empty(&asoc->ulpq.reasm) ||
1553 !skb_queue_empty(&asoc->ulpq.reasm_uo) ||
1554 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1555 struct sctp_chunk *chunk;
1557 chunk = sctp_make_abort_user(asoc, NULL, 0);
1558 sctp_primitive_ABORT(net, asoc, chunk);
1560 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1563 /* On a TCP-style socket, block for at most linger_time if set. */
1564 if (sctp_style(sk, TCP) && timeout)
1565 sctp_wait_for_close(sk, timeout);
1567 /* This will run the backlog queue. */
1570 /* Supposedly, no process has access to the socket, but
1571 * the net layers still may.
1572 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1573 * held and that should be grabbed before socket lock.
1575 spin_lock_bh(&net->sctp.addr_wq_lock);
1576 bh_lock_sock_nested(sk);
1578 /* Hold the sock, since sk_common_release() will put sock_put()
1579 * and we have just a little more cleanup.
1582 sk_common_release(sk);
1585 spin_unlock_bh(&net->sctp.addr_wq_lock);
1589 SCTP_DBG_OBJCNT_DEC(sock);
1592 /* Handle EPIPE error. */
1593 static int sctp_error(struct sock *sk, int flags, int err)
1596 err = sock_error(sk) ? : -EPIPE;
1597 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1598 send_sig(SIGPIPE, current, 0);
1602 /* API 3.1.3 sendmsg() - UDP Style Syntax
1604 * An application uses sendmsg() and recvmsg() calls to transmit data to
1605 * and receive data from its peer.
1607 * ssize_t sendmsg(int socket, const struct msghdr *message,
1610 * socket - the socket descriptor of the endpoint.
1611 * message - pointer to the msghdr structure which contains a single
1612 * user message and possibly some ancillary data.
1614 * See Section 5 for complete description of the data
1617 * flags - flags sent or received with the user message, see Section
1618 * 5 for complete description of the flags.
1620 * Note: This function could use a rewrite especially when explicit
1621 * connect support comes in.
1623 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1625 static int sctp_msghdr_parse(const struct msghdr *msg,
1626 struct sctp_cmsgs *cmsgs);
1628 static int sctp_sendmsg_parse(struct sock *sk, struct sctp_cmsgs *cmsgs,
1629 struct sctp_sndrcvinfo *srinfo,
1630 const struct msghdr *msg, size_t msg_len)
1635 if (sctp_sstate(sk, LISTENING) && sctp_style(sk, TCP))
1638 if (msg_len > sk->sk_sndbuf)
1641 memset(cmsgs, 0, sizeof(*cmsgs));
1642 err = sctp_msghdr_parse(msg, cmsgs);
1644 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1648 memset(srinfo, 0, sizeof(*srinfo));
1649 if (cmsgs->srinfo) {
1650 srinfo->sinfo_stream = cmsgs->srinfo->sinfo_stream;
1651 srinfo->sinfo_flags = cmsgs->srinfo->sinfo_flags;
1652 srinfo->sinfo_ppid = cmsgs->srinfo->sinfo_ppid;
1653 srinfo->sinfo_context = cmsgs->srinfo->sinfo_context;
1654 srinfo->sinfo_assoc_id = cmsgs->srinfo->sinfo_assoc_id;
1655 srinfo->sinfo_timetolive = cmsgs->srinfo->sinfo_timetolive;
1659 srinfo->sinfo_stream = cmsgs->sinfo->snd_sid;
1660 srinfo->sinfo_flags = cmsgs->sinfo->snd_flags;
1661 srinfo->sinfo_ppid = cmsgs->sinfo->snd_ppid;
1662 srinfo->sinfo_context = cmsgs->sinfo->snd_context;
1663 srinfo->sinfo_assoc_id = cmsgs->sinfo->snd_assoc_id;
1666 if (cmsgs->prinfo) {
1667 srinfo->sinfo_timetolive = cmsgs->prinfo->pr_value;
1668 SCTP_PR_SET_POLICY(srinfo->sinfo_flags,
1669 cmsgs->prinfo->pr_policy);
1672 sflags = srinfo->sinfo_flags;
1673 if (!sflags && msg_len)
1676 if (sctp_style(sk, TCP) && (sflags & (SCTP_EOF | SCTP_ABORT)))
1679 if (((sflags & SCTP_EOF) && msg_len > 0) ||
1680 (!(sflags & (SCTP_EOF | SCTP_ABORT)) && msg_len == 0))
1683 if ((sflags & SCTP_ADDR_OVER) && !msg->msg_name)
1689 static int sctp_sendmsg_new_asoc(struct sock *sk, __u16 sflags,
1690 struct sctp_cmsgs *cmsgs,
1691 union sctp_addr *daddr,
1692 struct sctp_transport **tp)
1694 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
1695 struct net *net = sock_net(sk);
1696 struct sctp_association *asoc;
1697 enum sctp_scope scope;
1698 struct cmsghdr *cmsg;
1699 __be32 flowinfo = 0;
1705 if (sflags & (SCTP_EOF | SCTP_ABORT))
1708 if (sctp_style(sk, TCP) && (sctp_sstate(sk, ESTABLISHED) ||
1709 sctp_sstate(sk, CLOSING)))
1710 return -EADDRNOTAVAIL;
1712 if (sctp_endpoint_is_peeled_off(ep, daddr))
1713 return -EADDRNOTAVAIL;
1715 if (!ep->base.bind_addr.port) {
1716 if (sctp_autobind(sk))
1719 if (ep->base.bind_addr.port < inet_prot_sock(net) &&
1720 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
1724 scope = sctp_scope(daddr);
1726 /* Label connection socket for first association 1-to-many
1727 * style for client sequence socket()->sendmsg(). This
1728 * needs to be done before sctp_assoc_add_peer() as that will
1729 * set up the initial packet that needs to account for any
1730 * security ip options (CIPSO/CALIPSO) added to the packet.
1732 af = sctp_get_af_specific(daddr->sa.sa_family);
1735 err = security_sctp_bind_connect(sk, SCTP_SENDMSG_CONNECT,
1736 (struct sockaddr *)daddr,
1741 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1745 if (sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL) < 0) {
1751 struct sctp_initmsg *init = cmsgs->init;
1753 if (init->sinit_num_ostreams) {
1754 __u16 outcnt = init->sinit_num_ostreams;
1756 asoc->c.sinit_num_ostreams = outcnt;
1757 /* outcnt has been changed, need to re-init stream */
1758 err = sctp_stream_init(&asoc->stream, outcnt, 0,
1764 if (init->sinit_max_instreams)
1765 asoc->c.sinit_max_instreams = init->sinit_max_instreams;
1767 if (init->sinit_max_attempts)
1768 asoc->max_init_attempts = init->sinit_max_attempts;
1770 if (init->sinit_max_init_timeo)
1771 asoc->max_init_timeo =
1772 msecs_to_jiffies(init->sinit_max_init_timeo);
1775 *tp = sctp_assoc_add_peer(asoc, daddr, GFP_KERNEL, SCTP_UNKNOWN);
1781 if (!cmsgs->addrs_msg)
1784 if (daddr->sa.sa_family == AF_INET6)
1785 flowinfo = daddr->v6.sin6_flowinfo;
1787 /* sendv addr list parse */
1788 for_each_cmsghdr(cmsg, cmsgs->addrs_msg) {
1789 struct sctp_transport *transport;
1790 struct sctp_association *old;
1791 union sctp_addr _daddr;
1794 if (cmsg->cmsg_level != IPPROTO_SCTP ||
1795 (cmsg->cmsg_type != SCTP_DSTADDRV4 &&
1796 cmsg->cmsg_type != SCTP_DSTADDRV6))
1800 memset(daddr, 0, sizeof(*daddr));
1801 dlen = cmsg->cmsg_len - sizeof(struct cmsghdr);
1802 if (cmsg->cmsg_type == SCTP_DSTADDRV4) {
1803 if (dlen < sizeof(struct in_addr)) {
1808 dlen = sizeof(struct in_addr);
1809 daddr->v4.sin_family = AF_INET;
1810 daddr->v4.sin_port = htons(asoc->peer.port);
1811 memcpy(&daddr->v4.sin_addr, CMSG_DATA(cmsg), dlen);
1813 if (dlen < sizeof(struct in6_addr)) {
1818 dlen = sizeof(struct in6_addr);
1819 daddr->v6.sin6_flowinfo = flowinfo;
1820 daddr->v6.sin6_family = AF_INET6;
1821 daddr->v6.sin6_port = htons(asoc->peer.port);
1822 memcpy(&daddr->v6.sin6_addr, CMSG_DATA(cmsg), dlen);
1824 err = sctp_verify_addr(sk, daddr, sizeof(*daddr));
1828 old = sctp_endpoint_lookup_assoc(ep, daddr, &transport);
1829 if (old && old != asoc) {
1830 if (old->state >= SCTP_STATE_ESTABLISHED)
1837 if (sctp_endpoint_is_peeled_off(ep, daddr)) {
1838 err = -EADDRNOTAVAIL;
1842 transport = sctp_assoc_add_peer(asoc, daddr, GFP_KERNEL,
1853 sctp_association_free(asoc);
1857 static int sctp_sendmsg_check_sflags(struct sctp_association *asoc,
1858 __u16 sflags, struct msghdr *msg,
1861 struct sock *sk = asoc->base.sk;
1862 struct net *net = sock_net(sk);
1864 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP))
1867 if ((sflags & SCTP_SENDALL) && sctp_style(sk, UDP) &&
1868 !sctp_state(asoc, ESTABLISHED))
1871 if (sflags & SCTP_EOF) {
1872 pr_debug("%s: shutting down association:%p\n", __func__, asoc);
1873 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1878 if (sflags & SCTP_ABORT) {
1879 struct sctp_chunk *chunk;
1881 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1885 pr_debug("%s: aborting association:%p\n", __func__, asoc);
1886 sctp_primitive_ABORT(net, asoc, chunk);
1894 static int sctp_sendmsg_to_asoc(struct sctp_association *asoc,
1895 struct msghdr *msg, size_t msg_len,
1896 struct sctp_transport *transport,
1897 struct sctp_sndrcvinfo *sinfo)
1899 struct sock *sk = asoc->base.sk;
1900 struct sctp_sock *sp = sctp_sk(sk);
1901 struct net *net = sock_net(sk);
1902 struct sctp_datamsg *datamsg;
1903 bool wait_connect = false;
1904 struct sctp_chunk *chunk;
1908 if (sinfo->sinfo_stream >= asoc->stream.outcnt) {
1913 if (unlikely(!SCTP_SO(&asoc->stream, sinfo->sinfo_stream)->ext)) {
1914 err = sctp_stream_init_ext(&asoc->stream, sinfo->sinfo_stream);
1919 if (sp->disable_fragments && msg_len > asoc->frag_point) {
1924 if (asoc->pmtu_pending) {
1925 if (sp->param_flags & SPP_PMTUD_ENABLE)
1926 sctp_assoc_sync_pmtu(asoc);
1927 asoc->pmtu_pending = 0;
1930 if (sctp_wspace(asoc) < msg_len)
1931 sctp_prsctp_prune(asoc, sinfo, msg_len - sctp_wspace(asoc));
1933 if (!sctp_wspace(asoc)) {
1934 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1935 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1940 if (sctp_state(asoc, CLOSED)) {
1941 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1945 if (sp->strm_interleave) {
1946 timeo = sock_sndtimeo(sk, 0);
1947 err = sctp_wait_for_connect(asoc, &timeo);
1953 wait_connect = true;
1956 pr_debug("%s: we associated primitively\n", __func__);
1959 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
1960 if (IS_ERR(datamsg)) {
1961 err = PTR_ERR(datamsg);
1965 asoc->force_delay = !!(msg->msg_flags & MSG_MORE);
1967 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1968 sctp_chunk_hold(chunk);
1969 sctp_set_owner_w(chunk);
1970 chunk->transport = transport;
1973 err = sctp_primitive_SEND(net, asoc, datamsg);
1975 sctp_datamsg_free(datamsg);
1979 pr_debug("%s: we sent primitively\n", __func__);
1981 sctp_datamsg_put(datamsg);
1983 if (unlikely(wait_connect)) {
1984 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1985 sctp_wait_for_connect(asoc, &timeo);
1994 static union sctp_addr *sctp_sendmsg_get_daddr(struct sock *sk,
1995 const struct msghdr *msg,
1996 struct sctp_cmsgs *cmsgs)
1998 union sctp_addr *daddr = NULL;
2001 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
2002 int len = msg->msg_namelen;
2004 if (len > sizeof(*daddr))
2005 len = sizeof(*daddr);
2007 daddr = (union sctp_addr *)msg->msg_name;
2009 err = sctp_verify_addr(sk, daddr, len);
2011 return ERR_PTR(err);
2017 static void sctp_sendmsg_update_sinfo(struct sctp_association *asoc,
2018 struct sctp_sndrcvinfo *sinfo,
2019 struct sctp_cmsgs *cmsgs)
2021 if (!cmsgs->srinfo && !cmsgs->sinfo) {
2022 sinfo->sinfo_stream = asoc->default_stream;
2023 sinfo->sinfo_ppid = asoc->default_ppid;
2024 sinfo->sinfo_context = asoc->default_context;
2025 sinfo->sinfo_assoc_id = sctp_assoc2id(asoc);
2028 sinfo->sinfo_flags = asoc->default_flags;
2031 if (!cmsgs->srinfo && !cmsgs->prinfo)
2032 sinfo->sinfo_timetolive = asoc->default_timetolive;
2034 if (cmsgs->authinfo) {
2035 /* Reuse sinfo_tsn to indicate that authinfo was set and
2036 * sinfo_ssn to save the keyid on tx path.
2038 sinfo->sinfo_tsn = 1;
2039 sinfo->sinfo_ssn = cmsgs->authinfo->auth_keynumber;
2043 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
2045 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
2046 struct sctp_transport *transport = NULL;
2047 struct sctp_sndrcvinfo _sinfo, *sinfo;
2048 struct sctp_association *asoc;
2049 struct sctp_cmsgs cmsgs;
2050 union sctp_addr *daddr;
2055 /* Parse and get snd_info */
2056 err = sctp_sendmsg_parse(sk, &cmsgs, &_sinfo, msg, msg_len);
2061 sflags = sinfo->sinfo_flags;
2063 /* Get daddr from msg */
2064 daddr = sctp_sendmsg_get_daddr(sk, msg, &cmsgs);
2065 if (IS_ERR(daddr)) {
2066 err = PTR_ERR(daddr);
2072 /* SCTP_SENDALL process */
2073 if ((sflags & SCTP_SENDALL) && sctp_style(sk, UDP)) {
2074 list_for_each_entry(asoc, &ep->asocs, asocs) {
2075 err = sctp_sendmsg_check_sflags(asoc, sflags, msg,
2082 sctp_sendmsg_update_sinfo(asoc, sinfo, &cmsgs);
2084 err = sctp_sendmsg_to_asoc(asoc, msg, msg_len,
2089 iov_iter_revert(&msg->msg_iter, err);
2095 /* Get and check or create asoc */
2097 asoc = sctp_endpoint_lookup_assoc(ep, daddr, &transport);
2099 err = sctp_sendmsg_check_sflags(asoc, sflags, msg,
2104 err = sctp_sendmsg_new_asoc(sk, sflags, &cmsgs, daddr,
2109 asoc = transport->asoc;
2113 if (!sctp_style(sk, TCP) && !(sflags & SCTP_ADDR_OVER))
2116 asoc = sctp_id2assoc(sk, sinfo->sinfo_assoc_id);
2122 err = sctp_sendmsg_check_sflags(asoc, sflags, msg, msg_len);
2127 /* Update snd_info with the asoc */
2128 sctp_sendmsg_update_sinfo(asoc, sinfo, &cmsgs);
2130 /* Send msg to the asoc */
2131 err = sctp_sendmsg_to_asoc(asoc, msg, msg_len, transport, sinfo);
2132 if (err < 0 && err != -ESRCH && new)
2133 sctp_association_free(asoc);
2138 return sctp_error(sk, msg->msg_flags, err);
2141 /* This is an extended version of skb_pull() that removes the data from the
2142 * start of a skb even when data is spread across the list of skb's in the
2143 * frag_list. len specifies the total amount of data that needs to be removed.
2144 * when 'len' bytes could be removed from the skb, it returns 0.
2145 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2146 * could not be removed.
2148 static int sctp_skb_pull(struct sk_buff *skb, int len)
2150 struct sk_buff *list;
2151 int skb_len = skb_headlen(skb);
2154 if (len <= skb_len) {
2155 __skb_pull(skb, len);
2159 __skb_pull(skb, skb_len);
2161 skb_walk_frags(skb, list) {
2162 rlen = sctp_skb_pull(list, len);
2163 skb->len -= (len-rlen);
2164 skb->data_len -= (len-rlen);
2175 /* API 3.1.3 recvmsg() - UDP Style Syntax
2177 * ssize_t recvmsg(int socket, struct msghdr *message,
2180 * socket - the socket descriptor of the endpoint.
2181 * message - pointer to the msghdr structure which contains a single
2182 * user message and possibly some ancillary data.
2184 * See Section 5 for complete description of the data
2187 * flags - flags sent or received with the user message, see Section
2188 * 5 for complete description of the flags.
2190 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2191 int noblock, int flags, int *addr_len)
2193 struct sctp_ulpevent *event = NULL;
2194 struct sctp_sock *sp = sctp_sk(sk);
2195 struct sk_buff *skb, *head_skb;
2200 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2201 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2206 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED) &&
2207 !sctp_sstate(sk, CLOSING) && !sctp_sstate(sk, CLOSED)) {
2212 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2216 /* Get the total length of the skb including any skb's in the
2225 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2227 event = sctp_skb2event(skb);
2232 if (event->chunk && event->chunk->head_skb)
2233 head_skb = event->chunk->head_skb;
2236 sock_recv_ts_and_drops(msg, sk, head_skb);
2237 if (sctp_ulpevent_is_notification(event)) {
2238 msg->msg_flags |= MSG_NOTIFICATION;
2239 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2241 sp->pf->skb_msgname(head_skb, msg->msg_name, addr_len);
2244 /* Check if we allow SCTP_NXTINFO. */
2245 if (sp->recvnxtinfo)
2246 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2247 /* Check if we allow SCTP_RCVINFO. */
2248 if (sp->recvrcvinfo)
2249 sctp_ulpevent_read_rcvinfo(event, msg);
2250 /* Check if we allow SCTP_SNDRCVINFO. */
2251 if (sp->subscribe.sctp_data_io_event)
2252 sctp_ulpevent_read_sndrcvinfo(event, msg);
2256 /* If skb's length exceeds the user's buffer, update the skb and
2257 * push it back to the receive_queue so that the next call to
2258 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2260 if (skb_len > copied) {
2261 msg->msg_flags &= ~MSG_EOR;
2262 if (flags & MSG_PEEK)
2264 sctp_skb_pull(skb, copied);
2265 skb_queue_head(&sk->sk_receive_queue, skb);
2267 /* When only partial message is copied to the user, increase
2268 * rwnd by that amount. If all the data in the skb is read,
2269 * rwnd is updated when the event is freed.
2271 if (!sctp_ulpevent_is_notification(event))
2272 sctp_assoc_rwnd_increase(event->asoc, copied);
2274 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2275 (event->msg_flags & MSG_EOR))
2276 msg->msg_flags |= MSG_EOR;
2278 msg->msg_flags &= ~MSG_EOR;
2281 if (flags & MSG_PEEK) {
2282 /* Release the skb reference acquired after peeking the skb in
2283 * sctp_skb_recv_datagram().
2287 /* Free the event which includes releasing the reference to
2288 * the owner of the skb, freeing the skb and updating the
2291 sctp_ulpevent_free(event);
2298 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2300 * This option is a on/off flag. If enabled no SCTP message
2301 * fragmentation will be performed. Instead if a message being sent
2302 * exceeds the current PMTU size, the message will NOT be sent and
2303 * instead a error will be indicated to the user.
2305 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2306 char __user *optval,
2307 unsigned int optlen)
2311 if (optlen < sizeof(int))
2314 if (get_user(val, (int __user *)optval))
2317 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2322 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2323 unsigned int optlen)
2325 struct sctp_association *asoc;
2326 struct sctp_ulpevent *event;
2328 if (optlen > sizeof(struct sctp_event_subscribe))
2330 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2333 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2334 * if there is no data to be sent or retransmit, the stack will
2335 * immediately send up this notification.
2337 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2338 &sctp_sk(sk)->subscribe)) {
2339 asoc = sctp_id2assoc(sk, 0);
2341 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2342 event = sctp_ulpevent_make_sender_dry_event(asoc,
2343 GFP_USER | __GFP_NOWARN);
2347 asoc->stream.si->enqueue_event(&asoc->ulpq, event);
2354 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2356 * This socket option is applicable to the UDP-style socket only. When
2357 * set it will cause associations that are idle for more than the
2358 * specified number of seconds to automatically close. An association
2359 * being idle is defined an association that has NOT sent or received
2360 * user data. The special value of '0' indicates that no automatic
2361 * close of any associations should be performed. The option expects an
2362 * integer defining the number of seconds of idle time before an
2363 * association is closed.
2365 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2366 unsigned int optlen)
2368 struct sctp_sock *sp = sctp_sk(sk);
2369 struct net *net = sock_net(sk);
2371 /* Applicable to UDP-style socket only */
2372 if (sctp_style(sk, TCP))
2374 if (optlen != sizeof(int))
2376 if (copy_from_user(&sp->autoclose, optval, optlen))
2379 if (sp->autoclose > net->sctp.max_autoclose)
2380 sp->autoclose = net->sctp.max_autoclose;
2385 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2387 * Applications can enable or disable heartbeats for any peer address of
2388 * an association, modify an address's heartbeat interval, force a
2389 * heartbeat to be sent immediately, and adjust the address's maximum
2390 * number of retransmissions sent before an address is considered
2391 * unreachable. The following structure is used to access and modify an
2392 * address's parameters:
2394 * struct sctp_paddrparams {
2395 * sctp_assoc_t spp_assoc_id;
2396 * struct sockaddr_storage spp_address;
2397 * uint32_t spp_hbinterval;
2398 * uint16_t spp_pathmaxrxt;
2399 * uint32_t spp_pathmtu;
2400 * uint32_t spp_sackdelay;
2401 * uint32_t spp_flags;
2402 * uint32_t spp_ipv6_flowlabel;
2406 * spp_assoc_id - (one-to-many style socket) This is filled in the
2407 * application, and identifies the association for
2409 * spp_address - This specifies which address is of interest.
2410 * spp_hbinterval - This contains the value of the heartbeat interval,
2411 * in milliseconds. If a value of zero
2412 * is present in this field then no changes are to
2413 * be made to this parameter.
2414 * spp_pathmaxrxt - This contains the maximum number of
2415 * retransmissions before this address shall be
2416 * considered unreachable. If a value of zero
2417 * is present in this field then no changes are to
2418 * be made to this parameter.
2419 * spp_pathmtu - When Path MTU discovery is disabled the value
2420 * specified here will be the "fixed" path mtu.
2421 * Note that if the spp_address field is empty
2422 * then all associations on this address will
2423 * have this fixed path mtu set upon them.
2425 * spp_sackdelay - When delayed sack is enabled, this value specifies
2426 * the number of milliseconds that sacks will be delayed
2427 * for. This value will apply to all addresses of an
2428 * association if the spp_address field is empty. Note
2429 * also, that if delayed sack is enabled and this
2430 * value is set to 0, no change is made to the last
2431 * recorded delayed sack timer value.
2433 * spp_flags - These flags are used to control various features
2434 * on an association. The flag field may contain
2435 * zero or more of the following options.
2437 * SPP_HB_ENABLE - Enable heartbeats on the
2438 * specified address. Note that if the address
2439 * field is empty all addresses for the association
2440 * have heartbeats enabled upon them.
2442 * SPP_HB_DISABLE - Disable heartbeats on the
2443 * speicifed address. Note that if the address
2444 * field is empty all addresses for the association
2445 * will have their heartbeats disabled. Note also
2446 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2447 * mutually exclusive, only one of these two should
2448 * be specified. Enabling both fields will have
2449 * undetermined results.
2451 * SPP_HB_DEMAND - Request a user initiated heartbeat
2452 * to be made immediately.
2454 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2455 * heartbeat delayis to be set to the value of 0
2458 * SPP_PMTUD_ENABLE - This field will enable PMTU
2459 * discovery upon the specified address. Note that
2460 * if the address feild is empty then all addresses
2461 * on the association are effected.
2463 * SPP_PMTUD_DISABLE - This field will disable PMTU
2464 * discovery upon the specified address. Note that
2465 * if the address feild is empty then all addresses
2466 * on the association are effected. Not also that
2467 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2468 * exclusive. Enabling both will have undetermined
2471 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2472 * on delayed sack. The time specified in spp_sackdelay
2473 * is used to specify the sack delay for this address. Note
2474 * that if spp_address is empty then all addresses will
2475 * enable delayed sack and take on the sack delay
2476 * value specified in spp_sackdelay.
2477 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2478 * off delayed sack. If the spp_address field is blank then
2479 * delayed sack is disabled for the entire association. Note
2480 * also that this field is mutually exclusive to
2481 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2484 * SPP_IPV6_FLOWLABEL: Setting this flag enables the
2485 * setting of the IPV6 flow label value. The value is
2486 * contained in the spp_ipv6_flowlabel field.
2487 * Upon retrieval, this flag will be set to indicate that
2488 * the spp_ipv6_flowlabel field has a valid value returned.
2489 * If a specific destination address is set (in the
2490 * spp_address field), then the value returned is that of
2491 * the address. If just an association is specified (and
2492 * no address), then the association's default flow label
2493 * is returned. If neither an association nor a destination
2494 * is specified, then the socket's default flow label is
2495 * returned. For non-IPv6 sockets, this flag will be left
2498 * SPP_DSCP: Setting this flag enables the setting of the
2499 * Differentiated Services Code Point (DSCP) value
2500 * associated with either the association or a specific
2501 * address. The value is obtained in the spp_dscp field.
2502 * Upon retrieval, this flag will be set to indicate that
2503 * the spp_dscp field has a valid value returned. If a
2504 * specific destination address is set when called (in the
2505 * spp_address field), then that specific destination
2506 * address's DSCP value is returned. If just an association
2507 * is specified, then the association's default DSCP is
2508 * returned. If neither an association nor a destination is
2509 * specified, then the socket's default DSCP is returned.
2511 * spp_ipv6_flowlabel
2512 * - This field is used in conjunction with the
2513 * SPP_IPV6_FLOWLABEL flag and contains the IPv6 flow label.
2514 * The 20 least significant bits are used for the flow
2515 * label. This setting has precedence over any IPv6-layer
2518 * spp_dscp - This field is used in conjunction with the SPP_DSCP flag
2519 * and contains the DSCP. The 6 most significant bits are
2520 * used for the DSCP. This setting has precedence over any
2521 * IPv4- or IPv6- layer setting.
2523 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2524 struct sctp_transport *trans,
2525 struct sctp_association *asoc,
2526 struct sctp_sock *sp,
2529 int sackdelay_change)
2533 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2534 struct net *net = sock_net(trans->asoc->base.sk);
2536 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2541 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2542 * this field is ignored. Note also that a value of zero indicates
2543 * the current setting should be left unchanged.
2545 if (params->spp_flags & SPP_HB_ENABLE) {
2547 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2548 * set. This lets us use 0 value when this flag
2551 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2552 params->spp_hbinterval = 0;
2554 if (params->spp_hbinterval ||
2555 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2558 msecs_to_jiffies(params->spp_hbinterval);
2561 msecs_to_jiffies(params->spp_hbinterval);
2563 sp->hbinterval = params->spp_hbinterval;
2570 trans->param_flags =
2571 (trans->param_flags & ~SPP_HB) | hb_change;
2574 (asoc->param_flags & ~SPP_HB) | hb_change;
2577 (sp->param_flags & ~SPP_HB) | hb_change;
2581 /* When Path MTU discovery is disabled the value specified here will
2582 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2583 * include the flag SPP_PMTUD_DISABLE for this field to have any
2586 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2588 trans->pathmtu = params->spp_pathmtu;
2589 sctp_assoc_sync_pmtu(asoc);
2591 sctp_assoc_set_pmtu(asoc, params->spp_pathmtu);
2593 sp->pathmtu = params->spp_pathmtu;
2599 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2600 (params->spp_flags & SPP_PMTUD_ENABLE);
2601 trans->param_flags =
2602 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2604 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2605 sctp_assoc_sync_pmtu(asoc);
2609 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2612 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2616 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2617 * value of this field is ignored. Note also that a value of zero
2618 * indicates the current setting should be left unchanged.
2620 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2623 msecs_to_jiffies(params->spp_sackdelay);
2626 msecs_to_jiffies(params->spp_sackdelay);
2628 sp->sackdelay = params->spp_sackdelay;
2632 if (sackdelay_change) {
2634 trans->param_flags =
2635 (trans->param_flags & ~SPP_SACKDELAY) |
2639 (asoc->param_flags & ~SPP_SACKDELAY) |
2643 (sp->param_flags & ~SPP_SACKDELAY) |
2648 /* Note that a value of zero indicates the current setting should be
2651 if (params->spp_pathmaxrxt) {
2653 trans->pathmaxrxt = params->spp_pathmaxrxt;
2655 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2657 sp->pathmaxrxt = params->spp_pathmaxrxt;
2661 if (params->spp_flags & SPP_IPV6_FLOWLABEL) {
2663 if (trans->ipaddr.sa.sa_family == AF_INET6) {
2664 trans->flowlabel = params->spp_ipv6_flowlabel &
2665 SCTP_FLOWLABEL_VAL_MASK;
2666 trans->flowlabel |= SCTP_FLOWLABEL_SET_MASK;
2669 struct sctp_transport *t;
2671 list_for_each_entry(t, &asoc->peer.transport_addr_list,
2673 if (t->ipaddr.sa.sa_family != AF_INET6)
2675 t->flowlabel = params->spp_ipv6_flowlabel &
2676 SCTP_FLOWLABEL_VAL_MASK;
2677 t->flowlabel |= SCTP_FLOWLABEL_SET_MASK;
2679 asoc->flowlabel = params->spp_ipv6_flowlabel &
2680 SCTP_FLOWLABEL_VAL_MASK;
2681 asoc->flowlabel |= SCTP_FLOWLABEL_SET_MASK;
2682 } else if (sctp_opt2sk(sp)->sk_family == AF_INET6) {
2683 sp->flowlabel = params->spp_ipv6_flowlabel &
2684 SCTP_FLOWLABEL_VAL_MASK;
2685 sp->flowlabel |= SCTP_FLOWLABEL_SET_MASK;
2689 if (params->spp_flags & SPP_DSCP) {
2691 trans->dscp = params->spp_dscp & SCTP_DSCP_VAL_MASK;
2692 trans->dscp |= SCTP_DSCP_SET_MASK;
2694 struct sctp_transport *t;
2696 list_for_each_entry(t, &asoc->peer.transport_addr_list,
2698 t->dscp = params->spp_dscp &
2700 t->dscp |= SCTP_DSCP_SET_MASK;
2702 asoc->dscp = params->spp_dscp & SCTP_DSCP_VAL_MASK;
2703 asoc->dscp |= SCTP_DSCP_SET_MASK;
2705 sp->dscp = params->spp_dscp & SCTP_DSCP_VAL_MASK;
2706 sp->dscp |= SCTP_DSCP_SET_MASK;
2713 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2714 char __user *optval,
2715 unsigned int optlen)
2717 struct sctp_paddrparams params;
2718 struct sctp_transport *trans = NULL;
2719 struct sctp_association *asoc = NULL;
2720 struct sctp_sock *sp = sctp_sk(sk);
2722 int hb_change, pmtud_change, sackdelay_change;
2724 if (optlen == sizeof(params)) {
2725 if (copy_from_user(¶ms, optval, optlen))
2727 } else if (optlen == ALIGN(offsetof(struct sctp_paddrparams,
2728 spp_ipv6_flowlabel), 4)) {
2729 if (copy_from_user(¶ms, optval, optlen))
2731 if (params.spp_flags & (SPP_DSCP | SPP_IPV6_FLOWLABEL))
2737 /* Validate flags and value parameters. */
2738 hb_change = params.spp_flags & SPP_HB;
2739 pmtud_change = params.spp_flags & SPP_PMTUD;
2740 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2742 if (hb_change == SPP_HB ||
2743 pmtud_change == SPP_PMTUD ||
2744 sackdelay_change == SPP_SACKDELAY ||
2745 params.spp_sackdelay > 500 ||
2746 (params.spp_pathmtu &&
2747 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2750 /* If an address other than INADDR_ANY is specified, and
2751 * no transport is found, then the request is invalid.
2753 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
2754 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2755 params.spp_assoc_id);
2760 /* Get association, if assoc_id != 0 and the socket is a one
2761 * to many style socket, and an association was not found, then
2762 * the id was invalid.
2764 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2765 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2768 /* Heartbeat demand can only be sent on a transport or
2769 * association, but not a socket.
2771 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2774 /* Process parameters. */
2775 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2776 hb_change, pmtud_change,
2782 /* If changes are for association, also apply parameters to each
2785 if (!trans && asoc) {
2786 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2788 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2789 hb_change, pmtud_change,
2797 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2799 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2802 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2804 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2808 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2810 * This option will effect the way delayed acks are performed. This
2811 * option allows you to get or set the delayed ack time, in
2812 * milliseconds. It also allows changing the delayed ack frequency.
2813 * Changing the frequency to 1 disables the delayed sack algorithm. If
2814 * the assoc_id is 0, then this sets or gets the endpoints default
2815 * values. If the assoc_id field is non-zero, then the set or get
2816 * effects the specified association for the one to many model (the
2817 * assoc_id field is ignored by the one to one model). Note that if
2818 * sack_delay or sack_freq are 0 when setting this option, then the
2819 * current values will remain unchanged.
2821 * struct sctp_sack_info {
2822 * sctp_assoc_t sack_assoc_id;
2823 * uint32_t sack_delay;
2824 * uint32_t sack_freq;
2827 * sack_assoc_id - This parameter, indicates which association the user
2828 * is performing an action upon. Note that if this field's value is
2829 * zero then the endpoints default value is changed (effecting future
2830 * associations only).
2832 * sack_delay - This parameter contains the number of milliseconds that
2833 * the user is requesting the delayed ACK timer be set to. Note that
2834 * this value is defined in the standard to be between 200 and 500
2837 * sack_freq - This parameter contains the number of packets that must
2838 * be received before a sack is sent without waiting for the delay
2839 * timer to expire. The default value for this is 2, setting this
2840 * value to 1 will disable the delayed sack algorithm.
2843 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2844 char __user *optval, unsigned int optlen)
2846 struct sctp_sack_info params;
2847 struct sctp_transport *trans = NULL;
2848 struct sctp_association *asoc = NULL;
2849 struct sctp_sock *sp = sctp_sk(sk);
2851 if (optlen == sizeof(struct sctp_sack_info)) {
2852 if (copy_from_user(¶ms, optval, optlen))
2855 if (params.sack_delay == 0 && params.sack_freq == 0)
2857 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2858 pr_warn_ratelimited(DEPRECATED
2860 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2861 "Use struct sctp_sack_info instead\n",
2862 current->comm, task_pid_nr(current));
2863 if (copy_from_user(¶ms, optval, optlen))
2866 if (params.sack_delay == 0)
2867 params.sack_freq = 1;
2869 params.sack_freq = 0;
2873 /* Validate value parameter. */
2874 if (params.sack_delay > 500)
2877 /* Get association, if sack_assoc_id != 0 and the socket is a one
2878 * to many style socket, and an association was not found, then
2879 * the id was invalid.
2881 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2882 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2885 if (params.sack_delay) {
2888 msecs_to_jiffies(params.sack_delay);
2890 sctp_spp_sackdelay_enable(asoc->param_flags);
2892 sp->sackdelay = params.sack_delay;
2894 sctp_spp_sackdelay_enable(sp->param_flags);
2898 if (params.sack_freq == 1) {
2901 sctp_spp_sackdelay_disable(asoc->param_flags);
2904 sctp_spp_sackdelay_disable(sp->param_flags);
2906 } else if (params.sack_freq > 1) {
2908 asoc->sackfreq = params.sack_freq;
2910 sctp_spp_sackdelay_enable(asoc->param_flags);
2912 sp->sackfreq = params.sack_freq;
2914 sctp_spp_sackdelay_enable(sp->param_flags);
2918 /* If change is for association, also apply to each transport. */
2920 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2922 if (params.sack_delay) {
2924 msecs_to_jiffies(params.sack_delay);
2925 trans->param_flags =
2926 sctp_spp_sackdelay_enable(trans->param_flags);
2928 if (params.sack_freq == 1) {
2929 trans->param_flags =
2930 sctp_spp_sackdelay_disable(trans->param_flags);
2931 } else if (params.sack_freq > 1) {
2932 trans->sackfreq = params.sack_freq;
2933 trans->param_flags =
2934 sctp_spp_sackdelay_enable(trans->param_flags);
2942 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2944 * Applications can specify protocol parameters for the default association
2945 * initialization. The option name argument to setsockopt() and getsockopt()
2948 * Setting initialization parameters is effective only on an unconnected
2949 * socket (for UDP-style sockets only future associations are effected
2950 * by the change). With TCP-style sockets, this option is inherited by
2951 * sockets derived from a listener socket.
2953 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2955 struct sctp_initmsg sinit;
2956 struct sctp_sock *sp = sctp_sk(sk);
2958 if (optlen != sizeof(struct sctp_initmsg))
2960 if (copy_from_user(&sinit, optval, optlen))
2963 if (sinit.sinit_num_ostreams)
2964 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2965 if (sinit.sinit_max_instreams)
2966 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2967 if (sinit.sinit_max_attempts)
2968 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2969 if (sinit.sinit_max_init_timeo)
2970 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2976 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2978 * Applications that wish to use the sendto() system call may wish to
2979 * specify a default set of parameters that would normally be supplied
2980 * through the inclusion of ancillary data. This socket option allows
2981 * such an application to set the default sctp_sndrcvinfo structure.
2982 * The application that wishes to use this socket option simply passes
2983 * in to this call the sctp_sndrcvinfo structure defined in Section
2984 * 5.2.2) The input parameters accepted by this call include
2985 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2986 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2987 * to this call if the caller is using the UDP model.
2989 static int sctp_setsockopt_default_send_param(struct sock *sk,
2990 char __user *optval,
2991 unsigned int optlen)
2993 struct sctp_sock *sp = sctp_sk(sk);
2994 struct sctp_association *asoc;
2995 struct sctp_sndrcvinfo info;
2997 if (optlen != sizeof(info))
2999 if (copy_from_user(&info, optval, optlen))
3001 if (info.sinfo_flags &
3002 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
3003 SCTP_ABORT | SCTP_EOF))
3006 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
3007 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
3010 asoc->default_stream = info.sinfo_stream;
3011 asoc->default_flags = info.sinfo_flags;
3012 asoc->default_ppid = info.sinfo_ppid;
3013 asoc->default_context = info.sinfo_context;
3014 asoc->default_timetolive = info.sinfo_timetolive;
3016 sp->default_stream = info.sinfo_stream;
3017 sp->default_flags = info.sinfo_flags;
3018 sp->default_ppid = info.sinfo_ppid;
3019 sp->default_context = info.sinfo_context;
3020 sp->default_timetolive = info.sinfo_timetolive;
3026 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
3027 * (SCTP_DEFAULT_SNDINFO)
3029 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
3030 char __user *optval,
3031 unsigned int optlen)
3033 struct sctp_sock *sp = sctp_sk(sk);
3034 struct sctp_association *asoc;
3035 struct sctp_sndinfo info;
3037 if (optlen != sizeof(info))
3039 if (copy_from_user(&info, optval, optlen))
3041 if (info.snd_flags &
3042 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
3043 SCTP_ABORT | SCTP_EOF))
3046 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
3047 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
3050 asoc->default_stream = info.snd_sid;
3051 asoc->default_flags = info.snd_flags;
3052 asoc->default_ppid = info.snd_ppid;
3053 asoc->default_context = info.snd_context;
3055 sp->default_stream = info.snd_sid;
3056 sp->default_flags = info.snd_flags;
3057 sp->default_ppid = info.snd_ppid;
3058 sp->default_context = info.snd_context;
3064 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
3066 * Requests that the local SCTP stack use the enclosed peer address as
3067 * the association primary. The enclosed address must be one of the
3068 * association peer's addresses.
3070 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
3071 unsigned int optlen)
3073 struct sctp_prim prim;
3074 struct sctp_transport *trans;
3078 if (optlen != sizeof(struct sctp_prim))
3081 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
3084 /* Allow security module to validate address but need address len. */
3085 af = sctp_get_af_specific(prim.ssp_addr.ss_family);
3089 err = security_sctp_bind_connect(sk, SCTP_PRIMARY_ADDR,
3090 (struct sockaddr *)&prim.ssp_addr,
3095 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
3099 sctp_assoc_set_primary(trans->asoc, trans);
3105 * 7.1.5 SCTP_NODELAY
3107 * Turn on/off any Nagle-like algorithm. This means that packets are
3108 * generally sent as soon as possible and no unnecessary delays are
3109 * introduced, at the cost of more packets in the network. Expects an
3110 * integer boolean flag.
3112 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
3113 unsigned int optlen)
3117 if (optlen < sizeof(int))
3119 if (get_user(val, (int __user *)optval))
3122 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
3128 * 7.1.1 SCTP_RTOINFO
3130 * The protocol parameters used to initialize and bound retransmission
3131 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
3132 * and modify these parameters.
3133 * All parameters are time values, in milliseconds. A value of 0, when
3134 * modifying the parameters, indicates that the current value should not
3138 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
3140 struct sctp_rtoinfo rtoinfo;
3141 struct sctp_association *asoc;
3142 unsigned long rto_min, rto_max;
3143 struct sctp_sock *sp = sctp_sk(sk);
3145 if (optlen != sizeof (struct sctp_rtoinfo))
3148 if (copy_from_user(&rtoinfo, optval, optlen))
3151 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
3153 /* Set the values to the specific association */
3154 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
3157 rto_max = rtoinfo.srto_max;
3158 rto_min = rtoinfo.srto_min;
3161 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
3163 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
3166 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
3168 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
3170 if (rto_min > rto_max)
3174 if (rtoinfo.srto_initial != 0)
3176 msecs_to_jiffies(rtoinfo.srto_initial);
3177 asoc->rto_max = rto_max;
3178 asoc->rto_min = rto_min;
3180 /* If there is no association or the association-id = 0
3181 * set the values to the endpoint.
3183 if (rtoinfo.srto_initial != 0)
3184 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
3185 sp->rtoinfo.srto_max = rto_max;
3186 sp->rtoinfo.srto_min = rto_min;
3194 * 7.1.2 SCTP_ASSOCINFO
3196 * This option is used to tune the maximum retransmission attempts
3197 * of the association.
3198 * Returns an error if the new association retransmission value is
3199 * greater than the sum of the retransmission value of the peer.
3200 * See [SCTP] for more information.
3203 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
3206 struct sctp_assocparams assocparams;
3207 struct sctp_association *asoc;
3209 if (optlen != sizeof(struct sctp_assocparams))
3211 if (copy_from_user(&assocparams, optval, optlen))
3214 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
3216 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
3219 /* Set the values to the specific association */
3221 if (assocparams.sasoc_asocmaxrxt != 0) {
3224 struct sctp_transport *peer_addr;
3226 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
3228 path_sum += peer_addr->pathmaxrxt;
3232 /* Only validate asocmaxrxt if we have more than
3233 * one path/transport. We do this because path
3234 * retransmissions are only counted when we have more
3238 assocparams.sasoc_asocmaxrxt > path_sum)
3241 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3244 if (assocparams.sasoc_cookie_life != 0)
3245 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3247 /* Set the values to the endpoint */
3248 struct sctp_sock *sp = sctp_sk(sk);
3250 if (assocparams.sasoc_asocmaxrxt != 0)
3251 sp->assocparams.sasoc_asocmaxrxt =
3252 assocparams.sasoc_asocmaxrxt;
3253 if (assocparams.sasoc_cookie_life != 0)
3254 sp->assocparams.sasoc_cookie_life =
3255 assocparams.sasoc_cookie_life;
3261 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3263 * This socket option is a boolean flag which turns on or off mapped V4
3264 * addresses. If this option is turned on and the socket is type
3265 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3266 * If this option is turned off, then no mapping will be done of V4
3267 * addresses and a user will receive both PF_INET6 and PF_INET type
3268 * addresses on the socket.
3270 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3273 struct sctp_sock *sp = sctp_sk(sk);
3275 if (optlen < sizeof(int))
3277 if (get_user(val, (int __user *)optval))
3288 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3289 * This option will get or set the maximum size to put in any outgoing
3290 * SCTP DATA chunk. If a message is larger than this size it will be
3291 * fragmented by SCTP into the specified size. Note that the underlying
3292 * SCTP implementation may fragment into smaller sized chunks when the
3293 * PMTU of the underlying association is smaller than the value set by
3294 * the user. The default value for this option is '0' which indicates
3295 * the user is NOT limiting fragmentation and only the PMTU will effect
3296 * SCTP's choice of DATA chunk size. Note also that values set larger
3297 * than the maximum size of an IP datagram will effectively let SCTP
3298 * control fragmentation (i.e. the same as setting this option to 0).
3300 * The following structure is used to access and modify this parameter:
3302 * struct sctp_assoc_value {
3303 * sctp_assoc_t assoc_id;
3304 * uint32_t assoc_value;
3307 * assoc_id: This parameter is ignored for one-to-one style sockets.
3308 * For one-to-many style sockets this parameter indicates which
3309 * association the user is performing an action upon. Note that if
3310 * this field's value is zero then the endpoints default value is
3311 * changed (effecting future associations only).
3312 * assoc_value: This parameter specifies the maximum size in bytes.
3314 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3316 struct sctp_sock *sp = sctp_sk(sk);
3317 struct sctp_assoc_value params;
3318 struct sctp_association *asoc;
3321 if (optlen == sizeof(int)) {
3322 pr_warn_ratelimited(DEPRECATED
3324 "Use of int in maxseg socket option.\n"
3325 "Use struct sctp_assoc_value instead\n",
3326 current->comm, task_pid_nr(current));
3327 if (copy_from_user(&val, optval, optlen))
3329 params.assoc_id = 0;
3330 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3331 if (copy_from_user(¶ms, optval, optlen))
3333 val = params.assoc_value;
3338 asoc = sctp_id2assoc(sk, params.assoc_id);
3341 int min_len, max_len;
3342 __u16 datasize = asoc ? sctp_datachk_len(&asoc->stream) :
3343 sizeof(struct sctp_data_chunk);
3345 min_len = sctp_mtu_payload(sp, SCTP_DEFAULT_MINSEGMENT,
3347 max_len = SCTP_MAX_CHUNK_LEN - datasize;
3349 if (val < min_len || val > max_len)
3354 asoc->user_frag = val;
3355 sctp_assoc_update_frag_point(asoc);
3357 if (params.assoc_id && sctp_style(sk, UDP))
3359 sp->user_frag = val;
3367 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3369 * Requests that the peer mark the enclosed address as the association
3370 * primary. The enclosed address must be one of the association's
3371 * locally bound addresses. The following structure is used to make a
3372 * set primary request:
3374 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3375 unsigned int optlen)
3377 struct net *net = sock_net(sk);
3378 struct sctp_sock *sp;
3379 struct sctp_association *asoc = NULL;
3380 struct sctp_setpeerprim prim;
3381 struct sctp_chunk *chunk;
3387 if (!net->sctp.addip_enable)
3390 if (optlen != sizeof(struct sctp_setpeerprim))
3393 if (copy_from_user(&prim, optval, optlen))
3396 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3400 if (!asoc->peer.asconf_capable)
3403 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3406 if (!sctp_state(asoc, ESTABLISHED))
3409 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3413 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3414 return -EADDRNOTAVAIL;
3416 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3417 return -EADDRNOTAVAIL;
3419 /* Allow security module to validate address. */
3420 err = security_sctp_bind_connect(sk, SCTP_SET_PEER_PRIMARY_ADDR,
3421 (struct sockaddr *)&prim.sspp_addr,
3426 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3427 chunk = sctp_make_asconf_set_prim(asoc,
3428 (union sctp_addr *)&prim.sspp_addr);
3432 err = sctp_send_asconf(asoc, chunk);
3434 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3439 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3440 unsigned int optlen)
3442 struct sctp_setadaptation adaptation;
3444 if (optlen != sizeof(struct sctp_setadaptation))
3446 if (copy_from_user(&adaptation, optval, optlen))
3449 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3455 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3457 * The context field in the sctp_sndrcvinfo structure is normally only
3458 * used when a failed message is retrieved holding the value that was
3459 * sent down on the actual send call. This option allows the setting of
3460 * a default context on an association basis that will be received on
3461 * reading messages from the peer. This is especially helpful in the
3462 * one-2-many model for an application to keep some reference to an
3463 * internal state machine that is processing messages on the
3464 * association. Note that the setting of this value only effects
3465 * received messages from the peer and does not effect the value that is
3466 * saved with outbound messages.
3468 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3469 unsigned int optlen)
3471 struct sctp_assoc_value params;
3472 struct sctp_sock *sp;
3473 struct sctp_association *asoc;
3475 if (optlen != sizeof(struct sctp_assoc_value))
3477 if (copy_from_user(¶ms, optval, optlen))
3482 if (params.assoc_id != 0) {
3483 asoc = sctp_id2assoc(sk, params.assoc_id);
3486 asoc->default_rcv_context = params.assoc_value;
3488 sp->default_rcv_context = params.assoc_value;
3495 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3497 * This options will at a minimum specify if the implementation is doing
3498 * fragmented interleave. Fragmented interleave, for a one to many
3499 * socket, is when subsequent calls to receive a message may return
3500 * parts of messages from different associations. Some implementations
3501 * may allow you to turn this value on or off. If so, when turned off,
3502 * no fragment interleave will occur (which will cause a head of line
3503 * blocking amongst multiple associations sharing the same one to many
3504 * socket). When this option is turned on, then each receive call may
3505 * come from a different association (thus the user must receive data
3506 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3507 * association each receive belongs to.
3509 * This option takes a boolean value. A non-zero value indicates that
3510 * fragmented interleave is on. A value of zero indicates that
3511 * fragmented interleave is off.
3513 * Note that it is important that an implementation that allows this
3514 * option to be turned on, have it off by default. Otherwise an unaware
3515 * application using the one to many model may become confused and act
3518 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3519 char __user *optval,
3520 unsigned int optlen)
3524 if (optlen != sizeof(int))
3526 if (get_user(val, (int __user *)optval))
3529 sctp_sk(sk)->frag_interleave = !!val;
3531 if (!sctp_sk(sk)->frag_interleave)
3532 sctp_sk(sk)->strm_interleave = 0;
3538 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3539 * (SCTP_PARTIAL_DELIVERY_POINT)
3541 * This option will set or get the SCTP partial delivery point. This
3542 * point is the size of a message where the partial delivery API will be
3543 * invoked to help free up rwnd space for the peer. Setting this to a
3544 * lower value will cause partial deliveries to happen more often. The
3545 * calls argument is an integer that sets or gets the partial delivery
3546 * point. Note also that the call will fail if the user attempts to set
3547 * this value larger than the socket receive buffer size.
3549 * Note that any single message having a length smaller than or equal to
3550 * the SCTP partial delivery point will be delivered in one single read
3551 * call as long as the user provided buffer is large enough to hold the
3554 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3555 char __user *optval,
3556 unsigned int optlen)
3560 if (optlen != sizeof(u32))
3562 if (get_user(val, (int __user *)optval))
3565 /* Note: We double the receive buffer from what the user sets
3566 * it to be, also initial rwnd is based on rcvbuf/2.
3568 if (val > (sk->sk_rcvbuf >> 1))
3571 sctp_sk(sk)->pd_point = val;
3573 return 0; /* is this the right error code? */
3577 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3579 * This option will allow a user to change the maximum burst of packets
3580 * that can be emitted by this association. Note that the default value
3581 * is 4, and some implementations may restrict this setting so that it
3582 * can only be lowered.
3584 * NOTE: This text doesn't seem right. Do this on a socket basis with
3585 * future associations inheriting the socket value.
3587 static int sctp_setsockopt_maxburst(struct sock *sk,
3588 char __user *optval,
3589 unsigned int optlen)
3591 struct sctp_assoc_value params;
3592 struct sctp_sock *sp;
3593 struct sctp_association *asoc;
3597 if (optlen == sizeof(int)) {
3598 pr_warn_ratelimited(DEPRECATED
3600 "Use of int in max_burst socket option deprecated.\n"
3601 "Use struct sctp_assoc_value instead\n",
3602 current->comm, task_pid_nr(current));
3603 if (copy_from_user(&val, optval, optlen))
3605 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3606 if (copy_from_user(¶ms, optval, optlen))
3608 val = params.assoc_value;
3609 assoc_id = params.assoc_id;
3615 if (assoc_id != 0) {
3616 asoc = sctp_id2assoc(sk, assoc_id);
3619 asoc->max_burst = val;
3621 sp->max_burst = val;
3627 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3629 * This set option adds a chunk type that the user is requesting to be
3630 * received only in an authenticated way. Changes to the list of chunks
3631 * will only effect future associations on the socket.
3633 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3634 char __user *optval,
3635 unsigned int optlen)
3637 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3638 struct sctp_authchunk val;
3640 if (!ep->auth_enable)
3643 if (optlen != sizeof(struct sctp_authchunk))
3645 if (copy_from_user(&val, optval, optlen))
3648 switch (val.sauth_chunk) {
3650 case SCTP_CID_INIT_ACK:
3651 case SCTP_CID_SHUTDOWN_COMPLETE:
3656 /* add this chunk id to the endpoint */
3657 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3661 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3663 * This option gets or sets the list of HMAC algorithms that the local
3664 * endpoint requires the peer to use.
3666 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3667 char __user *optval,
3668 unsigned int optlen)
3670 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3671 struct sctp_hmacalgo *hmacs;
3675 if (!ep->auth_enable)
3678 if (optlen < sizeof(struct sctp_hmacalgo))
3680 optlen = min_t(unsigned int, optlen, sizeof(struct sctp_hmacalgo) +
3681 SCTP_AUTH_NUM_HMACS * sizeof(u16));
3683 hmacs = memdup_user(optval, optlen);
3685 return PTR_ERR(hmacs);
3687 idents = hmacs->shmac_num_idents;
3688 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3689 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3694 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3701 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3703 * This option will set a shared secret key which is used to build an
3704 * association shared key.
3706 static int sctp_setsockopt_auth_key(struct sock *sk,
3707 char __user *optval,
3708 unsigned int optlen)
3710 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3711 struct sctp_authkey *authkey;
3712 struct sctp_association *asoc;
3715 if (!ep->auth_enable)
3718 if (optlen <= sizeof(struct sctp_authkey))
3720 /* authkey->sca_keylength is u16, so optlen can't be bigger than
3723 optlen = min_t(unsigned int, optlen, USHRT_MAX +
3724 sizeof(struct sctp_authkey));
3726 authkey = memdup_user(optval, optlen);
3727 if (IS_ERR(authkey))
3728 return PTR_ERR(authkey);
3730 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3735 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3736 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3741 ret = sctp_auth_set_key(ep, asoc, authkey);
3748 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3750 * This option will get or set the active shared key to be used to build
3751 * the association shared key.
3753 static int sctp_setsockopt_active_key(struct sock *sk,
3754 char __user *optval,
3755 unsigned int optlen)
3757 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3758 struct sctp_authkeyid val;
3759 struct sctp_association *asoc;
3761 if (!ep->auth_enable)
3764 if (optlen != sizeof(struct sctp_authkeyid))
3766 if (copy_from_user(&val, optval, optlen))
3769 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3770 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3773 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3777 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3779 * This set option will delete a shared secret key from use.
3781 static int sctp_setsockopt_del_key(struct sock *sk,
3782 char __user *optval,
3783 unsigned int optlen)
3785 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3786 struct sctp_authkeyid val;
3787 struct sctp_association *asoc;
3789 if (!ep->auth_enable)
3792 if (optlen != sizeof(struct sctp_authkeyid))
3794 if (copy_from_user(&val, optval, optlen))
3797 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3798 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3801 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3806 * 8.3.4 Deactivate a Shared Key (SCTP_AUTH_DEACTIVATE_KEY)
3808 * This set option will deactivate a shared secret key.
3810 static int sctp_setsockopt_deactivate_key(struct sock *sk, char __user *optval,
3811 unsigned int optlen)
3813 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3814 struct sctp_authkeyid val;
3815 struct sctp_association *asoc;
3817 if (!ep->auth_enable)
3820 if (optlen != sizeof(struct sctp_authkeyid))
3822 if (copy_from_user(&val, optval, optlen))
3825 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3826 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3829 return sctp_auth_deact_key_id(ep, asoc, val.scact_keynumber);
3833 * 8.1.23 SCTP_AUTO_ASCONF
3835 * This option will enable or disable the use of the automatic generation of
3836 * ASCONF chunks to add and delete addresses to an existing association. Note
3837 * that this option has two caveats namely: a) it only affects sockets that
3838 * are bound to all addresses available to the SCTP stack, and b) the system
3839 * administrator may have an overriding control that turns the ASCONF feature
3840 * off no matter what setting the socket option may have.
3841 * This option expects an integer boolean flag, where a non-zero value turns on
3842 * the option, and a zero value turns off the option.
3843 * Note. In this implementation, socket operation overrides default parameter
3844 * being set by sysctl as well as FreeBSD implementation
3846 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3847 unsigned int optlen)
3850 struct sctp_sock *sp = sctp_sk(sk);
3852 if (optlen < sizeof(int))
3854 if (get_user(val, (int __user *)optval))
3856 if (!sctp_is_ep_boundall(sk) && val)
3858 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3861 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3862 if (val == 0 && sp->do_auto_asconf) {
3863 list_del(&sp->auto_asconf_list);
3864 sp->do_auto_asconf = 0;
3865 } else if (val && !sp->do_auto_asconf) {
3866 list_add_tail(&sp->auto_asconf_list,
3867 &sock_net(sk)->sctp.auto_asconf_splist);
3868 sp->do_auto_asconf = 1;
3870 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3875 * SCTP_PEER_ADDR_THLDS
3877 * This option allows us to alter the partially failed threshold for one or all
3878 * transports in an association. See Section 6.1 of:
3879 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3881 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3882 char __user *optval,
3883 unsigned int optlen)
3885 struct sctp_paddrthlds val;
3886 struct sctp_transport *trans;
3887 struct sctp_association *asoc;
3889 if (optlen < sizeof(struct sctp_paddrthlds))
3891 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3892 sizeof(struct sctp_paddrthlds)))
3896 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3897 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3900 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3902 if (val.spt_pathmaxrxt)
3903 trans->pathmaxrxt = val.spt_pathmaxrxt;
3904 trans->pf_retrans = val.spt_pathpfthld;
3907 if (val.spt_pathmaxrxt)
3908 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3909 asoc->pf_retrans = val.spt_pathpfthld;
3911 trans = sctp_addr_id2transport(sk, &val.spt_address,
3916 if (val.spt_pathmaxrxt)
3917 trans->pathmaxrxt = val.spt_pathmaxrxt;
3918 trans->pf_retrans = val.spt_pathpfthld;
3924 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3925 char __user *optval,
3926 unsigned int optlen)
3930 if (optlen < sizeof(int))
3932 if (get_user(val, (int __user *) optval))
3935 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3940 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3941 char __user *optval,
3942 unsigned int optlen)
3946 if (optlen < sizeof(int))
3948 if (get_user(val, (int __user *) optval))
3951 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3956 static int sctp_setsockopt_pr_supported(struct sock *sk,
3957 char __user *optval,
3958 unsigned int optlen)
3960 struct sctp_assoc_value params;
3961 struct sctp_association *asoc;
3962 int retval = -EINVAL;
3964 if (optlen != sizeof(params))
3967 if (copy_from_user(¶ms, optval, optlen)) {
3972 asoc = sctp_id2assoc(sk, params.assoc_id);
3974 asoc->prsctp_enable = !!params.assoc_value;
3975 } else if (!params.assoc_id) {
3976 struct sctp_sock *sp = sctp_sk(sk);
3978 sp->ep->prsctp_enable = !!params.assoc_value;
3989 static int sctp_setsockopt_default_prinfo(struct sock *sk,
3990 char __user *optval,
3991 unsigned int optlen)
3993 struct sctp_default_prinfo info;
3994 struct sctp_association *asoc;
3995 int retval = -EINVAL;
3997 if (optlen != sizeof(info))
4000 if (copy_from_user(&info, optval, sizeof(info))) {
4005 if (info.pr_policy & ~SCTP_PR_SCTP_MASK)
4008 if (info.pr_policy == SCTP_PR_SCTP_NONE)
4011 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
4013 SCTP_PR_SET_POLICY(asoc->default_flags, info.pr_policy);
4014 asoc->default_timetolive = info.pr_value;
4015 } else if (!info.pr_assoc_id) {
4016 struct sctp_sock *sp = sctp_sk(sk);
4018 SCTP_PR_SET_POLICY(sp->default_flags, info.pr_policy);
4019 sp->default_timetolive = info.pr_value;
4030 static int sctp_setsockopt_reconfig_supported(struct sock *sk,
4031 char __user *optval,
4032 unsigned int optlen)
4034 struct sctp_assoc_value params;
4035 struct sctp_association *asoc;
4036 int retval = -EINVAL;
4038 if (optlen != sizeof(params))
4041 if (copy_from_user(¶ms, optval, optlen)) {
4046 asoc = sctp_id2assoc(sk, params.assoc_id);
4048 asoc->reconf_enable = !!params.assoc_value;
4049 } else if (!params.assoc_id) {
4050 struct sctp_sock *sp = sctp_sk(sk);
4052 sp->ep->reconf_enable = !!params.assoc_value;
4063 static int sctp_setsockopt_enable_strreset(struct sock *sk,
4064 char __user *optval,
4065 unsigned int optlen)
4067 struct sctp_assoc_value params;
4068 struct sctp_association *asoc;
4069 int retval = -EINVAL;
4071 if (optlen != sizeof(params))
4074 if (copy_from_user(¶ms, optval, optlen)) {
4079 if (params.assoc_value & (~SCTP_ENABLE_STRRESET_MASK))
4082 asoc = sctp_id2assoc(sk, params.assoc_id);
4084 asoc->strreset_enable = params.assoc_value;
4085 } else if (!params.assoc_id) {
4086 struct sctp_sock *sp = sctp_sk(sk);
4088 sp->ep->strreset_enable = params.assoc_value;
4099 static int sctp_setsockopt_reset_streams(struct sock *sk,
4100 char __user *optval,
4101 unsigned int optlen)
4103 struct sctp_reset_streams *params;
4104 struct sctp_association *asoc;
4105 int retval = -EINVAL;
4107 if (optlen < sizeof(*params))
4109 /* srs_number_streams is u16, so optlen can't be bigger than this. */
4110 optlen = min_t(unsigned int, optlen, USHRT_MAX +
4111 sizeof(__u16) * sizeof(*params));
4113 params = memdup_user(optval, optlen);
4115 return PTR_ERR(params);
4117 if (params->srs_number_streams * sizeof(__u16) >
4118 optlen - sizeof(*params))
4121 asoc = sctp_id2assoc(sk, params->srs_assoc_id);
4125 retval = sctp_send_reset_streams(asoc, params);
4132 static int sctp_setsockopt_reset_assoc(struct sock *sk,
4133 char __user *optval,
4134 unsigned int optlen)
4136 struct sctp_association *asoc;
4137 sctp_assoc_t associd;
4138 int retval = -EINVAL;
4140 if (optlen != sizeof(associd))
4143 if (copy_from_user(&associd, optval, optlen)) {
4148 asoc = sctp_id2assoc(sk, associd);
4152 retval = sctp_send_reset_assoc(asoc);
4158 static int sctp_setsockopt_add_streams(struct sock *sk,
4159 char __user *optval,
4160 unsigned int optlen)
4162 struct sctp_association *asoc;
4163 struct sctp_add_streams params;
4164 int retval = -EINVAL;
4166 if (optlen != sizeof(params))
4169 if (copy_from_user(¶ms, optval, optlen)) {
4174 asoc = sctp_id2assoc(sk, params.sas_assoc_id);
4178 retval = sctp_send_add_streams(asoc, ¶ms);
4184 static int sctp_setsockopt_scheduler(struct sock *sk,
4185 char __user *optval,
4186 unsigned int optlen)
4188 struct sctp_association *asoc;
4189 struct sctp_assoc_value params;
4190 int retval = -EINVAL;
4192 if (optlen < sizeof(params))
4195 optlen = sizeof(params);
4196 if (copy_from_user(¶ms, optval, optlen)) {
4201 if (params.assoc_value > SCTP_SS_MAX)
4204 asoc = sctp_id2assoc(sk, params.assoc_id);
4208 retval = sctp_sched_set_sched(asoc, params.assoc_value);
4214 static int sctp_setsockopt_scheduler_value(struct sock *sk,
4215 char __user *optval,
4216 unsigned int optlen)
4218 struct sctp_association *asoc;
4219 struct sctp_stream_value params;
4220 int retval = -EINVAL;
4222 if (optlen < sizeof(params))
4225 optlen = sizeof(params);
4226 if (copy_from_user(¶ms, optval, optlen)) {
4231 asoc = sctp_id2assoc(sk, params.assoc_id);
4235 retval = sctp_sched_set_value(asoc, params.stream_id,
4236 params.stream_value, GFP_KERNEL);
4242 static int sctp_setsockopt_interleaving_supported(struct sock *sk,
4243 char __user *optval,
4244 unsigned int optlen)
4246 struct sctp_sock *sp = sctp_sk(sk);
4247 struct net *net = sock_net(sk);
4248 struct sctp_assoc_value params;
4249 int retval = -EINVAL;
4251 if (optlen < sizeof(params))
4254 optlen = sizeof(params);
4255 if (copy_from_user(¶ms, optval, optlen)) {
4260 if (params.assoc_id)
4263 if (!net->sctp.intl_enable || !sp->frag_interleave) {
4268 sp->strm_interleave = !!params.assoc_value;
4276 static int sctp_setsockopt_reuse_port(struct sock *sk, char __user *optval,
4277 unsigned int optlen)
4281 if (!sctp_style(sk, TCP))
4284 if (sctp_sk(sk)->ep->base.bind_addr.port)
4287 if (optlen < sizeof(int))
4290 if (get_user(val, (int __user *)optval))
4293 sctp_sk(sk)->reuse = !!val;
4298 /* API 6.2 setsockopt(), getsockopt()
4300 * Applications use setsockopt() and getsockopt() to set or retrieve
4301 * socket options. Socket options are used to change the default
4302 * behavior of sockets calls. They are described in Section 7.
4306 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
4307 * int __user *optlen);
4308 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
4311 * sd - the socket descript.
4312 * level - set to IPPROTO_SCTP for all SCTP options.
4313 * optname - the option name.
4314 * optval - the buffer to store the value of the option.
4315 * optlen - the size of the buffer.
4317 static int sctp_setsockopt(struct sock *sk, int level, int optname,
4318 char __user *optval, unsigned int optlen)
4322 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
4324 /* I can hardly begin to describe how wrong this is. This is
4325 * so broken as to be worse than useless. The API draft
4326 * REALLY is NOT helpful here... I am not convinced that the
4327 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
4328 * are at all well-founded.
4330 if (level != SOL_SCTP) {
4331 struct sctp_af *af = sctp_sk(sk)->pf->af;
4332 retval = af->setsockopt(sk, level, optname, optval, optlen);
4339 case SCTP_SOCKOPT_BINDX_ADD:
4340 /* 'optlen' is the size of the addresses buffer. */
4341 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
4342 optlen, SCTP_BINDX_ADD_ADDR);
4345 case SCTP_SOCKOPT_BINDX_REM:
4346 /* 'optlen' is the size of the addresses buffer. */
4347 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
4348 optlen, SCTP_BINDX_REM_ADDR);
4351 case SCTP_SOCKOPT_CONNECTX_OLD:
4352 /* 'optlen' is the size of the addresses buffer. */
4353 retval = sctp_setsockopt_connectx_old(sk,
4354 (struct sockaddr __user *)optval,
4358 case SCTP_SOCKOPT_CONNECTX:
4359 /* 'optlen' is the size of the addresses buffer. */
4360 retval = sctp_setsockopt_connectx(sk,
4361 (struct sockaddr __user *)optval,
4365 case SCTP_DISABLE_FRAGMENTS:
4366 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
4370 retval = sctp_setsockopt_events(sk, optval, optlen);
4373 case SCTP_AUTOCLOSE:
4374 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
4377 case SCTP_PEER_ADDR_PARAMS:
4378 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
4381 case SCTP_DELAYED_SACK:
4382 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
4384 case SCTP_PARTIAL_DELIVERY_POINT:
4385 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
4389 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
4391 case SCTP_DEFAULT_SEND_PARAM:
4392 retval = sctp_setsockopt_default_send_param(sk, optval,
4395 case SCTP_DEFAULT_SNDINFO:
4396 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
4398 case SCTP_PRIMARY_ADDR:
4399 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
4401 case SCTP_SET_PEER_PRIMARY_ADDR:
4402 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
4405 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
4408 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
4410 case SCTP_ASSOCINFO:
4411 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
4413 case SCTP_I_WANT_MAPPED_V4_ADDR:
4414 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
4417 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
4419 case SCTP_ADAPTATION_LAYER:
4420 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
4423 retval = sctp_setsockopt_context(sk, optval, optlen);
4425 case SCTP_FRAGMENT_INTERLEAVE:
4426 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
4428 case SCTP_MAX_BURST:
4429 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
4431 case SCTP_AUTH_CHUNK:
4432 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
4434 case SCTP_HMAC_IDENT:
4435 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
4438 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
4440 case SCTP_AUTH_ACTIVE_KEY:
4441 retval = sctp_setsockopt_active_key(sk, optval, optlen);
4443 case SCTP_AUTH_DELETE_KEY:
4444 retval = sctp_setsockopt_del_key(sk, optval, optlen);
4446 case SCTP_AUTH_DEACTIVATE_KEY:
4447 retval = sctp_setsockopt_deactivate_key(sk, optval, optlen);
4449 case SCTP_AUTO_ASCONF:
4450 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
4452 case SCTP_PEER_ADDR_THLDS:
4453 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
4455 case SCTP_RECVRCVINFO:
4456 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
4458 case SCTP_RECVNXTINFO:
4459 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
4461 case SCTP_PR_SUPPORTED:
4462 retval = sctp_setsockopt_pr_supported(sk, optval, optlen);
4464 case SCTP_DEFAULT_PRINFO:
4465 retval = sctp_setsockopt_default_prinfo(sk, optval, optlen);
4467 case SCTP_RECONFIG_SUPPORTED:
4468 retval = sctp_setsockopt_reconfig_supported(sk, optval, optlen);
4470 case SCTP_ENABLE_STREAM_RESET:
4471 retval = sctp_setsockopt_enable_strreset(sk, optval, optlen);
4473 case SCTP_RESET_STREAMS:
4474 retval = sctp_setsockopt_reset_streams(sk, optval, optlen);
4476 case SCTP_RESET_ASSOC:
4477 retval = sctp_setsockopt_reset_assoc(sk, optval, optlen);
4479 case SCTP_ADD_STREAMS:
4480 retval = sctp_setsockopt_add_streams(sk, optval, optlen);
4482 case SCTP_STREAM_SCHEDULER:
4483 retval = sctp_setsockopt_scheduler(sk, optval, optlen);
4485 case SCTP_STREAM_SCHEDULER_VALUE:
4486 retval = sctp_setsockopt_scheduler_value(sk, optval, optlen);
4488 case SCTP_INTERLEAVING_SUPPORTED:
4489 retval = sctp_setsockopt_interleaving_supported(sk, optval,
4492 case SCTP_REUSE_PORT:
4493 retval = sctp_setsockopt_reuse_port(sk, optval, optlen);
4496 retval = -ENOPROTOOPT;
4506 /* API 3.1.6 connect() - UDP Style Syntax
4508 * An application may use the connect() call in the UDP model to initiate an
4509 * association without sending data.
4513 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
4515 * sd: the socket descriptor to have a new association added to.
4517 * nam: the address structure (either struct sockaddr_in or struct
4518 * sockaddr_in6 defined in RFC2553 [7]).
4520 * len: the size of the address.
4522 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
4523 int addr_len, int flags)
4525 struct inet_sock *inet = inet_sk(sk);
4531 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
4534 /* We may need to bind the socket. */
4535 if (!inet->inet_num) {
4536 if (sk->sk_prot->get_port(sk, 0)) {
4540 inet->inet_sport = htons(inet->inet_num);
4543 /* Validate addr_len before calling common connect/connectx routine. */
4544 af = sctp_get_af_specific(addr->sa_family);
4545 if (!af || addr_len < af->sockaddr_len) {
4548 /* Pass correct addr len to common routine (so it knows there
4549 * is only one address being passed.
4551 err = __sctp_connect(sk, addr, af->sockaddr_len, flags, NULL);
4558 int sctp_inet_connect(struct socket *sock, struct sockaddr *uaddr,
4559 int addr_len, int flags)
4561 if (addr_len < sizeof(uaddr->sa_family))
4564 if (uaddr->sa_family == AF_UNSPEC)
4567 return sctp_connect(sock->sk, uaddr, addr_len, flags);
4570 /* FIXME: Write comments. */
4571 static int sctp_disconnect(struct sock *sk, int flags)
4573 return -EOPNOTSUPP; /* STUB */
4576 /* 4.1.4 accept() - TCP Style Syntax
4578 * Applications use accept() call to remove an established SCTP
4579 * association from the accept queue of the endpoint. A new socket
4580 * descriptor will be returned from accept() to represent the newly
4581 * formed association.
4583 static struct sock *sctp_accept(struct sock *sk, int flags, int *err, bool kern)
4585 struct sctp_sock *sp;
4586 struct sctp_endpoint *ep;
4587 struct sock *newsk = NULL;
4588 struct sctp_association *asoc;
4597 if (!sctp_style(sk, TCP)) {
4598 error = -EOPNOTSUPP;
4602 if (!sctp_sstate(sk, LISTENING)) {
4607 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
4609 error = sctp_wait_for_accept(sk, timeo);
4613 /* We treat the list of associations on the endpoint as the accept
4614 * queue and pick the first association on the list.
4616 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
4618 newsk = sp->pf->create_accept_sk(sk, asoc, kern);
4624 /* Populate the fields of the newsk from the oldsk and migrate the
4625 * asoc to the newsk.
4627 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
4635 /* The SCTP ioctl handler. */
4636 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
4643 * SEQPACKET-style sockets in LISTENING state are valid, for
4644 * SCTP, so only discard TCP-style sockets in LISTENING state.
4646 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4651 struct sk_buff *skb;
4652 unsigned int amount = 0;
4654 skb = skb_peek(&sk->sk_receive_queue);
4657 * We will only return the amount of this packet since
4658 * that is all that will be read.
4662 rc = put_user(amount, (int __user *)arg);
4674 /* This is the function which gets called during socket creation to
4675 * initialized the SCTP-specific portion of the sock.
4676 * The sock structure should already be zero-filled memory.
4678 static int sctp_init_sock(struct sock *sk)
4680 struct net *net = sock_net(sk);
4681 struct sctp_sock *sp;
4683 pr_debug("%s: sk:%p\n", __func__, sk);
4687 /* Initialize the SCTP per socket area. */
4688 switch (sk->sk_type) {
4689 case SOCK_SEQPACKET:
4690 sp->type = SCTP_SOCKET_UDP;
4693 sp->type = SCTP_SOCKET_TCP;
4696 return -ESOCKTNOSUPPORT;
4699 sk->sk_gso_type = SKB_GSO_SCTP;
4701 /* Initialize default send parameters. These parameters can be
4702 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4704 sp->default_stream = 0;
4705 sp->default_ppid = 0;
4706 sp->default_flags = 0;
4707 sp->default_context = 0;
4708 sp->default_timetolive = 0;
4710 sp->default_rcv_context = 0;
4711 sp->max_burst = net->sctp.max_burst;
4713 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4715 /* Initialize default setup parameters. These parameters
4716 * can be modified with the SCTP_INITMSG socket option or
4717 * overridden by the SCTP_INIT CMSG.
4719 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4720 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4721 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4722 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4724 /* Initialize default RTO related parameters. These parameters can
4725 * be modified for with the SCTP_RTOINFO socket option.
4727 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4728 sp->rtoinfo.srto_max = net->sctp.rto_max;
4729 sp->rtoinfo.srto_min = net->sctp.rto_min;
4731 /* Initialize default association related parameters. These parameters
4732 * can be modified with the SCTP_ASSOCINFO socket option.
4734 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4735 sp->assocparams.sasoc_number_peer_destinations = 0;
4736 sp->assocparams.sasoc_peer_rwnd = 0;
4737 sp->assocparams.sasoc_local_rwnd = 0;
4738 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4740 /* Initialize default event subscriptions. By default, all the
4743 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4745 /* Default Peer Address Parameters. These defaults can
4746 * be modified via SCTP_PEER_ADDR_PARAMS
4748 sp->hbinterval = net->sctp.hb_interval;
4749 sp->pathmaxrxt = net->sctp.max_retrans_path;
4750 sp->pathmtu = 0; /* allow default discovery */
4751 sp->sackdelay = net->sctp.sack_timeout;
4753 sp->param_flags = SPP_HB_ENABLE |
4755 SPP_SACKDELAY_ENABLE;
4757 /* If enabled no SCTP message fragmentation will be performed.
4758 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4760 sp->disable_fragments = 0;
4762 /* Enable Nagle algorithm by default. */
4765 sp->recvrcvinfo = 0;
4766 sp->recvnxtinfo = 0;
4768 /* Enable by default. */
4771 /* Auto-close idle associations after the configured
4772 * number of seconds. A value of 0 disables this
4773 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4774 * for UDP-style sockets only.
4778 /* User specified fragmentation limit. */
4781 sp->adaptation_ind = 0;
4783 sp->pf = sctp_get_pf_specific(sk->sk_family);
4785 /* Control variables for partial data delivery. */
4786 atomic_set(&sp->pd_mode, 0);
4787 skb_queue_head_init(&sp->pd_lobby);
4788 sp->frag_interleave = 0;
4790 /* Create a per socket endpoint structure. Even if we
4791 * change the data structure relationships, this may still
4792 * be useful for storing pre-connect address information.
4794 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4800 sk->sk_destruct = sctp_destruct_sock;
4802 SCTP_DBG_OBJCNT_INC(sock);
4805 sk_sockets_allocated_inc(sk);
4806 sock_prot_inuse_add(net, sk->sk_prot, 1);
4808 /* Nothing can fail after this block, otherwise
4809 * sctp_destroy_sock() will be called without addr_wq_lock held
4811 if (net->sctp.default_auto_asconf) {
4812 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
4813 list_add_tail(&sp->auto_asconf_list,
4814 &net->sctp.auto_asconf_splist);
4815 sp->do_auto_asconf = 1;
4816 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4818 sp->do_auto_asconf = 0;
4826 /* Cleanup any SCTP per socket resources. Must be called with
4827 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4829 static void sctp_destroy_sock(struct sock *sk)
4831 struct sctp_sock *sp;
4833 pr_debug("%s: sk:%p\n", __func__, sk);
4835 /* Release our hold on the endpoint. */
4837 /* This could happen during socket init, thus we bail out
4838 * early, since the rest of the below is not setup either.
4843 if (sp->do_auto_asconf) {
4844 sp->do_auto_asconf = 0;
4845 list_del(&sp->auto_asconf_list);
4847 sctp_endpoint_free(sp->ep);
4849 sk_sockets_allocated_dec(sk);
4850 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4854 /* Triggered when there are no references on the socket anymore */
4855 static void sctp_destruct_sock(struct sock *sk)
4857 struct sctp_sock *sp = sctp_sk(sk);
4859 /* Free up the HMAC transform. */
4860 crypto_free_shash(sp->hmac);
4862 inet_sock_destruct(sk);
4865 /* API 4.1.7 shutdown() - TCP Style Syntax
4866 * int shutdown(int socket, int how);
4868 * sd - the socket descriptor of the association to be closed.
4869 * how - Specifies the type of shutdown. The values are
4872 * Disables further receive operations. No SCTP
4873 * protocol action is taken.
4875 * Disables further send operations, and initiates
4876 * the SCTP shutdown sequence.
4878 * Disables further send and receive operations
4879 * and initiates the SCTP shutdown sequence.
4881 static void sctp_shutdown(struct sock *sk, int how)
4883 struct net *net = sock_net(sk);
4884 struct sctp_endpoint *ep;
4886 if (!sctp_style(sk, TCP))
4889 ep = sctp_sk(sk)->ep;
4890 if (how & SEND_SHUTDOWN && !list_empty(&ep->asocs)) {
4891 struct sctp_association *asoc;
4893 inet_sk_set_state(sk, SCTP_SS_CLOSING);
4894 asoc = list_entry(ep->asocs.next,
4895 struct sctp_association, asocs);
4896 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4900 int sctp_get_sctp_info(struct sock *sk, struct sctp_association *asoc,
4901 struct sctp_info *info)
4903 struct sctp_transport *prim;
4904 struct list_head *pos;
4907 memset(info, 0, sizeof(*info));
4909 struct sctp_sock *sp = sctp_sk(sk);
4911 info->sctpi_s_autoclose = sp->autoclose;
4912 info->sctpi_s_adaptation_ind = sp->adaptation_ind;
4913 info->sctpi_s_pd_point = sp->pd_point;
4914 info->sctpi_s_nodelay = sp->nodelay;
4915 info->sctpi_s_disable_fragments = sp->disable_fragments;
4916 info->sctpi_s_v4mapped = sp->v4mapped;
4917 info->sctpi_s_frag_interleave = sp->frag_interleave;
4918 info->sctpi_s_type = sp->type;
4923 info->sctpi_tag = asoc->c.my_vtag;
4924 info->sctpi_state = asoc->state;
4925 info->sctpi_rwnd = asoc->a_rwnd;
4926 info->sctpi_unackdata = asoc->unack_data;
4927 info->sctpi_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4928 info->sctpi_instrms = asoc->stream.incnt;
4929 info->sctpi_outstrms = asoc->stream.outcnt;
4930 list_for_each(pos, &asoc->base.inqueue.in_chunk_list)
4931 info->sctpi_inqueue++;
4932 list_for_each(pos, &asoc->outqueue.out_chunk_list)
4933 info->sctpi_outqueue++;
4934 info->sctpi_overall_error = asoc->overall_error_count;
4935 info->sctpi_max_burst = asoc->max_burst;
4936 info->sctpi_maxseg = asoc->frag_point;
4937 info->sctpi_peer_rwnd = asoc->peer.rwnd;
4938 info->sctpi_peer_tag = asoc->c.peer_vtag;
4940 mask = asoc->peer.ecn_capable << 1;
4941 mask = (mask | asoc->peer.ipv4_address) << 1;
4942 mask = (mask | asoc->peer.ipv6_address) << 1;
4943 mask = (mask | asoc->peer.hostname_address) << 1;
4944 mask = (mask | asoc->peer.asconf_capable) << 1;
4945 mask = (mask | asoc->peer.prsctp_capable) << 1;
4946 mask = (mask | asoc->peer.auth_capable);
4947 info->sctpi_peer_capable = mask;
4948 mask = asoc->peer.sack_needed << 1;
4949 mask = (mask | asoc->peer.sack_generation) << 1;
4950 mask = (mask | asoc->peer.zero_window_announced);
4951 info->sctpi_peer_sack = mask;
4953 info->sctpi_isacks = asoc->stats.isacks;
4954 info->sctpi_osacks = asoc->stats.osacks;
4955 info->sctpi_opackets = asoc->stats.opackets;
4956 info->sctpi_ipackets = asoc->stats.ipackets;
4957 info->sctpi_rtxchunks = asoc->stats.rtxchunks;
4958 info->sctpi_outofseqtsns = asoc->stats.outofseqtsns;
4959 info->sctpi_idupchunks = asoc->stats.idupchunks;
4960 info->sctpi_gapcnt = asoc->stats.gapcnt;
4961 info->sctpi_ouodchunks = asoc->stats.ouodchunks;
4962 info->sctpi_iuodchunks = asoc->stats.iuodchunks;
4963 info->sctpi_oodchunks = asoc->stats.oodchunks;
4964 info->sctpi_iodchunks = asoc->stats.iodchunks;
4965 info->sctpi_octrlchunks = asoc->stats.octrlchunks;
4966 info->sctpi_ictrlchunks = asoc->stats.ictrlchunks;
4968 prim = asoc->peer.primary_path;
4969 memcpy(&info->sctpi_p_address, &prim->ipaddr, sizeof(prim->ipaddr));
4970 info->sctpi_p_state = prim->state;
4971 info->sctpi_p_cwnd = prim->cwnd;
4972 info->sctpi_p_srtt = prim->srtt;
4973 info->sctpi_p_rto = jiffies_to_msecs(prim->rto);
4974 info->sctpi_p_hbinterval = prim->hbinterval;
4975 info->sctpi_p_pathmaxrxt = prim->pathmaxrxt;
4976 info->sctpi_p_sackdelay = jiffies_to_msecs(prim->sackdelay);
4977 info->sctpi_p_ssthresh = prim->ssthresh;
4978 info->sctpi_p_partial_bytes_acked = prim->partial_bytes_acked;
4979 info->sctpi_p_flight_size = prim->flight_size;
4980 info->sctpi_p_error = prim->error_count;
4984 EXPORT_SYMBOL_GPL(sctp_get_sctp_info);
4986 /* use callback to avoid exporting the core structure */
4987 void sctp_transport_walk_start(struct rhashtable_iter *iter)
4989 rhltable_walk_enter(&sctp_transport_hashtable, iter);
4991 rhashtable_walk_start(iter);
4994 void sctp_transport_walk_stop(struct rhashtable_iter *iter)
4996 rhashtable_walk_stop(iter);
4997 rhashtable_walk_exit(iter);
5000 struct sctp_transport *sctp_transport_get_next(struct net *net,
5001 struct rhashtable_iter *iter)
5003 struct sctp_transport *t;
5005 t = rhashtable_walk_next(iter);
5006 for (; t; t = rhashtable_walk_next(iter)) {
5008 if (PTR_ERR(t) == -EAGAIN)
5013 if (!sctp_transport_hold(t))
5016 if (net_eq(sock_net(t->asoc->base.sk), net) &&
5017 t->asoc->peer.primary_path == t)
5020 sctp_transport_put(t);
5026 struct sctp_transport *sctp_transport_get_idx(struct net *net,
5027 struct rhashtable_iter *iter,
5030 struct sctp_transport *t;
5033 return SEQ_START_TOKEN;
5035 while ((t = sctp_transport_get_next(net, iter)) && !IS_ERR(t)) {
5038 sctp_transport_put(t);
5044 int sctp_for_each_endpoint(int (*cb)(struct sctp_endpoint *, void *),
5048 struct sctp_ep_common *epb;
5049 struct sctp_hashbucket *head;
5051 for (head = sctp_ep_hashtable; hash < sctp_ep_hashsize;
5053 read_lock_bh(&head->lock);
5054 sctp_for_each_hentry(epb, &head->chain) {
5055 err = cb(sctp_ep(epb), p);
5059 read_unlock_bh(&head->lock);
5064 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint);
5066 int sctp_transport_lookup_process(int (*cb)(struct sctp_transport *, void *),
5068 const union sctp_addr *laddr,
5069 const union sctp_addr *paddr, void *p)
5071 struct sctp_transport *transport;
5075 transport = sctp_addrs_lookup_transport(net, laddr, paddr);
5080 err = cb(transport, p);
5081 sctp_transport_put(transport);
5085 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process);
5087 int sctp_for_each_transport(int (*cb)(struct sctp_transport *, void *),
5088 int (*cb_done)(struct sctp_transport *, void *),
5089 struct net *net, int *pos, void *p) {
5090 struct rhashtable_iter hti;
5091 struct sctp_transport *tsp;
5096 sctp_transport_walk_start(&hti);
5098 tsp = sctp_transport_get_idx(net, &hti, *pos + 1);
5099 for (; !IS_ERR_OR_NULL(tsp); tsp = sctp_transport_get_next(net, &hti)) {
5104 sctp_transport_put(tsp);
5106 sctp_transport_walk_stop(&hti);
5109 if (cb_done && !cb_done(tsp, p)) {
5111 sctp_transport_put(tsp);
5114 sctp_transport_put(tsp);
5119 EXPORT_SYMBOL_GPL(sctp_for_each_transport);
5121 /* 7.2.1 Association Status (SCTP_STATUS)
5123 * Applications can retrieve current status information about an
5124 * association, including association state, peer receiver window size,
5125 * number of unacked data chunks, and number of data chunks pending
5126 * receipt. This information is read-only.
5128 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
5129 char __user *optval,
5132 struct sctp_status status;
5133 struct sctp_association *asoc = NULL;
5134 struct sctp_transport *transport;
5135 sctp_assoc_t associd;
5138 if (len < sizeof(status)) {
5143 len = sizeof(status);
5144 if (copy_from_user(&status, optval, len)) {
5149 associd = status.sstat_assoc_id;
5150 asoc = sctp_id2assoc(sk, associd);
5156 transport = asoc->peer.primary_path;
5158 status.sstat_assoc_id = sctp_assoc2id(asoc);
5159 status.sstat_state = sctp_assoc_to_state(asoc);
5160 status.sstat_rwnd = asoc->peer.rwnd;
5161 status.sstat_unackdata = asoc->unack_data;
5163 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
5164 status.sstat_instrms = asoc->stream.incnt;
5165 status.sstat_outstrms = asoc->stream.outcnt;
5166 status.sstat_fragmentation_point = asoc->frag_point;
5167 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
5168 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
5169 transport->af_specific->sockaddr_len);
5170 /* Map ipv4 address into v4-mapped-on-v6 address. */
5171 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
5172 (union sctp_addr *)&status.sstat_primary.spinfo_address);
5173 status.sstat_primary.spinfo_state = transport->state;
5174 status.sstat_primary.spinfo_cwnd = transport->cwnd;
5175 status.sstat_primary.spinfo_srtt = transport->srtt;
5176 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
5177 status.sstat_primary.spinfo_mtu = transport->pathmtu;
5179 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
5180 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
5182 if (put_user(len, optlen)) {
5187 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
5188 __func__, len, status.sstat_state, status.sstat_rwnd,
5189 status.sstat_assoc_id);
5191 if (copy_to_user(optval, &status, len)) {
5201 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
5203 * Applications can retrieve information about a specific peer address
5204 * of an association, including its reachability state, congestion
5205 * window, and retransmission timer values. This information is
5208 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
5209 char __user *optval,
5212 struct sctp_paddrinfo pinfo;
5213 struct sctp_transport *transport;
5216 if (len < sizeof(pinfo)) {
5221 len = sizeof(pinfo);
5222 if (copy_from_user(&pinfo, optval, len)) {
5227 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
5228 pinfo.spinfo_assoc_id);
5232 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
5233 pinfo.spinfo_state = transport->state;
5234 pinfo.spinfo_cwnd = transport->cwnd;
5235 pinfo.spinfo_srtt = transport->srtt;
5236 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
5237 pinfo.spinfo_mtu = transport->pathmtu;
5239 if (pinfo.spinfo_state == SCTP_UNKNOWN)
5240 pinfo.spinfo_state = SCTP_ACTIVE;
5242 if (put_user(len, optlen)) {
5247 if (copy_to_user(optval, &pinfo, len)) {
5256 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
5258 * This option is a on/off flag. If enabled no SCTP message
5259 * fragmentation will be performed. Instead if a message being sent
5260 * exceeds the current PMTU size, the message will NOT be sent and
5261 * instead a error will be indicated to the user.
5263 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
5264 char __user *optval, int __user *optlen)
5268 if (len < sizeof(int))
5272 val = (sctp_sk(sk)->disable_fragments == 1);
5273 if (put_user(len, optlen))
5275 if (copy_to_user(optval, &val, len))
5280 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
5282 * This socket option is used to specify various notifications and
5283 * ancillary data the user wishes to receive.
5285 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
5290 if (len > sizeof(struct sctp_event_subscribe))
5291 len = sizeof(struct sctp_event_subscribe);
5292 if (put_user(len, optlen))
5294 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
5299 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
5301 * This socket option is applicable to the UDP-style socket only. When
5302 * set it will cause associations that are idle for more than the
5303 * specified number of seconds to automatically close. An association
5304 * being idle is defined an association that has NOT sent or received
5305 * user data. The special value of '0' indicates that no automatic
5306 * close of any associations should be performed. The option expects an
5307 * integer defining the number of seconds of idle time before an
5308 * association is closed.
5310 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
5312 /* Applicable to UDP-style socket only */
5313 if (sctp_style(sk, TCP))
5315 if (len < sizeof(int))
5318 if (put_user(len, optlen))
5320 if (put_user(sctp_sk(sk)->autoclose, (int __user *)optval))
5325 /* Helper routine to branch off an association to a new socket. */
5326 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
5328 struct sctp_association *asoc = sctp_id2assoc(sk, id);
5329 struct sctp_sock *sp = sctp_sk(sk);
5330 struct socket *sock;
5333 /* Do not peel off from one netns to another one. */
5334 if (!net_eq(current->nsproxy->net_ns, sock_net(sk)))
5340 /* An association cannot be branched off from an already peeled-off
5341 * socket, nor is this supported for tcp style sockets.
5343 if (!sctp_style(sk, UDP))
5346 /* Create a new socket. */
5347 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
5351 sctp_copy_sock(sock->sk, sk, asoc);
5353 /* Make peeled-off sockets more like 1-1 accepted sockets.
5354 * Set the daddr and initialize id to something more random and also
5355 * copy over any ip options.
5357 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
5358 sp->pf->copy_ip_options(sk, sock->sk);
5360 /* Populate the fields of the newsk from the oldsk and migrate the
5361 * asoc to the newsk.
5363 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
5369 EXPORT_SYMBOL(sctp_do_peeloff);
5371 static int sctp_getsockopt_peeloff_common(struct sock *sk, sctp_peeloff_arg_t *peeloff,
5372 struct file **newfile, unsigned flags)
5374 struct socket *newsock;
5377 retval = sctp_do_peeloff(sk, peeloff->associd, &newsock);
5381 /* Map the socket to an unused fd that can be returned to the user. */
5382 retval = get_unused_fd_flags(flags & SOCK_CLOEXEC);
5384 sock_release(newsock);
5388 *newfile = sock_alloc_file(newsock, 0, NULL);
5389 if (IS_ERR(*newfile)) {
5390 put_unused_fd(retval);
5391 retval = PTR_ERR(*newfile);
5396 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
5399 peeloff->sd = retval;
5401 if (flags & SOCK_NONBLOCK)
5402 (*newfile)->f_flags |= O_NONBLOCK;
5407 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
5409 sctp_peeloff_arg_t peeloff;
5410 struct file *newfile = NULL;
5413 if (len < sizeof(sctp_peeloff_arg_t))
5415 len = sizeof(sctp_peeloff_arg_t);
5416 if (copy_from_user(&peeloff, optval, len))
5419 retval = sctp_getsockopt_peeloff_common(sk, &peeloff, &newfile, 0);
5423 /* Return the fd mapped to the new socket. */
5424 if (put_user(len, optlen)) {
5426 put_unused_fd(retval);
5430 if (copy_to_user(optval, &peeloff, len)) {
5432 put_unused_fd(retval);
5435 fd_install(retval, newfile);
5440 static int sctp_getsockopt_peeloff_flags(struct sock *sk, int len,
5441 char __user *optval, int __user *optlen)
5443 sctp_peeloff_flags_arg_t peeloff;
5444 struct file *newfile = NULL;
5447 if (len < sizeof(sctp_peeloff_flags_arg_t))
5449 len = sizeof(sctp_peeloff_flags_arg_t);
5450 if (copy_from_user(&peeloff, optval, len))
5453 retval = sctp_getsockopt_peeloff_common(sk, &peeloff.p_arg,
5454 &newfile, peeloff.flags);
5458 /* Return the fd mapped to the new socket. */
5459 if (put_user(len, optlen)) {
5461 put_unused_fd(retval);
5465 if (copy_to_user(optval, &peeloff, len)) {
5467 put_unused_fd(retval);
5470 fd_install(retval, newfile);
5475 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
5477 * Applications can enable or disable heartbeats for any peer address of
5478 * an association, modify an address's heartbeat interval, force a
5479 * heartbeat to be sent immediately, and adjust the address's maximum
5480 * number of retransmissions sent before an address is considered
5481 * unreachable. The following structure is used to access and modify an
5482 * address's parameters:
5484 * struct sctp_paddrparams {
5485 * sctp_assoc_t spp_assoc_id;
5486 * struct sockaddr_storage spp_address;
5487 * uint32_t spp_hbinterval;
5488 * uint16_t spp_pathmaxrxt;
5489 * uint32_t spp_pathmtu;
5490 * uint32_t spp_sackdelay;
5491 * uint32_t spp_flags;
5494 * spp_assoc_id - (one-to-many style socket) This is filled in the
5495 * application, and identifies the association for
5497 * spp_address - This specifies which address is of interest.
5498 * spp_hbinterval - This contains the value of the heartbeat interval,
5499 * in milliseconds. If a value of zero
5500 * is present in this field then no changes are to
5501 * be made to this parameter.
5502 * spp_pathmaxrxt - This contains the maximum number of
5503 * retransmissions before this address shall be
5504 * considered unreachable. If a value of zero
5505 * is present in this field then no changes are to
5506 * be made to this parameter.
5507 * spp_pathmtu - When Path MTU discovery is disabled the value
5508 * specified here will be the "fixed" path mtu.
5509 * Note that if the spp_address field is empty
5510 * then all associations on this address will
5511 * have this fixed path mtu set upon them.
5513 * spp_sackdelay - When delayed sack is enabled, this value specifies
5514 * the number of milliseconds that sacks will be delayed
5515 * for. This value will apply to all addresses of an
5516 * association if the spp_address field is empty. Note
5517 * also, that if delayed sack is enabled and this
5518 * value is set to 0, no change is made to the last
5519 * recorded delayed sack timer value.
5521 * spp_flags - These flags are used to control various features
5522 * on an association. The flag field may contain
5523 * zero or more of the following options.
5525 * SPP_HB_ENABLE - Enable heartbeats on the
5526 * specified address. Note that if the address
5527 * field is empty all addresses for the association
5528 * have heartbeats enabled upon them.
5530 * SPP_HB_DISABLE - Disable heartbeats on the
5531 * speicifed address. Note that if the address
5532 * field is empty all addresses for the association
5533 * will have their heartbeats disabled. Note also
5534 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
5535 * mutually exclusive, only one of these two should
5536 * be specified. Enabling both fields will have
5537 * undetermined results.
5539 * SPP_HB_DEMAND - Request a user initiated heartbeat
5540 * to be made immediately.
5542 * SPP_PMTUD_ENABLE - This field will enable PMTU
5543 * discovery upon the specified address. Note that
5544 * if the address feild is empty then all addresses
5545 * on the association are effected.
5547 * SPP_PMTUD_DISABLE - This field will disable PMTU
5548 * discovery upon the specified address. Note that
5549 * if the address feild is empty then all addresses
5550 * on the association are effected. Not also that
5551 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
5552 * exclusive. Enabling both will have undetermined
5555 * SPP_SACKDELAY_ENABLE - Setting this flag turns
5556 * on delayed sack. The time specified in spp_sackdelay
5557 * is used to specify the sack delay for this address. Note
5558 * that if spp_address is empty then all addresses will
5559 * enable delayed sack and take on the sack delay
5560 * value specified in spp_sackdelay.
5561 * SPP_SACKDELAY_DISABLE - Setting this flag turns
5562 * off delayed sack. If the spp_address field is blank then
5563 * delayed sack is disabled for the entire association. Note
5564 * also that this field is mutually exclusive to
5565 * SPP_SACKDELAY_ENABLE, setting both will have undefined
5568 * SPP_IPV6_FLOWLABEL: Setting this flag enables the
5569 * setting of the IPV6 flow label value. The value is
5570 * contained in the spp_ipv6_flowlabel field.
5571 * Upon retrieval, this flag will be set to indicate that
5572 * the spp_ipv6_flowlabel field has a valid value returned.
5573 * If a specific destination address is set (in the
5574 * spp_address field), then the value returned is that of
5575 * the address. If just an association is specified (and
5576 * no address), then the association's default flow label
5577 * is returned. If neither an association nor a destination
5578 * is specified, then the socket's default flow label is
5579 * returned. For non-IPv6 sockets, this flag will be left
5582 * SPP_DSCP: Setting this flag enables the setting of the
5583 * Differentiated Services Code Point (DSCP) value
5584 * associated with either the association or a specific
5585 * address. The value is obtained in the spp_dscp field.
5586 * Upon retrieval, this flag will be set to indicate that
5587 * the spp_dscp field has a valid value returned. If a
5588 * specific destination address is set when called (in the
5589 * spp_address field), then that specific destination
5590 * address's DSCP value is returned. If just an association
5591 * is specified, then the association's default DSCP is
5592 * returned. If neither an association nor a destination is
5593 * specified, then the socket's default DSCP is returned.
5595 * spp_ipv6_flowlabel
5596 * - This field is used in conjunction with the
5597 * SPP_IPV6_FLOWLABEL flag and contains the IPv6 flow label.
5598 * The 20 least significant bits are used for the flow
5599 * label. This setting has precedence over any IPv6-layer
5602 * spp_dscp - This field is used in conjunction with the SPP_DSCP flag
5603 * and contains the DSCP. The 6 most significant bits are
5604 * used for the DSCP. This setting has precedence over any
5605 * IPv4- or IPv6- layer setting.
5607 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
5608 char __user *optval, int __user *optlen)
5610 struct sctp_paddrparams params;
5611 struct sctp_transport *trans = NULL;
5612 struct sctp_association *asoc = NULL;
5613 struct sctp_sock *sp = sctp_sk(sk);
5615 if (len >= sizeof(params))
5616 len = sizeof(params);
5617 else if (len >= ALIGN(offsetof(struct sctp_paddrparams,
5618 spp_ipv6_flowlabel), 4))
5619 len = ALIGN(offsetof(struct sctp_paddrparams,
5620 spp_ipv6_flowlabel), 4);
5624 if (copy_from_user(¶ms, optval, len))
5627 /* If an address other than INADDR_ANY is specified, and
5628 * no transport is found, then the request is invalid.
5630 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
5631 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
5632 params.spp_assoc_id);
5634 pr_debug("%s: failed no transport\n", __func__);
5639 /* Get association, if assoc_id != 0 and the socket is a one
5640 * to many style socket, and an association was not found, then
5641 * the id was invalid.
5643 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
5644 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
5645 pr_debug("%s: failed no association\n", __func__);
5650 /* Fetch transport values. */
5651 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
5652 params.spp_pathmtu = trans->pathmtu;
5653 params.spp_pathmaxrxt = trans->pathmaxrxt;
5654 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
5656 /*draft-11 doesn't say what to return in spp_flags*/
5657 params.spp_flags = trans->param_flags;
5658 if (trans->flowlabel & SCTP_FLOWLABEL_SET_MASK) {
5659 params.spp_ipv6_flowlabel = trans->flowlabel &
5660 SCTP_FLOWLABEL_VAL_MASK;
5661 params.spp_flags |= SPP_IPV6_FLOWLABEL;
5663 if (trans->dscp & SCTP_DSCP_SET_MASK) {
5664 params.spp_dscp = trans->dscp & SCTP_DSCP_VAL_MASK;
5665 params.spp_flags |= SPP_DSCP;
5668 /* Fetch association values. */
5669 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
5670 params.spp_pathmtu = asoc->pathmtu;
5671 params.spp_pathmaxrxt = asoc->pathmaxrxt;
5672 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
5674 /*draft-11 doesn't say what to return in spp_flags*/
5675 params.spp_flags = asoc->param_flags;
5676 if (asoc->flowlabel & SCTP_FLOWLABEL_SET_MASK) {
5677 params.spp_ipv6_flowlabel = asoc->flowlabel &
5678 SCTP_FLOWLABEL_VAL_MASK;
5679 params.spp_flags |= SPP_IPV6_FLOWLABEL;
5681 if (asoc->dscp & SCTP_DSCP_SET_MASK) {
5682 params.spp_dscp = asoc->dscp & SCTP_DSCP_VAL_MASK;
5683 params.spp_flags |= SPP_DSCP;
5686 /* Fetch socket values. */
5687 params.spp_hbinterval = sp->hbinterval;
5688 params.spp_pathmtu = sp->pathmtu;
5689 params.spp_sackdelay = sp->sackdelay;
5690 params.spp_pathmaxrxt = sp->pathmaxrxt;
5692 /*draft-11 doesn't say what to return in spp_flags*/
5693 params.spp_flags = sp->param_flags;
5694 if (sp->flowlabel & SCTP_FLOWLABEL_SET_MASK) {
5695 params.spp_ipv6_flowlabel = sp->flowlabel &
5696 SCTP_FLOWLABEL_VAL_MASK;
5697 params.spp_flags |= SPP_IPV6_FLOWLABEL;
5699 if (sp->dscp & SCTP_DSCP_SET_MASK) {
5700 params.spp_dscp = sp->dscp & SCTP_DSCP_VAL_MASK;
5701 params.spp_flags |= SPP_DSCP;
5705 if (copy_to_user(optval, ¶ms, len))
5708 if (put_user(len, optlen))
5715 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
5717 * This option will effect the way delayed acks are performed. This
5718 * option allows you to get or set the delayed ack time, in
5719 * milliseconds. It also allows changing the delayed ack frequency.
5720 * Changing the frequency to 1 disables the delayed sack algorithm. If
5721 * the assoc_id is 0, then this sets or gets the endpoints default
5722 * values. If the assoc_id field is non-zero, then the set or get
5723 * effects the specified association for the one to many model (the
5724 * assoc_id field is ignored by the one to one model). Note that if
5725 * sack_delay or sack_freq are 0 when setting this option, then the
5726 * current values will remain unchanged.
5728 * struct sctp_sack_info {
5729 * sctp_assoc_t sack_assoc_id;
5730 * uint32_t sack_delay;
5731 * uint32_t sack_freq;
5734 * sack_assoc_id - This parameter, indicates which association the user
5735 * is performing an action upon. Note that if this field's value is
5736 * zero then the endpoints default value is changed (effecting future
5737 * associations only).
5739 * sack_delay - This parameter contains the number of milliseconds that
5740 * the user is requesting the delayed ACK timer be set to. Note that
5741 * this value is defined in the standard to be between 200 and 500
5744 * sack_freq - This parameter contains the number of packets that must
5745 * be received before a sack is sent without waiting for the delay
5746 * timer to expire. The default value for this is 2, setting this
5747 * value to 1 will disable the delayed sack algorithm.
5749 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
5750 char __user *optval,
5753 struct sctp_sack_info params;
5754 struct sctp_association *asoc = NULL;
5755 struct sctp_sock *sp = sctp_sk(sk);
5757 if (len >= sizeof(struct sctp_sack_info)) {
5758 len = sizeof(struct sctp_sack_info);
5760 if (copy_from_user(¶ms, optval, len))
5762 } else if (len == sizeof(struct sctp_assoc_value)) {
5763 pr_warn_ratelimited(DEPRECATED
5765 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
5766 "Use struct sctp_sack_info instead\n",
5767 current->comm, task_pid_nr(current));
5768 if (copy_from_user(¶ms, optval, len))
5773 /* Get association, if sack_assoc_id != 0 and the socket is a one
5774 * to many style socket, and an association was not found, then
5775 * the id was invalid.
5777 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
5778 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
5782 /* Fetch association values. */
5783 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
5784 params.sack_delay = jiffies_to_msecs(
5786 params.sack_freq = asoc->sackfreq;
5789 params.sack_delay = 0;
5790 params.sack_freq = 1;
5793 /* Fetch socket values. */
5794 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
5795 params.sack_delay = sp->sackdelay;
5796 params.sack_freq = sp->sackfreq;
5798 params.sack_delay = 0;
5799 params.sack_freq = 1;
5803 if (copy_to_user(optval, ¶ms, len))
5806 if (put_user(len, optlen))
5812 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5814 * Applications can specify protocol parameters for the default association
5815 * initialization. The option name argument to setsockopt() and getsockopt()
5818 * Setting initialization parameters is effective only on an unconnected
5819 * socket (for UDP-style sockets only future associations are effected
5820 * by the change). With TCP-style sockets, this option is inherited by
5821 * sockets derived from a listener socket.
5823 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
5825 if (len < sizeof(struct sctp_initmsg))
5827 len = sizeof(struct sctp_initmsg);
5828 if (put_user(len, optlen))
5830 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
5836 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
5837 char __user *optval, int __user *optlen)
5839 struct sctp_association *asoc;
5841 struct sctp_getaddrs getaddrs;
5842 struct sctp_transport *from;
5844 union sctp_addr temp;
5845 struct sctp_sock *sp = sctp_sk(sk);
5850 if (len < sizeof(struct sctp_getaddrs))
5853 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5856 /* For UDP-style sockets, id specifies the association to query. */
5857 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5861 to = optval + offsetof(struct sctp_getaddrs, addrs);
5862 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5864 list_for_each_entry(from, &asoc->peer.transport_addr_list,
5866 memcpy(&temp, &from->ipaddr, sizeof(temp));
5867 addrlen = sctp_get_pf_specific(sk->sk_family)
5868 ->addr_to_user(sp, &temp);
5869 if (space_left < addrlen)
5871 if (copy_to_user(to, &temp, addrlen))
5875 space_left -= addrlen;
5878 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
5880 bytes_copied = ((char __user *)to) - optval;
5881 if (put_user(bytes_copied, optlen))
5887 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
5888 size_t space_left, int *bytes_copied)
5890 struct sctp_sockaddr_entry *addr;
5891 union sctp_addr temp;
5894 struct net *net = sock_net(sk);
5897 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
5901 if ((PF_INET == sk->sk_family) &&
5902 (AF_INET6 == addr->a.sa.sa_family))
5904 if ((PF_INET6 == sk->sk_family) &&
5905 inet_v6_ipv6only(sk) &&
5906 (AF_INET == addr->a.sa.sa_family))
5908 memcpy(&temp, &addr->a, sizeof(temp));
5909 if (!temp.v4.sin_port)
5910 temp.v4.sin_port = htons(port);
5912 addrlen = sctp_get_pf_specific(sk->sk_family)
5913 ->addr_to_user(sctp_sk(sk), &temp);
5915 if (space_left < addrlen) {
5919 memcpy(to, &temp, addrlen);
5923 space_left -= addrlen;
5924 *bytes_copied += addrlen;
5932 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
5933 char __user *optval, int __user *optlen)
5935 struct sctp_bind_addr *bp;
5936 struct sctp_association *asoc;
5938 struct sctp_getaddrs getaddrs;
5939 struct sctp_sockaddr_entry *addr;
5941 union sctp_addr temp;
5942 struct sctp_sock *sp = sctp_sk(sk);
5946 int bytes_copied = 0;
5950 if (len < sizeof(struct sctp_getaddrs))
5953 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5957 * For UDP-style sockets, id specifies the association to query.
5958 * If the id field is set to the value '0' then the locally bound
5959 * addresses are returned without regard to any particular
5962 if (0 == getaddrs.assoc_id) {
5963 bp = &sctp_sk(sk)->ep->base.bind_addr;
5965 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5968 bp = &asoc->base.bind_addr;
5971 to = optval + offsetof(struct sctp_getaddrs, addrs);
5972 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5974 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN);
5978 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5979 * addresses from the global local address list.
5981 if (sctp_list_single_entry(&bp->address_list)) {
5982 addr = list_entry(bp->address_list.next,
5983 struct sctp_sockaddr_entry, list);
5984 if (sctp_is_any(sk, &addr->a)) {
5985 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
5986 space_left, &bytes_copied);
5996 /* Protection on the bound address list is not needed since
5997 * in the socket option context we hold a socket lock and
5998 * thus the bound address list can't change.
6000 list_for_each_entry(addr, &bp->address_list, list) {
6001 memcpy(&temp, &addr->a, sizeof(temp));
6002 addrlen = sctp_get_pf_specific(sk->sk_family)
6003 ->addr_to_user(sp, &temp);
6004 if (space_left < addrlen) {
6005 err = -ENOMEM; /*fixme: right error?*/
6008 memcpy(buf, &temp, addrlen);
6010 bytes_copied += addrlen;
6012 space_left -= addrlen;
6016 if (copy_to_user(to, addrs, bytes_copied)) {
6020 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
6024 /* XXX: We should have accounted for sizeof(struct sctp_getaddrs) too,
6025 * but we can't change it anymore.
6027 if (put_user(bytes_copied, optlen))
6034 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
6036 * Requests that the local SCTP stack use the enclosed peer address as
6037 * the association primary. The enclosed address must be one of the
6038 * association peer's addresses.
6040 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
6041 char __user *optval, int __user *optlen)
6043 struct sctp_prim prim;
6044 struct sctp_association *asoc;
6045 struct sctp_sock *sp = sctp_sk(sk);
6047 if (len < sizeof(struct sctp_prim))
6050 len = sizeof(struct sctp_prim);
6052 if (copy_from_user(&prim, optval, len))
6055 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
6059 if (!asoc->peer.primary_path)
6062 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
6063 asoc->peer.primary_path->af_specific->sockaddr_len);
6065 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
6066 (union sctp_addr *)&prim.ssp_addr);
6068 if (put_user(len, optlen))
6070 if (copy_to_user(optval, &prim, len))
6077 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
6079 * Requests that the local endpoint set the specified Adaptation Layer
6080 * Indication parameter for all future INIT and INIT-ACK exchanges.
6082 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
6083 char __user *optval, int __user *optlen)
6085 struct sctp_setadaptation adaptation;
6087 if (len < sizeof(struct sctp_setadaptation))
6090 len = sizeof(struct sctp_setadaptation);
6092 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
6094 if (put_user(len, optlen))
6096 if (copy_to_user(optval, &adaptation, len))
6104 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
6106 * Applications that wish to use the sendto() system call may wish to
6107 * specify a default set of parameters that would normally be supplied
6108 * through the inclusion of ancillary data. This socket option allows
6109 * such an application to set the default sctp_sndrcvinfo structure.
6112 * The application that wishes to use this socket option simply passes
6113 * in to this call the sctp_sndrcvinfo structure defined in Section
6114 * 5.2.2) The input parameters accepted by this call include
6115 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
6116 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
6117 * to this call if the caller is using the UDP model.
6119 * For getsockopt, it get the default sctp_sndrcvinfo structure.
6121 static int sctp_getsockopt_default_send_param(struct sock *sk,
6122 int len, char __user *optval,
6125 struct sctp_sock *sp = sctp_sk(sk);
6126 struct sctp_association *asoc;
6127 struct sctp_sndrcvinfo info;
6129 if (len < sizeof(info))
6134 if (copy_from_user(&info, optval, len))
6137 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
6138 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
6141 info.sinfo_stream = asoc->default_stream;
6142 info.sinfo_flags = asoc->default_flags;
6143 info.sinfo_ppid = asoc->default_ppid;
6144 info.sinfo_context = asoc->default_context;
6145 info.sinfo_timetolive = asoc->default_timetolive;
6147 info.sinfo_stream = sp->default_stream;
6148 info.sinfo_flags = sp->default_flags;
6149 info.sinfo_ppid = sp->default_ppid;
6150 info.sinfo_context = sp->default_context;
6151 info.sinfo_timetolive = sp->default_timetolive;
6154 if (put_user(len, optlen))
6156 if (copy_to_user(optval, &info, len))
6162 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
6163 * (SCTP_DEFAULT_SNDINFO)
6165 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
6166 char __user *optval,
6169 struct sctp_sock *sp = sctp_sk(sk);
6170 struct sctp_association *asoc;
6171 struct sctp_sndinfo info;
6173 if (len < sizeof(info))
6178 if (copy_from_user(&info, optval, len))
6181 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
6182 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
6185 info.snd_sid = asoc->default_stream;
6186 info.snd_flags = asoc->default_flags;
6187 info.snd_ppid = asoc->default_ppid;
6188 info.snd_context = asoc->default_context;
6190 info.snd_sid = sp->default_stream;
6191 info.snd_flags = sp->default_flags;
6192 info.snd_ppid = sp->default_ppid;
6193 info.snd_context = sp->default_context;
6196 if (put_user(len, optlen))
6198 if (copy_to_user(optval, &info, len))
6206 * 7.1.5 SCTP_NODELAY
6208 * Turn on/off any Nagle-like algorithm. This means that packets are
6209 * generally sent as soon as possible and no unnecessary delays are
6210 * introduced, at the cost of more packets in the network. Expects an
6211 * integer boolean flag.
6214 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
6215 char __user *optval, int __user *optlen)
6219 if (len < sizeof(int))
6223 val = (sctp_sk(sk)->nodelay == 1);
6224 if (put_user(len, optlen))
6226 if (copy_to_user(optval, &val, len))
6233 * 7.1.1 SCTP_RTOINFO
6235 * The protocol parameters used to initialize and bound retransmission
6236 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
6237 * and modify these parameters.
6238 * All parameters are time values, in milliseconds. A value of 0, when
6239 * modifying the parameters, indicates that the current value should not
6243 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
6244 char __user *optval,
6245 int __user *optlen) {
6246 struct sctp_rtoinfo rtoinfo;
6247 struct sctp_association *asoc;
6249 if (len < sizeof (struct sctp_rtoinfo))
6252 len = sizeof(struct sctp_rtoinfo);
6254 if (copy_from_user(&rtoinfo, optval, len))
6257 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
6259 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
6262 /* Values corresponding to the specific association. */
6264 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
6265 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
6266 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
6268 /* Values corresponding to the endpoint. */
6269 struct sctp_sock *sp = sctp_sk(sk);
6271 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
6272 rtoinfo.srto_max = sp->rtoinfo.srto_max;
6273 rtoinfo.srto_min = sp->rtoinfo.srto_min;
6276 if (put_user(len, optlen))
6279 if (copy_to_user(optval, &rtoinfo, len))
6287 * 7.1.2 SCTP_ASSOCINFO
6289 * This option is used to tune the maximum retransmission attempts
6290 * of the association.
6291 * Returns an error if the new association retransmission value is
6292 * greater than the sum of the retransmission value of the peer.
6293 * See [SCTP] for more information.
6296 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
6297 char __user *optval,
6301 struct sctp_assocparams assocparams;
6302 struct sctp_association *asoc;
6303 struct list_head *pos;
6306 if (len < sizeof (struct sctp_assocparams))
6309 len = sizeof(struct sctp_assocparams);
6311 if (copy_from_user(&assocparams, optval, len))
6314 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
6316 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
6319 /* Values correspoinding to the specific association */
6321 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
6322 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
6323 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
6324 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
6326 list_for_each(pos, &asoc->peer.transport_addr_list) {
6330 assocparams.sasoc_number_peer_destinations = cnt;
6332 /* Values corresponding to the endpoint */
6333 struct sctp_sock *sp = sctp_sk(sk);
6335 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
6336 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
6337 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
6338 assocparams.sasoc_cookie_life =
6339 sp->assocparams.sasoc_cookie_life;
6340 assocparams.sasoc_number_peer_destinations =
6342 sasoc_number_peer_destinations;
6345 if (put_user(len, optlen))
6348 if (copy_to_user(optval, &assocparams, len))
6355 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
6357 * This socket option is a boolean flag which turns on or off mapped V4
6358 * addresses. If this option is turned on and the socket is type
6359 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
6360 * If this option is turned off, then no mapping will be done of V4
6361 * addresses and a user will receive both PF_INET6 and PF_INET type
6362 * addresses on the socket.
6364 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
6365 char __user *optval, int __user *optlen)
6368 struct sctp_sock *sp = sctp_sk(sk);
6370 if (len < sizeof(int))
6375 if (put_user(len, optlen))
6377 if (copy_to_user(optval, &val, len))
6384 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
6385 * (chapter and verse is quoted at sctp_setsockopt_context())
6387 static int sctp_getsockopt_context(struct sock *sk, int len,
6388 char __user *optval, int __user *optlen)
6390 struct sctp_assoc_value params;
6391 struct sctp_sock *sp;
6392 struct sctp_association *asoc;
6394 if (len < sizeof(struct sctp_assoc_value))
6397 len = sizeof(struct sctp_assoc_value);
6399 if (copy_from_user(¶ms, optval, len))
6404 if (params.assoc_id != 0) {
6405 asoc = sctp_id2assoc(sk, params.assoc_id);
6408 params.assoc_value = asoc->default_rcv_context;
6410 params.assoc_value = sp->default_rcv_context;
6413 if (put_user(len, optlen))
6415 if (copy_to_user(optval, ¶ms, len))
6422 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
6423 * This option will get or set the maximum size to put in any outgoing
6424 * SCTP DATA chunk. If a message is larger than this size it will be
6425 * fragmented by SCTP into the specified size. Note that the underlying
6426 * SCTP implementation may fragment into smaller sized chunks when the
6427 * PMTU of the underlying association is smaller than the value set by
6428 * the user. The default value for this option is '0' which indicates
6429 * the user is NOT limiting fragmentation and only the PMTU will effect
6430 * SCTP's choice of DATA chunk size. Note also that values set larger
6431 * than the maximum size of an IP datagram will effectively let SCTP
6432 * control fragmentation (i.e. the same as setting this option to 0).
6434 * The following structure is used to access and modify this parameter:
6436 * struct sctp_assoc_value {
6437 * sctp_assoc_t assoc_id;
6438 * uint32_t assoc_value;
6441 * assoc_id: This parameter is ignored for one-to-one style sockets.
6442 * For one-to-many style sockets this parameter indicates which
6443 * association the user is performing an action upon. Note that if
6444 * this field's value is zero then the endpoints default value is
6445 * changed (effecting future associations only).
6446 * assoc_value: This parameter specifies the maximum size in bytes.
6448 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
6449 char __user *optval, int __user *optlen)
6451 struct sctp_assoc_value params;
6452 struct sctp_association *asoc;
6454 if (len == sizeof(int)) {
6455 pr_warn_ratelimited(DEPRECATED
6457 "Use of int in maxseg socket option.\n"
6458 "Use struct sctp_assoc_value instead\n",
6459 current->comm, task_pid_nr(current));
6460 params.assoc_id = 0;
6461 } else if (len >= sizeof(struct sctp_assoc_value)) {
6462 len = sizeof(struct sctp_assoc_value);
6463 if (copy_from_user(¶ms, optval, len))
6468 asoc = sctp_id2assoc(sk, params.assoc_id);
6469 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
6473 params.assoc_value = asoc->frag_point;
6475 params.assoc_value = sctp_sk(sk)->user_frag;
6477 if (put_user(len, optlen))
6479 if (len == sizeof(int)) {
6480 if (copy_to_user(optval, ¶ms.assoc_value, len))
6483 if (copy_to_user(optval, ¶ms, len))
6491 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
6492 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
6494 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
6495 char __user *optval, int __user *optlen)
6499 if (len < sizeof(int))
6504 val = sctp_sk(sk)->frag_interleave;
6505 if (put_user(len, optlen))
6507 if (copy_to_user(optval, &val, len))
6514 * 7.1.25. Set or Get the sctp partial delivery point
6515 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
6517 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
6518 char __user *optval,
6523 if (len < sizeof(u32))
6528 val = sctp_sk(sk)->pd_point;
6529 if (put_user(len, optlen))
6531 if (copy_to_user(optval, &val, len))
6538 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
6539 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
6541 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
6542 char __user *optval,
6545 struct sctp_assoc_value params;
6546 struct sctp_sock *sp;
6547 struct sctp_association *asoc;
6549 if (len == sizeof(int)) {
6550 pr_warn_ratelimited(DEPRECATED
6552 "Use of int in max_burst socket option.\n"
6553 "Use struct sctp_assoc_value instead\n",
6554 current->comm, task_pid_nr(current));
6555 params.assoc_id = 0;
6556 } else if (len >= sizeof(struct sctp_assoc_value)) {
6557 len = sizeof(struct sctp_assoc_value);
6558 if (copy_from_user(¶ms, optval, len))
6565 if (params.assoc_id != 0) {
6566 asoc = sctp_id2assoc(sk, params.assoc_id);
6569 params.assoc_value = asoc->max_burst;
6571 params.assoc_value = sp->max_burst;
6573 if (len == sizeof(int)) {
6574 if (copy_to_user(optval, ¶ms.assoc_value, len))
6577 if (copy_to_user(optval, ¶ms, len))
6585 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
6586 char __user *optval, int __user *optlen)
6588 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6589 struct sctp_hmacalgo __user *p = (void __user *)optval;
6590 struct sctp_hmac_algo_param *hmacs;
6595 if (!ep->auth_enable)
6598 hmacs = ep->auth_hmacs_list;
6599 data_len = ntohs(hmacs->param_hdr.length) -
6600 sizeof(struct sctp_paramhdr);
6602 if (len < sizeof(struct sctp_hmacalgo) + data_len)
6605 len = sizeof(struct sctp_hmacalgo) + data_len;
6606 num_idents = data_len / sizeof(u16);
6608 if (put_user(len, optlen))
6610 if (put_user(num_idents, &p->shmac_num_idents))
6612 for (i = 0; i < num_idents; i++) {
6613 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
6615 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
6621 static int sctp_getsockopt_active_key(struct sock *sk, int len,
6622 char __user *optval, int __user *optlen)
6624 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6625 struct sctp_authkeyid val;
6626 struct sctp_association *asoc;
6628 if (!ep->auth_enable)
6631 if (len < sizeof(struct sctp_authkeyid))
6634 len = sizeof(struct sctp_authkeyid);
6635 if (copy_from_user(&val, optval, len))
6638 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
6639 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
6643 val.scact_keynumber = asoc->active_key_id;
6645 val.scact_keynumber = ep->active_key_id;
6647 if (put_user(len, optlen))
6649 if (copy_to_user(optval, &val, len))
6655 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
6656 char __user *optval, int __user *optlen)
6658 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6659 struct sctp_authchunks __user *p = (void __user *)optval;
6660 struct sctp_authchunks val;
6661 struct sctp_association *asoc;
6662 struct sctp_chunks_param *ch;
6666 if (!ep->auth_enable)
6669 if (len < sizeof(struct sctp_authchunks))
6672 if (copy_from_user(&val, optval, sizeof(val)))
6675 to = p->gauth_chunks;
6676 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6680 ch = asoc->peer.peer_chunks;
6684 /* See if the user provided enough room for all the data */
6685 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6686 if (len < num_chunks)
6689 if (copy_to_user(to, ch->chunks, num_chunks))
6692 len = sizeof(struct sctp_authchunks) + num_chunks;
6693 if (put_user(len, optlen))
6695 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6700 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
6701 char __user *optval, int __user *optlen)
6703 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6704 struct sctp_authchunks __user *p = (void __user *)optval;
6705 struct sctp_authchunks val;
6706 struct sctp_association *asoc;
6707 struct sctp_chunks_param *ch;
6711 if (!ep->auth_enable)
6714 if (len < sizeof(struct sctp_authchunks))
6717 if (copy_from_user(&val, optval, sizeof(val)))
6720 to = p->gauth_chunks;
6721 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6722 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
6726 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
6728 ch = ep->auth_chunk_list;
6733 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6734 if (len < sizeof(struct sctp_authchunks) + num_chunks)
6737 if (copy_to_user(to, ch->chunks, num_chunks))
6740 len = sizeof(struct sctp_authchunks) + num_chunks;
6741 if (put_user(len, optlen))
6743 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6750 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
6751 * This option gets the current number of associations that are attached
6752 * to a one-to-many style socket. The option value is an uint32_t.
6754 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
6755 char __user *optval, int __user *optlen)
6757 struct sctp_sock *sp = sctp_sk(sk);
6758 struct sctp_association *asoc;
6761 if (sctp_style(sk, TCP))
6764 if (len < sizeof(u32))
6769 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6773 if (put_user(len, optlen))
6775 if (copy_to_user(optval, &val, len))
6782 * 8.1.23 SCTP_AUTO_ASCONF
6783 * See the corresponding setsockopt entry as description
6785 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
6786 char __user *optval, int __user *optlen)
6790 if (len < sizeof(int))
6794 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
6796 if (put_user(len, optlen))
6798 if (copy_to_user(optval, &val, len))
6804 * 8.2.6. Get the Current Identifiers of Associations
6805 * (SCTP_GET_ASSOC_ID_LIST)
6807 * This option gets the current list of SCTP association identifiers of
6808 * the SCTP associations handled by a one-to-many style socket.
6810 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
6811 char __user *optval, int __user *optlen)
6813 struct sctp_sock *sp = sctp_sk(sk);
6814 struct sctp_association *asoc;
6815 struct sctp_assoc_ids *ids;
6818 if (sctp_style(sk, TCP))
6821 if (len < sizeof(struct sctp_assoc_ids))
6824 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6828 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
6831 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
6833 ids = kmalloc(len, GFP_USER | __GFP_NOWARN);
6837 ids->gaids_number_of_ids = num;
6839 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6840 ids->gaids_assoc_id[num++] = asoc->assoc_id;
6843 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
6853 * SCTP_PEER_ADDR_THLDS
6855 * This option allows us to fetch the partially failed threshold for one or all
6856 * transports in an association. See Section 6.1 of:
6857 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6859 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
6860 char __user *optval,
6864 struct sctp_paddrthlds val;
6865 struct sctp_transport *trans;
6866 struct sctp_association *asoc;
6868 if (len < sizeof(struct sctp_paddrthlds))
6870 len = sizeof(struct sctp_paddrthlds);
6871 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
6874 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
6875 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
6879 val.spt_pathpfthld = asoc->pf_retrans;
6880 val.spt_pathmaxrxt = asoc->pathmaxrxt;
6882 trans = sctp_addr_id2transport(sk, &val.spt_address,
6887 val.spt_pathmaxrxt = trans->pathmaxrxt;
6888 val.spt_pathpfthld = trans->pf_retrans;
6891 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
6898 * SCTP_GET_ASSOC_STATS
6900 * This option retrieves local per endpoint statistics. It is modeled
6901 * after OpenSolaris' implementation
6903 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
6904 char __user *optval,
6907 struct sctp_assoc_stats sas;
6908 struct sctp_association *asoc = NULL;
6910 /* User must provide at least the assoc id */
6911 if (len < sizeof(sctp_assoc_t))
6914 /* Allow the struct to grow and fill in as much as possible */
6915 len = min_t(size_t, len, sizeof(sas));
6917 if (copy_from_user(&sas, optval, len))
6920 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
6924 sas.sas_rtxchunks = asoc->stats.rtxchunks;
6925 sas.sas_gapcnt = asoc->stats.gapcnt;
6926 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
6927 sas.sas_osacks = asoc->stats.osacks;
6928 sas.sas_isacks = asoc->stats.isacks;
6929 sas.sas_octrlchunks = asoc->stats.octrlchunks;
6930 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
6931 sas.sas_oodchunks = asoc->stats.oodchunks;
6932 sas.sas_iodchunks = asoc->stats.iodchunks;
6933 sas.sas_ouodchunks = asoc->stats.ouodchunks;
6934 sas.sas_iuodchunks = asoc->stats.iuodchunks;
6935 sas.sas_idupchunks = asoc->stats.idupchunks;
6936 sas.sas_opackets = asoc->stats.opackets;
6937 sas.sas_ipackets = asoc->stats.ipackets;
6939 /* New high max rto observed, will return 0 if not a single
6940 * RTO update took place. obs_rto_ipaddr will be bogus
6943 sas.sas_maxrto = asoc->stats.max_obs_rto;
6944 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
6945 sizeof(struct sockaddr_storage));
6947 /* Mark beginning of a new observation period */
6948 asoc->stats.max_obs_rto = asoc->rto_min;
6950 if (put_user(len, optlen))
6953 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
6955 if (copy_to_user(optval, &sas, len))
6961 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
6962 char __user *optval,
6967 if (len < sizeof(int))
6971 if (sctp_sk(sk)->recvrcvinfo)
6973 if (put_user(len, optlen))
6975 if (copy_to_user(optval, &val, len))
6981 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
6982 char __user *optval,
6987 if (len < sizeof(int))
6991 if (sctp_sk(sk)->recvnxtinfo)
6993 if (put_user(len, optlen))
6995 if (copy_to_user(optval, &val, len))
7001 static int sctp_getsockopt_pr_supported(struct sock *sk, int len,
7002 char __user *optval,
7005 struct sctp_assoc_value params;
7006 struct sctp_association *asoc;
7007 int retval = -EFAULT;
7009 if (len < sizeof(params)) {
7014 len = sizeof(params);
7015 if (copy_from_user(¶ms, optval, len))
7018 asoc = sctp_id2assoc(sk, params.assoc_id);
7020 params.assoc_value = asoc->prsctp_enable;
7021 } else if (!params.assoc_id) {
7022 struct sctp_sock *sp = sctp_sk(sk);
7024 params.assoc_value = sp->ep->prsctp_enable;
7030 if (put_user(len, optlen))
7033 if (copy_to_user(optval, ¶ms, len))
7042 static int sctp_getsockopt_default_prinfo(struct sock *sk, int len,
7043 char __user *optval,
7046 struct sctp_default_prinfo info;
7047 struct sctp_association *asoc;
7048 int retval = -EFAULT;
7050 if (len < sizeof(info)) {
7056 if (copy_from_user(&info, optval, len))
7059 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
7061 info.pr_policy = SCTP_PR_POLICY(asoc->default_flags);
7062 info.pr_value = asoc->default_timetolive;
7063 } else if (!info.pr_assoc_id) {
7064 struct sctp_sock *sp = sctp_sk(sk);
7066 info.pr_policy = SCTP_PR_POLICY(sp->default_flags);
7067 info.pr_value = sp->default_timetolive;
7073 if (put_user(len, optlen))
7076 if (copy_to_user(optval, &info, len))
7085 static int sctp_getsockopt_pr_assocstatus(struct sock *sk, int len,
7086 char __user *optval,
7089 struct sctp_prstatus params;
7090 struct sctp_association *asoc;
7092 int retval = -EINVAL;
7094 if (len < sizeof(params))
7097 len = sizeof(params);
7098 if (copy_from_user(¶ms, optval, len)) {
7103 policy = params.sprstat_policy;
7104 if (!policy || (policy & ~(SCTP_PR_SCTP_MASK | SCTP_PR_SCTP_ALL)))
7107 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
7111 if (policy & SCTP_PR_SCTP_ALL) {
7112 params.sprstat_abandoned_unsent = 0;
7113 params.sprstat_abandoned_sent = 0;
7114 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
7115 params.sprstat_abandoned_unsent +=
7116 asoc->abandoned_unsent[policy];
7117 params.sprstat_abandoned_sent +=
7118 asoc->abandoned_sent[policy];
7121 params.sprstat_abandoned_unsent =
7122 asoc->abandoned_unsent[__SCTP_PR_INDEX(policy)];
7123 params.sprstat_abandoned_sent =
7124 asoc->abandoned_sent[__SCTP_PR_INDEX(policy)];
7127 if (put_user(len, optlen)) {
7132 if (copy_to_user(optval, ¶ms, len)) {
7143 static int sctp_getsockopt_pr_streamstatus(struct sock *sk, int len,
7144 char __user *optval,
7147 struct sctp_stream_out_ext *streamoute;
7148 struct sctp_association *asoc;
7149 struct sctp_prstatus params;
7150 int retval = -EINVAL;
7153 if (len < sizeof(params))
7156 len = sizeof(params);
7157 if (copy_from_user(¶ms, optval, len)) {
7162 policy = params.sprstat_policy;
7163 if (!policy || (policy & ~(SCTP_PR_SCTP_MASK | SCTP_PR_SCTP_ALL)))
7166 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
7167 if (!asoc || params.sprstat_sid >= asoc->stream.outcnt)
7170 streamoute = SCTP_SO(&asoc->stream, params.sprstat_sid)->ext;
7172 /* Not allocated yet, means all stats are 0 */
7173 params.sprstat_abandoned_unsent = 0;
7174 params.sprstat_abandoned_sent = 0;
7179 if (policy == SCTP_PR_SCTP_ALL) {
7180 params.sprstat_abandoned_unsent = 0;
7181 params.sprstat_abandoned_sent = 0;
7182 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
7183 params.sprstat_abandoned_unsent +=
7184 streamoute->abandoned_unsent[policy];
7185 params.sprstat_abandoned_sent +=
7186 streamoute->abandoned_sent[policy];
7189 params.sprstat_abandoned_unsent =
7190 streamoute->abandoned_unsent[__SCTP_PR_INDEX(policy)];
7191 params.sprstat_abandoned_sent =
7192 streamoute->abandoned_sent[__SCTP_PR_INDEX(policy)];
7195 if (put_user(len, optlen) || copy_to_user(optval, ¶ms, len)) {
7206 static int sctp_getsockopt_reconfig_supported(struct sock *sk, int len,
7207 char __user *optval,
7210 struct sctp_assoc_value params;
7211 struct sctp_association *asoc;
7212 int retval = -EFAULT;
7214 if (len < sizeof(params)) {
7219 len = sizeof(params);
7220 if (copy_from_user(¶ms, optval, len))
7223 asoc = sctp_id2assoc(sk, params.assoc_id);
7225 params.assoc_value = asoc->reconf_enable;
7226 } else if (!params.assoc_id) {
7227 struct sctp_sock *sp = sctp_sk(sk);
7229 params.assoc_value = sp->ep->reconf_enable;
7235 if (put_user(len, optlen))
7238 if (copy_to_user(optval, ¶ms, len))
7247 static int sctp_getsockopt_enable_strreset(struct sock *sk, int len,
7248 char __user *optval,
7251 struct sctp_assoc_value params;
7252 struct sctp_association *asoc;
7253 int retval = -EFAULT;
7255 if (len < sizeof(params)) {
7260 len = sizeof(params);
7261 if (copy_from_user(¶ms, optval, len))
7264 asoc = sctp_id2assoc(sk, params.assoc_id);
7266 params.assoc_value = asoc->strreset_enable;
7267 } else if (!params.assoc_id) {
7268 struct sctp_sock *sp = sctp_sk(sk);
7270 params.assoc_value = sp->ep->strreset_enable;
7276 if (put_user(len, optlen))
7279 if (copy_to_user(optval, ¶ms, len))
7288 static int sctp_getsockopt_scheduler(struct sock *sk, int len,
7289 char __user *optval,
7292 struct sctp_assoc_value params;
7293 struct sctp_association *asoc;
7294 int retval = -EFAULT;
7296 if (len < sizeof(params)) {
7301 len = sizeof(params);
7302 if (copy_from_user(¶ms, optval, len))
7305 asoc = sctp_id2assoc(sk, params.assoc_id);
7311 params.assoc_value = sctp_sched_get_sched(asoc);
7313 if (put_user(len, optlen))
7316 if (copy_to_user(optval, ¶ms, len))
7325 static int sctp_getsockopt_scheduler_value(struct sock *sk, int len,
7326 char __user *optval,
7329 struct sctp_stream_value params;
7330 struct sctp_association *asoc;
7331 int retval = -EFAULT;
7333 if (len < sizeof(params)) {
7338 len = sizeof(params);
7339 if (copy_from_user(¶ms, optval, len))
7342 asoc = sctp_id2assoc(sk, params.assoc_id);
7348 retval = sctp_sched_get_value(asoc, params.stream_id,
7349 ¶ms.stream_value);
7353 if (put_user(len, optlen)) {
7358 if (copy_to_user(optval, ¶ms, len)) {
7367 static int sctp_getsockopt_interleaving_supported(struct sock *sk, int len,
7368 char __user *optval,
7371 struct sctp_assoc_value params;
7372 struct sctp_association *asoc;
7373 int retval = -EFAULT;
7375 if (len < sizeof(params)) {
7380 len = sizeof(params);
7381 if (copy_from_user(¶ms, optval, len))
7384 asoc = sctp_id2assoc(sk, params.assoc_id);
7386 params.assoc_value = asoc->intl_enable;
7387 } else if (!params.assoc_id) {
7388 struct sctp_sock *sp = sctp_sk(sk);
7390 params.assoc_value = sp->strm_interleave;
7396 if (put_user(len, optlen))
7399 if (copy_to_user(optval, ¶ms, len))
7408 static int sctp_getsockopt_reuse_port(struct sock *sk, int len,
7409 char __user *optval,
7414 if (len < sizeof(int))
7418 val = sctp_sk(sk)->reuse;
7419 if (put_user(len, optlen))
7422 if (copy_to_user(optval, &val, len))
7428 static int sctp_getsockopt(struct sock *sk, int level, int optname,
7429 char __user *optval, int __user *optlen)
7434 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
7436 /* I can hardly begin to describe how wrong this is. This is
7437 * so broken as to be worse than useless. The API draft
7438 * REALLY is NOT helpful here... I am not convinced that the
7439 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
7440 * are at all well-founded.
7442 if (level != SOL_SCTP) {
7443 struct sctp_af *af = sctp_sk(sk)->pf->af;
7445 retval = af->getsockopt(sk, level, optname, optval, optlen);
7449 if (get_user(len, optlen))
7459 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
7461 case SCTP_DISABLE_FRAGMENTS:
7462 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
7466 retval = sctp_getsockopt_events(sk, len, optval, optlen);
7468 case SCTP_AUTOCLOSE:
7469 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
7471 case SCTP_SOCKOPT_PEELOFF:
7472 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
7474 case SCTP_SOCKOPT_PEELOFF_FLAGS:
7475 retval = sctp_getsockopt_peeloff_flags(sk, len, optval, optlen);
7477 case SCTP_PEER_ADDR_PARAMS:
7478 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
7481 case SCTP_DELAYED_SACK:
7482 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
7486 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
7488 case SCTP_GET_PEER_ADDRS:
7489 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
7492 case SCTP_GET_LOCAL_ADDRS:
7493 retval = sctp_getsockopt_local_addrs(sk, len, optval,
7496 case SCTP_SOCKOPT_CONNECTX3:
7497 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
7499 case SCTP_DEFAULT_SEND_PARAM:
7500 retval = sctp_getsockopt_default_send_param(sk, len,
7503 case SCTP_DEFAULT_SNDINFO:
7504 retval = sctp_getsockopt_default_sndinfo(sk, len,
7507 case SCTP_PRIMARY_ADDR:
7508 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
7511 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
7514 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
7516 case SCTP_ASSOCINFO:
7517 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
7519 case SCTP_I_WANT_MAPPED_V4_ADDR:
7520 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
7523 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
7525 case SCTP_GET_PEER_ADDR_INFO:
7526 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
7529 case SCTP_ADAPTATION_LAYER:
7530 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
7534 retval = sctp_getsockopt_context(sk, len, optval, optlen);
7536 case SCTP_FRAGMENT_INTERLEAVE:
7537 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
7540 case SCTP_PARTIAL_DELIVERY_POINT:
7541 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
7544 case SCTP_MAX_BURST:
7545 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
7548 case SCTP_AUTH_CHUNK:
7549 case SCTP_AUTH_DELETE_KEY:
7550 case SCTP_AUTH_DEACTIVATE_KEY:
7551 retval = -EOPNOTSUPP;
7553 case SCTP_HMAC_IDENT:
7554 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
7556 case SCTP_AUTH_ACTIVE_KEY:
7557 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
7559 case SCTP_PEER_AUTH_CHUNKS:
7560 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
7563 case SCTP_LOCAL_AUTH_CHUNKS:
7564 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
7567 case SCTP_GET_ASSOC_NUMBER:
7568 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
7570 case SCTP_GET_ASSOC_ID_LIST:
7571 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
7573 case SCTP_AUTO_ASCONF:
7574 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
7576 case SCTP_PEER_ADDR_THLDS:
7577 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
7579 case SCTP_GET_ASSOC_STATS:
7580 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
7582 case SCTP_RECVRCVINFO:
7583 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
7585 case SCTP_RECVNXTINFO:
7586 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
7588 case SCTP_PR_SUPPORTED:
7589 retval = sctp_getsockopt_pr_supported(sk, len, optval, optlen);
7591 case SCTP_DEFAULT_PRINFO:
7592 retval = sctp_getsockopt_default_prinfo(sk, len, optval,
7595 case SCTP_PR_ASSOC_STATUS:
7596 retval = sctp_getsockopt_pr_assocstatus(sk, len, optval,
7599 case SCTP_PR_STREAM_STATUS:
7600 retval = sctp_getsockopt_pr_streamstatus(sk, len, optval,
7603 case SCTP_RECONFIG_SUPPORTED:
7604 retval = sctp_getsockopt_reconfig_supported(sk, len, optval,
7607 case SCTP_ENABLE_STREAM_RESET:
7608 retval = sctp_getsockopt_enable_strreset(sk, len, optval,
7611 case SCTP_STREAM_SCHEDULER:
7612 retval = sctp_getsockopt_scheduler(sk, len, optval,
7615 case SCTP_STREAM_SCHEDULER_VALUE:
7616 retval = sctp_getsockopt_scheduler_value(sk, len, optval,
7619 case SCTP_INTERLEAVING_SUPPORTED:
7620 retval = sctp_getsockopt_interleaving_supported(sk, len, optval,
7623 case SCTP_REUSE_PORT:
7624 retval = sctp_getsockopt_reuse_port(sk, len, optval, optlen);
7627 retval = -ENOPROTOOPT;
7635 static int sctp_hash(struct sock *sk)
7641 static void sctp_unhash(struct sock *sk)
7646 /* Check if port is acceptable. Possibly find first available port.
7648 * The port hash table (contained in the 'global' SCTP protocol storage
7649 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
7650 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
7651 * list (the list number is the port number hashed out, so as you
7652 * would expect from a hash function, all the ports in a given list have
7653 * such a number that hashes out to the same list number; you were
7654 * expecting that, right?); so each list has a set of ports, with a
7655 * link to the socket (struct sock) that uses it, the port number and
7656 * a fastreuse flag (FIXME: NPI ipg).
7658 static struct sctp_bind_bucket *sctp_bucket_create(
7659 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
7661 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
7663 bool reuse = (sk->sk_reuse || sctp_sk(sk)->reuse);
7664 struct sctp_bind_hashbucket *head; /* hash list */
7665 struct sctp_bind_bucket *pp;
7666 unsigned short snum;
7669 snum = ntohs(addr->v4.sin_port);
7671 pr_debug("%s: begins, snum:%d\n", __func__, snum);
7676 /* Search for an available port. */
7677 int low, high, remaining, index;
7679 struct net *net = sock_net(sk);
7681 inet_get_local_port_range(net, &low, &high);
7682 remaining = (high - low) + 1;
7683 rover = prandom_u32() % remaining + low;
7687 if ((rover < low) || (rover > high))
7689 if (inet_is_local_reserved_port(net, rover))
7691 index = sctp_phashfn(sock_net(sk), rover);
7692 head = &sctp_port_hashtable[index];
7693 spin_lock(&head->lock);
7694 sctp_for_each_hentry(pp, &head->chain)
7695 if ((pp->port == rover) &&
7696 net_eq(sock_net(sk), pp->net))
7700 spin_unlock(&head->lock);
7701 } while (--remaining > 0);
7703 /* Exhausted local port range during search? */
7708 /* OK, here is the one we will use. HEAD (the port
7709 * hash table list entry) is non-NULL and we hold it's
7714 /* We are given an specific port number; we verify
7715 * that it is not being used. If it is used, we will
7716 * exahust the search in the hash list corresponding
7717 * to the port number (snum) - we detect that with the
7718 * port iterator, pp being NULL.
7720 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
7721 spin_lock(&head->lock);
7722 sctp_for_each_hentry(pp, &head->chain) {
7723 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
7730 if (!hlist_empty(&pp->owner)) {
7731 /* We had a port hash table hit - there is an
7732 * available port (pp != NULL) and it is being
7733 * used by other socket (pp->owner not empty); that other
7734 * socket is going to be sk2.
7738 pr_debug("%s: found a possible match\n", __func__);
7740 if (pp->fastreuse && reuse && sk->sk_state != SCTP_SS_LISTENING)
7743 /* Run through the list of sockets bound to the port
7744 * (pp->port) [via the pointers bind_next and
7745 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
7746 * we get the endpoint they describe and run through
7747 * the endpoint's list of IP (v4 or v6) addresses,
7748 * comparing each of the addresses with the address of
7749 * the socket sk. If we find a match, then that means
7750 * that this port/socket (sk) combination are already
7753 sk_for_each_bound(sk2, &pp->owner) {
7754 struct sctp_endpoint *ep2;
7755 ep2 = sctp_sk(sk2)->ep;
7758 (reuse && (sk2->sk_reuse || sctp_sk(sk2)->reuse) &&
7759 sk2->sk_state != SCTP_SS_LISTENING))
7762 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
7763 sctp_sk(sk2), sctp_sk(sk))) {
7769 pr_debug("%s: found a match\n", __func__);
7772 /* If there was a hash table miss, create a new port. */
7774 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
7777 /* In either case (hit or miss), make sure fastreuse is 1 only
7778 * if sk->sk_reuse is too (that is, if the caller requested
7779 * SO_REUSEADDR on this socket -sk-).
7781 if (hlist_empty(&pp->owner)) {
7782 if (reuse && sk->sk_state != SCTP_SS_LISTENING)
7786 } else if (pp->fastreuse &&
7787 (!reuse || sk->sk_state == SCTP_SS_LISTENING))
7790 /* We are set, so fill up all the data in the hash table
7791 * entry, tie the socket list information with the rest of the
7792 * sockets FIXME: Blurry, NPI (ipg).
7795 if (!sctp_sk(sk)->bind_hash) {
7796 inet_sk(sk)->inet_num = snum;
7797 sk_add_bind_node(sk, &pp->owner);
7798 sctp_sk(sk)->bind_hash = pp;
7803 spin_unlock(&head->lock);
7810 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
7811 * port is requested.
7813 static int sctp_get_port(struct sock *sk, unsigned short snum)
7815 union sctp_addr addr;
7816 struct sctp_af *af = sctp_sk(sk)->pf->af;
7818 /* Set up a dummy address struct from the sk. */
7819 af->from_sk(&addr, sk);
7820 addr.v4.sin_port = htons(snum);
7822 /* Note: sk->sk_num gets filled in if ephemeral port request. */
7823 return !!sctp_get_port_local(sk, &addr);
7827 * Move a socket to LISTENING state.
7829 static int sctp_listen_start(struct sock *sk, int backlog)
7831 struct sctp_sock *sp = sctp_sk(sk);
7832 struct sctp_endpoint *ep = sp->ep;
7833 struct crypto_shash *tfm = NULL;
7836 /* Allocate HMAC for generating cookie. */
7837 if (!sp->hmac && sp->sctp_hmac_alg) {
7838 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
7839 tfm = crypto_alloc_shash(alg, 0, 0);
7841 net_info_ratelimited("failed to load transform for %s: %ld\n",
7842 sp->sctp_hmac_alg, PTR_ERR(tfm));
7845 sctp_sk(sk)->hmac = tfm;
7849 * If a bind() or sctp_bindx() is not called prior to a listen()
7850 * call that allows new associations to be accepted, the system
7851 * picks an ephemeral port and will choose an address set equivalent
7852 * to binding with a wildcard address.
7854 * This is not currently spelled out in the SCTP sockets
7855 * extensions draft, but follows the practice as seen in TCP
7859 inet_sk_set_state(sk, SCTP_SS_LISTENING);
7860 if (!ep->base.bind_addr.port) {
7861 if (sctp_autobind(sk))
7864 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
7865 inet_sk_set_state(sk, SCTP_SS_CLOSED);
7870 sk->sk_max_ack_backlog = backlog;
7871 sctp_hash_endpoint(ep);
7876 * 4.1.3 / 5.1.3 listen()
7878 * By default, new associations are not accepted for UDP style sockets.
7879 * An application uses listen() to mark a socket as being able to
7880 * accept new associations.
7882 * On TCP style sockets, applications use listen() to ready the SCTP
7883 * endpoint for accepting inbound associations.
7885 * On both types of endpoints a backlog of '0' disables listening.
7887 * Move a socket to LISTENING state.
7889 int sctp_inet_listen(struct socket *sock, int backlog)
7891 struct sock *sk = sock->sk;
7892 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
7895 if (unlikely(backlog < 0))
7900 /* Peeled-off sockets are not allowed to listen(). */
7901 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
7904 if (sock->state != SS_UNCONNECTED)
7907 if (!sctp_sstate(sk, LISTENING) && !sctp_sstate(sk, CLOSED))
7910 /* If backlog is zero, disable listening. */
7912 if (sctp_sstate(sk, CLOSED))
7916 sctp_unhash_endpoint(ep);
7917 sk->sk_state = SCTP_SS_CLOSED;
7918 if (sk->sk_reuse || sctp_sk(sk)->reuse)
7919 sctp_sk(sk)->bind_hash->fastreuse = 1;
7923 /* If we are already listening, just update the backlog */
7924 if (sctp_sstate(sk, LISTENING))
7925 sk->sk_max_ack_backlog = backlog;
7927 err = sctp_listen_start(sk, backlog);
7939 * This function is done by modeling the current datagram_poll() and the
7940 * tcp_poll(). Note that, based on these implementations, we don't
7941 * lock the socket in this function, even though it seems that,
7942 * ideally, locking or some other mechanisms can be used to ensure
7943 * the integrity of the counters (sndbuf and wmem_alloc) used
7944 * in this place. We assume that we don't need locks either until proven
7947 * Another thing to note is that we include the Async I/O support
7948 * here, again, by modeling the current TCP/UDP code. We don't have
7949 * a good way to test with it yet.
7951 __poll_t sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
7953 struct sock *sk = sock->sk;
7954 struct sctp_sock *sp = sctp_sk(sk);
7957 poll_wait(file, sk_sleep(sk), wait);
7959 sock_rps_record_flow(sk);
7961 /* A TCP-style listening socket becomes readable when the accept queue
7964 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
7965 return (!list_empty(&sp->ep->asocs)) ?
7966 (EPOLLIN | EPOLLRDNORM) : 0;
7970 /* Is there any exceptional events? */
7971 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
7973 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? EPOLLPRI : 0);
7974 if (sk->sk_shutdown & RCV_SHUTDOWN)
7975 mask |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
7976 if (sk->sk_shutdown == SHUTDOWN_MASK)
7979 /* Is it readable? Reconsider this code with TCP-style support. */
7980 if (!skb_queue_empty(&sk->sk_receive_queue))
7981 mask |= EPOLLIN | EPOLLRDNORM;
7983 /* The association is either gone or not ready. */
7984 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
7987 /* Is it writable? */
7988 if (sctp_writeable(sk)) {
7989 mask |= EPOLLOUT | EPOLLWRNORM;
7991 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
7993 * Since the socket is not locked, the buffer
7994 * might be made available after the writeable check and
7995 * before the bit is set. This could cause a lost I/O
7996 * signal. tcp_poll() has a race breaker for this race
7997 * condition. Based on their implementation, we put
7998 * in the following code to cover it as well.
8000 if (sctp_writeable(sk))
8001 mask |= EPOLLOUT | EPOLLWRNORM;
8006 /********************************************************************
8007 * 2nd Level Abstractions
8008 ********************************************************************/
8010 static struct sctp_bind_bucket *sctp_bucket_create(
8011 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
8013 struct sctp_bind_bucket *pp;
8015 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
8017 SCTP_DBG_OBJCNT_INC(bind_bucket);
8020 INIT_HLIST_HEAD(&pp->owner);
8022 hlist_add_head(&pp->node, &head->chain);
8027 /* Caller must hold hashbucket lock for this tb with local BH disabled */
8028 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
8030 if (pp && hlist_empty(&pp->owner)) {
8031 __hlist_del(&pp->node);
8032 kmem_cache_free(sctp_bucket_cachep, pp);
8033 SCTP_DBG_OBJCNT_DEC(bind_bucket);
8037 /* Release this socket's reference to a local port. */
8038 static inline void __sctp_put_port(struct sock *sk)
8040 struct sctp_bind_hashbucket *head =
8041 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
8042 inet_sk(sk)->inet_num)];
8043 struct sctp_bind_bucket *pp;
8045 spin_lock(&head->lock);
8046 pp = sctp_sk(sk)->bind_hash;
8047 __sk_del_bind_node(sk);
8048 sctp_sk(sk)->bind_hash = NULL;
8049 inet_sk(sk)->inet_num = 0;
8050 sctp_bucket_destroy(pp);
8051 spin_unlock(&head->lock);
8054 void sctp_put_port(struct sock *sk)
8057 __sctp_put_port(sk);
8062 * The system picks an ephemeral port and choose an address set equivalent
8063 * to binding with a wildcard address.
8064 * One of those addresses will be the primary address for the association.
8065 * This automatically enables the multihoming capability of SCTP.
8067 static int sctp_autobind(struct sock *sk)
8069 union sctp_addr autoaddr;
8073 /* Initialize a local sockaddr structure to INADDR_ANY. */
8074 af = sctp_sk(sk)->pf->af;
8076 port = htons(inet_sk(sk)->inet_num);
8077 af->inaddr_any(&autoaddr, port);
8079 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
8082 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
8085 * 4.2 The cmsghdr Structure *
8087 * When ancillary data is sent or received, any number of ancillary data
8088 * objects can be specified by the msg_control and msg_controllen members of
8089 * the msghdr structure, because each object is preceded by
8090 * a cmsghdr structure defining the object's length (the cmsg_len member).
8091 * Historically Berkeley-derived implementations have passed only one object
8092 * at a time, but this API allows multiple objects to be
8093 * passed in a single call to sendmsg() or recvmsg(). The following example
8094 * shows two ancillary data objects in a control buffer.
8096 * |<--------------------------- msg_controllen -------------------------->|
8099 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
8101 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
8104 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
8106 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
8109 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
8110 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
8112 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
8114 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
8121 static int sctp_msghdr_parse(const struct msghdr *msg, struct sctp_cmsgs *cmsgs)
8123 struct msghdr *my_msg = (struct msghdr *)msg;
8124 struct cmsghdr *cmsg;
8126 for_each_cmsghdr(cmsg, my_msg) {
8127 if (!CMSG_OK(my_msg, cmsg))
8130 /* Should we parse this header or ignore? */
8131 if (cmsg->cmsg_level != IPPROTO_SCTP)
8134 /* Strictly check lengths following example in SCM code. */
8135 switch (cmsg->cmsg_type) {
8137 /* SCTP Socket API Extension
8138 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
8140 * This cmsghdr structure provides information for
8141 * initializing new SCTP associations with sendmsg().
8142 * The SCTP_INITMSG socket option uses this same data
8143 * structure. This structure is not used for
8146 * cmsg_level cmsg_type cmsg_data[]
8147 * ------------ ------------ ----------------------
8148 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
8150 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
8153 cmsgs->init = CMSG_DATA(cmsg);
8157 /* SCTP Socket API Extension
8158 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
8160 * This cmsghdr structure specifies SCTP options for
8161 * sendmsg() and describes SCTP header information
8162 * about a received message through recvmsg().
8164 * cmsg_level cmsg_type cmsg_data[]
8165 * ------------ ------------ ----------------------
8166 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
8168 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
8171 cmsgs->srinfo = CMSG_DATA(cmsg);
8173 if (cmsgs->srinfo->sinfo_flags &
8174 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
8175 SCTP_SACK_IMMEDIATELY | SCTP_SENDALL |
8176 SCTP_PR_SCTP_MASK | SCTP_ABORT | SCTP_EOF))
8181 /* SCTP Socket API Extension
8182 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
8184 * This cmsghdr structure specifies SCTP options for
8185 * sendmsg(). This structure and SCTP_RCVINFO replaces
8186 * SCTP_SNDRCV which has been deprecated.
8188 * cmsg_level cmsg_type cmsg_data[]
8189 * ------------ ------------ ---------------------
8190 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
8192 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
8195 cmsgs->sinfo = CMSG_DATA(cmsg);
8197 if (cmsgs->sinfo->snd_flags &
8198 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
8199 SCTP_SACK_IMMEDIATELY | SCTP_SENDALL |
8200 SCTP_PR_SCTP_MASK | SCTP_ABORT | SCTP_EOF))
8204 /* SCTP Socket API Extension
8205 * 5.3.7 SCTP PR-SCTP Information Structure (SCTP_PRINFO)
8207 * This cmsghdr structure specifies SCTP options for sendmsg().
8209 * cmsg_level cmsg_type cmsg_data[]
8210 * ------------ ------------ ---------------------
8211 * IPPROTO_SCTP SCTP_PRINFO struct sctp_prinfo
8213 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_prinfo)))
8216 cmsgs->prinfo = CMSG_DATA(cmsg);
8217 if (cmsgs->prinfo->pr_policy & ~SCTP_PR_SCTP_MASK)
8220 if (cmsgs->prinfo->pr_policy == SCTP_PR_SCTP_NONE)
8221 cmsgs->prinfo->pr_value = 0;
8224 /* SCTP Socket API Extension
8225 * 5.3.8 SCTP AUTH Information Structure (SCTP_AUTHINFO)
8227 * This cmsghdr structure specifies SCTP options for sendmsg().
8229 * cmsg_level cmsg_type cmsg_data[]
8230 * ------------ ------------ ---------------------
8231 * IPPROTO_SCTP SCTP_AUTHINFO struct sctp_authinfo
8233 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_authinfo)))
8236 cmsgs->authinfo = CMSG_DATA(cmsg);
8238 case SCTP_DSTADDRV4:
8239 case SCTP_DSTADDRV6:
8240 /* SCTP Socket API Extension
8241 * 5.3.9/10 SCTP Destination IPv4/6 Address Structure (SCTP_DSTADDRV4/6)
8243 * This cmsghdr structure specifies SCTP options for sendmsg().
8245 * cmsg_level cmsg_type cmsg_data[]
8246 * ------------ ------------ ---------------------
8247 * IPPROTO_SCTP SCTP_DSTADDRV4 struct in_addr
8248 * ------------ ------------ ---------------------
8249 * IPPROTO_SCTP SCTP_DSTADDRV6 struct in6_addr
8251 cmsgs->addrs_msg = my_msg;
8262 * Wait for a packet..
8263 * Note: This function is the same function as in core/datagram.c
8264 * with a few modifications to make lksctp work.
8266 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
8271 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
8273 /* Socket errors? */
8274 error = sock_error(sk);
8278 if (!skb_queue_empty(&sk->sk_receive_queue))
8281 /* Socket shut down? */
8282 if (sk->sk_shutdown & RCV_SHUTDOWN)
8285 /* Sequenced packets can come disconnected. If so we report the
8290 /* Is there a good reason to think that we may receive some data? */
8291 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
8294 /* Handle signals. */
8295 if (signal_pending(current))
8298 /* Let another process have a go. Since we are going to sleep
8299 * anyway. Note: This may cause odd behaviors if the message
8300 * does not fit in the user's buffer, but this seems to be the
8301 * only way to honor MSG_DONTWAIT realistically.
8304 *timeo_p = schedule_timeout(*timeo_p);
8308 finish_wait(sk_sleep(sk), &wait);
8312 error = sock_intr_errno(*timeo_p);
8315 finish_wait(sk_sleep(sk), &wait);
8320 /* Receive a datagram.
8321 * Note: This is pretty much the same routine as in core/datagram.c
8322 * with a few changes to make lksctp work.
8324 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
8325 int noblock, int *err)
8328 struct sk_buff *skb;
8331 timeo = sock_rcvtimeo(sk, noblock);
8333 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
8334 MAX_SCHEDULE_TIMEOUT);
8337 /* Again only user level code calls this function,
8338 * so nothing interrupt level
8339 * will suddenly eat the receive_queue.
8341 * Look at current nfs client by the way...
8342 * However, this function was correct in any case. 8)
8344 if (flags & MSG_PEEK) {
8345 skb = skb_peek(&sk->sk_receive_queue);
8347 refcount_inc(&skb->users);
8349 skb = __skb_dequeue(&sk->sk_receive_queue);
8355 /* Caller is allowed not to check sk->sk_err before calling. */
8356 error = sock_error(sk);
8360 if (sk->sk_shutdown & RCV_SHUTDOWN)
8363 if (sk_can_busy_loop(sk)) {
8364 sk_busy_loop(sk, noblock);
8366 if (!skb_queue_empty(&sk->sk_receive_queue))
8370 /* User doesn't want to wait. */
8374 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
8383 /* If sndbuf has changed, wake up per association sndbuf waiters. */
8384 static void __sctp_write_space(struct sctp_association *asoc)
8386 struct sock *sk = asoc->base.sk;
8388 if (sctp_wspace(asoc) <= 0)
8391 if (waitqueue_active(&asoc->wait))
8392 wake_up_interruptible(&asoc->wait);
8394 if (sctp_writeable(sk)) {
8395 struct socket_wq *wq;
8398 wq = rcu_dereference(sk->sk_wq);
8400 if (waitqueue_active(&wq->wait))
8401 wake_up_interruptible(&wq->wait);
8403 /* Note that we try to include the Async I/O support
8404 * here by modeling from the current TCP/UDP code.
8405 * We have not tested with it yet.
8407 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
8408 sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT);
8414 static void sctp_wake_up_waiters(struct sock *sk,
8415 struct sctp_association *asoc)
8417 struct sctp_association *tmp = asoc;
8419 /* We do accounting for the sndbuf space per association,
8420 * so we only need to wake our own association.
8422 if (asoc->ep->sndbuf_policy)
8423 return __sctp_write_space(asoc);
8425 /* If association goes down and is just flushing its
8426 * outq, then just normally notify others.
8428 if (asoc->base.dead)
8429 return sctp_write_space(sk);
8431 /* Accounting for the sndbuf space is per socket, so we
8432 * need to wake up others, try to be fair and in case of
8433 * other associations, let them have a go first instead
8434 * of just doing a sctp_write_space() call.
8436 * Note that we reach sctp_wake_up_waiters() only when
8437 * associations free up queued chunks, thus we are under
8438 * lock and the list of associations on a socket is
8439 * guaranteed not to change.
8441 for (tmp = list_next_entry(tmp, asocs); 1;
8442 tmp = list_next_entry(tmp, asocs)) {
8443 /* Manually skip the head element. */
8444 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
8446 /* Wake up association. */
8447 __sctp_write_space(tmp);
8448 /* We've reached the end. */
8454 /* Do accounting for the sndbuf space.
8455 * Decrement the used sndbuf space of the corresponding association by the
8456 * data size which was just transmitted(freed).
8458 static void sctp_wfree(struct sk_buff *skb)
8460 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
8461 struct sctp_association *asoc = chunk->asoc;
8462 struct sock *sk = asoc->base.sk;
8464 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
8465 sizeof(struct sk_buff) +
8466 sizeof(struct sctp_chunk);
8468 WARN_ON(refcount_sub_and_test(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc));
8471 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
8473 sk->sk_wmem_queued -= skb->truesize;
8474 sk_mem_uncharge(sk, skb->truesize);
8477 struct sctp_shared_key *shkey = chunk->shkey;
8479 /* refcnt == 2 and !list_empty mean after this release, it's
8480 * not being used anywhere, and it's time to notify userland
8481 * that this shkey can be freed if it's been deactivated.
8483 if (shkey->deactivated && !list_empty(&shkey->key_list) &&
8484 refcount_read(&shkey->refcnt) == 2) {
8485 struct sctp_ulpevent *ev;
8487 ev = sctp_ulpevent_make_authkey(asoc, shkey->key_id,
8491 asoc->stream.si->enqueue_event(&asoc->ulpq, ev);
8493 sctp_auth_shkey_release(chunk->shkey);
8497 sctp_wake_up_waiters(sk, asoc);
8499 sctp_association_put(asoc);
8502 /* Do accounting for the receive space on the socket.
8503 * Accounting for the association is done in ulpevent.c
8504 * We set this as a destructor for the cloned data skbs so that
8505 * accounting is done at the correct time.
8507 void sctp_sock_rfree(struct sk_buff *skb)
8509 struct sock *sk = skb->sk;
8510 struct sctp_ulpevent *event = sctp_skb2event(skb);
8512 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
8515 * Mimic the behavior of sock_rfree
8517 sk_mem_uncharge(sk, event->rmem_len);
8521 /* Helper function to wait for space in the sndbuf. */
8522 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
8525 struct sock *sk = asoc->base.sk;
8526 long current_timeo = *timeo_p;
8530 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
8533 /* Increment the association's refcnt. */
8534 sctp_association_hold(asoc);
8536 /* Wait on the association specific sndbuf space. */
8538 prepare_to_wait_exclusive(&asoc->wait, &wait,
8539 TASK_INTERRUPTIBLE);
8540 if (asoc->base.dead)
8544 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING)
8546 if (signal_pending(current))
8547 goto do_interrupted;
8548 if (msg_len <= sctp_wspace(asoc))
8551 /* Let another process have a go. Since we are going
8555 current_timeo = schedule_timeout(current_timeo);
8557 if (sk != asoc->base.sk)
8560 *timeo_p = current_timeo;
8564 finish_wait(&asoc->wait, &wait);
8566 /* Release the association's refcnt. */
8567 sctp_association_put(asoc);
8580 err = sock_intr_errno(*timeo_p);
8588 void sctp_data_ready(struct sock *sk)
8590 struct socket_wq *wq;
8593 wq = rcu_dereference(sk->sk_wq);
8594 if (skwq_has_sleeper(wq))
8595 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
8596 EPOLLRDNORM | EPOLLRDBAND);
8597 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
8601 /* If socket sndbuf has changed, wake up all per association waiters. */
8602 void sctp_write_space(struct sock *sk)
8604 struct sctp_association *asoc;
8606 /* Wake up the tasks in each wait queue. */
8607 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
8608 __sctp_write_space(asoc);
8612 /* Is there any sndbuf space available on the socket?
8614 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
8615 * associations on the same socket. For a UDP-style socket with
8616 * multiple associations, it is possible for it to be "unwriteable"
8617 * prematurely. I assume that this is acceptable because
8618 * a premature "unwriteable" is better than an accidental "writeable" which
8619 * would cause an unwanted block under certain circumstances. For the 1-1
8620 * UDP-style sockets or TCP-style sockets, this code should work.
8623 static int sctp_writeable(struct sock *sk)
8627 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
8633 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
8634 * returns immediately with EINPROGRESS.
8636 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
8638 struct sock *sk = asoc->base.sk;
8640 long current_timeo = *timeo_p;
8643 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
8645 /* Increment the association's refcnt. */
8646 sctp_association_hold(asoc);
8649 prepare_to_wait_exclusive(&asoc->wait, &wait,
8650 TASK_INTERRUPTIBLE);
8653 if (sk->sk_shutdown & RCV_SHUTDOWN)
8655 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
8658 if (signal_pending(current))
8659 goto do_interrupted;
8661 if (sctp_state(asoc, ESTABLISHED))
8664 /* Let another process have a go. Since we are going
8668 current_timeo = schedule_timeout(current_timeo);
8671 *timeo_p = current_timeo;
8675 finish_wait(&asoc->wait, &wait);
8677 /* Release the association's refcnt. */
8678 sctp_association_put(asoc);
8683 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
8686 err = -ECONNREFUSED;
8690 err = sock_intr_errno(*timeo_p);
8698 static int sctp_wait_for_accept(struct sock *sk, long timeo)
8700 struct sctp_endpoint *ep;
8704 ep = sctp_sk(sk)->ep;
8708 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
8709 TASK_INTERRUPTIBLE);
8711 if (list_empty(&ep->asocs)) {
8713 timeo = schedule_timeout(timeo);
8718 if (!sctp_sstate(sk, LISTENING))
8722 if (!list_empty(&ep->asocs))
8725 err = sock_intr_errno(timeo);
8726 if (signal_pending(current))
8734 finish_wait(sk_sleep(sk), &wait);
8739 static void sctp_wait_for_close(struct sock *sk, long timeout)
8744 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
8745 if (list_empty(&sctp_sk(sk)->ep->asocs))
8748 timeout = schedule_timeout(timeout);
8750 } while (!signal_pending(current) && timeout);
8752 finish_wait(sk_sleep(sk), &wait);
8755 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
8757 struct sk_buff *frag;
8762 /* Don't forget the fragments. */
8763 skb_walk_frags(skb, frag)
8764 sctp_skb_set_owner_r_frag(frag, sk);
8767 sctp_skb_set_owner_r(skb, sk);
8770 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
8771 struct sctp_association *asoc)
8773 struct inet_sock *inet = inet_sk(sk);
8774 struct inet_sock *newinet;
8775 struct sctp_sock *sp = sctp_sk(sk);
8776 struct sctp_endpoint *ep = sp->ep;
8778 newsk->sk_type = sk->sk_type;
8779 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
8780 newsk->sk_flags = sk->sk_flags;
8781 newsk->sk_tsflags = sk->sk_tsflags;
8782 newsk->sk_no_check_tx = sk->sk_no_check_tx;
8783 newsk->sk_no_check_rx = sk->sk_no_check_rx;
8784 newsk->sk_reuse = sk->sk_reuse;
8785 sctp_sk(newsk)->reuse = sp->reuse;
8787 newsk->sk_shutdown = sk->sk_shutdown;
8788 newsk->sk_destruct = sctp_destruct_sock;
8789 newsk->sk_family = sk->sk_family;
8790 newsk->sk_protocol = IPPROTO_SCTP;
8791 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
8792 newsk->sk_sndbuf = sk->sk_sndbuf;
8793 newsk->sk_rcvbuf = sk->sk_rcvbuf;
8794 newsk->sk_lingertime = sk->sk_lingertime;
8795 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
8796 newsk->sk_sndtimeo = sk->sk_sndtimeo;
8797 newsk->sk_rxhash = sk->sk_rxhash;
8799 newinet = inet_sk(newsk);
8801 /* Initialize sk's sport, dport, rcv_saddr and daddr for
8802 * getsockname() and getpeername()
8804 newinet->inet_sport = inet->inet_sport;
8805 newinet->inet_saddr = inet->inet_saddr;
8806 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
8807 newinet->inet_dport = htons(asoc->peer.port);
8808 newinet->pmtudisc = inet->pmtudisc;
8809 newinet->inet_id = asoc->next_tsn ^ jiffies;
8811 newinet->uc_ttl = inet->uc_ttl;
8812 newinet->mc_loop = 1;
8813 newinet->mc_ttl = 1;
8814 newinet->mc_index = 0;
8815 newinet->mc_list = NULL;
8817 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
8818 net_enable_timestamp();
8820 /* Set newsk security attributes from orginal sk and connection
8821 * security attribute from ep.
8823 security_sctp_sk_clone(ep, sk, newsk);
8826 static inline void sctp_copy_descendant(struct sock *sk_to,
8827 const struct sock *sk_from)
8829 int ancestor_size = sizeof(struct inet_sock) +
8830 sizeof(struct sctp_sock) -
8831 offsetof(struct sctp_sock, auto_asconf_list);
8833 if (sk_from->sk_family == PF_INET6)
8834 ancestor_size += sizeof(struct ipv6_pinfo);
8836 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
8839 /* Populate the fields of the newsk from the oldsk and migrate the assoc
8840 * and its messages to the newsk.
8842 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
8843 struct sctp_association *assoc,
8844 enum sctp_socket_type type)
8846 struct sctp_sock *oldsp = sctp_sk(oldsk);
8847 struct sctp_sock *newsp = sctp_sk(newsk);
8848 struct sctp_bind_bucket *pp; /* hash list port iterator */
8849 struct sctp_endpoint *newep = newsp->ep;
8850 struct sk_buff *skb, *tmp;
8851 struct sctp_ulpevent *event;
8852 struct sctp_bind_hashbucket *head;
8854 /* Migrate socket buffer sizes and all the socket level options to the
8857 newsk->sk_sndbuf = oldsk->sk_sndbuf;
8858 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
8859 /* Brute force copy old sctp opt. */
8860 sctp_copy_descendant(newsk, oldsk);
8862 /* Restore the ep value that was overwritten with the above structure
8868 /* Hook this new socket in to the bind_hash list. */
8869 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
8870 inet_sk(oldsk)->inet_num)];
8871 spin_lock_bh(&head->lock);
8872 pp = sctp_sk(oldsk)->bind_hash;
8873 sk_add_bind_node(newsk, &pp->owner);
8874 sctp_sk(newsk)->bind_hash = pp;
8875 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
8876 spin_unlock_bh(&head->lock);
8878 /* Copy the bind_addr list from the original endpoint to the new
8879 * endpoint so that we can handle restarts properly
8881 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
8882 &oldsp->ep->base.bind_addr, GFP_KERNEL);
8884 /* Move any messages in the old socket's receive queue that are for the
8885 * peeled off association to the new socket's receive queue.
8887 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
8888 event = sctp_skb2event(skb);
8889 if (event->asoc == assoc) {
8890 __skb_unlink(skb, &oldsk->sk_receive_queue);
8891 __skb_queue_tail(&newsk->sk_receive_queue, skb);
8892 sctp_skb_set_owner_r_frag(skb, newsk);
8896 /* Clean up any messages pending delivery due to partial
8897 * delivery. Three cases:
8898 * 1) No partial deliver; no work.
8899 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
8900 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
8902 skb_queue_head_init(&newsp->pd_lobby);
8903 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
8905 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
8906 struct sk_buff_head *queue;
8908 /* Decide which queue to move pd_lobby skbs to. */
8909 if (assoc->ulpq.pd_mode) {
8910 queue = &newsp->pd_lobby;
8912 queue = &newsk->sk_receive_queue;
8914 /* Walk through the pd_lobby, looking for skbs that
8915 * need moved to the new socket.
8917 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
8918 event = sctp_skb2event(skb);
8919 if (event->asoc == assoc) {
8920 __skb_unlink(skb, &oldsp->pd_lobby);
8921 __skb_queue_tail(queue, skb);
8922 sctp_skb_set_owner_r_frag(skb, newsk);
8926 /* Clear up any skbs waiting for the partial
8927 * delivery to finish.
8929 if (assoc->ulpq.pd_mode)
8930 sctp_clear_pd(oldsk, NULL);
8934 sctp_for_each_rx_skb(assoc, newsk, sctp_skb_set_owner_r_frag);
8936 /* Set the type of socket to indicate that it is peeled off from the
8937 * original UDP-style socket or created with the accept() call on a
8938 * TCP-style socket..
8942 /* Mark the new socket "in-use" by the user so that any packets
8943 * that may arrive on the association after we've moved it are
8944 * queued to the backlog. This prevents a potential race between
8945 * backlog processing on the old socket and new-packet processing
8946 * on the new socket.
8948 * The caller has just allocated newsk so we can guarantee that other
8949 * paths won't try to lock it and then oldsk.
8951 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
8952 sctp_for_each_tx_datachunk(assoc, sctp_clear_owner_w);
8953 sctp_assoc_migrate(assoc, newsk);
8954 sctp_for_each_tx_datachunk(assoc, sctp_set_owner_w);
8956 /* If the association on the newsk is already closed before accept()
8957 * is called, set RCV_SHUTDOWN flag.
8959 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP)) {
8960 inet_sk_set_state(newsk, SCTP_SS_CLOSED);
8961 newsk->sk_shutdown |= RCV_SHUTDOWN;
8963 inet_sk_set_state(newsk, SCTP_SS_ESTABLISHED);
8966 release_sock(newsk);
8970 /* This proto struct describes the ULP interface for SCTP. */
8971 struct proto sctp_prot = {
8973 .owner = THIS_MODULE,
8974 .close = sctp_close,
8975 .disconnect = sctp_disconnect,
8976 .accept = sctp_accept,
8977 .ioctl = sctp_ioctl,
8978 .init = sctp_init_sock,
8979 .destroy = sctp_destroy_sock,
8980 .shutdown = sctp_shutdown,
8981 .setsockopt = sctp_setsockopt,
8982 .getsockopt = sctp_getsockopt,
8983 .sendmsg = sctp_sendmsg,
8984 .recvmsg = sctp_recvmsg,
8986 .backlog_rcv = sctp_backlog_rcv,
8988 .unhash = sctp_unhash,
8989 .get_port = sctp_get_port,
8990 .obj_size = sizeof(struct sctp_sock),
8991 .useroffset = offsetof(struct sctp_sock, subscribe),
8992 .usersize = offsetof(struct sctp_sock, initmsg) -
8993 offsetof(struct sctp_sock, subscribe) +
8994 sizeof_field(struct sctp_sock, initmsg),
8995 .sysctl_mem = sysctl_sctp_mem,
8996 .sysctl_rmem = sysctl_sctp_rmem,
8997 .sysctl_wmem = sysctl_sctp_wmem,
8998 .memory_pressure = &sctp_memory_pressure,
8999 .enter_memory_pressure = sctp_enter_memory_pressure,
9000 .memory_allocated = &sctp_memory_allocated,
9001 .sockets_allocated = &sctp_sockets_allocated,
9004 #if IS_ENABLED(CONFIG_IPV6)
9006 #include <net/transp_v6.h>
9007 static void sctp_v6_destroy_sock(struct sock *sk)
9009 sctp_destroy_sock(sk);
9010 inet6_destroy_sock(sk);
9013 struct proto sctpv6_prot = {
9015 .owner = THIS_MODULE,
9016 .close = sctp_close,
9017 .disconnect = sctp_disconnect,
9018 .accept = sctp_accept,
9019 .ioctl = sctp_ioctl,
9020 .init = sctp_init_sock,
9021 .destroy = sctp_v6_destroy_sock,
9022 .shutdown = sctp_shutdown,
9023 .setsockopt = sctp_setsockopt,
9024 .getsockopt = sctp_getsockopt,
9025 .sendmsg = sctp_sendmsg,
9026 .recvmsg = sctp_recvmsg,
9028 .backlog_rcv = sctp_backlog_rcv,
9030 .unhash = sctp_unhash,
9031 .get_port = sctp_get_port,
9032 .obj_size = sizeof(struct sctp6_sock),
9033 .useroffset = offsetof(struct sctp6_sock, sctp.subscribe),
9034 .usersize = offsetof(struct sctp6_sock, sctp.initmsg) -
9035 offsetof(struct sctp6_sock, sctp.subscribe) +
9036 sizeof_field(struct sctp6_sock, sctp.initmsg),
9037 .sysctl_mem = sysctl_sctp_mem,
9038 .sysctl_rmem = sysctl_sctp_rmem,
9039 .sysctl_wmem = sysctl_sctp_wmem,
9040 .memory_pressure = &sctp_memory_pressure,
9041 .enter_memory_pressure = sctp_enter_memory_pressure,
9042 .memory_allocated = &sctp_memory_allocated,
9043 .sockets_allocated = &sctp_sockets_allocated,
9045 #endif /* IS_ENABLED(CONFIG_IPV6) */