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>
61 #include <linux/capability.h>
62 #include <linux/fcntl.h>
63 #include <linux/poll.h>
64 #include <linux/init.h>
65 #include <linux/slab.h>
66 #include <linux/file.h>
67 #include <linux/compat.h>
71 #include <net/route.h>
73 #include <net/inet_common.h>
74 #include <net/busy_poll.h>
76 #include <linux/socket.h> /* for sa_family_t */
77 #include <linux/export.h>
79 #include <net/sctp/sctp.h>
80 #include <net/sctp/sm.h>
82 /* Forward declarations for internal helper functions. */
83 static int sctp_writeable(struct sock
*sk
);
84 static void sctp_wfree(struct sk_buff
*skb
);
85 static int sctp_wait_for_sndbuf(struct sctp_association
*, long *timeo_p
,
87 static int sctp_wait_for_packet(struct sock
*sk
, int *err
, long *timeo_p
);
88 static int sctp_wait_for_connect(struct sctp_association
*, long *timeo_p
);
89 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
);
90 static void sctp_wait_for_close(struct sock
*sk
, long timeo
);
91 static void sctp_destruct_sock(struct sock
*sk
);
92 static struct sctp_af
*sctp_sockaddr_af(struct sctp_sock
*opt
,
93 union sctp_addr
*addr
, int len
);
94 static int sctp_bindx_add(struct sock
*, struct sockaddr
*, int);
95 static int sctp_bindx_rem(struct sock
*, struct sockaddr
*, int);
96 static int sctp_send_asconf_add_ip(struct sock
*, struct sockaddr
*, int);
97 static int sctp_send_asconf_del_ip(struct sock
*, struct sockaddr
*, int);
98 static int sctp_send_asconf(struct sctp_association
*asoc
,
99 struct sctp_chunk
*chunk
);
100 static int sctp_do_bind(struct sock
*, union sctp_addr
*, int);
101 static int sctp_autobind(struct sock
*sk
);
102 static void sctp_sock_migrate(struct sock
*, struct sock
*,
103 struct sctp_association
*, sctp_socket_type_t
);
105 static int sctp_memory_pressure
;
106 static atomic_long_t sctp_memory_allocated
;
107 struct percpu_counter sctp_sockets_allocated
;
109 static void sctp_enter_memory_pressure(struct sock
*sk
)
111 sctp_memory_pressure
= 1;
115 /* Get the sndbuf space available at the time on the association. */
116 static inline int sctp_wspace(struct sctp_association
*asoc
)
120 if (asoc
->ep
->sndbuf_policy
)
121 amt
= asoc
->sndbuf_used
;
123 amt
= sk_wmem_alloc_get(asoc
->base
.sk
);
125 if (amt
>= asoc
->base
.sk
->sk_sndbuf
) {
126 if (asoc
->base
.sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)
129 amt
= sk_stream_wspace(asoc
->base
.sk
);
134 amt
= asoc
->base
.sk
->sk_sndbuf
- amt
;
139 /* Increment the used sndbuf space count of the corresponding association by
140 * the size of the outgoing data chunk.
141 * Also, set the skb destructor for sndbuf accounting later.
143 * Since it is always 1-1 between chunk and skb, and also a new skb is always
144 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
145 * destructor in the data chunk skb for the purpose of the sndbuf space
148 static inline void sctp_set_owner_w(struct sctp_chunk
*chunk
)
150 struct sctp_association
*asoc
= chunk
->asoc
;
151 struct sock
*sk
= asoc
->base
.sk
;
153 /* The sndbuf space is tracked per association. */
154 sctp_association_hold(asoc
);
156 skb_set_owner_w(chunk
->skb
, sk
);
158 chunk
->skb
->destructor
= sctp_wfree
;
159 /* Save the chunk pointer in skb for sctp_wfree to use later. */
160 skb_shinfo(chunk
->skb
)->destructor_arg
= chunk
;
162 asoc
->sndbuf_used
+= SCTP_DATA_SNDSIZE(chunk
) +
163 sizeof(struct sk_buff
) +
164 sizeof(struct sctp_chunk
);
166 atomic_add(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
167 sk
->sk_wmem_queued
+= chunk
->skb
->truesize
;
168 sk_mem_charge(sk
, chunk
->skb
->truesize
);
171 /* Verify that this is a valid address. */
172 static inline int sctp_verify_addr(struct sock
*sk
, union sctp_addr
*addr
,
177 /* Verify basic sockaddr. */
178 af
= sctp_sockaddr_af(sctp_sk(sk
), addr
, len
);
182 /* Is this a valid SCTP address? */
183 if (!af
->addr_valid(addr
, sctp_sk(sk
), NULL
))
186 if (!sctp_sk(sk
)->pf
->send_verify(sctp_sk(sk
), (addr
)))
192 /* Look up the association by its id. If this is not a UDP-style
193 * socket, the ID field is always ignored.
195 struct sctp_association
*sctp_id2assoc(struct sock
*sk
, sctp_assoc_t id
)
197 struct sctp_association
*asoc
= NULL
;
199 /* If this is not a UDP-style socket, assoc id should be ignored. */
200 if (!sctp_style(sk
, UDP
)) {
201 /* Return NULL if the socket state is not ESTABLISHED. It
202 * could be a TCP-style listening socket or a socket which
203 * hasn't yet called connect() to establish an association.
205 if (!sctp_sstate(sk
, ESTABLISHED
) && !sctp_sstate(sk
, CLOSING
))
208 /* Get the first and the only association from the list. */
209 if (!list_empty(&sctp_sk(sk
)->ep
->asocs
))
210 asoc
= list_entry(sctp_sk(sk
)->ep
->asocs
.next
,
211 struct sctp_association
, asocs
);
215 /* Otherwise this is a UDP-style socket. */
216 if (!id
|| (id
== (sctp_assoc_t
)-1))
219 spin_lock_bh(&sctp_assocs_id_lock
);
220 asoc
= (struct sctp_association
*)idr_find(&sctp_assocs_id
, (int)id
);
221 spin_unlock_bh(&sctp_assocs_id_lock
);
223 if (!asoc
|| (asoc
->base
.sk
!= sk
) || asoc
->base
.dead
)
229 /* Look up the transport from an address and an assoc id. If both address and
230 * id are specified, the associations matching the address and the id should be
233 static struct sctp_transport
*sctp_addr_id2transport(struct sock
*sk
,
234 struct sockaddr_storage
*addr
,
237 struct sctp_association
*addr_asoc
= NULL
, *id_asoc
= NULL
;
238 struct sctp_transport
*transport
;
239 union sctp_addr
*laddr
= (union sctp_addr
*)addr
;
241 addr_asoc
= sctp_endpoint_lookup_assoc(sctp_sk(sk
)->ep
,
248 id_asoc
= sctp_id2assoc(sk
, id
);
249 if (id_asoc
&& (id_asoc
!= addr_asoc
))
252 sctp_get_pf_specific(sk
->sk_family
)->addr_to_user(sctp_sk(sk
),
253 (union sctp_addr
*)addr
);
258 /* API 3.1.2 bind() - UDP Style Syntax
259 * The syntax of bind() is,
261 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
263 * sd - the socket descriptor returned by socket().
264 * addr - the address structure (struct sockaddr_in or struct
265 * sockaddr_in6 [RFC 2553]),
266 * addr_len - the size of the address structure.
268 static int sctp_bind(struct sock
*sk
, struct sockaddr
*addr
, int addr_len
)
274 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__
, sk
,
277 /* Disallow binding twice. */
278 if (!sctp_sk(sk
)->ep
->base
.bind_addr
.port
)
279 retval
= sctp_do_bind(sk
, (union sctp_addr
*)addr
,
289 static long sctp_get_port_local(struct sock
*, union sctp_addr
*);
291 /* Verify this is a valid sockaddr. */
292 static struct sctp_af
*sctp_sockaddr_af(struct sctp_sock
*opt
,
293 union sctp_addr
*addr
, int len
)
297 /* Check minimum size. */
298 if (len
< sizeof (struct sockaddr
))
301 /* V4 mapped address are really of AF_INET family */
302 if (addr
->sa
.sa_family
== AF_INET6
&&
303 ipv6_addr_v4mapped(&addr
->v6
.sin6_addr
)) {
304 if (!opt
->pf
->af_supported(AF_INET
, opt
))
307 /* Does this PF support this AF? */
308 if (!opt
->pf
->af_supported(addr
->sa
.sa_family
, opt
))
312 /* If we get this far, af is valid. */
313 af
= sctp_get_af_specific(addr
->sa
.sa_family
);
315 if (len
< af
->sockaddr_len
)
321 /* Bind a local address either to an endpoint or to an association. */
322 static int sctp_do_bind(struct sock
*sk
, union sctp_addr
*addr
, int len
)
324 struct net
*net
= sock_net(sk
);
325 struct sctp_sock
*sp
= sctp_sk(sk
);
326 struct sctp_endpoint
*ep
= sp
->ep
;
327 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
332 /* Common sockaddr verification. */
333 af
= sctp_sockaddr_af(sp
, addr
, len
);
335 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
336 __func__
, sk
, addr
, len
);
340 snum
= ntohs(addr
->v4
.sin_port
);
342 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
343 __func__
, sk
, &addr
->sa
, bp
->port
, snum
, len
);
345 /* PF specific bind() address verification. */
346 if (!sp
->pf
->bind_verify(sp
, addr
))
347 return -EADDRNOTAVAIL
;
349 /* We must either be unbound, or bind to the same port.
350 * It's OK to allow 0 ports if we are already bound.
351 * We'll just inhert an already bound port in this case
356 else if (snum
!= bp
->port
) {
357 pr_debug("%s: new port %d doesn't match existing port "
358 "%d\n", __func__
, snum
, bp
->port
);
363 if (snum
&& snum
< inet_prot_sock(net
) &&
364 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
))
367 /* See if the address matches any of the addresses we may have
368 * already bound before checking against other endpoints.
370 if (sctp_bind_addr_match(bp
, addr
, sp
))
373 /* Make sure we are allowed to bind here.
374 * The function sctp_get_port_local() does duplicate address
377 addr
->v4
.sin_port
= htons(snum
);
378 if ((ret
= sctp_get_port_local(sk
, addr
))) {
382 /* Refresh ephemeral port. */
384 bp
->port
= inet_sk(sk
)->inet_num
;
386 /* Add the address to the bind address list.
387 * Use GFP_ATOMIC since BHs will be disabled.
389 ret
= sctp_add_bind_addr(bp
, addr
, af
->sockaddr_len
,
390 SCTP_ADDR_SRC
, GFP_ATOMIC
);
392 /* Copy back into socket for getsockname() use. */
394 inet_sk(sk
)->inet_sport
= htons(inet_sk(sk
)->inet_num
);
395 sp
->pf
->to_sk_saddr(addr
, sk
);
401 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
403 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
404 * at any one time. If a sender, after sending an ASCONF chunk, decides
405 * it needs to transfer another ASCONF Chunk, it MUST wait until the
406 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
407 * subsequent ASCONF. Note this restriction binds each side, so at any
408 * time two ASCONF may be in-transit on any given association (one sent
409 * from each endpoint).
411 static int sctp_send_asconf(struct sctp_association
*asoc
,
412 struct sctp_chunk
*chunk
)
414 struct net
*net
= sock_net(asoc
->base
.sk
);
417 /* If there is an outstanding ASCONF chunk, queue it for later
420 if (asoc
->addip_last_asconf
) {
421 list_add_tail(&chunk
->list
, &asoc
->addip_chunk_list
);
425 /* Hold the chunk until an ASCONF_ACK is received. */
426 sctp_chunk_hold(chunk
);
427 retval
= sctp_primitive_ASCONF(net
, asoc
, chunk
);
429 sctp_chunk_free(chunk
);
431 asoc
->addip_last_asconf
= chunk
;
437 /* Add a list of addresses as bind addresses to local endpoint or
440 * Basically run through each address specified in the addrs/addrcnt
441 * array/length pair, determine if it is IPv6 or IPv4 and call
442 * sctp_do_bind() on it.
444 * If any of them fails, then the operation will be reversed and the
445 * ones that were added will be removed.
447 * Only sctp_setsockopt_bindx() is supposed to call this function.
449 static int sctp_bindx_add(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
454 struct sockaddr
*sa_addr
;
457 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__
, sk
,
461 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
462 /* The list may contain either IPv4 or IPv6 address;
463 * determine the address length for walking thru the list.
466 af
= sctp_get_af_specific(sa_addr
->sa_family
);
472 retval
= sctp_do_bind(sk
, (union sctp_addr
*)sa_addr
,
475 addr_buf
+= af
->sockaddr_len
;
479 /* Failed. Cleanup the ones that have been added */
481 sctp_bindx_rem(sk
, addrs
, cnt
);
489 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
490 * associations that are part of the endpoint indicating that a list of local
491 * addresses are added to the endpoint.
493 * If any of the addresses is already in the bind address list of the
494 * association, we do not send the chunk for that association. But it will not
495 * affect other associations.
497 * Only sctp_setsockopt_bindx() is supposed to call this function.
499 static int sctp_send_asconf_add_ip(struct sock
*sk
,
500 struct sockaddr
*addrs
,
503 struct net
*net
= sock_net(sk
);
504 struct sctp_sock
*sp
;
505 struct sctp_endpoint
*ep
;
506 struct sctp_association
*asoc
;
507 struct sctp_bind_addr
*bp
;
508 struct sctp_chunk
*chunk
;
509 struct sctp_sockaddr_entry
*laddr
;
510 union sctp_addr
*addr
;
511 union sctp_addr saveaddr
;
518 if (!net
->sctp
.addip_enable
)
524 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
525 __func__
, sk
, addrs
, addrcnt
);
527 list_for_each_entry(asoc
, &ep
->asocs
, asocs
) {
528 if (!asoc
->peer
.asconf_capable
)
531 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_ADD_IP
)
534 if (!sctp_state(asoc
, ESTABLISHED
))
537 /* Check if any address in the packed array of addresses is
538 * in the bind address list of the association. If so,
539 * do not send the asconf chunk to its peer, but continue with
540 * other associations.
543 for (i
= 0; i
< addrcnt
; i
++) {
545 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
551 if (sctp_assoc_lookup_laddr(asoc
, addr
))
554 addr_buf
+= af
->sockaddr_len
;
559 /* Use the first valid address in bind addr list of
560 * association as Address Parameter of ASCONF CHUNK.
562 bp
= &asoc
->base
.bind_addr
;
563 p
= bp
->address_list
.next
;
564 laddr
= list_entry(p
, struct sctp_sockaddr_entry
, list
);
565 chunk
= sctp_make_asconf_update_ip(asoc
, &laddr
->a
, addrs
,
566 addrcnt
, SCTP_PARAM_ADD_IP
);
572 /* Add the new addresses to the bind address list with
573 * use_as_src set to 0.
576 for (i
= 0; i
< addrcnt
; i
++) {
578 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
579 memcpy(&saveaddr
, addr
, af
->sockaddr_len
);
580 retval
= sctp_add_bind_addr(bp
, &saveaddr
,
582 SCTP_ADDR_NEW
, GFP_ATOMIC
);
583 addr_buf
+= af
->sockaddr_len
;
585 if (asoc
->src_out_of_asoc_ok
) {
586 struct sctp_transport
*trans
;
588 list_for_each_entry(trans
,
589 &asoc
->peer
.transport_addr_list
, transports
) {
590 /* Clear the source and route cache */
591 dst_release(trans
->dst
);
592 trans
->cwnd
= min(4*asoc
->pathmtu
, max_t(__u32
,
593 2*asoc
->pathmtu
, 4380));
594 trans
->ssthresh
= asoc
->peer
.i
.a_rwnd
;
595 trans
->rto
= asoc
->rto_initial
;
596 sctp_max_rto(asoc
, trans
);
597 trans
->rtt
= trans
->srtt
= trans
->rttvar
= 0;
598 sctp_transport_route(trans
, NULL
,
599 sctp_sk(asoc
->base
.sk
));
602 retval
= sctp_send_asconf(asoc
, chunk
);
609 /* Remove a list of addresses from bind addresses list. Do not remove the
612 * Basically run through each address specified in the addrs/addrcnt
613 * array/length pair, determine if it is IPv6 or IPv4 and call
614 * sctp_del_bind() on it.
616 * If any of them fails, then the operation will be reversed and the
617 * ones that were removed will be added back.
619 * At least one address has to be left; if only one address is
620 * available, the operation will return -EBUSY.
622 * Only sctp_setsockopt_bindx() is supposed to call this function.
624 static int sctp_bindx_rem(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
626 struct sctp_sock
*sp
= sctp_sk(sk
);
627 struct sctp_endpoint
*ep
= sp
->ep
;
629 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
632 union sctp_addr
*sa_addr
;
635 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
636 __func__
, sk
, addrs
, addrcnt
);
639 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
640 /* If the bind address list is empty or if there is only one
641 * bind address, there is nothing more to be removed (we need
642 * at least one address here).
644 if (list_empty(&bp
->address_list
) ||
645 (sctp_list_single_entry(&bp
->address_list
))) {
651 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
657 if (!af
->addr_valid(sa_addr
, sp
, NULL
)) {
658 retval
= -EADDRNOTAVAIL
;
662 if (sa_addr
->v4
.sin_port
&&
663 sa_addr
->v4
.sin_port
!= htons(bp
->port
)) {
668 if (!sa_addr
->v4
.sin_port
)
669 sa_addr
->v4
.sin_port
= htons(bp
->port
);
671 /* FIXME - There is probably a need to check if sk->sk_saddr and
672 * sk->sk_rcv_addr are currently set to one of the addresses to
673 * be removed. This is something which needs to be looked into
674 * when we are fixing the outstanding issues with multi-homing
675 * socket routing and failover schemes. Refer to comments in
676 * sctp_do_bind(). -daisy
678 retval
= sctp_del_bind_addr(bp
, sa_addr
);
680 addr_buf
+= af
->sockaddr_len
;
683 /* Failed. Add the ones that has been removed back */
685 sctp_bindx_add(sk
, addrs
, cnt
);
693 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
694 * the associations that are part of the endpoint indicating that a list of
695 * local addresses are removed from the endpoint.
697 * If any of the addresses is already in the bind address list of the
698 * association, we do not send the chunk for that association. But it will not
699 * affect other associations.
701 * Only sctp_setsockopt_bindx() is supposed to call this function.
703 static int sctp_send_asconf_del_ip(struct sock
*sk
,
704 struct sockaddr
*addrs
,
707 struct net
*net
= sock_net(sk
);
708 struct sctp_sock
*sp
;
709 struct sctp_endpoint
*ep
;
710 struct sctp_association
*asoc
;
711 struct sctp_transport
*transport
;
712 struct sctp_bind_addr
*bp
;
713 struct sctp_chunk
*chunk
;
714 union sctp_addr
*laddr
;
717 struct sctp_sockaddr_entry
*saddr
;
723 if (!net
->sctp
.addip_enable
)
729 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
730 __func__
, sk
, addrs
, addrcnt
);
732 list_for_each_entry(asoc
, &ep
->asocs
, asocs
) {
734 if (!asoc
->peer
.asconf_capable
)
737 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_DEL_IP
)
740 if (!sctp_state(asoc
, ESTABLISHED
))
743 /* Check if any address in the packed array of addresses is
744 * not present in the bind address list of the association.
745 * If so, do not send the asconf chunk to its peer, but
746 * continue with other associations.
749 for (i
= 0; i
< addrcnt
; i
++) {
751 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
757 if (!sctp_assoc_lookup_laddr(asoc
, laddr
))
760 addr_buf
+= af
->sockaddr_len
;
765 /* Find one address in the association's bind address list
766 * that is not in the packed array of addresses. This is to
767 * make sure that we do not delete all the addresses in the
770 bp
= &asoc
->base
.bind_addr
;
771 laddr
= sctp_find_unmatch_addr(bp
, (union sctp_addr
*)addrs
,
773 if ((laddr
== NULL
) && (addrcnt
== 1)) {
774 if (asoc
->asconf_addr_del_pending
)
776 asoc
->asconf_addr_del_pending
=
777 kzalloc(sizeof(union sctp_addr
), GFP_ATOMIC
);
778 if (asoc
->asconf_addr_del_pending
== NULL
) {
782 asoc
->asconf_addr_del_pending
->sa
.sa_family
=
784 asoc
->asconf_addr_del_pending
->v4
.sin_port
=
786 if (addrs
->sa_family
== AF_INET
) {
787 struct sockaddr_in
*sin
;
789 sin
= (struct sockaddr_in
*)addrs
;
790 asoc
->asconf_addr_del_pending
->v4
.sin_addr
.s_addr
= sin
->sin_addr
.s_addr
;
791 } else if (addrs
->sa_family
== AF_INET6
) {
792 struct sockaddr_in6
*sin6
;
794 sin6
= (struct sockaddr_in6
*)addrs
;
795 asoc
->asconf_addr_del_pending
->v6
.sin6_addr
= sin6
->sin6_addr
;
798 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
799 __func__
, asoc
, &asoc
->asconf_addr_del_pending
->sa
,
800 asoc
->asconf_addr_del_pending
);
802 asoc
->src_out_of_asoc_ok
= 1;
810 /* We do not need RCU protection throughout this loop
811 * because this is done under a socket lock from the
814 chunk
= sctp_make_asconf_update_ip(asoc
, laddr
, addrs
, addrcnt
,
822 /* Reset use_as_src flag for the addresses in the bind address
823 * list that are to be deleted.
826 for (i
= 0; i
< addrcnt
; i
++) {
828 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
829 list_for_each_entry(saddr
, &bp
->address_list
, list
) {
830 if (sctp_cmp_addr_exact(&saddr
->a
, laddr
))
831 saddr
->state
= SCTP_ADDR_DEL
;
833 addr_buf
+= af
->sockaddr_len
;
836 /* Update the route and saddr entries for all the transports
837 * as some of the addresses in the bind address list are
838 * about to be deleted and cannot be used as source addresses.
840 list_for_each_entry(transport
, &asoc
->peer
.transport_addr_list
,
842 dst_release(transport
->dst
);
843 sctp_transport_route(transport
, NULL
,
844 sctp_sk(asoc
->base
.sk
));
848 /* We don't need to transmit ASCONF */
850 retval
= sctp_send_asconf(asoc
, chunk
);
856 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
857 int sctp_asconf_mgmt(struct sctp_sock
*sp
, struct sctp_sockaddr_entry
*addrw
)
859 struct sock
*sk
= sctp_opt2sk(sp
);
860 union sctp_addr
*addr
;
863 /* It is safe to write port space in caller. */
865 addr
->v4
.sin_port
= htons(sp
->ep
->base
.bind_addr
.port
);
866 af
= sctp_get_af_specific(addr
->sa
.sa_family
);
869 if (sctp_verify_addr(sk
, addr
, af
->sockaddr_len
))
872 if (addrw
->state
== SCTP_ADDR_NEW
)
873 return sctp_send_asconf_add_ip(sk
, (struct sockaddr
*)addr
, 1);
875 return sctp_send_asconf_del_ip(sk
, (struct sockaddr
*)addr
, 1);
878 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
881 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
884 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
885 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
888 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
889 * Section 3.1.2 for this usage.
891 * addrs is a pointer to an array of one or more socket addresses. Each
892 * address is contained in its appropriate structure (i.e. struct
893 * sockaddr_in or struct sockaddr_in6) the family of the address type
894 * must be used to distinguish the address length (note that this
895 * representation is termed a "packed array" of addresses). The caller
896 * specifies the number of addresses in the array with addrcnt.
898 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
899 * -1, and sets errno to the appropriate error code.
901 * For SCTP, the port given in each socket address must be the same, or
902 * sctp_bindx() will fail, setting errno to EINVAL.
904 * The flags parameter is formed from the bitwise OR of zero or more of
905 * the following currently defined flags:
907 * SCTP_BINDX_ADD_ADDR
909 * SCTP_BINDX_REM_ADDR
911 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
912 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
913 * addresses from the association. The two flags are mutually exclusive;
914 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
915 * not remove all addresses from an association; sctp_bindx() will
916 * reject such an attempt with EINVAL.
918 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
919 * additional addresses with an endpoint after calling bind(). Or use
920 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
921 * socket is associated with so that no new association accepted will be
922 * associated with those addresses. If the endpoint supports dynamic
923 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
924 * endpoint to send the appropriate message to the peer to change the
925 * peers address lists.
927 * Adding and removing addresses from a connected association is
928 * optional functionality. Implementations that do not support this
929 * functionality should return EOPNOTSUPP.
931 * Basically do nothing but copying the addresses from user to kernel
932 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
933 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
936 * We don't use copy_from_user() for optimization: we first do the
937 * sanity checks (buffer size -fast- and access check-healthy
938 * pointer); if all of those succeed, then we can alloc the memory
939 * (expensive operation) needed to copy the data to kernel. Then we do
940 * the copying without checking the user space area
941 * (__copy_from_user()).
943 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
946 * sk The sk of the socket
947 * addrs The pointer to the addresses in user land
948 * addrssize Size of the addrs buffer
949 * op Operation to perform (add or remove, see the flags of
952 * Returns 0 if ok, <0 errno code on error.
954 static int sctp_setsockopt_bindx(struct sock
*sk
,
955 struct sockaddr __user
*addrs
,
956 int addrs_size
, int op
)
958 struct sockaddr
*kaddrs
;
962 struct sockaddr
*sa_addr
;
966 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
967 __func__
, sk
, addrs
, addrs_size
, op
);
969 if (unlikely(addrs_size
<= 0))
972 /* Check the user passed a healthy pointer. */
973 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
976 /* Alloc space for the address array in kernel memory. */
977 kaddrs
= kmalloc(addrs_size
, GFP_USER
| __GFP_NOWARN
);
978 if (unlikely(!kaddrs
))
981 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
986 /* Walk through the addrs buffer and count the number of addresses. */
988 while (walk_size
< addrs_size
) {
989 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
995 af
= sctp_get_af_specific(sa_addr
->sa_family
);
997 /* If the address family is not supported or if this address
998 * causes the address buffer to overflow return EINVAL.
1000 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
1005 addr_buf
+= af
->sockaddr_len
;
1006 walk_size
+= af
->sockaddr_len
;
1011 case SCTP_BINDX_ADD_ADDR
:
1012 err
= sctp_bindx_add(sk
, kaddrs
, addrcnt
);
1015 err
= sctp_send_asconf_add_ip(sk
, kaddrs
, addrcnt
);
1018 case SCTP_BINDX_REM_ADDR
:
1019 err
= sctp_bindx_rem(sk
, kaddrs
, addrcnt
);
1022 err
= sctp_send_asconf_del_ip(sk
, kaddrs
, addrcnt
);
1036 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1038 * Common routine for handling connect() and sctp_connectx().
1039 * Connect will come in with just a single address.
1041 static int __sctp_connect(struct sock
*sk
,
1042 struct sockaddr
*kaddrs
,
1044 sctp_assoc_t
*assoc_id
)
1046 struct net
*net
= sock_net(sk
);
1047 struct sctp_sock
*sp
;
1048 struct sctp_endpoint
*ep
;
1049 struct sctp_association
*asoc
= NULL
;
1050 struct sctp_association
*asoc2
;
1051 struct sctp_transport
*transport
;
1058 union sctp_addr
*sa_addr
= NULL
;
1060 unsigned short port
;
1061 unsigned int f_flags
= 0;
1066 /* connect() cannot be done on a socket that is already in ESTABLISHED
1067 * state - UDP-style peeled off socket or a TCP-style socket that
1068 * is already connected.
1069 * It cannot be done even on a TCP-style listening socket.
1071 if (sctp_sstate(sk
, ESTABLISHED
) || sctp_sstate(sk
, CLOSING
) ||
1072 (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))) {
1077 /* Walk through the addrs buffer and count the number of addresses. */
1079 while (walk_size
< addrs_size
) {
1082 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
1088 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
1090 /* If the address family is not supported or if this address
1091 * causes the address buffer to overflow return EINVAL.
1093 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
1098 port
= ntohs(sa_addr
->v4
.sin_port
);
1100 /* Save current address so we can work with it */
1101 memcpy(&to
, sa_addr
, af
->sockaddr_len
);
1103 err
= sctp_verify_addr(sk
, &to
, af
->sockaddr_len
);
1107 /* Make sure the destination port is correctly set
1110 if (asoc
&& asoc
->peer
.port
&& asoc
->peer
.port
!= port
) {
1115 /* Check if there already is a matching association on the
1116 * endpoint (other than the one created here).
1118 asoc2
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1119 if (asoc2
&& asoc2
!= asoc
) {
1120 if (asoc2
->state
>= SCTP_STATE_ESTABLISHED
)
1127 /* If we could not find a matching association on the endpoint,
1128 * make sure that there is no peeled-off association matching
1129 * the peer address even on another socket.
1131 if (sctp_endpoint_is_peeled_off(ep
, &to
)) {
1132 err
= -EADDRNOTAVAIL
;
1137 /* If a bind() or sctp_bindx() is not called prior to
1138 * an sctp_connectx() call, the system picks an
1139 * ephemeral port and will choose an address set
1140 * equivalent to binding with a wildcard address.
1142 if (!ep
->base
.bind_addr
.port
) {
1143 if (sctp_autobind(sk
)) {
1149 * If an unprivileged user inherits a 1-many
1150 * style socket with open associations on a
1151 * privileged port, it MAY be permitted to
1152 * accept new associations, but it SHOULD NOT
1153 * be permitted to open new associations.
1155 if (ep
->base
.bind_addr
.port
<
1156 inet_prot_sock(net
) &&
1157 !ns_capable(net
->user_ns
,
1158 CAP_NET_BIND_SERVICE
)) {
1164 scope
= sctp_scope(&to
);
1165 asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1171 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
,
1179 /* Prime the peer's transport structures. */
1180 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
,
1188 addr_buf
+= af
->sockaddr_len
;
1189 walk_size
+= af
->sockaddr_len
;
1192 /* In case the user of sctp_connectx() wants an association
1193 * id back, assign one now.
1196 err
= sctp_assoc_set_id(asoc
, GFP_KERNEL
);
1201 err
= sctp_primitive_ASSOCIATE(net
, asoc
, NULL
);
1206 /* Initialize sk's dport and daddr for getpeername() */
1207 inet_sk(sk
)->inet_dport
= htons(asoc
->peer
.port
);
1208 sp
->pf
->to_sk_daddr(sa_addr
, sk
);
1211 /* in-kernel sockets don't generally have a file allocated to them
1212 * if all they do is call sock_create_kern().
1214 if (sk
->sk_socket
->file
)
1215 f_flags
= sk
->sk_socket
->file
->f_flags
;
1217 timeo
= sock_sndtimeo(sk
, f_flags
& O_NONBLOCK
);
1220 *assoc_id
= asoc
->assoc_id
;
1221 err
= sctp_wait_for_connect(asoc
, &timeo
);
1222 /* Note: the asoc may be freed after the return of
1223 * sctp_wait_for_connect.
1226 /* Don't free association on exit. */
1230 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1231 __func__
, asoc
, kaddrs
, err
);
1234 /* sctp_primitive_ASSOCIATE may have added this association
1235 * To the hash table, try to unhash it, just in case, its a noop
1236 * if it wasn't hashed so we're safe
1238 sctp_association_free(asoc
);
1243 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1246 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1247 * sctp_assoc_t *asoc);
1249 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1250 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1251 * or IPv6 addresses.
1253 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1254 * Section 3.1.2 for this usage.
1256 * addrs is a pointer to an array of one or more socket addresses. Each
1257 * address is contained in its appropriate structure (i.e. struct
1258 * sockaddr_in or struct sockaddr_in6) the family of the address type
1259 * must be used to distengish the address length (note that this
1260 * representation is termed a "packed array" of addresses). The caller
1261 * specifies the number of addresses in the array with addrcnt.
1263 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1264 * the association id of the new association. On failure, sctp_connectx()
1265 * returns -1, and sets errno to the appropriate error code. The assoc_id
1266 * is not touched by the kernel.
1268 * For SCTP, the port given in each socket address must be the same, or
1269 * sctp_connectx() will fail, setting errno to EINVAL.
1271 * An application can use sctp_connectx to initiate an association with
1272 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1273 * allows a caller to specify multiple addresses at which a peer can be
1274 * reached. The way the SCTP stack uses the list of addresses to set up
1275 * the association is implementation dependent. This function only
1276 * specifies that the stack will try to make use of all the addresses in
1277 * the list when needed.
1279 * Note that the list of addresses passed in is only used for setting up
1280 * the association. It does not necessarily equal the set of addresses
1281 * the peer uses for the resulting association. If the caller wants to
1282 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1283 * retrieve them after the association has been set up.
1285 * Basically do nothing but copying the addresses from user to kernel
1286 * land and invoking either sctp_connectx(). This is used for tunneling
1287 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1289 * We don't use copy_from_user() for optimization: we first do the
1290 * sanity checks (buffer size -fast- and access check-healthy
1291 * pointer); if all of those succeed, then we can alloc the memory
1292 * (expensive operation) needed to copy the data to kernel. Then we do
1293 * the copying without checking the user space area
1294 * (__copy_from_user()).
1296 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1299 * sk The sk of the socket
1300 * addrs The pointer to the addresses in user land
1301 * addrssize Size of the addrs buffer
1303 * Returns >=0 if ok, <0 errno code on error.
1305 static int __sctp_setsockopt_connectx(struct sock
*sk
,
1306 struct sockaddr __user
*addrs
,
1308 sctp_assoc_t
*assoc_id
)
1310 struct sockaddr
*kaddrs
;
1311 gfp_t gfp
= GFP_KERNEL
;
1314 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1315 __func__
, sk
, addrs
, addrs_size
);
1317 if (unlikely(addrs_size
<= 0))
1320 /* Check the user passed a healthy pointer. */
1321 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
1324 /* Alloc space for the address array in kernel memory. */
1325 if (sk
->sk_socket
->file
)
1326 gfp
= GFP_USER
| __GFP_NOWARN
;
1327 kaddrs
= kmalloc(addrs_size
, gfp
);
1328 if (unlikely(!kaddrs
))
1331 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
1334 err
= __sctp_connect(sk
, kaddrs
, addrs_size
, assoc_id
);
1343 * This is an older interface. It's kept for backward compatibility
1344 * to the option that doesn't provide association id.
1346 static int sctp_setsockopt_connectx_old(struct sock
*sk
,
1347 struct sockaddr __user
*addrs
,
1350 return __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, NULL
);
1354 * New interface for the API. The since the API is done with a socket
1355 * option, to make it simple we feed back the association id is as a return
1356 * indication to the call. Error is always negative and association id is
1359 static int sctp_setsockopt_connectx(struct sock
*sk
,
1360 struct sockaddr __user
*addrs
,
1363 sctp_assoc_t assoc_id
= 0;
1366 err
= __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, &assoc_id
);
1375 * New (hopefully final) interface for the API.
1376 * We use the sctp_getaddrs_old structure so that use-space library
1377 * can avoid any unnecessary allocations. The only different part
1378 * is that we store the actual length of the address buffer into the
1379 * addrs_num structure member. That way we can re-use the existing
1382 #ifdef CONFIG_COMPAT
1383 struct compat_sctp_getaddrs_old
{
1384 sctp_assoc_t assoc_id
;
1386 compat_uptr_t addrs
; /* struct sockaddr * */
1390 static int sctp_getsockopt_connectx3(struct sock
*sk
, int len
,
1391 char __user
*optval
,
1394 struct sctp_getaddrs_old param
;
1395 sctp_assoc_t assoc_id
= 0;
1398 #ifdef CONFIG_COMPAT
1399 if (in_compat_syscall()) {
1400 struct compat_sctp_getaddrs_old param32
;
1402 if (len
< sizeof(param32
))
1404 if (copy_from_user(¶m32
, optval
, sizeof(param32
)))
1407 param
.assoc_id
= param32
.assoc_id
;
1408 param
.addr_num
= param32
.addr_num
;
1409 param
.addrs
= compat_ptr(param32
.addrs
);
1413 if (len
< sizeof(param
))
1415 if (copy_from_user(¶m
, optval
, sizeof(param
)))
1419 err
= __sctp_setsockopt_connectx(sk
, (struct sockaddr __user
*)
1420 param
.addrs
, param
.addr_num
,
1422 if (err
== 0 || err
== -EINPROGRESS
) {
1423 if (copy_to_user(optval
, &assoc_id
, sizeof(assoc_id
)))
1425 if (put_user(sizeof(assoc_id
), optlen
))
1432 /* API 3.1.4 close() - UDP Style Syntax
1433 * Applications use close() to perform graceful shutdown (as described in
1434 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1435 * by a UDP-style socket.
1439 * ret = close(int sd);
1441 * sd - the socket descriptor of the associations to be closed.
1443 * To gracefully shutdown a specific association represented by the
1444 * UDP-style socket, an application should use the sendmsg() call,
1445 * passing no user data, but including the appropriate flag in the
1446 * ancillary data (see Section xxxx).
1448 * If sd in the close() call is a branched-off socket representing only
1449 * one association, the shutdown is performed on that association only.
1451 * 4.1.6 close() - TCP Style Syntax
1453 * Applications use close() to gracefully close down an association.
1457 * int close(int sd);
1459 * sd - the socket descriptor of the association to be closed.
1461 * After an application calls close() on a socket descriptor, no further
1462 * socket operations will succeed on that descriptor.
1464 * API 7.1.4 SO_LINGER
1466 * An application using the TCP-style socket can use this option to
1467 * perform the SCTP ABORT primitive. The linger option structure is:
1470 * int l_onoff; // option on/off
1471 * int l_linger; // linger time
1474 * To enable the option, set l_onoff to 1. If the l_linger value is set
1475 * to 0, calling close() is the same as the ABORT primitive. If the
1476 * value is set to a negative value, the setsockopt() call will return
1477 * an error. If the value is set to a positive value linger_time, the
1478 * close() can be blocked for at most linger_time ms. If the graceful
1479 * shutdown phase does not finish during this period, close() will
1480 * return but the graceful shutdown phase continues in the system.
1482 static void sctp_close(struct sock
*sk
, long timeout
)
1484 struct net
*net
= sock_net(sk
);
1485 struct sctp_endpoint
*ep
;
1486 struct sctp_association
*asoc
;
1487 struct list_head
*pos
, *temp
;
1488 unsigned int data_was_unread
;
1490 pr_debug("%s: sk:%p, timeout:%ld\n", __func__
, sk
, timeout
);
1493 sk
->sk_shutdown
= SHUTDOWN_MASK
;
1494 sk
->sk_state
= SCTP_SS_CLOSING
;
1496 ep
= sctp_sk(sk
)->ep
;
1498 /* Clean up any skbs sitting on the receive queue. */
1499 data_was_unread
= sctp_queue_purge_ulpevents(&sk
->sk_receive_queue
);
1500 data_was_unread
+= sctp_queue_purge_ulpevents(&sctp_sk(sk
)->pd_lobby
);
1502 /* Walk all associations on an endpoint. */
1503 list_for_each_safe(pos
, temp
, &ep
->asocs
) {
1504 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
1506 if (sctp_style(sk
, TCP
)) {
1507 /* A closed association can still be in the list if
1508 * it belongs to a TCP-style listening socket that is
1509 * not yet accepted. If so, free it. If not, send an
1510 * ABORT or SHUTDOWN based on the linger options.
1512 if (sctp_state(asoc
, CLOSED
)) {
1513 sctp_association_free(asoc
);
1518 if (data_was_unread
|| !skb_queue_empty(&asoc
->ulpq
.lobby
) ||
1519 !skb_queue_empty(&asoc
->ulpq
.reasm
) ||
1520 (sock_flag(sk
, SOCK_LINGER
) && !sk
->sk_lingertime
)) {
1521 struct sctp_chunk
*chunk
;
1523 chunk
= sctp_make_abort_user(asoc
, NULL
, 0);
1524 sctp_primitive_ABORT(net
, asoc
, chunk
);
1526 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
1529 /* On a TCP-style socket, block for at most linger_time if set. */
1530 if (sctp_style(sk
, TCP
) && timeout
)
1531 sctp_wait_for_close(sk
, timeout
);
1533 /* This will run the backlog queue. */
1536 /* Supposedly, no process has access to the socket, but
1537 * the net layers still may.
1538 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1539 * held and that should be grabbed before socket lock.
1541 spin_lock_bh(&net
->sctp
.addr_wq_lock
);
1544 /* Hold the sock, since sk_common_release() will put sock_put()
1545 * and we have just a little more cleanup.
1548 sk_common_release(sk
);
1551 spin_unlock_bh(&net
->sctp
.addr_wq_lock
);
1555 SCTP_DBG_OBJCNT_DEC(sock
);
1558 /* Handle EPIPE error. */
1559 static int sctp_error(struct sock
*sk
, int flags
, int err
)
1562 err
= sock_error(sk
) ? : -EPIPE
;
1563 if (err
== -EPIPE
&& !(flags
& MSG_NOSIGNAL
))
1564 send_sig(SIGPIPE
, current
, 0);
1568 /* API 3.1.3 sendmsg() - UDP Style Syntax
1570 * An application uses sendmsg() and recvmsg() calls to transmit data to
1571 * and receive data from its peer.
1573 * ssize_t sendmsg(int socket, const struct msghdr *message,
1576 * socket - the socket descriptor of the endpoint.
1577 * message - pointer to the msghdr structure which contains a single
1578 * user message and possibly some ancillary data.
1580 * See Section 5 for complete description of the data
1583 * flags - flags sent or received with the user message, see Section
1584 * 5 for complete description of the flags.
1586 * Note: This function could use a rewrite especially when explicit
1587 * connect support comes in.
1589 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1591 static int sctp_msghdr_parse(const struct msghdr
*, sctp_cmsgs_t
*);
1593 static int sctp_sendmsg(struct sock
*sk
, struct msghdr
*msg
, size_t msg_len
)
1595 struct net
*net
= sock_net(sk
);
1596 struct sctp_sock
*sp
;
1597 struct sctp_endpoint
*ep
;
1598 struct sctp_association
*new_asoc
= NULL
, *asoc
= NULL
;
1599 struct sctp_transport
*transport
, *chunk_tp
;
1600 struct sctp_chunk
*chunk
;
1602 struct sockaddr
*msg_name
= NULL
;
1603 struct sctp_sndrcvinfo default_sinfo
;
1604 struct sctp_sndrcvinfo
*sinfo
;
1605 struct sctp_initmsg
*sinit
;
1606 sctp_assoc_t associd
= 0;
1607 sctp_cmsgs_t cmsgs
= { NULL
};
1609 bool fill_sinfo_ttl
= false, wait_connect
= false;
1610 struct sctp_datamsg
*datamsg
;
1611 int msg_flags
= msg
->msg_flags
;
1612 __u16 sinfo_flags
= 0;
1620 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__
, sk
,
1623 /* We cannot send a message over a TCP-style listening socket. */
1624 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
)) {
1629 /* Parse out the SCTP CMSGs. */
1630 err
= sctp_msghdr_parse(msg
, &cmsgs
);
1632 pr_debug("%s: msghdr parse err:%x\n", __func__
, err
);
1636 /* Fetch the destination address for this packet. This
1637 * address only selects the association--it is not necessarily
1638 * the address we will send to.
1639 * For a peeled-off socket, msg_name is ignored.
1641 if (!sctp_style(sk
, UDP_HIGH_BANDWIDTH
) && msg
->msg_name
) {
1642 int msg_namelen
= msg
->msg_namelen
;
1644 err
= sctp_verify_addr(sk
, (union sctp_addr
*)msg
->msg_name
,
1649 if (msg_namelen
> sizeof(to
))
1650 msg_namelen
= sizeof(to
);
1651 memcpy(&to
, msg
->msg_name
, msg_namelen
);
1652 msg_name
= msg
->msg_name
;
1656 if (cmsgs
.sinfo
!= NULL
) {
1657 memset(&default_sinfo
, 0, sizeof(default_sinfo
));
1658 default_sinfo
.sinfo_stream
= cmsgs
.sinfo
->snd_sid
;
1659 default_sinfo
.sinfo_flags
= cmsgs
.sinfo
->snd_flags
;
1660 default_sinfo
.sinfo_ppid
= cmsgs
.sinfo
->snd_ppid
;
1661 default_sinfo
.sinfo_context
= cmsgs
.sinfo
->snd_context
;
1662 default_sinfo
.sinfo_assoc_id
= cmsgs
.sinfo
->snd_assoc_id
;
1664 sinfo
= &default_sinfo
;
1665 fill_sinfo_ttl
= true;
1667 sinfo
= cmsgs
.srinfo
;
1669 /* Did the user specify SNDINFO/SNDRCVINFO? */
1671 sinfo_flags
= sinfo
->sinfo_flags
;
1672 associd
= sinfo
->sinfo_assoc_id
;
1675 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__
,
1676 msg_len
, sinfo_flags
);
1678 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1679 if (sctp_style(sk
, TCP
) && (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
))) {
1684 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1685 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1686 * If SCTP_ABORT is set, the message length could be non zero with
1687 * the msg_iov set to the user abort reason.
1689 if (((sinfo_flags
& SCTP_EOF
) && (msg_len
> 0)) ||
1690 (!(sinfo_flags
& (SCTP_EOF
|SCTP_ABORT
)) && (msg_len
== 0))) {
1695 /* If SCTP_ADDR_OVER is set, there must be an address
1696 * specified in msg_name.
1698 if ((sinfo_flags
& SCTP_ADDR_OVER
) && (!msg
->msg_name
)) {
1705 pr_debug("%s: about to look up association\n", __func__
);
1709 /* If a msg_name has been specified, assume this is to be used. */
1711 /* Look for a matching association on the endpoint. */
1712 asoc
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1714 /* If we could not find a matching association on the
1715 * endpoint, make sure that it is not a TCP-style
1716 * socket that already has an association or there is
1717 * no peeled-off association on another socket.
1720 ((sctp_style(sk
, TCP
) &&
1721 (sctp_sstate(sk
, ESTABLISHED
) ||
1722 sctp_sstate(sk
, CLOSING
))) ||
1723 sctp_endpoint_is_peeled_off(ep
, &to
))) {
1724 err
= -EADDRNOTAVAIL
;
1728 asoc
= sctp_id2assoc(sk
, associd
);
1736 pr_debug("%s: just looked up association:%p\n", __func__
, asoc
);
1738 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1739 * socket that has an association in CLOSED state. This can
1740 * happen when an accepted socket has an association that is
1743 if (sctp_state(asoc
, CLOSED
) && sctp_style(sk
, TCP
)) {
1748 if (sinfo_flags
& SCTP_EOF
) {
1749 pr_debug("%s: shutting down association:%p\n",
1752 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
1756 if (sinfo_flags
& SCTP_ABORT
) {
1758 chunk
= sctp_make_abort_user(asoc
, msg
, msg_len
);
1764 pr_debug("%s: aborting association:%p\n",
1767 sctp_primitive_ABORT(net
, asoc
, chunk
);
1773 /* Do we need to create the association? */
1775 pr_debug("%s: there is no association yet\n", __func__
);
1777 if (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
)) {
1782 /* Check for invalid stream against the stream counts,
1783 * either the default or the user specified stream counts.
1786 if (!sinit
|| !sinit
->sinit_num_ostreams
) {
1787 /* Check against the defaults. */
1788 if (sinfo
->sinfo_stream
>=
1789 sp
->initmsg
.sinit_num_ostreams
) {
1794 /* Check against the requested. */
1795 if (sinfo
->sinfo_stream
>=
1796 sinit
->sinit_num_ostreams
) {
1804 * API 3.1.2 bind() - UDP Style Syntax
1805 * If a bind() or sctp_bindx() is not called prior to a
1806 * sendmsg() call that initiates a new association, the
1807 * system picks an ephemeral port and will choose an address
1808 * set equivalent to binding with a wildcard address.
1810 if (!ep
->base
.bind_addr
.port
) {
1811 if (sctp_autobind(sk
)) {
1817 * If an unprivileged user inherits a one-to-many
1818 * style socket with open associations on a privileged
1819 * port, it MAY be permitted to accept new associations,
1820 * but it SHOULD NOT be permitted to open new
1823 if (ep
->base
.bind_addr
.port
< inet_prot_sock(net
) &&
1824 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
)) {
1830 scope
= sctp_scope(&to
);
1831 new_asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1837 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
, GFP_KERNEL
);
1843 /* If the SCTP_INIT ancillary data is specified, set all
1844 * the association init values accordingly.
1847 if (sinit
->sinit_num_ostreams
) {
1848 asoc
->c
.sinit_num_ostreams
=
1849 sinit
->sinit_num_ostreams
;
1851 if (sinit
->sinit_max_instreams
) {
1852 asoc
->c
.sinit_max_instreams
=
1853 sinit
->sinit_max_instreams
;
1855 if (sinit
->sinit_max_attempts
) {
1856 asoc
->max_init_attempts
1857 = sinit
->sinit_max_attempts
;
1859 if (sinit
->sinit_max_init_timeo
) {
1860 asoc
->max_init_timeo
=
1861 msecs_to_jiffies(sinit
->sinit_max_init_timeo
);
1865 /* Prime the peer's transport structures. */
1866 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
, SCTP_UNKNOWN
);
1873 /* ASSERT: we have a valid association at this point. */
1874 pr_debug("%s: we have a valid association\n", __func__
);
1877 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1878 * one with some defaults.
1880 memset(&default_sinfo
, 0, sizeof(default_sinfo
));
1881 default_sinfo
.sinfo_stream
= asoc
->default_stream
;
1882 default_sinfo
.sinfo_flags
= asoc
->default_flags
;
1883 default_sinfo
.sinfo_ppid
= asoc
->default_ppid
;
1884 default_sinfo
.sinfo_context
= asoc
->default_context
;
1885 default_sinfo
.sinfo_timetolive
= asoc
->default_timetolive
;
1886 default_sinfo
.sinfo_assoc_id
= sctp_assoc2id(asoc
);
1888 sinfo
= &default_sinfo
;
1889 } else if (fill_sinfo_ttl
) {
1890 /* In case SNDINFO was specified, we still need to fill
1891 * it with a default ttl from the assoc here.
1893 sinfo
->sinfo_timetolive
= asoc
->default_timetolive
;
1896 /* API 7.1.7, the sndbuf size per association bounds the
1897 * maximum size of data that can be sent in a single send call.
1899 if (msg_len
> sk
->sk_sndbuf
) {
1904 if (asoc
->pmtu_pending
)
1905 sctp_assoc_pending_pmtu(sk
, asoc
);
1907 /* If fragmentation is disabled and the message length exceeds the
1908 * association fragmentation point, return EMSGSIZE. The I-D
1909 * does not specify what this error is, but this looks like
1912 if (sctp_sk(sk
)->disable_fragments
&& (msg_len
> asoc
->frag_point
)) {
1917 /* Check for invalid stream. */
1918 if (sinfo
->sinfo_stream
>= asoc
->c
.sinit_num_ostreams
) {
1923 if (sctp_wspace(asoc
) < msg_len
)
1924 sctp_prsctp_prune(asoc
, sinfo
, msg_len
- sctp_wspace(asoc
));
1926 timeo
= sock_sndtimeo(sk
, msg
->msg_flags
& MSG_DONTWAIT
);
1927 if (!sctp_wspace(asoc
)) {
1928 err
= sctp_wait_for_sndbuf(asoc
, &timeo
, msg_len
);
1933 /* If an address is passed with the sendto/sendmsg call, it is used
1934 * to override the primary destination address in the TCP model, or
1935 * when SCTP_ADDR_OVER flag is set in the UDP model.
1937 if ((sctp_style(sk
, TCP
) && msg_name
) ||
1938 (sinfo_flags
& SCTP_ADDR_OVER
)) {
1939 chunk_tp
= sctp_assoc_lookup_paddr(asoc
, &to
);
1947 /* Auto-connect, if we aren't connected already. */
1948 if (sctp_state(asoc
, CLOSED
)) {
1949 err
= sctp_primitive_ASSOCIATE(net
, asoc
, NULL
);
1953 wait_connect
= true;
1954 pr_debug("%s: we associated primitively\n", __func__
);
1957 /* Break the message into multiple chunks of maximum size. */
1958 datamsg
= sctp_datamsg_from_user(asoc
, sinfo
, &msg
->msg_iter
);
1959 if (IS_ERR(datamsg
)) {
1960 err
= PTR_ERR(datamsg
);
1964 /* Now send the (possibly) fragmented message. */
1965 list_for_each_entry(chunk
, &datamsg
->chunks
, frag_list
) {
1966 sctp_chunk_hold(chunk
);
1968 /* Do accounting for the write space. */
1969 sctp_set_owner_w(chunk
);
1971 chunk
->transport
= chunk_tp
;
1974 /* Send it to the lower layers. Note: all chunks
1975 * must either fail or succeed. The lower layer
1976 * works that way today. Keep it that way or this
1979 err
= sctp_primitive_SEND(net
, asoc
, datamsg
);
1980 /* Did the lower layer accept the chunk? */
1982 sctp_datamsg_free(datamsg
);
1986 pr_debug("%s: we sent primitively\n", __func__
);
1988 sctp_datamsg_put(datamsg
);
1991 if (unlikely(wait_connect
)) {
1992 timeo
= sock_sndtimeo(sk
, msg_flags
& MSG_DONTWAIT
);
1993 sctp_wait_for_connect(asoc
, &timeo
);
1996 /* If we are already past ASSOCIATE, the lower
1997 * layers are responsible for association cleanup.
2003 sctp_association_free(asoc
);
2008 return sctp_error(sk
, msg_flags
, err
);
2015 err
= sock_error(sk
);
2025 /* This is an extended version of skb_pull() that removes the data from the
2026 * start of a skb even when data is spread across the list of skb's in the
2027 * frag_list. len specifies the total amount of data that needs to be removed.
2028 * when 'len' bytes could be removed from the skb, it returns 0.
2029 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2030 * could not be removed.
2032 static int sctp_skb_pull(struct sk_buff
*skb
, int len
)
2034 struct sk_buff
*list
;
2035 int skb_len
= skb_headlen(skb
);
2038 if (len
<= skb_len
) {
2039 __skb_pull(skb
, len
);
2043 __skb_pull(skb
, skb_len
);
2045 skb_walk_frags(skb
, list
) {
2046 rlen
= sctp_skb_pull(list
, len
);
2047 skb
->len
-= (len
-rlen
);
2048 skb
->data_len
-= (len
-rlen
);
2059 /* API 3.1.3 recvmsg() - UDP Style Syntax
2061 * ssize_t recvmsg(int socket, struct msghdr *message,
2064 * socket - the socket descriptor of the endpoint.
2065 * message - pointer to the msghdr structure which contains a single
2066 * user message and possibly some ancillary data.
2068 * See Section 5 for complete description of the data
2071 * flags - flags sent or received with the user message, see Section
2072 * 5 for complete description of the flags.
2074 static int sctp_recvmsg(struct sock
*sk
, struct msghdr
*msg
, size_t len
,
2075 int noblock
, int flags
, int *addr_len
)
2077 struct sctp_ulpevent
*event
= NULL
;
2078 struct sctp_sock
*sp
= sctp_sk(sk
);
2079 struct sk_buff
*skb
, *head_skb
;
2084 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2085 "addr_len:%p)\n", __func__
, sk
, msg
, len
, noblock
, flags
,
2090 if (sctp_style(sk
, TCP
) && !sctp_sstate(sk
, ESTABLISHED
) &&
2091 !sctp_sstate(sk
, CLOSING
) && !sctp_sstate(sk
, CLOSED
)) {
2096 skb
= sctp_skb_recv_datagram(sk
, flags
, noblock
, &err
);
2100 /* Get the total length of the skb including any skb's in the
2109 err
= skb_copy_datagram_msg(skb
, 0, msg
, copied
);
2111 event
= sctp_skb2event(skb
);
2116 if (event
->chunk
&& event
->chunk
->head_skb
)
2117 head_skb
= event
->chunk
->head_skb
;
2120 sock_recv_ts_and_drops(msg
, sk
, head_skb
);
2121 if (sctp_ulpevent_is_notification(event
)) {
2122 msg
->msg_flags
|= MSG_NOTIFICATION
;
2123 sp
->pf
->event_msgname(event
, msg
->msg_name
, addr_len
);
2125 sp
->pf
->skb_msgname(head_skb
, msg
->msg_name
, addr_len
);
2128 /* Check if we allow SCTP_NXTINFO. */
2129 if (sp
->recvnxtinfo
)
2130 sctp_ulpevent_read_nxtinfo(event
, msg
, sk
);
2131 /* Check if we allow SCTP_RCVINFO. */
2132 if (sp
->recvrcvinfo
)
2133 sctp_ulpevent_read_rcvinfo(event
, msg
);
2134 /* Check if we allow SCTP_SNDRCVINFO. */
2135 if (sp
->subscribe
.sctp_data_io_event
)
2136 sctp_ulpevent_read_sndrcvinfo(event
, msg
);
2140 /* If skb's length exceeds the user's buffer, update the skb and
2141 * push it back to the receive_queue so that the next call to
2142 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2144 if (skb_len
> copied
) {
2145 msg
->msg_flags
&= ~MSG_EOR
;
2146 if (flags
& MSG_PEEK
)
2148 sctp_skb_pull(skb
, copied
);
2149 skb_queue_head(&sk
->sk_receive_queue
, skb
);
2151 /* When only partial message is copied to the user, increase
2152 * rwnd by that amount. If all the data in the skb is read,
2153 * rwnd is updated when the event is freed.
2155 if (!sctp_ulpevent_is_notification(event
))
2156 sctp_assoc_rwnd_increase(event
->asoc
, copied
);
2158 } else if ((event
->msg_flags
& MSG_NOTIFICATION
) ||
2159 (event
->msg_flags
& MSG_EOR
))
2160 msg
->msg_flags
|= MSG_EOR
;
2162 msg
->msg_flags
&= ~MSG_EOR
;
2165 if (flags
& MSG_PEEK
) {
2166 /* Release the skb reference acquired after peeking the skb in
2167 * sctp_skb_recv_datagram().
2171 /* Free the event which includes releasing the reference to
2172 * the owner of the skb, freeing the skb and updating the
2175 sctp_ulpevent_free(event
);
2182 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2184 * This option is a on/off flag. If enabled no SCTP message
2185 * fragmentation will be performed. Instead if a message being sent
2186 * exceeds the current PMTU size, the message will NOT be sent and
2187 * instead a error will be indicated to the user.
2189 static int sctp_setsockopt_disable_fragments(struct sock
*sk
,
2190 char __user
*optval
,
2191 unsigned int optlen
)
2195 if (optlen
< sizeof(int))
2198 if (get_user(val
, (int __user
*)optval
))
2201 sctp_sk(sk
)->disable_fragments
= (val
== 0) ? 0 : 1;
2206 static int sctp_setsockopt_events(struct sock
*sk
, char __user
*optval
,
2207 unsigned int optlen
)
2209 struct sctp_association
*asoc
;
2210 struct sctp_ulpevent
*event
;
2212 if (optlen
> sizeof(struct sctp_event_subscribe
))
2214 if (copy_from_user(&sctp_sk(sk
)->subscribe
, optval
, optlen
))
2217 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2218 * if there is no data to be sent or retransmit, the stack will
2219 * immediately send up this notification.
2221 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT
,
2222 &sctp_sk(sk
)->subscribe
)) {
2223 asoc
= sctp_id2assoc(sk
, 0);
2225 if (asoc
&& sctp_outq_is_empty(&asoc
->outqueue
)) {
2226 event
= sctp_ulpevent_make_sender_dry_event(asoc
,
2231 sctp_ulpq_tail_event(&asoc
->ulpq
, event
);
2238 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2240 * This socket option is applicable to the UDP-style socket only. When
2241 * set it will cause associations that are idle for more than the
2242 * specified number of seconds to automatically close. An association
2243 * being idle is defined an association that has NOT sent or received
2244 * user data. The special value of '0' indicates that no automatic
2245 * close of any associations should be performed. The option expects an
2246 * integer defining the number of seconds of idle time before an
2247 * association is closed.
2249 static int sctp_setsockopt_autoclose(struct sock
*sk
, char __user
*optval
,
2250 unsigned int optlen
)
2252 struct sctp_sock
*sp
= sctp_sk(sk
);
2253 struct net
*net
= sock_net(sk
);
2255 /* Applicable to UDP-style socket only */
2256 if (sctp_style(sk
, TCP
))
2258 if (optlen
!= sizeof(int))
2260 if (copy_from_user(&sp
->autoclose
, optval
, optlen
))
2263 if (sp
->autoclose
> net
->sctp
.max_autoclose
)
2264 sp
->autoclose
= net
->sctp
.max_autoclose
;
2269 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2271 * Applications can enable or disable heartbeats for any peer address of
2272 * an association, modify an address's heartbeat interval, force a
2273 * heartbeat to be sent immediately, and adjust the address's maximum
2274 * number of retransmissions sent before an address is considered
2275 * unreachable. The following structure is used to access and modify an
2276 * address's parameters:
2278 * struct sctp_paddrparams {
2279 * sctp_assoc_t spp_assoc_id;
2280 * struct sockaddr_storage spp_address;
2281 * uint32_t spp_hbinterval;
2282 * uint16_t spp_pathmaxrxt;
2283 * uint32_t spp_pathmtu;
2284 * uint32_t spp_sackdelay;
2285 * uint32_t spp_flags;
2288 * spp_assoc_id - (one-to-many style socket) This is filled in the
2289 * application, and identifies the association for
2291 * spp_address - This specifies which address is of interest.
2292 * spp_hbinterval - This contains the value of the heartbeat interval,
2293 * in milliseconds. If a value of zero
2294 * is present in this field then no changes are to
2295 * be made to this parameter.
2296 * spp_pathmaxrxt - This contains the maximum number of
2297 * retransmissions before this address shall be
2298 * considered unreachable. If a value of zero
2299 * is present in this field then no changes are to
2300 * be made to this parameter.
2301 * spp_pathmtu - When Path MTU discovery is disabled the value
2302 * specified here will be the "fixed" path mtu.
2303 * Note that if the spp_address field is empty
2304 * then all associations on this address will
2305 * have this fixed path mtu set upon them.
2307 * spp_sackdelay - When delayed sack is enabled, this value specifies
2308 * the number of milliseconds that sacks will be delayed
2309 * for. This value will apply to all addresses of an
2310 * association if the spp_address field is empty. Note
2311 * also, that if delayed sack is enabled and this
2312 * value is set to 0, no change is made to the last
2313 * recorded delayed sack timer value.
2315 * spp_flags - These flags are used to control various features
2316 * on an association. The flag field may contain
2317 * zero or more of the following options.
2319 * SPP_HB_ENABLE - Enable heartbeats on the
2320 * specified address. Note that if the address
2321 * field is empty all addresses for the association
2322 * have heartbeats enabled upon them.
2324 * SPP_HB_DISABLE - Disable heartbeats on the
2325 * speicifed address. Note that if the address
2326 * field is empty all addresses for the association
2327 * will have their heartbeats disabled. Note also
2328 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2329 * mutually exclusive, only one of these two should
2330 * be specified. Enabling both fields will have
2331 * undetermined results.
2333 * SPP_HB_DEMAND - Request a user initiated heartbeat
2334 * to be made immediately.
2336 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2337 * heartbeat delayis to be set to the value of 0
2340 * SPP_PMTUD_ENABLE - This field will enable PMTU
2341 * discovery upon the specified address. Note that
2342 * if the address feild is empty then all addresses
2343 * on the association are effected.
2345 * SPP_PMTUD_DISABLE - This field will disable PMTU
2346 * discovery upon the specified address. Note that
2347 * if the address feild is empty then all addresses
2348 * on the association are effected. Not also that
2349 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2350 * exclusive. Enabling both will have undetermined
2353 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2354 * on delayed sack. The time specified in spp_sackdelay
2355 * is used to specify the sack delay for this address. Note
2356 * that if spp_address is empty then all addresses will
2357 * enable delayed sack and take on the sack delay
2358 * value specified in spp_sackdelay.
2359 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2360 * off delayed sack. If the spp_address field is blank then
2361 * delayed sack is disabled for the entire association. Note
2362 * also that this field is mutually exclusive to
2363 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2366 static int sctp_apply_peer_addr_params(struct sctp_paddrparams
*params
,
2367 struct sctp_transport
*trans
,
2368 struct sctp_association
*asoc
,
2369 struct sctp_sock
*sp
,
2372 int sackdelay_change
)
2376 if (params
->spp_flags
& SPP_HB_DEMAND
&& trans
) {
2377 struct net
*net
= sock_net(trans
->asoc
->base
.sk
);
2379 error
= sctp_primitive_REQUESTHEARTBEAT(net
, trans
->asoc
, trans
);
2384 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2385 * this field is ignored. Note also that a value of zero indicates
2386 * the current setting should be left unchanged.
2388 if (params
->spp_flags
& SPP_HB_ENABLE
) {
2390 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2391 * set. This lets us use 0 value when this flag
2394 if (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)
2395 params
->spp_hbinterval
= 0;
2397 if (params
->spp_hbinterval
||
2398 (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)) {
2401 msecs_to_jiffies(params
->spp_hbinterval
);
2404 msecs_to_jiffies(params
->spp_hbinterval
);
2406 sp
->hbinterval
= params
->spp_hbinterval
;
2413 trans
->param_flags
=
2414 (trans
->param_flags
& ~SPP_HB
) | hb_change
;
2417 (asoc
->param_flags
& ~SPP_HB
) | hb_change
;
2420 (sp
->param_flags
& ~SPP_HB
) | hb_change
;
2424 /* When Path MTU discovery is disabled the value specified here will
2425 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2426 * include the flag SPP_PMTUD_DISABLE for this field to have any
2429 if ((params
->spp_flags
& SPP_PMTUD_DISABLE
) && params
->spp_pathmtu
) {
2431 trans
->pathmtu
= params
->spp_pathmtu
;
2432 sctp_assoc_sync_pmtu(sctp_opt2sk(sp
), asoc
);
2434 asoc
->pathmtu
= params
->spp_pathmtu
;
2436 sp
->pathmtu
= params
->spp_pathmtu
;
2442 int update
= (trans
->param_flags
& SPP_PMTUD_DISABLE
) &&
2443 (params
->spp_flags
& SPP_PMTUD_ENABLE
);
2444 trans
->param_flags
=
2445 (trans
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2447 sctp_transport_pmtu(trans
, sctp_opt2sk(sp
));
2448 sctp_assoc_sync_pmtu(sctp_opt2sk(sp
), asoc
);
2452 (asoc
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2455 (sp
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2459 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2460 * value of this field is ignored. Note also that a value of zero
2461 * indicates the current setting should be left unchanged.
2463 if ((params
->spp_flags
& SPP_SACKDELAY_ENABLE
) && params
->spp_sackdelay
) {
2466 msecs_to_jiffies(params
->spp_sackdelay
);
2469 msecs_to_jiffies(params
->spp_sackdelay
);
2471 sp
->sackdelay
= params
->spp_sackdelay
;
2475 if (sackdelay_change
) {
2477 trans
->param_flags
=
2478 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2482 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2486 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2491 /* Note that a value of zero indicates the current setting should be
2494 if (params
->spp_pathmaxrxt
) {
2496 trans
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2498 asoc
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2500 sp
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2507 static int sctp_setsockopt_peer_addr_params(struct sock
*sk
,
2508 char __user
*optval
,
2509 unsigned int optlen
)
2511 struct sctp_paddrparams params
;
2512 struct sctp_transport
*trans
= NULL
;
2513 struct sctp_association
*asoc
= NULL
;
2514 struct sctp_sock
*sp
= sctp_sk(sk
);
2516 int hb_change
, pmtud_change
, sackdelay_change
;
2518 if (optlen
!= sizeof(struct sctp_paddrparams
))
2521 if (copy_from_user(¶ms
, optval
, optlen
))
2524 /* Validate flags and value parameters. */
2525 hb_change
= params
.spp_flags
& SPP_HB
;
2526 pmtud_change
= params
.spp_flags
& SPP_PMTUD
;
2527 sackdelay_change
= params
.spp_flags
& SPP_SACKDELAY
;
2529 if (hb_change
== SPP_HB
||
2530 pmtud_change
== SPP_PMTUD
||
2531 sackdelay_change
== SPP_SACKDELAY
||
2532 params
.spp_sackdelay
> 500 ||
2533 (params
.spp_pathmtu
&&
2534 params
.spp_pathmtu
< SCTP_DEFAULT_MINSEGMENT
))
2537 /* If an address other than INADDR_ANY is specified, and
2538 * no transport is found, then the request is invalid.
2540 if (!sctp_is_any(sk
, (union sctp_addr
*)¶ms
.spp_address
)) {
2541 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
2542 params
.spp_assoc_id
);
2547 /* Get association, if assoc_id != 0 and the socket is a one
2548 * to many style socket, and an association was not found, then
2549 * the id was invalid.
2551 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
2552 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
))
2555 /* Heartbeat demand can only be sent on a transport or
2556 * association, but not a socket.
2558 if (params
.spp_flags
& SPP_HB_DEMAND
&& !trans
&& !asoc
)
2561 /* Process parameters. */
2562 error
= sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2563 hb_change
, pmtud_change
,
2569 /* If changes are for association, also apply parameters to each
2572 if (!trans
&& asoc
) {
2573 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2575 sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2576 hb_change
, pmtud_change
,
2584 static inline __u32
sctp_spp_sackdelay_enable(__u32 param_flags
)
2586 return (param_flags
& ~SPP_SACKDELAY
) | SPP_SACKDELAY_ENABLE
;
2589 static inline __u32
sctp_spp_sackdelay_disable(__u32 param_flags
)
2591 return (param_flags
& ~SPP_SACKDELAY
) | SPP_SACKDELAY_DISABLE
;
2595 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2597 * This option will effect the way delayed acks are performed. This
2598 * option allows you to get or set the delayed ack time, in
2599 * milliseconds. It also allows changing the delayed ack frequency.
2600 * Changing the frequency to 1 disables the delayed sack algorithm. If
2601 * the assoc_id is 0, then this sets or gets the endpoints default
2602 * values. If the assoc_id field is non-zero, then the set or get
2603 * effects the specified association for the one to many model (the
2604 * assoc_id field is ignored by the one to one model). Note that if
2605 * sack_delay or sack_freq are 0 when setting this option, then the
2606 * current values will remain unchanged.
2608 * struct sctp_sack_info {
2609 * sctp_assoc_t sack_assoc_id;
2610 * uint32_t sack_delay;
2611 * uint32_t sack_freq;
2614 * sack_assoc_id - This parameter, indicates which association the user
2615 * is performing an action upon. Note that if this field's value is
2616 * zero then the endpoints default value is changed (effecting future
2617 * associations only).
2619 * sack_delay - This parameter contains the number of milliseconds that
2620 * the user is requesting the delayed ACK timer be set to. Note that
2621 * this value is defined in the standard to be between 200 and 500
2624 * sack_freq - This parameter contains the number of packets that must
2625 * be received before a sack is sent without waiting for the delay
2626 * timer to expire. The default value for this is 2, setting this
2627 * value to 1 will disable the delayed sack algorithm.
2630 static int sctp_setsockopt_delayed_ack(struct sock
*sk
,
2631 char __user
*optval
, unsigned int optlen
)
2633 struct sctp_sack_info params
;
2634 struct sctp_transport
*trans
= NULL
;
2635 struct sctp_association
*asoc
= NULL
;
2636 struct sctp_sock
*sp
= sctp_sk(sk
);
2638 if (optlen
== sizeof(struct sctp_sack_info
)) {
2639 if (copy_from_user(¶ms
, optval
, optlen
))
2642 if (params
.sack_delay
== 0 && params
.sack_freq
== 0)
2644 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
2645 pr_warn_ratelimited(DEPRECATED
2647 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2648 "Use struct sctp_sack_info instead\n",
2649 current
->comm
, task_pid_nr(current
));
2650 if (copy_from_user(¶ms
, optval
, optlen
))
2653 if (params
.sack_delay
== 0)
2654 params
.sack_freq
= 1;
2656 params
.sack_freq
= 0;
2660 /* Validate value parameter. */
2661 if (params
.sack_delay
> 500)
2664 /* Get association, if sack_assoc_id != 0 and the socket is a one
2665 * to many style socket, and an association was not found, then
2666 * the id was invalid.
2668 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
2669 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
2672 if (params
.sack_delay
) {
2675 msecs_to_jiffies(params
.sack_delay
);
2677 sctp_spp_sackdelay_enable(asoc
->param_flags
);
2679 sp
->sackdelay
= params
.sack_delay
;
2681 sctp_spp_sackdelay_enable(sp
->param_flags
);
2685 if (params
.sack_freq
== 1) {
2688 sctp_spp_sackdelay_disable(asoc
->param_flags
);
2691 sctp_spp_sackdelay_disable(sp
->param_flags
);
2693 } else if (params
.sack_freq
> 1) {
2695 asoc
->sackfreq
= params
.sack_freq
;
2697 sctp_spp_sackdelay_enable(asoc
->param_flags
);
2699 sp
->sackfreq
= params
.sack_freq
;
2701 sctp_spp_sackdelay_enable(sp
->param_flags
);
2705 /* If change is for association, also apply to each transport. */
2707 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2709 if (params
.sack_delay
) {
2711 msecs_to_jiffies(params
.sack_delay
);
2712 trans
->param_flags
=
2713 sctp_spp_sackdelay_enable(trans
->param_flags
);
2715 if (params
.sack_freq
== 1) {
2716 trans
->param_flags
=
2717 sctp_spp_sackdelay_disable(trans
->param_flags
);
2718 } else if (params
.sack_freq
> 1) {
2719 trans
->sackfreq
= params
.sack_freq
;
2720 trans
->param_flags
=
2721 sctp_spp_sackdelay_enable(trans
->param_flags
);
2729 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2731 * Applications can specify protocol parameters for the default association
2732 * initialization. The option name argument to setsockopt() and getsockopt()
2735 * Setting initialization parameters is effective only on an unconnected
2736 * socket (for UDP-style sockets only future associations are effected
2737 * by the change). With TCP-style sockets, this option is inherited by
2738 * sockets derived from a listener socket.
2740 static int sctp_setsockopt_initmsg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2742 struct sctp_initmsg sinit
;
2743 struct sctp_sock
*sp
= sctp_sk(sk
);
2745 if (optlen
!= sizeof(struct sctp_initmsg
))
2747 if (copy_from_user(&sinit
, optval
, optlen
))
2750 if (sinit
.sinit_num_ostreams
)
2751 sp
->initmsg
.sinit_num_ostreams
= sinit
.sinit_num_ostreams
;
2752 if (sinit
.sinit_max_instreams
)
2753 sp
->initmsg
.sinit_max_instreams
= sinit
.sinit_max_instreams
;
2754 if (sinit
.sinit_max_attempts
)
2755 sp
->initmsg
.sinit_max_attempts
= sinit
.sinit_max_attempts
;
2756 if (sinit
.sinit_max_init_timeo
)
2757 sp
->initmsg
.sinit_max_init_timeo
= sinit
.sinit_max_init_timeo
;
2763 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2765 * Applications that wish to use the sendto() system call may wish to
2766 * specify a default set of parameters that would normally be supplied
2767 * through the inclusion of ancillary data. This socket option allows
2768 * such an application to set the default sctp_sndrcvinfo structure.
2769 * The application that wishes to use this socket option simply passes
2770 * in to this call the sctp_sndrcvinfo structure defined in Section
2771 * 5.2.2) The input parameters accepted by this call include
2772 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2773 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2774 * to this call if the caller is using the UDP model.
2776 static int sctp_setsockopt_default_send_param(struct sock
*sk
,
2777 char __user
*optval
,
2778 unsigned int optlen
)
2780 struct sctp_sock
*sp
= sctp_sk(sk
);
2781 struct sctp_association
*asoc
;
2782 struct sctp_sndrcvinfo info
;
2784 if (optlen
!= sizeof(info
))
2786 if (copy_from_user(&info
, optval
, optlen
))
2788 if (info
.sinfo_flags
&
2789 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
2790 SCTP_ABORT
| SCTP_EOF
))
2793 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
2794 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
2797 asoc
->default_stream
= info
.sinfo_stream
;
2798 asoc
->default_flags
= info
.sinfo_flags
;
2799 asoc
->default_ppid
= info
.sinfo_ppid
;
2800 asoc
->default_context
= info
.sinfo_context
;
2801 asoc
->default_timetolive
= info
.sinfo_timetolive
;
2803 sp
->default_stream
= info
.sinfo_stream
;
2804 sp
->default_flags
= info
.sinfo_flags
;
2805 sp
->default_ppid
= info
.sinfo_ppid
;
2806 sp
->default_context
= info
.sinfo_context
;
2807 sp
->default_timetolive
= info
.sinfo_timetolive
;
2813 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2814 * (SCTP_DEFAULT_SNDINFO)
2816 static int sctp_setsockopt_default_sndinfo(struct sock
*sk
,
2817 char __user
*optval
,
2818 unsigned int optlen
)
2820 struct sctp_sock
*sp
= sctp_sk(sk
);
2821 struct sctp_association
*asoc
;
2822 struct sctp_sndinfo info
;
2824 if (optlen
!= sizeof(info
))
2826 if (copy_from_user(&info
, optval
, optlen
))
2828 if (info
.snd_flags
&
2829 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
2830 SCTP_ABORT
| SCTP_EOF
))
2833 asoc
= sctp_id2assoc(sk
, info
.snd_assoc_id
);
2834 if (!asoc
&& info
.snd_assoc_id
&& sctp_style(sk
, UDP
))
2837 asoc
->default_stream
= info
.snd_sid
;
2838 asoc
->default_flags
= info
.snd_flags
;
2839 asoc
->default_ppid
= info
.snd_ppid
;
2840 asoc
->default_context
= info
.snd_context
;
2842 sp
->default_stream
= info
.snd_sid
;
2843 sp
->default_flags
= info
.snd_flags
;
2844 sp
->default_ppid
= info
.snd_ppid
;
2845 sp
->default_context
= info
.snd_context
;
2851 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2853 * Requests that the local SCTP stack use the enclosed peer address as
2854 * the association primary. The enclosed address must be one of the
2855 * association peer's addresses.
2857 static int sctp_setsockopt_primary_addr(struct sock
*sk
, char __user
*optval
,
2858 unsigned int optlen
)
2860 struct sctp_prim prim
;
2861 struct sctp_transport
*trans
;
2863 if (optlen
!= sizeof(struct sctp_prim
))
2866 if (copy_from_user(&prim
, optval
, sizeof(struct sctp_prim
)))
2869 trans
= sctp_addr_id2transport(sk
, &prim
.ssp_addr
, prim
.ssp_assoc_id
);
2873 sctp_assoc_set_primary(trans
->asoc
, trans
);
2879 * 7.1.5 SCTP_NODELAY
2881 * Turn on/off any Nagle-like algorithm. This means that packets are
2882 * generally sent as soon as possible and no unnecessary delays are
2883 * introduced, at the cost of more packets in the network. Expects an
2884 * integer boolean flag.
2886 static int sctp_setsockopt_nodelay(struct sock
*sk
, char __user
*optval
,
2887 unsigned int optlen
)
2891 if (optlen
< sizeof(int))
2893 if (get_user(val
, (int __user
*)optval
))
2896 sctp_sk(sk
)->nodelay
= (val
== 0) ? 0 : 1;
2902 * 7.1.1 SCTP_RTOINFO
2904 * The protocol parameters used to initialize and bound retransmission
2905 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2906 * and modify these parameters.
2907 * All parameters are time values, in milliseconds. A value of 0, when
2908 * modifying the parameters, indicates that the current value should not
2912 static int sctp_setsockopt_rtoinfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2914 struct sctp_rtoinfo rtoinfo
;
2915 struct sctp_association
*asoc
;
2916 unsigned long rto_min
, rto_max
;
2917 struct sctp_sock
*sp
= sctp_sk(sk
);
2919 if (optlen
!= sizeof (struct sctp_rtoinfo
))
2922 if (copy_from_user(&rtoinfo
, optval
, optlen
))
2925 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
2927 /* Set the values to the specific association */
2928 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
2931 rto_max
= rtoinfo
.srto_max
;
2932 rto_min
= rtoinfo
.srto_min
;
2935 rto_max
= asoc
? msecs_to_jiffies(rto_max
) : rto_max
;
2937 rto_max
= asoc
? asoc
->rto_max
: sp
->rtoinfo
.srto_max
;
2940 rto_min
= asoc
? msecs_to_jiffies(rto_min
) : rto_min
;
2942 rto_min
= asoc
? asoc
->rto_min
: sp
->rtoinfo
.srto_min
;
2944 if (rto_min
> rto_max
)
2948 if (rtoinfo
.srto_initial
!= 0)
2950 msecs_to_jiffies(rtoinfo
.srto_initial
);
2951 asoc
->rto_max
= rto_max
;
2952 asoc
->rto_min
= rto_min
;
2954 /* If there is no association or the association-id = 0
2955 * set the values to the endpoint.
2957 if (rtoinfo
.srto_initial
!= 0)
2958 sp
->rtoinfo
.srto_initial
= rtoinfo
.srto_initial
;
2959 sp
->rtoinfo
.srto_max
= rto_max
;
2960 sp
->rtoinfo
.srto_min
= rto_min
;
2968 * 7.1.2 SCTP_ASSOCINFO
2970 * This option is used to tune the maximum retransmission attempts
2971 * of the association.
2972 * Returns an error if the new association retransmission value is
2973 * greater than the sum of the retransmission value of the peer.
2974 * See [SCTP] for more information.
2977 static int sctp_setsockopt_associnfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2980 struct sctp_assocparams assocparams
;
2981 struct sctp_association
*asoc
;
2983 if (optlen
!= sizeof(struct sctp_assocparams
))
2985 if (copy_from_user(&assocparams
, optval
, optlen
))
2988 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
2990 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
2993 /* Set the values to the specific association */
2995 if (assocparams
.sasoc_asocmaxrxt
!= 0) {
2998 struct sctp_transport
*peer_addr
;
3000 list_for_each_entry(peer_addr
, &asoc
->peer
.transport_addr_list
,
3002 path_sum
+= peer_addr
->pathmaxrxt
;
3006 /* Only validate asocmaxrxt if we have more than
3007 * one path/transport. We do this because path
3008 * retransmissions are only counted when we have more
3012 assocparams
.sasoc_asocmaxrxt
> path_sum
)
3015 asoc
->max_retrans
= assocparams
.sasoc_asocmaxrxt
;
3018 if (assocparams
.sasoc_cookie_life
!= 0)
3019 asoc
->cookie_life
= ms_to_ktime(assocparams
.sasoc_cookie_life
);
3021 /* Set the values to the endpoint */
3022 struct sctp_sock
*sp
= sctp_sk(sk
);
3024 if (assocparams
.sasoc_asocmaxrxt
!= 0)
3025 sp
->assocparams
.sasoc_asocmaxrxt
=
3026 assocparams
.sasoc_asocmaxrxt
;
3027 if (assocparams
.sasoc_cookie_life
!= 0)
3028 sp
->assocparams
.sasoc_cookie_life
=
3029 assocparams
.sasoc_cookie_life
;
3035 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3037 * This socket option is a boolean flag which turns on or off mapped V4
3038 * addresses. If this option is turned on and the socket is type
3039 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3040 * If this option is turned off, then no mapping will be done of V4
3041 * addresses and a user will receive both PF_INET6 and PF_INET type
3042 * addresses on the socket.
3044 static int sctp_setsockopt_mappedv4(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
3047 struct sctp_sock
*sp
= sctp_sk(sk
);
3049 if (optlen
< sizeof(int))
3051 if (get_user(val
, (int __user
*)optval
))
3062 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3063 * This option will get or set the maximum size to put in any outgoing
3064 * SCTP DATA chunk. If a message is larger than this size it will be
3065 * fragmented by SCTP into the specified size. Note that the underlying
3066 * SCTP implementation may fragment into smaller sized chunks when the
3067 * PMTU of the underlying association is smaller than the value set by
3068 * the user. The default value for this option is '0' which indicates
3069 * the user is NOT limiting fragmentation and only the PMTU will effect
3070 * SCTP's choice of DATA chunk size. Note also that values set larger
3071 * than the maximum size of an IP datagram will effectively let SCTP
3072 * control fragmentation (i.e. the same as setting this option to 0).
3074 * The following structure is used to access and modify this parameter:
3076 * struct sctp_assoc_value {
3077 * sctp_assoc_t assoc_id;
3078 * uint32_t assoc_value;
3081 * assoc_id: This parameter is ignored for one-to-one style sockets.
3082 * For one-to-many style sockets this parameter indicates which
3083 * association the user is performing an action upon. Note that if
3084 * this field's value is zero then the endpoints default value is
3085 * changed (effecting future associations only).
3086 * assoc_value: This parameter specifies the maximum size in bytes.
3088 static int sctp_setsockopt_maxseg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
3090 struct sctp_assoc_value params
;
3091 struct sctp_association
*asoc
;
3092 struct sctp_sock
*sp
= sctp_sk(sk
);
3095 if (optlen
== sizeof(int)) {
3096 pr_warn_ratelimited(DEPRECATED
3098 "Use of int in maxseg socket option.\n"
3099 "Use struct sctp_assoc_value instead\n",
3100 current
->comm
, task_pid_nr(current
));
3101 if (copy_from_user(&val
, optval
, optlen
))
3103 params
.assoc_id
= 0;
3104 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
3105 if (copy_from_user(¶ms
, optval
, optlen
))
3107 val
= params
.assoc_value
;
3111 if ((val
!= 0) && ((val
< 8) || (val
> SCTP_MAX_CHUNK_LEN
)))
3114 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3115 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
3120 val
= asoc
->pathmtu
;
3121 val
-= sp
->pf
->af
->net_header_len
;
3122 val
-= sizeof(struct sctphdr
) +
3123 sizeof(struct sctp_data_chunk
);
3125 asoc
->user_frag
= val
;
3126 asoc
->frag_point
= sctp_frag_point(asoc
, asoc
->pathmtu
);
3128 sp
->user_frag
= val
;
3136 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3138 * Requests that the peer mark the enclosed address as the association
3139 * primary. The enclosed address must be one of the association's
3140 * locally bound addresses. The following structure is used to make a
3141 * set primary request:
3143 static int sctp_setsockopt_peer_primary_addr(struct sock
*sk
, char __user
*optval
,
3144 unsigned int optlen
)
3146 struct net
*net
= sock_net(sk
);
3147 struct sctp_sock
*sp
;
3148 struct sctp_association
*asoc
= NULL
;
3149 struct sctp_setpeerprim prim
;
3150 struct sctp_chunk
*chunk
;
3156 if (!net
->sctp
.addip_enable
)
3159 if (optlen
!= sizeof(struct sctp_setpeerprim
))
3162 if (copy_from_user(&prim
, optval
, optlen
))
3165 asoc
= sctp_id2assoc(sk
, prim
.sspp_assoc_id
);
3169 if (!asoc
->peer
.asconf_capable
)
3172 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_SET_PRIMARY
)
3175 if (!sctp_state(asoc
, ESTABLISHED
))
3178 af
= sctp_get_af_specific(prim
.sspp_addr
.ss_family
);
3182 if (!af
->addr_valid((union sctp_addr
*)&prim
.sspp_addr
, sp
, NULL
))
3183 return -EADDRNOTAVAIL
;
3185 if (!sctp_assoc_lookup_laddr(asoc
, (union sctp_addr
*)&prim
.sspp_addr
))
3186 return -EADDRNOTAVAIL
;
3188 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3189 chunk
= sctp_make_asconf_set_prim(asoc
,
3190 (union sctp_addr
*)&prim
.sspp_addr
);
3194 err
= sctp_send_asconf(asoc
, chunk
);
3196 pr_debug("%s: we set peer primary addr primitively\n", __func__
);
3201 static int sctp_setsockopt_adaptation_layer(struct sock
*sk
, char __user
*optval
,
3202 unsigned int optlen
)
3204 struct sctp_setadaptation adaptation
;
3206 if (optlen
!= sizeof(struct sctp_setadaptation
))
3208 if (copy_from_user(&adaptation
, optval
, optlen
))
3211 sctp_sk(sk
)->adaptation_ind
= adaptation
.ssb_adaptation_ind
;
3217 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3219 * The context field in the sctp_sndrcvinfo structure is normally only
3220 * used when a failed message is retrieved holding the value that was
3221 * sent down on the actual send call. This option allows the setting of
3222 * a default context on an association basis that will be received on
3223 * reading messages from the peer. This is especially helpful in the
3224 * one-2-many model for an application to keep some reference to an
3225 * internal state machine that is processing messages on the
3226 * association. Note that the setting of this value only effects
3227 * received messages from the peer and does not effect the value that is
3228 * saved with outbound messages.
3230 static int sctp_setsockopt_context(struct sock
*sk
, char __user
*optval
,
3231 unsigned int optlen
)
3233 struct sctp_assoc_value params
;
3234 struct sctp_sock
*sp
;
3235 struct sctp_association
*asoc
;
3237 if (optlen
!= sizeof(struct sctp_assoc_value
))
3239 if (copy_from_user(¶ms
, optval
, optlen
))
3244 if (params
.assoc_id
!= 0) {
3245 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3248 asoc
->default_rcv_context
= params
.assoc_value
;
3250 sp
->default_rcv_context
= params
.assoc_value
;
3257 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3259 * This options will at a minimum specify if the implementation is doing
3260 * fragmented interleave. Fragmented interleave, for a one to many
3261 * socket, is when subsequent calls to receive a message may return
3262 * parts of messages from different associations. Some implementations
3263 * may allow you to turn this value on or off. If so, when turned off,
3264 * no fragment interleave will occur (which will cause a head of line
3265 * blocking amongst multiple associations sharing the same one to many
3266 * socket). When this option is turned on, then each receive call may
3267 * come from a different association (thus the user must receive data
3268 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3269 * association each receive belongs to.
3271 * This option takes a boolean value. A non-zero value indicates that
3272 * fragmented interleave is on. A value of zero indicates that
3273 * fragmented interleave is off.
3275 * Note that it is important that an implementation that allows this
3276 * option to be turned on, have it off by default. Otherwise an unaware
3277 * application using the one to many model may become confused and act
3280 static int sctp_setsockopt_fragment_interleave(struct sock
*sk
,
3281 char __user
*optval
,
3282 unsigned int optlen
)
3286 if (optlen
!= sizeof(int))
3288 if (get_user(val
, (int __user
*)optval
))
3291 sctp_sk(sk
)->frag_interleave
= (val
== 0) ? 0 : 1;
3297 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3298 * (SCTP_PARTIAL_DELIVERY_POINT)
3300 * This option will set or get the SCTP partial delivery point. This
3301 * point is the size of a message where the partial delivery API will be
3302 * invoked to help free up rwnd space for the peer. Setting this to a
3303 * lower value will cause partial deliveries to happen more often. The
3304 * calls argument is an integer that sets or gets the partial delivery
3305 * point. Note also that the call will fail if the user attempts to set
3306 * this value larger than the socket receive buffer size.
3308 * Note that any single message having a length smaller than or equal to
3309 * the SCTP partial delivery point will be delivered in one single read
3310 * call as long as the user provided buffer is large enough to hold the
3313 static int sctp_setsockopt_partial_delivery_point(struct sock
*sk
,
3314 char __user
*optval
,
3315 unsigned int optlen
)
3319 if (optlen
!= sizeof(u32
))
3321 if (get_user(val
, (int __user
*)optval
))
3324 /* Note: We double the receive buffer from what the user sets
3325 * it to be, also initial rwnd is based on rcvbuf/2.
3327 if (val
> (sk
->sk_rcvbuf
>> 1))
3330 sctp_sk(sk
)->pd_point
= val
;
3332 return 0; /* is this the right error code? */
3336 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3338 * This option will allow a user to change the maximum burst of packets
3339 * that can be emitted by this association. Note that the default value
3340 * is 4, and some implementations may restrict this setting so that it
3341 * can only be lowered.
3343 * NOTE: This text doesn't seem right. Do this on a socket basis with
3344 * future associations inheriting the socket value.
3346 static int sctp_setsockopt_maxburst(struct sock
*sk
,
3347 char __user
*optval
,
3348 unsigned int optlen
)
3350 struct sctp_assoc_value params
;
3351 struct sctp_sock
*sp
;
3352 struct sctp_association
*asoc
;
3356 if (optlen
== sizeof(int)) {
3357 pr_warn_ratelimited(DEPRECATED
3359 "Use of int in max_burst socket option deprecated.\n"
3360 "Use struct sctp_assoc_value instead\n",
3361 current
->comm
, task_pid_nr(current
));
3362 if (copy_from_user(&val
, optval
, optlen
))
3364 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
3365 if (copy_from_user(¶ms
, optval
, optlen
))
3367 val
= params
.assoc_value
;
3368 assoc_id
= params
.assoc_id
;
3374 if (assoc_id
!= 0) {
3375 asoc
= sctp_id2assoc(sk
, assoc_id
);
3378 asoc
->max_burst
= val
;
3380 sp
->max_burst
= val
;
3386 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3388 * This set option adds a chunk type that the user is requesting to be
3389 * received only in an authenticated way. Changes to the list of chunks
3390 * will only effect future associations on the socket.
3392 static int sctp_setsockopt_auth_chunk(struct sock
*sk
,
3393 char __user
*optval
,
3394 unsigned int optlen
)
3396 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3397 struct sctp_authchunk val
;
3399 if (!ep
->auth_enable
)
3402 if (optlen
!= sizeof(struct sctp_authchunk
))
3404 if (copy_from_user(&val
, optval
, optlen
))
3407 switch (val
.sauth_chunk
) {
3409 case SCTP_CID_INIT_ACK
:
3410 case SCTP_CID_SHUTDOWN_COMPLETE
:
3415 /* add this chunk id to the endpoint */
3416 return sctp_auth_ep_add_chunkid(ep
, val
.sauth_chunk
);
3420 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3422 * This option gets or sets the list of HMAC algorithms that the local
3423 * endpoint requires the peer to use.
3425 static int sctp_setsockopt_hmac_ident(struct sock
*sk
,
3426 char __user
*optval
,
3427 unsigned int optlen
)
3429 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3430 struct sctp_hmacalgo
*hmacs
;
3434 if (!ep
->auth_enable
)
3437 if (optlen
< sizeof(struct sctp_hmacalgo
))
3440 hmacs
= memdup_user(optval
, optlen
);
3442 return PTR_ERR(hmacs
);
3444 idents
= hmacs
->shmac_num_idents
;
3445 if (idents
== 0 || idents
> SCTP_AUTH_NUM_HMACS
||
3446 (idents
* sizeof(u16
)) > (optlen
- sizeof(struct sctp_hmacalgo
))) {
3451 err
= sctp_auth_ep_set_hmacs(ep
, hmacs
);
3458 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3460 * This option will set a shared secret key which is used to build an
3461 * association shared key.
3463 static int sctp_setsockopt_auth_key(struct sock
*sk
,
3464 char __user
*optval
,
3465 unsigned int optlen
)
3467 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3468 struct sctp_authkey
*authkey
;
3469 struct sctp_association
*asoc
;
3472 if (!ep
->auth_enable
)
3475 if (optlen
<= sizeof(struct sctp_authkey
))
3478 authkey
= memdup_user(optval
, optlen
);
3479 if (IS_ERR(authkey
))
3480 return PTR_ERR(authkey
);
3482 if (authkey
->sca_keylength
> optlen
- sizeof(struct sctp_authkey
)) {
3487 asoc
= sctp_id2assoc(sk
, authkey
->sca_assoc_id
);
3488 if (!asoc
&& authkey
->sca_assoc_id
&& sctp_style(sk
, UDP
)) {
3493 ret
= sctp_auth_set_key(ep
, asoc
, authkey
);
3500 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3502 * This option will get or set the active shared key to be used to build
3503 * the association shared key.
3505 static int sctp_setsockopt_active_key(struct sock
*sk
,
3506 char __user
*optval
,
3507 unsigned int optlen
)
3509 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3510 struct sctp_authkeyid val
;
3511 struct sctp_association
*asoc
;
3513 if (!ep
->auth_enable
)
3516 if (optlen
!= sizeof(struct sctp_authkeyid
))
3518 if (copy_from_user(&val
, optval
, optlen
))
3521 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3522 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3525 return sctp_auth_set_active_key(ep
, asoc
, val
.scact_keynumber
);
3529 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3531 * This set option will delete a shared secret key from use.
3533 static int sctp_setsockopt_del_key(struct sock
*sk
,
3534 char __user
*optval
,
3535 unsigned int optlen
)
3537 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3538 struct sctp_authkeyid val
;
3539 struct sctp_association
*asoc
;
3541 if (!ep
->auth_enable
)
3544 if (optlen
!= sizeof(struct sctp_authkeyid
))
3546 if (copy_from_user(&val
, optval
, optlen
))
3549 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3550 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3553 return sctp_auth_del_key_id(ep
, asoc
, val
.scact_keynumber
);
3558 * 8.1.23 SCTP_AUTO_ASCONF
3560 * This option will enable or disable the use of the automatic generation of
3561 * ASCONF chunks to add and delete addresses to an existing association. Note
3562 * that this option has two caveats namely: a) it only affects sockets that
3563 * are bound to all addresses available to the SCTP stack, and b) the system
3564 * administrator may have an overriding control that turns the ASCONF feature
3565 * off no matter what setting the socket option may have.
3566 * This option expects an integer boolean flag, where a non-zero value turns on
3567 * the option, and a zero value turns off the option.
3568 * Note. In this implementation, socket operation overrides default parameter
3569 * being set by sysctl as well as FreeBSD implementation
3571 static int sctp_setsockopt_auto_asconf(struct sock
*sk
, char __user
*optval
,
3572 unsigned int optlen
)
3575 struct sctp_sock
*sp
= sctp_sk(sk
);
3577 if (optlen
< sizeof(int))
3579 if (get_user(val
, (int __user
*)optval
))
3581 if (!sctp_is_ep_boundall(sk
) && val
)
3583 if ((val
&& sp
->do_auto_asconf
) || (!val
&& !sp
->do_auto_asconf
))
3586 spin_lock_bh(&sock_net(sk
)->sctp
.addr_wq_lock
);
3587 if (val
== 0 && sp
->do_auto_asconf
) {
3588 list_del(&sp
->auto_asconf_list
);
3589 sp
->do_auto_asconf
= 0;
3590 } else if (val
&& !sp
->do_auto_asconf
) {
3591 list_add_tail(&sp
->auto_asconf_list
,
3592 &sock_net(sk
)->sctp
.auto_asconf_splist
);
3593 sp
->do_auto_asconf
= 1;
3595 spin_unlock_bh(&sock_net(sk
)->sctp
.addr_wq_lock
);
3600 * SCTP_PEER_ADDR_THLDS
3602 * This option allows us to alter the partially failed threshold for one or all
3603 * transports in an association. See Section 6.1 of:
3604 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3606 static int sctp_setsockopt_paddr_thresholds(struct sock
*sk
,
3607 char __user
*optval
,
3608 unsigned int optlen
)
3610 struct sctp_paddrthlds val
;
3611 struct sctp_transport
*trans
;
3612 struct sctp_association
*asoc
;
3614 if (optlen
< sizeof(struct sctp_paddrthlds
))
3616 if (copy_from_user(&val
, (struct sctp_paddrthlds __user
*)optval
,
3617 sizeof(struct sctp_paddrthlds
)))
3621 if (sctp_is_any(sk
, (const union sctp_addr
*)&val
.spt_address
)) {
3622 asoc
= sctp_id2assoc(sk
, val
.spt_assoc_id
);
3625 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
3627 if (val
.spt_pathmaxrxt
)
3628 trans
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3629 trans
->pf_retrans
= val
.spt_pathpfthld
;
3632 if (val
.spt_pathmaxrxt
)
3633 asoc
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3634 asoc
->pf_retrans
= val
.spt_pathpfthld
;
3636 trans
= sctp_addr_id2transport(sk
, &val
.spt_address
,
3641 if (val
.spt_pathmaxrxt
)
3642 trans
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3643 trans
->pf_retrans
= val
.spt_pathpfthld
;
3649 static int sctp_setsockopt_recvrcvinfo(struct sock
*sk
,
3650 char __user
*optval
,
3651 unsigned int optlen
)
3655 if (optlen
< sizeof(int))
3657 if (get_user(val
, (int __user
*) optval
))
3660 sctp_sk(sk
)->recvrcvinfo
= (val
== 0) ? 0 : 1;
3665 static int sctp_setsockopt_recvnxtinfo(struct sock
*sk
,
3666 char __user
*optval
,
3667 unsigned int optlen
)
3671 if (optlen
< sizeof(int))
3673 if (get_user(val
, (int __user
*) optval
))
3676 sctp_sk(sk
)->recvnxtinfo
= (val
== 0) ? 0 : 1;
3681 static int sctp_setsockopt_pr_supported(struct sock
*sk
,
3682 char __user
*optval
,
3683 unsigned int optlen
)
3685 struct sctp_assoc_value params
;
3686 struct sctp_association
*asoc
;
3687 int retval
= -EINVAL
;
3689 if (optlen
!= sizeof(params
))
3692 if (copy_from_user(¶ms
, optval
, optlen
)) {
3697 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3699 asoc
->prsctp_enable
= !!params
.assoc_value
;
3700 } else if (!params
.assoc_id
) {
3701 struct sctp_sock
*sp
= sctp_sk(sk
);
3703 sp
->ep
->prsctp_enable
= !!params
.assoc_value
;
3714 static int sctp_setsockopt_default_prinfo(struct sock
*sk
,
3715 char __user
*optval
,
3716 unsigned int optlen
)
3718 struct sctp_default_prinfo info
;
3719 struct sctp_association
*asoc
;
3720 int retval
= -EINVAL
;
3722 if (optlen
!= sizeof(info
))
3725 if (copy_from_user(&info
, optval
, sizeof(info
))) {
3730 if (info
.pr_policy
& ~SCTP_PR_SCTP_MASK
)
3733 if (info
.pr_policy
== SCTP_PR_SCTP_NONE
)
3736 asoc
= sctp_id2assoc(sk
, info
.pr_assoc_id
);
3738 SCTP_PR_SET_POLICY(asoc
->default_flags
, info
.pr_policy
);
3739 asoc
->default_timetolive
= info
.pr_value
;
3740 } else if (!info
.pr_assoc_id
) {
3741 struct sctp_sock
*sp
= sctp_sk(sk
);
3743 SCTP_PR_SET_POLICY(sp
->default_flags
, info
.pr_policy
);
3744 sp
->default_timetolive
= info
.pr_value
;
3755 static int sctp_setsockopt_enable_strreset(struct sock
*sk
,
3756 char __user
*optval
,
3757 unsigned int optlen
)
3759 struct sctp_assoc_value params
;
3760 struct sctp_association
*asoc
;
3761 int retval
= -EINVAL
;
3763 if (optlen
!= sizeof(params
))
3766 if (copy_from_user(¶ms
, optval
, optlen
)) {
3771 if (params
.assoc_value
& (~SCTP_ENABLE_STRRESET_MASK
))
3774 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3776 asoc
->strreset_enable
= params
.assoc_value
;
3777 } else if (!params
.assoc_id
) {
3778 struct sctp_sock
*sp
= sctp_sk(sk
);
3780 sp
->ep
->strreset_enable
= params
.assoc_value
;
3791 static int sctp_setsockopt_reset_streams(struct sock
*sk
,
3792 char __user
*optval
,
3793 unsigned int optlen
)
3795 struct sctp_reset_streams
*params
;
3796 struct sctp_association
*asoc
;
3797 int retval
= -EINVAL
;
3799 if (optlen
< sizeof(struct sctp_reset_streams
))
3802 params
= memdup_user(optval
, optlen
);
3804 return PTR_ERR(params
);
3806 asoc
= sctp_id2assoc(sk
, params
->srs_assoc_id
);
3810 retval
= sctp_send_reset_streams(asoc
, params
);
3817 /* API 6.2 setsockopt(), getsockopt()
3819 * Applications use setsockopt() and getsockopt() to set or retrieve
3820 * socket options. Socket options are used to change the default
3821 * behavior of sockets calls. They are described in Section 7.
3825 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3826 * int __user *optlen);
3827 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3830 * sd - the socket descript.
3831 * level - set to IPPROTO_SCTP for all SCTP options.
3832 * optname - the option name.
3833 * optval - the buffer to store the value of the option.
3834 * optlen - the size of the buffer.
3836 static int sctp_setsockopt(struct sock
*sk
, int level
, int optname
,
3837 char __user
*optval
, unsigned int optlen
)
3841 pr_debug("%s: sk:%p, optname:%d\n", __func__
, sk
, optname
);
3843 /* I can hardly begin to describe how wrong this is. This is
3844 * so broken as to be worse than useless. The API draft
3845 * REALLY is NOT helpful here... I am not convinced that the
3846 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3847 * are at all well-founded.
3849 if (level
!= SOL_SCTP
) {
3850 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
3851 retval
= af
->setsockopt(sk
, level
, optname
, optval
, optlen
);
3858 case SCTP_SOCKOPT_BINDX_ADD
:
3859 /* 'optlen' is the size of the addresses buffer. */
3860 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3861 optlen
, SCTP_BINDX_ADD_ADDR
);
3864 case SCTP_SOCKOPT_BINDX_REM
:
3865 /* 'optlen' is the size of the addresses buffer. */
3866 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3867 optlen
, SCTP_BINDX_REM_ADDR
);
3870 case SCTP_SOCKOPT_CONNECTX_OLD
:
3871 /* 'optlen' is the size of the addresses buffer. */
3872 retval
= sctp_setsockopt_connectx_old(sk
,
3873 (struct sockaddr __user
*)optval
,
3877 case SCTP_SOCKOPT_CONNECTX
:
3878 /* 'optlen' is the size of the addresses buffer. */
3879 retval
= sctp_setsockopt_connectx(sk
,
3880 (struct sockaddr __user
*)optval
,
3884 case SCTP_DISABLE_FRAGMENTS
:
3885 retval
= sctp_setsockopt_disable_fragments(sk
, optval
, optlen
);
3889 retval
= sctp_setsockopt_events(sk
, optval
, optlen
);
3892 case SCTP_AUTOCLOSE
:
3893 retval
= sctp_setsockopt_autoclose(sk
, optval
, optlen
);
3896 case SCTP_PEER_ADDR_PARAMS
:
3897 retval
= sctp_setsockopt_peer_addr_params(sk
, optval
, optlen
);
3900 case SCTP_DELAYED_SACK
:
3901 retval
= sctp_setsockopt_delayed_ack(sk
, optval
, optlen
);
3903 case SCTP_PARTIAL_DELIVERY_POINT
:
3904 retval
= sctp_setsockopt_partial_delivery_point(sk
, optval
, optlen
);
3908 retval
= sctp_setsockopt_initmsg(sk
, optval
, optlen
);
3910 case SCTP_DEFAULT_SEND_PARAM
:
3911 retval
= sctp_setsockopt_default_send_param(sk
, optval
,
3914 case SCTP_DEFAULT_SNDINFO
:
3915 retval
= sctp_setsockopt_default_sndinfo(sk
, optval
, optlen
);
3917 case SCTP_PRIMARY_ADDR
:
3918 retval
= sctp_setsockopt_primary_addr(sk
, optval
, optlen
);
3920 case SCTP_SET_PEER_PRIMARY_ADDR
:
3921 retval
= sctp_setsockopt_peer_primary_addr(sk
, optval
, optlen
);
3924 retval
= sctp_setsockopt_nodelay(sk
, optval
, optlen
);
3927 retval
= sctp_setsockopt_rtoinfo(sk
, optval
, optlen
);
3929 case SCTP_ASSOCINFO
:
3930 retval
= sctp_setsockopt_associnfo(sk
, optval
, optlen
);
3932 case SCTP_I_WANT_MAPPED_V4_ADDR
:
3933 retval
= sctp_setsockopt_mappedv4(sk
, optval
, optlen
);
3936 retval
= sctp_setsockopt_maxseg(sk
, optval
, optlen
);
3938 case SCTP_ADAPTATION_LAYER
:
3939 retval
= sctp_setsockopt_adaptation_layer(sk
, optval
, optlen
);
3942 retval
= sctp_setsockopt_context(sk
, optval
, optlen
);
3944 case SCTP_FRAGMENT_INTERLEAVE
:
3945 retval
= sctp_setsockopt_fragment_interleave(sk
, optval
, optlen
);
3947 case SCTP_MAX_BURST
:
3948 retval
= sctp_setsockopt_maxburst(sk
, optval
, optlen
);
3950 case SCTP_AUTH_CHUNK
:
3951 retval
= sctp_setsockopt_auth_chunk(sk
, optval
, optlen
);
3953 case SCTP_HMAC_IDENT
:
3954 retval
= sctp_setsockopt_hmac_ident(sk
, optval
, optlen
);
3957 retval
= sctp_setsockopt_auth_key(sk
, optval
, optlen
);
3959 case SCTP_AUTH_ACTIVE_KEY
:
3960 retval
= sctp_setsockopt_active_key(sk
, optval
, optlen
);
3962 case SCTP_AUTH_DELETE_KEY
:
3963 retval
= sctp_setsockopt_del_key(sk
, optval
, optlen
);
3965 case SCTP_AUTO_ASCONF
:
3966 retval
= sctp_setsockopt_auto_asconf(sk
, optval
, optlen
);
3968 case SCTP_PEER_ADDR_THLDS
:
3969 retval
= sctp_setsockopt_paddr_thresholds(sk
, optval
, optlen
);
3971 case SCTP_RECVRCVINFO
:
3972 retval
= sctp_setsockopt_recvrcvinfo(sk
, optval
, optlen
);
3974 case SCTP_RECVNXTINFO
:
3975 retval
= sctp_setsockopt_recvnxtinfo(sk
, optval
, optlen
);
3977 case SCTP_PR_SUPPORTED
:
3978 retval
= sctp_setsockopt_pr_supported(sk
, optval
, optlen
);
3980 case SCTP_DEFAULT_PRINFO
:
3981 retval
= sctp_setsockopt_default_prinfo(sk
, optval
, optlen
);
3983 case SCTP_ENABLE_STREAM_RESET
:
3984 retval
= sctp_setsockopt_enable_strreset(sk
, optval
, optlen
);
3986 case SCTP_RESET_STREAMS
:
3987 retval
= sctp_setsockopt_reset_streams(sk
, optval
, optlen
);
3990 retval
= -ENOPROTOOPT
;
4000 /* API 3.1.6 connect() - UDP Style Syntax
4002 * An application may use the connect() call in the UDP model to initiate an
4003 * association without sending data.
4007 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
4009 * sd: the socket descriptor to have a new association added to.
4011 * nam: the address structure (either struct sockaddr_in or struct
4012 * sockaddr_in6 defined in RFC2553 [7]).
4014 * len: the size of the address.
4016 static int sctp_connect(struct sock
*sk
, struct sockaddr
*addr
,
4024 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__
, sk
,
4027 /* Validate addr_len before calling common connect/connectx routine. */
4028 af
= sctp_get_af_specific(addr
->sa_family
);
4029 if (!af
|| addr_len
< af
->sockaddr_len
) {
4032 /* Pass correct addr len to common routine (so it knows there
4033 * is only one address being passed.
4035 err
= __sctp_connect(sk
, addr
, af
->sockaddr_len
, NULL
);
4042 /* FIXME: Write comments. */
4043 static int sctp_disconnect(struct sock
*sk
, int flags
)
4045 return -EOPNOTSUPP
; /* STUB */
4048 /* 4.1.4 accept() - TCP Style Syntax
4050 * Applications use accept() call to remove an established SCTP
4051 * association from the accept queue of the endpoint. A new socket
4052 * descriptor will be returned from accept() to represent the newly
4053 * formed association.
4055 static struct sock
*sctp_accept(struct sock
*sk
, int flags
, int *err
)
4057 struct sctp_sock
*sp
;
4058 struct sctp_endpoint
*ep
;
4059 struct sock
*newsk
= NULL
;
4060 struct sctp_association
*asoc
;
4069 if (!sctp_style(sk
, TCP
)) {
4070 error
= -EOPNOTSUPP
;
4074 if (!sctp_sstate(sk
, LISTENING
)) {
4079 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
4081 error
= sctp_wait_for_accept(sk
, timeo
);
4085 /* We treat the list of associations on the endpoint as the accept
4086 * queue and pick the first association on the list.
4088 asoc
= list_entry(ep
->asocs
.next
, struct sctp_association
, asocs
);
4090 newsk
= sp
->pf
->create_accept_sk(sk
, asoc
);
4096 /* Populate the fields of the newsk from the oldsk and migrate the
4097 * asoc to the newsk.
4099 sctp_sock_migrate(sk
, newsk
, asoc
, SCTP_SOCKET_TCP
);
4107 /* The SCTP ioctl handler. */
4108 static int sctp_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
4115 * SEQPACKET-style sockets in LISTENING state are valid, for
4116 * SCTP, so only discard TCP-style sockets in LISTENING state.
4118 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
4123 struct sk_buff
*skb
;
4124 unsigned int amount
= 0;
4126 skb
= skb_peek(&sk
->sk_receive_queue
);
4129 * We will only return the amount of this packet since
4130 * that is all that will be read.
4134 rc
= put_user(amount
, (int __user
*)arg
);
4146 /* This is the function which gets called during socket creation to
4147 * initialized the SCTP-specific portion of the sock.
4148 * The sock structure should already be zero-filled memory.
4150 static int sctp_init_sock(struct sock
*sk
)
4152 struct net
*net
= sock_net(sk
);
4153 struct sctp_sock
*sp
;
4155 pr_debug("%s: sk:%p\n", __func__
, sk
);
4159 /* Initialize the SCTP per socket area. */
4160 switch (sk
->sk_type
) {
4161 case SOCK_SEQPACKET
:
4162 sp
->type
= SCTP_SOCKET_UDP
;
4165 sp
->type
= SCTP_SOCKET_TCP
;
4168 return -ESOCKTNOSUPPORT
;
4171 sk
->sk_gso_type
= SKB_GSO_SCTP
;
4173 /* Initialize default send parameters. These parameters can be
4174 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4176 sp
->default_stream
= 0;
4177 sp
->default_ppid
= 0;
4178 sp
->default_flags
= 0;
4179 sp
->default_context
= 0;
4180 sp
->default_timetolive
= 0;
4182 sp
->default_rcv_context
= 0;
4183 sp
->max_burst
= net
->sctp
.max_burst
;
4185 sp
->sctp_hmac_alg
= net
->sctp
.sctp_hmac_alg
;
4187 /* Initialize default setup parameters. These parameters
4188 * can be modified with the SCTP_INITMSG socket option or
4189 * overridden by the SCTP_INIT CMSG.
4191 sp
->initmsg
.sinit_num_ostreams
= sctp_max_outstreams
;
4192 sp
->initmsg
.sinit_max_instreams
= sctp_max_instreams
;
4193 sp
->initmsg
.sinit_max_attempts
= net
->sctp
.max_retrans_init
;
4194 sp
->initmsg
.sinit_max_init_timeo
= net
->sctp
.rto_max
;
4196 /* Initialize default RTO related parameters. These parameters can
4197 * be modified for with the SCTP_RTOINFO socket option.
4199 sp
->rtoinfo
.srto_initial
= net
->sctp
.rto_initial
;
4200 sp
->rtoinfo
.srto_max
= net
->sctp
.rto_max
;
4201 sp
->rtoinfo
.srto_min
= net
->sctp
.rto_min
;
4203 /* Initialize default association related parameters. These parameters
4204 * can be modified with the SCTP_ASSOCINFO socket option.
4206 sp
->assocparams
.sasoc_asocmaxrxt
= net
->sctp
.max_retrans_association
;
4207 sp
->assocparams
.sasoc_number_peer_destinations
= 0;
4208 sp
->assocparams
.sasoc_peer_rwnd
= 0;
4209 sp
->assocparams
.sasoc_local_rwnd
= 0;
4210 sp
->assocparams
.sasoc_cookie_life
= net
->sctp
.valid_cookie_life
;
4212 /* Initialize default event subscriptions. By default, all the
4215 memset(&sp
->subscribe
, 0, sizeof(struct sctp_event_subscribe
));
4217 /* Default Peer Address Parameters. These defaults can
4218 * be modified via SCTP_PEER_ADDR_PARAMS
4220 sp
->hbinterval
= net
->sctp
.hb_interval
;
4221 sp
->pathmaxrxt
= net
->sctp
.max_retrans_path
;
4222 sp
->pathmtu
= 0; /* allow default discovery */
4223 sp
->sackdelay
= net
->sctp
.sack_timeout
;
4225 sp
->param_flags
= SPP_HB_ENABLE
|
4227 SPP_SACKDELAY_ENABLE
;
4229 /* If enabled no SCTP message fragmentation will be performed.
4230 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4232 sp
->disable_fragments
= 0;
4234 /* Enable Nagle algorithm by default. */
4237 sp
->recvrcvinfo
= 0;
4238 sp
->recvnxtinfo
= 0;
4240 /* Enable by default. */
4243 /* Auto-close idle associations after the configured
4244 * number of seconds. A value of 0 disables this
4245 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4246 * for UDP-style sockets only.
4250 /* User specified fragmentation limit. */
4253 sp
->adaptation_ind
= 0;
4255 sp
->pf
= sctp_get_pf_specific(sk
->sk_family
);
4257 /* Control variables for partial data delivery. */
4258 atomic_set(&sp
->pd_mode
, 0);
4259 skb_queue_head_init(&sp
->pd_lobby
);
4260 sp
->frag_interleave
= 0;
4262 /* Create a per socket endpoint structure. Even if we
4263 * change the data structure relationships, this may still
4264 * be useful for storing pre-connect address information.
4266 sp
->ep
= sctp_endpoint_new(sk
, GFP_KERNEL
);
4272 sk
->sk_destruct
= sctp_destruct_sock
;
4274 SCTP_DBG_OBJCNT_INC(sock
);
4277 percpu_counter_inc(&sctp_sockets_allocated
);
4278 sock_prot_inuse_add(net
, sk
->sk_prot
, 1);
4280 /* Nothing can fail after this block, otherwise
4281 * sctp_destroy_sock() will be called without addr_wq_lock held
4283 if (net
->sctp
.default_auto_asconf
) {
4284 spin_lock(&sock_net(sk
)->sctp
.addr_wq_lock
);
4285 list_add_tail(&sp
->auto_asconf_list
,
4286 &net
->sctp
.auto_asconf_splist
);
4287 sp
->do_auto_asconf
= 1;
4288 spin_unlock(&sock_net(sk
)->sctp
.addr_wq_lock
);
4290 sp
->do_auto_asconf
= 0;
4298 /* Cleanup any SCTP per socket resources. Must be called with
4299 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4301 static void sctp_destroy_sock(struct sock
*sk
)
4303 struct sctp_sock
*sp
;
4305 pr_debug("%s: sk:%p\n", __func__
, sk
);
4307 /* Release our hold on the endpoint. */
4309 /* This could happen during socket init, thus we bail out
4310 * early, since the rest of the below is not setup either.
4315 if (sp
->do_auto_asconf
) {
4316 sp
->do_auto_asconf
= 0;
4317 list_del(&sp
->auto_asconf_list
);
4319 sctp_endpoint_free(sp
->ep
);
4321 percpu_counter_dec(&sctp_sockets_allocated
);
4322 sock_prot_inuse_add(sock_net(sk
), sk
->sk_prot
, -1);
4326 /* Triggered when there are no references on the socket anymore */
4327 static void sctp_destruct_sock(struct sock
*sk
)
4329 struct sctp_sock
*sp
= sctp_sk(sk
);
4331 /* Free up the HMAC transform. */
4332 crypto_free_shash(sp
->hmac
);
4334 inet_sock_destruct(sk
);
4337 /* API 4.1.7 shutdown() - TCP Style Syntax
4338 * int shutdown(int socket, int how);
4340 * sd - the socket descriptor of the association to be closed.
4341 * how - Specifies the type of shutdown. The values are
4344 * Disables further receive operations. No SCTP
4345 * protocol action is taken.
4347 * Disables further send operations, and initiates
4348 * the SCTP shutdown sequence.
4350 * Disables further send and receive operations
4351 * and initiates the SCTP shutdown sequence.
4353 static void sctp_shutdown(struct sock
*sk
, int how
)
4355 struct net
*net
= sock_net(sk
);
4356 struct sctp_endpoint
*ep
;
4358 if (!sctp_style(sk
, TCP
))
4361 ep
= sctp_sk(sk
)->ep
;
4362 if (how
& SEND_SHUTDOWN
&& !list_empty(&ep
->asocs
)) {
4363 struct sctp_association
*asoc
;
4365 sk
->sk_state
= SCTP_SS_CLOSING
;
4366 asoc
= list_entry(ep
->asocs
.next
,
4367 struct sctp_association
, asocs
);
4368 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
4372 int sctp_get_sctp_info(struct sock
*sk
, struct sctp_association
*asoc
,
4373 struct sctp_info
*info
)
4375 struct sctp_transport
*prim
;
4376 struct list_head
*pos
;
4379 memset(info
, 0, sizeof(*info
));
4381 struct sctp_sock
*sp
= sctp_sk(sk
);
4383 info
->sctpi_s_autoclose
= sp
->autoclose
;
4384 info
->sctpi_s_adaptation_ind
= sp
->adaptation_ind
;
4385 info
->sctpi_s_pd_point
= sp
->pd_point
;
4386 info
->sctpi_s_nodelay
= sp
->nodelay
;
4387 info
->sctpi_s_disable_fragments
= sp
->disable_fragments
;
4388 info
->sctpi_s_v4mapped
= sp
->v4mapped
;
4389 info
->sctpi_s_frag_interleave
= sp
->frag_interleave
;
4390 info
->sctpi_s_type
= sp
->type
;
4395 info
->sctpi_tag
= asoc
->c
.my_vtag
;
4396 info
->sctpi_state
= asoc
->state
;
4397 info
->sctpi_rwnd
= asoc
->a_rwnd
;
4398 info
->sctpi_unackdata
= asoc
->unack_data
;
4399 info
->sctpi_penddata
= sctp_tsnmap_pending(&asoc
->peer
.tsn_map
);
4400 info
->sctpi_instrms
= asoc
->c
.sinit_max_instreams
;
4401 info
->sctpi_outstrms
= asoc
->c
.sinit_num_ostreams
;
4402 list_for_each(pos
, &asoc
->base
.inqueue
.in_chunk_list
)
4403 info
->sctpi_inqueue
++;
4404 list_for_each(pos
, &asoc
->outqueue
.out_chunk_list
)
4405 info
->sctpi_outqueue
++;
4406 info
->sctpi_overall_error
= asoc
->overall_error_count
;
4407 info
->sctpi_max_burst
= asoc
->max_burst
;
4408 info
->sctpi_maxseg
= asoc
->frag_point
;
4409 info
->sctpi_peer_rwnd
= asoc
->peer
.rwnd
;
4410 info
->sctpi_peer_tag
= asoc
->c
.peer_vtag
;
4412 mask
= asoc
->peer
.ecn_capable
<< 1;
4413 mask
= (mask
| asoc
->peer
.ipv4_address
) << 1;
4414 mask
= (mask
| asoc
->peer
.ipv6_address
) << 1;
4415 mask
= (mask
| asoc
->peer
.hostname_address
) << 1;
4416 mask
= (mask
| asoc
->peer
.asconf_capable
) << 1;
4417 mask
= (mask
| asoc
->peer
.prsctp_capable
) << 1;
4418 mask
= (mask
| asoc
->peer
.auth_capable
);
4419 info
->sctpi_peer_capable
= mask
;
4420 mask
= asoc
->peer
.sack_needed
<< 1;
4421 mask
= (mask
| asoc
->peer
.sack_generation
) << 1;
4422 mask
= (mask
| asoc
->peer
.zero_window_announced
);
4423 info
->sctpi_peer_sack
= mask
;
4425 info
->sctpi_isacks
= asoc
->stats
.isacks
;
4426 info
->sctpi_osacks
= asoc
->stats
.osacks
;
4427 info
->sctpi_opackets
= asoc
->stats
.opackets
;
4428 info
->sctpi_ipackets
= asoc
->stats
.ipackets
;
4429 info
->sctpi_rtxchunks
= asoc
->stats
.rtxchunks
;
4430 info
->sctpi_outofseqtsns
= asoc
->stats
.outofseqtsns
;
4431 info
->sctpi_idupchunks
= asoc
->stats
.idupchunks
;
4432 info
->sctpi_gapcnt
= asoc
->stats
.gapcnt
;
4433 info
->sctpi_ouodchunks
= asoc
->stats
.ouodchunks
;
4434 info
->sctpi_iuodchunks
= asoc
->stats
.iuodchunks
;
4435 info
->sctpi_oodchunks
= asoc
->stats
.oodchunks
;
4436 info
->sctpi_iodchunks
= asoc
->stats
.iodchunks
;
4437 info
->sctpi_octrlchunks
= asoc
->stats
.octrlchunks
;
4438 info
->sctpi_ictrlchunks
= asoc
->stats
.ictrlchunks
;
4440 prim
= asoc
->peer
.primary_path
;
4441 memcpy(&info
->sctpi_p_address
, &prim
->ipaddr
,
4442 sizeof(struct sockaddr_storage
));
4443 info
->sctpi_p_state
= prim
->state
;
4444 info
->sctpi_p_cwnd
= prim
->cwnd
;
4445 info
->sctpi_p_srtt
= prim
->srtt
;
4446 info
->sctpi_p_rto
= jiffies_to_msecs(prim
->rto
);
4447 info
->sctpi_p_hbinterval
= prim
->hbinterval
;
4448 info
->sctpi_p_pathmaxrxt
= prim
->pathmaxrxt
;
4449 info
->sctpi_p_sackdelay
= jiffies_to_msecs(prim
->sackdelay
);
4450 info
->sctpi_p_ssthresh
= prim
->ssthresh
;
4451 info
->sctpi_p_partial_bytes_acked
= prim
->partial_bytes_acked
;
4452 info
->sctpi_p_flight_size
= prim
->flight_size
;
4453 info
->sctpi_p_error
= prim
->error_count
;
4457 EXPORT_SYMBOL_GPL(sctp_get_sctp_info
);
4459 /* use callback to avoid exporting the core structure */
4460 int sctp_transport_walk_start(struct rhashtable_iter
*iter
)
4464 rhltable_walk_enter(&sctp_transport_hashtable
, iter
);
4466 err
= rhashtable_walk_start(iter
);
4467 if (err
&& err
!= -EAGAIN
) {
4468 rhashtable_walk_stop(iter
);
4469 rhashtable_walk_exit(iter
);
4476 void sctp_transport_walk_stop(struct rhashtable_iter
*iter
)
4478 rhashtable_walk_stop(iter
);
4479 rhashtable_walk_exit(iter
);
4482 struct sctp_transport
*sctp_transport_get_next(struct net
*net
,
4483 struct rhashtable_iter
*iter
)
4485 struct sctp_transport
*t
;
4487 t
= rhashtable_walk_next(iter
);
4488 for (; t
; t
= rhashtable_walk_next(iter
)) {
4490 if (PTR_ERR(t
) == -EAGAIN
)
4495 if (net_eq(sock_net(t
->asoc
->base
.sk
), net
) &&
4496 t
->asoc
->peer
.primary_path
== t
)
4503 struct sctp_transport
*sctp_transport_get_idx(struct net
*net
,
4504 struct rhashtable_iter
*iter
,
4507 void *obj
= SEQ_START_TOKEN
;
4509 while (pos
&& (obj
= sctp_transport_get_next(net
, iter
)) &&
4516 int sctp_for_each_endpoint(int (*cb
)(struct sctp_endpoint
*, void *),
4520 struct sctp_ep_common
*epb
;
4521 struct sctp_hashbucket
*head
;
4523 for (head
= sctp_ep_hashtable
; hash
< sctp_ep_hashsize
;
4525 read_lock(&head
->lock
);
4526 sctp_for_each_hentry(epb
, &head
->chain
) {
4527 err
= cb(sctp_ep(epb
), p
);
4531 read_unlock(&head
->lock
);
4536 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint
);
4538 int sctp_transport_lookup_process(int (*cb
)(struct sctp_transport
*, void *),
4540 const union sctp_addr
*laddr
,
4541 const union sctp_addr
*paddr
, void *p
)
4543 struct sctp_transport
*transport
;
4547 transport
= sctp_addrs_lookup_transport(net
, laddr
, paddr
);
4552 err
= cb(transport
, p
);
4553 sctp_transport_put(transport
);
4557 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process
);
4559 int sctp_for_each_transport(int (*cb
)(struct sctp_transport
*, void *),
4560 struct net
*net
, int pos
, void *p
) {
4561 struct rhashtable_iter hti
;
4565 err
= sctp_transport_walk_start(&hti
);
4569 sctp_transport_get_idx(net
, &hti
, pos
);
4570 obj
= sctp_transport_get_next(net
, &hti
);
4571 for (; obj
&& !IS_ERR(obj
); obj
= sctp_transport_get_next(net
, &hti
)) {
4572 struct sctp_transport
*transport
= obj
;
4574 if (!sctp_transport_hold(transport
))
4576 err
= cb(transport
, p
);
4577 sctp_transport_put(transport
);
4581 sctp_transport_walk_stop(&hti
);
4585 EXPORT_SYMBOL_GPL(sctp_for_each_transport
);
4587 /* 7.2.1 Association Status (SCTP_STATUS)
4589 * Applications can retrieve current status information about an
4590 * association, including association state, peer receiver window size,
4591 * number of unacked data chunks, and number of data chunks pending
4592 * receipt. This information is read-only.
4594 static int sctp_getsockopt_sctp_status(struct sock
*sk
, int len
,
4595 char __user
*optval
,
4598 struct sctp_status status
;
4599 struct sctp_association
*asoc
= NULL
;
4600 struct sctp_transport
*transport
;
4601 sctp_assoc_t associd
;
4604 if (len
< sizeof(status
)) {
4609 len
= sizeof(status
);
4610 if (copy_from_user(&status
, optval
, len
)) {
4615 associd
= status
.sstat_assoc_id
;
4616 asoc
= sctp_id2assoc(sk
, associd
);
4622 transport
= asoc
->peer
.primary_path
;
4624 status
.sstat_assoc_id
= sctp_assoc2id(asoc
);
4625 status
.sstat_state
= sctp_assoc_to_state(asoc
);
4626 status
.sstat_rwnd
= asoc
->peer
.rwnd
;
4627 status
.sstat_unackdata
= asoc
->unack_data
;
4629 status
.sstat_penddata
= sctp_tsnmap_pending(&asoc
->peer
.tsn_map
);
4630 status
.sstat_instrms
= asoc
->c
.sinit_max_instreams
;
4631 status
.sstat_outstrms
= asoc
->c
.sinit_num_ostreams
;
4632 status
.sstat_fragmentation_point
= asoc
->frag_point
;
4633 status
.sstat_primary
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
4634 memcpy(&status
.sstat_primary
.spinfo_address
, &transport
->ipaddr
,
4635 transport
->af_specific
->sockaddr_len
);
4636 /* Map ipv4 address into v4-mapped-on-v6 address. */
4637 sctp_get_pf_specific(sk
->sk_family
)->addr_to_user(sctp_sk(sk
),
4638 (union sctp_addr
*)&status
.sstat_primary
.spinfo_address
);
4639 status
.sstat_primary
.spinfo_state
= transport
->state
;
4640 status
.sstat_primary
.spinfo_cwnd
= transport
->cwnd
;
4641 status
.sstat_primary
.spinfo_srtt
= transport
->srtt
;
4642 status
.sstat_primary
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
4643 status
.sstat_primary
.spinfo_mtu
= transport
->pathmtu
;
4645 if (status
.sstat_primary
.spinfo_state
== SCTP_UNKNOWN
)
4646 status
.sstat_primary
.spinfo_state
= SCTP_ACTIVE
;
4648 if (put_user(len
, optlen
)) {
4653 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4654 __func__
, len
, status
.sstat_state
, status
.sstat_rwnd
,
4655 status
.sstat_assoc_id
);
4657 if (copy_to_user(optval
, &status
, len
)) {
4667 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4669 * Applications can retrieve information about a specific peer address
4670 * of an association, including its reachability state, congestion
4671 * window, and retransmission timer values. This information is
4674 static int sctp_getsockopt_peer_addr_info(struct sock
*sk
, int len
,
4675 char __user
*optval
,
4678 struct sctp_paddrinfo pinfo
;
4679 struct sctp_transport
*transport
;
4682 if (len
< sizeof(pinfo
)) {
4687 len
= sizeof(pinfo
);
4688 if (copy_from_user(&pinfo
, optval
, len
)) {
4693 transport
= sctp_addr_id2transport(sk
, &pinfo
.spinfo_address
,
4694 pinfo
.spinfo_assoc_id
);
4698 pinfo
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
4699 pinfo
.spinfo_state
= transport
->state
;
4700 pinfo
.spinfo_cwnd
= transport
->cwnd
;
4701 pinfo
.spinfo_srtt
= transport
->srtt
;
4702 pinfo
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
4703 pinfo
.spinfo_mtu
= transport
->pathmtu
;
4705 if (pinfo
.spinfo_state
== SCTP_UNKNOWN
)
4706 pinfo
.spinfo_state
= SCTP_ACTIVE
;
4708 if (put_user(len
, optlen
)) {
4713 if (copy_to_user(optval
, &pinfo
, len
)) {
4722 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4724 * This option is a on/off flag. If enabled no SCTP message
4725 * fragmentation will be performed. Instead if a message being sent
4726 * exceeds the current PMTU size, the message will NOT be sent and
4727 * instead a error will be indicated to the user.
4729 static int sctp_getsockopt_disable_fragments(struct sock
*sk
, int len
,
4730 char __user
*optval
, int __user
*optlen
)
4734 if (len
< sizeof(int))
4738 val
= (sctp_sk(sk
)->disable_fragments
== 1);
4739 if (put_user(len
, optlen
))
4741 if (copy_to_user(optval
, &val
, len
))
4746 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4748 * This socket option is used to specify various notifications and
4749 * ancillary data the user wishes to receive.
4751 static int sctp_getsockopt_events(struct sock
*sk
, int len
, char __user
*optval
,
4756 if (len
> sizeof(struct sctp_event_subscribe
))
4757 len
= sizeof(struct sctp_event_subscribe
);
4758 if (put_user(len
, optlen
))
4760 if (copy_to_user(optval
, &sctp_sk(sk
)->subscribe
, len
))
4765 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4767 * This socket option is applicable to the UDP-style socket only. When
4768 * set it will cause associations that are idle for more than the
4769 * specified number of seconds to automatically close. An association
4770 * being idle is defined an association that has NOT sent or received
4771 * user data. The special value of '0' indicates that no automatic
4772 * close of any associations should be performed. The option expects an
4773 * integer defining the number of seconds of idle time before an
4774 * association is closed.
4776 static int sctp_getsockopt_autoclose(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4778 /* Applicable to UDP-style socket only */
4779 if (sctp_style(sk
, TCP
))
4781 if (len
< sizeof(int))
4784 if (put_user(len
, optlen
))
4786 if (copy_to_user(optval
, &sctp_sk(sk
)->autoclose
, sizeof(int)))
4791 /* Helper routine to branch off an association to a new socket. */
4792 int sctp_do_peeloff(struct sock
*sk
, sctp_assoc_t id
, struct socket
**sockp
)
4794 struct sctp_association
*asoc
= sctp_id2assoc(sk
, id
);
4795 struct sctp_sock
*sp
= sctp_sk(sk
);
4796 struct socket
*sock
;
4802 /* An association cannot be branched off from an already peeled-off
4803 * socket, nor is this supported for tcp style sockets.
4805 if (!sctp_style(sk
, UDP
))
4808 /* Create a new socket. */
4809 err
= sock_create(sk
->sk_family
, SOCK_SEQPACKET
, IPPROTO_SCTP
, &sock
);
4813 sctp_copy_sock(sock
->sk
, sk
, asoc
);
4815 /* Make peeled-off sockets more like 1-1 accepted sockets.
4816 * Set the daddr and initialize id to something more random
4818 sp
->pf
->to_sk_daddr(&asoc
->peer
.primary_addr
, sk
);
4820 /* Populate the fields of the newsk from the oldsk and migrate the
4821 * asoc to the newsk.
4823 sctp_sock_migrate(sk
, sock
->sk
, asoc
, SCTP_SOCKET_UDP_HIGH_BANDWIDTH
);
4829 EXPORT_SYMBOL(sctp_do_peeloff
);
4831 static int sctp_getsockopt_peeloff(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4833 sctp_peeloff_arg_t peeloff
;
4834 struct socket
*newsock
;
4835 struct file
*newfile
;
4838 if (len
< sizeof(sctp_peeloff_arg_t
))
4840 len
= sizeof(sctp_peeloff_arg_t
);
4841 if (copy_from_user(&peeloff
, optval
, len
))
4844 retval
= sctp_do_peeloff(sk
, peeloff
.associd
, &newsock
);
4848 /* Map the socket to an unused fd that can be returned to the user. */
4849 retval
= get_unused_fd_flags(0);
4851 sock_release(newsock
);
4855 newfile
= sock_alloc_file(newsock
, 0, NULL
);
4856 if (IS_ERR(newfile
)) {
4857 put_unused_fd(retval
);
4858 sock_release(newsock
);
4859 return PTR_ERR(newfile
);
4862 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__
, sk
, newsock
->sk
,
4865 /* Return the fd mapped to the new socket. */
4866 if (put_user(len
, optlen
)) {
4868 put_unused_fd(retval
);
4871 peeloff
.sd
= retval
;
4872 if (copy_to_user(optval
, &peeloff
, len
)) {
4874 put_unused_fd(retval
);
4877 fd_install(retval
, newfile
);
4882 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4884 * Applications can enable or disable heartbeats for any peer address of
4885 * an association, modify an address's heartbeat interval, force a
4886 * heartbeat to be sent immediately, and adjust the address's maximum
4887 * number of retransmissions sent before an address is considered
4888 * unreachable. The following structure is used to access and modify an
4889 * address's parameters:
4891 * struct sctp_paddrparams {
4892 * sctp_assoc_t spp_assoc_id;
4893 * struct sockaddr_storage spp_address;
4894 * uint32_t spp_hbinterval;
4895 * uint16_t spp_pathmaxrxt;
4896 * uint32_t spp_pathmtu;
4897 * uint32_t spp_sackdelay;
4898 * uint32_t spp_flags;
4901 * spp_assoc_id - (one-to-many style socket) This is filled in the
4902 * application, and identifies the association for
4904 * spp_address - This specifies which address is of interest.
4905 * spp_hbinterval - This contains the value of the heartbeat interval,
4906 * in milliseconds. If a value of zero
4907 * is present in this field then no changes are to
4908 * be made to this parameter.
4909 * spp_pathmaxrxt - This contains the maximum number of
4910 * retransmissions before this address shall be
4911 * considered unreachable. If a value of zero
4912 * is present in this field then no changes are to
4913 * be made to this parameter.
4914 * spp_pathmtu - When Path MTU discovery is disabled the value
4915 * specified here will be the "fixed" path mtu.
4916 * Note that if the spp_address field is empty
4917 * then all associations on this address will
4918 * have this fixed path mtu set upon them.
4920 * spp_sackdelay - When delayed sack is enabled, this value specifies
4921 * the number of milliseconds that sacks will be delayed
4922 * for. This value will apply to all addresses of an
4923 * association if the spp_address field is empty. Note
4924 * also, that if delayed sack is enabled and this
4925 * value is set to 0, no change is made to the last
4926 * recorded delayed sack timer value.
4928 * spp_flags - These flags are used to control various features
4929 * on an association. The flag field may contain
4930 * zero or more of the following options.
4932 * SPP_HB_ENABLE - Enable heartbeats on the
4933 * specified address. Note that if the address
4934 * field is empty all addresses for the association
4935 * have heartbeats enabled upon them.
4937 * SPP_HB_DISABLE - Disable heartbeats on the
4938 * speicifed address. Note that if the address
4939 * field is empty all addresses for the association
4940 * will have their heartbeats disabled. Note also
4941 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4942 * mutually exclusive, only one of these two should
4943 * be specified. Enabling both fields will have
4944 * undetermined results.
4946 * SPP_HB_DEMAND - Request a user initiated heartbeat
4947 * to be made immediately.
4949 * SPP_PMTUD_ENABLE - This field will enable PMTU
4950 * discovery upon the specified address. Note that
4951 * if the address feild is empty then all addresses
4952 * on the association are effected.
4954 * SPP_PMTUD_DISABLE - This field will disable PMTU
4955 * discovery upon the specified address. Note that
4956 * if the address feild is empty then all addresses
4957 * on the association are effected. Not also that
4958 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4959 * exclusive. Enabling both will have undetermined
4962 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4963 * on delayed sack. The time specified in spp_sackdelay
4964 * is used to specify the sack delay for this address. Note
4965 * that if spp_address is empty then all addresses will
4966 * enable delayed sack and take on the sack delay
4967 * value specified in spp_sackdelay.
4968 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4969 * off delayed sack. If the spp_address field is blank then
4970 * delayed sack is disabled for the entire association. Note
4971 * also that this field is mutually exclusive to
4972 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4975 static int sctp_getsockopt_peer_addr_params(struct sock
*sk
, int len
,
4976 char __user
*optval
, int __user
*optlen
)
4978 struct sctp_paddrparams params
;
4979 struct sctp_transport
*trans
= NULL
;
4980 struct sctp_association
*asoc
= NULL
;
4981 struct sctp_sock
*sp
= sctp_sk(sk
);
4983 if (len
< sizeof(struct sctp_paddrparams
))
4985 len
= sizeof(struct sctp_paddrparams
);
4986 if (copy_from_user(¶ms
, optval
, len
))
4989 /* If an address other than INADDR_ANY is specified, and
4990 * no transport is found, then the request is invalid.
4992 if (!sctp_is_any(sk
, (union sctp_addr
*)¶ms
.spp_address
)) {
4993 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
4994 params
.spp_assoc_id
);
4996 pr_debug("%s: failed no transport\n", __func__
);
5001 /* Get association, if assoc_id != 0 and the socket is a one
5002 * to many style socket, and an association was not found, then
5003 * the id was invalid.
5005 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
5006 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
)) {
5007 pr_debug("%s: failed no association\n", __func__
);
5012 /* Fetch transport values. */
5013 params
.spp_hbinterval
= jiffies_to_msecs(trans
->hbinterval
);
5014 params
.spp_pathmtu
= trans
->pathmtu
;
5015 params
.spp_pathmaxrxt
= trans
->pathmaxrxt
;
5016 params
.spp_sackdelay
= jiffies_to_msecs(trans
->sackdelay
);
5018 /*draft-11 doesn't say what to return in spp_flags*/
5019 params
.spp_flags
= trans
->param_flags
;
5021 /* Fetch association values. */
5022 params
.spp_hbinterval
= jiffies_to_msecs(asoc
->hbinterval
);
5023 params
.spp_pathmtu
= asoc
->pathmtu
;
5024 params
.spp_pathmaxrxt
= asoc
->pathmaxrxt
;
5025 params
.spp_sackdelay
= jiffies_to_msecs(asoc
->sackdelay
);
5027 /*draft-11 doesn't say what to return in spp_flags*/
5028 params
.spp_flags
= asoc
->param_flags
;
5030 /* Fetch socket values. */
5031 params
.spp_hbinterval
= sp
->hbinterval
;
5032 params
.spp_pathmtu
= sp
->pathmtu
;
5033 params
.spp_sackdelay
= sp
->sackdelay
;
5034 params
.spp_pathmaxrxt
= sp
->pathmaxrxt
;
5036 /*draft-11 doesn't say what to return in spp_flags*/
5037 params
.spp_flags
= sp
->param_flags
;
5040 if (copy_to_user(optval
, ¶ms
, len
))
5043 if (put_user(len
, optlen
))
5050 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
5052 * This option will effect the way delayed acks are performed. This
5053 * option allows you to get or set the delayed ack time, in
5054 * milliseconds. It also allows changing the delayed ack frequency.
5055 * Changing the frequency to 1 disables the delayed sack algorithm. If
5056 * the assoc_id is 0, then this sets or gets the endpoints default
5057 * values. If the assoc_id field is non-zero, then the set or get
5058 * effects the specified association for the one to many model (the
5059 * assoc_id field is ignored by the one to one model). Note that if
5060 * sack_delay or sack_freq are 0 when setting this option, then the
5061 * current values will remain unchanged.
5063 * struct sctp_sack_info {
5064 * sctp_assoc_t sack_assoc_id;
5065 * uint32_t sack_delay;
5066 * uint32_t sack_freq;
5069 * sack_assoc_id - This parameter, indicates which association the user
5070 * is performing an action upon. Note that if this field's value is
5071 * zero then the endpoints default value is changed (effecting future
5072 * associations only).
5074 * sack_delay - This parameter contains the number of milliseconds that
5075 * the user is requesting the delayed ACK timer be set to. Note that
5076 * this value is defined in the standard to be between 200 and 500
5079 * sack_freq - This parameter contains the number of packets that must
5080 * be received before a sack is sent without waiting for the delay
5081 * timer to expire. The default value for this is 2, setting this
5082 * value to 1 will disable the delayed sack algorithm.
5084 static int sctp_getsockopt_delayed_ack(struct sock
*sk
, int len
,
5085 char __user
*optval
,
5088 struct sctp_sack_info params
;
5089 struct sctp_association
*asoc
= NULL
;
5090 struct sctp_sock
*sp
= sctp_sk(sk
);
5092 if (len
>= sizeof(struct sctp_sack_info
)) {
5093 len
= sizeof(struct sctp_sack_info
);
5095 if (copy_from_user(¶ms
, optval
, len
))
5097 } else if (len
== sizeof(struct sctp_assoc_value
)) {
5098 pr_warn_ratelimited(DEPRECATED
5100 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
5101 "Use struct sctp_sack_info instead\n",
5102 current
->comm
, task_pid_nr(current
));
5103 if (copy_from_user(¶ms
, optval
, len
))
5108 /* Get association, if sack_assoc_id != 0 and the socket is a one
5109 * to many style socket, and an association was not found, then
5110 * the id was invalid.
5112 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
5113 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
5117 /* Fetch association values. */
5118 if (asoc
->param_flags
& SPP_SACKDELAY_ENABLE
) {
5119 params
.sack_delay
= jiffies_to_msecs(
5121 params
.sack_freq
= asoc
->sackfreq
;
5124 params
.sack_delay
= 0;
5125 params
.sack_freq
= 1;
5128 /* Fetch socket values. */
5129 if (sp
->param_flags
& SPP_SACKDELAY_ENABLE
) {
5130 params
.sack_delay
= sp
->sackdelay
;
5131 params
.sack_freq
= sp
->sackfreq
;
5133 params
.sack_delay
= 0;
5134 params
.sack_freq
= 1;
5138 if (copy_to_user(optval
, ¶ms
, len
))
5141 if (put_user(len
, optlen
))
5147 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5149 * Applications can specify protocol parameters for the default association
5150 * initialization. The option name argument to setsockopt() and getsockopt()
5153 * Setting initialization parameters is effective only on an unconnected
5154 * socket (for UDP-style sockets only future associations are effected
5155 * by the change). With TCP-style sockets, this option is inherited by
5156 * sockets derived from a listener socket.
5158 static int sctp_getsockopt_initmsg(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
5160 if (len
< sizeof(struct sctp_initmsg
))
5162 len
= sizeof(struct sctp_initmsg
);
5163 if (put_user(len
, optlen
))
5165 if (copy_to_user(optval
, &sctp_sk(sk
)->initmsg
, len
))
5171 static int sctp_getsockopt_peer_addrs(struct sock
*sk
, int len
,
5172 char __user
*optval
, int __user
*optlen
)
5174 struct sctp_association
*asoc
;
5176 struct sctp_getaddrs getaddrs
;
5177 struct sctp_transport
*from
;
5179 union sctp_addr temp
;
5180 struct sctp_sock
*sp
= sctp_sk(sk
);
5185 if (len
< sizeof(struct sctp_getaddrs
))
5188 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
5191 /* For UDP-style sockets, id specifies the association to query. */
5192 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
5196 to
= optval
+ offsetof(struct sctp_getaddrs
, addrs
);
5197 space_left
= len
- offsetof(struct sctp_getaddrs
, addrs
);
5199 list_for_each_entry(from
, &asoc
->peer
.transport_addr_list
,
5201 memcpy(&temp
, &from
->ipaddr
, sizeof(temp
));
5202 addrlen
= sctp_get_pf_specific(sk
->sk_family
)
5203 ->addr_to_user(sp
, &temp
);
5204 if (space_left
< addrlen
)
5206 if (copy_to_user(to
, &temp
, addrlen
))
5210 space_left
-= addrlen
;
5213 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
))
5215 bytes_copied
= ((char __user
*)to
) - optval
;
5216 if (put_user(bytes_copied
, optlen
))
5222 static int sctp_copy_laddrs(struct sock
*sk
, __u16 port
, void *to
,
5223 size_t space_left
, int *bytes_copied
)
5225 struct sctp_sockaddr_entry
*addr
;
5226 union sctp_addr temp
;
5229 struct net
*net
= sock_net(sk
);
5232 list_for_each_entry_rcu(addr
, &net
->sctp
.local_addr_list
, list
) {
5236 if ((PF_INET
== sk
->sk_family
) &&
5237 (AF_INET6
== addr
->a
.sa
.sa_family
))
5239 if ((PF_INET6
== sk
->sk_family
) &&
5240 inet_v6_ipv6only(sk
) &&
5241 (AF_INET
== addr
->a
.sa
.sa_family
))
5243 memcpy(&temp
, &addr
->a
, sizeof(temp
));
5244 if (!temp
.v4
.sin_port
)
5245 temp
.v4
.sin_port
= htons(port
);
5247 addrlen
= sctp_get_pf_specific(sk
->sk_family
)
5248 ->addr_to_user(sctp_sk(sk
), &temp
);
5250 if (space_left
< addrlen
) {
5254 memcpy(to
, &temp
, addrlen
);
5258 space_left
-= addrlen
;
5259 *bytes_copied
+= addrlen
;
5267 static int sctp_getsockopt_local_addrs(struct sock
*sk
, int len
,
5268 char __user
*optval
, int __user
*optlen
)
5270 struct sctp_bind_addr
*bp
;
5271 struct sctp_association
*asoc
;
5273 struct sctp_getaddrs getaddrs
;
5274 struct sctp_sockaddr_entry
*addr
;
5276 union sctp_addr temp
;
5277 struct sctp_sock
*sp
= sctp_sk(sk
);
5281 int bytes_copied
= 0;
5285 if (len
< sizeof(struct sctp_getaddrs
))
5288 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
5292 * For UDP-style sockets, id specifies the association to query.
5293 * If the id field is set to the value '0' then the locally bound
5294 * addresses are returned without regard to any particular
5297 if (0 == getaddrs
.assoc_id
) {
5298 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
5300 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
5303 bp
= &asoc
->base
.bind_addr
;
5306 to
= optval
+ offsetof(struct sctp_getaddrs
, addrs
);
5307 space_left
= len
- offsetof(struct sctp_getaddrs
, addrs
);
5309 addrs
= kmalloc(space_left
, GFP_USER
| __GFP_NOWARN
);
5313 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5314 * addresses from the global local address list.
5316 if (sctp_list_single_entry(&bp
->address_list
)) {
5317 addr
= list_entry(bp
->address_list
.next
,
5318 struct sctp_sockaddr_entry
, list
);
5319 if (sctp_is_any(sk
, &addr
->a
)) {
5320 cnt
= sctp_copy_laddrs(sk
, bp
->port
, addrs
,
5321 space_left
, &bytes_copied
);
5331 /* Protection on the bound address list is not needed since
5332 * in the socket option context we hold a socket lock and
5333 * thus the bound address list can't change.
5335 list_for_each_entry(addr
, &bp
->address_list
, list
) {
5336 memcpy(&temp
, &addr
->a
, sizeof(temp
));
5337 addrlen
= sctp_get_pf_specific(sk
->sk_family
)
5338 ->addr_to_user(sp
, &temp
);
5339 if (space_left
< addrlen
) {
5340 err
= -ENOMEM
; /*fixme: right error?*/
5343 memcpy(buf
, &temp
, addrlen
);
5345 bytes_copied
+= addrlen
;
5347 space_left
-= addrlen
;
5351 if (copy_to_user(to
, addrs
, bytes_copied
)) {
5355 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
)) {
5359 if (put_user(bytes_copied
, optlen
))
5366 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
5368 * Requests that the local SCTP stack use the enclosed peer address as
5369 * the association primary. The enclosed address must be one of the
5370 * association peer's addresses.
5372 static int sctp_getsockopt_primary_addr(struct sock
*sk
, int len
,
5373 char __user
*optval
, int __user
*optlen
)
5375 struct sctp_prim prim
;
5376 struct sctp_association
*asoc
;
5377 struct sctp_sock
*sp
= sctp_sk(sk
);
5379 if (len
< sizeof(struct sctp_prim
))
5382 len
= sizeof(struct sctp_prim
);
5384 if (copy_from_user(&prim
, optval
, len
))
5387 asoc
= sctp_id2assoc(sk
, prim
.ssp_assoc_id
);
5391 if (!asoc
->peer
.primary_path
)
5394 memcpy(&prim
.ssp_addr
, &asoc
->peer
.primary_path
->ipaddr
,
5395 asoc
->peer
.primary_path
->af_specific
->sockaddr_len
);
5397 sctp_get_pf_specific(sk
->sk_family
)->addr_to_user(sp
,
5398 (union sctp_addr
*)&prim
.ssp_addr
);
5400 if (put_user(len
, optlen
))
5402 if (copy_to_user(optval
, &prim
, len
))
5409 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5411 * Requests that the local endpoint set the specified Adaptation Layer
5412 * Indication parameter for all future INIT and INIT-ACK exchanges.
5414 static int sctp_getsockopt_adaptation_layer(struct sock
*sk
, int len
,
5415 char __user
*optval
, int __user
*optlen
)
5417 struct sctp_setadaptation adaptation
;
5419 if (len
< sizeof(struct sctp_setadaptation
))
5422 len
= sizeof(struct sctp_setadaptation
);
5424 adaptation
.ssb_adaptation_ind
= sctp_sk(sk
)->adaptation_ind
;
5426 if (put_user(len
, optlen
))
5428 if (copy_to_user(optval
, &adaptation
, len
))
5436 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5438 * Applications that wish to use the sendto() system call may wish to
5439 * specify a default set of parameters that would normally be supplied
5440 * through the inclusion of ancillary data. This socket option allows
5441 * such an application to set the default sctp_sndrcvinfo structure.
5444 * The application that wishes to use this socket option simply passes
5445 * in to this call the sctp_sndrcvinfo structure defined in Section
5446 * 5.2.2) The input parameters accepted by this call include
5447 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5448 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5449 * to this call if the caller is using the UDP model.
5451 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5453 static int sctp_getsockopt_default_send_param(struct sock
*sk
,
5454 int len
, char __user
*optval
,
5457 struct sctp_sock
*sp
= sctp_sk(sk
);
5458 struct sctp_association
*asoc
;
5459 struct sctp_sndrcvinfo info
;
5461 if (len
< sizeof(info
))
5466 if (copy_from_user(&info
, optval
, len
))
5469 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
5470 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
5473 info
.sinfo_stream
= asoc
->default_stream
;
5474 info
.sinfo_flags
= asoc
->default_flags
;
5475 info
.sinfo_ppid
= asoc
->default_ppid
;
5476 info
.sinfo_context
= asoc
->default_context
;
5477 info
.sinfo_timetolive
= asoc
->default_timetolive
;
5479 info
.sinfo_stream
= sp
->default_stream
;
5480 info
.sinfo_flags
= sp
->default_flags
;
5481 info
.sinfo_ppid
= sp
->default_ppid
;
5482 info
.sinfo_context
= sp
->default_context
;
5483 info
.sinfo_timetolive
= sp
->default_timetolive
;
5486 if (put_user(len
, optlen
))
5488 if (copy_to_user(optval
, &info
, len
))
5494 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5495 * (SCTP_DEFAULT_SNDINFO)
5497 static int sctp_getsockopt_default_sndinfo(struct sock
*sk
, int len
,
5498 char __user
*optval
,
5501 struct sctp_sock
*sp
= sctp_sk(sk
);
5502 struct sctp_association
*asoc
;
5503 struct sctp_sndinfo info
;
5505 if (len
< sizeof(info
))
5510 if (copy_from_user(&info
, optval
, len
))
5513 asoc
= sctp_id2assoc(sk
, info
.snd_assoc_id
);
5514 if (!asoc
&& info
.snd_assoc_id
&& sctp_style(sk
, UDP
))
5517 info
.snd_sid
= asoc
->default_stream
;
5518 info
.snd_flags
= asoc
->default_flags
;
5519 info
.snd_ppid
= asoc
->default_ppid
;
5520 info
.snd_context
= asoc
->default_context
;
5522 info
.snd_sid
= sp
->default_stream
;
5523 info
.snd_flags
= sp
->default_flags
;
5524 info
.snd_ppid
= sp
->default_ppid
;
5525 info
.snd_context
= sp
->default_context
;
5528 if (put_user(len
, optlen
))
5530 if (copy_to_user(optval
, &info
, len
))
5538 * 7.1.5 SCTP_NODELAY
5540 * Turn on/off any Nagle-like algorithm. This means that packets are
5541 * generally sent as soon as possible and no unnecessary delays are
5542 * introduced, at the cost of more packets in the network. Expects an
5543 * integer boolean flag.
5546 static int sctp_getsockopt_nodelay(struct sock
*sk
, int len
,
5547 char __user
*optval
, int __user
*optlen
)
5551 if (len
< sizeof(int))
5555 val
= (sctp_sk(sk
)->nodelay
== 1);
5556 if (put_user(len
, optlen
))
5558 if (copy_to_user(optval
, &val
, len
))
5565 * 7.1.1 SCTP_RTOINFO
5567 * The protocol parameters used to initialize and bound retransmission
5568 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5569 * and modify these parameters.
5570 * All parameters are time values, in milliseconds. A value of 0, when
5571 * modifying the parameters, indicates that the current value should not
5575 static int sctp_getsockopt_rtoinfo(struct sock
*sk
, int len
,
5576 char __user
*optval
,
5577 int __user
*optlen
) {
5578 struct sctp_rtoinfo rtoinfo
;
5579 struct sctp_association
*asoc
;
5581 if (len
< sizeof (struct sctp_rtoinfo
))
5584 len
= sizeof(struct sctp_rtoinfo
);
5586 if (copy_from_user(&rtoinfo
, optval
, len
))
5589 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
5591 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
5594 /* Values corresponding to the specific association. */
5596 rtoinfo
.srto_initial
= jiffies_to_msecs(asoc
->rto_initial
);
5597 rtoinfo
.srto_max
= jiffies_to_msecs(asoc
->rto_max
);
5598 rtoinfo
.srto_min
= jiffies_to_msecs(asoc
->rto_min
);
5600 /* Values corresponding to the endpoint. */
5601 struct sctp_sock
*sp
= sctp_sk(sk
);
5603 rtoinfo
.srto_initial
= sp
->rtoinfo
.srto_initial
;
5604 rtoinfo
.srto_max
= sp
->rtoinfo
.srto_max
;
5605 rtoinfo
.srto_min
= sp
->rtoinfo
.srto_min
;
5608 if (put_user(len
, optlen
))
5611 if (copy_to_user(optval
, &rtoinfo
, len
))
5619 * 7.1.2 SCTP_ASSOCINFO
5621 * This option is used to tune the maximum retransmission attempts
5622 * of the association.
5623 * Returns an error if the new association retransmission value is
5624 * greater than the sum of the retransmission value of the peer.
5625 * See [SCTP] for more information.
5628 static int sctp_getsockopt_associnfo(struct sock
*sk
, int len
,
5629 char __user
*optval
,
5633 struct sctp_assocparams assocparams
;
5634 struct sctp_association
*asoc
;
5635 struct list_head
*pos
;
5638 if (len
< sizeof (struct sctp_assocparams
))
5641 len
= sizeof(struct sctp_assocparams
);
5643 if (copy_from_user(&assocparams
, optval
, len
))
5646 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
5648 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
5651 /* Values correspoinding to the specific association */
5653 assocparams
.sasoc_asocmaxrxt
= asoc
->max_retrans
;
5654 assocparams
.sasoc_peer_rwnd
= asoc
->peer
.rwnd
;
5655 assocparams
.sasoc_local_rwnd
= asoc
->a_rwnd
;
5656 assocparams
.sasoc_cookie_life
= ktime_to_ms(asoc
->cookie_life
);
5658 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
5662 assocparams
.sasoc_number_peer_destinations
= cnt
;
5664 /* Values corresponding to the endpoint */
5665 struct sctp_sock
*sp
= sctp_sk(sk
);
5667 assocparams
.sasoc_asocmaxrxt
= sp
->assocparams
.sasoc_asocmaxrxt
;
5668 assocparams
.sasoc_peer_rwnd
= sp
->assocparams
.sasoc_peer_rwnd
;
5669 assocparams
.sasoc_local_rwnd
= sp
->assocparams
.sasoc_local_rwnd
;
5670 assocparams
.sasoc_cookie_life
=
5671 sp
->assocparams
.sasoc_cookie_life
;
5672 assocparams
.sasoc_number_peer_destinations
=
5674 sasoc_number_peer_destinations
;
5677 if (put_user(len
, optlen
))
5680 if (copy_to_user(optval
, &assocparams
, len
))
5687 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5689 * This socket option is a boolean flag which turns on or off mapped V4
5690 * addresses. If this option is turned on and the socket is type
5691 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5692 * If this option is turned off, then no mapping will be done of V4
5693 * addresses and a user will receive both PF_INET6 and PF_INET type
5694 * addresses on the socket.
5696 static int sctp_getsockopt_mappedv4(struct sock
*sk
, int len
,
5697 char __user
*optval
, int __user
*optlen
)
5700 struct sctp_sock
*sp
= sctp_sk(sk
);
5702 if (len
< sizeof(int))
5707 if (put_user(len
, optlen
))
5709 if (copy_to_user(optval
, &val
, len
))
5716 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5717 * (chapter and verse is quoted at sctp_setsockopt_context())
5719 static int sctp_getsockopt_context(struct sock
*sk
, int len
,
5720 char __user
*optval
, int __user
*optlen
)
5722 struct sctp_assoc_value params
;
5723 struct sctp_sock
*sp
;
5724 struct sctp_association
*asoc
;
5726 if (len
< sizeof(struct sctp_assoc_value
))
5729 len
= sizeof(struct sctp_assoc_value
);
5731 if (copy_from_user(¶ms
, optval
, len
))
5736 if (params
.assoc_id
!= 0) {
5737 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5740 params
.assoc_value
= asoc
->default_rcv_context
;
5742 params
.assoc_value
= sp
->default_rcv_context
;
5745 if (put_user(len
, optlen
))
5747 if (copy_to_user(optval
, ¶ms
, len
))
5754 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5755 * This option will get or set the maximum size to put in any outgoing
5756 * SCTP DATA chunk. If a message is larger than this size it will be
5757 * fragmented by SCTP into the specified size. Note that the underlying
5758 * SCTP implementation may fragment into smaller sized chunks when the
5759 * PMTU of the underlying association is smaller than the value set by
5760 * the user. The default value for this option is '0' which indicates
5761 * the user is NOT limiting fragmentation and only the PMTU will effect
5762 * SCTP's choice of DATA chunk size. Note also that values set larger
5763 * than the maximum size of an IP datagram will effectively let SCTP
5764 * control fragmentation (i.e. the same as setting this option to 0).
5766 * The following structure is used to access and modify this parameter:
5768 * struct sctp_assoc_value {
5769 * sctp_assoc_t assoc_id;
5770 * uint32_t assoc_value;
5773 * assoc_id: This parameter is ignored for one-to-one style sockets.
5774 * For one-to-many style sockets this parameter indicates which
5775 * association the user is performing an action upon. Note that if
5776 * this field's value is zero then the endpoints default value is
5777 * changed (effecting future associations only).
5778 * assoc_value: This parameter specifies the maximum size in bytes.
5780 static int sctp_getsockopt_maxseg(struct sock
*sk
, int len
,
5781 char __user
*optval
, int __user
*optlen
)
5783 struct sctp_assoc_value params
;
5784 struct sctp_association
*asoc
;
5786 if (len
== sizeof(int)) {
5787 pr_warn_ratelimited(DEPRECATED
5789 "Use of int in maxseg socket option.\n"
5790 "Use struct sctp_assoc_value instead\n",
5791 current
->comm
, task_pid_nr(current
));
5792 params
.assoc_id
= 0;
5793 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5794 len
= sizeof(struct sctp_assoc_value
);
5795 if (copy_from_user(¶ms
, optval
, sizeof(params
)))
5800 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5801 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
5805 params
.assoc_value
= asoc
->frag_point
;
5807 params
.assoc_value
= sctp_sk(sk
)->user_frag
;
5809 if (put_user(len
, optlen
))
5811 if (len
== sizeof(int)) {
5812 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5815 if (copy_to_user(optval
, ¶ms
, len
))
5823 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5824 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5826 static int sctp_getsockopt_fragment_interleave(struct sock
*sk
, int len
,
5827 char __user
*optval
, int __user
*optlen
)
5831 if (len
< sizeof(int))
5836 val
= sctp_sk(sk
)->frag_interleave
;
5837 if (put_user(len
, optlen
))
5839 if (copy_to_user(optval
, &val
, len
))
5846 * 7.1.25. Set or Get the sctp partial delivery point
5847 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5849 static int sctp_getsockopt_partial_delivery_point(struct sock
*sk
, int len
,
5850 char __user
*optval
,
5855 if (len
< sizeof(u32
))
5860 val
= sctp_sk(sk
)->pd_point
;
5861 if (put_user(len
, optlen
))
5863 if (copy_to_user(optval
, &val
, len
))
5870 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5871 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5873 static int sctp_getsockopt_maxburst(struct sock
*sk
, int len
,
5874 char __user
*optval
,
5877 struct sctp_assoc_value params
;
5878 struct sctp_sock
*sp
;
5879 struct sctp_association
*asoc
;
5881 if (len
== sizeof(int)) {
5882 pr_warn_ratelimited(DEPRECATED
5884 "Use of int in max_burst socket option.\n"
5885 "Use struct sctp_assoc_value instead\n",
5886 current
->comm
, task_pid_nr(current
));
5887 params
.assoc_id
= 0;
5888 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5889 len
= sizeof(struct sctp_assoc_value
);
5890 if (copy_from_user(¶ms
, optval
, len
))
5897 if (params
.assoc_id
!= 0) {
5898 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5901 params
.assoc_value
= asoc
->max_burst
;
5903 params
.assoc_value
= sp
->max_burst
;
5905 if (len
== sizeof(int)) {
5906 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5909 if (copy_to_user(optval
, ¶ms
, len
))
5917 static int sctp_getsockopt_hmac_ident(struct sock
*sk
, int len
,
5918 char __user
*optval
, int __user
*optlen
)
5920 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5921 struct sctp_hmacalgo __user
*p
= (void __user
*)optval
;
5922 struct sctp_hmac_algo_param
*hmacs
;
5927 if (!ep
->auth_enable
)
5930 hmacs
= ep
->auth_hmacs_list
;
5931 data_len
= ntohs(hmacs
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5933 if (len
< sizeof(struct sctp_hmacalgo
) + data_len
)
5936 len
= sizeof(struct sctp_hmacalgo
) + data_len
;
5937 num_idents
= data_len
/ sizeof(u16
);
5939 if (put_user(len
, optlen
))
5941 if (put_user(num_idents
, &p
->shmac_num_idents
))
5943 for (i
= 0; i
< num_idents
; i
++) {
5944 __u16 hmacid
= ntohs(hmacs
->hmac_ids
[i
]);
5946 if (copy_to_user(&p
->shmac_idents
[i
], &hmacid
, sizeof(__u16
)))
5952 static int sctp_getsockopt_active_key(struct sock
*sk
, int len
,
5953 char __user
*optval
, int __user
*optlen
)
5955 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5956 struct sctp_authkeyid val
;
5957 struct sctp_association
*asoc
;
5959 if (!ep
->auth_enable
)
5962 if (len
< sizeof(struct sctp_authkeyid
))
5964 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authkeyid
)))
5967 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
5968 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
5972 val
.scact_keynumber
= asoc
->active_key_id
;
5974 val
.scact_keynumber
= ep
->active_key_id
;
5976 len
= sizeof(struct sctp_authkeyid
);
5977 if (put_user(len
, optlen
))
5979 if (copy_to_user(optval
, &val
, len
))
5985 static int sctp_getsockopt_peer_auth_chunks(struct sock
*sk
, int len
,
5986 char __user
*optval
, int __user
*optlen
)
5988 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5989 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5990 struct sctp_authchunks val
;
5991 struct sctp_association
*asoc
;
5992 struct sctp_chunks_param
*ch
;
5996 if (!ep
->auth_enable
)
5999 if (len
< sizeof(struct sctp_authchunks
))
6002 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
6005 to
= p
->gauth_chunks
;
6006 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
6010 ch
= asoc
->peer
.peer_chunks
;
6014 /* See if the user provided enough room for all the data */
6015 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
6016 if (len
< num_chunks
)
6019 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
6022 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
6023 if (put_user(len
, optlen
))
6025 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
6030 static int sctp_getsockopt_local_auth_chunks(struct sock
*sk
, int len
,
6031 char __user
*optval
, int __user
*optlen
)
6033 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
6034 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
6035 struct sctp_authchunks val
;
6036 struct sctp_association
*asoc
;
6037 struct sctp_chunks_param
*ch
;
6041 if (!ep
->auth_enable
)
6044 if (len
< sizeof(struct sctp_authchunks
))
6047 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
6050 to
= p
->gauth_chunks
;
6051 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
6052 if (!asoc
&& val
.gauth_assoc_id
&& sctp_style(sk
, UDP
))
6056 ch
= (struct sctp_chunks_param
*)asoc
->c
.auth_chunks
;
6058 ch
= ep
->auth_chunk_list
;
6063 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
6064 if (len
< sizeof(struct sctp_authchunks
) + num_chunks
)
6067 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
6070 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
6071 if (put_user(len
, optlen
))
6073 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
6080 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
6081 * This option gets the current number of associations that are attached
6082 * to a one-to-many style socket. The option value is an uint32_t.
6084 static int sctp_getsockopt_assoc_number(struct sock
*sk
, int len
,
6085 char __user
*optval
, int __user
*optlen
)
6087 struct sctp_sock
*sp
= sctp_sk(sk
);
6088 struct sctp_association
*asoc
;
6091 if (sctp_style(sk
, TCP
))
6094 if (len
< sizeof(u32
))
6099 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
6103 if (put_user(len
, optlen
))
6105 if (copy_to_user(optval
, &val
, len
))
6112 * 8.1.23 SCTP_AUTO_ASCONF
6113 * See the corresponding setsockopt entry as description
6115 static int sctp_getsockopt_auto_asconf(struct sock
*sk
, int len
,
6116 char __user
*optval
, int __user
*optlen
)
6120 if (len
< sizeof(int))
6124 if (sctp_sk(sk
)->do_auto_asconf
&& sctp_is_ep_boundall(sk
))
6126 if (put_user(len
, optlen
))
6128 if (copy_to_user(optval
, &val
, len
))
6134 * 8.2.6. Get the Current Identifiers of Associations
6135 * (SCTP_GET_ASSOC_ID_LIST)
6137 * This option gets the current list of SCTP association identifiers of
6138 * the SCTP associations handled by a one-to-many style socket.
6140 static int sctp_getsockopt_assoc_ids(struct sock
*sk
, int len
,
6141 char __user
*optval
, int __user
*optlen
)
6143 struct sctp_sock
*sp
= sctp_sk(sk
);
6144 struct sctp_association
*asoc
;
6145 struct sctp_assoc_ids
*ids
;
6148 if (sctp_style(sk
, TCP
))
6151 if (len
< sizeof(struct sctp_assoc_ids
))
6154 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
6158 if (len
< sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
)
6161 len
= sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
;
6163 ids
= kmalloc(len
, GFP_USER
| __GFP_NOWARN
);
6167 ids
->gaids_number_of_ids
= num
;
6169 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
6170 ids
->gaids_assoc_id
[num
++] = asoc
->assoc_id
;
6173 if (put_user(len
, optlen
) || copy_to_user(optval
, ids
, len
)) {
6183 * SCTP_PEER_ADDR_THLDS
6185 * This option allows us to fetch the partially failed threshold for one or all
6186 * transports in an association. See Section 6.1 of:
6187 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6189 static int sctp_getsockopt_paddr_thresholds(struct sock
*sk
,
6190 char __user
*optval
,
6194 struct sctp_paddrthlds val
;
6195 struct sctp_transport
*trans
;
6196 struct sctp_association
*asoc
;
6198 if (len
< sizeof(struct sctp_paddrthlds
))
6200 len
= sizeof(struct sctp_paddrthlds
);
6201 if (copy_from_user(&val
, (struct sctp_paddrthlds __user
*)optval
, len
))
6204 if (sctp_is_any(sk
, (const union sctp_addr
*)&val
.spt_address
)) {
6205 asoc
= sctp_id2assoc(sk
, val
.spt_assoc_id
);
6209 val
.spt_pathpfthld
= asoc
->pf_retrans
;
6210 val
.spt_pathmaxrxt
= asoc
->pathmaxrxt
;
6212 trans
= sctp_addr_id2transport(sk
, &val
.spt_address
,
6217 val
.spt_pathmaxrxt
= trans
->pathmaxrxt
;
6218 val
.spt_pathpfthld
= trans
->pf_retrans
;
6221 if (put_user(len
, optlen
) || copy_to_user(optval
, &val
, len
))
6228 * SCTP_GET_ASSOC_STATS
6230 * This option retrieves local per endpoint statistics. It is modeled
6231 * after OpenSolaris' implementation
6233 static int sctp_getsockopt_assoc_stats(struct sock
*sk
, int len
,
6234 char __user
*optval
,
6237 struct sctp_assoc_stats sas
;
6238 struct sctp_association
*asoc
= NULL
;
6240 /* User must provide at least the assoc id */
6241 if (len
< sizeof(sctp_assoc_t
))
6244 /* Allow the struct to grow and fill in as much as possible */
6245 len
= min_t(size_t, len
, sizeof(sas
));
6247 if (copy_from_user(&sas
, optval
, len
))
6250 asoc
= sctp_id2assoc(sk
, sas
.sas_assoc_id
);
6254 sas
.sas_rtxchunks
= asoc
->stats
.rtxchunks
;
6255 sas
.sas_gapcnt
= asoc
->stats
.gapcnt
;
6256 sas
.sas_outofseqtsns
= asoc
->stats
.outofseqtsns
;
6257 sas
.sas_osacks
= asoc
->stats
.osacks
;
6258 sas
.sas_isacks
= asoc
->stats
.isacks
;
6259 sas
.sas_octrlchunks
= asoc
->stats
.octrlchunks
;
6260 sas
.sas_ictrlchunks
= asoc
->stats
.ictrlchunks
;
6261 sas
.sas_oodchunks
= asoc
->stats
.oodchunks
;
6262 sas
.sas_iodchunks
= asoc
->stats
.iodchunks
;
6263 sas
.sas_ouodchunks
= asoc
->stats
.ouodchunks
;
6264 sas
.sas_iuodchunks
= asoc
->stats
.iuodchunks
;
6265 sas
.sas_idupchunks
= asoc
->stats
.idupchunks
;
6266 sas
.sas_opackets
= asoc
->stats
.opackets
;
6267 sas
.sas_ipackets
= asoc
->stats
.ipackets
;
6269 /* New high max rto observed, will return 0 if not a single
6270 * RTO update took place. obs_rto_ipaddr will be bogus
6273 sas
.sas_maxrto
= asoc
->stats
.max_obs_rto
;
6274 memcpy(&sas
.sas_obs_rto_ipaddr
, &asoc
->stats
.obs_rto_ipaddr
,
6275 sizeof(struct sockaddr_storage
));
6277 /* Mark beginning of a new observation period */
6278 asoc
->stats
.max_obs_rto
= asoc
->rto_min
;
6280 if (put_user(len
, optlen
))
6283 pr_debug("%s: len:%d, assoc_id:%d\n", __func__
, len
, sas
.sas_assoc_id
);
6285 if (copy_to_user(optval
, &sas
, len
))
6291 static int sctp_getsockopt_recvrcvinfo(struct sock
*sk
, int len
,
6292 char __user
*optval
,
6297 if (len
< sizeof(int))
6301 if (sctp_sk(sk
)->recvrcvinfo
)
6303 if (put_user(len
, optlen
))
6305 if (copy_to_user(optval
, &val
, len
))
6311 static int sctp_getsockopt_recvnxtinfo(struct sock
*sk
, int len
,
6312 char __user
*optval
,
6317 if (len
< sizeof(int))
6321 if (sctp_sk(sk
)->recvnxtinfo
)
6323 if (put_user(len
, optlen
))
6325 if (copy_to_user(optval
, &val
, len
))
6331 static int sctp_getsockopt_pr_supported(struct sock
*sk
, int len
,
6332 char __user
*optval
,
6335 struct sctp_assoc_value params
;
6336 struct sctp_association
*asoc
;
6337 int retval
= -EFAULT
;
6339 if (len
< sizeof(params
)) {
6344 len
= sizeof(params
);
6345 if (copy_from_user(¶ms
, optval
, len
))
6348 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
6350 params
.assoc_value
= asoc
->prsctp_enable
;
6351 } else if (!params
.assoc_id
) {
6352 struct sctp_sock
*sp
= sctp_sk(sk
);
6354 params
.assoc_value
= sp
->ep
->prsctp_enable
;
6360 if (put_user(len
, optlen
))
6363 if (copy_to_user(optval
, ¶ms
, len
))
6372 static int sctp_getsockopt_default_prinfo(struct sock
*sk
, int len
,
6373 char __user
*optval
,
6376 struct sctp_default_prinfo info
;
6377 struct sctp_association
*asoc
;
6378 int retval
= -EFAULT
;
6380 if (len
< sizeof(info
)) {
6386 if (copy_from_user(&info
, optval
, len
))
6389 asoc
= sctp_id2assoc(sk
, info
.pr_assoc_id
);
6391 info
.pr_policy
= SCTP_PR_POLICY(asoc
->default_flags
);
6392 info
.pr_value
= asoc
->default_timetolive
;
6393 } else if (!info
.pr_assoc_id
) {
6394 struct sctp_sock
*sp
= sctp_sk(sk
);
6396 info
.pr_policy
= SCTP_PR_POLICY(sp
->default_flags
);
6397 info
.pr_value
= sp
->default_timetolive
;
6403 if (put_user(len
, optlen
))
6406 if (copy_to_user(optval
, &info
, len
))
6415 static int sctp_getsockopt_pr_assocstatus(struct sock
*sk
, int len
,
6416 char __user
*optval
,
6419 struct sctp_prstatus params
;
6420 struct sctp_association
*asoc
;
6422 int retval
= -EINVAL
;
6424 if (len
< sizeof(params
))
6427 len
= sizeof(params
);
6428 if (copy_from_user(¶ms
, optval
, len
)) {
6433 policy
= params
.sprstat_policy
;
6434 if (policy
& ~SCTP_PR_SCTP_MASK
)
6437 asoc
= sctp_id2assoc(sk
, params
.sprstat_assoc_id
);
6441 if (policy
== SCTP_PR_SCTP_NONE
) {
6442 params
.sprstat_abandoned_unsent
= 0;
6443 params
.sprstat_abandoned_sent
= 0;
6444 for (policy
= 0; policy
<= SCTP_PR_INDEX(MAX
); policy
++) {
6445 params
.sprstat_abandoned_unsent
+=
6446 asoc
->abandoned_unsent
[policy
];
6447 params
.sprstat_abandoned_sent
+=
6448 asoc
->abandoned_sent
[policy
];
6451 params
.sprstat_abandoned_unsent
=
6452 asoc
->abandoned_unsent
[__SCTP_PR_INDEX(policy
)];
6453 params
.sprstat_abandoned_sent
=
6454 asoc
->abandoned_sent
[__SCTP_PR_INDEX(policy
)];
6457 if (put_user(len
, optlen
)) {
6462 if (copy_to_user(optval
, ¶ms
, len
)) {
6473 static int sctp_getsockopt_enable_strreset(struct sock
*sk
, int len
,
6474 char __user
*optval
,
6477 struct sctp_assoc_value params
;
6478 struct sctp_association
*asoc
;
6479 int retval
= -EFAULT
;
6481 if (len
< sizeof(params
)) {
6486 len
= sizeof(params
);
6487 if (copy_from_user(¶ms
, optval
, len
))
6490 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
6492 params
.assoc_value
= asoc
->strreset_enable
;
6493 } else if (!params
.assoc_id
) {
6494 struct sctp_sock
*sp
= sctp_sk(sk
);
6496 params
.assoc_value
= sp
->ep
->strreset_enable
;
6502 if (put_user(len
, optlen
))
6505 if (copy_to_user(optval
, ¶ms
, len
))
6514 static int sctp_getsockopt(struct sock
*sk
, int level
, int optname
,
6515 char __user
*optval
, int __user
*optlen
)
6520 pr_debug("%s: sk:%p, optname:%d\n", __func__
, sk
, optname
);
6522 /* I can hardly begin to describe how wrong this is. This is
6523 * so broken as to be worse than useless. The API draft
6524 * REALLY is NOT helpful here... I am not convinced that the
6525 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
6526 * are at all well-founded.
6528 if (level
!= SOL_SCTP
) {
6529 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
6531 retval
= af
->getsockopt(sk
, level
, optname
, optval
, optlen
);
6535 if (get_user(len
, optlen
))
6545 retval
= sctp_getsockopt_sctp_status(sk
, len
, optval
, optlen
);
6547 case SCTP_DISABLE_FRAGMENTS
:
6548 retval
= sctp_getsockopt_disable_fragments(sk
, len
, optval
,
6552 retval
= sctp_getsockopt_events(sk
, len
, optval
, optlen
);
6554 case SCTP_AUTOCLOSE
:
6555 retval
= sctp_getsockopt_autoclose(sk
, len
, optval
, optlen
);
6557 case SCTP_SOCKOPT_PEELOFF
:
6558 retval
= sctp_getsockopt_peeloff(sk
, len
, optval
, optlen
);
6560 case SCTP_PEER_ADDR_PARAMS
:
6561 retval
= sctp_getsockopt_peer_addr_params(sk
, len
, optval
,
6564 case SCTP_DELAYED_SACK
:
6565 retval
= sctp_getsockopt_delayed_ack(sk
, len
, optval
,
6569 retval
= sctp_getsockopt_initmsg(sk
, len
, optval
, optlen
);
6571 case SCTP_GET_PEER_ADDRS
:
6572 retval
= sctp_getsockopt_peer_addrs(sk
, len
, optval
,
6575 case SCTP_GET_LOCAL_ADDRS
:
6576 retval
= sctp_getsockopt_local_addrs(sk
, len
, optval
,
6579 case SCTP_SOCKOPT_CONNECTX3
:
6580 retval
= sctp_getsockopt_connectx3(sk
, len
, optval
, optlen
);
6582 case SCTP_DEFAULT_SEND_PARAM
:
6583 retval
= sctp_getsockopt_default_send_param(sk
, len
,
6586 case SCTP_DEFAULT_SNDINFO
:
6587 retval
= sctp_getsockopt_default_sndinfo(sk
, len
,
6590 case SCTP_PRIMARY_ADDR
:
6591 retval
= sctp_getsockopt_primary_addr(sk
, len
, optval
, optlen
);
6594 retval
= sctp_getsockopt_nodelay(sk
, len
, optval
, optlen
);
6597 retval
= sctp_getsockopt_rtoinfo(sk
, len
, optval
, optlen
);
6599 case SCTP_ASSOCINFO
:
6600 retval
= sctp_getsockopt_associnfo(sk
, len
, optval
, optlen
);
6602 case SCTP_I_WANT_MAPPED_V4_ADDR
:
6603 retval
= sctp_getsockopt_mappedv4(sk
, len
, optval
, optlen
);
6606 retval
= sctp_getsockopt_maxseg(sk
, len
, optval
, optlen
);
6608 case SCTP_GET_PEER_ADDR_INFO
:
6609 retval
= sctp_getsockopt_peer_addr_info(sk
, len
, optval
,
6612 case SCTP_ADAPTATION_LAYER
:
6613 retval
= sctp_getsockopt_adaptation_layer(sk
, len
, optval
,
6617 retval
= sctp_getsockopt_context(sk
, len
, optval
, optlen
);
6619 case SCTP_FRAGMENT_INTERLEAVE
:
6620 retval
= sctp_getsockopt_fragment_interleave(sk
, len
, optval
,
6623 case SCTP_PARTIAL_DELIVERY_POINT
:
6624 retval
= sctp_getsockopt_partial_delivery_point(sk
, len
, optval
,
6627 case SCTP_MAX_BURST
:
6628 retval
= sctp_getsockopt_maxburst(sk
, len
, optval
, optlen
);
6631 case SCTP_AUTH_CHUNK
:
6632 case SCTP_AUTH_DELETE_KEY
:
6633 retval
= -EOPNOTSUPP
;
6635 case SCTP_HMAC_IDENT
:
6636 retval
= sctp_getsockopt_hmac_ident(sk
, len
, optval
, optlen
);
6638 case SCTP_AUTH_ACTIVE_KEY
:
6639 retval
= sctp_getsockopt_active_key(sk
, len
, optval
, optlen
);
6641 case SCTP_PEER_AUTH_CHUNKS
:
6642 retval
= sctp_getsockopt_peer_auth_chunks(sk
, len
, optval
,
6645 case SCTP_LOCAL_AUTH_CHUNKS
:
6646 retval
= sctp_getsockopt_local_auth_chunks(sk
, len
, optval
,
6649 case SCTP_GET_ASSOC_NUMBER
:
6650 retval
= sctp_getsockopt_assoc_number(sk
, len
, optval
, optlen
);
6652 case SCTP_GET_ASSOC_ID_LIST
:
6653 retval
= sctp_getsockopt_assoc_ids(sk
, len
, optval
, optlen
);
6655 case SCTP_AUTO_ASCONF
:
6656 retval
= sctp_getsockopt_auto_asconf(sk
, len
, optval
, optlen
);
6658 case SCTP_PEER_ADDR_THLDS
:
6659 retval
= sctp_getsockopt_paddr_thresholds(sk
, optval
, len
, optlen
);
6661 case SCTP_GET_ASSOC_STATS
:
6662 retval
= sctp_getsockopt_assoc_stats(sk
, len
, optval
, optlen
);
6664 case SCTP_RECVRCVINFO
:
6665 retval
= sctp_getsockopt_recvrcvinfo(sk
, len
, optval
, optlen
);
6667 case SCTP_RECVNXTINFO
:
6668 retval
= sctp_getsockopt_recvnxtinfo(sk
, len
, optval
, optlen
);
6670 case SCTP_PR_SUPPORTED
:
6671 retval
= sctp_getsockopt_pr_supported(sk
, len
, optval
, optlen
);
6673 case SCTP_DEFAULT_PRINFO
:
6674 retval
= sctp_getsockopt_default_prinfo(sk
, len
, optval
,
6677 case SCTP_PR_ASSOC_STATUS
:
6678 retval
= sctp_getsockopt_pr_assocstatus(sk
, len
, optval
,
6681 case SCTP_ENABLE_STREAM_RESET
:
6682 retval
= sctp_getsockopt_enable_strreset(sk
, len
, optval
,
6686 retval
= -ENOPROTOOPT
;
6694 static int sctp_hash(struct sock
*sk
)
6700 static void sctp_unhash(struct sock
*sk
)
6705 /* Check if port is acceptable. Possibly find first available port.
6707 * The port hash table (contained in the 'global' SCTP protocol storage
6708 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
6709 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
6710 * list (the list number is the port number hashed out, so as you
6711 * would expect from a hash function, all the ports in a given list have
6712 * such a number that hashes out to the same list number; you were
6713 * expecting that, right?); so each list has a set of ports, with a
6714 * link to the socket (struct sock) that uses it, the port number and
6715 * a fastreuse flag (FIXME: NPI ipg).
6717 static struct sctp_bind_bucket
*sctp_bucket_create(
6718 struct sctp_bind_hashbucket
*head
, struct net
*, unsigned short snum
);
6720 static long sctp_get_port_local(struct sock
*sk
, union sctp_addr
*addr
)
6722 struct sctp_bind_hashbucket
*head
; /* hash list */
6723 struct sctp_bind_bucket
*pp
;
6724 unsigned short snum
;
6727 snum
= ntohs(addr
->v4
.sin_port
);
6729 pr_debug("%s: begins, snum:%d\n", __func__
, snum
);
6734 /* Search for an available port. */
6735 int low
, high
, remaining
, index
;
6737 struct net
*net
= sock_net(sk
);
6739 inet_get_local_port_range(net
, &low
, &high
);
6740 remaining
= (high
- low
) + 1;
6741 rover
= prandom_u32() % remaining
+ low
;
6745 if ((rover
< low
) || (rover
> high
))
6747 if (inet_is_local_reserved_port(net
, rover
))
6749 index
= sctp_phashfn(sock_net(sk
), rover
);
6750 head
= &sctp_port_hashtable
[index
];
6751 spin_lock(&head
->lock
);
6752 sctp_for_each_hentry(pp
, &head
->chain
)
6753 if ((pp
->port
== rover
) &&
6754 net_eq(sock_net(sk
), pp
->net
))
6758 spin_unlock(&head
->lock
);
6759 } while (--remaining
> 0);
6761 /* Exhausted local port range during search? */
6766 /* OK, here is the one we will use. HEAD (the port
6767 * hash table list entry) is non-NULL and we hold it's
6772 /* We are given an specific port number; we verify
6773 * that it is not being used. If it is used, we will
6774 * exahust the search in the hash list corresponding
6775 * to the port number (snum) - we detect that with the
6776 * port iterator, pp being NULL.
6778 head
= &sctp_port_hashtable
[sctp_phashfn(sock_net(sk
), snum
)];
6779 spin_lock(&head
->lock
);
6780 sctp_for_each_hentry(pp
, &head
->chain
) {
6781 if ((pp
->port
== snum
) && net_eq(pp
->net
, sock_net(sk
)))
6788 if (!hlist_empty(&pp
->owner
)) {
6789 /* We had a port hash table hit - there is an
6790 * available port (pp != NULL) and it is being
6791 * used by other socket (pp->owner not empty); that other
6792 * socket is going to be sk2.
6794 int reuse
= sk
->sk_reuse
;
6797 pr_debug("%s: found a possible match\n", __func__
);
6799 if (pp
->fastreuse
&& sk
->sk_reuse
&&
6800 sk
->sk_state
!= SCTP_SS_LISTENING
)
6803 /* Run through the list of sockets bound to the port
6804 * (pp->port) [via the pointers bind_next and
6805 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6806 * we get the endpoint they describe and run through
6807 * the endpoint's list of IP (v4 or v6) addresses,
6808 * comparing each of the addresses with the address of
6809 * the socket sk. If we find a match, then that means
6810 * that this port/socket (sk) combination are already
6813 sk_for_each_bound(sk2
, &pp
->owner
) {
6814 struct sctp_endpoint
*ep2
;
6815 ep2
= sctp_sk(sk2
)->ep
;
6818 (reuse
&& sk2
->sk_reuse
&&
6819 sk2
->sk_state
!= SCTP_SS_LISTENING
))
6822 if (sctp_bind_addr_conflict(&ep2
->base
.bind_addr
, addr
,
6823 sctp_sk(sk2
), sctp_sk(sk
))) {
6829 pr_debug("%s: found a match\n", __func__
);
6832 /* If there was a hash table miss, create a new port. */
6834 if (!pp
&& !(pp
= sctp_bucket_create(head
, sock_net(sk
), snum
)))
6837 /* In either case (hit or miss), make sure fastreuse is 1 only
6838 * if sk->sk_reuse is too (that is, if the caller requested
6839 * SO_REUSEADDR on this socket -sk-).
6841 if (hlist_empty(&pp
->owner
)) {
6842 if (sk
->sk_reuse
&& sk
->sk_state
!= SCTP_SS_LISTENING
)
6846 } else if (pp
->fastreuse
&&
6847 (!sk
->sk_reuse
|| sk
->sk_state
== SCTP_SS_LISTENING
))
6850 /* We are set, so fill up all the data in the hash table
6851 * entry, tie the socket list information with the rest of the
6852 * sockets FIXME: Blurry, NPI (ipg).
6855 if (!sctp_sk(sk
)->bind_hash
) {
6856 inet_sk(sk
)->inet_num
= snum
;
6857 sk_add_bind_node(sk
, &pp
->owner
);
6858 sctp_sk(sk
)->bind_hash
= pp
;
6863 spin_unlock(&head
->lock
);
6870 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6871 * port is requested.
6873 static int sctp_get_port(struct sock
*sk
, unsigned short snum
)
6875 union sctp_addr addr
;
6876 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
6878 /* Set up a dummy address struct from the sk. */
6879 af
->from_sk(&addr
, sk
);
6880 addr
.v4
.sin_port
= htons(snum
);
6882 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6883 return !!sctp_get_port_local(sk
, &addr
);
6887 * Move a socket to LISTENING state.
6889 static int sctp_listen_start(struct sock
*sk
, int backlog
)
6891 struct sctp_sock
*sp
= sctp_sk(sk
);
6892 struct sctp_endpoint
*ep
= sp
->ep
;
6893 struct crypto_shash
*tfm
= NULL
;
6896 /* Allocate HMAC for generating cookie. */
6897 if (!sp
->hmac
&& sp
->sctp_hmac_alg
) {
6898 sprintf(alg
, "hmac(%s)", sp
->sctp_hmac_alg
);
6899 tfm
= crypto_alloc_shash(alg
, 0, 0);
6901 net_info_ratelimited("failed to load transform for %s: %ld\n",
6902 sp
->sctp_hmac_alg
, PTR_ERR(tfm
));
6905 sctp_sk(sk
)->hmac
= tfm
;
6909 * If a bind() or sctp_bindx() is not called prior to a listen()
6910 * call that allows new associations to be accepted, the system
6911 * picks an ephemeral port and will choose an address set equivalent
6912 * to binding with a wildcard address.
6914 * This is not currently spelled out in the SCTP sockets
6915 * extensions draft, but follows the practice as seen in TCP
6919 sk
->sk_state
= SCTP_SS_LISTENING
;
6920 if (!ep
->base
.bind_addr
.port
) {
6921 if (sctp_autobind(sk
))
6924 if (sctp_get_port(sk
, inet_sk(sk
)->inet_num
)) {
6925 sk
->sk_state
= SCTP_SS_CLOSED
;
6930 sk
->sk_max_ack_backlog
= backlog
;
6931 sctp_hash_endpoint(ep
);
6936 * 4.1.3 / 5.1.3 listen()
6938 * By default, new associations are not accepted for UDP style sockets.
6939 * An application uses listen() to mark a socket as being able to
6940 * accept new associations.
6942 * On TCP style sockets, applications use listen() to ready the SCTP
6943 * endpoint for accepting inbound associations.
6945 * On both types of endpoints a backlog of '0' disables listening.
6947 * Move a socket to LISTENING state.
6949 int sctp_inet_listen(struct socket
*sock
, int backlog
)
6951 struct sock
*sk
= sock
->sk
;
6952 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
6955 if (unlikely(backlog
< 0))
6960 /* Peeled-off sockets are not allowed to listen(). */
6961 if (sctp_style(sk
, UDP_HIGH_BANDWIDTH
))
6964 if (sock
->state
!= SS_UNCONNECTED
)
6967 /* If backlog is zero, disable listening. */
6969 if (sctp_sstate(sk
, CLOSED
))
6973 sctp_unhash_endpoint(ep
);
6974 sk
->sk_state
= SCTP_SS_CLOSED
;
6976 sctp_sk(sk
)->bind_hash
->fastreuse
= 1;
6980 /* If we are already listening, just update the backlog */
6981 if (sctp_sstate(sk
, LISTENING
))
6982 sk
->sk_max_ack_backlog
= backlog
;
6984 err
= sctp_listen_start(sk
, backlog
);
6996 * This function is done by modeling the current datagram_poll() and the
6997 * tcp_poll(). Note that, based on these implementations, we don't
6998 * lock the socket in this function, even though it seems that,
6999 * ideally, locking or some other mechanisms can be used to ensure
7000 * the integrity of the counters (sndbuf and wmem_alloc) used
7001 * in this place. We assume that we don't need locks either until proven
7004 * Another thing to note is that we include the Async I/O support
7005 * here, again, by modeling the current TCP/UDP code. We don't have
7006 * a good way to test with it yet.
7008 unsigned int sctp_poll(struct file
*file
, struct socket
*sock
, poll_table
*wait
)
7010 struct sock
*sk
= sock
->sk
;
7011 struct sctp_sock
*sp
= sctp_sk(sk
);
7014 poll_wait(file
, sk_sleep(sk
), wait
);
7016 sock_rps_record_flow(sk
);
7018 /* A TCP-style listening socket becomes readable when the accept queue
7021 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
7022 return (!list_empty(&sp
->ep
->asocs
)) ?
7023 (POLLIN
| POLLRDNORM
) : 0;
7027 /* Is there any exceptional events? */
7028 if (sk
->sk_err
|| !skb_queue_empty(&sk
->sk_error_queue
))
7030 (sock_flag(sk
, SOCK_SELECT_ERR_QUEUE
) ? POLLPRI
: 0);
7031 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
7032 mask
|= POLLRDHUP
| POLLIN
| POLLRDNORM
;
7033 if (sk
->sk_shutdown
== SHUTDOWN_MASK
)
7036 /* Is it readable? Reconsider this code with TCP-style support. */
7037 if (!skb_queue_empty(&sk
->sk_receive_queue
))
7038 mask
|= POLLIN
| POLLRDNORM
;
7040 /* The association is either gone or not ready. */
7041 if (!sctp_style(sk
, UDP
) && sctp_sstate(sk
, CLOSED
))
7044 /* Is it writable? */
7045 if (sctp_writeable(sk
)) {
7046 mask
|= POLLOUT
| POLLWRNORM
;
7048 sk_set_bit(SOCKWQ_ASYNC_NOSPACE
, sk
);
7050 * Since the socket is not locked, the buffer
7051 * might be made available after the writeable check and
7052 * before the bit is set. This could cause a lost I/O
7053 * signal. tcp_poll() has a race breaker for this race
7054 * condition. Based on their implementation, we put
7055 * in the following code to cover it as well.
7057 if (sctp_writeable(sk
))
7058 mask
|= POLLOUT
| POLLWRNORM
;
7063 /********************************************************************
7064 * 2nd Level Abstractions
7065 ********************************************************************/
7067 static struct sctp_bind_bucket
*sctp_bucket_create(
7068 struct sctp_bind_hashbucket
*head
, struct net
*net
, unsigned short snum
)
7070 struct sctp_bind_bucket
*pp
;
7072 pp
= kmem_cache_alloc(sctp_bucket_cachep
, GFP_ATOMIC
);
7074 SCTP_DBG_OBJCNT_INC(bind_bucket
);
7077 INIT_HLIST_HEAD(&pp
->owner
);
7079 hlist_add_head(&pp
->node
, &head
->chain
);
7084 /* Caller must hold hashbucket lock for this tb with local BH disabled */
7085 static void sctp_bucket_destroy(struct sctp_bind_bucket
*pp
)
7087 if (pp
&& hlist_empty(&pp
->owner
)) {
7088 __hlist_del(&pp
->node
);
7089 kmem_cache_free(sctp_bucket_cachep
, pp
);
7090 SCTP_DBG_OBJCNT_DEC(bind_bucket
);
7094 /* Release this socket's reference to a local port. */
7095 static inline void __sctp_put_port(struct sock
*sk
)
7097 struct sctp_bind_hashbucket
*head
=
7098 &sctp_port_hashtable
[sctp_phashfn(sock_net(sk
),
7099 inet_sk(sk
)->inet_num
)];
7100 struct sctp_bind_bucket
*pp
;
7102 spin_lock(&head
->lock
);
7103 pp
= sctp_sk(sk
)->bind_hash
;
7104 __sk_del_bind_node(sk
);
7105 sctp_sk(sk
)->bind_hash
= NULL
;
7106 inet_sk(sk
)->inet_num
= 0;
7107 sctp_bucket_destroy(pp
);
7108 spin_unlock(&head
->lock
);
7111 void sctp_put_port(struct sock
*sk
)
7114 __sctp_put_port(sk
);
7119 * The system picks an ephemeral port and choose an address set equivalent
7120 * to binding with a wildcard address.
7121 * One of those addresses will be the primary address for the association.
7122 * This automatically enables the multihoming capability of SCTP.
7124 static int sctp_autobind(struct sock
*sk
)
7126 union sctp_addr autoaddr
;
7130 /* Initialize a local sockaddr structure to INADDR_ANY. */
7131 af
= sctp_sk(sk
)->pf
->af
;
7133 port
= htons(inet_sk(sk
)->inet_num
);
7134 af
->inaddr_any(&autoaddr
, port
);
7136 return sctp_do_bind(sk
, &autoaddr
, af
->sockaddr_len
);
7139 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
7142 * 4.2 The cmsghdr Structure *
7144 * When ancillary data is sent or received, any number of ancillary data
7145 * objects can be specified by the msg_control and msg_controllen members of
7146 * the msghdr structure, because each object is preceded by
7147 * a cmsghdr structure defining the object's length (the cmsg_len member).
7148 * Historically Berkeley-derived implementations have passed only one object
7149 * at a time, but this API allows multiple objects to be
7150 * passed in a single call to sendmsg() or recvmsg(). The following example
7151 * shows two ancillary data objects in a control buffer.
7153 * |<--------------------------- msg_controllen -------------------------->|
7156 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
7158 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
7161 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
7163 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
7166 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7167 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
7169 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
7171 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7178 static int sctp_msghdr_parse(const struct msghdr
*msg
, sctp_cmsgs_t
*cmsgs
)
7180 struct cmsghdr
*cmsg
;
7181 struct msghdr
*my_msg
= (struct msghdr
*)msg
;
7183 for_each_cmsghdr(cmsg
, my_msg
) {
7184 if (!CMSG_OK(my_msg
, cmsg
))
7187 /* Should we parse this header or ignore? */
7188 if (cmsg
->cmsg_level
!= IPPROTO_SCTP
)
7191 /* Strictly check lengths following example in SCM code. */
7192 switch (cmsg
->cmsg_type
) {
7194 /* SCTP Socket API Extension
7195 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
7197 * This cmsghdr structure provides information for
7198 * initializing new SCTP associations with sendmsg().
7199 * The SCTP_INITMSG socket option uses this same data
7200 * structure. This structure is not used for
7203 * cmsg_level cmsg_type cmsg_data[]
7204 * ------------ ------------ ----------------------
7205 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
7207 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(struct sctp_initmsg
)))
7210 cmsgs
->init
= CMSG_DATA(cmsg
);
7214 /* SCTP Socket API Extension
7215 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
7217 * This cmsghdr structure specifies SCTP options for
7218 * sendmsg() and describes SCTP header information
7219 * about a received message through recvmsg().
7221 * cmsg_level cmsg_type cmsg_data[]
7222 * ------------ ------------ ----------------------
7223 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
7225 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(struct sctp_sndrcvinfo
)))
7228 cmsgs
->srinfo
= CMSG_DATA(cmsg
);
7230 if (cmsgs
->srinfo
->sinfo_flags
&
7231 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
7232 SCTP_SACK_IMMEDIATELY
| SCTP_PR_SCTP_MASK
|
7233 SCTP_ABORT
| SCTP_EOF
))
7238 /* SCTP Socket API Extension
7239 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
7241 * This cmsghdr structure specifies SCTP options for
7242 * sendmsg(). This structure and SCTP_RCVINFO replaces
7243 * SCTP_SNDRCV which has been deprecated.
7245 * cmsg_level cmsg_type cmsg_data[]
7246 * ------------ ------------ ---------------------
7247 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
7249 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(struct sctp_sndinfo
)))
7252 cmsgs
->sinfo
= CMSG_DATA(cmsg
);
7254 if (cmsgs
->sinfo
->snd_flags
&
7255 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
7256 SCTP_SACK_IMMEDIATELY
| SCTP_PR_SCTP_MASK
|
7257 SCTP_ABORT
| SCTP_EOF
))
7269 * Wait for a packet..
7270 * Note: This function is the same function as in core/datagram.c
7271 * with a few modifications to make lksctp work.
7273 static int sctp_wait_for_packet(struct sock
*sk
, int *err
, long *timeo_p
)
7278 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
7280 /* Socket errors? */
7281 error
= sock_error(sk
);
7285 if (!skb_queue_empty(&sk
->sk_receive_queue
))
7288 /* Socket shut down? */
7289 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
7292 /* Sequenced packets can come disconnected. If so we report the
7297 /* Is there a good reason to think that we may receive some data? */
7298 if (list_empty(&sctp_sk(sk
)->ep
->asocs
) && !sctp_sstate(sk
, LISTENING
))
7301 /* Handle signals. */
7302 if (signal_pending(current
))
7305 /* Let another process have a go. Since we are going to sleep
7306 * anyway. Note: This may cause odd behaviors if the message
7307 * does not fit in the user's buffer, but this seems to be the
7308 * only way to honor MSG_DONTWAIT realistically.
7311 *timeo_p
= schedule_timeout(*timeo_p
);
7315 finish_wait(sk_sleep(sk
), &wait
);
7319 error
= sock_intr_errno(*timeo_p
);
7322 finish_wait(sk_sleep(sk
), &wait
);
7327 /* Receive a datagram.
7328 * Note: This is pretty much the same routine as in core/datagram.c
7329 * with a few changes to make lksctp work.
7331 struct sk_buff
*sctp_skb_recv_datagram(struct sock
*sk
, int flags
,
7332 int noblock
, int *err
)
7335 struct sk_buff
*skb
;
7338 timeo
= sock_rcvtimeo(sk
, noblock
);
7340 pr_debug("%s: timeo:%ld, max:%ld\n", __func__
, timeo
,
7341 MAX_SCHEDULE_TIMEOUT
);
7344 /* Again only user level code calls this function,
7345 * so nothing interrupt level
7346 * will suddenly eat the receive_queue.
7348 * Look at current nfs client by the way...
7349 * However, this function was correct in any case. 8)
7351 if (flags
& MSG_PEEK
) {
7352 skb
= skb_peek(&sk
->sk_receive_queue
);
7354 atomic_inc(&skb
->users
);
7356 skb
= __skb_dequeue(&sk
->sk_receive_queue
);
7362 /* Caller is allowed not to check sk->sk_err before calling. */
7363 error
= sock_error(sk
);
7367 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
7370 if (sk_can_busy_loop(sk
) &&
7371 sk_busy_loop(sk
, noblock
))
7374 /* User doesn't want to wait. */
7378 } while (sctp_wait_for_packet(sk
, err
, &timeo
) == 0);
7387 /* If sndbuf has changed, wake up per association sndbuf waiters. */
7388 static void __sctp_write_space(struct sctp_association
*asoc
)
7390 struct sock
*sk
= asoc
->base
.sk
;
7392 if (sctp_wspace(asoc
) <= 0)
7395 if (waitqueue_active(&asoc
->wait
))
7396 wake_up_interruptible(&asoc
->wait
);
7398 if (sctp_writeable(sk
)) {
7399 struct socket_wq
*wq
;
7402 wq
= rcu_dereference(sk
->sk_wq
);
7404 if (waitqueue_active(&wq
->wait
))
7405 wake_up_interruptible(&wq
->wait
);
7407 /* Note that we try to include the Async I/O support
7408 * here by modeling from the current TCP/UDP code.
7409 * We have not tested with it yet.
7411 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
))
7412 sock_wake_async(wq
, SOCK_WAKE_SPACE
, POLL_OUT
);
7418 static void sctp_wake_up_waiters(struct sock
*sk
,
7419 struct sctp_association
*asoc
)
7421 struct sctp_association
*tmp
= asoc
;
7423 /* We do accounting for the sndbuf space per association,
7424 * so we only need to wake our own association.
7426 if (asoc
->ep
->sndbuf_policy
)
7427 return __sctp_write_space(asoc
);
7429 /* If association goes down and is just flushing its
7430 * outq, then just normally notify others.
7432 if (asoc
->base
.dead
)
7433 return sctp_write_space(sk
);
7435 /* Accounting for the sndbuf space is per socket, so we
7436 * need to wake up others, try to be fair and in case of
7437 * other associations, let them have a go first instead
7438 * of just doing a sctp_write_space() call.
7440 * Note that we reach sctp_wake_up_waiters() only when
7441 * associations free up queued chunks, thus we are under
7442 * lock and the list of associations on a socket is
7443 * guaranteed not to change.
7445 for (tmp
= list_next_entry(tmp
, asocs
); 1;
7446 tmp
= list_next_entry(tmp
, asocs
)) {
7447 /* Manually skip the head element. */
7448 if (&tmp
->asocs
== &((sctp_sk(sk
))->ep
->asocs
))
7450 /* Wake up association. */
7451 __sctp_write_space(tmp
);
7452 /* We've reached the end. */
7458 /* Do accounting for the sndbuf space.
7459 * Decrement the used sndbuf space of the corresponding association by the
7460 * data size which was just transmitted(freed).
7462 static void sctp_wfree(struct sk_buff
*skb
)
7464 struct sctp_chunk
*chunk
= skb_shinfo(skb
)->destructor_arg
;
7465 struct sctp_association
*asoc
= chunk
->asoc
;
7466 struct sock
*sk
= asoc
->base
.sk
;
7468 asoc
->sndbuf_used
-= SCTP_DATA_SNDSIZE(chunk
) +
7469 sizeof(struct sk_buff
) +
7470 sizeof(struct sctp_chunk
);
7472 atomic_sub(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
7475 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
7477 sk
->sk_wmem_queued
-= skb
->truesize
;
7478 sk_mem_uncharge(sk
, skb
->truesize
);
7481 sctp_wake_up_waiters(sk
, asoc
);
7483 sctp_association_put(asoc
);
7486 /* Do accounting for the receive space on the socket.
7487 * Accounting for the association is done in ulpevent.c
7488 * We set this as a destructor for the cloned data skbs so that
7489 * accounting is done at the correct time.
7491 void sctp_sock_rfree(struct sk_buff
*skb
)
7493 struct sock
*sk
= skb
->sk
;
7494 struct sctp_ulpevent
*event
= sctp_skb2event(skb
);
7496 atomic_sub(event
->rmem_len
, &sk
->sk_rmem_alloc
);
7499 * Mimic the behavior of sock_rfree
7501 sk_mem_uncharge(sk
, event
->rmem_len
);
7505 /* Helper function to wait for space in the sndbuf. */
7506 static int sctp_wait_for_sndbuf(struct sctp_association
*asoc
, long *timeo_p
,
7509 struct sock
*sk
= asoc
->base
.sk
;
7511 long current_timeo
= *timeo_p
;
7514 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__
, asoc
,
7517 /* Increment the association's refcnt. */
7518 sctp_association_hold(asoc
);
7520 /* Wait on the association specific sndbuf space. */
7522 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
7523 TASK_INTERRUPTIBLE
);
7526 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
7529 if (signal_pending(current
))
7530 goto do_interrupted
;
7531 if (msg_len
<= sctp_wspace(asoc
))
7534 /* Let another process have a go. Since we are going
7538 current_timeo
= schedule_timeout(current_timeo
);
7539 BUG_ON(sk
!= asoc
->base
.sk
);
7542 *timeo_p
= current_timeo
;
7546 finish_wait(&asoc
->wait
, &wait
);
7548 /* Release the association's refcnt. */
7549 sctp_association_put(asoc
);
7558 err
= sock_intr_errno(*timeo_p
);
7566 void sctp_data_ready(struct sock
*sk
)
7568 struct socket_wq
*wq
;
7571 wq
= rcu_dereference(sk
->sk_wq
);
7572 if (skwq_has_sleeper(wq
))
7573 wake_up_interruptible_sync_poll(&wq
->wait
, POLLIN
|
7574 POLLRDNORM
| POLLRDBAND
);
7575 sk_wake_async(sk
, SOCK_WAKE_WAITD
, POLL_IN
);
7579 /* If socket sndbuf has changed, wake up all per association waiters. */
7580 void sctp_write_space(struct sock
*sk
)
7582 struct sctp_association
*asoc
;
7584 /* Wake up the tasks in each wait queue. */
7585 list_for_each_entry(asoc
, &((sctp_sk(sk
))->ep
->asocs
), asocs
) {
7586 __sctp_write_space(asoc
);
7590 /* Is there any sndbuf space available on the socket?
7592 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
7593 * associations on the same socket. For a UDP-style socket with
7594 * multiple associations, it is possible for it to be "unwriteable"
7595 * prematurely. I assume that this is acceptable because
7596 * a premature "unwriteable" is better than an accidental "writeable" which
7597 * would cause an unwanted block under certain circumstances. For the 1-1
7598 * UDP-style sockets or TCP-style sockets, this code should work.
7601 static int sctp_writeable(struct sock
*sk
)
7605 amt
= sk
->sk_sndbuf
- sk_wmem_alloc_get(sk
);
7611 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
7612 * returns immediately with EINPROGRESS.
7614 static int sctp_wait_for_connect(struct sctp_association
*asoc
, long *timeo_p
)
7616 struct sock
*sk
= asoc
->base
.sk
;
7618 long current_timeo
= *timeo_p
;
7621 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__
, asoc
, *timeo_p
);
7623 /* Increment the association's refcnt. */
7624 sctp_association_hold(asoc
);
7627 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
7628 TASK_INTERRUPTIBLE
);
7631 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
7633 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
7636 if (signal_pending(current
))
7637 goto do_interrupted
;
7639 if (sctp_state(asoc
, ESTABLISHED
))
7642 /* Let another process have a go. Since we are going
7646 current_timeo
= schedule_timeout(current_timeo
);
7649 *timeo_p
= current_timeo
;
7653 finish_wait(&asoc
->wait
, &wait
);
7655 /* Release the association's refcnt. */
7656 sctp_association_put(asoc
);
7661 if (asoc
->init_err_counter
+ 1 > asoc
->max_init_attempts
)
7664 err
= -ECONNREFUSED
;
7668 err
= sock_intr_errno(*timeo_p
);
7676 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
)
7678 struct sctp_endpoint
*ep
;
7682 ep
= sctp_sk(sk
)->ep
;
7686 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
,
7687 TASK_INTERRUPTIBLE
);
7689 if (list_empty(&ep
->asocs
)) {
7691 timeo
= schedule_timeout(timeo
);
7696 if (!sctp_sstate(sk
, LISTENING
))
7700 if (!list_empty(&ep
->asocs
))
7703 err
= sock_intr_errno(timeo
);
7704 if (signal_pending(current
))
7712 finish_wait(sk_sleep(sk
), &wait
);
7717 static void sctp_wait_for_close(struct sock
*sk
, long timeout
)
7722 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
7723 if (list_empty(&sctp_sk(sk
)->ep
->asocs
))
7726 timeout
= schedule_timeout(timeout
);
7728 } while (!signal_pending(current
) && timeout
);
7730 finish_wait(sk_sleep(sk
), &wait
);
7733 static void sctp_skb_set_owner_r_frag(struct sk_buff
*skb
, struct sock
*sk
)
7735 struct sk_buff
*frag
;
7740 /* Don't forget the fragments. */
7741 skb_walk_frags(skb
, frag
)
7742 sctp_skb_set_owner_r_frag(frag
, sk
);
7745 sctp_skb_set_owner_r(skb
, sk
);
7748 void sctp_copy_sock(struct sock
*newsk
, struct sock
*sk
,
7749 struct sctp_association
*asoc
)
7751 struct inet_sock
*inet
= inet_sk(sk
);
7752 struct inet_sock
*newinet
;
7754 newsk
->sk_type
= sk
->sk_type
;
7755 newsk
->sk_bound_dev_if
= sk
->sk_bound_dev_if
;
7756 newsk
->sk_flags
= sk
->sk_flags
;
7757 newsk
->sk_tsflags
= sk
->sk_tsflags
;
7758 newsk
->sk_no_check_tx
= sk
->sk_no_check_tx
;
7759 newsk
->sk_no_check_rx
= sk
->sk_no_check_rx
;
7760 newsk
->sk_reuse
= sk
->sk_reuse
;
7762 newsk
->sk_shutdown
= sk
->sk_shutdown
;
7763 newsk
->sk_destruct
= sctp_destruct_sock
;
7764 newsk
->sk_family
= sk
->sk_family
;
7765 newsk
->sk_protocol
= IPPROTO_SCTP
;
7766 newsk
->sk_backlog_rcv
= sk
->sk_prot
->backlog_rcv
;
7767 newsk
->sk_sndbuf
= sk
->sk_sndbuf
;
7768 newsk
->sk_rcvbuf
= sk
->sk_rcvbuf
;
7769 newsk
->sk_lingertime
= sk
->sk_lingertime
;
7770 newsk
->sk_rcvtimeo
= sk
->sk_rcvtimeo
;
7771 newsk
->sk_sndtimeo
= sk
->sk_sndtimeo
;
7772 newsk
->sk_rxhash
= sk
->sk_rxhash
;
7774 newinet
= inet_sk(newsk
);
7776 /* Initialize sk's sport, dport, rcv_saddr and daddr for
7777 * getsockname() and getpeername()
7779 newinet
->inet_sport
= inet
->inet_sport
;
7780 newinet
->inet_saddr
= inet
->inet_saddr
;
7781 newinet
->inet_rcv_saddr
= inet
->inet_rcv_saddr
;
7782 newinet
->inet_dport
= htons(asoc
->peer
.port
);
7783 newinet
->pmtudisc
= inet
->pmtudisc
;
7784 newinet
->inet_id
= asoc
->next_tsn
^ jiffies
;
7786 newinet
->uc_ttl
= inet
->uc_ttl
;
7787 newinet
->mc_loop
= 1;
7788 newinet
->mc_ttl
= 1;
7789 newinet
->mc_index
= 0;
7790 newinet
->mc_list
= NULL
;
7792 if (newsk
->sk_flags
& SK_FLAGS_TIMESTAMP
)
7793 net_enable_timestamp();
7795 security_sk_clone(sk
, newsk
);
7798 static inline void sctp_copy_descendant(struct sock
*sk_to
,
7799 const struct sock
*sk_from
)
7801 int ancestor_size
= sizeof(struct inet_sock
) +
7802 sizeof(struct sctp_sock
) -
7803 offsetof(struct sctp_sock
, auto_asconf_list
);
7805 if (sk_from
->sk_family
== PF_INET6
)
7806 ancestor_size
+= sizeof(struct ipv6_pinfo
);
7808 __inet_sk_copy_descendant(sk_to
, sk_from
, ancestor_size
);
7811 /* Populate the fields of the newsk from the oldsk and migrate the assoc
7812 * and its messages to the newsk.
7814 static void sctp_sock_migrate(struct sock
*oldsk
, struct sock
*newsk
,
7815 struct sctp_association
*assoc
,
7816 sctp_socket_type_t type
)
7818 struct sctp_sock
*oldsp
= sctp_sk(oldsk
);
7819 struct sctp_sock
*newsp
= sctp_sk(newsk
);
7820 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
7821 struct sctp_endpoint
*newep
= newsp
->ep
;
7822 struct sk_buff
*skb
, *tmp
;
7823 struct sctp_ulpevent
*event
;
7824 struct sctp_bind_hashbucket
*head
;
7826 /* Migrate socket buffer sizes and all the socket level options to the
7829 newsk
->sk_sndbuf
= oldsk
->sk_sndbuf
;
7830 newsk
->sk_rcvbuf
= oldsk
->sk_rcvbuf
;
7831 /* Brute force copy old sctp opt. */
7832 sctp_copy_descendant(newsk
, oldsk
);
7834 /* Restore the ep value that was overwritten with the above structure
7840 /* Hook this new socket in to the bind_hash list. */
7841 head
= &sctp_port_hashtable
[sctp_phashfn(sock_net(oldsk
),
7842 inet_sk(oldsk
)->inet_num
)];
7843 spin_lock_bh(&head
->lock
);
7844 pp
= sctp_sk(oldsk
)->bind_hash
;
7845 sk_add_bind_node(newsk
, &pp
->owner
);
7846 sctp_sk(newsk
)->bind_hash
= pp
;
7847 inet_sk(newsk
)->inet_num
= inet_sk(oldsk
)->inet_num
;
7848 spin_unlock_bh(&head
->lock
);
7850 /* Copy the bind_addr list from the original endpoint to the new
7851 * endpoint so that we can handle restarts properly
7853 sctp_bind_addr_dup(&newsp
->ep
->base
.bind_addr
,
7854 &oldsp
->ep
->base
.bind_addr
, GFP_KERNEL
);
7856 /* Move any messages in the old socket's receive queue that are for the
7857 * peeled off association to the new socket's receive queue.
7859 sctp_skb_for_each(skb
, &oldsk
->sk_receive_queue
, tmp
) {
7860 event
= sctp_skb2event(skb
);
7861 if (event
->asoc
== assoc
) {
7862 __skb_unlink(skb
, &oldsk
->sk_receive_queue
);
7863 __skb_queue_tail(&newsk
->sk_receive_queue
, skb
);
7864 sctp_skb_set_owner_r_frag(skb
, newsk
);
7868 /* Clean up any messages pending delivery due to partial
7869 * delivery. Three cases:
7870 * 1) No partial deliver; no work.
7871 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7872 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7874 skb_queue_head_init(&newsp
->pd_lobby
);
7875 atomic_set(&sctp_sk(newsk
)->pd_mode
, assoc
->ulpq
.pd_mode
);
7877 if (atomic_read(&sctp_sk(oldsk
)->pd_mode
)) {
7878 struct sk_buff_head
*queue
;
7880 /* Decide which queue to move pd_lobby skbs to. */
7881 if (assoc
->ulpq
.pd_mode
) {
7882 queue
= &newsp
->pd_lobby
;
7884 queue
= &newsk
->sk_receive_queue
;
7886 /* Walk through the pd_lobby, looking for skbs that
7887 * need moved to the new socket.
7889 sctp_skb_for_each(skb
, &oldsp
->pd_lobby
, tmp
) {
7890 event
= sctp_skb2event(skb
);
7891 if (event
->asoc
== assoc
) {
7892 __skb_unlink(skb
, &oldsp
->pd_lobby
);
7893 __skb_queue_tail(queue
, skb
);
7894 sctp_skb_set_owner_r_frag(skb
, newsk
);
7898 /* Clear up any skbs waiting for the partial
7899 * delivery to finish.
7901 if (assoc
->ulpq
.pd_mode
)
7902 sctp_clear_pd(oldsk
, NULL
);
7906 sctp_skb_for_each(skb
, &assoc
->ulpq
.reasm
, tmp
)
7907 sctp_skb_set_owner_r_frag(skb
, newsk
);
7909 sctp_skb_for_each(skb
, &assoc
->ulpq
.lobby
, tmp
)
7910 sctp_skb_set_owner_r_frag(skb
, newsk
);
7912 /* Set the type of socket to indicate that it is peeled off from the
7913 * original UDP-style socket or created with the accept() call on a
7914 * TCP-style socket..
7918 /* Mark the new socket "in-use" by the user so that any packets
7919 * that may arrive on the association after we've moved it are
7920 * queued to the backlog. This prevents a potential race between
7921 * backlog processing on the old socket and new-packet processing
7922 * on the new socket.
7924 * The caller has just allocated newsk so we can guarantee that other
7925 * paths won't try to lock it and then oldsk.
7927 lock_sock_nested(newsk
, SINGLE_DEPTH_NESTING
);
7928 sctp_assoc_migrate(assoc
, newsk
);
7930 /* If the association on the newsk is already closed before accept()
7931 * is called, set RCV_SHUTDOWN flag.
7933 if (sctp_state(assoc
, CLOSED
) && sctp_style(newsk
, TCP
)) {
7934 newsk
->sk_state
= SCTP_SS_CLOSED
;
7935 newsk
->sk_shutdown
|= RCV_SHUTDOWN
;
7937 newsk
->sk_state
= SCTP_SS_ESTABLISHED
;
7940 release_sock(newsk
);
7944 /* This proto struct describes the ULP interface for SCTP. */
7945 struct proto sctp_prot
= {
7947 .owner
= THIS_MODULE
,
7948 .close
= sctp_close
,
7949 .connect
= sctp_connect
,
7950 .disconnect
= sctp_disconnect
,
7951 .accept
= sctp_accept
,
7952 .ioctl
= sctp_ioctl
,
7953 .init
= sctp_init_sock
,
7954 .destroy
= sctp_destroy_sock
,
7955 .shutdown
= sctp_shutdown
,
7956 .setsockopt
= sctp_setsockopt
,
7957 .getsockopt
= sctp_getsockopt
,
7958 .sendmsg
= sctp_sendmsg
,
7959 .recvmsg
= sctp_recvmsg
,
7961 .backlog_rcv
= sctp_backlog_rcv
,
7963 .unhash
= sctp_unhash
,
7964 .get_port
= sctp_get_port
,
7965 .obj_size
= sizeof(struct sctp_sock
),
7966 .sysctl_mem
= sysctl_sctp_mem
,
7967 .sysctl_rmem
= sysctl_sctp_rmem
,
7968 .sysctl_wmem
= sysctl_sctp_wmem
,
7969 .memory_pressure
= &sctp_memory_pressure
,
7970 .enter_memory_pressure
= sctp_enter_memory_pressure
,
7971 .memory_allocated
= &sctp_memory_allocated
,
7972 .sockets_allocated
= &sctp_sockets_allocated
,
7975 #if IS_ENABLED(CONFIG_IPV6)
7977 #include <net/transp_v6.h>
7978 static void sctp_v6_destroy_sock(struct sock
*sk
)
7980 sctp_destroy_sock(sk
);
7981 inet6_destroy_sock(sk
);
7984 struct proto sctpv6_prot
= {
7986 .owner
= THIS_MODULE
,
7987 .close
= sctp_close
,
7988 .connect
= sctp_connect
,
7989 .disconnect
= sctp_disconnect
,
7990 .accept
= sctp_accept
,
7991 .ioctl
= sctp_ioctl
,
7992 .init
= sctp_init_sock
,
7993 .destroy
= sctp_v6_destroy_sock
,
7994 .shutdown
= sctp_shutdown
,
7995 .setsockopt
= sctp_setsockopt
,
7996 .getsockopt
= sctp_getsockopt
,
7997 .sendmsg
= sctp_sendmsg
,
7998 .recvmsg
= sctp_recvmsg
,
8000 .backlog_rcv
= sctp_backlog_rcv
,
8002 .unhash
= sctp_unhash
,
8003 .get_port
= sctp_get_port
,
8004 .obj_size
= sizeof(struct sctp6_sock
),
8005 .sysctl_mem
= sysctl_sctp_mem
,
8006 .sysctl_rmem
= sysctl_sctp_rmem
,
8007 .sysctl_wmem
= sysctl_sctp_wmem
,
8008 .memory_pressure
= &sctp_memory_pressure
,
8009 .enter_memory_pressure
= sctp_enter_memory_pressure
,
8010 .memory_allocated
= &sctp_memory_allocated
,
8011 .sockets_allocated
= &sctp_sockets_allocated
,
8013 #endif /* IS_ENABLED(CONFIG_IPV6) */