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
< PROT_SOCK
&&
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
< PROT_SOCK
&&
1156 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
)) {
1162 scope
= sctp_scope(&to
);
1163 asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1169 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
,
1177 /* Prime the peer's transport structures. */
1178 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
,
1186 addr_buf
+= af
->sockaddr_len
;
1187 walk_size
+= af
->sockaddr_len
;
1190 /* In case the user of sctp_connectx() wants an association
1191 * id back, assign one now.
1194 err
= sctp_assoc_set_id(asoc
, GFP_KERNEL
);
1199 err
= sctp_primitive_ASSOCIATE(net
, asoc
, NULL
);
1204 /* Initialize sk's dport and daddr for getpeername() */
1205 inet_sk(sk
)->inet_dport
= htons(asoc
->peer
.port
);
1206 sp
->pf
->to_sk_daddr(sa_addr
, sk
);
1209 /* in-kernel sockets don't generally have a file allocated to them
1210 * if all they do is call sock_create_kern().
1212 if (sk
->sk_socket
->file
)
1213 f_flags
= sk
->sk_socket
->file
->f_flags
;
1215 timeo
= sock_sndtimeo(sk
, f_flags
& O_NONBLOCK
);
1217 err
= sctp_wait_for_connect(asoc
, &timeo
);
1218 if ((err
== 0 || err
== -EINPROGRESS
) && assoc_id
)
1219 *assoc_id
= asoc
->assoc_id
;
1221 /* Don't free association on exit. */
1225 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1226 __func__
, asoc
, kaddrs
, err
);
1229 /* sctp_primitive_ASSOCIATE may have added this association
1230 * To the hash table, try to unhash it, just in case, its a noop
1231 * if it wasn't hashed so we're safe
1233 sctp_association_free(asoc
);
1238 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1241 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1242 * sctp_assoc_t *asoc);
1244 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1245 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1246 * or IPv6 addresses.
1248 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1249 * Section 3.1.2 for this usage.
1251 * addrs is a pointer to an array of one or more socket addresses. Each
1252 * address is contained in its appropriate structure (i.e. struct
1253 * sockaddr_in or struct sockaddr_in6) the family of the address type
1254 * must be used to distengish the address length (note that this
1255 * representation is termed a "packed array" of addresses). The caller
1256 * specifies the number of addresses in the array with addrcnt.
1258 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1259 * the association id of the new association. On failure, sctp_connectx()
1260 * returns -1, and sets errno to the appropriate error code. The assoc_id
1261 * is not touched by the kernel.
1263 * For SCTP, the port given in each socket address must be the same, or
1264 * sctp_connectx() will fail, setting errno to EINVAL.
1266 * An application can use sctp_connectx to initiate an association with
1267 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1268 * allows a caller to specify multiple addresses at which a peer can be
1269 * reached. The way the SCTP stack uses the list of addresses to set up
1270 * the association is implementation dependent. This function only
1271 * specifies that the stack will try to make use of all the addresses in
1272 * the list when needed.
1274 * Note that the list of addresses passed in is only used for setting up
1275 * the association. It does not necessarily equal the set of addresses
1276 * the peer uses for the resulting association. If the caller wants to
1277 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1278 * retrieve them after the association has been set up.
1280 * Basically do nothing but copying the addresses from user to kernel
1281 * land and invoking either sctp_connectx(). This is used for tunneling
1282 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1284 * We don't use copy_from_user() for optimization: we first do the
1285 * sanity checks (buffer size -fast- and access check-healthy
1286 * pointer); if all of those succeed, then we can alloc the memory
1287 * (expensive operation) needed to copy the data to kernel. Then we do
1288 * the copying without checking the user space area
1289 * (__copy_from_user()).
1291 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1294 * sk The sk of the socket
1295 * addrs The pointer to the addresses in user land
1296 * addrssize Size of the addrs buffer
1298 * Returns >=0 if ok, <0 errno code on error.
1300 static int __sctp_setsockopt_connectx(struct sock
*sk
,
1301 struct sockaddr __user
*addrs
,
1303 sctp_assoc_t
*assoc_id
)
1305 struct sockaddr
*kaddrs
;
1306 gfp_t gfp
= GFP_KERNEL
;
1309 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1310 __func__
, sk
, addrs
, addrs_size
);
1312 if (unlikely(addrs_size
<= 0))
1315 /* Check the user passed a healthy pointer. */
1316 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
1319 /* Alloc space for the address array in kernel memory. */
1320 if (sk
->sk_socket
->file
)
1321 gfp
= GFP_USER
| __GFP_NOWARN
;
1322 kaddrs
= kmalloc(addrs_size
, gfp
);
1323 if (unlikely(!kaddrs
))
1326 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
1329 err
= __sctp_connect(sk
, kaddrs
, addrs_size
, assoc_id
);
1338 * This is an older interface. It's kept for backward compatibility
1339 * to the option that doesn't provide association id.
1341 static int sctp_setsockopt_connectx_old(struct sock
*sk
,
1342 struct sockaddr __user
*addrs
,
1345 return __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, NULL
);
1349 * New interface for the API. The since the API is done with a socket
1350 * option, to make it simple we feed back the association id is as a return
1351 * indication to the call. Error is always negative and association id is
1354 static int sctp_setsockopt_connectx(struct sock
*sk
,
1355 struct sockaddr __user
*addrs
,
1358 sctp_assoc_t assoc_id
= 0;
1361 err
= __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, &assoc_id
);
1370 * New (hopefully final) interface for the API.
1371 * We use the sctp_getaddrs_old structure so that use-space library
1372 * can avoid any unnecessary allocations. The only different part
1373 * is that we store the actual length of the address buffer into the
1374 * addrs_num structure member. That way we can re-use the existing
1377 #ifdef CONFIG_COMPAT
1378 struct compat_sctp_getaddrs_old
{
1379 sctp_assoc_t assoc_id
;
1381 compat_uptr_t addrs
; /* struct sockaddr * */
1385 static int sctp_getsockopt_connectx3(struct sock
*sk
, int len
,
1386 char __user
*optval
,
1389 struct sctp_getaddrs_old param
;
1390 sctp_assoc_t assoc_id
= 0;
1393 #ifdef CONFIG_COMPAT
1394 if (in_compat_syscall()) {
1395 struct compat_sctp_getaddrs_old param32
;
1397 if (len
< sizeof(param32
))
1399 if (copy_from_user(¶m32
, optval
, sizeof(param32
)))
1402 param
.assoc_id
= param32
.assoc_id
;
1403 param
.addr_num
= param32
.addr_num
;
1404 param
.addrs
= compat_ptr(param32
.addrs
);
1408 if (len
< sizeof(param
))
1410 if (copy_from_user(¶m
, optval
, sizeof(param
)))
1414 err
= __sctp_setsockopt_connectx(sk
, (struct sockaddr __user
*)
1415 param
.addrs
, param
.addr_num
,
1417 if (err
== 0 || err
== -EINPROGRESS
) {
1418 if (copy_to_user(optval
, &assoc_id
, sizeof(assoc_id
)))
1420 if (put_user(sizeof(assoc_id
), optlen
))
1427 /* API 3.1.4 close() - UDP Style Syntax
1428 * Applications use close() to perform graceful shutdown (as described in
1429 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1430 * by a UDP-style socket.
1434 * ret = close(int sd);
1436 * sd - the socket descriptor of the associations to be closed.
1438 * To gracefully shutdown a specific association represented by the
1439 * UDP-style socket, an application should use the sendmsg() call,
1440 * passing no user data, but including the appropriate flag in the
1441 * ancillary data (see Section xxxx).
1443 * If sd in the close() call is a branched-off socket representing only
1444 * one association, the shutdown is performed on that association only.
1446 * 4.1.6 close() - TCP Style Syntax
1448 * Applications use close() to gracefully close down an association.
1452 * int close(int sd);
1454 * sd - the socket descriptor of the association to be closed.
1456 * After an application calls close() on a socket descriptor, no further
1457 * socket operations will succeed on that descriptor.
1459 * API 7.1.4 SO_LINGER
1461 * An application using the TCP-style socket can use this option to
1462 * perform the SCTP ABORT primitive. The linger option structure is:
1465 * int l_onoff; // option on/off
1466 * int l_linger; // linger time
1469 * To enable the option, set l_onoff to 1. If the l_linger value is set
1470 * to 0, calling close() is the same as the ABORT primitive. If the
1471 * value is set to a negative value, the setsockopt() call will return
1472 * an error. If the value is set to a positive value linger_time, the
1473 * close() can be blocked for at most linger_time ms. If the graceful
1474 * shutdown phase does not finish during this period, close() will
1475 * return but the graceful shutdown phase continues in the system.
1477 static void sctp_close(struct sock
*sk
, long timeout
)
1479 struct net
*net
= sock_net(sk
);
1480 struct sctp_endpoint
*ep
;
1481 struct sctp_association
*asoc
;
1482 struct list_head
*pos
, *temp
;
1483 unsigned int data_was_unread
;
1485 pr_debug("%s: sk:%p, timeout:%ld\n", __func__
, sk
, timeout
);
1488 sk
->sk_shutdown
= SHUTDOWN_MASK
;
1489 sk
->sk_state
= SCTP_SS_CLOSING
;
1491 ep
= sctp_sk(sk
)->ep
;
1493 /* Clean up any skbs sitting on the receive queue. */
1494 data_was_unread
= sctp_queue_purge_ulpevents(&sk
->sk_receive_queue
);
1495 data_was_unread
+= sctp_queue_purge_ulpevents(&sctp_sk(sk
)->pd_lobby
);
1497 /* Walk all associations on an endpoint. */
1498 list_for_each_safe(pos
, temp
, &ep
->asocs
) {
1499 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
1501 if (sctp_style(sk
, TCP
)) {
1502 /* A closed association can still be in the list if
1503 * it belongs to a TCP-style listening socket that is
1504 * not yet accepted. If so, free it. If not, send an
1505 * ABORT or SHUTDOWN based on the linger options.
1507 if (sctp_state(asoc
, CLOSED
)) {
1508 sctp_association_free(asoc
);
1513 if (data_was_unread
|| !skb_queue_empty(&asoc
->ulpq
.lobby
) ||
1514 !skb_queue_empty(&asoc
->ulpq
.reasm
) ||
1515 (sock_flag(sk
, SOCK_LINGER
) && !sk
->sk_lingertime
)) {
1516 struct sctp_chunk
*chunk
;
1518 chunk
= sctp_make_abort_user(asoc
, NULL
, 0);
1519 sctp_primitive_ABORT(net
, asoc
, chunk
);
1521 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
1524 /* On a TCP-style socket, block for at most linger_time if set. */
1525 if (sctp_style(sk
, TCP
) && timeout
)
1526 sctp_wait_for_close(sk
, timeout
);
1528 /* This will run the backlog queue. */
1531 /* Supposedly, no process has access to the socket, but
1532 * the net layers still may.
1533 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1534 * held and that should be grabbed before socket lock.
1536 spin_lock_bh(&net
->sctp
.addr_wq_lock
);
1539 /* Hold the sock, since sk_common_release() will put sock_put()
1540 * and we have just a little more cleanup.
1543 sk_common_release(sk
);
1546 spin_unlock_bh(&net
->sctp
.addr_wq_lock
);
1550 SCTP_DBG_OBJCNT_DEC(sock
);
1553 /* Handle EPIPE error. */
1554 static int sctp_error(struct sock
*sk
, int flags
, int err
)
1557 err
= sock_error(sk
) ? : -EPIPE
;
1558 if (err
== -EPIPE
&& !(flags
& MSG_NOSIGNAL
))
1559 send_sig(SIGPIPE
, current
, 0);
1563 /* API 3.1.3 sendmsg() - UDP Style Syntax
1565 * An application uses sendmsg() and recvmsg() calls to transmit data to
1566 * and receive data from its peer.
1568 * ssize_t sendmsg(int socket, const struct msghdr *message,
1571 * socket - the socket descriptor of the endpoint.
1572 * message - pointer to the msghdr structure which contains a single
1573 * user message and possibly some ancillary data.
1575 * See Section 5 for complete description of the data
1578 * flags - flags sent or received with the user message, see Section
1579 * 5 for complete description of the flags.
1581 * Note: This function could use a rewrite especially when explicit
1582 * connect support comes in.
1584 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1586 static int sctp_msghdr_parse(const struct msghdr
*, sctp_cmsgs_t
*);
1588 static int sctp_sendmsg(struct sock
*sk
, struct msghdr
*msg
, size_t msg_len
)
1590 struct net
*net
= sock_net(sk
);
1591 struct sctp_sock
*sp
;
1592 struct sctp_endpoint
*ep
;
1593 struct sctp_association
*new_asoc
= NULL
, *asoc
= NULL
;
1594 struct sctp_transport
*transport
, *chunk_tp
;
1595 struct sctp_chunk
*chunk
;
1597 struct sockaddr
*msg_name
= NULL
;
1598 struct sctp_sndrcvinfo default_sinfo
;
1599 struct sctp_sndrcvinfo
*sinfo
;
1600 struct sctp_initmsg
*sinit
;
1601 sctp_assoc_t associd
= 0;
1602 sctp_cmsgs_t cmsgs
= { NULL
};
1604 bool fill_sinfo_ttl
= false, wait_connect
= false;
1605 struct sctp_datamsg
*datamsg
;
1606 int msg_flags
= msg
->msg_flags
;
1607 __u16 sinfo_flags
= 0;
1615 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__
, sk
,
1618 /* We cannot send a message over a TCP-style listening socket. */
1619 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
)) {
1624 /* Parse out the SCTP CMSGs. */
1625 err
= sctp_msghdr_parse(msg
, &cmsgs
);
1627 pr_debug("%s: msghdr parse err:%x\n", __func__
, err
);
1631 /* Fetch the destination address for this packet. This
1632 * address only selects the association--it is not necessarily
1633 * the address we will send to.
1634 * For a peeled-off socket, msg_name is ignored.
1636 if (!sctp_style(sk
, UDP_HIGH_BANDWIDTH
) && msg
->msg_name
) {
1637 int msg_namelen
= msg
->msg_namelen
;
1639 err
= sctp_verify_addr(sk
, (union sctp_addr
*)msg
->msg_name
,
1644 if (msg_namelen
> sizeof(to
))
1645 msg_namelen
= sizeof(to
);
1646 memcpy(&to
, msg
->msg_name
, msg_namelen
);
1647 msg_name
= msg
->msg_name
;
1651 if (cmsgs
.sinfo
!= NULL
) {
1652 memset(&default_sinfo
, 0, sizeof(default_sinfo
));
1653 default_sinfo
.sinfo_stream
= cmsgs
.sinfo
->snd_sid
;
1654 default_sinfo
.sinfo_flags
= cmsgs
.sinfo
->snd_flags
;
1655 default_sinfo
.sinfo_ppid
= cmsgs
.sinfo
->snd_ppid
;
1656 default_sinfo
.sinfo_context
= cmsgs
.sinfo
->snd_context
;
1657 default_sinfo
.sinfo_assoc_id
= cmsgs
.sinfo
->snd_assoc_id
;
1659 sinfo
= &default_sinfo
;
1660 fill_sinfo_ttl
= true;
1662 sinfo
= cmsgs
.srinfo
;
1664 /* Did the user specify SNDINFO/SNDRCVINFO? */
1666 sinfo_flags
= sinfo
->sinfo_flags
;
1667 associd
= sinfo
->sinfo_assoc_id
;
1670 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__
,
1671 msg_len
, sinfo_flags
);
1673 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1674 if (sctp_style(sk
, TCP
) && (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
))) {
1679 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1680 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1681 * If SCTP_ABORT is set, the message length could be non zero with
1682 * the msg_iov set to the user abort reason.
1684 if (((sinfo_flags
& SCTP_EOF
) && (msg_len
> 0)) ||
1685 (!(sinfo_flags
& (SCTP_EOF
|SCTP_ABORT
)) && (msg_len
== 0))) {
1690 /* If SCTP_ADDR_OVER is set, there must be an address
1691 * specified in msg_name.
1693 if ((sinfo_flags
& SCTP_ADDR_OVER
) && (!msg
->msg_name
)) {
1700 pr_debug("%s: about to look up association\n", __func__
);
1704 /* If a msg_name has been specified, assume this is to be used. */
1706 /* Look for a matching association on the endpoint. */
1707 asoc
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1709 /* If we could not find a matching association on the
1710 * endpoint, make sure that it is not a TCP-style
1711 * socket that already has an association or there is
1712 * no peeled-off association on another socket.
1715 ((sctp_style(sk
, TCP
) &&
1716 (sctp_sstate(sk
, ESTABLISHED
) ||
1717 sctp_sstate(sk
, CLOSING
))) ||
1718 sctp_endpoint_is_peeled_off(ep
, &to
))) {
1719 err
= -EADDRNOTAVAIL
;
1723 asoc
= sctp_id2assoc(sk
, associd
);
1731 pr_debug("%s: just looked up association:%p\n", __func__
, asoc
);
1733 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1734 * socket that has an association in CLOSED state. This can
1735 * happen when an accepted socket has an association that is
1738 if (sctp_state(asoc
, CLOSED
) && sctp_style(sk
, TCP
)) {
1743 if (sinfo_flags
& SCTP_EOF
) {
1744 pr_debug("%s: shutting down association:%p\n",
1747 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
1751 if (sinfo_flags
& SCTP_ABORT
) {
1753 chunk
= sctp_make_abort_user(asoc
, msg
, msg_len
);
1759 pr_debug("%s: aborting association:%p\n",
1762 sctp_primitive_ABORT(net
, asoc
, chunk
);
1768 /* Do we need to create the association? */
1770 pr_debug("%s: there is no association yet\n", __func__
);
1772 if (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
)) {
1777 /* Check for invalid stream against the stream counts,
1778 * either the default or the user specified stream counts.
1781 if (!sinit
|| !sinit
->sinit_num_ostreams
) {
1782 /* Check against the defaults. */
1783 if (sinfo
->sinfo_stream
>=
1784 sp
->initmsg
.sinit_num_ostreams
) {
1789 /* Check against the requested. */
1790 if (sinfo
->sinfo_stream
>=
1791 sinit
->sinit_num_ostreams
) {
1799 * API 3.1.2 bind() - UDP Style Syntax
1800 * If a bind() or sctp_bindx() is not called prior to a
1801 * sendmsg() call that initiates a new association, the
1802 * system picks an ephemeral port and will choose an address
1803 * set equivalent to binding with a wildcard address.
1805 if (!ep
->base
.bind_addr
.port
) {
1806 if (sctp_autobind(sk
)) {
1812 * If an unprivileged user inherits a one-to-many
1813 * style socket with open associations on a privileged
1814 * port, it MAY be permitted to accept new associations,
1815 * but it SHOULD NOT be permitted to open new
1818 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1819 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
)) {
1825 scope
= sctp_scope(&to
);
1826 new_asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1832 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
, GFP_KERNEL
);
1838 /* If the SCTP_INIT ancillary data is specified, set all
1839 * the association init values accordingly.
1842 if (sinit
->sinit_num_ostreams
) {
1843 asoc
->c
.sinit_num_ostreams
=
1844 sinit
->sinit_num_ostreams
;
1846 if (sinit
->sinit_max_instreams
) {
1847 asoc
->c
.sinit_max_instreams
=
1848 sinit
->sinit_max_instreams
;
1850 if (sinit
->sinit_max_attempts
) {
1851 asoc
->max_init_attempts
1852 = sinit
->sinit_max_attempts
;
1854 if (sinit
->sinit_max_init_timeo
) {
1855 asoc
->max_init_timeo
=
1856 msecs_to_jiffies(sinit
->sinit_max_init_timeo
);
1860 /* Prime the peer's transport structures. */
1861 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
, SCTP_UNKNOWN
);
1868 /* ASSERT: we have a valid association at this point. */
1869 pr_debug("%s: we have a valid association\n", __func__
);
1872 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1873 * one with some defaults.
1875 memset(&default_sinfo
, 0, sizeof(default_sinfo
));
1876 default_sinfo
.sinfo_stream
= asoc
->default_stream
;
1877 default_sinfo
.sinfo_flags
= asoc
->default_flags
;
1878 default_sinfo
.sinfo_ppid
= asoc
->default_ppid
;
1879 default_sinfo
.sinfo_context
= asoc
->default_context
;
1880 default_sinfo
.sinfo_timetolive
= asoc
->default_timetolive
;
1881 default_sinfo
.sinfo_assoc_id
= sctp_assoc2id(asoc
);
1883 sinfo
= &default_sinfo
;
1884 } else if (fill_sinfo_ttl
) {
1885 /* In case SNDINFO was specified, we still need to fill
1886 * it with a default ttl from the assoc here.
1888 sinfo
->sinfo_timetolive
= asoc
->default_timetolive
;
1891 /* API 7.1.7, the sndbuf size per association bounds the
1892 * maximum size of data that can be sent in a single send call.
1894 if (msg_len
> sk
->sk_sndbuf
) {
1899 if (asoc
->pmtu_pending
)
1900 sctp_assoc_pending_pmtu(sk
, asoc
);
1902 /* If fragmentation is disabled and the message length exceeds the
1903 * association fragmentation point, return EMSGSIZE. The I-D
1904 * does not specify what this error is, but this looks like
1907 if (sctp_sk(sk
)->disable_fragments
&& (msg_len
> asoc
->frag_point
)) {
1912 /* Check for invalid stream. */
1913 if (sinfo
->sinfo_stream
>= asoc
->c
.sinit_num_ostreams
) {
1918 if (sctp_wspace(asoc
) < msg_len
)
1919 sctp_prsctp_prune(asoc
, sinfo
, msg_len
- sctp_wspace(asoc
));
1921 timeo
= sock_sndtimeo(sk
, msg
->msg_flags
& MSG_DONTWAIT
);
1922 if (!sctp_wspace(asoc
)) {
1923 err
= sctp_wait_for_sndbuf(asoc
, &timeo
, msg_len
);
1928 /* If an address is passed with the sendto/sendmsg call, it is used
1929 * to override the primary destination address in the TCP model, or
1930 * when SCTP_ADDR_OVER flag is set in the UDP model.
1932 if ((sctp_style(sk
, TCP
) && msg_name
) ||
1933 (sinfo_flags
& SCTP_ADDR_OVER
)) {
1934 chunk_tp
= sctp_assoc_lookup_paddr(asoc
, &to
);
1942 /* Auto-connect, if we aren't connected already. */
1943 if (sctp_state(asoc
, CLOSED
)) {
1944 err
= sctp_primitive_ASSOCIATE(net
, asoc
, NULL
);
1948 wait_connect
= true;
1949 pr_debug("%s: we associated primitively\n", __func__
);
1952 /* Break the message into multiple chunks of maximum size. */
1953 datamsg
= sctp_datamsg_from_user(asoc
, sinfo
, &msg
->msg_iter
);
1954 if (IS_ERR(datamsg
)) {
1955 err
= PTR_ERR(datamsg
);
1959 /* Now send the (possibly) fragmented message. */
1960 list_for_each_entry(chunk
, &datamsg
->chunks
, frag_list
) {
1961 /* Do accounting for the write space. */
1962 sctp_set_owner_w(chunk
);
1964 chunk
->transport
= chunk_tp
;
1967 /* Send it to the lower layers. Note: all chunks
1968 * must either fail or succeed. The lower layer
1969 * works that way today. Keep it that way or this
1972 err
= sctp_primitive_SEND(net
, asoc
, datamsg
);
1973 sctp_datamsg_put(datamsg
);
1974 /* Did the lower layer accept the chunk? */
1978 pr_debug("%s: we sent primitively\n", __func__
);
1982 if (unlikely(wait_connect
)) {
1983 timeo
= sock_sndtimeo(sk
, msg_flags
& MSG_DONTWAIT
);
1984 sctp_wait_for_connect(asoc
, &timeo
);
1987 /* If we are already past ASSOCIATE, the lower
1988 * layers are responsible for association cleanup.
1994 sctp_association_free(asoc
);
1999 return sctp_error(sk
, msg_flags
, err
);
2006 err
= sock_error(sk
);
2016 /* This is an extended version of skb_pull() that removes the data from the
2017 * start of a skb even when data is spread across the list of skb's in the
2018 * frag_list. len specifies the total amount of data that needs to be removed.
2019 * when 'len' bytes could be removed from the skb, it returns 0.
2020 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2021 * could not be removed.
2023 static int sctp_skb_pull(struct sk_buff
*skb
, int len
)
2025 struct sk_buff
*list
;
2026 int skb_len
= skb_headlen(skb
);
2029 if (len
<= skb_len
) {
2030 __skb_pull(skb
, len
);
2034 __skb_pull(skb
, skb_len
);
2036 skb_walk_frags(skb
, list
) {
2037 rlen
= sctp_skb_pull(list
, len
);
2038 skb
->len
-= (len
-rlen
);
2039 skb
->data_len
-= (len
-rlen
);
2050 /* API 3.1.3 recvmsg() - UDP Style Syntax
2052 * ssize_t recvmsg(int socket, struct msghdr *message,
2055 * socket - the socket descriptor of the endpoint.
2056 * message - pointer to the msghdr structure which contains a single
2057 * user message and possibly some ancillary data.
2059 * See Section 5 for complete description of the data
2062 * flags - flags sent or received with the user message, see Section
2063 * 5 for complete description of the flags.
2065 static int sctp_recvmsg(struct sock
*sk
, struct msghdr
*msg
, size_t len
,
2066 int noblock
, int flags
, int *addr_len
)
2068 struct sctp_ulpevent
*event
= NULL
;
2069 struct sctp_sock
*sp
= sctp_sk(sk
);
2070 struct sk_buff
*skb
, *head_skb
;
2075 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2076 "addr_len:%p)\n", __func__
, sk
, msg
, len
, noblock
, flags
,
2081 if (sctp_style(sk
, TCP
) && !sctp_sstate(sk
, ESTABLISHED
) &&
2082 !sctp_sstate(sk
, CLOSING
) && !sctp_sstate(sk
, CLOSED
)) {
2087 skb
= sctp_skb_recv_datagram(sk
, flags
, noblock
, &err
);
2091 /* Get the total length of the skb including any skb's in the
2100 err
= skb_copy_datagram_msg(skb
, 0, msg
, copied
);
2102 event
= sctp_skb2event(skb
);
2107 if (event
->chunk
&& event
->chunk
->head_skb
)
2108 head_skb
= event
->chunk
->head_skb
;
2111 sock_recv_ts_and_drops(msg
, sk
, head_skb
);
2112 if (sctp_ulpevent_is_notification(event
)) {
2113 msg
->msg_flags
|= MSG_NOTIFICATION
;
2114 sp
->pf
->event_msgname(event
, msg
->msg_name
, addr_len
);
2116 sp
->pf
->skb_msgname(head_skb
, msg
->msg_name
, addr_len
);
2119 /* Check if we allow SCTP_NXTINFO. */
2120 if (sp
->recvnxtinfo
)
2121 sctp_ulpevent_read_nxtinfo(event
, msg
, sk
);
2122 /* Check if we allow SCTP_RCVINFO. */
2123 if (sp
->recvrcvinfo
)
2124 sctp_ulpevent_read_rcvinfo(event
, msg
);
2125 /* Check if we allow SCTP_SNDRCVINFO. */
2126 if (sp
->subscribe
.sctp_data_io_event
)
2127 sctp_ulpevent_read_sndrcvinfo(event
, msg
);
2131 /* If skb's length exceeds the user's buffer, update the skb and
2132 * push it back to the receive_queue so that the next call to
2133 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2135 if (skb_len
> copied
) {
2136 msg
->msg_flags
&= ~MSG_EOR
;
2137 if (flags
& MSG_PEEK
)
2139 sctp_skb_pull(skb
, copied
);
2140 skb_queue_head(&sk
->sk_receive_queue
, skb
);
2142 /* When only partial message is copied to the user, increase
2143 * rwnd by that amount. If all the data in the skb is read,
2144 * rwnd is updated when the event is freed.
2146 if (!sctp_ulpevent_is_notification(event
))
2147 sctp_assoc_rwnd_increase(event
->asoc
, copied
);
2149 } else if ((event
->msg_flags
& MSG_NOTIFICATION
) ||
2150 (event
->msg_flags
& MSG_EOR
))
2151 msg
->msg_flags
|= MSG_EOR
;
2153 msg
->msg_flags
&= ~MSG_EOR
;
2156 if (flags
& MSG_PEEK
) {
2157 /* Release the skb reference acquired after peeking the skb in
2158 * sctp_skb_recv_datagram().
2162 /* Free the event which includes releasing the reference to
2163 * the owner of the skb, freeing the skb and updating the
2166 sctp_ulpevent_free(event
);
2173 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2175 * This option is a on/off flag. If enabled no SCTP message
2176 * fragmentation will be performed. Instead if a message being sent
2177 * exceeds the current PMTU size, the message will NOT be sent and
2178 * instead a error will be indicated to the user.
2180 static int sctp_setsockopt_disable_fragments(struct sock
*sk
,
2181 char __user
*optval
,
2182 unsigned int optlen
)
2186 if (optlen
< sizeof(int))
2189 if (get_user(val
, (int __user
*)optval
))
2192 sctp_sk(sk
)->disable_fragments
= (val
== 0) ? 0 : 1;
2197 static int sctp_setsockopt_events(struct sock
*sk
, char __user
*optval
,
2198 unsigned int optlen
)
2200 struct sctp_association
*asoc
;
2201 struct sctp_ulpevent
*event
;
2203 if (optlen
> sizeof(struct sctp_event_subscribe
))
2205 if (copy_from_user(&sctp_sk(sk
)->subscribe
, optval
, optlen
))
2208 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2209 * if there is no data to be sent or retransmit, the stack will
2210 * immediately send up this notification.
2212 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT
,
2213 &sctp_sk(sk
)->subscribe
)) {
2214 asoc
= sctp_id2assoc(sk
, 0);
2216 if (asoc
&& sctp_outq_is_empty(&asoc
->outqueue
)) {
2217 event
= sctp_ulpevent_make_sender_dry_event(asoc
,
2222 sctp_ulpq_tail_event(&asoc
->ulpq
, event
);
2229 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2231 * This socket option is applicable to the UDP-style socket only. When
2232 * set it will cause associations that are idle for more than the
2233 * specified number of seconds to automatically close. An association
2234 * being idle is defined an association that has NOT sent or received
2235 * user data. The special value of '0' indicates that no automatic
2236 * close of any associations should be performed. The option expects an
2237 * integer defining the number of seconds of idle time before an
2238 * association is closed.
2240 static int sctp_setsockopt_autoclose(struct sock
*sk
, char __user
*optval
,
2241 unsigned int optlen
)
2243 struct sctp_sock
*sp
= sctp_sk(sk
);
2244 struct net
*net
= sock_net(sk
);
2246 /* Applicable to UDP-style socket only */
2247 if (sctp_style(sk
, TCP
))
2249 if (optlen
!= sizeof(int))
2251 if (copy_from_user(&sp
->autoclose
, optval
, optlen
))
2254 if (sp
->autoclose
> net
->sctp
.max_autoclose
)
2255 sp
->autoclose
= net
->sctp
.max_autoclose
;
2260 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2262 * Applications can enable or disable heartbeats for any peer address of
2263 * an association, modify an address's heartbeat interval, force a
2264 * heartbeat to be sent immediately, and adjust the address's maximum
2265 * number of retransmissions sent before an address is considered
2266 * unreachable. The following structure is used to access and modify an
2267 * address's parameters:
2269 * struct sctp_paddrparams {
2270 * sctp_assoc_t spp_assoc_id;
2271 * struct sockaddr_storage spp_address;
2272 * uint32_t spp_hbinterval;
2273 * uint16_t spp_pathmaxrxt;
2274 * uint32_t spp_pathmtu;
2275 * uint32_t spp_sackdelay;
2276 * uint32_t spp_flags;
2279 * spp_assoc_id - (one-to-many style socket) This is filled in the
2280 * application, and identifies the association for
2282 * spp_address - This specifies which address is of interest.
2283 * spp_hbinterval - This contains the value of the heartbeat interval,
2284 * in milliseconds. If a value of zero
2285 * is present in this field then no changes are to
2286 * be made to this parameter.
2287 * spp_pathmaxrxt - This contains the maximum number of
2288 * retransmissions before this address shall be
2289 * considered unreachable. If a value of zero
2290 * is present in this field then no changes are to
2291 * be made to this parameter.
2292 * spp_pathmtu - When Path MTU discovery is disabled the value
2293 * specified here will be the "fixed" path mtu.
2294 * Note that if the spp_address field is empty
2295 * then all associations on this address will
2296 * have this fixed path mtu set upon them.
2298 * spp_sackdelay - When delayed sack is enabled, this value specifies
2299 * the number of milliseconds that sacks will be delayed
2300 * for. This value will apply to all addresses of an
2301 * association if the spp_address field is empty. Note
2302 * also, that if delayed sack is enabled and this
2303 * value is set to 0, no change is made to the last
2304 * recorded delayed sack timer value.
2306 * spp_flags - These flags are used to control various features
2307 * on an association. The flag field may contain
2308 * zero or more of the following options.
2310 * SPP_HB_ENABLE - Enable heartbeats on the
2311 * specified address. Note that if the address
2312 * field is empty all addresses for the association
2313 * have heartbeats enabled upon them.
2315 * SPP_HB_DISABLE - Disable heartbeats on the
2316 * speicifed address. Note that if the address
2317 * field is empty all addresses for the association
2318 * will have their heartbeats disabled. Note also
2319 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2320 * mutually exclusive, only one of these two should
2321 * be specified. Enabling both fields will have
2322 * undetermined results.
2324 * SPP_HB_DEMAND - Request a user initiated heartbeat
2325 * to be made immediately.
2327 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2328 * heartbeat delayis to be set to the value of 0
2331 * SPP_PMTUD_ENABLE - This field will enable PMTU
2332 * discovery upon the specified address. Note that
2333 * if the address feild is empty then all addresses
2334 * on the association are effected.
2336 * SPP_PMTUD_DISABLE - This field will disable PMTU
2337 * discovery upon the specified address. Note that
2338 * if the address feild is empty then all addresses
2339 * on the association are effected. Not also that
2340 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2341 * exclusive. Enabling both will have undetermined
2344 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2345 * on delayed sack. The time specified in spp_sackdelay
2346 * is used to specify the sack delay for this address. Note
2347 * that if spp_address is empty then all addresses will
2348 * enable delayed sack and take on the sack delay
2349 * value specified in spp_sackdelay.
2350 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2351 * off delayed sack. If the spp_address field is blank then
2352 * delayed sack is disabled for the entire association. Note
2353 * also that this field is mutually exclusive to
2354 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2357 static int sctp_apply_peer_addr_params(struct sctp_paddrparams
*params
,
2358 struct sctp_transport
*trans
,
2359 struct sctp_association
*asoc
,
2360 struct sctp_sock
*sp
,
2363 int sackdelay_change
)
2367 if (params
->spp_flags
& SPP_HB_DEMAND
&& trans
) {
2368 struct net
*net
= sock_net(trans
->asoc
->base
.sk
);
2370 error
= sctp_primitive_REQUESTHEARTBEAT(net
, trans
->asoc
, trans
);
2375 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2376 * this field is ignored. Note also that a value of zero indicates
2377 * the current setting should be left unchanged.
2379 if (params
->spp_flags
& SPP_HB_ENABLE
) {
2381 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2382 * set. This lets us use 0 value when this flag
2385 if (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)
2386 params
->spp_hbinterval
= 0;
2388 if (params
->spp_hbinterval
||
2389 (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)) {
2392 msecs_to_jiffies(params
->spp_hbinterval
);
2395 msecs_to_jiffies(params
->spp_hbinterval
);
2397 sp
->hbinterval
= params
->spp_hbinterval
;
2404 trans
->param_flags
=
2405 (trans
->param_flags
& ~SPP_HB
) | hb_change
;
2408 (asoc
->param_flags
& ~SPP_HB
) | hb_change
;
2411 (sp
->param_flags
& ~SPP_HB
) | hb_change
;
2415 /* When Path MTU discovery is disabled the value specified here will
2416 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2417 * include the flag SPP_PMTUD_DISABLE for this field to have any
2420 if ((params
->spp_flags
& SPP_PMTUD_DISABLE
) && params
->spp_pathmtu
) {
2422 trans
->pathmtu
= params
->spp_pathmtu
;
2423 sctp_assoc_sync_pmtu(sctp_opt2sk(sp
), asoc
);
2425 asoc
->pathmtu
= params
->spp_pathmtu
;
2426 sctp_frag_point(asoc
, params
->spp_pathmtu
);
2428 sp
->pathmtu
= params
->spp_pathmtu
;
2434 int update
= (trans
->param_flags
& SPP_PMTUD_DISABLE
) &&
2435 (params
->spp_flags
& SPP_PMTUD_ENABLE
);
2436 trans
->param_flags
=
2437 (trans
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2439 sctp_transport_pmtu(trans
, sctp_opt2sk(sp
));
2440 sctp_assoc_sync_pmtu(sctp_opt2sk(sp
), asoc
);
2444 (asoc
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2447 (sp
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2451 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2452 * value of this field is ignored. Note also that a value of zero
2453 * indicates the current setting should be left unchanged.
2455 if ((params
->spp_flags
& SPP_SACKDELAY_ENABLE
) && params
->spp_sackdelay
) {
2458 msecs_to_jiffies(params
->spp_sackdelay
);
2461 msecs_to_jiffies(params
->spp_sackdelay
);
2463 sp
->sackdelay
= params
->spp_sackdelay
;
2467 if (sackdelay_change
) {
2469 trans
->param_flags
=
2470 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2474 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2478 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2483 /* Note that a value of zero indicates the current setting should be
2486 if (params
->spp_pathmaxrxt
) {
2488 trans
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2490 asoc
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2492 sp
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2499 static int sctp_setsockopt_peer_addr_params(struct sock
*sk
,
2500 char __user
*optval
,
2501 unsigned int optlen
)
2503 struct sctp_paddrparams params
;
2504 struct sctp_transport
*trans
= NULL
;
2505 struct sctp_association
*asoc
= NULL
;
2506 struct sctp_sock
*sp
= sctp_sk(sk
);
2508 int hb_change
, pmtud_change
, sackdelay_change
;
2510 if (optlen
!= sizeof(struct sctp_paddrparams
))
2513 if (copy_from_user(¶ms
, optval
, optlen
))
2516 /* Validate flags and value parameters. */
2517 hb_change
= params
.spp_flags
& SPP_HB
;
2518 pmtud_change
= params
.spp_flags
& SPP_PMTUD
;
2519 sackdelay_change
= params
.spp_flags
& SPP_SACKDELAY
;
2521 if (hb_change
== SPP_HB
||
2522 pmtud_change
== SPP_PMTUD
||
2523 sackdelay_change
== SPP_SACKDELAY
||
2524 params
.spp_sackdelay
> 500 ||
2525 (params
.spp_pathmtu
&&
2526 params
.spp_pathmtu
< SCTP_DEFAULT_MINSEGMENT
))
2529 /* If an address other than INADDR_ANY is specified, and
2530 * no transport is found, then the request is invalid.
2532 if (!sctp_is_any(sk
, (union sctp_addr
*)¶ms
.spp_address
)) {
2533 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
2534 params
.spp_assoc_id
);
2539 /* Get association, if assoc_id != 0 and the socket is a one
2540 * to many style socket, and an association was not found, then
2541 * the id was invalid.
2543 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
2544 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
))
2547 /* Heartbeat demand can only be sent on a transport or
2548 * association, but not a socket.
2550 if (params
.spp_flags
& SPP_HB_DEMAND
&& !trans
&& !asoc
)
2553 /* Process parameters. */
2554 error
= sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2555 hb_change
, pmtud_change
,
2561 /* If changes are for association, also apply parameters to each
2564 if (!trans
&& asoc
) {
2565 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2567 sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2568 hb_change
, pmtud_change
,
2576 static inline __u32
sctp_spp_sackdelay_enable(__u32 param_flags
)
2578 return (param_flags
& ~SPP_SACKDELAY
) | SPP_SACKDELAY_ENABLE
;
2581 static inline __u32
sctp_spp_sackdelay_disable(__u32 param_flags
)
2583 return (param_flags
& ~SPP_SACKDELAY
) | SPP_SACKDELAY_DISABLE
;
2587 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2589 * This option will effect the way delayed acks are performed. This
2590 * option allows you to get or set the delayed ack time, in
2591 * milliseconds. It also allows changing the delayed ack frequency.
2592 * Changing the frequency to 1 disables the delayed sack algorithm. If
2593 * the assoc_id is 0, then this sets or gets the endpoints default
2594 * values. If the assoc_id field is non-zero, then the set or get
2595 * effects the specified association for the one to many model (the
2596 * assoc_id field is ignored by the one to one model). Note that if
2597 * sack_delay or sack_freq are 0 when setting this option, then the
2598 * current values will remain unchanged.
2600 * struct sctp_sack_info {
2601 * sctp_assoc_t sack_assoc_id;
2602 * uint32_t sack_delay;
2603 * uint32_t sack_freq;
2606 * sack_assoc_id - This parameter, indicates which association the user
2607 * is performing an action upon. Note that if this field's value is
2608 * zero then the endpoints default value is changed (effecting future
2609 * associations only).
2611 * sack_delay - This parameter contains the number of milliseconds that
2612 * the user is requesting the delayed ACK timer be set to. Note that
2613 * this value is defined in the standard to be between 200 and 500
2616 * sack_freq - This parameter contains the number of packets that must
2617 * be received before a sack is sent without waiting for the delay
2618 * timer to expire. The default value for this is 2, setting this
2619 * value to 1 will disable the delayed sack algorithm.
2622 static int sctp_setsockopt_delayed_ack(struct sock
*sk
,
2623 char __user
*optval
, unsigned int optlen
)
2625 struct sctp_sack_info params
;
2626 struct sctp_transport
*trans
= NULL
;
2627 struct sctp_association
*asoc
= NULL
;
2628 struct sctp_sock
*sp
= sctp_sk(sk
);
2630 if (optlen
== sizeof(struct sctp_sack_info
)) {
2631 if (copy_from_user(¶ms
, optval
, optlen
))
2634 if (params
.sack_delay
== 0 && params
.sack_freq
== 0)
2636 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
2637 pr_warn_ratelimited(DEPRECATED
2639 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2640 "Use struct sctp_sack_info instead\n",
2641 current
->comm
, task_pid_nr(current
));
2642 if (copy_from_user(¶ms
, optval
, optlen
))
2645 if (params
.sack_delay
== 0)
2646 params
.sack_freq
= 1;
2648 params
.sack_freq
= 0;
2652 /* Validate value parameter. */
2653 if (params
.sack_delay
> 500)
2656 /* Get association, if sack_assoc_id != 0 and the socket is a one
2657 * to many style socket, and an association was not found, then
2658 * the id was invalid.
2660 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
2661 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
2664 if (params
.sack_delay
) {
2667 msecs_to_jiffies(params
.sack_delay
);
2669 sctp_spp_sackdelay_enable(asoc
->param_flags
);
2671 sp
->sackdelay
= params
.sack_delay
;
2673 sctp_spp_sackdelay_enable(sp
->param_flags
);
2677 if (params
.sack_freq
== 1) {
2680 sctp_spp_sackdelay_disable(asoc
->param_flags
);
2683 sctp_spp_sackdelay_disable(sp
->param_flags
);
2685 } else if (params
.sack_freq
> 1) {
2687 asoc
->sackfreq
= params
.sack_freq
;
2689 sctp_spp_sackdelay_enable(asoc
->param_flags
);
2691 sp
->sackfreq
= params
.sack_freq
;
2693 sctp_spp_sackdelay_enable(sp
->param_flags
);
2697 /* If change is for association, also apply to each transport. */
2699 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2701 if (params
.sack_delay
) {
2703 msecs_to_jiffies(params
.sack_delay
);
2704 trans
->param_flags
=
2705 sctp_spp_sackdelay_enable(trans
->param_flags
);
2707 if (params
.sack_freq
== 1) {
2708 trans
->param_flags
=
2709 sctp_spp_sackdelay_disable(trans
->param_flags
);
2710 } else if (params
.sack_freq
> 1) {
2711 trans
->sackfreq
= params
.sack_freq
;
2712 trans
->param_flags
=
2713 sctp_spp_sackdelay_enable(trans
->param_flags
);
2721 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2723 * Applications can specify protocol parameters for the default association
2724 * initialization. The option name argument to setsockopt() and getsockopt()
2727 * Setting initialization parameters is effective only on an unconnected
2728 * socket (for UDP-style sockets only future associations are effected
2729 * by the change). With TCP-style sockets, this option is inherited by
2730 * sockets derived from a listener socket.
2732 static int sctp_setsockopt_initmsg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2734 struct sctp_initmsg sinit
;
2735 struct sctp_sock
*sp
= sctp_sk(sk
);
2737 if (optlen
!= sizeof(struct sctp_initmsg
))
2739 if (copy_from_user(&sinit
, optval
, optlen
))
2742 if (sinit
.sinit_num_ostreams
)
2743 sp
->initmsg
.sinit_num_ostreams
= sinit
.sinit_num_ostreams
;
2744 if (sinit
.sinit_max_instreams
)
2745 sp
->initmsg
.sinit_max_instreams
= sinit
.sinit_max_instreams
;
2746 if (sinit
.sinit_max_attempts
)
2747 sp
->initmsg
.sinit_max_attempts
= sinit
.sinit_max_attempts
;
2748 if (sinit
.sinit_max_init_timeo
)
2749 sp
->initmsg
.sinit_max_init_timeo
= sinit
.sinit_max_init_timeo
;
2755 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2757 * Applications that wish to use the sendto() system call may wish to
2758 * specify a default set of parameters that would normally be supplied
2759 * through the inclusion of ancillary data. This socket option allows
2760 * such an application to set the default sctp_sndrcvinfo structure.
2761 * The application that wishes to use this socket option simply passes
2762 * in to this call the sctp_sndrcvinfo structure defined in Section
2763 * 5.2.2) The input parameters accepted by this call include
2764 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2765 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2766 * to this call if the caller is using the UDP model.
2768 static int sctp_setsockopt_default_send_param(struct sock
*sk
,
2769 char __user
*optval
,
2770 unsigned int optlen
)
2772 struct sctp_sock
*sp
= sctp_sk(sk
);
2773 struct sctp_association
*asoc
;
2774 struct sctp_sndrcvinfo info
;
2776 if (optlen
!= sizeof(info
))
2778 if (copy_from_user(&info
, optval
, optlen
))
2780 if (info
.sinfo_flags
&
2781 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
2782 SCTP_ABORT
| SCTP_EOF
))
2785 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
2786 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
2789 asoc
->default_stream
= info
.sinfo_stream
;
2790 asoc
->default_flags
= info
.sinfo_flags
;
2791 asoc
->default_ppid
= info
.sinfo_ppid
;
2792 asoc
->default_context
= info
.sinfo_context
;
2793 asoc
->default_timetolive
= info
.sinfo_timetolive
;
2795 sp
->default_stream
= info
.sinfo_stream
;
2796 sp
->default_flags
= info
.sinfo_flags
;
2797 sp
->default_ppid
= info
.sinfo_ppid
;
2798 sp
->default_context
= info
.sinfo_context
;
2799 sp
->default_timetolive
= info
.sinfo_timetolive
;
2805 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2806 * (SCTP_DEFAULT_SNDINFO)
2808 static int sctp_setsockopt_default_sndinfo(struct sock
*sk
,
2809 char __user
*optval
,
2810 unsigned int optlen
)
2812 struct sctp_sock
*sp
= sctp_sk(sk
);
2813 struct sctp_association
*asoc
;
2814 struct sctp_sndinfo info
;
2816 if (optlen
!= sizeof(info
))
2818 if (copy_from_user(&info
, optval
, optlen
))
2820 if (info
.snd_flags
&
2821 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
2822 SCTP_ABORT
| SCTP_EOF
))
2825 asoc
= sctp_id2assoc(sk
, info
.snd_assoc_id
);
2826 if (!asoc
&& info
.snd_assoc_id
&& sctp_style(sk
, UDP
))
2829 asoc
->default_stream
= info
.snd_sid
;
2830 asoc
->default_flags
= info
.snd_flags
;
2831 asoc
->default_ppid
= info
.snd_ppid
;
2832 asoc
->default_context
= info
.snd_context
;
2834 sp
->default_stream
= info
.snd_sid
;
2835 sp
->default_flags
= info
.snd_flags
;
2836 sp
->default_ppid
= info
.snd_ppid
;
2837 sp
->default_context
= info
.snd_context
;
2843 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2845 * Requests that the local SCTP stack use the enclosed peer address as
2846 * the association primary. The enclosed address must be one of the
2847 * association peer's addresses.
2849 static int sctp_setsockopt_primary_addr(struct sock
*sk
, char __user
*optval
,
2850 unsigned int optlen
)
2852 struct sctp_prim prim
;
2853 struct sctp_transport
*trans
;
2855 if (optlen
!= sizeof(struct sctp_prim
))
2858 if (copy_from_user(&prim
, optval
, sizeof(struct sctp_prim
)))
2861 trans
= sctp_addr_id2transport(sk
, &prim
.ssp_addr
, prim
.ssp_assoc_id
);
2865 sctp_assoc_set_primary(trans
->asoc
, trans
);
2871 * 7.1.5 SCTP_NODELAY
2873 * Turn on/off any Nagle-like algorithm. This means that packets are
2874 * generally sent as soon as possible and no unnecessary delays are
2875 * introduced, at the cost of more packets in the network. Expects an
2876 * integer boolean flag.
2878 static int sctp_setsockopt_nodelay(struct sock
*sk
, char __user
*optval
,
2879 unsigned int optlen
)
2883 if (optlen
< sizeof(int))
2885 if (get_user(val
, (int __user
*)optval
))
2888 sctp_sk(sk
)->nodelay
= (val
== 0) ? 0 : 1;
2894 * 7.1.1 SCTP_RTOINFO
2896 * The protocol parameters used to initialize and bound retransmission
2897 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2898 * and modify these parameters.
2899 * All parameters are time values, in milliseconds. A value of 0, when
2900 * modifying the parameters, indicates that the current value should not
2904 static int sctp_setsockopt_rtoinfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2906 struct sctp_rtoinfo rtoinfo
;
2907 struct sctp_association
*asoc
;
2908 unsigned long rto_min
, rto_max
;
2909 struct sctp_sock
*sp
= sctp_sk(sk
);
2911 if (optlen
!= sizeof (struct sctp_rtoinfo
))
2914 if (copy_from_user(&rtoinfo
, optval
, optlen
))
2917 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
2919 /* Set the values to the specific association */
2920 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
2923 rto_max
= rtoinfo
.srto_max
;
2924 rto_min
= rtoinfo
.srto_min
;
2927 rto_max
= asoc
? msecs_to_jiffies(rto_max
) : rto_max
;
2929 rto_max
= asoc
? asoc
->rto_max
: sp
->rtoinfo
.srto_max
;
2932 rto_min
= asoc
? msecs_to_jiffies(rto_min
) : rto_min
;
2934 rto_min
= asoc
? asoc
->rto_min
: sp
->rtoinfo
.srto_min
;
2936 if (rto_min
> rto_max
)
2940 if (rtoinfo
.srto_initial
!= 0)
2942 msecs_to_jiffies(rtoinfo
.srto_initial
);
2943 asoc
->rto_max
= rto_max
;
2944 asoc
->rto_min
= rto_min
;
2946 /* If there is no association or the association-id = 0
2947 * set the values to the endpoint.
2949 if (rtoinfo
.srto_initial
!= 0)
2950 sp
->rtoinfo
.srto_initial
= rtoinfo
.srto_initial
;
2951 sp
->rtoinfo
.srto_max
= rto_max
;
2952 sp
->rtoinfo
.srto_min
= rto_min
;
2960 * 7.1.2 SCTP_ASSOCINFO
2962 * This option is used to tune the maximum retransmission attempts
2963 * of the association.
2964 * Returns an error if the new association retransmission value is
2965 * greater than the sum of the retransmission value of the peer.
2966 * See [SCTP] for more information.
2969 static int sctp_setsockopt_associnfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2972 struct sctp_assocparams assocparams
;
2973 struct sctp_association
*asoc
;
2975 if (optlen
!= sizeof(struct sctp_assocparams
))
2977 if (copy_from_user(&assocparams
, optval
, optlen
))
2980 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
2982 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
2985 /* Set the values to the specific association */
2987 if (assocparams
.sasoc_asocmaxrxt
!= 0) {
2990 struct sctp_transport
*peer_addr
;
2992 list_for_each_entry(peer_addr
, &asoc
->peer
.transport_addr_list
,
2994 path_sum
+= peer_addr
->pathmaxrxt
;
2998 /* Only validate asocmaxrxt if we have more than
2999 * one path/transport. We do this because path
3000 * retransmissions are only counted when we have more
3004 assocparams
.sasoc_asocmaxrxt
> path_sum
)
3007 asoc
->max_retrans
= assocparams
.sasoc_asocmaxrxt
;
3010 if (assocparams
.sasoc_cookie_life
!= 0)
3011 asoc
->cookie_life
= ms_to_ktime(assocparams
.sasoc_cookie_life
);
3013 /* Set the values to the endpoint */
3014 struct sctp_sock
*sp
= sctp_sk(sk
);
3016 if (assocparams
.sasoc_asocmaxrxt
!= 0)
3017 sp
->assocparams
.sasoc_asocmaxrxt
=
3018 assocparams
.sasoc_asocmaxrxt
;
3019 if (assocparams
.sasoc_cookie_life
!= 0)
3020 sp
->assocparams
.sasoc_cookie_life
=
3021 assocparams
.sasoc_cookie_life
;
3027 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3029 * This socket option is a boolean flag which turns on or off mapped V4
3030 * addresses. If this option is turned on and the socket is type
3031 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3032 * If this option is turned off, then no mapping will be done of V4
3033 * addresses and a user will receive both PF_INET6 and PF_INET type
3034 * addresses on the socket.
3036 static int sctp_setsockopt_mappedv4(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
3039 struct sctp_sock
*sp
= sctp_sk(sk
);
3041 if (optlen
< sizeof(int))
3043 if (get_user(val
, (int __user
*)optval
))
3054 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3055 * This option will get or set the maximum size to put in any outgoing
3056 * SCTP DATA chunk. If a message is larger than this size it will be
3057 * fragmented by SCTP into the specified size. Note that the underlying
3058 * SCTP implementation may fragment into smaller sized chunks when the
3059 * PMTU of the underlying association is smaller than the value set by
3060 * the user. The default value for this option is '0' which indicates
3061 * the user is NOT limiting fragmentation and only the PMTU will effect
3062 * SCTP's choice of DATA chunk size. Note also that values set larger
3063 * than the maximum size of an IP datagram will effectively let SCTP
3064 * control fragmentation (i.e. the same as setting this option to 0).
3066 * The following structure is used to access and modify this parameter:
3068 * struct sctp_assoc_value {
3069 * sctp_assoc_t assoc_id;
3070 * uint32_t assoc_value;
3073 * assoc_id: This parameter is ignored for one-to-one style sockets.
3074 * For one-to-many style sockets this parameter indicates which
3075 * association the user is performing an action upon. Note that if
3076 * this field's value is zero then the endpoints default value is
3077 * changed (effecting future associations only).
3078 * assoc_value: This parameter specifies the maximum size in bytes.
3080 static int sctp_setsockopt_maxseg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
3082 struct sctp_assoc_value params
;
3083 struct sctp_association
*asoc
;
3084 struct sctp_sock
*sp
= sctp_sk(sk
);
3087 if (optlen
== sizeof(int)) {
3088 pr_warn_ratelimited(DEPRECATED
3090 "Use of int in maxseg socket option.\n"
3091 "Use struct sctp_assoc_value instead\n",
3092 current
->comm
, task_pid_nr(current
));
3093 if (copy_from_user(&val
, optval
, optlen
))
3095 params
.assoc_id
= 0;
3096 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
3097 if (copy_from_user(¶ms
, optval
, optlen
))
3099 val
= params
.assoc_value
;
3103 if ((val
!= 0) && ((val
< 8) || (val
> SCTP_MAX_CHUNK_LEN
)))
3106 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3107 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
3112 val
= asoc
->pathmtu
;
3113 val
-= sp
->pf
->af
->net_header_len
;
3114 val
-= sizeof(struct sctphdr
) +
3115 sizeof(struct sctp_data_chunk
);
3117 asoc
->user_frag
= val
;
3118 asoc
->frag_point
= sctp_frag_point(asoc
, asoc
->pathmtu
);
3120 sp
->user_frag
= val
;
3128 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3130 * Requests that the peer mark the enclosed address as the association
3131 * primary. The enclosed address must be one of the association's
3132 * locally bound addresses. The following structure is used to make a
3133 * set primary request:
3135 static int sctp_setsockopt_peer_primary_addr(struct sock
*sk
, char __user
*optval
,
3136 unsigned int optlen
)
3138 struct net
*net
= sock_net(sk
);
3139 struct sctp_sock
*sp
;
3140 struct sctp_association
*asoc
= NULL
;
3141 struct sctp_setpeerprim prim
;
3142 struct sctp_chunk
*chunk
;
3148 if (!net
->sctp
.addip_enable
)
3151 if (optlen
!= sizeof(struct sctp_setpeerprim
))
3154 if (copy_from_user(&prim
, optval
, optlen
))
3157 asoc
= sctp_id2assoc(sk
, prim
.sspp_assoc_id
);
3161 if (!asoc
->peer
.asconf_capable
)
3164 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_SET_PRIMARY
)
3167 if (!sctp_state(asoc
, ESTABLISHED
))
3170 af
= sctp_get_af_specific(prim
.sspp_addr
.ss_family
);
3174 if (!af
->addr_valid((union sctp_addr
*)&prim
.sspp_addr
, sp
, NULL
))
3175 return -EADDRNOTAVAIL
;
3177 if (!sctp_assoc_lookup_laddr(asoc
, (union sctp_addr
*)&prim
.sspp_addr
))
3178 return -EADDRNOTAVAIL
;
3180 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3181 chunk
= sctp_make_asconf_set_prim(asoc
,
3182 (union sctp_addr
*)&prim
.sspp_addr
);
3186 err
= sctp_send_asconf(asoc
, chunk
);
3188 pr_debug("%s: we set peer primary addr primitively\n", __func__
);
3193 static int sctp_setsockopt_adaptation_layer(struct sock
*sk
, char __user
*optval
,
3194 unsigned int optlen
)
3196 struct sctp_setadaptation adaptation
;
3198 if (optlen
!= sizeof(struct sctp_setadaptation
))
3200 if (copy_from_user(&adaptation
, optval
, optlen
))
3203 sctp_sk(sk
)->adaptation_ind
= adaptation
.ssb_adaptation_ind
;
3209 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3211 * The context field in the sctp_sndrcvinfo structure is normally only
3212 * used when a failed message is retrieved holding the value that was
3213 * sent down on the actual send call. This option allows the setting of
3214 * a default context on an association basis that will be received on
3215 * reading messages from the peer. This is especially helpful in the
3216 * one-2-many model for an application to keep some reference to an
3217 * internal state machine that is processing messages on the
3218 * association. Note that the setting of this value only effects
3219 * received messages from the peer and does not effect the value that is
3220 * saved with outbound messages.
3222 static int sctp_setsockopt_context(struct sock
*sk
, char __user
*optval
,
3223 unsigned int optlen
)
3225 struct sctp_assoc_value params
;
3226 struct sctp_sock
*sp
;
3227 struct sctp_association
*asoc
;
3229 if (optlen
!= sizeof(struct sctp_assoc_value
))
3231 if (copy_from_user(¶ms
, optval
, optlen
))
3236 if (params
.assoc_id
!= 0) {
3237 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3240 asoc
->default_rcv_context
= params
.assoc_value
;
3242 sp
->default_rcv_context
= params
.assoc_value
;
3249 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3251 * This options will at a minimum specify if the implementation is doing
3252 * fragmented interleave. Fragmented interleave, for a one to many
3253 * socket, is when subsequent calls to receive a message may return
3254 * parts of messages from different associations. Some implementations
3255 * may allow you to turn this value on or off. If so, when turned off,
3256 * no fragment interleave will occur (which will cause a head of line
3257 * blocking amongst multiple associations sharing the same one to many
3258 * socket). When this option is turned on, then each receive call may
3259 * come from a different association (thus the user must receive data
3260 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3261 * association each receive belongs to.
3263 * This option takes a boolean value. A non-zero value indicates that
3264 * fragmented interleave is on. A value of zero indicates that
3265 * fragmented interleave is off.
3267 * Note that it is important that an implementation that allows this
3268 * option to be turned on, have it off by default. Otherwise an unaware
3269 * application using the one to many model may become confused and act
3272 static int sctp_setsockopt_fragment_interleave(struct sock
*sk
,
3273 char __user
*optval
,
3274 unsigned int optlen
)
3278 if (optlen
!= sizeof(int))
3280 if (get_user(val
, (int __user
*)optval
))
3283 sctp_sk(sk
)->frag_interleave
= (val
== 0) ? 0 : 1;
3289 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3290 * (SCTP_PARTIAL_DELIVERY_POINT)
3292 * This option will set or get the SCTP partial delivery point. This
3293 * point is the size of a message where the partial delivery API will be
3294 * invoked to help free up rwnd space for the peer. Setting this to a
3295 * lower value will cause partial deliveries to happen more often. The
3296 * calls argument is an integer that sets or gets the partial delivery
3297 * point. Note also that the call will fail if the user attempts to set
3298 * this value larger than the socket receive buffer size.
3300 * Note that any single message having a length smaller than or equal to
3301 * the SCTP partial delivery point will be delivered in one single read
3302 * call as long as the user provided buffer is large enough to hold the
3305 static int sctp_setsockopt_partial_delivery_point(struct sock
*sk
,
3306 char __user
*optval
,
3307 unsigned int optlen
)
3311 if (optlen
!= sizeof(u32
))
3313 if (get_user(val
, (int __user
*)optval
))
3316 /* Note: We double the receive buffer from what the user sets
3317 * it to be, also initial rwnd is based on rcvbuf/2.
3319 if (val
> (sk
->sk_rcvbuf
>> 1))
3322 sctp_sk(sk
)->pd_point
= val
;
3324 return 0; /* is this the right error code? */
3328 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3330 * This option will allow a user to change the maximum burst of packets
3331 * that can be emitted by this association. Note that the default value
3332 * is 4, and some implementations may restrict this setting so that it
3333 * can only be lowered.
3335 * NOTE: This text doesn't seem right. Do this on a socket basis with
3336 * future associations inheriting the socket value.
3338 static int sctp_setsockopt_maxburst(struct sock
*sk
,
3339 char __user
*optval
,
3340 unsigned int optlen
)
3342 struct sctp_assoc_value params
;
3343 struct sctp_sock
*sp
;
3344 struct sctp_association
*asoc
;
3348 if (optlen
== sizeof(int)) {
3349 pr_warn_ratelimited(DEPRECATED
3351 "Use of int in max_burst socket option deprecated.\n"
3352 "Use struct sctp_assoc_value instead\n",
3353 current
->comm
, task_pid_nr(current
));
3354 if (copy_from_user(&val
, optval
, optlen
))
3356 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
3357 if (copy_from_user(¶ms
, optval
, optlen
))
3359 val
= params
.assoc_value
;
3360 assoc_id
= params
.assoc_id
;
3366 if (assoc_id
!= 0) {
3367 asoc
= sctp_id2assoc(sk
, assoc_id
);
3370 asoc
->max_burst
= val
;
3372 sp
->max_burst
= val
;
3378 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3380 * This set option adds a chunk type that the user is requesting to be
3381 * received only in an authenticated way. Changes to the list of chunks
3382 * will only effect future associations on the socket.
3384 static int sctp_setsockopt_auth_chunk(struct sock
*sk
,
3385 char __user
*optval
,
3386 unsigned int optlen
)
3388 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3389 struct sctp_authchunk val
;
3391 if (!ep
->auth_enable
)
3394 if (optlen
!= sizeof(struct sctp_authchunk
))
3396 if (copy_from_user(&val
, optval
, optlen
))
3399 switch (val
.sauth_chunk
) {
3401 case SCTP_CID_INIT_ACK
:
3402 case SCTP_CID_SHUTDOWN_COMPLETE
:
3407 /* add this chunk id to the endpoint */
3408 return sctp_auth_ep_add_chunkid(ep
, val
.sauth_chunk
);
3412 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3414 * This option gets or sets the list of HMAC algorithms that the local
3415 * endpoint requires the peer to use.
3417 static int sctp_setsockopt_hmac_ident(struct sock
*sk
,
3418 char __user
*optval
,
3419 unsigned int optlen
)
3421 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3422 struct sctp_hmacalgo
*hmacs
;
3426 if (!ep
->auth_enable
)
3429 if (optlen
< sizeof(struct sctp_hmacalgo
))
3432 hmacs
= memdup_user(optval
, optlen
);
3434 return PTR_ERR(hmacs
);
3436 idents
= hmacs
->shmac_num_idents
;
3437 if (idents
== 0 || idents
> SCTP_AUTH_NUM_HMACS
||
3438 (idents
* sizeof(u16
)) > (optlen
- sizeof(struct sctp_hmacalgo
))) {
3443 err
= sctp_auth_ep_set_hmacs(ep
, hmacs
);
3450 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3452 * This option will set a shared secret key which is used to build an
3453 * association shared key.
3455 static int sctp_setsockopt_auth_key(struct sock
*sk
,
3456 char __user
*optval
,
3457 unsigned int optlen
)
3459 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3460 struct sctp_authkey
*authkey
;
3461 struct sctp_association
*asoc
;
3464 if (!ep
->auth_enable
)
3467 if (optlen
<= sizeof(struct sctp_authkey
))
3470 authkey
= memdup_user(optval
, optlen
);
3471 if (IS_ERR(authkey
))
3472 return PTR_ERR(authkey
);
3474 if (authkey
->sca_keylength
> optlen
- sizeof(struct sctp_authkey
)) {
3479 asoc
= sctp_id2assoc(sk
, authkey
->sca_assoc_id
);
3480 if (!asoc
&& authkey
->sca_assoc_id
&& sctp_style(sk
, UDP
)) {
3485 ret
= sctp_auth_set_key(ep
, asoc
, authkey
);
3492 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3494 * This option will get or set the active shared key to be used to build
3495 * the association shared key.
3497 static int sctp_setsockopt_active_key(struct sock
*sk
,
3498 char __user
*optval
,
3499 unsigned int optlen
)
3501 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3502 struct sctp_authkeyid val
;
3503 struct sctp_association
*asoc
;
3505 if (!ep
->auth_enable
)
3508 if (optlen
!= sizeof(struct sctp_authkeyid
))
3510 if (copy_from_user(&val
, optval
, optlen
))
3513 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3514 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3517 return sctp_auth_set_active_key(ep
, asoc
, val
.scact_keynumber
);
3521 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3523 * This set option will delete a shared secret key from use.
3525 static int sctp_setsockopt_del_key(struct sock
*sk
,
3526 char __user
*optval
,
3527 unsigned int optlen
)
3529 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3530 struct sctp_authkeyid val
;
3531 struct sctp_association
*asoc
;
3533 if (!ep
->auth_enable
)
3536 if (optlen
!= sizeof(struct sctp_authkeyid
))
3538 if (copy_from_user(&val
, optval
, optlen
))
3541 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3542 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3545 return sctp_auth_del_key_id(ep
, asoc
, val
.scact_keynumber
);
3550 * 8.1.23 SCTP_AUTO_ASCONF
3552 * This option will enable or disable the use of the automatic generation of
3553 * ASCONF chunks to add and delete addresses to an existing association. Note
3554 * that this option has two caveats namely: a) it only affects sockets that
3555 * are bound to all addresses available to the SCTP stack, and b) the system
3556 * administrator may have an overriding control that turns the ASCONF feature
3557 * off no matter what setting the socket option may have.
3558 * This option expects an integer boolean flag, where a non-zero value turns on
3559 * the option, and a zero value turns off the option.
3560 * Note. In this implementation, socket operation overrides default parameter
3561 * being set by sysctl as well as FreeBSD implementation
3563 static int sctp_setsockopt_auto_asconf(struct sock
*sk
, char __user
*optval
,
3564 unsigned int optlen
)
3567 struct sctp_sock
*sp
= sctp_sk(sk
);
3569 if (optlen
< sizeof(int))
3571 if (get_user(val
, (int __user
*)optval
))
3573 if (!sctp_is_ep_boundall(sk
) && val
)
3575 if ((val
&& sp
->do_auto_asconf
) || (!val
&& !sp
->do_auto_asconf
))
3578 spin_lock_bh(&sock_net(sk
)->sctp
.addr_wq_lock
);
3579 if (val
== 0 && sp
->do_auto_asconf
) {
3580 list_del(&sp
->auto_asconf_list
);
3581 sp
->do_auto_asconf
= 0;
3582 } else if (val
&& !sp
->do_auto_asconf
) {
3583 list_add_tail(&sp
->auto_asconf_list
,
3584 &sock_net(sk
)->sctp
.auto_asconf_splist
);
3585 sp
->do_auto_asconf
= 1;
3587 spin_unlock_bh(&sock_net(sk
)->sctp
.addr_wq_lock
);
3592 * SCTP_PEER_ADDR_THLDS
3594 * This option allows us to alter the partially failed threshold for one or all
3595 * transports in an association. See Section 6.1 of:
3596 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3598 static int sctp_setsockopt_paddr_thresholds(struct sock
*sk
,
3599 char __user
*optval
,
3600 unsigned int optlen
)
3602 struct sctp_paddrthlds val
;
3603 struct sctp_transport
*trans
;
3604 struct sctp_association
*asoc
;
3606 if (optlen
< sizeof(struct sctp_paddrthlds
))
3608 if (copy_from_user(&val
, (struct sctp_paddrthlds __user
*)optval
,
3609 sizeof(struct sctp_paddrthlds
)))
3613 if (sctp_is_any(sk
, (const union sctp_addr
*)&val
.spt_address
)) {
3614 asoc
= sctp_id2assoc(sk
, val
.spt_assoc_id
);
3617 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
3619 if (val
.spt_pathmaxrxt
)
3620 trans
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3621 trans
->pf_retrans
= val
.spt_pathpfthld
;
3624 if (val
.spt_pathmaxrxt
)
3625 asoc
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3626 asoc
->pf_retrans
= val
.spt_pathpfthld
;
3628 trans
= sctp_addr_id2transport(sk
, &val
.spt_address
,
3633 if (val
.spt_pathmaxrxt
)
3634 trans
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3635 trans
->pf_retrans
= val
.spt_pathpfthld
;
3641 static int sctp_setsockopt_recvrcvinfo(struct sock
*sk
,
3642 char __user
*optval
,
3643 unsigned int optlen
)
3647 if (optlen
< sizeof(int))
3649 if (get_user(val
, (int __user
*) optval
))
3652 sctp_sk(sk
)->recvrcvinfo
= (val
== 0) ? 0 : 1;
3657 static int sctp_setsockopt_recvnxtinfo(struct sock
*sk
,
3658 char __user
*optval
,
3659 unsigned int optlen
)
3663 if (optlen
< sizeof(int))
3665 if (get_user(val
, (int __user
*) optval
))
3668 sctp_sk(sk
)->recvnxtinfo
= (val
== 0) ? 0 : 1;
3673 static int sctp_setsockopt_pr_supported(struct sock
*sk
,
3674 char __user
*optval
,
3675 unsigned int optlen
)
3677 struct sctp_assoc_value params
;
3678 struct sctp_association
*asoc
;
3679 int retval
= -EINVAL
;
3681 if (optlen
!= sizeof(params
))
3684 if (copy_from_user(¶ms
, optval
, optlen
)) {
3689 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3691 asoc
->prsctp_enable
= !!params
.assoc_value
;
3692 } else if (!params
.assoc_id
) {
3693 struct sctp_sock
*sp
= sctp_sk(sk
);
3695 sp
->ep
->prsctp_enable
= !!params
.assoc_value
;
3706 static int sctp_setsockopt_default_prinfo(struct sock
*sk
,
3707 char __user
*optval
,
3708 unsigned int optlen
)
3710 struct sctp_default_prinfo info
;
3711 struct sctp_association
*asoc
;
3712 int retval
= -EINVAL
;
3714 if (optlen
!= sizeof(info
))
3717 if (copy_from_user(&info
, optval
, sizeof(info
))) {
3722 if (info
.pr_policy
& ~SCTP_PR_SCTP_MASK
)
3725 if (info
.pr_policy
== SCTP_PR_SCTP_NONE
)
3728 asoc
= sctp_id2assoc(sk
, info
.pr_assoc_id
);
3730 SCTP_PR_SET_POLICY(asoc
->default_flags
, info
.pr_policy
);
3731 asoc
->default_timetolive
= info
.pr_value
;
3732 } else if (!info
.pr_assoc_id
) {
3733 struct sctp_sock
*sp
= sctp_sk(sk
);
3735 SCTP_PR_SET_POLICY(sp
->default_flags
, info
.pr_policy
);
3736 sp
->default_timetolive
= info
.pr_value
;
3747 /* API 6.2 setsockopt(), getsockopt()
3749 * Applications use setsockopt() and getsockopt() to set or retrieve
3750 * socket options. Socket options are used to change the default
3751 * behavior of sockets calls. They are described in Section 7.
3755 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3756 * int __user *optlen);
3757 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3760 * sd - the socket descript.
3761 * level - set to IPPROTO_SCTP for all SCTP options.
3762 * optname - the option name.
3763 * optval - the buffer to store the value of the option.
3764 * optlen - the size of the buffer.
3766 static int sctp_setsockopt(struct sock
*sk
, int level
, int optname
,
3767 char __user
*optval
, unsigned int optlen
)
3771 pr_debug("%s: sk:%p, optname:%d\n", __func__
, sk
, optname
);
3773 /* I can hardly begin to describe how wrong this is. This is
3774 * so broken as to be worse than useless. The API draft
3775 * REALLY is NOT helpful here... I am not convinced that the
3776 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3777 * are at all well-founded.
3779 if (level
!= SOL_SCTP
) {
3780 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
3781 retval
= af
->setsockopt(sk
, level
, optname
, optval
, optlen
);
3788 case SCTP_SOCKOPT_BINDX_ADD
:
3789 /* 'optlen' is the size of the addresses buffer. */
3790 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3791 optlen
, SCTP_BINDX_ADD_ADDR
);
3794 case SCTP_SOCKOPT_BINDX_REM
:
3795 /* 'optlen' is the size of the addresses buffer. */
3796 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3797 optlen
, SCTP_BINDX_REM_ADDR
);
3800 case SCTP_SOCKOPT_CONNECTX_OLD
:
3801 /* 'optlen' is the size of the addresses buffer. */
3802 retval
= sctp_setsockopt_connectx_old(sk
,
3803 (struct sockaddr __user
*)optval
,
3807 case SCTP_SOCKOPT_CONNECTX
:
3808 /* 'optlen' is the size of the addresses buffer. */
3809 retval
= sctp_setsockopt_connectx(sk
,
3810 (struct sockaddr __user
*)optval
,
3814 case SCTP_DISABLE_FRAGMENTS
:
3815 retval
= sctp_setsockopt_disable_fragments(sk
, optval
, optlen
);
3819 retval
= sctp_setsockopt_events(sk
, optval
, optlen
);
3822 case SCTP_AUTOCLOSE
:
3823 retval
= sctp_setsockopt_autoclose(sk
, optval
, optlen
);
3826 case SCTP_PEER_ADDR_PARAMS
:
3827 retval
= sctp_setsockopt_peer_addr_params(sk
, optval
, optlen
);
3830 case SCTP_DELAYED_SACK
:
3831 retval
= sctp_setsockopt_delayed_ack(sk
, optval
, optlen
);
3833 case SCTP_PARTIAL_DELIVERY_POINT
:
3834 retval
= sctp_setsockopt_partial_delivery_point(sk
, optval
, optlen
);
3838 retval
= sctp_setsockopt_initmsg(sk
, optval
, optlen
);
3840 case SCTP_DEFAULT_SEND_PARAM
:
3841 retval
= sctp_setsockopt_default_send_param(sk
, optval
,
3844 case SCTP_DEFAULT_SNDINFO
:
3845 retval
= sctp_setsockopt_default_sndinfo(sk
, optval
, optlen
);
3847 case SCTP_PRIMARY_ADDR
:
3848 retval
= sctp_setsockopt_primary_addr(sk
, optval
, optlen
);
3850 case SCTP_SET_PEER_PRIMARY_ADDR
:
3851 retval
= sctp_setsockopt_peer_primary_addr(sk
, optval
, optlen
);
3854 retval
= sctp_setsockopt_nodelay(sk
, optval
, optlen
);
3857 retval
= sctp_setsockopt_rtoinfo(sk
, optval
, optlen
);
3859 case SCTP_ASSOCINFO
:
3860 retval
= sctp_setsockopt_associnfo(sk
, optval
, optlen
);
3862 case SCTP_I_WANT_MAPPED_V4_ADDR
:
3863 retval
= sctp_setsockopt_mappedv4(sk
, optval
, optlen
);
3866 retval
= sctp_setsockopt_maxseg(sk
, optval
, optlen
);
3868 case SCTP_ADAPTATION_LAYER
:
3869 retval
= sctp_setsockopt_adaptation_layer(sk
, optval
, optlen
);
3872 retval
= sctp_setsockopt_context(sk
, optval
, optlen
);
3874 case SCTP_FRAGMENT_INTERLEAVE
:
3875 retval
= sctp_setsockopt_fragment_interleave(sk
, optval
, optlen
);
3877 case SCTP_MAX_BURST
:
3878 retval
= sctp_setsockopt_maxburst(sk
, optval
, optlen
);
3880 case SCTP_AUTH_CHUNK
:
3881 retval
= sctp_setsockopt_auth_chunk(sk
, optval
, optlen
);
3883 case SCTP_HMAC_IDENT
:
3884 retval
= sctp_setsockopt_hmac_ident(sk
, optval
, optlen
);
3887 retval
= sctp_setsockopt_auth_key(sk
, optval
, optlen
);
3889 case SCTP_AUTH_ACTIVE_KEY
:
3890 retval
= sctp_setsockopt_active_key(sk
, optval
, optlen
);
3892 case SCTP_AUTH_DELETE_KEY
:
3893 retval
= sctp_setsockopt_del_key(sk
, optval
, optlen
);
3895 case SCTP_AUTO_ASCONF
:
3896 retval
= sctp_setsockopt_auto_asconf(sk
, optval
, optlen
);
3898 case SCTP_PEER_ADDR_THLDS
:
3899 retval
= sctp_setsockopt_paddr_thresholds(sk
, optval
, optlen
);
3901 case SCTP_RECVRCVINFO
:
3902 retval
= sctp_setsockopt_recvrcvinfo(sk
, optval
, optlen
);
3904 case SCTP_RECVNXTINFO
:
3905 retval
= sctp_setsockopt_recvnxtinfo(sk
, optval
, optlen
);
3907 case SCTP_PR_SUPPORTED
:
3908 retval
= sctp_setsockopt_pr_supported(sk
, optval
, optlen
);
3910 case SCTP_DEFAULT_PRINFO
:
3911 retval
= sctp_setsockopt_default_prinfo(sk
, optval
, optlen
);
3914 retval
= -ENOPROTOOPT
;
3924 /* API 3.1.6 connect() - UDP Style Syntax
3926 * An application may use the connect() call in the UDP model to initiate an
3927 * association without sending data.
3931 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3933 * sd: the socket descriptor to have a new association added to.
3935 * nam: the address structure (either struct sockaddr_in or struct
3936 * sockaddr_in6 defined in RFC2553 [7]).
3938 * len: the size of the address.
3940 static int sctp_connect(struct sock
*sk
, struct sockaddr
*addr
,
3948 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__
, sk
,
3951 /* Validate addr_len before calling common connect/connectx routine. */
3952 af
= sctp_get_af_specific(addr
->sa_family
);
3953 if (!af
|| addr_len
< af
->sockaddr_len
) {
3956 /* Pass correct addr len to common routine (so it knows there
3957 * is only one address being passed.
3959 err
= __sctp_connect(sk
, addr
, af
->sockaddr_len
, NULL
);
3966 /* FIXME: Write comments. */
3967 static int sctp_disconnect(struct sock
*sk
, int flags
)
3969 return -EOPNOTSUPP
; /* STUB */
3972 /* 4.1.4 accept() - TCP Style Syntax
3974 * Applications use accept() call to remove an established SCTP
3975 * association from the accept queue of the endpoint. A new socket
3976 * descriptor will be returned from accept() to represent the newly
3977 * formed association.
3979 static struct sock
*sctp_accept(struct sock
*sk
, int flags
, int *err
)
3981 struct sctp_sock
*sp
;
3982 struct sctp_endpoint
*ep
;
3983 struct sock
*newsk
= NULL
;
3984 struct sctp_association
*asoc
;
3993 if (!sctp_style(sk
, TCP
)) {
3994 error
= -EOPNOTSUPP
;
3998 if (!sctp_sstate(sk
, LISTENING
)) {
4003 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
4005 error
= sctp_wait_for_accept(sk
, timeo
);
4009 /* We treat the list of associations on the endpoint as the accept
4010 * queue and pick the first association on the list.
4012 asoc
= list_entry(ep
->asocs
.next
, struct sctp_association
, asocs
);
4014 newsk
= sp
->pf
->create_accept_sk(sk
, asoc
);
4020 /* Populate the fields of the newsk from the oldsk and migrate the
4021 * asoc to the newsk.
4023 sctp_sock_migrate(sk
, newsk
, asoc
, SCTP_SOCKET_TCP
);
4031 /* The SCTP ioctl handler. */
4032 static int sctp_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
4039 * SEQPACKET-style sockets in LISTENING state are valid, for
4040 * SCTP, so only discard TCP-style sockets in LISTENING state.
4042 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
4047 struct sk_buff
*skb
;
4048 unsigned int amount
= 0;
4050 skb
= skb_peek(&sk
->sk_receive_queue
);
4053 * We will only return the amount of this packet since
4054 * that is all that will be read.
4058 rc
= put_user(amount
, (int __user
*)arg
);
4070 /* This is the function which gets called during socket creation to
4071 * initialized the SCTP-specific portion of the sock.
4072 * The sock structure should already be zero-filled memory.
4074 static int sctp_init_sock(struct sock
*sk
)
4076 struct net
*net
= sock_net(sk
);
4077 struct sctp_sock
*sp
;
4079 pr_debug("%s: sk:%p\n", __func__
, sk
);
4083 /* Initialize the SCTP per socket area. */
4084 switch (sk
->sk_type
) {
4085 case SOCK_SEQPACKET
:
4086 sp
->type
= SCTP_SOCKET_UDP
;
4089 sp
->type
= SCTP_SOCKET_TCP
;
4092 return -ESOCKTNOSUPPORT
;
4095 sk
->sk_gso_type
= SKB_GSO_SCTP
;
4097 /* Initialize default send parameters. These parameters can be
4098 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4100 sp
->default_stream
= 0;
4101 sp
->default_ppid
= 0;
4102 sp
->default_flags
= 0;
4103 sp
->default_context
= 0;
4104 sp
->default_timetolive
= 0;
4106 sp
->default_rcv_context
= 0;
4107 sp
->max_burst
= net
->sctp
.max_burst
;
4109 sp
->sctp_hmac_alg
= net
->sctp
.sctp_hmac_alg
;
4111 /* Initialize default setup parameters. These parameters
4112 * can be modified with the SCTP_INITMSG socket option or
4113 * overridden by the SCTP_INIT CMSG.
4115 sp
->initmsg
.sinit_num_ostreams
= sctp_max_outstreams
;
4116 sp
->initmsg
.sinit_max_instreams
= sctp_max_instreams
;
4117 sp
->initmsg
.sinit_max_attempts
= net
->sctp
.max_retrans_init
;
4118 sp
->initmsg
.sinit_max_init_timeo
= net
->sctp
.rto_max
;
4120 /* Initialize default RTO related parameters. These parameters can
4121 * be modified for with the SCTP_RTOINFO socket option.
4123 sp
->rtoinfo
.srto_initial
= net
->sctp
.rto_initial
;
4124 sp
->rtoinfo
.srto_max
= net
->sctp
.rto_max
;
4125 sp
->rtoinfo
.srto_min
= net
->sctp
.rto_min
;
4127 /* Initialize default association related parameters. These parameters
4128 * can be modified with the SCTP_ASSOCINFO socket option.
4130 sp
->assocparams
.sasoc_asocmaxrxt
= net
->sctp
.max_retrans_association
;
4131 sp
->assocparams
.sasoc_number_peer_destinations
= 0;
4132 sp
->assocparams
.sasoc_peer_rwnd
= 0;
4133 sp
->assocparams
.sasoc_local_rwnd
= 0;
4134 sp
->assocparams
.sasoc_cookie_life
= net
->sctp
.valid_cookie_life
;
4136 /* Initialize default event subscriptions. By default, all the
4139 memset(&sp
->subscribe
, 0, sizeof(struct sctp_event_subscribe
));
4141 /* Default Peer Address Parameters. These defaults can
4142 * be modified via SCTP_PEER_ADDR_PARAMS
4144 sp
->hbinterval
= net
->sctp
.hb_interval
;
4145 sp
->pathmaxrxt
= net
->sctp
.max_retrans_path
;
4146 sp
->pathmtu
= 0; /* allow default discovery */
4147 sp
->sackdelay
= net
->sctp
.sack_timeout
;
4149 sp
->param_flags
= SPP_HB_ENABLE
|
4151 SPP_SACKDELAY_ENABLE
;
4153 /* If enabled no SCTP message fragmentation will be performed.
4154 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4156 sp
->disable_fragments
= 0;
4158 /* Enable Nagle algorithm by default. */
4161 sp
->recvrcvinfo
= 0;
4162 sp
->recvnxtinfo
= 0;
4164 /* Enable by default. */
4167 /* Auto-close idle associations after the configured
4168 * number of seconds. A value of 0 disables this
4169 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4170 * for UDP-style sockets only.
4174 /* User specified fragmentation limit. */
4177 sp
->adaptation_ind
= 0;
4179 sp
->pf
= sctp_get_pf_specific(sk
->sk_family
);
4181 /* Control variables for partial data delivery. */
4182 atomic_set(&sp
->pd_mode
, 0);
4183 skb_queue_head_init(&sp
->pd_lobby
);
4184 sp
->frag_interleave
= 0;
4186 /* Create a per socket endpoint structure. Even if we
4187 * change the data structure relationships, this may still
4188 * be useful for storing pre-connect address information.
4190 sp
->ep
= sctp_endpoint_new(sk
, GFP_KERNEL
);
4196 sk
->sk_destruct
= sctp_destruct_sock
;
4198 SCTP_DBG_OBJCNT_INC(sock
);
4201 percpu_counter_inc(&sctp_sockets_allocated
);
4202 sock_prot_inuse_add(net
, sk
->sk_prot
, 1);
4204 /* Nothing can fail after this block, otherwise
4205 * sctp_destroy_sock() will be called without addr_wq_lock held
4207 if (net
->sctp
.default_auto_asconf
) {
4208 spin_lock(&sock_net(sk
)->sctp
.addr_wq_lock
);
4209 list_add_tail(&sp
->auto_asconf_list
,
4210 &net
->sctp
.auto_asconf_splist
);
4211 sp
->do_auto_asconf
= 1;
4212 spin_unlock(&sock_net(sk
)->sctp
.addr_wq_lock
);
4214 sp
->do_auto_asconf
= 0;
4222 /* Cleanup any SCTP per socket resources. Must be called with
4223 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4225 static void sctp_destroy_sock(struct sock
*sk
)
4227 struct sctp_sock
*sp
;
4229 pr_debug("%s: sk:%p\n", __func__
, sk
);
4231 /* Release our hold on the endpoint. */
4233 /* This could happen during socket init, thus we bail out
4234 * early, since the rest of the below is not setup either.
4239 if (sp
->do_auto_asconf
) {
4240 sp
->do_auto_asconf
= 0;
4241 list_del(&sp
->auto_asconf_list
);
4243 sctp_endpoint_free(sp
->ep
);
4245 percpu_counter_dec(&sctp_sockets_allocated
);
4246 sock_prot_inuse_add(sock_net(sk
), sk
->sk_prot
, -1);
4250 /* Triggered when there are no references on the socket anymore */
4251 static void sctp_destruct_sock(struct sock
*sk
)
4253 struct sctp_sock
*sp
= sctp_sk(sk
);
4255 /* Free up the HMAC transform. */
4256 crypto_free_shash(sp
->hmac
);
4258 inet_sock_destruct(sk
);
4261 /* API 4.1.7 shutdown() - TCP Style Syntax
4262 * int shutdown(int socket, int how);
4264 * sd - the socket descriptor of the association to be closed.
4265 * how - Specifies the type of shutdown. The values are
4268 * Disables further receive operations. No SCTP
4269 * protocol action is taken.
4271 * Disables further send operations, and initiates
4272 * the SCTP shutdown sequence.
4274 * Disables further send and receive operations
4275 * and initiates the SCTP shutdown sequence.
4277 static void sctp_shutdown(struct sock
*sk
, int how
)
4279 struct net
*net
= sock_net(sk
);
4280 struct sctp_endpoint
*ep
;
4281 struct sctp_association
*asoc
;
4283 if (!sctp_style(sk
, TCP
))
4286 if (how
& SEND_SHUTDOWN
) {
4287 sk
->sk_state
= SCTP_SS_CLOSING
;
4288 ep
= sctp_sk(sk
)->ep
;
4289 if (!list_empty(&ep
->asocs
)) {
4290 asoc
= list_entry(ep
->asocs
.next
,
4291 struct sctp_association
, asocs
);
4292 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
4297 int sctp_get_sctp_info(struct sock
*sk
, struct sctp_association
*asoc
,
4298 struct sctp_info
*info
)
4300 struct sctp_transport
*prim
;
4301 struct list_head
*pos
;
4304 memset(info
, 0, sizeof(*info
));
4306 struct sctp_sock
*sp
= sctp_sk(sk
);
4308 info
->sctpi_s_autoclose
= sp
->autoclose
;
4309 info
->sctpi_s_adaptation_ind
= sp
->adaptation_ind
;
4310 info
->sctpi_s_pd_point
= sp
->pd_point
;
4311 info
->sctpi_s_nodelay
= sp
->nodelay
;
4312 info
->sctpi_s_disable_fragments
= sp
->disable_fragments
;
4313 info
->sctpi_s_v4mapped
= sp
->v4mapped
;
4314 info
->sctpi_s_frag_interleave
= sp
->frag_interleave
;
4315 info
->sctpi_s_type
= sp
->type
;
4320 info
->sctpi_tag
= asoc
->c
.my_vtag
;
4321 info
->sctpi_state
= asoc
->state
;
4322 info
->sctpi_rwnd
= asoc
->a_rwnd
;
4323 info
->sctpi_unackdata
= asoc
->unack_data
;
4324 info
->sctpi_penddata
= sctp_tsnmap_pending(&asoc
->peer
.tsn_map
);
4325 info
->sctpi_instrms
= asoc
->c
.sinit_max_instreams
;
4326 info
->sctpi_outstrms
= asoc
->c
.sinit_num_ostreams
;
4327 list_for_each(pos
, &asoc
->base
.inqueue
.in_chunk_list
)
4328 info
->sctpi_inqueue
++;
4329 list_for_each(pos
, &asoc
->outqueue
.out_chunk_list
)
4330 info
->sctpi_outqueue
++;
4331 info
->sctpi_overall_error
= asoc
->overall_error_count
;
4332 info
->sctpi_max_burst
= asoc
->max_burst
;
4333 info
->sctpi_maxseg
= asoc
->frag_point
;
4334 info
->sctpi_peer_rwnd
= asoc
->peer
.rwnd
;
4335 info
->sctpi_peer_tag
= asoc
->c
.peer_vtag
;
4337 mask
= asoc
->peer
.ecn_capable
<< 1;
4338 mask
= (mask
| asoc
->peer
.ipv4_address
) << 1;
4339 mask
= (mask
| asoc
->peer
.ipv6_address
) << 1;
4340 mask
= (mask
| asoc
->peer
.hostname_address
) << 1;
4341 mask
= (mask
| asoc
->peer
.asconf_capable
) << 1;
4342 mask
= (mask
| asoc
->peer
.prsctp_capable
) << 1;
4343 mask
= (mask
| asoc
->peer
.auth_capable
);
4344 info
->sctpi_peer_capable
= mask
;
4345 mask
= asoc
->peer
.sack_needed
<< 1;
4346 mask
= (mask
| asoc
->peer
.sack_generation
) << 1;
4347 mask
= (mask
| asoc
->peer
.zero_window_announced
);
4348 info
->sctpi_peer_sack
= mask
;
4350 info
->sctpi_isacks
= asoc
->stats
.isacks
;
4351 info
->sctpi_osacks
= asoc
->stats
.osacks
;
4352 info
->sctpi_opackets
= asoc
->stats
.opackets
;
4353 info
->sctpi_ipackets
= asoc
->stats
.ipackets
;
4354 info
->sctpi_rtxchunks
= asoc
->stats
.rtxchunks
;
4355 info
->sctpi_outofseqtsns
= asoc
->stats
.outofseqtsns
;
4356 info
->sctpi_idupchunks
= asoc
->stats
.idupchunks
;
4357 info
->sctpi_gapcnt
= asoc
->stats
.gapcnt
;
4358 info
->sctpi_ouodchunks
= asoc
->stats
.ouodchunks
;
4359 info
->sctpi_iuodchunks
= asoc
->stats
.iuodchunks
;
4360 info
->sctpi_oodchunks
= asoc
->stats
.oodchunks
;
4361 info
->sctpi_iodchunks
= asoc
->stats
.iodchunks
;
4362 info
->sctpi_octrlchunks
= asoc
->stats
.octrlchunks
;
4363 info
->sctpi_ictrlchunks
= asoc
->stats
.ictrlchunks
;
4365 prim
= asoc
->peer
.primary_path
;
4366 memcpy(&info
->sctpi_p_address
, &prim
->ipaddr
,
4367 sizeof(struct sockaddr_storage
));
4368 info
->sctpi_p_state
= prim
->state
;
4369 info
->sctpi_p_cwnd
= prim
->cwnd
;
4370 info
->sctpi_p_srtt
= prim
->srtt
;
4371 info
->sctpi_p_rto
= jiffies_to_msecs(prim
->rto
);
4372 info
->sctpi_p_hbinterval
= prim
->hbinterval
;
4373 info
->sctpi_p_pathmaxrxt
= prim
->pathmaxrxt
;
4374 info
->sctpi_p_sackdelay
= jiffies_to_msecs(prim
->sackdelay
);
4375 info
->sctpi_p_ssthresh
= prim
->ssthresh
;
4376 info
->sctpi_p_partial_bytes_acked
= prim
->partial_bytes_acked
;
4377 info
->sctpi_p_flight_size
= prim
->flight_size
;
4378 info
->sctpi_p_error
= prim
->error_count
;
4382 EXPORT_SYMBOL_GPL(sctp_get_sctp_info
);
4384 /* use callback to avoid exporting the core structure */
4385 int sctp_transport_walk_start(struct rhashtable_iter
*iter
)
4389 err
= rhashtable_walk_init(&sctp_transport_hashtable
, iter
,
4394 err
= rhashtable_walk_start(iter
);
4395 if (err
&& err
!= -EAGAIN
) {
4396 rhashtable_walk_stop(iter
);
4397 rhashtable_walk_exit(iter
);
4404 void sctp_transport_walk_stop(struct rhashtable_iter
*iter
)
4406 rhashtable_walk_stop(iter
);
4407 rhashtable_walk_exit(iter
);
4410 struct sctp_transport
*sctp_transport_get_next(struct net
*net
,
4411 struct rhashtable_iter
*iter
)
4413 struct sctp_transport
*t
;
4415 t
= rhashtable_walk_next(iter
);
4416 for (; t
; t
= rhashtable_walk_next(iter
)) {
4418 if (PTR_ERR(t
) == -EAGAIN
)
4423 if (net_eq(sock_net(t
->asoc
->base
.sk
), net
) &&
4424 t
->asoc
->peer
.primary_path
== t
)
4431 struct sctp_transport
*sctp_transport_get_idx(struct net
*net
,
4432 struct rhashtable_iter
*iter
,
4435 void *obj
= SEQ_START_TOKEN
;
4437 while (pos
&& (obj
= sctp_transport_get_next(net
, iter
)) &&
4444 int sctp_for_each_endpoint(int (*cb
)(struct sctp_endpoint
*, void *),
4448 struct sctp_ep_common
*epb
;
4449 struct sctp_hashbucket
*head
;
4451 for (head
= sctp_ep_hashtable
; hash
< sctp_ep_hashsize
;
4453 read_lock(&head
->lock
);
4454 sctp_for_each_hentry(epb
, &head
->chain
) {
4455 err
= cb(sctp_ep(epb
), p
);
4459 read_unlock(&head
->lock
);
4464 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint
);
4466 int sctp_transport_lookup_process(int (*cb
)(struct sctp_transport
*, void *),
4468 const union sctp_addr
*laddr
,
4469 const union sctp_addr
*paddr
, void *p
)
4471 struct sctp_transport
*transport
;
4475 transport
= sctp_addrs_lookup_transport(net
, laddr
, paddr
);
4476 if (!transport
|| !sctp_transport_hold(transport
))
4478 err
= cb(transport
, p
);
4479 sctp_transport_put(transport
);
4485 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process
);
4487 int sctp_for_each_transport(int (*cb
)(struct sctp_transport
*, void *),
4488 struct net
*net
, int pos
, void *p
) {
4489 struct rhashtable_iter hti
;
4493 err
= sctp_transport_walk_start(&hti
);
4497 sctp_transport_get_idx(net
, &hti
, pos
);
4498 obj
= sctp_transport_get_next(net
, &hti
);
4499 for (; obj
&& !IS_ERR(obj
); obj
= sctp_transport_get_next(net
, &hti
)) {
4500 struct sctp_transport
*transport
= obj
;
4502 if (!sctp_transport_hold(transport
))
4504 err
= cb(transport
, p
);
4505 sctp_transport_put(transport
);
4509 sctp_transport_walk_stop(&hti
);
4513 EXPORT_SYMBOL_GPL(sctp_for_each_transport
);
4515 /* 7.2.1 Association Status (SCTP_STATUS)
4517 * Applications can retrieve current status information about an
4518 * association, including association state, peer receiver window size,
4519 * number of unacked data chunks, and number of data chunks pending
4520 * receipt. This information is read-only.
4522 static int sctp_getsockopt_sctp_status(struct sock
*sk
, int len
,
4523 char __user
*optval
,
4526 struct sctp_status status
;
4527 struct sctp_association
*asoc
= NULL
;
4528 struct sctp_transport
*transport
;
4529 sctp_assoc_t associd
;
4532 if (len
< sizeof(status
)) {
4537 len
= sizeof(status
);
4538 if (copy_from_user(&status
, optval
, len
)) {
4543 associd
= status
.sstat_assoc_id
;
4544 asoc
= sctp_id2assoc(sk
, associd
);
4550 transport
= asoc
->peer
.primary_path
;
4552 status
.sstat_assoc_id
= sctp_assoc2id(asoc
);
4553 status
.sstat_state
= sctp_assoc_to_state(asoc
);
4554 status
.sstat_rwnd
= asoc
->peer
.rwnd
;
4555 status
.sstat_unackdata
= asoc
->unack_data
;
4557 status
.sstat_penddata
= sctp_tsnmap_pending(&asoc
->peer
.tsn_map
);
4558 status
.sstat_instrms
= asoc
->c
.sinit_max_instreams
;
4559 status
.sstat_outstrms
= asoc
->c
.sinit_num_ostreams
;
4560 status
.sstat_fragmentation_point
= asoc
->frag_point
;
4561 status
.sstat_primary
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
4562 memcpy(&status
.sstat_primary
.spinfo_address
, &transport
->ipaddr
,
4563 transport
->af_specific
->sockaddr_len
);
4564 /* Map ipv4 address into v4-mapped-on-v6 address. */
4565 sctp_get_pf_specific(sk
->sk_family
)->addr_to_user(sctp_sk(sk
),
4566 (union sctp_addr
*)&status
.sstat_primary
.spinfo_address
);
4567 status
.sstat_primary
.spinfo_state
= transport
->state
;
4568 status
.sstat_primary
.spinfo_cwnd
= transport
->cwnd
;
4569 status
.sstat_primary
.spinfo_srtt
= transport
->srtt
;
4570 status
.sstat_primary
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
4571 status
.sstat_primary
.spinfo_mtu
= transport
->pathmtu
;
4573 if (status
.sstat_primary
.spinfo_state
== SCTP_UNKNOWN
)
4574 status
.sstat_primary
.spinfo_state
= SCTP_ACTIVE
;
4576 if (put_user(len
, optlen
)) {
4581 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4582 __func__
, len
, status
.sstat_state
, status
.sstat_rwnd
,
4583 status
.sstat_assoc_id
);
4585 if (copy_to_user(optval
, &status
, len
)) {
4595 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4597 * Applications can retrieve information about a specific peer address
4598 * of an association, including its reachability state, congestion
4599 * window, and retransmission timer values. This information is
4602 static int sctp_getsockopt_peer_addr_info(struct sock
*sk
, int len
,
4603 char __user
*optval
,
4606 struct sctp_paddrinfo pinfo
;
4607 struct sctp_transport
*transport
;
4610 if (len
< sizeof(pinfo
)) {
4615 len
= sizeof(pinfo
);
4616 if (copy_from_user(&pinfo
, optval
, len
)) {
4621 transport
= sctp_addr_id2transport(sk
, &pinfo
.spinfo_address
,
4622 pinfo
.spinfo_assoc_id
);
4626 pinfo
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
4627 pinfo
.spinfo_state
= transport
->state
;
4628 pinfo
.spinfo_cwnd
= transport
->cwnd
;
4629 pinfo
.spinfo_srtt
= transport
->srtt
;
4630 pinfo
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
4631 pinfo
.spinfo_mtu
= transport
->pathmtu
;
4633 if (pinfo
.spinfo_state
== SCTP_UNKNOWN
)
4634 pinfo
.spinfo_state
= SCTP_ACTIVE
;
4636 if (put_user(len
, optlen
)) {
4641 if (copy_to_user(optval
, &pinfo
, len
)) {
4650 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4652 * This option is a on/off flag. If enabled no SCTP message
4653 * fragmentation will be performed. Instead if a message being sent
4654 * exceeds the current PMTU size, the message will NOT be sent and
4655 * instead a error will be indicated to the user.
4657 static int sctp_getsockopt_disable_fragments(struct sock
*sk
, int len
,
4658 char __user
*optval
, int __user
*optlen
)
4662 if (len
< sizeof(int))
4666 val
= (sctp_sk(sk
)->disable_fragments
== 1);
4667 if (put_user(len
, optlen
))
4669 if (copy_to_user(optval
, &val
, len
))
4674 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4676 * This socket option is used to specify various notifications and
4677 * ancillary data the user wishes to receive.
4679 static int sctp_getsockopt_events(struct sock
*sk
, int len
, char __user
*optval
,
4684 if (len
> sizeof(struct sctp_event_subscribe
))
4685 len
= sizeof(struct sctp_event_subscribe
);
4686 if (put_user(len
, optlen
))
4688 if (copy_to_user(optval
, &sctp_sk(sk
)->subscribe
, len
))
4693 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4695 * This socket option is applicable to the UDP-style socket only. When
4696 * set it will cause associations that are idle for more than the
4697 * specified number of seconds to automatically close. An association
4698 * being idle is defined an association that has NOT sent or received
4699 * user data. The special value of '0' indicates that no automatic
4700 * close of any associations should be performed. The option expects an
4701 * integer defining the number of seconds of idle time before an
4702 * association is closed.
4704 static int sctp_getsockopt_autoclose(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4706 /* Applicable to UDP-style socket only */
4707 if (sctp_style(sk
, TCP
))
4709 if (len
< sizeof(int))
4712 if (put_user(len
, optlen
))
4714 if (copy_to_user(optval
, &sctp_sk(sk
)->autoclose
, sizeof(int)))
4719 /* Helper routine to branch off an association to a new socket. */
4720 int sctp_do_peeloff(struct sock
*sk
, sctp_assoc_t id
, struct socket
**sockp
)
4722 struct sctp_association
*asoc
= sctp_id2assoc(sk
, id
);
4723 struct sctp_sock
*sp
= sctp_sk(sk
);
4724 struct socket
*sock
;
4730 /* An association cannot be branched off from an already peeled-off
4731 * socket, nor is this supported for tcp style sockets.
4733 if (!sctp_style(sk
, UDP
))
4736 /* Create a new socket. */
4737 err
= sock_create(sk
->sk_family
, SOCK_SEQPACKET
, IPPROTO_SCTP
, &sock
);
4741 sctp_copy_sock(sock
->sk
, sk
, asoc
);
4743 /* Make peeled-off sockets more like 1-1 accepted sockets.
4744 * Set the daddr and initialize id to something more random
4746 sp
->pf
->to_sk_daddr(&asoc
->peer
.primary_addr
, sk
);
4748 /* Populate the fields of the newsk from the oldsk and migrate the
4749 * asoc to the newsk.
4751 sctp_sock_migrate(sk
, sock
->sk
, asoc
, SCTP_SOCKET_UDP_HIGH_BANDWIDTH
);
4757 EXPORT_SYMBOL(sctp_do_peeloff
);
4759 static int sctp_getsockopt_peeloff(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4761 sctp_peeloff_arg_t peeloff
;
4762 struct socket
*newsock
;
4763 struct file
*newfile
;
4766 if (len
< sizeof(sctp_peeloff_arg_t
))
4768 len
= sizeof(sctp_peeloff_arg_t
);
4769 if (copy_from_user(&peeloff
, optval
, len
))
4772 retval
= sctp_do_peeloff(sk
, peeloff
.associd
, &newsock
);
4776 /* Map the socket to an unused fd that can be returned to the user. */
4777 retval
= get_unused_fd_flags(0);
4779 sock_release(newsock
);
4783 newfile
= sock_alloc_file(newsock
, 0, NULL
);
4784 if (IS_ERR(newfile
)) {
4785 put_unused_fd(retval
);
4786 sock_release(newsock
);
4787 return PTR_ERR(newfile
);
4790 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__
, sk
, newsock
->sk
,
4793 /* Return the fd mapped to the new socket. */
4794 if (put_user(len
, optlen
)) {
4796 put_unused_fd(retval
);
4799 peeloff
.sd
= retval
;
4800 if (copy_to_user(optval
, &peeloff
, len
)) {
4802 put_unused_fd(retval
);
4805 fd_install(retval
, newfile
);
4810 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4812 * Applications can enable or disable heartbeats for any peer address of
4813 * an association, modify an address's heartbeat interval, force a
4814 * heartbeat to be sent immediately, and adjust the address's maximum
4815 * number of retransmissions sent before an address is considered
4816 * unreachable. The following structure is used to access and modify an
4817 * address's parameters:
4819 * struct sctp_paddrparams {
4820 * sctp_assoc_t spp_assoc_id;
4821 * struct sockaddr_storage spp_address;
4822 * uint32_t spp_hbinterval;
4823 * uint16_t spp_pathmaxrxt;
4824 * uint32_t spp_pathmtu;
4825 * uint32_t spp_sackdelay;
4826 * uint32_t spp_flags;
4829 * spp_assoc_id - (one-to-many style socket) This is filled in the
4830 * application, and identifies the association for
4832 * spp_address - This specifies which address is of interest.
4833 * spp_hbinterval - This contains the value of the heartbeat interval,
4834 * in milliseconds. If a value of zero
4835 * is present in this field then no changes are to
4836 * be made to this parameter.
4837 * spp_pathmaxrxt - This contains the maximum number of
4838 * retransmissions before this address shall be
4839 * considered unreachable. If a value of zero
4840 * is present in this field then no changes are to
4841 * be made to this parameter.
4842 * spp_pathmtu - When Path MTU discovery is disabled the value
4843 * specified here will be the "fixed" path mtu.
4844 * Note that if the spp_address field is empty
4845 * then all associations on this address will
4846 * have this fixed path mtu set upon them.
4848 * spp_sackdelay - When delayed sack is enabled, this value specifies
4849 * the number of milliseconds that sacks will be delayed
4850 * for. This value will apply to all addresses of an
4851 * association if the spp_address field is empty. Note
4852 * also, that if delayed sack is enabled and this
4853 * value is set to 0, no change is made to the last
4854 * recorded delayed sack timer value.
4856 * spp_flags - These flags are used to control various features
4857 * on an association. The flag field may contain
4858 * zero or more of the following options.
4860 * SPP_HB_ENABLE - Enable heartbeats on the
4861 * specified address. Note that if the address
4862 * field is empty all addresses for the association
4863 * have heartbeats enabled upon them.
4865 * SPP_HB_DISABLE - Disable heartbeats on the
4866 * speicifed address. Note that if the address
4867 * field is empty all addresses for the association
4868 * will have their heartbeats disabled. Note also
4869 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4870 * mutually exclusive, only one of these two should
4871 * be specified. Enabling both fields will have
4872 * undetermined results.
4874 * SPP_HB_DEMAND - Request a user initiated heartbeat
4875 * to be made immediately.
4877 * SPP_PMTUD_ENABLE - This field will enable PMTU
4878 * discovery upon the specified address. Note that
4879 * if the address feild is empty then all addresses
4880 * on the association are effected.
4882 * SPP_PMTUD_DISABLE - This field will disable PMTU
4883 * discovery upon the specified address. Note that
4884 * if the address feild is empty then all addresses
4885 * on the association are effected. Not also that
4886 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4887 * exclusive. Enabling both will have undetermined
4890 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4891 * on delayed sack. The time specified in spp_sackdelay
4892 * is used to specify the sack delay for this address. Note
4893 * that if spp_address is empty then all addresses will
4894 * enable delayed sack and take on the sack delay
4895 * value specified in spp_sackdelay.
4896 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4897 * off delayed sack. If the spp_address field is blank then
4898 * delayed sack is disabled for the entire association. Note
4899 * also that this field is mutually exclusive to
4900 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4903 static int sctp_getsockopt_peer_addr_params(struct sock
*sk
, int len
,
4904 char __user
*optval
, int __user
*optlen
)
4906 struct sctp_paddrparams params
;
4907 struct sctp_transport
*trans
= NULL
;
4908 struct sctp_association
*asoc
= NULL
;
4909 struct sctp_sock
*sp
= sctp_sk(sk
);
4911 if (len
< sizeof(struct sctp_paddrparams
))
4913 len
= sizeof(struct sctp_paddrparams
);
4914 if (copy_from_user(¶ms
, optval
, len
))
4917 /* If an address other than INADDR_ANY is specified, and
4918 * no transport is found, then the request is invalid.
4920 if (!sctp_is_any(sk
, (union sctp_addr
*)¶ms
.spp_address
)) {
4921 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
4922 params
.spp_assoc_id
);
4924 pr_debug("%s: failed no transport\n", __func__
);
4929 /* Get association, if assoc_id != 0 and the socket is a one
4930 * to many style socket, and an association was not found, then
4931 * the id was invalid.
4933 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
4934 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
)) {
4935 pr_debug("%s: failed no association\n", __func__
);
4940 /* Fetch transport values. */
4941 params
.spp_hbinterval
= jiffies_to_msecs(trans
->hbinterval
);
4942 params
.spp_pathmtu
= trans
->pathmtu
;
4943 params
.spp_pathmaxrxt
= trans
->pathmaxrxt
;
4944 params
.spp_sackdelay
= jiffies_to_msecs(trans
->sackdelay
);
4946 /*draft-11 doesn't say what to return in spp_flags*/
4947 params
.spp_flags
= trans
->param_flags
;
4949 /* Fetch association values. */
4950 params
.spp_hbinterval
= jiffies_to_msecs(asoc
->hbinterval
);
4951 params
.spp_pathmtu
= asoc
->pathmtu
;
4952 params
.spp_pathmaxrxt
= asoc
->pathmaxrxt
;
4953 params
.spp_sackdelay
= jiffies_to_msecs(asoc
->sackdelay
);
4955 /*draft-11 doesn't say what to return in spp_flags*/
4956 params
.spp_flags
= asoc
->param_flags
;
4958 /* Fetch socket values. */
4959 params
.spp_hbinterval
= sp
->hbinterval
;
4960 params
.spp_pathmtu
= sp
->pathmtu
;
4961 params
.spp_sackdelay
= sp
->sackdelay
;
4962 params
.spp_pathmaxrxt
= sp
->pathmaxrxt
;
4964 /*draft-11 doesn't say what to return in spp_flags*/
4965 params
.spp_flags
= sp
->param_flags
;
4968 if (copy_to_user(optval
, ¶ms
, len
))
4971 if (put_user(len
, optlen
))
4978 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4980 * This option will effect the way delayed acks are performed. This
4981 * option allows you to get or set the delayed ack time, in
4982 * milliseconds. It also allows changing the delayed ack frequency.
4983 * Changing the frequency to 1 disables the delayed sack algorithm. If
4984 * the assoc_id is 0, then this sets or gets the endpoints default
4985 * values. If the assoc_id field is non-zero, then the set or get
4986 * effects the specified association for the one to many model (the
4987 * assoc_id field is ignored by the one to one model). Note that if
4988 * sack_delay or sack_freq are 0 when setting this option, then the
4989 * current values will remain unchanged.
4991 * struct sctp_sack_info {
4992 * sctp_assoc_t sack_assoc_id;
4993 * uint32_t sack_delay;
4994 * uint32_t sack_freq;
4997 * sack_assoc_id - This parameter, indicates which association the user
4998 * is performing an action upon. Note that if this field's value is
4999 * zero then the endpoints default value is changed (effecting future
5000 * associations only).
5002 * sack_delay - This parameter contains the number of milliseconds that
5003 * the user is requesting the delayed ACK timer be set to. Note that
5004 * this value is defined in the standard to be between 200 and 500
5007 * sack_freq - This parameter contains the number of packets that must
5008 * be received before a sack is sent without waiting for the delay
5009 * timer to expire. The default value for this is 2, setting this
5010 * value to 1 will disable the delayed sack algorithm.
5012 static int sctp_getsockopt_delayed_ack(struct sock
*sk
, int len
,
5013 char __user
*optval
,
5016 struct sctp_sack_info params
;
5017 struct sctp_association
*asoc
= NULL
;
5018 struct sctp_sock
*sp
= sctp_sk(sk
);
5020 if (len
>= sizeof(struct sctp_sack_info
)) {
5021 len
= sizeof(struct sctp_sack_info
);
5023 if (copy_from_user(¶ms
, optval
, len
))
5025 } else if (len
== sizeof(struct sctp_assoc_value
)) {
5026 pr_warn_ratelimited(DEPRECATED
5028 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
5029 "Use struct sctp_sack_info instead\n",
5030 current
->comm
, task_pid_nr(current
));
5031 if (copy_from_user(¶ms
, optval
, len
))
5036 /* Get association, if sack_assoc_id != 0 and the socket is a one
5037 * to many style socket, and an association was not found, then
5038 * the id was invalid.
5040 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
5041 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
5045 /* Fetch association values. */
5046 if (asoc
->param_flags
& SPP_SACKDELAY_ENABLE
) {
5047 params
.sack_delay
= jiffies_to_msecs(
5049 params
.sack_freq
= asoc
->sackfreq
;
5052 params
.sack_delay
= 0;
5053 params
.sack_freq
= 1;
5056 /* Fetch socket values. */
5057 if (sp
->param_flags
& SPP_SACKDELAY_ENABLE
) {
5058 params
.sack_delay
= sp
->sackdelay
;
5059 params
.sack_freq
= sp
->sackfreq
;
5061 params
.sack_delay
= 0;
5062 params
.sack_freq
= 1;
5066 if (copy_to_user(optval
, ¶ms
, len
))
5069 if (put_user(len
, optlen
))
5075 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5077 * Applications can specify protocol parameters for the default association
5078 * initialization. The option name argument to setsockopt() and getsockopt()
5081 * Setting initialization parameters is effective only on an unconnected
5082 * socket (for UDP-style sockets only future associations are effected
5083 * by the change). With TCP-style sockets, this option is inherited by
5084 * sockets derived from a listener socket.
5086 static int sctp_getsockopt_initmsg(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
5088 if (len
< sizeof(struct sctp_initmsg
))
5090 len
= sizeof(struct sctp_initmsg
);
5091 if (put_user(len
, optlen
))
5093 if (copy_to_user(optval
, &sctp_sk(sk
)->initmsg
, len
))
5099 static int sctp_getsockopt_peer_addrs(struct sock
*sk
, int len
,
5100 char __user
*optval
, int __user
*optlen
)
5102 struct sctp_association
*asoc
;
5104 struct sctp_getaddrs getaddrs
;
5105 struct sctp_transport
*from
;
5107 union sctp_addr temp
;
5108 struct sctp_sock
*sp
= sctp_sk(sk
);
5113 if (len
< sizeof(struct sctp_getaddrs
))
5116 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
5119 /* For UDP-style sockets, id specifies the association to query. */
5120 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
5124 to
= optval
+ offsetof(struct sctp_getaddrs
, addrs
);
5125 space_left
= len
- offsetof(struct sctp_getaddrs
, addrs
);
5127 list_for_each_entry(from
, &asoc
->peer
.transport_addr_list
,
5129 memcpy(&temp
, &from
->ipaddr
, sizeof(temp
));
5130 addrlen
= sctp_get_pf_specific(sk
->sk_family
)
5131 ->addr_to_user(sp
, &temp
);
5132 if (space_left
< addrlen
)
5134 if (copy_to_user(to
, &temp
, addrlen
))
5138 space_left
-= addrlen
;
5141 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
))
5143 bytes_copied
= ((char __user
*)to
) - optval
;
5144 if (put_user(bytes_copied
, optlen
))
5150 static int sctp_copy_laddrs(struct sock
*sk
, __u16 port
, void *to
,
5151 size_t space_left
, int *bytes_copied
)
5153 struct sctp_sockaddr_entry
*addr
;
5154 union sctp_addr temp
;
5157 struct net
*net
= sock_net(sk
);
5160 list_for_each_entry_rcu(addr
, &net
->sctp
.local_addr_list
, list
) {
5164 if ((PF_INET
== sk
->sk_family
) &&
5165 (AF_INET6
== addr
->a
.sa
.sa_family
))
5167 if ((PF_INET6
== sk
->sk_family
) &&
5168 inet_v6_ipv6only(sk
) &&
5169 (AF_INET
== addr
->a
.sa
.sa_family
))
5171 memcpy(&temp
, &addr
->a
, sizeof(temp
));
5172 if (!temp
.v4
.sin_port
)
5173 temp
.v4
.sin_port
= htons(port
);
5175 addrlen
= sctp_get_pf_specific(sk
->sk_family
)
5176 ->addr_to_user(sctp_sk(sk
), &temp
);
5178 if (space_left
< addrlen
) {
5182 memcpy(to
, &temp
, addrlen
);
5186 space_left
-= addrlen
;
5187 *bytes_copied
+= addrlen
;
5195 static int sctp_getsockopt_local_addrs(struct sock
*sk
, int len
,
5196 char __user
*optval
, int __user
*optlen
)
5198 struct sctp_bind_addr
*bp
;
5199 struct sctp_association
*asoc
;
5201 struct sctp_getaddrs getaddrs
;
5202 struct sctp_sockaddr_entry
*addr
;
5204 union sctp_addr temp
;
5205 struct sctp_sock
*sp
= sctp_sk(sk
);
5209 int bytes_copied
= 0;
5213 if (len
< sizeof(struct sctp_getaddrs
))
5216 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
5220 * For UDP-style sockets, id specifies the association to query.
5221 * If the id field is set to the value '0' then the locally bound
5222 * addresses are returned without regard to any particular
5225 if (0 == getaddrs
.assoc_id
) {
5226 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
5228 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
5231 bp
= &asoc
->base
.bind_addr
;
5234 to
= optval
+ offsetof(struct sctp_getaddrs
, addrs
);
5235 space_left
= len
- offsetof(struct sctp_getaddrs
, addrs
);
5237 addrs
= kmalloc(space_left
, GFP_USER
| __GFP_NOWARN
);
5241 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5242 * addresses from the global local address list.
5244 if (sctp_list_single_entry(&bp
->address_list
)) {
5245 addr
= list_entry(bp
->address_list
.next
,
5246 struct sctp_sockaddr_entry
, list
);
5247 if (sctp_is_any(sk
, &addr
->a
)) {
5248 cnt
= sctp_copy_laddrs(sk
, bp
->port
, addrs
,
5249 space_left
, &bytes_copied
);
5259 /* Protection on the bound address list is not needed since
5260 * in the socket option context we hold a socket lock and
5261 * thus the bound address list can't change.
5263 list_for_each_entry(addr
, &bp
->address_list
, list
) {
5264 memcpy(&temp
, &addr
->a
, sizeof(temp
));
5265 addrlen
= sctp_get_pf_specific(sk
->sk_family
)
5266 ->addr_to_user(sp
, &temp
);
5267 if (space_left
< addrlen
) {
5268 err
= -ENOMEM
; /*fixme: right error?*/
5271 memcpy(buf
, &temp
, addrlen
);
5273 bytes_copied
+= addrlen
;
5275 space_left
-= addrlen
;
5279 if (copy_to_user(to
, addrs
, bytes_copied
)) {
5283 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
)) {
5287 if (put_user(bytes_copied
, optlen
))
5294 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
5296 * Requests that the local SCTP stack use the enclosed peer address as
5297 * the association primary. The enclosed address must be one of the
5298 * association peer's addresses.
5300 static int sctp_getsockopt_primary_addr(struct sock
*sk
, int len
,
5301 char __user
*optval
, int __user
*optlen
)
5303 struct sctp_prim prim
;
5304 struct sctp_association
*asoc
;
5305 struct sctp_sock
*sp
= sctp_sk(sk
);
5307 if (len
< sizeof(struct sctp_prim
))
5310 len
= sizeof(struct sctp_prim
);
5312 if (copy_from_user(&prim
, optval
, len
))
5315 asoc
= sctp_id2assoc(sk
, prim
.ssp_assoc_id
);
5319 if (!asoc
->peer
.primary_path
)
5322 memcpy(&prim
.ssp_addr
, &asoc
->peer
.primary_path
->ipaddr
,
5323 asoc
->peer
.primary_path
->af_specific
->sockaddr_len
);
5325 sctp_get_pf_specific(sk
->sk_family
)->addr_to_user(sp
,
5326 (union sctp_addr
*)&prim
.ssp_addr
);
5328 if (put_user(len
, optlen
))
5330 if (copy_to_user(optval
, &prim
, len
))
5337 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5339 * Requests that the local endpoint set the specified Adaptation Layer
5340 * Indication parameter for all future INIT and INIT-ACK exchanges.
5342 static int sctp_getsockopt_adaptation_layer(struct sock
*sk
, int len
,
5343 char __user
*optval
, int __user
*optlen
)
5345 struct sctp_setadaptation adaptation
;
5347 if (len
< sizeof(struct sctp_setadaptation
))
5350 len
= sizeof(struct sctp_setadaptation
);
5352 adaptation
.ssb_adaptation_ind
= sctp_sk(sk
)->adaptation_ind
;
5354 if (put_user(len
, optlen
))
5356 if (copy_to_user(optval
, &adaptation
, len
))
5364 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5366 * Applications that wish to use the sendto() system call may wish to
5367 * specify a default set of parameters that would normally be supplied
5368 * through the inclusion of ancillary data. This socket option allows
5369 * such an application to set the default sctp_sndrcvinfo structure.
5372 * The application that wishes to use this socket option simply passes
5373 * in to this call the sctp_sndrcvinfo structure defined in Section
5374 * 5.2.2) The input parameters accepted by this call include
5375 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5376 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5377 * to this call if the caller is using the UDP model.
5379 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5381 static int sctp_getsockopt_default_send_param(struct sock
*sk
,
5382 int len
, char __user
*optval
,
5385 struct sctp_sock
*sp
= sctp_sk(sk
);
5386 struct sctp_association
*asoc
;
5387 struct sctp_sndrcvinfo info
;
5389 if (len
< sizeof(info
))
5394 if (copy_from_user(&info
, optval
, len
))
5397 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
5398 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
5401 info
.sinfo_stream
= asoc
->default_stream
;
5402 info
.sinfo_flags
= asoc
->default_flags
;
5403 info
.sinfo_ppid
= asoc
->default_ppid
;
5404 info
.sinfo_context
= asoc
->default_context
;
5405 info
.sinfo_timetolive
= asoc
->default_timetolive
;
5407 info
.sinfo_stream
= sp
->default_stream
;
5408 info
.sinfo_flags
= sp
->default_flags
;
5409 info
.sinfo_ppid
= sp
->default_ppid
;
5410 info
.sinfo_context
= sp
->default_context
;
5411 info
.sinfo_timetolive
= sp
->default_timetolive
;
5414 if (put_user(len
, optlen
))
5416 if (copy_to_user(optval
, &info
, len
))
5422 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5423 * (SCTP_DEFAULT_SNDINFO)
5425 static int sctp_getsockopt_default_sndinfo(struct sock
*sk
, int len
,
5426 char __user
*optval
,
5429 struct sctp_sock
*sp
= sctp_sk(sk
);
5430 struct sctp_association
*asoc
;
5431 struct sctp_sndinfo info
;
5433 if (len
< sizeof(info
))
5438 if (copy_from_user(&info
, optval
, len
))
5441 asoc
= sctp_id2assoc(sk
, info
.snd_assoc_id
);
5442 if (!asoc
&& info
.snd_assoc_id
&& sctp_style(sk
, UDP
))
5445 info
.snd_sid
= asoc
->default_stream
;
5446 info
.snd_flags
= asoc
->default_flags
;
5447 info
.snd_ppid
= asoc
->default_ppid
;
5448 info
.snd_context
= asoc
->default_context
;
5450 info
.snd_sid
= sp
->default_stream
;
5451 info
.snd_flags
= sp
->default_flags
;
5452 info
.snd_ppid
= sp
->default_ppid
;
5453 info
.snd_context
= sp
->default_context
;
5456 if (put_user(len
, optlen
))
5458 if (copy_to_user(optval
, &info
, len
))
5466 * 7.1.5 SCTP_NODELAY
5468 * Turn on/off any Nagle-like algorithm. This means that packets are
5469 * generally sent as soon as possible and no unnecessary delays are
5470 * introduced, at the cost of more packets in the network. Expects an
5471 * integer boolean flag.
5474 static int sctp_getsockopt_nodelay(struct sock
*sk
, int len
,
5475 char __user
*optval
, int __user
*optlen
)
5479 if (len
< sizeof(int))
5483 val
= (sctp_sk(sk
)->nodelay
== 1);
5484 if (put_user(len
, optlen
))
5486 if (copy_to_user(optval
, &val
, len
))
5493 * 7.1.1 SCTP_RTOINFO
5495 * The protocol parameters used to initialize and bound retransmission
5496 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5497 * and modify these parameters.
5498 * All parameters are time values, in milliseconds. A value of 0, when
5499 * modifying the parameters, indicates that the current value should not
5503 static int sctp_getsockopt_rtoinfo(struct sock
*sk
, int len
,
5504 char __user
*optval
,
5505 int __user
*optlen
) {
5506 struct sctp_rtoinfo rtoinfo
;
5507 struct sctp_association
*asoc
;
5509 if (len
< sizeof (struct sctp_rtoinfo
))
5512 len
= sizeof(struct sctp_rtoinfo
);
5514 if (copy_from_user(&rtoinfo
, optval
, len
))
5517 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
5519 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
5522 /* Values corresponding to the specific association. */
5524 rtoinfo
.srto_initial
= jiffies_to_msecs(asoc
->rto_initial
);
5525 rtoinfo
.srto_max
= jiffies_to_msecs(asoc
->rto_max
);
5526 rtoinfo
.srto_min
= jiffies_to_msecs(asoc
->rto_min
);
5528 /* Values corresponding to the endpoint. */
5529 struct sctp_sock
*sp
= sctp_sk(sk
);
5531 rtoinfo
.srto_initial
= sp
->rtoinfo
.srto_initial
;
5532 rtoinfo
.srto_max
= sp
->rtoinfo
.srto_max
;
5533 rtoinfo
.srto_min
= sp
->rtoinfo
.srto_min
;
5536 if (put_user(len
, optlen
))
5539 if (copy_to_user(optval
, &rtoinfo
, len
))
5547 * 7.1.2 SCTP_ASSOCINFO
5549 * This option is used to tune the maximum retransmission attempts
5550 * of the association.
5551 * Returns an error if the new association retransmission value is
5552 * greater than the sum of the retransmission value of the peer.
5553 * See [SCTP] for more information.
5556 static int sctp_getsockopt_associnfo(struct sock
*sk
, int len
,
5557 char __user
*optval
,
5561 struct sctp_assocparams assocparams
;
5562 struct sctp_association
*asoc
;
5563 struct list_head
*pos
;
5566 if (len
< sizeof (struct sctp_assocparams
))
5569 len
= sizeof(struct sctp_assocparams
);
5571 if (copy_from_user(&assocparams
, optval
, len
))
5574 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
5576 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
5579 /* Values correspoinding to the specific association */
5581 assocparams
.sasoc_asocmaxrxt
= asoc
->max_retrans
;
5582 assocparams
.sasoc_peer_rwnd
= asoc
->peer
.rwnd
;
5583 assocparams
.sasoc_local_rwnd
= asoc
->a_rwnd
;
5584 assocparams
.sasoc_cookie_life
= ktime_to_ms(asoc
->cookie_life
);
5586 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
5590 assocparams
.sasoc_number_peer_destinations
= cnt
;
5592 /* Values corresponding to the endpoint */
5593 struct sctp_sock
*sp
= sctp_sk(sk
);
5595 assocparams
.sasoc_asocmaxrxt
= sp
->assocparams
.sasoc_asocmaxrxt
;
5596 assocparams
.sasoc_peer_rwnd
= sp
->assocparams
.sasoc_peer_rwnd
;
5597 assocparams
.sasoc_local_rwnd
= sp
->assocparams
.sasoc_local_rwnd
;
5598 assocparams
.sasoc_cookie_life
=
5599 sp
->assocparams
.sasoc_cookie_life
;
5600 assocparams
.sasoc_number_peer_destinations
=
5602 sasoc_number_peer_destinations
;
5605 if (put_user(len
, optlen
))
5608 if (copy_to_user(optval
, &assocparams
, len
))
5615 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5617 * This socket option is a boolean flag which turns on or off mapped V4
5618 * addresses. If this option is turned on and the socket is type
5619 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5620 * If this option is turned off, then no mapping will be done of V4
5621 * addresses and a user will receive both PF_INET6 and PF_INET type
5622 * addresses on the socket.
5624 static int sctp_getsockopt_mappedv4(struct sock
*sk
, int len
,
5625 char __user
*optval
, int __user
*optlen
)
5628 struct sctp_sock
*sp
= sctp_sk(sk
);
5630 if (len
< sizeof(int))
5635 if (put_user(len
, optlen
))
5637 if (copy_to_user(optval
, &val
, len
))
5644 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5645 * (chapter and verse is quoted at sctp_setsockopt_context())
5647 static int sctp_getsockopt_context(struct sock
*sk
, int len
,
5648 char __user
*optval
, int __user
*optlen
)
5650 struct sctp_assoc_value params
;
5651 struct sctp_sock
*sp
;
5652 struct sctp_association
*asoc
;
5654 if (len
< sizeof(struct sctp_assoc_value
))
5657 len
= sizeof(struct sctp_assoc_value
);
5659 if (copy_from_user(¶ms
, optval
, len
))
5664 if (params
.assoc_id
!= 0) {
5665 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5668 params
.assoc_value
= asoc
->default_rcv_context
;
5670 params
.assoc_value
= sp
->default_rcv_context
;
5673 if (put_user(len
, optlen
))
5675 if (copy_to_user(optval
, ¶ms
, len
))
5682 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5683 * This option will get or set the maximum size to put in any outgoing
5684 * SCTP DATA chunk. If a message is larger than this size it will be
5685 * fragmented by SCTP into the specified size. Note that the underlying
5686 * SCTP implementation may fragment into smaller sized chunks when the
5687 * PMTU of the underlying association is smaller than the value set by
5688 * the user. The default value for this option is '0' which indicates
5689 * the user is NOT limiting fragmentation and only the PMTU will effect
5690 * SCTP's choice of DATA chunk size. Note also that values set larger
5691 * than the maximum size of an IP datagram will effectively let SCTP
5692 * control fragmentation (i.e. the same as setting this option to 0).
5694 * The following structure is used to access and modify this parameter:
5696 * struct sctp_assoc_value {
5697 * sctp_assoc_t assoc_id;
5698 * uint32_t assoc_value;
5701 * assoc_id: This parameter is ignored for one-to-one style sockets.
5702 * For one-to-many style sockets this parameter indicates which
5703 * association the user is performing an action upon. Note that if
5704 * this field's value is zero then the endpoints default value is
5705 * changed (effecting future associations only).
5706 * assoc_value: This parameter specifies the maximum size in bytes.
5708 static int sctp_getsockopt_maxseg(struct sock
*sk
, int len
,
5709 char __user
*optval
, int __user
*optlen
)
5711 struct sctp_assoc_value params
;
5712 struct sctp_association
*asoc
;
5714 if (len
== sizeof(int)) {
5715 pr_warn_ratelimited(DEPRECATED
5717 "Use of int in maxseg socket option.\n"
5718 "Use struct sctp_assoc_value instead\n",
5719 current
->comm
, task_pid_nr(current
));
5720 params
.assoc_id
= 0;
5721 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5722 len
= sizeof(struct sctp_assoc_value
);
5723 if (copy_from_user(¶ms
, optval
, sizeof(params
)))
5728 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5729 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
5733 params
.assoc_value
= asoc
->frag_point
;
5735 params
.assoc_value
= sctp_sk(sk
)->user_frag
;
5737 if (put_user(len
, optlen
))
5739 if (len
== sizeof(int)) {
5740 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5743 if (copy_to_user(optval
, ¶ms
, len
))
5751 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5752 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5754 static int sctp_getsockopt_fragment_interleave(struct sock
*sk
, int len
,
5755 char __user
*optval
, int __user
*optlen
)
5759 if (len
< sizeof(int))
5764 val
= sctp_sk(sk
)->frag_interleave
;
5765 if (put_user(len
, optlen
))
5767 if (copy_to_user(optval
, &val
, len
))
5774 * 7.1.25. Set or Get the sctp partial delivery point
5775 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5777 static int sctp_getsockopt_partial_delivery_point(struct sock
*sk
, int len
,
5778 char __user
*optval
,
5783 if (len
< sizeof(u32
))
5788 val
= sctp_sk(sk
)->pd_point
;
5789 if (put_user(len
, optlen
))
5791 if (copy_to_user(optval
, &val
, len
))
5798 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5799 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5801 static int sctp_getsockopt_maxburst(struct sock
*sk
, int len
,
5802 char __user
*optval
,
5805 struct sctp_assoc_value params
;
5806 struct sctp_sock
*sp
;
5807 struct sctp_association
*asoc
;
5809 if (len
== sizeof(int)) {
5810 pr_warn_ratelimited(DEPRECATED
5812 "Use of int in max_burst socket option.\n"
5813 "Use struct sctp_assoc_value instead\n",
5814 current
->comm
, task_pid_nr(current
));
5815 params
.assoc_id
= 0;
5816 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5817 len
= sizeof(struct sctp_assoc_value
);
5818 if (copy_from_user(¶ms
, optval
, len
))
5825 if (params
.assoc_id
!= 0) {
5826 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5829 params
.assoc_value
= asoc
->max_burst
;
5831 params
.assoc_value
= sp
->max_burst
;
5833 if (len
== sizeof(int)) {
5834 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5837 if (copy_to_user(optval
, ¶ms
, len
))
5845 static int sctp_getsockopt_hmac_ident(struct sock
*sk
, int len
,
5846 char __user
*optval
, int __user
*optlen
)
5848 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5849 struct sctp_hmacalgo __user
*p
= (void __user
*)optval
;
5850 struct sctp_hmac_algo_param
*hmacs
;
5855 if (!ep
->auth_enable
)
5858 hmacs
= ep
->auth_hmacs_list
;
5859 data_len
= ntohs(hmacs
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5861 if (len
< sizeof(struct sctp_hmacalgo
) + data_len
)
5864 len
= sizeof(struct sctp_hmacalgo
) + data_len
;
5865 num_idents
= data_len
/ sizeof(u16
);
5867 if (put_user(len
, optlen
))
5869 if (put_user(num_idents
, &p
->shmac_num_idents
))
5871 for (i
= 0; i
< num_idents
; i
++) {
5872 __u16 hmacid
= ntohs(hmacs
->hmac_ids
[i
]);
5874 if (copy_to_user(&p
->shmac_idents
[i
], &hmacid
, sizeof(__u16
)))
5880 static int sctp_getsockopt_active_key(struct sock
*sk
, int len
,
5881 char __user
*optval
, int __user
*optlen
)
5883 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5884 struct sctp_authkeyid val
;
5885 struct sctp_association
*asoc
;
5887 if (!ep
->auth_enable
)
5890 if (len
< sizeof(struct sctp_authkeyid
))
5892 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authkeyid
)))
5895 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
5896 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
5900 val
.scact_keynumber
= asoc
->active_key_id
;
5902 val
.scact_keynumber
= ep
->active_key_id
;
5904 len
= sizeof(struct sctp_authkeyid
);
5905 if (put_user(len
, optlen
))
5907 if (copy_to_user(optval
, &val
, len
))
5913 static int sctp_getsockopt_peer_auth_chunks(struct sock
*sk
, int len
,
5914 char __user
*optval
, int __user
*optlen
)
5916 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5917 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5918 struct sctp_authchunks val
;
5919 struct sctp_association
*asoc
;
5920 struct sctp_chunks_param
*ch
;
5924 if (!ep
->auth_enable
)
5927 if (len
< sizeof(struct sctp_authchunks
))
5930 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5933 to
= p
->gauth_chunks
;
5934 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5938 ch
= asoc
->peer
.peer_chunks
;
5942 /* See if the user provided enough room for all the data */
5943 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5944 if (len
< num_chunks
)
5947 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
5950 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
5951 if (put_user(len
, optlen
))
5953 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
5958 static int sctp_getsockopt_local_auth_chunks(struct sock
*sk
, int len
,
5959 char __user
*optval
, int __user
*optlen
)
5961 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5962 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5963 struct sctp_authchunks val
;
5964 struct sctp_association
*asoc
;
5965 struct sctp_chunks_param
*ch
;
5969 if (!ep
->auth_enable
)
5972 if (len
< sizeof(struct sctp_authchunks
))
5975 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5978 to
= p
->gauth_chunks
;
5979 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5980 if (!asoc
&& val
.gauth_assoc_id
&& sctp_style(sk
, UDP
))
5984 ch
= (struct sctp_chunks_param
*)asoc
->c
.auth_chunks
;
5986 ch
= ep
->auth_chunk_list
;
5991 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5992 if (len
< sizeof(struct sctp_authchunks
) + num_chunks
)
5995 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
5998 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
5999 if (put_user(len
, optlen
))
6001 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
6008 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
6009 * This option gets the current number of associations that are attached
6010 * to a one-to-many style socket. The option value is an uint32_t.
6012 static int sctp_getsockopt_assoc_number(struct sock
*sk
, int len
,
6013 char __user
*optval
, int __user
*optlen
)
6015 struct sctp_sock
*sp
= sctp_sk(sk
);
6016 struct sctp_association
*asoc
;
6019 if (sctp_style(sk
, TCP
))
6022 if (len
< sizeof(u32
))
6027 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
6031 if (put_user(len
, optlen
))
6033 if (copy_to_user(optval
, &val
, len
))
6040 * 8.1.23 SCTP_AUTO_ASCONF
6041 * See the corresponding setsockopt entry as description
6043 static int sctp_getsockopt_auto_asconf(struct sock
*sk
, int len
,
6044 char __user
*optval
, int __user
*optlen
)
6048 if (len
< sizeof(int))
6052 if (sctp_sk(sk
)->do_auto_asconf
&& sctp_is_ep_boundall(sk
))
6054 if (put_user(len
, optlen
))
6056 if (copy_to_user(optval
, &val
, len
))
6062 * 8.2.6. Get the Current Identifiers of Associations
6063 * (SCTP_GET_ASSOC_ID_LIST)
6065 * This option gets the current list of SCTP association identifiers of
6066 * the SCTP associations handled by a one-to-many style socket.
6068 static int sctp_getsockopt_assoc_ids(struct sock
*sk
, int len
,
6069 char __user
*optval
, int __user
*optlen
)
6071 struct sctp_sock
*sp
= sctp_sk(sk
);
6072 struct sctp_association
*asoc
;
6073 struct sctp_assoc_ids
*ids
;
6076 if (sctp_style(sk
, TCP
))
6079 if (len
< sizeof(struct sctp_assoc_ids
))
6082 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
6086 if (len
< sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
)
6089 len
= sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
;
6091 ids
= kmalloc(len
, GFP_USER
| __GFP_NOWARN
);
6095 ids
->gaids_number_of_ids
= num
;
6097 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
6098 ids
->gaids_assoc_id
[num
++] = asoc
->assoc_id
;
6101 if (put_user(len
, optlen
) || copy_to_user(optval
, ids
, len
)) {
6111 * SCTP_PEER_ADDR_THLDS
6113 * This option allows us to fetch the partially failed threshold for one or all
6114 * transports in an association. See Section 6.1 of:
6115 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6117 static int sctp_getsockopt_paddr_thresholds(struct sock
*sk
,
6118 char __user
*optval
,
6122 struct sctp_paddrthlds val
;
6123 struct sctp_transport
*trans
;
6124 struct sctp_association
*asoc
;
6126 if (len
< sizeof(struct sctp_paddrthlds
))
6128 len
= sizeof(struct sctp_paddrthlds
);
6129 if (copy_from_user(&val
, (struct sctp_paddrthlds __user
*)optval
, len
))
6132 if (sctp_is_any(sk
, (const union sctp_addr
*)&val
.spt_address
)) {
6133 asoc
= sctp_id2assoc(sk
, val
.spt_assoc_id
);
6137 val
.spt_pathpfthld
= asoc
->pf_retrans
;
6138 val
.spt_pathmaxrxt
= asoc
->pathmaxrxt
;
6140 trans
= sctp_addr_id2transport(sk
, &val
.spt_address
,
6145 val
.spt_pathmaxrxt
= trans
->pathmaxrxt
;
6146 val
.spt_pathpfthld
= trans
->pf_retrans
;
6149 if (put_user(len
, optlen
) || copy_to_user(optval
, &val
, len
))
6156 * SCTP_GET_ASSOC_STATS
6158 * This option retrieves local per endpoint statistics. It is modeled
6159 * after OpenSolaris' implementation
6161 static int sctp_getsockopt_assoc_stats(struct sock
*sk
, int len
,
6162 char __user
*optval
,
6165 struct sctp_assoc_stats sas
;
6166 struct sctp_association
*asoc
= NULL
;
6168 /* User must provide at least the assoc id */
6169 if (len
< sizeof(sctp_assoc_t
))
6172 /* Allow the struct to grow and fill in as much as possible */
6173 len
= min_t(size_t, len
, sizeof(sas
));
6175 if (copy_from_user(&sas
, optval
, len
))
6178 asoc
= sctp_id2assoc(sk
, sas
.sas_assoc_id
);
6182 sas
.sas_rtxchunks
= asoc
->stats
.rtxchunks
;
6183 sas
.sas_gapcnt
= asoc
->stats
.gapcnt
;
6184 sas
.sas_outofseqtsns
= asoc
->stats
.outofseqtsns
;
6185 sas
.sas_osacks
= asoc
->stats
.osacks
;
6186 sas
.sas_isacks
= asoc
->stats
.isacks
;
6187 sas
.sas_octrlchunks
= asoc
->stats
.octrlchunks
;
6188 sas
.sas_ictrlchunks
= asoc
->stats
.ictrlchunks
;
6189 sas
.sas_oodchunks
= asoc
->stats
.oodchunks
;
6190 sas
.sas_iodchunks
= asoc
->stats
.iodchunks
;
6191 sas
.sas_ouodchunks
= asoc
->stats
.ouodchunks
;
6192 sas
.sas_iuodchunks
= asoc
->stats
.iuodchunks
;
6193 sas
.sas_idupchunks
= asoc
->stats
.idupchunks
;
6194 sas
.sas_opackets
= asoc
->stats
.opackets
;
6195 sas
.sas_ipackets
= asoc
->stats
.ipackets
;
6197 /* New high max rto observed, will return 0 if not a single
6198 * RTO update took place. obs_rto_ipaddr will be bogus
6201 sas
.sas_maxrto
= asoc
->stats
.max_obs_rto
;
6202 memcpy(&sas
.sas_obs_rto_ipaddr
, &asoc
->stats
.obs_rto_ipaddr
,
6203 sizeof(struct sockaddr_storage
));
6205 /* Mark beginning of a new observation period */
6206 asoc
->stats
.max_obs_rto
= asoc
->rto_min
;
6208 if (put_user(len
, optlen
))
6211 pr_debug("%s: len:%d, assoc_id:%d\n", __func__
, len
, sas
.sas_assoc_id
);
6213 if (copy_to_user(optval
, &sas
, len
))
6219 static int sctp_getsockopt_recvrcvinfo(struct sock
*sk
, int len
,
6220 char __user
*optval
,
6225 if (len
< sizeof(int))
6229 if (sctp_sk(sk
)->recvrcvinfo
)
6231 if (put_user(len
, optlen
))
6233 if (copy_to_user(optval
, &val
, len
))
6239 static int sctp_getsockopt_recvnxtinfo(struct sock
*sk
, int len
,
6240 char __user
*optval
,
6245 if (len
< sizeof(int))
6249 if (sctp_sk(sk
)->recvnxtinfo
)
6251 if (put_user(len
, optlen
))
6253 if (copy_to_user(optval
, &val
, len
))
6259 static int sctp_getsockopt_pr_supported(struct sock
*sk
, int len
,
6260 char __user
*optval
,
6263 struct sctp_assoc_value params
;
6264 struct sctp_association
*asoc
;
6265 int retval
= -EFAULT
;
6267 if (len
< sizeof(params
)) {
6272 len
= sizeof(params
);
6273 if (copy_from_user(¶ms
, optval
, len
))
6276 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
6278 params
.assoc_value
= asoc
->prsctp_enable
;
6279 } else if (!params
.assoc_id
) {
6280 struct sctp_sock
*sp
= sctp_sk(sk
);
6282 params
.assoc_value
= sp
->ep
->prsctp_enable
;
6288 if (put_user(len
, optlen
))
6291 if (copy_to_user(optval
, ¶ms
, len
))
6300 static int sctp_getsockopt_default_prinfo(struct sock
*sk
, int len
,
6301 char __user
*optval
,
6304 struct sctp_default_prinfo info
;
6305 struct sctp_association
*asoc
;
6306 int retval
= -EFAULT
;
6308 if (len
< sizeof(info
)) {
6314 if (copy_from_user(&info
, optval
, len
))
6317 asoc
= sctp_id2assoc(sk
, info
.pr_assoc_id
);
6319 info
.pr_policy
= SCTP_PR_POLICY(asoc
->default_flags
);
6320 info
.pr_value
= asoc
->default_timetolive
;
6321 } else if (!info
.pr_assoc_id
) {
6322 struct sctp_sock
*sp
= sctp_sk(sk
);
6324 info
.pr_policy
= SCTP_PR_POLICY(sp
->default_flags
);
6325 info
.pr_value
= sp
->default_timetolive
;
6331 if (put_user(len
, optlen
))
6334 if (copy_to_user(optval
, &info
, len
))
6343 static int sctp_getsockopt_pr_assocstatus(struct sock
*sk
, int len
,
6344 char __user
*optval
,
6347 struct sctp_prstatus params
;
6348 struct sctp_association
*asoc
;
6350 int retval
= -EINVAL
;
6352 if (len
< sizeof(params
))
6355 len
= sizeof(params
);
6356 if (copy_from_user(¶ms
, optval
, len
)) {
6361 policy
= params
.sprstat_policy
;
6362 if (policy
& ~SCTP_PR_SCTP_MASK
)
6365 asoc
= sctp_id2assoc(sk
, params
.sprstat_assoc_id
);
6369 if (policy
== SCTP_PR_SCTP_NONE
) {
6370 params
.sprstat_abandoned_unsent
= 0;
6371 params
.sprstat_abandoned_sent
= 0;
6372 for (policy
= 0; policy
<= SCTP_PR_INDEX(MAX
); policy
++) {
6373 params
.sprstat_abandoned_unsent
+=
6374 asoc
->abandoned_unsent
[policy
];
6375 params
.sprstat_abandoned_sent
+=
6376 asoc
->abandoned_sent
[policy
];
6379 params
.sprstat_abandoned_unsent
=
6380 asoc
->abandoned_unsent
[__SCTP_PR_INDEX(policy
)];
6381 params
.sprstat_abandoned_sent
=
6382 asoc
->abandoned_sent
[__SCTP_PR_INDEX(policy
)];
6385 if (put_user(len
, optlen
)) {
6390 if (copy_to_user(optval
, ¶ms
, len
)) {
6401 static int sctp_getsockopt(struct sock
*sk
, int level
, int optname
,
6402 char __user
*optval
, int __user
*optlen
)
6407 pr_debug("%s: sk:%p, optname:%d\n", __func__
, sk
, optname
);
6409 /* I can hardly begin to describe how wrong this is. This is
6410 * so broken as to be worse than useless. The API draft
6411 * REALLY is NOT helpful here... I am not convinced that the
6412 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
6413 * are at all well-founded.
6415 if (level
!= SOL_SCTP
) {
6416 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
6418 retval
= af
->getsockopt(sk
, level
, optname
, optval
, optlen
);
6422 if (get_user(len
, optlen
))
6429 retval
= sctp_getsockopt_sctp_status(sk
, len
, optval
, optlen
);
6431 case SCTP_DISABLE_FRAGMENTS
:
6432 retval
= sctp_getsockopt_disable_fragments(sk
, len
, optval
,
6436 retval
= sctp_getsockopt_events(sk
, len
, optval
, optlen
);
6438 case SCTP_AUTOCLOSE
:
6439 retval
= sctp_getsockopt_autoclose(sk
, len
, optval
, optlen
);
6441 case SCTP_SOCKOPT_PEELOFF
:
6442 retval
= sctp_getsockopt_peeloff(sk
, len
, optval
, optlen
);
6444 case SCTP_PEER_ADDR_PARAMS
:
6445 retval
= sctp_getsockopt_peer_addr_params(sk
, len
, optval
,
6448 case SCTP_DELAYED_SACK
:
6449 retval
= sctp_getsockopt_delayed_ack(sk
, len
, optval
,
6453 retval
= sctp_getsockopt_initmsg(sk
, len
, optval
, optlen
);
6455 case SCTP_GET_PEER_ADDRS
:
6456 retval
= sctp_getsockopt_peer_addrs(sk
, len
, optval
,
6459 case SCTP_GET_LOCAL_ADDRS
:
6460 retval
= sctp_getsockopt_local_addrs(sk
, len
, optval
,
6463 case SCTP_SOCKOPT_CONNECTX3
:
6464 retval
= sctp_getsockopt_connectx3(sk
, len
, optval
, optlen
);
6466 case SCTP_DEFAULT_SEND_PARAM
:
6467 retval
= sctp_getsockopt_default_send_param(sk
, len
,
6470 case SCTP_DEFAULT_SNDINFO
:
6471 retval
= sctp_getsockopt_default_sndinfo(sk
, len
,
6474 case SCTP_PRIMARY_ADDR
:
6475 retval
= sctp_getsockopt_primary_addr(sk
, len
, optval
, optlen
);
6478 retval
= sctp_getsockopt_nodelay(sk
, len
, optval
, optlen
);
6481 retval
= sctp_getsockopt_rtoinfo(sk
, len
, optval
, optlen
);
6483 case SCTP_ASSOCINFO
:
6484 retval
= sctp_getsockopt_associnfo(sk
, len
, optval
, optlen
);
6486 case SCTP_I_WANT_MAPPED_V4_ADDR
:
6487 retval
= sctp_getsockopt_mappedv4(sk
, len
, optval
, optlen
);
6490 retval
= sctp_getsockopt_maxseg(sk
, len
, optval
, optlen
);
6492 case SCTP_GET_PEER_ADDR_INFO
:
6493 retval
= sctp_getsockopt_peer_addr_info(sk
, len
, optval
,
6496 case SCTP_ADAPTATION_LAYER
:
6497 retval
= sctp_getsockopt_adaptation_layer(sk
, len
, optval
,
6501 retval
= sctp_getsockopt_context(sk
, len
, optval
, optlen
);
6503 case SCTP_FRAGMENT_INTERLEAVE
:
6504 retval
= sctp_getsockopt_fragment_interleave(sk
, len
, optval
,
6507 case SCTP_PARTIAL_DELIVERY_POINT
:
6508 retval
= sctp_getsockopt_partial_delivery_point(sk
, len
, optval
,
6511 case SCTP_MAX_BURST
:
6512 retval
= sctp_getsockopt_maxburst(sk
, len
, optval
, optlen
);
6515 case SCTP_AUTH_CHUNK
:
6516 case SCTP_AUTH_DELETE_KEY
:
6517 retval
= -EOPNOTSUPP
;
6519 case SCTP_HMAC_IDENT
:
6520 retval
= sctp_getsockopt_hmac_ident(sk
, len
, optval
, optlen
);
6522 case SCTP_AUTH_ACTIVE_KEY
:
6523 retval
= sctp_getsockopt_active_key(sk
, len
, optval
, optlen
);
6525 case SCTP_PEER_AUTH_CHUNKS
:
6526 retval
= sctp_getsockopt_peer_auth_chunks(sk
, len
, optval
,
6529 case SCTP_LOCAL_AUTH_CHUNKS
:
6530 retval
= sctp_getsockopt_local_auth_chunks(sk
, len
, optval
,
6533 case SCTP_GET_ASSOC_NUMBER
:
6534 retval
= sctp_getsockopt_assoc_number(sk
, len
, optval
, optlen
);
6536 case SCTP_GET_ASSOC_ID_LIST
:
6537 retval
= sctp_getsockopt_assoc_ids(sk
, len
, optval
, optlen
);
6539 case SCTP_AUTO_ASCONF
:
6540 retval
= sctp_getsockopt_auto_asconf(sk
, len
, optval
, optlen
);
6542 case SCTP_PEER_ADDR_THLDS
:
6543 retval
= sctp_getsockopt_paddr_thresholds(sk
, optval
, len
, optlen
);
6545 case SCTP_GET_ASSOC_STATS
:
6546 retval
= sctp_getsockopt_assoc_stats(sk
, len
, optval
, optlen
);
6548 case SCTP_RECVRCVINFO
:
6549 retval
= sctp_getsockopt_recvrcvinfo(sk
, len
, optval
, optlen
);
6551 case SCTP_RECVNXTINFO
:
6552 retval
= sctp_getsockopt_recvnxtinfo(sk
, len
, optval
, optlen
);
6554 case SCTP_PR_SUPPORTED
:
6555 retval
= sctp_getsockopt_pr_supported(sk
, len
, optval
, optlen
);
6557 case SCTP_DEFAULT_PRINFO
:
6558 retval
= sctp_getsockopt_default_prinfo(sk
, len
, optval
,
6561 case SCTP_PR_ASSOC_STATUS
:
6562 retval
= sctp_getsockopt_pr_assocstatus(sk
, len
, optval
,
6566 retval
= -ENOPROTOOPT
;
6574 static int sctp_hash(struct sock
*sk
)
6580 static void sctp_unhash(struct sock
*sk
)
6585 /* Check if port is acceptable. Possibly find first available port.
6587 * The port hash table (contained in the 'global' SCTP protocol storage
6588 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
6589 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
6590 * list (the list number is the port number hashed out, so as you
6591 * would expect from a hash function, all the ports in a given list have
6592 * such a number that hashes out to the same list number; you were
6593 * expecting that, right?); so each list has a set of ports, with a
6594 * link to the socket (struct sock) that uses it, the port number and
6595 * a fastreuse flag (FIXME: NPI ipg).
6597 static struct sctp_bind_bucket
*sctp_bucket_create(
6598 struct sctp_bind_hashbucket
*head
, struct net
*, unsigned short snum
);
6600 static long sctp_get_port_local(struct sock
*sk
, union sctp_addr
*addr
)
6602 struct sctp_bind_hashbucket
*head
; /* hash list */
6603 struct sctp_bind_bucket
*pp
;
6604 unsigned short snum
;
6607 snum
= ntohs(addr
->v4
.sin_port
);
6609 pr_debug("%s: begins, snum:%d\n", __func__
, snum
);
6614 /* Search for an available port. */
6615 int low
, high
, remaining
, index
;
6617 struct net
*net
= sock_net(sk
);
6619 inet_get_local_port_range(net
, &low
, &high
);
6620 remaining
= (high
- low
) + 1;
6621 rover
= prandom_u32() % remaining
+ low
;
6625 if ((rover
< low
) || (rover
> high
))
6627 if (inet_is_local_reserved_port(net
, rover
))
6629 index
= sctp_phashfn(sock_net(sk
), rover
);
6630 head
= &sctp_port_hashtable
[index
];
6631 spin_lock(&head
->lock
);
6632 sctp_for_each_hentry(pp
, &head
->chain
)
6633 if ((pp
->port
== rover
) &&
6634 net_eq(sock_net(sk
), pp
->net
))
6638 spin_unlock(&head
->lock
);
6639 } while (--remaining
> 0);
6641 /* Exhausted local port range during search? */
6646 /* OK, here is the one we will use. HEAD (the port
6647 * hash table list entry) is non-NULL and we hold it's
6652 /* We are given an specific port number; we verify
6653 * that it is not being used. If it is used, we will
6654 * exahust the search in the hash list corresponding
6655 * to the port number (snum) - we detect that with the
6656 * port iterator, pp being NULL.
6658 head
= &sctp_port_hashtable
[sctp_phashfn(sock_net(sk
), snum
)];
6659 spin_lock(&head
->lock
);
6660 sctp_for_each_hentry(pp
, &head
->chain
) {
6661 if ((pp
->port
== snum
) && net_eq(pp
->net
, sock_net(sk
)))
6668 if (!hlist_empty(&pp
->owner
)) {
6669 /* We had a port hash table hit - there is an
6670 * available port (pp != NULL) and it is being
6671 * used by other socket (pp->owner not empty); that other
6672 * socket is going to be sk2.
6674 int reuse
= sk
->sk_reuse
;
6677 pr_debug("%s: found a possible match\n", __func__
);
6679 if (pp
->fastreuse
&& sk
->sk_reuse
&&
6680 sk
->sk_state
!= SCTP_SS_LISTENING
)
6683 /* Run through the list of sockets bound to the port
6684 * (pp->port) [via the pointers bind_next and
6685 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6686 * we get the endpoint they describe and run through
6687 * the endpoint's list of IP (v4 or v6) addresses,
6688 * comparing each of the addresses with the address of
6689 * the socket sk. If we find a match, then that means
6690 * that this port/socket (sk) combination are already
6693 sk_for_each_bound(sk2
, &pp
->owner
) {
6694 struct sctp_endpoint
*ep2
;
6695 ep2
= sctp_sk(sk2
)->ep
;
6698 (reuse
&& sk2
->sk_reuse
&&
6699 sk2
->sk_state
!= SCTP_SS_LISTENING
))
6702 if (sctp_bind_addr_conflict(&ep2
->base
.bind_addr
, addr
,
6703 sctp_sk(sk2
), sctp_sk(sk
))) {
6709 pr_debug("%s: found a match\n", __func__
);
6712 /* If there was a hash table miss, create a new port. */
6714 if (!pp
&& !(pp
= sctp_bucket_create(head
, sock_net(sk
), snum
)))
6717 /* In either case (hit or miss), make sure fastreuse is 1 only
6718 * if sk->sk_reuse is too (that is, if the caller requested
6719 * SO_REUSEADDR on this socket -sk-).
6721 if (hlist_empty(&pp
->owner
)) {
6722 if (sk
->sk_reuse
&& sk
->sk_state
!= SCTP_SS_LISTENING
)
6726 } else if (pp
->fastreuse
&&
6727 (!sk
->sk_reuse
|| sk
->sk_state
== SCTP_SS_LISTENING
))
6730 /* We are set, so fill up all the data in the hash table
6731 * entry, tie the socket list information with the rest of the
6732 * sockets FIXME: Blurry, NPI (ipg).
6735 if (!sctp_sk(sk
)->bind_hash
) {
6736 inet_sk(sk
)->inet_num
= snum
;
6737 sk_add_bind_node(sk
, &pp
->owner
);
6738 sctp_sk(sk
)->bind_hash
= pp
;
6743 spin_unlock(&head
->lock
);
6750 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6751 * port is requested.
6753 static int sctp_get_port(struct sock
*sk
, unsigned short snum
)
6755 union sctp_addr addr
;
6756 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
6758 /* Set up a dummy address struct from the sk. */
6759 af
->from_sk(&addr
, sk
);
6760 addr
.v4
.sin_port
= htons(snum
);
6762 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6763 return !!sctp_get_port_local(sk
, &addr
);
6767 * Move a socket to LISTENING state.
6769 static int sctp_listen_start(struct sock
*sk
, int backlog
)
6771 struct sctp_sock
*sp
= sctp_sk(sk
);
6772 struct sctp_endpoint
*ep
= sp
->ep
;
6773 struct crypto_shash
*tfm
= NULL
;
6776 /* Allocate HMAC for generating cookie. */
6777 if (!sp
->hmac
&& sp
->sctp_hmac_alg
) {
6778 sprintf(alg
, "hmac(%s)", sp
->sctp_hmac_alg
);
6779 tfm
= crypto_alloc_shash(alg
, 0, 0);
6781 net_info_ratelimited("failed to load transform for %s: %ld\n",
6782 sp
->sctp_hmac_alg
, PTR_ERR(tfm
));
6785 sctp_sk(sk
)->hmac
= tfm
;
6789 * If a bind() or sctp_bindx() is not called prior to a listen()
6790 * call that allows new associations to be accepted, the system
6791 * picks an ephemeral port and will choose an address set equivalent
6792 * to binding with a wildcard address.
6794 * This is not currently spelled out in the SCTP sockets
6795 * extensions draft, but follows the practice as seen in TCP
6799 sk
->sk_state
= SCTP_SS_LISTENING
;
6800 if (!ep
->base
.bind_addr
.port
) {
6801 if (sctp_autobind(sk
))
6804 if (sctp_get_port(sk
, inet_sk(sk
)->inet_num
)) {
6805 sk
->sk_state
= SCTP_SS_CLOSED
;
6810 sk
->sk_max_ack_backlog
= backlog
;
6811 sctp_hash_endpoint(ep
);
6816 * 4.1.3 / 5.1.3 listen()
6818 * By default, new associations are not accepted for UDP style sockets.
6819 * An application uses listen() to mark a socket as being able to
6820 * accept new associations.
6822 * On TCP style sockets, applications use listen() to ready the SCTP
6823 * endpoint for accepting inbound associations.
6825 * On both types of endpoints a backlog of '0' disables listening.
6827 * Move a socket to LISTENING state.
6829 int sctp_inet_listen(struct socket
*sock
, int backlog
)
6831 struct sock
*sk
= sock
->sk
;
6832 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
6835 if (unlikely(backlog
< 0))
6840 /* Peeled-off sockets are not allowed to listen(). */
6841 if (sctp_style(sk
, UDP_HIGH_BANDWIDTH
))
6844 if (sock
->state
!= SS_UNCONNECTED
)
6847 /* If backlog is zero, disable listening. */
6849 if (sctp_sstate(sk
, CLOSED
))
6853 sctp_unhash_endpoint(ep
);
6854 sk
->sk_state
= SCTP_SS_CLOSED
;
6856 sctp_sk(sk
)->bind_hash
->fastreuse
= 1;
6860 /* If we are already listening, just update the backlog */
6861 if (sctp_sstate(sk
, LISTENING
))
6862 sk
->sk_max_ack_backlog
= backlog
;
6864 err
= sctp_listen_start(sk
, backlog
);
6876 * This function is done by modeling the current datagram_poll() and the
6877 * tcp_poll(). Note that, based on these implementations, we don't
6878 * lock the socket in this function, even though it seems that,
6879 * ideally, locking or some other mechanisms can be used to ensure
6880 * the integrity of the counters (sndbuf and wmem_alloc) used
6881 * in this place. We assume that we don't need locks either until proven
6884 * Another thing to note is that we include the Async I/O support
6885 * here, again, by modeling the current TCP/UDP code. We don't have
6886 * a good way to test with it yet.
6888 unsigned int sctp_poll(struct file
*file
, struct socket
*sock
, poll_table
*wait
)
6890 struct sock
*sk
= sock
->sk
;
6891 struct sctp_sock
*sp
= sctp_sk(sk
);
6894 poll_wait(file
, sk_sleep(sk
), wait
);
6896 sock_rps_record_flow(sk
);
6898 /* A TCP-style listening socket becomes readable when the accept queue
6901 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
6902 return (!list_empty(&sp
->ep
->asocs
)) ?
6903 (POLLIN
| POLLRDNORM
) : 0;
6907 /* Is there any exceptional events? */
6908 if (sk
->sk_err
|| !skb_queue_empty(&sk
->sk_error_queue
))
6910 (sock_flag(sk
, SOCK_SELECT_ERR_QUEUE
) ? POLLPRI
: 0);
6911 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6912 mask
|= POLLRDHUP
| POLLIN
| POLLRDNORM
;
6913 if (sk
->sk_shutdown
== SHUTDOWN_MASK
)
6916 /* Is it readable? Reconsider this code with TCP-style support. */
6917 if (!skb_queue_empty(&sk
->sk_receive_queue
))
6918 mask
|= POLLIN
| POLLRDNORM
;
6920 /* The association is either gone or not ready. */
6921 if (!sctp_style(sk
, UDP
) && sctp_sstate(sk
, CLOSED
))
6924 /* Is it writable? */
6925 if (sctp_writeable(sk
)) {
6926 mask
|= POLLOUT
| POLLWRNORM
;
6928 sk_set_bit(SOCKWQ_ASYNC_NOSPACE
, sk
);
6930 * Since the socket is not locked, the buffer
6931 * might be made available after the writeable check and
6932 * before the bit is set. This could cause a lost I/O
6933 * signal. tcp_poll() has a race breaker for this race
6934 * condition. Based on their implementation, we put
6935 * in the following code to cover it as well.
6937 if (sctp_writeable(sk
))
6938 mask
|= POLLOUT
| POLLWRNORM
;
6943 /********************************************************************
6944 * 2nd Level Abstractions
6945 ********************************************************************/
6947 static struct sctp_bind_bucket
*sctp_bucket_create(
6948 struct sctp_bind_hashbucket
*head
, struct net
*net
, unsigned short snum
)
6950 struct sctp_bind_bucket
*pp
;
6952 pp
= kmem_cache_alloc(sctp_bucket_cachep
, GFP_ATOMIC
);
6954 SCTP_DBG_OBJCNT_INC(bind_bucket
);
6957 INIT_HLIST_HEAD(&pp
->owner
);
6959 hlist_add_head(&pp
->node
, &head
->chain
);
6964 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6965 static void sctp_bucket_destroy(struct sctp_bind_bucket
*pp
)
6967 if (pp
&& hlist_empty(&pp
->owner
)) {
6968 __hlist_del(&pp
->node
);
6969 kmem_cache_free(sctp_bucket_cachep
, pp
);
6970 SCTP_DBG_OBJCNT_DEC(bind_bucket
);
6974 /* Release this socket's reference to a local port. */
6975 static inline void __sctp_put_port(struct sock
*sk
)
6977 struct sctp_bind_hashbucket
*head
=
6978 &sctp_port_hashtable
[sctp_phashfn(sock_net(sk
),
6979 inet_sk(sk
)->inet_num
)];
6980 struct sctp_bind_bucket
*pp
;
6982 spin_lock(&head
->lock
);
6983 pp
= sctp_sk(sk
)->bind_hash
;
6984 __sk_del_bind_node(sk
);
6985 sctp_sk(sk
)->bind_hash
= NULL
;
6986 inet_sk(sk
)->inet_num
= 0;
6987 sctp_bucket_destroy(pp
);
6988 spin_unlock(&head
->lock
);
6991 void sctp_put_port(struct sock
*sk
)
6994 __sctp_put_port(sk
);
6999 * The system picks an ephemeral port and choose an address set equivalent
7000 * to binding with a wildcard address.
7001 * One of those addresses will be the primary address for the association.
7002 * This automatically enables the multihoming capability of SCTP.
7004 static int sctp_autobind(struct sock
*sk
)
7006 union sctp_addr autoaddr
;
7010 /* Initialize a local sockaddr structure to INADDR_ANY. */
7011 af
= sctp_sk(sk
)->pf
->af
;
7013 port
= htons(inet_sk(sk
)->inet_num
);
7014 af
->inaddr_any(&autoaddr
, port
);
7016 return sctp_do_bind(sk
, &autoaddr
, af
->sockaddr_len
);
7019 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
7022 * 4.2 The cmsghdr Structure *
7024 * When ancillary data is sent or received, any number of ancillary data
7025 * objects can be specified by the msg_control and msg_controllen members of
7026 * the msghdr structure, because each object is preceded by
7027 * a cmsghdr structure defining the object's length (the cmsg_len member).
7028 * Historically Berkeley-derived implementations have passed only one object
7029 * at a time, but this API allows multiple objects to be
7030 * passed in a single call to sendmsg() or recvmsg(). The following example
7031 * shows two ancillary data objects in a control buffer.
7033 * |<--------------------------- msg_controllen -------------------------->|
7036 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
7038 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
7041 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
7043 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
7046 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7047 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
7049 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
7051 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7058 static int sctp_msghdr_parse(const struct msghdr
*msg
, sctp_cmsgs_t
*cmsgs
)
7060 struct cmsghdr
*cmsg
;
7061 struct msghdr
*my_msg
= (struct msghdr
*)msg
;
7063 for_each_cmsghdr(cmsg
, my_msg
) {
7064 if (!CMSG_OK(my_msg
, cmsg
))
7067 /* Should we parse this header or ignore? */
7068 if (cmsg
->cmsg_level
!= IPPROTO_SCTP
)
7071 /* Strictly check lengths following example in SCM code. */
7072 switch (cmsg
->cmsg_type
) {
7074 /* SCTP Socket API Extension
7075 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
7077 * This cmsghdr structure provides information for
7078 * initializing new SCTP associations with sendmsg().
7079 * The SCTP_INITMSG socket option uses this same data
7080 * structure. This structure is not used for
7083 * cmsg_level cmsg_type cmsg_data[]
7084 * ------------ ------------ ----------------------
7085 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
7087 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(struct sctp_initmsg
)))
7090 cmsgs
->init
= CMSG_DATA(cmsg
);
7094 /* SCTP Socket API Extension
7095 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
7097 * This cmsghdr structure specifies SCTP options for
7098 * sendmsg() and describes SCTP header information
7099 * about a received message through recvmsg().
7101 * cmsg_level cmsg_type cmsg_data[]
7102 * ------------ ------------ ----------------------
7103 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
7105 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(struct sctp_sndrcvinfo
)))
7108 cmsgs
->srinfo
= CMSG_DATA(cmsg
);
7110 if (cmsgs
->srinfo
->sinfo_flags
&
7111 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
7112 SCTP_SACK_IMMEDIATELY
| SCTP_PR_SCTP_MASK
|
7113 SCTP_ABORT
| SCTP_EOF
))
7118 /* SCTP Socket API Extension
7119 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
7121 * This cmsghdr structure specifies SCTP options for
7122 * sendmsg(). This structure and SCTP_RCVINFO replaces
7123 * SCTP_SNDRCV which has been deprecated.
7125 * cmsg_level cmsg_type cmsg_data[]
7126 * ------------ ------------ ---------------------
7127 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
7129 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(struct sctp_sndinfo
)))
7132 cmsgs
->sinfo
= CMSG_DATA(cmsg
);
7134 if (cmsgs
->sinfo
->snd_flags
&
7135 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
7136 SCTP_SACK_IMMEDIATELY
| SCTP_PR_SCTP_MASK
|
7137 SCTP_ABORT
| SCTP_EOF
))
7149 * Wait for a packet..
7150 * Note: This function is the same function as in core/datagram.c
7151 * with a few modifications to make lksctp work.
7153 static int sctp_wait_for_packet(struct sock
*sk
, int *err
, long *timeo_p
)
7158 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
7160 /* Socket errors? */
7161 error
= sock_error(sk
);
7165 if (!skb_queue_empty(&sk
->sk_receive_queue
))
7168 /* Socket shut down? */
7169 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
7172 /* Sequenced packets can come disconnected. If so we report the
7177 /* Is there a good reason to think that we may receive some data? */
7178 if (list_empty(&sctp_sk(sk
)->ep
->asocs
) && !sctp_sstate(sk
, LISTENING
))
7181 /* Handle signals. */
7182 if (signal_pending(current
))
7185 /* Let another process have a go. Since we are going to sleep
7186 * anyway. Note: This may cause odd behaviors if the message
7187 * does not fit in the user's buffer, but this seems to be the
7188 * only way to honor MSG_DONTWAIT realistically.
7191 *timeo_p
= schedule_timeout(*timeo_p
);
7195 finish_wait(sk_sleep(sk
), &wait
);
7199 error
= sock_intr_errno(*timeo_p
);
7202 finish_wait(sk_sleep(sk
), &wait
);
7207 /* Receive a datagram.
7208 * Note: This is pretty much the same routine as in core/datagram.c
7209 * with a few changes to make lksctp work.
7211 struct sk_buff
*sctp_skb_recv_datagram(struct sock
*sk
, int flags
,
7212 int noblock
, int *err
)
7215 struct sk_buff
*skb
;
7218 timeo
= sock_rcvtimeo(sk
, noblock
);
7220 pr_debug("%s: timeo:%ld, max:%ld\n", __func__
, timeo
,
7221 MAX_SCHEDULE_TIMEOUT
);
7224 /* Again only user level code calls this function,
7225 * so nothing interrupt level
7226 * will suddenly eat the receive_queue.
7228 * Look at current nfs client by the way...
7229 * However, this function was correct in any case. 8)
7231 if (flags
& MSG_PEEK
) {
7232 skb
= skb_peek(&sk
->sk_receive_queue
);
7234 atomic_inc(&skb
->users
);
7236 skb
= __skb_dequeue(&sk
->sk_receive_queue
);
7242 /* Caller is allowed not to check sk->sk_err before calling. */
7243 error
= sock_error(sk
);
7247 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
7250 if (sk_can_busy_loop(sk
) &&
7251 sk_busy_loop(sk
, noblock
))
7254 /* User doesn't want to wait. */
7258 } while (sctp_wait_for_packet(sk
, err
, &timeo
) == 0);
7267 /* If sndbuf has changed, wake up per association sndbuf waiters. */
7268 static void __sctp_write_space(struct sctp_association
*asoc
)
7270 struct sock
*sk
= asoc
->base
.sk
;
7272 if (sctp_wspace(asoc
) <= 0)
7275 if (waitqueue_active(&asoc
->wait
))
7276 wake_up_interruptible(&asoc
->wait
);
7278 if (sctp_writeable(sk
)) {
7279 struct socket_wq
*wq
;
7282 wq
= rcu_dereference(sk
->sk_wq
);
7284 if (waitqueue_active(&wq
->wait
))
7285 wake_up_interruptible(&wq
->wait
);
7287 /* Note that we try to include the Async I/O support
7288 * here by modeling from the current TCP/UDP code.
7289 * We have not tested with it yet.
7291 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
))
7292 sock_wake_async(wq
, SOCK_WAKE_SPACE
, POLL_OUT
);
7298 static void sctp_wake_up_waiters(struct sock
*sk
,
7299 struct sctp_association
*asoc
)
7301 struct sctp_association
*tmp
= asoc
;
7303 /* We do accounting for the sndbuf space per association,
7304 * so we only need to wake our own association.
7306 if (asoc
->ep
->sndbuf_policy
)
7307 return __sctp_write_space(asoc
);
7309 /* If association goes down and is just flushing its
7310 * outq, then just normally notify others.
7312 if (asoc
->base
.dead
)
7313 return sctp_write_space(sk
);
7315 /* Accounting for the sndbuf space is per socket, so we
7316 * need to wake up others, try to be fair and in case of
7317 * other associations, let them have a go first instead
7318 * of just doing a sctp_write_space() call.
7320 * Note that we reach sctp_wake_up_waiters() only when
7321 * associations free up queued chunks, thus we are under
7322 * lock and the list of associations on a socket is
7323 * guaranteed not to change.
7325 for (tmp
= list_next_entry(tmp
, asocs
); 1;
7326 tmp
= list_next_entry(tmp
, asocs
)) {
7327 /* Manually skip the head element. */
7328 if (&tmp
->asocs
== &((sctp_sk(sk
))->ep
->asocs
))
7330 /* Wake up association. */
7331 __sctp_write_space(tmp
);
7332 /* We've reached the end. */
7338 /* Do accounting for the sndbuf space.
7339 * Decrement the used sndbuf space of the corresponding association by the
7340 * data size which was just transmitted(freed).
7342 static void sctp_wfree(struct sk_buff
*skb
)
7344 struct sctp_chunk
*chunk
= skb_shinfo(skb
)->destructor_arg
;
7345 struct sctp_association
*asoc
= chunk
->asoc
;
7346 struct sock
*sk
= asoc
->base
.sk
;
7348 asoc
->sndbuf_used
-= SCTP_DATA_SNDSIZE(chunk
) +
7349 sizeof(struct sk_buff
) +
7350 sizeof(struct sctp_chunk
);
7352 atomic_sub(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
7355 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
7357 sk
->sk_wmem_queued
-= skb
->truesize
;
7358 sk_mem_uncharge(sk
, skb
->truesize
);
7361 sctp_wake_up_waiters(sk
, asoc
);
7363 sctp_association_put(asoc
);
7366 /* Do accounting for the receive space on the socket.
7367 * Accounting for the association is done in ulpevent.c
7368 * We set this as a destructor for the cloned data skbs so that
7369 * accounting is done at the correct time.
7371 void sctp_sock_rfree(struct sk_buff
*skb
)
7373 struct sock
*sk
= skb
->sk
;
7374 struct sctp_ulpevent
*event
= sctp_skb2event(skb
);
7376 atomic_sub(event
->rmem_len
, &sk
->sk_rmem_alloc
);
7379 * Mimic the behavior of sock_rfree
7381 sk_mem_uncharge(sk
, event
->rmem_len
);
7385 /* Helper function to wait for space in the sndbuf. */
7386 static int sctp_wait_for_sndbuf(struct sctp_association
*asoc
, long *timeo_p
,
7389 struct sock
*sk
= asoc
->base
.sk
;
7391 long current_timeo
= *timeo_p
;
7394 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__
, asoc
,
7397 /* Increment the association's refcnt. */
7398 sctp_association_hold(asoc
);
7400 /* Wait on the association specific sndbuf space. */
7402 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
7403 TASK_INTERRUPTIBLE
);
7406 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
7409 if (signal_pending(current
))
7410 goto do_interrupted
;
7411 if (msg_len
<= sctp_wspace(asoc
))
7414 /* Let another process have a go. Since we are going
7418 current_timeo
= schedule_timeout(current_timeo
);
7419 BUG_ON(sk
!= asoc
->base
.sk
);
7422 *timeo_p
= current_timeo
;
7426 finish_wait(&asoc
->wait
, &wait
);
7428 /* Release the association's refcnt. */
7429 sctp_association_put(asoc
);
7438 err
= sock_intr_errno(*timeo_p
);
7446 void sctp_data_ready(struct sock
*sk
)
7448 struct socket_wq
*wq
;
7451 wq
= rcu_dereference(sk
->sk_wq
);
7452 if (skwq_has_sleeper(wq
))
7453 wake_up_interruptible_sync_poll(&wq
->wait
, POLLIN
|
7454 POLLRDNORM
| POLLRDBAND
);
7455 sk_wake_async(sk
, SOCK_WAKE_WAITD
, POLL_IN
);
7459 /* If socket sndbuf has changed, wake up all per association waiters. */
7460 void sctp_write_space(struct sock
*sk
)
7462 struct sctp_association
*asoc
;
7464 /* Wake up the tasks in each wait queue. */
7465 list_for_each_entry(asoc
, &((sctp_sk(sk
))->ep
->asocs
), asocs
) {
7466 __sctp_write_space(asoc
);
7470 /* Is there any sndbuf space available on the socket?
7472 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
7473 * associations on the same socket. For a UDP-style socket with
7474 * multiple associations, it is possible for it to be "unwriteable"
7475 * prematurely. I assume that this is acceptable because
7476 * a premature "unwriteable" is better than an accidental "writeable" which
7477 * would cause an unwanted block under certain circumstances. For the 1-1
7478 * UDP-style sockets or TCP-style sockets, this code should work.
7481 static int sctp_writeable(struct sock
*sk
)
7485 amt
= sk
->sk_sndbuf
- sk_wmem_alloc_get(sk
);
7491 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
7492 * returns immediately with EINPROGRESS.
7494 static int sctp_wait_for_connect(struct sctp_association
*asoc
, long *timeo_p
)
7496 struct sock
*sk
= asoc
->base
.sk
;
7498 long current_timeo
= *timeo_p
;
7501 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__
, asoc
, *timeo_p
);
7503 /* Increment the association's refcnt. */
7504 sctp_association_hold(asoc
);
7507 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
7508 TASK_INTERRUPTIBLE
);
7511 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
7513 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
7516 if (signal_pending(current
))
7517 goto do_interrupted
;
7519 if (sctp_state(asoc
, ESTABLISHED
))
7522 /* Let another process have a go. Since we are going
7526 current_timeo
= schedule_timeout(current_timeo
);
7529 *timeo_p
= current_timeo
;
7533 finish_wait(&asoc
->wait
, &wait
);
7535 /* Release the association's refcnt. */
7536 sctp_association_put(asoc
);
7541 if (asoc
->init_err_counter
+ 1 > asoc
->max_init_attempts
)
7544 err
= -ECONNREFUSED
;
7548 err
= sock_intr_errno(*timeo_p
);
7556 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
)
7558 struct sctp_endpoint
*ep
;
7562 ep
= sctp_sk(sk
)->ep
;
7566 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
,
7567 TASK_INTERRUPTIBLE
);
7569 if (list_empty(&ep
->asocs
)) {
7571 timeo
= schedule_timeout(timeo
);
7576 if (!sctp_sstate(sk
, LISTENING
))
7580 if (!list_empty(&ep
->asocs
))
7583 err
= sock_intr_errno(timeo
);
7584 if (signal_pending(current
))
7592 finish_wait(sk_sleep(sk
), &wait
);
7597 static void sctp_wait_for_close(struct sock
*sk
, long timeout
)
7602 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
7603 if (list_empty(&sctp_sk(sk
)->ep
->asocs
))
7606 timeout
= schedule_timeout(timeout
);
7608 } while (!signal_pending(current
) && timeout
);
7610 finish_wait(sk_sleep(sk
), &wait
);
7613 static void sctp_skb_set_owner_r_frag(struct sk_buff
*skb
, struct sock
*sk
)
7615 struct sk_buff
*frag
;
7620 /* Don't forget the fragments. */
7621 skb_walk_frags(skb
, frag
)
7622 sctp_skb_set_owner_r_frag(frag
, sk
);
7625 sctp_skb_set_owner_r(skb
, sk
);
7628 void sctp_copy_sock(struct sock
*newsk
, struct sock
*sk
,
7629 struct sctp_association
*asoc
)
7631 struct inet_sock
*inet
= inet_sk(sk
);
7632 struct inet_sock
*newinet
;
7634 newsk
->sk_type
= sk
->sk_type
;
7635 newsk
->sk_bound_dev_if
= sk
->sk_bound_dev_if
;
7636 newsk
->sk_flags
= sk
->sk_flags
;
7637 newsk
->sk_tsflags
= sk
->sk_tsflags
;
7638 newsk
->sk_no_check_tx
= sk
->sk_no_check_tx
;
7639 newsk
->sk_no_check_rx
= sk
->sk_no_check_rx
;
7640 newsk
->sk_reuse
= sk
->sk_reuse
;
7642 newsk
->sk_shutdown
= sk
->sk_shutdown
;
7643 newsk
->sk_destruct
= sctp_destruct_sock
;
7644 newsk
->sk_family
= sk
->sk_family
;
7645 newsk
->sk_protocol
= IPPROTO_SCTP
;
7646 newsk
->sk_backlog_rcv
= sk
->sk_prot
->backlog_rcv
;
7647 newsk
->sk_sndbuf
= sk
->sk_sndbuf
;
7648 newsk
->sk_rcvbuf
= sk
->sk_rcvbuf
;
7649 newsk
->sk_lingertime
= sk
->sk_lingertime
;
7650 newsk
->sk_rcvtimeo
= sk
->sk_rcvtimeo
;
7651 newsk
->sk_sndtimeo
= sk
->sk_sndtimeo
;
7652 newsk
->sk_rxhash
= sk
->sk_rxhash
;
7654 newinet
= inet_sk(newsk
);
7656 /* Initialize sk's sport, dport, rcv_saddr and daddr for
7657 * getsockname() and getpeername()
7659 newinet
->inet_sport
= inet
->inet_sport
;
7660 newinet
->inet_saddr
= inet
->inet_saddr
;
7661 newinet
->inet_rcv_saddr
= inet
->inet_rcv_saddr
;
7662 newinet
->inet_dport
= htons(asoc
->peer
.port
);
7663 newinet
->pmtudisc
= inet
->pmtudisc
;
7664 newinet
->inet_id
= asoc
->next_tsn
^ jiffies
;
7666 newinet
->uc_ttl
= inet
->uc_ttl
;
7667 newinet
->mc_loop
= 1;
7668 newinet
->mc_ttl
= 1;
7669 newinet
->mc_index
= 0;
7670 newinet
->mc_list
= NULL
;
7672 if (newsk
->sk_flags
& SK_FLAGS_TIMESTAMP
)
7673 net_enable_timestamp();
7675 security_sk_clone(sk
, newsk
);
7678 static inline void sctp_copy_descendant(struct sock
*sk_to
,
7679 const struct sock
*sk_from
)
7681 int ancestor_size
= sizeof(struct inet_sock
) +
7682 sizeof(struct sctp_sock
) -
7683 offsetof(struct sctp_sock
, auto_asconf_list
);
7685 if (sk_from
->sk_family
== PF_INET6
)
7686 ancestor_size
+= sizeof(struct ipv6_pinfo
);
7688 __inet_sk_copy_descendant(sk_to
, sk_from
, ancestor_size
);
7691 /* Populate the fields of the newsk from the oldsk and migrate the assoc
7692 * and its messages to the newsk.
7694 static void sctp_sock_migrate(struct sock
*oldsk
, struct sock
*newsk
,
7695 struct sctp_association
*assoc
,
7696 sctp_socket_type_t type
)
7698 struct sctp_sock
*oldsp
= sctp_sk(oldsk
);
7699 struct sctp_sock
*newsp
= sctp_sk(newsk
);
7700 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
7701 struct sctp_endpoint
*newep
= newsp
->ep
;
7702 struct sk_buff
*skb
, *tmp
;
7703 struct sctp_ulpevent
*event
;
7704 struct sctp_bind_hashbucket
*head
;
7706 /* Migrate socket buffer sizes and all the socket level options to the
7709 newsk
->sk_sndbuf
= oldsk
->sk_sndbuf
;
7710 newsk
->sk_rcvbuf
= oldsk
->sk_rcvbuf
;
7711 /* Brute force copy old sctp opt. */
7712 sctp_copy_descendant(newsk
, oldsk
);
7714 /* Restore the ep value that was overwritten with the above structure
7720 /* Hook this new socket in to the bind_hash list. */
7721 head
= &sctp_port_hashtable
[sctp_phashfn(sock_net(oldsk
),
7722 inet_sk(oldsk
)->inet_num
)];
7723 spin_lock_bh(&head
->lock
);
7724 pp
= sctp_sk(oldsk
)->bind_hash
;
7725 sk_add_bind_node(newsk
, &pp
->owner
);
7726 sctp_sk(newsk
)->bind_hash
= pp
;
7727 inet_sk(newsk
)->inet_num
= inet_sk(oldsk
)->inet_num
;
7728 spin_unlock_bh(&head
->lock
);
7730 /* Copy the bind_addr list from the original endpoint to the new
7731 * endpoint so that we can handle restarts properly
7733 sctp_bind_addr_dup(&newsp
->ep
->base
.bind_addr
,
7734 &oldsp
->ep
->base
.bind_addr
, GFP_KERNEL
);
7736 /* Move any messages in the old socket's receive queue that are for the
7737 * peeled off association to the new socket's receive queue.
7739 sctp_skb_for_each(skb
, &oldsk
->sk_receive_queue
, tmp
) {
7740 event
= sctp_skb2event(skb
);
7741 if (event
->asoc
== assoc
) {
7742 __skb_unlink(skb
, &oldsk
->sk_receive_queue
);
7743 __skb_queue_tail(&newsk
->sk_receive_queue
, skb
);
7744 sctp_skb_set_owner_r_frag(skb
, newsk
);
7748 /* Clean up any messages pending delivery due to partial
7749 * delivery. Three cases:
7750 * 1) No partial deliver; no work.
7751 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7752 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7754 skb_queue_head_init(&newsp
->pd_lobby
);
7755 atomic_set(&sctp_sk(newsk
)->pd_mode
, assoc
->ulpq
.pd_mode
);
7757 if (atomic_read(&sctp_sk(oldsk
)->pd_mode
)) {
7758 struct sk_buff_head
*queue
;
7760 /* Decide which queue to move pd_lobby skbs to. */
7761 if (assoc
->ulpq
.pd_mode
) {
7762 queue
= &newsp
->pd_lobby
;
7764 queue
= &newsk
->sk_receive_queue
;
7766 /* Walk through the pd_lobby, looking for skbs that
7767 * need moved to the new socket.
7769 sctp_skb_for_each(skb
, &oldsp
->pd_lobby
, tmp
) {
7770 event
= sctp_skb2event(skb
);
7771 if (event
->asoc
== assoc
) {
7772 __skb_unlink(skb
, &oldsp
->pd_lobby
);
7773 __skb_queue_tail(queue
, skb
);
7774 sctp_skb_set_owner_r_frag(skb
, newsk
);
7778 /* Clear up any skbs waiting for the partial
7779 * delivery to finish.
7781 if (assoc
->ulpq
.pd_mode
)
7782 sctp_clear_pd(oldsk
, NULL
);
7786 sctp_skb_for_each(skb
, &assoc
->ulpq
.reasm
, tmp
)
7787 sctp_skb_set_owner_r_frag(skb
, newsk
);
7789 sctp_skb_for_each(skb
, &assoc
->ulpq
.lobby
, tmp
)
7790 sctp_skb_set_owner_r_frag(skb
, newsk
);
7792 /* Set the type of socket to indicate that it is peeled off from the
7793 * original UDP-style socket or created with the accept() call on a
7794 * TCP-style socket..
7798 /* Mark the new socket "in-use" by the user so that any packets
7799 * that may arrive on the association after we've moved it are
7800 * queued to the backlog. This prevents a potential race between
7801 * backlog processing on the old socket and new-packet processing
7802 * on the new socket.
7804 * The caller has just allocated newsk so we can guarantee that other
7805 * paths won't try to lock it and then oldsk.
7807 lock_sock_nested(newsk
, SINGLE_DEPTH_NESTING
);
7808 sctp_assoc_migrate(assoc
, newsk
);
7810 /* If the association on the newsk is already closed before accept()
7811 * is called, set RCV_SHUTDOWN flag.
7813 if (sctp_state(assoc
, CLOSED
) && sctp_style(newsk
, TCP
)) {
7814 newsk
->sk_state
= SCTP_SS_CLOSED
;
7815 newsk
->sk_shutdown
|= RCV_SHUTDOWN
;
7817 newsk
->sk_state
= SCTP_SS_ESTABLISHED
;
7820 release_sock(newsk
);
7824 /* This proto struct describes the ULP interface for SCTP. */
7825 struct proto sctp_prot
= {
7827 .owner
= THIS_MODULE
,
7828 .close
= sctp_close
,
7829 .connect
= sctp_connect
,
7830 .disconnect
= sctp_disconnect
,
7831 .accept
= sctp_accept
,
7832 .ioctl
= sctp_ioctl
,
7833 .init
= sctp_init_sock
,
7834 .destroy
= sctp_destroy_sock
,
7835 .shutdown
= sctp_shutdown
,
7836 .setsockopt
= sctp_setsockopt
,
7837 .getsockopt
= sctp_getsockopt
,
7838 .sendmsg
= sctp_sendmsg
,
7839 .recvmsg
= sctp_recvmsg
,
7841 .backlog_rcv
= sctp_backlog_rcv
,
7843 .unhash
= sctp_unhash
,
7844 .get_port
= sctp_get_port
,
7845 .obj_size
= sizeof(struct sctp_sock
),
7846 .sysctl_mem
= sysctl_sctp_mem
,
7847 .sysctl_rmem
= sysctl_sctp_rmem
,
7848 .sysctl_wmem
= sysctl_sctp_wmem
,
7849 .memory_pressure
= &sctp_memory_pressure
,
7850 .enter_memory_pressure
= sctp_enter_memory_pressure
,
7851 .memory_allocated
= &sctp_memory_allocated
,
7852 .sockets_allocated
= &sctp_sockets_allocated
,
7855 #if IS_ENABLED(CONFIG_IPV6)
7857 #include <net/transp_v6.h>
7858 static void sctp_v6_destroy_sock(struct sock
*sk
)
7860 sctp_destroy_sock(sk
);
7861 inet6_destroy_sock(sk
);
7864 struct proto sctpv6_prot
= {
7866 .owner
= THIS_MODULE
,
7867 .close
= sctp_close
,
7868 .connect
= sctp_connect
,
7869 .disconnect
= sctp_disconnect
,
7870 .accept
= sctp_accept
,
7871 .ioctl
= sctp_ioctl
,
7872 .init
= sctp_init_sock
,
7873 .destroy
= sctp_v6_destroy_sock
,
7874 .shutdown
= sctp_shutdown
,
7875 .setsockopt
= sctp_setsockopt
,
7876 .getsockopt
= sctp_getsockopt
,
7877 .sendmsg
= sctp_sendmsg
,
7878 .recvmsg
= sctp_recvmsg
,
7880 .backlog_rcv
= sctp_backlog_rcv
,
7882 .unhash
= sctp_unhash
,
7883 .get_port
= sctp_get_port
,
7884 .obj_size
= sizeof(struct sctp6_sock
),
7885 .sysctl_mem
= sysctl_sctp_mem
,
7886 .sysctl_rmem
= sysctl_sctp_rmem
,
7887 .sysctl_wmem
= sysctl_sctp_wmem
,
7888 .memory_pressure
= &sctp_memory_pressure
,
7889 .enter_memory_pressure
= sctp_enter_memory_pressure
,
7890 .memory_allocated
= &sctp_memory_allocated
,
7891 .sockets_allocated
= &sctp_sockets_allocated
,
7893 #endif /* IS_ENABLED(CONFIG_IPV6) */