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 <linux/types.h>
56 #include <linux/kernel.h>
57 #include <linux/wait.h>
58 #include <linux/time.h>
60 #include <linux/capability.h>
61 #include <linux/fcntl.h>
62 #include <linux/poll.h>
63 #include <linux/init.h>
64 #include <linux/crypto.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>
75 #include <linux/socket.h> /* for sa_family_t */
76 #include <linux/export.h>
78 #include <net/sctp/sctp.h>
79 #include <net/sctp/sm.h>
81 /* Forward declarations for internal helper functions. */
82 static int sctp_writeable(struct sock
*sk
);
83 static void sctp_wfree(struct sk_buff
*skb
);
84 static int sctp_wait_for_sndbuf(struct sctp_association
*, long *timeo_p
,
86 static int sctp_wait_for_packet(struct sock
*sk
, int *err
, long *timeo_p
);
87 static int sctp_wait_for_connect(struct sctp_association
*, long *timeo_p
);
88 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
);
89 static void sctp_wait_for_close(struct sock
*sk
, long timeo
);
90 static void sctp_destruct_sock(struct sock
*sk
);
91 static struct sctp_af
*sctp_sockaddr_af(struct sctp_sock
*opt
,
92 union sctp_addr
*addr
, int len
);
93 static int sctp_bindx_add(struct sock
*, struct sockaddr
*, int);
94 static int sctp_bindx_rem(struct sock
*, struct sockaddr
*, int);
95 static int sctp_send_asconf_add_ip(struct sock
*, struct sockaddr
*, int);
96 static int sctp_send_asconf_del_ip(struct sock
*, struct sockaddr
*, int);
97 static int sctp_send_asconf(struct sctp_association
*asoc
,
98 struct sctp_chunk
*chunk
);
99 static int sctp_do_bind(struct sock
*, union sctp_addr
*, int);
100 static int sctp_autobind(struct sock
*sk
);
101 static void sctp_sock_migrate(struct sock
*, struct sock
*,
102 struct sctp_association
*, sctp_socket_type_t
);
104 extern struct kmem_cache
*sctp_bucket_cachep
;
105 extern long sysctl_sctp_mem
[3];
106 extern int sysctl_sctp_rmem
[3];
107 extern int sysctl_sctp_wmem
[3];
109 static int sctp_memory_pressure
;
110 static atomic_long_t sctp_memory_allocated
;
111 struct percpu_counter sctp_sockets_allocated
;
113 static void sctp_enter_memory_pressure(struct sock
*sk
)
115 sctp_memory_pressure
= 1;
119 /* Get the sndbuf space available at the time on the association. */
120 static inline int sctp_wspace(struct sctp_association
*asoc
)
124 if (asoc
->ep
->sndbuf_policy
)
125 amt
= asoc
->sndbuf_used
;
127 amt
= sk_wmem_alloc_get(asoc
->base
.sk
);
129 if (amt
>= asoc
->base
.sk
->sk_sndbuf
) {
130 if (asoc
->base
.sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)
133 amt
= sk_stream_wspace(asoc
->base
.sk
);
138 amt
= asoc
->base
.sk
->sk_sndbuf
- amt
;
143 /* Increment the used sndbuf space count of the corresponding association by
144 * the size of the outgoing data chunk.
145 * Also, set the skb destructor for sndbuf accounting later.
147 * Since it is always 1-1 between chunk and skb, and also a new skb is always
148 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
149 * destructor in the data chunk skb for the purpose of the sndbuf space
152 static inline void sctp_set_owner_w(struct sctp_chunk
*chunk
)
154 struct sctp_association
*asoc
= chunk
->asoc
;
155 struct sock
*sk
= asoc
->base
.sk
;
157 /* The sndbuf space is tracked per association. */
158 sctp_association_hold(asoc
);
160 skb_set_owner_w(chunk
->skb
, sk
);
162 chunk
->skb
->destructor
= sctp_wfree
;
163 /* Save the chunk pointer in skb for sctp_wfree to use later. */
164 *((struct sctp_chunk
**)(chunk
->skb
->cb
)) = chunk
;
166 asoc
->sndbuf_used
+= SCTP_DATA_SNDSIZE(chunk
) +
167 sizeof(struct sk_buff
) +
168 sizeof(struct sctp_chunk
);
170 atomic_add(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
171 sk
->sk_wmem_queued
+= chunk
->skb
->truesize
;
172 sk_mem_charge(sk
, chunk
->skb
->truesize
);
175 /* Verify that this is a valid address. */
176 static inline int sctp_verify_addr(struct sock
*sk
, union sctp_addr
*addr
,
181 /* Verify basic sockaddr. */
182 af
= sctp_sockaddr_af(sctp_sk(sk
), addr
, len
);
186 /* Is this a valid SCTP address? */
187 if (!af
->addr_valid(addr
, sctp_sk(sk
), NULL
))
190 if (!sctp_sk(sk
)->pf
->send_verify(sctp_sk(sk
), (addr
)))
196 /* Look up the association by its id. If this is not a UDP-style
197 * socket, the ID field is always ignored.
199 struct sctp_association
*sctp_id2assoc(struct sock
*sk
, sctp_assoc_t id
)
201 struct sctp_association
*asoc
= NULL
;
203 /* If this is not a UDP-style socket, assoc id should be ignored. */
204 if (!sctp_style(sk
, UDP
)) {
205 /* Return NULL if the socket state is not ESTABLISHED. It
206 * could be a TCP-style listening socket or a socket which
207 * hasn't yet called connect() to establish an association.
209 if (!sctp_sstate(sk
, ESTABLISHED
))
212 /* Get the first and the only association from the list. */
213 if (!list_empty(&sctp_sk(sk
)->ep
->asocs
))
214 asoc
= list_entry(sctp_sk(sk
)->ep
->asocs
.next
,
215 struct sctp_association
, asocs
);
219 /* Otherwise this is a UDP-style socket. */
220 if (!id
|| (id
== (sctp_assoc_t
)-1))
223 spin_lock_bh(&sctp_assocs_id_lock
);
224 asoc
= (struct sctp_association
*)idr_find(&sctp_assocs_id
, (int)id
);
225 spin_unlock_bh(&sctp_assocs_id_lock
);
227 if (!asoc
|| (asoc
->base
.sk
!= sk
) || asoc
->base
.dead
)
233 /* Look up the transport from an address and an assoc id. If both address and
234 * id are specified, the associations matching the address and the id should be
237 static struct sctp_transport
*sctp_addr_id2transport(struct sock
*sk
,
238 struct sockaddr_storage
*addr
,
241 struct sctp_association
*addr_asoc
= NULL
, *id_asoc
= NULL
;
242 struct sctp_transport
*transport
;
243 union sctp_addr
*laddr
= (union sctp_addr
*)addr
;
245 addr_asoc
= sctp_endpoint_lookup_assoc(sctp_sk(sk
)->ep
,
252 id_asoc
= sctp_id2assoc(sk
, id
);
253 if (id_asoc
&& (id_asoc
!= addr_asoc
))
256 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
257 (union sctp_addr
*)addr
);
262 /* API 3.1.2 bind() - UDP Style Syntax
263 * The syntax of bind() is,
265 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
267 * sd - the socket descriptor returned by socket().
268 * addr - the address structure (struct sockaddr_in or struct
269 * sockaddr_in6 [RFC 2553]),
270 * addr_len - the size of the address structure.
272 static int sctp_bind(struct sock
*sk
, struct sockaddr
*addr
, int addr_len
)
278 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__
, sk
,
281 /* Disallow binding twice. */
282 if (!sctp_sk(sk
)->ep
->base
.bind_addr
.port
)
283 retval
= sctp_do_bind(sk
, (union sctp_addr
*)addr
,
293 static long sctp_get_port_local(struct sock
*, union sctp_addr
*);
295 /* Verify this is a valid sockaddr. */
296 static struct sctp_af
*sctp_sockaddr_af(struct sctp_sock
*opt
,
297 union sctp_addr
*addr
, int len
)
301 /* Check minimum size. */
302 if (len
< sizeof (struct sockaddr
))
305 /* V4 mapped address are really of AF_INET family */
306 if (addr
->sa
.sa_family
== AF_INET6
&&
307 ipv6_addr_v4mapped(&addr
->v6
.sin6_addr
)) {
308 if (!opt
->pf
->af_supported(AF_INET
, opt
))
311 /* Does this PF support this AF? */
312 if (!opt
->pf
->af_supported(addr
->sa
.sa_family
, opt
))
316 /* If we get this far, af is valid. */
317 af
= sctp_get_af_specific(addr
->sa
.sa_family
);
319 if (len
< af
->sockaddr_len
)
325 /* Bind a local address either to an endpoint or to an association. */
326 static int sctp_do_bind(struct sock
*sk
, union sctp_addr
*addr
, int len
)
328 struct net
*net
= sock_net(sk
);
329 struct sctp_sock
*sp
= sctp_sk(sk
);
330 struct sctp_endpoint
*ep
= sp
->ep
;
331 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
336 /* Common sockaddr verification. */
337 af
= sctp_sockaddr_af(sp
, addr
, len
);
339 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
340 __func__
, sk
, addr
, len
);
344 snum
= ntohs(addr
->v4
.sin_port
);
346 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
347 __func__
, sk
, &addr
->sa
, bp
->port
, snum
, len
);
349 /* PF specific bind() address verification. */
350 if (!sp
->pf
->bind_verify(sp
, addr
))
351 return -EADDRNOTAVAIL
;
353 /* We must either be unbound, or bind to the same port.
354 * It's OK to allow 0 ports if we are already bound.
355 * We'll just inhert an already bound port in this case
360 else if (snum
!= bp
->port
) {
361 pr_debug("%s: new port %d doesn't match existing port "
362 "%d\n", __func__
, snum
, bp
->port
);
367 if (snum
&& snum
< PROT_SOCK
&&
368 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
))
371 /* See if the address matches any of the addresses we may have
372 * already bound before checking against other endpoints.
374 if (sctp_bind_addr_match(bp
, addr
, sp
))
377 /* Make sure we are allowed to bind here.
378 * The function sctp_get_port_local() does duplicate address
381 addr
->v4
.sin_port
= htons(snum
);
382 if ((ret
= sctp_get_port_local(sk
, addr
))) {
386 /* Refresh ephemeral port. */
388 bp
->port
= inet_sk(sk
)->inet_num
;
390 /* Add the address to the bind address list.
391 * Use GFP_ATOMIC since BHs will be disabled.
393 ret
= sctp_add_bind_addr(bp
, addr
, SCTP_ADDR_SRC
, GFP_ATOMIC
);
395 /* Copy back into socket for getsockname() use. */
397 inet_sk(sk
)->inet_sport
= htons(inet_sk(sk
)->inet_num
);
398 af
->to_sk_saddr(addr
, sk
);
404 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
406 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
407 * at any one time. If a sender, after sending an ASCONF chunk, decides
408 * it needs to transfer another ASCONF Chunk, it MUST wait until the
409 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
410 * subsequent ASCONF. Note this restriction binds each side, so at any
411 * time two ASCONF may be in-transit on any given association (one sent
412 * from each endpoint).
414 static int sctp_send_asconf(struct sctp_association
*asoc
,
415 struct sctp_chunk
*chunk
)
417 struct net
*net
= sock_net(asoc
->base
.sk
);
420 /* If there is an outstanding ASCONF chunk, queue it for later
423 if (asoc
->addip_last_asconf
) {
424 list_add_tail(&chunk
->list
, &asoc
->addip_chunk_list
);
428 /* Hold the chunk until an ASCONF_ACK is received. */
429 sctp_chunk_hold(chunk
);
430 retval
= sctp_primitive_ASCONF(net
, asoc
, chunk
);
432 sctp_chunk_free(chunk
);
434 asoc
->addip_last_asconf
= chunk
;
440 /* Add a list of addresses as bind addresses to local endpoint or
443 * Basically run through each address specified in the addrs/addrcnt
444 * array/length pair, determine if it is IPv6 or IPv4 and call
445 * sctp_do_bind() on it.
447 * If any of them fails, then the operation will be reversed and the
448 * ones that were added will be removed.
450 * Only sctp_setsockopt_bindx() is supposed to call this function.
452 static int sctp_bindx_add(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
457 struct sockaddr
*sa_addr
;
460 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__
, sk
,
464 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
465 /* The list may contain either IPv4 or IPv6 address;
466 * determine the address length for walking thru the list.
469 af
= sctp_get_af_specific(sa_addr
->sa_family
);
475 retval
= sctp_do_bind(sk
, (union sctp_addr
*)sa_addr
,
478 addr_buf
+= af
->sockaddr_len
;
482 /* Failed. Cleanup the ones that have been added */
484 sctp_bindx_rem(sk
, addrs
, cnt
);
492 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
493 * associations that are part of the endpoint indicating that a list of local
494 * addresses are added to the endpoint.
496 * If any of the addresses is already in the bind address list of the
497 * association, we do not send the chunk for that association. But it will not
498 * affect other associations.
500 * Only sctp_setsockopt_bindx() is supposed to call this function.
502 static int sctp_send_asconf_add_ip(struct sock
*sk
,
503 struct sockaddr
*addrs
,
506 struct net
*net
= sock_net(sk
);
507 struct sctp_sock
*sp
;
508 struct sctp_endpoint
*ep
;
509 struct sctp_association
*asoc
;
510 struct sctp_bind_addr
*bp
;
511 struct sctp_chunk
*chunk
;
512 struct sctp_sockaddr_entry
*laddr
;
513 union sctp_addr
*addr
;
514 union sctp_addr saveaddr
;
521 if (!net
->sctp
.addip_enable
)
527 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
528 __func__
, sk
, addrs
, addrcnt
);
530 list_for_each_entry(asoc
, &ep
->asocs
, asocs
) {
531 if (!asoc
->peer
.asconf_capable
)
534 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_ADD_IP
)
537 if (!sctp_state(asoc
, ESTABLISHED
))
540 /* Check if any address in the packed array of addresses is
541 * in the bind address list of the association. If so,
542 * do not send the asconf chunk to its peer, but continue with
543 * other associations.
546 for (i
= 0; i
< addrcnt
; i
++) {
548 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
554 if (sctp_assoc_lookup_laddr(asoc
, addr
))
557 addr_buf
+= af
->sockaddr_len
;
562 /* Use the first valid address in bind addr list of
563 * association as Address Parameter of ASCONF CHUNK.
565 bp
= &asoc
->base
.bind_addr
;
566 p
= bp
->address_list
.next
;
567 laddr
= list_entry(p
, struct sctp_sockaddr_entry
, list
);
568 chunk
= sctp_make_asconf_update_ip(asoc
, &laddr
->a
, addrs
,
569 addrcnt
, SCTP_PARAM_ADD_IP
);
575 /* Add the new addresses to the bind address list with
576 * use_as_src set to 0.
579 for (i
= 0; i
< addrcnt
; i
++) {
581 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
582 memcpy(&saveaddr
, addr
, af
->sockaddr_len
);
583 retval
= sctp_add_bind_addr(bp
, &saveaddr
,
584 SCTP_ADDR_NEW
, GFP_ATOMIC
);
585 addr_buf
+= af
->sockaddr_len
;
587 if (asoc
->src_out_of_asoc_ok
) {
588 struct sctp_transport
*trans
;
590 list_for_each_entry(trans
,
591 &asoc
->peer
.transport_addr_list
, transports
) {
592 /* Clear the source and route cache */
593 dst_release(trans
->dst
);
594 trans
->cwnd
= min(4*asoc
->pathmtu
, max_t(__u32
,
595 2*asoc
->pathmtu
, 4380));
596 trans
->ssthresh
= asoc
->peer
.i
.a_rwnd
;
597 trans
->rto
= asoc
->rto_initial
;
598 sctp_max_rto(asoc
, trans
);
599 trans
->rtt
= trans
->srtt
= trans
->rttvar
= 0;
600 sctp_transport_route(trans
, NULL
,
601 sctp_sk(asoc
->base
.sk
));
604 retval
= sctp_send_asconf(asoc
, chunk
);
611 /* Remove a list of addresses from bind addresses list. Do not remove the
614 * Basically run through each address specified in the addrs/addrcnt
615 * array/length pair, determine if it is IPv6 or IPv4 and call
616 * sctp_del_bind() on it.
618 * If any of them fails, then the operation will be reversed and the
619 * ones that were removed will be added back.
621 * At least one address has to be left; if only one address is
622 * available, the operation will return -EBUSY.
624 * Only sctp_setsockopt_bindx() is supposed to call this function.
626 static int sctp_bindx_rem(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
628 struct sctp_sock
*sp
= sctp_sk(sk
);
629 struct sctp_endpoint
*ep
= sp
->ep
;
631 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
634 union sctp_addr
*sa_addr
;
637 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
638 __func__
, sk
, addrs
, addrcnt
);
641 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
642 /* If the bind address list is empty or if there is only one
643 * bind address, there is nothing more to be removed (we need
644 * at least one address here).
646 if (list_empty(&bp
->address_list
) ||
647 (sctp_list_single_entry(&bp
->address_list
))) {
653 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
659 if (!af
->addr_valid(sa_addr
, sp
, NULL
)) {
660 retval
= -EADDRNOTAVAIL
;
664 if (sa_addr
->v4
.sin_port
&&
665 sa_addr
->v4
.sin_port
!= htons(bp
->port
)) {
670 if (!sa_addr
->v4
.sin_port
)
671 sa_addr
->v4
.sin_port
= htons(bp
->port
);
673 /* FIXME - There is probably a need to check if sk->sk_saddr and
674 * sk->sk_rcv_addr are currently set to one of the addresses to
675 * be removed. This is something which needs to be looked into
676 * when we are fixing the outstanding issues with multi-homing
677 * socket routing and failover schemes. Refer to comments in
678 * sctp_do_bind(). -daisy
680 retval
= sctp_del_bind_addr(bp
, sa_addr
);
682 addr_buf
+= af
->sockaddr_len
;
685 /* Failed. Add the ones that has been removed back */
687 sctp_bindx_add(sk
, addrs
, cnt
);
695 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
696 * the associations that are part of the endpoint indicating that a list of
697 * local addresses are removed from the endpoint.
699 * If any of the addresses is already in the bind address list of the
700 * association, we do not send the chunk for that association. But it will not
701 * affect other associations.
703 * Only sctp_setsockopt_bindx() is supposed to call this function.
705 static int sctp_send_asconf_del_ip(struct sock
*sk
,
706 struct sockaddr
*addrs
,
709 struct net
*net
= sock_net(sk
);
710 struct sctp_sock
*sp
;
711 struct sctp_endpoint
*ep
;
712 struct sctp_association
*asoc
;
713 struct sctp_transport
*transport
;
714 struct sctp_bind_addr
*bp
;
715 struct sctp_chunk
*chunk
;
716 union sctp_addr
*laddr
;
719 struct sctp_sockaddr_entry
*saddr
;
725 if (!net
->sctp
.addip_enable
)
731 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
732 __func__
, sk
, addrs
, addrcnt
);
734 list_for_each_entry(asoc
, &ep
->asocs
, asocs
) {
736 if (!asoc
->peer
.asconf_capable
)
739 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_DEL_IP
)
742 if (!sctp_state(asoc
, ESTABLISHED
))
745 /* Check if any address in the packed array of addresses is
746 * not present in the bind address list of the association.
747 * If so, do not send the asconf chunk to its peer, but
748 * continue with other associations.
751 for (i
= 0; i
< addrcnt
; i
++) {
753 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
759 if (!sctp_assoc_lookup_laddr(asoc
, laddr
))
762 addr_buf
+= af
->sockaddr_len
;
767 /* Find one address in the association's bind address list
768 * that is not in the packed array of addresses. This is to
769 * make sure that we do not delete all the addresses in the
772 bp
= &asoc
->base
.bind_addr
;
773 laddr
= sctp_find_unmatch_addr(bp
, (union sctp_addr
*)addrs
,
775 if ((laddr
== NULL
) && (addrcnt
== 1)) {
776 if (asoc
->asconf_addr_del_pending
)
778 asoc
->asconf_addr_del_pending
=
779 kzalloc(sizeof(union sctp_addr
), GFP_ATOMIC
);
780 if (asoc
->asconf_addr_del_pending
== NULL
) {
784 asoc
->asconf_addr_del_pending
->sa
.sa_family
=
786 asoc
->asconf_addr_del_pending
->v4
.sin_port
=
788 if (addrs
->sa_family
== AF_INET
) {
789 struct sockaddr_in
*sin
;
791 sin
= (struct sockaddr_in
*)addrs
;
792 asoc
->asconf_addr_del_pending
->v4
.sin_addr
.s_addr
= sin
->sin_addr
.s_addr
;
793 } else if (addrs
->sa_family
== AF_INET6
) {
794 struct sockaddr_in6
*sin6
;
796 sin6
= (struct sockaddr_in6
*)addrs
;
797 asoc
->asconf_addr_del_pending
->v6
.sin6_addr
= sin6
->sin6_addr
;
800 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
801 __func__
, asoc
, &asoc
->asconf_addr_del_pending
->sa
,
802 asoc
->asconf_addr_del_pending
);
804 asoc
->src_out_of_asoc_ok
= 1;
812 /* We do not need RCU protection throughout this loop
813 * because this is done under a socket lock from the
816 chunk
= sctp_make_asconf_update_ip(asoc
, laddr
, addrs
, addrcnt
,
824 /* Reset use_as_src flag for the addresses in the bind address
825 * list that are to be deleted.
828 for (i
= 0; i
< addrcnt
; i
++) {
830 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
831 list_for_each_entry(saddr
, &bp
->address_list
, list
) {
832 if (sctp_cmp_addr_exact(&saddr
->a
, laddr
))
833 saddr
->state
= SCTP_ADDR_DEL
;
835 addr_buf
+= af
->sockaddr_len
;
838 /* Update the route and saddr entries for all the transports
839 * as some of the addresses in the bind address list are
840 * about to be deleted and cannot be used as source addresses.
842 list_for_each_entry(transport
, &asoc
->peer
.transport_addr_list
,
844 dst_release(transport
->dst
);
845 sctp_transport_route(transport
, NULL
,
846 sctp_sk(asoc
->base
.sk
));
850 /* We don't need to transmit ASCONF */
852 retval
= sctp_send_asconf(asoc
, chunk
);
858 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
859 int sctp_asconf_mgmt(struct sctp_sock
*sp
, struct sctp_sockaddr_entry
*addrw
)
861 struct sock
*sk
= sctp_opt2sk(sp
);
862 union sctp_addr
*addr
;
865 /* It is safe to write port space in caller. */
867 addr
->v4
.sin_port
= htons(sp
->ep
->base
.bind_addr
.port
);
868 af
= sctp_get_af_specific(addr
->sa
.sa_family
);
871 if (sctp_verify_addr(sk
, addr
, af
->sockaddr_len
))
874 if (addrw
->state
== SCTP_ADDR_NEW
)
875 return sctp_send_asconf_add_ip(sk
, (struct sockaddr
*)addr
, 1);
877 return sctp_send_asconf_del_ip(sk
, (struct sockaddr
*)addr
, 1);
880 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
883 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
886 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
887 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
890 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
891 * Section 3.1.2 for this usage.
893 * addrs is a pointer to an array of one or more socket addresses. Each
894 * address is contained in its appropriate structure (i.e. struct
895 * sockaddr_in or struct sockaddr_in6) the family of the address type
896 * must be used to distinguish the address length (note that this
897 * representation is termed a "packed array" of addresses). The caller
898 * specifies the number of addresses in the array with addrcnt.
900 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
901 * -1, and sets errno to the appropriate error code.
903 * For SCTP, the port given in each socket address must be the same, or
904 * sctp_bindx() will fail, setting errno to EINVAL.
906 * The flags parameter is formed from the bitwise OR of zero or more of
907 * the following currently defined flags:
909 * SCTP_BINDX_ADD_ADDR
911 * SCTP_BINDX_REM_ADDR
913 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
914 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
915 * addresses from the association. The two flags are mutually exclusive;
916 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
917 * not remove all addresses from an association; sctp_bindx() will
918 * reject such an attempt with EINVAL.
920 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
921 * additional addresses with an endpoint after calling bind(). Or use
922 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
923 * socket is associated with so that no new association accepted will be
924 * associated with those addresses. If the endpoint supports dynamic
925 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
926 * endpoint to send the appropriate message to the peer to change the
927 * peers address lists.
929 * Adding and removing addresses from a connected association is
930 * optional functionality. Implementations that do not support this
931 * functionality should return EOPNOTSUPP.
933 * Basically do nothing but copying the addresses from user to kernel
934 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
935 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
938 * We don't use copy_from_user() for optimization: we first do the
939 * sanity checks (buffer size -fast- and access check-healthy
940 * pointer); if all of those succeed, then we can alloc the memory
941 * (expensive operation) needed to copy the data to kernel. Then we do
942 * the copying without checking the user space area
943 * (__copy_from_user()).
945 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
948 * sk The sk of the socket
949 * addrs The pointer to the addresses in user land
950 * addrssize Size of the addrs buffer
951 * op Operation to perform (add or remove, see the flags of
954 * Returns 0 if ok, <0 errno code on error.
956 static int sctp_setsockopt_bindx(struct sock
*sk
,
957 struct sockaddr __user
*addrs
,
958 int addrs_size
, int op
)
960 struct sockaddr
*kaddrs
;
964 struct sockaddr
*sa_addr
;
968 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
969 __func__
, sk
, addrs
, addrs_size
, op
);
971 if (unlikely(addrs_size
<= 0))
974 /* Check the user passed a healthy pointer. */
975 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
978 /* Alloc space for the address array in kernel memory. */
979 kaddrs
= kmalloc(addrs_size
, GFP_KERNEL
);
980 if (unlikely(!kaddrs
))
983 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
988 /* Walk through the addrs buffer and count the number of addresses. */
990 while (walk_size
< addrs_size
) {
991 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
997 af
= sctp_get_af_specific(sa_addr
->sa_family
);
999 /* If the address family is not supported or if this address
1000 * causes the address buffer to overflow return EINVAL.
1002 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
1007 addr_buf
+= af
->sockaddr_len
;
1008 walk_size
+= af
->sockaddr_len
;
1013 case SCTP_BINDX_ADD_ADDR
:
1014 err
= sctp_bindx_add(sk
, kaddrs
, addrcnt
);
1017 err
= sctp_send_asconf_add_ip(sk
, kaddrs
, addrcnt
);
1020 case SCTP_BINDX_REM_ADDR
:
1021 err
= sctp_bindx_rem(sk
, kaddrs
, addrcnt
);
1024 err
= sctp_send_asconf_del_ip(sk
, kaddrs
, addrcnt
);
1038 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1040 * Common routine for handling connect() and sctp_connectx().
1041 * Connect will come in with just a single address.
1043 static int __sctp_connect(struct sock
*sk
,
1044 struct sockaddr
*kaddrs
,
1046 sctp_assoc_t
*assoc_id
)
1048 struct net
*net
= sock_net(sk
);
1049 struct sctp_sock
*sp
;
1050 struct sctp_endpoint
*ep
;
1051 struct sctp_association
*asoc
= NULL
;
1052 struct sctp_association
*asoc2
;
1053 struct sctp_transport
*transport
;
1061 union sctp_addr
*sa_addr
= NULL
;
1063 unsigned short port
;
1064 unsigned int f_flags
= 0;
1069 /* connect() cannot be done on a socket that is already in ESTABLISHED
1070 * state - UDP-style peeled off socket or a TCP-style socket that
1071 * is already connected.
1072 * It cannot be done even on a TCP-style listening socket.
1074 if (sctp_sstate(sk
, ESTABLISHED
) ||
1075 (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))) {
1080 /* Walk through the addrs buffer and count the number of addresses. */
1082 while (walk_size
< addrs_size
) {
1083 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
1089 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
1091 /* If the address family is not supported or if this address
1092 * causes the address buffer to overflow return EINVAL.
1094 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
1099 port
= ntohs(sa_addr
->v4
.sin_port
);
1101 /* Save current address so we can work with it */
1102 memcpy(&to
, sa_addr
, af
->sockaddr_len
);
1104 err
= sctp_verify_addr(sk
, &to
, af
->sockaddr_len
);
1108 /* Make sure the destination port is correctly set
1111 if (asoc
&& asoc
->peer
.port
&& asoc
->peer
.port
!= port
) {
1116 /* Check if there already is a matching association on the
1117 * endpoint (other than the one created here).
1119 asoc2
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1120 if (asoc2
&& asoc2
!= asoc
) {
1121 if (asoc2
->state
>= SCTP_STATE_ESTABLISHED
)
1128 /* If we could not find a matching association on the endpoint,
1129 * make sure that there is no peeled-off association matching
1130 * the peer address even on another socket.
1132 if (sctp_endpoint_is_peeled_off(ep
, &to
)) {
1133 err
= -EADDRNOTAVAIL
;
1138 /* If a bind() or sctp_bindx() is not called prior to
1139 * an sctp_connectx() call, the system picks an
1140 * ephemeral port and will choose an address set
1141 * equivalent to binding with a wildcard address.
1143 if (!ep
->base
.bind_addr
.port
) {
1144 if (sctp_autobind(sk
)) {
1150 * If an unprivileged user inherits a 1-many
1151 * style socket with open associations on a
1152 * privileged port, it MAY be permitted to
1153 * accept new associations, but it SHOULD NOT
1154 * be permitted to open new associations.
1156 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1157 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
)) {
1163 scope
= sctp_scope(&to
);
1164 asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1170 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
,
1178 /* Prime the peer's transport structures. */
1179 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
,
1187 addr_buf
+= af
->sockaddr_len
;
1188 walk_size
+= af
->sockaddr_len
;
1191 /* In case the user of sctp_connectx() wants an association
1192 * id back, assign one now.
1195 err
= sctp_assoc_set_id(asoc
, GFP_KERNEL
);
1200 err
= sctp_primitive_ASSOCIATE(net
, asoc
, NULL
);
1205 /* Initialize sk's dport and daddr for getpeername() */
1206 inet_sk(sk
)->inet_dport
= htons(asoc
->peer
.port
);
1207 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
1208 af
->to_sk_daddr(sa_addr
, sk
);
1211 /* in-kernel sockets don't generally have a file allocated to them
1212 * if all they do is call sock_create_kern().
1214 if (sk
->sk_socket
->file
)
1215 f_flags
= sk
->sk_socket
->file
->f_flags
;
1217 timeo
= sock_sndtimeo(sk
, f_flags
& O_NONBLOCK
);
1219 err
= sctp_wait_for_connect(asoc
, &timeo
);
1220 if ((err
== 0 || err
== -EINPROGRESS
) && assoc_id
)
1221 *assoc_id
= asoc
->assoc_id
;
1223 /* Don't free association on exit. */
1227 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1228 __func__
, asoc
, kaddrs
, err
);
1231 /* sctp_primitive_ASSOCIATE may have added this association
1232 * To the hash table, try to unhash it, just in case, its a noop
1233 * if it wasn't hashed so we're safe
1235 sctp_unhash_established(asoc
);
1236 sctp_association_free(asoc
);
1241 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1244 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1245 * sctp_assoc_t *asoc);
1247 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1248 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1249 * or IPv6 addresses.
1251 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1252 * Section 3.1.2 for this usage.
1254 * addrs is a pointer to an array of one or more socket addresses. Each
1255 * address is contained in its appropriate structure (i.e. struct
1256 * sockaddr_in or struct sockaddr_in6) the family of the address type
1257 * must be used to distengish the address length (note that this
1258 * representation is termed a "packed array" of addresses). The caller
1259 * specifies the number of addresses in the array with addrcnt.
1261 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1262 * the association id of the new association. On failure, sctp_connectx()
1263 * returns -1, and sets errno to the appropriate error code. The assoc_id
1264 * is not touched by the kernel.
1266 * For SCTP, the port given in each socket address must be the same, or
1267 * sctp_connectx() will fail, setting errno to EINVAL.
1269 * An application can use sctp_connectx to initiate an association with
1270 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1271 * allows a caller to specify multiple addresses at which a peer can be
1272 * reached. The way the SCTP stack uses the list of addresses to set up
1273 * the association is implementation dependent. This function only
1274 * specifies that the stack will try to make use of all the addresses in
1275 * the list when needed.
1277 * Note that the list of addresses passed in is only used for setting up
1278 * the association. It does not necessarily equal the set of addresses
1279 * the peer uses for the resulting association. If the caller wants to
1280 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1281 * retrieve them after the association has been set up.
1283 * Basically do nothing but copying the addresses from user to kernel
1284 * land and invoking either sctp_connectx(). This is used for tunneling
1285 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1287 * We don't use copy_from_user() for optimization: we first do the
1288 * sanity checks (buffer size -fast- and access check-healthy
1289 * pointer); if all of those succeed, then we can alloc the memory
1290 * (expensive operation) needed to copy the data to kernel. Then we do
1291 * the copying without checking the user space area
1292 * (__copy_from_user()).
1294 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1297 * sk The sk of the socket
1298 * addrs The pointer to the addresses in user land
1299 * addrssize Size of the addrs buffer
1301 * Returns >=0 if ok, <0 errno code on error.
1303 static int __sctp_setsockopt_connectx(struct sock
*sk
,
1304 struct sockaddr __user
*addrs
,
1306 sctp_assoc_t
*assoc_id
)
1309 struct sockaddr
*kaddrs
;
1311 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1312 __func__
, sk
, addrs
, addrs_size
);
1314 if (unlikely(addrs_size
<= 0))
1317 /* Check the user passed a healthy pointer. */
1318 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
1321 /* Alloc space for the address array in kernel memory. */
1322 kaddrs
= kmalloc(addrs_size
, GFP_KERNEL
);
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 (is_compat_task()) {
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_unhash_established(asoc
);
1509 sctp_association_free(asoc
);
1514 if (data_was_unread
|| !skb_queue_empty(&asoc
->ulpq
.lobby
) ||
1515 !skb_queue_empty(&asoc
->ulpq
.reasm
) ||
1516 (sock_flag(sk
, SOCK_LINGER
) && !sk
->sk_lingertime
)) {
1517 struct sctp_chunk
*chunk
;
1519 chunk
= sctp_make_abort_user(asoc
, NULL
, 0);
1521 sctp_primitive_ABORT(net
, asoc
, chunk
);
1523 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
1526 /* On a TCP-style socket, block for at most linger_time if set. */
1527 if (sctp_style(sk
, TCP
) && timeout
)
1528 sctp_wait_for_close(sk
, timeout
);
1530 /* This will run the backlog queue. */
1533 /* Supposedly, no process has access to the socket, but
1534 * the net layers still may.
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
);
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 kiocb
*iocb
, struct sock
*sk
,
1589 struct msghdr
*msg
, size_t msg_len
)
1591 struct net
*net
= sock_net(sk
);
1592 struct sctp_sock
*sp
;
1593 struct sctp_endpoint
*ep
;
1594 struct sctp_association
*new_asoc
= NULL
, *asoc
= NULL
;
1595 struct sctp_transport
*transport
, *chunk_tp
;
1596 struct sctp_chunk
*chunk
;
1598 struct sockaddr
*msg_name
= NULL
;
1599 struct sctp_sndrcvinfo default_sinfo
;
1600 struct sctp_sndrcvinfo
*sinfo
;
1601 struct sctp_initmsg
*sinit
;
1602 sctp_assoc_t associd
= 0;
1603 sctp_cmsgs_t cmsgs
= { NULL
};
1607 __u16 sinfo_flags
= 0;
1608 struct sctp_datamsg
*datamsg
;
1609 int msg_flags
= msg
->msg_flags
;
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
;
1653 /* Did the user specify SNDRCVINFO? */
1655 sinfo_flags
= sinfo
->sinfo_flags
;
1656 associd
= sinfo
->sinfo_assoc_id
;
1659 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__
,
1660 msg_len
, sinfo_flags
);
1662 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1663 if (sctp_style(sk
, TCP
) && (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
))) {
1668 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1669 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1670 * If SCTP_ABORT is set, the message length could be non zero with
1671 * the msg_iov set to the user abort reason.
1673 if (((sinfo_flags
& SCTP_EOF
) && (msg_len
> 0)) ||
1674 (!(sinfo_flags
& (SCTP_EOF
|SCTP_ABORT
)) && (msg_len
== 0))) {
1679 /* If SCTP_ADDR_OVER is set, there must be an address
1680 * specified in msg_name.
1682 if ((sinfo_flags
& SCTP_ADDR_OVER
) && (!msg
->msg_name
)) {
1689 pr_debug("%s: about to look up association\n", __func__
);
1693 /* If a msg_name has been specified, assume this is to be used. */
1695 /* Look for a matching association on the endpoint. */
1696 asoc
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1698 /* If we could not find a matching association on the
1699 * endpoint, make sure that it is not a TCP-style
1700 * socket that already has an association or there is
1701 * no peeled-off association on another socket.
1703 if ((sctp_style(sk
, TCP
) &&
1704 sctp_sstate(sk
, ESTABLISHED
)) ||
1705 sctp_endpoint_is_peeled_off(ep
, &to
)) {
1706 err
= -EADDRNOTAVAIL
;
1711 asoc
= sctp_id2assoc(sk
, associd
);
1719 pr_debug("%s: just looked up association:%p\n", __func__
, asoc
);
1721 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1722 * socket that has an association in CLOSED state. This can
1723 * happen when an accepted socket has an association that is
1726 if (sctp_state(asoc
, CLOSED
) && sctp_style(sk
, TCP
)) {
1731 if (sinfo_flags
& SCTP_EOF
) {
1732 pr_debug("%s: shutting down association:%p\n",
1735 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
1739 if (sinfo_flags
& SCTP_ABORT
) {
1741 chunk
= sctp_make_abort_user(asoc
, msg
, msg_len
);
1747 pr_debug("%s: aborting association:%p\n",
1750 sctp_primitive_ABORT(net
, asoc
, chunk
);
1756 /* Do we need to create the association? */
1758 pr_debug("%s: there is no association yet\n", __func__
);
1760 if (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
)) {
1765 /* Check for invalid stream against the stream counts,
1766 * either the default or the user specified stream counts.
1769 if (!sinit
|| !sinit
->sinit_num_ostreams
) {
1770 /* Check against the defaults. */
1771 if (sinfo
->sinfo_stream
>=
1772 sp
->initmsg
.sinit_num_ostreams
) {
1777 /* Check against the requested. */
1778 if (sinfo
->sinfo_stream
>=
1779 sinit
->sinit_num_ostreams
) {
1787 * API 3.1.2 bind() - UDP Style Syntax
1788 * If a bind() or sctp_bindx() is not called prior to a
1789 * sendmsg() call that initiates a new association, the
1790 * system picks an ephemeral port and will choose an address
1791 * set equivalent to binding with a wildcard address.
1793 if (!ep
->base
.bind_addr
.port
) {
1794 if (sctp_autobind(sk
)) {
1800 * If an unprivileged user inherits a one-to-many
1801 * style socket with open associations on a privileged
1802 * port, it MAY be permitted to accept new associations,
1803 * but it SHOULD NOT be permitted to open new
1806 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1807 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
)) {
1813 scope
= sctp_scope(&to
);
1814 new_asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1820 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
, GFP_KERNEL
);
1826 /* If the SCTP_INIT ancillary data is specified, set all
1827 * the association init values accordingly.
1830 if (sinit
->sinit_num_ostreams
) {
1831 asoc
->c
.sinit_num_ostreams
=
1832 sinit
->sinit_num_ostreams
;
1834 if (sinit
->sinit_max_instreams
) {
1835 asoc
->c
.sinit_max_instreams
=
1836 sinit
->sinit_max_instreams
;
1838 if (sinit
->sinit_max_attempts
) {
1839 asoc
->max_init_attempts
1840 = sinit
->sinit_max_attempts
;
1842 if (sinit
->sinit_max_init_timeo
) {
1843 asoc
->max_init_timeo
=
1844 msecs_to_jiffies(sinit
->sinit_max_init_timeo
);
1848 /* Prime the peer's transport structures. */
1849 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
, SCTP_UNKNOWN
);
1856 /* ASSERT: we have a valid association at this point. */
1857 pr_debug("%s: we have a valid association\n", __func__
);
1860 /* If the user didn't specify SNDRCVINFO, make up one with
1863 memset(&default_sinfo
, 0, sizeof(default_sinfo
));
1864 default_sinfo
.sinfo_stream
= asoc
->default_stream
;
1865 default_sinfo
.sinfo_flags
= asoc
->default_flags
;
1866 default_sinfo
.sinfo_ppid
= asoc
->default_ppid
;
1867 default_sinfo
.sinfo_context
= asoc
->default_context
;
1868 default_sinfo
.sinfo_timetolive
= asoc
->default_timetolive
;
1869 default_sinfo
.sinfo_assoc_id
= sctp_assoc2id(asoc
);
1870 sinfo
= &default_sinfo
;
1873 /* API 7.1.7, the sndbuf size per association bounds the
1874 * maximum size of data that can be sent in a single send call.
1876 if (msg_len
> sk
->sk_sndbuf
) {
1881 if (asoc
->pmtu_pending
)
1882 sctp_assoc_pending_pmtu(sk
, asoc
);
1884 /* If fragmentation is disabled and the message length exceeds the
1885 * association fragmentation point, return EMSGSIZE. The I-D
1886 * does not specify what this error is, but this looks like
1889 if (sctp_sk(sk
)->disable_fragments
&& (msg_len
> asoc
->frag_point
)) {
1894 /* Check for invalid stream. */
1895 if (sinfo
->sinfo_stream
>= asoc
->c
.sinit_num_ostreams
) {
1900 timeo
= sock_sndtimeo(sk
, msg
->msg_flags
& MSG_DONTWAIT
);
1901 if (!sctp_wspace(asoc
)) {
1902 err
= sctp_wait_for_sndbuf(asoc
, &timeo
, msg_len
);
1907 /* If an address is passed with the sendto/sendmsg call, it is used
1908 * to override the primary destination address in the TCP model, or
1909 * when SCTP_ADDR_OVER flag is set in the UDP model.
1911 if ((sctp_style(sk
, TCP
) && msg_name
) ||
1912 (sinfo_flags
& SCTP_ADDR_OVER
)) {
1913 chunk_tp
= sctp_assoc_lookup_paddr(asoc
, &to
);
1921 /* Auto-connect, if we aren't connected already. */
1922 if (sctp_state(asoc
, CLOSED
)) {
1923 err
= sctp_primitive_ASSOCIATE(net
, asoc
, NULL
);
1927 pr_debug("%s: we associated primitively\n", __func__
);
1930 /* Break the message into multiple chunks of maximum size. */
1931 datamsg
= sctp_datamsg_from_user(asoc
, sinfo
, msg
, msg_len
);
1932 if (IS_ERR(datamsg
)) {
1933 err
= PTR_ERR(datamsg
);
1937 /* Now send the (possibly) fragmented message. */
1938 list_for_each_entry(chunk
, &datamsg
->chunks
, frag_list
) {
1939 sctp_chunk_hold(chunk
);
1941 /* Do accounting for the write space. */
1942 sctp_set_owner_w(chunk
);
1944 chunk
->transport
= chunk_tp
;
1947 /* Send it to the lower layers. Note: all chunks
1948 * must either fail or succeed. The lower layer
1949 * works that way today. Keep it that way or this
1952 err
= sctp_primitive_SEND(net
, asoc
, datamsg
);
1953 /* Did the lower layer accept the chunk? */
1955 sctp_datamsg_free(datamsg
);
1959 pr_debug("%s: we sent primitively\n", __func__
);
1961 sctp_datamsg_put(datamsg
);
1964 /* If we are already past ASSOCIATE, the lower
1965 * layers are responsible for association cleanup.
1971 sctp_unhash_established(asoc
);
1972 sctp_association_free(asoc
);
1978 return sctp_error(sk
, msg_flags
, err
);
1985 err
= sock_error(sk
);
1995 /* This is an extended version of skb_pull() that removes the data from the
1996 * start of a skb even when data is spread across the list of skb's in the
1997 * frag_list. len specifies the total amount of data that needs to be removed.
1998 * when 'len' bytes could be removed from the skb, it returns 0.
1999 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2000 * could not be removed.
2002 static int sctp_skb_pull(struct sk_buff
*skb
, int len
)
2004 struct sk_buff
*list
;
2005 int skb_len
= skb_headlen(skb
);
2008 if (len
<= skb_len
) {
2009 __skb_pull(skb
, len
);
2013 __skb_pull(skb
, skb_len
);
2015 skb_walk_frags(skb
, list
) {
2016 rlen
= sctp_skb_pull(list
, len
);
2017 skb
->len
-= (len
-rlen
);
2018 skb
->data_len
-= (len
-rlen
);
2029 /* API 3.1.3 recvmsg() - UDP Style Syntax
2031 * ssize_t recvmsg(int socket, struct msghdr *message,
2034 * socket - the socket descriptor of the endpoint.
2035 * message - pointer to the msghdr structure which contains a single
2036 * user message and possibly some ancillary data.
2038 * See Section 5 for complete description of the data
2041 * flags - flags sent or received with the user message, see Section
2042 * 5 for complete description of the flags.
2044 static struct sk_buff
*sctp_skb_recv_datagram(struct sock
*, int, int, int *);
2046 static int sctp_recvmsg(struct kiocb
*iocb
, struct sock
*sk
,
2047 struct msghdr
*msg
, size_t len
, int noblock
,
2048 int flags
, int *addr_len
)
2050 struct sctp_ulpevent
*event
= NULL
;
2051 struct sctp_sock
*sp
= sctp_sk(sk
);
2052 struct sk_buff
*skb
;
2057 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2058 "addr_len:%p)\n", __func__
, sk
, msg
, len
, noblock
, flags
,
2063 if (sctp_style(sk
, TCP
) && !sctp_sstate(sk
, ESTABLISHED
)) {
2068 skb
= sctp_skb_recv_datagram(sk
, flags
, noblock
, &err
);
2072 /* Get the total length of the skb including any skb's in the
2081 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
2083 event
= sctp_skb2event(skb
);
2088 sock_recv_ts_and_drops(msg
, sk
, skb
);
2089 if (sctp_ulpevent_is_notification(event
)) {
2090 msg
->msg_flags
|= MSG_NOTIFICATION
;
2091 sp
->pf
->event_msgname(event
, msg
->msg_name
, addr_len
);
2093 sp
->pf
->skb_msgname(skb
, msg
->msg_name
, addr_len
);
2096 /* Check if we allow SCTP_SNDRCVINFO. */
2097 if (sp
->subscribe
.sctp_data_io_event
)
2098 sctp_ulpevent_read_sndrcvinfo(event
, msg
);
2100 /* FIXME: we should be calling IP/IPv6 layers. */
2101 if (sk
->sk_protinfo
.af_inet
.cmsg_flags
)
2102 ip_cmsg_recv(msg
, skb
);
2107 /* If skb's length exceeds the user's buffer, update the skb and
2108 * push it back to the receive_queue so that the next call to
2109 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2111 if (skb_len
> copied
) {
2112 msg
->msg_flags
&= ~MSG_EOR
;
2113 if (flags
& MSG_PEEK
)
2115 sctp_skb_pull(skb
, copied
);
2116 skb_queue_head(&sk
->sk_receive_queue
, skb
);
2118 /* When only partial message is copied to the user, increase
2119 * rwnd by that amount. If all the data in the skb is read,
2120 * rwnd is updated when the event is freed.
2122 if (!sctp_ulpevent_is_notification(event
))
2123 sctp_assoc_rwnd_increase(event
->asoc
, copied
);
2125 } else if ((event
->msg_flags
& MSG_NOTIFICATION
) ||
2126 (event
->msg_flags
& MSG_EOR
))
2127 msg
->msg_flags
|= MSG_EOR
;
2129 msg
->msg_flags
&= ~MSG_EOR
;
2132 if (flags
& MSG_PEEK
) {
2133 /* Release the skb reference acquired after peeking the skb in
2134 * sctp_skb_recv_datagram().
2138 /* Free the event which includes releasing the reference to
2139 * the owner of the skb, freeing the skb and updating the
2142 sctp_ulpevent_free(event
);
2149 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2151 * This option is a on/off flag. If enabled no SCTP message
2152 * fragmentation will be performed. Instead if a message being sent
2153 * exceeds the current PMTU size, the message will NOT be sent and
2154 * instead a error will be indicated to the user.
2156 static int sctp_setsockopt_disable_fragments(struct sock
*sk
,
2157 char __user
*optval
,
2158 unsigned int optlen
)
2162 if (optlen
< sizeof(int))
2165 if (get_user(val
, (int __user
*)optval
))
2168 sctp_sk(sk
)->disable_fragments
= (val
== 0) ? 0 : 1;
2173 static int sctp_setsockopt_events(struct sock
*sk
, char __user
*optval
,
2174 unsigned int optlen
)
2176 struct sctp_association
*asoc
;
2177 struct sctp_ulpevent
*event
;
2179 if (optlen
> sizeof(struct sctp_event_subscribe
))
2181 if (copy_from_user(&sctp_sk(sk
)->subscribe
, optval
, optlen
))
2185 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2186 * if there is no data to be sent or retransmit, the stack will
2187 * immediately send up this notification.
2189 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT
,
2190 &sctp_sk(sk
)->subscribe
)) {
2191 asoc
= sctp_id2assoc(sk
, 0);
2193 if (asoc
&& sctp_outq_is_empty(&asoc
->outqueue
)) {
2194 event
= sctp_ulpevent_make_sender_dry_event(asoc
,
2199 sctp_ulpq_tail_event(&asoc
->ulpq
, event
);
2206 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2208 * This socket option is applicable to the UDP-style socket only. When
2209 * set it will cause associations that are idle for more than the
2210 * specified number of seconds to automatically close. An association
2211 * being idle is defined an association that has NOT sent or received
2212 * user data. The special value of '0' indicates that no automatic
2213 * close of any associations should be performed. The option expects an
2214 * integer defining the number of seconds of idle time before an
2215 * association is closed.
2217 static int sctp_setsockopt_autoclose(struct sock
*sk
, char __user
*optval
,
2218 unsigned int optlen
)
2220 struct sctp_sock
*sp
= sctp_sk(sk
);
2221 struct net
*net
= sock_net(sk
);
2223 /* Applicable to UDP-style socket only */
2224 if (sctp_style(sk
, TCP
))
2226 if (optlen
!= sizeof(int))
2228 if (copy_from_user(&sp
->autoclose
, optval
, optlen
))
2231 if (sp
->autoclose
> net
->sctp
.max_autoclose
)
2232 sp
->autoclose
= net
->sctp
.max_autoclose
;
2237 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2239 * Applications can enable or disable heartbeats for any peer address of
2240 * an association, modify an address's heartbeat interval, force a
2241 * heartbeat to be sent immediately, and adjust the address's maximum
2242 * number of retransmissions sent before an address is considered
2243 * unreachable. The following structure is used to access and modify an
2244 * address's parameters:
2246 * struct sctp_paddrparams {
2247 * sctp_assoc_t spp_assoc_id;
2248 * struct sockaddr_storage spp_address;
2249 * uint32_t spp_hbinterval;
2250 * uint16_t spp_pathmaxrxt;
2251 * uint32_t spp_pathmtu;
2252 * uint32_t spp_sackdelay;
2253 * uint32_t spp_flags;
2256 * spp_assoc_id - (one-to-many style socket) This is filled in the
2257 * application, and identifies the association for
2259 * spp_address - This specifies which address is of interest.
2260 * spp_hbinterval - This contains the value of the heartbeat interval,
2261 * in milliseconds. If a value of zero
2262 * is present in this field then no changes are to
2263 * be made to this parameter.
2264 * spp_pathmaxrxt - This contains the maximum number of
2265 * retransmissions before this address shall be
2266 * considered unreachable. If a value of zero
2267 * is present in this field then no changes are to
2268 * be made to this parameter.
2269 * spp_pathmtu - When Path MTU discovery is disabled the value
2270 * specified here will be the "fixed" path mtu.
2271 * Note that if the spp_address field is empty
2272 * then all associations on this address will
2273 * have this fixed path mtu set upon them.
2275 * spp_sackdelay - When delayed sack is enabled, this value specifies
2276 * the number of milliseconds that sacks will be delayed
2277 * for. This value will apply to all addresses of an
2278 * association if the spp_address field is empty. Note
2279 * also, that if delayed sack is enabled and this
2280 * value is set to 0, no change is made to the last
2281 * recorded delayed sack timer value.
2283 * spp_flags - These flags are used to control various features
2284 * on an association. The flag field may contain
2285 * zero or more of the following options.
2287 * SPP_HB_ENABLE - Enable heartbeats on the
2288 * specified address. Note that if the address
2289 * field is empty all addresses for the association
2290 * have heartbeats enabled upon them.
2292 * SPP_HB_DISABLE - Disable heartbeats on the
2293 * speicifed address. Note that if the address
2294 * field is empty all addresses for the association
2295 * will have their heartbeats disabled. Note also
2296 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2297 * mutually exclusive, only one of these two should
2298 * be specified. Enabling both fields will have
2299 * undetermined results.
2301 * SPP_HB_DEMAND - Request a user initiated heartbeat
2302 * to be made immediately.
2304 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2305 * heartbeat delayis to be set to the value of 0
2308 * SPP_PMTUD_ENABLE - This field will enable PMTU
2309 * discovery upon the specified address. Note that
2310 * if the address feild is empty then all addresses
2311 * on the association are effected.
2313 * SPP_PMTUD_DISABLE - This field will disable PMTU
2314 * discovery upon the specified address. Note that
2315 * if the address feild is empty then all addresses
2316 * on the association are effected. Not also that
2317 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2318 * exclusive. Enabling both will have undetermined
2321 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2322 * on delayed sack. The time specified in spp_sackdelay
2323 * is used to specify the sack delay for this address. Note
2324 * that if spp_address is empty then all addresses will
2325 * enable delayed sack and take on the sack delay
2326 * value specified in spp_sackdelay.
2327 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2328 * off delayed sack. If the spp_address field is blank then
2329 * delayed sack is disabled for the entire association. Note
2330 * also that this field is mutually exclusive to
2331 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2334 static int sctp_apply_peer_addr_params(struct sctp_paddrparams
*params
,
2335 struct sctp_transport
*trans
,
2336 struct sctp_association
*asoc
,
2337 struct sctp_sock
*sp
,
2340 int sackdelay_change
)
2344 if (params
->spp_flags
& SPP_HB_DEMAND
&& trans
) {
2345 struct net
*net
= sock_net(trans
->asoc
->base
.sk
);
2347 error
= sctp_primitive_REQUESTHEARTBEAT(net
, trans
->asoc
, trans
);
2352 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2353 * this field is ignored. Note also that a value of zero indicates
2354 * the current setting should be left unchanged.
2356 if (params
->spp_flags
& SPP_HB_ENABLE
) {
2358 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2359 * set. This lets us use 0 value when this flag
2362 if (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)
2363 params
->spp_hbinterval
= 0;
2365 if (params
->spp_hbinterval
||
2366 (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)) {
2369 msecs_to_jiffies(params
->spp_hbinterval
);
2372 msecs_to_jiffies(params
->spp_hbinterval
);
2374 sp
->hbinterval
= params
->spp_hbinterval
;
2381 trans
->param_flags
=
2382 (trans
->param_flags
& ~SPP_HB
) | hb_change
;
2385 (asoc
->param_flags
& ~SPP_HB
) | hb_change
;
2388 (sp
->param_flags
& ~SPP_HB
) | hb_change
;
2392 /* When Path MTU discovery is disabled the value specified here will
2393 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2394 * include the flag SPP_PMTUD_DISABLE for this field to have any
2397 if ((params
->spp_flags
& SPP_PMTUD_DISABLE
) && params
->spp_pathmtu
) {
2399 trans
->pathmtu
= params
->spp_pathmtu
;
2400 sctp_assoc_sync_pmtu(sctp_opt2sk(sp
), asoc
);
2402 asoc
->pathmtu
= params
->spp_pathmtu
;
2403 sctp_frag_point(asoc
, params
->spp_pathmtu
);
2405 sp
->pathmtu
= params
->spp_pathmtu
;
2411 int update
= (trans
->param_flags
& SPP_PMTUD_DISABLE
) &&
2412 (params
->spp_flags
& SPP_PMTUD_ENABLE
);
2413 trans
->param_flags
=
2414 (trans
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2416 sctp_transport_pmtu(trans
, sctp_opt2sk(sp
));
2417 sctp_assoc_sync_pmtu(sctp_opt2sk(sp
), asoc
);
2421 (asoc
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2424 (sp
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2428 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2429 * value of this field is ignored. Note also that a value of zero
2430 * indicates the current setting should be left unchanged.
2432 if ((params
->spp_flags
& SPP_SACKDELAY_ENABLE
) && params
->spp_sackdelay
) {
2435 msecs_to_jiffies(params
->spp_sackdelay
);
2438 msecs_to_jiffies(params
->spp_sackdelay
);
2440 sp
->sackdelay
= params
->spp_sackdelay
;
2444 if (sackdelay_change
) {
2446 trans
->param_flags
=
2447 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2451 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2455 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2460 /* Note that a value of zero indicates the current setting should be
2463 if (params
->spp_pathmaxrxt
) {
2465 trans
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2467 asoc
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2469 sp
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2476 static int sctp_setsockopt_peer_addr_params(struct sock
*sk
,
2477 char __user
*optval
,
2478 unsigned int optlen
)
2480 struct sctp_paddrparams params
;
2481 struct sctp_transport
*trans
= NULL
;
2482 struct sctp_association
*asoc
= NULL
;
2483 struct sctp_sock
*sp
= sctp_sk(sk
);
2485 int hb_change
, pmtud_change
, sackdelay_change
;
2487 if (optlen
!= sizeof(struct sctp_paddrparams
))
2490 if (copy_from_user(¶ms
, optval
, optlen
))
2493 /* Validate flags and value parameters. */
2494 hb_change
= params
.spp_flags
& SPP_HB
;
2495 pmtud_change
= params
.spp_flags
& SPP_PMTUD
;
2496 sackdelay_change
= params
.spp_flags
& SPP_SACKDELAY
;
2498 if (hb_change
== SPP_HB
||
2499 pmtud_change
== SPP_PMTUD
||
2500 sackdelay_change
== SPP_SACKDELAY
||
2501 params
.spp_sackdelay
> 500 ||
2502 (params
.spp_pathmtu
&&
2503 params
.spp_pathmtu
< SCTP_DEFAULT_MINSEGMENT
))
2506 /* If an address other than INADDR_ANY is specified, and
2507 * no transport is found, then the request is invalid.
2509 if (!sctp_is_any(sk
, (union sctp_addr
*)¶ms
.spp_address
)) {
2510 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
2511 params
.spp_assoc_id
);
2516 /* Get association, if assoc_id != 0 and the socket is a one
2517 * to many style socket, and an association was not found, then
2518 * the id was invalid.
2520 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
2521 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
))
2524 /* Heartbeat demand can only be sent on a transport or
2525 * association, but not a socket.
2527 if (params
.spp_flags
& SPP_HB_DEMAND
&& !trans
&& !asoc
)
2530 /* Process parameters. */
2531 error
= sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2532 hb_change
, pmtud_change
,
2538 /* If changes are for association, also apply parameters to each
2541 if (!trans
&& asoc
) {
2542 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2544 sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2545 hb_change
, pmtud_change
,
2553 static inline __u32
sctp_spp_sackdelay_enable(__u32 param_flags
)
2555 return (param_flags
& ~SPP_SACKDELAY
) | SPP_SACKDELAY_ENABLE
;
2558 static inline __u32
sctp_spp_sackdelay_disable(__u32 param_flags
)
2560 return (param_flags
& ~SPP_SACKDELAY
) | SPP_SACKDELAY_DISABLE
;
2564 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2566 * This option will effect the way delayed acks are performed. This
2567 * option allows you to get or set the delayed ack time, in
2568 * milliseconds. It also allows changing the delayed ack frequency.
2569 * Changing the frequency to 1 disables the delayed sack algorithm. If
2570 * the assoc_id is 0, then this sets or gets the endpoints default
2571 * values. If the assoc_id field is non-zero, then the set or get
2572 * effects the specified association for the one to many model (the
2573 * assoc_id field is ignored by the one to one model). Note that if
2574 * sack_delay or sack_freq are 0 when setting this option, then the
2575 * current values will remain unchanged.
2577 * struct sctp_sack_info {
2578 * sctp_assoc_t sack_assoc_id;
2579 * uint32_t sack_delay;
2580 * uint32_t sack_freq;
2583 * sack_assoc_id - This parameter, indicates which association the user
2584 * is performing an action upon. Note that if this field's value is
2585 * zero then the endpoints default value is changed (effecting future
2586 * associations only).
2588 * sack_delay - This parameter contains the number of milliseconds that
2589 * the user is requesting the delayed ACK timer be set to. Note that
2590 * this value is defined in the standard to be between 200 and 500
2593 * sack_freq - This parameter contains the number of packets that must
2594 * be received before a sack is sent without waiting for the delay
2595 * timer to expire. The default value for this is 2, setting this
2596 * value to 1 will disable the delayed sack algorithm.
2599 static int sctp_setsockopt_delayed_ack(struct sock
*sk
,
2600 char __user
*optval
, unsigned int optlen
)
2602 struct sctp_sack_info params
;
2603 struct sctp_transport
*trans
= NULL
;
2604 struct sctp_association
*asoc
= NULL
;
2605 struct sctp_sock
*sp
= sctp_sk(sk
);
2607 if (optlen
== sizeof(struct sctp_sack_info
)) {
2608 if (copy_from_user(¶ms
, optval
, optlen
))
2611 if (params
.sack_delay
== 0 && params
.sack_freq
== 0)
2613 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
2614 pr_warn_ratelimited(DEPRECATED
2616 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2617 "Use struct sctp_sack_info instead\n",
2618 current
->comm
, task_pid_nr(current
));
2619 if (copy_from_user(¶ms
, optval
, optlen
))
2622 if (params
.sack_delay
== 0)
2623 params
.sack_freq
= 1;
2625 params
.sack_freq
= 0;
2629 /* Validate value parameter. */
2630 if (params
.sack_delay
> 500)
2633 /* Get association, if sack_assoc_id != 0 and the socket is a one
2634 * to many style socket, and an association was not found, then
2635 * the id was invalid.
2637 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
2638 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
2641 if (params
.sack_delay
) {
2644 msecs_to_jiffies(params
.sack_delay
);
2646 sctp_spp_sackdelay_enable(asoc
->param_flags
);
2648 sp
->sackdelay
= params
.sack_delay
;
2650 sctp_spp_sackdelay_enable(sp
->param_flags
);
2654 if (params
.sack_freq
== 1) {
2657 sctp_spp_sackdelay_disable(asoc
->param_flags
);
2660 sctp_spp_sackdelay_disable(sp
->param_flags
);
2662 } else if (params
.sack_freq
> 1) {
2664 asoc
->sackfreq
= params
.sack_freq
;
2666 sctp_spp_sackdelay_enable(asoc
->param_flags
);
2668 sp
->sackfreq
= params
.sack_freq
;
2670 sctp_spp_sackdelay_enable(sp
->param_flags
);
2674 /* If change is for association, also apply to each transport. */
2676 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2678 if (params
.sack_delay
) {
2680 msecs_to_jiffies(params
.sack_delay
);
2681 trans
->param_flags
=
2682 sctp_spp_sackdelay_enable(trans
->param_flags
);
2684 if (params
.sack_freq
== 1) {
2685 trans
->param_flags
=
2686 sctp_spp_sackdelay_disable(trans
->param_flags
);
2687 } else if (params
.sack_freq
> 1) {
2688 trans
->sackfreq
= params
.sack_freq
;
2689 trans
->param_flags
=
2690 sctp_spp_sackdelay_enable(trans
->param_flags
);
2698 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2700 * Applications can specify protocol parameters for the default association
2701 * initialization. The option name argument to setsockopt() and getsockopt()
2704 * Setting initialization parameters is effective only on an unconnected
2705 * socket (for UDP-style sockets only future associations are effected
2706 * by the change). With TCP-style sockets, this option is inherited by
2707 * sockets derived from a listener socket.
2709 static int sctp_setsockopt_initmsg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2711 struct sctp_initmsg sinit
;
2712 struct sctp_sock
*sp
= sctp_sk(sk
);
2714 if (optlen
!= sizeof(struct sctp_initmsg
))
2716 if (copy_from_user(&sinit
, optval
, optlen
))
2719 if (sinit
.sinit_num_ostreams
)
2720 sp
->initmsg
.sinit_num_ostreams
= sinit
.sinit_num_ostreams
;
2721 if (sinit
.sinit_max_instreams
)
2722 sp
->initmsg
.sinit_max_instreams
= sinit
.sinit_max_instreams
;
2723 if (sinit
.sinit_max_attempts
)
2724 sp
->initmsg
.sinit_max_attempts
= sinit
.sinit_max_attempts
;
2725 if (sinit
.sinit_max_init_timeo
)
2726 sp
->initmsg
.sinit_max_init_timeo
= sinit
.sinit_max_init_timeo
;
2732 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2734 * Applications that wish to use the sendto() system call may wish to
2735 * specify a default set of parameters that would normally be supplied
2736 * through the inclusion of ancillary data. This socket option allows
2737 * such an application to set the default sctp_sndrcvinfo structure.
2738 * The application that wishes to use this socket option simply passes
2739 * in to this call the sctp_sndrcvinfo structure defined in Section
2740 * 5.2.2) The input parameters accepted by this call include
2741 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2742 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2743 * to this call if the caller is using the UDP model.
2745 static int sctp_setsockopt_default_send_param(struct sock
*sk
,
2746 char __user
*optval
,
2747 unsigned int optlen
)
2749 struct sctp_sndrcvinfo info
;
2750 struct sctp_association
*asoc
;
2751 struct sctp_sock
*sp
= sctp_sk(sk
);
2753 if (optlen
!= sizeof(struct sctp_sndrcvinfo
))
2755 if (copy_from_user(&info
, optval
, optlen
))
2758 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
2759 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
2763 asoc
->default_stream
= info
.sinfo_stream
;
2764 asoc
->default_flags
= info
.sinfo_flags
;
2765 asoc
->default_ppid
= info
.sinfo_ppid
;
2766 asoc
->default_context
= info
.sinfo_context
;
2767 asoc
->default_timetolive
= info
.sinfo_timetolive
;
2769 sp
->default_stream
= info
.sinfo_stream
;
2770 sp
->default_flags
= info
.sinfo_flags
;
2771 sp
->default_ppid
= info
.sinfo_ppid
;
2772 sp
->default_context
= info
.sinfo_context
;
2773 sp
->default_timetolive
= info
.sinfo_timetolive
;
2779 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2781 * Requests that the local SCTP stack use the enclosed peer address as
2782 * the association primary. The enclosed address must be one of the
2783 * association peer's addresses.
2785 static int sctp_setsockopt_primary_addr(struct sock
*sk
, char __user
*optval
,
2786 unsigned int optlen
)
2788 struct sctp_prim prim
;
2789 struct sctp_transport
*trans
;
2791 if (optlen
!= sizeof(struct sctp_prim
))
2794 if (copy_from_user(&prim
, optval
, sizeof(struct sctp_prim
)))
2797 trans
= sctp_addr_id2transport(sk
, &prim
.ssp_addr
, prim
.ssp_assoc_id
);
2801 sctp_assoc_set_primary(trans
->asoc
, trans
);
2807 * 7.1.5 SCTP_NODELAY
2809 * Turn on/off any Nagle-like algorithm. This means that packets are
2810 * generally sent as soon as possible and no unnecessary delays are
2811 * introduced, at the cost of more packets in the network. Expects an
2812 * integer boolean flag.
2814 static int sctp_setsockopt_nodelay(struct sock
*sk
, char __user
*optval
,
2815 unsigned int optlen
)
2819 if (optlen
< sizeof(int))
2821 if (get_user(val
, (int __user
*)optval
))
2824 sctp_sk(sk
)->nodelay
= (val
== 0) ? 0 : 1;
2830 * 7.1.1 SCTP_RTOINFO
2832 * The protocol parameters used to initialize and bound retransmission
2833 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2834 * and modify these parameters.
2835 * All parameters are time values, in milliseconds. A value of 0, when
2836 * modifying the parameters, indicates that the current value should not
2840 static int sctp_setsockopt_rtoinfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2842 struct sctp_rtoinfo rtoinfo
;
2843 struct sctp_association
*asoc
;
2844 unsigned long rto_min
, rto_max
;
2845 struct sctp_sock
*sp
= sctp_sk(sk
);
2847 if (optlen
!= sizeof (struct sctp_rtoinfo
))
2850 if (copy_from_user(&rtoinfo
, optval
, optlen
))
2853 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
2855 /* Set the values to the specific association */
2856 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
2859 rto_max
= rtoinfo
.srto_max
;
2860 rto_min
= rtoinfo
.srto_min
;
2863 rto_max
= asoc
? msecs_to_jiffies(rto_max
) : rto_max
;
2865 rto_max
= asoc
? asoc
->rto_max
: sp
->rtoinfo
.srto_max
;
2868 rto_min
= asoc
? msecs_to_jiffies(rto_min
) : rto_min
;
2870 rto_min
= asoc
? asoc
->rto_min
: sp
->rtoinfo
.srto_min
;
2872 if (rto_min
> rto_max
)
2876 if (rtoinfo
.srto_initial
!= 0)
2878 msecs_to_jiffies(rtoinfo
.srto_initial
);
2879 asoc
->rto_max
= rto_max
;
2880 asoc
->rto_min
= rto_min
;
2882 /* If there is no association or the association-id = 0
2883 * set the values to the endpoint.
2885 if (rtoinfo
.srto_initial
!= 0)
2886 sp
->rtoinfo
.srto_initial
= rtoinfo
.srto_initial
;
2887 sp
->rtoinfo
.srto_max
= rto_max
;
2888 sp
->rtoinfo
.srto_min
= rto_min
;
2896 * 7.1.2 SCTP_ASSOCINFO
2898 * This option is used to tune the maximum retransmission attempts
2899 * of the association.
2900 * Returns an error if the new association retransmission value is
2901 * greater than the sum of the retransmission value of the peer.
2902 * See [SCTP] for more information.
2905 static int sctp_setsockopt_associnfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2908 struct sctp_assocparams assocparams
;
2909 struct sctp_association
*asoc
;
2911 if (optlen
!= sizeof(struct sctp_assocparams
))
2913 if (copy_from_user(&assocparams
, optval
, optlen
))
2916 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
2918 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
2921 /* Set the values to the specific association */
2923 if (assocparams
.sasoc_asocmaxrxt
!= 0) {
2926 struct sctp_transport
*peer_addr
;
2928 list_for_each_entry(peer_addr
, &asoc
->peer
.transport_addr_list
,
2930 path_sum
+= peer_addr
->pathmaxrxt
;
2934 /* Only validate asocmaxrxt if we have more than
2935 * one path/transport. We do this because path
2936 * retransmissions are only counted when we have more
2940 assocparams
.sasoc_asocmaxrxt
> path_sum
)
2943 asoc
->max_retrans
= assocparams
.sasoc_asocmaxrxt
;
2946 if (assocparams
.sasoc_cookie_life
!= 0)
2947 asoc
->cookie_life
= ms_to_ktime(assocparams
.sasoc_cookie_life
);
2949 /* Set the values to the endpoint */
2950 struct sctp_sock
*sp
= sctp_sk(sk
);
2952 if (assocparams
.sasoc_asocmaxrxt
!= 0)
2953 sp
->assocparams
.sasoc_asocmaxrxt
=
2954 assocparams
.sasoc_asocmaxrxt
;
2955 if (assocparams
.sasoc_cookie_life
!= 0)
2956 sp
->assocparams
.sasoc_cookie_life
=
2957 assocparams
.sasoc_cookie_life
;
2963 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2965 * This socket option is a boolean flag which turns on or off mapped V4
2966 * addresses. If this option is turned on and the socket is type
2967 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2968 * If this option is turned off, then no mapping will be done of V4
2969 * addresses and a user will receive both PF_INET6 and PF_INET type
2970 * addresses on the socket.
2972 static int sctp_setsockopt_mappedv4(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2975 struct sctp_sock
*sp
= sctp_sk(sk
);
2977 if (optlen
< sizeof(int))
2979 if (get_user(val
, (int __user
*)optval
))
2990 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2991 * This option will get or set the maximum size to put in any outgoing
2992 * SCTP DATA chunk. If a message is larger than this size it will be
2993 * fragmented by SCTP into the specified size. Note that the underlying
2994 * SCTP implementation may fragment into smaller sized chunks when the
2995 * PMTU of the underlying association is smaller than the value set by
2996 * the user. The default value for this option is '0' which indicates
2997 * the user is NOT limiting fragmentation and only the PMTU will effect
2998 * SCTP's choice of DATA chunk size. Note also that values set larger
2999 * than the maximum size of an IP datagram will effectively let SCTP
3000 * control fragmentation (i.e. the same as setting this option to 0).
3002 * The following structure is used to access and modify this parameter:
3004 * struct sctp_assoc_value {
3005 * sctp_assoc_t assoc_id;
3006 * uint32_t assoc_value;
3009 * assoc_id: This parameter is ignored for one-to-one style sockets.
3010 * For one-to-many style sockets this parameter indicates which
3011 * association the user is performing an action upon. Note that if
3012 * this field's value is zero then the endpoints default value is
3013 * changed (effecting future associations only).
3014 * assoc_value: This parameter specifies the maximum size in bytes.
3016 static int sctp_setsockopt_maxseg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
3018 struct sctp_assoc_value params
;
3019 struct sctp_association
*asoc
;
3020 struct sctp_sock
*sp
= sctp_sk(sk
);
3023 if (optlen
== sizeof(int)) {
3024 pr_warn_ratelimited(DEPRECATED
3026 "Use of int in maxseg socket option.\n"
3027 "Use struct sctp_assoc_value instead\n",
3028 current
->comm
, task_pid_nr(current
));
3029 if (copy_from_user(&val
, optval
, optlen
))
3031 params
.assoc_id
= 0;
3032 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
3033 if (copy_from_user(¶ms
, optval
, optlen
))
3035 val
= params
.assoc_value
;
3039 if ((val
!= 0) && ((val
< 8) || (val
> SCTP_MAX_CHUNK_LEN
)))
3042 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3043 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
3048 val
= asoc
->pathmtu
;
3049 val
-= sp
->pf
->af
->net_header_len
;
3050 val
-= sizeof(struct sctphdr
) +
3051 sizeof(struct sctp_data_chunk
);
3053 asoc
->user_frag
= val
;
3054 asoc
->frag_point
= sctp_frag_point(asoc
, asoc
->pathmtu
);
3056 sp
->user_frag
= val
;
3064 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3066 * Requests that the peer mark the enclosed address as the association
3067 * primary. The enclosed address must be one of the association's
3068 * locally bound addresses. The following structure is used to make a
3069 * set primary request:
3071 static int sctp_setsockopt_peer_primary_addr(struct sock
*sk
, char __user
*optval
,
3072 unsigned int optlen
)
3074 struct net
*net
= sock_net(sk
);
3075 struct sctp_sock
*sp
;
3076 struct sctp_association
*asoc
= NULL
;
3077 struct sctp_setpeerprim prim
;
3078 struct sctp_chunk
*chunk
;
3084 if (!net
->sctp
.addip_enable
)
3087 if (optlen
!= sizeof(struct sctp_setpeerprim
))
3090 if (copy_from_user(&prim
, optval
, optlen
))
3093 asoc
= sctp_id2assoc(sk
, prim
.sspp_assoc_id
);
3097 if (!asoc
->peer
.asconf_capable
)
3100 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_SET_PRIMARY
)
3103 if (!sctp_state(asoc
, ESTABLISHED
))
3106 af
= sctp_get_af_specific(prim
.sspp_addr
.ss_family
);
3110 if (!af
->addr_valid((union sctp_addr
*)&prim
.sspp_addr
, sp
, NULL
))
3111 return -EADDRNOTAVAIL
;
3113 if (!sctp_assoc_lookup_laddr(asoc
, (union sctp_addr
*)&prim
.sspp_addr
))
3114 return -EADDRNOTAVAIL
;
3116 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3117 chunk
= sctp_make_asconf_set_prim(asoc
,
3118 (union sctp_addr
*)&prim
.sspp_addr
);
3122 err
= sctp_send_asconf(asoc
, chunk
);
3124 pr_debug("%s: we set peer primary addr primitively\n", __func__
);
3129 static int sctp_setsockopt_adaptation_layer(struct sock
*sk
, char __user
*optval
,
3130 unsigned int optlen
)
3132 struct sctp_setadaptation adaptation
;
3134 if (optlen
!= sizeof(struct sctp_setadaptation
))
3136 if (copy_from_user(&adaptation
, optval
, optlen
))
3139 sctp_sk(sk
)->adaptation_ind
= adaptation
.ssb_adaptation_ind
;
3145 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3147 * The context field in the sctp_sndrcvinfo structure is normally only
3148 * used when a failed message is retrieved holding the value that was
3149 * sent down on the actual send call. This option allows the setting of
3150 * a default context on an association basis that will be received on
3151 * reading messages from the peer. This is especially helpful in the
3152 * one-2-many model for an application to keep some reference to an
3153 * internal state machine that is processing messages on the
3154 * association. Note that the setting of this value only effects
3155 * received messages from the peer and does not effect the value that is
3156 * saved with outbound messages.
3158 static int sctp_setsockopt_context(struct sock
*sk
, char __user
*optval
,
3159 unsigned int optlen
)
3161 struct sctp_assoc_value params
;
3162 struct sctp_sock
*sp
;
3163 struct sctp_association
*asoc
;
3165 if (optlen
!= sizeof(struct sctp_assoc_value
))
3167 if (copy_from_user(¶ms
, optval
, optlen
))
3172 if (params
.assoc_id
!= 0) {
3173 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3176 asoc
->default_rcv_context
= params
.assoc_value
;
3178 sp
->default_rcv_context
= params
.assoc_value
;
3185 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3187 * This options will at a minimum specify if the implementation is doing
3188 * fragmented interleave. Fragmented interleave, for a one to many
3189 * socket, is when subsequent calls to receive a message may return
3190 * parts of messages from different associations. Some implementations
3191 * may allow you to turn this value on or off. If so, when turned off,
3192 * no fragment interleave will occur (which will cause a head of line
3193 * blocking amongst multiple associations sharing the same one to many
3194 * socket). When this option is turned on, then each receive call may
3195 * come from a different association (thus the user must receive data
3196 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3197 * association each receive belongs to.
3199 * This option takes a boolean value. A non-zero value indicates that
3200 * fragmented interleave is on. A value of zero indicates that
3201 * fragmented interleave is off.
3203 * Note that it is important that an implementation that allows this
3204 * option to be turned on, have it off by default. Otherwise an unaware
3205 * application using the one to many model may become confused and act
3208 static int sctp_setsockopt_fragment_interleave(struct sock
*sk
,
3209 char __user
*optval
,
3210 unsigned int optlen
)
3214 if (optlen
!= sizeof(int))
3216 if (get_user(val
, (int __user
*)optval
))
3219 sctp_sk(sk
)->frag_interleave
= (val
== 0) ? 0 : 1;
3225 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3226 * (SCTP_PARTIAL_DELIVERY_POINT)
3228 * This option will set or get the SCTP partial delivery point. This
3229 * point is the size of a message where the partial delivery API will be
3230 * invoked to help free up rwnd space for the peer. Setting this to a
3231 * lower value will cause partial deliveries to happen more often. The
3232 * calls argument is an integer that sets or gets the partial delivery
3233 * point. Note also that the call will fail if the user attempts to set
3234 * this value larger than the socket receive buffer size.
3236 * Note that any single message having a length smaller than or equal to
3237 * the SCTP partial delivery point will be delivered in one single read
3238 * call as long as the user provided buffer is large enough to hold the
3241 static int sctp_setsockopt_partial_delivery_point(struct sock
*sk
,
3242 char __user
*optval
,
3243 unsigned int optlen
)
3247 if (optlen
!= sizeof(u32
))
3249 if (get_user(val
, (int __user
*)optval
))
3252 /* Note: We double the receive buffer from what the user sets
3253 * it to be, also initial rwnd is based on rcvbuf/2.
3255 if (val
> (sk
->sk_rcvbuf
>> 1))
3258 sctp_sk(sk
)->pd_point
= val
;
3260 return 0; /* is this the right error code? */
3264 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3266 * This option will allow a user to change the maximum burst of packets
3267 * that can be emitted by this association. Note that the default value
3268 * is 4, and some implementations may restrict this setting so that it
3269 * can only be lowered.
3271 * NOTE: This text doesn't seem right. Do this on a socket basis with
3272 * future associations inheriting the socket value.
3274 static int sctp_setsockopt_maxburst(struct sock
*sk
,
3275 char __user
*optval
,
3276 unsigned int optlen
)
3278 struct sctp_assoc_value params
;
3279 struct sctp_sock
*sp
;
3280 struct sctp_association
*asoc
;
3284 if (optlen
== sizeof(int)) {
3285 pr_warn_ratelimited(DEPRECATED
3287 "Use of int in max_burst socket option deprecated.\n"
3288 "Use struct sctp_assoc_value instead\n",
3289 current
->comm
, task_pid_nr(current
));
3290 if (copy_from_user(&val
, optval
, optlen
))
3292 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
3293 if (copy_from_user(¶ms
, optval
, optlen
))
3295 val
= params
.assoc_value
;
3296 assoc_id
= params
.assoc_id
;
3302 if (assoc_id
!= 0) {
3303 asoc
= sctp_id2assoc(sk
, assoc_id
);
3306 asoc
->max_burst
= val
;
3308 sp
->max_burst
= val
;
3314 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3316 * This set option adds a chunk type that the user is requesting to be
3317 * received only in an authenticated way. Changes to the list of chunks
3318 * will only effect future associations on the socket.
3320 static int sctp_setsockopt_auth_chunk(struct sock
*sk
,
3321 char __user
*optval
,
3322 unsigned int optlen
)
3324 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3325 struct sctp_authchunk val
;
3327 if (!ep
->auth_enable
)
3330 if (optlen
!= sizeof(struct sctp_authchunk
))
3332 if (copy_from_user(&val
, optval
, optlen
))
3335 switch (val
.sauth_chunk
) {
3337 case SCTP_CID_INIT_ACK
:
3338 case SCTP_CID_SHUTDOWN_COMPLETE
:
3343 /* add this chunk id to the endpoint */
3344 return sctp_auth_ep_add_chunkid(ep
, val
.sauth_chunk
);
3348 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3350 * This option gets or sets the list of HMAC algorithms that the local
3351 * endpoint requires the peer to use.
3353 static int sctp_setsockopt_hmac_ident(struct sock
*sk
,
3354 char __user
*optval
,
3355 unsigned int optlen
)
3357 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3358 struct sctp_hmacalgo
*hmacs
;
3362 if (!ep
->auth_enable
)
3365 if (optlen
< sizeof(struct sctp_hmacalgo
))
3368 hmacs
= memdup_user(optval
, optlen
);
3370 return PTR_ERR(hmacs
);
3372 idents
= hmacs
->shmac_num_idents
;
3373 if (idents
== 0 || idents
> SCTP_AUTH_NUM_HMACS
||
3374 (idents
* sizeof(u16
)) > (optlen
- sizeof(struct sctp_hmacalgo
))) {
3379 err
= sctp_auth_ep_set_hmacs(ep
, hmacs
);
3386 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3388 * This option will set a shared secret key which is used to build an
3389 * association shared key.
3391 static int sctp_setsockopt_auth_key(struct sock
*sk
,
3392 char __user
*optval
,
3393 unsigned int optlen
)
3395 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3396 struct sctp_authkey
*authkey
;
3397 struct sctp_association
*asoc
;
3400 if (!ep
->auth_enable
)
3403 if (optlen
<= sizeof(struct sctp_authkey
))
3406 authkey
= memdup_user(optval
, optlen
);
3407 if (IS_ERR(authkey
))
3408 return PTR_ERR(authkey
);
3410 if (authkey
->sca_keylength
> optlen
- sizeof(struct sctp_authkey
)) {
3415 asoc
= sctp_id2assoc(sk
, authkey
->sca_assoc_id
);
3416 if (!asoc
&& authkey
->sca_assoc_id
&& sctp_style(sk
, UDP
)) {
3421 ret
= sctp_auth_set_key(ep
, asoc
, authkey
);
3428 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3430 * This option will get or set the active shared key to be used to build
3431 * the association shared key.
3433 static int sctp_setsockopt_active_key(struct sock
*sk
,
3434 char __user
*optval
,
3435 unsigned int optlen
)
3437 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3438 struct sctp_authkeyid val
;
3439 struct sctp_association
*asoc
;
3441 if (!ep
->auth_enable
)
3444 if (optlen
!= sizeof(struct sctp_authkeyid
))
3446 if (copy_from_user(&val
, optval
, optlen
))
3449 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3450 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3453 return sctp_auth_set_active_key(ep
, asoc
, val
.scact_keynumber
);
3457 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3459 * This set option will delete a shared secret key from use.
3461 static int sctp_setsockopt_del_key(struct sock
*sk
,
3462 char __user
*optval
,
3463 unsigned int optlen
)
3465 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3466 struct sctp_authkeyid val
;
3467 struct sctp_association
*asoc
;
3469 if (!ep
->auth_enable
)
3472 if (optlen
!= sizeof(struct sctp_authkeyid
))
3474 if (copy_from_user(&val
, optval
, optlen
))
3477 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3478 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3481 return sctp_auth_del_key_id(ep
, asoc
, val
.scact_keynumber
);
3486 * 8.1.23 SCTP_AUTO_ASCONF
3488 * This option will enable or disable the use of the automatic generation of
3489 * ASCONF chunks to add and delete addresses to an existing association. Note
3490 * that this option has two caveats namely: a) it only affects sockets that
3491 * are bound to all addresses available to the SCTP stack, and b) the system
3492 * administrator may have an overriding control that turns the ASCONF feature
3493 * off no matter what setting the socket option may have.
3494 * This option expects an integer boolean flag, where a non-zero value turns on
3495 * the option, and a zero value turns off the option.
3496 * Note. In this implementation, socket operation overrides default parameter
3497 * being set by sysctl as well as FreeBSD implementation
3499 static int sctp_setsockopt_auto_asconf(struct sock
*sk
, char __user
*optval
,
3500 unsigned int optlen
)
3503 struct sctp_sock
*sp
= sctp_sk(sk
);
3505 if (optlen
< sizeof(int))
3507 if (get_user(val
, (int __user
*)optval
))
3509 if (!sctp_is_ep_boundall(sk
) && val
)
3511 if ((val
&& sp
->do_auto_asconf
) || (!val
&& !sp
->do_auto_asconf
))
3514 if (val
== 0 && sp
->do_auto_asconf
) {
3515 list_del(&sp
->auto_asconf_list
);
3516 sp
->do_auto_asconf
= 0;
3517 } else if (val
&& !sp
->do_auto_asconf
) {
3518 list_add_tail(&sp
->auto_asconf_list
,
3519 &sock_net(sk
)->sctp
.auto_asconf_splist
);
3520 sp
->do_auto_asconf
= 1;
3527 * SCTP_PEER_ADDR_THLDS
3529 * This option allows us to alter the partially failed threshold for one or all
3530 * transports in an association. See Section 6.1 of:
3531 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3533 static int sctp_setsockopt_paddr_thresholds(struct sock
*sk
,
3534 char __user
*optval
,
3535 unsigned int optlen
)
3537 struct sctp_paddrthlds val
;
3538 struct sctp_transport
*trans
;
3539 struct sctp_association
*asoc
;
3541 if (optlen
< sizeof(struct sctp_paddrthlds
))
3543 if (copy_from_user(&val
, (struct sctp_paddrthlds __user
*)optval
,
3544 sizeof(struct sctp_paddrthlds
)))
3548 if (sctp_is_any(sk
, (const union sctp_addr
*)&val
.spt_address
)) {
3549 asoc
= sctp_id2assoc(sk
, val
.spt_assoc_id
);
3552 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
3554 if (val
.spt_pathmaxrxt
)
3555 trans
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3556 trans
->pf_retrans
= val
.spt_pathpfthld
;
3559 if (val
.spt_pathmaxrxt
)
3560 asoc
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3561 asoc
->pf_retrans
= val
.spt_pathpfthld
;
3563 trans
= sctp_addr_id2transport(sk
, &val
.spt_address
,
3568 if (val
.spt_pathmaxrxt
)
3569 trans
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3570 trans
->pf_retrans
= val
.spt_pathpfthld
;
3576 /* API 6.2 setsockopt(), getsockopt()
3578 * Applications use setsockopt() and getsockopt() to set or retrieve
3579 * socket options. Socket options are used to change the default
3580 * behavior of sockets calls. They are described in Section 7.
3584 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3585 * int __user *optlen);
3586 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3589 * sd - the socket descript.
3590 * level - set to IPPROTO_SCTP for all SCTP options.
3591 * optname - the option name.
3592 * optval - the buffer to store the value of the option.
3593 * optlen - the size of the buffer.
3595 static int sctp_setsockopt(struct sock
*sk
, int level
, int optname
,
3596 char __user
*optval
, unsigned int optlen
)
3600 pr_debug("%s: sk:%p, optname:%d\n", __func__
, sk
, optname
);
3602 /* I can hardly begin to describe how wrong this is. This is
3603 * so broken as to be worse than useless. The API draft
3604 * REALLY is NOT helpful here... I am not convinced that the
3605 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3606 * are at all well-founded.
3608 if (level
!= SOL_SCTP
) {
3609 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
3610 retval
= af
->setsockopt(sk
, level
, optname
, optval
, optlen
);
3617 case SCTP_SOCKOPT_BINDX_ADD
:
3618 /* 'optlen' is the size of the addresses buffer. */
3619 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3620 optlen
, SCTP_BINDX_ADD_ADDR
);
3623 case SCTP_SOCKOPT_BINDX_REM
:
3624 /* 'optlen' is the size of the addresses buffer. */
3625 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3626 optlen
, SCTP_BINDX_REM_ADDR
);
3629 case SCTP_SOCKOPT_CONNECTX_OLD
:
3630 /* 'optlen' is the size of the addresses buffer. */
3631 retval
= sctp_setsockopt_connectx_old(sk
,
3632 (struct sockaddr __user
*)optval
,
3636 case SCTP_SOCKOPT_CONNECTX
:
3637 /* 'optlen' is the size of the addresses buffer. */
3638 retval
= sctp_setsockopt_connectx(sk
,
3639 (struct sockaddr __user
*)optval
,
3643 case SCTP_DISABLE_FRAGMENTS
:
3644 retval
= sctp_setsockopt_disable_fragments(sk
, optval
, optlen
);
3648 retval
= sctp_setsockopt_events(sk
, optval
, optlen
);
3651 case SCTP_AUTOCLOSE
:
3652 retval
= sctp_setsockopt_autoclose(sk
, optval
, optlen
);
3655 case SCTP_PEER_ADDR_PARAMS
:
3656 retval
= sctp_setsockopt_peer_addr_params(sk
, optval
, optlen
);
3659 case SCTP_DELAYED_SACK
:
3660 retval
= sctp_setsockopt_delayed_ack(sk
, optval
, optlen
);
3662 case SCTP_PARTIAL_DELIVERY_POINT
:
3663 retval
= sctp_setsockopt_partial_delivery_point(sk
, optval
, optlen
);
3667 retval
= sctp_setsockopt_initmsg(sk
, optval
, optlen
);
3669 case SCTP_DEFAULT_SEND_PARAM
:
3670 retval
= sctp_setsockopt_default_send_param(sk
, optval
,
3673 case SCTP_PRIMARY_ADDR
:
3674 retval
= sctp_setsockopt_primary_addr(sk
, optval
, optlen
);
3676 case SCTP_SET_PEER_PRIMARY_ADDR
:
3677 retval
= sctp_setsockopt_peer_primary_addr(sk
, optval
, optlen
);
3680 retval
= sctp_setsockopt_nodelay(sk
, optval
, optlen
);
3683 retval
= sctp_setsockopt_rtoinfo(sk
, optval
, optlen
);
3685 case SCTP_ASSOCINFO
:
3686 retval
= sctp_setsockopt_associnfo(sk
, optval
, optlen
);
3688 case SCTP_I_WANT_MAPPED_V4_ADDR
:
3689 retval
= sctp_setsockopt_mappedv4(sk
, optval
, optlen
);
3692 retval
= sctp_setsockopt_maxseg(sk
, optval
, optlen
);
3694 case SCTP_ADAPTATION_LAYER
:
3695 retval
= sctp_setsockopt_adaptation_layer(sk
, optval
, optlen
);
3698 retval
= sctp_setsockopt_context(sk
, optval
, optlen
);
3700 case SCTP_FRAGMENT_INTERLEAVE
:
3701 retval
= sctp_setsockopt_fragment_interleave(sk
, optval
, optlen
);
3703 case SCTP_MAX_BURST
:
3704 retval
= sctp_setsockopt_maxburst(sk
, optval
, optlen
);
3706 case SCTP_AUTH_CHUNK
:
3707 retval
= sctp_setsockopt_auth_chunk(sk
, optval
, optlen
);
3709 case SCTP_HMAC_IDENT
:
3710 retval
= sctp_setsockopt_hmac_ident(sk
, optval
, optlen
);
3713 retval
= sctp_setsockopt_auth_key(sk
, optval
, optlen
);
3715 case SCTP_AUTH_ACTIVE_KEY
:
3716 retval
= sctp_setsockopt_active_key(sk
, optval
, optlen
);
3718 case SCTP_AUTH_DELETE_KEY
:
3719 retval
= sctp_setsockopt_del_key(sk
, optval
, optlen
);
3721 case SCTP_AUTO_ASCONF
:
3722 retval
= sctp_setsockopt_auto_asconf(sk
, optval
, optlen
);
3724 case SCTP_PEER_ADDR_THLDS
:
3725 retval
= sctp_setsockopt_paddr_thresholds(sk
, optval
, optlen
);
3728 retval
= -ENOPROTOOPT
;
3738 /* API 3.1.6 connect() - UDP Style Syntax
3740 * An application may use the connect() call in the UDP model to initiate an
3741 * association without sending data.
3745 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3747 * sd: the socket descriptor to have a new association added to.
3749 * nam: the address structure (either struct sockaddr_in or struct
3750 * sockaddr_in6 defined in RFC2553 [7]).
3752 * len: the size of the address.
3754 static int sctp_connect(struct sock
*sk
, struct sockaddr
*addr
,
3762 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__
, sk
,
3765 /* Validate addr_len before calling common connect/connectx routine. */
3766 af
= sctp_get_af_specific(addr
->sa_family
);
3767 if (!af
|| addr_len
< af
->sockaddr_len
) {
3770 /* Pass correct addr len to common routine (so it knows there
3771 * is only one address being passed.
3773 err
= __sctp_connect(sk
, addr
, af
->sockaddr_len
, NULL
);
3780 /* FIXME: Write comments. */
3781 static int sctp_disconnect(struct sock
*sk
, int flags
)
3783 return -EOPNOTSUPP
; /* STUB */
3786 /* 4.1.4 accept() - TCP Style Syntax
3788 * Applications use accept() call to remove an established SCTP
3789 * association from the accept queue of the endpoint. A new socket
3790 * descriptor will be returned from accept() to represent the newly
3791 * formed association.
3793 static struct sock
*sctp_accept(struct sock
*sk
, int flags
, int *err
)
3795 struct sctp_sock
*sp
;
3796 struct sctp_endpoint
*ep
;
3797 struct sock
*newsk
= NULL
;
3798 struct sctp_association
*asoc
;
3807 if (!sctp_style(sk
, TCP
)) {
3808 error
= -EOPNOTSUPP
;
3812 if (!sctp_sstate(sk
, LISTENING
)) {
3817 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
3819 error
= sctp_wait_for_accept(sk
, timeo
);
3823 /* We treat the list of associations on the endpoint as the accept
3824 * queue and pick the first association on the list.
3826 asoc
= list_entry(ep
->asocs
.next
, struct sctp_association
, asocs
);
3828 newsk
= sp
->pf
->create_accept_sk(sk
, asoc
);
3834 /* Populate the fields of the newsk from the oldsk and migrate the
3835 * asoc to the newsk.
3837 sctp_sock_migrate(sk
, newsk
, asoc
, SCTP_SOCKET_TCP
);
3845 /* The SCTP ioctl handler. */
3846 static int sctp_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
3853 * SEQPACKET-style sockets in LISTENING state are valid, for
3854 * SCTP, so only discard TCP-style sockets in LISTENING state.
3856 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
3861 struct sk_buff
*skb
;
3862 unsigned int amount
= 0;
3864 skb
= skb_peek(&sk
->sk_receive_queue
);
3867 * We will only return the amount of this packet since
3868 * that is all that will be read.
3872 rc
= put_user(amount
, (int __user
*)arg
);
3884 /* This is the function which gets called during socket creation to
3885 * initialized the SCTP-specific portion of the sock.
3886 * The sock structure should already be zero-filled memory.
3888 static int sctp_init_sock(struct sock
*sk
)
3890 struct net
*net
= sock_net(sk
);
3891 struct sctp_sock
*sp
;
3893 pr_debug("%s: sk:%p\n", __func__
, sk
);
3897 /* Initialize the SCTP per socket area. */
3898 switch (sk
->sk_type
) {
3899 case SOCK_SEQPACKET
:
3900 sp
->type
= SCTP_SOCKET_UDP
;
3903 sp
->type
= SCTP_SOCKET_TCP
;
3906 return -ESOCKTNOSUPPORT
;
3909 /* Initialize default send parameters. These parameters can be
3910 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3912 sp
->default_stream
= 0;
3913 sp
->default_ppid
= 0;
3914 sp
->default_flags
= 0;
3915 sp
->default_context
= 0;
3916 sp
->default_timetolive
= 0;
3918 sp
->default_rcv_context
= 0;
3919 sp
->max_burst
= net
->sctp
.max_burst
;
3921 sp
->sctp_hmac_alg
= net
->sctp
.sctp_hmac_alg
;
3923 /* Initialize default setup parameters. These parameters
3924 * can be modified with the SCTP_INITMSG socket option or
3925 * overridden by the SCTP_INIT CMSG.
3927 sp
->initmsg
.sinit_num_ostreams
= sctp_max_outstreams
;
3928 sp
->initmsg
.sinit_max_instreams
= sctp_max_instreams
;
3929 sp
->initmsg
.sinit_max_attempts
= net
->sctp
.max_retrans_init
;
3930 sp
->initmsg
.sinit_max_init_timeo
= net
->sctp
.rto_max
;
3932 /* Initialize default RTO related parameters. These parameters can
3933 * be modified for with the SCTP_RTOINFO socket option.
3935 sp
->rtoinfo
.srto_initial
= net
->sctp
.rto_initial
;
3936 sp
->rtoinfo
.srto_max
= net
->sctp
.rto_max
;
3937 sp
->rtoinfo
.srto_min
= net
->sctp
.rto_min
;
3939 /* Initialize default association related parameters. These parameters
3940 * can be modified with the SCTP_ASSOCINFO socket option.
3942 sp
->assocparams
.sasoc_asocmaxrxt
= net
->sctp
.max_retrans_association
;
3943 sp
->assocparams
.sasoc_number_peer_destinations
= 0;
3944 sp
->assocparams
.sasoc_peer_rwnd
= 0;
3945 sp
->assocparams
.sasoc_local_rwnd
= 0;
3946 sp
->assocparams
.sasoc_cookie_life
= net
->sctp
.valid_cookie_life
;
3948 /* Initialize default event subscriptions. By default, all the
3951 memset(&sp
->subscribe
, 0, sizeof(struct sctp_event_subscribe
));
3953 /* Default Peer Address Parameters. These defaults can
3954 * be modified via SCTP_PEER_ADDR_PARAMS
3956 sp
->hbinterval
= net
->sctp
.hb_interval
;
3957 sp
->pathmaxrxt
= net
->sctp
.max_retrans_path
;
3958 sp
->pathmtu
= 0; /* allow default discovery */
3959 sp
->sackdelay
= net
->sctp
.sack_timeout
;
3961 sp
->param_flags
= SPP_HB_ENABLE
|
3963 SPP_SACKDELAY_ENABLE
;
3965 /* If enabled no SCTP message fragmentation will be performed.
3966 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3968 sp
->disable_fragments
= 0;
3970 /* Enable Nagle algorithm by default. */
3973 /* Enable by default. */
3976 /* Auto-close idle associations after the configured
3977 * number of seconds. A value of 0 disables this
3978 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3979 * for UDP-style sockets only.
3983 /* User specified fragmentation limit. */
3986 sp
->adaptation_ind
= 0;
3988 sp
->pf
= sctp_get_pf_specific(sk
->sk_family
);
3990 /* Control variables for partial data delivery. */
3991 atomic_set(&sp
->pd_mode
, 0);
3992 skb_queue_head_init(&sp
->pd_lobby
);
3993 sp
->frag_interleave
= 0;
3995 /* Create a per socket endpoint structure. Even if we
3996 * change the data structure relationships, this may still
3997 * be useful for storing pre-connect address information.
3999 sp
->ep
= sctp_endpoint_new(sk
, GFP_KERNEL
);
4005 sk
->sk_destruct
= sctp_destruct_sock
;
4007 SCTP_DBG_OBJCNT_INC(sock
);
4010 percpu_counter_inc(&sctp_sockets_allocated
);
4011 sock_prot_inuse_add(net
, sk
->sk_prot
, 1);
4012 if (net
->sctp
.default_auto_asconf
) {
4013 list_add_tail(&sp
->auto_asconf_list
,
4014 &net
->sctp
.auto_asconf_splist
);
4015 sp
->do_auto_asconf
= 1;
4017 sp
->do_auto_asconf
= 0;
4023 /* Cleanup any SCTP per socket resources. */
4024 static void sctp_destroy_sock(struct sock
*sk
)
4026 struct sctp_sock
*sp
;
4028 pr_debug("%s: sk:%p\n", __func__
, sk
);
4030 /* Release our hold on the endpoint. */
4032 /* This could happen during socket init, thus we bail out
4033 * early, since the rest of the below is not setup either.
4038 if (sp
->do_auto_asconf
) {
4039 sp
->do_auto_asconf
= 0;
4040 list_del(&sp
->auto_asconf_list
);
4042 sctp_endpoint_free(sp
->ep
);
4044 percpu_counter_dec(&sctp_sockets_allocated
);
4045 sock_prot_inuse_add(sock_net(sk
), sk
->sk_prot
, -1);
4049 /* Triggered when there are no references on the socket anymore */
4050 static void sctp_destruct_sock(struct sock
*sk
)
4052 struct sctp_sock
*sp
= sctp_sk(sk
);
4054 /* Free up the HMAC transform. */
4055 crypto_free_hash(sp
->hmac
);
4057 inet_sock_destruct(sk
);
4060 /* API 4.1.7 shutdown() - TCP Style Syntax
4061 * int shutdown(int socket, int how);
4063 * sd - the socket descriptor of the association to be closed.
4064 * how - Specifies the type of shutdown. The values are
4067 * Disables further receive operations. No SCTP
4068 * protocol action is taken.
4070 * Disables further send operations, and initiates
4071 * the SCTP shutdown sequence.
4073 * Disables further send and receive operations
4074 * and initiates the SCTP shutdown sequence.
4076 static void sctp_shutdown(struct sock
*sk
, int how
)
4078 struct net
*net
= sock_net(sk
);
4079 struct sctp_endpoint
*ep
;
4080 struct sctp_association
*asoc
;
4082 if (!sctp_style(sk
, TCP
))
4085 if (how
& SEND_SHUTDOWN
) {
4086 ep
= sctp_sk(sk
)->ep
;
4087 if (!list_empty(&ep
->asocs
)) {
4088 asoc
= list_entry(ep
->asocs
.next
,
4089 struct sctp_association
, asocs
);
4090 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
4095 /* 7.2.1 Association Status (SCTP_STATUS)
4097 * Applications can retrieve current status information about an
4098 * association, including association state, peer receiver window size,
4099 * number of unacked data chunks, and number of data chunks pending
4100 * receipt. This information is read-only.
4102 static int sctp_getsockopt_sctp_status(struct sock
*sk
, int len
,
4103 char __user
*optval
,
4106 struct sctp_status status
;
4107 struct sctp_association
*asoc
= NULL
;
4108 struct sctp_transport
*transport
;
4109 sctp_assoc_t associd
;
4112 if (len
< sizeof(status
)) {
4117 len
= sizeof(status
);
4118 if (copy_from_user(&status
, optval
, len
)) {
4123 associd
= status
.sstat_assoc_id
;
4124 asoc
= sctp_id2assoc(sk
, associd
);
4130 transport
= asoc
->peer
.primary_path
;
4132 status
.sstat_assoc_id
= sctp_assoc2id(asoc
);
4133 status
.sstat_state
= asoc
->state
;
4134 status
.sstat_rwnd
= asoc
->peer
.rwnd
;
4135 status
.sstat_unackdata
= asoc
->unack_data
;
4137 status
.sstat_penddata
= sctp_tsnmap_pending(&asoc
->peer
.tsn_map
);
4138 status
.sstat_instrms
= asoc
->c
.sinit_max_instreams
;
4139 status
.sstat_outstrms
= asoc
->c
.sinit_num_ostreams
;
4140 status
.sstat_fragmentation_point
= asoc
->frag_point
;
4141 status
.sstat_primary
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
4142 memcpy(&status
.sstat_primary
.spinfo_address
, &transport
->ipaddr
,
4143 transport
->af_specific
->sockaddr_len
);
4144 /* Map ipv4 address into v4-mapped-on-v6 address. */
4145 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
4146 (union sctp_addr
*)&status
.sstat_primary
.spinfo_address
);
4147 status
.sstat_primary
.spinfo_state
= transport
->state
;
4148 status
.sstat_primary
.spinfo_cwnd
= transport
->cwnd
;
4149 status
.sstat_primary
.spinfo_srtt
= transport
->srtt
;
4150 status
.sstat_primary
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
4151 status
.sstat_primary
.spinfo_mtu
= transport
->pathmtu
;
4153 if (status
.sstat_primary
.spinfo_state
== SCTP_UNKNOWN
)
4154 status
.sstat_primary
.spinfo_state
= SCTP_ACTIVE
;
4156 if (put_user(len
, optlen
)) {
4161 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4162 __func__
, len
, status
.sstat_state
, status
.sstat_rwnd
,
4163 status
.sstat_assoc_id
);
4165 if (copy_to_user(optval
, &status
, len
)) {
4175 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4177 * Applications can retrieve information about a specific peer address
4178 * of an association, including its reachability state, congestion
4179 * window, and retransmission timer values. This information is
4182 static int sctp_getsockopt_peer_addr_info(struct sock
*sk
, int len
,
4183 char __user
*optval
,
4186 struct sctp_paddrinfo pinfo
;
4187 struct sctp_transport
*transport
;
4190 if (len
< sizeof(pinfo
)) {
4195 len
= sizeof(pinfo
);
4196 if (copy_from_user(&pinfo
, optval
, len
)) {
4201 transport
= sctp_addr_id2transport(sk
, &pinfo
.spinfo_address
,
4202 pinfo
.spinfo_assoc_id
);
4206 pinfo
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
4207 pinfo
.spinfo_state
= transport
->state
;
4208 pinfo
.spinfo_cwnd
= transport
->cwnd
;
4209 pinfo
.spinfo_srtt
= transport
->srtt
;
4210 pinfo
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
4211 pinfo
.spinfo_mtu
= transport
->pathmtu
;
4213 if (pinfo
.spinfo_state
== SCTP_UNKNOWN
)
4214 pinfo
.spinfo_state
= SCTP_ACTIVE
;
4216 if (put_user(len
, optlen
)) {
4221 if (copy_to_user(optval
, &pinfo
, len
)) {
4230 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4232 * This option is a on/off flag. If enabled no SCTP message
4233 * fragmentation will be performed. Instead if a message being sent
4234 * exceeds the current PMTU size, the message will NOT be sent and
4235 * instead a error will be indicated to the user.
4237 static int sctp_getsockopt_disable_fragments(struct sock
*sk
, int len
,
4238 char __user
*optval
, int __user
*optlen
)
4242 if (len
< sizeof(int))
4246 val
= (sctp_sk(sk
)->disable_fragments
== 1);
4247 if (put_user(len
, optlen
))
4249 if (copy_to_user(optval
, &val
, len
))
4254 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4256 * This socket option is used to specify various notifications and
4257 * ancillary data the user wishes to receive.
4259 static int sctp_getsockopt_events(struct sock
*sk
, int len
, char __user
*optval
,
4264 if (len
> sizeof(struct sctp_event_subscribe
))
4265 len
= sizeof(struct sctp_event_subscribe
);
4266 if (put_user(len
, optlen
))
4268 if (copy_to_user(optval
, &sctp_sk(sk
)->subscribe
, len
))
4273 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4275 * This socket option is applicable to the UDP-style socket only. When
4276 * set it will cause associations that are idle for more than the
4277 * specified number of seconds to automatically close. An association
4278 * being idle is defined an association that has NOT sent or received
4279 * user data. The special value of '0' indicates that no automatic
4280 * close of any associations should be performed. The option expects an
4281 * integer defining the number of seconds of idle time before an
4282 * association is closed.
4284 static int sctp_getsockopt_autoclose(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4286 /* Applicable to UDP-style socket only */
4287 if (sctp_style(sk
, TCP
))
4289 if (len
< sizeof(int))
4292 if (put_user(len
, optlen
))
4294 if (copy_to_user(optval
, &sctp_sk(sk
)->autoclose
, sizeof(int)))
4299 /* Helper routine to branch off an association to a new socket. */
4300 int sctp_do_peeloff(struct sock
*sk
, sctp_assoc_t id
, struct socket
**sockp
)
4302 struct sctp_association
*asoc
= sctp_id2assoc(sk
, id
);
4303 struct socket
*sock
;
4310 /* An association cannot be branched off from an already peeled-off
4311 * socket, nor is this supported for tcp style sockets.
4313 if (!sctp_style(sk
, UDP
))
4316 /* Create a new socket. */
4317 err
= sock_create(sk
->sk_family
, SOCK_SEQPACKET
, IPPROTO_SCTP
, &sock
);
4321 sctp_copy_sock(sock
->sk
, sk
, asoc
);
4323 /* Make peeled-off sockets more like 1-1 accepted sockets.
4324 * Set the daddr and initialize id to something more random
4326 af
= sctp_get_af_specific(asoc
->peer
.primary_addr
.sa
.sa_family
);
4327 af
->to_sk_daddr(&asoc
->peer
.primary_addr
, sk
);
4329 /* Populate the fields of the newsk from the oldsk and migrate the
4330 * asoc to the newsk.
4332 sctp_sock_migrate(sk
, sock
->sk
, asoc
, SCTP_SOCKET_UDP_HIGH_BANDWIDTH
);
4338 EXPORT_SYMBOL(sctp_do_peeloff
);
4340 static int sctp_getsockopt_peeloff(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4342 sctp_peeloff_arg_t peeloff
;
4343 struct socket
*newsock
;
4344 struct file
*newfile
;
4347 if (len
< sizeof(sctp_peeloff_arg_t
))
4349 len
= sizeof(sctp_peeloff_arg_t
);
4350 if (copy_from_user(&peeloff
, optval
, len
))
4353 retval
= sctp_do_peeloff(sk
, peeloff
.associd
, &newsock
);
4357 /* Map the socket to an unused fd that can be returned to the user. */
4358 retval
= get_unused_fd_flags(0);
4360 sock_release(newsock
);
4364 newfile
= sock_alloc_file(newsock
, 0, NULL
);
4365 if (unlikely(IS_ERR(newfile
))) {
4366 put_unused_fd(retval
);
4367 sock_release(newsock
);
4368 return PTR_ERR(newfile
);
4371 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__
, sk
, newsock
->sk
,
4374 /* Return the fd mapped to the new socket. */
4375 if (put_user(len
, optlen
)) {
4377 put_unused_fd(retval
);
4380 peeloff
.sd
= retval
;
4381 if (copy_to_user(optval
, &peeloff
, len
)) {
4383 put_unused_fd(retval
);
4386 fd_install(retval
, newfile
);
4391 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4393 * Applications can enable or disable heartbeats for any peer address of
4394 * an association, modify an address's heartbeat interval, force a
4395 * heartbeat to be sent immediately, and adjust the address's maximum
4396 * number of retransmissions sent before an address is considered
4397 * unreachable. The following structure is used to access and modify an
4398 * address's parameters:
4400 * struct sctp_paddrparams {
4401 * sctp_assoc_t spp_assoc_id;
4402 * struct sockaddr_storage spp_address;
4403 * uint32_t spp_hbinterval;
4404 * uint16_t spp_pathmaxrxt;
4405 * uint32_t spp_pathmtu;
4406 * uint32_t spp_sackdelay;
4407 * uint32_t spp_flags;
4410 * spp_assoc_id - (one-to-many style socket) This is filled in the
4411 * application, and identifies the association for
4413 * spp_address - This specifies which address is of interest.
4414 * spp_hbinterval - This contains the value of the heartbeat interval,
4415 * in milliseconds. If a value of zero
4416 * is present in this field then no changes are to
4417 * be made to this parameter.
4418 * spp_pathmaxrxt - This contains the maximum number of
4419 * retransmissions before this address shall be
4420 * considered unreachable. If a value of zero
4421 * is present in this field then no changes are to
4422 * be made to this parameter.
4423 * spp_pathmtu - When Path MTU discovery is disabled the value
4424 * specified here will be the "fixed" path mtu.
4425 * Note that if the spp_address field is empty
4426 * then all associations on this address will
4427 * have this fixed path mtu set upon them.
4429 * spp_sackdelay - When delayed sack is enabled, this value specifies
4430 * the number of milliseconds that sacks will be delayed
4431 * for. This value will apply to all addresses of an
4432 * association if the spp_address field is empty. Note
4433 * also, that if delayed sack is enabled and this
4434 * value is set to 0, no change is made to the last
4435 * recorded delayed sack timer value.
4437 * spp_flags - These flags are used to control various features
4438 * on an association. The flag field may contain
4439 * zero or more of the following options.
4441 * SPP_HB_ENABLE - Enable heartbeats on the
4442 * specified address. Note that if the address
4443 * field is empty all addresses for the association
4444 * have heartbeats enabled upon them.
4446 * SPP_HB_DISABLE - Disable heartbeats on the
4447 * speicifed address. Note that if the address
4448 * field is empty all addresses for the association
4449 * will have their heartbeats disabled. Note also
4450 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4451 * mutually exclusive, only one of these two should
4452 * be specified. Enabling both fields will have
4453 * undetermined results.
4455 * SPP_HB_DEMAND - Request a user initiated heartbeat
4456 * to be made immediately.
4458 * SPP_PMTUD_ENABLE - This field will enable PMTU
4459 * discovery upon the specified address. Note that
4460 * if the address feild is empty then all addresses
4461 * on the association are effected.
4463 * SPP_PMTUD_DISABLE - This field will disable PMTU
4464 * discovery upon the specified address. Note that
4465 * if the address feild is empty then all addresses
4466 * on the association are effected. Not also that
4467 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4468 * exclusive. Enabling both will have undetermined
4471 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4472 * on delayed sack. The time specified in spp_sackdelay
4473 * is used to specify the sack delay for this address. Note
4474 * that if spp_address is empty then all addresses will
4475 * enable delayed sack and take on the sack delay
4476 * value specified in spp_sackdelay.
4477 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4478 * off delayed sack. If the spp_address field is blank then
4479 * delayed sack is disabled for the entire association. Note
4480 * also that this field is mutually exclusive to
4481 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4484 static int sctp_getsockopt_peer_addr_params(struct sock
*sk
, int len
,
4485 char __user
*optval
, int __user
*optlen
)
4487 struct sctp_paddrparams params
;
4488 struct sctp_transport
*trans
= NULL
;
4489 struct sctp_association
*asoc
= NULL
;
4490 struct sctp_sock
*sp
= sctp_sk(sk
);
4492 if (len
< sizeof(struct sctp_paddrparams
))
4494 len
= sizeof(struct sctp_paddrparams
);
4495 if (copy_from_user(¶ms
, optval
, len
))
4498 /* If an address other than INADDR_ANY is specified, and
4499 * no transport is found, then the request is invalid.
4501 if (!sctp_is_any(sk
, (union sctp_addr
*)¶ms
.spp_address
)) {
4502 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
4503 params
.spp_assoc_id
);
4505 pr_debug("%s: failed no transport\n", __func__
);
4510 /* Get association, if assoc_id != 0 and the socket is a one
4511 * to many style socket, and an association was not found, then
4512 * the id was invalid.
4514 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
4515 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
)) {
4516 pr_debug("%s: failed no association\n", __func__
);
4521 /* Fetch transport values. */
4522 params
.spp_hbinterval
= jiffies_to_msecs(trans
->hbinterval
);
4523 params
.spp_pathmtu
= trans
->pathmtu
;
4524 params
.spp_pathmaxrxt
= trans
->pathmaxrxt
;
4525 params
.spp_sackdelay
= jiffies_to_msecs(trans
->sackdelay
);
4527 /*draft-11 doesn't say what to return in spp_flags*/
4528 params
.spp_flags
= trans
->param_flags
;
4530 /* Fetch association values. */
4531 params
.spp_hbinterval
= jiffies_to_msecs(asoc
->hbinterval
);
4532 params
.spp_pathmtu
= asoc
->pathmtu
;
4533 params
.spp_pathmaxrxt
= asoc
->pathmaxrxt
;
4534 params
.spp_sackdelay
= jiffies_to_msecs(asoc
->sackdelay
);
4536 /*draft-11 doesn't say what to return in spp_flags*/
4537 params
.spp_flags
= asoc
->param_flags
;
4539 /* Fetch socket values. */
4540 params
.spp_hbinterval
= sp
->hbinterval
;
4541 params
.spp_pathmtu
= sp
->pathmtu
;
4542 params
.spp_sackdelay
= sp
->sackdelay
;
4543 params
.spp_pathmaxrxt
= sp
->pathmaxrxt
;
4545 /*draft-11 doesn't say what to return in spp_flags*/
4546 params
.spp_flags
= sp
->param_flags
;
4549 if (copy_to_user(optval
, ¶ms
, len
))
4552 if (put_user(len
, optlen
))
4559 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4561 * This option will effect the way delayed acks are performed. This
4562 * option allows you to get or set the delayed ack time, in
4563 * milliseconds. It also allows changing the delayed ack frequency.
4564 * Changing the frequency to 1 disables the delayed sack algorithm. If
4565 * the assoc_id is 0, then this sets or gets the endpoints default
4566 * values. If the assoc_id field is non-zero, then the set or get
4567 * effects the specified association for the one to many model (the
4568 * assoc_id field is ignored by the one to one model). Note that if
4569 * sack_delay or sack_freq are 0 when setting this option, then the
4570 * current values will remain unchanged.
4572 * struct sctp_sack_info {
4573 * sctp_assoc_t sack_assoc_id;
4574 * uint32_t sack_delay;
4575 * uint32_t sack_freq;
4578 * sack_assoc_id - This parameter, indicates which association the user
4579 * is performing an action upon. Note that if this field's value is
4580 * zero then the endpoints default value is changed (effecting future
4581 * associations only).
4583 * sack_delay - This parameter contains the number of milliseconds that
4584 * the user is requesting the delayed ACK timer be set to. Note that
4585 * this value is defined in the standard to be between 200 and 500
4588 * sack_freq - This parameter contains the number of packets that must
4589 * be received before a sack is sent without waiting for the delay
4590 * timer to expire. The default value for this is 2, setting this
4591 * value to 1 will disable the delayed sack algorithm.
4593 static int sctp_getsockopt_delayed_ack(struct sock
*sk
, int len
,
4594 char __user
*optval
,
4597 struct sctp_sack_info params
;
4598 struct sctp_association
*asoc
= NULL
;
4599 struct sctp_sock
*sp
= sctp_sk(sk
);
4601 if (len
>= sizeof(struct sctp_sack_info
)) {
4602 len
= sizeof(struct sctp_sack_info
);
4604 if (copy_from_user(¶ms
, optval
, len
))
4606 } else if (len
== sizeof(struct sctp_assoc_value
)) {
4607 pr_warn_ratelimited(DEPRECATED
4609 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
4610 "Use struct sctp_sack_info instead\n",
4611 current
->comm
, task_pid_nr(current
));
4612 if (copy_from_user(¶ms
, optval
, len
))
4617 /* Get association, if sack_assoc_id != 0 and the socket is a one
4618 * to many style socket, and an association was not found, then
4619 * the id was invalid.
4621 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
4622 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
4626 /* Fetch association values. */
4627 if (asoc
->param_flags
& SPP_SACKDELAY_ENABLE
) {
4628 params
.sack_delay
= jiffies_to_msecs(
4630 params
.sack_freq
= asoc
->sackfreq
;
4633 params
.sack_delay
= 0;
4634 params
.sack_freq
= 1;
4637 /* Fetch socket values. */
4638 if (sp
->param_flags
& SPP_SACKDELAY_ENABLE
) {
4639 params
.sack_delay
= sp
->sackdelay
;
4640 params
.sack_freq
= sp
->sackfreq
;
4642 params
.sack_delay
= 0;
4643 params
.sack_freq
= 1;
4647 if (copy_to_user(optval
, ¶ms
, len
))
4650 if (put_user(len
, optlen
))
4656 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4658 * Applications can specify protocol parameters for the default association
4659 * initialization. The option name argument to setsockopt() and getsockopt()
4662 * Setting initialization parameters is effective only on an unconnected
4663 * socket (for UDP-style sockets only future associations are effected
4664 * by the change). With TCP-style sockets, this option is inherited by
4665 * sockets derived from a listener socket.
4667 static int sctp_getsockopt_initmsg(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4669 if (len
< sizeof(struct sctp_initmsg
))
4671 len
= sizeof(struct sctp_initmsg
);
4672 if (put_user(len
, optlen
))
4674 if (copy_to_user(optval
, &sctp_sk(sk
)->initmsg
, len
))
4680 static int sctp_getsockopt_peer_addrs(struct sock
*sk
, int len
,
4681 char __user
*optval
, int __user
*optlen
)
4683 struct sctp_association
*asoc
;
4685 struct sctp_getaddrs getaddrs
;
4686 struct sctp_transport
*from
;
4688 union sctp_addr temp
;
4689 struct sctp_sock
*sp
= sctp_sk(sk
);
4694 if (len
< sizeof(struct sctp_getaddrs
))
4697 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
4700 /* For UDP-style sockets, id specifies the association to query. */
4701 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
4705 to
= optval
+ offsetof(struct sctp_getaddrs
, addrs
);
4706 space_left
= len
- offsetof(struct sctp_getaddrs
, addrs
);
4708 list_for_each_entry(from
, &asoc
->peer
.transport_addr_list
,
4710 memcpy(&temp
, &from
->ipaddr
, sizeof(temp
));
4711 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
4712 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4713 if (space_left
< addrlen
)
4715 if (copy_to_user(to
, &temp
, addrlen
))
4719 space_left
-= addrlen
;
4722 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
))
4724 bytes_copied
= ((char __user
*)to
) - optval
;
4725 if (put_user(bytes_copied
, optlen
))
4731 static int sctp_copy_laddrs(struct sock
*sk
, __u16 port
, void *to
,
4732 size_t space_left
, int *bytes_copied
)
4734 struct sctp_sockaddr_entry
*addr
;
4735 union sctp_addr temp
;
4738 struct net
*net
= sock_net(sk
);
4741 list_for_each_entry_rcu(addr
, &net
->sctp
.local_addr_list
, list
) {
4745 if ((PF_INET
== sk
->sk_family
) &&
4746 (AF_INET6
== addr
->a
.sa
.sa_family
))
4748 if ((PF_INET6
== sk
->sk_family
) &&
4749 inet_v6_ipv6only(sk
) &&
4750 (AF_INET
== addr
->a
.sa
.sa_family
))
4752 memcpy(&temp
, &addr
->a
, sizeof(temp
));
4753 if (!temp
.v4
.sin_port
)
4754 temp
.v4
.sin_port
= htons(port
);
4756 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
4758 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4759 if (space_left
< addrlen
) {
4763 memcpy(to
, &temp
, addrlen
);
4767 space_left
-= addrlen
;
4768 *bytes_copied
+= addrlen
;
4776 static int sctp_getsockopt_local_addrs(struct sock
*sk
, int len
,
4777 char __user
*optval
, int __user
*optlen
)
4779 struct sctp_bind_addr
*bp
;
4780 struct sctp_association
*asoc
;
4782 struct sctp_getaddrs getaddrs
;
4783 struct sctp_sockaddr_entry
*addr
;
4785 union sctp_addr temp
;
4786 struct sctp_sock
*sp
= sctp_sk(sk
);
4790 int bytes_copied
= 0;
4794 if (len
< sizeof(struct sctp_getaddrs
))
4797 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
4801 * For UDP-style sockets, id specifies the association to query.
4802 * If the id field is set to the value '0' then the locally bound
4803 * addresses are returned without regard to any particular
4806 if (0 == getaddrs
.assoc_id
) {
4807 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
4809 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
4812 bp
= &asoc
->base
.bind_addr
;
4815 to
= optval
+ offsetof(struct sctp_getaddrs
, addrs
);
4816 space_left
= len
- offsetof(struct sctp_getaddrs
, addrs
);
4818 addrs
= kmalloc(space_left
, GFP_KERNEL
);
4822 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4823 * addresses from the global local address list.
4825 if (sctp_list_single_entry(&bp
->address_list
)) {
4826 addr
= list_entry(bp
->address_list
.next
,
4827 struct sctp_sockaddr_entry
, list
);
4828 if (sctp_is_any(sk
, &addr
->a
)) {
4829 cnt
= sctp_copy_laddrs(sk
, bp
->port
, addrs
,
4830 space_left
, &bytes_copied
);
4840 /* Protection on the bound address list is not needed since
4841 * in the socket option context we hold a socket lock and
4842 * thus the bound address list can't change.
4844 list_for_each_entry(addr
, &bp
->address_list
, list
) {
4845 memcpy(&temp
, &addr
->a
, sizeof(temp
));
4846 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
4847 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4848 if (space_left
< addrlen
) {
4849 err
= -ENOMEM
; /*fixme: right error?*/
4852 memcpy(buf
, &temp
, addrlen
);
4854 bytes_copied
+= addrlen
;
4856 space_left
-= addrlen
;
4860 if (copy_to_user(to
, addrs
, bytes_copied
)) {
4864 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
)) {
4868 if (put_user(bytes_copied
, optlen
))
4875 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4877 * Requests that the local SCTP stack use the enclosed peer address as
4878 * the association primary. The enclosed address must be one of the
4879 * association peer's addresses.
4881 static int sctp_getsockopt_primary_addr(struct sock
*sk
, int len
,
4882 char __user
*optval
, int __user
*optlen
)
4884 struct sctp_prim prim
;
4885 struct sctp_association
*asoc
;
4886 struct sctp_sock
*sp
= sctp_sk(sk
);
4888 if (len
< sizeof(struct sctp_prim
))
4891 len
= sizeof(struct sctp_prim
);
4893 if (copy_from_user(&prim
, optval
, len
))
4896 asoc
= sctp_id2assoc(sk
, prim
.ssp_assoc_id
);
4900 if (!asoc
->peer
.primary_path
)
4903 memcpy(&prim
.ssp_addr
, &asoc
->peer
.primary_path
->ipaddr
,
4904 asoc
->peer
.primary_path
->af_specific
->sockaddr_len
);
4906 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
,
4907 (union sctp_addr
*)&prim
.ssp_addr
);
4909 if (put_user(len
, optlen
))
4911 if (copy_to_user(optval
, &prim
, len
))
4918 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4920 * Requests that the local endpoint set the specified Adaptation Layer
4921 * Indication parameter for all future INIT and INIT-ACK exchanges.
4923 static int sctp_getsockopt_adaptation_layer(struct sock
*sk
, int len
,
4924 char __user
*optval
, int __user
*optlen
)
4926 struct sctp_setadaptation adaptation
;
4928 if (len
< sizeof(struct sctp_setadaptation
))
4931 len
= sizeof(struct sctp_setadaptation
);
4933 adaptation
.ssb_adaptation_ind
= sctp_sk(sk
)->adaptation_ind
;
4935 if (put_user(len
, optlen
))
4937 if (copy_to_user(optval
, &adaptation
, len
))
4945 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4947 * Applications that wish to use the sendto() system call may wish to
4948 * specify a default set of parameters that would normally be supplied
4949 * through the inclusion of ancillary data. This socket option allows
4950 * such an application to set the default sctp_sndrcvinfo structure.
4953 * The application that wishes to use this socket option simply passes
4954 * in to this call the sctp_sndrcvinfo structure defined in Section
4955 * 5.2.2) The input parameters accepted by this call include
4956 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4957 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4958 * to this call if the caller is using the UDP model.
4960 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4962 static int sctp_getsockopt_default_send_param(struct sock
*sk
,
4963 int len
, char __user
*optval
,
4966 struct sctp_sndrcvinfo info
;
4967 struct sctp_association
*asoc
;
4968 struct sctp_sock
*sp
= sctp_sk(sk
);
4970 if (len
< sizeof(struct sctp_sndrcvinfo
))
4973 len
= sizeof(struct sctp_sndrcvinfo
);
4975 if (copy_from_user(&info
, optval
, len
))
4978 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
4979 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
4983 info
.sinfo_stream
= asoc
->default_stream
;
4984 info
.sinfo_flags
= asoc
->default_flags
;
4985 info
.sinfo_ppid
= asoc
->default_ppid
;
4986 info
.sinfo_context
= asoc
->default_context
;
4987 info
.sinfo_timetolive
= asoc
->default_timetolive
;
4989 info
.sinfo_stream
= sp
->default_stream
;
4990 info
.sinfo_flags
= sp
->default_flags
;
4991 info
.sinfo_ppid
= sp
->default_ppid
;
4992 info
.sinfo_context
= sp
->default_context
;
4993 info
.sinfo_timetolive
= sp
->default_timetolive
;
4996 if (put_user(len
, optlen
))
4998 if (copy_to_user(optval
, &info
, len
))
5006 * 7.1.5 SCTP_NODELAY
5008 * Turn on/off any Nagle-like algorithm. This means that packets are
5009 * generally sent as soon as possible and no unnecessary delays are
5010 * introduced, at the cost of more packets in the network. Expects an
5011 * integer boolean flag.
5014 static int sctp_getsockopt_nodelay(struct sock
*sk
, int len
,
5015 char __user
*optval
, int __user
*optlen
)
5019 if (len
< sizeof(int))
5023 val
= (sctp_sk(sk
)->nodelay
== 1);
5024 if (put_user(len
, optlen
))
5026 if (copy_to_user(optval
, &val
, len
))
5033 * 7.1.1 SCTP_RTOINFO
5035 * The protocol parameters used to initialize and bound retransmission
5036 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5037 * and modify these parameters.
5038 * All parameters are time values, in milliseconds. A value of 0, when
5039 * modifying the parameters, indicates that the current value should not
5043 static int sctp_getsockopt_rtoinfo(struct sock
*sk
, int len
,
5044 char __user
*optval
,
5045 int __user
*optlen
) {
5046 struct sctp_rtoinfo rtoinfo
;
5047 struct sctp_association
*asoc
;
5049 if (len
< sizeof (struct sctp_rtoinfo
))
5052 len
= sizeof(struct sctp_rtoinfo
);
5054 if (copy_from_user(&rtoinfo
, optval
, len
))
5057 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
5059 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
5062 /* Values corresponding to the specific association. */
5064 rtoinfo
.srto_initial
= jiffies_to_msecs(asoc
->rto_initial
);
5065 rtoinfo
.srto_max
= jiffies_to_msecs(asoc
->rto_max
);
5066 rtoinfo
.srto_min
= jiffies_to_msecs(asoc
->rto_min
);
5068 /* Values corresponding to the endpoint. */
5069 struct sctp_sock
*sp
= sctp_sk(sk
);
5071 rtoinfo
.srto_initial
= sp
->rtoinfo
.srto_initial
;
5072 rtoinfo
.srto_max
= sp
->rtoinfo
.srto_max
;
5073 rtoinfo
.srto_min
= sp
->rtoinfo
.srto_min
;
5076 if (put_user(len
, optlen
))
5079 if (copy_to_user(optval
, &rtoinfo
, len
))
5087 * 7.1.2 SCTP_ASSOCINFO
5089 * This option is used to tune the maximum retransmission attempts
5090 * of the association.
5091 * Returns an error if the new association retransmission value is
5092 * greater than the sum of the retransmission value of the peer.
5093 * See [SCTP] for more information.
5096 static int sctp_getsockopt_associnfo(struct sock
*sk
, int len
,
5097 char __user
*optval
,
5101 struct sctp_assocparams assocparams
;
5102 struct sctp_association
*asoc
;
5103 struct list_head
*pos
;
5106 if (len
< sizeof (struct sctp_assocparams
))
5109 len
= sizeof(struct sctp_assocparams
);
5111 if (copy_from_user(&assocparams
, optval
, len
))
5114 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
5116 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
5119 /* Values correspoinding to the specific association */
5121 assocparams
.sasoc_asocmaxrxt
= asoc
->max_retrans
;
5122 assocparams
.sasoc_peer_rwnd
= asoc
->peer
.rwnd
;
5123 assocparams
.sasoc_local_rwnd
= asoc
->a_rwnd
;
5124 assocparams
.sasoc_cookie_life
= ktime_to_ms(asoc
->cookie_life
);
5126 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
5130 assocparams
.sasoc_number_peer_destinations
= cnt
;
5132 /* Values corresponding to the endpoint */
5133 struct sctp_sock
*sp
= sctp_sk(sk
);
5135 assocparams
.sasoc_asocmaxrxt
= sp
->assocparams
.sasoc_asocmaxrxt
;
5136 assocparams
.sasoc_peer_rwnd
= sp
->assocparams
.sasoc_peer_rwnd
;
5137 assocparams
.sasoc_local_rwnd
= sp
->assocparams
.sasoc_local_rwnd
;
5138 assocparams
.sasoc_cookie_life
=
5139 sp
->assocparams
.sasoc_cookie_life
;
5140 assocparams
.sasoc_number_peer_destinations
=
5142 sasoc_number_peer_destinations
;
5145 if (put_user(len
, optlen
))
5148 if (copy_to_user(optval
, &assocparams
, len
))
5155 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5157 * This socket option is a boolean flag which turns on or off mapped V4
5158 * addresses. If this option is turned on and the socket is type
5159 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5160 * If this option is turned off, then no mapping will be done of V4
5161 * addresses and a user will receive both PF_INET6 and PF_INET type
5162 * addresses on the socket.
5164 static int sctp_getsockopt_mappedv4(struct sock
*sk
, int len
,
5165 char __user
*optval
, int __user
*optlen
)
5168 struct sctp_sock
*sp
= sctp_sk(sk
);
5170 if (len
< sizeof(int))
5175 if (put_user(len
, optlen
))
5177 if (copy_to_user(optval
, &val
, len
))
5184 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5185 * (chapter and verse is quoted at sctp_setsockopt_context())
5187 static int sctp_getsockopt_context(struct sock
*sk
, int len
,
5188 char __user
*optval
, int __user
*optlen
)
5190 struct sctp_assoc_value params
;
5191 struct sctp_sock
*sp
;
5192 struct sctp_association
*asoc
;
5194 if (len
< sizeof(struct sctp_assoc_value
))
5197 len
= sizeof(struct sctp_assoc_value
);
5199 if (copy_from_user(¶ms
, optval
, len
))
5204 if (params
.assoc_id
!= 0) {
5205 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5208 params
.assoc_value
= asoc
->default_rcv_context
;
5210 params
.assoc_value
= sp
->default_rcv_context
;
5213 if (put_user(len
, optlen
))
5215 if (copy_to_user(optval
, ¶ms
, len
))
5222 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5223 * This option will get or set the maximum size to put in any outgoing
5224 * SCTP DATA chunk. If a message is larger than this size it will be
5225 * fragmented by SCTP into the specified size. Note that the underlying
5226 * SCTP implementation may fragment into smaller sized chunks when the
5227 * PMTU of the underlying association is smaller than the value set by
5228 * the user. The default value for this option is '0' which indicates
5229 * the user is NOT limiting fragmentation and only the PMTU will effect
5230 * SCTP's choice of DATA chunk size. Note also that values set larger
5231 * than the maximum size of an IP datagram will effectively let SCTP
5232 * control fragmentation (i.e. the same as setting this option to 0).
5234 * The following structure is used to access and modify this parameter:
5236 * struct sctp_assoc_value {
5237 * sctp_assoc_t assoc_id;
5238 * uint32_t assoc_value;
5241 * assoc_id: This parameter is ignored for one-to-one style sockets.
5242 * For one-to-many style sockets this parameter indicates which
5243 * association the user is performing an action upon. Note that if
5244 * this field's value is zero then the endpoints default value is
5245 * changed (effecting future associations only).
5246 * assoc_value: This parameter specifies the maximum size in bytes.
5248 static int sctp_getsockopt_maxseg(struct sock
*sk
, int len
,
5249 char __user
*optval
, int __user
*optlen
)
5251 struct sctp_assoc_value params
;
5252 struct sctp_association
*asoc
;
5254 if (len
== sizeof(int)) {
5255 pr_warn_ratelimited(DEPRECATED
5257 "Use of int in maxseg socket option.\n"
5258 "Use struct sctp_assoc_value instead\n",
5259 current
->comm
, task_pid_nr(current
));
5260 params
.assoc_id
= 0;
5261 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5262 len
= sizeof(struct sctp_assoc_value
);
5263 if (copy_from_user(¶ms
, optval
, sizeof(params
)))
5268 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5269 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
5273 params
.assoc_value
= asoc
->frag_point
;
5275 params
.assoc_value
= sctp_sk(sk
)->user_frag
;
5277 if (put_user(len
, optlen
))
5279 if (len
== sizeof(int)) {
5280 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5283 if (copy_to_user(optval
, ¶ms
, len
))
5291 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5292 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5294 static int sctp_getsockopt_fragment_interleave(struct sock
*sk
, int len
,
5295 char __user
*optval
, int __user
*optlen
)
5299 if (len
< sizeof(int))
5304 val
= sctp_sk(sk
)->frag_interleave
;
5305 if (put_user(len
, optlen
))
5307 if (copy_to_user(optval
, &val
, len
))
5314 * 7.1.25. Set or Get the sctp partial delivery point
5315 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5317 static int sctp_getsockopt_partial_delivery_point(struct sock
*sk
, int len
,
5318 char __user
*optval
,
5323 if (len
< sizeof(u32
))
5328 val
= sctp_sk(sk
)->pd_point
;
5329 if (put_user(len
, optlen
))
5331 if (copy_to_user(optval
, &val
, len
))
5338 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5339 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5341 static int sctp_getsockopt_maxburst(struct sock
*sk
, int len
,
5342 char __user
*optval
,
5345 struct sctp_assoc_value params
;
5346 struct sctp_sock
*sp
;
5347 struct sctp_association
*asoc
;
5349 if (len
== sizeof(int)) {
5350 pr_warn_ratelimited(DEPRECATED
5352 "Use of int in max_burst socket option.\n"
5353 "Use struct sctp_assoc_value instead\n",
5354 current
->comm
, task_pid_nr(current
));
5355 params
.assoc_id
= 0;
5356 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5357 len
= sizeof(struct sctp_assoc_value
);
5358 if (copy_from_user(¶ms
, optval
, len
))
5365 if (params
.assoc_id
!= 0) {
5366 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5369 params
.assoc_value
= asoc
->max_burst
;
5371 params
.assoc_value
= sp
->max_burst
;
5373 if (len
== sizeof(int)) {
5374 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5377 if (copy_to_user(optval
, ¶ms
, len
))
5385 static int sctp_getsockopt_hmac_ident(struct sock
*sk
, int len
,
5386 char __user
*optval
, int __user
*optlen
)
5388 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5389 struct sctp_hmacalgo __user
*p
= (void __user
*)optval
;
5390 struct sctp_hmac_algo_param
*hmacs
;
5394 if (!ep
->auth_enable
)
5397 hmacs
= ep
->auth_hmacs_list
;
5398 data_len
= ntohs(hmacs
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5400 if (len
< sizeof(struct sctp_hmacalgo
) + data_len
)
5403 len
= sizeof(struct sctp_hmacalgo
) + data_len
;
5404 num_idents
= data_len
/ sizeof(u16
);
5406 if (put_user(len
, optlen
))
5408 if (put_user(num_idents
, &p
->shmac_num_idents
))
5410 if (copy_to_user(p
->shmac_idents
, hmacs
->hmac_ids
, data_len
))
5415 static int sctp_getsockopt_active_key(struct sock
*sk
, int len
,
5416 char __user
*optval
, int __user
*optlen
)
5418 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5419 struct sctp_authkeyid val
;
5420 struct sctp_association
*asoc
;
5422 if (!ep
->auth_enable
)
5425 if (len
< sizeof(struct sctp_authkeyid
))
5427 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authkeyid
)))
5430 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
5431 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
5435 val
.scact_keynumber
= asoc
->active_key_id
;
5437 val
.scact_keynumber
= ep
->active_key_id
;
5439 len
= sizeof(struct sctp_authkeyid
);
5440 if (put_user(len
, optlen
))
5442 if (copy_to_user(optval
, &val
, len
))
5448 static int sctp_getsockopt_peer_auth_chunks(struct sock
*sk
, int len
,
5449 char __user
*optval
, int __user
*optlen
)
5451 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5452 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5453 struct sctp_authchunks val
;
5454 struct sctp_association
*asoc
;
5455 struct sctp_chunks_param
*ch
;
5459 if (!ep
->auth_enable
)
5462 if (len
< sizeof(struct sctp_authchunks
))
5465 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5468 to
= p
->gauth_chunks
;
5469 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5473 ch
= asoc
->peer
.peer_chunks
;
5477 /* See if the user provided enough room for all the data */
5478 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5479 if (len
< num_chunks
)
5482 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
5485 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
5486 if (put_user(len
, optlen
))
5488 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
5493 static int sctp_getsockopt_local_auth_chunks(struct sock
*sk
, int len
,
5494 char __user
*optval
, int __user
*optlen
)
5496 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5497 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5498 struct sctp_authchunks val
;
5499 struct sctp_association
*asoc
;
5500 struct sctp_chunks_param
*ch
;
5504 if (!ep
->auth_enable
)
5507 if (len
< sizeof(struct sctp_authchunks
))
5510 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5513 to
= p
->gauth_chunks
;
5514 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5515 if (!asoc
&& val
.gauth_assoc_id
&& sctp_style(sk
, UDP
))
5519 ch
= (struct sctp_chunks_param
*)asoc
->c
.auth_chunks
;
5521 ch
= ep
->auth_chunk_list
;
5526 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5527 if (len
< sizeof(struct sctp_authchunks
) + num_chunks
)
5530 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
5533 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
5534 if (put_user(len
, optlen
))
5536 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
5543 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5544 * This option gets the current number of associations that are attached
5545 * to a one-to-many style socket. The option value is an uint32_t.
5547 static int sctp_getsockopt_assoc_number(struct sock
*sk
, int len
,
5548 char __user
*optval
, int __user
*optlen
)
5550 struct sctp_sock
*sp
= sctp_sk(sk
);
5551 struct sctp_association
*asoc
;
5554 if (sctp_style(sk
, TCP
))
5557 if (len
< sizeof(u32
))
5562 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5566 if (put_user(len
, optlen
))
5568 if (copy_to_user(optval
, &val
, len
))
5575 * 8.1.23 SCTP_AUTO_ASCONF
5576 * See the corresponding setsockopt entry as description
5578 static int sctp_getsockopt_auto_asconf(struct sock
*sk
, int len
,
5579 char __user
*optval
, int __user
*optlen
)
5583 if (len
< sizeof(int))
5587 if (sctp_sk(sk
)->do_auto_asconf
&& sctp_is_ep_boundall(sk
))
5589 if (put_user(len
, optlen
))
5591 if (copy_to_user(optval
, &val
, len
))
5597 * 8.2.6. Get the Current Identifiers of Associations
5598 * (SCTP_GET_ASSOC_ID_LIST)
5600 * This option gets the current list of SCTP association identifiers of
5601 * the SCTP associations handled by a one-to-many style socket.
5603 static int sctp_getsockopt_assoc_ids(struct sock
*sk
, int len
,
5604 char __user
*optval
, int __user
*optlen
)
5606 struct sctp_sock
*sp
= sctp_sk(sk
);
5607 struct sctp_association
*asoc
;
5608 struct sctp_assoc_ids
*ids
;
5611 if (sctp_style(sk
, TCP
))
5614 if (len
< sizeof(struct sctp_assoc_ids
))
5617 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5621 if (len
< sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
)
5624 len
= sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
;
5626 ids
= kmalloc(len
, GFP_KERNEL
);
5630 ids
->gaids_number_of_ids
= num
;
5632 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5633 ids
->gaids_assoc_id
[num
++] = asoc
->assoc_id
;
5636 if (put_user(len
, optlen
) || copy_to_user(optval
, ids
, len
)) {
5646 * SCTP_PEER_ADDR_THLDS
5648 * This option allows us to fetch the partially failed threshold for one or all
5649 * transports in an association. See Section 6.1 of:
5650 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5652 static int sctp_getsockopt_paddr_thresholds(struct sock
*sk
,
5653 char __user
*optval
,
5657 struct sctp_paddrthlds val
;
5658 struct sctp_transport
*trans
;
5659 struct sctp_association
*asoc
;
5661 if (len
< sizeof(struct sctp_paddrthlds
))
5663 len
= sizeof(struct sctp_paddrthlds
);
5664 if (copy_from_user(&val
, (struct sctp_paddrthlds __user
*)optval
, len
))
5667 if (sctp_is_any(sk
, (const union sctp_addr
*)&val
.spt_address
)) {
5668 asoc
= sctp_id2assoc(sk
, val
.spt_assoc_id
);
5672 val
.spt_pathpfthld
= asoc
->pf_retrans
;
5673 val
.spt_pathmaxrxt
= asoc
->pathmaxrxt
;
5675 trans
= sctp_addr_id2transport(sk
, &val
.spt_address
,
5680 val
.spt_pathmaxrxt
= trans
->pathmaxrxt
;
5681 val
.spt_pathpfthld
= trans
->pf_retrans
;
5684 if (put_user(len
, optlen
) || copy_to_user(optval
, &val
, len
))
5691 * SCTP_GET_ASSOC_STATS
5693 * This option retrieves local per endpoint statistics. It is modeled
5694 * after OpenSolaris' implementation
5696 static int sctp_getsockopt_assoc_stats(struct sock
*sk
, int len
,
5697 char __user
*optval
,
5700 struct sctp_assoc_stats sas
;
5701 struct sctp_association
*asoc
= NULL
;
5703 /* User must provide at least the assoc id */
5704 if (len
< sizeof(sctp_assoc_t
))
5707 /* Allow the struct to grow and fill in as much as possible */
5708 len
= min_t(size_t, len
, sizeof(sas
));
5710 if (copy_from_user(&sas
, optval
, len
))
5713 asoc
= sctp_id2assoc(sk
, sas
.sas_assoc_id
);
5717 sas
.sas_rtxchunks
= asoc
->stats
.rtxchunks
;
5718 sas
.sas_gapcnt
= asoc
->stats
.gapcnt
;
5719 sas
.sas_outofseqtsns
= asoc
->stats
.outofseqtsns
;
5720 sas
.sas_osacks
= asoc
->stats
.osacks
;
5721 sas
.sas_isacks
= asoc
->stats
.isacks
;
5722 sas
.sas_octrlchunks
= asoc
->stats
.octrlchunks
;
5723 sas
.sas_ictrlchunks
= asoc
->stats
.ictrlchunks
;
5724 sas
.sas_oodchunks
= asoc
->stats
.oodchunks
;
5725 sas
.sas_iodchunks
= asoc
->stats
.iodchunks
;
5726 sas
.sas_ouodchunks
= asoc
->stats
.ouodchunks
;
5727 sas
.sas_iuodchunks
= asoc
->stats
.iuodchunks
;
5728 sas
.sas_idupchunks
= asoc
->stats
.idupchunks
;
5729 sas
.sas_opackets
= asoc
->stats
.opackets
;
5730 sas
.sas_ipackets
= asoc
->stats
.ipackets
;
5732 /* New high max rto observed, will return 0 if not a single
5733 * RTO update took place. obs_rto_ipaddr will be bogus
5736 sas
.sas_maxrto
= asoc
->stats
.max_obs_rto
;
5737 memcpy(&sas
.sas_obs_rto_ipaddr
, &asoc
->stats
.obs_rto_ipaddr
,
5738 sizeof(struct sockaddr_storage
));
5740 /* Mark beginning of a new observation period */
5741 asoc
->stats
.max_obs_rto
= asoc
->rto_min
;
5743 if (put_user(len
, optlen
))
5746 pr_debug("%s: len:%d, assoc_id:%d\n", __func__
, len
, sas
.sas_assoc_id
);
5748 if (copy_to_user(optval
, &sas
, len
))
5754 static int sctp_getsockopt(struct sock
*sk
, int level
, int optname
,
5755 char __user
*optval
, int __user
*optlen
)
5760 pr_debug("%s: sk:%p, optname:%d\n", __func__
, sk
, optname
);
5762 /* I can hardly begin to describe how wrong this is. This is
5763 * so broken as to be worse than useless. The API draft
5764 * REALLY is NOT helpful here... I am not convinced that the
5765 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5766 * are at all well-founded.
5768 if (level
!= SOL_SCTP
) {
5769 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
5771 retval
= af
->getsockopt(sk
, level
, optname
, optval
, optlen
);
5775 if (get_user(len
, optlen
))
5782 retval
= sctp_getsockopt_sctp_status(sk
, len
, optval
, optlen
);
5784 case SCTP_DISABLE_FRAGMENTS
:
5785 retval
= sctp_getsockopt_disable_fragments(sk
, len
, optval
,
5789 retval
= sctp_getsockopt_events(sk
, len
, optval
, optlen
);
5791 case SCTP_AUTOCLOSE
:
5792 retval
= sctp_getsockopt_autoclose(sk
, len
, optval
, optlen
);
5794 case SCTP_SOCKOPT_PEELOFF
:
5795 retval
= sctp_getsockopt_peeloff(sk
, len
, optval
, optlen
);
5797 case SCTP_PEER_ADDR_PARAMS
:
5798 retval
= sctp_getsockopt_peer_addr_params(sk
, len
, optval
,
5801 case SCTP_DELAYED_SACK
:
5802 retval
= sctp_getsockopt_delayed_ack(sk
, len
, optval
,
5806 retval
= sctp_getsockopt_initmsg(sk
, len
, optval
, optlen
);
5808 case SCTP_GET_PEER_ADDRS
:
5809 retval
= sctp_getsockopt_peer_addrs(sk
, len
, optval
,
5812 case SCTP_GET_LOCAL_ADDRS
:
5813 retval
= sctp_getsockopt_local_addrs(sk
, len
, optval
,
5816 case SCTP_SOCKOPT_CONNECTX3
:
5817 retval
= sctp_getsockopt_connectx3(sk
, len
, optval
, optlen
);
5819 case SCTP_DEFAULT_SEND_PARAM
:
5820 retval
= sctp_getsockopt_default_send_param(sk
, len
,
5823 case SCTP_PRIMARY_ADDR
:
5824 retval
= sctp_getsockopt_primary_addr(sk
, len
, optval
, optlen
);
5827 retval
= sctp_getsockopt_nodelay(sk
, len
, optval
, optlen
);
5830 retval
= sctp_getsockopt_rtoinfo(sk
, len
, optval
, optlen
);
5832 case SCTP_ASSOCINFO
:
5833 retval
= sctp_getsockopt_associnfo(sk
, len
, optval
, optlen
);
5835 case SCTP_I_WANT_MAPPED_V4_ADDR
:
5836 retval
= sctp_getsockopt_mappedv4(sk
, len
, optval
, optlen
);
5839 retval
= sctp_getsockopt_maxseg(sk
, len
, optval
, optlen
);
5841 case SCTP_GET_PEER_ADDR_INFO
:
5842 retval
= sctp_getsockopt_peer_addr_info(sk
, len
, optval
,
5845 case SCTP_ADAPTATION_LAYER
:
5846 retval
= sctp_getsockopt_adaptation_layer(sk
, len
, optval
,
5850 retval
= sctp_getsockopt_context(sk
, len
, optval
, optlen
);
5852 case SCTP_FRAGMENT_INTERLEAVE
:
5853 retval
= sctp_getsockopt_fragment_interleave(sk
, len
, optval
,
5856 case SCTP_PARTIAL_DELIVERY_POINT
:
5857 retval
= sctp_getsockopt_partial_delivery_point(sk
, len
, optval
,
5860 case SCTP_MAX_BURST
:
5861 retval
= sctp_getsockopt_maxburst(sk
, len
, optval
, optlen
);
5864 case SCTP_AUTH_CHUNK
:
5865 case SCTP_AUTH_DELETE_KEY
:
5866 retval
= -EOPNOTSUPP
;
5868 case SCTP_HMAC_IDENT
:
5869 retval
= sctp_getsockopt_hmac_ident(sk
, len
, optval
, optlen
);
5871 case SCTP_AUTH_ACTIVE_KEY
:
5872 retval
= sctp_getsockopt_active_key(sk
, len
, optval
, optlen
);
5874 case SCTP_PEER_AUTH_CHUNKS
:
5875 retval
= sctp_getsockopt_peer_auth_chunks(sk
, len
, optval
,
5878 case SCTP_LOCAL_AUTH_CHUNKS
:
5879 retval
= sctp_getsockopt_local_auth_chunks(sk
, len
, optval
,
5882 case SCTP_GET_ASSOC_NUMBER
:
5883 retval
= sctp_getsockopt_assoc_number(sk
, len
, optval
, optlen
);
5885 case SCTP_GET_ASSOC_ID_LIST
:
5886 retval
= sctp_getsockopt_assoc_ids(sk
, len
, optval
, optlen
);
5888 case SCTP_AUTO_ASCONF
:
5889 retval
= sctp_getsockopt_auto_asconf(sk
, len
, optval
, optlen
);
5891 case SCTP_PEER_ADDR_THLDS
:
5892 retval
= sctp_getsockopt_paddr_thresholds(sk
, optval
, len
, optlen
);
5894 case SCTP_GET_ASSOC_STATS
:
5895 retval
= sctp_getsockopt_assoc_stats(sk
, len
, optval
, optlen
);
5898 retval
= -ENOPROTOOPT
;
5906 static void sctp_hash(struct sock
*sk
)
5911 static void sctp_unhash(struct sock
*sk
)
5916 /* Check if port is acceptable. Possibly find first available port.
5918 * The port hash table (contained in the 'global' SCTP protocol storage
5919 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5920 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5921 * list (the list number is the port number hashed out, so as you
5922 * would expect from a hash function, all the ports in a given list have
5923 * such a number that hashes out to the same list number; you were
5924 * expecting that, right?); so each list has a set of ports, with a
5925 * link to the socket (struct sock) that uses it, the port number and
5926 * a fastreuse flag (FIXME: NPI ipg).
5928 static struct sctp_bind_bucket
*sctp_bucket_create(
5929 struct sctp_bind_hashbucket
*head
, struct net
*, unsigned short snum
);
5931 static long sctp_get_port_local(struct sock
*sk
, union sctp_addr
*addr
)
5933 struct sctp_bind_hashbucket
*head
; /* hash list */
5934 struct sctp_bind_bucket
*pp
;
5935 unsigned short snum
;
5938 snum
= ntohs(addr
->v4
.sin_port
);
5940 pr_debug("%s: begins, snum:%d\n", __func__
, snum
);
5945 /* Search for an available port. */
5946 int low
, high
, remaining
, index
;
5949 inet_get_local_port_range(sock_net(sk
), &low
, &high
);
5950 remaining
= (high
- low
) + 1;
5951 rover
= prandom_u32() % remaining
+ low
;
5955 if ((rover
< low
) || (rover
> high
))
5957 if (inet_is_reserved_local_port(rover
))
5959 index
= sctp_phashfn(sock_net(sk
), rover
);
5960 head
= &sctp_port_hashtable
[index
];
5961 spin_lock(&head
->lock
);
5962 sctp_for_each_hentry(pp
, &head
->chain
)
5963 if ((pp
->port
== rover
) &&
5964 net_eq(sock_net(sk
), pp
->net
))
5968 spin_unlock(&head
->lock
);
5969 } while (--remaining
> 0);
5971 /* Exhausted local port range during search? */
5976 /* OK, here is the one we will use. HEAD (the port
5977 * hash table list entry) is non-NULL and we hold it's
5982 /* We are given an specific port number; we verify
5983 * that it is not being used. If it is used, we will
5984 * exahust the search in the hash list corresponding
5985 * to the port number (snum) - we detect that with the
5986 * port iterator, pp being NULL.
5988 head
= &sctp_port_hashtable
[sctp_phashfn(sock_net(sk
), snum
)];
5989 spin_lock(&head
->lock
);
5990 sctp_for_each_hentry(pp
, &head
->chain
) {
5991 if ((pp
->port
== snum
) && net_eq(pp
->net
, sock_net(sk
)))
5998 if (!hlist_empty(&pp
->owner
)) {
5999 /* We had a port hash table hit - there is an
6000 * available port (pp != NULL) and it is being
6001 * used by other socket (pp->owner not empty); that other
6002 * socket is going to be sk2.
6004 int reuse
= sk
->sk_reuse
;
6007 pr_debug("%s: found a possible match\n", __func__
);
6009 if (pp
->fastreuse
&& sk
->sk_reuse
&&
6010 sk
->sk_state
!= SCTP_SS_LISTENING
)
6013 /* Run through the list of sockets bound to the port
6014 * (pp->port) [via the pointers bind_next and
6015 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6016 * we get the endpoint they describe and run through
6017 * the endpoint's list of IP (v4 or v6) addresses,
6018 * comparing each of the addresses with the address of
6019 * the socket sk. If we find a match, then that means
6020 * that this port/socket (sk) combination are already
6023 sk_for_each_bound(sk2
, &pp
->owner
) {
6024 struct sctp_endpoint
*ep2
;
6025 ep2
= sctp_sk(sk2
)->ep
;
6028 (reuse
&& sk2
->sk_reuse
&&
6029 sk2
->sk_state
!= SCTP_SS_LISTENING
))
6032 if (sctp_bind_addr_conflict(&ep2
->base
.bind_addr
, addr
,
6033 sctp_sk(sk2
), sctp_sk(sk
))) {
6039 pr_debug("%s: found a match\n", __func__
);
6042 /* If there was a hash table miss, create a new port. */
6044 if (!pp
&& !(pp
= sctp_bucket_create(head
, sock_net(sk
), snum
)))
6047 /* In either case (hit or miss), make sure fastreuse is 1 only
6048 * if sk->sk_reuse is too (that is, if the caller requested
6049 * SO_REUSEADDR on this socket -sk-).
6051 if (hlist_empty(&pp
->owner
)) {
6052 if (sk
->sk_reuse
&& sk
->sk_state
!= SCTP_SS_LISTENING
)
6056 } else if (pp
->fastreuse
&&
6057 (!sk
->sk_reuse
|| sk
->sk_state
== SCTP_SS_LISTENING
))
6060 /* We are set, so fill up all the data in the hash table
6061 * entry, tie the socket list information with the rest of the
6062 * sockets FIXME: Blurry, NPI (ipg).
6065 if (!sctp_sk(sk
)->bind_hash
) {
6066 inet_sk(sk
)->inet_num
= snum
;
6067 sk_add_bind_node(sk
, &pp
->owner
);
6068 sctp_sk(sk
)->bind_hash
= pp
;
6073 spin_unlock(&head
->lock
);
6080 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6081 * port is requested.
6083 static int sctp_get_port(struct sock
*sk
, unsigned short snum
)
6085 union sctp_addr addr
;
6086 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
6088 /* Set up a dummy address struct from the sk. */
6089 af
->from_sk(&addr
, sk
);
6090 addr
.v4
.sin_port
= htons(snum
);
6092 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6093 return !!sctp_get_port_local(sk
, &addr
);
6097 * Move a socket to LISTENING state.
6099 static int sctp_listen_start(struct sock
*sk
, int backlog
)
6101 struct sctp_sock
*sp
= sctp_sk(sk
);
6102 struct sctp_endpoint
*ep
= sp
->ep
;
6103 struct crypto_hash
*tfm
= NULL
;
6106 /* Allocate HMAC for generating cookie. */
6107 if (!sp
->hmac
&& sp
->sctp_hmac_alg
) {
6108 sprintf(alg
, "hmac(%s)", sp
->sctp_hmac_alg
);
6109 tfm
= crypto_alloc_hash(alg
, 0, CRYPTO_ALG_ASYNC
);
6111 net_info_ratelimited("failed to load transform for %s: %ld\n",
6112 sp
->sctp_hmac_alg
, PTR_ERR(tfm
));
6115 sctp_sk(sk
)->hmac
= tfm
;
6119 * If a bind() or sctp_bindx() is not called prior to a listen()
6120 * call that allows new associations to be accepted, the system
6121 * picks an ephemeral port and will choose an address set equivalent
6122 * to binding with a wildcard address.
6124 * This is not currently spelled out in the SCTP sockets
6125 * extensions draft, but follows the practice as seen in TCP
6129 sk
->sk_state
= SCTP_SS_LISTENING
;
6130 if (!ep
->base
.bind_addr
.port
) {
6131 if (sctp_autobind(sk
))
6134 if (sctp_get_port(sk
, inet_sk(sk
)->inet_num
)) {
6135 sk
->sk_state
= SCTP_SS_CLOSED
;
6140 sk
->sk_max_ack_backlog
= backlog
;
6141 sctp_hash_endpoint(ep
);
6146 * 4.1.3 / 5.1.3 listen()
6148 * By default, new associations are not accepted for UDP style sockets.
6149 * An application uses listen() to mark a socket as being able to
6150 * accept new associations.
6152 * On TCP style sockets, applications use listen() to ready the SCTP
6153 * endpoint for accepting inbound associations.
6155 * On both types of endpoints a backlog of '0' disables listening.
6157 * Move a socket to LISTENING state.
6159 int sctp_inet_listen(struct socket
*sock
, int backlog
)
6161 struct sock
*sk
= sock
->sk
;
6162 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
6165 if (unlikely(backlog
< 0))
6170 /* Peeled-off sockets are not allowed to listen(). */
6171 if (sctp_style(sk
, UDP_HIGH_BANDWIDTH
))
6174 if (sock
->state
!= SS_UNCONNECTED
)
6177 /* If backlog is zero, disable listening. */
6179 if (sctp_sstate(sk
, CLOSED
))
6183 sctp_unhash_endpoint(ep
);
6184 sk
->sk_state
= SCTP_SS_CLOSED
;
6186 sctp_sk(sk
)->bind_hash
->fastreuse
= 1;
6190 /* If we are already listening, just update the backlog */
6191 if (sctp_sstate(sk
, LISTENING
))
6192 sk
->sk_max_ack_backlog
= backlog
;
6194 err
= sctp_listen_start(sk
, backlog
);
6206 * This function is done by modeling the current datagram_poll() and the
6207 * tcp_poll(). Note that, based on these implementations, we don't
6208 * lock the socket in this function, even though it seems that,
6209 * ideally, locking or some other mechanisms can be used to ensure
6210 * the integrity of the counters (sndbuf and wmem_alloc) used
6211 * in this place. We assume that we don't need locks either until proven
6214 * Another thing to note is that we include the Async I/O support
6215 * here, again, by modeling the current TCP/UDP code. We don't have
6216 * a good way to test with it yet.
6218 unsigned int sctp_poll(struct file
*file
, struct socket
*sock
, poll_table
*wait
)
6220 struct sock
*sk
= sock
->sk
;
6221 struct sctp_sock
*sp
= sctp_sk(sk
);
6224 poll_wait(file
, sk_sleep(sk
), wait
);
6226 /* A TCP-style listening socket becomes readable when the accept queue
6229 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
6230 return (!list_empty(&sp
->ep
->asocs
)) ?
6231 (POLLIN
| POLLRDNORM
) : 0;
6235 /* Is there any exceptional events? */
6236 if (sk
->sk_err
|| !skb_queue_empty(&sk
->sk_error_queue
))
6238 (sock_flag(sk
, SOCK_SELECT_ERR_QUEUE
) ? POLLPRI
: 0);
6239 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6240 mask
|= POLLRDHUP
| POLLIN
| POLLRDNORM
;
6241 if (sk
->sk_shutdown
== SHUTDOWN_MASK
)
6244 /* Is it readable? Reconsider this code with TCP-style support. */
6245 if (!skb_queue_empty(&sk
->sk_receive_queue
))
6246 mask
|= POLLIN
| POLLRDNORM
;
6248 /* The association is either gone or not ready. */
6249 if (!sctp_style(sk
, UDP
) && sctp_sstate(sk
, CLOSED
))
6252 /* Is it writable? */
6253 if (sctp_writeable(sk
)) {
6254 mask
|= POLLOUT
| POLLWRNORM
;
6256 set_bit(SOCK_ASYNC_NOSPACE
, &sk
->sk_socket
->flags
);
6258 * Since the socket is not locked, the buffer
6259 * might be made available after the writeable check and
6260 * before the bit is set. This could cause a lost I/O
6261 * signal. tcp_poll() has a race breaker for this race
6262 * condition. Based on their implementation, we put
6263 * in the following code to cover it as well.
6265 if (sctp_writeable(sk
))
6266 mask
|= POLLOUT
| POLLWRNORM
;
6271 /********************************************************************
6272 * 2nd Level Abstractions
6273 ********************************************************************/
6275 static struct sctp_bind_bucket
*sctp_bucket_create(
6276 struct sctp_bind_hashbucket
*head
, struct net
*net
, unsigned short snum
)
6278 struct sctp_bind_bucket
*pp
;
6280 pp
= kmem_cache_alloc(sctp_bucket_cachep
, GFP_ATOMIC
);
6282 SCTP_DBG_OBJCNT_INC(bind_bucket
);
6285 INIT_HLIST_HEAD(&pp
->owner
);
6287 hlist_add_head(&pp
->node
, &head
->chain
);
6292 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6293 static void sctp_bucket_destroy(struct sctp_bind_bucket
*pp
)
6295 if (pp
&& hlist_empty(&pp
->owner
)) {
6296 __hlist_del(&pp
->node
);
6297 kmem_cache_free(sctp_bucket_cachep
, pp
);
6298 SCTP_DBG_OBJCNT_DEC(bind_bucket
);
6302 /* Release this socket's reference to a local port. */
6303 static inline void __sctp_put_port(struct sock
*sk
)
6305 struct sctp_bind_hashbucket
*head
=
6306 &sctp_port_hashtable
[sctp_phashfn(sock_net(sk
),
6307 inet_sk(sk
)->inet_num
)];
6308 struct sctp_bind_bucket
*pp
;
6310 spin_lock(&head
->lock
);
6311 pp
= sctp_sk(sk
)->bind_hash
;
6312 __sk_del_bind_node(sk
);
6313 sctp_sk(sk
)->bind_hash
= NULL
;
6314 inet_sk(sk
)->inet_num
= 0;
6315 sctp_bucket_destroy(pp
);
6316 spin_unlock(&head
->lock
);
6319 void sctp_put_port(struct sock
*sk
)
6322 __sctp_put_port(sk
);
6327 * The system picks an ephemeral port and choose an address set equivalent
6328 * to binding with a wildcard address.
6329 * One of those addresses will be the primary address for the association.
6330 * This automatically enables the multihoming capability of SCTP.
6332 static int sctp_autobind(struct sock
*sk
)
6334 union sctp_addr autoaddr
;
6338 /* Initialize a local sockaddr structure to INADDR_ANY. */
6339 af
= sctp_sk(sk
)->pf
->af
;
6341 port
= htons(inet_sk(sk
)->inet_num
);
6342 af
->inaddr_any(&autoaddr
, port
);
6344 return sctp_do_bind(sk
, &autoaddr
, af
->sockaddr_len
);
6347 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6350 * 4.2 The cmsghdr Structure *
6352 * When ancillary data is sent or received, any number of ancillary data
6353 * objects can be specified by the msg_control and msg_controllen members of
6354 * the msghdr structure, because each object is preceded by
6355 * a cmsghdr structure defining the object's length (the cmsg_len member).
6356 * Historically Berkeley-derived implementations have passed only one object
6357 * at a time, but this API allows multiple objects to be
6358 * passed in a single call to sendmsg() or recvmsg(). The following example
6359 * shows two ancillary data objects in a control buffer.
6361 * |<--------------------------- msg_controllen -------------------------->|
6364 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6366 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6369 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6371 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6374 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6375 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6377 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6379 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6386 static int sctp_msghdr_parse(const struct msghdr
*msg
, sctp_cmsgs_t
*cmsgs
)
6388 struct cmsghdr
*cmsg
;
6389 struct msghdr
*my_msg
= (struct msghdr
*)msg
;
6391 for (cmsg
= CMSG_FIRSTHDR(msg
);
6393 cmsg
= CMSG_NXTHDR(my_msg
, cmsg
)) {
6394 if (!CMSG_OK(my_msg
, cmsg
))
6397 /* Should we parse this header or ignore? */
6398 if (cmsg
->cmsg_level
!= IPPROTO_SCTP
)
6401 /* Strictly check lengths following example in SCM code. */
6402 switch (cmsg
->cmsg_type
) {
6404 /* SCTP Socket API Extension
6405 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6407 * This cmsghdr structure provides information for
6408 * initializing new SCTP associations with sendmsg().
6409 * The SCTP_INITMSG socket option uses this same data
6410 * structure. This structure is not used for
6413 * cmsg_level cmsg_type cmsg_data[]
6414 * ------------ ------------ ----------------------
6415 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6417 if (cmsg
->cmsg_len
!=
6418 CMSG_LEN(sizeof(struct sctp_initmsg
)))
6420 cmsgs
->init
= (struct sctp_initmsg
*)CMSG_DATA(cmsg
);
6424 /* SCTP Socket API Extension
6425 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6427 * This cmsghdr structure specifies SCTP options for
6428 * sendmsg() and describes SCTP header information
6429 * about a received message through recvmsg().
6431 * cmsg_level cmsg_type cmsg_data[]
6432 * ------------ ------------ ----------------------
6433 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6435 if (cmsg
->cmsg_len
!=
6436 CMSG_LEN(sizeof(struct sctp_sndrcvinfo
)))
6440 (struct sctp_sndrcvinfo
*)CMSG_DATA(cmsg
);
6442 /* Minimally, validate the sinfo_flags. */
6443 if (cmsgs
->info
->sinfo_flags
&
6444 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
6445 SCTP_ABORT
| SCTP_EOF
))
6457 * Wait for a packet..
6458 * Note: This function is the same function as in core/datagram.c
6459 * with a few modifications to make lksctp work.
6461 static int sctp_wait_for_packet(struct sock
*sk
, int *err
, long *timeo_p
)
6466 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
6468 /* Socket errors? */
6469 error
= sock_error(sk
);
6473 if (!skb_queue_empty(&sk
->sk_receive_queue
))
6476 /* Socket shut down? */
6477 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6480 /* Sequenced packets can come disconnected. If so we report the
6485 /* Is there a good reason to think that we may receive some data? */
6486 if (list_empty(&sctp_sk(sk
)->ep
->asocs
) && !sctp_sstate(sk
, LISTENING
))
6489 /* Handle signals. */
6490 if (signal_pending(current
))
6493 /* Let another process have a go. Since we are going to sleep
6494 * anyway. Note: This may cause odd behaviors if the message
6495 * does not fit in the user's buffer, but this seems to be the
6496 * only way to honor MSG_DONTWAIT realistically.
6499 *timeo_p
= schedule_timeout(*timeo_p
);
6503 finish_wait(sk_sleep(sk
), &wait
);
6507 error
= sock_intr_errno(*timeo_p
);
6510 finish_wait(sk_sleep(sk
), &wait
);
6515 /* Receive a datagram.
6516 * Note: This is pretty much the same routine as in core/datagram.c
6517 * with a few changes to make lksctp work.
6519 static struct sk_buff
*sctp_skb_recv_datagram(struct sock
*sk
, int flags
,
6520 int noblock
, int *err
)
6523 struct sk_buff
*skb
;
6526 timeo
= sock_rcvtimeo(sk
, noblock
);
6528 pr_debug("%s: timeo:%ld, max:%ld\n", __func__
, timeo
,
6529 MAX_SCHEDULE_TIMEOUT
);
6532 /* Again only user level code calls this function,
6533 * so nothing interrupt level
6534 * will suddenly eat the receive_queue.
6536 * Look at current nfs client by the way...
6537 * However, this function was correct in any case. 8)
6539 if (flags
& MSG_PEEK
) {
6540 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
6541 skb
= skb_peek(&sk
->sk_receive_queue
);
6543 atomic_inc(&skb
->users
);
6544 spin_unlock_bh(&sk
->sk_receive_queue
.lock
);
6546 skb
= skb_dequeue(&sk
->sk_receive_queue
);
6552 /* Caller is allowed not to check sk->sk_err before calling. */
6553 error
= sock_error(sk
);
6557 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6560 /* User doesn't want to wait. */
6564 } while (sctp_wait_for_packet(sk
, err
, &timeo
) == 0);
6573 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6574 static void __sctp_write_space(struct sctp_association
*asoc
)
6576 struct sock
*sk
= asoc
->base
.sk
;
6577 struct socket
*sock
= sk
->sk_socket
;
6579 if ((sctp_wspace(asoc
) > 0) && sock
) {
6580 if (waitqueue_active(&asoc
->wait
))
6581 wake_up_interruptible(&asoc
->wait
);
6583 if (sctp_writeable(sk
)) {
6584 wait_queue_head_t
*wq
= sk_sleep(sk
);
6586 if (wq
&& waitqueue_active(wq
))
6587 wake_up_interruptible(wq
);
6589 /* Note that we try to include the Async I/O support
6590 * here by modeling from the current TCP/UDP code.
6591 * We have not tested with it yet.
6593 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
))
6594 sock_wake_async(sock
,
6595 SOCK_WAKE_SPACE
, POLL_OUT
);
6600 static void sctp_wake_up_waiters(struct sock
*sk
,
6601 struct sctp_association
*asoc
)
6603 struct sctp_association
*tmp
= asoc
;
6605 /* We do accounting for the sndbuf space per association,
6606 * so we only need to wake our own association.
6608 if (asoc
->ep
->sndbuf_policy
)
6609 return __sctp_write_space(asoc
);
6611 /* If association goes down and is just flushing its
6612 * outq, then just normally notify others.
6614 if (asoc
->base
.dead
)
6615 return sctp_write_space(sk
);
6617 /* Accounting for the sndbuf space is per socket, so we
6618 * need to wake up others, try to be fair and in case of
6619 * other associations, let them have a go first instead
6620 * of just doing a sctp_write_space() call.
6622 * Note that we reach sctp_wake_up_waiters() only when
6623 * associations free up queued chunks, thus we are under
6624 * lock and the list of associations on a socket is
6625 * guaranteed not to change.
6627 for (tmp
= list_next_entry(tmp
, asocs
); 1;
6628 tmp
= list_next_entry(tmp
, asocs
)) {
6629 /* Manually skip the head element. */
6630 if (&tmp
->asocs
== &((sctp_sk(sk
))->ep
->asocs
))
6632 /* Wake up association. */
6633 __sctp_write_space(tmp
);
6634 /* We've reached the end. */
6640 /* Do accounting for the sndbuf space.
6641 * Decrement the used sndbuf space of the corresponding association by the
6642 * data size which was just transmitted(freed).
6644 static void sctp_wfree(struct sk_buff
*skb
)
6646 struct sctp_association
*asoc
;
6647 struct sctp_chunk
*chunk
;
6650 /* Get the saved chunk pointer. */
6651 chunk
= *((struct sctp_chunk
**)(skb
->cb
));
6654 asoc
->sndbuf_used
-= SCTP_DATA_SNDSIZE(chunk
) +
6655 sizeof(struct sk_buff
) +
6656 sizeof(struct sctp_chunk
);
6658 atomic_sub(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
6661 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6663 sk
->sk_wmem_queued
-= skb
->truesize
;
6664 sk_mem_uncharge(sk
, skb
->truesize
);
6667 sctp_wake_up_waiters(sk
, asoc
);
6669 sctp_association_put(asoc
);
6672 /* Do accounting for the receive space on the socket.
6673 * Accounting for the association is done in ulpevent.c
6674 * We set this as a destructor for the cloned data skbs so that
6675 * accounting is done at the correct time.
6677 void sctp_sock_rfree(struct sk_buff
*skb
)
6679 struct sock
*sk
= skb
->sk
;
6680 struct sctp_ulpevent
*event
= sctp_skb2event(skb
);
6682 atomic_sub(event
->rmem_len
, &sk
->sk_rmem_alloc
);
6685 * Mimic the behavior of sock_rfree
6687 sk_mem_uncharge(sk
, event
->rmem_len
);
6691 /* Helper function to wait for space in the sndbuf. */
6692 static int sctp_wait_for_sndbuf(struct sctp_association
*asoc
, long *timeo_p
,
6695 struct sock
*sk
= asoc
->base
.sk
;
6697 long current_timeo
= *timeo_p
;
6700 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__
, asoc
,
6703 /* Increment the association's refcnt. */
6704 sctp_association_hold(asoc
);
6706 /* Wait on the association specific sndbuf space. */
6708 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
6709 TASK_INTERRUPTIBLE
);
6712 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
6715 if (signal_pending(current
))
6716 goto do_interrupted
;
6717 if (msg_len
<= sctp_wspace(asoc
))
6720 /* Let another process have a go. Since we are going
6724 current_timeo
= schedule_timeout(current_timeo
);
6725 BUG_ON(sk
!= asoc
->base
.sk
);
6728 *timeo_p
= current_timeo
;
6732 finish_wait(&asoc
->wait
, &wait
);
6734 /* Release the association's refcnt. */
6735 sctp_association_put(asoc
);
6744 err
= sock_intr_errno(*timeo_p
);
6752 void sctp_data_ready(struct sock
*sk
)
6754 struct socket_wq
*wq
;
6757 wq
= rcu_dereference(sk
->sk_wq
);
6758 if (wq_has_sleeper(wq
))
6759 wake_up_interruptible_sync_poll(&wq
->wait
, POLLIN
|
6760 POLLRDNORM
| POLLRDBAND
);
6761 sk_wake_async(sk
, SOCK_WAKE_WAITD
, POLL_IN
);
6765 /* If socket sndbuf has changed, wake up all per association waiters. */
6766 void sctp_write_space(struct sock
*sk
)
6768 struct sctp_association
*asoc
;
6770 /* Wake up the tasks in each wait queue. */
6771 list_for_each_entry(asoc
, &((sctp_sk(sk
))->ep
->asocs
), asocs
) {
6772 __sctp_write_space(asoc
);
6776 /* Is there any sndbuf space available on the socket?
6778 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6779 * associations on the same socket. For a UDP-style socket with
6780 * multiple associations, it is possible for it to be "unwriteable"
6781 * prematurely. I assume that this is acceptable because
6782 * a premature "unwriteable" is better than an accidental "writeable" which
6783 * would cause an unwanted block under certain circumstances. For the 1-1
6784 * UDP-style sockets or TCP-style sockets, this code should work.
6787 static int sctp_writeable(struct sock
*sk
)
6791 amt
= sk
->sk_sndbuf
- sk_wmem_alloc_get(sk
);
6797 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6798 * returns immediately with EINPROGRESS.
6800 static int sctp_wait_for_connect(struct sctp_association
*asoc
, long *timeo_p
)
6802 struct sock
*sk
= asoc
->base
.sk
;
6804 long current_timeo
= *timeo_p
;
6807 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__
, asoc
, *timeo_p
);
6809 /* Increment the association's refcnt. */
6810 sctp_association_hold(asoc
);
6813 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
6814 TASK_INTERRUPTIBLE
);
6817 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6819 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
6822 if (signal_pending(current
))
6823 goto do_interrupted
;
6825 if (sctp_state(asoc
, ESTABLISHED
))
6828 /* Let another process have a go. Since we are going
6832 current_timeo
= schedule_timeout(current_timeo
);
6835 *timeo_p
= current_timeo
;
6839 finish_wait(&asoc
->wait
, &wait
);
6841 /* Release the association's refcnt. */
6842 sctp_association_put(asoc
);
6847 if (asoc
->init_err_counter
+ 1 > asoc
->max_init_attempts
)
6850 err
= -ECONNREFUSED
;
6854 err
= sock_intr_errno(*timeo_p
);
6862 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
)
6864 struct sctp_endpoint
*ep
;
6868 ep
= sctp_sk(sk
)->ep
;
6872 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
,
6873 TASK_INTERRUPTIBLE
);
6875 if (list_empty(&ep
->asocs
)) {
6877 timeo
= schedule_timeout(timeo
);
6882 if (!sctp_sstate(sk
, LISTENING
))
6886 if (!list_empty(&ep
->asocs
))
6889 err
= sock_intr_errno(timeo
);
6890 if (signal_pending(current
))
6898 finish_wait(sk_sleep(sk
), &wait
);
6903 static void sctp_wait_for_close(struct sock
*sk
, long timeout
)
6908 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
6909 if (list_empty(&sctp_sk(sk
)->ep
->asocs
))
6912 timeout
= schedule_timeout(timeout
);
6914 } while (!signal_pending(current
) && timeout
);
6916 finish_wait(sk_sleep(sk
), &wait
);
6919 static void sctp_skb_set_owner_r_frag(struct sk_buff
*skb
, struct sock
*sk
)
6921 struct sk_buff
*frag
;
6926 /* Don't forget the fragments. */
6927 skb_walk_frags(skb
, frag
)
6928 sctp_skb_set_owner_r_frag(frag
, sk
);
6931 sctp_skb_set_owner_r(skb
, sk
);
6934 void sctp_copy_sock(struct sock
*newsk
, struct sock
*sk
,
6935 struct sctp_association
*asoc
)
6937 struct inet_sock
*inet
= inet_sk(sk
);
6938 struct inet_sock
*newinet
;
6940 newsk
->sk_type
= sk
->sk_type
;
6941 newsk
->sk_bound_dev_if
= sk
->sk_bound_dev_if
;
6942 newsk
->sk_flags
= sk
->sk_flags
;
6943 newsk
->sk_no_check
= sk
->sk_no_check
;
6944 newsk
->sk_reuse
= sk
->sk_reuse
;
6946 newsk
->sk_shutdown
= sk
->sk_shutdown
;
6947 newsk
->sk_destruct
= sctp_destruct_sock
;
6948 newsk
->sk_family
= sk
->sk_family
;
6949 newsk
->sk_protocol
= IPPROTO_SCTP
;
6950 newsk
->sk_backlog_rcv
= sk
->sk_prot
->backlog_rcv
;
6951 newsk
->sk_sndbuf
= sk
->sk_sndbuf
;
6952 newsk
->sk_rcvbuf
= sk
->sk_rcvbuf
;
6953 newsk
->sk_lingertime
= sk
->sk_lingertime
;
6954 newsk
->sk_rcvtimeo
= sk
->sk_rcvtimeo
;
6955 newsk
->sk_sndtimeo
= sk
->sk_sndtimeo
;
6957 newinet
= inet_sk(newsk
);
6959 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6960 * getsockname() and getpeername()
6962 newinet
->inet_sport
= inet
->inet_sport
;
6963 newinet
->inet_saddr
= inet
->inet_saddr
;
6964 newinet
->inet_rcv_saddr
= inet
->inet_rcv_saddr
;
6965 newinet
->inet_dport
= htons(asoc
->peer
.port
);
6966 newinet
->pmtudisc
= inet
->pmtudisc
;
6967 newinet
->inet_id
= asoc
->next_tsn
^ jiffies
;
6969 newinet
->uc_ttl
= inet
->uc_ttl
;
6970 newinet
->mc_loop
= 1;
6971 newinet
->mc_ttl
= 1;
6972 newinet
->mc_index
= 0;
6973 newinet
->mc_list
= NULL
;
6976 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6977 * and its messages to the newsk.
6979 static void sctp_sock_migrate(struct sock
*oldsk
, struct sock
*newsk
,
6980 struct sctp_association
*assoc
,
6981 sctp_socket_type_t type
)
6983 struct sctp_sock
*oldsp
= sctp_sk(oldsk
);
6984 struct sctp_sock
*newsp
= sctp_sk(newsk
);
6985 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
6986 struct sctp_endpoint
*newep
= newsp
->ep
;
6987 struct sk_buff
*skb
, *tmp
;
6988 struct sctp_ulpevent
*event
;
6989 struct sctp_bind_hashbucket
*head
;
6990 struct list_head tmplist
;
6992 /* Migrate socket buffer sizes and all the socket level options to the
6995 newsk
->sk_sndbuf
= oldsk
->sk_sndbuf
;
6996 newsk
->sk_rcvbuf
= oldsk
->sk_rcvbuf
;
6997 /* Brute force copy old sctp opt. */
6998 if (oldsp
->do_auto_asconf
) {
6999 memcpy(&tmplist
, &newsp
->auto_asconf_list
, sizeof(tmplist
));
7000 inet_sk_copy_descendant(newsk
, oldsk
);
7001 memcpy(&newsp
->auto_asconf_list
, &tmplist
, sizeof(tmplist
));
7003 inet_sk_copy_descendant(newsk
, oldsk
);
7005 /* Restore the ep value that was overwritten with the above structure
7011 /* Hook this new socket in to the bind_hash list. */
7012 head
= &sctp_port_hashtable
[sctp_phashfn(sock_net(oldsk
),
7013 inet_sk(oldsk
)->inet_num
)];
7015 spin_lock(&head
->lock
);
7016 pp
= sctp_sk(oldsk
)->bind_hash
;
7017 sk_add_bind_node(newsk
, &pp
->owner
);
7018 sctp_sk(newsk
)->bind_hash
= pp
;
7019 inet_sk(newsk
)->inet_num
= inet_sk(oldsk
)->inet_num
;
7020 spin_unlock(&head
->lock
);
7023 /* Copy the bind_addr list from the original endpoint to the new
7024 * endpoint so that we can handle restarts properly
7026 sctp_bind_addr_dup(&newsp
->ep
->base
.bind_addr
,
7027 &oldsp
->ep
->base
.bind_addr
, GFP_KERNEL
);
7029 /* Move any messages in the old socket's receive queue that are for the
7030 * peeled off association to the new socket's receive queue.
7032 sctp_skb_for_each(skb
, &oldsk
->sk_receive_queue
, tmp
) {
7033 event
= sctp_skb2event(skb
);
7034 if (event
->asoc
== assoc
) {
7035 __skb_unlink(skb
, &oldsk
->sk_receive_queue
);
7036 __skb_queue_tail(&newsk
->sk_receive_queue
, skb
);
7037 sctp_skb_set_owner_r_frag(skb
, newsk
);
7041 /* Clean up any messages pending delivery due to partial
7042 * delivery. Three cases:
7043 * 1) No partial deliver; no work.
7044 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7045 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7047 skb_queue_head_init(&newsp
->pd_lobby
);
7048 atomic_set(&sctp_sk(newsk
)->pd_mode
, assoc
->ulpq
.pd_mode
);
7050 if (atomic_read(&sctp_sk(oldsk
)->pd_mode
)) {
7051 struct sk_buff_head
*queue
;
7053 /* Decide which queue to move pd_lobby skbs to. */
7054 if (assoc
->ulpq
.pd_mode
) {
7055 queue
= &newsp
->pd_lobby
;
7057 queue
= &newsk
->sk_receive_queue
;
7059 /* Walk through the pd_lobby, looking for skbs that
7060 * need moved to the new socket.
7062 sctp_skb_for_each(skb
, &oldsp
->pd_lobby
, tmp
) {
7063 event
= sctp_skb2event(skb
);
7064 if (event
->asoc
== assoc
) {
7065 __skb_unlink(skb
, &oldsp
->pd_lobby
);
7066 __skb_queue_tail(queue
, skb
);
7067 sctp_skb_set_owner_r_frag(skb
, newsk
);
7071 /* Clear up any skbs waiting for the partial
7072 * delivery to finish.
7074 if (assoc
->ulpq
.pd_mode
)
7075 sctp_clear_pd(oldsk
, NULL
);
7079 sctp_skb_for_each(skb
, &assoc
->ulpq
.reasm
, tmp
)
7080 sctp_skb_set_owner_r_frag(skb
, newsk
);
7082 sctp_skb_for_each(skb
, &assoc
->ulpq
.lobby
, tmp
)
7083 sctp_skb_set_owner_r_frag(skb
, newsk
);
7085 /* Set the type of socket to indicate that it is peeled off from the
7086 * original UDP-style socket or created with the accept() call on a
7087 * TCP-style socket..
7091 /* Mark the new socket "in-use" by the user so that any packets
7092 * that may arrive on the association after we've moved it are
7093 * queued to the backlog. This prevents a potential race between
7094 * backlog processing on the old socket and new-packet processing
7095 * on the new socket.
7097 * The caller has just allocated newsk so we can guarantee that other
7098 * paths won't try to lock it and then oldsk.
7100 lock_sock_nested(newsk
, SINGLE_DEPTH_NESTING
);
7101 sctp_assoc_migrate(assoc
, newsk
);
7103 /* If the association on the newsk is already closed before accept()
7104 * is called, set RCV_SHUTDOWN flag.
7106 if (sctp_state(assoc
, CLOSED
) && sctp_style(newsk
, TCP
))
7107 newsk
->sk_shutdown
|= RCV_SHUTDOWN
;
7109 newsk
->sk_state
= SCTP_SS_ESTABLISHED
;
7110 release_sock(newsk
);
7114 /* This proto struct describes the ULP interface for SCTP. */
7115 struct proto sctp_prot
= {
7117 .owner
= THIS_MODULE
,
7118 .close
= sctp_close
,
7119 .connect
= sctp_connect
,
7120 .disconnect
= sctp_disconnect
,
7121 .accept
= sctp_accept
,
7122 .ioctl
= sctp_ioctl
,
7123 .init
= sctp_init_sock
,
7124 .destroy
= sctp_destroy_sock
,
7125 .shutdown
= sctp_shutdown
,
7126 .setsockopt
= sctp_setsockopt
,
7127 .getsockopt
= sctp_getsockopt
,
7128 .sendmsg
= sctp_sendmsg
,
7129 .recvmsg
= sctp_recvmsg
,
7131 .backlog_rcv
= sctp_backlog_rcv
,
7133 .unhash
= sctp_unhash
,
7134 .get_port
= sctp_get_port
,
7135 .obj_size
= sizeof(struct sctp_sock
),
7136 .sysctl_mem
= sysctl_sctp_mem
,
7137 .sysctl_rmem
= sysctl_sctp_rmem
,
7138 .sysctl_wmem
= sysctl_sctp_wmem
,
7139 .memory_pressure
= &sctp_memory_pressure
,
7140 .enter_memory_pressure
= sctp_enter_memory_pressure
,
7141 .memory_allocated
= &sctp_memory_allocated
,
7142 .sockets_allocated
= &sctp_sockets_allocated
,
7145 #if IS_ENABLED(CONFIG_IPV6)
7147 struct proto sctpv6_prot
= {
7149 .owner
= THIS_MODULE
,
7150 .close
= sctp_close
,
7151 .connect
= sctp_connect
,
7152 .disconnect
= sctp_disconnect
,
7153 .accept
= sctp_accept
,
7154 .ioctl
= sctp_ioctl
,
7155 .init
= sctp_init_sock
,
7156 .destroy
= sctp_destroy_sock
,
7157 .shutdown
= sctp_shutdown
,
7158 .setsockopt
= sctp_setsockopt
,
7159 .getsockopt
= sctp_getsockopt
,
7160 .sendmsg
= sctp_sendmsg
,
7161 .recvmsg
= sctp_recvmsg
,
7163 .backlog_rcv
= sctp_backlog_rcv
,
7165 .unhash
= sctp_unhash
,
7166 .get_port
= sctp_get_port
,
7167 .obj_size
= sizeof(struct sctp6_sock
),
7168 .sysctl_mem
= sysctl_sctp_mem
,
7169 .sysctl_rmem
= sysctl_sctp_rmem
,
7170 .sysctl_wmem
= sysctl_sctp_wmem
,
7171 .memory_pressure
= &sctp_memory_pressure
,
7172 .enter_memory_pressure
= sctp_enter_memory_pressure
,
7173 .memory_allocated
= &sctp_memory_allocated
,
7174 .sockets_allocated
= &sctp_sockets_allocated
,
7176 #endif /* IS_ENABLED(CONFIG_IPV6) */