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>
70 #include <net/route.h>
72 #include <net/inet_common.h>
74 #include <linux/socket.h> /* for sa_family_t */
75 #include <linux/export.h>
77 #include <net/sctp/sctp.h>
78 #include <net/sctp/sm.h>
80 /* Forward declarations for internal helper functions. */
81 static int sctp_writeable(struct sock
*sk
);
82 static void sctp_wfree(struct sk_buff
*skb
);
83 static int sctp_wait_for_sndbuf(struct sctp_association
*, long *timeo_p
,
85 static int sctp_wait_for_packet(struct sock
*sk
, int *err
, long *timeo_p
);
86 static int sctp_wait_for_connect(struct sctp_association
*, long *timeo_p
);
87 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
);
88 static void sctp_wait_for_close(struct sock
*sk
, long timeo
);
89 static void sctp_destruct_sock(struct sock
*sk
);
90 static struct sctp_af
*sctp_sockaddr_af(struct sctp_sock
*opt
,
91 union sctp_addr
*addr
, int len
);
92 static int sctp_bindx_add(struct sock
*, struct sockaddr
*, int);
93 static int sctp_bindx_rem(struct sock
*, struct sockaddr
*, int);
94 static int sctp_send_asconf_add_ip(struct sock
*, struct sockaddr
*, int);
95 static int sctp_send_asconf_del_ip(struct sock
*, struct sockaddr
*, int);
96 static int sctp_send_asconf(struct sctp_association
*asoc
,
97 struct sctp_chunk
*chunk
);
98 static int sctp_do_bind(struct sock
*, union sctp_addr
*, int);
99 static int sctp_autobind(struct sock
*sk
);
100 static void sctp_sock_migrate(struct sock
*, struct sock
*,
101 struct sctp_association
*, sctp_socket_type_t
);
103 extern struct kmem_cache
*sctp_bucket_cachep
;
104 extern long sysctl_sctp_mem
[3];
105 extern int sysctl_sctp_rmem
[3];
106 extern int sysctl_sctp_wmem
[3];
108 static int sctp_memory_pressure
;
109 static atomic_long_t sctp_memory_allocated
;
110 struct percpu_counter sctp_sockets_allocated
;
112 static void sctp_enter_memory_pressure(struct sock
*sk
)
114 sctp_memory_pressure
= 1;
118 /* Get the sndbuf space available at the time on the association. */
119 static inline int sctp_wspace(struct sctp_association
*asoc
)
123 if (asoc
->ep
->sndbuf_policy
)
124 amt
= asoc
->sndbuf_used
;
126 amt
= sk_wmem_alloc_get(asoc
->base
.sk
);
128 if (amt
>= asoc
->base
.sk
->sk_sndbuf
) {
129 if (asoc
->base
.sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)
132 amt
= sk_stream_wspace(asoc
->base
.sk
);
137 amt
= asoc
->base
.sk
->sk_sndbuf
- amt
;
142 /* Increment the used sndbuf space count of the corresponding association by
143 * the size of the outgoing data chunk.
144 * Also, set the skb destructor for sndbuf accounting later.
146 * Since it is always 1-1 between chunk and skb, and also a new skb is always
147 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
148 * destructor in the data chunk skb for the purpose of the sndbuf space
151 static inline void sctp_set_owner_w(struct sctp_chunk
*chunk
)
153 struct sctp_association
*asoc
= chunk
->asoc
;
154 struct sock
*sk
= asoc
->base
.sk
;
156 /* The sndbuf space is tracked per association. */
157 sctp_association_hold(asoc
);
159 skb_set_owner_w(chunk
->skb
, sk
);
161 chunk
->skb
->destructor
= sctp_wfree
;
162 /* Save the chunk pointer in skb for sctp_wfree to use later. */
163 *((struct sctp_chunk
**)(chunk
->skb
->cb
)) = chunk
;
165 asoc
->sndbuf_used
+= SCTP_DATA_SNDSIZE(chunk
) +
166 sizeof(struct sk_buff
) +
167 sizeof(struct sctp_chunk
);
169 atomic_add(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
170 sk
->sk_wmem_queued
+= chunk
->skb
->truesize
;
171 sk_mem_charge(sk
, chunk
->skb
->truesize
);
174 /* Verify that this is a valid address. */
175 static inline int sctp_verify_addr(struct sock
*sk
, union sctp_addr
*addr
,
180 /* Verify basic sockaddr. */
181 af
= sctp_sockaddr_af(sctp_sk(sk
), addr
, len
);
185 /* Is this a valid SCTP address? */
186 if (!af
->addr_valid(addr
, sctp_sk(sk
), NULL
))
189 if (!sctp_sk(sk
)->pf
->send_verify(sctp_sk(sk
), (addr
)))
195 /* Look up the association by its id. If this is not a UDP-style
196 * socket, the ID field is always ignored.
198 struct sctp_association
*sctp_id2assoc(struct sock
*sk
, sctp_assoc_t id
)
200 struct sctp_association
*asoc
= NULL
;
202 /* If this is not a UDP-style socket, assoc id should be ignored. */
203 if (!sctp_style(sk
, UDP
)) {
204 /* Return NULL if the socket state is not ESTABLISHED. It
205 * could be a TCP-style listening socket or a socket which
206 * hasn't yet called connect() to establish an association.
208 if (!sctp_sstate(sk
, ESTABLISHED
))
211 /* Get the first and the only association from the list. */
212 if (!list_empty(&sctp_sk(sk
)->ep
->asocs
))
213 asoc
= list_entry(sctp_sk(sk
)->ep
->asocs
.next
,
214 struct sctp_association
, asocs
);
218 /* Otherwise this is a UDP-style socket. */
219 if (!id
|| (id
== (sctp_assoc_t
)-1))
222 spin_lock_bh(&sctp_assocs_id_lock
);
223 asoc
= (struct sctp_association
*)idr_find(&sctp_assocs_id
, (int)id
);
224 spin_unlock_bh(&sctp_assocs_id_lock
);
226 if (!asoc
|| (asoc
->base
.sk
!= sk
) || asoc
->base
.dead
)
232 /* Look up the transport from an address and an assoc id. If both address and
233 * id are specified, the associations matching the address and the id should be
236 static struct sctp_transport
*sctp_addr_id2transport(struct sock
*sk
,
237 struct sockaddr_storage
*addr
,
240 struct sctp_association
*addr_asoc
= NULL
, *id_asoc
= NULL
;
241 struct sctp_transport
*transport
;
242 union sctp_addr
*laddr
= (union sctp_addr
*)addr
;
244 addr_asoc
= sctp_endpoint_lookup_assoc(sctp_sk(sk
)->ep
,
251 id_asoc
= sctp_id2assoc(sk
, id
);
252 if (id_asoc
&& (id_asoc
!= addr_asoc
))
255 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
256 (union sctp_addr
*)addr
);
261 /* API 3.1.2 bind() - UDP Style Syntax
262 * The syntax of bind() is,
264 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
266 * sd - the socket descriptor returned by socket().
267 * addr - the address structure (struct sockaddr_in or struct
268 * sockaddr_in6 [RFC 2553]),
269 * addr_len - the size of the address structure.
271 static int sctp_bind(struct sock
*sk
, struct sockaddr
*addr
, int addr_len
)
277 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__
, sk
,
280 /* Disallow binding twice. */
281 if (!sctp_sk(sk
)->ep
->base
.bind_addr
.port
)
282 retval
= sctp_do_bind(sk
, (union sctp_addr
*)addr
,
292 static long sctp_get_port_local(struct sock
*, union sctp_addr
*);
294 /* Verify this is a valid sockaddr. */
295 static struct sctp_af
*sctp_sockaddr_af(struct sctp_sock
*opt
,
296 union sctp_addr
*addr
, int len
)
300 /* Check minimum size. */
301 if (len
< sizeof (struct sockaddr
))
304 /* V4 mapped address are really of AF_INET family */
305 if (addr
->sa
.sa_family
== AF_INET6
&&
306 ipv6_addr_v4mapped(&addr
->v6
.sin6_addr
)) {
307 if (!opt
->pf
->af_supported(AF_INET
, opt
))
310 /* Does this PF support this AF? */
311 if (!opt
->pf
->af_supported(addr
->sa
.sa_family
, opt
))
315 /* If we get this far, af is valid. */
316 af
= sctp_get_af_specific(addr
->sa
.sa_family
);
318 if (len
< af
->sockaddr_len
)
324 /* Bind a local address either to an endpoint or to an association. */
325 static int sctp_do_bind(struct sock
*sk
, union sctp_addr
*addr
, int len
)
327 struct net
*net
= sock_net(sk
);
328 struct sctp_sock
*sp
= sctp_sk(sk
);
329 struct sctp_endpoint
*ep
= sp
->ep
;
330 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
335 /* Common sockaddr verification. */
336 af
= sctp_sockaddr_af(sp
, addr
, len
);
338 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
339 __func__
, sk
, addr
, len
);
343 snum
= ntohs(addr
->v4
.sin_port
);
345 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
346 __func__
, sk
, &addr
->sa
, bp
->port
, snum
, len
);
348 /* PF specific bind() address verification. */
349 if (!sp
->pf
->bind_verify(sp
, addr
))
350 return -EADDRNOTAVAIL
;
352 /* We must either be unbound, or bind to the same port.
353 * It's OK to allow 0 ports if we are already bound.
354 * We'll just inhert an already bound port in this case
359 else if (snum
!= bp
->port
) {
360 pr_debug("%s: new port %d doesn't match existing port "
361 "%d\n", __func__
, snum
, bp
->port
);
366 if (snum
&& snum
< PROT_SOCK
&&
367 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
))
370 /* See if the address matches any of the addresses we may have
371 * already bound before checking against other endpoints.
373 if (sctp_bind_addr_match(bp
, addr
, sp
))
376 /* Make sure we are allowed to bind here.
377 * The function sctp_get_port_local() does duplicate address
380 addr
->v4
.sin_port
= htons(snum
);
381 if ((ret
= sctp_get_port_local(sk
, addr
))) {
385 /* Refresh ephemeral port. */
387 bp
->port
= inet_sk(sk
)->inet_num
;
389 /* Add the address to the bind address list.
390 * Use GFP_ATOMIC since BHs will be disabled.
392 ret
= sctp_add_bind_addr(bp
, addr
, SCTP_ADDR_SRC
, GFP_ATOMIC
);
394 /* Copy back into socket for getsockname() use. */
396 inet_sk(sk
)->inet_sport
= htons(inet_sk(sk
)->inet_num
);
397 af
->to_sk_saddr(addr
, sk
);
403 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
405 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
406 * at any one time. If a sender, after sending an ASCONF chunk, decides
407 * it needs to transfer another ASCONF Chunk, it MUST wait until the
408 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
409 * subsequent ASCONF. Note this restriction binds each side, so at any
410 * time two ASCONF may be in-transit on any given association (one sent
411 * from each endpoint).
413 static int sctp_send_asconf(struct sctp_association
*asoc
,
414 struct sctp_chunk
*chunk
)
416 struct net
*net
= sock_net(asoc
->base
.sk
);
419 /* If there is an outstanding ASCONF chunk, queue it for later
422 if (asoc
->addip_last_asconf
) {
423 list_add_tail(&chunk
->list
, &asoc
->addip_chunk_list
);
427 /* Hold the chunk until an ASCONF_ACK is received. */
428 sctp_chunk_hold(chunk
);
429 retval
= sctp_primitive_ASCONF(net
, asoc
, chunk
);
431 sctp_chunk_free(chunk
);
433 asoc
->addip_last_asconf
= chunk
;
439 /* Add a list of addresses as bind addresses to local endpoint or
442 * Basically run through each address specified in the addrs/addrcnt
443 * array/length pair, determine if it is IPv6 or IPv4 and call
444 * sctp_do_bind() on it.
446 * If any of them fails, then the operation will be reversed and the
447 * ones that were added will be removed.
449 * Only sctp_setsockopt_bindx() is supposed to call this function.
451 static int sctp_bindx_add(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
456 struct sockaddr
*sa_addr
;
459 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__
, sk
,
463 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
464 /* The list may contain either IPv4 or IPv6 address;
465 * determine the address length for walking thru the list.
468 af
= sctp_get_af_specific(sa_addr
->sa_family
);
474 retval
= sctp_do_bind(sk
, (union sctp_addr
*)sa_addr
,
477 addr_buf
+= af
->sockaddr_len
;
481 /* Failed. Cleanup the ones that have been added */
483 sctp_bindx_rem(sk
, addrs
, cnt
);
491 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
492 * associations that are part of the endpoint indicating that a list of local
493 * addresses are added to the endpoint.
495 * If any of the addresses is already in the bind address list of the
496 * association, we do not send the chunk for that association. But it will not
497 * affect other associations.
499 * Only sctp_setsockopt_bindx() is supposed to call this function.
501 static int sctp_send_asconf_add_ip(struct sock
*sk
,
502 struct sockaddr
*addrs
,
505 struct net
*net
= sock_net(sk
);
506 struct sctp_sock
*sp
;
507 struct sctp_endpoint
*ep
;
508 struct sctp_association
*asoc
;
509 struct sctp_bind_addr
*bp
;
510 struct sctp_chunk
*chunk
;
511 struct sctp_sockaddr_entry
*laddr
;
512 union sctp_addr
*addr
;
513 union sctp_addr saveaddr
;
520 if (!net
->sctp
.addip_enable
)
526 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
527 __func__
, sk
, addrs
, addrcnt
);
529 list_for_each_entry(asoc
, &ep
->asocs
, asocs
) {
530 if (!asoc
->peer
.asconf_capable
)
533 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_ADD_IP
)
536 if (!sctp_state(asoc
, ESTABLISHED
))
539 /* Check if any address in the packed array of addresses is
540 * in the bind address list of the association. If so,
541 * do not send the asconf chunk to its peer, but continue with
542 * other associations.
545 for (i
= 0; i
< addrcnt
; i
++) {
547 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
553 if (sctp_assoc_lookup_laddr(asoc
, addr
))
556 addr_buf
+= af
->sockaddr_len
;
561 /* Use the first valid address in bind addr list of
562 * association as Address Parameter of ASCONF CHUNK.
564 bp
= &asoc
->base
.bind_addr
;
565 p
= bp
->address_list
.next
;
566 laddr
= list_entry(p
, struct sctp_sockaddr_entry
, list
);
567 chunk
= sctp_make_asconf_update_ip(asoc
, &laddr
->a
, addrs
,
568 addrcnt
, SCTP_PARAM_ADD_IP
);
574 /* Add the new addresses to the bind address list with
575 * use_as_src set to 0.
578 for (i
= 0; i
< addrcnt
; i
++) {
580 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
581 memcpy(&saveaddr
, addr
, af
->sockaddr_len
);
582 retval
= sctp_add_bind_addr(bp
, &saveaddr
,
583 SCTP_ADDR_NEW
, GFP_ATOMIC
);
584 addr_buf
+= af
->sockaddr_len
;
586 if (asoc
->src_out_of_asoc_ok
) {
587 struct sctp_transport
*trans
;
589 list_for_each_entry(trans
,
590 &asoc
->peer
.transport_addr_list
, transports
) {
591 /* Clear the source and route cache */
592 dst_release(trans
->dst
);
593 trans
->cwnd
= min(4*asoc
->pathmtu
, max_t(__u32
,
594 2*asoc
->pathmtu
, 4380));
595 trans
->ssthresh
= asoc
->peer
.i
.a_rwnd
;
596 trans
->rto
= asoc
->rto_initial
;
597 sctp_max_rto(asoc
, trans
);
598 trans
->rtt
= trans
->srtt
= trans
->rttvar
= 0;
599 sctp_transport_route(trans
, NULL
,
600 sctp_sk(asoc
->base
.sk
));
603 retval
= sctp_send_asconf(asoc
, chunk
);
610 /* Remove a list of addresses from bind addresses list. Do not remove the
613 * Basically run through each address specified in the addrs/addrcnt
614 * array/length pair, determine if it is IPv6 or IPv4 and call
615 * sctp_del_bind() on it.
617 * If any of them fails, then the operation will be reversed and the
618 * ones that were removed will be added back.
620 * At least one address has to be left; if only one address is
621 * available, the operation will return -EBUSY.
623 * Only sctp_setsockopt_bindx() is supposed to call this function.
625 static int sctp_bindx_rem(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
627 struct sctp_sock
*sp
= sctp_sk(sk
);
628 struct sctp_endpoint
*ep
= sp
->ep
;
630 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
633 union sctp_addr
*sa_addr
;
636 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
637 __func__
, sk
, addrs
, addrcnt
);
640 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
641 /* If the bind address list is empty or if there is only one
642 * bind address, there is nothing more to be removed (we need
643 * at least one address here).
645 if (list_empty(&bp
->address_list
) ||
646 (sctp_list_single_entry(&bp
->address_list
))) {
652 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
658 if (!af
->addr_valid(sa_addr
, sp
, NULL
)) {
659 retval
= -EADDRNOTAVAIL
;
663 if (sa_addr
->v4
.sin_port
&&
664 sa_addr
->v4
.sin_port
!= htons(bp
->port
)) {
669 if (!sa_addr
->v4
.sin_port
)
670 sa_addr
->v4
.sin_port
= htons(bp
->port
);
672 /* FIXME - There is probably a need to check if sk->sk_saddr and
673 * sk->sk_rcv_addr are currently set to one of the addresses to
674 * be removed. This is something which needs to be looked into
675 * when we are fixing the outstanding issues with multi-homing
676 * socket routing and failover schemes. Refer to comments in
677 * sctp_do_bind(). -daisy
679 retval
= sctp_del_bind_addr(bp
, sa_addr
);
681 addr_buf
+= af
->sockaddr_len
;
684 /* Failed. Add the ones that has been removed back */
686 sctp_bindx_add(sk
, addrs
, cnt
);
694 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
695 * the associations that are part of the endpoint indicating that a list of
696 * local addresses are removed from the endpoint.
698 * If any of the addresses is already in the bind address list of the
699 * association, we do not send the chunk for that association. But it will not
700 * affect other associations.
702 * Only sctp_setsockopt_bindx() is supposed to call this function.
704 static int sctp_send_asconf_del_ip(struct sock
*sk
,
705 struct sockaddr
*addrs
,
708 struct net
*net
= sock_net(sk
);
709 struct sctp_sock
*sp
;
710 struct sctp_endpoint
*ep
;
711 struct sctp_association
*asoc
;
712 struct sctp_transport
*transport
;
713 struct sctp_bind_addr
*bp
;
714 struct sctp_chunk
*chunk
;
715 union sctp_addr
*laddr
;
718 struct sctp_sockaddr_entry
*saddr
;
724 if (!net
->sctp
.addip_enable
)
730 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
731 __func__
, sk
, addrs
, addrcnt
);
733 list_for_each_entry(asoc
, &ep
->asocs
, asocs
) {
735 if (!asoc
->peer
.asconf_capable
)
738 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_DEL_IP
)
741 if (!sctp_state(asoc
, ESTABLISHED
))
744 /* Check if any address in the packed array of addresses is
745 * not present in the bind address list of the association.
746 * If so, do not send the asconf chunk to its peer, but
747 * continue with other associations.
750 for (i
= 0; i
< addrcnt
; i
++) {
752 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
758 if (!sctp_assoc_lookup_laddr(asoc
, laddr
))
761 addr_buf
+= af
->sockaddr_len
;
766 /* Find one address in the association's bind address list
767 * that is not in the packed array of addresses. This is to
768 * make sure that we do not delete all the addresses in the
771 bp
= &asoc
->base
.bind_addr
;
772 laddr
= sctp_find_unmatch_addr(bp
, (union sctp_addr
*)addrs
,
774 if ((laddr
== NULL
) && (addrcnt
== 1)) {
775 if (asoc
->asconf_addr_del_pending
)
777 asoc
->asconf_addr_del_pending
=
778 kzalloc(sizeof(union sctp_addr
), GFP_ATOMIC
);
779 if (asoc
->asconf_addr_del_pending
== NULL
) {
783 asoc
->asconf_addr_del_pending
->sa
.sa_family
=
785 asoc
->asconf_addr_del_pending
->v4
.sin_port
=
787 if (addrs
->sa_family
== AF_INET
) {
788 struct sockaddr_in
*sin
;
790 sin
= (struct sockaddr_in
*)addrs
;
791 asoc
->asconf_addr_del_pending
->v4
.sin_addr
.s_addr
= sin
->sin_addr
.s_addr
;
792 } else if (addrs
->sa_family
== AF_INET6
) {
793 struct sockaddr_in6
*sin6
;
795 sin6
= (struct sockaddr_in6
*)addrs
;
796 asoc
->asconf_addr_del_pending
->v6
.sin6_addr
= sin6
->sin6_addr
;
799 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
800 __func__
, asoc
, &asoc
->asconf_addr_del_pending
->sa
,
801 asoc
->asconf_addr_del_pending
);
803 asoc
->src_out_of_asoc_ok
= 1;
811 /* We do not need RCU protection throughout this loop
812 * because this is done under a socket lock from the
815 chunk
= sctp_make_asconf_update_ip(asoc
, laddr
, addrs
, addrcnt
,
823 /* Reset use_as_src flag for the addresses in the bind address
824 * list that are to be deleted.
827 for (i
= 0; i
< addrcnt
; i
++) {
829 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
830 list_for_each_entry(saddr
, &bp
->address_list
, list
) {
831 if (sctp_cmp_addr_exact(&saddr
->a
, laddr
))
832 saddr
->state
= SCTP_ADDR_DEL
;
834 addr_buf
+= af
->sockaddr_len
;
837 /* Update the route and saddr entries for all the transports
838 * as some of the addresses in the bind address list are
839 * about to be deleted and cannot be used as source addresses.
841 list_for_each_entry(transport
, &asoc
->peer
.transport_addr_list
,
843 dst_release(transport
->dst
);
844 sctp_transport_route(transport
, NULL
,
845 sctp_sk(asoc
->base
.sk
));
849 /* We don't need to transmit ASCONF */
851 retval
= sctp_send_asconf(asoc
, chunk
);
857 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
858 int sctp_asconf_mgmt(struct sctp_sock
*sp
, struct sctp_sockaddr_entry
*addrw
)
860 struct sock
*sk
= sctp_opt2sk(sp
);
861 union sctp_addr
*addr
;
864 /* It is safe to write port space in caller. */
866 addr
->v4
.sin_port
= htons(sp
->ep
->base
.bind_addr
.port
);
867 af
= sctp_get_af_specific(addr
->sa
.sa_family
);
870 if (sctp_verify_addr(sk
, addr
, af
->sockaddr_len
))
873 if (addrw
->state
== SCTP_ADDR_NEW
)
874 return sctp_send_asconf_add_ip(sk
, (struct sockaddr
*)addr
, 1);
876 return sctp_send_asconf_del_ip(sk
, (struct sockaddr
*)addr
, 1);
879 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
882 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
885 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
886 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
889 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
890 * Section 3.1.2 for this usage.
892 * addrs is a pointer to an array of one or more socket addresses. Each
893 * address is contained in its appropriate structure (i.e. struct
894 * sockaddr_in or struct sockaddr_in6) the family of the address type
895 * must be used to distinguish the address length (note that this
896 * representation is termed a "packed array" of addresses). The caller
897 * specifies the number of addresses in the array with addrcnt.
899 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
900 * -1, and sets errno to the appropriate error code.
902 * For SCTP, the port given in each socket address must be the same, or
903 * sctp_bindx() will fail, setting errno to EINVAL.
905 * The flags parameter is formed from the bitwise OR of zero or more of
906 * the following currently defined flags:
908 * SCTP_BINDX_ADD_ADDR
910 * SCTP_BINDX_REM_ADDR
912 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
913 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
914 * addresses from the association. The two flags are mutually exclusive;
915 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
916 * not remove all addresses from an association; sctp_bindx() will
917 * reject such an attempt with EINVAL.
919 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
920 * additional addresses with an endpoint after calling bind(). Or use
921 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
922 * socket is associated with so that no new association accepted will be
923 * associated with those addresses. If the endpoint supports dynamic
924 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
925 * endpoint to send the appropriate message to the peer to change the
926 * peers address lists.
928 * Adding and removing addresses from a connected association is
929 * optional functionality. Implementations that do not support this
930 * functionality should return EOPNOTSUPP.
932 * Basically do nothing but copying the addresses from user to kernel
933 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
934 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
937 * We don't use copy_from_user() for optimization: we first do the
938 * sanity checks (buffer size -fast- and access check-healthy
939 * pointer); if all of those succeed, then we can alloc the memory
940 * (expensive operation) needed to copy the data to kernel. Then we do
941 * the copying without checking the user space area
942 * (__copy_from_user()).
944 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
947 * sk The sk of the socket
948 * addrs The pointer to the addresses in user land
949 * addrssize Size of the addrs buffer
950 * op Operation to perform (add or remove, see the flags of
953 * Returns 0 if ok, <0 errno code on error.
955 static int sctp_setsockopt_bindx(struct sock
*sk
,
956 struct sockaddr __user
*addrs
,
957 int addrs_size
, int op
)
959 struct sockaddr
*kaddrs
;
963 struct sockaddr
*sa_addr
;
967 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
968 __func__
, sk
, addrs
, addrs_size
, op
);
970 if (unlikely(addrs_size
<= 0))
973 /* Check the user passed a healthy pointer. */
974 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
977 /* Alloc space for the address array in kernel memory. */
978 kaddrs
= kmalloc(addrs_size
, GFP_KERNEL
);
979 if (unlikely(!kaddrs
))
982 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
987 /* Walk through the addrs buffer and count the number of addresses. */
989 while (walk_size
< addrs_size
) {
990 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
996 af
= sctp_get_af_specific(sa_addr
->sa_family
);
998 /* If the address family is not supported or if this address
999 * causes the address buffer to overflow return EINVAL.
1001 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
1006 addr_buf
+= af
->sockaddr_len
;
1007 walk_size
+= af
->sockaddr_len
;
1012 case SCTP_BINDX_ADD_ADDR
:
1013 err
= sctp_bindx_add(sk
, kaddrs
, addrcnt
);
1016 err
= sctp_send_asconf_add_ip(sk
, kaddrs
, addrcnt
);
1019 case SCTP_BINDX_REM_ADDR
:
1020 err
= sctp_bindx_rem(sk
, kaddrs
, addrcnt
);
1023 err
= sctp_send_asconf_del_ip(sk
, kaddrs
, addrcnt
);
1037 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1039 * Common routine for handling connect() and sctp_connectx().
1040 * Connect will come in with just a single address.
1042 static int __sctp_connect(struct sock
*sk
,
1043 struct sockaddr
*kaddrs
,
1045 sctp_assoc_t
*assoc_id
)
1047 struct net
*net
= sock_net(sk
);
1048 struct sctp_sock
*sp
;
1049 struct sctp_endpoint
*ep
;
1050 struct sctp_association
*asoc
= NULL
;
1051 struct sctp_association
*asoc2
;
1052 struct sctp_transport
*transport
;
1060 union sctp_addr
*sa_addr
= NULL
;
1062 unsigned short port
;
1063 unsigned int f_flags
= 0;
1068 /* connect() cannot be done on a socket that is already in ESTABLISHED
1069 * state - UDP-style peeled off socket or a TCP-style socket that
1070 * is already connected.
1071 * It cannot be done even on a TCP-style listening socket.
1073 if (sctp_sstate(sk
, ESTABLISHED
) ||
1074 (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))) {
1079 /* Walk through the addrs buffer and count the number of addresses. */
1081 while (walk_size
< addrs_size
) {
1082 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
1088 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
1090 /* If the address family is not supported or if this address
1091 * causes the address buffer to overflow return EINVAL.
1093 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
1098 port
= ntohs(sa_addr
->v4
.sin_port
);
1100 /* Save current address so we can work with it */
1101 memcpy(&to
, sa_addr
, af
->sockaddr_len
);
1103 err
= sctp_verify_addr(sk
, &to
, af
->sockaddr_len
);
1107 /* Make sure the destination port is correctly set
1110 if (asoc
&& asoc
->peer
.port
&& asoc
->peer
.port
!= port
) {
1115 /* Check if there already is a matching association on the
1116 * endpoint (other than the one created here).
1118 asoc2
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1119 if (asoc2
&& asoc2
!= asoc
) {
1120 if (asoc2
->state
>= SCTP_STATE_ESTABLISHED
)
1127 /* If we could not find a matching association on the endpoint,
1128 * make sure that there is no peeled-off association matching
1129 * the peer address even on another socket.
1131 if (sctp_endpoint_is_peeled_off(ep
, &to
)) {
1132 err
= -EADDRNOTAVAIL
;
1137 /* If a bind() or sctp_bindx() is not called prior to
1138 * an sctp_connectx() call, the system picks an
1139 * ephemeral port and will choose an address set
1140 * equivalent to binding with a wildcard address.
1142 if (!ep
->base
.bind_addr
.port
) {
1143 if (sctp_autobind(sk
)) {
1149 * If an unprivileged user inherits a 1-many
1150 * style socket with open associations on a
1151 * privileged port, it MAY be permitted to
1152 * accept new associations, but it SHOULD NOT
1153 * be permitted to open new associations.
1155 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1156 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
)) {
1162 scope
= sctp_scope(&to
);
1163 asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1169 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
,
1177 /* Prime the peer's transport structures. */
1178 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
,
1186 addr_buf
+= af
->sockaddr_len
;
1187 walk_size
+= af
->sockaddr_len
;
1190 /* In case the user of sctp_connectx() wants an association
1191 * id back, assign one now.
1194 err
= sctp_assoc_set_id(asoc
, GFP_KERNEL
);
1199 err
= sctp_primitive_ASSOCIATE(net
, asoc
, NULL
);
1204 /* Initialize sk's dport and daddr for getpeername() */
1205 inet_sk(sk
)->inet_dport
= htons(asoc
->peer
.port
);
1206 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
1207 af
->to_sk_daddr(sa_addr
, sk
);
1210 /* in-kernel sockets don't generally have a file allocated to them
1211 * if all they do is call sock_create_kern().
1213 if (sk
->sk_socket
->file
)
1214 f_flags
= sk
->sk_socket
->file
->f_flags
;
1216 timeo
= sock_sndtimeo(sk
, f_flags
& O_NONBLOCK
);
1218 err
= sctp_wait_for_connect(asoc
, &timeo
);
1219 if ((err
== 0 || err
== -EINPROGRESS
) && assoc_id
)
1220 *assoc_id
= asoc
->assoc_id
;
1222 /* Don't free association on exit. */
1226 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1227 __func__
, asoc
, kaddrs
, err
);
1230 /* sctp_primitive_ASSOCIATE may have added this association
1231 * To the hash table, try to unhash it, just in case, its a noop
1232 * if it wasn't hashed so we're safe
1234 sctp_unhash_established(asoc
);
1235 sctp_association_free(asoc
);
1240 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1243 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1244 * sctp_assoc_t *asoc);
1246 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1247 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1248 * or IPv6 addresses.
1250 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1251 * Section 3.1.2 for this usage.
1253 * addrs is a pointer to an array of one or more socket addresses. Each
1254 * address is contained in its appropriate structure (i.e. struct
1255 * sockaddr_in or struct sockaddr_in6) the family of the address type
1256 * must be used to distengish the address length (note that this
1257 * representation is termed a "packed array" of addresses). The caller
1258 * specifies the number of addresses in the array with addrcnt.
1260 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1261 * the association id of the new association. On failure, sctp_connectx()
1262 * returns -1, and sets errno to the appropriate error code. The assoc_id
1263 * is not touched by the kernel.
1265 * For SCTP, the port given in each socket address must be the same, or
1266 * sctp_connectx() will fail, setting errno to EINVAL.
1268 * An application can use sctp_connectx to initiate an association with
1269 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1270 * allows a caller to specify multiple addresses at which a peer can be
1271 * reached. The way the SCTP stack uses the list of addresses to set up
1272 * the association is implementation dependent. This function only
1273 * specifies that the stack will try to make use of all the addresses in
1274 * the list when needed.
1276 * Note that the list of addresses passed in is only used for setting up
1277 * the association. It does not necessarily equal the set of addresses
1278 * the peer uses for the resulting association. If the caller wants to
1279 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1280 * retrieve them after the association has been set up.
1282 * Basically do nothing but copying the addresses from user to kernel
1283 * land and invoking either sctp_connectx(). This is used for tunneling
1284 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1286 * We don't use copy_from_user() for optimization: we first do the
1287 * sanity checks (buffer size -fast- and access check-healthy
1288 * pointer); if all of those succeed, then we can alloc the memory
1289 * (expensive operation) needed to copy the data to kernel. Then we do
1290 * the copying without checking the user space area
1291 * (__copy_from_user()).
1293 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1296 * sk The sk of the socket
1297 * addrs The pointer to the addresses in user land
1298 * addrssize Size of the addrs buffer
1300 * Returns >=0 if ok, <0 errno code on error.
1302 static int __sctp_setsockopt_connectx(struct sock
*sk
,
1303 struct sockaddr __user
*addrs
,
1305 sctp_assoc_t
*assoc_id
)
1308 struct sockaddr
*kaddrs
;
1310 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1311 __func__
, sk
, addrs
, addrs_size
);
1313 if (unlikely(addrs_size
<= 0))
1316 /* Check the user passed a healthy pointer. */
1317 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
1320 /* Alloc space for the address array in kernel memory. */
1321 kaddrs
= kmalloc(addrs_size
, GFP_KERNEL
);
1322 if (unlikely(!kaddrs
))
1325 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
1328 err
= __sctp_connect(sk
, kaddrs
, addrs_size
, assoc_id
);
1337 * This is an older interface. It's kept for backward compatibility
1338 * to the option that doesn't provide association id.
1340 static int sctp_setsockopt_connectx_old(struct sock
*sk
,
1341 struct sockaddr __user
*addrs
,
1344 return __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, NULL
);
1348 * New interface for the API. The since the API is done with a socket
1349 * option, to make it simple we feed back the association id is as a return
1350 * indication to the call. Error is always negative and association id is
1353 static int sctp_setsockopt_connectx(struct sock
*sk
,
1354 struct sockaddr __user
*addrs
,
1357 sctp_assoc_t assoc_id
= 0;
1360 err
= __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, &assoc_id
);
1369 * New (hopefully final) interface for the API.
1370 * We use the sctp_getaddrs_old structure so that use-space library
1371 * can avoid any unnecessary allocations. The only defferent part
1372 * is that we store the actual length of the address buffer into the
1373 * addrs_num structure member. That way we can re-use the existing
1376 static int sctp_getsockopt_connectx3(struct sock
*sk
, int len
,
1377 char __user
*optval
,
1380 struct sctp_getaddrs_old param
;
1381 sctp_assoc_t assoc_id
= 0;
1384 if (len
< sizeof(param
))
1387 if (copy_from_user(¶m
, optval
, sizeof(param
)))
1390 err
= __sctp_setsockopt_connectx(sk
,
1391 (struct sockaddr __user
*)param
.addrs
,
1392 param
.addr_num
, &assoc_id
);
1394 if (err
== 0 || err
== -EINPROGRESS
) {
1395 if (copy_to_user(optval
, &assoc_id
, sizeof(assoc_id
)))
1397 if (put_user(sizeof(assoc_id
), optlen
))
1404 /* API 3.1.4 close() - UDP Style Syntax
1405 * Applications use close() to perform graceful shutdown (as described in
1406 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1407 * by a UDP-style socket.
1411 * ret = close(int sd);
1413 * sd - the socket descriptor of the associations to be closed.
1415 * To gracefully shutdown a specific association represented by the
1416 * UDP-style socket, an application should use the sendmsg() call,
1417 * passing no user data, but including the appropriate flag in the
1418 * ancillary data (see Section xxxx).
1420 * If sd in the close() call is a branched-off socket representing only
1421 * one association, the shutdown is performed on that association only.
1423 * 4.1.6 close() - TCP Style Syntax
1425 * Applications use close() to gracefully close down an association.
1429 * int close(int sd);
1431 * sd - the socket descriptor of the association to be closed.
1433 * After an application calls close() on a socket descriptor, no further
1434 * socket operations will succeed on that descriptor.
1436 * API 7.1.4 SO_LINGER
1438 * An application using the TCP-style socket can use this option to
1439 * perform the SCTP ABORT primitive. The linger option structure is:
1442 * int l_onoff; // option on/off
1443 * int l_linger; // linger time
1446 * To enable the option, set l_onoff to 1. If the l_linger value is set
1447 * to 0, calling close() is the same as the ABORT primitive. If the
1448 * value is set to a negative value, the setsockopt() call will return
1449 * an error. If the value is set to a positive value linger_time, the
1450 * close() can be blocked for at most linger_time ms. If the graceful
1451 * shutdown phase does not finish during this period, close() will
1452 * return but the graceful shutdown phase continues in the system.
1454 static void sctp_close(struct sock
*sk
, long timeout
)
1456 struct net
*net
= sock_net(sk
);
1457 struct sctp_endpoint
*ep
;
1458 struct sctp_association
*asoc
;
1459 struct list_head
*pos
, *temp
;
1460 unsigned int data_was_unread
;
1462 pr_debug("%s: sk:%p, timeout:%ld\n", __func__
, sk
, timeout
);
1465 sk
->sk_shutdown
= SHUTDOWN_MASK
;
1466 sk
->sk_state
= SCTP_SS_CLOSING
;
1468 ep
= sctp_sk(sk
)->ep
;
1470 /* Clean up any skbs sitting on the receive queue. */
1471 data_was_unread
= sctp_queue_purge_ulpevents(&sk
->sk_receive_queue
);
1472 data_was_unread
+= sctp_queue_purge_ulpevents(&sctp_sk(sk
)->pd_lobby
);
1474 /* Walk all associations on an endpoint. */
1475 list_for_each_safe(pos
, temp
, &ep
->asocs
) {
1476 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
1478 if (sctp_style(sk
, TCP
)) {
1479 /* A closed association can still be in the list if
1480 * it belongs to a TCP-style listening socket that is
1481 * not yet accepted. If so, free it. If not, send an
1482 * ABORT or SHUTDOWN based on the linger options.
1484 if (sctp_state(asoc
, CLOSED
)) {
1485 sctp_unhash_established(asoc
);
1486 sctp_association_free(asoc
);
1491 if (data_was_unread
|| !skb_queue_empty(&asoc
->ulpq
.lobby
) ||
1492 !skb_queue_empty(&asoc
->ulpq
.reasm
) ||
1493 (sock_flag(sk
, SOCK_LINGER
) && !sk
->sk_lingertime
)) {
1494 struct sctp_chunk
*chunk
;
1496 chunk
= sctp_make_abort_user(asoc
, NULL
, 0);
1498 sctp_primitive_ABORT(net
, asoc
, chunk
);
1500 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
1503 /* On a TCP-style socket, block for at most linger_time if set. */
1504 if (sctp_style(sk
, TCP
) && timeout
)
1505 sctp_wait_for_close(sk
, timeout
);
1507 /* This will run the backlog queue. */
1510 /* Supposedly, no process has access to the socket, but
1511 * the net layers still may.
1516 /* Hold the sock, since sk_common_release() will put sock_put()
1517 * and we have just a little more cleanup.
1520 sk_common_release(sk
);
1527 SCTP_DBG_OBJCNT_DEC(sock
);
1530 /* Handle EPIPE error. */
1531 static int sctp_error(struct sock
*sk
, int flags
, int err
)
1534 err
= sock_error(sk
) ? : -EPIPE
;
1535 if (err
== -EPIPE
&& !(flags
& MSG_NOSIGNAL
))
1536 send_sig(SIGPIPE
, current
, 0);
1540 /* API 3.1.3 sendmsg() - UDP Style Syntax
1542 * An application uses sendmsg() and recvmsg() calls to transmit data to
1543 * and receive data from its peer.
1545 * ssize_t sendmsg(int socket, const struct msghdr *message,
1548 * socket - the socket descriptor of the endpoint.
1549 * message - pointer to the msghdr structure which contains a single
1550 * user message and possibly some ancillary data.
1552 * See Section 5 for complete description of the data
1555 * flags - flags sent or received with the user message, see Section
1556 * 5 for complete description of the flags.
1558 * Note: This function could use a rewrite especially when explicit
1559 * connect support comes in.
1561 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1563 static int sctp_msghdr_parse(const struct msghdr
*, sctp_cmsgs_t
*);
1565 static int sctp_sendmsg(struct kiocb
*iocb
, struct sock
*sk
,
1566 struct msghdr
*msg
, size_t msg_len
)
1568 struct net
*net
= sock_net(sk
);
1569 struct sctp_sock
*sp
;
1570 struct sctp_endpoint
*ep
;
1571 struct sctp_association
*new_asoc
= NULL
, *asoc
= NULL
;
1572 struct sctp_transport
*transport
, *chunk_tp
;
1573 struct sctp_chunk
*chunk
;
1575 struct sockaddr
*msg_name
= NULL
;
1576 struct sctp_sndrcvinfo default_sinfo
;
1577 struct sctp_sndrcvinfo
*sinfo
;
1578 struct sctp_initmsg
*sinit
;
1579 sctp_assoc_t associd
= 0;
1580 sctp_cmsgs_t cmsgs
= { NULL
};
1584 __u16 sinfo_flags
= 0;
1585 struct sctp_datamsg
*datamsg
;
1586 int msg_flags
= msg
->msg_flags
;
1592 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__
, sk
,
1595 /* We cannot send a message over a TCP-style listening socket. */
1596 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
)) {
1601 /* Parse out the SCTP CMSGs. */
1602 err
= sctp_msghdr_parse(msg
, &cmsgs
);
1604 pr_debug("%s: msghdr parse err:%x\n", __func__
, err
);
1608 /* Fetch the destination address for this packet. This
1609 * address only selects the association--it is not necessarily
1610 * the address we will send to.
1611 * For a peeled-off socket, msg_name is ignored.
1613 if (!sctp_style(sk
, UDP_HIGH_BANDWIDTH
) && msg
->msg_name
) {
1614 int msg_namelen
= msg
->msg_namelen
;
1616 err
= sctp_verify_addr(sk
, (union sctp_addr
*)msg
->msg_name
,
1621 if (msg_namelen
> sizeof(to
))
1622 msg_namelen
= sizeof(to
);
1623 memcpy(&to
, msg
->msg_name
, msg_namelen
);
1624 msg_name
= msg
->msg_name
;
1630 /* Did the user specify SNDRCVINFO? */
1632 sinfo_flags
= sinfo
->sinfo_flags
;
1633 associd
= sinfo
->sinfo_assoc_id
;
1636 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__
,
1637 msg_len
, sinfo_flags
);
1639 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1640 if (sctp_style(sk
, TCP
) && (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
))) {
1645 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1646 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1647 * If SCTP_ABORT is set, the message length could be non zero with
1648 * the msg_iov set to the user abort reason.
1650 if (((sinfo_flags
& SCTP_EOF
) && (msg_len
> 0)) ||
1651 (!(sinfo_flags
& (SCTP_EOF
|SCTP_ABORT
)) && (msg_len
== 0))) {
1656 /* If SCTP_ADDR_OVER is set, there must be an address
1657 * specified in msg_name.
1659 if ((sinfo_flags
& SCTP_ADDR_OVER
) && (!msg
->msg_name
)) {
1666 pr_debug("%s: about to look up association\n", __func__
);
1670 /* If a msg_name has been specified, assume this is to be used. */
1672 /* Look for a matching association on the endpoint. */
1673 asoc
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1675 /* If we could not find a matching association on the
1676 * endpoint, make sure that it is not a TCP-style
1677 * socket that already has an association or there is
1678 * no peeled-off association on another socket.
1680 if ((sctp_style(sk
, TCP
) &&
1681 sctp_sstate(sk
, ESTABLISHED
)) ||
1682 sctp_endpoint_is_peeled_off(ep
, &to
)) {
1683 err
= -EADDRNOTAVAIL
;
1688 asoc
= sctp_id2assoc(sk
, associd
);
1696 pr_debug("%s: just looked up association:%p\n", __func__
, asoc
);
1698 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1699 * socket that has an association in CLOSED state. This can
1700 * happen when an accepted socket has an association that is
1703 if (sctp_state(asoc
, CLOSED
) && sctp_style(sk
, TCP
)) {
1708 if (sinfo_flags
& SCTP_EOF
) {
1709 pr_debug("%s: shutting down association:%p\n",
1712 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
1716 if (sinfo_flags
& SCTP_ABORT
) {
1718 chunk
= sctp_make_abort_user(asoc
, msg
, msg_len
);
1724 pr_debug("%s: aborting association:%p\n",
1727 sctp_primitive_ABORT(net
, asoc
, chunk
);
1733 /* Do we need to create the association? */
1735 pr_debug("%s: there is no association yet\n", __func__
);
1737 if (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
)) {
1742 /* Check for invalid stream against the stream counts,
1743 * either the default or the user specified stream counts.
1746 if (!sinit
|| !sinit
->sinit_num_ostreams
) {
1747 /* Check against the defaults. */
1748 if (sinfo
->sinfo_stream
>=
1749 sp
->initmsg
.sinit_num_ostreams
) {
1754 /* Check against the requested. */
1755 if (sinfo
->sinfo_stream
>=
1756 sinit
->sinit_num_ostreams
) {
1764 * API 3.1.2 bind() - UDP Style Syntax
1765 * If a bind() or sctp_bindx() is not called prior to a
1766 * sendmsg() call that initiates a new association, the
1767 * system picks an ephemeral port and will choose an address
1768 * set equivalent to binding with a wildcard address.
1770 if (!ep
->base
.bind_addr
.port
) {
1771 if (sctp_autobind(sk
)) {
1777 * If an unprivileged user inherits a one-to-many
1778 * style socket with open associations on a privileged
1779 * port, it MAY be permitted to accept new associations,
1780 * but it SHOULD NOT be permitted to open new
1783 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1784 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
)) {
1790 scope
= sctp_scope(&to
);
1791 new_asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1797 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
, GFP_KERNEL
);
1803 /* If the SCTP_INIT ancillary data is specified, set all
1804 * the association init values accordingly.
1807 if (sinit
->sinit_num_ostreams
) {
1808 asoc
->c
.sinit_num_ostreams
=
1809 sinit
->sinit_num_ostreams
;
1811 if (sinit
->sinit_max_instreams
) {
1812 asoc
->c
.sinit_max_instreams
=
1813 sinit
->sinit_max_instreams
;
1815 if (sinit
->sinit_max_attempts
) {
1816 asoc
->max_init_attempts
1817 = sinit
->sinit_max_attempts
;
1819 if (sinit
->sinit_max_init_timeo
) {
1820 asoc
->max_init_timeo
=
1821 msecs_to_jiffies(sinit
->sinit_max_init_timeo
);
1825 /* Prime the peer's transport structures. */
1826 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
, SCTP_UNKNOWN
);
1833 /* ASSERT: we have a valid association at this point. */
1834 pr_debug("%s: we have a valid association\n", __func__
);
1837 /* If the user didn't specify SNDRCVINFO, make up one with
1840 memset(&default_sinfo
, 0, sizeof(default_sinfo
));
1841 default_sinfo
.sinfo_stream
= asoc
->default_stream
;
1842 default_sinfo
.sinfo_flags
= asoc
->default_flags
;
1843 default_sinfo
.sinfo_ppid
= asoc
->default_ppid
;
1844 default_sinfo
.sinfo_context
= asoc
->default_context
;
1845 default_sinfo
.sinfo_timetolive
= asoc
->default_timetolive
;
1846 default_sinfo
.sinfo_assoc_id
= sctp_assoc2id(asoc
);
1847 sinfo
= &default_sinfo
;
1850 /* API 7.1.7, the sndbuf size per association bounds the
1851 * maximum size of data that can be sent in a single send call.
1853 if (msg_len
> sk
->sk_sndbuf
) {
1858 if (asoc
->pmtu_pending
)
1859 sctp_assoc_pending_pmtu(sk
, asoc
);
1861 /* If fragmentation is disabled and the message length exceeds the
1862 * association fragmentation point, return EMSGSIZE. The I-D
1863 * does not specify what this error is, but this looks like
1866 if (sctp_sk(sk
)->disable_fragments
&& (msg_len
> asoc
->frag_point
)) {
1871 /* Check for invalid stream. */
1872 if (sinfo
->sinfo_stream
>= asoc
->c
.sinit_num_ostreams
) {
1877 timeo
= sock_sndtimeo(sk
, msg
->msg_flags
& MSG_DONTWAIT
);
1878 if (!sctp_wspace(asoc
)) {
1879 err
= sctp_wait_for_sndbuf(asoc
, &timeo
, msg_len
);
1884 /* If an address is passed with the sendto/sendmsg call, it is used
1885 * to override the primary destination address in the TCP model, or
1886 * when SCTP_ADDR_OVER flag is set in the UDP model.
1888 if ((sctp_style(sk
, TCP
) && msg_name
) ||
1889 (sinfo_flags
& SCTP_ADDR_OVER
)) {
1890 chunk_tp
= sctp_assoc_lookup_paddr(asoc
, &to
);
1898 /* Auto-connect, if we aren't connected already. */
1899 if (sctp_state(asoc
, CLOSED
)) {
1900 err
= sctp_primitive_ASSOCIATE(net
, asoc
, NULL
);
1904 pr_debug("%s: we associated primitively\n", __func__
);
1907 /* Break the message into multiple chunks of maximum size. */
1908 datamsg
= sctp_datamsg_from_user(asoc
, sinfo
, msg
, msg_len
);
1909 if (IS_ERR(datamsg
)) {
1910 err
= PTR_ERR(datamsg
);
1914 /* Now send the (possibly) fragmented message. */
1915 list_for_each_entry(chunk
, &datamsg
->chunks
, frag_list
) {
1916 sctp_chunk_hold(chunk
);
1918 /* Do accounting for the write space. */
1919 sctp_set_owner_w(chunk
);
1921 chunk
->transport
= chunk_tp
;
1924 /* Send it to the lower layers. Note: all chunks
1925 * must either fail or succeed. The lower layer
1926 * works that way today. Keep it that way or this
1929 err
= sctp_primitive_SEND(net
, asoc
, datamsg
);
1930 /* Did the lower layer accept the chunk? */
1932 sctp_datamsg_free(datamsg
);
1936 pr_debug("%s: we sent primitively\n", __func__
);
1938 sctp_datamsg_put(datamsg
);
1941 /* If we are already past ASSOCIATE, the lower
1942 * layers are responsible for association cleanup.
1948 sctp_unhash_established(asoc
);
1949 sctp_association_free(asoc
);
1955 return sctp_error(sk
, msg_flags
, err
);
1962 err
= sock_error(sk
);
1972 /* This is an extended version of skb_pull() that removes the data from the
1973 * start of a skb even when data is spread across the list of skb's in the
1974 * frag_list. len specifies the total amount of data that needs to be removed.
1975 * when 'len' bytes could be removed from the skb, it returns 0.
1976 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1977 * could not be removed.
1979 static int sctp_skb_pull(struct sk_buff
*skb
, int len
)
1981 struct sk_buff
*list
;
1982 int skb_len
= skb_headlen(skb
);
1985 if (len
<= skb_len
) {
1986 __skb_pull(skb
, len
);
1990 __skb_pull(skb
, skb_len
);
1992 skb_walk_frags(skb
, list
) {
1993 rlen
= sctp_skb_pull(list
, len
);
1994 skb
->len
-= (len
-rlen
);
1995 skb
->data_len
-= (len
-rlen
);
2006 /* API 3.1.3 recvmsg() - UDP Style Syntax
2008 * ssize_t recvmsg(int socket, struct msghdr *message,
2011 * socket - the socket descriptor of the endpoint.
2012 * message - pointer to the msghdr structure which contains a single
2013 * user message and possibly some ancillary data.
2015 * See Section 5 for complete description of the data
2018 * flags - flags sent or received with the user message, see Section
2019 * 5 for complete description of the flags.
2021 static struct sk_buff
*sctp_skb_recv_datagram(struct sock
*, int, int, int *);
2023 static int sctp_recvmsg(struct kiocb
*iocb
, struct sock
*sk
,
2024 struct msghdr
*msg
, size_t len
, int noblock
,
2025 int flags
, int *addr_len
)
2027 struct sctp_ulpevent
*event
= NULL
;
2028 struct sctp_sock
*sp
= sctp_sk(sk
);
2029 struct sk_buff
*skb
;
2034 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2035 "addr_len:%p)\n", __func__
, sk
, msg
, len
, noblock
, flags
,
2040 if (sctp_style(sk
, TCP
) && !sctp_sstate(sk
, ESTABLISHED
)) {
2045 skb
= sctp_skb_recv_datagram(sk
, flags
, noblock
, &err
);
2049 /* Get the total length of the skb including any skb's in the
2058 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
2060 event
= sctp_skb2event(skb
);
2065 sock_recv_ts_and_drops(msg
, sk
, skb
);
2066 if (sctp_ulpevent_is_notification(event
)) {
2067 msg
->msg_flags
|= MSG_NOTIFICATION
;
2068 sp
->pf
->event_msgname(event
, msg
->msg_name
, addr_len
);
2070 sp
->pf
->skb_msgname(skb
, msg
->msg_name
, addr_len
);
2073 /* Check if we allow SCTP_SNDRCVINFO. */
2074 if (sp
->subscribe
.sctp_data_io_event
)
2075 sctp_ulpevent_read_sndrcvinfo(event
, msg
);
2077 /* FIXME: we should be calling IP/IPv6 layers. */
2078 if (sk
->sk_protinfo
.af_inet
.cmsg_flags
)
2079 ip_cmsg_recv(msg
, skb
);
2084 /* If skb's length exceeds the user's buffer, update the skb and
2085 * push it back to the receive_queue so that the next call to
2086 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2088 if (skb_len
> copied
) {
2089 msg
->msg_flags
&= ~MSG_EOR
;
2090 if (flags
& MSG_PEEK
)
2092 sctp_skb_pull(skb
, copied
);
2093 skb_queue_head(&sk
->sk_receive_queue
, skb
);
2095 /* When only partial message is copied to the user, increase
2096 * rwnd by that amount. If all the data in the skb is read,
2097 * rwnd is updated when the event is freed.
2099 if (!sctp_ulpevent_is_notification(event
))
2100 sctp_assoc_rwnd_increase(event
->asoc
, copied
);
2102 } else if ((event
->msg_flags
& MSG_NOTIFICATION
) ||
2103 (event
->msg_flags
& MSG_EOR
))
2104 msg
->msg_flags
|= MSG_EOR
;
2106 msg
->msg_flags
&= ~MSG_EOR
;
2109 if (flags
& MSG_PEEK
) {
2110 /* Release the skb reference acquired after peeking the skb in
2111 * sctp_skb_recv_datagram().
2115 /* Free the event which includes releasing the reference to
2116 * the owner of the skb, freeing the skb and updating the
2119 sctp_ulpevent_free(event
);
2126 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2128 * This option is a on/off flag. If enabled no SCTP message
2129 * fragmentation will be performed. Instead if a message being sent
2130 * exceeds the current PMTU size, the message will NOT be sent and
2131 * instead a error will be indicated to the user.
2133 static int sctp_setsockopt_disable_fragments(struct sock
*sk
,
2134 char __user
*optval
,
2135 unsigned int optlen
)
2139 if (optlen
< sizeof(int))
2142 if (get_user(val
, (int __user
*)optval
))
2145 sctp_sk(sk
)->disable_fragments
= (val
== 0) ? 0 : 1;
2150 static int sctp_setsockopt_events(struct sock
*sk
, char __user
*optval
,
2151 unsigned int optlen
)
2153 struct sctp_association
*asoc
;
2154 struct sctp_ulpevent
*event
;
2156 if (optlen
> sizeof(struct sctp_event_subscribe
))
2158 if (copy_from_user(&sctp_sk(sk
)->subscribe
, optval
, optlen
))
2162 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2163 * if there is no data to be sent or retransmit, the stack will
2164 * immediately send up this notification.
2166 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT
,
2167 &sctp_sk(sk
)->subscribe
)) {
2168 asoc
= sctp_id2assoc(sk
, 0);
2170 if (asoc
&& sctp_outq_is_empty(&asoc
->outqueue
)) {
2171 event
= sctp_ulpevent_make_sender_dry_event(asoc
,
2176 sctp_ulpq_tail_event(&asoc
->ulpq
, event
);
2183 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2185 * This socket option is applicable to the UDP-style socket only. When
2186 * set it will cause associations that are idle for more than the
2187 * specified number of seconds to automatically close. An association
2188 * being idle is defined an association that has NOT sent or received
2189 * user data. The special value of '0' indicates that no automatic
2190 * close of any associations should be performed. The option expects an
2191 * integer defining the number of seconds of idle time before an
2192 * association is closed.
2194 static int sctp_setsockopt_autoclose(struct sock
*sk
, char __user
*optval
,
2195 unsigned int optlen
)
2197 struct sctp_sock
*sp
= sctp_sk(sk
);
2198 struct net
*net
= sock_net(sk
);
2200 /* Applicable to UDP-style socket only */
2201 if (sctp_style(sk
, TCP
))
2203 if (optlen
!= sizeof(int))
2205 if (copy_from_user(&sp
->autoclose
, optval
, optlen
))
2208 if (sp
->autoclose
> net
->sctp
.max_autoclose
)
2209 sp
->autoclose
= net
->sctp
.max_autoclose
;
2214 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2216 * Applications can enable or disable heartbeats for any peer address of
2217 * an association, modify an address's heartbeat interval, force a
2218 * heartbeat to be sent immediately, and adjust the address's maximum
2219 * number of retransmissions sent before an address is considered
2220 * unreachable. The following structure is used to access and modify an
2221 * address's parameters:
2223 * struct sctp_paddrparams {
2224 * sctp_assoc_t spp_assoc_id;
2225 * struct sockaddr_storage spp_address;
2226 * uint32_t spp_hbinterval;
2227 * uint16_t spp_pathmaxrxt;
2228 * uint32_t spp_pathmtu;
2229 * uint32_t spp_sackdelay;
2230 * uint32_t spp_flags;
2233 * spp_assoc_id - (one-to-many style socket) This is filled in the
2234 * application, and identifies the association for
2236 * spp_address - This specifies which address is of interest.
2237 * spp_hbinterval - This contains the value of the heartbeat interval,
2238 * in milliseconds. If a value of zero
2239 * is present in this field then no changes are to
2240 * be made to this parameter.
2241 * spp_pathmaxrxt - This contains the maximum number of
2242 * retransmissions before this address shall be
2243 * considered unreachable. If a value of zero
2244 * is present in this field then no changes are to
2245 * be made to this parameter.
2246 * spp_pathmtu - When Path MTU discovery is disabled the value
2247 * specified here will be the "fixed" path mtu.
2248 * Note that if the spp_address field is empty
2249 * then all associations on this address will
2250 * have this fixed path mtu set upon them.
2252 * spp_sackdelay - When delayed sack is enabled, this value specifies
2253 * the number of milliseconds that sacks will be delayed
2254 * for. This value will apply to all addresses of an
2255 * association if the spp_address field is empty. Note
2256 * also, that if delayed sack is enabled and this
2257 * value is set to 0, no change is made to the last
2258 * recorded delayed sack timer value.
2260 * spp_flags - These flags are used to control various features
2261 * on an association. The flag field may contain
2262 * zero or more of the following options.
2264 * SPP_HB_ENABLE - Enable heartbeats on the
2265 * specified address. Note that if the address
2266 * field is empty all addresses for the association
2267 * have heartbeats enabled upon them.
2269 * SPP_HB_DISABLE - Disable heartbeats on the
2270 * speicifed address. Note that if the address
2271 * field is empty all addresses for the association
2272 * will have their heartbeats disabled. Note also
2273 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2274 * mutually exclusive, only one of these two should
2275 * be specified. Enabling both fields will have
2276 * undetermined results.
2278 * SPP_HB_DEMAND - Request a user initiated heartbeat
2279 * to be made immediately.
2281 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2282 * heartbeat delayis to be set to the value of 0
2285 * SPP_PMTUD_ENABLE - This field will enable PMTU
2286 * discovery upon the specified address. Note that
2287 * if the address feild is empty then all addresses
2288 * on the association are effected.
2290 * SPP_PMTUD_DISABLE - This field will disable PMTU
2291 * discovery upon the specified address. Note that
2292 * if the address feild is empty then all addresses
2293 * on the association are effected. Not also that
2294 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2295 * exclusive. Enabling both will have undetermined
2298 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2299 * on delayed sack. The time specified in spp_sackdelay
2300 * is used to specify the sack delay for this address. Note
2301 * that if spp_address is empty then all addresses will
2302 * enable delayed sack and take on the sack delay
2303 * value specified in spp_sackdelay.
2304 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2305 * off delayed sack. If the spp_address field is blank then
2306 * delayed sack is disabled for the entire association. Note
2307 * also that this field is mutually exclusive to
2308 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2311 static int sctp_apply_peer_addr_params(struct sctp_paddrparams
*params
,
2312 struct sctp_transport
*trans
,
2313 struct sctp_association
*asoc
,
2314 struct sctp_sock
*sp
,
2317 int sackdelay_change
)
2321 if (params
->spp_flags
& SPP_HB_DEMAND
&& trans
) {
2322 struct net
*net
= sock_net(trans
->asoc
->base
.sk
);
2324 error
= sctp_primitive_REQUESTHEARTBEAT(net
, trans
->asoc
, trans
);
2329 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2330 * this field is ignored. Note also that a value of zero indicates
2331 * the current setting should be left unchanged.
2333 if (params
->spp_flags
& SPP_HB_ENABLE
) {
2335 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2336 * set. This lets us use 0 value when this flag
2339 if (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)
2340 params
->spp_hbinterval
= 0;
2342 if (params
->spp_hbinterval
||
2343 (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)) {
2346 msecs_to_jiffies(params
->spp_hbinterval
);
2349 msecs_to_jiffies(params
->spp_hbinterval
);
2351 sp
->hbinterval
= params
->spp_hbinterval
;
2358 trans
->param_flags
=
2359 (trans
->param_flags
& ~SPP_HB
) | hb_change
;
2362 (asoc
->param_flags
& ~SPP_HB
) | hb_change
;
2365 (sp
->param_flags
& ~SPP_HB
) | hb_change
;
2369 /* When Path MTU discovery is disabled the value specified here will
2370 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2371 * include the flag SPP_PMTUD_DISABLE for this field to have any
2374 if ((params
->spp_flags
& SPP_PMTUD_DISABLE
) && params
->spp_pathmtu
) {
2376 trans
->pathmtu
= params
->spp_pathmtu
;
2377 sctp_assoc_sync_pmtu(sctp_opt2sk(sp
), asoc
);
2379 asoc
->pathmtu
= params
->spp_pathmtu
;
2380 sctp_frag_point(asoc
, params
->spp_pathmtu
);
2382 sp
->pathmtu
= params
->spp_pathmtu
;
2388 int update
= (trans
->param_flags
& SPP_PMTUD_DISABLE
) &&
2389 (params
->spp_flags
& SPP_PMTUD_ENABLE
);
2390 trans
->param_flags
=
2391 (trans
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2393 sctp_transport_pmtu(trans
, sctp_opt2sk(sp
));
2394 sctp_assoc_sync_pmtu(sctp_opt2sk(sp
), asoc
);
2398 (asoc
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2401 (sp
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2405 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2406 * value of this field is ignored. Note also that a value of zero
2407 * indicates the current setting should be left unchanged.
2409 if ((params
->spp_flags
& SPP_SACKDELAY_ENABLE
) && params
->spp_sackdelay
) {
2412 msecs_to_jiffies(params
->spp_sackdelay
);
2415 msecs_to_jiffies(params
->spp_sackdelay
);
2417 sp
->sackdelay
= params
->spp_sackdelay
;
2421 if (sackdelay_change
) {
2423 trans
->param_flags
=
2424 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2428 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2432 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2437 /* Note that a value of zero indicates the current setting should be
2440 if (params
->spp_pathmaxrxt
) {
2442 trans
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2444 asoc
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2446 sp
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2453 static int sctp_setsockopt_peer_addr_params(struct sock
*sk
,
2454 char __user
*optval
,
2455 unsigned int optlen
)
2457 struct sctp_paddrparams params
;
2458 struct sctp_transport
*trans
= NULL
;
2459 struct sctp_association
*asoc
= NULL
;
2460 struct sctp_sock
*sp
= sctp_sk(sk
);
2462 int hb_change
, pmtud_change
, sackdelay_change
;
2464 if (optlen
!= sizeof(struct sctp_paddrparams
))
2467 if (copy_from_user(¶ms
, optval
, optlen
))
2470 /* Validate flags and value parameters. */
2471 hb_change
= params
.spp_flags
& SPP_HB
;
2472 pmtud_change
= params
.spp_flags
& SPP_PMTUD
;
2473 sackdelay_change
= params
.spp_flags
& SPP_SACKDELAY
;
2475 if (hb_change
== SPP_HB
||
2476 pmtud_change
== SPP_PMTUD
||
2477 sackdelay_change
== SPP_SACKDELAY
||
2478 params
.spp_sackdelay
> 500 ||
2479 (params
.spp_pathmtu
&&
2480 params
.spp_pathmtu
< SCTP_DEFAULT_MINSEGMENT
))
2483 /* If an address other than INADDR_ANY is specified, and
2484 * no transport is found, then the request is invalid.
2486 if (!sctp_is_any(sk
, (union sctp_addr
*)¶ms
.spp_address
)) {
2487 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
2488 params
.spp_assoc_id
);
2493 /* Get association, if assoc_id != 0 and the socket is a one
2494 * to many style socket, and an association was not found, then
2495 * the id was invalid.
2497 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
2498 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
))
2501 /* Heartbeat demand can only be sent on a transport or
2502 * association, but not a socket.
2504 if (params
.spp_flags
& SPP_HB_DEMAND
&& !trans
&& !asoc
)
2507 /* Process parameters. */
2508 error
= sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2509 hb_change
, pmtud_change
,
2515 /* If changes are for association, also apply parameters to each
2518 if (!trans
&& asoc
) {
2519 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2521 sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2522 hb_change
, pmtud_change
,
2530 static inline __u32
sctp_spp_sackdelay_enable(__u32 param_flags
)
2532 return (param_flags
& ~SPP_SACKDELAY
) | SPP_SACKDELAY_ENABLE
;
2535 static inline __u32
sctp_spp_sackdelay_disable(__u32 param_flags
)
2537 return (param_flags
& ~SPP_SACKDELAY
) | SPP_SACKDELAY_DISABLE
;
2541 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2543 * This option will effect the way delayed acks are performed. This
2544 * option allows you to get or set the delayed ack time, in
2545 * milliseconds. It also allows changing the delayed ack frequency.
2546 * Changing the frequency to 1 disables the delayed sack algorithm. If
2547 * the assoc_id is 0, then this sets or gets the endpoints default
2548 * values. If the assoc_id field is non-zero, then the set or get
2549 * effects the specified association for the one to many model (the
2550 * assoc_id field is ignored by the one to one model). Note that if
2551 * sack_delay or sack_freq are 0 when setting this option, then the
2552 * current values will remain unchanged.
2554 * struct sctp_sack_info {
2555 * sctp_assoc_t sack_assoc_id;
2556 * uint32_t sack_delay;
2557 * uint32_t sack_freq;
2560 * sack_assoc_id - This parameter, indicates which association the user
2561 * is performing an action upon. Note that if this field's value is
2562 * zero then the endpoints default value is changed (effecting future
2563 * associations only).
2565 * sack_delay - This parameter contains the number of milliseconds that
2566 * the user is requesting the delayed ACK timer be set to. Note that
2567 * this value is defined in the standard to be between 200 and 500
2570 * sack_freq - This parameter contains the number of packets that must
2571 * be received before a sack is sent without waiting for the delay
2572 * timer to expire. The default value for this is 2, setting this
2573 * value to 1 will disable the delayed sack algorithm.
2576 static int sctp_setsockopt_delayed_ack(struct sock
*sk
,
2577 char __user
*optval
, unsigned int optlen
)
2579 struct sctp_sack_info params
;
2580 struct sctp_transport
*trans
= NULL
;
2581 struct sctp_association
*asoc
= NULL
;
2582 struct sctp_sock
*sp
= sctp_sk(sk
);
2584 if (optlen
== sizeof(struct sctp_sack_info
)) {
2585 if (copy_from_user(¶ms
, optval
, optlen
))
2588 if (params
.sack_delay
== 0 && params
.sack_freq
== 0)
2590 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
2591 pr_warn_ratelimited(DEPRECATED
2593 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2594 "Use struct sctp_sack_info instead\n",
2595 current
->comm
, task_pid_nr(current
));
2596 if (copy_from_user(¶ms
, optval
, optlen
))
2599 if (params
.sack_delay
== 0)
2600 params
.sack_freq
= 1;
2602 params
.sack_freq
= 0;
2606 /* Validate value parameter. */
2607 if (params
.sack_delay
> 500)
2610 /* Get association, if sack_assoc_id != 0 and the socket is a one
2611 * to many style socket, and an association was not found, then
2612 * the id was invalid.
2614 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
2615 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
2618 if (params
.sack_delay
) {
2621 msecs_to_jiffies(params
.sack_delay
);
2623 sctp_spp_sackdelay_enable(asoc
->param_flags
);
2625 sp
->sackdelay
= params
.sack_delay
;
2627 sctp_spp_sackdelay_enable(sp
->param_flags
);
2631 if (params
.sack_freq
== 1) {
2634 sctp_spp_sackdelay_disable(asoc
->param_flags
);
2637 sctp_spp_sackdelay_disable(sp
->param_flags
);
2639 } else if (params
.sack_freq
> 1) {
2641 asoc
->sackfreq
= params
.sack_freq
;
2643 sctp_spp_sackdelay_enable(asoc
->param_flags
);
2645 sp
->sackfreq
= params
.sack_freq
;
2647 sctp_spp_sackdelay_enable(sp
->param_flags
);
2651 /* If change is for association, also apply to each transport. */
2653 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2655 if (params
.sack_delay
) {
2657 msecs_to_jiffies(params
.sack_delay
);
2658 trans
->param_flags
=
2659 sctp_spp_sackdelay_enable(trans
->param_flags
);
2661 if (params
.sack_freq
== 1) {
2662 trans
->param_flags
=
2663 sctp_spp_sackdelay_disable(trans
->param_flags
);
2664 } else if (params
.sack_freq
> 1) {
2665 trans
->sackfreq
= params
.sack_freq
;
2666 trans
->param_flags
=
2667 sctp_spp_sackdelay_enable(trans
->param_flags
);
2675 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2677 * Applications can specify protocol parameters for the default association
2678 * initialization. The option name argument to setsockopt() and getsockopt()
2681 * Setting initialization parameters is effective only on an unconnected
2682 * socket (for UDP-style sockets only future associations are effected
2683 * by the change). With TCP-style sockets, this option is inherited by
2684 * sockets derived from a listener socket.
2686 static int sctp_setsockopt_initmsg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2688 struct sctp_initmsg sinit
;
2689 struct sctp_sock
*sp
= sctp_sk(sk
);
2691 if (optlen
!= sizeof(struct sctp_initmsg
))
2693 if (copy_from_user(&sinit
, optval
, optlen
))
2696 if (sinit
.sinit_num_ostreams
)
2697 sp
->initmsg
.sinit_num_ostreams
= sinit
.sinit_num_ostreams
;
2698 if (sinit
.sinit_max_instreams
)
2699 sp
->initmsg
.sinit_max_instreams
= sinit
.sinit_max_instreams
;
2700 if (sinit
.sinit_max_attempts
)
2701 sp
->initmsg
.sinit_max_attempts
= sinit
.sinit_max_attempts
;
2702 if (sinit
.sinit_max_init_timeo
)
2703 sp
->initmsg
.sinit_max_init_timeo
= sinit
.sinit_max_init_timeo
;
2709 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2711 * Applications that wish to use the sendto() system call may wish to
2712 * specify a default set of parameters that would normally be supplied
2713 * through the inclusion of ancillary data. This socket option allows
2714 * such an application to set the default sctp_sndrcvinfo structure.
2715 * The application that wishes to use this socket option simply passes
2716 * in to this call the sctp_sndrcvinfo structure defined in Section
2717 * 5.2.2) The input parameters accepted by this call include
2718 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2719 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2720 * to this call if the caller is using the UDP model.
2722 static int sctp_setsockopt_default_send_param(struct sock
*sk
,
2723 char __user
*optval
,
2724 unsigned int optlen
)
2726 struct sctp_sndrcvinfo info
;
2727 struct sctp_association
*asoc
;
2728 struct sctp_sock
*sp
= sctp_sk(sk
);
2730 if (optlen
!= sizeof(struct sctp_sndrcvinfo
))
2732 if (copy_from_user(&info
, optval
, optlen
))
2735 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
2736 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
2740 asoc
->default_stream
= info
.sinfo_stream
;
2741 asoc
->default_flags
= info
.sinfo_flags
;
2742 asoc
->default_ppid
= info
.sinfo_ppid
;
2743 asoc
->default_context
= info
.sinfo_context
;
2744 asoc
->default_timetolive
= info
.sinfo_timetolive
;
2746 sp
->default_stream
= info
.sinfo_stream
;
2747 sp
->default_flags
= info
.sinfo_flags
;
2748 sp
->default_ppid
= info
.sinfo_ppid
;
2749 sp
->default_context
= info
.sinfo_context
;
2750 sp
->default_timetolive
= info
.sinfo_timetolive
;
2756 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2758 * Requests that the local SCTP stack use the enclosed peer address as
2759 * the association primary. The enclosed address must be one of the
2760 * association peer's addresses.
2762 static int sctp_setsockopt_primary_addr(struct sock
*sk
, char __user
*optval
,
2763 unsigned int optlen
)
2765 struct sctp_prim prim
;
2766 struct sctp_transport
*trans
;
2768 if (optlen
!= sizeof(struct sctp_prim
))
2771 if (copy_from_user(&prim
, optval
, sizeof(struct sctp_prim
)))
2774 trans
= sctp_addr_id2transport(sk
, &prim
.ssp_addr
, prim
.ssp_assoc_id
);
2778 sctp_assoc_set_primary(trans
->asoc
, trans
);
2784 * 7.1.5 SCTP_NODELAY
2786 * Turn on/off any Nagle-like algorithm. This means that packets are
2787 * generally sent as soon as possible and no unnecessary delays are
2788 * introduced, at the cost of more packets in the network. Expects an
2789 * integer boolean flag.
2791 static int sctp_setsockopt_nodelay(struct sock
*sk
, char __user
*optval
,
2792 unsigned int optlen
)
2796 if (optlen
< sizeof(int))
2798 if (get_user(val
, (int __user
*)optval
))
2801 sctp_sk(sk
)->nodelay
= (val
== 0) ? 0 : 1;
2807 * 7.1.1 SCTP_RTOINFO
2809 * The protocol parameters used to initialize and bound retransmission
2810 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2811 * and modify these parameters.
2812 * All parameters are time values, in milliseconds. A value of 0, when
2813 * modifying the parameters, indicates that the current value should not
2817 static int sctp_setsockopt_rtoinfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2819 struct sctp_rtoinfo rtoinfo
;
2820 struct sctp_association
*asoc
;
2821 unsigned long rto_min
, rto_max
;
2822 struct sctp_sock
*sp
= sctp_sk(sk
);
2824 if (optlen
!= sizeof (struct sctp_rtoinfo
))
2827 if (copy_from_user(&rtoinfo
, optval
, optlen
))
2830 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
2832 /* Set the values to the specific association */
2833 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
2836 rto_max
= rtoinfo
.srto_max
;
2837 rto_min
= rtoinfo
.srto_min
;
2840 rto_max
= asoc
? msecs_to_jiffies(rto_max
) : rto_max
;
2842 rto_max
= asoc
? asoc
->rto_max
: sp
->rtoinfo
.srto_max
;
2845 rto_min
= asoc
? msecs_to_jiffies(rto_min
) : rto_min
;
2847 rto_min
= asoc
? asoc
->rto_min
: sp
->rtoinfo
.srto_min
;
2849 if (rto_min
> rto_max
)
2853 if (rtoinfo
.srto_initial
!= 0)
2855 msecs_to_jiffies(rtoinfo
.srto_initial
);
2856 asoc
->rto_max
= rto_max
;
2857 asoc
->rto_min
= rto_min
;
2859 /* If there is no association or the association-id = 0
2860 * set the values to the endpoint.
2862 if (rtoinfo
.srto_initial
!= 0)
2863 sp
->rtoinfo
.srto_initial
= rtoinfo
.srto_initial
;
2864 sp
->rtoinfo
.srto_max
= rto_max
;
2865 sp
->rtoinfo
.srto_min
= rto_min
;
2873 * 7.1.2 SCTP_ASSOCINFO
2875 * This option is used to tune the maximum retransmission attempts
2876 * of the association.
2877 * Returns an error if the new association retransmission value is
2878 * greater than the sum of the retransmission value of the peer.
2879 * See [SCTP] for more information.
2882 static int sctp_setsockopt_associnfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2885 struct sctp_assocparams assocparams
;
2886 struct sctp_association
*asoc
;
2888 if (optlen
!= sizeof(struct sctp_assocparams
))
2890 if (copy_from_user(&assocparams
, optval
, optlen
))
2893 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
2895 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
2898 /* Set the values to the specific association */
2900 if (assocparams
.sasoc_asocmaxrxt
!= 0) {
2903 struct sctp_transport
*peer_addr
;
2905 list_for_each_entry(peer_addr
, &asoc
->peer
.transport_addr_list
,
2907 path_sum
+= peer_addr
->pathmaxrxt
;
2911 /* Only validate asocmaxrxt if we have more than
2912 * one path/transport. We do this because path
2913 * retransmissions are only counted when we have more
2917 assocparams
.sasoc_asocmaxrxt
> path_sum
)
2920 asoc
->max_retrans
= assocparams
.sasoc_asocmaxrxt
;
2923 if (assocparams
.sasoc_cookie_life
!= 0)
2924 asoc
->cookie_life
= ms_to_ktime(assocparams
.sasoc_cookie_life
);
2926 /* Set the values to the endpoint */
2927 struct sctp_sock
*sp
= sctp_sk(sk
);
2929 if (assocparams
.sasoc_asocmaxrxt
!= 0)
2930 sp
->assocparams
.sasoc_asocmaxrxt
=
2931 assocparams
.sasoc_asocmaxrxt
;
2932 if (assocparams
.sasoc_cookie_life
!= 0)
2933 sp
->assocparams
.sasoc_cookie_life
=
2934 assocparams
.sasoc_cookie_life
;
2940 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2942 * This socket option is a boolean flag which turns on or off mapped V4
2943 * addresses. If this option is turned on and the socket is type
2944 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2945 * If this option is turned off, then no mapping will be done of V4
2946 * addresses and a user will receive both PF_INET6 and PF_INET type
2947 * addresses on the socket.
2949 static int sctp_setsockopt_mappedv4(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2952 struct sctp_sock
*sp
= sctp_sk(sk
);
2954 if (optlen
< sizeof(int))
2956 if (get_user(val
, (int __user
*)optval
))
2967 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2968 * This option will get or set the maximum size to put in any outgoing
2969 * SCTP DATA chunk. If a message is larger than this size it will be
2970 * fragmented by SCTP into the specified size. Note that the underlying
2971 * SCTP implementation may fragment into smaller sized chunks when the
2972 * PMTU of the underlying association is smaller than the value set by
2973 * the user. The default value for this option is '0' which indicates
2974 * the user is NOT limiting fragmentation and only the PMTU will effect
2975 * SCTP's choice of DATA chunk size. Note also that values set larger
2976 * than the maximum size of an IP datagram will effectively let SCTP
2977 * control fragmentation (i.e. the same as setting this option to 0).
2979 * The following structure is used to access and modify this parameter:
2981 * struct sctp_assoc_value {
2982 * sctp_assoc_t assoc_id;
2983 * uint32_t assoc_value;
2986 * assoc_id: This parameter is ignored for one-to-one style sockets.
2987 * For one-to-many style sockets this parameter indicates which
2988 * association the user is performing an action upon. Note that if
2989 * this field's value is zero then the endpoints default value is
2990 * changed (effecting future associations only).
2991 * assoc_value: This parameter specifies the maximum size in bytes.
2993 static int sctp_setsockopt_maxseg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2995 struct sctp_assoc_value params
;
2996 struct sctp_association
*asoc
;
2997 struct sctp_sock
*sp
= sctp_sk(sk
);
3000 if (optlen
== sizeof(int)) {
3001 pr_warn_ratelimited(DEPRECATED
3003 "Use of int in maxseg socket option.\n"
3004 "Use struct sctp_assoc_value instead\n",
3005 current
->comm
, task_pid_nr(current
));
3006 if (copy_from_user(&val
, optval
, optlen
))
3008 params
.assoc_id
= 0;
3009 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
3010 if (copy_from_user(¶ms
, optval
, optlen
))
3012 val
= params
.assoc_value
;
3016 if ((val
!= 0) && ((val
< 8) || (val
> SCTP_MAX_CHUNK_LEN
)))
3019 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3020 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
3025 val
= asoc
->pathmtu
;
3026 val
-= sp
->pf
->af
->net_header_len
;
3027 val
-= sizeof(struct sctphdr
) +
3028 sizeof(struct sctp_data_chunk
);
3030 asoc
->user_frag
= val
;
3031 asoc
->frag_point
= sctp_frag_point(asoc
, asoc
->pathmtu
);
3033 sp
->user_frag
= val
;
3041 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3043 * Requests that the peer mark the enclosed address as the association
3044 * primary. The enclosed address must be one of the association's
3045 * locally bound addresses. The following structure is used to make a
3046 * set primary request:
3048 static int sctp_setsockopt_peer_primary_addr(struct sock
*sk
, char __user
*optval
,
3049 unsigned int optlen
)
3051 struct net
*net
= sock_net(sk
);
3052 struct sctp_sock
*sp
;
3053 struct sctp_association
*asoc
= NULL
;
3054 struct sctp_setpeerprim prim
;
3055 struct sctp_chunk
*chunk
;
3061 if (!net
->sctp
.addip_enable
)
3064 if (optlen
!= sizeof(struct sctp_setpeerprim
))
3067 if (copy_from_user(&prim
, optval
, optlen
))
3070 asoc
= sctp_id2assoc(sk
, prim
.sspp_assoc_id
);
3074 if (!asoc
->peer
.asconf_capable
)
3077 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_SET_PRIMARY
)
3080 if (!sctp_state(asoc
, ESTABLISHED
))
3083 af
= sctp_get_af_specific(prim
.sspp_addr
.ss_family
);
3087 if (!af
->addr_valid((union sctp_addr
*)&prim
.sspp_addr
, sp
, NULL
))
3088 return -EADDRNOTAVAIL
;
3090 if (!sctp_assoc_lookup_laddr(asoc
, (union sctp_addr
*)&prim
.sspp_addr
))
3091 return -EADDRNOTAVAIL
;
3093 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3094 chunk
= sctp_make_asconf_set_prim(asoc
,
3095 (union sctp_addr
*)&prim
.sspp_addr
);
3099 err
= sctp_send_asconf(asoc
, chunk
);
3101 pr_debug("%s: we set peer primary addr primitively\n", __func__
);
3106 static int sctp_setsockopt_adaptation_layer(struct sock
*sk
, char __user
*optval
,
3107 unsigned int optlen
)
3109 struct sctp_setadaptation adaptation
;
3111 if (optlen
!= sizeof(struct sctp_setadaptation
))
3113 if (copy_from_user(&adaptation
, optval
, optlen
))
3116 sctp_sk(sk
)->adaptation_ind
= adaptation
.ssb_adaptation_ind
;
3122 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3124 * The context field in the sctp_sndrcvinfo structure is normally only
3125 * used when a failed message is retrieved holding the value that was
3126 * sent down on the actual send call. This option allows the setting of
3127 * a default context on an association basis that will be received on
3128 * reading messages from the peer. This is especially helpful in the
3129 * one-2-many model for an application to keep some reference to an
3130 * internal state machine that is processing messages on the
3131 * association. Note that the setting of this value only effects
3132 * received messages from the peer and does not effect the value that is
3133 * saved with outbound messages.
3135 static int sctp_setsockopt_context(struct sock
*sk
, char __user
*optval
,
3136 unsigned int optlen
)
3138 struct sctp_assoc_value params
;
3139 struct sctp_sock
*sp
;
3140 struct sctp_association
*asoc
;
3142 if (optlen
!= sizeof(struct sctp_assoc_value
))
3144 if (copy_from_user(¶ms
, optval
, optlen
))
3149 if (params
.assoc_id
!= 0) {
3150 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3153 asoc
->default_rcv_context
= params
.assoc_value
;
3155 sp
->default_rcv_context
= params
.assoc_value
;
3162 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3164 * This options will at a minimum specify if the implementation is doing
3165 * fragmented interleave. Fragmented interleave, for a one to many
3166 * socket, is when subsequent calls to receive a message may return
3167 * parts of messages from different associations. Some implementations
3168 * may allow you to turn this value on or off. If so, when turned off,
3169 * no fragment interleave will occur (which will cause a head of line
3170 * blocking amongst multiple associations sharing the same one to many
3171 * socket). When this option is turned on, then each receive call may
3172 * come from a different association (thus the user must receive data
3173 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3174 * association each receive belongs to.
3176 * This option takes a boolean value. A non-zero value indicates that
3177 * fragmented interleave is on. A value of zero indicates that
3178 * fragmented interleave is off.
3180 * Note that it is important that an implementation that allows this
3181 * option to be turned on, have it off by default. Otherwise an unaware
3182 * application using the one to many model may become confused and act
3185 static int sctp_setsockopt_fragment_interleave(struct sock
*sk
,
3186 char __user
*optval
,
3187 unsigned int optlen
)
3191 if (optlen
!= sizeof(int))
3193 if (get_user(val
, (int __user
*)optval
))
3196 sctp_sk(sk
)->frag_interleave
= (val
== 0) ? 0 : 1;
3202 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3203 * (SCTP_PARTIAL_DELIVERY_POINT)
3205 * This option will set or get the SCTP partial delivery point. This
3206 * point is the size of a message where the partial delivery API will be
3207 * invoked to help free up rwnd space for the peer. Setting this to a
3208 * lower value will cause partial deliveries to happen more often. The
3209 * calls argument is an integer that sets or gets the partial delivery
3210 * point. Note also that the call will fail if the user attempts to set
3211 * this value larger than the socket receive buffer size.
3213 * Note that any single message having a length smaller than or equal to
3214 * the SCTP partial delivery point will be delivered in one single read
3215 * call as long as the user provided buffer is large enough to hold the
3218 static int sctp_setsockopt_partial_delivery_point(struct sock
*sk
,
3219 char __user
*optval
,
3220 unsigned int optlen
)
3224 if (optlen
!= sizeof(u32
))
3226 if (get_user(val
, (int __user
*)optval
))
3229 /* Note: We double the receive buffer from what the user sets
3230 * it to be, also initial rwnd is based on rcvbuf/2.
3232 if (val
> (sk
->sk_rcvbuf
>> 1))
3235 sctp_sk(sk
)->pd_point
= val
;
3237 return 0; /* is this the right error code? */
3241 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3243 * This option will allow a user to change the maximum burst of packets
3244 * that can be emitted by this association. Note that the default value
3245 * is 4, and some implementations may restrict this setting so that it
3246 * can only be lowered.
3248 * NOTE: This text doesn't seem right. Do this on a socket basis with
3249 * future associations inheriting the socket value.
3251 static int sctp_setsockopt_maxburst(struct sock
*sk
,
3252 char __user
*optval
,
3253 unsigned int optlen
)
3255 struct sctp_assoc_value params
;
3256 struct sctp_sock
*sp
;
3257 struct sctp_association
*asoc
;
3261 if (optlen
== sizeof(int)) {
3262 pr_warn_ratelimited(DEPRECATED
3264 "Use of int in max_burst socket option deprecated.\n"
3265 "Use struct sctp_assoc_value instead\n",
3266 current
->comm
, task_pid_nr(current
));
3267 if (copy_from_user(&val
, optval
, optlen
))
3269 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
3270 if (copy_from_user(¶ms
, optval
, optlen
))
3272 val
= params
.assoc_value
;
3273 assoc_id
= params
.assoc_id
;
3279 if (assoc_id
!= 0) {
3280 asoc
= sctp_id2assoc(sk
, assoc_id
);
3283 asoc
->max_burst
= val
;
3285 sp
->max_burst
= val
;
3291 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3293 * This set option adds a chunk type that the user is requesting to be
3294 * received only in an authenticated way. Changes to the list of chunks
3295 * will only effect future associations on the socket.
3297 static int sctp_setsockopt_auth_chunk(struct sock
*sk
,
3298 char __user
*optval
,
3299 unsigned int optlen
)
3301 struct net
*net
= sock_net(sk
);
3302 struct sctp_authchunk val
;
3304 if (!net
->sctp
.auth_enable
)
3307 if (optlen
!= sizeof(struct sctp_authchunk
))
3309 if (copy_from_user(&val
, optval
, optlen
))
3312 switch (val
.sauth_chunk
) {
3314 case SCTP_CID_INIT_ACK
:
3315 case SCTP_CID_SHUTDOWN_COMPLETE
:
3320 /* add this chunk id to the endpoint */
3321 return sctp_auth_ep_add_chunkid(sctp_sk(sk
)->ep
, val
.sauth_chunk
);
3325 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3327 * This option gets or sets the list of HMAC algorithms that the local
3328 * endpoint requires the peer to use.
3330 static int sctp_setsockopt_hmac_ident(struct sock
*sk
,
3331 char __user
*optval
,
3332 unsigned int optlen
)
3334 struct net
*net
= sock_net(sk
);
3335 struct sctp_hmacalgo
*hmacs
;
3339 if (!net
->sctp
.auth_enable
)
3342 if (optlen
< sizeof(struct sctp_hmacalgo
))
3345 hmacs
= memdup_user(optval
, optlen
);
3347 return PTR_ERR(hmacs
);
3349 idents
= hmacs
->shmac_num_idents
;
3350 if (idents
== 0 || idents
> SCTP_AUTH_NUM_HMACS
||
3351 (idents
* sizeof(u16
)) > (optlen
- sizeof(struct sctp_hmacalgo
))) {
3356 err
= sctp_auth_ep_set_hmacs(sctp_sk(sk
)->ep
, hmacs
);
3363 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3365 * This option will set a shared secret key which is used to build an
3366 * association shared key.
3368 static int sctp_setsockopt_auth_key(struct sock
*sk
,
3369 char __user
*optval
,
3370 unsigned int optlen
)
3372 struct net
*net
= sock_net(sk
);
3373 struct sctp_authkey
*authkey
;
3374 struct sctp_association
*asoc
;
3377 if (!net
->sctp
.auth_enable
)
3380 if (optlen
<= sizeof(struct sctp_authkey
))
3383 authkey
= memdup_user(optval
, optlen
);
3384 if (IS_ERR(authkey
))
3385 return PTR_ERR(authkey
);
3387 if (authkey
->sca_keylength
> optlen
- sizeof(struct sctp_authkey
)) {
3392 asoc
= sctp_id2assoc(sk
, authkey
->sca_assoc_id
);
3393 if (!asoc
&& authkey
->sca_assoc_id
&& sctp_style(sk
, UDP
)) {
3398 ret
= sctp_auth_set_key(sctp_sk(sk
)->ep
, asoc
, authkey
);
3405 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3407 * This option will get or set the active shared key to be used to build
3408 * the association shared key.
3410 static int sctp_setsockopt_active_key(struct sock
*sk
,
3411 char __user
*optval
,
3412 unsigned int optlen
)
3414 struct net
*net
= sock_net(sk
);
3415 struct sctp_authkeyid val
;
3416 struct sctp_association
*asoc
;
3418 if (!net
->sctp
.auth_enable
)
3421 if (optlen
!= sizeof(struct sctp_authkeyid
))
3423 if (copy_from_user(&val
, optval
, optlen
))
3426 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3427 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3430 return sctp_auth_set_active_key(sctp_sk(sk
)->ep
, asoc
,
3431 val
.scact_keynumber
);
3435 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3437 * This set option will delete a shared secret key from use.
3439 static int sctp_setsockopt_del_key(struct sock
*sk
,
3440 char __user
*optval
,
3441 unsigned int optlen
)
3443 struct net
*net
= sock_net(sk
);
3444 struct sctp_authkeyid val
;
3445 struct sctp_association
*asoc
;
3447 if (!net
->sctp
.auth_enable
)
3450 if (optlen
!= sizeof(struct sctp_authkeyid
))
3452 if (copy_from_user(&val
, optval
, optlen
))
3455 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3456 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3459 return sctp_auth_del_key_id(sctp_sk(sk
)->ep
, asoc
,
3460 val
.scact_keynumber
);
3465 * 8.1.23 SCTP_AUTO_ASCONF
3467 * This option will enable or disable the use of the automatic generation of
3468 * ASCONF chunks to add and delete addresses to an existing association. Note
3469 * that this option has two caveats namely: a) it only affects sockets that
3470 * are bound to all addresses available to the SCTP stack, and b) the system
3471 * administrator may have an overriding control that turns the ASCONF feature
3472 * off no matter what setting the socket option may have.
3473 * This option expects an integer boolean flag, where a non-zero value turns on
3474 * the option, and a zero value turns off the option.
3475 * Note. In this implementation, socket operation overrides default parameter
3476 * being set by sysctl as well as FreeBSD implementation
3478 static int sctp_setsockopt_auto_asconf(struct sock
*sk
, char __user
*optval
,
3479 unsigned int optlen
)
3482 struct sctp_sock
*sp
= sctp_sk(sk
);
3484 if (optlen
< sizeof(int))
3486 if (get_user(val
, (int __user
*)optval
))
3488 if (!sctp_is_ep_boundall(sk
) && val
)
3490 if ((val
&& sp
->do_auto_asconf
) || (!val
&& !sp
->do_auto_asconf
))
3493 if (val
== 0 && sp
->do_auto_asconf
) {
3494 list_del(&sp
->auto_asconf_list
);
3495 sp
->do_auto_asconf
= 0;
3496 } else if (val
&& !sp
->do_auto_asconf
) {
3497 list_add_tail(&sp
->auto_asconf_list
,
3498 &sock_net(sk
)->sctp
.auto_asconf_splist
);
3499 sp
->do_auto_asconf
= 1;
3506 * SCTP_PEER_ADDR_THLDS
3508 * This option allows us to alter the partially failed threshold for one or all
3509 * transports in an association. See Section 6.1 of:
3510 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3512 static int sctp_setsockopt_paddr_thresholds(struct sock
*sk
,
3513 char __user
*optval
,
3514 unsigned int optlen
)
3516 struct sctp_paddrthlds val
;
3517 struct sctp_transport
*trans
;
3518 struct sctp_association
*asoc
;
3520 if (optlen
< sizeof(struct sctp_paddrthlds
))
3522 if (copy_from_user(&val
, (struct sctp_paddrthlds __user
*)optval
,
3523 sizeof(struct sctp_paddrthlds
)))
3527 if (sctp_is_any(sk
, (const union sctp_addr
*)&val
.spt_address
)) {
3528 asoc
= sctp_id2assoc(sk
, val
.spt_assoc_id
);
3531 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
3533 if (val
.spt_pathmaxrxt
)
3534 trans
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3535 trans
->pf_retrans
= val
.spt_pathpfthld
;
3538 if (val
.spt_pathmaxrxt
)
3539 asoc
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3540 asoc
->pf_retrans
= val
.spt_pathpfthld
;
3542 trans
= sctp_addr_id2transport(sk
, &val
.spt_address
,
3547 if (val
.spt_pathmaxrxt
)
3548 trans
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3549 trans
->pf_retrans
= val
.spt_pathpfthld
;
3555 /* API 6.2 setsockopt(), getsockopt()
3557 * Applications use setsockopt() and getsockopt() to set or retrieve
3558 * socket options. Socket options are used to change the default
3559 * behavior of sockets calls. They are described in Section 7.
3563 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3564 * int __user *optlen);
3565 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3568 * sd - the socket descript.
3569 * level - set to IPPROTO_SCTP for all SCTP options.
3570 * optname - the option name.
3571 * optval - the buffer to store the value of the option.
3572 * optlen - the size of the buffer.
3574 static int sctp_setsockopt(struct sock
*sk
, int level
, int optname
,
3575 char __user
*optval
, unsigned int optlen
)
3579 pr_debug("%s: sk:%p, optname:%d\n", __func__
, sk
, optname
);
3581 /* I can hardly begin to describe how wrong this is. This is
3582 * so broken as to be worse than useless. The API draft
3583 * REALLY is NOT helpful here... I am not convinced that the
3584 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3585 * are at all well-founded.
3587 if (level
!= SOL_SCTP
) {
3588 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
3589 retval
= af
->setsockopt(sk
, level
, optname
, optval
, optlen
);
3596 case SCTP_SOCKOPT_BINDX_ADD
:
3597 /* 'optlen' is the size of the addresses buffer. */
3598 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3599 optlen
, SCTP_BINDX_ADD_ADDR
);
3602 case SCTP_SOCKOPT_BINDX_REM
:
3603 /* 'optlen' is the size of the addresses buffer. */
3604 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3605 optlen
, SCTP_BINDX_REM_ADDR
);
3608 case SCTP_SOCKOPT_CONNECTX_OLD
:
3609 /* 'optlen' is the size of the addresses buffer. */
3610 retval
= sctp_setsockopt_connectx_old(sk
,
3611 (struct sockaddr __user
*)optval
,
3615 case SCTP_SOCKOPT_CONNECTX
:
3616 /* 'optlen' is the size of the addresses buffer. */
3617 retval
= sctp_setsockopt_connectx(sk
,
3618 (struct sockaddr __user
*)optval
,
3622 case SCTP_DISABLE_FRAGMENTS
:
3623 retval
= sctp_setsockopt_disable_fragments(sk
, optval
, optlen
);
3627 retval
= sctp_setsockopt_events(sk
, optval
, optlen
);
3630 case SCTP_AUTOCLOSE
:
3631 retval
= sctp_setsockopt_autoclose(sk
, optval
, optlen
);
3634 case SCTP_PEER_ADDR_PARAMS
:
3635 retval
= sctp_setsockopt_peer_addr_params(sk
, optval
, optlen
);
3638 case SCTP_DELAYED_SACK
:
3639 retval
= sctp_setsockopt_delayed_ack(sk
, optval
, optlen
);
3641 case SCTP_PARTIAL_DELIVERY_POINT
:
3642 retval
= sctp_setsockopt_partial_delivery_point(sk
, optval
, optlen
);
3646 retval
= sctp_setsockopt_initmsg(sk
, optval
, optlen
);
3648 case SCTP_DEFAULT_SEND_PARAM
:
3649 retval
= sctp_setsockopt_default_send_param(sk
, optval
,
3652 case SCTP_PRIMARY_ADDR
:
3653 retval
= sctp_setsockopt_primary_addr(sk
, optval
, optlen
);
3655 case SCTP_SET_PEER_PRIMARY_ADDR
:
3656 retval
= sctp_setsockopt_peer_primary_addr(sk
, optval
, optlen
);
3659 retval
= sctp_setsockopt_nodelay(sk
, optval
, optlen
);
3662 retval
= sctp_setsockopt_rtoinfo(sk
, optval
, optlen
);
3664 case SCTP_ASSOCINFO
:
3665 retval
= sctp_setsockopt_associnfo(sk
, optval
, optlen
);
3667 case SCTP_I_WANT_MAPPED_V4_ADDR
:
3668 retval
= sctp_setsockopt_mappedv4(sk
, optval
, optlen
);
3671 retval
= sctp_setsockopt_maxseg(sk
, optval
, optlen
);
3673 case SCTP_ADAPTATION_LAYER
:
3674 retval
= sctp_setsockopt_adaptation_layer(sk
, optval
, optlen
);
3677 retval
= sctp_setsockopt_context(sk
, optval
, optlen
);
3679 case SCTP_FRAGMENT_INTERLEAVE
:
3680 retval
= sctp_setsockopt_fragment_interleave(sk
, optval
, optlen
);
3682 case SCTP_MAX_BURST
:
3683 retval
= sctp_setsockopt_maxburst(sk
, optval
, optlen
);
3685 case SCTP_AUTH_CHUNK
:
3686 retval
= sctp_setsockopt_auth_chunk(sk
, optval
, optlen
);
3688 case SCTP_HMAC_IDENT
:
3689 retval
= sctp_setsockopt_hmac_ident(sk
, optval
, optlen
);
3692 retval
= sctp_setsockopt_auth_key(sk
, optval
, optlen
);
3694 case SCTP_AUTH_ACTIVE_KEY
:
3695 retval
= sctp_setsockopt_active_key(sk
, optval
, optlen
);
3697 case SCTP_AUTH_DELETE_KEY
:
3698 retval
= sctp_setsockopt_del_key(sk
, optval
, optlen
);
3700 case SCTP_AUTO_ASCONF
:
3701 retval
= sctp_setsockopt_auto_asconf(sk
, optval
, optlen
);
3703 case SCTP_PEER_ADDR_THLDS
:
3704 retval
= sctp_setsockopt_paddr_thresholds(sk
, optval
, optlen
);
3707 retval
= -ENOPROTOOPT
;
3717 /* API 3.1.6 connect() - UDP Style Syntax
3719 * An application may use the connect() call in the UDP model to initiate an
3720 * association without sending data.
3724 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3726 * sd: the socket descriptor to have a new association added to.
3728 * nam: the address structure (either struct sockaddr_in or struct
3729 * sockaddr_in6 defined in RFC2553 [7]).
3731 * len: the size of the address.
3733 static int sctp_connect(struct sock
*sk
, struct sockaddr
*addr
,
3741 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__
, sk
,
3744 /* Validate addr_len before calling common connect/connectx routine. */
3745 af
= sctp_get_af_specific(addr
->sa_family
);
3746 if (!af
|| addr_len
< af
->sockaddr_len
) {
3749 /* Pass correct addr len to common routine (so it knows there
3750 * is only one address being passed.
3752 err
= __sctp_connect(sk
, addr
, af
->sockaddr_len
, NULL
);
3759 /* FIXME: Write comments. */
3760 static int sctp_disconnect(struct sock
*sk
, int flags
)
3762 return -EOPNOTSUPP
; /* STUB */
3765 /* 4.1.4 accept() - TCP Style Syntax
3767 * Applications use accept() call to remove an established SCTP
3768 * association from the accept queue of the endpoint. A new socket
3769 * descriptor will be returned from accept() to represent the newly
3770 * formed association.
3772 static struct sock
*sctp_accept(struct sock
*sk
, int flags
, int *err
)
3774 struct sctp_sock
*sp
;
3775 struct sctp_endpoint
*ep
;
3776 struct sock
*newsk
= NULL
;
3777 struct sctp_association
*asoc
;
3786 if (!sctp_style(sk
, TCP
)) {
3787 error
= -EOPNOTSUPP
;
3791 if (!sctp_sstate(sk
, LISTENING
)) {
3796 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
3798 error
= sctp_wait_for_accept(sk
, timeo
);
3802 /* We treat the list of associations on the endpoint as the accept
3803 * queue and pick the first association on the list.
3805 asoc
= list_entry(ep
->asocs
.next
, struct sctp_association
, asocs
);
3807 newsk
= sp
->pf
->create_accept_sk(sk
, asoc
);
3813 /* Populate the fields of the newsk from the oldsk and migrate the
3814 * asoc to the newsk.
3816 sctp_sock_migrate(sk
, newsk
, asoc
, SCTP_SOCKET_TCP
);
3824 /* The SCTP ioctl handler. */
3825 static int sctp_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
3832 * SEQPACKET-style sockets in LISTENING state are valid, for
3833 * SCTP, so only discard TCP-style sockets in LISTENING state.
3835 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
3840 struct sk_buff
*skb
;
3841 unsigned int amount
= 0;
3843 skb
= skb_peek(&sk
->sk_receive_queue
);
3846 * We will only return the amount of this packet since
3847 * that is all that will be read.
3851 rc
= put_user(amount
, (int __user
*)arg
);
3863 /* This is the function which gets called during socket creation to
3864 * initialized the SCTP-specific portion of the sock.
3865 * The sock structure should already be zero-filled memory.
3867 static int sctp_init_sock(struct sock
*sk
)
3869 struct net
*net
= sock_net(sk
);
3870 struct sctp_sock
*sp
;
3872 pr_debug("%s: sk:%p\n", __func__
, sk
);
3876 /* Initialize the SCTP per socket area. */
3877 switch (sk
->sk_type
) {
3878 case SOCK_SEQPACKET
:
3879 sp
->type
= SCTP_SOCKET_UDP
;
3882 sp
->type
= SCTP_SOCKET_TCP
;
3885 return -ESOCKTNOSUPPORT
;
3888 /* Initialize default send parameters. These parameters can be
3889 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3891 sp
->default_stream
= 0;
3892 sp
->default_ppid
= 0;
3893 sp
->default_flags
= 0;
3894 sp
->default_context
= 0;
3895 sp
->default_timetolive
= 0;
3897 sp
->default_rcv_context
= 0;
3898 sp
->max_burst
= net
->sctp
.max_burst
;
3900 sp
->sctp_hmac_alg
= net
->sctp
.sctp_hmac_alg
;
3902 /* Initialize default setup parameters. These parameters
3903 * can be modified with the SCTP_INITMSG socket option or
3904 * overridden by the SCTP_INIT CMSG.
3906 sp
->initmsg
.sinit_num_ostreams
= sctp_max_outstreams
;
3907 sp
->initmsg
.sinit_max_instreams
= sctp_max_instreams
;
3908 sp
->initmsg
.sinit_max_attempts
= net
->sctp
.max_retrans_init
;
3909 sp
->initmsg
.sinit_max_init_timeo
= net
->sctp
.rto_max
;
3911 /* Initialize default RTO related parameters. These parameters can
3912 * be modified for with the SCTP_RTOINFO socket option.
3914 sp
->rtoinfo
.srto_initial
= net
->sctp
.rto_initial
;
3915 sp
->rtoinfo
.srto_max
= net
->sctp
.rto_max
;
3916 sp
->rtoinfo
.srto_min
= net
->sctp
.rto_min
;
3918 /* Initialize default association related parameters. These parameters
3919 * can be modified with the SCTP_ASSOCINFO socket option.
3921 sp
->assocparams
.sasoc_asocmaxrxt
= net
->sctp
.max_retrans_association
;
3922 sp
->assocparams
.sasoc_number_peer_destinations
= 0;
3923 sp
->assocparams
.sasoc_peer_rwnd
= 0;
3924 sp
->assocparams
.sasoc_local_rwnd
= 0;
3925 sp
->assocparams
.sasoc_cookie_life
= net
->sctp
.valid_cookie_life
;
3927 /* Initialize default event subscriptions. By default, all the
3930 memset(&sp
->subscribe
, 0, sizeof(struct sctp_event_subscribe
));
3932 /* Default Peer Address Parameters. These defaults can
3933 * be modified via SCTP_PEER_ADDR_PARAMS
3935 sp
->hbinterval
= net
->sctp
.hb_interval
;
3936 sp
->pathmaxrxt
= net
->sctp
.max_retrans_path
;
3937 sp
->pathmtu
= 0; /* allow default discovery */
3938 sp
->sackdelay
= net
->sctp
.sack_timeout
;
3940 sp
->param_flags
= SPP_HB_ENABLE
|
3942 SPP_SACKDELAY_ENABLE
;
3944 /* If enabled no SCTP message fragmentation will be performed.
3945 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3947 sp
->disable_fragments
= 0;
3949 /* Enable Nagle algorithm by default. */
3952 /* Enable by default. */
3955 /* Auto-close idle associations after the configured
3956 * number of seconds. A value of 0 disables this
3957 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3958 * for UDP-style sockets only.
3962 /* User specified fragmentation limit. */
3965 sp
->adaptation_ind
= 0;
3967 sp
->pf
= sctp_get_pf_specific(sk
->sk_family
);
3969 /* Control variables for partial data delivery. */
3970 atomic_set(&sp
->pd_mode
, 0);
3971 skb_queue_head_init(&sp
->pd_lobby
);
3972 sp
->frag_interleave
= 0;
3974 /* Create a per socket endpoint structure. Even if we
3975 * change the data structure relationships, this may still
3976 * be useful for storing pre-connect address information.
3978 sp
->ep
= sctp_endpoint_new(sk
, GFP_KERNEL
);
3984 sk
->sk_destruct
= sctp_destruct_sock
;
3986 SCTP_DBG_OBJCNT_INC(sock
);
3989 percpu_counter_inc(&sctp_sockets_allocated
);
3990 sock_prot_inuse_add(net
, sk
->sk_prot
, 1);
3991 if (net
->sctp
.default_auto_asconf
) {
3992 list_add_tail(&sp
->auto_asconf_list
,
3993 &net
->sctp
.auto_asconf_splist
);
3994 sp
->do_auto_asconf
= 1;
3996 sp
->do_auto_asconf
= 0;
4002 /* Cleanup any SCTP per socket resources. */
4003 static void sctp_destroy_sock(struct sock
*sk
)
4005 struct sctp_sock
*sp
;
4007 pr_debug("%s: sk:%p\n", __func__
, sk
);
4009 /* Release our hold on the endpoint. */
4011 /* This could happen during socket init, thus we bail out
4012 * early, since the rest of the below is not setup either.
4017 if (sp
->do_auto_asconf
) {
4018 sp
->do_auto_asconf
= 0;
4019 list_del(&sp
->auto_asconf_list
);
4021 sctp_endpoint_free(sp
->ep
);
4023 percpu_counter_dec(&sctp_sockets_allocated
);
4024 sock_prot_inuse_add(sock_net(sk
), sk
->sk_prot
, -1);
4028 /* Triggered when there are no references on the socket anymore */
4029 static void sctp_destruct_sock(struct sock
*sk
)
4031 struct sctp_sock
*sp
= sctp_sk(sk
);
4033 /* Free up the HMAC transform. */
4034 crypto_free_hash(sp
->hmac
);
4036 inet_sock_destruct(sk
);
4039 /* API 4.1.7 shutdown() - TCP Style Syntax
4040 * int shutdown(int socket, int how);
4042 * sd - the socket descriptor of the association to be closed.
4043 * how - Specifies the type of shutdown. The values are
4046 * Disables further receive operations. No SCTP
4047 * protocol action is taken.
4049 * Disables further send operations, and initiates
4050 * the SCTP shutdown sequence.
4052 * Disables further send and receive operations
4053 * and initiates the SCTP shutdown sequence.
4055 static void sctp_shutdown(struct sock
*sk
, int how
)
4057 struct net
*net
= sock_net(sk
);
4058 struct sctp_endpoint
*ep
;
4059 struct sctp_association
*asoc
;
4061 if (!sctp_style(sk
, TCP
))
4064 if (how
& SEND_SHUTDOWN
) {
4065 ep
= sctp_sk(sk
)->ep
;
4066 if (!list_empty(&ep
->asocs
)) {
4067 asoc
= list_entry(ep
->asocs
.next
,
4068 struct sctp_association
, asocs
);
4069 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
4074 /* 7.2.1 Association Status (SCTP_STATUS)
4076 * Applications can retrieve current status information about an
4077 * association, including association state, peer receiver window size,
4078 * number of unacked data chunks, and number of data chunks pending
4079 * receipt. This information is read-only.
4081 static int sctp_getsockopt_sctp_status(struct sock
*sk
, int len
,
4082 char __user
*optval
,
4085 struct sctp_status status
;
4086 struct sctp_association
*asoc
= NULL
;
4087 struct sctp_transport
*transport
;
4088 sctp_assoc_t associd
;
4091 if (len
< sizeof(status
)) {
4096 len
= sizeof(status
);
4097 if (copy_from_user(&status
, optval
, len
)) {
4102 associd
= status
.sstat_assoc_id
;
4103 asoc
= sctp_id2assoc(sk
, associd
);
4109 transport
= asoc
->peer
.primary_path
;
4111 status
.sstat_assoc_id
= sctp_assoc2id(asoc
);
4112 status
.sstat_state
= asoc
->state
;
4113 status
.sstat_rwnd
= asoc
->peer
.rwnd
;
4114 status
.sstat_unackdata
= asoc
->unack_data
;
4116 status
.sstat_penddata
= sctp_tsnmap_pending(&asoc
->peer
.tsn_map
);
4117 status
.sstat_instrms
= asoc
->c
.sinit_max_instreams
;
4118 status
.sstat_outstrms
= asoc
->c
.sinit_num_ostreams
;
4119 status
.sstat_fragmentation_point
= asoc
->frag_point
;
4120 status
.sstat_primary
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
4121 memcpy(&status
.sstat_primary
.spinfo_address
, &transport
->ipaddr
,
4122 transport
->af_specific
->sockaddr_len
);
4123 /* Map ipv4 address into v4-mapped-on-v6 address. */
4124 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
4125 (union sctp_addr
*)&status
.sstat_primary
.spinfo_address
);
4126 status
.sstat_primary
.spinfo_state
= transport
->state
;
4127 status
.sstat_primary
.spinfo_cwnd
= transport
->cwnd
;
4128 status
.sstat_primary
.spinfo_srtt
= transport
->srtt
;
4129 status
.sstat_primary
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
4130 status
.sstat_primary
.spinfo_mtu
= transport
->pathmtu
;
4132 if (status
.sstat_primary
.spinfo_state
== SCTP_UNKNOWN
)
4133 status
.sstat_primary
.spinfo_state
= SCTP_ACTIVE
;
4135 if (put_user(len
, optlen
)) {
4140 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4141 __func__
, len
, status
.sstat_state
, status
.sstat_rwnd
,
4142 status
.sstat_assoc_id
);
4144 if (copy_to_user(optval
, &status
, len
)) {
4154 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4156 * Applications can retrieve information about a specific peer address
4157 * of an association, including its reachability state, congestion
4158 * window, and retransmission timer values. This information is
4161 static int sctp_getsockopt_peer_addr_info(struct sock
*sk
, int len
,
4162 char __user
*optval
,
4165 struct sctp_paddrinfo pinfo
;
4166 struct sctp_transport
*transport
;
4169 if (len
< sizeof(pinfo
)) {
4174 len
= sizeof(pinfo
);
4175 if (copy_from_user(&pinfo
, optval
, len
)) {
4180 transport
= sctp_addr_id2transport(sk
, &pinfo
.spinfo_address
,
4181 pinfo
.spinfo_assoc_id
);
4185 pinfo
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
4186 pinfo
.spinfo_state
= transport
->state
;
4187 pinfo
.spinfo_cwnd
= transport
->cwnd
;
4188 pinfo
.spinfo_srtt
= transport
->srtt
;
4189 pinfo
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
4190 pinfo
.spinfo_mtu
= transport
->pathmtu
;
4192 if (pinfo
.spinfo_state
== SCTP_UNKNOWN
)
4193 pinfo
.spinfo_state
= SCTP_ACTIVE
;
4195 if (put_user(len
, optlen
)) {
4200 if (copy_to_user(optval
, &pinfo
, len
)) {
4209 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4211 * This option is a on/off flag. If enabled no SCTP message
4212 * fragmentation will be performed. Instead if a message being sent
4213 * exceeds the current PMTU size, the message will NOT be sent and
4214 * instead a error will be indicated to the user.
4216 static int sctp_getsockopt_disable_fragments(struct sock
*sk
, int len
,
4217 char __user
*optval
, int __user
*optlen
)
4221 if (len
< sizeof(int))
4225 val
= (sctp_sk(sk
)->disable_fragments
== 1);
4226 if (put_user(len
, optlen
))
4228 if (copy_to_user(optval
, &val
, len
))
4233 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4235 * This socket option is used to specify various notifications and
4236 * ancillary data the user wishes to receive.
4238 static int sctp_getsockopt_events(struct sock
*sk
, int len
, char __user
*optval
,
4243 if (len
> sizeof(struct sctp_event_subscribe
))
4244 len
= sizeof(struct sctp_event_subscribe
);
4245 if (put_user(len
, optlen
))
4247 if (copy_to_user(optval
, &sctp_sk(sk
)->subscribe
, len
))
4252 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4254 * This socket option is applicable to the UDP-style socket only. When
4255 * set it will cause associations that are idle for more than the
4256 * specified number of seconds to automatically close. An association
4257 * being idle is defined an association that has NOT sent or received
4258 * user data. The special value of '0' indicates that no automatic
4259 * close of any associations should be performed. The option expects an
4260 * integer defining the number of seconds of idle time before an
4261 * association is closed.
4263 static int sctp_getsockopt_autoclose(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4265 /* Applicable to UDP-style socket only */
4266 if (sctp_style(sk
, TCP
))
4268 if (len
< sizeof(int))
4271 if (put_user(len
, optlen
))
4273 if (copy_to_user(optval
, &sctp_sk(sk
)->autoclose
, sizeof(int)))
4278 /* Helper routine to branch off an association to a new socket. */
4279 int sctp_do_peeloff(struct sock
*sk
, sctp_assoc_t id
, struct socket
**sockp
)
4281 struct sctp_association
*asoc
= sctp_id2assoc(sk
, id
);
4282 struct socket
*sock
;
4289 /* An association cannot be branched off from an already peeled-off
4290 * socket, nor is this supported for tcp style sockets.
4292 if (!sctp_style(sk
, UDP
))
4295 /* Create a new socket. */
4296 err
= sock_create(sk
->sk_family
, SOCK_SEQPACKET
, IPPROTO_SCTP
, &sock
);
4300 sctp_copy_sock(sock
->sk
, sk
, asoc
);
4302 /* Make peeled-off sockets more like 1-1 accepted sockets.
4303 * Set the daddr and initialize id to something more random
4305 af
= sctp_get_af_specific(asoc
->peer
.primary_addr
.sa
.sa_family
);
4306 af
->to_sk_daddr(&asoc
->peer
.primary_addr
, sk
);
4308 /* Populate the fields of the newsk from the oldsk and migrate the
4309 * asoc to the newsk.
4311 sctp_sock_migrate(sk
, sock
->sk
, asoc
, SCTP_SOCKET_UDP_HIGH_BANDWIDTH
);
4317 EXPORT_SYMBOL(sctp_do_peeloff
);
4319 static int sctp_getsockopt_peeloff(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4321 sctp_peeloff_arg_t peeloff
;
4322 struct socket
*newsock
;
4323 struct file
*newfile
;
4326 if (len
< sizeof(sctp_peeloff_arg_t
))
4328 len
= sizeof(sctp_peeloff_arg_t
);
4329 if (copy_from_user(&peeloff
, optval
, len
))
4332 retval
= sctp_do_peeloff(sk
, peeloff
.associd
, &newsock
);
4336 /* Map the socket to an unused fd that can be returned to the user. */
4337 retval
= get_unused_fd_flags(0);
4339 sock_release(newsock
);
4343 newfile
= sock_alloc_file(newsock
, 0, NULL
);
4344 if (unlikely(IS_ERR(newfile
))) {
4345 put_unused_fd(retval
);
4346 sock_release(newsock
);
4347 return PTR_ERR(newfile
);
4350 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__
, sk
, newsock
->sk
,
4353 /* Return the fd mapped to the new socket. */
4354 if (put_user(len
, optlen
)) {
4356 put_unused_fd(retval
);
4359 peeloff
.sd
= retval
;
4360 if (copy_to_user(optval
, &peeloff
, len
)) {
4362 put_unused_fd(retval
);
4365 fd_install(retval
, newfile
);
4370 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4372 * Applications can enable or disable heartbeats for any peer address of
4373 * an association, modify an address's heartbeat interval, force a
4374 * heartbeat to be sent immediately, and adjust the address's maximum
4375 * number of retransmissions sent before an address is considered
4376 * unreachable. The following structure is used to access and modify an
4377 * address's parameters:
4379 * struct sctp_paddrparams {
4380 * sctp_assoc_t spp_assoc_id;
4381 * struct sockaddr_storage spp_address;
4382 * uint32_t spp_hbinterval;
4383 * uint16_t spp_pathmaxrxt;
4384 * uint32_t spp_pathmtu;
4385 * uint32_t spp_sackdelay;
4386 * uint32_t spp_flags;
4389 * spp_assoc_id - (one-to-many style socket) This is filled in the
4390 * application, and identifies the association for
4392 * spp_address - This specifies which address is of interest.
4393 * spp_hbinterval - This contains the value of the heartbeat interval,
4394 * in milliseconds. If a value of zero
4395 * is present in this field then no changes are to
4396 * be made to this parameter.
4397 * spp_pathmaxrxt - This contains the maximum number of
4398 * retransmissions before this address shall be
4399 * considered unreachable. If a value of zero
4400 * is present in this field then no changes are to
4401 * be made to this parameter.
4402 * spp_pathmtu - When Path MTU discovery is disabled the value
4403 * specified here will be the "fixed" path mtu.
4404 * Note that if the spp_address field is empty
4405 * then all associations on this address will
4406 * have this fixed path mtu set upon them.
4408 * spp_sackdelay - When delayed sack is enabled, this value specifies
4409 * the number of milliseconds that sacks will be delayed
4410 * for. This value will apply to all addresses of an
4411 * association if the spp_address field is empty. Note
4412 * also, that if delayed sack is enabled and this
4413 * value is set to 0, no change is made to the last
4414 * recorded delayed sack timer value.
4416 * spp_flags - These flags are used to control various features
4417 * on an association. The flag field may contain
4418 * zero or more of the following options.
4420 * SPP_HB_ENABLE - Enable heartbeats on the
4421 * specified address. Note that if the address
4422 * field is empty all addresses for the association
4423 * have heartbeats enabled upon them.
4425 * SPP_HB_DISABLE - Disable heartbeats on the
4426 * speicifed address. Note that if the address
4427 * field is empty all addresses for the association
4428 * will have their heartbeats disabled. Note also
4429 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4430 * mutually exclusive, only one of these two should
4431 * be specified. Enabling both fields will have
4432 * undetermined results.
4434 * SPP_HB_DEMAND - Request a user initiated heartbeat
4435 * to be made immediately.
4437 * SPP_PMTUD_ENABLE - This field will enable PMTU
4438 * discovery upon the specified address. Note that
4439 * if the address feild is empty then all addresses
4440 * on the association are effected.
4442 * SPP_PMTUD_DISABLE - This field will disable PMTU
4443 * discovery upon the specified address. Note that
4444 * if the address feild is empty then all addresses
4445 * on the association are effected. Not also that
4446 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4447 * exclusive. Enabling both will have undetermined
4450 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4451 * on delayed sack. The time specified in spp_sackdelay
4452 * is used to specify the sack delay for this address. Note
4453 * that if spp_address is empty then all addresses will
4454 * enable delayed sack and take on the sack delay
4455 * value specified in spp_sackdelay.
4456 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4457 * off delayed sack. If the spp_address field is blank then
4458 * delayed sack is disabled for the entire association. Note
4459 * also that this field is mutually exclusive to
4460 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4463 static int sctp_getsockopt_peer_addr_params(struct sock
*sk
, int len
,
4464 char __user
*optval
, int __user
*optlen
)
4466 struct sctp_paddrparams params
;
4467 struct sctp_transport
*trans
= NULL
;
4468 struct sctp_association
*asoc
= NULL
;
4469 struct sctp_sock
*sp
= sctp_sk(sk
);
4471 if (len
< sizeof(struct sctp_paddrparams
))
4473 len
= sizeof(struct sctp_paddrparams
);
4474 if (copy_from_user(¶ms
, optval
, len
))
4477 /* If an address other than INADDR_ANY is specified, and
4478 * no transport is found, then the request is invalid.
4480 if (!sctp_is_any(sk
, (union sctp_addr
*)¶ms
.spp_address
)) {
4481 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
4482 params
.spp_assoc_id
);
4484 pr_debug("%s: failed no transport\n", __func__
);
4489 /* Get association, if assoc_id != 0 and the socket is a one
4490 * to many style socket, and an association was not found, then
4491 * the id was invalid.
4493 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
4494 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
)) {
4495 pr_debug("%s: failed no association\n", __func__
);
4500 /* Fetch transport values. */
4501 params
.spp_hbinterval
= jiffies_to_msecs(trans
->hbinterval
);
4502 params
.spp_pathmtu
= trans
->pathmtu
;
4503 params
.spp_pathmaxrxt
= trans
->pathmaxrxt
;
4504 params
.spp_sackdelay
= jiffies_to_msecs(trans
->sackdelay
);
4506 /*draft-11 doesn't say what to return in spp_flags*/
4507 params
.spp_flags
= trans
->param_flags
;
4509 /* Fetch association values. */
4510 params
.spp_hbinterval
= jiffies_to_msecs(asoc
->hbinterval
);
4511 params
.spp_pathmtu
= asoc
->pathmtu
;
4512 params
.spp_pathmaxrxt
= asoc
->pathmaxrxt
;
4513 params
.spp_sackdelay
= jiffies_to_msecs(asoc
->sackdelay
);
4515 /*draft-11 doesn't say what to return in spp_flags*/
4516 params
.spp_flags
= asoc
->param_flags
;
4518 /* Fetch socket values. */
4519 params
.spp_hbinterval
= sp
->hbinterval
;
4520 params
.spp_pathmtu
= sp
->pathmtu
;
4521 params
.spp_sackdelay
= sp
->sackdelay
;
4522 params
.spp_pathmaxrxt
= sp
->pathmaxrxt
;
4524 /*draft-11 doesn't say what to return in spp_flags*/
4525 params
.spp_flags
= sp
->param_flags
;
4528 if (copy_to_user(optval
, ¶ms
, len
))
4531 if (put_user(len
, optlen
))
4538 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4540 * This option will effect the way delayed acks are performed. This
4541 * option allows you to get or set the delayed ack time, in
4542 * milliseconds. It also allows changing the delayed ack frequency.
4543 * Changing the frequency to 1 disables the delayed sack algorithm. If
4544 * the assoc_id is 0, then this sets or gets the endpoints default
4545 * values. If the assoc_id field is non-zero, then the set or get
4546 * effects the specified association for the one to many model (the
4547 * assoc_id field is ignored by the one to one model). Note that if
4548 * sack_delay or sack_freq are 0 when setting this option, then the
4549 * current values will remain unchanged.
4551 * struct sctp_sack_info {
4552 * sctp_assoc_t sack_assoc_id;
4553 * uint32_t sack_delay;
4554 * uint32_t sack_freq;
4557 * sack_assoc_id - This parameter, indicates which association the user
4558 * is performing an action upon. Note that if this field's value is
4559 * zero then the endpoints default value is changed (effecting future
4560 * associations only).
4562 * sack_delay - This parameter contains the number of milliseconds that
4563 * the user is requesting the delayed ACK timer be set to. Note that
4564 * this value is defined in the standard to be between 200 and 500
4567 * sack_freq - This parameter contains the number of packets that must
4568 * be received before a sack is sent without waiting for the delay
4569 * timer to expire. The default value for this is 2, setting this
4570 * value to 1 will disable the delayed sack algorithm.
4572 static int sctp_getsockopt_delayed_ack(struct sock
*sk
, int len
,
4573 char __user
*optval
,
4576 struct sctp_sack_info params
;
4577 struct sctp_association
*asoc
= NULL
;
4578 struct sctp_sock
*sp
= sctp_sk(sk
);
4580 if (len
>= sizeof(struct sctp_sack_info
)) {
4581 len
= sizeof(struct sctp_sack_info
);
4583 if (copy_from_user(¶ms
, optval
, len
))
4585 } else if (len
== sizeof(struct sctp_assoc_value
)) {
4586 pr_warn_ratelimited(DEPRECATED
4588 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
4589 "Use struct sctp_sack_info instead\n",
4590 current
->comm
, task_pid_nr(current
));
4591 if (copy_from_user(¶ms
, optval
, len
))
4596 /* Get association, if sack_assoc_id != 0 and the socket is a one
4597 * to many style socket, and an association was not found, then
4598 * the id was invalid.
4600 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
4601 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
4605 /* Fetch association values. */
4606 if (asoc
->param_flags
& SPP_SACKDELAY_ENABLE
) {
4607 params
.sack_delay
= jiffies_to_msecs(
4609 params
.sack_freq
= asoc
->sackfreq
;
4612 params
.sack_delay
= 0;
4613 params
.sack_freq
= 1;
4616 /* Fetch socket values. */
4617 if (sp
->param_flags
& SPP_SACKDELAY_ENABLE
) {
4618 params
.sack_delay
= sp
->sackdelay
;
4619 params
.sack_freq
= sp
->sackfreq
;
4621 params
.sack_delay
= 0;
4622 params
.sack_freq
= 1;
4626 if (copy_to_user(optval
, ¶ms
, len
))
4629 if (put_user(len
, optlen
))
4635 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4637 * Applications can specify protocol parameters for the default association
4638 * initialization. The option name argument to setsockopt() and getsockopt()
4641 * Setting initialization parameters is effective only on an unconnected
4642 * socket (for UDP-style sockets only future associations are effected
4643 * by the change). With TCP-style sockets, this option is inherited by
4644 * sockets derived from a listener socket.
4646 static int sctp_getsockopt_initmsg(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4648 if (len
< sizeof(struct sctp_initmsg
))
4650 len
= sizeof(struct sctp_initmsg
);
4651 if (put_user(len
, optlen
))
4653 if (copy_to_user(optval
, &sctp_sk(sk
)->initmsg
, len
))
4659 static int sctp_getsockopt_peer_addrs(struct sock
*sk
, int len
,
4660 char __user
*optval
, int __user
*optlen
)
4662 struct sctp_association
*asoc
;
4664 struct sctp_getaddrs getaddrs
;
4665 struct sctp_transport
*from
;
4667 union sctp_addr temp
;
4668 struct sctp_sock
*sp
= sctp_sk(sk
);
4673 if (len
< sizeof(struct sctp_getaddrs
))
4676 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
4679 /* For UDP-style sockets, id specifies the association to query. */
4680 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
4684 to
= optval
+ offsetof(struct sctp_getaddrs
, addrs
);
4685 space_left
= len
- offsetof(struct sctp_getaddrs
, addrs
);
4687 list_for_each_entry(from
, &asoc
->peer
.transport_addr_list
,
4689 memcpy(&temp
, &from
->ipaddr
, sizeof(temp
));
4690 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
4691 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4692 if (space_left
< addrlen
)
4694 if (copy_to_user(to
, &temp
, addrlen
))
4698 space_left
-= addrlen
;
4701 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
))
4703 bytes_copied
= ((char __user
*)to
) - optval
;
4704 if (put_user(bytes_copied
, optlen
))
4710 static int sctp_copy_laddrs(struct sock
*sk
, __u16 port
, void *to
,
4711 size_t space_left
, int *bytes_copied
)
4713 struct sctp_sockaddr_entry
*addr
;
4714 union sctp_addr temp
;
4717 struct net
*net
= sock_net(sk
);
4720 list_for_each_entry_rcu(addr
, &net
->sctp
.local_addr_list
, list
) {
4724 if ((PF_INET
== sk
->sk_family
) &&
4725 (AF_INET6
== addr
->a
.sa
.sa_family
))
4727 if ((PF_INET6
== sk
->sk_family
) &&
4728 inet_v6_ipv6only(sk
) &&
4729 (AF_INET
== addr
->a
.sa
.sa_family
))
4731 memcpy(&temp
, &addr
->a
, sizeof(temp
));
4732 if (!temp
.v4
.sin_port
)
4733 temp
.v4
.sin_port
= htons(port
);
4735 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
4737 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4738 if (space_left
< addrlen
) {
4742 memcpy(to
, &temp
, addrlen
);
4746 space_left
-= addrlen
;
4747 *bytes_copied
+= addrlen
;
4755 static int sctp_getsockopt_local_addrs(struct sock
*sk
, int len
,
4756 char __user
*optval
, int __user
*optlen
)
4758 struct sctp_bind_addr
*bp
;
4759 struct sctp_association
*asoc
;
4761 struct sctp_getaddrs getaddrs
;
4762 struct sctp_sockaddr_entry
*addr
;
4764 union sctp_addr temp
;
4765 struct sctp_sock
*sp
= sctp_sk(sk
);
4769 int bytes_copied
= 0;
4773 if (len
< sizeof(struct sctp_getaddrs
))
4776 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
4780 * For UDP-style sockets, id specifies the association to query.
4781 * If the id field is set to the value '0' then the locally bound
4782 * addresses are returned without regard to any particular
4785 if (0 == getaddrs
.assoc_id
) {
4786 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
4788 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
4791 bp
= &asoc
->base
.bind_addr
;
4794 to
= optval
+ offsetof(struct sctp_getaddrs
, addrs
);
4795 space_left
= len
- offsetof(struct sctp_getaddrs
, addrs
);
4797 addrs
= kmalloc(space_left
, GFP_KERNEL
);
4801 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4802 * addresses from the global local address list.
4804 if (sctp_list_single_entry(&bp
->address_list
)) {
4805 addr
= list_entry(bp
->address_list
.next
,
4806 struct sctp_sockaddr_entry
, list
);
4807 if (sctp_is_any(sk
, &addr
->a
)) {
4808 cnt
= sctp_copy_laddrs(sk
, bp
->port
, addrs
,
4809 space_left
, &bytes_copied
);
4819 /* Protection on the bound address list is not needed since
4820 * in the socket option context we hold a socket lock and
4821 * thus the bound address list can't change.
4823 list_for_each_entry(addr
, &bp
->address_list
, list
) {
4824 memcpy(&temp
, &addr
->a
, sizeof(temp
));
4825 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
4826 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4827 if (space_left
< addrlen
) {
4828 err
= -ENOMEM
; /*fixme: right error?*/
4831 memcpy(buf
, &temp
, addrlen
);
4833 bytes_copied
+= addrlen
;
4835 space_left
-= addrlen
;
4839 if (copy_to_user(to
, addrs
, bytes_copied
)) {
4843 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
)) {
4847 if (put_user(bytes_copied
, optlen
))
4854 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4856 * Requests that the local SCTP stack use the enclosed peer address as
4857 * the association primary. The enclosed address must be one of the
4858 * association peer's addresses.
4860 static int sctp_getsockopt_primary_addr(struct sock
*sk
, int len
,
4861 char __user
*optval
, int __user
*optlen
)
4863 struct sctp_prim prim
;
4864 struct sctp_association
*asoc
;
4865 struct sctp_sock
*sp
= sctp_sk(sk
);
4867 if (len
< sizeof(struct sctp_prim
))
4870 len
= sizeof(struct sctp_prim
);
4872 if (copy_from_user(&prim
, optval
, len
))
4875 asoc
= sctp_id2assoc(sk
, prim
.ssp_assoc_id
);
4879 if (!asoc
->peer
.primary_path
)
4882 memcpy(&prim
.ssp_addr
, &asoc
->peer
.primary_path
->ipaddr
,
4883 asoc
->peer
.primary_path
->af_specific
->sockaddr_len
);
4885 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
,
4886 (union sctp_addr
*)&prim
.ssp_addr
);
4888 if (put_user(len
, optlen
))
4890 if (copy_to_user(optval
, &prim
, len
))
4897 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4899 * Requests that the local endpoint set the specified Adaptation Layer
4900 * Indication parameter for all future INIT and INIT-ACK exchanges.
4902 static int sctp_getsockopt_adaptation_layer(struct sock
*sk
, int len
,
4903 char __user
*optval
, int __user
*optlen
)
4905 struct sctp_setadaptation adaptation
;
4907 if (len
< sizeof(struct sctp_setadaptation
))
4910 len
= sizeof(struct sctp_setadaptation
);
4912 adaptation
.ssb_adaptation_ind
= sctp_sk(sk
)->adaptation_ind
;
4914 if (put_user(len
, optlen
))
4916 if (copy_to_user(optval
, &adaptation
, len
))
4924 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4926 * Applications that wish to use the sendto() system call may wish to
4927 * specify a default set of parameters that would normally be supplied
4928 * through the inclusion of ancillary data. This socket option allows
4929 * such an application to set the default sctp_sndrcvinfo structure.
4932 * The application that wishes to use this socket option simply passes
4933 * in to this call the sctp_sndrcvinfo structure defined in Section
4934 * 5.2.2) The input parameters accepted by this call include
4935 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4936 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4937 * to this call if the caller is using the UDP model.
4939 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4941 static int sctp_getsockopt_default_send_param(struct sock
*sk
,
4942 int len
, char __user
*optval
,
4945 struct sctp_sndrcvinfo info
;
4946 struct sctp_association
*asoc
;
4947 struct sctp_sock
*sp
= sctp_sk(sk
);
4949 if (len
< sizeof(struct sctp_sndrcvinfo
))
4952 len
= sizeof(struct sctp_sndrcvinfo
);
4954 if (copy_from_user(&info
, optval
, len
))
4957 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
4958 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
4962 info
.sinfo_stream
= asoc
->default_stream
;
4963 info
.sinfo_flags
= asoc
->default_flags
;
4964 info
.sinfo_ppid
= asoc
->default_ppid
;
4965 info
.sinfo_context
= asoc
->default_context
;
4966 info
.sinfo_timetolive
= asoc
->default_timetolive
;
4968 info
.sinfo_stream
= sp
->default_stream
;
4969 info
.sinfo_flags
= sp
->default_flags
;
4970 info
.sinfo_ppid
= sp
->default_ppid
;
4971 info
.sinfo_context
= sp
->default_context
;
4972 info
.sinfo_timetolive
= sp
->default_timetolive
;
4975 if (put_user(len
, optlen
))
4977 if (copy_to_user(optval
, &info
, len
))
4985 * 7.1.5 SCTP_NODELAY
4987 * Turn on/off any Nagle-like algorithm. This means that packets are
4988 * generally sent as soon as possible and no unnecessary delays are
4989 * introduced, at the cost of more packets in the network. Expects an
4990 * integer boolean flag.
4993 static int sctp_getsockopt_nodelay(struct sock
*sk
, int len
,
4994 char __user
*optval
, int __user
*optlen
)
4998 if (len
< sizeof(int))
5002 val
= (sctp_sk(sk
)->nodelay
== 1);
5003 if (put_user(len
, optlen
))
5005 if (copy_to_user(optval
, &val
, len
))
5012 * 7.1.1 SCTP_RTOINFO
5014 * The protocol parameters used to initialize and bound retransmission
5015 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5016 * and modify these parameters.
5017 * All parameters are time values, in milliseconds. A value of 0, when
5018 * modifying the parameters, indicates that the current value should not
5022 static int sctp_getsockopt_rtoinfo(struct sock
*sk
, int len
,
5023 char __user
*optval
,
5024 int __user
*optlen
) {
5025 struct sctp_rtoinfo rtoinfo
;
5026 struct sctp_association
*asoc
;
5028 if (len
< sizeof (struct sctp_rtoinfo
))
5031 len
= sizeof(struct sctp_rtoinfo
);
5033 if (copy_from_user(&rtoinfo
, optval
, len
))
5036 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
5038 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
5041 /* Values corresponding to the specific association. */
5043 rtoinfo
.srto_initial
= jiffies_to_msecs(asoc
->rto_initial
);
5044 rtoinfo
.srto_max
= jiffies_to_msecs(asoc
->rto_max
);
5045 rtoinfo
.srto_min
= jiffies_to_msecs(asoc
->rto_min
);
5047 /* Values corresponding to the endpoint. */
5048 struct sctp_sock
*sp
= sctp_sk(sk
);
5050 rtoinfo
.srto_initial
= sp
->rtoinfo
.srto_initial
;
5051 rtoinfo
.srto_max
= sp
->rtoinfo
.srto_max
;
5052 rtoinfo
.srto_min
= sp
->rtoinfo
.srto_min
;
5055 if (put_user(len
, optlen
))
5058 if (copy_to_user(optval
, &rtoinfo
, len
))
5066 * 7.1.2 SCTP_ASSOCINFO
5068 * This option is used to tune the maximum retransmission attempts
5069 * of the association.
5070 * Returns an error if the new association retransmission value is
5071 * greater than the sum of the retransmission value of the peer.
5072 * See [SCTP] for more information.
5075 static int sctp_getsockopt_associnfo(struct sock
*sk
, int len
,
5076 char __user
*optval
,
5080 struct sctp_assocparams assocparams
;
5081 struct sctp_association
*asoc
;
5082 struct list_head
*pos
;
5085 if (len
< sizeof (struct sctp_assocparams
))
5088 len
= sizeof(struct sctp_assocparams
);
5090 if (copy_from_user(&assocparams
, optval
, len
))
5093 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
5095 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
5098 /* Values correspoinding to the specific association */
5100 assocparams
.sasoc_asocmaxrxt
= asoc
->max_retrans
;
5101 assocparams
.sasoc_peer_rwnd
= asoc
->peer
.rwnd
;
5102 assocparams
.sasoc_local_rwnd
= asoc
->a_rwnd
;
5103 assocparams
.sasoc_cookie_life
= ktime_to_ms(asoc
->cookie_life
);
5105 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
5109 assocparams
.sasoc_number_peer_destinations
= cnt
;
5111 /* Values corresponding to the endpoint */
5112 struct sctp_sock
*sp
= sctp_sk(sk
);
5114 assocparams
.sasoc_asocmaxrxt
= sp
->assocparams
.sasoc_asocmaxrxt
;
5115 assocparams
.sasoc_peer_rwnd
= sp
->assocparams
.sasoc_peer_rwnd
;
5116 assocparams
.sasoc_local_rwnd
= sp
->assocparams
.sasoc_local_rwnd
;
5117 assocparams
.sasoc_cookie_life
=
5118 sp
->assocparams
.sasoc_cookie_life
;
5119 assocparams
.sasoc_number_peer_destinations
=
5121 sasoc_number_peer_destinations
;
5124 if (put_user(len
, optlen
))
5127 if (copy_to_user(optval
, &assocparams
, len
))
5134 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5136 * This socket option is a boolean flag which turns on or off mapped V4
5137 * addresses. If this option is turned on and the socket is type
5138 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5139 * If this option is turned off, then no mapping will be done of V4
5140 * addresses and a user will receive both PF_INET6 and PF_INET type
5141 * addresses on the socket.
5143 static int sctp_getsockopt_mappedv4(struct sock
*sk
, int len
,
5144 char __user
*optval
, int __user
*optlen
)
5147 struct sctp_sock
*sp
= sctp_sk(sk
);
5149 if (len
< sizeof(int))
5154 if (put_user(len
, optlen
))
5156 if (copy_to_user(optval
, &val
, len
))
5163 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5164 * (chapter and verse is quoted at sctp_setsockopt_context())
5166 static int sctp_getsockopt_context(struct sock
*sk
, int len
,
5167 char __user
*optval
, int __user
*optlen
)
5169 struct sctp_assoc_value params
;
5170 struct sctp_sock
*sp
;
5171 struct sctp_association
*asoc
;
5173 if (len
< sizeof(struct sctp_assoc_value
))
5176 len
= sizeof(struct sctp_assoc_value
);
5178 if (copy_from_user(¶ms
, optval
, len
))
5183 if (params
.assoc_id
!= 0) {
5184 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5187 params
.assoc_value
= asoc
->default_rcv_context
;
5189 params
.assoc_value
= sp
->default_rcv_context
;
5192 if (put_user(len
, optlen
))
5194 if (copy_to_user(optval
, ¶ms
, len
))
5201 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5202 * This option will get or set the maximum size to put in any outgoing
5203 * SCTP DATA chunk. If a message is larger than this size it will be
5204 * fragmented by SCTP into the specified size. Note that the underlying
5205 * SCTP implementation may fragment into smaller sized chunks when the
5206 * PMTU of the underlying association is smaller than the value set by
5207 * the user. The default value for this option is '0' which indicates
5208 * the user is NOT limiting fragmentation and only the PMTU will effect
5209 * SCTP's choice of DATA chunk size. Note also that values set larger
5210 * than the maximum size of an IP datagram will effectively let SCTP
5211 * control fragmentation (i.e. the same as setting this option to 0).
5213 * The following structure is used to access and modify this parameter:
5215 * struct sctp_assoc_value {
5216 * sctp_assoc_t assoc_id;
5217 * uint32_t assoc_value;
5220 * assoc_id: This parameter is ignored for one-to-one style sockets.
5221 * For one-to-many style sockets this parameter indicates which
5222 * association the user is performing an action upon. Note that if
5223 * this field's value is zero then the endpoints default value is
5224 * changed (effecting future associations only).
5225 * assoc_value: This parameter specifies the maximum size in bytes.
5227 static int sctp_getsockopt_maxseg(struct sock
*sk
, int len
,
5228 char __user
*optval
, int __user
*optlen
)
5230 struct sctp_assoc_value params
;
5231 struct sctp_association
*asoc
;
5233 if (len
== sizeof(int)) {
5234 pr_warn_ratelimited(DEPRECATED
5236 "Use of int in maxseg socket option.\n"
5237 "Use struct sctp_assoc_value instead\n",
5238 current
->comm
, task_pid_nr(current
));
5239 params
.assoc_id
= 0;
5240 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5241 len
= sizeof(struct sctp_assoc_value
);
5242 if (copy_from_user(¶ms
, optval
, sizeof(params
)))
5247 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5248 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
5252 params
.assoc_value
= asoc
->frag_point
;
5254 params
.assoc_value
= sctp_sk(sk
)->user_frag
;
5256 if (put_user(len
, optlen
))
5258 if (len
== sizeof(int)) {
5259 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5262 if (copy_to_user(optval
, ¶ms
, len
))
5270 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5271 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5273 static int sctp_getsockopt_fragment_interleave(struct sock
*sk
, int len
,
5274 char __user
*optval
, int __user
*optlen
)
5278 if (len
< sizeof(int))
5283 val
= sctp_sk(sk
)->frag_interleave
;
5284 if (put_user(len
, optlen
))
5286 if (copy_to_user(optval
, &val
, len
))
5293 * 7.1.25. Set or Get the sctp partial delivery point
5294 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5296 static int sctp_getsockopt_partial_delivery_point(struct sock
*sk
, int len
,
5297 char __user
*optval
,
5302 if (len
< sizeof(u32
))
5307 val
= sctp_sk(sk
)->pd_point
;
5308 if (put_user(len
, optlen
))
5310 if (copy_to_user(optval
, &val
, len
))
5317 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5318 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5320 static int sctp_getsockopt_maxburst(struct sock
*sk
, int len
,
5321 char __user
*optval
,
5324 struct sctp_assoc_value params
;
5325 struct sctp_sock
*sp
;
5326 struct sctp_association
*asoc
;
5328 if (len
== sizeof(int)) {
5329 pr_warn_ratelimited(DEPRECATED
5331 "Use of int in max_burst socket option.\n"
5332 "Use struct sctp_assoc_value instead\n",
5333 current
->comm
, task_pid_nr(current
));
5334 params
.assoc_id
= 0;
5335 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5336 len
= sizeof(struct sctp_assoc_value
);
5337 if (copy_from_user(¶ms
, optval
, len
))
5344 if (params
.assoc_id
!= 0) {
5345 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5348 params
.assoc_value
= asoc
->max_burst
;
5350 params
.assoc_value
= sp
->max_burst
;
5352 if (len
== sizeof(int)) {
5353 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5356 if (copy_to_user(optval
, ¶ms
, len
))
5364 static int sctp_getsockopt_hmac_ident(struct sock
*sk
, int len
,
5365 char __user
*optval
, int __user
*optlen
)
5367 struct net
*net
= sock_net(sk
);
5368 struct sctp_hmacalgo __user
*p
= (void __user
*)optval
;
5369 struct sctp_hmac_algo_param
*hmacs
;
5373 if (!net
->sctp
.auth_enable
)
5376 hmacs
= sctp_sk(sk
)->ep
->auth_hmacs_list
;
5377 data_len
= ntohs(hmacs
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5379 if (len
< sizeof(struct sctp_hmacalgo
) + data_len
)
5382 len
= sizeof(struct sctp_hmacalgo
) + data_len
;
5383 num_idents
= data_len
/ sizeof(u16
);
5385 if (put_user(len
, optlen
))
5387 if (put_user(num_idents
, &p
->shmac_num_idents
))
5389 if (copy_to_user(p
->shmac_idents
, hmacs
->hmac_ids
, data_len
))
5394 static int sctp_getsockopt_active_key(struct sock
*sk
, int len
,
5395 char __user
*optval
, int __user
*optlen
)
5397 struct net
*net
= sock_net(sk
);
5398 struct sctp_authkeyid val
;
5399 struct sctp_association
*asoc
;
5401 if (!net
->sctp
.auth_enable
)
5404 if (len
< sizeof(struct sctp_authkeyid
))
5406 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authkeyid
)))
5409 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
5410 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
5414 val
.scact_keynumber
= asoc
->active_key_id
;
5416 val
.scact_keynumber
= sctp_sk(sk
)->ep
->active_key_id
;
5418 len
= sizeof(struct sctp_authkeyid
);
5419 if (put_user(len
, optlen
))
5421 if (copy_to_user(optval
, &val
, len
))
5427 static int sctp_getsockopt_peer_auth_chunks(struct sock
*sk
, int len
,
5428 char __user
*optval
, int __user
*optlen
)
5430 struct net
*net
= sock_net(sk
);
5431 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5432 struct sctp_authchunks val
;
5433 struct sctp_association
*asoc
;
5434 struct sctp_chunks_param
*ch
;
5438 if (!net
->sctp
.auth_enable
)
5441 if (len
< sizeof(struct sctp_authchunks
))
5444 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5447 to
= p
->gauth_chunks
;
5448 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5452 ch
= asoc
->peer
.peer_chunks
;
5456 /* See if the user provided enough room for all the data */
5457 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5458 if (len
< num_chunks
)
5461 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
5464 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
5465 if (put_user(len
, optlen
))
5467 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
5472 static int sctp_getsockopt_local_auth_chunks(struct sock
*sk
, int len
,
5473 char __user
*optval
, int __user
*optlen
)
5475 struct net
*net
= sock_net(sk
);
5476 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5477 struct sctp_authchunks val
;
5478 struct sctp_association
*asoc
;
5479 struct sctp_chunks_param
*ch
;
5483 if (!net
->sctp
.auth_enable
)
5486 if (len
< sizeof(struct sctp_authchunks
))
5489 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5492 to
= p
->gauth_chunks
;
5493 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5494 if (!asoc
&& val
.gauth_assoc_id
&& sctp_style(sk
, UDP
))
5498 ch
= (struct sctp_chunks_param
*)asoc
->c
.auth_chunks
;
5500 ch
= sctp_sk(sk
)->ep
->auth_chunk_list
;
5505 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5506 if (len
< sizeof(struct sctp_authchunks
) + num_chunks
)
5509 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
5512 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
5513 if (put_user(len
, optlen
))
5515 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
5522 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5523 * This option gets the current number of associations that are attached
5524 * to a one-to-many style socket. The option value is an uint32_t.
5526 static int sctp_getsockopt_assoc_number(struct sock
*sk
, int len
,
5527 char __user
*optval
, int __user
*optlen
)
5529 struct sctp_sock
*sp
= sctp_sk(sk
);
5530 struct sctp_association
*asoc
;
5533 if (sctp_style(sk
, TCP
))
5536 if (len
< sizeof(u32
))
5541 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5545 if (put_user(len
, optlen
))
5547 if (copy_to_user(optval
, &val
, len
))
5554 * 8.1.23 SCTP_AUTO_ASCONF
5555 * See the corresponding setsockopt entry as description
5557 static int sctp_getsockopt_auto_asconf(struct sock
*sk
, int len
,
5558 char __user
*optval
, int __user
*optlen
)
5562 if (len
< sizeof(int))
5566 if (sctp_sk(sk
)->do_auto_asconf
&& sctp_is_ep_boundall(sk
))
5568 if (put_user(len
, optlen
))
5570 if (copy_to_user(optval
, &val
, len
))
5576 * 8.2.6. Get the Current Identifiers of Associations
5577 * (SCTP_GET_ASSOC_ID_LIST)
5579 * This option gets the current list of SCTP association identifiers of
5580 * the SCTP associations handled by a one-to-many style socket.
5582 static int sctp_getsockopt_assoc_ids(struct sock
*sk
, int len
,
5583 char __user
*optval
, int __user
*optlen
)
5585 struct sctp_sock
*sp
= sctp_sk(sk
);
5586 struct sctp_association
*asoc
;
5587 struct sctp_assoc_ids
*ids
;
5590 if (sctp_style(sk
, TCP
))
5593 if (len
< sizeof(struct sctp_assoc_ids
))
5596 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5600 if (len
< sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
)
5603 len
= sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
;
5605 ids
= kmalloc(len
, GFP_KERNEL
);
5609 ids
->gaids_number_of_ids
= num
;
5611 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5612 ids
->gaids_assoc_id
[num
++] = asoc
->assoc_id
;
5615 if (put_user(len
, optlen
) || copy_to_user(optval
, ids
, len
)) {
5625 * SCTP_PEER_ADDR_THLDS
5627 * This option allows us to fetch the partially failed threshold for one or all
5628 * transports in an association. See Section 6.1 of:
5629 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5631 static int sctp_getsockopt_paddr_thresholds(struct sock
*sk
,
5632 char __user
*optval
,
5636 struct sctp_paddrthlds val
;
5637 struct sctp_transport
*trans
;
5638 struct sctp_association
*asoc
;
5640 if (len
< sizeof(struct sctp_paddrthlds
))
5642 len
= sizeof(struct sctp_paddrthlds
);
5643 if (copy_from_user(&val
, (struct sctp_paddrthlds __user
*)optval
, len
))
5646 if (sctp_is_any(sk
, (const union sctp_addr
*)&val
.spt_address
)) {
5647 asoc
= sctp_id2assoc(sk
, val
.spt_assoc_id
);
5651 val
.spt_pathpfthld
= asoc
->pf_retrans
;
5652 val
.spt_pathmaxrxt
= asoc
->pathmaxrxt
;
5654 trans
= sctp_addr_id2transport(sk
, &val
.spt_address
,
5659 val
.spt_pathmaxrxt
= trans
->pathmaxrxt
;
5660 val
.spt_pathpfthld
= trans
->pf_retrans
;
5663 if (put_user(len
, optlen
) || copy_to_user(optval
, &val
, len
))
5670 * SCTP_GET_ASSOC_STATS
5672 * This option retrieves local per endpoint statistics. It is modeled
5673 * after OpenSolaris' implementation
5675 static int sctp_getsockopt_assoc_stats(struct sock
*sk
, int len
,
5676 char __user
*optval
,
5679 struct sctp_assoc_stats sas
;
5680 struct sctp_association
*asoc
= NULL
;
5682 /* User must provide at least the assoc id */
5683 if (len
< sizeof(sctp_assoc_t
))
5686 /* Allow the struct to grow and fill in as much as possible */
5687 len
= min_t(size_t, len
, sizeof(sas
));
5689 if (copy_from_user(&sas
, optval
, len
))
5692 asoc
= sctp_id2assoc(sk
, sas
.sas_assoc_id
);
5696 sas
.sas_rtxchunks
= asoc
->stats
.rtxchunks
;
5697 sas
.sas_gapcnt
= asoc
->stats
.gapcnt
;
5698 sas
.sas_outofseqtsns
= asoc
->stats
.outofseqtsns
;
5699 sas
.sas_osacks
= asoc
->stats
.osacks
;
5700 sas
.sas_isacks
= asoc
->stats
.isacks
;
5701 sas
.sas_octrlchunks
= asoc
->stats
.octrlchunks
;
5702 sas
.sas_ictrlchunks
= asoc
->stats
.ictrlchunks
;
5703 sas
.sas_oodchunks
= asoc
->stats
.oodchunks
;
5704 sas
.sas_iodchunks
= asoc
->stats
.iodchunks
;
5705 sas
.sas_ouodchunks
= asoc
->stats
.ouodchunks
;
5706 sas
.sas_iuodchunks
= asoc
->stats
.iuodchunks
;
5707 sas
.sas_idupchunks
= asoc
->stats
.idupchunks
;
5708 sas
.sas_opackets
= asoc
->stats
.opackets
;
5709 sas
.sas_ipackets
= asoc
->stats
.ipackets
;
5711 /* New high max rto observed, will return 0 if not a single
5712 * RTO update took place. obs_rto_ipaddr will be bogus
5715 sas
.sas_maxrto
= asoc
->stats
.max_obs_rto
;
5716 memcpy(&sas
.sas_obs_rto_ipaddr
, &asoc
->stats
.obs_rto_ipaddr
,
5717 sizeof(struct sockaddr_storage
));
5719 /* Mark beginning of a new observation period */
5720 asoc
->stats
.max_obs_rto
= asoc
->rto_min
;
5722 if (put_user(len
, optlen
))
5725 pr_debug("%s: len:%d, assoc_id:%d\n", __func__
, len
, sas
.sas_assoc_id
);
5727 if (copy_to_user(optval
, &sas
, len
))
5733 static int sctp_getsockopt(struct sock
*sk
, int level
, int optname
,
5734 char __user
*optval
, int __user
*optlen
)
5739 pr_debug("%s: sk:%p, optname:%d\n", __func__
, sk
, optname
);
5741 /* I can hardly begin to describe how wrong this is. This is
5742 * so broken as to be worse than useless. The API draft
5743 * REALLY is NOT helpful here... I am not convinced that the
5744 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5745 * are at all well-founded.
5747 if (level
!= SOL_SCTP
) {
5748 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
5750 retval
= af
->getsockopt(sk
, level
, optname
, optval
, optlen
);
5754 if (get_user(len
, optlen
))
5761 retval
= sctp_getsockopt_sctp_status(sk
, len
, optval
, optlen
);
5763 case SCTP_DISABLE_FRAGMENTS
:
5764 retval
= sctp_getsockopt_disable_fragments(sk
, len
, optval
,
5768 retval
= sctp_getsockopt_events(sk
, len
, optval
, optlen
);
5770 case SCTP_AUTOCLOSE
:
5771 retval
= sctp_getsockopt_autoclose(sk
, len
, optval
, optlen
);
5773 case SCTP_SOCKOPT_PEELOFF
:
5774 retval
= sctp_getsockopt_peeloff(sk
, len
, optval
, optlen
);
5776 case SCTP_PEER_ADDR_PARAMS
:
5777 retval
= sctp_getsockopt_peer_addr_params(sk
, len
, optval
,
5780 case SCTP_DELAYED_SACK
:
5781 retval
= sctp_getsockopt_delayed_ack(sk
, len
, optval
,
5785 retval
= sctp_getsockopt_initmsg(sk
, len
, optval
, optlen
);
5787 case SCTP_GET_PEER_ADDRS
:
5788 retval
= sctp_getsockopt_peer_addrs(sk
, len
, optval
,
5791 case SCTP_GET_LOCAL_ADDRS
:
5792 retval
= sctp_getsockopt_local_addrs(sk
, len
, optval
,
5795 case SCTP_SOCKOPT_CONNECTX3
:
5796 retval
= sctp_getsockopt_connectx3(sk
, len
, optval
, optlen
);
5798 case SCTP_DEFAULT_SEND_PARAM
:
5799 retval
= sctp_getsockopt_default_send_param(sk
, len
,
5802 case SCTP_PRIMARY_ADDR
:
5803 retval
= sctp_getsockopt_primary_addr(sk
, len
, optval
, optlen
);
5806 retval
= sctp_getsockopt_nodelay(sk
, len
, optval
, optlen
);
5809 retval
= sctp_getsockopt_rtoinfo(sk
, len
, optval
, optlen
);
5811 case SCTP_ASSOCINFO
:
5812 retval
= sctp_getsockopt_associnfo(sk
, len
, optval
, optlen
);
5814 case SCTP_I_WANT_MAPPED_V4_ADDR
:
5815 retval
= sctp_getsockopt_mappedv4(sk
, len
, optval
, optlen
);
5818 retval
= sctp_getsockopt_maxseg(sk
, len
, optval
, optlen
);
5820 case SCTP_GET_PEER_ADDR_INFO
:
5821 retval
= sctp_getsockopt_peer_addr_info(sk
, len
, optval
,
5824 case SCTP_ADAPTATION_LAYER
:
5825 retval
= sctp_getsockopt_adaptation_layer(sk
, len
, optval
,
5829 retval
= sctp_getsockopt_context(sk
, len
, optval
, optlen
);
5831 case SCTP_FRAGMENT_INTERLEAVE
:
5832 retval
= sctp_getsockopt_fragment_interleave(sk
, len
, optval
,
5835 case SCTP_PARTIAL_DELIVERY_POINT
:
5836 retval
= sctp_getsockopt_partial_delivery_point(sk
, len
, optval
,
5839 case SCTP_MAX_BURST
:
5840 retval
= sctp_getsockopt_maxburst(sk
, len
, optval
, optlen
);
5843 case SCTP_AUTH_CHUNK
:
5844 case SCTP_AUTH_DELETE_KEY
:
5845 retval
= -EOPNOTSUPP
;
5847 case SCTP_HMAC_IDENT
:
5848 retval
= sctp_getsockopt_hmac_ident(sk
, len
, optval
, optlen
);
5850 case SCTP_AUTH_ACTIVE_KEY
:
5851 retval
= sctp_getsockopt_active_key(sk
, len
, optval
, optlen
);
5853 case SCTP_PEER_AUTH_CHUNKS
:
5854 retval
= sctp_getsockopt_peer_auth_chunks(sk
, len
, optval
,
5857 case SCTP_LOCAL_AUTH_CHUNKS
:
5858 retval
= sctp_getsockopt_local_auth_chunks(sk
, len
, optval
,
5861 case SCTP_GET_ASSOC_NUMBER
:
5862 retval
= sctp_getsockopt_assoc_number(sk
, len
, optval
, optlen
);
5864 case SCTP_GET_ASSOC_ID_LIST
:
5865 retval
= sctp_getsockopt_assoc_ids(sk
, len
, optval
, optlen
);
5867 case SCTP_AUTO_ASCONF
:
5868 retval
= sctp_getsockopt_auto_asconf(sk
, len
, optval
, optlen
);
5870 case SCTP_PEER_ADDR_THLDS
:
5871 retval
= sctp_getsockopt_paddr_thresholds(sk
, optval
, len
, optlen
);
5873 case SCTP_GET_ASSOC_STATS
:
5874 retval
= sctp_getsockopt_assoc_stats(sk
, len
, optval
, optlen
);
5877 retval
= -ENOPROTOOPT
;
5885 static void sctp_hash(struct sock
*sk
)
5890 static void sctp_unhash(struct sock
*sk
)
5895 /* Check if port is acceptable. Possibly find first available port.
5897 * The port hash table (contained in the 'global' SCTP protocol storage
5898 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5899 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5900 * list (the list number is the port number hashed out, so as you
5901 * would expect from a hash function, all the ports in a given list have
5902 * such a number that hashes out to the same list number; you were
5903 * expecting that, right?); so each list has a set of ports, with a
5904 * link to the socket (struct sock) that uses it, the port number and
5905 * a fastreuse flag (FIXME: NPI ipg).
5907 static struct sctp_bind_bucket
*sctp_bucket_create(
5908 struct sctp_bind_hashbucket
*head
, struct net
*, unsigned short snum
);
5910 static long sctp_get_port_local(struct sock
*sk
, union sctp_addr
*addr
)
5912 struct sctp_bind_hashbucket
*head
; /* hash list */
5913 struct sctp_bind_bucket
*pp
;
5914 unsigned short snum
;
5917 snum
= ntohs(addr
->v4
.sin_port
);
5919 pr_debug("%s: begins, snum:%d\n", __func__
, snum
);
5924 /* Search for an available port. */
5925 int low
, high
, remaining
, index
;
5928 inet_get_local_port_range(sock_net(sk
), &low
, &high
);
5929 remaining
= (high
- low
) + 1;
5930 rover
= prandom_u32() % remaining
+ low
;
5934 if ((rover
< low
) || (rover
> high
))
5936 if (inet_is_reserved_local_port(rover
))
5938 index
= sctp_phashfn(sock_net(sk
), rover
);
5939 head
= &sctp_port_hashtable
[index
];
5940 spin_lock(&head
->lock
);
5941 sctp_for_each_hentry(pp
, &head
->chain
)
5942 if ((pp
->port
== rover
) &&
5943 net_eq(sock_net(sk
), pp
->net
))
5947 spin_unlock(&head
->lock
);
5948 } while (--remaining
> 0);
5950 /* Exhausted local port range during search? */
5955 /* OK, here is the one we will use. HEAD (the port
5956 * hash table list entry) is non-NULL and we hold it's
5961 /* We are given an specific port number; we verify
5962 * that it is not being used. If it is used, we will
5963 * exahust the search in the hash list corresponding
5964 * to the port number (snum) - we detect that with the
5965 * port iterator, pp being NULL.
5967 head
= &sctp_port_hashtable
[sctp_phashfn(sock_net(sk
), snum
)];
5968 spin_lock(&head
->lock
);
5969 sctp_for_each_hentry(pp
, &head
->chain
) {
5970 if ((pp
->port
== snum
) && net_eq(pp
->net
, sock_net(sk
)))
5977 if (!hlist_empty(&pp
->owner
)) {
5978 /* We had a port hash table hit - there is an
5979 * available port (pp != NULL) and it is being
5980 * used by other socket (pp->owner not empty); that other
5981 * socket is going to be sk2.
5983 int reuse
= sk
->sk_reuse
;
5986 pr_debug("%s: found a possible match\n", __func__
);
5988 if (pp
->fastreuse
&& sk
->sk_reuse
&&
5989 sk
->sk_state
!= SCTP_SS_LISTENING
)
5992 /* Run through the list of sockets bound to the port
5993 * (pp->port) [via the pointers bind_next and
5994 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5995 * we get the endpoint they describe and run through
5996 * the endpoint's list of IP (v4 or v6) addresses,
5997 * comparing each of the addresses with the address of
5998 * the socket sk. If we find a match, then that means
5999 * that this port/socket (sk) combination are already
6002 sk_for_each_bound(sk2
, &pp
->owner
) {
6003 struct sctp_endpoint
*ep2
;
6004 ep2
= sctp_sk(sk2
)->ep
;
6007 (reuse
&& sk2
->sk_reuse
&&
6008 sk2
->sk_state
!= SCTP_SS_LISTENING
))
6011 if (sctp_bind_addr_conflict(&ep2
->base
.bind_addr
, addr
,
6012 sctp_sk(sk2
), sctp_sk(sk
))) {
6018 pr_debug("%s: found a match\n", __func__
);
6021 /* If there was a hash table miss, create a new port. */
6023 if (!pp
&& !(pp
= sctp_bucket_create(head
, sock_net(sk
), snum
)))
6026 /* In either case (hit or miss), make sure fastreuse is 1 only
6027 * if sk->sk_reuse is too (that is, if the caller requested
6028 * SO_REUSEADDR on this socket -sk-).
6030 if (hlist_empty(&pp
->owner
)) {
6031 if (sk
->sk_reuse
&& sk
->sk_state
!= SCTP_SS_LISTENING
)
6035 } else if (pp
->fastreuse
&&
6036 (!sk
->sk_reuse
|| sk
->sk_state
== SCTP_SS_LISTENING
))
6039 /* We are set, so fill up all the data in the hash table
6040 * entry, tie the socket list information with the rest of the
6041 * sockets FIXME: Blurry, NPI (ipg).
6044 if (!sctp_sk(sk
)->bind_hash
) {
6045 inet_sk(sk
)->inet_num
= snum
;
6046 sk_add_bind_node(sk
, &pp
->owner
);
6047 sctp_sk(sk
)->bind_hash
= pp
;
6052 spin_unlock(&head
->lock
);
6059 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6060 * port is requested.
6062 static int sctp_get_port(struct sock
*sk
, unsigned short snum
)
6064 union sctp_addr addr
;
6065 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
6067 /* Set up a dummy address struct from the sk. */
6068 af
->from_sk(&addr
, sk
);
6069 addr
.v4
.sin_port
= htons(snum
);
6071 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6072 return !!sctp_get_port_local(sk
, &addr
);
6076 * Move a socket to LISTENING state.
6078 static int sctp_listen_start(struct sock
*sk
, int backlog
)
6080 struct sctp_sock
*sp
= sctp_sk(sk
);
6081 struct sctp_endpoint
*ep
= sp
->ep
;
6082 struct crypto_hash
*tfm
= NULL
;
6085 /* Allocate HMAC for generating cookie. */
6086 if (!sp
->hmac
&& sp
->sctp_hmac_alg
) {
6087 sprintf(alg
, "hmac(%s)", sp
->sctp_hmac_alg
);
6088 tfm
= crypto_alloc_hash(alg
, 0, CRYPTO_ALG_ASYNC
);
6090 net_info_ratelimited("failed to load transform for %s: %ld\n",
6091 sp
->sctp_hmac_alg
, PTR_ERR(tfm
));
6094 sctp_sk(sk
)->hmac
= tfm
;
6098 * If a bind() or sctp_bindx() is not called prior to a listen()
6099 * call that allows new associations to be accepted, the system
6100 * picks an ephemeral port and will choose an address set equivalent
6101 * to binding with a wildcard address.
6103 * This is not currently spelled out in the SCTP sockets
6104 * extensions draft, but follows the practice as seen in TCP
6108 sk
->sk_state
= SCTP_SS_LISTENING
;
6109 if (!ep
->base
.bind_addr
.port
) {
6110 if (sctp_autobind(sk
))
6113 if (sctp_get_port(sk
, inet_sk(sk
)->inet_num
)) {
6114 sk
->sk_state
= SCTP_SS_CLOSED
;
6119 sk
->sk_max_ack_backlog
= backlog
;
6120 sctp_hash_endpoint(ep
);
6125 * 4.1.3 / 5.1.3 listen()
6127 * By default, new associations are not accepted for UDP style sockets.
6128 * An application uses listen() to mark a socket as being able to
6129 * accept new associations.
6131 * On TCP style sockets, applications use listen() to ready the SCTP
6132 * endpoint for accepting inbound associations.
6134 * On both types of endpoints a backlog of '0' disables listening.
6136 * Move a socket to LISTENING state.
6138 int sctp_inet_listen(struct socket
*sock
, int backlog
)
6140 struct sock
*sk
= sock
->sk
;
6141 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
6144 if (unlikely(backlog
< 0))
6149 /* Peeled-off sockets are not allowed to listen(). */
6150 if (sctp_style(sk
, UDP_HIGH_BANDWIDTH
))
6153 if (sock
->state
!= SS_UNCONNECTED
)
6156 /* If backlog is zero, disable listening. */
6158 if (sctp_sstate(sk
, CLOSED
))
6162 sctp_unhash_endpoint(ep
);
6163 sk
->sk_state
= SCTP_SS_CLOSED
;
6165 sctp_sk(sk
)->bind_hash
->fastreuse
= 1;
6169 /* If we are already listening, just update the backlog */
6170 if (sctp_sstate(sk
, LISTENING
))
6171 sk
->sk_max_ack_backlog
= backlog
;
6173 err
= sctp_listen_start(sk
, backlog
);
6185 * This function is done by modeling the current datagram_poll() and the
6186 * tcp_poll(). Note that, based on these implementations, we don't
6187 * lock the socket in this function, even though it seems that,
6188 * ideally, locking or some other mechanisms can be used to ensure
6189 * the integrity of the counters (sndbuf and wmem_alloc) used
6190 * in this place. We assume that we don't need locks either until proven
6193 * Another thing to note is that we include the Async I/O support
6194 * here, again, by modeling the current TCP/UDP code. We don't have
6195 * a good way to test with it yet.
6197 unsigned int sctp_poll(struct file
*file
, struct socket
*sock
, poll_table
*wait
)
6199 struct sock
*sk
= sock
->sk
;
6200 struct sctp_sock
*sp
= sctp_sk(sk
);
6203 poll_wait(file
, sk_sleep(sk
), wait
);
6205 /* A TCP-style listening socket becomes readable when the accept queue
6208 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
6209 return (!list_empty(&sp
->ep
->asocs
)) ?
6210 (POLLIN
| POLLRDNORM
) : 0;
6214 /* Is there any exceptional events? */
6215 if (sk
->sk_err
|| !skb_queue_empty(&sk
->sk_error_queue
))
6217 (sock_flag(sk
, SOCK_SELECT_ERR_QUEUE
) ? POLLPRI
: 0);
6218 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6219 mask
|= POLLRDHUP
| POLLIN
| POLLRDNORM
;
6220 if (sk
->sk_shutdown
== SHUTDOWN_MASK
)
6223 /* Is it readable? Reconsider this code with TCP-style support. */
6224 if (!skb_queue_empty(&sk
->sk_receive_queue
))
6225 mask
|= POLLIN
| POLLRDNORM
;
6227 /* The association is either gone or not ready. */
6228 if (!sctp_style(sk
, UDP
) && sctp_sstate(sk
, CLOSED
))
6231 /* Is it writable? */
6232 if (sctp_writeable(sk
)) {
6233 mask
|= POLLOUT
| POLLWRNORM
;
6235 set_bit(SOCK_ASYNC_NOSPACE
, &sk
->sk_socket
->flags
);
6237 * Since the socket is not locked, the buffer
6238 * might be made available after the writeable check and
6239 * before the bit is set. This could cause a lost I/O
6240 * signal. tcp_poll() has a race breaker for this race
6241 * condition. Based on their implementation, we put
6242 * in the following code to cover it as well.
6244 if (sctp_writeable(sk
))
6245 mask
|= POLLOUT
| POLLWRNORM
;
6250 /********************************************************************
6251 * 2nd Level Abstractions
6252 ********************************************************************/
6254 static struct sctp_bind_bucket
*sctp_bucket_create(
6255 struct sctp_bind_hashbucket
*head
, struct net
*net
, unsigned short snum
)
6257 struct sctp_bind_bucket
*pp
;
6259 pp
= kmem_cache_alloc(sctp_bucket_cachep
, GFP_ATOMIC
);
6261 SCTP_DBG_OBJCNT_INC(bind_bucket
);
6264 INIT_HLIST_HEAD(&pp
->owner
);
6266 hlist_add_head(&pp
->node
, &head
->chain
);
6271 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6272 static void sctp_bucket_destroy(struct sctp_bind_bucket
*pp
)
6274 if (pp
&& hlist_empty(&pp
->owner
)) {
6275 __hlist_del(&pp
->node
);
6276 kmem_cache_free(sctp_bucket_cachep
, pp
);
6277 SCTP_DBG_OBJCNT_DEC(bind_bucket
);
6281 /* Release this socket's reference to a local port. */
6282 static inline void __sctp_put_port(struct sock
*sk
)
6284 struct sctp_bind_hashbucket
*head
=
6285 &sctp_port_hashtable
[sctp_phashfn(sock_net(sk
),
6286 inet_sk(sk
)->inet_num
)];
6287 struct sctp_bind_bucket
*pp
;
6289 spin_lock(&head
->lock
);
6290 pp
= sctp_sk(sk
)->bind_hash
;
6291 __sk_del_bind_node(sk
);
6292 sctp_sk(sk
)->bind_hash
= NULL
;
6293 inet_sk(sk
)->inet_num
= 0;
6294 sctp_bucket_destroy(pp
);
6295 spin_unlock(&head
->lock
);
6298 void sctp_put_port(struct sock
*sk
)
6301 __sctp_put_port(sk
);
6306 * The system picks an ephemeral port and choose an address set equivalent
6307 * to binding with a wildcard address.
6308 * One of those addresses will be the primary address for the association.
6309 * This automatically enables the multihoming capability of SCTP.
6311 static int sctp_autobind(struct sock
*sk
)
6313 union sctp_addr autoaddr
;
6317 /* Initialize a local sockaddr structure to INADDR_ANY. */
6318 af
= sctp_sk(sk
)->pf
->af
;
6320 port
= htons(inet_sk(sk
)->inet_num
);
6321 af
->inaddr_any(&autoaddr
, port
);
6323 return sctp_do_bind(sk
, &autoaddr
, af
->sockaddr_len
);
6326 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6329 * 4.2 The cmsghdr Structure *
6331 * When ancillary data is sent or received, any number of ancillary data
6332 * objects can be specified by the msg_control and msg_controllen members of
6333 * the msghdr structure, because each object is preceded by
6334 * a cmsghdr structure defining the object's length (the cmsg_len member).
6335 * Historically Berkeley-derived implementations have passed only one object
6336 * at a time, but this API allows multiple objects to be
6337 * passed in a single call to sendmsg() or recvmsg(). The following example
6338 * shows two ancillary data objects in a control buffer.
6340 * |<--------------------------- msg_controllen -------------------------->|
6343 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6345 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6348 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6350 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6353 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6354 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6356 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6358 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6365 static int sctp_msghdr_parse(const struct msghdr
*msg
, sctp_cmsgs_t
*cmsgs
)
6367 struct cmsghdr
*cmsg
;
6368 struct msghdr
*my_msg
= (struct msghdr
*)msg
;
6370 for (cmsg
= CMSG_FIRSTHDR(msg
);
6372 cmsg
= CMSG_NXTHDR(my_msg
, cmsg
)) {
6373 if (!CMSG_OK(my_msg
, cmsg
))
6376 /* Should we parse this header or ignore? */
6377 if (cmsg
->cmsg_level
!= IPPROTO_SCTP
)
6380 /* Strictly check lengths following example in SCM code. */
6381 switch (cmsg
->cmsg_type
) {
6383 /* SCTP Socket API Extension
6384 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6386 * This cmsghdr structure provides information for
6387 * initializing new SCTP associations with sendmsg().
6388 * The SCTP_INITMSG socket option uses this same data
6389 * structure. This structure is not used for
6392 * cmsg_level cmsg_type cmsg_data[]
6393 * ------------ ------------ ----------------------
6394 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6396 if (cmsg
->cmsg_len
!=
6397 CMSG_LEN(sizeof(struct sctp_initmsg
)))
6399 cmsgs
->init
= (struct sctp_initmsg
*)CMSG_DATA(cmsg
);
6403 /* SCTP Socket API Extension
6404 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6406 * This cmsghdr structure specifies SCTP options for
6407 * sendmsg() and describes SCTP header information
6408 * about a received message through recvmsg().
6410 * cmsg_level cmsg_type cmsg_data[]
6411 * ------------ ------------ ----------------------
6412 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6414 if (cmsg
->cmsg_len
!=
6415 CMSG_LEN(sizeof(struct sctp_sndrcvinfo
)))
6419 (struct sctp_sndrcvinfo
*)CMSG_DATA(cmsg
);
6421 /* Minimally, validate the sinfo_flags. */
6422 if (cmsgs
->info
->sinfo_flags
&
6423 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
6424 SCTP_ABORT
| SCTP_EOF
))
6436 * Wait for a packet..
6437 * Note: This function is the same function as in core/datagram.c
6438 * with a few modifications to make lksctp work.
6440 static int sctp_wait_for_packet(struct sock
*sk
, int *err
, long *timeo_p
)
6445 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
6447 /* Socket errors? */
6448 error
= sock_error(sk
);
6452 if (!skb_queue_empty(&sk
->sk_receive_queue
))
6455 /* Socket shut down? */
6456 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6459 /* Sequenced packets can come disconnected. If so we report the
6464 /* Is there a good reason to think that we may receive some data? */
6465 if (list_empty(&sctp_sk(sk
)->ep
->asocs
) && !sctp_sstate(sk
, LISTENING
))
6468 /* Handle signals. */
6469 if (signal_pending(current
))
6472 /* Let another process have a go. Since we are going to sleep
6473 * anyway. Note: This may cause odd behaviors if the message
6474 * does not fit in the user's buffer, but this seems to be the
6475 * only way to honor MSG_DONTWAIT realistically.
6478 *timeo_p
= schedule_timeout(*timeo_p
);
6482 finish_wait(sk_sleep(sk
), &wait
);
6486 error
= sock_intr_errno(*timeo_p
);
6489 finish_wait(sk_sleep(sk
), &wait
);
6494 /* Receive a datagram.
6495 * Note: This is pretty much the same routine as in core/datagram.c
6496 * with a few changes to make lksctp work.
6498 static struct sk_buff
*sctp_skb_recv_datagram(struct sock
*sk
, int flags
,
6499 int noblock
, int *err
)
6502 struct sk_buff
*skb
;
6505 timeo
= sock_rcvtimeo(sk
, noblock
);
6507 pr_debug("%s: timeo:%ld, max:%ld\n", __func__
, timeo
,
6508 MAX_SCHEDULE_TIMEOUT
);
6511 /* Again only user level code calls this function,
6512 * so nothing interrupt level
6513 * will suddenly eat the receive_queue.
6515 * Look at current nfs client by the way...
6516 * However, this function was correct in any case. 8)
6518 if (flags
& MSG_PEEK
) {
6519 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
6520 skb
= skb_peek(&sk
->sk_receive_queue
);
6522 atomic_inc(&skb
->users
);
6523 spin_unlock_bh(&sk
->sk_receive_queue
.lock
);
6525 skb
= skb_dequeue(&sk
->sk_receive_queue
);
6531 /* Caller is allowed not to check sk->sk_err before calling. */
6532 error
= sock_error(sk
);
6536 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6539 /* User doesn't want to wait. */
6543 } while (sctp_wait_for_packet(sk
, err
, &timeo
) == 0);
6552 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6553 static void __sctp_write_space(struct sctp_association
*asoc
)
6555 struct sock
*sk
= asoc
->base
.sk
;
6556 struct socket
*sock
= sk
->sk_socket
;
6558 if ((sctp_wspace(asoc
) > 0) && sock
) {
6559 if (waitqueue_active(&asoc
->wait
))
6560 wake_up_interruptible(&asoc
->wait
);
6562 if (sctp_writeable(sk
)) {
6563 wait_queue_head_t
*wq
= sk_sleep(sk
);
6565 if (wq
&& waitqueue_active(wq
))
6566 wake_up_interruptible(wq
);
6568 /* Note that we try to include the Async I/O support
6569 * here by modeling from the current TCP/UDP code.
6570 * We have not tested with it yet.
6572 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
))
6573 sock_wake_async(sock
,
6574 SOCK_WAKE_SPACE
, POLL_OUT
);
6579 /* Do accounting for the sndbuf space.
6580 * Decrement the used sndbuf space of the corresponding association by the
6581 * data size which was just transmitted(freed).
6583 static void sctp_wfree(struct sk_buff
*skb
)
6585 struct sctp_association
*asoc
;
6586 struct sctp_chunk
*chunk
;
6589 /* Get the saved chunk pointer. */
6590 chunk
= *((struct sctp_chunk
**)(skb
->cb
));
6593 asoc
->sndbuf_used
-= SCTP_DATA_SNDSIZE(chunk
) +
6594 sizeof(struct sk_buff
) +
6595 sizeof(struct sctp_chunk
);
6597 atomic_sub(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
6600 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6602 sk
->sk_wmem_queued
-= skb
->truesize
;
6603 sk_mem_uncharge(sk
, skb
->truesize
);
6606 __sctp_write_space(asoc
);
6608 sctp_association_put(asoc
);
6611 /* Do accounting for the receive space on the socket.
6612 * Accounting for the association is done in ulpevent.c
6613 * We set this as a destructor for the cloned data skbs so that
6614 * accounting is done at the correct time.
6616 void sctp_sock_rfree(struct sk_buff
*skb
)
6618 struct sock
*sk
= skb
->sk
;
6619 struct sctp_ulpevent
*event
= sctp_skb2event(skb
);
6621 atomic_sub(event
->rmem_len
, &sk
->sk_rmem_alloc
);
6624 * Mimic the behavior of sock_rfree
6626 sk_mem_uncharge(sk
, event
->rmem_len
);
6630 /* Helper function to wait for space in the sndbuf. */
6631 static int sctp_wait_for_sndbuf(struct sctp_association
*asoc
, long *timeo_p
,
6634 struct sock
*sk
= asoc
->base
.sk
;
6636 long current_timeo
= *timeo_p
;
6639 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__
, asoc
,
6642 /* Increment the association's refcnt. */
6643 sctp_association_hold(asoc
);
6645 /* Wait on the association specific sndbuf space. */
6647 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
6648 TASK_INTERRUPTIBLE
);
6651 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
6654 if (signal_pending(current
))
6655 goto do_interrupted
;
6656 if (msg_len
<= sctp_wspace(asoc
))
6659 /* Let another process have a go. Since we are going
6663 current_timeo
= schedule_timeout(current_timeo
);
6664 BUG_ON(sk
!= asoc
->base
.sk
);
6667 *timeo_p
= current_timeo
;
6671 finish_wait(&asoc
->wait
, &wait
);
6673 /* Release the association's refcnt. */
6674 sctp_association_put(asoc
);
6683 err
= sock_intr_errno(*timeo_p
);
6691 void sctp_data_ready(struct sock
*sk
, int len
)
6693 struct socket_wq
*wq
;
6696 wq
= rcu_dereference(sk
->sk_wq
);
6697 if (wq_has_sleeper(wq
))
6698 wake_up_interruptible_sync_poll(&wq
->wait
, POLLIN
|
6699 POLLRDNORM
| POLLRDBAND
);
6700 sk_wake_async(sk
, SOCK_WAKE_WAITD
, POLL_IN
);
6704 /* If socket sndbuf has changed, wake up all per association waiters. */
6705 void sctp_write_space(struct sock
*sk
)
6707 struct sctp_association
*asoc
;
6709 /* Wake up the tasks in each wait queue. */
6710 list_for_each_entry(asoc
, &((sctp_sk(sk
))->ep
->asocs
), asocs
) {
6711 __sctp_write_space(asoc
);
6715 /* Is there any sndbuf space available on the socket?
6717 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6718 * associations on the same socket. For a UDP-style socket with
6719 * multiple associations, it is possible for it to be "unwriteable"
6720 * prematurely. I assume that this is acceptable because
6721 * a premature "unwriteable" is better than an accidental "writeable" which
6722 * would cause an unwanted block under certain circumstances. For the 1-1
6723 * UDP-style sockets or TCP-style sockets, this code should work.
6726 static int sctp_writeable(struct sock
*sk
)
6730 amt
= sk
->sk_sndbuf
- sk_wmem_alloc_get(sk
);
6736 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6737 * returns immediately with EINPROGRESS.
6739 static int sctp_wait_for_connect(struct sctp_association
*asoc
, long *timeo_p
)
6741 struct sock
*sk
= asoc
->base
.sk
;
6743 long current_timeo
= *timeo_p
;
6746 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__
, asoc
, *timeo_p
);
6748 /* Increment the association's refcnt. */
6749 sctp_association_hold(asoc
);
6752 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
6753 TASK_INTERRUPTIBLE
);
6756 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6758 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
6761 if (signal_pending(current
))
6762 goto do_interrupted
;
6764 if (sctp_state(asoc
, ESTABLISHED
))
6767 /* Let another process have a go. Since we are going
6771 current_timeo
= schedule_timeout(current_timeo
);
6774 *timeo_p
= current_timeo
;
6778 finish_wait(&asoc
->wait
, &wait
);
6780 /* Release the association's refcnt. */
6781 sctp_association_put(asoc
);
6786 if (asoc
->init_err_counter
+ 1 > asoc
->max_init_attempts
)
6789 err
= -ECONNREFUSED
;
6793 err
= sock_intr_errno(*timeo_p
);
6801 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
)
6803 struct sctp_endpoint
*ep
;
6807 ep
= sctp_sk(sk
)->ep
;
6811 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
,
6812 TASK_INTERRUPTIBLE
);
6814 if (list_empty(&ep
->asocs
)) {
6816 timeo
= schedule_timeout(timeo
);
6821 if (!sctp_sstate(sk
, LISTENING
))
6825 if (!list_empty(&ep
->asocs
))
6828 err
= sock_intr_errno(timeo
);
6829 if (signal_pending(current
))
6837 finish_wait(sk_sleep(sk
), &wait
);
6842 static void sctp_wait_for_close(struct sock
*sk
, long timeout
)
6847 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
6848 if (list_empty(&sctp_sk(sk
)->ep
->asocs
))
6851 timeout
= schedule_timeout(timeout
);
6853 } while (!signal_pending(current
) && timeout
);
6855 finish_wait(sk_sleep(sk
), &wait
);
6858 static void sctp_skb_set_owner_r_frag(struct sk_buff
*skb
, struct sock
*sk
)
6860 struct sk_buff
*frag
;
6865 /* Don't forget the fragments. */
6866 skb_walk_frags(skb
, frag
)
6867 sctp_skb_set_owner_r_frag(frag
, sk
);
6870 sctp_skb_set_owner_r(skb
, sk
);
6873 void sctp_copy_sock(struct sock
*newsk
, struct sock
*sk
,
6874 struct sctp_association
*asoc
)
6876 struct inet_sock
*inet
= inet_sk(sk
);
6877 struct inet_sock
*newinet
;
6879 newsk
->sk_type
= sk
->sk_type
;
6880 newsk
->sk_bound_dev_if
= sk
->sk_bound_dev_if
;
6881 newsk
->sk_flags
= sk
->sk_flags
;
6882 newsk
->sk_no_check
= sk
->sk_no_check
;
6883 newsk
->sk_reuse
= sk
->sk_reuse
;
6885 newsk
->sk_shutdown
= sk
->sk_shutdown
;
6886 newsk
->sk_destruct
= sctp_destruct_sock
;
6887 newsk
->sk_family
= sk
->sk_family
;
6888 newsk
->sk_protocol
= IPPROTO_SCTP
;
6889 newsk
->sk_backlog_rcv
= sk
->sk_prot
->backlog_rcv
;
6890 newsk
->sk_sndbuf
= sk
->sk_sndbuf
;
6891 newsk
->sk_rcvbuf
= sk
->sk_rcvbuf
;
6892 newsk
->sk_lingertime
= sk
->sk_lingertime
;
6893 newsk
->sk_rcvtimeo
= sk
->sk_rcvtimeo
;
6894 newsk
->sk_sndtimeo
= sk
->sk_sndtimeo
;
6896 newinet
= inet_sk(newsk
);
6898 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6899 * getsockname() and getpeername()
6901 newinet
->inet_sport
= inet
->inet_sport
;
6902 newinet
->inet_saddr
= inet
->inet_saddr
;
6903 newinet
->inet_rcv_saddr
= inet
->inet_rcv_saddr
;
6904 newinet
->inet_dport
= htons(asoc
->peer
.port
);
6905 newinet
->pmtudisc
= inet
->pmtudisc
;
6906 newinet
->inet_id
= asoc
->next_tsn
^ jiffies
;
6908 newinet
->uc_ttl
= inet
->uc_ttl
;
6909 newinet
->mc_loop
= 1;
6910 newinet
->mc_ttl
= 1;
6911 newinet
->mc_index
= 0;
6912 newinet
->mc_list
= NULL
;
6915 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6916 * and its messages to the newsk.
6918 static void sctp_sock_migrate(struct sock
*oldsk
, struct sock
*newsk
,
6919 struct sctp_association
*assoc
,
6920 sctp_socket_type_t type
)
6922 struct sctp_sock
*oldsp
= sctp_sk(oldsk
);
6923 struct sctp_sock
*newsp
= sctp_sk(newsk
);
6924 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
6925 struct sctp_endpoint
*newep
= newsp
->ep
;
6926 struct sk_buff
*skb
, *tmp
;
6927 struct sctp_ulpevent
*event
;
6928 struct sctp_bind_hashbucket
*head
;
6929 struct list_head tmplist
;
6931 /* Migrate socket buffer sizes and all the socket level options to the
6934 newsk
->sk_sndbuf
= oldsk
->sk_sndbuf
;
6935 newsk
->sk_rcvbuf
= oldsk
->sk_rcvbuf
;
6936 /* Brute force copy old sctp opt. */
6937 if (oldsp
->do_auto_asconf
) {
6938 memcpy(&tmplist
, &newsp
->auto_asconf_list
, sizeof(tmplist
));
6939 inet_sk_copy_descendant(newsk
, oldsk
);
6940 memcpy(&newsp
->auto_asconf_list
, &tmplist
, sizeof(tmplist
));
6942 inet_sk_copy_descendant(newsk
, oldsk
);
6944 /* Restore the ep value that was overwritten with the above structure
6950 /* Hook this new socket in to the bind_hash list. */
6951 head
= &sctp_port_hashtable
[sctp_phashfn(sock_net(oldsk
),
6952 inet_sk(oldsk
)->inet_num
)];
6954 spin_lock(&head
->lock
);
6955 pp
= sctp_sk(oldsk
)->bind_hash
;
6956 sk_add_bind_node(newsk
, &pp
->owner
);
6957 sctp_sk(newsk
)->bind_hash
= pp
;
6958 inet_sk(newsk
)->inet_num
= inet_sk(oldsk
)->inet_num
;
6959 spin_unlock(&head
->lock
);
6962 /* Copy the bind_addr list from the original endpoint to the new
6963 * endpoint so that we can handle restarts properly
6965 sctp_bind_addr_dup(&newsp
->ep
->base
.bind_addr
,
6966 &oldsp
->ep
->base
.bind_addr
, GFP_KERNEL
);
6968 /* Move any messages in the old socket's receive queue that are for the
6969 * peeled off association to the new socket's receive queue.
6971 sctp_skb_for_each(skb
, &oldsk
->sk_receive_queue
, tmp
) {
6972 event
= sctp_skb2event(skb
);
6973 if (event
->asoc
== assoc
) {
6974 __skb_unlink(skb
, &oldsk
->sk_receive_queue
);
6975 __skb_queue_tail(&newsk
->sk_receive_queue
, skb
);
6976 sctp_skb_set_owner_r_frag(skb
, newsk
);
6980 /* Clean up any messages pending delivery due to partial
6981 * delivery. Three cases:
6982 * 1) No partial deliver; no work.
6983 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6984 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6986 skb_queue_head_init(&newsp
->pd_lobby
);
6987 atomic_set(&sctp_sk(newsk
)->pd_mode
, assoc
->ulpq
.pd_mode
);
6989 if (atomic_read(&sctp_sk(oldsk
)->pd_mode
)) {
6990 struct sk_buff_head
*queue
;
6992 /* Decide which queue to move pd_lobby skbs to. */
6993 if (assoc
->ulpq
.pd_mode
) {
6994 queue
= &newsp
->pd_lobby
;
6996 queue
= &newsk
->sk_receive_queue
;
6998 /* Walk through the pd_lobby, looking for skbs that
6999 * need moved to the new socket.
7001 sctp_skb_for_each(skb
, &oldsp
->pd_lobby
, tmp
) {
7002 event
= sctp_skb2event(skb
);
7003 if (event
->asoc
== assoc
) {
7004 __skb_unlink(skb
, &oldsp
->pd_lobby
);
7005 __skb_queue_tail(queue
, skb
);
7006 sctp_skb_set_owner_r_frag(skb
, newsk
);
7010 /* Clear up any skbs waiting for the partial
7011 * delivery to finish.
7013 if (assoc
->ulpq
.pd_mode
)
7014 sctp_clear_pd(oldsk
, NULL
);
7018 sctp_skb_for_each(skb
, &assoc
->ulpq
.reasm
, tmp
)
7019 sctp_skb_set_owner_r_frag(skb
, newsk
);
7021 sctp_skb_for_each(skb
, &assoc
->ulpq
.lobby
, tmp
)
7022 sctp_skb_set_owner_r_frag(skb
, newsk
);
7024 /* Set the type of socket to indicate that it is peeled off from the
7025 * original UDP-style socket or created with the accept() call on a
7026 * TCP-style socket..
7030 /* Mark the new socket "in-use" by the user so that any packets
7031 * that may arrive on the association after we've moved it are
7032 * queued to the backlog. This prevents a potential race between
7033 * backlog processing on the old socket and new-packet processing
7034 * on the new socket.
7036 * The caller has just allocated newsk so we can guarantee that other
7037 * paths won't try to lock it and then oldsk.
7039 lock_sock_nested(newsk
, SINGLE_DEPTH_NESTING
);
7040 sctp_assoc_migrate(assoc
, newsk
);
7042 /* If the association on the newsk is already closed before accept()
7043 * is called, set RCV_SHUTDOWN flag.
7045 if (sctp_state(assoc
, CLOSED
) && sctp_style(newsk
, TCP
))
7046 newsk
->sk_shutdown
|= RCV_SHUTDOWN
;
7048 newsk
->sk_state
= SCTP_SS_ESTABLISHED
;
7049 release_sock(newsk
);
7053 /* This proto struct describes the ULP interface for SCTP. */
7054 struct proto sctp_prot
= {
7056 .owner
= THIS_MODULE
,
7057 .close
= sctp_close
,
7058 .connect
= sctp_connect
,
7059 .disconnect
= sctp_disconnect
,
7060 .accept
= sctp_accept
,
7061 .ioctl
= sctp_ioctl
,
7062 .init
= sctp_init_sock
,
7063 .destroy
= sctp_destroy_sock
,
7064 .shutdown
= sctp_shutdown
,
7065 .setsockopt
= sctp_setsockopt
,
7066 .getsockopt
= sctp_getsockopt
,
7067 .sendmsg
= sctp_sendmsg
,
7068 .recvmsg
= sctp_recvmsg
,
7070 .backlog_rcv
= sctp_backlog_rcv
,
7072 .unhash
= sctp_unhash
,
7073 .get_port
= sctp_get_port
,
7074 .obj_size
= sizeof(struct sctp_sock
),
7075 .sysctl_mem
= sysctl_sctp_mem
,
7076 .sysctl_rmem
= sysctl_sctp_rmem
,
7077 .sysctl_wmem
= sysctl_sctp_wmem
,
7078 .memory_pressure
= &sctp_memory_pressure
,
7079 .enter_memory_pressure
= sctp_enter_memory_pressure
,
7080 .memory_allocated
= &sctp_memory_allocated
,
7081 .sockets_allocated
= &sctp_sockets_allocated
,
7084 #if IS_ENABLED(CONFIG_IPV6)
7086 struct proto sctpv6_prot
= {
7088 .owner
= THIS_MODULE
,
7089 .close
= sctp_close
,
7090 .connect
= sctp_connect
,
7091 .disconnect
= sctp_disconnect
,
7092 .accept
= sctp_accept
,
7093 .ioctl
= sctp_ioctl
,
7094 .init
= sctp_init_sock
,
7095 .destroy
= sctp_destroy_sock
,
7096 .shutdown
= sctp_shutdown
,
7097 .setsockopt
= sctp_setsockopt
,
7098 .getsockopt
= sctp_getsockopt
,
7099 .sendmsg
= sctp_sendmsg
,
7100 .recvmsg
= sctp_recvmsg
,
7102 .backlog_rcv
= sctp_backlog_rcv
,
7104 .unhash
= sctp_unhash
,
7105 .get_port
= sctp_get_port
,
7106 .obj_size
= sizeof(struct sctp6_sock
),
7107 .sysctl_mem
= sysctl_sctp_mem
,
7108 .sysctl_rmem
= sysctl_sctp_rmem
,
7109 .sysctl_wmem
= sysctl_sctp_wmem
,
7110 .memory_pressure
= &sctp_memory_pressure
,
7111 .enter_memory_pressure
= sctp_enter_memory_pressure
,
7112 .memory_allocated
= &sctp_memory_allocated
,
7113 .sockets_allocated
= &sctp_sockets_allocated
,
7115 #endif /* IS_ENABLED(CONFIG_IPV6) */