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, write to
32 * the Free Software Foundation, 59 Temple Place - Suite 330,
33 * Boston, MA 02111-1307, USA.
35 * Please send any bug reports or fixes you make to the
37 * lksctp developers <lksctp-developers@lists.sourceforge.net>
39 * Or submit a bug report through the following website:
40 * http://www.sf.net/projects/lksctp
42 * Written or modified by:
43 * La Monte H.P. Yarroll <piggy@acm.org>
44 * Narasimha Budihal <narsi@refcode.org>
45 * Karl Knutson <karl@athena.chicago.il.us>
46 * Jon Grimm <jgrimm@us.ibm.com>
47 * Xingang Guo <xingang.guo@intel.com>
48 * Daisy Chang <daisyc@us.ibm.com>
49 * Sridhar Samudrala <samudrala@us.ibm.com>
50 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
51 * Ardelle Fan <ardelle.fan@intel.com>
52 * Ryan Layer <rmlayer@us.ibm.com>
53 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
54 * Kevin Gao <kevin.gao@intel.com>
56 * Any bugs reported given to us we will try to fix... any fixes shared will
57 * be incorporated into the next SCTP release.
60 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
62 #include <linux/types.h>
63 #include <linux/kernel.h>
64 #include <linux/wait.h>
65 #include <linux/time.h>
67 #include <linux/capability.h>
68 #include <linux/fcntl.h>
69 #include <linux/poll.h>
70 #include <linux/init.h>
71 #include <linux/crypto.h>
72 #include <linux/slab.h>
73 #include <linux/file.h>
77 #include <net/route.h>
79 #include <net/inet_common.h>
81 #include <linux/socket.h> /* for sa_family_t */
82 #include <linux/export.h>
84 #include <net/sctp/sctp.h>
85 #include <net/sctp/sm.h>
87 /* WARNING: Please do not remove the SCTP_STATIC attribute to
88 * any of the functions below as they are used to export functions
89 * used by a project regression testsuite.
92 /* Forward declarations for internal helper functions. */
93 static int sctp_writeable(struct sock
*sk
);
94 static void sctp_wfree(struct sk_buff
*skb
);
95 static int sctp_wait_for_sndbuf(struct sctp_association
*, long *timeo_p
,
97 static int sctp_wait_for_packet(struct sock
* sk
, int *err
, long *timeo_p
);
98 static int sctp_wait_for_connect(struct sctp_association
*, long *timeo_p
);
99 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
);
100 static void sctp_wait_for_close(struct sock
*sk
, long timeo
);
101 static struct sctp_af
*sctp_sockaddr_af(struct sctp_sock
*opt
,
102 union sctp_addr
*addr
, int len
);
103 static int sctp_bindx_add(struct sock
*, struct sockaddr
*, int);
104 static int sctp_bindx_rem(struct sock
*, struct sockaddr
*, int);
105 static int sctp_send_asconf_add_ip(struct sock
*, struct sockaddr
*, int);
106 static int sctp_send_asconf_del_ip(struct sock
*, struct sockaddr
*, int);
107 static int sctp_send_asconf(struct sctp_association
*asoc
,
108 struct sctp_chunk
*chunk
);
109 static int sctp_do_bind(struct sock
*, union sctp_addr
*, int);
110 static int sctp_autobind(struct sock
*sk
);
111 static void sctp_sock_migrate(struct sock
*, struct sock
*,
112 struct sctp_association
*, sctp_socket_type_t
);
114 extern struct kmem_cache
*sctp_bucket_cachep
;
115 extern long sysctl_sctp_mem
[3];
116 extern int sysctl_sctp_rmem
[3];
117 extern int sysctl_sctp_wmem
[3];
119 static int sctp_memory_pressure
;
120 static atomic_long_t sctp_memory_allocated
;
121 struct percpu_counter sctp_sockets_allocated
;
123 static void sctp_enter_memory_pressure(struct sock
*sk
)
125 sctp_memory_pressure
= 1;
129 /* Get the sndbuf space available at the time on the association. */
130 static inline int sctp_wspace(struct sctp_association
*asoc
)
134 if (asoc
->ep
->sndbuf_policy
)
135 amt
= asoc
->sndbuf_used
;
137 amt
= sk_wmem_alloc_get(asoc
->base
.sk
);
139 if (amt
>= asoc
->base
.sk
->sk_sndbuf
) {
140 if (asoc
->base
.sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)
143 amt
= sk_stream_wspace(asoc
->base
.sk
);
148 amt
= asoc
->base
.sk
->sk_sndbuf
- amt
;
153 /* Increment the used sndbuf space count of the corresponding association by
154 * the size of the outgoing data chunk.
155 * Also, set the skb destructor for sndbuf accounting later.
157 * Since it is always 1-1 between chunk and skb, and also a new skb is always
158 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
159 * destructor in the data chunk skb for the purpose of the sndbuf space
162 static inline void sctp_set_owner_w(struct sctp_chunk
*chunk
)
164 struct sctp_association
*asoc
= chunk
->asoc
;
165 struct sock
*sk
= asoc
->base
.sk
;
167 /* The sndbuf space is tracked per association. */
168 sctp_association_hold(asoc
);
170 skb_set_owner_w(chunk
->skb
, sk
);
172 chunk
->skb
->destructor
= sctp_wfree
;
173 /* Save the chunk pointer in skb for sctp_wfree to use later. */
174 *((struct sctp_chunk
**)(chunk
->skb
->cb
)) = chunk
;
176 asoc
->sndbuf_used
+= SCTP_DATA_SNDSIZE(chunk
) +
177 sizeof(struct sk_buff
) +
178 sizeof(struct sctp_chunk
);
180 atomic_add(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
181 sk
->sk_wmem_queued
+= chunk
->skb
->truesize
;
182 sk_mem_charge(sk
, chunk
->skb
->truesize
);
185 /* Verify that this is a valid address. */
186 static inline int sctp_verify_addr(struct sock
*sk
, union sctp_addr
*addr
,
191 /* Verify basic sockaddr. */
192 af
= sctp_sockaddr_af(sctp_sk(sk
), addr
, len
);
196 /* Is this a valid SCTP address? */
197 if (!af
->addr_valid(addr
, sctp_sk(sk
), NULL
))
200 if (!sctp_sk(sk
)->pf
->send_verify(sctp_sk(sk
), (addr
)))
206 /* Look up the association by its id. If this is not a UDP-style
207 * socket, the ID field is always ignored.
209 struct sctp_association
*sctp_id2assoc(struct sock
*sk
, sctp_assoc_t id
)
211 struct sctp_association
*asoc
= NULL
;
213 /* If this is not a UDP-style socket, assoc id should be ignored. */
214 if (!sctp_style(sk
, UDP
)) {
215 /* Return NULL if the socket state is not ESTABLISHED. It
216 * could be a TCP-style listening socket or a socket which
217 * hasn't yet called connect() to establish an association.
219 if (!sctp_sstate(sk
, ESTABLISHED
))
222 /* Get the first and the only association from the list. */
223 if (!list_empty(&sctp_sk(sk
)->ep
->asocs
))
224 asoc
= list_entry(sctp_sk(sk
)->ep
->asocs
.next
,
225 struct sctp_association
, asocs
);
229 /* Otherwise this is a UDP-style socket. */
230 if (!id
|| (id
== (sctp_assoc_t
)-1))
233 spin_lock_bh(&sctp_assocs_id_lock
);
234 asoc
= (struct sctp_association
*)idr_find(&sctp_assocs_id
, (int)id
);
235 spin_unlock_bh(&sctp_assocs_id_lock
);
237 if (!asoc
|| (asoc
->base
.sk
!= sk
) || asoc
->base
.dead
)
243 /* Look up the transport from an address and an assoc id. If both address and
244 * id are specified, the associations matching the address and the id should be
247 static struct sctp_transport
*sctp_addr_id2transport(struct sock
*sk
,
248 struct sockaddr_storage
*addr
,
251 struct sctp_association
*addr_asoc
= NULL
, *id_asoc
= NULL
;
252 struct sctp_transport
*transport
;
253 union sctp_addr
*laddr
= (union sctp_addr
*)addr
;
255 addr_asoc
= sctp_endpoint_lookup_assoc(sctp_sk(sk
)->ep
,
262 id_asoc
= sctp_id2assoc(sk
, id
);
263 if (id_asoc
&& (id_asoc
!= addr_asoc
))
266 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
267 (union sctp_addr
*)addr
);
272 /* API 3.1.2 bind() - UDP Style Syntax
273 * The syntax of bind() is,
275 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
277 * sd - the socket descriptor returned by socket().
278 * addr - the address structure (struct sockaddr_in or struct
279 * sockaddr_in6 [RFC 2553]),
280 * addr_len - the size of the address structure.
282 SCTP_STATIC
int sctp_bind(struct sock
*sk
, struct sockaddr
*addr
, int addr_len
)
288 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
291 /* Disallow binding twice. */
292 if (!sctp_sk(sk
)->ep
->base
.bind_addr
.port
)
293 retval
= sctp_do_bind(sk
, (union sctp_addr
*)addr
,
298 sctp_release_sock(sk
);
303 static long sctp_get_port_local(struct sock
*, union sctp_addr
*);
305 /* Verify this is a valid sockaddr. */
306 static struct sctp_af
*sctp_sockaddr_af(struct sctp_sock
*opt
,
307 union sctp_addr
*addr
, int len
)
311 /* Check minimum size. */
312 if (len
< sizeof (struct sockaddr
))
315 /* V4 mapped address are really of AF_INET family */
316 if (addr
->sa
.sa_family
== AF_INET6
&&
317 ipv6_addr_v4mapped(&addr
->v6
.sin6_addr
)) {
318 if (!opt
->pf
->af_supported(AF_INET
, opt
))
321 /* Does this PF support this AF? */
322 if (!opt
->pf
->af_supported(addr
->sa
.sa_family
, opt
))
326 /* If we get this far, af is valid. */
327 af
= sctp_get_af_specific(addr
->sa
.sa_family
);
329 if (len
< af
->sockaddr_len
)
335 /* Bind a local address either to an endpoint or to an association. */
336 SCTP_STATIC
int sctp_do_bind(struct sock
*sk
, union sctp_addr
*addr
, int len
)
338 struct net
*net
= sock_net(sk
);
339 struct sctp_sock
*sp
= sctp_sk(sk
);
340 struct sctp_endpoint
*ep
= sp
->ep
;
341 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
346 /* Common sockaddr verification. */
347 af
= sctp_sockaddr_af(sp
, addr
, len
);
349 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
354 snum
= ntohs(addr
->v4
.sin_port
);
356 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
357 ", port: %d, new port: %d, len: %d)\n",
363 /* PF specific bind() address verification. */
364 if (!sp
->pf
->bind_verify(sp
, addr
))
365 return -EADDRNOTAVAIL
;
367 /* We must either be unbound, or bind to the same port.
368 * It's OK to allow 0 ports if we are already bound.
369 * We'll just inhert an already bound port in this case
374 else if (snum
!= bp
->port
) {
375 SCTP_DEBUG_PRINTK("sctp_do_bind:"
376 " New port %d does not match existing port "
377 "%d.\n", snum
, bp
->port
);
382 if (snum
&& snum
< PROT_SOCK
&&
383 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
))
386 /* See if the address matches any of the addresses we may have
387 * already bound before checking against other endpoints.
389 if (sctp_bind_addr_match(bp
, addr
, sp
))
392 /* Make sure we are allowed to bind here.
393 * The function sctp_get_port_local() does duplicate address
396 addr
->v4
.sin_port
= htons(snum
);
397 if ((ret
= sctp_get_port_local(sk
, addr
))) {
401 /* Refresh ephemeral port. */
403 bp
->port
= inet_sk(sk
)->inet_num
;
405 /* Add the address to the bind address list.
406 * Use GFP_ATOMIC since BHs will be disabled.
408 ret
= sctp_add_bind_addr(bp
, addr
, SCTP_ADDR_SRC
, GFP_ATOMIC
);
410 /* Copy back into socket for getsockname() use. */
412 inet_sk(sk
)->inet_sport
= htons(inet_sk(sk
)->inet_num
);
413 af
->to_sk_saddr(addr
, sk
);
419 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
421 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
422 * at any one time. If a sender, after sending an ASCONF chunk, decides
423 * it needs to transfer another ASCONF Chunk, it MUST wait until the
424 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
425 * subsequent ASCONF. Note this restriction binds each side, so at any
426 * time two ASCONF may be in-transit on any given association (one sent
427 * from each endpoint).
429 static int sctp_send_asconf(struct sctp_association
*asoc
,
430 struct sctp_chunk
*chunk
)
432 struct net
*net
= sock_net(asoc
->base
.sk
);
435 /* If there is an outstanding ASCONF chunk, queue it for later
438 if (asoc
->addip_last_asconf
) {
439 list_add_tail(&chunk
->list
, &asoc
->addip_chunk_list
);
443 /* Hold the chunk until an ASCONF_ACK is received. */
444 sctp_chunk_hold(chunk
);
445 retval
= sctp_primitive_ASCONF(net
, asoc
, chunk
);
447 sctp_chunk_free(chunk
);
449 asoc
->addip_last_asconf
= chunk
;
455 /* Add a list of addresses as bind addresses to local endpoint or
458 * Basically run through each address specified in the addrs/addrcnt
459 * array/length pair, determine if it is IPv6 or IPv4 and call
460 * sctp_do_bind() on it.
462 * If any of them fails, then the operation will be reversed and the
463 * ones that were added will be removed.
465 * Only sctp_setsockopt_bindx() is supposed to call this function.
467 static int sctp_bindx_add(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
472 struct sockaddr
*sa_addr
;
475 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
479 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
480 /* The list may contain either IPv4 or IPv6 address;
481 * determine the address length for walking thru the list.
484 af
= sctp_get_af_specific(sa_addr
->sa_family
);
490 retval
= sctp_do_bind(sk
, (union sctp_addr
*)sa_addr
,
493 addr_buf
+= af
->sockaddr_len
;
497 /* Failed. Cleanup the ones that have been added */
499 sctp_bindx_rem(sk
, addrs
, cnt
);
507 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
508 * associations that are part of the endpoint indicating that a list of local
509 * addresses are added to the endpoint.
511 * If any of the addresses is already in the bind address list of the
512 * association, we do not send the chunk for that association. But it will not
513 * affect other associations.
515 * Only sctp_setsockopt_bindx() is supposed to call this function.
517 static int sctp_send_asconf_add_ip(struct sock
*sk
,
518 struct sockaddr
*addrs
,
521 struct net
*net
= sock_net(sk
);
522 struct sctp_sock
*sp
;
523 struct sctp_endpoint
*ep
;
524 struct sctp_association
*asoc
;
525 struct sctp_bind_addr
*bp
;
526 struct sctp_chunk
*chunk
;
527 struct sctp_sockaddr_entry
*laddr
;
528 union sctp_addr
*addr
;
529 union sctp_addr saveaddr
;
536 if (!net
->sctp
.addip_enable
)
542 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
543 __func__
, sk
, addrs
, addrcnt
);
545 list_for_each_entry(asoc
, &ep
->asocs
, asocs
) {
547 if (!asoc
->peer
.asconf_capable
)
550 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_ADD_IP
)
553 if (!sctp_state(asoc
, ESTABLISHED
))
556 /* Check if any address in the packed array of addresses is
557 * in the bind address list of the association. If so,
558 * do not send the asconf chunk to its peer, but continue with
559 * other associations.
562 for (i
= 0; i
< addrcnt
; i
++) {
564 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
570 if (sctp_assoc_lookup_laddr(asoc
, addr
))
573 addr_buf
+= af
->sockaddr_len
;
578 /* Use the first valid address in bind addr list of
579 * association as Address Parameter of ASCONF CHUNK.
581 bp
= &asoc
->base
.bind_addr
;
582 p
= bp
->address_list
.next
;
583 laddr
= list_entry(p
, struct sctp_sockaddr_entry
, list
);
584 chunk
= sctp_make_asconf_update_ip(asoc
, &laddr
->a
, addrs
,
585 addrcnt
, SCTP_PARAM_ADD_IP
);
591 /* Add the new addresses to the bind address list with
592 * use_as_src set to 0.
595 for (i
= 0; i
< addrcnt
; i
++) {
597 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
598 memcpy(&saveaddr
, addr
, af
->sockaddr_len
);
599 retval
= sctp_add_bind_addr(bp
, &saveaddr
,
600 SCTP_ADDR_NEW
, GFP_ATOMIC
);
601 addr_buf
+= af
->sockaddr_len
;
603 if (asoc
->src_out_of_asoc_ok
) {
604 struct sctp_transport
*trans
;
606 list_for_each_entry(trans
,
607 &asoc
->peer
.transport_addr_list
, transports
) {
608 /* Clear the source and route cache */
609 dst_release(trans
->dst
);
610 trans
->cwnd
= min(4*asoc
->pathmtu
, max_t(__u32
,
611 2*asoc
->pathmtu
, 4380));
612 trans
->ssthresh
= asoc
->peer
.i
.a_rwnd
;
613 trans
->rto
= asoc
->rto_initial
;
614 sctp_max_rto(asoc
, trans
);
615 trans
->rtt
= trans
->srtt
= trans
->rttvar
= 0;
616 sctp_transport_route(trans
, NULL
,
617 sctp_sk(asoc
->base
.sk
));
620 retval
= sctp_send_asconf(asoc
, chunk
);
627 /* Remove a list of addresses from bind addresses list. Do not remove the
630 * Basically run through each address specified in the addrs/addrcnt
631 * array/length pair, determine if it is IPv6 or IPv4 and call
632 * sctp_del_bind() on it.
634 * If any of them fails, then the operation will be reversed and the
635 * ones that were removed will be added back.
637 * At least one address has to be left; if only one address is
638 * available, the operation will return -EBUSY.
640 * Only sctp_setsockopt_bindx() is supposed to call this function.
642 static int sctp_bindx_rem(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
644 struct sctp_sock
*sp
= sctp_sk(sk
);
645 struct sctp_endpoint
*ep
= sp
->ep
;
647 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
650 union sctp_addr
*sa_addr
;
653 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
657 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
658 /* If the bind address list is empty or if there is only one
659 * bind address, there is nothing more to be removed (we need
660 * at least one address here).
662 if (list_empty(&bp
->address_list
) ||
663 (sctp_list_single_entry(&bp
->address_list
))) {
669 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
675 if (!af
->addr_valid(sa_addr
, sp
, NULL
)) {
676 retval
= -EADDRNOTAVAIL
;
680 if (sa_addr
->v4
.sin_port
&&
681 sa_addr
->v4
.sin_port
!= htons(bp
->port
)) {
686 if (!sa_addr
->v4
.sin_port
)
687 sa_addr
->v4
.sin_port
= htons(bp
->port
);
689 /* FIXME - There is probably a need to check if sk->sk_saddr and
690 * sk->sk_rcv_addr are currently set to one of the addresses to
691 * be removed. This is something which needs to be looked into
692 * when we are fixing the outstanding issues with multi-homing
693 * socket routing and failover schemes. Refer to comments in
694 * sctp_do_bind(). -daisy
696 retval
= sctp_del_bind_addr(bp
, sa_addr
);
698 addr_buf
+= af
->sockaddr_len
;
701 /* Failed. Add the ones that has been removed back */
703 sctp_bindx_add(sk
, addrs
, cnt
);
711 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
712 * the associations that are part of the endpoint indicating that a list of
713 * local addresses are removed from the endpoint.
715 * If any of the addresses is already in the bind address list of the
716 * association, we do not send the chunk for that association. But it will not
717 * affect other associations.
719 * Only sctp_setsockopt_bindx() is supposed to call this function.
721 static int sctp_send_asconf_del_ip(struct sock
*sk
,
722 struct sockaddr
*addrs
,
725 struct net
*net
= sock_net(sk
);
726 struct sctp_sock
*sp
;
727 struct sctp_endpoint
*ep
;
728 struct sctp_association
*asoc
;
729 struct sctp_transport
*transport
;
730 struct sctp_bind_addr
*bp
;
731 struct sctp_chunk
*chunk
;
732 union sctp_addr
*laddr
;
735 struct sctp_sockaddr_entry
*saddr
;
741 if (!net
->sctp
.addip_enable
)
747 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
748 __func__
, sk
, addrs
, addrcnt
);
750 list_for_each_entry(asoc
, &ep
->asocs
, asocs
) {
752 if (!asoc
->peer
.asconf_capable
)
755 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_DEL_IP
)
758 if (!sctp_state(asoc
, ESTABLISHED
))
761 /* Check if any address in the packed array of addresses is
762 * not present in the bind address list of the association.
763 * If so, do not send the asconf chunk to its peer, but
764 * continue with other associations.
767 for (i
= 0; i
< addrcnt
; i
++) {
769 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
775 if (!sctp_assoc_lookup_laddr(asoc
, laddr
))
778 addr_buf
+= af
->sockaddr_len
;
783 /* Find one address in the association's bind address list
784 * that is not in the packed array of addresses. This is to
785 * make sure that we do not delete all the addresses in the
788 bp
= &asoc
->base
.bind_addr
;
789 laddr
= sctp_find_unmatch_addr(bp
, (union sctp_addr
*)addrs
,
791 if ((laddr
== NULL
) && (addrcnt
== 1)) {
792 if (asoc
->asconf_addr_del_pending
)
794 asoc
->asconf_addr_del_pending
=
795 kzalloc(sizeof(union sctp_addr
), GFP_ATOMIC
);
796 if (asoc
->asconf_addr_del_pending
== NULL
) {
800 asoc
->asconf_addr_del_pending
->sa
.sa_family
=
802 asoc
->asconf_addr_del_pending
->v4
.sin_port
=
804 if (addrs
->sa_family
== AF_INET
) {
805 struct sockaddr_in
*sin
;
807 sin
= (struct sockaddr_in
*)addrs
;
808 asoc
->asconf_addr_del_pending
->v4
.sin_addr
.s_addr
= sin
->sin_addr
.s_addr
;
809 } else if (addrs
->sa_family
== AF_INET6
) {
810 struct sockaddr_in6
*sin6
;
812 sin6
= (struct sockaddr_in6
*)addrs
;
813 asoc
->asconf_addr_del_pending
->v6
.sin6_addr
= sin6
->sin6_addr
;
815 SCTP_DEBUG_PRINTK_IPADDR("send_asconf_del_ip: keep the last address asoc: %p ",
816 " at %p\n", asoc
, asoc
->asconf_addr_del_pending
,
817 asoc
->asconf_addr_del_pending
);
818 asoc
->src_out_of_asoc_ok
= 1;
823 /* We do not need RCU protection throughout this loop
824 * because this is done under a socket lock from the
827 chunk
= sctp_make_asconf_update_ip(asoc
, laddr
, addrs
, addrcnt
,
835 /* Reset use_as_src flag for the addresses in the bind address
836 * list that are to be deleted.
839 for (i
= 0; i
< addrcnt
; i
++) {
841 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
842 list_for_each_entry(saddr
, &bp
->address_list
, list
) {
843 if (sctp_cmp_addr_exact(&saddr
->a
, laddr
))
844 saddr
->state
= SCTP_ADDR_DEL
;
846 addr_buf
+= af
->sockaddr_len
;
849 /* Update the route and saddr entries for all the transports
850 * as some of the addresses in the bind address list are
851 * about to be deleted and cannot be used as source addresses.
853 list_for_each_entry(transport
, &asoc
->peer
.transport_addr_list
,
855 dst_release(transport
->dst
);
856 sctp_transport_route(transport
, NULL
,
857 sctp_sk(asoc
->base
.sk
));
861 /* We don't need to transmit ASCONF */
863 retval
= sctp_send_asconf(asoc
, chunk
);
869 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
870 int sctp_asconf_mgmt(struct sctp_sock
*sp
, struct sctp_sockaddr_entry
*addrw
)
872 struct sock
*sk
= sctp_opt2sk(sp
);
873 union sctp_addr
*addr
;
876 /* It is safe to write port space in caller. */
878 addr
->v4
.sin_port
= htons(sp
->ep
->base
.bind_addr
.port
);
879 af
= sctp_get_af_specific(addr
->sa
.sa_family
);
882 if (sctp_verify_addr(sk
, addr
, af
->sockaddr_len
))
885 if (addrw
->state
== SCTP_ADDR_NEW
)
886 return sctp_send_asconf_add_ip(sk
, (struct sockaddr
*)addr
, 1);
888 return sctp_send_asconf_del_ip(sk
, (struct sockaddr
*)addr
, 1);
891 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
894 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
897 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
898 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
901 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
902 * Section 3.1.2 for this usage.
904 * addrs is a pointer to an array of one or more socket addresses. Each
905 * address is contained in its appropriate structure (i.e. struct
906 * sockaddr_in or struct sockaddr_in6) the family of the address type
907 * must be used to distinguish the address length (note that this
908 * representation is termed a "packed array" of addresses). The caller
909 * specifies the number of addresses in the array with addrcnt.
911 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
912 * -1, and sets errno to the appropriate error code.
914 * For SCTP, the port given in each socket address must be the same, or
915 * sctp_bindx() will fail, setting errno to EINVAL.
917 * The flags parameter is formed from the bitwise OR of zero or more of
918 * the following currently defined flags:
920 * SCTP_BINDX_ADD_ADDR
922 * SCTP_BINDX_REM_ADDR
924 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
925 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
926 * addresses from the association. The two flags are mutually exclusive;
927 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
928 * not remove all addresses from an association; sctp_bindx() will
929 * reject such an attempt with EINVAL.
931 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
932 * additional addresses with an endpoint after calling bind(). Or use
933 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
934 * socket is associated with so that no new association accepted will be
935 * associated with those addresses. If the endpoint supports dynamic
936 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
937 * endpoint to send the appropriate message to the peer to change the
938 * peers address lists.
940 * Adding and removing addresses from a connected association is
941 * optional functionality. Implementations that do not support this
942 * functionality should return EOPNOTSUPP.
944 * Basically do nothing but copying the addresses from user to kernel
945 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
946 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
949 * We don't use copy_from_user() for optimization: we first do the
950 * sanity checks (buffer size -fast- and access check-healthy
951 * pointer); if all of those succeed, then we can alloc the memory
952 * (expensive operation) needed to copy the data to kernel. Then we do
953 * the copying without checking the user space area
954 * (__copy_from_user()).
956 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
959 * sk The sk of the socket
960 * addrs The pointer to the addresses in user land
961 * addrssize Size of the addrs buffer
962 * op Operation to perform (add or remove, see the flags of
965 * Returns 0 if ok, <0 errno code on error.
967 SCTP_STATIC
int sctp_setsockopt_bindx(struct sock
* sk
,
968 struct sockaddr __user
*addrs
,
969 int addrs_size
, int op
)
971 struct sockaddr
*kaddrs
;
975 struct sockaddr
*sa_addr
;
979 SCTP_DEBUG_PRINTK("sctp_setsockopt_bindx: sk %p addrs %p"
980 " addrs_size %d opt %d\n", sk
, addrs
, addrs_size
, op
);
982 if (unlikely(addrs_size
<= 0))
985 /* Check the user passed a healthy pointer. */
986 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
989 /* Alloc space for the address array in kernel memory. */
990 kaddrs
= kmalloc(addrs_size
, GFP_KERNEL
);
991 if (unlikely(!kaddrs
))
994 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
999 /* Walk through the addrs buffer and count the number of addresses. */
1001 while (walk_size
< addrs_size
) {
1002 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
1008 af
= sctp_get_af_specific(sa_addr
->sa_family
);
1010 /* If the address family is not supported or if this address
1011 * causes the address buffer to overflow return EINVAL.
1013 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
1018 addr_buf
+= af
->sockaddr_len
;
1019 walk_size
+= af
->sockaddr_len
;
1024 case SCTP_BINDX_ADD_ADDR
:
1025 err
= sctp_bindx_add(sk
, kaddrs
, addrcnt
);
1028 err
= sctp_send_asconf_add_ip(sk
, kaddrs
, addrcnt
);
1031 case SCTP_BINDX_REM_ADDR
:
1032 err
= sctp_bindx_rem(sk
, kaddrs
, addrcnt
);
1035 err
= sctp_send_asconf_del_ip(sk
, kaddrs
, addrcnt
);
1049 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1051 * Common routine for handling connect() and sctp_connectx().
1052 * Connect will come in with just a single address.
1054 static int __sctp_connect(struct sock
* sk
,
1055 struct sockaddr
*kaddrs
,
1057 sctp_assoc_t
*assoc_id
)
1059 struct net
*net
= sock_net(sk
);
1060 struct sctp_sock
*sp
;
1061 struct sctp_endpoint
*ep
;
1062 struct sctp_association
*asoc
= NULL
;
1063 struct sctp_association
*asoc2
;
1064 struct sctp_transport
*transport
;
1072 union sctp_addr
*sa_addr
= NULL
;
1074 unsigned short port
;
1075 unsigned int f_flags
= 0;
1080 /* connect() cannot be done on a socket that is already in ESTABLISHED
1081 * state - UDP-style peeled off socket or a TCP-style socket that
1082 * is already connected.
1083 * It cannot be done even on a TCP-style listening socket.
1085 if (sctp_sstate(sk
, ESTABLISHED
) ||
1086 (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))) {
1091 /* Walk through the addrs buffer and count the number of addresses. */
1093 while (walk_size
< addrs_size
) {
1094 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
1100 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
1102 /* If the address family is not supported or if this address
1103 * causes the address buffer to overflow return EINVAL.
1105 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
1110 port
= ntohs(sa_addr
->v4
.sin_port
);
1112 /* Save current address so we can work with it */
1113 memcpy(&to
, sa_addr
, af
->sockaddr_len
);
1115 err
= sctp_verify_addr(sk
, &to
, af
->sockaddr_len
);
1119 /* Make sure the destination port is correctly set
1122 if (asoc
&& asoc
->peer
.port
&& asoc
->peer
.port
!= port
) {
1127 /* Check if there already is a matching association on the
1128 * endpoint (other than the one created here).
1130 asoc2
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1131 if (asoc2
&& asoc2
!= asoc
) {
1132 if (asoc2
->state
>= SCTP_STATE_ESTABLISHED
)
1139 /* If we could not find a matching association on the endpoint,
1140 * make sure that there is no peeled-off association matching
1141 * the peer address even on another socket.
1143 if (sctp_endpoint_is_peeled_off(ep
, &to
)) {
1144 err
= -EADDRNOTAVAIL
;
1149 /* If a bind() or sctp_bindx() is not called prior to
1150 * an sctp_connectx() call, the system picks an
1151 * ephemeral port and will choose an address set
1152 * equivalent to binding with a wildcard address.
1154 if (!ep
->base
.bind_addr
.port
) {
1155 if (sctp_autobind(sk
)) {
1161 * If an unprivileged user inherits a 1-many
1162 * style socket with open associations on a
1163 * privileged port, it MAY be permitted to
1164 * accept new associations, but it SHOULD NOT
1165 * be permitted to open new associations.
1167 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1168 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
)) {
1174 scope
= sctp_scope(&to
);
1175 asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1181 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
,
1189 /* Prime the peer's transport structures. */
1190 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
,
1198 addr_buf
+= af
->sockaddr_len
;
1199 walk_size
+= af
->sockaddr_len
;
1202 /* In case the user of sctp_connectx() wants an association
1203 * id back, assign one now.
1206 err
= sctp_assoc_set_id(asoc
, GFP_KERNEL
);
1211 err
= sctp_primitive_ASSOCIATE(net
, asoc
, NULL
);
1216 /* Initialize sk's dport and daddr for getpeername() */
1217 inet_sk(sk
)->inet_dport
= htons(asoc
->peer
.port
);
1218 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
1219 af
->to_sk_daddr(sa_addr
, sk
);
1222 /* in-kernel sockets don't generally have a file allocated to them
1223 * if all they do is call sock_create_kern().
1225 if (sk
->sk_socket
->file
)
1226 f_flags
= sk
->sk_socket
->file
->f_flags
;
1228 timeo
= sock_sndtimeo(sk
, f_flags
& O_NONBLOCK
);
1230 err
= sctp_wait_for_connect(asoc
, &timeo
);
1231 if ((err
== 0 || err
== -EINPROGRESS
) && assoc_id
)
1232 *assoc_id
= asoc
->assoc_id
;
1234 /* Don't free association on exit. */
1239 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1240 " kaddrs: %p err: %d\n",
1243 /* sctp_primitive_ASSOCIATE may have added this association
1244 * To the hash table, try to unhash it, just in case, its a noop
1245 * if it wasn't hashed so we're safe
1247 sctp_unhash_established(asoc
);
1248 sctp_association_free(asoc
);
1253 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1256 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1257 * sctp_assoc_t *asoc);
1259 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1260 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1261 * or IPv6 addresses.
1263 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1264 * Section 3.1.2 for this usage.
1266 * addrs is a pointer to an array of one or more socket addresses. Each
1267 * address is contained in its appropriate structure (i.e. struct
1268 * sockaddr_in or struct sockaddr_in6) the family of the address type
1269 * must be used to distengish the address length (note that this
1270 * representation is termed a "packed array" of addresses). The caller
1271 * specifies the number of addresses in the array with addrcnt.
1273 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1274 * the association id of the new association. On failure, sctp_connectx()
1275 * returns -1, and sets errno to the appropriate error code. The assoc_id
1276 * is not touched by the kernel.
1278 * For SCTP, the port given in each socket address must be the same, or
1279 * sctp_connectx() will fail, setting errno to EINVAL.
1281 * An application can use sctp_connectx to initiate an association with
1282 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1283 * allows a caller to specify multiple addresses at which a peer can be
1284 * reached. The way the SCTP stack uses the list of addresses to set up
1285 * the association is implementation dependent. This function only
1286 * specifies that the stack will try to make use of all the addresses in
1287 * the list when needed.
1289 * Note that the list of addresses passed in is only used for setting up
1290 * the association. It does not necessarily equal the set of addresses
1291 * the peer uses for the resulting association. If the caller wants to
1292 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1293 * retrieve them after the association has been set up.
1295 * Basically do nothing but copying the addresses from user to kernel
1296 * land and invoking either sctp_connectx(). This is used for tunneling
1297 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1299 * We don't use copy_from_user() for optimization: we first do the
1300 * sanity checks (buffer size -fast- and access check-healthy
1301 * pointer); if all of those succeed, then we can alloc the memory
1302 * (expensive operation) needed to copy the data to kernel. Then we do
1303 * the copying without checking the user space area
1304 * (__copy_from_user()).
1306 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1309 * sk The sk of the socket
1310 * addrs The pointer to the addresses in user land
1311 * addrssize Size of the addrs buffer
1313 * Returns >=0 if ok, <0 errno code on error.
1315 SCTP_STATIC
int __sctp_setsockopt_connectx(struct sock
* sk
,
1316 struct sockaddr __user
*addrs
,
1318 sctp_assoc_t
*assoc_id
)
1321 struct sockaddr
*kaddrs
;
1323 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1324 __func__
, sk
, addrs
, addrs_size
);
1326 if (unlikely(addrs_size
<= 0))
1329 /* Check the user passed a healthy pointer. */
1330 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
1333 /* Alloc space for the address array in kernel memory. */
1334 kaddrs
= kmalloc(addrs_size
, GFP_KERNEL
);
1335 if (unlikely(!kaddrs
))
1338 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
1341 err
= __sctp_connect(sk
, kaddrs
, addrs_size
, assoc_id
);
1350 * This is an older interface. It's kept for backward compatibility
1351 * to the option that doesn't provide association id.
1353 SCTP_STATIC
int sctp_setsockopt_connectx_old(struct sock
* sk
,
1354 struct sockaddr __user
*addrs
,
1357 return __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, NULL
);
1361 * New interface for the API. The since the API is done with a socket
1362 * option, to make it simple we feed back the association id is as a return
1363 * indication to the call. Error is always negative and association id is
1366 SCTP_STATIC
int sctp_setsockopt_connectx(struct sock
* sk
,
1367 struct sockaddr __user
*addrs
,
1370 sctp_assoc_t assoc_id
= 0;
1373 err
= __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, &assoc_id
);
1382 * New (hopefully final) interface for the API.
1383 * We use the sctp_getaddrs_old structure so that use-space library
1384 * can avoid any unnecessary allocations. The only defferent part
1385 * is that we store the actual length of the address buffer into the
1386 * addrs_num structure member. That way we can re-use the existing
1389 SCTP_STATIC
int sctp_getsockopt_connectx3(struct sock
* sk
, int len
,
1390 char __user
*optval
,
1393 struct sctp_getaddrs_old param
;
1394 sctp_assoc_t assoc_id
= 0;
1397 if (len
< sizeof(param
))
1400 if (copy_from_user(¶m
, optval
, sizeof(param
)))
1403 err
= __sctp_setsockopt_connectx(sk
,
1404 (struct sockaddr __user
*)param
.addrs
,
1405 param
.addr_num
, &assoc_id
);
1407 if (err
== 0 || err
== -EINPROGRESS
) {
1408 if (copy_to_user(optval
, &assoc_id
, sizeof(assoc_id
)))
1410 if (put_user(sizeof(assoc_id
), optlen
))
1417 /* API 3.1.4 close() - UDP Style Syntax
1418 * Applications use close() to perform graceful shutdown (as described in
1419 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1420 * by a UDP-style socket.
1424 * ret = close(int sd);
1426 * sd - the socket descriptor of the associations to be closed.
1428 * To gracefully shutdown a specific association represented by the
1429 * UDP-style socket, an application should use the sendmsg() call,
1430 * passing no user data, but including the appropriate flag in the
1431 * ancillary data (see Section xxxx).
1433 * If sd in the close() call is a branched-off socket representing only
1434 * one association, the shutdown is performed on that association only.
1436 * 4.1.6 close() - TCP Style Syntax
1438 * Applications use close() to gracefully close down an association.
1442 * int close(int sd);
1444 * sd - the socket descriptor of the association to be closed.
1446 * After an application calls close() on a socket descriptor, no further
1447 * socket operations will succeed on that descriptor.
1449 * API 7.1.4 SO_LINGER
1451 * An application using the TCP-style socket can use this option to
1452 * perform the SCTP ABORT primitive. The linger option structure is:
1455 * int l_onoff; // option on/off
1456 * int l_linger; // linger time
1459 * To enable the option, set l_onoff to 1. If the l_linger value is set
1460 * to 0, calling close() is the same as the ABORT primitive. If the
1461 * value is set to a negative value, the setsockopt() call will return
1462 * an error. If the value is set to a positive value linger_time, the
1463 * close() can be blocked for at most linger_time ms. If the graceful
1464 * shutdown phase does not finish during this period, close() will
1465 * return but the graceful shutdown phase continues in the system.
1467 SCTP_STATIC
void sctp_close(struct sock
*sk
, long timeout
)
1469 struct net
*net
= sock_net(sk
);
1470 struct sctp_endpoint
*ep
;
1471 struct sctp_association
*asoc
;
1472 struct list_head
*pos
, *temp
;
1473 unsigned int data_was_unread
;
1475 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk
, timeout
);
1478 sk
->sk_shutdown
= SHUTDOWN_MASK
;
1479 sk
->sk_state
= SCTP_SS_CLOSING
;
1481 ep
= sctp_sk(sk
)->ep
;
1483 /* Clean up any skbs sitting on the receive queue. */
1484 data_was_unread
= sctp_queue_purge_ulpevents(&sk
->sk_receive_queue
);
1485 data_was_unread
+= sctp_queue_purge_ulpevents(&sctp_sk(sk
)->pd_lobby
);
1487 /* Walk all associations on an endpoint. */
1488 list_for_each_safe(pos
, temp
, &ep
->asocs
) {
1489 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
1491 if (sctp_style(sk
, TCP
)) {
1492 /* A closed association can still be in the list if
1493 * it belongs to a TCP-style listening socket that is
1494 * not yet accepted. If so, free it. If not, send an
1495 * ABORT or SHUTDOWN based on the linger options.
1497 if (sctp_state(asoc
, CLOSED
)) {
1498 sctp_unhash_established(asoc
);
1499 sctp_association_free(asoc
);
1504 if (data_was_unread
|| !skb_queue_empty(&asoc
->ulpq
.lobby
) ||
1505 !skb_queue_empty(&asoc
->ulpq
.reasm
) ||
1506 (sock_flag(sk
, SOCK_LINGER
) && !sk
->sk_lingertime
)) {
1507 struct sctp_chunk
*chunk
;
1509 chunk
= sctp_make_abort_user(asoc
, NULL
, 0);
1511 sctp_primitive_ABORT(net
, asoc
, chunk
);
1513 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
1516 /* On a TCP-style socket, block for at most linger_time if set. */
1517 if (sctp_style(sk
, TCP
) && timeout
)
1518 sctp_wait_for_close(sk
, timeout
);
1520 /* This will run the backlog queue. */
1521 sctp_release_sock(sk
);
1523 /* Supposedly, no process has access to the socket, but
1524 * the net layers still may.
1526 sctp_local_bh_disable();
1527 sctp_bh_lock_sock(sk
);
1529 /* Hold the sock, since sk_common_release() will put sock_put()
1530 * and we have just a little more cleanup.
1533 sk_common_release(sk
);
1535 sctp_bh_unlock_sock(sk
);
1536 sctp_local_bh_enable();
1540 SCTP_DBG_OBJCNT_DEC(sock
);
1543 /* Handle EPIPE error. */
1544 static int sctp_error(struct sock
*sk
, int flags
, int err
)
1547 err
= sock_error(sk
) ? : -EPIPE
;
1548 if (err
== -EPIPE
&& !(flags
& MSG_NOSIGNAL
))
1549 send_sig(SIGPIPE
, current
, 0);
1553 /* API 3.1.3 sendmsg() - UDP Style Syntax
1555 * An application uses sendmsg() and recvmsg() calls to transmit data to
1556 * and receive data from its peer.
1558 * ssize_t sendmsg(int socket, const struct msghdr *message,
1561 * socket - the socket descriptor of the endpoint.
1562 * message - pointer to the msghdr structure which contains a single
1563 * user message and possibly some ancillary data.
1565 * See Section 5 for complete description of the data
1568 * flags - flags sent or received with the user message, see Section
1569 * 5 for complete description of the flags.
1571 * Note: This function could use a rewrite especially when explicit
1572 * connect support comes in.
1574 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1576 SCTP_STATIC
int sctp_msghdr_parse(const struct msghdr
*, sctp_cmsgs_t
*);
1578 SCTP_STATIC
int sctp_sendmsg(struct kiocb
*iocb
, struct sock
*sk
,
1579 struct msghdr
*msg
, size_t msg_len
)
1581 struct net
*net
= sock_net(sk
);
1582 struct sctp_sock
*sp
;
1583 struct sctp_endpoint
*ep
;
1584 struct sctp_association
*new_asoc
=NULL
, *asoc
=NULL
;
1585 struct sctp_transport
*transport
, *chunk_tp
;
1586 struct sctp_chunk
*chunk
;
1588 struct sockaddr
*msg_name
= NULL
;
1589 struct sctp_sndrcvinfo default_sinfo
;
1590 struct sctp_sndrcvinfo
*sinfo
;
1591 struct sctp_initmsg
*sinit
;
1592 sctp_assoc_t associd
= 0;
1593 sctp_cmsgs_t cmsgs
= { NULL
};
1597 __u16 sinfo_flags
= 0;
1598 struct sctp_datamsg
*datamsg
;
1599 int msg_flags
= msg
->msg_flags
;
1601 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1608 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep
);
1610 /* We cannot send a message over a TCP-style listening socket. */
1611 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
)) {
1616 /* Parse out the SCTP CMSGs. */
1617 err
= sctp_msghdr_parse(msg
, &cmsgs
);
1620 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err
);
1624 /* Fetch the destination address for this packet. This
1625 * address only selects the association--it is not necessarily
1626 * the address we will send to.
1627 * For a peeled-off socket, msg_name is ignored.
1629 if (!sctp_style(sk
, UDP_HIGH_BANDWIDTH
) && msg
->msg_name
) {
1630 int msg_namelen
= msg
->msg_namelen
;
1632 err
= sctp_verify_addr(sk
, (union sctp_addr
*)msg
->msg_name
,
1637 if (msg_namelen
> sizeof(to
))
1638 msg_namelen
= sizeof(to
);
1639 memcpy(&to
, msg
->msg_name
, msg_namelen
);
1640 msg_name
= msg
->msg_name
;
1646 /* Did the user specify SNDRCVINFO? */
1648 sinfo_flags
= sinfo
->sinfo_flags
;
1649 associd
= sinfo
->sinfo_assoc_id
;
1652 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1653 msg_len
, sinfo_flags
);
1655 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1656 if (sctp_style(sk
, TCP
) && (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
))) {
1661 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1662 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1663 * If SCTP_ABORT is set, the message length could be non zero with
1664 * the msg_iov set to the user abort reason.
1666 if (((sinfo_flags
& SCTP_EOF
) && (msg_len
> 0)) ||
1667 (!(sinfo_flags
& (SCTP_EOF
|SCTP_ABORT
)) && (msg_len
== 0))) {
1672 /* If SCTP_ADDR_OVER is set, there must be an address
1673 * specified in msg_name.
1675 if ((sinfo_flags
& SCTP_ADDR_OVER
) && (!msg
->msg_name
)) {
1682 SCTP_DEBUG_PRINTK("About to look up association.\n");
1686 /* If a msg_name has been specified, assume this is to be used. */
1688 /* Look for a matching association on the endpoint. */
1689 asoc
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1691 /* If we could not find a matching association on the
1692 * endpoint, make sure that it is not a TCP-style
1693 * socket that already has an association or there is
1694 * no peeled-off association on another socket.
1696 if ((sctp_style(sk
, TCP
) &&
1697 sctp_sstate(sk
, ESTABLISHED
)) ||
1698 sctp_endpoint_is_peeled_off(ep
, &to
)) {
1699 err
= -EADDRNOTAVAIL
;
1704 asoc
= sctp_id2assoc(sk
, associd
);
1712 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc
);
1714 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1715 * socket that has an association in CLOSED state. This can
1716 * happen when an accepted socket has an association that is
1719 if (sctp_state(asoc
, CLOSED
) && sctp_style(sk
, TCP
)) {
1724 if (sinfo_flags
& SCTP_EOF
) {
1725 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1727 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
1731 if (sinfo_flags
& SCTP_ABORT
) {
1733 chunk
= sctp_make_abort_user(asoc
, msg
, msg_len
);
1739 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc
);
1740 sctp_primitive_ABORT(net
, asoc
, chunk
);
1746 /* Do we need to create the association? */
1748 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1750 if (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
)) {
1755 /* Check for invalid stream against the stream counts,
1756 * either the default or the user specified stream counts.
1759 if (!sinit
|| (sinit
&& !sinit
->sinit_num_ostreams
)) {
1760 /* Check against the defaults. */
1761 if (sinfo
->sinfo_stream
>=
1762 sp
->initmsg
.sinit_num_ostreams
) {
1767 /* Check against the requested. */
1768 if (sinfo
->sinfo_stream
>=
1769 sinit
->sinit_num_ostreams
) {
1777 * API 3.1.2 bind() - UDP Style Syntax
1778 * If a bind() or sctp_bindx() is not called prior to a
1779 * sendmsg() call that initiates a new association, the
1780 * system picks an ephemeral port and will choose an address
1781 * set equivalent to binding with a wildcard address.
1783 if (!ep
->base
.bind_addr
.port
) {
1784 if (sctp_autobind(sk
)) {
1790 * If an unprivileged user inherits a one-to-many
1791 * style socket with open associations on a privileged
1792 * port, it MAY be permitted to accept new associations,
1793 * but it SHOULD NOT be permitted to open new
1796 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1797 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
)) {
1803 scope
= sctp_scope(&to
);
1804 new_asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1810 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
, GFP_KERNEL
);
1816 /* If the SCTP_INIT ancillary data is specified, set all
1817 * the association init values accordingly.
1820 if (sinit
->sinit_num_ostreams
) {
1821 asoc
->c
.sinit_num_ostreams
=
1822 sinit
->sinit_num_ostreams
;
1824 if (sinit
->sinit_max_instreams
) {
1825 asoc
->c
.sinit_max_instreams
=
1826 sinit
->sinit_max_instreams
;
1828 if (sinit
->sinit_max_attempts
) {
1829 asoc
->max_init_attempts
1830 = sinit
->sinit_max_attempts
;
1832 if (sinit
->sinit_max_init_timeo
) {
1833 asoc
->max_init_timeo
=
1834 msecs_to_jiffies(sinit
->sinit_max_init_timeo
);
1838 /* Prime the peer's transport structures. */
1839 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
, SCTP_UNKNOWN
);
1846 /* ASSERT: we have a valid association at this point. */
1847 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1850 /* If the user didn't specify SNDRCVINFO, make up one with
1853 memset(&default_sinfo
, 0, sizeof(default_sinfo
));
1854 default_sinfo
.sinfo_stream
= asoc
->default_stream
;
1855 default_sinfo
.sinfo_flags
= asoc
->default_flags
;
1856 default_sinfo
.sinfo_ppid
= asoc
->default_ppid
;
1857 default_sinfo
.sinfo_context
= asoc
->default_context
;
1858 default_sinfo
.sinfo_timetolive
= asoc
->default_timetolive
;
1859 default_sinfo
.sinfo_assoc_id
= sctp_assoc2id(asoc
);
1860 sinfo
= &default_sinfo
;
1863 /* API 7.1.7, the sndbuf size per association bounds the
1864 * maximum size of data that can be sent in a single send call.
1866 if (msg_len
> sk
->sk_sndbuf
) {
1871 if (asoc
->pmtu_pending
)
1872 sctp_assoc_pending_pmtu(sk
, asoc
);
1874 /* If fragmentation is disabled and the message length exceeds the
1875 * association fragmentation point, return EMSGSIZE. The I-D
1876 * does not specify what this error is, but this looks like
1879 if (sctp_sk(sk
)->disable_fragments
&& (msg_len
> asoc
->frag_point
)) {
1884 /* Check for invalid stream. */
1885 if (sinfo
->sinfo_stream
>= asoc
->c
.sinit_num_ostreams
) {
1890 timeo
= sock_sndtimeo(sk
, msg
->msg_flags
& MSG_DONTWAIT
);
1891 if (!sctp_wspace(asoc
)) {
1892 err
= sctp_wait_for_sndbuf(asoc
, &timeo
, msg_len
);
1897 /* If an address is passed with the sendto/sendmsg call, it is used
1898 * to override the primary destination address in the TCP model, or
1899 * when SCTP_ADDR_OVER flag is set in the UDP model.
1901 if ((sctp_style(sk
, TCP
) && msg_name
) ||
1902 (sinfo_flags
& SCTP_ADDR_OVER
)) {
1903 chunk_tp
= sctp_assoc_lookup_paddr(asoc
, &to
);
1911 /* Auto-connect, if we aren't connected already. */
1912 if (sctp_state(asoc
, CLOSED
)) {
1913 err
= sctp_primitive_ASSOCIATE(net
, asoc
, NULL
);
1916 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1919 /* Break the message into multiple chunks of maximum size. */
1920 datamsg
= sctp_datamsg_from_user(asoc
, sinfo
, msg
, msg_len
);
1921 if (IS_ERR(datamsg
)) {
1922 err
= PTR_ERR(datamsg
);
1926 /* Now send the (possibly) fragmented message. */
1927 list_for_each_entry(chunk
, &datamsg
->chunks
, frag_list
) {
1928 sctp_chunk_hold(chunk
);
1930 /* Do accounting for the write space. */
1931 sctp_set_owner_w(chunk
);
1933 chunk
->transport
= chunk_tp
;
1936 /* Send it to the lower layers. Note: all chunks
1937 * must either fail or succeed. The lower layer
1938 * works that way today. Keep it that way or this
1941 err
= sctp_primitive_SEND(net
, asoc
, datamsg
);
1942 /* Did the lower layer accept the chunk? */
1944 sctp_datamsg_free(datamsg
);
1946 sctp_datamsg_put(datamsg
);
1948 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1955 /* If we are already past ASSOCIATE, the lower
1956 * layers are responsible for association cleanup.
1962 sctp_unhash_established(asoc
);
1963 sctp_association_free(asoc
);
1966 sctp_release_sock(sk
);
1969 return sctp_error(sk
, msg_flags
, err
);
1976 err
= sock_error(sk
);
1986 /* This is an extended version of skb_pull() that removes the data from the
1987 * start of a skb even when data is spread across the list of skb's in the
1988 * frag_list. len specifies the total amount of data that needs to be removed.
1989 * when 'len' bytes could be removed from the skb, it returns 0.
1990 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1991 * could not be removed.
1993 static int sctp_skb_pull(struct sk_buff
*skb
, int len
)
1995 struct sk_buff
*list
;
1996 int skb_len
= skb_headlen(skb
);
1999 if (len
<= skb_len
) {
2000 __skb_pull(skb
, len
);
2004 __skb_pull(skb
, skb_len
);
2006 skb_walk_frags(skb
, list
) {
2007 rlen
= sctp_skb_pull(list
, len
);
2008 skb
->len
-= (len
-rlen
);
2009 skb
->data_len
-= (len
-rlen
);
2020 /* API 3.1.3 recvmsg() - UDP Style Syntax
2022 * ssize_t recvmsg(int socket, struct msghdr *message,
2025 * socket - the socket descriptor of the endpoint.
2026 * message - pointer to the msghdr structure which contains a single
2027 * user message and possibly some ancillary data.
2029 * See Section 5 for complete description of the data
2032 * flags - flags sent or received with the user message, see Section
2033 * 5 for complete description of the flags.
2035 static struct sk_buff
*sctp_skb_recv_datagram(struct sock
*, int, int, int *);
2037 SCTP_STATIC
int sctp_recvmsg(struct kiocb
*iocb
, struct sock
*sk
,
2038 struct msghdr
*msg
, size_t len
, int noblock
,
2039 int flags
, int *addr_len
)
2041 struct sctp_ulpevent
*event
= NULL
;
2042 struct sctp_sock
*sp
= sctp_sk(sk
);
2043 struct sk_buff
*skb
;
2048 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
2049 "0x%x, %s: %p)\n", "sk", sk
, "msghdr", msg
,
2050 "len", len
, "knoblauch", noblock
,
2051 "flags", flags
, "addr_len", addr_len
);
2055 if (sctp_style(sk
, TCP
) && !sctp_sstate(sk
, ESTABLISHED
)) {
2060 skb
= sctp_skb_recv_datagram(sk
, flags
, noblock
, &err
);
2064 /* Get the total length of the skb including any skb's in the
2073 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
2075 event
= sctp_skb2event(skb
);
2080 sock_recv_ts_and_drops(msg
, sk
, skb
);
2081 if (sctp_ulpevent_is_notification(event
)) {
2082 msg
->msg_flags
|= MSG_NOTIFICATION
;
2083 sp
->pf
->event_msgname(event
, msg
->msg_name
, addr_len
);
2085 sp
->pf
->skb_msgname(skb
, msg
->msg_name
, addr_len
);
2088 /* Check if we allow SCTP_SNDRCVINFO. */
2089 if (sp
->subscribe
.sctp_data_io_event
)
2090 sctp_ulpevent_read_sndrcvinfo(event
, msg
);
2092 /* FIXME: we should be calling IP/IPv6 layers. */
2093 if (sk
->sk_protinfo
.af_inet
.cmsg_flags
)
2094 ip_cmsg_recv(msg
, skb
);
2099 /* If skb's length exceeds the user's buffer, update the skb and
2100 * push it back to the receive_queue so that the next call to
2101 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2103 if (skb_len
> copied
) {
2104 msg
->msg_flags
&= ~MSG_EOR
;
2105 if (flags
& MSG_PEEK
)
2107 sctp_skb_pull(skb
, copied
);
2108 skb_queue_head(&sk
->sk_receive_queue
, skb
);
2110 /* When only partial message is copied to the user, increase
2111 * rwnd by that amount. If all the data in the skb is read,
2112 * rwnd is updated when the event is freed.
2114 if (!sctp_ulpevent_is_notification(event
))
2115 sctp_assoc_rwnd_increase(event
->asoc
, copied
);
2117 } else if ((event
->msg_flags
& MSG_NOTIFICATION
) ||
2118 (event
->msg_flags
& MSG_EOR
))
2119 msg
->msg_flags
|= MSG_EOR
;
2121 msg
->msg_flags
&= ~MSG_EOR
;
2124 if (flags
& MSG_PEEK
) {
2125 /* Release the skb reference acquired after peeking the skb in
2126 * sctp_skb_recv_datagram().
2130 /* Free the event which includes releasing the reference to
2131 * the owner of the skb, freeing the skb and updating the
2134 sctp_ulpevent_free(event
);
2137 sctp_release_sock(sk
);
2141 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2143 * This option is a on/off flag. If enabled no SCTP message
2144 * fragmentation will be performed. Instead if a message being sent
2145 * exceeds the current PMTU size, the message will NOT be sent and
2146 * instead a error will be indicated to the user.
2148 static int sctp_setsockopt_disable_fragments(struct sock
*sk
,
2149 char __user
*optval
,
2150 unsigned int optlen
)
2154 if (optlen
< sizeof(int))
2157 if (get_user(val
, (int __user
*)optval
))
2160 sctp_sk(sk
)->disable_fragments
= (val
== 0) ? 0 : 1;
2165 static int sctp_setsockopt_events(struct sock
*sk
, char __user
*optval
,
2166 unsigned int optlen
)
2168 struct sctp_association
*asoc
;
2169 struct sctp_ulpevent
*event
;
2171 if (optlen
> sizeof(struct sctp_event_subscribe
))
2173 if (copy_from_user(&sctp_sk(sk
)->subscribe
, optval
, optlen
))
2177 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2178 * if there is no data to be sent or retransmit, the stack will
2179 * immediately send up this notification.
2181 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT
,
2182 &sctp_sk(sk
)->subscribe
)) {
2183 asoc
= sctp_id2assoc(sk
, 0);
2185 if (asoc
&& sctp_outq_is_empty(&asoc
->outqueue
)) {
2186 event
= sctp_ulpevent_make_sender_dry_event(asoc
,
2191 sctp_ulpq_tail_event(&asoc
->ulpq
, event
);
2198 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2200 * This socket option is applicable to the UDP-style socket only. When
2201 * set it will cause associations that are idle for more than the
2202 * specified number of seconds to automatically close. An association
2203 * being idle is defined an association that has NOT sent or received
2204 * user data. The special value of '0' indicates that no automatic
2205 * close of any associations should be performed. The option expects an
2206 * integer defining the number of seconds of idle time before an
2207 * association is closed.
2209 static int sctp_setsockopt_autoclose(struct sock
*sk
, char __user
*optval
,
2210 unsigned int optlen
)
2212 struct sctp_sock
*sp
= sctp_sk(sk
);
2214 /* Applicable to UDP-style socket only */
2215 if (sctp_style(sk
, TCP
))
2217 if (optlen
!= sizeof(int))
2219 if (copy_from_user(&sp
->autoclose
, optval
, optlen
))
2225 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2227 * Applications can enable or disable heartbeats for any peer address of
2228 * an association, modify an address's heartbeat interval, force a
2229 * heartbeat to be sent immediately, and adjust the address's maximum
2230 * number of retransmissions sent before an address is considered
2231 * unreachable. The following structure is used to access and modify an
2232 * address's parameters:
2234 * struct sctp_paddrparams {
2235 * sctp_assoc_t spp_assoc_id;
2236 * struct sockaddr_storage spp_address;
2237 * uint32_t spp_hbinterval;
2238 * uint16_t spp_pathmaxrxt;
2239 * uint32_t spp_pathmtu;
2240 * uint32_t spp_sackdelay;
2241 * uint32_t spp_flags;
2244 * spp_assoc_id - (one-to-many style socket) This is filled in the
2245 * application, and identifies the association for
2247 * spp_address - This specifies which address is of interest.
2248 * spp_hbinterval - This contains the value of the heartbeat interval,
2249 * in milliseconds. If a value of zero
2250 * is present in this field then no changes are to
2251 * be made to this parameter.
2252 * spp_pathmaxrxt - This contains the maximum number of
2253 * retransmissions before this address shall be
2254 * considered unreachable. If a value of zero
2255 * is present in this field then no changes are to
2256 * be made to this parameter.
2257 * spp_pathmtu - When Path MTU discovery is disabled the value
2258 * specified here will be the "fixed" path mtu.
2259 * Note that if the spp_address field is empty
2260 * then all associations on this address will
2261 * have this fixed path mtu set upon them.
2263 * spp_sackdelay - When delayed sack is enabled, this value specifies
2264 * the number of milliseconds that sacks will be delayed
2265 * for. This value will apply to all addresses of an
2266 * association if the spp_address field is empty. Note
2267 * also, that if delayed sack is enabled and this
2268 * value is set to 0, no change is made to the last
2269 * recorded delayed sack timer value.
2271 * spp_flags - These flags are used to control various features
2272 * on an association. The flag field may contain
2273 * zero or more of the following options.
2275 * SPP_HB_ENABLE - Enable heartbeats on the
2276 * specified address. Note that if the address
2277 * field is empty all addresses for the association
2278 * have heartbeats enabled upon them.
2280 * SPP_HB_DISABLE - Disable heartbeats on the
2281 * speicifed address. Note that if the address
2282 * field is empty all addresses for the association
2283 * will have their heartbeats disabled. Note also
2284 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2285 * mutually exclusive, only one of these two should
2286 * be specified. Enabling both fields will have
2287 * undetermined results.
2289 * SPP_HB_DEMAND - Request a user initiated heartbeat
2290 * to be made immediately.
2292 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2293 * heartbeat delayis to be set to the value of 0
2296 * SPP_PMTUD_ENABLE - This field will enable PMTU
2297 * discovery upon the specified address. Note that
2298 * if the address feild is empty then all addresses
2299 * on the association are effected.
2301 * SPP_PMTUD_DISABLE - This field will disable PMTU
2302 * discovery upon the specified address. Note that
2303 * if the address feild is empty then all addresses
2304 * on the association are effected. Not also that
2305 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2306 * exclusive. Enabling both will have undetermined
2309 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2310 * on delayed sack. The time specified in spp_sackdelay
2311 * is used to specify the sack delay for this address. Note
2312 * that if spp_address is empty then all addresses will
2313 * enable delayed sack and take on the sack delay
2314 * value specified in spp_sackdelay.
2315 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2316 * off delayed sack. If the spp_address field is blank then
2317 * delayed sack is disabled for the entire association. Note
2318 * also that this field is mutually exclusive to
2319 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2322 static int sctp_apply_peer_addr_params(struct sctp_paddrparams
*params
,
2323 struct sctp_transport
*trans
,
2324 struct sctp_association
*asoc
,
2325 struct sctp_sock
*sp
,
2328 int sackdelay_change
)
2332 if (params
->spp_flags
& SPP_HB_DEMAND
&& trans
) {
2333 struct net
*net
= sock_net(trans
->asoc
->base
.sk
);
2335 error
= sctp_primitive_REQUESTHEARTBEAT(net
, trans
->asoc
, trans
);
2340 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2341 * this field is ignored. Note also that a value of zero indicates
2342 * the current setting should be left unchanged.
2344 if (params
->spp_flags
& SPP_HB_ENABLE
) {
2346 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2347 * set. This lets us use 0 value when this flag
2350 if (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)
2351 params
->spp_hbinterval
= 0;
2353 if (params
->spp_hbinterval
||
2354 (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)) {
2357 msecs_to_jiffies(params
->spp_hbinterval
);
2360 msecs_to_jiffies(params
->spp_hbinterval
);
2362 sp
->hbinterval
= params
->spp_hbinterval
;
2369 trans
->param_flags
=
2370 (trans
->param_flags
& ~SPP_HB
) | hb_change
;
2373 (asoc
->param_flags
& ~SPP_HB
) | hb_change
;
2376 (sp
->param_flags
& ~SPP_HB
) | hb_change
;
2380 /* When Path MTU discovery is disabled the value specified here will
2381 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2382 * include the flag SPP_PMTUD_DISABLE for this field to have any
2385 if ((params
->spp_flags
& SPP_PMTUD_DISABLE
) && params
->spp_pathmtu
) {
2387 trans
->pathmtu
= params
->spp_pathmtu
;
2388 sctp_assoc_sync_pmtu(sctp_opt2sk(sp
), asoc
);
2390 asoc
->pathmtu
= params
->spp_pathmtu
;
2391 sctp_frag_point(asoc
, params
->spp_pathmtu
);
2393 sp
->pathmtu
= params
->spp_pathmtu
;
2399 int update
= (trans
->param_flags
& SPP_PMTUD_DISABLE
) &&
2400 (params
->spp_flags
& SPP_PMTUD_ENABLE
);
2401 trans
->param_flags
=
2402 (trans
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2404 sctp_transport_pmtu(trans
, sctp_opt2sk(sp
));
2405 sctp_assoc_sync_pmtu(sctp_opt2sk(sp
), asoc
);
2409 (asoc
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2412 (sp
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2416 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2417 * value of this field is ignored. Note also that a value of zero
2418 * indicates the current setting should be left unchanged.
2420 if ((params
->spp_flags
& SPP_SACKDELAY_ENABLE
) && params
->spp_sackdelay
) {
2423 msecs_to_jiffies(params
->spp_sackdelay
);
2426 msecs_to_jiffies(params
->spp_sackdelay
);
2428 sp
->sackdelay
= params
->spp_sackdelay
;
2432 if (sackdelay_change
) {
2434 trans
->param_flags
=
2435 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2439 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2443 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2448 /* Note that a value of zero indicates the current setting should be
2451 if (params
->spp_pathmaxrxt
) {
2453 trans
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2455 asoc
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2457 sp
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2464 static int sctp_setsockopt_peer_addr_params(struct sock
*sk
,
2465 char __user
*optval
,
2466 unsigned int optlen
)
2468 struct sctp_paddrparams params
;
2469 struct sctp_transport
*trans
= NULL
;
2470 struct sctp_association
*asoc
= NULL
;
2471 struct sctp_sock
*sp
= sctp_sk(sk
);
2473 int hb_change
, pmtud_change
, sackdelay_change
;
2475 if (optlen
!= sizeof(struct sctp_paddrparams
))
2478 if (copy_from_user(¶ms
, optval
, optlen
))
2481 /* Validate flags and value parameters. */
2482 hb_change
= params
.spp_flags
& SPP_HB
;
2483 pmtud_change
= params
.spp_flags
& SPP_PMTUD
;
2484 sackdelay_change
= params
.spp_flags
& SPP_SACKDELAY
;
2486 if (hb_change
== SPP_HB
||
2487 pmtud_change
== SPP_PMTUD
||
2488 sackdelay_change
== SPP_SACKDELAY
||
2489 params
.spp_sackdelay
> 500 ||
2490 (params
.spp_pathmtu
&&
2491 params
.spp_pathmtu
< SCTP_DEFAULT_MINSEGMENT
))
2494 /* If an address other than INADDR_ANY is specified, and
2495 * no transport is found, then the request is invalid.
2497 if (!sctp_is_any(sk
, ( union sctp_addr
*)¶ms
.spp_address
)) {
2498 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
2499 params
.spp_assoc_id
);
2504 /* Get association, if assoc_id != 0 and the socket is a one
2505 * to many style socket, and an association was not found, then
2506 * the id was invalid.
2508 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
2509 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
))
2512 /* Heartbeat demand can only be sent on a transport or
2513 * association, but not a socket.
2515 if (params
.spp_flags
& SPP_HB_DEMAND
&& !trans
&& !asoc
)
2518 /* Process parameters. */
2519 error
= sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2520 hb_change
, pmtud_change
,
2526 /* If changes are for association, also apply parameters to each
2529 if (!trans
&& asoc
) {
2530 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2532 sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2533 hb_change
, pmtud_change
,
2542 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2544 * This option will effect the way delayed acks are performed. This
2545 * option allows you to get or set the delayed ack time, in
2546 * milliseconds. It also allows changing the delayed ack frequency.
2547 * Changing the frequency to 1 disables the delayed sack algorithm. If
2548 * the assoc_id is 0, then this sets or gets the endpoints default
2549 * values. If the assoc_id field is non-zero, then the set or get
2550 * effects the specified association for the one to many model (the
2551 * assoc_id field is ignored by the one to one model). Note that if
2552 * sack_delay or sack_freq are 0 when setting this option, then the
2553 * current values will remain unchanged.
2555 * struct sctp_sack_info {
2556 * sctp_assoc_t sack_assoc_id;
2557 * uint32_t sack_delay;
2558 * uint32_t sack_freq;
2561 * sack_assoc_id - This parameter, indicates which association the user
2562 * is performing an action upon. Note that if this field's value is
2563 * zero then the endpoints default value is changed (effecting future
2564 * associations only).
2566 * sack_delay - This parameter contains the number of milliseconds that
2567 * the user is requesting the delayed ACK timer be set to. Note that
2568 * this value is defined in the standard to be between 200 and 500
2571 * sack_freq - This parameter contains the number of packets that must
2572 * be received before a sack is sent without waiting for the delay
2573 * timer to expire. The default value for this is 2, setting this
2574 * value to 1 will disable the delayed sack algorithm.
2577 static int sctp_setsockopt_delayed_ack(struct sock
*sk
,
2578 char __user
*optval
, unsigned int optlen
)
2580 struct sctp_sack_info params
;
2581 struct sctp_transport
*trans
= NULL
;
2582 struct sctp_association
*asoc
= NULL
;
2583 struct sctp_sock
*sp
= sctp_sk(sk
);
2585 if (optlen
== sizeof(struct sctp_sack_info
)) {
2586 if (copy_from_user(¶ms
, optval
, optlen
))
2589 if (params
.sack_delay
== 0 && params
.sack_freq
== 0)
2591 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
2592 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2593 pr_warn("Use struct sctp_sack_info instead\n");
2594 if (copy_from_user(¶ms
, optval
, optlen
))
2597 if (params
.sack_delay
== 0)
2598 params
.sack_freq
= 1;
2600 params
.sack_freq
= 0;
2604 /* Validate value parameter. */
2605 if (params
.sack_delay
> 500)
2608 /* Get association, if sack_assoc_id != 0 and the socket is a one
2609 * to many style socket, and an association was not found, then
2610 * the id was invalid.
2612 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
2613 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
2616 if (params
.sack_delay
) {
2619 msecs_to_jiffies(params
.sack_delay
);
2621 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2622 SPP_SACKDELAY_ENABLE
;
2624 sp
->sackdelay
= params
.sack_delay
;
2626 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2627 SPP_SACKDELAY_ENABLE
;
2631 if (params
.sack_freq
== 1) {
2634 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2635 SPP_SACKDELAY_DISABLE
;
2638 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2639 SPP_SACKDELAY_DISABLE
;
2641 } else if (params
.sack_freq
> 1) {
2643 asoc
->sackfreq
= params
.sack_freq
;
2645 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2646 SPP_SACKDELAY_ENABLE
;
2648 sp
->sackfreq
= params
.sack_freq
;
2650 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2651 SPP_SACKDELAY_ENABLE
;
2655 /* If change is for association, also apply to each transport. */
2657 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2659 if (params
.sack_delay
) {
2661 msecs_to_jiffies(params
.sack_delay
);
2662 trans
->param_flags
=
2663 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2664 SPP_SACKDELAY_ENABLE
;
2666 if (params
.sack_freq
== 1) {
2667 trans
->param_flags
=
2668 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2669 SPP_SACKDELAY_DISABLE
;
2670 } else if (params
.sack_freq
> 1) {
2671 trans
->sackfreq
= params
.sack_freq
;
2672 trans
->param_flags
=
2673 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2674 SPP_SACKDELAY_ENABLE
;
2682 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2684 * Applications can specify protocol parameters for the default association
2685 * initialization. The option name argument to setsockopt() and getsockopt()
2688 * Setting initialization parameters is effective only on an unconnected
2689 * socket (for UDP-style sockets only future associations are effected
2690 * by the change). With TCP-style sockets, this option is inherited by
2691 * sockets derived from a listener socket.
2693 static int sctp_setsockopt_initmsg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2695 struct sctp_initmsg sinit
;
2696 struct sctp_sock
*sp
= sctp_sk(sk
);
2698 if (optlen
!= sizeof(struct sctp_initmsg
))
2700 if (copy_from_user(&sinit
, optval
, optlen
))
2703 if (sinit
.sinit_num_ostreams
)
2704 sp
->initmsg
.sinit_num_ostreams
= sinit
.sinit_num_ostreams
;
2705 if (sinit
.sinit_max_instreams
)
2706 sp
->initmsg
.sinit_max_instreams
= sinit
.sinit_max_instreams
;
2707 if (sinit
.sinit_max_attempts
)
2708 sp
->initmsg
.sinit_max_attempts
= sinit
.sinit_max_attempts
;
2709 if (sinit
.sinit_max_init_timeo
)
2710 sp
->initmsg
.sinit_max_init_timeo
= sinit
.sinit_max_init_timeo
;
2716 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2718 * Applications that wish to use the sendto() system call may wish to
2719 * specify a default set of parameters that would normally be supplied
2720 * through the inclusion of ancillary data. This socket option allows
2721 * such an application to set the default sctp_sndrcvinfo structure.
2722 * The application that wishes to use this socket option simply passes
2723 * in to this call the sctp_sndrcvinfo structure defined in Section
2724 * 5.2.2) The input parameters accepted by this call include
2725 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2726 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2727 * to this call if the caller is using the UDP model.
2729 static int sctp_setsockopt_default_send_param(struct sock
*sk
,
2730 char __user
*optval
,
2731 unsigned int optlen
)
2733 struct sctp_sndrcvinfo info
;
2734 struct sctp_association
*asoc
;
2735 struct sctp_sock
*sp
= sctp_sk(sk
);
2737 if (optlen
!= sizeof(struct sctp_sndrcvinfo
))
2739 if (copy_from_user(&info
, optval
, optlen
))
2742 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
2743 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
2747 asoc
->default_stream
= info
.sinfo_stream
;
2748 asoc
->default_flags
= info
.sinfo_flags
;
2749 asoc
->default_ppid
= info
.sinfo_ppid
;
2750 asoc
->default_context
= info
.sinfo_context
;
2751 asoc
->default_timetolive
= info
.sinfo_timetolive
;
2753 sp
->default_stream
= info
.sinfo_stream
;
2754 sp
->default_flags
= info
.sinfo_flags
;
2755 sp
->default_ppid
= info
.sinfo_ppid
;
2756 sp
->default_context
= info
.sinfo_context
;
2757 sp
->default_timetolive
= info
.sinfo_timetolive
;
2763 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2765 * Requests that the local SCTP stack use the enclosed peer address as
2766 * the association primary. The enclosed address must be one of the
2767 * association peer's addresses.
2769 static int sctp_setsockopt_primary_addr(struct sock
*sk
, char __user
*optval
,
2770 unsigned int optlen
)
2772 struct sctp_prim prim
;
2773 struct sctp_transport
*trans
;
2775 if (optlen
!= sizeof(struct sctp_prim
))
2778 if (copy_from_user(&prim
, optval
, sizeof(struct sctp_prim
)))
2781 trans
= sctp_addr_id2transport(sk
, &prim
.ssp_addr
, prim
.ssp_assoc_id
);
2785 sctp_assoc_set_primary(trans
->asoc
, trans
);
2791 * 7.1.5 SCTP_NODELAY
2793 * Turn on/off any Nagle-like algorithm. This means that packets are
2794 * generally sent as soon as possible and no unnecessary delays are
2795 * introduced, at the cost of more packets in the network. Expects an
2796 * integer boolean flag.
2798 static int sctp_setsockopt_nodelay(struct sock
*sk
, char __user
*optval
,
2799 unsigned int optlen
)
2803 if (optlen
< sizeof(int))
2805 if (get_user(val
, (int __user
*)optval
))
2808 sctp_sk(sk
)->nodelay
= (val
== 0) ? 0 : 1;
2814 * 7.1.1 SCTP_RTOINFO
2816 * The protocol parameters used to initialize and bound retransmission
2817 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2818 * and modify these parameters.
2819 * All parameters are time values, in milliseconds. A value of 0, when
2820 * modifying the parameters, indicates that the current value should not
2824 static int sctp_setsockopt_rtoinfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2826 struct sctp_rtoinfo rtoinfo
;
2827 struct sctp_association
*asoc
;
2829 if (optlen
!= sizeof (struct sctp_rtoinfo
))
2832 if (copy_from_user(&rtoinfo
, optval
, optlen
))
2835 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
2837 /* Set the values to the specific association */
2838 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
2842 if (rtoinfo
.srto_initial
!= 0)
2844 msecs_to_jiffies(rtoinfo
.srto_initial
);
2845 if (rtoinfo
.srto_max
!= 0)
2846 asoc
->rto_max
= msecs_to_jiffies(rtoinfo
.srto_max
);
2847 if (rtoinfo
.srto_min
!= 0)
2848 asoc
->rto_min
= msecs_to_jiffies(rtoinfo
.srto_min
);
2850 /* If there is no association or the association-id = 0
2851 * set the values to the endpoint.
2853 struct sctp_sock
*sp
= sctp_sk(sk
);
2855 if (rtoinfo
.srto_initial
!= 0)
2856 sp
->rtoinfo
.srto_initial
= rtoinfo
.srto_initial
;
2857 if (rtoinfo
.srto_max
!= 0)
2858 sp
->rtoinfo
.srto_max
= rtoinfo
.srto_max
;
2859 if (rtoinfo
.srto_min
!= 0)
2860 sp
->rtoinfo
.srto_min
= rtoinfo
.srto_min
;
2868 * 7.1.2 SCTP_ASSOCINFO
2870 * This option is used to tune the maximum retransmission attempts
2871 * of the association.
2872 * Returns an error if the new association retransmission value is
2873 * greater than the sum of the retransmission value of the peer.
2874 * See [SCTP] for more information.
2877 static int sctp_setsockopt_associnfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2880 struct sctp_assocparams assocparams
;
2881 struct sctp_association
*asoc
;
2883 if (optlen
!= sizeof(struct sctp_assocparams
))
2885 if (copy_from_user(&assocparams
, optval
, optlen
))
2888 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
2890 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
2893 /* Set the values to the specific association */
2895 if (assocparams
.sasoc_asocmaxrxt
!= 0) {
2898 struct sctp_transport
*peer_addr
;
2900 list_for_each_entry(peer_addr
, &asoc
->peer
.transport_addr_list
,
2902 path_sum
+= peer_addr
->pathmaxrxt
;
2906 /* Only validate asocmaxrxt if we have more than
2907 * one path/transport. We do this because path
2908 * retransmissions are only counted when we have more
2912 assocparams
.sasoc_asocmaxrxt
> path_sum
)
2915 asoc
->max_retrans
= assocparams
.sasoc_asocmaxrxt
;
2918 if (assocparams
.sasoc_cookie_life
!= 0) {
2919 asoc
->cookie_life
.tv_sec
=
2920 assocparams
.sasoc_cookie_life
/ 1000;
2921 asoc
->cookie_life
.tv_usec
=
2922 (assocparams
.sasoc_cookie_life
% 1000)
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("Use of int in maxseg socket option deprecated\n");
3002 pr_warn("Use struct sctp_assoc_value instead\n");
3003 if (copy_from_user(&val
, optval
, optlen
))
3005 params
.assoc_id
= 0;
3006 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
3007 if (copy_from_user(¶ms
, optval
, optlen
))
3009 val
= params
.assoc_value
;
3013 if ((val
!= 0) && ((val
< 8) || (val
> SCTP_MAX_CHUNK_LEN
)))
3016 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3017 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
3022 val
= asoc
->pathmtu
;
3023 val
-= sp
->pf
->af
->net_header_len
;
3024 val
-= sizeof(struct sctphdr
) +
3025 sizeof(struct sctp_data_chunk
);
3027 asoc
->user_frag
= val
;
3028 asoc
->frag_point
= sctp_frag_point(asoc
, asoc
->pathmtu
);
3030 sp
->user_frag
= val
;
3038 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3040 * Requests that the peer mark the enclosed address as the association
3041 * primary. The enclosed address must be one of the association's
3042 * locally bound addresses. The following structure is used to make a
3043 * set primary request:
3045 static int sctp_setsockopt_peer_primary_addr(struct sock
*sk
, char __user
*optval
,
3046 unsigned int optlen
)
3048 struct net
*net
= sock_net(sk
);
3049 struct sctp_sock
*sp
;
3050 struct sctp_association
*asoc
= NULL
;
3051 struct sctp_setpeerprim prim
;
3052 struct sctp_chunk
*chunk
;
3058 if (!net
->sctp
.addip_enable
)
3061 if (optlen
!= sizeof(struct sctp_setpeerprim
))
3064 if (copy_from_user(&prim
, optval
, optlen
))
3067 asoc
= sctp_id2assoc(sk
, prim
.sspp_assoc_id
);
3071 if (!asoc
->peer
.asconf_capable
)
3074 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_SET_PRIMARY
)
3077 if (!sctp_state(asoc
, ESTABLISHED
))
3080 af
= sctp_get_af_specific(prim
.sspp_addr
.ss_family
);
3084 if (!af
->addr_valid((union sctp_addr
*)&prim
.sspp_addr
, sp
, NULL
))
3085 return -EADDRNOTAVAIL
;
3087 if (!sctp_assoc_lookup_laddr(asoc
, (union sctp_addr
*)&prim
.sspp_addr
))
3088 return -EADDRNOTAVAIL
;
3090 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3091 chunk
= sctp_make_asconf_set_prim(asoc
,
3092 (union sctp_addr
*)&prim
.sspp_addr
);
3096 err
= sctp_send_asconf(asoc
, chunk
);
3098 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
3103 static int sctp_setsockopt_adaptation_layer(struct sock
*sk
, char __user
*optval
,
3104 unsigned int optlen
)
3106 struct sctp_setadaptation adaptation
;
3108 if (optlen
!= sizeof(struct sctp_setadaptation
))
3110 if (copy_from_user(&adaptation
, optval
, optlen
))
3113 sctp_sk(sk
)->adaptation_ind
= adaptation
.ssb_adaptation_ind
;
3119 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3121 * The context field in the sctp_sndrcvinfo structure is normally only
3122 * used when a failed message is retrieved holding the value that was
3123 * sent down on the actual send call. This option allows the setting of
3124 * a default context on an association basis that will be received on
3125 * reading messages from the peer. This is especially helpful in the
3126 * one-2-many model for an application to keep some reference to an
3127 * internal state machine that is processing messages on the
3128 * association. Note that the setting of this value only effects
3129 * received messages from the peer and does not effect the value that is
3130 * saved with outbound messages.
3132 static int sctp_setsockopt_context(struct sock
*sk
, char __user
*optval
,
3133 unsigned int optlen
)
3135 struct sctp_assoc_value params
;
3136 struct sctp_sock
*sp
;
3137 struct sctp_association
*asoc
;
3139 if (optlen
!= sizeof(struct sctp_assoc_value
))
3141 if (copy_from_user(¶ms
, optval
, optlen
))
3146 if (params
.assoc_id
!= 0) {
3147 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3150 asoc
->default_rcv_context
= params
.assoc_value
;
3152 sp
->default_rcv_context
= params
.assoc_value
;
3159 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3161 * This options will at a minimum specify if the implementation is doing
3162 * fragmented interleave. Fragmented interleave, for a one to many
3163 * socket, is when subsequent calls to receive a message may return
3164 * parts of messages from different associations. Some implementations
3165 * may allow you to turn this value on or off. If so, when turned off,
3166 * no fragment interleave will occur (which will cause a head of line
3167 * blocking amongst multiple associations sharing the same one to many
3168 * socket). When this option is turned on, then each receive call may
3169 * come from a different association (thus the user must receive data
3170 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3171 * association each receive belongs to.
3173 * This option takes a boolean value. A non-zero value indicates that
3174 * fragmented interleave is on. A value of zero indicates that
3175 * fragmented interleave is off.
3177 * Note that it is important that an implementation that allows this
3178 * option to be turned on, have it off by default. Otherwise an unaware
3179 * application using the one to many model may become confused and act
3182 static int sctp_setsockopt_fragment_interleave(struct sock
*sk
,
3183 char __user
*optval
,
3184 unsigned int optlen
)
3188 if (optlen
!= sizeof(int))
3190 if (get_user(val
, (int __user
*)optval
))
3193 sctp_sk(sk
)->frag_interleave
= (val
== 0) ? 0 : 1;
3199 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3200 * (SCTP_PARTIAL_DELIVERY_POINT)
3202 * This option will set or get the SCTP partial delivery point. This
3203 * point is the size of a message where the partial delivery API will be
3204 * invoked to help free up rwnd space for the peer. Setting this to a
3205 * lower value will cause partial deliveries to happen more often. The
3206 * calls argument is an integer that sets or gets the partial delivery
3207 * point. Note also that the call will fail if the user attempts to set
3208 * this value larger than the socket receive buffer size.
3210 * Note that any single message having a length smaller than or equal to
3211 * the SCTP partial delivery point will be delivered in one single read
3212 * call as long as the user provided buffer is large enough to hold the
3215 static int sctp_setsockopt_partial_delivery_point(struct sock
*sk
,
3216 char __user
*optval
,
3217 unsigned int optlen
)
3221 if (optlen
!= sizeof(u32
))
3223 if (get_user(val
, (int __user
*)optval
))
3226 /* Note: We double the receive buffer from what the user sets
3227 * it to be, also initial rwnd is based on rcvbuf/2.
3229 if (val
> (sk
->sk_rcvbuf
>> 1))
3232 sctp_sk(sk
)->pd_point
= val
;
3234 return 0; /* is this the right error code? */
3238 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3240 * This option will allow a user to change the maximum burst of packets
3241 * that can be emitted by this association. Note that the default value
3242 * is 4, and some implementations may restrict this setting so that it
3243 * can only be lowered.
3245 * NOTE: This text doesn't seem right. Do this on a socket basis with
3246 * future associations inheriting the socket value.
3248 static int sctp_setsockopt_maxburst(struct sock
*sk
,
3249 char __user
*optval
,
3250 unsigned int optlen
)
3252 struct sctp_assoc_value params
;
3253 struct sctp_sock
*sp
;
3254 struct sctp_association
*asoc
;
3258 if (optlen
== sizeof(int)) {
3259 pr_warn("Use of int in max_burst socket option deprecated\n");
3260 pr_warn("Use struct sctp_assoc_value instead\n");
3261 if (copy_from_user(&val
, optval
, optlen
))
3263 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
3264 if (copy_from_user(¶ms
, optval
, optlen
))
3266 val
= params
.assoc_value
;
3267 assoc_id
= params
.assoc_id
;
3273 if (assoc_id
!= 0) {
3274 asoc
= sctp_id2assoc(sk
, assoc_id
);
3277 asoc
->max_burst
= val
;
3279 sp
->max_burst
= val
;
3285 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3287 * This set option adds a chunk type that the user is requesting to be
3288 * received only in an authenticated way. Changes to the list of chunks
3289 * will only effect future associations on the socket.
3291 static int sctp_setsockopt_auth_chunk(struct sock
*sk
,
3292 char __user
*optval
,
3293 unsigned int optlen
)
3295 struct net
*net
= sock_net(sk
);
3296 struct sctp_authchunk val
;
3298 if (!net
->sctp
.auth_enable
)
3301 if (optlen
!= sizeof(struct sctp_authchunk
))
3303 if (copy_from_user(&val
, optval
, optlen
))
3306 switch (val
.sauth_chunk
) {
3308 case SCTP_CID_INIT_ACK
:
3309 case SCTP_CID_SHUTDOWN_COMPLETE
:
3314 /* add this chunk id to the endpoint */
3315 return sctp_auth_ep_add_chunkid(sctp_sk(sk
)->ep
, val
.sauth_chunk
);
3319 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3321 * This option gets or sets the list of HMAC algorithms that the local
3322 * endpoint requires the peer to use.
3324 static int sctp_setsockopt_hmac_ident(struct sock
*sk
,
3325 char __user
*optval
,
3326 unsigned int optlen
)
3328 struct net
*net
= sock_net(sk
);
3329 struct sctp_hmacalgo
*hmacs
;
3333 if (!net
->sctp
.auth_enable
)
3336 if (optlen
< sizeof(struct sctp_hmacalgo
))
3339 hmacs
= memdup_user(optval
, optlen
);
3341 return PTR_ERR(hmacs
);
3343 idents
= hmacs
->shmac_num_idents
;
3344 if (idents
== 0 || idents
> SCTP_AUTH_NUM_HMACS
||
3345 (idents
* sizeof(u16
)) > (optlen
- sizeof(struct sctp_hmacalgo
))) {
3350 err
= sctp_auth_ep_set_hmacs(sctp_sk(sk
)->ep
, hmacs
);
3357 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3359 * This option will set a shared secret key which is used to build an
3360 * association shared key.
3362 static int sctp_setsockopt_auth_key(struct sock
*sk
,
3363 char __user
*optval
,
3364 unsigned int optlen
)
3366 struct net
*net
= sock_net(sk
);
3367 struct sctp_authkey
*authkey
;
3368 struct sctp_association
*asoc
;
3371 if (!net
->sctp
.auth_enable
)
3374 if (optlen
<= sizeof(struct sctp_authkey
))
3377 authkey
= memdup_user(optval
, optlen
);
3378 if (IS_ERR(authkey
))
3379 return PTR_ERR(authkey
);
3381 if (authkey
->sca_keylength
> optlen
- sizeof(struct sctp_authkey
)) {
3386 asoc
= sctp_id2assoc(sk
, authkey
->sca_assoc_id
);
3387 if (!asoc
&& authkey
->sca_assoc_id
&& sctp_style(sk
, UDP
)) {
3392 ret
= sctp_auth_set_key(sctp_sk(sk
)->ep
, asoc
, authkey
);
3399 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3401 * This option will get or set the active shared key to be used to build
3402 * the association shared key.
3404 static int sctp_setsockopt_active_key(struct sock
*sk
,
3405 char __user
*optval
,
3406 unsigned int optlen
)
3408 struct net
*net
= sock_net(sk
);
3409 struct sctp_authkeyid val
;
3410 struct sctp_association
*asoc
;
3412 if (!net
->sctp
.auth_enable
)
3415 if (optlen
!= sizeof(struct sctp_authkeyid
))
3417 if (copy_from_user(&val
, optval
, optlen
))
3420 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3421 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3424 return sctp_auth_set_active_key(sctp_sk(sk
)->ep
, asoc
,
3425 val
.scact_keynumber
);
3429 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3431 * This set option will delete a shared secret key from use.
3433 static int sctp_setsockopt_del_key(struct sock
*sk
,
3434 char __user
*optval
,
3435 unsigned int optlen
)
3437 struct net
*net
= sock_net(sk
);
3438 struct sctp_authkeyid val
;
3439 struct sctp_association
*asoc
;
3441 if (!net
->sctp
.auth_enable
)
3444 if (optlen
!= sizeof(struct sctp_authkeyid
))
3446 if (copy_from_user(&val
, optval
, optlen
))
3449 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3450 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3453 return sctp_auth_del_key_id(sctp_sk(sk
)->ep
, asoc
,
3454 val
.scact_keynumber
);
3459 * 8.1.23 SCTP_AUTO_ASCONF
3461 * This option will enable or disable the use of the automatic generation of
3462 * ASCONF chunks to add and delete addresses to an existing association. Note
3463 * that this option has two caveats namely: a) it only affects sockets that
3464 * are bound to all addresses available to the SCTP stack, and b) the system
3465 * administrator may have an overriding control that turns the ASCONF feature
3466 * off no matter what setting the socket option may have.
3467 * This option expects an integer boolean flag, where a non-zero value turns on
3468 * the option, and a zero value turns off the option.
3469 * Note. In this implementation, socket operation overrides default parameter
3470 * being set by sysctl as well as FreeBSD implementation
3472 static int sctp_setsockopt_auto_asconf(struct sock
*sk
, char __user
*optval
,
3473 unsigned int optlen
)
3476 struct sctp_sock
*sp
= sctp_sk(sk
);
3478 if (optlen
< sizeof(int))
3480 if (get_user(val
, (int __user
*)optval
))
3482 if (!sctp_is_ep_boundall(sk
) && val
)
3484 if ((val
&& sp
->do_auto_asconf
) || (!val
&& !sp
->do_auto_asconf
))
3487 if (val
== 0 && sp
->do_auto_asconf
) {
3488 list_del(&sp
->auto_asconf_list
);
3489 sp
->do_auto_asconf
= 0;
3490 } else if (val
&& !sp
->do_auto_asconf
) {
3491 list_add_tail(&sp
->auto_asconf_list
,
3492 &sock_net(sk
)->sctp
.auto_asconf_splist
);
3493 sp
->do_auto_asconf
= 1;
3500 * SCTP_PEER_ADDR_THLDS
3502 * This option allows us to alter the partially failed threshold for one or all
3503 * transports in an association. See Section 6.1 of:
3504 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3506 static int sctp_setsockopt_paddr_thresholds(struct sock
*sk
,
3507 char __user
*optval
,
3508 unsigned int optlen
)
3510 struct sctp_paddrthlds val
;
3511 struct sctp_transport
*trans
;
3512 struct sctp_association
*asoc
;
3514 if (optlen
< sizeof(struct sctp_paddrthlds
))
3516 if (copy_from_user(&val
, (struct sctp_paddrthlds __user
*)optval
,
3517 sizeof(struct sctp_paddrthlds
)))
3521 if (sctp_is_any(sk
, (const union sctp_addr
*)&val
.spt_address
)) {
3522 asoc
= sctp_id2assoc(sk
, val
.spt_assoc_id
);
3525 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
3527 if (val
.spt_pathmaxrxt
)
3528 trans
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3529 trans
->pf_retrans
= val
.spt_pathpfthld
;
3532 if (val
.spt_pathmaxrxt
)
3533 asoc
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3534 asoc
->pf_retrans
= val
.spt_pathpfthld
;
3536 trans
= sctp_addr_id2transport(sk
, &val
.spt_address
,
3541 if (val
.spt_pathmaxrxt
)
3542 trans
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3543 trans
->pf_retrans
= val
.spt_pathpfthld
;
3549 /* API 6.2 setsockopt(), getsockopt()
3551 * Applications use setsockopt() and getsockopt() to set or retrieve
3552 * socket options. Socket options are used to change the default
3553 * behavior of sockets calls. They are described in Section 7.
3557 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3558 * int __user *optlen);
3559 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3562 * sd - the socket descript.
3563 * level - set to IPPROTO_SCTP for all SCTP options.
3564 * optname - the option name.
3565 * optval - the buffer to store the value of the option.
3566 * optlen - the size of the buffer.
3568 SCTP_STATIC
int sctp_setsockopt(struct sock
*sk
, int level
, int optname
,
3569 char __user
*optval
, unsigned int optlen
)
3573 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3576 /* I can hardly begin to describe how wrong this is. This is
3577 * so broken as to be worse than useless. The API draft
3578 * REALLY is NOT helpful here... I am not convinced that the
3579 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3580 * are at all well-founded.
3582 if (level
!= SOL_SCTP
) {
3583 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
3584 retval
= af
->setsockopt(sk
, level
, optname
, optval
, optlen
);
3591 case SCTP_SOCKOPT_BINDX_ADD
:
3592 /* 'optlen' is the size of the addresses buffer. */
3593 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3594 optlen
, SCTP_BINDX_ADD_ADDR
);
3597 case SCTP_SOCKOPT_BINDX_REM
:
3598 /* 'optlen' is the size of the addresses buffer. */
3599 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3600 optlen
, SCTP_BINDX_REM_ADDR
);
3603 case SCTP_SOCKOPT_CONNECTX_OLD
:
3604 /* 'optlen' is the size of the addresses buffer. */
3605 retval
= sctp_setsockopt_connectx_old(sk
,
3606 (struct sockaddr __user
*)optval
,
3610 case SCTP_SOCKOPT_CONNECTX
:
3611 /* 'optlen' is the size of the addresses buffer. */
3612 retval
= sctp_setsockopt_connectx(sk
,
3613 (struct sockaddr __user
*)optval
,
3617 case SCTP_DISABLE_FRAGMENTS
:
3618 retval
= sctp_setsockopt_disable_fragments(sk
, optval
, optlen
);
3622 retval
= sctp_setsockopt_events(sk
, optval
, optlen
);
3625 case SCTP_AUTOCLOSE
:
3626 retval
= sctp_setsockopt_autoclose(sk
, optval
, optlen
);
3629 case SCTP_PEER_ADDR_PARAMS
:
3630 retval
= sctp_setsockopt_peer_addr_params(sk
, optval
, optlen
);
3633 case SCTP_DELAYED_SACK
:
3634 retval
= sctp_setsockopt_delayed_ack(sk
, optval
, optlen
);
3636 case SCTP_PARTIAL_DELIVERY_POINT
:
3637 retval
= sctp_setsockopt_partial_delivery_point(sk
, optval
, optlen
);
3641 retval
= sctp_setsockopt_initmsg(sk
, optval
, optlen
);
3643 case SCTP_DEFAULT_SEND_PARAM
:
3644 retval
= sctp_setsockopt_default_send_param(sk
, optval
,
3647 case SCTP_PRIMARY_ADDR
:
3648 retval
= sctp_setsockopt_primary_addr(sk
, optval
, optlen
);
3650 case SCTP_SET_PEER_PRIMARY_ADDR
:
3651 retval
= sctp_setsockopt_peer_primary_addr(sk
, optval
, optlen
);
3654 retval
= sctp_setsockopt_nodelay(sk
, optval
, optlen
);
3657 retval
= sctp_setsockopt_rtoinfo(sk
, optval
, optlen
);
3659 case SCTP_ASSOCINFO
:
3660 retval
= sctp_setsockopt_associnfo(sk
, optval
, optlen
);
3662 case SCTP_I_WANT_MAPPED_V4_ADDR
:
3663 retval
= sctp_setsockopt_mappedv4(sk
, optval
, optlen
);
3666 retval
= sctp_setsockopt_maxseg(sk
, optval
, optlen
);
3668 case SCTP_ADAPTATION_LAYER
:
3669 retval
= sctp_setsockopt_adaptation_layer(sk
, optval
, optlen
);
3672 retval
= sctp_setsockopt_context(sk
, optval
, optlen
);
3674 case SCTP_FRAGMENT_INTERLEAVE
:
3675 retval
= sctp_setsockopt_fragment_interleave(sk
, optval
, optlen
);
3677 case SCTP_MAX_BURST
:
3678 retval
= sctp_setsockopt_maxburst(sk
, optval
, optlen
);
3680 case SCTP_AUTH_CHUNK
:
3681 retval
= sctp_setsockopt_auth_chunk(sk
, optval
, optlen
);
3683 case SCTP_HMAC_IDENT
:
3684 retval
= sctp_setsockopt_hmac_ident(sk
, optval
, optlen
);
3687 retval
= sctp_setsockopt_auth_key(sk
, optval
, optlen
);
3689 case SCTP_AUTH_ACTIVE_KEY
:
3690 retval
= sctp_setsockopt_active_key(sk
, optval
, optlen
);
3692 case SCTP_AUTH_DELETE_KEY
:
3693 retval
= sctp_setsockopt_del_key(sk
, optval
, optlen
);
3695 case SCTP_AUTO_ASCONF
:
3696 retval
= sctp_setsockopt_auto_asconf(sk
, optval
, optlen
);
3698 case SCTP_PEER_ADDR_THLDS
:
3699 retval
= sctp_setsockopt_paddr_thresholds(sk
, optval
, optlen
);
3702 retval
= -ENOPROTOOPT
;
3706 sctp_release_sock(sk
);
3712 /* API 3.1.6 connect() - UDP Style Syntax
3714 * An application may use the connect() call in the UDP model to initiate an
3715 * association without sending data.
3719 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3721 * sd: the socket descriptor to have a new association added to.
3723 * nam: the address structure (either struct sockaddr_in or struct
3724 * sockaddr_in6 defined in RFC2553 [7]).
3726 * len: the size of the address.
3728 SCTP_STATIC
int sctp_connect(struct sock
*sk
, struct sockaddr
*addr
,
3736 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3737 __func__
, sk
, addr
, addr_len
);
3739 /* Validate addr_len before calling common connect/connectx routine. */
3740 af
= sctp_get_af_specific(addr
->sa_family
);
3741 if (!af
|| addr_len
< af
->sockaddr_len
) {
3744 /* Pass correct addr len to common routine (so it knows there
3745 * is only one address being passed.
3747 err
= __sctp_connect(sk
, addr
, af
->sockaddr_len
, NULL
);
3750 sctp_release_sock(sk
);
3754 /* FIXME: Write comments. */
3755 SCTP_STATIC
int sctp_disconnect(struct sock
*sk
, int flags
)
3757 return -EOPNOTSUPP
; /* STUB */
3760 /* 4.1.4 accept() - TCP Style Syntax
3762 * Applications use accept() call to remove an established SCTP
3763 * association from the accept queue of the endpoint. A new socket
3764 * descriptor will be returned from accept() to represent the newly
3765 * formed association.
3767 SCTP_STATIC
struct sock
*sctp_accept(struct sock
*sk
, int flags
, int *err
)
3769 struct sctp_sock
*sp
;
3770 struct sctp_endpoint
*ep
;
3771 struct sock
*newsk
= NULL
;
3772 struct sctp_association
*asoc
;
3781 if (!sctp_style(sk
, TCP
)) {
3782 error
= -EOPNOTSUPP
;
3786 if (!sctp_sstate(sk
, LISTENING
)) {
3791 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
3793 error
= sctp_wait_for_accept(sk
, timeo
);
3797 /* We treat the list of associations on the endpoint as the accept
3798 * queue and pick the first association on the list.
3800 asoc
= list_entry(ep
->asocs
.next
, struct sctp_association
, asocs
);
3802 newsk
= sp
->pf
->create_accept_sk(sk
, asoc
);
3808 /* Populate the fields of the newsk from the oldsk and migrate the
3809 * asoc to the newsk.
3811 sctp_sock_migrate(sk
, newsk
, asoc
, SCTP_SOCKET_TCP
);
3814 sctp_release_sock(sk
);
3819 /* The SCTP ioctl handler. */
3820 SCTP_STATIC
int sctp_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
3827 * SEQPACKET-style sockets in LISTENING state are valid, for
3828 * SCTP, so only discard TCP-style sockets in LISTENING state.
3830 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
3835 struct sk_buff
*skb
;
3836 unsigned int amount
= 0;
3838 skb
= skb_peek(&sk
->sk_receive_queue
);
3841 * We will only return the amount of this packet since
3842 * that is all that will be read.
3846 rc
= put_user(amount
, (int __user
*)arg
);
3854 sctp_release_sock(sk
);
3858 /* This is the function which gets called during socket creation to
3859 * initialized the SCTP-specific portion of the sock.
3860 * The sock structure should already be zero-filled memory.
3862 SCTP_STATIC
int sctp_init_sock(struct sock
*sk
)
3864 struct net
*net
= sock_net(sk
);
3865 struct sctp_endpoint
*ep
;
3866 struct sctp_sock
*sp
;
3868 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk
);
3872 /* Initialize the SCTP per socket area. */
3873 switch (sk
->sk_type
) {
3874 case SOCK_SEQPACKET
:
3875 sp
->type
= SCTP_SOCKET_UDP
;
3878 sp
->type
= SCTP_SOCKET_TCP
;
3881 return -ESOCKTNOSUPPORT
;
3884 /* Initialize default send parameters. These parameters can be
3885 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3887 sp
->default_stream
= 0;
3888 sp
->default_ppid
= 0;
3889 sp
->default_flags
= 0;
3890 sp
->default_context
= 0;
3891 sp
->default_timetolive
= 0;
3893 sp
->default_rcv_context
= 0;
3894 sp
->max_burst
= net
->sctp
.max_burst
;
3896 sp
->sctp_hmac_alg
= net
->sctp
.sctp_hmac_alg
;
3898 /* Initialize default setup parameters. These parameters
3899 * can be modified with the SCTP_INITMSG socket option or
3900 * overridden by the SCTP_INIT CMSG.
3902 sp
->initmsg
.sinit_num_ostreams
= sctp_max_outstreams
;
3903 sp
->initmsg
.sinit_max_instreams
= sctp_max_instreams
;
3904 sp
->initmsg
.sinit_max_attempts
= net
->sctp
.max_retrans_init
;
3905 sp
->initmsg
.sinit_max_init_timeo
= net
->sctp
.rto_max
;
3907 /* Initialize default RTO related parameters. These parameters can
3908 * be modified for with the SCTP_RTOINFO socket option.
3910 sp
->rtoinfo
.srto_initial
= net
->sctp
.rto_initial
;
3911 sp
->rtoinfo
.srto_max
= net
->sctp
.rto_max
;
3912 sp
->rtoinfo
.srto_min
= net
->sctp
.rto_min
;
3914 /* Initialize default association related parameters. These parameters
3915 * can be modified with the SCTP_ASSOCINFO socket option.
3917 sp
->assocparams
.sasoc_asocmaxrxt
= net
->sctp
.max_retrans_association
;
3918 sp
->assocparams
.sasoc_number_peer_destinations
= 0;
3919 sp
->assocparams
.sasoc_peer_rwnd
= 0;
3920 sp
->assocparams
.sasoc_local_rwnd
= 0;
3921 sp
->assocparams
.sasoc_cookie_life
= net
->sctp
.valid_cookie_life
;
3923 /* Initialize default event subscriptions. By default, all the
3926 memset(&sp
->subscribe
, 0, sizeof(struct sctp_event_subscribe
));
3928 /* Default Peer Address Parameters. These defaults can
3929 * be modified via SCTP_PEER_ADDR_PARAMS
3931 sp
->hbinterval
= net
->sctp
.hb_interval
;
3932 sp
->pathmaxrxt
= net
->sctp
.max_retrans_path
;
3933 sp
->pathmtu
= 0; // allow default discovery
3934 sp
->sackdelay
= net
->sctp
.sack_timeout
;
3936 sp
->param_flags
= SPP_HB_ENABLE
|
3938 SPP_SACKDELAY_ENABLE
;
3940 /* If enabled no SCTP message fragmentation will be performed.
3941 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3943 sp
->disable_fragments
= 0;
3945 /* Enable Nagle algorithm by default. */
3948 /* Enable by default. */
3951 /* Auto-close idle associations after the configured
3952 * number of seconds. A value of 0 disables this
3953 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3954 * for UDP-style sockets only.
3958 /* User specified fragmentation limit. */
3961 sp
->adaptation_ind
= 0;
3963 sp
->pf
= sctp_get_pf_specific(sk
->sk_family
);
3965 /* Control variables for partial data delivery. */
3966 atomic_set(&sp
->pd_mode
, 0);
3967 skb_queue_head_init(&sp
->pd_lobby
);
3968 sp
->frag_interleave
= 0;
3970 /* Create a per socket endpoint structure. Even if we
3971 * change the data structure relationships, this may still
3972 * be useful for storing pre-connect address information.
3974 ep
= sctp_endpoint_new(sk
, GFP_KERNEL
);
3981 SCTP_DBG_OBJCNT_INC(sock
);
3984 percpu_counter_inc(&sctp_sockets_allocated
);
3985 sock_prot_inuse_add(net
, sk
->sk_prot
, 1);
3986 if (net
->sctp
.default_auto_asconf
) {
3987 list_add_tail(&sp
->auto_asconf_list
,
3988 &net
->sctp
.auto_asconf_splist
);
3989 sp
->do_auto_asconf
= 1;
3991 sp
->do_auto_asconf
= 0;
3997 /* Cleanup any SCTP per socket resources. */
3998 SCTP_STATIC
void sctp_destroy_sock(struct sock
*sk
)
4000 struct sctp_sock
*sp
;
4002 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk
);
4004 /* Release our hold on the endpoint. */
4006 if (sp
->do_auto_asconf
) {
4007 sp
->do_auto_asconf
= 0;
4008 list_del(&sp
->auto_asconf_list
);
4010 sctp_endpoint_free(sp
->ep
);
4012 percpu_counter_dec(&sctp_sockets_allocated
);
4013 sock_prot_inuse_add(sock_net(sk
), sk
->sk_prot
, -1);
4017 /* API 4.1.7 shutdown() - TCP Style Syntax
4018 * int shutdown(int socket, int how);
4020 * sd - the socket descriptor of the association to be closed.
4021 * how - Specifies the type of shutdown. The values are
4024 * Disables further receive operations. No SCTP
4025 * protocol action is taken.
4027 * Disables further send operations, and initiates
4028 * the SCTP shutdown sequence.
4030 * Disables further send and receive operations
4031 * and initiates the SCTP shutdown sequence.
4033 SCTP_STATIC
void sctp_shutdown(struct sock
*sk
, int how
)
4035 struct net
*net
= sock_net(sk
);
4036 struct sctp_endpoint
*ep
;
4037 struct sctp_association
*asoc
;
4039 if (!sctp_style(sk
, TCP
))
4042 if (how
& SEND_SHUTDOWN
) {
4043 ep
= sctp_sk(sk
)->ep
;
4044 if (!list_empty(&ep
->asocs
)) {
4045 asoc
= list_entry(ep
->asocs
.next
,
4046 struct sctp_association
, asocs
);
4047 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
4052 /* 7.2.1 Association Status (SCTP_STATUS)
4054 * Applications can retrieve current status information about an
4055 * association, including association state, peer receiver window size,
4056 * number of unacked data chunks, and number of data chunks pending
4057 * receipt. This information is read-only.
4059 static int sctp_getsockopt_sctp_status(struct sock
*sk
, int len
,
4060 char __user
*optval
,
4063 struct sctp_status status
;
4064 struct sctp_association
*asoc
= NULL
;
4065 struct sctp_transport
*transport
;
4066 sctp_assoc_t associd
;
4069 if (len
< sizeof(status
)) {
4074 len
= sizeof(status
);
4075 if (copy_from_user(&status
, optval
, len
)) {
4080 associd
= status
.sstat_assoc_id
;
4081 asoc
= sctp_id2assoc(sk
, associd
);
4087 transport
= asoc
->peer
.primary_path
;
4089 status
.sstat_assoc_id
= sctp_assoc2id(asoc
);
4090 status
.sstat_state
= asoc
->state
;
4091 status
.sstat_rwnd
= asoc
->peer
.rwnd
;
4092 status
.sstat_unackdata
= asoc
->unack_data
;
4094 status
.sstat_penddata
= sctp_tsnmap_pending(&asoc
->peer
.tsn_map
);
4095 status
.sstat_instrms
= asoc
->c
.sinit_max_instreams
;
4096 status
.sstat_outstrms
= asoc
->c
.sinit_num_ostreams
;
4097 status
.sstat_fragmentation_point
= asoc
->frag_point
;
4098 status
.sstat_primary
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
4099 memcpy(&status
.sstat_primary
.spinfo_address
, &transport
->ipaddr
,
4100 transport
->af_specific
->sockaddr_len
);
4101 /* Map ipv4 address into v4-mapped-on-v6 address. */
4102 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
4103 (union sctp_addr
*)&status
.sstat_primary
.spinfo_address
);
4104 status
.sstat_primary
.spinfo_state
= transport
->state
;
4105 status
.sstat_primary
.spinfo_cwnd
= transport
->cwnd
;
4106 status
.sstat_primary
.spinfo_srtt
= transport
->srtt
;
4107 status
.sstat_primary
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
4108 status
.sstat_primary
.spinfo_mtu
= transport
->pathmtu
;
4110 if (status
.sstat_primary
.spinfo_state
== SCTP_UNKNOWN
)
4111 status
.sstat_primary
.spinfo_state
= SCTP_ACTIVE
;
4113 if (put_user(len
, optlen
)) {
4118 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
4119 len
, status
.sstat_state
, status
.sstat_rwnd
,
4120 status
.sstat_assoc_id
);
4122 if (copy_to_user(optval
, &status
, len
)) {
4132 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4134 * Applications can retrieve information about a specific peer address
4135 * of an association, including its reachability state, congestion
4136 * window, and retransmission timer values. This information is
4139 static int sctp_getsockopt_peer_addr_info(struct sock
*sk
, int len
,
4140 char __user
*optval
,
4143 struct sctp_paddrinfo pinfo
;
4144 struct sctp_transport
*transport
;
4147 if (len
< sizeof(pinfo
)) {
4152 len
= sizeof(pinfo
);
4153 if (copy_from_user(&pinfo
, optval
, len
)) {
4158 transport
= sctp_addr_id2transport(sk
, &pinfo
.spinfo_address
,
4159 pinfo
.spinfo_assoc_id
);
4163 pinfo
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
4164 pinfo
.spinfo_state
= transport
->state
;
4165 pinfo
.spinfo_cwnd
= transport
->cwnd
;
4166 pinfo
.spinfo_srtt
= transport
->srtt
;
4167 pinfo
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
4168 pinfo
.spinfo_mtu
= transport
->pathmtu
;
4170 if (pinfo
.spinfo_state
== SCTP_UNKNOWN
)
4171 pinfo
.spinfo_state
= SCTP_ACTIVE
;
4173 if (put_user(len
, optlen
)) {
4178 if (copy_to_user(optval
, &pinfo
, len
)) {
4187 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4189 * This option is a on/off flag. If enabled no SCTP message
4190 * fragmentation will be performed. Instead if a message being sent
4191 * exceeds the current PMTU size, the message will NOT be sent and
4192 * instead a error will be indicated to the user.
4194 static int sctp_getsockopt_disable_fragments(struct sock
*sk
, int len
,
4195 char __user
*optval
, int __user
*optlen
)
4199 if (len
< sizeof(int))
4203 val
= (sctp_sk(sk
)->disable_fragments
== 1);
4204 if (put_user(len
, optlen
))
4206 if (copy_to_user(optval
, &val
, len
))
4211 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4213 * This socket option is used to specify various notifications and
4214 * ancillary data the user wishes to receive.
4216 static int sctp_getsockopt_events(struct sock
*sk
, int len
, char __user
*optval
,
4221 if (len
> sizeof(struct sctp_event_subscribe
))
4222 len
= sizeof(struct sctp_event_subscribe
);
4223 if (put_user(len
, optlen
))
4225 if (copy_to_user(optval
, &sctp_sk(sk
)->subscribe
, len
))
4230 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4232 * This socket option is applicable to the UDP-style socket only. When
4233 * set it will cause associations that are idle for more than the
4234 * specified number of seconds to automatically close. An association
4235 * being idle is defined an association that has NOT sent or received
4236 * user data. The special value of '0' indicates that no automatic
4237 * close of any associations should be performed. The option expects an
4238 * integer defining the number of seconds of idle time before an
4239 * association is closed.
4241 static int sctp_getsockopt_autoclose(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4243 /* Applicable to UDP-style socket only */
4244 if (sctp_style(sk
, TCP
))
4246 if (len
< sizeof(int))
4249 if (put_user(len
, optlen
))
4251 if (copy_to_user(optval
, &sctp_sk(sk
)->autoclose
, sizeof(int)))
4256 /* Helper routine to branch off an association to a new socket. */
4257 int sctp_do_peeloff(struct sock
*sk
, sctp_assoc_t id
, struct socket
**sockp
)
4259 struct sctp_association
*asoc
= sctp_id2assoc(sk
, id
);
4260 struct socket
*sock
;
4267 /* An association cannot be branched off from an already peeled-off
4268 * socket, nor is this supported for tcp style sockets.
4270 if (!sctp_style(sk
, UDP
))
4273 /* Create a new socket. */
4274 err
= sock_create(sk
->sk_family
, SOCK_SEQPACKET
, IPPROTO_SCTP
, &sock
);
4278 sctp_copy_sock(sock
->sk
, sk
, asoc
);
4280 /* Make peeled-off sockets more like 1-1 accepted sockets.
4281 * Set the daddr and initialize id to something more random
4283 af
= sctp_get_af_specific(asoc
->peer
.primary_addr
.sa
.sa_family
);
4284 af
->to_sk_daddr(&asoc
->peer
.primary_addr
, sk
);
4286 /* Populate the fields of the newsk from the oldsk and migrate the
4287 * asoc to the newsk.
4289 sctp_sock_migrate(sk
, sock
->sk
, asoc
, SCTP_SOCKET_UDP_HIGH_BANDWIDTH
);
4295 EXPORT_SYMBOL(sctp_do_peeloff
);
4297 static int sctp_getsockopt_peeloff(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4299 sctp_peeloff_arg_t peeloff
;
4300 struct socket
*newsock
;
4301 struct file
*newfile
;
4304 if (len
< sizeof(sctp_peeloff_arg_t
))
4306 len
= sizeof(sctp_peeloff_arg_t
);
4307 if (copy_from_user(&peeloff
, optval
, len
))
4310 retval
= sctp_do_peeloff(sk
, peeloff
.associd
, &newsock
);
4314 /* Map the socket to an unused fd that can be returned to the user. */
4315 retval
= get_unused_fd();
4317 sock_release(newsock
);
4321 newfile
= sock_alloc_file(newsock
, 0, NULL
);
4322 if (unlikely(IS_ERR(newfile
))) {
4323 put_unused_fd(retval
);
4324 sock_release(newsock
);
4325 return PTR_ERR(newfile
);
4328 SCTP_DEBUG_PRINTK("%s: sk: %p newsk: %p sd: %d\n",
4329 __func__
, sk
, newsock
->sk
, retval
);
4331 /* Return the fd mapped to the new socket. */
4332 if (put_user(len
, optlen
)) {
4334 put_unused_fd(retval
);
4337 peeloff
.sd
= retval
;
4338 if (copy_to_user(optval
, &peeloff
, len
)) {
4340 put_unused_fd(retval
);
4343 fd_install(retval
, newfile
);
4348 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4350 * Applications can enable or disable heartbeats for any peer address of
4351 * an association, modify an address's heartbeat interval, force a
4352 * heartbeat to be sent immediately, and adjust the address's maximum
4353 * number of retransmissions sent before an address is considered
4354 * unreachable. The following structure is used to access and modify an
4355 * address's parameters:
4357 * struct sctp_paddrparams {
4358 * sctp_assoc_t spp_assoc_id;
4359 * struct sockaddr_storage spp_address;
4360 * uint32_t spp_hbinterval;
4361 * uint16_t spp_pathmaxrxt;
4362 * uint32_t spp_pathmtu;
4363 * uint32_t spp_sackdelay;
4364 * uint32_t spp_flags;
4367 * spp_assoc_id - (one-to-many style socket) This is filled in the
4368 * application, and identifies the association for
4370 * spp_address - This specifies which address is of interest.
4371 * spp_hbinterval - This contains the value of the heartbeat interval,
4372 * in milliseconds. If a value of zero
4373 * is present in this field then no changes are to
4374 * be made to this parameter.
4375 * spp_pathmaxrxt - This contains the maximum number of
4376 * retransmissions before this address shall be
4377 * considered unreachable. If a value of zero
4378 * is present in this field then no changes are to
4379 * be made to this parameter.
4380 * spp_pathmtu - When Path MTU discovery is disabled the value
4381 * specified here will be the "fixed" path mtu.
4382 * Note that if the spp_address field is empty
4383 * then all associations on this address will
4384 * have this fixed path mtu set upon them.
4386 * spp_sackdelay - When delayed sack is enabled, this value specifies
4387 * the number of milliseconds that sacks will be delayed
4388 * for. This value will apply to all addresses of an
4389 * association if the spp_address field is empty. Note
4390 * also, that if delayed sack is enabled and this
4391 * value is set to 0, no change is made to the last
4392 * recorded delayed sack timer value.
4394 * spp_flags - These flags are used to control various features
4395 * on an association. The flag field may contain
4396 * zero or more of the following options.
4398 * SPP_HB_ENABLE - Enable heartbeats on the
4399 * specified address. Note that if the address
4400 * field is empty all addresses for the association
4401 * have heartbeats enabled upon them.
4403 * SPP_HB_DISABLE - Disable heartbeats on the
4404 * speicifed address. Note that if the address
4405 * field is empty all addresses for the association
4406 * will have their heartbeats disabled. Note also
4407 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4408 * mutually exclusive, only one of these two should
4409 * be specified. Enabling both fields will have
4410 * undetermined results.
4412 * SPP_HB_DEMAND - Request a user initiated heartbeat
4413 * to be made immediately.
4415 * SPP_PMTUD_ENABLE - This field will enable PMTU
4416 * discovery upon the specified address. Note that
4417 * if the address feild is empty then all addresses
4418 * on the association are effected.
4420 * SPP_PMTUD_DISABLE - This field will disable PMTU
4421 * discovery upon the specified address. Note that
4422 * if the address feild is empty then all addresses
4423 * on the association are effected. Not also that
4424 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4425 * exclusive. Enabling both will have undetermined
4428 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4429 * on delayed sack. The time specified in spp_sackdelay
4430 * is used to specify the sack delay for this address. Note
4431 * that if spp_address is empty then all addresses will
4432 * enable delayed sack and take on the sack delay
4433 * value specified in spp_sackdelay.
4434 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4435 * off delayed sack. If the spp_address field is blank then
4436 * delayed sack is disabled for the entire association. Note
4437 * also that this field is mutually exclusive to
4438 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4441 static int sctp_getsockopt_peer_addr_params(struct sock
*sk
, int len
,
4442 char __user
*optval
, int __user
*optlen
)
4444 struct sctp_paddrparams params
;
4445 struct sctp_transport
*trans
= NULL
;
4446 struct sctp_association
*asoc
= NULL
;
4447 struct sctp_sock
*sp
= sctp_sk(sk
);
4449 if (len
< sizeof(struct sctp_paddrparams
))
4451 len
= sizeof(struct sctp_paddrparams
);
4452 if (copy_from_user(¶ms
, optval
, len
))
4455 /* If an address other than INADDR_ANY is specified, and
4456 * no transport is found, then the request is invalid.
4458 if (!sctp_is_any(sk
, ( union sctp_addr
*)¶ms
.spp_address
)) {
4459 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
4460 params
.spp_assoc_id
);
4462 SCTP_DEBUG_PRINTK("Failed no transport\n");
4467 /* Get association, if assoc_id != 0 and the socket is a one
4468 * to many style socket, and an association was not found, then
4469 * the id was invalid.
4471 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
4472 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
)) {
4473 SCTP_DEBUG_PRINTK("Failed no association\n");
4478 /* Fetch transport values. */
4479 params
.spp_hbinterval
= jiffies_to_msecs(trans
->hbinterval
);
4480 params
.spp_pathmtu
= trans
->pathmtu
;
4481 params
.spp_pathmaxrxt
= trans
->pathmaxrxt
;
4482 params
.spp_sackdelay
= jiffies_to_msecs(trans
->sackdelay
);
4484 /*draft-11 doesn't say what to return in spp_flags*/
4485 params
.spp_flags
= trans
->param_flags
;
4487 /* Fetch association values. */
4488 params
.spp_hbinterval
= jiffies_to_msecs(asoc
->hbinterval
);
4489 params
.spp_pathmtu
= asoc
->pathmtu
;
4490 params
.spp_pathmaxrxt
= asoc
->pathmaxrxt
;
4491 params
.spp_sackdelay
= jiffies_to_msecs(asoc
->sackdelay
);
4493 /*draft-11 doesn't say what to return in spp_flags*/
4494 params
.spp_flags
= asoc
->param_flags
;
4496 /* Fetch socket values. */
4497 params
.spp_hbinterval
= sp
->hbinterval
;
4498 params
.spp_pathmtu
= sp
->pathmtu
;
4499 params
.spp_sackdelay
= sp
->sackdelay
;
4500 params
.spp_pathmaxrxt
= sp
->pathmaxrxt
;
4502 /*draft-11 doesn't say what to return in spp_flags*/
4503 params
.spp_flags
= sp
->param_flags
;
4506 if (copy_to_user(optval
, ¶ms
, len
))
4509 if (put_user(len
, optlen
))
4516 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4518 * This option will effect the way delayed acks are performed. This
4519 * option allows you to get or set the delayed ack time, in
4520 * milliseconds. It also allows changing the delayed ack frequency.
4521 * Changing the frequency to 1 disables the delayed sack algorithm. If
4522 * the assoc_id is 0, then this sets or gets the endpoints default
4523 * values. If the assoc_id field is non-zero, then the set or get
4524 * effects the specified association for the one to many model (the
4525 * assoc_id field is ignored by the one to one model). Note that if
4526 * sack_delay or sack_freq are 0 when setting this option, then the
4527 * current values will remain unchanged.
4529 * struct sctp_sack_info {
4530 * sctp_assoc_t sack_assoc_id;
4531 * uint32_t sack_delay;
4532 * uint32_t sack_freq;
4535 * sack_assoc_id - This parameter, indicates which association the user
4536 * is performing an action upon. Note that if this field's value is
4537 * zero then the endpoints default value is changed (effecting future
4538 * associations only).
4540 * sack_delay - This parameter contains the number of milliseconds that
4541 * the user is requesting the delayed ACK timer be set to. Note that
4542 * this value is defined in the standard to be between 200 and 500
4545 * sack_freq - This parameter contains the number of packets that must
4546 * be received before a sack is sent without waiting for the delay
4547 * timer to expire. The default value for this is 2, setting this
4548 * value to 1 will disable the delayed sack algorithm.
4550 static int sctp_getsockopt_delayed_ack(struct sock
*sk
, int len
,
4551 char __user
*optval
,
4554 struct sctp_sack_info params
;
4555 struct sctp_association
*asoc
= NULL
;
4556 struct sctp_sock
*sp
= sctp_sk(sk
);
4558 if (len
>= sizeof(struct sctp_sack_info
)) {
4559 len
= sizeof(struct sctp_sack_info
);
4561 if (copy_from_user(¶ms
, optval
, len
))
4563 } else if (len
== sizeof(struct sctp_assoc_value
)) {
4564 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4565 pr_warn("Use struct sctp_sack_info instead\n");
4566 if (copy_from_user(¶ms
, optval
, len
))
4571 /* Get association, if sack_assoc_id != 0 and the socket is a one
4572 * to many style socket, and an association was not found, then
4573 * the id was invalid.
4575 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
4576 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
4580 /* Fetch association values. */
4581 if (asoc
->param_flags
& SPP_SACKDELAY_ENABLE
) {
4582 params
.sack_delay
= jiffies_to_msecs(
4584 params
.sack_freq
= asoc
->sackfreq
;
4587 params
.sack_delay
= 0;
4588 params
.sack_freq
= 1;
4591 /* Fetch socket values. */
4592 if (sp
->param_flags
& SPP_SACKDELAY_ENABLE
) {
4593 params
.sack_delay
= sp
->sackdelay
;
4594 params
.sack_freq
= sp
->sackfreq
;
4596 params
.sack_delay
= 0;
4597 params
.sack_freq
= 1;
4601 if (copy_to_user(optval
, ¶ms
, len
))
4604 if (put_user(len
, optlen
))
4610 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4612 * Applications can specify protocol parameters for the default association
4613 * initialization. The option name argument to setsockopt() and getsockopt()
4616 * Setting initialization parameters is effective only on an unconnected
4617 * socket (for UDP-style sockets only future associations are effected
4618 * by the change). With TCP-style sockets, this option is inherited by
4619 * sockets derived from a listener socket.
4621 static int sctp_getsockopt_initmsg(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4623 if (len
< sizeof(struct sctp_initmsg
))
4625 len
= sizeof(struct sctp_initmsg
);
4626 if (put_user(len
, optlen
))
4628 if (copy_to_user(optval
, &sctp_sk(sk
)->initmsg
, len
))
4634 static int sctp_getsockopt_peer_addrs(struct sock
*sk
, int len
,
4635 char __user
*optval
, int __user
*optlen
)
4637 struct sctp_association
*asoc
;
4639 struct sctp_getaddrs getaddrs
;
4640 struct sctp_transport
*from
;
4642 union sctp_addr temp
;
4643 struct sctp_sock
*sp
= sctp_sk(sk
);
4648 if (len
< sizeof(struct sctp_getaddrs
))
4651 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
4654 /* For UDP-style sockets, id specifies the association to query. */
4655 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
4659 to
= optval
+ offsetof(struct sctp_getaddrs
,addrs
);
4660 space_left
= len
- offsetof(struct sctp_getaddrs
,addrs
);
4662 list_for_each_entry(from
, &asoc
->peer
.transport_addr_list
,
4664 memcpy(&temp
, &from
->ipaddr
, sizeof(temp
));
4665 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
4666 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4667 if (space_left
< addrlen
)
4669 if (copy_to_user(to
, &temp
, addrlen
))
4673 space_left
-= addrlen
;
4676 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
))
4678 bytes_copied
= ((char __user
*)to
) - optval
;
4679 if (put_user(bytes_copied
, optlen
))
4685 static int sctp_copy_laddrs(struct sock
*sk
, __u16 port
, void *to
,
4686 size_t space_left
, int *bytes_copied
)
4688 struct sctp_sockaddr_entry
*addr
;
4689 union sctp_addr temp
;
4692 struct net
*net
= sock_net(sk
);
4695 list_for_each_entry_rcu(addr
, &net
->sctp
.local_addr_list
, list
) {
4699 if ((PF_INET
== sk
->sk_family
) &&
4700 (AF_INET6
== addr
->a
.sa
.sa_family
))
4702 if ((PF_INET6
== sk
->sk_family
) &&
4703 inet_v6_ipv6only(sk
) &&
4704 (AF_INET
== addr
->a
.sa
.sa_family
))
4706 memcpy(&temp
, &addr
->a
, sizeof(temp
));
4707 if (!temp
.v4
.sin_port
)
4708 temp
.v4
.sin_port
= htons(port
);
4710 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
4712 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4713 if (space_left
< addrlen
) {
4717 memcpy(to
, &temp
, addrlen
);
4721 space_left
-= addrlen
;
4722 *bytes_copied
+= addrlen
;
4730 static int sctp_getsockopt_local_addrs(struct sock
*sk
, int len
,
4731 char __user
*optval
, int __user
*optlen
)
4733 struct sctp_bind_addr
*bp
;
4734 struct sctp_association
*asoc
;
4736 struct sctp_getaddrs getaddrs
;
4737 struct sctp_sockaddr_entry
*addr
;
4739 union sctp_addr temp
;
4740 struct sctp_sock
*sp
= sctp_sk(sk
);
4744 int bytes_copied
= 0;
4748 if (len
< sizeof(struct sctp_getaddrs
))
4751 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
4755 * For UDP-style sockets, id specifies the association to query.
4756 * If the id field is set to the value '0' then the locally bound
4757 * addresses are returned without regard to any particular
4760 if (0 == getaddrs
.assoc_id
) {
4761 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
4763 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
4766 bp
= &asoc
->base
.bind_addr
;
4769 to
= optval
+ offsetof(struct sctp_getaddrs
,addrs
);
4770 space_left
= len
- offsetof(struct sctp_getaddrs
,addrs
);
4772 addrs
= kmalloc(space_left
, GFP_KERNEL
);
4776 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4777 * addresses from the global local address list.
4779 if (sctp_list_single_entry(&bp
->address_list
)) {
4780 addr
= list_entry(bp
->address_list
.next
,
4781 struct sctp_sockaddr_entry
, list
);
4782 if (sctp_is_any(sk
, &addr
->a
)) {
4783 cnt
= sctp_copy_laddrs(sk
, bp
->port
, addrs
,
4784 space_left
, &bytes_copied
);
4794 /* Protection on the bound address list is not needed since
4795 * in the socket option context we hold a socket lock and
4796 * thus the bound address list can't change.
4798 list_for_each_entry(addr
, &bp
->address_list
, list
) {
4799 memcpy(&temp
, &addr
->a
, sizeof(temp
));
4800 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
4801 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4802 if (space_left
< addrlen
) {
4803 err
= -ENOMEM
; /*fixme: right error?*/
4806 memcpy(buf
, &temp
, addrlen
);
4808 bytes_copied
+= addrlen
;
4810 space_left
-= addrlen
;
4814 if (copy_to_user(to
, addrs
, bytes_copied
)) {
4818 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
)) {
4822 if (put_user(bytes_copied
, optlen
))
4829 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4831 * Requests that the local SCTP stack use the enclosed peer address as
4832 * the association primary. The enclosed address must be one of the
4833 * association peer's addresses.
4835 static int sctp_getsockopt_primary_addr(struct sock
*sk
, int len
,
4836 char __user
*optval
, int __user
*optlen
)
4838 struct sctp_prim prim
;
4839 struct sctp_association
*asoc
;
4840 struct sctp_sock
*sp
= sctp_sk(sk
);
4842 if (len
< sizeof(struct sctp_prim
))
4845 len
= sizeof(struct sctp_prim
);
4847 if (copy_from_user(&prim
, optval
, len
))
4850 asoc
= sctp_id2assoc(sk
, prim
.ssp_assoc_id
);
4854 if (!asoc
->peer
.primary_path
)
4857 memcpy(&prim
.ssp_addr
, &asoc
->peer
.primary_path
->ipaddr
,
4858 asoc
->peer
.primary_path
->af_specific
->sockaddr_len
);
4860 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
,
4861 (union sctp_addr
*)&prim
.ssp_addr
);
4863 if (put_user(len
, optlen
))
4865 if (copy_to_user(optval
, &prim
, len
))
4872 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4874 * Requests that the local endpoint set the specified Adaptation Layer
4875 * Indication parameter for all future INIT and INIT-ACK exchanges.
4877 static int sctp_getsockopt_adaptation_layer(struct sock
*sk
, int len
,
4878 char __user
*optval
, int __user
*optlen
)
4880 struct sctp_setadaptation adaptation
;
4882 if (len
< sizeof(struct sctp_setadaptation
))
4885 len
= sizeof(struct sctp_setadaptation
);
4887 adaptation
.ssb_adaptation_ind
= sctp_sk(sk
)->adaptation_ind
;
4889 if (put_user(len
, optlen
))
4891 if (copy_to_user(optval
, &adaptation
, len
))
4899 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4901 * Applications that wish to use the sendto() system call may wish to
4902 * specify a default set of parameters that would normally be supplied
4903 * through the inclusion of ancillary data. This socket option allows
4904 * such an application to set the default sctp_sndrcvinfo structure.
4907 * The application that wishes to use this socket option simply passes
4908 * in to this call the sctp_sndrcvinfo structure defined in Section
4909 * 5.2.2) The input parameters accepted by this call include
4910 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4911 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4912 * to this call if the caller is using the UDP model.
4914 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4916 static int sctp_getsockopt_default_send_param(struct sock
*sk
,
4917 int len
, char __user
*optval
,
4920 struct sctp_sndrcvinfo info
;
4921 struct sctp_association
*asoc
;
4922 struct sctp_sock
*sp
= sctp_sk(sk
);
4924 if (len
< sizeof(struct sctp_sndrcvinfo
))
4927 len
= sizeof(struct sctp_sndrcvinfo
);
4929 if (copy_from_user(&info
, optval
, len
))
4932 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
4933 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
4937 info
.sinfo_stream
= asoc
->default_stream
;
4938 info
.sinfo_flags
= asoc
->default_flags
;
4939 info
.sinfo_ppid
= asoc
->default_ppid
;
4940 info
.sinfo_context
= asoc
->default_context
;
4941 info
.sinfo_timetolive
= asoc
->default_timetolive
;
4943 info
.sinfo_stream
= sp
->default_stream
;
4944 info
.sinfo_flags
= sp
->default_flags
;
4945 info
.sinfo_ppid
= sp
->default_ppid
;
4946 info
.sinfo_context
= sp
->default_context
;
4947 info
.sinfo_timetolive
= sp
->default_timetolive
;
4950 if (put_user(len
, optlen
))
4952 if (copy_to_user(optval
, &info
, len
))
4960 * 7.1.5 SCTP_NODELAY
4962 * Turn on/off any Nagle-like algorithm. This means that packets are
4963 * generally sent as soon as possible and no unnecessary delays are
4964 * introduced, at the cost of more packets in the network. Expects an
4965 * integer boolean flag.
4968 static int sctp_getsockopt_nodelay(struct sock
*sk
, int len
,
4969 char __user
*optval
, int __user
*optlen
)
4973 if (len
< sizeof(int))
4977 val
= (sctp_sk(sk
)->nodelay
== 1);
4978 if (put_user(len
, optlen
))
4980 if (copy_to_user(optval
, &val
, len
))
4987 * 7.1.1 SCTP_RTOINFO
4989 * The protocol parameters used to initialize and bound retransmission
4990 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4991 * and modify these parameters.
4992 * All parameters are time values, in milliseconds. A value of 0, when
4993 * modifying the parameters, indicates that the current value should not
4997 static int sctp_getsockopt_rtoinfo(struct sock
*sk
, int len
,
4998 char __user
*optval
,
4999 int __user
*optlen
) {
5000 struct sctp_rtoinfo rtoinfo
;
5001 struct sctp_association
*asoc
;
5003 if (len
< sizeof (struct sctp_rtoinfo
))
5006 len
= sizeof(struct sctp_rtoinfo
);
5008 if (copy_from_user(&rtoinfo
, optval
, len
))
5011 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
5013 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
5016 /* Values corresponding to the specific association. */
5018 rtoinfo
.srto_initial
= jiffies_to_msecs(asoc
->rto_initial
);
5019 rtoinfo
.srto_max
= jiffies_to_msecs(asoc
->rto_max
);
5020 rtoinfo
.srto_min
= jiffies_to_msecs(asoc
->rto_min
);
5022 /* Values corresponding to the endpoint. */
5023 struct sctp_sock
*sp
= sctp_sk(sk
);
5025 rtoinfo
.srto_initial
= sp
->rtoinfo
.srto_initial
;
5026 rtoinfo
.srto_max
= sp
->rtoinfo
.srto_max
;
5027 rtoinfo
.srto_min
= sp
->rtoinfo
.srto_min
;
5030 if (put_user(len
, optlen
))
5033 if (copy_to_user(optval
, &rtoinfo
, len
))
5041 * 7.1.2 SCTP_ASSOCINFO
5043 * This option is used to tune the maximum retransmission attempts
5044 * of the association.
5045 * Returns an error if the new association retransmission value is
5046 * greater than the sum of the retransmission value of the peer.
5047 * See [SCTP] for more information.
5050 static int sctp_getsockopt_associnfo(struct sock
*sk
, int len
,
5051 char __user
*optval
,
5055 struct sctp_assocparams assocparams
;
5056 struct sctp_association
*asoc
;
5057 struct list_head
*pos
;
5060 if (len
< sizeof (struct sctp_assocparams
))
5063 len
= sizeof(struct sctp_assocparams
);
5065 if (copy_from_user(&assocparams
, optval
, len
))
5068 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
5070 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
5073 /* Values correspoinding to the specific association */
5075 assocparams
.sasoc_asocmaxrxt
= asoc
->max_retrans
;
5076 assocparams
.sasoc_peer_rwnd
= asoc
->peer
.rwnd
;
5077 assocparams
.sasoc_local_rwnd
= asoc
->a_rwnd
;
5078 assocparams
.sasoc_cookie_life
= (asoc
->cookie_life
.tv_sec
5080 (asoc
->cookie_life
.tv_usec
5083 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
5087 assocparams
.sasoc_number_peer_destinations
= cnt
;
5089 /* Values corresponding to the endpoint */
5090 struct sctp_sock
*sp
= sctp_sk(sk
);
5092 assocparams
.sasoc_asocmaxrxt
= sp
->assocparams
.sasoc_asocmaxrxt
;
5093 assocparams
.sasoc_peer_rwnd
= sp
->assocparams
.sasoc_peer_rwnd
;
5094 assocparams
.sasoc_local_rwnd
= sp
->assocparams
.sasoc_local_rwnd
;
5095 assocparams
.sasoc_cookie_life
=
5096 sp
->assocparams
.sasoc_cookie_life
;
5097 assocparams
.sasoc_number_peer_destinations
=
5099 sasoc_number_peer_destinations
;
5102 if (put_user(len
, optlen
))
5105 if (copy_to_user(optval
, &assocparams
, len
))
5112 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5114 * This socket option is a boolean flag which turns on or off mapped V4
5115 * addresses. If this option is turned on and the socket is type
5116 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5117 * If this option is turned off, then no mapping will be done of V4
5118 * addresses and a user will receive both PF_INET6 and PF_INET type
5119 * addresses on the socket.
5121 static int sctp_getsockopt_mappedv4(struct sock
*sk
, int len
,
5122 char __user
*optval
, int __user
*optlen
)
5125 struct sctp_sock
*sp
= sctp_sk(sk
);
5127 if (len
< sizeof(int))
5132 if (put_user(len
, optlen
))
5134 if (copy_to_user(optval
, &val
, len
))
5141 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5142 * (chapter and verse is quoted at sctp_setsockopt_context())
5144 static int sctp_getsockopt_context(struct sock
*sk
, int len
,
5145 char __user
*optval
, int __user
*optlen
)
5147 struct sctp_assoc_value params
;
5148 struct sctp_sock
*sp
;
5149 struct sctp_association
*asoc
;
5151 if (len
< sizeof(struct sctp_assoc_value
))
5154 len
= sizeof(struct sctp_assoc_value
);
5156 if (copy_from_user(¶ms
, optval
, len
))
5161 if (params
.assoc_id
!= 0) {
5162 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5165 params
.assoc_value
= asoc
->default_rcv_context
;
5167 params
.assoc_value
= sp
->default_rcv_context
;
5170 if (put_user(len
, optlen
))
5172 if (copy_to_user(optval
, ¶ms
, len
))
5179 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5180 * This option will get or set the maximum size to put in any outgoing
5181 * SCTP DATA chunk. If a message is larger than this size it will be
5182 * fragmented by SCTP into the specified size. Note that the underlying
5183 * SCTP implementation may fragment into smaller sized chunks when the
5184 * PMTU of the underlying association is smaller than the value set by
5185 * the user. The default value for this option is '0' which indicates
5186 * the user is NOT limiting fragmentation and only the PMTU will effect
5187 * SCTP's choice of DATA chunk size. Note also that values set larger
5188 * than the maximum size of an IP datagram will effectively let SCTP
5189 * control fragmentation (i.e. the same as setting this option to 0).
5191 * The following structure is used to access and modify this parameter:
5193 * struct sctp_assoc_value {
5194 * sctp_assoc_t assoc_id;
5195 * uint32_t assoc_value;
5198 * assoc_id: This parameter is ignored for one-to-one style sockets.
5199 * For one-to-many style sockets this parameter indicates which
5200 * association the user is performing an action upon. Note that if
5201 * this field's value is zero then the endpoints default value is
5202 * changed (effecting future associations only).
5203 * assoc_value: This parameter specifies the maximum size in bytes.
5205 static int sctp_getsockopt_maxseg(struct sock
*sk
, int len
,
5206 char __user
*optval
, int __user
*optlen
)
5208 struct sctp_assoc_value params
;
5209 struct sctp_association
*asoc
;
5211 if (len
== sizeof(int)) {
5212 pr_warn("Use of int in maxseg socket option deprecated\n");
5213 pr_warn("Use struct sctp_assoc_value instead\n");
5214 params
.assoc_id
= 0;
5215 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5216 len
= sizeof(struct sctp_assoc_value
);
5217 if (copy_from_user(¶ms
, optval
, sizeof(params
)))
5222 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5223 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
5227 params
.assoc_value
= asoc
->frag_point
;
5229 params
.assoc_value
= sctp_sk(sk
)->user_frag
;
5231 if (put_user(len
, optlen
))
5233 if (len
== sizeof(int)) {
5234 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5237 if (copy_to_user(optval
, ¶ms
, len
))
5245 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5246 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5248 static int sctp_getsockopt_fragment_interleave(struct sock
*sk
, int len
,
5249 char __user
*optval
, int __user
*optlen
)
5253 if (len
< sizeof(int))
5258 val
= sctp_sk(sk
)->frag_interleave
;
5259 if (put_user(len
, optlen
))
5261 if (copy_to_user(optval
, &val
, len
))
5268 * 7.1.25. Set or Get the sctp partial delivery point
5269 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5271 static int sctp_getsockopt_partial_delivery_point(struct sock
*sk
, int len
,
5272 char __user
*optval
,
5277 if (len
< sizeof(u32
))
5282 val
= sctp_sk(sk
)->pd_point
;
5283 if (put_user(len
, optlen
))
5285 if (copy_to_user(optval
, &val
, len
))
5292 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5293 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5295 static int sctp_getsockopt_maxburst(struct sock
*sk
, int len
,
5296 char __user
*optval
,
5299 struct sctp_assoc_value params
;
5300 struct sctp_sock
*sp
;
5301 struct sctp_association
*asoc
;
5303 if (len
== sizeof(int)) {
5304 pr_warn("Use of int in max_burst socket option deprecated\n");
5305 pr_warn("Use struct sctp_assoc_value instead\n");
5306 params
.assoc_id
= 0;
5307 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5308 len
= sizeof(struct sctp_assoc_value
);
5309 if (copy_from_user(¶ms
, optval
, len
))
5316 if (params
.assoc_id
!= 0) {
5317 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5320 params
.assoc_value
= asoc
->max_burst
;
5322 params
.assoc_value
= sp
->max_burst
;
5324 if (len
== sizeof(int)) {
5325 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5328 if (copy_to_user(optval
, ¶ms
, len
))
5336 static int sctp_getsockopt_hmac_ident(struct sock
*sk
, int len
,
5337 char __user
*optval
, int __user
*optlen
)
5339 struct net
*net
= sock_net(sk
);
5340 struct sctp_hmacalgo __user
*p
= (void __user
*)optval
;
5341 struct sctp_hmac_algo_param
*hmacs
;
5345 if (!net
->sctp
.auth_enable
)
5348 hmacs
= sctp_sk(sk
)->ep
->auth_hmacs_list
;
5349 data_len
= ntohs(hmacs
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5351 if (len
< sizeof(struct sctp_hmacalgo
) + data_len
)
5354 len
= sizeof(struct sctp_hmacalgo
) + data_len
;
5355 num_idents
= data_len
/ sizeof(u16
);
5357 if (put_user(len
, optlen
))
5359 if (put_user(num_idents
, &p
->shmac_num_idents
))
5361 if (copy_to_user(p
->shmac_idents
, hmacs
->hmac_ids
, data_len
))
5366 static int sctp_getsockopt_active_key(struct sock
*sk
, int len
,
5367 char __user
*optval
, int __user
*optlen
)
5369 struct net
*net
= sock_net(sk
);
5370 struct sctp_authkeyid val
;
5371 struct sctp_association
*asoc
;
5373 if (!net
->sctp
.auth_enable
)
5376 if (len
< sizeof(struct sctp_authkeyid
))
5378 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authkeyid
)))
5381 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
5382 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
5386 val
.scact_keynumber
= asoc
->active_key_id
;
5388 val
.scact_keynumber
= sctp_sk(sk
)->ep
->active_key_id
;
5390 len
= sizeof(struct sctp_authkeyid
);
5391 if (put_user(len
, optlen
))
5393 if (copy_to_user(optval
, &val
, len
))
5399 static int sctp_getsockopt_peer_auth_chunks(struct sock
*sk
, int len
,
5400 char __user
*optval
, int __user
*optlen
)
5402 struct net
*net
= sock_net(sk
);
5403 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5404 struct sctp_authchunks val
;
5405 struct sctp_association
*asoc
;
5406 struct sctp_chunks_param
*ch
;
5410 if (!net
->sctp
.auth_enable
)
5413 if (len
< sizeof(struct sctp_authchunks
))
5416 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5419 to
= p
->gauth_chunks
;
5420 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5424 ch
= asoc
->peer
.peer_chunks
;
5428 /* See if the user provided enough room for all the data */
5429 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5430 if (len
< num_chunks
)
5433 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
5436 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
5437 if (put_user(len
, optlen
)) return -EFAULT
;
5438 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
5443 static int sctp_getsockopt_local_auth_chunks(struct sock
*sk
, int len
,
5444 char __user
*optval
, int __user
*optlen
)
5446 struct net
*net
= sock_net(sk
);
5447 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5448 struct sctp_authchunks val
;
5449 struct sctp_association
*asoc
;
5450 struct sctp_chunks_param
*ch
;
5454 if (!net
->sctp
.auth_enable
)
5457 if (len
< sizeof(struct sctp_authchunks
))
5460 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5463 to
= p
->gauth_chunks
;
5464 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5465 if (!asoc
&& val
.gauth_assoc_id
&& sctp_style(sk
, UDP
))
5469 ch
= (struct sctp_chunks_param
*)asoc
->c
.auth_chunks
;
5471 ch
= sctp_sk(sk
)->ep
->auth_chunk_list
;
5476 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5477 if (len
< sizeof(struct sctp_authchunks
) + num_chunks
)
5480 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
5483 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
5484 if (put_user(len
, optlen
))
5486 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
5493 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5494 * This option gets the current number of associations that are attached
5495 * to a one-to-many style socket. The option value is an uint32_t.
5497 static int sctp_getsockopt_assoc_number(struct sock
*sk
, int len
,
5498 char __user
*optval
, int __user
*optlen
)
5500 struct sctp_sock
*sp
= sctp_sk(sk
);
5501 struct sctp_association
*asoc
;
5504 if (sctp_style(sk
, TCP
))
5507 if (len
< sizeof(u32
))
5512 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5516 if (put_user(len
, optlen
))
5518 if (copy_to_user(optval
, &val
, len
))
5525 * 8.1.23 SCTP_AUTO_ASCONF
5526 * See the corresponding setsockopt entry as description
5528 static int sctp_getsockopt_auto_asconf(struct sock
*sk
, int len
,
5529 char __user
*optval
, int __user
*optlen
)
5533 if (len
< sizeof(int))
5537 if (sctp_sk(sk
)->do_auto_asconf
&& sctp_is_ep_boundall(sk
))
5539 if (put_user(len
, optlen
))
5541 if (copy_to_user(optval
, &val
, len
))
5547 * 8.2.6. Get the Current Identifiers of Associations
5548 * (SCTP_GET_ASSOC_ID_LIST)
5550 * This option gets the current list of SCTP association identifiers of
5551 * the SCTP associations handled by a one-to-many style socket.
5553 static int sctp_getsockopt_assoc_ids(struct sock
*sk
, int len
,
5554 char __user
*optval
, int __user
*optlen
)
5556 struct sctp_sock
*sp
= sctp_sk(sk
);
5557 struct sctp_association
*asoc
;
5558 struct sctp_assoc_ids
*ids
;
5561 if (sctp_style(sk
, TCP
))
5564 if (len
< sizeof(struct sctp_assoc_ids
))
5567 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5571 if (len
< sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
)
5574 len
= sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
;
5576 ids
= kmalloc(len
, GFP_KERNEL
);
5580 ids
->gaids_number_of_ids
= num
;
5582 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5583 ids
->gaids_assoc_id
[num
++] = asoc
->assoc_id
;
5586 if (put_user(len
, optlen
) || copy_to_user(optval
, ids
, len
)) {
5596 * SCTP_PEER_ADDR_THLDS
5598 * This option allows us to fetch the partially failed threshold for one or all
5599 * transports in an association. See Section 6.1 of:
5600 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5602 static int sctp_getsockopt_paddr_thresholds(struct sock
*sk
,
5603 char __user
*optval
,
5607 struct sctp_paddrthlds val
;
5608 struct sctp_transport
*trans
;
5609 struct sctp_association
*asoc
;
5611 if (len
< sizeof(struct sctp_paddrthlds
))
5613 len
= sizeof(struct sctp_paddrthlds
);
5614 if (copy_from_user(&val
, (struct sctp_paddrthlds __user
*)optval
, len
))
5617 if (sctp_is_any(sk
, (const union sctp_addr
*)&val
.spt_address
)) {
5618 asoc
= sctp_id2assoc(sk
, val
.spt_assoc_id
);
5622 val
.spt_pathpfthld
= asoc
->pf_retrans
;
5623 val
.spt_pathmaxrxt
= asoc
->pathmaxrxt
;
5625 trans
= sctp_addr_id2transport(sk
, &val
.spt_address
,
5630 val
.spt_pathmaxrxt
= trans
->pathmaxrxt
;
5631 val
.spt_pathpfthld
= trans
->pf_retrans
;
5634 if (put_user(len
, optlen
) || copy_to_user(optval
, &val
, len
))
5641 * SCTP_GET_ASSOC_STATS
5643 * This option retrieves local per endpoint statistics. It is modeled
5644 * after OpenSolaris' implementation
5646 static int sctp_getsockopt_assoc_stats(struct sock
*sk
, int len
,
5647 char __user
*optval
,
5650 struct sctp_assoc_stats sas
;
5651 struct sctp_association
*asoc
= NULL
;
5653 /* User must provide at least the assoc id */
5654 if (len
< sizeof(sctp_assoc_t
))
5657 /* Allow the struct to grow and fill in as much as possible */
5658 len
= min_t(size_t, len
, sizeof(sas
));
5660 if (copy_from_user(&sas
, optval
, len
))
5663 asoc
= sctp_id2assoc(sk
, sas
.sas_assoc_id
);
5667 sas
.sas_rtxchunks
= asoc
->stats
.rtxchunks
;
5668 sas
.sas_gapcnt
= asoc
->stats
.gapcnt
;
5669 sas
.sas_outofseqtsns
= asoc
->stats
.outofseqtsns
;
5670 sas
.sas_osacks
= asoc
->stats
.osacks
;
5671 sas
.sas_isacks
= asoc
->stats
.isacks
;
5672 sas
.sas_octrlchunks
= asoc
->stats
.octrlchunks
;
5673 sas
.sas_ictrlchunks
= asoc
->stats
.ictrlchunks
;
5674 sas
.sas_oodchunks
= asoc
->stats
.oodchunks
;
5675 sas
.sas_iodchunks
= asoc
->stats
.iodchunks
;
5676 sas
.sas_ouodchunks
= asoc
->stats
.ouodchunks
;
5677 sas
.sas_iuodchunks
= asoc
->stats
.iuodchunks
;
5678 sas
.sas_idupchunks
= asoc
->stats
.idupchunks
;
5679 sas
.sas_opackets
= asoc
->stats
.opackets
;
5680 sas
.sas_ipackets
= asoc
->stats
.ipackets
;
5682 /* New high max rto observed, will return 0 if not a single
5683 * RTO update took place. obs_rto_ipaddr will be bogus
5686 sas
.sas_maxrto
= asoc
->stats
.max_obs_rto
;
5687 memcpy(&sas
.sas_obs_rto_ipaddr
, &asoc
->stats
.obs_rto_ipaddr
,
5688 sizeof(struct sockaddr_storage
));
5690 /* Mark beginning of a new observation period */
5691 asoc
->stats
.max_obs_rto
= asoc
->rto_min
;
5693 if (put_user(len
, optlen
))
5696 SCTP_DEBUG_PRINTK("sctp_getsockopt_assoc_stat(%d): %d\n",
5697 len
, sas
.sas_assoc_id
);
5699 if (copy_to_user(optval
, &sas
, len
))
5705 SCTP_STATIC
int sctp_getsockopt(struct sock
*sk
, int level
, int optname
,
5706 char __user
*optval
, int __user
*optlen
)
5711 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5714 /* I can hardly begin to describe how wrong this is. This is
5715 * so broken as to be worse than useless. The API draft
5716 * REALLY is NOT helpful here... I am not convinced that the
5717 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5718 * are at all well-founded.
5720 if (level
!= SOL_SCTP
) {
5721 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
5723 retval
= af
->getsockopt(sk
, level
, optname
, optval
, optlen
);
5727 if (get_user(len
, optlen
))
5734 retval
= sctp_getsockopt_sctp_status(sk
, len
, optval
, optlen
);
5736 case SCTP_DISABLE_FRAGMENTS
:
5737 retval
= sctp_getsockopt_disable_fragments(sk
, len
, optval
,
5741 retval
= sctp_getsockopt_events(sk
, len
, optval
, optlen
);
5743 case SCTP_AUTOCLOSE
:
5744 retval
= sctp_getsockopt_autoclose(sk
, len
, optval
, optlen
);
5746 case SCTP_SOCKOPT_PEELOFF
:
5747 retval
= sctp_getsockopt_peeloff(sk
, len
, optval
, optlen
);
5749 case SCTP_PEER_ADDR_PARAMS
:
5750 retval
= sctp_getsockopt_peer_addr_params(sk
, len
, optval
,
5753 case SCTP_DELAYED_SACK
:
5754 retval
= sctp_getsockopt_delayed_ack(sk
, len
, optval
,
5758 retval
= sctp_getsockopt_initmsg(sk
, len
, optval
, optlen
);
5760 case SCTP_GET_PEER_ADDRS
:
5761 retval
= sctp_getsockopt_peer_addrs(sk
, len
, optval
,
5764 case SCTP_GET_LOCAL_ADDRS
:
5765 retval
= sctp_getsockopt_local_addrs(sk
, len
, optval
,
5768 case SCTP_SOCKOPT_CONNECTX3
:
5769 retval
= sctp_getsockopt_connectx3(sk
, len
, optval
, optlen
);
5771 case SCTP_DEFAULT_SEND_PARAM
:
5772 retval
= sctp_getsockopt_default_send_param(sk
, len
,
5775 case SCTP_PRIMARY_ADDR
:
5776 retval
= sctp_getsockopt_primary_addr(sk
, len
, optval
, optlen
);
5779 retval
= sctp_getsockopt_nodelay(sk
, len
, optval
, optlen
);
5782 retval
= sctp_getsockopt_rtoinfo(sk
, len
, optval
, optlen
);
5784 case SCTP_ASSOCINFO
:
5785 retval
= sctp_getsockopt_associnfo(sk
, len
, optval
, optlen
);
5787 case SCTP_I_WANT_MAPPED_V4_ADDR
:
5788 retval
= sctp_getsockopt_mappedv4(sk
, len
, optval
, optlen
);
5791 retval
= sctp_getsockopt_maxseg(sk
, len
, optval
, optlen
);
5793 case SCTP_GET_PEER_ADDR_INFO
:
5794 retval
= sctp_getsockopt_peer_addr_info(sk
, len
, optval
,
5797 case SCTP_ADAPTATION_LAYER
:
5798 retval
= sctp_getsockopt_adaptation_layer(sk
, len
, optval
,
5802 retval
= sctp_getsockopt_context(sk
, len
, optval
, optlen
);
5804 case SCTP_FRAGMENT_INTERLEAVE
:
5805 retval
= sctp_getsockopt_fragment_interleave(sk
, len
, optval
,
5808 case SCTP_PARTIAL_DELIVERY_POINT
:
5809 retval
= sctp_getsockopt_partial_delivery_point(sk
, len
, optval
,
5812 case SCTP_MAX_BURST
:
5813 retval
= sctp_getsockopt_maxburst(sk
, len
, optval
, optlen
);
5816 case SCTP_AUTH_CHUNK
:
5817 case SCTP_AUTH_DELETE_KEY
:
5818 retval
= -EOPNOTSUPP
;
5820 case SCTP_HMAC_IDENT
:
5821 retval
= sctp_getsockopt_hmac_ident(sk
, len
, optval
, optlen
);
5823 case SCTP_AUTH_ACTIVE_KEY
:
5824 retval
= sctp_getsockopt_active_key(sk
, len
, optval
, optlen
);
5826 case SCTP_PEER_AUTH_CHUNKS
:
5827 retval
= sctp_getsockopt_peer_auth_chunks(sk
, len
, optval
,
5830 case SCTP_LOCAL_AUTH_CHUNKS
:
5831 retval
= sctp_getsockopt_local_auth_chunks(sk
, len
, optval
,
5834 case SCTP_GET_ASSOC_NUMBER
:
5835 retval
= sctp_getsockopt_assoc_number(sk
, len
, optval
, optlen
);
5837 case SCTP_GET_ASSOC_ID_LIST
:
5838 retval
= sctp_getsockopt_assoc_ids(sk
, len
, optval
, optlen
);
5840 case SCTP_AUTO_ASCONF
:
5841 retval
= sctp_getsockopt_auto_asconf(sk
, len
, optval
, optlen
);
5843 case SCTP_PEER_ADDR_THLDS
:
5844 retval
= sctp_getsockopt_paddr_thresholds(sk
, optval
, len
, optlen
);
5846 case SCTP_GET_ASSOC_STATS
:
5847 retval
= sctp_getsockopt_assoc_stats(sk
, len
, optval
, optlen
);
5850 retval
= -ENOPROTOOPT
;
5854 sctp_release_sock(sk
);
5858 static void sctp_hash(struct sock
*sk
)
5863 static void sctp_unhash(struct sock
*sk
)
5868 /* Check if port is acceptable. Possibly find first available port.
5870 * The port hash table (contained in the 'global' SCTP protocol storage
5871 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5872 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5873 * list (the list number is the port number hashed out, so as you
5874 * would expect from a hash function, all the ports in a given list have
5875 * such a number that hashes out to the same list number; you were
5876 * expecting that, right?); so each list has a set of ports, with a
5877 * link to the socket (struct sock) that uses it, the port number and
5878 * a fastreuse flag (FIXME: NPI ipg).
5880 static struct sctp_bind_bucket
*sctp_bucket_create(
5881 struct sctp_bind_hashbucket
*head
, struct net
*, unsigned short snum
);
5883 static long sctp_get_port_local(struct sock
*sk
, union sctp_addr
*addr
)
5885 struct sctp_bind_hashbucket
*head
; /* hash list */
5886 struct sctp_bind_bucket
*pp
;
5887 unsigned short snum
;
5890 snum
= ntohs(addr
->v4
.sin_port
);
5892 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum
);
5893 sctp_local_bh_disable();
5896 /* Search for an available port. */
5897 int low
, high
, remaining
, index
;
5900 inet_get_local_port_range(&low
, &high
);
5901 remaining
= (high
- low
) + 1;
5902 rover
= net_random() % remaining
+ low
;
5906 if ((rover
< low
) || (rover
> high
))
5908 if (inet_is_reserved_local_port(rover
))
5910 index
= sctp_phashfn(sock_net(sk
), rover
);
5911 head
= &sctp_port_hashtable
[index
];
5912 sctp_spin_lock(&head
->lock
);
5913 sctp_for_each_hentry(pp
, &head
->chain
)
5914 if ((pp
->port
== rover
) &&
5915 net_eq(sock_net(sk
), pp
->net
))
5919 sctp_spin_unlock(&head
->lock
);
5920 } while (--remaining
> 0);
5922 /* Exhausted local port range during search? */
5927 /* OK, here is the one we will use. HEAD (the port
5928 * hash table list entry) is non-NULL and we hold it's
5933 /* We are given an specific port number; we verify
5934 * that it is not being used. If it is used, we will
5935 * exahust the search in the hash list corresponding
5936 * to the port number (snum) - we detect that with the
5937 * port iterator, pp being NULL.
5939 head
= &sctp_port_hashtable
[sctp_phashfn(sock_net(sk
), snum
)];
5940 sctp_spin_lock(&head
->lock
);
5941 sctp_for_each_hentry(pp
, &head
->chain
) {
5942 if ((pp
->port
== snum
) && net_eq(pp
->net
, sock_net(sk
)))
5949 if (!hlist_empty(&pp
->owner
)) {
5950 /* We had a port hash table hit - there is an
5951 * available port (pp != NULL) and it is being
5952 * used by other socket (pp->owner not empty); that other
5953 * socket is going to be sk2.
5955 int reuse
= sk
->sk_reuse
;
5958 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5959 if (pp
->fastreuse
&& sk
->sk_reuse
&&
5960 sk
->sk_state
!= SCTP_SS_LISTENING
)
5963 /* Run through the list of sockets bound to the port
5964 * (pp->port) [via the pointers bind_next and
5965 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5966 * we get the endpoint they describe and run through
5967 * the endpoint's list of IP (v4 or v6) addresses,
5968 * comparing each of the addresses with the address of
5969 * the socket sk. If we find a match, then that means
5970 * that this port/socket (sk) combination are already
5973 sk_for_each_bound(sk2
, &pp
->owner
) {
5974 struct sctp_endpoint
*ep2
;
5975 ep2
= sctp_sk(sk2
)->ep
;
5978 (reuse
&& sk2
->sk_reuse
&&
5979 sk2
->sk_state
!= SCTP_SS_LISTENING
))
5982 if (sctp_bind_addr_conflict(&ep2
->base
.bind_addr
, addr
,
5983 sctp_sk(sk2
), sctp_sk(sk
))) {
5988 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5991 /* If there was a hash table miss, create a new port. */
5993 if (!pp
&& !(pp
= sctp_bucket_create(head
, sock_net(sk
), snum
)))
5996 /* In either case (hit or miss), make sure fastreuse is 1 only
5997 * if sk->sk_reuse is too (that is, if the caller requested
5998 * SO_REUSEADDR on this socket -sk-).
6000 if (hlist_empty(&pp
->owner
)) {
6001 if (sk
->sk_reuse
&& sk
->sk_state
!= SCTP_SS_LISTENING
)
6005 } else if (pp
->fastreuse
&&
6006 (!sk
->sk_reuse
|| sk
->sk_state
== SCTP_SS_LISTENING
))
6009 /* We are set, so fill up all the data in the hash table
6010 * entry, tie the socket list information with the rest of the
6011 * sockets FIXME: Blurry, NPI (ipg).
6014 if (!sctp_sk(sk
)->bind_hash
) {
6015 inet_sk(sk
)->inet_num
= snum
;
6016 sk_add_bind_node(sk
, &pp
->owner
);
6017 sctp_sk(sk
)->bind_hash
= pp
;
6022 sctp_spin_unlock(&head
->lock
);
6025 sctp_local_bh_enable();
6029 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6030 * port is requested.
6032 static int sctp_get_port(struct sock
*sk
, unsigned short snum
)
6035 union sctp_addr addr
;
6036 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
6038 /* Set up a dummy address struct from the sk. */
6039 af
->from_sk(&addr
, sk
);
6040 addr
.v4
.sin_port
= htons(snum
);
6042 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6043 ret
= sctp_get_port_local(sk
, &addr
);
6049 * Move a socket to LISTENING state.
6051 SCTP_STATIC
int sctp_listen_start(struct sock
*sk
, int backlog
)
6053 struct sctp_sock
*sp
= sctp_sk(sk
);
6054 struct sctp_endpoint
*ep
= sp
->ep
;
6055 struct crypto_hash
*tfm
= NULL
;
6058 /* Allocate HMAC for generating cookie. */
6059 if (!sp
->hmac
&& sp
->sctp_hmac_alg
) {
6060 sprintf(alg
, "hmac(%s)", sp
->sctp_hmac_alg
);
6061 tfm
= crypto_alloc_hash(alg
, 0, CRYPTO_ALG_ASYNC
);
6063 net_info_ratelimited("failed to load transform for %s: %ld\n",
6064 sp
->sctp_hmac_alg
, PTR_ERR(tfm
));
6067 sctp_sk(sk
)->hmac
= tfm
;
6071 * If a bind() or sctp_bindx() is not called prior to a listen()
6072 * call that allows new associations to be accepted, the system
6073 * picks an ephemeral port and will choose an address set equivalent
6074 * to binding with a wildcard address.
6076 * This is not currently spelled out in the SCTP sockets
6077 * extensions draft, but follows the practice as seen in TCP
6081 sk
->sk_state
= SCTP_SS_LISTENING
;
6082 if (!ep
->base
.bind_addr
.port
) {
6083 if (sctp_autobind(sk
))
6086 if (sctp_get_port(sk
, inet_sk(sk
)->inet_num
)) {
6087 sk
->sk_state
= SCTP_SS_CLOSED
;
6092 sk
->sk_max_ack_backlog
= backlog
;
6093 sctp_hash_endpoint(ep
);
6098 * 4.1.3 / 5.1.3 listen()
6100 * By default, new associations are not accepted for UDP style sockets.
6101 * An application uses listen() to mark a socket as being able to
6102 * accept new associations.
6104 * On TCP style sockets, applications use listen() to ready the SCTP
6105 * endpoint for accepting inbound associations.
6107 * On both types of endpoints a backlog of '0' disables listening.
6109 * Move a socket to LISTENING state.
6111 int sctp_inet_listen(struct socket
*sock
, int backlog
)
6113 struct sock
*sk
= sock
->sk
;
6114 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
6117 if (unlikely(backlog
< 0))
6122 /* Peeled-off sockets are not allowed to listen(). */
6123 if (sctp_style(sk
, UDP_HIGH_BANDWIDTH
))
6126 if (sock
->state
!= SS_UNCONNECTED
)
6129 /* If backlog is zero, disable listening. */
6131 if (sctp_sstate(sk
, CLOSED
))
6135 sctp_unhash_endpoint(ep
);
6136 sk
->sk_state
= SCTP_SS_CLOSED
;
6138 sctp_sk(sk
)->bind_hash
->fastreuse
= 1;
6142 /* If we are already listening, just update the backlog */
6143 if (sctp_sstate(sk
, LISTENING
))
6144 sk
->sk_max_ack_backlog
= backlog
;
6146 err
= sctp_listen_start(sk
, backlog
);
6153 sctp_release_sock(sk
);
6158 * This function is done by modeling the current datagram_poll() and the
6159 * tcp_poll(). Note that, based on these implementations, we don't
6160 * lock the socket in this function, even though it seems that,
6161 * ideally, locking or some other mechanisms can be used to ensure
6162 * the integrity of the counters (sndbuf and wmem_alloc) used
6163 * in this place. We assume that we don't need locks either until proven
6166 * Another thing to note is that we include the Async I/O support
6167 * here, again, by modeling the current TCP/UDP code. We don't have
6168 * a good way to test with it yet.
6170 unsigned int sctp_poll(struct file
*file
, struct socket
*sock
, poll_table
*wait
)
6172 struct sock
*sk
= sock
->sk
;
6173 struct sctp_sock
*sp
= sctp_sk(sk
);
6176 poll_wait(file
, sk_sleep(sk
), wait
);
6178 /* A TCP-style listening socket becomes readable when the accept queue
6181 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
6182 return (!list_empty(&sp
->ep
->asocs
)) ?
6183 (POLLIN
| POLLRDNORM
) : 0;
6187 /* Is there any exceptional events? */
6188 if (sk
->sk_err
|| !skb_queue_empty(&sk
->sk_error_queue
))
6190 sock_flag(sk
, SOCK_SELECT_ERR_QUEUE
) ? POLLPRI
: 0;
6191 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6192 mask
|= POLLRDHUP
| POLLIN
| POLLRDNORM
;
6193 if (sk
->sk_shutdown
== SHUTDOWN_MASK
)
6196 /* Is it readable? Reconsider this code with TCP-style support. */
6197 if (!skb_queue_empty(&sk
->sk_receive_queue
))
6198 mask
|= POLLIN
| POLLRDNORM
;
6200 /* The association is either gone or not ready. */
6201 if (!sctp_style(sk
, UDP
) && sctp_sstate(sk
, CLOSED
))
6204 /* Is it writable? */
6205 if (sctp_writeable(sk
)) {
6206 mask
|= POLLOUT
| POLLWRNORM
;
6208 set_bit(SOCK_ASYNC_NOSPACE
, &sk
->sk_socket
->flags
);
6210 * Since the socket is not locked, the buffer
6211 * might be made available after the writeable check and
6212 * before the bit is set. This could cause a lost I/O
6213 * signal. tcp_poll() has a race breaker for this race
6214 * condition. Based on their implementation, we put
6215 * in the following code to cover it as well.
6217 if (sctp_writeable(sk
))
6218 mask
|= POLLOUT
| POLLWRNORM
;
6223 /********************************************************************
6224 * 2nd Level Abstractions
6225 ********************************************************************/
6227 static struct sctp_bind_bucket
*sctp_bucket_create(
6228 struct sctp_bind_hashbucket
*head
, struct net
*net
, unsigned short snum
)
6230 struct sctp_bind_bucket
*pp
;
6232 pp
= kmem_cache_alloc(sctp_bucket_cachep
, GFP_ATOMIC
);
6234 SCTP_DBG_OBJCNT_INC(bind_bucket
);
6237 INIT_HLIST_HEAD(&pp
->owner
);
6239 hlist_add_head(&pp
->node
, &head
->chain
);
6244 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6245 static void sctp_bucket_destroy(struct sctp_bind_bucket
*pp
)
6247 if (pp
&& hlist_empty(&pp
->owner
)) {
6248 __hlist_del(&pp
->node
);
6249 kmem_cache_free(sctp_bucket_cachep
, pp
);
6250 SCTP_DBG_OBJCNT_DEC(bind_bucket
);
6254 /* Release this socket's reference to a local port. */
6255 static inline void __sctp_put_port(struct sock
*sk
)
6257 struct sctp_bind_hashbucket
*head
=
6258 &sctp_port_hashtable
[sctp_phashfn(sock_net(sk
),
6259 inet_sk(sk
)->inet_num
)];
6260 struct sctp_bind_bucket
*pp
;
6262 sctp_spin_lock(&head
->lock
);
6263 pp
= sctp_sk(sk
)->bind_hash
;
6264 __sk_del_bind_node(sk
);
6265 sctp_sk(sk
)->bind_hash
= NULL
;
6266 inet_sk(sk
)->inet_num
= 0;
6267 sctp_bucket_destroy(pp
);
6268 sctp_spin_unlock(&head
->lock
);
6271 void sctp_put_port(struct sock
*sk
)
6273 sctp_local_bh_disable();
6274 __sctp_put_port(sk
);
6275 sctp_local_bh_enable();
6279 * The system picks an ephemeral port and choose an address set equivalent
6280 * to binding with a wildcard address.
6281 * One of those addresses will be the primary address for the association.
6282 * This automatically enables the multihoming capability of SCTP.
6284 static int sctp_autobind(struct sock
*sk
)
6286 union sctp_addr autoaddr
;
6290 /* Initialize a local sockaddr structure to INADDR_ANY. */
6291 af
= sctp_sk(sk
)->pf
->af
;
6293 port
= htons(inet_sk(sk
)->inet_num
);
6294 af
->inaddr_any(&autoaddr
, port
);
6296 return sctp_do_bind(sk
, &autoaddr
, af
->sockaddr_len
);
6299 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6302 * 4.2 The cmsghdr Structure *
6304 * When ancillary data is sent or received, any number of ancillary data
6305 * objects can be specified by the msg_control and msg_controllen members of
6306 * the msghdr structure, because each object is preceded by
6307 * a cmsghdr structure defining the object's length (the cmsg_len member).
6308 * Historically Berkeley-derived implementations have passed only one object
6309 * at a time, but this API allows multiple objects to be
6310 * passed in a single call to sendmsg() or recvmsg(). The following example
6311 * shows two ancillary data objects in a control buffer.
6313 * |<--------------------------- msg_controllen -------------------------->|
6316 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6318 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6321 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6323 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6326 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6327 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6329 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6331 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6338 SCTP_STATIC
int sctp_msghdr_parse(const struct msghdr
*msg
,
6339 sctp_cmsgs_t
*cmsgs
)
6341 struct cmsghdr
*cmsg
;
6342 struct msghdr
*my_msg
= (struct msghdr
*)msg
;
6344 for (cmsg
= CMSG_FIRSTHDR(msg
);
6346 cmsg
= CMSG_NXTHDR(my_msg
, cmsg
)) {
6347 if (!CMSG_OK(my_msg
, cmsg
))
6350 /* Should we parse this header or ignore? */
6351 if (cmsg
->cmsg_level
!= IPPROTO_SCTP
)
6354 /* Strictly check lengths following example in SCM code. */
6355 switch (cmsg
->cmsg_type
) {
6357 /* SCTP Socket API Extension
6358 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6360 * This cmsghdr structure provides information for
6361 * initializing new SCTP associations with sendmsg().
6362 * The SCTP_INITMSG socket option uses this same data
6363 * structure. This structure is not used for
6366 * cmsg_level cmsg_type cmsg_data[]
6367 * ------------ ------------ ----------------------
6368 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6370 if (cmsg
->cmsg_len
!=
6371 CMSG_LEN(sizeof(struct sctp_initmsg
)))
6373 cmsgs
->init
= (struct sctp_initmsg
*)CMSG_DATA(cmsg
);
6377 /* SCTP Socket API Extension
6378 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6380 * This cmsghdr structure specifies SCTP options for
6381 * sendmsg() and describes SCTP header information
6382 * about a received message through recvmsg().
6384 * cmsg_level cmsg_type cmsg_data[]
6385 * ------------ ------------ ----------------------
6386 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6388 if (cmsg
->cmsg_len
!=
6389 CMSG_LEN(sizeof(struct sctp_sndrcvinfo
)))
6393 (struct sctp_sndrcvinfo
*)CMSG_DATA(cmsg
);
6395 /* Minimally, validate the sinfo_flags. */
6396 if (cmsgs
->info
->sinfo_flags
&
6397 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
6398 SCTP_ABORT
| SCTP_EOF
))
6410 * Wait for a packet..
6411 * Note: This function is the same function as in core/datagram.c
6412 * with a few modifications to make lksctp work.
6414 static int sctp_wait_for_packet(struct sock
* sk
, int *err
, long *timeo_p
)
6419 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
6421 /* Socket errors? */
6422 error
= sock_error(sk
);
6426 if (!skb_queue_empty(&sk
->sk_receive_queue
))
6429 /* Socket shut down? */
6430 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6433 /* Sequenced packets can come disconnected. If so we report the
6438 /* Is there a good reason to think that we may receive some data? */
6439 if (list_empty(&sctp_sk(sk
)->ep
->asocs
) && !sctp_sstate(sk
, LISTENING
))
6442 /* Handle signals. */
6443 if (signal_pending(current
))
6446 /* Let another process have a go. Since we are going to sleep
6447 * anyway. Note: This may cause odd behaviors if the message
6448 * does not fit in the user's buffer, but this seems to be the
6449 * only way to honor MSG_DONTWAIT realistically.
6451 sctp_release_sock(sk
);
6452 *timeo_p
= schedule_timeout(*timeo_p
);
6456 finish_wait(sk_sleep(sk
), &wait
);
6460 error
= sock_intr_errno(*timeo_p
);
6463 finish_wait(sk_sleep(sk
), &wait
);
6468 /* Receive a datagram.
6469 * Note: This is pretty much the same routine as in core/datagram.c
6470 * with a few changes to make lksctp work.
6472 static struct sk_buff
*sctp_skb_recv_datagram(struct sock
*sk
, int flags
,
6473 int noblock
, int *err
)
6476 struct sk_buff
*skb
;
6479 timeo
= sock_rcvtimeo(sk
, noblock
);
6481 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6482 timeo
, MAX_SCHEDULE_TIMEOUT
);
6485 /* Again only user level code calls this function,
6486 * so nothing interrupt level
6487 * will suddenly eat the receive_queue.
6489 * Look at current nfs client by the way...
6490 * However, this function was correct in any case. 8)
6492 if (flags
& MSG_PEEK
) {
6493 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
6494 skb
= skb_peek(&sk
->sk_receive_queue
);
6496 atomic_inc(&skb
->users
);
6497 spin_unlock_bh(&sk
->sk_receive_queue
.lock
);
6499 skb
= skb_dequeue(&sk
->sk_receive_queue
);
6505 /* Caller is allowed not to check sk->sk_err before calling. */
6506 error
= sock_error(sk
);
6510 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6513 /* User doesn't want to wait. */
6517 } while (sctp_wait_for_packet(sk
, err
, &timeo
) == 0);
6526 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6527 static void __sctp_write_space(struct sctp_association
*asoc
)
6529 struct sock
*sk
= asoc
->base
.sk
;
6530 struct socket
*sock
= sk
->sk_socket
;
6532 if ((sctp_wspace(asoc
) > 0) && sock
) {
6533 if (waitqueue_active(&asoc
->wait
))
6534 wake_up_interruptible(&asoc
->wait
);
6536 if (sctp_writeable(sk
)) {
6537 wait_queue_head_t
*wq
= sk_sleep(sk
);
6539 if (wq
&& waitqueue_active(wq
))
6540 wake_up_interruptible(wq
);
6542 /* Note that we try to include the Async I/O support
6543 * here by modeling from the current TCP/UDP code.
6544 * We have not tested with it yet.
6546 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
))
6547 sock_wake_async(sock
,
6548 SOCK_WAKE_SPACE
, POLL_OUT
);
6553 /* Do accounting for the sndbuf space.
6554 * Decrement the used sndbuf space of the corresponding association by the
6555 * data size which was just transmitted(freed).
6557 static void sctp_wfree(struct sk_buff
*skb
)
6559 struct sctp_association
*asoc
;
6560 struct sctp_chunk
*chunk
;
6563 /* Get the saved chunk pointer. */
6564 chunk
= *((struct sctp_chunk
**)(skb
->cb
));
6567 asoc
->sndbuf_used
-= SCTP_DATA_SNDSIZE(chunk
) +
6568 sizeof(struct sk_buff
) +
6569 sizeof(struct sctp_chunk
);
6571 atomic_sub(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
6574 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6576 sk
->sk_wmem_queued
-= skb
->truesize
;
6577 sk_mem_uncharge(sk
, skb
->truesize
);
6580 __sctp_write_space(asoc
);
6582 sctp_association_put(asoc
);
6585 /* Do accounting for the receive space on the socket.
6586 * Accounting for the association is done in ulpevent.c
6587 * We set this as a destructor for the cloned data skbs so that
6588 * accounting is done at the correct time.
6590 void sctp_sock_rfree(struct sk_buff
*skb
)
6592 struct sock
*sk
= skb
->sk
;
6593 struct sctp_ulpevent
*event
= sctp_skb2event(skb
);
6595 atomic_sub(event
->rmem_len
, &sk
->sk_rmem_alloc
);
6598 * Mimic the behavior of sock_rfree
6600 sk_mem_uncharge(sk
, event
->rmem_len
);
6604 /* Helper function to wait for space in the sndbuf. */
6605 static int sctp_wait_for_sndbuf(struct sctp_association
*asoc
, long *timeo_p
,
6608 struct sock
*sk
= asoc
->base
.sk
;
6610 long current_timeo
= *timeo_p
;
6613 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6614 asoc
, (long)(*timeo_p
), msg_len
);
6616 /* Increment the association's refcnt. */
6617 sctp_association_hold(asoc
);
6619 /* Wait on the association specific sndbuf space. */
6621 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
6622 TASK_INTERRUPTIBLE
);
6625 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
6628 if (signal_pending(current
))
6629 goto do_interrupted
;
6630 if (msg_len
<= sctp_wspace(asoc
))
6633 /* Let another process have a go. Since we are going
6636 sctp_release_sock(sk
);
6637 current_timeo
= schedule_timeout(current_timeo
);
6638 BUG_ON(sk
!= asoc
->base
.sk
);
6641 *timeo_p
= current_timeo
;
6645 finish_wait(&asoc
->wait
, &wait
);
6647 /* Release the association's refcnt. */
6648 sctp_association_put(asoc
);
6657 err
= sock_intr_errno(*timeo_p
);
6665 void sctp_data_ready(struct sock
*sk
, int len
)
6667 struct socket_wq
*wq
;
6670 wq
= rcu_dereference(sk
->sk_wq
);
6671 if (wq_has_sleeper(wq
))
6672 wake_up_interruptible_sync_poll(&wq
->wait
, POLLIN
|
6673 POLLRDNORM
| POLLRDBAND
);
6674 sk_wake_async(sk
, SOCK_WAKE_WAITD
, POLL_IN
);
6678 /* If socket sndbuf has changed, wake up all per association waiters. */
6679 void sctp_write_space(struct sock
*sk
)
6681 struct sctp_association
*asoc
;
6683 /* Wake up the tasks in each wait queue. */
6684 list_for_each_entry(asoc
, &((sctp_sk(sk
))->ep
->asocs
), asocs
) {
6685 __sctp_write_space(asoc
);
6689 /* Is there any sndbuf space available on the socket?
6691 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6692 * associations on the same socket. For a UDP-style socket with
6693 * multiple associations, it is possible for it to be "unwriteable"
6694 * prematurely. I assume that this is acceptable because
6695 * a premature "unwriteable" is better than an accidental "writeable" which
6696 * would cause an unwanted block under certain circumstances. For the 1-1
6697 * UDP-style sockets or TCP-style sockets, this code should work.
6700 static int sctp_writeable(struct sock
*sk
)
6704 amt
= sk
->sk_sndbuf
- sk_wmem_alloc_get(sk
);
6710 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6711 * returns immediately with EINPROGRESS.
6713 static int sctp_wait_for_connect(struct sctp_association
*asoc
, long *timeo_p
)
6715 struct sock
*sk
= asoc
->base
.sk
;
6717 long current_timeo
= *timeo_p
;
6720 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__
, asoc
,
6723 /* Increment the association's refcnt. */
6724 sctp_association_hold(asoc
);
6727 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
6728 TASK_INTERRUPTIBLE
);
6731 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6733 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
6736 if (signal_pending(current
))
6737 goto do_interrupted
;
6739 if (sctp_state(asoc
, ESTABLISHED
))
6742 /* Let another process have a go. Since we are going
6745 sctp_release_sock(sk
);
6746 current_timeo
= schedule_timeout(current_timeo
);
6749 *timeo_p
= current_timeo
;
6753 finish_wait(&asoc
->wait
, &wait
);
6755 /* Release the association's refcnt. */
6756 sctp_association_put(asoc
);
6761 if (asoc
->init_err_counter
+ 1 > asoc
->max_init_attempts
)
6764 err
= -ECONNREFUSED
;
6768 err
= sock_intr_errno(*timeo_p
);
6776 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
)
6778 struct sctp_endpoint
*ep
;
6782 ep
= sctp_sk(sk
)->ep
;
6786 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
,
6787 TASK_INTERRUPTIBLE
);
6789 if (list_empty(&ep
->asocs
)) {
6790 sctp_release_sock(sk
);
6791 timeo
= schedule_timeout(timeo
);
6796 if (!sctp_sstate(sk
, LISTENING
))
6800 if (!list_empty(&ep
->asocs
))
6803 err
= sock_intr_errno(timeo
);
6804 if (signal_pending(current
))
6812 finish_wait(sk_sleep(sk
), &wait
);
6817 static void sctp_wait_for_close(struct sock
*sk
, long timeout
)
6822 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
6823 if (list_empty(&sctp_sk(sk
)->ep
->asocs
))
6825 sctp_release_sock(sk
);
6826 timeout
= schedule_timeout(timeout
);
6828 } while (!signal_pending(current
) && timeout
);
6830 finish_wait(sk_sleep(sk
), &wait
);
6833 static void sctp_skb_set_owner_r_frag(struct sk_buff
*skb
, struct sock
*sk
)
6835 struct sk_buff
*frag
;
6840 /* Don't forget the fragments. */
6841 skb_walk_frags(skb
, frag
)
6842 sctp_skb_set_owner_r_frag(frag
, sk
);
6845 sctp_skb_set_owner_r(skb
, sk
);
6848 void sctp_copy_sock(struct sock
*newsk
, struct sock
*sk
,
6849 struct sctp_association
*asoc
)
6851 struct inet_sock
*inet
= inet_sk(sk
);
6852 struct inet_sock
*newinet
;
6854 newsk
->sk_type
= sk
->sk_type
;
6855 newsk
->sk_bound_dev_if
= sk
->sk_bound_dev_if
;
6856 newsk
->sk_flags
= sk
->sk_flags
;
6857 newsk
->sk_no_check
= sk
->sk_no_check
;
6858 newsk
->sk_reuse
= sk
->sk_reuse
;
6860 newsk
->sk_shutdown
= sk
->sk_shutdown
;
6861 newsk
->sk_destruct
= inet_sock_destruct
;
6862 newsk
->sk_family
= sk
->sk_family
;
6863 newsk
->sk_protocol
= IPPROTO_SCTP
;
6864 newsk
->sk_backlog_rcv
= sk
->sk_prot
->backlog_rcv
;
6865 newsk
->sk_sndbuf
= sk
->sk_sndbuf
;
6866 newsk
->sk_rcvbuf
= sk
->sk_rcvbuf
;
6867 newsk
->sk_lingertime
= sk
->sk_lingertime
;
6868 newsk
->sk_rcvtimeo
= sk
->sk_rcvtimeo
;
6869 newsk
->sk_sndtimeo
= sk
->sk_sndtimeo
;
6871 newinet
= inet_sk(newsk
);
6873 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6874 * getsockname() and getpeername()
6876 newinet
->inet_sport
= inet
->inet_sport
;
6877 newinet
->inet_saddr
= inet
->inet_saddr
;
6878 newinet
->inet_rcv_saddr
= inet
->inet_rcv_saddr
;
6879 newinet
->inet_dport
= htons(asoc
->peer
.port
);
6880 newinet
->pmtudisc
= inet
->pmtudisc
;
6881 newinet
->inet_id
= asoc
->next_tsn
^ jiffies
;
6883 newinet
->uc_ttl
= inet
->uc_ttl
;
6884 newinet
->mc_loop
= 1;
6885 newinet
->mc_ttl
= 1;
6886 newinet
->mc_index
= 0;
6887 newinet
->mc_list
= NULL
;
6890 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6891 * and its messages to the newsk.
6893 static void sctp_sock_migrate(struct sock
*oldsk
, struct sock
*newsk
,
6894 struct sctp_association
*assoc
,
6895 sctp_socket_type_t type
)
6897 struct sctp_sock
*oldsp
= sctp_sk(oldsk
);
6898 struct sctp_sock
*newsp
= sctp_sk(newsk
);
6899 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
6900 struct sctp_endpoint
*newep
= newsp
->ep
;
6901 struct sk_buff
*skb
, *tmp
;
6902 struct sctp_ulpevent
*event
;
6903 struct sctp_bind_hashbucket
*head
;
6904 struct list_head tmplist
;
6906 /* Migrate socket buffer sizes and all the socket level options to the
6909 newsk
->sk_sndbuf
= oldsk
->sk_sndbuf
;
6910 newsk
->sk_rcvbuf
= oldsk
->sk_rcvbuf
;
6911 /* Brute force copy old sctp opt. */
6912 if (oldsp
->do_auto_asconf
) {
6913 memcpy(&tmplist
, &newsp
->auto_asconf_list
, sizeof(tmplist
));
6914 inet_sk_copy_descendant(newsk
, oldsk
);
6915 memcpy(&newsp
->auto_asconf_list
, &tmplist
, sizeof(tmplist
));
6917 inet_sk_copy_descendant(newsk
, oldsk
);
6919 /* Restore the ep value that was overwritten with the above structure
6925 /* Hook this new socket in to the bind_hash list. */
6926 head
= &sctp_port_hashtable
[sctp_phashfn(sock_net(oldsk
),
6927 inet_sk(oldsk
)->inet_num
)];
6928 sctp_local_bh_disable();
6929 sctp_spin_lock(&head
->lock
);
6930 pp
= sctp_sk(oldsk
)->bind_hash
;
6931 sk_add_bind_node(newsk
, &pp
->owner
);
6932 sctp_sk(newsk
)->bind_hash
= pp
;
6933 inet_sk(newsk
)->inet_num
= inet_sk(oldsk
)->inet_num
;
6934 sctp_spin_unlock(&head
->lock
);
6935 sctp_local_bh_enable();
6937 /* Copy the bind_addr list from the original endpoint to the new
6938 * endpoint so that we can handle restarts properly
6940 sctp_bind_addr_dup(&newsp
->ep
->base
.bind_addr
,
6941 &oldsp
->ep
->base
.bind_addr
, GFP_KERNEL
);
6943 /* Move any messages in the old socket's receive queue that are for the
6944 * peeled off association to the new socket's receive queue.
6946 sctp_skb_for_each(skb
, &oldsk
->sk_receive_queue
, tmp
) {
6947 event
= sctp_skb2event(skb
);
6948 if (event
->asoc
== assoc
) {
6949 __skb_unlink(skb
, &oldsk
->sk_receive_queue
);
6950 __skb_queue_tail(&newsk
->sk_receive_queue
, skb
);
6951 sctp_skb_set_owner_r_frag(skb
, newsk
);
6955 /* Clean up any messages pending delivery due to partial
6956 * delivery. Three cases:
6957 * 1) No partial deliver; no work.
6958 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6959 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6961 skb_queue_head_init(&newsp
->pd_lobby
);
6962 atomic_set(&sctp_sk(newsk
)->pd_mode
, assoc
->ulpq
.pd_mode
);
6964 if (atomic_read(&sctp_sk(oldsk
)->pd_mode
)) {
6965 struct sk_buff_head
*queue
;
6967 /* Decide which queue to move pd_lobby skbs to. */
6968 if (assoc
->ulpq
.pd_mode
) {
6969 queue
= &newsp
->pd_lobby
;
6971 queue
= &newsk
->sk_receive_queue
;
6973 /* Walk through the pd_lobby, looking for skbs that
6974 * need moved to the new socket.
6976 sctp_skb_for_each(skb
, &oldsp
->pd_lobby
, tmp
) {
6977 event
= sctp_skb2event(skb
);
6978 if (event
->asoc
== assoc
) {
6979 __skb_unlink(skb
, &oldsp
->pd_lobby
);
6980 __skb_queue_tail(queue
, skb
);
6981 sctp_skb_set_owner_r_frag(skb
, newsk
);
6985 /* Clear up any skbs waiting for the partial
6986 * delivery to finish.
6988 if (assoc
->ulpq
.pd_mode
)
6989 sctp_clear_pd(oldsk
, NULL
);
6993 sctp_skb_for_each(skb
, &assoc
->ulpq
.reasm
, tmp
)
6994 sctp_skb_set_owner_r_frag(skb
, newsk
);
6996 sctp_skb_for_each(skb
, &assoc
->ulpq
.lobby
, tmp
)
6997 sctp_skb_set_owner_r_frag(skb
, newsk
);
6999 /* Set the type of socket to indicate that it is peeled off from the
7000 * original UDP-style socket or created with the accept() call on a
7001 * TCP-style socket..
7005 /* Mark the new socket "in-use" by the user so that any packets
7006 * that may arrive on the association after we've moved it are
7007 * queued to the backlog. This prevents a potential race between
7008 * backlog processing on the old socket and new-packet processing
7009 * on the new socket.
7011 * The caller has just allocated newsk so we can guarantee that other
7012 * paths won't try to lock it and then oldsk.
7014 lock_sock_nested(newsk
, SINGLE_DEPTH_NESTING
);
7015 sctp_assoc_migrate(assoc
, newsk
);
7017 /* If the association on the newsk is already closed before accept()
7018 * is called, set RCV_SHUTDOWN flag.
7020 if (sctp_state(assoc
, CLOSED
) && sctp_style(newsk
, TCP
))
7021 newsk
->sk_shutdown
|= RCV_SHUTDOWN
;
7023 newsk
->sk_state
= SCTP_SS_ESTABLISHED
;
7024 sctp_release_sock(newsk
);
7028 /* This proto struct describes the ULP interface for SCTP. */
7029 struct proto sctp_prot
= {
7031 .owner
= THIS_MODULE
,
7032 .close
= sctp_close
,
7033 .connect
= sctp_connect
,
7034 .disconnect
= sctp_disconnect
,
7035 .accept
= sctp_accept
,
7036 .ioctl
= sctp_ioctl
,
7037 .init
= sctp_init_sock
,
7038 .destroy
= sctp_destroy_sock
,
7039 .shutdown
= sctp_shutdown
,
7040 .setsockopt
= sctp_setsockopt
,
7041 .getsockopt
= sctp_getsockopt
,
7042 .sendmsg
= sctp_sendmsg
,
7043 .recvmsg
= sctp_recvmsg
,
7045 .backlog_rcv
= sctp_backlog_rcv
,
7047 .unhash
= sctp_unhash
,
7048 .get_port
= sctp_get_port
,
7049 .obj_size
= sizeof(struct sctp_sock
),
7050 .sysctl_mem
= sysctl_sctp_mem
,
7051 .sysctl_rmem
= sysctl_sctp_rmem
,
7052 .sysctl_wmem
= sysctl_sctp_wmem
,
7053 .memory_pressure
= &sctp_memory_pressure
,
7054 .enter_memory_pressure
= sctp_enter_memory_pressure
,
7055 .memory_allocated
= &sctp_memory_allocated
,
7056 .sockets_allocated
= &sctp_sockets_allocated
,
7059 #if IS_ENABLED(CONFIG_IPV6)
7061 struct proto sctpv6_prot
= {
7063 .owner
= THIS_MODULE
,
7064 .close
= sctp_close
,
7065 .connect
= sctp_connect
,
7066 .disconnect
= sctp_disconnect
,
7067 .accept
= sctp_accept
,
7068 .ioctl
= sctp_ioctl
,
7069 .init
= sctp_init_sock
,
7070 .destroy
= sctp_destroy_sock
,
7071 .shutdown
= sctp_shutdown
,
7072 .setsockopt
= sctp_setsockopt
,
7073 .getsockopt
= sctp_getsockopt
,
7074 .sendmsg
= sctp_sendmsg
,
7075 .recvmsg
= sctp_recvmsg
,
7077 .backlog_rcv
= sctp_backlog_rcv
,
7079 .unhash
= sctp_unhash
,
7080 .get_port
= sctp_get_port
,
7081 .obj_size
= sizeof(struct sctp6_sock
),
7082 .sysctl_mem
= sysctl_sctp_mem
,
7083 .sysctl_rmem
= sysctl_sctp_rmem
,
7084 .sysctl_wmem
= sysctl_sctp_wmem
,
7085 .memory_pressure
= &sctp_memory_pressure
,
7086 .enter_memory_pressure
= sctp_enter_memory_pressure
,
7087 .memory_allocated
= &sctp_memory_allocated
,
7088 .sockets_allocated
= &sctp_sockets_allocated
,
7090 #endif /* IS_ENABLED(CONFIG_IPV6) */