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>
76 #include <net/route.h>
78 #include <net/inet_common.h>
80 #include <linux/socket.h> /* for sa_family_t */
82 #include <net/sctp/sctp.h>
83 #include <net/sctp/sm.h>
85 /* WARNING: Please do not remove the SCTP_STATIC attribute to
86 * any of the functions below as they are used to export functions
87 * used by a project regression testsuite.
90 /* Forward declarations for internal helper functions. */
91 static int sctp_writeable(struct sock
*sk
);
92 static void sctp_wfree(struct sk_buff
*skb
);
93 static int sctp_wait_for_sndbuf(struct sctp_association
*, long *timeo_p
,
95 static int sctp_wait_for_packet(struct sock
* sk
, int *err
, long *timeo_p
);
96 static int sctp_wait_for_connect(struct sctp_association
*, long *timeo_p
);
97 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
);
98 static void sctp_wait_for_close(struct sock
*sk
, long timeo
);
99 static struct sctp_af
*sctp_sockaddr_af(struct sctp_sock
*opt
,
100 union sctp_addr
*addr
, int len
);
101 static int sctp_bindx_add(struct sock
*, struct sockaddr
*, int);
102 static int sctp_bindx_rem(struct sock
*, struct sockaddr
*, int);
103 static int sctp_send_asconf_add_ip(struct sock
*, struct sockaddr
*, int);
104 static int sctp_send_asconf_del_ip(struct sock
*, struct sockaddr
*, int);
105 static int sctp_send_asconf(struct sctp_association
*asoc
,
106 struct sctp_chunk
*chunk
);
107 static int sctp_do_bind(struct sock
*, union sctp_addr
*, int);
108 static int sctp_autobind(struct sock
*sk
);
109 static void sctp_sock_migrate(struct sock
*, struct sock
*,
110 struct sctp_association
*, sctp_socket_type_t
);
111 static char *sctp_hmac_alg
= SCTP_COOKIE_HMAC_ALG
;
113 extern struct kmem_cache
*sctp_bucket_cachep
;
114 extern long sysctl_sctp_mem
[3];
115 extern int sysctl_sctp_rmem
[3];
116 extern int sysctl_sctp_wmem
[3];
118 static int sctp_memory_pressure
;
119 static atomic_long_t sctp_memory_allocated
;
120 struct percpu_counter sctp_sockets_allocated
;
122 static void sctp_enter_memory_pressure(struct sock
*sk
)
124 sctp_memory_pressure
= 1;
128 /* Get the sndbuf space available at the time on the association. */
129 static inline int sctp_wspace(struct sctp_association
*asoc
)
133 if (asoc
->ep
->sndbuf_policy
)
134 amt
= asoc
->sndbuf_used
;
136 amt
= sk_wmem_alloc_get(asoc
->base
.sk
);
138 if (amt
>= asoc
->base
.sk
->sk_sndbuf
) {
139 if (asoc
->base
.sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)
142 amt
= sk_stream_wspace(asoc
->base
.sk
);
147 amt
= asoc
->base
.sk
->sk_sndbuf
- amt
;
152 /* Increment the used sndbuf space count of the corresponding association by
153 * the size of the outgoing data chunk.
154 * Also, set the skb destructor for sndbuf accounting later.
156 * Since it is always 1-1 between chunk and skb, and also a new skb is always
157 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
158 * destructor in the data chunk skb for the purpose of the sndbuf space
161 static inline void sctp_set_owner_w(struct sctp_chunk
*chunk
)
163 struct sctp_association
*asoc
= chunk
->asoc
;
164 struct sock
*sk
= asoc
->base
.sk
;
166 /* The sndbuf space is tracked per association. */
167 sctp_association_hold(asoc
);
169 skb_set_owner_w(chunk
->skb
, sk
);
171 chunk
->skb
->destructor
= sctp_wfree
;
172 /* Save the chunk pointer in skb for sctp_wfree to use later. */
173 *((struct sctp_chunk
**)(chunk
->skb
->cb
)) = chunk
;
175 asoc
->sndbuf_used
+= SCTP_DATA_SNDSIZE(chunk
) +
176 sizeof(struct sk_buff
) +
177 sizeof(struct sctp_chunk
);
179 atomic_add(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
180 sk
->sk_wmem_queued
+= chunk
->skb
->truesize
;
181 sk_mem_charge(sk
, chunk
->skb
->truesize
);
184 /* Verify that this is a valid address. */
185 static inline int sctp_verify_addr(struct sock
*sk
, union sctp_addr
*addr
,
190 /* Verify basic sockaddr. */
191 af
= sctp_sockaddr_af(sctp_sk(sk
), addr
, len
);
195 /* Is this a valid SCTP address? */
196 if (!af
->addr_valid(addr
, sctp_sk(sk
), NULL
))
199 if (!sctp_sk(sk
)->pf
->send_verify(sctp_sk(sk
), (addr
)))
205 /* Look up the association by its id. If this is not a UDP-style
206 * socket, the ID field is always ignored.
208 struct sctp_association
*sctp_id2assoc(struct sock
*sk
, sctp_assoc_t id
)
210 struct sctp_association
*asoc
= NULL
;
212 /* If this is not a UDP-style socket, assoc id should be ignored. */
213 if (!sctp_style(sk
, UDP
)) {
214 /* Return NULL if the socket state is not ESTABLISHED. It
215 * could be a TCP-style listening socket or a socket which
216 * hasn't yet called connect() to establish an association.
218 if (!sctp_sstate(sk
, ESTABLISHED
))
221 /* Get the first and the only association from the list. */
222 if (!list_empty(&sctp_sk(sk
)->ep
->asocs
))
223 asoc
= list_entry(sctp_sk(sk
)->ep
->asocs
.next
,
224 struct sctp_association
, asocs
);
228 /* Otherwise this is a UDP-style socket. */
229 if (!id
|| (id
== (sctp_assoc_t
)-1))
232 spin_lock_bh(&sctp_assocs_id_lock
);
233 asoc
= (struct sctp_association
*)idr_find(&sctp_assocs_id
, (int)id
);
234 spin_unlock_bh(&sctp_assocs_id_lock
);
236 if (!asoc
|| (asoc
->base
.sk
!= sk
) || asoc
->base
.dead
)
242 /* Look up the transport from an address and an assoc id. If both address and
243 * id are specified, the associations matching the address and the id should be
246 static struct sctp_transport
*sctp_addr_id2transport(struct sock
*sk
,
247 struct sockaddr_storage
*addr
,
250 struct sctp_association
*addr_asoc
= NULL
, *id_asoc
= NULL
;
251 struct sctp_transport
*transport
;
252 union sctp_addr
*laddr
= (union sctp_addr
*)addr
;
254 addr_asoc
= sctp_endpoint_lookup_assoc(sctp_sk(sk
)->ep
,
261 id_asoc
= sctp_id2assoc(sk
, id
);
262 if (id_asoc
&& (id_asoc
!= addr_asoc
))
265 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
266 (union sctp_addr
*)addr
);
271 /* API 3.1.2 bind() - UDP Style Syntax
272 * The syntax of bind() is,
274 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
276 * sd - the socket descriptor returned by socket().
277 * addr - the address structure (struct sockaddr_in or struct
278 * sockaddr_in6 [RFC 2553]),
279 * addr_len - the size of the address structure.
281 SCTP_STATIC
int sctp_bind(struct sock
*sk
, struct sockaddr
*addr
, int addr_len
)
287 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
290 /* Disallow binding twice. */
291 if (!sctp_sk(sk
)->ep
->base
.bind_addr
.port
)
292 retval
= sctp_do_bind(sk
, (union sctp_addr
*)addr
,
297 sctp_release_sock(sk
);
302 static long sctp_get_port_local(struct sock
*, union sctp_addr
*);
304 /* Verify this is a valid sockaddr. */
305 static struct sctp_af
*sctp_sockaddr_af(struct sctp_sock
*opt
,
306 union sctp_addr
*addr
, int len
)
310 /* Check minimum size. */
311 if (len
< sizeof (struct sockaddr
))
314 /* V4 mapped address are really of AF_INET family */
315 if (addr
->sa
.sa_family
== AF_INET6
&&
316 ipv6_addr_v4mapped(&addr
->v6
.sin6_addr
)) {
317 if (!opt
->pf
->af_supported(AF_INET
, opt
))
320 /* Does this PF support this AF? */
321 if (!opt
->pf
->af_supported(addr
->sa
.sa_family
, opt
))
325 /* If we get this far, af is valid. */
326 af
= sctp_get_af_specific(addr
->sa
.sa_family
);
328 if (len
< af
->sockaddr_len
)
334 /* Bind a local address either to an endpoint or to an association. */
335 SCTP_STATIC
int sctp_do_bind(struct sock
*sk
, union sctp_addr
*addr
, int len
)
337 struct sctp_sock
*sp
= sctp_sk(sk
);
338 struct sctp_endpoint
*ep
= sp
->ep
;
339 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
344 /* Common sockaddr verification. */
345 af
= sctp_sockaddr_af(sp
, addr
, len
);
347 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
352 snum
= ntohs(addr
->v4
.sin_port
);
354 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
355 ", port: %d, new port: %d, len: %d)\n",
361 /* PF specific bind() address verification. */
362 if (!sp
->pf
->bind_verify(sp
, addr
))
363 return -EADDRNOTAVAIL
;
365 /* We must either be unbound, or bind to the same port.
366 * It's OK to allow 0 ports if we are already bound.
367 * We'll just inhert an already bound port in this case
372 else if (snum
!= bp
->port
) {
373 SCTP_DEBUG_PRINTK("sctp_do_bind:"
374 " New port %d does not match existing port "
375 "%d.\n", snum
, bp
->port
);
380 if (snum
&& snum
< PROT_SOCK
&& !capable(CAP_NET_BIND_SERVICE
))
383 /* See if the address matches any of the addresses we may have
384 * already bound before checking against other endpoints.
386 if (sctp_bind_addr_match(bp
, addr
, sp
))
389 /* Make sure we are allowed to bind here.
390 * The function sctp_get_port_local() does duplicate address
393 addr
->v4
.sin_port
= htons(snum
);
394 if ((ret
= sctp_get_port_local(sk
, addr
))) {
398 /* Refresh ephemeral port. */
400 bp
->port
= inet_sk(sk
)->inet_num
;
402 /* Add the address to the bind address list.
403 * Use GFP_ATOMIC since BHs will be disabled.
405 ret
= sctp_add_bind_addr(bp
, addr
, SCTP_ADDR_SRC
, GFP_ATOMIC
);
407 /* Copy back into socket for getsockname() use. */
409 inet_sk(sk
)->inet_sport
= htons(inet_sk(sk
)->inet_num
);
410 af
->to_sk_saddr(addr
, sk
);
416 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
418 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
419 * at any one time. If a sender, after sending an ASCONF chunk, decides
420 * it needs to transfer another ASCONF Chunk, it MUST wait until the
421 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
422 * subsequent ASCONF. Note this restriction binds each side, so at any
423 * time two ASCONF may be in-transit on any given association (one sent
424 * from each endpoint).
426 static int sctp_send_asconf(struct sctp_association
*asoc
,
427 struct sctp_chunk
*chunk
)
431 /* If there is an outstanding ASCONF chunk, queue it for later
434 if (asoc
->addip_last_asconf
) {
435 list_add_tail(&chunk
->list
, &asoc
->addip_chunk_list
);
439 /* Hold the chunk until an ASCONF_ACK is received. */
440 sctp_chunk_hold(chunk
);
441 retval
= sctp_primitive_ASCONF(asoc
, chunk
);
443 sctp_chunk_free(chunk
);
445 asoc
->addip_last_asconf
= chunk
;
451 /* Add a list of addresses as bind addresses to local endpoint or
454 * Basically run through each address specified in the addrs/addrcnt
455 * array/length pair, determine if it is IPv6 or IPv4 and call
456 * sctp_do_bind() on it.
458 * If any of them fails, then the operation will be reversed and the
459 * ones that were added will be removed.
461 * Only sctp_setsockopt_bindx() is supposed to call this function.
463 static int sctp_bindx_add(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
468 struct sockaddr
*sa_addr
;
471 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
475 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
476 /* The list may contain either IPv4 or IPv6 address;
477 * determine the address length for walking thru the list.
479 sa_addr
= (struct sockaddr
*)addr_buf
;
480 af
= sctp_get_af_specific(sa_addr
->sa_family
);
486 retval
= sctp_do_bind(sk
, (union sctp_addr
*)sa_addr
,
489 addr_buf
+= af
->sockaddr_len
;
493 /* Failed. Cleanup the ones that have been added */
495 sctp_bindx_rem(sk
, addrs
, cnt
);
503 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
504 * associations that are part of the endpoint indicating that a list of local
505 * addresses are added to the endpoint.
507 * If any of the addresses is already in the bind address list of the
508 * association, we do not send the chunk for that association. But it will not
509 * affect other associations.
511 * Only sctp_setsockopt_bindx() is supposed to call this function.
513 static int sctp_send_asconf_add_ip(struct sock
*sk
,
514 struct sockaddr
*addrs
,
517 struct sctp_sock
*sp
;
518 struct sctp_endpoint
*ep
;
519 struct sctp_association
*asoc
;
520 struct sctp_bind_addr
*bp
;
521 struct sctp_chunk
*chunk
;
522 struct sctp_sockaddr_entry
*laddr
;
523 union sctp_addr
*addr
;
524 union sctp_addr saveaddr
;
531 if (!sctp_addip_enable
)
537 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
538 __func__
, sk
, addrs
, addrcnt
);
540 list_for_each_entry(asoc
, &ep
->asocs
, asocs
) {
542 if (!asoc
->peer
.asconf_capable
)
545 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_ADD_IP
)
548 if (!sctp_state(asoc
, ESTABLISHED
))
551 /* Check if any address in the packed array of addresses is
552 * in the bind address list of the association. If so,
553 * do not send the asconf chunk to its peer, but continue with
554 * other associations.
557 for (i
= 0; i
< addrcnt
; i
++) {
558 addr
= (union sctp_addr
*)addr_buf
;
559 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
565 if (sctp_assoc_lookup_laddr(asoc
, addr
))
568 addr_buf
+= af
->sockaddr_len
;
573 /* Use the first valid address in bind addr list of
574 * association as Address Parameter of ASCONF CHUNK.
576 bp
= &asoc
->base
.bind_addr
;
577 p
= bp
->address_list
.next
;
578 laddr
= list_entry(p
, struct sctp_sockaddr_entry
, list
);
579 chunk
= sctp_make_asconf_update_ip(asoc
, &laddr
->a
, addrs
,
580 addrcnt
, SCTP_PARAM_ADD_IP
);
586 /* Add the new addresses to the bind address list with
587 * use_as_src set to 0.
590 for (i
= 0; i
< addrcnt
; i
++) {
591 addr
= (union sctp_addr
*)addr_buf
;
592 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
593 memcpy(&saveaddr
, addr
, af
->sockaddr_len
);
594 retval
= sctp_add_bind_addr(bp
, &saveaddr
,
595 SCTP_ADDR_NEW
, GFP_ATOMIC
);
596 addr_buf
+= af
->sockaddr_len
;
598 if (asoc
->src_out_of_asoc_ok
) {
599 struct sctp_transport
*trans
;
601 list_for_each_entry(trans
,
602 &asoc
->peer
.transport_addr_list
, transports
) {
603 /* Clear the source and route cache */
604 dst_release(trans
->dst
);
605 trans
->cwnd
= min(4*asoc
->pathmtu
, max_t(__u32
,
606 2*asoc
->pathmtu
, 4380));
607 trans
->ssthresh
= asoc
->peer
.i
.a_rwnd
;
608 trans
->rto
= asoc
->rto_initial
;
609 trans
->rtt
= trans
->srtt
= trans
->rttvar
= 0;
610 sctp_transport_route(trans
, NULL
,
611 sctp_sk(asoc
->base
.sk
));
614 retval
= sctp_send_asconf(asoc
, chunk
);
621 /* Remove a list of addresses from bind addresses list. Do not remove the
624 * Basically run through each address specified in the addrs/addrcnt
625 * array/length pair, determine if it is IPv6 or IPv4 and call
626 * sctp_del_bind() on it.
628 * If any of them fails, then the operation will be reversed and the
629 * ones that were removed will be added back.
631 * At least one address has to be left; if only one address is
632 * available, the operation will return -EBUSY.
634 * Only sctp_setsockopt_bindx() is supposed to call this function.
636 static int sctp_bindx_rem(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
638 struct sctp_sock
*sp
= sctp_sk(sk
);
639 struct sctp_endpoint
*ep
= sp
->ep
;
641 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
644 union sctp_addr
*sa_addr
;
647 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
651 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
652 /* If the bind address list is empty or if there is only one
653 * bind address, there is nothing more to be removed (we need
654 * at least one address here).
656 if (list_empty(&bp
->address_list
) ||
657 (sctp_list_single_entry(&bp
->address_list
))) {
662 sa_addr
= (union sctp_addr
*)addr_buf
;
663 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
669 if (!af
->addr_valid(sa_addr
, sp
, NULL
)) {
670 retval
= -EADDRNOTAVAIL
;
674 if (sa_addr
->v4
.sin_port
&&
675 sa_addr
->v4
.sin_port
!= htons(bp
->port
)) {
680 if (!sa_addr
->v4
.sin_port
)
681 sa_addr
->v4
.sin_port
= htons(bp
->port
);
683 /* FIXME - There is probably a need to check if sk->sk_saddr and
684 * sk->sk_rcv_addr are currently set to one of the addresses to
685 * be removed. This is something which needs to be looked into
686 * when we are fixing the outstanding issues with multi-homing
687 * socket routing and failover schemes. Refer to comments in
688 * sctp_do_bind(). -daisy
690 retval
= sctp_del_bind_addr(bp
, sa_addr
);
692 addr_buf
+= af
->sockaddr_len
;
695 /* Failed. Add the ones that has been removed back */
697 sctp_bindx_add(sk
, addrs
, cnt
);
705 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
706 * the associations that are part of the endpoint indicating that a list of
707 * local addresses are removed from the endpoint.
709 * If any of the addresses is already in the bind address list of the
710 * association, we do not send the chunk for that association. But it will not
711 * affect other associations.
713 * Only sctp_setsockopt_bindx() is supposed to call this function.
715 static int sctp_send_asconf_del_ip(struct sock
*sk
,
716 struct sockaddr
*addrs
,
719 struct sctp_sock
*sp
;
720 struct sctp_endpoint
*ep
;
721 struct sctp_association
*asoc
;
722 struct sctp_transport
*transport
;
723 struct sctp_bind_addr
*bp
;
724 struct sctp_chunk
*chunk
;
725 union sctp_addr
*laddr
;
728 struct sctp_sockaddr_entry
*saddr
;
734 if (!sctp_addip_enable
)
740 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
741 __func__
, sk
, addrs
, addrcnt
);
743 list_for_each_entry(asoc
, &ep
->asocs
, asocs
) {
745 if (!asoc
->peer
.asconf_capable
)
748 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_DEL_IP
)
751 if (!sctp_state(asoc
, ESTABLISHED
))
754 /* Check if any address in the packed array of addresses is
755 * not present in the bind address list of the association.
756 * If so, do not send the asconf chunk to its peer, but
757 * continue with other associations.
760 for (i
= 0; i
< addrcnt
; i
++) {
761 laddr
= (union sctp_addr
*)addr_buf
;
762 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
768 if (!sctp_assoc_lookup_laddr(asoc
, laddr
))
771 addr_buf
+= af
->sockaddr_len
;
776 /* Find one address in the association's bind address list
777 * that is not in the packed array of addresses. This is to
778 * make sure that we do not delete all the addresses in the
781 bp
= &asoc
->base
.bind_addr
;
782 laddr
= sctp_find_unmatch_addr(bp
, (union sctp_addr
*)addrs
,
784 if ((laddr
== NULL
) && (addrcnt
== 1)) {
785 if (asoc
->asconf_addr_del_pending
)
787 asoc
->asconf_addr_del_pending
=
788 kzalloc(sizeof(union sctp_addr
), GFP_ATOMIC
);
789 asoc
->asconf_addr_del_pending
->sa
.sa_family
=
791 asoc
->asconf_addr_del_pending
->v4
.sin_port
=
793 if (addrs
->sa_family
== AF_INET
) {
794 struct sockaddr_in
*sin
;
796 sin
= (struct sockaddr_in
*)addrs
;
797 asoc
->asconf_addr_del_pending
->v4
.sin_addr
.s_addr
= sin
->sin_addr
.s_addr
;
798 } else if (addrs
->sa_family
== AF_INET6
) {
799 struct sockaddr_in6
*sin6
;
801 sin6
= (struct sockaddr_in6
*)addrs
;
802 ipv6_addr_copy(&asoc
->asconf_addr_del_pending
->v6
.sin6_addr
, &sin6
->sin6_addr
);
804 SCTP_DEBUG_PRINTK_IPADDR("send_asconf_del_ip: keep the last address asoc: %p ",
805 " at %p\n", asoc
, asoc
->asconf_addr_del_pending
,
806 asoc
->asconf_addr_del_pending
);
807 asoc
->src_out_of_asoc_ok
= 1;
812 /* We do not need RCU protection throughout this loop
813 * because this is done under a socket lock from the
816 chunk
= sctp_make_asconf_update_ip(asoc
, laddr
, addrs
, addrcnt
,
824 /* Reset use_as_src flag for the addresses in the bind address
825 * list that are to be deleted.
828 for (i
= 0; i
< addrcnt
; i
++) {
829 laddr
= (union sctp_addr
*)addr_buf
;
830 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
831 list_for_each_entry(saddr
, &bp
->address_list
, list
) {
832 if (sctp_cmp_addr_exact(&saddr
->a
, laddr
))
833 saddr
->state
= SCTP_ADDR_DEL
;
835 addr_buf
+= af
->sockaddr_len
;
838 /* Update the route and saddr entries for all the transports
839 * as some of the addresses in the bind address list are
840 * about to be deleted and cannot be used as source addresses.
842 list_for_each_entry(transport
, &asoc
->peer
.transport_addr_list
,
844 dst_release(transport
->dst
);
845 sctp_transport_route(transport
, NULL
,
846 sctp_sk(asoc
->base
.sk
));
850 /* We don't need to transmit ASCONF */
852 retval
= sctp_send_asconf(asoc
, chunk
);
858 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
859 int sctp_asconf_mgmt(struct sctp_sock
*sp
, struct sctp_sockaddr_entry
*addrw
)
861 struct sock
*sk
= sctp_opt2sk(sp
);
862 union sctp_addr
*addr
;
865 /* It is safe to write port space in caller. */
867 addr
->v4
.sin_port
= htons(sp
->ep
->base
.bind_addr
.port
);
868 af
= sctp_get_af_specific(addr
->sa
.sa_family
);
871 if (sctp_verify_addr(sk
, addr
, af
->sockaddr_len
))
874 if (addrw
->state
== SCTP_ADDR_NEW
)
875 return sctp_send_asconf_add_ip(sk
, (struct sockaddr
*)addr
, 1);
877 return sctp_send_asconf_del_ip(sk
, (struct sockaddr
*)addr
, 1);
880 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
883 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
886 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
887 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
890 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
891 * Section 3.1.2 for this usage.
893 * addrs is a pointer to an array of one or more socket addresses. Each
894 * address is contained in its appropriate structure (i.e. struct
895 * sockaddr_in or struct sockaddr_in6) the family of the address type
896 * must be used to distinguish the address length (note that this
897 * representation is termed a "packed array" of addresses). The caller
898 * specifies the number of addresses in the array with addrcnt.
900 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
901 * -1, and sets errno to the appropriate error code.
903 * For SCTP, the port given in each socket address must be the same, or
904 * sctp_bindx() will fail, setting errno to EINVAL.
906 * The flags parameter is formed from the bitwise OR of zero or more of
907 * the following currently defined flags:
909 * SCTP_BINDX_ADD_ADDR
911 * SCTP_BINDX_REM_ADDR
913 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
914 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
915 * addresses from the association. The two flags are mutually exclusive;
916 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
917 * not remove all addresses from an association; sctp_bindx() will
918 * reject such an attempt with EINVAL.
920 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
921 * additional addresses with an endpoint after calling bind(). Or use
922 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
923 * socket is associated with so that no new association accepted will be
924 * associated with those addresses. If the endpoint supports dynamic
925 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
926 * endpoint to send the appropriate message to the peer to change the
927 * peers address lists.
929 * Adding and removing addresses from a connected association is
930 * optional functionality. Implementations that do not support this
931 * functionality should return EOPNOTSUPP.
933 * Basically do nothing but copying the addresses from user to kernel
934 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
935 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
938 * We don't use copy_from_user() for optimization: we first do the
939 * sanity checks (buffer size -fast- and access check-healthy
940 * pointer); if all of those succeed, then we can alloc the memory
941 * (expensive operation) needed to copy the data to kernel. Then we do
942 * the copying without checking the user space area
943 * (__copy_from_user()).
945 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
948 * sk The sk of the socket
949 * addrs The pointer to the addresses in user land
950 * addrssize Size of the addrs buffer
951 * op Operation to perform (add or remove, see the flags of
954 * Returns 0 if ok, <0 errno code on error.
956 SCTP_STATIC
int sctp_setsockopt_bindx(struct sock
* sk
,
957 struct sockaddr __user
*addrs
,
958 int addrs_size
, int op
)
960 struct sockaddr
*kaddrs
;
964 struct sockaddr
*sa_addr
;
968 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
969 " addrs_size %d opt %d\n", sk
, addrs
, addrs_size
, op
);
971 if (unlikely(addrs_size
<= 0))
974 /* Check the user passed a healthy pointer. */
975 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
978 /* Alloc space for the address array in kernel memory. */
979 kaddrs
= kmalloc(addrs_size
, GFP_KERNEL
);
980 if (unlikely(!kaddrs
))
983 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
988 /* Walk through the addrs buffer and count the number of addresses. */
990 while (walk_size
< addrs_size
) {
991 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
996 sa_addr
= (struct sockaddr
*)addr_buf
;
997 af
= sctp_get_af_specific(sa_addr
->sa_family
);
999 /* If the address family is not supported or if this address
1000 * causes the address buffer to overflow return EINVAL.
1002 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
1007 addr_buf
+= af
->sockaddr_len
;
1008 walk_size
+= af
->sockaddr_len
;
1013 case SCTP_BINDX_ADD_ADDR
:
1014 err
= sctp_bindx_add(sk
, kaddrs
, addrcnt
);
1017 err
= sctp_send_asconf_add_ip(sk
, kaddrs
, addrcnt
);
1020 case SCTP_BINDX_REM_ADDR
:
1021 err
= sctp_bindx_rem(sk
, kaddrs
, addrcnt
);
1024 err
= sctp_send_asconf_del_ip(sk
, kaddrs
, addrcnt
);
1038 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1040 * Common routine for handling connect() and sctp_connectx().
1041 * Connect will come in with just a single address.
1043 static int __sctp_connect(struct sock
* sk
,
1044 struct sockaddr
*kaddrs
,
1046 sctp_assoc_t
*assoc_id
)
1048 struct sctp_sock
*sp
;
1049 struct sctp_endpoint
*ep
;
1050 struct sctp_association
*asoc
= NULL
;
1051 struct sctp_association
*asoc2
;
1052 struct sctp_transport
*transport
;
1060 union sctp_addr
*sa_addr
= NULL
;
1062 unsigned short port
;
1063 unsigned int f_flags
= 0;
1068 /* connect() cannot be done on a socket that is already in ESTABLISHED
1069 * state - UDP-style peeled off socket or a TCP-style socket that
1070 * is already connected.
1071 * It cannot be done even on a TCP-style listening socket.
1073 if (sctp_sstate(sk
, ESTABLISHED
) ||
1074 (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))) {
1079 /* Walk through the addrs buffer and count the number of addresses. */
1081 while (walk_size
< addrs_size
) {
1082 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
1087 sa_addr
= (union sctp_addr
*)addr_buf
;
1088 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
1090 /* If the address family is not supported or if this address
1091 * causes the address buffer to overflow return EINVAL.
1093 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
1098 port
= ntohs(sa_addr
->v4
.sin_port
);
1100 /* Save current address so we can work with it */
1101 memcpy(&to
, sa_addr
, af
->sockaddr_len
);
1103 err
= sctp_verify_addr(sk
, &to
, af
->sockaddr_len
);
1107 /* Make sure the destination port is correctly set
1110 if (asoc
&& asoc
->peer
.port
&& asoc
->peer
.port
!= port
)
1114 /* Check if there already is a matching association on the
1115 * endpoint (other than the one created here).
1117 asoc2
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1118 if (asoc2
&& asoc2
!= asoc
) {
1119 if (asoc2
->state
>= SCTP_STATE_ESTABLISHED
)
1126 /* If we could not find a matching association on the endpoint,
1127 * make sure that there is no peeled-off association matching
1128 * the peer address even on another socket.
1130 if (sctp_endpoint_is_peeled_off(ep
, &to
)) {
1131 err
= -EADDRNOTAVAIL
;
1136 /* If a bind() or sctp_bindx() is not called prior to
1137 * an sctp_connectx() call, the system picks an
1138 * ephemeral port and will choose an address set
1139 * equivalent to binding with a wildcard address.
1141 if (!ep
->base
.bind_addr
.port
) {
1142 if (sctp_autobind(sk
)) {
1148 * If an unprivileged user inherits a 1-many
1149 * style socket with open associations on a
1150 * privileged port, it MAY be permitted to
1151 * accept new associations, but it SHOULD NOT
1152 * be permitted to open new associations.
1154 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1155 !capable(CAP_NET_BIND_SERVICE
)) {
1161 scope
= sctp_scope(&to
);
1162 asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1168 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
,
1176 /* Prime the peer's transport structures. */
1177 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
,
1185 addr_buf
+= af
->sockaddr_len
;
1186 walk_size
+= af
->sockaddr_len
;
1189 /* In case the user of sctp_connectx() wants an association
1190 * id back, assign one now.
1193 err
= sctp_assoc_set_id(asoc
, GFP_KERNEL
);
1198 err
= sctp_primitive_ASSOCIATE(asoc
, NULL
);
1203 /* Initialize sk's dport and daddr for getpeername() */
1204 inet_sk(sk
)->inet_dport
= htons(asoc
->peer
.port
);
1205 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
1206 af
->to_sk_daddr(sa_addr
, sk
);
1209 /* in-kernel sockets don't generally have a file allocated to them
1210 * if all they do is call sock_create_kern().
1212 if (sk
->sk_socket
->file
)
1213 f_flags
= sk
->sk_socket
->file
->f_flags
;
1215 timeo
= sock_sndtimeo(sk
, f_flags
& O_NONBLOCK
);
1217 err
= sctp_wait_for_connect(asoc
, &timeo
);
1218 if ((err
== 0 || err
== -EINPROGRESS
) && assoc_id
)
1219 *assoc_id
= asoc
->assoc_id
;
1221 /* Don't free association on exit. */
1226 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1227 " kaddrs: %p err: %d\n",
1230 sctp_association_free(asoc
);
1234 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1237 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1238 * sctp_assoc_t *asoc);
1240 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1241 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1242 * or IPv6 addresses.
1244 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1245 * Section 3.1.2 for this usage.
1247 * addrs is a pointer to an array of one or more socket addresses. Each
1248 * address is contained in its appropriate structure (i.e. struct
1249 * sockaddr_in or struct sockaddr_in6) the family of the address type
1250 * must be used to distengish the address length (note that this
1251 * representation is termed a "packed array" of addresses). The caller
1252 * specifies the number of addresses in the array with addrcnt.
1254 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1255 * the association id of the new association. On failure, sctp_connectx()
1256 * returns -1, and sets errno to the appropriate error code. The assoc_id
1257 * is not touched by the kernel.
1259 * For SCTP, the port given in each socket address must be the same, or
1260 * sctp_connectx() will fail, setting errno to EINVAL.
1262 * An application can use sctp_connectx to initiate an association with
1263 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1264 * allows a caller to specify multiple addresses at which a peer can be
1265 * reached. The way the SCTP stack uses the list of addresses to set up
1266 * the association is implementation dependent. This function only
1267 * specifies that the stack will try to make use of all the addresses in
1268 * the list when needed.
1270 * Note that the list of addresses passed in is only used for setting up
1271 * the association. It does not necessarily equal the set of addresses
1272 * the peer uses for the resulting association. If the caller wants to
1273 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1274 * retrieve them after the association has been set up.
1276 * Basically do nothing but copying the addresses from user to kernel
1277 * land and invoking either sctp_connectx(). This is used for tunneling
1278 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1280 * We don't use copy_from_user() for optimization: we first do the
1281 * sanity checks (buffer size -fast- and access check-healthy
1282 * pointer); if all of those succeed, then we can alloc the memory
1283 * (expensive operation) needed to copy the data to kernel. Then we do
1284 * the copying without checking the user space area
1285 * (__copy_from_user()).
1287 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1290 * sk The sk of the socket
1291 * addrs The pointer to the addresses in user land
1292 * addrssize Size of the addrs buffer
1294 * Returns >=0 if ok, <0 errno code on error.
1296 SCTP_STATIC
int __sctp_setsockopt_connectx(struct sock
* sk
,
1297 struct sockaddr __user
*addrs
,
1299 sctp_assoc_t
*assoc_id
)
1302 struct sockaddr
*kaddrs
;
1304 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1305 __func__
, sk
, addrs
, addrs_size
);
1307 if (unlikely(addrs_size
<= 0))
1310 /* Check the user passed a healthy pointer. */
1311 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
1314 /* Alloc space for the address array in kernel memory. */
1315 kaddrs
= kmalloc(addrs_size
, GFP_KERNEL
);
1316 if (unlikely(!kaddrs
))
1319 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
1322 err
= __sctp_connect(sk
, kaddrs
, addrs_size
, assoc_id
);
1331 * This is an older interface. It's kept for backward compatibility
1332 * to the option that doesn't provide association id.
1334 SCTP_STATIC
int sctp_setsockopt_connectx_old(struct sock
* sk
,
1335 struct sockaddr __user
*addrs
,
1338 return __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, NULL
);
1342 * New interface for the API. The since the API is done with a socket
1343 * option, to make it simple we feed back the association id is as a return
1344 * indication to the call. Error is always negative and association id is
1347 SCTP_STATIC
int sctp_setsockopt_connectx(struct sock
* sk
,
1348 struct sockaddr __user
*addrs
,
1351 sctp_assoc_t assoc_id
= 0;
1354 err
= __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, &assoc_id
);
1363 * New (hopefully final) interface for the API.
1364 * We use the sctp_getaddrs_old structure so that use-space library
1365 * can avoid any unnecessary allocations. The only defferent part
1366 * is that we store the actual length of the address buffer into the
1367 * addrs_num structure member. That way we can re-use the existing
1370 SCTP_STATIC
int sctp_getsockopt_connectx3(struct sock
* sk
, int len
,
1371 char __user
*optval
,
1374 struct sctp_getaddrs_old param
;
1375 sctp_assoc_t assoc_id
= 0;
1378 if (len
< sizeof(param
))
1381 if (copy_from_user(¶m
, optval
, sizeof(param
)))
1384 err
= __sctp_setsockopt_connectx(sk
,
1385 (struct sockaddr __user
*)param
.addrs
,
1386 param
.addr_num
, &assoc_id
);
1388 if (err
== 0 || err
== -EINPROGRESS
) {
1389 if (copy_to_user(optval
, &assoc_id
, sizeof(assoc_id
)))
1391 if (put_user(sizeof(assoc_id
), optlen
))
1398 /* API 3.1.4 close() - UDP Style Syntax
1399 * Applications use close() to perform graceful shutdown (as described in
1400 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1401 * by a UDP-style socket.
1405 * ret = close(int sd);
1407 * sd - the socket descriptor of the associations to be closed.
1409 * To gracefully shutdown a specific association represented by the
1410 * UDP-style socket, an application should use the sendmsg() call,
1411 * passing no user data, but including the appropriate flag in the
1412 * ancillary data (see Section xxxx).
1414 * If sd in the close() call is a branched-off socket representing only
1415 * one association, the shutdown is performed on that association only.
1417 * 4.1.6 close() - TCP Style Syntax
1419 * Applications use close() to gracefully close down an association.
1423 * int close(int sd);
1425 * sd - the socket descriptor of the association to be closed.
1427 * After an application calls close() on a socket descriptor, no further
1428 * socket operations will succeed on that descriptor.
1430 * API 7.1.4 SO_LINGER
1432 * An application using the TCP-style socket can use this option to
1433 * perform the SCTP ABORT primitive. The linger option structure is:
1436 * int l_onoff; // option on/off
1437 * int l_linger; // linger time
1440 * To enable the option, set l_onoff to 1. If the l_linger value is set
1441 * to 0, calling close() is the same as the ABORT primitive. If the
1442 * value is set to a negative value, the setsockopt() call will return
1443 * an error. If the value is set to a positive value linger_time, the
1444 * close() can be blocked for at most linger_time ms. If the graceful
1445 * shutdown phase does not finish during this period, close() will
1446 * return but the graceful shutdown phase continues in the system.
1448 SCTP_STATIC
void sctp_close(struct sock
*sk
, long timeout
)
1450 struct sctp_endpoint
*ep
;
1451 struct sctp_association
*asoc
;
1452 struct list_head
*pos
, *temp
;
1454 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk
, timeout
);
1457 sk
->sk_shutdown
= SHUTDOWN_MASK
;
1458 sk
->sk_state
= SCTP_SS_CLOSING
;
1460 ep
= sctp_sk(sk
)->ep
;
1462 /* Walk all associations on an endpoint. */
1463 list_for_each_safe(pos
, temp
, &ep
->asocs
) {
1464 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
1466 if (sctp_style(sk
, TCP
)) {
1467 /* A closed association can still be in the list if
1468 * it belongs to a TCP-style listening socket that is
1469 * not yet accepted. If so, free it. If not, send an
1470 * ABORT or SHUTDOWN based on the linger options.
1472 if (sctp_state(asoc
, CLOSED
)) {
1473 sctp_unhash_established(asoc
);
1474 sctp_association_free(asoc
);
1479 if (sock_flag(sk
, SOCK_LINGER
) && !sk
->sk_lingertime
) {
1480 struct sctp_chunk
*chunk
;
1482 chunk
= sctp_make_abort_user(asoc
, NULL
, 0);
1484 sctp_primitive_ABORT(asoc
, chunk
);
1486 sctp_primitive_SHUTDOWN(asoc
, NULL
);
1489 /* Clean up any skbs sitting on the receive queue. */
1490 sctp_queue_purge_ulpevents(&sk
->sk_receive_queue
);
1491 sctp_queue_purge_ulpevents(&sctp_sk(sk
)->pd_lobby
);
1493 /* On a TCP-style socket, block for at most linger_time if set. */
1494 if (sctp_style(sk
, TCP
) && timeout
)
1495 sctp_wait_for_close(sk
, timeout
);
1497 /* This will run the backlog queue. */
1498 sctp_release_sock(sk
);
1500 /* Supposedly, no process has access to the socket, but
1501 * the net layers still may.
1503 sctp_local_bh_disable();
1504 sctp_bh_lock_sock(sk
);
1506 /* Hold the sock, since sk_common_release() will put sock_put()
1507 * and we have just a little more cleanup.
1510 sk_common_release(sk
);
1512 sctp_bh_unlock_sock(sk
);
1513 sctp_local_bh_enable();
1517 SCTP_DBG_OBJCNT_DEC(sock
);
1520 /* Handle EPIPE error. */
1521 static int sctp_error(struct sock
*sk
, int flags
, int err
)
1524 err
= sock_error(sk
) ? : -EPIPE
;
1525 if (err
== -EPIPE
&& !(flags
& MSG_NOSIGNAL
))
1526 send_sig(SIGPIPE
, current
, 0);
1530 /* API 3.1.3 sendmsg() - UDP Style Syntax
1532 * An application uses sendmsg() and recvmsg() calls to transmit data to
1533 * and receive data from its peer.
1535 * ssize_t sendmsg(int socket, const struct msghdr *message,
1538 * socket - the socket descriptor of the endpoint.
1539 * message - pointer to the msghdr structure which contains a single
1540 * user message and possibly some ancillary data.
1542 * See Section 5 for complete description of the data
1545 * flags - flags sent or received with the user message, see Section
1546 * 5 for complete description of the flags.
1548 * Note: This function could use a rewrite especially when explicit
1549 * connect support comes in.
1551 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1553 SCTP_STATIC
int sctp_msghdr_parse(const struct msghdr
*, sctp_cmsgs_t
*);
1555 SCTP_STATIC
int sctp_sendmsg(struct kiocb
*iocb
, struct sock
*sk
,
1556 struct msghdr
*msg
, size_t msg_len
)
1558 struct sctp_sock
*sp
;
1559 struct sctp_endpoint
*ep
;
1560 struct sctp_association
*new_asoc
=NULL
, *asoc
=NULL
;
1561 struct sctp_transport
*transport
, *chunk_tp
;
1562 struct sctp_chunk
*chunk
;
1564 struct sockaddr
*msg_name
= NULL
;
1565 struct sctp_sndrcvinfo default_sinfo
;
1566 struct sctp_sndrcvinfo
*sinfo
;
1567 struct sctp_initmsg
*sinit
;
1568 sctp_assoc_t associd
= 0;
1569 sctp_cmsgs_t cmsgs
= { NULL
};
1573 __u16 sinfo_flags
= 0;
1574 struct sctp_datamsg
*datamsg
;
1575 int msg_flags
= msg
->msg_flags
;
1577 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1584 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep
);
1586 /* We cannot send a message over a TCP-style listening socket. */
1587 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
)) {
1592 /* Parse out the SCTP CMSGs. */
1593 err
= sctp_msghdr_parse(msg
, &cmsgs
);
1596 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err
);
1600 /* Fetch the destination address for this packet. This
1601 * address only selects the association--it is not necessarily
1602 * the address we will send to.
1603 * For a peeled-off socket, msg_name is ignored.
1605 if (!sctp_style(sk
, UDP_HIGH_BANDWIDTH
) && msg
->msg_name
) {
1606 int msg_namelen
= msg
->msg_namelen
;
1608 err
= sctp_verify_addr(sk
, (union sctp_addr
*)msg
->msg_name
,
1613 if (msg_namelen
> sizeof(to
))
1614 msg_namelen
= sizeof(to
);
1615 memcpy(&to
, msg
->msg_name
, msg_namelen
);
1616 msg_name
= msg
->msg_name
;
1622 /* Did the user specify SNDRCVINFO? */
1624 sinfo_flags
= sinfo
->sinfo_flags
;
1625 associd
= sinfo
->sinfo_assoc_id
;
1628 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1629 msg_len
, sinfo_flags
);
1631 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1632 if (sctp_style(sk
, TCP
) && (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
))) {
1637 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1638 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1639 * If SCTP_ABORT is set, the message length could be non zero with
1640 * the msg_iov set to the user abort reason.
1642 if (((sinfo_flags
& SCTP_EOF
) && (msg_len
> 0)) ||
1643 (!(sinfo_flags
& (SCTP_EOF
|SCTP_ABORT
)) && (msg_len
== 0))) {
1648 /* If SCTP_ADDR_OVER is set, there must be an address
1649 * specified in msg_name.
1651 if ((sinfo_flags
& SCTP_ADDR_OVER
) && (!msg
->msg_name
)) {
1658 SCTP_DEBUG_PRINTK("About to look up association.\n");
1662 /* If a msg_name has been specified, assume this is to be used. */
1664 /* Look for a matching association on the endpoint. */
1665 asoc
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1667 /* If we could not find a matching association on the
1668 * endpoint, make sure that it is not a TCP-style
1669 * socket that already has an association or there is
1670 * no peeled-off association on another socket.
1672 if ((sctp_style(sk
, TCP
) &&
1673 sctp_sstate(sk
, ESTABLISHED
)) ||
1674 sctp_endpoint_is_peeled_off(ep
, &to
)) {
1675 err
= -EADDRNOTAVAIL
;
1680 asoc
= sctp_id2assoc(sk
, associd
);
1688 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc
);
1690 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1691 * socket that has an association in CLOSED state. This can
1692 * happen when an accepted socket has an association that is
1695 if (sctp_state(asoc
, CLOSED
) && sctp_style(sk
, TCP
)) {
1700 if (sinfo_flags
& SCTP_EOF
) {
1701 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1703 sctp_primitive_SHUTDOWN(asoc
, NULL
);
1707 if (sinfo_flags
& SCTP_ABORT
) {
1709 chunk
= sctp_make_abort_user(asoc
, msg
, msg_len
);
1715 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc
);
1716 sctp_primitive_ABORT(asoc
, chunk
);
1722 /* Do we need to create the association? */
1724 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1726 if (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
)) {
1731 /* Check for invalid stream against the stream counts,
1732 * either the default or the user specified stream counts.
1735 if (!sinit
|| (sinit
&& !sinit
->sinit_num_ostreams
)) {
1736 /* Check against the defaults. */
1737 if (sinfo
->sinfo_stream
>=
1738 sp
->initmsg
.sinit_num_ostreams
) {
1743 /* Check against the requested. */
1744 if (sinfo
->sinfo_stream
>=
1745 sinit
->sinit_num_ostreams
) {
1753 * API 3.1.2 bind() - UDP Style Syntax
1754 * If a bind() or sctp_bindx() is not called prior to a
1755 * sendmsg() call that initiates a new association, the
1756 * system picks an ephemeral port and will choose an address
1757 * set equivalent to binding with a wildcard address.
1759 if (!ep
->base
.bind_addr
.port
) {
1760 if (sctp_autobind(sk
)) {
1766 * If an unprivileged user inherits a one-to-many
1767 * style socket with open associations on a privileged
1768 * port, it MAY be permitted to accept new associations,
1769 * but it SHOULD NOT be permitted to open new
1772 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1773 !capable(CAP_NET_BIND_SERVICE
)) {
1779 scope
= sctp_scope(&to
);
1780 new_asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1786 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
, GFP_KERNEL
);
1792 /* If the SCTP_INIT ancillary data is specified, set all
1793 * the association init values accordingly.
1796 if (sinit
->sinit_num_ostreams
) {
1797 asoc
->c
.sinit_num_ostreams
=
1798 sinit
->sinit_num_ostreams
;
1800 if (sinit
->sinit_max_instreams
) {
1801 asoc
->c
.sinit_max_instreams
=
1802 sinit
->sinit_max_instreams
;
1804 if (sinit
->sinit_max_attempts
) {
1805 asoc
->max_init_attempts
1806 = sinit
->sinit_max_attempts
;
1808 if (sinit
->sinit_max_init_timeo
) {
1809 asoc
->max_init_timeo
=
1810 msecs_to_jiffies(sinit
->sinit_max_init_timeo
);
1814 /* Prime the peer's transport structures. */
1815 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
, SCTP_UNKNOWN
);
1822 /* ASSERT: we have a valid association at this point. */
1823 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1826 /* If the user didn't specify SNDRCVINFO, make up one with
1829 memset(&default_sinfo
, 0, sizeof(default_sinfo
));
1830 default_sinfo
.sinfo_stream
= asoc
->default_stream
;
1831 default_sinfo
.sinfo_flags
= asoc
->default_flags
;
1832 default_sinfo
.sinfo_ppid
= asoc
->default_ppid
;
1833 default_sinfo
.sinfo_context
= asoc
->default_context
;
1834 default_sinfo
.sinfo_timetolive
= asoc
->default_timetolive
;
1835 default_sinfo
.sinfo_assoc_id
= sctp_assoc2id(asoc
);
1836 sinfo
= &default_sinfo
;
1839 /* API 7.1.7, the sndbuf size per association bounds the
1840 * maximum size of data that can be sent in a single send call.
1842 if (msg_len
> sk
->sk_sndbuf
) {
1847 if (asoc
->pmtu_pending
)
1848 sctp_assoc_pending_pmtu(asoc
);
1850 /* If fragmentation is disabled and the message length exceeds the
1851 * association fragmentation point, return EMSGSIZE. The I-D
1852 * does not specify what this error is, but this looks like
1855 if (sctp_sk(sk
)->disable_fragments
&& (msg_len
> asoc
->frag_point
)) {
1860 /* Check for invalid stream. */
1861 if (sinfo
->sinfo_stream
>= asoc
->c
.sinit_num_ostreams
) {
1866 timeo
= sock_sndtimeo(sk
, msg
->msg_flags
& MSG_DONTWAIT
);
1867 if (!sctp_wspace(asoc
)) {
1868 err
= sctp_wait_for_sndbuf(asoc
, &timeo
, msg_len
);
1873 /* If an address is passed with the sendto/sendmsg call, it is used
1874 * to override the primary destination address in the TCP model, or
1875 * when SCTP_ADDR_OVER flag is set in the UDP model.
1877 if ((sctp_style(sk
, TCP
) && msg_name
) ||
1878 (sinfo_flags
& SCTP_ADDR_OVER
)) {
1879 chunk_tp
= sctp_assoc_lookup_paddr(asoc
, &to
);
1887 /* Auto-connect, if we aren't connected already. */
1888 if (sctp_state(asoc
, CLOSED
)) {
1889 err
= sctp_primitive_ASSOCIATE(asoc
, NULL
);
1892 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1895 /* Break the message into multiple chunks of maximum size. */
1896 datamsg
= sctp_datamsg_from_user(asoc
, sinfo
, msg
, msg_len
);
1902 /* Now send the (possibly) fragmented message. */
1903 list_for_each_entry(chunk
, &datamsg
->chunks
, frag_list
) {
1904 sctp_chunk_hold(chunk
);
1906 /* Do accounting for the write space. */
1907 sctp_set_owner_w(chunk
);
1909 chunk
->transport
= chunk_tp
;
1912 /* Send it to the lower layers. Note: all chunks
1913 * must either fail or succeed. The lower layer
1914 * works that way today. Keep it that way or this
1917 err
= sctp_primitive_SEND(asoc
, datamsg
);
1918 /* Did the lower layer accept the chunk? */
1920 sctp_datamsg_free(datamsg
);
1922 sctp_datamsg_put(datamsg
);
1924 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1931 /* If we are already past ASSOCIATE, the lower
1932 * layers are responsible for association cleanup.
1938 sctp_association_free(asoc
);
1940 sctp_release_sock(sk
);
1943 return sctp_error(sk
, msg_flags
, err
);
1950 err
= sock_error(sk
);
1960 /* This is an extended version of skb_pull() that removes the data from the
1961 * start of a skb even when data is spread across the list of skb's in the
1962 * frag_list. len specifies the total amount of data that needs to be removed.
1963 * when 'len' bytes could be removed from the skb, it returns 0.
1964 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1965 * could not be removed.
1967 static int sctp_skb_pull(struct sk_buff
*skb
, int len
)
1969 struct sk_buff
*list
;
1970 int skb_len
= skb_headlen(skb
);
1973 if (len
<= skb_len
) {
1974 __skb_pull(skb
, len
);
1978 __skb_pull(skb
, skb_len
);
1980 skb_walk_frags(skb
, list
) {
1981 rlen
= sctp_skb_pull(list
, len
);
1982 skb
->len
-= (len
-rlen
);
1983 skb
->data_len
-= (len
-rlen
);
1994 /* API 3.1.3 recvmsg() - UDP Style Syntax
1996 * ssize_t recvmsg(int socket, struct msghdr *message,
1999 * socket - the socket descriptor of the endpoint.
2000 * message - pointer to the msghdr structure which contains a single
2001 * user message and possibly some ancillary data.
2003 * See Section 5 for complete description of the data
2006 * flags - flags sent or received with the user message, see Section
2007 * 5 for complete description of the flags.
2009 static struct sk_buff
*sctp_skb_recv_datagram(struct sock
*, int, int, int *);
2011 SCTP_STATIC
int sctp_recvmsg(struct kiocb
*iocb
, struct sock
*sk
,
2012 struct msghdr
*msg
, size_t len
, int noblock
,
2013 int flags
, int *addr_len
)
2015 struct sctp_ulpevent
*event
= NULL
;
2016 struct sctp_sock
*sp
= sctp_sk(sk
);
2017 struct sk_buff
*skb
;
2022 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
2023 "0x%x, %s: %p)\n", "sk", sk
, "msghdr", msg
,
2024 "len", len
, "knoblauch", noblock
,
2025 "flags", flags
, "addr_len", addr_len
);
2029 if (sctp_style(sk
, TCP
) && !sctp_sstate(sk
, ESTABLISHED
)) {
2034 skb
= sctp_skb_recv_datagram(sk
, flags
, noblock
, &err
);
2038 /* Get the total length of the skb including any skb's in the
2047 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
2049 event
= sctp_skb2event(skb
);
2054 sock_recv_ts_and_drops(msg
, sk
, skb
);
2055 if (sctp_ulpevent_is_notification(event
)) {
2056 msg
->msg_flags
|= MSG_NOTIFICATION
;
2057 sp
->pf
->event_msgname(event
, msg
->msg_name
, addr_len
);
2059 sp
->pf
->skb_msgname(skb
, msg
->msg_name
, addr_len
);
2062 /* Check if we allow SCTP_SNDRCVINFO. */
2063 if (sp
->subscribe
.sctp_data_io_event
)
2064 sctp_ulpevent_read_sndrcvinfo(event
, msg
);
2066 /* FIXME: we should be calling IP/IPv6 layers. */
2067 if (sk
->sk_protinfo
.af_inet
.cmsg_flags
)
2068 ip_cmsg_recv(msg
, skb
);
2073 /* If skb's length exceeds the user's buffer, update the skb and
2074 * push it back to the receive_queue so that the next call to
2075 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2077 if (skb_len
> copied
) {
2078 msg
->msg_flags
&= ~MSG_EOR
;
2079 if (flags
& MSG_PEEK
)
2081 sctp_skb_pull(skb
, copied
);
2082 skb_queue_head(&sk
->sk_receive_queue
, skb
);
2084 /* When only partial message is copied to the user, increase
2085 * rwnd by that amount. If all the data in the skb is read,
2086 * rwnd is updated when the event is freed.
2088 if (!sctp_ulpevent_is_notification(event
))
2089 sctp_assoc_rwnd_increase(event
->asoc
, copied
);
2091 } else if ((event
->msg_flags
& MSG_NOTIFICATION
) ||
2092 (event
->msg_flags
& MSG_EOR
))
2093 msg
->msg_flags
|= MSG_EOR
;
2095 msg
->msg_flags
&= ~MSG_EOR
;
2098 if (flags
& MSG_PEEK
) {
2099 /* Release the skb reference acquired after peeking the skb in
2100 * sctp_skb_recv_datagram().
2104 /* Free the event which includes releasing the reference to
2105 * the owner of the skb, freeing the skb and updating the
2108 sctp_ulpevent_free(event
);
2111 sctp_release_sock(sk
);
2115 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2117 * This option is a on/off flag. If enabled no SCTP message
2118 * fragmentation will be performed. Instead if a message being sent
2119 * exceeds the current PMTU size, the message will NOT be sent and
2120 * instead a error will be indicated to the user.
2122 static int sctp_setsockopt_disable_fragments(struct sock
*sk
,
2123 char __user
*optval
,
2124 unsigned int optlen
)
2128 if (optlen
< sizeof(int))
2131 if (get_user(val
, (int __user
*)optval
))
2134 sctp_sk(sk
)->disable_fragments
= (val
== 0) ? 0 : 1;
2139 static int sctp_setsockopt_events(struct sock
*sk
, char __user
*optval
,
2140 unsigned int optlen
)
2142 if (optlen
> sizeof(struct sctp_event_subscribe
))
2144 if (copy_from_user(&sctp_sk(sk
)->subscribe
, optval
, optlen
))
2149 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2151 * This socket option is applicable to the UDP-style socket only. When
2152 * set it will cause associations that are idle for more than the
2153 * specified number of seconds to automatically close. An association
2154 * being idle is defined an association that has NOT sent or received
2155 * user data. The special value of '0' indicates that no automatic
2156 * close of any associations should be performed. The option expects an
2157 * integer defining the number of seconds of idle time before an
2158 * association is closed.
2160 static int sctp_setsockopt_autoclose(struct sock
*sk
, char __user
*optval
,
2161 unsigned int optlen
)
2163 struct sctp_sock
*sp
= sctp_sk(sk
);
2165 /* Applicable to UDP-style socket only */
2166 if (sctp_style(sk
, TCP
))
2168 if (optlen
!= sizeof(int))
2170 if (copy_from_user(&sp
->autoclose
, optval
, optlen
))
2172 /* make sure it won't exceed MAX_SCHEDULE_TIMEOUT */
2173 sp
->autoclose
= min_t(long, sp
->autoclose
, MAX_SCHEDULE_TIMEOUT
/ HZ
);
2178 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2180 * Applications can enable or disable heartbeats for any peer address of
2181 * an association, modify an address's heartbeat interval, force a
2182 * heartbeat to be sent immediately, and adjust the address's maximum
2183 * number of retransmissions sent before an address is considered
2184 * unreachable. The following structure is used to access and modify an
2185 * address's parameters:
2187 * struct sctp_paddrparams {
2188 * sctp_assoc_t spp_assoc_id;
2189 * struct sockaddr_storage spp_address;
2190 * uint32_t spp_hbinterval;
2191 * uint16_t spp_pathmaxrxt;
2192 * uint32_t spp_pathmtu;
2193 * uint32_t spp_sackdelay;
2194 * uint32_t spp_flags;
2197 * spp_assoc_id - (one-to-many style socket) This is filled in the
2198 * application, and identifies the association for
2200 * spp_address - This specifies which address is of interest.
2201 * spp_hbinterval - This contains the value of the heartbeat interval,
2202 * in milliseconds. If a value of zero
2203 * is present in this field then no changes are to
2204 * be made to this parameter.
2205 * spp_pathmaxrxt - This contains the maximum number of
2206 * retransmissions before this address shall be
2207 * considered unreachable. If a value of zero
2208 * is present in this field then no changes are to
2209 * be made to this parameter.
2210 * spp_pathmtu - When Path MTU discovery is disabled the value
2211 * specified here will be the "fixed" path mtu.
2212 * Note that if the spp_address field is empty
2213 * then all associations on this address will
2214 * have this fixed path mtu set upon them.
2216 * spp_sackdelay - When delayed sack is enabled, this value specifies
2217 * the number of milliseconds that sacks will be delayed
2218 * for. This value will apply to all addresses of an
2219 * association if the spp_address field is empty. Note
2220 * also, that if delayed sack is enabled and this
2221 * value is set to 0, no change is made to the last
2222 * recorded delayed sack timer value.
2224 * spp_flags - These flags are used to control various features
2225 * on an association. The flag field may contain
2226 * zero or more of the following options.
2228 * SPP_HB_ENABLE - Enable heartbeats on the
2229 * specified address. Note that if the address
2230 * field is empty all addresses for the association
2231 * have heartbeats enabled upon them.
2233 * SPP_HB_DISABLE - Disable heartbeats on the
2234 * speicifed address. Note that if the address
2235 * field is empty all addresses for the association
2236 * will have their heartbeats disabled. Note also
2237 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2238 * mutually exclusive, only one of these two should
2239 * be specified. Enabling both fields will have
2240 * undetermined results.
2242 * SPP_HB_DEMAND - Request a user initiated heartbeat
2243 * to be made immediately.
2245 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2246 * heartbeat delayis to be set to the value of 0
2249 * SPP_PMTUD_ENABLE - This field will enable PMTU
2250 * discovery upon the specified address. Note that
2251 * if the address feild is empty then all addresses
2252 * on the association are effected.
2254 * SPP_PMTUD_DISABLE - This field will disable PMTU
2255 * discovery upon the specified address. Note that
2256 * if the address feild is empty then all addresses
2257 * on the association are effected. Not also that
2258 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2259 * exclusive. Enabling both will have undetermined
2262 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2263 * on delayed sack. The time specified in spp_sackdelay
2264 * is used to specify the sack delay for this address. Note
2265 * that if spp_address is empty then all addresses will
2266 * enable delayed sack and take on the sack delay
2267 * value specified in spp_sackdelay.
2268 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2269 * off delayed sack. If the spp_address field is blank then
2270 * delayed sack is disabled for the entire association. Note
2271 * also that this field is mutually exclusive to
2272 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2275 static int sctp_apply_peer_addr_params(struct sctp_paddrparams
*params
,
2276 struct sctp_transport
*trans
,
2277 struct sctp_association
*asoc
,
2278 struct sctp_sock
*sp
,
2281 int sackdelay_change
)
2285 if (params
->spp_flags
& SPP_HB_DEMAND
&& trans
) {
2286 error
= sctp_primitive_REQUESTHEARTBEAT (trans
->asoc
, trans
);
2291 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2292 * this field is ignored. Note also that a value of zero indicates
2293 * the current setting should be left unchanged.
2295 if (params
->spp_flags
& SPP_HB_ENABLE
) {
2297 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2298 * set. This lets us use 0 value when this flag
2301 if (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)
2302 params
->spp_hbinterval
= 0;
2304 if (params
->spp_hbinterval
||
2305 (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)) {
2308 msecs_to_jiffies(params
->spp_hbinterval
);
2311 msecs_to_jiffies(params
->spp_hbinterval
);
2313 sp
->hbinterval
= params
->spp_hbinterval
;
2320 trans
->param_flags
=
2321 (trans
->param_flags
& ~SPP_HB
) | hb_change
;
2324 (asoc
->param_flags
& ~SPP_HB
) | hb_change
;
2327 (sp
->param_flags
& ~SPP_HB
) | hb_change
;
2331 /* When Path MTU discovery is disabled the value specified here will
2332 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2333 * include the flag SPP_PMTUD_DISABLE for this field to have any
2336 if ((params
->spp_flags
& SPP_PMTUD_DISABLE
) && params
->spp_pathmtu
) {
2338 trans
->pathmtu
= params
->spp_pathmtu
;
2339 sctp_assoc_sync_pmtu(asoc
);
2341 asoc
->pathmtu
= params
->spp_pathmtu
;
2342 sctp_frag_point(asoc
, params
->spp_pathmtu
);
2344 sp
->pathmtu
= params
->spp_pathmtu
;
2350 int update
= (trans
->param_flags
& SPP_PMTUD_DISABLE
) &&
2351 (params
->spp_flags
& SPP_PMTUD_ENABLE
);
2352 trans
->param_flags
=
2353 (trans
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2355 sctp_transport_pmtu(trans
, sctp_opt2sk(sp
));
2356 sctp_assoc_sync_pmtu(asoc
);
2360 (asoc
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2363 (sp
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2367 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2368 * value of this field is ignored. Note also that a value of zero
2369 * indicates the current setting should be left unchanged.
2371 if ((params
->spp_flags
& SPP_SACKDELAY_ENABLE
) && params
->spp_sackdelay
) {
2374 msecs_to_jiffies(params
->spp_sackdelay
);
2377 msecs_to_jiffies(params
->spp_sackdelay
);
2379 sp
->sackdelay
= params
->spp_sackdelay
;
2383 if (sackdelay_change
) {
2385 trans
->param_flags
=
2386 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2390 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2394 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2399 /* Note that a value of zero indicates the current setting should be
2402 if (params
->spp_pathmaxrxt
) {
2404 trans
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2406 asoc
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2408 sp
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2415 static int sctp_setsockopt_peer_addr_params(struct sock
*sk
,
2416 char __user
*optval
,
2417 unsigned int optlen
)
2419 struct sctp_paddrparams params
;
2420 struct sctp_transport
*trans
= NULL
;
2421 struct sctp_association
*asoc
= NULL
;
2422 struct sctp_sock
*sp
= sctp_sk(sk
);
2424 int hb_change
, pmtud_change
, sackdelay_change
;
2426 if (optlen
!= sizeof(struct sctp_paddrparams
))
2429 if (copy_from_user(¶ms
, optval
, optlen
))
2432 /* Validate flags and value parameters. */
2433 hb_change
= params
.spp_flags
& SPP_HB
;
2434 pmtud_change
= params
.spp_flags
& SPP_PMTUD
;
2435 sackdelay_change
= params
.spp_flags
& SPP_SACKDELAY
;
2437 if (hb_change
== SPP_HB
||
2438 pmtud_change
== SPP_PMTUD
||
2439 sackdelay_change
== SPP_SACKDELAY
||
2440 params
.spp_sackdelay
> 500 ||
2441 (params
.spp_pathmtu
&&
2442 params
.spp_pathmtu
< SCTP_DEFAULT_MINSEGMENT
))
2445 /* If an address other than INADDR_ANY is specified, and
2446 * no transport is found, then the request is invalid.
2448 if (!sctp_is_any(sk
, ( union sctp_addr
*)¶ms
.spp_address
)) {
2449 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
2450 params
.spp_assoc_id
);
2455 /* Get association, if assoc_id != 0 and the socket is a one
2456 * to many style socket, and an association was not found, then
2457 * the id was invalid.
2459 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
2460 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
))
2463 /* Heartbeat demand can only be sent on a transport or
2464 * association, but not a socket.
2466 if (params
.spp_flags
& SPP_HB_DEMAND
&& !trans
&& !asoc
)
2469 /* Process parameters. */
2470 error
= sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2471 hb_change
, pmtud_change
,
2477 /* If changes are for association, also apply parameters to each
2480 if (!trans
&& asoc
) {
2481 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2483 sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2484 hb_change
, pmtud_change
,
2493 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2495 * This option will effect the way delayed acks are performed. This
2496 * option allows you to get or set the delayed ack time, in
2497 * milliseconds. It also allows changing the delayed ack frequency.
2498 * Changing the frequency to 1 disables the delayed sack algorithm. If
2499 * the assoc_id is 0, then this sets or gets the endpoints default
2500 * values. If the assoc_id field is non-zero, then the set or get
2501 * effects the specified association for the one to many model (the
2502 * assoc_id field is ignored by the one to one model). Note that if
2503 * sack_delay or sack_freq are 0 when setting this option, then the
2504 * current values will remain unchanged.
2506 * struct sctp_sack_info {
2507 * sctp_assoc_t sack_assoc_id;
2508 * uint32_t sack_delay;
2509 * uint32_t sack_freq;
2512 * sack_assoc_id - This parameter, indicates which association the user
2513 * is performing an action upon. Note that if this field's value is
2514 * zero then the endpoints default value is changed (effecting future
2515 * associations only).
2517 * sack_delay - This parameter contains the number of milliseconds that
2518 * the user is requesting the delayed ACK timer be set to. Note that
2519 * this value is defined in the standard to be between 200 and 500
2522 * sack_freq - This parameter contains the number of packets that must
2523 * be received before a sack is sent without waiting for the delay
2524 * timer to expire. The default value for this is 2, setting this
2525 * value to 1 will disable the delayed sack algorithm.
2528 static int sctp_setsockopt_delayed_ack(struct sock
*sk
,
2529 char __user
*optval
, unsigned int optlen
)
2531 struct sctp_sack_info params
;
2532 struct sctp_transport
*trans
= NULL
;
2533 struct sctp_association
*asoc
= NULL
;
2534 struct sctp_sock
*sp
= sctp_sk(sk
);
2536 if (optlen
== sizeof(struct sctp_sack_info
)) {
2537 if (copy_from_user(¶ms
, optval
, optlen
))
2540 if (params
.sack_delay
== 0 && params
.sack_freq
== 0)
2542 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
2543 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2544 pr_warn("Use struct sctp_sack_info instead\n");
2545 if (copy_from_user(¶ms
, optval
, optlen
))
2548 if (params
.sack_delay
== 0)
2549 params
.sack_freq
= 1;
2551 params
.sack_freq
= 0;
2555 /* Validate value parameter. */
2556 if (params
.sack_delay
> 500)
2559 /* Get association, if sack_assoc_id != 0 and the socket is a one
2560 * to many style socket, and an association was not found, then
2561 * the id was invalid.
2563 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
2564 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
2567 if (params
.sack_delay
) {
2570 msecs_to_jiffies(params
.sack_delay
);
2572 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2573 SPP_SACKDELAY_ENABLE
;
2575 sp
->sackdelay
= params
.sack_delay
;
2577 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2578 SPP_SACKDELAY_ENABLE
;
2582 if (params
.sack_freq
== 1) {
2585 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2586 SPP_SACKDELAY_DISABLE
;
2589 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2590 SPP_SACKDELAY_DISABLE
;
2592 } else if (params
.sack_freq
> 1) {
2594 asoc
->sackfreq
= params
.sack_freq
;
2596 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2597 SPP_SACKDELAY_ENABLE
;
2599 sp
->sackfreq
= params
.sack_freq
;
2601 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2602 SPP_SACKDELAY_ENABLE
;
2606 /* If change is for association, also apply to each transport. */
2608 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2610 if (params
.sack_delay
) {
2612 msecs_to_jiffies(params
.sack_delay
);
2613 trans
->param_flags
=
2614 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2615 SPP_SACKDELAY_ENABLE
;
2617 if (params
.sack_freq
== 1) {
2618 trans
->param_flags
=
2619 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2620 SPP_SACKDELAY_DISABLE
;
2621 } else if (params
.sack_freq
> 1) {
2622 trans
->sackfreq
= params
.sack_freq
;
2623 trans
->param_flags
=
2624 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2625 SPP_SACKDELAY_ENABLE
;
2633 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2635 * Applications can specify protocol parameters for the default association
2636 * initialization. The option name argument to setsockopt() and getsockopt()
2639 * Setting initialization parameters is effective only on an unconnected
2640 * socket (for UDP-style sockets only future associations are effected
2641 * by the change). With TCP-style sockets, this option is inherited by
2642 * sockets derived from a listener socket.
2644 static int sctp_setsockopt_initmsg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2646 struct sctp_initmsg sinit
;
2647 struct sctp_sock
*sp
= sctp_sk(sk
);
2649 if (optlen
!= sizeof(struct sctp_initmsg
))
2651 if (copy_from_user(&sinit
, optval
, optlen
))
2654 if (sinit
.sinit_num_ostreams
)
2655 sp
->initmsg
.sinit_num_ostreams
= sinit
.sinit_num_ostreams
;
2656 if (sinit
.sinit_max_instreams
)
2657 sp
->initmsg
.sinit_max_instreams
= sinit
.sinit_max_instreams
;
2658 if (sinit
.sinit_max_attempts
)
2659 sp
->initmsg
.sinit_max_attempts
= sinit
.sinit_max_attempts
;
2660 if (sinit
.sinit_max_init_timeo
)
2661 sp
->initmsg
.sinit_max_init_timeo
= sinit
.sinit_max_init_timeo
;
2667 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2669 * Applications that wish to use the sendto() system call may wish to
2670 * specify a default set of parameters that would normally be supplied
2671 * through the inclusion of ancillary data. This socket option allows
2672 * such an application to set the default sctp_sndrcvinfo structure.
2673 * The application that wishes to use this socket option simply passes
2674 * in to this call the sctp_sndrcvinfo structure defined in Section
2675 * 5.2.2) The input parameters accepted by this call include
2676 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2677 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2678 * to this call if the caller is using the UDP model.
2680 static int sctp_setsockopt_default_send_param(struct sock
*sk
,
2681 char __user
*optval
,
2682 unsigned int optlen
)
2684 struct sctp_sndrcvinfo info
;
2685 struct sctp_association
*asoc
;
2686 struct sctp_sock
*sp
= sctp_sk(sk
);
2688 if (optlen
!= sizeof(struct sctp_sndrcvinfo
))
2690 if (copy_from_user(&info
, optval
, optlen
))
2693 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
2694 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
2698 asoc
->default_stream
= info
.sinfo_stream
;
2699 asoc
->default_flags
= info
.sinfo_flags
;
2700 asoc
->default_ppid
= info
.sinfo_ppid
;
2701 asoc
->default_context
= info
.sinfo_context
;
2702 asoc
->default_timetolive
= info
.sinfo_timetolive
;
2704 sp
->default_stream
= info
.sinfo_stream
;
2705 sp
->default_flags
= info
.sinfo_flags
;
2706 sp
->default_ppid
= info
.sinfo_ppid
;
2707 sp
->default_context
= info
.sinfo_context
;
2708 sp
->default_timetolive
= info
.sinfo_timetolive
;
2714 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2716 * Requests that the local SCTP stack use the enclosed peer address as
2717 * the association primary. The enclosed address must be one of the
2718 * association peer's addresses.
2720 static int sctp_setsockopt_primary_addr(struct sock
*sk
, char __user
*optval
,
2721 unsigned int optlen
)
2723 struct sctp_prim prim
;
2724 struct sctp_transport
*trans
;
2726 if (optlen
!= sizeof(struct sctp_prim
))
2729 if (copy_from_user(&prim
, optval
, sizeof(struct sctp_prim
)))
2732 trans
= sctp_addr_id2transport(sk
, &prim
.ssp_addr
, prim
.ssp_assoc_id
);
2736 sctp_assoc_set_primary(trans
->asoc
, trans
);
2742 * 7.1.5 SCTP_NODELAY
2744 * Turn on/off any Nagle-like algorithm. This means that packets are
2745 * generally sent as soon as possible and no unnecessary delays are
2746 * introduced, at the cost of more packets in the network. Expects an
2747 * integer boolean flag.
2749 static int sctp_setsockopt_nodelay(struct sock
*sk
, char __user
*optval
,
2750 unsigned int optlen
)
2754 if (optlen
< sizeof(int))
2756 if (get_user(val
, (int __user
*)optval
))
2759 sctp_sk(sk
)->nodelay
= (val
== 0) ? 0 : 1;
2765 * 7.1.1 SCTP_RTOINFO
2767 * The protocol parameters used to initialize and bound retransmission
2768 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2769 * and modify these parameters.
2770 * All parameters are time values, in milliseconds. A value of 0, when
2771 * modifying the parameters, indicates that the current value should not
2775 static int sctp_setsockopt_rtoinfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2777 struct sctp_rtoinfo rtoinfo
;
2778 struct sctp_association
*asoc
;
2780 if (optlen
!= sizeof (struct sctp_rtoinfo
))
2783 if (copy_from_user(&rtoinfo
, optval
, optlen
))
2786 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
2788 /* Set the values to the specific association */
2789 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
2793 if (rtoinfo
.srto_initial
!= 0)
2795 msecs_to_jiffies(rtoinfo
.srto_initial
);
2796 if (rtoinfo
.srto_max
!= 0)
2797 asoc
->rto_max
= msecs_to_jiffies(rtoinfo
.srto_max
);
2798 if (rtoinfo
.srto_min
!= 0)
2799 asoc
->rto_min
= msecs_to_jiffies(rtoinfo
.srto_min
);
2801 /* If there is no association or the association-id = 0
2802 * set the values to the endpoint.
2804 struct sctp_sock
*sp
= sctp_sk(sk
);
2806 if (rtoinfo
.srto_initial
!= 0)
2807 sp
->rtoinfo
.srto_initial
= rtoinfo
.srto_initial
;
2808 if (rtoinfo
.srto_max
!= 0)
2809 sp
->rtoinfo
.srto_max
= rtoinfo
.srto_max
;
2810 if (rtoinfo
.srto_min
!= 0)
2811 sp
->rtoinfo
.srto_min
= rtoinfo
.srto_min
;
2819 * 7.1.2 SCTP_ASSOCINFO
2821 * This option is used to tune the maximum retransmission attempts
2822 * of the association.
2823 * Returns an error if the new association retransmission value is
2824 * greater than the sum of the retransmission value of the peer.
2825 * See [SCTP] for more information.
2828 static int sctp_setsockopt_associnfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2831 struct sctp_assocparams assocparams
;
2832 struct sctp_association
*asoc
;
2834 if (optlen
!= sizeof(struct sctp_assocparams
))
2836 if (copy_from_user(&assocparams
, optval
, optlen
))
2839 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
2841 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
2844 /* Set the values to the specific association */
2846 if (assocparams
.sasoc_asocmaxrxt
!= 0) {
2849 struct sctp_transport
*peer_addr
;
2851 list_for_each_entry(peer_addr
, &asoc
->peer
.transport_addr_list
,
2853 path_sum
+= peer_addr
->pathmaxrxt
;
2857 /* Only validate asocmaxrxt if we have more than
2858 * one path/transport. We do this because path
2859 * retransmissions are only counted when we have more
2863 assocparams
.sasoc_asocmaxrxt
> path_sum
)
2866 asoc
->max_retrans
= assocparams
.sasoc_asocmaxrxt
;
2869 if (assocparams
.sasoc_cookie_life
!= 0) {
2870 asoc
->cookie_life
.tv_sec
=
2871 assocparams
.sasoc_cookie_life
/ 1000;
2872 asoc
->cookie_life
.tv_usec
=
2873 (assocparams
.sasoc_cookie_life
% 1000)
2877 /* Set the values to the endpoint */
2878 struct sctp_sock
*sp
= sctp_sk(sk
);
2880 if (assocparams
.sasoc_asocmaxrxt
!= 0)
2881 sp
->assocparams
.sasoc_asocmaxrxt
=
2882 assocparams
.sasoc_asocmaxrxt
;
2883 if (assocparams
.sasoc_cookie_life
!= 0)
2884 sp
->assocparams
.sasoc_cookie_life
=
2885 assocparams
.sasoc_cookie_life
;
2891 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2893 * This socket option is a boolean flag which turns on or off mapped V4
2894 * addresses. If this option is turned on and the socket is type
2895 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2896 * If this option is turned off, then no mapping will be done of V4
2897 * addresses and a user will receive both PF_INET6 and PF_INET type
2898 * addresses on the socket.
2900 static int sctp_setsockopt_mappedv4(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2903 struct sctp_sock
*sp
= sctp_sk(sk
);
2905 if (optlen
< sizeof(int))
2907 if (get_user(val
, (int __user
*)optval
))
2918 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2919 * This option will get or set the maximum size to put in any outgoing
2920 * SCTP DATA chunk. If a message is larger than this size it will be
2921 * fragmented by SCTP into the specified size. Note that the underlying
2922 * SCTP implementation may fragment into smaller sized chunks when the
2923 * PMTU of the underlying association is smaller than the value set by
2924 * the user. The default value for this option is '0' which indicates
2925 * the user is NOT limiting fragmentation and only the PMTU will effect
2926 * SCTP's choice of DATA chunk size. Note also that values set larger
2927 * than the maximum size of an IP datagram will effectively let SCTP
2928 * control fragmentation (i.e. the same as setting this option to 0).
2930 * The following structure is used to access and modify this parameter:
2932 * struct sctp_assoc_value {
2933 * sctp_assoc_t assoc_id;
2934 * uint32_t assoc_value;
2937 * assoc_id: This parameter is ignored for one-to-one style sockets.
2938 * For one-to-many style sockets this parameter indicates which
2939 * association the user is performing an action upon. Note that if
2940 * this field's value is zero then the endpoints default value is
2941 * changed (effecting future associations only).
2942 * assoc_value: This parameter specifies the maximum size in bytes.
2944 static int sctp_setsockopt_maxseg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2946 struct sctp_assoc_value params
;
2947 struct sctp_association
*asoc
;
2948 struct sctp_sock
*sp
= sctp_sk(sk
);
2951 if (optlen
== sizeof(int)) {
2952 pr_warn("Use of int in maxseg socket option deprecated\n");
2953 pr_warn("Use struct sctp_assoc_value instead\n");
2954 if (copy_from_user(&val
, optval
, optlen
))
2956 params
.assoc_id
= 0;
2957 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
2958 if (copy_from_user(¶ms
, optval
, optlen
))
2960 val
= params
.assoc_value
;
2964 if ((val
!= 0) && ((val
< 8) || (val
> SCTP_MAX_CHUNK_LEN
)))
2967 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
2968 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
2973 val
= asoc
->pathmtu
;
2974 val
-= sp
->pf
->af
->net_header_len
;
2975 val
-= sizeof(struct sctphdr
) +
2976 sizeof(struct sctp_data_chunk
);
2978 asoc
->user_frag
= val
;
2979 asoc
->frag_point
= sctp_frag_point(asoc
, asoc
->pathmtu
);
2981 sp
->user_frag
= val
;
2989 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2991 * Requests that the peer mark the enclosed address as the association
2992 * primary. The enclosed address must be one of the association's
2993 * locally bound addresses. The following structure is used to make a
2994 * set primary request:
2996 static int sctp_setsockopt_peer_primary_addr(struct sock
*sk
, char __user
*optval
,
2997 unsigned int optlen
)
2999 struct sctp_sock
*sp
;
3000 struct sctp_association
*asoc
= NULL
;
3001 struct sctp_setpeerprim prim
;
3002 struct sctp_chunk
*chunk
;
3008 if (!sctp_addip_enable
)
3011 if (optlen
!= sizeof(struct sctp_setpeerprim
))
3014 if (copy_from_user(&prim
, optval
, optlen
))
3017 asoc
= sctp_id2assoc(sk
, prim
.sspp_assoc_id
);
3021 if (!asoc
->peer
.asconf_capable
)
3024 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_SET_PRIMARY
)
3027 if (!sctp_state(asoc
, ESTABLISHED
))
3030 af
= sctp_get_af_specific(prim
.sspp_addr
.ss_family
);
3034 if (!af
->addr_valid((union sctp_addr
*)&prim
.sspp_addr
, sp
, NULL
))
3035 return -EADDRNOTAVAIL
;
3037 if (!sctp_assoc_lookup_laddr(asoc
, (union sctp_addr
*)&prim
.sspp_addr
))
3038 return -EADDRNOTAVAIL
;
3040 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3041 chunk
= sctp_make_asconf_set_prim(asoc
,
3042 (union sctp_addr
*)&prim
.sspp_addr
);
3046 err
= sctp_send_asconf(asoc
, chunk
);
3048 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
3053 static int sctp_setsockopt_adaptation_layer(struct sock
*sk
, char __user
*optval
,
3054 unsigned int optlen
)
3056 struct sctp_setadaptation adaptation
;
3058 if (optlen
!= sizeof(struct sctp_setadaptation
))
3060 if (copy_from_user(&adaptation
, optval
, optlen
))
3063 sctp_sk(sk
)->adaptation_ind
= adaptation
.ssb_adaptation_ind
;
3069 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3071 * The context field in the sctp_sndrcvinfo structure is normally only
3072 * used when a failed message is retrieved holding the value that was
3073 * sent down on the actual send call. This option allows the setting of
3074 * a default context on an association basis that will be received on
3075 * reading messages from the peer. This is especially helpful in the
3076 * one-2-many model for an application to keep some reference to an
3077 * internal state machine that is processing messages on the
3078 * association. Note that the setting of this value only effects
3079 * received messages from the peer and does not effect the value that is
3080 * saved with outbound messages.
3082 static int sctp_setsockopt_context(struct sock
*sk
, char __user
*optval
,
3083 unsigned int optlen
)
3085 struct sctp_assoc_value params
;
3086 struct sctp_sock
*sp
;
3087 struct sctp_association
*asoc
;
3089 if (optlen
!= sizeof(struct sctp_assoc_value
))
3091 if (copy_from_user(¶ms
, optval
, optlen
))
3096 if (params
.assoc_id
!= 0) {
3097 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3100 asoc
->default_rcv_context
= params
.assoc_value
;
3102 sp
->default_rcv_context
= params
.assoc_value
;
3109 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3111 * This options will at a minimum specify if the implementation is doing
3112 * fragmented interleave. Fragmented interleave, for a one to many
3113 * socket, is when subsequent calls to receive a message may return
3114 * parts of messages from different associations. Some implementations
3115 * may allow you to turn this value on or off. If so, when turned off,
3116 * no fragment interleave will occur (which will cause a head of line
3117 * blocking amongst multiple associations sharing the same one to many
3118 * socket). When this option is turned on, then each receive call may
3119 * come from a different association (thus the user must receive data
3120 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3121 * association each receive belongs to.
3123 * This option takes a boolean value. A non-zero value indicates that
3124 * fragmented interleave is on. A value of zero indicates that
3125 * fragmented interleave is off.
3127 * Note that it is important that an implementation that allows this
3128 * option to be turned on, have it off by default. Otherwise an unaware
3129 * application using the one to many model may become confused and act
3132 static int sctp_setsockopt_fragment_interleave(struct sock
*sk
,
3133 char __user
*optval
,
3134 unsigned int optlen
)
3138 if (optlen
!= sizeof(int))
3140 if (get_user(val
, (int __user
*)optval
))
3143 sctp_sk(sk
)->frag_interleave
= (val
== 0) ? 0 : 1;
3149 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3150 * (SCTP_PARTIAL_DELIVERY_POINT)
3152 * This option will set or get the SCTP partial delivery point. This
3153 * point is the size of a message where the partial delivery API will be
3154 * invoked to help free up rwnd space for the peer. Setting this to a
3155 * lower value will cause partial deliveries to happen more often. The
3156 * calls argument is an integer that sets or gets the partial delivery
3157 * point. Note also that the call will fail if the user attempts to set
3158 * this value larger than the socket receive buffer size.
3160 * Note that any single message having a length smaller than or equal to
3161 * the SCTP partial delivery point will be delivered in one single read
3162 * call as long as the user provided buffer is large enough to hold the
3165 static int sctp_setsockopt_partial_delivery_point(struct sock
*sk
,
3166 char __user
*optval
,
3167 unsigned int optlen
)
3171 if (optlen
!= sizeof(u32
))
3173 if (get_user(val
, (int __user
*)optval
))
3176 /* Note: We double the receive buffer from what the user sets
3177 * it to be, also initial rwnd is based on rcvbuf/2.
3179 if (val
> (sk
->sk_rcvbuf
>> 1))
3182 sctp_sk(sk
)->pd_point
= val
;
3184 return 0; /* is this the right error code? */
3188 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3190 * This option will allow a user to change the maximum burst of packets
3191 * that can be emitted by this association. Note that the default value
3192 * is 4, and some implementations may restrict this setting so that it
3193 * can only be lowered.
3195 * NOTE: This text doesn't seem right. Do this on a socket basis with
3196 * future associations inheriting the socket value.
3198 static int sctp_setsockopt_maxburst(struct sock
*sk
,
3199 char __user
*optval
,
3200 unsigned int optlen
)
3202 struct sctp_assoc_value params
;
3203 struct sctp_sock
*sp
;
3204 struct sctp_association
*asoc
;
3208 if (optlen
== sizeof(int)) {
3209 pr_warn("Use of int in max_burst socket option deprecated\n");
3210 pr_warn("Use struct sctp_assoc_value instead\n");
3211 if (copy_from_user(&val
, optval
, optlen
))
3213 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
3214 if (copy_from_user(¶ms
, optval
, optlen
))
3216 val
= params
.assoc_value
;
3217 assoc_id
= params
.assoc_id
;
3223 if (assoc_id
!= 0) {
3224 asoc
= sctp_id2assoc(sk
, assoc_id
);
3227 asoc
->max_burst
= val
;
3229 sp
->max_burst
= val
;
3235 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3237 * This set option adds a chunk type that the user is requesting to be
3238 * received only in an authenticated way. Changes to the list of chunks
3239 * will only effect future associations on the socket.
3241 static int sctp_setsockopt_auth_chunk(struct sock
*sk
,
3242 char __user
*optval
,
3243 unsigned int optlen
)
3245 struct sctp_authchunk val
;
3247 if (!sctp_auth_enable
)
3250 if (optlen
!= sizeof(struct sctp_authchunk
))
3252 if (copy_from_user(&val
, optval
, optlen
))
3255 switch (val
.sauth_chunk
) {
3257 case SCTP_CID_INIT_ACK
:
3258 case SCTP_CID_SHUTDOWN_COMPLETE
:
3263 /* add this chunk id to the endpoint */
3264 return sctp_auth_ep_add_chunkid(sctp_sk(sk
)->ep
, val
.sauth_chunk
);
3268 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3270 * This option gets or sets the list of HMAC algorithms that the local
3271 * endpoint requires the peer to use.
3273 static int sctp_setsockopt_hmac_ident(struct sock
*sk
,
3274 char __user
*optval
,
3275 unsigned int optlen
)
3277 struct sctp_hmacalgo
*hmacs
;
3281 if (!sctp_auth_enable
)
3284 if (optlen
< sizeof(struct sctp_hmacalgo
))
3287 hmacs
= memdup_user(optval
, optlen
);
3289 return PTR_ERR(hmacs
);
3291 idents
= hmacs
->shmac_num_idents
;
3292 if (idents
== 0 || idents
> SCTP_AUTH_NUM_HMACS
||
3293 (idents
* sizeof(u16
)) > (optlen
- sizeof(struct sctp_hmacalgo
))) {
3298 err
= sctp_auth_ep_set_hmacs(sctp_sk(sk
)->ep
, hmacs
);
3305 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3307 * This option will set a shared secret key which is used to build an
3308 * association shared key.
3310 static int sctp_setsockopt_auth_key(struct sock
*sk
,
3311 char __user
*optval
,
3312 unsigned int optlen
)
3314 struct sctp_authkey
*authkey
;
3315 struct sctp_association
*asoc
;
3318 if (!sctp_auth_enable
)
3321 if (optlen
<= sizeof(struct sctp_authkey
))
3324 authkey
= memdup_user(optval
, optlen
);
3325 if (IS_ERR(authkey
))
3326 return PTR_ERR(authkey
);
3328 if (authkey
->sca_keylength
> optlen
- sizeof(struct sctp_authkey
)) {
3333 asoc
= sctp_id2assoc(sk
, authkey
->sca_assoc_id
);
3334 if (!asoc
&& authkey
->sca_assoc_id
&& sctp_style(sk
, UDP
)) {
3339 ret
= sctp_auth_set_key(sctp_sk(sk
)->ep
, asoc
, authkey
);
3346 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3348 * This option will get or set the active shared key to be used to build
3349 * the association shared key.
3351 static int sctp_setsockopt_active_key(struct sock
*sk
,
3352 char __user
*optval
,
3353 unsigned int optlen
)
3355 struct sctp_authkeyid val
;
3356 struct sctp_association
*asoc
;
3358 if (!sctp_auth_enable
)
3361 if (optlen
!= sizeof(struct sctp_authkeyid
))
3363 if (copy_from_user(&val
, optval
, optlen
))
3366 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3367 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3370 return sctp_auth_set_active_key(sctp_sk(sk
)->ep
, asoc
,
3371 val
.scact_keynumber
);
3375 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3377 * This set option will delete a shared secret key from use.
3379 static int sctp_setsockopt_del_key(struct sock
*sk
,
3380 char __user
*optval
,
3381 unsigned int optlen
)
3383 struct sctp_authkeyid val
;
3384 struct sctp_association
*asoc
;
3386 if (!sctp_auth_enable
)
3389 if (optlen
!= sizeof(struct sctp_authkeyid
))
3391 if (copy_from_user(&val
, optval
, optlen
))
3394 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3395 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3398 return sctp_auth_del_key_id(sctp_sk(sk
)->ep
, asoc
,
3399 val
.scact_keynumber
);
3404 * 8.1.23 SCTP_AUTO_ASCONF
3406 * This option will enable or disable the use of the automatic generation of
3407 * ASCONF chunks to add and delete addresses to an existing association. Note
3408 * that this option has two caveats namely: a) it only affects sockets that
3409 * are bound to all addresses available to the SCTP stack, and b) the system
3410 * administrator may have an overriding control that turns the ASCONF feature
3411 * off no matter what setting the socket option may have.
3412 * This option expects an integer boolean flag, where a non-zero value turns on
3413 * the option, and a zero value turns off the option.
3414 * Note. In this implementation, socket operation overrides default parameter
3415 * being set by sysctl as well as FreeBSD implementation
3417 static int sctp_setsockopt_auto_asconf(struct sock
*sk
, char __user
*optval
,
3418 unsigned int optlen
)
3421 struct sctp_sock
*sp
= sctp_sk(sk
);
3423 if (optlen
< sizeof(int))
3425 if (get_user(val
, (int __user
*)optval
))
3427 if (!sctp_is_ep_boundall(sk
) && val
)
3429 if ((val
&& sp
->do_auto_asconf
) || (!val
&& !sp
->do_auto_asconf
))
3432 if (val
== 0 && sp
->do_auto_asconf
) {
3433 list_del(&sp
->auto_asconf_list
);
3434 sp
->do_auto_asconf
= 0;
3435 } else if (val
&& !sp
->do_auto_asconf
) {
3436 list_add_tail(&sp
->auto_asconf_list
,
3437 &sctp_auto_asconf_splist
);
3438 sp
->do_auto_asconf
= 1;
3444 /* API 6.2 setsockopt(), getsockopt()
3446 * Applications use setsockopt() and getsockopt() to set or retrieve
3447 * socket options. Socket options are used to change the default
3448 * behavior of sockets calls. They are described in Section 7.
3452 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3453 * int __user *optlen);
3454 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3457 * sd - the socket descript.
3458 * level - set to IPPROTO_SCTP for all SCTP options.
3459 * optname - the option name.
3460 * optval - the buffer to store the value of the option.
3461 * optlen - the size of the buffer.
3463 SCTP_STATIC
int sctp_setsockopt(struct sock
*sk
, int level
, int optname
,
3464 char __user
*optval
, unsigned int optlen
)
3468 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3471 /* I can hardly begin to describe how wrong this is. This is
3472 * so broken as to be worse than useless. The API draft
3473 * REALLY is NOT helpful here... I am not convinced that the
3474 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3475 * are at all well-founded.
3477 if (level
!= SOL_SCTP
) {
3478 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
3479 retval
= af
->setsockopt(sk
, level
, optname
, optval
, optlen
);
3486 case SCTP_SOCKOPT_BINDX_ADD
:
3487 /* 'optlen' is the size of the addresses buffer. */
3488 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3489 optlen
, SCTP_BINDX_ADD_ADDR
);
3492 case SCTP_SOCKOPT_BINDX_REM
:
3493 /* 'optlen' is the size of the addresses buffer. */
3494 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3495 optlen
, SCTP_BINDX_REM_ADDR
);
3498 case SCTP_SOCKOPT_CONNECTX_OLD
:
3499 /* 'optlen' is the size of the addresses buffer. */
3500 retval
= sctp_setsockopt_connectx_old(sk
,
3501 (struct sockaddr __user
*)optval
,
3505 case SCTP_SOCKOPT_CONNECTX
:
3506 /* 'optlen' is the size of the addresses buffer. */
3507 retval
= sctp_setsockopt_connectx(sk
,
3508 (struct sockaddr __user
*)optval
,
3512 case SCTP_DISABLE_FRAGMENTS
:
3513 retval
= sctp_setsockopt_disable_fragments(sk
, optval
, optlen
);
3517 retval
= sctp_setsockopt_events(sk
, optval
, optlen
);
3520 case SCTP_AUTOCLOSE
:
3521 retval
= sctp_setsockopt_autoclose(sk
, optval
, optlen
);
3524 case SCTP_PEER_ADDR_PARAMS
:
3525 retval
= sctp_setsockopt_peer_addr_params(sk
, optval
, optlen
);
3528 case SCTP_DELAYED_SACK
:
3529 retval
= sctp_setsockopt_delayed_ack(sk
, optval
, optlen
);
3531 case SCTP_PARTIAL_DELIVERY_POINT
:
3532 retval
= sctp_setsockopt_partial_delivery_point(sk
, optval
, optlen
);
3536 retval
= sctp_setsockopt_initmsg(sk
, optval
, optlen
);
3538 case SCTP_DEFAULT_SEND_PARAM
:
3539 retval
= sctp_setsockopt_default_send_param(sk
, optval
,
3542 case SCTP_PRIMARY_ADDR
:
3543 retval
= sctp_setsockopt_primary_addr(sk
, optval
, optlen
);
3545 case SCTP_SET_PEER_PRIMARY_ADDR
:
3546 retval
= sctp_setsockopt_peer_primary_addr(sk
, optval
, optlen
);
3549 retval
= sctp_setsockopt_nodelay(sk
, optval
, optlen
);
3552 retval
= sctp_setsockopt_rtoinfo(sk
, optval
, optlen
);
3554 case SCTP_ASSOCINFO
:
3555 retval
= sctp_setsockopt_associnfo(sk
, optval
, optlen
);
3557 case SCTP_I_WANT_MAPPED_V4_ADDR
:
3558 retval
= sctp_setsockopt_mappedv4(sk
, optval
, optlen
);
3561 retval
= sctp_setsockopt_maxseg(sk
, optval
, optlen
);
3563 case SCTP_ADAPTATION_LAYER
:
3564 retval
= sctp_setsockopt_adaptation_layer(sk
, optval
, optlen
);
3567 retval
= sctp_setsockopt_context(sk
, optval
, optlen
);
3569 case SCTP_FRAGMENT_INTERLEAVE
:
3570 retval
= sctp_setsockopt_fragment_interleave(sk
, optval
, optlen
);
3572 case SCTP_MAX_BURST
:
3573 retval
= sctp_setsockopt_maxburst(sk
, optval
, optlen
);
3575 case SCTP_AUTH_CHUNK
:
3576 retval
= sctp_setsockopt_auth_chunk(sk
, optval
, optlen
);
3578 case SCTP_HMAC_IDENT
:
3579 retval
= sctp_setsockopt_hmac_ident(sk
, optval
, optlen
);
3582 retval
= sctp_setsockopt_auth_key(sk
, optval
, optlen
);
3584 case SCTP_AUTH_ACTIVE_KEY
:
3585 retval
= sctp_setsockopt_active_key(sk
, optval
, optlen
);
3587 case SCTP_AUTH_DELETE_KEY
:
3588 retval
= sctp_setsockopt_del_key(sk
, optval
, optlen
);
3590 case SCTP_AUTO_ASCONF
:
3591 retval
= sctp_setsockopt_auto_asconf(sk
, optval
, optlen
);
3594 retval
= -ENOPROTOOPT
;
3598 sctp_release_sock(sk
);
3604 /* API 3.1.6 connect() - UDP Style Syntax
3606 * An application may use the connect() call in the UDP model to initiate an
3607 * association without sending data.
3611 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3613 * sd: the socket descriptor to have a new association added to.
3615 * nam: the address structure (either struct sockaddr_in or struct
3616 * sockaddr_in6 defined in RFC2553 [7]).
3618 * len: the size of the address.
3620 SCTP_STATIC
int sctp_connect(struct sock
*sk
, struct sockaddr
*addr
,
3628 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3629 __func__
, sk
, addr
, addr_len
);
3631 /* Validate addr_len before calling common connect/connectx routine. */
3632 af
= sctp_get_af_specific(addr
->sa_family
);
3633 if (!af
|| addr_len
< af
->sockaddr_len
) {
3636 /* Pass correct addr len to common routine (so it knows there
3637 * is only one address being passed.
3639 err
= __sctp_connect(sk
, addr
, af
->sockaddr_len
, NULL
);
3642 sctp_release_sock(sk
);
3646 /* FIXME: Write comments. */
3647 SCTP_STATIC
int sctp_disconnect(struct sock
*sk
, int flags
)
3649 return -EOPNOTSUPP
; /* STUB */
3652 /* 4.1.4 accept() - TCP Style Syntax
3654 * Applications use accept() call to remove an established SCTP
3655 * association from the accept queue of the endpoint. A new socket
3656 * descriptor will be returned from accept() to represent the newly
3657 * formed association.
3659 SCTP_STATIC
struct sock
*sctp_accept(struct sock
*sk
, int flags
, int *err
)
3661 struct sctp_sock
*sp
;
3662 struct sctp_endpoint
*ep
;
3663 struct sock
*newsk
= NULL
;
3664 struct sctp_association
*asoc
;
3673 if (!sctp_style(sk
, TCP
)) {
3674 error
= -EOPNOTSUPP
;
3678 if (!sctp_sstate(sk
, LISTENING
)) {
3683 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
3685 error
= sctp_wait_for_accept(sk
, timeo
);
3689 /* We treat the list of associations on the endpoint as the accept
3690 * queue and pick the first association on the list.
3692 asoc
= list_entry(ep
->asocs
.next
, struct sctp_association
, asocs
);
3694 newsk
= sp
->pf
->create_accept_sk(sk
, asoc
);
3700 /* Populate the fields of the newsk from the oldsk and migrate the
3701 * asoc to the newsk.
3703 sctp_sock_migrate(sk
, newsk
, asoc
, SCTP_SOCKET_TCP
);
3706 sctp_release_sock(sk
);
3711 /* The SCTP ioctl handler. */
3712 SCTP_STATIC
int sctp_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
3719 * SEQPACKET-style sockets in LISTENING state are valid, for
3720 * SCTP, so only discard TCP-style sockets in LISTENING state.
3722 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
3727 struct sk_buff
*skb
;
3728 unsigned int amount
= 0;
3730 skb
= skb_peek(&sk
->sk_receive_queue
);
3733 * We will only return the amount of this packet since
3734 * that is all that will be read.
3738 rc
= put_user(amount
, (int __user
*)arg
);
3746 sctp_release_sock(sk
);
3750 /* This is the function which gets called during socket creation to
3751 * initialized the SCTP-specific portion of the sock.
3752 * The sock structure should already be zero-filled memory.
3754 SCTP_STATIC
int sctp_init_sock(struct sock
*sk
)
3756 struct sctp_endpoint
*ep
;
3757 struct sctp_sock
*sp
;
3759 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk
);
3763 /* Initialize the SCTP per socket area. */
3764 switch (sk
->sk_type
) {
3765 case SOCK_SEQPACKET
:
3766 sp
->type
= SCTP_SOCKET_UDP
;
3769 sp
->type
= SCTP_SOCKET_TCP
;
3772 return -ESOCKTNOSUPPORT
;
3775 /* Initialize default send parameters. These parameters can be
3776 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3778 sp
->default_stream
= 0;
3779 sp
->default_ppid
= 0;
3780 sp
->default_flags
= 0;
3781 sp
->default_context
= 0;
3782 sp
->default_timetolive
= 0;
3784 sp
->default_rcv_context
= 0;
3785 sp
->max_burst
= sctp_max_burst
;
3787 /* Initialize default setup parameters. These parameters
3788 * can be modified with the SCTP_INITMSG socket option or
3789 * overridden by the SCTP_INIT CMSG.
3791 sp
->initmsg
.sinit_num_ostreams
= sctp_max_outstreams
;
3792 sp
->initmsg
.sinit_max_instreams
= sctp_max_instreams
;
3793 sp
->initmsg
.sinit_max_attempts
= sctp_max_retrans_init
;
3794 sp
->initmsg
.sinit_max_init_timeo
= sctp_rto_max
;
3796 /* Initialize default RTO related parameters. These parameters can
3797 * be modified for with the SCTP_RTOINFO socket option.
3799 sp
->rtoinfo
.srto_initial
= sctp_rto_initial
;
3800 sp
->rtoinfo
.srto_max
= sctp_rto_max
;
3801 sp
->rtoinfo
.srto_min
= sctp_rto_min
;
3803 /* Initialize default association related parameters. These parameters
3804 * can be modified with the SCTP_ASSOCINFO socket option.
3806 sp
->assocparams
.sasoc_asocmaxrxt
= sctp_max_retrans_association
;
3807 sp
->assocparams
.sasoc_number_peer_destinations
= 0;
3808 sp
->assocparams
.sasoc_peer_rwnd
= 0;
3809 sp
->assocparams
.sasoc_local_rwnd
= 0;
3810 sp
->assocparams
.sasoc_cookie_life
= sctp_valid_cookie_life
;
3812 /* Initialize default event subscriptions. By default, all the
3815 memset(&sp
->subscribe
, 0, sizeof(struct sctp_event_subscribe
));
3817 /* Default Peer Address Parameters. These defaults can
3818 * be modified via SCTP_PEER_ADDR_PARAMS
3820 sp
->hbinterval
= sctp_hb_interval
;
3821 sp
->pathmaxrxt
= sctp_max_retrans_path
;
3822 sp
->pathmtu
= 0; // allow default discovery
3823 sp
->sackdelay
= sctp_sack_timeout
;
3825 sp
->param_flags
= SPP_HB_ENABLE
|
3827 SPP_SACKDELAY_ENABLE
;
3829 /* If enabled no SCTP message fragmentation will be performed.
3830 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3832 sp
->disable_fragments
= 0;
3834 /* Enable Nagle algorithm by default. */
3837 /* Enable by default. */
3840 /* Auto-close idle associations after the configured
3841 * number of seconds. A value of 0 disables this
3842 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3843 * for UDP-style sockets only.
3847 /* User specified fragmentation limit. */
3850 sp
->adaptation_ind
= 0;
3852 sp
->pf
= sctp_get_pf_specific(sk
->sk_family
);
3854 /* Control variables for partial data delivery. */
3855 atomic_set(&sp
->pd_mode
, 0);
3856 skb_queue_head_init(&sp
->pd_lobby
);
3857 sp
->frag_interleave
= 0;
3859 /* Create a per socket endpoint structure. Even if we
3860 * change the data structure relationships, this may still
3861 * be useful for storing pre-connect address information.
3863 ep
= sctp_endpoint_new(sk
, GFP_KERNEL
);
3870 SCTP_DBG_OBJCNT_INC(sock
);
3873 percpu_counter_inc(&sctp_sockets_allocated
);
3874 sock_prot_inuse_add(sock_net(sk
), sk
->sk_prot
, 1);
3875 if (sctp_default_auto_asconf
) {
3876 list_add_tail(&sp
->auto_asconf_list
,
3877 &sctp_auto_asconf_splist
);
3878 sp
->do_auto_asconf
= 1;
3880 sp
->do_auto_asconf
= 0;
3886 /* Cleanup any SCTP per socket resources. */
3887 SCTP_STATIC
void sctp_destroy_sock(struct sock
*sk
)
3889 struct sctp_sock
*sp
;
3891 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk
);
3893 /* Release our hold on the endpoint. */
3895 if (sp
->do_auto_asconf
) {
3896 sp
->do_auto_asconf
= 0;
3897 list_del(&sp
->auto_asconf_list
);
3899 sctp_endpoint_free(sp
->ep
);
3901 percpu_counter_dec(&sctp_sockets_allocated
);
3902 sock_prot_inuse_add(sock_net(sk
), sk
->sk_prot
, -1);
3906 /* API 4.1.7 shutdown() - TCP Style Syntax
3907 * int shutdown(int socket, int how);
3909 * sd - the socket descriptor of the association to be closed.
3910 * how - Specifies the type of shutdown. The values are
3913 * Disables further receive operations. No SCTP
3914 * protocol action is taken.
3916 * Disables further send operations, and initiates
3917 * the SCTP shutdown sequence.
3919 * Disables further send and receive operations
3920 * and initiates the SCTP shutdown sequence.
3922 SCTP_STATIC
void sctp_shutdown(struct sock
*sk
, int how
)
3924 struct sctp_endpoint
*ep
;
3925 struct sctp_association
*asoc
;
3927 if (!sctp_style(sk
, TCP
))
3930 if (how
& SEND_SHUTDOWN
) {
3931 ep
= sctp_sk(sk
)->ep
;
3932 if (!list_empty(&ep
->asocs
)) {
3933 asoc
= list_entry(ep
->asocs
.next
,
3934 struct sctp_association
, asocs
);
3935 sctp_primitive_SHUTDOWN(asoc
, NULL
);
3940 /* 7.2.1 Association Status (SCTP_STATUS)
3942 * Applications can retrieve current status information about an
3943 * association, including association state, peer receiver window size,
3944 * number of unacked data chunks, and number of data chunks pending
3945 * receipt. This information is read-only.
3947 static int sctp_getsockopt_sctp_status(struct sock
*sk
, int len
,
3948 char __user
*optval
,
3951 struct sctp_status status
;
3952 struct sctp_association
*asoc
= NULL
;
3953 struct sctp_transport
*transport
;
3954 sctp_assoc_t associd
;
3957 if (len
< sizeof(status
)) {
3962 len
= sizeof(status
);
3963 if (copy_from_user(&status
, optval
, len
)) {
3968 associd
= status
.sstat_assoc_id
;
3969 asoc
= sctp_id2assoc(sk
, associd
);
3975 transport
= asoc
->peer
.primary_path
;
3977 status
.sstat_assoc_id
= sctp_assoc2id(asoc
);
3978 status
.sstat_state
= asoc
->state
;
3979 status
.sstat_rwnd
= asoc
->peer
.rwnd
;
3980 status
.sstat_unackdata
= asoc
->unack_data
;
3982 status
.sstat_penddata
= sctp_tsnmap_pending(&asoc
->peer
.tsn_map
);
3983 status
.sstat_instrms
= asoc
->c
.sinit_max_instreams
;
3984 status
.sstat_outstrms
= asoc
->c
.sinit_num_ostreams
;
3985 status
.sstat_fragmentation_point
= asoc
->frag_point
;
3986 status
.sstat_primary
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
3987 memcpy(&status
.sstat_primary
.spinfo_address
, &transport
->ipaddr
,
3988 transport
->af_specific
->sockaddr_len
);
3989 /* Map ipv4 address into v4-mapped-on-v6 address. */
3990 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
3991 (union sctp_addr
*)&status
.sstat_primary
.spinfo_address
);
3992 status
.sstat_primary
.spinfo_state
= transport
->state
;
3993 status
.sstat_primary
.spinfo_cwnd
= transport
->cwnd
;
3994 status
.sstat_primary
.spinfo_srtt
= transport
->srtt
;
3995 status
.sstat_primary
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
3996 status
.sstat_primary
.spinfo_mtu
= transport
->pathmtu
;
3998 if (status
.sstat_primary
.spinfo_state
== SCTP_UNKNOWN
)
3999 status
.sstat_primary
.spinfo_state
= SCTP_ACTIVE
;
4001 if (put_user(len
, optlen
)) {
4006 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
4007 len
, status
.sstat_state
, status
.sstat_rwnd
,
4008 status
.sstat_assoc_id
);
4010 if (copy_to_user(optval
, &status
, len
)) {
4020 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4022 * Applications can retrieve information about a specific peer address
4023 * of an association, including its reachability state, congestion
4024 * window, and retransmission timer values. This information is
4027 static int sctp_getsockopt_peer_addr_info(struct sock
*sk
, int len
,
4028 char __user
*optval
,
4031 struct sctp_paddrinfo pinfo
;
4032 struct sctp_transport
*transport
;
4035 if (len
< sizeof(pinfo
)) {
4040 len
= sizeof(pinfo
);
4041 if (copy_from_user(&pinfo
, optval
, len
)) {
4046 transport
= sctp_addr_id2transport(sk
, &pinfo
.spinfo_address
,
4047 pinfo
.spinfo_assoc_id
);
4051 pinfo
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
4052 pinfo
.spinfo_state
= transport
->state
;
4053 pinfo
.spinfo_cwnd
= transport
->cwnd
;
4054 pinfo
.spinfo_srtt
= transport
->srtt
;
4055 pinfo
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
4056 pinfo
.spinfo_mtu
= transport
->pathmtu
;
4058 if (pinfo
.spinfo_state
== SCTP_UNKNOWN
)
4059 pinfo
.spinfo_state
= SCTP_ACTIVE
;
4061 if (put_user(len
, optlen
)) {
4066 if (copy_to_user(optval
, &pinfo
, len
)) {
4075 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4077 * This option is a on/off flag. If enabled no SCTP message
4078 * fragmentation will be performed. Instead if a message being sent
4079 * exceeds the current PMTU size, the message will NOT be sent and
4080 * instead a error will be indicated to the user.
4082 static int sctp_getsockopt_disable_fragments(struct sock
*sk
, int len
,
4083 char __user
*optval
, int __user
*optlen
)
4087 if (len
< sizeof(int))
4091 val
= (sctp_sk(sk
)->disable_fragments
== 1);
4092 if (put_user(len
, optlen
))
4094 if (copy_to_user(optval
, &val
, len
))
4099 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4101 * This socket option is used to specify various notifications and
4102 * ancillary data the user wishes to receive.
4104 static int sctp_getsockopt_events(struct sock
*sk
, int len
, char __user
*optval
,
4107 if (len
< sizeof(struct sctp_event_subscribe
))
4109 len
= sizeof(struct sctp_event_subscribe
);
4110 if (put_user(len
, optlen
))
4112 if (copy_to_user(optval
, &sctp_sk(sk
)->subscribe
, len
))
4117 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4119 * This socket option is applicable to the UDP-style socket only. When
4120 * set it will cause associations that are idle for more than the
4121 * specified number of seconds to automatically close. An association
4122 * being idle is defined an association that has NOT sent or received
4123 * user data. The special value of '0' indicates that no automatic
4124 * close of any associations should be performed. The option expects an
4125 * integer defining the number of seconds of idle time before an
4126 * association is closed.
4128 static int sctp_getsockopt_autoclose(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4130 /* Applicable to UDP-style socket only */
4131 if (sctp_style(sk
, TCP
))
4133 if (len
< sizeof(int))
4136 if (put_user(len
, optlen
))
4138 if (copy_to_user(optval
, &sctp_sk(sk
)->autoclose
, sizeof(int)))
4143 /* Helper routine to branch off an association to a new socket. */
4144 SCTP_STATIC
int sctp_do_peeloff(struct sctp_association
*asoc
,
4145 struct socket
**sockp
)
4147 struct sock
*sk
= asoc
->base
.sk
;
4148 struct socket
*sock
;
4152 /* An association cannot be branched off from an already peeled-off
4153 * socket, nor is this supported for tcp style sockets.
4155 if (!sctp_style(sk
, UDP
))
4158 /* Create a new socket. */
4159 err
= sock_create(sk
->sk_family
, SOCK_SEQPACKET
, IPPROTO_SCTP
, &sock
);
4163 sctp_copy_sock(sock
->sk
, sk
, asoc
);
4165 /* Make peeled-off sockets more like 1-1 accepted sockets.
4166 * Set the daddr and initialize id to something more random
4168 af
= sctp_get_af_specific(asoc
->peer
.primary_addr
.sa
.sa_family
);
4169 af
->to_sk_daddr(&asoc
->peer
.primary_addr
, sk
);
4171 /* Populate the fields of the newsk from the oldsk and migrate the
4172 * asoc to the newsk.
4174 sctp_sock_migrate(sk
, sock
->sk
, asoc
, SCTP_SOCKET_UDP_HIGH_BANDWIDTH
);
4181 static int sctp_getsockopt_peeloff(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4183 sctp_peeloff_arg_t peeloff
;
4184 struct socket
*newsock
;
4186 struct sctp_association
*asoc
;
4188 if (len
< sizeof(sctp_peeloff_arg_t
))
4190 len
= sizeof(sctp_peeloff_arg_t
);
4191 if (copy_from_user(&peeloff
, optval
, len
))
4194 asoc
= sctp_id2assoc(sk
, peeloff
.associd
);
4200 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__
, sk
, asoc
);
4202 retval
= sctp_do_peeloff(asoc
, &newsock
);
4206 /* Map the socket to an unused fd that can be returned to the user. */
4207 retval
= sock_map_fd(newsock
, 0);
4209 sock_release(newsock
);
4213 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
4214 __func__
, sk
, asoc
, newsock
->sk
, retval
);
4216 /* Return the fd mapped to the new socket. */
4217 peeloff
.sd
= retval
;
4218 if (put_user(len
, optlen
))
4220 if (copy_to_user(optval
, &peeloff
, len
))
4227 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4229 * Applications can enable or disable heartbeats for any peer address of
4230 * an association, modify an address's heartbeat interval, force a
4231 * heartbeat to be sent immediately, and adjust the address's maximum
4232 * number of retransmissions sent before an address is considered
4233 * unreachable. The following structure is used to access and modify an
4234 * address's parameters:
4236 * struct sctp_paddrparams {
4237 * sctp_assoc_t spp_assoc_id;
4238 * struct sockaddr_storage spp_address;
4239 * uint32_t spp_hbinterval;
4240 * uint16_t spp_pathmaxrxt;
4241 * uint32_t spp_pathmtu;
4242 * uint32_t spp_sackdelay;
4243 * uint32_t spp_flags;
4246 * spp_assoc_id - (one-to-many style socket) This is filled in the
4247 * application, and identifies the association for
4249 * spp_address - This specifies which address is of interest.
4250 * spp_hbinterval - This contains the value of the heartbeat interval,
4251 * in milliseconds. If a value of zero
4252 * is present in this field then no changes are to
4253 * be made to this parameter.
4254 * spp_pathmaxrxt - This contains the maximum number of
4255 * retransmissions before this address shall be
4256 * considered unreachable. If a value of zero
4257 * is present in this field then no changes are to
4258 * be made to this parameter.
4259 * spp_pathmtu - When Path MTU discovery is disabled the value
4260 * specified here will be the "fixed" path mtu.
4261 * Note that if the spp_address field is empty
4262 * then all associations on this address will
4263 * have this fixed path mtu set upon them.
4265 * spp_sackdelay - When delayed sack is enabled, this value specifies
4266 * the number of milliseconds that sacks will be delayed
4267 * for. This value will apply to all addresses of an
4268 * association if the spp_address field is empty. Note
4269 * also, that if delayed sack is enabled and this
4270 * value is set to 0, no change is made to the last
4271 * recorded delayed sack timer value.
4273 * spp_flags - These flags are used to control various features
4274 * on an association. The flag field may contain
4275 * zero or more of the following options.
4277 * SPP_HB_ENABLE - Enable heartbeats on the
4278 * specified address. Note that if the address
4279 * field is empty all addresses for the association
4280 * have heartbeats enabled upon them.
4282 * SPP_HB_DISABLE - Disable heartbeats on the
4283 * speicifed address. Note that if the address
4284 * field is empty all addresses for the association
4285 * will have their heartbeats disabled. Note also
4286 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4287 * mutually exclusive, only one of these two should
4288 * be specified. Enabling both fields will have
4289 * undetermined results.
4291 * SPP_HB_DEMAND - Request a user initiated heartbeat
4292 * to be made immediately.
4294 * SPP_PMTUD_ENABLE - This field will enable PMTU
4295 * discovery upon the specified address. Note that
4296 * if the address feild is empty then all addresses
4297 * on the association are effected.
4299 * SPP_PMTUD_DISABLE - This field will disable PMTU
4300 * discovery upon the specified address. Note that
4301 * if the address feild is empty then all addresses
4302 * on the association are effected. Not also that
4303 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4304 * exclusive. Enabling both will have undetermined
4307 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4308 * on delayed sack. The time specified in spp_sackdelay
4309 * is used to specify the sack delay for this address. Note
4310 * that if spp_address is empty then all addresses will
4311 * enable delayed sack and take on the sack delay
4312 * value specified in spp_sackdelay.
4313 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4314 * off delayed sack. If the spp_address field is blank then
4315 * delayed sack is disabled for the entire association. Note
4316 * also that this field is mutually exclusive to
4317 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4320 static int sctp_getsockopt_peer_addr_params(struct sock
*sk
, int len
,
4321 char __user
*optval
, int __user
*optlen
)
4323 struct sctp_paddrparams params
;
4324 struct sctp_transport
*trans
= NULL
;
4325 struct sctp_association
*asoc
= NULL
;
4326 struct sctp_sock
*sp
= sctp_sk(sk
);
4328 if (len
< sizeof(struct sctp_paddrparams
))
4330 len
= sizeof(struct sctp_paddrparams
);
4331 if (copy_from_user(¶ms
, optval
, len
))
4334 /* If an address other than INADDR_ANY is specified, and
4335 * no transport is found, then the request is invalid.
4337 if (!sctp_is_any(sk
, ( union sctp_addr
*)¶ms
.spp_address
)) {
4338 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
4339 params
.spp_assoc_id
);
4341 SCTP_DEBUG_PRINTK("Failed no transport\n");
4346 /* Get association, if assoc_id != 0 and the socket is a one
4347 * to many style socket, and an association was not found, then
4348 * the id was invalid.
4350 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
4351 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
)) {
4352 SCTP_DEBUG_PRINTK("Failed no association\n");
4357 /* Fetch transport values. */
4358 params
.spp_hbinterval
= jiffies_to_msecs(trans
->hbinterval
);
4359 params
.spp_pathmtu
= trans
->pathmtu
;
4360 params
.spp_pathmaxrxt
= trans
->pathmaxrxt
;
4361 params
.spp_sackdelay
= jiffies_to_msecs(trans
->sackdelay
);
4363 /*draft-11 doesn't say what to return in spp_flags*/
4364 params
.spp_flags
= trans
->param_flags
;
4366 /* Fetch association values. */
4367 params
.spp_hbinterval
= jiffies_to_msecs(asoc
->hbinterval
);
4368 params
.spp_pathmtu
= asoc
->pathmtu
;
4369 params
.spp_pathmaxrxt
= asoc
->pathmaxrxt
;
4370 params
.spp_sackdelay
= jiffies_to_msecs(asoc
->sackdelay
);
4372 /*draft-11 doesn't say what to return in spp_flags*/
4373 params
.spp_flags
= asoc
->param_flags
;
4375 /* Fetch socket values. */
4376 params
.spp_hbinterval
= sp
->hbinterval
;
4377 params
.spp_pathmtu
= sp
->pathmtu
;
4378 params
.spp_sackdelay
= sp
->sackdelay
;
4379 params
.spp_pathmaxrxt
= sp
->pathmaxrxt
;
4381 /*draft-11 doesn't say what to return in spp_flags*/
4382 params
.spp_flags
= sp
->param_flags
;
4385 if (copy_to_user(optval
, ¶ms
, len
))
4388 if (put_user(len
, optlen
))
4395 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4397 * This option will effect the way delayed acks are performed. This
4398 * option allows you to get or set the delayed ack time, in
4399 * milliseconds. It also allows changing the delayed ack frequency.
4400 * Changing the frequency to 1 disables the delayed sack algorithm. If
4401 * the assoc_id is 0, then this sets or gets the endpoints default
4402 * values. If the assoc_id field is non-zero, then the set or get
4403 * effects the specified association for the one to many model (the
4404 * assoc_id field is ignored by the one to one model). Note that if
4405 * sack_delay or sack_freq are 0 when setting this option, then the
4406 * current values will remain unchanged.
4408 * struct sctp_sack_info {
4409 * sctp_assoc_t sack_assoc_id;
4410 * uint32_t sack_delay;
4411 * uint32_t sack_freq;
4414 * sack_assoc_id - This parameter, indicates which association the user
4415 * is performing an action upon. Note that if this field's value is
4416 * zero then the endpoints default value is changed (effecting future
4417 * associations only).
4419 * sack_delay - This parameter contains the number of milliseconds that
4420 * the user is requesting the delayed ACK timer be set to. Note that
4421 * this value is defined in the standard to be between 200 and 500
4424 * sack_freq - This parameter contains the number of packets that must
4425 * be received before a sack is sent without waiting for the delay
4426 * timer to expire. The default value for this is 2, setting this
4427 * value to 1 will disable the delayed sack algorithm.
4429 static int sctp_getsockopt_delayed_ack(struct sock
*sk
, int len
,
4430 char __user
*optval
,
4433 struct sctp_sack_info params
;
4434 struct sctp_association
*asoc
= NULL
;
4435 struct sctp_sock
*sp
= sctp_sk(sk
);
4437 if (len
>= sizeof(struct sctp_sack_info
)) {
4438 len
= sizeof(struct sctp_sack_info
);
4440 if (copy_from_user(¶ms
, optval
, len
))
4442 } else if (len
== sizeof(struct sctp_assoc_value
)) {
4443 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4444 pr_warn("Use struct sctp_sack_info instead\n");
4445 if (copy_from_user(¶ms
, optval
, len
))
4450 /* Get association, if sack_assoc_id != 0 and the socket is a one
4451 * to many style socket, and an association was not found, then
4452 * the id was invalid.
4454 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
4455 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
4459 /* Fetch association values. */
4460 if (asoc
->param_flags
& SPP_SACKDELAY_ENABLE
) {
4461 params
.sack_delay
= jiffies_to_msecs(
4463 params
.sack_freq
= asoc
->sackfreq
;
4466 params
.sack_delay
= 0;
4467 params
.sack_freq
= 1;
4470 /* Fetch socket values. */
4471 if (sp
->param_flags
& SPP_SACKDELAY_ENABLE
) {
4472 params
.sack_delay
= sp
->sackdelay
;
4473 params
.sack_freq
= sp
->sackfreq
;
4475 params
.sack_delay
= 0;
4476 params
.sack_freq
= 1;
4480 if (copy_to_user(optval
, ¶ms
, len
))
4483 if (put_user(len
, optlen
))
4489 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4491 * Applications can specify protocol parameters for the default association
4492 * initialization. The option name argument to setsockopt() and getsockopt()
4495 * Setting initialization parameters is effective only on an unconnected
4496 * socket (for UDP-style sockets only future associations are effected
4497 * by the change). With TCP-style sockets, this option is inherited by
4498 * sockets derived from a listener socket.
4500 static int sctp_getsockopt_initmsg(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4502 if (len
< sizeof(struct sctp_initmsg
))
4504 len
= sizeof(struct sctp_initmsg
);
4505 if (put_user(len
, optlen
))
4507 if (copy_to_user(optval
, &sctp_sk(sk
)->initmsg
, len
))
4513 static int sctp_getsockopt_peer_addrs(struct sock
*sk
, int len
,
4514 char __user
*optval
, int __user
*optlen
)
4516 struct sctp_association
*asoc
;
4518 struct sctp_getaddrs getaddrs
;
4519 struct sctp_transport
*from
;
4521 union sctp_addr temp
;
4522 struct sctp_sock
*sp
= sctp_sk(sk
);
4527 if (len
< sizeof(struct sctp_getaddrs
))
4530 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
4533 /* For UDP-style sockets, id specifies the association to query. */
4534 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
4538 to
= optval
+ offsetof(struct sctp_getaddrs
,addrs
);
4539 space_left
= len
- offsetof(struct sctp_getaddrs
,addrs
);
4541 list_for_each_entry(from
, &asoc
->peer
.transport_addr_list
,
4543 memcpy(&temp
, &from
->ipaddr
, sizeof(temp
));
4544 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
4545 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4546 if (space_left
< addrlen
)
4548 if (copy_to_user(to
, &temp
, addrlen
))
4552 space_left
-= addrlen
;
4555 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
))
4557 bytes_copied
= ((char __user
*)to
) - optval
;
4558 if (put_user(bytes_copied
, optlen
))
4564 static int sctp_copy_laddrs(struct sock
*sk
, __u16 port
, void *to
,
4565 size_t space_left
, int *bytes_copied
)
4567 struct sctp_sockaddr_entry
*addr
;
4568 union sctp_addr temp
;
4573 list_for_each_entry_rcu(addr
, &sctp_local_addr_list
, list
) {
4577 if ((PF_INET
== sk
->sk_family
) &&
4578 (AF_INET6
== addr
->a
.sa
.sa_family
))
4580 if ((PF_INET6
== sk
->sk_family
) &&
4581 inet_v6_ipv6only(sk
) &&
4582 (AF_INET
== addr
->a
.sa
.sa_family
))
4584 memcpy(&temp
, &addr
->a
, sizeof(temp
));
4585 if (!temp
.v4
.sin_port
)
4586 temp
.v4
.sin_port
= htons(port
);
4588 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
4590 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4591 if (space_left
< addrlen
) {
4595 memcpy(to
, &temp
, addrlen
);
4599 space_left
-= addrlen
;
4600 *bytes_copied
+= addrlen
;
4608 static int sctp_getsockopt_local_addrs(struct sock
*sk
, int len
,
4609 char __user
*optval
, int __user
*optlen
)
4611 struct sctp_bind_addr
*bp
;
4612 struct sctp_association
*asoc
;
4614 struct sctp_getaddrs getaddrs
;
4615 struct sctp_sockaddr_entry
*addr
;
4617 union sctp_addr temp
;
4618 struct sctp_sock
*sp
= sctp_sk(sk
);
4622 int bytes_copied
= 0;
4626 if (len
< sizeof(struct sctp_getaddrs
))
4629 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
4633 * For UDP-style sockets, id specifies the association to query.
4634 * If the id field is set to the value '0' then the locally bound
4635 * addresses are returned without regard to any particular
4638 if (0 == getaddrs
.assoc_id
) {
4639 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
4641 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
4644 bp
= &asoc
->base
.bind_addr
;
4647 to
= optval
+ offsetof(struct sctp_getaddrs
,addrs
);
4648 space_left
= len
- offsetof(struct sctp_getaddrs
,addrs
);
4650 addrs
= kmalloc(space_left
, GFP_KERNEL
);
4654 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4655 * addresses from the global local address list.
4657 if (sctp_list_single_entry(&bp
->address_list
)) {
4658 addr
= list_entry(bp
->address_list
.next
,
4659 struct sctp_sockaddr_entry
, list
);
4660 if (sctp_is_any(sk
, &addr
->a
)) {
4661 cnt
= sctp_copy_laddrs(sk
, bp
->port
, addrs
,
4662 space_left
, &bytes_copied
);
4672 /* Protection on the bound address list is not needed since
4673 * in the socket option context we hold a socket lock and
4674 * thus the bound address list can't change.
4676 list_for_each_entry(addr
, &bp
->address_list
, list
) {
4677 memcpy(&temp
, &addr
->a
, sizeof(temp
));
4678 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
4679 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4680 if (space_left
< addrlen
) {
4681 err
= -ENOMEM
; /*fixme: right error?*/
4684 memcpy(buf
, &temp
, addrlen
);
4686 bytes_copied
+= addrlen
;
4688 space_left
-= addrlen
;
4692 if (copy_to_user(to
, addrs
, bytes_copied
)) {
4696 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
)) {
4700 if (put_user(bytes_copied
, optlen
))
4707 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4709 * Requests that the local SCTP stack use the enclosed peer address as
4710 * the association primary. The enclosed address must be one of the
4711 * association peer's addresses.
4713 static int sctp_getsockopt_primary_addr(struct sock
*sk
, int len
,
4714 char __user
*optval
, int __user
*optlen
)
4716 struct sctp_prim prim
;
4717 struct sctp_association
*asoc
;
4718 struct sctp_sock
*sp
= sctp_sk(sk
);
4720 if (len
< sizeof(struct sctp_prim
))
4723 len
= sizeof(struct sctp_prim
);
4725 if (copy_from_user(&prim
, optval
, len
))
4728 asoc
= sctp_id2assoc(sk
, prim
.ssp_assoc_id
);
4732 if (!asoc
->peer
.primary_path
)
4735 memcpy(&prim
.ssp_addr
, &asoc
->peer
.primary_path
->ipaddr
,
4736 asoc
->peer
.primary_path
->af_specific
->sockaddr_len
);
4738 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
,
4739 (union sctp_addr
*)&prim
.ssp_addr
);
4741 if (put_user(len
, optlen
))
4743 if (copy_to_user(optval
, &prim
, len
))
4750 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4752 * Requests that the local endpoint set the specified Adaptation Layer
4753 * Indication parameter for all future INIT and INIT-ACK exchanges.
4755 static int sctp_getsockopt_adaptation_layer(struct sock
*sk
, int len
,
4756 char __user
*optval
, int __user
*optlen
)
4758 struct sctp_setadaptation adaptation
;
4760 if (len
< sizeof(struct sctp_setadaptation
))
4763 len
= sizeof(struct sctp_setadaptation
);
4765 adaptation
.ssb_adaptation_ind
= sctp_sk(sk
)->adaptation_ind
;
4767 if (put_user(len
, optlen
))
4769 if (copy_to_user(optval
, &adaptation
, len
))
4777 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4779 * Applications that wish to use the sendto() system call may wish to
4780 * specify a default set of parameters that would normally be supplied
4781 * through the inclusion of ancillary data. This socket option allows
4782 * such an application to set the default sctp_sndrcvinfo structure.
4785 * The application that wishes to use this socket option simply passes
4786 * in to this call the sctp_sndrcvinfo structure defined in Section
4787 * 5.2.2) The input parameters accepted by this call include
4788 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4789 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4790 * to this call if the caller is using the UDP model.
4792 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4794 static int sctp_getsockopt_default_send_param(struct sock
*sk
,
4795 int len
, char __user
*optval
,
4798 struct sctp_sndrcvinfo info
;
4799 struct sctp_association
*asoc
;
4800 struct sctp_sock
*sp
= sctp_sk(sk
);
4802 if (len
< sizeof(struct sctp_sndrcvinfo
))
4805 len
= sizeof(struct sctp_sndrcvinfo
);
4807 if (copy_from_user(&info
, optval
, len
))
4810 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
4811 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
4815 info
.sinfo_stream
= asoc
->default_stream
;
4816 info
.sinfo_flags
= asoc
->default_flags
;
4817 info
.sinfo_ppid
= asoc
->default_ppid
;
4818 info
.sinfo_context
= asoc
->default_context
;
4819 info
.sinfo_timetolive
= asoc
->default_timetolive
;
4821 info
.sinfo_stream
= sp
->default_stream
;
4822 info
.sinfo_flags
= sp
->default_flags
;
4823 info
.sinfo_ppid
= sp
->default_ppid
;
4824 info
.sinfo_context
= sp
->default_context
;
4825 info
.sinfo_timetolive
= sp
->default_timetolive
;
4828 if (put_user(len
, optlen
))
4830 if (copy_to_user(optval
, &info
, len
))
4838 * 7.1.5 SCTP_NODELAY
4840 * Turn on/off any Nagle-like algorithm. This means that packets are
4841 * generally sent as soon as possible and no unnecessary delays are
4842 * introduced, at the cost of more packets in the network. Expects an
4843 * integer boolean flag.
4846 static int sctp_getsockopt_nodelay(struct sock
*sk
, int len
,
4847 char __user
*optval
, int __user
*optlen
)
4851 if (len
< sizeof(int))
4855 val
= (sctp_sk(sk
)->nodelay
== 1);
4856 if (put_user(len
, optlen
))
4858 if (copy_to_user(optval
, &val
, len
))
4865 * 7.1.1 SCTP_RTOINFO
4867 * The protocol parameters used to initialize and bound retransmission
4868 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4869 * and modify these parameters.
4870 * All parameters are time values, in milliseconds. A value of 0, when
4871 * modifying the parameters, indicates that the current value should not
4875 static int sctp_getsockopt_rtoinfo(struct sock
*sk
, int len
,
4876 char __user
*optval
,
4877 int __user
*optlen
) {
4878 struct sctp_rtoinfo rtoinfo
;
4879 struct sctp_association
*asoc
;
4881 if (len
< sizeof (struct sctp_rtoinfo
))
4884 len
= sizeof(struct sctp_rtoinfo
);
4886 if (copy_from_user(&rtoinfo
, optval
, len
))
4889 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
4891 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
4894 /* Values corresponding to the specific association. */
4896 rtoinfo
.srto_initial
= jiffies_to_msecs(asoc
->rto_initial
);
4897 rtoinfo
.srto_max
= jiffies_to_msecs(asoc
->rto_max
);
4898 rtoinfo
.srto_min
= jiffies_to_msecs(asoc
->rto_min
);
4900 /* Values corresponding to the endpoint. */
4901 struct sctp_sock
*sp
= sctp_sk(sk
);
4903 rtoinfo
.srto_initial
= sp
->rtoinfo
.srto_initial
;
4904 rtoinfo
.srto_max
= sp
->rtoinfo
.srto_max
;
4905 rtoinfo
.srto_min
= sp
->rtoinfo
.srto_min
;
4908 if (put_user(len
, optlen
))
4911 if (copy_to_user(optval
, &rtoinfo
, len
))
4919 * 7.1.2 SCTP_ASSOCINFO
4921 * This option is used to tune the maximum retransmission attempts
4922 * of the association.
4923 * Returns an error if the new association retransmission value is
4924 * greater than the sum of the retransmission value of the peer.
4925 * See [SCTP] for more information.
4928 static int sctp_getsockopt_associnfo(struct sock
*sk
, int len
,
4929 char __user
*optval
,
4933 struct sctp_assocparams assocparams
;
4934 struct sctp_association
*asoc
;
4935 struct list_head
*pos
;
4938 if (len
< sizeof (struct sctp_assocparams
))
4941 len
= sizeof(struct sctp_assocparams
);
4943 if (copy_from_user(&assocparams
, optval
, len
))
4946 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
4948 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
4951 /* Values correspoinding to the specific association */
4953 assocparams
.sasoc_asocmaxrxt
= asoc
->max_retrans
;
4954 assocparams
.sasoc_peer_rwnd
= asoc
->peer
.rwnd
;
4955 assocparams
.sasoc_local_rwnd
= asoc
->a_rwnd
;
4956 assocparams
.sasoc_cookie_life
= (asoc
->cookie_life
.tv_sec
4958 (asoc
->cookie_life
.tv_usec
4961 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
4965 assocparams
.sasoc_number_peer_destinations
= cnt
;
4967 /* Values corresponding to the endpoint */
4968 struct sctp_sock
*sp
= sctp_sk(sk
);
4970 assocparams
.sasoc_asocmaxrxt
= sp
->assocparams
.sasoc_asocmaxrxt
;
4971 assocparams
.sasoc_peer_rwnd
= sp
->assocparams
.sasoc_peer_rwnd
;
4972 assocparams
.sasoc_local_rwnd
= sp
->assocparams
.sasoc_local_rwnd
;
4973 assocparams
.sasoc_cookie_life
=
4974 sp
->assocparams
.sasoc_cookie_life
;
4975 assocparams
.sasoc_number_peer_destinations
=
4977 sasoc_number_peer_destinations
;
4980 if (put_user(len
, optlen
))
4983 if (copy_to_user(optval
, &assocparams
, len
))
4990 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4992 * This socket option is a boolean flag which turns on or off mapped V4
4993 * addresses. If this option is turned on and the socket is type
4994 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4995 * If this option is turned off, then no mapping will be done of V4
4996 * addresses and a user will receive both PF_INET6 and PF_INET type
4997 * addresses on the socket.
4999 static int sctp_getsockopt_mappedv4(struct sock
*sk
, int len
,
5000 char __user
*optval
, int __user
*optlen
)
5003 struct sctp_sock
*sp
= sctp_sk(sk
);
5005 if (len
< sizeof(int))
5010 if (put_user(len
, optlen
))
5012 if (copy_to_user(optval
, &val
, len
))
5019 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5020 * (chapter and verse is quoted at sctp_setsockopt_context())
5022 static int sctp_getsockopt_context(struct sock
*sk
, int len
,
5023 char __user
*optval
, int __user
*optlen
)
5025 struct sctp_assoc_value params
;
5026 struct sctp_sock
*sp
;
5027 struct sctp_association
*asoc
;
5029 if (len
< sizeof(struct sctp_assoc_value
))
5032 len
= sizeof(struct sctp_assoc_value
);
5034 if (copy_from_user(¶ms
, optval
, len
))
5039 if (params
.assoc_id
!= 0) {
5040 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5043 params
.assoc_value
= asoc
->default_rcv_context
;
5045 params
.assoc_value
= sp
->default_rcv_context
;
5048 if (put_user(len
, optlen
))
5050 if (copy_to_user(optval
, ¶ms
, len
))
5057 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5058 * This option will get or set the maximum size to put in any outgoing
5059 * SCTP DATA chunk. If a message is larger than this size it will be
5060 * fragmented by SCTP into the specified size. Note that the underlying
5061 * SCTP implementation may fragment into smaller sized chunks when the
5062 * PMTU of the underlying association is smaller than the value set by
5063 * the user. The default value for this option is '0' which indicates
5064 * the user is NOT limiting fragmentation and only the PMTU will effect
5065 * SCTP's choice of DATA chunk size. Note also that values set larger
5066 * than the maximum size of an IP datagram will effectively let SCTP
5067 * control fragmentation (i.e. the same as setting this option to 0).
5069 * The following structure is used to access and modify this parameter:
5071 * struct sctp_assoc_value {
5072 * sctp_assoc_t assoc_id;
5073 * uint32_t assoc_value;
5076 * assoc_id: This parameter is ignored for one-to-one style sockets.
5077 * For one-to-many style sockets this parameter indicates which
5078 * association the user is performing an action upon. Note that if
5079 * this field's value is zero then the endpoints default value is
5080 * changed (effecting future associations only).
5081 * assoc_value: This parameter specifies the maximum size in bytes.
5083 static int sctp_getsockopt_maxseg(struct sock
*sk
, int len
,
5084 char __user
*optval
, int __user
*optlen
)
5086 struct sctp_assoc_value params
;
5087 struct sctp_association
*asoc
;
5089 if (len
== sizeof(int)) {
5090 pr_warn("Use of int in maxseg socket option deprecated\n");
5091 pr_warn("Use struct sctp_assoc_value instead\n");
5092 params
.assoc_id
= 0;
5093 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5094 len
= sizeof(struct sctp_assoc_value
);
5095 if (copy_from_user(¶ms
, optval
, sizeof(params
)))
5100 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5101 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
5105 params
.assoc_value
= asoc
->frag_point
;
5107 params
.assoc_value
= sctp_sk(sk
)->user_frag
;
5109 if (put_user(len
, optlen
))
5111 if (len
== sizeof(int)) {
5112 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5115 if (copy_to_user(optval
, ¶ms
, len
))
5123 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5124 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5126 static int sctp_getsockopt_fragment_interleave(struct sock
*sk
, int len
,
5127 char __user
*optval
, int __user
*optlen
)
5131 if (len
< sizeof(int))
5136 val
= sctp_sk(sk
)->frag_interleave
;
5137 if (put_user(len
, optlen
))
5139 if (copy_to_user(optval
, &val
, len
))
5146 * 7.1.25. Set or Get the sctp partial delivery point
5147 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5149 static int sctp_getsockopt_partial_delivery_point(struct sock
*sk
, int len
,
5150 char __user
*optval
,
5155 if (len
< sizeof(u32
))
5160 val
= sctp_sk(sk
)->pd_point
;
5161 if (put_user(len
, optlen
))
5163 if (copy_to_user(optval
, &val
, len
))
5170 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5171 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5173 static int sctp_getsockopt_maxburst(struct sock
*sk
, int len
,
5174 char __user
*optval
,
5177 struct sctp_assoc_value params
;
5178 struct sctp_sock
*sp
;
5179 struct sctp_association
*asoc
;
5181 if (len
== sizeof(int)) {
5182 pr_warn("Use of int in max_burst socket option deprecated\n");
5183 pr_warn("Use struct sctp_assoc_value instead\n");
5184 params
.assoc_id
= 0;
5185 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5186 len
= sizeof(struct sctp_assoc_value
);
5187 if (copy_from_user(¶ms
, optval
, len
))
5194 if (params
.assoc_id
!= 0) {
5195 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5198 params
.assoc_value
= asoc
->max_burst
;
5200 params
.assoc_value
= sp
->max_burst
;
5202 if (len
== sizeof(int)) {
5203 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5206 if (copy_to_user(optval
, ¶ms
, len
))
5214 static int sctp_getsockopt_hmac_ident(struct sock
*sk
, int len
,
5215 char __user
*optval
, int __user
*optlen
)
5217 struct sctp_hmacalgo __user
*p
= (void __user
*)optval
;
5218 struct sctp_hmac_algo_param
*hmacs
;
5222 if (!sctp_auth_enable
)
5225 hmacs
= sctp_sk(sk
)->ep
->auth_hmacs_list
;
5226 data_len
= ntohs(hmacs
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5228 if (len
< sizeof(struct sctp_hmacalgo
) + data_len
)
5231 len
= sizeof(struct sctp_hmacalgo
) + data_len
;
5232 num_idents
= data_len
/ sizeof(u16
);
5234 if (put_user(len
, optlen
))
5236 if (put_user(num_idents
, &p
->shmac_num_idents
))
5238 if (copy_to_user(p
->shmac_idents
, hmacs
->hmac_ids
, data_len
))
5243 static int sctp_getsockopt_active_key(struct sock
*sk
, int len
,
5244 char __user
*optval
, int __user
*optlen
)
5246 struct sctp_authkeyid val
;
5247 struct sctp_association
*asoc
;
5249 if (!sctp_auth_enable
)
5252 if (len
< sizeof(struct sctp_authkeyid
))
5254 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authkeyid
)))
5257 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
5258 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
5262 val
.scact_keynumber
= asoc
->active_key_id
;
5264 val
.scact_keynumber
= sctp_sk(sk
)->ep
->active_key_id
;
5266 len
= sizeof(struct sctp_authkeyid
);
5267 if (put_user(len
, optlen
))
5269 if (copy_to_user(optval
, &val
, len
))
5275 static int sctp_getsockopt_peer_auth_chunks(struct sock
*sk
, int len
,
5276 char __user
*optval
, int __user
*optlen
)
5278 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5279 struct sctp_authchunks val
;
5280 struct sctp_association
*asoc
;
5281 struct sctp_chunks_param
*ch
;
5285 if (!sctp_auth_enable
)
5288 if (len
< sizeof(struct sctp_authchunks
))
5291 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5294 to
= p
->gauth_chunks
;
5295 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5299 ch
= asoc
->peer
.peer_chunks
;
5303 /* See if the user provided enough room for all the data */
5304 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5305 if (len
< num_chunks
)
5308 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
5311 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
5312 if (put_user(len
, optlen
)) return -EFAULT
;
5313 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
5318 static int sctp_getsockopt_local_auth_chunks(struct sock
*sk
, int len
,
5319 char __user
*optval
, int __user
*optlen
)
5321 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5322 struct sctp_authchunks val
;
5323 struct sctp_association
*asoc
;
5324 struct sctp_chunks_param
*ch
;
5328 if (!sctp_auth_enable
)
5331 if (len
< sizeof(struct sctp_authchunks
))
5334 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5337 to
= p
->gauth_chunks
;
5338 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5339 if (!asoc
&& val
.gauth_assoc_id
&& sctp_style(sk
, UDP
))
5343 ch
= (struct sctp_chunks_param
*)asoc
->c
.auth_chunks
;
5345 ch
= sctp_sk(sk
)->ep
->auth_chunk_list
;
5350 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5351 if (len
< sizeof(struct sctp_authchunks
) + num_chunks
)
5354 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
5357 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
5358 if (put_user(len
, optlen
))
5360 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
5367 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5368 * This option gets the current number of associations that are attached
5369 * to a one-to-many style socket. The option value is an uint32_t.
5371 static int sctp_getsockopt_assoc_number(struct sock
*sk
, int len
,
5372 char __user
*optval
, int __user
*optlen
)
5374 struct sctp_sock
*sp
= sctp_sk(sk
);
5375 struct sctp_association
*asoc
;
5378 if (sctp_style(sk
, TCP
))
5381 if (len
< sizeof(u32
))
5386 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5390 if (put_user(len
, optlen
))
5392 if (copy_to_user(optval
, &val
, len
))
5399 * 8.1.23 SCTP_AUTO_ASCONF
5400 * See the corresponding setsockopt entry as description
5402 static int sctp_getsockopt_auto_asconf(struct sock
*sk
, int len
,
5403 char __user
*optval
, int __user
*optlen
)
5407 if (len
< sizeof(int))
5411 if (sctp_sk(sk
)->do_auto_asconf
&& sctp_is_ep_boundall(sk
))
5413 if (put_user(len
, optlen
))
5415 if (copy_to_user(optval
, &val
, len
))
5421 * 8.2.6. Get the Current Identifiers of Associations
5422 * (SCTP_GET_ASSOC_ID_LIST)
5424 * This option gets the current list of SCTP association identifiers of
5425 * the SCTP associations handled by a one-to-many style socket.
5427 static int sctp_getsockopt_assoc_ids(struct sock
*sk
, int len
,
5428 char __user
*optval
, int __user
*optlen
)
5430 struct sctp_sock
*sp
= sctp_sk(sk
);
5431 struct sctp_association
*asoc
;
5432 struct sctp_assoc_ids
*ids
;
5435 if (sctp_style(sk
, TCP
))
5438 if (len
< sizeof(struct sctp_assoc_ids
))
5441 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5445 if (len
< sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
)
5448 len
= sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
;
5450 ids
= kmalloc(len
, GFP_KERNEL
);
5454 ids
->gaids_number_of_ids
= num
;
5456 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5457 ids
->gaids_assoc_id
[num
++] = asoc
->assoc_id
;
5460 if (put_user(len
, optlen
) || copy_to_user(optval
, ids
, len
)) {
5469 SCTP_STATIC
int sctp_getsockopt(struct sock
*sk
, int level
, int optname
,
5470 char __user
*optval
, int __user
*optlen
)
5475 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5478 /* I can hardly begin to describe how wrong this is. This is
5479 * so broken as to be worse than useless. The API draft
5480 * REALLY is NOT helpful here... I am not convinced that the
5481 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5482 * are at all well-founded.
5484 if (level
!= SOL_SCTP
) {
5485 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
5487 retval
= af
->getsockopt(sk
, level
, optname
, optval
, optlen
);
5491 if (get_user(len
, optlen
))
5498 retval
= sctp_getsockopt_sctp_status(sk
, len
, optval
, optlen
);
5500 case SCTP_DISABLE_FRAGMENTS
:
5501 retval
= sctp_getsockopt_disable_fragments(sk
, len
, optval
,
5505 retval
= sctp_getsockopt_events(sk
, len
, optval
, optlen
);
5507 case SCTP_AUTOCLOSE
:
5508 retval
= sctp_getsockopt_autoclose(sk
, len
, optval
, optlen
);
5510 case SCTP_SOCKOPT_PEELOFF
:
5511 retval
= sctp_getsockopt_peeloff(sk
, len
, optval
, optlen
);
5513 case SCTP_PEER_ADDR_PARAMS
:
5514 retval
= sctp_getsockopt_peer_addr_params(sk
, len
, optval
,
5517 case SCTP_DELAYED_SACK
:
5518 retval
= sctp_getsockopt_delayed_ack(sk
, len
, optval
,
5522 retval
= sctp_getsockopt_initmsg(sk
, len
, optval
, optlen
);
5524 case SCTP_GET_PEER_ADDRS
:
5525 retval
= sctp_getsockopt_peer_addrs(sk
, len
, optval
,
5528 case SCTP_GET_LOCAL_ADDRS
:
5529 retval
= sctp_getsockopt_local_addrs(sk
, len
, optval
,
5532 case SCTP_SOCKOPT_CONNECTX3
:
5533 retval
= sctp_getsockopt_connectx3(sk
, len
, optval
, optlen
);
5535 case SCTP_DEFAULT_SEND_PARAM
:
5536 retval
= sctp_getsockopt_default_send_param(sk
, len
,
5539 case SCTP_PRIMARY_ADDR
:
5540 retval
= sctp_getsockopt_primary_addr(sk
, len
, optval
, optlen
);
5543 retval
= sctp_getsockopt_nodelay(sk
, len
, optval
, optlen
);
5546 retval
= sctp_getsockopt_rtoinfo(sk
, len
, optval
, optlen
);
5548 case SCTP_ASSOCINFO
:
5549 retval
= sctp_getsockopt_associnfo(sk
, len
, optval
, optlen
);
5551 case SCTP_I_WANT_MAPPED_V4_ADDR
:
5552 retval
= sctp_getsockopt_mappedv4(sk
, len
, optval
, optlen
);
5555 retval
= sctp_getsockopt_maxseg(sk
, len
, optval
, optlen
);
5557 case SCTP_GET_PEER_ADDR_INFO
:
5558 retval
= sctp_getsockopt_peer_addr_info(sk
, len
, optval
,
5561 case SCTP_ADAPTATION_LAYER
:
5562 retval
= sctp_getsockopt_adaptation_layer(sk
, len
, optval
,
5566 retval
= sctp_getsockopt_context(sk
, len
, optval
, optlen
);
5568 case SCTP_FRAGMENT_INTERLEAVE
:
5569 retval
= sctp_getsockopt_fragment_interleave(sk
, len
, optval
,
5572 case SCTP_PARTIAL_DELIVERY_POINT
:
5573 retval
= sctp_getsockopt_partial_delivery_point(sk
, len
, optval
,
5576 case SCTP_MAX_BURST
:
5577 retval
= sctp_getsockopt_maxburst(sk
, len
, optval
, optlen
);
5580 case SCTP_AUTH_CHUNK
:
5581 case SCTP_AUTH_DELETE_KEY
:
5582 retval
= -EOPNOTSUPP
;
5584 case SCTP_HMAC_IDENT
:
5585 retval
= sctp_getsockopt_hmac_ident(sk
, len
, optval
, optlen
);
5587 case SCTP_AUTH_ACTIVE_KEY
:
5588 retval
= sctp_getsockopt_active_key(sk
, len
, optval
, optlen
);
5590 case SCTP_PEER_AUTH_CHUNKS
:
5591 retval
= sctp_getsockopt_peer_auth_chunks(sk
, len
, optval
,
5594 case SCTP_LOCAL_AUTH_CHUNKS
:
5595 retval
= sctp_getsockopt_local_auth_chunks(sk
, len
, optval
,
5598 case SCTP_GET_ASSOC_NUMBER
:
5599 retval
= sctp_getsockopt_assoc_number(sk
, len
, optval
, optlen
);
5601 case SCTP_GET_ASSOC_ID_LIST
:
5602 retval
= sctp_getsockopt_assoc_ids(sk
, len
, optval
, optlen
);
5604 case SCTP_AUTO_ASCONF
:
5605 retval
= sctp_getsockopt_auto_asconf(sk
, len
, optval
, optlen
);
5608 retval
= -ENOPROTOOPT
;
5612 sctp_release_sock(sk
);
5616 static void sctp_hash(struct sock
*sk
)
5621 static void sctp_unhash(struct sock
*sk
)
5626 /* Check if port is acceptable. Possibly find first available port.
5628 * The port hash table (contained in the 'global' SCTP protocol storage
5629 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5630 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5631 * list (the list number is the port number hashed out, so as you
5632 * would expect from a hash function, all the ports in a given list have
5633 * such a number that hashes out to the same list number; you were
5634 * expecting that, right?); so each list has a set of ports, with a
5635 * link to the socket (struct sock) that uses it, the port number and
5636 * a fastreuse flag (FIXME: NPI ipg).
5638 static struct sctp_bind_bucket
*sctp_bucket_create(
5639 struct sctp_bind_hashbucket
*head
, unsigned short snum
);
5641 static long sctp_get_port_local(struct sock
*sk
, union sctp_addr
*addr
)
5643 struct sctp_bind_hashbucket
*head
; /* hash list */
5644 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
5645 struct hlist_node
*node
;
5646 unsigned short snum
;
5649 snum
= ntohs(addr
->v4
.sin_port
);
5651 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum
);
5652 sctp_local_bh_disable();
5655 /* Search for an available port. */
5656 int low
, high
, remaining
, index
;
5659 inet_get_local_port_range(&low
, &high
);
5660 remaining
= (high
- low
) + 1;
5661 rover
= net_random() % remaining
+ low
;
5665 if ((rover
< low
) || (rover
> high
))
5667 if (inet_is_reserved_local_port(rover
))
5669 index
= sctp_phashfn(rover
);
5670 head
= &sctp_port_hashtable
[index
];
5671 sctp_spin_lock(&head
->lock
);
5672 sctp_for_each_hentry(pp
, node
, &head
->chain
)
5673 if (pp
->port
== rover
)
5677 sctp_spin_unlock(&head
->lock
);
5678 } while (--remaining
> 0);
5680 /* Exhausted local port range during search? */
5685 /* OK, here is the one we will use. HEAD (the port
5686 * hash table list entry) is non-NULL and we hold it's
5691 /* We are given an specific port number; we verify
5692 * that it is not being used. If it is used, we will
5693 * exahust the search in the hash list corresponding
5694 * to the port number (snum) - we detect that with the
5695 * port iterator, pp being NULL.
5697 head
= &sctp_port_hashtable
[sctp_phashfn(snum
)];
5698 sctp_spin_lock(&head
->lock
);
5699 sctp_for_each_hentry(pp
, node
, &head
->chain
) {
5700 if (pp
->port
== snum
)
5707 if (!hlist_empty(&pp
->owner
)) {
5708 /* We had a port hash table hit - there is an
5709 * available port (pp != NULL) and it is being
5710 * used by other socket (pp->owner not empty); that other
5711 * socket is going to be sk2.
5713 int reuse
= sk
->sk_reuse
;
5716 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5717 if (pp
->fastreuse
&& sk
->sk_reuse
&&
5718 sk
->sk_state
!= SCTP_SS_LISTENING
)
5721 /* Run through the list of sockets bound to the port
5722 * (pp->port) [via the pointers bind_next and
5723 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5724 * we get the endpoint they describe and run through
5725 * the endpoint's list of IP (v4 or v6) addresses,
5726 * comparing each of the addresses with the address of
5727 * the socket sk. If we find a match, then that means
5728 * that this port/socket (sk) combination are already
5731 sk_for_each_bound(sk2
, node
, &pp
->owner
) {
5732 struct sctp_endpoint
*ep2
;
5733 ep2
= sctp_sk(sk2
)->ep
;
5736 (reuse
&& sk2
->sk_reuse
&&
5737 sk2
->sk_state
!= SCTP_SS_LISTENING
))
5740 if (sctp_bind_addr_conflict(&ep2
->base
.bind_addr
, addr
,
5741 sctp_sk(sk2
), sctp_sk(sk
))) {
5746 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5749 /* If there was a hash table miss, create a new port. */
5751 if (!pp
&& !(pp
= sctp_bucket_create(head
, snum
)))
5754 /* In either case (hit or miss), make sure fastreuse is 1 only
5755 * if sk->sk_reuse is too (that is, if the caller requested
5756 * SO_REUSEADDR on this socket -sk-).
5758 if (hlist_empty(&pp
->owner
)) {
5759 if (sk
->sk_reuse
&& sk
->sk_state
!= SCTP_SS_LISTENING
)
5763 } else if (pp
->fastreuse
&&
5764 (!sk
->sk_reuse
|| sk
->sk_state
== SCTP_SS_LISTENING
))
5767 /* We are set, so fill up all the data in the hash table
5768 * entry, tie the socket list information with the rest of the
5769 * sockets FIXME: Blurry, NPI (ipg).
5772 if (!sctp_sk(sk
)->bind_hash
) {
5773 inet_sk(sk
)->inet_num
= snum
;
5774 sk_add_bind_node(sk
, &pp
->owner
);
5775 sctp_sk(sk
)->bind_hash
= pp
;
5780 sctp_spin_unlock(&head
->lock
);
5783 sctp_local_bh_enable();
5787 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5788 * port is requested.
5790 static int sctp_get_port(struct sock
*sk
, unsigned short snum
)
5793 union sctp_addr addr
;
5794 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
5796 /* Set up a dummy address struct from the sk. */
5797 af
->from_sk(&addr
, sk
);
5798 addr
.v4
.sin_port
= htons(snum
);
5800 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5801 ret
= sctp_get_port_local(sk
, &addr
);
5807 * Move a socket to LISTENING state.
5809 SCTP_STATIC
int sctp_listen_start(struct sock
*sk
, int backlog
)
5811 struct sctp_sock
*sp
= sctp_sk(sk
);
5812 struct sctp_endpoint
*ep
= sp
->ep
;
5813 struct crypto_hash
*tfm
= NULL
;
5815 /* Allocate HMAC for generating cookie. */
5816 if (!sctp_sk(sk
)->hmac
&& sctp_hmac_alg
) {
5817 tfm
= crypto_alloc_hash(sctp_hmac_alg
, 0, CRYPTO_ALG_ASYNC
);
5819 if (net_ratelimit()) {
5820 pr_info("failed to load transform for %s: %ld\n",
5821 sctp_hmac_alg
, PTR_ERR(tfm
));
5825 sctp_sk(sk
)->hmac
= tfm
;
5829 * If a bind() or sctp_bindx() is not called prior to a listen()
5830 * call that allows new associations to be accepted, the system
5831 * picks an ephemeral port and will choose an address set equivalent
5832 * to binding with a wildcard address.
5834 * This is not currently spelled out in the SCTP sockets
5835 * extensions draft, but follows the practice as seen in TCP
5839 sk
->sk_state
= SCTP_SS_LISTENING
;
5840 if (!ep
->base
.bind_addr
.port
) {
5841 if (sctp_autobind(sk
))
5844 if (sctp_get_port(sk
, inet_sk(sk
)->inet_num
)) {
5845 sk
->sk_state
= SCTP_SS_CLOSED
;
5850 sk
->sk_max_ack_backlog
= backlog
;
5851 sctp_hash_endpoint(ep
);
5856 * 4.1.3 / 5.1.3 listen()
5858 * By default, new associations are not accepted for UDP style sockets.
5859 * An application uses listen() to mark a socket as being able to
5860 * accept new associations.
5862 * On TCP style sockets, applications use listen() to ready the SCTP
5863 * endpoint for accepting inbound associations.
5865 * On both types of endpoints a backlog of '0' disables listening.
5867 * Move a socket to LISTENING state.
5869 int sctp_inet_listen(struct socket
*sock
, int backlog
)
5871 struct sock
*sk
= sock
->sk
;
5872 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5875 if (unlikely(backlog
< 0))
5880 /* Peeled-off sockets are not allowed to listen(). */
5881 if (sctp_style(sk
, UDP_HIGH_BANDWIDTH
))
5884 if (sock
->state
!= SS_UNCONNECTED
)
5887 /* If backlog is zero, disable listening. */
5889 if (sctp_sstate(sk
, CLOSED
))
5893 sctp_unhash_endpoint(ep
);
5894 sk
->sk_state
= SCTP_SS_CLOSED
;
5896 sctp_sk(sk
)->bind_hash
->fastreuse
= 1;
5900 /* If we are already listening, just update the backlog */
5901 if (sctp_sstate(sk
, LISTENING
))
5902 sk
->sk_max_ack_backlog
= backlog
;
5904 err
= sctp_listen_start(sk
, backlog
);
5911 sctp_release_sock(sk
);
5916 * This function is done by modeling the current datagram_poll() and the
5917 * tcp_poll(). Note that, based on these implementations, we don't
5918 * lock the socket in this function, even though it seems that,
5919 * ideally, locking or some other mechanisms can be used to ensure
5920 * the integrity of the counters (sndbuf and wmem_alloc) used
5921 * in this place. We assume that we don't need locks either until proven
5924 * Another thing to note is that we include the Async I/O support
5925 * here, again, by modeling the current TCP/UDP code. We don't have
5926 * a good way to test with it yet.
5928 unsigned int sctp_poll(struct file
*file
, struct socket
*sock
, poll_table
*wait
)
5930 struct sock
*sk
= sock
->sk
;
5931 struct sctp_sock
*sp
= sctp_sk(sk
);
5934 poll_wait(file
, sk_sleep(sk
), wait
);
5936 /* A TCP-style listening socket becomes readable when the accept queue
5939 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
5940 return (!list_empty(&sp
->ep
->asocs
)) ?
5941 (POLLIN
| POLLRDNORM
) : 0;
5945 /* Is there any exceptional events? */
5946 if (sk
->sk_err
|| !skb_queue_empty(&sk
->sk_error_queue
))
5948 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
5949 mask
|= POLLRDHUP
| POLLIN
| POLLRDNORM
;
5950 if (sk
->sk_shutdown
== SHUTDOWN_MASK
)
5953 /* Is it readable? Reconsider this code with TCP-style support. */
5954 if (!skb_queue_empty(&sk
->sk_receive_queue
))
5955 mask
|= POLLIN
| POLLRDNORM
;
5957 /* The association is either gone or not ready. */
5958 if (!sctp_style(sk
, UDP
) && sctp_sstate(sk
, CLOSED
))
5961 /* Is it writable? */
5962 if (sctp_writeable(sk
)) {
5963 mask
|= POLLOUT
| POLLWRNORM
;
5965 set_bit(SOCK_ASYNC_NOSPACE
, &sk
->sk_socket
->flags
);
5967 * Since the socket is not locked, the buffer
5968 * might be made available after the writeable check and
5969 * before the bit is set. This could cause a lost I/O
5970 * signal. tcp_poll() has a race breaker for this race
5971 * condition. Based on their implementation, we put
5972 * in the following code to cover it as well.
5974 if (sctp_writeable(sk
))
5975 mask
|= POLLOUT
| POLLWRNORM
;
5980 /********************************************************************
5981 * 2nd Level Abstractions
5982 ********************************************************************/
5984 static struct sctp_bind_bucket
*sctp_bucket_create(
5985 struct sctp_bind_hashbucket
*head
, unsigned short snum
)
5987 struct sctp_bind_bucket
*pp
;
5989 pp
= kmem_cache_alloc(sctp_bucket_cachep
, GFP_ATOMIC
);
5991 SCTP_DBG_OBJCNT_INC(bind_bucket
);
5994 INIT_HLIST_HEAD(&pp
->owner
);
5995 hlist_add_head(&pp
->node
, &head
->chain
);
6000 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6001 static void sctp_bucket_destroy(struct sctp_bind_bucket
*pp
)
6003 if (pp
&& hlist_empty(&pp
->owner
)) {
6004 __hlist_del(&pp
->node
);
6005 kmem_cache_free(sctp_bucket_cachep
, pp
);
6006 SCTP_DBG_OBJCNT_DEC(bind_bucket
);
6010 /* Release this socket's reference to a local port. */
6011 static inline void __sctp_put_port(struct sock
*sk
)
6013 struct sctp_bind_hashbucket
*head
=
6014 &sctp_port_hashtable
[sctp_phashfn(inet_sk(sk
)->inet_num
)];
6015 struct sctp_bind_bucket
*pp
;
6017 sctp_spin_lock(&head
->lock
);
6018 pp
= sctp_sk(sk
)->bind_hash
;
6019 __sk_del_bind_node(sk
);
6020 sctp_sk(sk
)->bind_hash
= NULL
;
6021 inet_sk(sk
)->inet_num
= 0;
6022 sctp_bucket_destroy(pp
);
6023 sctp_spin_unlock(&head
->lock
);
6026 void sctp_put_port(struct sock
*sk
)
6028 sctp_local_bh_disable();
6029 __sctp_put_port(sk
);
6030 sctp_local_bh_enable();
6034 * The system picks an ephemeral port and choose an address set equivalent
6035 * to binding with a wildcard address.
6036 * One of those addresses will be the primary address for the association.
6037 * This automatically enables the multihoming capability of SCTP.
6039 static int sctp_autobind(struct sock
*sk
)
6041 union sctp_addr autoaddr
;
6045 /* Initialize a local sockaddr structure to INADDR_ANY. */
6046 af
= sctp_sk(sk
)->pf
->af
;
6048 port
= htons(inet_sk(sk
)->inet_num
);
6049 af
->inaddr_any(&autoaddr
, port
);
6051 return sctp_do_bind(sk
, &autoaddr
, af
->sockaddr_len
);
6054 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6057 * 4.2 The cmsghdr Structure *
6059 * When ancillary data is sent or received, any number of ancillary data
6060 * objects can be specified by the msg_control and msg_controllen members of
6061 * the msghdr structure, because each object is preceded by
6062 * a cmsghdr structure defining the object's length (the cmsg_len member).
6063 * Historically Berkeley-derived implementations have passed only one object
6064 * at a time, but this API allows multiple objects to be
6065 * passed in a single call to sendmsg() or recvmsg(). The following example
6066 * shows two ancillary data objects in a control buffer.
6068 * |<--------------------------- msg_controllen -------------------------->|
6071 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6073 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6076 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6078 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6081 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6082 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6084 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6086 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6093 SCTP_STATIC
int sctp_msghdr_parse(const struct msghdr
*msg
,
6094 sctp_cmsgs_t
*cmsgs
)
6096 struct cmsghdr
*cmsg
;
6097 struct msghdr
*my_msg
= (struct msghdr
*)msg
;
6099 for (cmsg
= CMSG_FIRSTHDR(msg
);
6101 cmsg
= CMSG_NXTHDR(my_msg
, cmsg
)) {
6102 if (!CMSG_OK(my_msg
, cmsg
))
6105 /* Should we parse this header or ignore? */
6106 if (cmsg
->cmsg_level
!= IPPROTO_SCTP
)
6109 /* Strictly check lengths following example in SCM code. */
6110 switch (cmsg
->cmsg_type
) {
6112 /* SCTP Socket API Extension
6113 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6115 * This cmsghdr structure provides information for
6116 * initializing new SCTP associations with sendmsg().
6117 * The SCTP_INITMSG socket option uses this same data
6118 * structure. This structure is not used for
6121 * cmsg_level cmsg_type cmsg_data[]
6122 * ------------ ------------ ----------------------
6123 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6125 if (cmsg
->cmsg_len
!=
6126 CMSG_LEN(sizeof(struct sctp_initmsg
)))
6128 cmsgs
->init
= (struct sctp_initmsg
*)CMSG_DATA(cmsg
);
6132 /* SCTP Socket API Extension
6133 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6135 * This cmsghdr structure specifies SCTP options for
6136 * sendmsg() and describes SCTP header information
6137 * about a received message through recvmsg().
6139 * cmsg_level cmsg_type cmsg_data[]
6140 * ------------ ------------ ----------------------
6141 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6143 if (cmsg
->cmsg_len
!=
6144 CMSG_LEN(sizeof(struct sctp_sndrcvinfo
)))
6148 (struct sctp_sndrcvinfo
*)CMSG_DATA(cmsg
);
6150 /* Minimally, validate the sinfo_flags. */
6151 if (cmsgs
->info
->sinfo_flags
&
6152 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
6153 SCTP_ABORT
| SCTP_EOF
))
6165 * Wait for a packet..
6166 * Note: This function is the same function as in core/datagram.c
6167 * with a few modifications to make lksctp work.
6169 static int sctp_wait_for_packet(struct sock
* sk
, int *err
, long *timeo_p
)
6174 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
6176 /* Socket errors? */
6177 error
= sock_error(sk
);
6181 if (!skb_queue_empty(&sk
->sk_receive_queue
))
6184 /* Socket shut down? */
6185 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6188 /* Sequenced packets can come disconnected. If so we report the
6193 /* Is there a good reason to think that we may receive some data? */
6194 if (list_empty(&sctp_sk(sk
)->ep
->asocs
) && !sctp_sstate(sk
, LISTENING
))
6197 /* Handle signals. */
6198 if (signal_pending(current
))
6201 /* Let another process have a go. Since we are going to sleep
6202 * anyway. Note: This may cause odd behaviors if the message
6203 * does not fit in the user's buffer, but this seems to be the
6204 * only way to honor MSG_DONTWAIT realistically.
6206 sctp_release_sock(sk
);
6207 *timeo_p
= schedule_timeout(*timeo_p
);
6211 finish_wait(sk_sleep(sk
), &wait
);
6215 error
= sock_intr_errno(*timeo_p
);
6218 finish_wait(sk_sleep(sk
), &wait
);
6223 /* Receive a datagram.
6224 * Note: This is pretty much the same routine as in core/datagram.c
6225 * with a few changes to make lksctp work.
6227 static struct sk_buff
*sctp_skb_recv_datagram(struct sock
*sk
, int flags
,
6228 int noblock
, int *err
)
6231 struct sk_buff
*skb
;
6234 timeo
= sock_rcvtimeo(sk
, noblock
);
6236 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6237 timeo
, MAX_SCHEDULE_TIMEOUT
);
6240 /* Again only user level code calls this function,
6241 * so nothing interrupt level
6242 * will suddenly eat the receive_queue.
6244 * Look at current nfs client by the way...
6245 * However, this function was correct in any case. 8)
6247 if (flags
& MSG_PEEK
) {
6248 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
6249 skb
= skb_peek(&sk
->sk_receive_queue
);
6251 atomic_inc(&skb
->users
);
6252 spin_unlock_bh(&sk
->sk_receive_queue
.lock
);
6254 skb
= skb_dequeue(&sk
->sk_receive_queue
);
6260 /* Caller is allowed not to check sk->sk_err before calling. */
6261 error
= sock_error(sk
);
6265 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6268 /* User doesn't want to wait. */
6272 } while (sctp_wait_for_packet(sk
, err
, &timeo
) == 0);
6281 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6282 static void __sctp_write_space(struct sctp_association
*asoc
)
6284 struct sock
*sk
= asoc
->base
.sk
;
6285 struct socket
*sock
= sk
->sk_socket
;
6287 if ((sctp_wspace(asoc
) > 0) && sock
) {
6288 if (waitqueue_active(&asoc
->wait
))
6289 wake_up_interruptible(&asoc
->wait
);
6291 if (sctp_writeable(sk
)) {
6292 wait_queue_head_t
*wq
= sk_sleep(sk
);
6294 if (wq
&& waitqueue_active(wq
))
6295 wake_up_interruptible(wq
);
6297 /* Note that we try to include the Async I/O support
6298 * here by modeling from the current TCP/UDP code.
6299 * We have not tested with it yet.
6301 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
))
6302 sock_wake_async(sock
,
6303 SOCK_WAKE_SPACE
, POLL_OUT
);
6308 /* Do accounting for the sndbuf space.
6309 * Decrement the used sndbuf space of the corresponding association by the
6310 * data size which was just transmitted(freed).
6312 static void sctp_wfree(struct sk_buff
*skb
)
6314 struct sctp_association
*asoc
;
6315 struct sctp_chunk
*chunk
;
6318 /* Get the saved chunk pointer. */
6319 chunk
= *((struct sctp_chunk
**)(skb
->cb
));
6322 asoc
->sndbuf_used
-= SCTP_DATA_SNDSIZE(chunk
) +
6323 sizeof(struct sk_buff
) +
6324 sizeof(struct sctp_chunk
);
6326 atomic_sub(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
6329 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6331 sk
->sk_wmem_queued
-= skb
->truesize
;
6332 sk_mem_uncharge(sk
, skb
->truesize
);
6335 __sctp_write_space(asoc
);
6337 sctp_association_put(asoc
);
6340 /* Do accounting for the receive space on the socket.
6341 * Accounting for the association is done in ulpevent.c
6342 * We set this as a destructor for the cloned data skbs so that
6343 * accounting is done at the correct time.
6345 void sctp_sock_rfree(struct sk_buff
*skb
)
6347 struct sock
*sk
= skb
->sk
;
6348 struct sctp_ulpevent
*event
= sctp_skb2event(skb
);
6350 atomic_sub(event
->rmem_len
, &sk
->sk_rmem_alloc
);
6353 * Mimic the behavior of sock_rfree
6355 sk_mem_uncharge(sk
, event
->rmem_len
);
6359 /* Helper function to wait for space in the sndbuf. */
6360 static int sctp_wait_for_sndbuf(struct sctp_association
*asoc
, long *timeo_p
,
6363 struct sock
*sk
= asoc
->base
.sk
;
6365 long current_timeo
= *timeo_p
;
6368 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6369 asoc
, (long)(*timeo_p
), msg_len
);
6371 /* Increment the association's refcnt. */
6372 sctp_association_hold(asoc
);
6374 /* Wait on the association specific sndbuf space. */
6376 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
6377 TASK_INTERRUPTIBLE
);
6380 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
6383 if (signal_pending(current
))
6384 goto do_interrupted
;
6385 if (msg_len
<= sctp_wspace(asoc
))
6388 /* Let another process have a go. Since we are going
6391 sctp_release_sock(sk
);
6392 current_timeo
= schedule_timeout(current_timeo
);
6393 BUG_ON(sk
!= asoc
->base
.sk
);
6396 *timeo_p
= current_timeo
;
6400 finish_wait(&asoc
->wait
, &wait
);
6402 /* Release the association's refcnt. */
6403 sctp_association_put(asoc
);
6412 err
= sock_intr_errno(*timeo_p
);
6420 void sctp_data_ready(struct sock
*sk
, int len
)
6422 struct socket_wq
*wq
;
6425 wq
= rcu_dereference(sk
->sk_wq
);
6426 if (wq_has_sleeper(wq
))
6427 wake_up_interruptible_sync_poll(&wq
->wait
, POLLIN
|
6428 POLLRDNORM
| POLLRDBAND
);
6429 sk_wake_async(sk
, SOCK_WAKE_WAITD
, POLL_IN
);
6433 /* If socket sndbuf has changed, wake up all per association waiters. */
6434 void sctp_write_space(struct sock
*sk
)
6436 struct sctp_association
*asoc
;
6438 /* Wake up the tasks in each wait queue. */
6439 list_for_each_entry(asoc
, &((sctp_sk(sk
))->ep
->asocs
), asocs
) {
6440 __sctp_write_space(asoc
);
6444 /* Is there any sndbuf space available on the socket?
6446 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6447 * associations on the same socket. For a UDP-style socket with
6448 * multiple associations, it is possible for it to be "unwriteable"
6449 * prematurely. I assume that this is acceptable because
6450 * a premature "unwriteable" is better than an accidental "writeable" which
6451 * would cause an unwanted block under certain circumstances. For the 1-1
6452 * UDP-style sockets or TCP-style sockets, this code should work.
6455 static int sctp_writeable(struct sock
*sk
)
6459 amt
= sk
->sk_sndbuf
- sk_wmem_alloc_get(sk
);
6465 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6466 * returns immediately with EINPROGRESS.
6468 static int sctp_wait_for_connect(struct sctp_association
*asoc
, long *timeo_p
)
6470 struct sock
*sk
= asoc
->base
.sk
;
6472 long current_timeo
= *timeo_p
;
6475 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__
, asoc
,
6478 /* Increment the association's refcnt. */
6479 sctp_association_hold(asoc
);
6482 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
6483 TASK_INTERRUPTIBLE
);
6486 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6488 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
6491 if (signal_pending(current
))
6492 goto do_interrupted
;
6494 if (sctp_state(asoc
, ESTABLISHED
))
6497 /* Let another process have a go. Since we are going
6500 sctp_release_sock(sk
);
6501 current_timeo
= schedule_timeout(current_timeo
);
6504 *timeo_p
= current_timeo
;
6508 finish_wait(&asoc
->wait
, &wait
);
6510 /* Release the association's refcnt. */
6511 sctp_association_put(asoc
);
6516 if (asoc
->init_err_counter
+ 1 > asoc
->max_init_attempts
)
6519 err
= -ECONNREFUSED
;
6523 err
= sock_intr_errno(*timeo_p
);
6531 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
)
6533 struct sctp_endpoint
*ep
;
6537 ep
= sctp_sk(sk
)->ep
;
6541 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
,
6542 TASK_INTERRUPTIBLE
);
6544 if (list_empty(&ep
->asocs
)) {
6545 sctp_release_sock(sk
);
6546 timeo
= schedule_timeout(timeo
);
6551 if (!sctp_sstate(sk
, LISTENING
))
6555 if (!list_empty(&ep
->asocs
))
6558 err
= sock_intr_errno(timeo
);
6559 if (signal_pending(current
))
6567 finish_wait(sk_sleep(sk
), &wait
);
6572 static void sctp_wait_for_close(struct sock
*sk
, long timeout
)
6577 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
6578 if (list_empty(&sctp_sk(sk
)->ep
->asocs
))
6580 sctp_release_sock(sk
);
6581 timeout
= schedule_timeout(timeout
);
6583 } while (!signal_pending(current
) && timeout
);
6585 finish_wait(sk_sleep(sk
), &wait
);
6588 static void sctp_skb_set_owner_r_frag(struct sk_buff
*skb
, struct sock
*sk
)
6590 struct sk_buff
*frag
;
6595 /* Don't forget the fragments. */
6596 skb_walk_frags(skb
, frag
)
6597 sctp_skb_set_owner_r_frag(frag
, sk
);
6600 sctp_skb_set_owner_r(skb
, sk
);
6603 void sctp_copy_sock(struct sock
*newsk
, struct sock
*sk
,
6604 struct sctp_association
*asoc
)
6606 struct inet_sock
*inet
= inet_sk(sk
);
6607 struct inet_sock
*newinet
;
6609 newsk
->sk_type
= sk
->sk_type
;
6610 newsk
->sk_bound_dev_if
= sk
->sk_bound_dev_if
;
6611 newsk
->sk_flags
= sk
->sk_flags
;
6612 newsk
->sk_no_check
= sk
->sk_no_check
;
6613 newsk
->sk_reuse
= sk
->sk_reuse
;
6615 newsk
->sk_shutdown
= sk
->sk_shutdown
;
6616 newsk
->sk_destruct
= inet_sock_destruct
;
6617 newsk
->sk_family
= sk
->sk_family
;
6618 newsk
->sk_protocol
= IPPROTO_SCTP
;
6619 newsk
->sk_backlog_rcv
= sk
->sk_prot
->backlog_rcv
;
6620 newsk
->sk_sndbuf
= sk
->sk_sndbuf
;
6621 newsk
->sk_rcvbuf
= sk
->sk_rcvbuf
;
6622 newsk
->sk_lingertime
= sk
->sk_lingertime
;
6623 newsk
->sk_rcvtimeo
= sk
->sk_rcvtimeo
;
6624 newsk
->sk_sndtimeo
= sk
->sk_sndtimeo
;
6626 newinet
= inet_sk(newsk
);
6628 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6629 * getsockname() and getpeername()
6631 newinet
->inet_sport
= inet
->inet_sport
;
6632 newinet
->inet_saddr
= inet
->inet_saddr
;
6633 newinet
->inet_rcv_saddr
= inet
->inet_rcv_saddr
;
6634 newinet
->inet_dport
= htons(asoc
->peer
.port
);
6635 newinet
->pmtudisc
= inet
->pmtudisc
;
6636 newinet
->inet_id
= asoc
->next_tsn
^ jiffies
;
6638 newinet
->uc_ttl
= inet
->uc_ttl
;
6639 newinet
->mc_loop
= 1;
6640 newinet
->mc_ttl
= 1;
6641 newinet
->mc_index
= 0;
6642 newinet
->mc_list
= NULL
;
6645 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6646 * and its messages to the newsk.
6648 static void sctp_sock_migrate(struct sock
*oldsk
, struct sock
*newsk
,
6649 struct sctp_association
*assoc
,
6650 sctp_socket_type_t type
)
6652 struct sctp_sock
*oldsp
= sctp_sk(oldsk
);
6653 struct sctp_sock
*newsp
= sctp_sk(newsk
);
6654 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
6655 struct sctp_endpoint
*newep
= newsp
->ep
;
6656 struct sk_buff
*skb
, *tmp
;
6657 struct sctp_ulpevent
*event
;
6658 struct sctp_bind_hashbucket
*head
;
6659 struct list_head tmplist
;
6661 /* Migrate socket buffer sizes and all the socket level options to the
6664 newsk
->sk_sndbuf
= oldsk
->sk_sndbuf
;
6665 newsk
->sk_rcvbuf
= oldsk
->sk_rcvbuf
;
6666 /* Brute force copy old sctp opt. */
6667 if (oldsp
->do_auto_asconf
) {
6668 memcpy(&tmplist
, &newsp
->auto_asconf_list
, sizeof(tmplist
));
6669 inet_sk_copy_descendant(newsk
, oldsk
);
6670 memcpy(&newsp
->auto_asconf_list
, &tmplist
, sizeof(tmplist
));
6672 inet_sk_copy_descendant(newsk
, oldsk
);
6674 /* Restore the ep value that was overwritten with the above structure
6680 /* Hook this new socket in to the bind_hash list. */
6681 head
= &sctp_port_hashtable
[sctp_phashfn(inet_sk(oldsk
)->inet_num
)];
6682 sctp_local_bh_disable();
6683 sctp_spin_lock(&head
->lock
);
6684 pp
= sctp_sk(oldsk
)->bind_hash
;
6685 sk_add_bind_node(newsk
, &pp
->owner
);
6686 sctp_sk(newsk
)->bind_hash
= pp
;
6687 inet_sk(newsk
)->inet_num
= inet_sk(oldsk
)->inet_num
;
6688 sctp_spin_unlock(&head
->lock
);
6689 sctp_local_bh_enable();
6691 /* Copy the bind_addr list from the original endpoint to the new
6692 * endpoint so that we can handle restarts properly
6694 sctp_bind_addr_dup(&newsp
->ep
->base
.bind_addr
,
6695 &oldsp
->ep
->base
.bind_addr
, GFP_KERNEL
);
6697 /* Move any messages in the old socket's receive queue that are for the
6698 * peeled off association to the new socket's receive queue.
6700 sctp_skb_for_each(skb
, &oldsk
->sk_receive_queue
, tmp
) {
6701 event
= sctp_skb2event(skb
);
6702 if (event
->asoc
== assoc
) {
6703 __skb_unlink(skb
, &oldsk
->sk_receive_queue
);
6704 __skb_queue_tail(&newsk
->sk_receive_queue
, skb
);
6705 sctp_skb_set_owner_r_frag(skb
, newsk
);
6709 /* Clean up any messages pending delivery due to partial
6710 * delivery. Three cases:
6711 * 1) No partial deliver; no work.
6712 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6713 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6715 skb_queue_head_init(&newsp
->pd_lobby
);
6716 atomic_set(&sctp_sk(newsk
)->pd_mode
, assoc
->ulpq
.pd_mode
);
6718 if (atomic_read(&sctp_sk(oldsk
)->pd_mode
)) {
6719 struct sk_buff_head
*queue
;
6721 /* Decide which queue to move pd_lobby skbs to. */
6722 if (assoc
->ulpq
.pd_mode
) {
6723 queue
= &newsp
->pd_lobby
;
6725 queue
= &newsk
->sk_receive_queue
;
6727 /* Walk through the pd_lobby, looking for skbs that
6728 * need moved to the new socket.
6730 sctp_skb_for_each(skb
, &oldsp
->pd_lobby
, tmp
) {
6731 event
= sctp_skb2event(skb
);
6732 if (event
->asoc
== assoc
) {
6733 __skb_unlink(skb
, &oldsp
->pd_lobby
);
6734 __skb_queue_tail(queue
, skb
);
6735 sctp_skb_set_owner_r_frag(skb
, newsk
);
6739 /* Clear up any skbs waiting for the partial
6740 * delivery to finish.
6742 if (assoc
->ulpq
.pd_mode
)
6743 sctp_clear_pd(oldsk
, NULL
);
6747 sctp_skb_for_each(skb
, &assoc
->ulpq
.reasm
, tmp
)
6748 sctp_skb_set_owner_r_frag(skb
, newsk
);
6750 sctp_skb_for_each(skb
, &assoc
->ulpq
.lobby
, tmp
)
6751 sctp_skb_set_owner_r_frag(skb
, newsk
);
6753 /* Set the type of socket to indicate that it is peeled off from the
6754 * original UDP-style socket or created with the accept() call on a
6755 * TCP-style socket..
6759 /* Mark the new socket "in-use" by the user so that any packets
6760 * that may arrive on the association after we've moved it are
6761 * queued to the backlog. This prevents a potential race between
6762 * backlog processing on the old socket and new-packet processing
6763 * on the new socket.
6765 * The caller has just allocated newsk so we can guarantee that other
6766 * paths won't try to lock it and then oldsk.
6768 lock_sock_nested(newsk
, SINGLE_DEPTH_NESTING
);
6769 sctp_assoc_migrate(assoc
, newsk
);
6771 /* If the association on the newsk is already closed before accept()
6772 * is called, set RCV_SHUTDOWN flag.
6774 if (sctp_state(assoc
, CLOSED
) && sctp_style(newsk
, TCP
))
6775 newsk
->sk_shutdown
|= RCV_SHUTDOWN
;
6777 newsk
->sk_state
= SCTP_SS_ESTABLISHED
;
6778 sctp_release_sock(newsk
);
6782 /* This proto struct describes the ULP interface for SCTP. */
6783 struct proto sctp_prot
= {
6785 .owner
= THIS_MODULE
,
6786 .close
= sctp_close
,
6787 .connect
= sctp_connect
,
6788 .disconnect
= sctp_disconnect
,
6789 .accept
= sctp_accept
,
6790 .ioctl
= sctp_ioctl
,
6791 .init
= sctp_init_sock
,
6792 .destroy
= sctp_destroy_sock
,
6793 .shutdown
= sctp_shutdown
,
6794 .setsockopt
= sctp_setsockopt
,
6795 .getsockopt
= sctp_getsockopt
,
6796 .sendmsg
= sctp_sendmsg
,
6797 .recvmsg
= sctp_recvmsg
,
6799 .backlog_rcv
= sctp_backlog_rcv
,
6801 .unhash
= sctp_unhash
,
6802 .get_port
= sctp_get_port
,
6803 .obj_size
= sizeof(struct sctp_sock
),
6804 .sysctl_mem
= sysctl_sctp_mem
,
6805 .sysctl_rmem
= sysctl_sctp_rmem
,
6806 .sysctl_wmem
= sysctl_sctp_wmem
,
6807 .memory_pressure
= &sctp_memory_pressure
,
6808 .enter_memory_pressure
= sctp_enter_memory_pressure
,
6809 .memory_allocated
= &sctp_memory_allocated
,
6810 .sockets_allocated
= &sctp_sockets_allocated
,
6813 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6815 struct proto sctpv6_prot
= {
6817 .owner
= THIS_MODULE
,
6818 .close
= sctp_close
,
6819 .connect
= sctp_connect
,
6820 .disconnect
= sctp_disconnect
,
6821 .accept
= sctp_accept
,
6822 .ioctl
= sctp_ioctl
,
6823 .init
= sctp_init_sock
,
6824 .destroy
= sctp_destroy_sock
,
6825 .shutdown
= sctp_shutdown
,
6826 .setsockopt
= sctp_setsockopt
,
6827 .getsockopt
= sctp_getsockopt
,
6828 .sendmsg
= sctp_sendmsg
,
6829 .recvmsg
= sctp_recvmsg
,
6831 .backlog_rcv
= sctp_backlog_rcv
,
6833 .unhash
= sctp_unhash
,
6834 .get_port
= sctp_get_port
,
6835 .obj_size
= sizeof(struct sctp6_sock
),
6836 .sysctl_mem
= sysctl_sctp_mem
,
6837 .sysctl_rmem
= sysctl_sctp_rmem
,
6838 .sysctl_wmem
= sysctl_sctp_wmem
,
6839 .memory_pressure
= &sctp_memory_pressure
,
6840 .enter_memory_pressure
= sctp_enter_memory_pressure
,
6841 .memory_allocated
= &sctp_memory_allocated
,
6842 .sockets_allocated
= &sctp_sockets_allocated
,
6844 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */