1 /* SCTP kernel reference 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 reference 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 * The SCTP reference 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 * The SCTP reference 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 #include <linux/types.h>
61 #include <linux/kernel.h>
62 #include <linux/wait.h>
63 #include <linux/time.h>
65 #include <linux/capability.h>
66 #include <linux/fcntl.h>
67 #include <linux/poll.h>
68 #include <linux/init.h>
69 #include <linux/crypto.h>
73 #include <net/route.h>
75 #include <net/inet_common.h>
77 #include <linux/socket.h> /* for sa_family_t */
79 #include <net/sctp/sctp.h>
80 #include <net/sctp/sm.h>
82 /* WARNING: Please do not remove the SCTP_STATIC attribute to
83 * any of the functions below as they are used to export functions
84 * used by a project regression testsuite.
87 /* Forward declarations for internal helper functions. */
88 static int sctp_writeable(struct sock
*sk
);
89 static void sctp_wfree(struct sk_buff
*skb
);
90 static int sctp_wait_for_sndbuf(struct sctp_association
*, long *timeo_p
,
92 static int sctp_wait_for_packet(struct sock
* sk
, int *err
, long *timeo_p
);
93 static int sctp_wait_for_connect(struct sctp_association
*, long *timeo_p
);
94 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
);
95 static void sctp_wait_for_close(struct sock
*sk
, long timeo
);
96 static struct sctp_af
*sctp_sockaddr_af(struct sctp_sock
*opt
,
97 union sctp_addr
*addr
, int len
);
98 static int sctp_bindx_add(struct sock
*, struct sockaddr
*, int);
99 static int sctp_bindx_rem(struct sock
*, struct sockaddr
*, int);
100 static int sctp_send_asconf_add_ip(struct sock
*, struct sockaddr
*, int);
101 static int sctp_send_asconf_del_ip(struct sock
*, struct sockaddr
*, int);
102 static int sctp_send_asconf(struct sctp_association
*asoc
,
103 struct sctp_chunk
*chunk
);
104 static int sctp_do_bind(struct sock
*, union sctp_addr
*, int);
105 static int sctp_autobind(struct sock
*sk
);
106 static void sctp_sock_migrate(struct sock
*, struct sock
*,
107 struct sctp_association
*, sctp_socket_type_t
);
108 static char *sctp_hmac_alg
= SCTP_COOKIE_HMAC_ALG
;
110 extern kmem_cache_t
*sctp_bucket_cachep
;
112 /* Get the sndbuf space available at the time on the association. */
113 static inline int sctp_wspace(struct sctp_association
*asoc
)
115 struct sock
*sk
= asoc
->base
.sk
;
118 if (asoc
->ep
->sndbuf_policy
) {
119 /* make sure that no association uses more than sk_sndbuf */
120 amt
= sk
->sk_sndbuf
- asoc
->sndbuf_used
;
122 /* do socket level accounting */
123 amt
= sk
->sk_sndbuf
- atomic_read(&sk
->sk_wmem_alloc
);
132 /* Increment the used sndbuf space count of the corresponding association by
133 * the size of the outgoing data chunk.
134 * Also, set the skb destructor for sndbuf accounting later.
136 * Since it is always 1-1 between chunk and skb, and also a new skb is always
137 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
138 * destructor in the data chunk skb for the purpose of the sndbuf space
141 static inline void sctp_set_owner_w(struct sctp_chunk
*chunk
)
143 struct sctp_association
*asoc
= chunk
->asoc
;
144 struct sock
*sk
= asoc
->base
.sk
;
146 /* The sndbuf space is tracked per association. */
147 sctp_association_hold(asoc
);
149 skb_set_owner_w(chunk
->skb
, sk
);
151 chunk
->skb
->destructor
= sctp_wfree
;
152 /* Save the chunk pointer in skb for sctp_wfree to use later. */
153 *((struct sctp_chunk
**)(chunk
->skb
->cb
)) = chunk
;
155 asoc
->sndbuf_used
+= SCTP_DATA_SNDSIZE(chunk
) +
156 sizeof(struct sk_buff
) +
157 sizeof(struct sctp_chunk
);
159 atomic_add(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
162 /* Verify that this is a valid address. */
163 static inline int sctp_verify_addr(struct sock
*sk
, union sctp_addr
*addr
,
168 /* Verify basic sockaddr. */
169 af
= sctp_sockaddr_af(sctp_sk(sk
), addr
, len
);
173 /* Is this a valid SCTP address? */
174 if (!af
->addr_valid(addr
, sctp_sk(sk
), NULL
))
177 if (!sctp_sk(sk
)->pf
->send_verify(sctp_sk(sk
), (addr
)))
183 /* Look up the association by its id. If this is not a UDP-style
184 * socket, the ID field is always ignored.
186 struct sctp_association
*sctp_id2assoc(struct sock
*sk
, sctp_assoc_t id
)
188 struct sctp_association
*asoc
= NULL
;
190 /* If this is not a UDP-style socket, assoc id should be ignored. */
191 if (!sctp_style(sk
, UDP
)) {
192 /* Return NULL if the socket state is not ESTABLISHED. It
193 * could be a TCP-style listening socket or a socket which
194 * hasn't yet called connect() to establish an association.
196 if (!sctp_sstate(sk
, ESTABLISHED
))
199 /* Get the first and the only association from the list. */
200 if (!list_empty(&sctp_sk(sk
)->ep
->asocs
))
201 asoc
= list_entry(sctp_sk(sk
)->ep
->asocs
.next
,
202 struct sctp_association
, asocs
);
206 /* Otherwise this is a UDP-style socket. */
207 if (!id
|| (id
== (sctp_assoc_t
)-1))
210 spin_lock_bh(&sctp_assocs_id_lock
);
211 asoc
= (struct sctp_association
*)idr_find(&sctp_assocs_id
, (int)id
);
212 spin_unlock_bh(&sctp_assocs_id_lock
);
214 if (!asoc
|| (asoc
->base
.sk
!= sk
) || asoc
->base
.dead
)
220 /* Look up the transport from an address and an assoc id. If both address and
221 * id are specified, the associations matching the address and the id should be
224 static struct sctp_transport
*sctp_addr_id2transport(struct sock
*sk
,
225 struct sockaddr_storage
*addr
,
228 struct sctp_association
*addr_asoc
= NULL
, *id_asoc
= NULL
;
229 struct sctp_transport
*transport
;
230 union sctp_addr
*laddr
= (union sctp_addr
*)addr
;
233 flip_to_h(&tmp
, laddr
);
234 addr_asoc
= sctp_endpoint_lookup_assoc(sctp_sk(sk
)->ep
,
241 id_asoc
= sctp_id2assoc(sk
, id
);
242 if (id_asoc
&& (id_asoc
!= addr_asoc
))
245 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
246 (union sctp_addr
*)addr
);
251 /* API 3.1.2 bind() - UDP Style Syntax
252 * The syntax of bind() is,
254 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
256 * sd - the socket descriptor returned by socket().
257 * addr - the address structure (struct sockaddr_in or struct
258 * sockaddr_in6 [RFC 2553]),
259 * addr_len - the size of the address structure.
261 SCTP_STATIC
int sctp_bind(struct sock
*sk
, struct sockaddr
*addr
, int addr_len
)
267 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
270 /* Disallow binding twice. */
271 if (!sctp_sk(sk
)->ep
->base
.bind_addr
.port
)
272 retval
= sctp_do_bind(sk
, (union sctp_addr
*)addr
,
277 sctp_release_sock(sk
);
282 static long sctp_get_port_local(struct sock
*, union sctp_addr
*);
284 /* Verify this is a valid sockaddr. */
285 static struct sctp_af
*sctp_sockaddr_af(struct sctp_sock
*opt
,
286 union sctp_addr
*addr
, int len
)
290 /* Check minimum size. */
291 if (len
< sizeof (struct sockaddr
))
294 /* Does this PF support this AF? */
295 if (!opt
->pf
->af_supported(addr
->sa
.sa_family
, opt
))
298 /* If we get this far, af is valid. */
299 af
= sctp_get_af_specific(addr
->sa
.sa_family
);
301 if (len
< af
->sockaddr_len
)
307 /* Bind a local address either to an endpoint or to an association. */
308 SCTP_STATIC
int sctp_do_bind(struct sock
*sk
, union sctp_addr
*addr
, int len
)
310 struct sctp_sock
*sp
= sctp_sk(sk
);
311 struct sctp_endpoint
*ep
= sp
->ep
;
312 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
318 /* Common sockaddr verification. */
319 af
= sctp_sockaddr_af(sp
, addr
, len
);
321 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
326 snum
= ntohs(addr
->v4
.sin_port
);
328 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
329 ", port: %d, new port: %d, len: %d)\n",
335 /* PF specific bind() address verification. */
336 if (!sp
->pf
->bind_verify(sp
, addr
))
337 return -EADDRNOTAVAIL
;
339 /* We must either be unbound, or bind to the same port. */
340 if (bp
->port
&& (snum
!= bp
->port
)) {
341 SCTP_DEBUG_PRINTK("sctp_do_bind:"
342 " New port %d does not match existing port "
343 "%d.\n", snum
, bp
->port
);
347 if (snum
&& snum
< PROT_SOCK
&& !capable(CAP_NET_BIND_SERVICE
))
350 /* Make sure we are allowed to bind here.
351 * The function sctp_get_port_local() does duplicate address
354 if ((ret
= sctp_get_port_local(sk
, addr
))) {
355 if (ret
== (long) sk
) {
356 /* This endpoint has a conflicting address. */
363 /* Refresh ephemeral port. */
365 bp
->port
= inet_sk(sk
)->num
;
367 /* Add the address to the bind address list. */
368 sctp_local_bh_disable();
369 sctp_write_lock(&ep
->base
.addr_lock
);
371 /* Use GFP_ATOMIC since BHs are disabled. */
372 flip_to_h(&tmp
, addr
);
373 ret
= sctp_add_bind_addr(bp
, &tmp
, 1, GFP_ATOMIC
);
374 sctp_write_unlock(&ep
->base
.addr_lock
);
375 sctp_local_bh_enable();
377 /* Copy back into socket for getsockname() use. */
379 inet_sk(sk
)->sport
= htons(inet_sk(sk
)->num
);
380 af
->to_sk_saddr(addr
, sk
);
386 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
388 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
389 * at any one time. If a sender, after sending an ASCONF chunk, decides
390 * it needs to transfer another ASCONF Chunk, it MUST wait until the
391 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
392 * subsequent ASCONF. Note this restriction binds each side, so at any
393 * time two ASCONF may be in-transit on any given association (one sent
394 * from each endpoint).
396 static int sctp_send_asconf(struct sctp_association
*asoc
,
397 struct sctp_chunk
*chunk
)
401 /* If there is an outstanding ASCONF chunk, queue it for later
404 if (asoc
->addip_last_asconf
) {
405 list_add_tail(&chunk
->list
, &asoc
->addip_chunk_list
);
409 /* Hold the chunk until an ASCONF_ACK is received. */
410 sctp_chunk_hold(chunk
);
411 retval
= sctp_primitive_ASCONF(asoc
, chunk
);
413 sctp_chunk_free(chunk
);
415 asoc
->addip_last_asconf
= chunk
;
421 /* Add a list of addresses as bind addresses to local endpoint or
424 * Basically run through each address specified in the addrs/addrcnt
425 * array/length pair, determine if it is IPv6 or IPv4 and call
426 * sctp_do_bind() on it.
428 * If any of them fails, then the operation will be reversed and the
429 * ones that were added will be removed.
431 * Only sctp_setsockopt_bindx() is supposed to call this function.
433 int sctp_bindx_add(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
438 struct sockaddr
*sa_addr
;
441 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
445 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
446 /* The list may contain either IPv4 or IPv6 address;
447 * determine the address length for walking thru the list.
449 sa_addr
= (struct sockaddr
*)addr_buf
;
450 af
= sctp_get_af_specific(sa_addr
->sa_family
);
456 retval
= sctp_do_bind(sk
, (union sctp_addr
*)sa_addr
,
459 addr_buf
+= af
->sockaddr_len
;
463 /* Failed. Cleanup the ones that have been added */
465 sctp_bindx_rem(sk
, addrs
, cnt
);
473 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
474 * associations that are part of the endpoint indicating that a list of local
475 * addresses are added to the endpoint.
477 * If any of the addresses is already in the bind address list of the
478 * association, we do not send the chunk for that association. But it will not
479 * affect other associations.
481 * Only sctp_setsockopt_bindx() is supposed to call this function.
483 static int sctp_send_asconf_add_ip(struct sock
*sk
,
484 struct sockaddr
*addrs
,
487 struct sctp_sock
*sp
;
488 struct sctp_endpoint
*ep
;
489 struct sctp_association
*asoc
;
490 struct sctp_bind_addr
*bp
;
491 struct sctp_chunk
*chunk
;
492 struct sctp_sockaddr_entry
*laddr
;
493 union sctp_addr
*addr
;
494 union sctp_addr saveaddr
;
497 struct list_head
*pos
;
502 if (!sctp_addip_enable
)
508 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
509 __FUNCTION__
, sk
, addrs
, addrcnt
);
511 list_for_each(pos
, &ep
->asocs
) {
512 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
514 if (!asoc
->peer
.asconf_capable
)
517 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_ADD_IP
)
520 if (!sctp_state(asoc
, ESTABLISHED
))
523 /* Check if any address in the packed array of addresses is
524 * in the bind address list of the association. If so,
525 * do not send the asconf chunk to its peer, but continue with
526 * other associations.
529 for (i
= 0; i
< addrcnt
; i
++) {
530 addr
= (union sctp_addr
*)addr_buf
;
531 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
537 if (sctp_assoc_lookup_laddr(asoc
, addr
))
540 addr_buf
+= af
->sockaddr_len
;
545 /* Use the first address in bind addr list of association as
546 * Address Parameter of ASCONF CHUNK.
548 sctp_read_lock(&asoc
->base
.addr_lock
);
549 bp
= &asoc
->base
.bind_addr
;
550 p
= bp
->address_list
.next
;
551 laddr
= list_entry(p
, struct sctp_sockaddr_entry
, list
);
552 sctp_read_unlock(&asoc
->base
.addr_lock
);
554 chunk
= sctp_make_asconf_update_ip(asoc
, &laddr
->a_h
, addrs
,
555 addrcnt
, SCTP_PARAM_ADD_IP
);
561 retval
= sctp_send_asconf(asoc
, chunk
);
565 /* Add the new addresses to the bind address list with
566 * use_as_src set to 0.
568 sctp_local_bh_disable();
569 sctp_write_lock(&asoc
->base
.addr_lock
);
571 for (i
= 0; i
< addrcnt
; i
++) {
572 addr
= (union sctp_addr
*)addr_buf
;
573 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
574 memcpy(&saveaddr
, addr
, af
->sockaddr_len
);
575 saveaddr
.v4
.sin_port
= ntohs(saveaddr
.v4
.sin_port
);
576 retval
= sctp_add_bind_addr(bp
, &saveaddr
, 0,
578 addr_buf
+= af
->sockaddr_len
;
580 sctp_write_unlock(&asoc
->base
.addr_lock
);
581 sctp_local_bh_enable();
588 /* Remove a list of addresses from bind addresses list. Do not remove the
591 * Basically run through each address specified in the addrs/addrcnt
592 * array/length pair, determine if it is IPv6 or IPv4 and call
593 * sctp_del_bind() on it.
595 * If any of them fails, then the operation will be reversed and the
596 * ones that were removed will be added back.
598 * At least one address has to be left; if only one address is
599 * available, the operation will return -EBUSY.
601 * Only sctp_setsockopt_bindx() is supposed to call this function.
603 int sctp_bindx_rem(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
605 struct sctp_sock
*sp
= sctp_sk(sk
);
606 struct sctp_endpoint
*ep
= sp
->ep
;
608 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
610 union sctp_addr saveaddr
;
612 struct sockaddr
*sa_addr
;
615 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
619 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
620 /* If the bind address list is empty or if there is only one
621 * bind address, there is nothing more to be removed (we need
622 * at least one address here).
624 if (list_empty(&bp
->address_list
) ||
625 (sctp_list_single_entry(&bp
->address_list
))) {
630 /* The list may contain either IPv4 or IPv6 address;
631 * determine the address length to copy the address to
634 sa_addr
= (struct sockaddr
*)addr_buf
;
635 af
= sctp_get_af_specific(sa_addr
->sa_family
);
640 memcpy(&saveaddr
, sa_addr
, af
->sockaddr_len
);
641 saveaddr
.v4
.sin_port
= ntohs(saveaddr
.v4
.sin_port
);
642 if (saveaddr
.v4
.sin_port
!= bp
->port
) {
647 /* FIXME - There is probably a need to check if sk->sk_saddr and
648 * sk->sk_rcv_addr are currently set to one of the addresses to
649 * be removed. This is something which needs to be looked into
650 * when we are fixing the outstanding issues with multi-homing
651 * socket routing and failover schemes. Refer to comments in
652 * sctp_do_bind(). -daisy
654 sctp_local_bh_disable();
655 sctp_write_lock(&ep
->base
.addr_lock
);
657 retval
= sctp_del_bind_addr(bp
, &saveaddr
);
659 sctp_write_unlock(&ep
->base
.addr_lock
);
660 sctp_local_bh_enable();
662 addr_buf
+= af
->sockaddr_len
;
665 /* Failed. Add the ones that has been removed back */
667 sctp_bindx_add(sk
, addrs
, cnt
);
675 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
676 * the associations that are part of the endpoint indicating that a list of
677 * local addresses are removed from the endpoint.
679 * If any of the addresses is already in the bind address list of the
680 * association, we do not send the chunk for that association. But it will not
681 * affect other associations.
683 * Only sctp_setsockopt_bindx() is supposed to call this function.
685 static int sctp_send_asconf_del_ip(struct sock
*sk
,
686 struct sockaddr
*addrs
,
689 struct sctp_sock
*sp
;
690 struct sctp_endpoint
*ep
;
691 struct sctp_association
*asoc
;
692 struct sctp_transport
*transport
;
693 struct sctp_bind_addr
*bp
;
694 struct sctp_chunk
*chunk
;
695 union sctp_addr
*laddr
;
698 struct list_head
*pos
, *pos1
;
699 struct sctp_sockaddr_entry
*saddr
;
703 if (!sctp_addip_enable
)
709 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
710 __FUNCTION__
, sk
, addrs
, addrcnt
);
712 list_for_each(pos
, &ep
->asocs
) {
713 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
715 if (!asoc
->peer
.asconf_capable
)
718 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_DEL_IP
)
721 if (!sctp_state(asoc
, ESTABLISHED
))
724 /* Check if any address in the packed array of addresses is
725 * not present in the bind address list of the association.
726 * If so, do not send the asconf chunk to its peer, but
727 * continue with other associations.
730 for (i
= 0; i
< addrcnt
; i
++) {
731 laddr
= (union sctp_addr
*)addr_buf
;
732 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
738 if (!sctp_assoc_lookup_laddr(asoc
, laddr
))
741 addr_buf
+= af
->sockaddr_len
;
746 /* Find one address in the association's bind address list
747 * that is not in the packed array of addresses. This is to
748 * make sure that we do not delete all the addresses in the
751 sctp_read_lock(&asoc
->base
.addr_lock
);
752 bp
= &asoc
->base
.bind_addr
;
753 laddr
= sctp_find_unmatch_addr(bp
, (union sctp_addr
*)addrs
,
755 sctp_read_unlock(&asoc
->base
.addr_lock
);
759 chunk
= sctp_make_asconf_update_ip(asoc
, laddr
, addrs
, addrcnt
,
766 /* Reset use_as_src flag for the addresses in the bind address
767 * list that are to be deleted.
769 sctp_local_bh_disable();
770 sctp_write_lock(&asoc
->base
.addr_lock
);
772 for (i
= 0; i
< addrcnt
; i
++) {
773 laddr
= (union sctp_addr
*)addr_buf
;
774 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
775 list_for_each(pos1
, &bp
->address_list
) {
776 saddr
= list_entry(pos1
,
777 struct sctp_sockaddr_entry
,
779 if (sctp_cmp_addr_exact(&saddr
->a
, laddr
))
780 saddr
->use_as_src
= 0;
782 addr_buf
+= af
->sockaddr_len
;
784 sctp_write_unlock(&asoc
->base
.addr_lock
);
785 sctp_local_bh_enable();
787 /* Update the route and saddr entries for all the transports
788 * as some of the addresses in the bind address list are
789 * about to be deleted and cannot be used as source addresses.
791 list_for_each(pos1
, &asoc
->peer
.transport_addr_list
) {
792 transport
= list_entry(pos1
, struct sctp_transport
,
794 dst_release(transport
->dst
);
795 sctp_transport_route(transport
, NULL
,
796 sctp_sk(asoc
->base
.sk
));
799 retval
= sctp_send_asconf(asoc
, chunk
);
805 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
808 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
811 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
812 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
815 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
816 * Section 3.1.2 for this usage.
818 * addrs is a pointer to an array of one or more socket addresses. Each
819 * address is contained in its appropriate structure (i.e. struct
820 * sockaddr_in or struct sockaddr_in6) the family of the address type
821 * must be used to distinguish the address length (note that this
822 * representation is termed a "packed array" of addresses). The caller
823 * specifies the number of addresses in the array with addrcnt.
825 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
826 * -1, and sets errno to the appropriate error code.
828 * For SCTP, the port given in each socket address must be the same, or
829 * sctp_bindx() will fail, setting errno to EINVAL.
831 * The flags parameter is formed from the bitwise OR of zero or more of
832 * the following currently defined flags:
834 * SCTP_BINDX_ADD_ADDR
836 * SCTP_BINDX_REM_ADDR
838 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
839 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
840 * addresses from the association. The two flags are mutually exclusive;
841 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
842 * not remove all addresses from an association; sctp_bindx() will
843 * reject such an attempt with EINVAL.
845 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
846 * additional addresses with an endpoint after calling bind(). Or use
847 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
848 * socket is associated with so that no new association accepted will be
849 * associated with those addresses. If the endpoint supports dynamic
850 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
851 * endpoint to send the appropriate message to the peer to change the
852 * peers address lists.
854 * Adding and removing addresses from a connected association is
855 * optional functionality. Implementations that do not support this
856 * functionality should return EOPNOTSUPP.
858 * Basically do nothing but copying the addresses from user to kernel
859 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
860 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
863 * We don't use copy_from_user() for optimization: we first do the
864 * sanity checks (buffer size -fast- and access check-healthy
865 * pointer); if all of those succeed, then we can alloc the memory
866 * (expensive operation) needed to copy the data to kernel. Then we do
867 * the copying without checking the user space area
868 * (__copy_from_user()).
870 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
873 * sk The sk of the socket
874 * addrs The pointer to the addresses in user land
875 * addrssize Size of the addrs buffer
876 * op Operation to perform (add or remove, see the flags of
879 * Returns 0 if ok, <0 errno code on error.
881 SCTP_STATIC
int sctp_setsockopt_bindx(struct sock
* sk
,
882 struct sockaddr __user
*addrs
,
883 int addrs_size
, int op
)
885 struct sockaddr
*kaddrs
;
889 struct sockaddr
*sa_addr
;
893 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
894 " addrs_size %d opt %d\n", sk
, addrs
, addrs_size
, op
);
896 if (unlikely(addrs_size
<= 0))
899 /* Check the user passed a healthy pointer. */
900 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
903 /* Alloc space for the address array in kernel memory. */
904 kaddrs
= kmalloc(addrs_size
, GFP_KERNEL
);
905 if (unlikely(!kaddrs
))
908 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
913 /* Walk through the addrs buffer and count the number of addresses. */
915 while (walk_size
< addrs_size
) {
916 sa_addr
= (struct sockaddr
*)addr_buf
;
917 af
= sctp_get_af_specific(sa_addr
->sa_family
);
919 /* If the address family is not supported or if this address
920 * causes the address buffer to overflow return EINVAL.
922 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
927 addr_buf
+= af
->sockaddr_len
;
928 walk_size
+= af
->sockaddr_len
;
933 case SCTP_BINDX_ADD_ADDR
:
934 err
= sctp_bindx_add(sk
, kaddrs
, addrcnt
);
937 err
= sctp_send_asconf_add_ip(sk
, kaddrs
, addrcnt
);
940 case SCTP_BINDX_REM_ADDR
:
941 err
= sctp_bindx_rem(sk
, kaddrs
, addrcnt
);
944 err
= sctp_send_asconf_del_ip(sk
, kaddrs
, addrcnt
);
958 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
960 * Common routine for handling connect() and sctp_connectx().
961 * Connect will come in with just a single address.
963 static int __sctp_connect(struct sock
* sk
,
964 struct sockaddr
*kaddrs
,
967 struct sctp_sock
*sp
;
968 struct sctp_endpoint
*ep
;
969 struct sctp_association
*asoc
= NULL
;
970 struct sctp_association
*asoc2
;
971 struct sctp_transport
*transport
;
979 struct sockaddr
*sa_addr
;
985 /* connect() cannot be done on a socket that is already in ESTABLISHED
986 * state - UDP-style peeled off socket or a TCP-style socket that
987 * is already connected.
988 * It cannot be done even on a TCP-style listening socket.
990 if (sctp_sstate(sk
, ESTABLISHED
) ||
991 (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))) {
996 /* Walk through the addrs buffer and count the number of addresses. */
998 while (walk_size
< addrs_size
) {
999 sa_addr
= (struct sockaddr
*)addr_buf
;
1000 af
= sctp_get_af_specific(sa_addr
->sa_family
);
1002 /* If the address family is not supported or if this address
1003 * causes the address buffer to overflow return EINVAL.
1005 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
1010 err
= sctp_verify_addr(sk
, (union sctp_addr
*)sa_addr
,
1015 memcpy(&to
, sa_addr
, af
->sockaddr_len
);
1016 to
.v4
.sin_port
= ntohs(to
.v4
.sin_port
);
1018 /* Check if there already is a matching association on the
1019 * endpoint (other than the one created here).
1021 asoc2
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1022 if (asoc2
&& asoc2
!= asoc
) {
1023 if (asoc2
->state
>= SCTP_STATE_ESTABLISHED
)
1030 /* If we could not find a matching association on the endpoint,
1031 * make sure that there is no peeled-off association matching
1032 * the peer address even on another socket.
1034 if (sctp_endpoint_is_peeled_off(ep
, &to
)) {
1035 err
= -EADDRNOTAVAIL
;
1040 /* If a bind() or sctp_bindx() is not called prior to
1041 * an sctp_connectx() call, the system picks an
1042 * ephemeral port and will choose an address set
1043 * equivalent to binding with a wildcard address.
1045 if (!ep
->base
.bind_addr
.port
) {
1046 if (sctp_autobind(sk
)) {
1052 * If an unprivileged user inherits a 1-many
1053 * style socket with open associations on a
1054 * privileged port, it MAY be permitted to
1055 * accept new associations, but it SHOULD NOT
1056 * be permitted to open new associations.
1058 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1059 !capable(CAP_NET_BIND_SERVICE
)) {
1065 scope
= sctp_scope(&to
);
1066 asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1073 /* Prime the peer's transport structures. */
1074 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
,
1082 addr_buf
+= af
->sockaddr_len
;
1083 walk_size
+= af
->sockaddr_len
;
1086 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, GFP_KERNEL
);
1091 err
= sctp_primitive_ASSOCIATE(asoc
, NULL
);
1096 /* Initialize sk's dport and daddr for getpeername() */
1097 inet_sk(sk
)->dport
= htons(asoc
->peer
.port
);
1098 af
= sctp_get_af_specific(to
.sa
.sa_family
);
1099 af
->to_sk_daddr(&to
, sk
);
1102 timeo
= sock_sndtimeo(sk
, sk
->sk_socket
->file
->f_flags
& O_NONBLOCK
);
1103 err
= sctp_wait_for_connect(asoc
, &timeo
);
1105 /* Don't free association on exit. */
1110 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1111 " kaddrs: %p err: %d\n",
1114 sctp_association_free(asoc
);
1118 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1121 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt);
1123 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1124 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1125 * or IPv6 addresses.
1127 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1128 * Section 3.1.2 for this usage.
1130 * addrs is a pointer to an array of one or more socket addresses. Each
1131 * address is contained in its appropriate structure (i.e. struct
1132 * sockaddr_in or struct sockaddr_in6) the family of the address type
1133 * must be used to distengish the address length (note that this
1134 * representation is termed a "packed array" of addresses). The caller
1135 * specifies the number of addresses in the array with addrcnt.
1137 * On success, sctp_connectx() returns 0. On failure, sctp_connectx() returns
1138 * -1, and sets errno to the appropriate error code.
1140 * For SCTP, the port given in each socket address must be the same, or
1141 * sctp_connectx() will fail, setting errno to EINVAL.
1143 * An application can use sctp_connectx to initiate an association with
1144 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1145 * allows a caller to specify multiple addresses at which a peer can be
1146 * reached. The way the SCTP stack uses the list of addresses to set up
1147 * the association is implementation dependant. This function only
1148 * specifies that the stack will try to make use of all the addresses in
1149 * the list when needed.
1151 * Note that the list of addresses passed in is only used for setting up
1152 * the association. It does not necessarily equal the set of addresses
1153 * the peer uses for the resulting association. If the caller wants to
1154 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1155 * retrieve them after the association has been set up.
1157 * Basically do nothing but copying the addresses from user to kernel
1158 * land and invoking either sctp_connectx(). This is used for tunneling
1159 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1161 * We don't use copy_from_user() for optimization: we first do the
1162 * sanity checks (buffer size -fast- and access check-healthy
1163 * pointer); if all of those succeed, then we can alloc the memory
1164 * (expensive operation) needed to copy the data to kernel. Then we do
1165 * the copying without checking the user space area
1166 * (__copy_from_user()).
1168 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1171 * sk The sk of the socket
1172 * addrs The pointer to the addresses in user land
1173 * addrssize Size of the addrs buffer
1175 * Returns 0 if ok, <0 errno code on error.
1177 SCTP_STATIC
int sctp_setsockopt_connectx(struct sock
* sk
,
1178 struct sockaddr __user
*addrs
,
1182 struct sockaddr
*kaddrs
;
1184 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1185 __FUNCTION__
, sk
, addrs
, addrs_size
);
1187 if (unlikely(addrs_size
<= 0))
1190 /* Check the user passed a healthy pointer. */
1191 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
1194 /* Alloc space for the address array in kernel memory. */
1195 kaddrs
= kmalloc(addrs_size
, GFP_KERNEL
);
1196 if (unlikely(!kaddrs
))
1199 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
1202 err
= __sctp_connect(sk
, kaddrs
, addrs_size
);
1209 /* API 3.1.4 close() - UDP Style Syntax
1210 * Applications use close() to perform graceful shutdown (as described in
1211 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1212 * by a UDP-style socket.
1216 * ret = close(int sd);
1218 * sd - the socket descriptor of the associations to be closed.
1220 * To gracefully shutdown a specific association represented by the
1221 * UDP-style socket, an application should use the sendmsg() call,
1222 * passing no user data, but including the appropriate flag in the
1223 * ancillary data (see Section xxxx).
1225 * If sd in the close() call is a branched-off socket representing only
1226 * one association, the shutdown is performed on that association only.
1228 * 4.1.6 close() - TCP Style Syntax
1230 * Applications use close() to gracefully close down an association.
1234 * int close(int sd);
1236 * sd - the socket descriptor of the association to be closed.
1238 * After an application calls close() on a socket descriptor, no further
1239 * socket operations will succeed on that descriptor.
1241 * API 7.1.4 SO_LINGER
1243 * An application using the TCP-style socket can use this option to
1244 * perform the SCTP ABORT primitive. The linger option structure is:
1247 * int l_onoff; // option on/off
1248 * int l_linger; // linger time
1251 * To enable the option, set l_onoff to 1. If the l_linger value is set
1252 * to 0, calling close() is the same as the ABORT primitive. If the
1253 * value is set to a negative value, the setsockopt() call will return
1254 * an error. If the value is set to a positive value linger_time, the
1255 * close() can be blocked for at most linger_time ms. If the graceful
1256 * shutdown phase does not finish during this period, close() will
1257 * return but the graceful shutdown phase continues in the system.
1259 SCTP_STATIC
void sctp_close(struct sock
*sk
, long timeout
)
1261 struct sctp_endpoint
*ep
;
1262 struct sctp_association
*asoc
;
1263 struct list_head
*pos
, *temp
;
1265 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk
, timeout
);
1268 sk
->sk_shutdown
= SHUTDOWN_MASK
;
1270 ep
= sctp_sk(sk
)->ep
;
1272 /* Walk all associations on an endpoint. */
1273 list_for_each_safe(pos
, temp
, &ep
->asocs
) {
1274 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
1276 if (sctp_style(sk
, TCP
)) {
1277 /* A closed association can still be in the list if
1278 * it belongs to a TCP-style listening socket that is
1279 * not yet accepted. If so, free it. If not, send an
1280 * ABORT or SHUTDOWN based on the linger options.
1282 if (sctp_state(asoc
, CLOSED
)) {
1283 sctp_unhash_established(asoc
);
1284 sctp_association_free(asoc
);
1289 if (sock_flag(sk
, SOCK_LINGER
) && !sk
->sk_lingertime
) {
1290 struct sctp_chunk
*chunk
;
1292 chunk
= sctp_make_abort_user(asoc
, NULL
, 0);
1294 sctp_primitive_ABORT(asoc
, chunk
);
1296 sctp_primitive_SHUTDOWN(asoc
, NULL
);
1299 /* Clean up any skbs sitting on the receive queue. */
1300 sctp_queue_purge_ulpevents(&sk
->sk_receive_queue
);
1301 sctp_queue_purge_ulpevents(&sctp_sk(sk
)->pd_lobby
);
1303 /* On a TCP-style socket, block for at most linger_time if set. */
1304 if (sctp_style(sk
, TCP
) && timeout
)
1305 sctp_wait_for_close(sk
, timeout
);
1307 /* This will run the backlog queue. */
1308 sctp_release_sock(sk
);
1310 /* Supposedly, no process has access to the socket, but
1311 * the net layers still may.
1313 sctp_local_bh_disable();
1314 sctp_bh_lock_sock(sk
);
1316 /* Hold the sock, since sk_common_release() will put sock_put()
1317 * and we have just a little more cleanup.
1320 sk_common_release(sk
);
1322 sctp_bh_unlock_sock(sk
);
1323 sctp_local_bh_enable();
1327 SCTP_DBG_OBJCNT_DEC(sock
);
1330 /* Handle EPIPE error. */
1331 static int sctp_error(struct sock
*sk
, int flags
, int err
)
1334 err
= sock_error(sk
) ? : -EPIPE
;
1335 if (err
== -EPIPE
&& !(flags
& MSG_NOSIGNAL
))
1336 send_sig(SIGPIPE
, current
, 0);
1340 /* API 3.1.3 sendmsg() - UDP Style Syntax
1342 * An application uses sendmsg() and recvmsg() calls to transmit data to
1343 * and receive data from its peer.
1345 * ssize_t sendmsg(int socket, const struct msghdr *message,
1348 * socket - the socket descriptor of the endpoint.
1349 * message - pointer to the msghdr structure which contains a single
1350 * user message and possibly some ancillary data.
1352 * See Section 5 for complete description of the data
1355 * flags - flags sent or received with the user message, see Section
1356 * 5 for complete description of the flags.
1358 * Note: This function could use a rewrite especially when explicit
1359 * connect support comes in.
1361 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1363 SCTP_STATIC
int sctp_msghdr_parse(const struct msghdr
*, sctp_cmsgs_t
*);
1365 SCTP_STATIC
int sctp_sendmsg(struct kiocb
*iocb
, struct sock
*sk
,
1366 struct msghdr
*msg
, size_t msg_len
)
1368 struct sctp_sock
*sp
;
1369 struct sctp_endpoint
*ep
;
1370 struct sctp_association
*new_asoc
=NULL
, *asoc
=NULL
;
1371 struct sctp_transport
*transport
, *chunk_tp
;
1372 struct sctp_chunk
*chunk
;
1373 union sctp_addr to
, tmp
;
1374 struct sockaddr
*msg_name
= NULL
;
1375 struct sctp_sndrcvinfo default_sinfo
= { 0 };
1376 struct sctp_sndrcvinfo
*sinfo
;
1377 struct sctp_initmsg
*sinit
;
1378 sctp_assoc_t associd
= 0;
1379 sctp_cmsgs_t cmsgs
= { NULL
};
1383 __u16 sinfo_flags
= 0;
1384 struct sctp_datamsg
*datamsg
;
1385 struct list_head
*pos
;
1386 int msg_flags
= msg
->msg_flags
;
1388 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1395 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep
);
1397 /* We cannot send a message over a TCP-style listening socket. */
1398 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
)) {
1403 /* Parse out the SCTP CMSGs. */
1404 err
= sctp_msghdr_parse(msg
, &cmsgs
);
1407 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err
);
1411 /* Fetch the destination address for this packet. This
1412 * address only selects the association--it is not necessarily
1413 * the address we will send to.
1414 * For a peeled-off socket, msg_name is ignored.
1416 if (!sctp_style(sk
, UDP_HIGH_BANDWIDTH
) && msg
->msg_name
) {
1417 int msg_namelen
= msg
->msg_namelen
;
1419 err
= sctp_verify_addr(sk
, (union sctp_addr
*)msg
->msg_name
,
1424 if (msg_namelen
> sizeof(to
))
1425 msg_namelen
= sizeof(to
);
1426 memcpy(&to
, msg
->msg_name
, msg_namelen
);
1427 memcpy(&tmp
, msg
->msg_name
, msg_namelen
);
1428 SCTP_DEBUG_PRINTK("Just memcpy'd. msg_name is "
1430 to
.v4
.sin_addr
.s_addr
, to
.v4
.sin_port
);
1432 to
.v4
.sin_port
= ntohs(to
.v4
.sin_port
);
1433 msg_name
= msg
->msg_name
;
1439 /* Did the user specify SNDRCVINFO? */
1441 sinfo_flags
= sinfo
->sinfo_flags
;
1442 associd
= sinfo
->sinfo_assoc_id
;
1445 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1446 msg_len
, sinfo_flags
);
1448 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1449 if (sctp_style(sk
, TCP
) && (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
))) {
1454 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1455 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1456 * If SCTP_ABORT is set, the message length could be non zero with
1457 * the msg_iov set to the user abort reason.
1459 if (((sinfo_flags
& SCTP_EOF
) && (msg_len
> 0)) ||
1460 (!(sinfo_flags
& (SCTP_EOF
|SCTP_ABORT
)) && (msg_len
== 0))) {
1465 /* If SCTP_ADDR_OVER is set, there must be an address
1466 * specified in msg_name.
1468 if ((sinfo_flags
& SCTP_ADDR_OVER
) && (!msg
->msg_name
)) {
1475 SCTP_DEBUG_PRINTK("About to look up association.\n");
1479 /* If a msg_name has been specified, assume this is to be used. */
1481 /* Look for a matching association on the endpoint. */
1482 asoc
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1484 /* If we could not find a matching association on the
1485 * endpoint, make sure that it is not a TCP-style
1486 * socket that already has an association or there is
1487 * no peeled-off association on another socket.
1489 if ((sctp_style(sk
, TCP
) &&
1490 sctp_sstate(sk
, ESTABLISHED
)) ||
1491 sctp_endpoint_is_peeled_off(ep
, &to
)) {
1492 err
= -EADDRNOTAVAIL
;
1497 asoc
= sctp_id2assoc(sk
, associd
);
1505 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc
);
1507 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1508 * socket that has an association in CLOSED state. This can
1509 * happen when an accepted socket has an association that is
1512 if (sctp_state(asoc
, CLOSED
) && sctp_style(sk
, TCP
)) {
1517 if (sinfo_flags
& SCTP_EOF
) {
1518 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1520 sctp_primitive_SHUTDOWN(asoc
, NULL
);
1524 if (sinfo_flags
& SCTP_ABORT
) {
1525 struct sctp_chunk
*chunk
;
1527 chunk
= sctp_make_abort_user(asoc
, msg
, msg_len
);
1533 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc
);
1534 sctp_primitive_ABORT(asoc
, chunk
);
1540 /* Do we need to create the association? */
1542 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1544 if (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
)) {
1549 /* Check for invalid stream against the stream counts,
1550 * either the default or the user specified stream counts.
1553 if (!sinit
|| (sinit
&& !sinit
->sinit_num_ostreams
)) {
1554 /* Check against the defaults. */
1555 if (sinfo
->sinfo_stream
>=
1556 sp
->initmsg
.sinit_num_ostreams
) {
1561 /* Check against the requested. */
1562 if (sinfo
->sinfo_stream
>=
1563 sinit
->sinit_num_ostreams
) {
1571 * API 3.1.2 bind() - UDP Style Syntax
1572 * If a bind() or sctp_bindx() is not called prior to a
1573 * sendmsg() call that initiates a new association, the
1574 * system picks an ephemeral port and will choose an address
1575 * set equivalent to binding with a wildcard address.
1577 if (!ep
->base
.bind_addr
.port
) {
1578 if (sctp_autobind(sk
)) {
1584 * If an unprivileged user inherits a one-to-many
1585 * style socket with open associations on a privileged
1586 * port, it MAY be permitted to accept new associations,
1587 * but it SHOULD NOT be permitted to open new
1590 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1591 !capable(CAP_NET_BIND_SERVICE
)) {
1597 scope
= sctp_scope(&to
);
1598 new_asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1605 /* If the SCTP_INIT ancillary data is specified, set all
1606 * the association init values accordingly.
1609 if (sinit
->sinit_num_ostreams
) {
1610 asoc
->c
.sinit_num_ostreams
=
1611 sinit
->sinit_num_ostreams
;
1613 if (sinit
->sinit_max_instreams
) {
1614 asoc
->c
.sinit_max_instreams
=
1615 sinit
->sinit_max_instreams
;
1617 if (sinit
->sinit_max_attempts
) {
1618 asoc
->max_init_attempts
1619 = sinit
->sinit_max_attempts
;
1621 if (sinit
->sinit_max_init_timeo
) {
1622 asoc
->max_init_timeo
=
1623 msecs_to_jiffies(sinit
->sinit_max_init_timeo
);
1627 /* Prime the peer's transport structures. */
1628 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
, SCTP_UNKNOWN
);
1633 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, GFP_KERNEL
);
1640 /* ASSERT: we have a valid association at this point. */
1641 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1644 /* If the user didn't specify SNDRCVINFO, make up one with
1647 default_sinfo
.sinfo_stream
= asoc
->default_stream
;
1648 default_sinfo
.sinfo_flags
= asoc
->default_flags
;
1649 default_sinfo
.sinfo_ppid
= asoc
->default_ppid
;
1650 default_sinfo
.sinfo_context
= asoc
->default_context
;
1651 default_sinfo
.sinfo_timetolive
= asoc
->default_timetolive
;
1652 default_sinfo
.sinfo_assoc_id
= sctp_assoc2id(asoc
);
1653 sinfo
= &default_sinfo
;
1656 /* API 7.1.7, the sndbuf size per association bounds the
1657 * maximum size of data that can be sent in a single send call.
1659 if (msg_len
> sk
->sk_sndbuf
) {
1664 /* If fragmentation is disabled and the message length exceeds the
1665 * association fragmentation point, return EMSGSIZE. The I-D
1666 * does not specify what this error is, but this looks like
1669 if (sctp_sk(sk
)->disable_fragments
&& (msg_len
> asoc
->frag_point
)) {
1675 /* Check for invalid stream. */
1676 if (sinfo
->sinfo_stream
>= asoc
->c
.sinit_num_ostreams
) {
1682 timeo
= sock_sndtimeo(sk
, msg
->msg_flags
& MSG_DONTWAIT
);
1683 if (!sctp_wspace(asoc
)) {
1684 err
= sctp_wait_for_sndbuf(asoc
, &timeo
, msg_len
);
1689 /* If an address is passed with the sendto/sendmsg call, it is used
1690 * to override the primary destination address in the TCP model, or
1691 * when SCTP_ADDR_OVER flag is set in the UDP model.
1693 if ((sctp_style(sk
, TCP
) && msg_name
) ||
1694 (sinfo_flags
& SCTP_ADDR_OVER
)) {
1695 chunk_tp
= sctp_assoc_lookup_paddr(asoc
, &tmp
);
1703 /* Auto-connect, if we aren't connected already. */
1704 if (sctp_state(asoc
, CLOSED
)) {
1705 err
= sctp_primitive_ASSOCIATE(asoc
, NULL
);
1708 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1711 /* Break the message into multiple chunks of maximum size. */
1712 datamsg
= sctp_datamsg_from_user(asoc
, sinfo
, msg
, msg_len
);
1718 /* Now send the (possibly) fragmented message. */
1719 list_for_each(pos
, &datamsg
->chunks
) {
1720 chunk
= list_entry(pos
, struct sctp_chunk
, frag_list
);
1721 sctp_datamsg_track(chunk
);
1723 /* Do accounting for the write space. */
1724 sctp_set_owner_w(chunk
);
1726 chunk
->transport
= chunk_tp
;
1728 /* Send it to the lower layers. Note: all chunks
1729 * must either fail or succeed. The lower layer
1730 * works that way today. Keep it that way or this
1733 err
= sctp_primitive_SEND(asoc
, chunk
);
1734 /* Did the lower layer accept the chunk? */
1736 sctp_chunk_free(chunk
);
1737 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1740 sctp_datamsg_free(datamsg
);
1746 /* If we are already past ASSOCIATE, the lower
1747 * layers are responsible for association cleanup.
1753 sctp_association_free(asoc
);
1755 sctp_release_sock(sk
);
1758 return sctp_error(sk
, msg_flags
, err
);
1765 err
= sock_error(sk
);
1775 /* This is an extended version of skb_pull() that removes the data from the
1776 * start of a skb even when data is spread across the list of skb's in the
1777 * frag_list. len specifies the total amount of data that needs to be removed.
1778 * when 'len' bytes could be removed from the skb, it returns 0.
1779 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1780 * could not be removed.
1782 static int sctp_skb_pull(struct sk_buff
*skb
, int len
)
1784 struct sk_buff
*list
;
1785 int skb_len
= skb_headlen(skb
);
1788 if (len
<= skb_len
) {
1789 __skb_pull(skb
, len
);
1793 __skb_pull(skb
, skb_len
);
1795 for (list
= skb_shinfo(skb
)->frag_list
; list
; list
= list
->next
) {
1796 rlen
= sctp_skb_pull(list
, len
);
1797 skb
->len
-= (len
-rlen
);
1798 skb
->data_len
-= (len
-rlen
);
1809 /* API 3.1.3 recvmsg() - UDP Style Syntax
1811 * ssize_t recvmsg(int socket, struct msghdr *message,
1814 * socket - the socket descriptor of the endpoint.
1815 * message - pointer to the msghdr structure which contains a single
1816 * user message and possibly some ancillary data.
1818 * See Section 5 for complete description of the data
1821 * flags - flags sent or received with the user message, see Section
1822 * 5 for complete description of the flags.
1824 static struct sk_buff
*sctp_skb_recv_datagram(struct sock
*, int, int, int *);
1826 SCTP_STATIC
int sctp_recvmsg(struct kiocb
*iocb
, struct sock
*sk
,
1827 struct msghdr
*msg
, size_t len
, int noblock
,
1828 int flags
, int *addr_len
)
1830 struct sctp_ulpevent
*event
= NULL
;
1831 struct sctp_sock
*sp
= sctp_sk(sk
);
1832 struct sk_buff
*skb
;
1837 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1838 "0x%x, %s: %p)\n", "sk", sk
, "msghdr", msg
,
1839 "len", len
, "knoblauch", noblock
,
1840 "flags", flags
, "addr_len", addr_len
);
1844 if (sctp_style(sk
, TCP
) && !sctp_sstate(sk
, ESTABLISHED
)) {
1849 skb
= sctp_skb_recv_datagram(sk
, flags
, noblock
, &err
);
1853 /* Get the total length of the skb including any skb's in the
1862 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
1864 event
= sctp_skb2event(skb
);
1869 sock_recv_timestamp(msg
, sk
, skb
);
1870 if (sctp_ulpevent_is_notification(event
)) {
1871 msg
->msg_flags
|= MSG_NOTIFICATION
;
1872 sp
->pf
->event_msgname(event
, msg
->msg_name
, addr_len
);
1874 sp
->pf
->skb_msgname(skb
, msg
->msg_name
, addr_len
);
1877 /* Check if we allow SCTP_SNDRCVINFO. */
1878 if (sp
->subscribe
.sctp_data_io_event
)
1879 sctp_ulpevent_read_sndrcvinfo(event
, msg
);
1881 /* FIXME: we should be calling IP/IPv6 layers. */
1882 if (sk
->sk_protinfo
.af_inet
.cmsg_flags
)
1883 ip_cmsg_recv(msg
, skb
);
1888 /* If skb's length exceeds the user's buffer, update the skb and
1889 * push it back to the receive_queue so that the next call to
1890 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1892 if (skb_len
> copied
) {
1893 msg
->msg_flags
&= ~MSG_EOR
;
1894 if (flags
& MSG_PEEK
)
1896 sctp_skb_pull(skb
, copied
);
1897 skb_queue_head(&sk
->sk_receive_queue
, skb
);
1899 /* When only partial message is copied to the user, increase
1900 * rwnd by that amount. If all the data in the skb is read,
1901 * rwnd is updated when the event is freed.
1903 sctp_assoc_rwnd_increase(event
->asoc
, copied
);
1905 } else if ((event
->msg_flags
& MSG_NOTIFICATION
) ||
1906 (event
->msg_flags
& MSG_EOR
))
1907 msg
->msg_flags
|= MSG_EOR
;
1909 msg
->msg_flags
&= ~MSG_EOR
;
1912 if (flags
& MSG_PEEK
) {
1913 /* Release the skb reference acquired after peeking the skb in
1914 * sctp_skb_recv_datagram().
1918 /* Free the event which includes releasing the reference to
1919 * the owner of the skb, freeing the skb and updating the
1922 sctp_ulpevent_free(event
);
1925 sctp_release_sock(sk
);
1929 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
1931 * This option is a on/off flag. If enabled no SCTP message
1932 * fragmentation will be performed. Instead if a message being sent
1933 * exceeds the current PMTU size, the message will NOT be sent and
1934 * instead a error will be indicated to the user.
1936 static int sctp_setsockopt_disable_fragments(struct sock
*sk
,
1937 char __user
*optval
, int optlen
)
1941 if (optlen
< sizeof(int))
1944 if (get_user(val
, (int __user
*)optval
))
1947 sctp_sk(sk
)->disable_fragments
= (val
== 0) ? 0 : 1;
1952 static int sctp_setsockopt_events(struct sock
*sk
, char __user
*optval
,
1955 if (optlen
!= sizeof(struct sctp_event_subscribe
))
1957 if (copy_from_user(&sctp_sk(sk
)->subscribe
, optval
, optlen
))
1962 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
1964 * This socket option is applicable to the UDP-style socket only. When
1965 * set it will cause associations that are idle for more than the
1966 * specified number of seconds to automatically close. An association
1967 * being idle is defined an association that has NOT sent or received
1968 * user data. The special value of '0' indicates that no automatic
1969 * close of any associations should be performed. The option expects an
1970 * integer defining the number of seconds of idle time before an
1971 * association is closed.
1973 static int sctp_setsockopt_autoclose(struct sock
*sk
, char __user
*optval
,
1976 struct sctp_sock
*sp
= sctp_sk(sk
);
1978 /* Applicable to UDP-style socket only */
1979 if (sctp_style(sk
, TCP
))
1981 if (optlen
!= sizeof(int))
1983 if (copy_from_user(&sp
->autoclose
, optval
, optlen
))
1989 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
1991 * Applications can enable or disable heartbeats for any peer address of
1992 * an association, modify an address's heartbeat interval, force a
1993 * heartbeat to be sent immediately, and adjust the address's maximum
1994 * number of retransmissions sent before an address is considered
1995 * unreachable. The following structure is used to access and modify an
1996 * address's parameters:
1998 * struct sctp_paddrparams {
1999 * sctp_assoc_t spp_assoc_id;
2000 * struct sockaddr_storage spp_address;
2001 * uint32_t spp_hbinterval;
2002 * uint16_t spp_pathmaxrxt;
2003 * uint32_t spp_pathmtu;
2004 * uint32_t spp_sackdelay;
2005 * uint32_t spp_flags;
2008 * spp_assoc_id - (one-to-many style socket) This is filled in the
2009 * application, and identifies the association for
2011 * spp_address - This specifies which address is of interest.
2012 * spp_hbinterval - This contains the value of the heartbeat interval,
2013 * in milliseconds. If a value of zero
2014 * is present in this field then no changes are to
2015 * be made to this parameter.
2016 * spp_pathmaxrxt - This contains the maximum number of
2017 * retransmissions before this address shall be
2018 * considered unreachable. If a value of zero
2019 * is present in this field then no changes are to
2020 * be made to this parameter.
2021 * spp_pathmtu - When Path MTU discovery is disabled the value
2022 * specified here will be the "fixed" path mtu.
2023 * Note that if the spp_address field is empty
2024 * then all associations on this address will
2025 * have this fixed path mtu set upon them.
2027 * spp_sackdelay - When delayed sack is enabled, this value specifies
2028 * the number of milliseconds that sacks will be delayed
2029 * for. This value will apply to all addresses of an
2030 * association if the spp_address field is empty. Note
2031 * also, that if delayed sack is enabled and this
2032 * value is set to 0, no change is made to the last
2033 * recorded delayed sack timer value.
2035 * spp_flags - These flags are used to control various features
2036 * on an association. The flag field may contain
2037 * zero or more of the following options.
2039 * SPP_HB_ENABLE - Enable heartbeats on the
2040 * specified address. Note that if the address
2041 * field is empty all addresses for the association
2042 * have heartbeats enabled upon them.
2044 * SPP_HB_DISABLE - Disable heartbeats on the
2045 * speicifed address. Note that if the address
2046 * field is empty all addresses for the association
2047 * will have their heartbeats disabled. Note also
2048 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2049 * mutually exclusive, only one of these two should
2050 * be specified. Enabling both fields will have
2051 * undetermined results.
2053 * SPP_HB_DEMAND - Request a user initiated heartbeat
2054 * to be made immediately.
2056 * SPP_PMTUD_ENABLE - This field will enable PMTU
2057 * discovery upon the specified address. Note that
2058 * if the address feild is empty then all addresses
2059 * on the association are effected.
2061 * SPP_PMTUD_DISABLE - This field will disable PMTU
2062 * discovery upon the specified address. Note that
2063 * if the address feild is empty then all addresses
2064 * on the association are effected. Not also that
2065 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2066 * exclusive. Enabling both will have undetermined
2069 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2070 * on delayed sack. The time specified in spp_sackdelay
2071 * is used to specify the sack delay for this address. Note
2072 * that if spp_address is empty then all addresses will
2073 * enable delayed sack and take on the sack delay
2074 * value specified in spp_sackdelay.
2075 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2076 * off delayed sack. If the spp_address field is blank then
2077 * delayed sack is disabled for the entire association. Note
2078 * also that this field is mutually exclusive to
2079 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2082 static int sctp_apply_peer_addr_params(struct sctp_paddrparams
*params
,
2083 struct sctp_transport
*trans
,
2084 struct sctp_association
*asoc
,
2085 struct sctp_sock
*sp
,
2088 int sackdelay_change
)
2092 if (params
->spp_flags
& SPP_HB_DEMAND
&& trans
) {
2093 error
= sctp_primitive_REQUESTHEARTBEAT (trans
->asoc
, trans
);
2098 if (params
->spp_hbinterval
) {
2100 trans
->hbinterval
= msecs_to_jiffies(params
->spp_hbinterval
);
2102 asoc
->hbinterval
= msecs_to_jiffies(params
->spp_hbinterval
);
2104 sp
->hbinterval
= params
->spp_hbinterval
;
2110 trans
->param_flags
=
2111 (trans
->param_flags
& ~SPP_HB
) | hb_change
;
2114 (asoc
->param_flags
& ~SPP_HB
) | hb_change
;
2117 (sp
->param_flags
& ~SPP_HB
) | hb_change
;
2121 if (params
->spp_pathmtu
) {
2123 trans
->pathmtu
= params
->spp_pathmtu
;
2124 sctp_assoc_sync_pmtu(asoc
);
2126 asoc
->pathmtu
= params
->spp_pathmtu
;
2127 sctp_frag_point(sp
, params
->spp_pathmtu
);
2129 sp
->pathmtu
= params
->spp_pathmtu
;
2135 int update
= (trans
->param_flags
& SPP_PMTUD_DISABLE
) &&
2136 (params
->spp_flags
& SPP_PMTUD_ENABLE
);
2137 trans
->param_flags
=
2138 (trans
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2140 sctp_transport_pmtu(trans
);
2141 sctp_assoc_sync_pmtu(asoc
);
2145 (asoc
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2148 (sp
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2152 if (params
->spp_sackdelay
) {
2155 msecs_to_jiffies(params
->spp_sackdelay
);
2158 msecs_to_jiffies(params
->spp_sackdelay
);
2160 sp
->sackdelay
= params
->spp_sackdelay
;
2164 if (sackdelay_change
) {
2166 trans
->param_flags
=
2167 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2171 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2175 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2180 if (params
->spp_pathmaxrxt
) {
2182 trans
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2184 asoc
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2186 sp
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2193 static int sctp_setsockopt_peer_addr_params(struct sock
*sk
,
2194 char __user
*optval
, int optlen
)
2196 struct sctp_paddrparams params
;
2197 struct sctp_transport
*trans
= NULL
;
2198 struct sctp_association
*asoc
= NULL
;
2199 struct sctp_sock
*sp
= sctp_sk(sk
);
2201 int hb_change
, pmtud_change
, sackdelay_change
;
2203 if (optlen
!= sizeof(struct sctp_paddrparams
))
2206 if (copy_from_user(¶ms
, optval
, optlen
))
2209 /* Validate flags and value parameters. */
2210 hb_change
= params
.spp_flags
& SPP_HB
;
2211 pmtud_change
= params
.spp_flags
& SPP_PMTUD
;
2212 sackdelay_change
= params
.spp_flags
& SPP_SACKDELAY
;
2214 if (hb_change
== SPP_HB
||
2215 pmtud_change
== SPP_PMTUD
||
2216 sackdelay_change
== SPP_SACKDELAY
||
2217 params
.spp_sackdelay
> 500 ||
2219 && params
.spp_pathmtu
< SCTP_DEFAULT_MINSEGMENT
))
2222 /* If an address other than INADDR_ANY is specified, and
2223 * no transport is found, then the request is invalid.
2225 if (!sctp_is_any(( union sctp_addr
*)¶ms
.spp_address
)) {
2226 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
2227 params
.spp_assoc_id
);
2232 /* Get association, if assoc_id != 0 and the socket is a one
2233 * to many style socket, and an association was not found, then
2234 * the id was invalid.
2236 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
2237 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
))
2240 /* Heartbeat demand can only be sent on a transport or
2241 * association, but not a socket.
2243 if (params
.spp_flags
& SPP_HB_DEMAND
&& !trans
&& !asoc
)
2246 /* Process parameters. */
2247 error
= sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2248 hb_change
, pmtud_change
,
2254 /* If changes are for association, also apply parameters to each
2257 if (!trans
&& asoc
) {
2258 struct list_head
*pos
;
2260 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
2261 trans
= list_entry(pos
, struct sctp_transport
,
2263 sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2264 hb_change
, pmtud_change
,
2272 /* 7.1.24. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
2274 * This options will get or set the delayed ack timer. The time is set
2275 * in milliseconds. If the assoc_id is 0, then this sets or gets the
2276 * endpoints default delayed ack timer value. If the assoc_id field is
2277 * non-zero, then the set or get effects the specified association.
2279 * struct sctp_assoc_value {
2280 * sctp_assoc_t assoc_id;
2281 * uint32_t assoc_value;
2284 * assoc_id - This parameter, indicates which association the
2285 * user is preforming an action upon. Note that if
2286 * this field's value is zero then the endpoints
2287 * default value is changed (effecting future
2288 * associations only).
2290 * assoc_value - This parameter contains the number of milliseconds
2291 * that the user is requesting the delayed ACK timer
2292 * be set to. Note that this value is defined in
2293 * the standard to be between 200 and 500 milliseconds.
2295 * Note: a value of zero will leave the value alone,
2296 * but disable SACK delay. A non-zero value will also
2297 * enable SACK delay.
2300 static int sctp_setsockopt_delayed_ack_time(struct sock
*sk
,
2301 char __user
*optval
, int optlen
)
2303 struct sctp_assoc_value params
;
2304 struct sctp_transport
*trans
= NULL
;
2305 struct sctp_association
*asoc
= NULL
;
2306 struct sctp_sock
*sp
= sctp_sk(sk
);
2308 if (optlen
!= sizeof(struct sctp_assoc_value
))
2311 if (copy_from_user(¶ms
, optval
, optlen
))
2314 /* Validate value parameter. */
2315 if (params
.assoc_value
> 500)
2318 /* Get association, if assoc_id != 0 and the socket is a one
2319 * to many style socket, and an association was not found, then
2320 * the id was invalid.
2322 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
2323 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
2326 if (params
.assoc_value
) {
2329 msecs_to_jiffies(params
.assoc_value
);
2331 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2332 SPP_SACKDELAY_ENABLE
;
2334 sp
->sackdelay
= params
.assoc_value
;
2336 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2337 SPP_SACKDELAY_ENABLE
;
2342 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2343 SPP_SACKDELAY_DISABLE
;
2346 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2347 SPP_SACKDELAY_DISABLE
;
2351 /* If change is for association, also apply to each transport. */
2353 struct list_head
*pos
;
2355 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
2356 trans
= list_entry(pos
, struct sctp_transport
,
2358 if (params
.assoc_value
) {
2360 msecs_to_jiffies(params
.assoc_value
);
2361 trans
->param_flags
=
2362 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2363 SPP_SACKDELAY_ENABLE
;
2365 trans
->param_flags
=
2366 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2367 SPP_SACKDELAY_DISABLE
;
2375 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2377 * Applications can specify protocol parameters for the default association
2378 * initialization. The option name argument to setsockopt() and getsockopt()
2381 * Setting initialization parameters is effective only on an unconnected
2382 * socket (for UDP-style sockets only future associations are effected
2383 * by the change). With TCP-style sockets, this option is inherited by
2384 * sockets derived from a listener socket.
2386 static int sctp_setsockopt_initmsg(struct sock
*sk
, char __user
*optval
, int optlen
)
2388 struct sctp_initmsg sinit
;
2389 struct sctp_sock
*sp
= sctp_sk(sk
);
2391 if (optlen
!= sizeof(struct sctp_initmsg
))
2393 if (copy_from_user(&sinit
, optval
, optlen
))
2396 if (sinit
.sinit_num_ostreams
)
2397 sp
->initmsg
.sinit_num_ostreams
= sinit
.sinit_num_ostreams
;
2398 if (sinit
.sinit_max_instreams
)
2399 sp
->initmsg
.sinit_max_instreams
= sinit
.sinit_max_instreams
;
2400 if (sinit
.sinit_max_attempts
)
2401 sp
->initmsg
.sinit_max_attempts
= sinit
.sinit_max_attempts
;
2402 if (sinit
.sinit_max_init_timeo
)
2403 sp
->initmsg
.sinit_max_init_timeo
= sinit
.sinit_max_init_timeo
;
2409 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2411 * Applications that wish to use the sendto() system call may wish to
2412 * specify a default set of parameters that would normally be supplied
2413 * through the inclusion of ancillary data. This socket option allows
2414 * such an application to set the default sctp_sndrcvinfo structure.
2415 * The application that wishes to use this socket option simply passes
2416 * in to this call the sctp_sndrcvinfo structure defined in Section
2417 * 5.2.2) The input parameters accepted by this call include
2418 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2419 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2420 * to this call if the caller is using the UDP model.
2422 static int sctp_setsockopt_default_send_param(struct sock
*sk
,
2423 char __user
*optval
, int optlen
)
2425 struct sctp_sndrcvinfo info
;
2426 struct sctp_association
*asoc
;
2427 struct sctp_sock
*sp
= sctp_sk(sk
);
2429 if (optlen
!= sizeof(struct sctp_sndrcvinfo
))
2431 if (copy_from_user(&info
, optval
, optlen
))
2434 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
2435 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
2439 asoc
->default_stream
= info
.sinfo_stream
;
2440 asoc
->default_flags
= info
.sinfo_flags
;
2441 asoc
->default_ppid
= info
.sinfo_ppid
;
2442 asoc
->default_context
= info
.sinfo_context
;
2443 asoc
->default_timetolive
= info
.sinfo_timetolive
;
2445 sp
->default_stream
= info
.sinfo_stream
;
2446 sp
->default_flags
= info
.sinfo_flags
;
2447 sp
->default_ppid
= info
.sinfo_ppid
;
2448 sp
->default_context
= info
.sinfo_context
;
2449 sp
->default_timetolive
= info
.sinfo_timetolive
;
2455 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2457 * Requests that the local SCTP stack use the enclosed peer address as
2458 * the association primary. The enclosed address must be one of the
2459 * association peer's addresses.
2461 static int sctp_setsockopt_primary_addr(struct sock
*sk
, char __user
*optval
,
2464 struct sctp_prim prim
;
2465 struct sctp_transport
*trans
;
2467 if (optlen
!= sizeof(struct sctp_prim
))
2470 if (copy_from_user(&prim
, optval
, sizeof(struct sctp_prim
)))
2473 trans
= sctp_addr_id2transport(sk
, &prim
.ssp_addr
, prim
.ssp_assoc_id
);
2477 sctp_assoc_set_primary(trans
->asoc
, trans
);
2483 * 7.1.5 SCTP_NODELAY
2485 * Turn on/off any Nagle-like algorithm. This means that packets are
2486 * generally sent as soon as possible and no unnecessary delays are
2487 * introduced, at the cost of more packets in the network. Expects an
2488 * integer boolean flag.
2490 static int sctp_setsockopt_nodelay(struct sock
*sk
, char __user
*optval
,
2495 if (optlen
< sizeof(int))
2497 if (get_user(val
, (int __user
*)optval
))
2500 sctp_sk(sk
)->nodelay
= (val
== 0) ? 0 : 1;
2506 * 7.1.1 SCTP_RTOINFO
2508 * The protocol parameters used to initialize and bound retransmission
2509 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2510 * and modify these parameters.
2511 * All parameters are time values, in milliseconds. A value of 0, when
2512 * modifying the parameters, indicates that the current value should not
2516 static int sctp_setsockopt_rtoinfo(struct sock
*sk
, char __user
*optval
, int optlen
) {
2517 struct sctp_rtoinfo rtoinfo
;
2518 struct sctp_association
*asoc
;
2520 if (optlen
!= sizeof (struct sctp_rtoinfo
))
2523 if (copy_from_user(&rtoinfo
, optval
, optlen
))
2526 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
2528 /* Set the values to the specific association */
2529 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
2533 if (rtoinfo
.srto_initial
!= 0)
2535 msecs_to_jiffies(rtoinfo
.srto_initial
);
2536 if (rtoinfo
.srto_max
!= 0)
2537 asoc
->rto_max
= msecs_to_jiffies(rtoinfo
.srto_max
);
2538 if (rtoinfo
.srto_min
!= 0)
2539 asoc
->rto_min
= msecs_to_jiffies(rtoinfo
.srto_min
);
2541 /* If there is no association or the association-id = 0
2542 * set the values to the endpoint.
2544 struct sctp_sock
*sp
= sctp_sk(sk
);
2546 if (rtoinfo
.srto_initial
!= 0)
2547 sp
->rtoinfo
.srto_initial
= rtoinfo
.srto_initial
;
2548 if (rtoinfo
.srto_max
!= 0)
2549 sp
->rtoinfo
.srto_max
= rtoinfo
.srto_max
;
2550 if (rtoinfo
.srto_min
!= 0)
2551 sp
->rtoinfo
.srto_min
= rtoinfo
.srto_min
;
2559 * 7.1.2 SCTP_ASSOCINFO
2561 * This option is used to tune the the maximum retransmission attempts
2562 * of the association.
2563 * Returns an error if the new association retransmission value is
2564 * greater than the sum of the retransmission value of the peer.
2565 * See [SCTP] for more information.
2568 static int sctp_setsockopt_associnfo(struct sock
*sk
, char __user
*optval
, int optlen
)
2571 struct sctp_assocparams assocparams
;
2572 struct sctp_association
*asoc
;
2574 if (optlen
!= sizeof(struct sctp_assocparams
))
2576 if (copy_from_user(&assocparams
, optval
, optlen
))
2579 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
2581 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
2584 /* Set the values to the specific association */
2586 if (assocparams
.sasoc_asocmaxrxt
!= 0) {
2589 struct list_head
*pos
;
2590 struct sctp_transport
*peer_addr
;
2592 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
2593 peer_addr
= list_entry(pos
,
2594 struct sctp_transport
,
2596 path_sum
+= peer_addr
->pathmaxrxt
;
2600 /* Only validate asocmaxrxt if we have more then
2601 * one path/transport. We do this because path
2602 * retransmissions are only counted when we have more
2606 assocparams
.sasoc_asocmaxrxt
> path_sum
)
2609 asoc
->max_retrans
= assocparams
.sasoc_asocmaxrxt
;
2612 if (assocparams
.sasoc_cookie_life
!= 0) {
2613 asoc
->cookie_life
.tv_sec
=
2614 assocparams
.sasoc_cookie_life
/ 1000;
2615 asoc
->cookie_life
.tv_usec
=
2616 (assocparams
.sasoc_cookie_life
% 1000)
2620 /* Set the values to the endpoint */
2621 struct sctp_sock
*sp
= sctp_sk(sk
);
2623 if (assocparams
.sasoc_asocmaxrxt
!= 0)
2624 sp
->assocparams
.sasoc_asocmaxrxt
=
2625 assocparams
.sasoc_asocmaxrxt
;
2626 if (assocparams
.sasoc_cookie_life
!= 0)
2627 sp
->assocparams
.sasoc_cookie_life
=
2628 assocparams
.sasoc_cookie_life
;
2634 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2636 * This socket option is a boolean flag which turns on or off mapped V4
2637 * addresses. If this option is turned on and the socket is type
2638 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2639 * If this option is turned off, then no mapping will be done of V4
2640 * addresses and a user will receive both PF_INET6 and PF_INET type
2641 * addresses on the socket.
2643 static int sctp_setsockopt_mappedv4(struct sock
*sk
, char __user
*optval
, int optlen
)
2646 struct sctp_sock
*sp
= sctp_sk(sk
);
2648 if (optlen
< sizeof(int))
2650 if (get_user(val
, (int __user
*)optval
))
2661 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
2663 * This socket option specifies the maximum size to put in any outgoing
2664 * SCTP chunk. If a message is larger than this size it will be
2665 * fragmented by SCTP into the specified size. Note that the underlying
2666 * SCTP implementation may fragment into smaller sized chunks when the
2667 * PMTU of the underlying association is smaller than the value set by
2670 static int sctp_setsockopt_maxseg(struct sock
*sk
, char __user
*optval
, int optlen
)
2672 struct sctp_association
*asoc
;
2673 struct list_head
*pos
;
2674 struct sctp_sock
*sp
= sctp_sk(sk
);
2677 if (optlen
< sizeof(int))
2679 if (get_user(val
, (int __user
*)optval
))
2681 if ((val
!= 0) && ((val
< 8) || (val
> SCTP_MAX_CHUNK_LEN
)))
2683 sp
->user_frag
= val
;
2685 /* Update the frag_point of the existing associations. */
2686 list_for_each(pos
, &(sp
->ep
->asocs
)) {
2687 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
2688 asoc
->frag_point
= sctp_frag_point(sp
, asoc
->pathmtu
);
2696 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2698 * Requests that the peer mark the enclosed address as the association
2699 * primary. The enclosed address must be one of the association's
2700 * locally bound addresses. The following structure is used to make a
2701 * set primary request:
2703 static int sctp_setsockopt_peer_primary_addr(struct sock
*sk
, char __user
*optval
,
2706 struct sctp_sock
*sp
;
2707 struct sctp_endpoint
*ep
;
2708 struct sctp_association
*asoc
= NULL
;
2709 struct sctp_setpeerprim prim
;
2710 struct sctp_chunk
*chunk
;
2716 if (!sctp_addip_enable
)
2719 if (optlen
!= sizeof(struct sctp_setpeerprim
))
2722 if (copy_from_user(&prim
, optval
, optlen
))
2725 asoc
= sctp_id2assoc(sk
, prim
.sspp_assoc_id
);
2729 if (!asoc
->peer
.asconf_capable
)
2732 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_SET_PRIMARY
)
2735 if (!sctp_state(asoc
, ESTABLISHED
))
2738 if (!sctp_assoc_lookup_laddr(asoc
, (union sctp_addr
*)&prim
.sspp_addr
))
2739 return -EADDRNOTAVAIL
;
2741 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2742 chunk
= sctp_make_asconf_set_prim(asoc
,
2743 (union sctp_addr
*)&prim
.sspp_addr
);
2747 err
= sctp_send_asconf(asoc
, chunk
);
2749 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2754 static int sctp_setsockopt_adaption_layer(struct sock
*sk
, char __user
*optval
,
2757 struct sctp_setadaption adaption
;
2759 if (optlen
!= sizeof(struct sctp_setadaption
))
2761 if (copy_from_user(&adaption
, optval
, optlen
))
2764 sctp_sk(sk
)->adaption_ind
= adaption
.ssb_adaption_ind
;
2769 /* API 6.2 setsockopt(), getsockopt()
2771 * Applications use setsockopt() and getsockopt() to set or retrieve
2772 * socket options. Socket options are used to change the default
2773 * behavior of sockets calls. They are described in Section 7.
2777 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
2778 * int __user *optlen);
2779 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
2782 * sd - the socket descript.
2783 * level - set to IPPROTO_SCTP for all SCTP options.
2784 * optname - the option name.
2785 * optval - the buffer to store the value of the option.
2786 * optlen - the size of the buffer.
2788 SCTP_STATIC
int sctp_setsockopt(struct sock
*sk
, int level
, int optname
,
2789 char __user
*optval
, int optlen
)
2793 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
2796 /* I can hardly begin to describe how wrong this is. This is
2797 * so broken as to be worse than useless. The API draft
2798 * REALLY is NOT helpful here... I am not convinced that the
2799 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
2800 * are at all well-founded.
2802 if (level
!= SOL_SCTP
) {
2803 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
2804 retval
= af
->setsockopt(sk
, level
, optname
, optval
, optlen
);
2811 case SCTP_SOCKOPT_BINDX_ADD
:
2812 /* 'optlen' is the size of the addresses buffer. */
2813 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
2814 optlen
, SCTP_BINDX_ADD_ADDR
);
2817 case SCTP_SOCKOPT_BINDX_REM
:
2818 /* 'optlen' is the size of the addresses buffer. */
2819 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
2820 optlen
, SCTP_BINDX_REM_ADDR
);
2823 case SCTP_SOCKOPT_CONNECTX
:
2824 /* 'optlen' is the size of the addresses buffer. */
2825 retval
= sctp_setsockopt_connectx(sk
, (struct sockaddr __user
*)optval
,
2829 case SCTP_DISABLE_FRAGMENTS
:
2830 retval
= sctp_setsockopt_disable_fragments(sk
, optval
, optlen
);
2834 retval
= sctp_setsockopt_events(sk
, optval
, optlen
);
2837 case SCTP_AUTOCLOSE
:
2838 retval
= sctp_setsockopt_autoclose(sk
, optval
, optlen
);
2841 case SCTP_PEER_ADDR_PARAMS
:
2842 retval
= sctp_setsockopt_peer_addr_params(sk
, optval
, optlen
);
2845 case SCTP_DELAYED_ACK_TIME
:
2846 retval
= sctp_setsockopt_delayed_ack_time(sk
, optval
, optlen
);
2850 retval
= sctp_setsockopt_initmsg(sk
, optval
, optlen
);
2852 case SCTP_DEFAULT_SEND_PARAM
:
2853 retval
= sctp_setsockopt_default_send_param(sk
, optval
,
2856 case SCTP_PRIMARY_ADDR
:
2857 retval
= sctp_setsockopt_primary_addr(sk
, optval
, optlen
);
2859 case SCTP_SET_PEER_PRIMARY_ADDR
:
2860 retval
= sctp_setsockopt_peer_primary_addr(sk
, optval
, optlen
);
2863 retval
= sctp_setsockopt_nodelay(sk
, optval
, optlen
);
2866 retval
= sctp_setsockopt_rtoinfo(sk
, optval
, optlen
);
2868 case SCTP_ASSOCINFO
:
2869 retval
= sctp_setsockopt_associnfo(sk
, optval
, optlen
);
2871 case SCTP_I_WANT_MAPPED_V4_ADDR
:
2872 retval
= sctp_setsockopt_mappedv4(sk
, optval
, optlen
);
2875 retval
= sctp_setsockopt_maxseg(sk
, optval
, optlen
);
2877 case SCTP_ADAPTION_LAYER
:
2878 retval
= sctp_setsockopt_adaption_layer(sk
, optval
, optlen
);
2882 retval
= -ENOPROTOOPT
;
2886 sctp_release_sock(sk
);
2892 /* API 3.1.6 connect() - UDP Style Syntax
2894 * An application may use the connect() call in the UDP model to initiate an
2895 * association without sending data.
2899 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
2901 * sd: the socket descriptor to have a new association added to.
2903 * nam: the address structure (either struct sockaddr_in or struct
2904 * sockaddr_in6 defined in RFC2553 [7]).
2906 * len: the size of the address.
2908 SCTP_STATIC
int sctp_connect(struct sock
*sk
, struct sockaddr
*addr
,
2916 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
2917 __FUNCTION__
, sk
, addr
, addr_len
);
2919 /* Validate addr_len before calling common connect/connectx routine. */
2920 af
= sctp_get_af_specific(addr
->sa_family
);
2921 if (!af
|| addr_len
< af
->sockaddr_len
) {
2924 /* Pass correct addr len to common routine (so it knows there
2925 * is only one address being passed.
2927 err
= __sctp_connect(sk
, addr
, af
->sockaddr_len
);
2930 sctp_release_sock(sk
);
2934 /* FIXME: Write comments. */
2935 SCTP_STATIC
int sctp_disconnect(struct sock
*sk
, int flags
)
2937 return -EOPNOTSUPP
; /* STUB */
2940 /* 4.1.4 accept() - TCP Style Syntax
2942 * Applications use accept() call to remove an established SCTP
2943 * association from the accept queue of the endpoint. A new socket
2944 * descriptor will be returned from accept() to represent the newly
2945 * formed association.
2947 SCTP_STATIC
struct sock
*sctp_accept(struct sock
*sk
, int flags
, int *err
)
2949 struct sctp_sock
*sp
;
2950 struct sctp_endpoint
*ep
;
2951 struct sock
*newsk
= NULL
;
2952 struct sctp_association
*asoc
;
2961 if (!sctp_style(sk
, TCP
)) {
2962 error
= -EOPNOTSUPP
;
2966 if (!sctp_sstate(sk
, LISTENING
)) {
2971 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
2973 error
= sctp_wait_for_accept(sk
, timeo
);
2977 /* We treat the list of associations on the endpoint as the accept
2978 * queue and pick the first association on the list.
2980 asoc
= list_entry(ep
->asocs
.next
, struct sctp_association
, asocs
);
2982 newsk
= sp
->pf
->create_accept_sk(sk
, asoc
);
2988 /* Populate the fields of the newsk from the oldsk and migrate the
2989 * asoc to the newsk.
2991 sctp_sock_migrate(sk
, newsk
, asoc
, SCTP_SOCKET_TCP
);
2994 sctp_release_sock(sk
);
2999 /* The SCTP ioctl handler. */
3000 SCTP_STATIC
int sctp_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
3002 return -ENOIOCTLCMD
;
3005 /* This is the function which gets called during socket creation to
3006 * initialized the SCTP-specific portion of the sock.
3007 * The sock structure should already be zero-filled memory.
3009 SCTP_STATIC
int sctp_init_sock(struct sock
*sk
)
3011 struct sctp_endpoint
*ep
;
3012 struct sctp_sock
*sp
;
3014 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk
);
3018 /* Initialize the SCTP per socket area. */
3019 switch (sk
->sk_type
) {
3020 case SOCK_SEQPACKET
:
3021 sp
->type
= SCTP_SOCKET_UDP
;
3024 sp
->type
= SCTP_SOCKET_TCP
;
3027 return -ESOCKTNOSUPPORT
;
3030 /* Initialize default send parameters. These parameters can be
3031 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3033 sp
->default_stream
= 0;
3034 sp
->default_ppid
= 0;
3035 sp
->default_flags
= 0;
3036 sp
->default_context
= 0;
3037 sp
->default_timetolive
= 0;
3039 /* Initialize default setup parameters. These parameters
3040 * can be modified with the SCTP_INITMSG socket option or
3041 * overridden by the SCTP_INIT CMSG.
3043 sp
->initmsg
.sinit_num_ostreams
= sctp_max_outstreams
;
3044 sp
->initmsg
.sinit_max_instreams
= sctp_max_instreams
;
3045 sp
->initmsg
.sinit_max_attempts
= sctp_max_retrans_init
;
3046 sp
->initmsg
.sinit_max_init_timeo
= sctp_rto_max
;
3048 /* Initialize default RTO related parameters. These parameters can
3049 * be modified for with the SCTP_RTOINFO socket option.
3051 sp
->rtoinfo
.srto_initial
= sctp_rto_initial
;
3052 sp
->rtoinfo
.srto_max
= sctp_rto_max
;
3053 sp
->rtoinfo
.srto_min
= sctp_rto_min
;
3055 /* Initialize default association related parameters. These parameters
3056 * can be modified with the SCTP_ASSOCINFO socket option.
3058 sp
->assocparams
.sasoc_asocmaxrxt
= sctp_max_retrans_association
;
3059 sp
->assocparams
.sasoc_number_peer_destinations
= 0;
3060 sp
->assocparams
.sasoc_peer_rwnd
= 0;
3061 sp
->assocparams
.sasoc_local_rwnd
= 0;
3062 sp
->assocparams
.sasoc_cookie_life
= sctp_valid_cookie_life
;
3064 /* Initialize default event subscriptions. By default, all the
3067 memset(&sp
->subscribe
, 0, sizeof(struct sctp_event_subscribe
));
3069 /* Default Peer Address Parameters. These defaults can
3070 * be modified via SCTP_PEER_ADDR_PARAMS
3072 sp
->hbinterval
= sctp_hb_interval
;
3073 sp
->pathmaxrxt
= sctp_max_retrans_path
;
3074 sp
->pathmtu
= 0; // allow default discovery
3075 sp
->sackdelay
= sctp_sack_timeout
;
3076 sp
->param_flags
= SPP_HB_ENABLE
|
3078 SPP_SACKDELAY_ENABLE
;
3080 /* If enabled no SCTP message fragmentation will be performed.
3081 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3083 sp
->disable_fragments
= 0;
3085 /* Enable Nagle algorithm by default. */
3088 /* Enable by default. */
3091 /* Auto-close idle associations after the configured
3092 * number of seconds. A value of 0 disables this
3093 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3094 * for UDP-style sockets only.
3098 /* User specified fragmentation limit. */
3101 sp
->adaption_ind
= 0;
3103 sp
->pf
= sctp_get_pf_specific(sk
->sk_family
);
3105 /* Control variables for partial data delivery. */
3107 skb_queue_head_init(&sp
->pd_lobby
);
3109 /* Create a per socket endpoint structure. Even if we
3110 * change the data structure relationships, this may still
3111 * be useful for storing pre-connect address information.
3113 ep
= sctp_endpoint_new(sk
, GFP_KERNEL
);
3120 SCTP_DBG_OBJCNT_INC(sock
);
3124 /* Cleanup any SCTP per socket resources. */
3125 SCTP_STATIC
int sctp_destroy_sock(struct sock
*sk
)
3127 struct sctp_endpoint
*ep
;
3129 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk
);
3131 /* Release our hold on the endpoint. */
3132 ep
= sctp_sk(sk
)->ep
;
3133 sctp_endpoint_free(ep
);
3138 /* API 4.1.7 shutdown() - TCP Style Syntax
3139 * int shutdown(int socket, int how);
3141 * sd - the socket descriptor of the association to be closed.
3142 * how - Specifies the type of shutdown. The values are
3145 * Disables further receive operations. No SCTP
3146 * protocol action is taken.
3148 * Disables further send operations, and initiates
3149 * the SCTP shutdown sequence.
3151 * Disables further send and receive operations
3152 * and initiates the SCTP shutdown sequence.
3154 SCTP_STATIC
void sctp_shutdown(struct sock
*sk
, int how
)
3156 struct sctp_endpoint
*ep
;
3157 struct sctp_association
*asoc
;
3159 if (!sctp_style(sk
, TCP
))
3162 if (how
& SEND_SHUTDOWN
) {
3163 ep
= sctp_sk(sk
)->ep
;
3164 if (!list_empty(&ep
->asocs
)) {
3165 asoc
= list_entry(ep
->asocs
.next
,
3166 struct sctp_association
, asocs
);
3167 sctp_primitive_SHUTDOWN(asoc
, NULL
);
3172 /* 7.2.1 Association Status (SCTP_STATUS)
3174 * Applications can retrieve current status information about an
3175 * association, including association state, peer receiver window size,
3176 * number of unacked data chunks, and number of data chunks pending
3177 * receipt. This information is read-only.
3179 static int sctp_getsockopt_sctp_status(struct sock
*sk
, int len
,
3180 char __user
*optval
,
3183 struct sctp_status status
;
3184 struct sctp_association
*asoc
= NULL
;
3185 struct sctp_transport
*transport
;
3186 sctp_assoc_t associd
;
3189 if (len
!= sizeof(status
)) {
3194 if (copy_from_user(&status
, optval
, sizeof(status
))) {
3199 associd
= status
.sstat_assoc_id
;
3200 asoc
= sctp_id2assoc(sk
, associd
);
3206 transport
= asoc
->peer
.primary_path
;
3208 status
.sstat_assoc_id
= sctp_assoc2id(asoc
);
3209 status
.sstat_state
= asoc
->state
;
3210 status
.sstat_rwnd
= asoc
->peer
.rwnd
;
3211 status
.sstat_unackdata
= asoc
->unack_data
;
3213 status
.sstat_penddata
= sctp_tsnmap_pending(&asoc
->peer
.tsn_map
);
3214 status
.sstat_instrms
= asoc
->c
.sinit_max_instreams
;
3215 status
.sstat_outstrms
= asoc
->c
.sinit_num_ostreams
;
3216 status
.sstat_fragmentation_point
= asoc
->frag_point
;
3217 status
.sstat_primary
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
3218 flip_to_n((union sctp_addr
*)&status
.sstat_primary
.spinfo_address
,
3219 &transport
->ipaddr_h
);
3220 /* Map ipv4 address into v4-mapped-on-v6 address. */
3221 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
3222 (union sctp_addr
*)&status
.sstat_primary
.spinfo_address
);
3223 status
.sstat_primary
.spinfo_state
= transport
->state
;
3224 status
.sstat_primary
.spinfo_cwnd
= transport
->cwnd
;
3225 status
.sstat_primary
.spinfo_srtt
= transport
->srtt
;
3226 status
.sstat_primary
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
3227 status
.sstat_primary
.spinfo_mtu
= transport
->pathmtu
;
3229 if (status
.sstat_primary
.spinfo_state
== SCTP_UNKNOWN
)
3230 status
.sstat_primary
.spinfo_state
= SCTP_ACTIVE
;
3232 if (put_user(len
, optlen
)) {
3237 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3238 len
, status
.sstat_state
, status
.sstat_rwnd
,
3239 status
.sstat_assoc_id
);
3241 if (copy_to_user(optval
, &status
, len
)) {
3251 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3253 * Applications can retrieve information about a specific peer address
3254 * of an association, including its reachability state, congestion
3255 * window, and retransmission timer values. This information is
3258 static int sctp_getsockopt_peer_addr_info(struct sock
*sk
, int len
,
3259 char __user
*optval
,
3262 struct sctp_paddrinfo pinfo
;
3263 struct sctp_transport
*transport
;
3266 if (len
!= sizeof(pinfo
)) {
3271 if (copy_from_user(&pinfo
, optval
, sizeof(pinfo
))) {
3276 transport
= sctp_addr_id2transport(sk
, &pinfo
.spinfo_address
,
3277 pinfo
.spinfo_assoc_id
);
3281 pinfo
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
3282 pinfo
.spinfo_state
= transport
->state
;
3283 pinfo
.spinfo_cwnd
= transport
->cwnd
;
3284 pinfo
.spinfo_srtt
= transport
->srtt
;
3285 pinfo
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
3286 pinfo
.spinfo_mtu
= transport
->pathmtu
;
3288 if (pinfo
.spinfo_state
== SCTP_UNKNOWN
)
3289 pinfo
.spinfo_state
= SCTP_ACTIVE
;
3291 if (put_user(len
, optlen
)) {
3296 if (copy_to_user(optval
, &pinfo
, len
)) {
3305 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3307 * This option is a on/off flag. If enabled no SCTP message
3308 * fragmentation will be performed. Instead if a message being sent
3309 * exceeds the current PMTU size, the message will NOT be sent and
3310 * instead a error will be indicated to the user.
3312 static int sctp_getsockopt_disable_fragments(struct sock
*sk
, int len
,
3313 char __user
*optval
, int __user
*optlen
)
3317 if (len
< sizeof(int))
3321 val
= (sctp_sk(sk
)->disable_fragments
== 1);
3322 if (put_user(len
, optlen
))
3324 if (copy_to_user(optval
, &val
, len
))
3329 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3331 * This socket option is used to specify various notifications and
3332 * ancillary data the user wishes to receive.
3334 static int sctp_getsockopt_events(struct sock
*sk
, int len
, char __user
*optval
,
3337 if (len
!= sizeof(struct sctp_event_subscribe
))
3339 if (copy_to_user(optval
, &sctp_sk(sk
)->subscribe
, len
))
3344 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3346 * This socket option is applicable to the UDP-style socket only. When
3347 * set it will cause associations that are idle for more than the
3348 * specified number of seconds to automatically close. An association
3349 * being idle is defined an association that has NOT sent or received
3350 * user data. The special value of '0' indicates that no automatic
3351 * close of any associations should be performed. The option expects an
3352 * integer defining the number of seconds of idle time before an
3353 * association is closed.
3355 static int sctp_getsockopt_autoclose(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
3357 /* Applicable to UDP-style socket only */
3358 if (sctp_style(sk
, TCP
))
3360 if (len
!= sizeof(int))
3362 if (copy_to_user(optval
, &sctp_sk(sk
)->autoclose
, len
))
3367 /* Helper routine to branch off an association to a new socket. */
3368 SCTP_STATIC
int sctp_do_peeloff(struct sctp_association
*asoc
,
3369 struct socket
**sockp
)
3371 struct sock
*sk
= asoc
->base
.sk
;
3372 struct socket
*sock
;
3373 struct inet_sock
*inetsk
;
3376 /* An association cannot be branched off from an already peeled-off
3377 * socket, nor is this supported for tcp style sockets.
3379 if (!sctp_style(sk
, UDP
))
3382 /* Create a new socket. */
3383 err
= sock_create(sk
->sk_family
, SOCK_SEQPACKET
, IPPROTO_SCTP
, &sock
);
3387 /* Populate the fields of the newsk from the oldsk and migrate the
3388 * asoc to the newsk.
3390 sctp_sock_migrate(sk
, sock
->sk
, asoc
, SCTP_SOCKET_UDP_HIGH_BANDWIDTH
);
3392 /* Make peeled-off sockets more like 1-1 accepted sockets.
3393 * Set the daddr and initialize id to something more random
3395 inetsk
= inet_sk(sock
->sk
);
3396 inetsk
->daddr
= asoc
->peer
.primary_addr
.v4
.sin_addr
.s_addr
;
3397 inetsk
->id
= asoc
->next_tsn
^ jiffies
;
3404 static int sctp_getsockopt_peeloff(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
3406 sctp_peeloff_arg_t peeloff
;
3407 struct socket
*newsock
;
3409 struct sctp_association
*asoc
;
3411 if (len
!= sizeof(sctp_peeloff_arg_t
))
3413 if (copy_from_user(&peeloff
, optval
, len
))
3416 asoc
= sctp_id2assoc(sk
, peeloff
.associd
);
3422 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __FUNCTION__
, sk
, asoc
);
3424 retval
= sctp_do_peeloff(asoc
, &newsock
);
3428 /* Map the socket to an unused fd that can be returned to the user. */
3429 retval
= sock_map_fd(newsock
);
3431 sock_release(newsock
);
3435 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3436 __FUNCTION__
, sk
, asoc
, newsock
->sk
, retval
);
3438 /* Return the fd mapped to the new socket. */
3439 peeloff
.sd
= retval
;
3440 if (copy_to_user(optval
, &peeloff
, len
))
3447 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
3449 * Applications can enable or disable heartbeats for any peer address of
3450 * an association, modify an address's heartbeat interval, force a
3451 * heartbeat to be sent immediately, and adjust the address's maximum
3452 * number of retransmissions sent before an address is considered
3453 * unreachable. The following structure is used to access and modify an
3454 * address's parameters:
3456 * struct sctp_paddrparams {
3457 * sctp_assoc_t spp_assoc_id;
3458 * struct sockaddr_storage spp_address;
3459 * uint32_t spp_hbinterval;
3460 * uint16_t spp_pathmaxrxt;
3461 * uint32_t spp_pathmtu;
3462 * uint32_t spp_sackdelay;
3463 * uint32_t spp_flags;
3466 * spp_assoc_id - (one-to-many style socket) This is filled in the
3467 * application, and identifies the association for
3469 * spp_address - This specifies which address is of interest.
3470 * spp_hbinterval - This contains the value of the heartbeat interval,
3471 * in milliseconds. If a value of zero
3472 * is present in this field then no changes are to
3473 * be made to this parameter.
3474 * spp_pathmaxrxt - This contains the maximum number of
3475 * retransmissions before this address shall be
3476 * considered unreachable. If a value of zero
3477 * is present in this field then no changes are to
3478 * be made to this parameter.
3479 * spp_pathmtu - When Path MTU discovery is disabled the value
3480 * specified here will be the "fixed" path mtu.
3481 * Note that if the spp_address field is empty
3482 * then all associations on this address will
3483 * have this fixed path mtu set upon them.
3485 * spp_sackdelay - When delayed sack is enabled, this value specifies
3486 * the number of milliseconds that sacks will be delayed
3487 * for. This value will apply to all addresses of an
3488 * association if the spp_address field is empty. Note
3489 * also, that if delayed sack is enabled and this
3490 * value is set to 0, no change is made to the last
3491 * recorded delayed sack timer value.
3493 * spp_flags - These flags are used to control various features
3494 * on an association. The flag field may contain
3495 * zero or more of the following options.
3497 * SPP_HB_ENABLE - Enable heartbeats on the
3498 * specified address. Note that if the address
3499 * field is empty all addresses for the association
3500 * have heartbeats enabled upon them.
3502 * SPP_HB_DISABLE - Disable heartbeats on the
3503 * speicifed address. Note that if the address
3504 * field is empty all addresses for the association
3505 * will have their heartbeats disabled. Note also
3506 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
3507 * mutually exclusive, only one of these two should
3508 * be specified. Enabling both fields will have
3509 * undetermined results.
3511 * SPP_HB_DEMAND - Request a user initiated heartbeat
3512 * to be made immediately.
3514 * SPP_PMTUD_ENABLE - This field will enable PMTU
3515 * discovery upon the specified address. Note that
3516 * if the address feild is empty then all addresses
3517 * on the association are effected.
3519 * SPP_PMTUD_DISABLE - This field will disable PMTU
3520 * discovery upon the specified address. Note that
3521 * if the address feild is empty then all addresses
3522 * on the association are effected. Not also that
3523 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
3524 * exclusive. Enabling both will have undetermined
3527 * SPP_SACKDELAY_ENABLE - Setting this flag turns
3528 * on delayed sack. The time specified in spp_sackdelay
3529 * is used to specify the sack delay for this address. Note
3530 * that if spp_address is empty then all addresses will
3531 * enable delayed sack and take on the sack delay
3532 * value specified in spp_sackdelay.
3533 * SPP_SACKDELAY_DISABLE - Setting this flag turns
3534 * off delayed sack. If the spp_address field is blank then
3535 * delayed sack is disabled for the entire association. Note
3536 * also that this field is mutually exclusive to
3537 * SPP_SACKDELAY_ENABLE, setting both will have undefined
3540 static int sctp_getsockopt_peer_addr_params(struct sock
*sk
, int len
,
3541 char __user
*optval
, int __user
*optlen
)
3543 struct sctp_paddrparams params
;
3544 struct sctp_transport
*trans
= NULL
;
3545 struct sctp_association
*asoc
= NULL
;
3546 struct sctp_sock
*sp
= sctp_sk(sk
);
3548 if (len
!= sizeof(struct sctp_paddrparams
))
3551 if (copy_from_user(¶ms
, optval
, len
))
3554 /* If an address other than INADDR_ANY is specified, and
3555 * no transport is found, then the request is invalid.
3557 if (!sctp_is_any(( union sctp_addr
*)¶ms
.spp_address
)) {
3558 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
3559 params
.spp_assoc_id
);
3561 SCTP_DEBUG_PRINTK("Failed no transport\n");
3566 /* Get association, if assoc_id != 0 and the socket is a one
3567 * to many style socket, and an association was not found, then
3568 * the id was invalid.
3570 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
3571 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
)) {
3572 SCTP_DEBUG_PRINTK("Failed no association\n");
3577 /* Fetch transport values. */
3578 params
.spp_hbinterval
= jiffies_to_msecs(trans
->hbinterval
);
3579 params
.spp_pathmtu
= trans
->pathmtu
;
3580 params
.spp_pathmaxrxt
= trans
->pathmaxrxt
;
3581 params
.spp_sackdelay
= jiffies_to_msecs(trans
->sackdelay
);
3583 /*draft-11 doesn't say what to return in spp_flags*/
3584 params
.spp_flags
= trans
->param_flags
;
3586 /* Fetch association values. */
3587 params
.spp_hbinterval
= jiffies_to_msecs(asoc
->hbinterval
);
3588 params
.spp_pathmtu
= asoc
->pathmtu
;
3589 params
.spp_pathmaxrxt
= asoc
->pathmaxrxt
;
3590 params
.spp_sackdelay
= jiffies_to_msecs(asoc
->sackdelay
);
3592 /*draft-11 doesn't say what to return in spp_flags*/
3593 params
.spp_flags
= asoc
->param_flags
;
3595 /* Fetch socket values. */
3596 params
.spp_hbinterval
= sp
->hbinterval
;
3597 params
.spp_pathmtu
= sp
->pathmtu
;
3598 params
.spp_sackdelay
= sp
->sackdelay
;
3599 params
.spp_pathmaxrxt
= sp
->pathmaxrxt
;
3601 /*draft-11 doesn't say what to return in spp_flags*/
3602 params
.spp_flags
= sp
->param_flags
;
3605 if (copy_to_user(optval
, ¶ms
, len
))
3608 if (put_user(len
, optlen
))
3614 /* 7.1.24. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
3616 * This options will get or set the delayed ack timer. The time is set
3617 * in milliseconds. If the assoc_id is 0, then this sets or gets the
3618 * endpoints default delayed ack timer value. If the assoc_id field is
3619 * non-zero, then the set or get effects the specified association.
3621 * struct sctp_assoc_value {
3622 * sctp_assoc_t assoc_id;
3623 * uint32_t assoc_value;
3626 * assoc_id - This parameter, indicates which association the
3627 * user is preforming an action upon. Note that if
3628 * this field's value is zero then the endpoints
3629 * default value is changed (effecting future
3630 * associations only).
3632 * assoc_value - This parameter contains the number of milliseconds
3633 * that the user is requesting the delayed ACK timer
3634 * be set to. Note that this value is defined in
3635 * the standard to be between 200 and 500 milliseconds.
3637 * Note: a value of zero will leave the value alone,
3638 * but disable SACK delay. A non-zero value will also
3639 * enable SACK delay.
3641 static int sctp_getsockopt_delayed_ack_time(struct sock
*sk
, int len
,
3642 char __user
*optval
,
3645 struct sctp_assoc_value params
;
3646 struct sctp_association
*asoc
= NULL
;
3647 struct sctp_sock
*sp
= sctp_sk(sk
);
3649 if (len
!= sizeof(struct sctp_assoc_value
))
3652 if (copy_from_user(¶ms
, optval
, len
))
3655 /* Get association, if assoc_id != 0 and the socket is a one
3656 * to many style socket, and an association was not found, then
3657 * the id was invalid.
3659 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3660 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
3664 /* Fetch association values. */
3665 if (asoc
->param_flags
& SPP_SACKDELAY_ENABLE
)
3666 params
.assoc_value
= jiffies_to_msecs(
3669 params
.assoc_value
= 0;
3671 /* Fetch socket values. */
3672 if (sp
->param_flags
& SPP_SACKDELAY_ENABLE
)
3673 params
.assoc_value
= sp
->sackdelay
;
3675 params
.assoc_value
= 0;
3678 if (copy_to_user(optval
, ¶ms
, len
))
3681 if (put_user(len
, optlen
))
3687 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
3689 * Applications can specify protocol parameters for the default association
3690 * initialization. The option name argument to setsockopt() and getsockopt()
3693 * Setting initialization parameters is effective only on an unconnected
3694 * socket (for UDP-style sockets only future associations are effected
3695 * by the change). With TCP-style sockets, this option is inherited by
3696 * sockets derived from a listener socket.
3698 static int sctp_getsockopt_initmsg(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
3700 if (len
!= sizeof(struct sctp_initmsg
))
3702 if (copy_to_user(optval
, &sctp_sk(sk
)->initmsg
, len
))
3707 static int sctp_getsockopt_peer_addrs_num_old(struct sock
*sk
, int len
,
3708 char __user
*optval
,
3712 struct sctp_association
*asoc
;
3713 struct list_head
*pos
;
3716 if (len
!= sizeof(sctp_assoc_t
))
3719 if (copy_from_user(&id
, optval
, sizeof(sctp_assoc_t
)))
3722 /* For UDP-style sockets, id specifies the association to query. */
3723 asoc
= sctp_id2assoc(sk
, id
);
3727 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
3735 * Old API for getting list of peer addresses. Does not work for 32-bit
3736 * programs running on a 64-bit kernel
3738 static int sctp_getsockopt_peer_addrs_old(struct sock
*sk
, int len
,
3739 char __user
*optval
,
3742 struct sctp_association
*asoc
;
3743 struct list_head
*pos
;
3745 struct sctp_getaddrs_old getaddrs
;
3746 struct sctp_transport
*from
;
3748 union sctp_addr temp
;
3749 struct sctp_sock
*sp
= sctp_sk(sk
);
3752 if (len
!= sizeof(struct sctp_getaddrs_old
))
3755 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs_old
)))
3758 if (getaddrs
.addr_num
<= 0) return -EINVAL
;
3760 /* For UDP-style sockets, id specifies the association to query. */
3761 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
3765 to
= (void __user
*)getaddrs
.addrs
;
3766 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
3767 from
= list_entry(pos
, struct sctp_transport
, transports
);
3768 memcpy(&temp
, &from
->ipaddr_h
, sizeof(temp
));
3769 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
3770 addrlen
= sctp_get_af_specific(sk
->sk_family
)->sockaddr_len
;
3771 temp
.v4
.sin_port
= htons(temp
.v4
.sin_port
);
3772 if (copy_to_user(to
, &temp
, addrlen
))
3776 if (cnt
>= getaddrs
.addr_num
) break;
3778 getaddrs
.addr_num
= cnt
;
3779 if (copy_to_user(optval
, &getaddrs
, sizeof(struct sctp_getaddrs_old
)))
3785 static int sctp_getsockopt_peer_addrs(struct sock
*sk
, int len
,
3786 char __user
*optval
, int __user
*optlen
)
3788 struct sctp_association
*asoc
;
3789 struct list_head
*pos
;
3791 struct sctp_getaddrs getaddrs
;
3792 struct sctp_transport
*from
;
3794 union sctp_addr temp
;
3795 struct sctp_sock
*sp
= sctp_sk(sk
);
3800 if (len
< sizeof(struct sctp_getaddrs
))
3803 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
3806 /* For UDP-style sockets, id specifies the association to query. */
3807 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
3811 to
= optval
+ offsetof(struct sctp_getaddrs
,addrs
);
3812 space_left
= len
- sizeof(struct sctp_getaddrs
) -
3813 offsetof(struct sctp_getaddrs
,addrs
);
3815 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
3816 from
= list_entry(pos
, struct sctp_transport
, transports
);
3817 memcpy(&temp
, &from
->ipaddr_h
, sizeof(temp
));
3818 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
3819 addrlen
= sctp_get_af_specific(sk
->sk_family
)->sockaddr_len
;
3820 if(space_left
< addrlen
)
3822 temp
.v4
.sin_port
= htons(temp
.v4
.sin_port
);
3823 if (copy_to_user(to
, &temp
, addrlen
))
3827 space_left
-= addrlen
;
3830 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
))
3832 bytes_copied
= ((char __user
*)to
) - optval
;
3833 if (put_user(bytes_copied
, optlen
))
3839 static int sctp_getsockopt_local_addrs_num_old(struct sock
*sk
, int len
,
3840 char __user
*optval
,
3844 struct sctp_bind_addr
*bp
;
3845 struct sctp_association
*asoc
;
3846 struct list_head
*pos
;
3847 struct sctp_sockaddr_entry
*addr
;
3848 rwlock_t
*addr_lock
;
3849 unsigned long flags
;
3852 if (len
!= sizeof(sctp_assoc_t
))
3855 if (copy_from_user(&id
, optval
, sizeof(sctp_assoc_t
)))
3859 * For UDP-style sockets, id specifies the association to query.
3860 * If the id field is set to the value '0' then the locally bound
3861 * addresses are returned without regard to any particular
3865 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
3866 addr_lock
= &sctp_sk(sk
)->ep
->base
.addr_lock
;
3868 asoc
= sctp_id2assoc(sk
, id
);
3871 bp
= &asoc
->base
.bind_addr
;
3872 addr_lock
= &asoc
->base
.addr_lock
;
3875 sctp_read_lock(addr_lock
);
3877 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
3878 * addresses from the global local address list.
3880 if (sctp_list_single_entry(&bp
->address_list
)) {
3881 addr
= list_entry(bp
->address_list
.next
,
3882 struct sctp_sockaddr_entry
, list
);
3883 if (sctp_is_any(&addr
->a_h
)) {
3884 sctp_spin_lock_irqsave(&sctp_local_addr_lock
, flags
);
3885 list_for_each(pos
, &sctp_local_addr_list
) {
3886 addr
= list_entry(pos
,
3887 struct sctp_sockaddr_entry
,
3889 if ((PF_INET
== sk
->sk_family
) &&
3890 (AF_INET6
== addr
->a_h
.sa
.sa_family
))
3894 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock
,
3902 list_for_each(pos
, &bp
->address_list
) {
3907 sctp_read_unlock(addr_lock
);
3911 /* Helper function that copies local addresses to user and returns the number
3912 * of addresses copied.
3914 static int sctp_copy_laddrs_to_user_old(struct sock
*sk
, __u16 port
, int max_addrs
,
3917 struct list_head
*pos
;
3918 struct sctp_sockaddr_entry
*addr
;
3919 unsigned long flags
;
3920 union sctp_addr temp
;
3924 sctp_spin_lock_irqsave(&sctp_local_addr_lock
, flags
);
3925 list_for_each(pos
, &sctp_local_addr_list
) {
3926 addr
= list_entry(pos
, struct sctp_sockaddr_entry
, list
);
3927 if ((PF_INET
== sk
->sk_family
) &&
3928 (AF_INET6
== addr
->a_h
.sa
.sa_family
))
3930 memcpy(&temp
, &addr
->a_h
, sizeof(temp
));
3931 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
3933 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
3934 temp
.v4
.sin_port
= htons(port
);
3935 if (copy_to_user(to
, &temp
, addrlen
)) {
3936 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock
,
3942 if (cnt
>= max_addrs
) break;
3944 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock
, flags
);
3949 static int sctp_copy_laddrs_to_user(struct sock
*sk
, __u16 port
,
3950 void __user
**to
, size_t space_left
)
3952 struct list_head
*pos
;
3953 struct sctp_sockaddr_entry
*addr
;
3954 unsigned long flags
;
3955 union sctp_addr temp
;
3959 sctp_spin_lock_irqsave(&sctp_local_addr_lock
, flags
);
3960 list_for_each(pos
, &sctp_local_addr_list
) {
3961 addr
= list_entry(pos
, struct sctp_sockaddr_entry
, list
);
3962 if ((PF_INET
== sk
->sk_family
) &&
3963 (AF_INET6
== addr
->a_h
.sa
.sa_family
))
3965 memcpy(&temp
, &addr
->a_h
, sizeof(temp
));
3966 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
3968 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
3969 if(space_left
<addrlen
)
3971 temp
.v4
.sin_port
= htons(port
);
3972 if (copy_to_user(*to
, &temp
, addrlen
)) {
3973 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock
,
3979 space_left
-= addrlen
;
3981 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock
, flags
);
3986 /* Old API for getting list of local addresses. Does not work for 32-bit
3987 * programs running on a 64-bit kernel
3989 static int sctp_getsockopt_local_addrs_old(struct sock
*sk
, int len
,
3990 char __user
*optval
, int __user
*optlen
)
3992 struct sctp_bind_addr
*bp
;
3993 struct sctp_association
*asoc
;
3994 struct list_head
*pos
;
3996 struct sctp_getaddrs_old getaddrs
;
3997 struct sctp_sockaddr_entry
*addr
;
3999 union sctp_addr temp
;
4000 struct sctp_sock
*sp
= sctp_sk(sk
);
4002 rwlock_t
*addr_lock
;
4005 if (len
!= sizeof(struct sctp_getaddrs_old
))
4008 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs_old
)))
4011 if (getaddrs
.addr_num
<= 0) return -EINVAL
;
4013 * For UDP-style sockets, id specifies the association to query.
4014 * If the id field is set to the value '0' then the locally bound
4015 * addresses are returned without regard to any particular
4018 if (0 == getaddrs
.assoc_id
) {
4019 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
4020 addr_lock
= &sctp_sk(sk
)->ep
->base
.addr_lock
;
4022 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
4025 bp
= &asoc
->base
.bind_addr
;
4026 addr_lock
= &asoc
->base
.addr_lock
;
4029 to
= getaddrs
.addrs
;
4031 sctp_read_lock(addr_lock
);
4033 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4034 * addresses from the global local address list.
4036 if (sctp_list_single_entry(&bp
->address_list
)) {
4037 addr
= list_entry(bp
->address_list
.next
,
4038 struct sctp_sockaddr_entry
, list
);
4039 if (sctp_is_any(&addr
->a_h
)) {
4040 cnt
= sctp_copy_laddrs_to_user_old(sk
, bp
->port
,
4051 list_for_each(pos
, &bp
->address_list
) {
4052 addr
= list_entry(pos
, struct sctp_sockaddr_entry
, list
);
4053 memcpy(&temp
, &addr
->a_h
, sizeof(temp
));
4054 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
4055 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4056 temp
.v4
.sin_port
= htons(temp
.v4
.sin_port
);
4057 if (copy_to_user(to
, &temp
, addrlen
)) {
4063 if (cnt
>= getaddrs
.addr_num
) break;
4067 getaddrs
.addr_num
= cnt
;
4068 if (copy_to_user(optval
, &getaddrs
, sizeof(struct sctp_getaddrs_old
)))
4072 sctp_read_unlock(addr_lock
);
4076 static int sctp_getsockopt_local_addrs(struct sock
*sk
, int len
,
4077 char __user
*optval
, int __user
*optlen
)
4079 struct sctp_bind_addr
*bp
;
4080 struct sctp_association
*asoc
;
4081 struct list_head
*pos
;
4083 struct sctp_getaddrs getaddrs
;
4084 struct sctp_sockaddr_entry
*addr
;
4086 union sctp_addr temp
;
4087 struct sctp_sock
*sp
= sctp_sk(sk
);
4089 rwlock_t
*addr_lock
;
4094 if (len
<= sizeof(struct sctp_getaddrs
))
4097 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
4101 * For UDP-style sockets, id specifies the association to query.
4102 * If the id field is set to the value '0' then the locally bound
4103 * addresses are returned without regard to any particular
4106 if (0 == getaddrs
.assoc_id
) {
4107 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
4108 addr_lock
= &sctp_sk(sk
)->ep
->base
.addr_lock
;
4110 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
4113 bp
= &asoc
->base
.bind_addr
;
4114 addr_lock
= &asoc
->base
.addr_lock
;
4117 to
= optval
+ offsetof(struct sctp_getaddrs
,addrs
);
4118 space_left
= len
- sizeof(struct sctp_getaddrs
) -
4119 offsetof(struct sctp_getaddrs
,addrs
);
4121 sctp_read_lock(addr_lock
);
4123 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4124 * addresses from the global local address list.
4126 if (sctp_list_single_entry(&bp
->address_list
)) {
4127 addr
= list_entry(bp
->address_list
.next
,
4128 struct sctp_sockaddr_entry
, list
);
4129 if (sctp_is_any(&addr
->a_h
)) {
4130 cnt
= sctp_copy_laddrs_to_user(sk
, bp
->port
,
4140 list_for_each(pos
, &bp
->address_list
) {
4141 addr
= list_entry(pos
, struct sctp_sockaddr_entry
, list
);
4142 memcpy(&temp
, &addr
->a_h
, sizeof(temp
));
4143 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
4144 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4145 if(space_left
< addrlen
)
4146 return -ENOMEM
; /*fixme: right error?*/
4147 temp
.v4
.sin_port
= htons(temp
.v4
.sin_port
);
4148 if (copy_to_user(to
, &temp
, addrlen
)) {
4154 space_left
-= addrlen
;
4158 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
))
4160 bytes_copied
= ((char __user
*)to
) - optval
;
4161 if (put_user(bytes_copied
, optlen
))
4165 sctp_read_unlock(addr_lock
);
4169 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4171 * Requests that the local SCTP stack use the enclosed peer address as
4172 * the association primary. The enclosed address must be one of the
4173 * association peer's addresses.
4175 static int sctp_getsockopt_primary_addr(struct sock
*sk
, int len
,
4176 char __user
*optval
, int __user
*optlen
)
4178 struct sctp_prim prim
;
4179 struct sctp_association
*asoc
;
4180 struct sctp_sock
*sp
= sctp_sk(sk
);
4182 if (len
!= sizeof(struct sctp_prim
))
4185 if (copy_from_user(&prim
, optval
, sizeof(struct sctp_prim
)))
4188 asoc
= sctp_id2assoc(sk
, prim
.ssp_assoc_id
);
4192 if (!asoc
->peer
.primary_path
)
4195 flip_to_n((union sctp_addr
*)&prim
.ssp_addr
,
4196 &asoc
->peer
.primary_path
->ipaddr_h
);
4198 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
,
4199 (union sctp_addr
*)&prim
.ssp_addr
);
4201 if (copy_to_user(optval
, &prim
, sizeof(struct sctp_prim
)))
4208 * 7.1.11 Set Adaption Layer Indicator (SCTP_ADAPTION_LAYER)
4210 * Requests that the local endpoint set the specified Adaption Layer
4211 * Indication parameter for all future INIT and INIT-ACK exchanges.
4213 static int sctp_getsockopt_adaption_layer(struct sock
*sk
, int len
,
4214 char __user
*optval
, int __user
*optlen
)
4216 struct sctp_setadaption adaption
;
4218 if (len
!= sizeof(struct sctp_setadaption
))
4221 adaption
.ssb_adaption_ind
= sctp_sk(sk
)->adaption_ind
;
4222 if (copy_to_user(optval
, &adaption
, len
))
4230 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4232 * Applications that wish to use the sendto() system call may wish to
4233 * specify a default set of parameters that would normally be supplied
4234 * through the inclusion of ancillary data. This socket option allows
4235 * such an application to set the default sctp_sndrcvinfo structure.
4238 * The application that wishes to use this socket option simply passes
4239 * in to this call the sctp_sndrcvinfo structure defined in Section
4240 * 5.2.2) The input parameters accepted by this call include
4241 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4242 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4243 * to this call if the caller is using the UDP model.
4245 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4247 static int sctp_getsockopt_default_send_param(struct sock
*sk
,
4248 int len
, char __user
*optval
,
4251 struct sctp_sndrcvinfo info
;
4252 struct sctp_association
*asoc
;
4253 struct sctp_sock
*sp
= sctp_sk(sk
);
4255 if (len
!= sizeof(struct sctp_sndrcvinfo
))
4257 if (copy_from_user(&info
, optval
, sizeof(struct sctp_sndrcvinfo
)))
4260 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
4261 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
4265 info
.sinfo_stream
= asoc
->default_stream
;
4266 info
.sinfo_flags
= asoc
->default_flags
;
4267 info
.sinfo_ppid
= asoc
->default_ppid
;
4268 info
.sinfo_context
= asoc
->default_context
;
4269 info
.sinfo_timetolive
= asoc
->default_timetolive
;
4271 info
.sinfo_stream
= sp
->default_stream
;
4272 info
.sinfo_flags
= sp
->default_flags
;
4273 info
.sinfo_ppid
= sp
->default_ppid
;
4274 info
.sinfo_context
= sp
->default_context
;
4275 info
.sinfo_timetolive
= sp
->default_timetolive
;
4278 if (copy_to_user(optval
, &info
, sizeof(struct sctp_sndrcvinfo
)))
4286 * 7.1.5 SCTP_NODELAY
4288 * Turn on/off any Nagle-like algorithm. This means that packets are
4289 * generally sent as soon as possible and no unnecessary delays are
4290 * introduced, at the cost of more packets in the network. Expects an
4291 * integer boolean flag.
4294 static int sctp_getsockopt_nodelay(struct sock
*sk
, int len
,
4295 char __user
*optval
, int __user
*optlen
)
4299 if (len
< sizeof(int))
4303 val
= (sctp_sk(sk
)->nodelay
== 1);
4304 if (put_user(len
, optlen
))
4306 if (copy_to_user(optval
, &val
, len
))
4313 * 7.1.1 SCTP_RTOINFO
4315 * The protocol parameters used to initialize and bound retransmission
4316 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4317 * and modify these parameters.
4318 * All parameters are time values, in milliseconds. A value of 0, when
4319 * modifying the parameters, indicates that the current value should not
4323 static int sctp_getsockopt_rtoinfo(struct sock
*sk
, int len
,
4324 char __user
*optval
,
4325 int __user
*optlen
) {
4326 struct sctp_rtoinfo rtoinfo
;
4327 struct sctp_association
*asoc
;
4329 if (len
!= sizeof (struct sctp_rtoinfo
))
4332 if (copy_from_user(&rtoinfo
, optval
, sizeof (struct sctp_rtoinfo
)))
4335 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
4337 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
4340 /* Values corresponding to the specific association. */
4342 rtoinfo
.srto_initial
= jiffies_to_msecs(asoc
->rto_initial
);
4343 rtoinfo
.srto_max
= jiffies_to_msecs(asoc
->rto_max
);
4344 rtoinfo
.srto_min
= jiffies_to_msecs(asoc
->rto_min
);
4346 /* Values corresponding to the endpoint. */
4347 struct sctp_sock
*sp
= sctp_sk(sk
);
4349 rtoinfo
.srto_initial
= sp
->rtoinfo
.srto_initial
;
4350 rtoinfo
.srto_max
= sp
->rtoinfo
.srto_max
;
4351 rtoinfo
.srto_min
= sp
->rtoinfo
.srto_min
;
4354 if (put_user(len
, optlen
))
4357 if (copy_to_user(optval
, &rtoinfo
, len
))
4365 * 7.1.2 SCTP_ASSOCINFO
4367 * This option is used to tune the the maximum retransmission attempts
4368 * of the association.
4369 * Returns an error if the new association retransmission value is
4370 * greater than the sum of the retransmission value of the peer.
4371 * See [SCTP] for more information.
4374 static int sctp_getsockopt_associnfo(struct sock
*sk
, int len
,
4375 char __user
*optval
,
4379 struct sctp_assocparams assocparams
;
4380 struct sctp_association
*asoc
;
4381 struct list_head
*pos
;
4384 if (len
!= sizeof (struct sctp_assocparams
))
4387 if (copy_from_user(&assocparams
, optval
,
4388 sizeof (struct sctp_assocparams
)))
4391 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
4393 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
4396 /* Values correspoinding to the specific association */
4398 assocparams
.sasoc_asocmaxrxt
= asoc
->max_retrans
;
4399 assocparams
.sasoc_peer_rwnd
= asoc
->peer
.rwnd
;
4400 assocparams
.sasoc_local_rwnd
= asoc
->a_rwnd
;
4401 assocparams
.sasoc_cookie_life
= (asoc
->cookie_life
.tv_sec
4403 (asoc
->cookie_life
.tv_usec
4406 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
4410 assocparams
.sasoc_number_peer_destinations
= cnt
;
4412 /* Values corresponding to the endpoint */
4413 struct sctp_sock
*sp
= sctp_sk(sk
);
4415 assocparams
.sasoc_asocmaxrxt
= sp
->assocparams
.sasoc_asocmaxrxt
;
4416 assocparams
.sasoc_peer_rwnd
= sp
->assocparams
.sasoc_peer_rwnd
;
4417 assocparams
.sasoc_local_rwnd
= sp
->assocparams
.sasoc_local_rwnd
;
4418 assocparams
.sasoc_cookie_life
=
4419 sp
->assocparams
.sasoc_cookie_life
;
4420 assocparams
.sasoc_number_peer_destinations
=
4422 sasoc_number_peer_destinations
;
4425 if (put_user(len
, optlen
))
4428 if (copy_to_user(optval
, &assocparams
, len
))
4435 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4437 * This socket option is a boolean flag which turns on or off mapped V4
4438 * addresses. If this option is turned on and the socket is type
4439 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4440 * If this option is turned off, then no mapping will be done of V4
4441 * addresses and a user will receive both PF_INET6 and PF_INET type
4442 * addresses on the socket.
4444 static int sctp_getsockopt_mappedv4(struct sock
*sk
, int len
,
4445 char __user
*optval
, int __user
*optlen
)
4448 struct sctp_sock
*sp
= sctp_sk(sk
);
4450 if (len
< sizeof(int))
4455 if (put_user(len
, optlen
))
4457 if (copy_to_user(optval
, &val
, len
))
4464 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
4466 * This socket option specifies the maximum size to put in any outgoing
4467 * SCTP chunk. If a message is larger than this size it will be
4468 * fragmented by SCTP into the specified size. Note that the underlying
4469 * SCTP implementation may fragment into smaller sized chunks when the
4470 * PMTU of the underlying association is smaller than the value set by
4473 static int sctp_getsockopt_maxseg(struct sock
*sk
, int len
,
4474 char __user
*optval
, int __user
*optlen
)
4478 if (len
< sizeof(int))
4483 val
= sctp_sk(sk
)->user_frag
;
4484 if (put_user(len
, optlen
))
4486 if (copy_to_user(optval
, &val
, len
))
4492 SCTP_STATIC
int sctp_getsockopt(struct sock
*sk
, int level
, int optname
,
4493 char __user
*optval
, int __user
*optlen
)
4498 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
4501 /* I can hardly begin to describe how wrong this is. This is
4502 * so broken as to be worse than useless. The API draft
4503 * REALLY is NOT helpful here... I am not convinced that the
4504 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
4505 * are at all well-founded.
4507 if (level
!= SOL_SCTP
) {
4508 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
4510 retval
= af
->getsockopt(sk
, level
, optname
, optval
, optlen
);
4514 if (get_user(len
, optlen
))
4521 retval
= sctp_getsockopt_sctp_status(sk
, len
, optval
, optlen
);
4523 case SCTP_DISABLE_FRAGMENTS
:
4524 retval
= sctp_getsockopt_disable_fragments(sk
, len
, optval
,
4528 retval
= sctp_getsockopt_events(sk
, len
, optval
, optlen
);
4530 case SCTP_AUTOCLOSE
:
4531 retval
= sctp_getsockopt_autoclose(sk
, len
, optval
, optlen
);
4533 case SCTP_SOCKOPT_PEELOFF
:
4534 retval
= sctp_getsockopt_peeloff(sk
, len
, optval
, optlen
);
4536 case SCTP_PEER_ADDR_PARAMS
:
4537 retval
= sctp_getsockopt_peer_addr_params(sk
, len
, optval
,
4540 case SCTP_DELAYED_ACK_TIME
:
4541 retval
= sctp_getsockopt_delayed_ack_time(sk
, len
, optval
,
4545 retval
= sctp_getsockopt_initmsg(sk
, len
, optval
, optlen
);
4547 case SCTP_GET_PEER_ADDRS_NUM_OLD
:
4548 retval
= sctp_getsockopt_peer_addrs_num_old(sk
, len
, optval
,
4551 case SCTP_GET_LOCAL_ADDRS_NUM_OLD
:
4552 retval
= sctp_getsockopt_local_addrs_num_old(sk
, len
, optval
,
4555 case SCTP_GET_PEER_ADDRS_OLD
:
4556 retval
= sctp_getsockopt_peer_addrs_old(sk
, len
, optval
,
4559 case SCTP_GET_LOCAL_ADDRS_OLD
:
4560 retval
= sctp_getsockopt_local_addrs_old(sk
, len
, optval
,
4563 case SCTP_GET_PEER_ADDRS
:
4564 retval
= sctp_getsockopt_peer_addrs(sk
, len
, optval
,
4567 case SCTP_GET_LOCAL_ADDRS
:
4568 retval
= sctp_getsockopt_local_addrs(sk
, len
, optval
,
4571 case SCTP_DEFAULT_SEND_PARAM
:
4572 retval
= sctp_getsockopt_default_send_param(sk
, len
,
4575 case SCTP_PRIMARY_ADDR
:
4576 retval
= sctp_getsockopt_primary_addr(sk
, len
, optval
, optlen
);
4579 retval
= sctp_getsockopt_nodelay(sk
, len
, optval
, optlen
);
4582 retval
= sctp_getsockopt_rtoinfo(sk
, len
, optval
, optlen
);
4584 case SCTP_ASSOCINFO
:
4585 retval
= sctp_getsockopt_associnfo(sk
, len
, optval
, optlen
);
4587 case SCTP_I_WANT_MAPPED_V4_ADDR
:
4588 retval
= sctp_getsockopt_mappedv4(sk
, len
, optval
, optlen
);
4591 retval
= sctp_getsockopt_maxseg(sk
, len
, optval
, optlen
);
4593 case SCTP_GET_PEER_ADDR_INFO
:
4594 retval
= sctp_getsockopt_peer_addr_info(sk
, len
, optval
,
4597 case SCTP_ADAPTION_LAYER
:
4598 retval
= sctp_getsockopt_adaption_layer(sk
, len
, optval
,
4602 retval
= -ENOPROTOOPT
;
4606 sctp_release_sock(sk
);
4610 static void sctp_hash(struct sock
*sk
)
4615 static void sctp_unhash(struct sock
*sk
)
4620 /* Check if port is acceptable. Possibly find first available port.
4622 * The port hash table (contained in the 'global' SCTP protocol storage
4623 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
4624 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
4625 * list (the list number is the port number hashed out, so as you
4626 * would expect from a hash function, all the ports in a given list have
4627 * such a number that hashes out to the same list number; you were
4628 * expecting that, right?); so each list has a set of ports, with a
4629 * link to the socket (struct sock) that uses it, the port number and
4630 * a fastreuse flag (FIXME: NPI ipg).
4632 static struct sctp_bind_bucket
*sctp_bucket_create(
4633 struct sctp_bind_hashbucket
*head
, unsigned short snum
);
4635 static long sctp_get_port_local(struct sock
*sk
, union sctp_addr
*addr
)
4637 struct sctp_bind_hashbucket
*head
; /* hash list */
4638 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
4639 unsigned short snum
;
4642 snum
= ntohs(addr
->v4
.sin_port
);
4644 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum
);
4645 sctp_local_bh_disable();
4648 /* Search for an available port.
4650 * 'sctp_port_rover' was the last port assigned, so
4651 * we start to search from 'sctp_port_rover +
4652 * 1'. What we do is first check if port 'rover' is
4653 * already in the hash table; if not, we use that; if
4654 * it is, we try next.
4656 int low
= sysctl_local_port_range
[0];
4657 int high
= sysctl_local_port_range
[1];
4658 int remaining
= (high
- low
) + 1;
4662 sctp_spin_lock(&sctp_port_alloc_lock
);
4663 rover
= sctp_port_rover
;
4666 if ((rover
< low
) || (rover
> high
))
4668 index
= sctp_phashfn(rover
);
4669 head
= &sctp_port_hashtable
[index
];
4670 sctp_spin_lock(&head
->lock
);
4671 for (pp
= head
->chain
; pp
; pp
= pp
->next
)
4672 if (pp
->port
== rover
)
4676 sctp_spin_unlock(&head
->lock
);
4677 } while (--remaining
> 0);
4678 sctp_port_rover
= rover
;
4679 sctp_spin_unlock(&sctp_port_alloc_lock
);
4681 /* Exhausted local port range during search? */
4686 /* OK, here is the one we will use. HEAD (the port
4687 * hash table list entry) is non-NULL and we hold it's
4692 /* We are given an specific port number; we verify
4693 * that it is not being used. If it is used, we will
4694 * exahust the search in the hash list corresponding
4695 * to the port number (snum) - we detect that with the
4696 * port iterator, pp being NULL.
4698 head
= &sctp_port_hashtable
[sctp_phashfn(snum
)];
4699 sctp_spin_lock(&head
->lock
);
4700 for (pp
= head
->chain
; pp
; pp
= pp
->next
) {
4701 if (pp
->port
== snum
)
4708 if (!hlist_empty(&pp
->owner
)) {
4709 /* We had a port hash table hit - there is an
4710 * available port (pp != NULL) and it is being
4711 * used by other socket (pp->owner not empty); that other
4712 * socket is going to be sk2.
4714 int reuse
= sk
->sk_reuse
;
4716 struct hlist_node
*node
;
4718 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
4719 if (pp
->fastreuse
&& sk
->sk_reuse
)
4722 /* Run through the list of sockets bound to the port
4723 * (pp->port) [via the pointers bind_next and
4724 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
4725 * we get the endpoint they describe and run through
4726 * the endpoint's list of IP (v4 or v6) addresses,
4727 * comparing each of the addresses with the address of
4728 * the socket sk. If we find a match, then that means
4729 * that this port/socket (sk) combination are already
4732 sk_for_each_bound(sk2
, node
, &pp
->owner
) {
4733 struct sctp_endpoint
*ep2
;
4734 ep2
= sctp_sk(sk2
)->ep
;
4736 if (reuse
&& sk2
->sk_reuse
)
4739 if (sctp_bind_addr_match(&ep2
->base
.bind_addr
, addr
,
4745 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
4748 /* If there was a hash table miss, create a new port. */
4750 if (!pp
&& !(pp
= sctp_bucket_create(head
, snum
)))
4753 /* In either case (hit or miss), make sure fastreuse is 1 only
4754 * if sk->sk_reuse is too (that is, if the caller requested
4755 * SO_REUSEADDR on this socket -sk-).
4757 if (hlist_empty(&pp
->owner
))
4758 pp
->fastreuse
= sk
->sk_reuse
? 1 : 0;
4759 else if (pp
->fastreuse
&& !sk
->sk_reuse
)
4762 /* We are set, so fill up all the data in the hash table
4763 * entry, tie the socket list information with the rest of the
4764 * sockets FIXME: Blurry, NPI (ipg).
4767 inet_sk(sk
)->num
= snum
;
4768 if (!sctp_sk(sk
)->bind_hash
) {
4769 sk_add_bind_node(sk
, &pp
->owner
);
4770 sctp_sk(sk
)->bind_hash
= pp
;
4775 sctp_spin_unlock(&head
->lock
);
4778 sctp_local_bh_enable();
4782 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
4783 * port is requested.
4785 static int sctp_get_port(struct sock
*sk
, unsigned short snum
)
4788 union sctp_addr addr
;
4789 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
4791 /* Set up a dummy address struct from the sk. */
4792 af
->from_sk(&addr
, sk
);
4793 addr
.v4
.sin_port
= htons(snum
);
4795 /* Note: sk->sk_num gets filled in if ephemeral port request. */
4796 ret
= sctp_get_port_local(sk
, &addr
);
4798 return (ret
? 1 : 0);
4802 * 3.1.3 listen() - UDP Style Syntax
4804 * By default, new associations are not accepted for UDP style sockets.
4805 * An application uses listen() to mark a socket as being able to
4806 * accept new associations.
4808 SCTP_STATIC
int sctp_seqpacket_listen(struct sock
*sk
, int backlog
)
4810 struct sctp_sock
*sp
= sctp_sk(sk
);
4811 struct sctp_endpoint
*ep
= sp
->ep
;
4813 /* Only UDP style sockets that are not peeled off are allowed to
4816 if (!sctp_style(sk
, UDP
))
4819 /* If backlog is zero, disable listening. */
4821 if (sctp_sstate(sk
, CLOSED
))
4824 sctp_unhash_endpoint(ep
);
4825 sk
->sk_state
= SCTP_SS_CLOSED
;
4828 /* Return if we are already listening. */
4829 if (sctp_sstate(sk
, LISTENING
))
4833 * If a bind() or sctp_bindx() is not called prior to a listen()
4834 * call that allows new associations to be accepted, the system
4835 * picks an ephemeral port and will choose an address set equivalent
4836 * to binding with a wildcard address.
4838 * This is not currently spelled out in the SCTP sockets
4839 * extensions draft, but follows the practice as seen in TCP
4842 if (!ep
->base
.bind_addr
.port
) {
4843 if (sctp_autobind(sk
))
4846 sk
->sk_state
= SCTP_SS_LISTENING
;
4847 sctp_hash_endpoint(ep
);
4852 * 4.1.3 listen() - TCP Style Syntax
4854 * Applications uses listen() to ready the SCTP endpoint for accepting
4855 * inbound associations.
4857 SCTP_STATIC
int sctp_stream_listen(struct sock
*sk
, int backlog
)
4859 struct sctp_sock
*sp
= sctp_sk(sk
);
4860 struct sctp_endpoint
*ep
= sp
->ep
;
4862 /* If backlog is zero, disable listening. */
4864 if (sctp_sstate(sk
, CLOSED
))
4867 sctp_unhash_endpoint(ep
);
4868 sk
->sk_state
= SCTP_SS_CLOSED
;
4871 if (sctp_sstate(sk
, LISTENING
))
4875 * If a bind() or sctp_bindx() is not called prior to a listen()
4876 * call that allows new associations to be accepted, the system
4877 * picks an ephemeral port and will choose an address set equivalent
4878 * to binding with a wildcard address.
4880 * This is not currently spelled out in the SCTP sockets
4881 * extensions draft, but follows the practice as seen in TCP
4884 if (!ep
->base
.bind_addr
.port
) {
4885 if (sctp_autobind(sk
))
4888 sk
->sk_state
= SCTP_SS_LISTENING
;
4889 sk
->sk_max_ack_backlog
= backlog
;
4890 sctp_hash_endpoint(ep
);
4895 * Move a socket to LISTENING state.
4897 int sctp_inet_listen(struct socket
*sock
, int backlog
)
4899 struct sock
*sk
= sock
->sk
;
4900 struct crypto_hash
*tfm
= NULL
;
4903 if (unlikely(backlog
< 0))
4908 if (sock
->state
!= SS_UNCONNECTED
)
4911 /* Allocate HMAC for generating cookie. */
4912 if (sctp_hmac_alg
) {
4913 tfm
= crypto_alloc_hash(sctp_hmac_alg
, 0, CRYPTO_ALG_ASYNC
);
4920 switch (sock
->type
) {
4921 case SOCK_SEQPACKET
:
4922 err
= sctp_seqpacket_listen(sk
, backlog
);
4925 err
= sctp_stream_listen(sk
, backlog
);
4933 /* Store away the transform reference. */
4934 sctp_sk(sk
)->hmac
= tfm
;
4936 sctp_release_sock(sk
);
4939 crypto_free_hash(tfm
);
4944 * This function is done by modeling the current datagram_poll() and the
4945 * tcp_poll(). Note that, based on these implementations, we don't
4946 * lock the socket in this function, even though it seems that,
4947 * ideally, locking or some other mechanisms can be used to ensure
4948 * the integrity of the counters (sndbuf and wmem_alloc) used
4949 * in this place. We assume that we don't need locks either until proven
4952 * Another thing to note is that we include the Async I/O support
4953 * here, again, by modeling the current TCP/UDP code. We don't have
4954 * a good way to test with it yet.
4956 unsigned int sctp_poll(struct file
*file
, struct socket
*sock
, poll_table
*wait
)
4958 struct sock
*sk
= sock
->sk
;
4959 struct sctp_sock
*sp
= sctp_sk(sk
);
4962 poll_wait(file
, sk
->sk_sleep
, wait
);
4964 /* A TCP-style listening socket becomes readable when the accept queue
4967 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
4968 return (!list_empty(&sp
->ep
->asocs
)) ?
4969 (POLLIN
| POLLRDNORM
) : 0;
4973 /* Is there any exceptional events? */
4974 if (sk
->sk_err
|| !skb_queue_empty(&sk
->sk_error_queue
))
4976 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
4978 if (sk
->sk_shutdown
== SHUTDOWN_MASK
)
4981 /* Is it readable? Reconsider this code with TCP-style support. */
4982 if (!skb_queue_empty(&sk
->sk_receive_queue
) ||
4983 (sk
->sk_shutdown
& RCV_SHUTDOWN
))
4984 mask
|= POLLIN
| POLLRDNORM
;
4986 /* The association is either gone or not ready. */
4987 if (!sctp_style(sk
, UDP
) && sctp_sstate(sk
, CLOSED
))
4990 /* Is it writable? */
4991 if (sctp_writeable(sk
)) {
4992 mask
|= POLLOUT
| POLLWRNORM
;
4994 set_bit(SOCK_ASYNC_NOSPACE
, &sk
->sk_socket
->flags
);
4996 * Since the socket is not locked, the buffer
4997 * might be made available after the writeable check and
4998 * before the bit is set. This could cause a lost I/O
4999 * signal. tcp_poll() has a race breaker for this race
5000 * condition. Based on their implementation, we put
5001 * in the following code to cover it as well.
5003 if (sctp_writeable(sk
))
5004 mask
|= POLLOUT
| POLLWRNORM
;
5009 /********************************************************************
5010 * 2nd Level Abstractions
5011 ********************************************************************/
5013 static struct sctp_bind_bucket
*sctp_bucket_create(
5014 struct sctp_bind_hashbucket
*head
, unsigned short snum
)
5016 struct sctp_bind_bucket
*pp
;
5018 pp
= kmem_cache_alloc(sctp_bucket_cachep
, SLAB_ATOMIC
);
5019 SCTP_DBG_OBJCNT_INC(bind_bucket
);
5023 INIT_HLIST_HEAD(&pp
->owner
);
5024 if ((pp
->next
= head
->chain
) != NULL
)
5025 pp
->next
->pprev
= &pp
->next
;
5027 pp
->pprev
= &head
->chain
;
5032 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5033 static void sctp_bucket_destroy(struct sctp_bind_bucket
*pp
)
5035 if (pp
&& hlist_empty(&pp
->owner
)) {
5037 pp
->next
->pprev
= pp
->pprev
;
5038 *(pp
->pprev
) = pp
->next
;
5039 kmem_cache_free(sctp_bucket_cachep
, pp
);
5040 SCTP_DBG_OBJCNT_DEC(bind_bucket
);
5044 /* Release this socket's reference to a local port. */
5045 static inline void __sctp_put_port(struct sock
*sk
)
5047 struct sctp_bind_hashbucket
*head
=
5048 &sctp_port_hashtable
[sctp_phashfn(inet_sk(sk
)->num
)];
5049 struct sctp_bind_bucket
*pp
;
5051 sctp_spin_lock(&head
->lock
);
5052 pp
= sctp_sk(sk
)->bind_hash
;
5053 __sk_del_bind_node(sk
);
5054 sctp_sk(sk
)->bind_hash
= NULL
;
5055 inet_sk(sk
)->num
= 0;
5056 sctp_bucket_destroy(pp
);
5057 sctp_spin_unlock(&head
->lock
);
5060 void sctp_put_port(struct sock
*sk
)
5062 sctp_local_bh_disable();
5063 __sctp_put_port(sk
);
5064 sctp_local_bh_enable();
5068 * The system picks an ephemeral port and choose an address set equivalent
5069 * to binding with a wildcard address.
5070 * One of those addresses will be the primary address for the association.
5071 * This automatically enables the multihoming capability of SCTP.
5073 static int sctp_autobind(struct sock
*sk
)
5075 union sctp_addr autoaddr
;
5077 unsigned short port
;
5079 /* Initialize a local sockaddr structure to INADDR_ANY. */
5080 af
= sctp_sk(sk
)->pf
->af
;
5082 port
= htons(inet_sk(sk
)->num
);
5083 af
->inaddr_any(&autoaddr
, port
);
5085 return sctp_do_bind(sk
, &autoaddr
, af
->sockaddr_len
);
5088 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5091 * 4.2 The cmsghdr Structure *
5093 * When ancillary data is sent or received, any number of ancillary data
5094 * objects can be specified by the msg_control and msg_controllen members of
5095 * the msghdr structure, because each object is preceded by
5096 * a cmsghdr structure defining the object's length (the cmsg_len member).
5097 * Historically Berkeley-derived implementations have passed only one object
5098 * at a time, but this API allows multiple objects to be
5099 * passed in a single call to sendmsg() or recvmsg(). The following example
5100 * shows two ancillary data objects in a control buffer.
5102 * |<--------------------------- msg_controllen -------------------------->|
5105 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5107 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5110 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5112 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5115 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5116 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5118 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5120 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5127 SCTP_STATIC
int sctp_msghdr_parse(const struct msghdr
*msg
,
5128 sctp_cmsgs_t
*cmsgs
)
5130 struct cmsghdr
*cmsg
;
5132 for (cmsg
= CMSG_FIRSTHDR(msg
);
5134 cmsg
= CMSG_NXTHDR((struct msghdr
*)msg
, cmsg
)) {
5135 if (!CMSG_OK(msg
, cmsg
))
5138 /* Should we parse this header or ignore? */
5139 if (cmsg
->cmsg_level
!= IPPROTO_SCTP
)
5142 /* Strictly check lengths following example in SCM code. */
5143 switch (cmsg
->cmsg_type
) {
5145 /* SCTP Socket API Extension
5146 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5148 * This cmsghdr structure provides information for
5149 * initializing new SCTP associations with sendmsg().
5150 * The SCTP_INITMSG socket option uses this same data
5151 * structure. This structure is not used for
5154 * cmsg_level cmsg_type cmsg_data[]
5155 * ------------ ------------ ----------------------
5156 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5158 if (cmsg
->cmsg_len
!=
5159 CMSG_LEN(sizeof(struct sctp_initmsg
)))
5161 cmsgs
->init
= (struct sctp_initmsg
*)CMSG_DATA(cmsg
);
5165 /* SCTP Socket API Extension
5166 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5168 * This cmsghdr structure specifies SCTP options for
5169 * sendmsg() and describes SCTP header information
5170 * about a received message through recvmsg().
5172 * cmsg_level cmsg_type cmsg_data[]
5173 * ------------ ------------ ----------------------
5174 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5176 if (cmsg
->cmsg_len
!=
5177 CMSG_LEN(sizeof(struct sctp_sndrcvinfo
)))
5181 (struct sctp_sndrcvinfo
*)CMSG_DATA(cmsg
);
5183 /* Minimally, validate the sinfo_flags. */
5184 if (cmsgs
->info
->sinfo_flags
&
5185 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
5186 SCTP_ABORT
| SCTP_EOF
))
5198 * Wait for a packet..
5199 * Note: This function is the same function as in core/datagram.c
5200 * with a few modifications to make lksctp work.
5202 static int sctp_wait_for_packet(struct sock
* sk
, int *err
, long *timeo_p
)
5207 prepare_to_wait_exclusive(sk
->sk_sleep
, &wait
, TASK_INTERRUPTIBLE
);
5209 /* Socket errors? */
5210 error
= sock_error(sk
);
5214 if (!skb_queue_empty(&sk
->sk_receive_queue
))
5217 /* Socket shut down? */
5218 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
5221 /* Sequenced packets can come disconnected. If so we report the
5226 /* Is there a good reason to think that we may receive some data? */
5227 if (list_empty(&sctp_sk(sk
)->ep
->asocs
) && !sctp_sstate(sk
, LISTENING
))
5230 /* Handle signals. */
5231 if (signal_pending(current
))
5234 /* Let another process have a go. Since we are going to sleep
5235 * anyway. Note: This may cause odd behaviors if the message
5236 * does not fit in the user's buffer, but this seems to be the
5237 * only way to honor MSG_DONTWAIT realistically.
5239 sctp_release_sock(sk
);
5240 *timeo_p
= schedule_timeout(*timeo_p
);
5244 finish_wait(sk
->sk_sleep
, &wait
);
5248 error
= sock_intr_errno(*timeo_p
);
5251 finish_wait(sk
->sk_sleep
, &wait
);
5256 /* Receive a datagram.
5257 * Note: This is pretty much the same routine as in core/datagram.c
5258 * with a few changes to make lksctp work.
5260 static struct sk_buff
*sctp_skb_recv_datagram(struct sock
*sk
, int flags
,
5261 int noblock
, int *err
)
5264 struct sk_buff
*skb
;
5267 timeo
= sock_rcvtimeo(sk
, noblock
);
5269 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
5270 timeo
, MAX_SCHEDULE_TIMEOUT
);
5273 /* Again only user level code calls this function,
5274 * so nothing interrupt level
5275 * will suddenly eat the receive_queue.
5277 * Look at current nfs client by the way...
5278 * However, this function was corrent in any case. 8)
5280 if (flags
& MSG_PEEK
) {
5281 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
5282 skb
= skb_peek(&sk
->sk_receive_queue
);
5284 atomic_inc(&skb
->users
);
5285 spin_unlock_bh(&sk
->sk_receive_queue
.lock
);
5287 skb
= skb_dequeue(&sk
->sk_receive_queue
);
5293 /* Caller is allowed not to check sk->sk_err before calling. */
5294 error
= sock_error(sk
);
5298 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
5301 /* User doesn't want to wait. */
5305 } while (sctp_wait_for_packet(sk
, err
, &timeo
) == 0);
5314 /* If sndbuf has changed, wake up per association sndbuf waiters. */
5315 static void __sctp_write_space(struct sctp_association
*asoc
)
5317 struct sock
*sk
= asoc
->base
.sk
;
5318 struct socket
*sock
= sk
->sk_socket
;
5320 if ((sctp_wspace(asoc
) > 0) && sock
) {
5321 if (waitqueue_active(&asoc
->wait
))
5322 wake_up_interruptible(&asoc
->wait
);
5324 if (sctp_writeable(sk
)) {
5325 if (sk
->sk_sleep
&& waitqueue_active(sk
->sk_sleep
))
5326 wake_up_interruptible(sk
->sk_sleep
);
5328 /* Note that we try to include the Async I/O support
5329 * here by modeling from the current TCP/UDP code.
5330 * We have not tested with it yet.
5332 if (sock
->fasync_list
&&
5333 !(sk
->sk_shutdown
& SEND_SHUTDOWN
))
5334 sock_wake_async(sock
, 2, POLL_OUT
);
5339 /* Do accounting for the sndbuf space.
5340 * Decrement the used sndbuf space of the corresponding association by the
5341 * data size which was just transmitted(freed).
5343 static void sctp_wfree(struct sk_buff
*skb
)
5345 struct sctp_association
*asoc
;
5346 struct sctp_chunk
*chunk
;
5349 /* Get the saved chunk pointer. */
5350 chunk
= *((struct sctp_chunk
**)(skb
->cb
));
5353 asoc
->sndbuf_used
-= SCTP_DATA_SNDSIZE(chunk
) +
5354 sizeof(struct sk_buff
) +
5355 sizeof(struct sctp_chunk
);
5357 atomic_sub(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
5360 __sctp_write_space(asoc
);
5362 sctp_association_put(asoc
);
5365 /* Do accounting for the receive space on the socket.
5366 * Accounting for the association is done in ulpevent.c
5367 * We set this as a destructor for the cloned data skbs so that
5368 * accounting is done at the correct time.
5370 void sctp_sock_rfree(struct sk_buff
*skb
)
5372 struct sock
*sk
= skb
->sk
;
5373 struct sctp_ulpevent
*event
= sctp_skb2event(skb
);
5375 atomic_sub(event
->rmem_len
, &sk
->sk_rmem_alloc
);
5379 /* Helper function to wait for space in the sndbuf. */
5380 static int sctp_wait_for_sndbuf(struct sctp_association
*asoc
, long *timeo_p
,
5383 struct sock
*sk
= asoc
->base
.sk
;
5385 long current_timeo
= *timeo_p
;
5388 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
5389 asoc
, (long)(*timeo_p
), msg_len
);
5391 /* Increment the association's refcnt. */
5392 sctp_association_hold(asoc
);
5394 /* Wait on the association specific sndbuf space. */
5396 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
5397 TASK_INTERRUPTIBLE
);
5400 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
5403 if (signal_pending(current
))
5404 goto do_interrupted
;
5405 if (msg_len
<= sctp_wspace(asoc
))
5408 /* Let another process have a go. Since we are going
5411 sctp_release_sock(sk
);
5412 current_timeo
= schedule_timeout(current_timeo
);
5413 BUG_ON(sk
!= asoc
->base
.sk
);
5416 *timeo_p
= current_timeo
;
5420 finish_wait(&asoc
->wait
, &wait
);
5422 /* Release the association's refcnt. */
5423 sctp_association_put(asoc
);
5432 err
= sock_intr_errno(*timeo_p
);
5440 /* If socket sndbuf has changed, wake up all per association waiters. */
5441 void sctp_write_space(struct sock
*sk
)
5443 struct sctp_association
*asoc
;
5444 struct list_head
*pos
;
5446 /* Wake up the tasks in each wait queue. */
5447 list_for_each(pos
, &((sctp_sk(sk
))->ep
->asocs
)) {
5448 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
5449 __sctp_write_space(asoc
);
5453 /* Is there any sndbuf space available on the socket?
5455 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
5456 * associations on the same socket. For a UDP-style socket with
5457 * multiple associations, it is possible for it to be "unwriteable"
5458 * prematurely. I assume that this is acceptable because
5459 * a premature "unwriteable" is better than an accidental "writeable" which
5460 * would cause an unwanted block under certain circumstances. For the 1-1
5461 * UDP-style sockets or TCP-style sockets, this code should work.
5464 static int sctp_writeable(struct sock
*sk
)
5468 amt
= sk
->sk_sndbuf
- atomic_read(&sk
->sk_wmem_alloc
);
5474 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
5475 * returns immediately with EINPROGRESS.
5477 static int sctp_wait_for_connect(struct sctp_association
*asoc
, long *timeo_p
)
5479 struct sock
*sk
= asoc
->base
.sk
;
5481 long current_timeo
= *timeo_p
;
5484 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __FUNCTION__
, asoc
,
5487 /* Increment the association's refcnt. */
5488 sctp_association_hold(asoc
);
5491 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
5492 TASK_INTERRUPTIBLE
);
5495 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
5497 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
5500 if (signal_pending(current
))
5501 goto do_interrupted
;
5503 if (sctp_state(asoc
, ESTABLISHED
))
5506 /* Let another process have a go. Since we are going
5509 sctp_release_sock(sk
);
5510 current_timeo
= schedule_timeout(current_timeo
);
5513 *timeo_p
= current_timeo
;
5517 finish_wait(&asoc
->wait
, &wait
);
5519 /* Release the association's refcnt. */
5520 sctp_association_put(asoc
);
5525 if (asoc
->init_err_counter
+ 1 > asoc
->max_init_attempts
)
5528 err
= -ECONNREFUSED
;
5532 err
= sock_intr_errno(*timeo_p
);
5540 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
)
5542 struct sctp_endpoint
*ep
;
5546 ep
= sctp_sk(sk
)->ep
;
5550 prepare_to_wait_exclusive(sk
->sk_sleep
, &wait
,
5551 TASK_INTERRUPTIBLE
);
5553 if (list_empty(&ep
->asocs
)) {
5554 sctp_release_sock(sk
);
5555 timeo
= schedule_timeout(timeo
);
5560 if (!sctp_sstate(sk
, LISTENING
))
5564 if (!list_empty(&ep
->asocs
))
5567 err
= sock_intr_errno(timeo
);
5568 if (signal_pending(current
))
5576 finish_wait(sk
->sk_sleep
, &wait
);
5581 void sctp_wait_for_close(struct sock
*sk
, long timeout
)
5586 prepare_to_wait(sk
->sk_sleep
, &wait
, TASK_INTERRUPTIBLE
);
5587 if (list_empty(&sctp_sk(sk
)->ep
->asocs
))
5589 sctp_release_sock(sk
);
5590 timeout
= schedule_timeout(timeout
);
5592 } while (!signal_pending(current
) && timeout
);
5594 finish_wait(sk
->sk_sleep
, &wait
);
5597 /* Populate the fields of the newsk from the oldsk and migrate the assoc
5598 * and its messages to the newsk.
5600 static void sctp_sock_migrate(struct sock
*oldsk
, struct sock
*newsk
,
5601 struct sctp_association
*assoc
,
5602 sctp_socket_type_t type
)
5604 struct sctp_sock
*oldsp
= sctp_sk(oldsk
);
5605 struct sctp_sock
*newsp
= sctp_sk(newsk
);
5606 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
5607 struct sctp_endpoint
*newep
= newsp
->ep
;
5608 struct sk_buff
*skb
, *tmp
;
5609 struct sctp_ulpevent
*event
;
5612 /* Migrate socket buffer sizes and all the socket level options to the
5615 newsk
->sk_sndbuf
= oldsk
->sk_sndbuf
;
5616 newsk
->sk_rcvbuf
= oldsk
->sk_rcvbuf
;
5617 /* Brute force copy old sctp opt. */
5618 inet_sk_copy_descendant(newsk
, oldsk
);
5620 /* Restore the ep value that was overwritten with the above structure
5626 /* Hook this new socket in to the bind_hash list. */
5627 pp
= sctp_sk(oldsk
)->bind_hash
;
5628 sk_add_bind_node(newsk
, &pp
->owner
);
5629 sctp_sk(newsk
)->bind_hash
= pp
;
5630 inet_sk(newsk
)->num
= inet_sk(oldsk
)->num
;
5632 /* Copy the bind_addr list from the original endpoint to the new
5633 * endpoint so that we can handle restarts properly
5635 if (PF_INET6
== assoc
->base
.sk
->sk_family
)
5636 flags
= SCTP_ADDR6_ALLOWED
;
5637 if (assoc
->peer
.ipv4_address
)
5638 flags
|= SCTP_ADDR4_PEERSUPP
;
5639 if (assoc
->peer
.ipv6_address
)
5640 flags
|= SCTP_ADDR6_PEERSUPP
;
5641 sctp_bind_addr_copy(&newsp
->ep
->base
.bind_addr
,
5642 &oldsp
->ep
->base
.bind_addr
,
5643 SCTP_SCOPE_GLOBAL
, GFP_KERNEL
, flags
);
5645 /* Move any messages in the old socket's receive queue that are for the
5646 * peeled off association to the new socket's receive queue.
5648 sctp_skb_for_each(skb
, &oldsk
->sk_receive_queue
, tmp
) {
5649 event
= sctp_skb2event(skb
);
5650 if (event
->asoc
== assoc
) {
5651 sctp_sock_rfree(skb
);
5652 __skb_unlink(skb
, &oldsk
->sk_receive_queue
);
5653 __skb_queue_tail(&newsk
->sk_receive_queue
, skb
);
5654 sctp_skb_set_owner_r(skb
, newsk
);
5658 /* Clean up any messages pending delivery due to partial
5659 * delivery. Three cases:
5660 * 1) No partial deliver; no work.
5661 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
5662 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
5664 skb_queue_head_init(&newsp
->pd_lobby
);
5665 sctp_sk(newsk
)->pd_mode
= assoc
->ulpq
.pd_mode
;
5667 if (sctp_sk(oldsk
)->pd_mode
) {
5668 struct sk_buff_head
*queue
;
5670 /* Decide which queue to move pd_lobby skbs to. */
5671 if (assoc
->ulpq
.pd_mode
) {
5672 queue
= &newsp
->pd_lobby
;
5674 queue
= &newsk
->sk_receive_queue
;
5676 /* Walk through the pd_lobby, looking for skbs that
5677 * need moved to the new socket.
5679 sctp_skb_for_each(skb
, &oldsp
->pd_lobby
, tmp
) {
5680 event
= sctp_skb2event(skb
);
5681 if (event
->asoc
== assoc
) {
5682 sctp_sock_rfree(skb
);
5683 __skb_unlink(skb
, &oldsp
->pd_lobby
);
5684 __skb_queue_tail(queue
, skb
);
5685 sctp_skb_set_owner_r(skb
, newsk
);
5689 /* Clear up any skbs waiting for the partial
5690 * delivery to finish.
5692 if (assoc
->ulpq
.pd_mode
)
5693 sctp_clear_pd(oldsk
);
5697 /* Set the type of socket to indicate that it is peeled off from the
5698 * original UDP-style socket or created with the accept() call on a
5699 * TCP-style socket..
5703 /* Mark the new socket "in-use" by the user so that any packets
5704 * that may arrive on the association after we've moved it are
5705 * queued to the backlog. This prevents a potential race between
5706 * backlog processing on the old socket and new-packet processing
5707 * on the new socket.
5709 sctp_lock_sock(newsk
);
5710 sctp_assoc_migrate(assoc
, newsk
);
5712 /* If the association on the newsk is already closed before accept()
5713 * is called, set RCV_SHUTDOWN flag.
5715 if (sctp_state(assoc
, CLOSED
) && sctp_style(newsk
, TCP
))
5716 newsk
->sk_shutdown
|= RCV_SHUTDOWN
;
5718 newsk
->sk_state
= SCTP_SS_ESTABLISHED
;
5719 sctp_release_sock(newsk
);
5722 /* This proto struct describes the ULP interface for SCTP. */
5723 struct proto sctp_prot
= {
5725 .owner
= THIS_MODULE
,
5726 .close
= sctp_close
,
5727 .connect
= sctp_connect
,
5728 .disconnect
= sctp_disconnect
,
5729 .accept
= sctp_accept
,
5730 .ioctl
= sctp_ioctl
,
5731 .init
= sctp_init_sock
,
5732 .destroy
= sctp_destroy_sock
,
5733 .shutdown
= sctp_shutdown
,
5734 .setsockopt
= sctp_setsockopt
,
5735 .getsockopt
= sctp_getsockopt
,
5736 .sendmsg
= sctp_sendmsg
,
5737 .recvmsg
= sctp_recvmsg
,
5739 .backlog_rcv
= sctp_backlog_rcv
,
5741 .unhash
= sctp_unhash
,
5742 .get_port
= sctp_get_port
,
5743 .obj_size
= sizeof(struct sctp_sock
),
5746 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
5747 struct proto sctpv6_prot
= {
5749 .owner
= THIS_MODULE
,
5750 .close
= sctp_close
,
5751 .connect
= sctp_connect
,
5752 .disconnect
= sctp_disconnect
,
5753 .accept
= sctp_accept
,
5754 .ioctl
= sctp_ioctl
,
5755 .init
= sctp_init_sock
,
5756 .destroy
= sctp_destroy_sock
,
5757 .shutdown
= sctp_shutdown
,
5758 .setsockopt
= sctp_setsockopt
,
5759 .getsockopt
= sctp_getsockopt
,
5760 .sendmsg
= sctp_sendmsg
,
5761 .recvmsg
= sctp_recvmsg
,
5763 .backlog_rcv
= sctp_backlog_rcv
,
5765 .unhash
= sctp_unhash
,
5766 .get_port
= sctp_get_port
,
5767 .obj_size
= sizeof(struct sctp6_sock
),
5769 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */