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
;
611 union sctp_addr
*sa_addr
;
614 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
618 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
619 /* If the bind address list is empty or if there is only one
620 * bind address, there is nothing more to be removed (we need
621 * at least one address here).
623 if (list_empty(&bp
->address_list
) ||
624 (sctp_list_single_entry(&bp
->address_list
))) {
629 sa_addr
= (union sctp_addr
*)addr_buf
;
630 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
635 if (sa_addr
->v4
.sin_port
!= htons(bp
->port
)) {
640 /* FIXME - There is probably a need to check if sk->sk_saddr and
641 * sk->sk_rcv_addr are currently set to one of the addresses to
642 * be removed. This is something which needs to be looked into
643 * when we are fixing the outstanding issues with multi-homing
644 * socket routing and failover schemes. Refer to comments in
645 * sctp_do_bind(). -daisy
647 sctp_local_bh_disable();
648 sctp_write_lock(&ep
->base
.addr_lock
);
650 retval
= sctp_del_bind_addr(bp
, sa_addr
);
652 sctp_write_unlock(&ep
->base
.addr_lock
);
653 sctp_local_bh_enable();
655 addr_buf
+= af
->sockaddr_len
;
658 /* Failed. Add the ones that has been removed back */
660 sctp_bindx_add(sk
, addrs
, cnt
);
668 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
669 * the associations that are part of the endpoint indicating that a list of
670 * local addresses are removed from the endpoint.
672 * If any of the addresses is already in the bind address list of the
673 * association, we do not send the chunk for that association. But it will not
674 * affect other associations.
676 * Only sctp_setsockopt_bindx() is supposed to call this function.
678 static int sctp_send_asconf_del_ip(struct sock
*sk
,
679 struct sockaddr
*addrs
,
682 struct sctp_sock
*sp
;
683 struct sctp_endpoint
*ep
;
684 struct sctp_association
*asoc
;
685 struct sctp_transport
*transport
;
686 struct sctp_bind_addr
*bp
;
687 struct sctp_chunk
*chunk
;
688 union sctp_addr
*laddr
;
691 struct list_head
*pos
, *pos1
;
692 struct sctp_sockaddr_entry
*saddr
;
696 if (!sctp_addip_enable
)
702 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
703 __FUNCTION__
, sk
, addrs
, addrcnt
);
705 list_for_each(pos
, &ep
->asocs
) {
706 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
708 if (!asoc
->peer
.asconf_capable
)
711 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_DEL_IP
)
714 if (!sctp_state(asoc
, ESTABLISHED
))
717 /* Check if any address in the packed array of addresses is
718 * not present in the bind address list of the association.
719 * If so, do not send the asconf chunk to its peer, but
720 * continue with other associations.
723 for (i
= 0; i
< addrcnt
; i
++) {
724 laddr
= (union sctp_addr
*)addr_buf
;
725 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
731 if (!sctp_assoc_lookup_laddr(asoc
, laddr
))
734 addr_buf
+= af
->sockaddr_len
;
739 /* Find one address in the association's bind address list
740 * that is not in the packed array of addresses. This is to
741 * make sure that we do not delete all the addresses in the
744 sctp_read_lock(&asoc
->base
.addr_lock
);
745 bp
= &asoc
->base
.bind_addr
;
746 laddr
= sctp_find_unmatch_addr(bp
, (union sctp_addr
*)addrs
,
748 sctp_read_unlock(&asoc
->base
.addr_lock
);
752 chunk
= sctp_make_asconf_update_ip(asoc
, laddr
, addrs
, addrcnt
,
759 /* Reset use_as_src flag for the addresses in the bind address
760 * list that are to be deleted.
762 sctp_local_bh_disable();
763 sctp_write_lock(&asoc
->base
.addr_lock
);
765 for (i
= 0; i
< addrcnt
; i
++) {
766 laddr
= (union sctp_addr
*)addr_buf
;
767 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
768 list_for_each(pos1
, &bp
->address_list
) {
769 saddr
= list_entry(pos1
,
770 struct sctp_sockaddr_entry
,
772 if (sctp_cmp_addr_exact(&saddr
->a
, laddr
))
773 saddr
->use_as_src
= 0;
775 addr_buf
+= af
->sockaddr_len
;
777 sctp_write_unlock(&asoc
->base
.addr_lock
);
778 sctp_local_bh_enable();
780 /* Update the route and saddr entries for all the transports
781 * as some of the addresses in the bind address list are
782 * about to be deleted and cannot be used as source addresses.
784 list_for_each(pos1
, &asoc
->peer
.transport_addr_list
) {
785 transport
= list_entry(pos1
, struct sctp_transport
,
787 dst_release(transport
->dst
);
788 sctp_transport_route(transport
, NULL
,
789 sctp_sk(asoc
->base
.sk
));
792 retval
= sctp_send_asconf(asoc
, chunk
);
798 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
801 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
804 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
805 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
808 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
809 * Section 3.1.2 for this usage.
811 * addrs is a pointer to an array of one or more socket addresses. Each
812 * address is contained in its appropriate structure (i.e. struct
813 * sockaddr_in or struct sockaddr_in6) the family of the address type
814 * must be used to distinguish the address length (note that this
815 * representation is termed a "packed array" of addresses). The caller
816 * specifies the number of addresses in the array with addrcnt.
818 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
819 * -1, and sets errno to the appropriate error code.
821 * For SCTP, the port given in each socket address must be the same, or
822 * sctp_bindx() will fail, setting errno to EINVAL.
824 * The flags parameter is formed from the bitwise OR of zero or more of
825 * the following currently defined flags:
827 * SCTP_BINDX_ADD_ADDR
829 * SCTP_BINDX_REM_ADDR
831 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
832 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
833 * addresses from the association. The two flags are mutually exclusive;
834 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
835 * not remove all addresses from an association; sctp_bindx() will
836 * reject such an attempt with EINVAL.
838 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
839 * additional addresses with an endpoint after calling bind(). Or use
840 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
841 * socket is associated with so that no new association accepted will be
842 * associated with those addresses. If the endpoint supports dynamic
843 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
844 * endpoint to send the appropriate message to the peer to change the
845 * peers address lists.
847 * Adding and removing addresses from a connected association is
848 * optional functionality. Implementations that do not support this
849 * functionality should return EOPNOTSUPP.
851 * Basically do nothing but copying the addresses from user to kernel
852 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
853 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
856 * We don't use copy_from_user() for optimization: we first do the
857 * sanity checks (buffer size -fast- and access check-healthy
858 * pointer); if all of those succeed, then we can alloc the memory
859 * (expensive operation) needed to copy the data to kernel. Then we do
860 * the copying without checking the user space area
861 * (__copy_from_user()).
863 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
866 * sk The sk of the socket
867 * addrs The pointer to the addresses in user land
868 * addrssize Size of the addrs buffer
869 * op Operation to perform (add or remove, see the flags of
872 * Returns 0 if ok, <0 errno code on error.
874 SCTP_STATIC
int sctp_setsockopt_bindx(struct sock
* sk
,
875 struct sockaddr __user
*addrs
,
876 int addrs_size
, int op
)
878 struct sockaddr
*kaddrs
;
882 struct sockaddr
*sa_addr
;
886 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
887 " addrs_size %d opt %d\n", sk
, addrs
, addrs_size
, op
);
889 if (unlikely(addrs_size
<= 0))
892 /* Check the user passed a healthy pointer. */
893 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
896 /* Alloc space for the address array in kernel memory. */
897 kaddrs
= kmalloc(addrs_size
, GFP_KERNEL
);
898 if (unlikely(!kaddrs
))
901 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
906 /* Walk through the addrs buffer and count the number of addresses. */
908 while (walk_size
< addrs_size
) {
909 sa_addr
= (struct sockaddr
*)addr_buf
;
910 af
= sctp_get_af_specific(sa_addr
->sa_family
);
912 /* If the address family is not supported or if this address
913 * causes the address buffer to overflow return EINVAL.
915 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
920 addr_buf
+= af
->sockaddr_len
;
921 walk_size
+= af
->sockaddr_len
;
926 case SCTP_BINDX_ADD_ADDR
:
927 err
= sctp_bindx_add(sk
, kaddrs
, addrcnt
);
930 err
= sctp_send_asconf_add_ip(sk
, kaddrs
, addrcnt
);
933 case SCTP_BINDX_REM_ADDR
:
934 err
= sctp_bindx_rem(sk
, kaddrs
, addrcnt
);
937 err
= sctp_send_asconf_del_ip(sk
, kaddrs
, addrcnt
);
951 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
953 * Common routine for handling connect() and sctp_connectx().
954 * Connect will come in with just a single address.
956 static int __sctp_connect(struct sock
* sk
,
957 struct sockaddr
*kaddrs
,
960 struct sctp_sock
*sp
;
961 struct sctp_endpoint
*ep
;
962 struct sctp_association
*asoc
= NULL
;
963 struct sctp_association
*asoc2
;
964 struct sctp_transport
*transport
;
972 struct sockaddr
*sa_addr
;
978 /* connect() cannot be done on a socket that is already in ESTABLISHED
979 * state - UDP-style peeled off socket or a TCP-style socket that
980 * is already connected.
981 * It cannot be done even on a TCP-style listening socket.
983 if (sctp_sstate(sk
, ESTABLISHED
) ||
984 (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))) {
989 /* Walk through the addrs buffer and count the number of addresses. */
991 while (walk_size
< addrs_size
) {
992 sa_addr
= (struct sockaddr
*)addr_buf
;
993 af
= sctp_get_af_specific(sa_addr
->sa_family
);
995 /* If the address family is not supported or if this address
996 * causes the address buffer to overflow return EINVAL.
998 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
1003 err
= sctp_verify_addr(sk
, (union sctp_addr
*)sa_addr
,
1008 memcpy(&to
, sa_addr
, af
->sockaddr_len
);
1009 to
.v4
.sin_port
= ntohs(to
.v4
.sin_port
);
1011 /* Check if there already is a matching association on the
1012 * endpoint (other than the one created here).
1014 asoc2
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1015 if (asoc2
&& asoc2
!= asoc
) {
1016 if (asoc2
->state
>= SCTP_STATE_ESTABLISHED
)
1023 /* If we could not find a matching association on the endpoint,
1024 * make sure that there is no peeled-off association matching
1025 * the peer address even on another socket.
1027 if (sctp_endpoint_is_peeled_off(ep
, &to
)) {
1028 err
= -EADDRNOTAVAIL
;
1033 /* If a bind() or sctp_bindx() is not called prior to
1034 * an sctp_connectx() call, the system picks an
1035 * ephemeral port and will choose an address set
1036 * equivalent to binding with a wildcard address.
1038 if (!ep
->base
.bind_addr
.port
) {
1039 if (sctp_autobind(sk
)) {
1045 * If an unprivileged user inherits a 1-many
1046 * style socket with open associations on a
1047 * privileged port, it MAY be permitted to
1048 * accept new associations, but it SHOULD NOT
1049 * be permitted to open new associations.
1051 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1052 !capable(CAP_NET_BIND_SERVICE
)) {
1058 scope
= sctp_scope(&to
);
1059 asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1066 /* Prime the peer's transport structures. */
1067 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
,
1075 addr_buf
+= af
->sockaddr_len
;
1076 walk_size
+= af
->sockaddr_len
;
1079 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, GFP_KERNEL
);
1084 err
= sctp_primitive_ASSOCIATE(asoc
, NULL
);
1089 /* Initialize sk's dport and daddr for getpeername() */
1090 inet_sk(sk
)->dport
= htons(asoc
->peer
.port
);
1091 af
= sctp_get_af_specific(to
.sa
.sa_family
);
1092 af
->to_sk_daddr(&to
, sk
);
1095 timeo
= sock_sndtimeo(sk
, sk
->sk_socket
->file
->f_flags
& O_NONBLOCK
);
1096 err
= sctp_wait_for_connect(asoc
, &timeo
);
1098 /* Don't free association on exit. */
1103 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1104 " kaddrs: %p err: %d\n",
1107 sctp_association_free(asoc
);
1111 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1114 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt);
1116 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1117 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1118 * or IPv6 addresses.
1120 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1121 * Section 3.1.2 for this usage.
1123 * addrs is a pointer to an array of one or more socket addresses. Each
1124 * address is contained in its appropriate structure (i.e. struct
1125 * sockaddr_in or struct sockaddr_in6) the family of the address type
1126 * must be used to distengish the address length (note that this
1127 * representation is termed a "packed array" of addresses). The caller
1128 * specifies the number of addresses in the array with addrcnt.
1130 * On success, sctp_connectx() returns 0. On failure, sctp_connectx() returns
1131 * -1, and sets errno to the appropriate error code.
1133 * For SCTP, the port given in each socket address must be the same, or
1134 * sctp_connectx() will fail, setting errno to EINVAL.
1136 * An application can use sctp_connectx to initiate an association with
1137 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1138 * allows a caller to specify multiple addresses at which a peer can be
1139 * reached. The way the SCTP stack uses the list of addresses to set up
1140 * the association is implementation dependant. This function only
1141 * specifies that the stack will try to make use of all the addresses in
1142 * the list when needed.
1144 * Note that the list of addresses passed in is only used for setting up
1145 * the association. It does not necessarily equal the set of addresses
1146 * the peer uses for the resulting association. If the caller wants to
1147 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1148 * retrieve them after the association has been set up.
1150 * Basically do nothing but copying the addresses from user to kernel
1151 * land and invoking either sctp_connectx(). This is used for tunneling
1152 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1154 * We don't use copy_from_user() for optimization: we first do the
1155 * sanity checks (buffer size -fast- and access check-healthy
1156 * pointer); if all of those succeed, then we can alloc the memory
1157 * (expensive operation) needed to copy the data to kernel. Then we do
1158 * the copying without checking the user space area
1159 * (__copy_from_user()).
1161 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1164 * sk The sk of the socket
1165 * addrs The pointer to the addresses in user land
1166 * addrssize Size of the addrs buffer
1168 * Returns 0 if ok, <0 errno code on error.
1170 SCTP_STATIC
int sctp_setsockopt_connectx(struct sock
* sk
,
1171 struct sockaddr __user
*addrs
,
1175 struct sockaddr
*kaddrs
;
1177 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1178 __FUNCTION__
, sk
, addrs
, addrs_size
);
1180 if (unlikely(addrs_size
<= 0))
1183 /* Check the user passed a healthy pointer. */
1184 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
1187 /* Alloc space for the address array in kernel memory. */
1188 kaddrs
= kmalloc(addrs_size
, GFP_KERNEL
);
1189 if (unlikely(!kaddrs
))
1192 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
1195 err
= __sctp_connect(sk
, kaddrs
, addrs_size
);
1202 /* API 3.1.4 close() - UDP Style Syntax
1203 * Applications use close() to perform graceful shutdown (as described in
1204 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1205 * by a UDP-style socket.
1209 * ret = close(int sd);
1211 * sd - the socket descriptor of the associations to be closed.
1213 * To gracefully shutdown a specific association represented by the
1214 * UDP-style socket, an application should use the sendmsg() call,
1215 * passing no user data, but including the appropriate flag in the
1216 * ancillary data (see Section xxxx).
1218 * If sd in the close() call is a branched-off socket representing only
1219 * one association, the shutdown is performed on that association only.
1221 * 4.1.6 close() - TCP Style Syntax
1223 * Applications use close() to gracefully close down an association.
1227 * int close(int sd);
1229 * sd - the socket descriptor of the association to be closed.
1231 * After an application calls close() on a socket descriptor, no further
1232 * socket operations will succeed on that descriptor.
1234 * API 7.1.4 SO_LINGER
1236 * An application using the TCP-style socket can use this option to
1237 * perform the SCTP ABORT primitive. The linger option structure is:
1240 * int l_onoff; // option on/off
1241 * int l_linger; // linger time
1244 * To enable the option, set l_onoff to 1. If the l_linger value is set
1245 * to 0, calling close() is the same as the ABORT primitive. If the
1246 * value is set to a negative value, the setsockopt() call will return
1247 * an error. If the value is set to a positive value linger_time, the
1248 * close() can be blocked for at most linger_time ms. If the graceful
1249 * shutdown phase does not finish during this period, close() will
1250 * return but the graceful shutdown phase continues in the system.
1252 SCTP_STATIC
void sctp_close(struct sock
*sk
, long timeout
)
1254 struct sctp_endpoint
*ep
;
1255 struct sctp_association
*asoc
;
1256 struct list_head
*pos
, *temp
;
1258 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk
, timeout
);
1261 sk
->sk_shutdown
= SHUTDOWN_MASK
;
1263 ep
= sctp_sk(sk
)->ep
;
1265 /* Walk all associations on an endpoint. */
1266 list_for_each_safe(pos
, temp
, &ep
->asocs
) {
1267 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
1269 if (sctp_style(sk
, TCP
)) {
1270 /* A closed association can still be in the list if
1271 * it belongs to a TCP-style listening socket that is
1272 * not yet accepted. If so, free it. If not, send an
1273 * ABORT or SHUTDOWN based on the linger options.
1275 if (sctp_state(asoc
, CLOSED
)) {
1276 sctp_unhash_established(asoc
);
1277 sctp_association_free(asoc
);
1282 if (sock_flag(sk
, SOCK_LINGER
) && !sk
->sk_lingertime
) {
1283 struct sctp_chunk
*chunk
;
1285 chunk
= sctp_make_abort_user(asoc
, NULL
, 0);
1287 sctp_primitive_ABORT(asoc
, chunk
);
1289 sctp_primitive_SHUTDOWN(asoc
, NULL
);
1292 /* Clean up any skbs sitting on the receive queue. */
1293 sctp_queue_purge_ulpevents(&sk
->sk_receive_queue
);
1294 sctp_queue_purge_ulpevents(&sctp_sk(sk
)->pd_lobby
);
1296 /* On a TCP-style socket, block for at most linger_time if set. */
1297 if (sctp_style(sk
, TCP
) && timeout
)
1298 sctp_wait_for_close(sk
, timeout
);
1300 /* This will run the backlog queue. */
1301 sctp_release_sock(sk
);
1303 /* Supposedly, no process has access to the socket, but
1304 * the net layers still may.
1306 sctp_local_bh_disable();
1307 sctp_bh_lock_sock(sk
);
1309 /* Hold the sock, since sk_common_release() will put sock_put()
1310 * and we have just a little more cleanup.
1313 sk_common_release(sk
);
1315 sctp_bh_unlock_sock(sk
);
1316 sctp_local_bh_enable();
1320 SCTP_DBG_OBJCNT_DEC(sock
);
1323 /* Handle EPIPE error. */
1324 static int sctp_error(struct sock
*sk
, int flags
, int err
)
1327 err
= sock_error(sk
) ? : -EPIPE
;
1328 if (err
== -EPIPE
&& !(flags
& MSG_NOSIGNAL
))
1329 send_sig(SIGPIPE
, current
, 0);
1333 /* API 3.1.3 sendmsg() - UDP Style Syntax
1335 * An application uses sendmsg() and recvmsg() calls to transmit data to
1336 * and receive data from its peer.
1338 * ssize_t sendmsg(int socket, const struct msghdr *message,
1341 * socket - the socket descriptor of the endpoint.
1342 * message - pointer to the msghdr structure which contains a single
1343 * user message and possibly some ancillary data.
1345 * See Section 5 for complete description of the data
1348 * flags - flags sent or received with the user message, see Section
1349 * 5 for complete description of the flags.
1351 * Note: This function could use a rewrite especially when explicit
1352 * connect support comes in.
1354 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1356 SCTP_STATIC
int sctp_msghdr_parse(const struct msghdr
*, sctp_cmsgs_t
*);
1358 SCTP_STATIC
int sctp_sendmsg(struct kiocb
*iocb
, struct sock
*sk
,
1359 struct msghdr
*msg
, size_t msg_len
)
1361 struct sctp_sock
*sp
;
1362 struct sctp_endpoint
*ep
;
1363 struct sctp_association
*new_asoc
=NULL
, *asoc
=NULL
;
1364 struct sctp_transport
*transport
, *chunk_tp
;
1365 struct sctp_chunk
*chunk
;
1366 union sctp_addr to
, tmp
;
1367 struct sockaddr
*msg_name
= NULL
;
1368 struct sctp_sndrcvinfo default_sinfo
= { 0 };
1369 struct sctp_sndrcvinfo
*sinfo
;
1370 struct sctp_initmsg
*sinit
;
1371 sctp_assoc_t associd
= 0;
1372 sctp_cmsgs_t cmsgs
= { NULL
};
1376 __u16 sinfo_flags
= 0;
1377 struct sctp_datamsg
*datamsg
;
1378 struct list_head
*pos
;
1379 int msg_flags
= msg
->msg_flags
;
1381 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1388 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep
);
1390 /* We cannot send a message over a TCP-style listening socket. */
1391 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
)) {
1396 /* Parse out the SCTP CMSGs. */
1397 err
= sctp_msghdr_parse(msg
, &cmsgs
);
1400 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err
);
1404 /* Fetch the destination address for this packet. This
1405 * address only selects the association--it is not necessarily
1406 * the address we will send to.
1407 * For a peeled-off socket, msg_name is ignored.
1409 if (!sctp_style(sk
, UDP_HIGH_BANDWIDTH
) && msg
->msg_name
) {
1410 int msg_namelen
= msg
->msg_namelen
;
1412 err
= sctp_verify_addr(sk
, (union sctp_addr
*)msg
->msg_name
,
1417 if (msg_namelen
> sizeof(to
))
1418 msg_namelen
= sizeof(to
);
1419 memcpy(&to
, msg
->msg_name
, msg_namelen
);
1420 memcpy(&tmp
, msg
->msg_name
, msg_namelen
);
1421 SCTP_DEBUG_PRINTK("Just memcpy'd. msg_name is "
1423 to
.v4
.sin_addr
.s_addr
, to
.v4
.sin_port
);
1425 to
.v4
.sin_port
= ntohs(to
.v4
.sin_port
);
1426 msg_name
= msg
->msg_name
;
1432 /* Did the user specify SNDRCVINFO? */
1434 sinfo_flags
= sinfo
->sinfo_flags
;
1435 associd
= sinfo
->sinfo_assoc_id
;
1438 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1439 msg_len
, sinfo_flags
);
1441 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1442 if (sctp_style(sk
, TCP
) && (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
))) {
1447 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1448 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1449 * If SCTP_ABORT is set, the message length could be non zero with
1450 * the msg_iov set to the user abort reason.
1452 if (((sinfo_flags
& SCTP_EOF
) && (msg_len
> 0)) ||
1453 (!(sinfo_flags
& (SCTP_EOF
|SCTP_ABORT
)) && (msg_len
== 0))) {
1458 /* If SCTP_ADDR_OVER is set, there must be an address
1459 * specified in msg_name.
1461 if ((sinfo_flags
& SCTP_ADDR_OVER
) && (!msg
->msg_name
)) {
1468 SCTP_DEBUG_PRINTK("About to look up association.\n");
1472 /* If a msg_name has been specified, assume this is to be used. */
1474 /* Look for a matching association on the endpoint. */
1475 asoc
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1477 /* If we could not find a matching association on the
1478 * endpoint, make sure that it is not a TCP-style
1479 * socket that already has an association or there is
1480 * no peeled-off association on another socket.
1482 if ((sctp_style(sk
, TCP
) &&
1483 sctp_sstate(sk
, ESTABLISHED
)) ||
1484 sctp_endpoint_is_peeled_off(ep
, &to
)) {
1485 err
= -EADDRNOTAVAIL
;
1490 asoc
= sctp_id2assoc(sk
, associd
);
1498 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc
);
1500 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1501 * socket that has an association in CLOSED state. This can
1502 * happen when an accepted socket has an association that is
1505 if (sctp_state(asoc
, CLOSED
) && sctp_style(sk
, TCP
)) {
1510 if (sinfo_flags
& SCTP_EOF
) {
1511 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1513 sctp_primitive_SHUTDOWN(asoc
, NULL
);
1517 if (sinfo_flags
& SCTP_ABORT
) {
1518 struct sctp_chunk
*chunk
;
1520 chunk
= sctp_make_abort_user(asoc
, msg
, msg_len
);
1526 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc
);
1527 sctp_primitive_ABORT(asoc
, chunk
);
1533 /* Do we need to create the association? */
1535 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1537 if (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
)) {
1542 /* Check for invalid stream against the stream counts,
1543 * either the default or the user specified stream counts.
1546 if (!sinit
|| (sinit
&& !sinit
->sinit_num_ostreams
)) {
1547 /* Check against the defaults. */
1548 if (sinfo
->sinfo_stream
>=
1549 sp
->initmsg
.sinit_num_ostreams
) {
1554 /* Check against the requested. */
1555 if (sinfo
->sinfo_stream
>=
1556 sinit
->sinit_num_ostreams
) {
1564 * API 3.1.2 bind() - UDP Style Syntax
1565 * If a bind() or sctp_bindx() is not called prior to a
1566 * sendmsg() call that initiates a new association, the
1567 * system picks an ephemeral port and will choose an address
1568 * set equivalent to binding with a wildcard address.
1570 if (!ep
->base
.bind_addr
.port
) {
1571 if (sctp_autobind(sk
)) {
1577 * If an unprivileged user inherits a one-to-many
1578 * style socket with open associations on a privileged
1579 * port, it MAY be permitted to accept new associations,
1580 * but it SHOULD NOT be permitted to open new
1583 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1584 !capable(CAP_NET_BIND_SERVICE
)) {
1590 scope
= sctp_scope(&to
);
1591 new_asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1598 /* If the SCTP_INIT ancillary data is specified, set all
1599 * the association init values accordingly.
1602 if (sinit
->sinit_num_ostreams
) {
1603 asoc
->c
.sinit_num_ostreams
=
1604 sinit
->sinit_num_ostreams
;
1606 if (sinit
->sinit_max_instreams
) {
1607 asoc
->c
.sinit_max_instreams
=
1608 sinit
->sinit_max_instreams
;
1610 if (sinit
->sinit_max_attempts
) {
1611 asoc
->max_init_attempts
1612 = sinit
->sinit_max_attempts
;
1614 if (sinit
->sinit_max_init_timeo
) {
1615 asoc
->max_init_timeo
=
1616 msecs_to_jiffies(sinit
->sinit_max_init_timeo
);
1620 /* Prime the peer's transport structures. */
1621 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
, SCTP_UNKNOWN
);
1626 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, GFP_KERNEL
);
1633 /* ASSERT: we have a valid association at this point. */
1634 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1637 /* If the user didn't specify SNDRCVINFO, make up one with
1640 default_sinfo
.sinfo_stream
= asoc
->default_stream
;
1641 default_sinfo
.sinfo_flags
= asoc
->default_flags
;
1642 default_sinfo
.sinfo_ppid
= asoc
->default_ppid
;
1643 default_sinfo
.sinfo_context
= asoc
->default_context
;
1644 default_sinfo
.sinfo_timetolive
= asoc
->default_timetolive
;
1645 default_sinfo
.sinfo_assoc_id
= sctp_assoc2id(asoc
);
1646 sinfo
= &default_sinfo
;
1649 /* API 7.1.7, the sndbuf size per association bounds the
1650 * maximum size of data that can be sent in a single send call.
1652 if (msg_len
> sk
->sk_sndbuf
) {
1657 /* If fragmentation is disabled and the message length exceeds the
1658 * association fragmentation point, return EMSGSIZE. The I-D
1659 * does not specify what this error is, but this looks like
1662 if (sctp_sk(sk
)->disable_fragments
&& (msg_len
> asoc
->frag_point
)) {
1668 /* Check for invalid stream. */
1669 if (sinfo
->sinfo_stream
>= asoc
->c
.sinit_num_ostreams
) {
1675 timeo
= sock_sndtimeo(sk
, msg
->msg_flags
& MSG_DONTWAIT
);
1676 if (!sctp_wspace(asoc
)) {
1677 err
= sctp_wait_for_sndbuf(asoc
, &timeo
, msg_len
);
1682 /* If an address is passed with the sendto/sendmsg call, it is used
1683 * to override the primary destination address in the TCP model, or
1684 * when SCTP_ADDR_OVER flag is set in the UDP model.
1686 if ((sctp_style(sk
, TCP
) && msg_name
) ||
1687 (sinfo_flags
& SCTP_ADDR_OVER
)) {
1688 chunk_tp
= sctp_assoc_lookup_paddr(asoc
, &tmp
);
1696 /* Auto-connect, if we aren't connected already. */
1697 if (sctp_state(asoc
, CLOSED
)) {
1698 err
= sctp_primitive_ASSOCIATE(asoc
, NULL
);
1701 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1704 /* Break the message into multiple chunks of maximum size. */
1705 datamsg
= sctp_datamsg_from_user(asoc
, sinfo
, msg
, msg_len
);
1711 /* Now send the (possibly) fragmented message. */
1712 list_for_each(pos
, &datamsg
->chunks
) {
1713 chunk
= list_entry(pos
, struct sctp_chunk
, frag_list
);
1714 sctp_datamsg_track(chunk
);
1716 /* Do accounting for the write space. */
1717 sctp_set_owner_w(chunk
);
1719 chunk
->transport
= chunk_tp
;
1721 /* Send it to the lower layers. Note: all chunks
1722 * must either fail or succeed. The lower layer
1723 * works that way today. Keep it that way or this
1726 err
= sctp_primitive_SEND(asoc
, chunk
);
1727 /* Did the lower layer accept the chunk? */
1729 sctp_chunk_free(chunk
);
1730 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1733 sctp_datamsg_free(datamsg
);
1739 /* If we are already past ASSOCIATE, the lower
1740 * layers are responsible for association cleanup.
1746 sctp_association_free(asoc
);
1748 sctp_release_sock(sk
);
1751 return sctp_error(sk
, msg_flags
, err
);
1758 err
= sock_error(sk
);
1768 /* This is an extended version of skb_pull() that removes the data from the
1769 * start of a skb even when data is spread across the list of skb's in the
1770 * frag_list. len specifies the total amount of data that needs to be removed.
1771 * when 'len' bytes could be removed from the skb, it returns 0.
1772 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1773 * could not be removed.
1775 static int sctp_skb_pull(struct sk_buff
*skb
, int len
)
1777 struct sk_buff
*list
;
1778 int skb_len
= skb_headlen(skb
);
1781 if (len
<= skb_len
) {
1782 __skb_pull(skb
, len
);
1786 __skb_pull(skb
, skb_len
);
1788 for (list
= skb_shinfo(skb
)->frag_list
; list
; list
= list
->next
) {
1789 rlen
= sctp_skb_pull(list
, len
);
1790 skb
->len
-= (len
-rlen
);
1791 skb
->data_len
-= (len
-rlen
);
1802 /* API 3.1.3 recvmsg() - UDP Style Syntax
1804 * ssize_t recvmsg(int socket, struct msghdr *message,
1807 * socket - the socket descriptor of the endpoint.
1808 * message - pointer to the msghdr structure which contains a single
1809 * user message and possibly some ancillary data.
1811 * See Section 5 for complete description of the data
1814 * flags - flags sent or received with the user message, see Section
1815 * 5 for complete description of the flags.
1817 static struct sk_buff
*sctp_skb_recv_datagram(struct sock
*, int, int, int *);
1819 SCTP_STATIC
int sctp_recvmsg(struct kiocb
*iocb
, struct sock
*sk
,
1820 struct msghdr
*msg
, size_t len
, int noblock
,
1821 int flags
, int *addr_len
)
1823 struct sctp_ulpevent
*event
= NULL
;
1824 struct sctp_sock
*sp
= sctp_sk(sk
);
1825 struct sk_buff
*skb
;
1830 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1831 "0x%x, %s: %p)\n", "sk", sk
, "msghdr", msg
,
1832 "len", len
, "knoblauch", noblock
,
1833 "flags", flags
, "addr_len", addr_len
);
1837 if (sctp_style(sk
, TCP
) && !sctp_sstate(sk
, ESTABLISHED
)) {
1842 skb
= sctp_skb_recv_datagram(sk
, flags
, noblock
, &err
);
1846 /* Get the total length of the skb including any skb's in the
1855 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
1857 event
= sctp_skb2event(skb
);
1862 sock_recv_timestamp(msg
, sk
, skb
);
1863 if (sctp_ulpevent_is_notification(event
)) {
1864 msg
->msg_flags
|= MSG_NOTIFICATION
;
1865 sp
->pf
->event_msgname(event
, msg
->msg_name
, addr_len
);
1867 sp
->pf
->skb_msgname(skb
, msg
->msg_name
, addr_len
);
1870 /* Check if we allow SCTP_SNDRCVINFO. */
1871 if (sp
->subscribe
.sctp_data_io_event
)
1872 sctp_ulpevent_read_sndrcvinfo(event
, msg
);
1874 /* FIXME: we should be calling IP/IPv6 layers. */
1875 if (sk
->sk_protinfo
.af_inet
.cmsg_flags
)
1876 ip_cmsg_recv(msg
, skb
);
1881 /* If skb's length exceeds the user's buffer, update the skb and
1882 * push it back to the receive_queue so that the next call to
1883 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1885 if (skb_len
> copied
) {
1886 msg
->msg_flags
&= ~MSG_EOR
;
1887 if (flags
& MSG_PEEK
)
1889 sctp_skb_pull(skb
, copied
);
1890 skb_queue_head(&sk
->sk_receive_queue
, skb
);
1892 /* When only partial message is copied to the user, increase
1893 * rwnd by that amount. If all the data in the skb is read,
1894 * rwnd is updated when the event is freed.
1896 sctp_assoc_rwnd_increase(event
->asoc
, copied
);
1898 } else if ((event
->msg_flags
& MSG_NOTIFICATION
) ||
1899 (event
->msg_flags
& MSG_EOR
))
1900 msg
->msg_flags
|= MSG_EOR
;
1902 msg
->msg_flags
&= ~MSG_EOR
;
1905 if (flags
& MSG_PEEK
) {
1906 /* Release the skb reference acquired after peeking the skb in
1907 * sctp_skb_recv_datagram().
1911 /* Free the event which includes releasing the reference to
1912 * the owner of the skb, freeing the skb and updating the
1915 sctp_ulpevent_free(event
);
1918 sctp_release_sock(sk
);
1922 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
1924 * This option is a on/off flag. If enabled no SCTP message
1925 * fragmentation will be performed. Instead if a message being sent
1926 * exceeds the current PMTU size, the message will NOT be sent and
1927 * instead a error will be indicated to the user.
1929 static int sctp_setsockopt_disable_fragments(struct sock
*sk
,
1930 char __user
*optval
, int optlen
)
1934 if (optlen
< sizeof(int))
1937 if (get_user(val
, (int __user
*)optval
))
1940 sctp_sk(sk
)->disable_fragments
= (val
== 0) ? 0 : 1;
1945 static int sctp_setsockopt_events(struct sock
*sk
, char __user
*optval
,
1948 if (optlen
!= sizeof(struct sctp_event_subscribe
))
1950 if (copy_from_user(&sctp_sk(sk
)->subscribe
, optval
, optlen
))
1955 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
1957 * This socket option is applicable to the UDP-style socket only. When
1958 * set it will cause associations that are idle for more than the
1959 * specified number of seconds to automatically close. An association
1960 * being idle is defined an association that has NOT sent or received
1961 * user data. The special value of '0' indicates that no automatic
1962 * close of any associations should be performed. The option expects an
1963 * integer defining the number of seconds of idle time before an
1964 * association is closed.
1966 static int sctp_setsockopt_autoclose(struct sock
*sk
, char __user
*optval
,
1969 struct sctp_sock
*sp
= sctp_sk(sk
);
1971 /* Applicable to UDP-style socket only */
1972 if (sctp_style(sk
, TCP
))
1974 if (optlen
!= sizeof(int))
1976 if (copy_from_user(&sp
->autoclose
, optval
, optlen
))
1982 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
1984 * Applications can enable or disable heartbeats for any peer address of
1985 * an association, modify an address's heartbeat interval, force a
1986 * heartbeat to be sent immediately, and adjust the address's maximum
1987 * number of retransmissions sent before an address is considered
1988 * unreachable. The following structure is used to access and modify an
1989 * address's parameters:
1991 * struct sctp_paddrparams {
1992 * sctp_assoc_t spp_assoc_id;
1993 * struct sockaddr_storage spp_address;
1994 * uint32_t spp_hbinterval;
1995 * uint16_t spp_pathmaxrxt;
1996 * uint32_t spp_pathmtu;
1997 * uint32_t spp_sackdelay;
1998 * uint32_t spp_flags;
2001 * spp_assoc_id - (one-to-many style socket) This is filled in the
2002 * application, and identifies the association for
2004 * spp_address - This specifies which address is of interest.
2005 * spp_hbinterval - This contains the value of the heartbeat interval,
2006 * in milliseconds. If a value of zero
2007 * is present in this field then no changes are to
2008 * be made to this parameter.
2009 * spp_pathmaxrxt - This contains the maximum number of
2010 * retransmissions before this address shall be
2011 * considered unreachable. If a value of zero
2012 * is present in this field then no changes are to
2013 * be made to this parameter.
2014 * spp_pathmtu - When Path MTU discovery is disabled the value
2015 * specified here will be the "fixed" path mtu.
2016 * Note that if the spp_address field is empty
2017 * then all associations on this address will
2018 * have this fixed path mtu set upon them.
2020 * spp_sackdelay - When delayed sack is enabled, this value specifies
2021 * the number of milliseconds that sacks will be delayed
2022 * for. This value will apply to all addresses of an
2023 * association if the spp_address field is empty. Note
2024 * also, that if delayed sack is enabled and this
2025 * value is set to 0, no change is made to the last
2026 * recorded delayed sack timer value.
2028 * spp_flags - These flags are used to control various features
2029 * on an association. The flag field may contain
2030 * zero or more of the following options.
2032 * SPP_HB_ENABLE - Enable heartbeats on the
2033 * specified address. Note that if the address
2034 * field is empty all addresses for the association
2035 * have heartbeats enabled upon them.
2037 * SPP_HB_DISABLE - Disable heartbeats on the
2038 * speicifed address. Note that if the address
2039 * field is empty all addresses for the association
2040 * will have their heartbeats disabled. Note also
2041 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2042 * mutually exclusive, only one of these two should
2043 * be specified. Enabling both fields will have
2044 * undetermined results.
2046 * SPP_HB_DEMAND - Request a user initiated heartbeat
2047 * to be made immediately.
2049 * SPP_PMTUD_ENABLE - This field will enable PMTU
2050 * discovery upon the specified address. Note that
2051 * if the address feild is empty then all addresses
2052 * on the association are effected.
2054 * SPP_PMTUD_DISABLE - This field will disable PMTU
2055 * discovery upon the specified address. Note that
2056 * if the address feild is empty then all addresses
2057 * on the association are effected. Not also that
2058 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2059 * exclusive. Enabling both will have undetermined
2062 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2063 * on delayed sack. The time specified in spp_sackdelay
2064 * is used to specify the sack delay for this address. Note
2065 * that if spp_address is empty then all addresses will
2066 * enable delayed sack and take on the sack delay
2067 * value specified in spp_sackdelay.
2068 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2069 * off delayed sack. If the spp_address field is blank then
2070 * delayed sack is disabled for the entire association. Note
2071 * also that this field is mutually exclusive to
2072 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2075 static int sctp_apply_peer_addr_params(struct sctp_paddrparams
*params
,
2076 struct sctp_transport
*trans
,
2077 struct sctp_association
*asoc
,
2078 struct sctp_sock
*sp
,
2081 int sackdelay_change
)
2085 if (params
->spp_flags
& SPP_HB_DEMAND
&& trans
) {
2086 error
= sctp_primitive_REQUESTHEARTBEAT (trans
->asoc
, trans
);
2091 if (params
->spp_hbinterval
) {
2093 trans
->hbinterval
= msecs_to_jiffies(params
->spp_hbinterval
);
2095 asoc
->hbinterval
= msecs_to_jiffies(params
->spp_hbinterval
);
2097 sp
->hbinterval
= params
->spp_hbinterval
;
2103 trans
->param_flags
=
2104 (trans
->param_flags
& ~SPP_HB
) | hb_change
;
2107 (asoc
->param_flags
& ~SPP_HB
) | hb_change
;
2110 (sp
->param_flags
& ~SPP_HB
) | hb_change
;
2114 if (params
->spp_pathmtu
) {
2116 trans
->pathmtu
= params
->spp_pathmtu
;
2117 sctp_assoc_sync_pmtu(asoc
);
2119 asoc
->pathmtu
= params
->spp_pathmtu
;
2120 sctp_frag_point(sp
, params
->spp_pathmtu
);
2122 sp
->pathmtu
= params
->spp_pathmtu
;
2128 int update
= (trans
->param_flags
& SPP_PMTUD_DISABLE
) &&
2129 (params
->spp_flags
& SPP_PMTUD_ENABLE
);
2130 trans
->param_flags
=
2131 (trans
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2133 sctp_transport_pmtu(trans
);
2134 sctp_assoc_sync_pmtu(asoc
);
2138 (asoc
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2141 (sp
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2145 if (params
->spp_sackdelay
) {
2148 msecs_to_jiffies(params
->spp_sackdelay
);
2151 msecs_to_jiffies(params
->spp_sackdelay
);
2153 sp
->sackdelay
= params
->spp_sackdelay
;
2157 if (sackdelay_change
) {
2159 trans
->param_flags
=
2160 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2164 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2168 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2173 if (params
->spp_pathmaxrxt
) {
2175 trans
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2177 asoc
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2179 sp
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2186 static int sctp_setsockopt_peer_addr_params(struct sock
*sk
,
2187 char __user
*optval
, int optlen
)
2189 struct sctp_paddrparams params
;
2190 struct sctp_transport
*trans
= NULL
;
2191 struct sctp_association
*asoc
= NULL
;
2192 struct sctp_sock
*sp
= sctp_sk(sk
);
2194 int hb_change
, pmtud_change
, sackdelay_change
;
2196 if (optlen
!= sizeof(struct sctp_paddrparams
))
2199 if (copy_from_user(¶ms
, optval
, optlen
))
2202 /* Validate flags and value parameters. */
2203 hb_change
= params
.spp_flags
& SPP_HB
;
2204 pmtud_change
= params
.spp_flags
& SPP_PMTUD
;
2205 sackdelay_change
= params
.spp_flags
& SPP_SACKDELAY
;
2207 if (hb_change
== SPP_HB
||
2208 pmtud_change
== SPP_PMTUD
||
2209 sackdelay_change
== SPP_SACKDELAY
||
2210 params
.spp_sackdelay
> 500 ||
2212 && params
.spp_pathmtu
< SCTP_DEFAULT_MINSEGMENT
))
2215 /* If an address other than INADDR_ANY is specified, and
2216 * no transport is found, then the request is invalid.
2218 if (!sctp_is_any(( union sctp_addr
*)¶ms
.spp_address
)) {
2219 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
2220 params
.spp_assoc_id
);
2225 /* Get association, if assoc_id != 0 and the socket is a one
2226 * to many style socket, and an association was not found, then
2227 * the id was invalid.
2229 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
2230 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
))
2233 /* Heartbeat demand can only be sent on a transport or
2234 * association, but not a socket.
2236 if (params
.spp_flags
& SPP_HB_DEMAND
&& !trans
&& !asoc
)
2239 /* Process parameters. */
2240 error
= sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2241 hb_change
, pmtud_change
,
2247 /* If changes are for association, also apply parameters to each
2250 if (!trans
&& asoc
) {
2251 struct list_head
*pos
;
2253 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
2254 trans
= list_entry(pos
, struct sctp_transport
,
2256 sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2257 hb_change
, pmtud_change
,
2265 /* 7.1.24. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
2267 * This options will get or set the delayed ack timer. The time is set
2268 * in milliseconds. If the assoc_id is 0, then this sets or gets the
2269 * endpoints default delayed ack timer value. If the assoc_id field is
2270 * non-zero, then the set or get effects the specified association.
2272 * struct sctp_assoc_value {
2273 * sctp_assoc_t assoc_id;
2274 * uint32_t assoc_value;
2277 * assoc_id - This parameter, indicates which association the
2278 * user is preforming an action upon. Note that if
2279 * this field's value is zero then the endpoints
2280 * default value is changed (effecting future
2281 * associations only).
2283 * assoc_value - This parameter contains the number of milliseconds
2284 * that the user is requesting the delayed ACK timer
2285 * be set to. Note that this value is defined in
2286 * the standard to be between 200 and 500 milliseconds.
2288 * Note: a value of zero will leave the value alone,
2289 * but disable SACK delay. A non-zero value will also
2290 * enable SACK delay.
2293 static int sctp_setsockopt_delayed_ack_time(struct sock
*sk
,
2294 char __user
*optval
, int optlen
)
2296 struct sctp_assoc_value params
;
2297 struct sctp_transport
*trans
= NULL
;
2298 struct sctp_association
*asoc
= NULL
;
2299 struct sctp_sock
*sp
= sctp_sk(sk
);
2301 if (optlen
!= sizeof(struct sctp_assoc_value
))
2304 if (copy_from_user(¶ms
, optval
, optlen
))
2307 /* Validate value parameter. */
2308 if (params
.assoc_value
> 500)
2311 /* Get association, if assoc_id != 0 and the socket is a one
2312 * to many style socket, and an association was not found, then
2313 * the id was invalid.
2315 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
2316 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
2319 if (params
.assoc_value
) {
2322 msecs_to_jiffies(params
.assoc_value
);
2324 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2325 SPP_SACKDELAY_ENABLE
;
2327 sp
->sackdelay
= params
.assoc_value
;
2329 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2330 SPP_SACKDELAY_ENABLE
;
2335 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2336 SPP_SACKDELAY_DISABLE
;
2339 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2340 SPP_SACKDELAY_DISABLE
;
2344 /* If change is for association, also apply to each transport. */
2346 struct list_head
*pos
;
2348 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
2349 trans
= list_entry(pos
, struct sctp_transport
,
2351 if (params
.assoc_value
) {
2353 msecs_to_jiffies(params
.assoc_value
);
2354 trans
->param_flags
=
2355 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2356 SPP_SACKDELAY_ENABLE
;
2358 trans
->param_flags
=
2359 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2360 SPP_SACKDELAY_DISABLE
;
2368 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2370 * Applications can specify protocol parameters for the default association
2371 * initialization. The option name argument to setsockopt() and getsockopt()
2374 * Setting initialization parameters is effective only on an unconnected
2375 * socket (for UDP-style sockets only future associations are effected
2376 * by the change). With TCP-style sockets, this option is inherited by
2377 * sockets derived from a listener socket.
2379 static int sctp_setsockopt_initmsg(struct sock
*sk
, char __user
*optval
, int optlen
)
2381 struct sctp_initmsg sinit
;
2382 struct sctp_sock
*sp
= sctp_sk(sk
);
2384 if (optlen
!= sizeof(struct sctp_initmsg
))
2386 if (copy_from_user(&sinit
, optval
, optlen
))
2389 if (sinit
.sinit_num_ostreams
)
2390 sp
->initmsg
.sinit_num_ostreams
= sinit
.sinit_num_ostreams
;
2391 if (sinit
.sinit_max_instreams
)
2392 sp
->initmsg
.sinit_max_instreams
= sinit
.sinit_max_instreams
;
2393 if (sinit
.sinit_max_attempts
)
2394 sp
->initmsg
.sinit_max_attempts
= sinit
.sinit_max_attempts
;
2395 if (sinit
.sinit_max_init_timeo
)
2396 sp
->initmsg
.sinit_max_init_timeo
= sinit
.sinit_max_init_timeo
;
2402 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2404 * Applications that wish to use the sendto() system call may wish to
2405 * specify a default set of parameters that would normally be supplied
2406 * through the inclusion of ancillary data. This socket option allows
2407 * such an application to set the default sctp_sndrcvinfo structure.
2408 * The application that wishes to use this socket option simply passes
2409 * in to this call the sctp_sndrcvinfo structure defined in Section
2410 * 5.2.2) The input parameters accepted by this call include
2411 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2412 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2413 * to this call if the caller is using the UDP model.
2415 static int sctp_setsockopt_default_send_param(struct sock
*sk
,
2416 char __user
*optval
, int optlen
)
2418 struct sctp_sndrcvinfo info
;
2419 struct sctp_association
*asoc
;
2420 struct sctp_sock
*sp
= sctp_sk(sk
);
2422 if (optlen
!= sizeof(struct sctp_sndrcvinfo
))
2424 if (copy_from_user(&info
, optval
, optlen
))
2427 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
2428 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
2432 asoc
->default_stream
= info
.sinfo_stream
;
2433 asoc
->default_flags
= info
.sinfo_flags
;
2434 asoc
->default_ppid
= info
.sinfo_ppid
;
2435 asoc
->default_context
= info
.sinfo_context
;
2436 asoc
->default_timetolive
= info
.sinfo_timetolive
;
2438 sp
->default_stream
= info
.sinfo_stream
;
2439 sp
->default_flags
= info
.sinfo_flags
;
2440 sp
->default_ppid
= info
.sinfo_ppid
;
2441 sp
->default_context
= info
.sinfo_context
;
2442 sp
->default_timetolive
= info
.sinfo_timetolive
;
2448 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2450 * Requests that the local SCTP stack use the enclosed peer address as
2451 * the association primary. The enclosed address must be one of the
2452 * association peer's addresses.
2454 static int sctp_setsockopt_primary_addr(struct sock
*sk
, char __user
*optval
,
2457 struct sctp_prim prim
;
2458 struct sctp_transport
*trans
;
2460 if (optlen
!= sizeof(struct sctp_prim
))
2463 if (copy_from_user(&prim
, optval
, sizeof(struct sctp_prim
)))
2466 trans
= sctp_addr_id2transport(sk
, &prim
.ssp_addr
, prim
.ssp_assoc_id
);
2470 sctp_assoc_set_primary(trans
->asoc
, trans
);
2476 * 7.1.5 SCTP_NODELAY
2478 * Turn on/off any Nagle-like algorithm. This means that packets are
2479 * generally sent as soon as possible and no unnecessary delays are
2480 * introduced, at the cost of more packets in the network. Expects an
2481 * integer boolean flag.
2483 static int sctp_setsockopt_nodelay(struct sock
*sk
, char __user
*optval
,
2488 if (optlen
< sizeof(int))
2490 if (get_user(val
, (int __user
*)optval
))
2493 sctp_sk(sk
)->nodelay
= (val
== 0) ? 0 : 1;
2499 * 7.1.1 SCTP_RTOINFO
2501 * The protocol parameters used to initialize and bound retransmission
2502 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2503 * and modify these parameters.
2504 * All parameters are time values, in milliseconds. A value of 0, when
2505 * modifying the parameters, indicates that the current value should not
2509 static int sctp_setsockopt_rtoinfo(struct sock
*sk
, char __user
*optval
, int optlen
) {
2510 struct sctp_rtoinfo rtoinfo
;
2511 struct sctp_association
*asoc
;
2513 if (optlen
!= sizeof (struct sctp_rtoinfo
))
2516 if (copy_from_user(&rtoinfo
, optval
, optlen
))
2519 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
2521 /* Set the values to the specific association */
2522 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
2526 if (rtoinfo
.srto_initial
!= 0)
2528 msecs_to_jiffies(rtoinfo
.srto_initial
);
2529 if (rtoinfo
.srto_max
!= 0)
2530 asoc
->rto_max
= msecs_to_jiffies(rtoinfo
.srto_max
);
2531 if (rtoinfo
.srto_min
!= 0)
2532 asoc
->rto_min
= msecs_to_jiffies(rtoinfo
.srto_min
);
2534 /* If there is no association or the association-id = 0
2535 * set the values to the endpoint.
2537 struct sctp_sock
*sp
= sctp_sk(sk
);
2539 if (rtoinfo
.srto_initial
!= 0)
2540 sp
->rtoinfo
.srto_initial
= rtoinfo
.srto_initial
;
2541 if (rtoinfo
.srto_max
!= 0)
2542 sp
->rtoinfo
.srto_max
= rtoinfo
.srto_max
;
2543 if (rtoinfo
.srto_min
!= 0)
2544 sp
->rtoinfo
.srto_min
= rtoinfo
.srto_min
;
2552 * 7.1.2 SCTP_ASSOCINFO
2554 * This option is used to tune the the maximum retransmission attempts
2555 * of the association.
2556 * Returns an error if the new association retransmission value is
2557 * greater than the sum of the retransmission value of the peer.
2558 * See [SCTP] for more information.
2561 static int sctp_setsockopt_associnfo(struct sock
*sk
, char __user
*optval
, int optlen
)
2564 struct sctp_assocparams assocparams
;
2565 struct sctp_association
*asoc
;
2567 if (optlen
!= sizeof(struct sctp_assocparams
))
2569 if (copy_from_user(&assocparams
, optval
, optlen
))
2572 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
2574 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
2577 /* Set the values to the specific association */
2579 if (assocparams
.sasoc_asocmaxrxt
!= 0) {
2582 struct list_head
*pos
;
2583 struct sctp_transport
*peer_addr
;
2585 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
2586 peer_addr
= list_entry(pos
,
2587 struct sctp_transport
,
2589 path_sum
+= peer_addr
->pathmaxrxt
;
2593 /* Only validate asocmaxrxt if we have more then
2594 * one path/transport. We do this because path
2595 * retransmissions are only counted when we have more
2599 assocparams
.sasoc_asocmaxrxt
> path_sum
)
2602 asoc
->max_retrans
= assocparams
.sasoc_asocmaxrxt
;
2605 if (assocparams
.sasoc_cookie_life
!= 0) {
2606 asoc
->cookie_life
.tv_sec
=
2607 assocparams
.sasoc_cookie_life
/ 1000;
2608 asoc
->cookie_life
.tv_usec
=
2609 (assocparams
.sasoc_cookie_life
% 1000)
2613 /* Set the values to the endpoint */
2614 struct sctp_sock
*sp
= sctp_sk(sk
);
2616 if (assocparams
.sasoc_asocmaxrxt
!= 0)
2617 sp
->assocparams
.sasoc_asocmaxrxt
=
2618 assocparams
.sasoc_asocmaxrxt
;
2619 if (assocparams
.sasoc_cookie_life
!= 0)
2620 sp
->assocparams
.sasoc_cookie_life
=
2621 assocparams
.sasoc_cookie_life
;
2627 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2629 * This socket option is a boolean flag which turns on or off mapped V4
2630 * addresses. If this option is turned on and the socket is type
2631 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2632 * If this option is turned off, then no mapping will be done of V4
2633 * addresses and a user will receive both PF_INET6 and PF_INET type
2634 * addresses on the socket.
2636 static int sctp_setsockopt_mappedv4(struct sock
*sk
, char __user
*optval
, int optlen
)
2639 struct sctp_sock
*sp
= sctp_sk(sk
);
2641 if (optlen
< sizeof(int))
2643 if (get_user(val
, (int __user
*)optval
))
2654 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
2656 * This socket option specifies the maximum size to put in any outgoing
2657 * SCTP chunk. If a message is larger than this size it will be
2658 * fragmented by SCTP into the specified size. Note that the underlying
2659 * SCTP implementation may fragment into smaller sized chunks when the
2660 * PMTU of the underlying association is smaller than the value set by
2663 static int sctp_setsockopt_maxseg(struct sock
*sk
, char __user
*optval
, int optlen
)
2665 struct sctp_association
*asoc
;
2666 struct list_head
*pos
;
2667 struct sctp_sock
*sp
= sctp_sk(sk
);
2670 if (optlen
< sizeof(int))
2672 if (get_user(val
, (int __user
*)optval
))
2674 if ((val
!= 0) && ((val
< 8) || (val
> SCTP_MAX_CHUNK_LEN
)))
2676 sp
->user_frag
= val
;
2678 /* Update the frag_point of the existing associations. */
2679 list_for_each(pos
, &(sp
->ep
->asocs
)) {
2680 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
2681 asoc
->frag_point
= sctp_frag_point(sp
, asoc
->pathmtu
);
2689 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2691 * Requests that the peer mark the enclosed address as the association
2692 * primary. The enclosed address must be one of the association's
2693 * locally bound addresses. The following structure is used to make a
2694 * set primary request:
2696 static int sctp_setsockopt_peer_primary_addr(struct sock
*sk
, char __user
*optval
,
2699 struct sctp_sock
*sp
;
2700 struct sctp_endpoint
*ep
;
2701 struct sctp_association
*asoc
= NULL
;
2702 struct sctp_setpeerprim prim
;
2703 struct sctp_chunk
*chunk
;
2709 if (!sctp_addip_enable
)
2712 if (optlen
!= sizeof(struct sctp_setpeerprim
))
2715 if (copy_from_user(&prim
, optval
, optlen
))
2718 asoc
= sctp_id2assoc(sk
, prim
.sspp_assoc_id
);
2722 if (!asoc
->peer
.asconf_capable
)
2725 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_SET_PRIMARY
)
2728 if (!sctp_state(asoc
, ESTABLISHED
))
2731 if (!sctp_assoc_lookup_laddr(asoc
, (union sctp_addr
*)&prim
.sspp_addr
))
2732 return -EADDRNOTAVAIL
;
2734 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2735 chunk
= sctp_make_asconf_set_prim(asoc
,
2736 (union sctp_addr
*)&prim
.sspp_addr
);
2740 err
= sctp_send_asconf(asoc
, chunk
);
2742 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2747 static int sctp_setsockopt_adaption_layer(struct sock
*sk
, char __user
*optval
,
2750 struct sctp_setadaption adaption
;
2752 if (optlen
!= sizeof(struct sctp_setadaption
))
2754 if (copy_from_user(&adaption
, optval
, optlen
))
2757 sctp_sk(sk
)->adaption_ind
= adaption
.ssb_adaption_ind
;
2762 /* API 6.2 setsockopt(), getsockopt()
2764 * Applications use setsockopt() and getsockopt() to set or retrieve
2765 * socket options. Socket options are used to change the default
2766 * behavior of sockets calls. They are described in Section 7.
2770 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
2771 * int __user *optlen);
2772 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
2775 * sd - the socket descript.
2776 * level - set to IPPROTO_SCTP for all SCTP options.
2777 * optname - the option name.
2778 * optval - the buffer to store the value of the option.
2779 * optlen - the size of the buffer.
2781 SCTP_STATIC
int sctp_setsockopt(struct sock
*sk
, int level
, int optname
,
2782 char __user
*optval
, int optlen
)
2786 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
2789 /* I can hardly begin to describe how wrong this is. This is
2790 * so broken as to be worse than useless. The API draft
2791 * REALLY is NOT helpful here... I am not convinced that the
2792 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
2793 * are at all well-founded.
2795 if (level
!= SOL_SCTP
) {
2796 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
2797 retval
= af
->setsockopt(sk
, level
, optname
, optval
, optlen
);
2804 case SCTP_SOCKOPT_BINDX_ADD
:
2805 /* 'optlen' is the size of the addresses buffer. */
2806 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
2807 optlen
, SCTP_BINDX_ADD_ADDR
);
2810 case SCTP_SOCKOPT_BINDX_REM
:
2811 /* 'optlen' is the size of the addresses buffer. */
2812 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
2813 optlen
, SCTP_BINDX_REM_ADDR
);
2816 case SCTP_SOCKOPT_CONNECTX
:
2817 /* 'optlen' is the size of the addresses buffer. */
2818 retval
= sctp_setsockopt_connectx(sk
, (struct sockaddr __user
*)optval
,
2822 case SCTP_DISABLE_FRAGMENTS
:
2823 retval
= sctp_setsockopt_disable_fragments(sk
, optval
, optlen
);
2827 retval
= sctp_setsockopt_events(sk
, optval
, optlen
);
2830 case SCTP_AUTOCLOSE
:
2831 retval
= sctp_setsockopt_autoclose(sk
, optval
, optlen
);
2834 case SCTP_PEER_ADDR_PARAMS
:
2835 retval
= sctp_setsockopt_peer_addr_params(sk
, optval
, optlen
);
2838 case SCTP_DELAYED_ACK_TIME
:
2839 retval
= sctp_setsockopt_delayed_ack_time(sk
, optval
, optlen
);
2843 retval
= sctp_setsockopt_initmsg(sk
, optval
, optlen
);
2845 case SCTP_DEFAULT_SEND_PARAM
:
2846 retval
= sctp_setsockopt_default_send_param(sk
, optval
,
2849 case SCTP_PRIMARY_ADDR
:
2850 retval
= sctp_setsockopt_primary_addr(sk
, optval
, optlen
);
2852 case SCTP_SET_PEER_PRIMARY_ADDR
:
2853 retval
= sctp_setsockopt_peer_primary_addr(sk
, optval
, optlen
);
2856 retval
= sctp_setsockopt_nodelay(sk
, optval
, optlen
);
2859 retval
= sctp_setsockopt_rtoinfo(sk
, optval
, optlen
);
2861 case SCTP_ASSOCINFO
:
2862 retval
= sctp_setsockopt_associnfo(sk
, optval
, optlen
);
2864 case SCTP_I_WANT_MAPPED_V4_ADDR
:
2865 retval
= sctp_setsockopt_mappedv4(sk
, optval
, optlen
);
2868 retval
= sctp_setsockopt_maxseg(sk
, optval
, optlen
);
2870 case SCTP_ADAPTION_LAYER
:
2871 retval
= sctp_setsockopt_adaption_layer(sk
, optval
, optlen
);
2875 retval
= -ENOPROTOOPT
;
2879 sctp_release_sock(sk
);
2885 /* API 3.1.6 connect() - UDP Style Syntax
2887 * An application may use the connect() call in the UDP model to initiate an
2888 * association without sending data.
2892 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
2894 * sd: the socket descriptor to have a new association added to.
2896 * nam: the address structure (either struct sockaddr_in or struct
2897 * sockaddr_in6 defined in RFC2553 [7]).
2899 * len: the size of the address.
2901 SCTP_STATIC
int sctp_connect(struct sock
*sk
, struct sockaddr
*addr
,
2909 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
2910 __FUNCTION__
, sk
, addr
, addr_len
);
2912 /* Validate addr_len before calling common connect/connectx routine. */
2913 af
= sctp_get_af_specific(addr
->sa_family
);
2914 if (!af
|| addr_len
< af
->sockaddr_len
) {
2917 /* Pass correct addr len to common routine (so it knows there
2918 * is only one address being passed.
2920 err
= __sctp_connect(sk
, addr
, af
->sockaddr_len
);
2923 sctp_release_sock(sk
);
2927 /* FIXME: Write comments. */
2928 SCTP_STATIC
int sctp_disconnect(struct sock
*sk
, int flags
)
2930 return -EOPNOTSUPP
; /* STUB */
2933 /* 4.1.4 accept() - TCP Style Syntax
2935 * Applications use accept() call to remove an established SCTP
2936 * association from the accept queue of the endpoint. A new socket
2937 * descriptor will be returned from accept() to represent the newly
2938 * formed association.
2940 SCTP_STATIC
struct sock
*sctp_accept(struct sock
*sk
, int flags
, int *err
)
2942 struct sctp_sock
*sp
;
2943 struct sctp_endpoint
*ep
;
2944 struct sock
*newsk
= NULL
;
2945 struct sctp_association
*asoc
;
2954 if (!sctp_style(sk
, TCP
)) {
2955 error
= -EOPNOTSUPP
;
2959 if (!sctp_sstate(sk
, LISTENING
)) {
2964 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
2966 error
= sctp_wait_for_accept(sk
, timeo
);
2970 /* We treat the list of associations on the endpoint as the accept
2971 * queue and pick the first association on the list.
2973 asoc
= list_entry(ep
->asocs
.next
, struct sctp_association
, asocs
);
2975 newsk
= sp
->pf
->create_accept_sk(sk
, asoc
);
2981 /* Populate the fields of the newsk from the oldsk and migrate the
2982 * asoc to the newsk.
2984 sctp_sock_migrate(sk
, newsk
, asoc
, SCTP_SOCKET_TCP
);
2987 sctp_release_sock(sk
);
2992 /* The SCTP ioctl handler. */
2993 SCTP_STATIC
int sctp_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
2995 return -ENOIOCTLCMD
;
2998 /* This is the function which gets called during socket creation to
2999 * initialized the SCTP-specific portion of the sock.
3000 * The sock structure should already be zero-filled memory.
3002 SCTP_STATIC
int sctp_init_sock(struct sock
*sk
)
3004 struct sctp_endpoint
*ep
;
3005 struct sctp_sock
*sp
;
3007 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk
);
3011 /* Initialize the SCTP per socket area. */
3012 switch (sk
->sk_type
) {
3013 case SOCK_SEQPACKET
:
3014 sp
->type
= SCTP_SOCKET_UDP
;
3017 sp
->type
= SCTP_SOCKET_TCP
;
3020 return -ESOCKTNOSUPPORT
;
3023 /* Initialize default send parameters. These parameters can be
3024 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3026 sp
->default_stream
= 0;
3027 sp
->default_ppid
= 0;
3028 sp
->default_flags
= 0;
3029 sp
->default_context
= 0;
3030 sp
->default_timetolive
= 0;
3032 /* Initialize default setup parameters. These parameters
3033 * can be modified with the SCTP_INITMSG socket option or
3034 * overridden by the SCTP_INIT CMSG.
3036 sp
->initmsg
.sinit_num_ostreams
= sctp_max_outstreams
;
3037 sp
->initmsg
.sinit_max_instreams
= sctp_max_instreams
;
3038 sp
->initmsg
.sinit_max_attempts
= sctp_max_retrans_init
;
3039 sp
->initmsg
.sinit_max_init_timeo
= sctp_rto_max
;
3041 /* Initialize default RTO related parameters. These parameters can
3042 * be modified for with the SCTP_RTOINFO socket option.
3044 sp
->rtoinfo
.srto_initial
= sctp_rto_initial
;
3045 sp
->rtoinfo
.srto_max
= sctp_rto_max
;
3046 sp
->rtoinfo
.srto_min
= sctp_rto_min
;
3048 /* Initialize default association related parameters. These parameters
3049 * can be modified with the SCTP_ASSOCINFO socket option.
3051 sp
->assocparams
.sasoc_asocmaxrxt
= sctp_max_retrans_association
;
3052 sp
->assocparams
.sasoc_number_peer_destinations
= 0;
3053 sp
->assocparams
.sasoc_peer_rwnd
= 0;
3054 sp
->assocparams
.sasoc_local_rwnd
= 0;
3055 sp
->assocparams
.sasoc_cookie_life
= sctp_valid_cookie_life
;
3057 /* Initialize default event subscriptions. By default, all the
3060 memset(&sp
->subscribe
, 0, sizeof(struct sctp_event_subscribe
));
3062 /* Default Peer Address Parameters. These defaults can
3063 * be modified via SCTP_PEER_ADDR_PARAMS
3065 sp
->hbinterval
= sctp_hb_interval
;
3066 sp
->pathmaxrxt
= sctp_max_retrans_path
;
3067 sp
->pathmtu
= 0; // allow default discovery
3068 sp
->sackdelay
= sctp_sack_timeout
;
3069 sp
->param_flags
= SPP_HB_ENABLE
|
3071 SPP_SACKDELAY_ENABLE
;
3073 /* If enabled no SCTP message fragmentation will be performed.
3074 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3076 sp
->disable_fragments
= 0;
3078 /* Enable Nagle algorithm by default. */
3081 /* Enable by default. */
3084 /* Auto-close idle associations after the configured
3085 * number of seconds. A value of 0 disables this
3086 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3087 * for UDP-style sockets only.
3091 /* User specified fragmentation limit. */
3094 sp
->adaption_ind
= 0;
3096 sp
->pf
= sctp_get_pf_specific(sk
->sk_family
);
3098 /* Control variables for partial data delivery. */
3100 skb_queue_head_init(&sp
->pd_lobby
);
3102 /* Create a per socket endpoint structure. Even if we
3103 * change the data structure relationships, this may still
3104 * be useful for storing pre-connect address information.
3106 ep
= sctp_endpoint_new(sk
, GFP_KERNEL
);
3113 SCTP_DBG_OBJCNT_INC(sock
);
3117 /* Cleanup any SCTP per socket resources. */
3118 SCTP_STATIC
int sctp_destroy_sock(struct sock
*sk
)
3120 struct sctp_endpoint
*ep
;
3122 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk
);
3124 /* Release our hold on the endpoint. */
3125 ep
= sctp_sk(sk
)->ep
;
3126 sctp_endpoint_free(ep
);
3131 /* API 4.1.7 shutdown() - TCP Style Syntax
3132 * int shutdown(int socket, int how);
3134 * sd - the socket descriptor of the association to be closed.
3135 * how - Specifies the type of shutdown. The values are
3138 * Disables further receive operations. No SCTP
3139 * protocol action is taken.
3141 * Disables further send operations, and initiates
3142 * the SCTP shutdown sequence.
3144 * Disables further send and receive operations
3145 * and initiates the SCTP shutdown sequence.
3147 SCTP_STATIC
void sctp_shutdown(struct sock
*sk
, int how
)
3149 struct sctp_endpoint
*ep
;
3150 struct sctp_association
*asoc
;
3152 if (!sctp_style(sk
, TCP
))
3155 if (how
& SEND_SHUTDOWN
) {
3156 ep
= sctp_sk(sk
)->ep
;
3157 if (!list_empty(&ep
->asocs
)) {
3158 asoc
= list_entry(ep
->asocs
.next
,
3159 struct sctp_association
, asocs
);
3160 sctp_primitive_SHUTDOWN(asoc
, NULL
);
3165 /* 7.2.1 Association Status (SCTP_STATUS)
3167 * Applications can retrieve current status information about an
3168 * association, including association state, peer receiver window size,
3169 * number of unacked data chunks, and number of data chunks pending
3170 * receipt. This information is read-only.
3172 static int sctp_getsockopt_sctp_status(struct sock
*sk
, int len
,
3173 char __user
*optval
,
3176 struct sctp_status status
;
3177 struct sctp_association
*asoc
= NULL
;
3178 struct sctp_transport
*transport
;
3179 sctp_assoc_t associd
;
3182 if (len
!= sizeof(status
)) {
3187 if (copy_from_user(&status
, optval
, sizeof(status
))) {
3192 associd
= status
.sstat_assoc_id
;
3193 asoc
= sctp_id2assoc(sk
, associd
);
3199 transport
= asoc
->peer
.primary_path
;
3201 status
.sstat_assoc_id
= sctp_assoc2id(asoc
);
3202 status
.sstat_state
= asoc
->state
;
3203 status
.sstat_rwnd
= asoc
->peer
.rwnd
;
3204 status
.sstat_unackdata
= asoc
->unack_data
;
3206 status
.sstat_penddata
= sctp_tsnmap_pending(&asoc
->peer
.tsn_map
);
3207 status
.sstat_instrms
= asoc
->c
.sinit_max_instreams
;
3208 status
.sstat_outstrms
= asoc
->c
.sinit_num_ostreams
;
3209 status
.sstat_fragmentation_point
= asoc
->frag_point
;
3210 status
.sstat_primary
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
3211 flip_to_n((union sctp_addr
*)&status
.sstat_primary
.spinfo_address
,
3212 &transport
->ipaddr_h
);
3213 /* Map ipv4 address into v4-mapped-on-v6 address. */
3214 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
3215 (union sctp_addr
*)&status
.sstat_primary
.spinfo_address
);
3216 status
.sstat_primary
.spinfo_state
= transport
->state
;
3217 status
.sstat_primary
.spinfo_cwnd
= transport
->cwnd
;
3218 status
.sstat_primary
.spinfo_srtt
= transport
->srtt
;
3219 status
.sstat_primary
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
3220 status
.sstat_primary
.spinfo_mtu
= transport
->pathmtu
;
3222 if (status
.sstat_primary
.spinfo_state
== SCTP_UNKNOWN
)
3223 status
.sstat_primary
.spinfo_state
= SCTP_ACTIVE
;
3225 if (put_user(len
, optlen
)) {
3230 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3231 len
, status
.sstat_state
, status
.sstat_rwnd
,
3232 status
.sstat_assoc_id
);
3234 if (copy_to_user(optval
, &status
, len
)) {
3244 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3246 * Applications can retrieve information about a specific peer address
3247 * of an association, including its reachability state, congestion
3248 * window, and retransmission timer values. This information is
3251 static int sctp_getsockopt_peer_addr_info(struct sock
*sk
, int len
,
3252 char __user
*optval
,
3255 struct sctp_paddrinfo pinfo
;
3256 struct sctp_transport
*transport
;
3259 if (len
!= sizeof(pinfo
)) {
3264 if (copy_from_user(&pinfo
, optval
, sizeof(pinfo
))) {
3269 transport
= sctp_addr_id2transport(sk
, &pinfo
.spinfo_address
,
3270 pinfo
.spinfo_assoc_id
);
3274 pinfo
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
3275 pinfo
.spinfo_state
= transport
->state
;
3276 pinfo
.spinfo_cwnd
= transport
->cwnd
;
3277 pinfo
.spinfo_srtt
= transport
->srtt
;
3278 pinfo
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
3279 pinfo
.spinfo_mtu
= transport
->pathmtu
;
3281 if (pinfo
.spinfo_state
== SCTP_UNKNOWN
)
3282 pinfo
.spinfo_state
= SCTP_ACTIVE
;
3284 if (put_user(len
, optlen
)) {
3289 if (copy_to_user(optval
, &pinfo
, len
)) {
3298 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3300 * This option is a on/off flag. If enabled no SCTP message
3301 * fragmentation will be performed. Instead if a message being sent
3302 * exceeds the current PMTU size, the message will NOT be sent and
3303 * instead a error will be indicated to the user.
3305 static int sctp_getsockopt_disable_fragments(struct sock
*sk
, int len
,
3306 char __user
*optval
, int __user
*optlen
)
3310 if (len
< sizeof(int))
3314 val
= (sctp_sk(sk
)->disable_fragments
== 1);
3315 if (put_user(len
, optlen
))
3317 if (copy_to_user(optval
, &val
, len
))
3322 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3324 * This socket option is used to specify various notifications and
3325 * ancillary data the user wishes to receive.
3327 static int sctp_getsockopt_events(struct sock
*sk
, int len
, char __user
*optval
,
3330 if (len
!= sizeof(struct sctp_event_subscribe
))
3332 if (copy_to_user(optval
, &sctp_sk(sk
)->subscribe
, len
))
3337 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3339 * This socket option is applicable to the UDP-style socket only. When
3340 * set it will cause associations that are idle for more than the
3341 * specified number of seconds to automatically close. An association
3342 * being idle is defined an association that has NOT sent or received
3343 * user data. The special value of '0' indicates that no automatic
3344 * close of any associations should be performed. The option expects an
3345 * integer defining the number of seconds of idle time before an
3346 * association is closed.
3348 static int sctp_getsockopt_autoclose(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
3350 /* Applicable to UDP-style socket only */
3351 if (sctp_style(sk
, TCP
))
3353 if (len
!= sizeof(int))
3355 if (copy_to_user(optval
, &sctp_sk(sk
)->autoclose
, len
))
3360 /* Helper routine to branch off an association to a new socket. */
3361 SCTP_STATIC
int sctp_do_peeloff(struct sctp_association
*asoc
,
3362 struct socket
**sockp
)
3364 struct sock
*sk
= asoc
->base
.sk
;
3365 struct socket
*sock
;
3366 struct inet_sock
*inetsk
;
3369 /* An association cannot be branched off from an already peeled-off
3370 * socket, nor is this supported for tcp style sockets.
3372 if (!sctp_style(sk
, UDP
))
3375 /* Create a new socket. */
3376 err
= sock_create(sk
->sk_family
, SOCK_SEQPACKET
, IPPROTO_SCTP
, &sock
);
3380 /* Populate the fields of the newsk from the oldsk and migrate the
3381 * asoc to the newsk.
3383 sctp_sock_migrate(sk
, sock
->sk
, asoc
, SCTP_SOCKET_UDP_HIGH_BANDWIDTH
);
3385 /* Make peeled-off sockets more like 1-1 accepted sockets.
3386 * Set the daddr and initialize id to something more random
3388 inetsk
= inet_sk(sock
->sk
);
3389 inetsk
->daddr
= asoc
->peer
.primary_addr
.v4
.sin_addr
.s_addr
;
3390 inetsk
->id
= asoc
->next_tsn
^ jiffies
;
3397 static int sctp_getsockopt_peeloff(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
3399 sctp_peeloff_arg_t peeloff
;
3400 struct socket
*newsock
;
3402 struct sctp_association
*asoc
;
3404 if (len
!= sizeof(sctp_peeloff_arg_t
))
3406 if (copy_from_user(&peeloff
, optval
, len
))
3409 asoc
= sctp_id2assoc(sk
, peeloff
.associd
);
3415 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __FUNCTION__
, sk
, asoc
);
3417 retval
= sctp_do_peeloff(asoc
, &newsock
);
3421 /* Map the socket to an unused fd that can be returned to the user. */
3422 retval
= sock_map_fd(newsock
);
3424 sock_release(newsock
);
3428 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3429 __FUNCTION__
, sk
, asoc
, newsock
->sk
, retval
);
3431 /* Return the fd mapped to the new socket. */
3432 peeloff
.sd
= retval
;
3433 if (copy_to_user(optval
, &peeloff
, len
))
3440 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
3442 * Applications can enable or disable heartbeats for any peer address of
3443 * an association, modify an address's heartbeat interval, force a
3444 * heartbeat to be sent immediately, and adjust the address's maximum
3445 * number of retransmissions sent before an address is considered
3446 * unreachable. The following structure is used to access and modify an
3447 * address's parameters:
3449 * struct sctp_paddrparams {
3450 * sctp_assoc_t spp_assoc_id;
3451 * struct sockaddr_storage spp_address;
3452 * uint32_t spp_hbinterval;
3453 * uint16_t spp_pathmaxrxt;
3454 * uint32_t spp_pathmtu;
3455 * uint32_t spp_sackdelay;
3456 * uint32_t spp_flags;
3459 * spp_assoc_id - (one-to-many style socket) This is filled in the
3460 * application, and identifies the association for
3462 * spp_address - This specifies which address is of interest.
3463 * spp_hbinterval - This contains the value of the heartbeat interval,
3464 * in milliseconds. If a value of zero
3465 * is present in this field then no changes are to
3466 * be made to this parameter.
3467 * spp_pathmaxrxt - This contains the maximum number of
3468 * retransmissions before this address shall be
3469 * considered unreachable. If a value of zero
3470 * is present in this field then no changes are to
3471 * be made to this parameter.
3472 * spp_pathmtu - When Path MTU discovery is disabled the value
3473 * specified here will be the "fixed" path mtu.
3474 * Note that if the spp_address field is empty
3475 * then all associations on this address will
3476 * have this fixed path mtu set upon them.
3478 * spp_sackdelay - When delayed sack is enabled, this value specifies
3479 * the number of milliseconds that sacks will be delayed
3480 * for. This value will apply to all addresses of an
3481 * association if the spp_address field is empty. Note
3482 * also, that if delayed sack is enabled and this
3483 * value is set to 0, no change is made to the last
3484 * recorded delayed sack timer value.
3486 * spp_flags - These flags are used to control various features
3487 * on an association. The flag field may contain
3488 * zero or more of the following options.
3490 * SPP_HB_ENABLE - Enable heartbeats on the
3491 * specified address. Note that if the address
3492 * field is empty all addresses for the association
3493 * have heartbeats enabled upon them.
3495 * SPP_HB_DISABLE - Disable heartbeats on the
3496 * speicifed address. Note that if the address
3497 * field is empty all addresses for the association
3498 * will have their heartbeats disabled. Note also
3499 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
3500 * mutually exclusive, only one of these two should
3501 * be specified. Enabling both fields will have
3502 * undetermined results.
3504 * SPP_HB_DEMAND - Request a user initiated heartbeat
3505 * to be made immediately.
3507 * SPP_PMTUD_ENABLE - This field will enable PMTU
3508 * discovery upon the specified address. Note that
3509 * if the address feild is empty then all addresses
3510 * on the association are effected.
3512 * SPP_PMTUD_DISABLE - This field will disable PMTU
3513 * discovery upon the specified address. Note that
3514 * if the address feild is empty then all addresses
3515 * on the association are effected. Not also that
3516 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
3517 * exclusive. Enabling both will have undetermined
3520 * SPP_SACKDELAY_ENABLE - Setting this flag turns
3521 * on delayed sack. The time specified in spp_sackdelay
3522 * is used to specify the sack delay for this address. Note
3523 * that if spp_address is empty then all addresses will
3524 * enable delayed sack and take on the sack delay
3525 * value specified in spp_sackdelay.
3526 * SPP_SACKDELAY_DISABLE - Setting this flag turns
3527 * off delayed sack. If the spp_address field is blank then
3528 * delayed sack is disabled for the entire association. Note
3529 * also that this field is mutually exclusive to
3530 * SPP_SACKDELAY_ENABLE, setting both will have undefined
3533 static int sctp_getsockopt_peer_addr_params(struct sock
*sk
, int len
,
3534 char __user
*optval
, int __user
*optlen
)
3536 struct sctp_paddrparams params
;
3537 struct sctp_transport
*trans
= NULL
;
3538 struct sctp_association
*asoc
= NULL
;
3539 struct sctp_sock
*sp
= sctp_sk(sk
);
3541 if (len
!= sizeof(struct sctp_paddrparams
))
3544 if (copy_from_user(¶ms
, optval
, len
))
3547 /* If an address other than INADDR_ANY is specified, and
3548 * no transport is found, then the request is invalid.
3550 if (!sctp_is_any(( union sctp_addr
*)¶ms
.spp_address
)) {
3551 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
3552 params
.spp_assoc_id
);
3554 SCTP_DEBUG_PRINTK("Failed no transport\n");
3559 /* Get association, if assoc_id != 0 and the socket is a one
3560 * to many style socket, and an association was not found, then
3561 * the id was invalid.
3563 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
3564 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
)) {
3565 SCTP_DEBUG_PRINTK("Failed no association\n");
3570 /* Fetch transport values. */
3571 params
.spp_hbinterval
= jiffies_to_msecs(trans
->hbinterval
);
3572 params
.spp_pathmtu
= trans
->pathmtu
;
3573 params
.spp_pathmaxrxt
= trans
->pathmaxrxt
;
3574 params
.spp_sackdelay
= jiffies_to_msecs(trans
->sackdelay
);
3576 /*draft-11 doesn't say what to return in spp_flags*/
3577 params
.spp_flags
= trans
->param_flags
;
3579 /* Fetch association values. */
3580 params
.spp_hbinterval
= jiffies_to_msecs(asoc
->hbinterval
);
3581 params
.spp_pathmtu
= asoc
->pathmtu
;
3582 params
.spp_pathmaxrxt
= asoc
->pathmaxrxt
;
3583 params
.spp_sackdelay
= jiffies_to_msecs(asoc
->sackdelay
);
3585 /*draft-11 doesn't say what to return in spp_flags*/
3586 params
.spp_flags
= asoc
->param_flags
;
3588 /* Fetch socket values. */
3589 params
.spp_hbinterval
= sp
->hbinterval
;
3590 params
.spp_pathmtu
= sp
->pathmtu
;
3591 params
.spp_sackdelay
= sp
->sackdelay
;
3592 params
.spp_pathmaxrxt
= sp
->pathmaxrxt
;
3594 /*draft-11 doesn't say what to return in spp_flags*/
3595 params
.spp_flags
= sp
->param_flags
;
3598 if (copy_to_user(optval
, ¶ms
, len
))
3601 if (put_user(len
, optlen
))
3607 /* 7.1.24. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
3609 * This options will get or set the delayed ack timer. The time is set
3610 * in milliseconds. If the assoc_id is 0, then this sets or gets the
3611 * endpoints default delayed ack timer value. If the assoc_id field is
3612 * non-zero, then the set or get effects the specified association.
3614 * struct sctp_assoc_value {
3615 * sctp_assoc_t assoc_id;
3616 * uint32_t assoc_value;
3619 * assoc_id - This parameter, indicates which association the
3620 * user is preforming an action upon. Note that if
3621 * this field's value is zero then the endpoints
3622 * default value is changed (effecting future
3623 * associations only).
3625 * assoc_value - This parameter contains the number of milliseconds
3626 * that the user is requesting the delayed ACK timer
3627 * be set to. Note that this value is defined in
3628 * the standard to be between 200 and 500 milliseconds.
3630 * Note: a value of zero will leave the value alone,
3631 * but disable SACK delay. A non-zero value will also
3632 * enable SACK delay.
3634 static int sctp_getsockopt_delayed_ack_time(struct sock
*sk
, int len
,
3635 char __user
*optval
,
3638 struct sctp_assoc_value params
;
3639 struct sctp_association
*asoc
= NULL
;
3640 struct sctp_sock
*sp
= sctp_sk(sk
);
3642 if (len
!= sizeof(struct sctp_assoc_value
))
3645 if (copy_from_user(¶ms
, optval
, len
))
3648 /* Get association, if assoc_id != 0 and the socket is a one
3649 * to many style socket, and an association was not found, then
3650 * the id was invalid.
3652 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3653 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
3657 /* Fetch association values. */
3658 if (asoc
->param_flags
& SPP_SACKDELAY_ENABLE
)
3659 params
.assoc_value
= jiffies_to_msecs(
3662 params
.assoc_value
= 0;
3664 /* Fetch socket values. */
3665 if (sp
->param_flags
& SPP_SACKDELAY_ENABLE
)
3666 params
.assoc_value
= sp
->sackdelay
;
3668 params
.assoc_value
= 0;
3671 if (copy_to_user(optval
, ¶ms
, len
))
3674 if (put_user(len
, optlen
))
3680 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
3682 * Applications can specify protocol parameters for the default association
3683 * initialization. The option name argument to setsockopt() and getsockopt()
3686 * Setting initialization parameters is effective only on an unconnected
3687 * socket (for UDP-style sockets only future associations are effected
3688 * by the change). With TCP-style sockets, this option is inherited by
3689 * sockets derived from a listener socket.
3691 static int sctp_getsockopt_initmsg(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
3693 if (len
!= sizeof(struct sctp_initmsg
))
3695 if (copy_to_user(optval
, &sctp_sk(sk
)->initmsg
, len
))
3700 static int sctp_getsockopt_peer_addrs_num_old(struct sock
*sk
, int len
,
3701 char __user
*optval
,
3705 struct sctp_association
*asoc
;
3706 struct list_head
*pos
;
3709 if (len
!= sizeof(sctp_assoc_t
))
3712 if (copy_from_user(&id
, optval
, sizeof(sctp_assoc_t
)))
3715 /* For UDP-style sockets, id specifies the association to query. */
3716 asoc
= sctp_id2assoc(sk
, id
);
3720 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
3728 * Old API for getting list of peer addresses. Does not work for 32-bit
3729 * programs running on a 64-bit kernel
3731 static int sctp_getsockopt_peer_addrs_old(struct sock
*sk
, int len
,
3732 char __user
*optval
,
3735 struct sctp_association
*asoc
;
3736 struct list_head
*pos
;
3738 struct sctp_getaddrs_old getaddrs
;
3739 struct sctp_transport
*from
;
3741 union sctp_addr temp
;
3742 struct sctp_sock
*sp
= sctp_sk(sk
);
3745 if (len
!= sizeof(struct sctp_getaddrs_old
))
3748 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs_old
)))
3751 if (getaddrs
.addr_num
<= 0) return -EINVAL
;
3753 /* For UDP-style sockets, id specifies the association to query. */
3754 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
3758 to
= (void __user
*)getaddrs
.addrs
;
3759 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
3760 from
= list_entry(pos
, struct sctp_transport
, transports
);
3761 memcpy(&temp
, &from
->ipaddr_h
, sizeof(temp
));
3762 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
3763 addrlen
= sctp_get_af_specific(sk
->sk_family
)->sockaddr_len
;
3764 temp
.v4
.sin_port
= htons(temp
.v4
.sin_port
);
3765 if (copy_to_user(to
, &temp
, addrlen
))
3769 if (cnt
>= getaddrs
.addr_num
) break;
3771 getaddrs
.addr_num
= cnt
;
3772 if (copy_to_user(optval
, &getaddrs
, sizeof(struct sctp_getaddrs_old
)))
3778 static int sctp_getsockopt_peer_addrs(struct sock
*sk
, int len
,
3779 char __user
*optval
, int __user
*optlen
)
3781 struct sctp_association
*asoc
;
3782 struct list_head
*pos
;
3784 struct sctp_getaddrs getaddrs
;
3785 struct sctp_transport
*from
;
3787 union sctp_addr temp
;
3788 struct sctp_sock
*sp
= sctp_sk(sk
);
3793 if (len
< sizeof(struct sctp_getaddrs
))
3796 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
3799 /* For UDP-style sockets, id specifies the association to query. */
3800 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
3804 to
= optval
+ offsetof(struct sctp_getaddrs
,addrs
);
3805 space_left
= len
- sizeof(struct sctp_getaddrs
) -
3806 offsetof(struct sctp_getaddrs
,addrs
);
3808 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
3809 from
= list_entry(pos
, struct sctp_transport
, transports
);
3810 memcpy(&temp
, &from
->ipaddr_h
, sizeof(temp
));
3811 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
3812 addrlen
= sctp_get_af_specific(sk
->sk_family
)->sockaddr_len
;
3813 if(space_left
< addrlen
)
3815 temp
.v4
.sin_port
= htons(temp
.v4
.sin_port
);
3816 if (copy_to_user(to
, &temp
, addrlen
))
3820 space_left
-= addrlen
;
3823 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
))
3825 bytes_copied
= ((char __user
*)to
) - optval
;
3826 if (put_user(bytes_copied
, optlen
))
3832 static int sctp_getsockopt_local_addrs_num_old(struct sock
*sk
, int len
,
3833 char __user
*optval
,
3837 struct sctp_bind_addr
*bp
;
3838 struct sctp_association
*asoc
;
3839 struct list_head
*pos
;
3840 struct sctp_sockaddr_entry
*addr
;
3841 rwlock_t
*addr_lock
;
3842 unsigned long flags
;
3845 if (len
!= sizeof(sctp_assoc_t
))
3848 if (copy_from_user(&id
, optval
, sizeof(sctp_assoc_t
)))
3852 * For UDP-style sockets, id specifies the association to query.
3853 * If the id field is set to the value '0' then the locally bound
3854 * addresses are returned without regard to any particular
3858 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
3859 addr_lock
= &sctp_sk(sk
)->ep
->base
.addr_lock
;
3861 asoc
= sctp_id2assoc(sk
, id
);
3864 bp
= &asoc
->base
.bind_addr
;
3865 addr_lock
= &asoc
->base
.addr_lock
;
3868 sctp_read_lock(addr_lock
);
3870 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
3871 * addresses from the global local address list.
3873 if (sctp_list_single_entry(&bp
->address_list
)) {
3874 addr
= list_entry(bp
->address_list
.next
,
3875 struct sctp_sockaddr_entry
, list
);
3876 if (sctp_is_any(&addr
->a_h
)) {
3877 sctp_spin_lock_irqsave(&sctp_local_addr_lock
, flags
);
3878 list_for_each(pos
, &sctp_local_addr_list
) {
3879 addr
= list_entry(pos
,
3880 struct sctp_sockaddr_entry
,
3882 if ((PF_INET
== sk
->sk_family
) &&
3883 (AF_INET6
== addr
->a_h
.sa
.sa_family
))
3887 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock
,
3895 list_for_each(pos
, &bp
->address_list
) {
3900 sctp_read_unlock(addr_lock
);
3904 /* Helper function that copies local addresses to user and returns the number
3905 * of addresses copied.
3907 static int sctp_copy_laddrs_to_user_old(struct sock
*sk
, __u16 port
, int max_addrs
,
3910 struct list_head
*pos
;
3911 struct sctp_sockaddr_entry
*addr
;
3912 unsigned long flags
;
3913 union sctp_addr temp
;
3917 sctp_spin_lock_irqsave(&sctp_local_addr_lock
, flags
);
3918 list_for_each(pos
, &sctp_local_addr_list
) {
3919 addr
= list_entry(pos
, struct sctp_sockaddr_entry
, list
);
3920 if ((PF_INET
== sk
->sk_family
) &&
3921 (AF_INET6
== addr
->a_h
.sa
.sa_family
))
3923 memcpy(&temp
, &addr
->a_h
, sizeof(temp
));
3924 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
3926 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
3927 temp
.v4
.sin_port
= htons(port
);
3928 if (copy_to_user(to
, &temp
, addrlen
)) {
3929 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock
,
3935 if (cnt
>= max_addrs
) break;
3937 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock
, flags
);
3942 static int sctp_copy_laddrs_to_user(struct sock
*sk
, __u16 port
,
3943 void __user
**to
, size_t space_left
)
3945 struct list_head
*pos
;
3946 struct sctp_sockaddr_entry
*addr
;
3947 unsigned long flags
;
3948 union sctp_addr temp
;
3952 sctp_spin_lock_irqsave(&sctp_local_addr_lock
, flags
);
3953 list_for_each(pos
, &sctp_local_addr_list
) {
3954 addr
= list_entry(pos
, struct sctp_sockaddr_entry
, list
);
3955 if ((PF_INET
== sk
->sk_family
) &&
3956 (AF_INET6
== addr
->a_h
.sa
.sa_family
))
3958 memcpy(&temp
, &addr
->a_h
, sizeof(temp
));
3959 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
3961 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
3962 if(space_left
<addrlen
)
3964 temp
.v4
.sin_port
= htons(port
);
3965 if (copy_to_user(*to
, &temp
, addrlen
)) {
3966 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock
,
3972 space_left
-= addrlen
;
3974 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock
, flags
);
3979 /* Old API for getting list of local addresses. Does not work for 32-bit
3980 * programs running on a 64-bit kernel
3982 static int sctp_getsockopt_local_addrs_old(struct sock
*sk
, int len
,
3983 char __user
*optval
, int __user
*optlen
)
3985 struct sctp_bind_addr
*bp
;
3986 struct sctp_association
*asoc
;
3987 struct list_head
*pos
;
3989 struct sctp_getaddrs_old getaddrs
;
3990 struct sctp_sockaddr_entry
*addr
;
3992 union sctp_addr temp
;
3993 struct sctp_sock
*sp
= sctp_sk(sk
);
3995 rwlock_t
*addr_lock
;
3998 if (len
!= sizeof(struct sctp_getaddrs_old
))
4001 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs_old
)))
4004 if (getaddrs
.addr_num
<= 0) return -EINVAL
;
4006 * For UDP-style sockets, id specifies the association to query.
4007 * If the id field is set to the value '0' then the locally bound
4008 * addresses are returned without regard to any particular
4011 if (0 == getaddrs
.assoc_id
) {
4012 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
4013 addr_lock
= &sctp_sk(sk
)->ep
->base
.addr_lock
;
4015 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
4018 bp
= &asoc
->base
.bind_addr
;
4019 addr_lock
= &asoc
->base
.addr_lock
;
4022 to
= getaddrs
.addrs
;
4024 sctp_read_lock(addr_lock
);
4026 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4027 * addresses from the global local address list.
4029 if (sctp_list_single_entry(&bp
->address_list
)) {
4030 addr
= list_entry(bp
->address_list
.next
,
4031 struct sctp_sockaddr_entry
, list
);
4032 if (sctp_is_any(&addr
->a_h
)) {
4033 cnt
= sctp_copy_laddrs_to_user_old(sk
, bp
->port
,
4044 list_for_each(pos
, &bp
->address_list
) {
4045 addr
= list_entry(pos
, struct sctp_sockaddr_entry
, list
);
4046 memcpy(&temp
, &addr
->a_h
, sizeof(temp
));
4047 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
4048 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4049 temp
.v4
.sin_port
= htons(temp
.v4
.sin_port
);
4050 if (copy_to_user(to
, &temp
, addrlen
)) {
4056 if (cnt
>= getaddrs
.addr_num
) break;
4060 getaddrs
.addr_num
= cnt
;
4061 if (copy_to_user(optval
, &getaddrs
, sizeof(struct sctp_getaddrs_old
)))
4065 sctp_read_unlock(addr_lock
);
4069 static int sctp_getsockopt_local_addrs(struct sock
*sk
, int len
,
4070 char __user
*optval
, int __user
*optlen
)
4072 struct sctp_bind_addr
*bp
;
4073 struct sctp_association
*asoc
;
4074 struct list_head
*pos
;
4076 struct sctp_getaddrs getaddrs
;
4077 struct sctp_sockaddr_entry
*addr
;
4079 union sctp_addr temp
;
4080 struct sctp_sock
*sp
= sctp_sk(sk
);
4082 rwlock_t
*addr_lock
;
4087 if (len
<= sizeof(struct sctp_getaddrs
))
4090 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
4094 * For UDP-style sockets, id specifies the association to query.
4095 * If the id field is set to the value '0' then the locally bound
4096 * addresses are returned without regard to any particular
4099 if (0 == getaddrs
.assoc_id
) {
4100 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
4101 addr_lock
= &sctp_sk(sk
)->ep
->base
.addr_lock
;
4103 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
4106 bp
= &asoc
->base
.bind_addr
;
4107 addr_lock
= &asoc
->base
.addr_lock
;
4110 to
= optval
+ offsetof(struct sctp_getaddrs
,addrs
);
4111 space_left
= len
- sizeof(struct sctp_getaddrs
) -
4112 offsetof(struct sctp_getaddrs
,addrs
);
4114 sctp_read_lock(addr_lock
);
4116 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4117 * addresses from the global local address list.
4119 if (sctp_list_single_entry(&bp
->address_list
)) {
4120 addr
= list_entry(bp
->address_list
.next
,
4121 struct sctp_sockaddr_entry
, list
);
4122 if (sctp_is_any(&addr
->a_h
)) {
4123 cnt
= sctp_copy_laddrs_to_user(sk
, bp
->port
,
4133 list_for_each(pos
, &bp
->address_list
) {
4134 addr
= list_entry(pos
, struct sctp_sockaddr_entry
, list
);
4135 memcpy(&temp
, &addr
->a_h
, sizeof(temp
));
4136 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
4137 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4138 if(space_left
< addrlen
)
4139 return -ENOMEM
; /*fixme: right error?*/
4140 temp
.v4
.sin_port
= htons(temp
.v4
.sin_port
);
4141 if (copy_to_user(to
, &temp
, addrlen
)) {
4147 space_left
-= addrlen
;
4151 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
))
4153 bytes_copied
= ((char __user
*)to
) - optval
;
4154 if (put_user(bytes_copied
, optlen
))
4158 sctp_read_unlock(addr_lock
);
4162 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4164 * Requests that the local SCTP stack use the enclosed peer address as
4165 * the association primary. The enclosed address must be one of the
4166 * association peer's addresses.
4168 static int sctp_getsockopt_primary_addr(struct sock
*sk
, int len
,
4169 char __user
*optval
, int __user
*optlen
)
4171 struct sctp_prim prim
;
4172 struct sctp_association
*asoc
;
4173 struct sctp_sock
*sp
= sctp_sk(sk
);
4175 if (len
!= sizeof(struct sctp_prim
))
4178 if (copy_from_user(&prim
, optval
, sizeof(struct sctp_prim
)))
4181 asoc
= sctp_id2assoc(sk
, prim
.ssp_assoc_id
);
4185 if (!asoc
->peer
.primary_path
)
4188 flip_to_n((union sctp_addr
*)&prim
.ssp_addr
,
4189 &asoc
->peer
.primary_path
->ipaddr_h
);
4191 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
,
4192 (union sctp_addr
*)&prim
.ssp_addr
);
4194 if (copy_to_user(optval
, &prim
, sizeof(struct sctp_prim
)))
4201 * 7.1.11 Set Adaption Layer Indicator (SCTP_ADAPTION_LAYER)
4203 * Requests that the local endpoint set the specified Adaption Layer
4204 * Indication parameter for all future INIT and INIT-ACK exchanges.
4206 static int sctp_getsockopt_adaption_layer(struct sock
*sk
, int len
,
4207 char __user
*optval
, int __user
*optlen
)
4209 struct sctp_setadaption adaption
;
4211 if (len
!= sizeof(struct sctp_setadaption
))
4214 adaption
.ssb_adaption_ind
= sctp_sk(sk
)->adaption_ind
;
4215 if (copy_to_user(optval
, &adaption
, len
))
4223 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4225 * Applications that wish to use the sendto() system call may wish to
4226 * specify a default set of parameters that would normally be supplied
4227 * through the inclusion of ancillary data. This socket option allows
4228 * such an application to set the default sctp_sndrcvinfo structure.
4231 * The application that wishes to use this socket option simply passes
4232 * in to this call the sctp_sndrcvinfo structure defined in Section
4233 * 5.2.2) The input parameters accepted by this call include
4234 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4235 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4236 * to this call if the caller is using the UDP model.
4238 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4240 static int sctp_getsockopt_default_send_param(struct sock
*sk
,
4241 int len
, char __user
*optval
,
4244 struct sctp_sndrcvinfo info
;
4245 struct sctp_association
*asoc
;
4246 struct sctp_sock
*sp
= sctp_sk(sk
);
4248 if (len
!= sizeof(struct sctp_sndrcvinfo
))
4250 if (copy_from_user(&info
, optval
, sizeof(struct sctp_sndrcvinfo
)))
4253 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
4254 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
4258 info
.sinfo_stream
= asoc
->default_stream
;
4259 info
.sinfo_flags
= asoc
->default_flags
;
4260 info
.sinfo_ppid
= asoc
->default_ppid
;
4261 info
.sinfo_context
= asoc
->default_context
;
4262 info
.sinfo_timetolive
= asoc
->default_timetolive
;
4264 info
.sinfo_stream
= sp
->default_stream
;
4265 info
.sinfo_flags
= sp
->default_flags
;
4266 info
.sinfo_ppid
= sp
->default_ppid
;
4267 info
.sinfo_context
= sp
->default_context
;
4268 info
.sinfo_timetolive
= sp
->default_timetolive
;
4271 if (copy_to_user(optval
, &info
, sizeof(struct sctp_sndrcvinfo
)))
4279 * 7.1.5 SCTP_NODELAY
4281 * Turn on/off any Nagle-like algorithm. This means that packets are
4282 * generally sent as soon as possible and no unnecessary delays are
4283 * introduced, at the cost of more packets in the network. Expects an
4284 * integer boolean flag.
4287 static int sctp_getsockopt_nodelay(struct sock
*sk
, int len
,
4288 char __user
*optval
, int __user
*optlen
)
4292 if (len
< sizeof(int))
4296 val
= (sctp_sk(sk
)->nodelay
== 1);
4297 if (put_user(len
, optlen
))
4299 if (copy_to_user(optval
, &val
, len
))
4306 * 7.1.1 SCTP_RTOINFO
4308 * The protocol parameters used to initialize and bound retransmission
4309 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4310 * and modify these parameters.
4311 * All parameters are time values, in milliseconds. A value of 0, when
4312 * modifying the parameters, indicates that the current value should not
4316 static int sctp_getsockopt_rtoinfo(struct sock
*sk
, int len
,
4317 char __user
*optval
,
4318 int __user
*optlen
) {
4319 struct sctp_rtoinfo rtoinfo
;
4320 struct sctp_association
*asoc
;
4322 if (len
!= sizeof (struct sctp_rtoinfo
))
4325 if (copy_from_user(&rtoinfo
, optval
, sizeof (struct sctp_rtoinfo
)))
4328 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
4330 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
4333 /* Values corresponding to the specific association. */
4335 rtoinfo
.srto_initial
= jiffies_to_msecs(asoc
->rto_initial
);
4336 rtoinfo
.srto_max
= jiffies_to_msecs(asoc
->rto_max
);
4337 rtoinfo
.srto_min
= jiffies_to_msecs(asoc
->rto_min
);
4339 /* Values corresponding to the endpoint. */
4340 struct sctp_sock
*sp
= sctp_sk(sk
);
4342 rtoinfo
.srto_initial
= sp
->rtoinfo
.srto_initial
;
4343 rtoinfo
.srto_max
= sp
->rtoinfo
.srto_max
;
4344 rtoinfo
.srto_min
= sp
->rtoinfo
.srto_min
;
4347 if (put_user(len
, optlen
))
4350 if (copy_to_user(optval
, &rtoinfo
, len
))
4358 * 7.1.2 SCTP_ASSOCINFO
4360 * This option is used to tune the the maximum retransmission attempts
4361 * of the association.
4362 * Returns an error if the new association retransmission value is
4363 * greater than the sum of the retransmission value of the peer.
4364 * See [SCTP] for more information.
4367 static int sctp_getsockopt_associnfo(struct sock
*sk
, int len
,
4368 char __user
*optval
,
4372 struct sctp_assocparams assocparams
;
4373 struct sctp_association
*asoc
;
4374 struct list_head
*pos
;
4377 if (len
!= sizeof (struct sctp_assocparams
))
4380 if (copy_from_user(&assocparams
, optval
,
4381 sizeof (struct sctp_assocparams
)))
4384 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
4386 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
4389 /* Values correspoinding to the specific association */
4391 assocparams
.sasoc_asocmaxrxt
= asoc
->max_retrans
;
4392 assocparams
.sasoc_peer_rwnd
= asoc
->peer
.rwnd
;
4393 assocparams
.sasoc_local_rwnd
= asoc
->a_rwnd
;
4394 assocparams
.sasoc_cookie_life
= (asoc
->cookie_life
.tv_sec
4396 (asoc
->cookie_life
.tv_usec
4399 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
4403 assocparams
.sasoc_number_peer_destinations
= cnt
;
4405 /* Values corresponding to the endpoint */
4406 struct sctp_sock
*sp
= sctp_sk(sk
);
4408 assocparams
.sasoc_asocmaxrxt
= sp
->assocparams
.sasoc_asocmaxrxt
;
4409 assocparams
.sasoc_peer_rwnd
= sp
->assocparams
.sasoc_peer_rwnd
;
4410 assocparams
.sasoc_local_rwnd
= sp
->assocparams
.sasoc_local_rwnd
;
4411 assocparams
.sasoc_cookie_life
=
4412 sp
->assocparams
.sasoc_cookie_life
;
4413 assocparams
.sasoc_number_peer_destinations
=
4415 sasoc_number_peer_destinations
;
4418 if (put_user(len
, optlen
))
4421 if (copy_to_user(optval
, &assocparams
, len
))
4428 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4430 * This socket option is a boolean flag which turns on or off mapped V4
4431 * addresses. If this option is turned on and the socket is type
4432 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4433 * If this option is turned off, then no mapping will be done of V4
4434 * addresses and a user will receive both PF_INET6 and PF_INET type
4435 * addresses on the socket.
4437 static int sctp_getsockopt_mappedv4(struct sock
*sk
, int len
,
4438 char __user
*optval
, int __user
*optlen
)
4441 struct sctp_sock
*sp
= sctp_sk(sk
);
4443 if (len
< sizeof(int))
4448 if (put_user(len
, optlen
))
4450 if (copy_to_user(optval
, &val
, len
))
4457 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
4459 * This socket option specifies the maximum size to put in any outgoing
4460 * SCTP chunk. If a message is larger than this size it will be
4461 * fragmented by SCTP into the specified size. Note that the underlying
4462 * SCTP implementation may fragment into smaller sized chunks when the
4463 * PMTU of the underlying association is smaller than the value set by
4466 static int sctp_getsockopt_maxseg(struct sock
*sk
, int len
,
4467 char __user
*optval
, int __user
*optlen
)
4471 if (len
< sizeof(int))
4476 val
= sctp_sk(sk
)->user_frag
;
4477 if (put_user(len
, optlen
))
4479 if (copy_to_user(optval
, &val
, len
))
4485 SCTP_STATIC
int sctp_getsockopt(struct sock
*sk
, int level
, int optname
,
4486 char __user
*optval
, int __user
*optlen
)
4491 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
4494 /* I can hardly begin to describe how wrong this is. This is
4495 * so broken as to be worse than useless. The API draft
4496 * REALLY is NOT helpful here... I am not convinced that the
4497 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
4498 * are at all well-founded.
4500 if (level
!= SOL_SCTP
) {
4501 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
4503 retval
= af
->getsockopt(sk
, level
, optname
, optval
, optlen
);
4507 if (get_user(len
, optlen
))
4514 retval
= sctp_getsockopt_sctp_status(sk
, len
, optval
, optlen
);
4516 case SCTP_DISABLE_FRAGMENTS
:
4517 retval
= sctp_getsockopt_disable_fragments(sk
, len
, optval
,
4521 retval
= sctp_getsockopt_events(sk
, len
, optval
, optlen
);
4523 case SCTP_AUTOCLOSE
:
4524 retval
= sctp_getsockopt_autoclose(sk
, len
, optval
, optlen
);
4526 case SCTP_SOCKOPT_PEELOFF
:
4527 retval
= sctp_getsockopt_peeloff(sk
, len
, optval
, optlen
);
4529 case SCTP_PEER_ADDR_PARAMS
:
4530 retval
= sctp_getsockopt_peer_addr_params(sk
, len
, optval
,
4533 case SCTP_DELAYED_ACK_TIME
:
4534 retval
= sctp_getsockopt_delayed_ack_time(sk
, len
, optval
,
4538 retval
= sctp_getsockopt_initmsg(sk
, len
, optval
, optlen
);
4540 case SCTP_GET_PEER_ADDRS_NUM_OLD
:
4541 retval
= sctp_getsockopt_peer_addrs_num_old(sk
, len
, optval
,
4544 case SCTP_GET_LOCAL_ADDRS_NUM_OLD
:
4545 retval
= sctp_getsockopt_local_addrs_num_old(sk
, len
, optval
,
4548 case SCTP_GET_PEER_ADDRS_OLD
:
4549 retval
= sctp_getsockopt_peer_addrs_old(sk
, len
, optval
,
4552 case SCTP_GET_LOCAL_ADDRS_OLD
:
4553 retval
= sctp_getsockopt_local_addrs_old(sk
, len
, optval
,
4556 case SCTP_GET_PEER_ADDRS
:
4557 retval
= sctp_getsockopt_peer_addrs(sk
, len
, optval
,
4560 case SCTP_GET_LOCAL_ADDRS
:
4561 retval
= sctp_getsockopt_local_addrs(sk
, len
, optval
,
4564 case SCTP_DEFAULT_SEND_PARAM
:
4565 retval
= sctp_getsockopt_default_send_param(sk
, len
,
4568 case SCTP_PRIMARY_ADDR
:
4569 retval
= sctp_getsockopt_primary_addr(sk
, len
, optval
, optlen
);
4572 retval
= sctp_getsockopt_nodelay(sk
, len
, optval
, optlen
);
4575 retval
= sctp_getsockopt_rtoinfo(sk
, len
, optval
, optlen
);
4577 case SCTP_ASSOCINFO
:
4578 retval
= sctp_getsockopt_associnfo(sk
, len
, optval
, optlen
);
4580 case SCTP_I_WANT_MAPPED_V4_ADDR
:
4581 retval
= sctp_getsockopt_mappedv4(sk
, len
, optval
, optlen
);
4584 retval
= sctp_getsockopt_maxseg(sk
, len
, optval
, optlen
);
4586 case SCTP_GET_PEER_ADDR_INFO
:
4587 retval
= sctp_getsockopt_peer_addr_info(sk
, len
, optval
,
4590 case SCTP_ADAPTION_LAYER
:
4591 retval
= sctp_getsockopt_adaption_layer(sk
, len
, optval
,
4595 retval
= -ENOPROTOOPT
;
4599 sctp_release_sock(sk
);
4603 static void sctp_hash(struct sock
*sk
)
4608 static void sctp_unhash(struct sock
*sk
)
4613 /* Check if port is acceptable. Possibly find first available port.
4615 * The port hash table (contained in the 'global' SCTP protocol storage
4616 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
4617 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
4618 * list (the list number is the port number hashed out, so as you
4619 * would expect from a hash function, all the ports in a given list have
4620 * such a number that hashes out to the same list number; you were
4621 * expecting that, right?); so each list has a set of ports, with a
4622 * link to the socket (struct sock) that uses it, the port number and
4623 * a fastreuse flag (FIXME: NPI ipg).
4625 static struct sctp_bind_bucket
*sctp_bucket_create(
4626 struct sctp_bind_hashbucket
*head
, unsigned short snum
);
4628 static long sctp_get_port_local(struct sock
*sk
, union sctp_addr
*addr
)
4630 struct sctp_bind_hashbucket
*head
; /* hash list */
4631 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
4632 unsigned short snum
;
4635 snum
= ntohs(addr
->v4
.sin_port
);
4637 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum
);
4638 sctp_local_bh_disable();
4641 /* Search for an available port.
4643 * 'sctp_port_rover' was the last port assigned, so
4644 * we start to search from 'sctp_port_rover +
4645 * 1'. What we do is first check if port 'rover' is
4646 * already in the hash table; if not, we use that; if
4647 * it is, we try next.
4649 int low
= sysctl_local_port_range
[0];
4650 int high
= sysctl_local_port_range
[1];
4651 int remaining
= (high
- low
) + 1;
4655 sctp_spin_lock(&sctp_port_alloc_lock
);
4656 rover
= sctp_port_rover
;
4659 if ((rover
< low
) || (rover
> high
))
4661 index
= sctp_phashfn(rover
);
4662 head
= &sctp_port_hashtable
[index
];
4663 sctp_spin_lock(&head
->lock
);
4664 for (pp
= head
->chain
; pp
; pp
= pp
->next
)
4665 if (pp
->port
== rover
)
4669 sctp_spin_unlock(&head
->lock
);
4670 } while (--remaining
> 0);
4671 sctp_port_rover
= rover
;
4672 sctp_spin_unlock(&sctp_port_alloc_lock
);
4674 /* Exhausted local port range during search? */
4679 /* OK, here is the one we will use. HEAD (the port
4680 * hash table list entry) is non-NULL and we hold it's
4685 /* We are given an specific port number; we verify
4686 * that it is not being used. If it is used, we will
4687 * exahust the search in the hash list corresponding
4688 * to the port number (snum) - we detect that with the
4689 * port iterator, pp being NULL.
4691 head
= &sctp_port_hashtable
[sctp_phashfn(snum
)];
4692 sctp_spin_lock(&head
->lock
);
4693 for (pp
= head
->chain
; pp
; pp
= pp
->next
) {
4694 if (pp
->port
== snum
)
4701 if (!hlist_empty(&pp
->owner
)) {
4702 /* We had a port hash table hit - there is an
4703 * available port (pp != NULL) and it is being
4704 * used by other socket (pp->owner not empty); that other
4705 * socket is going to be sk2.
4707 int reuse
= sk
->sk_reuse
;
4709 struct hlist_node
*node
;
4711 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
4712 if (pp
->fastreuse
&& sk
->sk_reuse
)
4715 /* Run through the list of sockets bound to the port
4716 * (pp->port) [via the pointers bind_next and
4717 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
4718 * we get the endpoint they describe and run through
4719 * the endpoint's list of IP (v4 or v6) addresses,
4720 * comparing each of the addresses with the address of
4721 * the socket sk. If we find a match, then that means
4722 * that this port/socket (sk) combination are already
4725 sk_for_each_bound(sk2
, node
, &pp
->owner
) {
4726 struct sctp_endpoint
*ep2
;
4727 ep2
= sctp_sk(sk2
)->ep
;
4729 if (reuse
&& sk2
->sk_reuse
)
4732 if (sctp_bind_addr_match(&ep2
->base
.bind_addr
, addr
,
4738 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
4741 /* If there was a hash table miss, create a new port. */
4743 if (!pp
&& !(pp
= sctp_bucket_create(head
, snum
)))
4746 /* In either case (hit or miss), make sure fastreuse is 1 only
4747 * if sk->sk_reuse is too (that is, if the caller requested
4748 * SO_REUSEADDR on this socket -sk-).
4750 if (hlist_empty(&pp
->owner
))
4751 pp
->fastreuse
= sk
->sk_reuse
? 1 : 0;
4752 else if (pp
->fastreuse
&& !sk
->sk_reuse
)
4755 /* We are set, so fill up all the data in the hash table
4756 * entry, tie the socket list information with the rest of the
4757 * sockets FIXME: Blurry, NPI (ipg).
4760 inet_sk(sk
)->num
= snum
;
4761 if (!sctp_sk(sk
)->bind_hash
) {
4762 sk_add_bind_node(sk
, &pp
->owner
);
4763 sctp_sk(sk
)->bind_hash
= pp
;
4768 sctp_spin_unlock(&head
->lock
);
4771 sctp_local_bh_enable();
4775 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
4776 * port is requested.
4778 static int sctp_get_port(struct sock
*sk
, unsigned short snum
)
4781 union sctp_addr addr
;
4782 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
4784 /* Set up a dummy address struct from the sk. */
4785 af
->from_sk(&addr
, sk
);
4786 addr
.v4
.sin_port
= htons(snum
);
4788 /* Note: sk->sk_num gets filled in if ephemeral port request. */
4789 ret
= sctp_get_port_local(sk
, &addr
);
4791 return (ret
? 1 : 0);
4795 * 3.1.3 listen() - UDP Style Syntax
4797 * By default, new associations are not accepted for UDP style sockets.
4798 * An application uses listen() to mark a socket as being able to
4799 * accept new associations.
4801 SCTP_STATIC
int sctp_seqpacket_listen(struct sock
*sk
, int backlog
)
4803 struct sctp_sock
*sp
= sctp_sk(sk
);
4804 struct sctp_endpoint
*ep
= sp
->ep
;
4806 /* Only UDP style sockets that are not peeled off are allowed to
4809 if (!sctp_style(sk
, UDP
))
4812 /* If backlog is zero, disable listening. */
4814 if (sctp_sstate(sk
, CLOSED
))
4817 sctp_unhash_endpoint(ep
);
4818 sk
->sk_state
= SCTP_SS_CLOSED
;
4821 /* Return if we are already listening. */
4822 if (sctp_sstate(sk
, LISTENING
))
4826 * If a bind() or sctp_bindx() is not called prior to a listen()
4827 * call that allows new associations to be accepted, the system
4828 * picks an ephemeral port and will choose an address set equivalent
4829 * to binding with a wildcard address.
4831 * This is not currently spelled out in the SCTP sockets
4832 * extensions draft, but follows the practice as seen in TCP
4835 if (!ep
->base
.bind_addr
.port
) {
4836 if (sctp_autobind(sk
))
4839 sk
->sk_state
= SCTP_SS_LISTENING
;
4840 sctp_hash_endpoint(ep
);
4845 * 4.1.3 listen() - TCP Style Syntax
4847 * Applications uses listen() to ready the SCTP endpoint for accepting
4848 * inbound associations.
4850 SCTP_STATIC
int sctp_stream_listen(struct sock
*sk
, int backlog
)
4852 struct sctp_sock
*sp
= sctp_sk(sk
);
4853 struct sctp_endpoint
*ep
= sp
->ep
;
4855 /* If backlog is zero, disable listening. */
4857 if (sctp_sstate(sk
, CLOSED
))
4860 sctp_unhash_endpoint(ep
);
4861 sk
->sk_state
= SCTP_SS_CLOSED
;
4864 if (sctp_sstate(sk
, LISTENING
))
4868 * If a bind() or sctp_bindx() is not called prior to a listen()
4869 * call that allows new associations to be accepted, the system
4870 * picks an ephemeral port and will choose an address set equivalent
4871 * to binding with a wildcard address.
4873 * This is not currently spelled out in the SCTP sockets
4874 * extensions draft, but follows the practice as seen in TCP
4877 if (!ep
->base
.bind_addr
.port
) {
4878 if (sctp_autobind(sk
))
4881 sk
->sk_state
= SCTP_SS_LISTENING
;
4882 sk
->sk_max_ack_backlog
= backlog
;
4883 sctp_hash_endpoint(ep
);
4888 * Move a socket to LISTENING state.
4890 int sctp_inet_listen(struct socket
*sock
, int backlog
)
4892 struct sock
*sk
= sock
->sk
;
4893 struct crypto_hash
*tfm
= NULL
;
4896 if (unlikely(backlog
< 0))
4901 if (sock
->state
!= SS_UNCONNECTED
)
4904 /* Allocate HMAC for generating cookie. */
4905 if (sctp_hmac_alg
) {
4906 tfm
= crypto_alloc_hash(sctp_hmac_alg
, 0, CRYPTO_ALG_ASYNC
);
4913 switch (sock
->type
) {
4914 case SOCK_SEQPACKET
:
4915 err
= sctp_seqpacket_listen(sk
, backlog
);
4918 err
= sctp_stream_listen(sk
, backlog
);
4926 /* Store away the transform reference. */
4927 sctp_sk(sk
)->hmac
= tfm
;
4929 sctp_release_sock(sk
);
4932 crypto_free_hash(tfm
);
4937 * This function is done by modeling the current datagram_poll() and the
4938 * tcp_poll(). Note that, based on these implementations, we don't
4939 * lock the socket in this function, even though it seems that,
4940 * ideally, locking or some other mechanisms can be used to ensure
4941 * the integrity of the counters (sndbuf and wmem_alloc) used
4942 * in this place. We assume that we don't need locks either until proven
4945 * Another thing to note is that we include the Async I/O support
4946 * here, again, by modeling the current TCP/UDP code. We don't have
4947 * a good way to test with it yet.
4949 unsigned int sctp_poll(struct file
*file
, struct socket
*sock
, poll_table
*wait
)
4951 struct sock
*sk
= sock
->sk
;
4952 struct sctp_sock
*sp
= sctp_sk(sk
);
4955 poll_wait(file
, sk
->sk_sleep
, wait
);
4957 /* A TCP-style listening socket becomes readable when the accept queue
4960 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
4961 return (!list_empty(&sp
->ep
->asocs
)) ?
4962 (POLLIN
| POLLRDNORM
) : 0;
4966 /* Is there any exceptional events? */
4967 if (sk
->sk_err
|| !skb_queue_empty(&sk
->sk_error_queue
))
4969 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
4971 if (sk
->sk_shutdown
== SHUTDOWN_MASK
)
4974 /* Is it readable? Reconsider this code with TCP-style support. */
4975 if (!skb_queue_empty(&sk
->sk_receive_queue
) ||
4976 (sk
->sk_shutdown
& RCV_SHUTDOWN
))
4977 mask
|= POLLIN
| POLLRDNORM
;
4979 /* The association is either gone or not ready. */
4980 if (!sctp_style(sk
, UDP
) && sctp_sstate(sk
, CLOSED
))
4983 /* Is it writable? */
4984 if (sctp_writeable(sk
)) {
4985 mask
|= POLLOUT
| POLLWRNORM
;
4987 set_bit(SOCK_ASYNC_NOSPACE
, &sk
->sk_socket
->flags
);
4989 * Since the socket is not locked, the buffer
4990 * might be made available after the writeable check and
4991 * before the bit is set. This could cause a lost I/O
4992 * signal. tcp_poll() has a race breaker for this race
4993 * condition. Based on their implementation, we put
4994 * in the following code to cover it as well.
4996 if (sctp_writeable(sk
))
4997 mask
|= POLLOUT
| POLLWRNORM
;
5002 /********************************************************************
5003 * 2nd Level Abstractions
5004 ********************************************************************/
5006 static struct sctp_bind_bucket
*sctp_bucket_create(
5007 struct sctp_bind_hashbucket
*head
, unsigned short snum
)
5009 struct sctp_bind_bucket
*pp
;
5011 pp
= kmem_cache_alloc(sctp_bucket_cachep
, SLAB_ATOMIC
);
5012 SCTP_DBG_OBJCNT_INC(bind_bucket
);
5016 INIT_HLIST_HEAD(&pp
->owner
);
5017 if ((pp
->next
= head
->chain
) != NULL
)
5018 pp
->next
->pprev
= &pp
->next
;
5020 pp
->pprev
= &head
->chain
;
5025 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5026 static void sctp_bucket_destroy(struct sctp_bind_bucket
*pp
)
5028 if (pp
&& hlist_empty(&pp
->owner
)) {
5030 pp
->next
->pprev
= pp
->pprev
;
5031 *(pp
->pprev
) = pp
->next
;
5032 kmem_cache_free(sctp_bucket_cachep
, pp
);
5033 SCTP_DBG_OBJCNT_DEC(bind_bucket
);
5037 /* Release this socket's reference to a local port. */
5038 static inline void __sctp_put_port(struct sock
*sk
)
5040 struct sctp_bind_hashbucket
*head
=
5041 &sctp_port_hashtable
[sctp_phashfn(inet_sk(sk
)->num
)];
5042 struct sctp_bind_bucket
*pp
;
5044 sctp_spin_lock(&head
->lock
);
5045 pp
= sctp_sk(sk
)->bind_hash
;
5046 __sk_del_bind_node(sk
);
5047 sctp_sk(sk
)->bind_hash
= NULL
;
5048 inet_sk(sk
)->num
= 0;
5049 sctp_bucket_destroy(pp
);
5050 sctp_spin_unlock(&head
->lock
);
5053 void sctp_put_port(struct sock
*sk
)
5055 sctp_local_bh_disable();
5056 __sctp_put_port(sk
);
5057 sctp_local_bh_enable();
5061 * The system picks an ephemeral port and choose an address set equivalent
5062 * to binding with a wildcard address.
5063 * One of those addresses will be the primary address for the association.
5064 * This automatically enables the multihoming capability of SCTP.
5066 static int sctp_autobind(struct sock
*sk
)
5068 union sctp_addr autoaddr
;
5070 unsigned short port
;
5072 /* Initialize a local sockaddr structure to INADDR_ANY. */
5073 af
= sctp_sk(sk
)->pf
->af
;
5075 port
= htons(inet_sk(sk
)->num
);
5076 af
->inaddr_any(&autoaddr
, port
);
5078 return sctp_do_bind(sk
, &autoaddr
, af
->sockaddr_len
);
5081 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5084 * 4.2 The cmsghdr Structure *
5086 * When ancillary data is sent or received, any number of ancillary data
5087 * objects can be specified by the msg_control and msg_controllen members of
5088 * the msghdr structure, because each object is preceded by
5089 * a cmsghdr structure defining the object's length (the cmsg_len member).
5090 * Historically Berkeley-derived implementations have passed only one object
5091 * at a time, but this API allows multiple objects to be
5092 * passed in a single call to sendmsg() or recvmsg(). The following example
5093 * shows two ancillary data objects in a control buffer.
5095 * |<--------------------------- msg_controllen -------------------------->|
5098 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5100 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5103 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5105 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5108 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5109 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5111 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5113 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5120 SCTP_STATIC
int sctp_msghdr_parse(const struct msghdr
*msg
,
5121 sctp_cmsgs_t
*cmsgs
)
5123 struct cmsghdr
*cmsg
;
5125 for (cmsg
= CMSG_FIRSTHDR(msg
);
5127 cmsg
= CMSG_NXTHDR((struct msghdr
*)msg
, cmsg
)) {
5128 if (!CMSG_OK(msg
, cmsg
))
5131 /* Should we parse this header or ignore? */
5132 if (cmsg
->cmsg_level
!= IPPROTO_SCTP
)
5135 /* Strictly check lengths following example in SCM code. */
5136 switch (cmsg
->cmsg_type
) {
5138 /* SCTP Socket API Extension
5139 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5141 * This cmsghdr structure provides information for
5142 * initializing new SCTP associations with sendmsg().
5143 * The SCTP_INITMSG socket option uses this same data
5144 * structure. This structure is not used for
5147 * cmsg_level cmsg_type cmsg_data[]
5148 * ------------ ------------ ----------------------
5149 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5151 if (cmsg
->cmsg_len
!=
5152 CMSG_LEN(sizeof(struct sctp_initmsg
)))
5154 cmsgs
->init
= (struct sctp_initmsg
*)CMSG_DATA(cmsg
);
5158 /* SCTP Socket API Extension
5159 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5161 * This cmsghdr structure specifies SCTP options for
5162 * sendmsg() and describes SCTP header information
5163 * about a received message through recvmsg().
5165 * cmsg_level cmsg_type cmsg_data[]
5166 * ------------ ------------ ----------------------
5167 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5169 if (cmsg
->cmsg_len
!=
5170 CMSG_LEN(sizeof(struct sctp_sndrcvinfo
)))
5174 (struct sctp_sndrcvinfo
*)CMSG_DATA(cmsg
);
5176 /* Minimally, validate the sinfo_flags. */
5177 if (cmsgs
->info
->sinfo_flags
&
5178 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
5179 SCTP_ABORT
| SCTP_EOF
))
5191 * Wait for a packet..
5192 * Note: This function is the same function as in core/datagram.c
5193 * with a few modifications to make lksctp work.
5195 static int sctp_wait_for_packet(struct sock
* sk
, int *err
, long *timeo_p
)
5200 prepare_to_wait_exclusive(sk
->sk_sleep
, &wait
, TASK_INTERRUPTIBLE
);
5202 /* Socket errors? */
5203 error
= sock_error(sk
);
5207 if (!skb_queue_empty(&sk
->sk_receive_queue
))
5210 /* Socket shut down? */
5211 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
5214 /* Sequenced packets can come disconnected. If so we report the
5219 /* Is there a good reason to think that we may receive some data? */
5220 if (list_empty(&sctp_sk(sk
)->ep
->asocs
) && !sctp_sstate(sk
, LISTENING
))
5223 /* Handle signals. */
5224 if (signal_pending(current
))
5227 /* Let another process have a go. Since we are going to sleep
5228 * anyway. Note: This may cause odd behaviors if the message
5229 * does not fit in the user's buffer, but this seems to be the
5230 * only way to honor MSG_DONTWAIT realistically.
5232 sctp_release_sock(sk
);
5233 *timeo_p
= schedule_timeout(*timeo_p
);
5237 finish_wait(sk
->sk_sleep
, &wait
);
5241 error
= sock_intr_errno(*timeo_p
);
5244 finish_wait(sk
->sk_sleep
, &wait
);
5249 /* Receive a datagram.
5250 * Note: This is pretty much the same routine as in core/datagram.c
5251 * with a few changes to make lksctp work.
5253 static struct sk_buff
*sctp_skb_recv_datagram(struct sock
*sk
, int flags
,
5254 int noblock
, int *err
)
5257 struct sk_buff
*skb
;
5260 timeo
= sock_rcvtimeo(sk
, noblock
);
5262 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
5263 timeo
, MAX_SCHEDULE_TIMEOUT
);
5266 /* Again only user level code calls this function,
5267 * so nothing interrupt level
5268 * will suddenly eat the receive_queue.
5270 * Look at current nfs client by the way...
5271 * However, this function was corrent in any case. 8)
5273 if (flags
& MSG_PEEK
) {
5274 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
5275 skb
= skb_peek(&sk
->sk_receive_queue
);
5277 atomic_inc(&skb
->users
);
5278 spin_unlock_bh(&sk
->sk_receive_queue
.lock
);
5280 skb
= skb_dequeue(&sk
->sk_receive_queue
);
5286 /* Caller is allowed not to check sk->sk_err before calling. */
5287 error
= sock_error(sk
);
5291 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
5294 /* User doesn't want to wait. */
5298 } while (sctp_wait_for_packet(sk
, err
, &timeo
) == 0);
5307 /* If sndbuf has changed, wake up per association sndbuf waiters. */
5308 static void __sctp_write_space(struct sctp_association
*asoc
)
5310 struct sock
*sk
= asoc
->base
.sk
;
5311 struct socket
*sock
= sk
->sk_socket
;
5313 if ((sctp_wspace(asoc
) > 0) && sock
) {
5314 if (waitqueue_active(&asoc
->wait
))
5315 wake_up_interruptible(&asoc
->wait
);
5317 if (sctp_writeable(sk
)) {
5318 if (sk
->sk_sleep
&& waitqueue_active(sk
->sk_sleep
))
5319 wake_up_interruptible(sk
->sk_sleep
);
5321 /* Note that we try to include the Async I/O support
5322 * here by modeling from the current TCP/UDP code.
5323 * We have not tested with it yet.
5325 if (sock
->fasync_list
&&
5326 !(sk
->sk_shutdown
& SEND_SHUTDOWN
))
5327 sock_wake_async(sock
, 2, POLL_OUT
);
5332 /* Do accounting for the sndbuf space.
5333 * Decrement the used sndbuf space of the corresponding association by the
5334 * data size which was just transmitted(freed).
5336 static void sctp_wfree(struct sk_buff
*skb
)
5338 struct sctp_association
*asoc
;
5339 struct sctp_chunk
*chunk
;
5342 /* Get the saved chunk pointer. */
5343 chunk
= *((struct sctp_chunk
**)(skb
->cb
));
5346 asoc
->sndbuf_used
-= SCTP_DATA_SNDSIZE(chunk
) +
5347 sizeof(struct sk_buff
) +
5348 sizeof(struct sctp_chunk
);
5350 atomic_sub(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
5353 __sctp_write_space(asoc
);
5355 sctp_association_put(asoc
);
5358 /* Do accounting for the receive space on the socket.
5359 * Accounting for the association is done in ulpevent.c
5360 * We set this as a destructor for the cloned data skbs so that
5361 * accounting is done at the correct time.
5363 void sctp_sock_rfree(struct sk_buff
*skb
)
5365 struct sock
*sk
= skb
->sk
;
5366 struct sctp_ulpevent
*event
= sctp_skb2event(skb
);
5368 atomic_sub(event
->rmem_len
, &sk
->sk_rmem_alloc
);
5372 /* Helper function to wait for space in the sndbuf. */
5373 static int sctp_wait_for_sndbuf(struct sctp_association
*asoc
, long *timeo_p
,
5376 struct sock
*sk
= asoc
->base
.sk
;
5378 long current_timeo
= *timeo_p
;
5381 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
5382 asoc
, (long)(*timeo_p
), msg_len
);
5384 /* Increment the association's refcnt. */
5385 sctp_association_hold(asoc
);
5387 /* Wait on the association specific sndbuf space. */
5389 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
5390 TASK_INTERRUPTIBLE
);
5393 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
5396 if (signal_pending(current
))
5397 goto do_interrupted
;
5398 if (msg_len
<= sctp_wspace(asoc
))
5401 /* Let another process have a go. Since we are going
5404 sctp_release_sock(sk
);
5405 current_timeo
= schedule_timeout(current_timeo
);
5406 BUG_ON(sk
!= asoc
->base
.sk
);
5409 *timeo_p
= current_timeo
;
5413 finish_wait(&asoc
->wait
, &wait
);
5415 /* Release the association's refcnt. */
5416 sctp_association_put(asoc
);
5425 err
= sock_intr_errno(*timeo_p
);
5433 /* If socket sndbuf has changed, wake up all per association waiters. */
5434 void sctp_write_space(struct sock
*sk
)
5436 struct sctp_association
*asoc
;
5437 struct list_head
*pos
;
5439 /* Wake up the tasks in each wait queue. */
5440 list_for_each(pos
, &((sctp_sk(sk
))->ep
->asocs
)) {
5441 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
5442 __sctp_write_space(asoc
);
5446 /* Is there any sndbuf space available on the socket?
5448 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
5449 * associations on the same socket. For a UDP-style socket with
5450 * multiple associations, it is possible for it to be "unwriteable"
5451 * prematurely. I assume that this is acceptable because
5452 * a premature "unwriteable" is better than an accidental "writeable" which
5453 * would cause an unwanted block under certain circumstances. For the 1-1
5454 * UDP-style sockets or TCP-style sockets, this code should work.
5457 static int sctp_writeable(struct sock
*sk
)
5461 amt
= sk
->sk_sndbuf
- atomic_read(&sk
->sk_wmem_alloc
);
5467 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
5468 * returns immediately with EINPROGRESS.
5470 static int sctp_wait_for_connect(struct sctp_association
*asoc
, long *timeo_p
)
5472 struct sock
*sk
= asoc
->base
.sk
;
5474 long current_timeo
= *timeo_p
;
5477 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __FUNCTION__
, asoc
,
5480 /* Increment the association's refcnt. */
5481 sctp_association_hold(asoc
);
5484 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
5485 TASK_INTERRUPTIBLE
);
5488 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
5490 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
5493 if (signal_pending(current
))
5494 goto do_interrupted
;
5496 if (sctp_state(asoc
, ESTABLISHED
))
5499 /* Let another process have a go. Since we are going
5502 sctp_release_sock(sk
);
5503 current_timeo
= schedule_timeout(current_timeo
);
5506 *timeo_p
= current_timeo
;
5510 finish_wait(&asoc
->wait
, &wait
);
5512 /* Release the association's refcnt. */
5513 sctp_association_put(asoc
);
5518 if (asoc
->init_err_counter
+ 1 > asoc
->max_init_attempts
)
5521 err
= -ECONNREFUSED
;
5525 err
= sock_intr_errno(*timeo_p
);
5533 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
)
5535 struct sctp_endpoint
*ep
;
5539 ep
= sctp_sk(sk
)->ep
;
5543 prepare_to_wait_exclusive(sk
->sk_sleep
, &wait
,
5544 TASK_INTERRUPTIBLE
);
5546 if (list_empty(&ep
->asocs
)) {
5547 sctp_release_sock(sk
);
5548 timeo
= schedule_timeout(timeo
);
5553 if (!sctp_sstate(sk
, LISTENING
))
5557 if (!list_empty(&ep
->asocs
))
5560 err
= sock_intr_errno(timeo
);
5561 if (signal_pending(current
))
5569 finish_wait(sk
->sk_sleep
, &wait
);
5574 void sctp_wait_for_close(struct sock
*sk
, long timeout
)
5579 prepare_to_wait(sk
->sk_sleep
, &wait
, TASK_INTERRUPTIBLE
);
5580 if (list_empty(&sctp_sk(sk
)->ep
->asocs
))
5582 sctp_release_sock(sk
);
5583 timeout
= schedule_timeout(timeout
);
5585 } while (!signal_pending(current
) && timeout
);
5587 finish_wait(sk
->sk_sleep
, &wait
);
5590 /* Populate the fields of the newsk from the oldsk and migrate the assoc
5591 * and its messages to the newsk.
5593 static void sctp_sock_migrate(struct sock
*oldsk
, struct sock
*newsk
,
5594 struct sctp_association
*assoc
,
5595 sctp_socket_type_t type
)
5597 struct sctp_sock
*oldsp
= sctp_sk(oldsk
);
5598 struct sctp_sock
*newsp
= sctp_sk(newsk
);
5599 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
5600 struct sctp_endpoint
*newep
= newsp
->ep
;
5601 struct sk_buff
*skb
, *tmp
;
5602 struct sctp_ulpevent
*event
;
5605 /* Migrate socket buffer sizes and all the socket level options to the
5608 newsk
->sk_sndbuf
= oldsk
->sk_sndbuf
;
5609 newsk
->sk_rcvbuf
= oldsk
->sk_rcvbuf
;
5610 /* Brute force copy old sctp opt. */
5611 inet_sk_copy_descendant(newsk
, oldsk
);
5613 /* Restore the ep value that was overwritten with the above structure
5619 /* Hook this new socket in to the bind_hash list. */
5620 pp
= sctp_sk(oldsk
)->bind_hash
;
5621 sk_add_bind_node(newsk
, &pp
->owner
);
5622 sctp_sk(newsk
)->bind_hash
= pp
;
5623 inet_sk(newsk
)->num
= inet_sk(oldsk
)->num
;
5625 /* Copy the bind_addr list from the original endpoint to the new
5626 * endpoint so that we can handle restarts properly
5628 if (PF_INET6
== assoc
->base
.sk
->sk_family
)
5629 flags
= SCTP_ADDR6_ALLOWED
;
5630 if (assoc
->peer
.ipv4_address
)
5631 flags
|= SCTP_ADDR4_PEERSUPP
;
5632 if (assoc
->peer
.ipv6_address
)
5633 flags
|= SCTP_ADDR6_PEERSUPP
;
5634 sctp_bind_addr_copy(&newsp
->ep
->base
.bind_addr
,
5635 &oldsp
->ep
->base
.bind_addr
,
5636 SCTP_SCOPE_GLOBAL
, GFP_KERNEL
, flags
);
5638 /* Move any messages in the old socket's receive queue that are for the
5639 * peeled off association to the new socket's receive queue.
5641 sctp_skb_for_each(skb
, &oldsk
->sk_receive_queue
, tmp
) {
5642 event
= sctp_skb2event(skb
);
5643 if (event
->asoc
== assoc
) {
5644 sctp_sock_rfree(skb
);
5645 __skb_unlink(skb
, &oldsk
->sk_receive_queue
);
5646 __skb_queue_tail(&newsk
->sk_receive_queue
, skb
);
5647 sctp_skb_set_owner_r(skb
, newsk
);
5651 /* Clean up any messages pending delivery due to partial
5652 * delivery. Three cases:
5653 * 1) No partial deliver; no work.
5654 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
5655 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
5657 skb_queue_head_init(&newsp
->pd_lobby
);
5658 sctp_sk(newsk
)->pd_mode
= assoc
->ulpq
.pd_mode
;
5660 if (sctp_sk(oldsk
)->pd_mode
) {
5661 struct sk_buff_head
*queue
;
5663 /* Decide which queue to move pd_lobby skbs to. */
5664 if (assoc
->ulpq
.pd_mode
) {
5665 queue
= &newsp
->pd_lobby
;
5667 queue
= &newsk
->sk_receive_queue
;
5669 /* Walk through the pd_lobby, looking for skbs that
5670 * need moved to the new socket.
5672 sctp_skb_for_each(skb
, &oldsp
->pd_lobby
, tmp
) {
5673 event
= sctp_skb2event(skb
);
5674 if (event
->asoc
== assoc
) {
5675 sctp_sock_rfree(skb
);
5676 __skb_unlink(skb
, &oldsp
->pd_lobby
);
5677 __skb_queue_tail(queue
, skb
);
5678 sctp_skb_set_owner_r(skb
, newsk
);
5682 /* Clear up any skbs waiting for the partial
5683 * delivery to finish.
5685 if (assoc
->ulpq
.pd_mode
)
5686 sctp_clear_pd(oldsk
);
5690 /* Set the type of socket to indicate that it is peeled off from the
5691 * original UDP-style socket or created with the accept() call on a
5692 * TCP-style socket..
5696 /* Mark the new socket "in-use" by the user so that any packets
5697 * that may arrive on the association after we've moved it are
5698 * queued to the backlog. This prevents a potential race between
5699 * backlog processing on the old socket and new-packet processing
5700 * on the new socket.
5702 sctp_lock_sock(newsk
);
5703 sctp_assoc_migrate(assoc
, newsk
);
5705 /* If the association on the newsk is already closed before accept()
5706 * is called, set RCV_SHUTDOWN flag.
5708 if (sctp_state(assoc
, CLOSED
) && sctp_style(newsk
, TCP
))
5709 newsk
->sk_shutdown
|= RCV_SHUTDOWN
;
5711 newsk
->sk_state
= SCTP_SS_ESTABLISHED
;
5712 sctp_release_sock(newsk
);
5715 /* This proto struct describes the ULP interface for SCTP. */
5716 struct proto sctp_prot
= {
5718 .owner
= THIS_MODULE
,
5719 .close
= sctp_close
,
5720 .connect
= sctp_connect
,
5721 .disconnect
= sctp_disconnect
,
5722 .accept
= sctp_accept
,
5723 .ioctl
= sctp_ioctl
,
5724 .init
= sctp_init_sock
,
5725 .destroy
= sctp_destroy_sock
,
5726 .shutdown
= sctp_shutdown
,
5727 .setsockopt
= sctp_setsockopt
,
5728 .getsockopt
= sctp_getsockopt
,
5729 .sendmsg
= sctp_sendmsg
,
5730 .recvmsg
= sctp_recvmsg
,
5732 .backlog_rcv
= sctp_backlog_rcv
,
5734 .unhash
= sctp_unhash
,
5735 .get_port
= sctp_get_port
,
5736 .obj_size
= sizeof(struct sctp_sock
),
5739 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
5740 struct proto sctpv6_prot
= {
5742 .owner
= THIS_MODULE
,
5743 .close
= sctp_close
,
5744 .connect
= sctp_connect
,
5745 .disconnect
= sctp_disconnect
,
5746 .accept
= sctp_accept
,
5747 .ioctl
= sctp_ioctl
,
5748 .init
= sctp_init_sock
,
5749 .destroy
= sctp_destroy_sock
,
5750 .shutdown
= sctp_shutdown
,
5751 .setsockopt
= sctp_setsockopt
,
5752 .getsockopt
= sctp_getsockopt
,
5753 .sendmsg
= sctp_sendmsg
,
5754 .recvmsg
= sctp_recvmsg
,
5756 .backlog_rcv
= sctp_backlog_rcv
,
5758 .unhash
= sctp_unhash
,
5759 .get_port
= sctp_get_port
,
5760 .obj_size
= sizeof(struct sctp6_sock
),
5762 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */