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/config.h>
61 #include <linux/types.h>
62 #include <linux/kernel.h>
63 #include <linux/wait.h>
64 #include <linux/time.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
)))
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
;
232 laddr
->v4
.sin_port
= ntohs(laddr
->v4
.sin_port
);
233 addr_asoc
= sctp_endpoint_lookup_assoc(sctp_sk(sk
)->ep
,
234 (union sctp_addr
*)addr
,
236 laddr
->v4
.sin_port
= htons(laddr
->v4
.sin_port
);
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
;
317 /* Common sockaddr verification. */
318 af
= sctp_sockaddr_af(sp
, addr
, len
);
320 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
325 snum
= ntohs(addr
->v4
.sin_port
);
327 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
328 ", port: %d, new port: %d, len: %d)\n",
334 /* PF specific bind() address verification. */
335 if (!sp
->pf
->bind_verify(sp
, addr
))
336 return -EADDRNOTAVAIL
;
338 /* We must either be unbound, or bind to the same port. */
339 if (bp
->port
&& (snum
!= bp
->port
)) {
340 SCTP_DEBUG_PRINTK("sctp_do_bind:"
341 " New port %d does not match existing port "
342 "%d.\n", snum
, bp
->port
);
346 if (snum
&& snum
< PROT_SOCK
&& !capable(CAP_NET_BIND_SERVICE
))
349 /* Make sure we are allowed to bind here.
350 * The function sctp_get_port_local() does duplicate address
353 if ((ret
= sctp_get_port_local(sk
, addr
))) {
354 if (ret
== (long) sk
) {
355 /* This endpoint has a conflicting address. */
362 /* Refresh ephemeral port. */
364 bp
->port
= inet_sk(sk
)->num
;
366 /* Add the address to the bind address list. */
367 sctp_local_bh_disable();
368 sctp_write_lock(&ep
->base
.addr_lock
);
370 /* Use GFP_ATOMIC since BHs are disabled. */
371 addr
->v4
.sin_port
= ntohs(addr
->v4
.sin_port
);
372 ret
= sctp_add_bind_addr(bp
, addr
, GFP_ATOMIC
);
373 addr
->v4
.sin_port
= htons(addr
->v4
.sin_port
);
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
;
496 struct list_head
*pos
;
501 if (!sctp_addip_enable
)
507 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
508 __FUNCTION__
, sk
, addrs
, addrcnt
);
510 list_for_each(pos
, &ep
->asocs
) {
511 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
513 if (!asoc
->peer
.asconf_capable
)
516 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_ADD_IP
)
519 if (!sctp_state(asoc
, ESTABLISHED
))
522 /* Check if any address in the packed array of addresses is
523 * in the bind address list of the association. If so,
524 * do not send the asconf chunk to its peer, but continue with
525 * other associations.
528 for (i
= 0; i
< addrcnt
; i
++) {
529 addr
= (union sctp_addr
*)addr_buf
;
530 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
536 if (sctp_assoc_lookup_laddr(asoc
, addr
))
539 addr_buf
+= af
->sockaddr_len
;
544 /* Use the first address in bind addr list of association as
545 * Address Parameter of ASCONF CHUNK.
547 sctp_read_lock(&asoc
->base
.addr_lock
);
548 bp
= &asoc
->base
.bind_addr
;
549 p
= bp
->address_list
.next
;
550 laddr
= list_entry(p
, struct sctp_sockaddr_entry
, list
);
551 sctp_read_unlock(&asoc
->base
.addr_lock
);
553 chunk
= sctp_make_asconf_update_ip(asoc
, &laddr
->a
, addrs
,
554 addrcnt
, SCTP_PARAM_ADD_IP
);
560 retval
= sctp_send_asconf(asoc
, chunk
);
562 /* FIXME: After sending the add address ASCONF chunk, we
563 * cannot append the address to the association's binding
564 * address list, because the new address may be used as the
565 * source of a message sent to the peer before the ASCONF
566 * chunk is received by the peer. So we should wait until
567 * ASCONF_ACK is received.
575 /* Remove a list of addresses from bind addresses list. Do not remove the
578 * Basically run through each address specified in the addrs/addrcnt
579 * array/length pair, determine if it is IPv6 or IPv4 and call
580 * sctp_del_bind() on it.
582 * If any of them fails, then the operation will be reversed and the
583 * ones that were removed will be added back.
585 * At least one address has to be left; if only one address is
586 * available, the operation will return -EBUSY.
588 * Only sctp_setsockopt_bindx() is supposed to call this function.
590 int sctp_bindx_rem(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
592 struct sctp_sock
*sp
= sctp_sk(sk
);
593 struct sctp_endpoint
*ep
= sp
->ep
;
595 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
597 union sctp_addr saveaddr
;
599 struct sockaddr
*sa_addr
;
602 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
606 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
607 /* If the bind address list is empty or if there is only one
608 * bind address, there is nothing more to be removed (we need
609 * at least one address here).
611 if (list_empty(&bp
->address_list
) ||
612 (sctp_list_single_entry(&bp
->address_list
))) {
617 /* The list may contain either IPv4 or IPv6 address;
618 * determine the address length to copy the address to
621 sa_addr
= (struct sockaddr
*)addr_buf
;
622 af
= sctp_get_af_specific(sa_addr
->sa_family
);
627 memcpy(&saveaddr
, sa_addr
, af
->sockaddr_len
);
628 saveaddr
.v4
.sin_port
= ntohs(saveaddr
.v4
.sin_port
);
629 if (saveaddr
.v4
.sin_port
!= bp
->port
) {
634 /* FIXME - There is probably a need to check if sk->sk_saddr and
635 * sk->sk_rcv_addr are currently set to one of the addresses to
636 * be removed. This is something which needs to be looked into
637 * when we are fixing the outstanding issues with multi-homing
638 * socket routing and failover schemes. Refer to comments in
639 * sctp_do_bind(). -daisy
641 sctp_local_bh_disable();
642 sctp_write_lock(&ep
->base
.addr_lock
);
644 retval
= sctp_del_bind_addr(bp
, &saveaddr
);
646 sctp_write_unlock(&ep
->base
.addr_lock
);
647 sctp_local_bh_enable();
649 addr_buf
+= af
->sockaddr_len
;
652 /* Failed. Add the ones that has been removed back */
654 sctp_bindx_add(sk
, addrs
, cnt
);
662 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
663 * the associations that are part of the endpoint indicating that a list of
664 * local addresses are removed from the endpoint.
666 * If any of the addresses is already in the bind address list of the
667 * association, we do not send the chunk for that association. But it will not
668 * affect other associations.
670 * Only sctp_setsockopt_bindx() is supposed to call this function.
672 static int sctp_send_asconf_del_ip(struct sock
*sk
,
673 struct sockaddr
*addrs
,
676 struct sctp_sock
*sp
;
677 struct sctp_endpoint
*ep
;
678 struct sctp_association
*asoc
;
679 struct sctp_bind_addr
*bp
;
680 struct sctp_chunk
*chunk
;
681 union sctp_addr
*laddr
;
684 struct list_head
*pos
;
688 if (!sctp_addip_enable
)
694 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
695 __FUNCTION__
, sk
, addrs
, addrcnt
);
697 list_for_each(pos
, &ep
->asocs
) {
698 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
700 if (!asoc
->peer
.asconf_capable
)
703 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_DEL_IP
)
706 if (!sctp_state(asoc
, ESTABLISHED
))
709 /* Check if any address in the packed array of addresses is
710 * not present in the bind address list of the association.
711 * If so, do not send the asconf chunk to its peer, but
712 * continue with other associations.
715 for (i
= 0; i
< addrcnt
; i
++) {
716 laddr
= (union sctp_addr
*)addr_buf
;
717 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
723 if (!sctp_assoc_lookup_laddr(asoc
, laddr
))
726 addr_buf
+= af
->sockaddr_len
;
731 /* Find one address in the association's bind address list
732 * that is not in the packed array of addresses. This is to
733 * make sure that we do not delete all the addresses in the
736 sctp_read_lock(&asoc
->base
.addr_lock
);
737 bp
= &asoc
->base
.bind_addr
;
738 laddr
= sctp_find_unmatch_addr(bp
, (union sctp_addr
*)addrs
,
740 sctp_read_unlock(&asoc
->base
.addr_lock
);
744 chunk
= sctp_make_asconf_update_ip(asoc
, laddr
, addrs
, addrcnt
,
751 retval
= sctp_send_asconf(asoc
, chunk
);
753 /* FIXME: After sending the delete address ASCONF chunk, we
754 * cannot remove the addresses from the association's bind
755 * address list, because there maybe some packet send to
756 * the delete addresses, so we should wait until ASCONF_ACK
757 * packet is received.
764 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
767 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
770 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
771 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
774 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
775 * Section 3.1.2 for this usage.
777 * addrs is a pointer to an array of one or more socket addresses. Each
778 * address is contained in its appropriate structure (i.e. struct
779 * sockaddr_in or struct sockaddr_in6) the family of the address type
780 * must be used to distengish the address length (note that this
781 * representation is termed a "packed array" of addresses). The caller
782 * specifies the number of addresses in the array with addrcnt.
784 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
785 * -1, and sets errno to the appropriate error code.
787 * For SCTP, the port given in each socket address must be the same, or
788 * sctp_bindx() will fail, setting errno to EINVAL.
790 * The flags parameter is formed from the bitwise OR of zero or more of
791 * the following currently defined flags:
793 * SCTP_BINDX_ADD_ADDR
795 * SCTP_BINDX_REM_ADDR
797 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
798 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
799 * addresses from the association. The two flags are mutually exclusive;
800 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
801 * not remove all addresses from an association; sctp_bindx() will
802 * reject such an attempt with EINVAL.
804 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
805 * additional addresses with an endpoint after calling bind(). Or use
806 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
807 * socket is associated with so that no new association accepted will be
808 * associated with those addresses. If the endpoint supports dynamic
809 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
810 * endpoint to send the appropriate message to the peer to change the
811 * peers address lists.
813 * Adding and removing addresses from a connected association is
814 * optional functionality. Implementations that do not support this
815 * functionality should return EOPNOTSUPP.
817 * Basically do nothing but copying the addresses from user to kernel
818 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
819 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
822 * We don't use copy_from_user() for optimization: we first do the
823 * sanity checks (buffer size -fast- and access check-healthy
824 * pointer); if all of those succeed, then we can alloc the memory
825 * (expensive operation) needed to copy the data to kernel. Then we do
826 * the copying without checking the user space area
827 * (__copy_from_user()).
829 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
832 * sk The sk of the socket
833 * addrs The pointer to the addresses in user land
834 * addrssize Size of the addrs buffer
835 * op Operation to perform (add or remove, see the flags of
838 * Returns 0 if ok, <0 errno code on error.
840 SCTP_STATIC
int sctp_setsockopt_bindx(struct sock
* sk
,
841 struct sockaddr __user
*addrs
,
842 int addrs_size
, int op
)
844 struct sockaddr
*kaddrs
;
848 struct sockaddr
*sa_addr
;
852 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
853 " addrs_size %d opt %d\n", sk
, addrs
, addrs_size
, op
);
855 if (unlikely(addrs_size
<= 0))
858 /* Check the user passed a healthy pointer. */
859 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
862 /* Alloc space for the address array in kernel memory. */
863 kaddrs
= (struct sockaddr
*)kmalloc(addrs_size
, GFP_KERNEL
);
864 if (unlikely(!kaddrs
))
867 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
872 /* Walk through the addrs buffer and count the number of addresses. */
874 while (walk_size
< addrs_size
) {
875 sa_addr
= (struct sockaddr
*)addr_buf
;
876 af
= sctp_get_af_specific(sa_addr
->sa_family
);
878 /* If the address family is not supported or if this address
879 * causes the address buffer to overflow return EINVAL.
881 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
886 addr_buf
+= af
->sockaddr_len
;
887 walk_size
+= af
->sockaddr_len
;
892 case SCTP_BINDX_ADD_ADDR
:
893 err
= sctp_bindx_add(sk
, kaddrs
, addrcnt
);
896 err
= sctp_send_asconf_add_ip(sk
, kaddrs
, addrcnt
);
899 case SCTP_BINDX_REM_ADDR
:
900 err
= sctp_bindx_rem(sk
, kaddrs
, addrcnt
);
903 err
= sctp_send_asconf_del_ip(sk
, kaddrs
, addrcnt
);
917 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
919 * Common routine for handling connect() and sctp_connectx().
920 * Connect will come in with just a single address.
922 static int __sctp_connect(struct sock
* sk
,
923 struct sockaddr
*kaddrs
,
926 struct sctp_sock
*sp
;
927 struct sctp_endpoint
*ep
;
928 struct sctp_association
*asoc
= NULL
;
929 struct sctp_association
*asoc2
;
930 struct sctp_transport
*transport
;
938 struct sockaddr
*sa_addr
;
944 /* connect() cannot be done on a socket that is already in ESTABLISHED
945 * state - UDP-style peeled off socket or a TCP-style socket that
946 * is already connected.
947 * It cannot be done even on a TCP-style listening socket.
949 if (sctp_sstate(sk
, ESTABLISHED
) ||
950 (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))) {
955 /* Walk through the addrs buffer and count the number of addresses. */
957 while (walk_size
< addrs_size
) {
958 sa_addr
= (struct sockaddr
*)addr_buf
;
959 af
= sctp_get_af_specific(sa_addr
->sa_family
);
961 /* If the address family is not supported or if this address
962 * causes the address buffer to overflow return EINVAL.
964 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
969 err
= sctp_verify_addr(sk
, (union sctp_addr
*)sa_addr
,
974 memcpy(&to
, sa_addr
, af
->sockaddr_len
);
975 to
.v4
.sin_port
= ntohs(to
.v4
.sin_port
);
977 /* Check if there already is a matching association on the
978 * endpoint (other than the one created here).
980 asoc2
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
981 if (asoc2
&& asoc2
!= asoc
) {
982 if (asoc2
->state
>= SCTP_STATE_ESTABLISHED
)
989 /* If we could not find a matching association on the endpoint,
990 * make sure that there is no peeled-off association matching
991 * the peer address even on another socket.
993 if (sctp_endpoint_is_peeled_off(ep
, &to
)) {
994 err
= -EADDRNOTAVAIL
;
999 /* If a bind() or sctp_bindx() is not called prior to
1000 * an sctp_connectx() call, the system picks an
1001 * ephemeral port and will choose an address set
1002 * equivalent to binding with a wildcard address.
1004 if (!ep
->base
.bind_addr
.port
) {
1005 if (sctp_autobind(sk
)) {
1011 * If an unprivileged user inherits a 1-many
1012 * style socket with open associations on a
1013 * privileged port, it MAY be permitted to
1014 * accept new associations, but it SHOULD NOT
1015 * be permitted to open new associations.
1017 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1018 !capable(CAP_NET_BIND_SERVICE
)) {
1024 scope
= sctp_scope(&to
);
1025 asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1032 /* Prime the peer's transport structures. */
1033 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
,
1041 addr_buf
+= af
->sockaddr_len
;
1042 walk_size
+= af
->sockaddr_len
;
1045 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, GFP_KERNEL
);
1050 err
= sctp_primitive_ASSOCIATE(asoc
, NULL
);
1055 /* Initialize sk's dport and daddr for getpeername() */
1056 inet_sk(sk
)->dport
= htons(asoc
->peer
.port
);
1057 af
= sctp_get_af_specific(to
.sa
.sa_family
);
1058 af
->to_sk_daddr(&to
, sk
);
1060 timeo
= sock_sndtimeo(sk
, sk
->sk_socket
->file
->f_flags
& O_NONBLOCK
);
1061 err
= sctp_wait_for_connect(asoc
, &timeo
);
1063 /* Don't free association on exit. */
1068 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1069 " kaddrs: %p err: %d\n",
1072 sctp_association_free(asoc
);
1076 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1079 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt);
1081 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1082 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1083 * or IPv6 addresses.
1085 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1086 * Section 3.1.2 for this usage.
1088 * addrs is a pointer to an array of one or more socket addresses. Each
1089 * address is contained in its appropriate structure (i.e. struct
1090 * sockaddr_in or struct sockaddr_in6) the family of the address type
1091 * must be used to distengish the address length (note that this
1092 * representation is termed a "packed array" of addresses). The caller
1093 * specifies the number of addresses in the array with addrcnt.
1095 * On success, sctp_connectx() returns 0. On failure, sctp_connectx() returns
1096 * -1, and sets errno to the appropriate error code.
1098 * For SCTP, the port given in each socket address must be the same, or
1099 * sctp_connectx() will fail, setting errno to EINVAL.
1101 * An application can use sctp_connectx to initiate an association with
1102 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1103 * allows a caller to specify multiple addresses at which a peer can be
1104 * reached. The way the SCTP stack uses the list of addresses to set up
1105 * the association is implementation dependant. This function only
1106 * specifies that the stack will try to make use of all the addresses in
1107 * the list when needed.
1109 * Note that the list of addresses passed in is only used for setting up
1110 * the association. It does not necessarily equal the set of addresses
1111 * the peer uses for the resulting association. If the caller wants to
1112 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1113 * retrieve them after the association has been set up.
1115 * Basically do nothing but copying the addresses from user to kernel
1116 * land and invoking either sctp_connectx(). This is used for tunneling
1117 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1119 * We don't use copy_from_user() for optimization: we first do the
1120 * sanity checks (buffer size -fast- and access check-healthy
1121 * pointer); if all of those succeed, then we can alloc the memory
1122 * (expensive operation) needed to copy the data to kernel. Then we do
1123 * the copying without checking the user space area
1124 * (__copy_from_user()).
1126 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1129 * sk The sk of the socket
1130 * addrs The pointer to the addresses in user land
1131 * addrssize Size of the addrs buffer
1133 * Returns 0 if ok, <0 errno code on error.
1135 SCTP_STATIC
int sctp_setsockopt_connectx(struct sock
* sk
,
1136 struct sockaddr __user
*addrs
,
1140 struct sockaddr
*kaddrs
;
1142 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1143 __FUNCTION__
, sk
, addrs
, addrs_size
);
1145 if (unlikely(addrs_size
<= 0))
1148 /* Check the user passed a healthy pointer. */
1149 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
1152 /* Alloc space for the address array in kernel memory. */
1153 kaddrs
= (struct sockaddr
*)kmalloc(addrs_size
, GFP_KERNEL
);
1154 if (unlikely(!kaddrs
))
1157 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
1160 err
= __sctp_connect(sk
, kaddrs
, addrs_size
);
1167 /* API 3.1.4 close() - UDP Style Syntax
1168 * Applications use close() to perform graceful shutdown (as described in
1169 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1170 * by a UDP-style socket.
1174 * ret = close(int sd);
1176 * sd - the socket descriptor of the associations to be closed.
1178 * To gracefully shutdown a specific association represented by the
1179 * UDP-style socket, an application should use the sendmsg() call,
1180 * passing no user data, but including the appropriate flag in the
1181 * ancillary data (see Section xxxx).
1183 * If sd in the close() call is a branched-off socket representing only
1184 * one association, the shutdown is performed on that association only.
1186 * 4.1.6 close() - TCP Style Syntax
1188 * Applications use close() to gracefully close down an association.
1192 * int close(int sd);
1194 * sd - the socket descriptor of the association to be closed.
1196 * After an application calls close() on a socket descriptor, no further
1197 * socket operations will succeed on that descriptor.
1199 * API 7.1.4 SO_LINGER
1201 * An application using the TCP-style socket can use this option to
1202 * perform the SCTP ABORT primitive. The linger option structure is:
1205 * int l_onoff; // option on/off
1206 * int l_linger; // linger time
1209 * To enable the option, set l_onoff to 1. If the l_linger value is set
1210 * to 0, calling close() is the same as the ABORT primitive. If the
1211 * value is set to a negative value, the setsockopt() call will return
1212 * an error. If the value is set to a positive value linger_time, the
1213 * close() can be blocked for at most linger_time ms. If the graceful
1214 * shutdown phase does not finish during this period, close() will
1215 * return but the graceful shutdown phase continues in the system.
1217 SCTP_STATIC
void sctp_close(struct sock
*sk
, long timeout
)
1219 struct sctp_endpoint
*ep
;
1220 struct sctp_association
*asoc
;
1221 struct list_head
*pos
, *temp
;
1223 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk
, timeout
);
1226 sk
->sk_shutdown
= SHUTDOWN_MASK
;
1228 ep
= sctp_sk(sk
)->ep
;
1230 /* Walk all associations on a socket, not on an endpoint. */
1231 list_for_each_safe(pos
, temp
, &ep
->asocs
) {
1232 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
1234 if (sctp_style(sk
, TCP
)) {
1235 /* A closed association can still be in the list if
1236 * it belongs to a TCP-style listening socket that is
1237 * not yet accepted. If so, free it. If not, send an
1238 * ABORT or SHUTDOWN based on the linger options.
1240 if (sctp_state(asoc
, CLOSED
)) {
1241 sctp_unhash_established(asoc
);
1242 sctp_association_free(asoc
);
1244 } else if (sock_flag(sk
, SOCK_LINGER
) &&
1246 sctp_primitive_ABORT(asoc
, NULL
);
1248 sctp_primitive_SHUTDOWN(asoc
, NULL
);
1250 sctp_primitive_SHUTDOWN(asoc
, NULL
);
1253 /* Clean up any skbs sitting on the receive queue. */
1254 sctp_queue_purge_ulpevents(&sk
->sk_receive_queue
);
1255 sctp_queue_purge_ulpevents(&sctp_sk(sk
)->pd_lobby
);
1257 /* On a TCP-style socket, block for at most linger_time if set. */
1258 if (sctp_style(sk
, TCP
) && timeout
)
1259 sctp_wait_for_close(sk
, timeout
);
1261 /* This will run the backlog queue. */
1262 sctp_release_sock(sk
);
1264 /* Supposedly, no process has access to the socket, but
1265 * the net layers still may.
1267 sctp_local_bh_disable();
1268 sctp_bh_lock_sock(sk
);
1270 /* Hold the sock, since sk_common_release() will put sock_put()
1271 * and we have just a little more cleanup.
1274 sk_common_release(sk
);
1276 sctp_bh_unlock_sock(sk
);
1277 sctp_local_bh_enable();
1281 SCTP_DBG_OBJCNT_DEC(sock
);
1284 /* Handle EPIPE error. */
1285 static int sctp_error(struct sock
*sk
, int flags
, int err
)
1288 err
= sock_error(sk
) ? : -EPIPE
;
1289 if (err
== -EPIPE
&& !(flags
& MSG_NOSIGNAL
))
1290 send_sig(SIGPIPE
, current
, 0);
1294 /* API 3.1.3 sendmsg() - UDP Style Syntax
1296 * An application uses sendmsg() and recvmsg() calls to transmit data to
1297 * and receive data from its peer.
1299 * ssize_t sendmsg(int socket, const struct msghdr *message,
1302 * socket - the socket descriptor of the endpoint.
1303 * message - pointer to the msghdr structure which contains a single
1304 * user message and possibly some ancillary data.
1306 * See Section 5 for complete description of the data
1309 * flags - flags sent or received with the user message, see Section
1310 * 5 for complete description of the flags.
1312 * Note: This function could use a rewrite especially when explicit
1313 * connect support comes in.
1315 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1317 SCTP_STATIC
int sctp_msghdr_parse(const struct msghdr
*, sctp_cmsgs_t
*);
1319 SCTP_STATIC
int sctp_sendmsg(struct kiocb
*iocb
, struct sock
*sk
,
1320 struct msghdr
*msg
, size_t msg_len
)
1322 struct sctp_sock
*sp
;
1323 struct sctp_endpoint
*ep
;
1324 struct sctp_association
*new_asoc
=NULL
, *asoc
=NULL
;
1325 struct sctp_transport
*transport
, *chunk_tp
;
1326 struct sctp_chunk
*chunk
;
1328 struct sockaddr
*msg_name
= NULL
;
1329 struct sctp_sndrcvinfo default_sinfo
= { 0 };
1330 struct sctp_sndrcvinfo
*sinfo
;
1331 struct sctp_initmsg
*sinit
;
1332 sctp_assoc_t associd
= 0;
1333 sctp_cmsgs_t cmsgs
= { NULL
};
1337 __u16 sinfo_flags
= 0;
1338 struct sctp_datamsg
*datamsg
;
1339 struct list_head
*pos
;
1340 int msg_flags
= msg
->msg_flags
;
1342 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1349 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep
);
1351 /* We cannot send a message over a TCP-style listening socket. */
1352 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
)) {
1357 /* Parse out the SCTP CMSGs. */
1358 err
= sctp_msghdr_parse(msg
, &cmsgs
);
1361 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err
);
1365 /* Fetch the destination address for this packet. This
1366 * address only selects the association--it is not necessarily
1367 * the address we will send to.
1368 * For a peeled-off socket, msg_name is ignored.
1370 if (!sctp_style(sk
, UDP_HIGH_BANDWIDTH
) && msg
->msg_name
) {
1371 int msg_namelen
= msg
->msg_namelen
;
1373 err
= sctp_verify_addr(sk
, (union sctp_addr
*)msg
->msg_name
,
1378 if (msg_namelen
> sizeof(to
))
1379 msg_namelen
= sizeof(to
);
1380 memcpy(&to
, msg
->msg_name
, msg_namelen
);
1381 SCTP_DEBUG_PRINTK("Just memcpy'd. msg_name is "
1383 to
.v4
.sin_addr
.s_addr
, to
.v4
.sin_port
);
1385 to
.v4
.sin_port
= ntohs(to
.v4
.sin_port
);
1386 msg_name
= msg
->msg_name
;
1392 /* Did the user specify SNDRCVINFO? */
1394 sinfo_flags
= sinfo
->sinfo_flags
;
1395 associd
= sinfo
->sinfo_assoc_id
;
1398 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1399 msg_len
, sinfo_flags
);
1401 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1402 if (sctp_style(sk
, TCP
) && (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
))) {
1407 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1408 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1409 * If SCTP_ABORT is set, the message length could be non zero with
1410 * the msg_iov set to the user abort reason.
1412 if (((sinfo_flags
& SCTP_EOF
) && (msg_len
> 0)) ||
1413 (!(sinfo_flags
& (SCTP_EOF
|SCTP_ABORT
)) && (msg_len
== 0))) {
1418 /* If SCTP_ADDR_OVER is set, there must be an address
1419 * specified in msg_name.
1421 if ((sinfo_flags
& SCTP_ADDR_OVER
) && (!msg
->msg_name
)) {
1428 SCTP_DEBUG_PRINTK("About to look up association.\n");
1432 /* If a msg_name has been specified, assume this is to be used. */
1434 /* Look for a matching association on the endpoint. */
1435 asoc
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1437 /* If we could not find a matching association on the
1438 * endpoint, make sure that it is not a TCP-style
1439 * socket that already has an association or there is
1440 * no peeled-off association on another socket.
1442 if ((sctp_style(sk
, TCP
) &&
1443 sctp_sstate(sk
, ESTABLISHED
)) ||
1444 sctp_endpoint_is_peeled_off(ep
, &to
)) {
1445 err
= -EADDRNOTAVAIL
;
1450 asoc
= sctp_id2assoc(sk
, associd
);
1458 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc
);
1460 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1461 * socket that has an association in CLOSED state. This can
1462 * happen when an accepted socket has an association that is
1465 if (sctp_state(asoc
, CLOSED
) && sctp_style(sk
, TCP
)) {
1470 if (sinfo_flags
& SCTP_EOF
) {
1471 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1473 sctp_primitive_SHUTDOWN(asoc
, NULL
);
1477 if (sinfo_flags
& SCTP_ABORT
) {
1478 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc
);
1479 sctp_primitive_ABORT(asoc
, msg
);
1485 /* Do we need to create the association? */
1487 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1489 if (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
)) {
1494 /* Check for invalid stream against the stream counts,
1495 * either the default or the user specified stream counts.
1498 if (!sinit
|| (sinit
&& !sinit
->sinit_num_ostreams
)) {
1499 /* Check against the defaults. */
1500 if (sinfo
->sinfo_stream
>=
1501 sp
->initmsg
.sinit_num_ostreams
) {
1506 /* Check against the requested. */
1507 if (sinfo
->sinfo_stream
>=
1508 sinit
->sinit_num_ostreams
) {
1516 * API 3.1.2 bind() - UDP Style Syntax
1517 * If a bind() or sctp_bindx() is not called prior to a
1518 * sendmsg() call that initiates a new association, the
1519 * system picks an ephemeral port and will choose an address
1520 * set equivalent to binding with a wildcard address.
1522 if (!ep
->base
.bind_addr
.port
) {
1523 if (sctp_autobind(sk
)) {
1529 * If an unprivileged user inherits a one-to-many
1530 * style socket with open associations on a privileged
1531 * port, it MAY be permitted to accept new associations,
1532 * but it SHOULD NOT be permitted to open new
1535 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1536 !capable(CAP_NET_BIND_SERVICE
)) {
1542 scope
= sctp_scope(&to
);
1543 new_asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1550 /* If the SCTP_INIT ancillary data is specified, set all
1551 * the association init values accordingly.
1554 if (sinit
->sinit_num_ostreams
) {
1555 asoc
->c
.sinit_num_ostreams
=
1556 sinit
->sinit_num_ostreams
;
1558 if (sinit
->sinit_max_instreams
) {
1559 asoc
->c
.sinit_max_instreams
=
1560 sinit
->sinit_max_instreams
;
1562 if (sinit
->sinit_max_attempts
) {
1563 asoc
->max_init_attempts
1564 = sinit
->sinit_max_attempts
;
1566 if (sinit
->sinit_max_init_timeo
) {
1567 asoc
->max_init_timeo
=
1568 msecs_to_jiffies(sinit
->sinit_max_init_timeo
);
1572 /* Prime the peer's transport structures. */
1573 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
, SCTP_UNKNOWN
);
1578 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, GFP_KERNEL
);
1585 /* ASSERT: we have a valid association at this point. */
1586 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1589 /* If the user didn't specify SNDRCVINFO, make up one with
1592 default_sinfo
.sinfo_stream
= asoc
->default_stream
;
1593 default_sinfo
.sinfo_flags
= asoc
->default_flags
;
1594 default_sinfo
.sinfo_ppid
= asoc
->default_ppid
;
1595 default_sinfo
.sinfo_context
= asoc
->default_context
;
1596 default_sinfo
.sinfo_timetolive
= asoc
->default_timetolive
;
1597 default_sinfo
.sinfo_assoc_id
= sctp_assoc2id(asoc
);
1598 sinfo
= &default_sinfo
;
1601 /* API 7.1.7, the sndbuf size per association bounds the
1602 * maximum size of data that can be sent in a single send call.
1604 if (msg_len
> sk
->sk_sndbuf
) {
1609 /* If fragmentation is disabled and the message length exceeds the
1610 * association fragmentation point, return EMSGSIZE. The I-D
1611 * does not specify what this error is, but this looks like
1614 if (sctp_sk(sk
)->disable_fragments
&& (msg_len
> asoc
->frag_point
)) {
1620 /* Check for invalid stream. */
1621 if (sinfo
->sinfo_stream
>= asoc
->c
.sinit_num_ostreams
) {
1627 timeo
= sock_sndtimeo(sk
, msg
->msg_flags
& MSG_DONTWAIT
);
1628 if (!sctp_wspace(asoc
)) {
1629 err
= sctp_wait_for_sndbuf(asoc
, &timeo
, msg_len
);
1634 /* If an address is passed with the sendto/sendmsg call, it is used
1635 * to override the primary destination address in the TCP model, or
1636 * when SCTP_ADDR_OVER flag is set in the UDP model.
1638 if ((sctp_style(sk
, TCP
) && msg_name
) ||
1639 (sinfo_flags
& SCTP_ADDR_OVER
)) {
1640 chunk_tp
= sctp_assoc_lookup_paddr(asoc
, &to
);
1648 /* Auto-connect, if we aren't connected already. */
1649 if (sctp_state(asoc
, CLOSED
)) {
1650 err
= sctp_primitive_ASSOCIATE(asoc
, NULL
);
1653 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1656 /* Break the message into multiple chunks of maximum size. */
1657 datamsg
= sctp_datamsg_from_user(asoc
, sinfo
, msg
, msg_len
);
1663 /* Now send the (possibly) fragmented message. */
1664 list_for_each(pos
, &datamsg
->chunks
) {
1665 chunk
= list_entry(pos
, struct sctp_chunk
, frag_list
);
1666 sctp_datamsg_track(chunk
);
1668 /* Do accounting for the write space. */
1669 sctp_set_owner_w(chunk
);
1671 chunk
->transport
= chunk_tp
;
1673 /* Send it to the lower layers. Note: all chunks
1674 * must either fail or succeed. The lower layer
1675 * works that way today. Keep it that way or this
1678 err
= sctp_primitive_SEND(asoc
, chunk
);
1679 /* Did the lower layer accept the chunk? */
1681 sctp_chunk_free(chunk
);
1682 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1685 sctp_datamsg_free(datamsg
);
1691 /* If we are already past ASSOCIATE, the lower
1692 * layers are responsible for association cleanup.
1698 sctp_association_free(asoc
);
1700 sctp_release_sock(sk
);
1703 return sctp_error(sk
, msg_flags
, err
);
1710 err
= sock_error(sk
);
1720 /* This is an extended version of skb_pull() that removes the data from the
1721 * start of a skb even when data is spread across the list of skb's in the
1722 * frag_list. len specifies the total amount of data that needs to be removed.
1723 * when 'len' bytes could be removed from the skb, it returns 0.
1724 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1725 * could not be removed.
1727 static int sctp_skb_pull(struct sk_buff
*skb
, int len
)
1729 struct sk_buff
*list
;
1730 int skb_len
= skb_headlen(skb
);
1733 if (len
<= skb_len
) {
1734 __skb_pull(skb
, len
);
1738 __skb_pull(skb
, skb_len
);
1740 for (list
= skb_shinfo(skb
)->frag_list
; list
; list
= list
->next
) {
1741 rlen
= sctp_skb_pull(list
, len
);
1742 skb
->len
-= (len
-rlen
);
1743 skb
->data_len
-= (len
-rlen
);
1754 /* API 3.1.3 recvmsg() - UDP Style Syntax
1756 * ssize_t recvmsg(int socket, struct msghdr *message,
1759 * socket - the socket descriptor of the endpoint.
1760 * message - pointer to the msghdr structure which contains a single
1761 * user message and possibly some ancillary data.
1763 * See Section 5 for complete description of the data
1766 * flags - flags sent or received with the user message, see Section
1767 * 5 for complete description of the flags.
1769 static struct sk_buff
*sctp_skb_recv_datagram(struct sock
*, int, int, int *);
1771 SCTP_STATIC
int sctp_recvmsg(struct kiocb
*iocb
, struct sock
*sk
,
1772 struct msghdr
*msg
, size_t len
, int noblock
,
1773 int flags
, int *addr_len
)
1775 struct sctp_ulpevent
*event
= NULL
;
1776 struct sctp_sock
*sp
= sctp_sk(sk
);
1777 struct sk_buff
*skb
;
1782 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1783 "0x%x, %s: %p)\n", "sk", sk
, "msghdr", msg
,
1784 "len", len
, "knoblauch", noblock
,
1785 "flags", flags
, "addr_len", addr_len
);
1789 if (sctp_style(sk
, TCP
) && !sctp_sstate(sk
, ESTABLISHED
)) {
1794 skb
= sctp_skb_recv_datagram(sk
, flags
, noblock
, &err
);
1798 /* Get the total length of the skb including any skb's in the
1807 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
1809 event
= sctp_skb2event(skb
);
1814 sock_recv_timestamp(msg
, sk
, skb
);
1815 if (sctp_ulpevent_is_notification(event
)) {
1816 msg
->msg_flags
|= MSG_NOTIFICATION
;
1817 sp
->pf
->event_msgname(event
, msg
->msg_name
, addr_len
);
1819 sp
->pf
->skb_msgname(skb
, msg
->msg_name
, addr_len
);
1822 /* Check if we allow SCTP_SNDRCVINFO. */
1823 if (sp
->subscribe
.sctp_data_io_event
)
1824 sctp_ulpevent_read_sndrcvinfo(event
, msg
);
1826 /* FIXME: we should be calling IP/IPv6 layers. */
1827 if (sk
->sk_protinfo
.af_inet
.cmsg_flags
)
1828 ip_cmsg_recv(msg
, skb
);
1833 /* If skb's length exceeds the user's buffer, update the skb and
1834 * push it back to the receive_queue so that the next call to
1835 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1837 if (skb_len
> copied
) {
1838 msg
->msg_flags
&= ~MSG_EOR
;
1839 if (flags
& MSG_PEEK
)
1841 sctp_skb_pull(skb
, copied
);
1842 skb_queue_head(&sk
->sk_receive_queue
, skb
);
1844 /* When only partial message is copied to the user, increase
1845 * rwnd by that amount. If all the data in the skb is read,
1846 * rwnd is updated when the event is freed.
1848 sctp_assoc_rwnd_increase(event
->asoc
, copied
);
1850 } else if ((event
->msg_flags
& MSG_NOTIFICATION
) ||
1851 (event
->msg_flags
& MSG_EOR
))
1852 msg
->msg_flags
|= MSG_EOR
;
1854 msg
->msg_flags
&= ~MSG_EOR
;
1857 if (flags
& MSG_PEEK
) {
1858 /* Release the skb reference acquired after peeking the skb in
1859 * sctp_skb_recv_datagram().
1863 /* Free the event which includes releasing the reference to
1864 * the owner of the skb, freeing the skb and updating the
1867 sctp_ulpevent_free(event
);
1870 sctp_release_sock(sk
);
1874 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
1876 * This option is a on/off flag. If enabled no SCTP message
1877 * fragmentation will be performed. Instead if a message being sent
1878 * exceeds the current PMTU size, the message will NOT be sent and
1879 * instead a error will be indicated to the user.
1881 static int sctp_setsockopt_disable_fragments(struct sock
*sk
,
1882 char __user
*optval
, int optlen
)
1886 if (optlen
< sizeof(int))
1889 if (get_user(val
, (int __user
*)optval
))
1892 sctp_sk(sk
)->disable_fragments
= (val
== 0) ? 0 : 1;
1897 static int sctp_setsockopt_events(struct sock
*sk
, char __user
*optval
,
1900 if (optlen
!= sizeof(struct sctp_event_subscribe
))
1902 if (copy_from_user(&sctp_sk(sk
)->subscribe
, optval
, optlen
))
1907 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
1909 * This socket option is applicable to the UDP-style socket only. When
1910 * set it will cause associations that are idle for more than the
1911 * specified number of seconds to automatically close. An association
1912 * being idle is defined an association that has NOT sent or received
1913 * user data. The special value of '0' indicates that no automatic
1914 * close of any associations should be performed. The option expects an
1915 * integer defining the number of seconds of idle time before an
1916 * association is closed.
1918 static int sctp_setsockopt_autoclose(struct sock
*sk
, char __user
*optval
,
1921 struct sctp_sock
*sp
= sctp_sk(sk
);
1923 /* Applicable to UDP-style socket only */
1924 if (sctp_style(sk
, TCP
))
1926 if (optlen
!= sizeof(int))
1928 if (copy_from_user(&sp
->autoclose
, optval
, optlen
))
1934 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
1936 * Applications can enable or disable heartbeats for any peer address of
1937 * an association, modify an address's heartbeat interval, force a
1938 * heartbeat to be sent immediately, and adjust the address's maximum
1939 * number of retransmissions sent before an address is considered
1940 * unreachable. The following structure is used to access and modify an
1941 * address's parameters:
1943 * struct sctp_paddrparams {
1944 * sctp_assoc_t spp_assoc_id;
1945 * struct sockaddr_storage spp_address;
1946 * uint32_t spp_hbinterval;
1947 * uint16_t spp_pathmaxrxt;
1950 * spp_assoc_id - (UDP style socket) This is filled in the application,
1951 * and identifies the association for this query.
1952 * spp_address - This specifies which address is of interest.
1953 * spp_hbinterval - This contains the value of the heartbeat interval,
1954 * in milliseconds. A value of 0, when modifying the
1955 * parameter, specifies that the heartbeat on this
1956 * address should be disabled. A value of UINT32_MAX
1957 * (4294967295), when modifying the parameter,
1958 * specifies that a heartbeat should be sent
1959 * immediately to the peer address, and the current
1960 * interval should remain unchanged.
1961 * spp_pathmaxrxt - This contains the maximum number of
1962 * retransmissions before this address shall be
1963 * considered unreachable.
1965 static int sctp_setsockopt_peer_addr_params(struct sock
*sk
,
1966 char __user
*optval
, int optlen
)
1968 struct sctp_paddrparams params
;
1969 struct sctp_transport
*trans
;
1972 if (optlen
!= sizeof(struct sctp_paddrparams
))
1974 if (copy_from_user(¶ms
, optval
, optlen
))
1978 * API 7. Socket Options (setting the default value for the endpoint)
1979 * All options that support specific settings on an association by
1980 * filling in either an association id variable or a sockaddr_storage
1981 * SHOULD also support setting of the same value for the entire endpoint
1982 * (i.e. future associations). To accomplish this the following logic is
1983 * used when setting one of these options:
1985 * c) If neither the sockaddr_storage or association identification is
1986 * set i.e. the sockaddr_storage is set to all 0's (INADDR_ANY) and
1987 * the association identification is 0, the settings are a default
1988 * and to be applied to the endpoint (all future associations).
1991 /* update default value for endpoint (all future associations) */
1992 if (!params
.spp_assoc_id
&&
1993 sctp_is_any(( union sctp_addr
*)¶ms
.spp_address
)) {
1994 /* Manual heartbeat on an endpoint is invalid. */
1995 if (0xffffffff == params
.spp_hbinterval
)
1997 else if (params
.spp_hbinterval
)
1998 sctp_sk(sk
)->paddrparam
.spp_hbinterval
=
1999 params
.spp_hbinterval
;
2000 if (params
.spp_pathmaxrxt
)
2001 sctp_sk(sk
)->paddrparam
.spp_pathmaxrxt
=
2002 params
.spp_pathmaxrxt
;
2006 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
2007 params
.spp_assoc_id
);
2011 /* Applications can enable or disable heartbeats for any peer address
2012 * of an association, modify an address's heartbeat interval, force a
2013 * heartbeat to be sent immediately, and adjust the address's maximum
2014 * number of retransmissions sent before an address is considered
2017 * The value of the heartbeat interval, in milliseconds. A value of
2018 * UINT32_MAX (4294967295), when modifying the parameter, specifies
2019 * that a heartbeat should be sent immediately to the peer address,
2020 * and the current interval should remain unchanged.
2022 if (0xffffffff == params
.spp_hbinterval
) {
2023 error
= sctp_primitive_REQUESTHEARTBEAT (trans
->asoc
, trans
);
2027 /* The value of the heartbeat interval, in milliseconds. A value of 0,
2028 * when modifying the parameter, specifies that the heartbeat on this
2029 * address should be disabled.
2031 if (params
.spp_hbinterval
) {
2032 trans
->hb_allowed
= 1;
2033 trans
->hb_interval
=
2034 msecs_to_jiffies(params
.spp_hbinterval
);
2036 trans
->hb_allowed
= 0;
2039 /* spp_pathmaxrxt contains the maximum number of retransmissions
2040 * before this address shall be considered unreachable.
2042 if (params
.spp_pathmaxrxt
)
2043 trans
->max_retrans
= params
.spp_pathmaxrxt
;
2048 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2050 * Applications can specify protocol parameters for the default association
2051 * initialization. The option name argument to setsockopt() and getsockopt()
2054 * Setting initialization parameters is effective only on an unconnected
2055 * socket (for UDP-style sockets only future associations are effected
2056 * by the change). With TCP-style sockets, this option is inherited by
2057 * sockets derived from a listener socket.
2059 static int sctp_setsockopt_initmsg(struct sock
*sk
, char __user
*optval
, int optlen
)
2061 struct sctp_initmsg sinit
;
2062 struct sctp_sock
*sp
= sctp_sk(sk
);
2064 if (optlen
!= sizeof(struct sctp_initmsg
))
2066 if (copy_from_user(&sinit
, optval
, optlen
))
2069 if (sinit
.sinit_num_ostreams
)
2070 sp
->initmsg
.sinit_num_ostreams
= sinit
.sinit_num_ostreams
;
2071 if (sinit
.sinit_max_instreams
)
2072 sp
->initmsg
.sinit_max_instreams
= sinit
.sinit_max_instreams
;
2073 if (sinit
.sinit_max_attempts
)
2074 sp
->initmsg
.sinit_max_attempts
= sinit
.sinit_max_attempts
;
2075 if (sinit
.sinit_max_init_timeo
)
2076 sp
->initmsg
.sinit_max_init_timeo
= sinit
.sinit_max_init_timeo
;
2082 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2084 * Applications that wish to use the sendto() system call may wish to
2085 * specify a default set of parameters that would normally be supplied
2086 * through the inclusion of ancillary data. This socket option allows
2087 * such an application to set the default sctp_sndrcvinfo structure.
2088 * The application that wishes to use this socket option simply passes
2089 * in to this call the sctp_sndrcvinfo structure defined in Section
2090 * 5.2.2) The input parameters accepted by this call include
2091 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2092 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2093 * to this call if the caller is using the UDP model.
2095 static int sctp_setsockopt_default_send_param(struct sock
*sk
,
2096 char __user
*optval
, int optlen
)
2098 struct sctp_sndrcvinfo info
;
2099 struct sctp_association
*asoc
;
2100 struct sctp_sock
*sp
= sctp_sk(sk
);
2102 if (optlen
!= sizeof(struct sctp_sndrcvinfo
))
2104 if (copy_from_user(&info
, optval
, optlen
))
2107 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
2108 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
2112 asoc
->default_stream
= info
.sinfo_stream
;
2113 asoc
->default_flags
= info
.sinfo_flags
;
2114 asoc
->default_ppid
= info
.sinfo_ppid
;
2115 asoc
->default_context
= info
.sinfo_context
;
2116 asoc
->default_timetolive
= info
.sinfo_timetolive
;
2118 sp
->default_stream
= info
.sinfo_stream
;
2119 sp
->default_flags
= info
.sinfo_flags
;
2120 sp
->default_ppid
= info
.sinfo_ppid
;
2121 sp
->default_context
= info
.sinfo_context
;
2122 sp
->default_timetolive
= info
.sinfo_timetolive
;
2128 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2130 * Requests that the local SCTP stack use the enclosed peer address as
2131 * the association primary. The enclosed address must be one of the
2132 * association peer's addresses.
2134 static int sctp_setsockopt_primary_addr(struct sock
*sk
, char __user
*optval
,
2137 struct sctp_prim prim
;
2138 struct sctp_transport
*trans
;
2140 if (optlen
!= sizeof(struct sctp_prim
))
2143 if (copy_from_user(&prim
, optval
, sizeof(struct sctp_prim
)))
2146 trans
= sctp_addr_id2transport(sk
, &prim
.ssp_addr
, prim
.ssp_assoc_id
);
2150 sctp_assoc_set_primary(trans
->asoc
, trans
);
2156 * 7.1.5 SCTP_NODELAY
2158 * Turn on/off any Nagle-like algorithm. This means that packets are
2159 * generally sent as soon as possible and no unnecessary delays are
2160 * introduced, at the cost of more packets in the network. Expects an
2161 * integer boolean flag.
2163 static int sctp_setsockopt_nodelay(struct sock
*sk
, char __user
*optval
,
2168 if (optlen
< sizeof(int))
2170 if (get_user(val
, (int __user
*)optval
))
2173 sctp_sk(sk
)->nodelay
= (val
== 0) ? 0 : 1;
2179 * 7.1.1 SCTP_RTOINFO
2181 * The protocol parameters used to initialize and bound retransmission
2182 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2183 * and modify these parameters.
2184 * All parameters are time values, in milliseconds. A value of 0, when
2185 * modifying the parameters, indicates that the current value should not
2189 static int sctp_setsockopt_rtoinfo(struct sock
*sk
, char __user
*optval
, int optlen
) {
2190 struct sctp_rtoinfo rtoinfo
;
2191 struct sctp_association
*asoc
;
2193 if (optlen
!= sizeof (struct sctp_rtoinfo
))
2196 if (copy_from_user(&rtoinfo
, optval
, optlen
))
2199 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
2201 /* Set the values to the specific association */
2202 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
2206 if (rtoinfo
.srto_initial
!= 0)
2208 msecs_to_jiffies(rtoinfo
.srto_initial
);
2209 if (rtoinfo
.srto_max
!= 0)
2210 asoc
->rto_max
= msecs_to_jiffies(rtoinfo
.srto_max
);
2211 if (rtoinfo
.srto_min
!= 0)
2212 asoc
->rto_min
= msecs_to_jiffies(rtoinfo
.srto_min
);
2214 /* If there is no association or the association-id = 0
2215 * set the values to the endpoint.
2217 struct sctp_sock
*sp
= sctp_sk(sk
);
2219 if (rtoinfo
.srto_initial
!= 0)
2220 sp
->rtoinfo
.srto_initial
= rtoinfo
.srto_initial
;
2221 if (rtoinfo
.srto_max
!= 0)
2222 sp
->rtoinfo
.srto_max
= rtoinfo
.srto_max
;
2223 if (rtoinfo
.srto_min
!= 0)
2224 sp
->rtoinfo
.srto_min
= rtoinfo
.srto_min
;
2232 * 7.1.2 SCTP_ASSOCINFO
2234 * This option is used to tune the the maximum retransmission attempts
2235 * of the association.
2236 * Returns an error if the new association retransmission value is
2237 * greater than the sum of the retransmission value of the peer.
2238 * See [SCTP] for more information.
2241 static int sctp_setsockopt_associnfo(struct sock
*sk
, char __user
*optval
, int optlen
)
2244 struct sctp_assocparams assocparams
;
2245 struct sctp_association
*asoc
;
2247 if (optlen
!= sizeof(struct sctp_assocparams
))
2249 if (copy_from_user(&assocparams
, optval
, optlen
))
2252 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
2254 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
2257 /* Set the values to the specific association */
2259 if (assocparams
.sasoc_asocmaxrxt
!= 0)
2260 asoc
->max_retrans
= assocparams
.sasoc_asocmaxrxt
;
2261 if (assocparams
.sasoc_cookie_life
!= 0) {
2262 asoc
->cookie_life
.tv_sec
=
2263 assocparams
.sasoc_cookie_life
/ 1000;
2264 asoc
->cookie_life
.tv_usec
=
2265 (assocparams
.sasoc_cookie_life
% 1000)
2269 /* Set the values to the endpoint */
2270 struct sctp_sock
*sp
= sctp_sk(sk
);
2272 if (assocparams
.sasoc_asocmaxrxt
!= 0)
2273 sp
->assocparams
.sasoc_asocmaxrxt
=
2274 assocparams
.sasoc_asocmaxrxt
;
2275 if (assocparams
.sasoc_cookie_life
!= 0)
2276 sp
->assocparams
.sasoc_cookie_life
=
2277 assocparams
.sasoc_cookie_life
;
2283 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2285 * This socket option is a boolean flag which turns on or off mapped V4
2286 * addresses. If this option is turned on and the socket is type
2287 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2288 * If this option is turned off, then no mapping will be done of V4
2289 * addresses and a user will receive both PF_INET6 and PF_INET type
2290 * addresses on the socket.
2292 static int sctp_setsockopt_mappedv4(struct sock
*sk
, char __user
*optval
, int optlen
)
2295 struct sctp_sock
*sp
= sctp_sk(sk
);
2297 if (optlen
< sizeof(int))
2299 if (get_user(val
, (int __user
*)optval
))
2310 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
2312 * This socket option specifies the maximum size to put in any outgoing
2313 * SCTP chunk. If a message is larger than this size it will be
2314 * fragmented by SCTP into the specified size. Note that the underlying
2315 * SCTP implementation may fragment into smaller sized chunks when the
2316 * PMTU of the underlying association is smaller than the value set by
2319 static int sctp_setsockopt_maxseg(struct sock
*sk
, char __user
*optval
, int optlen
)
2321 struct sctp_association
*asoc
;
2322 struct list_head
*pos
;
2323 struct sctp_sock
*sp
= sctp_sk(sk
);
2326 if (optlen
< sizeof(int))
2328 if (get_user(val
, (int __user
*)optval
))
2330 if ((val
!= 0) && ((val
< 8) || (val
> SCTP_MAX_CHUNK_LEN
)))
2332 sp
->user_frag
= val
;
2334 /* Update the frag_point of the existing associations. */
2335 list_for_each(pos
, &(sp
->ep
->asocs
)) {
2336 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
2337 asoc
->frag_point
= sctp_frag_point(sp
, asoc
->pmtu
);
2345 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2347 * Requests that the peer mark the enclosed address as the association
2348 * primary. The enclosed address must be one of the association's
2349 * locally bound addresses. The following structure is used to make a
2350 * set primary request:
2352 static int sctp_setsockopt_peer_primary_addr(struct sock
*sk
, char __user
*optval
,
2355 struct sctp_sock
*sp
;
2356 struct sctp_endpoint
*ep
;
2357 struct sctp_association
*asoc
= NULL
;
2358 struct sctp_setpeerprim prim
;
2359 struct sctp_chunk
*chunk
;
2365 if (!sctp_addip_enable
)
2368 if (optlen
!= sizeof(struct sctp_setpeerprim
))
2371 if (copy_from_user(&prim
, optval
, optlen
))
2374 asoc
= sctp_id2assoc(sk
, prim
.sspp_assoc_id
);
2378 if (!asoc
->peer
.asconf_capable
)
2381 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_SET_PRIMARY
)
2384 if (!sctp_state(asoc
, ESTABLISHED
))
2387 if (!sctp_assoc_lookup_laddr(asoc
, (union sctp_addr
*)&prim
.sspp_addr
))
2388 return -EADDRNOTAVAIL
;
2390 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2391 chunk
= sctp_make_asconf_set_prim(asoc
,
2392 (union sctp_addr
*)&prim
.sspp_addr
);
2396 err
= sctp_send_asconf(asoc
, chunk
);
2398 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2403 static int sctp_setsockopt_adaption_layer(struct sock
*sk
, char __user
*optval
,
2406 struct sctp_setadaption adaption
;
2408 if (optlen
!= sizeof(struct sctp_setadaption
))
2410 if (copy_from_user(&adaption
, optval
, optlen
))
2413 sctp_sk(sk
)->adaption_ind
= adaption
.ssb_adaption_ind
;
2418 /* API 6.2 setsockopt(), getsockopt()
2420 * Applications use setsockopt() and getsockopt() to set or retrieve
2421 * socket options. Socket options are used to change the default
2422 * behavior of sockets calls. They are described in Section 7.
2426 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
2427 * int __user *optlen);
2428 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
2431 * sd - the socket descript.
2432 * level - set to IPPROTO_SCTP for all SCTP options.
2433 * optname - the option name.
2434 * optval - the buffer to store the value of the option.
2435 * optlen - the size of the buffer.
2437 SCTP_STATIC
int sctp_setsockopt(struct sock
*sk
, int level
, int optname
,
2438 char __user
*optval
, int optlen
)
2442 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
2445 /* I can hardly begin to describe how wrong this is. This is
2446 * so broken as to be worse than useless. The API draft
2447 * REALLY is NOT helpful here... I am not convinced that the
2448 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
2449 * are at all well-founded.
2451 if (level
!= SOL_SCTP
) {
2452 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
2453 retval
= af
->setsockopt(sk
, level
, optname
, optval
, optlen
);
2460 case SCTP_SOCKOPT_BINDX_ADD
:
2461 /* 'optlen' is the size of the addresses buffer. */
2462 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
2463 optlen
, SCTP_BINDX_ADD_ADDR
);
2466 case SCTP_SOCKOPT_BINDX_REM
:
2467 /* 'optlen' is the size of the addresses buffer. */
2468 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
2469 optlen
, SCTP_BINDX_REM_ADDR
);
2472 case SCTP_SOCKOPT_CONNECTX
:
2473 /* 'optlen' is the size of the addresses buffer. */
2474 retval
= sctp_setsockopt_connectx(sk
, (struct sockaddr __user
*)optval
,
2478 case SCTP_DISABLE_FRAGMENTS
:
2479 retval
= sctp_setsockopt_disable_fragments(sk
, optval
, optlen
);
2483 retval
= sctp_setsockopt_events(sk
, optval
, optlen
);
2486 case SCTP_AUTOCLOSE
:
2487 retval
= sctp_setsockopt_autoclose(sk
, optval
, optlen
);
2490 case SCTP_PEER_ADDR_PARAMS
:
2491 retval
= sctp_setsockopt_peer_addr_params(sk
, optval
, optlen
);
2495 retval
= sctp_setsockopt_initmsg(sk
, optval
, optlen
);
2497 case SCTP_DEFAULT_SEND_PARAM
:
2498 retval
= sctp_setsockopt_default_send_param(sk
, optval
,
2501 case SCTP_PRIMARY_ADDR
:
2502 retval
= sctp_setsockopt_primary_addr(sk
, optval
, optlen
);
2504 case SCTP_SET_PEER_PRIMARY_ADDR
:
2505 retval
= sctp_setsockopt_peer_primary_addr(sk
, optval
, optlen
);
2508 retval
= sctp_setsockopt_nodelay(sk
, optval
, optlen
);
2511 retval
= sctp_setsockopt_rtoinfo(sk
, optval
, optlen
);
2513 case SCTP_ASSOCINFO
:
2514 retval
= sctp_setsockopt_associnfo(sk
, optval
, optlen
);
2516 case SCTP_I_WANT_MAPPED_V4_ADDR
:
2517 retval
= sctp_setsockopt_mappedv4(sk
, optval
, optlen
);
2520 retval
= sctp_setsockopt_maxseg(sk
, optval
, optlen
);
2522 case SCTP_ADAPTION_LAYER
:
2523 retval
= sctp_setsockopt_adaption_layer(sk
, optval
, optlen
);
2527 retval
= -ENOPROTOOPT
;
2531 sctp_release_sock(sk
);
2537 /* API 3.1.6 connect() - UDP Style Syntax
2539 * An application may use the connect() call in the UDP model to initiate an
2540 * association without sending data.
2544 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
2546 * sd: the socket descriptor to have a new association added to.
2548 * nam: the address structure (either struct sockaddr_in or struct
2549 * sockaddr_in6 defined in RFC2553 [7]).
2551 * len: the size of the address.
2553 SCTP_STATIC
int sctp_connect(struct sock
*sk
, struct sockaddr
*addr
,
2561 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
2562 __FUNCTION__
, sk
, addr
, addr_len
);
2564 /* Validate addr_len before calling common connect/connectx routine. */
2565 af
= sctp_get_af_specific(addr
->sa_family
);
2566 if (!af
|| addr_len
< af
->sockaddr_len
) {
2569 /* Pass correct addr len to common routine (so it knows there
2570 * is only one address being passed.
2572 err
= __sctp_connect(sk
, addr
, af
->sockaddr_len
);
2575 sctp_release_sock(sk
);
2579 /* FIXME: Write comments. */
2580 SCTP_STATIC
int sctp_disconnect(struct sock
*sk
, int flags
)
2582 return -EOPNOTSUPP
; /* STUB */
2585 /* 4.1.4 accept() - TCP Style Syntax
2587 * Applications use accept() call to remove an established SCTP
2588 * association from the accept queue of the endpoint. A new socket
2589 * descriptor will be returned from accept() to represent the newly
2590 * formed association.
2592 SCTP_STATIC
struct sock
*sctp_accept(struct sock
*sk
, int flags
, int *err
)
2594 struct sctp_sock
*sp
;
2595 struct sctp_endpoint
*ep
;
2596 struct sock
*newsk
= NULL
;
2597 struct sctp_association
*asoc
;
2606 if (!sctp_style(sk
, TCP
)) {
2607 error
= -EOPNOTSUPP
;
2611 if (!sctp_sstate(sk
, LISTENING
)) {
2616 timeo
= sock_rcvtimeo(sk
, sk
->sk_socket
->file
->f_flags
& O_NONBLOCK
);
2618 error
= sctp_wait_for_accept(sk
, timeo
);
2622 /* We treat the list of associations on the endpoint as the accept
2623 * queue and pick the first association on the list.
2625 asoc
= list_entry(ep
->asocs
.next
, struct sctp_association
, asocs
);
2627 newsk
= sp
->pf
->create_accept_sk(sk
, asoc
);
2633 /* Populate the fields of the newsk from the oldsk and migrate the
2634 * asoc to the newsk.
2636 sctp_sock_migrate(sk
, newsk
, asoc
, SCTP_SOCKET_TCP
);
2639 sctp_release_sock(sk
);
2644 /* The SCTP ioctl handler. */
2645 SCTP_STATIC
int sctp_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
2647 return -ENOIOCTLCMD
;
2650 /* This is the function which gets called during socket creation to
2651 * initialized the SCTP-specific portion of the sock.
2652 * The sock structure should already be zero-filled memory.
2654 SCTP_STATIC
int sctp_init_sock(struct sock
*sk
)
2656 struct sctp_endpoint
*ep
;
2657 struct sctp_sock
*sp
;
2659 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk
);
2663 /* Initialize the SCTP per socket area. */
2664 switch (sk
->sk_type
) {
2665 case SOCK_SEQPACKET
:
2666 sp
->type
= SCTP_SOCKET_UDP
;
2669 sp
->type
= SCTP_SOCKET_TCP
;
2672 return -ESOCKTNOSUPPORT
;
2675 /* Initialize default send parameters. These parameters can be
2676 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
2678 sp
->default_stream
= 0;
2679 sp
->default_ppid
= 0;
2680 sp
->default_flags
= 0;
2681 sp
->default_context
= 0;
2682 sp
->default_timetolive
= 0;
2684 /* Initialize default setup parameters. These parameters
2685 * can be modified with the SCTP_INITMSG socket option or
2686 * overridden by the SCTP_INIT CMSG.
2688 sp
->initmsg
.sinit_num_ostreams
= sctp_max_outstreams
;
2689 sp
->initmsg
.sinit_max_instreams
= sctp_max_instreams
;
2690 sp
->initmsg
.sinit_max_attempts
= sctp_max_retrans_init
;
2691 sp
->initmsg
.sinit_max_init_timeo
= jiffies_to_msecs(sctp_rto_max
);
2693 /* Initialize default RTO related parameters. These parameters can
2694 * be modified for with the SCTP_RTOINFO socket option.
2696 sp
->rtoinfo
.srto_initial
= jiffies_to_msecs(sctp_rto_initial
);
2697 sp
->rtoinfo
.srto_max
= jiffies_to_msecs(sctp_rto_max
);
2698 sp
->rtoinfo
.srto_min
= jiffies_to_msecs(sctp_rto_min
);
2700 /* Initialize default association related parameters. These parameters
2701 * can be modified with the SCTP_ASSOCINFO socket option.
2703 sp
->assocparams
.sasoc_asocmaxrxt
= sctp_max_retrans_association
;
2704 sp
->assocparams
.sasoc_number_peer_destinations
= 0;
2705 sp
->assocparams
.sasoc_peer_rwnd
= 0;
2706 sp
->assocparams
.sasoc_local_rwnd
= 0;
2707 sp
->assocparams
.sasoc_cookie_life
=
2708 jiffies_to_msecs(sctp_valid_cookie_life
);
2710 /* Initialize default event subscriptions. By default, all the
2713 memset(&sp
->subscribe
, 0, sizeof(struct sctp_event_subscribe
));
2715 /* Default Peer Address Parameters. These defaults can
2716 * be modified via SCTP_PEER_ADDR_PARAMS
2718 sp
->paddrparam
.spp_hbinterval
= jiffies_to_msecs(sctp_hb_interval
);
2719 sp
->paddrparam
.spp_pathmaxrxt
= sctp_max_retrans_path
;
2721 /* If enabled no SCTP message fragmentation will be performed.
2722 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
2724 sp
->disable_fragments
= 0;
2726 /* Turn on/off any Nagle-like algorithm. */
2729 /* Enable by default. */
2732 /* Auto-close idle associations after the configured
2733 * number of seconds. A value of 0 disables this
2734 * feature. Configure through the SCTP_AUTOCLOSE socket option,
2735 * for UDP-style sockets only.
2739 /* User specified fragmentation limit. */
2742 sp
->adaption_ind
= 0;
2744 sp
->pf
= sctp_get_pf_specific(sk
->sk_family
);
2746 /* Control variables for partial data delivery. */
2748 skb_queue_head_init(&sp
->pd_lobby
);
2750 /* Create a per socket endpoint structure. Even if we
2751 * change the data structure relationships, this may still
2752 * be useful for storing pre-connect address information.
2754 ep
= sctp_endpoint_new(sk
, GFP_KERNEL
);
2761 SCTP_DBG_OBJCNT_INC(sock
);
2765 /* Cleanup any SCTP per socket resources. */
2766 SCTP_STATIC
int sctp_destroy_sock(struct sock
*sk
)
2768 struct sctp_endpoint
*ep
;
2770 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk
);
2772 /* Release our hold on the endpoint. */
2773 ep
= sctp_sk(sk
)->ep
;
2774 sctp_endpoint_free(ep
);
2779 /* API 4.1.7 shutdown() - TCP Style Syntax
2780 * int shutdown(int socket, int how);
2782 * sd - the socket descriptor of the association to be closed.
2783 * how - Specifies the type of shutdown. The values are
2786 * Disables further receive operations. No SCTP
2787 * protocol action is taken.
2789 * Disables further send operations, and initiates
2790 * the SCTP shutdown sequence.
2792 * Disables further send and receive operations
2793 * and initiates the SCTP shutdown sequence.
2795 SCTP_STATIC
void sctp_shutdown(struct sock
*sk
, int how
)
2797 struct sctp_endpoint
*ep
;
2798 struct sctp_association
*asoc
;
2800 if (!sctp_style(sk
, TCP
))
2803 if (how
& SEND_SHUTDOWN
) {
2804 ep
= sctp_sk(sk
)->ep
;
2805 if (!list_empty(&ep
->asocs
)) {
2806 asoc
= list_entry(ep
->asocs
.next
,
2807 struct sctp_association
, asocs
);
2808 sctp_primitive_SHUTDOWN(asoc
, NULL
);
2813 /* 7.2.1 Association Status (SCTP_STATUS)
2815 * Applications can retrieve current status information about an
2816 * association, including association state, peer receiver window size,
2817 * number of unacked data chunks, and number of data chunks pending
2818 * receipt. This information is read-only.
2820 static int sctp_getsockopt_sctp_status(struct sock
*sk
, int len
,
2821 char __user
*optval
,
2824 struct sctp_status status
;
2825 struct sctp_association
*asoc
= NULL
;
2826 struct sctp_transport
*transport
;
2827 sctp_assoc_t associd
;
2830 if (len
!= sizeof(status
)) {
2835 if (copy_from_user(&status
, optval
, sizeof(status
))) {
2840 associd
= status
.sstat_assoc_id
;
2841 asoc
= sctp_id2assoc(sk
, associd
);
2847 transport
= asoc
->peer
.primary_path
;
2849 status
.sstat_assoc_id
= sctp_assoc2id(asoc
);
2850 status
.sstat_state
= asoc
->state
;
2851 status
.sstat_rwnd
= asoc
->peer
.rwnd
;
2852 status
.sstat_unackdata
= asoc
->unack_data
;
2854 status
.sstat_penddata
= sctp_tsnmap_pending(&asoc
->peer
.tsn_map
);
2855 status
.sstat_instrms
= asoc
->c
.sinit_max_instreams
;
2856 status
.sstat_outstrms
= asoc
->c
.sinit_num_ostreams
;
2857 status
.sstat_fragmentation_point
= asoc
->frag_point
;
2858 status
.sstat_primary
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
2859 memcpy(&status
.sstat_primary
.spinfo_address
,
2860 &(transport
->ipaddr
), sizeof(union sctp_addr
));
2861 /* Map ipv4 address into v4-mapped-on-v6 address. */
2862 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
2863 (union sctp_addr
*)&status
.sstat_primary
.spinfo_address
);
2864 status
.sstat_primary
.spinfo_state
= transport
->state
;
2865 status
.sstat_primary
.spinfo_cwnd
= transport
->cwnd
;
2866 status
.sstat_primary
.spinfo_srtt
= transport
->srtt
;
2867 status
.sstat_primary
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
2868 status
.sstat_primary
.spinfo_mtu
= transport
->pmtu
;
2870 if (status
.sstat_primary
.spinfo_state
== SCTP_UNKNOWN
)
2871 status
.sstat_primary
.spinfo_state
= SCTP_ACTIVE
;
2873 if (put_user(len
, optlen
)) {
2878 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
2879 len
, status
.sstat_state
, status
.sstat_rwnd
,
2880 status
.sstat_assoc_id
);
2882 if (copy_to_user(optval
, &status
, len
)) {
2892 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
2894 * Applications can retrieve information about a specific peer address
2895 * of an association, including its reachability state, congestion
2896 * window, and retransmission timer values. This information is
2899 static int sctp_getsockopt_peer_addr_info(struct sock
*sk
, int len
,
2900 char __user
*optval
,
2903 struct sctp_paddrinfo pinfo
;
2904 struct sctp_transport
*transport
;
2907 if (len
!= sizeof(pinfo
)) {
2912 if (copy_from_user(&pinfo
, optval
, sizeof(pinfo
))) {
2917 transport
= sctp_addr_id2transport(sk
, &pinfo
.spinfo_address
,
2918 pinfo
.spinfo_assoc_id
);
2922 pinfo
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
2923 pinfo
.spinfo_state
= transport
->state
;
2924 pinfo
.spinfo_cwnd
= transport
->cwnd
;
2925 pinfo
.spinfo_srtt
= transport
->srtt
;
2926 pinfo
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
2927 pinfo
.spinfo_mtu
= transport
->pmtu
;
2929 if (pinfo
.spinfo_state
== SCTP_UNKNOWN
)
2930 pinfo
.spinfo_state
= SCTP_ACTIVE
;
2932 if (put_user(len
, optlen
)) {
2937 if (copy_to_user(optval
, &pinfo
, len
)) {
2946 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2948 * This option is a on/off flag. If enabled no SCTP message
2949 * fragmentation will be performed. Instead if a message being sent
2950 * exceeds the current PMTU size, the message will NOT be sent and
2951 * instead a error will be indicated to the user.
2953 static int sctp_getsockopt_disable_fragments(struct sock
*sk
, int len
,
2954 char __user
*optval
, int __user
*optlen
)
2958 if (len
< sizeof(int))
2962 val
= (sctp_sk(sk
)->disable_fragments
== 1);
2963 if (put_user(len
, optlen
))
2965 if (copy_to_user(optval
, &val
, len
))
2970 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
2972 * This socket option is used to specify various notifications and
2973 * ancillary data the user wishes to receive.
2975 static int sctp_getsockopt_events(struct sock
*sk
, int len
, char __user
*optval
,
2978 if (len
!= sizeof(struct sctp_event_subscribe
))
2980 if (copy_to_user(optval
, &sctp_sk(sk
)->subscribe
, len
))
2985 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2987 * This socket option is applicable to the UDP-style socket only. When
2988 * set it will cause associations that are idle for more than the
2989 * specified number of seconds to automatically close. An association
2990 * being idle is defined an association that has NOT sent or received
2991 * user data. The special value of '0' indicates that no automatic
2992 * close of any associations should be performed. The option expects an
2993 * integer defining the number of seconds of idle time before an
2994 * association is closed.
2996 static int sctp_getsockopt_autoclose(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
2998 /* Applicable to UDP-style socket only */
2999 if (sctp_style(sk
, TCP
))
3001 if (len
!= sizeof(int))
3003 if (copy_to_user(optval
, &sctp_sk(sk
)->autoclose
, len
))
3008 /* Helper routine to branch off an association to a new socket. */
3009 SCTP_STATIC
int sctp_do_peeloff(struct sctp_association
*asoc
,
3010 struct socket
**sockp
)
3012 struct sock
*sk
= asoc
->base
.sk
;
3013 struct socket
*sock
;
3016 /* An association cannot be branched off from an already peeled-off
3017 * socket, nor is this supported for tcp style sockets.
3019 if (!sctp_style(sk
, UDP
))
3022 /* Create a new socket. */
3023 err
= sock_create(sk
->sk_family
, SOCK_SEQPACKET
, IPPROTO_SCTP
, &sock
);
3027 /* Populate the fields of the newsk from the oldsk and migrate the
3028 * asoc to the newsk.
3030 sctp_sock_migrate(sk
, sock
->sk
, asoc
, SCTP_SOCKET_UDP_HIGH_BANDWIDTH
);
3036 static int sctp_getsockopt_peeloff(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
3038 sctp_peeloff_arg_t peeloff
;
3039 struct socket
*newsock
;
3041 struct sctp_association
*asoc
;
3043 if (len
!= sizeof(sctp_peeloff_arg_t
))
3045 if (copy_from_user(&peeloff
, optval
, len
))
3048 asoc
= sctp_id2assoc(sk
, peeloff
.associd
);
3054 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __FUNCTION__
, sk
, asoc
);
3056 retval
= sctp_do_peeloff(asoc
, &newsock
);
3060 /* Map the socket to an unused fd that can be returned to the user. */
3061 retval
= sock_map_fd(newsock
);
3063 sock_release(newsock
);
3067 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3068 __FUNCTION__
, sk
, asoc
, newsock
->sk
, retval
);
3070 /* Return the fd mapped to the new socket. */
3071 peeloff
.sd
= retval
;
3072 if (copy_to_user(optval
, &peeloff
, len
))
3079 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
3081 * Applications can enable or disable heartbeats for any peer address of
3082 * an association, modify an address's heartbeat interval, force a
3083 * heartbeat to be sent immediately, and adjust the address's maximum
3084 * number of retransmissions sent before an address is considered
3085 * unreachable. The following structure is used to access and modify an
3086 * address's parameters:
3088 * struct sctp_paddrparams {
3089 * sctp_assoc_t spp_assoc_id;
3090 * struct sockaddr_storage spp_address;
3091 * uint32_t spp_hbinterval;
3092 * uint16_t spp_pathmaxrxt;
3095 * spp_assoc_id - (UDP style socket) This is filled in the application,
3096 * and identifies the association for this query.
3097 * spp_address - This specifies which address is of interest.
3098 * spp_hbinterval - This contains the value of the heartbeat interval,
3099 * in milliseconds. A value of 0, when modifying the
3100 * parameter, specifies that the heartbeat on this
3101 * address should be disabled. A value of UINT32_MAX
3102 * (4294967295), when modifying the parameter,
3103 * specifies that a heartbeat should be sent
3104 * immediately to the peer address, and the current
3105 * interval should remain unchanged.
3106 * spp_pathmaxrxt - This contains the maximum number of
3107 * retransmissions before this address shall be
3108 * considered unreachable.
3110 static int sctp_getsockopt_peer_addr_params(struct sock
*sk
, int len
,
3111 char __user
*optval
, int __user
*optlen
)
3113 struct sctp_paddrparams params
;
3114 struct sctp_transport
*trans
;
3116 if (len
!= sizeof(struct sctp_paddrparams
))
3118 if (copy_from_user(¶ms
, optval
, len
))
3121 /* If no association id is specified retrieve the default value
3122 * for the endpoint that will be used for all future associations
3124 if (!params
.spp_assoc_id
&&
3125 sctp_is_any(( union sctp_addr
*)¶ms
.spp_address
)) {
3126 params
.spp_hbinterval
= sctp_sk(sk
)->paddrparam
.spp_hbinterval
;
3127 params
.spp_pathmaxrxt
= sctp_sk(sk
)->paddrparam
.spp_pathmaxrxt
;
3132 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
3133 params
.spp_assoc_id
);
3137 /* The value of the heartbeat interval, in milliseconds. A value of 0,
3138 * when modifying the parameter, specifies that the heartbeat on this
3139 * address should be disabled.
3141 if (!trans
->hb_allowed
)
3142 params
.spp_hbinterval
= 0;
3144 params
.spp_hbinterval
= jiffies_to_msecs(trans
->hb_interval
);
3146 /* spp_pathmaxrxt contains the maximum number of retransmissions
3147 * before this address shall be considered unreachable.
3149 params
.spp_pathmaxrxt
= trans
->max_retrans
;
3152 if (copy_to_user(optval
, ¶ms
, len
))
3155 if (put_user(len
, optlen
))
3161 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
3163 * Applications can specify protocol parameters for the default association
3164 * initialization. The option name argument to setsockopt() and getsockopt()
3167 * Setting initialization parameters is effective only on an unconnected
3168 * socket (for UDP-style sockets only future associations are effected
3169 * by the change). With TCP-style sockets, this option is inherited by
3170 * sockets derived from a listener socket.
3172 static int sctp_getsockopt_initmsg(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
3174 if (len
!= sizeof(struct sctp_initmsg
))
3176 if (copy_to_user(optval
, &sctp_sk(sk
)->initmsg
, len
))
3181 static int sctp_getsockopt_peer_addrs_num_old(struct sock
*sk
, int len
,
3182 char __user
*optval
,
3186 struct sctp_association
*asoc
;
3187 struct list_head
*pos
;
3190 if (len
!= sizeof(sctp_assoc_t
))
3193 if (copy_from_user(&id
, optval
, sizeof(sctp_assoc_t
)))
3196 /* For UDP-style sockets, id specifies the association to query. */
3197 asoc
= sctp_id2assoc(sk
, id
);
3201 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
3209 * Old API for getting list of peer addresses. Does not work for 32-bit
3210 * programs running on a 64-bit kernel
3212 static int sctp_getsockopt_peer_addrs_old(struct sock
*sk
, int len
,
3213 char __user
*optval
,
3216 struct sctp_association
*asoc
;
3217 struct list_head
*pos
;
3219 struct sctp_getaddrs_old getaddrs
;
3220 struct sctp_transport
*from
;
3222 union sctp_addr temp
;
3223 struct sctp_sock
*sp
= sctp_sk(sk
);
3226 if (len
!= sizeof(struct sctp_getaddrs_old
))
3229 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs_old
)))
3232 if (getaddrs
.addr_num
<= 0) return -EINVAL
;
3234 /* For UDP-style sockets, id specifies the association to query. */
3235 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
3239 to
= (void __user
*)getaddrs
.addrs
;
3240 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
3241 from
= list_entry(pos
, struct sctp_transport
, transports
);
3242 memcpy(&temp
, &from
->ipaddr
, sizeof(temp
));
3243 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
3244 addrlen
= sctp_get_af_specific(sk
->sk_family
)->sockaddr_len
;
3245 temp
.v4
.sin_port
= htons(temp
.v4
.sin_port
);
3246 if (copy_to_user(to
, &temp
, addrlen
))
3250 if (cnt
>= getaddrs
.addr_num
) break;
3252 getaddrs
.addr_num
= cnt
;
3253 if (copy_to_user(optval
, &getaddrs
, sizeof(struct sctp_getaddrs_old
)))
3259 static int sctp_getsockopt_peer_addrs(struct sock
*sk
, int len
,
3260 char __user
*optval
, int __user
*optlen
)
3262 struct sctp_association
*asoc
;
3263 struct list_head
*pos
;
3265 struct sctp_getaddrs getaddrs
;
3266 struct sctp_transport
*from
;
3268 union sctp_addr temp
;
3269 struct sctp_sock
*sp
= sctp_sk(sk
);
3274 if (len
< sizeof(struct sctp_getaddrs
))
3277 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
3280 /* For UDP-style sockets, id specifies the association to query. */
3281 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
3285 to
= optval
+ offsetof(struct sctp_getaddrs
,addrs
);
3286 space_left
= len
- sizeof(struct sctp_getaddrs
) -
3287 offsetof(struct sctp_getaddrs
,addrs
);
3289 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
3290 from
= list_entry(pos
, struct sctp_transport
, transports
);
3291 memcpy(&temp
, &from
->ipaddr
, sizeof(temp
));
3292 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
3293 addrlen
= sctp_get_af_specific(sk
->sk_family
)->sockaddr_len
;
3294 if(space_left
< addrlen
)
3296 temp
.v4
.sin_port
= htons(temp
.v4
.sin_port
);
3297 if (copy_to_user(to
, &temp
, addrlen
))
3301 space_left
-= addrlen
;
3304 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
))
3306 bytes_copied
= ((char __user
*)to
) - optval
;
3307 if (put_user(bytes_copied
, optlen
))
3313 static int sctp_getsockopt_local_addrs_num_old(struct sock
*sk
, int len
,
3314 char __user
*optval
,
3318 struct sctp_bind_addr
*bp
;
3319 struct sctp_association
*asoc
;
3320 struct list_head
*pos
;
3321 struct sctp_sockaddr_entry
*addr
;
3322 rwlock_t
*addr_lock
;
3323 unsigned long flags
;
3326 if (len
!= sizeof(sctp_assoc_t
))
3329 if (copy_from_user(&id
, optval
, sizeof(sctp_assoc_t
)))
3333 * For UDP-style sockets, id specifies the association to query.
3334 * If the id field is set to the value '0' then the locally bound
3335 * addresses are returned without regard to any particular
3339 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
3340 addr_lock
= &sctp_sk(sk
)->ep
->base
.addr_lock
;
3342 asoc
= sctp_id2assoc(sk
, id
);
3345 bp
= &asoc
->base
.bind_addr
;
3346 addr_lock
= &asoc
->base
.addr_lock
;
3349 sctp_read_lock(addr_lock
);
3351 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
3352 * addresses from the global local address list.
3354 if (sctp_list_single_entry(&bp
->address_list
)) {
3355 addr
= list_entry(bp
->address_list
.next
,
3356 struct sctp_sockaddr_entry
, list
);
3357 if (sctp_is_any(&addr
->a
)) {
3358 sctp_spin_lock_irqsave(&sctp_local_addr_lock
, flags
);
3359 list_for_each(pos
, &sctp_local_addr_list
) {
3360 addr
= list_entry(pos
,
3361 struct sctp_sockaddr_entry
,
3363 if ((PF_INET
== sk
->sk_family
) &&
3364 (AF_INET6
== addr
->a
.sa
.sa_family
))
3368 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock
,
3376 list_for_each(pos
, &bp
->address_list
) {
3381 sctp_read_unlock(addr_lock
);
3385 /* Helper function that copies local addresses to user and returns the number
3386 * of addresses copied.
3388 static int sctp_copy_laddrs_to_user_old(struct sock
*sk
, __u16 port
, int max_addrs
,
3391 struct list_head
*pos
;
3392 struct sctp_sockaddr_entry
*addr
;
3393 unsigned long flags
;
3394 union sctp_addr temp
;
3398 sctp_spin_lock_irqsave(&sctp_local_addr_lock
, flags
);
3399 list_for_each(pos
, &sctp_local_addr_list
) {
3400 addr
= list_entry(pos
, struct sctp_sockaddr_entry
, list
);
3401 if ((PF_INET
== sk
->sk_family
) &&
3402 (AF_INET6
== addr
->a
.sa
.sa_family
))
3404 memcpy(&temp
, &addr
->a
, sizeof(temp
));
3405 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
3407 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
3408 temp
.v4
.sin_port
= htons(port
);
3409 if (copy_to_user(to
, &temp
, addrlen
)) {
3410 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock
,
3416 if (cnt
>= max_addrs
) break;
3418 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock
, flags
);
3423 static int sctp_copy_laddrs_to_user(struct sock
*sk
, __u16 port
,
3424 void __user
**to
, size_t space_left
)
3426 struct list_head
*pos
;
3427 struct sctp_sockaddr_entry
*addr
;
3428 unsigned long flags
;
3429 union sctp_addr temp
;
3433 sctp_spin_lock_irqsave(&sctp_local_addr_lock
, flags
);
3434 list_for_each(pos
, &sctp_local_addr_list
) {
3435 addr
= list_entry(pos
, struct sctp_sockaddr_entry
, list
);
3436 if ((PF_INET
== sk
->sk_family
) &&
3437 (AF_INET6
== addr
->a
.sa
.sa_family
))
3439 memcpy(&temp
, &addr
->a
, sizeof(temp
));
3440 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
3442 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
3443 if(space_left
<addrlen
)
3445 temp
.v4
.sin_port
= htons(port
);
3446 if (copy_to_user(*to
, &temp
, addrlen
)) {
3447 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock
,
3453 space_left
-= addrlen
;
3455 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock
, flags
);
3460 /* Old API for getting list of local addresses. Does not work for 32-bit
3461 * programs running on a 64-bit kernel
3463 static int sctp_getsockopt_local_addrs_old(struct sock
*sk
, int len
,
3464 char __user
*optval
, int __user
*optlen
)
3466 struct sctp_bind_addr
*bp
;
3467 struct sctp_association
*asoc
;
3468 struct list_head
*pos
;
3470 struct sctp_getaddrs_old getaddrs
;
3471 struct sctp_sockaddr_entry
*addr
;
3473 union sctp_addr temp
;
3474 struct sctp_sock
*sp
= sctp_sk(sk
);
3476 rwlock_t
*addr_lock
;
3479 if (len
!= sizeof(struct sctp_getaddrs_old
))
3482 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs_old
)))
3485 if (getaddrs
.addr_num
<= 0) return -EINVAL
;
3487 * For UDP-style sockets, id specifies the association to query.
3488 * If the id field is set to the value '0' then the locally bound
3489 * addresses are returned without regard to any particular
3492 if (0 == getaddrs
.assoc_id
) {
3493 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
3494 addr_lock
= &sctp_sk(sk
)->ep
->base
.addr_lock
;
3496 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
3499 bp
= &asoc
->base
.bind_addr
;
3500 addr_lock
= &asoc
->base
.addr_lock
;
3503 to
= getaddrs
.addrs
;
3505 sctp_read_lock(addr_lock
);
3507 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
3508 * addresses from the global local address list.
3510 if (sctp_list_single_entry(&bp
->address_list
)) {
3511 addr
= list_entry(bp
->address_list
.next
,
3512 struct sctp_sockaddr_entry
, list
);
3513 if (sctp_is_any(&addr
->a
)) {
3514 cnt
= sctp_copy_laddrs_to_user_old(sk
, bp
->port
,
3525 list_for_each(pos
, &bp
->address_list
) {
3526 addr
= list_entry(pos
, struct sctp_sockaddr_entry
, list
);
3527 memcpy(&temp
, &addr
->a
, sizeof(temp
));
3528 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
3529 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
3530 temp
.v4
.sin_port
= htons(temp
.v4
.sin_port
);
3531 if (copy_to_user(to
, &temp
, addrlen
)) {
3537 if (cnt
>= getaddrs
.addr_num
) break;
3541 getaddrs
.addr_num
= cnt
;
3542 if (copy_to_user(optval
, &getaddrs
, sizeof(struct sctp_getaddrs_old
)))
3546 sctp_read_unlock(addr_lock
);
3550 static int sctp_getsockopt_local_addrs(struct sock
*sk
, int len
,
3551 char __user
*optval
, int __user
*optlen
)
3553 struct sctp_bind_addr
*bp
;
3554 struct sctp_association
*asoc
;
3555 struct list_head
*pos
;
3557 struct sctp_getaddrs getaddrs
;
3558 struct sctp_sockaddr_entry
*addr
;
3560 union sctp_addr temp
;
3561 struct sctp_sock
*sp
= sctp_sk(sk
);
3563 rwlock_t
*addr_lock
;
3568 if (len
<= sizeof(struct sctp_getaddrs
))
3571 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
3575 * For UDP-style sockets, id specifies the association to query.
3576 * If the id field is set to the value '0' then the locally bound
3577 * addresses are returned without regard to any particular
3580 if (0 == getaddrs
.assoc_id
) {
3581 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
3582 addr_lock
= &sctp_sk(sk
)->ep
->base
.addr_lock
;
3584 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
3587 bp
= &asoc
->base
.bind_addr
;
3588 addr_lock
= &asoc
->base
.addr_lock
;
3591 to
= optval
+ offsetof(struct sctp_getaddrs
,addrs
);
3592 space_left
= len
- sizeof(struct sctp_getaddrs
) -
3593 offsetof(struct sctp_getaddrs
,addrs
);
3595 sctp_read_lock(addr_lock
);
3597 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
3598 * addresses from the global local address list.
3600 if (sctp_list_single_entry(&bp
->address_list
)) {
3601 addr
= list_entry(bp
->address_list
.next
,
3602 struct sctp_sockaddr_entry
, list
);
3603 if (sctp_is_any(&addr
->a
)) {
3604 cnt
= sctp_copy_laddrs_to_user(sk
, bp
->port
,
3614 list_for_each(pos
, &bp
->address_list
) {
3615 addr
= list_entry(pos
, struct sctp_sockaddr_entry
, list
);
3616 memcpy(&temp
, &addr
->a
, sizeof(temp
));
3617 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
3618 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
3619 if(space_left
< addrlen
)
3620 return -ENOMEM
; /*fixme: right error?*/
3621 temp
.v4
.sin_port
= htons(temp
.v4
.sin_port
);
3622 if (copy_to_user(to
, &temp
, addrlen
)) {
3628 space_left
-= addrlen
;
3632 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
))
3634 bytes_copied
= ((char __user
*)to
) - optval
;
3635 if (put_user(bytes_copied
, optlen
))
3639 sctp_read_unlock(addr_lock
);
3643 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
3645 * Requests that the local SCTP stack use the enclosed peer address as
3646 * the association primary. The enclosed address must be one of the
3647 * association peer's addresses.
3649 static int sctp_getsockopt_primary_addr(struct sock
*sk
, int len
,
3650 char __user
*optval
, int __user
*optlen
)
3652 struct sctp_prim prim
;
3653 struct sctp_association
*asoc
;
3654 struct sctp_sock
*sp
= sctp_sk(sk
);
3656 if (len
!= sizeof(struct sctp_prim
))
3659 if (copy_from_user(&prim
, optval
, sizeof(struct sctp_prim
)))
3662 asoc
= sctp_id2assoc(sk
, prim
.ssp_assoc_id
);
3666 if (!asoc
->peer
.primary_path
)
3669 asoc
->peer
.primary_path
->ipaddr
.v4
.sin_port
=
3670 htons(asoc
->peer
.primary_path
->ipaddr
.v4
.sin_port
);
3671 memcpy(&prim
.ssp_addr
, &asoc
->peer
.primary_path
->ipaddr
,
3672 sizeof(union sctp_addr
));
3673 asoc
->peer
.primary_path
->ipaddr
.v4
.sin_port
=
3674 ntohs(asoc
->peer
.primary_path
->ipaddr
.v4
.sin_port
);
3676 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
,
3677 (union sctp_addr
*)&prim
.ssp_addr
);
3679 if (copy_to_user(optval
, &prim
, sizeof(struct sctp_prim
)))
3686 * 7.1.11 Set Adaption Layer Indicator (SCTP_ADAPTION_LAYER)
3688 * Requests that the local endpoint set the specified Adaption Layer
3689 * Indication parameter for all future INIT and INIT-ACK exchanges.
3691 static int sctp_getsockopt_adaption_layer(struct sock
*sk
, int len
,
3692 char __user
*optval
, int __user
*optlen
)
3694 struct sctp_setadaption adaption
;
3696 if (len
!= sizeof(struct sctp_setadaption
))
3699 adaption
.ssb_adaption_ind
= sctp_sk(sk
)->adaption_ind
;
3700 if (copy_to_user(optval
, &adaption
, len
))
3708 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
3710 * Applications that wish to use the sendto() system call may wish to
3711 * specify a default set of parameters that would normally be supplied
3712 * through the inclusion of ancillary data. This socket option allows
3713 * such an application to set the default sctp_sndrcvinfo structure.
3716 * The application that wishes to use this socket option simply passes
3717 * in to this call the sctp_sndrcvinfo structure defined in Section
3718 * 5.2.2) The input parameters accepted by this call include
3719 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
3720 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
3721 * to this call if the caller is using the UDP model.
3723 * For getsockopt, it get the default sctp_sndrcvinfo structure.
3725 static int sctp_getsockopt_default_send_param(struct sock
*sk
,
3726 int len
, char __user
*optval
,
3729 struct sctp_sndrcvinfo info
;
3730 struct sctp_association
*asoc
;
3731 struct sctp_sock
*sp
= sctp_sk(sk
);
3733 if (len
!= sizeof(struct sctp_sndrcvinfo
))
3735 if (copy_from_user(&info
, optval
, sizeof(struct sctp_sndrcvinfo
)))
3738 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
3739 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
3743 info
.sinfo_stream
= asoc
->default_stream
;
3744 info
.sinfo_flags
= asoc
->default_flags
;
3745 info
.sinfo_ppid
= asoc
->default_ppid
;
3746 info
.sinfo_context
= asoc
->default_context
;
3747 info
.sinfo_timetolive
= asoc
->default_timetolive
;
3749 info
.sinfo_stream
= sp
->default_stream
;
3750 info
.sinfo_flags
= sp
->default_flags
;
3751 info
.sinfo_ppid
= sp
->default_ppid
;
3752 info
.sinfo_context
= sp
->default_context
;
3753 info
.sinfo_timetolive
= sp
->default_timetolive
;
3756 if (copy_to_user(optval
, &info
, sizeof(struct sctp_sndrcvinfo
)))
3764 * 7.1.5 SCTP_NODELAY
3766 * Turn on/off any Nagle-like algorithm. This means that packets are
3767 * generally sent as soon as possible and no unnecessary delays are
3768 * introduced, at the cost of more packets in the network. Expects an
3769 * integer boolean flag.
3772 static int sctp_getsockopt_nodelay(struct sock
*sk
, int len
,
3773 char __user
*optval
, int __user
*optlen
)
3777 if (len
< sizeof(int))
3781 val
= (sctp_sk(sk
)->nodelay
== 1);
3782 if (put_user(len
, optlen
))
3784 if (copy_to_user(optval
, &val
, len
))
3791 * 7.1.1 SCTP_RTOINFO
3793 * The protocol parameters used to initialize and bound retransmission
3794 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
3795 * and modify these parameters.
3796 * All parameters are time values, in milliseconds. A value of 0, when
3797 * modifying the parameters, indicates that the current value should not
3801 static int sctp_getsockopt_rtoinfo(struct sock
*sk
, int len
,
3802 char __user
*optval
,
3803 int __user
*optlen
) {
3804 struct sctp_rtoinfo rtoinfo
;
3805 struct sctp_association
*asoc
;
3807 if (len
!= sizeof (struct sctp_rtoinfo
))
3810 if (copy_from_user(&rtoinfo
, optval
, sizeof (struct sctp_rtoinfo
)))
3813 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
3815 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
3818 /* Values corresponding to the specific association. */
3820 rtoinfo
.srto_initial
= jiffies_to_msecs(asoc
->rto_initial
);
3821 rtoinfo
.srto_max
= jiffies_to_msecs(asoc
->rto_max
);
3822 rtoinfo
.srto_min
= jiffies_to_msecs(asoc
->rto_min
);
3824 /* Values corresponding to the endpoint. */
3825 struct sctp_sock
*sp
= sctp_sk(sk
);
3827 rtoinfo
.srto_initial
= sp
->rtoinfo
.srto_initial
;
3828 rtoinfo
.srto_max
= sp
->rtoinfo
.srto_max
;
3829 rtoinfo
.srto_min
= sp
->rtoinfo
.srto_min
;
3832 if (put_user(len
, optlen
))
3835 if (copy_to_user(optval
, &rtoinfo
, len
))
3843 * 7.1.2 SCTP_ASSOCINFO
3845 * This option is used to tune the the maximum retransmission attempts
3846 * of the association.
3847 * Returns an error if the new association retransmission value is
3848 * greater than the sum of the retransmission value of the peer.
3849 * See [SCTP] for more information.
3852 static int sctp_getsockopt_associnfo(struct sock
*sk
, int len
,
3853 char __user
*optval
,
3857 struct sctp_assocparams assocparams
;
3858 struct sctp_association
*asoc
;
3859 struct list_head
*pos
;
3862 if (len
!= sizeof (struct sctp_assocparams
))
3865 if (copy_from_user(&assocparams
, optval
,
3866 sizeof (struct sctp_assocparams
)))
3869 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
3871 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
3874 /* Values correspoinding to the specific association */
3876 assocparams
.sasoc_asocmaxrxt
= asoc
->max_retrans
;
3877 assocparams
.sasoc_peer_rwnd
= asoc
->peer
.rwnd
;
3878 assocparams
.sasoc_local_rwnd
= asoc
->a_rwnd
;
3879 assocparams
.sasoc_cookie_life
= (asoc
->cookie_life
.tv_sec
3881 (asoc
->cookie_life
.tv_usec
3884 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
3888 assocparams
.sasoc_number_peer_destinations
= cnt
;
3890 /* Values corresponding to the endpoint */
3891 struct sctp_sock
*sp
= sctp_sk(sk
);
3893 assocparams
.sasoc_asocmaxrxt
= sp
->assocparams
.sasoc_asocmaxrxt
;
3894 assocparams
.sasoc_peer_rwnd
= sp
->assocparams
.sasoc_peer_rwnd
;
3895 assocparams
.sasoc_local_rwnd
= sp
->assocparams
.sasoc_local_rwnd
;
3896 assocparams
.sasoc_cookie_life
=
3897 sp
->assocparams
.sasoc_cookie_life
;
3898 assocparams
.sasoc_number_peer_destinations
=
3900 sasoc_number_peer_destinations
;
3903 if (put_user(len
, optlen
))
3906 if (copy_to_user(optval
, &assocparams
, len
))
3913 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3915 * This socket option is a boolean flag which turns on or off mapped V4
3916 * addresses. If this option is turned on and the socket is type
3917 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3918 * If this option is turned off, then no mapping will be done of V4
3919 * addresses and a user will receive both PF_INET6 and PF_INET type
3920 * addresses on the socket.
3922 static int sctp_getsockopt_mappedv4(struct sock
*sk
, int len
,
3923 char __user
*optval
, int __user
*optlen
)
3926 struct sctp_sock
*sp
= sctp_sk(sk
);
3928 if (len
< sizeof(int))
3933 if (put_user(len
, optlen
))
3935 if (copy_to_user(optval
, &val
, len
))
3942 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
3944 * This socket option specifies the maximum size to put in any outgoing
3945 * SCTP chunk. If a message is larger than this size it will be
3946 * fragmented by SCTP into the specified size. Note that the underlying
3947 * SCTP implementation may fragment into smaller sized chunks when the
3948 * PMTU of the underlying association is smaller than the value set by
3951 static int sctp_getsockopt_maxseg(struct sock
*sk
, int len
,
3952 char __user
*optval
, int __user
*optlen
)
3956 if (len
< sizeof(int))
3961 val
= sctp_sk(sk
)->user_frag
;
3962 if (put_user(len
, optlen
))
3964 if (copy_to_user(optval
, &val
, len
))
3970 SCTP_STATIC
int sctp_getsockopt(struct sock
*sk
, int level
, int optname
,
3971 char __user
*optval
, int __user
*optlen
)
3976 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
3979 /* I can hardly begin to describe how wrong this is. This is
3980 * so broken as to be worse than useless. The API draft
3981 * REALLY is NOT helpful here... I am not convinced that the
3982 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
3983 * are at all well-founded.
3985 if (level
!= SOL_SCTP
) {
3986 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
3988 retval
= af
->getsockopt(sk
, level
, optname
, optval
, optlen
);
3992 if (get_user(len
, optlen
))
3999 retval
= sctp_getsockopt_sctp_status(sk
, len
, optval
, optlen
);
4001 case SCTP_DISABLE_FRAGMENTS
:
4002 retval
= sctp_getsockopt_disable_fragments(sk
, len
, optval
,
4006 retval
= sctp_getsockopt_events(sk
, len
, optval
, optlen
);
4008 case SCTP_AUTOCLOSE
:
4009 retval
= sctp_getsockopt_autoclose(sk
, len
, optval
, optlen
);
4011 case SCTP_SOCKOPT_PEELOFF
:
4012 retval
= sctp_getsockopt_peeloff(sk
, len
, optval
, optlen
);
4014 case SCTP_PEER_ADDR_PARAMS
:
4015 retval
= sctp_getsockopt_peer_addr_params(sk
, len
, optval
,
4019 retval
= sctp_getsockopt_initmsg(sk
, len
, optval
, optlen
);
4021 case SCTP_GET_PEER_ADDRS_NUM_OLD
:
4022 retval
= sctp_getsockopt_peer_addrs_num_old(sk
, len
, optval
,
4025 case SCTP_GET_LOCAL_ADDRS_NUM_OLD
:
4026 retval
= sctp_getsockopt_local_addrs_num_old(sk
, len
, optval
,
4029 case SCTP_GET_PEER_ADDRS_OLD
:
4030 retval
= sctp_getsockopt_peer_addrs_old(sk
, len
, optval
,
4033 case SCTP_GET_LOCAL_ADDRS_OLD
:
4034 retval
= sctp_getsockopt_local_addrs_old(sk
, len
, optval
,
4037 case SCTP_GET_PEER_ADDRS
:
4038 retval
= sctp_getsockopt_peer_addrs(sk
, len
, optval
,
4041 case SCTP_GET_LOCAL_ADDRS
:
4042 retval
= sctp_getsockopt_local_addrs(sk
, len
, optval
,
4045 case SCTP_DEFAULT_SEND_PARAM
:
4046 retval
= sctp_getsockopt_default_send_param(sk
, len
,
4049 case SCTP_PRIMARY_ADDR
:
4050 retval
= sctp_getsockopt_primary_addr(sk
, len
, optval
, optlen
);
4053 retval
= sctp_getsockopt_nodelay(sk
, len
, optval
, optlen
);
4056 retval
= sctp_getsockopt_rtoinfo(sk
, len
, optval
, optlen
);
4058 case SCTP_ASSOCINFO
:
4059 retval
= sctp_getsockopt_associnfo(sk
, len
, optval
, optlen
);
4061 case SCTP_I_WANT_MAPPED_V4_ADDR
:
4062 retval
= sctp_getsockopt_mappedv4(sk
, len
, optval
, optlen
);
4065 retval
= sctp_getsockopt_maxseg(sk
, len
, optval
, optlen
);
4067 case SCTP_GET_PEER_ADDR_INFO
:
4068 retval
= sctp_getsockopt_peer_addr_info(sk
, len
, optval
,
4071 case SCTP_ADAPTION_LAYER
:
4072 retval
= sctp_getsockopt_adaption_layer(sk
, len
, optval
,
4076 retval
= -ENOPROTOOPT
;
4080 sctp_release_sock(sk
);
4084 static void sctp_hash(struct sock
*sk
)
4089 static void sctp_unhash(struct sock
*sk
)
4094 /* Check if port is acceptable. Possibly find first available port.
4096 * The port hash table (contained in the 'global' SCTP protocol storage
4097 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
4098 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
4099 * list (the list number is the port number hashed out, so as you
4100 * would expect from a hash function, all the ports in a given list have
4101 * such a number that hashes out to the same list number; you were
4102 * expecting that, right?); so each list has a set of ports, with a
4103 * link to the socket (struct sock) that uses it, the port number and
4104 * a fastreuse flag (FIXME: NPI ipg).
4106 static struct sctp_bind_bucket
*sctp_bucket_create(
4107 struct sctp_bind_hashbucket
*head
, unsigned short snum
);
4109 static long sctp_get_port_local(struct sock
*sk
, union sctp_addr
*addr
)
4111 struct sctp_bind_hashbucket
*head
; /* hash list */
4112 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
4113 unsigned short snum
;
4116 /* NOTE: Remember to put this back to net order. */
4117 addr
->v4
.sin_port
= ntohs(addr
->v4
.sin_port
);
4118 snum
= addr
->v4
.sin_port
;
4120 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum
);
4121 sctp_local_bh_disable();
4124 /* Search for an available port.
4126 * 'sctp_port_rover' was the last port assigned, so
4127 * we start to search from 'sctp_port_rover +
4128 * 1'. What we do is first check if port 'rover' is
4129 * already in the hash table; if not, we use that; if
4130 * it is, we try next.
4132 int low
= sysctl_local_port_range
[0];
4133 int high
= sysctl_local_port_range
[1];
4134 int remaining
= (high
- low
) + 1;
4138 sctp_spin_lock(&sctp_port_alloc_lock
);
4139 rover
= sctp_port_rover
;
4142 if ((rover
< low
) || (rover
> high
))
4144 index
= sctp_phashfn(rover
);
4145 head
= &sctp_port_hashtable
[index
];
4146 sctp_spin_lock(&head
->lock
);
4147 for (pp
= head
->chain
; pp
; pp
= pp
->next
)
4148 if (pp
->port
== rover
)
4152 sctp_spin_unlock(&head
->lock
);
4153 } while (--remaining
> 0);
4154 sctp_port_rover
= rover
;
4155 sctp_spin_unlock(&sctp_port_alloc_lock
);
4157 /* Exhausted local port range during search? */
4162 /* OK, here is the one we will use. HEAD (the port
4163 * hash table list entry) is non-NULL and we hold it's
4168 /* We are given an specific port number; we verify
4169 * that it is not being used. If it is used, we will
4170 * exahust the search in the hash list corresponding
4171 * to the port number (snum) - we detect that with the
4172 * port iterator, pp being NULL.
4174 head
= &sctp_port_hashtable
[sctp_phashfn(snum
)];
4175 sctp_spin_lock(&head
->lock
);
4176 for (pp
= head
->chain
; pp
; pp
= pp
->next
) {
4177 if (pp
->port
== snum
)
4184 if (!hlist_empty(&pp
->owner
)) {
4185 /* We had a port hash table hit - there is an
4186 * available port (pp != NULL) and it is being
4187 * used by other socket (pp->owner not empty); that other
4188 * socket is going to be sk2.
4190 int reuse
= sk
->sk_reuse
;
4192 struct hlist_node
*node
;
4194 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
4195 if (pp
->fastreuse
&& sk
->sk_reuse
)
4198 /* Run through the list of sockets bound to the port
4199 * (pp->port) [via the pointers bind_next and
4200 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
4201 * we get the endpoint they describe and run through
4202 * the endpoint's list of IP (v4 or v6) addresses,
4203 * comparing each of the addresses with the address of
4204 * the socket sk. If we find a match, then that means
4205 * that this port/socket (sk) combination are already
4208 sk_for_each_bound(sk2
, node
, &pp
->owner
) {
4209 struct sctp_endpoint
*ep2
;
4210 ep2
= sctp_sk(sk2
)->ep
;
4212 if (reuse
&& sk2
->sk_reuse
)
4215 if (sctp_bind_addr_match(&ep2
->base
.bind_addr
, addr
,
4221 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
4224 /* If there was a hash table miss, create a new port. */
4226 if (!pp
&& !(pp
= sctp_bucket_create(head
, snum
)))
4229 /* In either case (hit or miss), make sure fastreuse is 1 only
4230 * if sk->sk_reuse is too (that is, if the caller requested
4231 * SO_REUSEADDR on this socket -sk-).
4233 if (hlist_empty(&pp
->owner
))
4234 pp
->fastreuse
= sk
->sk_reuse
? 1 : 0;
4235 else if (pp
->fastreuse
&& !sk
->sk_reuse
)
4238 /* We are set, so fill up all the data in the hash table
4239 * entry, tie the socket list information with the rest of the
4240 * sockets FIXME: Blurry, NPI (ipg).
4243 inet_sk(sk
)->num
= snum
;
4244 if (!sctp_sk(sk
)->bind_hash
) {
4245 sk_add_bind_node(sk
, &pp
->owner
);
4246 sctp_sk(sk
)->bind_hash
= pp
;
4251 sctp_spin_unlock(&head
->lock
);
4254 sctp_local_bh_enable();
4255 addr
->v4
.sin_port
= htons(addr
->v4
.sin_port
);
4259 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
4260 * port is requested.
4262 static int sctp_get_port(struct sock
*sk
, unsigned short snum
)
4265 union sctp_addr addr
;
4266 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
4268 /* Set up a dummy address struct from the sk. */
4269 af
->from_sk(&addr
, sk
);
4270 addr
.v4
.sin_port
= htons(snum
);
4272 /* Note: sk->sk_num gets filled in if ephemeral port request. */
4273 ret
= sctp_get_port_local(sk
, &addr
);
4275 return (ret
? 1 : 0);
4279 * 3.1.3 listen() - UDP Style Syntax
4281 * By default, new associations are not accepted for UDP style sockets.
4282 * An application uses listen() to mark a socket as being able to
4283 * accept new associations.
4285 SCTP_STATIC
int sctp_seqpacket_listen(struct sock
*sk
, int backlog
)
4287 struct sctp_sock
*sp
= sctp_sk(sk
);
4288 struct sctp_endpoint
*ep
= sp
->ep
;
4290 /* Only UDP style sockets that are not peeled off are allowed to
4293 if (!sctp_style(sk
, UDP
))
4296 /* If backlog is zero, disable listening. */
4298 if (sctp_sstate(sk
, CLOSED
))
4301 sctp_unhash_endpoint(ep
);
4302 sk
->sk_state
= SCTP_SS_CLOSED
;
4305 /* Return if we are already listening. */
4306 if (sctp_sstate(sk
, LISTENING
))
4310 * If a bind() or sctp_bindx() is not called prior to a listen()
4311 * call that allows new associations to be accepted, the system
4312 * picks an ephemeral port and will choose an address set equivalent
4313 * to binding with a wildcard address.
4315 * This is not currently spelled out in the SCTP sockets
4316 * extensions draft, but follows the practice as seen in TCP
4319 if (!ep
->base
.bind_addr
.port
) {
4320 if (sctp_autobind(sk
))
4323 sk
->sk_state
= SCTP_SS_LISTENING
;
4324 sctp_hash_endpoint(ep
);
4329 * 4.1.3 listen() - TCP Style Syntax
4331 * Applications uses listen() to ready the SCTP endpoint for accepting
4332 * inbound associations.
4334 SCTP_STATIC
int sctp_stream_listen(struct sock
*sk
, int backlog
)
4336 struct sctp_sock
*sp
= sctp_sk(sk
);
4337 struct sctp_endpoint
*ep
= sp
->ep
;
4339 /* If backlog is zero, disable listening. */
4341 if (sctp_sstate(sk
, CLOSED
))
4344 sctp_unhash_endpoint(ep
);
4345 sk
->sk_state
= SCTP_SS_CLOSED
;
4348 if (sctp_sstate(sk
, LISTENING
))
4352 * If a bind() or sctp_bindx() is not called prior to a listen()
4353 * call that allows new associations to be accepted, the system
4354 * picks an ephemeral port and will choose an address set equivalent
4355 * to binding with a wildcard address.
4357 * This is not currently spelled out in the SCTP sockets
4358 * extensions draft, but follows the practice as seen in TCP
4361 if (!ep
->base
.bind_addr
.port
) {
4362 if (sctp_autobind(sk
))
4365 sk
->sk_state
= SCTP_SS_LISTENING
;
4366 sk
->sk_max_ack_backlog
= backlog
;
4367 sctp_hash_endpoint(ep
);
4372 * Move a socket to LISTENING state.
4374 int sctp_inet_listen(struct socket
*sock
, int backlog
)
4376 struct sock
*sk
= sock
->sk
;
4377 struct crypto_tfm
*tfm
=NULL
;
4380 if (unlikely(backlog
< 0))
4385 if (sock
->state
!= SS_UNCONNECTED
)
4388 /* Allocate HMAC for generating cookie. */
4389 if (sctp_hmac_alg
) {
4390 tfm
= sctp_crypto_alloc_tfm(sctp_hmac_alg
, 0);
4397 switch (sock
->type
) {
4398 case SOCK_SEQPACKET
:
4399 err
= sctp_seqpacket_listen(sk
, backlog
);
4402 err
= sctp_stream_listen(sk
, backlog
);
4410 /* Store away the transform reference. */
4411 sctp_sk(sk
)->hmac
= tfm
;
4413 sctp_release_sock(sk
);
4416 sctp_crypto_free_tfm(tfm
);
4421 * This function is done by modeling the current datagram_poll() and the
4422 * tcp_poll(). Note that, based on these implementations, we don't
4423 * lock the socket in this function, even though it seems that,
4424 * ideally, locking or some other mechanisms can be used to ensure
4425 * the integrity of the counters (sndbuf and wmem_alloc) used
4426 * in this place. We assume that we don't need locks either until proven
4429 * Another thing to note is that we include the Async I/O support
4430 * here, again, by modeling the current TCP/UDP code. We don't have
4431 * a good way to test with it yet.
4433 unsigned int sctp_poll(struct file
*file
, struct socket
*sock
, poll_table
*wait
)
4435 struct sock
*sk
= sock
->sk
;
4436 struct sctp_sock
*sp
= sctp_sk(sk
);
4439 poll_wait(file
, sk
->sk_sleep
, wait
);
4441 /* A TCP-style listening socket becomes readable when the accept queue
4444 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
4445 return (!list_empty(&sp
->ep
->asocs
)) ?
4446 (POLLIN
| POLLRDNORM
) : 0;
4450 /* Is there any exceptional events? */
4451 if (sk
->sk_err
|| !skb_queue_empty(&sk
->sk_error_queue
))
4453 if (sk
->sk_shutdown
== SHUTDOWN_MASK
)
4456 /* Is it readable? Reconsider this code with TCP-style support. */
4457 if (!skb_queue_empty(&sk
->sk_receive_queue
) ||
4458 (sk
->sk_shutdown
& RCV_SHUTDOWN
))
4459 mask
|= POLLIN
| POLLRDNORM
;
4461 /* The association is either gone or not ready. */
4462 if (!sctp_style(sk
, UDP
) && sctp_sstate(sk
, CLOSED
))
4465 /* Is it writable? */
4466 if (sctp_writeable(sk
)) {
4467 mask
|= POLLOUT
| POLLWRNORM
;
4469 set_bit(SOCK_ASYNC_NOSPACE
, &sk
->sk_socket
->flags
);
4471 * Since the socket is not locked, the buffer
4472 * might be made available after the writeable check and
4473 * before the bit is set. This could cause a lost I/O
4474 * signal. tcp_poll() has a race breaker for this race
4475 * condition. Based on their implementation, we put
4476 * in the following code to cover it as well.
4478 if (sctp_writeable(sk
))
4479 mask
|= POLLOUT
| POLLWRNORM
;
4484 /********************************************************************
4485 * 2nd Level Abstractions
4486 ********************************************************************/
4488 static struct sctp_bind_bucket
*sctp_bucket_create(
4489 struct sctp_bind_hashbucket
*head
, unsigned short snum
)
4491 struct sctp_bind_bucket
*pp
;
4493 pp
= kmem_cache_alloc(sctp_bucket_cachep
, SLAB_ATOMIC
);
4494 SCTP_DBG_OBJCNT_INC(bind_bucket
);
4498 INIT_HLIST_HEAD(&pp
->owner
);
4499 if ((pp
->next
= head
->chain
) != NULL
)
4500 pp
->next
->pprev
= &pp
->next
;
4502 pp
->pprev
= &head
->chain
;
4507 /* Caller must hold hashbucket lock for this tb with local BH disabled */
4508 static void sctp_bucket_destroy(struct sctp_bind_bucket
*pp
)
4510 if (hlist_empty(&pp
->owner
)) {
4512 pp
->next
->pprev
= pp
->pprev
;
4513 *(pp
->pprev
) = pp
->next
;
4514 kmem_cache_free(sctp_bucket_cachep
, pp
);
4515 SCTP_DBG_OBJCNT_DEC(bind_bucket
);
4519 /* Release this socket's reference to a local port. */
4520 static inline void __sctp_put_port(struct sock
*sk
)
4522 struct sctp_bind_hashbucket
*head
=
4523 &sctp_port_hashtable
[sctp_phashfn(inet_sk(sk
)->num
)];
4524 struct sctp_bind_bucket
*pp
;
4526 sctp_spin_lock(&head
->lock
);
4527 pp
= sctp_sk(sk
)->bind_hash
;
4528 __sk_del_bind_node(sk
);
4529 sctp_sk(sk
)->bind_hash
= NULL
;
4530 inet_sk(sk
)->num
= 0;
4531 sctp_bucket_destroy(pp
);
4532 sctp_spin_unlock(&head
->lock
);
4535 void sctp_put_port(struct sock
*sk
)
4537 sctp_local_bh_disable();
4538 __sctp_put_port(sk
);
4539 sctp_local_bh_enable();
4543 * The system picks an ephemeral port and choose an address set equivalent
4544 * to binding with a wildcard address.
4545 * One of those addresses will be the primary address for the association.
4546 * This automatically enables the multihoming capability of SCTP.
4548 static int sctp_autobind(struct sock
*sk
)
4550 union sctp_addr autoaddr
;
4552 unsigned short port
;
4554 /* Initialize a local sockaddr structure to INADDR_ANY. */
4555 af
= sctp_sk(sk
)->pf
->af
;
4557 port
= htons(inet_sk(sk
)->num
);
4558 af
->inaddr_any(&autoaddr
, port
);
4560 return sctp_do_bind(sk
, &autoaddr
, af
->sockaddr_len
);
4563 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
4566 * 4.2 The cmsghdr Structure *
4568 * When ancillary data is sent or received, any number of ancillary data
4569 * objects can be specified by the msg_control and msg_controllen members of
4570 * the msghdr structure, because each object is preceded by
4571 * a cmsghdr structure defining the object's length (the cmsg_len member).
4572 * Historically Berkeley-derived implementations have passed only one object
4573 * at a time, but this API allows multiple objects to be
4574 * passed in a single call to sendmsg() or recvmsg(). The following example
4575 * shows two ancillary data objects in a control buffer.
4577 * |<--------------------------- msg_controllen -------------------------->|
4580 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
4582 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
4585 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
4587 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
4590 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
4591 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
4593 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
4595 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
4602 SCTP_STATIC
int sctp_msghdr_parse(const struct msghdr
*msg
,
4603 sctp_cmsgs_t
*cmsgs
)
4605 struct cmsghdr
*cmsg
;
4607 for (cmsg
= CMSG_FIRSTHDR(msg
);
4609 cmsg
= CMSG_NXTHDR((struct msghdr
*)msg
, cmsg
)) {
4610 if (!CMSG_OK(msg
, cmsg
))
4613 /* Should we parse this header or ignore? */
4614 if (cmsg
->cmsg_level
!= IPPROTO_SCTP
)
4617 /* Strictly check lengths following example in SCM code. */
4618 switch (cmsg
->cmsg_type
) {
4620 /* SCTP Socket API Extension
4621 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
4623 * This cmsghdr structure provides information for
4624 * initializing new SCTP associations with sendmsg().
4625 * The SCTP_INITMSG socket option uses this same data
4626 * structure. This structure is not used for
4629 * cmsg_level cmsg_type cmsg_data[]
4630 * ------------ ------------ ----------------------
4631 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
4633 if (cmsg
->cmsg_len
!=
4634 CMSG_LEN(sizeof(struct sctp_initmsg
)))
4636 cmsgs
->init
= (struct sctp_initmsg
*)CMSG_DATA(cmsg
);
4640 /* SCTP Socket API Extension
4641 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
4643 * This cmsghdr structure specifies SCTP options for
4644 * sendmsg() and describes SCTP header information
4645 * about a received message through recvmsg().
4647 * cmsg_level cmsg_type cmsg_data[]
4648 * ------------ ------------ ----------------------
4649 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
4651 if (cmsg
->cmsg_len
!=
4652 CMSG_LEN(sizeof(struct sctp_sndrcvinfo
)))
4656 (struct sctp_sndrcvinfo
*)CMSG_DATA(cmsg
);
4658 /* Minimally, validate the sinfo_flags. */
4659 if (cmsgs
->info
->sinfo_flags
&
4660 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
4661 SCTP_ABORT
| SCTP_EOF
))
4673 * Wait for a packet..
4674 * Note: This function is the same function as in core/datagram.c
4675 * with a few modifications to make lksctp work.
4677 static int sctp_wait_for_packet(struct sock
* sk
, int *err
, long *timeo_p
)
4682 prepare_to_wait_exclusive(sk
->sk_sleep
, &wait
, TASK_INTERRUPTIBLE
);
4684 /* Socket errors? */
4685 error
= sock_error(sk
);
4689 if (!skb_queue_empty(&sk
->sk_receive_queue
))
4692 /* Socket shut down? */
4693 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
4696 /* Sequenced packets can come disconnected. If so we report the
4701 /* Is there a good reason to think that we may receive some data? */
4702 if (list_empty(&sctp_sk(sk
)->ep
->asocs
) && !sctp_sstate(sk
, LISTENING
))
4705 /* Handle signals. */
4706 if (signal_pending(current
))
4709 /* Let another process have a go. Since we are going to sleep
4710 * anyway. Note: This may cause odd behaviors if the message
4711 * does not fit in the user's buffer, but this seems to be the
4712 * only way to honor MSG_DONTWAIT realistically.
4714 sctp_release_sock(sk
);
4715 *timeo_p
= schedule_timeout(*timeo_p
);
4719 finish_wait(sk
->sk_sleep
, &wait
);
4723 error
= sock_intr_errno(*timeo_p
);
4726 finish_wait(sk
->sk_sleep
, &wait
);
4731 /* Receive a datagram.
4732 * Note: This is pretty much the same routine as in core/datagram.c
4733 * with a few changes to make lksctp work.
4735 static struct sk_buff
*sctp_skb_recv_datagram(struct sock
*sk
, int flags
,
4736 int noblock
, int *err
)
4739 struct sk_buff
*skb
;
4742 timeo
= sock_rcvtimeo(sk
, noblock
);
4744 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
4745 timeo
, MAX_SCHEDULE_TIMEOUT
);
4748 /* Again only user level code calls this function,
4749 * so nothing interrupt level
4750 * will suddenly eat the receive_queue.
4752 * Look at current nfs client by the way...
4753 * However, this function was corrent in any case. 8)
4755 if (flags
& MSG_PEEK
) {
4756 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
4757 skb
= skb_peek(&sk
->sk_receive_queue
);
4759 atomic_inc(&skb
->users
);
4760 spin_unlock_bh(&sk
->sk_receive_queue
.lock
);
4762 skb
= skb_dequeue(&sk
->sk_receive_queue
);
4768 /* Caller is allowed not to check sk->sk_err before calling. */
4769 error
= sock_error(sk
);
4773 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
4776 /* User doesn't want to wait. */
4780 } while (sctp_wait_for_packet(sk
, err
, &timeo
) == 0);
4789 /* If sndbuf has changed, wake up per association sndbuf waiters. */
4790 static void __sctp_write_space(struct sctp_association
*asoc
)
4792 struct sock
*sk
= asoc
->base
.sk
;
4793 struct socket
*sock
= sk
->sk_socket
;
4795 if ((sctp_wspace(asoc
) > 0) && sock
) {
4796 if (waitqueue_active(&asoc
->wait
))
4797 wake_up_interruptible(&asoc
->wait
);
4799 if (sctp_writeable(sk
)) {
4800 if (sk
->sk_sleep
&& waitqueue_active(sk
->sk_sleep
))
4801 wake_up_interruptible(sk
->sk_sleep
);
4803 /* Note that we try to include the Async I/O support
4804 * here by modeling from the current TCP/UDP code.
4805 * We have not tested with it yet.
4807 if (sock
->fasync_list
&&
4808 !(sk
->sk_shutdown
& SEND_SHUTDOWN
))
4809 sock_wake_async(sock
, 2, POLL_OUT
);
4814 /* Do accounting for the sndbuf space.
4815 * Decrement the used sndbuf space of the corresponding association by the
4816 * data size which was just transmitted(freed).
4818 static void sctp_wfree(struct sk_buff
*skb
)
4820 struct sctp_association
*asoc
;
4821 struct sctp_chunk
*chunk
;
4824 /* Get the saved chunk pointer. */
4825 chunk
= *((struct sctp_chunk
**)(skb
->cb
));
4828 asoc
->sndbuf_used
-= SCTP_DATA_SNDSIZE(chunk
) +
4829 sizeof(struct sk_buff
) +
4830 sizeof(struct sctp_chunk
);
4832 atomic_sub(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
4835 __sctp_write_space(asoc
);
4837 sctp_association_put(asoc
);
4840 /* Helper function to wait for space in the sndbuf. */
4841 static int sctp_wait_for_sndbuf(struct sctp_association
*asoc
, long *timeo_p
,
4844 struct sock
*sk
= asoc
->base
.sk
;
4846 long current_timeo
= *timeo_p
;
4849 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
4850 asoc
, (long)(*timeo_p
), msg_len
);
4852 /* Increment the association's refcnt. */
4853 sctp_association_hold(asoc
);
4855 /* Wait on the association specific sndbuf space. */
4857 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
4858 TASK_INTERRUPTIBLE
);
4861 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
4864 if (signal_pending(current
))
4865 goto do_interrupted
;
4866 if (msg_len
<= sctp_wspace(asoc
))
4869 /* Let another process have a go. Since we are going
4872 sctp_release_sock(sk
);
4873 current_timeo
= schedule_timeout(current_timeo
);
4876 *timeo_p
= current_timeo
;
4880 finish_wait(&asoc
->wait
, &wait
);
4882 /* Release the association's refcnt. */
4883 sctp_association_put(asoc
);
4892 err
= sock_intr_errno(*timeo_p
);
4900 /* If socket sndbuf has changed, wake up all per association waiters. */
4901 void sctp_write_space(struct sock
*sk
)
4903 struct sctp_association
*asoc
;
4904 struct list_head
*pos
;
4906 /* Wake up the tasks in each wait queue. */
4907 list_for_each(pos
, &((sctp_sk(sk
))->ep
->asocs
)) {
4908 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
4909 __sctp_write_space(asoc
);
4913 /* Is there any sndbuf space available on the socket?
4915 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
4916 * associations on the same socket. For a UDP-style socket with
4917 * multiple associations, it is possible for it to be "unwriteable"
4918 * prematurely. I assume that this is acceptable because
4919 * a premature "unwriteable" is better than an accidental "writeable" which
4920 * would cause an unwanted block under certain circumstances. For the 1-1
4921 * UDP-style sockets or TCP-style sockets, this code should work.
4924 static int sctp_writeable(struct sock
*sk
)
4928 amt
= sk
->sk_sndbuf
- atomic_read(&sk
->sk_wmem_alloc
);
4934 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
4935 * returns immediately with EINPROGRESS.
4937 static int sctp_wait_for_connect(struct sctp_association
*asoc
, long *timeo_p
)
4939 struct sock
*sk
= asoc
->base
.sk
;
4941 long current_timeo
= *timeo_p
;
4944 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __FUNCTION__
, asoc
,
4947 /* Increment the association's refcnt. */
4948 sctp_association_hold(asoc
);
4951 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
4952 TASK_INTERRUPTIBLE
);
4955 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
4957 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
4960 if (signal_pending(current
))
4961 goto do_interrupted
;
4963 if (sctp_state(asoc
, ESTABLISHED
))
4966 /* Let another process have a go. Since we are going
4969 sctp_release_sock(sk
);
4970 current_timeo
= schedule_timeout(current_timeo
);
4973 *timeo_p
= current_timeo
;
4977 finish_wait(&asoc
->wait
, &wait
);
4979 /* Release the association's refcnt. */
4980 sctp_association_put(asoc
);
4985 if (asoc
->init_err_counter
+ 1 >= asoc
->max_init_attempts
)
4988 err
= -ECONNREFUSED
;
4992 err
= sock_intr_errno(*timeo_p
);
5000 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
)
5002 struct sctp_endpoint
*ep
;
5006 ep
= sctp_sk(sk
)->ep
;
5010 prepare_to_wait_exclusive(sk
->sk_sleep
, &wait
,
5011 TASK_INTERRUPTIBLE
);
5013 if (list_empty(&ep
->asocs
)) {
5014 sctp_release_sock(sk
);
5015 timeo
= schedule_timeout(timeo
);
5020 if (!sctp_sstate(sk
, LISTENING
))
5024 if (!list_empty(&ep
->asocs
))
5027 err
= sock_intr_errno(timeo
);
5028 if (signal_pending(current
))
5036 finish_wait(sk
->sk_sleep
, &wait
);
5041 void sctp_wait_for_close(struct sock
*sk
, long timeout
)
5046 prepare_to_wait(sk
->sk_sleep
, &wait
, TASK_INTERRUPTIBLE
);
5047 if (list_empty(&sctp_sk(sk
)->ep
->asocs
))
5049 sctp_release_sock(sk
);
5050 timeout
= schedule_timeout(timeout
);
5052 } while (!signal_pending(current
) && timeout
);
5054 finish_wait(sk
->sk_sleep
, &wait
);
5057 /* Populate the fields of the newsk from the oldsk and migrate the assoc
5058 * and its messages to the newsk.
5060 static void sctp_sock_migrate(struct sock
*oldsk
, struct sock
*newsk
,
5061 struct sctp_association
*assoc
,
5062 sctp_socket_type_t type
)
5064 struct sctp_sock
*oldsp
= sctp_sk(oldsk
);
5065 struct sctp_sock
*newsp
= sctp_sk(newsk
);
5066 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
5067 struct sctp_endpoint
*newep
= newsp
->ep
;
5068 struct sk_buff
*skb
, *tmp
;
5069 struct sctp_ulpevent
*event
;
5072 /* Migrate socket buffer sizes and all the socket level options to the
5075 newsk
->sk_sndbuf
= oldsk
->sk_sndbuf
;
5076 newsk
->sk_rcvbuf
= oldsk
->sk_rcvbuf
;
5077 /* Brute force copy old sctp opt. */
5078 inet_sk_copy_descendant(newsk
, oldsk
);
5080 /* Restore the ep value that was overwritten with the above structure
5086 /* Hook this new socket in to the bind_hash list. */
5087 pp
= sctp_sk(oldsk
)->bind_hash
;
5088 sk_add_bind_node(newsk
, &pp
->owner
);
5089 sctp_sk(newsk
)->bind_hash
= pp
;
5090 inet_sk(newsk
)->num
= inet_sk(oldsk
)->num
;
5092 /* Copy the bind_addr list from the original endpoint to the new
5093 * endpoint so that we can handle restarts properly
5095 if (assoc
->peer
.ipv4_address
)
5096 flags
|= SCTP_ADDR4_PEERSUPP
;
5097 if (assoc
->peer
.ipv6_address
)
5098 flags
|= SCTP_ADDR6_PEERSUPP
;
5099 sctp_bind_addr_copy(&newsp
->ep
->base
.bind_addr
,
5100 &oldsp
->ep
->base
.bind_addr
,
5101 SCTP_SCOPE_GLOBAL
, GFP_KERNEL
, flags
);
5103 /* Move any messages in the old socket's receive queue that are for the
5104 * peeled off association to the new socket's receive queue.
5106 sctp_skb_for_each(skb
, &oldsk
->sk_receive_queue
, tmp
) {
5107 event
= sctp_skb2event(skb
);
5108 if (event
->asoc
== assoc
) {
5110 __skb_unlink(skb
, &oldsk
->sk_receive_queue
);
5111 __skb_queue_tail(&newsk
->sk_receive_queue
, skb
);
5112 skb_set_owner_r(skb
, newsk
);
5116 /* Clean up any messages pending delivery due to partial
5117 * delivery. Three cases:
5118 * 1) No partial deliver; no work.
5119 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
5120 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
5122 skb_queue_head_init(&newsp
->pd_lobby
);
5123 sctp_sk(newsk
)->pd_mode
= assoc
->ulpq
.pd_mode
;
5125 if (sctp_sk(oldsk
)->pd_mode
) {
5126 struct sk_buff_head
*queue
;
5128 /* Decide which queue to move pd_lobby skbs to. */
5129 if (assoc
->ulpq
.pd_mode
) {
5130 queue
= &newsp
->pd_lobby
;
5132 queue
= &newsk
->sk_receive_queue
;
5134 /* Walk through the pd_lobby, looking for skbs that
5135 * need moved to the new socket.
5137 sctp_skb_for_each(skb
, &oldsp
->pd_lobby
, tmp
) {
5138 event
= sctp_skb2event(skb
);
5139 if (event
->asoc
== assoc
) {
5141 __skb_unlink(skb
, &oldsp
->pd_lobby
);
5142 __skb_queue_tail(queue
, skb
);
5143 skb_set_owner_r(skb
, newsk
);
5147 /* Clear up any skbs waiting for the partial
5148 * delivery to finish.
5150 if (assoc
->ulpq
.pd_mode
)
5151 sctp_clear_pd(oldsk
);
5155 /* Set the type of socket to indicate that it is peeled off from the
5156 * original UDP-style socket or created with the accept() call on a
5157 * TCP-style socket..
5161 /* Migrate the association to the new socket. */
5162 sctp_assoc_migrate(assoc
, newsk
);
5164 /* If the association on the newsk is already closed before accept()
5165 * is called, set RCV_SHUTDOWN flag.
5167 if (sctp_state(assoc
, CLOSED
) && sctp_style(newsk
, TCP
))
5168 newsk
->sk_shutdown
|= RCV_SHUTDOWN
;
5170 newsk
->sk_state
= SCTP_SS_ESTABLISHED
;
5173 /* This proto struct describes the ULP interface for SCTP. */
5174 struct proto sctp_prot
= {
5176 .owner
= THIS_MODULE
,
5177 .close
= sctp_close
,
5178 .connect
= sctp_connect
,
5179 .disconnect
= sctp_disconnect
,
5180 .accept
= sctp_accept
,
5181 .ioctl
= sctp_ioctl
,
5182 .init
= sctp_init_sock
,
5183 .destroy
= sctp_destroy_sock
,
5184 .shutdown
= sctp_shutdown
,
5185 .setsockopt
= sctp_setsockopt
,
5186 .getsockopt
= sctp_getsockopt
,
5187 .sendmsg
= sctp_sendmsg
,
5188 .recvmsg
= sctp_recvmsg
,
5190 .backlog_rcv
= sctp_backlog_rcv
,
5192 .unhash
= sctp_unhash
,
5193 .get_port
= sctp_get_port
,
5194 .obj_size
= sizeof(struct sctp_sock
),
5197 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
5198 struct proto sctpv6_prot
= {
5200 .owner
= THIS_MODULE
,
5201 .close
= sctp_close
,
5202 .connect
= sctp_connect
,
5203 .disconnect
= sctp_disconnect
,
5204 .accept
= sctp_accept
,
5205 .ioctl
= sctp_ioctl
,
5206 .init
= sctp_init_sock
,
5207 .destroy
= sctp_destroy_sock
,
5208 .shutdown
= sctp_shutdown
,
5209 .setsockopt
= sctp_setsockopt
,
5210 .getsockopt
= sctp_getsockopt
,
5211 .sendmsg
= sctp_sendmsg
,
5212 .recvmsg
= sctp_recvmsg
,
5214 .backlog_rcv
= sctp_backlog_rcv
,
5216 .unhash
= sctp_unhash
,
5217 .get_port
= sctp_get_port
,
5218 .obj_size
= sizeof(struct sctp6_sock
),
5220 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */