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 sk
->sk_wmem_queued
+= SCTP_DATA_SNDSIZE(chunk
) +
160 sizeof(struct sk_buff
) +
161 sizeof(struct sctp_chunk
);
163 atomic_add(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
166 /* Verify that this is a valid address. */
167 static inline int sctp_verify_addr(struct sock
*sk
, union sctp_addr
*addr
,
172 /* Verify basic sockaddr. */
173 af
= sctp_sockaddr_af(sctp_sk(sk
), addr
, len
);
177 /* Is this a valid SCTP address? */
178 if (!af
->addr_valid(addr
, sctp_sk(sk
)))
181 if (!sctp_sk(sk
)->pf
->send_verify(sctp_sk(sk
), (addr
)))
187 /* Look up the association by its id. If this is not a UDP-style
188 * socket, the ID field is always ignored.
190 struct sctp_association
*sctp_id2assoc(struct sock
*sk
, sctp_assoc_t id
)
192 struct sctp_association
*asoc
= NULL
;
194 /* If this is not a UDP-style socket, assoc id should be ignored. */
195 if (!sctp_style(sk
, UDP
)) {
196 /* Return NULL if the socket state is not ESTABLISHED. It
197 * could be a TCP-style listening socket or a socket which
198 * hasn't yet called connect() to establish an association.
200 if (!sctp_sstate(sk
, ESTABLISHED
))
203 /* Get the first and the only association from the list. */
204 if (!list_empty(&sctp_sk(sk
)->ep
->asocs
))
205 asoc
= list_entry(sctp_sk(sk
)->ep
->asocs
.next
,
206 struct sctp_association
, asocs
);
210 /* Otherwise this is a UDP-style socket. */
211 if (!id
|| (id
== (sctp_assoc_t
)-1))
214 spin_lock_bh(&sctp_assocs_id_lock
);
215 asoc
= (struct sctp_association
*)idr_find(&sctp_assocs_id
, (int)id
);
216 spin_unlock_bh(&sctp_assocs_id_lock
);
218 if (!asoc
|| (asoc
->base
.sk
!= sk
) || asoc
->base
.dead
)
224 /* Look up the transport from an address and an assoc id. If both address and
225 * id are specified, the associations matching the address and the id should be
228 static struct sctp_transport
*sctp_addr_id2transport(struct sock
*sk
,
229 struct sockaddr_storage
*addr
,
232 struct sctp_association
*addr_asoc
= NULL
, *id_asoc
= NULL
;
233 struct sctp_transport
*transport
;
234 union sctp_addr
*laddr
= (union sctp_addr
*)addr
;
236 laddr
->v4
.sin_port
= ntohs(laddr
->v4
.sin_port
);
237 addr_asoc
= sctp_endpoint_lookup_assoc(sctp_sk(sk
)->ep
,
238 (union sctp_addr
*)addr
,
240 laddr
->v4
.sin_port
= htons(laddr
->v4
.sin_port
);
245 id_asoc
= sctp_id2assoc(sk
, id
);
246 if (id_asoc
&& (id_asoc
!= addr_asoc
))
249 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
250 (union sctp_addr
*)addr
);
255 /* API 3.1.2 bind() - UDP Style Syntax
256 * The syntax of bind() is,
258 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
260 * sd - the socket descriptor returned by socket().
261 * addr - the address structure (struct sockaddr_in or struct
262 * sockaddr_in6 [RFC 2553]),
263 * addr_len - the size of the address structure.
265 SCTP_STATIC
int sctp_bind(struct sock
*sk
, struct sockaddr
*addr
, int addr_len
)
271 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
274 /* Disallow binding twice. */
275 if (!sctp_sk(sk
)->ep
->base
.bind_addr
.port
)
276 retval
= sctp_do_bind(sk
, (union sctp_addr
*)addr
,
281 sctp_release_sock(sk
);
286 static long sctp_get_port_local(struct sock
*, union sctp_addr
*);
288 /* Verify this is a valid sockaddr. */
289 static struct sctp_af
*sctp_sockaddr_af(struct sctp_sock
*opt
,
290 union sctp_addr
*addr
, int len
)
294 /* Check minimum size. */
295 if (len
< sizeof (struct sockaddr
))
298 /* Does this PF support this AF? */
299 if (!opt
->pf
->af_supported(addr
->sa
.sa_family
, opt
))
302 /* If we get this far, af is valid. */
303 af
= sctp_get_af_specific(addr
->sa
.sa_family
);
305 if (len
< af
->sockaddr_len
)
311 /* Bind a local address either to an endpoint or to an association. */
312 SCTP_STATIC
int sctp_do_bind(struct sock
*sk
, union sctp_addr
*addr
, int len
)
314 struct sctp_sock
*sp
= sctp_sk(sk
);
315 struct sctp_endpoint
*ep
= sp
->ep
;
316 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
321 /* Common sockaddr verification. */
322 af
= sctp_sockaddr_af(sp
, addr
, len
);
324 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
329 snum
= ntohs(addr
->v4
.sin_port
);
331 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
332 ", port: %d, new port: %d, len: %d)\n",
338 /* PF specific bind() address verification. */
339 if (!sp
->pf
->bind_verify(sp
, addr
))
340 return -EADDRNOTAVAIL
;
342 /* We must either be unbound, or bind to the same port. */
343 if (bp
->port
&& (snum
!= bp
->port
)) {
344 SCTP_DEBUG_PRINTK("sctp_do_bind:"
345 " New port %d does not match existing port "
346 "%d.\n", snum
, bp
->port
);
350 if (snum
&& snum
< PROT_SOCK
&& !capable(CAP_NET_BIND_SERVICE
))
353 /* Make sure we are allowed to bind here.
354 * The function sctp_get_port_local() does duplicate address
357 if ((ret
= sctp_get_port_local(sk
, addr
))) {
358 if (ret
== (long) sk
) {
359 /* This endpoint has a conflicting address. */
366 /* Refresh ephemeral port. */
368 bp
->port
= inet_sk(sk
)->num
;
370 /* Add the address to the bind address list. */
371 sctp_local_bh_disable();
372 sctp_write_lock(&ep
->base
.addr_lock
);
374 /* Use GFP_ATOMIC since BHs are disabled. */
375 addr
->v4
.sin_port
= ntohs(addr
->v4
.sin_port
);
376 ret
= sctp_add_bind_addr(bp
, addr
, GFP_ATOMIC
);
377 addr
->v4
.sin_port
= htons(addr
->v4
.sin_port
);
378 sctp_write_unlock(&ep
->base
.addr_lock
);
379 sctp_local_bh_enable();
381 /* Copy back into socket for getsockname() use. */
383 inet_sk(sk
)->sport
= htons(inet_sk(sk
)->num
);
384 af
->to_sk_saddr(addr
, sk
);
390 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
392 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
393 * at any one time. If a sender, after sending an ASCONF chunk, decides
394 * it needs to transfer another ASCONF Chunk, it MUST wait until the
395 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
396 * subsequent ASCONF. Note this restriction binds each side, so at any
397 * time two ASCONF may be in-transit on any given association (one sent
398 * from each endpoint).
400 static int sctp_send_asconf(struct sctp_association
*asoc
,
401 struct sctp_chunk
*chunk
)
405 /* If there is an outstanding ASCONF chunk, queue it for later
408 if (asoc
->addip_last_asconf
) {
409 list_add_tail(&chunk
->list
, &asoc
->addip_chunk_list
);
413 /* Hold the chunk until an ASCONF_ACK is received. */
414 sctp_chunk_hold(chunk
);
415 retval
= sctp_primitive_ASCONF(asoc
, chunk
);
417 sctp_chunk_free(chunk
);
419 asoc
->addip_last_asconf
= chunk
;
425 /* Add a list of addresses as bind addresses to local endpoint or
428 * Basically run through each address specified in the addrs/addrcnt
429 * array/length pair, determine if it is IPv6 or IPv4 and call
430 * sctp_do_bind() on it.
432 * If any of them fails, then the operation will be reversed and the
433 * ones that were added will be removed.
435 * Only sctp_setsockopt_bindx() is supposed to call this function.
437 int sctp_bindx_add(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
442 struct sockaddr
*sa_addr
;
445 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
449 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
450 /* The list may contain either IPv4 or IPv6 address;
451 * determine the address length for walking thru the list.
453 sa_addr
= (struct sockaddr
*)addr_buf
;
454 af
= sctp_get_af_specific(sa_addr
->sa_family
);
460 retval
= sctp_do_bind(sk
, (union sctp_addr
*)sa_addr
,
463 addr_buf
+= af
->sockaddr_len
;
467 /* Failed. Cleanup the ones that have been added */
469 sctp_bindx_rem(sk
, addrs
, cnt
);
477 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
478 * associations that are part of the endpoint indicating that a list of local
479 * addresses are added to the endpoint.
481 * If any of the addresses is already in the bind address list of the
482 * association, we do not send the chunk for that association. But it will not
483 * affect other associations.
485 * Only sctp_setsockopt_bindx() is supposed to call this function.
487 static int sctp_send_asconf_add_ip(struct sock
*sk
,
488 struct sockaddr
*addrs
,
491 struct sctp_sock
*sp
;
492 struct sctp_endpoint
*ep
;
493 struct sctp_association
*asoc
;
494 struct sctp_bind_addr
*bp
;
495 struct sctp_chunk
*chunk
;
496 struct sctp_sockaddr_entry
*laddr
;
497 union sctp_addr
*addr
;
500 struct list_head
*pos
;
505 if (!sctp_addip_enable
)
511 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
512 __FUNCTION__
, sk
, addrs
, addrcnt
);
514 list_for_each(pos
, &ep
->asocs
) {
515 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
517 if (!asoc
->peer
.asconf_capable
)
520 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_ADD_IP
)
523 if (!sctp_state(asoc
, ESTABLISHED
))
526 /* Check if any address in the packed array of addresses is
527 * in the bind address list of the association. If so,
528 * do not send the asconf chunk to its peer, but continue with
529 * other associations.
532 for (i
= 0; i
< addrcnt
; i
++) {
533 addr
= (union sctp_addr
*)addr_buf
;
534 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
540 if (sctp_assoc_lookup_laddr(asoc
, addr
))
543 addr_buf
+= af
->sockaddr_len
;
548 /* Use the first address in bind addr list of association as
549 * Address Parameter of ASCONF CHUNK.
551 sctp_read_lock(&asoc
->base
.addr_lock
);
552 bp
= &asoc
->base
.bind_addr
;
553 p
= bp
->address_list
.next
;
554 laddr
= list_entry(p
, struct sctp_sockaddr_entry
, list
);
555 sctp_read_unlock(&asoc
->base
.addr_lock
);
557 chunk
= sctp_make_asconf_update_ip(asoc
, &laddr
->a
, addrs
,
558 addrcnt
, SCTP_PARAM_ADD_IP
);
564 retval
= sctp_send_asconf(asoc
, chunk
);
566 /* FIXME: After sending the add address ASCONF chunk, we
567 * cannot append the address to the association's binding
568 * address list, because the new address may be used as the
569 * source of a message sent to the peer before the ASCONF
570 * chunk is received by the peer. So we should wait until
571 * ASCONF_ACK is received.
579 /* Remove a list of addresses from bind addresses list. Do not remove the
582 * Basically run through each address specified in the addrs/addrcnt
583 * array/length pair, determine if it is IPv6 or IPv4 and call
584 * sctp_del_bind() on it.
586 * If any of them fails, then the operation will be reversed and the
587 * ones that were removed will be added back.
589 * At least one address has to be left; if only one address is
590 * available, the operation will return -EBUSY.
592 * Only sctp_setsockopt_bindx() is supposed to call this function.
594 int sctp_bindx_rem(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
596 struct sctp_sock
*sp
= sctp_sk(sk
);
597 struct sctp_endpoint
*ep
= sp
->ep
;
599 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
601 union sctp_addr saveaddr
;
603 struct sockaddr
*sa_addr
;
606 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
610 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
611 /* If the bind address list is empty or if there is only one
612 * bind address, there is nothing more to be removed (we need
613 * at least one address here).
615 if (list_empty(&bp
->address_list
) ||
616 (sctp_list_single_entry(&bp
->address_list
))) {
621 /* The list may contain either IPv4 or IPv6 address;
622 * determine the address length to copy the address to
625 sa_addr
= (struct sockaddr
*)addr_buf
;
626 af
= sctp_get_af_specific(sa_addr
->sa_family
);
631 memcpy(&saveaddr
, sa_addr
, af
->sockaddr_len
);
632 saveaddr
.v4
.sin_port
= ntohs(saveaddr
.v4
.sin_port
);
633 if (saveaddr
.v4
.sin_port
!= bp
->port
) {
638 /* FIXME - There is probably a need to check if sk->sk_saddr and
639 * sk->sk_rcv_addr are currently set to one of the addresses to
640 * be removed. This is something which needs to be looked into
641 * when we are fixing the outstanding issues with multi-homing
642 * socket routing and failover schemes. Refer to comments in
643 * sctp_do_bind(). -daisy
645 sctp_local_bh_disable();
646 sctp_write_lock(&ep
->base
.addr_lock
);
648 retval
= sctp_del_bind_addr(bp
, &saveaddr
);
650 sctp_write_unlock(&ep
->base
.addr_lock
);
651 sctp_local_bh_enable();
653 addr_buf
+= af
->sockaddr_len
;
656 /* Failed. Add the ones that has been removed back */
658 sctp_bindx_add(sk
, addrs
, cnt
);
666 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
667 * the associations that are part of the endpoint indicating that a list of
668 * local addresses are removed from the endpoint.
670 * If any of the addresses is already in the bind address list of the
671 * association, we do not send the chunk for that association. But it will not
672 * affect other associations.
674 * Only sctp_setsockopt_bindx() is supposed to call this function.
676 static int sctp_send_asconf_del_ip(struct sock
*sk
,
677 struct sockaddr
*addrs
,
680 struct sctp_sock
*sp
;
681 struct sctp_endpoint
*ep
;
682 struct sctp_association
*asoc
;
683 struct sctp_bind_addr
*bp
;
684 struct sctp_chunk
*chunk
;
685 union sctp_addr
*laddr
;
688 struct list_head
*pos
;
692 if (!sctp_addip_enable
)
698 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
699 __FUNCTION__
, sk
, addrs
, addrcnt
);
701 list_for_each(pos
, &ep
->asocs
) {
702 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
704 if (!asoc
->peer
.asconf_capable
)
707 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_DEL_IP
)
710 if (!sctp_state(asoc
, ESTABLISHED
))
713 /* Check if any address in the packed array of addresses is
714 * not present in the bind address list of the association.
715 * If so, do not send the asconf chunk to its peer, but
716 * continue with other associations.
719 for (i
= 0; i
< addrcnt
; i
++) {
720 laddr
= (union sctp_addr
*)addr_buf
;
721 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
727 if (!sctp_assoc_lookup_laddr(asoc
, laddr
))
730 addr_buf
+= af
->sockaddr_len
;
735 /* Find one address in the association's bind address list
736 * that is not in the packed array of addresses. This is to
737 * make sure that we do not delete all the addresses in the
740 sctp_read_lock(&asoc
->base
.addr_lock
);
741 bp
= &asoc
->base
.bind_addr
;
742 laddr
= sctp_find_unmatch_addr(bp
, (union sctp_addr
*)addrs
,
744 sctp_read_unlock(&asoc
->base
.addr_lock
);
748 chunk
= sctp_make_asconf_update_ip(asoc
, laddr
, addrs
, addrcnt
,
755 retval
= sctp_send_asconf(asoc
, chunk
);
757 /* FIXME: After sending the delete address ASCONF chunk, we
758 * cannot remove the addresses from the association's bind
759 * address list, because there maybe some packet send to
760 * the delete addresses, so we should wait until ASCONF_ACK
761 * packet is received.
768 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
771 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
774 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
775 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
778 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
779 * Section 3.1.2 for this usage.
781 * addrs is a pointer to an array of one or more socket addresses. Each
782 * address is contained in its appropriate structure (i.e. struct
783 * sockaddr_in or struct sockaddr_in6) the family of the address type
784 * must be used to distengish the address length (note that this
785 * representation is termed a "packed array" of addresses). The caller
786 * specifies the number of addresses in the array with addrcnt.
788 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
789 * -1, and sets errno to the appropriate error code.
791 * For SCTP, the port given in each socket address must be the same, or
792 * sctp_bindx() will fail, setting errno to EINVAL.
794 * The flags parameter is formed from the bitwise OR of zero or more of
795 * the following currently defined flags:
797 * SCTP_BINDX_ADD_ADDR
799 * SCTP_BINDX_REM_ADDR
801 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
802 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
803 * addresses from the association. The two flags are mutually exclusive;
804 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
805 * not remove all addresses from an association; sctp_bindx() will
806 * reject such an attempt with EINVAL.
808 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
809 * additional addresses with an endpoint after calling bind(). Or use
810 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
811 * socket is associated with so that no new association accepted will be
812 * associated with those addresses. If the endpoint supports dynamic
813 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
814 * endpoint to send the appropriate message to the peer to change the
815 * peers address lists.
817 * Adding and removing addresses from a connected association is
818 * optional functionality. Implementations that do not support this
819 * functionality should return EOPNOTSUPP.
821 * Basically do nothing but copying the addresses from user to kernel
822 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
823 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
826 * We don't use copy_from_user() for optimization: we first do the
827 * sanity checks (buffer size -fast- and access check-healthy
828 * pointer); if all of those succeed, then we can alloc the memory
829 * (expensive operation) needed to copy the data to kernel. Then we do
830 * the copying without checking the user space area
831 * (__copy_from_user()).
833 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
836 * sk The sk of the socket
837 * addrs The pointer to the addresses in user land
838 * addrssize Size of the addrs buffer
839 * op Operation to perform (add or remove, see the flags of
842 * Returns 0 if ok, <0 errno code on error.
844 SCTP_STATIC
int sctp_setsockopt_bindx(struct sock
* sk
,
845 struct sockaddr __user
*addrs
,
846 int addrs_size
, int op
)
848 struct sockaddr
*kaddrs
;
852 struct sockaddr
*sa_addr
;
856 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
857 " addrs_size %d opt %d\n", sk
, addrs
, addrs_size
, op
);
859 if (unlikely(addrs_size
<= 0))
862 /* Check the user passed a healthy pointer. */
863 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
866 /* Alloc space for the address array in kernel memory. */
867 kaddrs
= (struct sockaddr
*)kmalloc(addrs_size
, GFP_KERNEL
);
868 if (unlikely(!kaddrs
))
871 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
876 /* Walk through the addrs buffer and count the number of addresses. */
878 while (walk_size
< addrs_size
) {
879 sa_addr
= (struct sockaddr
*)addr_buf
;
880 af
= sctp_get_af_specific(sa_addr
->sa_family
);
882 /* If the address family is not supported or if this address
883 * causes the address buffer to overflow return EINVAL.
885 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
890 addr_buf
+= af
->sockaddr_len
;
891 walk_size
+= af
->sockaddr_len
;
896 case SCTP_BINDX_ADD_ADDR
:
897 err
= sctp_bindx_add(sk
, kaddrs
, addrcnt
);
900 err
= sctp_send_asconf_add_ip(sk
, kaddrs
, addrcnt
);
903 case SCTP_BINDX_REM_ADDR
:
904 err
= sctp_bindx_rem(sk
, kaddrs
, addrcnt
);
907 err
= sctp_send_asconf_del_ip(sk
, kaddrs
, addrcnt
);
921 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
923 * Common routine for handling connect() and sctp_connectx().
924 * Connect will come in with just a single address.
926 static int __sctp_connect(struct sock
* sk
,
927 struct sockaddr
*kaddrs
,
930 struct sctp_sock
*sp
;
931 struct sctp_endpoint
*ep
;
932 struct sctp_association
*asoc
= NULL
;
933 struct sctp_association
*asoc2
;
934 struct sctp_transport
*transport
;
942 struct sockaddr
*sa_addr
;
948 /* connect() cannot be done on a socket that is already in ESTABLISHED
949 * state - UDP-style peeled off socket or a TCP-style socket that
950 * is already connected.
951 * It cannot be done even on a TCP-style listening socket.
953 if (sctp_sstate(sk
, ESTABLISHED
) ||
954 (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))) {
959 /* Walk through the addrs buffer and count the number of addresses. */
961 while (walk_size
< addrs_size
) {
962 sa_addr
= (struct sockaddr
*)addr_buf
;
963 af
= sctp_get_af_specific(sa_addr
->sa_family
);
965 /* If the address family is not supported or if this address
966 * causes the address buffer to overflow return EINVAL.
968 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
973 err
= sctp_verify_addr(sk
, (union sctp_addr
*)sa_addr
,
978 memcpy(&to
, sa_addr
, af
->sockaddr_len
);
979 to
.v4
.sin_port
= ntohs(to
.v4
.sin_port
);
981 /* Check if there already is a matching association on the
982 * endpoint (other than the one created here).
984 asoc2
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
985 if (asoc2
&& asoc2
!= asoc
) {
986 if (asoc2
->state
>= SCTP_STATE_ESTABLISHED
)
993 /* If we could not find a matching association on the endpoint,
994 * make sure that there is no peeled-off association matching
995 * the peer address even on another socket.
997 if (sctp_endpoint_is_peeled_off(ep
, &to
)) {
998 err
= -EADDRNOTAVAIL
;
1003 /* If a bind() or sctp_bindx() is not called prior to
1004 * an sctp_connectx() call, the system picks an
1005 * ephemeral port and will choose an address set
1006 * equivalent to binding with a wildcard address.
1008 if (!ep
->base
.bind_addr
.port
) {
1009 if (sctp_autobind(sk
)) {
1015 * If an unprivileged user inherits a 1-many
1016 * style socket with open associations on a
1017 * privileged port, it MAY be permitted to
1018 * accept new associations, but it SHOULD NOT
1019 * be permitted to open new associations.
1021 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1022 !capable(CAP_NET_BIND_SERVICE
)) {
1028 scope
= sctp_scope(&to
);
1029 asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1036 /* Prime the peer's transport structures. */
1037 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
,
1045 addr_buf
+= af
->sockaddr_len
;
1046 walk_size
+= af
->sockaddr_len
;
1049 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, GFP_KERNEL
);
1054 err
= sctp_primitive_ASSOCIATE(asoc
, NULL
);
1059 /* Initialize sk's dport and daddr for getpeername() */
1060 inet_sk(sk
)->dport
= htons(asoc
->peer
.port
);
1061 af
= sctp_get_af_specific(to
.sa
.sa_family
);
1062 af
->to_sk_daddr(&to
, sk
);
1064 timeo
= sock_sndtimeo(sk
, sk
->sk_socket
->file
->f_flags
& O_NONBLOCK
);
1065 err
= sctp_wait_for_connect(asoc
, &timeo
);
1067 /* Don't free association on exit. */
1072 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1073 " kaddrs: %p err: %d\n",
1076 sctp_association_free(asoc
);
1080 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1083 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt);
1085 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1086 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1087 * or IPv6 addresses.
1089 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1090 * Section 3.1.2 for this usage.
1092 * addrs is a pointer to an array of one or more socket addresses. Each
1093 * address is contained in its appropriate structure (i.e. struct
1094 * sockaddr_in or struct sockaddr_in6) the family of the address type
1095 * must be used to distengish the address length (note that this
1096 * representation is termed a "packed array" of addresses). The caller
1097 * specifies the number of addresses in the array with addrcnt.
1099 * On success, sctp_connectx() returns 0. On failure, sctp_connectx() returns
1100 * -1, and sets errno to the appropriate error code.
1102 * For SCTP, the port given in each socket address must be the same, or
1103 * sctp_connectx() will fail, setting errno to EINVAL.
1105 * An application can use sctp_connectx to initiate an association with
1106 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1107 * allows a caller to specify multiple addresses at which a peer can be
1108 * reached. The way the SCTP stack uses the list of addresses to set up
1109 * the association is implementation dependant. This function only
1110 * specifies that the stack will try to make use of all the addresses in
1111 * the list when needed.
1113 * Note that the list of addresses passed in is only used for setting up
1114 * the association. It does not necessarily equal the set of addresses
1115 * the peer uses for the resulting association. If the caller wants to
1116 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1117 * retrieve them after the association has been set up.
1119 * Basically do nothing but copying the addresses from user to kernel
1120 * land and invoking either sctp_connectx(). This is used for tunneling
1121 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1123 * We don't use copy_from_user() for optimization: we first do the
1124 * sanity checks (buffer size -fast- and access check-healthy
1125 * pointer); if all of those succeed, then we can alloc the memory
1126 * (expensive operation) needed to copy the data to kernel. Then we do
1127 * the copying without checking the user space area
1128 * (__copy_from_user()).
1130 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1133 * sk The sk of the socket
1134 * addrs The pointer to the addresses in user land
1135 * addrssize Size of the addrs buffer
1137 * Returns 0 if ok, <0 errno code on error.
1139 SCTP_STATIC
int sctp_setsockopt_connectx(struct sock
* sk
,
1140 struct sockaddr __user
*addrs
,
1144 struct sockaddr
*kaddrs
;
1146 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1147 __FUNCTION__
, sk
, addrs
, addrs_size
);
1149 if (unlikely(addrs_size
<= 0))
1152 /* Check the user passed a healthy pointer. */
1153 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
1156 /* Alloc space for the address array in kernel memory. */
1157 kaddrs
= (struct sockaddr
*)kmalloc(addrs_size
, GFP_KERNEL
);
1158 if (unlikely(!kaddrs
))
1161 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
1164 err
= __sctp_connect(sk
, kaddrs
, addrs_size
);
1171 /* API 3.1.4 close() - UDP Style Syntax
1172 * Applications use close() to perform graceful shutdown (as described in
1173 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1174 * by a UDP-style socket.
1178 * ret = close(int sd);
1180 * sd - the socket descriptor of the associations to be closed.
1182 * To gracefully shutdown a specific association represented by the
1183 * UDP-style socket, an application should use the sendmsg() call,
1184 * passing no user data, but including the appropriate flag in the
1185 * ancillary data (see Section xxxx).
1187 * If sd in the close() call is a branched-off socket representing only
1188 * one association, the shutdown is performed on that association only.
1190 * 4.1.6 close() - TCP Style Syntax
1192 * Applications use close() to gracefully close down an association.
1196 * int close(int sd);
1198 * sd - the socket descriptor of the association to be closed.
1200 * After an application calls close() on a socket descriptor, no further
1201 * socket operations will succeed on that descriptor.
1203 * API 7.1.4 SO_LINGER
1205 * An application using the TCP-style socket can use this option to
1206 * perform the SCTP ABORT primitive. The linger option structure is:
1209 * int l_onoff; // option on/off
1210 * int l_linger; // linger time
1213 * To enable the option, set l_onoff to 1. If the l_linger value is set
1214 * to 0, calling close() is the same as the ABORT primitive. If the
1215 * value is set to a negative value, the setsockopt() call will return
1216 * an error. If the value is set to a positive value linger_time, the
1217 * close() can be blocked for at most linger_time ms. If the graceful
1218 * shutdown phase does not finish during this period, close() will
1219 * return but the graceful shutdown phase continues in the system.
1221 SCTP_STATIC
void sctp_close(struct sock
*sk
, long timeout
)
1223 struct sctp_endpoint
*ep
;
1224 struct sctp_association
*asoc
;
1225 struct list_head
*pos
, *temp
;
1227 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk
, timeout
);
1230 sk
->sk_shutdown
= SHUTDOWN_MASK
;
1232 ep
= sctp_sk(sk
)->ep
;
1234 /* Walk all associations on a socket, not on an endpoint. */
1235 list_for_each_safe(pos
, temp
, &ep
->asocs
) {
1236 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
1238 if (sctp_style(sk
, TCP
)) {
1239 /* A closed association can still be in the list if
1240 * it belongs to a TCP-style listening socket that is
1241 * not yet accepted. If so, free it. If not, send an
1242 * ABORT or SHUTDOWN based on the linger options.
1244 if (sctp_state(asoc
, CLOSED
)) {
1245 sctp_unhash_established(asoc
);
1246 sctp_association_free(asoc
);
1248 } else if (sock_flag(sk
, SOCK_LINGER
) &&
1250 sctp_primitive_ABORT(asoc
, NULL
);
1252 sctp_primitive_SHUTDOWN(asoc
, NULL
);
1254 sctp_primitive_SHUTDOWN(asoc
, NULL
);
1257 /* Clean up any skbs sitting on the receive queue. */
1258 sctp_queue_purge_ulpevents(&sk
->sk_receive_queue
);
1259 sctp_queue_purge_ulpevents(&sctp_sk(sk
)->pd_lobby
);
1261 /* On a TCP-style socket, block for at most linger_time if set. */
1262 if (sctp_style(sk
, TCP
) && timeout
)
1263 sctp_wait_for_close(sk
, timeout
);
1265 /* This will run the backlog queue. */
1266 sctp_release_sock(sk
);
1268 /* Supposedly, no process has access to the socket, but
1269 * the net layers still may.
1271 sctp_local_bh_disable();
1272 sctp_bh_lock_sock(sk
);
1274 /* Hold the sock, since sk_common_release() will put sock_put()
1275 * and we have just a little more cleanup.
1278 sk_common_release(sk
);
1280 sctp_bh_unlock_sock(sk
);
1281 sctp_local_bh_enable();
1285 SCTP_DBG_OBJCNT_DEC(sock
);
1288 /* Handle EPIPE error. */
1289 static int sctp_error(struct sock
*sk
, int flags
, int err
)
1292 err
= sock_error(sk
) ? : -EPIPE
;
1293 if (err
== -EPIPE
&& !(flags
& MSG_NOSIGNAL
))
1294 send_sig(SIGPIPE
, current
, 0);
1298 /* API 3.1.3 sendmsg() - UDP Style Syntax
1300 * An application uses sendmsg() and recvmsg() calls to transmit data to
1301 * and receive data from its peer.
1303 * ssize_t sendmsg(int socket, const struct msghdr *message,
1306 * socket - the socket descriptor of the endpoint.
1307 * message - pointer to the msghdr structure which contains a single
1308 * user message and possibly some ancillary data.
1310 * See Section 5 for complete description of the data
1313 * flags - flags sent or received with the user message, see Section
1314 * 5 for complete description of the flags.
1316 * Note: This function could use a rewrite especially when explicit
1317 * connect support comes in.
1319 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1321 SCTP_STATIC
int sctp_msghdr_parse(const struct msghdr
*, sctp_cmsgs_t
*);
1323 SCTP_STATIC
int sctp_sendmsg(struct kiocb
*iocb
, struct sock
*sk
,
1324 struct msghdr
*msg
, size_t msg_len
)
1326 struct sctp_sock
*sp
;
1327 struct sctp_endpoint
*ep
;
1328 struct sctp_association
*new_asoc
=NULL
, *asoc
=NULL
;
1329 struct sctp_transport
*transport
, *chunk_tp
;
1330 struct sctp_chunk
*chunk
;
1332 struct sockaddr
*msg_name
= NULL
;
1333 struct sctp_sndrcvinfo default_sinfo
= { 0 };
1334 struct sctp_sndrcvinfo
*sinfo
;
1335 struct sctp_initmsg
*sinit
;
1336 sctp_assoc_t associd
= 0;
1337 sctp_cmsgs_t cmsgs
= { NULL
};
1341 __u16 sinfo_flags
= 0;
1342 struct sctp_datamsg
*datamsg
;
1343 struct list_head
*pos
;
1344 int msg_flags
= msg
->msg_flags
;
1346 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1353 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep
);
1355 /* We cannot send a message over a TCP-style listening socket. */
1356 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
)) {
1361 /* Parse out the SCTP CMSGs. */
1362 err
= sctp_msghdr_parse(msg
, &cmsgs
);
1365 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err
);
1369 /* Fetch the destination address for this packet. This
1370 * address only selects the association--it is not necessarily
1371 * the address we will send to.
1372 * For a peeled-off socket, msg_name is ignored.
1374 if (!sctp_style(sk
, UDP_HIGH_BANDWIDTH
) && msg
->msg_name
) {
1375 int msg_namelen
= msg
->msg_namelen
;
1377 err
= sctp_verify_addr(sk
, (union sctp_addr
*)msg
->msg_name
,
1382 if (msg_namelen
> sizeof(to
))
1383 msg_namelen
= sizeof(to
);
1384 memcpy(&to
, msg
->msg_name
, msg_namelen
);
1385 SCTP_DEBUG_PRINTK("Just memcpy'd. msg_name is "
1387 to
.v4
.sin_addr
.s_addr
, to
.v4
.sin_port
);
1389 to
.v4
.sin_port
= ntohs(to
.v4
.sin_port
);
1390 msg_name
= msg
->msg_name
;
1396 /* Did the user specify SNDRCVINFO? */
1398 sinfo_flags
= sinfo
->sinfo_flags
;
1399 associd
= sinfo
->sinfo_assoc_id
;
1402 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1403 msg_len
, sinfo_flags
);
1405 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1406 if (sctp_style(sk
, TCP
) && (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
))) {
1411 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1412 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1413 * If SCTP_ABORT is set, the message length could be non zero with
1414 * the msg_iov set to the user abort reason.
1416 if (((sinfo_flags
& SCTP_EOF
) && (msg_len
> 0)) ||
1417 (!(sinfo_flags
& (SCTP_EOF
|SCTP_ABORT
)) && (msg_len
== 0))) {
1422 /* If SCTP_ADDR_OVER is set, there must be an address
1423 * specified in msg_name.
1425 if ((sinfo_flags
& SCTP_ADDR_OVER
) && (!msg
->msg_name
)) {
1432 SCTP_DEBUG_PRINTK("About to look up association.\n");
1436 /* If a msg_name has been specified, assume this is to be used. */
1438 /* Look for a matching association on the endpoint. */
1439 asoc
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1441 /* If we could not find a matching association on the
1442 * endpoint, make sure that it is not a TCP-style
1443 * socket that already has an association or there is
1444 * no peeled-off association on another socket.
1446 if ((sctp_style(sk
, TCP
) &&
1447 sctp_sstate(sk
, ESTABLISHED
)) ||
1448 sctp_endpoint_is_peeled_off(ep
, &to
)) {
1449 err
= -EADDRNOTAVAIL
;
1454 asoc
= sctp_id2assoc(sk
, associd
);
1462 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc
);
1464 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1465 * socket that has an association in CLOSED state. This can
1466 * happen when an accepted socket has an association that is
1469 if (sctp_state(asoc
, CLOSED
) && sctp_style(sk
, TCP
)) {
1474 if (sinfo_flags
& SCTP_EOF
) {
1475 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1477 sctp_primitive_SHUTDOWN(asoc
, NULL
);
1481 if (sinfo_flags
& SCTP_ABORT
) {
1482 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc
);
1483 sctp_primitive_ABORT(asoc
, msg
);
1489 /* Do we need to create the association? */
1491 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1493 if (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
)) {
1498 /* Check for invalid stream against the stream counts,
1499 * either the default or the user specified stream counts.
1502 if (!sinit
|| (sinit
&& !sinit
->sinit_num_ostreams
)) {
1503 /* Check against the defaults. */
1504 if (sinfo
->sinfo_stream
>=
1505 sp
->initmsg
.sinit_num_ostreams
) {
1510 /* Check against the requested. */
1511 if (sinfo
->sinfo_stream
>=
1512 sinit
->sinit_num_ostreams
) {
1520 * API 3.1.2 bind() - UDP Style Syntax
1521 * If a bind() or sctp_bindx() is not called prior to a
1522 * sendmsg() call that initiates a new association, the
1523 * system picks an ephemeral port and will choose an address
1524 * set equivalent to binding with a wildcard address.
1526 if (!ep
->base
.bind_addr
.port
) {
1527 if (sctp_autobind(sk
)) {
1533 * If an unprivileged user inherits a one-to-many
1534 * style socket with open associations on a privileged
1535 * port, it MAY be permitted to accept new associations,
1536 * but it SHOULD NOT be permitted to open new
1539 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1540 !capable(CAP_NET_BIND_SERVICE
)) {
1546 scope
= sctp_scope(&to
);
1547 new_asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1554 /* If the SCTP_INIT ancillary data is specified, set all
1555 * the association init values accordingly.
1558 if (sinit
->sinit_num_ostreams
) {
1559 asoc
->c
.sinit_num_ostreams
=
1560 sinit
->sinit_num_ostreams
;
1562 if (sinit
->sinit_max_instreams
) {
1563 asoc
->c
.sinit_max_instreams
=
1564 sinit
->sinit_max_instreams
;
1566 if (sinit
->sinit_max_attempts
) {
1567 asoc
->max_init_attempts
1568 = sinit
->sinit_max_attempts
;
1570 if (sinit
->sinit_max_init_timeo
) {
1571 asoc
->max_init_timeo
=
1572 msecs_to_jiffies(sinit
->sinit_max_init_timeo
);
1576 /* Prime the peer's transport structures. */
1577 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
, SCTP_UNKNOWN
);
1582 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, GFP_KERNEL
);
1589 /* ASSERT: we have a valid association at this point. */
1590 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1593 /* If the user didn't specify SNDRCVINFO, make up one with
1596 default_sinfo
.sinfo_stream
= asoc
->default_stream
;
1597 default_sinfo
.sinfo_flags
= asoc
->default_flags
;
1598 default_sinfo
.sinfo_ppid
= asoc
->default_ppid
;
1599 default_sinfo
.sinfo_context
= asoc
->default_context
;
1600 default_sinfo
.sinfo_timetolive
= asoc
->default_timetolive
;
1601 default_sinfo
.sinfo_assoc_id
= sctp_assoc2id(asoc
);
1602 sinfo
= &default_sinfo
;
1605 /* API 7.1.7, the sndbuf size per association bounds the
1606 * maximum size of data that can be sent in a single send call.
1608 if (msg_len
> sk
->sk_sndbuf
) {
1613 /* If fragmentation is disabled and the message length exceeds the
1614 * association fragmentation point, return EMSGSIZE. The I-D
1615 * does not specify what this error is, but this looks like
1618 if (sctp_sk(sk
)->disable_fragments
&& (msg_len
> asoc
->frag_point
)) {
1624 /* Check for invalid stream. */
1625 if (sinfo
->sinfo_stream
>= asoc
->c
.sinit_num_ostreams
) {
1631 timeo
= sock_sndtimeo(sk
, msg
->msg_flags
& MSG_DONTWAIT
);
1632 if (!sctp_wspace(asoc
)) {
1633 err
= sctp_wait_for_sndbuf(asoc
, &timeo
, msg_len
);
1638 /* If an address is passed with the sendto/sendmsg call, it is used
1639 * to override the primary destination address in the TCP model, or
1640 * when SCTP_ADDR_OVER flag is set in the UDP model.
1642 if ((sctp_style(sk
, TCP
) && msg_name
) ||
1643 (sinfo_flags
& SCTP_ADDR_OVER
)) {
1644 chunk_tp
= sctp_assoc_lookup_paddr(asoc
, &to
);
1652 /* Auto-connect, if we aren't connected already. */
1653 if (sctp_state(asoc
, CLOSED
)) {
1654 err
= sctp_primitive_ASSOCIATE(asoc
, NULL
);
1657 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1660 /* Break the message into multiple chunks of maximum size. */
1661 datamsg
= sctp_datamsg_from_user(asoc
, sinfo
, msg
, msg_len
);
1667 /* Now send the (possibly) fragmented message. */
1668 list_for_each(pos
, &datamsg
->chunks
) {
1669 chunk
= list_entry(pos
, struct sctp_chunk
, frag_list
);
1670 sctp_datamsg_track(chunk
);
1672 /* Do accounting for the write space. */
1673 sctp_set_owner_w(chunk
);
1675 chunk
->transport
= chunk_tp
;
1677 /* Send it to the lower layers. Note: all chunks
1678 * must either fail or succeed. The lower layer
1679 * works that way today. Keep it that way or this
1682 err
= sctp_primitive_SEND(asoc
, chunk
);
1683 /* Did the lower layer accept the chunk? */
1685 sctp_chunk_free(chunk
);
1686 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1689 sctp_datamsg_free(datamsg
);
1695 /* If we are already past ASSOCIATE, the lower
1696 * layers are responsible for association cleanup.
1702 sctp_association_free(asoc
);
1704 sctp_release_sock(sk
);
1707 return sctp_error(sk
, msg_flags
, err
);
1714 err
= sock_error(sk
);
1724 /* This is an extended version of skb_pull() that removes the data from the
1725 * start of a skb even when data is spread across the list of skb's in the
1726 * frag_list. len specifies the total amount of data that needs to be removed.
1727 * when 'len' bytes could be removed from the skb, it returns 0.
1728 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1729 * could not be removed.
1731 static int sctp_skb_pull(struct sk_buff
*skb
, int len
)
1733 struct sk_buff
*list
;
1734 int skb_len
= skb_headlen(skb
);
1737 if (len
<= skb_len
) {
1738 __skb_pull(skb
, len
);
1742 __skb_pull(skb
, skb_len
);
1744 for (list
= skb_shinfo(skb
)->frag_list
; list
; list
= list
->next
) {
1745 rlen
= sctp_skb_pull(list
, len
);
1746 skb
->len
-= (len
-rlen
);
1747 skb
->data_len
-= (len
-rlen
);
1758 /* API 3.1.3 recvmsg() - UDP Style Syntax
1760 * ssize_t recvmsg(int socket, struct msghdr *message,
1763 * socket - the socket descriptor of the endpoint.
1764 * message - pointer to the msghdr structure which contains a single
1765 * user message and possibly some ancillary data.
1767 * See Section 5 for complete description of the data
1770 * flags - flags sent or received with the user message, see Section
1771 * 5 for complete description of the flags.
1773 static struct sk_buff
*sctp_skb_recv_datagram(struct sock
*, int, int, int *);
1775 SCTP_STATIC
int sctp_recvmsg(struct kiocb
*iocb
, struct sock
*sk
,
1776 struct msghdr
*msg
, size_t len
, int noblock
,
1777 int flags
, int *addr_len
)
1779 struct sctp_ulpevent
*event
= NULL
;
1780 struct sctp_sock
*sp
= sctp_sk(sk
);
1781 struct sk_buff
*skb
;
1786 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1787 "0x%x, %s: %p)\n", "sk", sk
, "msghdr", msg
,
1788 "len", len
, "knoblauch", noblock
,
1789 "flags", flags
, "addr_len", addr_len
);
1793 if (sctp_style(sk
, TCP
) && !sctp_sstate(sk
, ESTABLISHED
)) {
1798 skb
= sctp_skb_recv_datagram(sk
, flags
, noblock
, &err
);
1802 /* Get the total length of the skb including any skb's in the
1811 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
1813 event
= sctp_skb2event(skb
);
1818 sock_recv_timestamp(msg
, sk
, skb
);
1819 if (sctp_ulpevent_is_notification(event
)) {
1820 msg
->msg_flags
|= MSG_NOTIFICATION
;
1821 sp
->pf
->event_msgname(event
, msg
->msg_name
, addr_len
);
1823 sp
->pf
->skb_msgname(skb
, msg
->msg_name
, addr_len
);
1826 /* Check if we allow SCTP_SNDRCVINFO. */
1827 if (sp
->subscribe
.sctp_data_io_event
)
1828 sctp_ulpevent_read_sndrcvinfo(event
, msg
);
1830 /* FIXME: we should be calling IP/IPv6 layers. */
1831 if (sk
->sk_protinfo
.af_inet
.cmsg_flags
)
1832 ip_cmsg_recv(msg
, skb
);
1837 /* If skb's length exceeds the user's buffer, update the skb and
1838 * push it back to the receive_queue so that the next call to
1839 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1841 if (skb_len
> copied
) {
1842 msg
->msg_flags
&= ~MSG_EOR
;
1843 if (flags
& MSG_PEEK
)
1845 sctp_skb_pull(skb
, copied
);
1846 skb_queue_head(&sk
->sk_receive_queue
, skb
);
1848 /* When only partial message is copied to the user, increase
1849 * rwnd by that amount. If all the data in the skb is read,
1850 * rwnd is updated when the event is freed.
1852 sctp_assoc_rwnd_increase(event
->asoc
, copied
);
1854 } else if ((event
->msg_flags
& MSG_NOTIFICATION
) ||
1855 (event
->msg_flags
& MSG_EOR
))
1856 msg
->msg_flags
|= MSG_EOR
;
1858 msg
->msg_flags
&= ~MSG_EOR
;
1861 if (flags
& MSG_PEEK
) {
1862 /* Release the skb reference acquired after peeking the skb in
1863 * sctp_skb_recv_datagram().
1867 /* Free the event which includes releasing the reference to
1868 * the owner of the skb, freeing the skb and updating the
1871 sctp_ulpevent_free(event
);
1874 sctp_release_sock(sk
);
1878 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
1880 * This option is a on/off flag. If enabled no SCTP message
1881 * fragmentation will be performed. Instead if a message being sent
1882 * exceeds the current PMTU size, the message will NOT be sent and
1883 * instead a error will be indicated to the user.
1885 static int sctp_setsockopt_disable_fragments(struct sock
*sk
,
1886 char __user
*optval
, int optlen
)
1890 if (optlen
< sizeof(int))
1893 if (get_user(val
, (int __user
*)optval
))
1896 sctp_sk(sk
)->disable_fragments
= (val
== 0) ? 0 : 1;
1901 static int sctp_setsockopt_events(struct sock
*sk
, char __user
*optval
,
1904 if (optlen
!= sizeof(struct sctp_event_subscribe
))
1906 if (copy_from_user(&sctp_sk(sk
)->subscribe
, optval
, optlen
))
1911 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
1913 * This socket option is applicable to the UDP-style socket only. When
1914 * set it will cause associations that are idle for more than the
1915 * specified number of seconds to automatically close. An association
1916 * being idle is defined an association that has NOT sent or received
1917 * user data. The special value of '0' indicates that no automatic
1918 * close of any associations should be performed. The option expects an
1919 * integer defining the number of seconds of idle time before an
1920 * association is closed.
1922 static int sctp_setsockopt_autoclose(struct sock
*sk
, char __user
*optval
,
1925 struct sctp_sock
*sp
= sctp_sk(sk
);
1927 /* Applicable to UDP-style socket only */
1928 if (sctp_style(sk
, TCP
))
1930 if (optlen
!= sizeof(int))
1932 if (copy_from_user(&sp
->autoclose
, optval
, optlen
))
1935 sp
->ep
->timeouts
[SCTP_EVENT_TIMEOUT_AUTOCLOSE
] = sp
->autoclose
* HZ
;
1939 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
1941 * Applications can enable or disable heartbeats for any peer address of
1942 * an association, modify an address's heartbeat interval, force a
1943 * heartbeat to be sent immediately, and adjust the address's maximum
1944 * number of retransmissions sent before an address is considered
1945 * unreachable. The following structure is used to access and modify an
1946 * address's parameters:
1948 * struct sctp_paddrparams {
1949 * sctp_assoc_t spp_assoc_id;
1950 * struct sockaddr_storage spp_address;
1951 * uint32_t spp_hbinterval;
1952 * uint16_t spp_pathmaxrxt;
1955 * spp_assoc_id - (UDP style socket) This is filled in the application,
1956 * and identifies the association for this query.
1957 * spp_address - This specifies which address is of interest.
1958 * spp_hbinterval - This contains the value of the heartbeat interval,
1959 * in milliseconds. A value of 0, when modifying the
1960 * parameter, specifies that the heartbeat on this
1961 * address should be disabled. A value of UINT32_MAX
1962 * (4294967295), when modifying the parameter,
1963 * specifies that a heartbeat should be sent
1964 * immediately to the peer address, and the current
1965 * interval should remain unchanged.
1966 * spp_pathmaxrxt - This contains the maximum number of
1967 * retransmissions before this address shall be
1968 * considered unreachable.
1970 static int sctp_setsockopt_peer_addr_params(struct sock
*sk
,
1971 char __user
*optval
, int optlen
)
1973 struct sctp_paddrparams params
;
1974 struct sctp_transport
*trans
;
1977 if (optlen
!= sizeof(struct sctp_paddrparams
))
1979 if (copy_from_user(¶ms
, optval
, optlen
))
1983 * API 7. Socket Options (setting the default value for the endpoint)
1984 * All options that support specific settings on an association by
1985 * filling in either an association id variable or a sockaddr_storage
1986 * SHOULD also support setting of the same value for the entire endpoint
1987 * (i.e. future associations). To accomplish this the following logic is
1988 * used when setting one of these options:
1990 * c) If neither the sockaddr_storage or association identification is
1991 * set i.e. the sockaddr_storage is set to all 0's (INADDR_ANY) and
1992 * the association identification is 0, the settings are a default
1993 * and to be applied to the endpoint (all future associations).
1996 /* update default value for endpoint (all future associations) */
1997 if (!params
.spp_assoc_id
&&
1998 sctp_is_any(( union sctp_addr
*)¶ms
.spp_address
)) {
1999 /* Manual heartbeat on an endpoint is invalid. */
2000 if (0xffffffff == params
.spp_hbinterval
)
2002 else if (params
.spp_hbinterval
)
2003 sctp_sk(sk
)->paddrparam
.spp_hbinterval
=
2004 params
.spp_hbinterval
;
2005 if (params
.spp_pathmaxrxt
)
2006 sctp_sk(sk
)->paddrparam
.spp_pathmaxrxt
=
2007 params
.spp_pathmaxrxt
;
2011 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
2012 params
.spp_assoc_id
);
2016 /* Applications can enable or disable heartbeats for any peer address
2017 * of an association, modify an address's heartbeat interval, force a
2018 * heartbeat to be sent immediately, and adjust the address's maximum
2019 * number of retransmissions sent before an address is considered
2022 * The value of the heartbeat interval, in milliseconds. A value of
2023 * UINT32_MAX (4294967295), when modifying the parameter, specifies
2024 * that a heartbeat should be sent immediately to the peer address,
2025 * and the current interval should remain unchanged.
2027 if (0xffffffff == params
.spp_hbinterval
) {
2028 error
= sctp_primitive_REQUESTHEARTBEAT (trans
->asoc
, trans
);
2032 /* The value of the heartbeat interval, in milliseconds. A value of 0,
2033 * when modifying the parameter, specifies that the heartbeat on this
2034 * address should be disabled.
2036 if (params
.spp_hbinterval
) {
2037 trans
->hb_allowed
= 1;
2038 trans
->hb_interval
=
2039 msecs_to_jiffies(params
.spp_hbinterval
);
2041 trans
->hb_allowed
= 0;
2044 /* spp_pathmaxrxt contains the maximum number of retransmissions
2045 * before this address shall be considered unreachable.
2047 if (params
.spp_pathmaxrxt
)
2048 trans
->max_retrans
= params
.spp_pathmaxrxt
;
2053 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2055 * Applications can specify protocol parameters for the default association
2056 * initialization. The option name argument to setsockopt() and getsockopt()
2059 * Setting initialization parameters is effective only on an unconnected
2060 * socket (for UDP-style sockets only future associations are effected
2061 * by the change). With TCP-style sockets, this option is inherited by
2062 * sockets derived from a listener socket.
2064 static int sctp_setsockopt_initmsg(struct sock
*sk
, char __user
*optval
, int optlen
)
2066 struct sctp_initmsg sinit
;
2067 struct sctp_sock
*sp
= sctp_sk(sk
);
2069 if (optlen
!= sizeof(struct sctp_initmsg
))
2071 if (copy_from_user(&sinit
, optval
, optlen
))
2074 if (sinit
.sinit_num_ostreams
)
2075 sp
->initmsg
.sinit_num_ostreams
= sinit
.sinit_num_ostreams
;
2076 if (sinit
.sinit_max_instreams
)
2077 sp
->initmsg
.sinit_max_instreams
= sinit
.sinit_max_instreams
;
2078 if (sinit
.sinit_max_attempts
)
2079 sp
->initmsg
.sinit_max_attempts
= sinit
.sinit_max_attempts
;
2080 if (sinit
.sinit_max_init_timeo
)
2081 sp
->initmsg
.sinit_max_init_timeo
= sinit
.sinit_max_init_timeo
;
2087 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2089 * Applications that wish to use the sendto() system call may wish to
2090 * specify a default set of parameters that would normally be supplied
2091 * through the inclusion of ancillary data. This socket option allows
2092 * such an application to set the default sctp_sndrcvinfo structure.
2093 * The application that wishes to use this socket option simply passes
2094 * in to this call the sctp_sndrcvinfo structure defined in Section
2095 * 5.2.2) The input parameters accepted by this call include
2096 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2097 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2098 * to this call if the caller is using the UDP model.
2100 static int sctp_setsockopt_default_send_param(struct sock
*sk
,
2101 char __user
*optval
, int optlen
)
2103 struct sctp_sndrcvinfo info
;
2104 struct sctp_association
*asoc
;
2105 struct sctp_sock
*sp
= sctp_sk(sk
);
2107 if (optlen
!= sizeof(struct sctp_sndrcvinfo
))
2109 if (copy_from_user(&info
, optval
, optlen
))
2112 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
2113 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
2117 asoc
->default_stream
= info
.sinfo_stream
;
2118 asoc
->default_flags
= info
.sinfo_flags
;
2119 asoc
->default_ppid
= info
.sinfo_ppid
;
2120 asoc
->default_context
= info
.sinfo_context
;
2121 asoc
->default_timetolive
= info
.sinfo_timetolive
;
2123 sp
->default_stream
= info
.sinfo_stream
;
2124 sp
->default_flags
= info
.sinfo_flags
;
2125 sp
->default_ppid
= info
.sinfo_ppid
;
2126 sp
->default_context
= info
.sinfo_context
;
2127 sp
->default_timetolive
= info
.sinfo_timetolive
;
2133 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2135 * Requests that the local SCTP stack use the enclosed peer address as
2136 * the association primary. The enclosed address must be one of the
2137 * association peer's addresses.
2139 static int sctp_setsockopt_primary_addr(struct sock
*sk
, char __user
*optval
,
2142 struct sctp_prim prim
;
2143 struct sctp_transport
*trans
;
2145 if (optlen
!= sizeof(struct sctp_prim
))
2148 if (copy_from_user(&prim
, optval
, sizeof(struct sctp_prim
)))
2151 trans
= sctp_addr_id2transport(sk
, &prim
.ssp_addr
, prim
.ssp_assoc_id
);
2155 sctp_assoc_set_primary(trans
->asoc
, trans
);
2161 * 7.1.5 SCTP_NODELAY
2163 * Turn on/off any Nagle-like algorithm. This means that packets are
2164 * generally sent as soon as possible and no unnecessary delays are
2165 * introduced, at the cost of more packets in the network. Expects an
2166 * integer boolean flag.
2168 static int sctp_setsockopt_nodelay(struct sock
*sk
, char __user
*optval
,
2173 if (optlen
< sizeof(int))
2175 if (get_user(val
, (int __user
*)optval
))
2178 sctp_sk(sk
)->nodelay
= (val
== 0) ? 0 : 1;
2184 * 7.1.1 SCTP_RTOINFO
2186 * The protocol parameters used to initialize and bound retransmission
2187 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2188 * and modify these parameters.
2189 * All parameters are time values, in milliseconds. A value of 0, when
2190 * modifying the parameters, indicates that the current value should not
2194 static int sctp_setsockopt_rtoinfo(struct sock
*sk
, char __user
*optval
, int optlen
) {
2195 struct sctp_rtoinfo rtoinfo
;
2196 struct sctp_association
*asoc
;
2198 if (optlen
!= sizeof (struct sctp_rtoinfo
))
2201 if (copy_from_user(&rtoinfo
, optval
, optlen
))
2204 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
2206 /* Set the values to the specific association */
2207 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
2211 if (rtoinfo
.srto_initial
!= 0)
2213 msecs_to_jiffies(rtoinfo
.srto_initial
);
2214 if (rtoinfo
.srto_max
!= 0)
2215 asoc
->rto_max
= msecs_to_jiffies(rtoinfo
.srto_max
);
2216 if (rtoinfo
.srto_min
!= 0)
2217 asoc
->rto_min
= msecs_to_jiffies(rtoinfo
.srto_min
);
2219 /* If there is no association or the association-id = 0
2220 * set the values to the endpoint.
2222 struct sctp_sock
*sp
= sctp_sk(sk
);
2224 if (rtoinfo
.srto_initial
!= 0)
2225 sp
->rtoinfo
.srto_initial
= rtoinfo
.srto_initial
;
2226 if (rtoinfo
.srto_max
!= 0)
2227 sp
->rtoinfo
.srto_max
= rtoinfo
.srto_max
;
2228 if (rtoinfo
.srto_min
!= 0)
2229 sp
->rtoinfo
.srto_min
= rtoinfo
.srto_min
;
2237 * 7.1.2 SCTP_ASSOCINFO
2239 * This option is used to tune the the maximum retransmission attempts
2240 * of the association.
2241 * Returns an error if the new association retransmission value is
2242 * greater than the sum of the retransmission value of the peer.
2243 * See [SCTP] for more information.
2246 static int sctp_setsockopt_associnfo(struct sock
*sk
, char __user
*optval
, int optlen
)
2249 struct sctp_assocparams assocparams
;
2250 struct sctp_association
*asoc
;
2252 if (optlen
!= sizeof(struct sctp_assocparams
))
2254 if (copy_from_user(&assocparams
, optval
, optlen
))
2257 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
2259 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
2262 /* Set the values to the specific association */
2264 if (assocparams
.sasoc_asocmaxrxt
!= 0)
2265 asoc
->max_retrans
= assocparams
.sasoc_asocmaxrxt
;
2266 if (assocparams
.sasoc_cookie_life
!= 0) {
2267 asoc
->cookie_life
.tv_sec
=
2268 assocparams
.sasoc_cookie_life
/ 1000;
2269 asoc
->cookie_life
.tv_usec
=
2270 (assocparams
.sasoc_cookie_life
% 1000)
2274 /* Set the values to the endpoint */
2275 struct sctp_sock
*sp
= sctp_sk(sk
);
2277 if (assocparams
.sasoc_asocmaxrxt
!= 0)
2278 sp
->assocparams
.sasoc_asocmaxrxt
=
2279 assocparams
.sasoc_asocmaxrxt
;
2280 if (assocparams
.sasoc_cookie_life
!= 0)
2281 sp
->assocparams
.sasoc_cookie_life
=
2282 assocparams
.sasoc_cookie_life
;
2288 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2290 * This socket option is a boolean flag which turns on or off mapped V4
2291 * addresses. If this option is turned on and the socket is type
2292 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2293 * If this option is turned off, then no mapping will be done of V4
2294 * addresses and a user will receive both PF_INET6 and PF_INET type
2295 * addresses on the socket.
2297 static int sctp_setsockopt_mappedv4(struct sock
*sk
, char __user
*optval
, int optlen
)
2300 struct sctp_sock
*sp
= sctp_sk(sk
);
2302 if (optlen
< sizeof(int))
2304 if (get_user(val
, (int __user
*)optval
))
2315 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
2317 * This socket option specifies the maximum size to put in any outgoing
2318 * SCTP chunk. If a message is larger than this size it will be
2319 * fragmented by SCTP into the specified size. Note that the underlying
2320 * SCTP implementation may fragment into smaller sized chunks when the
2321 * PMTU of the underlying association is smaller than the value set by
2324 static int sctp_setsockopt_maxseg(struct sock
*sk
, char __user
*optval
, int optlen
)
2326 struct sctp_association
*asoc
;
2327 struct list_head
*pos
;
2328 struct sctp_sock
*sp
= sctp_sk(sk
);
2331 if (optlen
< sizeof(int))
2333 if (get_user(val
, (int __user
*)optval
))
2335 if ((val
!= 0) && ((val
< 8) || (val
> SCTP_MAX_CHUNK_LEN
)))
2337 sp
->user_frag
= val
;
2339 /* Update the frag_point of the existing associations. */
2340 list_for_each(pos
, &(sp
->ep
->asocs
)) {
2341 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
2342 asoc
->frag_point
= sctp_frag_point(sp
, asoc
->pmtu
);
2350 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2352 * Requests that the peer mark the enclosed address as the association
2353 * primary. The enclosed address must be one of the association's
2354 * locally bound addresses. The following structure is used to make a
2355 * set primary request:
2357 static int sctp_setsockopt_peer_primary_addr(struct sock
*sk
, char __user
*optval
,
2360 struct sctp_sock
*sp
;
2361 struct sctp_endpoint
*ep
;
2362 struct sctp_association
*asoc
= NULL
;
2363 struct sctp_setpeerprim prim
;
2364 struct sctp_chunk
*chunk
;
2370 if (!sctp_addip_enable
)
2373 if (optlen
!= sizeof(struct sctp_setpeerprim
))
2376 if (copy_from_user(&prim
, optval
, optlen
))
2379 asoc
= sctp_id2assoc(sk
, prim
.sspp_assoc_id
);
2383 if (!asoc
->peer
.asconf_capable
)
2386 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_SET_PRIMARY
)
2389 if (!sctp_state(asoc
, ESTABLISHED
))
2392 if (!sctp_assoc_lookup_laddr(asoc
, (union sctp_addr
*)&prim
.sspp_addr
))
2393 return -EADDRNOTAVAIL
;
2395 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2396 chunk
= sctp_make_asconf_set_prim(asoc
,
2397 (union sctp_addr
*)&prim
.sspp_addr
);
2401 err
= sctp_send_asconf(asoc
, chunk
);
2403 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2408 static int sctp_setsockopt_adaption_layer(struct sock
*sk
, char __user
*optval
,
2411 struct sctp_setadaption adaption
;
2413 if (optlen
!= sizeof(struct sctp_setadaption
))
2415 if (copy_from_user(&adaption
, optval
, optlen
))
2418 sctp_sk(sk
)->adaption_ind
= adaption
.ssb_adaption_ind
;
2423 /* API 6.2 setsockopt(), getsockopt()
2425 * Applications use setsockopt() and getsockopt() to set or retrieve
2426 * socket options. Socket options are used to change the default
2427 * behavior of sockets calls. They are described in Section 7.
2431 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
2432 * int __user *optlen);
2433 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
2436 * sd - the socket descript.
2437 * level - set to IPPROTO_SCTP for all SCTP options.
2438 * optname - the option name.
2439 * optval - the buffer to store the value of the option.
2440 * optlen - the size of the buffer.
2442 SCTP_STATIC
int sctp_setsockopt(struct sock
*sk
, int level
, int optname
,
2443 char __user
*optval
, int optlen
)
2447 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
2450 /* I can hardly begin to describe how wrong this is. This is
2451 * so broken as to be worse than useless. The API draft
2452 * REALLY is NOT helpful here... I am not convinced that the
2453 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
2454 * are at all well-founded.
2456 if (level
!= SOL_SCTP
) {
2457 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
2458 retval
= af
->setsockopt(sk
, level
, optname
, optval
, optlen
);
2465 case SCTP_SOCKOPT_BINDX_ADD
:
2466 /* 'optlen' is the size of the addresses buffer. */
2467 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
2468 optlen
, SCTP_BINDX_ADD_ADDR
);
2471 case SCTP_SOCKOPT_BINDX_REM
:
2472 /* 'optlen' is the size of the addresses buffer. */
2473 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
2474 optlen
, SCTP_BINDX_REM_ADDR
);
2477 case SCTP_SOCKOPT_CONNECTX
:
2478 /* 'optlen' is the size of the addresses buffer. */
2479 retval
= sctp_setsockopt_connectx(sk
, (struct sockaddr __user
*)optval
,
2483 case SCTP_DISABLE_FRAGMENTS
:
2484 retval
= sctp_setsockopt_disable_fragments(sk
, optval
, optlen
);
2488 retval
= sctp_setsockopt_events(sk
, optval
, optlen
);
2491 case SCTP_AUTOCLOSE
:
2492 retval
= sctp_setsockopt_autoclose(sk
, optval
, optlen
);
2495 case SCTP_PEER_ADDR_PARAMS
:
2496 retval
= sctp_setsockopt_peer_addr_params(sk
, optval
, optlen
);
2500 retval
= sctp_setsockopt_initmsg(sk
, optval
, optlen
);
2502 case SCTP_DEFAULT_SEND_PARAM
:
2503 retval
= sctp_setsockopt_default_send_param(sk
, optval
,
2506 case SCTP_PRIMARY_ADDR
:
2507 retval
= sctp_setsockopt_primary_addr(sk
, optval
, optlen
);
2509 case SCTP_SET_PEER_PRIMARY_ADDR
:
2510 retval
= sctp_setsockopt_peer_primary_addr(sk
, optval
, optlen
);
2513 retval
= sctp_setsockopt_nodelay(sk
, optval
, optlen
);
2516 retval
= sctp_setsockopt_rtoinfo(sk
, optval
, optlen
);
2518 case SCTP_ASSOCINFO
:
2519 retval
= sctp_setsockopt_associnfo(sk
, optval
, optlen
);
2521 case SCTP_I_WANT_MAPPED_V4_ADDR
:
2522 retval
= sctp_setsockopt_mappedv4(sk
, optval
, optlen
);
2525 retval
= sctp_setsockopt_maxseg(sk
, optval
, optlen
);
2527 case SCTP_ADAPTION_LAYER
:
2528 retval
= sctp_setsockopt_adaption_layer(sk
, optval
, optlen
);
2532 retval
= -ENOPROTOOPT
;
2536 sctp_release_sock(sk
);
2542 /* API 3.1.6 connect() - UDP Style Syntax
2544 * An application may use the connect() call in the UDP model to initiate an
2545 * association without sending data.
2549 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
2551 * sd: the socket descriptor to have a new association added to.
2553 * nam: the address structure (either struct sockaddr_in or struct
2554 * sockaddr_in6 defined in RFC2553 [7]).
2556 * len: the size of the address.
2558 SCTP_STATIC
int sctp_connect(struct sock
*sk
, struct sockaddr
*addr
,
2566 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
2567 __FUNCTION__
, sk
, addr
, addr_len
);
2569 /* Validate addr_len before calling common connect/connectx routine. */
2570 af
= sctp_get_af_specific(addr
->sa_family
);
2571 if (!af
|| addr_len
< af
->sockaddr_len
) {
2574 /* Pass correct addr len to common routine (so it knows there
2575 * is only one address being passed.
2577 err
= __sctp_connect(sk
, addr
, af
->sockaddr_len
);
2580 sctp_release_sock(sk
);
2584 /* FIXME: Write comments. */
2585 SCTP_STATIC
int sctp_disconnect(struct sock
*sk
, int flags
)
2587 return -EOPNOTSUPP
; /* STUB */
2590 /* 4.1.4 accept() - TCP Style Syntax
2592 * Applications use accept() call to remove an established SCTP
2593 * association from the accept queue of the endpoint. A new socket
2594 * descriptor will be returned from accept() to represent the newly
2595 * formed association.
2597 SCTP_STATIC
struct sock
*sctp_accept(struct sock
*sk
, int flags
, int *err
)
2599 struct sctp_sock
*sp
;
2600 struct sctp_endpoint
*ep
;
2601 struct sock
*newsk
= NULL
;
2602 struct sctp_association
*asoc
;
2611 if (!sctp_style(sk
, TCP
)) {
2612 error
= -EOPNOTSUPP
;
2616 if (!sctp_sstate(sk
, LISTENING
)) {
2621 timeo
= sock_rcvtimeo(sk
, sk
->sk_socket
->file
->f_flags
& O_NONBLOCK
);
2623 error
= sctp_wait_for_accept(sk
, timeo
);
2627 /* We treat the list of associations on the endpoint as the accept
2628 * queue and pick the first association on the list.
2630 asoc
= list_entry(ep
->asocs
.next
, struct sctp_association
, asocs
);
2632 newsk
= sp
->pf
->create_accept_sk(sk
, asoc
);
2638 /* Populate the fields of the newsk from the oldsk and migrate the
2639 * asoc to the newsk.
2641 sctp_sock_migrate(sk
, newsk
, asoc
, SCTP_SOCKET_TCP
);
2644 sctp_release_sock(sk
);
2649 /* The SCTP ioctl handler. */
2650 SCTP_STATIC
int sctp_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
2652 return -ENOIOCTLCMD
;
2655 /* This is the function which gets called during socket creation to
2656 * initialized the SCTP-specific portion of the sock.
2657 * The sock structure should already be zero-filled memory.
2659 SCTP_STATIC
int sctp_init_sock(struct sock
*sk
)
2661 struct sctp_endpoint
*ep
;
2662 struct sctp_sock
*sp
;
2664 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk
);
2668 /* Initialize the SCTP per socket area. */
2669 switch (sk
->sk_type
) {
2670 case SOCK_SEQPACKET
:
2671 sp
->type
= SCTP_SOCKET_UDP
;
2674 sp
->type
= SCTP_SOCKET_TCP
;
2677 return -ESOCKTNOSUPPORT
;
2680 /* Initialize default send parameters. These parameters can be
2681 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
2683 sp
->default_stream
= 0;
2684 sp
->default_ppid
= 0;
2685 sp
->default_flags
= 0;
2686 sp
->default_context
= 0;
2687 sp
->default_timetolive
= 0;
2689 /* Initialize default setup parameters. These parameters
2690 * can be modified with the SCTP_INITMSG socket option or
2691 * overridden by the SCTP_INIT CMSG.
2693 sp
->initmsg
.sinit_num_ostreams
= sctp_max_outstreams
;
2694 sp
->initmsg
.sinit_max_instreams
= sctp_max_instreams
;
2695 sp
->initmsg
.sinit_max_attempts
= sctp_max_retrans_init
;
2696 sp
->initmsg
.sinit_max_init_timeo
= jiffies_to_msecs(sctp_rto_max
);
2698 /* Initialize default RTO related parameters. These parameters can
2699 * be modified for with the SCTP_RTOINFO socket option.
2701 sp
->rtoinfo
.srto_initial
= jiffies_to_msecs(sctp_rto_initial
);
2702 sp
->rtoinfo
.srto_max
= jiffies_to_msecs(sctp_rto_max
);
2703 sp
->rtoinfo
.srto_min
= jiffies_to_msecs(sctp_rto_min
);
2705 /* Initialize default association related parameters. These parameters
2706 * can be modified with the SCTP_ASSOCINFO socket option.
2708 sp
->assocparams
.sasoc_asocmaxrxt
= sctp_max_retrans_association
;
2709 sp
->assocparams
.sasoc_number_peer_destinations
= 0;
2710 sp
->assocparams
.sasoc_peer_rwnd
= 0;
2711 sp
->assocparams
.sasoc_local_rwnd
= 0;
2712 sp
->assocparams
.sasoc_cookie_life
=
2713 jiffies_to_msecs(sctp_valid_cookie_life
);
2715 /* Initialize default event subscriptions. By default, all the
2718 memset(&sp
->subscribe
, 0, sizeof(struct sctp_event_subscribe
));
2720 /* Default Peer Address Parameters. These defaults can
2721 * be modified via SCTP_PEER_ADDR_PARAMS
2723 sp
->paddrparam
.spp_hbinterval
= jiffies_to_msecs(sctp_hb_interval
);
2724 sp
->paddrparam
.spp_pathmaxrxt
= sctp_max_retrans_path
;
2726 /* If enabled no SCTP message fragmentation will be performed.
2727 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
2729 sp
->disable_fragments
= 0;
2731 /* Turn on/off any Nagle-like algorithm. */
2734 /* Enable by default. */
2737 /* Auto-close idle associations after the configured
2738 * number of seconds. A value of 0 disables this
2739 * feature. Configure through the SCTP_AUTOCLOSE socket option,
2740 * for UDP-style sockets only.
2744 /* User specified fragmentation limit. */
2747 sp
->adaption_ind
= 0;
2749 sp
->pf
= sctp_get_pf_specific(sk
->sk_family
);
2751 /* Control variables for partial data delivery. */
2753 skb_queue_head_init(&sp
->pd_lobby
);
2755 /* Create a per socket endpoint structure. Even if we
2756 * change the data structure relationships, this may still
2757 * be useful for storing pre-connect address information.
2759 ep
= sctp_endpoint_new(sk
, GFP_KERNEL
);
2766 SCTP_DBG_OBJCNT_INC(sock
);
2770 /* Cleanup any SCTP per socket resources. */
2771 SCTP_STATIC
int sctp_destroy_sock(struct sock
*sk
)
2773 struct sctp_endpoint
*ep
;
2775 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk
);
2777 /* Release our hold on the endpoint. */
2778 ep
= sctp_sk(sk
)->ep
;
2779 sctp_endpoint_free(ep
);
2784 /* API 4.1.7 shutdown() - TCP Style Syntax
2785 * int shutdown(int socket, int how);
2787 * sd - the socket descriptor of the association to be closed.
2788 * how - Specifies the type of shutdown. The values are
2791 * Disables further receive operations. No SCTP
2792 * protocol action is taken.
2794 * Disables further send operations, and initiates
2795 * the SCTP shutdown sequence.
2797 * Disables further send and receive operations
2798 * and initiates the SCTP shutdown sequence.
2800 SCTP_STATIC
void sctp_shutdown(struct sock
*sk
, int how
)
2802 struct sctp_endpoint
*ep
;
2803 struct sctp_association
*asoc
;
2805 if (!sctp_style(sk
, TCP
))
2808 if (how
& SEND_SHUTDOWN
) {
2809 ep
= sctp_sk(sk
)->ep
;
2810 if (!list_empty(&ep
->asocs
)) {
2811 asoc
= list_entry(ep
->asocs
.next
,
2812 struct sctp_association
, asocs
);
2813 sctp_primitive_SHUTDOWN(asoc
, NULL
);
2818 /* 7.2.1 Association Status (SCTP_STATUS)
2820 * Applications can retrieve current status information about an
2821 * association, including association state, peer receiver window size,
2822 * number of unacked data chunks, and number of data chunks pending
2823 * receipt. This information is read-only.
2825 static int sctp_getsockopt_sctp_status(struct sock
*sk
, int len
,
2826 char __user
*optval
,
2829 struct sctp_status status
;
2830 struct sctp_association
*asoc
= NULL
;
2831 struct sctp_transport
*transport
;
2832 sctp_assoc_t associd
;
2835 if (len
!= sizeof(status
)) {
2840 if (copy_from_user(&status
, optval
, sizeof(status
))) {
2845 associd
= status
.sstat_assoc_id
;
2846 asoc
= sctp_id2assoc(sk
, associd
);
2852 transport
= asoc
->peer
.primary_path
;
2854 status
.sstat_assoc_id
= sctp_assoc2id(asoc
);
2855 status
.sstat_state
= asoc
->state
;
2856 status
.sstat_rwnd
= asoc
->peer
.rwnd
;
2857 status
.sstat_unackdata
= asoc
->unack_data
;
2859 status
.sstat_penddata
= sctp_tsnmap_pending(&asoc
->peer
.tsn_map
);
2860 status
.sstat_instrms
= asoc
->c
.sinit_max_instreams
;
2861 status
.sstat_outstrms
= asoc
->c
.sinit_num_ostreams
;
2862 status
.sstat_fragmentation_point
= asoc
->frag_point
;
2863 status
.sstat_primary
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
2864 memcpy(&status
.sstat_primary
.spinfo_address
,
2865 &(transport
->ipaddr
), sizeof(union sctp_addr
));
2866 /* Map ipv4 address into v4-mapped-on-v6 address. */
2867 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
2868 (union sctp_addr
*)&status
.sstat_primary
.spinfo_address
);
2869 status
.sstat_primary
.spinfo_state
= transport
->state
;
2870 status
.sstat_primary
.spinfo_cwnd
= transport
->cwnd
;
2871 status
.sstat_primary
.spinfo_srtt
= transport
->srtt
;
2872 status
.sstat_primary
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
2873 status
.sstat_primary
.spinfo_mtu
= transport
->pmtu
;
2875 if (status
.sstat_primary
.spinfo_state
== SCTP_UNKNOWN
)
2876 status
.sstat_primary
.spinfo_state
= SCTP_ACTIVE
;
2878 if (put_user(len
, optlen
)) {
2883 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
2884 len
, status
.sstat_state
, status
.sstat_rwnd
,
2885 status
.sstat_assoc_id
);
2887 if (copy_to_user(optval
, &status
, len
)) {
2897 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
2899 * Applications can retrieve information about a specific peer address
2900 * of an association, including its reachability state, congestion
2901 * window, and retransmission timer values. This information is
2904 static int sctp_getsockopt_peer_addr_info(struct sock
*sk
, int len
,
2905 char __user
*optval
,
2908 struct sctp_paddrinfo pinfo
;
2909 struct sctp_transport
*transport
;
2912 if (len
!= sizeof(pinfo
)) {
2917 if (copy_from_user(&pinfo
, optval
, sizeof(pinfo
))) {
2922 transport
= sctp_addr_id2transport(sk
, &pinfo
.spinfo_address
,
2923 pinfo
.spinfo_assoc_id
);
2927 pinfo
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
2928 pinfo
.spinfo_state
= transport
->state
;
2929 pinfo
.spinfo_cwnd
= transport
->cwnd
;
2930 pinfo
.spinfo_srtt
= transport
->srtt
;
2931 pinfo
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
2932 pinfo
.spinfo_mtu
= transport
->pmtu
;
2934 if (pinfo
.spinfo_state
== SCTP_UNKNOWN
)
2935 pinfo
.spinfo_state
= SCTP_ACTIVE
;
2937 if (put_user(len
, optlen
)) {
2942 if (copy_to_user(optval
, &pinfo
, len
)) {
2951 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2953 * This option is a on/off flag. If enabled no SCTP message
2954 * fragmentation will be performed. Instead if a message being sent
2955 * exceeds the current PMTU size, the message will NOT be sent and
2956 * instead a error will be indicated to the user.
2958 static int sctp_getsockopt_disable_fragments(struct sock
*sk
, int len
,
2959 char __user
*optval
, int __user
*optlen
)
2963 if (len
< sizeof(int))
2967 val
= (sctp_sk(sk
)->disable_fragments
== 1);
2968 if (put_user(len
, optlen
))
2970 if (copy_to_user(optval
, &val
, len
))
2975 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
2977 * This socket option is used to specify various notifications and
2978 * ancillary data the user wishes to receive.
2980 static int sctp_getsockopt_events(struct sock
*sk
, int len
, char __user
*optval
,
2983 if (len
!= sizeof(struct sctp_event_subscribe
))
2985 if (copy_to_user(optval
, &sctp_sk(sk
)->subscribe
, len
))
2990 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2992 * This socket option is applicable to the UDP-style socket only. When
2993 * set it will cause associations that are idle for more than the
2994 * specified number of seconds to automatically close. An association
2995 * being idle is defined an association that has NOT sent or received
2996 * user data. The special value of '0' indicates that no automatic
2997 * close of any associations should be performed. The option expects an
2998 * integer defining the number of seconds of idle time before an
2999 * association is closed.
3001 static int sctp_getsockopt_autoclose(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
3003 /* Applicable to UDP-style socket only */
3004 if (sctp_style(sk
, TCP
))
3006 if (len
!= sizeof(int))
3008 if (copy_to_user(optval
, &sctp_sk(sk
)->autoclose
, len
))
3013 /* Helper routine to branch off an association to a new socket. */
3014 SCTP_STATIC
int sctp_do_peeloff(struct sctp_association
*asoc
,
3015 struct socket
**sockp
)
3017 struct sock
*sk
= asoc
->base
.sk
;
3018 struct socket
*sock
;
3021 /* An association cannot be branched off from an already peeled-off
3022 * socket, nor is this supported for tcp style sockets.
3024 if (!sctp_style(sk
, UDP
))
3027 /* Create a new socket. */
3028 err
= sock_create(sk
->sk_family
, SOCK_SEQPACKET
, IPPROTO_SCTP
, &sock
);
3032 /* Populate the fields of the newsk from the oldsk and migrate the
3033 * asoc to the newsk.
3035 sctp_sock_migrate(sk
, sock
->sk
, asoc
, SCTP_SOCKET_UDP_HIGH_BANDWIDTH
);
3041 static int sctp_getsockopt_peeloff(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
3043 sctp_peeloff_arg_t peeloff
;
3044 struct socket
*newsock
;
3046 struct sctp_association
*asoc
;
3048 if (len
!= sizeof(sctp_peeloff_arg_t
))
3050 if (copy_from_user(&peeloff
, optval
, len
))
3053 asoc
= sctp_id2assoc(sk
, peeloff
.associd
);
3059 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __FUNCTION__
, sk
, asoc
);
3061 retval
= sctp_do_peeloff(asoc
, &newsock
);
3065 /* Map the socket to an unused fd that can be returned to the user. */
3066 retval
= sock_map_fd(newsock
);
3068 sock_release(newsock
);
3072 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3073 __FUNCTION__
, sk
, asoc
, newsock
->sk
, retval
);
3075 /* Return the fd mapped to the new socket. */
3076 peeloff
.sd
= retval
;
3077 if (copy_to_user(optval
, &peeloff
, len
))
3084 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
3086 * Applications can enable or disable heartbeats for any peer address of
3087 * an association, modify an address's heartbeat interval, force a
3088 * heartbeat to be sent immediately, and adjust the address's maximum
3089 * number of retransmissions sent before an address is considered
3090 * unreachable. The following structure is used to access and modify an
3091 * address's parameters:
3093 * struct sctp_paddrparams {
3094 * sctp_assoc_t spp_assoc_id;
3095 * struct sockaddr_storage spp_address;
3096 * uint32_t spp_hbinterval;
3097 * uint16_t spp_pathmaxrxt;
3100 * spp_assoc_id - (UDP style socket) This is filled in the application,
3101 * and identifies the association for this query.
3102 * spp_address - This specifies which address is of interest.
3103 * spp_hbinterval - This contains the value of the heartbeat interval,
3104 * in milliseconds. A value of 0, when modifying the
3105 * parameter, specifies that the heartbeat on this
3106 * address should be disabled. A value of UINT32_MAX
3107 * (4294967295), when modifying the parameter,
3108 * specifies that a heartbeat should be sent
3109 * immediately to the peer address, and the current
3110 * interval should remain unchanged.
3111 * spp_pathmaxrxt - This contains the maximum number of
3112 * retransmissions before this address shall be
3113 * considered unreachable.
3115 static int sctp_getsockopt_peer_addr_params(struct sock
*sk
, int len
,
3116 char __user
*optval
, int __user
*optlen
)
3118 struct sctp_paddrparams params
;
3119 struct sctp_transport
*trans
;
3121 if (len
!= sizeof(struct sctp_paddrparams
))
3123 if (copy_from_user(¶ms
, optval
, len
))
3126 /* If no association id is specified retrieve the default value
3127 * for the endpoint that will be used for all future associations
3129 if (!params
.spp_assoc_id
&&
3130 sctp_is_any(( union sctp_addr
*)¶ms
.spp_address
)) {
3131 params
.spp_hbinterval
= sctp_sk(sk
)->paddrparam
.spp_hbinterval
;
3132 params
.spp_pathmaxrxt
= sctp_sk(sk
)->paddrparam
.spp_pathmaxrxt
;
3137 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
3138 params
.spp_assoc_id
);
3142 /* The value of the heartbeat interval, in milliseconds. A value of 0,
3143 * when modifying the parameter, specifies that the heartbeat on this
3144 * address should be disabled.
3146 if (!trans
->hb_allowed
)
3147 params
.spp_hbinterval
= 0;
3149 params
.spp_hbinterval
= jiffies_to_msecs(trans
->hb_interval
);
3151 /* spp_pathmaxrxt contains the maximum number of retransmissions
3152 * before this address shall be considered unreachable.
3154 params
.spp_pathmaxrxt
= trans
->max_retrans
;
3157 if (copy_to_user(optval
, ¶ms
, len
))
3160 if (put_user(len
, optlen
))
3166 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
3168 * Applications can specify protocol parameters for the default association
3169 * initialization. The option name argument to setsockopt() and getsockopt()
3172 * Setting initialization parameters is effective only on an unconnected
3173 * socket (for UDP-style sockets only future associations are effected
3174 * by the change). With TCP-style sockets, this option is inherited by
3175 * sockets derived from a listener socket.
3177 static int sctp_getsockopt_initmsg(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
3179 if (len
!= sizeof(struct sctp_initmsg
))
3181 if (copy_to_user(optval
, &sctp_sk(sk
)->initmsg
, len
))
3186 static int sctp_getsockopt_peer_addrs_num_old(struct sock
*sk
, int len
,
3187 char __user
*optval
,
3191 struct sctp_association
*asoc
;
3192 struct list_head
*pos
;
3195 if (len
!= sizeof(sctp_assoc_t
))
3198 if (copy_from_user(&id
, optval
, sizeof(sctp_assoc_t
)))
3201 /* For UDP-style sockets, id specifies the association to query. */
3202 asoc
= sctp_id2assoc(sk
, id
);
3206 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
3214 * Old API for getting list of peer addresses. Does not work for 32-bit
3215 * programs running on a 64-bit kernel
3217 static int sctp_getsockopt_peer_addrs_old(struct sock
*sk
, int len
,
3218 char __user
*optval
,
3221 struct sctp_association
*asoc
;
3222 struct list_head
*pos
;
3224 struct sctp_getaddrs_old getaddrs
;
3225 struct sctp_transport
*from
;
3227 union sctp_addr temp
;
3228 struct sctp_sock
*sp
= sctp_sk(sk
);
3231 if (len
!= sizeof(struct sctp_getaddrs_old
))
3234 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs_old
)))
3237 if (getaddrs
.addr_num
<= 0) return -EINVAL
;
3239 /* For UDP-style sockets, id specifies the association to query. */
3240 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
3244 to
= (void __user
*)getaddrs
.addrs
;
3245 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
3246 from
= list_entry(pos
, struct sctp_transport
, transports
);
3247 memcpy(&temp
, &from
->ipaddr
, sizeof(temp
));
3248 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
3249 addrlen
= sctp_get_af_specific(sk
->sk_family
)->sockaddr_len
;
3250 temp
.v4
.sin_port
= htons(temp
.v4
.sin_port
);
3251 if (copy_to_user(to
, &temp
, addrlen
))
3255 if (cnt
>= getaddrs
.addr_num
) break;
3257 getaddrs
.addr_num
= cnt
;
3258 if (copy_to_user(optval
, &getaddrs
, sizeof(struct sctp_getaddrs_old
)))
3264 static int sctp_getsockopt_peer_addrs(struct sock
*sk
, int len
,
3265 char __user
*optval
, int __user
*optlen
)
3267 struct sctp_association
*asoc
;
3268 struct list_head
*pos
;
3270 struct sctp_getaddrs getaddrs
;
3271 struct sctp_transport
*from
;
3273 union sctp_addr temp
;
3274 struct sctp_sock
*sp
= sctp_sk(sk
);
3279 if (len
< sizeof(struct sctp_getaddrs
))
3282 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
3285 /* For UDP-style sockets, id specifies the association to query. */
3286 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
3290 to
= optval
+ offsetof(struct sctp_getaddrs
,addrs
);
3291 space_left
= len
- sizeof(struct sctp_getaddrs
) -
3292 offsetof(struct sctp_getaddrs
,addrs
);
3294 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
3295 from
= list_entry(pos
, struct sctp_transport
, transports
);
3296 memcpy(&temp
, &from
->ipaddr
, sizeof(temp
));
3297 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
3298 addrlen
= sctp_get_af_specific(sk
->sk_family
)->sockaddr_len
;
3299 if(space_left
< addrlen
)
3301 temp
.v4
.sin_port
= htons(temp
.v4
.sin_port
);
3302 if (copy_to_user(to
, &temp
, addrlen
))
3306 space_left
-= addrlen
;
3309 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
))
3311 bytes_copied
= ((char __user
*)to
) - optval
;
3312 if (put_user(bytes_copied
, optlen
))
3318 static int sctp_getsockopt_local_addrs_num_old(struct sock
*sk
, int len
,
3319 char __user
*optval
,
3323 struct sctp_bind_addr
*bp
;
3324 struct sctp_association
*asoc
;
3325 struct list_head
*pos
;
3326 struct sctp_sockaddr_entry
*addr
;
3327 rwlock_t
*addr_lock
;
3328 unsigned long flags
;
3331 if (len
!= sizeof(sctp_assoc_t
))
3334 if (copy_from_user(&id
, optval
, sizeof(sctp_assoc_t
)))
3338 * For UDP-style sockets, id specifies the association to query.
3339 * If the id field is set to the value '0' then the locally bound
3340 * addresses are returned without regard to any particular
3344 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
3345 addr_lock
= &sctp_sk(sk
)->ep
->base
.addr_lock
;
3347 asoc
= sctp_id2assoc(sk
, id
);
3350 bp
= &asoc
->base
.bind_addr
;
3351 addr_lock
= &asoc
->base
.addr_lock
;
3354 sctp_read_lock(addr_lock
);
3356 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
3357 * addresses from the global local address list.
3359 if (sctp_list_single_entry(&bp
->address_list
)) {
3360 addr
= list_entry(bp
->address_list
.next
,
3361 struct sctp_sockaddr_entry
, list
);
3362 if (sctp_is_any(&addr
->a
)) {
3363 sctp_spin_lock_irqsave(&sctp_local_addr_lock
, flags
);
3364 list_for_each(pos
, &sctp_local_addr_list
) {
3365 addr
= list_entry(pos
,
3366 struct sctp_sockaddr_entry
,
3368 if ((PF_INET
== sk
->sk_family
) &&
3369 (AF_INET6
== addr
->a
.sa
.sa_family
))
3373 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock
,
3381 list_for_each(pos
, &bp
->address_list
) {
3386 sctp_read_unlock(addr_lock
);
3390 /* Helper function that copies local addresses to user and returns the number
3391 * of addresses copied.
3393 static int sctp_copy_laddrs_to_user_old(struct sock
*sk
, __u16 port
, int max_addrs
,
3396 struct list_head
*pos
;
3397 struct sctp_sockaddr_entry
*addr
;
3398 unsigned long flags
;
3399 union sctp_addr temp
;
3403 sctp_spin_lock_irqsave(&sctp_local_addr_lock
, flags
);
3404 list_for_each(pos
, &sctp_local_addr_list
) {
3405 addr
= list_entry(pos
, struct sctp_sockaddr_entry
, list
);
3406 if ((PF_INET
== sk
->sk_family
) &&
3407 (AF_INET6
== addr
->a
.sa
.sa_family
))
3409 memcpy(&temp
, &addr
->a
, sizeof(temp
));
3410 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
3412 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
3413 temp
.v4
.sin_port
= htons(port
);
3414 if (copy_to_user(to
, &temp
, addrlen
)) {
3415 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock
,
3421 if (cnt
>= max_addrs
) break;
3423 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock
, flags
);
3428 static int sctp_copy_laddrs_to_user(struct sock
*sk
, __u16 port
,
3429 void * __user
*to
, size_t space_left
)
3431 struct list_head
*pos
;
3432 struct sctp_sockaddr_entry
*addr
;
3433 unsigned long flags
;
3434 union sctp_addr temp
;
3438 sctp_spin_lock_irqsave(&sctp_local_addr_lock
, flags
);
3439 list_for_each(pos
, &sctp_local_addr_list
) {
3440 addr
= list_entry(pos
, struct sctp_sockaddr_entry
, list
);
3441 if ((PF_INET
== sk
->sk_family
) &&
3442 (AF_INET6
== addr
->a
.sa
.sa_family
))
3444 memcpy(&temp
, &addr
->a
, sizeof(temp
));
3445 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
3447 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
3448 if(space_left
<addrlen
)
3450 temp
.v4
.sin_port
= htons(port
);
3451 if (copy_to_user(*to
, &temp
, addrlen
)) {
3452 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock
,
3458 space_left
-= addrlen
;
3460 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock
, flags
);
3465 /* Old API for getting list of local addresses. Does not work for 32-bit
3466 * programs running on a 64-bit kernel
3468 static int sctp_getsockopt_local_addrs_old(struct sock
*sk
, int len
,
3469 char __user
*optval
, int __user
*optlen
)
3471 struct sctp_bind_addr
*bp
;
3472 struct sctp_association
*asoc
;
3473 struct list_head
*pos
;
3475 struct sctp_getaddrs_old getaddrs
;
3476 struct sctp_sockaddr_entry
*addr
;
3478 union sctp_addr temp
;
3479 struct sctp_sock
*sp
= sctp_sk(sk
);
3481 rwlock_t
*addr_lock
;
3484 if (len
!= sizeof(struct sctp_getaddrs_old
))
3487 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs_old
)))
3490 if (getaddrs
.addr_num
<= 0) return -EINVAL
;
3492 * For UDP-style sockets, id specifies the association to query.
3493 * If the id field is set to the value '0' then the locally bound
3494 * addresses are returned without regard to any particular
3497 if (0 == getaddrs
.assoc_id
) {
3498 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
3499 addr_lock
= &sctp_sk(sk
)->ep
->base
.addr_lock
;
3501 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
3504 bp
= &asoc
->base
.bind_addr
;
3505 addr_lock
= &asoc
->base
.addr_lock
;
3508 to
= getaddrs
.addrs
;
3510 sctp_read_lock(addr_lock
);
3512 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
3513 * addresses from the global local address list.
3515 if (sctp_list_single_entry(&bp
->address_list
)) {
3516 addr
= list_entry(bp
->address_list
.next
,
3517 struct sctp_sockaddr_entry
, list
);
3518 if (sctp_is_any(&addr
->a
)) {
3519 cnt
= sctp_copy_laddrs_to_user_old(sk
, bp
->port
,
3530 list_for_each(pos
, &bp
->address_list
) {
3531 addr
= list_entry(pos
, struct sctp_sockaddr_entry
, list
);
3532 memcpy(&temp
, &addr
->a
, sizeof(temp
));
3533 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
3534 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
3535 temp
.v4
.sin_port
= htons(temp
.v4
.sin_port
);
3536 if (copy_to_user(to
, &temp
, addrlen
)) {
3542 if (cnt
>= getaddrs
.addr_num
) break;
3546 getaddrs
.addr_num
= cnt
;
3547 if (copy_to_user(optval
, &getaddrs
, sizeof(struct sctp_getaddrs_old
)))
3551 sctp_read_unlock(addr_lock
);
3555 static int sctp_getsockopt_local_addrs(struct sock
*sk
, int len
,
3556 char __user
*optval
, int __user
*optlen
)
3558 struct sctp_bind_addr
*bp
;
3559 struct sctp_association
*asoc
;
3560 struct list_head
*pos
;
3562 struct sctp_getaddrs getaddrs
;
3563 struct sctp_sockaddr_entry
*addr
;
3565 union sctp_addr temp
;
3566 struct sctp_sock
*sp
= sctp_sk(sk
);
3568 rwlock_t
*addr_lock
;
3573 if (len
<= sizeof(struct sctp_getaddrs
))
3576 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
3580 * For UDP-style sockets, id specifies the association to query.
3581 * If the id field is set to the value '0' then the locally bound
3582 * addresses are returned without regard to any particular
3585 if (0 == getaddrs
.assoc_id
) {
3586 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
3587 addr_lock
= &sctp_sk(sk
)->ep
->base
.addr_lock
;
3589 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
3592 bp
= &asoc
->base
.bind_addr
;
3593 addr_lock
= &asoc
->base
.addr_lock
;
3596 to
= optval
+ offsetof(struct sctp_getaddrs
,addrs
);
3597 space_left
= len
- sizeof(struct sctp_getaddrs
) -
3598 offsetof(struct sctp_getaddrs
,addrs
);
3600 sctp_read_lock(addr_lock
);
3602 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
3603 * addresses from the global local address list.
3605 if (sctp_list_single_entry(&bp
->address_list
)) {
3606 addr
= list_entry(bp
->address_list
.next
,
3607 struct sctp_sockaddr_entry
, list
);
3608 if (sctp_is_any(&addr
->a
)) {
3609 cnt
= sctp_copy_laddrs_to_user(sk
, bp
->port
,
3619 list_for_each(pos
, &bp
->address_list
) {
3620 addr
= list_entry(pos
, struct sctp_sockaddr_entry
, list
);
3621 memcpy(&temp
, &addr
->a
, sizeof(temp
));
3622 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
3623 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
3624 if(space_left
< addrlen
)
3625 return -ENOMEM
; /*fixme: right error?*/
3626 temp
.v4
.sin_port
= htons(temp
.v4
.sin_port
);
3627 if (copy_to_user(to
, &temp
, addrlen
)) {
3633 space_left
-= addrlen
;
3637 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
))
3639 bytes_copied
= ((char __user
*)to
) - optval
;
3640 if (put_user(bytes_copied
, optlen
))
3644 sctp_read_unlock(addr_lock
);
3648 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
3650 * Requests that the local SCTP stack use the enclosed peer address as
3651 * the association primary. The enclosed address must be one of the
3652 * association peer's addresses.
3654 static int sctp_getsockopt_primary_addr(struct sock
*sk
, int len
,
3655 char __user
*optval
, int __user
*optlen
)
3657 struct sctp_prim prim
;
3658 struct sctp_association
*asoc
;
3659 struct sctp_sock
*sp
= sctp_sk(sk
);
3661 if (len
!= sizeof(struct sctp_prim
))
3664 if (copy_from_user(&prim
, optval
, sizeof(struct sctp_prim
)))
3667 asoc
= sctp_id2assoc(sk
, prim
.ssp_assoc_id
);
3671 if (!asoc
->peer
.primary_path
)
3674 asoc
->peer
.primary_path
->ipaddr
.v4
.sin_port
=
3675 htons(asoc
->peer
.primary_path
->ipaddr
.v4
.sin_port
);
3676 memcpy(&prim
.ssp_addr
, &asoc
->peer
.primary_path
->ipaddr
,
3677 sizeof(union sctp_addr
));
3678 asoc
->peer
.primary_path
->ipaddr
.v4
.sin_port
=
3679 ntohs(asoc
->peer
.primary_path
->ipaddr
.v4
.sin_port
);
3681 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
,
3682 (union sctp_addr
*)&prim
.ssp_addr
);
3684 if (copy_to_user(optval
, &prim
, sizeof(struct sctp_prim
)))
3691 * 7.1.11 Set Adaption Layer Indicator (SCTP_ADAPTION_LAYER)
3693 * Requests that the local endpoint set the specified Adaption Layer
3694 * Indication parameter for all future INIT and INIT-ACK exchanges.
3696 static int sctp_getsockopt_adaption_layer(struct sock
*sk
, int len
,
3697 char __user
*optval
, int __user
*optlen
)
3699 struct sctp_setadaption adaption
;
3701 if (len
!= sizeof(struct sctp_setadaption
))
3704 adaption
.ssb_adaption_ind
= sctp_sk(sk
)->adaption_ind
;
3705 if (copy_to_user(optval
, &adaption
, len
))
3713 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
3715 * Applications that wish to use the sendto() system call may wish to
3716 * specify a default set of parameters that would normally be supplied
3717 * through the inclusion of ancillary data. This socket option allows
3718 * such an application to set the default sctp_sndrcvinfo structure.
3721 * The application that wishes to use this socket option simply passes
3722 * in to this call the sctp_sndrcvinfo structure defined in Section
3723 * 5.2.2) The input parameters accepted by this call include
3724 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
3725 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
3726 * to this call if the caller is using the UDP model.
3728 * For getsockopt, it get the default sctp_sndrcvinfo structure.
3730 static int sctp_getsockopt_default_send_param(struct sock
*sk
,
3731 int len
, char __user
*optval
,
3734 struct sctp_sndrcvinfo info
;
3735 struct sctp_association
*asoc
;
3736 struct sctp_sock
*sp
= sctp_sk(sk
);
3738 if (len
!= sizeof(struct sctp_sndrcvinfo
))
3740 if (copy_from_user(&info
, optval
, sizeof(struct sctp_sndrcvinfo
)))
3743 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
3744 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
3748 info
.sinfo_stream
= asoc
->default_stream
;
3749 info
.sinfo_flags
= asoc
->default_flags
;
3750 info
.sinfo_ppid
= asoc
->default_ppid
;
3751 info
.sinfo_context
= asoc
->default_context
;
3752 info
.sinfo_timetolive
= asoc
->default_timetolive
;
3754 info
.sinfo_stream
= sp
->default_stream
;
3755 info
.sinfo_flags
= sp
->default_flags
;
3756 info
.sinfo_ppid
= sp
->default_ppid
;
3757 info
.sinfo_context
= sp
->default_context
;
3758 info
.sinfo_timetolive
= sp
->default_timetolive
;
3761 if (copy_to_user(optval
, &info
, sizeof(struct sctp_sndrcvinfo
)))
3769 * 7.1.5 SCTP_NODELAY
3771 * Turn on/off any Nagle-like algorithm. This means that packets are
3772 * generally sent as soon as possible and no unnecessary delays are
3773 * introduced, at the cost of more packets in the network. Expects an
3774 * integer boolean flag.
3777 static int sctp_getsockopt_nodelay(struct sock
*sk
, int len
,
3778 char __user
*optval
, int __user
*optlen
)
3782 if (len
< sizeof(int))
3786 val
= (sctp_sk(sk
)->nodelay
== 1);
3787 if (put_user(len
, optlen
))
3789 if (copy_to_user(optval
, &val
, len
))
3796 * 7.1.1 SCTP_RTOINFO
3798 * The protocol parameters used to initialize and bound retransmission
3799 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
3800 * and modify these parameters.
3801 * All parameters are time values, in milliseconds. A value of 0, when
3802 * modifying the parameters, indicates that the current value should not
3806 static int sctp_getsockopt_rtoinfo(struct sock
*sk
, int len
,
3807 char __user
*optval
,
3808 int __user
*optlen
) {
3809 struct sctp_rtoinfo rtoinfo
;
3810 struct sctp_association
*asoc
;
3812 if (len
!= sizeof (struct sctp_rtoinfo
))
3815 if (copy_from_user(&rtoinfo
, optval
, sizeof (struct sctp_rtoinfo
)))
3818 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
3820 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
3823 /* Values corresponding to the specific association. */
3825 rtoinfo
.srto_initial
= jiffies_to_msecs(asoc
->rto_initial
);
3826 rtoinfo
.srto_max
= jiffies_to_msecs(asoc
->rto_max
);
3827 rtoinfo
.srto_min
= jiffies_to_msecs(asoc
->rto_min
);
3829 /* Values corresponding to the endpoint. */
3830 struct sctp_sock
*sp
= sctp_sk(sk
);
3832 rtoinfo
.srto_initial
= sp
->rtoinfo
.srto_initial
;
3833 rtoinfo
.srto_max
= sp
->rtoinfo
.srto_max
;
3834 rtoinfo
.srto_min
= sp
->rtoinfo
.srto_min
;
3837 if (put_user(len
, optlen
))
3840 if (copy_to_user(optval
, &rtoinfo
, len
))
3848 * 7.1.2 SCTP_ASSOCINFO
3850 * This option is used to tune the the maximum retransmission attempts
3851 * of the association.
3852 * Returns an error if the new association retransmission value is
3853 * greater than the sum of the retransmission value of the peer.
3854 * See [SCTP] for more information.
3857 static int sctp_getsockopt_associnfo(struct sock
*sk
, int len
,
3858 char __user
*optval
,
3862 struct sctp_assocparams assocparams
;
3863 struct sctp_association
*asoc
;
3864 struct list_head
*pos
;
3867 if (len
!= sizeof (struct sctp_assocparams
))
3870 if (copy_from_user(&assocparams
, optval
,
3871 sizeof (struct sctp_assocparams
)))
3874 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
3876 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
3879 /* Values correspoinding to the specific association */
3881 assocparams
.sasoc_asocmaxrxt
= asoc
->max_retrans
;
3882 assocparams
.sasoc_peer_rwnd
= asoc
->peer
.rwnd
;
3883 assocparams
.sasoc_local_rwnd
= asoc
->a_rwnd
;
3884 assocparams
.sasoc_cookie_life
= (asoc
->cookie_life
.tv_sec
3886 (asoc
->cookie_life
.tv_usec
3889 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
3893 assocparams
.sasoc_number_peer_destinations
= cnt
;
3895 /* Values corresponding to the endpoint */
3896 struct sctp_sock
*sp
= sctp_sk(sk
);
3898 assocparams
.sasoc_asocmaxrxt
= sp
->assocparams
.sasoc_asocmaxrxt
;
3899 assocparams
.sasoc_peer_rwnd
= sp
->assocparams
.sasoc_peer_rwnd
;
3900 assocparams
.sasoc_local_rwnd
= sp
->assocparams
.sasoc_local_rwnd
;
3901 assocparams
.sasoc_cookie_life
=
3902 sp
->assocparams
.sasoc_cookie_life
;
3903 assocparams
.sasoc_number_peer_destinations
=
3905 sasoc_number_peer_destinations
;
3908 if (put_user(len
, optlen
))
3911 if (copy_to_user(optval
, &assocparams
, len
))
3918 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3920 * This socket option is a boolean flag which turns on or off mapped V4
3921 * addresses. If this option is turned on and the socket is type
3922 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3923 * If this option is turned off, then no mapping will be done of V4
3924 * addresses and a user will receive both PF_INET6 and PF_INET type
3925 * addresses on the socket.
3927 static int sctp_getsockopt_mappedv4(struct sock
*sk
, int len
,
3928 char __user
*optval
, int __user
*optlen
)
3931 struct sctp_sock
*sp
= sctp_sk(sk
);
3933 if (len
< sizeof(int))
3938 if (put_user(len
, optlen
))
3940 if (copy_to_user(optval
, &val
, len
))
3947 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
3949 * This socket option specifies the maximum size to put in any outgoing
3950 * SCTP chunk. If a message is larger than this size it will be
3951 * fragmented by SCTP into the specified size. Note that the underlying
3952 * SCTP implementation may fragment into smaller sized chunks when the
3953 * PMTU of the underlying association is smaller than the value set by
3956 static int sctp_getsockopt_maxseg(struct sock
*sk
, int len
,
3957 char __user
*optval
, int __user
*optlen
)
3961 if (len
< sizeof(int))
3966 val
= sctp_sk(sk
)->user_frag
;
3967 if (put_user(len
, optlen
))
3969 if (copy_to_user(optval
, &val
, len
))
3975 SCTP_STATIC
int sctp_getsockopt(struct sock
*sk
, int level
, int optname
,
3976 char __user
*optval
, int __user
*optlen
)
3981 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
3984 /* I can hardly begin to describe how wrong this is. This is
3985 * so broken as to be worse than useless. The API draft
3986 * REALLY is NOT helpful here... I am not convinced that the
3987 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
3988 * are at all well-founded.
3990 if (level
!= SOL_SCTP
) {
3991 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
3993 retval
= af
->getsockopt(sk
, level
, optname
, optval
, optlen
);
3997 if (get_user(len
, optlen
))
4004 retval
= sctp_getsockopt_sctp_status(sk
, len
, optval
, optlen
);
4006 case SCTP_DISABLE_FRAGMENTS
:
4007 retval
= sctp_getsockopt_disable_fragments(sk
, len
, optval
,
4011 retval
= sctp_getsockopt_events(sk
, len
, optval
, optlen
);
4013 case SCTP_AUTOCLOSE
:
4014 retval
= sctp_getsockopt_autoclose(sk
, len
, optval
, optlen
);
4016 case SCTP_SOCKOPT_PEELOFF
:
4017 retval
= sctp_getsockopt_peeloff(sk
, len
, optval
, optlen
);
4019 case SCTP_PEER_ADDR_PARAMS
:
4020 retval
= sctp_getsockopt_peer_addr_params(sk
, len
, optval
,
4024 retval
= sctp_getsockopt_initmsg(sk
, len
, optval
, optlen
);
4026 case SCTP_GET_PEER_ADDRS_NUM_OLD
:
4027 retval
= sctp_getsockopt_peer_addrs_num_old(sk
, len
, optval
,
4030 case SCTP_GET_LOCAL_ADDRS_NUM_OLD
:
4031 retval
= sctp_getsockopt_local_addrs_num_old(sk
, len
, optval
,
4034 case SCTP_GET_PEER_ADDRS_OLD
:
4035 retval
= sctp_getsockopt_peer_addrs_old(sk
, len
, optval
,
4038 case SCTP_GET_LOCAL_ADDRS_OLD
:
4039 retval
= sctp_getsockopt_local_addrs_old(sk
, len
, optval
,
4042 case SCTP_GET_PEER_ADDRS
:
4043 retval
= sctp_getsockopt_peer_addrs(sk
, len
, optval
,
4046 case SCTP_GET_LOCAL_ADDRS
:
4047 retval
= sctp_getsockopt_local_addrs(sk
, len
, optval
,
4050 case SCTP_DEFAULT_SEND_PARAM
:
4051 retval
= sctp_getsockopt_default_send_param(sk
, len
,
4054 case SCTP_PRIMARY_ADDR
:
4055 retval
= sctp_getsockopt_primary_addr(sk
, len
, optval
, optlen
);
4058 retval
= sctp_getsockopt_nodelay(sk
, len
, optval
, optlen
);
4061 retval
= sctp_getsockopt_rtoinfo(sk
, len
, optval
, optlen
);
4063 case SCTP_ASSOCINFO
:
4064 retval
= sctp_getsockopt_associnfo(sk
, len
, optval
, optlen
);
4066 case SCTP_I_WANT_MAPPED_V4_ADDR
:
4067 retval
= sctp_getsockopt_mappedv4(sk
, len
, optval
, optlen
);
4070 retval
= sctp_getsockopt_maxseg(sk
, len
, optval
, optlen
);
4072 case SCTP_GET_PEER_ADDR_INFO
:
4073 retval
= sctp_getsockopt_peer_addr_info(sk
, len
, optval
,
4076 case SCTP_ADAPTION_LAYER
:
4077 retval
= sctp_getsockopt_adaption_layer(sk
, len
, optval
,
4081 retval
= -ENOPROTOOPT
;
4085 sctp_release_sock(sk
);
4089 static void sctp_hash(struct sock
*sk
)
4094 static void sctp_unhash(struct sock
*sk
)
4099 /* Check if port is acceptable. Possibly find first available port.
4101 * The port hash table (contained in the 'global' SCTP protocol storage
4102 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
4103 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
4104 * list (the list number is the port number hashed out, so as you
4105 * would expect from a hash function, all the ports in a given list have
4106 * such a number that hashes out to the same list number; you were
4107 * expecting that, right?); so each list has a set of ports, with a
4108 * link to the socket (struct sock) that uses it, the port number and
4109 * a fastreuse flag (FIXME: NPI ipg).
4111 static struct sctp_bind_bucket
*sctp_bucket_create(
4112 struct sctp_bind_hashbucket
*head
, unsigned short snum
);
4114 static long sctp_get_port_local(struct sock
*sk
, union sctp_addr
*addr
)
4116 struct sctp_bind_hashbucket
*head
; /* hash list */
4117 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
4118 unsigned short snum
;
4121 /* NOTE: Remember to put this back to net order. */
4122 addr
->v4
.sin_port
= ntohs(addr
->v4
.sin_port
);
4123 snum
= addr
->v4
.sin_port
;
4125 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum
);
4126 sctp_local_bh_disable();
4129 /* Search for an available port.
4131 * 'sctp_port_rover' was the last port assigned, so
4132 * we start to search from 'sctp_port_rover +
4133 * 1'. What we do is first check if port 'rover' is
4134 * already in the hash table; if not, we use that; if
4135 * it is, we try next.
4137 int low
= sysctl_local_port_range
[0];
4138 int high
= sysctl_local_port_range
[1];
4139 int remaining
= (high
- low
) + 1;
4143 sctp_spin_lock(&sctp_port_alloc_lock
);
4144 rover
= sctp_port_rover
;
4147 if ((rover
< low
) || (rover
> high
))
4149 index
= sctp_phashfn(rover
);
4150 head
= &sctp_port_hashtable
[index
];
4151 sctp_spin_lock(&head
->lock
);
4152 for (pp
= head
->chain
; pp
; pp
= pp
->next
)
4153 if (pp
->port
== rover
)
4157 sctp_spin_unlock(&head
->lock
);
4158 } while (--remaining
> 0);
4159 sctp_port_rover
= rover
;
4160 sctp_spin_unlock(&sctp_port_alloc_lock
);
4162 /* Exhausted local port range during search? */
4167 /* OK, here is the one we will use. HEAD (the port
4168 * hash table list entry) is non-NULL and we hold it's
4173 /* We are given an specific port number; we verify
4174 * that it is not being used. If it is used, we will
4175 * exahust the search in the hash list corresponding
4176 * to the port number (snum) - we detect that with the
4177 * port iterator, pp being NULL.
4179 head
= &sctp_port_hashtable
[sctp_phashfn(snum
)];
4180 sctp_spin_lock(&head
->lock
);
4181 for (pp
= head
->chain
; pp
; pp
= pp
->next
) {
4182 if (pp
->port
== snum
)
4189 if (!hlist_empty(&pp
->owner
)) {
4190 /* We had a port hash table hit - there is an
4191 * available port (pp != NULL) and it is being
4192 * used by other socket (pp->owner not empty); that other
4193 * socket is going to be sk2.
4195 int reuse
= sk
->sk_reuse
;
4197 struct hlist_node
*node
;
4199 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
4200 if (pp
->fastreuse
&& sk
->sk_reuse
)
4203 /* Run through the list of sockets bound to the port
4204 * (pp->port) [via the pointers bind_next and
4205 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
4206 * we get the endpoint they describe and run through
4207 * the endpoint's list of IP (v4 or v6) addresses,
4208 * comparing each of the addresses with the address of
4209 * the socket sk. If we find a match, then that means
4210 * that this port/socket (sk) combination are already
4213 sk_for_each_bound(sk2
, node
, &pp
->owner
) {
4214 struct sctp_endpoint
*ep2
;
4215 ep2
= sctp_sk(sk2
)->ep
;
4217 if (reuse
&& sk2
->sk_reuse
)
4220 if (sctp_bind_addr_match(&ep2
->base
.bind_addr
, addr
,
4226 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
4229 /* If there was a hash table miss, create a new port. */
4231 if (!pp
&& !(pp
= sctp_bucket_create(head
, snum
)))
4234 /* In either case (hit or miss), make sure fastreuse is 1 only
4235 * if sk->sk_reuse is too (that is, if the caller requested
4236 * SO_REUSEADDR on this socket -sk-).
4238 if (hlist_empty(&pp
->owner
))
4239 pp
->fastreuse
= sk
->sk_reuse
? 1 : 0;
4240 else if (pp
->fastreuse
&& !sk
->sk_reuse
)
4243 /* We are set, so fill up all the data in the hash table
4244 * entry, tie the socket list information with the rest of the
4245 * sockets FIXME: Blurry, NPI (ipg).
4248 inet_sk(sk
)->num
= snum
;
4249 if (!sctp_sk(sk
)->bind_hash
) {
4250 sk_add_bind_node(sk
, &pp
->owner
);
4251 sctp_sk(sk
)->bind_hash
= pp
;
4256 sctp_spin_unlock(&head
->lock
);
4259 sctp_local_bh_enable();
4260 addr
->v4
.sin_port
= htons(addr
->v4
.sin_port
);
4264 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
4265 * port is requested.
4267 static int sctp_get_port(struct sock
*sk
, unsigned short snum
)
4270 union sctp_addr addr
;
4271 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
4273 /* Set up a dummy address struct from the sk. */
4274 af
->from_sk(&addr
, sk
);
4275 addr
.v4
.sin_port
= htons(snum
);
4277 /* Note: sk->sk_num gets filled in if ephemeral port request. */
4278 ret
= sctp_get_port_local(sk
, &addr
);
4280 return (ret
? 1 : 0);
4284 * 3.1.3 listen() - UDP Style Syntax
4286 * By default, new associations are not accepted for UDP style sockets.
4287 * An application uses listen() to mark a socket as being able to
4288 * accept new associations.
4290 SCTP_STATIC
int sctp_seqpacket_listen(struct sock
*sk
, int backlog
)
4292 struct sctp_sock
*sp
= sctp_sk(sk
);
4293 struct sctp_endpoint
*ep
= sp
->ep
;
4295 /* Only UDP style sockets that are not peeled off are allowed to
4298 if (!sctp_style(sk
, UDP
))
4301 /* If backlog is zero, disable listening. */
4303 if (sctp_sstate(sk
, CLOSED
))
4306 sctp_unhash_endpoint(ep
);
4307 sk
->sk_state
= SCTP_SS_CLOSED
;
4310 /* Return if we are already listening. */
4311 if (sctp_sstate(sk
, LISTENING
))
4315 * If a bind() or sctp_bindx() is not called prior to a listen()
4316 * call that allows new associations to be accepted, the system
4317 * picks an ephemeral port and will choose an address set equivalent
4318 * to binding with a wildcard address.
4320 * This is not currently spelled out in the SCTP sockets
4321 * extensions draft, but follows the practice as seen in TCP
4324 if (!ep
->base
.bind_addr
.port
) {
4325 if (sctp_autobind(sk
))
4328 sk
->sk_state
= SCTP_SS_LISTENING
;
4329 sctp_hash_endpoint(ep
);
4334 * 4.1.3 listen() - TCP Style Syntax
4336 * Applications uses listen() to ready the SCTP endpoint for accepting
4337 * inbound associations.
4339 SCTP_STATIC
int sctp_stream_listen(struct sock
*sk
, int backlog
)
4341 struct sctp_sock
*sp
= sctp_sk(sk
);
4342 struct sctp_endpoint
*ep
= sp
->ep
;
4344 /* If backlog is zero, disable listening. */
4346 if (sctp_sstate(sk
, CLOSED
))
4349 sctp_unhash_endpoint(ep
);
4350 sk
->sk_state
= SCTP_SS_CLOSED
;
4353 if (sctp_sstate(sk
, LISTENING
))
4357 * If a bind() or sctp_bindx() is not called prior to a listen()
4358 * call that allows new associations to be accepted, the system
4359 * picks an ephemeral port and will choose an address set equivalent
4360 * to binding with a wildcard address.
4362 * This is not currently spelled out in the SCTP sockets
4363 * extensions draft, but follows the practice as seen in TCP
4366 if (!ep
->base
.bind_addr
.port
) {
4367 if (sctp_autobind(sk
))
4370 sk
->sk_state
= SCTP_SS_LISTENING
;
4371 sk
->sk_max_ack_backlog
= backlog
;
4372 sctp_hash_endpoint(ep
);
4377 * Move a socket to LISTENING state.
4379 int sctp_inet_listen(struct socket
*sock
, int backlog
)
4381 struct sock
*sk
= sock
->sk
;
4382 struct crypto_tfm
*tfm
=NULL
;
4385 if (unlikely(backlog
< 0))
4390 if (sock
->state
!= SS_UNCONNECTED
)
4393 /* Allocate HMAC for generating cookie. */
4394 if (sctp_hmac_alg
) {
4395 tfm
= sctp_crypto_alloc_tfm(sctp_hmac_alg
, 0);
4402 switch (sock
->type
) {
4403 case SOCK_SEQPACKET
:
4404 err
= sctp_seqpacket_listen(sk
, backlog
);
4407 err
= sctp_stream_listen(sk
, backlog
);
4415 /* Store away the transform reference. */
4416 sctp_sk(sk
)->hmac
= tfm
;
4418 sctp_release_sock(sk
);
4421 sctp_crypto_free_tfm(tfm
);
4426 * This function is done by modeling the current datagram_poll() and the
4427 * tcp_poll(). Note that, based on these implementations, we don't
4428 * lock the socket in this function, even though it seems that,
4429 * ideally, locking or some other mechanisms can be used to ensure
4430 * the integrity of the counters (sndbuf and wmem_queued) used
4431 * in this place. We assume that we don't need locks either until proven
4434 * Another thing to note is that we include the Async I/O support
4435 * here, again, by modeling the current TCP/UDP code. We don't have
4436 * a good way to test with it yet.
4438 unsigned int sctp_poll(struct file
*file
, struct socket
*sock
, poll_table
*wait
)
4440 struct sock
*sk
= sock
->sk
;
4441 struct sctp_sock
*sp
= sctp_sk(sk
);
4444 poll_wait(file
, sk
->sk_sleep
, wait
);
4446 /* A TCP-style listening socket becomes readable when the accept queue
4449 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
4450 return (!list_empty(&sp
->ep
->asocs
)) ?
4451 (POLLIN
| POLLRDNORM
) : 0;
4455 /* Is there any exceptional events? */
4456 if (sk
->sk_err
|| !skb_queue_empty(&sk
->sk_error_queue
))
4458 if (sk
->sk_shutdown
== SHUTDOWN_MASK
)
4461 /* Is it readable? Reconsider this code with TCP-style support. */
4462 if (!skb_queue_empty(&sk
->sk_receive_queue
) ||
4463 (sk
->sk_shutdown
& RCV_SHUTDOWN
))
4464 mask
|= POLLIN
| POLLRDNORM
;
4466 /* The association is either gone or not ready. */
4467 if (!sctp_style(sk
, UDP
) && sctp_sstate(sk
, CLOSED
))
4470 /* Is it writable? */
4471 if (sctp_writeable(sk
)) {
4472 mask
|= POLLOUT
| POLLWRNORM
;
4474 set_bit(SOCK_ASYNC_NOSPACE
, &sk
->sk_socket
->flags
);
4476 * Since the socket is not locked, the buffer
4477 * might be made available after the writeable check and
4478 * before the bit is set. This could cause a lost I/O
4479 * signal. tcp_poll() has a race breaker for this race
4480 * condition. Based on their implementation, we put
4481 * in the following code to cover it as well.
4483 if (sctp_writeable(sk
))
4484 mask
|= POLLOUT
| POLLWRNORM
;
4489 /********************************************************************
4490 * 2nd Level Abstractions
4491 ********************************************************************/
4493 static struct sctp_bind_bucket
*sctp_bucket_create(
4494 struct sctp_bind_hashbucket
*head
, unsigned short snum
)
4496 struct sctp_bind_bucket
*pp
;
4498 pp
= kmem_cache_alloc(sctp_bucket_cachep
, SLAB_ATOMIC
);
4499 SCTP_DBG_OBJCNT_INC(bind_bucket
);
4503 INIT_HLIST_HEAD(&pp
->owner
);
4504 if ((pp
->next
= head
->chain
) != NULL
)
4505 pp
->next
->pprev
= &pp
->next
;
4507 pp
->pprev
= &head
->chain
;
4512 /* Caller must hold hashbucket lock for this tb with local BH disabled */
4513 static void sctp_bucket_destroy(struct sctp_bind_bucket
*pp
)
4515 if (hlist_empty(&pp
->owner
)) {
4517 pp
->next
->pprev
= pp
->pprev
;
4518 *(pp
->pprev
) = pp
->next
;
4519 kmem_cache_free(sctp_bucket_cachep
, pp
);
4520 SCTP_DBG_OBJCNT_DEC(bind_bucket
);
4524 /* Release this socket's reference to a local port. */
4525 static inline void __sctp_put_port(struct sock
*sk
)
4527 struct sctp_bind_hashbucket
*head
=
4528 &sctp_port_hashtable
[sctp_phashfn(inet_sk(sk
)->num
)];
4529 struct sctp_bind_bucket
*pp
;
4531 sctp_spin_lock(&head
->lock
);
4532 pp
= sctp_sk(sk
)->bind_hash
;
4533 __sk_del_bind_node(sk
);
4534 sctp_sk(sk
)->bind_hash
= NULL
;
4535 inet_sk(sk
)->num
= 0;
4536 sctp_bucket_destroy(pp
);
4537 sctp_spin_unlock(&head
->lock
);
4540 void sctp_put_port(struct sock
*sk
)
4542 sctp_local_bh_disable();
4543 __sctp_put_port(sk
);
4544 sctp_local_bh_enable();
4548 * The system picks an ephemeral port and choose an address set equivalent
4549 * to binding with a wildcard address.
4550 * One of those addresses will be the primary address for the association.
4551 * This automatically enables the multihoming capability of SCTP.
4553 static int sctp_autobind(struct sock
*sk
)
4555 union sctp_addr autoaddr
;
4557 unsigned short port
;
4559 /* Initialize a local sockaddr structure to INADDR_ANY. */
4560 af
= sctp_sk(sk
)->pf
->af
;
4562 port
= htons(inet_sk(sk
)->num
);
4563 af
->inaddr_any(&autoaddr
, port
);
4565 return sctp_do_bind(sk
, &autoaddr
, af
->sockaddr_len
);
4568 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
4571 * 4.2 The cmsghdr Structure *
4573 * When ancillary data is sent or received, any number of ancillary data
4574 * objects can be specified by the msg_control and msg_controllen members of
4575 * the msghdr structure, because each object is preceded by
4576 * a cmsghdr structure defining the object's length (the cmsg_len member).
4577 * Historically Berkeley-derived implementations have passed only one object
4578 * at a time, but this API allows multiple objects to be
4579 * passed in a single call to sendmsg() or recvmsg(). The following example
4580 * shows two ancillary data objects in a control buffer.
4582 * |<--------------------------- msg_controllen -------------------------->|
4585 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
4587 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
4590 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
4592 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
4595 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
4596 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
4598 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
4600 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
4607 SCTP_STATIC
int sctp_msghdr_parse(const struct msghdr
*msg
,
4608 sctp_cmsgs_t
*cmsgs
)
4610 struct cmsghdr
*cmsg
;
4612 for (cmsg
= CMSG_FIRSTHDR(msg
);
4614 cmsg
= CMSG_NXTHDR((struct msghdr
*)msg
, cmsg
)) {
4615 if (!CMSG_OK(msg
, cmsg
))
4618 /* Should we parse this header or ignore? */
4619 if (cmsg
->cmsg_level
!= IPPROTO_SCTP
)
4622 /* Strictly check lengths following example in SCM code. */
4623 switch (cmsg
->cmsg_type
) {
4625 /* SCTP Socket API Extension
4626 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
4628 * This cmsghdr structure provides information for
4629 * initializing new SCTP associations with sendmsg().
4630 * The SCTP_INITMSG socket option uses this same data
4631 * structure. This structure is not used for
4634 * cmsg_level cmsg_type cmsg_data[]
4635 * ------------ ------------ ----------------------
4636 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
4638 if (cmsg
->cmsg_len
!=
4639 CMSG_LEN(sizeof(struct sctp_initmsg
)))
4641 cmsgs
->init
= (struct sctp_initmsg
*)CMSG_DATA(cmsg
);
4645 /* SCTP Socket API Extension
4646 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
4648 * This cmsghdr structure specifies SCTP options for
4649 * sendmsg() and describes SCTP header information
4650 * about a received message through recvmsg().
4652 * cmsg_level cmsg_type cmsg_data[]
4653 * ------------ ------------ ----------------------
4654 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
4656 if (cmsg
->cmsg_len
!=
4657 CMSG_LEN(sizeof(struct sctp_sndrcvinfo
)))
4661 (struct sctp_sndrcvinfo
*)CMSG_DATA(cmsg
);
4663 /* Minimally, validate the sinfo_flags. */
4664 if (cmsgs
->info
->sinfo_flags
&
4665 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
4666 SCTP_ABORT
| SCTP_EOF
))
4678 * Wait for a packet..
4679 * Note: This function is the same function as in core/datagram.c
4680 * with a few modifications to make lksctp work.
4682 static int sctp_wait_for_packet(struct sock
* sk
, int *err
, long *timeo_p
)
4687 prepare_to_wait_exclusive(sk
->sk_sleep
, &wait
, TASK_INTERRUPTIBLE
);
4689 /* Socket errors? */
4690 error
= sock_error(sk
);
4694 if (!skb_queue_empty(&sk
->sk_receive_queue
))
4697 /* Socket shut down? */
4698 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
4701 /* Sequenced packets can come disconnected. If so we report the
4706 /* Is there a good reason to think that we may receive some data? */
4707 if (list_empty(&sctp_sk(sk
)->ep
->asocs
) && !sctp_sstate(sk
, LISTENING
))
4710 /* Handle signals. */
4711 if (signal_pending(current
))
4714 /* Let another process have a go. Since we are going to sleep
4715 * anyway. Note: This may cause odd behaviors if the message
4716 * does not fit in the user's buffer, but this seems to be the
4717 * only way to honor MSG_DONTWAIT realistically.
4719 sctp_release_sock(sk
);
4720 *timeo_p
= schedule_timeout(*timeo_p
);
4724 finish_wait(sk
->sk_sleep
, &wait
);
4728 error
= sock_intr_errno(*timeo_p
);
4731 finish_wait(sk
->sk_sleep
, &wait
);
4736 /* Receive a datagram.
4737 * Note: This is pretty much the same routine as in core/datagram.c
4738 * with a few changes to make lksctp work.
4740 static struct sk_buff
*sctp_skb_recv_datagram(struct sock
*sk
, int flags
,
4741 int noblock
, int *err
)
4744 struct sk_buff
*skb
;
4747 /* Caller is allowed not to check sk->sk_err before calling. */
4748 error
= sock_error(sk
);
4752 timeo
= sock_rcvtimeo(sk
, noblock
);
4754 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
4755 timeo
, MAX_SCHEDULE_TIMEOUT
);
4758 /* Again only user level code calls this function,
4759 * so nothing interrupt level
4760 * will suddenly eat the receive_queue.
4762 * Look at current nfs client by the way...
4763 * However, this function was corrent in any case. 8)
4765 if (flags
& MSG_PEEK
) {
4766 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
4767 skb
= skb_peek(&sk
->sk_receive_queue
);
4769 atomic_inc(&skb
->users
);
4770 spin_unlock_bh(&sk
->sk_receive_queue
.lock
);
4772 skb
= skb_dequeue(&sk
->sk_receive_queue
);
4778 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
4781 /* User doesn't want to wait. */
4785 } while (sctp_wait_for_packet(sk
, err
, &timeo
) == 0);
4794 /* If sndbuf has changed, wake up per association sndbuf waiters. */
4795 static void __sctp_write_space(struct sctp_association
*asoc
)
4797 struct sock
*sk
= asoc
->base
.sk
;
4798 struct socket
*sock
= sk
->sk_socket
;
4800 if ((sctp_wspace(asoc
) > 0) && sock
) {
4801 if (waitqueue_active(&asoc
->wait
))
4802 wake_up_interruptible(&asoc
->wait
);
4804 if (sctp_writeable(sk
)) {
4805 if (sk
->sk_sleep
&& waitqueue_active(sk
->sk_sleep
))
4806 wake_up_interruptible(sk
->sk_sleep
);
4808 /* Note that we try to include the Async I/O support
4809 * here by modeling from the current TCP/UDP code.
4810 * We have not tested with it yet.
4812 if (sock
->fasync_list
&&
4813 !(sk
->sk_shutdown
& SEND_SHUTDOWN
))
4814 sock_wake_async(sock
, 2, POLL_OUT
);
4819 /* Do accounting for the sndbuf space.
4820 * Decrement the used sndbuf space of the corresponding association by the
4821 * data size which was just transmitted(freed).
4823 static void sctp_wfree(struct sk_buff
*skb
)
4825 struct sctp_association
*asoc
;
4826 struct sctp_chunk
*chunk
;
4829 /* Get the saved chunk pointer. */
4830 chunk
= *((struct sctp_chunk
**)(skb
->cb
));
4833 asoc
->sndbuf_used
-= SCTP_DATA_SNDSIZE(chunk
) +
4834 sizeof(struct sk_buff
) +
4835 sizeof(struct sctp_chunk
);
4837 sk
->sk_wmem_queued
-= SCTP_DATA_SNDSIZE(chunk
) +
4838 sizeof(struct sk_buff
) +
4839 sizeof(struct sctp_chunk
);
4841 atomic_sub(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
4844 __sctp_write_space(asoc
);
4846 sctp_association_put(asoc
);
4849 /* Helper function to wait for space in the sndbuf. */
4850 static int sctp_wait_for_sndbuf(struct sctp_association
*asoc
, long *timeo_p
,
4853 struct sock
*sk
= asoc
->base
.sk
;
4855 long current_timeo
= *timeo_p
;
4858 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
4859 asoc
, (long)(*timeo_p
), msg_len
);
4861 /* Increment the association's refcnt. */
4862 sctp_association_hold(asoc
);
4864 /* Wait on the association specific sndbuf space. */
4866 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
4867 TASK_INTERRUPTIBLE
);
4870 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
4873 if (signal_pending(current
))
4874 goto do_interrupted
;
4875 if (msg_len
<= sctp_wspace(asoc
))
4878 /* Let another process have a go. Since we are going
4881 sctp_release_sock(sk
);
4882 current_timeo
= schedule_timeout(current_timeo
);
4885 *timeo_p
= current_timeo
;
4889 finish_wait(&asoc
->wait
, &wait
);
4891 /* Release the association's refcnt. */
4892 sctp_association_put(asoc
);
4901 err
= sock_intr_errno(*timeo_p
);
4909 /* If socket sndbuf has changed, wake up all per association waiters. */
4910 void sctp_write_space(struct sock
*sk
)
4912 struct sctp_association
*asoc
;
4913 struct list_head
*pos
;
4915 /* Wake up the tasks in each wait queue. */
4916 list_for_each(pos
, &((sctp_sk(sk
))->ep
->asocs
)) {
4917 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
4918 __sctp_write_space(asoc
);
4922 /* Is there any sndbuf space available on the socket?
4924 * Note that wmem_queued is the sum of the send buffers on all of the
4925 * associations on the same socket. For a UDP-style socket with
4926 * multiple associations, it is possible for it to be "unwriteable"
4927 * prematurely. I assume that this is acceptable because
4928 * a premature "unwriteable" is better than an accidental "writeable" which
4929 * would cause an unwanted block under certain circumstances. For the 1-1
4930 * UDP-style sockets or TCP-style sockets, this code should work.
4933 static int sctp_writeable(struct sock
*sk
)
4937 amt
= sk
->sk_sndbuf
- sk
->sk_wmem_queued
;
4943 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
4944 * returns immediately with EINPROGRESS.
4946 static int sctp_wait_for_connect(struct sctp_association
*asoc
, long *timeo_p
)
4948 struct sock
*sk
= asoc
->base
.sk
;
4950 long current_timeo
= *timeo_p
;
4953 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __FUNCTION__
, asoc
,
4956 /* Increment the association's refcnt. */
4957 sctp_association_hold(asoc
);
4960 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
4961 TASK_INTERRUPTIBLE
);
4964 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
4966 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
4969 if (signal_pending(current
))
4970 goto do_interrupted
;
4972 if (sctp_state(asoc
, ESTABLISHED
))
4975 /* Let another process have a go. Since we are going
4978 sctp_release_sock(sk
);
4979 current_timeo
= schedule_timeout(current_timeo
);
4982 *timeo_p
= current_timeo
;
4986 finish_wait(&asoc
->wait
, &wait
);
4988 /* Release the association's refcnt. */
4989 sctp_association_put(asoc
);
4994 if (asoc
->init_err_counter
+ 1 >= asoc
->max_init_attempts
)
4997 err
= -ECONNREFUSED
;
5001 err
= sock_intr_errno(*timeo_p
);
5009 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
)
5011 struct sctp_endpoint
*ep
;
5015 ep
= sctp_sk(sk
)->ep
;
5019 prepare_to_wait_exclusive(sk
->sk_sleep
, &wait
,
5020 TASK_INTERRUPTIBLE
);
5022 if (list_empty(&ep
->asocs
)) {
5023 sctp_release_sock(sk
);
5024 timeo
= schedule_timeout(timeo
);
5029 if (!sctp_sstate(sk
, LISTENING
))
5033 if (!list_empty(&ep
->asocs
))
5036 err
= sock_intr_errno(timeo
);
5037 if (signal_pending(current
))
5045 finish_wait(sk
->sk_sleep
, &wait
);
5050 void sctp_wait_for_close(struct sock
*sk
, long timeout
)
5055 prepare_to_wait(sk
->sk_sleep
, &wait
, TASK_INTERRUPTIBLE
);
5056 if (list_empty(&sctp_sk(sk
)->ep
->asocs
))
5058 sctp_release_sock(sk
);
5059 timeout
= schedule_timeout(timeout
);
5061 } while (!signal_pending(current
) && timeout
);
5063 finish_wait(sk
->sk_sleep
, &wait
);
5066 /* Populate the fields of the newsk from the oldsk and migrate the assoc
5067 * and its messages to the newsk.
5069 static void sctp_sock_migrate(struct sock
*oldsk
, struct sock
*newsk
,
5070 struct sctp_association
*assoc
,
5071 sctp_socket_type_t type
)
5073 struct sctp_sock
*oldsp
= sctp_sk(oldsk
);
5074 struct sctp_sock
*newsp
= sctp_sk(newsk
);
5075 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
5076 struct sctp_endpoint
*newep
= newsp
->ep
;
5077 struct sk_buff
*skb
, *tmp
;
5078 struct sctp_ulpevent
*event
;
5081 /* Migrate socket buffer sizes and all the socket level options to the
5084 newsk
->sk_sndbuf
= oldsk
->sk_sndbuf
;
5085 newsk
->sk_rcvbuf
= oldsk
->sk_rcvbuf
;
5086 /* Brute force copy old sctp opt. */
5087 inet_sk_copy_descendant(newsk
, oldsk
);
5089 /* Restore the ep value that was overwritten with the above structure
5095 /* Hook this new socket in to the bind_hash list. */
5096 pp
= sctp_sk(oldsk
)->bind_hash
;
5097 sk_add_bind_node(newsk
, &pp
->owner
);
5098 sctp_sk(newsk
)->bind_hash
= pp
;
5099 inet_sk(newsk
)->num
= inet_sk(oldsk
)->num
;
5101 /* Copy the bind_addr list from the original endpoint to the new
5102 * endpoint so that we can handle restarts properly
5104 if (assoc
->peer
.ipv4_address
)
5105 flags
|= SCTP_ADDR4_PEERSUPP
;
5106 if (assoc
->peer
.ipv6_address
)
5107 flags
|= SCTP_ADDR6_PEERSUPP
;
5108 sctp_bind_addr_copy(&newsp
->ep
->base
.bind_addr
,
5109 &oldsp
->ep
->base
.bind_addr
,
5110 SCTP_SCOPE_GLOBAL
, GFP_KERNEL
, flags
);
5112 /* Move any messages in the old socket's receive queue that are for the
5113 * peeled off association to the new socket's receive queue.
5115 sctp_skb_for_each(skb
, &oldsk
->sk_receive_queue
, tmp
) {
5116 event
= sctp_skb2event(skb
);
5117 if (event
->asoc
== assoc
) {
5118 __skb_unlink(skb
, &oldsk
->sk_receive_queue
);
5119 __skb_queue_tail(&newsk
->sk_receive_queue
, skb
);
5123 /* Clean up any messages pending delivery due to partial
5124 * delivery. Three cases:
5125 * 1) No partial deliver; no work.
5126 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
5127 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
5129 skb_queue_head_init(&newsp
->pd_lobby
);
5130 sctp_sk(newsk
)->pd_mode
= assoc
->ulpq
.pd_mode
;
5132 if (sctp_sk(oldsk
)->pd_mode
) {
5133 struct sk_buff_head
*queue
;
5135 /* Decide which queue to move pd_lobby skbs to. */
5136 if (assoc
->ulpq
.pd_mode
) {
5137 queue
= &newsp
->pd_lobby
;
5139 queue
= &newsk
->sk_receive_queue
;
5141 /* Walk through the pd_lobby, looking for skbs that
5142 * need moved to the new socket.
5144 sctp_skb_for_each(skb
, &oldsp
->pd_lobby
, tmp
) {
5145 event
= sctp_skb2event(skb
);
5146 if (event
->asoc
== assoc
) {
5147 __skb_unlink(skb
, &oldsp
->pd_lobby
);
5148 __skb_queue_tail(queue
, skb
);
5152 /* Clear up any skbs waiting for the partial
5153 * delivery to finish.
5155 if (assoc
->ulpq
.pd_mode
)
5156 sctp_clear_pd(oldsk
);
5160 /* Set the type of socket to indicate that it is peeled off from the
5161 * original UDP-style socket or created with the accept() call on a
5162 * TCP-style socket..
5166 /* Migrate the association to the new socket. */
5167 sctp_assoc_migrate(assoc
, newsk
);
5169 /* If the association on the newsk is already closed before accept()
5170 * is called, set RCV_SHUTDOWN flag.
5172 if (sctp_state(assoc
, CLOSED
) && sctp_style(newsk
, TCP
))
5173 newsk
->sk_shutdown
|= RCV_SHUTDOWN
;
5175 newsk
->sk_state
= SCTP_SS_ESTABLISHED
;
5178 /* This proto struct describes the ULP interface for SCTP. */
5179 struct proto sctp_prot
= {
5181 .owner
= THIS_MODULE
,
5182 .close
= sctp_close
,
5183 .connect
= sctp_connect
,
5184 .disconnect
= sctp_disconnect
,
5185 .accept
= sctp_accept
,
5186 .ioctl
= sctp_ioctl
,
5187 .init
= sctp_init_sock
,
5188 .destroy
= sctp_destroy_sock
,
5189 .shutdown
= sctp_shutdown
,
5190 .setsockopt
= sctp_setsockopt
,
5191 .getsockopt
= sctp_getsockopt
,
5192 .sendmsg
= sctp_sendmsg
,
5193 .recvmsg
= sctp_recvmsg
,
5195 .backlog_rcv
= sctp_backlog_rcv
,
5197 .unhash
= sctp_unhash
,
5198 .get_port
= sctp_get_port
,
5199 .obj_size
= sizeof(struct sctp_sock
),
5202 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
5203 struct proto sctpv6_prot
= {
5205 .owner
= THIS_MODULE
,
5206 .close
= sctp_close
,
5207 .connect
= sctp_connect
,
5208 .disconnect
= sctp_disconnect
,
5209 .accept
= sctp_accept
,
5210 .ioctl
= sctp_ioctl
,
5211 .init
= sctp_init_sock
,
5212 .destroy
= sctp_destroy_sock
,
5213 .shutdown
= sctp_shutdown
,
5214 .setsockopt
= sctp_setsockopt
,
5215 .getsockopt
= sctp_getsockopt
,
5216 .sendmsg
= sctp_sendmsg
,
5217 .recvmsg
= sctp_recvmsg
,
5219 .backlog_rcv
= sctp_backlog_rcv
,
5221 .unhash
= sctp_unhash
,
5222 .get_port
= sctp_get_port
,
5223 .obj_size
= sizeof(struct sctp6_sock
),
5225 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */