1 /* SCTP kernel implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
9 * This file is part of the SCTP kernel implementation
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
18 * This SCTP implementation is free software;
19 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
24 * This SCTP implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, write to
32 * the Free Software Foundation, 59 Temple Place - Suite 330,
33 * Boston, MA 02111-1307, USA.
35 * Please send any bug reports or fixes you make to the
37 * lksctp developers <lksctp-developers@lists.sourceforge.net>
39 * Or submit a bug report through the following website:
40 * http://www.sf.net/projects/lksctp
42 * Written or modified by:
43 * La Monte H.P. Yarroll <piggy@acm.org>
44 * Narasimha Budihal <narsi@refcode.org>
45 * Karl Knutson <karl@athena.chicago.il.us>
46 * Jon Grimm <jgrimm@us.ibm.com>
47 * Xingang Guo <xingang.guo@intel.com>
48 * Daisy Chang <daisyc@us.ibm.com>
49 * Sridhar Samudrala <samudrala@us.ibm.com>
50 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
51 * Ardelle Fan <ardelle.fan@intel.com>
52 * Ryan Layer <rmlayer@us.ibm.com>
53 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
54 * Kevin Gao <kevin.gao@intel.com>
56 * Any bugs reported given to us we will try to fix... any fixes shared will
57 * be incorporated into the next SCTP release.
60 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
62 #include <linux/types.h>
63 #include <linux/kernel.h>
64 #include <linux/wait.h>
65 #include <linux/time.h>
67 #include <linux/capability.h>
68 #include <linux/fcntl.h>
69 #include <linux/poll.h>
70 #include <linux/init.h>
71 #include <linux/crypto.h>
72 #include <linux/slab.h>
76 #include <net/route.h>
78 #include <net/inet_common.h>
80 #include <linux/socket.h> /* for sa_family_t */
82 #include <net/sctp/sctp.h>
83 #include <net/sctp/sm.h>
85 /* WARNING: Please do not remove the SCTP_STATIC attribute to
86 * any of the functions below as they are used to export functions
87 * used by a project regression testsuite.
90 /* Forward declarations for internal helper functions. */
91 static int sctp_writeable(struct sock
*sk
);
92 static void sctp_wfree(struct sk_buff
*skb
);
93 static int sctp_wait_for_sndbuf(struct sctp_association
*, long *timeo_p
,
95 static int sctp_wait_for_packet(struct sock
* sk
, int *err
, long *timeo_p
);
96 static int sctp_wait_for_connect(struct sctp_association
*, long *timeo_p
);
97 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
);
98 static void sctp_wait_for_close(struct sock
*sk
, long timeo
);
99 static struct sctp_af
*sctp_sockaddr_af(struct sctp_sock
*opt
,
100 union sctp_addr
*addr
, int len
);
101 static int sctp_bindx_add(struct sock
*, struct sockaddr
*, int);
102 static int sctp_bindx_rem(struct sock
*, struct sockaddr
*, int);
103 static int sctp_send_asconf_add_ip(struct sock
*, struct sockaddr
*, int);
104 static int sctp_send_asconf_del_ip(struct sock
*, struct sockaddr
*, int);
105 static int sctp_send_asconf(struct sctp_association
*asoc
,
106 struct sctp_chunk
*chunk
);
107 static int sctp_do_bind(struct sock
*, union sctp_addr
*, int);
108 static int sctp_autobind(struct sock
*sk
);
109 static void sctp_sock_migrate(struct sock
*, struct sock
*,
110 struct sctp_association
*, sctp_socket_type_t
);
111 static char *sctp_hmac_alg
= SCTP_COOKIE_HMAC_ALG
;
113 extern struct kmem_cache
*sctp_bucket_cachep
;
114 extern long sysctl_sctp_mem
[3];
115 extern int sysctl_sctp_rmem
[3];
116 extern int sysctl_sctp_wmem
[3];
118 static int sctp_memory_pressure
;
119 static atomic_long_t sctp_memory_allocated
;
120 struct percpu_counter sctp_sockets_allocated
;
122 static void sctp_enter_memory_pressure(struct sock
*sk
)
124 sctp_memory_pressure
= 1;
128 /* Get the sndbuf space available at the time on the association. */
129 static inline int sctp_wspace(struct sctp_association
*asoc
)
133 if (asoc
->ep
->sndbuf_policy
)
134 amt
= asoc
->sndbuf_used
;
136 amt
= sk_wmem_alloc_get(asoc
->base
.sk
);
138 if (amt
>= asoc
->base
.sk
->sk_sndbuf
) {
139 if (asoc
->base
.sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)
142 amt
= sk_stream_wspace(asoc
->base
.sk
);
147 amt
= asoc
->base
.sk
->sk_sndbuf
- amt
;
152 /* Increment the used sndbuf space count of the corresponding association by
153 * the size of the outgoing data chunk.
154 * Also, set the skb destructor for sndbuf accounting later.
156 * Since it is always 1-1 between chunk and skb, and also a new skb is always
157 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
158 * destructor in the data chunk skb for the purpose of the sndbuf space
161 static inline void sctp_set_owner_w(struct sctp_chunk
*chunk
)
163 struct sctp_association
*asoc
= chunk
->asoc
;
164 struct sock
*sk
= asoc
->base
.sk
;
166 /* The sndbuf space is tracked per association. */
167 sctp_association_hold(asoc
);
169 skb_set_owner_w(chunk
->skb
, sk
);
171 chunk
->skb
->destructor
= sctp_wfree
;
172 /* Save the chunk pointer in skb for sctp_wfree to use later. */
173 *((struct sctp_chunk
**)(chunk
->skb
->cb
)) = chunk
;
175 asoc
->sndbuf_used
+= SCTP_DATA_SNDSIZE(chunk
) +
176 sizeof(struct sk_buff
) +
177 sizeof(struct sctp_chunk
);
179 atomic_add(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
180 sk
->sk_wmem_queued
+= chunk
->skb
->truesize
;
181 sk_mem_charge(sk
, chunk
->skb
->truesize
);
184 /* Verify that this is a valid address. */
185 static inline int sctp_verify_addr(struct sock
*sk
, union sctp_addr
*addr
,
190 /* Verify basic sockaddr. */
191 af
= sctp_sockaddr_af(sctp_sk(sk
), addr
, len
);
195 /* Is this a valid SCTP address? */
196 if (!af
->addr_valid(addr
, sctp_sk(sk
), NULL
))
199 if (!sctp_sk(sk
)->pf
->send_verify(sctp_sk(sk
), (addr
)))
205 /* Look up the association by its id. If this is not a UDP-style
206 * socket, the ID field is always ignored.
208 struct sctp_association
*sctp_id2assoc(struct sock
*sk
, sctp_assoc_t id
)
210 struct sctp_association
*asoc
= NULL
;
212 /* If this is not a UDP-style socket, assoc id should be ignored. */
213 if (!sctp_style(sk
, UDP
)) {
214 /* Return NULL if the socket state is not ESTABLISHED. It
215 * could be a TCP-style listening socket or a socket which
216 * hasn't yet called connect() to establish an association.
218 if (!sctp_sstate(sk
, ESTABLISHED
))
221 /* Get the first and the only association from the list. */
222 if (!list_empty(&sctp_sk(sk
)->ep
->asocs
))
223 asoc
= list_entry(sctp_sk(sk
)->ep
->asocs
.next
,
224 struct sctp_association
, asocs
);
228 /* Otherwise this is a UDP-style socket. */
229 if (!id
|| (id
== (sctp_assoc_t
)-1))
232 spin_lock_bh(&sctp_assocs_id_lock
);
233 asoc
= (struct sctp_association
*)idr_find(&sctp_assocs_id
, (int)id
);
234 spin_unlock_bh(&sctp_assocs_id_lock
);
236 if (!asoc
|| (asoc
->base
.sk
!= sk
) || asoc
->base
.dead
)
242 /* Look up the transport from an address and an assoc id. If both address and
243 * id are specified, the associations matching the address and the id should be
246 static struct sctp_transport
*sctp_addr_id2transport(struct sock
*sk
,
247 struct sockaddr_storage
*addr
,
250 struct sctp_association
*addr_asoc
= NULL
, *id_asoc
= NULL
;
251 struct sctp_transport
*transport
;
252 union sctp_addr
*laddr
= (union sctp_addr
*)addr
;
254 addr_asoc
= sctp_endpoint_lookup_assoc(sctp_sk(sk
)->ep
,
261 id_asoc
= sctp_id2assoc(sk
, id
);
262 if (id_asoc
&& (id_asoc
!= addr_asoc
))
265 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
266 (union sctp_addr
*)addr
);
271 /* API 3.1.2 bind() - UDP Style Syntax
272 * The syntax of bind() is,
274 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
276 * sd - the socket descriptor returned by socket().
277 * addr - the address structure (struct sockaddr_in or struct
278 * sockaddr_in6 [RFC 2553]),
279 * addr_len - the size of the address structure.
281 SCTP_STATIC
int sctp_bind(struct sock
*sk
, struct sockaddr
*addr
, int addr_len
)
287 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
290 /* Disallow binding twice. */
291 if (!sctp_sk(sk
)->ep
->base
.bind_addr
.port
)
292 retval
= sctp_do_bind(sk
, (union sctp_addr
*)addr
,
297 sctp_release_sock(sk
);
302 static long sctp_get_port_local(struct sock
*, union sctp_addr
*);
304 /* Verify this is a valid sockaddr. */
305 static struct sctp_af
*sctp_sockaddr_af(struct sctp_sock
*opt
,
306 union sctp_addr
*addr
, int len
)
310 /* Check minimum size. */
311 if (len
< sizeof (struct sockaddr
))
314 /* V4 mapped address are really of AF_INET family */
315 if (addr
->sa
.sa_family
== AF_INET6
&&
316 ipv6_addr_v4mapped(&addr
->v6
.sin6_addr
)) {
317 if (!opt
->pf
->af_supported(AF_INET
, opt
))
320 /* Does this PF support this AF? */
321 if (!opt
->pf
->af_supported(addr
->sa
.sa_family
, opt
))
325 /* If we get this far, af is valid. */
326 af
= sctp_get_af_specific(addr
->sa
.sa_family
);
328 if (len
< af
->sockaddr_len
)
334 /* Bind a local address either to an endpoint or to an association. */
335 SCTP_STATIC
int sctp_do_bind(struct sock
*sk
, union sctp_addr
*addr
, int len
)
337 struct sctp_sock
*sp
= sctp_sk(sk
);
338 struct sctp_endpoint
*ep
= sp
->ep
;
339 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
344 /* Common sockaddr verification. */
345 af
= sctp_sockaddr_af(sp
, addr
, len
);
347 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
352 snum
= ntohs(addr
->v4
.sin_port
);
354 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
355 ", port: %d, new port: %d, len: %d)\n",
361 /* PF specific bind() address verification. */
362 if (!sp
->pf
->bind_verify(sp
, addr
))
363 return -EADDRNOTAVAIL
;
365 /* We must either be unbound, or bind to the same port.
366 * It's OK to allow 0 ports if we are already bound.
367 * We'll just inhert an already bound port in this case
372 else if (snum
!= bp
->port
) {
373 SCTP_DEBUG_PRINTK("sctp_do_bind:"
374 " New port %d does not match existing port "
375 "%d.\n", snum
, bp
->port
);
380 if (snum
&& snum
< PROT_SOCK
&& !capable(CAP_NET_BIND_SERVICE
))
383 /* See if the address matches any of the addresses we may have
384 * already bound before checking against other endpoints.
386 if (sctp_bind_addr_match(bp
, addr
, sp
))
389 /* Make sure we are allowed to bind here.
390 * The function sctp_get_port_local() does duplicate address
393 addr
->v4
.sin_port
= htons(snum
);
394 if ((ret
= sctp_get_port_local(sk
, addr
))) {
398 /* Refresh ephemeral port. */
400 bp
->port
= inet_sk(sk
)->inet_num
;
402 /* Add the address to the bind address list.
403 * Use GFP_ATOMIC since BHs will be disabled.
405 ret
= sctp_add_bind_addr(bp
, addr
, SCTP_ADDR_SRC
, GFP_ATOMIC
);
407 /* Copy back into socket for getsockname() use. */
409 inet_sk(sk
)->inet_sport
= htons(inet_sk(sk
)->inet_num
);
410 af
->to_sk_saddr(addr
, sk
);
416 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
418 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
419 * at any one time. If a sender, after sending an ASCONF chunk, decides
420 * it needs to transfer another ASCONF Chunk, it MUST wait until the
421 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
422 * subsequent ASCONF. Note this restriction binds each side, so at any
423 * time two ASCONF may be in-transit on any given association (one sent
424 * from each endpoint).
426 static int sctp_send_asconf(struct sctp_association
*asoc
,
427 struct sctp_chunk
*chunk
)
431 /* If there is an outstanding ASCONF chunk, queue it for later
434 if (asoc
->addip_last_asconf
) {
435 list_add_tail(&chunk
->list
, &asoc
->addip_chunk_list
);
439 /* Hold the chunk until an ASCONF_ACK is received. */
440 sctp_chunk_hold(chunk
);
441 retval
= sctp_primitive_ASCONF(asoc
, chunk
);
443 sctp_chunk_free(chunk
);
445 asoc
->addip_last_asconf
= chunk
;
451 /* Add a list of addresses as bind addresses to local endpoint or
454 * Basically run through each address specified in the addrs/addrcnt
455 * array/length pair, determine if it is IPv6 or IPv4 and call
456 * sctp_do_bind() on it.
458 * If any of them fails, then the operation will be reversed and the
459 * ones that were added will be removed.
461 * Only sctp_setsockopt_bindx() is supposed to call this function.
463 static int sctp_bindx_add(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
468 struct sockaddr
*sa_addr
;
471 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
475 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
476 /* The list may contain either IPv4 or IPv6 address;
477 * determine the address length for walking thru the list.
479 sa_addr
= (struct sockaddr
*)addr_buf
;
480 af
= sctp_get_af_specific(sa_addr
->sa_family
);
486 retval
= sctp_do_bind(sk
, (union sctp_addr
*)sa_addr
,
489 addr_buf
+= af
->sockaddr_len
;
493 /* Failed. Cleanup the ones that have been added */
495 sctp_bindx_rem(sk
, addrs
, cnt
);
503 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
504 * associations that are part of the endpoint indicating that a list of local
505 * addresses are added to the endpoint.
507 * If any of the addresses is already in the bind address list of the
508 * association, we do not send the chunk for that association. But it will not
509 * affect other associations.
511 * Only sctp_setsockopt_bindx() is supposed to call this function.
513 static int sctp_send_asconf_add_ip(struct sock
*sk
,
514 struct sockaddr
*addrs
,
517 struct sctp_sock
*sp
;
518 struct sctp_endpoint
*ep
;
519 struct sctp_association
*asoc
;
520 struct sctp_bind_addr
*bp
;
521 struct sctp_chunk
*chunk
;
522 struct sctp_sockaddr_entry
*laddr
;
523 union sctp_addr
*addr
;
524 union sctp_addr saveaddr
;
531 if (!sctp_addip_enable
)
537 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
538 __func__
, sk
, addrs
, addrcnt
);
540 list_for_each_entry(asoc
, &ep
->asocs
, asocs
) {
542 if (!asoc
->peer
.asconf_capable
)
545 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_ADD_IP
)
548 if (!sctp_state(asoc
, ESTABLISHED
))
551 /* Check if any address in the packed array of addresses is
552 * in the bind address list of the association. If so,
553 * do not send the asconf chunk to its peer, but continue with
554 * other associations.
557 for (i
= 0; i
< addrcnt
; i
++) {
558 addr
= (union sctp_addr
*)addr_buf
;
559 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
565 if (sctp_assoc_lookup_laddr(asoc
, addr
))
568 addr_buf
+= af
->sockaddr_len
;
573 /* Use the first valid address in bind addr list of
574 * association as Address Parameter of ASCONF CHUNK.
576 bp
= &asoc
->base
.bind_addr
;
577 p
= bp
->address_list
.next
;
578 laddr
= list_entry(p
, struct sctp_sockaddr_entry
, list
);
579 chunk
= sctp_make_asconf_update_ip(asoc
, &laddr
->a
, addrs
,
580 addrcnt
, SCTP_PARAM_ADD_IP
);
586 retval
= sctp_send_asconf(asoc
, chunk
);
590 /* Add the new addresses to the bind address list with
591 * use_as_src set to 0.
594 for (i
= 0; i
< addrcnt
; i
++) {
595 addr
= (union sctp_addr
*)addr_buf
;
596 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
597 memcpy(&saveaddr
, addr
, af
->sockaddr_len
);
598 retval
= sctp_add_bind_addr(bp
, &saveaddr
,
599 SCTP_ADDR_NEW
, GFP_ATOMIC
);
600 addr_buf
+= af
->sockaddr_len
;
608 /* Remove a list of addresses from bind addresses list. Do not remove the
611 * Basically run through each address specified in the addrs/addrcnt
612 * array/length pair, determine if it is IPv6 or IPv4 and call
613 * sctp_del_bind() on it.
615 * If any of them fails, then the operation will be reversed and the
616 * ones that were removed will be added back.
618 * At least one address has to be left; if only one address is
619 * available, the operation will return -EBUSY.
621 * Only sctp_setsockopt_bindx() is supposed to call this function.
623 static int sctp_bindx_rem(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
625 struct sctp_sock
*sp
= sctp_sk(sk
);
626 struct sctp_endpoint
*ep
= sp
->ep
;
628 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
631 union sctp_addr
*sa_addr
;
634 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
638 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
639 /* If the bind address list is empty or if there is only one
640 * bind address, there is nothing more to be removed (we need
641 * at least one address here).
643 if (list_empty(&bp
->address_list
) ||
644 (sctp_list_single_entry(&bp
->address_list
))) {
649 sa_addr
= (union sctp_addr
*)addr_buf
;
650 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
656 if (!af
->addr_valid(sa_addr
, sp
, NULL
)) {
657 retval
= -EADDRNOTAVAIL
;
661 if (sa_addr
->v4
.sin_port
&&
662 sa_addr
->v4
.sin_port
!= htons(bp
->port
)) {
667 if (!sa_addr
->v4
.sin_port
)
668 sa_addr
->v4
.sin_port
= htons(bp
->port
);
670 /* FIXME - There is probably a need to check if sk->sk_saddr and
671 * sk->sk_rcv_addr are currently set to one of the addresses to
672 * be removed. This is something which needs to be looked into
673 * when we are fixing the outstanding issues with multi-homing
674 * socket routing and failover schemes. Refer to comments in
675 * sctp_do_bind(). -daisy
677 retval
= sctp_del_bind_addr(bp
, sa_addr
);
679 addr_buf
+= af
->sockaddr_len
;
682 /* Failed. Add the ones that has been removed back */
684 sctp_bindx_add(sk
, addrs
, cnt
);
692 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
693 * the associations that are part of the endpoint indicating that a list of
694 * local addresses are removed from the endpoint.
696 * If any of the addresses is already in the bind address list of the
697 * association, we do not send the chunk for that association. But it will not
698 * affect other associations.
700 * Only sctp_setsockopt_bindx() is supposed to call this function.
702 static int sctp_send_asconf_del_ip(struct sock
*sk
,
703 struct sockaddr
*addrs
,
706 struct sctp_sock
*sp
;
707 struct sctp_endpoint
*ep
;
708 struct sctp_association
*asoc
;
709 struct sctp_transport
*transport
;
710 struct sctp_bind_addr
*bp
;
711 struct sctp_chunk
*chunk
;
712 union sctp_addr
*laddr
;
715 struct sctp_sockaddr_entry
*saddr
;
719 if (!sctp_addip_enable
)
725 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
726 __func__
, sk
, addrs
, addrcnt
);
728 list_for_each_entry(asoc
, &ep
->asocs
, asocs
) {
730 if (!asoc
->peer
.asconf_capable
)
733 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_DEL_IP
)
736 if (!sctp_state(asoc
, ESTABLISHED
))
739 /* Check if any address in the packed array of addresses is
740 * not present in the bind address list of the association.
741 * If so, do not send the asconf chunk to its peer, but
742 * continue with other associations.
745 for (i
= 0; i
< addrcnt
; i
++) {
746 laddr
= (union sctp_addr
*)addr_buf
;
747 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
753 if (!sctp_assoc_lookup_laddr(asoc
, laddr
))
756 addr_buf
+= af
->sockaddr_len
;
761 /* Find one address in the association's bind address list
762 * that is not in the packed array of addresses. This is to
763 * make sure that we do not delete all the addresses in the
766 bp
= &asoc
->base
.bind_addr
;
767 laddr
= sctp_find_unmatch_addr(bp
, (union sctp_addr
*)addrs
,
772 /* We do not need RCU protection throughout this loop
773 * because this is done under a socket lock from the
776 chunk
= sctp_make_asconf_update_ip(asoc
, laddr
, addrs
, addrcnt
,
783 /* Reset use_as_src flag for the addresses in the bind address
784 * list that are to be deleted.
787 for (i
= 0; i
< addrcnt
; i
++) {
788 laddr
= (union sctp_addr
*)addr_buf
;
789 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
790 list_for_each_entry(saddr
, &bp
->address_list
, list
) {
791 if (sctp_cmp_addr_exact(&saddr
->a
, laddr
))
792 saddr
->state
= SCTP_ADDR_DEL
;
794 addr_buf
+= af
->sockaddr_len
;
797 /* Update the route and saddr entries for all the transports
798 * as some of the addresses in the bind address list are
799 * about to be deleted and cannot be used as source addresses.
801 list_for_each_entry(transport
, &asoc
->peer
.transport_addr_list
,
803 dst_release(transport
->dst
);
804 sctp_transport_route(transport
, NULL
,
805 sctp_sk(asoc
->base
.sk
));
808 retval
= sctp_send_asconf(asoc
, chunk
);
814 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
815 int sctp_asconf_mgmt(struct sctp_sock
*sp
, struct sctp_sockaddr_entry
*addrw
)
817 struct sock
*sk
= sctp_opt2sk(sp
);
818 union sctp_addr
*addr
;
821 /* It is safe to write port space in caller. */
823 addr
->v4
.sin_port
= htons(sp
->ep
->base
.bind_addr
.port
);
824 af
= sctp_get_af_specific(addr
->sa
.sa_family
);
827 if (sctp_verify_addr(sk
, addr
, af
->sockaddr_len
))
830 if (addrw
->state
== SCTP_ADDR_NEW
)
831 return sctp_send_asconf_add_ip(sk
, (struct sockaddr
*)addr
, 1);
833 return sctp_send_asconf_del_ip(sk
, (struct sockaddr
*)addr
, 1);
836 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
839 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
842 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
843 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
846 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
847 * Section 3.1.2 for this usage.
849 * addrs is a pointer to an array of one or more socket addresses. Each
850 * address is contained in its appropriate structure (i.e. struct
851 * sockaddr_in or struct sockaddr_in6) the family of the address type
852 * must be used to distinguish the address length (note that this
853 * representation is termed a "packed array" of addresses). The caller
854 * specifies the number of addresses in the array with addrcnt.
856 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
857 * -1, and sets errno to the appropriate error code.
859 * For SCTP, the port given in each socket address must be the same, or
860 * sctp_bindx() will fail, setting errno to EINVAL.
862 * The flags parameter is formed from the bitwise OR of zero or more of
863 * the following currently defined flags:
865 * SCTP_BINDX_ADD_ADDR
867 * SCTP_BINDX_REM_ADDR
869 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
870 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
871 * addresses from the association. The two flags are mutually exclusive;
872 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
873 * not remove all addresses from an association; sctp_bindx() will
874 * reject such an attempt with EINVAL.
876 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
877 * additional addresses with an endpoint after calling bind(). Or use
878 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
879 * socket is associated with so that no new association accepted will be
880 * associated with those addresses. If the endpoint supports dynamic
881 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
882 * endpoint to send the appropriate message to the peer to change the
883 * peers address lists.
885 * Adding and removing addresses from a connected association is
886 * optional functionality. Implementations that do not support this
887 * functionality should return EOPNOTSUPP.
889 * Basically do nothing but copying the addresses from user to kernel
890 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
891 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
894 * We don't use copy_from_user() for optimization: we first do the
895 * sanity checks (buffer size -fast- and access check-healthy
896 * pointer); if all of those succeed, then we can alloc the memory
897 * (expensive operation) needed to copy the data to kernel. Then we do
898 * the copying without checking the user space area
899 * (__copy_from_user()).
901 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
904 * sk The sk of the socket
905 * addrs The pointer to the addresses in user land
906 * addrssize Size of the addrs buffer
907 * op Operation to perform (add or remove, see the flags of
910 * Returns 0 if ok, <0 errno code on error.
912 SCTP_STATIC
int sctp_setsockopt_bindx(struct sock
* sk
,
913 struct sockaddr __user
*addrs
,
914 int addrs_size
, int op
)
916 struct sockaddr
*kaddrs
;
920 struct sockaddr
*sa_addr
;
924 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
925 " addrs_size %d opt %d\n", sk
, addrs
, addrs_size
, op
);
927 if (unlikely(addrs_size
<= 0))
930 /* Check the user passed a healthy pointer. */
931 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
934 /* Alloc space for the address array in kernel memory. */
935 kaddrs
= kmalloc(addrs_size
, GFP_KERNEL
);
936 if (unlikely(!kaddrs
))
939 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
944 /* Walk through the addrs buffer and count the number of addresses. */
946 while (walk_size
< addrs_size
) {
947 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
952 sa_addr
= (struct sockaddr
*)addr_buf
;
953 af
= sctp_get_af_specific(sa_addr
->sa_family
);
955 /* If the address family is not supported or if this address
956 * causes the address buffer to overflow return EINVAL.
958 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
963 addr_buf
+= af
->sockaddr_len
;
964 walk_size
+= af
->sockaddr_len
;
969 case SCTP_BINDX_ADD_ADDR
:
970 err
= sctp_bindx_add(sk
, kaddrs
, addrcnt
);
973 err
= sctp_send_asconf_add_ip(sk
, kaddrs
, addrcnt
);
976 case SCTP_BINDX_REM_ADDR
:
977 err
= sctp_bindx_rem(sk
, kaddrs
, addrcnt
);
980 err
= sctp_send_asconf_del_ip(sk
, kaddrs
, addrcnt
);
994 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
996 * Common routine for handling connect() and sctp_connectx().
997 * Connect will come in with just a single address.
999 static int __sctp_connect(struct sock
* sk
,
1000 struct sockaddr
*kaddrs
,
1002 sctp_assoc_t
*assoc_id
)
1004 struct sctp_sock
*sp
;
1005 struct sctp_endpoint
*ep
;
1006 struct sctp_association
*asoc
= NULL
;
1007 struct sctp_association
*asoc2
;
1008 struct sctp_transport
*transport
;
1016 union sctp_addr
*sa_addr
= NULL
;
1018 unsigned short port
;
1019 unsigned int f_flags
= 0;
1024 /* connect() cannot be done on a socket that is already in ESTABLISHED
1025 * state - UDP-style peeled off socket or a TCP-style socket that
1026 * is already connected.
1027 * It cannot be done even on a TCP-style listening socket.
1029 if (sctp_sstate(sk
, ESTABLISHED
) ||
1030 (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))) {
1035 /* Walk through the addrs buffer and count the number of addresses. */
1037 while (walk_size
< addrs_size
) {
1038 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
1043 sa_addr
= (union sctp_addr
*)addr_buf
;
1044 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
1046 /* If the address family is not supported or if this address
1047 * causes the address buffer to overflow return EINVAL.
1049 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
1054 port
= ntohs(sa_addr
->v4
.sin_port
);
1056 /* Save current address so we can work with it */
1057 memcpy(&to
, sa_addr
, af
->sockaddr_len
);
1059 err
= sctp_verify_addr(sk
, &to
, af
->sockaddr_len
);
1063 /* Make sure the destination port is correctly set
1066 if (asoc
&& asoc
->peer
.port
&& asoc
->peer
.port
!= port
)
1070 /* Check if there already is a matching association on the
1071 * endpoint (other than the one created here).
1073 asoc2
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1074 if (asoc2
&& asoc2
!= asoc
) {
1075 if (asoc2
->state
>= SCTP_STATE_ESTABLISHED
)
1082 /* If we could not find a matching association on the endpoint,
1083 * make sure that there is no peeled-off association matching
1084 * the peer address even on another socket.
1086 if (sctp_endpoint_is_peeled_off(ep
, &to
)) {
1087 err
= -EADDRNOTAVAIL
;
1092 /* If a bind() or sctp_bindx() is not called prior to
1093 * an sctp_connectx() call, the system picks an
1094 * ephemeral port and will choose an address set
1095 * equivalent to binding with a wildcard address.
1097 if (!ep
->base
.bind_addr
.port
) {
1098 if (sctp_autobind(sk
)) {
1104 * If an unprivileged user inherits a 1-many
1105 * style socket with open associations on a
1106 * privileged port, it MAY be permitted to
1107 * accept new associations, but it SHOULD NOT
1108 * be permitted to open new associations.
1110 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1111 !capable(CAP_NET_BIND_SERVICE
)) {
1117 scope
= sctp_scope(&to
);
1118 asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1124 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
,
1132 /* Prime the peer's transport structures. */
1133 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
,
1141 addr_buf
+= af
->sockaddr_len
;
1142 walk_size
+= af
->sockaddr_len
;
1145 /* In case the user of sctp_connectx() wants an association
1146 * id back, assign one now.
1149 err
= sctp_assoc_set_id(asoc
, GFP_KERNEL
);
1154 err
= sctp_primitive_ASSOCIATE(asoc
, NULL
);
1159 /* Initialize sk's dport and daddr for getpeername() */
1160 inet_sk(sk
)->inet_dport
= htons(asoc
->peer
.port
);
1161 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
1162 af
->to_sk_daddr(sa_addr
, sk
);
1165 /* in-kernel sockets don't generally have a file allocated to them
1166 * if all they do is call sock_create_kern().
1168 if (sk
->sk_socket
->file
)
1169 f_flags
= sk
->sk_socket
->file
->f_flags
;
1171 timeo
= sock_sndtimeo(sk
, f_flags
& O_NONBLOCK
);
1173 err
= sctp_wait_for_connect(asoc
, &timeo
);
1174 if ((err
== 0 || err
== -EINPROGRESS
) && assoc_id
)
1175 *assoc_id
= asoc
->assoc_id
;
1177 /* Don't free association on exit. */
1182 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1183 " kaddrs: %p err: %d\n",
1186 sctp_association_free(asoc
);
1190 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1193 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1194 * sctp_assoc_t *asoc);
1196 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1197 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1198 * or IPv6 addresses.
1200 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1201 * Section 3.1.2 for this usage.
1203 * addrs is a pointer to an array of one or more socket addresses. Each
1204 * address is contained in its appropriate structure (i.e. struct
1205 * sockaddr_in or struct sockaddr_in6) the family of the address type
1206 * must be used to distengish the address length (note that this
1207 * representation is termed a "packed array" of addresses). The caller
1208 * specifies the number of addresses in the array with addrcnt.
1210 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1211 * the association id of the new association. On failure, sctp_connectx()
1212 * returns -1, and sets errno to the appropriate error code. The assoc_id
1213 * is not touched by the kernel.
1215 * For SCTP, the port given in each socket address must be the same, or
1216 * sctp_connectx() will fail, setting errno to EINVAL.
1218 * An application can use sctp_connectx to initiate an association with
1219 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1220 * allows a caller to specify multiple addresses at which a peer can be
1221 * reached. The way the SCTP stack uses the list of addresses to set up
1222 * the association is implementation dependent. This function only
1223 * specifies that the stack will try to make use of all the addresses in
1224 * the list when needed.
1226 * Note that the list of addresses passed in is only used for setting up
1227 * the association. It does not necessarily equal the set of addresses
1228 * the peer uses for the resulting association. If the caller wants to
1229 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1230 * retrieve them after the association has been set up.
1232 * Basically do nothing but copying the addresses from user to kernel
1233 * land and invoking either sctp_connectx(). This is used for tunneling
1234 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1236 * We don't use copy_from_user() for optimization: we first do the
1237 * sanity checks (buffer size -fast- and access check-healthy
1238 * pointer); if all of those succeed, then we can alloc the memory
1239 * (expensive operation) needed to copy the data to kernel. Then we do
1240 * the copying without checking the user space area
1241 * (__copy_from_user()).
1243 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1246 * sk The sk of the socket
1247 * addrs The pointer to the addresses in user land
1248 * addrssize Size of the addrs buffer
1250 * Returns >=0 if ok, <0 errno code on error.
1252 SCTP_STATIC
int __sctp_setsockopt_connectx(struct sock
* sk
,
1253 struct sockaddr __user
*addrs
,
1255 sctp_assoc_t
*assoc_id
)
1258 struct sockaddr
*kaddrs
;
1260 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1261 __func__
, sk
, addrs
, addrs_size
);
1263 if (unlikely(addrs_size
<= 0))
1266 /* Check the user passed a healthy pointer. */
1267 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
1270 /* Alloc space for the address array in kernel memory. */
1271 kaddrs
= kmalloc(addrs_size
, GFP_KERNEL
);
1272 if (unlikely(!kaddrs
))
1275 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
1278 err
= __sctp_connect(sk
, kaddrs
, addrs_size
, assoc_id
);
1287 * This is an older interface. It's kept for backward compatibility
1288 * to the option that doesn't provide association id.
1290 SCTP_STATIC
int sctp_setsockopt_connectx_old(struct sock
* sk
,
1291 struct sockaddr __user
*addrs
,
1294 return __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, NULL
);
1298 * New interface for the API. The since the API is done with a socket
1299 * option, to make it simple we feed back the association id is as a return
1300 * indication to the call. Error is always negative and association id is
1303 SCTP_STATIC
int sctp_setsockopt_connectx(struct sock
* sk
,
1304 struct sockaddr __user
*addrs
,
1307 sctp_assoc_t assoc_id
= 0;
1310 err
= __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, &assoc_id
);
1319 * New (hopefully final) interface for the API.
1320 * We use the sctp_getaddrs_old structure so that use-space library
1321 * can avoid any unnecessary allocations. The only defferent part
1322 * is that we store the actual length of the address buffer into the
1323 * addrs_num structure member. That way we can re-use the existing
1326 SCTP_STATIC
int sctp_getsockopt_connectx3(struct sock
* sk
, int len
,
1327 char __user
*optval
,
1330 struct sctp_getaddrs_old param
;
1331 sctp_assoc_t assoc_id
= 0;
1334 if (len
< sizeof(param
))
1337 if (copy_from_user(¶m
, optval
, sizeof(param
)))
1340 err
= __sctp_setsockopt_connectx(sk
,
1341 (struct sockaddr __user
*)param
.addrs
,
1342 param
.addr_num
, &assoc_id
);
1344 if (err
== 0 || err
== -EINPROGRESS
) {
1345 if (copy_to_user(optval
, &assoc_id
, sizeof(assoc_id
)))
1347 if (put_user(sizeof(assoc_id
), optlen
))
1354 /* API 3.1.4 close() - UDP Style Syntax
1355 * Applications use close() to perform graceful shutdown (as described in
1356 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1357 * by a UDP-style socket.
1361 * ret = close(int sd);
1363 * sd - the socket descriptor of the associations to be closed.
1365 * To gracefully shutdown a specific association represented by the
1366 * UDP-style socket, an application should use the sendmsg() call,
1367 * passing no user data, but including the appropriate flag in the
1368 * ancillary data (see Section xxxx).
1370 * If sd in the close() call is a branched-off socket representing only
1371 * one association, the shutdown is performed on that association only.
1373 * 4.1.6 close() - TCP Style Syntax
1375 * Applications use close() to gracefully close down an association.
1379 * int close(int sd);
1381 * sd - the socket descriptor of the association to be closed.
1383 * After an application calls close() on a socket descriptor, no further
1384 * socket operations will succeed on that descriptor.
1386 * API 7.1.4 SO_LINGER
1388 * An application using the TCP-style socket can use this option to
1389 * perform the SCTP ABORT primitive. The linger option structure is:
1392 * int l_onoff; // option on/off
1393 * int l_linger; // linger time
1396 * To enable the option, set l_onoff to 1. If the l_linger value is set
1397 * to 0, calling close() is the same as the ABORT primitive. If the
1398 * value is set to a negative value, the setsockopt() call will return
1399 * an error. If the value is set to a positive value linger_time, the
1400 * close() can be blocked for at most linger_time ms. If the graceful
1401 * shutdown phase does not finish during this period, close() will
1402 * return but the graceful shutdown phase continues in the system.
1404 SCTP_STATIC
void sctp_close(struct sock
*sk
, long timeout
)
1406 struct sctp_endpoint
*ep
;
1407 struct sctp_association
*asoc
;
1408 struct list_head
*pos
, *temp
;
1410 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk
, timeout
);
1413 sk
->sk_shutdown
= SHUTDOWN_MASK
;
1414 sk
->sk_state
= SCTP_SS_CLOSING
;
1416 ep
= sctp_sk(sk
)->ep
;
1418 /* Walk all associations on an endpoint. */
1419 list_for_each_safe(pos
, temp
, &ep
->asocs
) {
1420 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
1422 if (sctp_style(sk
, TCP
)) {
1423 /* A closed association can still be in the list if
1424 * it belongs to a TCP-style listening socket that is
1425 * not yet accepted. If so, free it. If not, send an
1426 * ABORT or SHUTDOWN based on the linger options.
1428 if (sctp_state(asoc
, CLOSED
)) {
1429 sctp_unhash_established(asoc
);
1430 sctp_association_free(asoc
);
1435 if (sock_flag(sk
, SOCK_LINGER
) && !sk
->sk_lingertime
) {
1436 struct sctp_chunk
*chunk
;
1438 chunk
= sctp_make_abort_user(asoc
, NULL
, 0);
1440 sctp_primitive_ABORT(asoc
, chunk
);
1442 sctp_primitive_SHUTDOWN(asoc
, NULL
);
1445 /* Clean up any skbs sitting on the receive queue. */
1446 sctp_queue_purge_ulpevents(&sk
->sk_receive_queue
);
1447 sctp_queue_purge_ulpevents(&sctp_sk(sk
)->pd_lobby
);
1449 /* On a TCP-style socket, block for at most linger_time if set. */
1450 if (sctp_style(sk
, TCP
) && timeout
)
1451 sctp_wait_for_close(sk
, timeout
);
1453 /* This will run the backlog queue. */
1454 sctp_release_sock(sk
);
1456 /* Supposedly, no process has access to the socket, but
1457 * the net layers still may.
1459 sctp_local_bh_disable();
1460 sctp_bh_lock_sock(sk
);
1462 /* Hold the sock, since sk_common_release() will put sock_put()
1463 * and we have just a little more cleanup.
1466 sk_common_release(sk
);
1468 sctp_bh_unlock_sock(sk
);
1469 sctp_local_bh_enable();
1473 SCTP_DBG_OBJCNT_DEC(sock
);
1476 /* Handle EPIPE error. */
1477 static int sctp_error(struct sock
*sk
, int flags
, int err
)
1480 err
= sock_error(sk
) ? : -EPIPE
;
1481 if (err
== -EPIPE
&& !(flags
& MSG_NOSIGNAL
))
1482 send_sig(SIGPIPE
, current
, 0);
1486 /* API 3.1.3 sendmsg() - UDP Style Syntax
1488 * An application uses sendmsg() and recvmsg() calls to transmit data to
1489 * and receive data from its peer.
1491 * ssize_t sendmsg(int socket, const struct msghdr *message,
1494 * socket - the socket descriptor of the endpoint.
1495 * message - pointer to the msghdr structure which contains a single
1496 * user message and possibly some ancillary data.
1498 * See Section 5 for complete description of the data
1501 * flags - flags sent or received with the user message, see Section
1502 * 5 for complete description of the flags.
1504 * Note: This function could use a rewrite especially when explicit
1505 * connect support comes in.
1507 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1509 SCTP_STATIC
int sctp_msghdr_parse(const struct msghdr
*, sctp_cmsgs_t
*);
1511 SCTP_STATIC
int sctp_sendmsg(struct kiocb
*iocb
, struct sock
*sk
,
1512 struct msghdr
*msg
, size_t msg_len
)
1514 struct sctp_sock
*sp
;
1515 struct sctp_endpoint
*ep
;
1516 struct sctp_association
*new_asoc
=NULL
, *asoc
=NULL
;
1517 struct sctp_transport
*transport
, *chunk_tp
;
1518 struct sctp_chunk
*chunk
;
1520 struct sockaddr
*msg_name
= NULL
;
1521 struct sctp_sndrcvinfo default_sinfo
;
1522 struct sctp_sndrcvinfo
*sinfo
;
1523 struct sctp_initmsg
*sinit
;
1524 sctp_assoc_t associd
= 0;
1525 sctp_cmsgs_t cmsgs
= { NULL
};
1529 __u16 sinfo_flags
= 0;
1530 struct sctp_datamsg
*datamsg
;
1531 int msg_flags
= msg
->msg_flags
;
1533 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1540 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep
);
1542 /* We cannot send a message over a TCP-style listening socket. */
1543 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
)) {
1548 /* Parse out the SCTP CMSGs. */
1549 err
= sctp_msghdr_parse(msg
, &cmsgs
);
1552 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err
);
1556 /* Fetch the destination address for this packet. This
1557 * address only selects the association--it is not necessarily
1558 * the address we will send to.
1559 * For a peeled-off socket, msg_name is ignored.
1561 if (!sctp_style(sk
, UDP_HIGH_BANDWIDTH
) && msg
->msg_name
) {
1562 int msg_namelen
= msg
->msg_namelen
;
1564 err
= sctp_verify_addr(sk
, (union sctp_addr
*)msg
->msg_name
,
1569 if (msg_namelen
> sizeof(to
))
1570 msg_namelen
= sizeof(to
);
1571 memcpy(&to
, msg
->msg_name
, msg_namelen
);
1572 msg_name
= msg
->msg_name
;
1578 /* Did the user specify SNDRCVINFO? */
1580 sinfo_flags
= sinfo
->sinfo_flags
;
1581 associd
= sinfo
->sinfo_assoc_id
;
1584 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1585 msg_len
, sinfo_flags
);
1587 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1588 if (sctp_style(sk
, TCP
) && (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
))) {
1593 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1594 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1595 * If SCTP_ABORT is set, the message length could be non zero with
1596 * the msg_iov set to the user abort reason.
1598 if (((sinfo_flags
& SCTP_EOF
) && (msg_len
> 0)) ||
1599 (!(sinfo_flags
& (SCTP_EOF
|SCTP_ABORT
)) && (msg_len
== 0))) {
1604 /* If SCTP_ADDR_OVER is set, there must be an address
1605 * specified in msg_name.
1607 if ((sinfo_flags
& SCTP_ADDR_OVER
) && (!msg
->msg_name
)) {
1614 SCTP_DEBUG_PRINTK("About to look up association.\n");
1618 /* If a msg_name has been specified, assume this is to be used. */
1620 /* Look for a matching association on the endpoint. */
1621 asoc
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1623 /* If we could not find a matching association on the
1624 * endpoint, make sure that it is not a TCP-style
1625 * socket that already has an association or there is
1626 * no peeled-off association on another socket.
1628 if ((sctp_style(sk
, TCP
) &&
1629 sctp_sstate(sk
, ESTABLISHED
)) ||
1630 sctp_endpoint_is_peeled_off(ep
, &to
)) {
1631 err
= -EADDRNOTAVAIL
;
1636 asoc
= sctp_id2assoc(sk
, associd
);
1644 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc
);
1646 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1647 * socket that has an association in CLOSED state. This can
1648 * happen when an accepted socket has an association that is
1651 if (sctp_state(asoc
, CLOSED
) && sctp_style(sk
, TCP
)) {
1656 if (sinfo_flags
& SCTP_EOF
) {
1657 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1659 sctp_primitive_SHUTDOWN(asoc
, NULL
);
1663 if (sinfo_flags
& SCTP_ABORT
) {
1665 chunk
= sctp_make_abort_user(asoc
, msg
, msg_len
);
1671 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc
);
1672 sctp_primitive_ABORT(asoc
, chunk
);
1678 /* Do we need to create the association? */
1680 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1682 if (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
)) {
1687 /* Check for invalid stream against the stream counts,
1688 * either the default or the user specified stream counts.
1691 if (!sinit
|| (sinit
&& !sinit
->sinit_num_ostreams
)) {
1692 /* Check against the defaults. */
1693 if (sinfo
->sinfo_stream
>=
1694 sp
->initmsg
.sinit_num_ostreams
) {
1699 /* Check against the requested. */
1700 if (sinfo
->sinfo_stream
>=
1701 sinit
->sinit_num_ostreams
) {
1709 * API 3.1.2 bind() - UDP Style Syntax
1710 * If a bind() or sctp_bindx() is not called prior to a
1711 * sendmsg() call that initiates a new association, the
1712 * system picks an ephemeral port and will choose an address
1713 * set equivalent to binding with a wildcard address.
1715 if (!ep
->base
.bind_addr
.port
) {
1716 if (sctp_autobind(sk
)) {
1722 * If an unprivileged user inherits a one-to-many
1723 * style socket with open associations on a privileged
1724 * port, it MAY be permitted to accept new associations,
1725 * but it SHOULD NOT be permitted to open new
1728 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1729 !capable(CAP_NET_BIND_SERVICE
)) {
1735 scope
= sctp_scope(&to
);
1736 new_asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1742 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
, GFP_KERNEL
);
1748 /* If the SCTP_INIT ancillary data is specified, set all
1749 * the association init values accordingly.
1752 if (sinit
->sinit_num_ostreams
) {
1753 asoc
->c
.sinit_num_ostreams
=
1754 sinit
->sinit_num_ostreams
;
1756 if (sinit
->sinit_max_instreams
) {
1757 asoc
->c
.sinit_max_instreams
=
1758 sinit
->sinit_max_instreams
;
1760 if (sinit
->sinit_max_attempts
) {
1761 asoc
->max_init_attempts
1762 = sinit
->sinit_max_attempts
;
1764 if (sinit
->sinit_max_init_timeo
) {
1765 asoc
->max_init_timeo
=
1766 msecs_to_jiffies(sinit
->sinit_max_init_timeo
);
1770 /* Prime the peer's transport structures. */
1771 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
, SCTP_UNKNOWN
);
1778 /* ASSERT: we have a valid association at this point. */
1779 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1782 /* If the user didn't specify SNDRCVINFO, make up one with
1785 memset(&default_sinfo
, 0, sizeof(default_sinfo
));
1786 default_sinfo
.sinfo_stream
= asoc
->default_stream
;
1787 default_sinfo
.sinfo_flags
= asoc
->default_flags
;
1788 default_sinfo
.sinfo_ppid
= asoc
->default_ppid
;
1789 default_sinfo
.sinfo_context
= asoc
->default_context
;
1790 default_sinfo
.sinfo_timetolive
= asoc
->default_timetolive
;
1791 default_sinfo
.sinfo_assoc_id
= sctp_assoc2id(asoc
);
1792 sinfo
= &default_sinfo
;
1795 /* API 7.1.7, the sndbuf size per association bounds the
1796 * maximum size of data that can be sent in a single send call.
1798 if (msg_len
> sk
->sk_sndbuf
) {
1803 if (asoc
->pmtu_pending
)
1804 sctp_assoc_pending_pmtu(asoc
);
1806 /* If fragmentation is disabled and the message length exceeds the
1807 * association fragmentation point, return EMSGSIZE. The I-D
1808 * does not specify what this error is, but this looks like
1811 if (sctp_sk(sk
)->disable_fragments
&& (msg_len
> asoc
->frag_point
)) {
1816 /* Check for invalid stream. */
1817 if (sinfo
->sinfo_stream
>= asoc
->c
.sinit_num_ostreams
) {
1822 timeo
= sock_sndtimeo(sk
, msg
->msg_flags
& MSG_DONTWAIT
);
1823 if (!sctp_wspace(asoc
)) {
1824 err
= sctp_wait_for_sndbuf(asoc
, &timeo
, msg_len
);
1829 /* If an address is passed with the sendto/sendmsg call, it is used
1830 * to override the primary destination address in the TCP model, or
1831 * when SCTP_ADDR_OVER flag is set in the UDP model.
1833 if ((sctp_style(sk
, TCP
) && msg_name
) ||
1834 (sinfo_flags
& SCTP_ADDR_OVER
)) {
1835 chunk_tp
= sctp_assoc_lookup_paddr(asoc
, &to
);
1843 /* Auto-connect, if we aren't connected already. */
1844 if (sctp_state(asoc
, CLOSED
)) {
1845 err
= sctp_primitive_ASSOCIATE(asoc
, NULL
);
1848 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1851 /* Break the message into multiple chunks of maximum size. */
1852 datamsg
= sctp_datamsg_from_user(asoc
, sinfo
, msg
, msg_len
);
1858 /* Now send the (possibly) fragmented message. */
1859 list_for_each_entry(chunk
, &datamsg
->chunks
, frag_list
) {
1860 sctp_chunk_hold(chunk
);
1862 /* Do accounting for the write space. */
1863 sctp_set_owner_w(chunk
);
1865 chunk
->transport
= chunk_tp
;
1868 /* Send it to the lower layers. Note: all chunks
1869 * must either fail or succeed. The lower layer
1870 * works that way today. Keep it that way or this
1873 err
= sctp_primitive_SEND(asoc
, datamsg
);
1874 /* Did the lower layer accept the chunk? */
1876 sctp_datamsg_free(datamsg
);
1878 sctp_datamsg_put(datamsg
);
1880 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1887 /* If we are already past ASSOCIATE, the lower
1888 * layers are responsible for association cleanup.
1894 sctp_association_free(asoc
);
1896 sctp_release_sock(sk
);
1899 return sctp_error(sk
, msg_flags
, err
);
1906 err
= sock_error(sk
);
1916 /* This is an extended version of skb_pull() that removes the data from the
1917 * start of a skb even when data is spread across the list of skb's in the
1918 * frag_list. len specifies the total amount of data that needs to be removed.
1919 * when 'len' bytes could be removed from the skb, it returns 0.
1920 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1921 * could not be removed.
1923 static int sctp_skb_pull(struct sk_buff
*skb
, int len
)
1925 struct sk_buff
*list
;
1926 int skb_len
= skb_headlen(skb
);
1929 if (len
<= skb_len
) {
1930 __skb_pull(skb
, len
);
1934 __skb_pull(skb
, skb_len
);
1936 skb_walk_frags(skb
, list
) {
1937 rlen
= sctp_skb_pull(list
, len
);
1938 skb
->len
-= (len
-rlen
);
1939 skb
->data_len
-= (len
-rlen
);
1950 /* API 3.1.3 recvmsg() - UDP Style Syntax
1952 * ssize_t recvmsg(int socket, struct msghdr *message,
1955 * socket - the socket descriptor of the endpoint.
1956 * message - pointer to the msghdr structure which contains a single
1957 * user message and possibly some ancillary data.
1959 * See Section 5 for complete description of the data
1962 * flags - flags sent or received with the user message, see Section
1963 * 5 for complete description of the flags.
1965 static struct sk_buff
*sctp_skb_recv_datagram(struct sock
*, int, int, int *);
1967 SCTP_STATIC
int sctp_recvmsg(struct kiocb
*iocb
, struct sock
*sk
,
1968 struct msghdr
*msg
, size_t len
, int noblock
,
1969 int flags
, int *addr_len
)
1971 struct sctp_ulpevent
*event
= NULL
;
1972 struct sctp_sock
*sp
= sctp_sk(sk
);
1973 struct sk_buff
*skb
;
1978 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1979 "0x%x, %s: %p)\n", "sk", sk
, "msghdr", msg
,
1980 "len", len
, "knoblauch", noblock
,
1981 "flags", flags
, "addr_len", addr_len
);
1985 if (sctp_style(sk
, TCP
) && !sctp_sstate(sk
, ESTABLISHED
)) {
1990 skb
= sctp_skb_recv_datagram(sk
, flags
, noblock
, &err
);
1994 /* Get the total length of the skb including any skb's in the
2003 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
2005 event
= sctp_skb2event(skb
);
2010 sock_recv_ts_and_drops(msg
, sk
, skb
);
2011 if (sctp_ulpevent_is_notification(event
)) {
2012 msg
->msg_flags
|= MSG_NOTIFICATION
;
2013 sp
->pf
->event_msgname(event
, msg
->msg_name
, addr_len
);
2015 sp
->pf
->skb_msgname(skb
, msg
->msg_name
, addr_len
);
2018 /* Check if we allow SCTP_SNDRCVINFO. */
2019 if (sp
->subscribe
.sctp_data_io_event
)
2020 sctp_ulpevent_read_sndrcvinfo(event
, msg
);
2022 /* FIXME: we should be calling IP/IPv6 layers. */
2023 if (sk
->sk_protinfo
.af_inet
.cmsg_flags
)
2024 ip_cmsg_recv(msg
, skb
);
2029 /* If skb's length exceeds the user's buffer, update the skb and
2030 * push it back to the receive_queue so that the next call to
2031 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2033 if (skb_len
> copied
) {
2034 msg
->msg_flags
&= ~MSG_EOR
;
2035 if (flags
& MSG_PEEK
)
2037 sctp_skb_pull(skb
, copied
);
2038 skb_queue_head(&sk
->sk_receive_queue
, skb
);
2040 /* When only partial message is copied to the user, increase
2041 * rwnd by that amount. If all the data in the skb is read,
2042 * rwnd is updated when the event is freed.
2044 if (!sctp_ulpevent_is_notification(event
))
2045 sctp_assoc_rwnd_increase(event
->asoc
, copied
);
2047 } else if ((event
->msg_flags
& MSG_NOTIFICATION
) ||
2048 (event
->msg_flags
& MSG_EOR
))
2049 msg
->msg_flags
|= MSG_EOR
;
2051 msg
->msg_flags
&= ~MSG_EOR
;
2054 if (flags
& MSG_PEEK
) {
2055 /* Release the skb reference acquired after peeking the skb in
2056 * sctp_skb_recv_datagram().
2060 /* Free the event which includes releasing the reference to
2061 * the owner of the skb, freeing the skb and updating the
2064 sctp_ulpevent_free(event
);
2067 sctp_release_sock(sk
);
2071 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2073 * This option is a on/off flag. If enabled no SCTP message
2074 * fragmentation will be performed. Instead if a message being sent
2075 * exceeds the current PMTU size, the message will NOT be sent and
2076 * instead a error will be indicated to the user.
2078 static int sctp_setsockopt_disable_fragments(struct sock
*sk
,
2079 char __user
*optval
,
2080 unsigned int optlen
)
2084 if (optlen
< sizeof(int))
2087 if (get_user(val
, (int __user
*)optval
))
2090 sctp_sk(sk
)->disable_fragments
= (val
== 0) ? 0 : 1;
2095 static int sctp_setsockopt_events(struct sock
*sk
, char __user
*optval
,
2096 unsigned int optlen
)
2098 if (optlen
> sizeof(struct sctp_event_subscribe
))
2100 if (copy_from_user(&sctp_sk(sk
)->subscribe
, optval
, optlen
))
2105 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2107 * This socket option is applicable to the UDP-style socket only. When
2108 * set it will cause associations that are idle for more than the
2109 * specified number of seconds to automatically close. An association
2110 * being idle is defined an association that has NOT sent or received
2111 * user data. The special value of '0' indicates that no automatic
2112 * close of any associations should be performed. The option expects an
2113 * integer defining the number of seconds of idle time before an
2114 * association is closed.
2116 static int sctp_setsockopt_autoclose(struct sock
*sk
, char __user
*optval
,
2117 unsigned int optlen
)
2119 struct sctp_sock
*sp
= sctp_sk(sk
);
2121 /* Applicable to UDP-style socket only */
2122 if (sctp_style(sk
, TCP
))
2124 if (optlen
!= sizeof(int))
2126 if (copy_from_user(&sp
->autoclose
, optval
, optlen
))
2128 /* make sure it won't exceed MAX_SCHEDULE_TIMEOUT */
2129 sp
->autoclose
= min_t(long, sp
->autoclose
, MAX_SCHEDULE_TIMEOUT
/ HZ
);
2134 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2136 * Applications can enable or disable heartbeats for any peer address of
2137 * an association, modify an address's heartbeat interval, force a
2138 * heartbeat to be sent immediately, and adjust the address's maximum
2139 * number of retransmissions sent before an address is considered
2140 * unreachable. The following structure is used to access and modify an
2141 * address's parameters:
2143 * struct sctp_paddrparams {
2144 * sctp_assoc_t spp_assoc_id;
2145 * struct sockaddr_storage spp_address;
2146 * uint32_t spp_hbinterval;
2147 * uint16_t spp_pathmaxrxt;
2148 * uint32_t spp_pathmtu;
2149 * uint32_t spp_sackdelay;
2150 * uint32_t spp_flags;
2153 * spp_assoc_id - (one-to-many style socket) This is filled in the
2154 * application, and identifies the association for
2156 * spp_address - This specifies which address is of interest.
2157 * spp_hbinterval - This contains the value of the heartbeat interval,
2158 * in milliseconds. If a value of zero
2159 * is present in this field then no changes are to
2160 * be made to this parameter.
2161 * spp_pathmaxrxt - This contains the maximum number of
2162 * retransmissions before this address shall be
2163 * considered unreachable. If a value of zero
2164 * is present in this field then no changes are to
2165 * be made to this parameter.
2166 * spp_pathmtu - When Path MTU discovery is disabled the value
2167 * specified here will be the "fixed" path mtu.
2168 * Note that if the spp_address field is empty
2169 * then all associations on this address will
2170 * have this fixed path mtu set upon them.
2172 * spp_sackdelay - When delayed sack is enabled, this value specifies
2173 * the number of milliseconds that sacks will be delayed
2174 * for. This value will apply to all addresses of an
2175 * association if the spp_address field is empty. Note
2176 * also, that if delayed sack is enabled and this
2177 * value is set to 0, no change is made to the last
2178 * recorded delayed sack timer value.
2180 * spp_flags - These flags are used to control various features
2181 * on an association. The flag field may contain
2182 * zero or more of the following options.
2184 * SPP_HB_ENABLE - Enable heartbeats on the
2185 * specified address. Note that if the address
2186 * field is empty all addresses for the association
2187 * have heartbeats enabled upon them.
2189 * SPP_HB_DISABLE - Disable heartbeats on the
2190 * speicifed address. Note that if the address
2191 * field is empty all addresses for the association
2192 * will have their heartbeats disabled. Note also
2193 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2194 * mutually exclusive, only one of these two should
2195 * be specified. Enabling both fields will have
2196 * undetermined results.
2198 * SPP_HB_DEMAND - Request a user initiated heartbeat
2199 * to be made immediately.
2201 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2202 * heartbeat delayis to be set to the value of 0
2205 * SPP_PMTUD_ENABLE - This field will enable PMTU
2206 * discovery upon the specified address. Note that
2207 * if the address feild is empty then all addresses
2208 * on the association are effected.
2210 * SPP_PMTUD_DISABLE - This field will disable PMTU
2211 * discovery upon the specified address. Note that
2212 * if the address feild is empty then all addresses
2213 * on the association are effected. Not also that
2214 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2215 * exclusive. Enabling both will have undetermined
2218 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2219 * on delayed sack. The time specified in spp_sackdelay
2220 * is used to specify the sack delay for this address. Note
2221 * that if spp_address is empty then all addresses will
2222 * enable delayed sack and take on the sack delay
2223 * value specified in spp_sackdelay.
2224 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2225 * off delayed sack. If the spp_address field is blank then
2226 * delayed sack is disabled for the entire association. Note
2227 * also that this field is mutually exclusive to
2228 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2231 static int sctp_apply_peer_addr_params(struct sctp_paddrparams
*params
,
2232 struct sctp_transport
*trans
,
2233 struct sctp_association
*asoc
,
2234 struct sctp_sock
*sp
,
2237 int sackdelay_change
)
2241 if (params
->spp_flags
& SPP_HB_DEMAND
&& trans
) {
2242 error
= sctp_primitive_REQUESTHEARTBEAT (trans
->asoc
, trans
);
2247 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2248 * this field is ignored. Note also that a value of zero indicates
2249 * the current setting should be left unchanged.
2251 if (params
->spp_flags
& SPP_HB_ENABLE
) {
2253 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2254 * set. This lets us use 0 value when this flag
2257 if (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)
2258 params
->spp_hbinterval
= 0;
2260 if (params
->spp_hbinterval
||
2261 (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)) {
2264 msecs_to_jiffies(params
->spp_hbinterval
);
2267 msecs_to_jiffies(params
->spp_hbinterval
);
2269 sp
->hbinterval
= params
->spp_hbinterval
;
2276 trans
->param_flags
=
2277 (trans
->param_flags
& ~SPP_HB
) | hb_change
;
2280 (asoc
->param_flags
& ~SPP_HB
) | hb_change
;
2283 (sp
->param_flags
& ~SPP_HB
) | hb_change
;
2287 /* When Path MTU discovery is disabled the value specified here will
2288 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2289 * include the flag SPP_PMTUD_DISABLE for this field to have any
2292 if ((params
->spp_flags
& SPP_PMTUD_DISABLE
) && params
->spp_pathmtu
) {
2294 trans
->pathmtu
= params
->spp_pathmtu
;
2295 sctp_assoc_sync_pmtu(asoc
);
2297 asoc
->pathmtu
= params
->spp_pathmtu
;
2298 sctp_frag_point(asoc
, params
->spp_pathmtu
);
2300 sp
->pathmtu
= params
->spp_pathmtu
;
2306 int update
= (trans
->param_flags
& SPP_PMTUD_DISABLE
) &&
2307 (params
->spp_flags
& SPP_PMTUD_ENABLE
);
2308 trans
->param_flags
=
2309 (trans
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2311 sctp_transport_pmtu(trans
, sctp_opt2sk(sp
));
2312 sctp_assoc_sync_pmtu(asoc
);
2316 (asoc
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2319 (sp
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2323 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2324 * value of this field is ignored. Note also that a value of zero
2325 * indicates the current setting should be left unchanged.
2327 if ((params
->spp_flags
& SPP_SACKDELAY_ENABLE
) && params
->spp_sackdelay
) {
2330 msecs_to_jiffies(params
->spp_sackdelay
);
2333 msecs_to_jiffies(params
->spp_sackdelay
);
2335 sp
->sackdelay
= params
->spp_sackdelay
;
2339 if (sackdelay_change
) {
2341 trans
->param_flags
=
2342 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2346 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2350 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2355 /* Note that a value of zero indicates the current setting should be
2358 if (params
->spp_pathmaxrxt
) {
2360 trans
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2362 asoc
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2364 sp
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2371 static int sctp_setsockopt_peer_addr_params(struct sock
*sk
,
2372 char __user
*optval
,
2373 unsigned int optlen
)
2375 struct sctp_paddrparams params
;
2376 struct sctp_transport
*trans
= NULL
;
2377 struct sctp_association
*asoc
= NULL
;
2378 struct sctp_sock
*sp
= sctp_sk(sk
);
2380 int hb_change
, pmtud_change
, sackdelay_change
;
2382 if (optlen
!= sizeof(struct sctp_paddrparams
))
2385 if (copy_from_user(¶ms
, optval
, optlen
))
2388 /* Validate flags and value parameters. */
2389 hb_change
= params
.spp_flags
& SPP_HB
;
2390 pmtud_change
= params
.spp_flags
& SPP_PMTUD
;
2391 sackdelay_change
= params
.spp_flags
& SPP_SACKDELAY
;
2393 if (hb_change
== SPP_HB
||
2394 pmtud_change
== SPP_PMTUD
||
2395 sackdelay_change
== SPP_SACKDELAY
||
2396 params
.spp_sackdelay
> 500 ||
2397 (params
.spp_pathmtu
&&
2398 params
.spp_pathmtu
< SCTP_DEFAULT_MINSEGMENT
))
2401 /* If an address other than INADDR_ANY is specified, and
2402 * no transport is found, then the request is invalid.
2404 if (!sctp_is_any(sk
, ( union sctp_addr
*)¶ms
.spp_address
)) {
2405 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
2406 params
.spp_assoc_id
);
2411 /* Get association, if assoc_id != 0 and the socket is a one
2412 * to many style socket, and an association was not found, then
2413 * the id was invalid.
2415 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
2416 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
))
2419 /* Heartbeat demand can only be sent on a transport or
2420 * association, but not a socket.
2422 if (params
.spp_flags
& SPP_HB_DEMAND
&& !trans
&& !asoc
)
2425 /* Process parameters. */
2426 error
= sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2427 hb_change
, pmtud_change
,
2433 /* If changes are for association, also apply parameters to each
2436 if (!trans
&& asoc
) {
2437 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2439 sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2440 hb_change
, pmtud_change
,
2449 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2451 * This option will effect the way delayed acks are performed. This
2452 * option allows you to get or set the delayed ack time, in
2453 * milliseconds. It also allows changing the delayed ack frequency.
2454 * Changing the frequency to 1 disables the delayed sack algorithm. If
2455 * the assoc_id is 0, then this sets or gets the endpoints default
2456 * values. If the assoc_id field is non-zero, then the set or get
2457 * effects the specified association for the one to many model (the
2458 * assoc_id field is ignored by the one to one model). Note that if
2459 * sack_delay or sack_freq are 0 when setting this option, then the
2460 * current values will remain unchanged.
2462 * struct sctp_sack_info {
2463 * sctp_assoc_t sack_assoc_id;
2464 * uint32_t sack_delay;
2465 * uint32_t sack_freq;
2468 * sack_assoc_id - This parameter, indicates which association the user
2469 * is performing an action upon. Note that if this field's value is
2470 * zero then the endpoints default value is changed (effecting future
2471 * associations only).
2473 * sack_delay - This parameter contains the number of milliseconds that
2474 * the user is requesting the delayed ACK timer be set to. Note that
2475 * this value is defined in the standard to be between 200 and 500
2478 * sack_freq - This parameter contains the number of packets that must
2479 * be received before a sack is sent without waiting for the delay
2480 * timer to expire. The default value for this is 2, setting this
2481 * value to 1 will disable the delayed sack algorithm.
2484 static int sctp_setsockopt_delayed_ack(struct sock
*sk
,
2485 char __user
*optval
, unsigned int optlen
)
2487 struct sctp_sack_info params
;
2488 struct sctp_transport
*trans
= NULL
;
2489 struct sctp_association
*asoc
= NULL
;
2490 struct sctp_sock
*sp
= sctp_sk(sk
);
2492 if (optlen
== sizeof(struct sctp_sack_info
)) {
2493 if (copy_from_user(¶ms
, optval
, optlen
))
2496 if (params
.sack_delay
== 0 && params
.sack_freq
== 0)
2498 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
2499 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2500 pr_warn("Use struct sctp_sack_info instead\n");
2501 if (copy_from_user(¶ms
, optval
, optlen
))
2504 if (params
.sack_delay
== 0)
2505 params
.sack_freq
= 1;
2507 params
.sack_freq
= 0;
2511 /* Validate value parameter. */
2512 if (params
.sack_delay
> 500)
2515 /* Get association, if sack_assoc_id != 0 and the socket is a one
2516 * to many style socket, and an association was not found, then
2517 * the id was invalid.
2519 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
2520 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
2523 if (params
.sack_delay
) {
2526 msecs_to_jiffies(params
.sack_delay
);
2528 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2529 SPP_SACKDELAY_ENABLE
;
2531 sp
->sackdelay
= params
.sack_delay
;
2533 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2534 SPP_SACKDELAY_ENABLE
;
2538 if (params
.sack_freq
== 1) {
2541 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2542 SPP_SACKDELAY_DISABLE
;
2545 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2546 SPP_SACKDELAY_DISABLE
;
2548 } else if (params
.sack_freq
> 1) {
2550 asoc
->sackfreq
= params
.sack_freq
;
2552 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2553 SPP_SACKDELAY_ENABLE
;
2555 sp
->sackfreq
= params
.sack_freq
;
2557 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2558 SPP_SACKDELAY_ENABLE
;
2562 /* If change is for association, also apply to each transport. */
2564 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2566 if (params
.sack_delay
) {
2568 msecs_to_jiffies(params
.sack_delay
);
2569 trans
->param_flags
=
2570 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2571 SPP_SACKDELAY_ENABLE
;
2573 if (params
.sack_freq
== 1) {
2574 trans
->param_flags
=
2575 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2576 SPP_SACKDELAY_DISABLE
;
2577 } else if (params
.sack_freq
> 1) {
2578 trans
->sackfreq
= params
.sack_freq
;
2579 trans
->param_flags
=
2580 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2581 SPP_SACKDELAY_ENABLE
;
2589 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2591 * Applications can specify protocol parameters for the default association
2592 * initialization. The option name argument to setsockopt() and getsockopt()
2595 * Setting initialization parameters is effective only on an unconnected
2596 * socket (for UDP-style sockets only future associations are effected
2597 * by the change). With TCP-style sockets, this option is inherited by
2598 * sockets derived from a listener socket.
2600 static int sctp_setsockopt_initmsg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2602 struct sctp_initmsg sinit
;
2603 struct sctp_sock
*sp
= sctp_sk(sk
);
2605 if (optlen
!= sizeof(struct sctp_initmsg
))
2607 if (copy_from_user(&sinit
, optval
, optlen
))
2610 if (sinit
.sinit_num_ostreams
)
2611 sp
->initmsg
.sinit_num_ostreams
= sinit
.sinit_num_ostreams
;
2612 if (sinit
.sinit_max_instreams
)
2613 sp
->initmsg
.sinit_max_instreams
= sinit
.sinit_max_instreams
;
2614 if (sinit
.sinit_max_attempts
)
2615 sp
->initmsg
.sinit_max_attempts
= sinit
.sinit_max_attempts
;
2616 if (sinit
.sinit_max_init_timeo
)
2617 sp
->initmsg
.sinit_max_init_timeo
= sinit
.sinit_max_init_timeo
;
2623 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2625 * Applications that wish to use the sendto() system call may wish to
2626 * specify a default set of parameters that would normally be supplied
2627 * through the inclusion of ancillary data. This socket option allows
2628 * such an application to set the default sctp_sndrcvinfo structure.
2629 * The application that wishes to use this socket option simply passes
2630 * in to this call the sctp_sndrcvinfo structure defined in Section
2631 * 5.2.2) The input parameters accepted by this call include
2632 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2633 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2634 * to this call if the caller is using the UDP model.
2636 static int sctp_setsockopt_default_send_param(struct sock
*sk
,
2637 char __user
*optval
,
2638 unsigned int optlen
)
2640 struct sctp_sndrcvinfo info
;
2641 struct sctp_association
*asoc
;
2642 struct sctp_sock
*sp
= sctp_sk(sk
);
2644 if (optlen
!= sizeof(struct sctp_sndrcvinfo
))
2646 if (copy_from_user(&info
, optval
, optlen
))
2649 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
2650 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
2654 asoc
->default_stream
= info
.sinfo_stream
;
2655 asoc
->default_flags
= info
.sinfo_flags
;
2656 asoc
->default_ppid
= info
.sinfo_ppid
;
2657 asoc
->default_context
= info
.sinfo_context
;
2658 asoc
->default_timetolive
= info
.sinfo_timetolive
;
2660 sp
->default_stream
= info
.sinfo_stream
;
2661 sp
->default_flags
= info
.sinfo_flags
;
2662 sp
->default_ppid
= info
.sinfo_ppid
;
2663 sp
->default_context
= info
.sinfo_context
;
2664 sp
->default_timetolive
= info
.sinfo_timetolive
;
2670 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2672 * Requests that the local SCTP stack use the enclosed peer address as
2673 * the association primary. The enclosed address must be one of the
2674 * association peer's addresses.
2676 static int sctp_setsockopt_primary_addr(struct sock
*sk
, char __user
*optval
,
2677 unsigned int optlen
)
2679 struct sctp_prim prim
;
2680 struct sctp_transport
*trans
;
2682 if (optlen
!= sizeof(struct sctp_prim
))
2685 if (copy_from_user(&prim
, optval
, sizeof(struct sctp_prim
)))
2688 trans
= sctp_addr_id2transport(sk
, &prim
.ssp_addr
, prim
.ssp_assoc_id
);
2692 sctp_assoc_set_primary(trans
->asoc
, trans
);
2698 * 7.1.5 SCTP_NODELAY
2700 * Turn on/off any Nagle-like algorithm. This means that packets are
2701 * generally sent as soon as possible and no unnecessary delays are
2702 * introduced, at the cost of more packets in the network. Expects an
2703 * integer boolean flag.
2705 static int sctp_setsockopt_nodelay(struct sock
*sk
, char __user
*optval
,
2706 unsigned int optlen
)
2710 if (optlen
< sizeof(int))
2712 if (get_user(val
, (int __user
*)optval
))
2715 sctp_sk(sk
)->nodelay
= (val
== 0) ? 0 : 1;
2721 * 7.1.1 SCTP_RTOINFO
2723 * The protocol parameters used to initialize and bound retransmission
2724 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2725 * and modify these parameters.
2726 * All parameters are time values, in milliseconds. A value of 0, when
2727 * modifying the parameters, indicates that the current value should not
2731 static int sctp_setsockopt_rtoinfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2733 struct sctp_rtoinfo rtoinfo
;
2734 struct sctp_association
*asoc
;
2736 if (optlen
!= sizeof (struct sctp_rtoinfo
))
2739 if (copy_from_user(&rtoinfo
, optval
, optlen
))
2742 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
2744 /* Set the values to the specific association */
2745 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
2749 if (rtoinfo
.srto_initial
!= 0)
2751 msecs_to_jiffies(rtoinfo
.srto_initial
);
2752 if (rtoinfo
.srto_max
!= 0)
2753 asoc
->rto_max
= msecs_to_jiffies(rtoinfo
.srto_max
);
2754 if (rtoinfo
.srto_min
!= 0)
2755 asoc
->rto_min
= msecs_to_jiffies(rtoinfo
.srto_min
);
2757 /* If there is no association or the association-id = 0
2758 * set the values to the endpoint.
2760 struct sctp_sock
*sp
= sctp_sk(sk
);
2762 if (rtoinfo
.srto_initial
!= 0)
2763 sp
->rtoinfo
.srto_initial
= rtoinfo
.srto_initial
;
2764 if (rtoinfo
.srto_max
!= 0)
2765 sp
->rtoinfo
.srto_max
= rtoinfo
.srto_max
;
2766 if (rtoinfo
.srto_min
!= 0)
2767 sp
->rtoinfo
.srto_min
= rtoinfo
.srto_min
;
2775 * 7.1.2 SCTP_ASSOCINFO
2777 * This option is used to tune the maximum retransmission attempts
2778 * of the association.
2779 * Returns an error if the new association retransmission value is
2780 * greater than the sum of the retransmission value of the peer.
2781 * See [SCTP] for more information.
2784 static int sctp_setsockopt_associnfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2787 struct sctp_assocparams assocparams
;
2788 struct sctp_association
*asoc
;
2790 if (optlen
!= sizeof(struct sctp_assocparams
))
2792 if (copy_from_user(&assocparams
, optval
, optlen
))
2795 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
2797 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
2800 /* Set the values to the specific association */
2802 if (assocparams
.sasoc_asocmaxrxt
!= 0) {
2805 struct sctp_transport
*peer_addr
;
2807 list_for_each_entry(peer_addr
, &asoc
->peer
.transport_addr_list
,
2809 path_sum
+= peer_addr
->pathmaxrxt
;
2813 /* Only validate asocmaxrxt if we have more than
2814 * one path/transport. We do this because path
2815 * retransmissions are only counted when we have more
2819 assocparams
.sasoc_asocmaxrxt
> path_sum
)
2822 asoc
->max_retrans
= assocparams
.sasoc_asocmaxrxt
;
2825 if (assocparams
.sasoc_cookie_life
!= 0) {
2826 asoc
->cookie_life
.tv_sec
=
2827 assocparams
.sasoc_cookie_life
/ 1000;
2828 asoc
->cookie_life
.tv_usec
=
2829 (assocparams
.sasoc_cookie_life
% 1000)
2833 /* Set the values to the endpoint */
2834 struct sctp_sock
*sp
= sctp_sk(sk
);
2836 if (assocparams
.sasoc_asocmaxrxt
!= 0)
2837 sp
->assocparams
.sasoc_asocmaxrxt
=
2838 assocparams
.sasoc_asocmaxrxt
;
2839 if (assocparams
.sasoc_cookie_life
!= 0)
2840 sp
->assocparams
.sasoc_cookie_life
=
2841 assocparams
.sasoc_cookie_life
;
2847 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2849 * This socket option is a boolean flag which turns on or off mapped V4
2850 * addresses. If this option is turned on and the socket is type
2851 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2852 * If this option is turned off, then no mapping will be done of V4
2853 * addresses and a user will receive both PF_INET6 and PF_INET type
2854 * addresses on the socket.
2856 static int sctp_setsockopt_mappedv4(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2859 struct sctp_sock
*sp
= sctp_sk(sk
);
2861 if (optlen
< sizeof(int))
2863 if (get_user(val
, (int __user
*)optval
))
2874 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2875 * This option will get or set the maximum size to put in any outgoing
2876 * SCTP DATA chunk. If a message is larger than this size it will be
2877 * fragmented by SCTP into the specified size. Note that the underlying
2878 * SCTP implementation may fragment into smaller sized chunks when the
2879 * PMTU of the underlying association is smaller than the value set by
2880 * the user. The default value for this option is '0' which indicates
2881 * the user is NOT limiting fragmentation and only the PMTU will effect
2882 * SCTP's choice of DATA chunk size. Note also that values set larger
2883 * than the maximum size of an IP datagram will effectively let SCTP
2884 * control fragmentation (i.e. the same as setting this option to 0).
2886 * The following structure is used to access and modify this parameter:
2888 * struct sctp_assoc_value {
2889 * sctp_assoc_t assoc_id;
2890 * uint32_t assoc_value;
2893 * assoc_id: This parameter is ignored for one-to-one style sockets.
2894 * For one-to-many style sockets this parameter indicates which
2895 * association the user is performing an action upon. Note that if
2896 * this field's value is zero then the endpoints default value is
2897 * changed (effecting future associations only).
2898 * assoc_value: This parameter specifies the maximum size in bytes.
2900 static int sctp_setsockopt_maxseg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2902 struct sctp_assoc_value params
;
2903 struct sctp_association
*asoc
;
2904 struct sctp_sock
*sp
= sctp_sk(sk
);
2907 if (optlen
== sizeof(int)) {
2908 pr_warn("Use of int in maxseg socket option deprecated\n");
2909 pr_warn("Use struct sctp_assoc_value instead\n");
2910 if (copy_from_user(&val
, optval
, optlen
))
2912 params
.assoc_id
= 0;
2913 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
2914 if (copy_from_user(¶ms
, optval
, optlen
))
2916 val
= params
.assoc_value
;
2920 if ((val
!= 0) && ((val
< 8) || (val
> SCTP_MAX_CHUNK_LEN
)))
2923 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
2924 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
2929 val
= asoc
->pathmtu
;
2930 val
-= sp
->pf
->af
->net_header_len
;
2931 val
-= sizeof(struct sctphdr
) +
2932 sizeof(struct sctp_data_chunk
);
2934 asoc
->user_frag
= val
;
2935 asoc
->frag_point
= sctp_frag_point(asoc
, asoc
->pathmtu
);
2937 sp
->user_frag
= val
;
2945 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2947 * Requests that the peer mark the enclosed address as the association
2948 * primary. The enclosed address must be one of the association's
2949 * locally bound addresses. The following structure is used to make a
2950 * set primary request:
2952 static int sctp_setsockopt_peer_primary_addr(struct sock
*sk
, char __user
*optval
,
2953 unsigned int optlen
)
2955 struct sctp_sock
*sp
;
2956 struct sctp_association
*asoc
= NULL
;
2957 struct sctp_setpeerprim prim
;
2958 struct sctp_chunk
*chunk
;
2964 if (!sctp_addip_enable
)
2967 if (optlen
!= sizeof(struct sctp_setpeerprim
))
2970 if (copy_from_user(&prim
, optval
, optlen
))
2973 asoc
= sctp_id2assoc(sk
, prim
.sspp_assoc_id
);
2977 if (!asoc
->peer
.asconf_capable
)
2980 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_SET_PRIMARY
)
2983 if (!sctp_state(asoc
, ESTABLISHED
))
2986 af
= sctp_get_af_specific(prim
.sspp_addr
.ss_family
);
2990 if (!af
->addr_valid((union sctp_addr
*)&prim
.sspp_addr
, sp
, NULL
))
2991 return -EADDRNOTAVAIL
;
2993 if (!sctp_assoc_lookup_laddr(asoc
, (union sctp_addr
*)&prim
.sspp_addr
))
2994 return -EADDRNOTAVAIL
;
2996 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2997 chunk
= sctp_make_asconf_set_prim(asoc
,
2998 (union sctp_addr
*)&prim
.sspp_addr
);
3002 err
= sctp_send_asconf(asoc
, chunk
);
3004 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
3009 static int sctp_setsockopt_adaptation_layer(struct sock
*sk
, char __user
*optval
,
3010 unsigned int optlen
)
3012 struct sctp_setadaptation adaptation
;
3014 if (optlen
!= sizeof(struct sctp_setadaptation
))
3016 if (copy_from_user(&adaptation
, optval
, optlen
))
3019 sctp_sk(sk
)->adaptation_ind
= adaptation
.ssb_adaptation_ind
;
3025 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3027 * The context field in the sctp_sndrcvinfo structure is normally only
3028 * used when a failed message is retrieved holding the value that was
3029 * sent down on the actual send call. This option allows the setting of
3030 * a default context on an association basis that will be received on
3031 * reading messages from the peer. This is especially helpful in the
3032 * one-2-many model for an application to keep some reference to an
3033 * internal state machine that is processing messages on the
3034 * association. Note that the setting of this value only effects
3035 * received messages from the peer and does not effect the value that is
3036 * saved with outbound messages.
3038 static int sctp_setsockopt_context(struct sock
*sk
, char __user
*optval
,
3039 unsigned int optlen
)
3041 struct sctp_assoc_value params
;
3042 struct sctp_sock
*sp
;
3043 struct sctp_association
*asoc
;
3045 if (optlen
!= sizeof(struct sctp_assoc_value
))
3047 if (copy_from_user(¶ms
, optval
, optlen
))
3052 if (params
.assoc_id
!= 0) {
3053 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3056 asoc
->default_rcv_context
= params
.assoc_value
;
3058 sp
->default_rcv_context
= params
.assoc_value
;
3065 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3067 * This options will at a minimum specify if the implementation is doing
3068 * fragmented interleave. Fragmented interleave, for a one to many
3069 * socket, is when subsequent calls to receive a message may return
3070 * parts of messages from different associations. Some implementations
3071 * may allow you to turn this value on or off. If so, when turned off,
3072 * no fragment interleave will occur (which will cause a head of line
3073 * blocking amongst multiple associations sharing the same one to many
3074 * socket). When this option is turned on, then each receive call may
3075 * come from a different association (thus the user must receive data
3076 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3077 * association each receive belongs to.
3079 * This option takes a boolean value. A non-zero value indicates that
3080 * fragmented interleave is on. A value of zero indicates that
3081 * fragmented interleave is off.
3083 * Note that it is important that an implementation that allows this
3084 * option to be turned on, have it off by default. Otherwise an unaware
3085 * application using the one to many model may become confused and act
3088 static int sctp_setsockopt_fragment_interleave(struct sock
*sk
,
3089 char __user
*optval
,
3090 unsigned int optlen
)
3094 if (optlen
!= sizeof(int))
3096 if (get_user(val
, (int __user
*)optval
))
3099 sctp_sk(sk
)->frag_interleave
= (val
== 0) ? 0 : 1;
3105 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3106 * (SCTP_PARTIAL_DELIVERY_POINT)
3108 * This option will set or get the SCTP partial delivery point. This
3109 * point is the size of a message where the partial delivery API will be
3110 * invoked to help free up rwnd space for the peer. Setting this to a
3111 * lower value will cause partial deliveries to happen more often. The
3112 * calls argument is an integer that sets or gets the partial delivery
3113 * point. Note also that the call will fail if the user attempts to set
3114 * this value larger than the socket receive buffer size.
3116 * Note that any single message having a length smaller than or equal to
3117 * the SCTP partial delivery point will be delivered in one single read
3118 * call as long as the user provided buffer is large enough to hold the
3121 static int sctp_setsockopt_partial_delivery_point(struct sock
*sk
,
3122 char __user
*optval
,
3123 unsigned int optlen
)
3127 if (optlen
!= sizeof(u32
))
3129 if (get_user(val
, (int __user
*)optval
))
3132 /* Note: We double the receive buffer from what the user sets
3133 * it to be, also initial rwnd is based on rcvbuf/2.
3135 if (val
> (sk
->sk_rcvbuf
>> 1))
3138 sctp_sk(sk
)->pd_point
= val
;
3140 return 0; /* is this the right error code? */
3144 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3146 * This option will allow a user to change the maximum burst of packets
3147 * that can be emitted by this association. Note that the default value
3148 * is 4, and some implementations may restrict this setting so that it
3149 * can only be lowered.
3151 * NOTE: This text doesn't seem right. Do this on a socket basis with
3152 * future associations inheriting the socket value.
3154 static int sctp_setsockopt_maxburst(struct sock
*sk
,
3155 char __user
*optval
,
3156 unsigned int optlen
)
3158 struct sctp_assoc_value params
;
3159 struct sctp_sock
*sp
;
3160 struct sctp_association
*asoc
;
3164 if (optlen
== sizeof(int)) {
3165 pr_warn("Use of int in max_burst socket option deprecated\n");
3166 pr_warn("Use struct sctp_assoc_value instead\n");
3167 if (copy_from_user(&val
, optval
, optlen
))
3169 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
3170 if (copy_from_user(¶ms
, optval
, optlen
))
3172 val
= params
.assoc_value
;
3173 assoc_id
= params
.assoc_id
;
3179 if (assoc_id
!= 0) {
3180 asoc
= sctp_id2assoc(sk
, assoc_id
);
3183 asoc
->max_burst
= val
;
3185 sp
->max_burst
= val
;
3191 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3193 * This set option adds a chunk type that the user is requesting to be
3194 * received only in an authenticated way. Changes to the list of chunks
3195 * will only effect future associations on the socket.
3197 static int sctp_setsockopt_auth_chunk(struct sock
*sk
,
3198 char __user
*optval
,
3199 unsigned int optlen
)
3201 struct sctp_authchunk val
;
3203 if (!sctp_auth_enable
)
3206 if (optlen
!= sizeof(struct sctp_authchunk
))
3208 if (copy_from_user(&val
, optval
, optlen
))
3211 switch (val
.sauth_chunk
) {
3213 case SCTP_CID_INIT_ACK
:
3214 case SCTP_CID_SHUTDOWN_COMPLETE
:
3219 /* add this chunk id to the endpoint */
3220 return sctp_auth_ep_add_chunkid(sctp_sk(sk
)->ep
, val
.sauth_chunk
);
3224 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3226 * This option gets or sets the list of HMAC algorithms that the local
3227 * endpoint requires the peer to use.
3229 static int sctp_setsockopt_hmac_ident(struct sock
*sk
,
3230 char __user
*optval
,
3231 unsigned int optlen
)
3233 struct sctp_hmacalgo
*hmacs
;
3237 if (!sctp_auth_enable
)
3240 if (optlen
< sizeof(struct sctp_hmacalgo
))
3243 hmacs
= memdup_user(optval
, optlen
);
3245 return PTR_ERR(hmacs
);
3247 idents
= hmacs
->shmac_num_idents
;
3248 if (idents
== 0 || idents
> SCTP_AUTH_NUM_HMACS
||
3249 (idents
* sizeof(u16
)) > (optlen
- sizeof(struct sctp_hmacalgo
))) {
3254 err
= sctp_auth_ep_set_hmacs(sctp_sk(sk
)->ep
, hmacs
);
3261 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3263 * This option will set a shared secret key which is used to build an
3264 * association shared key.
3266 static int sctp_setsockopt_auth_key(struct sock
*sk
,
3267 char __user
*optval
,
3268 unsigned int optlen
)
3270 struct sctp_authkey
*authkey
;
3271 struct sctp_association
*asoc
;
3274 if (!sctp_auth_enable
)
3277 if (optlen
<= sizeof(struct sctp_authkey
))
3280 authkey
= memdup_user(optval
, optlen
);
3281 if (IS_ERR(authkey
))
3282 return PTR_ERR(authkey
);
3284 if (authkey
->sca_keylength
> optlen
- sizeof(struct sctp_authkey
)) {
3289 asoc
= sctp_id2assoc(sk
, authkey
->sca_assoc_id
);
3290 if (!asoc
&& authkey
->sca_assoc_id
&& sctp_style(sk
, UDP
)) {
3295 ret
= sctp_auth_set_key(sctp_sk(sk
)->ep
, asoc
, authkey
);
3302 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3304 * This option will get or set the active shared key to be used to build
3305 * the association shared key.
3307 static int sctp_setsockopt_active_key(struct sock
*sk
,
3308 char __user
*optval
,
3309 unsigned int optlen
)
3311 struct sctp_authkeyid val
;
3312 struct sctp_association
*asoc
;
3314 if (!sctp_auth_enable
)
3317 if (optlen
!= sizeof(struct sctp_authkeyid
))
3319 if (copy_from_user(&val
, optval
, optlen
))
3322 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3323 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3326 return sctp_auth_set_active_key(sctp_sk(sk
)->ep
, asoc
,
3327 val
.scact_keynumber
);
3331 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3333 * This set option will delete a shared secret key from use.
3335 static int sctp_setsockopt_del_key(struct sock
*sk
,
3336 char __user
*optval
,
3337 unsigned int optlen
)
3339 struct sctp_authkeyid val
;
3340 struct sctp_association
*asoc
;
3342 if (!sctp_auth_enable
)
3345 if (optlen
!= sizeof(struct sctp_authkeyid
))
3347 if (copy_from_user(&val
, optval
, optlen
))
3350 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3351 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3354 return sctp_auth_del_key_id(sctp_sk(sk
)->ep
, asoc
,
3355 val
.scact_keynumber
);
3360 /* API 6.2 setsockopt(), getsockopt()
3362 * Applications use setsockopt() and getsockopt() to set or retrieve
3363 * socket options. Socket options are used to change the default
3364 * behavior of sockets calls. They are described in Section 7.
3368 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3369 * int __user *optlen);
3370 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3373 * sd - the socket descript.
3374 * level - set to IPPROTO_SCTP for all SCTP options.
3375 * optname - the option name.
3376 * optval - the buffer to store the value of the option.
3377 * optlen - the size of the buffer.
3379 SCTP_STATIC
int sctp_setsockopt(struct sock
*sk
, int level
, int optname
,
3380 char __user
*optval
, unsigned int optlen
)
3384 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3387 /* I can hardly begin to describe how wrong this is. This is
3388 * so broken as to be worse than useless. The API draft
3389 * REALLY is NOT helpful here... I am not convinced that the
3390 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3391 * are at all well-founded.
3393 if (level
!= SOL_SCTP
) {
3394 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
3395 retval
= af
->setsockopt(sk
, level
, optname
, optval
, optlen
);
3402 case SCTP_SOCKOPT_BINDX_ADD
:
3403 /* 'optlen' is the size of the addresses buffer. */
3404 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3405 optlen
, SCTP_BINDX_ADD_ADDR
);
3408 case SCTP_SOCKOPT_BINDX_REM
:
3409 /* 'optlen' is the size of the addresses buffer. */
3410 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3411 optlen
, SCTP_BINDX_REM_ADDR
);
3414 case SCTP_SOCKOPT_CONNECTX_OLD
:
3415 /* 'optlen' is the size of the addresses buffer. */
3416 retval
= sctp_setsockopt_connectx_old(sk
,
3417 (struct sockaddr __user
*)optval
,
3421 case SCTP_SOCKOPT_CONNECTX
:
3422 /* 'optlen' is the size of the addresses buffer. */
3423 retval
= sctp_setsockopt_connectx(sk
,
3424 (struct sockaddr __user
*)optval
,
3428 case SCTP_DISABLE_FRAGMENTS
:
3429 retval
= sctp_setsockopt_disable_fragments(sk
, optval
, optlen
);
3433 retval
= sctp_setsockopt_events(sk
, optval
, optlen
);
3436 case SCTP_AUTOCLOSE
:
3437 retval
= sctp_setsockopt_autoclose(sk
, optval
, optlen
);
3440 case SCTP_PEER_ADDR_PARAMS
:
3441 retval
= sctp_setsockopt_peer_addr_params(sk
, optval
, optlen
);
3444 case SCTP_DELAYED_SACK
:
3445 retval
= sctp_setsockopt_delayed_ack(sk
, optval
, optlen
);
3447 case SCTP_PARTIAL_DELIVERY_POINT
:
3448 retval
= sctp_setsockopt_partial_delivery_point(sk
, optval
, optlen
);
3452 retval
= sctp_setsockopt_initmsg(sk
, optval
, optlen
);
3454 case SCTP_DEFAULT_SEND_PARAM
:
3455 retval
= sctp_setsockopt_default_send_param(sk
, optval
,
3458 case SCTP_PRIMARY_ADDR
:
3459 retval
= sctp_setsockopt_primary_addr(sk
, optval
, optlen
);
3461 case SCTP_SET_PEER_PRIMARY_ADDR
:
3462 retval
= sctp_setsockopt_peer_primary_addr(sk
, optval
, optlen
);
3465 retval
= sctp_setsockopt_nodelay(sk
, optval
, optlen
);
3468 retval
= sctp_setsockopt_rtoinfo(sk
, optval
, optlen
);
3470 case SCTP_ASSOCINFO
:
3471 retval
= sctp_setsockopt_associnfo(sk
, optval
, optlen
);
3473 case SCTP_I_WANT_MAPPED_V4_ADDR
:
3474 retval
= sctp_setsockopt_mappedv4(sk
, optval
, optlen
);
3477 retval
= sctp_setsockopt_maxseg(sk
, optval
, optlen
);
3479 case SCTP_ADAPTATION_LAYER
:
3480 retval
= sctp_setsockopt_adaptation_layer(sk
, optval
, optlen
);
3483 retval
= sctp_setsockopt_context(sk
, optval
, optlen
);
3485 case SCTP_FRAGMENT_INTERLEAVE
:
3486 retval
= sctp_setsockopt_fragment_interleave(sk
, optval
, optlen
);
3488 case SCTP_MAX_BURST
:
3489 retval
= sctp_setsockopt_maxburst(sk
, optval
, optlen
);
3491 case SCTP_AUTH_CHUNK
:
3492 retval
= sctp_setsockopt_auth_chunk(sk
, optval
, optlen
);
3494 case SCTP_HMAC_IDENT
:
3495 retval
= sctp_setsockopt_hmac_ident(sk
, optval
, optlen
);
3498 retval
= sctp_setsockopt_auth_key(sk
, optval
, optlen
);
3500 case SCTP_AUTH_ACTIVE_KEY
:
3501 retval
= sctp_setsockopt_active_key(sk
, optval
, optlen
);
3503 case SCTP_AUTH_DELETE_KEY
:
3504 retval
= sctp_setsockopt_del_key(sk
, optval
, optlen
);
3507 retval
= -ENOPROTOOPT
;
3511 sctp_release_sock(sk
);
3517 /* API 3.1.6 connect() - UDP Style Syntax
3519 * An application may use the connect() call in the UDP model to initiate an
3520 * association without sending data.
3524 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3526 * sd: the socket descriptor to have a new association added to.
3528 * nam: the address structure (either struct sockaddr_in or struct
3529 * sockaddr_in6 defined in RFC2553 [7]).
3531 * len: the size of the address.
3533 SCTP_STATIC
int sctp_connect(struct sock
*sk
, struct sockaddr
*addr
,
3541 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3542 __func__
, sk
, addr
, addr_len
);
3544 /* Validate addr_len before calling common connect/connectx routine. */
3545 af
= sctp_get_af_specific(addr
->sa_family
);
3546 if (!af
|| addr_len
< af
->sockaddr_len
) {
3549 /* Pass correct addr len to common routine (so it knows there
3550 * is only one address being passed.
3552 err
= __sctp_connect(sk
, addr
, af
->sockaddr_len
, NULL
);
3555 sctp_release_sock(sk
);
3559 /* FIXME: Write comments. */
3560 SCTP_STATIC
int sctp_disconnect(struct sock
*sk
, int flags
)
3562 return -EOPNOTSUPP
; /* STUB */
3565 /* 4.1.4 accept() - TCP Style Syntax
3567 * Applications use accept() call to remove an established SCTP
3568 * association from the accept queue of the endpoint. A new socket
3569 * descriptor will be returned from accept() to represent the newly
3570 * formed association.
3572 SCTP_STATIC
struct sock
*sctp_accept(struct sock
*sk
, int flags
, int *err
)
3574 struct sctp_sock
*sp
;
3575 struct sctp_endpoint
*ep
;
3576 struct sock
*newsk
= NULL
;
3577 struct sctp_association
*asoc
;
3586 if (!sctp_style(sk
, TCP
)) {
3587 error
= -EOPNOTSUPP
;
3591 if (!sctp_sstate(sk
, LISTENING
)) {
3596 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
3598 error
= sctp_wait_for_accept(sk
, timeo
);
3602 /* We treat the list of associations on the endpoint as the accept
3603 * queue and pick the first association on the list.
3605 asoc
= list_entry(ep
->asocs
.next
, struct sctp_association
, asocs
);
3607 newsk
= sp
->pf
->create_accept_sk(sk
, asoc
);
3613 /* Populate the fields of the newsk from the oldsk and migrate the
3614 * asoc to the newsk.
3616 sctp_sock_migrate(sk
, newsk
, asoc
, SCTP_SOCKET_TCP
);
3619 sctp_release_sock(sk
);
3624 /* The SCTP ioctl handler. */
3625 SCTP_STATIC
int sctp_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
3632 * SEQPACKET-style sockets in LISTENING state are valid, for
3633 * SCTP, so only discard TCP-style sockets in LISTENING state.
3635 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
3640 struct sk_buff
*skb
;
3641 unsigned int amount
= 0;
3643 skb
= skb_peek(&sk
->sk_receive_queue
);
3646 * We will only return the amount of this packet since
3647 * that is all that will be read.
3651 rc
= put_user(amount
, (int __user
*)arg
);
3659 sctp_release_sock(sk
);
3663 /* This is the function which gets called during socket creation to
3664 * initialized the SCTP-specific portion of the sock.
3665 * The sock structure should already be zero-filled memory.
3667 SCTP_STATIC
int sctp_init_sock(struct sock
*sk
)
3669 struct sctp_endpoint
*ep
;
3670 struct sctp_sock
*sp
;
3672 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk
);
3676 /* Initialize the SCTP per socket area. */
3677 switch (sk
->sk_type
) {
3678 case SOCK_SEQPACKET
:
3679 sp
->type
= SCTP_SOCKET_UDP
;
3682 sp
->type
= SCTP_SOCKET_TCP
;
3685 return -ESOCKTNOSUPPORT
;
3688 /* Initialize default send parameters. These parameters can be
3689 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3691 sp
->default_stream
= 0;
3692 sp
->default_ppid
= 0;
3693 sp
->default_flags
= 0;
3694 sp
->default_context
= 0;
3695 sp
->default_timetolive
= 0;
3697 sp
->default_rcv_context
= 0;
3698 sp
->max_burst
= sctp_max_burst
;
3700 /* Initialize default setup parameters. These parameters
3701 * can be modified with the SCTP_INITMSG socket option or
3702 * overridden by the SCTP_INIT CMSG.
3704 sp
->initmsg
.sinit_num_ostreams
= sctp_max_outstreams
;
3705 sp
->initmsg
.sinit_max_instreams
= sctp_max_instreams
;
3706 sp
->initmsg
.sinit_max_attempts
= sctp_max_retrans_init
;
3707 sp
->initmsg
.sinit_max_init_timeo
= sctp_rto_max
;
3709 /* Initialize default RTO related parameters. These parameters can
3710 * be modified for with the SCTP_RTOINFO socket option.
3712 sp
->rtoinfo
.srto_initial
= sctp_rto_initial
;
3713 sp
->rtoinfo
.srto_max
= sctp_rto_max
;
3714 sp
->rtoinfo
.srto_min
= sctp_rto_min
;
3716 /* Initialize default association related parameters. These parameters
3717 * can be modified with the SCTP_ASSOCINFO socket option.
3719 sp
->assocparams
.sasoc_asocmaxrxt
= sctp_max_retrans_association
;
3720 sp
->assocparams
.sasoc_number_peer_destinations
= 0;
3721 sp
->assocparams
.sasoc_peer_rwnd
= 0;
3722 sp
->assocparams
.sasoc_local_rwnd
= 0;
3723 sp
->assocparams
.sasoc_cookie_life
= sctp_valid_cookie_life
;
3725 /* Initialize default event subscriptions. By default, all the
3728 memset(&sp
->subscribe
, 0, sizeof(struct sctp_event_subscribe
));
3730 /* Default Peer Address Parameters. These defaults can
3731 * be modified via SCTP_PEER_ADDR_PARAMS
3733 sp
->hbinterval
= sctp_hb_interval
;
3734 sp
->pathmaxrxt
= sctp_max_retrans_path
;
3735 sp
->pathmtu
= 0; // allow default discovery
3736 sp
->sackdelay
= sctp_sack_timeout
;
3738 sp
->param_flags
= SPP_HB_ENABLE
|
3740 SPP_SACKDELAY_ENABLE
;
3742 /* If enabled no SCTP message fragmentation will be performed.
3743 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3745 sp
->disable_fragments
= 0;
3747 /* Enable Nagle algorithm by default. */
3750 /* Enable by default. */
3753 /* Auto-close idle associations after the configured
3754 * number of seconds. A value of 0 disables this
3755 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3756 * for UDP-style sockets only.
3760 /* User specified fragmentation limit. */
3763 sp
->adaptation_ind
= 0;
3765 sp
->pf
= sctp_get_pf_specific(sk
->sk_family
);
3767 /* Control variables for partial data delivery. */
3768 atomic_set(&sp
->pd_mode
, 0);
3769 skb_queue_head_init(&sp
->pd_lobby
);
3770 sp
->frag_interleave
= 0;
3772 /* Create a per socket endpoint structure. Even if we
3773 * change the data structure relationships, this may still
3774 * be useful for storing pre-connect address information.
3776 ep
= sctp_endpoint_new(sk
, GFP_KERNEL
);
3783 SCTP_DBG_OBJCNT_INC(sock
);
3786 percpu_counter_inc(&sctp_sockets_allocated
);
3787 sock_prot_inuse_add(sock_net(sk
), sk
->sk_prot
, 1);
3788 if (sctp_default_auto_asconf
) {
3789 list_add_tail(&sp
->auto_asconf_list
,
3790 &sctp_auto_asconf_splist
);
3791 sp
->do_auto_asconf
= 1;
3793 sp
->do_auto_asconf
= 0;
3799 /* Cleanup any SCTP per socket resources. */
3800 SCTP_STATIC
void sctp_destroy_sock(struct sock
*sk
)
3802 struct sctp_sock
*sp
;
3804 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk
);
3806 /* Release our hold on the endpoint. */
3808 if (sp
->do_auto_asconf
) {
3809 sp
->do_auto_asconf
= 0;
3810 list_del(&sp
->auto_asconf_list
);
3812 sctp_endpoint_free(sp
->ep
);
3814 percpu_counter_dec(&sctp_sockets_allocated
);
3815 sock_prot_inuse_add(sock_net(sk
), sk
->sk_prot
, -1);
3819 /* API 4.1.7 shutdown() - TCP Style Syntax
3820 * int shutdown(int socket, int how);
3822 * sd - the socket descriptor of the association to be closed.
3823 * how - Specifies the type of shutdown. The values are
3826 * Disables further receive operations. No SCTP
3827 * protocol action is taken.
3829 * Disables further send operations, and initiates
3830 * the SCTP shutdown sequence.
3832 * Disables further send and receive operations
3833 * and initiates the SCTP shutdown sequence.
3835 SCTP_STATIC
void sctp_shutdown(struct sock
*sk
, int how
)
3837 struct sctp_endpoint
*ep
;
3838 struct sctp_association
*asoc
;
3840 if (!sctp_style(sk
, TCP
))
3843 if (how
& SEND_SHUTDOWN
) {
3844 ep
= sctp_sk(sk
)->ep
;
3845 if (!list_empty(&ep
->asocs
)) {
3846 asoc
= list_entry(ep
->asocs
.next
,
3847 struct sctp_association
, asocs
);
3848 sctp_primitive_SHUTDOWN(asoc
, NULL
);
3853 /* 7.2.1 Association Status (SCTP_STATUS)
3855 * Applications can retrieve current status information about an
3856 * association, including association state, peer receiver window size,
3857 * number of unacked data chunks, and number of data chunks pending
3858 * receipt. This information is read-only.
3860 static int sctp_getsockopt_sctp_status(struct sock
*sk
, int len
,
3861 char __user
*optval
,
3864 struct sctp_status status
;
3865 struct sctp_association
*asoc
= NULL
;
3866 struct sctp_transport
*transport
;
3867 sctp_assoc_t associd
;
3870 if (len
< sizeof(status
)) {
3875 len
= sizeof(status
);
3876 if (copy_from_user(&status
, optval
, len
)) {
3881 associd
= status
.sstat_assoc_id
;
3882 asoc
= sctp_id2assoc(sk
, associd
);
3888 transport
= asoc
->peer
.primary_path
;
3890 status
.sstat_assoc_id
= sctp_assoc2id(asoc
);
3891 status
.sstat_state
= asoc
->state
;
3892 status
.sstat_rwnd
= asoc
->peer
.rwnd
;
3893 status
.sstat_unackdata
= asoc
->unack_data
;
3895 status
.sstat_penddata
= sctp_tsnmap_pending(&asoc
->peer
.tsn_map
);
3896 status
.sstat_instrms
= asoc
->c
.sinit_max_instreams
;
3897 status
.sstat_outstrms
= asoc
->c
.sinit_num_ostreams
;
3898 status
.sstat_fragmentation_point
= asoc
->frag_point
;
3899 status
.sstat_primary
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
3900 memcpy(&status
.sstat_primary
.spinfo_address
, &transport
->ipaddr
,
3901 transport
->af_specific
->sockaddr_len
);
3902 /* Map ipv4 address into v4-mapped-on-v6 address. */
3903 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
3904 (union sctp_addr
*)&status
.sstat_primary
.spinfo_address
);
3905 status
.sstat_primary
.spinfo_state
= transport
->state
;
3906 status
.sstat_primary
.spinfo_cwnd
= transport
->cwnd
;
3907 status
.sstat_primary
.spinfo_srtt
= transport
->srtt
;
3908 status
.sstat_primary
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
3909 status
.sstat_primary
.spinfo_mtu
= transport
->pathmtu
;
3911 if (status
.sstat_primary
.spinfo_state
== SCTP_UNKNOWN
)
3912 status
.sstat_primary
.spinfo_state
= SCTP_ACTIVE
;
3914 if (put_user(len
, optlen
)) {
3919 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3920 len
, status
.sstat_state
, status
.sstat_rwnd
,
3921 status
.sstat_assoc_id
);
3923 if (copy_to_user(optval
, &status
, len
)) {
3933 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3935 * Applications can retrieve information about a specific peer address
3936 * of an association, including its reachability state, congestion
3937 * window, and retransmission timer values. This information is
3940 static int sctp_getsockopt_peer_addr_info(struct sock
*sk
, int len
,
3941 char __user
*optval
,
3944 struct sctp_paddrinfo pinfo
;
3945 struct sctp_transport
*transport
;
3948 if (len
< sizeof(pinfo
)) {
3953 len
= sizeof(pinfo
);
3954 if (copy_from_user(&pinfo
, optval
, len
)) {
3959 transport
= sctp_addr_id2transport(sk
, &pinfo
.spinfo_address
,
3960 pinfo
.spinfo_assoc_id
);
3964 pinfo
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
3965 pinfo
.spinfo_state
= transport
->state
;
3966 pinfo
.spinfo_cwnd
= transport
->cwnd
;
3967 pinfo
.spinfo_srtt
= transport
->srtt
;
3968 pinfo
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
3969 pinfo
.spinfo_mtu
= transport
->pathmtu
;
3971 if (pinfo
.spinfo_state
== SCTP_UNKNOWN
)
3972 pinfo
.spinfo_state
= SCTP_ACTIVE
;
3974 if (put_user(len
, optlen
)) {
3979 if (copy_to_user(optval
, &pinfo
, len
)) {
3988 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3990 * This option is a on/off flag. If enabled no SCTP message
3991 * fragmentation will be performed. Instead if a message being sent
3992 * exceeds the current PMTU size, the message will NOT be sent and
3993 * instead a error will be indicated to the user.
3995 static int sctp_getsockopt_disable_fragments(struct sock
*sk
, int len
,
3996 char __user
*optval
, int __user
*optlen
)
4000 if (len
< sizeof(int))
4004 val
= (sctp_sk(sk
)->disable_fragments
== 1);
4005 if (put_user(len
, optlen
))
4007 if (copy_to_user(optval
, &val
, len
))
4012 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4014 * This socket option is used to specify various notifications and
4015 * ancillary data the user wishes to receive.
4017 static int sctp_getsockopt_events(struct sock
*sk
, int len
, char __user
*optval
,
4020 if (len
< sizeof(struct sctp_event_subscribe
))
4022 len
= sizeof(struct sctp_event_subscribe
);
4023 if (put_user(len
, optlen
))
4025 if (copy_to_user(optval
, &sctp_sk(sk
)->subscribe
, len
))
4030 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4032 * This socket option is applicable to the UDP-style socket only. When
4033 * set it will cause associations that are idle for more than the
4034 * specified number of seconds to automatically close. An association
4035 * being idle is defined an association that has NOT sent or received
4036 * user data. The special value of '0' indicates that no automatic
4037 * close of any associations should be performed. The option expects an
4038 * integer defining the number of seconds of idle time before an
4039 * association is closed.
4041 static int sctp_getsockopt_autoclose(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4043 /* Applicable to UDP-style socket only */
4044 if (sctp_style(sk
, TCP
))
4046 if (len
< sizeof(int))
4049 if (put_user(len
, optlen
))
4051 if (copy_to_user(optval
, &sctp_sk(sk
)->autoclose
, sizeof(int)))
4056 /* Helper routine to branch off an association to a new socket. */
4057 SCTP_STATIC
int sctp_do_peeloff(struct sctp_association
*asoc
,
4058 struct socket
**sockp
)
4060 struct sock
*sk
= asoc
->base
.sk
;
4061 struct socket
*sock
;
4065 /* An association cannot be branched off from an already peeled-off
4066 * socket, nor is this supported for tcp style sockets.
4068 if (!sctp_style(sk
, UDP
))
4071 /* Create a new socket. */
4072 err
= sock_create(sk
->sk_family
, SOCK_SEQPACKET
, IPPROTO_SCTP
, &sock
);
4076 sctp_copy_sock(sock
->sk
, sk
, asoc
);
4078 /* Make peeled-off sockets more like 1-1 accepted sockets.
4079 * Set the daddr and initialize id to something more random
4081 af
= sctp_get_af_specific(asoc
->peer
.primary_addr
.sa
.sa_family
);
4082 af
->to_sk_daddr(&asoc
->peer
.primary_addr
, sk
);
4084 /* Populate the fields of the newsk from the oldsk and migrate the
4085 * asoc to the newsk.
4087 sctp_sock_migrate(sk
, sock
->sk
, asoc
, SCTP_SOCKET_UDP_HIGH_BANDWIDTH
);
4094 static int sctp_getsockopt_peeloff(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4096 sctp_peeloff_arg_t peeloff
;
4097 struct socket
*newsock
;
4099 struct sctp_association
*asoc
;
4101 if (len
< sizeof(sctp_peeloff_arg_t
))
4103 len
= sizeof(sctp_peeloff_arg_t
);
4104 if (copy_from_user(&peeloff
, optval
, len
))
4107 asoc
= sctp_id2assoc(sk
, peeloff
.associd
);
4113 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__
, sk
, asoc
);
4115 retval
= sctp_do_peeloff(asoc
, &newsock
);
4119 /* Map the socket to an unused fd that can be returned to the user. */
4120 retval
= sock_map_fd(newsock
, 0);
4122 sock_release(newsock
);
4126 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
4127 __func__
, sk
, asoc
, newsock
->sk
, retval
);
4129 /* Return the fd mapped to the new socket. */
4130 peeloff
.sd
= retval
;
4131 if (put_user(len
, optlen
))
4133 if (copy_to_user(optval
, &peeloff
, len
))
4140 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4142 * Applications can enable or disable heartbeats for any peer address of
4143 * an association, modify an address's heartbeat interval, force a
4144 * heartbeat to be sent immediately, and adjust the address's maximum
4145 * number of retransmissions sent before an address is considered
4146 * unreachable. The following structure is used to access and modify an
4147 * address's parameters:
4149 * struct sctp_paddrparams {
4150 * sctp_assoc_t spp_assoc_id;
4151 * struct sockaddr_storage spp_address;
4152 * uint32_t spp_hbinterval;
4153 * uint16_t spp_pathmaxrxt;
4154 * uint32_t spp_pathmtu;
4155 * uint32_t spp_sackdelay;
4156 * uint32_t spp_flags;
4159 * spp_assoc_id - (one-to-many style socket) This is filled in the
4160 * application, and identifies the association for
4162 * spp_address - This specifies which address is of interest.
4163 * spp_hbinterval - This contains the value of the heartbeat interval,
4164 * in milliseconds. If a value of zero
4165 * is present in this field then no changes are to
4166 * be made to this parameter.
4167 * spp_pathmaxrxt - This contains the maximum number of
4168 * retransmissions before this address shall be
4169 * considered unreachable. If a value of zero
4170 * is present in this field then no changes are to
4171 * be made to this parameter.
4172 * spp_pathmtu - When Path MTU discovery is disabled the value
4173 * specified here will be the "fixed" path mtu.
4174 * Note that if the spp_address field is empty
4175 * then all associations on this address will
4176 * have this fixed path mtu set upon them.
4178 * spp_sackdelay - When delayed sack is enabled, this value specifies
4179 * the number of milliseconds that sacks will be delayed
4180 * for. This value will apply to all addresses of an
4181 * association if the spp_address field is empty. Note
4182 * also, that if delayed sack is enabled and this
4183 * value is set to 0, no change is made to the last
4184 * recorded delayed sack timer value.
4186 * spp_flags - These flags are used to control various features
4187 * on an association. The flag field may contain
4188 * zero or more of the following options.
4190 * SPP_HB_ENABLE - Enable heartbeats on the
4191 * specified address. Note that if the address
4192 * field is empty all addresses for the association
4193 * have heartbeats enabled upon them.
4195 * SPP_HB_DISABLE - Disable heartbeats on the
4196 * speicifed address. Note that if the address
4197 * field is empty all addresses for the association
4198 * will have their heartbeats disabled. Note also
4199 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4200 * mutually exclusive, only one of these two should
4201 * be specified. Enabling both fields will have
4202 * undetermined results.
4204 * SPP_HB_DEMAND - Request a user initiated heartbeat
4205 * to be made immediately.
4207 * SPP_PMTUD_ENABLE - This field will enable PMTU
4208 * discovery upon the specified address. Note that
4209 * if the address feild is empty then all addresses
4210 * on the association are effected.
4212 * SPP_PMTUD_DISABLE - This field will disable PMTU
4213 * discovery upon the specified address. Note that
4214 * if the address feild is empty then all addresses
4215 * on the association are effected. Not also that
4216 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4217 * exclusive. Enabling both will have undetermined
4220 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4221 * on delayed sack. The time specified in spp_sackdelay
4222 * is used to specify the sack delay for this address. Note
4223 * that if spp_address is empty then all addresses will
4224 * enable delayed sack and take on the sack delay
4225 * value specified in spp_sackdelay.
4226 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4227 * off delayed sack. If the spp_address field is blank then
4228 * delayed sack is disabled for the entire association. Note
4229 * also that this field is mutually exclusive to
4230 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4233 static int sctp_getsockopt_peer_addr_params(struct sock
*sk
, int len
,
4234 char __user
*optval
, int __user
*optlen
)
4236 struct sctp_paddrparams params
;
4237 struct sctp_transport
*trans
= NULL
;
4238 struct sctp_association
*asoc
= NULL
;
4239 struct sctp_sock
*sp
= sctp_sk(sk
);
4241 if (len
< sizeof(struct sctp_paddrparams
))
4243 len
= sizeof(struct sctp_paddrparams
);
4244 if (copy_from_user(¶ms
, optval
, len
))
4247 /* If an address other than INADDR_ANY is specified, and
4248 * no transport is found, then the request is invalid.
4250 if (!sctp_is_any(sk
, ( union sctp_addr
*)¶ms
.spp_address
)) {
4251 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
4252 params
.spp_assoc_id
);
4254 SCTP_DEBUG_PRINTK("Failed no transport\n");
4259 /* Get association, if assoc_id != 0 and the socket is a one
4260 * to many style socket, and an association was not found, then
4261 * the id was invalid.
4263 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
4264 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
)) {
4265 SCTP_DEBUG_PRINTK("Failed no association\n");
4270 /* Fetch transport values. */
4271 params
.spp_hbinterval
= jiffies_to_msecs(trans
->hbinterval
);
4272 params
.spp_pathmtu
= trans
->pathmtu
;
4273 params
.spp_pathmaxrxt
= trans
->pathmaxrxt
;
4274 params
.spp_sackdelay
= jiffies_to_msecs(trans
->sackdelay
);
4276 /*draft-11 doesn't say what to return in spp_flags*/
4277 params
.spp_flags
= trans
->param_flags
;
4279 /* Fetch association values. */
4280 params
.spp_hbinterval
= jiffies_to_msecs(asoc
->hbinterval
);
4281 params
.spp_pathmtu
= asoc
->pathmtu
;
4282 params
.spp_pathmaxrxt
= asoc
->pathmaxrxt
;
4283 params
.spp_sackdelay
= jiffies_to_msecs(asoc
->sackdelay
);
4285 /*draft-11 doesn't say what to return in spp_flags*/
4286 params
.spp_flags
= asoc
->param_flags
;
4288 /* Fetch socket values. */
4289 params
.spp_hbinterval
= sp
->hbinterval
;
4290 params
.spp_pathmtu
= sp
->pathmtu
;
4291 params
.spp_sackdelay
= sp
->sackdelay
;
4292 params
.spp_pathmaxrxt
= sp
->pathmaxrxt
;
4294 /*draft-11 doesn't say what to return in spp_flags*/
4295 params
.spp_flags
= sp
->param_flags
;
4298 if (copy_to_user(optval
, ¶ms
, len
))
4301 if (put_user(len
, optlen
))
4308 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4310 * This option will effect the way delayed acks are performed. This
4311 * option allows you to get or set the delayed ack time, in
4312 * milliseconds. It also allows changing the delayed ack frequency.
4313 * Changing the frequency to 1 disables the delayed sack algorithm. If
4314 * the assoc_id is 0, then this sets or gets the endpoints default
4315 * values. If the assoc_id field is non-zero, then the set or get
4316 * effects the specified association for the one to many model (the
4317 * assoc_id field is ignored by the one to one model). Note that if
4318 * sack_delay or sack_freq are 0 when setting this option, then the
4319 * current values will remain unchanged.
4321 * struct sctp_sack_info {
4322 * sctp_assoc_t sack_assoc_id;
4323 * uint32_t sack_delay;
4324 * uint32_t sack_freq;
4327 * sack_assoc_id - This parameter, indicates which association the user
4328 * is performing an action upon. Note that if this field's value is
4329 * zero then the endpoints default value is changed (effecting future
4330 * associations only).
4332 * sack_delay - This parameter contains the number of milliseconds that
4333 * the user is requesting the delayed ACK timer be set to. Note that
4334 * this value is defined in the standard to be between 200 and 500
4337 * sack_freq - This parameter contains the number of packets that must
4338 * be received before a sack is sent without waiting for the delay
4339 * timer to expire. The default value for this is 2, setting this
4340 * value to 1 will disable the delayed sack algorithm.
4342 static int sctp_getsockopt_delayed_ack(struct sock
*sk
, int len
,
4343 char __user
*optval
,
4346 struct sctp_sack_info params
;
4347 struct sctp_association
*asoc
= NULL
;
4348 struct sctp_sock
*sp
= sctp_sk(sk
);
4350 if (len
>= sizeof(struct sctp_sack_info
)) {
4351 len
= sizeof(struct sctp_sack_info
);
4353 if (copy_from_user(¶ms
, optval
, len
))
4355 } else if (len
== sizeof(struct sctp_assoc_value
)) {
4356 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4357 pr_warn("Use struct sctp_sack_info instead\n");
4358 if (copy_from_user(¶ms
, optval
, len
))
4363 /* Get association, if sack_assoc_id != 0 and the socket is a one
4364 * to many style socket, and an association was not found, then
4365 * the id was invalid.
4367 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
4368 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
4372 /* Fetch association values. */
4373 if (asoc
->param_flags
& SPP_SACKDELAY_ENABLE
) {
4374 params
.sack_delay
= jiffies_to_msecs(
4376 params
.sack_freq
= asoc
->sackfreq
;
4379 params
.sack_delay
= 0;
4380 params
.sack_freq
= 1;
4383 /* Fetch socket values. */
4384 if (sp
->param_flags
& SPP_SACKDELAY_ENABLE
) {
4385 params
.sack_delay
= sp
->sackdelay
;
4386 params
.sack_freq
= sp
->sackfreq
;
4388 params
.sack_delay
= 0;
4389 params
.sack_freq
= 1;
4393 if (copy_to_user(optval
, ¶ms
, len
))
4396 if (put_user(len
, optlen
))
4402 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4404 * Applications can specify protocol parameters for the default association
4405 * initialization. The option name argument to setsockopt() and getsockopt()
4408 * Setting initialization parameters is effective only on an unconnected
4409 * socket (for UDP-style sockets only future associations are effected
4410 * by the change). With TCP-style sockets, this option is inherited by
4411 * sockets derived from a listener socket.
4413 static int sctp_getsockopt_initmsg(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4415 if (len
< sizeof(struct sctp_initmsg
))
4417 len
= sizeof(struct sctp_initmsg
);
4418 if (put_user(len
, optlen
))
4420 if (copy_to_user(optval
, &sctp_sk(sk
)->initmsg
, len
))
4426 static int sctp_getsockopt_peer_addrs(struct sock
*sk
, int len
,
4427 char __user
*optval
, int __user
*optlen
)
4429 struct sctp_association
*asoc
;
4431 struct sctp_getaddrs getaddrs
;
4432 struct sctp_transport
*from
;
4434 union sctp_addr temp
;
4435 struct sctp_sock
*sp
= sctp_sk(sk
);
4440 if (len
< sizeof(struct sctp_getaddrs
))
4443 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
4446 /* For UDP-style sockets, id specifies the association to query. */
4447 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
4451 to
= optval
+ offsetof(struct sctp_getaddrs
,addrs
);
4452 space_left
= len
- offsetof(struct sctp_getaddrs
,addrs
);
4454 list_for_each_entry(from
, &asoc
->peer
.transport_addr_list
,
4456 memcpy(&temp
, &from
->ipaddr
, sizeof(temp
));
4457 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
4458 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4459 if (space_left
< addrlen
)
4461 if (copy_to_user(to
, &temp
, addrlen
))
4465 space_left
-= addrlen
;
4468 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
))
4470 bytes_copied
= ((char __user
*)to
) - optval
;
4471 if (put_user(bytes_copied
, optlen
))
4477 static int sctp_copy_laddrs(struct sock
*sk
, __u16 port
, void *to
,
4478 size_t space_left
, int *bytes_copied
)
4480 struct sctp_sockaddr_entry
*addr
;
4481 union sctp_addr temp
;
4486 list_for_each_entry_rcu(addr
, &sctp_local_addr_list
, list
) {
4490 if ((PF_INET
== sk
->sk_family
) &&
4491 (AF_INET6
== addr
->a
.sa
.sa_family
))
4493 if ((PF_INET6
== sk
->sk_family
) &&
4494 inet_v6_ipv6only(sk
) &&
4495 (AF_INET
== addr
->a
.sa
.sa_family
))
4497 memcpy(&temp
, &addr
->a
, sizeof(temp
));
4498 if (!temp
.v4
.sin_port
)
4499 temp
.v4
.sin_port
= htons(port
);
4501 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
4503 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4504 if (space_left
< addrlen
) {
4508 memcpy(to
, &temp
, addrlen
);
4512 space_left
-= addrlen
;
4513 *bytes_copied
+= addrlen
;
4521 static int sctp_getsockopt_local_addrs(struct sock
*sk
, int len
,
4522 char __user
*optval
, int __user
*optlen
)
4524 struct sctp_bind_addr
*bp
;
4525 struct sctp_association
*asoc
;
4527 struct sctp_getaddrs getaddrs
;
4528 struct sctp_sockaddr_entry
*addr
;
4530 union sctp_addr temp
;
4531 struct sctp_sock
*sp
= sctp_sk(sk
);
4535 int bytes_copied
= 0;
4539 if (len
< sizeof(struct sctp_getaddrs
))
4542 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
4546 * For UDP-style sockets, id specifies the association to query.
4547 * If the id field is set to the value '0' then the locally bound
4548 * addresses are returned without regard to any particular
4551 if (0 == getaddrs
.assoc_id
) {
4552 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
4554 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
4557 bp
= &asoc
->base
.bind_addr
;
4560 to
= optval
+ offsetof(struct sctp_getaddrs
,addrs
);
4561 space_left
= len
- offsetof(struct sctp_getaddrs
,addrs
);
4563 addrs
= kmalloc(space_left
, GFP_KERNEL
);
4567 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4568 * addresses from the global local address list.
4570 if (sctp_list_single_entry(&bp
->address_list
)) {
4571 addr
= list_entry(bp
->address_list
.next
,
4572 struct sctp_sockaddr_entry
, list
);
4573 if (sctp_is_any(sk
, &addr
->a
)) {
4574 cnt
= sctp_copy_laddrs(sk
, bp
->port
, addrs
,
4575 space_left
, &bytes_copied
);
4585 /* Protection on the bound address list is not needed since
4586 * in the socket option context we hold a socket lock and
4587 * thus the bound address list can't change.
4589 list_for_each_entry(addr
, &bp
->address_list
, list
) {
4590 memcpy(&temp
, &addr
->a
, sizeof(temp
));
4591 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
4592 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4593 if (space_left
< addrlen
) {
4594 err
= -ENOMEM
; /*fixme: right error?*/
4597 memcpy(buf
, &temp
, addrlen
);
4599 bytes_copied
+= addrlen
;
4601 space_left
-= addrlen
;
4605 if (copy_to_user(to
, addrs
, bytes_copied
)) {
4609 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
)) {
4613 if (put_user(bytes_copied
, optlen
))
4620 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4622 * Requests that the local SCTP stack use the enclosed peer address as
4623 * the association primary. The enclosed address must be one of the
4624 * association peer's addresses.
4626 static int sctp_getsockopt_primary_addr(struct sock
*sk
, int len
,
4627 char __user
*optval
, int __user
*optlen
)
4629 struct sctp_prim prim
;
4630 struct sctp_association
*asoc
;
4631 struct sctp_sock
*sp
= sctp_sk(sk
);
4633 if (len
< sizeof(struct sctp_prim
))
4636 len
= sizeof(struct sctp_prim
);
4638 if (copy_from_user(&prim
, optval
, len
))
4641 asoc
= sctp_id2assoc(sk
, prim
.ssp_assoc_id
);
4645 if (!asoc
->peer
.primary_path
)
4648 memcpy(&prim
.ssp_addr
, &asoc
->peer
.primary_path
->ipaddr
,
4649 asoc
->peer
.primary_path
->af_specific
->sockaddr_len
);
4651 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
,
4652 (union sctp_addr
*)&prim
.ssp_addr
);
4654 if (put_user(len
, optlen
))
4656 if (copy_to_user(optval
, &prim
, len
))
4663 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4665 * Requests that the local endpoint set the specified Adaptation Layer
4666 * Indication parameter for all future INIT and INIT-ACK exchanges.
4668 static int sctp_getsockopt_adaptation_layer(struct sock
*sk
, int len
,
4669 char __user
*optval
, int __user
*optlen
)
4671 struct sctp_setadaptation adaptation
;
4673 if (len
< sizeof(struct sctp_setadaptation
))
4676 len
= sizeof(struct sctp_setadaptation
);
4678 adaptation
.ssb_adaptation_ind
= sctp_sk(sk
)->adaptation_ind
;
4680 if (put_user(len
, optlen
))
4682 if (copy_to_user(optval
, &adaptation
, len
))
4690 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4692 * Applications that wish to use the sendto() system call may wish to
4693 * specify a default set of parameters that would normally be supplied
4694 * through the inclusion of ancillary data. This socket option allows
4695 * such an application to set the default sctp_sndrcvinfo structure.
4698 * The application that wishes to use this socket option simply passes
4699 * in to this call the sctp_sndrcvinfo structure defined in Section
4700 * 5.2.2) The input parameters accepted by this call include
4701 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4702 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4703 * to this call if the caller is using the UDP model.
4705 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4707 static int sctp_getsockopt_default_send_param(struct sock
*sk
,
4708 int len
, char __user
*optval
,
4711 struct sctp_sndrcvinfo info
;
4712 struct sctp_association
*asoc
;
4713 struct sctp_sock
*sp
= sctp_sk(sk
);
4715 if (len
< sizeof(struct sctp_sndrcvinfo
))
4718 len
= sizeof(struct sctp_sndrcvinfo
);
4720 if (copy_from_user(&info
, optval
, len
))
4723 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
4724 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
4728 info
.sinfo_stream
= asoc
->default_stream
;
4729 info
.sinfo_flags
= asoc
->default_flags
;
4730 info
.sinfo_ppid
= asoc
->default_ppid
;
4731 info
.sinfo_context
= asoc
->default_context
;
4732 info
.sinfo_timetolive
= asoc
->default_timetolive
;
4734 info
.sinfo_stream
= sp
->default_stream
;
4735 info
.sinfo_flags
= sp
->default_flags
;
4736 info
.sinfo_ppid
= sp
->default_ppid
;
4737 info
.sinfo_context
= sp
->default_context
;
4738 info
.sinfo_timetolive
= sp
->default_timetolive
;
4741 if (put_user(len
, optlen
))
4743 if (copy_to_user(optval
, &info
, len
))
4751 * 7.1.5 SCTP_NODELAY
4753 * Turn on/off any Nagle-like algorithm. This means that packets are
4754 * generally sent as soon as possible and no unnecessary delays are
4755 * introduced, at the cost of more packets in the network. Expects an
4756 * integer boolean flag.
4759 static int sctp_getsockopt_nodelay(struct sock
*sk
, int len
,
4760 char __user
*optval
, int __user
*optlen
)
4764 if (len
< sizeof(int))
4768 val
= (sctp_sk(sk
)->nodelay
== 1);
4769 if (put_user(len
, optlen
))
4771 if (copy_to_user(optval
, &val
, len
))
4778 * 7.1.1 SCTP_RTOINFO
4780 * The protocol parameters used to initialize and bound retransmission
4781 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4782 * and modify these parameters.
4783 * All parameters are time values, in milliseconds. A value of 0, when
4784 * modifying the parameters, indicates that the current value should not
4788 static int sctp_getsockopt_rtoinfo(struct sock
*sk
, int len
,
4789 char __user
*optval
,
4790 int __user
*optlen
) {
4791 struct sctp_rtoinfo rtoinfo
;
4792 struct sctp_association
*asoc
;
4794 if (len
< sizeof (struct sctp_rtoinfo
))
4797 len
= sizeof(struct sctp_rtoinfo
);
4799 if (copy_from_user(&rtoinfo
, optval
, len
))
4802 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
4804 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
4807 /* Values corresponding to the specific association. */
4809 rtoinfo
.srto_initial
= jiffies_to_msecs(asoc
->rto_initial
);
4810 rtoinfo
.srto_max
= jiffies_to_msecs(asoc
->rto_max
);
4811 rtoinfo
.srto_min
= jiffies_to_msecs(asoc
->rto_min
);
4813 /* Values corresponding to the endpoint. */
4814 struct sctp_sock
*sp
= sctp_sk(sk
);
4816 rtoinfo
.srto_initial
= sp
->rtoinfo
.srto_initial
;
4817 rtoinfo
.srto_max
= sp
->rtoinfo
.srto_max
;
4818 rtoinfo
.srto_min
= sp
->rtoinfo
.srto_min
;
4821 if (put_user(len
, optlen
))
4824 if (copy_to_user(optval
, &rtoinfo
, len
))
4832 * 7.1.2 SCTP_ASSOCINFO
4834 * This option is used to tune the maximum retransmission attempts
4835 * of the association.
4836 * Returns an error if the new association retransmission value is
4837 * greater than the sum of the retransmission value of the peer.
4838 * See [SCTP] for more information.
4841 static int sctp_getsockopt_associnfo(struct sock
*sk
, int len
,
4842 char __user
*optval
,
4846 struct sctp_assocparams assocparams
;
4847 struct sctp_association
*asoc
;
4848 struct list_head
*pos
;
4851 if (len
< sizeof (struct sctp_assocparams
))
4854 len
= sizeof(struct sctp_assocparams
);
4856 if (copy_from_user(&assocparams
, optval
, len
))
4859 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
4861 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
4864 /* Values correspoinding to the specific association */
4866 assocparams
.sasoc_asocmaxrxt
= asoc
->max_retrans
;
4867 assocparams
.sasoc_peer_rwnd
= asoc
->peer
.rwnd
;
4868 assocparams
.sasoc_local_rwnd
= asoc
->a_rwnd
;
4869 assocparams
.sasoc_cookie_life
= (asoc
->cookie_life
.tv_sec
4871 (asoc
->cookie_life
.tv_usec
4874 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
4878 assocparams
.sasoc_number_peer_destinations
= cnt
;
4880 /* Values corresponding to the endpoint */
4881 struct sctp_sock
*sp
= sctp_sk(sk
);
4883 assocparams
.sasoc_asocmaxrxt
= sp
->assocparams
.sasoc_asocmaxrxt
;
4884 assocparams
.sasoc_peer_rwnd
= sp
->assocparams
.sasoc_peer_rwnd
;
4885 assocparams
.sasoc_local_rwnd
= sp
->assocparams
.sasoc_local_rwnd
;
4886 assocparams
.sasoc_cookie_life
=
4887 sp
->assocparams
.sasoc_cookie_life
;
4888 assocparams
.sasoc_number_peer_destinations
=
4890 sasoc_number_peer_destinations
;
4893 if (put_user(len
, optlen
))
4896 if (copy_to_user(optval
, &assocparams
, len
))
4903 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4905 * This socket option is a boolean flag which turns on or off mapped V4
4906 * addresses. If this option is turned on and the socket is type
4907 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4908 * If this option is turned off, then no mapping will be done of V4
4909 * addresses and a user will receive both PF_INET6 and PF_INET type
4910 * addresses on the socket.
4912 static int sctp_getsockopt_mappedv4(struct sock
*sk
, int len
,
4913 char __user
*optval
, int __user
*optlen
)
4916 struct sctp_sock
*sp
= sctp_sk(sk
);
4918 if (len
< sizeof(int))
4923 if (put_user(len
, optlen
))
4925 if (copy_to_user(optval
, &val
, len
))
4932 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
4933 * (chapter and verse is quoted at sctp_setsockopt_context())
4935 static int sctp_getsockopt_context(struct sock
*sk
, int len
,
4936 char __user
*optval
, int __user
*optlen
)
4938 struct sctp_assoc_value params
;
4939 struct sctp_sock
*sp
;
4940 struct sctp_association
*asoc
;
4942 if (len
< sizeof(struct sctp_assoc_value
))
4945 len
= sizeof(struct sctp_assoc_value
);
4947 if (copy_from_user(¶ms
, optval
, len
))
4952 if (params
.assoc_id
!= 0) {
4953 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
4956 params
.assoc_value
= asoc
->default_rcv_context
;
4958 params
.assoc_value
= sp
->default_rcv_context
;
4961 if (put_user(len
, optlen
))
4963 if (copy_to_user(optval
, ¶ms
, len
))
4970 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
4971 * This option will get or set the maximum size to put in any outgoing
4972 * SCTP DATA chunk. If a message is larger than this size it will be
4973 * fragmented by SCTP into the specified size. Note that the underlying
4974 * SCTP implementation may fragment into smaller sized chunks when the
4975 * PMTU of the underlying association is smaller than the value set by
4976 * the user. The default value for this option is '0' which indicates
4977 * the user is NOT limiting fragmentation and only the PMTU will effect
4978 * SCTP's choice of DATA chunk size. Note also that values set larger
4979 * than the maximum size of an IP datagram will effectively let SCTP
4980 * control fragmentation (i.e. the same as setting this option to 0).
4982 * The following structure is used to access and modify this parameter:
4984 * struct sctp_assoc_value {
4985 * sctp_assoc_t assoc_id;
4986 * uint32_t assoc_value;
4989 * assoc_id: This parameter is ignored for one-to-one style sockets.
4990 * For one-to-many style sockets this parameter indicates which
4991 * association the user is performing an action upon. Note that if
4992 * this field's value is zero then the endpoints default value is
4993 * changed (effecting future associations only).
4994 * assoc_value: This parameter specifies the maximum size in bytes.
4996 static int sctp_getsockopt_maxseg(struct sock
*sk
, int len
,
4997 char __user
*optval
, int __user
*optlen
)
4999 struct sctp_assoc_value params
;
5000 struct sctp_association
*asoc
;
5002 if (len
== sizeof(int)) {
5003 pr_warn("Use of int in maxseg socket option deprecated\n");
5004 pr_warn("Use struct sctp_assoc_value instead\n");
5005 params
.assoc_id
= 0;
5006 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5007 len
= sizeof(struct sctp_assoc_value
);
5008 if (copy_from_user(¶ms
, optval
, sizeof(params
)))
5013 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5014 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
5018 params
.assoc_value
= asoc
->frag_point
;
5020 params
.assoc_value
= sctp_sk(sk
)->user_frag
;
5022 if (put_user(len
, optlen
))
5024 if (len
== sizeof(int)) {
5025 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5028 if (copy_to_user(optval
, ¶ms
, len
))
5036 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5037 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5039 static int sctp_getsockopt_fragment_interleave(struct sock
*sk
, int len
,
5040 char __user
*optval
, int __user
*optlen
)
5044 if (len
< sizeof(int))
5049 val
= sctp_sk(sk
)->frag_interleave
;
5050 if (put_user(len
, optlen
))
5052 if (copy_to_user(optval
, &val
, len
))
5059 * 7.1.25. Set or Get the sctp partial delivery point
5060 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5062 static int sctp_getsockopt_partial_delivery_point(struct sock
*sk
, int len
,
5063 char __user
*optval
,
5068 if (len
< sizeof(u32
))
5073 val
= sctp_sk(sk
)->pd_point
;
5074 if (put_user(len
, optlen
))
5076 if (copy_to_user(optval
, &val
, len
))
5083 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5084 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5086 static int sctp_getsockopt_maxburst(struct sock
*sk
, int len
,
5087 char __user
*optval
,
5090 struct sctp_assoc_value params
;
5091 struct sctp_sock
*sp
;
5092 struct sctp_association
*asoc
;
5094 if (len
== sizeof(int)) {
5095 pr_warn("Use of int in max_burst socket option deprecated\n");
5096 pr_warn("Use struct sctp_assoc_value instead\n");
5097 params
.assoc_id
= 0;
5098 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5099 len
= sizeof(struct sctp_assoc_value
);
5100 if (copy_from_user(¶ms
, optval
, len
))
5107 if (params
.assoc_id
!= 0) {
5108 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5111 params
.assoc_value
= asoc
->max_burst
;
5113 params
.assoc_value
= sp
->max_burst
;
5115 if (len
== sizeof(int)) {
5116 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5119 if (copy_to_user(optval
, ¶ms
, len
))
5127 static int sctp_getsockopt_hmac_ident(struct sock
*sk
, int len
,
5128 char __user
*optval
, int __user
*optlen
)
5130 struct sctp_hmacalgo __user
*p
= (void __user
*)optval
;
5131 struct sctp_hmac_algo_param
*hmacs
;
5135 if (!sctp_auth_enable
)
5138 hmacs
= sctp_sk(sk
)->ep
->auth_hmacs_list
;
5139 data_len
= ntohs(hmacs
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5141 if (len
< sizeof(struct sctp_hmacalgo
) + data_len
)
5144 len
= sizeof(struct sctp_hmacalgo
) + data_len
;
5145 num_idents
= data_len
/ sizeof(u16
);
5147 if (put_user(len
, optlen
))
5149 if (put_user(num_idents
, &p
->shmac_num_idents
))
5151 if (copy_to_user(p
->shmac_idents
, hmacs
->hmac_ids
, data_len
))
5156 static int sctp_getsockopt_active_key(struct sock
*sk
, int len
,
5157 char __user
*optval
, int __user
*optlen
)
5159 struct sctp_authkeyid val
;
5160 struct sctp_association
*asoc
;
5162 if (!sctp_auth_enable
)
5165 if (len
< sizeof(struct sctp_authkeyid
))
5167 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authkeyid
)))
5170 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
5171 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
5175 val
.scact_keynumber
= asoc
->active_key_id
;
5177 val
.scact_keynumber
= sctp_sk(sk
)->ep
->active_key_id
;
5179 len
= sizeof(struct sctp_authkeyid
);
5180 if (put_user(len
, optlen
))
5182 if (copy_to_user(optval
, &val
, len
))
5188 static int sctp_getsockopt_peer_auth_chunks(struct sock
*sk
, int len
,
5189 char __user
*optval
, int __user
*optlen
)
5191 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5192 struct sctp_authchunks val
;
5193 struct sctp_association
*asoc
;
5194 struct sctp_chunks_param
*ch
;
5198 if (!sctp_auth_enable
)
5201 if (len
< sizeof(struct sctp_authchunks
))
5204 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5207 to
= p
->gauth_chunks
;
5208 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5212 ch
= asoc
->peer
.peer_chunks
;
5216 /* See if the user provided enough room for all the data */
5217 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5218 if (len
< num_chunks
)
5221 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
5224 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
5225 if (put_user(len
, optlen
)) return -EFAULT
;
5226 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
5231 static int sctp_getsockopt_local_auth_chunks(struct sock
*sk
, int len
,
5232 char __user
*optval
, int __user
*optlen
)
5234 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5235 struct sctp_authchunks val
;
5236 struct sctp_association
*asoc
;
5237 struct sctp_chunks_param
*ch
;
5241 if (!sctp_auth_enable
)
5244 if (len
< sizeof(struct sctp_authchunks
))
5247 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5250 to
= p
->gauth_chunks
;
5251 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5252 if (!asoc
&& val
.gauth_assoc_id
&& sctp_style(sk
, UDP
))
5256 ch
= (struct sctp_chunks_param
*)asoc
->c
.auth_chunks
;
5258 ch
= sctp_sk(sk
)->ep
->auth_chunk_list
;
5263 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5264 if (len
< sizeof(struct sctp_authchunks
) + num_chunks
)
5267 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
5270 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
5271 if (put_user(len
, optlen
))
5273 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
5280 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5281 * This option gets the current number of associations that are attached
5282 * to a one-to-many style socket. The option value is an uint32_t.
5284 static int sctp_getsockopt_assoc_number(struct sock
*sk
, int len
,
5285 char __user
*optval
, int __user
*optlen
)
5287 struct sctp_sock
*sp
= sctp_sk(sk
);
5288 struct sctp_association
*asoc
;
5291 if (sctp_style(sk
, TCP
))
5294 if (len
< sizeof(u32
))
5299 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5303 if (put_user(len
, optlen
))
5305 if (copy_to_user(optval
, &val
, len
))
5312 * 8.2.6. Get the Current Identifiers of Associations
5313 * (SCTP_GET_ASSOC_ID_LIST)
5315 * This option gets the current list of SCTP association identifiers of
5316 * the SCTP associations handled by a one-to-many style socket.
5318 static int sctp_getsockopt_assoc_ids(struct sock
*sk
, int len
,
5319 char __user
*optval
, int __user
*optlen
)
5321 struct sctp_sock
*sp
= sctp_sk(sk
);
5322 struct sctp_association
*asoc
;
5323 struct sctp_assoc_ids
*ids
;
5326 if (sctp_style(sk
, TCP
))
5329 if (len
< sizeof(struct sctp_assoc_ids
))
5332 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5336 if (len
< sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
)
5339 len
= sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
;
5341 ids
= kmalloc(len
, GFP_KERNEL
);
5345 ids
->gaids_number_of_ids
= num
;
5347 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5348 ids
->gaids_assoc_id
[num
++] = asoc
->assoc_id
;
5351 if (put_user(len
, optlen
) || copy_to_user(optval
, ids
, len
)) {
5360 SCTP_STATIC
int sctp_getsockopt(struct sock
*sk
, int level
, int optname
,
5361 char __user
*optval
, int __user
*optlen
)
5366 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5369 /* I can hardly begin to describe how wrong this is. This is
5370 * so broken as to be worse than useless. The API draft
5371 * REALLY is NOT helpful here... I am not convinced that the
5372 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5373 * are at all well-founded.
5375 if (level
!= SOL_SCTP
) {
5376 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
5378 retval
= af
->getsockopt(sk
, level
, optname
, optval
, optlen
);
5382 if (get_user(len
, optlen
))
5389 retval
= sctp_getsockopt_sctp_status(sk
, len
, optval
, optlen
);
5391 case SCTP_DISABLE_FRAGMENTS
:
5392 retval
= sctp_getsockopt_disable_fragments(sk
, len
, optval
,
5396 retval
= sctp_getsockopt_events(sk
, len
, optval
, optlen
);
5398 case SCTP_AUTOCLOSE
:
5399 retval
= sctp_getsockopt_autoclose(sk
, len
, optval
, optlen
);
5401 case SCTP_SOCKOPT_PEELOFF
:
5402 retval
= sctp_getsockopt_peeloff(sk
, len
, optval
, optlen
);
5404 case SCTP_PEER_ADDR_PARAMS
:
5405 retval
= sctp_getsockopt_peer_addr_params(sk
, len
, optval
,
5408 case SCTP_DELAYED_SACK
:
5409 retval
= sctp_getsockopt_delayed_ack(sk
, len
, optval
,
5413 retval
= sctp_getsockopt_initmsg(sk
, len
, optval
, optlen
);
5415 case SCTP_GET_PEER_ADDRS
:
5416 retval
= sctp_getsockopt_peer_addrs(sk
, len
, optval
,
5419 case SCTP_GET_LOCAL_ADDRS
:
5420 retval
= sctp_getsockopt_local_addrs(sk
, len
, optval
,
5423 case SCTP_SOCKOPT_CONNECTX3
:
5424 retval
= sctp_getsockopt_connectx3(sk
, len
, optval
, optlen
);
5426 case SCTP_DEFAULT_SEND_PARAM
:
5427 retval
= sctp_getsockopt_default_send_param(sk
, len
,
5430 case SCTP_PRIMARY_ADDR
:
5431 retval
= sctp_getsockopt_primary_addr(sk
, len
, optval
, optlen
);
5434 retval
= sctp_getsockopt_nodelay(sk
, len
, optval
, optlen
);
5437 retval
= sctp_getsockopt_rtoinfo(sk
, len
, optval
, optlen
);
5439 case SCTP_ASSOCINFO
:
5440 retval
= sctp_getsockopt_associnfo(sk
, len
, optval
, optlen
);
5442 case SCTP_I_WANT_MAPPED_V4_ADDR
:
5443 retval
= sctp_getsockopt_mappedv4(sk
, len
, optval
, optlen
);
5446 retval
= sctp_getsockopt_maxseg(sk
, len
, optval
, optlen
);
5448 case SCTP_GET_PEER_ADDR_INFO
:
5449 retval
= sctp_getsockopt_peer_addr_info(sk
, len
, optval
,
5452 case SCTP_ADAPTATION_LAYER
:
5453 retval
= sctp_getsockopt_adaptation_layer(sk
, len
, optval
,
5457 retval
= sctp_getsockopt_context(sk
, len
, optval
, optlen
);
5459 case SCTP_FRAGMENT_INTERLEAVE
:
5460 retval
= sctp_getsockopt_fragment_interleave(sk
, len
, optval
,
5463 case SCTP_PARTIAL_DELIVERY_POINT
:
5464 retval
= sctp_getsockopt_partial_delivery_point(sk
, len
, optval
,
5467 case SCTP_MAX_BURST
:
5468 retval
= sctp_getsockopt_maxburst(sk
, len
, optval
, optlen
);
5471 case SCTP_AUTH_CHUNK
:
5472 case SCTP_AUTH_DELETE_KEY
:
5473 retval
= -EOPNOTSUPP
;
5475 case SCTP_HMAC_IDENT
:
5476 retval
= sctp_getsockopt_hmac_ident(sk
, len
, optval
, optlen
);
5478 case SCTP_AUTH_ACTIVE_KEY
:
5479 retval
= sctp_getsockopt_active_key(sk
, len
, optval
, optlen
);
5481 case SCTP_PEER_AUTH_CHUNKS
:
5482 retval
= sctp_getsockopt_peer_auth_chunks(sk
, len
, optval
,
5485 case SCTP_LOCAL_AUTH_CHUNKS
:
5486 retval
= sctp_getsockopt_local_auth_chunks(sk
, len
, optval
,
5489 case SCTP_GET_ASSOC_NUMBER
:
5490 retval
= sctp_getsockopt_assoc_number(sk
, len
, optval
, optlen
);
5492 case SCTP_GET_ASSOC_ID_LIST
:
5493 retval
= sctp_getsockopt_assoc_ids(sk
, len
, optval
, optlen
);
5496 retval
= -ENOPROTOOPT
;
5500 sctp_release_sock(sk
);
5504 static void sctp_hash(struct sock
*sk
)
5509 static void sctp_unhash(struct sock
*sk
)
5514 /* Check if port is acceptable. Possibly find first available port.
5516 * The port hash table (contained in the 'global' SCTP protocol storage
5517 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5518 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5519 * list (the list number is the port number hashed out, so as you
5520 * would expect from a hash function, all the ports in a given list have
5521 * such a number that hashes out to the same list number; you were
5522 * expecting that, right?); so each list has a set of ports, with a
5523 * link to the socket (struct sock) that uses it, the port number and
5524 * a fastreuse flag (FIXME: NPI ipg).
5526 static struct sctp_bind_bucket
*sctp_bucket_create(
5527 struct sctp_bind_hashbucket
*head
, unsigned short snum
);
5529 static long sctp_get_port_local(struct sock
*sk
, union sctp_addr
*addr
)
5531 struct sctp_bind_hashbucket
*head
; /* hash list */
5532 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
5533 struct hlist_node
*node
;
5534 unsigned short snum
;
5537 snum
= ntohs(addr
->v4
.sin_port
);
5539 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum
);
5540 sctp_local_bh_disable();
5543 /* Search for an available port. */
5544 int low
, high
, remaining
, index
;
5547 inet_get_local_port_range(&low
, &high
);
5548 remaining
= (high
- low
) + 1;
5549 rover
= net_random() % remaining
+ low
;
5553 if ((rover
< low
) || (rover
> high
))
5555 if (inet_is_reserved_local_port(rover
))
5557 index
= sctp_phashfn(rover
);
5558 head
= &sctp_port_hashtable
[index
];
5559 sctp_spin_lock(&head
->lock
);
5560 sctp_for_each_hentry(pp
, node
, &head
->chain
)
5561 if (pp
->port
== rover
)
5565 sctp_spin_unlock(&head
->lock
);
5566 } while (--remaining
> 0);
5568 /* Exhausted local port range during search? */
5573 /* OK, here is the one we will use. HEAD (the port
5574 * hash table list entry) is non-NULL and we hold it's
5579 /* We are given an specific port number; we verify
5580 * that it is not being used. If it is used, we will
5581 * exahust the search in the hash list corresponding
5582 * to the port number (snum) - we detect that with the
5583 * port iterator, pp being NULL.
5585 head
= &sctp_port_hashtable
[sctp_phashfn(snum
)];
5586 sctp_spin_lock(&head
->lock
);
5587 sctp_for_each_hentry(pp
, node
, &head
->chain
) {
5588 if (pp
->port
== snum
)
5595 if (!hlist_empty(&pp
->owner
)) {
5596 /* We had a port hash table hit - there is an
5597 * available port (pp != NULL) and it is being
5598 * used by other socket (pp->owner not empty); that other
5599 * socket is going to be sk2.
5601 int reuse
= sk
->sk_reuse
;
5604 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5605 if (pp
->fastreuse
&& sk
->sk_reuse
&&
5606 sk
->sk_state
!= SCTP_SS_LISTENING
)
5609 /* Run through the list of sockets bound to the port
5610 * (pp->port) [via the pointers bind_next and
5611 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5612 * we get the endpoint they describe and run through
5613 * the endpoint's list of IP (v4 or v6) addresses,
5614 * comparing each of the addresses with the address of
5615 * the socket sk. If we find a match, then that means
5616 * that this port/socket (sk) combination are already
5619 sk_for_each_bound(sk2
, node
, &pp
->owner
) {
5620 struct sctp_endpoint
*ep2
;
5621 ep2
= sctp_sk(sk2
)->ep
;
5624 (reuse
&& sk2
->sk_reuse
&&
5625 sk2
->sk_state
!= SCTP_SS_LISTENING
))
5628 if (sctp_bind_addr_conflict(&ep2
->base
.bind_addr
, addr
,
5629 sctp_sk(sk2
), sctp_sk(sk
))) {
5634 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5637 /* If there was a hash table miss, create a new port. */
5639 if (!pp
&& !(pp
= sctp_bucket_create(head
, snum
)))
5642 /* In either case (hit or miss), make sure fastreuse is 1 only
5643 * if sk->sk_reuse is too (that is, if the caller requested
5644 * SO_REUSEADDR on this socket -sk-).
5646 if (hlist_empty(&pp
->owner
)) {
5647 if (sk
->sk_reuse
&& sk
->sk_state
!= SCTP_SS_LISTENING
)
5651 } else if (pp
->fastreuse
&&
5652 (!sk
->sk_reuse
|| sk
->sk_state
== SCTP_SS_LISTENING
))
5655 /* We are set, so fill up all the data in the hash table
5656 * entry, tie the socket list information with the rest of the
5657 * sockets FIXME: Blurry, NPI (ipg).
5660 if (!sctp_sk(sk
)->bind_hash
) {
5661 inet_sk(sk
)->inet_num
= snum
;
5662 sk_add_bind_node(sk
, &pp
->owner
);
5663 sctp_sk(sk
)->bind_hash
= pp
;
5668 sctp_spin_unlock(&head
->lock
);
5671 sctp_local_bh_enable();
5675 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5676 * port is requested.
5678 static int sctp_get_port(struct sock
*sk
, unsigned short snum
)
5681 union sctp_addr addr
;
5682 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
5684 /* Set up a dummy address struct from the sk. */
5685 af
->from_sk(&addr
, sk
);
5686 addr
.v4
.sin_port
= htons(snum
);
5688 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5689 ret
= sctp_get_port_local(sk
, &addr
);
5695 * Move a socket to LISTENING state.
5697 SCTP_STATIC
int sctp_listen_start(struct sock
*sk
, int backlog
)
5699 struct sctp_sock
*sp
= sctp_sk(sk
);
5700 struct sctp_endpoint
*ep
= sp
->ep
;
5701 struct crypto_hash
*tfm
= NULL
;
5703 /* Allocate HMAC for generating cookie. */
5704 if (!sctp_sk(sk
)->hmac
&& sctp_hmac_alg
) {
5705 tfm
= crypto_alloc_hash(sctp_hmac_alg
, 0, CRYPTO_ALG_ASYNC
);
5707 if (net_ratelimit()) {
5708 pr_info("failed to load transform for %s: %ld\n",
5709 sctp_hmac_alg
, PTR_ERR(tfm
));
5713 sctp_sk(sk
)->hmac
= tfm
;
5717 * If a bind() or sctp_bindx() is not called prior to a listen()
5718 * call that allows new associations to be accepted, the system
5719 * picks an ephemeral port and will choose an address set equivalent
5720 * to binding with a wildcard address.
5722 * This is not currently spelled out in the SCTP sockets
5723 * extensions draft, but follows the practice as seen in TCP
5727 sk
->sk_state
= SCTP_SS_LISTENING
;
5728 if (!ep
->base
.bind_addr
.port
) {
5729 if (sctp_autobind(sk
))
5732 if (sctp_get_port(sk
, inet_sk(sk
)->inet_num
)) {
5733 sk
->sk_state
= SCTP_SS_CLOSED
;
5738 sk
->sk_max_ack_backlog
= backlog
;
5739 sctp_hash_endpoint(ep
);
5744 * 4.1.3 / 5.1.3 listen()
5746 * By default, new associations are not accepted for UDP style sockets.
5747 * An application uses listen() to mark a socket as being able to
5748 * accept new associations.
5750 * On TCP style sockets, applications use listen() to ready the SCTP
5751 * endpoint for accepting inbound associations.
5753 * On both types of endpoints a backlog of '0' disables listening.
5755 * Move a socket to LISTENING state.
5757 int sctp_inet_listen(struct socket
*sock
, int backlog
)
5759 struct sock
*sk
= sock
->sk
;
5760 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5763 if (unlikely(backlog
< 0))
5768 /* Peeled-off sockets are not allowed to listen(). */
5769 if (sctp_style(sk
, UDP_HIGH_BANDWIDTH
))
5772 if (sock
->state
!= SS_UNCONNECTED
)
5775 /* If backlog is zero, disable listening. */
5777 if (sctp_sstate(sk
, CLOSED
))
5781 sctp_unhash_endpoint(ep
);
5782 sk
->sk_state
= SCTP_SS_CLOSED
;
5784 sctp_sk(sk
)->bind_hash
->fastreuse
= 1;
5788 /* If we are already listening, just update the backlog */
5789 if (sctp_sstate(sk
, LISTENING
))
5790 sk
->sk_max_ack_backlog
= backlog
;
5792 err
= sctp_listen_start(sk
, backlog
);
5799 sctp_release_sock(sk
);
5804 * This function is done by modeling the current datagram_poll() and the
5805 * tcp_poll(). Note that, based on these implementations, we don't
5806 * lock the socket in this function, even though it seems that,
5807 * ideally, locking or some other mechanisms can be used to ensure
5808 * the integrity of the counters (sndbuf and wmem_alloc) used
5809 * in this place. We assume that we don't need locks either until proven
5812 * Another thing to note is that we include the Async I/O support
5813 * here, again, by modeling the current TCP/UDP code. We don't have
5814 * a good way to test with it yet.
5816 unsigned int sctp_poll(struct file
*file
, struct socket
*sock
, poll_table
*wait
)
5818 struct sock
*sk
= sock
->sk
;
5819 struct sctp_sock
*sp
= sctp_sk(sk
);
5822 poll_wait(file
, sk_sleep(sk
), wait
);
5824 /* A TCP-style listening socket becomes readable when the accept queue
5827 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
5828 return (!list_empty(&sp
->ep
->asocs
)) ?
5829 (POLLIN
| POLLRDNORM
) : 0;
5833 /* Is there any exceptional events? */
5834 if (sk
->sk_err
|| !skb_queue_empty(&sk
->sk_error_queue
))
5836 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
5837 mask
|= POLLRDHUP
| POLLIN
| POLLRDNORM
;
5838 if (sk
->sk_shutdown
== SHUTDOWN_MASK
)
5841 /* Is it readable? Reconsider this code with TCP-style support. */
5842 if (!skb_queue_empty(&sk
->sk_receive_queue
))
5843 mask
|= POLLIN
| POLLRDNORM
;
5845 /* The association is either gone or not ready. */
5846 if (!sctp_style(sk
, UDP
) && sctp_sstate(sk
, CLOSED
))
5849 /* Is it writable? */
5850 if (sctp_writeable(sk
)) {
5851 mask
|= POLLOUT
| POLLWRNORM
;
5853 set_bit(SOCK_ASYNC_NOSPACE
, &sk
->sk_socket
->flags
);
5855 * Since the socket is not locked, the buffer
5856 * might be made available after the writeable check and
5857 * before the bit is set. This could cause a lost I/O
5858 * signal. tcp_poll() has a race breaker for this race
5859 * condition. Based on their implementation, we put
5860 * in the following code to cover it as well.
5862 if (sctp_writeable(sk
))
5863 mask
|= POLLOUT
| POLLWRNORM
;
5868 /********************************************************************
5869 * 2nd Level Abstractions
5870 ********************************************************************/
5872 static struct sctp_bind_bucket
*sctp_bucket_create(
5873 struct sctp_bind_hashbucket
*head
, unsigned short snum
)
5875 struct sctp_bind_bucket
*pp
;
5877 pp
= kmem_cache_alloc(sctp_bucket_cachep
, GFP_ATOMIC
);
5879 SCTP_DBG_OBJCNT_INC(bind_bucket
);
5882 INIT_HLIST_HEAD(&pp
->owner
);
5883 hlist_add_head(&pp
->node
, &head
->chain
);
5888 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5889 static void sctp_bucket_destroy(struct sctp_bind_bucket
*pp
)
5891 if (pp
&& hlist_empty(&pp
->owner
)) {
5892 __hlist_del(&pp
->node
);
5893 kmem_cache_free(sctp_bucket_cachep
, pp
);
5894 SCTP_DBG_OBJCNT_DEC(bind_bucket
);
5898 /* Release this socket's reference to a local port. */
5899 static inline void __sctp_put_port(struct sock
*sk
)
5901 struct sctp_bind_hashbucket
*head
=
5902 &sctp_port_hashtable
[sctp_phashfn(inet_sk(sk
)->inet_num
)];
5903 struct sctp_bind_bucket
*pp
;
5905 sctp_spin_lock(&head
->lock
);
5906 pp
= sctp_sk(sk
)->bind_hash
;
5907 __sk_del_bind_node(sk
);
5908 sctp_sk(sk
)->bind_hash
= NULL
;
5909 inet_sk(sk
)->inet_num
= 0;
5910 sctp_bucket_destroy(pp
);
5911 sctp_spin_unlock(&head
->lock
);
5914 void sctp_put_port(struct sock
*sk
)
5916 sctp_local_bh_disable();
5917 __sctp_put_port(sk
);
5918 sctp_local_bh_enable();
5922 * The system picks an ephemeral port and choose an address set equivalent
5923 * to binding with a wildcard address.
5924 * One of those addresses will be the primary address for the association.
5925 * This automatically enables the multihoming capability of SCTP.
5927 static int sctp_autobind(struct sock
*sk
)
5929 union sctp_addr autoaddr
;
5933 /* Initialize a local sockaddr structure to INADDR_ANY. */
5934 af
= sctp_sk(sk
)->pf
->af
;
5936 port
= htons(inet_sk(sk
)->inet_num
);
5937 af
->inaddr_any(&autoaddr
, port
);
5939 return sctp_do_bind(sk
, &autoaddr
, af
->sockaddr_len
);
5942 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5945 * 4.2 The cmsghdr Structure *
5947 * When ancillary data is sent or received, any number of ancillary data
5948 * objects can be specified by the msg_control and msg_controllen members of
5949 * the msghdr structure, because each object is preceded by
5950 * a cmsghdr structure defining the object's length (the cmsg_len member).
5951 * Historically Berkeley-derived implementations have passed only one object
5952 * at a time, but this API allows multiple objects to be
5953 * passed in a single call to sendmsg() or recvmsg(). The following example
5954 * shows two ancillary data objects in a control buffer.
5956 * |<--------------------------- msg_controllen -------------------------->|
5959 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5961 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5964 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5966 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5969 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5970 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5972 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5974 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5981 SCTP_STATIC
int sctp_msghdr_parse(const struct msghdr
*msg
,
5982 sctp_cmsgs_t
*cmsgs
)
5984 struct cmsghdr
*cmsg
;
5985 struct msghdr
*my_msg
= (struct msghdr
*)msg
;
5987 for (cmsg
= CMSG_FIRSTHDR(msg
);
5989 cmsg
= CMSG_NXTHDR(my_msg
, cmsg
)) {
5990 if (!CMSG_OK(my_msg
, cmsg
))
5993 /* Should we parse this header or ignore? */
5994 if (cmsg
->cmsg_level
!= IPPROTO_SCTP
)
5997 /* Strictly check lengths following example in SCM code. */
5998 switch (cmsg
->cmsg_type
) {
6000 /* SCTP Socket API Extension
6001 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6003 * This cmsghdr structure provides information for
6004 * initializing new SCTP associations with sendmsg().
6005 * The SCTP_INITMSG socket option uses this same data
6006 * structure. This structure is not used for
6009 * cmsg_level cmsg_type cmsg_data[]
6010 * ------------ ------------ ----------------------
6011 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6013 if (cmsg
->cmsg_len
!=
6014 CMSG_LEN(sizeof(struct sctp_initmsg
)))
6016 cmsgs
->init
= (struct sctp_initmsg
*)CMSG_DATA(cmsg
);
6020 /* SCTP Socket API Extension
6021 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6023 * This cmsghdr structure specifies SCTP options for
6024 * sendmsg() and describes SCTP header information
6025 * about a received message through recvmsg().
6027 * cmsg_level cmsg_type cmsg_data[]
6028 * ------------ ------------ ----------------------
6029 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6031 if (cmsg
->cmsg_len
!=
6032 CMSG_LEN(sizeof(struct sctp_sndrcvinfo
)))
6036 (struct sctp_sndrcvinfo
*)CMSG_DATA(cmsg
);
6038 /* Minimally, validate the sinfo_flags. */
6039 if (cmsgs
->info
->sinfo_flags
&
6040 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
6041 SCTP_ABORT
| SCTP_EOF
))
6053 * Wait for a packet..
6054 * Note: This function is the same function as in core/datagram.c
6055 * with a few modifications to make lksctp work.
6057 static int sctp_wait_for_packet(struct sock
* sk
, int *err
, long *timeo_p
)
6062 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
6064 /* Socket errors? */
6065 error
= sock_error(sk
);
6069 if (!skb_queue_empty(&sk
->sk_receive_queue
))
6072 /* Socket shut down? */
6073 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6076 /* Sequenced packets can come disconnected. If so we report the
6081 /* Is there a good reason to think that we may receive some data? */
6082 if (list_empty(&sctp_sk(sk
)->ep
->asocs
) && !sctp_sstate(sk
, LISTENING
))
6085 /* Handle signals. */
6086 if (signal_pending(current
))
6089 /* Let another process have a go. Since we are going to sleep
6090 * anyway. Note: This may cause odd behaviors if the message
6091 * does not fit in the user's buffer, but this seems to be the
6092 * only way to honor MSG_DONTWAIT realistically.
6094 sctp_release_sock(sk
);
6095 *timeo_p
= schedule_timeout(*timeo_p
);
6099 finish_wait(sk_sleep(sk
), &wait
);
6103 error
= sock_intr_errno(*timeo_p
);
6106 finish_wait(sk_sleep(sk
), &wait
);
6111 /* Receive a datagram.
6112 * Note: This is pretty much the same routine as in core/datagram.c
6113 * with a few changes to make lksctp work.
6115 static struct sk_buff
*sctp_skb_recv_datagram(struct sock
*sk
, int flags
,
6116 int noblock
, int *err
)
6119 struct sk_buff
*skb
;
6122 timeo
= sock_rcvtimeo(sk
, noblock
);
6124 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6125 timeo
, MAX_SCHEDULE_TIMEOUT
);
6128 /* Again only user level code calls this function,
6129 * so nothing interrupt level
6130 * will suddenly eat the receive_queue.
6132 * Look at current nfs client by the way...
6133 * However, this function was correct in any case. 8)
6135 if (flags
& MSG_PEEK
) {
6136 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
6137 skb
= skb_peek(&sk
->sk_receive_queue
);
6139 atomic_inc(&skb
->users
);
6140 spin_unlock_bh(&sk
->sk_receive_queue
.lock
);
6142 skb
= skb_dequeue(&sk
->sk_receive_queue
);
6148 /* Caller is allowed not to check sk->sk_err before calling. */
6149 error
= sock_error(sk
);
6153 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6156 /* User doesn't want to wait. */
6160 } while (sctp_wait_for_packet(sk
, err
, &timeo
) == 0);
6169 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6170 static void __sctp_write_space(struct sctp_association
*asoc
)
6172 struct sock
*sk
= asoc
->base
.sk
;
6173 struct socket
*sock
= sk
->sk_socket
;
6175 if ((sctp_wspace(asoc
) > 0) && sock
) {
6176 if (waitqueue_active(&asoc
->wait
))
6177 wake_up_interruptible(&asoc
->wait
);
6179 if (sctp_writeable(sk
)) {
6180 wait_queue_head_t
*wq
= sk_sleep(sk
);
6182 if (wq
&& waitqueue_active(wq
))
6183 wake_up_interruptible(wq
);
6185 /* Note that we try to include the Async I/O support
6186 * here by modeling from the current TCP/UDP code.
6187 * We have not tested with it yet.
6189 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
))
6190 sock_wake_async(sock
,
6191 SOCK_WAKE_SPACE
, POLL_OUT
);
6196 /* Do accounting for the sndbuf space.
6197 * Decrement the used sndbuf space of the corresponding association by the
6198 * data size which was just transmitted(freed).
6200 static void sctp_wfree(struct sk_buff
*skb
)
6202 struct sctp_association
*asoc
;
6203 struct sctp_chunk
*chunk
;
6206 /* Get the saved chunk pointer. */
6207 chunk
= *((struct sctp_chunk
**)(skb
->cb
));
6210 asoc
->sndbuf_used
-= SCTP_DATA_SNDSIZE(chunk
) +
6211 sizeof(struct sk_buff
) +
6212 sizeof(struct sctp_chunk
);
6214 atomic_sub(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
6217 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6219 sk
->sk_wmem_queued
-= skb
->truesize
;
6220 sk_mem_uncharge(sk
, skb
->truesize
);
6223 __sctp_write_space(asoc
);
6225 sctp_association_put(asoc
);
6228 /* Do accounting for the receive space on the socket.
6229 * Accounting for the association is done in ulpevent.c
6230 * We set this as a destructor for the cloned data skbs so that
6231 * accounting is done at the correct time.
6233 void sctp_sock_rfree(struct sk_buff
*skb
)
6235 struct sock
*sk
= skb
->sk
;
6236 struct sctp_ulpevent
*event
= sctp_skb2event(skb
);
6238 atomic_sub(event
->rmem_len
, &sk
->sk_rmem_alloc
);
6241 * Mimic the behavior of sock_rfree
6243 sk_mem_uncharge(sk
, event
->rmem_len
);
6247 /* Helper function to wait for space in the sndbuf. */
6248 static int sctp_wait_for_sndbuf(struct sctp_association
*asoc
, long *timeo_p
,
6251 struct sock
*sk
= asoc
->base
.sk
;
6253 long current_timeo
= *timeo_p
;
6256 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6257 asoc
, (long)(*timeo_p
), msg_len
);
6259 /* Increment the association's refcnt. */
6260 sctp_association_hold(asoc
);
6262 /* Wait on the association specific sndbuf space. */
6264 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
6265 TASK_INTERRUPTIBLE
);
6268 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
6271 if (signal_pending(current
))
6272 goto do_interrupted
;
6273 if (msg_len
<= sctp_wspace(asoc
))
6276 /* Let another process have a go. Since we are going
6279 sctp_release_sock(sk
);
6280 current_timeo
= schedule_timeout(current_timeo
);
6281 BUG_ON(sk
!= asoc
->base
.sk
);
6284 *timeo_p
= current_timeo
;
6288 finish_wait(&asoc
->wait
, &wait
);
6290 /* Release the association's refcnt. */
6291 sctp_association_put(asoc
);
6300 err
= sock_intr_errno(*timeo_p
);
6308 void sctp_data_ready(struct sock
*sk
, int len
)
6310 struct socket_wq
*wq
;
6313 wq
= rcu_dereference(sk
->sk_wq
);
6314 if (wq_has_sleeper(wq
))
6315 wake_up_interruptible_sync_poll(&wq
->wait
, POLLIN
|
6316 POLLRDNORM
| POLLRDBAND
);
6317 sk_wake_async(sk
, SOCK_WAKE_WAITD
, POLL_IN
);
6321 /* If socket sndbuf has changed, wake up all per association waiters. */
6322 void sctp_write_space(struct sock
*sk
)
6324 struct sctp_association
*asoc
;
6326 /* Wake up the tasks in each wait queue. */
6327 list_for_each_entry(asoc
, &((sctp_sk(sk
))->ep
->asocs
), asocs
) {
6328 __sctp_write_space(asoc
);
6332 /* Is there any sndbuf space available on the socket?
6334 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6335 * associations on the same socket. For a UDP-style socket with
6336 * multiple associations, it is possible for it to be "unwriteable"
6337 * prematurely. I assume that this is acceptable because
6338 * a premature "unwriteable" is better than an accidental "writeable" which
6339 * would cause an unwanted block under certain circumstances. For the 1-1
6340 * UDP-style sockets or TCP-style sockets, this code should work.
6343 static int sctp_writeable(struct sock
*sk
)
6347 amt
= sk
->sk_sndbuf
- sk_wmem_alloc_get(sk
);
6353 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6354 * returns immediately with EINPROGRESS.
6356 static int sctp_wait_for_connect(struct sctp_association
*asoc
, long *timeo_p
)
6358 struct sock
*sk
= asoc
->base
.sk
;
6360 long current_timeo
= *timeo_p
;
6363 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__
, asoc
,
6366 /* Increment the association's refcnt. */
6367 sctp_association_hold(asoc
);
6370 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
6371 TASK_INTERRUPTIBLE
);
6374 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6376 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
6379 if (signal_pending(current
))
6380 goto do_interrupted
;
6382 if (sctp_state(asoc
, ESTABLISHED
))
6385 /* Let another process have a go. Since we are going
6388 sctp_release_sock(sk
);
6389 current_timeo
= schedule_timeout(current_timeo
);
6392 *timeo_p
= current_timeo
;
6396 finish_wait(&asoc
->wait
, &wait
);
6398 /* Release the association's refcnt. */
6399 sctp_association_put(asoc
);
6404 if (asoc
->init_err_counter
+ 1 > asoc
->max_init_attempts
)
6407 err
= -ECONNREFUSED
;
6411 err
= sock_intr_errno(*timeo_p
);
6419 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
)
6421 struct sctp_endpoint
*ep
;
6425 ep
= sctp_sk(sk
)->ep
;
6429 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
,
6430 TASK_INTERRUPTIBLE
);
6432 if (list_empty(&ep
->asocs
)) {
6433 sctp_release_sock(sk
);
6434 timeo
= schedule_timeout(timeo
);
6439 if (!sctp_sstate(sk
, LISTENING
))
6443 if (!list_empty(&ep
->asocs
))
6446 err
= sock_intr_errno(timeo
);
6447 if (signal_pending(current
))
6455 finish_wait(sk_sleep(sk
), &wait
);
6460 static void sctp_wait_for_close(struct sock
*sk
, long timeout
)
6465 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
6466 if (list_empty(&sctp_sk(sk
)->ep
->asocs
))
6468 sctp_release_sock(sk
);
6469 timeout
= schedule_timeout(timeout
);
6471 } while (!signal_pending(current
) && timeout
);
6473 finish_wait(sk_sleep(sk
), &wait
);
6476 static void sctp_skb_set_owner_r_frag(struct sk_buff
*skb
, struct sock
*sk
)
6478 struct sk_buff
*frag
;
6483 /* Don't forget the fragments. */
6484 skb_walk_frags(skb
, frag
)
6485 sctp_skb_set_owner_r_frag(frag
, sk
);
6488 sctp_skb_set_owner_r(skb
, sk
);
6491 void sctp_copy_sock(struct sock
*newsk
, struct sock
*sk
,
6492 struct sctp_association
*asoc
)
6494 struct inet_sock
*inet
= inet_sk(sk
);
6495 struct inet_sock
*newinet
;
6497 newsk
->sk_type
= sk
->sk_type
;
6498 newsk
->sk_bound_dev_if
= sk
->sk_bound_dev_if
;
6499 newsk
->sk_flags
= sk
->sk_flags
;
6500 newsk
->sk_no_check
= sk
->sk_no_check
;
6501 newsk
->sk_reuse
= sk
->sk_reuse
;
6503 newsk
->sk_shutdown
= sk
->sk_shutdown
;
6504 newsk
->sk_destruct
= inet_sock_destruct
;
6505 newsk
->sk_family
= sk
->sk_family
;
6506 newsk
->sk_protocol
= IPPROTO_SCTP
;
6507 newsk
->sk_backlog_rcv
= sk
->sk_prot
->backlog_rcv
;
6508 newsk
->sk_sndbuf
= sk
->sk_sndbuf
;
6509 newsk
->sk_rcvbuf
= sk
->sk_rcvbuf
;
6510 newsk
->sk_lingertime
= sk
->sk_lingertime
;
6511 newsk
->sk_rcvtimeo
= sk
->sk_rcvtimeo
;
6512 newsk
->sk_sndtimeo
= sk
->sk_sndtimeo
;
6514 newinet
= inet_sk(newsk
);
6516 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6517 * getsockname() and getpeername()
6519 newinet
->inet_sport
= inet
->inet_sport
;
6520 newinet
->inet_saddr
= inet
->inet_saddr
;
6521 newinet
->inet_rcv_saddr
= inet
->inet_rcv_saddr
;
6522 newinet
->inet_dport
= htons(asoc
->peer
.port
);
6523 newinet
->pmtudisc
= inet
->pmtudisc
;
6524 newinet
->inet_id
= asoc
->next_tsn
^ jiffies
;
6526 newinet
->uc_ttl
= inet
->uc_ttl
;
6527 newinet
->mc_loop
= 1;
6528 newinet
->mc_ttl
= 1;
6529 newinet
->mc_index
= 0;
6530 newinet
->mc_list
= NULL
;
6533 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6534 * and its messages to the newsk.
6536 static void sctp_sock_migrate(struct sock
*oldsk
, struct sock
*newsk
,
6537 struct sctp_association
*assoc
,
6538 sctp_socket_type_t type
)
6540 struct sctp_sock
*oldsp
= sctp_sk(oldsk
);
6541 struct sctp_sock
*newsp
= sctp_sk(newsk
);
6542 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
6543 struct sctp_endpoint
*newep
= newsp
->ep
;
6544 struct sk_buff
*skb
, *tmp
;
6545 struct sctp_ulpevent
*event
;
6546 struct sctp_bind_hashbucket
*head
;
6547 struct list_head tmplist
;
6549 /* Migrate socket buffer sizes and all the socket level options to the
6552 newsk
->sk_sndbuf
= oldsk
->sk_sndbuf
;
6553 newsk
->sk_rcvbuf
= oldsk
->sk_rcvbuf
;
6554 /* Brute force copy old sctp opt. */
6555 if (oldsp
->do_auto_asconf
) {
6556 memcpy(&tmplist
, &newsp
->auto_asconf_list
, sizeof(tmplist
));
6557 inet_sk_copy_descendant(newsk
, oldsk
);
6558 memcpy(&newsp
->auto_asconf_list
, &tmplist
, sizeof(tmplist
));
6560 inet_sk_copy_descendant(newsk
, oldsk
);
6562 /* Restore the ep value that was overwritten with the above structure
6568 /* Hook this new socket in to the bind_hash list. */
6569 head
= &sctp_port_hashtable
[sctp_phashfn(inet_sk(oldsk
)->inet_num
)];
6570 sctp_local_bh_disable();
6571 sctp_spin_lock(&head
->lock
);
6572 pp
= sctp_sk(oldsk
)->bind_hash
;
6573 sk_add_bind_node(newsk
, &pp
->owner
);
6574 sctp_sk(newsk
)->bind_hash
= pp
;
6575 inet_sk(newsk
)->inet_num
= inet_sk(oldsk
)->inet_num
;
6576 sctp_spin_unlock(&head
->lock
);
6577 sctp_local_bh_enable();
6579 /* Copy the bind_addr list from the original endpoint to the new
6580 * endpoint so that we can handle restarts properly
6582 sctp_bind_addr_dup(&newsp
->ep
->base
.bind_addr
,
6583 &oldsp
->ep
->base
.bind_addr
, GFP_KERNEL
);
6585 /* Move any messages in the old socket's receive queue that are for the
6586 * peeled off association to the new socket's receive queue.
6588 sctp_skb_for_each(skb
, &oldsk
->sk_receive_queue
, tmp
) {
6589 event
= sctp_skb2event(skb
);
6590 if (event
->asoc
== assoc
) {
6591 __skb_unlink(skb
, &oldsk
->sk_receive_queue
);
6592 __skb_queue_tail(&newsk
->sk_receive_queue
, skb
);
6593 sctp_skb_set_owner_r_frag(skb
, newsk
);
6597 /* Clean up any messages pending delivery due to partial
6598 * delivery. Three cases:
6599 * 1) No partial deliver; no work.
6600 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6601 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6603 skb_queue_head_init(&newsp
->pd_lobby
);
6604 atomic_set(&sctp_sk(newsk
)->pd_mode
, assoc
->ulpq
.pd_mode
);
6606 if (atomic_read(&sctp_sk(oldsk
)->pd_mode
)) {
6607 struct sk_buff_head
*queue
;
6609 /* Decide which queue to move pd_lobby skbs to. */
6610 if (assoc
->ulpq
.pd_mode
) {
6611 queue
= &newsp
->pd_lobby
;
6613 queue
= &newsk
->sk_receive_queue
;
6615 /* Walk through the pd_lobby, looking for skbs that
6616 * need moved to the new socket.
6618 sctp_skb_for_each(skb
, &oldsp
->pd_lobby
, tmp
) {
6619 event
= sctp_skb2event(skb
);
6620 if (event
->asoc
== assoc
) {
6621 __skb_unlink(skb
, &oldsp
->pd_lobby
);
6622 __skb_queue_tail(queue
, skb
);
6623 sctp_skb_set_owner_r_frag(skb
, newsk
);
6627 /* Clear up any skbs waiting for the partial
6628 * delivery to finish.
6630 if (assoc
->ulpq
.pd_mode
)
6631 sctp_clear_pd(oldsk
, NULL
);
6635 sctp_skb_for_each(skb
, &assoc
->ulpq
.reasm
, tmp
)
6636 sctp_skb_set_owner_r_frag(skb
, newsk
);
6638 sctp_skb_for_each(skb
, &assoc
->ulpq
.lobby
, tmp
)
6639 sctp_skb_set_owner_r_frag(skb
, newsk
);
6641 /* Set the type of socket to indicate that it is peeled off from the
6642 * original UDP-style socket or created with the accept() call on a
6643 * TCP-style socket..
6647 /* Mark the new socket "in-use" by the user so that any packets
6648 * that may arrive on the association after we've moved it are
6649 * queued to the backlog. This prevents a potential race between
6650 * backlog processing on the old socket and new-packet processing
6651 * on the new socket.
6653 * The caller has just allocated newsk so we can guarantee that other
6654 * paths won't try to lock it and then oldsk.
6656 lock_sock_nested(newsk
, SINGLE_DEPTH_NESTING
);
6657 sctp_assoc_migrate(assoc
, newsk
);
6659 /* If the association on the newsk is already closed before accept()
6660 * is called, set RCV_SHUTDOWN flag.
6662 if (sctp_state(assoc
, CLOSED
) && sctp_style(newsk
, TCP
))
6663 newsk
->sk_shutdown
|= RCV_SHUTDOWN
;
6665 newsk
->sk_state
= SCTP_SS_ESTABLISHED
;
6666 sctp_release_sock(newsk
);
6670 /* This proto struct describes the ULP interface for SCTP. */
6671 struct proto sctp_prot
= {
6673 .owner
= THIS_MODULE
,
6674 .close
= sctp_close
,
6675 .connect
= sctp_connect
,
6676 .disconnect
= sctp_disconnect
,
6677 .accept
= sctp_accept
,
6678 .ioctl
= sctp_ioctl
,
6679 .init
= sctp_init_sock
,
6680 .destroy
= sctp_destroy_sock
,
6681 .shutdown
= sctp_shutdown
,
6682 .setsockopt
= sctp_setsockopt
,
6683 .getsockopt
= sctp_getsockopt
,
6684 .sendmsg
= sctp_sendmsg
,
6685 .recvmsg
= sctp_recvmsg
,
6687 .backlog_rcv
= sctp_backlog_rcv
,
6689 .unhash
= sctp_unhash
,
6690 .get_port
= sctp_get_port
,
6691 .obj_size
= sizeof(struct sctp_sock
),
6692 .sysctl_mem
= sysctl_sctp_mem
,
6693 .sysctl_rmem
= sysctl_sctp_rmem
,
6694 .sysctl_wmem
= sysctl_sctp_wmem
,
6695 .memory_pressure
= &sctp_memory_pressure
,
6696 .enter_memory_pressure
= sctp_enter_memory_pressure
,
6697 .memory_allocated
= &sctp_memory_allocated
,
6698 .sockets_allocated
= &sctp_sockets_allocated
,
6701 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6703 struct proto sctpv6_prot
= {
6705 .owner
= THIS_MODULE
,
6706 .close
= sctp_close
,
6707 .connect
= sctp_connect
,
6708 .disconnect
= sctp_disconnect
,
6709 .accept
= sctp_accept
,
6710 .ioctl
= sctp_ioctl
,
6711 .init
= sctp_init_sock
,
6712 .destroy
= sctp_destroy_sock
,
6713 .shutdown
= sctp_shutdown
,
6714 .setsockopt
= sctp_setsockopt
,
6715 .getsockopt
= sctp_getsockopt
,
6716 .sendmsg
= sctp_sendmsg
,
6717 .recvmsg
= sctp_recvmsg
,
6719 .backlog_rcv
= sctp_backlog_rcv
,
6721 .unhash
= sctp_unhash
,
6722 .get_port
= sctp_get_port
,
6723 .obj_size
= sizeof(struct sctp6_sock
),
6724 .sysctl_mem
= sysctl_sctp_mem
,
6725 .sysctl_rmem
= sysctl_sctp_rmem
,
6726 .sysctl_wmem
= sysctl_sctp_wmem
,
6727 .memory_pressure
= &sctp_memory_pressure
,
6728 .enter_memory_pressure
= sctp_enter_memory_pressure
,
6729 .memory_allocated
= &sctp_memory_allocated
,
6730 .sockets_allocated
= &sctp_sockets_allocated
,
6732 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */