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 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
817 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
820 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
821 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
824 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
825 * Section 3.1.2 for this usage.
827 * addrs is a pointer to an array of one or more socket addresses. Each
828 * address is contained in its appropriate structure (i.e. struct
829 * sockaddr_in or struct sockaddr_in6) the family of the address type
830 * must be used to distinguish the address length (note that this
831 * representation is termed a "packed array" of addresses). The caller
832 * specifies the number of addresses in the array with addrcnt.
834 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
835 * -1, and sets errno to the appropriate error code.
837 * For SCTP, the port given in each socket address must be the same, or
838 * sctp_bindx() will fail, setting errno to EINVAL.
840 * The flags parameter is formed from the bitwise OR of zero or more of
841 * the following currently defined flags:
843 * SCTP_BINDX_ADD_ADDR
845 * SCTP_BINDX_REM_ADDR
847 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
848 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
849 * addresses from the association. The two flags are mutually exclusive;
850 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
851 * not remove all addresses from an association; sctp_bindx() will
852 * reject such an attempt with EINVAL.
854 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
855 * additional addresses with an endpoint after calling bind(). Or use
856 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
857 * socket is associated with so that no new association accepted will be
858 * associated with those addresses. If the endpoint supports dynamic
859 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
860 * endpoint to send the appropriate message to the peer to change the
861 * peers address lists.
863 * Adding and removing addresses from a connected association is
864 * optional functionality. Implementations that do not support this
865 * functionality should return EOPNOTSUPP.
867 * Basically do nothing but copying the addresses from user to kernel
868 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
869 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
872 * We don't use copy_from_user() for optimization: we first do the
873 * sanity checks (buffer size -fast- and access check-healthy
874 * pointer); if all of those succeed, then we can alloc the memory
875 * (expensive operation) needed to copy the data to kernel. Then we do
876 * the copying without checking the user space area
877 * (__copy_from_user()).
879 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
882 * sk The sk of the socket
883 * addrs The pointer to the addresses in user land
884 * addrssize Size of the addrs buffer
885 * op Operation to perform (add or remove, see the flags of
888 * Returns 0 if ok, <0 errno code on error.
890 SCTP_STATIC
int sctp_setsockopt_bindx(struct sock
* sk
,
891 struct sockaddr __user
*addrs
,
892 int addrs_size
, int op
)
894 struct sockaddr
*kaddrs
;
898 struct sockaddr
*sa_addr
;
902 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
903 " addrs_size %d opt %d\n", sk
, addrs
, addrs_size
, op
);
905 if (unlikely(addrs_size
<= 0))
908 /* Check the user passed a healthy pointer. */
909 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
912 /* Alloc space for the address array in kernel memory. */
913 kaddrs
= kmalloc(addrs_size
, GFP_KERNEL
);
914 if (unlikely(!kaddrs
))
917 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
922 /* Walk through the addrs buffer and count the number of addresses. */
924 while (walk_size
< addrs_size
) {
925 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
930 sa_addr
= (struct sockaddr
*)addr_buf
;
931 af
= sctp_get_af_specific(sa_addr
->sa_family
);
933 /* If the address family is not supported or if this address
934 * causes the address buffer to overflow return EINVAL.
936 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
941 addr_buf
+= af
->sockaddr_len
;
942 walk_size
+= af
->sockaddr_len
;
947 case SCTP_BINDX_ADD_ADDR
:
948 err
= sctp_bindx_add(sk
, kaddrs
, addrcnt
);
951 err
= sctp_send_asconf_add_ip(sk
, kaddrs
, addrcnt
);
954 case SCTP_BINDX_REM_ADDR
:
955 err
= sctp_bindx_rem(sk
, kaddrs
, addrcnt
);
958 err
= sctp_send_asconf_del_ip(sk
, kaddrs
, addrcnt
);
972 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
974 * Common routine for handling connect() and sctp_connectx().
975 * Connect will come in with just a single address.
977 static int __sctp_connect(struct sock
* sk
,
978 struct sockaddr
*kaddrs
,
980 sctp_assoc_t
*assoc_id
)
982 struct sctp_sock
*sp
;
983 struct sctp_endpoint
*ep
;
984 struct sctp_association
*asoc
= NULL
;
985 struct sctp_association
*asoc2
;
986 struct sctp_transport
*transport
;
994 union sctp_addr
*sa_addr
= NULL
;
997 unsigned int f_flags
= 0;
1002 /* connect() cannot be done on a socket that is already in ESTABLISHED
1003 * state - UDP-style peeled off socket or a TCP-style socket that
1004 * is already connected.
1005 * It cannot be done even on a TCP-style listening socket.
1007 if (sctp_sstate(sk
, ESTABLISHED
) ||
1008 (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))) {
1013 /* Walk through the addrs buffer and count the number of addresses. */
1015 while (walk_size
< addrs_size
) {
1016 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
1021 sa_addr
= (union sctp_addr
*)addr_buf
;
1022 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
1024 /* If the address family is not supported or if this address
1025 * causes the address buffer to overflow return EINVAL.
1027 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
1032 port
= ntohs(sa_addr
->v4
.sin_port
);
1034 /* Save current address so we can work with it */
1035 memcpy(&to
, sa_addr
, af
->sockaddr_len
);
1037 err
= sctp_verify_addr(sk
, &to
, af
->sockaddr_len
);
1041 /* Make sure the destination port is correctly set
1044 if (asoc
&& asoc
->peer
.port
&& asoc
->peer
.port
!= port
)
1048 /* Check if there already is a matching association on the
1049 * endpoint (other than the one created here).
1051 asoc2
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1052 if (asoc2
&& asoc2
!= asoc
) {
1053 if (asoc2
->state
>= SCTP_STATE_ESTABLISHED
)
1060 /* If we could not find a matching association on the endpoint,
1061 * make sure that there is no peeled-off association matching
1062 * the peer address even on another socket.
1064 if (sctp_endpoint_is_peeled_off(ep
, &to
)) {
1065 err
= -EADDRNOTAVAIL
;
1070 /* If a bind() or sctp_bindx() is not called prior to
1071 * an sctp_connectx() call, the system picks an
1072 * ephemeral port and will choose an address set
1073 * equivalent to binding with a wildcard address.
1075 if (!ep
->base
.bind_addr
.port
) {
1076 if (sctp_autobind(sk
)) {
1082 * If an unprivileged user inherits a 1-many
1083 * style socket with open associations on a
1084 * privileged port, it MAY be permitted to
1085 * accept new associations, but it SHOULD NOT
1086 * be permitted to open new associations.
1088 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1089 !capable(CAP_NET_BIND_SERVICE
)) {
1095 scope
= sctp_scope(&to
);
1096 asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1102 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
,
1110 /* Prime the peer's transport structures. */
1111 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
,
1119 addr_buf
+= af
->sockaddr_len
;
1120 walk_size
+= af
->sockaddr_len
;
1123 /* In case the user of sctp_connectx() wants an association
1124 * id back, assign one now.
1127 err
= sctp_assoc_set_id(asoc
, GFP_KERNEL
);
1132 err
= sctp_primitive_ASSOCIATE(asoc
, NULL
);
1137 /* Initialize sk's dport and daddr for getpeername() */
1138 inet_sk(sk
)->inet_dport
= htons(asoc
->peer
.port
);
1139 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
1140 af
->to_sk_daddr(sa_addr
, sk
);
1143 /* in-kernel sockets don't generally have a file allocated to them
1144 * if all they do is call sock_create_kern().
1146 if (sk
->sk_socket
->file
)
1147 f_flags
= sk
->sk_socket
->file
->f_flags
;
1149 timeo
= sock_sndtimeo(sk
, f_flags
& O_NONBLOCK
);
1151 err
= sctp_wait_for_connect(asoc
, &timeo
);
1152 if ((err
== 0 || err
== -EINPROGRESS
) && assoc_id
)
1153 *assoc_id
= asoc
->assoc_id
;
1155 /* Don't free association on exit. */
1160 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1161 " kaddrs: %p err: %d\n",
1164 sctp_association_free(asoc
);
1168 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1171 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1172 * sctp_assoc_t *asoc);
1174 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1175 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1176 * or IPv6 addresses.
1178 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1179 * Section 3.1.2 for this usage.
1181 * addrs is a pointer to an array of one or more socket addresses. Each
1182 * address is contained in its appropriate structure (i.e. struct
1183 * sockaddr_in or struct sockaddr_in6) the family of the address type
1184 * must be used to distengish the address length (note that this
1185 * representation is termed a "packed array" of addresses). The caller
1186 * specifies the number of addresses in the array with addrcnt.
1188 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1189 * the association id of the new association. On failure, sctp_connectx()
1190 * returns -1, and sets errno to the appropriate error code. The assoc_id
1191 * is not touched by the kernel.
1193 * For SCTP, the port given in each socket address must be the same, or
1194 * sctp_connectx() will fail, setting errno to EINVAL.
1196 * An application can use sctp_connectx to initiate an association with
1197 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1198 * allows a caller to specify multiple addresses at which a peer can be
1199 * reached. The way the SCTP stack uses the list of addresses to set up
1200 * the association is implementation dependent. This function only
1201 * specifies that the stack will try to make use of all the addresses in
1202 * the list when needed.
1204 * Note that the list of addresses passed in is only used for setting up
1205 * the association. It does not necessarily equal the set of addresses
1206 * the peer uses for the resulting association. If the caller wants to
1207 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1208 * retrieve them after the association has been set up.
1210 * Basically do nothing but copying the addresses from user to kernel
1211 * land and invoking either sctp_connectx(). This is used for tunneling
1212 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1214 * We don't use copy_from_user() for optimization: we first do the
1215 * sanity checks (buffer size -fast- and access check-healthy
1216 * pointer); if all of those succeed, then we can alloc the memory
1217 * (expensive operation) needed to copy the data to kernel. Then we do
1218 * the copying without checking the user space area
1219 * (__copy_from_user()).
1221 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1224 * sk The sk of the socket
1225 * addrs The pointer to the addresses in user land
1226 * addrssize Size of the addrs buffer
1228 * Returns >=0 if ok, <0 errno code on error.
1230 SCTP_STATIC
int __sctp_setsockopt_connectx(struct sock
* sk
,
1231 struct sockaddr __user
*addrs
,
1233 sctp_assoc_t
*assoc_id
)
1236 struct sockaddr
*kaddrs
;
1238 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1239 __func__
, sk
, addrs
, addrs_size
);
1241 if (unlikely(addrs_size
<= 0))
1244 /* Check the user passed a healthy pointer. */
1245 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
1248 /* Alloc space for the address array in kernel memory. */
1249 kaddrs
= kmalloc(addrs_size
, GFP_KERNEL
);
1250 if (unlikely(!kaddrs
))
1253 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
1256 err
= __sctp_connect(sk
, kaddrs
, addrs_size
, assoc_id
);
1265 * This is an older interface. It's kept for backward compatibility
1266 * to the option that doesn't provide association id.
1268 SCTP_STATIC
int sctp_setsockopt_connectx_old(struct sock
* sk
,
1269 struct sockaddr __user
*addrs
,
1272 return __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, NULL
);
1276 * New interface for the API. The since the API is done with a socket
1277 * option, to make it simple we feed back the association id is as a return
1278 * indication to the call. Error is always negative and association id is
1281 SCTP_STATIC
int sctp_setsockopt_connectx(struct sock
* sk
,
1282 struct sockaddr __user
*addrs
,
1285 sctp_assoc_t assoc_id
= 0;
1288 err
= __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, &assoc_id
);
1297 * New (hopefully final) interface for the API.
1298 * We use the sctp_getaddrs_old structure so that use-space library
1299 * can avoid any unnecessary allocations. The only defferent part
1300 * is that we store the actual length of the address buffer into the
1301 * addrs_num structure member. That way we can re-use the existing
1304 SCTP_STATIC
int sctp_getsockopt_connectx3(struct sock
* sk
, int len
,
1305 char __user
*optval
,
1308 struct sctp_getaddrs_old param
;
1309 sctp_assoc_t assoc_id
= 0;
1312 if (len
< sizeof(param
))
1315 if (copy_from_user(¶m
, optval
, sizeof(param
)))
1318 err
= __sctp_setsockopt_connectx(sk
,
1319 (struct sockaddr __user
*)param
.addrs
,
1320 param
.addr_num
, &assoc_id
);
1322 if (err
== 0 || err
== -EINPROGRESS
) {
1323 if (copy_to_user(optval
, &assoc_id
, sizeof(assoc_id
)))
1325 if (put_user(sizeof(assoc_id
), optlen
))
1332 /* API 3.1.4 close() - UDP Style Syntax
1333 * Applications use close() to perform graceful shutdown (as described in
1334 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1335 * by a UDP-style socket.
1339 * ret = close(int sd);
1341 * sd - the socket descriptor of the associations to be closed.
1343 * To gracefully shutdown a specific association represented by the
1344 * UDP-style socket, an application should use the sendmsg() call,
1345 * passing no user data, but including the appropriate flag in the
1346 * ancillary data (see Section xxxx).
1348 * If sd in the close() call is a branched-off socket representing only
1349 * one association, the shutdown is performed on that association only.
1351 * 4.1.6 close() - TCP Style Syntax
1353 * Applications use close() to gracefully close down an association.
1357 * int close(int sd);
1359 * sd - the socket descriptor of the association to be closed.
1361 * After an application calls close() on a socket descriptor, no further
1362 * socket operations will succeed on that descriptor.
1364 * API 7.1.4 SO_LINGER
1366 * An application using the TCP-style socket can use this option to
1367 * perform the SCTP ABORT primitive. The linger option structure is:
1370 * int l_onoff; // option on/off
1371 * int l_linger; // linger time
1374 * To enable the option, set l_onoff to 1. If the l_linger value is set
1375 * to 0, calling close() is the same as the ABORT primitive. If the
1376 * value is set to a negative value, the setsockopt() call will return
1377 * an error. If the value is set to a positive value linger_time, the
1378 * close() can be blocked for at most linger_time ms. If the graceful
1379 * shutdown phase does not finish during this period, close() will
1380 * return but the graceful shutdown phase continues in the system.
1382 SCTP_STATIC
void sctp_close(struct sock
*sk
, long timeout
)
1384 struct sctp_endpoint
*ep
;
1385 struct sctp_association
*asoc
;
1386 struct list_head
*pos
, *temp
;
1388 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk
, timeout
);
1391 sk
->sk_shutdown
= SHUTDOWN_MASK
;
1392 sk
->sk_state
= SCTP_SS_CLOSING
;
1394 ep
= sctp_sk(sk
)->ep
;
1396 /* Walk all associations on an endpoint. */
1397 list_for_each_safe(pos
, temp
, &ep
->asocs
) {
1398 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
1400 if (sctp_style(sk
, TCP
)) {
1401 /* A closed association can still be in the list if
1402 * it belongs to a TCP-style listening socket that is
1403 * not yet accepted. If so, free it. If not, send an
1404 * ABORT or SHUTDOWN based on the linger options.
1406 if (sctp_state(asoc
, CLOSED
)) {
1407 sctp_unhash_established(asoc
);
1408 sctp_association_free(asoc
);
1413 if (sock_flag(sk
, SOCK_LINGER
) && !sk
->sk_lingertime
) {
1414 struct sctp_chunk
*chunk
;
1416 chunk
= sctp_make_abort_user(asoc
, NULL
, 0);
1418 sctp_primitive_ABORT(asoc
, chunk
);
1420 sctp_primitive_SHUTDOWN(asoc
, NULL
);
1423 /* Clean up any skbs sitting on the receive queue. */
1424 sctp_queue_purge_ulpevents(&sk
->sk_receive_queue
);
1425 sctp_queue_purge_ulpevents(&sctp_sk(sk
)->pd_lobby
);
1427 /* On a TCP-style socket, block for at most linger_time if set. */
1428 if (sctp_style(sk
, TCP
) && timeout
)
1429 sctp_wait_for_close(sk
, timeout
);
1431 /* This will run the backlog queue. */
1432 sctp_release_sock(sk
);
1434 /* Supposedly, no process has access to the socket, but
1435 * the net layers still may.
1437 sctp_local_bh_disable();
1438 sctp_bh_lock_sock(sk
);
1440 /* Hold the sock, since sk_common_release() will put sock_put()
1441 * and we have just a little more cleanup.
1444 sk_common_release(sk
);
1446 sctp_bh_unlock_sock(sk
);
1447 sctp_local_bh_enable();
1451 SCTP_DBG_OBJCNT_DEC(sock
);
1454 /* Handle EPIPE error. */
1455 static int sctp_error(struct sock
*sk
, int flags
, int err
)
1458 err
= sock_error(sk
) ? : -EPIPE
;
1459 if (err
== -EPIPE
&& !(flags
& MSG_NOSIGNAL
))
1460 send_sig(SIGPIPE
, current
, 0);
1464 /* API 3.1.3 sendmsg() - UDP Style Syntax
1466 * An application uses sendmsg() and recvmsg() calls to transmit data to
1467 * and receive data from its peer.
1469 * ssize_t sendmsg(int socket, const struct msghdr *message,
1472 * socket - the socket descriptor of the endpoint.
1473 * message - pointer to the msghdr structure which contains a single
1474 * user message and possibly some ancillary data.
1476 * See Section 5 for complete description of the data
1479 * flags - flags sent or received with the user message, see Section
1480 * 5 for complete description of the flags.
1482 * Note: This function could use a rewrite especially when explicit
1483 * connect support comes in.
1485 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1487 SCTP_STATIC
int sctp_msghdr_parse(const struct msghdr
*, sctp_cmsgs_t
*);
1489 SCTP_STATIC
int sctp_sendmsg(struct kiocb
*iocb
, struct sock
*sk
,
1490 struct msghdr
*msg
, size_t msg_len
)
1492 struct sctp_sock
*sp
;
1493 struct sctp_endpoint
*ep
;
1494 struct sctp_association
*new_asoc
=NULL
, *asoc
=NULL
;
1495 struct sctp_transport
*transport
, *chunk_tp
;
1496 struct sctp_chunk
*chunk
;
1498 struct sockaddr
*msg_name
= NULL
;
1499 struct sctp_sndrcvinfo default_sinfo
;
1500 struct sctp_sndrcvinfo
*sinfo
;
1501 struct sctp_initmsg
*sinit
;
1502 sctp_assoc_t associd
= 0;
1503 sctp_cmsgs_t cmsgs
= { NULL
};
1507 __u16 sinfo_flags
= 0;
1508 struct sctp_datamsg
*datamsg
;
1509 int msg_flags
= msg
->msg_flags
;
1511 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1518 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep
);
1520 /* We cannot send a message over a TCP-style listening socket. */
1521 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
)) {
1526 /* Parse out the SCTP CMSGs. */
1527 err
= sctp_msghdr_parse(msg
, &cmsgs
);
1530 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err
);
1534 /* Fetch the destination address for this packet. This
1535 * address only selects the association--it is not necessarily
1536 * the address we will send to.
1537 * For a peeled-off socket, msg_name is ignored.
1539 if (!sctp_style(sk
, UDP_HIGH_BANDWIDTH
) && msg
->msg_name
) {
1540 int msg_namelen
= msg
->msg_namelen
;
1542 err
= sctp_verify_addr(sk
, (union sctp_addr
*)msg
->msg_name
,
1547 if (msg_namelen
> sizeof(to
))
1548 msg_namelen
= sizeof(to
);
1549 memcpy(&to
, msg
->msg_name
, msg_namelen
);
1550 msg_name
= msg
->msg_name
;
1556 /* Did the user specify SNDRCVINFO? */
1558 sinfo_flags
= sinfo
->sinfo_flags
;
1559 associd
= sinfo
->sinfo_assoc_id
;
1562 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1563 msg_len
, sinfo_flags
);
1565 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1566 if (sctp_style(sk
, TCP
) && (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
))) {
1571 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1572 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1573 * If SCTP_ABORT is set, the message length could be non zero with
1574 * the msg_iov set to the user abort reason.
1576 if (((sinfo_flags
& SCTP_EOF
) && (msg_len
> 0)) ||
1577 (!(sinfo_flags
& (SCTP_EOF
|SCTP_ABORT
)) && (msg_len
== 0))) {
1582 /* If SCTP_ADDR_OVER is set, there must be an address
1583 * specified in msg_name.
1585 if ((sinfo_flags
& SCTP_ADDR_OVER
) && (!msg
->msg_name
)) {
1592 SCTP_DEBUG_PRINTK("About to look up association.\n");
1596 /* If a msg_name has been specified, assume this is to be used. */
1598 /* Look for a matching association on the endpoint. */
1599 asoc
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1601 /* If we could not find a matching association on the
1602 * endpoint, make sure that it is not a TCP-style
1603 * socket that already has an association or there is
1604 * no peeled-off association on another socket.
1606 if ((sctp_style(sk
, TCP
) &&
1607 sctp_sstate(sk
, ESTABLISHED
)) ||
1608 sctp_endpoint_is_peeled_off(ep
, &to
)) {
1609 err
= -EADDRNOTAVAIL
;
1614 asoc
= sctp_id2assoc(sk
, associd
);
1622 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc
);
1624 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1625 * socket that has an association in CLOSED state. This can
1626 * happen when an accepted socket has an association that is
1629 if (sctp_state(asoc
, CLOSED
) && sctp_style(sk
, TCP
)) {
1634 if (sinfo_flags
& SCTP_EOF
) {
1635 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1637 sctp_primitive_SHUTDOWN(asoc
, NULL
);
1641 if (sinfo_flags
& SCTP_ABORT
) {
1643 chunk
= sctp_make_abort_user(asoc
, msg
, msg_len
);
1649 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc
);
1650 sctp_primitive_ABORT(asoc
, chunk
);
1656 /* Do we need to create the association? */
1658 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1660 if (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
)) {
1665 /* Check for invalid stream against the stream counts,
1666 * either the default or the user specified stream counts.
1669 if (!sinit
|| (sinit
&& !sinit
->sinit_num_ostreams
)) {
1670 /* Check against the defaults. */
1671 if (sinfo
->sinfo_stream
>=
1672 sp
->initmsg
.sinit_num_ostreams
) {
1677 /* Check against the requested. */
1678 if (sinfo
->sinfo_stream
>=
1679 sinit
->sinit_num_ostreams
) {
1687 * API 3.1.2 bind() - UDP Style Syntax
1688 * If a bind() or sctp_bindx() is not called prior to a
1689 * sendmsg() call that initiates a new association, the
1690 * system picks an ephemeral port and will choose an address
1691 * set equivalent to binding with a wildcard address.
1693 if (!ep
->base
.bind_addr
.port
) {
1694 if (sctp_autobind(sk
)) {
1700 * If an unprivileged user inherits a one-to-many
1701 * style socket with open associations on a privileged
1702 * port, it MAY be permitted to accept new associations,
1703 * but it SHOULD NOT be permitted to open new
1706 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1707 !capable(CAP_NET_BIND_SERVICE
)) {
1713 scope
= sctp_scope(&to
);
1714 new_asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1720 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
, GFP_KERNEL
);
1726 /* If the SCTP_INIT ancillary data is specified, set all
1727 * the association init values accordingly.
1730 if (sinit
->sinit_num_ostreams
) {
1731 asoc
->c
.sinit_num_ostreams
=
1732 sinit
->sinit_num_ostreams
;
1734 if (sinit
->sinit_max_instreams
) {
1735 asoc
->c
.sinit_max_instreams
=
1736 sinit
->sinit_max_instreams
;
1738 if (sinit
->sinit_max_attempts
) {
1739 asoc
->max_init_attempts
1740 = sinit
->sinit_max_attempts
;
1742 if (sinit
->sinit_max_init_timeo
) {
1743 asoc
->max_init_timeo
=
1744 msecs_to_jiffies(sinit
->sinit_max_init_timeo
);
1748 /* Prime the peer's transport structures. */
1749 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
, SCTP_UNKNOWN
);
1756 /* ASSERT: we have a valid association at this point. */
1757 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1760 /* If the user didn't specify SNDRCVINFO, make up one with
1763 memset(&default_sinfo
, 0, sizeof(default_sinfo
));
1764 default_sinfo
.sinfo_stream
= asoc
->default_stream
;
1765 default_sinfo
.sinfo_flags
= asoc
->default_flags
;
1766 default_sinfo
.sinfo_ppid
= asoc
->default_ppid
;
1767 default_sinfo
.sinfo_context
= asoc
->default_context
;
1768 default_sinfo
.sinfo_timetolive
= asoc
->default_timetolive
;
1769 default_sinfo
.sinfo_assoc_id
= sctp_assoc2id(asoc
);
1770 sinfo
= &default_sinfo
;
1773 /* API 7.1.7, the sndbuf size per association bounds the
1774 * maximum size of data that can be sent in a single send call.
1776 if (msg_len
> sk
->sk_sndbuf
) {
1781 if (asoc
->pmtu_pending
)
1782 sctp_assoc_pending_pmtu(asoc
);
1784 /* If fragmentation is disabled and the message length exceeds the
1785 * association fragmentation point, return EMSGSIZE. The I-D
1786 * does not specify what this error is, but this looks like
1789 if (sctp_sk(sk
)->disable_fragments
&& (msg_len
> asoc
->frag_point
)) {
1795 /* Check for invalid stream. */
1796 if (sinfo
->sinfo_stream
>= asoc
->c
.sinit_num_ostreams
) {
1802 timeo
= sock_sndtimeo(sk
, msg
->msg_flags
& MSG_DONTWAIT
);
1803 if (!sctp_wspace(asoc
)) {
1804 err
= sctp_wait_for_sndbuf(asoc
, &timeo
, msg_len
);
1809 /* If an address is passed with the sendto/sendmsg call, it is used
1810 * to override the primary destination address in the TCP model, or
1811 * when SCTP_ADDR_OVER flag is set in the UDP model.
1813 if ((sctp_style(sk
, TCP
) && msg_name
) ||
1814 (sinfo_flags
& SCTP_ADDR_OVER
)) {
1815 chunk_tp
= sctp_assoc_lookup_paddr(asoc
, &to
);
1823 /* Auto-connect, if we aren't connected already. */
1824 if (sctp_state(asoc
, CLOSED
)) {
1825 err
= sctp_primitive_ASSOCIATE(asoc
, NULL
);
1828 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1831 /* Break the message into multiple chunks of maximum size. */
1832 datamsg
= sctp_datamsg_from_user(asoc
, sinfo
, msg
, msg_len
);
1838 /* Now send the (possibly) fragmented message. */
1839 list_for_each_entry(chunk
, &datamsg
->chunks
, frag_list
) {
1840 sctp_chunk_hold(chunk
);
1842 /* Do accounting for the write space. */
1843 sctp_set_owner_w(chunk
);
1845 chunk
->transport
= chunk_tp
;
1848 /* Send it to the lower layers. Note: all chunks
1849 * must either fail or succeed. The lower layer
1850 * works that way today. Keep it that way or this
1853 err
= sctp_primitive_SEND(asoc
, datamsg
);
1854 /* Did the lower layer accept the chunk? */
1856 sctp_datamsg_free(datamsg
);
1858 sctp_datamsg_put(datamsg
);
1860 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1867 /* If we are already past ASSOCIATE, the lower
1868 * layers are responsible for association cleanup.
1874 sctp_association_free(asoc
);
1876 sctp_release_sock(sk
);
1879 return sctp_error(sk
, msg_flags
, err
);
1886 err
= sock_error(sk
);
1896 /* This is an extended version of skb_pull() that removes the data from the
1897 * start of a skb even when data is spread across the list of skb's in the
1898 * frag_list. len specifies the total amount of data that needs to be removed.
1899 * when 'len' bytes could be removed from the skb, it returns 0.
1900 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1901 * could not be removed.
1903 static int sctp_skb_pull(struct sk_buff
*skb
, int len
)
1905 struct sk_buff
*list
;
1906 int skb_len
= skb_headlen(skb
);
1909 if (len
<= skb_len
) {
1910 __skb_pull(skb
, len
);
1914 __skb_pull(skb
, skb_len
);
1916 skb_walk_frags(skb
, list
) {
1917 rlen
= sctp_skb_pull(list
, len
);
1918 skb
->len
-= (len
-rlen
);
1919 skb
->data_len
-= (len
-rlen
);
1930 /* API 3.1.3 recvmsg() - UDP Style Syntax
1932 * ssize_t recvmsg(int socket, struct msghdr *message,
1935 * socket - the socket descriptor of the endpoint.
1936 * message - pointer to the msghdr structure which contains a single
1937 * user message and possibly some ancillary data.
1939 * See Section 5 for complete description of the data
1942 * flags - flags sent or received with the user message, see Section
1943 * 5 for complete description of the flags.
1945 static struct sk_buff
*sctp_skb_recv_datagram(struct sock
*, int, int, int *);
1947 SCTP_STATIC
int sctp_recvmsg(struct kiocb
*iocb
, struct sock
*sk
,
1948 struct msghdr
*msg
, size_t len
, int noblock
,
1949 int flags
, int *addr_len
)
1951 struct sctp_ulpevent
*event
= NULL
;
1952 struct sctp_sock
*sp
= sctp_sk(sk
);
1953 struct sk_buff
*skb
;
1958 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1959 "0x%x, %s: %p)\n", "sk", sk
, "msghdr", msg
,
1960 "len", len
, "knoblauch", noblock
,
1961 "flags", flags
, "addr_len", addr_len
);
1965 if (sctp_style(sk
, TCP
) && !sctp_sstate(sk
, ESTABLISHED
)) {
1970 skb
= sctp_skb_recv_datagram(sk
, flags
, noblock
, &err
);
1974 /* Get the total length of the skb including any skb's in the
1983 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
1985 event
= sctp_skb2event(skb
);
1990 sock_recv_ts_and_drops(msg
, sk
, skb
);
1991 if (sctp_ulpevent_is_notification(event
)) {
1992 msg
->msg_flags
|= MSG_NOTIFICATION
;
1993 sp
->pf
->event_msgname(event
, msg
->msg_name
, addr_len
);
1995 sp
->pf
->skb_msgname(skb
, msg
->msg_name
, addr_len
);
1998 /* Check if we allow SCTP_SNDRCVINFO. */
1999 if (sp
->subscribe
.sctp_data_io_event
)
2000 sctp_ulpevent_read_sndrcvinfo(event
, msg
);
2002 /* FIXME: we should be calling IP/IPv6 layers. */
2003 if (sk
->sk_protinfo
.af_inet
.cmsg_flags
)
2004 ip_cmsg_recv(msg
, skb
);
2009 /* If skb's length exceeds the user's buffer, update the skb and
2010 * push it back to the receive_queue so that the next call to
2011 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2013 if (skb_len
> copied
) {
2014 msg
->msg_flags
&= ~MSG_EOR
;
2015 if (flags
& MSG_PEEK
)
2017 sctp_skb_pull(skb
, copied
);
2018 skb_queue_head(&sk
->sk_receive_queue
, skb
);
2020 /* When only partial message is copied to the user, increase
2021 * rwnd by that amount. If all the data in the skb is read,
2022 * rwnd is updated when the event is freed.
2024 if (!sctp_ulpevent_is_notification(event
))
2025 sctp_assoc_rwnd_increase(event
->asoc
, copied
);
2027 } else if ((event
->msg_flags
& MSG_NOTIFICATION
) ||
2028 (event
->msg_flags
& MSG_EOR
))
2029 msg
->msg_flags
|= MSG_EOR
;
2031 msg
->msg_flags
&= ~MSG_EOR
;
2034 if (flags
& MSG_PEEK
) {
2035 /* Release the skb reference acquired after peeking the skb in
2036 * sctp_skb_recv_datagram().
2040 /* Free the event which includes releasing the reference to
2041 * the owner of the skb, freeing the skb and updating the
2044 sctp_ulpevent_free(event
);
2047 sctp_release_sock(sk
);
2051 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2053 * This option is a on/off flag. If enabled no SCTP message
2054 * fragmentation will be performed. Instead if a message being sent
2055 * exceeds the current PMTU size, the message will NOT be sent and
2056 * instead a error will be indicated to the user.
2058 static int sctp_setsockopt_disable_fragments(struct sock
*sk
,
2059 char __user
*optval
,
2060 unsigned int optlen
)
2064 if (optlen
< sizeof(int))
2067 if (get_user(val
, (int __user
*)optval
))
2070 sctp_sk(sk
)->disable_fragments
= (val
== 0) ? 0 : 1;
2075 static int sctp_setsockopt_events(struct sock
*sk
, char __user
*optval
,
2076 unsigned int optlen
)
2078 if (optlen
> sizeof(struct sctp_event_subscribe
))
2080 if (copy_from_user(&sctp_sk(sk
)->subscribe
, optval
, optlen
))
2085 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2087 * This socket option is applicable to the UDP-style socket only. When
2088 * set it will cause associations that are idle for more than the
2089 * specified number of seconds to automatically close. An association
2090 * being idle is defined an association that has NOT sent or received
2091 * user data. The special value of '0' indicates that no automatic
2092 * close of any associations should be performed. The option expects an
2093 * integer defining the number of seconds of idle time before an
2094 * association is closed.
2096 static int sctp_setsockopt_autoclose(struct sock
*sk
, char __user
*optval
,
2097 unsigned int optlen
)
2099 struct sctp_sock
*sp
= sctp_sk(sk
);
2101 /* Applicable to UDP-style socket only */
2102 if (sctp_style(sk
, TCP
))
2104 if (optlen
!= sizeof(int))
2106 if (copy_from_user(&sp
->autoclose
, optval
, optlen
))
2108 /* make sure it won't exceed MAX_SCHEDULE_TIMEOUT */
2109 sp
->autoclose
= min_t(long, sp
->autoclose
, MAX_SCHEDULE_TIMEOUT
/ HZ
);
2114 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2116 * Applications can enable or disable heartbeats for any peer address of
2117 * an association, modify an address's heartbeat interval, force a
2118 * heartbeat to be sent immediately, and adjust the address's maximum
2119 * number of retransmissions sent before an address is considered
2120 * unreachable. The following structure is used to access and modify an
2121 * address's parameters:
2123 * struct sctp_paddrparams {
2124 * sctp_assoc_t spp_assoc_id;
2125 * struct sockaddr_storage spp_address;
2126 * uint32_t spp_hbinterval;
2127 * uint16_t spp_pathmaxrxt;
2128 * uint32_t spp_pathmtu;
2129 * uint32_t spp_sackdelay;
2130 * uint32_t spp_flags;
2133 * spp_assoc_id - (one-to-many style socket) This is filled in the
2134 * application, and identifies the association for
2136 * spp_address - This specifies which address is of interest.
2137 * spp_hbinterval - This contains the value of the heartbeat interval,
2138 * in milliseconds. If a value of zero
2139 * is present in this field then no changes are to
2140 * be made to this parameter.
2141 * spp_pathmaxrxt - This contains the maximum number of
2142 * retransmissions before this address shall be
2143 * considered unreachable. If a value of zero
2144 * is present in this field then no changes are to
2145 * be made to this parameter.
2146 * spp_pathmtu - When Path MTU discovery is disabled the value
2147 * specified here will be the "fixed" path mtu.
2148 * Note that if the spp_address field is empty
2149 * then all associations on this address will
2150 * have this fixed path mtu set upon them.
2152 * spp_sackdelay - When delayed sack is enabled, this value specifies
2153 * the number of milliseconds that sacks will be delayed
2154 * for. This value will apply to all addresses of an
2155 * association if the spp_address field is empty. Note
2156 * also, that if delayed sack is enabled and this
2157 * value is set to 0, no change is made to the last
2158 * recorded delayed sack timer value.
2160 * spp_flags - These flags are used to control various features
2161 * on an association. The flag field may contain
2162 * zero or more of the following options.
2164 * SPP_HB_ENABLE - Enable heartbeats on the
2165 * specified address. Note that if the address
2166 * field is empty all addresses for the association
2167 * have heartbeats enabled upon them.
2169 * SPP_HB_DISABLE - Disable heartbeats on the
2170 * speicifed address. Note that if the address
2171 * field is empty all addresses for the association
2172 * will have their heartbeats disabled. Note also
2173 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2174 * mutually exclusive, only one of these two should
2175 * be specified. Enabling both fields will have
2176 * undetermined results.
2178 * SPP_HB_DEMAND - Request a user initiated heartbeat
2179 * to be made immediately.
2181 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2182 * heartbeat delayis to be set to the value of 0
2185 * SPP_PMTUD_ENABLE - This field will enable PMTU
2186 * discovery upon the specified address. Note that
2187 * if the address feild is empty then all addresses
2188 * on the association are effected.
2190 * SPP_PMTUD_DISABLE - This field will disable PMTU
2191 * discovery upon the specified address. Note that
2192 * if the address feild is empty then all addresses
2193 * on the association are effected. Not also that
2194 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2195 * exclusive. Enabling both will have undetermined
2198 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2199 * on delayed sack. The time specified in spp_sackdelay
2200 * is used to specify the sack delay for this address. Note
2201 * that if spp_address is empty then all addresses will
2202 * enable delayed sack and take on the sack delay
2203 * value specified in spp_sackdelay.
2204 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2205 * off delayed sack. If the spp_address field is blank then
2206 * delayed sack is disabled for the entire association. Note
2207 * also that this field is mutually exclusive to
2208 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2211 static int sctp_apply_peer_addr_params(struct sctp_paddrparams
*params
,
2212 struct sctp_transport
*trans
,
2213 struct sctp_association
*asoc
,
2214 struct sctp_sock
*sp
,
2217 int sackdelay_change
)
2221 if (params
->spp_flags
& SPP_HB_DEMAND
&& trans
) {
2222 error
= sctp_primitive_REQUESTHEARTBEAT (trans
->asoc
, trans
);
2227 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2228 * this field is ignored. Note also that a value of zero indicates
2229 * the current setting should be left unchanged.
2231 if (params
->spp_flags
& SPP_HB_ENABLE
) {
2233 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2234 * set. This lets us use 0 value when this flag
2237 if (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)
2238 params
->spp_hbinterval
= 0;
2240 if (params
->spp_hbinterval
||
2241 (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)) {
2244 msecs_to_jiffies(params
->spp_hbinterval
);
2247 msecs_to_jiffies(params
->spp_hbinterval
);
2249 sp
->hbinterval
= params
->spp_hbinterval
;
2256 trans
->param_flags
=
2257 (trans
->param_flags
& ~SPP_HB
) | hb_change
;
2260 (asoc
->param_flags
& ~SPP_HB
) | hb_change
;
2263 (sp
->param_flags
& ~SPP_HB
) | hb_change
;
2267 /* When Path MTU discovery is disabled the value specified here will
2268 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2269 * include the flag SPP_PMTUD_DISABLE for this field to have any
2272 if ((params
->spp_flags
& SPP_PMTUD_DISABLE
) && params
->spp_pathmtu
) {
2274 trans
->pathmtu
= params
->spp_pathmtu
;
2275 sctp_assoc_sync_pmtu(asoc
);
2277 asoc
->pathmtu
= params
->spp_pathmtu
;
2278 sctp_frag_point(asoc
, params
->spp_pathmtu
);
2280 sp
->pathmtu
= params
->spp_pathmtu
;
2286 int update
= (trans
->param_flags
& SPP_PMTUD_DISABLE
) &&
2287 (params
->spp_flags
& SPP_PMTUD_ENABLE
);
2288 trans
->param_flags
=
2289 (trans
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2291 sctp_transport_pmtu(trans
, sctp_opt2sk(sp
));
2292 sctp_assoc_sync_pmtu(asoc
);
2296 (asoc
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2299 (sp
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2303 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2304 * value of this field is ignored. Note also that a value of zero
2305 * indicates the current setting should be left unchanged.
2307 if ((params
->spp_flags
& SPP_SACKDELAY_ENABLE
) && params
->spp_sackdelay
) {
2310 msecs_to_jiffies(params
->spp_sackdelay
);
2313 msecs_to_jiffies(params
->spp_sackdelay
);
2315 sp
->sackdelay
= params
->spp_sackdelay
;
2319 if (sackdelay_change
) {
2321 trans
->param_flags
=
2322 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2326 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2330 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2335 /* Note that a value of zero indicates the current setting should be
2338 if (params
->spp_pathmaxrxt
) {
2340 trans
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2342 asoc
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2344 sp
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2351 static int sctp_setsockopt_peer_addr_params(struct sock
*sk
,
2352 char __user
*optval
,
2353 unsigned int optlen
)
2355 struct sctp_paddrparams params
;
2356 struct sctp_transport
*trans
= NULL
;
2357 struct sctp_association
*asoc
= NULL
;
2358 struct sctp_sock
*sp
= sctp_sk(sk
);
2360 int hb_change
, pmtud_change
, sackdelay_change
;
2362 if (optlen
!= sizeof(struct sctp_paddrparams
))
2365 if (copy_from_user(¶ms
, optval
, optlen
))
2368 /* Validate flags and value parameters. */
2369 hb_change
= params
.spp_flags
& SPP_HB
;
2370 pmtud_change
= params
.spp_flags
& SPP_PMTUD
;
2371 sackdelay_change
= params
.spp_flags
& SPP_SACKDELAY
;
2373 if (hb_change
== SPP_HB
||
2374 pmtud_change
== SPP_PMTUD
||
2375 sackdelay_change
== SPP_SACKDELAY
||
2376 params
.spp_sackdelay
> 500 ||
2377 (params
.spp_pathmtu
&&
2378 params
.spp_pathmtu
< SCTP_DEFAULT_MINSEGMENT
))
2381 /* If an address other than INADDR_ANY is specified, and
2382 * no transport is found, then the request is invalid.
2384 if (!sctp_is_any(sk
, ( union sctp_addr
*)¶ms
.spp_address
)) {
2385 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
2386 params
.spp_assoc_id
);
2391 /* Get association, if assoc_id != 0 and the socket is a one
2392 * to many style socket, and an association was not found, then
2393 * the id was invalid.
2395 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
2396 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
))
2399 /* Heartbeat demand can only be sent on a transport or
2400 * association, but not a socket.
2402 if (params
.spp_flags
& SPP_HB_DEMAND
&& !trans
&& !asoc
)
2405 /* Process parameters. */
2406 error
= sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2407 hb_change
, pmtud_change
,
2413 /* If changes are for association, also apply parameters to each
2416 if (!trans
&& asoc
) {
2417 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2419 sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2420 hb_change
, pmtud_change
,
2429 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2431 * This option will effect the way delayed acks are performed. This
2432 * option allows you to get or set the delayed ack time, in
2433 * milliseconds. It also allows changing the delayed ack frequency.
2434 * Changing the frequency to 1 disables the delayed sack algorithm. If
2435 * the assoc_id is 0, then this sets or gets the endpoints default
2436 * values. If the assoc_id field is non-zero, then the set or get
2437 * effects the specified association for the one to many model (the
2438 * assoc_id field is ignored by the one to one model). Note that if
2439 * sack_delay or sack_freq are 0 when setting this option, then the
2440 * current values will remain unchanged.
2442 * struct sctp_sack_info {
2443 * sctp_assoc_t sack_assoc_id;
2444 * uint32_t sack_delay;
2445 * uint32_t sack_freq;
2448 * sack_assoc_id - This parameter, indicates which association the user
2449 * is performing an action upon. Note that if this field's value is
2450 * zero then the endpoints default value is changed (effecting future
2451 * associations only).
2453 * sack_delay - This parameter contains the number of milliseconds that
2454 * the user is requesting the delayed ACK timer be set to. Note that
2455 * this value is defined in the standard to be between 200 and 500
2458 * sack_freq - This parameter contains the number of packets that must
2459 * be received before a sack is sent without waiting for the delay
2460 * timer to expire. The default value for this is 2, setting this
2461 * value to 1 will disable the delayed sack algorithm.
2464 static int sctp_setsockopt_delayed_ack(struct sock
*sk
,
2465 char __user
*optval
, unsigned int optlen
)
2467 struct sctp_sack_info params
;
2468 struct sctp_transport
*trans
= NULL
;
2469 struct sctp_association
*asoc
= NULL
;
2470 struct sctp_sock
*sp
= sctp_sk(sk
);
2472 if (optlen
== sizeof(struct sctp_sack_info
)) {
2473 if (copy_from_user(¶ms
, optval
, optlen
))
2476 if (params
.sack_delay
== 0 && params
.sack_freq
== 0)
2478 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
2479 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2480 pr_warn("Use struct sctp_sack_info instead\n");
2481 if (copy_from_user(¶ms
, optval
, optlen
))
2484 if (params
.sack_delay
== 0)
2485 params
.sack_freq
= 1;
2487 params
.sack_freq
= 0;
2491 /* Validate value parameter. */
2492 if (params
.sack_delay
> 500)
2495 /* Get association, if sack_assoc_id != 0 and the socket is a one
2496 * to many style socket, and an association was not found, then
2497 * the id was invalid.
2499 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
2500 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
2503 if (params
.sack_delay
) {
2506 msecs_to_jiffies(params
.sack_delay
);
2508 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2509 SPP_SACKDELAY_ENABLE
;
2511 sp
->sackdelay
= params
.sack_delay
;
2513 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2514 SPP_SACKDELAY_ENABLE
;
2518 if (params
.sack_freq
== 1) {
2521 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2522 SPP_SACKDELAY_DISABLE
;
2525 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2526 SPP_SACKDELAY_DISABLE
;
2528 } else if (params
.sack_freq
> 1) {
2530 asoc
->sackfreq
= params
.sack_freq
;
2532 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2533 SPP_SACKDELAY_ENABLE
;
2535 sp
->sackfreq
= params
.sack_freq
;
2537 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2538 SPP_SACKDELAY_ENABLE
;
2542 /* If change is for association, also apply to each transport. */
2544 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2546 if (params
.sack_delay
) {
2548 msecs_to_jiffies(params
.sack_delay
);
2549 trans
->param_flags
=
2550 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2551 SPP_SACKDELAY_ENABLE
;
2553 if (params
.sack_freq
== 1) {
2554 trans
->param_flags
=
2555 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2556 SPP_SACKDELAY_DISABLE
;
2557 } else if (params
.sack_freq
> 1) {
2558 trans
->sackfreq
= params
.sack_freq
;
2559 trans
->param_flags
=
2560 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2561 SPP_SACKDELAY_ENABLE
;
2569 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2571 * Applications can specify protocol parameters for the default association
2572 * initialization. The option name argument to setsockopt() and getsockopt()
2575 * Setting initialization parameters is effective only on an unconnected
2576 * socket (for UDP-style sockets only future associations are effected
2577 * by the change). With TCP-style sockets, this option is inherited by
2578 * sockets derived from a listener socket.
2580 static int sctp_setsockopt_initmsg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2582 struct sctp_initmsg sinit
;
2583 struct sctp_sock
*sp
= sctp_sk(sk
);
2585 if (optlen
!= sizeof(struct sctp_initmsg
))
2587 if (copy_from_user(&sinit
, optval
, optlen
))
2590 if (sinit
.sinit_num_ostreams
)
2591 sp
->initmsg
.sinit_num_ostreams
= sinit
.sinit_num_ostreams
;
2592 if (sinit
.sinit_max_instreams
)
2593 sp
->initmsg
.sinit_max_instreams
= sinit
.sinit_max_instreams
;
2594 if (sinit
.sinit_max_attempts
)
2595 sp
->initmsg
.sinit_max_attempts
= sinit
.sinit_max_attempts
;
2596 if (sinit
.sinit_max_init_timeo
)
2597 sp
->initmsg
.sinit_max_init_timeo
= sinit
.sinit_max_init_timeo
;
2603 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2605 * Applications that wish to use the sendto() system call may wish to
2606 * specify a default set of parameters that would normally be supplied
2607 * through the inclusion of ancillary data. This socket option allows
2608 * such an application to set the default sctp_sndrcvinfo structure.
2609 * The application that wishes to use this socket option simply passes
2610 * in to this call the sctp_sndrcvinfo structure defined in Section
2611 * 5.2.2) The input parameters accepted by this call include
2612 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2613 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2614 * to this call if the caller is using the UDP model.
2616 static int sctp_setsockopt_default_send_param(struct sock
*sk
,
2617 char __user
*optval
,
2618 unsigned int optlen
)
2620 struct sctp_sndrcvinfo info
;
2621 struct sctp_association
*asoc
;
2622 struct sctp_sock
*sp
= sctp_sk(sk
);
2624 if (optlen
!= sizeof(struct sctp_sndrcvinfo
))
2626 if (copy_from_user(&info
, optval
, optlen
))
2629 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
2630 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
2634 asoc
->default_stream
= info
.sinfo_stream
;
2635 asoc
->default_flags
= info
.sinfo_flags
;
2636 asoc
->default_ppid
= info
.sinfo_ppid
;
2637 asoc
->default_context
= info
.sinfo_context
;
2638 asoc
->default_timetolive
= info
.sinfo_timetolive
;
2640 sp
->default_stream
= info
.sinfo_stream
;
2641 sp
->default_flags
= info
.sinfo_flags
;
2642 sp
->default_ppid
= info
.sinfo_ppid
;
2643 sp
->default_context
= info
.sinfo_context
;
2644 sp
->default_timetolive
= info
.sinfo_timetolive
;
2650 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2652 * Requests that the local SCTP stack use the enclosed peer address as
2653 * the association primary. The enclosed address must be one of the
2654 * association peer's addresses.
2656 static int sctp_setsockopt_primary_addr(struct sock
*sk
, char __user
*optval
,
2657 unsigned int optlen
)
2659 struct sctp_prim prim
;
2660 struct sctp_transport
*trans
;
2662 if (optlen
!= sizeof(struct sctp_prim
))
2665 if (copy_from_user(&prim
, optval
, sizeof(struct sctp_prim
)))
2668 trans
= sctp_addr_id2transport(sk
, &prim
.ssp_addr
, prim
.ssp_assoc_id
);
2672 sctp_assoc_set_primary(trans
->asoc
, trans
);
2678 * 7.1.5 SCTP_NODELAY
2680 * Turn on/off any Nagle-like algorithm. This means that packets are
2681 * generally sent as soon as possible and no unnecessary delays are
2682 * introduced, at the cost of more packets in the network. Expects an
2683 * integer boolean flag.
2685 static int sctp_setsockopt_nodelay(struct sock
*sk
, char __user
*optval
,
2686 unsigned int optlen
)
2690 if (optlen
< sizeof(int))
2692 if (get_user(val
, (int __user
*)optval
))
2695 sctp_sk(sk
)->nodelay
= (val
== 0) ? 0 : 1;
2701 * 7.1.1 SCTP_RTOINFO
2703 * The protocol parameters used to initialize and bound retransmission
2704 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2705 * and modify these parameters.
2706 * All parameters are time values, in milliseconds. A value of 0, when
2707 * modifying the parameters, indicates that the current value should not
2711 static int sctp_setsockopt_rtoinfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2713 struct sctp_rtoinfo rtoinfo
;
2714 struct sctp_association
*asoc
;
2716 if (optlen
!= sizeof (struct sctp_rtoinfo
))
2719 if (copy_from_user(&rtoinfo
, optval
, optlen
))
2722 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
2724 /* Set the values to the specific association */
2725 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
2729 if (rtoinfo
.srto_initial
!= 0)
2731 msecs_to_jiffies(rtoinfo
.srto_initial
);
2732 if (rtoinfo
.srto_max
!= 0)
2733 asoc
->rto_max
= msecs_to_jiffies(rtoinfo
.srto_max
);
2734 if (rtoinfo
.srto_min
!= 0)
2735 asoc
->rto_min
= msecs_to_jiffies(rtoinfo
.srto_min
);
2737 /* If there is no association or the association-id = 0
2738 * set the values to the endpoint.
2740 struct sctp_sock
*sp
= sctp_sk(sk
);
2742 if (rtoinfo
.srto_initial
!= 0)
2743 sp
->rtoinfo
.srto_initial
= rtoinfo
.srto_initial
;
2744 if (rtoinfo
.srto_max
!= 0)
2745 sp
->rtoinfo
.srto_max
= rtoinfo
.srto_max
;
2746 if (rtoinfo
.srto_min
!= 0)
2747 sp
->rtoinfo
.srto_min
= rtoinfo
.srto_min
;
2755 * 7.1.2 SCTP_ASSOCINFO
2757 * This option is used to tune the maximum retransmission attempts
2758 * of the association.
2759 * Returns an error if the new association retransmission value is
2760 * greater than the sum of the retransmission value of the peer.
2761 * See [SCTP] for more information.
2764 static int sctp_setsockopt_associnfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2767 struct sctp_assocparams assocparams
;
2768 struct sctp_association
*asoc
;
2770 if (optlen
!= sizeof(struct sctp_assocparams
))
2772 if (copy_from_user(&assocparams
, optval
, optlen
))
2775 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
2777 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
2780 /* Set the values to the specific association */
2782 if (assocparams
.sasoc_asocmaxrxt
!= 0) {
2785 struct sctp_transport
*peer_addr
;
2787 list_for_each_entry(peer_addr
, &asoc
->peer
.transport_addr_list
,
2789 path_sum
+= peer_addr
->pathmaxrxt
;
2793 /* Only validate asocmaxrxt if we have more than
2794 * one path/transport. We do this because path
2795 * retransmissions are only counted when we have more
2799 assocparams
.sasoc_asocmaxrxt
> path_sum
)
2802 asoc
->max_retrans
= assocparams
.sasoc_asocmaxrxt
;
2805 if (assocparams
.sasoc_cookie_life
!= 0) {
2806 asoc
->cookie_life
.tv_sec
=
2807 assocparams
.sasoc_cookie_life
/ 1000;
2808 asoc
->cookie_life
.tv_usec
=
2809 (assocparams
.sasoc_cookie_life
% 1000)
2813 /* Set the values to the endpoint */
2814 struct sctp_sock
*sp
= sctp_sk(sk
);
2816 if (assocparams
.sasoc_asocmaxrxt
!= 0)
2817 sp
->assocparams
.sasoc_asocmaxrxt
=
2818 assocparams
.sasoc_asocmaxrxt
;
2819 if (assocparams
.sasoc_cookie_life
!= 0)
2820 sp
->assocparams
.sasoc_cookie_life
=
2821 assocparams
.sasoc_cookie_life
;
2827 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2829 * This socket option is a boolean flag which turns on or off mapped V4
2830 * addresses. If this option is turned on and the socket is type
2831 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2832 * If this option is turned off, then no mapping will be done of V4
2833 * addresses and a user will receive both PF_INET6 and PF_INET type
2834 * addresses on the socket.
2836 static int sctp_setsockopt_mappedv4(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2839 struct sctp_sock
*sp
= sctp_sk(sk
);
2841 if (optlen
< sizeof(int))
2843 if (get_user(val
, (int __user
*)optval
))
2854 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2855 * This option will get or set the maximum size to put in any outgoing
2856 * SCTP DATA chunk. If a message is larger than this size it will be
2857 * fragmented by SCTP into the specified size. Note that the underlying
2858 * SCTP implementation may fragment into smaller sized chunks when the
2859 * PMTU of the underlying association is smaller than the value set by
2860 * the user. The default value for this option is '0' which indicates
2861 * the user is NOT limiting fragmentation and only the PMTU will effect
2862 * SCTP's choice of DATA chunk size. Note also that values set larger
2863 * than the maximum size of an IP datagram will effectively let SCTP
2864 * control fragmentation (i.e. the same as setting this option to 0).
2866 * The following structure is used to access and modify this parameter:
2868 * struct sctp_assoc_value {
2869 * sctp_assoc_t assoc_id;
2870 * uint32_t assoc_value;
2873 * assoc_id: This parameter is ignored for one-to-one style sockets.
2874 * For one-to-many style sockets this parameter indicates which
2875 * association the user is performing an action upon. Note that if
2876 * this field's value is zero then the endpoints default value is
2877 * changed (effecting future associations only).
2878 * assoc_value: This parameter specifies the maximum size in bytes.
2880 static int sctp_setsockopt_maxseg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2882 struct sctp_assoc_value params
;
2883 struct sctp_association
*asoc
;
2884 struct sctp_sock
*sp
= sctp_sk(sk
);
2887 if (optlen
== sizeof(int)) {
2888 pr_warn("Use of int in maxseg socket option deprecated\n");
2889 pr_warn("Use struct sctp_assoc_value instead\n");
2890 if (copy_from_user(&val
, optval
, optlen
))
2892 params
.assoc_id
= 0;
2893 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
2894 if (copy_from_user(¶ms
, optval
, optlen
))
2896 val
= params
.assoc_value
;
2900 if ((val
!= 0) && ((val
< 8) || (val
> SCTP_MAX_CHUNK_LEN
)))
2903 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
2904 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
2909 val
= asoc
->pathmtu
;
2910 val
-= sp
->pf
->af
->net_header_len
;
2911 val
-= sizeof(struct sctphdr
) +
2912 sizeof(struct sctp_data_chunk
);
2914 asoc
->user_frag
= val
;
2915 asoc
->frag_point
= sctp_frag_point(asoc
, asoc
->pathmtu
);
2917 sp
->user_frag
= val
;
2925 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2927 * Requests that the peer mark the enclosed address as the association
2928 * primary. The enclosed address must be one of the association's
2929 * locally bound addresses. The following structure is used to make a
2930 * set primary request:
2932 static int sctp_setsockopt_peer_primary_addr(struct sock
*sk
, char __user
*optval
,
2933 unsigned int optlen
)
2935 struct sctp_sock
*sp
;
2936 struct sctp_association
*asoc
= NULL
;
2937 struct sctp_setpeerprim prim
;
2938 struct sctp_chunk
*chunk
;
2944 if (!sctp_addip_enable
)
2947 if (optlen
!= sizeof(struct sctp_setpeerprim
))
2950 if (copy_from_user(&prim
, optval
, optlen
))
2953 asoc
= sctp_id2assoc(sk
, prim
.sspp_assoc_id
);
2957 if (!asoc
->peer
.asconf_capable
)
2960 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_SET_PRIMARY
)
2963 if (!sctp_state(asoc
, ESTABLISHED
))
2966 af
= sctp_get_af_specific(prim
.sspp_addr
.ss_family
);
2970 if (!af
->addr_valid((union sctp_addr
*)&prim
.sspp_addr
, sp
, NULL
))
2971 return -EADDRNOTAVAIL
;
2973 if (!sctp_assoc_lookup_laddr(asoc
, (union sctp_addr
*)&prim
.sspp_addr
))
2974 return -EADDRNOTAVAIL
;
2976 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2977 chunk
= sctp_make_asconf_set_prim(asoc
,
2978 (union sctp_addr
*)&prim
.sspp_addr
);
2982 err
= sctp_send_asconf(asoc
, chunk
);
2984 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2989 static int sctp_setsockopt_adaptation_layer(struct sock
*sk
, char __user
*optval
,
2990 unsigned int optlen
)
2992 struct sctp_setadaptation adaptation
;
2994 if (optlen
!= sizeof(struct sctp_setadaptation
))
2996 if (copy_from_user(&adaptation
, optval
, optlen
))
2999 sctp_sk(sk
)->adaptation_ind
= adaptation
.ssb_adaptation_ind
;
3005 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3007 * The context field in the sctp_sndrcvinfo structure is normally only
3008 * used when a failed message is retrieved holding the value that was
3009 * sent down on the actual send call. This option allows the setting of
3010 * a default context on an association basis that will be received on
3011 * reading messages from the peer. This is especially helpful in the
3012 * one-2-many model for an application to keep some reference to an
3013 * internal state machine that is processing messages on the
3014 * association. Note that the setting of this value only effects
3015 * received messages from the peer and does not effect the value that is
3016 * saved with outbound messages.
3018 static int sctp_setsockopt_context(struct sock
*sk
, char __user
*optval
,
3019 unsigned int optlen
)
3021 struct sctp_assoc_value params
;
3022 struct sctp_sock
*sp
;
3023 struct sctp_association
*asoc
;
3025 if (optlen
!= sizeof(struct sctp_assoc_value
))
3027 if (copy_from_user(¶ms
, optval
, optlen
))
3032 if (params
.assoc_id
!= 0) {
3033 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3036 asoc
->default_rcv_context
= params
.assoc_value
;
3038 sp
->default_rcv_context
= params
.assoc_value
;
3045 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3047 * This options will at a minimum specify if the implementation is doing
3048 * fragmented interleave. Fragmented interleave, for a one to many
3049 * socket, is when subsequent calls to receive a message may return
3050 * parts of messages from different associations. Some implementations
3051 * may allow you to turn this value on or off. If so, when turned off,
3052 * no fragment interleave will occur (which will cause a head of line
3053 * blocking amongst multiple associations sharing the same one to many
3054 * socket). When this option is turned on, then each receive call may
3055 * come from a different association (thus the user must receive data
3056 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3057 * association each receive belongs to.
3059 * This option takes a boolean value. A non-zero value indicates that
3060 * fragmented interleave is on. A value of zero indicates that
3061 * fragmented interleave is off.
3063 * Note that it is important that an implementation that allows this
3064 * option to be turned on, have it off by default. Otherwise an unaware
3065 * application using the one to many model may become confused and act
3068 static int sctp_setsockopt_fragment_interleave(struct sock
*sk
,
3069 char __user
*optval
,
3070 unsigned int optlen
)
3074 if (optlen
!= sizeof(int))
3076 if (get_user(val
, (int __user
*)optval
))
3079 sctp_sk(sk
)->frag_interleave
= (val
== 0) ? 0 : 1;
3085 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3086 * (SCTP_PARTIAL_DELIVERY_POINT)
3088 * This option will set or get the SCTP partial delivery point. This
3089 * point is the size of a message where the partial delivery API will be
3090 * invoked to help free up rwnd space for the peer. Setting this to a
3091 * lower value will cause partial deliveries to happen more often. The
3092 * calls argument is an integer that sets or gets the partial delivery
3093 * point. Note also that the call will fail if the user attempts to set
3094 * this value larger than the socket receive buffer size.
3096 * Note that any single message having a length smaller than or equal to
3097 * the SCTP partial delivery point will be delivered in one single read
3098 * call as long as the user provided buffer is large enough to hold the
3101 static int sctp_setsockopt_partial_delivery_point(struct sock
*sk
,
3102 char __user
*optval
,
3103 unsigned int optlen
)
3107 if (optlen
!= sizeof(u32
))
3109 if (get_user(val
, (int __user
*)optval
))
3112 /* Note: We double the receive buffer from what the user sets
3113 * it to be, also initial rwnd is based on rcvbuf/2.
3115 if (val
> (sk
->sk_rcvbuf
>> 1))
3118 sctp_sk(sk
)->pd_point
= val
;
3120 return 0; /* is this the right error code? */
3124 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3126 * This option will allow a user to change the maximum burst of packets
3127 * that can be emitted by this association. Note that the default value
3128 * is 4, and some implementations may restrict this setting so that it
3129 * can only be lowered.
3131 * NOTE: This text doesn't seem right. Do this on a socket basis with
3132 * future associations inheriting the socket value.
3134 static int sctp_setsockopt_maxburst(struct sock
*sk
,
3135 char __user
*optval
,
3136 unsigned int optlen
)
3138 struct sctp_assoc_value params
;
3139 struct sctp_sock
*sp
;
3140 struct sctp_association
*asoc
;
3144 if (optlen
== sizeof(int)) {
3145 pr_warn("Use of int in max_burst socket option deprecated\n");
3146 pr_warn("Use struct sctp_assoc_value instead\n");
3147 if (copy_from_user(&val
, optval
, optlen
))
3149 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
3150 if (copy_from_user(¶ms
, optval
, optlen
))
3152 val
= params
.assoc_value
;
3153 assoc_id
= params
.assoc_id
;
3159 if (assoc_id
!= 0) {
3160 asoc
= sctp_id2assoc(sk
, assoc_id
);
3163 asoc
->max_burst
= val
;
3165 sp
->max_burst
= val
;
3171 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3173 * This set option adds a chunk type that the user is requesting to be
3174 * received only in an authenticated way. Changes to the list of chunks
3175 * will only effect future associations on the socket.
3177 static int sctp_setsockopt_auth_chunk(struct sock
*sk
,
3178 char __user
*optval
,
3179 unsigned int optlen
)
3181 struct sctp_authchunk val
;
3183 if (!sctp_auth_enable
)
3186 if (optlen
!= sizeof(struct sctp_authchunk
))
3188 if (copy_from_user(&val
, optval
, optlen
))
3191 switch (val
.sauth_chunk
) {
3193 case SCTP_CID_INIT_ACK
:
3194 case SCTP_CID_SHUTDOWN_COMPLETE
:
3199 /* add this chunk id to the endpoint */
3200 return sctp_auth_ep_add_chunkid(sctp_sk(sk
)->ep
, val
.sauth_chunk
);
3204 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3206 * This option gets or sets the list of HMAC algorithms that the local
3207 * endpoint requires the peer to use.
3209 static int sctp_setsockopt_hmac_ident(struct sock
*sk
,
3210 char __user
*optval
,
3211 unsigned int optlen
)
3213 struct sctp_hmacalgo
*hmacs
;
3217 if (!sctp_auth_enable
)
3220 if (optlen
< sizeof(struct sctp_hmacalgo
))
3223 hmacs
= memdup_user(optval
, optlen
);
3225 return PTR_ERR(hmacs
);
3227 idents
= hmacs
->shmac_num_idents
;
3228 if (idents
== 0 || idents
> SCTP_AUTH_NUM_HMACS
||
3229 (idents
* sizeof(u16
)) > (optlen
- sizeof(struct sctp_hmacalgo
))) {
3234 err
= sctp_auth_ep_set_hmacs(sctp_sk(sk
)->ep
, hmacs
);
3241 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3243 * This option will set a shared secret key which is used to build an
3244 * association shared key.
3246 static int sctp_setsockopt_auth_key(struct sock
*sk
,
3247 char __user
*optval
,
3248 unsigned int optlen
)
3250 struct sctp_authkey
*authkey
;
3251 struct sctp_association
*asoc
;
3254 if (!sctp_auth_enable
)
3257 if (optlen
<= sizeof(struct sctp_authkey
))
3260 authkey
= memdup_user(optval
, optlen
);
3261 if (IS_ERR(authkey
))
3262 return PTR_ERR(authkey
);
3264 if (authkey
->sca_keylength
> optlen
- sizeof(struct sctp_authkey
)) {
3269 asoc
= sctp_id2assoc(sk
, authkey
->sca_assoc_id
);
3270 if (!asoc
&& authkey
->sca_assoc_id
&& sctp_style(sk
, UDP
)) {
3275 ret
= sctp_auth_set_key(sctp_sk(sk
)->ep
, asoc
, authkey
);
3282 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3284 * This option will get or set the active shared key to be used to build
3285 * the association shared key.
3287 static int sctp_setsockopt_active_key(struct sock
*sk
,
3288 char __user
*optval
,
3289 unsigned int optlen
)
3291 struct sctp_authkeyid val
;
3292 struct sctp_association
*asoc
;
3294 if (!sctp_auth_enable
)
3297 if (optlen
!= sizeof(struct sctp_authkeyid
))
3299 if (copy_from_user(&val
, optval
, optlen
))
3302 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3303 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3306 return sctp_auth_set_active_key(sctp_sk(sk
)->ep
, asoc
,
3307 val
.scact_keynumber
);
3311 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3313 * This set option will delete a shared secret key from use.
3315 static int sctp_setsockopt_del_key(struct sock
*sk
,
3316 char __user
*optval
,
3317 unsigned int optlen
)
3319 struct sctp_authkeyid val
;
3320 struct sctp_association
*asoc
;
3322 if (!sctp_auth_enable
)
3325 if (optlen
!= sizeof(struct sctp_authkeyid
))
3327 if (copy_from_user(&val
, optval
, optlen
))
3330 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3331 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3334 return sctp_auth_del_key_id(sctp_sk(sk
)->ep
, asoc
,
3335 val
.scact_keynumber
);
3340 /* API 6.2 setsockopt(), getsockopt()
3342 * Applications use setsockopt() and getsockopt() to set or retrieve
3343 * socket options. Socket options are used to change the default
3344 * behavior of sockets calls. They are described in Section 7.
3348 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3349 * int __user *optlen);
3350 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3353 * sd - the socket descript.
3354 * level - set to IPPROTO_SCTP for all SCTP options.
3355 * optname - the option name.
3356 * optval - the buffer to store the value of the option.
3357 * optlen - the size of the buffer.
3359 SCTP_STATIC
int sctp_setsockopt(struct sock
*sk
, int level
, int optname
,
3360 char __user
*optval
, unsigned int optlen
)
3364 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3367 /* I can hardly begin to describe how wrong this is. This is
3368 * so broken as to be worse than useless. The API draft
3369 * REALLY is NOT helpful here... I am not convinced that the
3370 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3371 * are at all well-founded.
3373 if (level
!= SOL_SCTP
) {
3374 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
3375 retval
= af
->setsockopt(sk
, level
, optname
, optval
, optlen
);
3382 case SCTP_SOCKOPT_BINDX_ADD
:
3383 /* 'optlen' is the size of the addresses buffer. */
3384 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3385 optlen
, SCTP_BINDX_ADD_ADDR
);
3388 case SCTP_SOCKOPT_BINDX_REM
:
3389 /* 'optlen' is the size of the addresses buffer. */
3390 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3391 optlen
, SCTP_BINDX_REM_ADDR
);
3394 case SCTP_SOCKOPT_CONNECTX_OLD
:
3395 /* 'optlen' is the size of the addresses buffer. */
3396 retval
= sctp_setsockopt_connectx_old(sk
,
3397 (struct sockaddr __user
*)optval
,
3401 case SCTP_SOCKOPT_CONNECTX
:
3402 /* 'optlen' is the size of the addresses buffer. */
3403 retval
= sctp_setsockopt_connectx(sk
,
3404 (struct sockaddr __user
*)optval
,
3408 case SCTP_DISABLE_FRAGMENTS
:
3409 retval
= sctp_setsockopt_disable_fragments(sk
, optval
, optlen
);
3413 retval
= sctp_setsockopt_events(sk
, optval
, optlen
);
3416 case SCTP_AUTOCLOSE
:
3417 retval
= sctp_setsockopt_autoclose(sk
, optval
, optlen
);
3420 case SCTP_PEER_ADDR_PARAMS
:
3421 retval
= sctp_setsockopt_peer_addr_params(sk
, optval
, optlen
);
3424 case SCTP_DELAYED_SACK
:
3425 retval
= sctp_setsockopt_delayed_ack(sk
, optval
, optlen
);
3427 case SCTP_PARTIAL_DELIVERY_POINT
:
3428 retval
= sctp_setsockopt_partial_delivery_point(sk
, optval
, optlen
);
3432 retval
= sctp_setsockopt_initmsg(sk
, optval
, optlen
);
3434 case SCTP_DEFAULT_SEND_PARAM
:
3435 retval
= sctp_setsockopt_default_send_param(sk
, optval
,
3438 case SCTP_PRIMARY_ADDR
:
3439 retval
= sctp_setsockopt_primary_addr(sk
, optval
, optlen
);
3441 case SCTP_SET_PEER_PRIMARY_ADDR
:
3442 retval
= sctp_setsockopt_peer_primary_addr(sk
, optval
, optlen
);
3445 retval
= sctp_setsockopt_nodelay(sk
, optval
, optlen
);
3448 retval
= sctp_setsockopt_rtoinfo(sk
, optval
, optlen
);
3450 case SCTP_ASSOCINFO
:
3451 retval
= sctp_setsockopt_associnfo(sk
, optval
, optlen
);
3453 case SCTP_I_WANT_MAPPED_V4_ADDR
:
3454 retval
= sctp_setsockopt_mappedv4(sk
, optval
, optlen
);
3457 retval
= sctp_setsockopt_maxseg(sk
, optval
, optlen
);
3459 case SCTP_ADAPTATION_LAYER
:
3460 retval
= sctp_setsockopt_adaptation_layer(sk
, optval
, optlen
);
3463 retval
= sctp_setsockopt_context(sk
, optval
, optlen
);
3465 case SCTP_FRAGMENT_INTERLEAVE
:
3466 retval
= sctp_setsockopt_fragment_interleave(sk
, optval
, optlen
);
3468 case SCTP_MAX_BURST
:
3469 retval
= sctp_setsockopt_maxburst(sk
, optval
, optlen
);
3471 case SCTP_AUTH_CHUNK
:
3472 retval
= sctp_setsockopt_auth_chunk(sk
, optval
, optlen
);
3474 case SCTP_HMAC_IDENT
:
3475 retval
= sctp_setsockopt_hmac_ident(sk
, optval
, optlen
);
3478 retval
= sctp_setsockopt_auth_key(sk
, optval
, optlen
);
3480 case SCTP_AUTH_ACTIVE_KEY
:
3481 retval
= sctp_setsockopt_active_key(sk
, optval
, optlen
);
3483 case SCTP_AUTH_DELETE_KEY
:
3484 retval
= sctp_setsockopt_del_key(sk
, optval
, optlen
);
3487 retval
= -ENOPROTOOPT
;
3491 sctp_release_sock(sk
);
3497 /* API 3.1.6 connect() - UDP Style Syntax
3499 * An application may use the connect() call in the UDP model to initiate an
3500 * association without sending data.
3504 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3506 * sd: the socket descriptor to have a new association added to.
3508 * nam: the address structure (either struct sockaddr_in or struct
3509 * sockaddr_in6 defined in RFC2553 [7]).
3511 * len: the size of the address.
3513 SCTP_STATIC
int sctp_connect(struct sock
*sk
, struct sockaddr
*addr
,
3521 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3522 __func__
, sk
, addr
, addr_len
);
3524 /* Validate addr_len before calling common connect/connectx routine. */
3525 af
= sctp_get_af_specific(addr
->sa_family
);
3526 if (!af
|| addr_len
< af
->sockaddr_len
) {
3529 /* Pass correct addr len to common routine (so it knows there
3530 * is only one address being passed.
3532 err
= __sctp_connect(sk
, addr
, af
->sockaddr_len
, NULL
);
3535 sctp_release_sock(sk
);
3539 /* FIXME: Write comments. */
3540 SCTP_STATIC
int sctp_disconnect(struct sock
*sk
, int flags
)
3542 return -EOPNOTSUPP
; /* STUB */
3545 /* 4.1.4 accept() - TCP Style Syntax
3547 * Applications use accept() call to remove an established SCTP
3548 * association from the accept queue of the endpoint. A new socket
3549 * descriptor will be returned from accept() to represent the newly
3550 * formed association.
3552 SCTP_STATIC
struct sock
*sctp_accept(struct sock
*sk
, int flags
, int *err
)
3554 struct sctp_sock
*sp
;
3555 struct sctp_endpoint
*ep
;
3556 struct sock
*newsk
= NULL
;
3557 struct sctp_association
*asoc
;
3566 if (!sctp_style(sk
, TCP
)) {
3567 error
= -EOPNOTSUPP
;
3571 if (!sctp_sstate(sk
, LISTENING
)) {
3576 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
3578 error
= sctp_wait_for_accept(sk
, timeo
);
3582 /* We treat the list of associations on the endpoint as the accept
3583 * queue and pick the first association on the list.
3585 asoc
= list_entry(ep
->asocs
.next
, struct sctp_association
, asocs
);
3587 newsk
= sp
->pf
->create_accept_sk(sk
, asoc
);
3593 /* Populate the fields of the newsk from the oldsk and migrate the
3594 * asoc to the newsk.
3596 sctp_sock_migrate(sk
, newsk
, asoc
, SCTP_SOCKET_TCP
);
3599 sctp_release_sock(sk
);
3604 /* The SCTP ioctl handler. */
3605 SCTP_STATIC
int sctp_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
3612 * SEQPACKET-style sockets in LISTENING state are valid, for
3613 * SCTP, so only discard TCP-style sockets in LISTENING state.
3615 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
3620 struct sk_buff
*skb
;
3621 unsigned int amount
= 0;
3623 skb
= skb_peek(&sk
->sk_receive_queue
);
3626 * We will only return the amount of this packet since
3627 * that is all that will be read.
3631 rc
= put_user(amount
, (int __user
*)arg
);
3639 sctp_release_sock(sk
);
3643 /* This is the function which gets called during socket creation to
3644 * initialized the SCTP-specific portion of the sock.
3645 * The sock structure should already be zero-filled memory.
3647 SCTP_STATIC
int sctp_init_sock(struct sock
*sk
)
3649 struct sctp_endpoint
*ep
;
3650 struct sctp_sock
*sp
;
3652 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk
);
3656 /* Initialize the SCTP per socket area. */
3657 switch (sk
->sk_type
) {
3658 case SOCK_SEQPACKET
:
3659 sp
->type
= SCTP_SOCKET_UDP
;
3662 sp
->type
= SCTP_SOCKET_TCP
;
3665 return -ESOCKTNOSUPPORT
;
3668 /* Initialize default send parameters. These parameters can be
3669 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3671 sp
->default_stream
= 0;
3672 sp
->default_ppid
= 0;
3673 sp
->default_flags
= 0;
3674 sp
->default_context
= 0;
3675 sp
->default_timetolive
= 0;
3677 sp
->default_rcv_context
= 0;
3678 sp
->max_burst
= sctp_max_burst
;
3680 /* Initialize default setup parameters. These parameters
3681 * can be modified with the SCTP_INITMSG socket option or
3682 * overridden by the SCTP_INIT CMSG.
3684 sp
->initmsg
.sinit_num_ostreams
= sctp_max_outstreams
;
3685 sp
->initmsg
.sinit_max_instreams
= sctp_max_instreams
;
3686 sp
->initmsg
.sinit_max_attempts
= sctp_max_retrans_init
;
3687 sp
->initmsg
.sinit_max_init_timeo
= sctp_rto_max
;
3689 /* Initialize default RTO related parameters. These parameters can
3690 * be modified for with the SCTP_RTOINFO socket option.
3692 sp
->rtoinfo
.srto_initial
= sctp_rto_initial
;
3693 sp
->rtoinfo
.srto_max
= sctp_rto_max
;
3694 sp
->rtoinfo
.srto_min
= sctp_rto_min
;
3696 /* Initialize default association related parameters. These parameters
3697 * can be modified with the SCTP_ASSOCINFO socket option.
3699 sp
->assocparams
.sasoc_asocmaxrxt
= sctp_max_retrans_association
;
3700 sp
->assocparams
.sasoc_number_peer_destinations
= 0;
3701 sp
->assocparams
.sasoc_peer_rwnd
= 0;
3702 sp
->assocparams
.sasoc_local_rwnd
= 0;
3703 sp
->assocparams
.sasoc_cookie_life
= sctp_valid_cookie_life
;
3705 /* Initialize default event subscriptions. By default, all the
3708 memset(&sp
->subscribe
, 0, sizeof(struct sctp_event_subscribe
));
3710 /* Default Peer Address Parameters. These defaults can
3711 * be modified via SCTP_PEER_ADDR_PARAMS
3713 sp
->hbinterval
= sctp_hb_interval
;
3714 sp
->pathmaxrxt
= sctp_max_retrans_path
;
3715 sp
->pathmtu
= 0; // allow default discovery
3716 sp
->sackdelay
= sctp_sack_timeout
;
3718 sp
->param_flags
= SPP_HB_ENABLE
|
3720 SPP_SACKDELAY_ENABLE
;
3722 /* If enabled no SCTP message fragmentation will be performed.
3723 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3725 sp
->disable_fragments
= 0;
3727 /* Enable Nagle algorithm by default. */
3730 /* Enable by default. */
3733 /* Auto-close idle associations after the configured
3734 * number of seconds. A value of 0 disables this
3735 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3736 * for UDP-style sockets only.
3740 /* User specified fragmentation limit. */
3743 sp
->adaptation_ind
= 0;
3745 sp
->pf
= sctp_get_pf_specific(sk
->sk_family
);
3747 /* Control variables for partial data delivery. */
3748 atomic_set(&sp
->pd_mode
, 0);
3749 skb_queue_head_init(&sp
->pd_lobby
);
3750 sp
->frag_interleave
= 0;
3752 /* Create a per socket endpoint structure. Even if we
3753 * change the data structure relationships, this may still
3754 * be useful for storing pre-connect address information.
3756 ep
= sctp_endpoint_new(sk
, GFP_KERNEL
);
3763 SCTP_DBG_OBJCNT_INC(sock
);
3766 percpu_counter_inc(&sctp_sockets_allocated
);
3767 sock_prot_inuse_add(sock_net(sk
), sk
->sk_prot
, 1);
3773 /* Cleanup any SCTP per socket resources. */
3774 SCTP_STATIC
void sctp_destroy_sock(struct sock
*sk
)
3776 struct sctp_endpoint
*ep
;
3778 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk
);
3780 /* Release our hold on the endpoint. */
3781 ep
= sctp_sk(sk
)->ep
;
3782 sctp_endpoint_free(ep
);
3784 percpu_counter_dec(&sctp_sockets_allocated
);
3785 sock_prot_inuse_add(sock_net(sk
), sk
->sk_prot
, -1);
3789 /* API 4.1.7 shutdown() - TCP Style Syntax
3790 * int shutdown(int socket, int how);
3792 * sd - the socket descriptor of the association to be closed.
3793 * how - Specifies the type of shutdown. The values are
3796 * Disables further receive operations. No SCTP
3797 * protocol action is taken.
3799 * Disables further send operations, and initiates
3800 * the SCTP shutdown sequence.
3802 * Disables further send and receive operations
3803 * and initiates the SCTP shutdown sequence.
3805 SCTP_STATIC
void sctp_shutdown(struct sock
*sk
, int how
)
3807 struct sctp_endpoint
*ep
;
3808 struct sctp_association
*asoc
;
3810 if (!sctp_style(sk
, TCP
))
3813 if (how
& SEND_SHUTDOWN
) {
3814 ep
= sctp_sk(sk
)->ep
;
3815 if (!list_empty(&ep
->asocs
)) {
3816 asoc
= list_entry(ep
->asocs
.next
,
3817 struct sctp_association
, asocs
);
3818 sctp_primitive_SHUTDOWN(asoc
, NULL
);
3823 /* 7.2.1 Association Status (SCTP_STATUS)
3825 * Applications can retrieve current status information about an
3826 * association, including association state, peer receiver window size,
3827 * number of unacked data chunks, and number of data chunks pending
3828 * receipt. This information is read-only.
3830 static int sctp_getsockopt_sctp_status(struct sock
*sk
, int len
,
3831 char __user
*optval
,
3834 struct sctp_status status
;
3835 struct sctp_association
*asoc
= NULL
;
3836 struct sctp_transport
*transport
;
3837 sctp_assoc_t associd
;
3840 if (len
< sizeof(status
)) {
3845 len
= sizeof(status
);
3846 if (copy_from_user(&status
, optval
, len
)) {
3851 associd
= status
.sstat_assoc_id
;
3852 asoc
= sctp_id2assoc(sk
, associd
);
3858 transport
= asoc
->peer
.primary_path
;
3860 status
.sstat_assoc_id
= sctp_assoc2id(asoc
);
3861 status
.sstat_state
= asoc
->state
;
3862 status
.sstat_rwnd
= asoc
->peer
.rwnd
;
3863 status
.sstat_unackdata
= asoc
->unack_data
;
3865 status
.sstat_penddata
= sctp_tsnmap_pending(&asoc
->peer
.tsn_map
);
3866 status
.sstat_instrms
= asoc
->c
.sinit_max_instreams
;
3867 status
.sstat_outstrms
= asoc
->c
.sinit_num_ostreams
;
3868 status
.sstat_fragmentation_point
= asoc
->frag_point
;
3869 status
.sstat_primary
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
3870 memcpy(&status
.sstat_primary
.spinfo_address
, &transport
->ipaddr
,
3871 transport
->af_specific
->sockaddr_len
);
3872 /* Map ipv4 address into v4-mapped-on-v6 address. */
3873 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
3874 (union sctp_addr
*)&status
.sstat_primary
.spinfo_address
);
3875 status
.sstat_primary
.spinfo_state
= transport
->state
;
3876 status
.sstat_primary
.spinfo_cwnd
= transport
->cwnd
;
3877 status
.sstat_primary
.spinfo_srtt
= transport
->srtt
;
3878 status
.sstat_primary
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
3879 status
.sstat_primary
.spinfo_mtu
= transport
->pathmtu
;
3881 if (status
.sstat_primary
.spinfo_state
== SCTP_UNKNOWN
)
3882 status
.sstat_primary
.spinfo_state
= SCTP_ACTIVE
;
3884 if (put_user(len
, optlen
)) {
3889 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3890 len
, status
.sstat_state
, status
.sstat_rwnd
,
3891 status
.sstat_assoc_id
);
3893 if (copy_to_user(optval
, &status
, len
)) {
3903 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3905 * Applications can retrieve information about a specific peer address
3906 * of an association, including its reachability state, congestion
3907 * window, and retransmission timer values. This information is
3910 static int sctp_getsockopt_peer_addr_info(struct sock
*sk
, int len
,
3911 char __user
*optval
,
3914 struct sctp_paddrinfo pinfo
;
3915 struct sctp_transport
*transport
;
3918 if (len
< sizeof(pinfo
)) {
3923 len
= sizeof(pinfo
);
3924 if (copy_from_user(&pinfo
, optval
, len
)) {
3929 transport
= sctp_addr_id2transport(sk
, &pinfo
.spinfo_address
,
3930 pinfo
.spinfo_assoc_id
);
3934 pinfo
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
3935 pinfo
.spinfo_state
= transport
->state
;
3936 pinfo
.spinfo_cwnd
= transport
->cwnd
;
3937 pinfo
.spinfo_srtt
= transport
->srtt
;
3938 pinfo
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
3939 pinfo
.spinfo_mtu
= transport
->pathmtu
;
3941 if (pinfo
.spinfo_state
== SCTP_UNKNOWN
)
3942 pinfo
.spinfo_state
= SCTP_ACTIVE
;
3944 if (put_user(len
, optlen
)) {
3949 if (copy_to_user(optval
, &pinfo
, len
)) {
3958 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3960 * This option is a on/off flag. If enabled no SCTP message
3961 * fragmentation will be performed. Instead if a message being sent
3962 * exceeds the current PMTU size, the message will NOT be sent and
3963 * instead a error will be indicated to the user.
3965 static int sctp_getsockopt_disable_fragments(struct sock
*sk
, int len
,
3966 char __user
*optval
, int __user
*optlen
)
3970 if (len
< sizeof(int))
3974 val
= (sctp_sk(sk
)->disable_fragments
== 1);
3975 if (put_user(len
, optlen
))
3977 if (copy_to_user(optval
, &val
, len
))
3982 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3984 * This socket option is used to specify various notifications and
3985 * ancillary data the user wishes to receive.
3987 static int sctp_getsockopt_events(struct sock
*sk
, int len
, char __user
*optval
,
3990 if (len
< sizeof(struct sctp_event_subscribe
))
3992 len
= sizeof(struct sctp_event_subscribe
);
3993 if (put_user(len
, optlen
))
3995 if (copy_to_user(optval
, &sctp_sk(sk
)->subscribe
, len
))
4000 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4002 * This socket option is applicable to the UDP-style socket only. When
4003 * set it will cause associations that are idle for more than the
4004 * specified number of seconds to automatically close. An association
4005 * being idle is defined an association that has NOT sent or received
4006 * user data. The special value of '0' indicates that no automatic
4007 * close of any associations should be performed. The option expects an
4008 * integer defining the number of seconds of idle time before an
4009 * association is closed.
4011 static int sctp_getsockopt_autoclose(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4013 /* Applicable to UDP-style socket only */
4014 if (sctp_style(sk
, TCP
))
4016 if (len
< sizeof(int))
4019 if (put_user(len
, optlen
))
4021 if (copy_to_user(optval
, &sctp_sk(sk
)->autoclose
, sizeof(int)))
4026 /* Helper routine to branch off an association to a new socket. */
4027 SCTP_STATIC
int sctp_do_peeloff(struct sctp_association
*asoc
,
4028 struct socket
**sockp
)
4030 struct sock
*sk
= asoc
->base
.sk
;
4031 struct socket
*sock
;
4035 /* An association cannot be branched off from an already peeled-off
4036 * socket, nor is this supported for tcp style sockets.
4038 if (!sctp_style(sk
, UDP
))
4041 /* Create a new socket. */
4042 err
= sock_create(sk
->sk_family
, SOCK_SEQPACKET
, IPPROTO_SCTP
, &sock
);
4046 sctp_copy_sock(sock
->sk
, sk
, asoc
);
4048 /* Make peeled-off sockets more like 1-1 accepted sockets.
4049 * Set the daddr and initialize id to something more random
4051 af
= sctp_get_af_specific(asoc
->peer
.primary_addr
.sa
.sa_family
);
4052 af
->to_sk_daddr(&asoc
->peer
.primary_addr
, sk
);
4054 /* Populate the fields of the newsk from the oldsk and migrate the
4055 * asoc to the newsk.
4057 sctp_sock_migrate(sk
, sock
->sk
, asoc
, SCTP_SOCKET_UDP_HIGH_BANDWIDTH
);
4064 static int sctp_getsockopt_peeloff(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4066 sctp_peeloff_arg_t peeloff
;
4067 struct socket
*newsock
;
4069 struct sctp_association
*asoc
;
4071 if (len
< sizeof(sctp_peeloff_arg_t
))
4073 len
= sizeof(sctp_peeloff_arg_t
);
4074 if (copy_from_user(&peeloff
, optval
, len
))
4077 asoc
= sctp_id2assoc(sk
, peeloff
.associd
);
4083 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__
, sk
, asoc
);
4085 retval
= sctp_do_peeloff(asoc
, &newsock
);
4089 /* Map the socket to an unused fd that can be returned to the user. */
4090 retval
= sock_map_fd(newsock
, 0);
4092 sock_release(newsock
);
4096 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
4097 __func__
, sk
, asoc
, newsock
->sk
, retval
);
4099 /* Return the fd mapped to the new socket. */
4100 peeloff
.sd
= retval
;
4101 if (put_user(len
, optlen
))
4103 if (copy_to_user(optval
, &peeloff
, len
))
4110 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4112 * Applications can enable or disable heartbeats for any peer address of
4113 * an association, modify an address's heartbeat interval, force a
4114 * heartbeat to be sent immediately, and adjust the address's maximum
4115 * number of retransmissions sent before an address is considered
4116 * unreachable. The following structure is used to access and modify an
4117 * address's parameters:
4119 * struct sctp_paddrparams {
4120 * sctp_assoc_t spp_assoc_id;
4121 * struct sockaddr_storage spp_address;
4122 * uint32_t spp_hbinterval;
4123 * uint16_t spp_pathmaxrxt;
4124 * uint32_t spp_pathmtu;
4125 * uint32_t spp_sackdelay;
4126 * uint32_t spp_flags;
4129 * spp_assoc_id - (one-to-many style socket) This is filled in the
4130 * application, and identifies the association for
4132 * spp_address - This specifies which address is of interest.
4133 * spp_hbinterval - This contains the value of the heartbeat interval,
4134 * in milliseconds. If a value of zero
4135 * is present in this field then no changes are to
4136 * be made to this parameter.
4137 * spp_pathmaxrxt - This contains the maximum number of
4138 * retransmissions before this address shall be
4139 * considered unreachable. If a value of zero
4140 * is present in this field then no changes are to
4141 * be made to this parameter.
4142 * spp_pathmtu - When Path MTU discovery is disabled the value
4143 * specified here will be the "fixed" path mtu.
4144 * Note that if the spp_address field is empty
4145 * then all associations on this address will
4146 * have this fixed path mtu set upon them.
4148 * spp_sackdelay - When delayed sack is enabled, this value specifies
4149 * the number of milliseconds that sacks will be delayed
4150 * for. This value will apply to all addresses of an
4151 * association if the spp_address field is empty. Note
4152 * also, that if delayed sack is enabled and this
4153 * value is set to 0, no change is made to the last
4154 * recorded delayed sack timer value.
4156 * spp_flags - These flags are used to control various features
4157 * on an association. The flag field may contain
4158 * zero or more of the following options.
4160 * SPP_HB_ENABLE - Enable heartbeats on the
4161 * specified address. Note that if the address
4162 * field is empty all addresses for the association
4163 * have heartbeats enabled upon them.
4165 * SPP_HB_DISABLE - Disable heartbeats on the
4166 * speicifed address. Note that if the address
4167 * field is empty all addresses for the association
4168 * will have their heartbeats disabled. Note also
4169 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4170 * mutually exclusive, only one of these two should
4171 * be specified. Enabling both fields will have
4172 * undetermined results.
4174 * SPP_HB_DEMAND - Request a user initiated heartbeat
4175 * to be made immediately.
4177 * SPP_PMTUD_ENABLE - This field will enable PMTU
4178 * discovery upon the specified address. Note that
4179 * if the address feild is empty then all addresses
4180 * on the association are effected.
4182 * SPP_PMTUD_DISABLE - This field will disable PMTU
4183 * discovery upon the specified address. Note that
4184 * if the address feild is empty then all addresses
4185 * on the association are effected. Not also that
4186 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4187 * exclusive. Enabling both will have undetermined
4190 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4191 * on delayed sack. The time specified in spp_sackdelay
4192 * is used to specify the sack delay for this address. Note
4193 * that if spp_address is empty then all addresses will
4194 * enable delayed sack and take on the sack delay
4195 * value specified in spp_sackdelay.
4196 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4197 * off delayed sack. If the spp_address field is blank then
4198 * delayed sack is disabled for the entire association. Note
4199 * also that this field is mutually exclusive to
4200 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4203 static int sctp_getsockopt_peer_addr_params(struct sock
*sk
, int len
,
4204 char __user
*optval
, int __user
*optlen
)
4206 struct sctp_paddrparams params
;
4207 struct sctp_transport
*trans
= NULL
;
4208 struct sctp_association
*asoc
= NULL
;
4209 struct sctp_sock
*sp
= sctp_sk(sk
);
4211 if (len
< sizeof(struct sctp_paddrparams
))
4213 len
= sizeof(struct sctp_paddrparams
);
4214 if (copy_from_user(¶ms
, optval
, len
))
4217 /* If an address other than INADDR_ANY is specified, and
4218 * no transport is found, then the request is invalid.
4220 if (!sctp_is_any(sk
, ( union sctp_addr
*)¶ms
.spp_address
)) {
4221 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
4222 params
.spp_assoc_id
);
4224 SCTP_DEBUG_PRINTK("Failed no transport\n");
4229 /* Get association, if assoc_id != 0 and the socket is a one
4230 * to many style socket, and an association was not found, then
4231 * the id was invalid.
4233 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
4234 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
)) {
4235 SCTP_DEBUG_PRINTK("Failed no association\n");
4240 /* Fetch transport values. */
4241 params
.spp_hbinterval
= jiffies_to_msecs(trans
->hbinterval
);
4242 params
.spp_pathmtu
= trans
->pathmtu
;
4243 params
.spp_pathmaxrxt
= trans
->pathmaxrxt
;
4244 params
.spp_sackdelay
= jiffies_to_msecs(trans
->sackdelay
);
4246 /*draft-11 doesn't say what to return in spp_flags*/
4247 params
.spp_flags
= trans
->param_flags
;
4249 /* Fetch association values. */
4250 params
.spp_hbinterval
= jiffies_to_msecs(asoc
->hbinterval
);
4251 params
.spp_pathmtu
= asoc
->pathmtu
;
4252 params
.spp_pathmaxrxt
= asoc
->pathmaxrxt
;
4253 params
.spp_sackdelay
= jiffies_to_msecs(asoc
->sackdelay
);
4255 /*draft-11 doesn't say what to return in spp_flags*/
4256 params
.spp_flags
= asoc
->param_flags
;
4258 /* Fetch socket values. */
4259 params
.spp_hbinterval
= sp
->hbinterval
;
4260 params
.spp_pathmtu
= sp
->pathmtu
;
4261 params
.spp_sackdelay
= sp
->sackdelay
;
4262 params
.spp_pathmaxrxt
= sp
->pathmaxrxt
;
4264 /*draft-11 doesn't say what to return in spp_flags*/
4265 params
.spp_flags
= sp
->param_flags
;
4268 if (copy_to_user(optval
, ¶ms
, len
))
4271 if (put_user(len
, optlen
))
4278 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4280 * This option will effect the way delayed acks are performed. This
4281 * option allows you to get or set the delayed ack time, in
4282 * milliseconds. It also allows changing the delayed ack frequency.
4283 * Changing the frequency to 1 disables the delayed sack algorithm. If
4284 * the assoc_id is 0, then this sets or gets the endpoints default
4285 * values. If the assoc_id field is non-zero, then the set or get
4286 * effects the specified association for the one to many model (the
4287 * assoc_id field is ignored by the one to one model). Note that if
4288 * sack_delay or sack_freq are 0 when setting this option, then the
4289 * current values will remain unchanged.
4291 * struct sctp_sack_info {
4292 * sctp_assoc_t sack_assoc_id;
4293 * uint32_t sack_delay;
4294 * uint32_t sack_freq;
4297 * sack_assoc_id - This parameter, indicates which association the user
4298 * is performing an action upon. Note that if this field's value is
4299 * zero then the endpoints default value is changed (effecting future
4300 * associations only).
4302 * sack_delay - This parameter contains the number of milliseconds that
4303 * the user is requesting the delayed ACK timer be set to. Note that
4304 * this value is defined in the standard to be between 200 and 500
4307 * sack_freq - This parameter contains the number of packets that must
4308 * be received before a sack is sent without waiting for the delay
4309 * timer to expire. The default value for this is 2, setting this
4310 * value to 1 will disable the delayed sack algorithm.
4312 static int sctp_getsockopt_delayed_ack(struct sock
*sk
, int len
,
4313 char __user
*optval
,
4316 struct sctp_sack_info params
;
4317 struct sctp_association
*asoc
= NULL
;
4318 struct sctp_sock
*sp
= sctp_sk(sk
);
4320 if (len
>= sizeof(struct sctp_sack_info
)) {
4321 len
= sizeof(struct sctp_sack_info
);
4323 if (copy_from_user(¶ms
, optval
, len
))
4325 } else if (len
== sizeof(struct sctp_assoc_value
)) {
4326 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4327 pr_warn("Use struct sctp_sack_info instead\n");
4328 if (copy_from_user(¶ms
, optval
, len
))
4333 /* Get association, if sack_assoc_id != 0 and the socket is a one
4334 * to many style socket, and an association was not found, then
4335 * the id was invalid.
4337 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
4338 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
4342 /* Fetch association values. */
4343 if (asoc
->param_flags
& SPP_SACKDELAY_ENABLE
) {
4344 params
.sack_delay
= jiffies_to_msecs(
4346 params
.sack_freq
= asoc
->sackfreq
;
4349 params
.sack_delay
= 0;
4350 params
.sack_freq
= 1;
4353 /* Fetch socket values. */
4354 if (sp
->param_flags
& SPP_SACKDELAY_ENABLE
) {
4355 params
.sack_delay
= sp
->sackdelay
;
4356 params
.sack_freq
= sp
->sackfreq
;
4358 params
.sack_delay
= 0;
4359 params
.sack_freq
= 1;
4363 if (copy_to_user(optval
, ¶ms
, len
))
4366 if (put_user(len
, optlen
))
4372 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4374 * Applications can specify protocol parameters for the default association
4375 * initialization. The option name argument to setsockopt() and getsockopt()
4378 * Setting initialization parameters is effective only on an unconnected
4379 * socket (for UDP-style sockets only future associations are effected
4380 * by the change). With TCP-style sockets, this option is inherited by
4381 * sockets derived from a listener socket.
4383 static int sctp_getsockopt_initmsg(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4385 if (len
< sizeof(struct sctp_initmsg
))
4387 len
= sizeof(struct sctp_initmsg
);
4388 if (put_user(len
, optlen
))
4390 if (copy_to_user(optval
, &sctp_sk(sk
)->initmsg
, len
))
4396 static int sctp_getsockopt_peer_addrs(struct sock
*sk
, int len
,
4397 char __user
*optval
, int __user
*optlen
)
4399 struct sctp_association
*asoc
;
4401 struct sctp_getaddrs getaddrs
;
4402 struct sctp_transport
*from
;
4404 union sctp_addr temp
;
4405 struct sctp_sock
*sp
= sctp_sk(sk
);
4410 if (len
< sizeof(struct sctp_getaddrs
))
4413 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
4416 /* For UDP-style sockets, id specifies the association to query. */
4417 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
4421 to
= optval
+ offsetof(struct sctp_getaddrs
,addrs
);
4422 space_left
= len
- offsetof(struct sctp_getaddrs
,addrs
);
4424 list_for_each_entry(from
, &asoc
->peer
.transport_addr_list
,
4426 memcpy(&temp
, &from
->ipaddr
, sizeof(temp
));
4427 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
4428 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4429 if (space_left
< addrlen
)
4431 if (copy_to_user(to
, &temp
, addrlen
))
4435 space_left
-= addrlen
;
4438 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
))
4440 bytes_copied
= ((char __user
*)to
) - optval
;
4441 if (put_user(bytes_copied
, optlen
))
4447 static int sctp_copy_laddrs(struct sock
*sk
, __u16 port
, void *to
,
4448 size_t space_left
, int *bytes_copied
)
4450 struct sctp_sockaddr_entry
*addr
;
4451 union sctp_addr temp
;
4456 list_for_each_entry_rcu(addr
, &sctp_local_addr_list
, list
) {
4460 if ((PF_INET
== sk
->sk_family
) &&
4461 (AF_INET6
== addr
->a
.sa
.sa_family
))
4463 if ((PF_INET6
== sk
->sk_family
) &&
4464 inet_v6_ipv6only(sk
) &&
4465 (AF_INET
== addr
->a
.sa
.sa_family
))
4467 memcpy(&temp
, &addr
->a
, sizeof(temp
));
4468 if (!temp
.v4
.sin_port
)
4469 temp
.v4
.sin_port
= htons(port
);
4471 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
4473 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4474 if (space_left
< addrlen
) {
4478 memcpy(to
, &temp
, addrlen
);
4482 space_left
-= addrlen
;
4483 *bytes_copied
+= addrlen
;
4491 static int sctp_getsockopt_local_addrs(struct sock
*sk
, int len
,
4492 char __user
*optval
, int __user
*optlen
)
4494 struct sctp_bind_addr
*bp
;
4495 struct sctp_association
*asoc
;
4497 struct sctp_getaddrs getaddrs
;
4498 struct sctp_sockaddr_entry
*addr
;
4500 union sctp_addr temp
;
4501 struct sctp_sock
*sp
= sctp_sk(sk
);
4505 int bytes_copied
= 0;
4509 if (len
< sizeof(struct sctp_getaddrs
))
4512 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
4516 * For UDP-style sockets, id specifies the association to query.
4517 * If the id field is set to the value '0' then the locally bound
4518 * addresses are returned without regard to any particular
4521 if (0 == getaddrs
.assoc_id
) {
4522 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
4524 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
4527 bp
= &asoc
->base
.bind_addr
;
4530 to
= optval
+ offsetof(struct sctp_getaddrs
,addrs
);
4531 space_left
= len
- offsetof(struct sctp_getaddrs
,addrs
);
4533 addrs
= kmalloc(space_left
, GFP_KERNEL
);
4537 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4538 * addresses from the global local address list.
4540 if (sctp_list_single_entry(&bp
->address_list
)) {
4541 addr
= list_entry(bp
->address_list
.next
,
4542 struct sctp_sockaddr_entry
, list
);
4543 if (sctp_is_any(sk
, &addr
->a
)) {
4544 cnt
= sctp_copy_laddrs(sk
, bp
->port
, addrs
,
4545 space_left
, &bytes_copied
);
4555 /* Protection on the bound address list is not needed since
4556 * in the socket option context we hold a socket lock and
4557 * thus the bound address list can't change.
4559 list_for_each_entry(addr
, &bp
->address_list
, list
) {
4560 memcpy(&temp
, &addr
->a
, sizeof(temp
));
4561 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
4562 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4563 if (space_left
< addrlen
) {
4564 err
= -ENOMEM
; /*fixme: right error?*/
4567 memcpy(buf
, &temp
, addrlen
);
4569 bytes_copied
+= addrlen
;
4571 space_left
-= addrlen
;
4575 if (copy_to_user(to
, addrs
, bytes_copied
)) {
4579 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
)) {
4583 if (put_user(bytes_copied
, optlen
))
4590 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4592 * Requests that the local SCTP stack use the enclosed peer address as
4593 * the association primary. The enclosed address must be one of the
4594 * association peer's addresses.
4596 static int sctp_getsockopt_primary_addr(struct sock
*sk
, int len
,
4597 char __user
*optval
, int __user
*optlen
)
4599 struct sctp_prim prim
;
4600 struct sctp_association
*asoc
;
4601 struct sctp_sock
*sp
= sctp_sk(sk
);
4603 if (len
< sizeof(struct sctp_prim
))
4606 len
= sizeof(struct sctp_prim
);
4608 if (copy_from_user(&prim
, optval
, len
))
4611 asoc
= sctp_id2assoc(sk
, prim
.ssp_assoc_id
);
4615 if (!asoc
->peer
.primary_path
)
4618 memcpy(&prim
.ssp_addr
, &asoc
->peer
.primary_path
->ipaddr
,
4619 asoc
->peer
.primary_path
->af_specific
->sockaddr_len
);
4621 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
,
4622 (union sctp_addr
*)&prim
.ssp_addr
);
4624 if (put_user(len
, optlen
))
4626 if (copy_to_user(optval
, &prim
, len
))
4633 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4635 * Requests that the local endpoint set the specified Adaptation Layer
4636 * Indication parameter for all future INIT and INIT-ACK exchanges.
4638 static int sctp_getsockopt_adaptation_layer(struct sock
*sk
, int len
,
4639 char __user
*optval
, int __user
*optlen
)
4641 struct sctp_setadaptation adaptation
;
4643 if (len
< sizeof(struct sctp_setadaptation
))
4646 len
= sizeof(struct sctp_setadaptation
);
4648 adaptation
.ssb_adaptation_ind
= sctp_sk(sk
)->adaptation_ind
;
4650 if (put_user(len
, optlen
))
4652 if (copy_to_user(optval
, &adaptation
, len
))
4660 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4662 * Applications that wish to use the sendto() system call may wish to
4663 * specify a default set of parameters that would normally be supplied
4664 * through the inclusion of ancillary data. This socket option allows
4665 * such an application to set the default sctp_sndrcvinfo structure.
4668 * The application that wishes to use this socket option simply passes
4669 * in to this call the sctp_sndrcvinfo structure defined in Section
4670 * 5.2.2) The input parameters accepted by this call include
4671 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4672 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4673 * to this call if the caller is using the UDP model.
4675 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4677 static int sctp_getsockopt_default_send_param(struct sock
*sk
,
4678 int len
, char __user
*optval
,
4681 struct sctp_sndrcvinfo info
;
4682 struct sctp_association
*asoc
;
4683 struct sctp_sock
*sp
= sctp_sk(sk
);
4685 if (len
< sizeof(struct sctp_sndrcvinfo
))
4688 len
= sizeof(struct sctp_sndrcvinfo
);
4690 if (copy_from_user(&info
, optval
, len
))
4693 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
4694 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
4698 info
.sinfo_stream
= asoc
->default_stream
;
4699 info
.sinfo_flags
= asoc
->default_flags
;
4700 info
.sinfo_ppid
= asoc
->default_ppid
;
4701 info
.sinfo_context
= asoc
->default_context
;
4702 info
.sinfo_timetolive
= asoc
->default_timetolive
;
4704 info
.sinfo_stream
= sp
->default_stream
;
4705 info
.sinfo_flags
= sp
->default_flags
;
4706 info
.sinfo_ppid
= sp
->default_ppid
;
4707 info
.sinfo_context
= sp
->default_context
;
4708 info
.sinfo_timetolive
= sp
->default_timetolive
;
4711 if (put_user(len
, optlen
))
4713 if (copy_to_user(optval
, &info
, len
))
4721 * 7.1.5 SCTP_NODELAY
4723 * Turn on/off any Nagle-like algorithm. This means that packets are
4724 * generally sent as soon as possible and no unnecessary delays are
4725 * introduced, at the cost of more packets in the network. Expects an
4726 * integer boolean flag.
4729 static int sctp_getsockopt_nodelay(struct sock
*sk
, int len
,
4730 char __user
*optval
, int __user
*optlen
)
4734 if (len
< sizeof(int))
4738 val
= (sctp_sk(sk
)->nodelay
== 1);
4739 if (put_user(len
, optlen
))
4741 if (copy_to_user(optval
, &val
, len
))
4748 * 7.1.1 SCTP_RTOINFO
4750 * The protocol parameters used to initialize and bound retransmission
4751 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4752 * and modify these parameters.
4753 * All parameters are time values, in milliseconds. A value of 0, when
4754 * modifying the parameters, indicates that the current value should not
4758 static int sctp_getsockopt_rtoinfo(struct sock
*sk
, int len
,
4759 char __user
*optval
,
4760 int __user
*optlen
) {
4761 struct sctp_rtoinfo rtoinfo
;
4762 struct sctp_association
*asoc
;
4764 if (len
< sizeof (struct sctp_rtoinfo
))
4767 len
= sizeof(struct sctp_rtoinfo
);
4769 if (copy_from_user(&rtoinfo
, optval
, len
))
4772 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
4774 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
4777 /* Values corresponding to the specific association. */
4779 rtoinfo
.srto_initial
= jiffies_to_msecs(asoc
->rto_initial
);
4780 rtoinfo
.srto_max
= jiffies_to_msecs(asoc
->rto_max
);
4781 rtoinfo
.srto_min
= jiffies_to_msecs(asoc
->rto_min
);
4783 /* Values corresponding to the endpoint. */
4784 struct sctp_sock
*sp
= sctp_sk(sk
);
4786 rtoinfo
.srto_initial
= sp
->rtoinfo
.srto_initial
;
4787 rtoinfo
.srto_max
= sp
->rtoinfo
.srto_max
;
4788 rtoinfo
.srto_min
= sp
->rtoinfo
.srto_min
;
4791 if (put_user(len
, optlen
))
4794 if (copy_to_user(optval
, &rtoinfo
, len
))
4802 * 7.1.2 SCTP_ASSOCINFO
4804 * This option is used to tune the maximum retransmission attempts
4805 * of the association.
4806 * Returns an error if the new association retransmission value is
4807 * greater than the sum of the retransmission value of the peer.
4808 * See [SCTP] for more information.
4811 static int sctp_getsockopt_associnfo(struct sock
*sk
, int len
,
4812 char __user
*optval
,
4816 struct sctp_assocparams assocparams
;
4817 struct sctp_association
*asoc
;
4818 struct list_head
*pos
;
4821 if (len
< sizeof (struct sctp_assocparams
))
4824 len
= sizeof(struct sctp_assocparams
);
4826 if (copy_from_user(&assocparams
, optval
, len
))
4829 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
4831 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
4834 /* Values correspoinding to the specific association */
4836 assocparams
.sasoc_asocmaxrxt
= asoc
->max_retrans
;
4837 assocparams
.sasoc_peer_rwnd
= asoc
->peer
.rwnd
;
4838 assocparams
.sasoc_local_rwnd
= asoc
->a_rwnd
;
4839 assocparams
.sasoc_cookie_life
= (asoc
->cookie_life
.tv_sec
4841 (asoc
->cookie_life
.tv_usec
4844 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
4848 assocparams
.sasoc_number_peer_destinations
= cnt
;
4850 /* Values corresponding to the endpoint */
4851 struct sctp_sock
*sp
= sctp_sk(sk
);
4853 assocparams
.sasoc_asocmaxrxt
= sp
->assocparams
.sasoc_asocmaxrxt
;
4854 assocparams
.sasoc_peer_rwnd
= sp
->assocparams
.sasoc_peer_rwnd
;
4855 assocparams
.sasoc_local_rwnd
= sp
->assocparams
.sasoc_local_rwnd
;
4856 assocparams
.sasoc_cookie_life
=
4857 sp
->assocparams
.sasoc_cookie_life
;
4858 assocparams
.sasoc_number_peer_destinations
=
4860 sasoc_number_peer_destinations
;
4863 if (put_user(len
, optlen
))
4866 if (copy_to_user(optval
, &assocparams
, len
))
4873 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4875 * This socket option is a boolean flag which turns on or off mapped V4
4876 * addresses. If this option is turned on and the socket is type
4877 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4878 * If this option is turned off, then no mapping will be done of V4
4879 * addresses and a user will receive both PF_INET6 and PF_INET type
4880 * addresses on the socket.
4882 static int sctp_getsockopt_mappedv4(struct sock
*sk
, int len
,
4883 char __user
*optval
, int __user
*optlen
)
4886 struct sctp_sock
*sp
= sctp_sk(sk
);
4888 if (len
< sizeof(int))
4893 if (put_user(len
, optlen
))
4895 if (copy_to_user(optval
, &val
, len
))
4902 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
4903 * (chapter and verse is quoted at sctp_setsockopt_context())
4905 static int sctp_getsockopt_context(struct sock
*sk
, int len
,
4906 char __user
*optval
, int __user
*optlen
)
4908 struct sctp_assoc_value params
;
4909 struct sctp_sock
*sp
;
4910 struct sctp_association
*asoc
;
4912 if (len
< sizeof(struct sctp_assoc_value
))
4915 len
= sizeof(struct sctp_assoc_value
);
4917 if (copy_from_user(¶ms
, optval
, len
))
4922 if (params
.assoc_id
!= 0) {
4923 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
4926 params
.assoc_value
= asoc
->default_rcv_context
;
4928 params
.assoc_value
= sp
->default_rcv_context
;
4931 if (put_user(len
, optlen
))
4933 if (copy_to_user(optval
, ¶ms
, len
))
4940 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
4941 * This option will get or set the maximum size to put in any outgoing
4942 * SCTP DATA chunk. If a message is larger than this size it will be
4943 * fragmented by SCTP into the specified size. Note that the underlying
4944 * SCTP implementation may fragment into smaller sized chunks when the
4945 * PMTU of the underlying association is smaller than the value set by
4946 * the user. The default value for this option is '0' which indicates
4947 * the user is NOT limiting fragmentation and only the PMTU will effect
4948 * SCTP's choice of DATA chunk size. Note also that values set larger
4949 * than the maximum size of an IP datagram will effectively let SCTP
4950 * control fragmentation (i.e. the same as setting this option to 0).
4952 * The following structure is used to access and modify this parameter:
4954 * struct sctp_assoc_value {
4955 * sctp_assoc_t assoc_id;
4956 * uint32_t assoc_value;
4959 * assoc_id: This parameter is ignored for one-to-one style sockets.
4960 * For one-to-many style sockets this parameter indicates which
4961 * association the user is performing an action upon. Note that if
4962 * this field's value is zero then the endpoints default value is
4963 * changed (effecting future associations only).
4964 * assoc_value: This parameter specifies the maximum size in bytes.
4966 static int sctp_getsockopt_maxseg(struct sock
*sk
, int len
,
4967 char __user
*optval
, int __user
*optlen
)
4969 struct sctp_assoc_value params
;
4970 struct sctp_association
*asoc
;
4972 if (len
== sizeof(int)) {
4973 pr_warn("Use of int in maxseg socket option deprecated\n");
4974 pr_warn("Use struct sctp_assoc_value instead\n");
4975 params
.assoc_id
= 0;
4976 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
4977 len
= sizeof(struct sctp_assoc_value
);
4978 if (copy_from_user(¶ms
, optval
, sizeof(params
)))
4983 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
4984 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
4988 params
.assoc_value
= asoc
->frag_point
;
4990 params
.assoc_value
= sctp_sk(sk
)->user_frag
;
4992 if (put_user(len
, optlen
))
4994 if (len
== sizeof(int)) {
4995 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
4998 if (copy_to_user(optval
, ¶ms
, len
))
5006 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5007 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5009 static int sctp_getsockopt_fragment_interleave(struct sock
*sk
, int len
,
5010 char __user
*optval
, int __user
*optlen
)
5014 if (len
< sizeof(int))
5019 val
= sctp_sk(sk
)->frag_interleave
;
5020 if (put_user(len
, optlen
))
5022 if (copy_to_user(optval
, &val
, len
))
5029 * 7.1.25. Set or Get the sctp partial delivery point
5030 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5032 static int sctp_getsockopt_partial_delivery_point(struct sock
*sk
, int len
,
5033 char __user
*optval
,
5038 if (len
< sizeof(u32
))
5043 val
= sctp_sk(sk
)->pd_point
;
5044 if (put_user(len
, optlen
))
5046 if (copy_to_user(optval
, &val
, len
))
5053 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5054 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5056 static int sctp_getsockopt_maxburst(struct sock
*sk
, int len
,
5057 char __user
*optval
,
5060 struct sctp_assoc_value params
;
5061 struct sctp_sock
*sp
;
5062 struct sctp_association
*asoc
;
5064 if (len
== sizeof(int)) {
5065 pr_warn("Use of int in max_burst socket option deprecated\n");
5066 pr_warn("Use struct sctp_assoc_value instead\n");
5067 params
.assoc_id
= 0;
5068 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5069 len
= sizeof(struct sctp_assoc_value
);
5070 if (copy_from_user(¶ms
, optval
, len
))
5077 if (params
.assoc_id
!= 0) {
5078 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5081 params
.assoc_value
= asoc
->max_burst
;
5083 params
.assoc_value
= sp
->max_burst
;
5085 if (len
== sizeof(int)) {
5086 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5089 if (copy_to_user(optval
, ¶ms
, len
))
5097 static int sctp_getsockopt_hmac_ident(struct sock
*sk
, int len
,
5098 char __user
*optval
, int __user
*optlen
)
5100 struct sctp_hmacalgo __user
*p
= (void __user
*)optval
;
5101 struct sctp_hmac_algo_param
*hmacs
;
5105 if (!sctp_auth_enable
)
5108 hmacs
= sctp_sk(sk
)->ep
->auth_hmacs_list
;
5109 data_len
= ntohs(hmacs
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5111 if (len
< sizeof(struct sctp_hmacalgo
) + data_len
)
5114 len
= sizeof(struct sctp_hmacalgo
) + data_len
;
5115 num_idents
= data_len
/ sizeof(u16
);
5117 if (put_user(len
, optlen
))
5119 if (put_user(num_idents
, &p
->shmac_num_idents
))
5121 if (copy_to_user(p
->shmac_idents
, hmacs
->hmac_ids
, data_len
))
5126 static int sctp_getsockopt_active_key(struct sock
*sk
, int len
,
5127 char __user
*optval
, int __user
*optlen
)
5129 struct sctp_authkeyid val
;
5130 struct sctp_association
*asoc
;
5132 if (!sctp_auth_enable
)
5135 if (len
< sizeof(struct sctp_authkeyid
))
5137 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authkeyid
)))
5140 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
5141 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
5145 val
.scact_keynumber
= asoc
->active_key_id
;
5147 val
.scact_keynumber
= sctp_sk(sk
)->ep
->active_key_id
;
5149 len
= sizeof(struct sctp_authkeyid
);
5150 if (put_user(len
, optlen
))
5152 if (copy_to_user(optval
, &val
, len
))
5158 static int sctp_getsockopt_peer_auth_chunks(struct sock
*sk
, int len
,
5159 char __user
*optval
, int __user
*optlen
)
5161 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5162 struct sctp_authchunks val
;
5163 struct sctp_association
*asoc
;
5164 struct sctp_chunks_param
*ch
;
5168 if (!sctp_auth_enable
)
5171 if (len
< sizeof(struct sctp_authchunks
))
5174 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5177 to
= p
->gauth_chunks
;
5178 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5182 ch
= asoc
->peer
.peer_chunks
;
5186 /* See if the user provided enough room for all the data */
5187 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5188 if (len
< num_chunks
)
5191 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
5194 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
5195 if (put_user(len
, optlen
)) return -EFAULT
;
5196 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
5201 static int sctp_getsockopt_local_auth_chunks(struct sock
*sk
, int len
,
5202 char __user
*optval
, int __user
*optlen
)
5204 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5205 struct sctp_authchunks val
;
5206 struct sctp_association
*asoc
;
5207 struct sctp_chunks_param
*ch
;
5211 if (!sctp_auth_enable
)
5214 if (len
< sizeof(struct sctp_authchunks
))
5217 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5220 to
= p
->gauth_chunks
;
5221 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5222 if (!asoc
&& val
.gauth_assoc_id
&& sctp_style(sk
, UDP
))
5226 ch
= (struct sctp_chunks_param
*)asoc
->c
.auth_chunks
;
5228 ch
= sctp_sk(sk
)->ep
->auth_chunk_list
;
5233 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5234 if (len
< sizeof(struct sctp_authchunks
) + num_chunks
)
5237 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
5240 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
5241 if (put_user(len
, optlen
))
5243 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
5250 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5251 * This option gets the current number of associations that are attached
5252 * to a one-to-many style socket. The option value is an uint32_t.
5254 static int sctp_getsockopt_assoc_number(struct sock
*sk
, int len
,
5255 char __user
*optval
, int __user
*optlen
)
5257 struct sctp_sock
*sp
= sctp_sk(sk
);
5258 struct sctp_association
*asoc
;
5261 if (sctp_style(sk
, TCP
))
5264 if (len
< sizeof(u32
))
5269 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5273 if (put_user(len
, optlen
))
5275 if (copy_to_user(optval
, &val
, len
))
5282 * 8.2.6. Get the Current Identifiers of Associations
5283 * (SCTP_GET_ASSOC_ID_LIST)
5285 * This option gets the current list of SCTP association identifiers of
5286 * the SCTP associations handled by a one-to-many style socket.
5288 static int sctp_getsockopt_assoc_ids(struct sock
*sk
, int len
,
5289 char __user
*optval
, int __user
*optlen
)
5291 struct sctp_sock
*sp
= sctp_sk(sk
);
5292 struct sctp_association
*asoc
;
5293 struct sctp_assoc_ids
*ids
;
5296 if (sctp_style(sk
, TCP
))
5299 if (len
< sizeof(struct sctp_assoc_ids
))
5302 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5306 if (len
< sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
)
5309 len
= sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
;
5311 ids
= kmalloc(len
, GFP_KERNEL
);
5315 ids
->gaids_number_of_ids
= num
;
5317 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5318 ids
->gaids_assoc_id
[num
++] = asoc
->assoc_id
;
5321 if (put_user(len
, optlen
) || copy_to_user(optval
, ids
, len
)) {
5330 SCTP_STATIC
int sctp_getsockopt(struct sock
*sk
, int level
, int optname
,
5331 char __user
*optval
, int __user
*optlen
)
5336 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5339 /* I can hardly begin to describe how wrong this is. This is
5340 * so broken as to be worse than useless. The API draft
5341 * REALLY is NOT helpful here... I am not convinced that the
5342 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5343 * are at all well-founded.
5345 if (level
!= SOL_SCTP
) {
5346 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
5348 retval
= af
->getsockopt(sk
, level
, optname
, optval
, optlen
);
5352 if (get_user(len
, optlen
))
5359 retval
= sctp_getsockopt_sctp_status(sk
, len
, optval
, optlen
);
5361 case SCTP_DISABLE_FRAGMENTS
:
5362 retval
= sctp_getsockopt_disable_fragments(sk
, len
, optval
,
5366 retval
= sctp_getsockopt_events(sk
, len
, optval
, optlen
);
5368 case SCTP_AUTOCLOSE
:
5369 retval
= sctp_getsockopt_autoclose(sk
, len
, optval
, optlen
);
5371 case SCTP_SOCKOPT_PEELOFF
:
5372 retval
= sctp_getsockopt_peeloff(sk
, len
, optval
, optlen
);
5374 case SCTP_PEER_ADDR_PARAMS
:
5375 retval
= sctp_getsockopt_peer_addr_params(sk
, len
, optval
,
5378 case SCTP_DELAYED_SACK
:
5379 retval
= sctp_getsockopt_delayed_ack(sk
, len
, optval
,
5383 retval
= sctp_getsockopt_initmsg(sk
, len
, optval
, optlen
);
5385 case SCTP_GET_PEER_ADDRS
:
5386 retval
= sctp_getsockopt_peer_addrs(sk
, len
, optval
,
5389 case SCTP_GET_LOCAL_ADDRS
:
5390 retval
= sctp_getsockopt_local_addrs(sk
, len
, optval
,
5393 case SCTP_SOCKOPT_CONNECTX3
:
5394 retval
= sctp_getsockopt_connectx3(sk
, len
, optval
, optlen
);
5396 case SCTP_DEFAULT_SEND_PARAM
:
5397 retval
= sctp_getsockopt_default_send_param(sk
, len
,
5400 case SCTP_PRIMARY_ADDR
:
5401 retval
= sctp_getsockopt_primary_addr(sk
, len
, optval
, optlen
);
5404 retval
= sctp_getsockopt_nodelay(sk
, len
, optval
, optlen
);
5407 retval
= sctp_getsockopt_rtoinfo(sk
, len
, optval
, optlen
);
5409 case SCTP_ASSOCINFO
:
5410 retval
= sctp_getsockopt_associnfo(sk
, len
, optval
, optlen
);
5412 case SCTP_I_WANT_MAPPED_V4_ADDR
:
5413 retval
= sctp_getsockopt_mappedv4(sk
, len
, optval
, optlen
);
5416 retval
= sctp_getsockopt_maxseg(sk
, len
, optval
, optlen
);
5418 case SCTP_GET_PEER_ADDR_INFO
:
5419 retval
= sctp_getsockopt_peer_addr_info(sk
, len
, optval
,
5422 case SCTP_ADAPTATION_LAYER
:
5423 retval
= sctp_getsockopt_adaptation_layer(sk
, len
, optval
,
5427 retval
= sctp_getsockopt_context(sk
, len
, optval
, optlen
);
5429 case SCTP_FRAGMENT_INTERLEAVE
:
5430 retval
= sctp_getsockopt_fragment_interleave(sk
, len
, optval
,
5433 case SCTP_PARTIAL_DELIVERY_POINT
:
5434 retval
= sctp_getsockopt_partial_delivery_point(sk
, len
, optval
,
5437 case SCTP_MAX_BURST
:
5438 retval
= sctp_getsockopt_maxburst(sk
, len
, optval
, optlen
);
5441 case SCTP_AUTH_CHUNK
:
5442 case SCTP_AUTH_DELETE_KEY
:
5443 retval
= -EOPNOTSUPP
;
5445 case SCTP_HMAC_IDENT
:
5446 retval
= sctp_getsockopt_hmac_ident(sk
, len
, optval
, optlen
);
5448 case SCTP_AUTH_ACTIVE_KEY
:
5449 retval
= sctp_getsockopt_active_key(sk
, len
, optval
, optlen
);
5451 case SCTP_PEER_AUTH_CHUNKS
:
5452 retval
= sctp_getsockopt_peer_auth_chunks(sk
, len
, optval
,
5455 case SCTP_LOCAL_AUTH_CHUNKS
:
5456 retval
= sctp_getsockopt_local_auth_chunks(sk
, len
, optval
,
5459 case SCTP_GET_ASSOC_NUMBER
:
5460 retval
= sctp_getsockopt_assoc_number(sk
, len
, optval
, optlen
);
5462 case SCTP_GET_ASSOC_ID_LIST
:
5463 retval
= sctp_getsockopt_assoc_ids(sk
, len
, optval
, optlen
);
5466 retval
= -ENOPROTOOPT
;
5470 sctp_release_sock(sk
);
5474 static void sctp_hash(struct sock
*sk
)
5479 static void sctp_unhash(struct sock
*sk
)
5484 /* Check if port is acceptable. Possibly find first available port.
5486 * The port hash table (contained in the 'global' SCTP protocol storage
5487 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5488 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5489 * list (the list number is the port number hashed out, so as you
5490 * would expect from a hash function, all the ports in a given list have
5491 * such a number that hashes out to the same list number; you were
5492 * expecting that, right?); so each list has a set of ports, with a
5493 * link to the socket (struct sock) that uses it, the port number and
5494 * a fastreuse flag (FIXME: NPI ipg).
5496 static struct sctp_bind_bucket
*sctp_bucket_create(
5497 struct sctp_bind_hashbucket
*head
, unsigned short snum
);
5499 static long sctp_get_port_local(struct sock
*sk
, union sctp_addr
*addr
)
5501 struct sctp_bind_hashbucket
*head
; /* hash list */
5502 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
5503 struct hlist_node
*node
;
5504 unsigned short snum
;
5507 snum
= ntohs(addr
->v4
.sin_port
);
5509 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum
);
5510 sctp_local_bh_disable();
5513 /* Search for an available port. */
5514 int low
, high
, remaining
, index
;
5517 inet_get_local_port_range(&low
, &high
);
5518 remaining
= (high
- low
) + 1;
5519 rover
= net_random() % remaining
+ low
;
5523 if ((rover
< low
) || (rover
> high
))
5525 if (inet_is_reserved_local_port(rover
))
5527 index
= sctp_phashfn(rover
);
5528 head
= &sctp_port_hashtable
[index
];
5529 sctp_spin_lock(&head
->lock
);
5530 sctp_for_each_hentry(pp
, node
, &head
->chain
)
5531 if (pp
->port
== rover
)
5535 sctp_spin_unlock(&head
->lock
);
5536 } while (--remaining
> 0);
5538 /* Exhausted local port range during search? */
5543 /* OK, here is the one we will use. HEAD (the port
5544 * hash table list entry) is non-NULL and we hold it's
5549 /* We are given an specific port number; we verify
5550 * that it is not being used. If it is used, we will
5551 * exahust the search in the hash list corresponding
5552 * to the port number (snum) - we detect that with the
5553 * port iterator, pp being NULL.
5555 head
= &sctp_port_hashtable
[sctp_phashfn(snum
)];
5556 sctp_spin_lock(&head
->lock
);
5557 sctp_for_each_hentry(pp
, node
, &head
->chain
) {
5558 if (pp
->port
== snum
)
5565 if (!hlist_empty(&pp
->owner
)) {
5566 /* We had a port hash table hit - there is an
5567 * available port (pp != NULL) and it is being
5568 * used by other socket (pp->owner not empty); that other
5569 * socket is going to be sk2.
5571 int reuse
= sk
->sk_reuse
;
5574 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5575 if (pp
->fastreuse
&& sk
->sk_reuse
&&
5576 sk
->sk_state
!= SCTP_SS_LISTENING
)
5579 /* Run through the list of sockets bound to the port
5580 * (pp->port) [via the pointers bind_next and
5581 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5582 * we get the endpoint they describe and run through
5583 * the endpoint's list of IP (v4 or v6) addresses,
5584 * comparing each of the addresses with the address of
5585 * the socket sk. If we find a match, then that means
5586 * that this port/socket (sk) combination are already
5589 sk_for_each_bound(sk2
, node
, &pp
->owner
) {
5590 struct sctp_endpoint
*ep2
;
5591 ep2
= sctp_sk(sk2
)->ep
;
5594 (reuse
&& sk2
->sk_reuse
&&
5595 sk2
->sk_state
!= SCTP_SS_LISTENING
))
5598 if (sctp_bind_addr_conflict(&ep2
->base
.bind_addr
, addr
,
5599 sctp_sk(sk2
), sctp_sk(sk
))) {
5604 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5607 /* If there was a hash table miss, create a new port. */
5609 if (!pp
&& !(pp
= sctp_bucket_create(head
, snum
)))
5612 /* In either case (hit or miss), make sure fastreuse is 1 only
5613 * if sk->sk_reuse is too (that is, if the caller requested
5614 * SO_REUSEADDR on this socket -sk-).
5616 if (hlist_empty(&pp
->owner
)) {
5617 if (sk
->sk_reuse
&& sk
->sk_state
!= SCTP_SS_LISTENING
)
5621 } else if (pp
->fastreuse
&&
5622 (!sk
->sk_reuse
|| sk
->sk_state
== SCTP_SS_LISTENING
))
5625 /* We are set, so fill up all the data in the hash table
5626 * entry, tie the socket list information with the rest of the
5627 * sockets FIXME: Blurry, NPI (ipg).
5630 if (!sctp_sk(sk
)->bind_hash
) {
5631 inet_sk(sk
)->inet_num
= snum
;
5632 sk_add_bind_node(sk
, &pp
->owner
);
5633 sctp_sk(sk
)->bind_hash
= pp
;
5638 sctp_spin_unlock(&head
->lock
);
5641 sctp_local_bh_enable();
5645 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5646 * port is requested.
5648 static int sctp_get_port(struct sock
*sk
, unsigned short snum
)
5651 union sctp_addr addr
;
5652 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
5654 /* Set up a dummy address struct from the sk. */
5655 af
->from_sk(&addr
, sk
);
5656 addr
.v4
.sin_port
= htons(snum
);
5658 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5659 ret
= sctp_get_port_local(sk
, &addr
);
5665 * Move a socket to LISTENING state.
5667 SCTP_STATIC
int sctp_listen_start(struct sock
*sk
, int backlog
)
5669 struct sctp_sock
*sp
= sctp_sk(sk
);
5670 struct sctp_endpoint
*ep
= sp
->ep
;
5671 struct crypto_hash
*tfm
= NULL
;
5673 /* Allocate HMAC for generating cookie. */
5674 if (!sctp_sk(sk
)->hmac
&& sctp_hmac_alg
) {
5675 tfm
= crypto_alloc_hash(sctp_hmac_alg
, 0, CRYPTO_ALG_ASYNC
);
5677 if (net_ratelimit()) {
5678 pr_info("failed to load transform for %s: %ld\n",
5679 sctp_hmac_alg
, PTR_ERR(tfm
));
5683 sctp_sk(sk
)->hmac
= tfm
;
5687 * If a bind() or sctp_bindx() is not called prior to a listen()
5688 * call that allows new associations to be accepted, the system
5689 * picks an ephemeral port and will choose an address set equivalent
5690 * to binding with a wildcard address.
5692 * This is not currently spelled out in the SCTP sockets
5693 * extensions draft, but follows the practice as seen in TCP
5697 sk
->sk_state
= SCTP_SS_LISTENING
;
5698 if (!ep
->base
.bind_addr
.port
) {
5699 if (sctp_autobind(sk
))
5702 if (sctp_get_port(sk
, inet_sk(sk
)->inet_num
)) {
5703 sk
->sk_state
= SCTP_SS_CLOSED
;
5708 sk
->sk_max_ack_backlog
= backlog
;
5709 sctp_hash_endpoint(ep
);
5714 * 4.1.3 / 5.1.3 listen()
5716 * By default, new associations are not accepted for UDP style sockets.
5717 * An application uses listen() to mark a socket as being able to
5718 * accept new associations.
5720 * On TCP style sockets, applications use listen() to ready the SCTP
5721 * endpoint for accepting inbound associations.
5723 * On both types of endpoints a backlog of '0' disables listening.
5725 * Move a socket to LISTENING state.
5727 int sctp_inet_listen(struct socket
*sock
, int backlog
)
5729 struct sock
*sk
= sock
->sk
;
5730 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5733 if (unlikely(backlog
< 0))
5738 /* Peeled-off sockets are not allowed to listen(). */
5739 if (sctp_style(sk
, UDP_HIGH_BANDWIDTH
))
5742 if (sock
->state
!= SS_UNCONNECTED
)
5745 /* If backlog is zero, disable listening. */
5747 if (sctp_sstate(sk
, CLOSED
))
5751 sctp_unhash_endpoint(ep
);
5752 sk
->sk_state
= SCTP_SS_CLOSED
;
5754 sctp_sk(sk
)->bind_hash
->fastreuse
= 1;
5758 /* If we are already listening, just update the backlog */
5759 if (sctp_sstate(sk
, LISTENING
))
5760 sk
->sk_max_ack_backlog
= backlog
;
5762 err
= sctp_listen_start(sk
, backlog
);
5769 sctp_release_sock(sk
);
5774 * This function is done by modeling the current datagram_poll() and the
5775 * tcp_poll(). Note that, based on these implementations, we don't
5776 * lock the socket in this function, even though it seems that,
5777 * ideally, locking or some other mechanisms can be used to ensure
5778 * the integrity of the counters (sndbuf and wmem_alloc) used
5779 * in this place. We assume that we don't need locks either until proven
5782 * Another thing to note is that we include the Async I/O support
5783 * here, again, by modeling the current TCP/UDP code. We don't have
5784 * a good way to test with it yet.
5786 unsigned int sctp_poll(struct file
*file
, struct socket
*sock
, poll_table
*wait
)
5788 struct sock
*sk
= sock
->sk
;
5789 struct sctp_sock
*sp
= sctp_sk(sk
);
5792 poll_wait(file
, sk_sleep(sk
), wait
);
5794 /* A TCP-style listening socket becomes readable when the accept queue
5797 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
5798 return (!list_empty(&sp
->ep
->asocs
)) ?
5799 (POLLIN
| POLLRDNORM
) : 0;
5803 /* Is there any exceptional events? */
5804 if (sk
->sk_err
|| !skb_queue_empty(&sk
->sk_error_queue
))
5806 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
5807 mask
|= POLLRDHUP
| POLLIN
| POLLRDNORM
;
5808 if (sk
->sk_shutdown
== SHUTDOWN_MASK
)
5811 /* Is it readable? Reconsider this code with TCP-style support. */
5812 if (!skb_queue_empty(&sk
->sk_receive_queue
))
5813 mask
|= POLLIN
| POLLRDNORM
;
5815 /* The association is either gone or not ready. */
5816 if (!sctp_style(sk
, UDP
) && sctp_sstate(sk
, CLOSED
))
5819 /* Is it writable? */
5820 if (sctp_writeable(sk
)) {
5821 mask
|= POLLOUT
| POLLWRNORM
;
5823 set_bit(SOCK_ASYNC_NOSPACE
, &sk
->sk_socket
->flags
);
5825 * Since the socket is not locked, the buffer
5826 * might be made available after the writeable check and
5827 * before the bit is set. This could cause a lost I/O
5828 * signal. tcp_poll() has a race breaker for this race
5829 * condition. Based on their implementation, we put
5830 * in the following code to cover it as well.
5832 if (sctp_writeable(sk
))
5833 mask
|= POLLOUT
| POLLWRNORM
;
5838 /********************************************************************
5839 * 2nd Level Abstractions
5840 ********************************************************************/
5842 static struct sctp_bind_bucket
*sctp_bucket_create(
5843 struct sctp_bind_hashbucket
*head
, unsigned short snum
)
5845 struct sctp_bind_bucket
*pp
;
5847 pp
= kmem_cache_alloc(sctp_bucket_cachep
, GFP_ATOMIC
);
5849 SCTP_DBG_OBJCNT_INC(bind_bucket
);
5852 INIT_HLIST_HEAD(&pp
->owner
);
5853 hlist_add_head(&pp
->node
, &head
->chain
);
5858 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5859 static void sctp_bucket_destroy(struct sctp_bind_bucket
*pp
)
5861 if (pp
&& hlist_empty(&pp
->owner
)) {
5862 __hlist_del(&pp
->node
);
5863 kmem_cache_free(sctp_bucket_cachep
, pp
);
5864 SCTP_DBG_OBJCNT_DEC(bind_bucket
);
5868 /* Release this socket's reference to a local port. */
5869 static inline void __sctp_put_port(struct sock
*sk
)
5871 struct sctp_bind_hashbucket
*head
=
5872 &sctp_port_hashtable
[sctp_phashfn(inet_sk(sk
)->inet_num
)];
5873 struct sctp_bind_bucket
*pp
;
5875 sctp_spin_lock(&head
->lock
);
5876 pp
= sctp_sk(sk
)->bind_hash
;
5877 __sk_del_bind_node(sk
);
5878 sctp_sk(sk
)->bind_hash
= NULL
;
5879 inet_sk(sk
)->inet_num
= 0;
5880 sctp_bucket_destroy(pp
);
5881 sctp_spin_unlock(&head
->lock
);
5884 void sctp_put_port(struct sock
*sk
)
5886 sctp_local_bh_disable();
5887 __sctp_put_port(sk
);
5888 sctp_local_bh_enable();
5892 * The system picks an ephemeral port and choose an address set equivalent
5893 * to binding with a wildcard address.
5894 * One of those addresses will be the primary address for the association.
5895 * This automatically enables the multihoming capability of SCTP.
5897 static int sctp_autobind(struct sock
*sk
)
5899 union sctp_addr autoaddr
;
5903 /* Initialize a local sockaddr structure to INADDR_ANY. */
5904 af
= sctp_sk(sk
)->pf
->af
;
5906 port
= htons(inet_sk(sk
)->inet_num
);
5907 af
->inaddr_any(&autoaddr
, port
);
5909 return sctp_do_bind(sk
, &autoaddr
, af
->sockaddr_len
);
5912 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5915 * 4.2 The cmsghdr Structure *
5917 * When ancillary data is sent or received, any number of ancillary data
5918 * objects can be specified by the msg_control and msg_controllen members of
5919 * the msghdr structure, because each object is preceded by
5920 * a cmsghdr structure defining the object's length (the cmsg_len member).
5921 * Historically Berkeley-derived implementations have passed only one object
5922 * at a time, but this API allows multiple objects to be
5923 * passed in a single call to sendmsg() or recvmsg(). The following example
5924 * shows two ancillary data objects in a control buffer.
5926 * |<--------------------------- msg_controllen -------------------------->|
5929 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5931 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5934 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5936 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5939 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5940 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5942 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5944 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5951 SCTP_STATIC
int sctp_msghdr_parse(const struct msghdr
*msg
,
5952 sctp_cmsgs_t
*cmsgs
)
5954 struct cmsghdr
*cmsg
;
5955 struct msghdr
*my_msg
= (struct msghdr
*)msg
;
5957 for (cmsg
= CMSG_FIRSTHDR(msg
);
5959 cmsg
= CMSG_NXTHDR(my_msg
, cmsg
)) {
5960 if (!CMSG_OK(my_msg
, cmsg
))
5963 /* Should we parse this header or ignore? */
5964 if (cmsg
->cmsg_level
!= IPPROTO_SCTP
)
5967 /* Strictly check lengths following example in SCM code. */
5968 switch (cmsg
->cmsg_type
) {
5970 /* SCTP Socket API Extension
5971 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5973 * This cmsghdr structure provides information for
5974 * initializing new SCTP associations with sendmsg().
5975 * The SCTP_INITMSG socket option uses this same data
5976 * structure. This structure is not used for
5979 * cmsg_level cmsg_type cmsg_data[]
5980 * ------------ ------------ ----------------------
5981 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5983 if (cmsg
->cmsg_len
!=
5984 CMSG_LEN(sizeof(struct sctp_initmsg
)))
5986 cmsgs
->init
= (struct sctp_initmsg
*)CMSG_DATA(cmsg
);
5990 /* SCTP Socket API Extension
5991 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5993 * This cmsghdr structure specifies SCTP options for
5994 * sendmsg() and describes SCTP header information
5995 * about a received message through recvmsg().
5997 * cmsg_level cmsg_type cmsg_data[]
5998 * ------------ ------------ ----------------------
5999 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6001 if (cmsg
->cmsg_len
!=
6002 CMSG_LEN(sizeof(struct sctp_sndrcvinfo
)))
6006 (struct sctp_sndrcvinfo
*)CMSG_DATA(cmsg
);
6008 /* Minimally, validate the sinfo_flags. */
6009 if (cmsgs
->info
->sinfo_flags
&
6010 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
6011 SCTP_ABORT
| SCTP_EOF
))
6023 * Wait for a packet..
6024 * Note: This function is the same function as in core/datagram.c
6025 * with a few modifications to make lksctp work.
6027 static int sctp_wait_for_packet(struct sock
* sk
, int *err
, long *timeo_p
)
6032 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
6034 /* Socket errors? */
6035 error
= sock_error(sk
);
6039 if (!skb_queue_empty(&sk
->sk_receive_queue
))
6042 /* Socket shut down? */
6043 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6046 /* Sequenced packets can come disconnected. If so we report the
6051 /* Is there a good reason to think that we may receive some data? */
6052 if (list_empty(&sctp_sk(sk
)->ep
->asocs
) && !sctp_sstate(sk
, LISTENING
))
6055 /* Handle signals. */
6056 if (signal_pending(current
))
6059 /* Let another process have a go. Since we are going to sleep
6060 * anyway. Note: This may cause odd behaviors if the message
6061 * does not fit in the user's buffer, but this seems to be the
6062 * only way to honor MSG_DONTWAIT realistically.
6064 sctp_release_sock(sk
);
6065 *timeo_p
= schedule_timeout(*timeo_p
);
6069 finish_wait(sk_sleep(sk
), &wait
);
6073 error
= sock_intr_errno(*timeo_p
);
6076 finish_wait(sk_sleep(sk
), &wait
);
6081 /* Receive a datagram.
6082 * Note: This is pretty much the same routine as in core/datagram.c
6083 * with a few changes to make lksctp work.
6085 static struct sk_buff
*sctp_skb_recv_datagram(struct sock
*sk
, int flags
,
6086 int noblock
, int *err
)
6089 struct sk_buff
*skb
;
6092 timeo
= sock_rcvtimeo(sk
, noblock
);
6094 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6095 timeo
, MAX_SCHEDULE_TIMEOUT
);
6098 /* Again only user level code calls this function,
6099 * so nothing interrupt level
6100 * will suddenly eat the receive_queue.
6102 * Look at current nfs client by the way...
6103 * However, this function was correct in any case. 8)
6105 if (flags
& MSG_PEEK
) {
6106 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
6107 skb
= skb_peek(&sk
->sk_receive_queue
);
6109 atomic_inc(&skb
->users
);
6110 spin_unlock_bh(&sk
->sk_receive_queue
.lock
);
6112 skb
= skb_dequeue(&sk
->sk_receive_queue
);
6118 /* Caller is allowed not to check sk->sk_err before calling. */
6119 error
= sock_error(sk
);
6123 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6126 /* User doesn't want to wait. */
6130 } while (sctp_wait_for_packet(sk
, err
, &timeo
) == 0);
6139 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6140 static void __sctp_write_space(struct sctp_association
*asoc
)
6142 struct sock
*sk
= asoc
->base
.sk
;
6143 struct socket
*sock
= sk
->sk_socket
;
6145 if ((sctp_wspace(asoc
) > 0) && sock
) {
6146 if (waitqueue_active(&asoc
->wait
))
6147 wake_up_interruptible(&asoc
->wait
);
6149 if (sctp_writeable(sk
)) {
6150 wait_queue_head_t
*wq
= sk_sleep(sk
);
6152 if (wq
&& waitqueue_active(wq
))
6153 wake_up_interruptible(wq
);
6155 /* Note that we try to include the Async I/O support
6156 * here by modeling from the current TCP/UDP code.
6157 * We have not tested with it yet.
6159 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
))
6160 sock_wake_async(sock
,
6161 SOCK_WAKE_SPACE
, POLL_OUT
);
6166 /* Do accounting for the sndbuf space.
6167 * Decrement the used sndbuf space of the corresponding association by the
6168 * data size which was just transmitted(freed).
6170 static void sctp_wfree(struct sk_buff
*skb
)
6172 struct sctp_association
*asoc
;
6173 struct sctp_chunk
*chunk
;
6176 /* Get the saved chunk pointer. */
6177 chunk
= *((struct sctp_chunk
**)(skb
->cb
));
6180 asoc
->sndbuf_used
-= SCTP_DATA_SNDSIZE(chunk
) +
6181 sizeof(struct sk_buff
) +
6182 sizeof(struct sctp_chunk
);
6184 atomic_sub(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
6187 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6189 sk
->sk_wmem_queued
-= skb
->truesize
;
6190 sk_mem_uncharge(sk
, skb
->truesize
);
6193 __sctp_write_space(asoc
);
6195 sctp_association_put(asoc
);
6198 /* Do accounting for the receive space on the socket.
6199 * Accounting for the association is done in ulpevent.c
6200 * We set this as a destructor for the cloned data skbs so that
6201 * accounting is done at the correct time.
6203 void sctp_sock_rfree(struct sk_buff
*skb
)
6205 struct sock
*sk
= skb
->sk
;
6206 struct sctp_ulpevent
*event
= sctp_skb2event(skb
);
6208 atomic_sub(event
->rmem_len
, &sk
->sk_rmem_alloc
);
6211 * Mimic the behavior of sock_rfree
6213 sk_mem_uncharge(sk
, event
->rmem_len
);
6217 /* Helper function to wait for space in the sndbuf. */
6218 static int sctp_wait_for_sndbuf(struct sctp_association
*asoc
, long *timeo_p
,
6221 struct sock
*sk
= asoc
->base
.sk
;
6223 long current_timeo
= *timeo_p
;
6226 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6227 asoc
, (long)(*timeo_p
), msg_len
);
6229 /* Increment the association's refcnt. */
6230 sctp_association_hold(asoc
);
6232 /* Wait on the association specific sndbuf space. */
6234 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
6235 TASK_INTERRUPTIBLE
);
6238 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
6241 if (signal_pending(current
))
6242 goto do_interrupted
;
6243 if (msg_len
<= sctp_wspace(asoc
))
6246 /* Let another process have a go. Since we are going
6249 sctp_release_sock(sk
);
6250 current_timeo
= schedule_timeout(current_timeo
);
6251 BUG_ON(sk
!= asoc
->base
.sk
);
6254 *timeo_p
= current_timeo
;
6258 finish_wait(&asoc
->wait
, &wait
);
6260 /* Release the association's refcnt. */
6261 sctp_association_put(asoc
);
6270 err
= sock_intr_errno(*timeo_p
);
6278 void sctp_data_ready(struct sock
*sk
, int len
)
6280 struct socket_wq
*wq
;
6283 wq
= rcu_dereference(sk
->sk_wq
);
6284 if (wq_has_sleeper(wq
))
6285 wake_up_interruptible_sync_poll(&wq
->wait
, POLLIN
|
6286 POLLRDNORM
| POLLRDBAND
);
6287 sk_wake_async(sk
, SOCK_WAKE_WAITD
, POLL_IN
);
6291 /* If socket sndbuf has changed, wake up all per association waiters. */
6292 void sctp_write_space(struct sock
*sk
)
6294 struct sctp_association
*asoc
;
6296 /* Wake up the tasks in each wait queue. */
6297 list_for_each_entry(asoc
, &((sctp_sk(sk
))->ep
->asocs
), asocs
) {
6298 __sctp_write_space(asoc
);
6302 /* Is there any sndbuf space available on the socket?
6304 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6305 * associations on the same socket. For a UDP-style socket with
6306 * multiple associations, it is possible for it to be "unwriteable"
6307 * prematurely. I assume that this is acceptable because
6308 * a premature "unwriteable" is better than an accidental "writeable" which
6309 * would cause an unwanted block under certain circumstances. For the 1-1
6310 * UDP-style sockets or TCP-style sockets, this code should work.
6313 static int sctp_writeable(struct sock
*sk
)
6317 amt
= sk
->sk_sndbuf
- sk_wmem_alloc_get(sk
);
6323 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6324 * returns immediately with EINPROGRESS.
6326 static int sctp_wait_for_connect(struct sctp_association
*asoc
, long *timeo_p
)
6328 struct sock
*sk
= asoc
->base
.sk
;
6330 long current_timeo
= *timeo_p
;
6333 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__
, asoc
,
6336 /* Increment the association's refcnt. */
6337 sctp_association_hold(asoc
);
6340 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
6341 TASK_INTERRUPTIBLE
);
6344 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6346 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
6349 if (signal_pending(current
))
6350 goto do_interrupted
;
6352 if (sctp_state(asoc
, ESTABLISHED
))
6355 /* Let another process have a go. Since we are going
6358 sctp_release_sock(sk
);
6359 current_timeo
= schedule_timeout(current_timeo
);
6362 *timeo_p
= current_timeo
;
6366 finish_wait(&asoc
->wait
, &wait
);
6368 /* Release the association's refcnt. */
6369 sctp_association_put(asoc
);
6374 if (asoc
->init_err_counter
+ 1 > asoc
->max_init_attempts
)
6377 err
= -ECONNREFUSED
;
6381 err
= sock_intr_errno(*timeo_p
);
6389 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
)
6391 struct sctp_endpoint
*ep
;
6395 ep
= sctp_sk(sk
)->ep
;
6399 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
,
6400 TASK_INTERRUPTIBLE
);
6402 if (list_empty(&ep
->asocs
)) {
6403 sctp_release_sock(sk
);
6404 timeo
= schedule_timeout(timeo
);
6409 if (!sctp_sstate(sk
, LISTENING
))
6413 if (!list_empty(&ep
->asocs
))
6416 err
= sock_intr_errno(timeo
);
6417 if (signal_pending(current
))
6425 finish_wait(sk_sleep(sk
), &wait
);
6430 static void sctp_wait_for_close(struct sock
*sk
, long timeout
)
6435 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
6436 if (list_empty(&sctp_sk(sk
)->ep
->asocs
))
6438 sctp_release_sock(sk
);
6439 timeout
= schedule_timeout(timeout
);
6441 } while (!signal_pending(current
) && timeout
);
6443 finish_wait(sk_sleep(sk
), &wait
);
6446 static void sctp_skb_set_owner_r_frag(struct sk_buff
*skb
, struct sock
*sk
)
6448 struct sk_buff
*frag
;
6453 /* Don't forget the fragments. */
6454 skb_walk_frags(skb
, frag
)
6455 sctp_skb_set_owner_r_frag(frag
, sk
);
6458 sctp_skb_set_owner_r(skb
, sk
);
6461 void sctp_copy_sock(struct sock
*newsk
, struct sock
*sk
,
6462 struct sctp_association
*asoc
)
6464 struct inet_sock
*inet
= inet_sk(sk
);
6465 struct inet_sock
*newinet
;
6467 newsk
->sk_type
= sk
->sk_type
;
6468 newsk
->sk_bound_dev_if
= sk
->sk_bound_dev_if
;
6469 newsk
->sk_flags
= sk
->sk_flags
;
6470 newsk
->sk_no_check
= sk
->sk_no_check
;
6471 newsk
->sk_reuse
= sk
->sk_reuse
;
6473 newsk
->sk_shutdown
= sk
->sk_shutdown
;
6474 newsk
->sk_destruct
= inet_sock_destruct
;
6475 newsk
->sk_family
= sk
->sk_family
;
6476 newsk
->sk_protocol
= IPPROTO_SCTP
;
6477 newsk
->sk_backlog_rcv
= sk
->sk_prot
->backlog_rcv
;
6478 newsk
->sk_sndbuf
= sk
->sk_sndbuf
;
6479 newsk
->sk_rcvbuf
= sk
->sk_rcvbuf
;
6480 newsk
->sk_lingertime
= sk
->sk_lingertime
;
6481 newsk
->sk_rcvtimeo
= sk
->sk_rcvtimeo
;
6482 newsk
->sk_sndtimeo
= sk
->sk_sndtimeo
;
6484 newinet
= inet_sk(newsk
);
6486 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6487 * getsockname() and getpeername()
6489 newinet
->inet_sport
= inet
->inet_sport
;
6490 newinet
->inet_saddr
= inet
->inet_saddr
;
6491 newinet
->inet_rcv_saddr
= inet
->inet_rcv_saddr
;
6492 newinet
->inet_dport
= htons(asoc
->peer
.port
);
6493 newinet
->pmtudisc
= inet
->pmtudisc
;
6494 newinet
->inet_id
= asoc
->next_tsn
^ jiffies
;
6496 newinet
->uc_ttl
= inet
->uc_ttl
;
6497 newinet
->mc_loop
= 1;
6498 newinet
->mc_ttl
= 1;
6499 newinet
->mc_index
= 0;
6500 newinet
->mc_list
= NULL
;
6503 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6504 * and its messages to the newsk.
6506 static void sctp_sock_migrate(struct sock
*oldsk
, struct sock
*newsk
,
6507 struct sctp_association
*assoc
,
6508 sctp_socket_type_t type
)
6510 struct sctp_sock
*oldsp
= sctp_sk(oldsk
);
6511 struct sctp_sock
*newsp
= sctp_sk(newsk
);
6512 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
6513 struct sctp_endpoint
*newep
= newsp
->ep
;
6514 struct sk_buff
*skb
, *tmp
;
6515 struct sctp_ulpevent
*event
;
6516 struct sctp_bind_hashbucket
*head
;
6518 /* Migrate socket buffer sizes and all the socket level options to the
6521 newsk
->sk_sndbuf
= oldsk
->sk_sndbuf
;
6522 newsk
->sk_rcvbuf
= oldsk
->sk_rcvbuf
;
6523 /* Brute force copy old sctp opt. */
6524 inet_sk_copy_descendant(newsk
, oldsk
);
6526 /* Restore the ep value that was overwritten with the above structure
6532 /* Hook this new socket in to the bind_hash list. */
6533 head
= &sctp_port_hashtable
[sctp_phashfn(inet_sk(oldsk
)->inet_num
)];
6534 sctp_local_bh_disable();
6535 sctp_spin_lock(&head
->lock
);
6536 pp
= sctp_sk(oldsk
)->bind_hash
;
6537 sk_add_bind_node(newsk
, &pp
->owner
);
6538 sctp_sk(newsk
)->bind_hash
= pp
;
6539 inet_sk(newsk
)->inet_num
= inet_sk(oldsk
)->inet_num
;
6540 sctp_spin_unlock(&head
->lock
);
6541 sctp_local_bh_enable();
6543 /* Copy the bind_addr list from the original endpoint to the new
6544 * endpoint so that we can handle restarts properly
6546 sctp_bind_addr_dup(&newsp
->ep
->base
.bind_addr
,
6547 &oldsp
->ep
->base
.bind_addr
, GFP_KERNEL
);
6549 /* Move any messages in the old socket's receive queue that are for the
6550 * peeled off association to the new socket's receive queue.
6552 sctp_skb_for_each(skb
, &oldsk
->sk_receive_queue
, tmp
) {
6553 event
= sctp_skb2event(skb
);
6554 if (event
->asoc
== assoc
) {
6555 __skb_unlink(skb
, &oldsk
->sk_receive_queue
);
6556 __skb_queue_tail(&newsk
->sk_receive_queue
, skb
);
6557 sctp_skb_set_owner_r_frag(skb
, newsk
);
6561 /* Clean up any messages pending delivery due to partial
6562 * delivery. Three cases:
6563 * 1) No partial deliver; no work.
6564 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6565 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6567 skb_queue_head_init(&newsp
->pd_lobby
);
6568 atomic_set(&sctp_sk(newsk
)->pd_mode
, assoc
->ulpq
.pd_mode
);
6570 if (atomic_read(&sctp_sk(oldsk
)->pd_mode
)) {
6571 struct sk_buff_head
*queue
;
6573 /* Decide which queue to move pd_lobby skbs to. */
6574 if (assoc
->ulpq
.pd_mode
) {
6575 queue
= &newsp
->pd_lobby
;
6577 queue
= &newsk
->sk_receive_queue
;
6579 /* Walk through the pd_lobby, looking for skbs that
6580 * need moved to the new socket.
6582 sctp_skb_for_each(skb
, &oldsp
->pd_lobby
, tmp
) {
6583 event
= sctp_skb2event(skb
);
6584 if (event
->asoc
== assoc
) {
6585 __skb_unlink(skb
, &oldsp
->pd_lobby
);
6586 __skb_queue_tail(queue
, skb
);
6587 sctp_skb_set_owner_r_frag(skb
, newsk
);
6591 /* Clear up any skbs waiting for the partial
6592 * delivery to finish.
6594 if (assoc
->ulpq
.pd_mode
)
6595 sctp_clear_pd(oldsk
, NULL
);
6599 sctp_skb_for_each(skb
, &assoc
->ulpq
.reasm
, tmp
)
6600 sctp_skb_set_owner_r_frag(skb
, newsk
);
6602 sctp_skb_for_each(skb
, &assoc
->ulpq
.lobby
, tmp
)
6603 sctp_skb_set_owner_r_frag(skb
, newsk
);
6605 /* Set the type of socket to indicate that it is peeled off from the
6606 * original UDP-style socket or created with the accept() call on a
6607 * TCP-style socket..
6611 /* Mark the new socket "in-use" by the user so that any packets
6612 * that may arrive on the association after we've moved it are
6613 * queued to the backlog. This prevents a potential race between
6614 * backlog processing on the old socket and new-packet processing
6615 * on the new socket.
6617 * The caller has just allocated newsk so we can guarantee that other
6618 * paths won't try to lock it and then oldsk.
6620 lock_sock_nested(newsk
, SINGLE_DEPTH_NESTING
);
6621 sctp_assoc_migrate(assoc
, newsk
);
6623 /* If the association on the newsk is already closed before accept()
6624 * is called, set RCV_SHUTDOWN flag.
6626 if (sctp_state(assoc
, CLOSED
) && sctp_style(newsk
, TCP
))
6627 newsk
->sk_shutdown
|= RCV_SHUTDOWN
;
6629 newsk
->sk_state
= SCTP_SS_ESTABLISHED
;
6630 sctp_release_sock(newsk
);
6634 /* This proto struct describes the ULP interface for SCTP. */
6635 struct proto sctp_prot
= {
6637 .owner
= THIS_MODULE
,
6638 .close
= sctp_close
,
6639 .connect
= sctp_connect
,
6640 .disconnect
= sctp_disconnect
,
6641 .accept
= sctp_accept
,
6642 .ioctl
= sctp_ioctl
,
6643 .init
= sctp_init_sock
,
6644 .destroy
= sctp_destroy_sock
,
6645 .shutdown
= sctp_shutdown
,
6646 .setsockopt
= sctp_setsockopt
,
6647 .getsockopt
= sctp_getsockopt
,
6648 .sendmsg
= sctp_sendmsg
,
6649 .recvmsg
= sctp_recvmsg
,
6651 .backlog_rcv
= sctp_backlog_rcv
,
6653 .unhash
= sctp_unhash
,
6654 .get_port
= sctp_get_port
,
6655 .obj_size
= sizeof(struct sctp_sock
),
6656 .sysctl_mem
= sysctl_sctp_mem
,
6657 .sysctl_rmem
= sysctl_sctp_rmem
,
6658 .sysctl_wmem
= sysctl_sctp_wmem
,
6659 .memory_pressure
= &sctp_memory_pressure
,
6660 .enter_memory_pressure
= sctp_enter_memory_pressure
,
6661 .memory_allocated
= &sctp_memory_allocated
,
6662 .sockets_allocated
= &sctp_sockets_allocated
,
6665 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6667 struct proto sctpv6_prot
= {
6669 .owner
= THIS_MODULE
,
6670 .close
= sctp_close
,
6671 .connect
= sctp_connect
,
6672 .disconnect
= sctp_disconnect
,
6673 .accept
= sctp_accept
,
6674 .ioctl
= sctp_ioctl
,
6675 .init
= sctp_init_sock
,
6676 .destroy
= sctp_destroy_sock
,
6677 .shutdown
= sctp_shutdown
,
6678 .setsockopt
= sctp_setsockopt
,
6679 .getsockopt
= sctp_getsockopt
,
6680 .sendmsg
= sctp_sendmsg
,
6681 .recvmsg
= sctp_recvmsg
,
6683 .backlog_rcv
= sctp_backlog_rcv
,
6685 .unhash
= sctp_unhash
,
6686 .get_port
= sctp_get_port
,
6687 .obj_size
= sizeof(struct sctp6_sock
),
6688 .sysctl_mem
= sysctl_sctp_mem
,
6689 .sysctl_rmem
= sysctl_sctp_rmem
,
6690 .sysctl_wmem
= sysctl_sctp_wmem
,
6691 .memory_pressure
= &sctp_memory_pressure
,
6692 .enter_memory_pressure
= sctp_enter_memory_pressure
,
6693 .memory_allocated
= &sctp_memory_allocated
,
6694 .sockets_allocated
= &sctp_sockets_allocated
,
6696 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */