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
!= htons(bp
->port
)) {
666 /* FIXME - There is probably a need to check if sk->sk_saddr and
667 * sk->sk_rcv_addr are currently set to one of the addresses to
668 * be removed. This is something which needs to be looked into
669 * when we are fixing the outstanding issues with multi-homing
670 * socket routing and failover schemes. Refer to comments in
671 * sctp_do_bind(). -daisy
673 retval
= sctp_del_bind_addr(bp
, sa_addr
);
675 addr_buf
+= af
->sockaddr_len
;
678 /* Failed. Add the ones that has been removed back */
680 sctp_bindx_add(sk
, addrs
, cnt
);
688 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
689 * the associations that are part of the endpoint indicating that a list of
690 * local addresses are removed from the endpoint.
692 * If any of the addresses is already in the bind address list of the
693 * association, we do not send the chunk for that association. But it will not
694 * affect other associations.
696 * Only sctp_setsockopt_bindx() is supposed to call this function.
698 static int sctp_send_asconf_del_ip(struct sock
*sk
,
699 struct sockaddr
*addrs
,
702 struct sctp_sock
*sp
;
703 struct sctp_endpoint
*ep
;
704 struct sctp_association
*asoc
;
705 struct sctp_transport
*transport
;
706 struct sctp_bind_addr
*bp
;
707 struct sctp_chunk
*chunk
;
708 union sctp_addr
*laddr
;
711 struct sctp_sockaddr_entry
*saddr
;
715 if (!sctp_addip_enable
)
721 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
722 __func__
, sk
, addrs
, addrcnt
);
724 list_for_each_entry(asoc
, &ep
->asocs
, asocs
) {
726 if (!asoc
->peer
.asconf_capable
)
729 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_DEL_IP
)
732 if (!sctp_state(asoc
, ESTABLISHED
))
735 /* Check if any address in the packed array of addresses is
736 * not present in the bind address list of the association.
737 * If so, do not send the asconf chunk to its peer, but
738 * continue with other associations.
741 for (i
= 0; i
< addrcnt
; i
++) {
742 laddr
= (union sctp_addr
*)addr_buf
;
743 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
749 if (!sctp_assoc_lookup_laddr(asoc
, laddr
))
752 addr_buf
+= af
->sockaddr_len
;
757 /* Find one address in the association's bind address list
758 * that is not in the packed array of addresses. This is to
759 * make sure that we do not delete all the addresses in the
762 bp
= &asoc
->base
.bind_addr
;
763 laddr
= sctp_find_unmatch_addr(bp
, (union sctp_addr
*)addrs
,
768 /* We do not need RCU protection throughout this loop
769 * because this is done under a socket lock from the
772 chunk
= sctp_make_asconf_update_ip(asoc
, laddr
, addrs
, addrcnt
,
779 /* Reset use_as_src flag for the addresses in the bind address
780 * list that are to be deleted.
783 for (i
= 0; i
< addrcnt
; i
++) {
784 laddr
= (union sctp_addr
*)addr_buf
;
785 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
786 list_for_each_entry(saddr
, &bp
->address_list
, list
) {
787 if (sctp_cmp_addr_exact(&saddr
->a
, laddr
))
788 saddr
->state
= SCTP_ADDR_DEL
;
790 addr_buf
+= af
->sockaddr_len
;
793 /* Update the route and saddr entries for all the transports
794 * as some of the addresses in the bind address list are
795 * about to be deleted and cannot be used as source addresses.
797 list_for_each_entry(transport
, &asoc
->peer
.transport_addr_list
,
799 dst_release(transport
->dst
);
800 sctp_transport_route(transport
, NULL
,
801 sctp_sk(asoc
->base
.sk
));
804 retval
= sctp_send_asconf(asoc
, chunk
);
810 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
813 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
816 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
817 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
820 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
821 * Section 3.1.2 for this usage.
823 * addrs is a pointer to an array of one or more socket addresses. Each
824 * address is contained in its appropriate structure (i.e. struct
825 * sockaddr_in or struct sockaddr_in6) the family of the address type
826 * must be used to distinguish the address length (note that this
827 * representation is termed a "packed array" of addresses). The caller
828 * specifies the number of addresses in the array with addrcnt.
830 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
831 * -1, and sets errno to the appropriate error code.
833 * For SCTP, the port given in each socket address must be the same, or
834 * sctp_bindx() will fail, setting errno to EINVAL.
836 * The flags parameter is formed from the bitwise OR of zero or more of
837 * the following currently defined flags:
839 * SCTP_BINDX_ADD_ADDR
841 * SCTP_BINDX_REM_ADDR
843 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
844 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
845 * addresses from the association. The two flags are mutually exclusive;
846 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
847 * not remove all addresses from an association; sctp_bindx() will
848 * reject such an attempt with EINVAL.
850 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
851 * additional addresses with an endpoint after calling bind(). Or use
852 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
853 * socket is associated with so that no new association accepted will be
854 * associated with those addresses. If the endpoint supports dynamic
855 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
856 * endpoint to send the appropriate message to the peer to change the
857 * peers address lists.
859 * Adding and removing addresses from a connected association is
860 * optional functionality. Implementations that do not support this
861 * functionality should return EOPNOTSUPP.
863 * Basically do nothing but copying the addresses from user to kernel
864 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
865 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
868 * We don't use copy_from_user() for optimization: we first do the
869 * sanity checks (buffer size -fast- and access check-healthy
870 * pointer); if all of those succeed, then we can alloc the memory
871 * (expensive operation) needed to copy the data to kernel. Then we do
872 * the copying without checking the user space area
873 * (__copy_from_user()).
875 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
878 * sk The sk of the socket
879 * addrs The pointer to the addresses in user land
880 * addrssize Size of the addrs buffer
881 * op Operation to perform (add or remove, see the flags of
884 * Returns 0 if ok, <0 errno code on error.
886 SCTP_STATIC
int sctp_setsockopt_bindx(struct sock
* sk
,
887 struct sockaddr __user
*addrs
,
888 int addrs_size
, int op
)
890 struct sockaddr
*kaddrs
;
894 struct sockaddr
*sa_addr
;
898 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
899 " addrs_size %d opt %d\n", sk
, addrs
, addrs_size
, op
);
901 if (unlikely(addrs_size
<= 0))
904 /* Check the user passed a healthy pointer. */
905 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
908 /* Alloc space for the address array in kernel memory. */
909 kaddrs
= kmalloc(addrs_size
, GFP_KERNEL
);
910 if (unlikely(!kaddrs
))
913 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
918 /* Walk through the addrs buffer and count the number of addresses. */
920 while (walk_size
< addrs_size
) {
921 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
926 sa_addr
= (struct sockaddr
*)addr_buf
;
927 af
= sctp_get_af_specific(sa_addr
->sa_family
);
929 /* If the address family is not supported or if this address
930 * causes the address buffer to overflow return EINVAL.
932 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
937 addr_buf
+= af
->sockaddr_len
;
938 walk_size
+= af
->sockaddr_len
;
943 case SCTP_BINDX_ADD_ADDR
:
944 err
= sctp_bindx_add(sk
, kaddrs
, addrcnt
);
947 err
= sctp_send_asconf_add_ip(sk
, kaddrs
, addrcnt
);
950 case SCTP_BINDX_REM_ADDR
:
951 err
= sctp_bindx_rem(sk
, kaddrs
, addrcnt
);
954 err
= sctp_send_asconf_del_ip(sk
, kaddrs
, addrcnt
);
968 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
970 * Common routine for handling connect() and sctp_connectx().
971 * Connect will come in with just a single address.
973 static int __sctp_connect(struct sock
* sk
,
974 struct sockaddr
*kaddrs
,
976 sctp_assoc_t
*assoc_id
)
978 struct sctp_sock
*sp
;
979 struct sctp_endpoint
*ep
;
980 struct sctp_association
*asoc
= NULL
;
981 struct sctp_association
*asoc2
;
982 struct sctp_transport
*transport
;
990 union sctp_addr
*sa_addr
= NULL
;
993 unsigned int f_flags
= 0;
998 /* connect() cannot be done on a socket that is already in ESTABLISHED
999 * state - UDP-style peeled off socket or a TCP-style socket that
1000 * is already connected.
1001 * It cannot be done even on a TCP-style listening socket.
1003 if (sctp_sstate(sk
, ESTABLISHED
) ||
1004 (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))) {
1009 /* Walk through the addrs buffer and count the number of addresses. */
1011 while (walk_size
< addrs_size
) {
1012 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
1017 sa_addr
= (union sctp_addr
*)addr_buf
;
1018 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
1020 /* If the address family is not supported or if this address
1021 * causes the address buffer to overflow return EINVAL.
1023 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
1028 port
= ntohs(sa_addr
->v4
.sin_port
);
1030 /* Save current address so we can work with it */
1031 memcpy(&to
, sa_addr
, af
->sockaddr_len
);
1033 err
= sctp_verify_addr(sk
, &to
, af
->sockaddr_len
);
1037 /* Make sure the destination port is correctly set
1040 if (asoc
&& asoc
->peer
.port
&& asoc
->peer
.port
!= port
)
1044 /* Check if there already is a matching association on the
1045 * endpoint (other than the one created here).
1047 asoc2
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1048 if (asoc2
&& asoc2
!= asoc
) {
1049 if (asoc2
->state
>= SCTP_STATE_ESTABLISHED
)
1056 /* If we could not find a matching association on the endpoint,
1057 * make sure that there is no peeled-off association matching
1058 * the peer address even on another socket.
1060 if (sctp_endpoint_is_peeled_off(ep
, &to
)) {
1061 err
= -EADDRNOTAVAIL
;
1066 /* If a bind() or sctp_bindx() is not called prior to
1067 * an sctp_connectx() call, the system picks an
1068 * ephemeral port and will choose an address set
1069 * equivalent to binding with a wildcard address.
1071 if (!ep
->base
.bind_addr
.port
) {
1072 if (sctp_autobind(sk
)) {
1078 * If an unprivileged user inherits a 1-many
1079 * style socket with open associations on a
1080 * privileged port, it MAY be permitted to
1081 * accept new associations, but it SHOULD NOT
1082 * be permitted to open new associations.
1084 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1085 !capable(CAP_NET_BIND_SERVICE
)) {
1091 scope
= sctp_scope(&to
);
1092 asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1098 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
,
1106 /* Prime the peer's transport structures. */
1107 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
,
1115 addr_buf
+= af
->sockaddr_len
;
1116 walk_size
+= af
->sockaddr_len
;
1119 /* In case the user of sctp_connectx() wants an association
1120 * id back, assign one now.
1123 err
= sctp_assoc_set_id(asoc
, GFP_KERNEL
);
1128 err
= sctp_primitive_ASSOCIATE(asoc
, NULL
);
1133 /* Initialize sk's dport and daddr for getpeername() */
1134 inet_sk(sk
)->inet_dport
= htons(asoc
->peer
.port
);
1135 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
1136 af
->to_sk_daddr(sa_addr
, sk
);
1139 /* in-kernel sockets don't generally have a file allocated to them
1140 * if all they do is call sock_create_kern().
1142 if (sk
->sk_socket
->file
)
1143 f_flags
= sk
->sk_socket
->file
->f_flags
;
1145 timeo
= sock_sndtimeo(sk
, f_flags
& O_NONBLOCK
);
1147 err
= sctp_wait_for_connect(asoc
, &timeo
);
1148 if ((err
== 0 || err
== -EINPROGRESS
) && assoc_id
)
1149 *assoc_id
= asoc
->assoc_id
;
1151 /* Don't free association on exit. */
1156 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1157 " kaddrs: %p err: %d\n",
1160 sctp_association_free(asoc
);
1164 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1167 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1168 * sctp_assoc_t *asoc);
1170 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1171 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1172 * or IPv6 addresses.
1174 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1175 * Section 3.1.2 for this usage.
1177 * addrs is a pointer to an array of one or more socket addresses. Each
1178 * address is contained in its appropriate structure (i.e. struct
1179 * sockaddr_in or struct sockaddr_in6) the family of the address type
1180 * must be used to distengish the address length (note that this
1181 * representation is termed a "packed array" of addresses). The caller
1182 * specifies the number of addresses in the array with addrcnt.
1184 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1185 * the association id of the new association. On failure, sctp_connectx()
1186 * returns -1, and sets errno to the appropriate error code. The assoc_id
1187 * is not touched by the kernel.
1189 * For SCTP, the port given in each socket address must be the same, or
1190 * sctp_connectx() will fail, setting errno to EINVAL.
1192 * An application can use sctp_connectx to initiate an association with
1193 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1194 * allows a caller to specify multiple addresses at which a peer can be
1195 * reached. The way the SCTP stack uses the list of addresses to set up
1196 * the association is implementation dependant. This function only
1197 * specifies that the stack will try to make use of all the addresses in
1198 * the list when needed.
1200 * Note that the list of addresses passed in is only used for setting up
1201 * the association. It does not necessarily equal the set of addresses
1202 * the peer uses for the resulting association. If the caller wants to
1203 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1204 * retrieve them after the association has been set up.
1206 * Basically do nothing but copying the addresses from user to kernel
1207 * land and invoking either sctp_connectx(). This is used for tunneling
1208 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1210 * We don't use copy_from_user() for optimization: we first do the
1211 * sanity checks (buffer size -fast- and access check-healthy
1212 * pointer); if all of those succeed, then we can alloc the memory
1213 * (expensive operation) needed to copy the data to kernel. Then we do
1214 * the copying without checking the user space area
1215 * (__copy_from_user()).
1217 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1220 * sk The sk of the socket
1221 * addrs The pointer to the addresses in user land
1222 * addrssize Size of the addrs buffer
1224 * Returns >=0 if ok, <0 errno code on error.
1226 SCTP_STATIC
int __sctp_setsockopt_connectx(struct sock
* sk
,
1227 struct sockaddr __user
*addrs
,
1229 sctp_assoc_t
*assoc_id
)
1232 struct sockaddr
*kaddrs
;
1234 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1235 __func__
, sk
, addrs
, addrs_size
);
1237 if (unlikely(addrs_size
<= 0))
1240 /* Check the user passed a healthy pointer. */
1241 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
1244 /* Alloc space for the address array in kernel memory. */
1245 kaddrs
= kmalloc(addrs_size
, GFP_KERNEL
);
1246 if (unlikely(!kaddrs
))
1249 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
1252 err
= __sctp_connect(sk
, kaddrs
, addrs_size
, assoc_id
);
1261 * This is an older interface. It's kept for backward compatibility
1262 * to the option that doesn't provide association id.
1264 SCTP_STATIC
int sctp_setsockopt_connectx_old(struct sock
* sk
,
1265 struct sockaddr __user
*addrs
,
1268 return __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, NULL
);
1272 * New interface for the API. The since the API is done with a socket
1273 * option, to make it simple we feed back the association id is as a return
1274 * indication to the call. Error is always negative and association id is
1277 SCTP_STATIC
int sctp_setsockopt_connectx(struct sock
* sk
,
1278 struct sockaddr __user
*addrs
,
1281 sctp_assoc_t assoc_id
= 0;
1284 err
= __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, &assoc_id
);
1293 * New (hopefully final) interface for the API.
1294 * We use the sctp_getaddrs_old structure so that use-space library
1295 * can avoid any unnecessary allocations. The only defferent part
1296 * is that we store the actual length of the address buffer into the
1297 * addrs_num structure member. That way we can re-use the existing
1300 SCTP_STATIC
int sctp_getsockopt_connectx3(struct sock
* sk
, int len
,
1301 char __user
*optval
,
1304 struct sctp_getaddrs_old param
;
1305 sctp_assoc_t assoc_id
= 0;
1308 if (len
< sizeof(param
))
1311 if (copy_from_user(¶m
, optval
, sizeof(param
)))
1314 err
= __sctp_setsockopt_connectx(sk
,
1315 (struct sockaddr __user
*)param
.addrs
,
1316 param
.addr_num
, &assoc_id
);
1318 if (err
== 0 || err
== -EINPROGRESS
) {
1319 if (copy_to_user(optval
, &assoc_id
, sizeof(assoc_id
)))
1321 if (put_user(sizeof(assoc_id
), optlen
))
1328 /* API 3.1.4 close() - UDP Style Syntax
1329 * Applications use close() to perform graceful shutdown (as described in
1330 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1331 * by a UDP-style socket.
1335 * ret = close(int sd);
1337 * sd - the socket descriptor of the associations to be closed.
1339 * To gracefully shutdown a specific association represented by the
1340 * UDP-style socket, an application should use the sendmsg() call,
1341 * passing no user data, but including the appropriate flag in the
1342 * ancillary data (see Section xxxx).
1344 * If sd in the close() call is a branched-off socket representing only
1345 * one association, the shutdown is performed on that association only.
1347 * 4.1.6 close() - TCP Style Syntax
1349 * Applications use close() to gracefully close down an association.
1353 * int close(int sd);
1355 * sd - the socket descriptor of the association to be closed.
1357 * After an application calls close() on a socket descriptor, no further
1358 * socket operations will succeed on that descriptor.
1360 * API 7.1.4 SO_LINGER
1362 * An application using the TCP-style socket can use this option to
1363 * perform the SCTP ABORT primitive. The linger option structure is:
1366 * int l_onoff; // option on/off
1367 * int l_linger; // linger time
1370 * To enable the option, set l_onoff to 1. If the l_linger value is set
1371 * to 0, calling close() is the same as the ABORT primitive. If the
1372 * value is set to a negative value, the setsockopt() call will return
1373 * an error. If the value is set to a positive value linger_time, the
1374 * close() can be blocked for at most linger_time ms. If the graceful
1375 * shutdown phase does not finish during this period, close() will
1376 * return but the graceful shutdown phase continues in the system.
1378 SCTP_STATIC
void sctp_close(struct sock
*sk
, long timeout
)
1380 struct sctp_endpoint
*ep
;
1381 struct sctp_association
*asoc
;
1382 struct list_head
*pos
, *temp
;
1384 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk
, timeout
);
1387 sk
->sk_shutdown
= SHUTDOWN_MASK
;
1388 sk
->sk_state
= SCTP_SS_CLOSING
;
1390 ep
= sctp_sk(sk
)->ep
;
1392 /* Walk all associations on an endpoint. */
1393 list_for_each_safe(pos
, temp
, &ep
->asocs
) {
1394 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
1396 if (sctp_style(sk
, TCP
)) {
1397 /* A closed association can still be in the list if
1398 * it belongs to a TCP-style listening socket that is
1399 * not yet accepted. If so, free it. If not, send an
1400 * ABORT or SHUTDOWN based on the linger options.
1402 if (sctp_state(asoc
, CLOSED
)) {
1403 sctp_unhash_established(asoc
);
1404 sctp_association_free(asoc
);
1409 if (sock_flag(sk
, SOCK_LINGER
) && !sk
->sk_lingertime
) {
1410 struct sctp_chunk
*chunk
;
1412 chunk
= sctp_make_abort_user(asoc
, NULL
, 0);
1414 sctp_primitive_ABORT(asoc
, chunk
);
1416 sctp_primitive_SHUTDOWN(asoc
, NULL
);
1419 /* Clean up any skbs sitting on the receive queue. */
1420 sctp_queue_purge_ulpevents(&sk
->sk_receive_queue
);
1421 sctp_queue_purge_ulpevents(&sctp_sk(sk
)->pd_lobby
);
1423 /* On a TCP-style socket, block for at most linger_time if set. */
1424 if (sctp_style(sk
, TCP
) && timeout
)
1425 sctp_wait_for_close(sk
, timeout
);
1427 /* This will run the backlog queue. */
1428 sctp_release_sock(sk
);
1430 /* Supposedly, no process has access to the socket, but
1431 * the net layers still may.
1433 sctp_local_bh_disable();
1434 sctp_bh_lock_sock(sk
);
1436 /* Hold the sock, since sk_common_release() will put sock_put()
1437 * and we have just a little more cleanup.
1440 sk_common_release(sk
);
1442 sctp_bh_unlock_sock(sk
);
1443 sctp_local_bh_enable();
1447 SCTP_DBG_OBJCNT_DEC(sock
);
1450 /* Handle EPIPE error. */
1451 static int sctp_error(struct sock
*sk
, int flags
, int err
)
1454 err
= sock_error(sk
) ? : -EPIPE
;
1455 if (err
== -EPIPE
&& !(flags
& MSG_NOSIGNAL
))
1456 send_sig(SIGPIPE
, current
, 0);
1460 /* API 3.1.3 sendmsg() - UDP Style Syntax
1462 * An application uses sendmsg() and recvmsg() calls to transmit data to
1463 * and receive data from its peer.
1465 * ssize_t sendmsg(int socket, const struct msghdr *message,
1468 * socket - the socket descriptor of the endpoint.
1469 * message - pointer to the msghdr structure which contains a single
1470 * user message and possibly some ancillary data.
1472 * See Section 5 for complete description of the data
1475 * flags - flags sent or received with the user message, see Section
1476 * 5 for complete description of the flags.
1478 * Note: This function could use a rewrite especially when explicit
1479 * connect support comes in.
1481 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1483 SCTP_STATIC
int sctp_msghdr_parse(const struct msghdr
*, sctp_cmsgs_t
*);
1485 SCTP_STATIC
int sctp_sendmsg(struct kiocb
*iocb
, struct sock
*sk
,
1486 struct msghdr
*msg
, size_t msg_len
)
1488 struct sctp_sock
*sp
;
1489 struct sctp_endpoint
*ep
;
1490 struct sctp_association
*new_asoc
=NULL
, *asoc
=NULL
;
1491 struct sctp_transport
*transport
, *chunk_tp
;
1492 struct sctp_chunk
*chunk
;
1494 struct sockaddr
*msg_name
= NULL
;
1495 struct sctp_sndrcvinfo default_sinfo
= { 0 };
1496 struct sctp_sndrcvinfo
*sinfo
;
1497 struct sctp_initmsg
*sinit
;
1498 sctp_assoc_t associd
= 0;
1499 sctp_cmsgs_t cmsgs
= { NULL
};
1503 __u16 sinfo_flags
= 0;
1504 struct sctp_datamsg
*datamsg
;
1505 int msg_flags
= msg
->msg_flags
;
1507 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1514 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep
);
1516 /* We cannot send a message over a TCP-style listening socket. */
1517 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
)) {
1522 /* Parse out the SCTP CMSGs. */
1523 err
= sctp_msghdr_parse(msg
, &cmsgs
);
1526 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err
);
1530 /* Fetch the destination address for this packet. This
1531 * address only selects the association--it is not necessarily
1532 * the address we will send to.
1533 * For a peeled-off socket, msg_name is ignored.
1535 if (!sctp_style(sk
, UDP_HIGH_BANDWIDTH
) && msg
->msg_name
) {
1536 int msg_namelen
= msg
->msg_namelen
;
1538 err
= sctp_verify_addr(sk
, (union sctp_addr
*)msg
->msg_name
,
1543 if (msg_namelen
> sizeof(to
))
1544 msg_namelen
= sizeof(to
);
1545 memcpy(&to
, msg
->msg_name
, msg_namelen
);
1546 msg_name
= msg
->msg_name
;
1552 /* Did the user specify SNDRCVINFO? */
1554 sinfo_flags
= sinfo
->sinfo_flags
;
1555 associd
= sinfo
->sinfo_assoc_id
;
1558 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1559 msg_len
, sinfo_flags
);
1561 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1562 if (sctp_style(sk
, TCP
) && (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
))) {
1567 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1568 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1569 * If SCTP_ABORT is set, the message length could be non zero with
1570 * the msg_iov set to the user abort reason.
1572 if (((sinfo_flags
& SCTP_EOF
) && (msg_len
> 0)) ||
1573 (!(sinfo_flags
& (SCTP_EOF
|SCTP_ABORT
)) && (msg_len
== 0))) {
1578 /* If SCTP_ADDR_OVER is set, there must be an address
1579 * specified in msg_name.
1581 if ((sinfo_flags
& SCTP_ADDR_OVER
) && (!msg
->msg_name
)) {
1588 SCTP_DEBUG_PRINTK("About to look up association.\n");
1592 /* If a msg_name has been specified, assume this is to be used. */
1594 /* Look for a matching association on the endpoint. */
1595 asoc
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1597 /* If we could not find a matching association on the
1598 * endpoint, make sure that it is not a TCP-style
1599 * socket that already has an association or there is
1600 * no peeled-off association on another socket.
1602 if ((sctp_style(sk
, TCP
) &&
1603 sctp_sstate(sk
, ESTABLISHED
)) ||
1604 sctp_endpoint_is_peeled_off(ep
, &to
)) {
1605 err
= -EADDRNOTAVAIL
;
1610 asoc
= sctp_id2assoc(sk
, associd
);
1618 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc
);
1620 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1621 * socket that has an association in CLOSED state. This can
1622 * happen when an accepted socket has an association that is
1625 if (sctp_state(asoc
, CLOSED
) && sctp_style(sk
, TCP
)) {
1630 if (sinfo_flags
& SCTP_EOF
) {
1631 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1633 sctp_primitive_SHUTDOWN(asoc
, NULL
);
1637 if (sinfo_flags
& SCTP_ABORT
) {
1639 chunk
= sctp_make_abort_user(asoc
, msg
, msg_len
);
1645 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc
);
1646 sctp_primitive_ABORT(asoc
, chunk
);
1652 /* Do we need to create the association? */
1654 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1656 if (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
)) {
1661 /* Check for invalid stream against the stream counts,
1662 * either the default or the user specified stream counts.
1665 if (!sinit
|| (sinit
&& !sinit
->sinit_num_ostreams
)) {
1666 /* Check against the defaults. */
1667 if (sinfo
->sinfo_stream
>=
1668 sp
->initmsg
.sinit_num_ostreams
) {
1673 /* Check against the requested. */
1674 if (sinfo
->sinfo_stream
>=
1675 sinit
->sinit_num_ostreams
) {
1683 * API 3.1.2 bind() - UDP Style Syntax
1684 * If a bind() or sctp_bindx() is not called prior to a
1685 * sendmsg() call that initiates a new association, the
1686 * system picks an ephemeral port and will choose an address
1687 * set equivalent to binding with a wildcard address.
1689 if (!ep
->base
.bind_addr
.port
) {
1690 if (sctp_autobind(sk
)) {
1696 * If an unprivileged user inherits a one-to-many
1697 * style socket with open associations on a privileged
1698 * port, it MAY be permitted to accept new associations,
1699 * but it SHOULD NOT be permitted to open new
1702 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1703 !capable(CAP_NET_BIND_SERVICE
)) {
1709 scope
= sctp_scope(&to
);
1710 new_asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1716 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
, GFP_KERNEL
);
1722 /* If the SCTP_INIT ancillary data is specified, set all
1723 * the association init values accordingly.
1726 if (sinit
->sinit_num_ostreams
) {
1727 asoc
->c
.sinit_num_ostreams
=
1728 sinit
->sinit_num_ostreams
;
1730 if (sinit
->sinit_max_instreams
) {
1731 asoc
->c
.sinit_max_instreams
=
1732 sinit
->sinit_max_instreams
;
1734 if (sinit
->sinit_max_attempts
) {
1735 asoc
->max_init_attempts
1736 = sinit
->sinit_max_attempts
;
1738 if (sinit
->sinit_max_init_timeo
) {
1739 asoc
->max_init_timeo
=
1740 msecs_to_jiffies(sinit
->sinit_max_init_timeo
);
1744 /* Prime the peer's transport structures. */
1745 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
, SCTP_UNKNOWN
);
1752 /* ASSERT: we have a valid association at this point. */
1753 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1756 /* If the user didn't specify SNDRCVINFO, make up one with
1759 default_sinfo
.sinfo_stream
= asoc
->default_stream
;
1760 default_sinfo
.sinfo_flags
= asoc
->default_flags
;
1761 default_sinfo
.sinfo_ppid
= asoc
->default_ppid
;
1762 default_sinfo
.sinfo_context
= asoc
->default_context
;
1763 default_sinfo
.sinfo_timetolive
= asoc
->default_timetolive
;
1764 default_sinfo
.sinfo_assoc_id
= sctp_assoc2id(asoc
);
1765 sinfo
= &default_sinfo
;
1768 /* API 7.1.7, the sndbuf size per association bounds the
1769 * maximum size of data that can be sent in a single send call.
1771 if (msg_len
> sk
->sk_sndbuf
) {
1776 if (asoc
->pmtu_pending
)
1777 sctp_assoc_pending_pmtu(asoc
);
1779 /* If fragmentation is disabled and the message length exceeds the
1780 * association fragmentation point, return EMSGSIZE. The I-D
1781 * does not specify what this error is, but this looks like
1784 if (sctp_sk(sk
)->disable_fragments
&& (msg_len
> asoc
->frag_point
)) {
1790 /* Check for invalid stream. */
1791 if (sinfo
->sinfo_stream
>= asoc
->c
.sinit_num_ostreams
) {
1797 timeo
= sock_sndtimeo(sk
, msg
->msg_flags
& MSG_DONTWAIT
);
1798 if (!sctp_wspace(asoc
)) {
1799 err
= sctp_wait_for_sndbuf(asoc
, &timeo
, msg_len
);
1804 /* If an address is passed with the sendto/sendmsg call, it is used
1805 * to override the primary destination address in the TCP model, or
1806 * when SCTP_ADDR_OVER flag is set in the UDP model.
1808 if ((sctp_style(sk
, TCP
) && msg_name
) ||
1809 (sinfo_flags
& SCTP_ADDR_OVER
)) {
1810 chunk_tp
= sctp_assoc_lookup_paddr(asoc
, &to
);
1818 /* Auto-connect, if we aren't connected already. */
1819 if (sctp_state(asoc
, CLOSED
)) {
1820 err
= sctp_primitive_ASSOCIATE(asoc
, NULL
);
1823 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1826 /* Break the message into multiple chunks of maximum size. */
1827 datamsg
= sctp_datamsg_from_user(asoc
, sinfo
, msg
, msg_len
);
1833 /* Now send the (possibly) fragmented message. */
1834 list_for_each_entry(chunk
, &datamsg
->chunks
, frag_list
) {
1835 sctp_chunk_hold(chunk
);
1837 /* Do accounting for the write space. */
1838 sctp_set_owner_w(chunk
);
1840 chunk
->transport
= chunk_tp
;
1843 /* Send it to the lower layers. Note: all chunks
1844 * must either fail or succeed. The lower layer
1845 * works that way today. Keep it that way or this
1848 err
= sctp_primitive_SEND(asoc
, datamsg
);
1849 /* Did the lower layer accept the chunk? */
1851 sctp_datamsg_free(datamsg
);
1853 sctp_datamsg_put(datamsg
);
1855 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1862 /* If we are already past ASSOCIATE, the lower
1863 * layers are responsible for association cleanup.
1869 sctp_association_free(asoc
);
1871 sctp_release_sock(sk
);
1874 return sctp_error(sk
, msg_flags
, err
);
1881 err
= sock_error(sk
);
1891 /* This is an extended version of skb_pull() that removes the data from the
1892 * start of a skb even when data is spread across the list of skb's in the
1893 * frag_list. len specifies the total amount of data that needs to be removed.
1894 * when 'len' bytes could be removed from the skb, it returns 0.
1895 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1896 * could not be removed.
1898 static int sctp_skb_pull(struct sk_buff
*skb
, int len
)
1900 struct sk_buff
*list
;
1901 int skb_len
= skb_headlen(skb
);
1904 if (len
<= skb_len
) {
1905 __skb_pull(skb
, len
);
1909 __skb_pull(skb
, skb_len
);
1911 skb_walk_frags(skb
, list
) {
1912 rlen
= sctp_skb_pull(list
, len
);
1913 skb
->len
-= (len
-rlen
);
1914 skb
->data_len
-= (len
-rlen
);
1925 /* API 3.1.3 recvmsg() - UDP Style Syntax
1927 * ssize_t recvmsg(int socket, struct msghdr *message,
1930 * socket - the socket descriptor of the endpoint.
1931 * message - pointer to the msghdr structure which contains a single
1932 * user message and possibly some ancillary data.
1934 * See Section 5 for complete description of the data
1937 * flags - flags sent or received with the user message, see Section
1938 * 5 for complete description of the flags.
1940 static struct sk_buff
*sctp_skb_recv_datagram(struct sock
*, int, int, int *);
1942 SCTP_STATIC
int sctp_recvmsg(struct kiocb
*iocb
, struct sock
*sk
,
1943 struct msghdr
*msg
, size_t len
, int noblock
,
1944 int flags
, int *addr_len
)
1946 struct sctp_ulpevent
*event
= NULL
;
1947 struct sctp_sock
*sp
= sctp_sk(sk
);
1948 struct sk_buff
*skb
;
1953 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1954 "0x%x, %s: %p)\n", "sk", sk
, "msghdr", msg
,
1955 "len", len
, "knoblauch", noblock
,
1956 "flags", flags
, "addr_len", addr_len
);
1960 if (sctp_style(sk
, TCP
) && !sctp_sstate(sk
, ESTABLISHED
)) {
1965 skb
= sctp_skb_recv_datagram(sk
, flags
, noblock
, &err
);
1969 /* Get the total length of the skb including any skb's in the
1978 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
1980 event
= sctp_skb2event(skb
);
1985 sock_recv_ts_and_drops(msg
, sk
, skb
);
1986 if (sctp_ulpevent_is_notification(event
)) {
1987 msg
->msg_flags
|= MSG_NOTIFICATION
;
1988 sp
->pf
->event_msgname(event
, msg
->msg_name
, addr_len
);
1990 sp
->pf
->skb_msgname(skb
, msg
->msg_name
, addr_len
);
1993 /* Check if we allow SCTP_SNDRCVINFO. */
1994 if (sp
->subscribe
.sctp_data_io_event
)
1995 sctp_ulpevent_read_sndrcvinfo(event
, msg
);
1997 /* FIXME: we should be calling IP/IPv6 layers. */
1998 if (sk
->sk_protinfo
.af_inet
.cmsg_flags
)
1999 ip_cmsg_recv(msg
, skb
);
2004 /* If skb's length exceeds the user's buffer, update the skb and
2005 * push it back to the receive_queue so that the next call to
2006 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2008 if (skb_len
> copied
) {
2009 msg
->msg_flags
&= ~MSG_EOR
;
2010 if (flags
& MSG_PEEK
)
2012 sctp_skb_pull(skb
, copied
);
2013 skb_queue_head(&sk
->sk_receive_queue
, skb
);
2015 /* When only partial message is copied to the user, increase
2016 * rwnd by that amount. If all the data in the skb is read,
2017 * rwnd is updated when the event is freed.
2019 if (!sctp_ulpevent_is_notification(event
))
2020 sctp_assoc_rwnd_increase(event
->asoc
, copied
);
2022 } else if ((event
->msg_flags
& MSG_NOTIFICATION
) ||
2023 (event
->msg_flags
& MSG_EOR
))
2024 msg
->msg_flags
|= MSG_EOR
;
2026 msg
->msg_flags
&= ~MSG_EOR
;
2029 if (flags
& MSG_PEEK
) {
2030 /* Release the skb reference acquired after peeking the skb in
2031 * sctp_skb_recv_datagram().
2035 /* Free the event which includes releasing the reference to
2036 * the owner of the skb, freeing the skb and updating the
2039 sctp_ulpevent_free(event
);
2042 sctp_release_sock(sk
);
2046 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2048 * This option is a on/off flag. If enabled no SCTP message
2049 * fragmentation will be performed. Instead if a message being sent
2050 * exceeds the current PMTU size, the message will NOT be sent and
2051 * instead a error will be indicated to the user.
2053 static int sctp_setsockopt_disable_fragments(struct sock
*sk
,
2054 char __user
*optval
,
2055 unsigned int optlen
)
2059 if (optlen
< sizeof(int))
2062 if (get_user(val
, (int __user
*)optval
))
2065 sctp_sk(sk
)->disable_fragments
= (val
== 0) ? 0 : 1;
2070 static int sctp_setsockopt_events(struct sock
*sk
, char __user
*optval
,
2071 unsigned int optlen
)
2073 if (optlen
> sizeof(struct sctp_event_subscribe
))
2075 if (copy_from_user(&sctp_sk(sk
)->subscribe
, optval
, optlen
))
2080 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2082 * This socket option is applicable to the UDP-style socket only. When
2083 * set it will cause associations that are idle for more than the
2084 * specified number of seconds to automatically close. An association
2085 * being idle is defined an association that has NOT sent or received
2086 * user data. The special value of '0' indicates that no automatic
2087 * close of any associations should be performed. The option expects an
2088 * integer defining the number of seconds of idle time before an
2089 * association is closed.
2091 static int sctp_setsockopt_autoclose(struct sock
*sk
, char __user
*optval
,
2092 unsigned int optlen
)
2094 struct sctp_sock
*sp
= sctp_sk(sk
);
2096 /* Applicable to UDP-style socket only */
2097 if (sctp_style(sk
, TCP
))
2099 if (optlen
!= sizeof(int))
2101 if (copy_from_user(&sp
->autoclose
, optval
, optlen
))
2103 /* make sure it won't exceed MAX_SCHEDULE_TIMEOUT */
2104 sp
->autoclose
= min_t(long, sp
->autoclose
, MAX_SCHEDULE_TIMEOUT
/ HZ
);
2109 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2111 * Applications can enable or disable heartbeats for any peer address of
2112 * an association, modify an address's heartbeat interval, force a
2113 * heartbeat to be sent immediately, and adjust the address's maximum
2114 * number of retransmissions sent before an address is considered
2115 * unreachable. The following structure is used to access and modify an
2116 * address's parameters:
2118 * struct sctp_paddrparams {
2119 * sctp_assoc_t spp_assoc_id;
2120 * struct sockaddr_storage spp_address;
2121 * uint32_t spp_hbinterval;
2122 * uint16_t spp_pathmaxrxt;
2123 * uint32_t spp_pathmtu;
2124 * uint32_t spp_sackdelay;
2125 * uint32_t spp_flags;
2128 * spp_assoc_id - (one-to-many style socket) This is filled in the
2129 * application, and identifies the association for
2131 * spp_address - This specifies which address is of interest.
2132 * spp_hbinterval - This contains the value of the heartbeat interval,
2133 * in milliseconds. If a value of zero
2134 * is present in this field then no changes are to
2135 * be made to this parameter.
2136 * spp_pathmaxrxt - This contains the maximum number of
2137 * retransmissions before this address shall be
2138 * considered unreachable. If a value of zero
2139 * is present in this field then no changes are to
2140 * be made to this parameter.
2141 * spp_pathmtu - When Path MTU discovery is disabled the value
2142 * specified here will be the "fixed" path mtu.
2143 * Note that if the spp_address field is empty
2144 * then all associations on this address will
2145 * have this fixed path mtu set upon them.
2147 * spp_sackdelay - When delayed sack is enabled, this value specifies
2148 * the number of milliseconds that sacks will be delayed
2149 * for. This value will apply to all addresses of an
2150 * association if the spp_address field is empty. Note
2151 * also, that if delayed sack is enabled and this
2152 * value is set to 0, no change is made to the last
2153 * recorded delayed sack timer value.
2155 * spp_flags - These flags are used to control various features
2156 * on an association. The flag field may contain
2157 * zero or more of the following options.
2159 * SPP_HB_ENABLE - Enable heartbeats on the
2160 * specified address. Note that if the address
2161 * field is empty all addresses for the association
2162 * have heartbeats enabled upon them.
2164 * SPP_HB_DISABLE - Disable heartbeats on the
2165 * speicifed address. Note that if the address
2166 * field is empty all addresses for the association
2167 * will have their heartbeats disabled. Note also
2168 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2169 * mutually exclusive, only one of these two should
2170 * be specified. Enabling both fields will have
2171 * undetermined results.
2173 * SPP_HB_DEMAND - Request a user initiated heartbeat
2174 * to be made immediately.
2176 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2177 * heartbeat delayis to be set to the value of 0
2180 * SPP_PMTUD_ENABLE - This field will enable PMTU
2181 * discovery upon the specified address. Note that
2182 * if the address feild is empty then all addresses
2183 * on the association are effected.
2185 * SPP_PMTUD_DISABLE - This field will disable 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. Not also that
2189 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2190 * exclusive. Enabling both will have undetermined
2193 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2194 * on delayed sack. The time specified in spp_sackdelay
2195 * is used to specify the sack delay for this address. Note
2196 * that if spp_address is empty then all addresses will
2197 * enable delayed sack and take on the sack delay
2198 * value specified in spp_sackdelay.
2199 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2200 * off delayed sack. If the spp_address field is blank then
2201 * delayed sack is disabled for the entire association. Note
2202 * also that this field is mutually exclusive to
2203 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2206 static int sctp_apply_peer_addr_params(struct sctp_paddrparams
*params
,
2207 struct sctp_transport
*trans
,
2208 struct sctp_association
*asoc
,
2209 struct sctp_sock
*sp
,
2212 int sackdelay_change
)
2216 if (params
->spp_flags
& SPP_HB_DEMAND
&& trans
) {
2217 error
= sctp_primitive_REQUESTHEARTBEAT (trans
->asoc
, trans
);
2222 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2223 * this field is ignored. Note also that a value of zero indicates
2224 * the current setting should be left unchanged.
2226 if (params
->spp_flags
& SPP_HB_ENABLE
) {
2228 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2229 * set. This lets us use 0 value when this flag
2232 if (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)
2233 params
->spp_hbinterval
= 0;
2235 if (params
->spp_hbinterval
||
2236 (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)) {
2239 msecs_to_jiffies(params
->spp_hbinterval
);
2242 msecs_to_jiffies(params
->spp_hbinterval
);
2244 sp
->hbinterval
= params
->spp_hbinterval
;
2251 trans
->param_flags
=
2252 (trans
->param_flags
& ~SPP_HB
) | hb_change
;
2255 (asoc
->param_flags
& ~SPP_HB
) | hb_change
;
2258 (sp
->param_flags
& ~SPP_HB
) | hb_change
;
2262 /* When Path MTU discovery is disabled the value specified here will
2263 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2264 * include the flag SPP_PMTUD_DISABLE for this field to have any
2267 if ((params
->spp_flags
& SPP_PMTUD_DISABLE
) && params
->spp_pathmtu
) {
2269 trans
->pathmtu
= params
->spp_pathmtu
;
2270 sctp_assoc_sync_pmtu(asoc
);
2272 asoc
->pathmtu
= params
->spp_pathmtu
;
2273 sctp_frag_point(asoc
, params
->spp_pathmtu
);
2275 sp
->pathmtu
= params
->spp_pathmtu
;
2281 int update
= (trans
->param_flags
& SPP_PMTUD_DISABLE
) &&
2282 (params
->spp_flags
& SPP_PMTUD_ENABLE
);
2283 trans
->param_flags
=
2284 (trans
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2286 sctp_transport_pmtu(trans
);
2287 sctp_assoc_sync_pmtu(asoc
);
2291 (asoc
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2294 (sp
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2298 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2299 * value of this field is ignored. Note also that a value of zero
2300 * indicates the current setting should be left unchanged.
2302 if ((params
->spp_flags
& SPP_SACKDELAY_ENABLE
) && params
->spp_sackdelay
) {
2305 msecs_to_jiffies(params
->spp_sackdelay
);
2308 msecs_to_jiffies(params
->spp_sackdelay
);
2310 sp
->sackdelay
= params
->spp_sackdelay
;
2314 if (sackdelay_change
) {
2316 trans
->param_flags
=
2317 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2321 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2325 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2330 /* Note that a value of zero indicates the current setting should be
2333 if (params
->spp_pathmaxrxt
) {
2335 trans
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2337 asoc
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2339 sp
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2346 static int sctp_setsockopt_peer_addr_params(struct sock
*sk
,
2347 char __user
*optval
,
2348 unsigned int optlen
)
2350 struct sctp_paddrparams params
;
2351 struct sctp_transport
*trans
= NULL
;
2352 struct sctp_association
*asoc
= NULL
;
2353 struct sctp_sock
*sp
= sctp_sk(sk
);
2355 int hb_change
, pmtud_change
, sackdelay_change
;
2357 if (optlen
!= sizeof(struct sctp_paddrparams
))
2360 if (copy_from_user(¶ms
, optval
, optlen
))
2363 /* Validate flags and value parameters. */
2364 hb_change
= params
.spp_flags
& SPP_HB
;
2365 pmtud_change
= params
.spp_flags
& SPP_PMTUD
;
2366 sackdelay_change
= params
.spp_flags
& SPP_SACKDELAY
;
2368 if (hb_change
== SPP_HB
||
2369 pmtud_change
== SPP_PMTUD
||
2370 sackdelay_change
== SPP_SACKDELAY
||
2371 params
.spp_sackdelay
> 500 ||
2372 (params
.spp_pathmtu
&&
2373 params
.spp_pathmtu
< SCTP_DEFAULT_MINSEGMENT
))
2376 /* If an address other than INADDR_ANY is specified, and
2377 * no transport is found, then the request is invalid.
2379 if (!sctp_is_any(sk
, ( union sctp_addr
*)¶ms
.spp_address
)) {
2380 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
2381 params
.spp_assoc_id
);
2386 /* Get association, if assoc_id != 0 and the socket is a one
2387 * to many style socket, and an association was not found, then
2388 * the id was invalid.
2390 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
2391 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
))
2394 /* Heartbeat demand can only be sent on a transport or
2395 * association, but not a socket.
2397 if (params
.spp_flags
& SPP_HB_DEMAND
&& !trans
&& !asoc
)
2400 /* Process parameters. */
2401 error
= sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2402 hb_change
, pmtud_change
,
2408 /* If changes are for association, also apply parameters to each
2411 if (!trans
&& asoc
) {
2412 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2414 sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2415 hb_change
, pmtud_change
,
2424 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2426 * This option will effect the way delayed acks are performed. This
2427 * option allows you to get or set the delayed ack time, in
2428 * milliseconds. It also allows changing the delayed ack frequency.
2429 * Changing the frequency to 1 disables the delayed sack algorithm. If
2430 * the assoc_id is 0, then this sets or gets the endpoints default
2431 * values. If the assoc_id field is non-zero, then the set or get
2432 * effects the specified association for the one to many model (the
2433 * assoc_id field is ignored by the one to one model). Note that if
2434 * sack_delay or sack_freq are 0 when setting this option, then the
2435 * current values will remain unchanged.
2437 * struct sctp_sack_info {
2438 * sctp_assoc_t sack_assoc_id;
2439 * uint32_t sack_delay;
2440 * uint32_t sack_freq;
2443 * sack_assoc_id - This parameter, indicates which association the user
2444 * is performing an action upon. Note that if this field's value is
2445 * zero then the endpoints default value is changed (effecting future
2446 * associations only).
2448 * sack_delay - This parameter contains the number of milliseconds that
2449 * the user is requesting the delayed ACK timer be set to. Note that
2450 * this value is defined in the standard to be between 200 and 500
2453 * sack_freq - This parameter contains the number of packets that must
2454 * be received before a sack is sent without waiting for the delay
2455 * timer to expire. The default value for this is 2, setting this
2456 * value to 1 will disable the delayed sack algorithm.
2459 static int sctp_setsockopt_delayed_ack(struct sock
*sk
,
2460 char __user
*optval
, unsigned int optlen
)
2462 struct sctp_sack_info params
;
2463 struct sctp_transport
*trans
= NULL
;
2464 struct sctp_association
*asoc
= NULL
;
2465 struct sctp_sock
*sp
= sctp_sk(sk
);
2467 if (optlen
== sizeof(struct sctp_sack_info
)) {
2468 if (copy_from_user(¶ms
, optval
, optlen
))
2471 if (params
.sack_delay
== 0 && params
.sack_freq
== 0)
2473 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
2474 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2475 pr_warn("Use struct sctp_sack_info instead\n");
2476 if (copy_from_user(¶ms
, optval
, optlen
))
2479 if (params
.sack_delay
== 0)
2480 params
.sack_freq
= 1;
2482 params
.sack_freq
= 0;
2486 /* Validate value parameter. */
2487 if (params
.sack_delay
> 500)
2490 /* Get association, if sack_assoc_id != 0 and the socket is a one
2491 * to many style socket, and an association was not found, then
2492 * the id was invalid.
2494 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
2495 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
2498 if (params
.sack_delay
) {
2501 msecs_to_jiffies(params
.sack_delay
);
2503 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2504 SPP_SACKDELAY_ENABLE
;
2506 sp
->sackdelay
= params
.sack_delay
;
2508 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2509 SPP_SACKDELAY_ENABLE
;
2513 if (params
.sack_freq
== 1) {
2516 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2517 SPP_SACKDELAY_DISABLE
;
2520 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2521 SPP_SACKDELAY_DISABLE
;
2523 } else if (params
.sack_freq
> 1) {
2525 asoc
->sackfreq
= params
.sack_freq
;
2527 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2528 SPP_SACKDELAY_ENABLE
;
2530 sp
->sackfreq
= params
.sack_freq
;
2532 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2533 SPP_SACKDELAY_ENABLE
;
2537 /* If change is for association, also apply to each transport. */
2539 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2541 if (params
.sack_delay
) {
2543 msecs_to_jiffies(params
.sack_delay
);
2544 trans
->param_flags
=
2545 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2546 SPP_SACKDELAY_ENABLE
;
2548 if (params
.sack_freq
== 1) {
2549 trans
->param_flags
=
2550 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2551 SPP_SACKDELAY_DISABLE
;
2552 } else if (params
.sack_freq
> 1) {
2553 trans
->sackfreq
= params
.sack_freq
;
2554 trans
->param_flags
=
2555 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2556 SPP_SACKDELAY_ENABLE
;
2564 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2566 * Applications can specify protocol parameters for the default association
2567 * initialization. The option name argument to setsockopt() and getsockopt()
2570 * Setting initialization parameters is effective only on an unconnected
2571 * socket (for UDP-style sockets only future associations are effected
2572 * by the change). With TCP-style sockets, this option is inherited by
2573 * sockets derived from a listener socket.
2575 static int sctp_setsockopt_initmsg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2577 struct sctp_initmsg sinit
;
2578 struct sctp_sock
*sp
= sctp_sk(sk
);
2580 if (optlen
!= sizeof(struct sctp_initmsg
))
2582 if (copy_from_user(&sinit
, optval
, optlen
))
2585 if (sinit
.sinit_num_ostreams
)
2586 sp
->initmsg
.sinit_num_ostreams
= sinit
.sinit_num_ostreams
;
2587 if (sinit
.sinit_max_instreams
)
2588 sp
->initmsg
.sinit_max_instreams
= sinit
.sinit_max_instreams
;
2589 if (sinit
.sinit_max_attempts
)
2590 sp
->initmsg
.sinit_max_attempts
= sinit
.sinit_max_attempts
;
2591 if (sinit
.sinit_max_init_timeo
)
2592 sp
->initmsg
.sinit_max_init_timeo
= sinit
.sinit_max_init_timeo
;
2598 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2600 * Applications that wish to use the sendto() system call may wish to
2601 * specify a default set of parameters that would normally be supplied
2602 * through the inclusion of ancillary data. This socket option allows
2603 * such an application to set the default sctp_sndrcvinfo structure.
2604 * The application that wishes to use this socket option simply passes
2605 * in to this call the sctp_sndrcvinfo structure defined in Section
2606 * 5.2.2) The input parameters accepted by this call include
2607 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2608 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2609 * to this call if the caller is using the UDP model.
2611 static int sctp_setsockopt_default_send_param(struct sock
*sk
,
2612 char __user
*optval
,
2613 unsigned int optlen
)
2615 struct sctp_sndrcvinfo info
;
2616 struct sctp_association
*asoc
;
2617 struct sctp_sock
*sp
= sctp_sk(sk
);
2619 if (optlen
!= sizeof(struct sctp_sndrcvinfo
))
2621 if (copy_from_user(&info
, optval
, optlen
))
2624 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
2625 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
2629 asoc
->default_stream
= info
.sinfo_stream
;
2630 asoc
->default_flags
= info
.sinfo_flags
;
2631 asoc
->default_ppid
= info
.sinfo_ppid
;
2632 asoc
->default_context
= info
.sinfo_context
;
2633 asoc
->default_timetolive
= info
.sinfo_timetolive
;
2635 sp
->default_stream
= info
.sinfo_stream
;
2636 sp
->default_flags
= info
.sinfo_flags
;
2637 sp
->default_ppid
= info
.sinfo_ppid
;
2638 sp
->default_context
= info
.sinfo_context
;
2639 sp
->default_timetolive
= info
.sinfo_timetolive
;
2645 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2647 * Requests that the local SCTP stack use the enclosed peer address as
2648 * the association primary. The enclosed address must be one of the
2649 * association peer's addresses.
2651 static int sctp_setsockopt_primary_addr(struct sock
*sk
, char __user
*optval
,
2652 unsigned int optlen
)
2654 struct sctp_prim prim
;
2655 struct sctp_transport
*trans
;
2657 if (optlen
!= sizeof(struct sctp_prim
))
2660 if (copy_from_user(&prim
, optval
, sizeof(struct sctp_prim
)))
2663 trans
= sctp_addr_id2transport(sk
, &prim
.ssp_addr
, prim
.ssp_assoc_id
);
2667 sctp_assoc_set_primary(trans
->asoc
, trans
);
2673 * 7.1.5 SCTP_NODELAY
2675 * Turn on/off any Nagle-like algorithm. This means that packets are
2676 * generally sent as soon as possible and no unnecessary delays are
2677 * introduced, at the cost of more packets in the network. Expects an
2678 * integer boolean flag.
2680 static int sctp_setsockopt_nodelay(struct sock
*sk
, char __user
*optval
,
2681 unsigned int optlen
)
2685 if (optlen
< sizeof(int))
2687 if (get_user(val
, (int __user
*)optval
))
2690 sctp_sk(sk
)->nodelay
= (val
== 0) ? 0 : 1;
2696 * 7.1.1 SCTP_RTOINFO
2698 * The protocol parameters used to initialize and bound retransmission
2699 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2700 * and modify these parameters.
2701 * All parameters are time values, in milliseconds. A value of 0, when
2702 * modifying the parameters, indicates that the current value should not
2706 static int sctp_setsockopt_rtoinfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2708 struct sctp_rtoinfo rtoinfo
;
2709 struct sctp_association
*asoc
;
2711 if (optlen
!= sizeof (struct sctp_rtoinfo
))
2714 if (copy_from_user(&rtoinfo
, optval
, optlen
))
2717 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
2719 /* Set the values to the specific association */
2720 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
2724 if (rtoinfo
.srto_initial
!= 0)
2726 msecs_to_jiffies(rtoinfo
.srto_initial
);
2727 if (rtoinfo
.srto_max
!= 0)
2728 asoc
->rto_max
= msecs_to_jiffies(rtoinfo
.srto_max
);
2729 if (rtoinfo
.srto_min
!= 0)
2730 asoc
->rto_min
= msecs_to_jiffies(rtoinfo
.srto_min
);
2732 /* If there is no association or the association-id = 0
2733 * set the values to the endpoint.
2735 struct sctp_sock
*sp
= sctp_sk(sk
);
2737 if (rtoinfo
.srto_initial
!= 0)
2738 sp
->rtoinfo
.srto_initial
= rtoinfo
.srto_initial
;
2739 if (rtoinfo
.srto_max
!= 0)
2740 sp
->rtoinfo
.srto_max
= rtoinfo
.srto_max
;
2741 if (rtoinfo
.srto_min
!= 0)
2742 sp
->rtoinfo
.srto_min
= rtoinfo
.srto_min
;
2750 * 7.1.2 SCTP_ASSOCINFO
2752 * This option is used to tune the maximum retransmission attempts
2753 * of the association.
2754 * Returns an error if the new association retransmission value is
2755 * greater than the sum of the retransmission value of the peer.
2756 * See [SCTP] for more information.
2759 static int sctp_setsockopt_associnfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2762 struct sctp_assocparams assocparams
;
2763 struct sctp_association
*asoc
;
2765 if (optlen
!= sizeof(struct sctp_assocparams
))
2767 if (copy_from_user(&assocparams
, optval
, optlen
))
2770 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
2772 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
2775 /* Set the values to the specific association */
2777 if (assocparams
.sasoc_asocmaxrxt
!= 0) {
2780 struct sctp_transport
*peer_addr
;
2782 list_for_each_entry(peer_addr
, &asoc
->peer
.transport_addr_list
,
2784 path_sum
+= peer_addr
->pathmaxrxt
;
2788 /* Only validate asocmaxrxt if we have more than
2789 * one path/transport. We do this because path
2790 * retransmissions are only counted when we have more
2794 assocparams
.sasoc_asocmaxrxt
> path_sum
)
2797 asoc
->max_retrans
= assocparams
.sasoc_asocmaxrxt
;
2800 if (assocparams
.sasoc_cookie_life
!= 0) {
2801 asoc
->cookie_life
.tv_sec
=
2802 assocparams
.sasoc_cookie_life
/ 1000;
2803 asoc
->cookie_life
.tv_usec
=
2804 (assocparams
.sasoc_cookie_life
% 1000)
2808 /* Set the values to the endpoint */
2809 struct sctp_sock
*sp
= sctp_sk(sk
);
2811 if (assocparams
.sasoc_asocmaxrxt
!= 0)
2812 sp
->assocparams
.sasoc_asocmaxrxt
=
2813 assocparams
.sasoc_asocmaxrxt
;
2814 if (assocparams
.sasoc_cookie_life
!= 0)
2815 sp
->assocparams
.sasoc_cookie_life
=
2816 assocparams
.sasoc_cookie_life
;
2822 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2824 * This socket option is a boolean flag which turns on or off mapped V4
2825 * addresses. If this option is turned on and the socket is type
2826 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2827 * If this option is turned off, then no mapping will be done of V4
2828 * addresses and a user will receive both PF_INET6 and PF_INET type
2829 * addresses on the socket.
2831 static int sctp_setsockopt_mappedv4(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2834 struct sctp_sock
*sp
= sctp_sk(sk
);
2836 if (optlen
< sizeof(int))
2838 if (get_user(val
, (int __user
*)optval
))
2849 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2850 * This option will get or set the maximum size to put in any outgoing
2851 * SCTP DATA chunk. If a message is larger than this size it will be
2852 * fragmented by SCTP into the specified size. Note that the underlying
2853 * SCTP implementation may fragment into smaller sized chunks when the
2854 * PMTU of the underlying association is smaller than the value set by
2855 * the user. The default value for this option is '0' which indicates
2856 * the user is NOT limiting fragmentation and only the PMTU will effect
2857 * SCTP's choice of DATA chunk size. Note also that values set larger
2858 * than the maximum size of an IP datagram will effectively let SCTP
2859 * control fragmentation (i.e. the same as setting this option to 0).
2861 * The following structure is used to access and modify this parameter:
2863 * struct sctp_assoc_value {
2864 * sctp_assoc_t assoc_id;
2865 * uint32_t assoc_value;
2868 * assoc_id: This parameter is ignored for one-to-one style sockets.
2869 * For one-to-many style sockets this parameter indicates which
2870 * association the user is performing an action upon. Note that if
2871 * this field's value is zero then the endpoints default value is
2872 * changed (effecting future associations only).
2873 * assoc_value: This parameter specifies the maximum size in bytes.
2875 static int sctp_setsockopt_maxseg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2877 struct sctp_assoc_value params
;
2878 struct sctp_association
*asoc
;
2879 struct sctp_sock
*sp
= sctp_sk(sk
);
2882 if (optlen
== sizeof(int)) {
2883 pr_warn("Use of int in maxseg socket option deprecated\n");
2884 pr_warn("Use struct sctp_assoc_value instead\n");
2885 if (copy_from_user(&val
, optval
, optlen
))
2887 params
.assoc_id
= 0;
2888 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
2889 if (copy_from_user(¶ms
, optval
, optlen
))
2891 val
= params
.assoc_value
;
2895 if ((val
!= 0) && ((val
< 8) || (val
> SCTP_MAX_CHUNK_LEN
)))
2898 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
2899 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
2904 val
= asoc
->pathmtu
;
2905 val
-= sp
->pf
->af
->net_header_len
;
2906 val
-= sizeof(struct sctphdr
) +
2907 sizeof(struct sctp_data_chunk
);
2909 asoc
->user_frag
= val
;
2910 asoc
->frag_point
= sctp_frag_point(asoc
, asoc
->pathmtu
);
2912 sp
->user_frag
= val
;
2920 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2922 * Requests that the peer mark the enclosed address as the association
2923 * primary. The enclosed address must be one of the association's
2924 * locally bound addresses. The following structure is used to make a
2925 * set primary request:
2927 static int sctp_setsockopt_peer_primary_addr(struct sock
*sk
, char __user
*optval
,
2928 unsigned int optlen
)
2930 struct sctp_sock
*sp
;
2931 struct sctp_endpoint
*ep
;
2932 struct sctp_association
*asoc
= NULL
;
2933 struct sctp_setpeerprim prim
;
2934 struct sctp_chunk
*chunk
;
2940 if (!sctp_addip_enable
)
2943 if (optlen
!= sizeof(struct sctp_setpeerprim
))
2946 if (copy_from_user(&prim
, optval
, optlen
))
2949 asoc
= sctp_id2assoc(sk
, prim
.sspp_assoc_id
);
2953 if (!asoc
->peer
.asconf_capable
)
2956 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_SET_PRIMARY
)
2959 if (!sctp_state(asoc
, ESTABLISHED
))
2962 if (!sctp_assoc_lookup_laddr(asoc
, (union sctp_addr
*)&prim
.sspp_addr
))
2963 return -EADDRNOTAVAIL
;
2965 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2966 chunk
= sctp_make_asconf_set_prim(asoc
,
2967 (union sctp_addr
*)&prim
.sspp_addr
);
2971 err
= sctp_send_asconf(asoc
, chunk
);
2973 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2978 static int sctp_setsockopt_adaptation_layer(struct sock
*sk
, char __user
*optval
,
2979 unsigned int optlen
)
2981 struct sctp_setadaptation adaptation
;
2983 if (optlen
!= sizeof(struct sctp_setadaptation
))
2985 if (copy_from_user(&adaptation
, optval
, optlen
))
2988 sctp_sk(sk
)->adaptation_ind
= adaptation
.ssb_adaptation_ind
;
2994 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
2996 * The context field in the sctp_sndrcvinfo structure is normally only
2997 * used when a failed message is retrieved holding the value that was
2998 * sent down on the actual send call. This option allows the setting of
2999 * a default context on an association basis that will be received on
3000 * reading messages from the peer. This is especially helpful in the
3001 * one-2-many model for an application to keep some reference to an
3002 * internal state machine that is processing messages on the
3003 * association. Note that the setting of this value only effects
3004 * received messages from the peer and does not effect the value that is
3005 * saved with outbound messages.
3007 static int sctp_setsockopt_context(struct sock
*sk
, char __user
*optval
,
3008 unsigned int optlen
)
3010 struct sctp_assoc_value params
;
3011 struct sctp_sock
*sp
;
3012 struct sctp_association
*asoc
;
3014 if (optlen
!= sizeof(struct sctp_assoc_value
))
3016 if (copy_from_user(¶ms
, optval
, optlen
))
3021 if (params
.assoc_id
!= 0) {
3022 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3025 asoc
->default_rcv_context
= params
.assoc_value
;
3027 sp
->default_rcv_context
= params
.assoc_value
;
3034 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3036 * This options will at a minimum specify if the implementation is doing
3037 * fragmented interleave. Fragmented interleave, for a one to many
3038 * socket, is when subsequent calls to receive a message may return
3039 * parts of messages from different associations. Some implementations
3040 * may allow you to turn this value on or off. If so, when turned off,
3041 * no fragment interleave will occur (which will cause a head of line
3042 * blocking amongst multiple associations sharing the same one to many
3043 * socket). When this option is turned on, then each receive call may
3044 * come from a different association (thus the user must receive data
3045 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3046 * association each receive belongs to.
3048 * This option takes a boolean value. A non-zero value indicates that
3049 * fragmented interleave is on. A value of zero indicates that
3050 * fragmented interleave is off.
3052 * Note that it is important that an implementation that allows this
3053 * option to be turned on, have it off by default. Otherwise an unaware
3054 * application using the one to many model may become confused and act
3057 static int sctp_setsockopt_fragment_interleave(struct sock
*sk
,
3058 char __user
*optval
,
3059 unsigned int optlen
)
3063 if (optlen
!= sizeof(int))
3065 if (get_user(val
, (int __user
*)optval
))
3068 sctp_sk(sk
)->frag_interleave
= (val
== 0) ? 0 : 1;
3074 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3075 * (SCTP_PARTIAL_DELIVERY_POINT)
3077 * This option will set or get the SCTP partial delivery point. This
3078 * point is the size of a message where the partial delivery API will be
3079 * invoked to help free up rwnd space for the peer. Setting this to a
3080 * lower value will cause partial deliveries to happen more often. The
3081 * calls argument is an integer that sets or gets the partial delivery
3082 * point. Note also that the call will fail if the user attempts to set
3083 * this value larger than the socket receive buffer size.
3085 * Note that any single message having a length smaller than or equal to
3086 * the SCTP partial delivery point will be delivered in one single read
3087 * call as long as the user provided buffer is large enough to hold the
3090 static int sctp_setsockopt_partial_delivery_point(struct sock
*sk
,
3091 char __user
*optval
,
3092 unsigned int optlen
)
3096 if (optlen
!= sizeof(u32
))
3098 if (get_user(val
, (int __user
*)optval
))
3101 /* Note: We double the receive buffer from what the user sets
3102 * it to be, also initial rwnd is based on rcvbuf/2.
3104 if (val
> (sk
->sk_rcvbuf
>> 1))
3107 sctp_sk(sk
)->pd_point
= val
;
3109 return 0; /* is this the right error code? */
3113 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3115 * This option will allow a user to change the maximum burst of packets
3116 * that can be emitted by this association. Note that the default value
3117 * is 4, and some implementations may restrict this setting so that it
3118 * can only be lowered.
3120 * NOTE: This text doesn't seem right. Do this on a socket basis with
3121 * future associations inheriting the socket value.
3123 static int sctp_setsockopt_maxburst(struct sock
*sk
,
3124 char __user
*optval
,
3125 unsigned int optlen
)
3127 struct sctp_assoc_value params
;
3128 struct sctp_sock
*sp
;
3129 struct sctp_association
*asoc
;
3133 if (optlen
== sizeof(int)) {
3134 pr_warn("Use of int in max_burst socket option deprecated\n");
3135 pr_warn("Use struct sctp_assoc_value instead\n");
3136 if (copy_from_user(&val
, optval
, optlen
))
3138 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
3139 if (copy_from_user(¶ms
, optval
, optlen
))
3141 val
= params
.assoc_value
;
3142 assoc_id
= params
.assoc_id
;
3148 if (assoc_id
!= 0) {
3149 asoc
= sctp_id2assoc(sk
, assoc_id
);
3152 asoc
->max_burst
= val
;
3154 sp
->max_burst
= val
;
3160 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3162 * This set option adds a chunk type that the user is requesting to be
3163 * received only in an authenticated way. Changes to the list of chunks
3164 * will only effect future associations on the socket.
3166 static int sctp_setsockopt_auth_chunk(struct sock
*sk
,
3167 char __user
*optval
,
3168 unsigned int optlen
)
3170 struct sctp_authchunk val
;
3172 if (!sctp_auth_enable
)
3175 if (optlen
!= sizeof(struct sctp_authchunk
))
3177 if (copy_from_user(&val
, optval
, optlen
))
3180 switch (val
.sauth_chunk
) {
3182 case SCTP_CID_INIT_ACK
:
3183 case SCTP_CID_SHUTDOWN_COMPLETE
:
3188 /* add this chunk id to the endpoint */
3189 return sctp_auth_ep_add_chunkid(sctp_sk(sk
)->ep
, val
.sauth_chunk
);
3193 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3195 * This option gets or sets the list of HMAC algorithms that the local
3196 * endpoint requires the peer to use.
3198 static int sctp_setsockopt_hmac_ident(struct sock
*sk
,
3199 char __user
*optval
,
3200 unsigned int optlen
)
3202 struct sctp_hmacalgo
*hmacs
;
3206 if (!sctp_auth_enable
)
3209 if (optlen
< sizeof(struct sctp_hmacalgo
))
3212 hmacs
= kmalloc(optlen
, GFP_KERNEL
);
3216 if (copy_from_user(hmacs
, optval
, optlen
)) {
3221 idents
= hmacs
->shmac_num_idents
;
3222 if (idents
== 0 || idents
> SCTP_AUTH_NUM_HMACS
||
3223 (idents
* sizeof(u16
)) > (optlen
- sizeof(struct sctp_hmacalgo
))) {
3228 err
= sctp_auth_ep_set_hmacs(sctp_sk(sk
)->ep
, hmacs
);
3235 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3237 * This option will set a shared secret key which is used to build an
3238 * association shared key.
3240 static int sctp_setsockopt_auth_key(struct sock
*sk
,
3241 char __user
*optval
,
3242 unsigned int optlen
)
3244 struct sctp_authkey
*authkey
;
3245 struct sctp_association
*asoc
;
3248 if (!sctp_auth_enable
)
3251 if (optlen
<= sizeof(struct sctp_authkey
))
3254 authkey
= kmalloc(optlen
, GFP_KERNEL
);
3258 if (copy_from_user(authkey
, optval
, optlen
)) {
3263 if (authkey
->sca_keylength
> optlen
- sizeof(struct sctp_authkey
)) {
3268 asoc
= sctp_id2assoc(sk
, authkey
->sca_assoc_id
);
3269 if (!asoc
&& authkey
->sca_assoc_id
&& sctp_style(sk
, UDP
)) {
3274 ret
= sctp_auth_set_key(sctp_sk(sk
)->ep
, asoc
, authkey
);
3281 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3283 * This option will get or set the active shared key to be used to build
3284 * the association shared key.
3286 static int sctp_setsockopt_active_key(struct sock
*sk
,
3287 char __user
*optval
,
3288 unsigned int optlen
)
3290 struct sctp_authkeyid val
;
3291 struct sctp_association
*asoc
;
3293 if (!sctp_auth_enable
)
3296 if (optlen
!= sizeof(struct sctp_authkeyid
))
3298 if (copy_from_user(&val
, optval
, optlen
))
3301 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3302 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3305 return sctp_auth_set_active_key(sctp_sk(sk
)->ep
, asoc
,
3306 val
.scact_keynumber
);
3310 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3312 * This set option will delete a shared secret key from use.
3314 static int sctp_setsockopt_del_key(struct sock
*sk
,
3315 char __user
*optval
,
3316 unsigned int optlen
)
3318 struct sctp_authkeyid val
;
3319 struct sctp_association
*asoc
;
3321 if (!sctp_auth_enable
)
3324 if (optlen
!= sizeof(struct sctp_authkeyid
))
3326 if (copy_from_user(&val
, optval
, optlen
))
3329 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3330 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3333 return sctp_auth_del_key_id(sctp_sk(sk
)->ep
, asoc
,
3334 val
.scact_keynumber
);
3339 /* API 6.2 setsockopt(), getsockopt()
3341 * Applications use setsockopt() and getsockopt() to set or retrieve
3342 * socket options. Socket options are used to change the default
3343 * behavior of sockets calls. They are described in Section 7.
3347 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3348 * int __user *optlen);
3349 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3352 * sd - the socket descript.
3353 * level - set to IPPROTO_SCTP for all SCTP options.
3354 * optname - the option name.
3355 * optval - the buffer to store the value of the option.
3356 * optlen - the size of the buffer.
3358 SCTP_STATIC
int sctp_setsockopt(struct sock
*sk
, int level
, int optname
,
3359 char __user
*optval
, unsigned int optlen
)
3363 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3366 /* I can hardly begin to describe how wrong this is. This is
3367 * so broken as to be worse than useless. The API draft
3368 * REALLY is NOT helpful here... I am not convinced that the
3369 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3370 * are at all well-founded.
3372 if (level
!= SOL_SCTP
) {
3373 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
3374 retval
= af
->setsockopt(sk
, level
, optname
, optval
, optlen
);
3381 case SCTP_SOCKOPT_BINDX_ADD
:
3382 /* 'optlen' is the size of the addresses buffer. */
3383 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3384 optlen
, SCTP_BINDX_ADD_ADDR
);
3387 case SCTP_SOCKOPT_BINDX_REM
:
3388 /* 'optlen' is the size of the addresses buffer. */
3389 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3390 optlen
, SCTP_BINDX_REM_ADDR
);
3393 case SCTP_SOCKOPT_CONNECTX_OLD
:
3394 /* 'optlen' is the size of the addresses buffer. */
3395 retval
= sctp_setsockopt_connectx_old(sk
,
3396 (struct sockaddr __user
*)optval
,
3400 case SCTP_SOCKOPT_CONNECTX
:
3401 /* 'optlen' is the size of the addresses buffer. */
3402 retval
= sctp_setsockopt_connectx(sk
,
3403 (struct sockaddr __user
*)optval
,
3407 case SCTP_DISABLE_FRAGMENTS
:
3408 retval
= sctp_setsockopt_disable_fragments(sk
, optval
, optlen
);
3412 retval
= sctp_setsockopt_events(sk
, optval
, optlen
);
3415 case SCTP_AUTOCLOSE
:
3416 retval
= sctp_setsockopt_autoclose(sk
, optval
, optlen
);
3419 case SCTP_PEER_ADDR_PARAMS
:
3420 retval
= sctp_setsockopt_peer_addr_params(sk
, optval
, optlen
);
3423 case SCTP_DELAYED_ACK
:
3424 retval
= sctp_setsockopt_delayed_ack(sk
, optval
, optlen
);
3426 case SCTP_PARTIAL_DELIVERY_POINT
:
3427 retval
= sctp_setsockopt_partial_delivery_point(sk
, optval
, optlen
);
3431 retval
= sctp_setsockopt_initmsg(sk
, optval
, optlen
);
3433 case SCTP_DEFAULT_SEND_PARAM
:
3434 retval
= sctp_setsockopt_default_send_param(sk
, optval
,
3437 case SCTP_PRIMARY_ADDR
:
3438 retval
= sctp_setsockopt_primary_addr(sk
, optval
, optlen
);
3440 case SCTP_SET_PEER_PRIMARY_ADDR
:
3441 retval
= sctp_setsockopt_peer_primary_addr(sk
, optval
, optlen
);
3444 retval
= sctp_setsockopt_nodelay(sk
, optval
, optlen
);
3447 retval
= sctp_setsockopt_rtoinfo(sk
, optval
, optlen
);
3449 case SCTP_ASSOCINFO
:
3450 retval
= sctp_setsockopt_associnfo(sk
, optval
, optlen
);
3452 case SCTP_I_WANT_MAPPED_V4_ADDR
:
3453 retval
= sctp_setsockopt_mappedv4(sk
, optval
, optlen
);
3456 retval
= sctp_setsockopt_maxseg(sk
, optval
, optlen
);
3458 case SCTP_ADAPTATION_LAYER
:
3459 retval
= sctp_setsockopt_adaptation_layer(sk
, optval
, optlen
);
3462 retval
= sctp_setsockopt_context(sk
, optval
, optlen
);
3464 case SCTP_FRAGMENT_INTERLEAVE
:
3465 retval
= sctp_setsockopt_fragment_interleave(sk
, optval
, optlen
);
3467 case SCTP_MAX_BURST
:
3468 retval
= sctp_setsockopt_maxburst(sk
, optval
, optlen
);
3470 case SCTP_AUTH_CHUNK
:
3471 retval
= sctp_setsockopt_auth_chunk(sk
, optval
, optlen
);
3473 case SCTP_HMAC_IDENT
:
3474 retval
= sctp_setsockopt_hmac_ident(sk
, optval
, optlen
);
3477 retval
= sctp_setsockopt_auth_key(sk
, optval
, optlen
);
3479 case SCTP_AUTH_ACTIVE_KEY
:
3480 retval
= sctp_setsockopt_active_key(sk
, optval
, optlen
);
3482 case SCTP_AUTH_DELETE_KEY
:
3483 retval
= sctp_setsockopt_del_key(sk
, optval
, optlen
);
3486 retval
= -ENOPROTOOPT
;
3490 sctp_release_sock(sk
);
3496 /* API 3.1.6 connect() - UDP Style Syntax
3498 * An application may use the connect() call in the UDP model to initiate an
3499 * association without sending data.
3503 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3505 * sd: the socket descriptor to have a new association added to.
3507 * nam: the address structure (either struct sockaddr_in or struct
3508 * sockaddr_in6 defined in RFC2553 [7]).
3510 * len: the size of the address.
3512 SCTP_STATIC
int sctp_connect(struct sock
*sk
, struct sockaddr
*addr
,
3520 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3521 __func__
, sk
, addr
, addr_len
);
3523 /* Validate addr_len before calling common connect/connectx routine. */
3524 af
= sctp_get_af_specific(addr
->sa_family
);
3525 if (!af
|| addr_len
< af
->sockaddr_len
) {
3528 /* Pass correct addr len to common routine (so it knows there
3529 * is only one address being passed.
3531 err
= __sctp_connect(sk
, addr
, af
->sockaddr_len
, NULL
);
3534 sctp_release_sock(sk
);
3538 /* FIXME: Write comments. */
3539 SCTP_STATIC
int sctp_disconnect(struct sock
*sk
, int flags
)
3541 return -EOPNOTSUPP
; /* STUB */
3544 /* 4.1.4 accept() - TCP Style Syntax
3546 * Applications use accept() call to remove an established SCTP
3547 * association from the accept queue of the endpoint. A new socket
3548 * descriptor will be returned from accept() to represent the newly
3549 * formed association.
3551 SCTP_STATIC
struct sock
*sctp_accept(struct sock
*sk
, int flags
, int *err
)
3553 struct sctp_sock
*sp
;
3554 struct sctp_endpoint
*ep
;
3555 struct sock
*newsk
= NULL
;
3556 struct sctp_association
*asoc
;
3565 if (!sctp_style(sk
, TCP
)) {
3566 error
= -EOPNOTSUPP
;
3570 if (!sctp_sstate(sk
, LISTENING
)) {
3575 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
3577 error
= sctp_wait_for_accept(sk
, timeo
);
3581 /* We treat the list of associations on the endpoint as the accept
3582 * queue and pick the first association on the list.
3584 asoc
= list_entry(ep
->asocs
.next
, struct sctp_association
, asocs
);
3586 newsk
= sp
->pf
->create_accept_sk(sk
, asoc
);
3592 /* Populate the fields of the newsk from the oldsk and migrate the
3593 * asoc to the newsk.
3595 sctp_sock_migrate(sk
, newsk
, asoc
, SCTP_SOCKET_TCP
);
3598 sctp_release_sock(sk
);
3603 /* The SCTP ioctl handler. */
3604 SCTP_STATIC
int sctp_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
3611 * SEQPACKET-style sockets in LISTENING state are valid, for
3612 * SCTP, so only discard TCP-style sockets in LISTENING state.
3614 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
3619 struct sk_buff
*skb
;
3620 unsigned int amount
= 0;
3622 skb
= skb_peek(&sk
->sk_receive_queue
);
3625 * We will only return the amount of this packet since
3626 * that is all that will be read.
3630 rc
= put_user(amount
, (int __user
*)arg
);
3638 sctp_release_sock(sk
);
3642 /* This is the function which gets called during socket creation to
3643 * initialized the SCTP-specific portion of the sock.
3644 * The sock structure should already be zero-filled memory.
3646 SCTP_STATIC
int sctp_init_sock(struct sock
*sk
)
3648 struct sctp_endpoint
*ep
;
3649 struct sctp_sock
*sp
;
3651 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk
);
3655 /* Initialize the SCTP per socket area. */
3656 switch (sk
->sk_type
) {
3657 case SOCK_SEQPACKET
:
3658 sp
->type
= SCTP_SOCKET_UDP
;
3661 sp
->type
= SCTP_SOCKET_TCP
;
3664 return -ESOCKTNOSUPPORT
;
3667 /* Initialize default send parameters. These parameters can be
3668 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3670 sp
->default_stream
= 0;
3671 sp
->default_ppid
= 0;
3672 sp
->default_flags
= 0;
3673 sp
->default_context
= 0;
3674 sp
->default_timetolive
= 0;
3676 sp
->default_rcv_context
= 0;
3677 sp
->max_burst
= sctp_max_burst
;
3679 /* Initialize default setup parameters. These parameters
3680 * can be modified with the SCTP_INITMSG socket option or
3681 * overridden by the SCTP_INIT CMSG.
3683 sp
->initmsg
.sinit_num_ostreams
= sctp_max_outstreams
;
3684 sp
->initmsg
.sinit_max_instreams
= sctp_max_instreams
;
3685 sp
->initmsg
.sinit_max_attempts
= sctp_max_retrans_init
;
3686 sp
->initmsg
.sinit_max_init_timeo
= sctp_rto_max
;
3688 /* Initialize default RTO related parameters. These parameters can
3689 * be modified for with the SCTP_RTOINFO socket option.
3691 sp
->rtoinfo
.srto_initial
= sctp_rto_initial
;
3692 sp
->rtoinfo
.srto_max
= sctp_rto_max
;
3693 sp
->rtoinfo
.srto_min
= sctp_rto_min
;
3695 /* Initialize default association related parameters. These parameters
3696 * can be modified with the SCTP_ASSOCINFO socket option.
3698 sp
->assocparams
.sasoc_asocmaxrxt
= sctp_max_retrans_association
;
3699 sp
->assocparams
.sasoc_number_peer_destinations
= 0;
3700 sp
->assocparams
.sasoc_peer_rwnd
= 0;
3701 sp
->assocparams
.sasoc_local_rwnd
= 0;
3702 sp
->assocparams
.sasoc_cookie_life
= sctp_valid_cookie_life
;
3704 /* Initialize default event subscriptions. By default, all the
3707 memset(&sp
->subscribe
, 0, sizeof(struct sctp_event_subscribe
));
3709 /* Default Peer Address Parameters. These defaults can
3710 * be modified via SCTP_PEER_ADDR_PARAMS
3712 sp
->hbinterval
= sctp_hb_interval
;
3713 sp
->pathmaxrxt
= sctp_max_retrans_path
;
3714 sp
->pathmtu
= 0; // allow default discovery
3715 sp
->sackdelay
= sctp_sack_timeout
;
3717 sp
->param_flags
= SPP_HB_ENABLE
|
3719 SPP_SACKDELAY_ENABLE
;
3721 /* If enabled no SCTP message fragmentation will be performed.
3722 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3724 sp
->disable_fragments
= 0;
3726 /* Enable Nagle algorithm by default. */
3729 /* Enable by default. */
3732 /* Auto-close idle associations after the configured
3733 * number of seconds. A value of 0 disables this
3734 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3735 * for UDP-style sockets only.
3739 /* User specified fragmentation limit. */
3742 sp
->adaptation_ind
= 0;
3744 sp
->pf
= sctp_get_pf_specific(sk
->sk_family
);
3746 /* Control variables for partial data delivery. */
3747 atomic_set(&sp
->pd_mode
, 0);
3748 skb_queue_head_init(&sp
->pd_lobby
);
3749 sp
->frag_interleave
= 0;
3751 /* Create a per socket endpoint structure. Even if we
3752 * change the data structure relationships, this may still
3753 * be useful for storing pre-connect address information.
3755 ep
= sctp_endpoint_new(sk
, GFP_KERNEL
);
3762 SCTP_DBG_OBJCNT_INC(sock
);
3765 percpu_counter_inc(&sctp_sockets_allocated
);
3766 sock_prot_inuse_add(sock_net(sk
), sk
->sk_prot
, 1);
3772 /* Cleanup any SCTP per socket resources. */
3773 SCTP_STATIC
void sctp_destroy_sock(struct sock
*sk
)
3775 struct sctp_endpoint
*ep
;
3777 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk
);
3779 /* Release our hold on the endpoint. */
3780 ep
= sctp_sk(sk
)->ep
;
3781 sctp_endpoint_free(ep
);
3783 percpu_counter_dec(&sctp_sockets_allocated
);
3784 sock_prot_inuse_add(sock_net(sk
), sk
->sk_prot
, -1);
3788 /* API 4.1.7 shutdown() - TCP Style Syntax
3789 * int shutdown(int socket, int how);
3791 * sd - the socket descriptor of the association to be closed.
3792 * how - Specifies the type of shutdown. The values are
3795 * Disables further receive operations. No SCTP
3796 * protocol action is taken.
3798 * Disables further send operations, and initiates
3799 * the SCTP shutdown sequence.
3801 * Disables further send and receive operations
3802 * and initiates the SCTP shutdown sequence.
3804 SCTP_STATIC
void sctp_shutdown(struct sock
*sk
, int how
)
3806 struct sctp_endpoint
*ep
;
3807 struct sctp_association
*asoc
;
3809 if (!sctp_style(sk
, TCP
))
3812 if (how
& SEND_SHUTDOWN
) {
3813 ep
= sctp_sk(sk
)->ep
;
3814 if (!list_empty(&ep
->asocs
)) {
3815 asoc
= list_entry(ep
->asocs
.next
,
3816 struct sctp_association
, asocs
);
3817 sctp_primitive_SHUTDOWN(asoc
, NULL
);
3822 /* 7.2.1 Association Status (SCTP_STATUS)
3824 * Applications can retrieve current status information about an
3825 * association, including association state, peer receiver window size,
3826 * number of unacked data chunks, and number of data chunks pending
3827 * receipt. This information is read-only.
3829 static int sctp_getsockopt_sctp_status(struct sock
*sk
, int len
,
3830 char __user
*optval
,
3833 struct sctp_status status
;
3834 struct sctp_association
*asoc
= NULL
;
3835 struct sctp_transport
*transport
;
3836 sctp_assoc_t associd
;
3839 if (len
< sizeof(status
)) {
3844 len
= sizeof(status
);
3845 if (copy_from_user(&status
, optval
, len
)) {
3850 associd
= status
.sstat_assoc_id
;
3851 asoc
= sctp_id2assoc(sk
, associd
);
3857 transport
= asoc
->peer
.primary_path
;
3859 status
.sstat_assoc_id
= sctp_assoc2id(asoc
);
3860 status
.sstat_state
= asoc
->state
;
3861 status
.sstat_rwnd
= asoc
->peer
.rwnd
;
3862 status
.sstat_unackdata
= asoc
->unack_data
;
3864 status
.sstat_penddata
= sctp_tsnmap_pending(&asoc
->peer
.tsn_map
);
3865 status
.sstat_instrms
= asoc
->c
.sinit_max_instreams
;
3866 status
.sstat_outstrms
= asoc
->c
.sinit_num_ostreams
;
3867 status
.sstat_fragmentation_point
= asoc
->frag_point
;
3868 status
.sstat_primary
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
3869 memcpy(&status
.sstat_primary
.spinfo_address
, &transport
->ipaddr
,
3870 transport
->af_specific
->sockaddr_len
);
3871 /* Map ipv4 address into v4-mapped-on-v6 address. */
3872 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
3873 (union sctp_addr
*)&status
.sstat_primary
.spinfo_address
);
3874 status
.sstat_primary
.spinfo_state
= transport
->state
;
3875 status
.sstat_primary
.spinfo_cwnd
= transport
->cwnd
;
3876 status
.sstat_primary
.spinfo_srtt
= transport
->srtt
;
3877 status
.sstat_primary
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
3878 status
.sstat_primary
.spinfo_mtu
= transport
->pathmtu
;
3880 if (status
.sstat_primary
.spinfo_state
== SCTP_UNKNOWN
)
3881 status
.sstat_primary
.spinfo_state
= SCTP_ACTIVE
;
3883 if (put_user(len
, optlen
)) {
3888 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3889 len
, status
.sstat_state
, status
.sstat_rwnd
,
3890 status
.sstat_assoc_id
);
3892 if (copy_to_user(optval
, &status
, len
)) {
3902 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3904 * Applications can retrieve information about a specific peer address
3905 * of an association, including its reachability state, congestion
3906 * window, and retransmission timer values. This information is
3909 static int sctp_getsockopt_peer_addr_info(struct sock
*sk
, int len
,
3910 char __user
*optval
,
3913 struct sctp_paddrinfo pinfo
;
3914 struct sctp_transport
*transport
;
3917 if (len
< sizeof(pinfo
)) {
3922 len
= sizeof(pinfo
);
3923 if (copy_from_user(&pinfo
, optval
, len
)) {
3928 transport
= sctp_addr_id2transport(sk
, &pinfo
.spinfo_address
,
3929 pinfo
.spinfo_assoc_id
);
3933 pinfo
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
3934 pinfo
.spinfo_state
= transport
->state
;
3935 pinfo
.spinfo_cwnd
= transport
->cwnd
;
3936 pinfo
.spinfo_srtt
= transport
->srtt
;
3937 pinfo
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
3938 pinfo
.spinfo_mtu
= transport
->pathmtu
;
3940 if (pinfo
.spinfo_state
== SCTP_UNKNOWN
)
3941 pinfo
.spinfo_state
= SCTP_ACTIVE
;
3943 if (put_user(len
, optlen
)) {
3948 if (copy_to_user(optval
, &pinfo
, len
)) {
3957 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3959 * This option is a on/off flag. If enabled no SCTP message
3960 * fragmentation will be performed. Instead if a message being sent
3961 * exceeds the current PMTU size, the message will NOT be sent and
3962 * instead a error will be indicated to the user.
3964 static int sctp_getsockopt_disable_fragments(struct sock
*sk
, int len
,
3965 char __user
*optval
, int __user
*optlen
)
3969 if (len
< sizeof(int))
3973 val
= (sctp_sk(sk
)->disable_fragments
== 1);
3974 if (put_user(len
, optlen
))
3976 if (copy_to_user(optval
, &val
, len
))
3981 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3983 * This socket option is used to specify various notifications and
3984 * ancillary data the user wishes to receive.
3986 static int sctp_getsockopt_events(struct sock
*sk
, int len
, char __user
*optval
,
3989 if (len
< sizeof(struct sctp_event_subscribe
))
3991 len
= sizeof(struct sctp_event_subscribe
);
3992 if (put_user(len
, optlen
))
3994 if (copy_to_user(optval
, &sctp_sk(sk
)->subscribe
, len
))
3999 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4001 * This socket option is applicable to the UDP-style socket only. When
4002 * set it will cause associations that are idle for more than the
4003 * specified number of seconds to automatically close. An association
4004 * being idle is defined an association that has NOT sent or received
4005 * user data. The special value of '0' indicates that no automatic
4006 * close of any associations should be performed. The option expects an
4007 * integer defining the number of seconds of idle time before an
4008 * association is closed.
4010 static int sctp_getsockopt_autoclose(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4012 /* Applicable to UDP-style socket only */
4013 if (sctp_style(sk
, TCP
))
4015 if (len
< sizeof(int))
4018 if (put_user(len
, optlen
))
4020 if (copy_to_user(optval
, &sctp_sk(sk
)->autoclose
, sizeof(int)))
4025 /* Helper routine to branch off an association to a new socket. */
4026 SCTP_STATIC
int sctp_do_peeloff(struct sctp_association
*asoc
,
4027 struct socket
**sockp
)
4029 struct sock
*sk
= asoc
->base
.sk
;
4030 struct socket
*sock
;
4034 /* An association cannot be branched off from an already peeled-off
4035 * socket, nor is this supported for tcp style sockets.
4037 if (!sctp_style(sk
, UDP
))
4040 /* Create a new socket. */
4041 err
= sock_create(sk
->sk_family
, SOCK_SEQPACKET
, IPPROTO_SCTP
, &sock
);
4045 sctp_copy_sock(sock
->sk
, sk
, asoc
);
4047 /* Make peeled-off sockets more like 1-1 accepted sockets.
4048 * Set the daddr and initialize id to something more random
4050 af
= sctp_get_af_specific(asoc
->peer
.primary_addr
.sa
.sa_family
);
4051 af
->to_sk_daddr(&asoc
->peer
.primary_addr
, sk
);
4053 /* Populate the fields of the newsk from the oldsk and migrate the
4054 * asoc to the newsk.
4056 sctp_sock_migrate(sk
, sock
->sk
, asoc
, SCTP_SOCKET_UDP_HIGH_BANDWIDTH
);
4063 static int sctp_getsockopt_peeloff(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4065 sctp_peeloff_arg_t peeloff
;
4066 struct socket
*newsock
;
4068 struct sctp_association
*asoc
;
4070 if (len
< sizeof(sctp_peeloff_arg_t
))
4072 len
= sizeof(sctp_peeloff_arg_t
);
4073 if (copy_from_user(&peeloff
, optval
, len
))
4076 asoc
= sctp_id2assoc(sk
, peeloff
.associd
);
4082 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__
, sk
, asoc
);
4084 retval
= sctp_do_peeloff(asoc
, &newsock
);
4088 /* Map the socket to an unused fd that can be returned to the user. */
4089 retval
= sock_map_fd(newsock
, 0);
4091 sock_release(newsock
);
4095 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
4096 __func__
, sk
, asoc
, newsock
->sk
, retval
);
4098 /* Return the fd mapped to the new socket. */
4099 peeloff
.sd
= retval
;
4100 if (put_user(len
, optlen
))
4102 if (copy_to_user(optval
, &peeloff
, len
))
4109 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4111 * Applications can enable or disable heartbeats for any peer address of
4112 * an association, modify an address's heartbeat interval, force a
4113 * heartbeat to be sent immediately, and adjust the address's maximum
4114 * number of retransmissions sent before an address is considered
4115 * unreachable. The following structure is used to access and modify an
4116 * address's parameters:
4118 * struct sctp_paddrparams {
4119 * sctp_assoc_t spp_assoc_id;
4120 * struct sockaddr_storage spp_address;
4121 * uint32_t spp_hbinterval;
4122 * uint16_t spp_pathmaxrxt;
4123 * uint32_t spp_pathmtu;
4124 * uint32_t spp_sackdelay;
4125 * uint32_t spp_flags;
4128 * spp_assoc_id - (one-to-many style socket) This is filled in the
4129 * application, and identifies the association for
4131 * spp_address - This specifies which address is of interest.
4132 * spp_hbinterval - This contains the value of the heartbeat interval,
4133 * in milliseconds. If a value of zero
4134 * is present in this field then no changes are to
4135 * be made to this parameter.
4136 * spp_pathmaxrxt - This contains the maximum number of
4137 * retransmissions before this address shall be
4138 * considered unreachable. If a value of zero
4139 * is present in this field then no changes are to
4140 * be made to this parameter.
4141 * spp_pathmtu - When Path MTU discovery is disabled the value
4142 * specified here will be the "fixed" path mtu.
4143 * Note that if the spp_address field is empty
4144 * then all associations on this address will
4145 * have this fixed path mtu set upon them.
4147 * spp_sackdelay - When delayed sack is enabled, this value specifies
4148 * the number of milliseconds that sacks will be delayed
4149 * for. This value will apply to all addresses of an
4150 * association if the spp_address field is empty. Note
4151 * also, that if delayed sack is enabled and this
4152 * value is set to 0, no change is made to the last
4153 * recorded delayed sack timer value.
4155 * spp_flags - These flags are used to control various features
4156 * on an association. The flag field may contain
4157 * zero or more of the following options.
4159 * SPP_HB_ENABLE - Enable heartbeats on the
4160 * specified address. Note that if the address
4161 * field is empty all addresses for the association
4162 * have heartbeats enabled upon them.
4164 * SPP_HB_DISABLE - Disable heartbeats on the
4165 * speicifed address. Note that if the address
4166 * field is empty all addresses for the association
4167 * will have their heartbeats disabled. Note also
4168 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4169 * mutually exclusive, only one of these two should
4170 * be specified. Enabling both fields will have
4171 * undetermined results.
4173 * SPP_HB_DEMAND - Request a user initiated heartbeat
4174 * to be made immediately.
4176 * SPP_PMTUD_ENABLE - This field will enable PMTU
4177 * discovery upon the specified address. Note that
4178 * if the address feild is empty then all addresses
4179 * on the association are effected.
4181 * SPP_PMTUD_DISABLE - This field will disable PMTU
4182 * discovery upon the specified address. Note that
4183 * if the address feild is empty then all addresses
4184 * on the association are effected. Not also that
4185 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4186 * exclusive. Enabling both will have undetermined
4189 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4190 * on delayed sack. The time specified in spp_sackdelay
4191 * is used to specify the sack delay for this address. Note
4192 * that if spp_address is empty then all addresses will
4193 * enable delayed sack and take on the sack delay
4194 * value specified in spp_sackdelay.
4195 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4196 * off delayed sack. If the spp_address field is blank then
4197 * delayed sack is disabled for the entire association. Note
4198 * also that this field is mutually exclusive to
4199 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4202 static int sctp_getsockopt_peer_addr_params(struct sock
*sk
, int len
,
4203 char __user
*optval
, int __user
*optlen
)
4205 struct sctp_paddrparams params
;
4206 struct sctp_transport
*trans
= NULL
;
4207 struct sctp_association
*asoc
= NULL
;
4208 struct sctp_sock
*sp
= sctp_sk(sk
);
4210 if (len
< sizeof(struct sctp_paddrparams
))
4212 len
= sizeof(struct sctp_paddrparams
);
4213 if (copy_from_user(¶ms
, optval
, len
))
4216 /* If an address other than INADDR_ANY is specified, and
4217 * no transport is found, then the request is invalid.
4219 if (!sctp_is_any(sk
, ( union sctp_addr
*)¶ms
.spp_address
)) {
4220 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
4221 params
.spp_assoc_id
);
4223 SCTP_DEBUG_PRINTK("Failed no transport\n");
4228 /* Get association, if assoc_id != 0 and the socket is a one
4229 * to many style socket, and an association was not found, then
4230 * the id was invalid.
4232 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
4233 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
)) {
4234 SCTP_DEBUG_PRINTK("Failed no association\n");
4239 /* Fetch transport values. */
4240 params
.spp_hbinterval
= jiffies_to_msecs(trans
->hbinterval
);
4241 params
.spp_pathmtu
= trans
->pathmtu
;
4242 params
.spp_pathmaxrxt
= trans
->pathmaxrxt
;
4243 params
.spp_sackdelay
= jiffies_to_msecs(trans
->sackdelay
);
4245 /*draft-11 doesn't say what to return in spp_flags*/
4246 params
.spp_flags
= trans
->param_flags
;
4248 /* Fetch association values. */
4249 params
.spp_hbinterval
= jiffies_to_msecs(asoc
->hbinterval
);
4250 params
.spp_pathmtu
= asoc
->pathmtu
;
4251 params
.spp_pathmaxrxt
= asoc
->pathmaxrxt
;
4252 params
.spp_sackdelay
= jiffies_to_msecs(asoc
->sackdelay
);
4254 /*draft-11 doesn't say what to return in spp_flags*/
4255 params
.spp_flags
= asoc
->param_flags
;
4257 /* Fetch socket values. */
4258 params
.spp_hbinterval
= sp
->hbinterval
;
4259 params
.spp_pathmtu
= sp
->pathmtu
;
4260 params
.spp_sackdelay
= sp
->sackdelay
;
4261 params
.spp_pathmaxrxt
= sp
->pathmaxrxt
;
4263 /*draft-11 doesn't say what to return in spp_flags*/
4264 params
.spp_flags
= sp
->param_flags
;
4267 if (copy_to_user(optval
, ¶ms
, len
))
4270 if (put_user(len
, optlen
))
4277 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4279 * This option will effect the way delayed acks are performed. This
4280 * option allows you to get or set the delayed ack time, in
4281 * milliseconds. It also allows changing the delayed ack frequency.
4282 * Changing the frequency to 1 disables the delayed sack algorithm. If
4283 * the assoc_id is 0, then this sets or gets the endpoints default
4284 * values. If the assoc_id field is non-zero, then the set or get
4285 * effects the specified association for the one to many model (the
4286 * assoc_id field is ignored by the one to one model). Note that if
4287 * sack_delay or sack_freq are 0 when setting this option, then the
4288 * current values will remain unchanged.
4290 * struct sctp_sack_info {
4291 * sctp_assoc_t sack_assoc_id;
4292 * uint32_t sack_delay;
4293 * uint32_t sack_freq;
4296 * sack_assoc_id - This parameter, indicates which association the user
4297 * is performing an action upon. Note that if this field's value is
4298 * zero then the endpoints default value is changed (effecting future
4299 * associations only).
4301 * sack_delay - This parameter contains the number of milliseconds that
4302 * the user is requesting the delayed ACK timer be set to. Note that
4303 * this value is defined in the standard to be between 200 and 500
4306 * sack_freq - This parameter contains the number of packets that must
4307 * be received before a sack is sent without waiting for the delay
4308 * timer to expire. The default value for this is 2, setting this
4309 * value to 1 will disable the delayed sack algorithm.
4311 static int sctp_getsockopt_delayed_ack(struct sock
*sk
, int len
,
4312 char __user
*optval
,
4315 struct sctp_sack_info params
;
4316 struct sctp_association
*asoc
= NULL
;
4317 struct sctp_sock
*sp
= sctp_sk(sk
);
4319 if (len
>= sizeof(struct sctp_sack_info
)) {
4320 len
= sizeof(struct sctp_sack_info
);
4322 if (copy_from_user(¶ms
, optval
, len
))
4324 } else if (len
== sizeof(struct sctp_assoc_value
)) {
4325 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4326 pr_warn("Use struct sctp_sack_info instead\n");
4327 if (copy_from_user(¶ms
, optval
, len
))
4332 /* Get association, if sack_assoc_id != 0 and the socket is a one
4333 * to many style socket, and an association was not found, then
4334 * the id was invalid.
4336 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
4337 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
4341 /* Fetch association values. */
4342 if (asoc
->param_flags
& SPP_SACKDELAY_ENABLE
) {
4343 params
.sack_delay
= jiffies_to_msecs(
4345 params
.sack_freq
= asoc
->sackfreq
;
4348 params
.sack_delay
= 0;
4349 params
.sack_freq
= 1;
4352 /* Fetch socket values. */
4353 if (sp
->param_flags
& SPP_SACKDELAY_ENABLE
) {
4354 params
.sack_delay
= sp
->sackdelay
;
4355 params
.sack_freq
= sp
->sackfreq
;
4357 params
.sack_delay
= 0;
4358 params
.sack_freq
= 1;
4362 if (copy_to_user(optval
, ¶ms
, len
))
4365 if (put_user(len
, optlen
))
4371 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4373 * Applications can specify protocol parameters for the default association
4374 * initialization. The option name argument to setsockopt() and getsockopt()
4377 * Setting initialization parameters is effective only on an unconnected
4378 * socket (for UDP-style sockets only future associations are effected
4379 * by the change). With TCP-style sockets, this option is inherited by
4380 * sockets derived from a listener socket.
4382 static int sctp_getsockopt_initmsg(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4384 if (len
< sizeof(struct sctp_initmsg
))
4386 len
= sizeof(struct sctp_initmsg
);
4387 if (put_user(len
, optlen
))
4389 if (copy_to_user(optval
, &sctp_sk(sk
)->initmsg
, len
))
4395 static int sctp_getsockopt_peer_addrs(struct sock
*sk
, int len
,
4396 char __user
*optval
, int __user
*optlen
)
4398 struct sctp_association
*asoc
;
4400 struct sctp_getaddrs getaddrs
;
4401 struct sctp_transport
*from
;
4403 union sctp_addr temp
;
4404 struct sctp_sock
*sp
= sctp_sk(sk
);
4409 if (len
< sizeof(struct sctp_getaddrs
))
4412 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
4415 /* For UDP-style sockets, id specifies the association to query. */
4416 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
4420 to
= optval
+ offsetof(struct sctp_getaddrs
,addrs
);
4421 space_left
= len
- offsetof(struct sctp_getaddrs
,addrs
);
4423 list_for_each_entry(from
, &asoc
->peer
.transport_addr_list
,
4425 memcpy(&temp
, &from
->ipaddr
, sizeof(temp
));
4426 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
4427 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4428 if (space_left
< addrlen
)
4430 if (copy_to_user(to
, &temp
, addrlen
))
4434 space_left
-= addrlen
;
4437 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
))
4439 bytes_copied
= ((char __user
*)to
) - optval
;
4440 if (put_user(bytes_copied
, optlen
))
4446 static int sctp_copy_laddrs(struct sock
*sk
, __u16 port
, void *to
,
4447 size_t space_left
, int *bytes_copied
)
4449 struct sctp_sockaddr_entry
*addr
;
4450 union sctp_addr temp
;
4455 list_for_each_entry_rcu(addr
, &sctp_local_addr_list
, list
) {
4459 if ((PF_INET
== sk
->sk_family
) &&
4460 (AF_INET6
== addr
->a
.sa
.sa_family
))
4462 if ((PF_INET6
== sk
->sk_family
) &&
4463 inet_v6_ipv6only(sk
) &&
4464 (AF_INET
== addr
->a
.sa
.sa_family
))
4466 memcpy(&temp
, &addr
->a
, sizeof(temp
));
4467 if (!temp
.v4
.sin_port
)
4468 temp
.v4
.sin_port
= htons(port
);
4470 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
4472 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4473 if (space_left
< addrlen
) {
4477 memcpy(to
, &temp
, addrlen
);
4481 space_left
-= addrlen
;
4482 *bytes_copied
+= addrlen
;
4490 static int sctp_getsockopt_local_addrs(struct sock
*sk
, int len
,
4491 char __user
*optval
, int __user
*optlen
)
4493 struct sctp_bind_addr
*bp
;
4494 struct sctp_association
*asoc
;
4496 struct sctp_getaddrs getaddrs
;
4497 struct sctp_sockaddr_entry
*addr
;
4499 union sctp_addr temp
;
4500 struct sctp_sock
*sp
= sctp_sk(sk
);
4504 int bytes_copied
= 0;
4508 if (len
< sizeof(struct sctp_getaddrs
))
4511 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
4515 * For UDP-style sockets, id specifies the association to query.
4516 * If the id field is set to the value '0' then the locally bound
4517 * addresses are returned without regard to any particular
4520 if (0 == getaddrs
.assoc_id
) {
4521 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
4523 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
4526 bp
= &asoc
->base
.bind_addr
;
4529 to
= optval
+ offsetof(struct sctp_getaddrs
,addrs
);
4530 space_left
= len
- offsetof(struct sctp_getaddrs
,addrs
);
4532 addrs
= kmalloc(space_left
, GFP_KERNEL
);
4536 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4537 * addresses from the global local address list.
4539 if (sctp_list_single_entry(&bp
->address_list
)) {
4540 addr
= list_entry(bp
->address_list
.next
,
4541 struct sctp_sockaddr_entry
, list
);
4542 if (sctp_is_any(sk
, &addr
->a
)) {
4543 cnt
= sctp_copy_laddrs(sk
, bp
->port
, addrs
,
4544 space_left
, &bytes_copied
);
4554 /* Protection on the bound address list is not needed since
4555 * in the socket option context we hold a socket lock and
4556 * thus the bound address list can't change.
4558 list_for_each_entry(addr
, &bp
->address_list
, list
) {
4559 memcpy(&temp
, &addr
->a
, sizeof(temp
));
4560 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
4561 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4562 if (space_left
< addrlen
) {
4563 err
= -ENOMEM
; /*fixme: right error?*/
4566 memcpy(buf
, &temp
, addrlen
);
4568 bytes_copied
+= addrlen
;
4570 space_left
-= addrlen
;
4574 if (copy_to_user(to
, addrs
, bytes_copied
)) {
4578 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
)) {
4582 if (put_user(bytes_copied
, optlen
))
4589 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4591 * Requests that the local SCTP stack use the enclosed peer address as
4592 * the association primary. The enclosed address must be one of the
4593 * association peer's addresses.
4595 static int sctp_getsockopt_primary_addr(struct sock
*sk
, int len
,
4596 char __user
*optval
, int __user
*optlen
)
4598 struct sctp_prim prim
;
4599 struct sctp_association
*asoc
;
4600 struct sctp_sock
*sp
= sctp_sk(sk
);
4602 if (len
< sizeof(struct sctp_prim
))
4605 len
= sizeof(struct sctp_prim
);
4607 if (copy_from_user(&prim
, optval
, len
))
4610 asoc
= sctp_id2assoc(sk
, prim
.ssp_assoc_id
);
4614 if (!asoc
->peer
.primary_path
)
4617 memcpy(&prim
.ssp_addr
, &asoc
->peer
.primary_path
->ipaddr
,
4618 asoc
->peer
.primary_path
->af_specific
->sockaddr_len
);
4620 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
,
4621 (union sctp_addr
*)&prim
.ssp_addr
);
4623 if (put_user(len
, optlen
))
4625 if (copy_to_user(optval
, &prim
, len
))
4632 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4634 * Requests that the local endpoint set the specified Adaptation Layer
4635 * Indication parameter for all future INIT and INIT-ACK exchanges.
4637 static int sctp_getsockopt_adaptation_layer(struct sock
*sk
, int len
,
4638 char __user
*optval
, int __user
*optlen
)
4640 struct sctp_setadaptation adaptation
;
4642 if (len
< sizeof(struct sctp_setadaptation
))
4645 len
= sizeof(struct sctp_setadaptation
);
4647 adaptation
.ssb_adaptation_ind
= sctp_sk(sk
)->adaptation_ind
;
4649 if (put_user(len
, optlen
))
4651 if (copy_to_user(optval
, &adaptation
, len
))
4659 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4661 * Applications that wish to use the sendto() system call may wish to
4662 * specify a default set of parameters that would normally be supplied
4663 * through the inclusion of ancillary data. This socket option allows
4664 * such an application to set the default sctp_sndrcvinfo structure.
4667 * The application that wishes to use this socket option simply passes
4668 * in to this call the sctp_sndrcvinfo structure defined in Section
4669 * 5.2.2) The input parameters accepted by this call include
4670 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4671 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4672 * to this call if the caller is using the UDP model.
4674 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4676 static int sctp_getsockopt_default_send_param(struct sock
*sk
,
4677 int len
, char __user
*optval
,
4680 struct sctp_sndrcvinfo info
;
4681 struct sctp_association
*asoc
;
4682 struct sctp_sock
*sp
= sctp_sk(sk
);
4684 if (len
< sizeof(struct sctp_sndrcvinfo
))
4687 len
= sizeof(struct sctp_sndrcvinfo
);
4689 if (copy_from_user(&info
, optval
, len
))
4692 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
4693 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
4697 info
.sinfo_stream
= asoc
->default_stream
;
4698 info
.sinfo_flags
= asoc
->default_flags
;
4699 info
.sinfo_ppid
= asoc
->default_ppid
;
4700 info
.sinfo_context
= asoc
->default_context
;
4701 info
.sinfo_timetolive
= asoc
->default_timetolive
;
4703 info
.sinfo_stream
= sp
->default_stream
;
4704 info
.sinfo_flags
= sp
->default_flags
;
4705 info
.sinfo_ppid
= sp
->default_ppid
;
4706 info
.sinfo_context
= sp
->default_context
;
4707 info
.sinfo_timetolive
= sp
->default_timetolive
;
4710 if (put_user(len
, optlen
))
4712 if (copy_to_user(optval
, &info
, len
))
4720 * 7.1.5 SCTP_NODELAY
4722 * Turn on/off any Nagle-like algorithm. This means that packets are
4723 * generally sent as soon as possible and no unnecessary delays are
4724 * introduced, at the cost of more packets in the network. Expects an
4725 * integer boolean flag.
4728 static int sctp_getsockopt_nodelay(struct sock
*sk
, int len
,
4729 char __user
*optval
, int __user
*optlen
)
4733 if (len
< sizeof(int))
4737 val
= (sctp_sk(sk
)->nodelay
== 1);
4738 if (put_user(len
, optlen
))
4740 if (copy_to_user(optval
, &val
, len
))
4747 * 7.1.1 SCTP_RTOINFO
4749 * The protocol parameters used to initialize and bound retransmission
4750 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4751 * and modify these parameters.
4752 * All parameters are time values, in milliseconds. A value of 0, when
4753 * modifying the parameters, indicates that the current value should not
4757 static int sctp_getsockopt_rtoinfo(struct sock
*sk
, int len
,
4758 char __user
*optval
,
4759 int __user
*optlen
) {
4760 struct sctp_rtoinfo rtoinfo
;
4761 struct sctp_association
*asoc
;
4763 if (len
< sizeof (struct sctp_rtoinfo
))
4766 len
= sizeof(struct sctp_rtoinfo
);
4768 if (copy_from_user(&rtoinfo
, optval
, len
))
4771 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
4773 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
4776 /* Values corresponding to the specific association. */
4778 rtoinfo
.srto_initial
= jiffies_to_msecs(asoc
->rto_initial
);
4779 rtoinfo
.srto_max
= jiffies_to_msecs(asoc
->rto_max
);
4780 rtoinfo
.srto_min
= jiffies_to_msecs(asoc
->rto_min
);
4782 /* Values corresponding to the endpoint. */
4783 struct sctp_sock
*sp
= sctp_sk(sk
);
4785 rtoinfo
.srto_initial
= sp
->rtoinfo
.srto_initial
;
4786 rtoinfo
.srto_max
= sp
->rtoinfo
.srto_max
;
4787 rtoinfo
.srto_min
= sp
->rtoinfo
.srto_min
;
4790 if (put_user(len
, optlen
))
4793 if (copy_to_user(optval
, &rtoinfo
, len
))
4801 * 7.1.2 SCTP_ASSOCINFO
4803 * This option is used to tune the maximum retransmission attempts
4804 * of the association.
4805 * Returns an error if the new association retransmission value is
4806 * greater than the sum of the retransmission value of the peer.
4807 * See [SCTP] for more information.
4810 static int sctp_getsockopt_associnfo(struct sock
*sk
, int len
,
4811 char __user
*optval
,
4815 struct sctp_assocparams assocparams
;
4816 struct sctp_association
*asoc
;
4817 struct list_head
*pos
;
4820 if (len
< sizeof (struct sctp_assocparams
))
4823 len
= sizeof(struct sctp_assocparams
);
4825 if (copy_from_user(&assocparams
, optval
, len
))
4828 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
4830 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
4833 /* Values correspoinding to the specific association */
4835 assocparams
.sasoc_asocmaxrxt
= asoc
->max_retrans
;
4836 assocparams
.sasoc_peer_rwnd
= asoc
->peer
.rwnd
;
4837 assocparams
.sasoc_local_rwnd
= asoc
->a_rwnd
;
4838 assocparams
.sasoc_cookie_life
= (asoc
->cookie_life
.tv_sec
4840 (asoc
->cookie_life
.tv_usec
4843 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
4847 assocparams
.sasoc_number_peer_destinations
= cnt
;
4849 /* Values corresponding to the endpoint */
4850 struct sctp_sock
*sp
= sctp_sk(sk
);
4852 assocparams
.sasoc_asocmaxrxt
= sp
->assocparams
.sasoc_asocmaxrxt
;
4853 assocparams
.sasoc_peer_rwnd
= sp
->assocparams
.sasoc_peer_rwnd
;
4854 assocparams
.sasoc_local_rwnd
= sp
->assocparams
.sasoc_local_rwnd
;
4855 assocparams
.sasoc_cookie_life
=
4856 sp
->assocparams
.sasoc_cookie_life
;
4857 assocparams
.sasoc_number_peer_destinations
=
4859 sasoc_number_peer_destinations
;
4862 if (put_user(len
, optlen
))
4865 if (copy_to_user(optval
, &assocparams
, len
))
4872 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4874 * This socket option is a boolean flag which turns on or off mapped V4
4875 * addresses. If this option is turned on and the socket is type
4876 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4877 * If this option is turned off, then no mapping will be done of V4
4878 * addresses and a user will receive both PF_INET6 and PF_INET type
4879 * addresses on the socket.
4881 static int sctp_getsockopt_mappedv4(struct sock
*sk
, int len
,
4882 char __user
*optval
, int __user
*optlen
)
4885 struct sctp_sock
*sp
= sctp_sk(sk
);
4887 if (len
< sizeof(int))
4892 if (put_user(len
, optlen
))
4894 if (copy_to_user(optval
, &val
, len
))
4901 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
4902 * (chapter and verse is quoted at sctp_setsockopt_context())
4904 static int sctp_getsockopt_context(struct sock
*sk
, int len
,
4905 char __user
*optval
, int __user
*optlen
)
4907 struct sctp_assoc_value params
;
4908 struct sctp_sock
*sp
;
4909 struct sctp_association
*asoc
;
4911 if (len
< sizeof(struct sctp_assoc_value
))
4914 len
= sizeof(struct sctp_assoc_value
);
4916 if (copy_from_user(¶ms
, optval
, len
))
4921 if (params
.assoc_id
!= 0) {
4922 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
4925 params
.assoc_value
= asoc
->default_rcv_context
;
4927 params
.assoc_value
= sp
->default_rcv_context
;
4930 if (put_user(len
, optlen
))
4932 if (copy_to_user(optval
, ¶ms
, len
))
4939 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
4940 * This option will get or set the maximum size to put in any outgoing
4941 * SCTP DATA chunk. If a message is larger than this size it will be
4942 * fragmented by SCTP into the specified size. Note that the underlying
4943 * SCTP implementation may fragment into smaller sized chunks when the
4944 * PMTU of the underlying association is smaller than the value set by
4945 * the user. The default value for this option is '0' which indicates
4946 * the user is NOT limiting fragmentation and only the PMTU will effect
4947 * SCTP's choice of DATA chunk size. Note also that values set larger
4948 * than the maximum size of an IP datagram will effectively let SCTP
4949 * control fragmentation (i.e. the same as setting this option to 0).
4951 * The following structure is used to access and modify this parameter:
4953 * struct sctp_assoc_value {
4954 * sctp_assoc_t assoc_id;
4955 * uint32_t assoc_value;
4958 * assoc_id: This parameter is ignored for one-to-one style sockets.
4959 * For one-to-many style sockets this parameter indicates which
4960 * association the user is performing an action upon. Note that if
4961 * this field's value is zero then the endpoints default value is
4962 * changed (effecting future associations only).
4963 * assoc_value: This parameter specifies the maximum size in bytes.
4965 static int sctp_getsockopt_maxseg(struct sock
*sk
, int len
,
4966 char __user
*optval
, int __user
*optlen
)
4968 struct sctp_assoc_value params
;
4969 struct sctp_association
*asoc
;
4971 if (len
== sizeof(int)) {
4972 pr_warn("Use of int in maxseg socket option deprecated\n");
4973 pr_warn("Use struct sctp_assoc_value instead\n");
4974 params
.assoc_id
= 0;
4975 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
4976 len
= sizeof(struct sctp_assoc_value
);
4977 if (copy_from_user(¶ms
, optval
, sizeof(params
)))
4982 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
4983 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
4987 params
.assoc_value
= asoc
->frag_point
;
4989 params
.assoc_value
= sctp_sk(sk
)->user_frag
;
4991 if (put_user(len
, optlen
))
4993 if (len
== sizeof(int)) {
4994 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
4997 if (copy_to_user(optval
, ¶ms
, len
))
5005 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5006 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5008 static int sctp_getsockopt_fragment_interleave(struct sock
*sk
, int len
,
5009 char __user
*optval
, int __user
*optlen
)
5013 if (len
< sizeof(int))
5018 val
= sctp_sk(sk
)->frag_interleave
;
5019 if (put_user(len
, optlen
))
5021 if (copy_to_user(optval
, &val
, len
))
5028 * 7.1.25. Set or Get the sctp partial delivery point
5029 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5031 static int sctp_getsockopt_partial_delivery_point(struct sock
*sk
, int len
,
5032 char __user
*optval
,
5037 if (len
< sizeof(u32
))
5042 val
= sctp_sk(sk
)->pd_point
;
5043 if (put_user(len
, optlen
))
5045 if (copy_to_user(optval
, &val
, len
))
5052 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5053 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5055 static int sctp_getsockopt_maxburst(struct sock
*sk
, int len
,
5056 char __user
*optval
,
5059 struct sctp_assoc_value params
;
5060 struct sctp_sock
*sp
;
5061 struct sctp_association
*asoc
;
5063 if (len
== sizeof(int)) {
5064 pr_warn("Use of int in max_burst socket option deprecated\n");
5065 pr_warn("Use struct sctp_assoc_value instead\n");
5066 params
.assoc_id
= 0;
5067 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5068 len
= sizeof(struct sctp_assoc_value
);
5069 if (copy_from_user(¶ms
, optval
, len
))
5076 if (params
.assoc_id
!= 0) {
5077 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5080 params
.assoc_value
= asoc
->max_burst
;
5082 params
.assoc_value
= sp
->max_burst
;
5084 if (len
== sizeof(int)) {
5085 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5088 if (copy_to_user(optval
, ¶ms
, len
))
5096 static int sctp_getsockopt_hmac_ident(struct sock
*sk
, int len
,
5097 char __user
*optval
, int __user
*optlen
)
5099 struct sctp_hmacalgo __user
*p
= (void __user
*)optval
;
5100 struct sctp_hmac_algo_param
*hmacs
;
5104 if (!sctp_auth_enable
)
5107 hmacs
= sctp_sk(sk
)->ep
->auth_hmacs_list
;
5108 data_len
= ntohs(hmacs
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5110 if (len
< sizeof(struct sctp_hmacalgo
) + data_len
)
5113 len
= sizeof(struct sctp_hmacalgo
) + data_len
;
5114 num_idents
= data_len
/ sizeof(u16
);
5116 if (put_user(len
, optlen
))
5118 if (put_user(num_idents
, &p
->shmac_num_idents
))
5120 if (copy_to_user(p
->shmac_idents
, hmacs
->hmac_ids
, data_len
))
5125 static int sctp_getsockopt_active_key(struct sock
*sk
, int len
,
5126 char __user
*optval
, int __user
*optlen
)
5128 struct sctp_authkeyid val
;
5129 struct sctp_association
*asoc
;
5131 if (!sctp_auth_enable
)
5134 if (len
< sizeof(struct sctp_authkeyid
))
5136 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authkeyid
)))
5139 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
5140 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
5144 val
.scact_keynumber
= asoc
->active_key_id
;
5146 val
.scact_keynumber
= sctp_sk(sk
)->ep
->active_key_id
;
5148 len
= sizeof(struct sctp_authkeyid
);
5149 if (put_user(len
, optlen
))
5151 if (copy_to_user(optval
, &val
, len
))
5157 static int sctp_getsockopt_peer_auth_chunks(struct sock
*sk
, int len
,
5158 char __user
*optval
, int __user
*optlen
)
5160 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5161 struct sctp_authchunks val
;
5162 struct sctp_association
*asoc
;
5163 struct sctp_chunks_param
*ch
;
5167 if (!sctp_auth_enable
)
5170 if (len
< sizeof(struct sctp_authchunks
))
5173 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5176 to
= p
->gauth_chunks
;
5177 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5181 ch
= asoc
->peer
.peer_chunks
;
5185 /* See if the user provided enough room for all the data */
5186 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5187 if (len
< num_chunks
)
5190 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
5193 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
5194 if (put_user(len
, optlen
)) return -EFAULT
;
5195 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
5200 static int sctp_getsockopt_local_auth_chunks(struct sock
*sk
, int len
,
5201 char __user
*optval
, int __user
*optlen
)
5203 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5204 struct sctp_authchunks val
;
5205 struct sctp_association
*asoc
;
5206 struct sctp_chunks_param
*ch
;
5210 if (!sctp_auth_enable
)
5213 if (len
< sizeof(struct sctp_authchunks
))
5216 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5219 to
= p
->gauth_chunks
;
5220 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5221 if (!asoc
&& val
.gauth_assoc_id
&& sctp_style(sk
, UDP
))
5225 ch
= (struct sctp_chunks_param
*)asoc
->c
.auth_chunks
;
5227 ch
= sctp_sk(sk
)->ep
->auth_chunk_list
;
5232 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5233 if (len
< sizeof(struct sctp_authchunks
) + num_chunks
)
5236 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
5239 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
5240 if (put_user(len
, optlen
))
5242 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
5249 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5250 * This option gets the current number of associations that are attached
5251 * to a one-to-many style socket. The option value is an uint32_t.
5253 static int sctp_getsockopt_assoc_number(struct sock
*sk
, int len
,
5254 char __user
*optval
, int __user
*optlen
)
5256 struct sctp_sock
*sp
= sctp_sk(sk
);
5257 struct sctp_association
*asoc
;
5260 if (sctp_style(sk
, TCP
))
5263 if (len
< sizeof(u32
))
5268 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5272 if (put_user(len
, optlen
))
5274 if (copy_to_user(optval
, &val
, len
))
5280 SCTP_STATIC
int sctp_getsockopt(struct sock
*sk
, int level
, int optname
,
5281 char __user
*optval
, int __user
*optlen
)
5286 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5289 /* I can hardly begin to describe how wrong this is. This is
5290 * so broken as to be worse than useless. The API draft
5291 * REALLY is NOT helpful here... I am not convinced that the
5292 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5293 * are at all well-founded.
5295 if (level
!= SOL_SCTP
) {
5296 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
5298 retval
= af
->getsockopt(sk
, level
, optname
, optval
, optlen
);
5302 if (get_user(len
, optlen
))
5309 retval
= sctp_getsockopt_sctp_status(sk
, len
, optval
, optlen
);
5311 case SCTP_DISABLE_FRAGMENTS
:
5312 retval
= sctp_getsockopt_disable_fragments(sk
, len
, optval
,
5316 retval
= sctp_getsockopt_events(sk
, len
, optval
, optlen
);
5318 case SCTP_AUTOCLOSE
:
5319 retval
= sctp_getsockopt_autoclose(sk
, len
, optval
, optlen
);
5321 case SCTP_SOCKOPT_PEELOFF
:
5322 retval
= sctp_getsockopt_peeloff(sk
, len
, optval
, optlen
);
5324 case SCTP_PEER_ADDR_PARAMS
:
5325 retval
= sctp_getsockopt_peer_addr_params(sk
, len
, optval
,
5328 case SCTP_DELAYED_ACK
:
5329 retval
= sctp_getsockopt_delayed_ack(sk
, len
, optval
,
5333 retval
= sctp_getsockopt_initmsg(sk
, len
, optval
, optlen
);
5335 case SCTP_GET_PEER_ADDRS
:
5336 retval
= sctp_getsockopt_peer_addrs(sk
, len
, optval
,
5339 case SCTP_GET_LOCAL_ADDRS
:
5340 retval
= sctp_getsockopt_local_addrs(sk
, len
, optval
,
5343 case SCTP_SOCKOPT_CONNECTX3
:
5344 retval
= sctp_getsockopt_connectx3(sk
, len
, optval
, optlen
);
5346 case SCTP_DEFAULT_SEND_PARAM
:
5347 retval
= sctp_getsockopt_default_send_param(sk
, len
,
5350 case SCTP_PRIMARY_ADDR
:
5351 retval
= sctp_getsockopt_primary_addr(sk
, len
, optval
, optlen
);
5354 retval
= sctp_getsockopt_nodelay(sk
, len
, optval
, optlen
);
5357 retval
= sctp_getsockopt_rtoinfo(sk
, len
, optval
, optlen
);
5359 case SCTP_ASSOCINFO
:
5360 retval
= sctp_getsockopt_associnfo(sk
, len
, optval
, optlen
);
5362 case SCTP_I_WANT_MAPPED_V4_ADDR
:
5363 retval
= sctp_getsockopt_mappedv4(sk
, len
, optval
, optlen
);
5366 retval
= sctp_getsockopt_maxseg(sk
, len
, optval
, optlen
);
5368 case SCTP_GET_PEER_ADDR_INFO
:
5369 retval
= sctp_getsockopt_peer_addr_info(sk
, len
, optval
,
5372 case SCTP_ADAPTATION_LAYER
:
5373 retval
= sctp_getsockopt_adaptation_layer(sk
, len
, optval
,
5377 retval
= sctp_getsockopt_context(sk
, len
, optval
, optlen
);
5379 case SCTP_FRAGMENT_INTERLEAVE
:
5380 retval
= sctp_getsockopt_fragment_interleave(sk
, len
, optval
,
5383 case SCTP_PARTIAL_DELIVERY_POINT
:
5384 retval
= sctp_getsockopt_partial_delivery_point(sk
, len
, optval
,
5387 case SCTP_MAX_BURST
:
5388 retval
= sctp_getsockopt_maxburst(sk
, len
, optval
, optlen
);
5391 case SCTP_AUTH_CHUNK
:
5392 case SCTP_AUTH_DELETE_KEY
:
5393 retval
= -EOPNOTSUPP
;
5395 case SCTP_HMAC_IDENT
:
5396 retval
= sctp_getsockopt_hmac_ident(sk
, len
, optval
, optlen
);
5398 case SCTP_AUTH_ACTIVE_KEY
:
5399 retval
= sctp_getsockopt_active_key(sk
, len
, optval
, optlen
);
5401 case SCTP_PEER_AUTH_CHUNKS
:
5402 retval
= sctp_getsockopt_peer_auth_chunks(sk
, len
, optval
,
5405 case SCTP_LOCAL_AUTH_CHUNKS
:
5406 retval
= sctp_getsockopt_local_auth_chunks(sk
, len
, optval
,
5409 case SCTP_GET_ASSOC_NUMBER
:
5410 retval
= sctp_getsockopt_assoc_number(sk
, len
, optval
, optlen
);
5413 retval
= -ENOPROTOOPT
;
5417 sctp_release_sock(sk
);
5421 static void sctp_hash(struct sock
*sk
)
5426 static void sctp_unhash(struct sock
*sk
)
5431 /* Check if port is acceptable. Possibly find first available port.
5433 * The port hash table (contained in the 'global' SCTP protocol storage
5434 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5435 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5436 * list (the list number is the port number hashed out, so as you
5437 * would expect from a hash function, all the ports in a given list have
5438 * such a number that hashes out to the same list number; you were
5439 * expecting that, right?); so each list has a set of ports, with a
5440 * link to the socket (struct sock) that uses it, the port number and
5441 * a fastreuse flag (FIXME: NPI ipg).
5443 static struct sctp_bind_bucket
*sctp_bucket_create(
5444 struct sctp_bind_hashbucket
*head
, unsigned short snum
);
5446 static long sctp_get_port_local(struct sock
*sk
, union sctp_addr
*addr
)
5448 struct sctp_bind_hashbucket
*head
; /* hash list */
5449 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
5450 struct hlist_node
*node
;
5451 unsigned short snum
;
5454 snum
= ntohs(addr
->v4
.sin_port
);
5456 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum
);
5457 sctp_local_bh_disable();
5460 /* Search for an available port. */
5461 int low
, high
, remaining
, index
;
5464 inet_get_local_port_range(&low
, &high
);
5465 remaining
= (high
- low
) + 1;
5466 rover
= net_random() % remaining
+ low
;
5470 if ((rover
< low
) || (rover
> high
))
5472 if (inet_is_reserved_local_port(rover
))
5474 index
= sctp_phashfn(rover
);
5475 head
= &sctp_port_hashtable
[index
];
5476 sctp_spin_lock(&head
->lock
);
5477 sctp_for_each_hentry(pp
, node
, &head
->chain
)
5478 if (pp
->port
== rover
)
5482 sctp_spin_unlock(&head
->lock
);
5483 } while (--remaining
> 0);
5485 /* Exhausted local port range during search? */
5490 /* OK, here is the one we will use. HEAD (the port
5491 * hash table list entry) is non-NULL and we hold it's
5496 /* We are given an specific port number; we verify
5497 * that it is not being used. If it is used, we will
5498 * exahust the search in the hash list corresponding
5499 * to the port number (snum) - we detect that with the
5500 * port iterator, pp being NULL.
5502 head
= &sctp_port_hashtable
[sctp_phashfn(snum
)];
5503 sctp_spin_lock(&head
->lock
);
5504 sctp_for_each_hentry(pp
, node
, &head
->chain
) {
5505 if (pp
->port
== snum
)
5512 if (!hlist_empty(&pp
->owner
)) {
5513 /* We had a port hash table hit - there is an
5514 * available port (pp != NULL) and it is being
5515 * used by other socket (pp->owner not empty); that other
5516 * socket is going to be sk2.
5518 int reuse
= sk
->sk_reuse
;
5521 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5522 if (pp
->fastreuse
&& sk
->sk_reuse
&&
5523 sk
->sk_state
!= SCTP_SS_LISTENING
)
5526 /* Run through the list of sockets bound to the port
5527 * (pp->port) [via the pointers bind_next and
5528 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5529 * we get the endpoint they describe and run through
5530 * the endpoint's list of IP (v4 or v6) addresses,
5531 * comparing each of the addresses with the address of
5532 * the socket sk. If we find a match, then that means
5533 * that this port/socket (sk) combination are already
5536 sk_for_each_bound(sk2
, node
, &pp
->owner
) {
5537 struct sctp_endpoint
*ep2
;
5538 ep2
= sctp_sk(sk2
)->ep
;
5541 (reuse
&& sk2
->sk_reuse
&&
5542 sk2
->sk_state
!= SCTP_SS_LISTENING
))
5545 if (sctp_bind_addr_conflict(&ep2
->base
.bind_addr
, addr
,
5546 sctp_sk(sk2
), sctp_sk(sk
))) {
5551 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5554 /* If there was a hash table miss, create a new port. */
5556 if (!pp
&& !(pp
= sctp_bucket_create(head
, snum
)))
5559 /* In either case (hit or miss), make sure fastreuse is 1 only
5560 * if sk->sk_reuse is too (that is, if the caller requested
5561 * SO_REUSEADDR on this socket -sk-).
5563 if (hlist_empty(&pp
->owner
)) {
5564 if (sk
->sk_reuse
&& sk
->sk_state
!= SCTP_SS_LISTENING
)
5568 } else if (pp
->fastreuse
&&
5569 (!sk
->sk_reuse
|| sk
->sk_state
== SCTP_SS_LISTENING
))
5572 /* We are set, so fill up all the data in the hash table
5573 * entry, tie the socket list information with the rest of the
5574 * sockets FIXME: Blurry, NPI (ipg).
5577 if (!sctp_sk(sk
)->bind_hash
) {
5578 inet_sk(sk
)->inet_num
= snum
;
5579 sk_add_bind_node(sk
, &pp
->owner
);
5580 sctp_sk(sk
)->bind_hash
= pp
;
5585 sctp_spin_unlock(&head
->lock
);
5588 sctp_local_bh_enable();
5592 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5593 * port is requested.
5595 static int sctp_get_port(struct sock
*sk
, unsigned short snum
)
5598 union sctp_addr addr
;
5599 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
5601 /* Set up a dummy address struct from the sk. */
5602 af
->from_sk(&addr
, sk
);
5603 addr
.v4
.sin_port
= htons(snum
);
5605 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5606 ret
= sctp_get_port_local(sk
, &addr
);
5612 * Move a socket to LISTENING state.
5614 SCTP_STATIC
int sctp_listen_start(struct sock
*sk
, int backlog
)
5616 struct sctp_sock
*sp
= sctp_sk(sk
);
5617 struct sctp_endpoint
*ep
= sp
->ep
;
5618 struct crypto_hash
*tfm
= NULL
;
5620 /* Allocate HMAC for generating cookie. */
5621 if (!sctp_sk(sk
)->hmac
&& sctp_hmac_alg
) {
5622 tfm
= crypto_alloc_hash(sctp_hmac_alg
, 0, CRYPTO_ALG_ASYNC
);
5624 if (net_ratelimit()) {
5625 pr_info("failed to load transform for %s: %ld\n",
5626 sctp_hmac_alg
, PTR_ERR(tfm
));
5630 sctp_sk(sk
)->hmac
= tfm
;
5634 * If a bind() or sctp_bindx() is not called prior to a listen()
5635 * call that allows new associations to be accepted, the system
5636 * picks an ephemeral port and will choose an address set equivalent
5637 * to binding with a wildcard address.
5639 * This is not currently spelled out in the SCTP sockets
5640 * extensions draft, but follows the practice as seen in TCP
5644 sk
->sk_state
= SCTP_SS_LISTENING
;
5645 if (!ep
->base
.bind_addr
.port
) {
5646 if (sctp_autobind(sk
))
5649 if (sctp_get_port(sk
, inet_sk(sk
)->inet_num
)) {
5650 sk
->sk_state
= SCTP_SS_CLOSED
;
5655 sk
->sk_max_ack_backlog
= backlog
;
5656 sctp_hash_endpoint(ep
);
5661 * 4.1.3 / 5.1.3 listen()
5663 * By default, new associations are not accepted for UDP style sockets.
5664 * An application uses listen() to mark a socket as being able to
5665 * accept new associations.
5667 * On TCP style sockets, applications use listen() to ready the SCTP
5668 * endpoint for accepting inbound associations.
5670 * On both types of endpoints a backlog of '0' disables listening.
5672 * Move a socket to LISTENING state.
5674 int sctp_inet_listen(struct socket
*sock
, int backlog
)
5676 struct sock
*sk
= sock
->sk
;
5677 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5680 if (unlikely(backlog
< 0))
5685 /* Peeled-off sockets are not allowed to listen(). */
5686 if (sctp_style(sk
, UDP_HIGH_BANDWIDTH
))
5689 if (sock
->state
!= SS_UNCONNECTED
)
5692 /* If backlog is zero, disable listening. */
5694 if (sctp_sstate(sk
, CLOSED
))
5698 sctp_unhash_endpoint(ep
);
5699 sk
->sk_state
= SCTP_SS_CLOSED
;
5701 sctp_sk(sk
)->bind_hash
->fastreuse
= 1;
5705 /* If we are already listening, just update the backlog */
5706 if (sctp_sstate(sk
, LISTENING
))
5707 sk
->sk_max_ack_backlog
= backlog
;
5709 err
= sctp_listen_start(sk
, backlog
);
5716 sctp_release_sock(sk
);
5721 * This function is done by modeling the current datagram_poll() and the
5722 * tcp_poll(). Note that, based on these implementations, we don't
5723 * lock the socket in this function, even though it seems that,
5724 * ideally, locking or some other mechanisms can be used to ensure
5725 * the integrity of the counters (sndbuf and wmem_alloc) used
5726 * in this place. We assume that we don't need locks either until proven
5729 * Another thing to note is that we include the Async I/O support
5730 * here, again, by modeling the current TCP/UDP code. We don't have
5731 * a good way to test with it yet.
5733 unsigned int sctp_poll(struct file
*file
, struct socket
*sock
, poll_table
*wait
)
5735 struct sock
*sk
= sock
->sk
;
5736 struct sctp_sock
*sp
= sctp_sk(sk
);
5739 poll_wait(file
, sk_sleep(sk
), wait
);
5741 /* A TCP-style listening socket becomes readable when the accept queue
5744 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
5745 return (!list_empty(&sp
->ep
->asocs
)) ?
5746 (POLLIN
| POLLRDNORM
) : 0;
5750 /* Is there any exceptional events? */
5751 if (sk
->sk_err
|| !skb_queue_empty(&sk
->sk_error_queue
))
5753 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
5754 mask
|= POLLRDHUP
| POLLIN
| POLLRDNORM
;
5755 if (sk
->sk_shutdown
== SHUTDOWN_MASK
)
5758 /* Is it readable? Reconsider this code with TCP-style support. */
5759 if (!skb_queue_empty(&sk
->sk_receive_queue
))
5760 mask
|= POLLIN
| POLLRDNORM
;
5762 /* The association is either gone or not ready. */
5763 if (!sctp_style(sk
, UDP
) && sctp_sstate(sk
, CLOSED
))
5766 /* Is it writable? */
5767 if (sctp_writeable(sk
)) {
5768 mask
|= POLLOUT
| POLLWRNORM
;
5770 set_bit(SOCK_ASYNC_NOSPACE
, &sk
->sk_socket
->flags
);
5772 * Since the socket is not locked, the buffer
5773 * might be made available after the writeable check and
5774 * before the bit is set. This could cause a lost I/O
5775 * signal. tcp_poll() has a race breaker for this race
5776 * condition. Based on their implementation, we put
5777 * in the following code to cover it as well.
5779 if (sctp_writeable(sk
))
5780 mask
|= POLLOUT
| POLLWRNORM
;
5785 /********************************************************************
5786 * 2nd Level Abstractions
5787 ********************************************************************/
5789 static struct sctp_bind_bucket
*sctp_bucket_create(
5790 struct sctp_bind_hashbucket
*head
, unsigned short snum
)
5792 struct sctp_bind_bucket
*pp
;
5794 pp
= kmem_cache_alloc(sctp_bucket_cachep
, GFP_ATOMIC
);
5796 SCTP_DBG_OBJCNT_INC(bind_bucket
);
5799 INIT_HLIST_HEAD(&pp
->owner
);
5800 hlist_add_head(&pp
->node
, &head
->chain
);
5805 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5806 static void sctp_bucket_destroy(struct sctp_bind_bucket
*pp
)
5808 if (pp
&& hlist_empty(&pp
->owner
)) {
5809 __hlist_del(&pp
->node
);
5810 kmem_cache_free(sctp_bucket_cachep
, pp
);
5811 SCTP_DBG_OBJCNT_DEC(bind_bucket
);
5815 /* Release this socket's reference to a local port. */
5816 static inline void __sctp_put_port(struct sock
*sk
)
5818 struct sctp_bind_hashbucket
*head
=
5819 &sctp_port_hashtable
[sctp_phashfn(inet_sk(sk
)->inet_num
)];
5820 struct sctp_bind_bucket
*pp
;
5822 sctp_spin_lock(&head
->lock
);
5823 pp
= sctp_sk(sk
)->bind_hash
;
5824 __sk_del_bind_node(sk
);
5825 sctp_sk(sk
)->bind_hash
= NULL
;
5826 inet_sk(sk
)->inet_num
= 0;
5827 sctp_bucket_destroy(pp
);
5828 sctp_spin_unlock(&head
->lock
);
5831 void sctp_put_port(struct sock
*sk
)
5833 sctp_local_bh_disable();
5834 __sctp_put_port(sk
);
5835 sctp_local_bh_enable();
5839 * The system picks an ephemeral port and choose an address set equivalent
5840 * to binding with a wildcard address.
5841 * One of those addresses will be the primary address for the association.
5842 * This automatically enables the multihoming capability of SCTP.
5844 static int sctp_autobind(struct sock
*sk
)
5846 union sctp_addr autoaddr
;
5850 /* Initialize a local sockaddr structure to INADDR_ANY. */
5851 af
= sctp_sk(sk
)->pf
->af
;
5853 port
= htons(inet_sk(sk
)->inet_num
);
5854 af
->inaddr_any(&autoaddr
, port
);
5856 return sctp_do_bind(sk
, &autoaddr
, af
->sockaddr_len
);
5859 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5862 * 4.2 The cmsghdr Structure *
5864 * When ancillary data is sent or received, any number of ancillary data
5865 * objects can be specified by the msg_control and msg_controllen members of
5866 * the msghdr structure, because each object is preceded by
5867 * a cmsghdr structure defining the object's length (the cmsg_len member).
5868 * Historically Berkeley-derived implementations have passed only one object
5869 * at a time, but this API allows multiple objects to be
5870 * passed in a single call to sendmsg() or recvmsg(). The following example
5871 * shows two ancillary data objects in a control buffer.
5873 * |<--------------------------- msg_controllen -------------------------->|
5876 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5878 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5881 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5883 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5886 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5887 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5889 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5891 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5898 SCTP_STATIC
int sctp_msghdr_parse(const struct msghdr
*msg
,
5899 sctp_cmsgs_t
*cmsgs
)
5901 struct cmsghdr
*cmsg
;
5902 struct msghdr
*my_msg
= (struct msghdr
*)msg
;
5904 for (cmsg
= CMSG_FIRSTHDR(msg
);
5906 cmsg
= CMSG_NXTHDR(my_msg
, cmsg
)) {
5907 if (!CMSG_OK(my_msg
, cmsg
))
5910 /* Should we parse this header or ignore? */
5911 if (cmsg
->cmsg_level
!= IPPROTO_SCTP
)
5914 /* Strictly check lengths following example in SCM code. */
5915 switch (cmsg
->cmsg_type
) {
5917 /* SCTP Socket API Extension
5918 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5920 * This cmsghdr structure provides information for
5921 * initializing new SCTP associations with sendmsg().
5922 * The SCTP_INITMSG socket option uses this same data
5923 * structure. This structure is not used for
5926 * cmsg_level cmsg_type cmsg_data[]
5927 * ------------ ------------ ----------------------
5928 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5930 if (cmsg
->cmsg_len
!=
5931 CMSG_LEN(sizeof(struct sctp_initmsg
)))
5933 cmsgs
->init
= (struct sctp_initmsg
*)CMSG_DATA(cmsg
);
5937 /* SCTP Socket API Extension
5938 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5940 * This cmsghdr structure specifies SCTP options for
5941 * sendmsg() and describes SCTP header information
5942 * about a received message through recvmsg().
5944 * cmsg_level cmsg_type cmsg_data[]
5945 * ------------ ------------ ----------------------
5946 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5948 if (cmsg
->cmsg_len
!=
5949 CMSG_LEN(sizeof(struct sctp_sndrcvinfo
)))
5953 (struct sctp_sndrcvinfo
*)CMSG_DATA(cmsg
);
5955 /* Minimally, validate the sinfo_flags. */
5956 if (cmsgs
->info
->sinfo_flags
&
5957 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
5958 SCTP_ABORT
| SCTP_EOF
))
5970 * Wait for a packet..
5971 * Note: This function is the same function as in core/datagram.c
5972 * with a few modifications to make lksctp work.
5974 static int sctp_wait_for_packet(struct sock
* sk
, int *err
, long *timeo_p
)
5979 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
5981 /* Socket errors? */
5982 error
= sock_error(sk
);
5986 if (!skb_queue_empty(&sk
->sk_receive_queue
))
5989 /* Socket shut down? */
5990 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
5993 /* Sequenced packets can come disconnected. If so we report the
5998 /* Is there a good reason to think that we may receive some data? */
5999 if (list_empty(&sctp_sk(sk
)->ep
->asocs
) && !sctp_sstate(sk
, LISTENING
))
6002 /* Handle signals. */
6003 if (signal_pending(current
))
6006 /* Let another process have a go. Since we are going to sleep
6007 * anyway. Note: This may cause odd behaviors if the message
6008 * does not fit in the user's buffer, but this seems to be the
6009 * only way to honor MSG_DONTWAIT realistically.
6011 sctp_release_sock(sk
);
6012 *timeo_p
= schedule_timeout(*timeo_p
);
6016 finish_wait(sk_sleep(sk
), &wait
);
6020 error
= sock_intr_errno(*timeo_p
);
6023 finish_wait(sk_sleep(sk
), &wait
);
6028 /* Receive a datagram.
6029 * Note: This is pretty much the same routine as in core/datagram.c
6030 * with a few changes to make lksctp work.
6032 static struct sk_buff
*sctp_skb_recv_datagram(struct sock
*sk
, int flags
,
6033 int noblock
, int *err
)
6036 struct sk_buff
*skb
;
6039 timeo
= sock_rcvtimeo(sk
, noblock
);
6041 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6042 timeo
, MAX_SCHEDULE_TIMEOUT
);
6045 /* Again only user level code calls this function,
6046 * so nothing interrupt level
6047 * will suddenly eat the receive_queue.
6049 * Look at current nfs client by the way...
6050 * However, this function was corrent in any case. 8)
6052 if (flags
& MSG_PEEK
) {
6053 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
6054 skb
= skb_peek(&sk
->sk_receive_queue
);
6056 atomic_inc(&skb
->users
);
6057 spin_unlock_bh(&sk
->sk_receive_queue
.lock
);
6059 skb
= skb_dequeue(&sk
->sk_receive_queue
);
6065 /* Caller is allowed not to check sk->sk_err before calling. */
6066 error
= sock_error(sk
);
6070 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6073 /* User doesn't want to wait. */
6077 } while (sctp_wait_for_packet(sk
, err
, &timeo
) == 0);
6086 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6087 static void __sctp_write_space(struct sctp_association
*asoc
)
6089 struct sock
*sk
= asoc
->base
.sk
;
6090 struct socket
*sock
= sk
->sk_socket
;
6092 if ((sctp_wspace(asoc
) > 0) && sock
) {
6093 if (waitqueue_active(&asoc
->wait
))
6094 wake_up_interruptible(&asoc
->wait
);
6096 if (sctp_writeable(sk
)) {
6097 if (sk_sleep(sk
) && waitqueue_active(sk_sleep(sk
)))
6098 wake_up_interruptible(sk_sleep(sk
));
6100 /* Note that we try to include the Async I/O support
6101 * here by modeling from the current TCP/UDP code.
6102 * We have not tested with it yet.
6104 if (sock
->wq
->fasync_list
&&
6105 !(sk
->sk_shutdown
& SEND_SHUTDOWN
))
6106 sock_wake_async(sock
,
6107 SOCK_WAKE_SPACE
, POLL_OUT
);
6112 /* Do accounting for the sndbuf space.
6113 * Decrement the used sndbuf space of the corresponding association by the
6114 * data size which was just transmitted(freed).
6116 static void sctp_wfree(struct sk_buff
*skb
)
6118 struct sctp_association
*asoc
;
6119 struct sctp_chunk
*chunk
;
6122 /* Get the saved chunk pointer. */
6123 chunk
= *((struct sctp_chunk
**)(skb
->cb
));
6126 asoc
->sndbuf_used
-= SCTP_DATA_SNDSIZE(chunk
) +
6127 sizeof(struct sk_buff
) +
6128 sizeof(struct sctp_chunk
);
6130 atomic_sub(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
6133 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6135 sk
->sk_wmem_queued
-= skb
->truesize
;
6136 sk_mem_uncharge(sk
, skb
->truesize
);
6139 __sctp_write_space(asoc
);
6141 sctp_association_put(asoc
);
6144 /* Do accounting for the receive space on the socket.
6145 * Accounting for the association is done in ulpevent.c
6146 * We set this as a destructor for the cloned data skbs so that
6147 * accounting is done at the correct time.
6149 void sctp_sock_rfree(struct sk_buff
*skb
)
6151 struct sock
*sk
= skb
->sk
;
6152 struct sctp_ulpevent
*event
= sctp_skb2event(skb
);
6154 atomic_sub(event
->rmem_len
, &sk
->sk_rmem_alloc
);
6157 * Mimic the behavior of sock_rfree
6159 sk_mem_uncharge(sk
, event
->rmem_len
);
6163 /* Helper function to wait for space in the sndbuf. */
6164 static int sctp_wait_for_sndbuf(struct sctp_association
*asoc
, long *timeo_p
,
6167 struct sock
*sk
= asoc
->base
.sk
;
6169 long current_timeo
= *timeo_p
;
6172 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6173 asoc
, (long)(*timeo_p
), msg_len
);
6175 /* Increment the association's refcnt. */
6176 sctp_association_hold(asoc
);
6178 /* Wait on the association specific sndbuf space. */
6180 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
6181 TASK_INTERRUPTIBLE
);
6184 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
6187 if (signal_pending(current
))
6188 goto do_interrupted
;
6189 if (msg_len
<= sctp_wspace(asoc
))
6192 /* Let another process have a go. Since we are going
6195 sctp_release_sock(sk
);
6196 current_timeo
= schedule_timeout(current_timeo
);
6197 BUG_ON(sk
!= asoc
->base
.sk
);
6200 *timeo_p
= current_timeo
;
6204 finish_wait(&asoc
->wait
, &wait
);
6206 /* Release the association's refcnt. */
6207 sctp_association_put(asoc
);
6216 err
= sock_intr_errno(*timeo_p
);
6224 void sctp_data_ready(struct sock
*sk
, int len
)
6226 struct socket_wq
*wq
;
6229 wq
= rcu_dereference(sk
->sk_wq
);
6230 if (wq_has_sleeper(wq
))
6231 wake_up_interruptible_sync_poll(&wq
->wait
, POLLIN
|
6232 POLLRDNORM
| POLLRDBAND
);
6233 sk_wake_async(sk
, SOCK_WAKE_WAITD
, POLL_IN
);
6237 /* If socket sndbuf has changed, wake up all per association waiters. */
6238 void sctp_write_space(struct sock
*sk
)
6240 struct sctp_association
*asoc
;
6242 /* Wake up the tasks in each wait queue. */
6243 list_for_each_entry(asoc
, &((sctp_sk(sk
))->ep
->asocs
), asocs
) {
6244 __sctp_write_space(asoc
);
6248 /* Is there any sndbuf space available on the socket?
6250 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6251 * associations on the same socket. For a UDP-style socket with
6252 * multiple associations, it is possible for it to be "unwriteable"
6253 * prematurely. I assume that this is acceptable because
6254 * a premature "unwriteable" is better than an accidental "writeable" which
6255 * would cause an unwanted block under certain circumstances. For the 1-1
6256 * UDP-style sockets or TCP-style sockets, this code should work.
6259 static int sctp_writeable(struct sock
*sk
)
6263 amt
= sk
->sk_sndbuf
- sk_wmem_alloc_get(sk
);
6269 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6270 * returns immediately with EINPROGRESS.
6272 static int sctp_wait_for_connect(struct sctp_association
*asoc
, long *timeo_p
)
6274 struct sock
*sk
= asoc
->base
.sk
;
6276 long current_timeo
= *timeo_p
;
6279 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__
, asoc
,
6282 /* Increment the association's refcnt. */
6283 sctp_association_hold(asoc
);
6286 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
6287 TASK_INTERRUPTIBLE
);
6290 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6292 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
6295 if (signal_pending(current
))
6296 goto do_interrupted
;
6298 if (sctp_state(asoc
, ESTABLISHED
))
6301 /* Let another process have a go. Since we are going
6304 sctp_release_sock(sk
);
6305 current_timeo
= schedule_timeout(current_timeo
);
6308 *timeo_p
= current_timeo
;
6312 finish_wait(&asoc
->wait
, &wait
);
6314 /* Release the association's refcnt. */
6315 sctp_association_put(asoc
);
6320 if (asoc
->init_err_counter
+ 1 > asoc
->max_init_attempts
)
6323 err
= -ECONNREFUSED
;
6327 err
= sock_intr_errno(*timeo_p
);
6335 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
)
6337 struct sctp_endpoint
*ep
;
6341 ep
= sctp_sk(sk
)->ep
;
6345 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
,
6346 TASK_INTERRUPTIBLE
);
6348 if (list_empty(&ep
->asocs
)) {
6349 sctp_release_sock(sk
);
6350 timeo
= schedule_timeout(timeo
);
6355 if (!sctp_sstate(sk
, LISTENING
))
6359 if (!list_empty(&ep
->asocs
))
6362 err
= sock_intr_errno(timeo
);
6363 if (signal_pending(current
))
6371 finish_wait(sk_sleep(sk
), &wait
);
6376 static void sctp_wait_for_close(struct sock
*sk
, long timeout
)
6381 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
6382 if (list_empty(&sctp_sk(sk
)->ep
->asocs
))
6384 sctp_release_sock(sk
);
6385 timeout
= schedule_timeout(timeout
);
6387 } while (!signal_pending(current
) && timeout
);
6389 finish_wait(sk_sleep(sk
), &wait
);
6392 static void sctp_skb_set_owner_r_frag(struct sk_buff
*skb
, struct sock
*sk
)
6394 struct sk_buff
*frag
;
6399 /* Don't forget the fragments. */
6400 skb_walk_frags(skb
, frag
)
6401 sctp_skb_set_owner_r_frag(frag
, sk
);
6404 sctp_skb_set_owner_r(skb
, sk
);
6407 void sctp_copy_sock(struct sock
*newsk
, struct sock
*sk
,
6408 struct sctp_association
*asoc
)
6410 struct inet_sock
*inet
= inet_sk(sk
);
6411 struct inet_sock
*newinet
;
6413 newsk
->sk_type
= sk
->sk_type
;
6414 newsk
->sk_bound_dev_if
= sk
->sk_bound_dev_if
;
6415 newsk
->sk_flags
= sk
->sk_flags
;
6416 newsk
->sk_no_check
= sk
->sk_no_check
;
6417 newsk
->sk_reuse
= sk
->sk_reuse
;
6419 newsk
->sk_shutdown
= sk
->sk_shutdown
;
6420 newsk
->sk_destruct
= inet_sock_destruct
;
6421 newsk
->sk_family
= sk
->sk_family
;
6422 newsk
->sk_protocol
= IPPROTO_SCTP
;
6423 newsk
->sk_backlog_rcv
= sk
->sk_prot
->backlog_rcv
;
6424 newsk
->sk_sndbuf
= sk
->sk_sndbuf
;
6425 newsk
->sk_rcvbuf
= sk
->sk_rcvbuf
;
6426 newsk
->sk_lingertime
= sk
->sk_lingertime
;
6427 newsk
->sk_rcvtimeo
= sk
->sk_rcvtimeo
;
6428 newsk
->sk_sndtimeo
= sk
->sk_sndtimeo
;
6430 newinet
= inet_sk(newsk
);
6432 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6433 * getsockname() and getpeername()
6435 newinet
->inet_sport
= inet
->inet_sport
;
6436 newinet
->inet_saddr
= inet
->inet_saddr
;
6437 newinet
->inet_rcv_saddr
= inet
->inet_rcv_saddr
;
6438 newinet
->inet_dport
= htons(asoc
->peer
.port
);
6439 newinet
->pmtudisc
= inet
->pmtudisc
;
6440 newinet
->inet_id
= asoc
->next_tsn
^ jiffies
;
6442 newinet
->uc_ttl
= inet
->uc_ttl
;
6443 newinet
->mc_loop
= 1;
6444 newinet
->mc_ttl
= 1;
6445 newinet
->mc_index
= 0;
6446 newinet
->mc_list
= NULL
;
6449 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6450 * and its messages to the newsk.
6452 static void sctp_sock_migrate(struct sock
*oldsk
, struct sock
*newsk
,
6453 struct sctp_association
*assoc
,
6454 sctp_socket_type_t type
)
6456 struct sctp_sock
*oldsp
= sctp_sk(oldsk
);
6457 struct sctp_sock
*newsp
= sctp_sk(newsk
);
6458 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
6459 struct sctp_endpoint
*newep
= newsp
->ep
;
6460 struct sk_buff
*skb
, *tmp
;
6461 struct sctp_ulpevent
*event
;
6462 struct sctp_bind_hashbucket
*head
;
6464 /* Migrate socket buffer sizes and all the socket level options to the
6467 newsk
->sk_sndbuf
= oldsk
->sk_sndbuf
;
6468 newsk
->sk_rcvbuf
= oldsk
->sk_rcvbuf
;
6469 /* Brute force copy old sctp opt. */
6470 inet_sk_copy_descendant(newsk
, oldsk
);
6472 /* Restore the ep value that was overwritten with the above structure
6478 /* Hook this new socket in to the bind_hash list. */
6479 head
= &sctp_port_hashtable
[sctp_phashfn(inet_sk(oldsk
)->inet_num
)];
6480 sctp_local_bh_disable();
6481 sctp_spin_lock(&head
->lock
);
6482 pp
= sctp_sk(oldsk
)->bind_hash
;
6483 sk_add_bind_node(newsk
, &pp
->owner
);
6484 sctp_sk(newsk
)->bind_hash
= pp
;
6485 inet_sk(newsk
)->inet_num
= inet_sk(oldsk
)->inet_num
;
6486 sctp_spin_unlock(&head
->lock
);
6487 sctp_local_bh_enable();
6489 /* Copy the bind_addr list from the original endpoint to the new
6490 * endpoint so that we can handle restarts properly
6492 sctp_bind_addr_dup(&newsp
->ep
->base
.bind_addr
,
6493 &oldsp
->ep
->base
.bind_addr
, GFP_KERNEL
);
6495 /* Move any messages in the old socket's receive queue that are for the
6496 * peeled off association to the new socket's receive queue.
6498 sctp_skb_for_each(skb
, &oldsk
->sk_receive_queue
, tmp
) {
6499 event
= sctp_skb2event(skb
);
6500 if (event
->asoc
== assoc
) {
6501 __skb_unlink(skb
, &oldsk
->sk_receive_queue
);
6502 __skb_queue_tail(&newsk
->sk_receive_queue
, skb
);
6503 sctp_skb_set_owner_r_frag(skb
, newsk
);
6507 /* Clean up any messages pending delivery due to partial
6508 * delivery. Three cases:
6509 * 1) No partial deliver; no work.
6510 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6511 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6513 skb_queue_head_init(&newsp
->pd_lobby
);
6514 atomic_set(&sctp_sk(newsk
)->pd_mode
, assoc
->ulpq
.pd_mode
);
6516 if (atomic_read(&sctp_sk(oldsk
)->pd_mode
)) {
6517 struct sk_buff_head
*queue
;
6519 /* Decide which queue to move pd_lobby skbs to. */
6520 if (assoc
->ulpq
.pd_mode
) {
6521 queue
= &newsp
->pd_lobby
;
6523 queue
= &newsk
->sk_receive_queue
;
6525 /* Walk through the pd_lobby, looking for skbs that
6526 * need moved to the new socket.
6528 sctp_skb_for_each(skb
, &oldsp
->pd_lobby
, tmp
) {
6529 event
= sctp_skb2event(skb
);
6530 if (event
->asoc
== assoc
) {
6531 __skb_unlink(skb
, &oldsp
->pd_lobby
);
6532 __skb_queue_tail(queue
, skb
);
6533 sctp_skb_set_owner_r_frag(skb
, newsk
);
6537 /* Clear up any skbs waiting for the partial
6538 * delivery to finish.
6540 if (assoc
->ulpq
.pd_mode
)
6541 sctp_clear_pd(oldsk
, NULL
);
6545 sctp_skb_for_each(skb
, &assoc
->ulpq
.reasm
, tmp
)
6546 sctp_skb_set_owner_r_frag(skb
, newsk
);
6548 sctp_skb_for_each(skb
, &assoc
->ulpq
.lobby
, tmp
)
6549 sctp_skb_set_owner_r_frag(skb
, newsk
);
6551 /* Set the type of socket to indicate that it is peeled off from the
6552 * original UDP-style socket or created with the accept() call on a
6553 * TCP-style socket..
6557 /* Mark the new socket "in-use" by the user so that any packets
6558 * that may arrive on the association after we've moved it are
6559 * queued to the backlog. This prevents a potential race between
6560 * backlog processing on the old socket and new-packet processing
6561 * on the new socket.
6563 * The caller has just allocated newsk so we can guarantee that other
6564 * paths won't try to lock it and then oldsk.
6566 lock_sock_nested(newsk
, SINGLE_DEPTH_NESTING
);
6567 sctp_assoc_migrate(assoc
, newsk
);
6569 /* If the association on the newsk is already closed before accept()
6570 * is called, set RCV_SHUTDOWN flag.
6572 if (sctp_state(assoc
, CLOSED
) && sctp_style(newsk
, TCP
))
6573 newsk
->sk_shutdown
|= RCV_SHUTDOWN
;
6575 newsk
->sk_state
= SCTP_SS_ESTABLISHED
;
6576 sctp_release_sock(newsk
);
6580 /* This proto struct describes the ULP interface for SCTP. */
6581 struct proto sctp_prot
= {
6583 .owner
= THIS_MODULE
,
6584 .close
= sctp_close
,
6585 .connect
= sctp_connect
,
6586 .disconnect
= sctp_disconnect
,
6587 .accept
= sctp_accept
,
6588 .ioctl
= sctp_ioctl
,
6589 .init
= sctp_init_sock
,
6590 .destroy
= sctp_destroy_sock
,
6591 .shutdown
= sctp_shutdown
,
6592 .setsockopt
= sctp_setsockopt
,
6593 .getsockopt
= sctp_getsockopt
,
6594 .sendmsg
= sctp_sendmsg
,
6595 .recvmsg
= sctp_recvmsg
,
6597 .backlog_rcv
= sctp_backlog_rcv
,
6599 .unhash
= sctp_unhash
,
6600 .get_port
= sctp_get_port
,
6601 .obj_size
= sizeof(struct sctp_sock
),
6602 .sysctl_mem
= sysctl_sctp_mem
,
6603 .sysctl_rmem
= sysctl_sctp_rmem
,
6604 .sysctl_wmem
= sysctl_sctp_wmem
,
6605 .memory_pressure
= &sctp_memory_pressure
,
6606 .enter_memory_pressure
= sctp_enter_memory_pressure
,
6607 .memory_allocated
= &sctp_memory_allocated
,
6608 .sockets_allocated
= &sctp_sockets_allocated
,
6611 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6613 struct proto sctpv6_prot
= {
6615 .owner
= THIS_MODULE
,
6616 .close
= sctp_close
,
6617 .connect
= sctp_connect
,
6618 .disconnect
= sctp_disconnect
,
6619 .accept
= sctp_accept
,
6620 .ioctl
= sctp_ioctl
,
6621 .init
= sctp_init_sock
,
6622 .destroy
= sctp_destroy_sock
,
6623 .shutdown
= sctp_shutdown
,
6624 .setsockopt
= sctp_setsockopt
,
6625 .getsockopt
= sctp_getsockopt
,
6626 .sendmsg
= sctp_sendmsg
,
6627 .recvmsg
= sctp_recvmsg
,
6629 .backlog_rcv
= sctp_backlog_rcv
,
6631 .unhash
= sctp_unhash
,
6632 .get_port
= sctp_get_port
,
6633 .obj_size
= sizeof(struct sctp6_sock
),
6634 .sysctl_mem
= sysctl_sctp_mem
,
6635 .sysctl_rmem
= sysctl_sctp_rmem
,
6636 .sysctl_wmem
= sysctl_sctp_wmem
,
6637 .memory_pressure
= &sctp_memory_pressure
,
6638 .enter_memory_pressure
= sctp_enter_memory_pressure
,
6639 .memory_allocated
= &sctp_memory_allocated
,
6640 .sockets_allocated
= &sctp_sockets_allocated
,
6642 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */