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>
73 #include <linux/file.h>
74 #include <linux/compat.h>
78 #include <net/route.h>
80 #include <net/inet_common.h>
82 #include <linux/socket.h> /* for sa_family_t */
83 #include <linux/export.h>
85 #include <net/sctp/sctp.h>
86 #include <net/sctp/sm.h>
88 /* WARNING: Please do not remove the SCTP_STATIC attribute to
89 * any of the functions below as they are used to export functions
90 * used by a project regression testsuite.
93 /* Forward declarations for internal helper functions. */
94 static int sctp_writeable(struct sock
*sk
);
95 static void sctp_wfree(struct sk_buff
*skb
);
96 static int sctp_wait_for_sndbuf(struct sctp_association
*, long *timeo_p
,
98 static int sctp_wait_for_packet(struct sock
* sk
, int *err
, long *timeo_p
);
99 static int sctp_wait_for_connect(struct sctp_association
*, long *timeo_p
);
100 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
);
101 static void sctp_wait_for_close(struct sock
*sk
, long timeo
);
102 static struct sctp_af
*sctp_sockaddr_af(struct sctp_sock
*opt
,
103 union sctp_addr
*addr
, int len
);
104 static int sctp_bindx_add(struct sock
*, struct sockaddr
*, int);
105 static int sctp_bindx_rem(struct sock
*, struct sockaddr
*, int);
106 static int sctp_send_asconf_add_ip(struct sock
*, struct sockaddr
*, int);
107 static int sctp_send_asconf_del_ip(struct sock
*, struct sockaddr
*, int);
108 static int sctp_send_asconf(struct sctp_association
*asoc
,
109 struct sctp_chunk
*chunk
);
110 static int sctp_do_bind(struct sock
*, union sctp_addr
*, int);
111 static int sctp_autobind(struct sock
*sk
);
112 static void sctp_sock_migrate(struct sock
*, struct sock
*,
113 struct sctp_association
*, sctp_socket_type_t
);
115 extern struct kmem_cache
*sctp_bucket_cachep
;
116 extern long sysctl_sctp_mem
[3];
117 extern int sysctl_sctp_rmem
[3];
118 extern int sysctl_sctp_wmem
[3];
120 static int sctp_memory_pressure
;
121 static atomic_long_t sctp_memory_allocated
;
122 struct percpu_counter sctp_sockets_allocated
;
124 static void sctp_enter_memory_pressure(struct sock
*sk
)
126 sctp_memory_pressure
= 1;
130 /* Get the sndbuf space available at the time on the association. */
131 static inline int sctp_wspace(struct sctp_association
*asoc
)
135 if (asoc
->ep
->sndbuf_policy
)
136 amt
= asoc
->sndbuf_used
;
138 amt
= sk_wmem_alloc_get(asoc
->base
.sk
);
140 if (amt
>= asoc
->base
.sk
->sk_sndbuf
) {
141 if (asoc
->base
.sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)
144 amt
= sk_stream_wspace(asoc
->base
.sk
);
149 amt
= asoc
->base
.sk
->sk_sndbuf
- amt
;
154 /* Increment the used sndbuf space count of the corresponding association by
155 * the size of the outgoing data chunk.
156 * Also, set the skb destructor for sndbuf accounting later.
158 * Since it is always 1-1 between chunk and skb, and also a new skb is always
159 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
160 * destructor in the data chunk skb for the purpose of the sndbuf space
163 static inline void sctp_set_owner_w(struct sctp_chunk
*chunk
)
165 struct sctp_association
*asoc
= chunk
->asoc
;
166 struct sock
*sk
= asoc
->base
.sk
;
168 /* The sndbuf space is tracked per association. */
169 sctp_association_hold(asoc
);
171 skb_set_owner_w(chunk
->skb
, sk
);
173 chunk
->skb
->destructor
= sctp_wfree
;
174 /* Save the chunk pointer in skb for sctp_wfree to use later. */
175 *((struct sctp_chunk
**)(chunk
->skb
->cb
)) = chunk
;
177 asoc
->sndbuf_used
+= SCTP_DATA_SNDSIZE(chunk
) +
178 sizeof(struct sk_buff
) +
179 sizeof(struct sctp_chunk
);
181 atomic_add(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
182 sk
->sk_wmem_queued
+= chunk
->skb
->truesize
;
183 sk_mem_charge(sk
, chunk
->skb
->truesize
);
186 /* Verify that this is a valid address. */
187 static inline int sctp_verify_addr(struct sock
*sk
, union sctp_addr
*addr
,
192 /* Verify basic sockaddr. */
193 af
= sctp_sockaddr_af(sctp_sk(sk
), addr
, len
);
197 /* Is this a valid SCTP address? */
198 if (!af
->addr_valid(addr
, sctp_sk(sk
), NULL
))
201 if (!sctp_sk(sk
)->pf
->send_verify(sctp_sk(sk
), (addr
)))
207 /* Look up the association by its id. If this is not a UDP-style
208 * socket, the ID field is always ignored.
210 struct sctp_association
*sctp_id2assoc(struct sock
*sk
, sctp_assoc_t id
)
212 struct sctp_association
*asoc
= NULL
;
214 /* If this is not a UDP-style socket, assoc id should be ignored. */
215 if (!sctp_style(sk
, UDP
)) {
216 /* Return NULL if the socket state is not ESTABLISHED. It
217 * could be a TCP-style listening socket or a socket which
218 * hasn't yet called connect() to establish an association.
220 if (!sctp_sstate(sk
, ESTABLISHED
))
223 /* Get the first and the only association from the list. */
224 if (!list_empty(&sctp_sk(sk
)->ep
->asocs
))
225 asoc
= list_entry(sctp_sk(sk
)->ep
->asocs
.next
,
226 struct sctp_association
, asocs
);
230 /* Otherwise this is a UDP-style socket. */
231 if (!id
|| (id
== (sctp_assoc_t
)-1))
234 spin_lock_bh(&sctp_assocs_id_lock
);
235 asoc
= (struct sctp_association
*)idr_find(&sctp_assocs_id
, (int)id
);
236 spin_unlock_bh(&sctp_assocs_id_lock
);
238 if (!asoc
|| (asoc
->base
.sk
!= sk
) || asoc
->base
.dead
)
244 /* Look up the transport from an address and an assoc id. If both address and
245 * id are specified, the associations matching the address and the id should be
248 static struct sctp_transport
*sctp_addr_id2transport(struct sock
*sk
,
249 struct sockaddr_storage
*addr
,
252 struct sctp_association
*addr_asoc
= NULL
, *id_asoc
= NULL
;
253 struct sctp_transport
*transport
;
254 union sctp_addr
*laddr
= (union sctp_addr
*)addr
;
256 addr_asoc
= sctp_endpoint_lookup_assoc(sctp_sk(sk
)->ep
,
263 id_asoc
= sctp_id2assoc(sk
, id
);
264 if (id_asoc
&& (id_asoc
!= addr_asoc
))
267 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
268 (union sctp_addr
*)addr
);
273 /* API 3.1.2 bind() - UDP Style Syntax
274 * The syntax of bind() is,
276 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
278 * sd - the socket descriptor returned by socket().
279 * addr - the address structure (struct sockaddr_in or struct
280 * sockaddr_in6 [RFC 2553]),
281 * addr_len - the size of the address structure.
283 SCTP_STATIC
int sctp_bind(struct sock
*sk
, struct sockaddr
*addr
, int addr_len
)
289 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
292 /* Disallow binding twice. */
293 if (!sctp_sk(sk
)->ep
->base
.bind_addr
.port
)
294 retval
= sctp_do_bind(sk
, (union sctp_addr
*)addr
,
299 sctp_release_sock(sk
);
304 static long sctp_get_port_local(struct sock
*, union sctp_addr
*);
306 /* Verify this is a valid sockaddr. */
307 static struct sctp_af
*sctp_sockaddr_af(struct sctp_sock
*opt
,
308 union sctp_addr
*addr
, int len
)
312 /* Check minimum size. */
313 if (len
< sizeof (struct sockaddr
))
316 /* V4 mapped address are really of AF_INET family */
317 if (addr
->sa
.sa_family
== AF_INET6
&&
318 ipv6_addr_v4mapped(&addr
->v6
.sin6_addr
)) {
319 if (!opt
->pf
->af_supported(AF_INET
, opt
))
322 /* Does this PF support this AF? */
323 if (!opt
->pf
->af_supported(addr
->sa
.sa_family
, opt
))
327 /* If we get this far, af is valid. */
328 af
= sctp_get_af_specific(addr
->sa
.sa_family
);
330 if (len
< af
->sockaddr_len
)
336 /* Bind a local address either to an endpoint or to an association. */
337 SCTP_STATIC
int sctp_do_bind(struct sock
*sk
, union sctp_addr
*addr
, int len
)
339 struct net
*net
= sock_net(sk
);
340 struct sctp_sock
*sp
= sctp_sk(sk
);
341 struct sctp_endpoint
*ep
= sp
->ep
;
342 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
347 /* Common sockaddr verification. */
348 af
= sctp_sockaddr_af(sp
, addr
, len
);
350 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
355 snum
= ntohs(addr
->v4
.sin_port
);
357 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
358 ", port: %d, new port: %d, len: %d)\n",
364 /* PF specific bind() address verification. */
365 if (!sp
->pf
->bind_verify(sp
, addr
))
366 return -EADDRNOTAVAIL
;
368 /* We must either be unbound, or bind to the same port.
369 * It's OK to allow 0 ports if we are already bound.
370 * We'll just inhert an already bound port in this case
375 else if (snum
!= bp
->port
) {
376 SCTP_DEBUG_PRINTK("sctp_do_bind:"
377 " New port %d does not match existing port "
378 "%d.\n", snum
, bp
->port
);
383 if (snum
&& snum
< PROT_SOCK
&&
384 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
))
387 /* See if the address matches any of the addresses we may have
388 * already bound before checking against other endpoints.
390 if (sctp_bind_addr_match(bp
, addr
, sp
))
393 /* Make sure we are allowed to bind here.
394 * The function sctp_get_port_local() does duplicate address
397 addr
->v4
.sin_port
= htons(snum
);
398 if ((ret
= sctp_get_port_local(sk
, addr
))) {
402 /* Refresh ephemeral port. */
404 bp
->port
= inet_sk(sk
)->inet_num
;
406 /* Add the address to the bind address list.
407 * Use GFP_ATOMIC since BHs will be disabled.
409 ret
= sctp_add_bind_addr(bp
, addr
, SCTP_ADDR_SRC
, GFP_ATOMIC
);
411 /* Copy back into socket for getsockname() use. */
413 inet_sk(sk
)->inet_sport
= htons(inet_sk(sk
)->inet_num
);
414 af
->to_sk_saddr(addr
, sk
);
420 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
422 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
423 * at any one time. If a sender, after sending an ASCONF chunk, decides
424 * it needs to transfer another ASCONF Chunk, it MUST wait until the
425 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
426 * subsequent ASCONF. Note this restriction binds each side, so at any
427 * time two ASCONF may be in-transit on any given association (one sent
428 * from each endpoint).
430 static int sctp_send_asconf(struct sctp_association
*asoc
,
431 struct sctp_chunk
*chunk
)
433 struct net
*net
= sock_net(asoc
->base
.sk
);
436 /* If there is an outstanding ASCONF chunk, queue it for later
439 if (asoc
->addip_last_asconf
) {
440 list_add_tail(&chunk
->list
, &asoc
->addip_chunk_list
);
444 /* Hold the chunk until an ASCONF_ACK is received. */
445 sctp_chunk_hold(chunk
);
446 retval
= sctp_primitive_ASCONF(net
, asoc
, chunk
);
448 sctp_chunk_free(chunk
);
450 asoc
->addip_last_asconf
= chunk
;
456 /* Add a list of addresses as bind addresses to local endpoint or
459 * Basically run through each address specified in the addrs/addrcnt
460 * array/length pair, determine if it is IPv6 or IPv4 and call
461 * sctp_do_bind() on it.
463 * If any of them fails, then the operation will be reversed and the
464 * ones that were added will be removed.
466 * Only sctp_setsockopt_bindx() is supposed to call this function.
468 static int sctp_bindx_add(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
473 struct sockaddr
*sa_addr
;
476 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
480 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
481 /* The list may contain either IPv4 or IPv6 address;
482 * determine the address length for walking thru the list.
485 af
= sctp_get_af_specific(sa_addr
->sa_family
);
491 retval
= sctp_do_bind(sk
, (union sctp_addr
*)sa_addr
,
494 addr_buf
+= af
->sockaddr_len
;
498 /* Failed. Cleanup the ones that have been added */
500 sctp_bindx_rem(sk
, addrs
, cnt
);
508 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
509 * associations that are part of the endpoint indicating that a list of local
510 * addresses are added to the endpoint.
512 * If any of the addresses is already in the bind address list of the
513 * association, we do not send the chunk for that association. But it will not
514 * affect other associations.
516 * Only sctp_setsockopt_bindx() is supposed to call this function.
518 static int sctp_send_asconf_add_ip(struct sock
*sk
,
519 struct sockaddr
*addrs
,
522 struct net
*net
= sock_net(sk
);
523 struct sctp_sock
*sp
;
524 struct sctp_endpoint
*ep
;
525 struct sctp_association
*asoc
;
526 struct sctp_bind_addr
*bp
;
527 struct sctp_chunk
*chunk
;
528 struct sctp_sockaddr_entry
*laddr
;
529 union sctp_addr
*addr
;
530 union sctp_addr saveaddr
;
537 if (!net
->sctp
.addip_enable
)
543 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
544 __func__
, sk
, addrs
, addrcnt
);
546 list_for_each_entry(asoc
, &ep
->asocs
, asocs
) {
548 if (!asoc
->peer
.asconf_capable
)
551 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_ADD_IP
)
554 if (!sctp_state(asoc
, ESTABLISHED
))
557 /* Check if any address in the packed array of addresses is
558 * in the bind address list of the association. If so,
559 * do not send the asconf chunk to its peer, but continue with
560 * other associations.
563 for (i
= 0; i
< addrcnt
; i
++) {
565 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
571 if (sctp_assoc_lookup_laddr(asoc
, addr
))
574 addr_buf
+= af
->sockaddr_len
;
579 /* Use the first valid address in bind addr list of
580 * association as Address Parameter of ASCONF CHUNK.
582 bp
= &asoc
->base
.bind_addr
;
583 p
= bp
->address_list
.next
;
584 laddr
= list_entry(p
, struct sctp_sockaddr_entry
, list
);
585 chunk
= sctp_make_asconf_update_ip(asoc
, &laddr
->a
, addrs
,
586 addrcnt
, SCTP_PARAM_ADD_IP
);
592 /* Add the new addresses to the bind address list with
593 * use_as_src set to 0.
596 for (i
= 0; i
< addrcnt
; i
++) {
598 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
599 memcpy(&saveaddr
, addr
, af
->sockaddr_len
);
600 retval
= sctp_add_bind_addr(bp
, &saveaddr
,
601 SCTP_ADDR_NEW
, GFP_ATOMIC
);
602 addr_buf
+= af
->sockaddr_len
;
604 if (asoc
->src_out_of_asoc_ok
) {
605 struct sctp_transport
*trans
;
607 list_for_each_entry(trans
,
608 &asoc
->peer
.transport_addr_list
, transports
) {
609 /* Clear the source and route cache */
610 dst_release(trans
->dst
);
611 trans
->cwnd
= min(4*asoc
->pathmtu
, max_t(__u32
,
612 2*asoc
->pathmtu
, 4380));
613 trans
->ssthresh
= asoc
->peer
.i
.a_rwnd
;
614 trans
->rto
= asoc
->rto_initial
;
615 sctp_max_rto(asoc
, trans
);
616 trans
->rtt
= trans
->srtt
= trans
->rttvar
= 0;
617 sctp_transport_route(trans
, NULL
,
618 sctp_sk(asoc
->base
.sk
));
621 retval
= sctp_send_asconf(asoc
, chunk
);
628 /* Remove a list of addresses from bind addresses list. Do not remove the
631 * Basically run through each address specified in the addrs/addrcnt
632 * array/length pair, determine if it is IPv6 or IPv4 and call
633 * sctp_del_bind() on it.
635 * If any of them fails, then the operation will be reversed and the
636 * ones that were removed will be added back.
638 * At least one address has to be left; if only one address is
639 * available, the operation will return -EBUSY.
641 * Only sctp_setsockopt_bindx() is supposed to call this function.
643 static int sctp_bindx_rem(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
645 struct sctp_sock
*sp
= sctp_sk(sk
);
646 struct sctp_endpoint
*ep
= sp
->ep
;
648 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
651 union sctp_addr
*sa_addr
;
654 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
658 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
659 /* If the bind address list is empty or if there is only one
660 * bind address, there is nothing more to be removed (we need
661 * at least one address here).
663 if (list_empty(&bp
->address_list
) ||
664 (sctp_list_single_entry(&bp
->address_list
))) {
670 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
676 if (!af
->addr_valid(sa_addr
, sp
, NULL
)) {
677 retval
= -EADDRNOTAVAIL
;
681 if (sa_addr
->v4
.sin_port
&&
682 sa_addr
->v4
.sin_port
!= htons(bp
->port
)) {
687 if (!sa_addr
->v4
.sin_port
)
688 sa_addr
->v4
.sin_port
= htons(bp
->port
);
690 /* FIXME - There is probably a need to check if sk->sk_saddr and
691 * sk->sk_rcv_addr are currently set to one of the addresses to
692 * be removed. This is something which needs to be looked into
693 * when we are fixing the outstanding issues with multi-homing
694 * socket routing and failover schemes. Refer to comments in
695 * sctp_do_bind(). -daisy
697 retval
= sctp_del_bind_addr(bp
, sa_addr
);
699 addr_buf
+= af
->sockaddr_len
;
702 /* Failed. Add the ones that has been removed back */
704 sctp_bindx_add(sk
, addrs
, cnt
);
712 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
713 * the associations that are part of the endpoint indicating that a list of
714 * local addresses are removed from the endpoint.
716 * If any of the addresses is already in the bind address list of the
717 * association, we do not send the chunk for that association. But it will not
718 * affect other associations.
720 * Only sctp_setsockopt_bindx() is supposed to call this function.
722 static int sctp_send_asconf_del_ip(struct sock
*sk
,
723 struct sockaddr
*addrs
,
726 struct net
*net
= sock_net(sk
);
727 struct sctp_sock
*sp
;
728 struct sctp_endpoint
*ep
;
729 struct sctp_association
*asoc
;
730 struct sctp_transport
*transport
;
731 struct sctp_bind_addr
*bp
;
732 struct sctp_chunk
*chunk
;
733 union sctp_addr
*laddr
;
736 struct sctp_sockaddr_entry
*saddr
;
742 if (!net
->sctp
.addip_enable
)
748 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
749 __func__
, sk
, addrs
, addrcnt
);
751 list_for_each_entry(asoc
, &ep
->asocs
, asocs
) {
753 if (!asoc
->peer
.asconf_capable
)
756 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_DEL_IP
)
759 if (!sctp_state(asoc
, ESTABLISHED
))
762 /* Check if any address in the packed array of addresses is
763 * not present in the bind address list of the association.
764 * If so, do not send the asconf chunk to its peer, but
765 * continue with other associations.
768 for (i
= 0; i
< addrcnt
; i
++) {
770 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
776 if (!sctp_assoc_lookup_laddr(asoc
, laddr
))
779 addr_buf
+= af
->sockaddr_len
;
784 /* Find one address in the association's bind address list
785 * that is not in the packed array of addresses. This is to
786 * make sure that we do not delete all the addresses in the
789 bp
= &asoc
->base
.bind_addr
;
790 laddr
= sctp_find_unmatch_addr(bp
, (union sctp_addr
*)addrs
,
792 if ((laddr
== NULL
) && (addrcnt
== 1)) {
793 if (asoc
->asconf_addr_del_pending
)
795 asoc
->asconf_addr_del_pending
=
796 kzalloc(sizeof(union sctp_addr
), GFP_ATOMIC
);
797 if (asoc
->asconf_addr_del_pending
== NULL
) {
801 asoc
->asconf_addr_del_pending
->sa
.sa_family
=
803 asoc
->asconf_addr_del_pending
->v4
.sin_port
=
805 if (addrs
->sa_family
== AF_INET
) {
806 struct sockaddr_in
*sin
;
808 sin
= (struct sockaddr_in
*)addrs
;
809 asoc
->asconf_addr_del_pending
->v4
.sin_addr
.s_addr
= sin
->sin_addr
.s_addr
;
810 } else if (addrs
->sa_family
== AF_INET6
) {
811 struct sockaddr_in6
*sin6
;
813 sin6
= (struct sockaddr_in6
*)addrs
;
814 asoc
->asconf_addr_del_pending
->v6
.sin6_addr
= sin6
->sin6_addr
;
816 SCTP_DEBUG_PRINTK_IPADDR("send_asconf_del_ip: keep the last address asoc: %p ",
817 " at %p\n", asoc
, asoc
->asconf_addr_del_pending
,
818 asoc
->asconf_addr_del_pending
);
819 asoc
->src_out_of_asoc_ok
= 1;
827 /* We do not need RCU protection throughout this loop
828 * because this is done under a socket lock from the
831 chunk
= sctp_make_asconf_update_ip(asoc
, laddr
, addrs
, addrcnt
,
839 /* Reset use_as_src flag for the addresses in the bind address
840 * list that are to be deleted.
843 for (i
= 0; i
< addrcnt
; i
++) {
845 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
846 list_for_each_entry(saddr
, &bp
->address_list
, list
) {
847 if (sctp_cmp_addr_exact(&saddr
->a
, laddr
))
848 saddr
->state
= SCTP_ADDR_DEL
;
850 addr_buf
+= af
->sockaddr_len
;
853 /* Update the route and saddr entries for all the transports
854 * as some of the addresses in the bind address list are
855 * about to be deleted and cannot be used as source addresses.
857 list_for_each_entry(transport
, &asoc
->peer
.transport_addr_list
,
859 dst_release(transport
->dst
);
860 sctp_transport_route(transport
, NULL
,
861 sctp_sk(asoc
->base
.sk
));
865 /* We don't need to transmit ASCONF */
867 retval
= sctp_send_asconf(asoc
, chunk
);
873 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
874 int sctp_asconf_mgmt(struct sctp_sock
*sp
, struct sctp_sockaddr_entry
*addrw
)
876 struct sock
*sk
= sctp_opt2sk(sp
);
877 union sctp_addr
*addr
;
880 /* It is safe to write port space in caller. */
882 addr
->v4
.sin_port
= htons(sp
->ep
->base
.bind_addr
.port
);
883 af
= sctp_get_af_specific(addr
->sa
.sa_family
);
886 if (sctp_verify_addr(sk
, addr
, af
->sockaddr_len
))
889 if (addrw
->state
== SCTP_ADDR_NEW
)
890 return sctp_send_asconf_add_ip(sk
, (struct sockaddr
*)addr
, 1);
892 return sctp_send_asconf_del_ip(sk
, (struct sockaddr
*)addr
, 1);
895 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
898 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
901 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
902 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
905 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
906 * Section 3.1.2 for this usage.
908 * addrs is a pointer to an array of one or more socket addresses. Each
909 * address is contained in its appropriate structure (i.e. struct
910 * sockaddr_in or struct sockaddr_in6) the family of the address type
911 * must be used to distinguish the address length (note that this
912 * representation is termed a "packed array" of addresses). The caller
913 * specifies the number of addresses in the array with addrcnt.
915 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
916 * -1, and sets errno to the appropriate error code.
918 * For SCTP, the port given in each socket address must be the same, or
919 * sctp_bindx() will fail, setting errno to EINVAL.
921 * The flags parameter is formed from the bitwise OR of zero or more of
922 * the following currently defined flags:
924 * SCTP_BINDX_ADD_ADDR
926 * SCTP_BINDX_REM_ADDR
928 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
929 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
930 * addresses from the association. The two flags are mutually exclusive;
931 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
932 * not remove all addresses from an association; sctp_bindx() will
933 * reject such an attempt with EINVAL.
935 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
936 * additional addresses with an endpoint after calling bind(). Or use
937 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
938 * socket is associated with so that no new association accepted will be
939 * associated with those addresses. If the endpoint supports dynamic
940 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
941 * endpoint to send the appropriate message to the peer to change the
942 * peers address lists.
944 * Adding and removing addresses from a connected association is
945 * optional functionality. Implementations that do not support this
946 * functionality should return EOPNOTSUPP.
948 * Basically do nothing but copying the addresses from user to kernel
949 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
950 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
953 * We don't use copy_from_user() for optimization: we first do the
954 * sanity checks (buffer size -fast- and access check-healthy
955 * pointer); if all of those succeed, then we can alloc the memory
956 * (expensive operation) needed to copy the data to kernel. Then we do
957 * the copying without checking the user space area
958 * (__copy_from_user()).
960 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
963 * sk The sk of the socket
964 * addrs The pointer to the addresses in user land
965 * addrssize Size of the addrs buffer
966 * op Operation to perform (add or remove, see the flags of
969 * Returns 0 if ok, <0 errno code on error.
971 SCTP_STATIC
int sctp_setsockopt_bindx(struct sock
* sk
,
972 struct sockaddr __user
*addrs
,
973 int addrs_size
, int op
)
975 struct sockaddr
*kaddrs
;
979 struct sockaddr
*sa_addr
;
983 SCTP_DEBUG_PRINTK("sctp_setsockopt_bindx: sk %p addrs %p"
984 " addrs_size %d opt %d\n", sk
, addrs
, addrs_size
, op
);
986 if (unlikely(addrs_size
<= 0))
989 /* Check the user passed a healthy pointer. */
990 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
993 /* Alloc space for the address array in kernel memory. */
994 kaddrs
= kmalloc(addrs_size
, GFP_KERNEL
);
995 if (unlikely(!kaddrs
))
998 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
1003 /* Walk through the addrs buffer and count the number of addresses. */
1005 while (walk_size
< addrs_size
) {
1006 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
1012 af
= sctp_get_af_specific(sa_addr
->sa_family
);
1014 /* If the address family is not supported or if this address
1015 * causes the address buffer to overflow return EINVAL.
1017 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
1022 addr_buf
+= af
->sockaddr_len
;
1023 walk_size
+= af
->sockaddr_len
;
1028 case SCTP_BINDX_ADD_ADDR
:
1029 err
= sctp_bindx_add(sk
, kaddrs
, addrcnt
);
1032 err
= sctp_send_asconf_add_ip(sk
, kaddrs
, addrcnt
);
1035 case SCTP_BINDX_REM_ADDR
:
1036 err
= sctp_bindx_rem(sk
, kaddrs
, addrcnt
);
1039 err
= sctp_send_asconf_del_ip(sk
, kaddrs
, addrcnt
);
1053 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1055 * Common routine for handling connect() and sctp_connectx().
1056 * Connect will come in with just a single address.
1058 static int __sctp_connect(struct sock
* sk
,
1059 struct sockaddr
*kaddrs
,
1061 sctp_assoc_t
*assoc_id
)
1063 struct net
*net
= sock_net(sk
);
1064 struct sctp_sock
*sp
;
1065 struct sctp_endpoint
*ep
;
1066 struct sctp_association
*asoc
= NULL
;
1067 struct sctp_association
*asoc2
;
1068 struct sctp_transport
*transport
;
1076 union sctp_addr
*sa_addr
= NULL
;
1078 unsigned short port
;
1079 unsigned int f_flags
= 0;
1084 /* connect() cannot be done on a socket that is already in ESTABLISHED
1085 * state - UDP-style peeled off socket or a TCP-style socket that
1086 * is already connected.
1087 * It cannot be done even on a TCP-style listening socket.
1089 if (sctp_sstate(sk
, ESTABLISHED
) ||
1090 (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))) {
1095 /* Walk through the addrs buffer and count the number of addresses. */
1097 while (walk_size
< addrs_size
) {
1098 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
1104 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
1106 /* If the address family is not supported or if this address
1107 * causes the address buffer to overflow return EINVAL.
1109 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
1114 port
= ntohs(sa_addr
->v4
.sin_port
);
1116 /* Save current address so we can work with it */
1117 memcpy(&to
, sa_addr
, af
->sockaddr_len
);
1119 err
= sctp_verify_addr(sk
, &to
, af
->sockaddr_len
);
1123 /* Make sure the destination port is correctly set
1126 if (asoc
&& asoc
->peer
.port
&& asoc
->peer
.port
!= port
) {
1131 /* Check if there already is a matching association on the
1132 * endpoint (other than the one created here).
1134 asoc2
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1135 if (asoc2
&& asoc2
!= asoc
) {
1136 if (asoc2
->state
>= SCTP_STATE_ESTABLISHED
)
1143 /* If we could not find a matching association on the endpoint,
1144 * make sure that there is no peeled-off association matching
1145 * the peer address even on another socket.
1147 if (sctp_endpoint_is_peeled_off(ep
, &to
)) {
1148 err
= -EADDRNOTAVAIL
;
1153 /* If a bind() or sctp_bindx() is not called prior to
1154 * an sctp_connectx() call, the system picks an
1155 * ephemeral port and will choose an address set
1156 * equivalent to binding with a wildcard address.
1158 if (!ep
->base
.bind_addr
.port
) {
1159 if (sctp_autobind(sk
)) {
1165 * If an unprivileged user inherits a 1-many
1166 * style socket with open associations on a
1167 * privileged port, it MAY be permitted to
1168 * accept new associations, but it SHOULD NOT
1169 * be permitted to open new associations.
1171 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1172 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
)) {
1178 scope
= sctp_scope(&to
);
1179 asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1185 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
,
1193 /* Prime the peer's transport structures. */
1194 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
,
1202 addr_buf
+= af
->sockaddr_len
;
1203 walk_size
+= af
->sockaddr_len
;
1206 /* In case the user of sctp_connectx() wants an association
1207 * id back, assign one now.
1210 err
= sctp_assoc_set_id(asoc
, GFP_KERNEL
);
1215 err
= sctp_primitive_ASSOCIATE(net
, asoc
, NULL
);
1220 /* Initialize sk's dport and daddr for getpeername() */
1221 inet_sk(sk
)->inet_dport
= htons(asoc
->peer
.port
);
1222 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
1223 af
->to_sk_daddr(sa_addr
, sk
);
1226 /* in-kernel sockets don't generally have a file allocated to them
1227 * if all they do is call sock_create_kern().
1229 if (sk
->sk_socket
->file
)
1230 f_flags
= sk
->sk_socket
->file
->f_flags
;
1232 timeo
= sock_sndtimeo(sk
, f_flags
& O_NONBLOCK
);
1235 *assoc_id
= asoc
->assoc_id
;
1236 err
= sctp_wait_for_connect(asoc
, &timeo
);
1237 /* Note: the asoc may be freed after the return of
1238 * sctp_wait_for_connect.
1241 /* Don't free association on exit. */
1246 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1247 " kaddrs: %p err: %d\n",
1250 /* sctp_primitive_ASSOCIATE may have added this association
1251 * To the hash table, try to unhash it, just in case, its a noop
1252 * if it wasn't hashed so we're safe
1254 sctp_unhash_established(asoc
);
1255 sctp_association_free(asoc
);
1260 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1263 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1264 * sctp_assoc_t *asoc);
1266 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1267 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1268 * or IPv6 addresses.
1270 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1271 * Section 3.1.2 for this usage.
1273 * addrs is a pointer to an array of one or more socket addresses. Each
1274 * address is contained in its appropriate structure (i.e. struct
1275 * sockaddr_in or struct sockaddr_in6) the family of the address type
1276 * must be used to distengish the address length (note that this
1277 * representation is termed a "packed array" of addresses). The caller
1278 * specifies the number of addresses in the array with addrcnt.
1280 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1281 * the association id of the new association. On failure, sctp_connectx()
1282 * returns -1, and sets errno to the appropriate error code. The assoc_id
1283 * is not touched by the kernel.
1285 * For SCTP, the port given in each socket address must be the same, or
1286 * sctp_connectx() will fail, setting errno to EINVAL.
1288 * An application can use sctp_connectx to initiate an association with
1289 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1290 * allows a caller to specify multiple addresses at which a peer can be
1291 * reached. The way the SCTP stack uses the list of addresses to set up
1292 * the association is implementation dependent. This function only
1293 * specifies that the stack will try to make use of all the addresses in
1294 * the list when needed.
1296 * Note that the list of addresses passed in is only used for setting up
1297 * the association. It does not necessarily equal the set of addresses
1298 * the peer uses for the resulting association. If the caller wants to
1299 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1300 * retrieve them after the association has been set up.
1302 * Basically do nothing but copying the addresses from user to kernel
1303 * land and invoking either sctp_connectx(). This is used for tunneling
1304 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1306 * We don't use copy_from_user() for optimization: we first do the
1307 * sanity checks (buffer size -fast- and access check-healthy
1308 * pointer); if all of those succeed, then we can alloc the memory
1309 * (expensive operation) needed to copy the data to kernel. Then we do
1310 * the copying without checking the user space area
1311 * (__copy_from_user()).
1313 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1316 * sk The sk of the socket
1317 * addrs The pointer to the addresses in user land
1318 * addrssize Size of the addrs buffer
1320 * Returns >=0 if ok, <0 errno code on error.
1322 SCTP_STATIC
int __sctp_setsockopt_connectx(struct sock
* sk
,
1323 struct sockaddr __user
*addrs
,
1325 sctp_assoc_t
*assoc_id
)
1328 struct sockaddr
*kaddrs
;
1330 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1331 __func__
, sk
, addrs
, addrs_size
);
1333 if (unlikely(addrs_size
<= 0))
1336 /* Check the user passed a healthy pointer. */
1337 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
1340 /* Alloc space for the address array in kernel memory. */
1341 kaddrs
= kmalloc(addrs_size
, GFP_KERNEL
);
1342 if (unlikely(!kaddrs
))
1345 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
1348 err
= __sctp_connect(sk
, kaddrs
, addrs_size
, assoc_id
);
1357 * This is an older interface. It's kept for backward compatibility
1358 * to the option that doesn't provide association id.
1360 SCTP_STATIC
int sctp_setsockopt_connectx_old(struct sock
* sk
,
1361 struct sockaddr __user
*addrs
,
1364 return __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, NULL
);
1368 * New interface for the API. The since the API is done with a socket
1369 * option, to make it simple we feed back the association id is as a return
1370 * indication to the call. Error is always negative and association id is
1373 SCTP_STATIC
int sctp_setsockopt_connectx(struct sock
* sk
,
1374 struct sockaddr __user
*addrs
,
1377 sctp_assoc_t assoc_id
= 0;
1380 err
= __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, &assoc_id
);
1389 * New (hopefully final) interface for the API.
1390 * We use the sctp_getaddrs_old structure so that use-space library
1391 * can avoid any unnecessary allocations. The only different part
1392 * is that we store the actual length of the address buffer into the
1393 * addrs_num structure member. That way we can re-use the existing
1396 #ifdef CONFIG_COMPAT
1397 struct compat_sctp_getaddrs_old
{
1398 sctp_assoc_t assoc_id
;
1400 compat_uptr_t addrs
; /* struct sockaddr * */
1404 SCTP_STATIC
int sctp_getsockopt_connectx3(struct sock
* sk
, int len
,
1405 char __user
*optval
,
1408 struct sctp_getaddrs_old param
;
1409 sctp_assoc_t assoc_id
= 0;
1412 #ifdef CONFIG_COMPAT
1413 if (is_compat_task()) {
1414 struct compat_sctp_getaddrs_old param32
;
1416 if (len
< sizeof(param32
))
1418 if (copy_from_user(¶m32
, optval
, sizeof(param32
)))
1421 param
.assoc_id
= param32
.assoc_id
;
1422 param
.addr_num
= param32
.addr_num
;
1423 param
.addrs
= compat_ptr(param32
.addrs
);
1427 if (len
< sizeof(param
))
1429 if (copy_from_user(¶m
, optval
, sizeof(param
)))
1433 err
= __sctp_setsockopt_connectx(sk
, (struct sockaddr __user
*)
1434 param
.addrs
, param
.addr_num
,
1436 if (err
== 0 || err
== -EINPROGRESS
) {
1437 if (copy_to_user(optval
, &assoc_id
, sizeof(assoc_id
)))
1439 if (put_user(sizeof(assoc_id
), optlen
))
1446 /* API 3.1.4 close() - UDP Style Syntax
1447 * Applications use close() to perform graceful shutdown (as described in
1448 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1449 * by a UDP-style socket.
1453 * ret = close(int sd);
1455 * sd - the socket descriptor of the associations to be closed.
1457 * To gracefully shutdown a specific association represented by the
1458 * UDP-style socket, an application should use the sendmsg() call,
1459 * passing no user data, but including the appropriate flag in the
1460 * ancillary data (see Section xxxx).
1462 * If sd in the close() call is a branched-off socket representing only
1463 * one association, the shutdown is performed on that association only.
1465 * 4.1.6 close() - TCP Style Syntax
1467 * Applications use close() to gracefully close down an association.
1471 * int close(int sd);
1473 * sd - the socket descriptor of the association to be closed.
1475 * After an application calls close() on a socket descriptor, no further
1476 * socket operations will succeed on that descriptor.
1478 * API 7.1.4 SO_LINGER
1480 * An application using the TCP-style socket can use this option to
1481 * perform the SCTP ABORT primitive. The linger option structure is:
1484 * int l_onoff; // option on/off
1485 * int l_linger; // linger time
1488 * To enable the option, set l_onoff to 1. If the l_linger value is set
1489 * to 0, calling close() is the same as the ABORT primitive. If the
1490 * value is set to a negative value, the setsockopt() call will return
1491 * an error. If the value is set to a positive value linger_time, the
1492 * close() can be blocked for at most linger_time ms. If the graceful
1493 * shutdown phase does not finish during this period, close() will
1494 * return but the graceful shutdown phase continues in the system.
1496 SCTP_STATIC
void sctp_close(struct sock
*sk
, long timeout
)
1498 struct net
*net
= sock_net(sk
);
1499 struct sctp_endpoint
*ep
;
1500 struct sctp_association
*asoc
;
1501 struct list_head
*pos
, *temp
;
1502 unsigned int data_was_unread
;
1504 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk
, timeout
);
1507 sk
->sk_shutdown
= SHUTDOWN_MASK
;
1508 sk
->sk_state
= SCTP_SS_CLOSING
;
1510 ep
= sctp_sk(sk
)->ep
;
1512 /* Clean up any skbs sitting on the receive queue. */
1513 data_was_unread
= sctp_queue_purge_ulpevents(&sk
->sk_receive_queue
);
1514 data_was_unread
+= sctp_queue_purge_ulpevents(&sctp_sk(sk
)->pd_lobby
);
1516 /* Walk all associations on an endpoint. */
1517 list_for_each_safe(pos
, temp
, &ep
->asocs
) {
1518 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
1520 if (sctp_style(sk
, TCP
)) {
1521 /* A closed association can still be in the list if
1522 * it belongs to a TCP-style listening socket that is
1523 * not yet accepted. If so, free it. If not, send an
1524 * ABORT or SHUTDOWN based on the linger options.
1526 if (sctp_state(asoc
, CLOSED
)) {
1527 sctp_unhash_established(asoc
);
1528 sctp_association_free(asoc
);
1533 if (data_was_unread
|| !skb_queue_empty(&asoc
->ulpq
.lobby
) ||
1534 !skb_queue_empty(&asoc
->ulpq
.reasm
) ||
1535 (sock_flag(sk
, SOCK_LINGER
) && !sk
->sk_lingertime
)) {
1536 struct sctp_chunk
*chunk
;
1538 chunk
= sctp_make_abort_user(asoc
, NULL
, 0);
1539 sctp_primitive_ABORT(net
, asoc
, chunk
);
1541 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
1544 /* On a TCP-style socket, block for at most linger_time if set. */
1545 if (sctp_style(sk
, TCP
) && timeout
)
1546 sctp_wait_for_close(sk
, timeout
);
1548 /* This will run the backlog queue. */
1549 sctp_release_sock(sk
);
1551 /* Supposedly, no process has access to the socket, but
1552 * the net layers still may.
1553 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1554 * held and that should be grabbed before socket lock.
1556 spin_lock_bh(&net
->sctp
.addr_wq_lock
);
1557 sctp_bh_lock_sock(sk
);
1559 /* Hold the sock, since sk_common_release() will put sock_put()
1560 * and we have just a little more cleanup.
1563 sk_common_release(sk
);
1565 sctp_bh_unlock_sock(sk
);
1566 spin_unlock_bh(&net
->sctp
.addr_wq_lock
);
1570 SCTP_DBG_OBJCNT_DEC(sock
);
1573 /* Handle EPIPE error. */
1574 static int sctp_error(struct sock
*sk
, int flags
, int err
)
1577 err
= sock_error(sk
) ? : -EPIPE
;
1578 if (err
== -EPIPE
&& !(flags
& MSG_NOSIGNAL
))
1579 send_sig(SIGPIPE
, current
, 0);
1583 /* API 3.1.3 sendmsg() - UDP Style Syntax
1585 * An application uses sendmsg() and recvmsg() calls to transmit data to
1586 * and receive data from its peer.
1588 * ssize_t sendmsg(int socket, const struct msghdr *message,
1591 * socket - the socket descriptor of the endpoint.
1592 * message - pointer to the msghdr structure which contains a single
1593 * user message and possibly some ancillary data.
1595 * See Section 5 for complete description of the data
1598 * flags - flags sent or received with the user message, see Section
1599 * 5 for complete description of the flags.
1601 * Note: This function could use a rewrite especially when explicit
1602 * connect support comes in.
1604 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1606 SCTP_STATIC
int sctp_msghdr_parse(const struct msghdr
*, sctp_cmsgs_t
*);
1608 SCTP_STATIC
int sctp_sendmsg(struct kiocb
*iocb
, struct sock
*sk
,
1609 struct msghdr
*msg
, size_t msg_len
)
1611 struct net
*net
= sock_net(sk
);
1612 struct sctp_sock
*sp
;
1613 struct sctp_endpoint
*ep
;
1614 struct sctp_association
*new_asoc
=NULL
, *asoc
=NULL
;
1615 struct sctp_transport
*transport
, *chunk_tp
;
1616 struct sctp_chunk
*chunk
;
1618 struct sockaddr
*msg_name
= NULL
;
1619 struct sctp_sndrcvinfo default_sinfo
;
1620 struct sctp_sndrcvinfo
*sinfo
;
1621 struct sctp_initmsg
*sinit
;
1622 sctp_assoc_t associd
= 0;
1623 sctp_cmsgs_t cmsgs
= { NULL
};
1627 __u16 sinfo_flags
= 0;
1628 struct sctp_datamsg
*datamsg
;
1629 int msg_flags
= msg
->msg_flags
;
1631 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1638 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep
);
1640 /* We cannot send a message over a TCP-style listening socket. */
1641 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
)) {
1646 /* Parse out the SCTP CMSGs. */
1647 err
= sctp_msghdr_parse(msg
, &cmsgs
);
1650 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err
);
1654 /* Fetch the destination address for this packet. This
1655 * address only selects the association--it is not necessarily
1656 * the address we will send to.
1657 * For a peeled-off socket, msg_name is ignored.
1659 if (!sctp_style(sk
, UDP_HIGH_BANDWIDTH
) && msg
->msg_name
) {
1660 int msg_namelen
= msg
->msg_namelen
;
1662 err
= sctp_verify_addr(sk
, (union sctp_addr
*)msg
->msg_name
,
1667 if (msg_namelen
> sizeof(to
))
1668 msg_namelen
= sizeof(to
);
1669 memcpy(&to
, msg
->msg_name
, msg_namelen
);
1670 msg_name
= msg
->msg_name
;
1676 /* Did the user specify SNDRCVINFO? */
1678 sinfo_flags
= sinfo
->sinfo_flags
;
1679 associd
= sinfo
->sinfo_assoc_id
;
1682 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1683 msg_len
, sinfo_flags
);
1685 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1686 if (sctp_style(sk
, TCP
) && (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
))) {
1691 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1692 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1693 * If SCTP_ABORT is set, the message length could be non zero with
1694 * the msg_iov set to the user abort reason.
1696 if (((sinfo_flags
& SCTP_EOF
) && (msg_len
> 0)) ||
1697 (!(sinfo_flags
& (SCTP_EOF
|SCTP_ABORT
)) && (msg_len
== 0))) {
1702 /* If SCTP_ADDR_OVER is set, there must be an address
1703 * specified in msg_name.
1705 if ((sinfo_flags
& SCTP_ADDR_OVER
) && (!msg
->msg_name
)) {
1712 SCTP_DEBUG_PRINTK("About to look up association.\n");
1716 /* If a msg_name has been specified, assume this is to be used. */
1718 /* Look for a matching association on the endpoint. */
1719 asoc
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1721 /* If we could not find a matching association on the
1722 * endpoint, make sure that it is not a TCP-style
1723 * socket that already has an association or there is
1724 * no peeled-off association on another socket.
1726 if ((sctp_style(sk
, TCP
) &&
1727 sctp_sstate(sk
, ESTABLISHED
)) ||
1728 sctp_endpoint_is_peeled_off(ep
, &to
)) {
1729 err
= -EADDRNOTAVAIL
;
1734 asoc
= sctp_id2assoc(sk
, associd
);
1742 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc
);
1744 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1745 * socket that has an association in CLOSED state. This can
1746 * happen when an accepted socket has an association that is
1749 if (sctp_state(asoc
, CLOSED
) && sctp_style(sk
, TCP
)) {
1754 if (sinfo_flags
& SCTP_EOF
) {
1755 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1757 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
1761 if (sinfo_flags
& SCTP_ABORT
) {
1763 chunk
= sctp_make_abort_user(asoc
, msg
, msg_len
);
1769 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc
);
1770 sctp_primitive_ABORT(net
, asoc
, chunk
);
1776 /* Do we need to create the association? */
1778 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1780 if (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
)) {
1785 /* Check for invalid stream against the stream counts,
1786 * either the default or the user specified stream counts.
1789 if (!sinit
|| (sinit
&& !sinit
->sinit_num_ostreams
)) {
1790 /* Check against the defaults. */
1791 if (sinfo
->sinfo_stream
>=
1792 sp
->initmsg
.sinit_num_ostreams
) {
1797 /* Check against the requested. */
1798 if (sinfo
->sinfo_stream
>=
1799 sinit
->sinit_num_ostreams
) {
1807 * API 3.1.2 bind() - UDP Style Syntax
1808 * If a bind() or sctp_bindx() is not called prior to a
1809 * sendmsg() call that initiates a new association, the
1810 * system picks an ephemeral port and will choose an address
1811 * set equivalent to binding with a wildcard address.
1813 if (!ep
->base
.bind_addr
.port
) {
1814 if (sctp_autobind(sk
)) {
1820 * If an unprivileged user inherits a one-to-many
1821 * style socket with open associations on a privileged
1822 * port, it MAY be permitted to accept new associations,
1823 * but it SHOULD NOT be permitted to open new
1826 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1827 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
)) {
1833 scope
= sctp_scope(&to
);
1834 new_asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1840 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
, GFP_KERNEL
);
1846 /* If the SCTP_INIT ancillary data is specified, set all
1847 * the association init values accordingly.
1850 if (sinit
->sinit_num_ostreams
) {
1851 asoc
->c
.sinit_num_ostreams
=
1852 sinit
->sinit_num_ostreams
;
1854 if (sinit
->sinit_max_instreams
) {
1855 asoc
->c
.sinit_max_instreams
=
1856 sinit
->sinit_max_instreams
;
1858 if (sinit
->sinit_max_attempts
) {
1859 asoc
->max_init_attempts
1860 = sinit
->sinit_max_attempts
;
1862 if (sinit
->sinit_max_init_timeo
) {
1863 asoc
->max_init_timeo
=
1864 msecs_to_jiffies(sinit
->sinit_max_init_timeo
);
1868 /* Prime the peer's transport structures. */
1869 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
, SCTP_UNKNOWN
);
1876 /* ASSERT: we have a valid association at this point. */
1877 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1880 /* If the user didn't specify SNDRCVINFO, make up one with
1883 memset(&default_sinfo
, 0, sizeof(default_sinfo
));
1884 default_sinfo
.sinfo_stream
= asoc
->default_stream
;
1885 default_sinfo
.sinfo_flags
= asoc
->default_flags
;
1886 default_sinfo
.sinfo_ppid
= asoc
->default_ppid
;
1887 default_sinfo
.sinfo_context
= asoc
->default_context
;
1888 default_sinfo
.sinfo_timetolive
= asoc
->default_timetolive
;
1889 default_sinfo
.sinfo_assoc_id
= sctp_assoc2id(asoc
);
1890 sinfo
= &default_sinfo
;
1893 /* API 7.1.7, the sndbuf size per association bounds the
1894 * maximum size of data that can be sent in a single send call.
1896 if (msg_len
> sk
->sk_sndbuf
) {
1901 if (asoc
->pmtu_pending
)
1902 sctp_assoc_pending_pmtu(sk
, asoc
);
1904 /* If fragmentation is disabled and the message length exceeds the
1905 * association fragmentation point, return EMSGSIZE. The I-D
1906 * does not specify what this error is, but this looks like
1909 if (sctp_sk(sk
)->disable_fragments
&& (msg_len
> asoc
->frag_point
)) {
1914 /* Check for invalid stream. */
1915 if (sinfo
->sinfo_stream
>= asoc
->c
.sinit_num_ostreams
) {
1920 timeo
= sock_sndtimeo(sk
, msg
->msg_flags
& MSG_DONTWAIT
);
1921 if (!sctp_wspace(asoc
)) {
1922 err
= sctp_wait_for_sndbuf(asoc
, &timeo
, msg_len
);
1927 /* If an address is passed with the sendto/sendmsg call, it is used
1928 * to override the primary destination address in the TCP model, or
1929 * when SCTP_ADDR_OVER flag is set in the UDP model.
1931 if ((sctp_style(sk
, TCP
) && msg_name
) ||
1932 (sinfo_flags
& SCTP_ADDR_OVER
)) {
1933 chunk_tp
= sctp_assoc_lookup_paddr(asoc
, &to
);
1941 /* Auto-connect, if we aren't connected already. */
1942 if (sctp_state(asoc
, CLOSED
)) {
1943 err
= sctp_primitive_ASSOCIATE(net
, asoc
, NULL
);
1946 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1949 /* Break the message into multiple chunks of maximum size. */
1950 datamsg
= sctp_datamsg_from_user(asoc
, sinfo
, msg
, msg_len
);
1951 if (IS_ERR(datamsg
)) {
1952 err
= PTR_ERR(datamsg
);
1956 /* Now send the (possibly) fragmented message. */
1957 list_for_each_entry(chunk
, &datamsg
->chunks
, frag_list
) {
1958 sctp_chunk_hold(chunk
);
1960 /* Do accounting for the write space. */
1961 sctp_set_owner_w(chunk
);
1963 chunk
->transport
= chunk_tp
;
1966 /* Send it to the lower layers. Note: all chunks
1967 * must either fail or succeed. The lower layer
1968 * works that way today. Keep it that way or this
1971 err
= sctp_primitive_SEND(net
, asoc
, datamsg
);
1972 /* Did the lower layer accept the chunk? */
1974 sctp_datamsg_free(datamsg
);
1976 sctp_datamsg_put(datamsg
);
1978 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1985 /* If we are already past ASSOCIATE, the lower
1986 * layers are responsible for association cleanup.
1992 sctp_unhash_established(asoc
);
1993 sctp_association_free(asoc
);
1996 sctp_release_sock(sk
);
1999 return sctp_error(sk
, msg_flags
, err
);
2006 err
= sock_error(sk
);
2016 /* This is an extended version of skb_pull() that removes the data from the
2017 * start of a skb even when data is spread across the list of skb's in the
2018 * frag_list. len specifies the total amount of data that needs to be removed.
2019 * when 'len' bytes could be removed from the skb, it returns 0.
2020 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2021 * could not be removed.
2023 static int sctp_skb_pull(struct sk_buff
*skb
, int len
)
2025 struct sk_buff
*list
;
2026 int skb_len
= skb_headlen(skb
);
2029 if (len
<= skb_len
) {
2030 __skb_pull(skb
, len
);
2034 __skb_pull(skb
, skb_len
);
2036 skb_walk_frags(skb
, list
) {
2037 rlen
= sctp_skb_pull(list
, len
);
2038 skb
->len
-= (len
-rlen
);
2039 skb
->data_len
-= (len
-rlen
);
2050 /* API 3.1.3 recvmsg() - UDP Style Syntax
2052 * ssize_t recvmsg(int socket, struct msghdr *message,
2055 * socket - the socket descriptor of the endpoint.
2056 * message - pointer to the msghdr structure which contains a single
2057 * user message and possibly some ancillary data.
2059 * See Section 5 for complete description of the data
2062 * flags - flags sent or received with the user message, see Section
2063 * 5 for complete description of the flags.
2065 static struct sk_buff
*sctp_skb_recv_datagram(struct sock
*, int, int, int *);
2067 SCTP_STATIC
int sctp_recvmsg(struct kiocb
*iocb
, struct sock
*sk
,
2068 struct msghdr
*msg
, size_t len
, int noblock
,
2069 int flags
, int *addr_len
)
2071 struct sctp_ulpevent
*event
= NULL
;
2072 struct sctp_sock
*sp
= sctp_sk(sk
);
2073 struct sk_buff
*skb
;
2078 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
2079 "0x%x, %s: %p)\n", "sk", sk
, "msghdr", msg
,
2080 "len", len
, "knoblauch", noblock
,
2081 "flags", flags
, "addr_len", addr_len
);
2085 if (sctp_style(sk
, TCP
) && !sctp_sstate(sk
, ESTABLISHED
)) {
2090 skb
= sctp_skb_recv_datagram(sk
, flags
, noblock
, &err
);
2094 /* Get the total length of the skb including any skb's in the
2103 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
2105 event
= sctp_skb2event(skb
);
2110 sock_recv_ts_and_drops(msg
, sk
, skb
);
2111 if (sctp_ulpevent_is_notification(event
)) {
2112 msg
->msg_flags
|= MSG_NOTIFICATION
;
2113 sp
->pf
->event_msgname(event
, msg
->msg_name
, addr_len
);
2115 sp
->pf
->skb_msgname(skb
, msg
->msg_name
, addr_len
);
2118 /* Check if we allow SCTP_SNDRCVINFO. */
2119 if (sp
->subscribe
.sctp_data_io_event
)
2120 sctp_ulpevent_read_sndrcvinfo(event
, msg
);
2122 /* FIXME: we should be calling IP/IPv6 layers. */
2123 if (sk
->sk_protinfo
.af_inet
.cmsg_flags
)
2124 ip_cmsg_recv(msg
, skb
);
2129 /* If skb's length exceeds the user's buffer, update the skb and
2130 * push it back to the receive_queue so that the next call to
2131 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2133 if (skb_len
> copied
) {
2134 msg
->msg_flags
&= ~MSG_EOR
;
2135 if (flags
& MSG_PEEK
)
2137 sctp_skb_pull(skb
, copied
);
2138 skb_queue_head(&sk
->sk_receive_queue
, skb
);
2140 /* When only partial message is copied to the user, increase
2141 * rwnd by that amount. If all the data in the skb is read,
2142 * rwnd is updated when the event is freed.
2144 if (!sctp_ulpevent_is_notification(event
))
2145 sctp_assoc_rwnd_increase(event
->asoc
, copied
);
2147 } else if ((event
->msg_flags
& MSG_NOTIFICATION
) ||
2148 (event
->msg_flags
& MSG_EOR
))
2149 msg
->msg_flags
|= MSG_EOR
;
2151 msg
->msg_flags
&= ~MSG_EOR
;
2154 if (flags
& MSG_PEEK
) {
2155 /* Release the skb reference acquired after peeking the skb in
2156 * sctp_skb_recv_datagram().
2160 /* Free the event which includes releasing the reference to
2161 * the owner of the skb, freeing the skb and updating the
2164 sctp_ulpevent_free(event
);
2167 sctp_release_sock(sk
);
2171 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2173 * This option is a on/off flag. If enabled no SCTP message
2174 * fragmentation will be performed. Instead if a message being sent
2175 * exceeds the current PMTU size, the message will NOT be sent and
2176 * instead a error will be indicated to the user.
2178 static int sctp_setsockopt_disable_fragments(struct sock
*sk
,
2179 char __user
*optval
,
2180 unsigned int optlen
)
2184 if (optlen
< sizeof(int))
2187 if (get_user(val
, (int __user
*)optval
))
2190 sctp_sk(sk
)->disable_fragments
= (val
== 0) ? 0 : 1;
2195 static int sctp_setsockopt_events(struct sock
*sk
, char __user
*optval
,
2196 unsigned int optlen
)
2198 struct sctp_association
*asoc
;
2199 struct sctp_ulpevent
*event
;
2201 if (optlen
> sizeof(struct sctp_event_subscribe
))
2203 if (copy_from_user(&sctp_sk(sk
)->subscribe
, optval
, optlen
))
2207 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2208 * if there is no data to be sent or retransmit, the stack will
2209 * immediately send up this notification.
2211 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT
,
2212 &sctp_sk(sk
)->subscribe
)) {
2213 asoc
= sctp_id2assoc(sk
, 0);
2215 if (asoc
&& sctp_outq_is_empty(&asoc
->outqueue
)) {
2216 event
= sctp_ulpevent_make_sender_dry_event(asoc
,
2221 sctp_ulpq_tail_event(&asoc
->ulpq
, event
);
2228 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2230 * This socket option is applicable to the UDP-style socket only. When
2231 * set it will cause associations that are idle for more than the
2232 * specified number of seconds to automatically close. An association
2233 * being idle is defined an association that has NOT sent or received
2234 * user data. The special value of '0' indicates that no automatic
2235 * close of any associations should be performed. The option expects an
2236 * integer defining the number of seconds of idle time before an
2237 * association is closed.
2239 static int sctp_setsockopt_autoclose(struct sock
*sk
, char __user
*optval
,
2240 unsigned int optlen
)
2242 struct sctp_sock
*sp
= sctp_sk(sk
);
2244 /* Applicable to UDP-style socket only */
2245 if (sctp_style(sk
, TCP
))
2247 if (optlen
!= sizeof(int))
2249 if (copy_from_user(&sp
->autoclose
, optval
, optlen
))
2255 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2257 * Applications can enable or disable heartbeats for any peer address of
2258 * an association, modify an address's heartbeat interval, force a
2259 * heartbeat to be sent immediately, and adjust the address's maximum
2260 * number of retransmissions sent before an address is considered
2261 * unreachable. The following structure is used to access and modify an
2262 * address's parameters:
2264 * struct sctp_paddrparams {
2265 * sctp_assoc_t spp_assoc_id;
2266 * struct sockaddr_storage spp_address;
2267 * uint32_t spp_hbinterval;
2268 * uint16_t spp_pathmaxrxt;
2269 * uint32_t spp_pathmtu;
2270 * uint32_t spp_sackdelay;
2271 * uint32_t spp_flags;
2274 * spp_assoc_id - (one-to-many style socket) This is filled in the
2275 * application, and identifies the association for
2277 * spp_address - This specifies which address is of interest.
2278 * spp_hbinterval - This contains the value of the heartbeat interval,
2279 * in milliseconds. If a value of zero
2280 * is present in this field then no changes are to
2281 * be made to this parameter.
2282 * spp_pathmaxrxt - This contains the maximum number of
2283 * retransmissions before this address shall be
2284 * considered unreachable. If a value of zero
2285 * is present in this field then no changes are to
2286 * be made to this parameter.
2287 * spp_pathmtu - When Path MTU discovery is disabled the value
2288 * specified here will be the "fixed" path mtu.
2289 * Note that if the spp_address field is empty
2290 * then all associations on this address will
2291 * have this fixed path mtu set upon them.
2293 * spp_sackdelay - When delayed sack is enabled, this value specifies
2294 * the number of milliseconds that sacks will be delayed
2295 * for. This value will apply to all addresses of an
2296 * association if the spp_address field is empty. Note
2297 * also, that if delayed sack is enabled and this
2298 * value is set to 0, no change is made to the last
2299 * recorded delayed sack timer value.
2301 * spp_flags - These flags are used to control various features
2302 * on an association. The flag field may contain
2303 * zero or more of the following options.
2305 * SPP_HB_ENABLE - Enable heartbeats on the
2306 * specified address. Note that if the address
2307 * field is empty all addresses for the association
2308 * have heartbeats enabled upon them.
2310 * SPP_HB_DISABLE - Disable heartbeats on the
2311 * speicifed address. Note that if the address
2312 * field is empty all addresses for the association
2313 * will have their heartbeats disabled. Note also
2314 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2315 * mutually exclusive, only one of these two should
2316 * be specified. Enabling both fields will have
2317 * undetermined results.
2319 * SPP_HB_DEMAND - Request a user initiated heartbeat
2320 * to be made immediately.
2322 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2323 * heartbeat delayis to be set to the value of 0
2326 * SPP_PMTUD_ENABLE - This field will enable PMTU
2327 * discovery upon the specified address. Note that
2328 * if the address feild is empty then all addresses
2329 * on the association are effected.
2331 * SPP_PMTUD_DISABLE - This field will disable PMTU
2332 * discovery upon the specified address. Note that
2333 * if the address feild is empty then all addresses
2334 * on the association are effected. Not also that
2335 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2336 * exclusive. Enabling both will have undetermined
2339 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2340 * on delayed sack. The time specified in spp_sackdelay
2341 * is used to specify the sack delay for this address. Note
2342 * that if spp_address is empty then all addresses will
2343 * enable delayed sack and take on the sack delay
2344 * value specified in spp_sackdelay.
2345 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2346 * off delayed sack. If the spp_address field is blank then
2347 * delayed sack is disabled for the entire association. Note
2348 * also that this field is mutually exclusive to
2349 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2352 static int sctp_apply_peer_addr_params(struct sctp_paddrparams
*params
,
2353 struct sctp_transport
*trans
,
2354 struct sctp_association
*asoc
,
2355 struct sctp_sock
*sp
,
2358 int sackdelay_change
)
2362 if (params
->spp_flags
& SPP_HB_DEMAND
&& trans
) {
2363 struct net
*net
= sock_net(trans
->asoc
->base
.sk
);
2365 error
= sctp_primitive_REQUESTHEARTBEAT(net
, trans
->asoc
, trans
);
2370 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2371 * this field is ignored. Note also that a value of zero indicates
2372 * the current setting should be left unchanged.
2374 if (params
->spp_flags
& SPP_HB_ENABLE
) {
2376 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2377 * set. This lets us use 0 value when this flag
2380 if (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)
2381 params
->spp_hbinterval
= 0;
2383 if (params
->spp_hbinterval
||
2384 (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)) {
2387 msecs_to_jiffies(params
->spp_hbinterval
);
2390 msecs_to_jiffies(params
->spp_hbinterval
);
2392 sp
->hbinterval
= params
->spp_hbinterval
;
2399 trans
->param_flags
=
2400 (trans
->param_flags
& ~SPP_HB
) | hb_change
;
2403 (asoc
->param_flags
& ~SPP_HB
) | hb_change
;
2406 (sp
->param_flags
& ~SPP_HB
) | hb_change
;
2410 /* When Path MTU discovery is disabled the value specified here will
2411 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2412 * include the flag SPP_PMTUD_DISABLE for this field to have any
2415 if ((params
->spp_flags
& SPP_PMTUD_DISABLE
) && params
->spp_pathmtu
) {
2417 trans
->pathmtu
= params
->spp_pathmtu
;
2418 sctp_assoc_sync_pmtu(sctp_opt2sk(sp
), asoc
);
2420 asoc
->pathmtu
= params
->spp_pathmtu
;
2421 sctp_frag_point(asoc
, params
->spp_pathmtu
);
2423 sp
->pathmtu
= params
->spp_pathmtu
;
2429 int update
= (trans
->param_flags
& SPP_PMTUD_DISABLE
) &&
2430 (params
->spp_flags
& SPP_PMTUD_ENABLE
);
2431 trans
->param_flags
=
2432 (trans
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2434 sctp_transport_pmtu(trans
, sctp_opt2sk(sp
));
2435 sctp_assoc_sync_pmtu(sctp_opt2sk(sp
), asoc
);
2439 (asoc
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2442 (sp
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2446 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2447 * value of this field is ignored. Note also that a value of zero
2448 * indicates the current setting should be left unchanged.
2450 if ((params
->spp_flags
& SPP_SACKDELAY_ENABLE
) && params
->spp_sackdelay
) {
2453 msecs_to_jiffies(params
->spp_sackdelay
);
2456 msecs_to_jiffies(params
->spp_sackdelay
);
2458 sp
->sackdelay
= params
->spp_sackdelay
;
2462 if (sackdelay_change
) {
2464 trans
->param_flags
=
2465 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2469 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2473 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2478 /* Note that a value of zero indicates the current setting should be
2481 if (params
->spp_pathmaxrxt
) {
2483 trans
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2485 asoc
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2487 sp
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2494 static int sctp_setsockopt_peer_addr_params(struct sock
*sk
,
2495 char __user
*optval
,
2496 unsigned int optlen
)
2498 struct sctp_paddrparams params
;
2499 struct sctp_transport
*trans
= NULL
;
2500 struct sctp_association
*asoc
= NULL
;
2501 struct sctp_sock
*sp
= sctp_sk(sk
);
2503 int hb_change
, pmtud_change
, sackdelay_change
;
2505 if (optlen
!= sizeof(struct sctp_paddrparams
))
2508 if (copy_from_user(¶ms
, optval
, optlen
))
2511 /* Validate flags and value parameters. */
2512 hb_change
= params
.spp_flags
& SPP_HB
;
2513 pmtud_change
= params
.spp_flags
& SPP_PMTUD
;
2514 sackdelay_change
= params
.spp_flags
& SPP_SACKDELAY
;
2516 if (hb_change
== SPP_HB
||
2517 pmtud_change
== SPP_PMTUD
||
2518 sackdelay_change
== SPP_SACKDELAY
||
2519 params
.spp_sackdelay
> 500 ||
2520 (params
.spp_pathmtu
&&
2521 params
.spp_pathmtu
< SCTP_DEFAULT_MINSEGMENT
))
2524 /* If an address other than INADDR_ANY is specified, and
2525 * no transport is found, then the request is invalid.
2527 if (!sctp_is_any(sk
, ( union sctp_addr
*)¶ms
.spp_address
)) {
2528 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
2529 params
.spp_assoc_id
);
2534 /* Get association, if assoc_id != 0 and the socket is a one
2535 * to many style socket, and an association was not found, then
2536 * the id was invalid.
2538 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
2539 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
))
2542 /* Heartbeat demand can only be sent on a transport or
2543 * association, but not a socket.
2545 if (params
.spp_flags
& SPP_HB_DEMAND
&& !trans
&& !asoc
)
2548 /* Process parameters. */
2549 error
= sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2550 hb_change
, pmtud_change
,
2556 /* If changes are for association, also apply parameters to each
2559 if (!trans
&& asoc
) {
2560 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2562 sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2563 hb_change
, pmtud_change
,
2572 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2574 * This option will effect the way delayed acks are performed. This
2575 * option allows you to get or set the delayed ack time, in
2576 * milliseconds. It also allows changing the delayed ack frequency.
2577 * Changing the frequency to 1 disables the delayed sack algorithm. If
2578 * the assoc_id is 0, then this sets or gets the endpoints default
2579 * values. If the assoc_id field is non-zero, then the set or get
2580 * effects the specified association for the one to many model (the
2581 * assoc_id field is ignored by the one to one model). Note that if
2582 * sack_delay or sack_freq are 0 when setting this option, then the
2583 * current values will remain unchanged.
2585 * struct sctp_sack_info {
2586 * sctp_assoc_t sack_assoc_id;
2587 * uint32_t sack_delay;
2588 * uint32_t sack_freq;
2591 * sack_assoc_id - This parameter, indicates which association the user
2592 * is performing an action upon. Note that if this field's value is
2593 * zero then the endpoints default value is changed (effecting future
2594 * associations only).
2596 * sack_delay - This parameter contains the number of milliseconds that
2597 * the user is requesting the delayed ACK timer be set to. Note that
2598 * this value is defined in the standard to be between 200 and 500
2601 * sack_freq - This parameter contains the number of packets that must
2602 * be received before a sack is sent without waiting for the delay
2603 * timer to expire. The default value for this is 2, setting this
2604 * value to 1 will disable the delayed sack algorithm.
2607 static int sctp_setsockopt_delayed_ack(struct sock
*sk
,
2608 char __user
*optval
, unsigned int optlen
)
2610 struct sctp_sack_info params
;
2611 struct sctp_transport
*trans
= NULL
;
2612 struct sctp_association
*asoc
= NULL
;
2613 struct sctp_sock
*sp
= sctp_sk(sk
);
2615 if (optlen
== sizeof(struct sctp_sack_info
)) {
2616 if (copy_from_user(¶ms
, optval
, optlen
))
2619 if (params
.sack_delay
== 0 && params
.sack_freq
== 0)
2621 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
2622 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2623 pr_warn("Use struct sctp_sack_info instead\n");
2624 if (copy_from_user(¶ms
, optval
, optlen
))
2627 if (params
.sack_delay
== 0)
2628 params
.sack_freq
= 1;
2630 params
.sack_freq
= 0;
2634 /* Validate value parameter. */
2635 if (params
.sack_delay
> 500)
2638 /* Get association, if sack_assoc_id != 0 and the socket is a one
2639 * to many style socket, and an association was not found, then
2640 * the id was invalid.
2642 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
2643 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
2646 if (params
.sack_delay
) {
2649 msecs_to_jiffies(params
.sack_delay
);
2651 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2652 SPP_SACKDELAY_ENABLE
;
2654 sp
->sackdelay
= params
.sack_delay
;
2656 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2657 SPP_SACKDELAY_ENABLE
;
2661 if (params
.sack_freq
== 1) {
2664 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2665 SPP_SACKDELAY_DISABLE
;
2668 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2669 SPP_SACKDELAY_DISABLE
;
2671 } else if (params
.sack_freq
> 1) {
2673 asoc
->sackfreq
= params
.sack_freq
;
2675 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2676 SPP_SACKDELAY_ENABLE
;
2678 sp
->sackfreq
= params
.sack_freq
;
2680 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2681 SPP_SACKDELAY_ENABLE
;
2685 /* If change is for association, also apply to each transport. */
2687 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2689 if (params
.sack_delay
) {
2691 msecs_to_jiffies(params
.sack_delay
);
2692 trans
->param_flags
=
2693 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2694 SPP_SACKDELAY_ENABLE
;
2696 if (params
.sack_freq
== 1) {
2697 trans
->param_flags
=
2698 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2699 SPP_SACKDELAY_DISABLE
;
2700 } else if (params
.sack_freq
> 1) {
2701 trans
->sackfreq
= params
.sack_freq
;
2702 trans
->param_flags
=
2703 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2704 SPP_SACKDELAY_ENABLE
;
2712 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2714 * Applications can specify protocol parameters for the default association
2715 * initialization. The option name argument to setsockopt() and getsockopt()
2718 * Setting initialization parameters is effective only on an unconnected
2719 * socket (for UDP-style sockets only future associations are effected
2720 * by the change). With TCP-style sockets, this option is inherited by
2721 * sockets derived from a listener socket.
2723 static int sctp_setsockopt_initmsg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2725 struct sctp_initmsg sinit
;
2726 struct sctp_sock
*sp
= sctp_sk(sk
);
2728 if (optlen
!= sizeof(struct sctp_initmsg
))
2730 if (copy_from_user(&sinit
, optval
, optlen
))
2733 if (sinit
.sinit_num_ostreams
)
2734 sp
->initmsg
.sinit_num_ostreams
= sinit
.sinit_num_ostreams
;
2735 if (sinit
.sinit_max_instreams
)
2736 sp
->initmsg
.sinit_max_instreams
= sinit
.sinit_max_instreams
;
2737 if (sinit
.sinit_max_attempts
)
2738 sp
->initmsg
.sinit_max_attempts
= sinit
.sinit_max_attempts
;
2739 if (sinit
.sinit_max_init_timeo
)
2740 sp
->initmsg
.sinit_max_init_timeo
= sinit
.sinit_max_init_timeo
;
2746 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2748 * Applications that wish to use the sendto() system call may wish to
2749 * specify a default set of parameters that would normally be supplied
2750 * through the inclusion of ancillary data. This socket option allows
2751 * such an application to set the default sctp_sndrcvinfo structure.
2752 * The application that wishes to use this socket option simply passes
2753 * in to this call the sctp_sndrcvinfo structure defined in Section
2754 * 5.2.2) The input parameters accepted by this call include
2755 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2756 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2757 * to this call if the caller is using the UDP model.
2759 static int sctp_setsockopt_default_send_param(struct sock
*sk
,
2760 char __user
*optval
,
2761 unsigned int optlen
)
2763 struct sctp_sndrcvinfo info
;
2764 struct sctp_association
*asoc
;
2765 struct sctp_sock
*sp
= sctp_sk(sk
);
2767 if (optlen
!= sizeof(struct sctp_sndrcvinfo
))
2769 if (copy_from_user(&info
, optval
, optlen
))
2772 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
2773 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
2777 asoc
->default_stream
= info
.sinfo_stream
;
2778 asoc
->default_flags
= info
.sinfo_flags
;
2779 asoc
->default_ppid
= info
.sinfo_ppid
;
2780 asoc
->default_context
= info
.sinfo_context
;
2781 asoc
->default_timetolive
= info
.sinfo_timetolive
;
2783 sp
->default_stream
= info
.sinfo_stream
;
2784 sp
->default_flags
= info
.sinfo_flags
;
2785 sp
->default_ppid
= info
.sinfo_ppid
;
2786 sp
->default_context
= info
.sinfo_context
;
2787 sp
->default_timetolive
= info
.sinfo_timetolive
;
2793 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2795 * Requests that the local SCTP stack use the enclosed peer address as
2796 * the association primary. The enclosed address must be one of the
2797 * association peer's addresses.
2799 static int sctp_setsockopt_primary_addr(struct sock
*sk
, char __user
*optval
,
2800 unsigned int optlen
)
2802 struct sctp_prim prim
;
2803 struct sctp_transport
*trans
;
2805 if (optlen
!= sizeof(struct sctp_prim
))
2808 if (copy_from_user(&prim
, optval
, sizeof(struct sctp_prim
)))
2811 trans
= sctp_addr_id2transport(sk
, &prim
.ssp_addr
, prim
.ssp_assoc_id
);
2815 sctp_assoc_set_primary(trans
->asoc
, trans
);
2821 * 7.1.5 SCTP_NODELAY
2823 * Turn on/off any Nagle-like algorithm. This means that packets are
2824 * generally sent as soon as possible and no unnecessary delays are
2825 * introduced, at the cost of more packets in the network. Expects an
2826 * integer boolean flag.
2828 static int sctp_setsockopt_nodelay(struct sock
*sk
, char __user
*optval
,
2829 unsigned int optlen
)
2833 if (optlen
< sizeof(int))
2835 if (get_user(val
, (int __user
*)optval
))
2838 sctp_sk(sk
)->nodelay
= (val
== 0) ? 0 : 1;
2844 * 7.1.1 SCTP_RTOINFO
2846 * The protocol parameters used to initialize and bound retransmission
2847 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2848 * and modify these parameters.
2849 * All parameters are time values, in milliseconds. A value of 0, when
2850 * modifying the parameters, indicates that the current value should not
2854 static int sctp_setsockopt_rtoinfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2856 struct sctp_rtoinfo rtoinfo
;
2857 struct sctp_association
*asoc
;
2859 if (optlen
!= sizeof (struct sctp_rtoinfo
))
2862 if (copy_from_user(&rtoinfo
, optval
, optlen
))
2865 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
2867 /* Set the values to the specific association */
2868 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
2872 if (rtoinfo
.srto_initial
!= 0)
2874 msecs_to_jiffies(rtoinfo
.srto_initial
);
2875 if (rtoinfo
.srto_max
!= 0)
2876 asoc
->rto_max
= msecs_to_jiffies(rtoinfo
.srto_max
);
2877 if (rtoinfo
.srto_min
!= 0)
2878 asoc
->rto_min
= msecs_to_jiffies(rtoinfo
.srto_min
);
2880 /* If there is no association or the association-id = 0
2881 * set the values to the endpoint.
2883 struct sctp_sock
*sp
= sctp_sk(sk
);
2885 if (rtoinfo
.srto_initial
!= 0)
2886 sp
->rtoinfo
.srto_initial
= rtoinfo
.srto_initial
;
2887 if (rtoinfo
.srto_max
!= 0)
2888 sp
->rtoinfo
.srto_max
= rtoinfo
.srto_max
;
2889 if (rtoinfo
.srto_min
!= 0)
2890 sp
->rtoinfo
.srto_min
= rtoinfo
.srto_min
;
2898 * 7.1.2 SCTP_ASSOCINFO
2900 * This option is used to tune the maximum retransmission attempts
2901 * of the association.
2902 * Returns an error if the new association retransmission value is
2903 * greater than the sum of the retransmission value of the peer.
2904 * See [SCTP] for more information.
2907 static int sctp_setsockopt_associnfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2910 struct sctp_assocparams assocparams
;
2911 struct sctp_association
*asoc
;
2913 if (optlen
!= sizeof(struct sctp_assocparams
))
2915 if (copy_from_user(&assocparams
, optval
, optlen
))
2918 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
2920 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
2923 /* Set the values to the specific association */
2925 if (assocparams
.sasoc_asocmaxrxt
!= 0) {
2928 struct sctp_transport
*peer_addr
;
2930 list_for_each_entry(peer_addr
, &asoc
->peer
.transport_addr_list
,
2932 path_sum
+= peer_addr
->pathmaxrxt
;
2936 /* Only validate asocmaxrxt if we have more than
2937 * one path/transport. We do this because path
2938 * retransmissions are only counted when we have more
2942 assocparams
.sasoc_asocmaxrxt
> path_sum
)
2945 asoc
->max_retrans
= assocparams
.sasoc_asocmaxrxt
;
2948 if (assocparams
.sasoc_cookie_life
!= 0) {
2949 asoc
->cookie_life
.tv_sec
=
2950 assocparams
.sasoc_cookie_life
/ 1000;
2951 asoc
->cookie_life
.tv_usec
=
2952 (assocparams
.sasoc_cookie_life
% 1000)
2956 /* Set the values to the endpoint */
2957 struct sctp_sock
*sp
= sctp_sk(sk
);
2959 if (assocparams
.sasoc_asocmaxrxt
!= 0)
2960 sp
->assocparams
.sasoc_asocmaxrxt
=
2961 assocparams
.sasoc_asocmaxrxt
;
2962 if (assocparams
.sasoc_cookie_life
!= 0)
2963 sp
->assocparams
.sasoc_cookie_life
=
2964 assocparams
.sasoc_cookie_life
;
2970 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2972 * This socket option is a boolean flag which turns on or off mapped V4
2973 * addresses. If this option is turned on and the socket is type
2974 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2975 * If this option is turned off, then no mapping will be done of V4
2976 * addresses and a user will receive both PF_INET6 and PF_INET type
2977 * addresses on the socket.
2979 static int sctp_setsockopt_mappedv4(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2982 struct sctp_sock
*sp
= sctp_sk(sk
);
2984 if (optlen
< sizeof(int))
2986 if (get_user(val
, (int __user
*)optval
))
2997 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2998 * This option will get or set the maximum size to put in any outgoing
2999 * SCTP DATA chunk. If a message is larger than this size it will be
3000 * fragmented by SCTP into the specified size. Note that the underlying
3001 * SCTP implementation may fragment into smaller sized chunks when the
3002 * PMTU of the underlying association is smaller than the value set by
3003 * the user. The default value for this option is '0' which indicates
3004 * the user is NOT limiting fragmentation and only the PMTU will effect
3005 * SCTP's choice of DATA chunk size. Note also that values set larger
3006 * than the maximum size of an IP datagram will effectively let SCTP
3007 * control fragmentation (i.e. the same as setting this option to 0).
3009 * The following structure is used to access and modify this parameter:
3011 * struct sctp_assoc_value {
3012 * sctp_assoc_t assoc_id;
3013 * uint32_t assoc_value;
3016 * assoc_id: This parameter is ignored for one-to-one style sockets.
3017 * For one-to-many style sockets this parameter indicates which
3018 * association the user is performing an action upon. Note that if
3019 * this field's value is zero then the endpoints default value is
3020 * changed (effecting future associations only).
3021 * assoc_value: This parameter specifies the maximum size in bytes.
3023 static int sctp_setsockopt_maxseg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
3025 struct sctp_assoc_value params
;
3026 struct sctp_association
*asoc
;
3027 struct sctp_sock
*sp
= sctp_sk(sk
);
3030 if (optlen
== sizeof(int)) {
3031 pr_warn("Use of int in maxseg socket option deprecated\n");
3032 pr_warn("Use struct sctp_assoc_value instead\n");
3033 if (copy_from_user(&val
, optval
, optlen
))
3035 params
.assoc_id
= 0;
3036 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
3037 if (copy_from_user(¶ms
, optval
, optlen
))
3039 val
= params
.assoc_value
;
3043 if ((val
!= 0) && ((val
< 8) || (val
> SCTP_MAX_CHUNK_LEN
)))
3046 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3047 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
3052 val
= asoc
->pathmtu
;
3053 val
-= sp
->pf
->af
->net_header_len
;
3054 val
-= sizeof(struct sctphdr
) +
3055 sizeof(struct sctp_data_chunk
);
3057 asoc
->user_frag
= val
;
3058 asoc
->frag_point
= sctp_frag_point(asoc
, asoc
->pathmtu
);
3060 sp
->user_frag
= val
;
3068 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3070 * Requests that the peer mark the enclosed address as the association
3071 * primary. The enclosed address must be one of the association's
3072 * locally bound addresses. The following structure is used to make a
3073 * set primary request:
3075 static int sctp_setsockopt_peer_primary_addr(struct sock
*sk
, char __user
*optval
,
3076 unsigned int optlen
)
3078 struct net
*net
= sock_net(sk
);
3079 struct sctp_sock
*sp
;
3080 struct sctp_association
*asoc
= NULL
;
3081 struct sctp_setpeerprim prim
;
3082 struct sctp_chunk
*chunk
;
3088 if (!net
->sctp
.addip_enable
)
3091 if (optlen
!= sizeof(struct sctp_setpeerprim
))
3094 if (copy_from_user(&prim
, optval
, optlen
))
3097 asoc
= sctp_id2assoc(sk
, prim
.sspp_assoc_id
);
3101 if (!asoc
->peer
.asconf_capable
)
3104 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_SET_PRIMARY
)
3107 if (!sctp_state(asoc
, ESTABLISHED
))
3110 af
= sctp_get_af_specific(prim
.sspp_addr
.ss_family
);
3114 if (!af
->addr_valid((union sctp_addr
*)&prim
.sspp_addr
, sp
, NULL
))
3115 return -EADDRNOTAVAIL
;
3117 if (!sctp_assoc_lookup_laddr(asoc
, (union sctp_addr
*)&prim
.sspp_addr
))
3118 return -EADDRNOTAVAIL
;
3120 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3121 chunk
= sctp_make_asconf_set_prim(asoc
,
3122 (union sctp_addr
*)&prim
.sspp_addr
);
3126 err
= sctp_send_asconf(asoc
, chunk
);
3128 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
3133 static int sctp_setsockopt_adaptation_layer(struct sock
*sk
, char __user
*optval
,
3134 unsigned int optlen
)
3136 struct sctp_setadaptation adaptation
;
3138 if (optlen
!= sizeof(struct sctp_setadaptation
))
3140 if (copy_from_user(&adaptation
, optval
, optlen
))
3143 sctp_sk(sk
)->adaptation_ind
= adaptation
.ssb_adaptation_ind
;
3149 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3151 * The context field in the sctp_sndrcvinfo structure is normally only
3152 * used when a failed message is retrieved holding the value that was
3153 * sent down on the actual send call. This option allows the setting of
3154 * a default context on an association basis that will be received on
3155 * reading messages from the peer. This is especially helpful in the
3156 * one-2-many model for an application to keep some reference to an
3157 * internal state machine that is processing messages on the
3158 * association. Note that the setting of this value only effects
3159 * received messages from the peer and does not effect the value that is
3160 * saved with outbound messages.
3162 static int sctp_setsockopt_context(struct sock
*sk
, char __user
*optval
,
3163 unsigned int optlen
)
3165 struct sctp_assoc_value params
;
3166 struct sctp_sock
*sp
;
3167 struct sctp_association
*asoc
;
3169 if (optlen
!= sizeof(struct sctp_assoc_value
))
3171 if (copy_from_user(¶ms
, optval
, optlen
))
3176 if (params
.assoc_id
!= 0) {
3177 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3180 asoc
->default_rcv_context
= params
.assoc_value
;
3182 sp
->default_rcv_context
= params
.assoc_value
;
3189 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3191 * This options will at a minimum specify if the implementation is doing
3192 * fragmented interleave. Fragmented interleave, for a one to many
3193 * socket, is when subsequent calls to receive a message may return
3194 * parts of messages from different associations. Some implementations
3195 * may allow you to turn this value on or off. If so, when turned off,
3196 * no fragment interleave will occur (which will cause a head of line
3197 * blocking amongst multiple associations sharing the same one to many
3198 * socket). When this option is turned on, then each receive call may
3199 * come from a different association (thus the user must receive data
3200 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3201 * association each receive belongs to.
3203 * This option takes a boolean value. A non-zero value indicates that
3204 * fragmented interleave is on. A value of zero indicates that
3205 * fragmented interleave is off.
3207 * Note that it is important that an implementation that allows this
3208 * option to be turned on, have it off by default. Otherwise an unaware
3209 * application using the one to many model may become confused and act
3212 static int sctp_setsockopt_fragment_interleave(struct sock
*sk
,
3213 char __user
*optval
,
3214 unsigned int optlen
)
3218 if (optlen
!= sizeof(int))
3220 if (get_user(val
, (int __user
*)optval
))
3223 sctp_sk(sk
)->frag_interleave
= (val
== 0) ? 0 : 1;
3229 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3230 * (SCTP_PARTIAL_DELIVERY_POINT)
3232 * This option will set or get the SCTP partial delivery point. This
3233 * point is the size of a message where the partial delivery API will be
3234 * invoked to help free up rwnd space for the peer. Setting this to a
3235 * lower value will cause partial deliveries to happen more often. The
3236 * calls argument is an integer that sets or gets the partial delivery
3237 * point. Note also that the call will fail if the user attempts to set
3238 * this value larger than the socket receive buffer size.
3240 * Note that any single message having a length smaller than or equal to
3241 * the SCTP partial delivery point will be delivered in one single read
3242 * call as long as the user provided buffer is large enough to hold the
3245 static int sctp_setsockopt_partial_delivery_point(struct sock
*sk
,
3246 char __user
*optval
,
3247 unsigned int optlen
)
3251 if (optlen
!= sizeof(u32
))
3253 if (get_user(val
, (int __user
*)optval
))
3256 /* Note: We double the receive buffer from what the user sets
3257 * it to be, also initial rwnd is based on rcvbuf/2.
3259 if (val
> (sk
->sk_rcvbuf
>> 1))
3262 sctp_sk(sk
)->pd_point
= val
;
3264 return 0; /* is this the right error code? */
3268 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3270 * This option will allow a user to change the maximum burst of packets
3271 * that can be emitted by this association. Note that the default value
3272 * is 4, and some implementations may restrict this setting so that it
3273 * can only be lowered.
3275 * NOTE: This text doesn't seem right. Do this on a socket basis with
3276 * future associations inheriting the socket value.
3278 static int sctp_setsockopt_maxburst(struct sock
*sk
,
3279 char __user
*optval
,
3280 unsigned int optlen
)
3282 struct sctp_assoc_value params
;
3283 struct sctp_sock
*sp
;
3284 struct sctp_association
*asoc
;
3288 if (optlen
== sizeof(int)) {
3289 pr_warn("Use of int in max_burst socket option deprecated\n");
3290 pr_warn("Use struct sctp_assoc_value instead\n");
3291 if (copy_from_user(&val
, optval
, optlen
))
3293 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
3294 if (copy_from_user(¶ms
, optval
, optlen
))
3296 val
= params
.assoc_value
;
3297 assoc_id
= params
.assoc_id
;
3303 if (assoc_id
!= 0) {
3304 asoc
= sctp_id2assoc(sk
, assoc_id
);
3307 asoc
->max_burst
= val
;
3309 sp
->max_burst
= val
;
3315 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3317 * This set option adds a chunk type that the user is requesting to be
3318 * received only in an authenticated way. Changes to the list of chunks
3319 * will only effect future associations on the socket.
3321 static int sctp_setsockopt_auth_chunk(struct sock
*sk
,
3322 char __user
*optval
,
3323 unsigned int optlen
)
3325 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3326 struct sctp_authchunk val
;
3328 if (!ep
->auth_enable
)
3331 if (optlen
!= sizeof(struct sctp_authchunk
))
3333 if (copy_from_user(&val
, optval
, optlen
))
3336 switch (val
.sauth_chunk
) {
3338 case SCTP_CID_INIT_ACK
:
3339 case SCTP_CID_SHUTDOWN_COMPLETE
:
3344 /* add this chunk id to the endpoint */
3345 return sctp_auth_ep_add_chunkid(ep
, val
.sauth_chunk
);
3349 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3351 * This option gets or sets the list of HMAC algorithms that the local
3352 * endpoint requires the peer to use.
3354 static int sctp_setsockopt_hmac_ident(struct sock
*sk
,
3355 char __user
*optval
,
3356 unsigned int optlen
)
3358 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3359 struct sctp_hmacalgo
*hmacs
;
3363 if (!ep
->auth_enable
)
3366 if (optlen
< sizeof(struct sctp_hmacalgo
))
3369 hmacs
= memdup_user(optval
, optlen
);
3371 return PTR_ERR(hmacs
);
3373 idents
= hmacs
->shmac_num_idents
;
3374 if (idents
== 0 || idents
> SCTP_AUTH_NUM_HMACS
||
3375 (idents
* sizeof(u16
)) > (optlen
- sizeof(struct sctp_hmacalgo
))) {
3380 err
= sctp_auth_ep_set_hmacs(ep
, hmacs
);
3387 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3389 * This option will set a shared secret key which is used to build an
3390 * association shared key.
3392 static int sctp_setsockopt_auth_key(struct sock
*sk
,
3393 char __user
*optval
,
3394 unsigned int optlen
)
3396 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3397 struct sctp_authkey
*authkey
;
3398 struct sctp_association
*asoc
;
3401 if (!ep
->auth_enable
)
3404 if (optlen
<= sizeof(struct sctp_authkey
))
3407 authkey
= memdup_user(optval
, optlen
);
3408 if (IS_ERR(authkey
))
3409 return PTR_ERR(authkey
);
3411 if (authkey
->sca_keylength
> optlen
- sizeof(struct sctp_authkey
)) {
3416 asoc
= sctp_id2assoc(sk
, authkey
->sca_assoc_id
);
3417 if (!asoc
&& authkey
->sca_assoc_id
&& sctp_style(sk
, UDP
)) {
3422 ret
= sctp_auth_set_key(ep
, asoc
, authkey
);
3429 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3431 * This option will get or set the active shared key to be used to build
3432 * the association shared key.
3434 static int sctp_setsockopt_active_key(struct sock
*sk
,
3435 char __user
*optval
,
3436 unsigned int optlen
)
3438 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3439 struct sctp_authkeyid val
;
3440 struct sctp_association
*asoc
;
3442 if (!ep
->auth_enable
)
3445 if (optlen
!= sizeof(struct sctp_authkeyid
))
3447 if (copy_from_user(&val
, optval
, optlen
))
3450 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3451 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3454 return sctp_auth_set_active_key(ep
, asoc
, val
.scact_keynumber
);
3458 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3460 * This set option will delete a shared secret key from use.
3462 static int sctp_setsockopt_del_key(struct sock
*sk
,
3463 char __user
*optval
,
3464 unsigned int optlen
)
3466 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3467 struct sctp_authkeyid val
;
3468 struct sctp_association
*asoc
;
3470 if (!ep
->auth_enable
)
3473 if (optlen
!= sizeof(struct sctp_authkeyid
))
3475 if (copy_from_user(&val
, optval
, optlen
))
3478 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3479 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3482 return sctp_auth_del_key_id(ep
, asoc
, val
.scact_keynumber
);
3487 * 8.1.23 SCTP_AUTO_ASCONF
3489 * This option will enable or disable the use of the automatic generation of
3490 * ASCONF chunks to add and delete addresses to an existing association. Note
3491 * that this option has two caveats namely: a) it only affects sockets that
3492 * are bound to all addresses available to the SCTP stack, and b) the system
3493 * administrator may have an overriding control that turns the ASCONF feature
3494 * off no matter what setting the socket option may have.
3495 * This option expects an integer boolean flag, where a non-zero value turns on
3496 * the option, and a zero value turns off the option.
3497 * Note. In this implementation, socket operation overrides default parameter
3498 * being set by sysctl as well as FreeBSD implementation
3500 static int sctp_setsockopt_auto_asconf(struct sock
*sk
, char __user
*optval
,
3501 unsigned int optlen
)
3504 struct sctp_sock
*sp
= sctp_sk(sk
);
3506 if (optlen
< sizeof(int))
3508 if (get_user(val
, (int __user
*)optval
))
3510 if (!sctp_is_ep_boundall(sk
) && val
)
3512 if ((val
&& sp
->do_auto_asconf
) || (!val
&& !sp
->do_auto_asconf
))
3515 spin_lock_bh(&sock_net(sk
)->sctp
.addr_wq_lock
);
3516 if (val
== 0 && sp
->do_auto_asconf
) {
3517 list_del(&sp
->auto_asconf_list
);
3518 sp
->do_auto_asconf
= 0;
3519 } else if (val
&& !sp
->do_auto_asconf
) {
3520 list_add_tail(&sp
->auto_asconf_list
,
3521 &sock_net(sk
)->sctp
.auto_asconf_splist
);
3522 sp
->do_auto_asconf
= 1;
3524 spin_unlock_bh(&sock_net(sk
)->sctp
.addr_wq_lock
);
3530 * SCTP_PEER_ADDR_THLDS
3532 * This option allows us to alter the partially failed threshold for one or all
3533 * transports in an association. See Section 6.1 of:
3534 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3536 static int sctp_setsockopt_paddr_thresholds(struct sock
*sk
,
3537 char __user
*optval
,
3538 unsigned int optlen
)
3540 struct sctp_paddrthlds val
;
3541 struct sctp_transport
*trans
;
3542 struct sctp_association
*asoc
;
3544 if (optlen
< sizeof(struct sctp_paddrthlds
))
3546 if (copy_from_user(&val
, (struct sctp_paddrthlds __user
*)optval
,
3547 sizeof(struct sctp_paddrthlds
)))
3551 if (sctp_is_any(sk
, (const union sctp_addr
*)&val
.spt_address
)) {
3552 asoc
= sctp_id2assoc(sk
, val
.spt_assoc_id
);
3555 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
3557 if (val
.spt_pathmaxrxt
)
3558 trans
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3559 trans
->pf_retrans
= val
.spt_pathpfthld
;
3562 if (val
.spt_pathmaxrxt
)
3563 asoc
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3564 asoc
->pf_retrans
= val
.spt_pathpfthld
;
3566 trans
= sctp_addr_id2transport(sk
, &val
.spt_address
,
3571 if (val
.spt_pathmaxrxt
)
3572 trans
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3573 trans
->pf_retrans
= val
.spt_pathpfthld
;
3579 /* API 6.2 setsockopt(), getsockopt()
3581 * Applications use setsockopt() and getsockopt() to set or retrieve
3582 * socket options. Socket options are used to change the default
3583 * behavior of sockets calls. They are described in Section 7.
3587 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3588 * int __user *optlen);
3589 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3592 * sd - the socket descript.
3593 * level - set to IPPROTO_SCTP for all SCTP options.
3594 * optname - the option name.
3595 * optval - the buffer to store the value of the option.
3596 * optlen - the size of the buffer.
3598 SCTP_STATIC
int sctp_setsockopt(struct sock
*sk
, int level
, int optname
,
3599 char __user
*optval
, unsigned int optlen
)
3603 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3606 /* I can hardly begin to describe how wrong this is. This is
3607 * so broken as to be worse than useless. The API draft
3608 * REALLY is NOT helpful here... I am not convinced that the
3609 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3610 * are at all well-founded.
3612 if (level
!= SOL_SCTP
) {
3613 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
3614 retval
= af
->setsockopt(sk
, level
, optname
, optval
, optlen
);
3621 case SCTP_SOCKOPT_BINDX_ADD
:
3622 /* 'optlen' is the size of the addresses buffer. */
3623 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3624 optlen
, SCTP_BINDX_ADD_ADDR
);
3627 case SCTP_SOCKOPT_BINDX_REM
:
3628 /* 'optlen' is the size of the addresses buffer. */
3629 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3630 optlen
, SCTP_BINDX_REM_ADDR
);
3633 case SCTP_SOCKOPT_CONNECTX_OLD
:
3634 /* 'optlen' is the size of the addresses buffer. */
3635 retval
= sctp_setsockopt_connectx_old(sk
,
3636 (struct sockaddr __user
*)optval
,
3640 case SCTP_SOCKOPT_CONNECTX
:
3641 /* 'optlen' is the size of the addresses buffer. */
3642 retval
= sctp_setsockopt_connectx(sk
,
3643 (struct sockaddr __user
*)optval
,
3647 case SCTP_DISABLE_FRAGMENTS
:
3648 retval
= sctp_setsockopt_disable_fragments(sk
, optval
, optlen
);
3652 retval
= sctp_setsockopt_events(sk
, optval
, optlen
);
3655 case SCTP_AUTOCLOSE
:
3656 retval
= sctp_setsockopt_autoclose(sk
, optval
, optlen
);
3659 case SCTP_PEER_ADDR_PARAMS
:
3660 retval
= sctp_setsockopt_peer_addr_params(sk
, optval
, optlen
);
3663 case SCTP_DELAYED_SACK
:
3664 retval
= sctp_setsockopt_delayed_ack(sk
, optval
, optlen
);
3666 case SCTP_PARTIAL_DELIVERY_POINT
:
3667 retval
= sctp_setsockopt_partial_delivery_point(sk
, optval
, optlen
);
3671 retval
= sctp_setsockopt_initmsg(sk
, optval
, optlen
);
3673 case SCTP_DEFAULT_SEND_PARAM
:
3674 retval
= sctp_setsockopt_default_send_param(sk
, optval
,
3677 case SCTP_PRIMARY_ADDR
:
3678 retval
= sctp_setsockopt_primary_addr(sk
, optval
, optlen
);
3680 case SCTP_SET_PEER_PRIMARY_ADDR
:
3681 retval
= sctp_setsockopt_peer_primary_addr(sk
, optval
, optlen
);
3684 retval
= sctp_setsockopt_nodelay(sk
, optval
, optlen
);
3687 retval
= sctp_setsockopt_rtoinfo(sk
, optval
, optlen
);
3689 case SCTP_ASSOCINFO
:
3690 retval
= sctp_setsockopt_associnfo(sk
, optval
, optlen
);
3692 case SCTP_I_WANT_MAPPED_V4_ADDR
:
3693 retval
= sctp_setsockopt_mappedv4(sk
, optval
, optlen
);
3696 retval
= sctp_setsockopt_maxseg(sk
, optval
, optlen
);
3698 case SCTP_ADAPTATION_LAYER
:
3699 retval
= sctp_setsockopt_adaptation_layer(sk
, optval
, optlen
);
3702 retval
= sctp_setsockopt_context(sk
, optval
, optlen
);
3704 case SCTP_FRAGMENT_INTERLEAVE
:
3705 retval
= sctp_setsockopt_fragment_interleave(sk
, optval
, optlen
);
3707 case SCTP_MAX_BURST
:
3708 retval
= sctp_setsockopt_maxburst(sk
, optval
, optlen
);
3710 case SCTP_AUTH_CHUNK
:
3711 retval
= sctp_setsockopt_auth_chunk(sk
, optval
, optlen
);
3713 case SCTP_HMAC_IDENT
:
3714 retval
= sctp_setsockopt_hmac_ident(sk
, optval
, optlen
);
3717 retval
= sctp_setsockopt_auth_key(sk
, optval
, optlen
);
3719 case SCTP_AUTH_ACTIVE_KEY
:
3720 retval
= sctp_setsockopt_active_key(sk
, optval
, optlen
);
3722 case SCTP_AUTH_DELETE_KEY
:
3723 retval
= sctp_setsockopt_del_key(sk
, optval
, optlen
);
3725 case SCTP_AUTO_ASCONF
:
3726 retval
= sctp_setsockopt_auto_asconf(sk
, optval
, optlen
);
3728 case SCTP_PEER_ADDR_THLDS
:
3729 retval
= sctp_setsockopt_paddr_thresholds(sk
, optval
, optlen
);
3732 retval
= -ENOPROTOOPT
;
3736 sctp_release_sock(sk
);
3742 /* API 3.1.6 connect() - UDP Style Syntax
3744 * An application may use the connect() call in the UDP model to initiate an
3745 * association without sending data.
3749 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3751 * sd: the socket descriptor to have a new association added to.
3753 * nam: the address structure (either struct sockaddr_in or struct
3754 * sockaddr_in6 defined in RFC2553 [7]).
3756 * len: the size of the address.
3758 SCTP_STATIC
int sctp_connect(struct sock
*sk
, struct sockaddr
*addr
,
3766 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3767 __func__
, sk
, addr
, addr_len
);
3769 /* Validate addr_len before calling common connect/connectx routine. */
3770 af
= sctp_get_af_specific(addr
->sa_family
);
3771 if (!af
|| addr_len
< af
->sockaddr_len
) {
3774 /* Pass correct addr len to common routine (so it knows there
3775 * is only one address being passed.
3777 err
= __sctp_connect(sk
, addr
, af
->sockaddr_len
, NULL
);
3780 sctp_release_sock(sk
);
3784 /* FIXME: Write comments. */
3785 SCTP_STATIC
int sctp_disconnect(struct sock
*sk
, int flags
)
3787 return -EOPNOTSUPP
; /* STUB */
3790 /* 4.1.4 accept() - TCP Style Syntax
3792 * Applications use accept() call to remove an established SCTP
3793 * association from the accept queue of the endpoint. A new socket
3794 * descriptor will be returned from accept() to represent the newly
3795 * formed association.
3797 SCTP_STATIC
struct sock
*sctp_accept(struct sock
*sk
, int flags
, int *err
)
3799 struct sctp_sock
*sp
;
3800 struct sctp_endpoint
*ep
;
3801 struct sock
*newsk
= NULL
;
3802 struct sctp_association
*asoc
;
3811 if (!sctp_style(sk
, TCP
)) {
3812 error
= -EOPNOTSUPP
;
3816 if (!sctp_sstate(sk
, LISTENING
)) {
3821 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
3823 error
= sctp_wait_for_accept(sk
, timeo
);
3827 /* We treat the list of associations on the endpoint as the accept
3828 * queue and pick the first association on the list.
3830 asoc
= list_entry(ep
->asocs
.next
, struct sctp_association
, asocs
);
3832 newsk
= sp
->pf
->create_accept_sk(sk
, asoc
);
3838 /* Populate the fields of the newsk from the oldsk and migrate the
3839 * asoc to the newsk.
3841 sctp_sock_migrate(sk
, newsk
, asoc
, SCTP_SOCKET_TCP
);
3844 sctp_release_sock(sk
);
3849 /* The SCTP ioctl handler. */
3850 SCTP_STATIC
int sctp_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
3857 * SEQPACKET-style sockets in LISTENING state are valid, for
3858 * SCTP, so only discard TCP-style sockets in LISTENING state.
3860 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
3865 struct sk_buff
*skb
;
3866 unsigned int amount
= 0;
3868 skb
= skb_peek(&sk
->sk_receive_queue
);
3871 * We will only return the amount of this packet since
3872 * that is all that will be read.
3876 rc
= put_user(amount
, (int __user
*)arg
);
3884 sctp_release_sock(sk
);
3888 /* This is the function which gets called during socket creation to
3889 * initialized the SCTP-specific portion of the sock.
3890 * The sock structure should already be zero-filled memory.
3892 SCTP_STATIC
int sctp_init_sock(struct sock
*sk
)
3894 struct net
*net
= sock_net(sk
);
3895 struct sctp_endpoint
*ep
;
3896 struct sctp_sock
*sp
;
3898 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk
);
3902 /* Initialize the SCTP per socket area. */
3903 switch (sk
->sk_type
) {
3904 case SOCK_SEQPACKET
:
3905 sp
->type
= SCTP_SOCKET_UDP
;
3908 sp
->type
= SCTP_SOCKET_TCP
;
3911 return -ESOCKTNOSUPPORT
;
3914 /* Initialize default send parameters. These parameters can be
3915 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3917 sp
->default_stream
= 0;
3918 sp
->default_ppid
= 0;
3919 sp
->default_flags
= 0;
3920 sp
->default_context
= 0;
3921 sp
->default_timetolive
= 0;
3923 sp
->default_rcv_context
= 0;
3924 sp
->max_burst
= net
->sctp
.max_burst
;
3926 sp
->sctp_hmac_alg
= net
->sctp
.sctp_hmac_alg
;
3928 /* Initialize default setup parameters. These parameters
3929 * can be modified with the SCTP_INITMSG socket option or
3930 * overridden by the SCTP_INIT CMSG.
3932 sp
->initmsg
.sinit_num_ostreams
= sctp_max_outstreams
;
3933 sp
->initmsg
.sinit_max_instreams
= sctp_max_instreams
;
3934 sp
->initmsg
.sinit_max_attempts
= net
->sctp
.max_retrans_init
;
3935 sp
->initmsg
.sinit_max_init_timeo
= net
->sctp
.rto_max
;
3937 /* Initialize default RTO related parameters. These parameters can
3938 * be modified for with the SCTP_RTOINFO socket option.
3940 sp
->rtoinfo
.srto_initial
= net
->sctp
.rto_initial
;
3941 sp
->rtoinfo
.srto_max
= net
->sctp
.rto_max
;
3942 sp
->rtoinfo
.srto_min
= net
->sctp
.rto_min
;
3944 /* Initialize default association related parameters. These parameters
3945 * can be modified with the SCTP_ASSOCINFO socket option.
3947 sp
->assocparams
.sasoc_asocmaxrxt
= net
->sctp
.max_retrans_association
;
3948 sp
->assocparams
.sasoc_number_peer_destinations
= 0;
3949 sp
->assocparams
.sasoc_peer_rwnd
= 0;
3950 sp
->assocparams
.sasoc_local_rwnd
= 0;
3951 sp
->assocparams
.sasoc_cookie_life
= net
->sctp
.valid_cookie_life
;
3953 /* Initialize default event subscriptions. By default, all the
3956 memset(&sp
->subscribe
, 0, sizeof(struct sctp_event_subscribe
));
3958 /* Default Peer Address Parameters. These defaults can
3959 * be modified via SCTP_PEER_ADDR_PARAMS
3961 sp
->hbinterval
= net
->sctp
.hb_interval
;
3962 sp
->pathmaxrxt
= net
->sctp
.max_retrans_path
;
3963 sp
->pathmtu
= 0; // allow default discovery
3964 sp
->sackdelay
= net
->sctp
.sack_timeout
;
3966 sp
->param_flags
= SPP_HB_ENABLE
|
3968 SPP_SACKDELAY_ENABLE
;
3970 /* If enabled no SCTP message fragmentation will be performed.
3971 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3973 sp
->disable_fragments
= 0;
3975 /* Enable Nagle algorithm by default. */
3978 /* Enable by default. */
3981 /* Auto-close idle associations after the configured
3982 * number of seconds. A value of 0 disables this
3983 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3984 * for UDP-style sockets only.
3988 /* User specified fragmentation limit. */
3991 sp
->adaptation_ind
= 0;
3993 sp
->pf
= sctp_get_pf_specific(sk
->sk_family
);
3995 /* Control variables for partial data delivery. */
3996 atomic_set(&sp
->pd_mode
, 0);
3997 skb_queue_head_init(&sp
->pd_lobby
);
3998 sp
->frag_interleave
= 0;
4000 /* Create a per socket endpoint structure. Even if we
4001 * change the data structure relationships, this may still
4002 * be useful for storing pre-connect address information.
4004 ep
= sctp_endpoint_new(sk
, GFP_KERNEL
);
4011 SCTP_DBG_OBJCNT_INC(sock
);
4014 percpu_counter_inc(&sctp_sockets_allocated
);
4015 sock_prot_inuse_add(net
, sk
->sk_prot
, 1);
4017 /* Nothing can fail after this block, otherwise
4018 * sctp_destroy_sock() will be called without addr_wq_lock held
4020 if (net
->sctp
.default_auto_asconf
) {
4021 spin_lock(&sock_net(sk
)->sctp
.addr_wq_lock
);
4022 list_add_tail(&sp
->auto_asconf_list
,
4023 &net
->sctp
.auto_asconf_splist
);
4024 sp
->do_auto_asconf
= 1;
4025 spin_unlock(&sock_net(sk
)->sctp
.addr_wq_lock
);
4027 sp
->do_auto_asconf
= 0;
4035 /* Cleanup any SCTP per socket resources. Must be called with
4036 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4038 SCTP_STATIC
void sctp_destroy_sock(struct sock
*sk
)
4040 struct sctp_sock
*sp
;
4042 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk
);
4044 /* Release our hold on the endpoint. */
4046 /* This could happen during socket init, thus we bail out
4047 * early, since the rest of the below is not setup either.
4052 if (sp
->do_auto_asconf
) {
4053 sp
->do_auto_asconf
= 0;
4054 list_del(&sp
->auto_asconf_list
);
4056 sctp_endpoint_free(sp
->ep
);
4058 percpu_counter_dec(&sctp_sockets_allocated
);
4059 sock_prot_inuse_add(sock_net(sk
), sk
->sk_prot
, -1);
4063 /* API 4.1.7 shutdown() - TCP Style Syntax
4064 * int shutdown(int socket, int how);
4066 * sd - the socket descriptor of the association to be closed.
4067 * how - Specifies the type of shutdown. The values are
4070 * Disables further receive operations. No SCTP
4071 * protocol action is taken.
4073 * Disables further send operations, and initiates
4074 * the SCTP shutdown sequence.
4076 * Disables further send and receive operations
4077 * and initiates the SCTP shutdown sequence.
4079 SCTP_STATIC
void sctp_shutdown(struct sock
*sk
, int how
)
4081 struct net
*net
= sock_net(sk
);
4082 struct sctp_endpoint
*ep
;
4083 struct sctp_association
*asoc
;
4085 if (!sctp_style(sk
, TCP
))
4088 if (how
& SEND_SHUTDOWN
) {
4089 ep
= sctp_sk(sk
)->ep
;
4090 if (!list_empty(&ep
->asocs
)) {
4091 asoc
= list_entry(ep
->asocs
.next
,
4092 struct sctp_association
, asocs
);
4093 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
4098 /* 7.2.1 Association Status (SCTP_STATUS)
4100 * Applications can retrieve current status information about an
4101 * association, including association state, peer receiver window size,
4102 * number of unacked data chunks, and number of data chunks pending
4103 * receipt. This information is read-only.
4105 static int sctp_getsockopt_sctp_status(struct sock
*sk
, int len
,
4106 char __user
*optval
,
4109 struct sctp_status status
;
4110 struct sctp_association
*asoc
= NULL
;
4111 struct sctp_transport
*transport
;
4112 sctp_assoc_t associd
;
4115 if (len
< sizeof(status
)) {
4120 len
= sizeof(status
);
4121 if (copy_from_user(&status
, optval
, len
)) {
4126 associd
= status
.sstat_assoc_id
;
4127 asoc
= sctp_id2assoc(sk
, associd
);
4133 transport
= asoc
->peer
.primary_path
;
4135 status
.sstat_assoc_id
= sctp_assoc2id(asoc
);
4136 status
.sstat_state
= asoc
->state
;
4137 status
.sstat_rwnd
= asoc
->peer
.rwnd
;
4138 status
.sstat_unackdata
= asoc
->unack_data
;
4140 status
.sstat_penddata
= sctp_tsnmap_pending(&asoc
->peer
.tsn_map
);
4141 status
.sstat_instrms
= asoc
->c
.sinit_max_instreams
;
4142 status
.sstat_outstrms
= asoc
->c
.sinit_num_ostreams
;
4143 status
.sstat_fragmentation_point
= asoc
->frag_point
;
4144 status
.sstat_primary
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
4145 memcpy(&status
.sstat_primary
.spinfo_address
, &transport
->ipaddr
,
4146 transport
->af_specific
->sockaddr_len
);
4147 /* Map ipv4 address into v4-mapped-on-v6 address. */
4148 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
4149 (union sctp_addr
*)&status
.sstat_primary
.spinfo_address
);
4150 status
.sstat_primary
.spinfo_state
= transport
->state
;
4151 status
.sstat_primary
.spinfo_cwnd
= transport
->cwnd
;
4152 status
.sstat_primary
.spinfo_srtt
= transport
->srtt
;
4153 status
.sstat_primary
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
4154 status
.sstat_primary
.spinfo_mtu
= transport
->pathmtu
;
4156 if (status
.sstat_primary
.spinfo_state
== SCTP_UNKNOWN
)
4157 status
.sstat_primary
.spinfo_state
= SCTP_ACTIVE
;
4159 if (put_user(len
, optlen
)) {
4164 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
4165 len
, status
.sstat_state
, status
.sstat_rwnd
,
4166 status
.sstat_assoc_id
);
4168 if (copy_to_user(optval
, &status
, len
)) {
4178 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4180 * Applications can retrieve information about a specific peer address
4181 * of an association, including its reachability state, congestion
4182 * window, and retransmission timer values. This information is
4185 static int sctp_getsockopt_peer_addr_info(struct sock
*sk
, int len
,
4186 char __user
*optval
,
4189 struct sctp_paddrinfo pinfo
;
4190 struct sctp_transport
*transport
;
4193 if (len
< sizeof(pinfo
)) {
4198 len
= sizeof(pinfo
);
4199 if (copy_from_user(&pinfo
, optval
, len
)) {
4204 transport
= sctp_addr_id2transport(sk
, &pinfo
.spinfo_address
,
4205 pinfo
.spinfo_assoc_id
);
4209 pinfo
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
4210 pinfo
.spinfo_state
= transport
->state
;
4211 pinfo
.spinfo_cwnd
= transport
->cwnd
;
4212 pinfo
.spinfo_srtt
= transport
->srtt
;
4213 pinfo
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
4214 pinfo
.spinfo_mtu
= transport
->pathmtu
;
4216 if (pinfo
.spinfo_state
== SCTP_UNKNOWN
)
4217 pinfo
.spinfo_state
= SCTP_ACTIVE
;
4219 if (put_user(len
, optlen
)) {
4224 if (copy_to_user(optval
, &pinfo
, len
)) {
4233 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4235 * This option is a on/off flag. If enabled no SCTP message
4236 * fragmentation will be performed. Instead if a message being sent
4237 * exceeds the current PMTU size, the message will NOT be sent and
4238 * instead a error will be indicated to the user.
4240 static int sctp_getsockopt_disable_fragments(struct sock
*sk
, int len
,
4241 char __user
*optval
, int __user
*optlen
)
4245 if (len
< sizeof(int))
4249 val
= (sctp_sk(sk
)->disable_fragments
== 1);
4250 if (put_user(len
, optlen
))
4252 if (copy_to_user(optval
, &val
, len
))
4257 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4259 * This socket option is used to specify various notifications and
4260 * ancillary data the user wishes to receive.
4262 static int sctp_getsockopt_events(struct sock
*sk
, int len
, char __user
*optval
,
4267 if (len
> sizeof(struct sctp_event_subscribe
))
4268 len
= sizeof(struct sctp_event_subscribe
);
4269 if (put_user(len
, optlen
))
4271 if (copy_to_user(optval
, &sctp_sk(sk
)->subscribe
, len
))
4276 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4278 * This socket option is applicable to the UDP-style socket only. When
4279 * set it will cause associations that are idle for more than the
4280 * specified number of seconds to automatically close. An association
4281 * being idle is defined an association that has NOT sent or received
4282 * user data. The special value of '0' indicates that no automatic
4283 * close of any associations should be performed. The option expects an
4284 * integer defining the number of seconds of idle time before an
4285 * association is closed.
4287 static int sctp_getsockopt_autoclose(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4289 /* Applicable to UDP-style socket only */
4290 if (sctp_style(sk
, TCP
))
4292 if (len
< sizeof(int))
4295 if (put_user(len
, optlen
))
4297 if (copy_to_user(optval
, &sctp_sk(sk
)->autoclose
, sizeof(int)))
4302 /* Helper routine to branch off an association to a new socket. */
4303 int sctp_do_peeloff(struct sock
*sk
, sctp_assoc_t id
, struct socket
**sockp
)
4305 struct sctp_association
*asoc
= sctp_id2assoc(sk
, id
);
4306 struct socket
*sock
;
4313 /* If there is a thread waiting on more sndbuf space for
4314 * sending on this asoc, it cannot be peeled.
4316 if (waitqueue_active(&asoc
->wait
))
4319 /* An association cannot be branched off from an already peeled-off
4320 * socket, nor is this supported for tcp style sockets.
4322 if (!sctp_style(sk
, UDP
))
4325 /* Create a new socket. */
4326 err
= sock_create(sk
->sk_family
, SOCK_SEQPACKET
, IPPROTO_SCTP
, &sock
);
4330 sctp_copy_sock(sock
->sk
, sk
, asoc
);
4332 /* Make peeled-off sockets more like 1-1 accepted sockets.
4333 * Set the daddr and initialize id to something more random
4335 af
= sctp_get_af_specific(asoc
->peer
.primary_addr
.sa
.sa_family
);
4336 af
->to_sk_daddr(&asoc
->peer
.primary_addr
, sk
);
4338 /* Populate the fields of the newsk from the oldsk and migrate the
4339 * asoc to the newsk.
4341 sctp_sock_migrate(sk
, sock
->sk
, asoc
, SCTP_SOCKET_UDP_HIGH_BANDWIDTH
);
4347 EXPORT_SYMBOL(sctp_do_peeloff
);
4349 static int sctp_getsockopt_peeloff(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4351 sctp_peeloff_arg_t peeloff
;
4352 struct socket
*newsock
;
4353 struct file
*newfile
;
4356 if (len
< sizeof(sctp_peeloff_arg_t
))
4358 len
= sizeof(sctp_peeloff_arg_t
);
4359 if (copy_from_user(&peeloff
, optval
, len
))
4362 retval
= sctp_do_peeloff(sk
, peeloff
.associd
, &newsock
);
4366 /* Map the socket to an unused fd that can be returned to the user. */
4367 retval
= get_unused_fd();
4369 sock_release(newsock
);
4373 newfile
= sock_alloc_file(newsock
, 0, NULL
);
4374 if (unlikely(IS_ERR(newfile
))) {
4375 put_unused_fd(retval
);
4376 sock_release(newsock
);
4377 return PTR_ERR(newfile
);
4380 SCTP_DEBUG_PRINTK("%s: sk: %p newsk: %p sd: %d\n",
4381 __func__
, sk
, newsock
->sk
, retval
);
4383 /* Return the fd mapped to the new socket. */
4384 if (put_user(len
, optlen
)) {
4386 put_unused_fd(retval
);
4389 peeloff
.sd
= retval
;
4390 if (copy_to_user(optval
, &peeloff
, len
)) {
4392 put_unused_fd(retval
);
4395 fd_install(retval
, newfile
);
4400 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4402 * Applications can enable or disable heartbeats for any peer address of
4403 * an association, modify an address's heartbeat interval, force a
4404 * heartbeat to be sent immediately, and adjust the address's maximum
4405 * number of retransmissions sent before an address is considered
4406 * unreachable. The following structure is used to access and modify an
4407 * address's parameters:
4409 * struct sctp_paddrparams {
4410 * sctp_assoc_t spp_assoc_id;
4411 * struct sockaddr_storage spp_address;
4412 * uint32_t spp_hbinterval;
4413 * uint16_t spp_pathmaxrxt;
4414 * uint32_t spp_pathmtu;
4415 * uint32_t spp_sackdelay;
4416 * uint32_t spp_flags;
4419 * spp_assoc_id - (one-to-many style socket) This is filled in the
4420 * application, and identifies the association for
4422 * spp_address - This specifies which address is of interest.
4423 * spp_hbinterval - This contains the value of the heartbeat interval,
4424 * in milliseconds. If a value of zero
4425 * is present in this field then no changes are to
4426 * be made to this parameter.
4427 * spp_pathmaxrxt - This contains the maximum number of
4428 * retransmissions before this address shall be
4429 * considered unreachable. If a value of zero
4430 * is present in this field then no changes are to
4431 * be made to this parameter.
4432 * spp_pathmtu - When Path MTU discovery is disabled the value
4433 * specified here will be the "fixed" path mtu.
4434 * Note that if the spp_address field is empty
4435 * then all associations on this address will
4436 * have this fixed path mtu set upon them.
4438 * spp_sackdelay - When delayed sack is enabled, this value specifies
4439 * the number of milliseconds that sacks will be delayed
4440 * for. This value will apply to all addresses of an
4441 * association if the spp_address field is empty. Note
4442 * also, that if delayed sack is enabled and this
4443 * value is set to 0, no change is made to the last
4444 * recorded delayed sack timer value.
4446 * spp_flags - These flags are used to control various features
4447 * on an association. The flag field may contain
4448 * zero or more of the following options.
4450 * SPP_HB_ENABLE - Enable heartbeats on the
4451 * specified address. Note that if the address
4452 * field is empty all addresses for the association
4453 * have heartbeats enabled upon them.
4455 * SPP_HB_DISABLE - Disable heartbeats on the
4456 * speicifed address. Note that if the address
4457 * field is empty all addresses for the association
4458 * will have their heartbeats disabled. Note also
4459 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4460 * mutually exclusive, only one of these two should
4461 * be specified. Enabling both fields will have
4462 * undetermined results.
4464 * SPP_HB_DEMAND - Request a user initiated heartbeat
4465 * to be made immediately.
4467 * SPP_PMTUD_ENABLE - This field will enable PMTU
4468 * discovery upon the specified address. Note that
4469 * if the address feild is empty then all addresses
4470 * on the association are effected.
4472 * SPP_PMTUD_DISABLE - This field will disable PMTU
4473 * discovery upon the specified address. Note that
4474 * if the address feild is empty then all addresses
4475 * on the association are effected. Not also that
4476 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4477 * exclusive. Enabling both will have undetermined
4480 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4481 * on delayed sack. The time specified in spp_sackdelay
4482 * is used to specify the sack delay for this address. Note
4483 * that if spp_address is empty then all addresses will
4484 * enable delayed sack and take on the sack delay
4485 * value specified in spp_sackdelay.
4486 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4487 * off delayed sack. If the spp_address field is blank then
4488 * delayed sack is disabled for the entire association. Note
4489 * also that this field is mutually exclusive to
4490 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4493 static int sctp_getsockopt_peer_addr_params(struct sock
*sk
, int len
,
4494 char __user
*optval
, int __user
*optlen
)
4496 struct sctp_paddrparams params
;
4497 struct sctp_transport
*trans
= NULL
;
4498 struct sctp_association
*asoc
= NULL
;
4499 struct sctp_sock
*sp
= sctp_sk(sk
);
4501 if (len
< sizeof(struct sctp_paddrparams
))
4503 len
= sizeof(struct sctp_paddrparams
);
4504 if (copy_from_user(¶ms
, optval
, len
))
4507 /* If an address other than INADDR_ANY is specified, and
4508 * no transport is found, then the request is invalid.
4510 if (!sctp_is_any(sk
, ( union sctp_addr
*)¶ms
.spp_address
)) {
4511 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
4512 params
.spp_assoc_id
);
4514 SCTP_DEBUG_PRINTK("Failed no transport\n");
4519 /* Get association, if assoc_id != 0 and the socket is a one
4520 * to many style socket, and an association was not found, then
4521 * the id was invalid.
4523 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
4524 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
)) {
4525 SCTP_DEBUG_PRINTK("Failed no association\n");
4530 /* Fetch transport values. */
4531 params
.spp_hbinterval
= jiffies_to_msecs(trans
->hbinterval
);
4532 params
.spp_pathmtu
= trans
->pathmtu
;
4533 params
.spp_pathmaxrxt
= trans
->pathmaxrxt
;
4534 params
.spp_sackdelay
= jiffies_to_msecs(trans
->sackdelay
);
4536 /*draft-11 doesn't say what to return in spp_flags*/
4537 params
.spp_flags
= trans
->param_flags
;
4539 /* Fetch association values. */
4540 params
.spp_hbinterval
= jiffies_to_msecs(asoc
->hbinterval
);
4541 params
.spp_pathmtu
= asoc
->pathmtu
;
4542 params
.spp_pathmaxrxt
= asoc
->pathmaxrxt
;
4543 params
.spp_sackdelay
= jiffies_to_msecs(asoc
->sackdelay
);
4545 /*draft-11 doesn't say what to return in spp_flags*/
4546 params
.spp_flags
= asoc
->param_flags
;
4548 /* Fetch socket values. */
4549 params
.spp_hbinterval
= sp
->hbinterval
;
4550 params
.spp_pathmtu
= sp
->pathmtu
;
4551 params
.spp_sackdelay
= sp
->sackdelay
;
4552 params
.spp_pathmaxrxt
= sp
->pathmaxrxt
;
4554 /*draft-11 doesn't say what to return in spp_flags*/
4555 params
.spp_flags
= sp
->param_flags
;
4558 if (copy_to_user(optval
, ¶ms
, len
))
4561 if (put_user(len
, optlen
))
4568 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4570 * This option will effect the way delayed acks are performed. This
4571 * option allows you to get or set the delayed ack time, in
4572 * milliseconds. It also allows changing the delayed ack frequency.
4573 * Changing the frequency to 1 disables the delayed sack algorithm. If
4574 * the assoc_id is 0, then this sets or gets the endpoints default
4575 * values. If the assoc_id field is non-zero, then the set or get
4576 * effects the specified association for the one to many model (the
4577 * assoc_id field is ignored by the one to one model). Note that if
4578 * sack_delay or sack_freq are 0 when setting this option, then the
4579 * current values will remain unchanged.
4581 * struct sctp_sack_info {
4582 * sctp_assoc_t sack_assoc_id;
4583 * uint32_t sack_delay;
4584 * uint32_t sack_freq;
4587 * sack_assoc_id - This parameter, indicates which association the user
4588 * is performing an action upon. Note that if this field's value is
4589 * zero then the endpoints default value is changed (effecting future
4590 * associations only).
4592 * sack_delay - This parameter contains the number of milliseconds that
4593 * the user is requesting the delayed ACK timer be set to. Note that
4594 * this value is defined in the standard to be between 200 and 500
4597 * sack_freq - This parameter contains the number of packets that must
4598 * be received before a sack is sent without waiting for the delay
4599 * timer to expire. The default value for this is 2, setting this
4600 * value to 1 will disable the delayed sack algorithm.
4602 static int sctp_getsockopt_delayed_ack(struct sock
*sk
, int len
,
4603 char __user
*optval
,
4606 struct sctp_sack_info params
;
4607 struct sctp_association
*asoc
= NULL
;
4608 struct sctp_sock
*sp
= sctp_sk(sk
);
4610 if (len
>= sizeof(struct sctp_sack_info
)) {
4611 len
= sizeof(struct sctp_sack_info
);
4613 if (copy_from_user(¶ms
, optval
, len
))
4615 } else if (len
== sizeof(struct sctp_assoc_value
)) {
4616 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4617 pr_warn("Use struct sctp_sack_info instead\n");
4618 if (copy_from_user(¶ms
, optval
, len
))
4623 /* Get association, if sack_assoc_id != 0 and the socket is a one
4624 * to many style socket, and an association was not found, then
4625 * the id was invalid.
4627 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
4628 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
4632 /* Fetch association values. */
4633 if (asoc
->param_flags
& SPP_SACKDELAY_ENABLE
) {
4634 params
.sack_delay
= jiffies_to_msecs(
4636 params
.sack_freq
= asoc
->sackfreq
;
4639 params
.sack_delay
= 0;
4640 params
.sack_freq
= 1;
4643 /* Fetch socket values. */
4644 if (sp
->param_flags
& SPP_SACKDELAY_ENABLE
) {
4645 params
.sack_delay
= sp
->sackdelay
;
4646 params
.sack_freq
= sp
->sackfreq
;
4648 params
.sack_delay
= 0;
4649 params
.sack_freq
= 1;
4653 if (copy_to_user(optval
, ¶ms
, len
))
4656 if (put_user(len
, optlen
))
4662 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4664 * Applications can specify protocol parameters for the default association
4665 * initialization. The option name argument to setsockopt() and getsockopt()
4668 * Setting initialization parameters is effective only on an unconnected
4669 * socket (for UDP-style sockets only future associations are effected
4670 * by the change). With TCP-style sockets, this option is inherited by
4671 * sockets derived from a listener socket.
4673 static int sctp_getsockopt_initmsg(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4675 if (len
< sizeof(struct sctp_initmsg
))
4677 len
= sizeof(struct sctp_initmsg
);
4678 if (put_user(len
, optlen
))
4680 if (copy_to_user(optval
, &sctp_sk(sk
)->initmsg
, len
))
4686 static int sctp_getsockopt_peer_addrs(struct sock
*sk
, int len
,
4687 char __user
*optval
, int __user
*optlen
)
4689 struct sctp_association
*asoc
;
4691 struct sctp_getaddrs getaddrs
;
4692 struct sctp_transport
*from
;
4694 union sctp_addr temp
;
4695 struct sctp_sock
*sp
= sctp_sk(sk
);
4700 if (len
< sizeof(struct sctp_getaddrs
))
4703 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
4706 /* For UDP-style sockets, id specifies the association to query. */
4707 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
4711 to
= optval
+ offsetof(struct sctp_getaddrs
,addrs
);
4712 space_left
= len
- offsetof(struct sctp_getaddrs
,addrs
);
4714 list_for_each_entry(from
, &asoc
->peer
.transport_addr_list
,
4716 memcpy(&temp
, &from
->ipaddr
, sizeof(temp
));
4717 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
4718 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4719 if (space_left
< addrlen
)
4721 if (copy_to_user(to
, &temp
, addrlen
))
4725 space_left
-= addrlen
;
4728 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
))
4730 bytes_copied
= ((char __user
*)to
) - optval
;
4731 if (put_user(bytes_copied
, optlen
))
4737 static int sctp_copy_laddrs(struct sock
*sk
, __u16 port
, void *to
,
4738 size_t space_left
, int *bytes_copied
)
4740 struct sctp_sockaddr_entry
*addr
;
4741 union sctp_addr temp
;
4744 struct net
*net
= sock_net(sk
);
4747 list_for_each_entry_rcu(addr
, &net
->sctp
.local_addr_list
, list
) {
4751 if ((PF_INET
== sk
->sk_family
) &&
4752 (AF_INET6
== addr
->a
.sa
.sa_family
))
4754 if ((PF_INET6
== sk
->sk_family
) &&
4755 inet_v6_ipv6only(sk
) &&
4756 (AF_INET
== addr
->a
.sa
.sa_family
))
4758 memcpy(&temp
, &addr
->a
, sizeof(temp
));
4759 if (!temp
.v4
.sin_port
)
4760 temp
.v4
.sin_port
= htons(port
);
4762 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sctp_sk(sk
),
4764 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4765 if (space_left
< addrlen
) {
4769 memcpy(to
, &temp
, addrlen
);
4773 space_left
-= addrlen
;
4774 *bytes_copied
+= addrlen
;
4782 static int sctp_getsockopt_local_addrs(struct sock
*sk
, int len
,
4783 char __user
*optval
, int __user
*optlen
)
4785 struct sctp_bind_addr
*bp
;
4786 struct sctp_association
*asoc
;
4788 struct sctp_getaddrs getaddrs
;
4789 struct sctp_sockaddr_entry
*addr
;
4791 union sctp_addr temp
;
4792 struct sctp_sock
*sp
= sctp_sk(sk
);
4796 int bytes_copied
= 0;
4800 if (len
< sizeof(struct sctp_getaddrs
))
4803 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
4807 * For UDP-style sockets, id specifies the association to query.
4808 * If the id field is set to the value '0' then the locally bound
4809 * addresses are returned without regard to any particular
4812 if (0 == getaddrs
.assoc_id
) {
4813 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
4815 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
4818 bp
= &asoc
->base
.bind_addr
;
4821 to
= optval
+ offsetof(struct sctp_getaddrs
,addrs
);
4822 space_left
= len
- offsetof(struct sctp_getaddrs
,addrs
);
4824 addrs
= kmalloc(space_left
, GFP_KERNEL
);
4828 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4829 * addresses from the global local address list.
4831 if (sctp_list_single_entry(&bp
->address_list
)) {
4832 addr
= list_entry(bp
->address_list
.next
,
4833 struct sctp_sockaddr_entry
, list
);
4834 if (sctp_is_any(sk
, &addr
->a
)) {
4835 cnt
= sctp_copy_laddrs(sk
, bp
->port
, addrs
,
4836 space_left
, &bytes_copied
);
4846 /* Protection on the bound address list is not needed since
4847 * in the socket option context we hold a socket lock and
4848 * thus the bound address list can't change.
4850 list_for_each_entry(addr
, &bp
->address_list
, list
) {
4851 memcpy(&temp
, &addr
->a
, sizeof(temp
));
4852 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
, &temp
);
4853 addrlen
= sctp_get_af_specific(temp
.sa
.sa_family
)->sockaddr_len
;
4854 if (space_left
< addrlen
) {
4855 err
= -ENOMEM
; /*fixme: right error?*/
4858 memcpy(buf
, &temp
, addrlen
);
4860 bytes_copied
+= addrlen
;
4862 space_left
-= addrlen
;
4866 if (copy_to_user(to
, addrs
, bytes_copied
)) {
4870 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
)) {
4874 if (put_user(bytes_copied
, optlen
))
4881 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4883 * Requests that the local SCTP stack use the enclosed peer address as
4884 * the association primary. The enclosed address must be one of the
4885 * association peer's addresses.
4887 static int sctp_getsockopt_primary_addr(struct sock
*sk
, int len
,
4888 char __user
*optval
, int __user
*optlen
)
4890 struct sctp_prim prim
;
4891 struct sctp_association
*asoc
;
4892 struct sctp_sock
*sp
= sctp_sk(sk
);
4894 if (len
< sizeof(struct sctp_prim
))
4897 len
= sizeof(struct sctp_prim
);
4899 if (copy_from_user(&prim
, optval
, len
))
4902 asoc
= sctp_id2assoc(sk
, prim
.ssp_assoc_id
);
4906 if (!asoc
->peer
.primary_path
)
4909 memcpy(&prim
.ssp_addr
, &asoc
->peer
.primary_path
->ipaddr
,
4910 asoc
->peer
.primary_path
->af_specific
->sockaddr_len
);
4912 sctp_get_pf_specific(sk
->sk_family
)->addr_v4map(sp
,
4913 (union sctp_addr
*)&prim
.ssp_addr
);
4915 if (put_user(len
, optlen
))
4917 if (copy_to_user(optval
, &prim
, len
))
4924 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4926 * Requests that the local endpoint set the specified Adaptation Layer
4927 * Indication parameter for all future INIT and INIT-ACK exchanges.
4929 static int sctp_getsockopt_adaptation_layer(struct sock
*sk
, int len
,
4930 char __user
*optval
, int __user
*optlen
)
4932 struct sctp_setadaptation adaptation
;
4934 if (len
< sizeof(struct sctp_setadaptation
))
4937 len
= sizeof(struct sctp_setadaptation
);
4939 adaptation
.ssb_adaptation_ind
= sctp_sk(sk
)->adaptation_ind
;
4941 if (put_user(len
, optlen
))
4943 if (copy_to_user(optval
, &adaptation
, len
))
4951 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4953 * Applications that wish to use the sendto() system call may wish to
4954 * specify a default set of parameters that would normally be supplied
4955 * through the inclusion of ancillary data. This socket option allows
4956 * such an application to set the default sctp_sndrcvinfo structure.
4959 * The application that wishes to use this socket option simply passes
4960 * in to this call the sctp_sndrcvinfo structure defined in Section
4961 * 5.2.2) The input parameters accepted by this call include
4962 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4963 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4964 * to this call if the caller is using the UDP model.
4966 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4968 static int sctp_getsockopt_default_send_param(struct sock
*sk
,
4969 int len
, char __user
*optval
,
4972 struct sctp_sndrcvinfo info
;
4973 struct sctp_association
*asoc
;
4974 struct sctp_sock
*sp
= sctp_sk(sk
);
4976 if (len
< sizeof(struct sctp_sndrcvinfo
))
4979 len
= sizeof(struct sctp_sndrcvinfo
);
4981 if (copy_from_user(&info
, optval
, len
))
4984 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
4985 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
4989 info
.sinfo_stream
= asoc
->default_stream
;
4990 info
.sinfo_flags
= asoc
->default_flags
;
4991 info
.sinfo_ppid
= asoc
->default_ppid
;
4992 info
.sinfo_context
= asoc
->default_context
;
4993 info
.sinfo_timetolive
= asoc
->default_timetolive
;
4995 info
.sinfo_stream
= sp
->default_stream
;
4996 info
.sinfo_flags
= sp
->default_flags
;
4997 info
.sinfo_ppid
= sp
->default_ppid
;
4998 info
.sinfo_context
= sp
->default_context
;
4999 info
.sinfo_timetolive
= sp
->default_timetolive
;
5002 if (put_user(len
, optlen
))
5004 if (copy_to_user(optval
, &info
, len
))
5012 * 7.1.5 SCTP_NODELAY
5014 * Turn on/off any Nagle-like algorithm. This means that packets are
5015 * generally sent as soon as possible and no unnecessary delays are
5016 * introduced, at the cost of more packets in the network. Expects an
5017 * integer boolean flag.
5020 static int sctp_getsockopt_nodelay(struct sock
*sk
, int len
,
5021 char __user
*optval
, int __user
*optlen
)
5025 if (len
< sizeof(int))
5029 val
= (sctp_sk(sk
)->nodelay
== 1);
5030 if (put_user(len
, optlen
))
5032 if (copy_to_user(optval
, &val
, len
))
5039 * 7.1.1 SCTP_RTOINFO
5041 * The protocol parameters used to initialize and bound retransmission
5042 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5043 * and modify these parameters.
5044 * All parameters are time values, in milliseconds. A value of 0, when
5045 * modifying the parameters, indicates that the current value should not
5049 static int sctp_getsockopt_rtoinfo(struct sock
*sk
, int len
,
5050 char __user
*optval
,
5051 int __user
*optlen
) {
5052 struct sctp_rtoinfo rtoinfo
;
5053 struct sctp_association
*asoc
;
5055 if (len
< sizeof (struct sctp_rtoinfo
))
5058 len
= sizeof(struct sctp_rtoinfo
);
5060 if (copy_from_user(&rtoinfo
, optval
, len
))
5063 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
5065 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
5068 /* Values corresponding to the specific association. */
5070 rtoinfo
.srto_initial
= jiffies_to_msecs(asoc
->rto_initial
);
5071 rtoinfo
.srto_max
= jiffies_to_msecs(asoc
->rto_max
);
5072 rtoinfo
.srto_min
= jiffies_to_msecs(asoc
->rto_min
);
5074 /* Values corresponding to the endpoint. */
5075 struct sctp_sock
*sp
= sctp_sk(sk
);
5077 rtoinfo
.srto_initial
= sp
->rtoinfo
.srto_initial
;
5078 rtoinfo
.srto_max
= sp
->rtoinfo
.srto_max
;
5079 rtoinfo
.srto_min
= sp
->rtoinfo
.srto_min
;
5082 if (put_user(len
, optlen
))
5085 if (copy_to_user(optval
, &rtoinfo
, len
))
5093 * 7.1.2 SCTP_ASSOCINFO
5095 * This option is used to tune the maximum retransmission attempts
5096 * of the association.
5097 * Returns an error if the new association retransmission value is
5098 * greater than the sum of the retransmission value of the peer.
5099 * See [SCTP] for more information.
5102 static int sctp_getsockopt_associnfo(struct sock
*sk
, int len
,
5103 char __user
*optval
,
5107 struct sctp_assocparams assocparams
;
5108 struct sctp_association
*asoc
;
5109 struct list_head
*pos
;
5112 if (len
< sizeof (struct sctp_assocparams
))
5115 len
= sizeof(struct sctp_assocparams
);
5117 if (copy_from_user(&assocparams
, optval
, len
))
5120 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
5122 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
5125 /* Values correspoinding to the specific association */
5127 assocparams
.sasoc_asocmaxrxt
= asoc
->max_retrans
;
5128 assocparams
.sasoc_peer_rwnd
= asoc
->peer
.rwnd
;
5129 assocparams
.sasoc_local_rwnd
= asoc
->a_rwnd
;
5130 assocparams
.sasoc_cookie_life
= (asoc
->cookie_life
.tv_sec
5132 (asoc
->cookie_life
.tv_usec
5135 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
5139 assocparams
.sasoc_number_peer_destinations
= cnt
;
5141 /* Values corresponding to the endpoint */
5142 struct sctp_sock
*sp
= sctp_sk(sk
);
5144 assocparams
.sasoc_asocmaxrxt
= sp
->assocparams
.sasoc_asocmaxrxt
;
5145 assocparams
.sasoc_peer_rwnd
= sp
->assocparams
.sasoc_peer_rwnd
;
5146 assocparams
.sasoc_local_rwnd
= sp
->assocparams
.sasoc_local_rwnd
;
5147 assocparams
.sasoc_cookie_life
=
5148 sp
->assocparams
.sasoc_cookie_life
;
5149 assocparams
.sasoc_number_peer_destinations
=
5151 sasoc_number_peer_destinations
;
5154 if (put_user(len
, optlen
))
5157 if (copy_to_user(optval
, &assocparams
, len
))
5164 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5166 * This socket option is a boolean flag which turns on or off mapped V4
5167 * addresses. If this option is turned on and the socket is type
5168 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5169 * If this option is turned off, then no mapping will be done of V4
5170 * addresses and a user will receive both PF_INET6 and PF_INET type
5171 * addresses on the socket.
5173 static int sctp_getsockopt_mappedv4(struct sock
*sk
, int len
,
5174 char __user
*optval
, int __user
*optlen
)
5177 struct sctp_sock
*sp
= sctp_sk(sk
);
5179 if (len
< sizeof(int))
5184 if (put_user(len
, optlen
))
5186 if (copy_to_user(optval
, &val
, len
))
5193 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5194 * (chapter and verse is quoted at sctp_setsockopt_context())
5196 static int sctp_getsockopt_context(struct sock
*sk
, int len
,
5197 char __user
*optval
, int __user
*optlen
)
5199 struct sctp_assoc_value params
;
5200 struct sctp_sock
*sp
;
5201 struct sctp_association
*asoc
;
5203 if (len
< sizeof(struct sctp_assoc_value
))
5206 len
= sizeof(struct sctp_assoc_value
);
5208 if (copy_from_user(¶ms
, optval
, len
))
5213 if (params
.assoc_id
!= 0) {
5214 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5217 params
.assoc_value
= asoc
->default_rcv_context
;
5219 params
.assoc_value
= sp
->default_rcv_context
;
5222 if (put_user(len
, optlen
))
5224 if (copy_to_user(optval
, ¶ms
, len
))
5231 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5232 * This option will get or set the maximum size to put in any outgoing
5233 * SCTP DATA chunk. If a message is larger than this size it will be
5234 * fragmented by SCTP into the specified size. Note that the underlying
5235 * SCTP implementation may fragment into smaller sized chunks when the
5236 * PMTU of the underlying association is smaller than the value set by
5237 * the user. The default value for this option is '0' which indicates
5238 * the user is NOT limiting fragmentation and only the PMTU will effect
5239 * SCTP's choice of DATA chunk size. Note also that values set larger
5240 * than the maximum size of an IP datagram will effectively let SCTP
5241 * control fragmentation (i.e. the same as setting this option to 0).
5243 * The following structure is used to access and modify this parameter:
5245 * struct sctp_assoc_value {
5246 * sctp_assoc_t assoc_id;
5247 * uint32_t assoc_value;
5250 * assoc_id: This parameter is ignored for one-to-one style sockets.
5251 * For one-to-many style sockets this parameter indicates which
5252 * association the user is performing an action upon. Note that if
5253 * this field's value is zero then the endpoints default value is
5254 * changed (effecting future associations only).
5255 * assoc_value: This parameter specifies the maximum size in bytes.
5257 static int sctp_getsockopt_maxseg(struct sock
*sk
, int len
,
5258 char __user
*optval
, int __user
*optlen
)
5260 struct sctp_assoc_value params
;
5261 struct sctp_association
*asoc
;
5263 if (len
== sizeof(int)) {
5264 pr_warn("Use of int in maxseg socket option deprecated\n");
5265 pr_warn("Use struct sctp_assoc_value instead\n");
5266 params
.assoc_id
= 0;
5267 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5268 len
= sizeof(struct sctp_assoc_value
);
5269 if (copy_from_user(¶ms
, optval
, sizeof(params
)))
5274 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5275 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
5279 params
.assoc_value
= asoc
->frag_point
;
5281 params
.assoc_value
= sctp_sk(sk
)->user_frag
;
5283 if (put_user(len
, optlen
))
5285 if (len
== sizeof(int)) {
5286 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5289 if (copy_to_user(optval
, ¶ms
, len
))
5297 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5298 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5300 static int sctp_getsockopt_fragment_interleave(struct sock
*sk
, int len
,
5301 char __user
*optval
, int __user
*optlen
)
5305 if (len
< sizeof(int))
5310 val
= sctp_sk(sk
)->frag_interleave
;
5311 if (put_user(len
, optlen
))
5313 if (copy_to_user(optval
, &val
, len
))
5320 * 7.1.25. Set or Get the sctp partial delivery point
5321 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5323 static int sctp_getsockopt_partial_delivery_point(struct sock
*sk
, int len
,
5324 char __user
*optval
,
5329 if (len
< sizeof(u32
))
5334 val
= sctp_sk(sk
)->pd_point
;
5335 if (put_user(len
, optlen
))
5337 if (copy_to_user(optval
, &val
, len
))
5344 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5345 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5347 static int sctp_getsockopt_maxburst(struct sock
*sk
, int len
,
5348 char __user
*optval
,
5351 struct sctp_assoc_value params
;
5352 struct sctp_sock
*sp
;
5353 struct sctp_association
*asoc
;
5355 if (len
== sizeof(int)) {
5356 pr_warn("Use of int in max_burst socket option deprecated\n");
5357 pr_warn("Use struct sctp_assoc_value instead\n");
5358 params
.assoc_id
= 0;
5359 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5360 len
= sizeof(struct sctp_assoc_value
);
5361 if (copy_from_user(¶ms
, optval
, len
))
5368 if (params
.assoc_id
!= 0) {
5369 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5372 params
.assoc_value
= asoc
->max_burst
;
5374 params
.assoc_value
= sp
->max_burst
;
5376 if (len
== sizeof(int)) {
5377 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5380 if (copy_to_user(optval
, ¶ms
, len
))
5388 static int sctp_getsockopt_hmac_ident(struct sock
*sk
, int len
,
5389 char __user
*optval
, int __user
*optlen
)
5391 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5392 struct sctp_hmacalgo __user
*p
= (void __user
*)optval
;
5393 struct sctp_hmac_algo_param
*hmacs
;
5397 if (!ep
->auth_enable
)
5400 hmacs
= ep
->auth_hmacs_list
;
5401 data_len
= ntohs(hmacs
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5403 if (len
< sizeof(struct sctp_hmacalgo
) + data_len
)
5406 len
= sizeof(struct sctp_hmacalgo
) + data_len
;
5407 num_idents
= data_len
/ sizeof(u16
);
5409 if (put_user(len
, optlen
))
5411 if (put_user(num_idents
, &p
->shmac_num_idents
))
5413 if (copy_to_user(p
->shmac_idents
, hmacs
->hmac_ids
, data_len
))
5418 static int sctp_getsockopt_active_key(struct sock
*sk
, int len
,
5419 char __user
*optval
, int __user
*optlen
)
5421 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5422 struct sctp_authkeyid val
;
5423 struct sctp_association
*asoc
;
5425 if (!ep
->auth_enable
)
5428 if (len
< sizeof(struct sctp_authkeyid
))
5430 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authkeyid
)))
5433 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
5434 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
5438 val
.scact_keynumber
= asoc
->active_key_id
;
5440 val
.scact_keynumber
= ep
->active_key_id
;
5442 len
= sizeof(struct sctp_authkeyid
);
5443 if (put_user(len
, optlen
))
5445 if (copy_to_user(optval
, &val
, len
))
5451 static int sctp_getsockopt_peer_auth_chunks(struct sock
*sk
, int len
,
5452 char __user
*optval
, int __user
*optlen
)
5454 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5455 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5456 struct sctp_authchunks val
;
5457 struct sctp_association
*asoc
;
5458 struct sctp_chunks_param
*ch
;
5462 if (!ep
->auth_enable
)
5465 if (len
< sizeof(struct sctp_authchunks
))
5468 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5471 to
= p
->gauth_chunks
;
5472 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5476 ch
= asoc
->peer
.peer_chunks
;
5480 /* See if the user provided enough room for all the data */
5481 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5482 if (len
< num_chunks
)
5485 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
5488 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
5489 if (put_user(len
, optlen
)) return -EFAULT
;
5490 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
5495 static int sctp_getsockopt_local_auth_chunks(struct sock
*sk
, int len
,
5496 char __user
*optval
, int __user
*optlen
)
5498 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5499 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5500 struct sctp_authchunks val
;
5501 struct sctp_association
*asoc
;
5502 struct sctp_chunks_param
*ch
;
5506 if (!ep
->auth_enable
)
5509 if (len
< sizeof(struct sctp_authchunks
))
5512 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5515 to
= p
->gauth_chunks
;
5516 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5517 if (!asoc
&& val
.gauth_assoc_id
&& sctp_style(sk
, UDP
))
5521 ch
= (struct sctp_chunks_param
*)asoc
->c
.auth_chunks
;
5523 ch
= ep
->auth_chunk_list
;
5528 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5529 if (len
< sizeof(struct sctp_authchunks
) + num_chunks
)
5532 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
5535 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
5536 if (put_user(len
, optlen
))
5538 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
5545 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5546 * This option gets the current number of associations that are attached
5547 * to a one-to-many style socket. The option value is an uint32_t.
5549 static int sctp_getsockopt_assoc_number(struct sock
*sk
, int len
,
5550 char __user
*optval
, int __user
*optlen
)
5552 struct sctp_sock
*sp
= sctp_sk(sk
);
5553 struct sctp_association
*asoc
;
5556 if (sctp_style(sk
, TCP
))
5559 if (len
< sizeof(u32
))
5564 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5568 if (put_user(len
, optlen
))
5570 if (copy_to_user(optval
, &val
, len
))
5577 * 8.1.23 SCTP_AUTO_ASCONF
5578 * See the corresponding setsockopt entry as description
5580 static int sctp_getsockopt_auto_asconf(struct sock
*sk
, int len
,
5581 char __user
*optval
, int __user
*optlen
)
5585 if (len
< sizeof(int))
5589 if (sctp_sk(sk
)->do_auto_asconf
&& sctp_is_ep_boundall(sk
))
5591 if (put_user(len
, optlen
))
5593 if (copy_to_user(optval
, &val
, len
))
5599 * 8.2.6. Get the Current Identifiers of Associations
5600 * (SCTP_GET_ASSOC_ID_LIST)
5602 * This option gets the current list of SCTP association identifiers of
5603 * the SCTP associations handled by a one-to-many style socket.
5605 static int sctp_getsockopt_assoc_ids(struct sock
*sk
, int len
,
5606 char __user
*optval
, int __user
*optlen
)
5608 struct sctp_sock
*sp
= sctp_sk(sk
);
5609 struct sctp_association
*asoc
;
5610 struct sctp_assoc_ids
*ids
;
5613 if (sctp_style(sk
, TCP
))
5616 if (len
< sizeof(struct sctp_assoc_ids
))
5619 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5623 if (len
< sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
)
5626 len
= sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
;
5628 ids
= kmalloc(len
, GFP_KERNEL
);
5632 ids
->gaids_number_of_ids
= num
;
5634 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5635 ids
->gaids_assoc_id
[num
++] = asoc
->assoc_id
;
5638 if (put_user(len
, optlen
) || copy_to_user(optval
, ids
, len
)) {
5648 * SCTP_PEER_ADDR_THLDS
5650 * This option allows us to fetch the partially failed threshold for one or all
5651 * transports in an association. See Section 6.1 of:
5652 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5654 static int sctp_getsockopt_paddr_thresholds(struct sock
*sk
,
5655 char __user
*optval
,
5659 struct sctp_paddrthlds val
;
5660 struct sctp_transport
*trans
;
5661 struct sctp_association
*asoc
;
5663 if (len
< sizeof(struct sctp_paddrthlds
))
5665 len
= sizeof(struct sctp_paddrthlds
);
5666 if (copy_from_user(&val
, (struct sctp_paddrthlds __user
*)optval
, len
))
5669 if (sctp_is_any(sk
, (const union sctp_addr
*)&val
.spt_address
)) {
5670 asoc
= sctp_id2assoc(sk
, val
.spt_assoc_id
);
5674 val
.spt_pathpfthld
= asoc
->pf_retrans
;
5675 val
.spt_pathmaxrxt
= asoc
->pathmaxrxt
;
5677 trans
= sctp_addr_id2transport(sk
, &val
.spt_address
,
5682 val
.spt_pathmaxrxt
= trans
->pathmaxrxt
;
5683 val
.spt_pathpfthld
= trans
->pf_retrans
;
5686 if (put_user(len
, optlen
) || copy_to_user(optval
, &val
, len
))
5693 * SCTP_GET_ASSOC_STATS
5695 * This option retrieves local per endpoint statistics. It is modeled
5696 * after OpenSolaris' implementation
5698 static int sctp_getsockopt_assoc_stats(struct sock
*sk
, int len
,
5699 char __user
*optval
,
5702 struct sctp_assoc_stats sas
;
5703 struct sctp_association
*asoc
= NULL
;
5705 /* User must provide at least the assoc id */
5706 if (len
< sizeof(sctp_assoc_t
))
5709 /* Allow the struct to grow and fill in as much as possible */
5710 len
= min_t(size_t, len
, sizeof(sas
));
5712 if (copy_from_user(&sas
, optval
, len
))
5715 asoc
= sctp_id2assoc(sk
, sas
.sas_assoc_id
);
5719 sas
.sas_rtxchunks
= asoc
->stats
.rtxchunks
;
5720 sas
.sas_gapcnt
= asoc
->stats
.gapcnt
;
5721 sas
.sas_outofseqtsns
= asoc
->stats
.outofseqtsns
;
5722 sas
.sas_osacks
= asoc
->stats
.osacks
;
5723 sas
.sas_isacks
= asoc
->stats
.isacks
;
5724 sas
.sas_octrlchunks
= asoc
->stats
.octrlchunks
;
5725 sas
.sas_ictrlchunks
= asoc
->stats
.ictrlchunks
;
5726 sas
.sas_oodchunks
= asoc
->stats
.oodchunks
;
5727 sas
.sas_iodchunks
= asoc
->stats
.iodchunks
;
5728 sas
.sas_ouodchunks
= asoc
->stats
.ouodchunks
;
5729 sas
.sas_iuodchunks
= asoc
->stats
.iuodchunks
;
5730 sas
.sas_idupchunks
= asoc
->stats
.idupchunks
;
5731 sas
.sas_opackets
= asoc
->stats
.opackets
;
5732 sas
.sas_ipackets
= asoc
->stats
.ipackets
;
5734 /* New high max rto observed, will return 0 if not a single
5735 * RTO update took place. obs_rto_ipaddr will be bogus
5738 sas
.sas_maxrto
= asoc
->stats
.max_obs_rto
;
5739 memcpy(&sas
.sas_obs_rto_ipaddr
, &asoc
->stats
.obs_rto_ipaddr
,
5740 sizeof(struct sockaddr_storage
));
5742 /* Mark beginning of a new observation period */
5743 asoc
->stats
.max_obs_rto
= asoc
->rto_min
;
5745 if (put_user(len
, optlen
))
5748 SCTP_DEBUG_PRINTK("sctp_getsockopt_assoc_stat(%d): %d\n",
5749 len
, sas
.sas_assoc_id
);
5751 if (copy_to_user(optval
, &sas
, len
))
5757 SCTP_STATIC
int sctp_getsockopt(struct sock
*sk
, int level
, int optname
,
5758 char __user
*optval
, int __user
*optlen
)
5763 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5766 /* I can hardly begin to describe how wrong this is. This is
5767 * so broken as to be worse than useless. The API draft
5768 * REALLY is NOT helpful here... I am not convinced that the
5769 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5770 * are at all well-founded.
5772 if (level
!= SOL_SCTP
) {
5773 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
5775 retval
= af
->getsockopt(sk
, level
, optname
, optval
, optlen
);
5779 if (get_user(len
, optlen
))
5789 retval
= sctp_getsockopt_sctp_status(sk
, len
, optval
, optlen
);
5791 case SCTP_DISABLE_FRAGMENTS
:
5792 retval
= sctp_getsockopt_disable_fragments(sk
, len
, optval
,
5796 retval
= sctp_getsockopt_events(sk
, len
, optval
, optlen
);
5798 case SCTP_AUTOCLOSE
:
5799 retval
= sctp_getsockopt_autoclose(sk
, len
, optval
, optlen
);
5801 case SCTP_SOCKOPT_PEELOFF
:
5802 retval
= sctp_getsockopt_peeloff(sk
, len
, optval
, optlen
);
5804 case SCTP_PEER_ADDR_PARAMS
:
5805 retval
= sctp_getsockopt_peer_addr_params(sk
, len
, optval
,
5808 case SCTP_DELAYED_SACK
:
5809 retval
= sctp_getsockopt_delayed_ack(sk
, len
, optval
,
5813 retval
= sctp_getsockopt_initmsg(sk
, len
, optval
, optlen
);
5815 case SCTP_GET_PEER_ADDRS
:
5816 retval
= sctp_getsockopt_peer_addrs(sk
, len
, optval
,
5819 case SCTP_GET_LOCAL_ADDRS
:
5820 retval
= sctp_getsockopt_local_addrs(sk
, len
, optval
,
5823 case SCTP_SOCKOPT_CONNECTX3
:
5824 retval
= sctp_getsockopt_connectx3(sk
, len
, optval
, optlen
);
5826 case SCTP_DEFAULT_SEND_PARAM
:
5827 retval
= sctp_getsockopt_default_send_param(sk
, len
,
5830 case SCTP_PRIMARY_ADDR
:
5831 retval
= sctp_getsockopt_primary_addr(sk
, len
, optval
, optlen
);
5834 retval
= sctp_getsockopt_nodelay(sk
, len
, optval
, optlen
);
5837 retval
= sctp_getsockopt_rtoinfo(sk
, len
, optval
, optlen
);
5839 case SCTP_ASSOCINFO
:
5840 retval
= sctp_getsockopt_associnfo(sk
, len
, optval
, optlen
);
5842 case SCTP_I_WANT_MAPPED_V4_ADDR
:
5843 retval
= sctp_getsockopt_mappedv4(sk
, len
, optval
, optlen
);
5846 retval
= sctp_getsockopt_maxseg(sk
, len
, optval
, optlen
);
5848 case SCTP_GET_PEER_ADDR_INFO
:
5849 retval
= sctp_getsockopt_peer_addr_info(sk
, len
, optval
,
5852 case SCTP_ADAPTATION_LAYER
:
5853 retval
= sctp_getsockopt_adaptation_layer(sk
, len
, optval
,
5857 retval
= sctp_getsockopt_context(sk
, len
, optval
, optlen
);
5859 case SCTP_FRAGMENT_INTERLEAVE
:
5860 retval
= sctp_getsockopt_fragment_interleave(sk
, len
, optval
,
5863 case SCTP_PARTIAL_DELIVERY_POINT
:
5864 retval
= sctp_getsockopt_partial_delivery_point(sk
, len
, optval
,
5867 case SCTP_MAX_BURST
:
5868 retval
= sctp_getsockopt_maxburst(sk
, len
, optval
, optlen
);
5871 case SCTP_AUTH_CHUNK
:
5872 case SCTP_AUTH_DELETE_KEY
:
5873 retval
= -EOPNOTSUPP
;
5875 case SCTP_HMAC_IDENT
:
5876 retval
= sctp_getsockopt_hmac_ident(sk
, len
, optval
, optlen
);
5878 case SCTP_AUTH_ACTIVE_KEY
:
5879 retval
= sctp_getsockopt_active_key(sk
, len
, optval
, optlen
);
5881 case SCTP_PEER_AUTH_CHUNKS
:
5882 retval
= sctp_getsockopt_peer_auth_chunks(sk
, len
, optval
,
5885 case SCTP_LOCAL_AUTH_CHUNKS
:
5886 retval
= sctp_getsockopt_local_auth_chunks(sk
, len
, optval
,
5889 case SCTP_GET_ASSOC_NUMBER
:
5890 retval
= sctp_getsockopt_assoc_number(sk
, len
, optval
, optlen
);
5892 case SCTP_GET_ASSOC_ID_LIST
:
5893 retval
= sctp_getsockopt_assoc_ids(sk
, len
, optval
, optlen
);
5895 case SCTP_AUTO_ASCONF
:
5896 retval
= sctp_getsockopt_auto_asconf(sk
, len
, optval
, optlen
);
5898 case SCTP_PEER_ADDR_THLDS
:
5899 retval
= sctp_getsockopt_paddr_thresholds(sk
, optval
, len
, optlen
);
5901 case SCTP_GET_ASSOC_STATS
:
5902 retval
= sctp_getsockopt_assoc_stats(sk
, len
, optval
, optlen
);
5905 retval
= -ENOPROTOOPT
;
5909 sctp_release_sock(sk
);
5913 static void sctp_hash(struct sock
*sk
)
5918 static void sctp_unhash(struct sock
*sk
)
5923 /* Check if port is acceptable. Possibly find first available port.
5925 * The port hash table (contained in the 'global' SCTP protocol storage
5926 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5927 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5928 * list (the list number is the port number hashed out, so as you
5929 * would expect from a hash function, all the ports in a given list have
5930 * such a number that hashes out to the same list number; you were
5931 * expecting that, right?); so each list has a set of ports, with a
5932 * link to the socket (struct sock) that uses it, the port number and
5933 * a fastreuse flag (FIXME: NPI ipg).
5935 static struct sctp_bind_bucket
*sctp_bucket_create(
5936 struct sctp_bind_hashbucket
*head
, struct net
*, unsigned short snum
);
5938 static long sctp_get_port_local(struct sock
*sk
, union sctp_addr
*addr
)
5940 struct sctp_bind_hashbucket
*head
; /* hash list */
5941 struct sctp_bind_bucket
*pp
;
5942 unsigned short snum
;
5945 snum
= ntohs(addr
->v4
.sin_port
);
5947 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum
);
5948 sctp_local_bh_disable();
5951 /* Search for an available port. */
5952 int low
, high
, remaining
, index
;
5955 inet_get_local_port_range(&low
, &high
);
5956 remaining
= (high
- low
) + 1;
5957 rover
= net_random() % remaining
+ low
;
5961 if ((rover
< low
) || (rover
> high
))
5963 if (inet_is_reserved_local_port(rover
))
5965 index
= sctp_phashfn(sock_net(sk
), rover
);
5966 head
= &sctp_port_hashtable
[index
];
5967 sctp_spin_lock(&head
->lock
);
5968 sctp_for_each_hentry(pp
, &head
->chain
)
5969 if ((pp
->port
== rover
) &&
5970 net_eq(sock_net(sk
), pp
->net
))
5974 sctp_spin_unlock(&head
->lock
);
5975 } while (--remaining
> 0);
5977 /* Exhausted local port range during search? */
5982 /* OK, here is the one we will use. HEAD (the port
5983 * hash table list entry) is non-NULL and we hold it's
5988 /* We are given an specific port number; we verify
5989 * that it is not being used. If it is used, we will
5990 * exahust the search in the hash list corresponding
5991 * to the port number (snum) - we detect that with the
5992 * port iterator, pp being NULL.
5994 head
= &sctp_port_hashtable
[sctp_phashfn(sock_net(sk
), snum
)];
5995 sctp_spin_lock(&head
->lock
);
5996 sctp_for_each_hentry(pp
, &head
->chain
) {
5997 if ((pp
->port
== snum
) && net_eq(pp
->net
, sock_net(sk
)))
6004 if (!hlist_empty(&pp
->owner
)) {
6005 /* We had a port hash table hit - there is an
6006 * available port (pp != NULL) and it is being
6007 * used by other socket (pp->owner not empty); that other
6008 * socket is going to be sk2.
6010 int reuse
= sk
->sk_reuse
;
6013 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
6014 if (pp
->fastreuse
&& sk
->sk_reuse
&&
6015 sk
->sk_state
!= SCTP_SS_LISTENING
)
6018 /* Run through the list of sockets bound to the port
6019 * (pp->port) [via the pointers bind_next and
6020 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6021 * we get the endpoint they describe and run through
6022 * the endpoint's list of IP (v4 or v6) addresses,
6023 * comparing each of the addresses with the address of
6024 * the socket sk. If we find a match, then that means
6025 * that this port/socket (sk) combination are already
6028 sk_for_each_bound(sk2
, &pp
->owner
) {
6029 struct sctp_endpoint
*ep2
;
6030 ep2
= sctp_sk(sk2
)->ep
;
6033 (reuse
&& sk2
->sk_reuse
&&
6034 sk2
->sk_state
!= SCTP_SS_LISTENING
))
6037 if (sctp_bind_addr_conflict(&ep2
->base
.bind_addr
, addr
,
6038 sctp_sk(sk2
), sctp_sk(sk
))) {
6043 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
6046 /* If there was a hash table miss, create a new port. */
6048 if (!pp
&& !(pp
= sctp_bucket_create(head
, sock_net(sk
), snum
)))
6051 /* In either case (hit or miss), make sure fastreuse is 1 only
6052 * if sk->sk_reuse is too (that is, if the caller requested
6053 * SO_REUSEADDR on this socket -sk-).
6055 if (hlist_empty(&pp
->owner
)) {
6056 if (sk
->sk_reuse
&& sk
->sk_state
!= SCTP_SS_LISTENING
)
6060 } else if (pp
->fastreuse
&&
6061 (!sk
->sk_reuse
|| sk
->sk_state
== SCTP_SS_LISTENING
))
6064 /* We are set, so fill up all the data in the hash table
6065 * entry, tie the socket list information with the rest of the
6066 * sockets FIXME: Blurry, NPI (ipg).
6069 if (!sctp_sk(sk
)->bind_hash
) {
6070 inet_sk(sk
)->inet_num
= snum
;
6071 sk_add_bind_node(sk
, &pp
->owner
);
6072 sctp_sk(sk
)->bind_hash
= pp
;
6077 sctp_spin_unlock(&head
->lock
);
6080 sctp_local_bh_enable();
6084 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6085 * port is requested.
6087 static int sctp_get_port(struct sock
*sk
, unsigned short snum
)
6090 union sctp_addr addr
;
6091 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
6093 /* Set up a dummy address struct from the sk. */
6094 af
->from_sk(&addr
, sk
);
6095 addr
.v4
.sin_port
= htons(snum
);
6097 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6098 ret
= sctp_get_port_local(sk
, &addr
);
6104 * Move a socket to LISTENING state.
6106 SCTP_STATIC
int sctp_listen_start(struct sock
*sk
, int backlog
)
6108 struct sctp_sock
*sp
= sctp_sk(sk
);
6109 struct sctp_endpoint
*ep
= sp
->ep
;
6110 struct crypto_hash
*tfm
= NULL
;
6113 /* Allocate HMAC for generating cookie. */
6114 if (!sp
->hmac
&& sp
->sctp_hmac_alg
) {
6115 sprintf(alg
, "hmac(%s)", sp
->sctp_hmac_alg
);
6116 tfm
= crypto_alloc_hash(alg
, 0, CRYPTO_ALG_ASYNC
);
6118 net_info_ratelimited("failed to load transform for %s: %ld\n",
6119 sp
->sctp_hmac_alg
, PTR_ERR(tfm
));
6122 sctp_sk(sk
)->hmac
= tfm
;
6126 * If a bind() or sctp_bindx() is not called prior to a listen()
6127 * call that allows new associations to be accepted, the system
6128 * picks an ephemeral port and will choose an address set equivalent
6129 * to binding with a wildcard address.
6131 * This is not currently spelled out in the SCTP sockets
6132 * extensions draft, but follows the practice as seen in TCP
6136 sk
->sk_state
= SCTP_SS_LISTENING
;
6137 if (!ep
->base
.bind_addr
.port
) {
6138 if (sctp_autobind(sk
))
6141 if (sctp_get_port(sk
, inet_sk(sk
)->inet_num
)) {
6142 sk
->sk_state
= SCTP_SS_CLOSED
;
6147 sk
->sk_max_ack_backlog
= backlog
;
6148 sctp_hash_endpoint(ep
);
6153 * 4.1.3 / 5.1.3 listen()
6155 * By default, new associations are not accepted for UDP style sockets.
6156 * An application uses listen() to mark a socket as being able to
6157 * accept new associations.
6159 * On TCP style sockets, applications use listen() to ready the SCTP
6160 * endpoint for accepting inbound associations.
6162 * On both types of endpoints a backlog of '0' disables listening.
6164 * Move a socket to LISTENING state.
6166 int sctp_inet_listen(struct socket
*sock
, int backlog
)
6168 struct sock
*sk
= sock
->sk
;
6169 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
6172 if (unlikely(backlog
< 0))
6177 /* Peeled-off sockets are not allowed to listen(). */
6178 if (sctp_style(sk
, UDP_HIGH_BANDWIDTH
))
6181 if (sock
->state
!= SS_UNCONNECTED
)
6184 if (!sctp_sstate(sk
, LISTENING
) && !sctp_sstate(sk
, CLOSED
))
6187 /* If backlog is zero, disable listening. */
6189 if (sctp_sstate(sk
, CLOSED
))
6193 sctp_unhash_endpoint(ep
);
6194 sk
->sk_state
= SCTP_SS_CLOSED
;
6196 sctp_sk(sk
)->bind_hash
->fastreuse
= 1;
6200 /* If we are already listening, just update the backlog */
6201 if (sctp_sstate(sk
, LISTENING
))
6202 sk
->sk_max_ack_backlog
= backlog
;
6204 err
= sctp_listen_start(sk
, backlog
);
6211 sctp_release_sock(sk
);
6216 * This function is done by modeling the current datagram_poll() and the
6217 * tcp_poll(). Note that, based on these implementations, we don't
6218 * lock the socket in this function, even though it seems that,
6219 * ideally, locking or some other mechanisms can be used to ensure
6220 * the integrity of the counters (sndbuf and wmem_alloc) used
6221 * in this place. We assume that we don't need locks either until proven
6224 * Another thing to note is that we include the Async I/O support
6225 * here, again, by modeling the current TCP/UDP code. We don't have
6226 * a good way to test with it yet.
6228 unsigned int sctp_poll(struct file
*file
, struct socket
*sock
, poll_table
*wait
)
6230 struct sock
*sk
= sock
->sk
;
6231 struct sctp_sock
*sp
= sctp_sk(sk
);
6234 poll_wait(file
, sk_sleep(sk
), wait
);
6236 /* A TCP-style listening socket becomes readable when the accept queue
6239 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
6240 return (!list_empty(&sp
->ep
->asocs
)) ?
6241 (POLLIN
| POLLRDNORM
) : 0;
6245 /* Is there any exceptional events? */
6246 if (sk
->sk_err
|| !skb_queue_empty(&sk
->sk_error_queue
))
6248 (sock_flag(sk
, SOCK_SELECT_ERR_QUEUE
) ? POLLPRI
: 0);
6249 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6250 mask
|= POLLRDHUP
| POLLIN
| POLLRDNORM
;
6251 if (sk
->sk_shutdown
== SHUTDOWN_MASK
)
6254 /* Is it readable? Reconsider this code with TCP-style support. */
6255 if (!skb_queue_empty(&sk
->sk_receive_queue
))
6256 mask
|= POLLIN
| POLLRDNORM
;
6258 /* The association is either gone or not ready. */
6259 if (!sctp_style(sk
, UDP
) && sctp_sstate(sk
, CLOSED
))
6262 /* Is it writable? */
6263 if (sctp_writeable(sk
)) {
6264 mask
|= POLLOUT
| POLLWRNORM
;
6266 set_bit(SOCK_ASYNC_NOSPACE
, &sk
->sk_socket
->flags
);
6268 * Since the socket is not locked, the buffer
6269 * might be made available after the writeable check and
6270 * before the bit is set. This could cause a lost I/O
6271 * signal. tcp_poll() has a race breaker for this race
6272 * condition. Based on their implementation, we put
6273 * in the following code to cover it as well.
6275 if (sctp_writeable(sk
))
6276 mask
|= POLLOUT
| POLLWRNORM
;
6281 /********************************************************************
6282 * 2nd Level Abstractions
6283 ********************************************************************/
6285 static struct sctp_bind_bucket
*sctp_bucket_create(
6286 struct sctp_bind_hashbucket
*head
, struct net
*net
, unsigned short snum
)
6288 struct sctp_bind_bucket
*pp
;
6290 pp
= kmem_cache_alloc(sctp_bucket_cachep
, GFP_ATOMIC
);
6292 SCTP_DBG_OBJCNT_INC(bind_bucket
);
6295 INIT_HLIST_HEAD(&pp
->owner
);
6297 hlist_add_head(&pp
->node
, &head
->chain
);
6302 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6303 static void sctp_bucket_destroy(struct sctp_bind_bucket
*pp
)
6305 if (pp
&& hlist_empty(&pp
->owner
)) {
6306 __hlist_del(&pp
->node
);
6307 kmem_cache_free(sctp_bucket_cachep
, pp
);
6308 SCTP_DBG_OBJCNT_DEC(bind_bucket
);
6312 /* Release this socket's reference to a local port. */
6313 static inline void __sctp_put_port(struct sock
*sk
)
6315 struct sctp_bind_hashbucket
*head
=
6316 &sctp_port_hashtable
[sctp_phashfn(sock_net(sk
),
6317 inet_sk(sk
)->inet_num
)];
6318 struct sctp_bind_bucket
*pp
;
6320 sctp_spin_lock(&head
->lock
);
6321 pp
= sctp_sk(sk
)->bind_hash
;
6322 __sk_del_bind_node(sk
);
6323 sctp_sk(sk
)->bind_hash
= NULL
;
6324 inet_sk(sk
)->inet_num
= 0;
6325 sctp_bucket_destroy(pp
);
6326 sctp_spin_unlock(&head
->lock
);
6329 void sctp_put_port(struct sock
*sk
)
6331 sctp_local_bh_disable();
6332 __sctp_put_port(sk
);
6333 sctp_local_bh_enable();
6337 * The system picks an ephemeral port and choose an address set equivalent
6338 * to binding with a wildcard address.
6339 * One of those addresses will be the primary address for the association.
6340 * This automatically enables the multihoming capability of SCTP.
6342 static int sctp_autobind(struct sock
*sk
)
6344 union sctp_addr autoaddr
;
6348 /* Initialize a local sockaddr structure to INADDR_ANY. */
6349 af
= sctp_sk(sk
)->pf
->af
;
6351 port
= htons(inet_sk(sk
)->inet_num
);
6352 af
->inaddr_any(&autoaddr
, port
);
6354 return sctp_do_bind(sk
, &autoaddr
, af
->sockaddr_len
);
6357 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6360 * 4.2 The cmsghdr Structure *
6362 * When ancillary data is sent or received, any number of ancillary data
6363 * objects can be specified by the msg_control and msg_controllen members of
6364 * the msghdr structure, because each object is preceded by
6365 * a cmsghdr structure defining the object's length (the cmsg_len member).
6366 * Historically Berkeley-derived implementations have passed only one object
6367 * at a time, but this API allows multiple objects to be
6368 * passed in a single call to sendmsg() or recvmsg(). The following example
6369 * shows two ancillary data objects in a control buffer.
6371 * |<--------------------------- msg_controllen -------------------------->|
6374 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6376 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6379 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6381 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6384 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6385 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6387 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6389 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6396 SCTP_STATIC
int sctp_msghdr_parse(const struct msghdr
*msg
,
6397 sctp_cmsgs_t
*cmsgs
)
6399 struct cmsghdr
*cmsg
;
6400 struct msghdr
*my_msg
= (struct msghdr
*)msg
;
6402 for (cmsg
= CMSG_FIRSTHDR(msg
);
6404 cmsg
= CMSG_NXTHDR(my_msg
, cmsg
)) {
6405 if (!CMSG_OK(my_msg
, cmsg
))
6408 /* Should we parse this header or ignore? */
6409 if (cmsg
->cmsg_level
!= IPPROTO_SCTP
)
6412 /* Strictly check lengths following example in SCM code. */
6413 switch (cmsg
->cmsg_type
) {
6415 /* SCTP Socket API Extension
6416 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6418 * This cmsghdr structure provides information for
6419 * initializing new SCTP associations with sendmsg().
6420 * The SCTP_INITMSG socket option uses this same data
6421 * structure. This structure is not used for
6424 * cmsg_level cmsg_type cmsg_data[]
6425 * ------------ ------------ ----------------------
6426 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6428 if (cmsg
->cmsg_len
!=
6429 CMSG_LEN(sizeof(struct sctp_initmsg
)))
6431 cmsgs
->init
= (struct sctp_initmsg
*)CMSG_DATA(cmsg
);
6435 /* SCTP Socket API Extension
6436 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6438 * This cmsghdr structure specifies SCTP options for
6439 * sendmsg() and describes SCTP header information
6440 * about a received message through recvmsg().
6442 * cmsg_level cmsg_type cmsg_data[]
6443 * ------------ ------------ ----------------------
6444 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6446 if (cmsg
->cmsg_len
!=
6447 CMSG_LEN(sizeof(struct sctp_sndrcvinfo
)))
6451 (struct sctp_sndrcvinfo
*)CMSG_DATA(cmsg
);
6453 /* Minimally, validate the sinfo_flags. */
6454 if (cmsgs
->info
->sinfo_flags
&
6455 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
6456 SCTP_ABORT
| SCTP_EOF
))
6468 * Wait for a packet..
6469 * Note: This function is the same function as in core/datagram.c
6470 * with a few modifications to make lksctp work.
6472 static int sctp_wait_for_packet(struct sock
* sk
, int *err
, long *timeo_p
)
6477 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
6479 /* Socket errors? */
6480 error
= sock_error(sk
);
6484 if (!skb_queue_empty(&sk
->sk_receive_queue
))
6487 /* Socket shut down? */
6488 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6491 /* Sequenced packets can come disconnected. If so we report the
6496 /* Is there a good reason to think that we may receive some data? */
6497 if (list_empty(&sctp_sk(sk
)->ep
->asocs
) && !sctp_sstate(sk
, LISTENING
))
6500 /* Handle signals. */
6501 if (signal_pending(current
))
6504 /* Let another process have a go. Since we are going to sleep
6505 * anyway. Note: This may cause odd behaviors if the message
6506 * does not fit in the user's buffer, but this seems to be the
6507 * only way to honor MSG_DONTWAIT realistically.
6509 sctp_release_sock(sk
);
6510 *timeo_p
= schedule_timeout(*timeo_p
);
6514 finish_wait(sk_sleep(sk
), &wait
);
6518 error
= sock_intr_errno(*timeo_p
);
6521 finish_wait(sk_sleep(sk
), &wait
);
6526 /* Receive a datagram.
6527 * Note: This is pretty much the same routine as in core/datagram.c
6528 * with a few changes to make lksctp work.
6530 static struct sk_buff
*sctp_skb_recv_datagram(struct sock
*sk
, int flags
,
6531 int noblock
, int *err
)
6534 struct sk_buff
*skb
;
6537 timeo
= sock_rcvtimeo(sk
, noblock
);
6539 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6540 timeo
, MAX_SCHEDULE_TIMEOUT
);
6543 /* Again only user level code calls this function,
6544 * so nothing interrupt level
6545 * will suddenly eat the receive_queue.
6547 * Look at current nfs client by the way...
6548 * However, this function was correct in any case. 8)
6550 if (flags
& MSG_PEEK
) {
6551 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
6552 skb
= skb_peek(&sk
->sk_receive_queue
);
6554 atomic_inc(&skb
->users
);
6555 spin_unlock_bh(&sk
->sk_receive_queue
.lock
);
6557 skb
= skb_dequeue(&sk
->sk_receive_queue
);
6563 /* Caller is allowed not to check sk->sk_err before calling. */
6564 error
= sock_error(sk
);
6568 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6571 /* User doesn't want to wait. */
6575 } while (sctp_wait_for_packet(sk
, err
, &timeo
) == 0);
6584 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6585 static void __sctp_write_space(struct sctp_association
*asoc
)
6587 struct sock
*sk
= asoc
->base
.sk
;
6588 struct socket
*sock
= sk
->sk_socket
;
6590 if ((sctp_wspace(asoc
) > 0) && sock
) {
6591 if (waitqueue_active(&asoc
->wait
))
6592 wake_up_interruptible(&asoc
->wait
);
6594 if (sctp_writeable(sk
)) {
6595 wait_queue_head_t
*wq
= sk_sleep(sk
);
6597 if (wq
&& waitqueue_active(wq
))
6598 wake_up_interruptible(wq
);
6600 /* Note that we try to include the Async I/O support
6601 * here by modeling from the current TCP/UDP code.
6602 * We have not tested with it yet.
6604 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
))
6605 sock_wake_async(sock
,
6606 SOCK_WAKE_SPACE
, POLL_OUT
);
6611 static void sctp_wake_up_waiters(struct sock
*sk
,
6612 struct sctp_association
*asoc
)
6614 struct sctp_association
*tmp
= asoc
;
6616 /* We do accounting for the sndbuf space per association,
6617 * so we only need to wake our own association.
6619 if (asoc
->ep
->sndbuf_policy
)
6620 return __sctp_write_space(asoc
);
6622 /* If association goes down and is just flushing its
6623 * outq, then just normally notify others.
6625 if (asoc
->base
.dead
)
6626 return sctp_write_space(sk
);
6628 /* Accounting for the sndbuf space is per socket, so we
6629 * need to wake up others, try to be fair and in case of
6630 * other associations, let them have a go first instead
6631 * of just doing a sctp_write_space() call.
6633 * Note that we reach sctp_wake_up_waiters() only when
6634 * associations free up queued chunks, thus we are under
6635 * lock and the list of associations on a socket is
6636 * guaranteed not to change.
6638 for (tmp
= list_next_entry(tmp
, asocs
); 1;
6639 tmp
= list_next_entry(tmp
, asocs
)) {
6640 /* Manually skip the head element. */
6641 if (&tmp
->asocs
== &((sctp_sk(sk
))->ep
->asocs
))
6643 /* Wake up association. */
6644 __sctp_write_space(tmp
);
6645 /* We've reached the end. */
6651 /* Do accounting for the sndbuf space.
6652 * Decrement the used sndbuf space of the corresponding association by the
6653 * data size which was just transmitted(freed).
6655 static void sctp_wfree(struct sk_buff
*skb
)
6657 struct sctp_association
*asoc
;
6658 struct sctp_chunk
*chunk
;
6661 /* Get the saved chunk pointer. */
6662 chunk
= *((struct sctp_chunk
**)(skb
->cb
));
6665 asoc
->sndbuf_used
-= SCTP_DATA_SNDSIZE(chunk
) +
6666 sizeof(struct sk_buff
) +
6667 sizeof(struct sctp_chunk
);
6669 atomic_sub(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
6672 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6674 sk
->sk_wmem_queued
-= skb
->truesize
;
6675 sk_mem_uncharge(sk
, skb
->truesize
);
6678 sctp_wake_up_waiters(sk
, asoc
);
6680 sctp_association_put(asoc
);
6683 /* Do accounting for the receive space on the socket.
6684 * Accounting for the association is done in ulpevent.c
6685 * We set this as a destructor for the cloned data skbs so that
6686 * accounting is done at the correct time.
6688 void sctp_sock_rfree(struct sk_buff
*skb
)
6690 struct sock
*sk
= skb
->sk
;
6691 struct sctp_ulpevent
*event
= sctp_skb2event(skb
);
6693 atomic_sub(event
->rmem_len
, &sk
->sk_rmem_alloc
);
6696 * Mimic the behavior of sock_rfree
6698 sk_mem_uncharge(sk
, event
->rmem_len
);
6702 /* Helper function to wait for space in the sndbuf. */
6703 static int sctp_wait_for_sndbuf(struct sctp_association
*asoc
, long *timeo_p
,
6706 struct sock
*sk
= asoc
->base
.sk
;
6708 long current_timeo
= *timeo_p
;
6711 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6712 asoc
, (long)(*timeo_p
), msg_len
);
6714 /* Increment the association's refcnt. */
6715 sctp_association_hold(asoc
);
6717 /* Wait on the association specific sndbuf space. */
6719 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
6720 TASK_INTERRUPTIBLE
);
6723 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
6726 if (signal_pending(current
))
6727 goto do_interrupted
;
6728 if (msg_len
<= sctp_wspace(asoc
))
6731 /* Let another process have a go. Since we are going
6734 sctp_release_sock(sk
);
6735 current_timeo
= schedule_timeout(current_timeo
);
6738 *timeo_p
= current_timeo
;
6742 finish_wait(&asoc
->wait
, &wait
);
6744 /* Release the association's refcnt. */
6745 sctp_association_put(asoc
);
6754 err
= sock_intr_errno(*timeo_p
);
6762 void sctp_data_ready(struct sock
*sk
, int len
)
6764 struct socket_wq
*wq
;
6767 wq
= rcu_dereference(sk
->sk_wq
);
6768 if (wq_has_sleeper(wq
))
6769 wake_up_interruptible_sync_poll(&wq
->wait
, POLLIN
|
6770 POLLRDNORM
| POLLRDBAND
);
6771 sk_wake_async(sk
, SOCK_WAKE_WAITD
, POLL_IN
);
6775 /* If socket sndbuf has changed, wake up all per association waiters. */
6776 void sctp_write_space(struct sock
*sk
)
6778 struct sctp_association
*asoc
;
6780 /* Wake up the tasks in each wait queue. */
6781 list_for_each_entry(asoc
, &((sctp_sk(sk
))->ep
->asocs
), asocs
) {
6782 __sctp_write_space(asoc
);
6786 /* Is there any sndbuf space available on the socket?
6788 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6789 * associations on the same socket. For a UDP-style socket with
6790 * multiple associations, it is possible for it to be "unwriteable"
6791 * prematurely. I assume that this is acceptable because
6792 * a premature "unwriteable" is better than an accidental "writeable" which
6793 * would cause an unwanted block under certain circumstances. For the 1-1
6794 * UDP-style sockets or TCP-style sockets, this code should work.
6797 static int sctp_writeable(struct sock
*sk
)
6801 amt
= sk
->sk_sndbuf
- sk_wmem_alloc_get(sk
);
6807 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6808 * returns immediately with EINPROGRESS.
6810 static int sctp_wait_for_connect(struct sctp_association
*asoc
, long *timeo_p
)
6812 struct sock
*sk
= asoc
->base
.sk
;
6814 long current_timeo
= *timeo_p
;
6817 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__
, asoc
,
6820 /* Increment the association's refcnt. */
6821 sctp_association_hold(asoc
);
6824 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
6825 TASK_INTERRUPTIBLE
);
6828 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6830 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
6833 if (signal_pending(current
))
6834 goto do_interrupted
;
6836 if (sctp_state(asoc
, ESTABLISHED
))
6839 /* Let another process have a go. Since we are going
6842 sctp_release_sock(sk
);
6843 current_timeo
= schedule_timeout(current_timeo
);
6846 *timeo_p
= current_timeo
;
6850 finish_wait(&asoc
->wait
, &wait
);
6852 /* Release the association's refcnt. */
6853 sctp_association_put(asoc
);
6858 if (asoc
->init_err_counter
+ 1 > asoc
->max_init_attempts
)
6861 err
= -ECONNREFUSED
;
6865 err
= sock_intr_errno(*timeo_p
);
6873 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
)
6875 struct sctp_endpoint
*ep
;
6879 ep
= sctp_sk(sk
)->ep
;
6883 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
,
6884 TASK_INTERRUPTIBLE
);
6886 if (list_empty(&ep
->asocs
)) {
6887 sctp_release_sock(sk
);
6888 timeo
= schedule_timeout(timeo
);
6893 if (!sctp_sstate(sk
, LISTENING
))
6897 if (!list_empty(&ep
->asocs
))
6900 err
= sock_intr_errno(timeo
);
6901 if (signal_pending(current
))
6909 finish_wait(sk_sleep(sk
), &wait
);
6914 static void sctp_wait_for_close(struct sock
*sk
, long timeout
)
6919 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
6920 if (list_empty(&sctp_sk(sk
)->ep
->asocs
))
6922 sctp_release_sock(sk
);
6923 timeout
= schedule_timeout(timeout
);
6925 } while (!signal_pending(current
) && timeout
);
6927 finish_wait(sk_sleep(sk
), &wait
);
6930 static void sctp_skb_set_owner_r_frag(struct sk_buff
*skb
, struct sock
*sk
)
6932 struct sk_buff
*frag
;
6937 /* Don't forget the fragments. */
6938 skb_walk_frags(skb
, frag
)
6939 sctp_skb_set_owner_r_frag(frag
, sk
);
6942 sctp_skb_set_owner_r(skb
, sk
);
6945 void sctp_copy_sock(struct sock
*newsk
, struct sock
*sk
,
6946 struct sctp_association
*asoc
)
6948 struct inet_sock
*inet
= inet_sk(sk
);
6949 struct inet_sock
*newinet
;
6951 newsk
->sk_type
= sk
->sk_type
;
6952 newsk
->sk_bound_dev_if
= sk
->sk_bound_dev_if
;
6953 newsk
->sk_flags
= sk
->sk_flags
;
6954 newsk
->sk_no_check
= sk
->sk_no_check
;
6955 newsk
->sk_reuse
= sk
->sk_reuse
;
6957 newsk
->sk_shutdown
= sk
->sk_shutdown
;
6958 newsk
->sk_destruct
= inet_sock_destruct
;
6959 newsk
->sk_family
= sk
->sk_family
;
6960 newsk
->sk_protocol
= IPPROTO_SCTP
;
6961 newsk
->sk_backlog_rcv
= sk
->sk_prot
->backlog_rcv
;
6962 newsk
->sk_sndbuf
= sk
->sk_sndbuf
;
6963 newsk
->sk_rcvbuf
= sk
->sk_rcvbuf
;
6964 newsk
->sk_lingertime
= sk
->sk_lingertime
;
6965 newsk
->sk_rcvtimeo
= sk
->sk_rcvtimeo
;
6966 newsk
->sk_sndtimeo
= sk
->sk_sndtimeo
;
6968 newinet
= inet_sk(newsk
);
6970 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6971 * getsockname() and getpeername()
6973 newinet
->inet_sport
= inet
->inet_sport
;
6974 newinet
->inet_saddr
= inet
->inet_saddr
;
6975 newinet
->inet_rcv_saddr
= inet
->inet_rcv_saddr
;
6976 newinet
->inet_dport
= htons(asoc
->peer
.port
);
6977 newinet
->pmtudisc
= inet
->pmtudisc
;
6978 newinet
->inet_id
= asoc
->next_tsn
^ jiffies
;
6980 newinet
->uc_ttl
= inet
->uc_ttl
;
6981 newinet
->mc_loop
= 1;
6982 newinet
->mc_ttl
= 1;
6983 newinet
->mc_index
= 0;
6984 newinet
->mc_list
= NULL
;
6986 if (newsk
->sk_flags
& SK_FLAGS_TIMESTAMP
)
6987 net_enable_timestamp();
6990 static inline void sctp_copy_descendant(struct sock
*sk_to
,
6991 const struct sock
*sk_from
)
6993 int ancestor_size
= sizeof(struct inet_sock
) +
6994 sizeof(struct sctp_sock
) -
6995 offsetof(struct sctp_sock
, auto_asconf_list
);
6997 if (sk_from
->sk_family
== PF_INET6
)
6998 ancestor_size
+= sizeof(struct ipv6_pinfo
);
7000 __inet_sk_copy_descendant(sk_to
, sk_from
, ancestor_size
);
7003 /* Populate the fields of the newsk from the oldsk and migrate the assoc
7004 * and its messages to the newsk.
7006 static void sctp_sock_migrate(struct sock
*oldsk
, struct sock
*newsk
,
7007 struct sctp_association
*assoc
,
7008 sctp_socket_type_t type
)
7010 struct sctp_sock
*oldsp
= sctp_sk(oldsk
);
7011 struct sctp_sock
*newsp
= sctp_sk(newsk
);
7012 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
7013 struct sctp_endpoint
*newep
= newsp
->ep
;
7014 struct sk_buff
*skb
, *tmp
;
7015 struct sctp_ulpevent
*event
;
7016 struct sctp_bind_hashbucket
*head
;
7018 /* Migrate socket buffer sizes and all the socket level options to the
7021 newsk
->sk_sndbuf
= oldsk
->sk_sndbuf
;
7022 newsk
->sk_rcvbuf
= oldsk
->sk_rcvbuf
;
7023 /* Brute force copy old sctp opt. */
7024 sctp_copy_descendant(newsk
, oldsk
);
7026 /* Restore the ep value that was overwritten with the above structure
7032 /* Hook this new socket in to the bind_hash list. */
7033 head
= &sctp_port_hashtable
[sctp_phashfn(sock_net(oldsk
),
7034 inet_sk(oldsk
)->inet_num
)];
7035 sctp_local_bh_disable();
7036 sctp_spin_lock(&head
->lock
);
7037 pp
= sctp_sk(oldsk
)->bind_hash
;
7038 sk_add_bind_node(newsk
, &pp
->owner
);
7039 sctp_sk(newsk
)->bind_hash
= pp
;
7040 inet_sk(newsk
)->inet_num
= inet_sk(oldsk
)->inet_num
;
7041 sctp_spin_unlock(&head
->lock
);
7042 sctp_local_bh_enable();
7044 /* Copy the bind_addr list from the original endpoint to the new
7045 * endpoint so that we can handle restarts properly
7047 sctp_bind_addr_dup(&newsp
->ep
->base
.bind_addr
,
7048 &oldsp
->ep
->base
.bind_addr
, GFP_KERNEL
);
7050 /* Move any messages in the old socket's receive queue that are for the
7051 * peeled off association to the new socket's receive queue.
7053 sctp_skb_for_each(skb
, &oldsk
->sk_receive_queue
, tmp
) {
7054 event
= sctp_skb2event(skb
);
7055 if (event
->asoc
== assoc
) {
7056 __skb_unlink(skb
, &oldsk
->sk_receive_queue
);
7057 __skb_queue_tail(&newsk
->sk_receive_queue
, skb
);
7058 sctp_skb_set_owner_r_frag(skb
, newsk
);
7062 /* Clean up any messages pending delivery due to partial
7063 * delivery. Three cases:
7064 * 1) No partial deliver; no work.
7065 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7066 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7068 skb_queue_head_init(&newsp
->pd_lobby
);
7069 atomic_set(&sctp_sk(newsk
)->pd_mode
, assoc
->ulpq
.pd_mode
);
7071 if (atomic_read(&sctp_sk(oldsk
)->pd_mode
)) {
7072 struct sk_buff_head
*queue
;
7074 /* Decide which queue to move pd_lobby skbs to. */
7075 if (assoc
->ulpq
.pd_mode
) {
7076 queue
= &newsp
->pd_lobby
;
7078 queue
= &newsk
->sk_receive_queue
;
7080 /* Walk through the pd_lobby, looking for skbs that
7081 * need moved to the new socket.
7083 sctp_skb_for_each(skb
, &oldsp
->pd_lobby
, tmp
) {
7084 event
= sctp_skb2event(skb
);
7085 if (event
->asoc
== assoc
) {
7086 __skb_unlink(skb
, &oldsp
->pd_lobby
);
7087 __skb_queue_tail(queue
, skb
);
7088 sctp_skb_set_owner_r_frag(skb
, newsk
);
7092 /* Clear up any skbs waiting for the partial
7093 * delivery to finish.
7095 if (assoc
->ulpq
.pd_mode
)
7096 sctp_clear_pd(oldsk
, NULL
);
7100 sctp_skb_for_each(skb
, &assoc
->ulpq
.reasm
, tmp
)
7101 sctp_skb_set_owner_r_frag(skb
, newsk
);
7103 sctp_skb_for_each(skb
, &assoc
->ulpq
.lobby
, tmp
)
7104 sctp_skb_set_owner_r_frag(skb
, newsk
);
7106 /* Set the type of socket to indicate that it is peeled off from the
7107 * original UDP-style socket or created with the accept() call on a
7108 * TCP-style socket..
7112 /* Mark the new socket "in-use" by the user so that any packets
7113 * that may arrive on the association after we've moved it are
7114 * queued to the backlog. This prevents a potential race between
7115 * backlog processing on the old socket and new-packet processing
7116 * on the new socket.
7118 * The caller has just allocated newsk so we can guarantee that other
7119 * paths won't try to lock it and then oldsk.
7121 lock_sock_nested(newsk
, SINGLE_DEPTH_NESTING
);
7122 sctp_assoc_migrate(assoc
, newsk
);
7124 /* If the association on the newsk is already closed before accept()
7125 * is called, set RCV_SHUTDOWN flag.
7127 if (sctp_state(assoc
, CLOSED
) && sctp_style(newsk
, TCP
))
7128 newsk
->sk_shutdown
|= RCV_SHUTDOWN
;
7130 newsk
->sk_state
= SCTP_SS_ESTABLISHED
;
7131 sctp_release_sock(newsk
);
7135 /* This proto struct describes the ULP interface for SCTP. */
7136 struct proto sctp_prot
= {
7138 .owner
= THIS_MODULE
,
7139 .close
= sctp_close
,
7140 .connect
= sctp_connect
,
7141 .disconnect
= sctp_disconnect
,
7142 .accept
= sctp_accept
,
7143 .ioctl
= sctp_ioctl
,
7144 .init
= sctp_init_sock
,
7145 .destroy
= sctp_destroy_sock
,
7146 .shutdown
= sctp_shutdown
,
7147 .setsockopt
= sctp_setsockopt
,
7148 .getsockopt
= sctp_getsockopt
,
7149 .sendmsg
= sctp_sendmsg
,
7150 .recvmsg
= sctp_recvmsg
,
7152 .backlog_rcv
= sctp_backlog_rcv
,
7154 .unhash
= sctp_unhash
,
7155 .get_port
= sctp_get_port
,
7156 .obj_size
= sizeof(struct sctp_sock
),
7157 .sysctl_mem
= sysctl_sctp_mem
,
7158 .sysctl_rmem
= sysctl_sctp_rmem
,
7159 .sysctl_wmem
= sysctl_sctp_wmem
,
7160 .memory_pressure
= &sctp_memory_pressure
,
7161 .enter_memory_pressure
= sctp_enter_memory_pressure
,
7162 .memory_allocated
= &sctp_memory_allocated
,
7163 .sockets_allocated
= &sctp_sockets_allocated
,
7166 #if IS_ENABLED(CONFIG_IPV6)
7168 #include <net/transp_v6.h>
7169 static void sctp_v6_destroy_sock(struct sock
*sk
)
7171 sctp_destroy_sock(sk
);
7172 inet6_destroy_sock(sk
);
7175 struct proto sctpv6_prot
= {
7177 .owner
= THIS_MODULE
,
7178 .close
= sctp_close
,
7179 .connect
= sctp_connect
,
7180 .disconnect
= sctp_disconnect
,
7181 .accept
= sctp_accept
,
7182 .ioctl
= sctp_ioctl
,
7183 .init
= sctp_init_sock
,
7184 .destroy
= sctp_v6_destroy_sock
,
7185 .shutdown
= sctp_shutdown
,
7186 .setsockopt
= sctp_setsockopt
,
7187 .getsockopt
= sctp_getsockopt
,
7188 .sendmsg
= sctp_sendmsg
,
7189 .recvmsg
= sctp_recvmsg
,
7191 .backlog_rcv
= sctp_backlog_rcv
,
7193 .unhash
= sctp_unhash
,
7194 .get_port
= sctp_get_port
,
7195 .obj_size
= sizeof(struct sctp6_sock
),
7196 .sysctl_mem
= sysctl_sctp_mem
,
7197 .sysctl_rmem
= sysctl_sctp_rmem
,
7198 .sysctl_wmem
= sysctl_sctp_wmem
,
7199 .memory_pressure
= &sctp_memory_pressure
,
7200 .enter_memory_pressure
= sctp_enter_memory_pressure
,
7201 .memory_allocated
= &sctp_memory_allocated
,
7202 .sockets_allocated
= &sctp_sockets_allocated
,
7204 #endif /* IS_ENABLED(CONFIG_IPV6) */