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, see
32 * <http://www.gnu.org/licenses/>.
34 * Please send any bug reports or fixes you make to the
36 * lksctp developers <linux-sctp@vger.kernel.org>
38 * Written or modified by:
39 * La Monte H.P. Yarroll <piggy@acm.org>
40 * Narasimha Budihal <narsi@refcode.org>
41 * Karl Knutson <karl@athena.chicago.il.us>
42 * Jon Grimm <jgrimm@us.ibm.com>
43 * Xingang Guo <xingang.guo@intel.com>
44 * Daisy Chang <daisyc@us.ibm.com>
45 * Sridhar Samudrala <samudrala@us.ibm.com>
46 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
47 * Ardelle Fan <ardelle.fan@intel.com>
48 * Ryan Layer <rmlayer@us.ibm.com>
49 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
50 * Kevin Gao <kevin.gao@intel.com>
53 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
55 #include <linux/types.h>
56 #include <linux/kernel.h>
57 #include <linux/wait.h>
58 #include <linux/time.h>
60 #include <linux/capability.h>
61 #include <linux/fcntl.h>
62 #include <linux/poll.h>
63 #include <linux/init.h>
64 #include <linux/crypto.h>
65 #include <linux/slab.h>
66 #include <linux/file.h>
67 #include <linux/compat.h>
71 #include <net/route.h>
73 #include <net/inet_common.h>
74 #include <net/busy_poll.h>
76 #include <linux/socket.h> /* for sa_family_t */
77 #include <linux/export.h>
79 #include <net/sctp/sctp.h>
80 #include <net/sctp/sm.h>
82 /* Forward declarations for internal helper functions. */
83 static int sctp_writeable(struct sock
*sk
);
84 static void sctp_wfree(struct sk_buff
*skb
);
85 static int sctp_wait_for_sndbuf(struct sctp_association
*, long *timeo_p
,
87 static int sctp_wait_for_packet(struct sock
*sk
, int *err
, long *timeo_p
);
88 static int sctp_wait_for_connect(struct sctp_association
*, long *timeo_p
);
89 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
);
90 static void sctp_wait_for_close(struct sock
*sk
, long timeo
);
91 static void sctp_destruct_sock(struct sock
*sk
);
92 static struct sctp_af
*sctp_sockaddr_af(struct sctp_sock
*opt
,
93 union sctp_addr
*addr
, int len
);
94 static int sctp_bindx_add(struct sock
*, struct sockaddr
*, int);
95 static int sctp_bindx_rem(struct sock
*, struct sockaddr
*, int);
96 static int sctp_send_asconf_add_ip(struct sock
*, struct sockaddr
*, int);
97 static int sctp_send_asconf_del_ip(struct sock
*, struct sockaddr
*, int);
98 static int sctp_send_asconf(struct sctp_association
*asoc
,
99 struct sctp_chunk
*chunk
);
100 static int sctp_do_bind(struct sock
*, union sctp_addr
*, int);
101 static int sctp_autobind(struct sock
*sk
);
102 static void sctp_sock_migrate(struct sock
*, struct sock
*,
103 struct sctp_association
*, sctp_socket_type_t
);
105 extern struct kmem_cache
*sctp_bucket_cachep
;
106 extern long sysctl_sctp_mem
[3];
107 extern int sysctl_sctp_rmem
[3];
108 extern int sysctl_sctp_wmem
[3];
110 static int sctp_memory_pressure
;
111 static atomic_long_t sctp_memory_allocated
;
112 struct percpu_counter sctp_sockets_allocated
;
114 static void sctp_enter_memory_pressure(struct sock
*sk
)
116 sctp_memory_pressure
= 1;
120 /* Get the sndbuf space available at the time on the association. */
121 static inline int sctp_wspace(struct sctp_association
*asoc
)
125 if (asoc
->ep
->sndbuf_policy
)
126 amt
= asoc
->sndbuf_used
;
128 amt
= sk_wmem_alloc_get(asoc
->base
.sk
);
130 if (amt
>= asoc
->base
.sk
->sk_sndbuf
) {
131 if (asoc
->base
.sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)
134 amt
= sk_stream_wspace(asoc
->base
.sk
);
139 amt
= asoc
->base
.sk
->sk_sndbuf
- amt
;
144 /* Increment the used sndbuf space count of the corresponding association by
145 * the size of the outgoing data chunk.
146 * Also, set the skb destructor for sndbuf accounting later.
148 * Since it is always 1-1 between chunk and skb, and also a new skb is always
149 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
150 * destructor in the data chunk skb for the purpose of the sndbuf space
153 static inline void sctp_set_owner_w(struct sctp_chunk
*chunk
)
155 struct sctp_association
*asoc
= chunk
->asoc
;
156 struct sock
*sk
= asoc
->base
.sk
;
158 /* The sndbuf space is tracked per association. */
159 sctp_association_hold(asoc
);
161 skb_set_owner_w(chunk
->skb
, sk
);
163 chunk
->skb
->destructor
= sctp_wfree
;
164 /* Save the chunk pointer in skb for sctp_wfree to use later. */
165 skb_shinfo(chunk
->skb
)->destructor_arg
= chunk
;
167 asoc
->sndbuf_used
+= SCTP_DATA_SNDSIZE(chunk
) +
168 sizeof(struct sk_buff
) +
169 sizeof(struct sctp_chunk
);
171 atomic_add(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
172 sk
->sk_wmem_queued
+= chunk
->skb
->truesize
;
173 sk_mem_charge(sk
, chunk
->skb
->truesize
);
176 /* Verify that this is a valid address. */
177 static inline int sctp_verify_addr(struct sock
*sk
, union sctp_addr
*addr
,
182 /* Verify basic sockaddr. */
183 af
= sctp_sockaddr_af(sctp_sk(sk
), addr
, len
);
187 /* Is this a valid SCTP address? */
188 if (!af
->addr_valid(addr
, sctp_sk(sk
), NULL
))
191 if (!sctp_sk(sk
)->pf
->send_verify(sctp_sk(sk
), (addr
)))
197 /* Look up the association by its id. If this is not a UDP-style
198 * socket, the ID field is always ignored.
200 struct sctp_association
*sctp_id2assoc(struct sock
*sk
, sctp_assoc_t id
)
202 struct sctp_association
*asoc
= NULL
;
204 /* If this is not a UDP-style socket, assoc id should be ignored. */
205 if (!sctp_style(sk
, UDP
)) {
206 /* Return NULL if the socket state is not ESTABLISHED. It
207 * could be a TCP-style listening socket or a socket which
208 * hasn't yet called connect() to establish an association.
210 if (!sctp_sstate(sk
, ESTABLISHED
))
213 /* Get the first and the only association from the list. */
214 if (!list_empty(&sctp_sk(sk
)->ep
->asocs
))
215 asoc
= list_entry(sctp_sk(sk
)->ep
->asocs
.next
,
216 struct sctp_association
, asocs
);
220 /* Otherwise this is a UDP-style socket. */
221 if (!id
|| (id
== (sctp_assoc_t
)-1))
224 spin_lock_bh(&sctp_assocs_id_lock
);
225 asoc
= (struct sctp_association
*)idr_find(&sctp_assocs_id
, (int)id
);
226 spin_unlock_bh(&sctp_assocs_id_lock
);
228 if (!asoc
|| (asoc
->base
.sk
!= sk
) || asoc
->base
.dead
)
234 /* Look up the transport from an address and an assoc id. If both address and
235 * id are specified, the associations matching the address and the id should be
238 static struct sctp_transport
*sctp_addr_id2transport(struct sock
*sk
,
239 struct sockaddr_storage
*addr
,
242 struct sctp_association
*addr_asoc
= NULL
, *id_asoc
= NULL
;
243 struct sctp_transport
*transport
;
244 union sctp_addr
*laddr
= (union sctp_addr
*)addr
;
246 addr_asoc
= sctp_endpoint_lookup_assoc(sctp_sk(sk
)->ep
,
253 id_asoc
= sctp_id2assoc(sk
, id
);
254 if (id_asoc
&& (id_asoc
!= addr_asoc
))
257 sctp_get_pf_specific(sk
->sk_family
)->addr_to_user(sctp_sk(sk
),
258 (union sctp_addr
*)addr
);
263 /* API 3.1.2 bind() - UDP Style Syntax
264 * The syntax of bind() is,
266 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
268 * sd - the socket descriptor returned by socket().
269 * addr - the address structure (struct sockaddr_in or struct
270 * sockaddr_in6 [RFC 2553]),
271 * addr_len - the size of the address structure.
273 static int sctp_bind(struct sock
*sk
, struct sockaddr
*addr
, int addr_len
)
279 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__
, sk
,
282 /* Disallow binding twice. */
283 if (!sctp_sk(sk
)->ep
->base
.bind_addr
.port
)
284 retval
= sctp_do_bind(sk
, (union sctp_addr
*)addr
,
294 static long sctp_get_port_local(struct sock
*, union sctp_addr
*);
296 /* Verify this is a valid sockaddr. */
297 static struct sctp_af
*sctp_sockaddr_af(struct sctp_sock
*opt
,
298 union sctp_addr
*addr
, int len
)
302 /* Check minimum size. */
303 if (len
< sizeof (struct sockaddr
))
306 /* V4 mapped address are really of AF_INET family */
307 if (addr
->sa
.sa_family
== AF_INET6
&&
308 ipv6_addr_v4mapped(&addr
->v6
.sin6_addr
)) {
309 if (!opt
->pf
->af_supported(AF_INET
, opt
))
312 /* Does this PF support this AF? */
313 if (!opt
->pf
->af_supported(addr
->sa
.sa_family
, opt
))
317 /* If we get this far, af is valid. */
318 af
= sctp_get_af_specific(addr
->sa
.sa_family
);
320 if (len
< af
->sockaddr_len
)
326 /* Bind a local address either to an endpoint or to an association. */
327 static int sctp_do_bind(struct sock
*sk
, union sctp_addr
*addr
, int len
)
329 struct net
*net
= sock_net(sk
);
330 struct sctp_sock
*sp
= sctp_sk(sk
);
331 struct sctp_endpoint
*ep
= sp
->ep
;
332 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
337 /* Common sockaddr verification. */
338 af
= sctp_sockaddr_af(sp
, addr
, len
);
340 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
341 __func__
, sk
, addr
, len
);
345 snum
= ntohs(addr
->v4
.sin_port
);
347 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
348 __func__
, sk
, &addr
->sa
, bp
->port
, snum
, len
);
350 /* PF specific bind() address verification. */
351 if (!sp
->pf
->bind_verify(sp
, addr
))
352 return -EADDRNOTAVAIL
;
354 /* We must either be unbound, or bind to the same port.
355 * It's OK to allow 0 ports if we are already bound.
356 * We'll just inhert an already bound port in this case
361 else if (snum
!= bp
->port
) {
362 pr_debug("%s: new port %d doesn't match existing port "
363 "%d\n", __func__
, snum
, bp
->port
);
368 if (snum
&& snum
< PROT_SOCK
&&
369 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
))
372 /* See if the address matches any of the addresses we may have
373 * already bound before checking against other endpoints.
375 if (sctp_bind_addr_match(bp
, addr
, sp
))
378 /* Make sure we are allowed to bind here.
379 * The function sctp_get_port_local() does duplicate address
382 addr
->v4
.sin_port
= htons(snum
);
383 if ((ret
= sctp_get_port_local(sk
, addr
))) {
387 /* Refresh ephemeral port. */
389 bp
->port
= inet_sk(sk
)->inet_num
;
391 /* Add the address to the bind address list.
392 * Use GFP_ATOMIC since BHs will be disabled.
394 ret
= sctp_add_bind_addr(bp
, addr
, SCTP_ADDR_SRC
, GFP_ATOMIC
);
396 /* Copy back into socket for getsockname() use. */
398 inet_sk(sk
)->inet_sport
= htons(inet_sk(sk
)->inet_num
);
399 sp
->pf
->to_sk_saddr(addr
, sk
);
405 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
407 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
408 * at any one time. If a sender, after sending an ASCONF chunk, decides
409 * it needs to transfer another ASCONF Chunk, it MUST wait until the
410 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
411 * subsequent ASCONF. Note this restriction binds each side, so at any
412 * time two ASCONF may be in-transit on any given association (one sent
413 * from each endpoint).
415 static int sctp_send_asconf(struct sctp_association
*asoc
,
416 struct sctp_chunk
*chunk
)
418 struct net
*net
= sock_net(asoc
->base
.sk
);
421 /* If there is an outstanding ASCONF chunk, queue it for later
424 if (asoc
->addip_last_asconf
) {
425 list_add_tail(&chunk
->list
, &asoc
->addip_chunk_list
);
429 /* Hold the chunk until an ASCONF_ACK is received. */
430 sctp_chunk_hold(chunk
);
431 retval
= sctp_primitive_ASCONF(net
, asoc
, chunk
);
433 sctp_chunk_free(chunk
);
435 asoc
->addip_last_asconf
= chunk
;
441 /* Add a list of addresses as bind addresses to local endpoint or
444 * Basically run through each address specified in the addrs/addrcnt
445 * array/length pair, determine if it is IPv6 or IPv4 and call
446 * sctp_do_bind() on it.
448 * If any of them fails, then the operation will be reversed and the
449 * ones that were added will be removed.
451 * Only sctp_setsockopt_bindx() is supposed to call this function.
453 static int sctp_bindx_add(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
458 struct sockaddr
*sa_addr
;
461 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__
, sk
,
465 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
466 /* The list may contain either IPv4 or IPv6 address;
467 * determine the address length for walking thru the list.
470 af
= sctp_get_af_specific(sa_addr
->sa_family
);
476 retval
= sctp_do_bind(sk
, (union sctp_addr
*)sa_addr
,
479 addr_buf
+= af
->sockaddr_len
;
483 /* Failed. Cleanup the ones that have been added */
485 sctp_bindx_rem(sk
, addrs
, cnt
);
493 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
494 * associations that are part of the endpoint indicating that a list of local
495 * addresses are added to the endpoint.
497 * If any of the addresses is already in the bind address list of the
498 * association, we do not send the chunk for that association. But it will not
499 * affect other associations.
501 * Only sctp_setsockopt_bindx() is supposed to call this function.
503 static int sctp_send_asconf_add_ip(struct sock
*sk
,
504 struct sockaddr
*addrs
,
507 struct net
*net
= sock_net(sk
);
508 struct sctp_sock
*sp
;
509 struct sctp_endpoint
*ep
;
510 struct sctp_association
*asoc
;
511 struct sctp_bind_addr
*bp
;
512 struct sctp_chunk
*chunk
;
513 struct sctp_sockaddr_entry
*laddr
;
514 union sctp_addr
*addr
;
515 union sctp_addr saveaddr
;
522 if (!net
->sctp
.addip_enable
)
528 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
529 __func__
, sk
, addrs
, addrcnt
);
531 list_for_each_entry(asoc
, &ep
->asocs
, asocs
) {
532 if (!asoc
->peer
.asconf_capable
)
535 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_ADD_IP
)
538 if (!sctp_state(asoc
, ESTABLISHED
))
541 /* Check if any address in the packed array of addresses is
542 * in the bind address list of the association. If so,
543 * do not send the asconf chunk to its peer, but continue with
544 * other associations.
547 for (i
= 0; i
< addrcnt
; i
++) {
549 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
555 if (sctp_assoc_lookup_laddr(asoc
, addr
))
558 addr_buf
+= af
->sockaddr_len
;
563 /* Use the first valid address in bind addr list of
564 * association as Address Parameter of ASCONF CHUNK.
566 bp
= &asoc
->base
.bind_addr
;
567 p
= bp
->address_list
.next
;
568 laddr
= list_entry(p
, struct sctp_sockaddr_entry
, list
);
569 chunk
= sctp_make_asconf_update_ip(asoc
, &laddr
->a
, addrs
,
570 addrcnt
, SCTP_PARAM_ADD_IP
);
576 /* Add the new addresses to the bind address list with
577 * use_as_src set to 0.
580 for (i
= 0; i
< addrcnt
; i
++) {
582 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
583 memcpy(&saveaddr
, addr
, af
->sockaddr_len
);
584 retval
= sctp_add_bind_addr(bp
, &saveaddr
,
585 SCTP_ADDR_NEW
, GFP_ATOMIC
);
586 addr_buf
+= af
->sockaddr_len
;
588 if (asoc
->src_out_of_asoc_ok
) {
589 struct sctp_transport
*trans
;
591 list_for_each_entry(trans
,
592 &asoc
->peer
.transport_addr_list
, transports
) {
593 /* Clear the source and route cache */
594 dst_release(trans
->dst
);
595 trans
->cwnd
= min(4*asoc
->pathmtu
, max_t(__u32
,
596 2*asoc
->pathmtu
, 4380));
597 trans
->ssthresh
= asoc
->peer
.i
.a_rwnd
;
598 trans
->rto
= asoc
->rto_initial
;
599 sctp_max_rto(asoc
, trans
);
600 trans
->rtt
= trans
->srtt
= trans
->rttvar
= 0;
601 sctp_transport_route(trans
, NULL
,
602 sctp_sk(asoc
->base
.sk
));
605 retval
= sctp_send_asconf(asoc
, chunk
);
612 /* Remove a list of addresses from bind addresses list. Do not remove the
615 * Basically run through each address specified in the addrs/addrcnt
616 * array/length pair, determine if it is IPv6 or IPv4 and call
617 * sctp_del_bind() on it.
619 * If any of them fails, then the operation will be reversed and the
620 * ones that were removed will be added back.
622 * At least one address has to be left; if only one address is
623 * available, the operation will return -EBUSY.
625 * Only sctp_setsockopt_bindx() is supposed to call this function.
627 static int sctp_bindx_rem(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
629 struct sctp_sock
*sp
= sctp_sk(sk
);
630 struct sctp_endpoint
*ep
= sp
->ep
;
632 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
635 union sctp_addr
*sa_addr
;
638 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
639 __func__
, sk
, addrs
, addrcnt
);
642 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
643 /* If the bind address list is empty or if there is only one
644 * bind address, there is nothing more to be removed (we need
645 * at least one address here).
647 if (list_empty(&bp
->address_list
) ||
648 (sctp_list_single_entry(&bp
->address_list
))) {
654 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
660 if (!af
->addr_valid(sa_addr
, sp
, NULL
)) {
661 retval
= -EADDRNOTAVAIL
;
665 if (sa_addr
->v4
.sin_port
&&
666 sa_addr
->v4
.sin_port
!= htons(bp
->port
)) {
671 if (!sa_addr
->v4
.sin_port
)
672 sa_addr
->v4
.sin_port
= htons(bp
->port
);
674 /* FIXME - There is probably a need to check if sk->sk_saddr and
675 * sk->sk_rcv_addr are currently set to one of the addresses to
676 * be removed. This is something which needs to be looked into
677 * when we are fixing the outstanding issues with multi-homing
678 * socket routing and failover schemes. Refer to comments in
679 * sctp_do_bind(). -daisy
681 retval
= sctp_del_bind_addr(bp
, sa_addr
);
683 addr_buf
+= af
->sockaddr_len
;
686 /* Failed. Add the ones that has been removed back */
688 sctp_bindx_add(sk
, addrs
, cnt
);
696 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
697 * the associations that are part of the endpoint indicating that a list of
698 * local addresses are removed from the endpoint.
700 * If any of the addresses is already in the bind address list of the
701 * association, we do not send the chunk for that association. But it will not
702 * affect other associations.
704 * Only sctp_setsockopt_bindx() is supposed to call this function.
706 static int sctp_send_asconf_del_ip(struct sock
*sk
,
707 struct sockaddr
*addrs
,
710 struct net
*net
= sock_net(sk
);
711 struct sctp_sock
*sp
;
712 struct sctp_endpoint
*ep
;
713 struct sctp_association
*asoc
;
714 struct sctp_transport
*transport
;
715 struct sctp_bind_addr
*bp
;
716 struct sctp_chunk
*chunk
;
717 union sctp_addr
*laddr
;
720 struct sctp_sockaddr_entry
*saddr
;
726 if (!net
->sctp
.addip_enable
)
732 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
733 __func__
, sk
, addrs
, addrcnt
);
735 list_for_each_entry(asoc
, &ep
->asocs
, asocs
) {
737 if (!asoc
->peer
.asconf_capable
)
740 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_DEL_IP
)
743 if (!sctp_state(asoc
, ESTABLISHED
))
746 /* Check if any address in the packed array of addresses is
747 * not present in the bind address list of the association.
748 * If so, do not send the asconf chunk to its peer, but
749 * continue with other associations.
752 for (i
= 0; i
< addrcnt
; i
++) {
754 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
760 if (!sctp_assoc_lookup_laddr(asoc
, laddr
))
763 addr_buf
+= af
->sockaddr_len
;
768 /* Find one address in the association's bind address list
769 * that is not in the packed array of addresses. This is to
770 * make sure that we do not delete all the addresses in the
773 bp
= &asoc
->base
.bind_addr
;
774 laddr
= sctp_find_unmatch_addr(bp
, (union sctp_addr
*)addrs
,
776 if ((laddr
== NULL
) && (addrcnt
== 1)) {
777 if (asoc
->asconf_addr_del_pending
)
779 asoc
->asconf_addr_del_pending
=
780 kzalloc(sizeof(union sctp_addr
), GFP_ATOMIC
);
781 if (asoc
->asconf_addr_del_pending
== NULL
) {
785 asoc
->asconf_addr_del_pending
->sa
.sa_family
=
787 asoc
->asconf_addr_del_pending
->v4
.sin_port
=
789 if (addrs
->sa_family
== AF_INET
) {
790 struct sockaddr_in
*sin
;
792 sin
= (struct sockaddr_in
*)addrs
;
793 asoc
->asconf_addr_del_pending
->v4
.sin_addr
.s_addr
= sin
->sin_addr
.s_addr
;
794 } else if (addrs
->sa_family
== AF_INET6
) {
795 struct sockaddr_in6
*sin6
;
797 sin6
= (struct sockaddr_in6
*)addrs
;
798 asoc
->asconf_addr_del_pending
->v6
.sin6_addr
= sin6
->sin6_addr
;
801 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
802 __func__
, asoc
, &asoc
->asconf_addr_del_pending
->sa
,
803 asoc
->asconf_addr_del_pending
);
805 asoc
->src_out_of_asoc_ok
= 1;
813 /* We do not need RCU protection throughout this loop
814 * because this is done under a socket lock from the
817 chunk
= sctp_make_asconf_update_ip(asoc
, laddr
, addrs
, addrcnt
,
825 /* Reset use_as_src flag for the addresses in the bind address
826 * list that are to be deleted.
829 for (i
= 0; i
< addrcnt
; i
++) {
831 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
832 list_for_each_entry(saddr
, &bp
->address_list
, list
) {
833 if (sctp_cmp_addr_exact(&saddr
->a
, laddr
))
834 saddr
->state
= SCTP_ADDR_DEL
;
836 addr_buf
+= af
->sockaddr_len
;
839 /* Update the route and saddr entries for all the transports
840 * as some of the addresses in the bind address list are
841 * about to be deleted and cannot be used as source addresses.
843 list_for_each_entry(transport
, &asoc
->peer
.transport_addr_list
,
845 dst_release(transport
->dst
);
846 sctp_transport_route(transport
, NULL
,
847 sctp_sk(asoc
->base
.sk
));
851 /* We don't need to transmit ASCONF */
853 retval
= sctp_send_asconf(asoc
, chunk
);
859 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
860 int sctp_asconf_mgmt(struct sctp_sock
*sp
, struct sctp_sockaddr_entry
*addrw
)
862 struct sock
*sk
= sctp_opt2sk(sp
);
863 union sctp_addr
*addr
;
866 /* It is safe to write port space in caller. */
868 addr
->v4
.sin_port
= htons(sp
->ep
->base
.bind_addr
.port
);
869 af
= sctp_get_af_specific(addr
->sa
.sa_family
);
872 if (sctp_verify_addr(sk
, addr
, af
->sockaddr_len
))
875 if (addrw
->state
== SCTP_ADDR_NEW
)
876 return sctp_send_asconf_add_ip(sk
, (struct sockaddr
*)addr
, 1);
878 return sctp_send_asconf_del_ip(sk
, (struct sockaddr
*)addr
, 1);
881 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
884 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
887 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
888 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
891 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
892 * Section 3.1.2 for this usage.
894 * addrs is a pointer to an array of one or more socket addresses. Each
895 * address is contained in its appropriate structure (i.e. struct
896 * sockaddr_in or struct sockaddr_in6) the family of the address type
897 * must be used to distinguish the address length (note that this
898 * representation is termed a "packed array" of addresses). The caller
899 * specifies the number of addresses in the array with addrcnt.
901 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
902 * -1, and sets errno to the appropriate error code.
904 * For SCTP, the port given in each socket address must be the same, or
905 * sctp_bindx() will fail, setting errno to EINVAL.
907 * The flags parameter is formed from the bitwise OR of zero or more of
908 * the following currently defined flags:
910 * SCTP_BINDX_ADD_ADDR
912 * SCTP_BINDX_REM_ADDR
914 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
915 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
916 * addresses from the association. The two flags are mutually exclusive;
917 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
918 * not remove all addresses from an association; sctp_bindx() will
919 * reject such an attempt with EINVAL.
921 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
922 * additional addresses with an endpoint after calling bind(). Or use
923 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
924 * socket is associated with so that no new association accepted will be
925 * associated with those addresses. If the endpoint supports dynamic
926 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
927 * endpoint to send the appropriate message to the peer to change the
928 * peers address lists.
930 * Adding and removing addresses from a connected association is
931 * optional functionality. Implementations that do not support this
932 * functionality should return EOPNOTSUPP.
934 * Basically do nothing but copying the addresses from user to kernel
935 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
936 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
939 * We don't use copy_from_user() for optimization: we first do the
940 * sanity checks (buffer size -fast- and access check-healthy
941 * pointer); if all of those succeed, then we can alloc the memory
942 * (expensive operation) needed to copy the data to kernel. Then we do
943 * the copying without checking the user space area
944 * (__copy_from_user()).
946 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
949 * sk The sk of the socket
950 * addrs The pointer to the addresses in user land
951 * addrssize Size of the addrs buffer
952 * op Operation to perform (add or remove, see the flags of
955 * Returns 0 if ok, <0 errno code on error.
957 static int sctp_setsockopt_bindx(struct sock
*sk
,
958 struct sockaddr __user
*addrs
,
959 int addrs_size
, int op
)
961 struct sockaddr
*kaddrs
;
965 struct sockaddr
*sa_addr
;
969 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
970 __func__
, sk
, addrs
, addrs_size
, op
);
972 if (unlikely(addrs_size
<= 0))
975 /* Check the user passed a healthy pointer. */
976 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
979 /* Alloc space for the address array in kernel memory. */
980 kaddrs
= kmalloc(addrs_size
, GFP_KERNEL
);
981 if (unlikely(!kaddrs
))
984 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
989 /* Walk through the addrs buffer and count the number of addresses. */
991 while (walk_size
< addrs_size
) {
992 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
998 af
= sctp_get_af_specific(sa_addr
->sa_family
);
1000 /* If the address family is not supported or if this address
1001 * causes the address buffer to overflow return EINVAL.
1003 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
1008 addr_buf
+= af
->sockaddr_len
;
1009 walk_size
+= af
->sockaddr_len
;
1014 case SCTP_BINDX_ADD_ADDR
:
1015 err
= sctp_bindx_add(sk
, kaddrs
, addrcnt
);
1018 err
= sctp_send_asconf_add_ip(sk
, kaddrs
, addrcnt
);
1021 case SCTP_BINDX_REM_ADDR
:
1022 err
= sctp_bindx_rem(sk
, kaddrs
, addrcnt
);
1025 err
= sctp_send_asconf_del_ip(sk
, kaddrs
, addrcnt
);
1039 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1041 * Common routine for handling connect() and sctp_connectx().
1042 * Connect will come in with just a single address.
1044 static int __sctp_connect(struct sock
*sk
,
1045 struct sockaddr
*kaddrs
,
1047 sctp_assoc_t
*assoc_id
)
1049 struct net
*net
= sock_net(sk
);
1050 struct sctp_sock
*sp
;
1051 struct sctp_endpoint
*ep
;
1052 struct sctp_association
*asoc
= NULL
;
1053 struct sctp_association
*asoc2
;
1054 struct sctp_transport
*transport
;
1061 union sctp_addr
*sa_addr
= NULL
;
1063 unsigned short port
;
1064 unsigned int f_flags
= 0;
1069 /* connect() cannot be done on a socket that is already in ESTABLISHED
1070 * state - UDP-style peeled off socket or a TCP-style socket that
1071 * is already connected.
1072 * It cannot be done even on a TCP-style listening socket.
1074 if (sctp_sstate(sk
, ESTABLISHED
) ||
1075 (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))) {
1080 /* Walk through the addrs buffer and count the number of addresses. */
1082 while (walk_size
< addrs_size
) {
1085 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
1091 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
1093 /* If the address family is not supported or if this address
1094 * causes the address buffer to overflow return EINVAL.
1096 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
1101 port
= ntohs(sa_addr
->v4
.sin_port
);
1103 /* Save current address so we can work with it */
1104 memcpy(&to
, sa_addr
, af
->sockaddr_len
);
1106 err
= sctp_verify_addr(sk
, &to
, af
->sockaddr_len
);
1110 /* Make sure the destination port is correctly set
1113 if (asoc
&& asoc
->peer
.port
&& asoc
->peer
.port
!= port
) {
1118 /* Check if there already is a matching association on the
1119 * endpoint (other than the one created here).
1121 asoc2
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1122 if (asoc2
&& asoc2
!= asoc
) {
1123 if (asoc2
->state
>= SCTP_STATE_ESTABLISHED
)
1130 /* If we could not find a matching association on the endpoint,
1131 * make sure that there is no peeled-off association matching
1132 * the peer address even on another socket.
1134 if (sctp_endpoint_is_peeled_off(ep
, &to
)) {
1135 err
= -EADDRNOTAVAIL
;
1140 /* If a bind() or sctp_bindx() is not called prior to
1141 * an sctp_connectx() call, the system picks an
1142 * ephemeral port and will choose an address set
1143 * equivalent to binding with a wildcard address.
1145 if (!ep
->base
.bind_addr
.port
) {
1146 if (sctp_autobind(sk
)) {
1152 * If an unprivileged user inherits a 1-many
1153 * style socket with open associations on a
1154 * privileged port, it MAY be permitted to
1155 * accept new associations, but it SHOULD NOT
1156 * be permitted to open new associations.
1158 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1159 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
)) {
1165 scope
= sctp_scope(&to
);
1166 asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1172 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
,
1180 /* Prime the peer's transport structures. */
1181 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
,
1189 addr_buf
+= af
->sockaddr_len
;
1190 walk_size
+= af
->sockaddr_len
;
1193 /* In case the user of sctp_connectx() wants an association
1194 * id back, assign one now.
1197 err
= sctp_assoc_set_id(asoc
, GFP_KERNEL
);
1202 err
= sctp_primitive_ASSOCIATE(net
, asoc
, NULL
);
1207 /* Initialize sk's dport and daddr for getpeername() */
1208 inet_sk(sk
)->inet_dport
= htons(asoc
->peer
.port
);
1209 sp
->pf
->to_sk_daddr(sa_addr
, sk
);
1212 /* in-kernel sockets don't generally have a file allocated to them
1213 * if all they do is call sock_create_kern().
1215 if (sk
->sk_socket
->file
)
1216 f_flags
= sk
->sk_socket
->file
->f_flags
;
1218 timeo
= sock_sndtimeo(sk
, f_flags
& O_NONBLOCK
);
1220 err
= sctp_wait_for_connect(asoc
, &timeo
);
1221 if ((err
== 0 || err
== -EINPROGRESS
) && assoc_id
)
1222 *assoc_id
= asoc
->assoc_id
;
1224 /* Don't free association on exit. */
1228 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1229 __func__
, asoc
, kaddrs
, err
);
1232 /* sctp_primitive_ASSOCIATE may have added this association
1233 * To the hash table, try to unhash it, just in case, its a noop
1234 * if it wasn't hashed so we're safe
1236 sctp_unhash_established(asoc
);
1237 sctp_association_free(asoc
);
1242 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1245 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1246 * sctp_assoc_t *asoc);
1248 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1249 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1250 * or IPv6 addresses.
1252 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1253 * Section 3.1.2 for this usage.
1255 * addrs is a pointer to an array of one or more socket addresses. Each
1256 * address is contained in its appropriate structure (i.e. struct
1257 * sockaddr_in or struct sockaddr_in6) the family of the address type
1258 * must be used to distengish the address length (note that this
1259 * representation is termed a "packed array" of addresses). The caller
1260 * specifies the number of addresses in the array with addrcnt.
1262 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1263 * the association id of the new association. On failure, sctp_connectx()
1264 * returns -1, and sets errno to the appropriate error code. The assoc_id
1265 * is not touched by the kernel.
1267 * For SCTP, the port given in each socket address must be the same, or
1268 * sctp_connectx() will fail, setting errno to EINVAL.
1270 * An application can use sctp_connectx to initiate an association with
1271 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1272 * allows a caller to specify multiple addresses at which a peer can be
1273 * reached. The way the SCTP stack uses the list of addresses to set up
1274 * the association is implementation dependent. This function only
1275 * specifies that the stack will try to make use of all the addresses in
1276 * the list when needed.
1278 * Note that the list of addresses passed in is only used for setting up
1279 * the association. It does not necessarily equal the set of addresses
1280 * the peer uses for the resulting association. If the caller wants to
1281 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1282 * retrieve them after the association has been set up.
1284 * Basically do nothing but copying the addresses from user to kernel
1285 * land and invoking either sctp_connectx(). This is used for tunneling
1286 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1288 * We don't use copy_from_user() for optimization: we first do the
1289 * sanity checks (buffer size -fast- and access check-healthy
1290 * pointer); if all of those succeed, then we can alloc the memory
1291 * (expensive operation) needed to copy the data to kernel. Then we do
1292 * the copying without checking the user space area
1293 * (__copy_from_user()).
1295 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1298 * sk The sk of the socket
1299 * addrs The pointer to the addresses in user land
1300 * addrssize Size of the addrs buffer
1302 * Returns >=0 if ok, <0 errno code on error.
1304 static int __sctp_setsockopt_connectx(struct sock
*sk
,
1305 struct sockaddr __user
*addrs
,
1307 sctp_assoc_t
*assoc_id
)
1310 struct sockaddr
*kaddrs
;
1312 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1313 __func__
, sk
, addrs
, addrs_size
);
1315 if (unlikely(addrs_size
<= 0))
1318 /* Check the user passed a healthy pointer. */
1319 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
1322 /* Alloc space for the address array in kernel memory. */
1323 kaddrs
= kmalloc(addrs_size
, GFP_KERNEL
);
1324 if (unlikely(!kaddrs
))
1327 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
1330 err
= __sctp_connect(sk
, kaddrs
, addrs_size
, assoc_id
);
1339 * This is an older interface. It's kept for backward compatibility
1340 * to the option that doesn't provide association id.
1342 static int sctp_setsockopt_connectx_old(struct sock
*sk
,
1343 struct sockaddr __user
*addrs
,
1346 return __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, NULL
);
1350 * New interface for the API. The since the API is done with a socket
1351 * option, to make it simple we feed back the association id is as a return
1352 * indication to the call. Error is always negative and association id is
1355 static int sctp_setsockopt_connectx(struct sock
*sk
,
1356 struct sockaddr __user
*addrs
,
1359 sctp_assoc_t assoc_id
= 0;
1362 err
= __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, &assoc_id
);
1371 * New (hopefully final) interface for the API.
1372 * We use the sctp_getaddrs_old structure so that use-space library
1373 * can avoid any unnecessary allocations. The only different part
1374 * is that we store the actual length of the address buffer into the
1375 * addrs_num structure member. That way we can re-use the existing
1378 #ifdef CONFIG_COMPAT
1379 struct compat_sctp_getaddrs_old
{
1380 sctp_assoc_t assoc_id
;
1382 compat_uptr_t addrs
; /* struct sockaddr * */
1386 static int sctp_getsockopt_connectx3(struct sock
*sk
, int len
,
1387 char __user
*optval
,
1390 struct sctp_getaddrs_old param
;
1391 sctp_assoc_t assoc_id
= 0;
1394 #ifdef CONFIG_COMPAT
1395 if (is_compat_task()) {
1396 struct compat_sctp_getaddrs_old param32
;
1398 if (len
< sizeof(param32
))
1400 if (copy_from_user(¶m32
, optval
, sizeof(param32
)))
1403 param
.assoc_id
= param32
.assoc_id
;
1404 param
.addr_num
= param32
.addr_num
;
1405 param
.addrs
= compat_ptr(param32
.addrs
);
1409 if (len
< sizeof(param
))
1411 if (copy_from_user(¶m
, optval
, sizeof(param
)))
1415 err
= __sctp_setsockopt_connectx(sk
, (struct sockaddr __user
*)
1416 param
.addrs
, param
.addr_num
,
1418 if (err
== 0 || err
== -EINPROGRESS
) {
1419 if (copy_to_user(optval
, &assoc_id
, sizeof(assoc_id
)))
1421 if (put_user(sizeof(assoc_id
), optlen
))
1428 /* API 3.1.4 close() - UDP Style Syntax
1429 * Applications use close() to perform graceful shutdown (as described in
1430 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1431 * by a UDP-style socket.
1435 * ret = close(int sd);
1437 * sd - the socket descriptor of the associations to be closed.
1439 * To gracefully shutdown a specific association represented by the
1440 * UDP-style socket, an application should use the sendmsg() call,
1441 * passing no user data, but including the appropriate flag in the
1442 * ancillary data (see Section xxxx).
1444 * If sd in the close() call is a branched-off socket representing only
1445 * one association, the shutdown is performed on that association only.
1447 * 4.1.6 close() - TCP Style Syntax
1449 * Applications use close() to gracefully close down an association.
1453 * int close(int sd);
1455 * sd - the socket descriptor of the association to be closed.
1457 * After an application calls close() on a socket descriptor, no further
1458 * socket operations will succeed on that descriptor.
1460 * API 7.1.4 SO_LINGER
1462 * An application using the TCP-style socket can use this option to
1463 * perform the SCTP ABORT primitive. The linger option structure is:
1466 * int l_onoff; // option on/off
1467 * int l_linger; // linger time
1470 * To enable the option, set l_onoff to 1. If the l_linger value is set
1471 * to 0, calling close() is the same as the ABORT primitive. If the
1472 * value is set to a negative value, the setsockopt() call will return
1473 * an error. If the value is set to a positive value linger_time, the
1474 * close() can be blocked for at most linger_time ms. If the graceful
1475 * shutdown phase does not finish during this period, close() will
1476 * return but the graceful shutdown phase continues in the system.
1478 static void sctp_close(struct sock
*sk
, long timeout
)
1480 struct net
*net
= sock_net(sk
);
1481 struct sctp_endpoint
*ep
;
1482 struct sctp_association
*asoc
;
1483 struct list_head
*pos
, *temp
;
1484 unsigned int data_was_unread
;
1486 pr_debug("%s: sk:%p, timeout:%ld\n", __func__
, sk
, timeout
);
1489 sk
->sk_shutdown
= SHUTDOWN_MASK
;
1490 sk
->sk_state
= SCTP_SS_CLOSING
;
1492 ep
= sctp_sk(sk
)->ep
;
1494 /* Clean up any skbs sitting on the receive queue. */
1495 data_was_unread
= sctp_queue_purge_ulpevents(&sk
->sk_receive_queue
);
1496 data_was_unread
+= sctp_queue_purge_ulpevents(&sctp_sk(sk
)->pd_lobby
);
1498 /* Walk all associations on an endpoint. */
1499 list_for_each_safe(pos
, temp
, &ep
->asocs
) {
1500 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
1502 if (sctp_style(sk
, TCP
)) {
1503 /* A closed association can still be in the list if
1504 * it belongs to a TCP-style listening socket that is
1505 * not yet accepted. If so, free it. If not, send an
1506 * ABORT or SHUTDOWN based on the linger options.
1508 if (sctp_state(asoc
, CLOSED
)) {
1509 sctp_unhash_established(asoc
);
1510 sctp_association_free(asoc
);
1515 if (data_was_unread
|| !skb_queue_empty(&asoc
->ulpq
.lobby
) ||
1516 !skb_queue_empty(&asoc
->ulpq
.reasm
) ||
1517 (sock_flag(sk
, SOCK_LINGER
) && !sk
->sk_lingertime
)) {
1518 struct sctp_chunk
*chunk
;
1520 chunk
= sctp_make_abort_user(asoc
, NULL
, 0);
1522 sctp_primitive_ABORT(net
, asoc
, chunk
);
1524 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
1527 /* On a TCP-style socket, block for at most linger_time if set. */
1528 if (sctp_style(sk
, TCP
) && timeout
)
1529 sctp_wait_for_close(sk
, timeout
);
1531 /* This will run the backlog queue. */
1534 /* Supposedly, no process has access to the socket, but
1535 * the net layers still may.
1540 /* Hold the sock, since sk_common_release() will put sock_put()
1541 * and we have just a little more cleanup.
1544 sk_common_release(sk
);
1551 SCTP_DBG_OBJCNT_DEC(sock
);
1554 /* Handle EPIPE error. */
1555 static int sctp_error(struct sock
*sk
, int flags
, int err
)
1558 err
= sock_error(sk
) ? : -EPIPE
;
1559 if (err
== -EPIPE
&& !(flags
& MSG_NOSIGNAL
))
1560 send_sig(SIGPIPE
, current
, 0);
1564 /* API 3.1.3 sendmsg() - UDP Style Syntax
1566 * An application uses sendmsg() and recvmsg() calls to transmit data to
1567 * and receive data from its peer.
1569 * ssize_t sendmsg(int socket, const struct msghdr *message,
1572 * socket - the socket descriptor of the endpoint.
1573 * message - pointer to the msghdr structure which contains a single
1574 * user message and possibly some ancillary data.
1576 * See Section 5 for complete description of the data
1579 * flags - flags sent or received with the user message, see Section
1580 * 5 for complete description of the flags.
1582 * Note: This function could use a rewrite especially when explicit
1583 * connect support comes in.
1585 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1587 static int sctp_msghdr_parse(const struct msghdr
*, sctp_cmsgs_t
*);
1589 static int sctp_sendmsg(struct sock
*sk
, struct msghdr
*msg
, size_t msg_len
)
1591 struct net
*net
= sock_net(sk
);
1592 struct sctp_sock
*sp
;
1593 struct sctp_endpoint
*ep
;
1594 struct sctp_association
*new_asoc
= NULL
, *asoc
= NULL
;
1595 struct sctp_transport
*transport
, *chunk_tp
;
1596 struct sctp_chunk
*chunk
;
1598 struct sockaddr
*msg_name
= NULL
;
1599 struct sctp_sndrcvinfo default_sinfo
;
1600 struct sctp_sndrcvinfo
*sinfo
;
1601 struct sctp_initmsg
*sinit
;
1602 sctp_assoc_t associd
= 0;
1603 sctp_cmsgs_t cmsgs
= { NULL
};
1605 bool fill_sinfo_ttl
= false, wait_connect
= false;
1606 struct sctp_datamsg
*datamsg
;
1607 int msg_flags
= msg
->msg_flags
;
1608 __u16 sinfo_flags
= 0;
1616 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__
, sk
,
1619 /* We cannot send a message over a TCP-style listening socket. */
1620 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
)) {
1625 /* Parse out the SCTP CMSGs. */
1626 err
= sctp_msghdr_parse(msg
, &cmsgs
);
1628 pr_debug("%s: msghdr parse err:%x\n", __func__
, err
);
1632 /* Fetch the destination address for this packet. This
1633 * address only selects the association--it is not necessarily
1634 * the address we will send to.
1635 * For a peeled-off socket, msg_name is ignored.
1637 if (!sctp_style(sk
, UDP_HIGH_BANDWIDTH
) && msg
->msg_name
) {
1638 int msg_namelen
= msg
->msg_namelen
;
1640 err
= sctp_verify_addr(sk
, (union sctp_addr
*)msg
->msg_name
,
1645 if (msg_namelen
> sizeof(to
))
1646 msg_namelen
= sizeof(to
);
1647 memcpy(&to
, msg
->msg_name
, msg_namelen
);
1648 msg_name
= msg
->msg_name
;
1652 if (cmsgs
.sinfo
!= NULL
) {
1653 memset(&default_sinfo
, 0, sizeof(default_sinfo
));
1654 default_sinfo
.sinfo_stream
= cmsgs
.sinfo
->snd_sid
;
1655 default_sinfo
.sinfo_flags
= cmsgs
.sinfo
->snd_flags
;
1656 default_sinfo
.sinfo_ppid
= cmsgs
.sinfo
->snd_ppid
;
1657 default_sinfo
.sinfo_context
= cmsgs
.sinfo
->snd_context
;
1658 default_sinfo
.sinfo_assoc_id
= cmsgs
.sinfo
->snd_assoc_id
;
1660 sinfo
= &default_sinfo
;
1661 fill_sinfo_ttl
= true;
1663 sinfo
= cmsgs
.srinfo
;
1665 /* Did the user specify SNDINFO/SNDRCVINFO? */
1667 sinfo_flags
= sinfo
->sinfo_flags
;
1668 associd
= sinfo
->sinfo_assoc_id
;
1671 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__
,
1672 msg_len
, sinfo_flags
);
1674 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1675 if (sctp_style(sk
, TCP
) && (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
))) {
1680 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1681 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1682 * If SCTP_ABORT is set, the message length could be non zero with
1683 * the msg_iov set to the user abort reason.
1685 if (((sinfo_flags
& SCTP_EOF
) && (msg_len
> 0)) ||
1686 (!(sinfo_flags
& (SCTP_EOF
|SCTP_ABORT
)) && (msg_len
== 0))) {
1691 /* If SCTP_ADDR_OVER is set, there must be an address
1692 * specified in msg_name.
1694 if ((sinfo_flags
& SCTP_ADDR_OVER
) && (!msg
->msg_name
)) {
1701 pr_debug("%s: about to look up association\n", __func__
);
1705 /* If a msg_name has been specified, assume this is to be used. */
1707 /* Look for a matching association on the endpoint. */
1708 asoc
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1710 /* If we could not find a matching association on the
1711 * endpoint, make sure that it is not a TCP-style
1712 * socket that already has an association or there is
1713 * no peeled-off association on another socket.
1715 if ((sctp_style(sk
, TCP
) &&
1716 sctp_sstate(sk
, ESTABLISHED
)) ||
1717 sctp_endpoint_is_peeled_off(ep
, &to
)) {
1718 err
= -EADDRNOTAVAIL
;
1723 asoc
= sctp_id2assoc(sk
, associd
);
1731 pr_debug("%s: just looked up association:%p\n", __func__
, asoc
);
1733 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1734 * socket that has an association in CLOSED state. This can
1735 * happen when an accepted socket has an association that is
1738 if (sctp_state(asoc
, CLOSED
) && sctp_style(sk
, TCP
)) {
1743 if (sinfo_flags
& SCTP_EOF
) {
1744 pr_debug("%s: shutting down association:%p\n",
1747 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
1751 if (sinfo_flags
& SCTP_ABORT
) {
1753 chunk
= sctp_make_abort_user(asoc
, msg
, msg_len
);
1759 pr_debug("%s: aborting association:%p\n",
1762 sctp_primitive_ABORT(net
, asoc
, chunk
);
1768 /* Do we need to create the association? */
1770 pr_debug("%s: there is no association yet\n", __func__
);
1772 if (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
)) {
1777 /* Check for invalid stream against the stream counts,
1778 * either the default or the user specified stream counts.
1781 if (!sinit
|| !sinit
->sinit_num_ostreams
) {
1782 /* Check against the defaults. */
1783 if (sinfo
->sinfo_stream
>=
1784 sp
->initmsg
.sinit_num_ostreams
) {
1789 /* Check against the requested. */
1790 if (sinfo
->sinfo_stream
>=
1791 sinit
->sinit_num_ostreams
) {
1799 * API 3.1.2 bind() - UDP Style Syntax
1800 * If a bind() or sctp_bindx() is not called prior to a
1801 * sendmsg() call that initiates a new association, the
1802 * system picks an ephemeral port and will choose an address
1803 * set equivalent to binding with a wildcard address.
1805 if (!ep
->base
.bind_addr
.port
) {
1806 if (sctp_autobind(sk
)) {
1812 * If an unprivileged user inherits a one-to-many
1813 * style socket with open associations on a privileged
1814 * port, it MAY be permitted to accept new associations,
1815 * but it SHOULD NOT be permitted to open new
1818 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1819 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
)) {
1825 scope
= sctp_scope(&to
);
1826 new_asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1832 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
, GFP_KERNEL
);
1838 /* If the SCTP_INIT ancillary data is specified, set all
1839 * the association init values accordingly.
1842 if (sinit
->sinit_num_ostreams
) {
1843 asoc
->c
.sinit_num_ostreams
=
1844 sinit
->sinit_num_ostreams
;
1846 if (sinit
->sinit_max_instreams
) {
1847 asoc
->c
.sinit_max_instreams
=
1848 sinit
->sinit_max_instreams
;
1850 if (sinit
->sinit_max_attempts
) {
1851 asoc
->max_init_attempts
1852 = sinit
->sinit_max_attempts
;
1854 if (sinit
->sinit_max_init_timeo
) {
1855 asoc
->max_init_timeo
=
1856 msecs_to_jiffies(sinit
->sinit_max_init_timeo
);
1860 /* Prime the peer's transport structures. */
1861 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
, SCTP_UNKNOWN
);
1868 /* ASSERT: we have a valid association at this point. */
1869 pr_debug("%s: we have a valid association\n", __func__
);
1872 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1873 * one with some defaults.
1875 memset(&default_sinfo
, 0, sizeof(default_sinfo
));
1876 default_sinfo
.sinfo_stream
= asoc
->default_stream
;
1877 default_sinfo
.sinfo_flags
= asoc
->default_flags
;
1878 default_sinfo
.sinfo_ppid
= asoc
->default_ppid
;
1879 default_sinfo
.sinfo_context
= asoc
->default_context
;
1880 default_sinfo
.sinfo_timetolive
= asoc
->default_timetolive
;
1881 default_sinfo
.sinfo_assoc_id
= sctp_assoc2id(asoc
);
1883 sinfo
= &default_sinfo
;
1884 } else if (fill_sinfo_ttl
) {
1885 /* In case SNDINFO was specified, we still need to fill
1886 * it with a default ttl from the assoc here.
1888 sinfo
->sinfo_timetolive
= asoc
->default_timetolive
;
1891 /* API 7.1.7, the sndbuf size per association bounds the
1892 * maximum size of data that can be sent in a single send call.
1894 if (msg_len
> sk
->sk_sndbuf
) {
1899 if (asoc
->pmtu_pending
)
1900 sctp_assoc_pending_pmtu(sk
, asoc
);
1902 /* If fragmentation is disabled and the message length exceeds the
1903 * association fragmentation point, return EMSGSIZE. The I-D
1904 * does not specify what this error is, but this looks like
1907 if (sctp_sk(sk
)->disable_fragments
&& (msg_len
> asoc
->frag_point
)) {
1912 /* Check for invalid stream. */
1913 if (sinfo
->sinfo_stream
>= asoc
->c
.sinit_num_ostreams
) {
1918 timeo
= sock_sndtimeo(sk
, msg
->msg_flags
& MSG_DONTWAIT
);
1919 if (!sctp_wspace(asoc
)) {
1920 err
= sctp_wait_for_sndbuf(asoc
, &timeo
, msg_len
);
1925 /* If an address is passed with the sendto/sendmsg call, it is used
1926 * to override the primary destination address in the TCP model, or
1927 * when SCTP_ADDR_OVER flag is set in the UDP model.
1929 if ((sctp_style(sk
, TCP
) && msg_name
) ||
1930 (sinfo_flags
& SCTP_ADDR_OVER
)) {
1931 chunk_tp
= sctp_assoc_lookup_paddr(asoc
, &to
);
1939 /* Auto-connect, if we aren't connected already. */
1940 if (sctp_state(asoc
, CLOSED
)) {
1941 err
= sctp_primitive_ASSOCIATE(net
, asoc
, NULL
);
1945 wait_connect
= true;
1946 pr_debug("%s: we associated primitively\n", __func__
);
1949 /* Break the message into multiple chunks of maximum size. */
1950 datamsg
= sctp_datamsg_from_user(asoc
, sinfo
, &msg
->msg_iter
);
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
);
1978 pr_debug("%s: we sent primitively\n", __func__
);
1980 sctp_datamsg_put(datamsg
);
1983 if (unlikely(wait_connect
)) {
1984 timeo
= sock_sndtimeo(sk
, msg_flags
& MSG_DONTWAIT
);
1985 sctp_wait_for_connect(asoc
, &timeo
);
1988 /* If we are already past ASSOCIATE, the lower
1989 * layers are responsible for association cleanup.
1995 sctp_unhash_established(asoc
);
1996 sctp_association_free(asoc
);
2002 return sctp_error(sk
, msg_flags
, err
);
2009 err
= sock_error(sk
);
2019 /* This is an extended version of skb_pull() that removes the data from the
2020 * start of a skb even when data is spread across the list of skb's in the
2021 * frag_list. len specifies the total amount of data that needs to be removed.
2022 * when 'len' bytes could be removed from the skb, it returns 0.
2023 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2024 * could not be removed.
2026 static int sctp_skb_pull(struct sk_buff
*skb
, int len
)
2028 struct sk_buff
*list
;
2029 int skb_len
= skb_headlen(skb
);
2032 if (len
<= skb_len
) {
2033 __skb_pull(skb
, len
);
2037 __skb_pull(skb
, skb_len
);
2039 skb_walk_frags(skb
, list
) {
2040 rlen
= sctp_skb_pull(list
, len
);
2041 skb
->len
-= (len
-rlen
);
2042 skb
->data_len
-= (len
-rlen
);
2053 /* API 3.1.3 recvmsg() - UDP Style Syntax
2055 * ssize_t recvmsg(int socket, struct msghdr *message,
2058 * socket - the socket descriptor of the endpoint.
2059 * message - pointer to the msghdr structure which contains a single
2060 * user message and possibly some ancillary data.
2062 * See Section 5 for complete description of the data
2065 * flags - flags sent or received with the user message, see Section
2066 * 5 for complete description of the flags.
2068 static int sctp_recvmsg(struct sock
*sk
, struct msghdr
*msg
, size_t len
,
2069 int noblock
, 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 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2079 "addr_len:%p)\n", __func__
, sk
, msg
, len
, noblock
, flags
,
2084 if (sctp_style(sk
, TCP
) && !sctp_sstate(sk
, ESTABLISHED
)) {
2089 skb
= sctp_skb_recv_datagram(sk
, flags
, noblock
, &err
);
2093 /* Get the total length of the skb including any skb's in the
2102 err
= skb_copy_datagram_msg(skb
, 0, msg
, copied
);
2104 event
= sctp_skb2event(skb
);
2109 sock_recv_ts_and_drops(msg
, sk
, skb
);
2110 if (sctp_ulpevent_is_notification(event
)) {
2111 msg
->msg_flags
|= MSG_NOTIFICATION
;
2112 sp
->pf
->event_msgname(event
, msg
->msg_name
, addr_len
);
2114 sp
->pf
->skb_msgname(skb
, msg
->msg_name
, addr_len
);
2117 /* Check if we allow SCTP_NXTINFO. */
2118 if (sp
->recvnxtinfo
)
2119 sctp_ulpevent_read_nxtinfo(event
, msg
, sk
);
2120 /* Check if we allow SCTP_RCVINFO. */
2121 if (sp
->recvrcvinfo
)
2122 sctp_ulpevent_read_rcvinfo(event
, msg
);
2123 /* Check if we allow SCTP_SNDRCVINFO. */
2124 if (sp
->subscribe
.sctp_data_io_event
)
2125 sctp_ulpevent_read_sndrcvinfo(event
, msg
);
2128 /* FIXME: we should be calling IP/IPv6 layers. */
2129 if (sk
->sk_protinfo
.af_inet
.cmsg_flags
)
2130 ip_cmsg_recv(msg
, skb
);
2135 /* If skb's length exceeds the user's buffer, update the skb and
2136 * push it back to the receive_queue so that the next call to
2137 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2139 if (skb_len
> copied
) {
2140 msg
->msg_flags
&= ~MSG_EOR
;
2141 if (flags
& MSG_PEEK
)
2143 sctp_skb_pull(skb
, copied
);
2144 skb_queue_head(&sk
->sk_receive_queue
, skb
);
2146 /* When only partial message is copied to the user, increase
2147 * rwnd by that amount. If all the data in the skb is read,
2148 * rwnd is updated when the event is freed.
2150 if (!sctp_ulpevent_is_notification(event
))
2151 sctp_assoc_rwnd_increase(event
->asoc
, copied
);
2153 } else if ((event
->msg_flags
& MSG_NOTIFICATION
) ||
2154 (event
->msg_flags
& MSG_EOR
))
2155 msg
->msg_flags
|= MSG_EOR
;
2157 msg
->msg_flags
&= ~MSG_EOR
;
2160 if (flags
& MSG_PEEK
) {
2161 /* Release the skb reference acquired after peeking the skb in
2162 * sctp_skb_recv_datagram().
2166 /* Free the event which includes releasing the reference to
2167 * the owner of the skb, freeing the skb and updating the
2170 sctp_ulpevent_free(event
);
2177 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2179 * This option is a on/off flag. If enabled no SCTP message
2180 * fragmentation will be performed. Instead if a message being sent
2181 * exceeds the current PMTU size, the message will NOT be sent and
2182 * instead a error will be indicated to the user.
2184 static int sctp_setsockopt_disable_fragments(struct sock
*sk
,
2185 char __user
*optval
,
2186 unsigned int optlen
)
2190 if (optlen
< sizeof(int))
2193 if (get_user(val
, (int __user
*)optval
))
2196 sctp_sk(sk
)->disable_fragments
= (val
== 0) ? 0 : 1;
2201 static int sctp_setsockopt_events(struct sock
*sk
, char __user
*optval
,
2202 unsigned int optlen
)
2204 struct sctp_association
*asoc
;
2205 struct sctp_ulpevent
*event
;
2207 if (optlen
> sizeof(struct sctp_event_subscribe
))
2209 if (copy_from_user(&sctp_sk(sk
)->subscribe
, optval
, optlen
))
2212 if (sctp_sk(sk
)->subscribe
.sctp_data_io_event
)
2213 pr_warn_ratelimited(DEPRECATED
"%s (pid %d) "
2214 "Requested SCTP_SNDRCVINFO event.\n"
2215 "Use SCTP_RCVINFO through SCTP_RECVRCVINFO option instead.\n",
2216 current
->comm
, task_pid_nr(current
));
2218 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2219 * if there is no data to be sent or retransmit, the stack will
2220 * immediately send up this notification.
2222 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT
,
2223 &sctp_sk(sk
)->subscribe
)) {
2224 asoc
= sctp_id2assoc(sk
, 0);
2226 if (asoc
&& sctp_outq_is_empty(&asoc
->outqueue
)) {
2227 event
= sctp_ulpevent_make_sender_dry_event(asoc
,
2232 sctp_ulpq_tail_event(&asoc
->ulpq
, event
);
2239 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2241 * This socket option is applicable to the UDP-style socket only. When
2242 * set it will cause associations that are idle for more than the
2243 * specified number of seconds to automatically close. An association
2244 * being idle is defined an association that has NOT sent or received
2245 * user data. The special value of '0' indicates that no automatic
2246 * close of any associations should be performed. The option expects an
2247 * integer defining the number of seconds of idle time before an
2248 * association is closed.
2250 static int sctp_setsockopt_autoclose(struct sock
*sk
, char __user
*optval
,
2251 unsigned int optlen
)
2253 struct sctp_sock
*sp
= sctp_sk(sk
);
2254 struct net
*net
= sock_net(sk
);
2256 /* Applicable to UDP-style socket only */
2257 if (sctp_style(sk
, TCP
))
2259 if (optlen
!= sizeof(int))
2261 if (copy_from_user(&sp
->autoclose
, optval
, optlen
))
2264 if (sp
->autoclose
> net
->sctp
.max_autoclose
)
2265 sp
->autoclose
= net
->sctp
.max_autoclose
;
2270 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2272 * Applications can enable or disable heartbeats for any peer address of
2273 * an association, modify an address's heartbeat interval, force a
2274 * heartbeat to be sent immediately, and adjust the address's maximum
2275 * number of retransmissions sent before an address is considered
2276 * unreachable. The following structure is used to access and modify an
2277 * address's parameters:
2279 * struct sctp_paddrparams {
2280 * sctp_assoc_t spp_assoc_id;
2281 * struct sockaddr_storage spp_address;
2282 * uint32_t spp_hbinterval;
2283 * uint16_t spp_pathmaxrxt;
2284 * uint32_t spp_pathmtu;
2285 * uint32_t spp_sackdelay;
2286 * uint32_t spp_flags;
2289 * spp_assoc_id - (one-to-many style socket) This is filled in the
2290 * application, and identifies the association for
2292 * spp_address - This specifies which address is of interest.
2293 * spp_hbinterval - This contains the value of the heartbeat interval,
2294 * in milliseconds. If a value of zero
2295 * is present in this field then no changes are to
2296 * be made to this parameter.
2297 * spp_pathmaxrxt - This contains the maximum number of
2298 * retransmissions before this address shall be
2299 * considered unreachable. If a value of zero
2300 * is present in this field then no changes are to
2301 * be made to this parameter.
2302 * spp_pathmtu - When Path MTU discovery is disabled the value
2303 * specified here will be the "fixed" path mtu.
2304 * Note that if the spp_address field is empty
2305 * then all associations on this address will
2306 * have this fixed path mtu set upon them.
2308 * spp_sackdelay - When delayed sack is enabled, this value specifies
2309 * the number of milliseconds that sacks will be delayed
2310 * for. This value will apply to all addresses of an
2311 * association if the spp_address field is empty. Note
2312 * also, that if delayed sack is enabled and this
2313 * value is set to 0, no change is made to the last
2314 * recorded delayed sack timer value.
2316 * spp_flags - These flags are used to control various features
2317 * on an association. The flag field may contain
2318 * zero or more of the following options.
2320 * SPP_HB_ENABLE - Enable heartbeats on the
2321 * specified address. Note that if the address
2322 * field is empty all addresses for the association
2323 * have heartbeats enabled upon them.
2325 * SPP_HB_DISABLE - Disable heartbeats on the
2326 * speicifed address. Note that if the address
2327 * field is empty all addresses for the association
2328 * will have their heartbeats disabled. Note also
2329 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2330 * mutually exclusive, only one of these two should
2331 * be specified. Enabling both fields will have
2332 * undetermined results.
2334 * SPP_HB_DEMAND - Request a user initiated heartbeat
2335 * to be made immediately.
2337 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2338 * heartbeat delayis to be set to the value of 0
2341 * SPP_PMTUD_ENABLE - This field will enable PMTU
2342 * discovery upon the specified address. Note that
2343 * if the address feild is empty then all addresses
2344 * on the association are effected.
2346 * SPP_PMTUD_DISABLE - This field will disable PMTU
2347 * discovery upon the specified address. Note that
2348 * if the address feild is empty then all addresses
2349 * on the association are effected. Not also that
2350 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2351 * exclusive. Enabling both will have undetermined
2354 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2355 * on delayed sack. The time specified in spp_sackdelay
2356 * is used to specify the sack delay for this address. Note
2357 * that if spp_address is empty then all addresses will
2358 * enable delayed sack and take on the sack delay
2359 * value specified in spp_sackdelay.
2360 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2361 * off delayed sack. If the spp_address field is blank then
2362 * delayed sack is disabled for the entire association. Note
2363 * also that this field is mutually exclusive to
2364 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2367 static int sctp_apply_peer_addr_params(struct sctp_paddrparams
*params
,
2368 struct sctp_transport
*trans
,
2369 struct sctp_association
*asoc
,
2370 struct sctp_sock
*sp
,
2373 int sackdelay_change
)
2377 if (params
->spp_flags
& SPP_HB_DEMAND
&& trans
) {
2378 struct net
*net
= sock_net(trans
->asoc
->base
.sk
);
2380 error
= sctp_primitive_REQUESTHEARTBEAT(net
, trans
->asoc
, trans
);
2385 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2386 * this field is ignored. Note also that a value of zero indicates
2387 * the current setting should be left unchanged.
2389 if (params
->spp_flags
& SPP_HB_ENABLE
) {
2391 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2392 * set. This lets us use 0 value when this flag
2395 if (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)
2396 params
->spp_hbinterval
= 0;
2398 if (params
->spp_hbinterval
||
2399 (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)) {
2402 msecs_to_jiffies(params
->spp_hbinterval
);
2405 msecs_to_jiffies(params
->spp_hbinterval
);
2407 sp
->hbinterval
= params
->spp_hbinterval
;
2414 trans
->param_flags
=
2415 (trans
->param_flags
& ~SPP_HB
) | hb_change
;
2418 (asoc
->param_flags
& ~SPP_HB
) | hb_change
;
2421 (sp
->param_flags
& ~SPP_HB
) | hb_change
;
2425 /* When Path MTU discovery is disabled the value specified here will
2426 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2427 * include the flag SPP_PMTUD_DISABLE for this field to have any
2430 if ((params
->spp_flags
& SPP_PMTUD_DISABLE
) && params
->spp_pathmtu
) {
2432 trans
->pathmtu
= params
->spp_pathmtu
;
2433 sctp_assoc_sync_pmtu(sctp_opt2sk(sp
), asoc
);
2435 asoc
->pathmtu
= params
->spp_pathmtu
;
2436 sctp_frag_point(asoc
, params
->spp_pathmtu
);
2438 sp
->pathmtu
= params
->spp_pathmtu
;
2444 int update
= (trans
->param_flags
& SPP_PMTUD_DISABLE
) &&
2445 (params
->spp_flags
& SPP_PMTUD_ENABLE
);
2446 trans
->param_flags
=
2447 (trans
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2449 sctp_transport_pmtu(trans
, sctp_opt2sk(sp
));
2450 sctp_assoc_sync_pmtu(sctp_opt2sk(sp
), asoc
);
2454 (asoc
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2457 (sp
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2461 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2462 * value of this field is ignored. Note also that a value of zero
2463 * indicates the current setting should be left unchanged.
2465 if ((params
->spp_flags
& SPP_SACKDELAY_ENABLE
) && params
->spp_sackdelay
) {
2468 msecs_to_jiffies(params
->spp_sackdelay
);
2471 msecs_to_jiffies(params
->spp_sackdelay
);
2473 sp
->sackdelay
= params
->spp_sackdelay
;
2477 if (sackdelay_change
) {
2479 trans
->param_flags
=
2480 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2484 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2488 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2493 /* Note that a value of zero indicates the current setting should be
2496 if (params
->spp_pathmaxrxt
) {
2498 trans
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2500 asoc
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2502 sp
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2509 static int sctp_setsockopt_peer_addr_params(struct sock
*sk
,
2510 char __user
*optval
,
2511 unsigned int optlen
)
2513 struct sctp_paddrparams params
;
2514 struct sctp_transport
*trans
= NULL
;
2515 struct sctp_association
*asoc
= NULL
;
2516 struct sctp_sock
*sp
= sctp_sk(sk
);
2518 int hb_change
, pmtud_change
, sackdelay_change
;
2520 if (optlen
!= sizeof(struct sctp_paddrparams
))
2523 if (copy_from_user(¶ms
, optval
, optlen
))
2526 /* Validate flags and value parameters. */
2527 hb_change
= params
.spp_flags
& SPP_HB
;
2528 pmtud_change
= params
.spp_flags
& SPP_PMTUD
;
2529 sackdelay_change
= params
.spp_flags
& SPP_SACKDELAY
;
2531 if (hb_change
== SPP_HB
||
2532 pmtud_change
== SPP_PMTUD
||
2533 sackdelay_change
== SPP_SACKDELAY
||
2534 params
.spp_sackdelay
> 500 ||
2535 (params
.spp_pathmtu
&&
2536 params
.spp_pathmtu
< SCTP_DEFAULT_MINSEGMENT
))
2539 /* If an address other than INADDR_ANY is specified, and
2540 * no transport is found, then the request is invalid.
2542 if (!sctp_is_any(sk
, (union sctp_addr
*)¶ms
.spp_address
)) {
2543 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
2544 params
.spp_assoc_id
);
2549 /* Get association, if assoc_id != 0 and the socket is a one
2550 * to many style socket, and an association was not found, then
2551 * the id was invalid.
2553 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
2554 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
))
2557 /* Heartbeat demand can only be sent on a transport or
2558 * association, but not a socket.
2560 if (params
.spp_flags
& SPP_HB_DEMAND
&& !trans
&& !asoc
)
2563 /* Process parameters. */
2564 error
= sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2565 hb_change
, pmtud_change
,
2571 /* If changes are for association, also apply parameters to each
2574 if (!trans
&& asoc
) {
2575 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2577 sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2578 hb_change
, pmtud_change
,
2586 static inline __u32
sctp_spp_sackdelay_enable(__u32 param_flags
)
2588 return (param_flags
& ~SPP_SACKDELAY
) | SPP_SACKDELAY_ENABLE
;
2591 static inline __u32
sctp_spp_sackdelay_disable(__u32 param_flags
)
2593 return (param_flags
& ~SPP_SACKDELAY
) | SPP_SACKDELAY_DISABLE
;
2597 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2599 * This option will effect the way delayed acks are performed. This
2600 * option allows you to get or set the delayed ack time, in
2601 * milliseconds. It also allows changing the delayed ack frequency.
2602 * Changing the frequency to 1 disables the delayed sack algorithm. If
2603 * the assoc_id is 0, then this sets or gets the endpoints default
2604 * values. If the assoc_id field is non-zero, then the set or get
2605 * effects the specified association for the one to many model (the
2606 * assoc_id field is ignored by the one to one model). Note that if
2607 * sack_delay or sack_freq are 0 when setting this option, then the
2608 * current values will remain unchanged.
2610 * struct sctp_sack_info {
2611 * sctp_assoc_t sack_assoc_id;
2612 * uint32_t sack_delay;
2613 * uint32_t sack_freq;
2616 * sack_assoc_id - This parameter, indicates which association the user
2617 * is performing an action upon. Note that if this field's value is
2618 * zero then the endpoints default value is changed (effecting future
2619 * associations only).
2621 * sack_delay - This parameter contains the number of milliseconds that
2622 * the user is requesting the delayed ACK timer be set to. Note that
2623 * this value is defined in the standard to be between 200 and 500
2626 * sack_freq - This parameter contains the number of packets that must
2627 * be received before a sack is sent without waiting for the delay
2628 * timer to expire. The default value for this is 2, setting this
2629 * value to 1 will disable the delayed sack algorithm.
2632 static int sctp_setsockopt_delayed_ack(struct sock
*sk
,
2633 char __user
*optval
, unsigned int optlen
)
2635 struct sctp_sack_info params
;
2636 struct sctp_transport
*trans
= NULL
;
2637 struct sctp_association
*asoc
= NULL
;
2638 struct sctp_sock
*sp
= sctp_sk(sk
);
2640 if (optlen
== sizeof(struct sctp_sack_info
)) {
2641 if (copy_from_user(¶ms
, optval
, optlen
))
2644 if (params
.sack_delay
== 0 && params
.sack_freq
== 0)
2646 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
2647 pr_warn_ratelimited(DEPRECATED
2649 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2650 "Use struct sctp_sack_info instead\n",
2651 current
->comm
, task_pid_nr(current
));
2652 if (copy_from_user(¶ms
, optval
, optlen
))
2655 if (params
.sack_delay
== 0)
2656 params
.sack_freq
= 1;
2658 params
.sack_freq
= 0;
2662 /* Validate value parameter. */
2663 if (params
.sack_delay
> 500)
2666 /* Get association, if sack_assoc_id != 0 and the socket is a one
2667 * to many style socket, and an association was not found, then
2668 * the id was invalid.
2670 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
2671 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
2674 if (params
.sack_delay
) {
2677 msecs_to_jiffies(params
.sack_delay
);
2679 sctp_spp_sackdelay_enable(asoc
->param_flags
);
2681 sp
->sackdelay
= params
.sack_delay
;
2683 sctp_spp_sackdelay_enable(sp
->param_flags
);
2687 if (params
.sack_freq
== 1) {
2690 sctp_spp_sackdelay_disable(asoc
->param_flags
);
2693 sctp_spp_sackdelay_disable(sp
->param_flags
);
2695 } else if (params
.sack_freq
> 1) {
2697 asoc
->sackfreq
= params
.sack_freq
;
2699 sctp_spp_sackdelay_enable(asoc
->param_flags
);
2701 sp
->sackfreq
= params
.sack_freq
;
2703 sctp_spp_sackdelay_enable(sp
->param_flags
);
2707 /* If change is for association, also apply to each transport. */
2709 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2711 if (params
.sack_delay
) {
2713 msecs_to_jiffies(params
.sack_delay
);
2714 trans
->param_flags
=
2715 sctp_spp_sackdelay_enable(trans
->param_flags
);
2717 if (params
.sack_freq
== 1) {
2718 trans
->param_flags
=
2719 sctp_spp_sackdelay_disable(trans
->param_flags
);
2720 } else if (params
.sack_freq
> 1) {
2721 trans
->sackfreq
= params
.sack_freq
;
2722 trans
->param_flags
=
2723 sctp_spp_sackdelay_enable(trans
->param_flags
);
2731 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2733 * Applications can specify protocol parameters for the default association
2734 * initialization. The option name argument to setsockopt() and getsockopt()
2737 * Setting initialization parameters is effective only on an unconnected
2738 * socket (for UDP-style sockets only future associations are effected
2739 * by the change). With TCP-style sockets, this option is inherited by
2740 * sockets derived from a listener socket.
2742 static int sctp_setsockopt_initmsg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2744 struct sctp_initmsg sinit
;
2745 struct sctp_sock
*sp
= sctp_sk(sk
);
2747 if (optlen
!= sizeof(struct sctp_initmsg
))
2749 if (copy_from_user(&sinit
, optval
, optlen
))
2752 if (sinit
.sinit_num_ostreams
)
2753 sp
->initmsg
.sinit_num_ostreams
= sinit
.sinit_num_ostreams
;
2754 if (sinit
.sinit_max_instreams
)
2755 sp
->initmsg
.sinit_max_instreams
= sinit
.sinit_max_instreams
;
2756 if (sinit
.sinit_max_attempts
)
2757 sp
->initmsg
.sinit_max_attempts
= sinit
.sinit_max_attempts
;
2758 if (sinit
.sinit_max_init_timeo
)
2759 sp
->initmsg
.sinit_max_init_timeo
= sinit
.sinit_max_init_timeo
;
2765 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2767 * Applications that wish to use the sendto() system call may wish to
2768 * specify a default set of parameters that would normally be supplied
2769 * through the inclusion of ancillary data. This socket option allows
2770 * such an application to set the default sctp_sndrcvinfo structure.
2771 * The application that wishes to use this socket option simply passes
2772 * in to this call the sctp_sndrcvinfo structure defined in Section
2773 * 5.2.2) The input parameters accepted by this call include
2774 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2775 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2776 * to this call if the caller is using the UDP model.
2778 static int sctp_setsockopt_default_send_param(struct sock
*sk
,
2779 char __user
*optval
,
2780 unsigned int optlen
)
2782 struct sctp_sock
*sp
= sctp_sk(sk
);
2783 struct sctp_association
*asoc
;
2784 struct sctp_sndrcvinfo info
;
2786 if (optlen
!= sizeof(info
))
2788 if (copy_from_user(&info
, optval
, optlen
))
2790 if (info
.sinfo_flags
&
2791 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
2792 SCTP_ABORT
| SCTP_EOF
))
2795 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
2796 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
2799 asoc
->default_stream
= info
.sinfo_stream
;
2800 asoc
->default_flags
= info
.sinfo_flags
;
2801 asoc
->default_ppid
= info
.sinfo_ppid
;
2802 asoc
->default_context
= info
.sinfo_context
;
2803 asoc
->default_timetolive
= info
.sinfo_timetolive
;
2805 sp
->default_stream
= info
.sinfo_stream
;
2806 sp
->default_flags
= info
.sinfo_flags
;
2807 sp
->default_ppid
= info
.sinfo_ppid
;
2808 sp
->default_context
= info
.sinfo_context
;
2809 sp
->default_timetolive
= info
.sinfo_timetolive
;
2815 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2816 * (SCTP_DEFAULT_SNDINFO)
2818 static int sctp_setsockopt_default_sndinfo(struct sock
*sk
,
2819 char __user
*optval
,
2820 unsigned int optlen
)
2822 struct sctp_sock
*sp
= sctp_sk(sk
);
2823 struct sctp_association
*asoc
;
2824 struct sctp_sndinfo info
;
2826 if (optlen
!= sizeof(info
))
2828 if (copy_from_user(&info
, optval
, optlen
))
2830 if (info
.snd_flags
&
2831 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
2832 SCTP_ABORT
| SCTP_EOF
))
2835 asoc
= sctp_id2assoc(sk
, info
.snd_assoc_id
);
2836 if (!asoc
&& info
.snd_assoc_id
&& sctp_style(sk
, UDP
))
2839 asoc
->default_stream
= info
.snd_sid
;
2840 asoc
->default_flags
= info
.snd_flags
;
2841 asoc
->default_ppid
= info
.snd_ppid
;
2842 asoc
->default_context
= info
.snd_context
;
2844 sp
->default_stream
= info
.snd_sid
;
2845 sp
->default_flags
= info
.snd_flags
;
2846 sp
->default_ppid
= info
.snd_ppid
;
2847 sp
->default_context
= info
.snd_context
;
2853 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2855 * Requests that the local SCTP stack use the enclosed peer address as
2856 * the association primary. The enclosed address must be one of the
2857 * association peer's addresses.
2859 static int sctp_setsockopt_primary_addr(struct sock
*sk
, char __user
*optval
,
2860 unsigned int optlen
)
2862 struct sctp_prim prim
;
2863 struct sctp_transport
*trans
;
2865 if (optlen
!= sizeof(struct sctp_prim
))
2868 if (copy_from_user(&prim
, optval
, sizeof(struct sctp_prim
)))
2871 trans
= sctp_addr_id2transport(sk
, &prim
.ssp_addr
, prim
.ssp_assoc_id
);
2875 sctp_assoc_set_primary(trans
->asoc
, trans
);
2881 * 7.1.5 SCTP_NODELAY
2883 * Turn on/off any Nagle-like algorithm. This means that packets are
2884 * generally sent as soon as possible and no unnecessary delays are
2885 * introduced, at the cost of more packets in the network. Expects an
2886 * integer boolean flag.
2888 static int sctp_setsockopt_nodelay(struct sock
*sk
, char __user
*optval
,
2889 unsigned int optlen
)
2893 if (optlen
< sizeof(int))
2895 if (get_user(val
, (int __user
*)optval
))
2898 sctp_sk(sk
)->nodelay
= (val
== 0) ? 0 : 1;
2904 * 7.1.1 SCTP_RTOINFO
2906 * The protocol parameters used to initialize and bound retransmission
2907 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2908 * and modify these parameters.
2909 * All parameters are time values, in milliseconds. A value of 0, when
2910 * modifying the parameters, indicates that the current value should not
2914 static int sctp_setsockopt_rtoinfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2916 struct sctp_rtoinfo rtoinfo
;
2917 struct sctp_association
*asoc
;
2918 unsigned long rto_min
, rto_max
;
2919 struct sctp_sock
*sp
= sctp_sk(sk
);
2921 if (optlen
!= sizeof (struct sctp_rtoinfo
))
2924 if (copy_from_user(&rtoinfo
, optval
, optlen
))
2927 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
2929 /* Set the values to the specific association */
2930 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
2933 rto_max
= rtoinfo
.srto_max
;
2934 rto_min
= rtoinfo
.srto_min
;
2937 rto_max
= asoc
? msecs_to_jiffies(rto_max
) : rto_max
;
2939 rto_max
= asoc
? asoc
->rto_max
: sp
->rtoinfo
.srto_max
;
2942 rto_min
= asoc
? msecs_to_jiffies(rto_min
) : rto_min
;
2944 rto_min
= asoc
? asoc
->rto_min
: sp
->rtoinfo
.srto_min
;
2946 if (rto_min
> rto_max
)
2950 if (rtoinfo
.srto_initial
!= 0)
2952 msecs_to_jiffies(rtoinfo
.srto_initial
);
2953 asoc
->rto_max
= rto_max
;
2954 asoc
->rto_min
= rto_min
;
2956 /* If there is no association or the association-id = 0
2957 * set the values to the endpoint.
2959 if (rtoinfo
.srto_initial
!= 0)
2960 sp
->rtoinfo
.srto_initial
= rtoinfo
.srto_initial
;
2961 sp
->rtoinfo
.srto_max
= rto_max
;
2962 sp
->rtoinfo
.srto_min
= rto_min
;
2970 * 7.1.2 SCTP_ASSOCINFO
2972 * This option is used to tune the maximum retransmission attempts
2973 * of the association.
2974 * Returns an error if the new association retransmission value is
2975 * greater than the sum of the retransmission value of the peer.
2976 * See [SCTP] for more information.
2979 static int sctp_setsockopt_associnfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2982 struct sctp_assocparams assocparams
;
2983 struct sctp_association
*asoc
;
2985 if (optlen
!= sizeof(struct sctp_assocparams
))
2987 if (copy_from_user(&assocparams
, optval
, optlen
))
2990 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
2992 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
2995 /* Set the values to the specific association */
2997 if (assocparams
.sasoc_asocmaxrxt
!= 0) {
3000 struct sctp_transport
*peer_addr
;
3002 list_for_each_entry(peer_addr
, &asoc
->peer
.transport_addr_list
,
3004 path_sum
+= peer_addr
->pathmaxrxt
;
3008 /* Only validate asocmaxrxt if we have more than
3009 * one path/transport. We do this because path
3010 * retransmissions are only counted when we have more
3014 assocparams
.sasoc_asocmaxrxt
> path_sum
)
3017 asoc
->max_retrans
= assocparams
.sasoc_asocmaxrxt
;
3020 if (assocparams
.sasoc_cookie_life
!= 0)
3021 asoc
->cookie_life
= ms_to_ktime(assocparams
.sasoc_cookie_life
);
3023 /* Set the values to the endpoint */
3024 struct sctp_sock
*sp
= sctp_sk(sk
);
3026 if (assocparams
.sasoc_asocmaxrxt
!= 0)
3027 sp
->assocparams
.sasoc_asocmaxrxt
=
3028 assocparams
.sasoc_asocmaxrxt
;
3029 if (assocparams
.sasoc_cookie_life
!= 0)
3030 sp
->assocparams
.sasoc_cookie_life
=
3031 assocparams
.sasoc_cookie_life
;
3037 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3039 * This socket option is a boolean flag which turns on or off mapped V4
3040 * addresses. If this option is turned on and the socket is type
3041 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3042 * If this option is turned off, then no mapping will be done of V4
3043 * addresses and a user will receive both PF_INET6 and PF_INET type
3044 * addresses on the socket.
3046 static int sctp_setsockopt_mappedv4(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
3049 struct sctp_sock
*sp
= sctp_sk(sk
);
3051 if (optlen
< sizeof(int))
3053 if (get_user(val
, (int __user
*)optval
))
3064 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3065 * This option will get or set the maximum size to put in any outgoing
3066 * SCTP DATA chunk. If a message is larger than this size it will be
3067 * fragmented by SCTP into the specified size. Note that the underlying
3068 * SCTP implementation may fragment into smaller sized chunks when the
3069 * PMTU of the underlying association is smaller than the value set by
3070 * the user. The default value for this option is '0' which indicates
3071 * the user is NOT limiting fragmentation and only the PMTU will effect
3072 * SCTP's choice of DATA chunk size. Note also that values set larger
3073 * than the maximum size of an IP datagram will effectively let SCTP
3074 * control fragmentation (i.e. the same as setting this option to 0).
3076 * The following structure is used to access and modify this parameter:
3078 * struct sctp_assoc_value {
3079 * sctp_assoc_t assoc_id;
3080 * uint32_t assoc_value;
3083 * assoc_id: This parameter is ignored for one-to-one style sockets.
3084 * For one-to-many style sockets this parameter indicates which
3085 * association the user is performing an action upon. Note that if
3086 * this field's value is zero then the endpoints default value is
3087 * changed (effecting future associations only).
3088 * assoc_value: This parameter specifies the maximum size in bytes.
3090 static int sctp_setsockopt_maxseg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
3092 struct sctp_assoc_value params
;
3093 struct sctp_association
*asoc
;
3094 struct sctp_sock
*sp
= sctp_sk(sk
);
3097 if (optlen
== sizeof(int)) {
3098 pr_warn_ratelimited(DEPRECATED
3100 "Use of int in maxseg socket option.\n"
3101 "Use struct sctp_assoc_value instead\n",
3102 current
->comm
, task_pid_nr(current
));
3103 if (copy_from_user(&val
, optval
, optlen
))
3105 params
.assoc_id
= 0;
3106 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
3107 if (copy_from_user(¶ms
, optval
, optlen
))
3109 val
= params
.assoc_value
;
3113 if ((val
!= 0) && ((val
< 8) || (val
> SCTP_MAX_CHUNK_LEN
)))
3116 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3117 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
3122 val
= asoc
->pathmtu
;
3123 val
-= sp
->pf
->af
->net_header_len
;
3124 val
-= sizeof(struct sctphdr
) +
3125 sizeof(struct sctp_data_chunk
);
3127 asoc
->user_frag
= val
;
3128 asoc
->frag_point
= sctp_frag_point(asoc
, asoc
->pathmtu
);
3130 sp
->user_frag
= val
;
3138 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3140 * Requests that the peer mark the enclosed address as the association
3141 * primary. The enclosed address must be one of the association's
3142 * locally bound addresses. The following structure is used to make a
3143 * set primary request:
3145 static int sctp_setsockopt_peer_primary_addr(struct sock
*sk
, char __user
*optval
,
3146 unsigned int optlen
)
3148 struct net
*net
= sock_net(sk
);
3149 struct sctp_sock
*sp
;
3150 struct sctp_association
*asoc
= NULL
;
3151 struct sctp_setpeerprim prim
;
3152 struct sctp_chunk
*chunk
;
3158 if (!net
->sctp
.addip_enable
)
3161 if (optlen
!= sizeof(struct sctp_setpeerprim
))
3164 if (copy_from_user(&prim
, optval
, optlen
))
3167 asoc
= sctp_id2assoc(sk
, prim
.sspp_assoc_id
);
3171 if (!asoc
->peer
.asconf_capable
)
3174 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_SET_PRIMARY
)
3177 if (!sctp_state(asoc
, ESTABLISHED
))
3180 af
= sctp_get_af_specific(prim
.sspp_addr
.ss_family
);
3184 if (!af
->addr_valid((union sctp_addr
*)&prim
.sspp_addr
, sp
, NULL
))
3185 return -EADDRNOTAVAIL
;
3187 if (!sctp_assoc_lookup_laddr(asoc
, (union sctp_addr
*)&prim
.sspp_addr
))
3188 return -EADDRNOTAVAIL
;
3190 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3191 chunk
= sctp_make_asconf_set_prim(asoc
,
3192 (union sctp_addr
*)&prim
.sspp_addr
);
3196 err
= sctp_send_asconf(asoc
, chunk
);
3198 pr_debug("%s: we set peer primary addr primitively\n", __func__
);
3203 static int sctp_setsockopt_adaptation_layer(struct sock
*sk
, char __user
*optval
,
3204 unsigned int optlen
)
3206 struct sctp_setadaptation adaptation
;
3208 if (optlen
!= sizeof(struct sctp_setadaptation
))
3210 if (copy_from_user(&adaptation
, optval
, optlen
))
3213 sctp_sk(sk
)->adaptation_ind
= adaptation
.ssb_adaptation_ind
;
3219 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3221 * The context field in the sctp_sndrcvinfo structure is normally only
3222 * used when a failed message is retrieved holding the value that was
3223 * sent down on the actual send call. This option allows the setting of
3224 * a default context on an association basis that will be received on
3225 * reading messages from the peer. This is especially helpful in the
3226 * one-2-many model for an application to keep some reference to an
3227 * internal state machine that is processing messages on the
3228 * association. Note that the setting of this value only effects
3229 * received messages from the peer and does not effect the value that is
3230 * saved with outbound messages.
3232 static int sctp_setsockopt_context(struct sock
*sk
, char __user
*optval
,
3233 unsigned int optlen
)
3235 struct sctp_assoc_value params
;
3236 struct sctp_sock
*sp
;
3237 struct sctp_association
*asoc
;
3239 if (optlen
!= sizeof(struct sctp_assoc_value
))
3241 if (copy_from_user(¶ms
, optval
, optlen
))
3246 if (params
.assoc_id
!= 0) {
3247 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3250 asoc
->default_rcv_context
= params
.assoc_value
;
3252 sp
->default_rcv_context
= params
.assoc_value
;
3259 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3261 * This options will at a minimum specify if the implementation is doing
3262 * fragmented interleave. Fragmented interleave, for a one to many
3263 * socket, is when subsequent calls to receive a message may return
3264 * parts of messages from different associations. Some implementations
3265 * may allow you to turn this value on or off. If so, when turned off,
3266 * no fragment interleave will occur (which will cause a head of line
3267 * blocking amongst multiple associations sharing the same one to many
3268 * socket). When this option is turned on, then each receive call may
3269 * come from a different association (thus the user must receive data
3270 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3271 * association each receive belongs to.
3273 * This option takes a boolean value. A non-zero value indicates that
3274 * fragmented interleave is on. A value of zero indicates that
3275 * fragmented interleave is off.
3277 * Note that it is important that an implementation that allows this
3278 * option to be turned on, have it off by default. Otherwise an unaware
3279 * application using the one to many model may become confused and act
3282 static int sctp_setsockopt_fragment_interleave(struct sock
*sk
,
3283 char __user
*optval
,
3284 unsigned int optlen
)
3288 if (optlen
!= sizeof(int))
3290 if (get_user(val
, (int __user
*)optval
))
3293 sctp_sk(sk
)->frag_interleave
= (val
== 0) ? 0 : 1;
3299 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3300 * (SCTP_PARTIAL_DELIVERY_POINT)
3302 * This option will set or get the SCTP partial delivery point. This
3303 * point is the size of a message where the partial delivery API will be
3304 * invoked to help free up rwnd space for the peer. Setting this to a
3305 * lower value will cause partial deliveries to happen more often. The
3306 * calls argument is an integer that sets or gets the partial delivery
3307 * point. Note also that the call will fail if the user attempts to set
3308 * this value larger than the socket receive buffer size.
3310 * Note that any single message having a length smaller than or equal to
3311 * the SCTP partial delivery point will be delivered in one single read
3312 * call as long as the user provided buffer is large enough to hold the
3315 static int sctp_setsockopt_partial_delivery_point(struct sock
*sk
,
3316 char __user
*optval
,
3317 unsigned int optlen
)
3321 if (optlen
!= sizeof(u32
))
3323 if (get_user(val
, (int __user
*)optval
))
3326 /* Note: We double the receive buffer from what the user sets
3327 * it to be, also initial rwnd is based on rcvbuf/2.
3329 if (val
> (sk
->sk_rcvbuf
>> 1))
3332 sctp_sk(sk
)->pd_point
= val
;
3334 return 0; /* is this the right error code? */
3338 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3340 * This option will allow a user to change the maximum burst of packets
3341 * that can be emitted by this association. Note that the default value
3342 * is 4, and some implementations may restrict this setting so that it
3343 * can only be lowered.
3345 * NOTE: This text doesn't seem right. Do this on a socket basis with
3346 * future associations inheriting the socket value.
3348 static int sctp_setsockopt_maxburst(struct sock
*sk
,
3349 char __user
*optval
,
3350 unsigned int optlen
)
3352 struct sctp_assoc_value params
;
3353 struct sctp_sock
*sp
;
3354 struct sctp_association
*asoc
;
3358 if (optlen
== sizeof(int)) {
3359 pr_warn_ratelimited(DEPRECATED
3361 "Use of int in max_burst socket option deprecated.\n"
3362 "Use struct sctp_assoc_value instead\n",
3363 current
->comm
, task_pid_nr(current
));
3364 if (copy_from_user(&val
, optval
, optlen
))
3366 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
3367 if (copy_from_user(¶ms
, optval
, optlen
))
3369 val
= params
.assoc_value
;
3370 assoc_id
= params
.assoc_id
;
3376 if (assoc_id
!= 0) {
3377 asoc
= sctp_id2assoc(sk
, assoc_id
);
3380 asoc
->max_burst
= val
;
3382 sp
->max_burst
= val
;
3388 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3390 * This set option adds a chunk type that the user is requesting to be
3391 * received only in an authenticated way. Changes to the list of chunks
3392 * will only effect future associations on the socket.
3394 static int sctp_setsockopt_auth_chunk(struct sock
*sk
,
3395 char __user
*optval
,
3396 unsigned int optlen
)
3398 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3399 struct sctp_authchunk val
;
3401 if (!ep
->auth_enable
)
3404 if (optlen
!= sizeof(struct sctp_authchunk
))
3406 if (copy_from_user(&val
, optval
, optlen
))
3409 switch (val
.sauth_chunk
) {
3411 case SCTP_CID_INIT_ACK
:
3412 case SCTP_CID_SHUTDOWN_COMPLETE
:
3417 /* add this chunk id to the endpoint */
3418 return sctp_auth_ep_add_chunkid(ep
, val
.sauth_chunk
);
3422 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3424 * This option gets or sets the list of HMAC algorithms that the local
3425 * endpoint requires the peer to use.
3427 static int sctp_setsockopt_hmac_ident(struct sock
*sk
,
3428 char __user
*optval
,
3429 unsigned int optlen
)
3431 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3432 struct sctp_hmacalgo
*hmacs
;
3436 if (!ep
->auth_enable
)
3439 if (optlen
< sizeof(struct sctp_hmacalgo
))
3442 hmacs
= memdup_user(optval
, optlen
);
3444 return PTR_ERR(hmacs
);
3446 idents
= hmacs
->shmac_num_idents
;
3447 if (idents
== 0 || idents
> SCTP_AUTH_NUM_HMACS
||
3448 (idents
* sizeof(u16
)) > (optlen
- sizeof(struct sctp_hmacalgo
))) {
3453 err
= sctp_auth_ep_set_hmacs(ep
, hmacs
);
3460 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3462 * This option will set a shared secret key which is used to build an
3463 * association shared key.
3465 static int sctp_setsockopt_auth_key(struct sock
*sk
,
3466 char __user
*optval
,
3467 unsigned int optlen
)
3469 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3470 struct sctp_authkey
*authkey
;
3471 struct sctp_association
*asoc
;
3474 if (!ep
->auth_enable
)
3477 if (optlen
<= sizeof(struct sctp_authkey
))
3480 authkey
= memdup_user(optval
, optlen
);
3481 if (IS_ERR(authkey
))
3482 return PTR_ERR(authkey
);
3484 if (authkey
->sca_keylength
> optlen
- sizeof(struct sctp_authkey
)) {
3489 asoc
= sctp_id2assoc(sk
, authkey
->sca_assoc_id
);
3490 if (!asoc
&& authkey
->sca_assoc_id
&& sctp_style(sk
, UDP
)) {
3495 ret
= sctp_auth_set_key(ep
, asoc
, authkey
);
3502 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3504 * This option will get or set the active shared key to be used to build
3505 * the association shared key.
3507 static int sctp_setsockopt_active_key(struct sock
*sk
,
3508 char __user
*optval
,
3509 unsigned int optlen
)
3511 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3512 struct sctp_authkeyid val
;
3513 struct sctp_association
*asoc
;
3515 if (!ep
->auth_enable
)
3518 if (optlen
!= sizeof(struct sctp_authkeyid
))
3520 if (copy_from_user(&val
, optval
, optlen
))
3523 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3524 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3527 return sctp_auth_set_active_key(ep
, asoc
, val
.scact_keynumber
);
3531 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3533 * This set option will delete a shared secret key from use.
3535 static int sctp_setsockopt_del_key(struct sock
*sk
,
3536 char __user
*optval
,
3537 unsigned int optlen
)
3539 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3540 struct sctp_authkeyid val
;
3541 struct sctp_association
*asoc
;
3543 if (!ep
->auth_enable
)
3546 if (optlen
!= sizeof(struct sctp_authkeyid
))
3548 if (copy_from_user(&val
, optval
, optlen
))
3551 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3552 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3555 return sctp_auth_del_key_id(ep
, asoc
, val
.scact_keynumber
);
3560 * 8.1.23 SCTP_AUTO_ASCONF
3562 * This option will enable or disable the use of the automatic generation of
3563 * ASCONF chunks to add and delete addresses to an existing association. Note
3564 * that this option has two caveats namely: a) it only affects sockets that
3565 * are bound to all addresses available to the SCTP stack, and b) the system
3566 * administrator may have an overriding control that turns the ASCONF feature
3567 * off no matter what setting the socket option may have.
3568 * This option expects an integer boolean flag, where a non-zero value turns on
3569 * the option, and a zero value turns off the option.
3570 * Note. In this implementation, socket operation overrides default parameter
3571 * being set by sysctl as well as FreeBSD implementation
3573 static int sctp_setsockopt_auto_asconf(struct sock
*sk
, char __user
*optval
,
3574 unsigned int optlen
)
3577 struct sctp_sock
*sp
= sctp_sk(sk
);
3579 if (optlen
< sizeof(int))
3581 if (get_user(val
, (int __user
*)optval
))
3583 if (!sctp_is_ep_boundall(sk
) && val
)
3585 if ((val
&& sp
->do_auto_asconf
) || (!val
&& !sp
->do_auto_asconf
))
3588 if (val
== 0 && sp
->do_auto_asconf
) {
3589 list_del(&sp
->auto_asconf_list
);
3590 sp
->do_auto_asconf
= 0;
3591 } else if (val
&& !sp
->do_auto_asconf
) {
3592 list_add_tail(&sp
->auto_asconf_list
,
3593 &sock_net(sk
)->sctp
.auto_asconf_splist
);
3594 sp
->do_auto_asconf
= 1;
3600 * SCTP_PEER_ADDR_THLDS
3602 * This option allows us to alter the partially failed threshold for one or all
3603 * transports in an association. See Section 6.1 of:
3604 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3606 static int sctp_setsockopt_paddr_thresholds(struct sock
*sk
,
3607 char __user
*optval
,
3608 unsigned int optlen
)
3610 struct sctp_paddrthlds val
;
3611 struct sctp_transport
*trans
;
3612 struct sctp_association
*asoc
;
3614 if (optlen
< sizeof(struct sctp_paddrthlds
))
3616 if (copy_from_user(&val
, (struct sctp_paddrthlds __user
*)optval
,
3617 sizeof(struct sctp_paddrthlds
)))
3621 if (sctp_is_any(sk
, (const union sctp_addr
*)&val
.spt_address
)) {
3622 asoc
= sctp_id2assoc(sk
, val
.spt_assoc_id
);
3625 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
3627 if (val
.spt_pathmaxrxt
)
3628 trans
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3629 trans
->pf_retrans
= val
.spt_pathpfthld
;
3632 if (val
.spt_pathmaxrxt
)
3633 asoc
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3634 asoc
->pf_retrans
= val
.spt_pathpfthld
;
3636 trans
= sctp_addr_id2transport(sk
, &val
.spt_address
,
3641 if (val
.spt_pathmaxrxt
)
3642 trans
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3643 trans
->pf_retrans
= val
.spt_pathpfthld
;
3649 static int sctp_setsockopt_recvrcvinfo(struct sock
*sk
,
3650 char __user
*optval
,
3651 unsigned int optlen
)
3655 if (optlen
< sizeof(int))
3657 if (get_user(val
, (int __user
*) optval
))
3660 sctp_sk(sk
)->recvrcvinfo
= (val
== 0) ? 0 : 1;
3665 static int sctp_setsockopt_recvnxtinfo(struct sock
*sk
,
3666 char __user
*optval
,
3667 unsigned int optlen
)
3671 if (optlen
< sizeof(int))
3673 if (get_user(val
, (int __user
*) optval
))
3676 sctp_sk(sk
)->recvnxtinfo
= (val
== 0) ? 0 : 1;
3681 /* API 6.2 setsockopt(), getsockopt()
3683 * Applications use setsockopt() and getsockopt() to set or retrieve
3684 * socket options. Socket options are used to change the default
3685 * behavior of sockets calls. They are described in Section 7.
3689 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3690 * int __user *optlen);
3691 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3694 * sd - the socket descript.
3695 * level - set to IPPROTO_SCTP for all SCTP options.
3696 * optname - the option name.
3697 * optval - the buffer to store the value of the option.
3698 * optlen - the size of the buffer.
3700 static int sctp_setsockopt(struct sock
*sk
, int level
, int optname
,
3701 char __user
*optval
, unsigned int optlen
)
3705 pr_debug("%s: sk:%p, optname:%d\n", __func__
, sk
, optname
);
3707 /* I can hardly begin to describe how wrong this is. This is
3708 * so broken as to be worse than useless. The API draft
3709 * REALLY is NOT helpful here... I am not convinced that the
3710 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3711 * are at all well-founded.
3713 if (level
!= SOL_SCTP
) {
3714 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
3715 retval
= af
->setsockopt(sk
, level
, optname
, optval
, optlen
);
3722 case SCTP_SOCKOPT_BINDX_ADD
:
3723 /* 'optlen' is the size of the addresses buffer. */
3724 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3725 optlen
, SCTP_BINDX_ADD_ADDR
);
3728 case SCTP_SOCKOPT_BINDX_REM
:
3729 /* 'optlen' is the size of the addresses buffer. */
3730 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3731 optlen
, SCTP_BINDX_REM_ADDR
);
3734 case SCTP_SOCKOPT_CONNECTX_OLD
:
3735 /* 'optlen' is the size of the addresses buffer. */
3736 retval
= sctp_setsockopt_connectx_old(sk
,
3737 (struct sockaddr __user
*)optval
,
3741 case SCTP_SOCKOPT_CONNECTX
:
3742 /* 'optlen' is the size of the addresses buffer. */
3743 retval
= sctp_setsockopt_connectx(sk
,
3744 (struct sockaddr __user
*)optval
,
3748 case SCTP_DISABLE_FRAGMENTS
:
3749 retval
= sctp_setsockopt_disable_fragments(sk
, optval
, optlen
);
3753 retval
= sctp_setsockopt_events(sk
, optval
, optlen
);
3756 case SCTP_AUTOCLOSE
:
3757 retval
= sctp_setsockopt_autoclose(sk
, optval
, optlen
);
3760 case SCTP_PEER_ADDR_PARAMS
:
3761 retval
= sctp_setsockopt_peer_addr_params(sk
, optval
, optlen
);
3764 case SCTP_DELAYED_SACK
:
3765 retval
= sctp_setsockopt_delayed_ack(sk
, optval
, optlen
);
3767 case SCTP_PARTIAL_DELIVERY_POINT
:
3768 retval
= sctp_setsockopt_partial_delivery_point(sk
, optval
, optlen
);
3772 retval
= sctp_setsockopt_initmsg(sk
, optval
, optlen
);
3774 case SCTP_DEFAULT_SEND_PARAM
:
3775 retval
= sctp_setsockopt_default_send_param(sk
, optval
,
3778 case SCTP_DEFAULT_SNDINFO
:
3779 retval
= sctp_setsockopt_default_sndinfo(sk
, optval
, optlen
);
3781 case SCTP_PRIMARY_ADDR
:
3782 retval
= sctp_setsockopt_primary_addr(sk
, optval
, optlen
);
3784 case SCTP_SET_PEER_PRIMARY_ADDR
:
3785 retval
= sctp_setsockopt_peer_primary_addr(sk
, optval
, optlen
);
3788 retval
= sctp_setsockopt_nodelay(sk
, optval
, optlen
);
3791 retval
= sctp_setsockopt_rtoinfo(sk
, optval
, optlen
);
3793 case SCTP_ASSOCINFO
:
3794 retval
= sctp_setsockopt_associnfo(sk
, optval
, optlen
);
3796 case SCTP_I_WANT_MAPPED_V4_ADDR
:
3797 retval
= sctp_setsockopt_mappedv4(sk
, optval
, optlen
);
3800 retval
= sctp_setsockopt_maxseg(sk
, optval
, optlen
);
3802 case SCTP_ADAPTATION_LAYER
:
3803 retval
= sctp_setsockopt_adaptation_layer(sk
, optval
, optlen
);
3806 retval
= sctp_setsockopt_context(sk
, optval
, optlen
);
3808 case SCTP_FRAGMENT_INTERLEAVE
:
3809 retval
= sctp_setsockopt_fragment_interleave(sk
, optval
, optlen
);
3811 case SCTP_MAX_BURST
:
3812 retval
= sctp_setsockopt_maxburst(sk
, optval
, optlen
);
3814 case SCTP_AUTH_CHUNK
:
3815 retval
= sctp_setsockopt_auth_chunk(sk
, optval
, optlen
);
3817 case SCTP_HMAC_IDENT
:
3818 retval
= sctp_setsockopt_hmac_ident(sk
, optval
, optlen
);
3821 retval
= sctp_setsockopt_auth_key(sk
, optval
, optlen
);
3823 case SCTP_AUTH_ACTIVE_KEY
:
3824 retval
= sctp_setsockopt_active_key(sk
, optval
, optlen
);
3826 case SCTP_AUTH_DELETE_KEY
:
3827 retval
= sctp_setsockopt_del_key(sk
, optval
, optlen
);
3829 case SCTP_AUTO_ASCONF
:
3830 retval
= sctp_setsockopt_auto_asconf(sk
, optval
, optlen
);
3832 case SCTP_PEER_ADDR_THLDS
:
3833 retval
= sctp_setsockopt_paddr_thresholds(sk
, optval
, optlen
);
3835 case SCTP_RECVRCVINFO
:
3836 retval
= sctp_setsockopt_recvrcvinfo(sk
, optval
, optlen
);
3838 case SCTP_RECVNXTINFO
:
3839 retval
= sctp_setsockopt_recvnxtinfo(sk
, optval
, optlen
);
3842 retval
= -ENOPROTOOPT
;
3852 /* API 3.1.6 connect() - UDP Style Syntax
3854 * An application may use the connect() call in the UDP model to initiate an
3855 * association without sending data.
3859 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3861 * sd: the socket descriptor to have a new association added to.
3863 * nam: the address structure (either struct sockaddr_in or struct
3864 * sockaddr_in6 defined in RFC2553 [7]).
3866 * len: the size of the address.
3868 static int sctp_connect(struct sock
*sk
, struct sockaddr
*addr
,
3876 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__
, sk
,
3879 /* Validate addr_len before calling common connect/connectx routine. */
3880 af
= sctp_get_af_specific(addr
->sa_family
);
3881 if (!af
|| addr_len
< af
->sockaddr_len
) {
3884 /* Pass correct addr len to common routine (so it knows there
3885 * is only one address being passed.
3887 err
= __sctp_connect(sk
, addr
, af
->sockaddr_len
, NULL
);
3894 /* FIXME: Write comments. */
3895 static int sctp_disconnect(struct sock
*sk
, int flags
)
3897 return -EOPNOTSUPP
; /* STUB */
3900 /* 4.1.4 accept() - TCP Style Syntax
3902 * Applications use accept() call to remove an established SCTP
3903 * association from the accept queue of the endpoint. A new socket
3904 * descriptor will be returned from accept() to represent the newly
3905 * formed association.
3907 static struct sock
*sctp_accept(struct sock
*sk
, int flags
, int *err
)
3909 struct sctp_sock
*sp
;
3910 struct sctp_endpoint
*ep
;
3911 struct sock
*newsk
= NULL
;
3912 struct sctp_association
*asoc
;
3921 if (!sctp_style(sk
, TCP
)) {
3922 error
= -EOPNOTSUPP
;
3926 if (!sctp_sstate(sk
, LISTENING
)) {
3931 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
3933 error
= sctp_wait_for_accept(sk
, timeo
);
3937 /* We treat the list of associations on the endpoint as the accept
3938 * queue and pick the first association on the list.
3940 asoc
= list_entry(ep
->asocs
.next
, struct sctp_association
, asocs
);
3942 newsk
= sp
->pf
->create_accept_sk(sk
, asoc
);
3948 /* Populate the fields of the newsk from the oldsk and migrate the
3949 * asoc to the newsk.
3951 sctp_sock_migrate(sk
, newsk
, asoc
, SCTP_SOCKET_TCP
);
3959 /* The SCTP ioctl handler. */
3960 static int sctp_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
3967 * SEQPACKET-style sockets in LISTENING state are valid, for
3968 * SCTP, so only discard TCP-style sockets in LISTENING state.
3970 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
3975 struct sk_buff
*skb
;
3976 unsigned int amount
= 0;
3978 skb
= skb_peek(&sk
->sk_receive_queue
);
3981 * We will only return the amount of this packet since
3982 * that is all that will be read.
3986 rc
= put_user(amount
, (int __user
*)arg
);
3998 /* This is the function which gets called during socket creation to
3999 * initialized the SCTP-specific portion of the sock.
4000 * The sock structure should already be zero-filled memory.
4002 static int sctp_init_sock(struct sock
*sk
)
4004 struct net
*net
= sock_net(sk
);
4005 struct sctp_sock
*sp
;
4007 pr_debug("%s: sk:%p\n", __func__
, sk
);
4011 /* Initialize the SCTP per socket area. */
4012 switch (sk
->sk_type
) {
4013 case SOCK_SEQPACKET
:
4014 sp
->type
= SCTP_SOCKET_UDP
;
4017 sp
->type
= SCTP_SOCKET_TCP
;
4020 return -ESOCKTNOSUPPORT
;
4023 /* Initialize default send parameters. These parameters can be
4024 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4026 sp
->default_stream
= 0;
4027 sp
->default_ppid
= 0;
4028 sp
->default_flags
= 0;
4029 sp
->default_context
= 0;
4030 sp
->default_timetolive
= 0;
4032 sp
->default_rcv_context
= 0;
4033 sp
->max_burst
= net
->sctp
.max_burst
;
4035 sp
->sctp_hmac_alg
= net
->sctp
.sctp_hmac_alg
;
4037 /* Initialize default setup parameters. These parameters
4038 * can be modified with the SCTP_INITMSG socket option or
4039 * overridden by the SCTP_INIT CMSG.
4041 sp
->initmsg
.sinit_num_ostreams
= sctp_max_outstreams
;
4042 sp
->initmsg
.sinit_max_instreams
= sctp_max_instreams
;
4043 sp
->initmsg
.sinit_max_attempts
= net
->sctp
.max_retrans_init
;
4044 sp
->initmsg
.sinit_max_init_timeo
= net
->sctp
.rto_max
;
4046 /* Initialize default RTO related parameters. These parameters can
4047 * be modified for with the SCTP_RTOINFO socket option.
4049 sp
->rtoinfo
.srto_initial
= net
->sctp
.rto_initial
;
4050 sp
->rtoinfo
.srto_max
= net
->sctp
.rto_max
;
4051 sp
->rtoinfo
.srto_min
= net
->sctp
.rto_min
;
4053 /* Initialize default association related parameters. These parameters
4054 * can be modified with the SCTP_ASSOCINFO socket option.
4056 sp
->assocparams
.sasoc_asocmaxrxt
= net
->sctp
.max_retrans_association
;
4057 sp
->assocparams
.sasoc_number_peer_destinations
= 0;
4058 sp
->assocparams
.sasoc_peer_rwnd
= 0;
4059 sp
->assocparams
.sasoc_local_rwnd
= 0;
4060 sp
->assocparams
.sasoc_cookie_life
= net
->sctp
.valid_cookie_life
;
4062 /* Initialize default event subscriptions. By default, all the
4065 memset(&sp
->subscribe
, 0, sizeof(struct sctp_event_subscribe
));
4067 /* Default Peer Address Parameters. These defaults can
4068 * be modified via SCTP_PEER_ADDR_PARAMS
4070 sp
->hbinterval
= net
->sctp
.hb_interval
;
4071 sp
->pathmaxrxt
= net
->sctp
.max_retrans_path
;
4072 sp
->pathmtu
= 0; /* allow default discovery */
4073 sp
->sackdelay
= net
->sctp
.sack_timeout
;
4075 sp
->param_flags
= SPP_HB_ENABLE
|
4077 SPP_SACKDELAY_ENABLE
;
4079 /* If enabled no SCTP message fragmentation will be performed.
4080 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4082 sp
->disable_fragments
= 0;
4084 /* Enable Nagle algorithm by default. */
4087 sp
->recvrcvinfo
= 0;
4088 sp
->recvnxtinfo
= 0;
4090 /* Enable by default. */
4093 /* Auto-close idle associations after the configured
4094 * number of seconds. A value of 0 disables this
4095 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4096 * for UDP-style sockets only.
4100 /* User specified fragmentation limit. */
4103 sp
->adaptation_ind
= 0;
4105 sp
->pf
= sctp_get_pf_specific(sk
->sk_family
);
4107 /* Control variables for partial data delivery. */
4108 atomic_set(&sp
->pd_mode
, 0);
4109 skb_queue_head_init(&sp
->pd_lobby
);
4110 sp
->frag_interleave
= 0;
4112 /* Create a per socket endpoint structure. Even if we
4113 * change the data structure relationships, this may still
4114 * be useful for storing pre-connect address information.
4116 sp
->ep
= sctp_endpoint_new(sk
, GFP_KERNEL
);
4122 sk
->sk_destruct
= sctp_destruct_sock
;
4124 SCTP_DBG_OBJCNT_INC(sock
);
4127 percpu_counter_inc(&sctp_sockets_allocated
);
4128 sock_prot_inuse_add(net
, sk
->sk_prot
, 1);
4129 if (net
->sctp
.default_auto_asconf
) {
4130 list_add_tail(&sp
->auto_asconf_list
,
4131 &net
->sctp
.auto_asconf_splist
);
4132 sp
->do_auto_asconf
= 1;
4134 sp
->do_auto_asconf
= 0;
4140 /* Cleanup any SCTP per socket resources. */
4141 static void sctp_destroy_sock(struct sock
*sk
)
4143 struct sctp_sock
*sp
;
4145 pr_debug("%s: sk:%p\n", __func__
, sk
);
4147 /* Release our hold on the endpoint. */
4149 /* This could happen during socket init, thus we bail out
4150 * early, since the rest of the below is not setup either.
4155 if (sp
->do_auto_asconf
) {
4156 sp
->do_auto_asconf
= 0;
4157 list_del(&sp
->auto_asconf_list
);
4159 sctp_endpoint_free(sp
->ep
);
4161 percpu_counter_dec(&sctp_sockets_allocated
);
4162 sock_prot_inuse_add(sock_net(sk
), sk
->sk_prot
, -1);
4166 /* Triggered when there are no references on the socket anymore */
4167 static void sctp_destruct_sock(struct sock
*sk
)
4169 struct sctp_sock
*sp
= sctp_sk(sk
);
4171 /* Free up the HMAC transform. */
4172 crypto_free_hash(sp
->hmac
);
4174 inet_sock_destruct(sk
);
4177 /* API 4.1.7 shutdown() - TCP Style Syntax
4178 * int shutdown(int socket, int how);
4180 * sd - the socket descriptor of the association to be closed.
4181 * how - Specifies the type of shutdown. The values are
4184 * Disables further receive operations. No SCTP
4185 * protocol action is taken.
4187 * Disables further send operations, and initiates
4188 * the SCTP shutdown sequence.
4190 * Disables further send and receive operations
4191 * and initiates the SCTP shutdown sequence.
4193 static void sctp_shutdown(struct sock
*sk
, int how
)
4195 struct net
*net
= sock_net(sk
);
4196 struct sctp_endpoint
*ep
;
4197 struct sctp_association
*asoc
;
4199 if (!sctp_style(sk
, TCP
))
4202 if (how
& SEND_SHUTDOWN
) {
4203 ep
= sctp_sk(sk
)->ep
;
4204 if (!list_empty(&ep
->asocs
)) {
4205 asoc
= list_entry(ep
->asocs
.next
,
4206 struct sctp_association
, asocs
);
4207 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
4212 /* 7.2.1 Association Status (SCTP_STATUS)
4214 * Applications can retrieve current status information about an
4215 * association, including association state, peer receiver window size,
4216 * number of unacked data chunks, and number of data chunks pending
4217 * receipt. This information is read-only.
4219 static int sctp_getsockopt_sctp_status(struct sock
*sk
, int len
,
4220 char __user
*optval
,
4223 struct sctp_status status
;
4224 struct sctp_association
*asoc
= NULL
;
4225 struct sctp_transport
*transport
;
4226 sctp_assoc_t associd
;
4229 if (len
< sizeof(status
)) {
4234 len
= sizeof(status
);
4235 if (copy_from_user(&status
, optval
, len
)) {
4240 associd
= status
.sstat_assoc_id
;
4241 asoc
= sctp_id2assoc(sk
, associd
);
4247 transport
= asoc
->peer
.primary_path
;
4249 status
.sstat_assoc_id
= sctp_assoc2id(asoc
);
4250 status
.sstat_state
= sctp_assoc_to_state(asoc
);
4251 status
.sstat_rwnd
= asoc
->peer
.rwnd
;
4252 status
.sstat_unackdata
= asoc
->unack_data
;
4254 status
.sstat_penddata
= sctp_tsnmap_pending(&asoc
->peer
.tsn_map
);
4255 status
.sstat_instrms
= asoc
->c
.sinit_max_instreams
;
4256 status
.sstat_outstrms
= asoc
->c
.sinit_num_ostreams
;
4257 status
.sstat_fragmentation_point
= asoc
->frag_point
;
4258 status
.sstat_primary
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
4259 memcpy(&status
.sstat_primary
.spinfo_address
, &transport
->ipaddr
,
4260 transport
->af_specific
->sockaddr_len
);
4261 /* Map ipv4 address into v4-mapped-on-v6 address. */
4262 sctp_get_pf_specific(sk
->sk_family
)->addr_to_user(sctp_sk(sk
),
4263 (union sctp_addr
*)&status
.sstat_primary
.spinfo_address
);
4264 status
.sstat_primary
.spinfo_state
= transport
->state
;
4265 status
.sstat_primary
.spinfo_cwnd
= transport
->cwnd
;
4266 status
.sstat_primary
.spinfo_srtt
= transport
->srtt
;
4267 status
.sstat_primary
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
4268 status
.sstat_primary
.spinfo_mtu
= transport
->pathmtu
;
4270 if (status
.sstat_primary
.spinfo_state
== SCTP_UNKNOWN
)
4271 status
.sstat_primary
.spinfo_state
= SCTP_ACTIVE
;
4273 if (put_user(len
, optlen
)) {
4278 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4279 __func__
, len
, status
.sstat_state
, status
.sstat_rwnd
,
4280 status
.sstat_assoc_id
);
4282 if (copy_to_user(optval
, &status
, len
)) {
4292 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4294 * Applications can retrieve information about a specific peer address
4295 * of an association, including its reachability state, congestion
4296 * window, and retransmission timer values. This information is
4299 static int sctp_getsockopt_peer_addr_info(struct sock
*sk
, int len
,
4300 char __user
*optval
,
4303 struct sctp_paddrinfo pinfo
;
4304 struct sctp_transport
*transport
;
4307 if (len
< sizeof(pinfo
)) {
4312 len
= sizeof(pinfo
);
4313 if (copy_from_user(&pinfo
, optval
, len
)) {
4318 transport
= sctp_addr_id2transport(sk
, &pinfo
.spinfo_address
,
4319 pinfo
.spinfo_assoc_id
);
4323 pinfo
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
4324 pinfo
.spinfo_state
= transport
->state
;
4325 pinfo
.spinfo_cwnd
= transport
->cwnd
;
4326 pinfo
.spinfo_srtt
= transport
->srtt
;
4327 pinfo
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
4328 pinfo
.spinfo_mtu
= transport
->pathmtu
;
4330 if (pinfo
.spinfo_state
== SCTP_UNKNOWN
)
4331 pinfo
.spinfo_state
= SCTP_ACTIVE
;
4333 if (put_user(len
, optlen
)) {
4338 if (copy_to_user(optval
, &pinfo
, len
)) {
4347 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4349 * This option is a on/off flag. If enabled no SCTP message
4350 * fragmentation will be performed. Instead if a message being sent
4351 * exceeds the current PMTU size, the message will NOT be sent and
4352 * instead a error will be indicated to the user.
4354 static int sctp_getsockopt_disable_fragments(struct sock
*sk
, int len
,
4355 char __user
*optval
, int __user
*optlen
)
4359 if (len
< sizeof(int))
4363 val
= (sctp_sk(sk
)->disable_fragments
== 1);
4364 if (put_user(len
, optlen
))
4366 if (copy_to_user(optval
, &val
, len
))
4371 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4373 * This socket option is used to specify various notifications and
4374 * ancillary data the user wishes to receive.
4376 static int sctp_getsockopt_events(struct sock
*sk
, int len
, char __user
*optval
,
4381 if (len
> sizeof(struct sctp_event_subscribe
))
4382 len
= sizeof(struct sctp_event_subscribe
);
4383 if (put_user(len
, optlen
))
4385 if (copy_to_user(optval
, &sctp_sk(sk
)->subscribe
, len
))
4390 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4392 * This socket option is applicable to the UDP-style socket only. When
4393 * set it will cause associations that are idle for more than the
4394 * specified number of seconds to automatically close. An association
4395 * being idle is defined an association that has NOT sent or received
4396 * user data. The special value of '0' indicates that no automatic
4397 * close of any associations should be performed. The option expects an
4398 * integer defining the number of seconds of idle time before an
4399 * association is closed.
4401 static int sctp_getsockopt_autoclose(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4403 /* Applicable to UDP-style socket only */
4404 if (sctp_style(sk
, TCP
))
4406 if (len
< sizeof(int))
4409 if (put_user(len
, optlen
))
4411 if (copy_to_user(optval
, &sctp_sk(sk
)->autoclose
, sizeof(int)))
4416 /* Helper routine to branch off an association to a new socket. */
4417 int sctp_do_peeloff(struct sock
*sk
, sctp_assoc_t id
, struct socket
**sockp
)
4419 struct sctp_association
*asoc
= sctp_id2assoc(sk
, id
);
4420 struct sctp_sock
*sp
= sctp_sk(sk
);
4421 struct socket
*sock
;
4427 /* An association cannot be branched off from an already peeled-off
4428 * socket, nor is this supported for tcp style sockets.
4430 if (!sctp_style(sk
, UDP
))
4433 /* Create a new socket. */
4434 err
= sock_create(sk
->sk_family
, SOCK_SEQPACKET
, IPPROTO_SCTP
, &sock
);
4438 sctp_copy_sock(sock
->sk
, sk
, asoc
);
4440 /* Make peeled-off sockets more like 1-1 accepted sockets.
4441 * Set the daddr and initialize id to something more random
4443 sp
->pf
->to_sk_daddr(&asoc
->peer
.primary_addr
, sk
);
4445 /* Populate the fields of the newsk from the oldsk and migrate the
4446 * asoc to the newsk.
4448 sctp_sock_migrate(sk
, sock
->sk
, asoc
, SCTP_SOCKET_UDP_HIGH_BANDWIDTH
);
4454 EXPORT_SYMBOL(sctp_do_peeloff
);
4456 static int sctp_getsockopt_peeloff(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4458 sctp_peeloff_arg_t peeloff
;
4459 struct socket
*newsock
;
4460 struct file
*newfile
;
4463 if (len
< sizeof(sctp_peeloff_arg_t
))
4465 len
= sizeof(sctp_peeloff_arg_t
);
4466 if (copy_from_user(&peeloff
, optval
, len
))
4469 retval
= sctp_do_peeloff(sk
, peeloff
.associd
, &newsock
);
4473 /* Map the socket to an unused fd that can be returned to the user. */
4474 retval
= get_unused_fd_flags(0);
4476 sock_release(newsock
);
4480 newfile
= sock_alloc_file(newsock
, 0, NULL
);
4481 if (unlikely(IS_ERR(newfile
))) {
4482 put_unused_fd(retval
);
4483 sock_release(newsock
);
4484 return PTR_ERR(newfile
);
4487 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__
, sk
, newsock
->sk
,
4490 /* Return the fd mapped to the new socket. */
4491 if (put_user(len
, optlen
)) {
4493 put_unused_fd(retval
);
4496 peeloff
.sd
= retval
;
4497 if (copy_to_user(optval
, &peeloff
, len
)) {
4499 put_unused_fd(retval
);
4502 fd_install(retval
, newfile
);
4507 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4509 * Applications can enable or disable heartbeats for any peer address of
4510 * an association, modify an address's heartbeat interval, force a
4511 * heartbeat to be sent immediately, and adjust the address's maximum
4512 * number of retransmissions sent before an address is considered
4513 * unreachable. The following structure is used to access and modify an
4514 * address's parameters:
4516 * struct sctp_paddrparams {
4517 * sctp_assoc_t spp_assoc_id;
4518 * struct sockaddr_storage spp_address;
4519 * uint32_t spp_hbinterval;
4520 * uint16_t spp_pathmaxrxt;
4521 * uint32_t spp_pathmtu;
4522 * uint32_t spp_sackdelay;
4523 * uint32_t spp_flags;
4526 * spp_assoc_id - (one-to-many style socket) This is filled in the
4527 * application, and identifies the association for
4529 * spp_address - This specifies which address is of interest.
4530 * spp_hbinterval - This contains the value of the heartbeat interval,
4531 * in milliseconds. If a value of zero
4532 * is present in this field then no changes are to
4533 * be made to this parameter.
4534 * spp_pathmaxrxt - This contains the maximum number of
4535 * retransmissions before this address shall be
4536 * considered unreachable. If a value of zero
4537 * is present in this field then no changes are to
4538 * be made to this parameter.
4539 * spp_pathmtu - When Path MTU discovery is disabled the value
4540 * specified here will be the "fixed" path mtu.
4541 * Note that if the spp_address field is empty
4542 * then all associations on this address will
4543 * have this fixed path mtu set upon them.
4545 * spp_sackdelay - When delayed sack is enabled, this value specifies
4546 * the number of milliseconds that sacks will be delayed
4547 * for. This value will apply to all addresses of an
4548 * association if the spp_address field is empty. Note
4549 * also, that if delayed sack is enabled and this
4550 * value is set to 0, no change is made to the last
4551 * recorded delayed sack timer value.
4553 * spp_flags - These flags are used to control various features
4554 * on an association. The flag field may contain
4555 * zero or more of the following options.
4557 * SPP_HB_ENABLE - Enable heartbeats on the
4558 * specified address. Note that if the address
4559 * field is empty all addresses for the association
4560 * have heartbeats enabled upon them.
4562 * SPP_HB_DISABLE - Disable heartbeats on the
4563 * speicifed address. Note that if the address
4564 * field is empty all addresses for the association
4565 * will have their heartbeats disabled. Note also
4566 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4567 * mutually exclusive, only one of these two should
4568 * be specified. Enabling both fields will have
4569 * undetermined results.
4571 * SPP_HB_DEMAND - Request a user initiated heartbeat
4572 * to be made immediately.
4574 * SPP_PMTUD_ENABLE - This field will enable PMTU
4575 * discovery upon the specified address. Note that
4576 * if the address feild is empty then all addresses
4577 * on the association are effected.
4579 * SPP_PMTUD_DISABLE - This field will disable PMTU
4580 * discovery upon the specified address. Note that
4581 * if the address feild is empty then all addresses
4582 * on the association are effected. Not also that
4583 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4584 * exclusive. Enabling both will have undetermined
4587 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4588 * on delayed sack. The time specified in spp_sackdelay
4589 * is used to specify the sack delay for this address. Note
4590 * that if spp_address is empty then all addresses will
4591 * enable delayed sack and take on the sack delay
4592 * value specified in spp_sackdelay.
4593 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4594 * off delayed sack. If the spp_address field is blank then
4595 * delayed sack is disabled for the entire association. Note
4596 * also that this field is mutually exclusive to
4597 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4600 static int sctp_getsockopt_peer_addr_params(struct sock
*sk
, int len
,
4601 char __user
*optval
, int __user
*optlen
)
4603 struct sctp_paddrparams params
;
4604 struct sctp_transport
*trans
= NULL
;
4605 struct sctp_association
*asoc
= NULL
;
4606 struct sctp_sock
*sp
= sctp_sk(sk
);
4608 if (len
< sizeof(struct sctp_paddrparams
))
4610 len
= sizeof(struct sctp_paddrparams
);
4611 if (copy_from_user(¶ms
, optval
, len
))
4614 /* If an address other than INADDR_ANY is specified, and
4615 * no transport is found, then the request is invalid.
4617 if (!sctp_is_any(sk
, (union sctp_addr
*)¶ms
.spp_address
)) {
4618 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
4619 params
.spp_assoc_id
);
4621 pr_debug("%s: failed no transport\n", __func__
);
4626 /* Get association, if assoc_id != 0 and the socket is a one
4627 * to many style socket, and an association was not found, then
4628 * the id was invalid.
4630 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
4631 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
)) {
4632 pr_debug("%s: failed no association\n", __func__
);
4637 /* Fetch transport values. */
4638 params
.spp_hbinterval
= jiffies_to_msecs(trans
->hbinterval
);
4639 params
.spp_pathmtu
= trans
->pathmtu
;
4640 params
.spp_pathmaxrxt
= trans
->pathmaxrxt
;
4641 params
.spp_sackdelay
= jiffies_to_msecs(trans
->sackdelay
);
4643 /*draft-11 doesn't say what to return in spp_flags*/
4644 params
.spp_flags
= trans
->param_flags
;
4646 /* Fetch association values. */
4647 params
.spp_hbinterval
= jiffies_to_msecs(asoc
->hbinterval
);
4648 params
.spp_pathmtu
= asoc
->pathmtu
;
4649 params
.spp_pathmaxrxt
= asoc
->pathmaxrxt
;
4650 params
.spp_sackdelay
= jiffies_to_msecs(asoc
->sackdelay
);
4652 /*draft-11 doesn't say what to return in spp_flags*/
4653 params
.spp_flags
= asoc
->param_flags
;
4655 /* Fetch socket values. */
4656 params
.spp_hbinterval
= sp
->hbinterval
;
4657 params
.spp_pathmtu
= sp
->pathmtu
;
4658 params
.spp_sackdelay
= sp
->sackdelay
;
4659 params
.spp_pathmaxrxt
= sp
->pathmaxrxt
;
4661 /*draft-11 doesn't say what to return in spp_flags*/
4662 params
.spp_flags
= sp
->param_flags
;
4665 if (copy_to_user(optval
, ¶ms
, len
))
4668 if (put_user(len
, optlen
))
4675 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4677 * This option will effect the way delayed acks are performed. This
4678 * option allows you to get or set the delayed ack time, in
4679 * milliseconds. It also allows changing the delayed ack frequency.
4680 * Changing the frequency to 1 disables the delayed sack algorithm. If
4681 * the assoc_id is 0, then this sets or gets the endpoints default
4682 * values. If the assoc_id field is non-zero, then the set or get
4683 * effects the specified association for the one to many model (the
4684 * assoc_id field is ignored by the one to one model). Note that if
4685 * sack_delay or sack_freq are 0 when setting this option, then the
4686 * current values will remain unchanged.
4688 * struct sctp_sack_info {
4689 * sctp_assoc_t sack_assoc_id;
4690 * uint32_t sack_delay;
4691 * uint32_t sack_freq;
4694 * sack_assoc_id - This parameter, indicates which association the user
4695 * is performing an action upon. Note that if this field's value is
4696 * zero then the endpoints default value is changed (effecting future
4697 * associations only).
4699 * sack_delay - This parameter contains the number of milliseconds that
4700 * the user is requesting the delayed ACK timer be set to. Note that
4701 * this value is defined in the standard to be between 200 and 500
4704 * sack_freq - This parameter contains the number of packets that must
4705 * be received before a sack is sent without waiting for the delay
4706 * timer to expire. The default value for this is 2, setting this
4707 * value to 1 will disable the delayed sack algorithm.
4709 static int sctp_getsockopt_delayed_ack(struct sock
*sk
, int len
,
4710 char __user
*optval
,
4713 struct sctp_sack_info params
;
4714 struct sctp_association
*asoc
= NULL
;
4715 struct sctp_sock
*sp
= sctp_sk(sk
);
4717 if (len
>= sizeof(struct sctp_sack_info
)) {
4718 len
= sizeof(struct sctp_sack_info
);
4720 if (copy_from_user(¶ms
, optval
, len
))
4722 } else if (len
== sizeof(struct sctp_assoc_value
)) {
4723 pr_warn_ratelimited(DEPRECATED
4725 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
4726 "Use struct sctp_sack_info instead\n",
4727 current
->comm
, task_pid_nr(current
));
4728 if (copy_from_user(¶ms
, optval
, len
))
4733 /* Get association, if sack_assoc_id != 0 and the socket is a one
4734 * to many style socket, and an association was not found, then
4735 * the id was invalid.
4737 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
4738 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
4742 /* Fetch association values. */
4743 if (asoc
->param_flags
& SPP_SACKDELAY_ENABLE
) {
4744 params
.sack_delay
= jiffies_to_msecs(
4746 params
.sack_freq
= asoc
->sackfreq
;
4749 params
.sack_delay
= 0;
4750 params
.sack_freq
= 1;
4753 /* Fetch socket values. */
4754 if (sp
->param_flags
& SPP_SACKDELAY_ENABLE
) {
4755 params
.sack_delay
= sp
->sackdelay
;
4756 params
.sack_freq
= sp
->sackfreq
;
4758 params
.sack_delay
= 0;
4759 params
.sack_freq
= 1;
4763 if (copy_to_user(optval
, ¶ms
, len
))
4766 if (put_user(len
, optlen
))
4772 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4774 * Applications can specify protocol parameters for the default association
4775 * initialization. The option name argument to setsockopt() and getsockopt()
4778 * Setting initialization parameters is effective only on an unconnected
4779 * socket (for UDP-style sockets only future associations are effected
4780 * by the change). With TCP-style sockets, this option is inherited by
4781 * sockets derived from a listener socket.
4783 static int sctp_getsockopt_initmsg(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4785 if (len
< sizeof(struct sctp_initmsg
))
4787 len
= sizeof(struct sctp_initmsg
);
4788 if (put_user(len
, optlen
))
4790 if (copy_to_user(optval
, &sctp_sk(sk
)->initmsg
, len
))
4796 static int sctp_getsockopt_peer_addrs(struct sock
*sk
, int len
,
4797 char __user
*optval
, int __user
*optlen
)
4799 struct sctp_association
*asoc
;
4801 struct sctp_getaddrs getaddrs
;
4802 struct sctp_transport
*from
;
4804 union sctp_addr temp
;
4805 struct sctp_sock
*sp
= sctp_sk(sk
);
4810 if (len
< sizeof(struct sctp_getaddrs
))
4813 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
4816 /* For UDP-style sockets, id specifies the association to query. */
4817 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
4821 to
= optval
+ offsetof(struct sctp_getaddrs
, addrs
);
4822 space_left
= len
- offsetof(struct sctp_getaddrs
, addrs
);
4824 list_for_each_entry(from
, &asoc
->peer
.transport_addr_list
,
4826 memcpy(&temp
, &from
->ipaddr
, sizeof(temp
));
4827 addrlen
= sctp_get_pf_specific(sk
->sk_family
)
4828 ->addr_to_user(sp
, &temp
);
4829 if (space_left
< addrlen
)
4831 if (copy_to_user(to
, &temp
, addrlen
))
4835 space_left
-= addrlen
;
4838 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
))
4840 bytes_copied
= ((char __user
*)to
) - optval
;
4841 if (put_user(bytes_copied
, optlen
))
4847 static int sctp_copy_laddrs(struct sock
*sk
, __u16 port
, void *to
,
4848 size_t space_left
, int *bytes_copied
)
4850 struct sctp_sockaddr_entry
*addr
;
4851 union sctp_addr temp
;
4854 struct net
*net
= sock_net(sk
);
4857 list_for_each_entry_rcu(addr
, &net
->sctp
.local_addr_list
, list
) {
4861 if ((PF_INET
== sk
->sk_family
) &&
4862 (AF_INET6
== addr
->a
.sa
.sa_family
))
4864 if ((PF_INET6
== sk
->sk_family
) &&
4865 inet_v6_ipv6only(sk
) &&
4866 (AF_INET
== addr
->a
.sa
.sa_family
))
4868 memcpy(&temp
, &addr
->a
, sizeof(temp
));
4869 if (!temp
.v4
.sin_port
)
4870 temp
.v4
.sin_port
= htons(port
);
4872 addrlen
= sctp_get_pf_specific(sk
->sk_family
)
4873 ->addr_to_user(sctp_sk(sk
), &temp
);
4875 if (space_left
< addrlen
) {
4879 memcpy(to
, &temp
, addrlen
);
4883 space_left
-= addrlen
;
4884 *bytes_copied
+= addrlen
;
4892 static int sctp_getsockopt_local_addrs(struct sock
*sk
, int len
,
4893 char __user
*optval
, int __user
*optlen
)
4895 struct sctp_bind_addr
*bp
;
4896 struct sctp_association
*asoc
;
4898 struct sctp_getaddrs getaddrs
;
4899 struct sctp_sockaddr_entry
*addr
;
4901 union sctp_addr temp
;
4902 struct sctp_sock
*sp
= sctp_sk(sk
);
4906 int bytes_copied
= 0;
4910 if (len
< sizeof(struct sctp_getaddrs
))
4913 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
4917 * For UDP-style sockets, id specifies the association to query.
4918 * If the id field is set to the value '0' then the locally bound
4919 * addresses are returned without regard to any particular
4922 if (0 == getaddrs
.assoc_id
) {
4923 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
4925 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
4928 bp
= &asoc
->base
.bind_addr
;
4931 to
= optval
+ offsetof(struct sctp_getaddrs
, addrs
);
4932 space_left
= len
- offsetof(struct sctp_getaddrs
, addrs
);
4934 addrs
= kmalloc(space_left
, GFP_KERNEL
);
4938 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4939 * addresses from the global local address list.
4941 if (sctp_list_single_entry(&bp
->address_list
)) {
4942 addr
= list_entry(bp
->address_list
.next
,
4943 struct sctp_sockaddr_entry
, list
);
4944 if (sctp_is_any(sk
, &addr
->a
)) {
4945 cnt
= sctp_copy_laddrs(sk
, bp
->port
, addrs
,
4946 space_left
, &bytes_copied
);
4956 /* Protection on the bound address list is not needed since
4957 * in the socket option context we hold a socket lock and
4958 * thus the bound address list can't change.
4960 list_for_each_entry(addr
, &bp
->address_list
, list
) {
4961 memcpy(&temp
, &addr
->a
, sizeof(temp
));
4962 addrlen
= sctp_get_pf_specific(sk
->sk_family
)
4963 ->addr_to_user(sp
, &temp
);
4964 if (space_left
< addrlen
) {
4965 err
= -ENOMEM
; /*fixme: right error?*/
4968 memcpy(buf
, &temp
, addrlen
);
4970 bytes_copied
+= addrlen
;
4972 space_left
-= addrlen
;
4976 if (copy_to_user(to
, addrs
, bytes_copied
)) {
4980 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
)) {
4984 if (put_user(bytes_copied
, optlen
))
4991 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4993 * Requests that the local SCTP stack use the enclosed peer address as
4994 * the association primary. The enclosed address must be one of the
4995 * association peer's addresses.
4997 static int sctp_getsockopt_primary_addr(struct sock
*sk
, int len
,
4998 char __user
*optval
, int __user
*optlen
)
5000 struct sctp_prim prim
;
5001 struct sctp_association
*asoc
;
5002 struct sctp_sock
*sp
= sctp_sk(sk
);
5004 if (len
< sizeof(struct sctp_prim
))
5007 len
= sizeof(struct sctp_prim
);
5009 if (copy_from_user(&prim
, optval
, len
))
5012 asoc
= sctp_id2assoc(sk
, prim
.ssp_assoc_id
);
5016 if (!asoc
->peer
.primary_path
)
5019 memcpy(&prim
.ssp_addr
, &asoc
->peer
.primary_path
->ipaddr
,
5020 asoc
->peer
.primary_path
->af_specific
->sockaddr_len
);
5022 sctp_get_pf_specific(sk
->sk_family
)->addr_to_user(sp
,
5023 (union sctp_addr
*)&prim
.ssp_addr
);
5025 if (put_user(len
, optlen
))
5027 if (copy_to_user(optval
, &prim
, len
))
5034 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5036 * Requests that the local endpoint set the specified Adaptation Layer
5037 * Indication parameter for all future INIT and INIT-ACK exchanges.
5039 static int sctp_getsockopt_adaptation_layer(struct sock
*sk
, int len
,
5040 char __user
*optval
, int __user
*optlen
)
5042 struct sctp_setadaptation adaptation
;
5044 if (len
< sizeof(struct sctp_setadaptation
))
5047 len
= sizeof(struct sctp_setadaptation
);
5049 adaptation
.ssb_adaptation_ind
= sctp_sk(sk
)->adaptation_ind
;
5051 if (put_user(len
, optlen
))
5053 if (copy_to_user(optval
, &adaptation
, len
))
5061 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5063 * Applications that wish to use the sendto() system call may wish to
5064 * specify a default set of parameters that would normally be supplied
5065 * through the inclusion of ancillary data. This socket option allows
5066 * such an application to set the default sctp_sndrcvinfo structure.
5069 * The application that wishes to use this socket option simply passes
5070 * in to this call the sctp_sndrcvinfo structure defined in Section
5071 * 5.2.2) The input parameters accepted by this call include
5072 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5073 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5074 * to this call if the caller is using the UDP model.
5076 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5078 static int sctp_getsockopt_default_send_param(struct sock
*sk
,
5079 int len
, char __user
*optval
,
5082 struct sctp_sock
*sp
= sctp_sk(sk
);
5083 struct sctp_association
*asoc
;
5084 struct sctp_sndrcvinfo info
;
5086 if (len
< sizeof(info
))
5091 if (copy_from_user(&info
, optval
, len
))
5094 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
5095 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
5098 info
.sinfo_stream
= asoc
->default_stream
;
5099 info
.sinfo_flags
= asoc
->default_flags
;
5100 info
.sinfo_ppid
= asoc
->default_ppid
;
5101 info
.sinfo_context
= asoc
->default_context
;
5102 info
.sinfo_timetolive
= asoc
->default_timetolive
;
5104 info
.sinfo_stream
= sp
->default_stream
;
5105 info
.sinfo_flags
= sp
->default_flags
;
5106 info
.sinfo_ppid
= sp
->default_ppid
;
5107 info
.sinfo_context
= sp
->default_context
;
5108 info
.sinfo_timetolive
= sp
->default_timetolive
;
5111 if (put_user(len
, optlen
))
5113 if (copy_to_user(optval
, &info
, len
))
5119 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5120 * (SCTP_DEFAULT_SNDINFO)
5122 static int sctp_getsockopt_default_sndinfo(struct sock
*sk
, int len
,
5123 char __user
*optval
,
5126 struct sctp_sock
*sp
= sctp_sk(sk
);
5127 struct sctp_association
*asoc
;
5128 struct sctp_sndinfo info
;
5130 if (len
< sizeof(info
))
5135 if (copy_from_user(&info
, optval
, len
))
5138 asoc
= sctp_id2assoc(sk
, info
.snd_assoc_id
);
5139 if (!asoc
&& info
.snd_assoc_id
&& sctp_style(sk
, UDP
))
5142 info
.snd_sid
= asoc
->default_stream
;
5143 info
.snd_flags
= asoc
->default_flags
;
5144 info
.snd_ppid
= asoc
->default_ppid
;
5145 info
.snd_context
= asoc
->default_context
;
5147 info
.snd_sid
= sp
->default_stream
;
5148 info
.snd_flags
= sp
->default_flags
;
5149 info
.snd_ppid
= sp
->default_ppid
;
5150 info
.snd_context
= sp
->default_context
;
5153 if (put_user(len
, optlen
))
5155 if (copy_to_user(optval
, &info
, len
))
5163 * 7.1.5 SCTP_NODELAY
5165 * Turn on/off any Nagle-like algorithm. This means that packets are
5166 * generally sent as soon as possible and no unnecessary delays are
5167 * introduced, at the cost of more packets in the network. Expects an
5168 * integer boolean flag.
5171 static int sctp_getsockopt_nodelay(struct sock
*sk
, int len
,
5172 char __user
*optval
, int __user
*optlen
)
5176 if (len
< sizeof(int))
5180 val
= (sctp_sk(sk
)->nodelay
== 1);
5181 if (put_user(len
, optlen
))
5183 if (copy_to_user(optval
, &val
, len
))
5190 * 7.1.1 SCTP_RTOINFO
5192 * The protocol parameters used to initialize and bound retransmission
5193 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5194 * and modify these parameters.
5195 * All parameters are time values, in milliseconds. A value of 0, when
5196 * modifying the parameters, indicates that the current value should not
5200 static int sctp_getsockopt_rtoinfo(struct sock
*sk
, int len
,
5201 char __user
*optval
,
5202 int __user
*optlen
) {
5203 struct sctp_rtoinfo rtoinfo
;
5204 struct sctp_association
*asoc
;
5206 if (len
< sizeof (struct sctp_rtoinfo
))
5209 len
= sizeof(struct sctp_rtoinfo
);
5211 if (copy_from_user(&rtoinfo
, optval
, len
))
5214 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
5216 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
5219 /* Values corresponding to the specific association. */
5221 rtoinfo
.srto_initial
= jiffies_to_msecs(asoc
->rto_initial
);
5222 rtoinfo
.srto_max
= jiffies_to_msecs(asoc
->rto_max
);
5223 rtoinfo
.srto_min
= jiffies_to_msecs(asoc
->rto_min
);
5225 /* Values corresponding to the endpoint. */
5226 struct sctp_sock
*sp
= sctp_sk(sk
);
5228 rtoinfo
.srto_initial
= sp
->rtoinfo
.srto_initial
;
5229 rtoinfo
.srto_max
= sp
->rtoinfo
.srto_max
;
5230 rtoinfo
.srto_min
= sp
->rtoinfo
.srto_min
;
5233 if (put_user(len
, optlen
))
5236 if (copy_to_user(optval
, &rtoinfo
, len
))
5244 * 7.1.2 SCTP_ASSOCINFO
5246 * This option is used to tune the maximum retransmission attempts
5247 * of the association.
5248 * Returns an error if the new association retransmission value is
5249 * greater than the sum of the retransmission value of the peer.
5250 * See [SCTP] for more information.
5253 static int sctp_getsockopt_associnfo(struct sock
*sk
, int len
,
5254 char __user
*optval
,
5258 struct sctp_assocparams assocparams
;
5259 struct sctp_association
*asoc
;
5260 struct list_head
*pos
;
5263 if (len
< sizeof (struct sctp_assocparams
))
5266 len
= sizeof(struct sctp_assocparams
);
5268 if (copy_from_user(&assocparams
, optval
, len
))
5271 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
5273 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
5276 /* Values correspoinding to the specific association */
5278 assocparams
.sasoc_asocmaxrxt
= asoc
->max_retrans
;
5279 assocparams
.sasoc_peer_rwnd
= asoc
->peer
.rwnd
;
5280 assocparams
.sasoc_local_rwnd
= asoc
->a_rwnd
;
5281 assocparams
.sasoc_cookie_life
= ktime_to_ms(asoc
->cookie_life
);
5283 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
5287 assocparams
.sasoc_number_peer_destinations
= cnt
;
5289 /* Values corresponding to the endpoint */
5290 struct sctp_sock
*sp
= sctp_sk(sk
);
5292 assocparams
.sasoc_asocmaxrxt
= sp
->assocparams
.sasoc_asocmaxrxt
;
5293 assocparams
.sasoc_peer_rwnd
= sp
->assocparams
.sasoc_peer_rwnd
;
5294 assocparams
.sasoc_local_rwnd
= sp
->assocparams
.sasoc_local_rwnd
;
5295 assocparams
.sasoc_cookie_life
=
5296 sp
->assocparams
.sasoc_cookie_life
;
5297 assocparams
.sasoc_number_peer_destinations
=
5299 sasoc_number_peer_destinations
;
5302 if (put_user(len
, optlen
))
5305 if (copy_to_user(optval
, &assocparams
, len
))
5312 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5314 * This socket option is a boolean flag which turns on or off mapped V4
5315 * addresses. If this option is turned on and the socket is type
5316 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5317 * If this option is turned off, then no mapping will be done of V4
5318 * addresses and a user will receive both PF_INET6 and PF_INET type
5319 * addresses on the socket.
5321 static int sctp_getsockopt_mappedv4(struct sock
*sk
, int len
,
5322 char __user
*optval
, int __user
*optlen
)
5325 struct sctp_sock
*sp
= sctp_sk(sk
);
5327 if (len
< sizeof(int))
5332 if (put_user(len
, optlen
))
5334 if (copy_to_user(optval
, &val
, len
))
5341 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5342 * (chapter and verse is quoted at sctp_setsockopt_context())
5344 static int sctp_getsockopt_context(struct sock
*sk
, int len
,
5345 char __user
*optval
, int __user
*optlen
)
5347 struct sctp_assoc_value params
;
5348 struct sctp_sock
*sp
;
5349 struct sctp_association
*asoc
;
5351 if (len
< sizeof(struct sctp_assoc_value
))
5354 len
= sizeof(struct sctp_assoc_value
);
5356 if (copy_from_user(¶ms
, optval
, len
))
5361 if (params
.assoc_id
!= 0) {
5362 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5365 params
.assoc_value
= asoc
->default_rcv_context
;
5367 params
.assoc_value
= sp
->default_rcv_context
;
5370 if (put_user(len
, optlen
))
5372 if (copy_to_user(optval
, ¶ms
, len
))
5379 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5380 * This option will get or set the maximum size to put in any outgoing
5381 * SCTP DATA chunk. If a message is larger than this size it will be
5382 * fragmented by SCTP into the specified size. Note that the underlying
5383 * SCTP implementation may fragment into smaller sized chunks when the
5384 * PMTU of the underlying association is smaller than the value set by
5385 * the user. The default value for this option is '0' which indicates
5386 * the user is NOT limiting fragmentation and only the PMTU will effect
5387 * SCTP's choice of DATA chunk size. Note also that values set larger
5388 * than the maximum size of an IP datagram will effectively let SCTP
5389 * control fragmentation (i.e. the same as setting this option to 0).
5391 * The following structure is used to access and modify this parameter:
5393 * struct sctp_assoc_value {
5394 * sctp_assoc_t assoc_id;
5395 * uint32_t assoc_value;
5398 * assoc_id: This parameter is ignored for one-to-one style sockets.
5399 * For one-to-many style sockets this parameter indicates which
5400 * association the user is performing an action upon. Note that if
5401 * this field's value is zero then the endpoints default value is
5402 * changed (effecting future associations only).
5403 * assoc_value: This parameter specifies the maximum size in bytes.
5405 static int sctp_getsockopt_maxseg(struct sock
*sk
, int len
,
5406 char __user
*optval
, int __user
*optlen
)
5408 struct sctp_assoc_value params
;
5409 struct sctp_association
*asoc
;
5411 if (len
== sizeof(int)) {
5412 pr_warn_ratelimited(DEPRECATED
5414 "Use of int in maxseg socket option.\n"
5415 "Use struct sctp_assoc_value instead\n",
5416 current
->comm
, task_pid_nr(current
));
5417 params
.assoc_id
= 0;
5418 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5419 len
= sizeof(struct sctp_assoc_value
);
5420 if (copy_from_user(¶ms
, optval
, sizeof(params
)))
5425 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5426 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
5430 params
.assoc_value
= asoc
->frag_point
;
5432 params
.assoc_value
= sctp_sk(sk
)->user_frag
;
5434 if (put_user(len
, optlen
))
5436 if (len
== sizeof(int)) {
5437 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5440 if (copy_to_user(optval
, ¶ms
, len
))
5448 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5449 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5451 static int sctp_getsockopt_fragment_interleave(struct sock
*sk
, int len
,
5452 char __user
*optval
, int __user
*optlen
)
5456 if (len
< sizeof(int))
5461 val
= sctp_sk(sk
)->frag_interleave
;
5462 if (put_user(len
, optlen
))
5464 if (copy_to_user(optval
, &val
, len
))
5471 * 7.1.25. Set or Get the sctp partial delivery point
5472 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5474 static int sctp_getsockopt_partial_delivery_point(struct sock
*sk
, int len
,
5475 char __user
*optval
,
5480 if (len
< sizeof(u32
))
5485 val
= sctp_sk(sk
)->pd_point
;
5486 if (put_user(len
, optlen
))
5488 if (copy_to_user(optval
, &val
, len
))
5495 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5496 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5498 static int sctp_getsockopt_maxburst(struct sock
*sk
, int len
,
5499 char __user
*optval
,
5502 struct sctp_assoc_value params
;
5503 struct sctp_sock
*sp
;
5504 struct sctp_association
*asoc
;
5506 if (len
== sizeof(int)) {
5507 pr_warn_ratelimited(DEPRECATED
5509 "Use of int in max_burst socket option.\n"
5510 "Use struct sctp_assoc_value instead\n",
5511 current
->comm
, task_pid_nr(current
));
5512 params
.assoc_id
= 0;
5513 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5514 len
= sizeof(struct sctp_assoc_value
);
5515 if (copy_from_user(¶ms
, optval
, len
))
5522 if (params
.assoc_id
!= 0) {
5523 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5526 params
.assoc_value
= asoc
->max_burst
;
5528 params
.assoc_value
= sp
->max_burst
;
5530 if (len
== sizeof(int)) {
5531 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5534 if (copy_to_user(optval
, ¶ms
, len
))
5542 static int sctp_getsockopt_hmac_ident(struct sock
*sk
, int len
,
5543 char __user
*optval
, int __user
*optlen
)
5545 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5546 struct sctp_hmacalgo __user
*p
= (void __user
*)optval
;
5547 struct sctp_hmac_algo_param
*hmacs
;
5551 if (!ep
->auth_enable
)
5554 hmacs
= ep
->auth_hmacs_list
;
5555 data_len
= ntohs(hmacs
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5557 if (len
< sizeof(struct sctp_hmacalgo
) + data_len
)
5560 len
= sizeof(struct sctp_hmacalgo
) + data_len
;
5561 num_idents
= data_len
/ sizeof(u16
);
5563 if (put_user(len
, optlen
))
5565 if (put_user(num_idents
, &p
->shmac_num_idents
))
5567 if (copy_to_user(p
->shmac_idents
, hmacs
->hmac_ids
, data_len
))
5572 static int sctp_getsockopt_active_key(struct sock
*sk
, int len
,
5573 char __user
*optval
, int __user
*optlen
)
5575 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5576 struct sctp_authkeyid val
;
5577 struct sctp_association
*asoc
;
5579 if (!ep
->auth_enable
)
5582 if (len
< sizeof(struct sctp_authkeyid
))
5584 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authkeyid
)))
5587 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
5588 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
5592 val
.scact_keynumber
= asoc
->active_key_id
;
5594 val
.scact_keynumber
= ep
->active_key_id
;
5596 len
= sizeof(struct sctp_authkeyid
);
5597 if (put_user(len
, optlen
))
5599 if (copy_to_user(optval
, &val
, len
))
5605 static int sctp_getsockopt_peer_auth_chunks(struct sock
*sk
, int len
,
5606 char __user
*optval
, int __user
*optlen
)
5608 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5609 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5610 struct sctp_authchunks val
;
5611 struct sctp_association
*asoc
;
5612 struct sctp_chunks_param
*ch
;
5616 if (!ep
->auth_enable
)
5619 if (len
< sizeof(struct sctp_authchunks
))
5622 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5625 to
= p
->gauth_chunks
;
5626 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5630 ch
= asoc
->peer
.peer_chunks
;
5634 /* See if the user provided enough room for all the data */
5635 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5636 if (len
< num_chunks
)
5639 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
5642 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
5643 if (put_user(len
, optlen
))
5645 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
5650 static int sctp_getsockopt_local_auth_chunks(struct sock
*sk
, int len
,
5651 char __user
*optval
, int __user
*optlen
)
5653 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5654 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5655 struct sctp_authchunks val
;
5656 struct sctp_association
*asoc
;
5657 struct sctp_chunks_param
*ch
;
5661 if (!ep
->auth_enable
)
5664 if (len
< sizeof(struct sctp_authchunks
))
5667 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5670 to
= p
->gauth_chunks
;
5671 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5672 if (!asoc
&& val
.gauth_assoc_id
&& sctp_style(sk
, UDP
))
5676 ch
= (struct sctp_chunks_param
*)asoc
->c
.auth_chunks
;
5678 ch
= ep
->auth_chunk_list
;
5683 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5684 if (len
< sizeof(struct sctp_authchunks
) + num_chunks
)
5687 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
5690 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
5691 if (put_user(len
, optlen
))
5693 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
5700 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5701 * This option gets the current number of associations that are attached
5702 * to a one-to-many style socket. The option value is an uint32_t.
5704 static int sctp_getsockopt_assoc_number(struct sock
*sk
, int len
,
5705 char __user
*optval
, int __user
*optlen
)
5707 struct sctp_sock
*sp
= sctp_sk(sk
);
5708 struct sctp_association
*asoc
;
5711 if (sctp_style(sk
, TCP
))
5714 if (len
< sizeof(u32
))
5719 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5723 if (put_user(len
, optlen
))
5725 if (copy_to_user(optval
, &val
, len
))
5732 * 8.1.23 SCTP_AUTO_ASCONF
5733 * See the corresponding setsockopt entry as description
5735 static int sctp_getsockopt_auto_asconf(struct sock
*sk
, int len
,
5736 char __user
*optval
, int __user
*optlen
)
5740 if (len
< sizeof(int))
5744 if (sctp_sk(sk
)->do_auto_asconf
&& sctp_is_ep_boundall(sk
))
5746 if (put_user(len
, optlen
))
5748 if (copy_to_user(optval
, &val
, len
))
5754 * 8.2.6. Get the Current Identifiers of Associations
5755 * (SCTP_GET_ASSOC_ID_LIST)
5757 * This option gets the current list of SCTP association identifiers of
5758 * the SCTP associations handled by a one-to-many style socket.
5760 static int sctp_getsockopt_assoc_ids(struct sock
*sk
, int len
,
5761 char __user
*optval
, int __user
*optlen
)
5763 struct sctp_sock
*sp
= sctp_sk(sk
);
5764 struct sctp_association
*asoc
;
5765 struct sctp_assoc_ids
*ids
;
5768 if (sctp_style(sk
, TCP
))
5771 if (len
< sizeof(struct sctp_assoc_ids
))
5774 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5778 if (len
< sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
)
5781 len
= sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
;
5783 ids
= kmalloc(len
, GFP_KERNEL
);
5787 ids
->gaids_number_of_ids
= num
;
5789 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5790 ids
->gaids_assoc_id
[num
++] = asoc
->assoc_id
;
5793 if (put_user(len
, optlen
) || copy_to_user(optval
, ids
, len
)) {
5803 * SCTP_PEER_ADDR_THLDS
5805 * This option allows us to fetch the partially failed threshold for one or all
5806 * transports in an association. See Section 6.1 of:
5807 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5809 static int sctp_getsockopt_paddr_thresholds(struct sock
*sk
,
5810 char __user
*optval
,
5814 struct sctp_paddrthlds val
;
5815 struct sctp_transport
*trans
;
5816 struct sctp_association
*asoc
;
5818 if (len
< sizeof(struct sctp_paddrthlds
))
5820 len
= sizeof(struct sctp_paddrthlds
);
5821 if (copy_from_user(&val
, (struct sctp_paddrthlds __user
*)optval
, len
))
5824 if (sctp_is_any(sk
, (const union sctp_addr
*)&val
.spt_address
)) {
5825 asoc
= sctp_id2assoc(sk
, val
.spt_assoc_id
);
5829 val
.spt_pathpfthld
= asoc
->pf_retrans
;
5830 val
.spt_pathmaxrxt
= asoc
->pathmaxrxt
;
5832 trans
= sctp_addr_id2transport(sk
, &val
.spt_address
,
5837 val
.spt_pathmaxrxt
= trans
->pathmaxrxt
;
5838 val
.spt_pathpfthld
= trans
->pf_retrans
;
5841 if (put_user(len
, optlen
) || copy_to_user(optval
, &val
, len
))
5848 * SCTP_GET_ASSOC_STATS
5850 * This option retrieves local per endpoint statistics. It is modeled
5851 * after OpenSolaris' implementation
5853 static int sctp_getsockopt_assoc_stats(struct sock
*sk
, int len
,
5854 char __user
*optval
,
5857 struct sctp_assoc_stats sas
;
5858 struct sctp_association
*asoc
= NULL
;
5860 /* User must provide at least the assoc id */
5861 if (len
< sizeof(sctp_assoc_t
))
5864 /* Allow the struct to grow and fill in as much as possible */
5865 len
= min_t(size_t, len
, sizeof(sas
));
5867 if (copy_from_user(&sas
, optval
, len
))
5870 asoc
= sctp_id2assoc(sk
, sas
.sas_assoc_id
);
5874 sas
.sas_rtxchunks
= asoc
->stats
.rtxchunks
;
5875 sas
.sas_gapcnt
= asoc
->stats
.gapcnt
;
5876 sas
.sas_outofseqtsns
= asoc
->stats
.outofseqtsns
;
5877 sas
.sas_osacks
= asoc
->stats
.osacks
;
5878 sas
.sas_isacks
= asoc
->stats
.isacks
;
5879 sas
.sas_octrlchunks
= asoc
->stats
.octrlchunks
;
5880 sas
.sas_ictrlchunks
= asoc
->stats
.ictrlchunks
;
5881 sas
.sas_oodchunks
= asoc
->stats
.oodchunks
;
5882 sas
.sas_iodchunks
= asoc
->stats
.iodchunks
;
5883 sas
.sas_ouodchunks
= asoc
->stats
.ouodchunks
;
5884 sas
.sas_iuodchunks
= asoc
->stats
.iuodchunks
;
5885 sas
.sas_idupchunks
= asoc
->stats
.idupchunks
;
5886 sas
.sas_opackets
= asoc
->stats
.opackets
;
5887 sas
.sas_ipackets
= asoc
->stats
.ipackets
;
5889 /* New high max rto observed, will return 0 if not a single
5890 * RTO update took place. obs_rto_ipaddr will be bogus
5893 sas
.sas_maxrto
= asoc
->stats
.max_obs_rto
;
5894 memcpy(&sas
.sas_obs_rto_ipaddr
, &asoc
->stats
.obs_rto_ipaddr
,
5895 sizeof(struct sockaddr_storage
));
5897 /* Mark beginning of a new observation period */
5898 asoc
->stats
.max_obs_rto
= asoc
->rto_min
;
5900 if (put_user(len
, optlen
))
5903 pr_debug("%s: len:%d, assoc_id:%d\n", __func__
, len
, sas
.sas_assoc_id
);
5905 if (copy_to_user(optval
, &sas
, len
))
5911 static int sctp_getsockopt_recvrcvinfo(struct sock
*sk
, int len
,
5912 char __user
*optval
,
5917 if (len
< sizeof(int))
5921 if (sctp_sk(sk
)->recvrcvinfo
)
5923 if (put_user(len
, optlen
))
5925 if (copy_to_user(optval
, &val
, len
))
5931 static int sctp_getsockopt_recvnxtinfo(struct sock
*sk
, int len
,
5932 char __user
*optval
,
5937 if (len
< sizeof(int))
5941 if (sctp_sk(sk
)->recvnxtinfo
)
5943 if (put_user(len
, optlen
))
5945 if (copy_to_user(optval
, &val
, len
))
5951 static int sctp_getsockopt(struct sock
*sk
, int level
, int optname
,
5952 char __user
*optval
, int __user
*optlen
)
5957 pr_debug("%s: sk:%p, optname:%d\n", __func__
, sk
, optname
);
5959 /* I can hardly begin to describe how wrong this is. This is
5960 * so broken as to be worse than useless. The API draft
5961 * REALLY is NOT helpful here... I am not convinced that the
5962 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5963 * are at all well-founded.
5965 if (level
!= SOL_SCTP
) {
5966 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
5968 retval
= af
->getsockopt(sk
, level
, optname
, optval
, optlen
);
5972 if (get_user(len
, optlen
))
5979 retval
= sctp_getsockopt_sctp_status(sk
, len
, optval
, optlen
);
5981 case SCTP_DISABLE_FRAGMENTS
:
5982 retval
= sctp_getsockopt_disable_fragments(sk
, len
, optval
,
5986 retval
= sctp_getsockopt_events(sk
, len
, optval
, optlen
);
5988 case SCTP_AUTOCLOSE
:
5989 retval
= sctp_getsockopt_autoclose(sk
, len
, optval
, optlen
);
5991 case SCTP_SOCKOPT_PEELOFF
:
5992 retval
= sctp_getsockopt_peeloff(sk
, len
, optval
, optlen
);
5994 case SCTP_PEER_ADDR_PARAMS
:
5995 retval
= sctp_getsockopt_peer_addr_params(sk
, len
, optval
,
5998 case SCTP_DELAYED_SACK
:
5999 retval
= sctp_getsockopt_delayed_ack(sk
, len
, optval
,
6003 retval
= sctp_getsockopt_initmsg(sk
, len
, optval
, optlen
);
6005 case SCTP_GET_PEER_ADDRS
:
6006 retval
= sctp_getsockopt_peer_addrs(sk
, len
, optval
,
6009 case SCTP_GET_LOCAL_ADDRS
:
6010 retval
= sctp_getsockopt_local_addrs(sk
, len
, optval
,
6013 case SCTP_SOCKOPT_CONNECTX3
:
6014 retval
= sctp_getsockopt_connectx3(sk
, len
, optval
, optlen
);
6016 case SCTP_DEFAULT_SEND_PARAM
:
6017 retval
= sctp_getsockopt_default_send_param(sk
, len
,
6020 case SCTP_DEFAULT_SNDINFO
:
6021 retval
= sctp_getsockopt_default_sndinfo(sk
, len
,
6024 case SCTP_PRIMARY_ADDR
:
6025 retval
= sctp_getsockopt_primary_addr(sk
, len
, optval
, optlen
);
6028 retval
= sctp_getsockopt_nodelay(sk
, len
, optval
, optlen
);
6031 retval
= sctp_getsockopt_rtoinfo(sk
, len
, optval
, optlen
);
6033 case SCTP_ASSOCINFO
:
6034 retval
= sctp_getsockopt_associnfo(sk
, len
, optval
, optlen
);
6036 case SCTP_I_WANT_MAPPED_V4_ADDR
:
6037 retval
= sctp_getsockopt_mappedv4(sk
, len
, optval
, optlen
);
6040 retval
= sctp_getsockopt_maxseg(sk
, len
, optval
, optlen
);
6042 case SCTP_GET_PEER_ADDR_INFO
:
6043 retval
= sctp_getsockopt_peer_addr_info(sk
, len
, optval
,
6046 case SCTP_ADAPTATION_LAYER
:
6047 retval
= sctp_getsockopt_adaptation_layer(sk
, len
, optval
,
6051 retval
= sctp_getsockopt_context(sk
, len
, optval
, optlen
);
6053 case SCTP_FRAGMENT_INTERLEAVE
:
6054 retval
= sctp_getsockopt_fragment_interleave(sk
, len
, optval
,
6057 case SCTP_PARTIAL_DELIVERY_POINT
:
6058 retval
= sctp_getsockopt_partial_delivery_point(sk
, len
, optval
,
6061 case SCTP_MAX_BURST
:
6062 retval
= sctp_getsockopt_maxburst(sk
, len
, optval
, optlen
);
6065 case SCTP_AUTH_CHUNK
:
6066 case SCTP_AUTH_DELETE_KEY
:
6067 retval
= -EOPNOTSUPP
;
6069 case SCTP_HMAC_IDENT
:
6070 retval
= sctp_getsockopt_hmac_ident(sk
, len
, optval
, optlen
);
6072 case SCTP_AUTH_ACTIVE_KEY
:
6073 retval
= sctp_getsockopt_active_key(sk
, len
, optval
, optlen
);
6075 case SCTP_PEER_AUTH_CHUNKS
:
6076 retval
= sctp_getsockopt_peer_auth_chunks(sk
, len
, optval
,
6079 case SCTP_LOCAL_AUTH_CHUNKS
:
6080 retval
= sctp_getsockopt_local_auth_chunks(sk
, len
, optval
,
6083 case SCTP_GET_ASSOC_NUMBER
:
6084 retval
= sctp_getsockopt_assoc_number(sk
, len
, optval
, optlen
);
6086 case SCTP_GET_ASSOC_ID_LIST
:
6087 retval
= sctp_getsockopt_assoc_ids(sk
, len
, optval
, optlen
);
6089 case SCTP_AUTO_ASCONF
:
6090 retval
= sctp_getsockopt_auto_asconf(sk
, len
, optval
, optlen
);
6092 case SCTP_PEER_ADDR_THLDS
:
6093 retval
= sctp_getsockopt_paddr_thresholds(sk
, optval
, len
, optlen
);
6095 case SCTP_GET_ASSOC_STATS
:
6096 retval
= sctp_getsockopt_assoc_stats(sk
, len
, optval
, optlen
);
6098 case SCTP_RECVRCVINFO
:
6099 retval
= sctp_getsockopt_recvrcvinfo(sk
, len
, optval
, optlen
);
6101 case SCTP_RECVNXTINFO
:
6102 retval
= sctp_getsockopt_recvnxtinfo(sk
, len
, optval
, optlen
);
6105 retval
= -ENOPROTOOPT
;
6113 static void sctp_hash(struct sock
*sk
)
6118 static void sctp_unhash(struct sock
*sk
)
6123 /* Check if port is acceptable. Possibly find first available port.
6125 * The port hash table (contained in the 'global' SCTP protocol storage
6126 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
6127 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
6128 * list (the list number is the port number hashed out, so as you
6129 * would expect from a hash function, all the ports in a given list have
6130 * such a number that hashes out to the same list number; you were
6131 * expecting that, right?); so each list has a set of ports, with a
6132 * link to the socket (struct sock) that uses it, the port number and
6133 * a fastreuse flag (FIXME: NPI ipg).
6135 static struct sctp_bind_bucket
*sctp_bucket_create(
6136 struct sctp_bind_hashbucket
*head
, struct net
*, unsigned short snum
);
6138 static long sctp_get_port_local(struct sock
*sk
, union sctp_addr
*addr
)
6140 struct sctp_bind_hashbucket
*head
; /* hash list */
6141 struct sctp_bind_bucket
*pp
;
6142 unsigned short snum
;
6145 snum
= ntohs(addr
->v4
.sin_port
);
6147 pr_debug("%s: begins, snum:%d\n", __func__
, snum
);
6152 /* Search for an available port. */
6153 int low
, high
, remaining
, index
;
6155 struct net
*net
= sock_net(sk
);
6157 inet_get_local_port_range(net
, &low
, &high
);
6158 remaining
= (high
- low
) + 1;
6159 rover
= prandom_u32() % remaining
+ low
;
6163 if ((rover
< low
) || (rover
> high
))
6165 if (inet_is_local_reserved_port(net
, rover
))
6167 index
= sctp_phashfn(sock_net(sk
), rover
);
6168 head
= &sctp_port_hashtable
[index
];
6169 spin_lock(&head
->lock
);
6170 sctp_for_each_hentry(pp
, &head
->chain
)
6171 if ((pp
->port
== rover
) &&
6172 net_eq(sock_net(sk
), pp
->net
))
6176 spin_unlock(&head
->lock
);
6177 } while (--remaining
> 0);
6179 /* Exhausted local port range during search? */
6184 /* OK, here is the one we will use. HEAD (the port
6185 * hash table list entry) is non-NULL and we hold it's
6190 /* We are given an specific port number; we verify
6191 * that it is not being used. If it is used, we will
6192 * exahust the search in the hash list corresponding
6193 * to the port number (snum) - we detect that with the
6194 * port iterator, pp being NULL.
6196 head
= &sctp_port_hashtable
[sctp_phashfn(sock_net(sk
), snum
)];
6197 spin_lock(&head
->lock
);
6198 sctp_for_each_hentry(pp
, &head
->chain
) {
6199 if ((pp
->port
== snum
) && net_eq(pp
->net
, sock_net(sk
)))
6206 if (!hlist_empty(&pp
->owner
)) {
6207 /* We had a port hash table hit - there is an
6208 * available port (pp != NULL) and it is being
6209 * used by other socket (pp->owner not empty); that other
6210 * socket is going to be sk2.
6212 int reuse
= sk
->sk_reuse
;
6215 pr_debug("%s: found a possible match\n", __func__
);
6217 if (pp
->fastreuse
&& sk
->sk_reuse
&&
6218 sk
->sk_state
!= SCTP_SS_LISTENING
)
6221 /* Run through the list of sockets bound to the port
6222 * (pp->port) [via the pointers bind_next and
6223 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6224 * we get the endpoint they describe and run through
6225 * the endpoint's list of IP (v4 or v6) addresses,
6226 * comparing each of the addresses with the address of
6227 * the socket sk. If we find a match, then that means
6228 * that this port/socket (sk) combination are already
6231 sk_for_each_bound(sk2
, &pp
->owner
) {
6232 struct sctp_endpoint
*ep2
;
6233 ep2
= sctp_sk(sk2
)->ep
;
6236 (reuse
&& sk2
->sk_reuse
&&
6237 sk2
->sk_state
!= SCTP_SS_LISTENING
))
6240 if (sctp_bind_addr_conflict(&ep2
->base
.bind_addr
, addr
,
6241 sctp_sk(sk2
), sctp_sk(sk
))) {
6247 pr_debug("%s: found a match\n", __func__
);
6250 /* If there was a hash table miss, create a new port. */
6252 if (!pp
&& !(pp
= sctp_bucket_create(head
, sock_net(sk
), snum
)))
6255 /* In either case (hit or miss), make sure fastreuse is 1 only
6256 * if sk->sk_reuse is too (that is, if the caller requested
6257 * SO_REUSEADDR on this socket -sk-).
6259 if (hlist_empty(&pp
->owner
)) {
6260 if (sk
->sk_reuse
&& sk
->sk_state
!= SCTP_SS_LISTENING
)
6264 } else if (pp
->fastreuse
&&
6265 (!sk
->sk_reuse
|| sk
->sk_state
== SCTP_SS_LISTENING
))
6268 /* We are set, so fill up all the data in the hash table
6269 * entry, tie the socket list information with the rest of the
6270 * sockets FIXME: Blurry, NPI (ipg).
6273 if (!sctp_sk(sk
)->bind_hash
) {
6274 inet_sk(sk
)->inet_num
= snum
;
6275 sk_add_bind_node(sk
, &pp
->owner
);
6276 sctp_sk(sk
)->bind_hash
= pp
;
6281 spin_unlock(&head
->lock
);
6288 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6289 * port is requested.
6291 static int sctp_get_port(struct sock
*sk
, unsigned short snum
)
6293 union sctp_addr addr
;
6294 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
6296 /* Set up a dummy address struct from the sk. */
6297 af
->from_sk(&addr
, sk
);
6298 addr
.v4
.sin_port
= htons(snum
);
6300 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6301 return !!sctp_get_port_local(sk
, &addr
);
6305 * Move a socket to LISTENING state.
6307 static int sctp_listen_start(struct sock
*sk
, int backlog
)
6309 struct sctp_sock
*sp
= sctp_sk(sk
);
6310 struct sctp_endpoint
*ep
= sp
->ep
;
6311 struct crypto_hash
*tfm
= NULL
;
6314 /* Allocate HMAC for generating cookie. */
6315 if (!sp
->hmac
&& sp
->sctp_hmac_alg
) {
6316 sprintf(alg
, "hmac(%s)", sp
->sctp_hmac_alg
);
6317 tfm
= crypto_alloc_hash(alg
, 0, CRYPTO_ALG_ASYNC
);
6319 net_info_ratelimited("failed to load transform for %s: %ld\n",
6320 sp
->sctp_hmac_alg
, PTR_ERR(tfm
));
6323 sctp_sk(sk
)->hmac
= tfm
;
6327 * If a bind() or sctp_bindx() is not called prior to a listen()
6328 * call that allows new associations to be accepted, the system
6329 * picks an ephemeral port and will choose an address set equivalent
6330 * to binding with a wildcard address.
6332 * This is not currently spelled out in the SCTP sockets
6333 * extensions draft, but follows the practice as seen in TCP
6337 sk
->sk_state
= SCTP_SS_LISTENING
;
6338 if (!ep
->base
.bind_addr
.port
) {
6339 if (sctp_autobind(sk
))
6342 if (sctp_get_port(sk
, inet_sk(sk
)->inet_num
)) {
6343 sk
->sk_state
= SCTP_SS_CLOSED
;
6348 sk
->sk_max_ack_backlog
= backlog
;
6349 sctp_hash_endpoint(ep
);
6354 * 4.1.3 / 5.1.3 listen()
6356 * By default, new associations are not accepted for UDP style sockets.
6357 * An application uses listen() to mark a socket as being able to
6358 * accept new associations.
6360 * On TCP style sockets, applications use listen() to ready the SCTP
6361 * endpoint for accepting inbound associations.
6363 * On both types of endpoints a backlog of '0' disables listening.
6365 * Move a socket to LISTENING state.
6367 int sctp_inet_listen(struct socket
*sock
, int backlog
)
6369 struct sock
*sk
= sock
->sk
;
6370 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
6373 if (unlikely(backlog
< 0))
6378 /* Peeled-off sockets are not allowed to listen(). */
6379 if (sctp_style(sk
, UDP_HIGH_BANDWIDTH
))
6382 if (sock
->state
!= SS_UNCONNECTED
)
6385 /* If backlog is zero, disable listening. */
6387 if (sctp_sstate(sk
, CLOSED
))
6391 sctp_unhash_endpoint(ep
);
6392 sk
->sk_state
= SCTP_SS_CLOSED
;
6394 sctp_sk(sk
)->bind_hash
->fastreuse
= 1;
6398 /* If we are already listening, just update the backlog */
6399 if (sctp_sstate(sk
, LISTENING
))
6400 sk
->sk_max_ack_backlog
= backlog
;
6402 err
= sctp_listen_start(sk
, backlog
);
6414 * This function is done by modeling the current datagram_poll() and the
6415 * tcp_poll(). Note that, based on these implementations, we don't
6416 * lock the socket in this function, even though it seems that,
6417 * ideally, locking or some other mechanisms can be used to ensure
6418 * the integrity of the counters (sndbuf and wmem_alloc) used
6419 * in this place. We assume that we don't need locks either until proven
6422 * Another thing to note is that we include the Async I/O support
6423 * here, again, by modeling the current TCP/UDP code. We don't have
6424 * a good way to test with it yet.
6426 unsigned int sctp_poll(struct file
*file
, struct socket
*sock
, poll_table
*wait
)
6428 struct sock
*sk
= sock
->sk
;
6429 struct sctp_sock
*sp
= sctp_sk(sk
);
6432 poll_wait(file
, sk_sleep(sk
), wait
);
6434 /* A TCP-style listening socket becomes readable when the accept queue
6437 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
6438 return (!list_empty(&sp
->ep
->asocs
)) ?
6439 (POLLIN
| POLLRDNORM
) : 0;
6443 /* Is there any exceptional events? */
6444 if (sk
->sk_err
|| !skb_queue_empty(&sk
->sk_error_queue
))
6446 (sock_flag(sk
, SOCK_SELECT_ERR_QUEUE
) ? POLLPRI
: 0);
6447 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6448 mask
|= POLLRDHUP
| POLLIN
| POLLRDNORM
;
6449 if (sk
->sk_shutdown
== SHUTDOWN_MASK
)
6452 /* Is it readable? Reconsider this code with TCP-style support. */
6453 if (!skb_queue_empty(&sk
->sk_receive_queue
))
6454 mask
|= POLLIN
| POLLRDNORM
;
6456 /* The association is either gone or not ready. */
6457 if (!sctp_style(sk
, UDP
) && sctp_sstate(sk
, CLOSED
))
6460 /* Is it writable? */
6461 if (sctp_writeable(sk
)) {
6462 mask
|= POLLOUT
| POLLWRNORM
;
6464 set_bit(SOCK_ASYNC_NOSPACE
, &sk
->sk_socket
->flags
);
6466 * Since the socket is not locked, the buffer
6467 * might be made available after the writeable check and
6468 * before the bit is set. This could cause a lost I/O
6469 * signal. tcp_poll() has a race breaker for this race
6470 * condition. Based on their implementation, we put
6471 * in the following code to cover it as well.
6473 if (sctp_writeable(sk
))
6474 mask
|= POLLOUT
| POLLWRNORM
;
6479 /********************************************************************
6480 * 2nd Level Abstractions
6481 ********************************************************************/
6483 static struct sctp_bind_bucket
*sctp_bucket_create(
6484 struct sctp_bind_hashbucket
*head
, struct net
*net
, unsigned short snum
)
6486 struct sctp_bind_bucket
*pp
;
6488 pp
= kmem_cache_alloc(sctp_bucket_cachep
, GFP_ATOMIC
);
6490 SCTP_DBG_OBJCNT_INC(bind_bucket
);
6493 INIT_HLIST_HEAD(&pp
->owner
);
6495 hlist_add_head(&pp
->node
, &head
->chain
);
6500 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6501 static void sctp_bucket_destroy(struct sctp_bind_bucket
*pp
)
6503 if (pp
&& hlist_empty(&pp
->owner
)) {
6504 __hlist_del(&pp
->node
);
6505 kmem_cache_free(sctp_bucket_cachep
, pp
);
6506 SCTP_DBG_OBJCNT_DEC(bind_bucket
);
6510 /* Release this socket's reference to a local port. */
6511 static inline void __sctp_put_port(struct sock
*sk
)
6513 struct sctp_bind_hashbucket
*head
=
6514 &sctp_port_hashtable
[sctp_phashfn(sock_net(sk
),
6515 inet_sk(sk
)->inet_num
)];
6516 struct sctp_bind_bucket
*pp
;
6518 spin_lock(&head
->lock
);
6519 pp
= sctp_sk(sk
)->bind_hash
;
6520 __sk_del_bind_node(sk
);
6521 sctp_sk(sk
)->bind_hash
= NULL
;
6522 inet_sk(sk
)->inet_num
= 0;
6523 sctp_bucket_destroy(pp
);
6524 spin_unlock(&head
->lock
);
6527 void sctp_put_port(struct sock
*sk
)
6530 __sctp_put_port(sk
);
6535 * The system picks an ephemeral port and choose an address set equivalent
6536 * to binding with a wildcard address.
6537 * One of those addresses will be the primary address for the association.
6538 * This automatically enables the multihoming capability of SCTP.
6540 static int sctp_autobind(struct sock
*sk
)
6542 union sctp_addr autoaddr
;
6546 /* Initialize a local sockaddr structure to INADDR_ANY. */
6547 af
= sctp_sk(sk
)->pf
->af
;
6549 port
= htons(inet_sk(sk
)->inet_num
);
6550 af
->inaddr_any(&autoaddr
, port
);
6552 return sctp_do_bind(sk
, &autoaddr
, af
->sockaddr_len
);
6555 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6558 * 4.2 The cmsghdr Structure *
6560 * When ancillary data is sent or received, any number of ancillary data
6561 * objects can be specified by the msg_control and msg_controllen members of
6562 * the msghdr structure, because each object is preceded by
6563 * a cmsghdr structure defining the object's length (the cmsg_len member).
6564 * Historically Berkeley-derived implementations have passed only one object
6565 * at a time, but this API allows multiple objects to be
6566 * passed in a single call to sendmsg() or recvmsg(). The following example
6567 * shows two ancillary data objects in a control buffer.
6569 * |<--------------------------- msg_controllen -------------------------->|
6572 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6574 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6577 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6579 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6582 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6583 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6585 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6587 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6594 static int sctp_msghdr_parse(const struct msghdr
*msg
, sctp_cmsgs_t
*cmsgs
)
6596 struct cmsghdr
*cmsg
;
6597 struct msghdr
*my_msg
= (struct msghdr
*)msg
;
6599 for_each_cmsghdr(cmsg
, my_msg
) {
6600 if (!CMSG_OK(my_msg
, cmsg
))
6603 /* Should we parse this header or ignore? */
6604 if (cmsg
->cmsg_level
!= IPPROTO_SCTP
)
6607 /* Strictly check lengths following example in SCM code. */
6608 switch (cmsg
->cmsg_type
) {
6610 /* SCTP Socket API Extension
6611 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
6613 * This cmsghdr structure provides information for
6614 * initializing new SCTP associations with sendmsg().
6615 * The SCTP_INITMSG socket option uses this same data
6616 * structure. This structure is not used for
6619 * cmsg_level cmsg_type cmsg_data[]
6620 * ------------ ------------ ----------------------
6621 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6623 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(struct sctp_initmsg
)))
6626 cmsgs
->init
= CMSG_DATA(cmsg
);
6630 /* SCTP Socket API Extension
6631 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
6633 * This cmsghdr structure specifies SCTP options for
6634 * sendmsg() and describes SCTP header information
6635 * about a received message through recvmsg().
6637 * cmsg_level cmsg_type cmsg_data[]
6638 * ------------ ------------ ----------------------
6639 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6641 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(struct sctp_sndrcvinfo
)))
6644 cmsgs
->srinfo
= CMSG_DATA(cmsg
);
6646 if (cmsgs
->srinfo
->sinfo_flags
&
6647 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
6648 SCTP_ABORT
| SCTP_EOF
))
6653 /* SCTP Socket API Extension
6654 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
6656 * This cmsghdr structure specifies SCTP options for
6657 * sendmsg(). This structure and SCTP_RCVINFO replaces
6658 * SCTP_SNDRCV which has been deprecated.
6660 * cmsg_level cmsg_type cmsg_data[]
6661 * ------------ ------------ ---------------------
6662 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
6664 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(struct sctp_sndinfo
)))
6667 cmsgs
->sinfo
= CMSG_DATA(cmsg
);
6669 if (cmsgs
->sinfo
->snd_flags
&
6670 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
6671 SCTP_ABORT
| SCTP_EOF
))
6683 * Wait for a packet..
6684 * Note: This function is the same function as in core/datagram.c
6685 * with a few modifications to make lksctp work.
6687 static int sctp_wait_for_packet(struct sock
*sk
, int *err
, long *timeo_p
)
6692 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
6694 /* Socket errors? */
6695 error
= sock_error(sk
);
6699 if (!skb_queue_empty(&sk
->sk_receive_queue
))
6702 /* Socket shut down? */
6703 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6706 /* Sequenced packets can come disconnected. If so we report the
6711 /* Is there a good reason to think that we may receive some data? */
6712 if (list_empty(&sctp_sk(sk
)->ep
->asocs
) && !sctp_sstate(sk
, LISTENING
))
6715 /* Handle signals. */
6716 if (signal_pending(current
))
6719 /* Let another process have a go. Since we are going to sleep
6720 * anyway. Note: This may cause odd behaviors if the message
6721 * does not fit in the user's buffer, but this seems to be the
6722 * only way to honor MSG_DONTWAIT realistically.
6725 *timeo_p
= schedule_timeout(*timeo_p
);
6729 finish_wait(sk_sleep(sk
), &wait
);
6733 error
= sock_intr_errno(*timeo_p
);
6736 finish_wait(sk_sleep(sk
), &wait
);
6741 /* Receive a datagram.
6742 * Note: This is pretty much the same routine as in core/datagram.c
6743 * with a few changes to make lksctp work.
6745 struct sk_buff
*sctp_skb_recv_datagram(struct sock
*sk
, int flags
,
6746 int noblock
, int *err
)
6749 struct sk_buff
*skb
;
6752 timeo
= sock_rcvtimeo(sk
, noblock
);
6754 pr_debug("%s: timeo:%ld, max:%ld\n", __func__
, timeo
,
6755 MAX_SCHEDULE_TIMEOUT
);
6758 /* Again only user level code calls this function,
6759 * so nothing interrupt level
6760 * will suddenly eat the receive_queue.
6762 * Look at current nfs client by the way...
6763 * However, this function was correct in any case. 8)
6765 if (flags
& MSG_PEEK
) {
6766 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
6767 skb
= skb_peek(&sk
->sk_receive_queue
);
6769 atomic_inc(&skb
->users
);
6770 spin_unlock_bh(&sk
->sk_receive_queue
.lock
);
6772 skb
= skb_dequeue(&sk
->sk_receive_queue
);
6778 /* Caller is allowed not to check sk->sk_err before calling. */
6779 error
= sock_error(sk
);
6783 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6786 if (sk_can_busy_loop(sk
) &&
6787 sk_busy_loop(sk
, noblock
))
6790 /* User doesn't want to wait. */
6794 } while (sctp_wait_for_packet(sk
, err
, &timeo
) == 0);
6803 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6804 static void __sctp_write_space(struct sctp_association
*asoc
)
6806 struct sock
*sk
= asoc
->base
.sk
;
6807 struct socket
*sock
= sk
->sk_socket
;
6809 if ((sctp_wspace(asoc
) > 0) && sock
) {
6810 if (waitqueue_active(&asoc
->wait
))
6811 wake_up_interruptible(&asoc
->wait
);
6813 if (sctp_writeable(sk
)) {
6814 wait_queue_head_t
*wq
= sk_sleep(sk
);
6816 if (wq
&& waitqueue_active(wq
))
6817 wake_up_interruptible(wq
);
6819 /* Note that we try to include the Async I/O support
6820 * here by modeling from the current TCP/UDP code.
6821 * We have not tested with it yet.
6823 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
))
6824 sock_wake_async(sock
,
6825 SOCK_WAKE_SPACE
, POLL_OUT
);
6830 static void sctp_wake_up_waiters(struct sock
*sk
,
6831 struct sctp_association
*asoc
)
6833 struct sctp_association
*tmp
= asoc
;
6835 /* We do accounting for the sndbuf space per association,
6836 * so we only need to wake our own association.
6838 if (asoc
->ep
->sndbuf_policy
)
6839 return __sctp_write_space(asoc
);
6841 /* If association goes down and is just flushing its
6842 * outq, then just normally notify others.
6844 if (asoc
->base
.dead
)
6845 return sctp_write_space(sk
);
6847 /* Accounting for the sndbuf space is per socket, so we
6848 * need to wake up others, try to be fair and in case of
6849 * other associations, let them have a go first instead
6850 * of just doing a sctp_write_space() call.
6852 * Note that we reach sctp_wake_up_waiters() only when
6853 * associations free up queued chunks, thus we are under
6854 * lock and the list of associations on a socket is
6855 * guaranteed not to change.
6857 for (tmp
= list_next_entry(tmp
, asocs
); 1;
6858 tmp
= list_next_entry(tmp
, asocs
)) {
6859 /* Manually skip the head element. */
6860 if (&tmp
->asocs
== &((sctp_sk(sk
))->ep
->asocs
))
6862 /* Wake up association. */
6863 __sctp_write_space(tmp
);
6864 /* We've reached the end. */
6870 /* Do accounting for the sndbuf space.
6871 * Decrement the used sndbuf space of the corresponding association by the
6872 * data size which was just transmitted(freed).
6874 static void sctp_wfree(struct sk_buff
*skb
)
6876 struct sctp_chunk
*chunk
= skb_shinfo(skb
)->destructor_arg
;
6877 struct sctp_association
*asoc
= chunk
->asoc
;
6878 struct sock
*sk
= asoc
->base
.sk
;
6880 asoc
->sndbuf_used
-= SCTP_DATA_SNDSIZE(chunk
) +
6881 sizeof(struct sk_buff
) +
6882 sizeof(struct sctp_chunk
);
6884 atomic_sub(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
6887 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6889 sk
->sk_wmem_queued
-= skb
->truesize
;
6890 sk_mem_uncharge(sk
, skb
->truesize
);
6893 sctp_wake_up_waiters(sk
, asoc
);
6895 sctp_association_put(asoc
);
6898 /* Do accounting for the receive space on the socket.
6899 * Accounting for the association is done in ulpevent.c
6900 * We set this as a destructor for the cloned data skbs so that
6901 * accounting is done at the correct time.
6903 void sctp_sock_rfree(struct sk_buff
*skb
)
6905 struct sock
*sk
= skb
->sk
;
6906 struct sctp_ulpevent
*event
= sctp_skb2event(skb
);
6908 atomic_sub(event
->rmem_len
, &sk
->sk_rmem_alloc
);
6911 * Mimic the behavior of sock_rfree
6913 sk_mem_uncharge(sk
, event
->rmem_len
);
6917 /* Helper function to wait for space in the sndbuf. */
6918 static int sctp_wait_for_sndbuf(struct sctp_association
*asoc
, long *timeo_p
,
6921 struct sock
*sk
= asoc
->base
.sk
;
6923 long current_timeo
= *timeo_p
;
6926 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__
, asoc
,
6929 /* Increment the association's refcnt. */
6930 sctp_association_hold(asoc
);
6932 /* Wait on the association specific sndbuf space. */
6934 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
6935 TASK_INTERRUPTIBLE
);
6938 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
6941 if (signal_pending(current
))
6942 goto do_interrupted
;
6943 if (msg_len
<= sctp_wspace(asoc
))
6946 /* Let another process have a go. Since we are going
6950 current_timeo
= schedule_timeout(current_timeo
);
6951 BUG_ON(sk
!= asoc
->base
.sk
);
6954 *timeo_p
= current_timeo
;
6958 finish_wait(&asoc
->wait
, &wait
);
6960 /* Release the association's refcnt. */
6961 sctp_association_put(asoc
);
6970 err
= sock_intr_errno(*timeo_p
);
6978 void sctp_data_ready(struct sock
*sk
)
6980 struct socket_wq
*wq
;
6983 wq
= rcu_dereference(sk
->sk_wq
);
6984 if (wq_has_sleeper(wq
))
6985 wake_up_interruptible_sync_poll(&wq
->wait
, POLLIN
|
6986 POLLRDNORM
| POLLRDBAND
);
6987 sk_wake_async(sk
, SOCK_WAKE_WAITD
, POLL_IN
);
6991 /* If socket sndbuf has changed, wake up all per association waiters. */
6992 void sctp_write_space(struct sock
*sk
)
6994 struct sctp_association
*asoc
;
6996 /* Wake up the tasks in each wait queue. */
6997 list_for_each_entry(asoc
, &((sctp_sk(sk
))->ep
->asocs
), asocs
) {
6998 __sctp_write_space(asoc
);
7002 /* Is there any sndbuf space available on the socket?
7004 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
7005 * associations on the same socket. For a UDP-style socket with
7006 * multiple associations, it is possible for it to be "unwriteable"
7007 * prematurely. I assume that this is acceptable because
7008 * a premature "unwriteable" is better than an accidental "writeable" which
7009 * would cause an unwanted block under certain circumstances. For the 1-1
7010 * UDP-style sockets or TCP-style sockets, this code should work.
7013 static int sctp_writeable(struct sock
*sk
)
7017 amt
= sk
->sk_sndbuf
- sk_wmem_alloc_get(sk
);
7023 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
7024 * returns immediately with EINPROGRESS.
7026 static int sctp_wait_for_connect(struct sctp_association
*asoc
, long *timeo_p
)
7028 struct sock
*sk
= asoc
->base
.sk
;
7030 long current_timeo
= *timeo_p
;
7033 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__
, asoc
, *timeo_p
);
7035 /* Increment the association's refcnt. */
7036 sctp_association_hold(asoc
);
7039 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
7040 TASK_INTERRUPTIBLE
);
7043 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
7045 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
7048 if (signal_pending(current
))
7049 goto do_interrupted
;
7051 if (sctp_state(asoc
, ESTABLISHED
))
7054 /* Let another process have a go. Since we are going
7058 current_timeo
= schedule_timeout(current_timeo
);
7061 *timeo_p
= current_timeo
;
7065 finish_wait(&asoc
->wait
, &wait
);
7067 /* Release the association's refcnt. */
7068 sctp_association_put(asoc
);
7073 if (asoc
->init_err_counter
+ 1 > asoc
->max_init_attempts
)
7076 err
= -ECONNREFUSED
;
7080 err
= sock_intr_errno(*timeo_p
);
7088 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
)
7090 struct sctp_endpoint
*ep
;
7094 ep
= sctp_sk(sk
)->ep
;
7098 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
,
7099 TASK_INTERRUPTIBLE
);
7101 if (list_empty(&ep
->asocs
)) {
7103 timeo
= schedule_timeout(timeo
);
7108 if (!sctp_sstate(sk
, LISTENING
))
7112 if (!list_empty(&ep
->asocs
))
7115 err
= sock_intr_errno(timeo
);
7116 if (signal_pending(current
))
7124 finish_wait(sk_sleep(sk
), &wait
);
7129 static void sctp_wait_for_close(struct sock
*sk
, long timeout
)
7134 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
7135 if (list_empty(&sctp_sk(sk
)->ep
->asocs
))
7138 timeout
= schedule_timeout(timeout
);
7140 } while (!signal_pending(current
) && timeout
);
7142 finish_wait(sk_sleep(sk
), &wait
);
7145 static void sctp_skb_set_owner_r_frag(struct sk_buff
*skb
, struct sock
*sk
)
7147 struct sk_buff
*frag
;
7152 /* Don't forget the fragments. */
7153 skb_walk_frags(skb
, frag
)
7154 sctp_skb_set_owner_r_frag(frag
, sk
);
7157 sctp_skb_set_owner_r(skb
, sk
);
7160 void sctp_copy_sock(struct sock
*newsk
, struct sock
*sk
,
7161 struct sctp_association
*asoc
)
7163 struct inet_sock
*inet
= inet_sk(sk
);
7164 struct inet_sock
*newinet
;
7166 newsk
->sk_type
= sk
->sk_type
;
7167 newsk
->sk_bound_dev_if
= sk
->sk_bound_dev_if
;
7168 newsk
->sk_flags
= sk
->sk_flags
;
7169 newsk
->sk_no_check_tx
= sk
->sk_no_check_tx
;
7170 newsk
->sk_no_check_rx
= sk
->sk_no_check_rx
;
7171 newsk
->sk_reuse
= sk
->sk_reuse
;
7173 newsk
->sk_shutdown
= sk
->sk_shutdown
;
7174 newsk
->sk_destruct
= sctp_destruct_sock
;
7175 newsk
->sk_family
= sk
->sk_family
;
7176 newsk
->sk_protocol
= IPPROTO_SCTP
;
7177 newsk
->sk_backlog_rcv
= sk
->sk_prot
->backlog_rcv
;
7178 newsk
->sk_sndbuf
= sk
->sk_sndbuf
;
7179 newsk
->sk_rcvbuf
= sk
->sk_rcvbuf
;
7180 newsk
->sk_lingertime
= sk
->sk_lingertime
;
7181 newsk
->sk_rcvtimeo
= sk
->sk_rcvtimeo
;
7182 newsk
->sk_sndtimeo
= sk
->sk_sndtimeo
;
7184 newinet
= inet_sk(newsk
);
7186 /* Initialize sk's sport, dport, rcv_saddr and daddr for
7187 * getsockname() and getpeername()
7189 newinet
->inet_sport
= inet
->inet_sport
;
7190 newinet
->inet_saddr
= inet
->inet_saddr
;
7191 newinet
->inet_rcv_saddr
= inet
->inet_rcv_saddr
;
7192 newinet
->inet_dport
= htons(asoc
->peer
.port
);
7193 newinet
->pmtudisc
= inet
->pmtudisc
;
7194 newinet
->inet_id
= asoc
->next_tsn
^ jiffies
;
7196 newinet
->uc_ttl
= inet
->uc_ttl
;
7197 newinet
->mc_loop
= 1;
7198 newinet
->mc_ttl
= 1;
7199 newinet
->mc_index
= 0;
7200 newinet
->mc_list
= NULL
;
7203 /* Populate the fields of the newsk from the oldsk and migrate the assoc
7204 * and its messages to the newsk.
7206 static void sctp_sock_migrate(struct sock
*oldsk
, struct sock
*newsk
,
7207 struct sctp_association
*assoc
,
7208 sctp_socket_type_t type
)
7210 struct sctp_sock
*oldsp
= sctp_sk(oldsk
);
7211 struct sctp_sock
*newsp
= sctp_sk(newsk
);
7212 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
7213 struct sctp_endpoint
*newep
= newsp
->ep
;
7214 struct sk_buff
*skb
, *tmp
;
7215 struct sctp_ulpevent
*event
;
7216 struct sctp_bind_hashbucket
*head
;
7217 struct list_head tmplist
;
7219 /* Migrate socket buffer sizes and all the socket level options to the
7222 newsk
->sk_sndbuf
= oldsk
->sk_sndbuf
;
7223 newsk
->sk_rcvbuf
= oldsk
->sk_rcvbuf
;
7224 /* Brute force copy old sctp opt. */
7225 if (oldsp
->do_auto_asconf
) {
7226 memcpy(&tmplist
, &newsp
->auto_asconf_list
, sizeof(tmplist
));
7227 inet_sk_copy_descendant(newsk
, oldsk
);
7228 memcpy(&newsp
->auto_asconf_list
, &tmplist
, sizeof(tmplist
));
7230 inet_sk_copy_descendant(newsk
, oldsk
);
7232 /* Restore the ep value that was overwritten with the above structure
7238 /* Hook this new socket in to the bind_hash list. */
7239 head
= &sctp_port_hashtable
[sctp_phashfn(sock_net(oldsk
),
7240 inet_sk(oldsk
)->inet_num
)];
7242 spin_lock(&head
->lock
);
7243 pp
= sctp_sk(oldsk
)->bind_hash
;
7244 sk_add_bind_node(newsk
, &pp
->owner
);
7245 sctp_sk(newsk
)->bind_hash
= pp
;
7246 inet_sk(newsk
)->inet_num
= inet_sk(oldsk
)->inet_num
;
7247 spin_unlock(&head
->lock
);
7250 /* Copy the bind_addr list from the original endpoint to the new
7251 * endpoint so that we can handle restarts properly
7253 sctp_bind_addr_dup(&newsp
->ep
->base
.bind_addr
,
7254 &oldsp
->ep
->base
.bind_addr
, GFP_KERNEL
);
7256 /* Move any messages in the old socket's receive queue that are for the
7257 * peeled off association to the new socket's receive queue.
7259 sctp_skb_for_each(skb
, &oldsk
->sk_receive_queue
, tmp
) {
7260 event
= sctp_skb2event(skb
);
7261 if (event
->asoc
== assoc
) {
7262 __skb_unlink(skb
, &oldsk
->sk_receive_queue
);
7263 __skb_queue_tail(&newsk
->sk_receive_queue
, skb
);
7264 sctp_skb_set_owner_r_frag(skb
, newsk
);
7268 /* Clean up any messages pending delivery due to partial
7269 * delivery. Three cases:
7270 * 1) No partial deliver; no work.
7271 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7272 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7274 skb_queue_head_init(&newsp
->pd_lobby
);
7275 atomic_set(&sctp_sk(newsk
)->pd_mode
, assoc
->ulpq
.pd_mode
);
7277 if (atomic_read(&sctp_sk(oldsk
)->pd_mode
)) {
7278 struct sk_buff_head
*queue
;
7280 /* Decide which queue to move pd_lobby skbs to. */
7281 if (assoc
->ulpq
.pd_mode
) {
7282 queue
= &newsp
->pd_lobby
;
7284 queue
= &newsk
->sk_receive_queue
;
7286 /* Walk through the pd_lobby, looking for skbs that
7287 * need moved to the new socket.
7289 sctp_skb_for_each(skb
, &oldsp
->pd_lobby
, tmp
) {
7290 event
= sctp_skb2event(skb
);
7291 if (event
->asoc
== assoc
) {
7292 __skb_unlink(skb
, &oldsp
->pd_lobby
);
7293 __skb_queue_tail(queue
, skb
);
7294 sctp_skb_set_owner_r_frag(skb
, newsk
);
7298 /* Clear up any skbs waiting for the partial
7299 * delivery to finish.
7301 if (assoc
->ulpq
.pd_mode
)
7302 sctp_clear_pd(oldsk
, NULL
);
7306 sctp_skb_for_each(skb
, &assoc
->ulpq
.reasm
, tmp
)
7307 sctp_skb_set_owner_r_frag(skb
, newsk
);
7309 sctp_skb_for_each(skb
, &assoc
->ulpq
.lobby
, tmp
)
7310 sctp_skb_set_owner_r_frag(skb
, newsk
);
7312 /* Set the type of socket to indicate that it is peeled off from the
7313 * original UDP-style socket or created with the accept() call on a
7314 * TCP-style socket..
7318 /* Mark the new socket "in-use" by the user so that any packets
7319 * that may arrive on the association after we've moved it are
7320 * queued to the backlog. This prevents a potential race between
7321 * backlog processing on the old socket and new-packet processing
7322 * on the new socket.
7324 * The caller has just allocated newsk so we can guarantee that other
7325 * paths won't try to lock it and then oldsk.
7327 lock_sock_nested(newsk
, SINGLE_DEPTH_NESTING
);
7328 sctp_assoc_migrate(assoc
, newsk
);
7330 /* If the association on the newsk is already closed before accept()
7331 * is called, set RCV_SHUTDOWN flag.
7333 if (sctp_state(assoc
, CLOSED
) && sctp_style(newsk
, TCP
))
7334 newsk
->sk_shutdown
|= RCV_SHUTDOWN
;
7336 newsk
->sk_state
= SCTP_SS_ESTABLISHED
;
7337 release_sock(newsk
);
7341 /* This proto struct describes the ULP interface for SCTP. */
7342 struct proto sctp_prot
= {
7344 .owner
= THIS_MODULE
,
7345 .close
= sctp_close
,
7346 .connect
= sctp_connect
,
7347 .disconnect
= sctp_disconnect
,
7348 .accept
= sctp_accept
,
7349 .ioctl
= sctp_ioctl
,
7350 .init
= sctp_init_sock
,
7351 .destroy
= sctp_destroy_sock
,
7352 .shutdown
= sctp_shutdown
,
7353 .setsockopt
= sctp_setsockopt
,
7354 .getsockopt
= sctp_getsockopt
,
7355 .sendmsg
= sctp_sendmsg
,
7356 .recvmsg
= sctp_recvmsg
,
7358 .backlog_rcv
= sctp_backlog_rcv
,
7360 .unhash
= sctp_unhash
,
7361 .get_port
= sctp_get_port
,
7362 .obj_size
= sizeof(struct sctp_sock
),
7363 .sysctl_mem
= sysctl_sctp_mem
,
7364 .sysctl_rmem
= sysctl_sctp_rmem
,
7365 .sysctl_wmem
= sysctl_sctp_wmem
,
7366 .memory_pressure
= &sctp_memory_pressure
,
7367 .enter_memory_pressure
= sctp_enter_memory_pressure
,
7368 .memory_allocated
= &sctp_memory_allocated
,
7369 .sockets_allocated
= &sctp_sockets_allocated
,
7372 #if IS_ENABLED(CONFIG_IPV6)
7374 struct proto sctpv6_prot
= {
7376 .owner
= THIS_MODULE
,
7377 .close
= sctp_close
,
7378 .connect
= sctp_connect
,
7379 .disconnect
= sctp_disconnect
,
7380 .accept
= sctp_accept
,
7381 .ioctl
= sctp_ioctl
,
7382 .init
= sctp_init_sock
,
7383 .destroy
= sctp_destroy_sock
,
7384 .shutdown
= sctp_shutdown
,
7385 .setsockopt
= sctp_setsockopt
,
7386 .getsockopt
= sctp_getsockopt
,
7387 .sendmsg
= sctp_sendmsg
,
7388 .recvmsg
= sctp_recvmsg
,
7390 .backlog_rcv
= sctp_backlog_rcv
,
7392 .unhash
= sctp_unhash
,
7393 .get_port
= sctp_get_port
,
7394 .obj_size
= sizeof(struct sctp6_sock
),
7395 .sysctl_mem
= sysctl_sctp_mem
,
7396 .sysctl_rmem
= sysctl_sctp_rmem
,
7397 .sysctl_wmem
= sysctl_sctp_wmem
,
7398 .memory_pressure
= &sctp_memory_pressure
,
7399 .enter_memory_pressure
= sctp_enter_memory_pressure
,
7400 .memory_allocated
= &sctp_memory_allocated
,
7401 .sockets_allocated
= &sctp_sockets_allocated
,
7403 #endif /* IS_ENABLED(CONFIG_IPV6) */